Merge /spare/repo/netdev-2.6/ branch 'ieee80211'

61 files changed, 18546 insertions, 0 deletions

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
new file mode 100644
index 000000000000..3e89767cea72
--- /dev/null
+++ b/arch/xtensa/Kconfig
@@ -0,0 +1,258 @@
+# For a description of the syntax of this configuration file,
+# see Documentation/kbuild/config-language.txt.
+
+mainmenu "Linux/Xtensa Kernel Configuration"
+
+config FRAME_POINTER
+	bool
+	default n
+
+config XTENSA
+	bool
+	default y
+	help
+	  Xtensa processors are 32-bit RISC machines designed by Tensilica
+	  primarily for embedded systems.  These processors are both
+	  configurable and extensible.  The Linux port to the Xtensa
+	  architecture supports all processor configurations and extensions,
+	  with reasonable minimum requirements.  The Xtensa Linux project has
+	  a home page at <http://xtensa.sourceforge.net/>.
+
+config UID16
+	bool
+	default n
+
+config RWSEM_XCHGADD_ALGORITHM
+	bool
+	default y
+
+config HAVE_DEC_LOCK
+	bool
+	default y
+
+config GENERIC_HARDIRQS
+	bool
+	default y
+
+source "init/Kconfig"
+
+menu "Processor type and features"
+
+choice
+	prompt "Xtensa Processor Configuration"
+	default XTENSA_CPU_LINUX_BE
+
+config XTENSA_CPU_LINUX_BE
+	bool "linux_be"
+	---help---
+	  The linux_be processor configuration is the baseline Xtensa
+	  configurations included in this kernel and also used by
+	  binutils, gcc, and gdb. It contains no TIE, no coprocessors,
+	  and the following configuration options:
+
+	  Code Density Option                2 Misc Special Registers
+	  NSA/NSAU Instructions              128-bit Data Bus Width
+	  Processor ID                       8K, 2-way I and D Caches
+	  Zero-Overhead Loops                2 Inst Address Break Registers
+	  Big Endian                         2 Data Address Break Registers
+	  64 General-Purpose Registers       JTAG Interface and Trace Port
+	  17 Interrupts                      MMU w/ TLBs and Autorefill
+	  3 Interrupt Levels                 8 Autorefill Ways (I/D TLBs)
+	  3 Timers                           Unaligned Exceptions
+endchoice
+
+config MMU
+	bool
+	default y
+
+config XTENSA_UNALIGNED_USER
+	bool "Unaligned memory access in use space"
+	---help---
+	   The Xtensa architecture currently does not handle unaligned
+	   memory accesses in hardware but through an exception handler.
+	   Per default, unaligned memory accesses are disabled in user space.
+
+	   Say Y here to enable unaligned memory access in user space.
+
+config PREEMPT
+	bool "Preemptible Kernel"
+	---help---
+           This option reduces the latency of the kernel when reacting to
+           real-time or interactive events by allowing a low priority process to
+           be preempted even if it is in kernel mode executing a system call.
+           Unfortunately the kernel code has some race conditions if both
+           CONFIG_SMP and CONFIG_PREEMPT are enabled, so this option is
+           currently disabled if you are building an SMP kernel.
+
+           Say Y here if you are building a kernel for a desktop, embedded
+           or real-time system.  Say N if you are unsure.
+
+config MATH_EMULATION
+	bool "Math emulation"
+	help
+	Can we use information of configuration file?
+
+config HIGHMEM
+	bool "High memory support"
+
+endmenu
+
+menu "Platform options"
+
+choice
+	prompt "Xtensa System Type"
+	default XTENSA_PLATFORM_ISS
+
+config XTENSA_PLATFORM_ISS
+	bool "ISS"
+	help
+	  ISS is an acronym for Tensilica's Instruction Set Simulator.
+
+config XTENSA_PLATFORM_XT2000
+	bool "XT2000"
+	help
+	  XT2000 is the name of Tensilica's feature-rich emulation platform.
+	  This hardware is capable of running a full Linux distribution.
+
+endchoice
+
+
+config XTENSA_CALIBRATE_CCOUNT
+	bool "Auto calibration of the CPU clock rate"
+	---help---
+	  On some platforms (XT2000, for example), the CPU clock rate can
+	  vary.  The frequency can be determined, however, by measuring
+	  against a well known, fixed frequency, such as an UART oscillator.
+
+config XTENSA_CPU_CLOCK
+	int "CPU clock rate [MHz]"
+	depends on !XTENSA_CALIBRATE_CCOUNT
+	default "16"
+
+config GENERIC_CALIBRATE_DELAY
+	bool "Auto calibration of the BogoMIPS value"
+	---help---
+	  The BogoMIPS value can easily derived from the CPU frequency.
+
+config CMDLINE_BOOL
+	bool "Default bootloader kernel arguments"
+
+config CMDLINE
+	string "Initial kernel command string"
+	depends on CMDLINE_BOOL
+	default "console=ttyS0,38400 root=/dev/ram"
+	help
+	  On some architectures (EBSA110 and CATS), there is currently no way
+	  for the boot loader to pass arguments to the kernel. For these
+	  architectures, you should supply some command-line options at build
+	  time by entering them here. As a minimum, you should specify the
+	  memory size and the root device (e.g., mem=64M root=/dev/nfs).
+
+config SERIAL_CONSOLE
+	bool
+	depends on XTENSA_PLATFORM_ISS
+	default y
+
+config XTENSA_ISS_NETWORK
+	bool
+	depends on XTENSA_PLATFORM_ISS
+	default y
+
+endmenu
+
+menu "Bus options"
+
+config PCI
+	bool "PCI support" if !XTENSA_PLATFORM_ISS
+	depends on !XTENSA_PLATFORM_ISS
+	default y
+	help
+	  Find out whether you have a PCI motherboard. PCI is the name of a
+	  bus system, i.e. the way the CPU talks to the other stuff inside
+	  your box. Other bus systems are ISA, EISA, MicroChannel (MCA) or
+	  VESA. If you have PCI, say Y, otherwise N.
+
+	  The PCI-HOWTO, available from
+	  <http://www.linuxdoc.org/docs.html#howto>, contains valuable
+	  information about which PCI hardware does work under Linux and which
+	  doesn't
+
+source "drivers/pci/Kconfig"
+
+config HOTPLUG
+
+	bool "Support for hot-pluggable devices"
+	---help---
+	Say Y here if you want to plug devices into your computer while
+	the system is running, and be able to use them quickly.  In many
+	cases, the devices can likewise be unplugged at any time too.
+
+	One well known example of this is PCMCIA- or PC-cards, credit-card
+	size devices such as network cards, modems or hard drives which are
+	plugged into slots found on all modern laptop computers.  Another
+	example, used on modern desktops as well as laptops, is USB.
+
+	Enable HOTPLUG and KMOD, and build a modular kernel.  Get agent
+	software (at <http://linux-hotplug.sourceforge.net/>) and install it.
+	Then your kernel will automatically call out to a user mode "policy
+	agent" (/sbin/hotplug) to load modules and set up software needed
+	to use devices as you hotplug them.
+
+source "drivers/pcmcia/Kconfig"
+
+source "drivers/pci/hotplug/Kconfig"
+
+endmenu
+
+menu "Exectuable file formats"
+
+# only elf supported
+config KCORE_ELF
+        bool
+        depends on PROC_FS
+        default y
+        help
+          If you enabled support for /proc file system then the file
+          /proc/kcore will contain the kernel core image in ELF format. This
+          can be used in gdb:
+
+          $ cd /usr/src/linux ; gdb vmlinux /proc/kcore
+
+          This is especially useful if you have compiled the kernel with the
+          "-g" option to preserve debugging information. It is mainly used
+	  for examining kernel data structures on the live kernel.
+
+source "fs/Kconfig.binfmt"
+
+endmenu
+
+source "drivers/Kconfig"
+
+source "fs/Kconfig"
+
+menu "Xtensa initrd options"
+	depends on BLK_DEV_INITRD
+
+	config EMBEDDED_RAMDISK
+	bool "Embed root filesystem ramdisk into the kernel"
+
+config EMBEDDED_RAMDISK_IMAGE
+	string "Filename of gziped ramdisk image"
+	depends on EMBEDDED_RAMDISK
+	default "ramdisk.gz"
+	help
+	  This is the filename of the ramdisk image to be built into the
+	  kernel.  Relative pathnames are relative to arch/xtensa/boot/ramdisk/.
+	  The ramdisk image is not part of the kernel distribution; you must
+	  provide one yourself.
+endmenu
+
+source "arch/xtensa/Kconfig.debug"
+
+source "security/Kconfig"
+
+source "crypto/Kconfig"
+
+source "lib/Kconfig"
+
+
diff --git a/arch/xtensa/Kconfig.debug b/arch/xtensa/Kconfig.debug
new file mode 100644
index 000000000000..11c585295dd7
--- /dev/null
+++ b/arch/xtensa/Kconfig.debug
@@ -0,0 +1,7 @@
+menu "Kernel hacking"
+
+source "lib/Kconfig.debug"
+
+endmenu
+
+
diff --git a/arch/xtensa/Makefile b/arch/xtensa/Makefile
new file mode 100644
index 000000000000..4fa27453b1f9
--- /dev/null
+++ b/arch/xtensa/Makefile
@@ -0,0 +1,102 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+# Copyright (C) 2001 - 2005  Tensilica Inc.
+#
+# This file is included by the global makefile so that you can add your own
+# architecture-specific flags and dependencies. Remember to do have actions
+# for "archclean" and "archdep" for cleaning up and making dependencies for
+# this architecture
+
+# Core configuration.
+# (Use CPU=<xtensa_config> to use another default compiler.)
+
+cpu-$(CONFIG_XTENSA_CPU_LINUX_BE)	:= linux_be
+cpu-$(CONFIG_XTENSA_CPU_LINUX_CUSTOM)	:= linux_custom
+
+CPU = $(cpu-y)
+export CPU
+
+# Platform configuration
+
+platform-y					:= common
+platform-$(CONFIG_XTENSA_PLATFORM_XT2000)	:= xt2000
+platform-$(CONFIG_XTENSA_PLATFORM_ISS)		:= iss
+
+PLATFORM = $(platform-y)
+export PLATFORM
+
+#LDFLAGS_vmlinux		:= -T$(word 1,$(LINKSCRIPT))
+AFLAGS_vmlinux.lds.o	:= -Uxtensa
+CPPFLAGS		+= -Iarch/xtensa -Iinclude/asm  -mlongcalls  -g
+AFLAGS			+= -Iarch/xtensa -Iinclude/asm
+CPP			= $(CC) -E $(CFLAGS)
+
+cflags-y	+= -Iarch/xtensa -pipe -mlongcalls
+
+
+KBUILD_DEFCONFIG := common_defconfig
+
+# ramdisk/initrd support
+# You need a compressed ramdisk image, named ramdisk.gz in
+# arch/xtensa/boot/ramdisk
+
+core-$(CONFIG_EMBEDDED_RAMDISK)	+= arch/xtensa/boot/ramdisk/
+
+# Test for cross compiling
+
+ifneq ($(CPU),)
+  COMPILE_ARCH = $(shell uname -m)
+
+  ifneq ($(COMPILE_ARCH), xtensa)
+    ifndef CROSS_COMPILE
+      CROSS_COMPILE = xtensa_$(CPU)-
+    endif
+  endif
+endif
+
+#
+
+LIBGCC := $(shell $(CC) $(CFLAGS) -print-libgcc-file-name)
+
+head-y		:= arch/xtensa/kernel/head.o
+core-y		+= arch/xtensa/kernel/ \
+		   arch/xtensa/mm/ arch/xtensa/platform-$(PLATFORM)/
+libs-y		+= arch/xtensa/lib/ $(LIBGCC)
+
+boot	:= arch/xtensa/boot
+
+arch/xtensa/kernel/asm-offsets.s: \
+	arch/xtensa/kernel/asm-offsets.c \
+	include/asm-xtensa/.platform
+
+include/asm-xtensa/offsets.h: arch/xtensa/kernel/asm-offsets.s
+	$(call filechk,gen-asm-offsets)
+
+prepare: include/asm-xtensa/.platform include/asm-xtensa/offsets.h
+
+# Update machine cpu and platform symlinks if something which affects
+# them changed.
+
+include/asm-xtensa/.platform: $(wildcard include/config/arch/*.h)
+	@echo '  Setting up cpu ($(CPU)) and platform ($(PLATFORM)) symlinks'
+	$(Q)rm -f include/asm-xtensa/platform
+	$(Q)rm -f include/asm-xtensa/xtensa/config
+	$(Q)(cd include/asm-xtensa/; ln -sf platform-$(PLATFORM) platform)
+	$(Q)(cd include/asm-xtensa/xtensa; ln -sf config-$(CPU) config)
+
+all: zImage
+
+bzImage : zImage
+
+zImage zImage.initrd: vmlinux
+	$(Q)$(MAKE) $(build)=$(boot) $@
+
+CLEAN_FILES	+= arch/xtensa/vmlinux.lds include/asm-xtensa/offset.h
+
+define archhelp
+  @echo '* zImage      - Compressed kernel image (arch/xtensa/boot/images/zImage.*)'
+endef
+
diff --git a/arch/xtensa/boot/Makefile b/arch/xtensa/boot/Makefile
new file mode 100644
index 000000000000..260f456ccf0b
--- /dev/null
+++ b/arch/xtensa/boot/Makefile
@@ -0,0 +1,37 @@
+#
+# arch/xtensa/boot/Makefile
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+#
+
+
+CFLAGS 		+= -fno-builtin -Iarch/$(ARCH)/boot/include
+HOSTFLAGS	+= -Iarch/$(ARCH)/boot/include
+
+BIG_ENDIAN	:= $(shell echo -e "\#ifdef __XTENSA_EL__\nint little;\n\#else\nint big;\n\#endif" | $(CC) -E -|grep -c big)
+
+
+export CFLAGS
+export AFLAGS
+export BIG_ENDIAN
+
+# Subdirs for the boot loader(s)
+
+bootdir-$(CONFIG_XTENSA_PLATFORM_ISS)	 += boot-elf
+bootdir-$(CONFIG_XTENSA_PLATFORM_XT2000) += boot-redboot boot-elf
+
+subdir-y	:= lib/
+
+subdir-y	+= boot-elf/ boot-redboot/
+
+zImage zImage.initrd Image Image.initrd: $(bootdir-y)
+
+$(bootdir-y): $(addprefix $(obj)/,$(subdir-y)) \
+	      $(addprefix $(obj)/,$(host-progs))
+	$(Q)$(MAKE) $(build)=$(obj)/$@ $(MAKECMDGOALS)
+
+
+
diff --git a/arch/xtensa/boot/boot-elf/Makefile b/arch/xtensa/boot/boot-elf/Makefile
new file mode 100644
index 000000000000..f6ef6a369667
--- /dev/null
+++ b/arch/xtensa/boot/boot-elf/Makefile
@@ -0,0 +1,52 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+
+GZIP = gzip
+GZIP_FLAGS = -v9fc
+
+ifeq ($(BIG_ENDIAN),1)
+OBJCOPY_ARGS    := -O elf32-xtensa-be
+else
+OBJCOPY_ARGS    := -O elf32-xtensa-le
+endif
+
+export OBJCOPY_ARGS
+
+boot-y		:= bootstrap.o
+
+OBJS		:= $(addprefix $(obj)/,$(boot-y))
+
+Image:	vmlinux $(OBJS)
+	$(OBJCOPY) --strip-all -R .comment -R .xt.insn -O binary \
+		vmlinux vmlinux.tmp
+	$(OBJCOPY) $(OBJCOPY_ARGS) -R .comment \
+		--add-section image=vmlinux.tmp \
+		--set-section-flags image=contents,alloc,load,load,data \
+		$(OBJS) $@.tmp
+	$(LD) $(LDFLAGS) $(LDFLAGS_vmlinux) \
+		-T arch/$(ARCH)/boot/boot-elf/boot.ld \
+		-o arch/$(ARCH)/boot/$@.elf $@.tmp
+	rm -f $@.tmp vmlinux.tmp
+
+Image.initrd:	vmlinux $(OBJS)
+	$(OBJCOPY) --strip-all -R .comment -R .xt.insn -O binary \
+		--add-section .initrd=arch/$(ARCH)/boot/ramdisk \
+		--set-section-flags .initrd=contents,alloc,load,load,data \
+		vmlinux vmlinux.tmp
+	$(OBJCOPY) $(OBJCOPY_ARGS) -R .comment \
+		--add-section image=vmlinux.tmp \
+		--set-section-flags image=contents,alloc,load,load,data \
+		$(OBJS) $@.tmp
+	$(LD) $(LDFLAGS) $(LDFLAGS_vmlinux) \
+		-T arch/$(ARCH)/boot/boot-elf/boot.ld \
+		-o arch/$(ARCH)/boot/$@.elf $@.tmp
+	rm -f $@.tmp vmlinux.tmp
+
+
+zImage:	Image
+
+zImage.initrd:	Image.initrd
+
diff --git a/arch/xtensa/boot/boot-elf/boot.ld b/arch/xtensa/boot/boot-elf/boot.ld
new file mode 100644
index 000000000000..4ab06a0a7a6b
--- /dev/null
+++ b/arch/xtensa/boot/boot-elf/boot.ld
@@ -0,0 +1,71 @@
+OUTPUT_ARCH(xtensa)
+
+SECTIONS
+{
+	.start 0xD0000000 : { *(.start) }
+
+	.text 0xD0000000:
+	{
+		__reloc_start = . ;
+		_text_start = . ;
+		*(.literal .text.literal .text)
+		_text_end = . ;
+	}
+
+	.rodata ALIGN(0x04):
+	{
+		*(.rodata)
+		*(.rodata1)
+	}
+
+	.data ALIGN(0x04):
+	{
+		*(.data)
+		*(.data1)
+		*(.sdata)
+		*(.sdata2)
+		*(.got.plt)
+		*(.got)
+		*(.dynamic)
+	}
+
+	__reloc_end = . ;
+
+	.initrd ALIGN(0x10) :
+	{
+		boot_initrd_start = . ;
+		*(.initrd)
+		boot_initrd_end = .;
+	}
+
+	. = ALIGN(0x10);
+	__image_load = . ;
+	.image 0xd0001000:
+	{
+		_image_start = .;
+		*(image)
+		. = (. + 3) & ~ 3;
+		_image_end = .	;
+	}
+
+
+	.bss ((LOADADDR(.image) + SIZEOF(.image) + 3) & ~ 3):
+	{
+		__bss_start = .;
+		*(.sbss)
+		*(.scommon)
+		*(.dynbss)
+		*(.bss)
+		__bss_end = .;
+	}
+	_end = .;
+	_param_start = .;
+
+	.ResetVector.text 0xfe000020 :
+	{
+		*(.ResetVector.text)
+	}
+
+
+	PROVIDE (end = .);
+}
diff --git a/arch/xtensa/boot/boot-elf/bootstrap.S b/arch/xtensa/boot/boot-elf/bootstrap.S
new file mode 100644
index 000000000000..7cba94abdab8
--- /dev/null
+++ b/arch/xtensa/boot/boot-elf/bootstrap.S
@@ -0,0 +1,37 @@
+
+#include <xtensa/config/specreg.h>
+#include <xtensa/config/core.h>
+
+#include <linux/config.h>
+#include <asm/bootparam.h>
+
+
+/* ResetVector
+ */
+	.section        .ResetVector.text, "ax"
+	.global         _ResetVector
+_ResetVector:
+	_j reset
+	.align 4
+RomInitAddr:
+	.word 0xd0001000
+RomBootParam:
+	.word _bootparam
+reset:
+	l32r    a0, RomInitAddr
+	l32r	a2, RomBootParam
+	movi	a3, 0
+	movi	a4, 0
+	jx      a0
+
+	.align 4
+	.section .bootstrap.data, "aw"
+
+	.globl _bootparam
+_bootparam:
+	.short	BP_TAG_FIRST
+	.short	4
+	.long	BP_VERSION
+	.short	BP_TAG_LAST
+	.short	0
+	.long	0
diff --git a/arch/xtensa/boot/boot-redboot/Makefile b/arch/xtensa/boot/boot-redboot/Makefile
new file mode 100644
index 000000000000..ca8a68bc8472
--- /dev/null
+++ b/arch/xtensa/boot/boot-redboot/Makefile
@@ -0,0 +1,35 @@
+#
+# This file is subject to the terms and conditions of the GNU General Public
+# License.  See the file "COPYING" in the main directory of this archive
+# for more details.
+#
+
+GZIP = gzip
+GZIP_FLAGS = -v9fc
+ifeq ($(BIG_ENDIAN),1)
+OBJCOPY_ARGS 	:= -O elf32-xtensa-be
+else
+OBJCOPY_ARGS 	:= -O elf32-xtensa-le
+endif
+
+LD_ARGS	= -T $(obj)/boot.ld
+
+boot-y	:= bootstrap.o
+
+OBJS	:= $(addprefix $(obj)/,$(boot-y))
+LIBS	:= arch/$(ARCH)/boot/lib/lib.a arch/$(ARCH)/lib/lib.a
+
+LIBGCC := $(shell $(CC) $(CFLAGS) -print-libgcc-file-name)
+
+zImage: vmlinux $(OBJS) $(LIBS)
+	$(OBJCOPY) --strip-all -R .comment -R .xt.insn -O binary \
+		$(TOPDIR)/vmlinux vmlinux.tmp
+	gzip -vf9 vmlinux.tmp
+	$(OBJCOPY) $(OBJCOPY_ARGS) -R .comment \
+		--add-section image=vmlinux.tmp.gz \
+		--set-section-flags image=contents,alloc,load,load,data \
+		$(OBJS) $@.tmp
+	$(LD) $(LD_ARGS) -o $@.elf $@.tmp $(LIBS) -L/xtensa-elf/lib $(LIBGCC)
+	$(OBJCOPY) -S -O binary $@.elf arch/$(ARCH)/boot/images/$@.redboot
+#	rm -f $@.tmp $@.elf vmlinux.tmp.gz
+
diff --git a/arch/xtensa/boot/boot-redboot/boot.ld b/arch/xtensa/boot/boot-redboot/boot.ld
new file mode 100644
index 000000000000..65b726410e8a
--- /dev/null
+++ b/arch/xtensa/boot/boot-redboot/boot.ld
@@ -0,0 +1,66 @@
+OUTPUT_ARCH(xtensa)
+
+SECTIONS
+{
+	.start 0xD0200000 : { *(.start) }
+
+	.text :
+	{
+		__reloc_start = . ;
+		_text_start = . ;
+		*(.literal .text.literal .text)
+		_text_end = . ;
+	}
+
+	.rodata ALIGN(0x04):
+	{
+		*(.rodata)
+		*(.rodata1)
+	}
+
+	.data ALIGN(0x04):
+	{
+		*(.data)
+		*(.data1)
+		*(.sdata)
+		*(.sdata2)
+		*(.got.plt)
+		*(.got)
+		*(.dynamic)
+	}
+
+	__reloc_end = . ;
+
+	.initrd ALIGN(0x10) :
+	{
+		boot_initrd_start = . ;
+		*(.initrd)
+		boot_initrd_end = .;
+	}
+
+	. = ALIGN(0x10);
+	__image_load = . ;
+	.image 0xd0001000: AT(__image_load)
+	{
+		_image_start = .;
+		*(image)
+		. = (. + 3) & ~ 3;
+		_image_end = .	;
+	}
+
+
+	.bss ((LOADADDR(.image) + SIZEOF(.image) + 3) & ~ 3):
+	{
+		__bss_start = .;
+		*(.sbss)
+		*(.scommon)
+		*(.dynbss)
+		*(.bss)
+		__bss_end = .;
+	}
+	_end = .;
+	_param_start = .;
+
+
+	PROVIDE (end = .);
+}
diff --git a/arch/xtensa/boot/boot-redboot/bootstrap.S b/arch/xtensa/boot/boot-redboot/bootstrap.S
new file mode 100644
index 000000000000..ee636b0da81c
--- /dev/null
+++ b/arch/xtensa/boot/boot-redboot/bootstrap.S
@@ -0,0 +1,246 @@
+
+#define _ASMLANGUAGE
+#include <xtensa/config/specreg.h>
+#include <xtensa/config/core.h>
+#include <xtensa/cacheasm.h>
+
+	/*
+	 * RB-Data: RedBoot data/bss
+	 * P:	    Boot-Parameters
+	 * L:	    Kernel-Loader
+	 *
+	 * The Linux-Kernel image including the loader must be loaded
+	 * to a position so that the kernel and the boot parameters
+	 * can fit in the space before the load address.
+	 *  ______________________________________________________
+	 * |_RB-Data_|_P_|__________|_L_|___Linux-Kernel___|______|
+	 *                          ^
+	 *                          ^ Load address
+	 *  ______________________________________________________
+	 * |___Linux-Kernel___|_P_|_L_|___________________________|
+	 *
+	 * The loader copies the parameter to the position that will
+	 * be the end of the kernel and itself to the end of the
+	 * parameter list.
+	 */
+
+/* Make sure we have enough space for the 'uncompressor' */
+
+#define STACK_SIZE 32768
+#define HEAP_SIZE (131072*4)
+
+	# a2: Parameter list
+	# a3: Size of parameter list
+
+	.section .start, "ax"
+
+	.globl __start
+	/* this must be the first byte of the loader! */
+__start:
+	entry	sp, 32		# we do not intend to return
+	_call0	_start
+__start_a0:
+	.align 4
+
+	.section .text, "ax"
+	.begin literal_prefix .text
+
+	/* put literals in here! */
+
+	.globl _start
+_start:
+
+	/* 'reset' window registers */
+
+	movi	a4, 1
+	wsr	a4, PS
+	rsync
+
+	rsr	a5, WINDOWBASE
+	ssl	a5
+	sll	a4, a4
+	wsr	a4, WINDOWSTART
+	rsync
+
+	movi	a4, 0x00040000
+	wsr	a4, PS
+	rsync
+
+	/* copy the loader to its address
+	 * Note: The loader itself is a very small piece, so we assume we
+	 *       don't partially overlap. We also assume (even more important)
+	 *	 that the kernel image is out of the way. Usually, when the
+	 *	 load address of this image is not at an arbitrary address,
+	 *	 but aligned to some 10K's we shouldn't overlap.
+	 */
+
+	/* Note: The assembler cannot relax "addi a0, a0, ..." to an
+	   l32r, so we load to a4 first. */
+
+	addi	a4, a0, __start - __start_a0
+	mov	a0, a4
+	movi	a4, __start
+	movi	a5, __reloc_end
+
+	# a0: address where this code has been loaded
+	# a4: compiled address of __start
+	# a5: compiled end address
+
+	mov.n	a7, a0
+	mov.n	a8, a4
+
+1:
+	l32i	a10, a7, 0
+	l32i	a11, a7, 4
+	s32i	a10, a8, 0
+	s32i	a11, a8, 4
+	l32i	a10, a7, 8
+	l32i	a11, a7, 12
+	s32i	a10, a8, 8
+	s32i	a11, a8, 12
+	addi	a8, a8, 16
+	addi	a7, a7, 16
+	blt	a8, a5, 1b
+
+
+	/* We have to flush and invalidate the caches here before we jump. */
+
+#if XCHAL_DCACHE_IS_WRITEBACK
+	dcache_writeback_all  a5, a6
+#endif
+	icache_invalidate_all a5, a6
+
+	movi	a11, _reloc
+	jx	a11
+
+	.globl _reloc
+_reloc:
+
+	/* RedBoot is now at the end of the memory, so we don't have
+	 * to copy the parameter list. Keep the code around; in case
+	 * we need it again. */
+#if 0
+	# a0: load address
+	# a2: start address of parameter list
+	# a3: length of parameter list
+	# a4: __start
+
+	/* copy the parameter list out of the way */
+
+	movi	a6, _param_start
+	add	a3, a2, a3
+2:
+	l32i	a8, a2, 0
+	s32i	a8, a6, 0
+	addi	a2, a2, 4
+	addi	a6, a6, 4
+	blt	a2, a3, 2b
+#endif
+
+	/* clear BSS section */
+	movi	a6, __bss_start
+	movi	a7, __bss_end
+	movi.n	a5, 0
+3:
+	s32i	a5, a6, 0
+	addi	a6, a6, 4
+	blt	a6, a7, 3b
+
+	movi	a5, -16
+	movi	a1, _stack + STACK_SIZE
+	and	a1, a1, a5
+
+	/* Uncompress the kernel */
+
+	# a0: load address
+	# a2: boot parameter
+	# a4: __start
+
+	movi	a3, __image_load
+	sub	a4, a3, a4
+	add	a8, a0, a4
+
+	# a1  Stack
+	# a8(a4)  Load address of the image
+
+	movi	a6, _image_start
+	movi	a10, _image_end
+	movi	a7, 0x1000000
+	sub	a11, a10, a6
+	movi	a9, complen
+	s32i	a11, a9, 0
+
+	movi	a0, 0
+
+	# a6 destination
+	# a7 maximum size of destination
+	# a8 source
+	# a9 ptr to length
+
+	.extern gunzip
+	movi	a4, gunzip
+	beqz	a4, 1f
+
+	callx4	a4
+
+	j	2f
+
+
+	# a6 destination start
+	# a7 maximum size of destination
+	# a8 source start
+	# a9 ptr to length
+	# a10 destination end
+
+1:
+        l32i    a9, a8, 0
+        l32i    a11, a8, 4
+        s32i    a9, a6, 0
+        s32i    a11, a6, 4
+        l32i    a9, a8, 8
+        l32i    a11, a8, 12
+        s32i    a9, a6, 8
+        s32i    a11, a6, 12
+        addi    a6, a6, 16
+        addi    a8, a8, 16
+        blt     a6, a10, 1b
+
+
+	/* jump to the kernel */
+2:
+#if XCHAL_DCACHE_IS_WRITEBACK
+	dcache_writeback_all a5, a6
+#endif
+	icache_invalidate_all a5, a6
+
+	movi	a5, __start
+	movi	a3, boot_initrd_start
+	movi	a4, boot_initrd_end
+	sub	a3, a3, a5
+	sub	a4, a4, a5
+	add	a3, a0, a3
+	add	a4, a0, a4
+
+	# a2  Boot parameter list
+	# a3  initrd_start (virtual load address)
+	# a4  initrd_end   (virtual load address)
+
+	movi	a0, _image_start
+	jx	a0
+
+	.align 16
+	.data
+	.globl avail_ram
+avail_ram:
+	.long	_heap
+	.globl end_avail
+end_avail:
+	.long	_heap + HEAP_SIZE
+
+	.comm _stack, STACK_SIZE
+	.comm _heap, HEAP_SIZE
+
+	.globl end_avail
+	.comm complen, 4
+
+	.end	literal_prefix
diff --git a/arch/xtensa/boot/include/zlib.h b/arch/xtensa/boot/include/zlib.h
new file mode 100644
index 000000000000..ea29b6237852
--- /dev/null
+++ b/arch/xtensa/boot/include/zlib.h
@@ -0,0 +1,433 @@
+/*
+ * BK Id: SCCS/s.zlib.h 1.8 05/18/01 15:17:23 cort
+ */
+/*
+ * This file is derived from zlib.h and zconf.h from the zlib-0.95
+ * distribution by Jean-loup Gailly and Mark Adler, with some additions
+ * by Paul Mackerras to aid in implementing Deflate compression and
+ * decompression for PPP packets.
+ */
+
+/*
+ *  ==FILEVERSION 960122==
+ *
+ * This marker is used by the Linux installation script to determine
+ * whether an up-to-date version of this file is already installed.
+ */
+
+/* zlib.h -- interface of the 'zlib' general purpose compression library
+  version 0.95, Aug 16th, 1995.
+
+  Copyright (C) 1995 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  gzip@prep.ai.mit.edu    madler@alumni.caltech.edu
+ */
+
+#ifndef _ZLIB_H
+#define _ZLIB_H
+
+/* #include "zconf.h" */	/* included directly here */
+
+/* zconf.h -- configuration of the zlib compression library
+ * Copyright (C) 1995 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* From: zconf.h,v 1.12 1995/05/03 17:27:12 jloup Exp */
+
+/*
+     The library does not install any signal handler. It is recommended to
+  add at least a handler for SIGSEGV when decompressing; the library checks
+  the consistency of the input data whenever possible but may go nuts
+  for some forms of corrupted input.
+ */
+
+/*
+ * Compile with -DMAXSEG_64K if the alloc function cannot allocate more
+ * than 64k bytes at a time (needed on systems with 16-bit int).
+ * Compile with -DUNALIGNED_OK if it is OK to access shorts or ints
+ * at addresses which are not a multiple of their size.
+ * Under DOS, -DFAR=far or -DFAR=__far may be needed.
+ */
+
+#ifndef STDC
+#  if defined(MSDOS) || defined(__STDC__) || defined(__cplusplus)
+#    define STDC
+#  endif
+#endif
+
+#ifdef	__MWERKS__ /* Metrowerks CodeWarrior declares fileno() in unix.h */
+#  include <unix.h>
+#endif
+
+/* Maximum value for memLevel in deflateInit2 */
+#ifndef MAX_MEM_LEVEL
+#  ifdef MAXSEG_64K
+#    define MAX_MEM_LEVEL 8
+#  else
+#    define MAX_MEM_LEVEL 9
+#  endif
+#endif
+
+#ifndef FAR
+#  define FAR
+#endif
+
+/* Maximum value for windowBits in deflateInit2 and inflateInit2 */
+#ifndef MAX_WBITS
+#  define MAX_WBITS   15 /* 32K LZ77 window */
+#endif
+
+/* The memory requirements for deflate are (in bytes):
+            1 << (windowBits+2)   +  1 << (memLevel+9)
+ that is: 128K for windowBits=15  +  128K for memLevel = 8  (default values)
+ plus a few kilobytes for small objects. For example, if you want to reduce
+ the default memory requirements from 256K to 128K, compile with
+     make CFLAGS="-O -DMAX_WBITS=14 -DMAX_MEM_LEVEL=7"
+ Of course this will generally degrade compression (there's no free lunch).
+
+   The memory requirements for inflate are (in bytes) 1 << windowBits
+ that is, 32K for windowBits=15 (default value) plus a few kilobytes
+ for small objects.
+*/
+
+                        /* Type declarations */
+
+#ifndef OF /* function prototypes */
+#  ifdef STDC
+#    define OF(args)  args
+#  else
+#    define OF(args)  ()
+#  endif
+#endif
+
+typedef unsigned char  Byte;  /* 8 bits */
+typedef unsigned int   uInt;  /* 16 bits or more */
+typedef unsigned long  uLong; /* 32 bits or more */
+
+typedef Byte FAR Bytef;
+typedef char FAR charf;
+typedef int FAR intf;
+typedef uInt FAR uIntf;
+typedef uLong FAR uLongf;
+
+#ifdef STDC
+   typedef void FAR *voidpf;
+   typedef void     *voidp;
+#else
+   typedef Byte FAR *voidpf;
+   typedef Byte     *voidp;
+#endif
+
+/* end of original zconf.h */
+
+#define ZLIB_VERSION "0.95P"
+
+/*
+     The 'zlib' compression library provides in-memory compression and
+  decompression functions, including integrity checks of the uncompressed
+  data.  This version of the library supports only one compression method
+  (deflation) but other algorithms may be added later and will have the same
+  stream interface.
+
+     For compression the application must provide the output buffer and
+  may optionally provide the input buffer for optimization. For decompression,
+  the application must provide the input buffer and may optionally provide
+  the output buffer for optimization.
+
+     Compression can be done in a single step if the buffers are large
+  enough (for example if an input file is mmap'ed), or can be done by
+  repeated calls of the compression function.  In the latter case, the
+  application must provide more input and/or consume the output
+  (providing more output space) before each call.
+*/
+
+typedef voidpf (*alloc_func) OF((voidpf opaque, uInt items, uInt size));
+typedef void   (*free_func)  OF((voidpf opaque, voidpf address, uInt nbytes));
+
+struct internal_state;
+
+typedef struct z_stream_s {
+    Bytef    *next_in;  /* next input byte */
+    uInt     avail_in;  /* number of bytes available at next_in */
+    uLong    total_in;  /* total nb of input bytes read so far */
+
+    Bytef    *next_out; /* next output byte should be put there */
+    uInt     avail_out; /* remaining free space at next_out */
+    uLong    total_out; /* total nb of bytes output so far */
+
+    char     *msg;      /* last error message, NULL if no error */
+    struct internal_state FAR *state; /* not visible by applications */
+
+    alloc_func zalloc;  /* used to allocate the internal state */
+    free_func  zfree;   /* used to free the internal state */
+    voidp      opaque;  /* private data object passed to zalloc and zfree */
+
+    Byte     data_type; /* best guess about the data type: ascii or binary */
+
+} z_stream;
+
+/*
+   The application must update next_in and avail_in when avail_in has
+   dropped to zero. It must update next_out and avail_out when avail_out
+   has dropped to zero. The application must initialize zalloc, zfree and
+   opaque before calling the init function. All other fields are set by the
+   compression library and must not be updated by the application.
+
+   The opaque value provided by the application will be passed as the first
+   parameter for calls of zalloc and zfree. This can be useful for custom
+   memory management. The compression library attaches no meaning to the
+   opaque value.
+
+   zalloc must return Z_NULL if there is not enough memory for the object.
+   On 16-bit systems, the functions zalloc and zfree must be able to allocate
+   exactly 65536 bytes, but will not be required to allocate more than this
+   if the symbol MAXSEG_64K is defined (see zconf.h). WARNING: On MSDOS,
+   pointers returned by zalloc for objects of exactly 65536 bytes *must*
+   have their offset normalized to zero. The default allocation function
+   provided by this library ensures this (see zutil.c). To reduce memory
+   requirements and avoid any allocation of 64K objects, at the expense of
+   compression ratio, compile the library with -DMAX_WBITS=14 (see zconf.h).
+
+   The fields total_in and total_out can be used for statistics or
+   progress reports. After compression, total_in holds the total size of
+   the uncompressed data and may be saved for use in the decompressor
+   (particularly if the decompressor wants to decompress everything in
+   a single step).
+*/
+
+                        /* constants */
+
+#define Z_NO_FLUSH      0
+#define Z_PARTIAL_FLUSH 1
+#define Z_FULL_FLUSH    2
+#define Z_SYNC_FLUSH    3 /* experimental: partial_flush + byte align */
+#define Z_FINISH        4
+#define Z_PACKET_FLUSH	5
+/* See deflate() below for the usage of these constants */
+
+#define Z_OK            0
+#define Z_STREAM_END    1
+#define Z_ERRNO        (-1)
+#define Z_STREAM_ERROR (-2)
+#define Z_DATA_ERROR   (-3)
+#define Z_MEM_ERROR    (-4)
+#define Z_BUF_ERROR    (-5)
+/* error codes for the compression/decompression functions */
+
+#define Z_BEST_SPEED             1
+#define Z_BEST_COMPRESSION       9
+#define Z_DEFAULT_COMPRESSION  (-1)
+/* compression levels */
+
+#define Z_FILTERED            1
+#define Z_HUFFMAN_ONLY        2
+#define Z_DEFAULT_STRATEGY    0
+
+#define Z_BINARY   0
+#define Z_ASCII    1
+#define Z_UNKNOWN  2
+/* Used to set the data_type field */
+
+#define Z_NULL  0  /* for initializing zalloc, zfree, opaque */
+
+extern char *zlib_version;
+/* The application can compare zlib_version and ZLIB_VERSION for consistency.
+   If the first character differs, the library code actually used is
+   not compatible with the zlib.h header file used by the application.
+ */
+
+                        /* basic functions */
+
+extern int inflateInit OF((z_stream *strm));
+/*
+     Initializes the internal stream state for decompression. The fields
+   zalloc and zfree must be initialized before by the caller.  If zalloc and
+   zfree are set to Z_NULL, inflateInit updates them to use default allocation
+   functions.
+
+     inflateInit returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory.  msg is set to null if there is no error message.
+   inflateInit does not perform any decompression: this will be done by
+   inflate().
+*/
+
+
+extern int inflate OF((z_stream *strm, int flush));
+/*
+  Performs one or both of the following actions:
+
+  - Decompress more input starting at next_in and update next_in and avail_in
+    accordingly. If not all input can be processed (because there is not
+    enough room in the output buffer), next_in is updated and processing
+    will resume at this point for the next call of inflate().
+
+  - Provide more output starting at next_out and update next_out and avail_out
+    accordingly.  inflate() always provides as much output as possible
+    (until there is no more input data or no more space in the output buffer).
+
+  Before the call of inflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming
+  more output, and updating the next_* and avail_* values accordingly.
+  The application can consume the uncompressed output when it wants, for
+  example when the output buffer is full (avail_out == 0), or after each
+  call of inflate().
+
+    If the parameter flush is set to Z_PARTIAL_FLUSH or Z_PACKET_FLUSH,
+  inflate flushes as much output as possible to the output buffer. The
+  flushing behavior of inflate is not specified for values of the flush
+  parameter other than Z_PARTIAL_FLUSH, Z_PACKET_FLUSH or Z_FINISH, but the
+  current implementation actually flushes as much output as possible
+  anyway.  For Z_PACKET_FLUSH, inflate checks that once all the input data
+  has been consumed, it is expecting to see the length field of a stored
+  block; if not, it returns Z_DATA_ERROR.
+
+    inflate() should normally be called until it returns Z_STREAM_END or an
+  error. However if all decompression is to be performed in a single step
+  (a single call of inflate), the parameter flush should be set to
+  Z_FINISH. In this case all pending input is processed and all pending
+  output is flushed; avail_out must be large enough to hold all the
+  uncompressed data. (The size of the uncompressed data may have been saved
+  by the compressor for this purpose.) The next operation on this stream must
+  be inflateEnd to deallocate the decompression state. The use of Z_FINISH
+  is never required, but can be used to inform inflate that a faster routine
+  may be used for the single inflate() call.
+
+    inflate() returns Z_OK if some progress has been made (more input
+  processed or more output produced), Z_STREAM_END if the end of the
+  compressed data has been reached and all uncompressed output has been
+  produced, Z_DATA_ERROR if the input data was corrupted, Z_STREAM_ERROR if
+  the stream structure was inconsistent (for example if next_in or next_out
+  was NULL), Z_MEM_ERROR if there was not enough memory, Z_BUF_ERROR if no
+  progress is possible or if there was not enough room in the output buffer
+  when Z_FINISH is used. In the Z_DATA_ERROR case, the application may then
+  call inflateSync to look for a good compression block.  */
+
+
+extern int inflateEnd OF((z_stream *strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any
+   pending output.
+
+     inflateEnd returns Z_OK if success, Z_STREAM_ERROR if the stream state
+   was inconsistent. In the error case, msg may be set but then points to a
+   static string (which must not be deallocated).
+*/
+
+                        /* advanced functions */
+
+extern int inflateInit2 OF((z_stream *strm,
+                            int  windowBits));
+/*
+     This is another version of inflateInit with more compression options. The
+   fields next_out, zalloc and zfree must be initialized before by the caller.
+
+     The windowBits parameter is the base two logarithm of the maximum window
+   size (the size of the history buffer).  It should be in the range 8..15 for
+   this version of the library (the value 16 will be allowed soon). The
+   default value is 15 if inflateInit is used instead. If a compressed stream
+   with a larger window size is given as input, inflate() will return with
+   the error code Z_DATA_ERROR instead of trying to allocate a larger window.
+
+     If next_out is not null, the library will use this buffer for the history
+   buffer; the buffer must either be large enough to hold the entire output
+   data, or have at least 1<<windowBits bytes.  If next_out is null, the
+   library will allocate its own buffer (and leave next_out null). next_in
+   need not be provided here but must be provided by the application for the
+   next call of inflate().
+
+     If the history buffer is provided by the application, next_out must
+   never be changed by the application since the decompressor maintains
+   history information inside this buffer from call to call; the application
+   can only reset next_out to the beginning of the history buffer when
+   avail_out is zero and all output has been consumed.
+
+      inflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was
+   not enough memory, Z_STREAM_ERROR if a parameter is invalid (such as
+   windowBits < 8). msg is set to null if there is no error message.
+   inflateInit2 does not perform any decompression: this will be done by
+   inflate().
+*/
+
+extern int inflateSync OF((z_stream *strm));
+/*
+    Skips invalid compressed data until the special marker (see deflate()
+  above) can be found, or until all available input is skipped. No output
+  is provided.
+
+    inflateSync returns Z_OK if the special marker has been found, Z_BUF_ERROR
+  if no more input was provided, Z_DATA_ERROR if no marker has been found,
+  or Z_STREAM_ERROR if the stream structure was inconsistent. In the success
+  case, the application may save the current current value of total_in which
+  indicates where valid compressed data was found. In the error case, the
+  application may repeatedly call inflateSync, providing more input each time,
+  until success or end of the input data.
+*/
+
+extern int inflateReset OF((z_stream *strm));
+/*
+     This function is equivalent to inflateEnd followed by inflateInit,
+   but does not free and reallocate all the internal decompression state.
+   The stream will keep attributes that may have been set by inflateInit2.
+
+      inflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being NULL).
+*/
+
+extern int inflateIncomp OF((z_stream *strm));
+/*
+     This function adds the data at next_in (avail_in bytes) to the output
+   history without performing any output.  There must be no pending output,
+   and the decompressor must be expecting to see the start of a block.
+   Calling this function is equivalent to decompressing a stored block
+   containing the data at next_in (except that the data is not output).
+*/
+
+                        /* checksum functions */
+
+/*
+     This function is not related to compression but is exported
+   anyway because it might be useful in applications using the
+   compression library.
+*/
+
+extern uLong adler32 OF((uLong adler, Bytef *buf, uInt len));
+
+/*
+     Update a running Adler-32 checksum with the bytes buf[0..len-1] and
+   return the updated checksum. If buf is NULL, this function returns
+   the required initial value for the checksum.
+   An Adler-32 checksum is almost as reliable as a CRC32 but can be computed
+   much faster. Usage example:
+
+     uLong adler = adler32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       adler = adler32(adler, buffer, length);
+     }
+     if (adler != original_adler) error();
+*/
+
+#ifndef _Z_UTIL_H
+    struct internal_state {int dummy;}; /* hack for buggy compilers */
+#endif
+
+#endif /* _ZLIB_H */
diff --git a/arch/xtensa/boot/lib/Makefile b/arch/xtensa/boot/lib/Makefile
new file mode 100644
index 000000000000..c0a74dc3a0df
--- /dev/null
+++ b/arch/xtensa/boot/lib/Makefile
@@ -0,0 +1,6 @@
+#
+# Makefile for some libs needed by zImage.
+#
+
+
+lib-y := zlib.o zmem.o
diff --git a/arch/xtensa/boot/lib/memcpy.S b/arch/xtensa/boot/lib/memcpy.S
new file mode 100644
index 000000000000..a029f5df2d5c
--- /dev/null
+++ b/arch/xtensa/boot/lib/memcpy.S
@@ -0,0 +1,36 @@
+/*
+ *  arch/xtensa/lib/memcpy.S
+ *
+ *  ANSI C standard library function memcpy
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file "COPYING" in the main directory of
+ *  this archive for more details.
+ *
+ *  Copyright (C) 2002 Tensilica Inc.
+ */
+
+#define _ASMLANGUAGE
+#include <xtensa/config/core.h>
+
+.text
+.align	4
+.global	bcopy
+.type   bcopy,@function
+bcopy:
+        movi    a14, xthal_bcopy        // a14 safe to use regardless of whether caller
+                                        //   used call4 or call8 (can't have used call12)
+        jx      a14                     // let the Core HAL do the work
+
+.text
+.align	4
+.global	memcpy
+.type   memcpy,@function
+memcpy:
+.global	memmove
+.type   memmove,@function
+memmove:
+        movi    a14, xthal_memcpy      // a14 safe to use regardless of whether caller
+                                        //   used call4 or call8 (can't have used call12)
+        jx      a14                     // let the Core HAL do the work
+
diff --git a/arch/xtensa/boot/lib/zlib.c b/arch/xtensa/boot/lib/zlib.c
new file mode 100644
index 000000000000..e3859f631077
--- /dev/null
+++ b/arch/xtensa/boot/lib/zlib.c
@@ -0,0 +1,2150 @@
+/*
+ * BK Id: SCCS/s.zlib.c 1.8 05/18/01 15:17:24 cort
+ */
+/*
+ * This file is derived from various .h and .c files from the zlib-0.95
+ * distribution by Jean-loup Gailly and Mark Adler, with some additions
+ * by Paul Mackerras to aid in implementing Deflate compression and
+ * decompression for PPP packets.  See zlib.h for conditions of
+ * distribution and use.
+ *
+ * Changes that have been made include:
+ * - changed functions not used outside this file to "local"
+ * - added minCompression parameter to deflateInit2
+ * - added Z_PACKET_FLUSH (see zlib.h for details)
+ * - added inflateIncomp
+ *
+ */
+
+/*+++++*/
+/* zutil.h -- internal interface and configuration of the compression library
+ * Copyright (C) 1995 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* From: zutil.h,v 1.9 1995/05/03 17:27:12 jloup Exp */
+
+#define _Z_UTIL_H
+
+#include "zlib.h"
+
+#ifndef local
+#  define local static
+#endif
+/* compile with -Dlocal if your debugger can't find static symbols */
+
+#define FAR
+
+typedef unsigned char  uch;
+typedef uch FAR uchf;
+typedef unsigned short ush;
+typedef ush FAR ushf;
+typedef unsigned long  ulg;
+
+extern char *z_errmsg[]; /* indexed by 1-zlib_error */
+
+#define ERR_RETURN(strm,err) return (strm->msg=z_errmsg[1-err], err)
+/* To be used only when the state is known to be valid */
+
+#ifndef NULL
+#define NULL	((void *) 0)
+#endif
+
+        /* common constants */
+
+#define DEFLATED   8
+
+#ifndef DEF_WBITS
+#  define DEF_WBITS MAX_WBITS
+#endif
+/* default windowBits for decompression. MAX_WBITS is for compression only */
+
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+/* default memLevel */
+
+#define STORED_BLOCK 0
+#define STATIC_TREES 1
+#define DYN_TREES    2
+/* The three kinds of block type */
+
+#define MIN_MATCH  3
+#define MAX_MATCH  258
+/* The minimum and maximum match lengths */
+
+         /* functions */
+
+#include <linux/string.h>
+#define zmemcpy memcpy
+#define zmemzero(dest, len)	memset(dest, 0, len)
+
+/* Diagnostic functions */
+#ifdef DEBUG_ZLIB
+#  include <stdio.h>
+#  ifndef verbose
+#    define verbose 0
+#  endif
+#  define Assert(cond,msg) {if(!(cond)) z_error(msg);}
+#  define Trace(x) fprintf x
+#  define Tracev(x) {if (verbose) fprintf x ;}
+#  define Tracevv(x) {if (verbose>1) fprintf x ;}
+#  define Tracec(c,x) {if (verbose && (c)) fprintf x ;}
+#  define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;}
+#else
+#  define Assert(cond,msg)
+#  define Trace(x)
+#  define Tracev(x)
+#  define Tracevv(x)
+#  define Tracec(c,x)
+#  define Tracecv(c,x)
+#endif
+
+
+typedef uLong (*check_func) OF((uLong check, Bytef *buf, uInt len));
+
+/* voidpf zcalloc OF((voidpf opaque, unsigned items, unsigned size)); */
+/* void   zcfree  OF((voidpf opaque, voidpf ptr)); */
+
+#define ZALLOC(strm, items, size) \
+           (*((strm)->zalloc))((strm)->opaque, (items), (size))
+#define ZFREE(strm, addr, size)	\
+	   (*((strm)->zfree))((strm)->opaque, (voidpf)(addr), (size))
+#define TRY_FREE(s, p, n) {if (p) ZFREE(s, p, n);}
+
+/* deflate.h -- internal compression state
+ * Copyright (C) 1995 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/*+++++*/
+/* infblock.h -- header to use infblock.c
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+struct inflate_blocks_state;
+typedef struct inflate_blocks_state FAR inflate_blocks_statef;
+
+local inflate_blocks_statef * inflate_blocks_new OF((
+    z_stream *z,
+    check_func c,               /* check function */
+    uInt w));                   /* window size */
+
+local int inflate_blocks OF((
+    inflate_blocks_statef *,
+    z_stream *,
+    int));                      /* initial return code */
+
+local void inflate_blocks_reset OF((
+    inflate_blocks_statef *,
+    z_stream *,
+    uLongf *));                  /* check value on output */
+
+local int inflate_blocks_free OF((
+    inflate_blocks_statef *,
+    z_stream *,
+    uLongf *));                  /* check value on output */
+
+local int inflate_addhistory OF((
+    inflate_blocks_statef *,
+    z_stream *));
+
+local int inflate_packet_flush OF((
+    inflate_blocks_statef *));
+
+/*+++++*/
+/* inftrees.h -- header to use inftrees.c
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* Huffman code lookup table entry--this entry is four bytes for machines
+   that have 16-bit pointers (e.g. PC's in the small or medium model). */
+
+typedef struct inflate_huft_s FAR inflate_huft;
+
+struct inflate_huft_s {
+  union {
+    struct {
+      Byte Exop;        /* number of extra bits or operation */
+      Byte Bits;        /* number of bits in this code or subcode */
+    } what;
+    uInt Nalloc;	/* number of these allocated here */
+    Bytef *pad;         /* pad structure to a power of 2 (4 bytes for */
+  } word;               /*  16-bit, 8 bytes for 32-bit machines) */
+  union {
+    uInt Base;          /* literal, length base, or distance base */
+    inflate_huft *Next; /* pointer to next level of table */
+  } more;
+};
+
+#ifdef DEBUG_ZLIB
+  local uInt inflate_hufts;
+#endif
+
+local int inflate_trees_bits OF((
+    uIntf *,                    /* 19 code lengths */
+    uIntf *,                    /* bits tree desired/actual depth */
+    inflate_huft * FAR *,       /* bits tree result */
+    z_stream *));               /* for zalloc, zfree functions */
+
+local int inflate_trees_dynamic OF((
+    uInt,                       /* number of literal/length codes */
+    uInt,                       /* number of distance codes */
+    uIntf *,                    /* that many (total) code lengths */
+    uIntf *,                    /* literal desired/actual bit depth */
+    uIntf *,                    /* distance desired/actual bit depth */
+    inflate_huft * FAR *,       /* literal/length tree result */
+    inflate_huft * FAR *,       /* distance tree result */
+    z_stream *));               /* for zalloc, zfree functions */
+
+local int inflate_trees_fixed OF((
+    uIntf *,                    /* literal desired/actual bit depth */
+    uIntf *,                    /* distance desired/actual bit depth */
+    inflate_huft * FAR *,       /* literal/length tree result */
+    inflate_huft * FAR *));     /* distance tree result */
+
+local int inflate_trees_free OF((
+    inflate_huft *,             /* tables to free */
+    z_stream *));               /* for zfree function */
+
+
+/*+++++*/
+/* infcodes.h -- header to use infcodes.c
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+struct inflate_codes_state;
+typedef struct inflate_codes_state FAR inflate_codes_statef;
+
+local inflate_codes_statef *inflate_codes_new OF((
+    uInt, uInt,
+    inflate_huft *, inflate_huft *,
+    z_stream *));
+
+local int inflate_codes OF((
+    inflate_blocks_statef *,
+    z_stream *,
+    int));
+
+local void inflate_codes_free OF((
+    inflate_codes_statef *,
+    z_stream *));
+
+
+/*+++++*/
+/* inflate.c -- zlib interface to inflate modules
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* inflate private state */
+struct internal_state {
+
+  /* mode */
+  enum {
+      METHOD,   /* waiting for method byte */
+      FLAG,     /* waiting for flag byte */
+      BLOCKS,   /* decompressing blocks */
+      CHECK4,   /* four check bytes to go */
+      CHECK3,   /* three check bytes to go */
+      CHECK2,   /* two check bytes to go */
+      CHECK1,   /* one check byte to go */
+      DONE,     /* finished check, done */
+      BAD}      /* got an error--stay here */
+    mode;               /* current inflate mode */
+
+  /* mode dependent information */
+  union {
+    uInt method;        /* if FLAGS, method byte */
+    struct {
+      uLong was;                /* computed check value */
+      uLong need;               /* stream check value */
+    } check;            /* if CHECK, check values to compare */
+    uInt marker;        /* if BAD, inflateSync's marker bytes count */
+  } sub;        /* submode */
+
+  /* mode independent information */
+  int  nowrap;          /* flag for no wrapper */
+  uInt wbits;           /* log2(window size)  (8..15, defaults to 15) */
+  inflate_blocks_statef
+    *blocks;            /* current inflate_blocks state */
+
+};
+
+
+int inflateReset(z)
+z_stream *z;
+{
+  uLong c;
+
+  if (z == Z_NULL || z->state == Z_NULL)
+    return Z_STREAM_ERROR;
+  z->total_in = z->total_out = 0;
+  z->msg = Z_NULL;
+  z->state->mode = z->state->nowrap ? BLOCKS : METHOD;
+  inflate_blocks_reset(z->state->blocks, z, &c);
+  Trace((stderr, "inflate: reset\n"));
+  return Z_OK;
+}
+
+
+int inflateEnd(z)
+z_stream *z;
+{
+  uLong c;
+
+  if (z == Z_NULL || z->state == Z_NULL || z->zfree == Z_NULL)
+    return Z_STREAM_ERROR;
+  if (z->state->blocks != Z_NULL)
+    inflate_blocks_free(z->state->blocks, z, &c);
+  ZFREE(z, z->state, sizeof(struct internal_state));
+  z->state = Z_NULL;
+  Trace((stderr, "inflate: end\n"));
+  return Z_OK;
+}
+
+
+int inflateInit2(z, w)
+z_stream *z;
+int w;
+{
+  /* initialize state */
+  if (z == Z_NULL)
+    return Z_STREAM_ERROR;
+/*  if (z->zalloc == Z_NULL) z->zalloc = zcalloc; */
+/*  if (z->zfree == Z_NULL) z->zfree = zcfree; */
+  if ((z->state = (struct internal_state FAR *)
+       ZALLOC(z,1,sizeof(struct internal_state))) == Z_NULL)
+    return Z_MEM_ERROR;
+  z->state->blocks = Z_NULL;
+
+  /* handle undocumented nowrap option (no zlib header or check) */
+  z->state->nowrap = 0;
+  if (w < 0)
+  {
+    w = - w;
+    z->state->nowrap = 1;
+  }
+
+  /* set window size */
+  if (w < 8 || w > 15)
+  {
+    inflateEnd(z);
+    return Z_STREAM_ERROR;
+  }
+  z->state->wbits = (uInt)w;
+
+  /* create inflate_blocks state */
+  if ((z->state->blocks =
+       inflate_blocks_new(z, z->state->nowrap ? Z_NULL : adler32, 1 << w))
+      == Z_NULL)
+  {
+    inflateEnd(z);
+    return Z_MEM_ERROR;
+  }
+  Trace((stderr, "inflate: allocated\n"));
+
+  /* reset state */
+  inflateReset(z);
+  return Z_OK;
+}
+
+
+int inflateInit(z)
+z_stream *z;
+{
+  return inflateInit2(z, DEF_WBITS);
+}
+
+
+#define NEEDBYTE {if(z->avail_in==0)goto empty;r=Z_OK;}
+#define NEXTBYTE (z->avail_in--,z->total_in++,*z->next_in++)
+
+int inflate(z, f)
+z_stream *z;
+int f;
+{
+  int r;
+  uInt b;
+
+  if (z == Z_NULL || z->next_in == Z_NULL)
+    return Z_STREAM_ERROR;
+  r = Z_BUF_ERROR;
+  while (1) switch (z->state->mode)
+  {
+    case METHOD:
+      NEEDBYTE
+      if (((z->state->sub.method = NEXTBYTE) & 0xf) != DEFLATED)
+      {
+        z->state->mode = BAD;
+        z->msg = "unknown compression method";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      if ((z->state->sub.method >> 4) + 8 > z->state->wbits)
+      {
+        z->state->mode = BAD;
+        z->msg = "invalid window size";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      z->state->mode = FLAG;
+    case FLAG:
+      NEEDBYTE
+      if ((b = NEXTBYTE) & 0x20)
+      {
+        z->state->mode = BAD;
+        z->msg = "invalid reserved bit";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      if (((z->state->sub.method << 8) + b) % 31)
+      {
+        z->state->mode = BAD;
+        z->msg = "incorrect header check";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      Trace((stderr, "inflate: zlib header ok\n"));
+      z->state->mode = BLOCKS;
+    case BLOCKS:
+      r = inflate_blocks(z->state->blocks, z, r);
+      if (f == Z_PACKET_FLUSH && z->avail_in == 0 && z->avail_out != 0)
+	  r = inflate_packet_flush(z->state->blocks);
+      if (r == Z_DATA_ERROR)
+      {
+        z->state->mode = BAD;
+        z->state->sub.marker = 0;       /* can try inflateSync */
+        break;
+      }
+      if (r != Z_STREAM_END)
+        return r;
+      r = Z_OK;
+      inflate_blocks_reset(z->state->blocks, z, &z->state->sub.check.was);
+      if (z->state->nowrap)
+      {
+        z->state->mode = DONE;
+        break;
+      }
+      z->state->mode = CHECK4;
+    case CHECK4:
+      NEEDBYTE
+      z->state->sub.check.need = (uLong)NEXTBYTE << 24;
+      z->state->mode = CHECK3;
+    case CHECK3:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE << 16;
+      z->state->mode = CHECK2;
+    case CHECK2:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE << 8;
+      z->state->mode = CHECK1;
+    case CHECK1:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE;
+
+      if (z->state->sub.check.was != z->state->sub.check.need)
+      {
+        z->state->mode = BAD;
+        z->msg = "incorrect data check";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      Trace((stderr, "inflate: zlib check ok\n"));
+      z->state->mode = DONE;
+    case DONE:
+      return Z_STREAM_END;
+    case BAD:
+      return Z_DATA_ERROR;
+    default:
+      return Z_STREAM_ERROR;
+  }
+
+ empty:
+  if (f != Z_PACKET_FLUSH)
+    return r;
+  z->state->mode = BAD;
+  z->state->sub.marker = 0;       /* can try inflateSync */
+  return Z_DATA_ERROR;
+}
+
+/*
+ * This subroutine adds the data at next_in/avail_in to the output history
+ * without performing any output.  The output buffer must be "caught up";
+ * i.e. no pending output (hence s->read equals s->write), and the state must
+ * be BLOCKS (i.e. we should be willing to see the start of a series of
+ * BLOCKS).  On exit, the output will also be caught up, and the checksum
+ * will have been updated if need be.
+ */
+
+int inflateIncomp(z)
+z_stream *z;
+{
+    if (z->state->mode != BLOCKS)
+	return Z_DATA_ERROR;
+    return inflate_addhistory(z->state->blocks, z);
+}
+
+
+int inflateSync(z)
+z_stream *z;
+{
+  uInt n;       /* number of bytes to look at */
+  Bytef *p;     /* pointer to bytes */
+  uInt m;       /* number of marker bytes found in a row */
+  uLong r, w;   /* temporaries to save total_in and total_out */
+
+  /* set up */
+  if (z == Z_NULL || z->state == Z_NULL)
+    return Z_STREAM_ERROR;
+  if (z->state->mode != BAD)
+  {
+    z->state->mode = BAD;
+    z->state->sub.marker = 0;
+  }
+  if ((n = z->avail_in) == 0)
+    return Z_BUF_ERROR;
+  p = z->next_in;
+  m = z->state->sub.marker;
+
+  /* search */
+  while (n && m < 4)
+  {
+    if (*p == (Byte)(m < 2 ? 0 : 0xff))
+      m++;
+    else if (*p)
+      m = 0;
+    else
+      m = 4 - m;
+    p++, n--;
+  }
+
+  /* restore */
+  z->total_in += p - z->next_in;
+  z->next_in = p;
+  z->avail_in = n;
+  z->state->sub.marker = m;
+
+  /* return no joy or set up to restart on a new block */
+  if (m != 4)
+    return Z_DATA_ERROR;
+  r = z->total_in;  w = z->total_out;
+  inflateReset(z);
+  z->total_in = r;  z->total_out = w;
+  z->state->mode = BLOCKS;
+  return Z_OK;
+}
+
+#undef NEEDBYTE
+#undef NEXTBYTE
+
+/*+++++*/
+/* infutil.h -- types and macros common to blocks and codes
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* inflate blocks semi-private state */
+struct inflate_blocks_state {
+
+  /* mode */
+  enum {
+      TYPE,     /* get type bits (3, including end bit) */
+      LENS,     /* get lengths for stored */
+      STORED,   /* processing stored block */
+      TABLE,    /* get table lengths */
+      BTREE,    /* get bit lengths tree for a dynamic block */
+      DTREE,    /* get length, distance trees for a dynamic block */
+      CODES,    /* processing fixed or dynamic block */
+      DRY,      /* output remaining window bytes */
+      DONEB,     /* finished last block, done */
+      BADB}      /* got a data error--stuck here */
+    mode;               /* current inflate_block mode */
+
+  /* mode dependent information */
+  union {
+    uInt left;          /* if STORED, bytes left to copy */
+    struct {
+      uInt table;               /* table lengths (14 bits) */
+      uInt index;               /* index into blens (or border) */
+      uIntf *blens;             /* bit lengths of codes */
+      uInt bb;                  /* bit length tree depth */
+      inflate_huft *tb;         /* bit length decoding tree */
+      int nblens;		/* # elements allocated at blens */
+    } trees;            /* if DTREE, decoding info for trees */
+    struct {
+      inflate_huft *tl, *td;    /* trees to free */
+      inflate_codes_statef
+         *codes;
+    } decode;           /* if CODES, current state */
+  } sub;                /* submode */
+  uInt last;            /* true if this block is the last block */
+
+  /* mode independent information */
+  uInt bitk;            /* bits in bit buffer */
+  uLong bitb;           /* bit buffer */
+  Bytef *window;        /* sliding window */
+  Bytef *end;           /* one byte after sliding window */
+  Bytef *read;          /* window read pointer */
+  Bytef *write;         /* window write pointer */
+  check_func checkfn;   /* check function */
+  uLong check;          /* check on output */
+
+};
+
+
+/* defines for inflate input/output */
+/*   update pointers and return */
+#define UPDBITS {s->bitb=b;s->bitk=k;}
+#define UPDIN {z->avail_in=n;z->total_in+=p-z->next_in;z->next_in=p;}
+#define UPDOUT {s->write=q;}
+#define UPDATE {UPDBITS UPDIN UPDOUT}
+#define LEAVE {UPDATE return inflate_flush(s,z,r);}
+/*   get bytes and bits */
+#define LOADIN {p=z->next_in;n=z->avail_in;b=s->bitb;k=s->bitk;}
+#define NEEDBYTE {if(n)r=Z_OK;else LEAVE}
+#define NEXTBYTE (n--,*p++)
+#define NEEDBITS(j) {while(k<(j)){NEEDBYTE;b|=((uLong)NEXTBYTE)<<k;k+=8;}}
+#define DUMPBITS(j) {b>>=(j);k-=(j);}
+/*   output bytes */
+#define WAVAIL (q<s->read?s->read-q-1:s->end-q)
+#define LOADOUT {q=s->write;m=WAVAIL;}
+#define WRAP {if(q==s->end&&s->read!=s->window){q=s->window;m=WAVAIL;}}
+#define FLUSH {UPDOUT r=inflate_flush(s,z,r); LOADOUT}
+#define NEEDOUT {if(m==0){WRAP if(m==0){FLUSH WRAP if(m==0) LEAVE}}r=Z_OK;}
+#define OUTBYTE(a) {*q++=(Byte)(a);m--;}
+/*   load local pointers */
+#define LOAD {LOADIN LOADOUT}
+
+/*
+ * The IBM 150 firmware munges the data right after _etext[].  This
+ * protects it. -- Cort
+ */
+local uInt protect_mask[] = {0, 0, 0, 0, 0, 0, 0, 0, 0 ,0 ,0 ,0};
+/* And'ing with mask[n] masks the lower n bits */
+local uInt inflate_mask[] = {
+    0x0000,
+    0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
+    0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
+};
+
+/* copy as much as possible from the sliding window to the output area */
+local int inflate_flush OF((
+    inflate_blocks_statef *,
+    z_stream *,
+    int));
+
+/*+++++*/
+/* inffast.h -- header to use inffast.c
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+local int inflate_fast OF((
+    uInt,
+    uInt,
+    inflate_huft *,
+    inflate_huft *,
+    inflate_blocks_statef *,
+    z_stream *));
+
+
+/*+++++*/
+/* infblock.c -- interpret and process block types to last block
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* Table for deflate from PKZIP's appnote.txt. */
+local uInt border[] = { /* Order of the bit length code lengths */
+        16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+/*
+   Notes beyond the 1.93a appnote.txt:
+
+   1. Distance pointers never point before the beginning of the output
+      stream.
+   2. Distance pointers can point back across blocks, up to 32k away.
+   3. There is an implied maximum of 7 bits for the bit length table and
+      15 bits for the actual data.
+   4. If only one code exists, then it is encoded using one bit.  (Zero
+      would be more efficient, but perhaps a little confusing.)  If two
+      codes exist, they are coded using one bit each (0 and 1).
+   5. There is no way of sending zero distance codes--a dummy must be
+      sent if there are none.  (History: a pre 2.0 version of PKZIP would
+      store blocks with no distance codes, but this was discovered to be
+      too harsh a criterion.)  Valid only for 1.93a.  2.04c does allow
+      zero distance codes, which is sent as one code of zero bits in
+      length.
+   6. There are up to 286 literal/length codes.  Code 256 represents the
+      end-of-block.  Note however that the static length tree defines
+      288 codes just to fill out the Huffman codes.  Codes 286 and 287
+      cannot be used though, since there is no length base or extra bits
+      defined for them.  Similarily, there are up to 30 distance codes.
+      However, static trees define 32 codes (all 5 bits) to fill out the
+      Huffman codes, but the last two had better not show up in the data.
+   7. Unzip can check dynamic Huffman blocks for complete code sets.
+      The exception is that a single code would not be complete (see #4).
+   8. The five bits following the block type is really the number of
+      literal codes sent minus 257.
+   9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits
+      (1+6+6).  Therefore, to output three times the length, you output
+      three codes (1+1+1), whereas to output four times the same length,
+      you only need two codes (1+3).  Hmm.
+  10. In the tree reconstruction algorithm, Code = Code + Increment
+      only if BitLength(i) is not zero.  (Pretty obvious.)
+  11. Correction: 4 Bits: # of Bit Length codes - 4     (4 - 19)
+  12. Note: length code 284 can represent 227-258, but length code 285
+      really is 258.  The last length deserves its own, short code
+      since it gets used a lot in very redundant files.  The length
+      258 is special since 258 - 3 (the min match length) is 255.
+  13. The literal/length and distance code bit lengths are read as a
+      single stream of lengths.  It is possible (and advantageous) for
+      a repeat code (16, 17, or 18) to go across the boundary between
+      the two sets of lengths.
+ */
+
+
+local void inflate_blocks_reset(s, z, c)
+inflate_blocks_statef *s;
+z_stream *z;
+uLongf *c;
+{
+  if (s->checkfn != Z_NULL)
+    *c = s->check;
+  if (s->mode == BTREE || s->mode == DTREE)
+    ZFREE(z, s->sub.trees.blens, s->sub.trees.nblens * sizeof(uInt));
+  if (s->mode == CODES)
+  {
+    inflate_codes_free(s->sub.decode.codes, z);
+    inflate_trees_free(s->sub.decode.td, z);
+    inflate_trees_free(s->sub.decode.tl, z);
+  }
+  s->mode = TYPE;
+  s->bitk = 0;
+  s->bitb = 0;
+  s->read = s->write = s->window;
+  if (s->checkfn != Z_NULL)
+    s->check = (*s->checkfn)(0L, Z_NULL, 0);
+  Trace((stderr, "inflate:   blocks reset\n"));
+}
+
+
+local inflate_blocks_statef *inflate_blocks_new(z, c, w)
+z_stream *z;
+check_func c;
+uInt w;
+{
+  inflate_blocks_statef *s;
+
+  if ((s = (inflate_blocks_statef *)ZALLOC
+       (z,1,sizeof(struct inflate_blocks_state))) == Z_NULL)
+    return s;
+  if ((s->window = (Bytef *)ZALLOC(z, 1, w)) == Z_NULL)
+  {
+    ZFREE(z, s, sizeof(struct inflate_blocks_state));
+    return Z_NULL;
+  }
+  s->end = s->window + w;
+  s->checkfn = c;
+  s->mode = TYPE;
+  Trace((stderr, "inflate:   blocks allocated\n"));
+  inflate_blocks_reset(s, z, &s->check);
+  return s;
+}
+
+
+local int inflate_blocks(s, z, r)
+inflate_blocks_statef *s;
+z_stream *z;
+int r;
+{
+  uInt t;               /* temporary storage */
+  uLong b;              /* bit buffer */
+  uInt k;               /* bits in bit buffer */
+  Bytef *p;             /* input data pointer */
+  uInt n;               /* bytes available there */
+  Bytef *q;             /* output window write pointer */
+  uInt m;               /* bytes to end of window or read pointer */
+
+  /* copy input/output information to locals (UPDATE macro restores) */
+  LOAD
+
+  /* process input based on current state */
+  while (1) switch (s->mode)
+  {
+    case TYPE:
+      NEEDBITS(3)
+      t = (uInt)b & 7;
+      s->last = t & 1;
+      switch (t >> 1)
+      {
+        case 0:                         /* stored */
+          Trace((stderr, "inflate:     stored block%s\n",
+                 s->last ? " (last)" : ""));
+          DUMPBITS(3)
+          t = k & 7;                    /* go to byte boundary */
+          DUMPBITS(t)
+          s->mode = LENS;               /* get length of stored block */
+          break;
+        case 1:                         /* fixed */
+          Trace((stderr, "inflate:     fixed codes block%s\n",
+                 s->last ? " (last)" : ""));
+          {
+            uInt bl, bd;
+            inflate_huft *tl, *td;
+
+            inflate_trees_fixed(&bl, &bd, &tl, &td);
+            s->sub.decode.codes = inflate_codes_new(bl, bd, tl, td, z);
+            if (s->sub.decode.codes == Z_NULL)
+            {
+              r = Z_MEM_ERROR;
+              LEAVE
+            }
+            s->sub.decode.tl = Z_NULL;  /* don't try to free these */
+            s->sub.decode.td = Z_NULL;
+          }
+          DUMPBITS(3)
+          s->mode = CODES;
+          break;
+        case 2:                         /* dynamic */
+          Trace((stderr, "inflate:     dynamic codes block%s\n",
+                 s->last ? " (last)" : ""));
+          DUMPBITS(3)
+          s->mode = TABLE;
+          break;
+        case 3:                         /* illegal */
+          DUMPBITS(3)
+          s->mode = BADB;
+          z->msg = "invalid block type";
+          r = Z_DATA_ERROR;
+          LEAVE
+      }
+      break;
+    case LENS:
+      NEEDBITS(32)
+      if (((~b) >> 16) != (b & 0xffff))
+      {
+        s->mode = BADB;
+        z->msg = "invalid stored block lengths";
+        r = Z_DATA_ERROR;
+        LEAVE
+      }
+      s->sub.left = (uInt)b & 0xffff;
+      b = k = 0;                      /* dump bits */
+      Tracev((stderr, "inflate:       stored length %u\n", s->sub.left));
+      s->mode = s->sub.left ? STORED : TYPE;
+      break;
+    case STORED:
+      if (n == 0)
+        LEAVE
+      NEEDOUT
+      t = s->sub.left;
+      if (t > n) t = n;
+      if (t > m) t = m;
+      zmemcpy(q, p, t);
+      p += t;  n -= t;
+      q += t;  m -= t;
+      if ((s->sub.left -= t) != 0)
+        break;
+      Tracev((stderr, "inflate:       stored end, %lu total out\n",
+              z->total_out + (q >= s->read ? q - s->read :
+              (s->end - s->read) + (q - s->window))));
+      s->mode = s->last ? DRY : TYPE;
+      break;
+    case TABLE:
+      NEEDBITS(14)
+      s->sub.trees.table = t = (uInt)b & 0x3fff;
+#ifndef PKZIP_BUG_WORKAROUND
+      if ((t & 0x1f) > 29 || ((t >> 5) & 0x1f) > 29)
+      {
+        s->mode = BADB;
+        z->msg = "too many length or distance symbols";
+        r = Z_DATA_ERROR;
+        LEAVE
+      }
+#endif
+      t = 258 + (t & 0x1f) + ((t >> 5) & 0x1f);
+      if (t < 19)
+        t = 19;
+      if ((s->sub.trees.blens = (uIntf*)ZALLOC(z, t, sizeof(uInt))) == Z_NULL)
+      {
+        r = Z_MEM_ERROR;
+        LEAVE
+      }
+      s->sub.trees.nblens = t;
+      DUMPBITS(14)
+      s->sub.trees.index = 0;
+      Tracev((stderr, "inflate:       table sizes ok\n"));
+      s->mode = BTREE;
+    case BTREE:
+      while (s->sub.trees.index < 4 + (s->sub.trees.table >> 10))
+      {
+        NEEDBITS(3)
+        s->sub.trees.blens[border[s->sub.trees.index++]] = (uInt)b & 7;
+        DUMPBITS(3)
+      }
+      while (s->sub.trees.index < 19)
+        s->sub.trees.blens[border[s->sub.trees.index++]] = 0;
+      s->sub.trees.bb = 7;
+      t = inflate_trees_bits(s->sub.trees.blens, &s->sub.trees.bb,
+                             &s->sub.trees.tb, z);
+      if (t != Z_OK)
+      {
+        r = t;
+        if (r == Z_DATA_ERROR)
+          s->mode = BADB;
+        LEAVE
+      }
+      s->sub.trees.index = 0;
+      Tracev((stderr, "inflate:       bits tree ok\n"));
+      s->mode = DTREE;
+    case DTREE:
+      while (t = s->sub.trees.table,
+             s->sub.trees.index < 258 + (t & 0x1f) + ((t >> 5) & 0x1f))
+      {
+        inflate_huft *h;
+        uInt i, j, c;
+
+        t = s->sub.trees.bb;
+        NEEDBITS(t)
+        h = s->sub.trees.tb + ((uInt)b & inflate_mask[t]);
+        t = h->word.what.Bits;
+        c = h->more.Base;
+        if (c < 16)
+        {
+          DUMPBITS(t)
+          s->sub.trees.blens[s->sub.trees.index++] = c;
+        }
+        else /* c == 16..18 */
+        {
+          i = c == 18 ? 7 : c - 14;
+          j = c == 18 ? 11 : 3;
+          NEEDBITS(t + i)
+          DUMPBITS(t)
+          j += (uInt)b & inflate_mask[i];
+          DUMPBITS(i)
+          i = s->sub.trees.index;
+          t = s->sub.trees.table;
+          if (i + j > 258 + (t & 0x1f) + ((t >> 5) & 0x1f) ||
+              (c == 16 && i < 1))
+          {
+            s->mode = BADB;
+            z->msg = "invalid bit length repeat";
+            r = Z_DATA_ERROR;
+            LEAVE
+          }
+          c = c == 16 ? s->sub.trees.blens[i - 1] : 0;
+          do {
+            s->sub.trees.blens[i++] = c;
+          } while (--j);
+          s->sub.trees.index = i;
+        }
+      }
+      inflate_trees_free(s->sub.trees.tb, z);
+      s->sub.trees.tb = Z_NULL;
+      {
+        uInt bl, bd;
+        inflate_huft *tl, *td;
+        inflate_codes_statef *c;
+
+        bl = 9;         /* must be <= 9 for lookahead assumptions */
+        bd = 6;         /* must be <= 9 for lookahead assumptions */
+        t = s->sub.trees.table;
+        t = inflate_trees_dynamic(257 + (t & 0x1f), 1 + ((t >> 5) & 0x1f),
+                                  s->sub.trees.blens, &bl, &bd, &tl, &td, z);
+        if (t != Z_OK)
+        {
+          if (t == (uInt)Z_DATA_ERROR)
+            s->mode = BADB;
+          r = t;
+          LEAVE
+        }
+        Tracev((stderr, "inflate:       trees ok\n"));
+        if ((c = inflate_codes_new(bl, bd, tl, td, z)) == Z_NULL)
+        {
+          inflate_trees_free(td, z);
+          inflate_trees_free(tl, z);
+          r = Z_MEM_ERROR;
+          LEAVE
+        }
+        ZFREE(z, s->sub.trees.blens, s->sub.trees.nblens * sizeof(uInt));
+        s->sub.decode.codes = c;
+        s->sub.decode.tl = tl;
+        s->sub.decode.td = td;
+      }
+      s->mode = CODES;
+    case CODES:
+      UPDATE
+      if ((r = inflate_codes(s, z, r)) != Z_STREAM_END)
+        return inflate_flush(s, z, r);
+      r = Z_OK;
+      inflate_codes_free(s->sub.decode.codes, z);
+      inflate_trees_free(s->sub.decode.td, z);
+      inflate_trees_free(s->sub.decode.tl, z);
+      LOAD
+      Tracev((stderr, "inflate:       codes end, %lu total out\n",
+              z->total_out + (q >= s->read ? q - s->read :
+              (s->end - s->read) + (q - s->window))));
+      if (!s->last)
+      {
+        s->mode = TYPE;
+        break;
+      }
+      if (k > 7)              /* return unused byte, if any */
+      {
+        Assert(k < 16, "inflate_codes grabbed too many bytes")
+        k -= 8;
+        n++;
+        p--;                    /* can always return one */
+      }
+      s->mode = DRY;
+    case DRY:
+      FLUSH
+      if (s->read != s->write)
+        LEAVE
+      s->mode = DONEB;
+    case DONEB:
+      r = Z_STREAM_END;
+      LEAVE
+    case BADB:
+      r = Z_DATA_ERROR;
+      LEAVE
+    default:
+      r = Z_STREAM_ERROR;
+      LEAVE
+  }
+}
+
+
+local int inflate_blocks_free(s, z, c)
+inflate_blocks_statef *s;
+z_stream *z;
+uLongf *c;
+{
+  inflate_blocks_reset(s, z, c);
+  ZFREE(z, s->window, s->end - s->window);
+  ZFREE(z, s, sizeof(struct inflate_blocks_state));
+  Trace((stderr, "inflate:   blocks freed\n"));
+  return Z_OK;
+}
+
+/*
+ * This subroutine adds the data at next_in/avail_in to the output history
+ * without performing any output.  The output buffer must be "caught up";
+ * i.e. no pending output (hence s->read equals s->write), and the state must
+ * be BLOCKS (i.e. we should be willing to see the start of a series of
+ * BLOCKS).  On exit, the output will also be caught up, and the checksum
+ * will have been updated if need be.
+ */
+local int inflate_addhistory(s, z)
+inflate_blocks_statef *s;
+z_stream *z;
+{
+    uLong b;              /* bit buffer */  /* NOT USED HERE */
+    uInt k;               /* bits in bit buffer */ /* NOT USED HERE */
+    uInt t;               /* temporary storage */
+    Bytef *p;             /* input data pointer */
+    uInt n;               /* bytes available there */
+    Bytef *q;             /* output window write pointer */
+    uInt m;               /* bytes to end of window or read pointer */
+
+    if (s->read != s->write)
+	return Z_STREAM_ERROR;
+    if (s->mode != TYPE)
+	return Z_DATA_ERROR;
+
+    /* we're ready to rock */
+    LOAD
+    /* while there is input ready, copy to output buffer, moving
+     * pointers as needed.
+     */
+    while (n) {
+	t = n;  /* how many to do */
+	/* is there room until end of buffer? */
+	if (t > m) t = m;
+	/* update check information */
+	if (s->checkfn != Z_NULL)
+	    s->check = (*s->checkfn)(s->check, q, t);
+	zmemcpy(q, p, t);
+	q += t;
+	p += t;
+	n -= t;
+	z->total_out += t;
+	s->read = q;    /* drag read pointer forward */
+/*      WRAP  */ 	/* expand WRAP macro by hand to handle s->read */
+	if (q == s->end) {
+	    s->read = q = s->window;
+	    m = WAVAIL;
+	}
+    }
+    UPDATE
+    return Z_OK;
+}
+
+
+/*
+ * At the end of a Deflate-compressed PPP packet, we expect to have seen
+ * a `stored' block type value but not the (zero) length bytes.
+ */
+local int inflate_packet_flush(s)
+    inflate_blocks_statef *s;
+{
+    if (s->mode != LENS)
+	return Z_DATA_ERROR;
+    s->mode = TYPE;
+    return Z_OK;
+}
+
+
+/*+++++*/
+/* inftrees.c -- generate Huffman trees for efficient decoding
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* simplify the use of the inflate_huft type with some defines */
+#define base more.Base
+#define next more.Next
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+
+local int huft_build OF((
+    uIntf *,            /* code lengths in bits */
+    uInt,               /* number of codes */
+    uInt,               /* number of "simple" codes */
+    uIntf *,            /* list of base values for non-simple codes */
+    uIntf *,            /* list of extra bits for non-simple codes */
+    inflate_huft * FAR*,/* result: starting table */
+    uIntf *,            /* maximum lookup bits (returns actual) */
+    z_stream *));       /* for zalloc function */
+
+local voidpf falloc OF((
+    voidpf,             /* opaque pointer (not used) */
+    uInt,               /* number of items */
+    uInt));             /* size of item */
+
+local void ffree OF((
+    voidpf q,           /* opaque pointer (not used) */
+    voidpf p,           /* what to free (not used) */
+    uInt n));		/* number of bytes (not used) */
+
+/* Tables for deflate from PKZIP's appnote.txt. */
+local uInt cplens[] = { /* Copy lengths for literal codes 257..285 */
+        3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
+        35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
+        /* actually lengths - 2; also see note #13 above about 258 */
+local uInt cplext[] = { /* Extra bits for literal codes 257..285 */
+        0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
+        3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 192, 192}; /* 192==invalid */
+local uInt cpdist[] = { /* Copy offsets for distance codes 0..29 */
+        1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
+        257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
+        8193, 12289, 16385, 24577};
+local uInt cpdext[] = { /* Extra bits for distance codes */
+        0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
+        7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
+        12, 12, 13, 13};
+
+/*
+   Huffman code decoding is performed using a multi-level table lookup.
+   The fastest way to decode is to simply build a lookup table whose
+   size is determined by the longest code.  However, the time it takes
+   to build this table can also be a factor if the data being decoded
+   is not very long.  The most common codes are necessarily the
+   shortest codes, so those codes dominate the decoding time, and hence
+   the speed.  The idea is you can have a shorter table that decodes the
+   shorter, more probable codes, and then point to subsidiary tables for
+   the longer codes.  The time it costs to decode the longer codes is
+   then traded against the time it takes to make longer tables.
+
+   This results of this trade are in the variables lbits and dbits
+   below.  lbits is the number of bits the first level table for literal/
+   length codes can decode in one step, and dbits is the same thing for
+   the distance codes.  Subsequent tables are also less than or equal to
+   those sizes.  These values may be adjusted either when all of the
+   codes are shorter than that, in which case the longest code length in
+   bits is used, or when the shortest code is *longer* than the requested
+   table size, in which case the length of the shortest code in bits is
+   used.
+
+   There are two different values for the two tables, since they code a
+   different number of possibilities each.  The literal/length table
+   codes 286 possible values, or in a flat code, a little over eight
+   bits.  The distance table codes 30 possible values, or a little less
+   than five bits, flat.  The optimum values for speed end up being
+   about one bit more than those, so lbits is 8+1 and dbits is 5+1.
+   The optimum values may differ though from machine to machine, and
+   possibly even between compilers.  Your mileage may vary.
+ */
+
+
+/* If BMAX needs to be larger than 16, then h and x[] should be uLong. */
+#define BMAX 15         /* maximum bit length of any code */
+#define N_MAX 288       /* maximum number of codes in any set */
+
+#ifdef DEBUG_ZLIB
+  uInt inflate_hufts;
+#endif
+
+local int huft_build(b, n, s, d, e, t, m, zs)
+uIntf *b;               /* code lengths in bits (all assumed <= BMAX) */
+uInt n;                 /* number of codes (assumed <= N_MAX) */
+uInt s;                 /* number of simple-valued codes (0..s-1) */
+uIntf *d;               /* list of base values for non-simple codes */
+uIntf *e;               /* list of extra bits for non-simple codes */
+inflate_huft * FAR *t;  /* result: starting table */
+uIntf *m;               /* maximum lookup bits, returns actual */
+z_stream *zs;           /* for zalloc function */
+/* Given a list of code lengths and a maximum table size, make a set of
+   tables to decode that set of codes.  Return Z_OK on success, Z_BUF_ERROR
+   if the given code set is incomplete (the tables are still built in this
+   case), Z_DATA_ERROR if the input is invalid (all zero length codes or an
+   over-subscribed set of lengths), or Z_MEM_ERROR if not enough memory. */
+{
+
+  uInt a;                       /* counter for codes of length k */
+  uInt c[BMAX+1];               /* bit length count table */
+  uInt f;                       /* i repeats in table every f entries */
+  int g;                        /* maximum code length */
+  int h;                        /* table level */
+  register uInt i;              /* counter, current code */
+  register uInt j;              /* counter */
+  register int k;               /* number of bits in current code */
+  int l;                        /* bits per table (returned in m) */
+  register uIntf *p;            /* pointer into c[], b[], or v[] */
+  inflate_huft *q;              /* points to current table */
+  struct inflate_huft_s r;      /* table entry for structure assignment */
+  inflate_huft *u[BMAX];        /* table stack */
+  uInt v[N_MAX];                /* values in order of bit length */
+  register int w;               /* bits before this table == (l * h) */
+  uInt x[BMAX+1];               /* bit offsets, then code stack */
+  uIntf *xp;                    /* pointer into x */
+  int y;                        /* number of dummy codes added */
+  uInt z;                       /* number of entries in current table */
+
+
+  /* Generate counts for each bit length */
+  p = c;
+#define C0 *p++ = 0;
+#define C2 C0 C0 C0 C0
+#define C4 C2 C2 C2 C2
+  C4                            /* clear c[]--assume BMAX+1 is 16 */
+  p = b;  i = n;
+  do {
+    c[*p++]++;                  /* assume all entries <= BMAX */
+  } while (--i);
+  if (c[0] == n)                /* null input--all zero length codes */
+  {
+    *t = (inflate_huft *)Z_NULL;
+    *m = 0;
+    return Z_OK;
+  }
+
+
+  /* Find minimum and maximum length, bound *m by those */
+  l = *m;
+  for (j = 1; j <= BMAX; j++)
+    if (c[j])
+      break;
+  k = j;                        /* minimum code length */
+  if ((uInt)l < j)
+    l = j;
+  for (i = BMAX; i; i--)
+    if (c[i])
+      break;
+  g = i;                        /* maximum code length */
+  if ((uInt)l > i)
+    l = i;
+  *m = l;
+
+
+  /* Adjust last length count to fill out codes, if needed */
+  for (y = 1 << j; j < i; j++, y <<= 1)
+    if ((y -= c[j]) < 0)
+      return Z_DATA_ERROR;
+  if ((y -= c[i]) < 0)
+    return Z_DATA_ERROR;
+  c[i] += y;
+
+
+  /* Generate starting offsets into the value table for each length */
+  x[1] = j = 0;
+  p = c + 1;  xp = x + 2;
+  while (--i) {                 /* note that i == g from above */
+    *xp++ = (j += *p++);
+  }
+
+
+  /* Make a table of values in order of bit lengths */
+  p = b;  i = 0;
+  do {
+    if ((j = *p++) != 0)
+      v[x[j]++] = i;
+  } while (++i < n);
+
+
+  /* Generate the Huffman codes and for each, make the table entries */
+  x[0] = i = 0;                 /* first Huffman code is zero */
+  p = v;                        /* grab values in bit order */
+  h = -1;                       /* no tables yet--level -1 */
+  w = -l;                       /* bits decoded == (l * h) */
+  u[0] = (inflate_huft *)Z_NULL;        /* just to keep compilers happy */
+  q = (inflate_huft *)Z_NULL;   /* ditto */
+  z = 0;                        /* ditto */
+
+  /* go through the bit lengths (k already is bits in shortest code) */
+  for (; k <= g; k++)
+  {
+    a = c[k];
+    while (a--)
+    {
+      /* here i is the Huffman code of length k bits for value *p */
+      /* make tables up to required level */
+      while (k > w + l)
+      {
+        h++;
+        w += l;                 /* previous table always l bits */
+
+        /* compute minimum size table less than or equal to l bits */
+        z = (z = g - w) > (uInt)l ? l : z;      /* table size upper limit */
+        if ((f = 1 << (j = k - w)) > a + 1)     /* try a k-w bit table */
+        {                       /* too few codes for k-w bit table */
+          f -= a + 1;           /* deduct codes from patterns left */
+          xp = c + k;
+          if (j < z)
+            while (++j < z)     /* try smaller tables up to z bits */
+            {
+              if ((f <<= 1) <= *++xp)
+                break;          /* enough codes to use up j bits */
+              f -= *xp;         /* else deduct codes from patterns */
+            }
+        }
+        z = 1 << j;             /* table entries for j-bit table */
+
+        /* allocate and link in new table */
+        if ((q = (inflate_huft *)ZALLOC
+             (zs,z + 1,sizeof(inflate_huft))) == Z_NULL)
+        {
+          if (h)
+            inflate_trees_free(u[0], zs);
+          return Z_MEM_ERROR;   /* not enough memory */
+        }
+	q->word.Nalloc = z + 1;
+#ifdef DEBUG_ZLIB
+        inflate_hufts += z + 1;
+#endif
+        *t = q + 1;             /* link to list for huft_free() */
+        *(t = &(q->next)) = Z_NULL;
+        u[h] = ++q;             /* table starts after link */
+
+        /* connect to last table, if there is one */
+        if (h)
+        {
+          x[h] = i;             /* save pattern for backing up */
+          r.bits = (Byte)l;     /* bits to dump before this table */
+          r.exop = (Byte)j;     /* bits in this table */
+          r.next = q;           /* pointer to this table */
+          j = i >> (w - l);     /* (get around Turbo C bug) */
+          u[h-1][j] = r;        /* connect to last table */
+        }
+      }
+
+      /* set up table entry in r */
+      r.bits = (Byte)(k - w);
+      if (p >= v + n)
+        r.exop = 128 + 64;      /* out of values--invalid code */
+      else if (*p < s)
+      {
+        r.exop = (Byte)(*p < 256 ? 0 : 32 + 64);     /* 256 is end-of-block */
+        r.base = *p++;          /* simple code is just the value */
+      }
+      else
+      {
+        r.exop = (Byte)e[*p - s] + 16 + 64; /* non-simple--look up in lists */
+        r.base = d[*p++ - s];
+      }
+
+      /* fill code-like entries with r */
+      f = 1 << (k - w);
+      for (j = i >> w; j < z; j += f)
+        q[j] = r;
+
+      /* backwards increment the k-bit code i */
+      for (j = 1 << (k - 1); i & j; j >>= 1)
+        i ^= j;
+      i ^= j;
+
+      /* backup over finished tables */
+      while ((i & ((1 << w) - 1)) != x[h])
+      {
+        h--;                    /* don't need to update q */
+        w -= l;
+      }
+    }
+  }
+
+
+  /* Return Z_BUF_ERROR if we were given an incomplete table */
+  return y != 0 && g != 1 ? Z_BUF_ERROR : Z_OK;
+}
+
+
+local int inflate_trees_bits(c, bb, tb, z)
+uIntf *c;               /* 19 code lengths */
+uIntf *bb;              /* bits tree desired/actual depth */
+inflate_huft * FAR *tb; /* bits tree result */
+z_stream *z;            /* for zfree function */
+{
+  int r;
+
+  r = huft_build(c, 19, 19, (uIntf*)Z_NULL, (uIntf*)Z_NULL, tb, bb, z);
+  if (r == Z_DATA_ERROR)
+    z->msg = "oversubscribed dynamic bit lengths tree";
+  else if (r == Z_BUF_ERROR)
+  {
+    inflate_trees_free(*tb, z);
+    z->msg = "incomplete dynamic bit lengths tree";
+    r = Z_DATA_ERROR;
+  }
+  return r;
+}
+
+
+local int inflate_trees_dynamic(nl, nd, c, bl, bd, tl, td, z)
+uInt nl;                /* number of literal/length codes */
+uInt nd;                /* number of distance codes */
+uIntf *c;               /* that many (total) code lengths */
+uIntf *bl;              /* literal desired/actual bit depth */
+uIntf *bd;              /* distance desired/actual bit depth */
+inflate_huft * FAR *tl; /* literal/length tree result */
+inflate_huft * FAR *td; /* distance tree result */
+z_stream *z;            /* for zfree function */
+{
+  int r;
+
+  /* build literal/length tree */
+  if ((r = huft_build(c, nl, 257, cplens, cplext, tl, bl, z)) != Z_OK)
+  {
+    if (r == Z_DATA_ERROR)
+      z->msg = "oversubscribed literal/length tree";
+    else if (r == Z_BUF_ERROR)
+    {
+      inflate_trees_free(*tl, z);
+      z->msg = "incomplete literal/length tree";
+      r = Z_DATA_ERROR;
+    }
+    return r;
+  }
+
+  /* build distance tree */
+  if ((r = huft_build(c + nl, nd, 0, cpdist, cpdext, td, bd, z)) != Z_OK)
+  {
+    if (r == Z_DATA_ERROR)
+      z->msg = "oversubscribed literal/length tree";
+    else if (r == Z_BUF_ERROR) {
+#ifdef PKZIP_BUG_WORKAROUND
+      r = Z_OK;
+    }
+#else
+      inflate_trees_free(*td, z);
+      z->msg = "incomplete literal/length tree";
+      r = Z_DATA_ERROR;
+    }
+    inflate_trees_free(*tl, z);
+    return r;
+#endif
+  }
+
+  /* done */
+  return Z_OK;
+}
+
+
+/* build fixed tables only once--keep them here */
+local int fixed_lock = 0;
+local int fixed_built = 0;
+#define FIXEDH 530      /* number of hufts used by fixed tables */
+local uInt fixed_left = FIXEDH;
+local inflate_huft fixed_mem[FIXEDH];
+local uInt fixed_bl;
+local uInt fixed_bd;
+local inflate_huft *fixed_tl;
+local inflate_huft *fixed_td;
+
+
+local voidpf falloc(q, n, s)
+voidpf q;        /* opaque pointer (not used) */
+uInt n;         /* number of items */
+uInt s;         /* size of item */
+{
+  Assert(s == sizeof(inflate_huft) && n <= fixed_left,
+         "inflate_trees falloc overflow");
+  if (q) s++; /* to make some compilers happy */
+  fixed_left -= n;
+  return (voidpf)(fixed_mem + fixed_left);
+}
+
+
+local void ffree(q, p, n)
+voidpf q;
+voidpf p;
+uInt n;
+{
+  Assert(0, "inflate_trees ffree called!");
+  if (q) q = p; /* to make some compilers happy */
+}
+
+
+local int inflate_trees_fixed(bl, bd, tl, td)
+uIntf *bl;               /* literal desired/actual bit depth */
+uIntf *bd;               /* distance desired/actual bit depth */
+inflate_huft * FAR *tl;  /* literal/length tree result */
+inflate_huft * FAR *td;  /* distance tree result */
+{
+  /* build fixed tables if not built already--lock out other instances */
+  while (++fixed_lock > 1)
+    fixed_lock--;
+  if (!fixed_built)
+  {
+    int k;              /* temporary variable */
+    unsigned c[288];    /* length list for huft_build */
+    z_stream z;         /* for falloc function */
+
+    /* set up fake z_stream for memory routines */
+    z.zalloc = falloc;
+    z.zfree = ffree;
+    z.opaque = Z_NULL;
+
+    /* literal table */
+    for (k = 0; k < 144; k++)
+      c[k] = 8;
+    for (; k < 256; k++)
+      c[k] = 9;
+    for (; k < 280; k++)
+      c[k] = 7;
+    for (; k < 288; k++)
+      c[k] = 8;
+    fixed_bl = 7;
+    huft_build(c, 288, 257, cplens, cplext, &fixed_tl, &fixed_bl, &z);
+
+    /* distance table */
+    for (k = 0; k < 30; k++)
+      c[k] = 5;
+    fixed_bd = 5;
+    huft_build(c, 30, 0, cpdist, cpdext, &fixed_td, &fixed_bd, &z);
+
+    /* done */
+    fixed_built = 1;
+  }
+  fixed_lock--;
+  *bl = fixed_bl;
+  *bd = fixed_bd;
+  *tl = fixed_tl;
+  *td = fixed_td;
+  return Z_OK;
+}
+
+
+local int inflate_trees_free(t, z)
+inflate_huft *t;        /* table to free */
+z_stream *z;            /* for zfree function */
+/* Free the malloc'ed tables built by huft_build(), which makes a linked
+   list of the tables it made, with the links in a dummy first entry of
+   each table. */
+{
+  register inflate_huft *p, *q;
+
+  /* Go through linked list, freeing from the malloced (t[-1]) address. */
+  p = t;
+  while (p != Z_NULL)
+  {
+    q = (--p)->next;
+    ZFREE(z, p, p->word.Nalloc * sizeof(inflate_huft));
+    p = q;
+  }
+  return Z_OK;
+}
+
+/*+++++*/
+/* infcodes.c -- process literals and length/distance pairs
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* simplify the use of the inflate_huft type with some defines */
+#define base more.Base
+#define next more.Next
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+/* inflate codes private state */
+struct inflate_codes_state {
+
+  /* mode */
+  enum {        /* waiting for "i:"=input, "o:"=output, "x:"=nothing */
+      START,    /* x: set up for LEN */
+      LEN,      /* i: get length/literal/eob next */
+      LENEXT,   /* i: getting length extra (have base) */
+      DIST,     /* i: get distance next */
+      DISTEXT,  /* i: getting distance extra */
+      COPY,     /* o: copying bytes in window, waiting for space */
+      LIT,      /* o: got literal, waiting for output space */
+      WASH,     /* o: got eob, possibly still output waiting */
+      END,      /* x: got eob and all data flushed */
+      BADCODE}  /* x: got error */
+    mode;               /* current inflate_codes mode */
+
+  /* mode dependent information */
+  uInt len;
+  union {
+    struct {
+      inflate_huft *tree;       /* pointer into tree */
+      uInt need;                /* bits needed */
+    } code;             /* if LEN or DIST, where in tree */
+    uInt lit;           /* if LIT, literal */
+    struct {
+      uInt get;                 /* bits to get for extra */
+      uInt dist;                /* distance back to copy from */
+    } copy;             /* if EXT or COPY, where and how much */
+  } sub;                /* submode */
+
+  /* mode independent information */
+  Byte lbits;           /* ltree bits decoded per branch */
+  Byte dbits;           /* dtree bits decoder per branch */
+  inflate_huft *ltree;          /* literal/length/eob tree */
+  inflate_huft *dtree;          /* distance tree */
+
+};
+
+
+local inflate_codes_statef *inflate_codes_new(bl, bd, tl, td, z)
+uInt bl, bd;
+inflate_huft *tl, *td;
+z_stream *z;
+{
+  inflate_codes_statef *c;
+
+  if ((c = (inflate_codes_statef *)
+       ZALLOC(z,1,sizeof(struct inflate_codes_state))) != Z_NULL)
+  {
+    c->mode = START;
+    c->lbits = (Byte)bl;
+    c->dbits = (Byte)bd;
+    c->ltree = tl;
+    c->dtree = td;
+    Tracev((stderr, "inflate:       codes new\n"));
+  }
+  return c;
+}
+
+
+local int inflate_codes(s, z, r)
+inflate_blocks_statef *s;
+z_stream *z;
+int r;
+{
+  uInt j;               /* temporary storage */
+  inflate_huft *t;      /* temporary pointer */
+  uInt e;               /* extra bits or operation */
+  uLong b;              /* bit buffer */
+  uInt k;               /* bits in bit buffer */
+  Bytef *p;             /* input data pointer */
+  uInt n;               /* bytes available there */
+  Bytef *q;             /* output window write pointer */
+  uInt m;               /* bytes to end of window or read pointer */
+  Bytef *f;             /* pointer to copy strings from */
+  inflate_codes_statef *c = s->sub.decode.codes;  /* codes state */
+
+  /* copy input/output information to locals (UPDATE macro restores) */
+  LOAD
+
+  /* process input and output based on current state */
+  while (1) switch (c->mode)
+  {             /* waiting for "i:"=input, "o:"=output, "x:"=nothing */
+    case START:         /* x: set up for LEN */
+#ifndef SLOW
+      if (m >= 258 && n >= 10)
+      {
+        UPDATE
+        r = inflate_fast(c->lbits, c->dbits, c->ltree, c->dtree, s, z);
+        LOAD
+        if (r != Z_OK)
+        {
+          c->mode = r == Z_STREAM_END ? WASH : BADCODE;
+          break;
+        }
+      }
+#endif /* !SLOW */
+      c->sub.code.need = c->lbits;
+      c->sub.code.tree = c->ltree;
+      c->mode = LEN;
+    case LEN:           /* i: get length/literal/eob next */
+      j = c->sub.code.need;
+      NEEDBITS(j)
+      t = c->sub.code.tree + ((uInt)b & inflate_mask[j]);
+      DUMPBITS(t->bits)
+      e = (uInt)(t->exop);
+      if (e == 0)               /* literal */
+      {
+        c->sub.lit = t->base;
+        Tracevv((stderr, t->base >= 0x20 && t->base < 0x7f ?
+                 "inflate:         literal '%c'\n" :
+                 "inflate:         literal 0x%02x\n", t->base));
+        c->mode = LIT;
+        break;
+      }
+      if (e & 16)               /* length */
+      {
+        c->sub.copy.get = e & 15;
+        c->len = t->base;
+        c->mode = LENEXT;
+        break;
+      }
+      if ((e & 64) == 0)        /* next table */
+      {
+        c->sub.code.need = e;
+        c->sub.code.tree = t->next;
+        break;
+      }
+      if (e & 32)               /* end of block */
+      {
+        Tracevv((stderr, "inflate:         end of block\n"));
+        c->mode = WASH;
+        break;
+      }
+      c->mode = BADCODE;        /* invalid code */
+      z->msg = "invalid literal/length code";
+      r = Z_DATA_ERROR;
+      LEAVE
+    case LENEXT:        /* i: getting length extra (have base) */
+      j = c->sub.copy.get;
+      NEEDBITS(j)
+      c->len += (uInt)b & inflate_mask[j];
+      DUMPBITS(j)
+      c->sub.code.need = c->dbits;
+      c->sub.code.tree = c->dtree;
+      Tracevv((stderr, "inflate:         length %u\n", c->len));
+      c->mode = DIST;
+    case DIST:          /* i: get distance next */
+      j = c->sub.code.need;
+      NEEDBITS(j)
+      t = c->sub.code.tree + ((uInt)b & inflate_mask[j]);
+      DUMPBITS(t->bits)
+      e = (uInt)(t->exop);
+      if (e & 16)               /* distance */
+      {
+        c->sub.copy.get = e & 15;
+        c->sub.copy.dist = t->base;
+        c->mode = DISTEXT;
+        break;
+      }
+      if ((e & 64) == 0)        /* next table */
+      {
+        c->sub.code.need = e;
+        c->sub.code.tree = t->next;
+        break;
+      }
+      c->mode = BADCODE;        /* invalid code */
+      z->msg = "invalid distance code";
+      r = Z_DATA_ERROR;
+      LEAVE
+    case DISTEXT:       /* i: getting distance extra */
+      j = c->sub.copy.get;
+      NEEDBITS(j)
+      c->sub.copy.dist += (uInt)b & inflate_mask[j];
+      DUMPBITS(j)
+      Tracevv((stderr, "inflate:         distance %u\n", c->sub.copy.dist));
+      c->mode = COPY;
+    case COPY:          /* o: copying bytes in window, waiting for space */
+#ifndef __TURBOC__ /* Turbo C bug for following expression */
+      f = (uInt)(q - s->window) < c->sub.copy.dist ?
+          s->end - (c->sub.copy.dist - (q - s->window)) :
+          q - c->sub.copy.dist;
+#else
+      f = q - c->sub.copy.dist;
+      if ((uInt)(q - s->window) < c->sub.copy.dist)
+        f = s->end - (c->sub.copy.dist - (q - s->window));
+#endif
+      while (c->len)
+      {
+        NEEDOUT
+        OUTBYTE(*f++)
+        if (f == s->end)
+          f = s->window;
+        c->len--;
+      }
+      c->mode = START;
+      break;
+    case LIT:           /* o: got literal, waiting for output space */
+      NEEDOUT
+      OUTBYTE(c->sub.lit)
+      c->mode = START;
+      break;
+    case WASH:          /* o: got eob, possibly more output */
+      FLUSH
+      if (s->read != s->write)
+        LEAVE
+      c->mode = END;
+    case END:
+      r = Z_STREAM_END;
+      LEAVE
+    case BADCODE:       /* x: got error */
+      r = Z_DATA_ERROR;
+      LEAVE
+    default:
+      r = Z_STREAM_ERROR;
+      LEAVE
+  }
+}
+
+
+local void inflate_codes_free(c, z)
+inflate_codes_statef *c;
+z_stream *z;
+{
+  ZFREE(z, c, sizeof(struct inflate_codes_state));
+  Tracev((stderr, "inflate:       codes free\n"));
+}
+
+/*+++++*/
+/* inflate_util.c -- data and routines common to blocks and codes
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* copy as much as possible from the sliding window to the output area */
+local int inflate_flush(s, z, r)
+inflate_blocks_statef *s;
+z_stream *z;
+int r;
+{
+  uInt n;
+  Bytef *p, *q;
+
+  /* local copies of source and destination pointers */
+  p = z->next_out;
+  q = s->read;
+
+  /* compute number of bytes to copy as far as end of window */
+  n = (uInt)((q <= s->write ? s->write : s->end) - q);
+  if (n > z->avail_out) n = z->avail_out;
+  if (n && r == Z_BUF_ERROR) r = Z_OK;
+
+  /* update counters */
+  z->avail_out -= n;
+  z->total_out += n;
+
+  /* update check information */
+  if (s->checkfn != Z_NULL)
+    s->check = (*s->checkfn)(s->check, q, n);
+
+  /* copy as far as end of window */
+  zmemcpy(p, q, n);
+  p += n;
+  q += n;
+
+  /* see if more to copy at beginning of window */
+  if (q == s->end)
+  {
+    /* wrap pointers */
+    q = s->window;
+    if (s->write == s->end)
+      s->write = s->window;
+
+    /* compute bytes to copy */
+    n = (uInt)(s->write - q);
+    if (n > z->avail_out) n = z->avail_out;
+    if (n && r == Z_BUF_ERROR) r = Z_OK;
+
+    /* update counters */
+    z->avail_out -= n;
+    z->total_out += n;
+
+    /* update check information */
+    if (s->checkfn != Z_NULL)
+      s->check = (*s->checkfn)(s->check, q, n);
+
+    /* copy */
+    zmemcpy(p, q, n);
+    p += n;
+    q += n;
+  }
+
+  /* update pointers */
+  z->next_out = p;
+  s->read = q;
+
+  /* done */
+  return r;
+}
+
+
+/*+++++*/
+/* inffast.c -- process literals and length/distance pairs fast
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* simplify the use of the inflate_huft type with some defines */
+#define base more.Base
+#define next more.Next
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+/* macros for bit input with no checking and for returning unused bytes */
+#define GRABBITS(j) {while(k<(j)){b|=((uLong)NEXTBYTE)<<k;k+=8;}}
+#define UNGRAB {n+=(c=k>>3);p-=c;k&=7;}
+
+/* Called with number of bytes left to write in window at least 258
+   (the maximum string length) and number of input bytes available
+   at least ten.  The ten bytes are six bytes for the longest length/
+   distance pair plus four bytes for overloading the bit buffer. */
+
+local int inflate_fast(bl, bd, tl, td, s, z)
+uInt bl, bd;
+inflate_huft *tl, *td;
+inflate_blocks_statef *s;
+z_stream *z;
+{
+  inflate_huft *t;      /* temporary pointer */
+  uInt e;               /* extra bits or operation */
+  uLong b;              /* bit buffer */
+  uInt k;               /* bits in bit buffer */
+  Bytef *p;             /* input data pointer */
+  uInt n;               /* bytes available there */
+  Bytef *q;             /* output window write pointer */
+  uInt m;               /* bytes to end of window or read pointer */
+  uInt ml;              /* mask for literal/length tree */
+  uInt md;              /* mask for distance tree */
+  uInt c;               /* bytes to copy */
+  uInt d;               /* distance back to copy from */
+  Bytef *r;             /* copy source pointer */
+
+  /* load input, output, bit values */
+  LOAD
+
+  /* initialize masks */
+  ml = inflate_mask[bl];
+  md = inflate_mask[bd];
+
+  /* do until not enough input or output space for fast loop */
+  do {                          /* assume called with m >= 258 && n >= 10 */
+    /* get literal/length code */
+    GRABBITS(20)                /* max bits for literal/length code */
+    if ((e = (t = tl + ((uInt)b & ml))->exop) == 0)
+    {
+      DUMPBITS(t->bits)
+      Tracevv((stderr, t->base >= 0x20 && t->base < 0x7f ?
+                "inflate:         * literal '%c'\n" :
+                "inflate:         * literal 0x%02x\n", t->base));
+      *q++ = (Byte)t->base;
+      m--;
+      continue;
+    }
+    do {
+      DUMPBITS(t->bits)
+      if (e & 16)
+      {
+        /* get extra bits for length */
+        e &= 15;
+        c = t->base + ((uInt)b & inflate_mask[e]);
+        DUMPBITS(e)
+        Tracevv((stderr, "inflate:         * length %u\n", c));
+
+        /* decode distance base of block to copy */
+        GRABBITS(15);           /* max bits for distance code */
+        e = (t = td + ((uInt)b & md))->exop;
+        do {
+          DUMPBITS(t->bits)
+          if (e & 16)
+          {
+            /* get extra bits to add to distance base */
+            e &= 15;
+            GRABBITS(e)         /* get extra bits (up to 13) */
+            d = t->base + ((uInt)b & inflate_mask[e]);
+            DUMPBITS(e)
+            Tracevv((stderr, "inflate:         * distance %u\n", d));
+
+            /* do the copy */
+            m -= c;
+            if ((uInt)(q - s->window) >= d)     /* offset before dest */
+            {                                   /*  just copy */
+              r = q - d;
+              *q++ = *r++;  c--;        /* minimum count is three, */
+              *q++ = *r++;  c--;        /*  so unroll loop a little */
+            }
+            else                        /* else offset after destination */
+            {
+              e = d - (q - s->window);  /* bytes from offset to end */
+              r = s->end - e;           /* pointer to offset */
+              if (c > e)                /* if source crosses, */
+              {
+                c -= e;                 /* copy to end of window */
+                do {
+                  *q++ = *r++;
+                } while (--e);
+                r = s->window;          /* copy rest from start of window */
+              }
+            }
+            do {                        /* copy all or what's left */
+              *q++ = *r++;
+            } while (--c);
+            break;
+          }
+          else if ((e & 64) == 0)
+            e = (t = t->next + ((uInt)b & inflate_mask[e]))->exop;
+          else
+          {
+            z->msg = "invalid distance code";
+            UNGRAB
+            UPDATE
+            return Z_DATA_ERROR;
+          }
+        } while (1);
+        break;
+      }
+      if ((e & 64) == 0)
+      {
+        if ((e = (t = t->next + ((uInt)b & inflate_mask[e]))->exop) == 0)
+        {
+          DUMPBITS(t->bits)
+          Tracevv((stderr, t->base >= 0x20 && t->base < 0x7f ?
+                    "inflate:         * literal '%c'\n" :
+                    "inflate:         * literal 0x%02x\n", t->base));
+          *q++ = (Byte)t->base;
+          m--;
+          break;
+        }
+      }
+      else if (e & 32)
+      {
+        Tracevv((stderr, "inflate:         * end of block\n"));
+        UNGRAB
+        UPDATE
+        return Z_STREAM_END;
+      }
+      else
+      {
+        z->msg = "invalid literal/length code";
+        UNGRAB
+        UPDATE
+        return Z_DATA_ERROR;
+      }
+    } while (1);
+  } while (m >= 258 && n >= 10);
+
+  /* not enough input or output--restore pointers and return */
+  UNGRAB
+  UPDATE
+  return Z_OK;
+}
+
+
+/*+++++*/
+/* zutil.c -- target dependent utility functions for the compression library
+ * Copyright (C) 1995 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* From: zutil.c,v 1.8 1995/05/03 17:27:12 jloup Exp */
+
+char *zlib_version = ZLIB_VERSION;
+
+char *z_errmsg[] = {
+"stream end",          /* Z_STREAM_END    1 */
+"",                    /* Z_OK            0 */
+"file error",          /* Z_ERRNO        (-1) */
+"stream error",        /* Z_STREAM_ERROR (-2) */
+"data error",          /* Z_DATA_ERROR   (-3) */
+"insufficient memory", /* Z_MEM_ERROR    (-4) */
+"buffer error",        /* Z_BUF_ERROR    (-5) */
+""};
+
+
+/*+++++*/
+/* adler32.c -- compute the Adler-32 checksum of a data stream
+ * Copyright (C) 1995 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* From: adler32.c,v 1.6 1995/05/03 17:27:08 jloup Exp */
+
+#define BASE 65521L /* largest prime smaller than 65536 */
+#define NMAX 5552
+/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
+
+#define DO1(buf)  {s1 += *buf++; s2 += s1;}
+#define DO2(buf)  DO1(buf); DO1(buf);
+#define DO4(buf)  DO2(buf); DO2(buf);
+#define DO8(buf)  DO4(buf); DO4(buf);
+#define DO16(buf) DO8(buf); DO8(buf);
+
+/* ========================================================================= */
+uLong adler32(adler, buf, len)
+    uLong adler;
+    Bytef *buf;
+    uInt len;
+{
+    unsigned long s1 = adler & 0xffff;
+    unsigned long s2 = (adler >> 16) & 0xffff;
+    int k;
+
+    if (buf == Z_NULL) return 1L;
+
+    while (len > 0) {
+        k = len < NMAX ? len : NMAX;
+        len -= k;
+        while (k >= 16) {
+            DO16(buf);
+            k -= 16;
+        }
+        if (k != 0) do {
+            DO1(buf);
+        } while (--k);
+        s1 %= BASE;
+        s2 %= BASE;
+    }
+    return (s2 << 16) | s1;
+}
diff --git a/arch/xtensa/boot/lib/zmem.c b/arch/xtensa/boot/lib/zmem.c
new file mode 100644
index 000000000000..7848f126d67d
--- /dev/null
+++ b/arch/xtensa/boot/lib/zmem.c
@@ -0,0 +1,87 @@
+#include "zlib.h"
+
+/* bits taken from ppc */
+
+extern void *avail_ram, *end_avail;
+
+void exit (void)
+{
+  for (;;);
+}
+
+void *zalloc(void *x, unsigned items, unsigned size)
+{
+        void *p = avail_ram;
+
+        size *= items;
+        size = (size + 7) & -8;
+        avail_ram += size;
+        if (avail_ram > end_avail) {
+                //puts("oops... out of memory\n");
+                //pause();
+                exit ();
+        }
+        return p;
+}
+
+void zfree(void *x, void *addr, unsigned nb)
+{
+}
+
+
+#define HEAD_CRC        2
+#define EXTRA_FIELD     4
+#define ORIG_NAME       8
+#define COMMENT         0x10
+#define RESERVED        0xe0
+
+#define DEFLATED        8
+
+void gunzip (void *dst, int dstlen, unsigned char *src, int *lenp)
+{
+	z_stream s;
+	int r, i, flags;
+
+        /* skip header */
+
+        i = 10;
+        flags = src[3];
+        if (src[2] != DEFLATED || (flags & RESERVED) != 0) {
+                //puts("bad gzipped data\n");
+                exit();
+        }
+        if ((flags & EXTRA_FIELD) != 0)
+                i = 12 + src[10] + (src[11] << 8);
+        if ((flags & ORIG_NAME) != 0)
+                while (src[i++] != 0)
+                        ;
+        if ((flags & COMMENT) != 0)
+                while (src[i++] != 0)
+                        ;
+        if ((flags & HEAD_CRC) != 0)
+                i += 2;
+        if (i >= *lenp) {
+                //puts("gunzip: ran out of data in header\n");
+                exit();
+        }
+
+        s.zalloc = zalloc;
+        s.zfree = zfree;
+        r = inflateInit2(&s, -MAX_WBITS);
+        if (r != Z_OK) {
+                //puts("inflateInit2 returned "); puthex(r); puts("\n");
+                exit();
+        }
+        s.next_in = src + i;
+        s.avail_in = *lenp - i;
+        s.next_out = dst;
+        s.avail_out = dstlen;
+        r = inflate(&s, Z_FINISH);
+        if (r != Z_OK && r != Z_STREAM_END) {
+                //puts("inflate returned "); puthex(r); puts("\n");
+                exit();
+        }
+        *lenp = s.next_out - (unsigned char *) dst;
+        inflateEnd(&s);
+}
+
diff --git a/arch/xtensa/boot/ramdisk/Makefile b/arch/xtensa/boot/ramdisk/Makefile
new file mode 100644
index 000000000000..b12f76352438
--- /dev/null
+++ b/arch/xtensa/boot/ramdisk/Makefile
@@ -0,0 +1,23 @@
+#
+# Makefile for a ramdisk image
+#
+
+BIG_ENDIAN	:= $(shell echo -e "\#ifdef __XTENSA_EL__\nint little;\n\#else\nint big;\n\#endif" | $(CC) -E -|grep -c big)
+
+ifeq ($(BIG_ENDIAN),1)
+OBJCOPY_ARGS    := -O elf32-xtensa-be
+else
+OBJCOPY_ARGS    := -O elf32-xtensa-le
+endif
+
+obj-y = ramdisk.o
+
+RAMDISK_IMAGE = arch/$(ARCH)/boot/ramdisk/$(CONFIG_EMBEDDED_RAMDISK_IMAGE)
+
+arch/$(ARCH)/boot/ramdisk/ramdisk.o:
+	$(Q)echo -e "dummy:" | $(AS) -o $@;
+	$(Q)$(OBJCOPY) $(OBJCOPY_ARGS)					    \
+		--add-section .initrd=$(RAMDISK_IMAGE)			    \
+		--set-section-flags .initrd=contents,alloc,load,load,data   \
+		arch/$(ARCH)/boot/ramdisk/ramdisk.o $@
+
diff --git a/arch/xtensa/configs/common_defconfig b/arch/xtensa/configs/common_defconfig
new file mode 100644
index 000000000000..1d230ee081b4
--- /dev/null
+++ b/arch/xtensa/configs/common_defconfig
@@ -0,0 +1,662 @@
+#
+# Automatically generated make config: don't edit
+# Linux kernel version: 2.6.11-rc2
+# Tue Mar  1 16:36:53 2005
+#
+# CONFIG_FRAME_POINTER is not set
+CONFIG_XTENSA=y
+# CONFIG_UID16 is not set
+CONFIG_RWSEM_XCHGADD_ALGORITHM=y
+CONFIG_HAVE_DEC_LOCK=y
+CONFIG_GENERIC_HARDIRQS=y
+
+#
+# Code maturity level options
+#
+CONFIG_EXPERIMENTAL=y
+CONFIG_CLEAN_COMPILE=y
+CONFIG_BROKEN_ON_SMP=y
+
+#
+# General setup
+#
+CONFIG_LOCALVERSION=""
+CONFIG_SWAP=y
+CONFIG_SYSVIPC=y
+# CONFIG_POSIX_MQUEUE is not set
+CONFIG_BSD_PROCESS_ACCT=y
+# CONFIG_BSD_PROCESS_ACCT_V3 is not set
+CONFIG_SYSCTL=y
+# CONFIG_AUDIT is not set
+CONFIG_LOG_BUF_SHIFT=14
+# CONFIG_HOTPLUG is not set
+CONFIG_KOBJECT_UEVENT=y
+# CONFIG_IKCONFIG is not set
+# CONFIG_EMBEDDED is not set
+CONFIG_KALLSYMS=y
+# CONFIG_KALLSYMS_ALL is not set
+# CONFIG_KALLSYMS_EXTRA_PASS is not set
+CONFIG_FUTEX=y
+CONFIG_EPOLL=y
+# CONFIG_CC_OPTIMIZE_FOR_SIZE is not set
+CONFIG_SHMEM=y
+CONFIG_CC_ALIGN_FUNCTIONS=0
+CONFIG_CC_ALIGN_LABELS=0
+CONFIG_CC_ALIGN_LOOPS=0
+CONFIG_CC_ALIGN_JUMPS=0
+# CONFIG_TINY_SHMEM is not set
+
+#
+# Loadable module support
+#
+CONFIG_MODULES=y
+# CONFIG_MODULE_UNLOAD is not set
+CONFIG_OBSOLETE_MODPARM=y
+CONFIG_MODVERSIONS=y
+# CONFIG_MODULE_SRCVERSION_ALL is not set
+CONFIG_KMOD=y
+
+#
+# Processor type and features
+#
+CONFIG_XTENSA_ARCH_LINUX_BE=y
+# CONFIG_XTENSA_ARCH_LINUX_LE is not set
+# CONFIG_XTENSA_ARCH_LINUX_TEST is not set
+# CONFIG_XTENSA_ARCH_S5 is not set
+# CONFIG_XTENSA_CUSTOM is not set
+CONFIG_MMU=y
+# CONFIG_XTENSA_UNALIGNED_USER is not set
+# CONFIG_PREEMPT is not set
+# CONFIG_MATH_EMULATION is not set
+# CONFIG_HIGHMEM is not set
+
+#
+# Platform options
+#
+# CONFIG_XTENSA_PLATFORM_ISS is not set
+CONFIG_XTENSA_PLATFORM_XT2000=y
+CONFIG_XTENSA_CALIBRATE_CCOUNT=y
+CONFIG_GENERIC_CALIBRATE_DELAY=y
+CONFIG_CMDLINE_BOOL=y
+CONFIG_CMDLINE="console=ttyS0,38400 ip=bootp root=nfs nfsroot=/opt/montavista/pro/devkit/xtensa/linux_be/target"
+
+#
+# Bus options
+#
+CONFIG_PCI=y
+# CONFIG_PCI_LEGACY_PROC is not set
+# CONFIG_PCI_NAMES is not set
+
+#
+# PCCARD (PCMCIA/CardBus) support
+#
+# CONFIG_PCCARD is not set
+
+#
+# PC-card bridges
+#
+
+#
+# PCI Hotplug Support
+#
+# CONFIG_HOTPLUG_PCI is not set
+
+#
+# Exectuable file formats
+#
+CONFIG_KCORE_ELF=y
+CONFIG_BINFMT_ELF=y
+CONFIG_BINFMT_MISC=y
+
+#
+# Device Drivers
+#
+
+#
+# Generic Driver Options
+#
+CONFIG_STANDALONE=y
+CONFIG_PREVENT_FIRMWARE_BUILD=y
+# CONFIG_FW_LOADER is not set
+# CONFIG_DEBUG_DRIVER is not set
+
+#
+# Memory Technology Devices (MTD)
+#
+# CONFIG_MTD is not set
+
+#
+# Parallel port support
+#
+# CONFIG_PARPORT is not set
+
+#
+# Plug and Play support
+#
+
+#
+# Block devices
+#
+# CONFIG_BLK_DEV_FD is not set
+# CONFIG_BLK_CPQ_DA is not set
+# CONFIG_BLK_CPQ_CISS_DA is not set
+# CONFIG_BLK_DEV_DAC960 is not set
+# CONFIG_BLK_DEV_UMEM is not set
+# CONFIG_BLK_DEV_COW_COMMON is not set
+# CONFIG_BLK_DEV_LOOP is not set
+# CONFIG_BLK_DEV_NBD is not set
+# CONFIG_BLK_DEV_SX8 is not set
+# CONFIG_BLK_DEV_RAM is not set
+CONFIG_BLK_DEV_RAM_COUNT=16
+CONFIG_INITRAMFS_SOURCE=""
+# CONFIG_CDROM_PKTCDVD is not set
+
+#
+# IO Schedulers
+#
+CONFIG_IOSCHED_NOOP=y
+CONFIG_IOSCHED_AS=y
+CONFIG_IOSCHED_DEADLINE=y
+CONFIG_IOSCHED_CFQ=y
+# CONFIG_ATA_OVER_ETH is not set
+
+#
+# ATA/ATAPI/MFM/RLL support
+#
+# CONFIG_IDE is not set
+
+#
+# SCSI device support
+#
+# CONFIG_SCSI is not set
+
+#
+# Multi-device support (RAID and LVM)
+#
+# CONFIG_MD is not set
+
+#
+# Fusion MPT device support
+#
+
+#
+# IEEE 1394 (FireWire) support
+#
+# CONFIG_IEEE1394 is not set
+
+#
+# I2O device support
+#
+# CONFIG_I2O is not set
+
+#
+# Networking support
+#
+CONFIG_NET=y
+
+#
+# Networking options
+#
+# CONFIG_PACKET is not set
+# CONFIG_NETLINK_DEV is not set
+CONFIG_UNIX=y
+# CONFIG_NET_KEY is not set
+CONFIG_INET=y
+CONFIG_IP_MULTICAST=y
+CONFIG_IP_ADVANCED_ROUTER=y
+CONFIG_IP_MULTIPLE_TABLES=y
+CONFIG_IP_ROUTE_MULTIPATH=y
+CONFIG_IP_ROUTE_VERBOSE=y
+CONFIG_IP_PNP=y
+CONFIG_IP_PNP_DHCP=y
+CONFIG_IP_PNP_BOOTP=y
+CONFIG_IP_PNP_RARP=y
+# CONFIG_NET_IPIP is not set
+# CONFIG_NET_IPGRE is not set
+# CONFIG_IP_MROUTE is not set
+# CONFIG_ARPD is not set
+# CONFIG_SYN_COOKIES is not set
+# CONFIG_INET_AH is not set
+# CONFIG_INET_ESP is not set
+# CONFIG_INET_IPCOMP is not set
+# CONFIG_INET_TUNNEL is not set
+# CONFIG_IP_TCPDIAG is not set
+# CONFIG_IP_TCPDIAG_IPV6 is not set
+# CONFIG_IPV6 is not set
+# CONFIG_NETFILTER is not set
+
+#
+# SCTP Configuration (EXPERIMENTAL)
+#
+# CONFIG_IP_SCTP is not set
+# CONFIG_ATM is not set
+# CONFIG_BRIDGE is not set
+# CONFIG_VLAN_8021Q is not set
+# CONFIG_DECNET is not set
+# CONFIG_LLC2 is not set
+# CONFIG_IPX is not set
+# CONFIG_ATALK is not set
+# CONFIG_X25 is not set
+# CONFIG_LAPB is not set
+# CONFIG_NET_DIVERT is not set
+# CONFIG_ECONET is not set
+# CONFIG_WAN_ROUTER is not set
+
+#
+# QoS and/or fair queueing
+#
+CONFIG_NET_SCHED=y
+CONFIG_NET_SCH_CLK_JIFFIES=y
+# CONFIG_NET_SCH_CLK_GETTIMEOFDAY is not set
+# CONFIG_NET_SCH_CLK_CPU is not set
+CONFIG_NET_SCH_CBQ=m
+CONFIG_NET_SCH_HTB=m
+# CONFIG_NET_SCH_HFSC is not set
+CONFIG_NET_SCH_PRIO=m
+CONFIG_NET_SCH_RED=m
+CONFIG_NET_SCH_SFQ=m
+CONFIG_NET_SCH_TEQL=m
+CONFIG_NET_SCH_TBF=m
+CONFIG_NET_SCH_GRED=m
+CONFIG_NET_SCH_DSMARK=m
+# CONFIG_NET_SCH_NETEM is not set
+CONFIG_NET_SCH_INGRESS=m
+CONFIG_NET_QOS=y
+CONFIG_NET_ESTIMATOR=y
+CONFIG_NET_CLS=y
+CONFIG_NET_CLS_TCINDEX=m
+CONFIG_NET_CLS_ROUTE4=m
+CONFIG_NET_CLS_ROUTE=y
+CONFIG_NET_CLS_FW=m
+CONFIG_NET_CLS_U32=m
+# CONFIG_CLS_U32_PERF is not set
+# CONFIG_NET_CLS_IND is not set
+CONFIG_NET_CLS_RSVP=m
+CONFIG_NET_CLS_RSVP6=m
+# CONFIG_NET_CLS_ACT is not set
+CONFIG_NET_CLS_POLICE=y
+
+#
+# Network testing
+#
+# CONFIG_NET_PKTGEN is not set
+# CONFIG_NETPOLL is not set
+# CONFIG_NET_POLL_CONTROLLER is not set
+# CONFIG_HAMRADIO is not set
+# CONFIG_IRDA is not set
+# CONFIG_BT is not set
+CONFIG_NETDEVICES=y
+CONFIG_DUMMY=y
+# CONFIG_BONDING is not set
+# CONFIG_EQUALIZER is not set
+# CONFIG_TUN is not set
+
+#
+# ARCnet devices
+#
+# CONFIG_ARCNET is not set
+
+#
+# Ethernet (10 or 100Mbit)
+#
+CONFIG_NET_ETHERNET=y
+# CONFIG_MII is not set
+CONFIG_XT2000_SONIC=y
+# CONFIG_HAPPYMEAL is not set
+# CONFIG_SUNGEM is not set
+# CONFIG_NET_VENDOR_3COM is not set
+
+#
+# Tulip family network device support
+#
+# CONFIG_NET_TULIP is not set
+# CONFIG_HP100 is not set
+# CONFIG_NET_PCI is not set
+
+#
+# Ethernet (1000 Mbit)
+#
+# CONFIG_ACENIC is not set
+# CONFIG_DL2K is not set
+# CONFIG_E1000 is not set
+# CONFIG_NS83820 is not set
+# CONFIG_HAMACHI is not set
+# CONFIG_YELLOWFIN is not set
+# CONFIG_R8169 is not set
+# CONFIG_SK98LIN is not set
+# CONFIG_TIGON3 is not set
+
+#
+# Ethernet (10000 Mbit)
+#
+# CONFIG_IXGB is not set
+# CONFIG_S2IO is not set
+
+#
+# Token Ring devices
+#
+# CONFIG_TR is not set
+
+#
+# Wireless LAN (non-hamradio)
+#
+CONFIG_NET_RADIO=y
+
+#
+# Obsolete Wireless cards support (pre-802.11)
+#
+CONFIG_STRIP=m
+
+#
+# Wireless 802.11b ISA/PCI cards support
+#
+CONFIG_HERMES=m
+# CONFIG_PLX_HERMES is not set
+# CONFIG_TMD_HERMES is not set
+# CONFIG_PCI_HERMES is not set
+# CONFIG_ATMEL is not set
+
+#
+# Prism GT/Duette 802.11(a/b/g) PCI/Cardbus support
+#
+# CONFIG_PRISM54 is not set
+CONFIG_NET_WIRELESS=y
+
+#
+# Wan interfaces
+#
+# CONFIG_WAN is not set
+# CONFIG_FDDI is not set
+# CONFIG_HIPPI is not set
+# CONFIG_PPP is not set
+# CONFIG_SLIP is not set
+# CONFIG_SHAPER is not set
+# CONFIG_NETCONSOLE is not set
+
+#
+# ISDN subsystem
+#
+# CONFIG_ISDN is not set
+
+#
+# Telephony Support
+#
+# CONFIG_PHONE is not set
+
+#
+# Input device support
+#
+CONFIG_INPUT=y
+
+#
+# Userland interfaces
+#
+CONFIG_INPUT_MOUSEDEV=y
+# CONFIG_INPUT_MOUSEDEV_PSAUX is not set
+CONFIG_INPUT_MOUSEDEV_SCREEN_X=1024
+CONFIG_INPUT_MOUSEDEV_SCREEN_Y=768
+# CONFIG_INPUT_JOYDEV is not set
+# CONFIG_INPUT_TSDEV is not set
+# CONFIG_INPUT_EVDEV is not set
+# CONFIG_INPUT_EVBUG is not set
+
+#
+# Input I/O drivers
+#
+# CONFIG_GAMEPORT is not set
+CONFIG_SOUND_GAMEPORT=y
+CONFIG_SERIO=y
+# CONFIG_SERIO_I8042 is not set
+# CONFIG_SERIO_SERPORT is not set
+# CONFIG_SERIO_CT82C710 is not set
+# CONFIG_SERIO_PCIPS2 is not set
+# CONFIG_SERIO_RAW is not set
+
+#
+# Input Device Drivers
+#
+# CONFIG_INPUT_KEYBOARD is not set
+# CONFIG_INPUT_MOUSE is not set
+# CONFIG_INPUT_JOYSTICK is not set
+# CONFIG_INPUT_TOUCHSCREEN is not set
+# CONFIG_INPUT_MISC is not set
+
+#
+# Character devices
+#
+CONFIG_VT=y
+CONFIG_VT_CONSOLE=y
+CONFIG_HW_CONSOLE=y
+# CONFIG_SERIAL_NONSTANDARD is not set
+
+#
+# Serial drivers
+#
+CONFIG_SERIAL_8250=y
+CONFIG_SERIAL_8250_CONSOLE=y
+CONFIG_SERIAL_8250_NR_UARTS=4
+# CONFIG_SERIAL_8250_EXTENDED is not set
+
+#
+# Non-8250 serial port support
+#
+CONFIG_SERIAL_CORE=y
+CONFIG_SERIAL_CORE_CONSOLE=y
+CONFIG_UNIX98_PTYS=y
+CONFIG_LEGACY_PTYS=y
+CONFIG_LEGACY_PTY_COUNT=256
+
+#
+# IPMI
+#
+# CONFIG_IPMI_HANDLER is not set
+
+#
+# Watchdog Cards
+#
+# CONFIG_WATCHDOG is not set
+# CONFIG_RTC is not set
+# CONFIG_GEN_RTC is not set
+# CONFIG_DTLK is not set
+# CONFIG_R3964 is not set
+# CONFIG_APPLICOM is not set
+
+#
+# Ftape, the floppy tape device driver
+#
+# CONFIG_DRM is not set
+# CONFIG_RAW_DRIVER is not set
+
+#
+# I2C support
+#
+# CONFIG_I2C is not set
+
+#
+# Dallas's 1-wire bus
+#
+# CONFIG_W1 is not set
+
+#
+# Misc devices
+#
+
+#
+# Multimedia devices
+#
+# CONFIG_VIDEO_DEV is not set
+
+#
+# Digital Video Broadcasting Devices
+#
+# CONFIG_DVB is not set
+
+#
+# Graphics support
+#
+# CONFIG_FB is not set
+
+#
+# Console display driver support
+#
+# CONFIG_VGA_CONSOLE is not set
+CONFIG_DUMMY_CONSOLE=y
+# CONFIG_BACKLIGHT_LCD_SUPPORT is not set
+
+#
+# Sound
+#
+# CONFIG_SOUND is not set
+
+#
+# USB support
+#
+# CONFIG_USB is not set
+CONFIG_USB_ARCH_HAS_HCD=y
+CONFIG_USB_ARCH_HAS_OHCI=y
+
+#
+# NOTE: USB_STORAGE enables SCSI, and 'SCSI disk support' may also be needed; see USB_STORAGE Help for more information
+#
+
+#
+# USB Gadget Support
+#
+# CONFIG_USB_GADGET is not set
+
+#
+# MMC/SD Card support
+#
+# CONFIG_MMC is not set
+
+#
+# InfiniBand support
+#
+# CONFIG_INFINIBAND is not set
+
+#
+# File systems
+#
+# CONFIG_EXT2_FS is not set
+# CONFIG_EXT3_FS is not set
+# CONFIG_JBD is not set
+# CONFIG_REISERFS_FS is not set
+# CONFIG_JFS_FS is not set
+# CONFIG_XFS_FS is not set
+# CONFIG_MINIX_FS is not set
+# CONFIG_ROMFS_FS is not set
+# CONFIG_QUOTA is not set
+CONFIG_DNOTIFY=y
+# CONFIG_AUTOFS_FS is not set
+# CONFIG_AUTOFS4_FS is not set
+
+#
+# CD-ROM/DVD Filesystems
+#
+# CONFIG_ISO9660_FS is not set
+# CONFIG_UDF_FS is not set
+
+#
+# DOS/FAT/NT Filesystems
+#
+# CONFIG_MSDOS_FS is not set
+# CONFIG_VFAT_FS is not set
+# CONFIG_NTFS_FS is not set
+
+#
+# Pseudo filesystems
+#
+CONFIG_PROC_FS=y
+# CONFIG_PROC_KCORE is not set
+CONFIG_SYSFS=y
+CONFIG_DEVFS_FS=y
+# CONFIG_DEVFS_MOUNT is not set
+# CONFIG_DEVFS_DEBUG is not set
+# CONFIG_DEVPTS_FS_XATTR is not set
+# CONFIG_TMPFS is not set
+# CONFIG_HUGETLB_PAGE is not set
+CONFIG_RAMFS=y
+
+#
+# Miscellaneous filesystems
+#
+# CONFIG_ADFS_FS is not set
+# CONFIG_AFFS_FS is not set
+# CONFIG_HFS_FS is not set
+# CONFIG_HFSPLUS_FS is not set
+# CONFIG_BEFS_FS is not set
+# CONFIG_BFS_FS is not set
+# CONFIG_EFS_FS is not set
+# CONFIG_CRAMFS is not set
+# CONFIG_VXFS_FS is not set
+# CONFIG_HPFS_FS is not set
+# CONFIG_QNX4FS_FS is not set
+# CONFIG_SYSV_FS is not set
+# CONFIG_UFS_FS is not set
+
+#
+# Network File Systems
+#
+CONFIG_NFS_FS=y
+CONFIG_NFS_V3=y
+# CONFIG_NFS_V4 is not set
+# CONFIG_NFS_DIRECTIO is not set
+# CONFIG_NFSD is not set
+CONFIG_ROOT_NFS=y
+CONFIG_LOCKD=y
+CONFIG_LOCKD_V4=y
+# CONFIG_EXPORTFS is not set
+CONFIG_SUNRPC=y
+# CONFIG_RPCSEC_GSS_KRB5 is not set
+# CONFIG_RPCSEC_GSS_SPKM3 is not set
+# CONFIG_SMB_FS is not set
+# CONFIG_CIFS is not set
+# CONFIG_NCP_FS is not set
+# CONFIG_CODA_FS is not set
+# CONFIG_AFS_FS is not set
+
+#
+# Partition Types
+#
+# CONFIG_PARTITION_ADVANCED is not set
+CONFIG_MSDOS_PARTITION=y
+
+#
+# Native Language Support
+#
+# CONFIG_NLS is not set
+
+#
+# Kernel hacking
+#
+CONFIG_DEBUG_KERNEL=y
+# CONFIG_DEBUG_STACKOVERFLOW is not set
+# CONFIG_DEBUG_SLAB is not set
+CONFIG_MAGIC_SYSRQ=y
+# CONFIG_DEBUG_SPINLOCK is not set
+# CONFIG_DEBUG_PAGEALLOC is not set
+# CONFIG_DEBUG_INFO is not set
+# CONFIG_DEBUG_SPINLOCK_SLEEP is not set
+# CONFIG_KGDB is not set
+
+#
+# Security options
+#
+# CONFIG_KEYS is not set
+# CONFIG_SECURITY is not set
+
+#
+# Cryptographic options
+#
+# CONFIG_CRYPTO is not set
+
+#
+# Hardware crypto devices
+#
+
+#
+# Library routines
+#
+# CONFIG_CRC_CCITT is not set
+# CONFIG_CRC32 is not set
+# CONFIG_LIBCRC32C is not set
diff --git a/arch/xtensa/configs/iss_defconfig b/arch/xtensa/configs/iss_defconfig
new file mode 100644
index 000000000000..802621dd4867
--- /dev/null
+++ b/arch/xtensa/configs/iss_defconfig
@@ -0,0 +1,531 @@
+#
+# Automatically generated make config: don't edit
+# Linux kernel version: 2.6.11-rc2
+# Fri Feb 25 19:21:24 2005
+#
+CONFIG_FRAME_POINTER=y
+CONFIG_XTENSA=y
+# CONFIG_UID16 is not set
+CONFIG_RWSEM_XCHGADD_ALGORITHM=y
+CONFIG_HAVE_DEC_LOCK=y
+CONFIG_GENERIC_HARDIRQS=y
+
+#
+# Code maturity level options
+#
+CONFIG_EXPERIMENTAL=y
+CONFIG_CLEAN_COMPILE=y
+CONFIG_BROKEN_ON_SMP=y
+
+#
+# General setup
+#
+CONFIG_LOCALVERSION=""
+CONFIG_SWAP=y
+CONFIG_SYSVIPC=y
+# CONFIG_POSIX_MQUEUE is not set
+# CONFIG_BSD_PROCESS_ACCT is not set
+CONFIG_SYSCTL=y
+# CONFIG_AUDIT is not set
+CONFIG_LOG_BUF_SHIFT=14
+# CONFIG_HOTPLUG is not set
+# CONFIG_KOBJECT_UEVENT is not set
+# CONFIG_IKCONFIG is not set
+CONFIG_EMBEDDED=y
+CONFIG_KALLSYMS=y
+# CONFIG_KALLSYMS_ALL is not set
+# CONFIG_KALLSYMS_EXTRA_PASS is not set
+CONFIG_FUTEX=y
+CONFIG_EPOLL=y
+# CONFIG_CC_OPTIMIZE_FOR_SIZE is not set
+CONFIG_SHMEM=y
+CONFIG_CC_ALIGN_FUNCTIONS=0
+CONFIG_CC_ALIGN_LABELS=0
+CONFIG_CC_ALIGN_LOOPS=0
+CONFIG_CC_ALIGN_JUMPS=0
+# CONFIG_TINY_SHMEM is not set
+
+#
+# Loadable module support
+#
+# CONFIG_MODULES is not set
+
+#
+# Processor type and features
+#
+CONFIG_XTENSA_ARCH_LINUX_BE=y
+# CONFIG_XTENSA_ARCH_LINUX_LE is not set
+# CONFIG_XTENSA_ARCH_LINUX_TEST is not set
+# CONFIG_XTENSA_ARCH_S5 is not set
+# CONFIG_XTENSA_CUSTOM is not set
+CONFIG_MMU=y
+# CONFIG_XTENSA_UNALIGNED_USER is not set
+# CONFIG_PREEMPT is not set
+# CONFIG_MATH_EMULATION is not set
+# CONFIG_HIGHMEM is not set
+
+#
+# Platform options
+#
+CONFIG_XTENSA_PLATFORM_ISS=y
+# CONFIG_XTENSA_PLATFORM_XT2000 is not set
+# CONFIG_XTENSA_PLATFORM_ARUBA is not set
+# CONFIG_XTENSA_CALIBRATE_CCOUNT is not set
+CONFIG_XTENSA_CPU_CLOCK=10
+# CONFIG_GENERIC_CALIBRATE_DELAY is not set
+CONFIG_CMDLINE_BOOL=y
+CONFIG_CMDLINE="console=ttyS0,38400 eth0=tuntap,,tap0 ip=192.168.168.5:192.168.168.1 root=nfs nfsroot=192.168.168.1:/opt/montavista/pro/devkit/xtensa/linux_be/target"
+CONFIG_SERIAL_CONSOLE=y
+CONFIG_XTENSA_ISS_NETWORK=y
+
+#
+# Bus options
+#
+
+#
+# PCCARD (PCMCIA/CardBus) support
+#
+# CONFIG_PCCARD is not set
+
+#
+# PC-card bridges
+#
+
+#
+# PCI Hotplug Support
+#
+
+#
+# Exectuable file formats
+#
+CONFIG_KCORE_ELF=y
+CONFIG_BINFMT_ELF=y
+# CONFIG_BINFMT_MISC is not set
+
+#
+# Device Drivers
+#
+
+#
+# Generic Driver Options
+#
+# CONFIG_STANDALONE is not set
+CONFIG_PREVENT_FIRMWARE_BUILD=y
+# CONFIG_FW_LOADER is not set
+# CONFIG_DEBUG_DRIVER is not set
+
+#
+# Memory Technology Devices (MTD)
+#
+# CONFIG_MTD is not set
+
+#
+# Parallel port support
+#
+# CONFIG_PARPORT is not set
+
+#
+# Plug and Play support
+#
+
+#
+# Block devices
+#
+# CONFIG_BLK_DEV_FD is not set
+# CONFIG_BLK_DEV_COW_COMMON is not set
+# CONFIG_BLK_DEV_LOOP is not set
+# CONFIG_BLK_DEV_NBD is not set
+# CONFIG_BLK_DEV_RAM is not set
+CONFIG_BLK_DEV_RAM_COUNT=16
+CONFIG_INITRAMFS_SOURCE=""
+# CONFIG_CDROM_PKTCDVD is not set
+
+#
+# IO Schedulers
+#
+CONFIG_IOSCHED_NOOP=y
+# CONFIG_IOSCHED_AS is not set
+# CONFIG_IOSCHED_DEADLINE is not set
+# CONFIG_IOSCHED_CFQ is not set
+# CONFIG_ATA_OVER_ETH is not set
+
+#
+# ATA/ATAPI/MFM/RLL support
+#
+# CONFIG_IDE is not set
+
+#
+# SCSI device support
+#
+# CONFIG_SCSI is not set
+
+#
+# Multi-device support (RAID and LVM)
+#
+# CONFIG_MD is not set
+
+#
+# Fusion MPT device support
+#
+
+#
+# IEEE 1394 (FireWire) support
+#
+
+#
+# I2O device support
+#
+
+#
+# Networking support
+#
+CONFIG_NET=y
+
+#
+# Networking options
+#
+CONFIG_PACKET=y
+# CONFIG_PACKET_MMAP is not set
+# CONFIG_NETLINK_DEV is not set
+CONFIG_UNIX=y
+# CONFIG_NET_KEY is not set
+CONFIG_INET=y
+# CONFIG_IP_MULTICAST is not set
+# CONFIG_IP_ADVANCED_ROUTER is not set
+CONFIG_IP_PNP=y
+CONFIG_IP_PNP_DHCP=y
+CONFIG_IP_PNP_BOOTP=y
+CONFIG_IP_PNP_RARP=y
+# CONFIG_NET_IPIP is not set
+# CONFIG_NET_IPGRE is not set
+# CONFIG_ARPD is not set
+# CONFIG_SYN_COOKIES is not set
+# CONFIG_INET_AH is not set
+# CONFIG_INET_ESP is not set
+# CONFIG_INET_IPCOMP is not set
+# CONFIG_INET_TUNNEL is not set
+# CONFIG_IP_TCPDIAG is not set
+# CONFIG_IP_TCPDIAG_IPV6 is not set
+# CONFIG_IPV6 is not set
+# CONFIG_NETFILTER is not set
+
+#
+# SCTP Configuration (EXPERIMENTAL)
+#
+# CONFIG_IP_SCTP is not set
+# CONFIG_SCTP_HMAC_NONE is not set
+# CONFIG_SCTP_HMAC_SHA1 is not set
+# CONFIG_SCTP_HMAC_MD5 is not set
+# CONFIG_ATM is not set
+# CONFIG_BRIDGE is not set
+# CONFIG_VLAN_8021Q is not set
+# CONFIG_DECNET is not set
+# CONFIG_LLC2 is not set
+# CONFIG_IPX is not set
+# CONFIG_ATALK is not set
+# CONFIG_X25 is not set
+# CONFIG_LAPB is not set
+# CONFIG_NET_DIVERT is not set
+# CONFIG_ECONET is not set
+# CONFIG_WAN_ROUTER is not set
+
+#
+# QoS and/or fair queueing
+#
+# CONFIG_NET_SCHED is not set
+# CONFIG_NET_SCH_CLK_JIFFIES is not set
+# CONFIG_NET_SCH_CLK_GETTIMEOFDAY is not set
+# CONFIG_NET_SCH_CLK_CPU is not set
+# CONFIG_NET_CLS_ROUTE is not set
+
+#
+# Network testing
+#
+# CONFIG_NET_PKTGEN is not set
+# CONFIG_NETPOLL is not set
+# CONFIG_NET_POLL_CONTROLLER is not set
+# CONFIG_HAMRADIO is not set
+# CONFIG_IRDA is not set
+# CONFIG_BT is not set
+# CONFIG_NETDEVICES is not set
+
+#
+# ISDN subsystem
+#
+# CONFIG_ISDN is not set
+
+#
+# Telephony Support
+#
+# CONFIG_PHONE is not set
+
+#
+# Input device support
+#
+CONFIG_INPUT=y
+
+#
+# Userland interfaces
+#
+# CONFIG_INPUT_MOUSEDEV is not set
+# CONFIG_INPUT_JOYDEV is not set
+# CONFIG_INPUT_TSDEV is not set
+# CONFIG_INPUT_EVDEV is not set
+# CONFIG_INPUT_EVBUG is not set
+
+#
+# Input I/O drivers
+#
+# CONFIG_GAMEPORT is not set
+CONFIG_SOUND_GAMEPORT=y
+# CONFIG_SERIO is not set
+# CONFIG_SERIO_I8042 is not set
+
+#
+# Input Device Drivers
+#
+# CONFIG_INPUT_KEYBOARD is not set
+# CONFIG_INPUT_MOUSE is not set
+# CONFIG_INPUT_JOYSTICK is not set
+# CONFIG_INPUT_TOUCHSCREEN is not set
+# CONFIG_INPUT_MISC is not set
+
+#
+# Character devices
+#
+CONFIG_VT=y
+CONFIG_VT_CONSOLE=y
+CONFIG_HW_CONSOLE=y
+# CONFIG_SERIAL_NONSTANDARD is not set
+
+#
+# Serial drivers
+#
+# CONFIG_SERIAL_8250 is not set
+
+#
+# Non-8250 serial port support
+#
+CONFIG_UNIX98_PTYS=y
+CONFIG_LEGACY_PTYS=y
+CONFIG_LEGACY_PTY_COUNT=256
+
+#
+# IPMI
+#
+# CONFIG_IPMI_HANDLER is not set
+
+#
+# Watchdog Cards
+#
+CONFIG_WATCHDOG=y
+CONFIG_WATCHDOG_NOWAYOUT=y
+
+#
+# Watchdog Device Drivers
+#
+CONFIG_SOFT_WATCHDOG=y
+# CONFIG_RTC is not set
+# CONFIG_GEN_RTC is not set
+# CONFIG_DTLK is not set
+# CONFIG_R3964 is not set
+
+#
+# Ftape, the floppy tape device driver
+#
+# CONFIG_DRM is not set
+# CONFIG_RAW_DRIVER is not set
+
+#
+# I2C support
+#
+# CONFIG_I2C is not set
+
+#
+# Dallas's 1-wire bus
+#
+# CONFIG_W1 is not set
+
+#
+# Misc devices
+#
+
+#
+# Multimedia devices
+#
+# CONFIG_VIDEO_DEV is not set
+
+#
+# Digital Video Broadcasting Devices
+#
+# CONFIG_DVB is not set
+
+#
+# Graphics support
+#
+# CONFIG_FB is not set
+
+#
+# Console display driver support
+#
+# CONFIG_VGA_CONSOLE is not set
+CONFIG_DUMMY_CONSOLE=y
+# CONFIG_BACKLIGHT_LCD_SUPPORT is not set
+
+#
+# Sound
+#
+# CONFIG_SOUND is not set
+
+#
+# USB support
+#
+# CONFIG_USB_ARCH_HAS_HCD is not set
+# CONFIG_USB_ARCH_HAS_OHCI is not set
+
+#
+# NOTE: USB_STORAGE enables SCSI, and 'SCSI disk support' may also be needed; see USB_STORAGE Help for more information
+#
+
+#
+# USB Gadget Support
+#
+# CONFIG_USB_GADGET is not set
+
+#
+# MMC/SD Card support
+#
+# CONFIG_MMC is not set
+
+#
+# InfiniBand support
+#
+# CONFIG_INFINIBAND is not set
+
+#
+# File systems
+#
+# CONFIG_EXT2_FS is not set
+# CONFIG_EXT3_FS is not set
+# CONFIG_JBD is not set
+# CONFIG_REISERFS_FS is not set
+# CONFIG_JFS_FS is not set
+# CONFIG_XFS_FS is not set
+# CONFIG_MINIX_FS is not set
+# CONFIG_ROMFS_FS is not set
+# CONFIG_QUOTA is not set
+# CONFIG_DNOTIFY is not set
+# CONFIG_AUTOFS_FS is not set
+# CONFIG_AUTOFS4_FS is not set
+
+#
+# CD-ROM/DVD Filesystems
+#
+# CONFIG_ISO9660_FS is not set
+# CONFIG_UDF_FS is not set
+
+#
+# DOS/FAT/NT Filesystems
+#
+# CONFIG_MSDOS_FS is not set
+# CONFIG_VFAT_FS is not set
+# CONFIG_NTFS_FS is not set
+
+#
+# Pseudo filesystems
+#
+CONFIG_PROC_FS=y
+CONFIG_PROC_KCORE=y
+CONFIG_SYSFS=y
+CONFIG_DEVFS_FS=y
+CONFIG_DEVFS_MOUNT=y
+# CONFIG_DEVFS_DEBUG is not set
+# CONFIG_DEVPTS_FS_XATTR is not set
+CONFIG_TMPFS=y
+# CONFIG_TMPFS_XATTR is not set
+# CONFIG_HUGETLB_PAGE is not set
+CONFIG_RAMFS=y
+
+#
+# Miscellaneous filesystems
+#
+# CONFIG_ADFS_FS is not set
+# CONFIG_AFFS_FS is not set
+# CONFIG_HFS_FS is not set
+# CONFIG_HFSPLUS_FS is not set
+# CONFIG_BEFS_FS is not set
+# CONFIG_BFS_FS is not set
+# CONFIG_EFS_FS is not set
+# CONFIG_CRAMFS is not set
+# CONFIG_VXFS_FS is not set
+# CONFIG_HPFS_FS is not set
+# CONFIG_QNX4FS_FS is not set
+# CONFIG_SYSV_FS is not set
+# CONFIG_UFS_FS is not set
+
+#
+# Network File Systems
+#
+CONFIG_NFS_FS=y
+CONFIG_NFS_V3=y
+# CONFIG_NFS_V4 is not set
+CONFIG_NFS_DIRECTIO=y
+# CONFIG_NFSD is not set
+CONFIG_ROOT_NFS=y
+CONFIG_LOCKD=y
+CONFIG_LOCKD_V4=y
+# CONFIG_EXPORTFS is not set
+CONFIG_SUNRPC=y
+# CONFIG_RPCSEC_GSS_KRB5 is not set
+# CONFIG_RPCSEC_GSS_SPKM3 is not set
+# CONFIG_SMB_FS is not set
+# CONFIG_CIFS is not set
+# CONFIG_NCP_FS is not set
+# CONFIG_CODA_FS is not set
+# CONFIG_AFS_FS is not set
+
+#
+# Partition Types
+#
+# CONFIG_PARTITION_ADVANCED is not set
+CONFIG_MSDOS_PARTITION=y
+
+#
+# Native Language Support
+#
+# CONFIG_NLS is not set
+
+#
+# Kernel hacking
+#
+CONFIG_DEBUG_KERNEL=y
+# CONFIG_DEBUG_STACKOVERFLOW is not set
+# CONFIG_DEBUG_SLAB is not set
+# CONFIG_MAGIC_SYSRQ is not set
+# CONFIG_DEBUG_SPINLOCK is not set
+# CONFIG_DEBUG_PAGEALLOC is not set
+# CONFIG_DEBUG_INFO is not set
+# CONFIG_DEBUG_SPINLOCK_SLEEP is not set
+# CONFIG_KGDB is not set
+
+#
+# Security options
+#
+# CONFIG_KEYS is not set
+# CONFIG_SECURITY is not set
+
+#
+# Cryptographic options
+#
+# CONFIG_CRYPTO is not set
+
+#
+# Hardware crypto devices
+#
+
+#
+# Library routines
+#
+# CONFIG_CRC_CCITT is not set
+# CONFIG_CRC32 is not set
+# CONFIG_LIBCRC32C is not set
diff --git a/arch/xtensa/kernel/Makefile b/arch/xtensa/kernel/Makefile
new file mode 100644
index 000000000000..d573017a5dde
--- /dev/null
+++ b/arch/xtensa/kernel/Makefile
@@ -0,0 +1,18 @@
+#
+# Makefile for the Linux/Xtensa kernel.
+#
+
+extra-y := head.o vmlinux.lds
+
+
+obj-y := align.o entry.o irq.o coprocessor.o process.o ptrace.o semaphore.o  \
+	 setup.o signal.o syscalls.o time.o traps.o vectors.o platform.o  \
+	 pci-dma.o
+
+## windowspill.o
+
+obj-$(CONFIG_KGDB) += xtensa-stub.o
+obj-$(CONFIG_PCI) += pci.o
+obj-$(CONFIG_MODULES) += xtensa_ksyms.o module.o
+
+
diff --git a/arch/xtensa/kernel/align.S b/arch/xtensa/kernel/align.S
new file mode 100644
index 000000000000..74b1e90ef08c
--- /dev/null
+++ b/arch/xtensa/kernel/align.S
@@ -0,0 +1,459 @@
+/*
+ * arch/xtensa/kernel/align.S
+ *
+ * Handle unalignment exceptions in kernel space.
+ *
+ * This file is subject to the terms and conditions of the GNU General
+ * Public License.  See the file "COPYING" in the main directory of
+ * this archive for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica, Inc.
+ *
+ * Rewritten by Chris Zankel <chris@zankel.net>
+ *
+ * Based on work from Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * and Marc Gauthier <marc@tensilica.com, marc@alimni.uwaterloo.ca>
+ */
+
+#include <linux/linkage.h>
+#include <asm/ptrace.h>
+#include <asm/ptrace.h>
+#include <asm/current.h>
+#include <asm/offsets.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/page.h>
+#include <asm/thread_info.h>
+
+#if XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION
+
+/*  First-level exception handler for unaligned exceptions.
+ *
+ *  Note: This handler works only for kernel exceptions.  Unaligned user
+ *        access should get a seg fault.
+ */
+
+/* Big and little endian 16-bit values are located in
+ * different halves of a register.  HWORD_START helps to
+ * abstract the notion of extracting a 16-bit value from a
+ * register.
+ * We also have to define new shifting instructions because
+ * lsb and msb are on 'opposite' ends in a register for
+ * different endian machines.
+ *
+ * Assume a memory region in ascending address:
+ *   	0 1 2 3|4 5 6 7
+ *
+ * When loading one word into a register, the content of that register is:
+ *  LE	3 2 1 0, 7 6 5 4
+ *  BE  0 1 2 3, 4 5 6 7
+ *
+ * Masking the bits of the higher/lower address means:
+ *  LE  X X 0 0, 0 0 X X
+ *  BE	0 0 X X, X X 0 0
+ *
+ * Shifting to higher/lower addresses, means:
+ *  LE  shift left / shift right
+ *  BE  shift right / shift left
+ *
+ * Extracting 16 bits from a 32 bit reg. value to higher/lower address means:
+ *  LE  mask 0 0 X X / shift left
+ *  BE  shift left / mask 0 0 X X
+ */
+
+#define UNALIGNED_USER_EXCEPTION
+
+#if XCHAL_HAVE_BE
+
+#define HWORD_START	16
+#define	INSN_OP0	28
+#define	INSN_T		24
+#define	INSN_OP1	16
+
+.macro __src_b	r, w0, w1;	src	\r, \w0, \w1;	.endm
+.macro __ssa8	r;		ssa8b	\r;		.endm
+.macro __ssa8r	r;		ssa8l	\r;		.endm
+.macro __sh	r, s;		srl	\r, \s;		.endm
+.macro __sl	r, s;		sll	\r, \s;		.endm
+.macro __exth	r, s;		extui	\r, \s, 0, 16;	.endm
+.macro __extl	r, s;		slli	\r, \s, 16;	.endm
+
+#else
+
+#define HWORD_START	0
+#define	INSN_OP0	0
+#define	INSN_T		4
+#define	INSN_OP1	12
+
+.macro __src_b	r, w0, w1;	src	\r, \w1, \w0;	.endm
+.macro __ssa8	r;		ssa8l	\r;		.endm
+.macro __ssa8r	r;		ssa8b	\r;		.endm
+.macro __sh	r, s;		sll	\r, \s;		.endm
+.macro __sl	r, s;		srl	\r, \s;		.endm
+.macro __exth	r, s;		slli	\r, \s, 16;	.endm
+.macro __extl	r, s;		extui	\r, \s, 0, 16;	.endm
+
+#endif
+
+/*
+ *	xxxx xxxx = imm8 field
+ *	     yyyy = imm4 field
+ *	     ssss = s field
+ *	     tttt = t field
+ *
+ *	       		 16		    0
+ *		          -------------------
+ *	L32I.N		  yyyy ssss tttt 1000
+ *	S32I.N	          yyyy ssss tttt 1001
+ *
+ *	       23			    0
+ *		-----------------------------
+ *	res	          0000           0010
+ *	L16UI	xxxx xxxx 0001 ssss tttt 0010
+ *	L32I	xxxx xxxx 0010 ssss tttt 0010
+ *	XXX	          0011 ssss tttt 0010
+ *	XXX	          0100 ssss tttt 0010
+ *	S16I	xxxx xxxx 0101 ssss tttt 0010
+ *	S32I	xxxx xxxx 0110 ssss tttt 0010
+ *	XXX	          0111 ssss tttt 0010
+ *	XXX	          1000 ssss tttt 0010
+ *	L16SI	xxxx xxxx 1001 ssss tttt 0010
+ *	XXX	          1010           0010
+ *      **L32AI	xxxx xxxx 1011 ssss tttt 0010 unsupported
+ *	XXX	          1100           0010
+ *	XXX	          1101           0010
+ *	XXX	          1110           0010
+ *	**S32RI	xxxx xxxx 1111 ssss tttt 0010 unsupported
+ *		-----------------------------
+ *                           ^         ^    ^
+ *    sub-opcode (NIBBLE_R) -+         |    |
+ *       t field (NIBBLE_T) -----------+    |
+ *  major opcode (NIBBLE_OP0) --------------+
+ */
+
+#define OP0_L32I_N	0x8		/* load immediate narrow */
+#define OP0_S32I_N	0x9		/* store immediate narrow */
+#define OP1_SI_MASK	0x4		/* OP1 bit set for stores */
+#define OP1_SI_BIT	2		/* OP1 bit number for stores */
+
+#define OP1_L32I	0x2
+#define OP1_L16UI	0x1
+#define OP1_L16SI	0x9
+#define OP1_L32AI	0xb
+
+#define OP1_S32I	0x6
+#define OP1_S16I	0x5
+#define OP1_S32RI	0xf
+
+/*
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ */
+
+
+ENTRY(fast_unaligned)
+
+	/* Note: We don't expect the address to be aligned on a word
+	 *       boundary. After all, the processor generated that exception
+	 *       and it would be a hardware fault.
+	 */
+
+	/* Save some working register */
+
+	s32i	a4, a2, PT_AREG4
+	s32i	a5, a2, PT_AREG5
+	s32i	a6, a2, PT_AREG6
+	s32i	a7, a2, PT_AREG7
+	s32i	a8, a2, PT_AREG8
+
+	rsr	a0, DEPC
+	xsr	a3, EXCSAVE_1
+	s32i	a0, a2, PT_AREG2
+	s32i	a3, a2, PT_AREG3
+
+	/* Keep value of SAR in a0 */
+
+	rsr	a0, SAR
+	rsr	a8, EXCVADDR		# load unaligned memory address
+
+	/* Now, identify one of the following load/store instructions.
+	 *
+	 * The only possible danger of a double exception on the
+	 * following l32i instructions is kernel code in vmalloc
+	 * memory. The processor was just executing at the EPC_1
+	 * address, and indeed, already fetched the instruction.  That
+	 * guarantees a TLB mapping, which hasn't been replaced by
+	 * this unaligned exception handler that uses only static TLB
+	 * mappings. However, high-level interrupt handlers might
+	 * modify TLB entries, so for the generic case, we register a
+	 * TABLE_FIXUP handler here, too.
+	 */
+
+	/* a3...a6 saved on stack, a2 = SP */
+
+	/* Extract the instruction that caused the unaligned access. */
+
+	rsr	a7, EPC_1	# load exception address
+	movi	a3, ~3
+	and	a3, a3, a7	# mask lower bits
+
+	l32i	a4, a3, 0	# load 2 words
+	l32i	a5, a3, 4
+
+	__ssa8	a7
+	__src_b	a4, a4, a5	# a4 has the instruction
+
+	/* Analyze the instruction (load or store?). */
+
+	extui	a5, a4, INSN_OP0, 4	# get insn.op0 nibble
+
+#if XCHAL_HAVE_NARROW
+	_beqi	a5, OP0_L32I_N, .Lload	# L32I.N, jump
+	addi	a6, a5, -OP0_S32I_N
+	_beqz	a6, .Lstore		# S32I.N, do a store
+#endif
+	/* 'store indicator bit' not set, jump */
+	_bbci.l	a4, OP1_SI_BIT + INSN_OP1, .Lload
+
+	/* Store: Jump to table entry to get the value in the source register.*/
+
+.Lstore:movi	a5, .Lstore_table	# table
+	extui	a6, a4, INSN_T, 4	# get source register
+	addx8	a5, a6, a5
+	jx	a5			# jump into table
+
+	/* Invalid instruction, CRITICAL! */
+.Linvalid_instruction_load:
+	j	.Linvalid_instruction
+
+	/* Load: Load memory address. */
+
+.Lload: movi	a3, ~3
+	and	a3, a3, a8		# align memory address
+
+	__ssa8	a8
+#ifdef UNALIGNED_USER_EXCEPTION
+	addi	a3, a3, 8
+	l32e	a5, a3, -8
+	l32e	a6, a3, -4
+#else
+	l32i	a5, a3, 0
+	l32i	a6, a3, 4
+#endif
+	__src_b	a3, a5, a6		# a3 has the data word
+
+#if XCHAL_HAVE_NARROW
+	addi	a7, a7, 2		# increment PC (assume 16-bit insn)
+
+	extui	a5, a4, INSN_OP0, 4
+	_beqi	a5, OP0_L32I_N, 1f	# l32i.n: jump
+
+	addi	a7, a7, 1
+#else
+	addi	a7, a7, 3
+#endif
+
+	extui	a5, a4, INSN_OP1, 4
+	_beqi	a5, OP1_L32I, 1f	# l32i: jump
+
+	extui	a3, a3, 0, 16		# extract lower 16 bits
+	_beqi	a5, OP1_L16UI, 1f
+	addi	a5, a5, -OP1_L16SI
+	_bnez	a5, .Linvalid_instruction_load
+
+	/* sign extend value */
+
+	slli	a3, a3, 16
+	srai	a3, a3, 16
+
+	/* Set target register. */
+
+1:
+
+#if XCHAL_HAVE_LOOP
+	rsr	a3, LEND		# check if we reached LEND
+	bne	a7, a3, 1f
+	rsr	a3, LCOUNT		# and LCOUNT != 0
+	beqz	a3, 1f
+	addi	a3, a3, -1		# decrement LCOUNT and set
+	rsr	a7, LBEG		# set PC to LBEGIN
+	wsr	a3, LCOUNT
+#endif
+
+1:	wsr	a7, EPC_1		# skip load instruction
+	extui	a4, a4, INSN_T, 4	# extract target register
+	movi	a5, .Lload_table
+	addx8	a4, a4, a5
+	jx	a4			# jump to entry for target register
+
+	.align	8
+.Lload_table:
+	s32i	a3, a2, PT_AREG0;	_j .Lexit;	.align 8
+	mov	a1, a3;			_j .Lexit;	.align 8 # fishy??
+	s32i	a3, a2, PT_AREG2;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG3;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG4;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG5;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG6;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG7;	_j .Lexit;	.align 8
+	s32i	a3, a2, PT_AREG8;	_j .Lexit;	.align 8
+	mov	a9, a3		;	_j .Lexit;	.align 8
+	mov	a10, a3		;	_j .Lexit;	.align 8
+	mov	a11, a3		;	_j .Lexit;	.align 8
+	mov	a12, a3		;	_j .Lexit;	.align 8
+	mov	a13, a3		;	_j .Lexit;	.align 8
+	mov	a14, a3		;	_j .Lexit;	.align 8
+	mov	a15, a3		;	_j .Lexit;	.align 8
+
+.Lstore_table:
+	l32i	a3, a2, PT_AREG0;	_j 1f;	.align 8
+	mov	a3, a1;			_j 1f;	.align 8	# fishy??
+	l32i	a3, a2, PT_AREG2;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG3;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG4;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG5;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG6;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG7;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG8;	_j 1f;	.align 8
+	mov	a3, a9		;	_j 1f;	.align 8
+	mov	a3, a10		;	_j 1f;	.align 8
+	mov	a3, a11		;	_j 1f;	.align 8
+	mov	a3, a12		;	_j 1f;	.align 8
+	mov	a3, a13		;	_j 1f;	.align 8
+	mov	a3, a14		;	_j 1f;	.align 8
+	mov	a3, a15		;	_j 1f;	.align 8
+
+1: 	# a7: instruction pointer, a4: instruction, a3: value
+
+	movi	a6, 0			# mask: ffffffff:00000000
+
+#if XCHAL_HAVE_NARROW
+	addi	a7, a7, 2		# incr. PC,assume 16-bit instruction
+
+	extui	a5, a4, INSN_OP0, 4	# extract OP0
+	addi	a5, a5, -OP0_S32I_N
+	_beqz	a5, 1f			# s32i.n: jump
+
+	addi	a7, a7, 1		# increment PC, 32-bit instruction
+#else
+	addi	a7, a7, 3		# increment PC, 32-bit instruction
+#endif
+
+	extui	a5, a4, INSN_OP1, 4	# extract OP1
+	_beqi	a5, OP1_S32I, 1f	# jump if 32 bit store
+	_bnei	a5, OP1_S16I, .Linvalid_instruction_store
+
+	movi	a5, -1
+	__extl	a3, a3			# get 16-bit value
+	__exth	a6, a5			# get 16-bit mask ffffffff:ffff0000
+
+	/* Get memory address */
+
+1:
+#if XCHAL_HAVE_LOOP
+	rsr	a3, LEND		# check if we reached LEND
+	bne	a7, a3, 1f
+	rsr	a3, LCOUNT		# and LCOUNT != 0
+	beqz	a3, 1f
+	addi	a3, a3, -1		# decrement LCOUNT and set
+	rsr	a7, LBEG		# set PC to LBEGIN
+	wsr	a3, LCOUNT
+#endif
+
+1:	wsr	a7, EPC_1		# skip store instruction
+	movi	a4, ~3
+	and	a4, a4, a8		# align memory address
+
+	/* Insert value into memory */
+
+	movi	a5, -1			# mask: ffffffff:XXXX0000
+#ifdef UNALIGNED_USER_EXCEPTION
+	addi	a4, a4, 8
+#endif
+
+	__ssa8r a8
+	__src_b	a7, a5, a6		# lo-mask  F..F0..0 (BE) 0..0F..F (LE)
+	__src_b	a6, a6, a5		# hi-mask  0..0F..F (BE) F..F0..0 (LE)
+#ifdef UNALIGNED_USER_EXCEPTION
+	l32e	a5, a4, -8
+#else
+	l32i	a5, a4, 0		# load lower address word
+#endif
+	and	a5, a5, a7		# mask
+	__sh	a7, a3 			# shift value
+	or	a5, a5, a7		# or with original value
+#ifdef UNALIGNED_USER_EXCEPTION
+	s32e	a5, a4, -8
+	l32e	a7, a4, -4
+#else
+	s32i	a5, a4, 0		# store
+	l32i	a7, a4, 4		# same for upper address word
+#endif
+	__sl	a5, a3
+	and	a6, a7, a6
+	or	a6, a6, a5
+#ifdef UNALIGNED_USER_EXCEPTION
+	s32e	a6, a4, -4
+#else
+	s32i	a6, a4, 4
+#endif
+
+	/* Done. restore stack and return */
+
+.Lexit:
+	movi	a4, 0
+	rsr	a3, EXCSAVE_1
+	s32i	a4, a3, EXC_TABLE_FIXUP
+
+	/* Restore working register */
+
+	l32i	a7, a2, PT_AREG7
+	l32i	a6, a2, PT_AREG6
+	l32i	a5, a2, PT_AREG5
+	l32i	a4, a2, PT_AREG4
+	l32i	a3, a2, PT_AREG3
+
+	/* restore SAR and return */
+
+	wsr	a0, SAR
+	l32i	a0, a2, PT_AREG0
+	l32i	a2, a2, PT_AREG2
+	rfe
+
+	/* We cannot handle this exception. */
+
+	.extern _kernel_exception
+.Linvalid_instruction_store:
+.Linvalid_instruction:
+
+	/* Restore a4...a8 and SAR, set SP, and jump to default exception. */
+
+	l32i	a8, a2, PT_AREG8
+	l32i	a7, a2, PT_AREG7
+	l32i	a6, a2, PT_AREG6
+	l32i	a5, a2, PT_AREG5
+	l32i	a4, a2, PT_AREG4
+	wsr	a0, SAR
+	mov	a1, a2
+
+	rsr	a0, PS
+        bbsi.l  a2, PS_UM_SHIFT, 1f     # jump if user mode
+
+	movi	a0, _kernel_exception
+	jx	a0
+
+1:	movi	a0, _user_exception
+	jx	a0
+
+
+#endif /* XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION */
+
diff --git a/arch/xtensa/kernel/asm-offsets.c b/arch/xtensa/kernel/asm-offsets.c
new file mode 100644
index 000000000000..840cd9a1d3d2
--- /dev/null
+++ b/arch/xtensa/kernel/asm-offsets.c
@@ -0,0 +1,94 @@
+/*
+ * arch/xtensa/kernel/asm-offsets.c
+ *
+ * Generates definitions from c-type structures used by assembly sources.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ */
+
+#include <asm/processor.h>
+
+#include <linux/types.h>
+#include <linux/sched.h>
+#include <linux/stddef.h>
+#include <linux/thread_info.h>
+#include <linux/ptrace.h>
+#include <asm/ptrace.h>
+#include <asm/processor.h>
+#include <asm/uaccess.h>
+
+#define DEFINE(sym, val) asm volatile("\n->" #sym " %0 " #val : : "i" (val))
+#define BLANK() asm volatile("\n->" : : )
+
+int main(void)
+{
+	/* struct pt_regs */
+	DEFINE(PT_PC, offsetof (struct pt_regs, pc));
+	DEFINE(PT_PS, offsetof (struct pt_regs, ps));
+	DEFINE(PT_DEPC, offsetof (struct pt_regs, depc));
+	DEFINE(PT_EXCCAUSE, offsetof (struct pt_regs, exccause));
+	DEFINE(PT_EXCVADDR, offsetof (struct pt_regs, excvaddr));
+	DEFINE(PT_DEBUGCAUSE, offsetof (struct pt_regs, debugcause));
+	DEFINE(PT_WMASK, offsetof (struct pt_regs, wmask));
+	DEFINE(PT_LBEG, offsetof (struct pt_regs, lbeg));
+	DEFINE(PT_LEND, offsetof (struct pt_regs, lend));
+	DEFINE(PT_LCOUNT, offsetof (struct pt_regs, lcount));
+	DEFINE(PT_SAR, offsetof (struct pt_regs, sar));
+	DEFINE(PT_SYSCALL, offsetof (struct pt_regs, syscall));
+	DEFINE(PT_AREG, offsetof (struct pt_regs, areg[0]));
+	DEFINE(PT_AREG0, offsetof (struct pt_regs, areg[0]));
+	DEFINE(PT_AREG1, offsetof (struct pt_regs, areg[1]));
+	DEFINE(PT_AREG2, offsetof (struct pt_regs, areg[2]));
+	DEFINE(PT_AREG3, offsetof (struct pt_regs, areg[3]));
+	DEFINE(PT_AREG4, offsetof (struct pt_regs, areg[4]));
+	DEFINE(PT_AREG5, offsetof (struct pt_regs, areg[5]));
+	DEFINE(PT_AREG6, offsetof (struct pt_regs, areg[6]));
+	DEFINE(PT_AREG7, offsetof (struct pt_regs, areg[7]));
+	DEFINE(PT_AREG8, offsetof (struct pt_regs, areg[8]));
+	DEFINE(PT_AREG9, offsetof (struct pt_regs, areg[9]));
+	DEFINE(PT_AREG10, offsetof (struct pt_regs, areg[10]));
+	DEFINE(PT_AREG11, offsetof (struct pt_regs, areg[11]));
+	DEFINE(PT_AREG12, offsetof (struct pt_regs, areg[12]));
+	DEFINE(PT_AREG13, offsetof (struct pt_regs, areg[13]));
+	DEFINE(PT_AREG14, offsetof (struct pt_regs, areg[14]));
+	DEFINE(PT_AREG15, offsetof (struct pt_regs, areg[15]));
+	DEFINE(PT_WINDOWBASE, offsetof (struct pt_regs, windowbase));
+	DEFINE(PT_WINDOWSTART, offsetof(struct pt_regs, windowstart));
+	DEFINE(PT_SIZE, sizeof(struct pt_regs));
+	DEFINE(PT_AREG_END, offsetof (struct pt_regs, areg[XCHAL_NUM_AREGS]));
+	DEFINE(PT_USER_SIZE, offsetof(struct pt_regs, areg[XCHAL_NUM_AREGS]));
+	BLANK();
+
+	/* struct task_struct */
+	DEFINE(TASK_PTRACE, offsetof (struct task_struct, ptrace));
+	DEFINE(TASK_MM, offsetof (struct task_struct, mm));
+	DEFINE(TASK_ACTIVE_MM, offsetof (struct task_struct, active_mm));
+	DEFINE(TASK_PID, offsetof (struct task_struct, pid));
+	DEFINE(TASK_THREAD, offsetof (struct task_struct, thread));
+	DEFINE(TASK_THREAD_INFO, offsetof (struct task_struct, thread_info));
+	DEFINE(TASK_STRUCT_SIZE, sizeof (struct task_struct));
+	BLANK();
+
+	/* struct thread_info (offset from start_struct) */
+	DEFINE(THREAD_RA, offsetof (struct task_struct, thread.ra));
+	DEFINE(THREAD_SP, offsetof (struct task_struct, thread.sp));
+	DEFINE(THREAD_CP_SAVE, offsetof (struct task_struct, thread.cp_save));
+	DEFINE(THREAD_CURRENT_DS, offsetof (struct task_struct, thread.current_ds));
+	BLANK();
+
+	/* struct mm_struct */
+	DEFINE(MM_USERS, offsetof(struct mm_struct, mm_users));
+	DEFINE(MM_PGD, offsetof (struct mm_struct, pgd));
+	DEFINE(MM_CONTEXT, offsetof (struct mm_struct, context));
+	BLANK();
+	DEFINE(PT_SINGLESTEP_BIT, PT_SINGLESTEP_BIT);
+	return 0;
+}
+
+
diff --git a/arch/xtensa/kernel/coprocessor.S b/arch/xtensa/kernel/coprocessor.S
new file mode 100644
index 000000000000..356192a4d39d
--- /dev/null
+++ b/arch/xtensa/kernel/coprocessor.S
@@ -0,0 +1,201 @@
+/*
+ * arch/xtensa/kernel/coprocessor.S
+ *
+ * Xtensa processor configuration-specific table of coprocessor and
+ * other custom register layout information.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2003 - 2005 Tensilica Inc.
+ *
+ * Marc Gauthier <marc@tensilica.com> <marc@alumni.uwaterloo.ca>
+ */
+
+/*
+ * This module contains a table that describes the layout of the various
+ * custom registers and states associated with each coprocessor, as well
+ * as those not associated with any coprocessor ("extra state").
+ * This table is included with core dumps and is available via the ptrace
+ * interface, allowing the layout of such register/state information to
+ * be modified in the kernel without affecting the debugger.  Each
+ * register or state is identified using a 32-bit "libdb target number"
+ * assigned when the Xtensa processor is generated.
+ */
+
+#include <linux/config.h>
+#include <linux/linkage.h>
+#include <asm/processor.h>
+
+#if XCHAL_HAVE_CP
+
+#define CP_LAST ((XCHAL_CP_MAX - 1) * COPROCESSOR_INFO_SIZE)
+
+ENTRY(release_coprocessors)
+
+	entry	a1, 16
+						# a2: task
+	movi	a3, 1 << XCHAL_CP_MAX 		# a3: coprocessor-bit
+	movi	a4, coprocessor_info+CP_LAST	# a4: owner-table
+						# a5: tmp
+	movi	a6, 0				# a6: 0
+	rsil	a7, LOCKLEVEL			# a7: PS
+
+1:	/* Check if task is coprocessor owner of coprocessor[i]. */
+
+	l32i	a5, a4, COPROCESSOR_INFO_OWNER
+	srli	a3, a3, 1
+	beqz	a3, 1f
+	addi	a4, a4, -8
+	beq	a2, a5, 1b
+
+	/* Found an entry: Clear entry CPENABLE bit to disable CP. */
+
+	rsr	a5, CPENABLE
+	s32i	a6, a4, COPROCESSOR_INFO_OWNER
+	xor	a5, a3, a5
+	wsr	a5, CPENABLE
+
+	bnez	a3, 1b
+
+1:	wsr	a7, PS
+	rsync
+	retw
+
+
+ENTRY(disable_coprocessor)
+	entry	sp, 16
+	rsil	a7, LOCKLEVEL
+	rsr	a3, CPENABLE
+	movi	a4, 1
+	ssl	a2
+	sll	a4, a4
+	and	a4, a3, a4
+	xor	a3, a3, a4
+	wsr	a3, CPENABLE
+	wsr	a7, PS
+	rsync
+	retw
+
+ENTRY(enable_coprocessor)
+	entry	sp, 16
+	rsil	a7, LOCKLEVEL
+	rsr	a3, CPENABLE
+	movi	a4, 1
+	ssl	a2
+	sll	a4, a4
+	or	a3, a3, a4
+	wsr	a3, CPENABLE
+	wsr	a7, PS
+	rsync
+	retw
+
+#endif
+
+ENTRY(save_coprocessor_extra)
+	entry	sp, 16
+	xchal_extra_store_funcbody
+	retw
+
+ENTRY(restore_coprocessor_extra)
+	entry	sp, 16
+	xchal_extra_load_funcbody
+	retw
+
+ENTRY(save_coprocessor_registers)
+	entry	sp, 16
+	xchal_cpi_store_funcbody
+	retw
+
+ENTRY(restore_coprocessor_registers)
+	entry	sp, 16
+	xchal_cpi_load_funcbody
+	retw
+
+
+/*
+ *  The Xtensa compile-time HAL (core.h) XCHAL_*_SA_CONTENTS_LIBDB macros
+ *  describe the contents of coprocessor & extra save areas in terms of
+ *  undefined CONTENTS_LIBDB_{SREG,UREG,REGF} macros.  We define these
+ *  latter macros here; they expand into a table of the format we want.
+ *  The general format is:
+ *
+ *	CONTENTS_LIBDB_SREG(libdbnum, offset, size, align, rsv1, name, sregnum,
+ *			    bitmask, rsv2, rsv3)
+ *	CONTENTS_LIBDB_UREG(libdbnum, offset, size, align, rsv1, name, uregnum,
+ *			    bitmask, rsv2, rsv3)
+ *	CONTENTS_LIBDB_REGF(libdbnum, offset, size, align, rsv1, name, index,
+ *			    numentries, contentsize, regname_base,
+ *			    regfile_name, rsv2, rsv3)
+ *
+ *  For this table, we only care about the <libdbnum>, <offset> and <size>
+ *  fields.
+ */
+
+/*  Map all XCHAL CONTENTS macros to the reg_entry asm macro defined below:  */
+
+#define CONTENTS_LIBDB_SREG(libdbnum,offset,size,align,rsv1,name,sregnum,     \
+			    bitmask, rsv2, rsv3)			      \
+		reg_entry libdbnum, offset, size ;
+#define CONTENTS_LIBDB_UREG(libdbnum,offset,size,align,rsv1,name,uregnum,     \
+			    bitmask, rsv2, rsv3)			      \
+		reg_entry libdbnum, offset, size ;
+#define CONTENTS_LIBDB_REGF(libdbnum, offset, size, align, rsv1, name, index, \
+			    numentries, contentsize, regname_base,	      \
+			    regfile_name, rsv2, rsv3)			      \
+		reg_entry libdbnum, offset, size ;
+
+/* A single table entry: */
+	.macro	reg_entry	libdbnum, offset, size
+	 .ifne	(__last_offset-(__last_group_offset+\offset))
+	  /* padding entry */
+	  .word	(0xFC000000+__last_offset-(__last_group_offset+\offset))
+	 .endif
+	 .word	\libdbnum				/* actual entry */
+	 .set	__last_offset, __last_group_offset+\offset+\size
+	.endm	/* reg_entry */
+
+
+/* Table entry that marks the beginning of a group (coprocessor or "extra"): */
+	.macro	reg_group	cpnum, num_entries, align
+	 .set	__last_group_offset, (__last_offset + \align- 1) & -\align
+	 .ifne	\num_entries
+	  .word	0xFD000000+(\cpnum<<16)+\num_entries
+	 .endif
+	.endm	/* reg_group */
+
+/*
+ * Register info tables.
+ */
+
+	.section .rodata, "a"
+	.globl	_xtensa_reginfo_tables
+	.globl	_xtensa_reginfo_table_size
+	.align	4
+_xtensa_reginfo_table_size:
+	.word	_xtensa_reginfo_table_end - _xtensa_reginfo_tables
+
+_xtensa_reginfo_tables:
+	.set	__last_offset, 0
+	reg_group 0xFF, XCHAL_EXTRA_SA_CONTENTS_LIBDB_NUM, XCHAL_EXTRA_SA_ALIGN
+	XCHAL_EXTRA_SA_CONTENTS_LIBDB
+	reg_group 0, XCHAL_CP0_SA_CONTENTS_LIBDB_NUM, XCHAL_CP0_SA_ALIGN
+	XCHAL_CP0_SA_CONTENTS_LIBDB
+	reg_group 1, XCHAL_CP1_SA_CONTENTS_LIBDB_NUM, XCHAL_CP1_SA_ALIGN
+	XCHAL_CP1_SA_CONTENTS_LIBDB
+	reg_group 2, XCHAL_CP2_SA_CONTENTS_LIBDB_NUM, XCHAL_CP2_SA_ALIGN
+	XCHAL_CP2_SA_CONTENTS_LIBDB
+	reg_group 3, XCHAL_CP3_SA_CONTENTS_LIBDB_NUM, XCHAL_CP3_SA_ALIGN
+	XCHAL_CP3_SA_CONTENTS_LIBDB
+	reg_group 4, XCHAL_CP4_SA_CONTENTS_LIBDB_NUM, XCHAL_CP4_SA_ALIGN
+	XCHAL_CP4_SA_CONTENTS_LIBDB
+	reg_group 5, XCHAL_CP5_SA_CONTENTS_LIBDB_NUM, XCHAL_CP5_SA_ALIGN
+	XCHAL_CP5_SA_CONTENTS_LIBDB
+	reg_group 6, XCHAL_CP6_SA_CONTENTS_LIBDB_NUM, XCHAL_CP6_SA_ALIGN
+	XCHAL_CP6_SA_CONTENTS_LIBDB
+	reg_group 7, XCHAL_CP7_SA_CONTENTS_LIBDB_NUM, XCHAL_CP7_SA_ALIGN
+	XCHAL_CP7_SA_CONTENTS_LIBDB
+	.word	0xFC000000	/* invalid register number,marks end of table*/
+_xtensa_reginfo_table_end:
+
diff --git a/arch/xtensa/kernel/entry.S b/arch/xtensa/kernel/entry.S
new file mode 100644
index 000000000000..c64a01f71de6
--- /dev/null
+++ b/arch/xtensa/kernel/entry.S
@@ -0,0 +1,1996 @@
+/*
+ * arch/xtensa/kernel/entry.S
+ *
+ * Low-level exception handling
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2004-2005 by Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ *
+ */
+
+#include <linux/linkage.h>
+#include <asm/offsets.h>
+#include <asm/processor.h>
+#include <asm/thread_info.h>
+#include <asm/uaccess.h>
+#include <asm/unistd.h>
+#include <asm/ptrace.h>
+#include <asm/current.h>
+#include <asm/pgtable.h>
+#include <asm/page.h>
+#include <asm/signal.h>
+#include <xtensa/coreasm.h>
+
+/* Unimplemented features. */
+
+#undef SIGNAL_HANDLING_IN_DOUBLE_EXCEPTION
+#undef KERNEL_STACK_OVERFLOW_CHECK
+#undef PREEMPTIBLE_KERNEL
+#undef ALLOCA_EXCEPTION_IN_IRAM
+
+/* Not well tested.
+ *
+ * - fast_coprocessor
+ */
+
+/*
+ * Macro to find first bit set in WINDOWBASE from the left + 1
+ *
+ * 100....0 -> 1
+ * 010....0 -> 2
+ * 000....1 -> WSBITS
+ */
+
+	.macro ffs_ws bit mask
+
+#if XCHAL_HAVE_NSA
+	nsau    \bit, \mask			# 32-WSBITS ... 31 (32 iff 0)
+	addi    \bit, \bit, WSBITS - 32 + 1   	# uppest bit set -> return 1
+#else
+	movi    \bit, WSBITS
+#if WSBITS > 16
+	_bltui  \mask, 0x10000, 99f
+	addi    \bit, \bit, -16
+	extui   \mask, \mask, 16, 16
+#endif
+#if WSBITS > 8
+99:	_bltui  \mask, 0x100, 99f
+	addi    \bit, \bit, -8
+	srli    \mask, \mask, 8
+#endif
+99:	_bltui  \mask, 0x10, 99f
+	addi    \bit, \bit, -4
+	srli    \mask, \mask, 4
+99:	_bltui  \mask, 0x4, 99f
+	addi    \bit, \bit, -2
+	srli    \mask, \mask, 2
+99:	_bltui  \mask, 0x2, 99f
+	addi    \bit, \bit, -1
+99:
+
+#endif
+	.endm
+
+/* ----------------- DEFAULT FIRST LEVEL EXCEPTION HANDLERS ----------------- */
+
+/*
+ * First-level exception handler for user exceptions.
+ * Save some special registers, extra states and all registers in the AR
+ * register file that were in use in the user task, and jump to the common
+ * exception code.
+ * We save SAR (used to calculate WMASK), and WB and WS (we don't have to
+ * save them for kernel exceptions).
+ *
+ * Entry condition for user_exception:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original value in depc
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ *
+ * Entry condition for _user_exception:
+ *
+ *   a0-a3 and depc have been saved to PT_AREG0...PT_AREG3 and PT_DEPC
+ *   excsave has been restored, and
+ *   stack pointer (a1) has been set.
+ *
+ * Note: _user_exception might be at an odd adress. Don't use call0..call12
+ */
+
+ENTRY(user_exception)
+
+	/* Save a2, a3, and depc, restore excsave_1 and set SP. */
+
+	xsr	a3, EXCSAVE_1
+	rsr	a0, DEPC
+	s32i	a1, a2, PT_AREG1
+	s32i	a0, a2, PT_AREG2
+	s32i	a3, a2, PT_AREG3
+	mov	a1, a2
+
+	.globl _user_exception
+_user_exception:
+
+	/* Save SAR and turn off single stepping */
+
+	movi	a2, 0
+	rsr	a3, SAR
+	wsr	a2, ICOUNTLEVEL
+	s32i	a3, a1, PT_SAR
+
+	/* Rotate ws so that the current windowbase is at bit0. */
+	/* Assume ws = xxwww1yyyy. Rotate ws right, so that a2 = yyyyxxwww1 */
+
+	rsr	a2, WINDOWBASE
+	rsr	a3, WINDOWSTART
+	ssr	a2
+	s32i	a2, a1, PT_WINDOWBASE
+	s32i	a3, a1, PT_WINDOWSTART
+	slli	a2, a3, 32-WSBITS
+	src	a2, a3, a2
+	srli	a2, a2, 32-WSBITS
+	s32i	a2, a1, PT_WMASK	# needed for restoring registers
+
+	/* Save only live registers. */
+
+	_bbsi.l	a2, 1, 1f
+	s32i	a4, a1, PT_AREG4
+	s32i	a5, a1, PT_AREG5
+	s32i	a6, a1, PT_AREG6
+	s32i	a7, a1, PT_AREG7
+	_bbsi.l	a2, 2, 1f
+	s32i	a8, a1, PT_AREG8
+	s32i	a9, a1, PT_AREG9
+	s32i	a10, a1, PT_AREG10
+	s32i	a11, a1, PT_AREG11
+	_bbsi.l	a2, 3, 1f
+	s32i	a12, a1, PT_AREG12
+	s32i	a13, a1, PT_AREG13
+	s32i	a14, a1, PT_AREG14
+	s32i	a15, a1, PT_AREG15
+	_bnei	a2, 1, 1f		# only one valid frame?
+
+	/* Only one valid frame, skip saving regs. */
+
+	j	2f
+
+	/* Save the remaining registers.
+	 * We have to save all registers up to the first '1' from
+	 * the right, except the current frame (bit 0).
+	 * Assume a2 is:  001001000110001
+	 * All regiser frames starting from the top fiel to the marked '1'
+	 * must be saved.
+	 */
+
+1:	addi	a3, a2, -1		# eliminate '1' in bit 0: yyyyxxww0
+	neg	a3, a3			# yyyyxxww0 -> YYYYXXWW1+1
+	and	a3, a3, a2		# max. only one bit is set
+
+	/* Find number of frames to save */
+
+	ffs_ws	a0, a3			# number of frames to the '1' from left
+
+	/* Store information into WMASK:
+	 * bits 0..3: xxx1 masked lower 4 bits of the rotated windowstart,
+	 * bits 4...: number of valid 4-register frames
+	 */
+
+	slli	a3, a0, 4		# number of frames to save in bits 8..4
+	extui	a2, a2, 0, 4		# mask for the first 16 registers
+	or	a2, a3, a2
+	s32i	a2, a1, PT_WMASK	# needed when we restore the reg-file
+
+	/* Save 4 registers at a time */
+
+1:	rotw	-1
+	s32i	a0, a5, PT_AREG_END - 16
+	s32i	a1, a5, PT_AREG_END - 12
+	s32i	a2, a5, PT_AREG_END - 8
+	s32i	a3, a5, PT_AREG_END - 4
+	addi	a0, a4, -1
+	addi	a1, a5, -16
+	_bnez	a0, 1b
+
+	/* WINDOWBASE still in SAR! */
+
+	rsr	a2, SAR			# original WINDOWBASE
+	movi	a3, 1
+	ssl	a2
+	sll	a3, a3
+	wsr	a3, WINDOWSTART		# set corresponding WINDOWSTART bit
+	wsr	a2, WINDOWBASE		# and WINDOWSTART
+	rsync
+
+	/* We are back to the original stack pointer (a1) */
+
+2:
+#if XCHAL_EXTRA_SA_SIZE
+
+	/* For user exceptions, save the extra state into the user's TCB.
+	 * Note: We must assume that xchal_extra_store_funcbody destroys a2..a15
+	 */
+
+	GET_CURRENT(a2,a1)
+	addi	a2, a2, THREAD_CP_SAVE
+	xchal_extra_store_funcbody
+#endif
+
+	/* Now, jump to the common exception handler. */
+
+	j	common_exception
+
+
+/*
+ * First-level exit handler for kernel exceptions
+ * Save special registers and the live window frame.
+ * Note: Even though we changes the stack pointer, we don't have to do a
+ *	 MOVSP here, as we do that when we return from the exception.
+ *	 (See comment in the kernel exception exit code)
+ *
+ * Entry condition for kernel_exception:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ *
+ * Entry condition for _kernel_exception:
+ *
+ *   a0-a3 and depc have been saved to PT_AREG0...PT_AREG3 and PT_DEPC
+ *   excsave has been restored, and
+ *   stack pointer (a1) has been set.
+ *
+ * Note: _kernel_exception might be at an odd adress. Don't use call0..call12
+ */
+
+ENTRY(kernel_exception)
+
+	/* Save a0, a2, a3, DEPC and set SP. */
+
+	xsr	a3, EXCSAVE_1		# restore a3, excsave_1
+	rsr	a0, DEPC		# get a2
+	s32i	a1, a2, PT_AREG1
+	s32i	a0, a2, PT_AREG2
+	s32i	a3, a2, PT_AREG3
+	mov	a1, a2
+
+	.globl _kernel_exception
+_kernel_exception:
+
+	/* Save SAR and turn off single stepping */
+
+	movi	a2, 0
+	rsr	a3, SAR
+	wsr	a2, ICOUNTLEVEL
+	s32i	a3, a1, PT_SAR
+
+	/* Rotate ws so that the current windowbase is at bit0. */
+	/* Assume ws = xxwww1yyyy. Rotate ws right, so that a2 = yyyyxxwww1 */
+
+	rsr	a2, WINDOWBASE		# don't need to save these, we only
+	rsr	a3, WINDOWSTART		# need shifted windowstart: windowmask
+	ssr	a2
+	slli	a2, a3, 32-WSBITS
+	src	a2, a3, a2
+	srli	a2, a2, 32-WSBITS
+	s32i	a2, a1, PT_WMASK	# needed for kernel_exception_exit
+
+	/* Save only the live window-frame */
+
+	_bbsi.l	a2, 1, 1f
+	s32i	a4, a1, PT_AREG4
+	s32i	a5, a1, PT_AREG5
+	s32i	a6, a1, PT_AREG6
+	s32i	a7, a1, PT_AREG7
+	_bbsi.l	a2, 2, 1f
+	s32i	a8, a1, PT_AREG8
+	s32i	a9, a1, PT_AREG9
+	s32i	a10, a1, PT_AREG10
+	s32i	a11, a1, PT_AREG11
+	_bbsi.l	a2, 3, 1f
+	s32i	a12, a1, PT_AREG12
+	s32i	a13, a1, PT_AREG13
+	s32i	a14, a1, PT_AREG14
+	s32i	a15, a1, PT_AREG15
+
+1:
+
+#ifdef KERNEL_STACK_OVERFLOW_CHECK
+
+	/*  Stack overflow check, for debugging  */
+	extui	a2, a1, TASK_SIZE_BITS,XX
+	movi	a3, SIZE??
+	_bge	a2, a3, out_of_stack_panic
+
+#endif
+
+/*
+ * This is the common exception handler.
+ * We get here from the user exception handler or simply by falling through
+ * from the kernel exception handler.
+ * Save the remaining special registers, switch to kernel mode, and jump
+ * to the second-level exception handler.
+ *
+ */
+
+common_exception:
+
+	/* Save EXCVADDR, DEBUGCAUSE, and PC, and clear LCOUNT */
+
+	rsr	a2, DEBUGCAUSE
+	rsr	a3, EPC_1
+	s32i	a2, a1, PT_DEBUGCAUSE
+	s32i	a3, a1, PT_PC
+
+	rsr	a3, EXCVADDR
+	movi	a2, 0
+	s32i	a3, a1, PT_EXCVADDR
+	xsr	a2, LCOUNT
+	s32i	a2, a1, PT_LCOUNT
+
+	/* It is now save to restore the EXC_TABLE_FIXUP variable. */
+
+	rsr	a0, EXCCAUSE
+	movi	a3, 0
+	rsr	a2, EXCSAVE_1
+	s32i	a0, a1, PT_EXCCAUSE
+	s32i	a3, a2, EXC_TABLE_FIXUP
+
+	/* All unrecoverable states are saved on stack, now, and a1 is valid,
+	 * so we can allow exceptions and interrupts (*) again.
+	 * Set PS(EXCM = 0, UM = 0, RING = 0, OWB = 0, WOE = 1, INTLEVEL = X)
+	 *
+	 * (*) We only allow interrupts if PS.INTLEVEL was not set to 1 before
+	 *     (interrupts disabled) and if this exception is not an interrupt.
+	 */
+
+	rsr	a3, PS
+	addi	a0, a0, -4
+	movi	a2, 1
+	extui	a3, a3, 0, 1		# a3 = PS.INTLEVEL[0]
+	moveqz	a3, a2, a0		# a3 = 1 iff interrupt exception
+	movi	a2, PS_WOE_MASK
+	or	a3, a3, a2
+	rsr	a0, EXCCAUSE
+	xsr	a3, PS
+
+	s32i	a3, a1, PT_PS		# save ps
+
+	/* Save LBEG, LEND */
+
+	rsr	a2, LBEG
+	rsr	a3, LEND
+	s32i	a2, a1, PT_LBEG
+	s32i	a3, a1, PT_LEND
+
+	/* Go to second-level dispatcher. Set up parameters to pass to the
+	 * exception handler and call the exception handler.
+	 */
+
+	movi	a4, exc_table
+	mov	a6, a1			# pass stack frame
+	mov	a7, a0			# pass EXCCAUSE
+	addx4	a4, a0, a4
+	l32i	a4, a4, EXC_TABLE_DEFAULT		# load handler
+
+	/* Call the second-level handler */
+
+	callx4	a4
+
+	/* Jump here for exception exit */
+
+common_exception_return:
+
+	/* Jump if we are returning from kernel exceptions. */
+
+1:	l32i	a3, a1, PT_PS
+	_bbsi.l	a3, PS_UM_SHIFT, 2f
+	j	kernel_exception_exit
+
+	/* Specific to a user exception exit:
+	 * We need to check some flags for signal handling and rescheduling,
+	 * and have to restore WB and WS, extra states, and all registers
+	 * in the register file that were in use in the user task.
+	 */
+
+2:	wsr	a3, PS		/* disable interrupts */
+
+	/* Check for signals (keep interrupts disabled while we read TI_FLAGS)
+	 * Note: PS.INTLEVEL = 0, PS.EXCM = 1
+	 */
+
+	GET_THREAD_INFO(a2,a1)
+	l32i	a4, a2, TI_FLAGS
+
+	/* Enable interrupts again.
+	 * Note: When we get here, we certainly have handled any interrupts.
+	 *       (Hint: There is only one user exception frame on stack)
+	 */
+
+	movi	a3, PS_WOE_MASK
+
+	_bbsi.l	a4, TIF_NEED_RESCHED, 3f
+	_bbci.l	a4, TIF_SIGPENDING, 4f
+
+#ifndef SIGNAL_HANDLING_IN_DOUBLE_EXCEPTION
+	l32i	a4, a1, PT_DEPC
+	bgeui	a4, VALID_DOUBLE_EXCEPTION_ADDRESS, 4f
+#endif
+
+	/* Reenable interrupts and call do_signal() */
+
+	wsr	a3, PS
+	movi	a4, do_signal	# int do_signal(struct pt_regs*, sigset_t*)
+	mov	a6, a1
+	movi	a7, 0
+	callx4	a4
+	j	1b
+
+3:	/* Reenable interrupts and reschedule */
+
+	wsr	a3, PS
+	movi	a4, schedule	# void schedule (void)
+	callx4	a4
+	j	1b
+
+	/* Restore the state of the task and return from the exception. */
+
+
+	/* If we are returning from a user exception, and the process
+	 * to run next has PT_SINGLESTEP set, we want to setup
+	 * ICOUNT and ICOUNTLEVEL to step one instruction.
+	 * PT_SINGLESTEP is set by sys_ptrace (ptrace.c)
+	 */
+
+4:	/* a2 holds GET_CURRENT(a2,a1)  */
+
+	l32i	a3, a2, TI_TASK
+	l32i	a3, a3, TASK_PTRACE
+	bbci.l	a3, PT_SINGLESTEP_BIT, 1f # jump if single-step flag is not set
+
+	movi	a3, -2			# PT_SINGLESTEP flag is set,
+	movi	a4, 1			# icountlevel of 1 means it won't
+	wsr	a3, ICOUNT		# start counting until after rfe
+	wsr	a4, ICOUNTLEVEL		# so setup icount & icountlevel.
+	isync
+
+1:
+
+#if XCHAL_EXTRA_SA_SIZE
+
+	/* For user exceptions, restore the extra state from the user's TCB. */
+
+	/* Note: a2 still contains GET_CURRENT(a2,a1) */
+	addi	a2, a2, THREAD_CP_SAVE
+	xchal_extra_load_funcbody
+
+	/* We must assume that xchal_extra_store_funcbody destroys
+	 * registers a2..a15.  FIXME, this list can eventually be
+	 * reduced once real register requirements of the macro are
+	 * finalized. */
+
+#endif /* XCHAL_EXTRA_SA_SIZE */
+
+
+	/* Switch to the user thread WINDOWBASE. Save SP temporarily in DEPC */
+
+	l32i	a2, a1, PT_WINDOWBASE
+	l32i	a3, a1, PT_WINDOWSTART
+	wsr	a1, DEPC		# use DEPC as temp storage
+	wsr	a3, WINDOWSTART		# restore WINDOWSTART
+	ssr	a2			# preserve user's WB in the SAR
+	wsr	a2, WINDOWBASE		# switch to user's saved WB
+	rsync
+	rsr	a1, DEPC		# restore stack pointer
+	l32i	a2, a1, PT_WMASK	# register frames saved (in bits 4...9)
+	rotw	-1			# we restore a4..a7
+	_bltui	a6, 16, 1f		# only have to restore current window?
+
+	/* The working registers are a0 and a3.  We are restoring to
+	 * a4..a7.  Be careful not to destroy what we have just restored.
+	 * Note: wmask has the format YYYYM:
+	 *       Y: number of registers saved in groups of 4
+	 *       M: 4 bit mask of first 16 registers
+	 */
+
+	mov	a2, a6
+	mov	a3, a5
+
+2:	rotw	-1			# a0..a3 become a4..a7
+	addi	a3, a7, -4*4		# next iteration
+	addi	a2, a6, -16		# decrementing Y in WMASK
+	l32i	a4, a3, PT_AREG_END + 0
+	l32i	a5, a3, PT_AREG_END + 4
+	l32i	a6, a3, PT_AREG_END + 8
+	l32i	a7, a3, PT_AREG_END + 12
+	_bgeui	a2, 16, 2b
+
+	/* Clear unrestored registers (don't leak anything to user-land */
+
+1:	rsr	a0, WINDOWBASE
+	rsr	a3, SAR
+	sub	a3, a0, a3
+	beqz	a3, 2f
+	extui	a3, a3, 0, WBBITS
+
+1:	rotw	-1
+	addi	a3, a7, -1
+	movi	a4, 0
+	movi	a5, 0
+	movi	a6, 0
+	movi	a7, 0
+	bgei	a3, 1, 1b
+
+	/* We are back were we were when we started.
+	 * Note: a2 still contains WMASK (if we've returned to the original
+	 *	 frame where we had loaded a2), or at least the lower 4 bits
+	 *	 (if we have restored WSBITS-1 frames).
+	 */
+
+2:	j	common_exception_exit
+
+	/* This is the kernel exception exit.
+	 * We avoided to do a MOVSP when we entered the exception, but we
+	 * have to do it here.
+	 */
+
+kernel_exception_exit:
+
+	/* Disable interrupts (a3 holds PT_PS) */
+
+	wsr	a3, PS
+
+#ifdef PREEMPTIBLE_KERNEL
+
+#ifdef CONFIG_PREEMPT
+
+	/*
+	 * Note: We've just returned from a call4, so we have
+	 * at least 4 addt'l regs.
+	 */
+
+	/* Check current_thread_info->preempt_count */
+
+	GET_THREAD_INFO(a2)
+	l32i	a3, a2, TI_PREEMPT
+	bnez	a3, 1f
+
+	l32i	a2, a2, TI_FLAGS
+
+1:
+
+#endif
+
+#endif
+
+	/* Check if we have to do a movsp.
+	 *
+	 * We only have to do a movsp if the previous window-frame has
+	 * been spilled to the *temporary* exception stack instead of the
+	 * task's stack. This is the case if the corresponding bit in
+	 * WINDOWSTART for the previous window-frame was set before
+	 * (not spilled) but is zero now (spilled).
+	 * If this bit is zero, all other bits except the one for the
+	 * current window frame are also zero. So, we can use a simple test:
+	 * 'and' WINDOWSTART and WINDOWSTART-1:
+	 *
+	 *  (XXXXXX1[0]* - 1) AND XXXXXX1[0]* = XXXXXX0[0]*
+	 *
+	 * The result is zero only if one bit was set.
+	 *
+	 * (Note: We might have gone through several task switches before
+	 *        we come back to the current task, so WINDOWBASE might be
+	 *        different from the time the exception occurred.)
+	 */
+
+	/* Test WINDOWSTART before and after the exception.
+	 * We actually have WMASK, so we only have to test if it is 1 or not.
+	 */
+
+	l32i	a2, a1, PT_WMASK
+	_beqi	a2, 1, common_exception_exit	# Spilled before exception,jump
+
+	/* Test WINDOWSTART now. If spilled, do the movsp */
+
+	rsr     a3, WINDOWSTART
+	addi	a0, a3, -1
+	and     a3, a3, a0
+	_bnez	a3, common_exception_exit
+
+	/* Do a movsp (we returned from a call4, so we have at least a0..a7) */
+
+	addi    a0, a1, -16
+	l32i    a3, a0, 0
+	l32i    a4, a0, 4
+	s32i    a3, a1, PT_SIZE+0
+	s32i    a4, a1, PT_SIZE+4
+	l32i    a3, a0, 8
+	l32i    a4, a0, 12
+	s32i    a3, a1, PT_SIZE+8
+	s32i    a4, a1, PT_SIZE+12
+
+	/* Common exception exit.
+	 * We restore the special register and the current window frame, and
+	 * return from the exception.
+	 *
+	 * Note: We expect a2 to hold PT_WMASK
+	 */
+
+common_exception_exit:
+
+	_bbsi.l	a2, 1, 1f
+	l32i	a4,  a1, PT_AREG4
+	l32i	a5,  a1, PT_AREG5
+	l32i	a6,  a1, PT_AREG6
+	l32i	a7,  a1, PT_AREG7
+	_bbsi.l	a2, 2, 1f
+	l32i	a8,  a1, PT_AREG8
+	l32i	a9,  a1, PT_AREG9
+	l32i	a10, a1, PT_AREG10
+	l32i	a11, a1, PT_AREG11
+	_bbsi.l	a2, 3, 1f
+	l32i	a12, a1, PT_AREG12
+	l32i	a13, a1, PT_AREG13
+	l32i	a14, a1, PT_AREG14
+	l32i	a15, a1, PT_AREG15
+
+	/* Restore PC, SAR */
+
+1:	l32i	a2, a1, PT_PC
+	l32i	a3, a1, PT_SAR
+	wsr	a2, EPC_1
+	wsr	a3, SAR
+
+	/* Restore LBEG, LEND, LCOUNT */
+
+	l32i	a2, a1, PT_LBEG
+	l32i	a3, a1, PT_LEND
+	wsr	a2, LBEG
+	l32i	a2, a1, PT_LCOUNT
+	wsr	a3, LEND
+	wsr	a2, LCOUNT
+
+	/* Check if it was double exception. */
+
+	l32i	a0, a1, PT_DEPC
+	l32i	a3, a1, PT_AREG3
+	l32i	a2, a1, PT_AREG2
+	_bgeui	a0, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f
+
+	/* Restore a0...a3 and return */
+
+	l32i	a0, a1, PT_AREG0
+	l32i	a1, a1, PT_AREG1
+	rfe
+
+1:	wsr	a0, DEPC
+	l32i	a0, a1, PT_AREG0
+	l32i	a1, a1, PT_AREG1
+	rfde
+
+/*
+ * Debug exception handler.
+ *
+ * Currently, we don't support KGDB, so only user application can be debugged.
+ *
+ * When we get here,  a0 is trashed and saved to excsave[debuglevel]
+ */
+
+ENTRY(debug_exception)
+
+	rsr	a0, EPS + XCHAL_DEBUGLEVEL
+	bbsi.l	a0, PS_EXCM_SHIFT, 1f	# exception mode
+
+	/* Set EPC_1 and EXCCAUSE */
+
+	wsr	a2, DEPC		# save a2 temporarily
+	rsr	a2, EPC + XCHAL_DEBUGLEVEL
+	wsr	a2, EPC_1
+
+	movi	a2, EXCCAUSE_MAPPED_DEBUG
+	wsr	a2, EXCCAUSE
+
+	/* Restore PS to the value before the debug exc but with PS.EXCM set.*/
+
+	movi	a2, 1 << PS_EXCM_SHIFT
+	or	a2, a0, a2
+	movi	a0, debug_exception	# restore a3, debug jump vector
+	wsr	a2, PS
+	xsr	a0, EXCSAVE + XCHAL_DEBUGLEVEL
+
+	/* Switch to kernel/user stack, restore jump vector, and save a0 */
+
+	bbsi.l	a2, PS_UM_SHIFT, 2f	# jump if user mode
+
+	addi	a2, a1, -16-PT_SIZE	# assume kernel stack
+	s32i	a0, a2, PT_AREG0
+	movi	a0, 0
+	s32i	a1, a2, PT_AREG1
+	s32i	a0, a2, PT_DEPC		# mark it as a regular exception
+	xsr	a0, DEPC
+	s32i	a3, a2, PT_AREG3
+	s32i	a0, a2, PT_AREG2
+	mov	a1, a2
+	j	_kernel_exception
+
+2:	rsr	a2, EXCSAVE_1
+	l32i	a2, a2, EXC_TABLE_KSTK	# load kernel stack pointer
+	s32i	a0, a2, PT_AREG0
+	movi	a0, 0
+	s32i	a1, a2, PT_AREG1
+	s32i	a0, a2, PT_DEPC
+	xsr	a0, DEPC
+	s32i	a3, a2, PT_AREG3
+	s32i	a0, a2, PT_AREG2
+	mov	a1, a2
+	j	_user_exception
+
+	/* Debug exception while in exception mode. */
+1:	j	1b	// FIXME!!
+
+
+/*
+ * We get here in case of an unrecoverable exception.
+ * The only thing we can do is to be nice and print a panic message.
+ * We only produce a single stack frame for panic, so ???
+ *
+ *
+ * Entry conditions:
+ *
+ *   - a0 contains the caller address; original value saved in excsave1.
+ *   - the original a0 contains a valid return address (backtrace) or 0.
+ *   - a2 contains a valid stackpointer
+ *
+ * Notes:
+ *
+ *   - If the stack pointer could be invalid, the caller has to setup a
+ *     dummy stack pointer (e.g. the stack of the init_task)
+ *
+ *   - If the return address could be invalid, the caller has to set it
+ *     to 0, so the backtrace would stop.
+ *
+ */
+	.align 4
+unrecoverable_text:
+	.ascii "Unrecoverable error in exception handler\0"
+
+ENTRY(unrecoverable_exception)
+
+	movi	a0, 1
+	movi	a1, 0
+
+	wsr	a0, WINDOWSTART
+	wsr	a1, WINDOWBASE
+	rsync
+
+	movi	a1, PS_WOE_MASK | 1
+	wsr	a1, PS
+	rsync
+
+	movi	a1, init_task
+	movi	a0, 0
+	addi	a1, a1, PT_REGS_OFFSET
+
+	movi	a4, panic
+	movi	a6, unrecoverable_text
+
+	callx4	a4
+
+1:	j	1b
+
+
+/* -------------------------- FAST EXCEPTION HANDLERS ----------------------- */
+
+/*
+ * Fast-handler for alloca exceptions
+ *
+ *  The ALLOCA handler is entered when user code executes the MOVSP
+ *  instruction and the caller's frame is not in the register file.
+ *  In this case, the caller frame's a0..a3 are on the stack just
+ *  below sp (a1), and this handler moves them.
+ *
+ *  For "MOVSP <ar>,<as>" without destination register a1, this routine
+ *  simply moves the value from <as> to <ar> without moving the save area.
+ *
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ */
+
+#if XCHAL_HAVE_BE
+#define _EXTUI_MOVSP_SRC(ar)	extui ar, ar, 4, 4
+#define _EXTUI_MOVSP_DST(ar)	extui ar, ar, 0, 4
+#else
+#define _EXTUI_MOVSP_SRC(ar)	extui ar, ar, 0, 4
+#define _EXTUI_MOVSP_DST(ar)	extui ar, ar, 4, 4
+#endif
+
+ENTRY(fast_alloca)
+
+	/* We shouldn't be in a double exception. */
+
+	l32i	a0, a2, PT_DEPC
+	_bgeui	a0, VALID_DOUBLE_EXCEPTION_ADDRESS, .Lunhandled_double
+
+	rsr	a0, DEPC		# get a2
+	s32i	a4, a2, PT_AREG4	# save a4 and
+	s32i	a0, a2, PT_AREG2	# a2 to stack
+
+	/* Exit critical section. */
+
+	movi	a0, 0
+	s32i	a0, a3, EXC_TABLE_FIXUP
+
+	/* Restore a3, excsave_1 */
+
+	xsr	a3, EXCSAVE_1		# make sure excsave_1 is valid for dbl.
+	rsr	a4, EPC_1		# get exception address
+	s32i	a3, a2, PT_AREG3	# save a3 to stack
+
+#ifdef ALLOCA_EXCEPTION_IN_IRAM
+#error	iram not supported
+#else
+	/* Note: l8ui not allowed in IRAM/IROM!! */
+	l8ui	a0, a4, 1		# read as(src) from MOVSP instruction
+#endif
+	movi	a3, .Lmovsp_src
+	_EXTUI_MOVSP_SRC(a0)		# extract source register number
+	addx8	a3, a0, a3
+	jx	a3
+
+.Lunhandled_double:
+	wsr	a0, EXCSAVE_1
+	movi	a0, unrecoverable_exception
+	callx0	a0
+
+	.align 8
+.Lmovsp_src:
+	l32i	a3, a2, PT_AREG0;	_j 1f;	.align 8
+	mov	a3, a1;			_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG2;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG3;	_j 1f;	.align 8
+	l32i	a3, a2, PT_AREG4;	_j 1f;	.align 8
+	mov	a3, a5;			_j 1f;	.align 8
+	mov	a3, a6;			_j 1f;	.align 8
+	mov	a3, a7;			_j 1f;	.align 8
+	mov	a3, a8;			_j 1f;	.align 8
+	mov	a3, a9;			_j 1f;	.align 8
+	mov	a3, a10;		_j 1f;	.align 8
+	mov	a3, a11;		_j 1f;	.align 8
+	mov	a3, a12;		_j 1f;	.align 8
+	mov	a3, a13;		_j 1f;	.align 8
+	mov	a3, a14;		_j 1f;	.align 8
+	mov	a3, a15;		_j 1f;	.align 8
+
+1:
+
+#ifdef ALLOCA_EXCEPTION_IN_IRAM
+#error	iram not supported
+#else
+	l8ui	a0, a4, 0		# read ar(dst) from MOVSP instruction
+#endif
+	addi	a4, a4, 3		# step over movsp
+	_EXTUI_MOVSP_DST(a0)		# extract destination register
+	wsr	a4, EPC_1		# save new epc_1
+
+	_bnei	a0, 1, 1f		# no 'movsp a1, ax': jump
+
+        /* Move the save area. This implies the use of the L32E
+	 * and S32E instructions, because this move must be done with
+	 * the user's PS.RING privilege levels, not with ring 0
+	 * (kernel's) privileges currently active with PS.EXCM
+	 * set. Note that we have stil registered a fixup routine with the
+	 * double exception vector in case a double exception occurs.
+	 */
+
+	/* a0,a4:avail a1:old user stack a2:exc. stack a3:new user stack. */
+
+	l32e	a0, a1, -16
+	l32e	a4, a1, -12
+	s32e	a0, a3, -16
+	s32e	a4, a3, -12
+	l32e	a0, a1, -8
+	l32e	a4, a1, -4
+	s32e	a0, a3, -8
+	s32e	a4, a3, -4
+
+	/* Restore stack-pointer and all the other saved registers. */
+
+	mov	a1, a3
+
+	l32i	a4, a2, PT_AREG4
+	l32i	a3, a2, PT_AREG3
+	l32i	a0, a2, PT_AREG0
+	l32i	a2, a2, PT_AREG2
+	rfe
+
+	/*  MOVSP <at>,<as>  was invoked with <at> != a1.
+	 *  Because the stack pointer is not being modified,
+	 *  we should be able to just modify the pointer
+	 *  without moving any save area.
+	 *  The processor only traps these occurrences if the
+	 *  caller window isn't live, so unfortunately we can't
+	 *  use this as an alternate trap mechanism.
+	 *  So we just do the move.  This requires that we
+	 *  resolve the destination register, not just the source,
+	 *  so there's some extra work.
+	 *  (PERHAPS NOT REALLY NEEDED, BUT CLEANER...)
+	 */
+
+	/* a0 dst-reg, a1 user-stack, a2 stack, a3 value of src reg. */
+
+1:	movi	a4, .Lmovsp_dst
+	addx8	a4, a0, a4
+	jx	a4
+
+	.align 8
+.Lmovsp_dst:
+	s32i	a3, a2, PT_AREG0;	_j 1f;	.align 8
+	mov	a1, a3;			_j 1f;	.align 8
+	s32i	a3, a2, PT_AREG2;	_j 1f;	.align 8
+	s32i	a3, a2, PT_AREG3;	_j 1f;	.align 8
+	s32i	a3, a2, PT_AREG4;	_j 1f;	.align 8
+	mov	a5, a3;			_j 1f;	.align 8
+	mov	a6, a3;			_j 1f;	.align 8
+	mov	a7, a3;			_j 1f;	.align 8
+	mov	a8, a3;			_j 1f;	.align 8
+	mov	a9, a3;			_j 1f;	.align 8
+	mov	a10, a3;		_j 1f;	.align 8
+	mov	a11, a3;		_j 1f;	.align 8
+	mov	a12, a3;		_j 1f;	.align 8
+	mov	a13, a3;		_j 1f;	.align 8
+	mov	a14, a3;		_j 1f;	.align 8
+	mov	a15, a3;		_j 1f;	.align 8
+
+1:	l32i	a4, a2, PT_AREG4
+	l32i	a3, a2, PT_AREG3
+	l32i	a0, a2, PT_AREG0
+	l32i	a2, a2, PT_AREG2
+	rfe
+
+
+/*
+ * fast system calls.
+ *
+ * WARNING:  The kernel doesn't save the entire user context before
+ * handling a fast system call.  These functions are small and short,
+ * usually offering some functionality not available to user tasks.
+ *
+ * BE CAREFUL TO PRESERVE THE USER'S CONTEXT.
+ *
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ */
+
+ENTRY(fast_syscall_kernel)
+
+	/* Skip syscall. */
+
+	rsr	a0, EPC_1
+	addi	a0, a0, 3
+	wsr	a0, EPC_1
+
+	l32i	a0, a2, PT_DEPC
+	bgeui	a0, VALID_DOUBLE_EXCEPTION_ADDRESS, fast_syscall_unrecoverable
+
+	rsr	a0, DEPC			# get syscall-nr
+	_beqz	a0, fast_syscall_spill_registers
+
+	addi	a0, a0, -__NR_sysxtensa
+	_beqz	a0, fast_syscall_sysxtensa
+
+	j	kernel_exception
+
+
+ENTRY(fast_syscall_user)
+
+	/* Skip syscall. */
+
+	rsr	a0, EPC_1
+	addi	a0, a0, 3
+	wsr	a0, EPC_1
+
+	l32i	a0, a2, PT_DEPC
+	bgeui	a0, VALID_DOUBLE_EXCEPTION_ADDRESS, fast_syscall_unrecoverable
+
+	rsr	a0, DEPC			# get syscall-nr
+	_beqz	a0, fast_syscall_spill_registers
+
+	addi	a0, a0, -__NR_sysxtensa
+	_beqz	a0, fast_syscall_sysxtensa
+
+	j	user_exception
+
+ENTRY(fast_syscall_unrecoverable)
+
+        /* Restore all states. */
+
+        l32i    a0, a2, PT_AREG0        # restore a0
+        xsr     a2, DEPC                # restore a2, depc
+        rsr     a3, EXCSAVE_1
+
+        wsr     a0, EXCSAVE_1
+        movi    a0, unrecoverable_exception
+        callx0  a0
+
+
+
+/*
+ * sysxtensa syscall handler
+ *
+ * int sysxtensa (XTENSA_ATOMIC_SET, ptr, val, unused);
+ * int sysxtensa (XTENSA_ATOMIC_ADD, ptr, val, unused);
+ * int sysxtensa (XTENSA_ATOMIC_EXG_ADD, ptr, val, unused);
+ * int sysxtensa (XTENSA_ATOMIC_CMP_SWP, ptr, oldval, newval);
+ * a2                    a6              a3    a4      a5
+ *
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ *
+ * Note: we don't have to save a2; a2 holds the return value
+ *
+ * We use the two macros TRY and CATCH:
+ *
+ * TRY	 adds an entry to the __ex_table fixup table for the immediately
+ *	 following instruction.
+ *
+ * CATCH catches any exception that occurred at one of the preceeding TRY
+ *       statements and continues from there
+ *
+ * Usage TRY	l32i	a0, a1, 0
+ *		<other code>
+ *	 done:	rfe
+ *	 CATCH	<set return code>
+ *		j done
+ */
+
+#define TRY								\
+	.section __ex_table, "a";					\
+	.word	66f, 67f;						\
+	.text;								\
+66:
+
+#define CATCH								\
+67:
+
+ENTRY(fast_syscall_sysxtensa)
+
+	_beqz	a6, 1f
+	_blti	a6, SYSXTENSA_COUNT, 2f
+
+1:	j	user_exception
+
+2:	xsr	a3, EXCSAVE_1		# restore a3, excsave1
+	s32i	a7, a2, PT_AREG7
+
+	movi	a7, 4			# sizeof(unsigned int)
+	verify_area a3, a7, a0, a2, .Leac
+
+	_beqi	a6, SYSXTENSA_ATOMIC_SET, .Lset
+	_beqi	a6, SYSXTENSA_ATOMIC_EXG_ADD, .Lexg
+	_beqi	a6, SYSXTENSA_ATOMIC_ADD, .Ladd
+
+	/* Fall through for SYSXTENSA_ATOMIC_CMP_SWP */
+
+.Lswp:	/* Atomic compare and swap */
+
+TRY	l32i	a7, a3, 0		# read old value
+	bne	a7, a4, 1f		# same as old value? jump
+	s32i	a5, a3, 0		# different, modify value
+	movi	a7, 1			# and return 1
+	j	.Lret
+
+1:	movi	a7, 0			# same values: return 0
+	j	.Lret
+
+.Ladd:	/* Atomic add */
+.Lexg:	/* Atomic (exchange) add */
+
+TRY	l32i	a7, a3, 0
+	add	a4, a4, a7
+	s32i	a4, a3, 0
+	j	.Lret
+
+.Lset:	/* Atomic set */
+
+TRY	l32i	a7, a3, 0		# read old value as return value
+	s32i	a4, a3, 0		# write new value
+
+.Lret:	mov	a0, a2
+	mov	a2, a7
+	l32i	a7, a0, PT_AREG7
+	l32i	a3, a0, PT_AREG3
+	l32i	a0, a0, PT_AREG0
+	rfe
+
+CATCH
+.Leac:	movi	a7, -EFAULT
+	j	.Lret
+
+
+
+/* fast_syscall_spill_registers.
+ *
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ * Note: We assume the stack pointer is EXC_TABLE_KSTK in the fixup handler.
+ * Note: We don't need to save a2 in depc (return value)
+ */
+
+ENTRY(fast_syscall_spill_registers)
+
+	/* Register a FIXUP handler (pass current wb as a parameter) */
+
+	movi	a0, fast_syscall_spill_registers_fixup
+	s32i	a0, a3, EXC_TABLE_FIXUP
+	rsr	a0, WINDOWBASE
+	s32i	a0, a3, EXC_TABLE_PARAM
+
+	/* Save a3 and SAR on stack. */
+
+	rsr	a0, SAR
+	xsr	a3, EXCSAVE_1		# restore a3 and excsave_1
+	s32i	a0, a2, PT_AREG4	# store SAR to PT_AREG4
+	s32i	a3, a2, PT_AREG3
+
+	/* The spill routine might clobber a7, a11, and a15. */
+
+	s32i	a7, a2, PT_AREG5
+	s32i	a11, a2, PT_AREG6
+	s32i	a15, a2, PT_AREG7
+
+	call0	_spill_registers	# destroys a3, DEPC, and SAR
+
+	/* Advance PC, restore registers and SAR, and return from exception. */
+
+	l32i	a3, a2, PT_AREG4
+	l32i	a0, a2, PT_AREG0
+	wsr	a3, SAR
+	l32i	a3, a2, PT_AREG3
+
+	/* Restore clobbered registers. */
+
+	l32i	a7, a2, PT_AREG5
+	l32i	a11, a2, PT_AREG6
+	l32i	a15, a2, PT_AREG7
+
+	movi	a2, 0
+	rfe
+
+/* Fixup handler.
+ *
+ * We get here if the spill routine causes an exception, e.g. tlb miss.
+ * We basically restore WINDOWBASE and WINDOWSTART to the condition when
+ * we entered the spill routine and jump to the user exception handler.
+ *
+ * a0: value of depc, original value in depc
+ * a2: trashed, original value in EXC_TABLE_DOUBLE_SAVE
+ * a3: exctable, original value in excsave1
+ */
+
+fast_syscall_spill_registers_fixup:
+
+	rsr	a2, WINDOWBASE	# get current windowbase (a2 is saved)
+	xsr	a0, DEPC	# restore depc and a0
+	ssl	a2		# set shift (32 - WB)
+
+	/* We need to make sure the current registers (a0-a3) are preserved.
+	 * To do this, we simply set the bit for the current window frame
+	 * in WS, so that the exception handlers save them to the task stack.
+	 */
+
+	rsr	a3, EXCSAVE_1	# get spill-mask
+	slli	a2, a3, 1	# shift left by one
+
+	slli	a3, a2, 32-WSBITS
+	src	a2, a2, a3	# a1 = xxwww1yyxxxwww1yy......
+	wsr	a2, WINDOWSTART	# set corrected windowstart
+
+	movi	a3, exc_table
+	l32i	a2, a3, EXC_TABLE_DOUBLE_SAVE	# restore a2
+	l32i	a3, a3, EXC_TABLE_PARAM	# original WB (in user task)
+
+	/* Return to the original (user task) WINDOWBASE.
+	 * We leave the following frame behind:
+	 * a0, a1, a2	same
+	 * a3:		trashed (saved in excsave_1)
+	 * depc:	depc (we have to return to that address)
+	 * excsave_1:	a3
+	 */
+
+	wsr	a3, WINDOWBASE
+	rsync
+
+	/* We are now in the original frame when we entered _spill_registers:
+	 *  a0: return address
+	 *  a1: used, stack pointer
+	 *  a2: kernel stack pointer
+	 *  a3: available, saved in EXCSAVE_1
+	 *  depc: exception address
+	 *  excsave: a3
+	 * Note: This frame might be the same as above.
+	 */
+
+#ifdef SIGNAL_HANDLING_IN_DOUBLE_EXCEPTION
+	/* Restore registers we precautiously saved.
+	 * We have the value of the 'right' a3
+	 */
+
+	l32i	a7, a2, PT_AREG5
+	l32i	a11, a2, PT_AREG6
+	l32i	a15, a2, PT_AREG7
+#endif
+
+	/* Setup stack pointer. */
+
+	addi	a2, a2, -PT_USER_SIZE
+	s32i	a0, a2, PT_AREG0
+
+	/* Make sure we return to this fixup handler. */
+
+	movi	a3, fast_syscall_spill_registers_fixup_return
+	s32i	a3, a2, PT_DEPC		# setup depc
+
+	/* Jump to the exception handler. */
+
+	movi	a3, exc_table
+	rsr	a0, EXCCAUSE
+        addx4   a0, a0, a3              	# find entry in table
+        l32i    a0, a0, EXC_TABLE_FAST_USER     # load handler
+        jx      a0
+
+fast_syscall_spill_registers_fixup_return:
+
+	/* When we return here, all registers have been restored (a2: DEPC) */
+
+	wsr	a2, DEPC		# exception address
+
+	/* Restore fixup handler. */
+
+	xsr	a3, EXCSAVE_1
+	movi	a2, fast_syscall_spill_registers_fixup
+	s32i	a2, a3, EXC_TABLE_FIXUP
+	rsr	a2, WINDOWBASE
+	s32i	a2, a3, EXC_TABLE_PARAM
+	l32i	a2, a3, EXC_TABLE_KSTK
+
+#ifdef SIGNAL_HANDLING_IN_DOUBLE_EXCEPTION
+	/* Save registers again that might be clobbered. */
+
+	s32i	a7, a2, PT_AREG5
+	s32i	a11, a2, PT_AREG6
+	s32i	a15, a2, PT_AREG7
+#endif
+
+	/* Load WB at the time the exception occurred. */
+
+	rsr	a3, SAR			# WB is still in SAR
+	neg	a3, a3
+	wsr	a3, WINDOWBASE
+	rsync
+
+	/* Restore a3 and return. */
+
+	movi	a3, exc_table
+	xsr	a3, EXCSAVE_1
+
+	rfde
+
+
+/*
+ * spill all registers.
+ *
+ * This is not a real function. The following conditions must be met:
+ *
+ *  - must be called with call0.
+ *  - uses DEPC, a3 and SAR.
+ *  - the last 'valid' register of each frame are clobbered.
+ *  - the caller must have registered a fixup handler
+ *    (or be inside a critical section)
+ *  - PS_EXCM must be set (PS_WOE cleared?)
+ */
+
+ENTRY(_spill_registers)
+
+	/*
+	 * Rotate ws so that the current windowbase is at bit 0.
+	 * Assume ws = xxxwww1yy (www1 current window frame).
+	 * Rotate ws right so that a2 = yyxxxwww1.
+	 */
+
+	wsr	a2, DEPC		# preserve a2
+	rsr	a2, WINDOWBASE
+	rsr	a3, WINDOWSTART
+	ssr	a2			# holds WB
+	slli	a2, a3, WSBITS
+	or	a3, a3, a2		# a2 = xxxwww1yyxxxwww1yy
+	srl	a3, a3
+
+	/* We are done if there are no more than the current register frame. */
+
+	extui	a3, a3, 1, WSBITS-2	# a3 = 0yyxxxwww
+	movi	a2, (1 << (WSBITS-1))
+	_beqz	a3, .Lnospill		# only one active frame? jump
+
+	/* We want 1 at the top, so that we return to the current windowbase */
+
+	or	a3, a3, a2		# 1yyxxxwww
+
+	/* Skip empty frames - get 'oldest' WINDOWSTART-bit. */
+
+	wsr	a3, WINDOWSTART		# save shifted windowstart
+	neg	a2, a3
+	and	a3, a2, a3		# first bit set from right: 000010000
+
+	ffs_ws	a2, a3			# a2: shifts to skip empty frames
+	movi	a3, WSBITS
+	sub	a2, a3, a2		# WSBITS-a2:number of 0-bits from right
+	ssr	a2			# save in SAR for later.
+
+	rsr	a3, WINDOWBASE
+	add	a3, a3, a2
+	rsr	a2, DEPC		# restore a2
+	wsr	a3, WINDOWBASE
+	rsync
+
+	rsr	a3, WINDOWSTART
+	srl	a3, a3			# shift windowstart
+
+	/* WB is now just one frame below the oldest frame in the register
+	   window. WS is shifted so the oldest frame is in bit 0, thus, WB
+	   and WS differ by one 4-register frame. */
+
+	/* Save frames. Depending what call was used (call4, call8, call12),
+	 * we have to save 4,8. or 12 registers.
+	 */
+
+	_bbsi.l	a3, 1, .Lc4
+	_bbsi.l	a3, 2, .Lc8
+
+	/* Special case: we have a call12-frame starting at a4. */
+
+	_bbci.l	a3, 3, .Lc12	# bit 3 shouldn't be zero! (Jump to Lc12 first)
+
+	s32e	a4, a1, -16	# a1 is valid with an empty spill area
+	l32e	a4, a5, -12
+	s32e	a8, a4, -48
+	mov	a8, a4
+	l32e	a4, a1, -16
+	j	.Lc12c
+
+.Lloop: _bbsi.l	a3, 1, .Lc4
+	_bbci.l	a3, 2, .Lc12
+
+.Lc8:	s32e	a4, a13, -16
+	l32e	a4, a5, -12
+	s32e	a8, a4, -32
+	s32e	a5, a13, -12
+	s32e	a6, a13, -8
+	s32e	a7, a13, -4
+	s32e	a9, a4, -28
+	s32e	a10, a4, -24
+	s32e	a11, a4, -20
+
+	srli	a11, a3, 2		# shift windowbase by 2
+	rotw	2
+	_bnei	a3, 1, .Lloop
+
+.Lexit: /* Done. Do the final rotation, set WS, and return. */
+
+	rotw	1
+	rsr	a3, WINDOWBASE
+	ssl	a3
+	movi	a3, 1
+	sll	a3, a3
+	wsr	a3, WINDOWSTART
+
+.Lnospill:
+	jx	a0
+
+.Lc4:	s32e	a4, a9, -16
+	s32e	a5, a9, -12
+	s32e	a6, a9, -8
+	s32e	a7, a9, -4
+
+	srli	a7, a3, 1
+	rotw	1
+	_bnei	a3, 1, .Lloop
+	j	.Lexit
+
+.Lc12:	_bbci.l	a3, 3, .Linvalid_mask	# bit 2 shouldn't be zero!
+
+	/* 12-register frame (call12) */
+
+	l32e	a2, a5, -12
+	s32e	a8, a2, -48
+	mov	a8, a2
+
+.Lc12c: s32e	a9, a8, -44
+	s32e	a10, a8, -40
+	s32e	a11, a8, -36
+	s32e	a12, a8, -32
+	s32e	a13, a8, -28
+	s32e	a14, a8, -24
+	s32e	a15, a8, -20
+	srli	a15, a3, 3
+
+	/* The stack pointer for a4..a7 is out of reach, so we rotate the
+	 * window, grab the stackpointer, and rotate back.
+	 * Alternatively, we could also use the following approach, but that
+	 * makes the fixup routine much more complicated:
+	 * rotw	1
+	 * s32e	a0, a13, -16
+	 * ...
+	 * rotw 2
+	 */
+
+	rotw	1
+	mov	a5, a13
+	rotw	-1
+
+	s32e	a4, a9, -16
+	s32e	a5, a9, -12
+	s32e	a6, a9, -8
+	s32e	a7, a9, -4
+
+	rotw	3
+
+	_beqi	a3, 1, .Lexit
+	j	.Lloop
+
+.Linvalid_mask:
+
+	/* We get here because of an unrecoverable error in the window
+	 * registers. If we are in user space, we kill the application,
+	 * however, this condition is unrecoverable in kernel space.
+	 */
+
+	rsr	a0, PS
+	_bbci.l	a0, PS_UM_SHIFT, 1f
+
+ 	/* User space: Setup a dummy frame and kill application.
+	 * Note: We assume EXC_TABLE_KSTK contains a valid stack pointer.
+	 */
+
+	movi	a0, 1
+	movi	a1, 0
+
+	wsr	a0, WINDOWSTART
+	wsr	a1, WINDOWBASE
+	rsync
+
+	movi	a0, 0
+
+	movi	a3, exc_table
+	l32i	a1, a3, EXC_TABLE_KSTK
+	wsr	a3, EXCSAVE_1
+
+	movi	a4, PS_WOE_MASK | 1
+	wsr	a4, PS
+	rsync
+
+	movi	a6, SIGSEGV
+	movi	a4, do_exit
+	callx4	a4
+
+1:	/* Kernel space: PANIC! */
+
+	wsr	a0, EXCSAVE_1
+	movi	a0, unrecoverable_exception
+	callx0	a0		# should not return
+1:	j	1b
+
+/*
+ * We should never get here. Bail out!
+ */
+
+ENTRY(fast_second_level_miss_double_kernel)
+
+1:	movi	a0, unrecoverable_exception
+	callx0	a0		# should not return
+1:	j	1b
+
+/* First-level entry handler for user, kernel, and double 2nd-level
+ * TLB miss exceptions.  Note that for now, user and kernel miss
+ * exceptions share the same entry point and are handled identically.
+ *
+ * An old, less-efficient C version of this function used to exist.
+ * We include it below, interleaved as comments, for reference.
+ *
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ */
+
+ENTRY(fast_second_level_miss)
+
+	/* Save a1. Note: we don't expect a double exception. */
+
+	s32i	a1, a2, PT_AREG1
+
+	/* We need to map the page of PTEs for the user task.  Find
+	 * the pointer to that page.  Also, it's possible for tsk->mm
+	 * to be NULL while tsk->active_mm is nonzero if we faulted on
+	 * a vmalloc address.  In that rare case, we must use
+	 * active_mm instead to avoid a fault in this handler.  See
+	 *
+	 * http://mail.nl.linux.org/linux-mm/2002-08/msg00258.html
+	 *   (or search Internet on "mm vs. active_mm")
+	 *
+	 *	if (!mm)
+	 *		mm = tsk->active_mm;
+	 *	pgd = pgd_offset (mm, regs->excvaddr);
+	 *	pmd = pmd_offset (pgd, regs->excvaddr);
+	 *	pmdval = *pmd;
+	 */
+
+	GET_CURRENT(a1,a2)
+	l32i	a0, a1, TASK_MM		# tsk->mm
+	beqz	a0, 9f
+
+8:	rsr	a1, EXCVADDR		# fault address
+	_PGD_OFFSET(a0, a1, a1)
+	l32i	a0, a0, 0		# read pmdval
+	//beqi	a0, _PAGE_USER, 2f
+	beqz	a0, 2f
+
+	/* Read ptevaddr and convert to top of page-table page.
+	 *
+	 * 	vpnval = read_ptevaddr_register() & PAGE_MASK;
+	 * 	vpnval += DTLB_WAY_PGTABLE;
+	 *	pteval = mk_pte (virt_to_page(pmd_val(pmdval)), PAGE_KERNEL);
+	 *	write_dtlb_entry (pteval, vpnval);
+	 *
+	 * The messy computation for 'pteval' above really simplifies
+	 * into the following:
+	 *
+	 * pteval = ((pmdval - PAGE_OFFSET) & PAGE_MASK) | PAGE_KERNEL
+	 */
+
+	movi	a1, -PAGE_OFFSET
+	add	a0, a0, a1		# pmdval - PAGE_OFFSET
+	extui	a1, a0, 0, PAGE_SHIFT	# ... & PAGE_MASK
+	xor	a0, a0, a1
+
+
+	movi	a1, PAGE_DIRECTORY
+	or	a0, a0, a1		# ... | PAGE_DIRECTORY
+
+	rsr	a1, PTEVADDR
+	srli	a1, a1, PAGE_SHIFT
+	slli	a1, a1, PAGE_SHIFT	# ptevaddr & PAGE_MASK
+	addi	a1, a1, DTLB_WAY_PGTABLE	# ... + way_number
+
+	wdtlb	a0, a1
+	dsync
+
+	/* Exit critical section. */
+
+	movi	a0, 0
+	s32i	a0, a3, EXC_TABLE_FIXUP
+
+	/* Restore the working registers, and return. */
+
+	l32i	a0, a2, PT_AREG0
+	l32i	a1, a2, PT_AREG1
+	l32i	a2, a2, PT_DEPC
+	xsr	a3, EXCSAVE_1
+
+	bgeui	a2, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f
+
+	/* Restore excsave1 and return. */
+
+	rsr	a2, DEPC
+	rfe
+
+	/* Return from double exception. */
+
+1:	xsr	a2, DEPC
+	esync
+	rfde
+
+9:	l32i	a0, a1, TASK_ACTIVE_MM	# unlikely case mm == 0
+	j	8b
+
+2:	/* Invalid PGD, default exception handling */
+
+	rsr	a1, DEPC
+	xsr	a3, EXCSAVE_1
+	s32i	a1, a2, PT_AREG2
+	s32i	a3, a2, PT_AREG3
+	mov	a1, a2
+
+	rsr	a2, PS
+	bbsi.l	a2, PS_UM_SHIFT, 1f
+	j	_kernel_exception
+1:	j	_user_exception
+
+
+/*
+ * StoreProhibitedException
+ *
+ * Update the pte and invalidate the itlb mapping for this pte.
+ *
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ */
+
+ENTRY(fast_store_prohibited)
+
+	/* Save a1 and a4. */
+
+	s32i	a1, a2, PT_AREG1
+	s32i	a4, a2, PT_AREG4
+
+	GET_CURRENT(a1,a2)
+	l32i	a0, a1, TASK_MM		# tsk->mm
+	beqz	a0, 9f
+
+8:	rsr	a1, EXCVADDR		# fault address
+	_PGD_OFFSET(a0, a1, a4)
+	l32i	a0, a0, 0
+	//beqi	a0, _PAGE_USER, 2f	# FIXME use _PAGE_INVALID
+	beqz	a0, 2f
+
+	_PTE_OFFSET(a0, a1, a4)
+	l32i	a4, a0, 0		# read pteval
+	movi	a1, _PAGE_VALID | _PAGE_RW
+	bnall	a4, a1, 2f
+
+	movi	a1, _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_WRENABLE
+	or	a4, a4, a1
+	rsr	a1, EXCVADDR
+	s32i	a4, a0, 0
+
+	/* We need to flush the cache if we have page coloring. */
+#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK
+	dhwb	a0, 0
+#endif
+	pdtlb	a0, a1
+	beqz	a0, 1f
+	idtlb	a0		// FIXME do we need this?
+	wdtlb	a4, a0
+1:
+
+	/* Exit critical section. */
+
+	movi	a0, 0
+	s32i	a0, a3, EXC_TABLE_FIXUP
+
+	/* Restore the working registers, and return. */
+
+	l32i	a4, a2, PT_AREG4
+	l32i	a1, a2, PT_AREG1
+	l32i	a0, a2, PT_AREG0
+	l32i	a2, a2, PT_DEPC
+
+	/* Restore excsave1 and a3. */
+
+	xsr	a3, EXCSAVE_1
+	bgeui	a2, VALID_DOUBLE_EXCEPTION_ADDRESS, 1f
+
+	rsr	a2, DEPC
+	rfe
+
+	/* Double exception. Restore FIXUP handler and return. */
+
+1:	xsr	a2, DEPC
+	esync
+	rfde
+
+9:	l32i	a0, a1, TASK_ACTIVE_MM	# unlikely case mm == 0
+	j	8b
+
+2:	/* If there was a problem, handle fault in C */
+
+	rsr	a4, DEPC	# still holds a2
+	xsr	a3, EXCSAVE_1
+	s32i	a4, a2, PT_AREG2
+	s32i	a3, a2, PT_AREG3
+	l32i	a4, a2, PT_AREG4
+	mov	a1, a2
+
+	rsr	a2, PS
+	bbsi.l	a2, PS_UM_SHIFT, 1f
+	j	_kernel_exception
+1:	j	_user_exception
+
+
+#if XCHAL_EXTRA_SA_SIZE
+
+#warning fast_coprocessor untested
+
+/*
+ * Entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original in DEPC
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ *   PT_DEPC >= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception, DEPC
+ *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ */
+
+ENTRY(fast_coprocessor_double)
+	wsr	a0, EXCSAVE_1
+	movi	a0, unrecoverable_exception
+	callx0	a0
+
+ENTRY(fast_coprocessor)
+
+	/* Fatal if we are in a double exception. */
+
+	l32i	a0, a2, PT_DEPC
+	_bgeui	a0, VALID_DOUBLE_EXCEPTION_ADDRESS, fast_coprocessor_double
+
+	/* Save some registers a1, a3, a4, SAR */
+
+	xsr	a3, EXCSAVE_1
+	s32i	a3, a2, PT_AREG3
+	rsr	a3, SAR
+	s32i	a4, a2, PT_AREG4
+	s32i	a1, a2, PT_AREG1
+	s32i	a5, a1, PT_AREG5
+	s32i	a3, a2, PT_SAR
+	mov	a1, a2
+
+	/* Currently, the HAL macros only guarantee saving a0 and a1.
+	 * These can and will be refined in the future, but for now,
+	 * just save the remaining registers of a2...a15.
+	 */
+	s32i	a6, a1, PT_AREG6
+	s32i	a7, a1, PT_AREG7
+	s32i	a8, a1, PT_AREG8
+	s32i	a9, a1, PT_AREG9
+	s32i	a10, a1, PT_AREG10
+	s32i	a11, a1, PT_AREG11
+	s32i	a12, a1, PT_AREG12
+	s32i	a13, a1, PT_AREG13
+	s32i	a14, a1, PT_AREG14
+	s32i	a15, a1, PT_AREG15
+
+	/* Find coprocessor number. Subtract first CP EXCCAUSE from EXCCAUSE */
+
+	rsr	a0, EXCCAUSE
+	addi	a3, a0, -XCHAL_EXCCAUSE_COPROCESSOR0_DISABLED
+
+	/* Set corresponding CPENABLE bit */
+
+	movi	a4, 1
+	ssl	a3			# SAR: 32 - coprocessor_number
+	rsr	a5, CPENABLE
+	sll	a4, a4
+	or	a4, a5, a4
+	wsr	a4, CPENABLE
+	rsync
+	movi	a5, coprocessor_info	# list of owner and offset into cp_save
+	addx8	a0, a4, a5		# entry for CP
+
+	bne	a4, a5, .Lload		# bit wasn't set before, cp not in use
+
+	/* Now compare the current task with the owner of the coprocessor.
+	 * If they are the same, there is no reason to save or restore any
+	 * coprocessor state. Having already enabled the coprocessor,
+	 * branch ahead to return.
+	 */
+	GET_CURRENT(a5,a1)
+	l32i	a4, a0, COPROCESSOR_INFO_OWNER	# a4: current owner for this CP
+	beq	a4, a5, .Ldone
+
+	/* Find location to dump current coprocessor state:
+	 *  task_struct->task_cp_save_offset + coprocessor_offset[coprocessor]
+	 *
+	 * Note: a0 pointer to the entry in the coprocessor owner table,
+	 *	 a3 coprocessor number,
+         *	 a4 current owner of coprocessor.
+	 */
+	l32i	a5, a0, COPROCESSOR_INFO_OFFSET
+	addi	a2, a4, THREAD_CP_SAVE
+	add	a2, a2, a5
+
+	/* Store current coprocessor states. (a5 still has CP number) */
+
+	xchal_cpi_store_funcbody
+
+	/* The macro might have destroyed a3 (coprocessor number), but
+	 * SAR still has 32 - coprocessor_number!
+	 */
+	movi	a3, 32
+	rsr	a4, SAR
+	sub	a3, a3, a4
+
+.Lload:	/* A new task now owns the corpocessors. Save its TCB pointer into
+	 * the coprocessor owner table.
+	 *
+	 * Note: a0 pointer to the entry in the coprocessor owner table,
+	 *	 a3 coprocessor number.
+	 */
+	GET_CURRENT(a4,a1)
+	s32i	a4, a0, 0
+
+	/* Find location from where to restore the current coprocessor state.*/
+
+	l32i	a5, a0, COPROCESSOR_INFO_OFFSET
+	addi	a2, a4, THREAD_CP_SAVE
+	add	a2, a2, a4
+
+	xchal_cpi_load_funcbody
+
+	/* We must assume that the xchal_cpi_store_funcbody macro destroyed
+	 * registers a2..a15.
+	 */
+
+.Ldone:	l32i	a15, a1, PT_AREG15
+	l32i	a14, a1, PT_AREG14
+	l32i	a13, a1, PT_AREG13
+	l32i	a12, a1, PT_AREG12
+	l32i	a11, a1, PT_AREG11
+	l32i	a10, a1, PT_AREG10
+	l32i	a9, a1, PT_AREG9
+	l32i	a8, a1, PT_AREG8
+	l32i	a7, a1, PT_AREG7
+	l32i	a6, a1, PT_AREG6
+	l32i	a5, a1, PT_AREG5
+	l32i	a4, a1, PT_AREG4
+	l32i	a3, a1, PT_AREG3
+	l32i	a2, a1, PT_AREG2
+	l32i	a0, a1, PT_AREG0
+	l32i	a1, a1, PT_AREG1
+
+	rfe
+
+#endif /* XCHAL_EXTRA_SA_SIZE */
+
+/*
+ * Task switch.
+ *
+ * struct task*  _switch_to (struct task* prev, struct task* next)
+ *         a2                              a2                 a3
+ */
+
+ENTRY(_switch_to)
+
+	entry	a1, 16
+
+	mov	a4, a3			# preserve a3
+
+	s32i	a0, a2, THREAD_RA	# save return address
+	s32i	a1, a2, THREAD_SP	# save stack pointer
+
+	/* Disable ints while we manipulate the stack pointer; spill regs. */
+
+	movi	a5, PS_EXCM_MASK | LOCKLEVEL
+	xsr	a5, PS
+	rsr	a3, EXCSAVE_1
+	rsync
+	s32i	a3, a3, EXC_TABLE_FIXUP	/* enter critical section */
+
+	call0	_spill_registers
+
+	/* Set kernel stack (and leave critical section)
+	 * Note: It's save to set it here. The stack will not be overwritten
+	 *       because the kernel stack will only be loaded again after
+	 *       we return from kernel space.
+	 */
+
+	l32i	a0, a4, TASK_THREAD_INFO
+	rsr	a3, EXCSAVE_1		# exc_table
+	movi	a1, 0
+	addi	a0, a0, PT_REGS_OFFSET
+	s32i	a1, a3, EXC_TABLE_FIXUP
+	s32i	a0, a3, EXC_TABLE_KSTK
+
+	/* restore context of the task that 'next' addresses */
+
+	l32i	a0, a4, THREAD_RA	/* restore return address */
+	l32i	a1, a4, THREAD_SP	/* restore stack pointer */
+
+	wsr	a5, PS
+	rsync
+
+	retw
+
+
+ENTRY(ret_from_fork)
+
+	/* void schedule_tail (struct task_struct *prev)
+	 * Note: prev is still in a6 (return value from fake call4 frame)
+	 */
+	movi	a4, schedule_tail
+	callx4	a4
+
+	movi	a4, do_syscall_trace
+	callx4	a4
+
+	j	common_exception_return
+
+
+
+/*
+ * Table of syscalls
+ */
+
+.data
+.align  4
+.global sys_call_table
+sys_call_table:
+
+#define SYSCALL(call, narg) .word call
+#include "syscalls.h"
+
+/*
+ * Number of arguments of each syscall
+ */
+
+.global sys_narg_table
+sys_narg_table:
+
+#undef SYSCALL
+#define SYSCALL(call, narg) .byte narg
+#include "syscalls.h"
+
diff --git a/arch/xtensa/kernel/head.S b/arch/xtensa/kernel/head.S
new file mode 100644
index 000000000000..6e9b5225b8f6
--- /dev/null
+++ b/arch/xtensa/kernel/head.S
@@ -0,0 +1,237 @@
+/*
+ * arch/xtensa/kernel/head.S
+ *
+ * Xtensa Processor startup code.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * Kevin Chea
+ */
+
+#include <xtensa/cacheasm.h>
+#include <linux/config.h>
+#include <asm/processor.h>
+#include <asm/page.h>
+
+/*
+ * This module contains the entry code for kernel images. It performs the
+ * minimal setup needed to call the generic C routines.
+ *
+ * Prerequisites:
+ *
+ * - The kernel image has been loaded to the actual address where it was
+ *   compiled to.
+ * - a2 contains either 0 or a pointer to a list of boot parameters.
+ *   (see setup.c for more details)
+ *
+ */
+
+	.macro	iterate	from, to , cmd
+		.ifeq	((\to - \from) & ~0xfff)
+			\cmd	\from
+			iterate "(\from+1)", \to, \cmd
+		.endif
+	.endm
+
+/*
+ *  _start
+ *
+ *  The bootloader passes a pointer to a list of boot parameters in a2.
+ */
+
+	/* The first bytes of the kernel image must be an instruction, so we
+	 * manually allocate and define the literal constant we need for a jx
+	 * instruction.
+	 */
+
+	.section .head.text, "ax"
+	.globl _start
+_start:	_j	2f
+	.align	4
+1:	.word	_startup
+2:	l32r	a0, 1b
+	jx	a0
+
+	.text
+	.align 4
+_startup:
+
+	/* Disable interrupts and exceptions. */
+
+	movi	a0, XCHAL_PS_EXCM_MASK
+	wsr	a0, PS
+
+	/* Preserve the pointer to the boot parameter list in EXCSAVE_1 */
+
+	wsr	a2, EXCSAVE_1
+
+	/* Start with a fresh windowbase and windowstart.  */
+
+	movi	a1, 1
+	movi	a0, 0
+	wsr	a1, WINDOWSTART
+	wsr	a0, WINDOWBASE
+	rsync
+
+	/* Set a0 to 0 for the remaining initialization. */
+
+	movi	a0, 0
+
+	/* Clear debugging registers. */
+
+#if XCHAL_HAVE_DEBUG
+	wsr	a0, IBREAKENABLE
+	wsr	a0, ICOUNT
+	movi	a1, 15
+	wsr	a0, ICOUNTLEVEL
+
+	.macro reset_dbreak num
+	wsr	a0, DBREAKC + \num
+	.endm
+
+        iterate 0, XCHAL_NUM_IBREAK-1, reset_dbreak
+#endif
+
+	/* Clear CCOUNT (not really necessary, but nice) */
+
+	wsr	a0, CCOUNT	# not really necessary, but nice
+
+	/* Disable zero-loops. */
+
+#if XCHAL_HAVE_LOOPS
+	wsr	a0, LCOUNT
+#endif
+
+	/* Disable all timers. */
+
+	.macro	reset_timer	num
+	wsr	a0, CCOMPARE_0 + \num
+	.endm
+	iterate	0, XCHAL_NUM_TIMERS-1, reset_timer
+
+	/* Interrupt initialization. */
+
+	movi	a2, XCHAL_INTTYPE_MASK_SOFTWARE | XCHAL_INTTYPE_MASK_EXTERN_EDGE
+	wsr	a0, INTENABLE
+	wsr	a2, INTCLEAR
+
+	/* Disable coprocessors. */
+
+#if XCHAL_CP_NUM > 0
+	wsr	a0, CPENABLE
+#endif
+
+	/* Set PS.INTLEVEL=1, PS.WOE=0, kernel stack, PS.EXCM=0
+	 *
+	 * Note: PS.EXCM must be cleared before using any loop
+	 *	 instructions; otherwise, they are silently disabled, and
+	 * 	 at most one iteration of the loop is executed.
+	 */
+
+	movi	a1, 1
+	wsr	a1, PS
+	rsync
+
+	/*  Initialize the caches.
+	 *  Does not include flushing writeback d-cache.
+	 *  a6, a7 are just working registers (clobbered).
+	 */
+
+	icache_reset  a2, a3
+	dcache_reset  a2, a3
+
+	/* Unpack data sections
+	 *
+	 * The linker script used to build the Linux kernel image
+	 * creates a table located at __boot_reloc_table_start
+	 * that contans the information what data needs to be unpacked.
+	 *
+	 * Uses a2-a7.
+	 */
+
+	movi	a2, __boot_reloc_table_start
+	movi	a3, __boot_reloc_table_end
+
+1:	beq	a2, a3, 3f	# no more entries?
+	l32i	a4, a2, 0	# start destination (in RAM)
+	l32i	a5, a2, 4	# end desination (in RAM)
+	l32i	a6, a2, 8	# start source (in ROM)
+	addi	a2, a2, 12	# next entry
+	beq	a4, a5, 1b	# skip, empty entry
+	beq	a4, a6, 1b	# skip, source and dest. are the same
+
+2:	l32i	a7, a6, 0	# load word
+	addi	a6, a6, 4
+	s32i	a7, a4, 0	# store word
+	addi	a4, a4, 4
+	bltu	a4, a5, 2b
+	j	1b
+
+3:
+	/* All code and initialized data segments have been copied.
+	 * Now clear the BSS segment.
+	 */
+
+	movi	a2, _bss_start	# start of BSS
+	movi	a3, _bss_end	# end of BSS
+
+1:	addi	a2, a2, 4
+	s32i	a0, a2, 0
+	blt	a2, a3, 1b
+
+#if XCHAL_DCACHE_IS_WRITEBACK
+
+	/* After unpacking, flush the writeback cache to memory so the
+	 * instructions/data are available.
+	 */
+
+	dcache_writeback_all	a2, a3
+#endif
+
+	/* Setup stack and enable window exceptions (keep irqs disabled) */
+
+	movi	a1, init_thread_union
+	addi	a1, a1, KERNEL_STACK_SIZE
+
+	movi	a2, 0x00040001		# WOE=1, INTLEVEL=1, UM=0
+	wsr	a2, PS			# (enable reg-windows; progmode stack)
+	rsync
+
+	/* Set up EXCSAVE[DEBUGLEVEL] to point to the Debug Exception Handler.*/
+
+	movi	a2, debug_exception
+	wsr	a2, EXCSAVE + XCHAL_DEBUGLEVEL
+
+	/* Set up EXCSAVE[1] to point to the exc_table. */
+
+	movi	a6, exc_table
+	xsr	a6, EXCSAVE_1
+
+	/* init_arch kick-starts the linux kernel */
+
+	movi	a4, init_arch
+	callx4	a4
+
+	movi	a4, start_kernel
+	callx4	a4
+
+should_never_return:
+	j	should_never_return
+
+	/* Define some common data structures here.  We define them
+	 * here in this assembly file due to their unusual alignment
+	 * requirements.
+	 */
+
+	.comm   swapper_pg_dir,PAGE_SIZE,PAGE_SIZE
+	.comm	empty_bad_page_table,PAGE_SIZE,PAGE_SIZE
+	.comm	empty_bad_page,PAGE_SIZE,PAGE_SIZE
+	.comm	empty_zero_page,PAGE_SIZE,PAGE_SIZE
+
diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
new file mode 100644
index 000000000000..4cbf6d91571f
--- /dev/null
+++ b/arch/xtensa/kernel/irq.c
@@ -0,0 +1,192 @@
+/*
+ * linux/arch/xtensa/kernel/irq.c
+ *
+ * Xtensa built-in interrupt controller and some generic functions copied
+ * from i386.
+ *
+ * Copyright (C) 2002 - 2005 Tensilica, Inc.
+ * Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar
+ *
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Kevin Chea
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/seq_file.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel_stat.h>
+
+#include <asm/uaccess.h>
+#include <asm/platform.h>
+
+static void enable_xtensa_irq(unsigned int irq);
+static void disable_xtensa_irq(unsigned int irq);
+static void mask_and_ack_xtensa(unsigned int irq);
+static void end_xtensa_irq(unsigned int irq);
+
+static unsigned int cached_irq_mask;
+
+atomic_t irq_err_count;
+
+/*
+ * 'what should we do if we get a hw irq event on an illegal vector'.
+ * each architecture has to answer this themselves.
+ */
+void ack_bad_irq(unsigned int irq)
+{
+          printk("unexpected IRQ trap at vector %02x\n", irq);
+}
+
+/*
+ * do_IRQ handles all normal device IRQ's (the special
+ * SMP cross-CPU interrupts have their own specific
+ * handlers).
+ */
+
+unsigned int  do_IRQ(int irq, struct pt_regs *regs)
+{
+	irq_enter();
+
+#ifdef CONFIG_DEBUG_STACKOVERFLOW
+	/* Debugging check for stack overflow: is there less than 1KB free? */
+	{
+		unsigned long sp;
+
+		__asm__ __volatile__ ("mov %0, a1\n" : "=a" (sp));
+		sp &= THREAD_SIZE - 1;
+
+		if (unlikely(sp < (sizeof(thread_info) + 1024)))
+			printk("Stack overflow in do_IRQ: %ld\n",
+			       sp - sizeof(struct thread_info));
+	}
+#endif
+
+	__do_IRQ(irq, regs);
+
+	irq_exit();
+
+	return 1;
+}
+
+/*
+ * Generic, controller-independent functions:
+ */
+
+int show_interrupts(struct seq_file *p, void *v)
+{
+	int i = *(loff_t *) v, j;
+	struct irqaction * action;
+	unsigned long flags;
+
+	if (i == 0) {
+		seq_printf(p, "           ");
+		for (j=0; j<NR_CPUS; j++)
+			if (cpu_online(j))
+				seq_printf(p, "CPU%d       ",j);
+		seq_putc(p, '\n');
+	}
+
+	if (i < NR_IRQS) {
+		spin_lock_irqsave(&irq_desc[i].lock, flags);
+		action = irq_desc[i].action;
+		if (!action)
+			goto skip;
+		seq_printf(p, "%3d: ",i);
+#ifndef CONFIG_SMP
+		seq_printf(p, "%10u ", kstat_irqs(i));
+#else
+		for (j = 0; j < NR_CPUS; j++)
+			if (cpu_online(j))
+				seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+#endif
+		seq_printf(p, " %14s", irq_desc[i].handler->typename);
+		seq_printf(p, "  %s", action->name);
+
+		for (action=action->next; action; action = action->next)
+			seq_printf(p, ", %s", action->name);
+
+		seq_putc(p, '\n');
+skip:
+		spin_unlock_irqrestore(&irq_desc[i].lock, flags);
+	} else if (i == NR_IRQS) {
+		seq_printf(p, "NMI: ");
+		for (j = 0; j < NR_CPUS; j++)
+			if (cpu_online(j))
+				seq_printf(p, "%10u ", nmi_count(j));
+		seq_putc(p, '\n');
+		seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
+	}
+	return 0;
+}
+/* shutdown is same as "disable" */
+#define shutdown_xtensa_irq disable_xtensa_irq
+
+static unsigned int startup_xtensa_irq(unsigned int irq)
+{
+	enable_xtensa_irq(irq);
+	return 0;               /* never anything pending */
+}
+
+static struct hw_interrupt_type xtensa_irq_type = {
+	"Xtensa-IRQ",
+	startup_xtensa_irq,
+	shutdown_xtensa_irq,
+	enable_xtensa_irq,
+	disable_xtensa_irq,
+	mask_and_ack_xtensa,
+	end_xtensa_irq
+};
+
+static inline void mask_irq(unsigned int irq)
+{
+	cached_irq_mask &= ~(1 << irq);
+	set_sr (cached_irq_mask, INTENABLE);
+}
+
+static inline void unmask_irq(unsigned int irq)
+{
+	cached_irq_mask |= 1 << irq;
+	set_sr (cached_irq_mask, INTENABLE);
+}
+
+static void disable_xtensa_irq(unsigned int irq)
+{
+	unsigned long flags;
+	local_save_flags(flags);
+	mask_irq(irq);
+	local_irq_restore(flags);
+}
+
+static void enable_xtensa_irq(unsigned int irq)
+{
+	unsigned long flags;
+	local_save_flags(flags);
+	unmask_irq(irq);
+	local_irq_restore(flags);
+}
+
+static void mask_and_ack_xtensa(unsigned int irq)
+{
+        disable_xtensa_irq(irq);
+}
+
+static void end_xtensa_irq(unsigned int irq)
+{
+        enable_xtensa_irq(irq);
+}
+
+
+void __init init_IRQ(void)
+{
+	int i;
+
+	for (i=0; i < XTENSA_NR_IRQS; i++)
+		irq_desc[i].handler = &xtensa_irq_type;
+
+	cached_irq_mask = 0;
+
+	platform_init_irq();
+}
diff --git a/arch/xtensa/kernel/module.c b/arch/xtensa/kernel/module.c
new file mode 100644
index 000000000000..d1683cfa19a2
--- /dev/null
+++ b/arch/xtensa/kernel/module.c
@@ -0,0 +1,78 @@
+/*
+ * arch/xtensa/kernel/platform.c
+ *
+ * Module support.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/moduleloader.h>
+#include <linux/elf.h>
+#include <linux/vmalloc.h>
+#include <linux/fs.h>
+#include <linux/string.h>
+#include <linux/kernel.h>
+#include <linux/cache.h>
+
+LIST_HEAD(module_buf_list);
+
+void *module_alloc(unsigned long size)
+{
+  panic("module_alloc not implemented");
+}
+
+void module_free(struct module *mod, void *module_region)
+{
+  panic("module_free not implemented");
+}
+
+int module_frob_arch_sections(Elf32_Ehdr *hdr,
+    			      Elf32_Shdr *sechdrs,
+			      char *secstrings,
+			      struct module *me)
+{
+  panic("module_frob_arch_sections not implemented");
+}
+
+int apply_relocate(Elf32_Shdr *sechdrs,
+    		   const char *strtab,
+		   unsigned int symindex,
+		   unsigned int relsec,
+		   struct module *module)
+{
+  panic ("apply_relocate not implemented");
+}
+
+int apply_relocate_add(Elf32_Shdr *sechdrs,
+		       const char *strtab,
+		       unsigned int symindex,
+		       unsigned int relsec,
+		       struct module *module)
+{
+  panic("apply_relocate_add not implemented");
+}
+
+int module_finalize(const Elf_Ehdr *hdr,
+    		    const Elf_Shdr *sechdrs,
+		    struct module *me)
+{
+  panic ("module_finalize not implemented");
+}
+
+void module_arch_cleanup(struct module *mod)
+{
+  panic("module_arch_cleanup not implemented");
+}
+
+struct bug_entry *module_find_bug(unsigned long bugaddr)
+{
+  panic("module_find_bug not implemented");
+}
diff --git a/arch/xtensa/kernel/pci-dma.c b/arch/xtensa/kernel/pci-dma.c
new file mode 100644
index 000000000000..84fde258cf85
--- /dev/null
+++ b/arch/xtensa/kernel/pci-dma.c
@@ -0,0 +1,73 @@
+/*
+ * arch/xtensa/pci-dma.c
+ *
+ * DMA coherent memory allocation.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ * Copyright (C) 2002 - 2005 Tensilica Inc.
+ *
+ * Based on version for i386.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ */
+
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/string.h>
+#include <linux/pci.h>
+#include <asm/io.h>
+#include <asm/cacheflush.h>
+
+/*
+ * Note: We assume that the full memory space is always mapped to 'kseg'
+ *	 Otherwise we have to use page attributes (not implemented).
+ */
+
+void *
+dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *handle, int gfp)
+{
+	void *ret;
+
+	/* ignore region speicifiers */
+	gfp &= ~(__GFP_DMA | __GFP_HIGHMEM);
+
+	if (dev == NULL || (*dev->dma_mask < 0xffffffff))
+		gfp |= GFP_DMA;
+	ret = (void *)__get_free_pages(gfp, get_order(size));
+
+	if (ret != NULL) {
+		memset(ret, 0, size);
+		*handle = virt_to_bus(ret);
+	}
+	return (void*) BYPASS_ADDR((unsigned long)ret);
+}
+
+void dma_free_coherent(struct device *hwdev, size_t size,
+			 void *vaddr, dma_addr_t dma_handle)
+{
+	free_pages(CACHED_ADDR((unsigned long)vaddr), get_order(size));
+}
+
+
+void consistent_sync(void *vaddr, size_t size, int direction)
+{
+	switch (direction) {
+	case PCI_DMA_NONE:
+		BUG();
+	case PCI_DMA_FROMDEVICE:        /* invalidate only */
+		__invalidate_dcache_range((unsigned long)vaddr,
+				          (unsigned long)size);
+		break;
+
+	case PCI_DMA_TODEVICE:          /* writeback only */
+	case PCI_DMA_BIDIRECTIONAL:     /* writeback and invalidate */
+		__flush_invalidate_dcache_range((unsigned long)vaddr,
+				    		(unsigned long)size);
+		break;
+	}
+}
diff --git a/arch/xtensa/kernel/pci.c b/arch/xtensa/kernel/pci.c
new file mode 100644
index 000000000000..d29a81648637
--- /dev/null
+++ b/arch/xtensa/kernel/pci.c
@@ -0,0 +1,563 @@
+/*
+ * arch/xtensa/pcibios.c
+ *
+ * PCI bios-type initialisation for PCI machines
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ * Copyright (C) 2001-2005 Tensilica Inc.
+ *
+ * Based largely on work from Cort (ppc/kernel/pci.c)
+ * IO functions copied from sparc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ *
+ */
+
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/string.h>
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/errno.h>
+#include <linux/bootmem.h>
+
+#include <asm/pci-bridge.h>
+#include <asm/platform.h>
+
+#undef DEBUG
+
+#ifdef DEBUG
+#define DBG(x...) printk(x)
+#else
+#define DBG(x...)
+#endif
+
+/* PCI Controller */
+
+
+/*
+ * pcibios_alloc_controller
+ * pcibios_enable_device
+ * pcibios_fixups
+ * pcibios_align_resource
+ * pcibios_fixup_bus
+ * pcibios_setup
+ * pci_bus_add_device
+ * pci_mmap_page_range
+ */
+
+struct pci_controller* pci_ctrl_head;
+struct pci_controller** pci_ctrl_tail = &pci_ctrl_head;
+
+static int pci_bus_count;
+
+static void pcibios_fixup_resources(struct pci_dev* dev);
+
+#if 0 // FIXME
+struct pci_fixup pcibios_fixups[] = {
+	{ DECLARE_PCI_FIXUP_HEADER, PCI_ANY_ID, PCI_ANY_ID, pcibios_fixup_resources },
+	{ 0 }
+};
+#endif
+
+void
+pcibios_update_resource(struct pci_dev *dev, struct resource *root,
+			struct resource *res, int resource)
+{
+	u32 new, check, mask;
+	int reg;
+	struct pci_controller* pci_ctrl = dev->sysdata;
+
+	new = res->start;
+	if (pci_ctrl && res->flags & IORESOURCE_IO) {
+		new -= pci_ctrl->io_space.base;
+	}
+	new |= (res->flags & PCI_REGION_FLAG_MASK);
+	if (resource < 6) {
+		reg = PCI_BASE_ADDRESS_0 + 4*resource;
+	} else if (resource == PCI_ROM_RESOURCE) {
+		res->flags |= PCI_ROM_ADDRESS_ENABLE;
+		reg = dev->rom_base_reg;
+	} else {
+	/* Somebody might have asked allocation of a non-standard resource */
+		return;
+	}
+
+	pci_write_config_dword(dev, reg, new);
+	pci_read_config_dword(dev, reg, &check);
+	mask = (new & PCI_BASE_ADDRESS_SPACE_IO) ?
+		PCI_BASE_ADDRESS_IO_MASK : PCI_BASE_ADDRESS_MEM_MASK;
+
+	if ((new ^ check) & mask) {
+		printk(KERN_ERR "PCI: Error while updating region "
+		       "%s/%d (%08x != %08x)\n", dev->slot_name, resource,
+		       new, check);
+	}
+}
+
+/*
+ * We need to avoid collisions with `mirrored' VGA ports
+ * and other strange ISA hardware, so we always want the
+ * addresses to be allocated in the 0x000-0x0ff region
+ * modulo 0x400.
+ *
+ * Why? Because some silly external IO cards only decode
+ * the low 10 bits of the IO address. The 0x00-0xff region
+ * is reserved for motherboard devices that decode all 16
+ * bits, so it's ok to allocate at, say, 0x2800-0x28ff,
+ * but we want to try to avoid allocating at 0x2900-0x2bff
+ * which might have be mirrored at 0x0100-0x03ff..
+ */
+void
+pcibios_align_resource(void *data, struct resource *res, unsigned long size,
+    		       unsigned long align)
+{
+	struct pci_dev *dev = data;
+
+	if (res->flags & IORESOURCE_IO) {
+		unsigned long start = res->start;
+
+		if (size > 0x100) {
+			printk(KERN_ERR "PCI: I/O Region %s/%d too large"
+			       " (%ld bytes)\n", dev->slot_name,
+			       dev->resource - res, size);
+		}
+
+		if (start & 0x300) {
+			start = (start + 0x3ff) & ~0x3ff;
+			res->start = start;
+		}
+	}
+}
+
+int
+pcibios_enable_resources(struct pci_dev *dev, int mask)
+{
+	u16 cmd, old_cmd;
+	int idx;
+	struct resource *r;
+
+	pci_read_config_word(dev, PCI_COMMAND, &cmd);
+	old_cmd = cmd;
+	for(idx=0; idx<6; idx++) {
+		r = &dev->resource[idx];
+		if (!r->start && r->end) {
+			printk (KERN_ERR "PCI: Device %s not available because "
+				"of resource collisions\n", dev->slot_name);
+			return -EINVAL;
+		}
+		if (r->flags & IORESOURCE_IO)
+			cmd |= PCI_COMMAND_IO;
+		if (r->flags & IORESOURCE_MEM)
+			cmd |= PCI_COMMAND_MEMORY;
+	}
+	if (dev->resource[PCI_ROM_RESOURCE].start)
+		cmd |= PCI_COMMAND_MEMORY;
+	if (cmd != old_cmd) {
+		printk("PCI: Enabling device %s (%04x -> %04x)\n",
+			dev->slot_name, old_cmd, cmd);
+		pci_write_config_word(dev, PCI_COMMAND, cmd);
+	}
+	return 0;
+}
+
+struct pci_controller * __init pcibios_alloc_controller(void)
+{
+	struct pci_controller *pci_ctrl;
+
+	pci_ctrl = (struct pci_controller *)alloc_bootmem(sizeof(*pci_ctrl));
+	memset(pci_ctrl, 0, sizeof(struct pci_controller));
+
+	*pci_ctrl_tail = pci_ctrl;
+	pci_ctrl_tail = &pci_ctrl->next;
+
+	return pci_ctrl;
+}
+
+static int __init pcibios_init(void)
+{
+	struct pci_controller *pci_ctrl;
+	struct pci_bus *bus;
+	int next_busno = 0, i;
+
+	printk("PCI: Probing PCI hardware\n");
+
+	/* Scan all of the recorded PCI controllers.  */
+	for (pci_ctrl = pci_ctrl_head; pci_ctrl; pci_ctrl = pci_ctrl->next) {
+		pci_ctrl->last_busno = 0xff;
+		bus = pci_scan_bus(pci_ctrl->first_busno, pci_ctrl->ops,
+				   pci_ctrl);
+		if (pci_ctrl->io_resource.flags) {
+			unsigned long offs;
+
+			offs = (unsigned long)pci_ctrl->io_space.base;
+			pci_ctrl->io_resource.start += offs;
+			pci_ctrl->io_resource.end += offs;
+			bus->resource[0] = &pci_ctrl->io_resource;
+		}
+		for (i = 0; i < 3; ++i)
+			if (pci_ctrl->mem_resources[i].flags)
+				bus->resource[i+1] =&pci_ctrl->mem_resources[i];
+		pci_ctrl->bus = bus;
+		pci_ctrl->last_busno = bus->subordinate;
+		if (next_busno <= pci_ctrl->last_busno)
+			next_busno = pci_ctrl->last_busno+1;
+	}
+	pci_bus_count = next_busno;
+
+	return platform_pcibios_fixup();
+}
+
+subsys_initcall(pcibios_init);
+
+void __init pcibios_fixup_bus(struct pci_bus *bus)
+{
+	struct pci_controller *pci_ctrl = bus->sysdata;
+	struct resource *res;
+	unsigned long io_offset;
+	int i;
+
+	io_offset = (unsigned long)pci_ctrl->io_space.base;
+	if (bus->parent == NULL) {
+		/* this is a host bridge - fill in its resources */
+		pci_ctrl->bus = bus;
+
+		bus->resource[0] = res = &pci_ctrl->io_resource;
+		if (!res->flags) {
+			if (io_offset)
+				printk (KERN_ERR "I/O resource not set for host"
+					" bridge %d\n", pci_ctrl->index);
+			res->start = 0;
+			res->end = IO_SPACE_LIMIT;
+			res->flags = IORESOURCE_IO;
+		}
+		res->start += io_offset;
+		res->end += io_offset;
+
+		for (i = 0; i < 3; i++) {
+			res = &pci_ctrl->mem_resources[i];
+			if (!res->flags) {
+				if (i > 0)
+					continue;
+				printk(KERN_ERR "Memory resource not set for "
+				       "host bridge %d\n", pci_ctrl->index);
+				res->start = 0;
+				res->end = ~0U;
+				res->flags = IORESOURCE_MEM;
+			}
+			bus->resource[i+1] = res;
+		}
+	} else {
+		/* This is a subordinate bridge */
+		pci_read_bridge_bases(bus);
+
+		for (i = 0; i < 4; i++) {
+			if ((res = bus->resource[i]) == NULL || !res->flags)
+				continue;
+			if (io_offset && (res->flags & IORESOURCE_IO)) {
+				res->start += io_offset;
+				res->end += io_offset;
+			}
+		}
+	}
+}
+
+char __init *pcibios_setup(char *str)
+{
+	return str;
+}
+
+/* the next one is stolen from the alpha port... */
+
+void __init
+pcibios_update_irq(struct pci_dev *dev, int irq)
+{
+	pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq);
+}
+
+int pcibios_enable_device(struct pci_dev *dev, int mask)
+{
+	u16 cmd, old_cmd;
+	int idx;
+	struct resource *r;
+
+	pci_read_config_word(dev, PCI_COMMAND, &cmd);
+	old_cmd = cmd;
+	for (idx=0; idx<6; idx++) {
+		r = &dev->resource[idx];
+		if (!r->start && r->end) {
+			printk(KERN_ERR "PCI: Device %s not available because "
+			       "of resource collisions\n", dev->slot_name);
+			return -EINVAL;
+		}
+		if (r->flags & IORESOURCE_IO)
+			cmd |= PCI_COMMAND_IO;
+		if (r->flags & IORESOURCE_MEM)
+			cmd |= PCI_COMMAND_MEMORY;
+	}
+	if (cmd != old_cmd) {
+		printk("PCI: Enabling device %s (%04x -> %04x)\n",
+		       dev->slot_name, old_cmd, cmd);
+		pci_write_config_word(dev, PCI_COMMAND, cmd);
+	}
+
+	return 0;
+}
+
+#ifdef CONFIG_PROC_FS
+
+/*
+ * Return the index of the PCI controller for device pdev.
+ */
+
+int
+pci_controller_num(struct pci_dev *dev)
+{
+	struct pci_controller *pci_ctrl = (struct pci_controller*) dev->sysdata;
+	return pci_ctrl->index;
+}
+
+#endif /* CONFIG_PROC_FS */
+
+
+static void
+pcibios_fixup_resources(struct pci_dev *dev)
+{
+	struct pci_controller* pci_ctrl = (struct pci_controller *)dev->sysdata;
+	int i;
+	unsigned long offset;
+
+	if (!pci_ctrl) {
+		printk(KERN_ERR "No pci_ctrl for PCI dev %s!\n",dev->slot_name);
+		return;
+	}
+	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
+		struct resource *res = dev->resource + i;
+		if (!res->start || !res->flags)
+			continue;
+		if (res->end == 0xffffffff) {
+			DBG("PCI:%s Resource %d [%08lx-%08lx] is unassigned\n",
+			    dev->slot_name, i, res->start, res->end);
+			res->end -= res->start;
+			res->start = 0;
+			continue;
+		}
+		offset = 0;
+		if (res->flags & IORESOURCE_IO)
+			offset = (unsigned long) pci_ctrl->io_space.base;
+		else if (res->flags & IORESOURCE_MEM)
+			offset = (unsigned long) pci_ctrl->mem_space.base;
+
+		if (offset != 0) {
+			res->start += offset;
+			res->end += offset;
+#ifdef DEBUG
+			printk("Fixup res %d (%lx) of dev %s: %lx -> %lx\n",
+			       i, res->flags, dev->slot_name,
+			       res->start - offset, res->start);
+#endif
+		}
+	}
+}
+
+/*
+ * Platform support for /proc/bus/pci/X/Y mmap()s,
+ * modelled on the sparc64 implementation by Dave Miller.
+ *  -- paulus.
+ */
+
+/*
+ * Adjust vm_pgoff of VMA such that it is the physical page offset
+ * corresponding to the 32-bit pci bus offset for DEV requested by the user.
+ *
+ * Basically, the user finds the base address for his device which he wishes
+ * to mmap.  They read the 32-bit value from the config space base register,
+ * add whatever PAGE_SIZE multiple offset they wish, and feed this into the
+ * offset parameter of mmap on /proc/bus/pci/XXX for that device.
+ *
+ * Returns negative error code on failure, zero on success.
+ */
+static __inline__ int
+__pci_mmap_make_offset(struct pci_dev *dev, struct vm_area_struct *vma,
+		       enum pci_mmap_state mmap_state)
+{
+	struct pci_controller *pci_ctrl = (struct pci_controller*) dev->sysdata;
+	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
+	unsigned long io_offset = 0;
+	int i, res_bit;
+
+	if (pci_ctrl == 0)
+		return -EINVAL;		/* should never happen */
+
+	/* If memory, add on the PCI bridge address offset */
+	if (mmap_state == pci_mmap_mem) {
+		res_bit = IORESOURCE_MEM;
+	} else {
+		io_offset = (unsigned long)pci_ctrl->io_space.base;
+		offset += io_offset;
+		res_bit = IORESOURCE_IO;
+	}
+
+	/*
+	 * Check that the offset requested corresponds to one of the
+	 * resources of the device.
+	 */
+	for (i = 0; i <= PCI_ROM_RESOURCE; i++) {
+		struct resource *rp = &dev->resource[i];
+		int flags = rp->flags;
+
+		/* treat ROM as memory (should be already) */
+		if (i == PCI_ROM_RESOURCE)
+			flags |= IORESOURCE_MEM;
+
+		/* Active and same type? */
+		if ((flags & res_bit) == 0)
+			continue;
+
+		/* In the range of this resource? */
+		if (offset < (rp->start & PAGE_MASK) || offset > rp->end)
+			continue;
+
+		/* found it! construct the final physical address */
+		if (mmap_state == pci_mmap_io)
+			offset += pci_ctrl->io_space.start - io_offset;
+		vma->vm_pgoff = offset >> PAGE_SHIFT;
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+/*
+ * Set vm_flags of VMA, as appropriate for this architecture, for a pci device
+ * mapping.
+ */
+static __inline__ void
+__pci_mmap_set_flags(struct pci_dev *dev, struct vm_area_struct *vma,
+		     enum pci_mmap_state mmap_state)
+{
+	vma->vm_flags |= VM_SHM | VM_LOCKED | VM_IO;
+}
+
+/*
+ * Set vm_page_prot of VMA, as appropriate for this architecture, for a pci
+ * device mapping.
+ */
+static __inline__ void
+__pci_mmap_set_pgprot(struct pci_dev *dev, struct vm_area_struct *vma,
+		      enum pci_mmap_state mmap_state, int write_combine)
+{
+	int prot = pgprot_val(vma->vm_page_prot);
+
+	/* Set to write-through */
+	prot &= ~_PAGE_NO_CACHE;
+#if 0
+	if (!write_combine)
+		prot |= _PAGE_WRITETHRU;
+#endif
+	vma->vm_page_prot = __pgprot(prot);
+}
+
+/*
+ * Perform the actual remap of the pages for a PCI device mapping, as
+ * appropriate for this architecture.  The region in the process to map
+ * is described by vm_start and vm_end members of VMA, the base physical
+ * address is found in vm_pgoff.
+ * The pci device structure is provided so that architectures may make mapping
+ * decisions on a per-device or per-bus basis.
+ *
+ * Returns a negative error code on failure, zero on success.
+ */
+int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
+			enum pci_mmap_state mmap_state,
+			int write_combine)
+{
+	int ret;
+
+	ret = __pci_mmap_make_offset(dev, vma, mmap_state);
+	if (ret < 0)
+		return ret;
+
+	__pci_mmap_set_flags(dev, vma, mmap_state);
+	__pci_mmap_set_pgprot(dev, vma, mmap_state, write_combine);
+
+	ret = io_remap_page_range(vma, vma->vm_start, vma->vm_pgoff<<PAGE_SHIFT,
+			       vma->vm_end - vma->vm_start, vma->vm_page_prot);
+
+	return ret;
+}
+
+/*
+ * This probably belongs here rather than ioport.c because
+ * we do not want this crud linked into SBus kernels.
+ * Also, think for a moment about likes of floppy.c that
+ * include architecture specific parts. They may want to redefine ins/outs.
+ *
+ * We do not use horroble macroses here because we want to
+ * advance pointer by sizeof(size).
+ */
+void outsb(unsigned long addr, const void *src, unsigned long count) {
+        while (count) {
+                count -= 1;
+                writeb(*(const char *)src, addr);
+                src += 1;
+                addr += 1;
+        }
+}
+
+void outsw(unsigned long addr, const void *src, unsigned long count) {
+        while (count) {
+                count -= 2;
+                writew(*(const short *)src, addr);
+                src += 2;
+                addr += 2;
+        }
+}
+
+void outsl(unsigned long addr, const void *src, unsigned long count) {
+        while (count) {
+                count -= 4;
+                writel(*(const long *)src, addr);
+                src += 4;
+                addr += 4;
+        }
+}
+
+void insb(unsigned long addr, void *dst, unsigned long count) {
+        while (count) {
+                count -= 1;
+                *(unsigned char *)dst = readb(addr);
+                dst += 1;
+                addr += 1;
+        }
+}
+
+void insw(unsigned long addr, void *dst, unsigned long count) {
+        while (count) {
+                count -= 2;
+                *(unsigned short *)dst = readw(addr);
+                dst += 2;
+                addr += 2;
+        }
+}
+
+void insl(unsigned long addr, void *dst, unsigned long count) {
+        while (count) {
+                count -= 4;
+                /*
+                 * XXX I am sure we are in for an unaligned trap here.
+                 */
+                *(unsigned long *)dst = readl(addr);
+                dst += 4;
+                addr += 4;
+        }
+}
+
+
+
diff --git a/arch/xtensa/kernel/platform.c b/arch/xtensa/kernel/platform.c
new file mode 100644
index 000000000000..cf1362784443
--- /dev/null
+++ b/arch/xtensa/kernel/platform.c
@@ -0,0 +1,49 @@
+/*
+ * arch/xtensa/kernel/platform.c
+ *
+ * Default platform functions.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ */
+
+#include <linux/config.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/time.h>
+#include <asm/platform.h>
+#include <asm/timex.h>
+
+#define _F(r,f,a,b)							\
+	r __platform_##f a b;                                   	\
+	r platform_##f a __attribute__((weak, alias("__platform_"#f)))
+
+/*
+ * Default functions that are used if no platform specific function is defined.
+ * (Please, refer to include/asm-xtensa/platform.h for more information)
+ */
+
+_F(void, setup, (char** cmd), { });
+_F(void, init_irq, (void), { });
+_F(void, restart, (void), { while(1); });
+_F(void, halt, (void), { while(1); });
+_F(void, power_off, (void), { while(1); });
+_F(void, idle, (void), { __asm__ __volatile__ ("waiti 0" ::: "memory"); });
+_F(void, heartbeat, (void), { });
+_F(int,  pcibios_fixup, (void), { return 0; });
+_F(int, get_rtc_time, (time_t* t), { return 0; });
+_F(int, set_rtc_time, (time_t t), { return 0; });
+
+#if CONFIG_XTENSA_CALIBRATE_CCOUNT
+_F(void, calibrate_ccount, (void),
+{
+  printk ("ERROR: Cannot calibrate cpu frequency! Assuming 100MHz.\n");
+  ccount_per_jiffy = 100 * (1000000UL/HZ);
+});
+#endif
+
diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c
new file mode 100644
index 000000000000..4099703b14be
--- /dev/null
+++ b/arch/xtensa/kernel/process.c
@@ -0,0 +1,482 @@
+// TODO	verify coprocessor handling
+/*
+ * arch/xtensa/kernel/process.c
+ *
+ * Xtensa Processor version.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * Chris Zankel <chris@zankel.net>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Kevin Chea
+ */
+
+#include <linux/config.h>
+#include <linux/errno.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/stddef.h>
+#include <linux/unistd.h>
+#include <linux/ptrace.h>
+#include <linux/slab.h>
+#include <linux/elf.h>
+#include <linux/init.h>
+#include <linux/prctl.h>
+#include <linux/init_task.h>
+#include <linux/module.h>
+#include <linux/mqueue.h>
+
+#include <asm/pgtable.h>
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/io.h>
+#include <asm/processor.h>
+#include <asm/platform.h>
+#include <asm/mmu.h>
+#include <asm/irq.h>
+#include <asm/atomic.h>
+#include <asm/offsets.h>
+#include <asm/coprocessor.h>
+
+extern void ret_from_fork(void);
+
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS(init_signals);
+static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand);
+struct mm_struct init_mm = INIT_MM(init_mm);
+EXPORT_SYMBOL(init_mm);
+
+union thread_union init_thread_union
+	__attribute__((__section__(".data.init_task"))) =
+{ INIT_THREAD_INFO(init_task) };
+
+struct task_struct init_task = INIT_TASK(init_task);
+EXPORT_SYMBOL(init_task);
+
+struct task_struct *current_set[NR_CPUS] = {&init_task, };
+
+
+#if XCHAL_CP_NUM > 0
+
+/*
+ * Coprocessor ownership.
+ */
+
+coprocessor_info_t coprocessor_info[] = {
+	{ 0, XTENSA_CPE_CP0_OFFSET },
+	{ 0, XTENSA_CPE_CP1_OFFSET },
+	{ 0, XTENSA_CPE_CP2_OFFSET },
+	{ 0, XTENSA_CPE_CP3_OFFSET },
+	{ 0, XTENSA_CPE_CP4_OFFSET },
+	{ 0, XTENSA_CPE_CP5_OFFSET },
+	{ 0, XTENSA_CPE_CP6_OFFSET },
+	{ 0, XTENSA_CPE_CP7_OFFSET },
+};
+
+#endif
+
+/*
+ * Powermanagement idle function, if any is provided by the platform.
+ */
+
+void cpu_idle(void)
+{
+  	local_irq_enable();
+
+	/* endless idle loop with no priority at all */
+	while (1) {
+		while (!need_resched())
+			platform_idle();
+		preempt_enable();
+		schedule();
+	}
+}
+
+/*
+ * Free current thread data structures etc..
+ */
+
+void exit_thread(void)
+{
+	release_coprocessors(current);	/* Empty macro if no CPs are defined */
+}
+
+void flush_thread(void)
+{
+	release_coprocessors(current);	/* Empty macro if no CPs are defined */
+}
+
+/*
+ * Copy thread.
+ *
+ * The stack layout for the new thread looks like this:
+ *
+ *	+------------------------+ <- sp in childregs (= tos)
+ *	|       childregs        |
+ *	+------------------------+ <- thread.sp = sp in dummy-frame
+ *	|      dummy-frame       |    (saved in dummy-frame spill-area)
+ *	+------------------------+
+ *
+ * We create a dummy frame to return to ret_from_fork:
+ *   a0 points to ret_from_fork (simulating a call4)
+ *   sp points to itself (thread.sp)
+ *   a2, a3 are unused.
+ *
+ * Note: This is a pristine frame, so we don't need any spill region on top of
+ *       childregs.
+ */
+
+int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
+		unsigned long unused,
+                struct task_struct * p, struct pt_regs * regs)
+{
+	struct pt_regs *childregs;
+	unsigned long tos;
+	int user_mode = user_mode(regs);
+
+	/* Set up new TSS. */
+	tos = (unsigned long)p->thread_info + THREAD_SIZE;
+	if (user_mode)
+		childregs = (struct pt_regs*)(tos - PT_USER_SIZE);
+	else
+		childregs = (struct pt_regs*)tos - 1;
+
+	*childregs = *regs;
+
+	/* Create a call4 dummy-frame: a0 = 0, a1 = childregs. */
+	*((int*)childregs - 3) = (unsigned long)childregs;
+	*((int*)childregs - 4) = 0;
+
+	childregs->areg[1] = tos;
+	childregs->areg[2] = 0;
+	p->set_child_tid = p->clear_child_tid = NULL;
+	p->thread.ra = MAKE_RA_FOR_CALL((unsigned long)ret_from_fork, 0x1);
+	p->thread.sp = (unsigned long)childregs;
+	if (user_mode(regs)) {
+
+		int len = childregs->wmask & ~0xf;
+		childregs->areg[1] = usp;
+		memcpy(&childregs->areg[XCHAL_NUM_AREGS - len/4],
+		       &regs->areg[XCHAL_NUM_AREGS - len/4], len);
+
+		if (clone_flags & CLONE_SETTLS)
+			childregs->areg[2] = childregs->areg[6];
+
+	} else {
+		/* In kernel space, we start a new thread with a new stack. */
+		childregs->wmask = 1;
+	}
+	return 0;
+}
+
+
+/*
+ * Create a kernel thread
+ */
+
+int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
+{
+	long retval;
+	__asm__ __volatile__
+		("mov           a5, %4\n\t" /* preserve fn in a5 */
+		 "mov           a6, %3\n\t" /* preserve and setup arg in a6 */
+		 "movi		a2, %1\n\t" /* load __NR_clone for syscall*/
+		 "mov		a3, sp\n\t" /* sp check and sys_clone */
+		 "mov		a4, %5\n\t" /* load flags for syscall */
+		 "syscall\n\t"
+		 "beq		a3, sp, 1f\n\t" /* branch if parent */
+		 "callx4	a5\n\t"     /* call fn */
+		 "movi		a2, %2\n\t" /* load __NR_exit for syscall */
+		 "mov		a3, a6\n\t" /* load fn return value */
+		 "syscall\n"
+		 "1:\n\t"
+		 "mov		%0, a2\n\t" /* parent returns zero */
+		 :"=r" (retval)
+		 :"i" (__NR_clone), "i" (__NR_exit),
+		 "r" (arg), "r" (fn),
+		 "r" (flags | CLONE_VM)
+		 : "a2", "a3", "a4", "a5", "a6" );
+	return retval;
+}
+
+
+/*
+ * These bracket the sleeping functions..
+ */
+
+unsigned long get_wchan(struct task_struct *p)
+{
+	unsigned long sp, pc;
+	unsigned long stack_page = (unsigned long) p->thread_info;
+	int count = 0;
+
+	if (!p || p == current || p->state == TASK_RUNNING)
+		return 0;
+
+	sp = p->thread.sp;
+	pc = MAKE_PC_FROM_RA(p->thread.ra, p->thread.sp);
+
+	do {
+		if (sp < stack_page + sizeof(struct task_struct) ||
+		    sp >= (stack_page + THREAD_SIZE) ||
+		    pc == 0)
+			return 0;
+		if (!in_sched_functions(pc))
+			return pc;
+
+		/* Stack layout: sp-4: ra, sp-3: sp' */
+
+		pc = MAKE_PC_FROM_RA(*(unsigned long*)sp - 4, sp);
+		sp = *(unsigned long *)sp - 3;
+	} while (count++ < 16);
+	return 0;
+}
+
+/*
+ * do_copy_regs() gathers information from 'struct pt_regs' and
+ * 'current->thread.areg[]' to fill in the xtensa_gregset_t
+ * structure.
+ *
+ * xtensa_gregset_t and 'struct pt_regs' are vastly different formats
+ * of processor registers.  Besides different ordering,
+ * xtensa_gregset_t contains non-live register information that
+ * 'struct pt_regs' does not.  Exception handling (primarily) uses
+ * 'struct pt_regs'.  Core files and ptrace use xtensa_gregset_t.
+ *
+ */
+
+void do_copy_regs (xtensa_gregset_t *elfregs, struct pt_regs *regs,
+		   struct task_struct *tsk)
+{
+	int i, n, wb_offset;
+
+	elfregs->xchal_config_id0 = XCHAL_HW_CONFIGID0;
+	elfregs->xchal_config_id1 = XCHAL_HW_CONFIGID1;
+
+	__asm__ __volatile__ ("rsr  %0, 176\n" : "=a" (i));
+ 	elfregs->cpux = i;
+	__asm__ __volatile__ ("rsr  %0, 208\n" : "=a" (i));
+ 	elfregs->cpuy = i;
+
+	/* Note:  PS.EXCM is not set while user task is running; its
+	 * being set in regs->ps is for exception handling convenience.
+	 */
+
+	elfregs->pc		= regs->pc;
+	elfregs->ps		= (regs->ps & ~XCHAL_PS_EXCM_MASK);
+	elfregs->exccause	= regs->exccause;
+	elfregs->excvaddr	= regs->excvaddr;
+	elfregs->windowbase	= regs->windowbase;
+	elfregs->windowstart	= regs->windowstart;
+	elfregs->lbeg		= regs->lbeg;
+	elfregs->lend		= regs->lend;
+	elfregs->lcount		= regs->lcount;
+	elfregs->sar		= regs->sar;
+	elfregs->syscall	= regs->syscall;
+
+	/* Copy register file.
+	 * The layout looks like this:
+	 *
+	 * |  a0 ... a15  | Z ... Z |  arX ... arY  |
+	 *  current window  unused    saved frames
+	 */
+
+	memset (elfregs->ar, 0, sizeof(elfregs->ar));
+
+	wb_offset = regs->windowbase * 4;
+	n = (regs->wmask&1)? 4 : (regs->wmask&2)? 8 : (regs->wmask&4)? 12 : 16;
+
+	for (i = 0; i < n; i++)
+		elfregs->ar[(wb_offset + i) % XCHAL_NUM_AREGS] = regs->areg[i];
+
+	n = (regs->wmask >> 4) * 4;
+
+	for (i = XCHAL_NUM_AREGS - n; n > 0; i++, n--)
+		elfregs->ar[(wb_offset + i) % XCHAL_NUM_AREGS] = regs->areg[i];
+}
+
+void xtensa_elf_core_copy_regs (xtensa_gregset_t *elfregs, struct pt_regs *regs)
+{
+	do_copy_regs ((xtensa_gregset_t *)elfregs, regs, current);
+}
+
+
+/* The inverse of do_copy_regs().  No error or sanity checking. */
+
+void do_restore_regs (xtensa_gregset_t *elfregs, struct pt_regs *regs,
+		      struct task_struct *tsk)
+{
+	int i, n, wb_offset;
+
+	/* Note:  PS.EXCM is not set while user task is running; it
+	 * needs to be set in regs->ps is for exception handling convenience.
+	 */
+
+	regs->pc		= elfregs->pc;
+	regs->ps		= (elfregs->ps | XCHAL_PS_EXCM_MASK);
+	regs->exccause		= elfregs->exccause;
+	regs->excvaddr		= elfregs->excvaddr;
+	regs->windowbase	= elfregs->windowbase;
+	regs->windowstart	= elfregs->windowstart;
+	regs->lbeg		= elfregs->lbeg;
+	regs->lend		= elfregs->lend;
+	regs->lcount		= elfregs->lcount;
+	regs->sar		= elfregs->sar;
+	regs->syscall	= elfregs->syscall;
+
+	/* Clear everything. */
+
+	memset (regs->areg, 0, sizeof(regs->areg));
+
+	/* Copy regs from live window frame. */
+
+	wb_offset = regs->windowbase * 4;
+	n = (regs->wmask&1)? 4 : (regs->wmask&2)? 8 : (regs->wmask&4)? 12 : 16;
+
+	for (i = 0; i < n; i++)
+		regs->areg[(wb_offset+i) % XCHAL_NUM_AREGS] = elfregs->ar[i];
+
+	n = (regs->wmask >> 4) * 4;
+
+	for (i = XCHAL_NUM_AREGS - n; n > 0; i++, n--)
+		regs->areg[(wb_offset+i) % XCHAL_NUM_AREGS] = elfregs->ar[i];
+}
+
+/*
+ * do_save_fpregs() gathers information from 'struct pt_regs' and
+ * 'current->thread' to fill in the elf_fpregset_t structure.
+ *
+ * Core files and ptrace use elf_fpregset_t.
+ */
+
+void do_save_fpregs (elf_fpregset_t *fpregs, struct pt_regs *regs,
+		     struct task_struct *tsk)
+{
+#if XCHAL_HAVE_CP
+
+	extern unsigned char	_xtensa_reginfo_tables[];
+	extern unsigned		_xtensa_reginfo_table_size;
+	int i;
+	unsigned long flags;
+
+	/* Before dumping coprocessor state from memory,
+	 * ensure any live coprocessor contents for this
+	 * task are first saved to memory:
+	 */
+	local_irq_save(flags);
+
+	for (i = 0; i < XCHAL_CP_MAX; i++) {
+		if (tsk == coprocessor_info[i].owner) {
+			enable_coprocessor(i);
+			save_coprocessor_registers(
+			    tsk->thread.cp_save+coprocessor_info[i].offset,i);
+			disable_coprocessor(i);
+		}
+	}
+
+	local_irq_restore(flags);
+
+	/* Now dump coprocessor & extra state: */
+	memcpy((unsigned char*)fpregs,
+		_xtensa_reginfo_tables, _xtensa_reginfo_table_size);
+	memcpy((unsigned char*)fpregs + _xtensa_reginfo_table_size,
+		tsk->thread.cp_save, XTENSA_CP_EXTRA_SIZE);
+#endif
+}
+
+/*
+ * The inverse of do_save_fpregs().
+ * Copies coprocessor and extra state from fpregs into regs and tsk->thread.
+ * Returns 0 on success, non-zero if layout doesn't match.
+ */
+
+int  do_restore_fpregs (elf_fpregset_t *fpregs, struct pt_regs *regs,
+		        struct task_struct *tsk)
+{
+#if XCHAL_HAVE_CP
+
+	extern unsigned char	_xtensa_reginfo_tables[];
+	extern unsigned		_xtensa_reginfo_table_size;
+	int i;
+	unsigned long flags;
+
+	/* Make sure save area layouts match.
+	 * FIXME:  in the future we could allow restoring from
+	 * a different layout of the same registers, by comparing
+	 * fpregs' table with _xtensa_reginfo_tables and matching
+	 * entries and copying registers one at a time.
+	 * Not too sure yet whether that's very useful.
+	 */
+
+	if( memcmp((unsigned char*)fpregs,
+		_xtensa_reginfo_tables, _xtensa_reginfo_table_size) ) {
+	    return -1;
+	}
+
+	/* Before restoring coprocessor state from memory,
+	 * ensure any live coprocessor contents for this
+	 * task are first invalidated.
+	 */
+
+	local_irq_save(flags);
+
+	for (i = 0; i < XCHAL_CP_MAX; i++) {
+		if (tsk == coprocessor_info[i].owner) {
+			enable_coprocessor(i);
+			save_coprocessor_registers(
+			    tsk->thread.cp_save+coprocessor_info[i].offset,i);
+			coprocessor_info[i].owner = 0;
+			disable_coprocessor(i);
+		}
+	}
+
+	local_irq_restore(flags);
+
+	/*  Now restore coprocessor & extra state:  */
+
+	memcpy(tsk->thread.cp_save,
+		(unsigned char*)fpregs + _xtensa_reginfo_table_size,
+		XTENSA_CP_EXTRA_SIZE);
+#endif
+	return 0;
+}
+/*
+ * Fill in the CP structure for a core dump for a particular task.
+ */
+
+int
+dump_task_fpu(struct pt_regs *regs, struct task_struct *task, elf_fpregset_t *r)
+{
+/* see asm/coprocessor.h for this magic number 16 */
+#if TOTAL_CPEXTRA_SIZE > 16
+	do_save_fpregs (r, regs, task);
+
+	/*  For now, bit 16 means some extra state may be present:  */
+// FIXME!! need to track to return more accurate mask
+	return 0x10000 | XCHAL_CP_MASK;
+#else
+	return 0;	/* no coprocessors active on this processor */
+#endif
+}
+
+/*
+ * Fill in the CP structure for a core dump.
+ * This includes any FPU coprocessor.
+ * Here, we dump all coprocessors, and other ("extra") custom state.
+ *
+ * This function is called by elf_core_dump() in fs/binfmt_elf.c
+ * (in which case 'regs' comes from calls to do_coredump, see signals.c).
+ */
+int  dump_fpu(struct pt_regs *regs, elf_fpregset_t *r)
+{
+	return dump_task_fpu(regs, current, r);
+}
diff --git a/arch/xtensa/kernel/ptrace.c b/arch/xtensa/kernel/ptrace.c
new file mode 100644
index 000000000000..9ef07a4dd2a2
--- /dev/null
+++ b/arch/xtensa/kernel/ptrace.c
@@ -0,0 +1,407 @@
+// TODO some minor issues
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005  Tensilica Inc.
+ *
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ * Chris Zankel <chris@zankel.net>
+ * Scott Foehner<sfoehner@yahoo.com>,
+ * Kevin Chea
+ * Marc Gauthier<marc@tensilica.com> <marc@alumni.uwaterloo.ca>
+ */
+
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/errno.h>
+#include <linux/ptrace.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/security.h>
+
+#include <asm/pgtable.h>
+#include <asm/page.h>
+#include <asm/system.h>
+#include <asm/uaccess.h>
+#include <asm/ptrace.h>
+#include <asm/elf.h>
+
+#define TEST_KERNEL	// verify kernel operations FIXME: remove
+
+
+/*
+ * Called by kernel/ptrace.c when detaching..
+ *
+ * Make sure single step bits etc are not set.
+ */
+
+void ptrace_disable(struct task_struct *child)
+{
+	/* Nothing to do.. */
+}
+
+int sys_ptrace(long request, long pid, long addr, long data)
+{
+	struct task_struct *child;
+	int ret = -EPERM;
+
+	lock_kernel();
+
+#if 0
+	if ((int)request != 1)
+	printk("ptrace(r=%d,pid=%d,addr=%08lx,data=%08lx)\n",
+	       (int) request, (int) pid, (unsigned long) addr,
+	       (unsigned long) data);
+#endif
+
+	if (request == PTRACE_TRACEME) {
+
+		/* Are we already being traced? */
+
+		if (current->ptrace & PT_PTRACED)
+			goto out;
+
+		if ((ret = security_ptrace(current->parent, current)))
+			goto out;
+
+		/* Set the ptrace bit in the process flags. */
+
+		current->ptrace |= PT_PTRACED;
+		ret = 0;
+		goto out;
+	}
+
+	ret = -ESRCH;
+	read_lock(&tasklist_lock);
+	child = find_task_by_pid(pid);
+	if (child)
+		get_task_struct(child);
+	read_unlock(&tasklist_lock);
+	if (!child)
+		goto out;
+
+	ret = -EPERM;
+	if (pid == 1)		/* you may not mess with init */
+		goto out;
+
+	if (request == PTRACE_ATTACH) {
+		ret = ptrace_attach(child);
+		goto out_tsk;
+	}
+
+	if ((ret = ptrace_check_attach(child, request == PTRACE_KILL)) < 0)
+		goto out_tsk;
+
+	switch (request) {
+	case PTRACE_PEEKTEXT: /* read word at location addr. */
+	case PTRACE_PEEKDATA:
+	{
+		unsigned long tmp;
+		int copied;
+
+		copied = access_process_vm(child, addr, &tmp, sizeof(tmp), 0);
+		ret = -EIO;
+		if (copied != sizeof(tmp))
+			break;
+		ret = put_user(tmp,(unsigned long *) data);
+
+		goto out;
+	}
+
+	/* Read the word at location addr in the USER area.  */
+
+	case PTRACE_PEEKUSR:
+		{
+		struct pt_regs *regs;
+		unsigned long tmp;
+
+		regs = xtensa_pt_regs(child);
+		tmp = 0;  /* Default return value. */
+
+		switch(addr) {
+
+		case REG_AR_BASE ... REG_AR_BASE + XCHAL_NUM_AREGS - 1:
+			{
+			int ar = addr - REG_AR_BASE - regs->windowbase * 4;
+			ar &= (XCHAL_NUM_AREGS - 1);
+			if (ar < 16 && ar + (regs->wmask >> 4) * 4 >= 0)
+				tmp = regs->areg[ar];
+			else
+				ret = -EIO;
+			break;
+			}
+		case REG_A_BASE ... REG_A_BASE + 15:
+			tmp = regs->areg[addr - REG_A_BASE];
+			break;
+		case REG_PC:
+			tmp = regs->pc;
+			break;
+		case REG_PS:
+			/* Note:  PS.EXCM is not set while user task is running;
+			 * its being set in regs is for exception handling
+			 * convenience.  */
+			tmp = (regs->ps & ~XCHAL_PS_EXCM_MASK);
+			break;
+		case REG_WB:
+			tmp = regs->windowbase;
+			break;
+		case REG_WS:
+			tmp = regs->windowstart;
+			break;
+		case REG_LBEG:
+			tmp = regs->lbeg;
+			break;
+		case REG_LEND:
+			tmp = regs->lend;
+			break;
+		case REG_LCOUNT:
+			tmp = regs->lcount;
+			break;
+		case REG_SAR:
+			tmp = regs->sar;
+			break;
+		case REG_DEPC:
+			tmp = regs->depc;
+			break;
+		case REG_EXCCAUSE:
+			tmp = regs->exccause;
+			break;
+		case REG_EXCVADDR:
+			tmp = regs->excvaddr;
+			break;
+		case SYSCALL_NR:
+			tmp = regs->syscall;
+			break;
+		default:
+			tmp = 0;
+			ret = -EIO;
+			goto out;
+		}
+		ret = put_user(tmp, (unsigned long *) data);
+		goto out;
+		}
+
+	case PTRACE_POKETEXT: /* write the word at location addr. */
+	case PTRACE_POKEDATA:
+		if (access_process_vm(child, addr, &data, sizeof(data), 1)
+		    == sizeof(data))
+			break;
+		ret = -EIO;
+		goto out;
+
+	case PTRACE_POKEUSR:
+		{
+		struct pt_regs *regs;
+		regs = xtensa_pt_regs(child);
+
+		switch (addr) {
+		case REG_AR_BASE ... REG_AR_BASE + XCHAL_NUM_AREGS - 1:
+			{
+			int ar = addr - REG_AR_BASE - regs->windowbase * 4;
+			if (ar < 16 && ar + (regs->wmask >> 4) * 4 >= 0)
+				regs->areg[ar & (XCHAL_NUM_AREGS - 1)] = data;
+			else
+				ret = -EIO;
+			break;
+			}
+		case REG_A_BASE ... REG_A_BASE + 15:
+			regs->areg[addr - REG_A_BASE] = data;
+			break;
+		case REG_PC:
+			regs->pc = data;
+			break;
+		case SYSCALL_NR:
+			regs->syscall = data;
+			break;
+#ifdef TEST_KERNEL
+		case REG_WB:
+			regs->windowbase = data;
+			break;
+		case REG_WS:
+			regs->windowstart = data;
+			break;
+#endif
+
+		default:
+			/* The rest are not allowed. */
+			ret = -EIO;
+			break;
+		}
+		break;
+		}
+
+	/* continue and stop at next (return from) syscall */
+	case PTRACE_SYSCALL:
+	case PTRACE_CONT: /* restart after signal. */
+	{
+		ret = -EIO;
+		if ((unsigned long) data > _NSIG)
+			break;
+		if (request == PTRACE_SYSCALL)
+			set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+		else
+			clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+		child->exit_code = data;
+		/* Make sure the single step bit is not set. */
+		child->ptrace &= ~PT_SINGLESTEP;
+		wake_up_process(child);
+		ret = 0;
+		break;
+	}
+
+	/*
+	 * make the child exit.  Best I can do is send it a sigkill.
+	 * perhaps it should be put in the status that it wants to
+	 * exit.
+	 */
+	case PTRACE_KILL:
+		ret = 0;
+		if (child->state == EXIT_ZOMBIE)	/* already dead */
+			break;
+		child->exit_code = SIGKILL;
+		child->ptrace &= ~PT_SINGLESTEP;
+		wake_up_process(child);
+		break;
+
+	case PTRACE_SINGLESTEP:
+		ret = -EIO;
+		if ((unsigned long) data > _NSIG)
+			break;
+		clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+		child->ptrace |= PT_SINGLESTEP;
+		child->exit_code = data;
+		wake_up_process(child);
+		ret = 0;
+		break;
+
+	case PTRACE_GETREGS:
+	{
+		/* 'data' points to user memory in which to write.
+		 * Mainly due to the non-live register values, we
+		 * reformat the register values into something more
+		 * standard.  For convenience, we use the handy
+		 * elf_gregset_t format. */
+
+		xtensa_gregset_t format;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		do_copy_regs (&format, regs, child);
+
+		/* Now, copy to user space nice and easy... */
+		ret = 0;
+		if (copy_to_user((void *)data, &format, sizeof(elf_gregset_t)))
+			ret = -EFAULT;
+		break;
+	}
+
+	case PTRACE_SETREGS:
+	{
+		/* 'data' points to user memory that contains the new
+		 * values in the elf_gregset_t format. */
+
+		xtensa_gregset_t format;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		if (copy_from_user(&format,(void *)data,sizeof(elf_gregset_t))){
+			ret = -EFAULT;
+			break;
+		}
+
+		/* FIXME: Perhaps we want some sanity checks on
+		 * these user-space values?  See ARM version.  Are
+		 * debuggers a security concern? */
+
+		do_restore_regs (&format, regs, child);
+
+		ret = 0;
+		break;
+	}
+
+	case PTRACE_GETFPREGS:
+	{
+		/* 'data' points to user memory in which to write.
+		 * For convenience, we use the handy
+		 * elf_fpregset_t format. */
+
+		elf_fpregset_t fpregs;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		do_save_fpregs (&fpregs, regs, child);
+
+		/* Now, copy to user space nice and easy... */
+		ret = 0;
+		if (copy_to_user((void *)data, &fpregs, sizeof(elf_fpregset_t)))
+			ret = -EFAULT;
+
+		break;
+	}
+
+	case PTRACE_SETFPREGS:
+	{
+		/* 'data' points to user memory that contains the new
+		 * values in the elf_fpregset_t format.
+		 */
+		elf_fpregset_t fpregs;
+		struct pt_regs *regs = xtensa_pt_regs(child);
+
+		ret = 0;
+		if (copy_from_user(&fpregs, (void *)data, sizeof(elf_fpregset_t))) {
+			ret = -EFAULT;
+			break;
+		}
+
+		if (do_restore_fpregs (&fpregs, regs, child))
+			ret = -EIO;
+		break;
+	}
+
+	case PTRACE_GETFPREGSIZE:
+		/* 'data' points to 'unsigned long' set to the size
+		 * of elf_fpregset_t
+		 */
+		ret = put_user(sizeof(elf_fpregset_t), (unsigned long *) data);
+		break;
+
+	case PTRACE_DETACH: /* detach a process that was attached. */
+		ret = ptrace_detach(child, data);
+		break;
+
+	default:
+		ret = ptrace_request(child, request, addr, data);
+		goto out;
+	}
+out_tsk:
+	put_task_struct(child);
+out:
+	unlock_kernel();
+	return ret;
+}
+
+void do_syscall_trace(void)
+{
+	if (!test_thread_flag(TIF_SYSCALL_TRACE))
+		return;
+
+	if (!(current->ptrace & PT_PTRACED))
+		return;
+
+	/*
+	 * The 0x80 provides a way for the tracing parent to distinguish
+	 * between a syscall stop and SIGTRAP delivery
+	 */
+	ptrace_notify(SIGTRAP|((current->ptrace & PT_TRACESYSGOOD) ? 0x80 : 0));
+
+	/*
+	 * this isn't the same as continuing with a signal, but it will do
+	 * for normal use.  strace only continues with a signal if the
+	 * stopping signal is not SIGTRAP.  -brl
+	 */
+	if (current->exit_code) {
+		send_sig(current->exit_code, current, 1);
+		current->exit_code = 0;
+	}
+}
diff --git a/arch/xtensa/kernel/semaphore.c b/arch/xtensa/kernel/semaphore.c
new file mode 100644
index 000000000000..d40f4b1b75ac
--- /dev/null
+++ b/arch/xtensa/kernel/semaphore.c
@@ -0,0 +1,226 @@
+/*
+ * arch/xtensa/kernel/semaphore.c
+ *
+ * Generic semaphore code. Buyer beware. Do your own specific changes
+ * in <asm/semaphore-helper.h>
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ * Chris Zankel	<chris@zankel.net>
+ * Marc Gauthier<marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Kevin Chea
+ */
+
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/init.h>
+#include <asm/semaphore.h>
+#include <asm/errno.h>
+
+/*
+ * These two _must_ execute atomically wrt each other.
+ */
+
+static __inline__ void wake_one_more(struct semaphore * sem)
+{
+	atomic_inc((atomic_t *)&sem->sleepers);
+}
+
+static __inline__ int waking_non_zero(struct semaphore *sem)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&semaphore_wake_lock, flags);
+	if (sem->sleepers > 0) {
+		sem->sleepers--;
+		ret = 1;
+	}
+	spin_unlock_irqrestore(&semaphore_wake_lock, flags);
+	return ret;
+}
+
+/*
+ * waking_non_zero_interruptible:
+ *	1	got the lock
+ *	0	go to sleep
+ *	-EINTR	interrupted
+ *
+ * We must undo the sem->count down_interruptible() increment while we are
+ * protected by the spinlock in order to make atomic this atomic_inc() with the
+ * atomic_read() in wake_one_more(), otherwise we can race. -arca
+ */
+
+static __inline__ int waking_non_zero_interruptible(struct semaphore *sem,
+						struct task_struct *tsk)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&semaphore_wake_lock, flags);
+	if (sem->sleepers > 0) {
+		sem->sleepers--;
+		ret = 1;
+	} else if (signal_pending(tsk)) {
+		atomic_inc(&sem->count);
+		ret = -EINTR;
+	}
+	spin_unlock_irqrestore(&semaphore_wake_lock, flags);
+	return ret;
+}
+
+/*
+ * waking_non_zero_trylock:
+ *	1	failed to lock
+ *	0	got the lock
+ *
+ * We must undo the sem->count down_trylock() increment while we are
+ * protected by the spinlock in order to make atomic this atomic_inc() with the
+ * atomic_read() in wake_one_more(), otherwise we can race. -arca
+ */
+
+static __inline__ int waking_non_zero_trylock(struct semaphore *sem)
+{
+	unsigned long flags;
+	int ret = 1;
+
+	spin_lock_irqsave(&semaphore_wake_lock, flags);
+	if (sem->sleepers <= 0)
+		atomic_inc(&sem->count);
+	else {
+		sem->sleepers--;
+		ret = 0;
+	}
+	spin_unlock_irqrestore(&semaphore_wake_lock, flags);
+	return ret;
+}
+
+spinlock_t semaphore_wake_lock;
+
+/*
+ * Semaphores are implemented using a two-way counter:
+ * The "count" variable is decremented for each process
+ * that tries to sleep, while the "waking" variable is
+ * incremented when the "up()" code goes to wake up waiting
+ * processes.
+ *
+ * Notably, the inline "up()" and "down()" functions can
+ * efficiently test if they need to do any extra work (up
+ * needs to do something only if count was negative before
+ * the increment operation.
+ *
+ * waking_non_zero() (from asm/semaphore.h) must execute
+ * atomically.
+ *
+ * When __up() is called, the count was negative before
+ * incrementing it, and we need to wake up somebody.
+ *
+ * This routine adds one to the count of processes that need to
+ * wake up and exit.  ALL waiting processes actually wake up but
+ * only the one that gets to the "waking" field first will gate
+ * through and acquire the semaphore.  The others will go back
+ * to sleep.
+ *
+ * Note that these functions are only called when there is
+ * contention on the lock, and as such all this is the
+ * "non-critical" part of the whole semaphore business. The
+ * critical part is the inline stuff in <asm/semaphore.h>
+ * where we want to avoid any extra jumps and calls.
+ */
+
+void __up(struct semaphore *sem)
+{
+	wake_one_more(sem);
+	wake_up(&sem->wait);
+}
+
+/*
+ * Perform the "down" function.  Return zero for semaphore acquired,
+ * return negative for signalled out of the function.
+ *
+ * If called from __down, the return is ignored and the wait loop is
+ * not interruptible.  This means that a task waiting on a semaphore
+ * using "down()" cannot be killed until someone does an "up()" on
+ * the semaphore.
+ *
+ * If called from __down_interruptible, the return value gets checked
+ * upon return.  If the return value is negative then the task continues
+ * with the negative value in the return register (it can be tested by
+ * the caller).
+ *
+ * Either form may be used in conjunction with "up()".
+ *
+ */
+
+#define DOWN_VAR				\
+	struct task_struct *tsk = current;	\
+	wait_queue_t wait;			\
+	init_waitqueue_entry(&wait, tsk);
+
+#define DOWN_HEAD(task_state)						\
+									\
+									\
+	tsk->state = (task_state);					\
+	add_wait_queue(&sem->wait, &wait);				\
+									\
+	/*								\
+	 * Ok, we're set up.  sem->count is known to be less than zero	\
+	 * so we must wait.						\
+	 *								\
+	 * We can let go the lock for purposes of waiting.		\
+	 * We re-acquire it after awaking so as to protect		\
+	 * all semaphore operations.					\
+	 *								\
+	 * If "up()" is called before we call waking_non_zero() then	\
+	 * we will catch it right away.  If it is called later then	\
+	 * we will have to go through a wakeup cycle to catch it.	\
+	 *								\
+	 * Multiple waiters contend for the semaphore lock to see	\
+	 * who gets to gate through and who has to wait some more.	\
+	 */								\
+	for (;;) {
+
+#define DOWN_TAIL(task_state)			\
+		tsk->state = (task_state);	\
+	}					\
+	tsk->state = TASK_RUNNING;		\
+	remove_wait_queue(&sem->wait, &wait);
+
+void __sched __down(struct semaphore * sem)
+{
+	DOWN_VAR
+	DOWN_HEAD(TASK_UNINTERRUPTIBLE)
+	if (waking_non_zero(sem))
+		break;
+	schedule();
+	DOWN_TAIL(TASK_UNINTERRUPTIBLE)
+}
+
+int __sched __down_interruptible(struct semaphore * sem)
+{
+	int ret = 0;
+	DOWN_VAR
+	DOWN_HEAD(TASK_INTERRUPTIBLE)
+
+	ret = waking_non_zero_interruptible(sem, tsk);
+	if (ret)
+	{
+		if (ret == 1)
+			/* ret != 0 only if we get interrupted -arca */
+			ret = 0;
+		break;
+	}
+	schedule();
+	DOWN_TAIL(TASK_INTERRUPTIBLE)
+	return ret;
+}
+
+int __down_trylock(struct semaphore * sem)
+{
+	return waking_non_zero_trylock(sem);
+}
diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c
new file mode 100644
index 000000000000..1f5bf5d624e4
--- /dev/null
+++ b/arch/xtensa/kernel/setup.c
@@ -0,0 +1,520 @@
+/*
+ * arch/xtensa/setup.c
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1995  Linus Torvalds
+ * Copyright (C) 2001 - 2005  Tensilica Inc.
+ *
+ * Chris Zankel	<chris@zankel.net>
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ * Kevin Chea
+ * Marc Gauthier<marc@tensilica.com> <marc@alumni.uwaterloo.ca>
+ */
+
+#include <linux/config.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/proc_fs.h>
+#include <linux/tty.h>
+#include <linux/bootmem.h>
+#include <linux/kernel.h>
+
+#if defined(CONFIG_VGA_CONSOLE) || defined(CONFIG_DUMMY_CONSOLE)
+# include <linux/console.h>
+#endif
+
+#ifdef CONFIG_RTC
+# include <linux/timex.h>
+#endif
+
+#ifdef CONFIG_PROC_FS
+# include <linux/seq_file.h>
+#endif
+
+#include <asm/system.h>
+#include <asm/bootparam.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/timex.h>
+#include <asm/platform.h>
+#include <asm/page.h>
+#include <asm/setup.h>
+
+#include <xtensa/config/system.h>
+
+#if defined(CONFIG_VGA_CONSOLE) || defined(CONFIG_DUMMY_CONSOLE)
+struct screen_info screen_info = { 0, 24, 0, 0, 0, 80, 0, 0, 0, 24, 1, 16};
+#endif
+
+#ifdef CONFIG_BLK_DEV_FD
+extern struct fd_ops no_fd_ops;
+struct fd_ops *fd_ops;
+#endif
+
+#if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_IDE_MODULE)
+extern struct ide_ops no_ide_ops;
+struct ide_ops *ide_ops;
+#endif
+
+extern struct rtc_ops no_rtc_ops;
+struct rtc_ops *rtc_ops;
+
+#ifdef CONFIG_PC_KEYB
+extern struct kbd_ops no_kbd_ops;
+struct kbd_ops *kbd_ops;
+#endif
+
+#ifdef CONFIG_BLK_DEV_INITRD
+extern void *initrd_start;
+extern void *initrd_end;
+extern void *__initrd_start;
+extern void *__initrd_end;
+int initrd_is_mapped = 0;
+extern int initrd_below_start_ok;
+#endif
+
+unsigned char aux_device_present;
+extern unsigned long loops_per_jiffy;
+
+/* Command line specified as configuration option. */
+
+static char command_line[COMMAND_LINE_SIZE];
+
+#ifdef CONFIG_CMDLINE_BOOL
+static char default_command_line[COMMAND_LINE_SIZE] __initdata = CONFIG_CMDLINE;
+#endif
+
+sysmem_info_t __initdata sysmem;
+
+#ifdef CONFIG_BLK_DEV_INITRD
+int initrd_is_mapped;
+#endif
+
+extern void init_mmu(void);
+
+/*
+ * Boot parameter parsing.
+ *
+ * The Xtensa port uses a list of variable-sized tags to pass data to
+ * the kernel. The first tag must be a BP_TAG_FIRST tag for the list
+ * to be recognised. The list is terminated with a zero-sized
+ * BP_TAG_LAST tag.
+ */
+
+typedef struct tagtable {
+	u32 tag;
+	int (*parse)(const bp_tag_t*);
+} tagtable_t;
+
+#define __tagtable(tag, fn) static tagtable_t __tagtable_##fn 		\
+	__attribute__((unused, __section__(".taglist"))) = { tag, fn }
+
+/* parse current tag */
+
+static int __init parse_tag_mem(const bp_tag_t *tag)
+{
+	meminfo_t *mi = (meminfo_t*)(tag->data);
+
+	if (mi->type != MEMORY_TYPE_CONVENTIONAL)
+		return -1;
+
+	if (sysmem.nr_banks >= SYSMEM_BANKS_MAX) {
+		printk(KERN_WARNING
+		       "Ignoring memory bank 0x%08lx size %ldKB\n",
+		       (unsigned long)mi->start,
+		       (unsigned long)mi->end - (unsigned long)mi->start);
+		return -EINVAL;
+	}
+	sysmem.bank[sysmem.nr_banks].type  = mi->type;
+	sysmem.bank[sysmem.nr_banks].start = PAGE_ALIGN(mi->start);
+	sysmem.bank[sysmem.nr_banks].end   = mi->end & PAGE_SIZE;
+	sysmem.nr_banks++;
+
+	return 0;
+}
+
+__tagtable(BP_TAG_MEMORY, parse_tag_mem);
+
+#ifdef CONFIG_BLK_DEV_INITRD
+
+static int __init parse_tag_initrd(const bp_tag_t* tag)
+{
+	meminfo_t* mi;
+	mi = (meminfo_t*)(tag->data);
+	initrd_start = (void*)(mi->start);
+	initrd_end = (void*)(mi->end);
+
+	return 0;
+}
+
+__tagtable(BP_TAG_INITRD, parse_tag_initrd);
+
+#endif /* CONFIG_BLK_DEV_INITRD */
+
+static int __init parse_tag_cmdline(const bp_tag_t* tag)
+{
+	strncpy(command_line, (char*)(tag->data), COMMAND_LINE_SIZE);
+	command_line[COMMAND_LINE_SIZE - 1] = '\0';
+	return 0;
+}
+
+__tagtable(BP_TAG_COMMAND_LINE, parse_tag_cmdline);
+
+static int __init parse_bootparam(const bp_tag_t* tag)
+{
+	extern tagtable_t __tagtable_begin, __tagtable_end;
+	tagtable_t *t;
+
+	/* Boot parameters must start with a BP_TAG_FIRST tag. */
+
+	if (tag->id != BP_TAG_FIRST) {
+		printk(KERN_WARNING "Invalid boot parameters!\n");
+		return 0;
+	}
+
+	tag = (bp_tag_t*)((unsigned long)tag + sizeof(bp_tag_t) + tag->size);
+
+	/* Parse all tags. */
+
+	while (tag != NULL && tag->id != BP_TAG_LAST) {
+	 	for (t = &__tagtable_begin; t < &__tagtable_end; t++) {
+			if (tag->id == t->tag) {
+				t->parse(tag);
+				break;
+			}
+		}
+		if (t == &__tagtable_end)
+			printk(KERN_WARNING "Ignoring tag "
+			       "0x%08x\n", tag->id);
+		tag = (bp_tag_t*)((unsigned long)(tag + 1) + tag->size);
+	}
+
+	return 0;
+}
+
+/*
+ * Initialize architecture. (Early stage)
+ */
+
+void __init init_arch(bp_tag_t *bp_start)
+{
+
+#ifdef CONFIG_BLK_DEV_INITRD
+	initrd_start = &__initrd_start;
+	initrd_end = &__initrd_end;
+#endif
+
+	sysmem.nr_banks = 0;
+
+#ifdef CONFIG_CMDLINE_BOOL
+	strcpy(command_line, default_command_line);
+#endif
+
+	/* Parse boot parameters */
+
+        if (bp_start)
+	  parse_bootparam(bp_start);
+
+	if (sysmem.nr_banks == 0) {
+		sysmem.nr_banks = 1;
+		sysmem.bank[0].start = PLATFORM_DEFAULT_MEM_START;
+		sysmem.bank[0].end = PLATFORM_DEFAULT_MEM_START
+				     + PLATFORM_DEFAULT_MEM_SIZE;
+	}
+
+	/* Early hook for platforms */
+
+	platform_init(bp_start);
+
+	/* Initialize MMU. */
+
+	init_mmu();
+}
+
+/*
+ * Initialize system. Setup memory and reserve regions.
+ */
+
+extern char _end;
+extern char _stext;
+extern char _WindowVectors_text_start;
+extern char _WindowVectors_text_end;
+extern char _DebugInterruptVector_literal_start;
+extern char _DebugInterruptVector_text_end;
+extern char _KernelExceptionVector_literal_start;
+extern char _KernelExceptionVector_text_end;
+extern char _UserExceptionVector_literal_start;
+extern char _UserExceptionVector_text_end;
+extern char _DoubleExceptionVector_literal_start;
+extern char _DoubleExceptionVector_text_end;
+
+void __init setup_arch(char **cmdline_p)
+{
+	extern int mem_reserve(unsigned long, unsigned long, int);
+	extern void bootmem_init(void);
+
+	memcpy(saved_command_line, command_line, COMMAND_LINE_SIZE);
+	saved_command_line[COMMAND_LINE_SIZE-1] = '\0';
+	*cmdline_p = command_line;
+
+	/* Reserve some memory regions */
+
+#ifdef CONFIG_BLK_DEV_INITRD
+	if (initrd_start < initrd_end) {
+		initrd_is_mapped = mem_reserve(__pa(initrd_start),
+					       __pa(initrd_end), 0);
+		initrd_below_start_ok = 1;
+ 	} else {
+		initrd_start = 0;
+	}
+#endif
+
+	mem_reserve(__pa(&_stext),__pa(&_end), 1);
+
+	mem_reserve(__pa(&_WindowVectors_text_start),
+		    __pa(&_WindowVectors_text_end), 0);
+
+	mem_reserve(__pa(&_DebugInterruptVector_literal_start),
+		    __pa(&_DebugInterruptVector_text_end), 0);
+
+	mem_reserve(__pa(&_KernelExceptionVector_literal_start),
+		    __pa(&_KernelExceptionVector_text_end), 0);
+
+	mem_reserve(__pa(&_UserExceptionVector_literal_start),
+		    __pa(&_UserExceptionVector_text_end), 0);
+
+	mem_reserve(__pa(&_DoubleExceptionVector_literal_start),
+		    __pa(&_DoubleExceptionVector_text_end), 0);
+
+	bootmem_init();
+
+	platform_setup(cmdline_p);
+
+
+	paging_init();
+
+#ifdef CONFIG_VT
+# if defined(CONFIG_VGA_CONSOLE)
+	conswitchp = &vga_con;
+# elif defined(CONFIG_DUMMY_CONSOLE)
+	conswitchp = &dummy_con;
+# endif
+#endif
+
+#if CONFIG_PCI
+	platform_pcibios_init();
+#endif
+}
+
+void machine_restart(char * cmd)
+{
+	platform_restart();
+}
+
+void machine_halt(void)
+{
+	platform_halt();
+	while (1);
+}
+
+void machine_power_off(void)
+{
+	platform_power_off();
+	while (1);
+}
+#ifdef CONFIG_PROC_FS
+
+/*
+ * Display some core information through /proc/cpuinfo.
+ */
+
+static int
+c_show(struct seq_file *f, void *slot)
+{
+	/* high-level stuff */
+	seq_printf(f,"processor\t: 0\n"
+		     "vendor_id\t: Tensilica\n"
+		     "model\t\t: Xtensa " XCHAL_HW_RELEASE_NAME "\n"
+		     "core ID\t\t: " XCHAL_CORE_ID "\n"
+		     "build ID\t: 0x%x\n"
+		     "byte order\t: %s\n"
+ 		     "cpu MHz\t\t: %lu.%02lu\n"
+		     "bogomips\t: %lu.%02lu\n",
+		     XCHAL_BUILD_UNIQUE_ID,
+		     XCHAL_HAVE_BE ?  "big" : "little",
+		     CCOUNT_PER_JIFFY/(1000000/HZ),
+		     (CCOUNT_PER_JIFFY/(10000/HZ)) % 100,
+		     loops_per_jiffy/(500000/HZ),
+		     (loops_per_jiffy/(5000/HZ)) % 100);
+
+	seq_printf(f,"flags\t\t: "
+#if XCHAL_HAVE_NMI
+		     "nmi "
+#endif
+#if XCHAL_HAVE_DEBUG
+		     "debug "
+# if XCHAL_HAVE_OCD
+		     "ocd "
+# endif
+#endif
+#if XCHAL_HAVE_DENSITY
+	    	     "density "
+#endif
+#if XCHAL_HAVE_BOOLEANS
+		     "boolean "
+#endif
+#if XCHAL_HAVE_LOOPS
+		     "loop "
+#endif
+#if XCHAL_HAVE_NSA
+		     "nsa "
+#endif
+#if XCHAL_HAVE_MINMAX
+		     "minmax "
+#endif
+#if XCHAL_HAVE_SEXT
+		     "sext "
+#endif
+#if XCHAL_HAVE_CLAMPS
+		     "clamps "
+#endif
+#if XCHAL_HAVE_MAC16
+		     "mac16 "
+#endif
+#if XCHAL_HAVE_MUL16
+		     "mul16 "
+#endif
+#if XCHAL_HAVE_MUL32
+		     "mul32 "
+#endif
+#if XCHAL_HAVE_MUL32_HIGH
+		     "mul32h "
+#endif
+#if XCHAL_HAVE_FP
+		     "fpu "
+#endif
+		     "\n");
+
+	/* Registers. */
+	seq_printf(f,"physical aregs\t: %d\n"
+		     "misc regs\t: %d\n"
+		     "ibreak\t\t: %d\n"
+		     "dbreak\t\t: %d\n",
+		     XCHAL_NUM_AREGS,
+		     XCHAL_NUM_MISC_REGS,
+		     XCHAL_NUM_IBREAK,
+		     XCHAL_NUM_DBREAK);
+
+
+	/* Interrupt. */
+	seq_printf(f,"num ints\t: %d\n"
+		     "ext ints\t: %d\n"
+		     "int levels\t: %d\n"
+		     "timers\t\t: %d\n"
+		     "debug level\t: %d\n",
+		     XCHAL_NUM_INTERRUPTS,
+		     XCHAL_NUM_EXTINTERRUPTS,
+		     XCHAL_NUM_INTLEVELS,
+		     XCHAL_NUM_TIMERS,
+		     XCHAL_DEBUGLEVEL);
+
+	/* Coprocessors */
+#if XCHAL_HAVE_CP
+	seq_printf(f, "coprocessors\t: %d\n", XCHAL_CP_NUM);
+#else
+	seq_printf(f, "coprocessors\t: none\n");
+#endif
+
+	/* {I,D}{RAM,ROM} and XLMI */
+	seq_printf(f,"inst ROMs\t: %d\n"
+		     "inst RAMs\t: %d\n"
+		     "data ROMs\t: %d\n"
+		     "data RAMs\t: %d\n"
+		     "XLMI ports\t: %d\n",
+		     XCHAL_NUM_IROM,
+		     XCHAL_NUM_IRAM,
+		     XCHAL_NUM_DROM,
+		     XCHAL_NUM_DRAM,
+		     XCHAL_NUM_XLMI);
+
+	/* Cache */
+	seq_printf(f,"icache line size: %d\n"
+		     "icache ways\t: %d\n"
+		     "icache size\t: %d\n"
+		     "icache flags\t: "
+#if XCHAL_ICACHE_LINE_LOCKABLE
+		     "lock"
+#endif
+		     "\n"
+		     "dcache line size: %d\n"
+		     "dcache ways\t: %d\n"
+		     "dcache size\t: %d\n"
+		     "dcache flags\t: "
+#if XCHAL_DCACHE_IS_WRITEBACK
+		     "writeback"
+#endif
+#if XCHAL_DCACHE_LINE_LOCKABLE
+		     "lock"
+#endif
+		     "\n",
+		     XCHAL_ICACHE_LINESIZE,
+		     XCHAL_ICACHE_WAYS,
+		     XCHAL_ICACHE_SIZE,
+		     XCHAL_DCACHE_LINESIZE,
+		     XCHAL_DCACHE_WAYS,
+		     XCHAL_DCACHE_SIZE);
+
+	/* MMU */
+	seq_printf(f,"ASID bits\t: %d\n"
+		     "ASID invalid\t: %d\n"
+		     "ASID kernel\t: %d\n"
+		     "rings\t\t: %d\n"
+		     "itlb ways\t: %d\n"
+		     "itlb AR ways\t: %d\n"
+		     "dtlb ways\t: %d\n"
+		     "dtlb AR ways\t: %d\n",
+		     XCHAL_MMU_ASID_BITS,
+		     XCHAL_MMU_ASID_INVALID,
+		     XCHAL_MMU_ASID_KERNEL,
+		     XCHAL_MMU_RINGS,
+		     XCHAL_ITLB_WAYS,
+		     XCHAL_ITLB_ARF_WAYS,
+		     XCHAL_DTLB_WAYS,
+		     XCHAL_DTLB_ARF_WAYS);
+
+	return 0;
+}
+
+/*
+ * We show only CPU #0 info.
+ */
+static void *
+c_start(struct seq_file *f, loff_t *pos)
+{
+	return (void *) ((*pos == 0) ? (void *)1 : NULL);
+}
+
+static void *
+c_next(struct seq_file *f, void *v, loff_t *pos)
+{
+	return NULL;
+}
+
+static void
+c_stop(struct seq_file *f, void *v)
+{
+}
+
+struct seq_operations cpuinfo_op =
+{
+	start:  c_start,
+	next:   c_next,
+	stop:   c_stop,
+	show:   c_show
+};
+
+#endif /* CONFIG_PROC_FS */
+
diff --git a/arch/xtensa/kernel/signal.c b/arch/xtensa/kernel/signal.c
new file mode 100644
index 000000000000..df6e1e17b096
--- /dev/null
+++ b/arch/xtensa/kernel/signal.c
@@ -0,0 +1,713 @@
+// TODO coprocessor stuff
+/*
+ *  linux/arch/xtensa/kernel/signal.c
+ *
+ *  Copyright (C) 1991, 1992  Linus Torvalds
+ *  1997-11-28  Modified for POSIX.1b signals by Richard Henderson
+ *
+ *  Joe Taylor <joe@tensilica.com>
+ *  Chris Zankel <chris@zankel.net>
+ *
+ *
+ *
+ */
+
+#include <xtensa/config/core.h>
+#include <xtensa/hal.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/kernel.h>
+#include <linux/signal.h>
+#include <linux/errno.h>
+#include <linux/wait.h>
+#include <linux/ptrace.h>
+#include <linux/unistd.h>
+#include <linux/stddef.h>
+#include <linux/personality.h>
+#include <asm/ucontext.h>
+#include <asm/uaccess.h>
+#include <asm/pgtable.h>
+#include <asm/cacheflush.h>
+
+#define DEBUG_SIG  0
+
+#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
+
+asmlinkage long sys_wait4(pid_t pid,unsigned int * stat_addr, int options,
+			  struct rusage * ru);
+asmlinkage int do_signal(struct pt_regs *regs, sigset_t *oldset);
+
+extern struct task_struct *coproc_owners[];
+
+
+/*
+ * Atomically swap in the new signal mask, and wait for a signal.
+ */
+
+int sys_sigsuspend(struct pt_regs *regs)
+{
+	old_sigset_t mask = (old_sigset_t) regs->areg[3];
+	sigset_t saveset;
+
+	mask &= _BLOCKABLE;
+	spin_lock_irq(&current->sighand->siglock);
+	saveset = current->blocked;
+	siginitset(&current->blocked, mask);
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	regs->areg[2] = -EINTR;
+	while (1) {
+		current->state = TASK_INTERRUPTIBLE;
+		schedule();
+		if (do_signal(regs, &saveset))
+			return -EINTR;
+	}
+}
+
+asmlinkage int
+sys_rt_sigsuspend(struct pt_regs *regs)
+{
+	sigset_t *unewset = (sigset_t *) regs->areg[4];
+	size_t sigsetsize = (size_t) regs->areg[3];
+	sigset_t saveset, newset;
+	/* XXX: Don't preclude handling different sized sigset_t's.  */
+	if (sigsetsize != sizeof(sigset_t))
+		return -EINVAL;
+
+	if (copy_from_user(&newset, unewset, sizeof(newset)))
+		return -EFAULT;
+	sigdelsetmask(&newset, ~_BLOCKABLE);
+	spin_lock_irq(&current->sighand->siglock);
+	saveset = current->blocked;
+	current->blocked = newset;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	regs->areg[2] = -EINTR;
+	while (1) {
+		current->state = TASK_INTERRUPTIBLE;
+		schedule();
+		if (do_signal(regs, &saveset))
+			return -EINTR;
+	}
+}
+
+asmlinkage int
+sys_sigaction(int sig, const struct old_sigaction *act,
+	      struct old_sigaction *oact)
+{
+	struct k_sigaction new_ka, old_ka;
+	int ret;
+
+	if (act) {
+		old_sigset_t mask;
+		if (verify_area(VERIFY_READ, act, sizeof(*act)) ||
+		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
+		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer))
+			return -EFAULT;
+		__get_user(new_ka.sa.sa_flags, &act->sa_flags);
+		__get_user(mask, &act->sa_mask);
+		siginitset(&new_ka.sa.sa_mask, mask);
+	}
+
+	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
+
+	if (!ret && oact) {
+		if (verify_area(VERIFY_WRITE, oact, sizeof(*oact)) ||
+		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
+		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer))
+			return -EFAULT;
+		__put_user(old_ka.sa.sa_flags, &oact->sa_flags);
+		__put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask);
+	}
+
+	return ret;
+}
+
+asmlinkage int
+sys_sigaltstack(struct pt_regs *regs)
+{
+	const stack_t *uss = (stack_t *) regs->areg[4];
+	stack_t *uoss = (stack_t *) regs->areg[3];
+
+	if (regs->depc > 64)
+		panic ("Double exception sys_sigreturn\n");
+
+
+	return do_sigaltstack(uss, uoss, regs->areg[1]);
+}
+
+
+/*
+ * Do a signal return; undo the signal stack.
+ */
+
+struct sigframe
+{
+	struct sigcontext sc;
+	struct _cpstate cpstate;
+	unsigned long extramask[_NSIG_WORDS-1];
+	unsigned char retcode[6];
+	unsigned int reserved[4]; /* Reserved area for chaining */
+	unsigned int window[4]; /* Window of 4 registers for initial context */
+};
+
+struct rt_sigframe
+{
+	struct siginfo info;
+	struct ucontext uc;
+	struct _cpstate cpstate;
+	unsigned char retcode[6];
+	unsigned int reserved[4]; /* Reserved area for chaining */
+	unsigned int window[4]; /* Window of 4 registers for initial context */
+};
+
+extern void release_all_cp (struct task_struct *);
+
+
+// FIXME restore_cpextra
+static inline int
+restore_cpextra (struct _cpstate *buf)
+{
+#if 0
+	/* The signal handler may have used coprocessors in which
+	 * case they are still enabled.  We disable them to force a
+	 * reloading of the original task's CP state by the lazy
+	 * context-switching mechanisms of CP exception handling.
+	 * Also, we essentially discard any coprocessor state that the
+	 * signal handler created. */
+
+	struct task_struct *tsk = current;
+	release_all_cp(tsk);
+	return __copy_from_user(tsk->thread.cpextra, buf, TOTAL_CPEXTRA_SIZE);
+#endif
+	return 0;
+}
+
+/* Note: We don't copy double exception 'tregs', we have to finish double exc. first before we return to signal handler! This dbl.exc.handler might cause another double exception, but I think we are fine as the situation is the same as if we had returned to the signal handerl and got an interrupt immediately...
+ */
+
+
+static int
+restore_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
+{
+	struct thread_struct *thread;
+	unsigned int err = 0;
+	unsigned long ps;
+	struct _cpstate *buf;
+
+#define COPY(x)	err |= __get_user(regs->x, &sc->sc_##x)
+	COPY(pc);
+	COPY(depc);
+	COPY(wmask);
+	COPY(lbeg);
+	COPY(lend);
+	COPY(lcount);
+	COPY(sar);
+	COPY(windowbase);
+	COPY(windowstart);
+#undef COPY
+
+	/* For PS, restore only PS.CALLINC.
+	 * Assume that all other bits are either the same as for the signal
+	 * handler, or the user mode value doesn't matter (e.g. PS.OWB).
+	 */
+	err |= __get_user(ps, &sc->sc_ps);
+	regs->ps = (regs->ps & ~XCHAL_PS_CALLINC_MASK)
+		| (ps & XCHAL_PS_CALLINC_MASK);
+
+	/* Additional corruption checks */
+
+	if ((regs->windowbase >= (XCHAL_NUM_AREGS/4))
+	|| ((regs->windowstart & ~((1<<(XCHAL_NUM_AREGS/4)) - 1)) != 0) )
+		err = 1;
+	if ((regs->lcount > 0)
+	&& ((regs->lbeg > TASK_SIZE) || (regs->lend > TASK_SIZE)) )
+		err = 1;
+
+	/* Restore extended register state.
+	 * See struct thread_struct in processor.h.
+	 */
+	thread = &current->thread;
+
+	err |= __copy_from_user (regs->areg, sc->sc_areg, XCHAL_NUM_AREGS*4);
+	err |= __get_user(buf, &sc->sc_cpstate);
+	if (buf) {
+		if (verify_area(VERIFY_READ, buf, sizeof(*buf)))
+			goto badframe;
+		err |= restore_cpextra(buf);
+	}
+
+	regs->syscall = -1;		/* disable syscall checks */
+	return err;
+
+badframe:
+	return 1;
+}
+
+static inline void
+flush_my_cpstate(struct task_struct *tsk)
+{
+	unsigned long flags;
+	local_irq_save(flags);
+
+#if 0	// FIXME
+	for (i = 0; i < XCHAL_CP_NUM; i++) {
+		if (tsk == coproc_owners[i]) {
+			xthal_validate_cp(i);
+			xthal_save_cpregs(tsk->thread.cpregs_ptr[i], i);
+
+			/* Invalidate and "disown" the cp to allow
+			 * callers the chance to reset cp state in the
+			 * task_struct. */
+
+			xthal_invalidate_cp(i);
+			coproc_owners[i] = 0;
+		}
+	}
+#endif
+	local_irq_restore(flags);
+}
+
+/* Return codes:
+	0:  nothing saved
+	1:  stuff to save, successful
+       -1:  stuff to save, error happened
+*/
+static int
+save_cpextra (struct _cpstate *buf)
+{
+#if (XCHAL_EXTRA_SA_SIZE == 0) && (XCHAL_CP_NUM == 0)
+	return 0;
+#else
+
+	/* FIXME: If a task has never used a coprocessor, there is
+	 * no need to save and restore anything.  Tracking this
+	 * information would allow us to optimize this section.
+	 * Perhaps we can use current->used_math or (current->flags &
+	 * PF_USEDFPU) or define a new field in the thread
+	 * structure. */
+
+	/* We flush any live, task-owned cp state to the task_struct,
+	 * then copy it all to the sigframe.  Then we clear all
+	 * cp/extra state in the task_struct, effectively
+	 * clearing/resetting all cp/extra state for the signal
+	 * handler (cp-exception handling will load these new values
+	 * into the cp/extra registers.)  This step is important for
+	 * things like a floating-point cp, where the OS must reset
+	 * the FCR to the default rounding mode. */
+
+	int err = 0;
+	struct task_struct *tsk = current;
+
+	flush_my_cpstate(tsk);
+	/* Note that we just copy everything: 'extra' and 'cp' state together.*/
+	err |= __copy_to_user(buf, tsk->thread.cp_save, XTENSA_CP_EXTRA_SIZE);
+	memset(tsk->thread.cp_save, 0, XTENSA_CP_EXTRA_SIZE);
+
+#if (XTENSA_CP_EXTRA_SIZE == 0)
+#error Sanity check on memset above, cpextra_size should not be zero.
+#endif
+
+	return err ? -1 : 1;
+#endif
+}
+
+static int
+setup_sigcontext(struct sigcontext *sc, struct _cpstate *cpstate,
+		 struct pt_regs *regs, unsigned long mask)
+{
+	struct thread_struct *thread;
+	int err = 0;
+
+//printk("setup_sigcontext\n");
+#define COPY(x)	err |= __put_user(regs->x, &sc->sc_##x)
+	COPY(pc);
+	COPY(ps);
+	COPY(depc);
+	COPY(wmask);
+	COPY(lbeg);
+	COPY(lend);
+	COPY(lcount);
+	COPY(sar);
+	COPY(windowbase);
+	COPY(windowstart);
+#undef COPY
+
+	/* Save extended register state.
+	 * See struct thread_struct in processor.h.
+	 */
+	thread = &current->thread;
+	err |= __copy_to_user (sc->sc_areg, regs->areg, XCHAL_NUM_AREGS * 4);
+	err |= save_cpextra(cpstate);
+	err |= __put_user(err ? NULL : cpstate, &sc->sc_cpstate);
+	/* non-iBCS2 extensions.. */
+	err |= __put_user(mask, &sc->oldmask);
+
+	return err;
+}
+
+asmlinkage int sys_sigreturn(struct pt_regs *regs)
+{
+	struct sigframe *frame = (struct sigframe *)regs->areg[1];
+	sigset_t set;
+	if (regs->depc > 64)
+		panic ("Double exception sys_sigreturn\n");
+
+	if (verify_area(VERIFY_READ, frame, sizeof(*frame)))
+		goto badframe;
+
+	if (__get_user(set.sig[0], &frame->sc.oldmask)
+	    || (_NSIG_WORDS > 1
+		&& __copy_from_user(&set.sig[1], &frame->extramask,
+				    sizeof(frame->extramask))))
+		goto badframe;
+
+	sigdelsetmask(&set, ~_BLOCKABLE);
+
+	spin_lock_irq(&current->sighand->siglock);
+	current->blocked = set;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	if (restore_sigcontext(regs, &frame->sc))
+		goto badframe;
+	return regs->areg[2];
+
+badframe:
+	force_sig(SIGSEGV, current);
+	return 0;
+}
+
+asmlinkage int sys_rt_sigreturn(struct pt_regs *regs)
+{
+	struct rt_sigframe *frame = (struct rt_sigframe *)regs->areg[1];
+	sigset_t set;
+	stack_t st;
+	int ret;
+	if (regs->depc > 64)
+	{
+		printk("!!!!!!! DEPC !!!!!!!\n");
+		return 0;
+	}
+
+	if (verify_area(VERIFY_READ, frame, sizeof(*frame)))
+		goto badframe;
+
+	if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
+		goto badframe;
+
+	sigdelsetmask(&set, ~_BLOCKABLE);
+	spin_lock_irq(&current->sighand->siglock);
+	current->blocked = set;
+	recalc_sigpending();
+	spin_unlock_irq(&current->sighand->siglock);
+
+	if (restore_sigcontext(regs, &frame->uc.uc_mcontext))
+		goto badframe;
+	ret = regs->areg[2];
+
+	if (__copy_from_user(&st, &frame->uc.uc_stack, sizeof(st)))
+		goto badframe;
+	/* It is more difficult to avoid calling this function than to
+	   call it and ignore errors.  */
+	do_sigaltstack(&st, NULL, regs->areg[1]);
+
+	return ret;
+
+badframe:
+	force_sig(SIGSEGV, current);
+	return 0;
+}
+
+/*
+ * Set up a signal frame.
+ */
+
+/*
+ * Determine which stack to use..
+ */
+static inline void *
+get_sigframe(struct k_sigaction *ka, unsigned long sp, size_t frame_size)
+{
+	if ((ka->sa.sa_flags & SA_ONSTACK) != 0 && ! on_sig_stack(sp))
+		sp = current->sas_ss_sp + current->sas_ss_size;
+
+	return (void *)((sp - frame_size) & -16ul);
+}
+
+#define USE_SIGRETURN		0
+#define USE_RT_SIGRETURN	1
+
+static int
+gen_return_code(unsigned char *codemem, unsigned int use_rt_sigreturn)
+{
+	unsigned int retcall;
+	int err = 0;
+
+#if 0
+	/* Ignoring SA_RESTORER for now; it's supposed to be obsolete,
+	 * and the xtensa glibc doesn't use it.
+	 */
+	if (ka->sa.sa_flags & SA_RESTORER) {
+		regs->pr = (unsigned long) ka->sa.sa_restorer;
+	} else
+#endif /* 0 */
+	{
+
+#if (__NR_sigreturn > 255) || (__NR_rt_sigreturn > 255)
+
+/* The 12-bit immediate is really split up within the 24-bit MOVI
+ * instruction.  As long as the above system call numbers fit within
+ * 8-bits, the following code works fine. See the Xtensa ISA for
+ * details.
+ */
+
+#error Generating the MOVI instruction below breaks!
+#endif
+
+		retcall = use_rt_sigreturn ? __NR_rt_sigreturn : __NR_sigreturn;
+
+#ifdef __XTENSA_EB__   /* Big Endian version */
+		/* Generate instruction:  MOVI a2, retcall */
+		err |= __put_user(0x22, &codemem[0]);
+		err |= __put_user(0x0a, &codemem[1]);
+		err |= __put_user(retcall, &codemem[2]);
+		/* Generate instruction:  SYSCALL */
+		err |= __put_user(0x00, &codemem[3]);
+		err |= __put_user(0x05, &codemem[4]);
+		err |= __put_user(0x00, &codemem[5]);
+
+#elif defined __XTENSA_EL__   /* Little Endian version */
+		/* Generate instruction:  MOVI a2, retcall */
+		err |= __put_user(0x22, &codemem[0]);
+		err |= __put_user(0xa0, &codemem[1]);
+		err |= __put_user(retcall, &codemem[2]);
+		/* Generate instruction:  SYSCALL */
+		err |= __put_user(0x00, &codemem[3]);
+		err |= __put_user(0x50, &codemem[4]);
+		err |= __put_user(0x00, &codemem[5]);
+#else
+#error Must use compiler for Xtensa processors.
+#endif
+	}
+
+	/* Flush generated code out of the data cache */
+
+	if (err == 0)
+		__flush_invalidate_cache_range((unsigned long)codemem, 6UL);
+
+	return err;
+}
+
+static void
+set_thread_state(struct pt_regs *regs, void *stack, unsigned char *retaddr,
+	void *handler, unsigned long arg1, void *arg2, void *arg3)
+{
+	/* Set up registers for signal handler */
+	start_thread(regs, (unsigned long) handler, (unsigned long) stack);
+
+	/* Set up a stack frame for a call4
+	 * Note: PS.CALLINC is set to one by start_thread
+	 */
+	regs->areg[4] = (((unsigned long) retaddr) & 0x3fffffff) | 0x40000000;
+	regs->areg[6] = arg1;
+	regs->areg[7] = (unsigned long) arg2;
+	regs->areg[8] = (unsigned long) arg3;
+}
+
+static void setup_frame(int sig, struct k_sigaction *ka,
+			sigset_t *set, struct pt_regs *regs)
+{
+	struct sigframe *frame;
+	int err = 0;
+	int signal;
+
+	frame = get_sigframe(ka, regs->areg[1], sizeof(*frame));
+	if (regs->depc > 64)
+	{
+		printk("!!!!!!! DEPC !!!!!!!\n");
+		return;
+	}
+
+
+	if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
+		goto give_sigsegv;
+
+	signal = current_thread_info()->exec_domain
+		&& current_thread_info()->exec_domain->signal_invmap
+		&& sig < 32
+		? current_thread_info()->exec_domain->signal_invmap[sig]
+		: sig;
+
+	err |= setup_sigcontext(&frame->sc, &frame->cpstate, regs, set->sig[0]);
+
+	if (_NSIG_WORDS > 1) {
+		err |= __copy_to_user(frame->extramask, &set->sig[1],
+				      sizeof(frame->extramask));
+	}
+
+	/* Create sys_sigreturn syscall in stack frame */
+	err |= gen_return_code(frame->retcode, USE_SIGRETURN);
+
+	if (err)
+		goto give_sigsegv;
+
+	/* Create signal handler execution context.
+	 * Return context not modified until this point.
+	 */
+	set_thread_state(regs, frame, frame->retcode,
+		ka->sa.sa_handler, signal, &frame->sc, NULL);
+
+	/* Set access mode to USER_DS.  Nomenclature is outdated, but
+	 * functionality is used in uaccess.h
+	 */
+	set_fs(USER_DS);
+
+
+#if DEBUG_SIG
+	printk("SIG deliver (%s:%d): signal=%d sp=%p pc=%08x\n",
+		current->comm, current->pid, signal, frame, regs->pc);
+#endif
+
+	return;
+
+give_sigsegv:
+	if (sig == SIGSEGV)
+		ka->sa.sa_handler = SIG_DFL;
+	force_sig(SIGSEGV, current);
+}
+
+static void setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
+			   sigset_t *set, struct pt_regs *regs)
+{
+	struct rt_sigframe *frame;
+	int err = 0;
+	int signal;
+
+	frame = get_sigframe(ka, regs->areg[1], sizeof(*frame));
+	if (regs->depc > 64)
+		panic ("Double exception sys_sigreturn\n");
+
+	if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
+		goto give_sigsegv;
+
+	signal = current_thread_info()->exec_domain
+		&& current_thread_info()->exec_domain->signal_invmap
+		&& sig < 32
+		? current_thread_info()->exec_domain->signal_invmap[sig]
+		: sig;
+
+	err |= copy_siginfo_to_user(&frame->info, info);
+
+	/* Create the ucontext.  */
+	err |= __put_user(0, &frame->uc.uc_flags);
+	err |= __put_user(0, &frame->uc.uc_link);
+	err |= __put_user((void *)current->sas_ss_sp,
+			  &frame->uc.uc_stack.ss_sp);
+	err |= __put_user(sas_ss_flags(regs->areg[1]),
+			  &frame->uc.uc_stack.ss_flags);
+	err |= __put_user(current->sas_ss_size, &frame->uc.uc_stack.ss_size);
+	err |= setup_sigcontext(&frame->uc.uc_mcontext, &frame->cpstate,
+			        regs, set->sig[0]);
+	err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set));
+
+	/* Create sys_rt_sigreturn syscall in stack frame */
+	err |= gen_return_code(frame->retcode, USE_RT_SIGRETURN);
+
+	if (err)
+		goto give_sigsegv;
+
+	/* Create signal handler execution context.
+	 * Return context not modified until this point.
+	 */
+	set_thread_state(regs, frame, frame->retcode,
+		ka->sa.sa_handler, signal, &frame->info, &frame->uc);
+
+	/* Set access mode to USER_DS.  Nomenclature is outdated, but
+	 * functionality is used in uaccess.h
+	 */
+	set_fs(USER_DS);
+
+#if DEBUG_SIG
+	printk("SIG rt deliver (%s:%d): signal=%d sp=%p pc=%08x\n",
+		current->comm, current->pid, signal, frame, regs->pc);
+#endif
+
+	return;
+
+give_sigsegv:
+	if (sig == SIGSEGV)
+		ka->sa.sa_handler = SIG_DFL;
+	force_sig(SIGSEGV, current);
+}
+
+
+
+/*
+ * Note that 'init' is a special process: it doesn't get signals it doesn't
+ * want to handle. Thus you cannot kill init even with a SIGKILL even by
+ * mistake.
+ *
+ * Note that we go through the signals twice: once to check the signals that
+ * the kernel can handle, and then we build all the user-level signal handling
+ * stack-frames in one go after that.
+ */
+int do_signal(struct pt_regs *regs, sigset_t *oldset)
+{
+	siginfo_t info;
+	int signr;
+	struct k_sigaction ka;
+
+	if (!oldset)
+		oldset = &current->blocked;
+
+	signr = get_signal_to_deliver(&info, &ka, regs, NULL);
+
+	/* Are we from a system call? */
+	if (regs->syscall >= 0) {
+		/* If so, check system call restarting.. */
+		switch (regs->areg[2]) {
+			case ERESTARTNOHAND:
+			case ERESTART_RESTARTBLOCK:
+				regs->areg[2] = -EINTR;
+				break;
+
+			case ERESTARTSYS:
+				if (!(ka.sa.sa_flags & SA_RESTART)) {
+					regs->areg[2] = -EINTR;
+					break;
+				}
+			/* fallthrough */
+			case ERESTARTNOINTR:
+				regs->areg[2] = regs->syscall;
+				regs->pc -= 3;
+		}
+	}
+
+	if (signr == 0)
+		return 0;		/* no signals delivered */
+
+	/* Whee!  Actually deliver the signal.  */
+
+	/* Set up the stack frame */
+	if (ka.sa.sa_flags & SA_SIGINFO)
+		setup_rt_frame(signr, &ka, &info, oldset, regs);
+	else
+		setup_frame(signr, &ka, oldset, regs);
+
+	if (ka.sa.sa_flags & SA_ONESHOT)
+		ka.sa.sa_handler = SIG_DFL;
+
+	if (!(ka.sa.sa_flags & SA_NODEFER)) {
+		spin_lock_irq(&current->sighand->siglock);
+		sigorsets(&current->blocked, &current->blocked, &ka.sa.sa_mask);
+		sigaddset(&current->blocked, signr);
+		recalc_sigpending();
+		spin_unlock_irq(&current->sighand->siglock);
+	}
+	return 1;
+}
diff --git a/arch/xtensa/kernel/syscalls.c b/arch/xtensa/kernel/syscalls.c
new file mode 100644
index 000000000000..abc8ed6c7026
--- /dev/null
+++ b/arch/xtensa/kernel/syscalls.c
@@ -0,0 +1,418 @@
+/*
+ * arch/xtensa/kernel/syscall.c
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ * Copyright (C) 2000 Silicon Graphics, Inc.
+ * Copyright (C) 1995 - 2000 by Ralf Baechle
+ *
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Chris Zankel <chris@zankel.net>
+ * Kevin Chea
+ *
+ */
+
+#define DEBUG	0
+
+#include <linux/config.h>
+#include <linux/linkage.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/mman.h>
+#include <linux/sched.h>
+#include <linux/file.h>
+#include <linux/slab.h>
+#include <linux/utsname.h>
+#include <linux/unistd.h>
+#include <linux/stringify.h>
+#include <linux/syscalls.h>
+#include <linux/sem.h>
+#include <linux/msg.h>
+#include <linux/shm.h>
+#include <linux/errno.h>
+#include <asm/ptrace.h>
+#include <asm/signal.h>
+#include <asm/uaccess.h>
+#include <asm/hardirq.h>
+#include <asm/mman.h>
+#include <asm/shmparam.h>
+#include <asm/page.h>
+#include <asm/ipc.h>
+
+extern void do_syscall_trace(void);
+typedef int (*syscall_t)(void *a0,...);
+extern int (*do_syscalls)(struct pt_regs *regs, syscall_t fun,
+				     int narg);
+extern syscall_t sys_call_table[];
+extern unsigned char sys_narg_table[];
+
+/*
+ * sys_pipe() is the normal C calling standard for creating a pipe. It's not
+ * the way unix traditional does this, though.
+ */
+
+int sys_pipe(int __user *userfds)
+{
+	int fd[2];
+	int error;
+
+	error = do_pipe(fd);
+	if (!error) {
+		if (copy_to_user(userfds, fd, 2 * sizeof(int)))
+			error = -EFAULT;
+	}
+	return error;
+}
+
+/*
+ * Common code for old and new mmaps.
+ */
+
+static inline long do_mmap2(unsigned long addr, unsigned long len,
+    			    unsigned long prot, unsigned long flags,
+			    unsigned long fd, unsigned long pgoff)
+{
+	int error = -EBADF;
+	struct file * file = NULL;
+
+	flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
+	if (!(flags & MAP_ANONYMOUS)) {
+		file = fget(fd);
+		if (!file)
+			goto out;
+	}
+
+	down_write(&current->mm->mmap_sem);
+	error = do_mmap_pgoff(file, addr, len, prot, flags, pgoff);
+	up_write(&current->mm->mmap_sem);
+
+	if (file)
+		fput(file);
+out:
+	return error;
+}
+
+unsigned long old_mmap(unsigned long addr, size_t len, int prot,
+		       int flags, int fd, off_t offset)
+{
+	return do_mmap2(addr, len, prot, flags, fd, offset >> PAGE_SHIFT);
+}
+
+long sys_mmap2(unsigned long addr, unsigned long len, unsigned long prot,
+	       unsigned long flags, unsigned long fd, unsigned long pgoff)
+{
+	return do_mmap2(addr, len, prot, flags, fd, pgoff);
+}
+
+int sys_fork(struct pt_regs *regs)
+{
+	return do_fork(SIGCHLD, regs->areg[1], regs, 0, NULL, NULL);
+}
+
+int sys_vfork(struct pt_regs *regs)
+{
+	return do_fork(CLONE_VFORK|CLONE_VM|SIGCHLD, regs->areg[1],
+		       regs, 0, NULL, NULL);
+}
+
+int sys_clone(struct pt_regs *regs)
+{
+	unsigned long clone_flags;
+	unsigned long newsp;
+	int __user *parent_tidptr, *child_tidptr;
+	clone_flags = regs->areg[4];
+	newsp = regs->areg[3];
+	parent_tidptr = (int __user *)regs->areg[5];
+	child_tidptr = (int __user *)regs->areg[6];
+	if (!newsp)
+		newsp = regs->areg[1];
+	return do_fork(clone_flags,newsp,regs,0,parent_tidptr,child_tidptr);
+}
+
+/*
+ * sys_execve() executes a new program.
+ */
+
+int sys_execve(struct pt_regs *regs)
+{
+	int error;
+	char * filename;
+
+	filename = getname((char *) (long)regs->areg[5]);
+	error = PTR_ERR(filename);
+	if (IS_ERR(filename))
+		goto out;
+	error = do_execve(filename, (char **) (long)regs->areg[3],
+	                  (char **) (long)regs->areg[4], regs);
+	putname(filename);
+
+out:
+	return error;
+}
+
+int sys_uname(struct old_utsname * name)
+{
+	if (name && !copy_to_user(name, &system_utsname, sizeof (*name)))
+		return 0;
+	return -EFAULT;
+}
+
+int sys_olduname(struct oldold_utsname * name)
+{
+	int error;
+
+	if (!name)
+		return -EFAULT;
+	if (!access_ok(VERIFY_WRITE,name,sizeof(struct oldold_utsname)))
+		return -EFAULT;
+
+	error = __copy_to_user(&name->sysname,&system_utsname.sysname,__OLD_UTS_LEN);
+	error -= __put_user(0,name->sysname+__OLD_UTS_LEN);
+	error -= __copy_to_user(&name->nodename,&system_utsname.nodename,__OLD_UTS_LEN);
+	error -= __put_user(0,name->nodename+__OLD_UTS_LEN);
+	error -= __copy_to_user(&name->release,&system_utsname.release,__OLD_UTS_LEN);
+	error -= __put_user(0,name->release+__OLD_UTS_LEN);
+	error -= __copy_to_user(&name->version,&system_utsname.version,__OLD_UTS_LEN);
+	error -= __put_user(0,name->version+__OLD_UTS_LEN);
+	error -= __copy_to_user(&name->machine,&system_utsname.machine,__OLD_UTS_LEN);
+	error -= __put_user(0,name->machine+__OLD_UTS_LEN);
+
+	return error ? -EFAULT : 0;
+}
+
+
+/*
+ * Build the string table for the builtin "poor man's strace".
+ */
+
+#if DEBUG
+#define SYSCALL(fun, narg) #fun,
+static char *sfnames[] = {
+#include "syscalls.h"
+};
+#undef SYS
+#endif
+
+void system_call (struct pt_regs *regs)
+{
+	syscall_t syscall;
+	unsigned long parm0, parm1, parm2, parm3, parm4, parm5;
+	int nargs, res;
+	unsigned int syscallnr;
+	int ps;
+
+#if DEBUG
+	int i;
+	unsigned long parms[6];
+	char *sysname;
+#endif
+
+	regs->syscall = regs->areg[2];
+
+	do_syscall_trace();
+
+	/* Have to load after syscall_trace because strace
+	 * sometimes changes regs->syscall.
+	 */
+	syscallnr = regs->syscall;
+
+	parm0 = parm1 = parm2 = parm3 = parm4 = parm5 = 0;
+
+	/* Restore interrupt level to syscall invoker's.
+	 * If this were in assembly, we wouldn't disable
+	 * interrupts in the first place:
+	 */
+	local_save_flags (ps);
+	local_irq_restore((ps & ~XCHAL_PS_INTLEVEL_MASK) |
+			  (regs->ps & XCHAL_PS_INTLEVEL_MASK) );
+
+	if (syscallnr > __NR_Linux_syscalls) {
+		regs->areg[2] = -ENOSYS;
+		return;
+	}
+
+	syscall = sys_call_table[syscallnr];
+	nargs = sys_narg_table[syscallnr];
+
+	if (syscall == NULL) {
+		regs->areg[2] = -ENOSYS;
+		return;
+	}
+
+	/* There shouldn't be more than six arguments in the table! */
+
+	if (nargs > 6)
+		panic("Internal error - too many syscall arguments (%d)!\n",
+		      nargs);
+
+	/* Linux takes system-call arguments in registers.  The ABI
+         * and Xtensa software conventions require the system-call
+         * number in a2.  If an argument exists in a2, we move it to
+         * the next available register.  Note that for improved
+         * efficiency, we do NOT shift all parameters down one
+         * register to maintain the original order.
+	 *
+         * At best case (zero arguments), we just write the syscall
+         * number to a2.  At worst case (1 to 6 arguments), we move
+         * the argument in a2 to the next available register, then
+         * write the syscall number to a2.
+	 *
+         * For clarity, the following truth table enumerates all
+         * possibilities.
+	 *
+         * arguments	syscall number	arg0, arg1, arg2, arg3, arg4, arg5
+         * ---------	--------------	----------------------------------
+	 *	0	      a2
+	 *	1	      a2	a3
+	 *	2	      a2	a4,   a3
+	 *	3	      a2	a5,   a3,   a4
+	 *	4	      a2	a6,   a3,   a4,   a5
+	 *	5	      a2	a7,   a3,   a4,   a5,   a6
+	 *	6	      a2	a8,   a3,   a4,   a5,   a6,   a7
+	 */
+	if (nargs) {
+		parm0 = regs->areg[nargs+2];
+		parm1 = regs->areg[3];
+		parm2 = regs->areg[4];
+		parm3 = regs->areg[5];
+		parm4 = regs->areg[6];
+		parm5 = regs->areg[7];
+	} else /* nargs == 0 */
+		parm0 = (unsigned long) regs;
+
+#if DEBUG
+	parms[0] = parm0;
+	parms[1] = parm1;
+	parms[2] = parm2;
+	parms[3] = parm3;
+	parms[4] = parm4;
+	parms[5] = parm5;
+
+	sysname = sfnames[syscallnr];
+	if (strncmp(sysname, "sys_", 4) == 0)
+		sysname = sysname + 4;
+
+	printk("\017SYSCALL:I:%x:%d:%s  %s(", regs->pc, current->pid,
+	       current->comm, sysname);
+	for (i = 0; i < nargs; i++)
+		printk((i>0) ? ", %#lx" : "%#lx", parms[i]);
+	printk(")\n");
+#endif
+
+	res = syscall((void *)parm0, parm1, parm2, parm3, parm4, parm5);
+
+#if DEBUG
+	printk("\017SYSCALL:O:%d:%s  %s(",current->pid, current->comm, sysname);
+	for (i = 0; i < nargs; i++)
+		printk((i>0) ? ", %#lx" : "%#lx", parms[i]);
+	if (res < 4096)
+		printk(") = %d\n", res);
+	else
+		printk(") = %#x\n", res);
+#endif /* DEBUG */
+
+	regs->areg[2] = res;
+	do_syscall_trace();
+}
+
+/*
+ * sys_ipc() is the de-multiplexer for the SysV IPC calls..
+ *
+ * This is really horribly ugly.
+ */
+
+int sys_ipc (uint call, int first, int second,
+			int third, void __user *ptr, long fifth)
+{
+	int version, ret;
+
+	version = call >> 16; /* hack for backward compatibility */
+	call &= 0xffff;
+	ret = -ENOSYS;
+
+	switch (call) {
+	case SEMOP:
+		ret = sys_semtimedop (first, (struct sembuf __user *)ptr,
+				     second, NULL);
+		break;
+
+	case SEMTIMEDOP:
+		ret = sys_semtimedop (first, (struct sembuf __user *)ptr,
+				      second, (const struct timespec *) fifth);
+		break;
+
+	case SEMGET:
+		ret = sys_semget (first, second, third);
+		break;
+
+	case SEMCTL: {
+		union semun fourth;
+
+		if (ptr && !get_user(fourth.__pad, (void *__user *) ptr))
+			ret = sys_semctl (first, second, third, fourth);
+		break;
+		}
+
+	case MSGSND:
+		ret = sys_msgsnd (first, (struct msgbuf __user*) ptr,
+				  second, third);
+		break;
+
+	case MSGRCV:
+		switch (version) {
+		case 0: {
+			struct ipc_kludge tmp;
+
+			if (ptr && !copy_from_user(&tmp,
+					   (struct ipc_kludge *) ptr,
+					   sizeof (tmp)))
+				ret = sys_msgrcv (first, tmp.msgp, second,
+						  tmp.msgtyp, third);
+			break;
+			}
+
+		default:
+			ret = sys_msgrcv (first, (struct msgbuf __user *) ptr,
+					  second, 0, third);
+			break;
+		}
+		break;
+
+	case MSGGET:
+		ret = sys_msgget ((key_t) first, second);
+		break;
+
+	case MSGCTL:
+		ret = sys_msgctl (first, second, (struct msqid_ds __user*) ptr);
+		break;
+
+	case SHMAT: {
+		ulong raddr;
+		ret = do_shmat (first, (char __user *) ptr, second, &raddr);
+
+		if (!ret)
+			ret = put_user (raddr, (ulong __user *) third);
+
+		break;
+		}
+
+	case SHMDT:
+		ret = sys_shmdt ((char __user *)ptr);
+		break;
+
+	case SHMGET:
+		ret = sys_shmget (first, second, third);
+		break;
+
+	case SHMCTL:
+		ret = sys_shmctl (first, second, (struct shmid_ds __user*) ptr);
+		break;
+	}
+	return ret;
+}
+
diff --git a/arch/xtensa/kernel/syscalls.h b/arch/xtensa/kernel/syscalls.h
new file mode 100644
index 000000000000..5b3f75f50feb
--- /dev/null
+++ b/arch/xtensa/kernel/syscalls.h
@@ -0,0 +1,248 @@
+/*
+ * arch/xtensa/kernel/syscalls.h
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1995, 1996, 1997, 1998 by Ralf Baechle
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Changes by Joe Taylor <joe@tensilica.com>
+ */
+
+/*
+ * This file is being included twice - once to build a list of all
+ * syscalls and once to build a table of how many arguments each syscall
+ * accepts.  Syscalls that receive a pointer to the saved registers are
+ * marked as having zero arguments.
+ *
+ * The binary compatibility calls are in a separate list.
+ *
+ * Entry '0' used to be system_call.  It's removed to disable indirect
+ * system calls for now so user tasks can't recurse.  See mips'
+ * sys_syscall for a comparable example.
+ */
+
+SYSCALL(0, 0)		                /* 00 */
+
+SYSCALL(sys_exit, 1)
+SYSCALL(sys_fork, 0)
+SYSCALL(sys_read, 3)
+SYSCALL(sys_write, 3)
+SYSCALL(sys_open, 3)			/* 05 */
+SYSCALL(sys_close, 1)
+SYSCALL(sys_waitpid, 3)
+SYSCALL(sys_creat, 2)
+SYSCALL(sys_link, 2)
+SYSCALL(sys_unlink, 1)			/* 10 */
+SYSCALL(sys_execve, 0)
+SYSCALL(sys_chdir, 1)
+SYSCALL(sys_time, 1)
+SYSCALL(sys_mknod, 3)
+SYSCALL(sys_chmod, 2)			/* 15 */
+SYSCALL(sys_lchown, 3)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_stat, 2)
+SYSCALL(sys_lseek, 3)
+SYSCALL(sys_getpid, 0)			/* 20 */
+SYSCALL(sys_mount, 5)
+SYSCALL(sys_oldumount, 1)
+SYSCALL(sys_setuid, 1)
+SYSCALL(sys_getuid, 0)
+SYSCALL(sys_stime, 1)			/* 25 */
+SYSCALL(sys_ptrace, 4)
+SYSCALL(sys_alarm, 1)
+SYSCALL(sys_fstat, 2)
+SYSCALL(sys_pause, 0)
+SYSCALL(sys_utime, 2)			/* 30 */
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_access, 2)
+SYSCALL(sys_nice, 1)
+SYSCALL(sys_ni_syscall, 0)		/* 35 */
+SYSCALL(sys_sync, 0)
+SYSCALL(sys_kill, 2)
+SYSCALL(sys_rename, 2)
+SYSCALL(sys_mkdir, 2)
+SYSCALL(sys_rmdir, 1)			/* 40 */
+SYSCALL(sys_dup, 1)
+SYSCALL(sys_pipe, 1)
+SYSCALL(sys_times, 1)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_brk, 1)			/* 45 */
+SYSCALL(sys_setgid, 1)
+SYSCALL(sys_getgid, 0)
+SYSCALL(sys_ni_syscall, 0)		/* was signal(2) */
+SYSCALL(sys_geteuid, 0)
+SYSCALL(sys_getegid, 0)			/* 50 */
+SYSCALL(sys_acct, 1)
+SYSCALL(sys_umount, 2)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_ioctl, 3)
+SYSCALL(sys_fcntl, 3)			/* 55 */
+SYSCALL(sys_ni_syscall, 2)
+SYSCALL(sys_setpgid, 2)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_olduname, 1)
+SYSCALL(sys_umask, 1)			/* 60 */
+SYSCALL(sys_chroot, 1)
+SYSCALL(sys_ustat, 2)
+SYSCALL(sys_dup2, 2)
+SYSCALL(sys_getppid, 0)
+SYSCALL(sys_getpgrp, 0)			/* 65 */
+SYSCALL(sys_setsid, 0)
+SYSCALL(sys_sigaction, 3)
+SYSCALL(sys_sgetmask, 0)
+SYSCALL(sys_ssetmask, 1)
+SYSCALL(sys_setreuid, 2)		/* 70 */
+SYSCALL(sys_setregid, 2)
+SYSCALL(sys_sigsuspend, 0)
+SYSCALL(sys_sigpending, 1)
+SYSCALL(sys_sethostname, 2)
+SYSCALL(sys_setrlimit, 2)		/* 75 */
+SYSCALL(sys_getrlimit, 2)
+SYSCALL(sys_getrusage, 2)
+SYSCALL(sys_gettimeofday, 2)
+SYSCALL(sys_settimeofday, 2)
+SYSCALL(sys_getgroups, 2)		/* 80 */
+SYSCALL(sys_setgroups, 2)
+SYSCALL(sys_ni_syscall, 0)		 /* old_select */
+SYSCALL(sys_symlink, 2)
+SYSCALL(sys_lstat, 2)
+SYSCALL(sys_readlink, 3)		/* 85 */
+SYSCALL(sys_uselib, 1)
+SYSCALL(sys_swapon, 2)
+SYSCALL(sys_reboot, 3)
+SYSCALL(old_readdir, 3)
+SYSCALL(old_mmap, 6)			/* 90 */
+SYSCALL(sys_munmap, 2)
+SYSCALL(sys_truncate, 2)
+SYSCALL(sys_ftruncate, 2)
+SYSCALL(sys_fchmod, 2)
+SYSCALL(sys_fchown, 3)			/* 95 */
+SYSCALL(sys_getpriority, 2)
+SYSCALL(sys_setpriority, 3)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_statfs, 2)
+SYSCALL(sys_fstatfs, 2)			/* 100 */
+SYSCALL(sys_ni_syscall, 3)
+SYSCALL(sys_socketcall, 2)
+SYSCALL(sys_syslog, 3)
+SYSCALL(sys_setitimer, 3)
+SYSCALL(sys_getitimer, 2)		/* 105 */
+SYSCALL(sys_newstat, 2)
+SYSCALL(sys_newlstat, 2)
+SYSCALL(sys_newfstat, 2)
+SYSCALL(sys_uname, 1)
+SYSCALL(sys_ni_syscall, 0)		/* 110 */
+SYSCALL(sys_vhangup, 0)
+SYSCALL(sys_ni_syscall, 0)		/* was sys_idle() */
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_wait4, 4)
+SYSCALL(sys_swapoff, 1)			/* 115 */
+SYSCALL(sys_sysinfo, 1)
+SYSCALL(sys_ipc, 5)   			/* 6 really, but glibc uses only 5) */
+SYSCALL(sys_fsync, 1)
+SYSCALL(sys_sigreturn, 0)
+SYSCALL(sys_clone, 0)			/* 120 */
+SYSCALL(sys_setdomainname, 2)
+SYSCALL(sys_newuname, 1)
+SYSCALL(sys_ni_syscall, 0) 		/* sys_modify_ldt */
+SYSCALL(sys_adjtimex, 1)
+SYSCALL(sys_mprotect, 3)		/* 125 */
+SYSCALL(sys_sigprocmask, 3)
+SYSCALL(sys_ni_syscall, 2)		/* old sys_create_module */
+SYSCALL(sys_init_module, 2)
+SYSCALL(sys_delete_module, 1)
+SYSCALL(sys_ni_syscall, 1)		/* old sys_get_kernel_sysm */	/* 130 */
+SYSCALL(sys_quotactl, 0)
+SYSCALL(sys_getpgid, 1)
+SYSCALL(sys_fchdir, 1)
+SYSCALL(sys_bdflush, 2)
+SYSCALL(sys_sysfs, 3)			/* 135 */
+SYSCALL(sys_personality, 1)
+SYSCALL(sys_ni_syscall, 0)		/* for afs_syscall */
+SYSCALL(sys_setfsuid, 1)
+SYSCALL(sys_setfsgid, 1)
+SYSCALL(sys_llseek, 5)			/* 140 */
+SYSCALL(sys_getdents, 3)
+SYSCALL(sys_select, 5)
+SYSCALL(sys_flock, 2)
+SYSCALL(sys_msync, 3)
+SYSCALL(sys_readv, 3)			/* 145 */
+SYSCALL(sys_writev, 3)
+SYSCALL(sys_ni_syscall, 3)
+SYSCALL(sys_ni_syscall, 3)
+SYSCALL(sys_ni_syscall, 4)		/* handled in fast syscall handler. */
+SYSCALL(sys_ni_syscall, 0)		/* 150 */
+SYSCALL(sys_getsid, 1)
+SYSCALL(sys_fdatasync, 1)
+SYSCALL(sys_sysctl, 1)
+SYSCALL(sys_mlock, 2)
+SYSCALL(sys_munlock, 2)			/* 155 */
+SYSCALL(sys_mlockall, 1)
+SYSCALL(sys_munlockall, 0)
+SYSCALL(sys_sched_setparam,2)
+SYSCALL(sys_sched_getparam,2)
+SYSCALL(sys_sched_setscheduler,3)	/* 160 */
+SYSCALL(sys_sched_getscheduler,1)
+SYSCALL(sys_sched_yield,0)
+SYSCALL(sys_sched_get_priority_max,1)
+SYSCALL(sys_sched_get_priority_min,1)
+SYSCALL(sys_sched_rr_get_interval,2)	/* 165 */
+SYSCALL(sys_nanosleep,2)
+SYSCALL(sys_mremap,4)
+SYSCALL(sys_accept, 3)
+SYSCALL(sys_bind, 3)
+SYSCALL(sys_connect, 3)			/* 170 */
+SYSCALL(sys_getpeername, 3)
+SYSCALL(sys_getsockname, 3)
+SYSCALL(sys_getsockopt, 5)
+SYSCALL(sys_listen, 2)
+SYSCALL(sys_recv, 4)			/* 175 */
+SYSCALL(sys_recvfrom, 6)
+SYSCALL(sys_recvmsg, 3)
+SYSCALL(sys_send, 4)
+SYSCALL(sys_sendmsg, 3)
+SYSCALL(sys_sendto, 6)			/* 180 */
+SYSCALL(sys_setsockopt, 5)
+SYSCALL(sys_shutdown, 2)
+SYSCALL(sys_socket, 3)
+SYSCALL(sys_socketpair, 4)
+SYSCALL(sys_setresuid, 3)		/* 185 */
+SYSCALL(sys_getresuid, 3)
+SYSCALL(sys_ni_syscall, 5)		/* old sys_query_module */
+SYSCALL(sys_poll, 3)
+SYSCALL(sys_nfsservctl, 3)
+SYSCALL(sys_setresgid, 3)		/* 190 */
+SYSCALL(sys_getresgid, 3)
+SYSCALL(sys_prctl, 5)
+SYSCALL(sys_rt_sigreturn, 0)
+SYSCALL(sys_rt_sigaction, 4)
+SYSCALL(sys_rt_sigprocmask, 4)		/* 195 */
+SYSCALL(sys_rt_sigpending, 2)
+SYSCALL(sys_rt_sigtimedwait, 4)
+SYSCALL(sys_rt_sigqueueinfo, 3)
+SYSCALL(sys_rt_sigsuspend, 0)
+SYSCALL(sys_pread64, 5)			/* 200 */
+SYSCALL(sys_pwrite64, 5)
+SYSCALL(sys_chown, 3)
+SYSCALL(sys_getcwd, 2)
+SYSCALL(sys_capget, 2)
+SYSCALL(sys_capset, 2)			/* 205 */
+SYSCALL(sys_sigaltstack, 0)
+SYSCALL(sys_sendfile, 4)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_ni_syscall, 0)
+SYSCALL(sys_mmap2, 6)			/* 210 */
+SYSCALL(sys_truncate64, 2)
+SYSCALL(sys_ftruncate64, 2)
+SYSCALL(sys_stat64, 2)
+SYSCALL(sys_lstat64, 2)
+SYSCALL(sys_fstat64, 2)			/* 215 */
+SYSCALL(sys_pivot_root, 2)
+SYSCALL(sys_mincore, 3)
+SYSCALL(sys_madvise, 3)
+SYSCALL(sys_getdents64, 3)
+SYSCALL(sys_vfork, 0)			/* 220 */
diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c
new file mode 100644
index 000000000000..e07287db5a40
--- /dev/null
+++ b/arch/xtensa/kernel/time.c
@@ -0,0 +1,227 @@
+/*
+ * arch/xtensa/kernel/time.c
+ *
+ * Timer and clock support.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ */
+
+#include <linux/config.h>
+#include <linux/errno.h>
+#include <linux/time.h>
+#include <linux/timex.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/irq.h>
+#include <linux/profile.h>
+#include <linux/delay.h>
+
+#include <asm/timex.h>
+#include <asm/platform.h>
+
+
+extern volatile unsigned long wall_jiffies;
+
+u64 jiffies_64 = INITIAL_JIFFIES;
+EXPORT_SYMBOL(jiffies_64);
+
+spinlock_t rtc_lock = SPIN_LOCK_UNLOCKED;
+EXPORT_SYMBOL(rtc_lock);
+
+
+#ifdef CONFIG_XTENSA_CALIBRATE_CCOUNT
+unsigned long ccount_per_jiffy;		/* per 1/HZ */
+unsigned long ccount_nsec;		/* nsec per ccount increment */
+#endif
+
+unsigned int last_ccount_stamp;
+static long last_rtc_update = 0;
+
+/*
+ * Scheduler clock - returns current tim in nanosec units.
+ */
+
+unsigned long long sched_clock(void)
+{
+	return (unsigned long long)jiffies * (1000000000 / HZ);
+}
+
+static irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs);
+static struct irqaction timer_irqaction = {
+	.handler =	timer_interrupt,
+	.flags =	SA_INTERRUPT,
+	.name =		"timer",
+};
+
+void __init time_init(void)
+{
+	time_t sec_o, sec_n = 0;
+
+	/* The platform must provide a function to calibrate the processor
+	 * speed for the CALIBRATE.
+	 */
+
+#if CONFIG_XTENSA_CALIBRATE_CCOUNT
+	printk("Calibrating CPU frequency ");
+	platform_calibrate_ccount();
+	printk("%d.%02d MHz\n", (int)ccount_per_jiffy/(1000000/HZ),
+			(int)(ccount_per_jiffy/(10000/HZ))%100);
+#endif
+
+	/* Set time from RTC (if provided) */
+
+	if (platform_get_rtc_time(&sec_o) == 0)
+		while (platform_get_rtc_time(&sec_n))
+			if (sec_o != sec_n)
+				break;
+
+	xtime.tv_nsec = 0;
+	last_rtc_update = xtime.tv_sec = sec_n;
+	last_ccount_stamp = get_ccount();
+
+	set_normalized_timespec(&wall_to_monotonic,
+		-xtime.tv_sec, -xtime.tv_nsec);
+
+	/* Initialize the linux timer interrupt. */
+
+	setup_irq(LINUX_TIMER_INT, &timer_irqaction);
+	set_linux_timer(get_ccount() + CCOUNT_PER_JIFFY);
+}
+
+
+int do_settimeofday(struct timespec *tv)
+{
+	time_t wtm_sec, sec = tv->tv_sec;
+	long wtm_nsec, nsec = tv->tv_nsec;
+	unsigned long ccount;
+
+	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
+		return -EINVAL;
+
+	write_seqlock_irq(&xtime_lock);
+
+	/* This is revolting. We need to set "xtime" correctly. However, the
+	 * value in this location is the value at the most recent update of
+	 * wall time.  Discover what correction gettimeofday() would have
+	 * made, and then undo it!
+	 */
+	ccount = get_ccount();
+	nsec -= (ccount - last_ccount_stamp) * CCOUNT_NSEC;
+	nsec -= (jiffies - wall_jiffies) * CCOUNT_PER_JIFFY * CCOUNT_NSEC;
+
+	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
+	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
+
+	set_normalized_timespec(&xtime, sec, nsec);
+	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
+
+	time_adjust = 0;                /* stop active adjtime() */
+	time_status |= STA_UNSYNC;
+	time_maxerror = NTP_PHASE_LIMIT;
+	time_esterror = NTP_PHASE_LIMIT;
+	write_sequnlock_irq(&xtime_lock);
+	return 0;
+}
+
+EXPORT_SYMBOL(do_settimeofday);
+
+
+void do_gettimeofday(struct timeval *tv)
+{
+	unsigned long flags;
+	unsigned long sec, usec, delta, lost, seq;
+
+	do {
+		seq = read_seqbegin_irqsave(&xtime_lock, flags);
+
+		delta = get_ccount() - last_ccount_stamp;
+		sec = xtime.tv_sec;
+		usec = (xtime.tv_nsec / NSEC_PER_USEC);
+
+		lost = jiffies - wall_jiffies;
+
+	} while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
+
+	usec += lost * (1000000UL/HZ) + (delta * CCOUNT_NSEC) / NSEC_PER_USEC;
+	for (; usec >= 1000000; sec++, usec -= 1000000)
+		;
+
+	tv->tv_sec = sec;
+	tv->tv_usec = usec;
+}
+
+EXPORT_SYMBOL(do_gettimeofday);
+
+/*
+ * The timer interrupt is called HZ times per second.
+ */
+
+irqreturn_t timer_interrupt (int irq, void *dev_id, struct pt_regs *regs)
+{
+
+	unsigned long next;
+
+	next = get_linux_timer();
+
+again:
+	while ((signed long)(get_ccount() - next) > 0) {
+
+		profile_tick(CPU_PROFILING, regs);
+#ifndef CONFIG_SMP
+		update_process_times(user_mode(regs));
+#endif
+
+		write_seqlock(&xtime_lock);
+
+		last_ccount_stamp = next;
+		next += CCOUNT_PER_JIFFY;
+		do_timer (regs); /* Linux handler in kernel/timer.c */
+
+		if ((time_status & STA_UNSYNC) == 0 &&
+		    xtime.tv_sec - last_rtc_update >= 659 &&
+		    abs((xtime.tv_nsec/1000)-(1000000-1000000/HZ))<5000000/HZ &&
+		    jiffies - wall_jiffies == 1) {
+
+			if (platform_set_rtc_time(xtime.tv_sec+1) == 0)
+				last_rtc_update = xtime.tv_sec+1;
+			else
+				/* Do it again in 60 s */
+				last_rtc_update += 60;
+		}
+		write_sequnlock(&xtime_lock);
+	}
+
+	/* NOTE: writing CCOMPAREn clears the interrupt.  */
+
+	set_linux_timer (next);
+
+	/* Make sure we didn't miss any tick... */
+
+	if ((signed long)(get_ccount() - next) > 0)
+		goto again;
+
+	/* Allow platform to do something usefull (Wdog). */
+
+	platform_heartbeat();
+
+	return IRQ_HANDLED;
+}
+
+#ifndef CONFIG_GENERIC_CALIBRATE_DELAY
+void __devinit calibrate_delay(void)
+{
+	loops_per_jiffy = CCOUNT_PER_JIFFY;
+	printk("Calibrating delay loop (skipped)... "
+	       "%lu.%02lu BogoMIPS preset\n",
+	       loops_per_jiffy/(1000000/HZ),
+	       (loops_per_jiffy/(10000/HZ)) % 100);
+}
+#endif
+
diff --git a/arch/xtensa/kernel/traps.c b/arch/xtensa/kernel/traps.c
new file mode 100644
index 000000000000..804246e743b1
--- /dev/null
+++ b/arch/xtensa/kernel/traps.c
@@ -0,0 +1,498 @@
+/*
+ * arch/xtensa/kernel/traps.c
+ *
+ * Exception handling.
+ *
+ * Derived from code with the following copyrights:
+ * Copyright (C) 1994 - 1999 by Ralf Baechle
+ * Modified for R3000 by Paul M. Antoine, 1995, 1996
+ * Complete output from die() by Ulf Carlsson, 1998
+ * Copyright (C) 1999 Silicon Graphics, Inc.
+ *
+ * Essentially rewritten for the Xtensa architecture port.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ * Chris Zankel	<chris@zankel.net>
+ * Marc Gauthier<marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Kevin Chea
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/stringify.h>
+#include <linux/kallsyms.h>
+
+#include <asm/ptrace.h>
+#include <asm/timex.h>
+#include <asm/uaccess.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+
+#ifdef CONFIG_KGDB
+extern int gdb_enter;
+extern int return_from_debug_flag;
+#endif
+
+/*
+ * Machine specific interrupt handlers
+ */
+
+extern void kernel_exception(void);
+extern void user_exception(void);
+
+extern void fast_syscall_kernel(void);
+extern void fast_syscall_user(void);
+extern void fast_alloca(void);
+extern void fast_unaligned(void);
+extern void fast_second_level_miss(void);
+extern void fast_store_prohibited(void);
+extern void fast_coprocessor(void);
+
+extern void do_illegal_instruction (struct pt_regs*);
+extern void do_interrupt (struct pt_regs*);
+extern void do_unaligned_user (struct pt_regs*);
+extern void do_multihit (struct pt_regs*, unsigned long);
+extern void do_page_fault (struct pt_regs*, unsigned long);
+extern void do_debug (struct pt_regs*);
+extern void system_call (struct pt_regs*);
+
+/*
+ * The vector table must be preceded by a save area (which
+ * implies it must be in RAM, unless one places RAM immediately
+ * before a ROM and puts the vector at the start of the ROM (!))
+ */
+
+#define KRNL		0x01
+#define USER		0x02
+
+#define COPROCESSOR(x)							\
+{ XCHAL_EXCCAUSE_COPROCESSOR ## x ## _DISABLED, USER, fast_coprocessor }
+
+typedef struct {
+	int cause;
+	int fast;
+	void* handler;
+} dispatch_init_table_t;
+
+dispatch_init_table_t __init dispatch_init_table[] = {
+
+{ XCHAL_EXCCAUSE_ILLEGAL_INSTRUCTION,	0,	   do_illegal_instruction},
+{ XCHAL_EXCCAUSE_SYSTEM_CALL,		KRNL,	   fast_syscall_kernel },
+{ XCHAL_EXCCAUSE_SYSTEM_CALL,		USER,	   fast_syscall_user },
+{ XCHAL_EXCCAUSE_SYSTEM_CALL,		0,	   system_call },
+/* XCHAL_EXCCAUSE_INSTRUCTION_FETCH unhandled */
+/* XCHAL_EXCCAUSE_LOAD_STORE_ERROR unhandled*/
+{ XCHAL_EXCCAUSE_LEVEL1_INTERRUPT,	0,	   do_interrupt },
+{ XCHAL_EXCCAUSE_ALLOCA,		USER|KRNL, fast_alloca },
+/* XCHAL_EXCCAUSE_INTEGER_DIVIDE_BY_ZERO unhandled */
+/* XCHAL_EXCCAUSE_PRIVILEGED unhandled */
+#if XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION
+#ifdef CONFIG_UNALIGNED_USER
+{ XCHAL_EXCCAUSE_UNALIGNED,		USER,	   fast_unaligned },
+#else
+{ XCHAL_EXCCAUSE_UNALIGNED,		0,	   do_unaligned_user },
+#endif
+{ XCHAL_EXCCAUSE_UNALIGNED,		KRNL,	   fast_unaligned },
+#endif
+{ XCHAL_EXCCAUSE_ITLB_MISS,		0,	   do_page_fault },
+{ XCHAL_EXCCAUSE_ITLB_MISS,		USER|KRNL, fast_second_level_miss},
+{ XCHAL_EXCCAUSE_ITLB_MULTIHIT,		0,	   do_multihit },
+{ XCHAL_EXCCAUSE_ITLB_PRIVILEGE,	0,	   do_page_fault },
+/* XCHAL_EXCCAUSE_SIZE_RESTRICTION unhandled */
+{ XCHAL_EXCCAUSE_FETCH_CACHE_ATTRIBUTE,	0,	   do_page_fault },
+{ XCHAL_EXCCAUSE_DTLB_MISS,		USER|KRNL, fast_second_level_miss},
+{ XCHAL_EXCCAUSE_DTLB_MISS,		0,	   do_page_fault },
+{ XCHAL_EXCCAUSE_DTLB_MULTIHIT,		0,	   do_multihit },
+{ XCHAL_EXCCAUSE_DTLB_PRIVILEGE,	0,	   do_page_fault },
+/* XCHAL_EXCCAUSE_DTLB_SIZE_RESTRICTION unhandled */
+{ XCHAL_EXCCAUSE_STORE_CACHE_ATTRIBUTE,	USER|KRNL, fast_store_prohibited },
+{ XCHAL_EXCCAUSE_STORE_CACHE_ATTRIBUTE,	0,	   do_page_fault },
+{ XCHAL_EXCCAUSE_LOAD_CACHE_ATTRIBUTE,	0,	   do_page_fault },
+/* XCCHAL_EXCCAUSE_FLOATING_POINT unhandled */
+#if (XCHAL_CP_MASK & 1)
+COPROCESSOR(0),
+#endif
+#if (XCHAL_CP_MASK & 2)
+COPROCESSOR(1),
+#endif
+#if (XCHAL_CP_MASK & 4)
+COPROCESSOR(2),
+#endif
+#if (XCHAL_CP_MASK & 8)
+COPROCESSOR(3),
+#endif
+#if (XCHAL_CP_MASK & 16)
+COPROCESSOR(4),
+#endif
+#if (XCHAL_CP_MASK & 32)
+COPROCESSOR(5),
+#endif
+#if (XCHAL_CP_MASK & 64)
+COPROCESSOR(6),
+#endif
+#if (XCHAL_CP_MASK & 128)
+COPROCESSOR(7),
+#endif
+{ EXCCAUSE_MAPPED_DEBUG,		0,		do_debug },
+{ -1, -1, 0 }
+
+};
+
+/* The exception table <exc_table> serves two functions:
+ * 1. it contains three dispatch tables (fast_user, fast_kernel, default-c)
+ * 2. it is a temporary memory buffer for the exception handlers.
+ */
+
+unsigned long exc_table[EXC_TABLE_SIZE/4];
+
+void die(const char*, struct pt_regs*, long);
+
+static inline void
+__die_if_kernel(const char *str, struct pt_regs *regs, long err)
+{
+	if (!user_mode(regs))
+		die(str, regs, err);
+}
+
+/*
+ * Unhandled Exceptions. Kill user task or panic if in kernel space.
+ */
+
+void do_unhandled(struct pt_regs *regs, unsigned long exccause)
+{
+	__die_if_kernel("Caught unhandled exception - should not happen",
+	    		regs, SIGKILL);
+
+	/* If in user mode, send SIGILL signal to current process */
+	printk("Caught unhandled exception in '%s' "
+	       "(pid = %d, pc = %#010lx) - should not happen\n"
+	       "\tEXCCAUSE is %ld\n",
+	       current->comm, current->pid, regs->pc, exccause);
+	force_sig(SIGILL, current);
+}
+
+/*
+ * Multi-hit exception. This if fatal!
+ */
+
+void do_multihit(struct pt_regs *regs, unsigned long exccause)
+{
+	die("Caught multihit exception", regs, SIGKILL);
+}
+
+/*
+ * Level-1 interrupt.
+ * We currently have no priority encoding.
+ */
+
+unsigned long ignored_level1_interrupts;
+extern void do_IRQ(int, struct pt_regs *);
+
+void do_interrupt (struct pt_regs *regs)
+{
+	unsigned long intread = get_sr (INTREAD);
+	unsigned long intenable = get_sr (INTENABLE);
+	int i, mask;
+
+	/* Handle all interrupts (no priorities).
+	 * (Clear the interrupt before processing, in case it's
+	 *  edge-triggered or software-generated)
+	 */
+
+	for (i=0, mask = 1; i < XCHAL_NUM_INTERRUPTS; i++, mask <<= 1) {
+		if (mask & (intread & intenable)) {
+			set_sr (mask, INTCLEAR);
+			do_IRQ (i,regs);
+		}
+	}
+}
+
+/*
+ * Illegal instruction. Fatal if in kernel space.
+ */
+
+void
+do_illegal_instruction(struct pt_regs *regs)
+{
+	__die_if_kernel("Illegal instruction in kernel", regs, SIGKILL);
+
+	/* If in user mode, send SIGILL signal to current process. */
+
+	printk("Illegal Instruction in '%s' (pid = %d, pc = %#010lx)\n",
+	    current->comm, current->pid, regs->pc);
+	force_sig(SIGILL, current);
+}
+
+
+/*
+ * Handle unaligned memory accesses from user space. Kill task.
+ *
+ * If CONFIG_UNALIGNED_USER is not set, we don't allow unaligned memory
+ * accesses causes from user space.
+ */
+
+#if XCHAL_UNALIGNED_LOAD_EXCEPTION || XCHAL_UNALIGNED_STORE_EXCEPTION
+#ifndef CONFIG_UNALIGNED_USER
+void
+do_unaligned_user (struct pt_regs *regs)
+{
+	siginfo_t info;
+
+	__die_if_kernel("Unhandled unaligned exception in kernel",
+	    		regs, SIGKILL);
+
+	current->thread.bad_vaddr = regs->excvaddr;
+	current->thread.error_code = -3;
+	printk("Unaligned memory access to %08lx in '%s' "
+	       "(pid = %d, pc = %#010lx)\n",
+	       regs->excvaddr, current->comm, current->pid, regs->pc);
+	info.si_signo = SIGBUS;
+	info.si_errno = 0;
+	info.si_code = BUS_ADRALN;
+	info.si_addr = (void *) regs->excvaddr;
+	force_sig_info(SIGSEGV, &info, current);
+
+}
+#endif
+#endif
+
+void
+do_debug(struct pt_regs *regs)
+{
+#ifdef CONFIG_KGDB
+	/* If remote debugging is configured AND enabled, we give control to
+	 * kgdb.  Otherwise, we fall through, perhaps giving control to the
+	 * native debugger.
+	 */
+
+	if (gdb_enter) {
+		extern void gdb_handle_exception(struct pt_regs *);
+		gdb_handle_exception(regs);
+		return_from_debug_flag = 1;
+		return;
+	}
+#endif
+
+	__die_if_kernel("Breakpoint in kernel", regs, SIGKILL);
+
+	/* If in user mode, send SIGTRAP signal to current process */
+
+	force_sig(SIGTRAP, current);
+}
+
+
+/*
+ * Initialize dispatch tables.
+ *
+ * The exception vectors are stored compressed the __init section in the
+ * dispatch_init_table. This function initializes the following three tables
+ * from that compressed table:
+ * - fast user		first dispatch table for user exceptions
+ * - fast kernel	first dispatch table for kernel exceptions
+ * - default C-handler	C-handler called by the default fast handler.
+ *
+ * See vectors.S for more details.
+ */
+
+#define set_handler(idx,handler) (exc_table[idx] = (unsigned long) (handler))
+
+void trap_init(void)
+{
+	int i;
+
+	/* Setup default vectors. */
+
+	for(i = 0; i < 64; i++) {
+		set_handler(EXC_TABLE_FAST_USER/4   + i, user_exception);
+		set_handler(EXC_TABLE_FAST_KERNEL/4 + i, kernel_exception);
+		set_handler(EXC_TABLE_DEFAULT/4 + i, do_unhandled);
+	}
+
+	/* Setup specific handlers. */
+
+	for(i = 0; dispatch_init_table[i].cause >= 0; i++) {
+
+		int fast = dispatch_init_table[i].fast;
+		int cause = dispatch_init_table[i].cause;
+		void *handler = dispatch_init_table[i].handler;
+
+		if (fast == 0)
+			set_handler (EXC_TABLE_DEFAULT/4 + cause, handler);
+		if (fast && fast & USER)
+			set_handler (EXC_TABLE_FAST_USER/4 + cause, handler);
+		if (fast && fast & KRNL)
+			set_handler (EXC_TABLE_FAST_KERNEL/4 + cause, handler);
+	}
+
+	/* Initialize EXCSAVE_1 to hold the address of the exception table. */
+
+	i = (unsigned long)exc_table;
+	__asm__ __volatile__("wsr  %0, "__stringify(EXCSAVE_1)"\n" : : "a" (i));
+}
+
+/*
+ * This function dumps the current valid window frame and other base registers.
+ */
+
+void show_regs(struct pt_regs * regs)
+{
+	int i, wmask;
+
+	wmask = regs->wmask & ~1;
+
+	for (i = 0; i < 32; i++) {
+		if (wmask & (1 << (i / 4)))
+			break;
+		if ((i % 8) == 0)
+			printk ("\n" KERN_INFO "a%02d: ", i);
+		printk("%08lx ", regs->areg[i]);
+	}
+	printk("\n");
+
+	printk("pc: %08lx, ps: %08lx, depc: %08lx, excvaddr: %08lx\n",
+	       regs->pc, regs->ps, regs->depc, regs->excvaddr);
+	printk("lbeg: %08lx, lend: %08lx lcount: %08lx, sar: %08lx\n",
+	       regs->lbeg, regs->lend, regs->lcount, regs->sar);
+	if (user_mode(regs))
+		printk("wb: %08lx, ws: %08lx, wmask: %08lx, syscall: %ld\n",
+		       regs->windowbase, regs->windowstart, regs->wmask,
+		       regs->syscall);
+}
+
+void show_trace(struct task_struct *task, unsigned long *sp)
+{
+	unsigned long a0, a1, pc;
+	unsigned long sp_start, sp_end;
+
+	a1 = (unsigned long)sp;
+
+	if (a1 == 0)
+		__asm__ __volatile__ ("mov %0, a1\n" : "=a"(a1));
+
+
+	sp_start = a1 & ~(THREAD_SIZE-1);
+	sp_end = sp_start + THREAD_SIZE;
+
+	printk("Call Trace:");
+#ifdef CONFIG_KALLSYMS
+	printk("\n");
+#endif
+	spill_registers();
+
+	while (a1 > sp_start && a1 < sp_end) {
+		sp = (unsigned long*)a1;
+
+		a0 = *(sp - 4);
+		a1 = *(sp - 3);
+
+		if (a1 <= (unsigned long) sp)
+			break;
+
+		pc = MAKE_PC_FROM_RA(a0, a1);
+
+		if (kernel_text_address(pc)) {
+			printk(" [<%08lx>] ", pc);
+			print_symbol("%s\n", pc);
+		}
+	}
+	printk("\n");
+}
+
+/*
+ * This routine abuses get_user()/put_user() to reference pointers
+ * with at least a bit of error checking ...
+ */
+
+static int kstack_depth_to_print = 24;
+
+void show_stack(struct task_struct *task, unsigned long *sp)
+{
+	int i = 0;
+	unsigned long *stack;
+
+	if (sp == 0)
+		__asm__ __volatile__ ("mov %0, a1\n" : "=a"(sp));
+
+ 	stack = sp;
+
+	printk("\nStack: ");
+
+	for (i = 0; i < kstack_depth_to_print; i++) {
+		if (kstack_end(sp))
+			break;
+		if (i && ((i % 8) == 0))
+			printk("\n       ");
+		printk("%08lx ", *sp++);
+	}
+	printk("\n");
+	show_trace(task, stack);
+}
+
+void dump_stack(void)
+{
+	show_stack(current, NULL);
+}
+
+EXPORT_SYMBOL(dump_stack);
+
+
+void show_code(unsigned int *pc)
+{
+	long i;
+
+	printk("\nCode:");
+
+	for(i = -3 ; i < 6 ; i++) {
+		unsigned long insn;
+		if (__get_user(insn, pc + i)) {
+			printk(" (Bad address in pc)\n");
+			break;
+		}
+		printk("%c%08lx%c",(i?' ':'<'),insn,(i?' ':'>'));
+	}
+}
+
+spinlock_t die_lock = SPIN_LOCK_UNLOCKED;
+
+void die(const char * str, struct pt_regs * regs, long err)
+{
+	static int die_counter;
+	int nl = 0;
+
+	console_verbose();
+	spin_lock_irq(&die_lock);
+
+	printk("%s: sig: %ld [#%d]\n", str, err, ++die_counter);
+#ifdef CONFIG_PREEMPT
+	printk("PREEMPT ");
+	nl = 1;
+#endif
+	if (nl)
+		printk("\n");
+	show_regs(regs);
+	if (!user_mode(regs))
+		show_stack(NULL, (unsigned long*)regs->areg[1]);
+
+	spin_unlock_irq(&die_lock);
+
+	if (in_interrupt())
+		panic("Fatal exception in interrupt");
+
+	if (panic_on_oops) {
+		printk(KERN_EMERG "Fatal exception: panic in 5 seconds\n");
+		set_current_state(TASK_UNINTERRUPTIBLE);
+		schedule_timeout(5 * HZ);
+		panic("Fatal exception");
+	}
+	do_exit(err);
+}
+
+
diff --git a/arch/xtensa/kernel/vectors.S b/arch/xtensa/kernel/vectors.S
new file mode 100644
index 000000000000..81808f0c6742
--- /dev/null
+++ b/arch/xtensa/kernel/vectors.S
@@ -0,0 +1,464 @@
+/*
+ * arch/xtensa/kernel/vectors.S
+ *
+ * This file contains all exception vectors (user, kernel, and double),
+ * as well as the window vectors (overflow and underflow), and the debug
+ * vector. These are the primary vectors executed by the processor if an
+ * exception occurs.
+ *
+ * This file is subject to the terms and conditions of the GNU General
+ * Public License.  See the file "COPYING" in the main directory of
+ * this archive for more details.
+ *
+ * Copyright (C) 2005 Tensilica, Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ *
+ */
+
+/*
+ * We use a two-level table approach. The user and kernel exception vectors
+ * use a first-level dispatch table to dispatch the exception to a registered
+ * fast handler or the default handler, if no fast handler was registered.
+ * The default handler sets up a C-stack and dispatches the exception to a
+ * registerd C handler in the second-level dispatch table.
+ *
+ * Fast handler entry condition:
+ *
+ *   a0:	trashed, original value saved on stack (PT_AREG0)
+ *   a1:	a1
+ *   a2:	new stack pointer, original value in depc
+ *   a3:	dispatch table
+ *   depc:	a2, original value saved on stack (PT_DEPC)
+ *   excsave_1:	a3
+ *
+ * The value for PT_DEPC saved to stack also functions as a boolean to
+ * indicate that the exception is either a double or a regular exception:
+ *
+ *   PT_DEPC	>= VALID_DOUBLE_EXCEPTION_ADDRESS: double exception
+ *		<  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
+ *
+ * Note:  Neither the kernel nor the user exception handler generate literals.
+ *
+ */
+
+#include <linux/linkage.h>
+#include <asm/ptrace.h>
+#include <asm/ptrace.h>
+#include <asm/current.h>
+#include <asm/offsets.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/page.h>
+#include <asm/thread_info.h>
+#include <asm/processor.h>
+
+
+/*
+ * User exception vector. (Exceptions with PS.UM == 1, PS.EXCM == 0)
+ *
+ * We get here when an exception occurred while we were in userland.
+ * We switch to the kernel stack and jump to the first level handler
+ * associated to the exception cause.
+ *
+ * Note: the saved kernel stack pointer (EXC_TABLE_KSTK) is already
+ *       decremented by PT_USER_SIZE.
+ */
+
+	.section .UserExceptionVector.text, "ax"
+
+ENTRY(_UserExceptionVector)
+
+	xsr	a3, EXCSAVE_1		# save a3 and get dispatch table
+	wsr	a2, DEPC		# save a2
+	l32i	a2, a3, EXC_TABLE_KSTK	# load kernel stack to a2
+	s32i	a0, a2, PT_AREG0	# save a0 to ESF
+	rsr	a0, EXCCAUSE		# retrieve exception cause
+	s32i	a0, a2, PT_DEPC		# mark it as a regular exception
+	addx4	a0, a0, a3		# find entry in table
+	l32i	a0, a0, EXC_TABLE_FAST_USER	# load handler
+	jx	a0
+
+/*
+ * Kernel exception vector. (Exceptions with PS.UM == 0, PS.EXCM == 0)
+ *
+ * We get this exception when we were already in kernel space.
+ * We decrement the current stack pointer (kernel) by PT_SIZE and
+ * jump to the first-level handler associated with the exception cause.
+ *
+ * Note: we need to preserve space for the spill region.
+ */
+
+	.section .KernelExceptionVector.text, "ax"
+
+ENTRY(_KernelExceptionVector)
+
+	xsr	a3, EXCSAVE_1		# save a3, and get dispatch table
+	wsr	a2, DEPC		# save a2
+	addi	a2, a1, -16-PT_SIZE	# adjust stack pointer
+	s32i	a0, a2, PT_AREG0	# save a0 to ESF
+	rsr	a0, EXCCAUSE		# retrieve exception cause
+	s32i	a0, a2, PT_DEPC		# mark it as a regular exception
+	addx4	a0, a0, a3		# find entry in table
+	l32i	a0, a0, EXC_TABLE_FAST_KERNEL	# load handler address
+	jx	a0
+
+
+/*
+ * Double exception vector (Exceptions with PS.EXCM == 1)
+ * We get this exception when another exception occurs while were are
+ * already in an exception, such as window overflow/underflow exception,
+ * or 'expected' exceptions, for example memory exception when we were trying
+ * to read data from an invalid address in user space.
+ *
+ * Note that this vector is never invoked for level-1 interrupts, because such
+ * interrupts are disabled (masked) when PS.EXCM is set.
+ *
+ * We decode the exception and take the appropriate action.  However, the
+ * double exception vector is much more careful, because a lot more error
+ * cases go through the double exception vector than through the user and
+ * kernel exception vectors.
+ *
+ * Occasionally, the kernel expects a double exception to occur.  This usually
+ * happens when accessing user-space memory with the user's permissions
+ * (l32e/s32e instructions).  The kernel state, though, is not always suitable
+ * for immediate transfer of control to handle_double, where "normal" exception
+ * processing occurs. Also in kernel mode, TLB misses can occur if accessing
+ * vmalloc memory, possibly requiring repair in a double exception handler.
+ *
+ * The variable at TABLE_FIXUP offset from the pointer in EXCSAVE_1 doubles as
+ * a boolean variable and a pointer to a fixup routine. If the variable
+ * EXC_TABLE_FIXUP is non-zero, this handler jumps to that address. A value of
+ * zero indicates to use the default kernel/user exception handler.
+ * There is only one exception, when the value is identical to the exc_table
+ * label, the kernel is in trouble. This mechanism is used to protect critical
+ * sections, mainly when the handler writes to the stack to assert the stack
+ * pointer is valid. Once the fixup/default handler leaves that area, the
+ * EXC_TABLE_FIXUP variable is reset to the fixup handler or zero.
+ *
+ * Procedures wishing to use this mechanism should set EXC_TABLE_FIXUP to the
+ * nonzero address of a fixup routine before it could cause a double exception
+ * and reset it before it returns.
+ *
+ * Some other things to take care of when a fast exception handler doesn't
+ * specify a particular fixup handler but wants to use the default handlers:
+ *
+ *  - The original stack pointer (in a1) must not be modified. The fast
+ *    exception handler should only use a2 as the stack pointer.
+ *
+ *  - If the fast handler manipulates the stack pointer (in a2), it has to
+ *    register a valid fixup handler and cannot use the default handlers.
+ *
+ *  - The handler can use any other generic register from a3 to a15, but it
+ *    must save the content of these registers to stack (PT_AREG3...PT_AREGx)
+ *
+ *  - These registers must be saved before a double exception can occur.
+ *
+ *  - If we ever implement handling signals while in double exceptions, the
+ *    number of registers a fast handler has saved (excluding a0 and a1) must
+ *    be written to  PT_AREG1. (1 if only a3 is used, 2 for a3 and a4, etc. )
+ *
+ * The fixup handlers are special handlers:
+ *
+ *  - Fixup entry conditions differ from regular exceptions:
+ *
+ *	a0:	   DEPC
+ *	a1: 	   a1
+ *	a2:	   trashed, original value in EXC_TABLE_DOUBLE_A2
+ *	a3:	   exctable
+ *	depc:	   a0
+ *	excsave_1: a3
+ *
+ *  - When the kernel enters the fixup handler, it still assumes it is in a
+ *    critical section, so EXC_TABLE_FIXUP variable is set to exc_table.
+ *    The fixup handler, therefore, has to re-register itself as the fixup
+ *    handler before it returns from the double exception.
+ *
+ *  - Fixup handler can share the same exception frame with the fast handler.
+ *    The kernel stack pointer is not changed when entering the fixup handler.
+ *
+ *  - Fixup handlers can jump to the default kernel and user exception
+ *    handlers. Before it jumps, though, it has to setup a exception frame
+ *    on stack. Because the default handler resets the register fixup handler
+ *    the fixup handler must make sure that the default handler returns to
+ *    it instead of the exception address, so it can re-register itself as
+ *    the fixup handler.
+ *
+ * In case of a critical condition where the kernel cannot recover, we jump
+ * to unrecoverable_exception with the following entry conditions.
+ * All registers a0...a15 are unchanged from the last exception, except:
+ *
+ *	a0:	   last address before we jumped to the unrecoverable_exception.
+ *	excsave_1: a0
+ *
+ *
+ * See the handle_alloca_user and spill_registers routines for example clients.
+ *
+ * FIXME: Note: we currently don't allow signal handling coming from a double
+ *        exception, so the item markt with (*) is not required.
+ */
+
+	.section .DoubleExceptionVector.text, "ax"
+	.begin literal_prefix .DoubleExceptionVector
+
+ENTRY(_DoubleExceptionVector)
+
+	/* Deliberately destroy excsave (don't assume it's value was valid). */
+
+	wsr	a3, EXCSAVE_1		# save a3
+
+	/* Check for kernel double exception (usually fatal). */
+
+	rsr	a3, PS
+	_bbci.l	a3, PS_UM_SHIFT, .Lksp
+
+	/* Check if we are currently handling a window exception. */
+	/* Note: We don't need to indicate that we enter a critical section. */
+
+	xsr	a0, DEPC		# get DEPC, save a0
+
+	movi	a3, XCHAL_WINDOW_VECTORS_VADDR
+	_bltu	a0, a3, .Lfixup
+	addi	a3, a3, XSHAL_WINDOW_VECTORS_SIZE
+	_bgeu	a0, a3, .Lfixup
+
+	/* Window overflow/underflow exception. Get stack pointer. */
+
+	mov	a3, a2
+	movi	a2, exc_table
+	l32i	a2, a2, EXC_TABLE_KSTK
+
+	/* Check for overflow/underflow exception, jump if overflow. */
+
+	_bbci.l	a0, 6, .Lovfl
+
+	/* a0: depc, a1: a1, a2: kstk, a3: a2, depc: a0, excsave: a3  */
+
+	/* Restart window underflow exception.
+	 * We return to the instruction in user space that caused the window
+	 * underflow exception. Therefore, we change window base to the value
+	 * before we entered the window underflow exception and prepare the
+	 * registers to return as if we were coming from a regular exception
+	 * by changing depc (in a0).
+	 * Note: We can trash the current window frame (a0...a3) and depc!
+	 */
+
+	wsr	a2, DEPC		# save stack pointer temporarily
+	rsr	a0, PS
+	extui	a0, a0, XCHAL_PS_OWB_SHIFT, XCHAL_PS_OWB_BITS
+	wsr	a0, WINDOWBASE
+	rsync
+
+	/* We are now in the previous window frame. Save registers again. */
+
+	xsr	a2, DEPC		# save a2 and get stack pointer
+	s32i	a0, a2, PT_AREG0
+
+	wsr	a3, EXCSAVE_1		# save a3
+	movi	a3, exc_table
+
+	rsr	a0, EXCCAUSE
+	s32i	a0, a2, PT_DEPC		# mark it as a regular exception
+	addx4	a0, a0, a3
+	l32i	a0, a0, EXC_TABLE_FAST_USER
+	jx	a0
+
+.Lfixup:/* Check for a fixup handler or if we were in a critical section. */
+
+	/* a0: depc, a1: a1, a2: a2, a3: trashed, depc: a0, excsave1: a3 */
+
+	movi	a3, exc_table
+	s32i	a2, a3, EXC_TABLE_DOUBLE_SAVE	# temporary variable
+
+	/* Enter critical section. */
+
+	l32i	a2, a3, EXC_TABLE_FIXUP
+	s32i	a3, a3, EXC_TABLE_FIXUP
+	beq	a2, a3, .Lunrecoverable_fixup	# critical!
+	beqz	a2, .Ldflt			# no handler was registered
+
+	/* a0: depc, a1: a1, a2: trash, a3: exctable, depc: a0, excsave: a3 */
+
+	jx	a2
+
+.Ldflt:	/* Get stack pointer. */
+
+	l32i	a3, a3, EXC_TABLE_DOUBLE_SAVE
+	addi	a2, a3, -PT_USER_SIZE
+
+.Lovfl:	/* Jump to default handlers. */
+
+	/* a0: depc, a1: a1, a2: kstk, a3: a2, depc: a0, excsave: a3 */
+
+	xsr	a3, DEPC
+	s32i	a0, a2, PT_DEPC
+	s32i	a3, a2, PT_AREG0
+
+	/* a0: avail, a1: a1, a2: kstk, a3: avail, depc: a2, excsave: a3 */
+
+	movi	a3, exc_table
+	rsr	a0, EXCCAUSE
+	addx4	a0, a0, a3
+	l32i	a0, a0, EXC_TABLE_FAST_USER
+	jx	a0
+
+	/*
+	 * We only allow the ITLB miss exception if we are in kernel space.
+	 * All other exceptions are unexpected and thus unrecoverable!
+	 */
+
+	.extern fast_second_level_miss_double_kernel
+
+.Lksp:	/* a0: a0, a1: a1, a2: a2, a3: trashed, depc: depc, excsave: a3 */
+
+	rsr	a3, EXCCAUSE
+	beqi	a3, XCHAL_EXCCAUSE_ITLB_MISS, 1f
+	addi	a3, a3, -XCHAL_EXCCAUSE_DTLB_MISS
+	bnez	a3, .Lunrecoverable
+1:	movi	a3, fast_second_level_miss_double_kernel
+	jx	a3
+
+	/* Critical! We can't handle this situation. PANIC! */
+
+	.extern unrecoverable_exception
+
+.Lunrecoverable_fixup:
+	l32i	a2, a3, EXC_TABLE_DOUBLE_SAVE
+	xsr	a0, DEPC
+
+.Lunrecoverable:
+	rsr	a3, EXCSAVE_1
+	wsr	a0, EXCSAVE_1
+	movi	a0, unrecoverable_exception
+	callx0	a0
+
+	.end literal_prefix
+
+
+/*
+ * Debug interrupt vector
+ *
+ * There is not much space here, so simply jump to another handler.
+ * EXCSAVE[DEBUGLEVEL] has been set to that handler.
+ */
+
+	.section .DebugInterruptVector.text, "ax"
+
+ENTRY(_DebugInterruptVector)
+	xsr	a0, EXCSAVE + XCHAL_DEBUGLEVEL
+	jx	a0
+
+
+
+/* Window overflow and underflow handlers.
+ * The handlers must be 64 bytes apart, first starting with the underflow
+ * handlers underflow-4 to underflow-12, then the overflow handlers
+ * overflow-4 to overflow-12.
+ *
+ * Note: We rerun the underflow handlers if we hit an exception, so
+ *	 we try to access any page that would cause a page fault early.
+ */
+
+	.section		.WindowVectors.text, "ax"
+
+
+/* 4-Register Window Overflow Vector (Handler) */
+
+	.align 64
+.global _WindowOverflow4
+_WindowOverflow4:
+	s32e	a0, a5, -16
+	s32e	a1, a5, -12
+	s32e	a2, a5,  -8
+	s32e	a3, a5,  -4
+	rfwo
+
+
+/* 4-Register Window Underflow Vector (Handler) */
+
+	.align 64
+.global _WindowUnderflow4
+_WindowUnderflow4:
+	l32e	a0, a5, -16
+	l32e	a1, a5, -12
+	l32e	a2, a5,  -8
+	l32e	a3, a5,  -4
+	rfwu
+
+
+/* 8-Register Window Overflow Vector (Handler) */
+
+	.align 64
+.global _WindowOverflow8
+_WindowOverflow8:
+	s32e	a0, a9, -16
+	l32e	a0, a1, -12
+	s32e	a2, a9,  -8
+	s32e	a1, a9, -12
+	s32e	a3, a9,  -4
+	s32e	a4, a0, -32
+	s32e	a5, a0, -28
+	s32e	a6, a0, -24
+	s32e	a7, a0, -20
+	rfwo
+
+/* 8-Register Window Underflow Vector (Handler) */
+
+	.align 64
+.global _WindowUnderflow8
+_WindowUnderflow8:
+	l32e	a1, a9, -12
+	l32e	a0, a9, -16
+	l32e	a7, a1, -12
+	l32e	a2, a9,  -8
+	l32e	a4, a7, -32
+	l32e	a3, a9,  -4
+	l32e	a5, a7, -28
+	l32e	a6, a7, -24
+	l32e	a7, a7, -20
+	rfwu
+
+
+/* 12-Register Window Overflow Vector (Handler) */
+
+	.align 64
+.global _WindowOverflow12
+_WindowOverflow12:
+	s32e	a0,  a13, -16
+	l32e	a0,  a1,  -12
+	s32e	a1,  a13, -12
+	s32e	a2,  a13,  -8
+	s32e	a3,  a13,  -4
+	s32e	a4,  a0,  -48
+	s32e	a5,  a0,  -44
+	s32e	a6,  a0,  -40
+	s32e	a7,  a0,  -36
+	s32e	a8,  a0,  -32
+	s32e	a9,  a0,  -28
+	s32e	a10, a0,  -24
+	s32e	a11, a0,  -20
+	rfwo
+
+/* 12-Register Window Underflow Vector (Handler) */
+
+	.align 64
+.global _WindowUnderflow12
+_WindowUnderflow12:
+	l32e	a1,  a13, -12
+	l32e	a0,  a13, -16
+	l32e	a11, a1,  -12
+	l32e	a2,  a13,  -8
+	l32e	a4,  a11, -48
+	l32e	a8,  a11, -32
+	l32e	a3,  a13,  -4
+	l32e	a5,  a11, -44
+	l32e	a6,  a11, -40
+	l32e	a7,  a11, -36
+	l32e	a9,  a11, -28
+	l32e	a10, a11, -24
+	l32e	a11, a11, -20
+	rfwu
+
+	.text
+
+
diff --git a/arch/xtensa/kernel/vmlinux.lds.S b/arch/xtensa/kernel/vmlinux.lds.S
new file mode 100644
index 000000000000..476b2b53cd01
--- /dev/null
+++ b/arch/xtensa/kernel/vmlinux.lds.S
@@ -0,0 +1,341 @@
+/*
+ * arch/xtensa/kernel/vmlinux.lds.S
+ *
+ * Xtensa linker script
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
+ */
+
+#include <asm-generic/vmlinux.lds.h>
+
+#include <linux/config.h>
+#define _NOCLANGUAGE
+#include <xtensa/config/core.h>
+#include <xtensa/config/system.h>
+OUTPUT_ARCH(xtensa)
+ENTRY(_start)
+
+#if XCHAL_MEMORY_ORDER == XTHAL_BIGENDIAN
+jiffies = jiffies_64 + 4;
+#else
+jiffies = jiffies_64;
+#endif
+
+#define KERNELOFFSET 0x1000
+
+/* Note: In the following macros, it would be nice to specify only the
+   vector name and section kind and construct "sym" and "section" using
+   CPP concatenation, but that does not work reliably.  Concatenating a
+   string with "." produces an invalid token.  CPP will not print a
+   warning because it thinks this is an assembly file, but it leaves
+   them as multiple tokens and there may or may not be whitespace
+   between them.  */
+
+/* Macro for a relocation entry */
+
+#define RELOCATE_ENTRY(sym, section)		\
+	LONG(sym ## _start);			\
+	LONG(sym ## _end);			\
+	LONG(LOADADDR(section))
+
+/* Macro to define a section for a vector.
+ *
+ * Use of the MIN function catches the types of errors illustrated in
+ * the following example:
+ *
+ * Assume the section .DoubleExceptionVector.literal is completely
+ * full.  Then a programmer adds code to .DoubleExceptionVector.text
+ * that produces another literal.  The final literal position will
+ * overlay onto the first word of the adjacent code section
+ * .DoubleExceptionVector.text.  (In practice, the literals will
+ * overwrite the code, and the first few instructions will be
+ * garbage.)
+ */
+
+#define SECTION_VECTOR(sym, section, addr, max_prevsec_size, prevsec)       \
+  section addr : AT((MIN(LOADADDR(prevsec) + max_prevsec_size,		    \
+		         LOADADDR(prevsec) + SIZEOF(prevsec)) + 3) & ~ 3)   \
+  {									    \
+    . = ALIGN(4);							    \
+    sym ## _start = ABSOLUTE(.);		 			    \
+    *(section)								    \
+    sym ## _end = ABSOLUTE(.);						    \
+  }
+
+/*
+ *  Mapping of input sections to output sections when linking.
+ */
+
+SECTIONS
+{
+  . = XCHAL_KSEG_CACHED_VADDR + KERNELOFFSET;
+  /* .text section */
+
+  _text = .;
+  _stext = .;
+  _ftext = .;
+
+  .text :
+  {
+    /* The .head.text section must be the first section! */
+    *(.head.text)
+    *(.literal .text)
+    *(.srom.text)
+    VMLINUX_SYMBOL(__sched_text_start) = .;
+    *(.sched.text.literal .sched.text)
+    VMLINUX_SYMBOL(__sched_text_end) = .;
+    VMLINUX_SYMBOL(__lock_text_start) = .;
+    *(.spinlock.text.literal .spinlock.text)
+    VMLINUX_SYMBOL(__lock_text_end) = .;
+
+  }
+  _etext = .;
+
+  . = ALIGN(16);
+
+  RODATA
+
+  /*  Relocation table */
+
+  . = ALIGN(16);
+  __boot_reloc_table_start = ABSOLUTE(.);
+
+  __relocate : {
+
+    RELOCATE_ENTRY(_WindowVectors_text,
+		   .WindowVectors.text);
+#if 0
+    RELOCATE_ENTRY(_KernelExceptionVector_literal,
+		   .KernelExceptionVector.literal);
+#endif
+    RELOCATE_ENTRY(_KernelExceptionVector_text,
+		   .KernelExceptionVector.text);
+#if 0
+    RELOCATE_ENTRY(_UserExceptionVector_literal,
+		   .UserExceptionVector.literal);
+#endif
+    RELOCATE_ENTRY(_UserExceptionVector_text,
+		   .UserExceptionVector.text);
+    RELOCATE_ENTRY(_DoubleExceptionVector_literal,
+		   .DoubleExceptionVector.literal);
+    RELOCATE_ENTRY(_DoubleExceptionVector_text,
+		   .DoubleExceptionVector.text);
+  }
+  __boot_reloc_table_end = ABSOLUTE(.) ;
+
+  .fixup   : { *(.fixup) }
+
+  . = ALIGN(16);
+
+  __ex_table : {
+    __start___ex_table = .;
+    *(__ex_table)
+    __stop___ex_table = .;
+  }
+
+  /* Data section */
+
+  . = ALIGN(XCHAL_ICACHE_LINESIZE);
+  _fdata = .;
+  .data :
+  {
+    *(.data) CONSTRUCTORS
+    . = ALIGN(XCHAL_ICACHE_LINESIZE);
+    *(.data.cacheline_aligned)
+  }
+
+  _edata = .;
+
+  /* The initial task */
+  . = ALIGN(8192);
+  .data.init_task : { *(.data.init_task) }
+
+  /* Initialization code and data: */
+
+  . = ALIGN(1<<XCHAL_MMU_MIN_PTE_PAGE_SIZE);
+  __init_begin = .;
+  .init.text : {
+  	_sinittext = .;
+	*(.init.text.literal) *(.init.text)
+	_einittext = .;
+  }
+
+  .init.data :
+  {
+    *(.init.data)
+    . = ALIGN(0x4);
+    __tagtable_begin = .;
+    *(.taglist)
+    __tagtable_end = .;
+  }
+
+  . = ALIGN(XCHAL_ICACHE_LINESIZE);
+
+  __setup_start = .;
+  .init.setup : { *(.init.setup) }
+  __setup_end = .;
+
+  __initcall_start = .;
+  .initcall.init : {
+  	*(.initcall1.init)
+  	*(.initcall2.init)
+  	*(.initcall3.init)
+  	*(.initcall4.init)
+  	*(.initcall5.init)
+  	*(.initcall6.init)
+  	*(.initcall7.init)
+  }
+  __initcall_end = .;
+
+  __con_initcall_start = .;
+  .con_initcall.init : { *(.con_initcall.init) }
+  __con_initcall_end = .;
+
+  SECURITY_INIT
+
+  . = ALIGN(4);
+
+  __start___ftr_fixup = .;
+  __ftr_fixup : { *(__ftr_fixup) }
+  __stop___ftr_fixup = .;
+
+  . = ALIGN(32);
+  __per_cpu_start = .;
+  .data.percpu  : { *(.data.percpu) }
+  __per_cpu_end = .;
+
+  . = ALIGN(4096);
+  __initramfs_start =.;
+  .init.ramfs : { *(.init.ramfs) }
+  __initramfs_end = .;
+
+  /* We need this dummy segment here */
+
+  . = ALIGN(4);
+  .dummy : { LONG(0) }
+
+  /* The vectors are relocated to the real position at startup time */
+
+  SECTION_VECTOR (_WindowVectors_text,
+		  .WindowVectors.text,
+		  XCHAL_WINDOW_VECTORS_VADDR, 4,
+		  .dummy)
+  SECTION_VECTOR (_DebugInterruptVector_literal,
+		  .DebugInterruptVector.literal,
+		  XCHAL_INTLEVEL_VECTOR_VADDR(XCHAL_DEBUGLEVEL) - 4,
+		  SIZEOF(.WindowVectors.text),
+		  .WindowVectors.text)
+  SECTION_VECTOR (_DebugInterruptVector_text,
+		  .DebugInterruptVector.text,
+		  XCHAL_INTLEVEL_VECTOR_VADDR(XCHAL_DEBUGLEVEL),
+		  4,
+		  .DebugInterruptVector.literal)
+  SECTION_VECTOR (_KernelExceptionVector_literal,
+		  .KernelExceptionVector.literal,
+		  XCHAL_KERNELEXC_VECTOR_VADDR - 4,
+		  SIZEOF(.DebugInterruptVector.text),
+		  .DebugInterruptVector.text)
+  SECTION_VECTOR (_KernelExceptionVector_text,
+		  .KernelExceptionVector.text,
+		  XCHAL_KERNELEXC_VECTOR_VADDR,
+		  4,
+		  .KernelExceptionVector.literal)
+  SECTION_VECTOR (_UserExceptionVector_literal,
+		  .UserExceptionVector.literal,
+		  XCHAL_USEREXC_VECTOR_VADDR - 4,
+		  SIZEOF(.KernelExceptionVector.text),
+		  .KernelExceptionVector.text)
+  SECTION_VECTOR (_UserExceptionVector_text,
+		  .UserExceptionVector.text,
+		  XCHAL_USEREXC_VECTOR_VADDR,
+		  4,
+		  .UserExceptionVector.literal)
+  SECTION_VECTOR (_DoubleExceptionVector_literal,
+		  .DoubleExceptionVector.literal,
+		  XCHAL_DOUBLEEXC_VECTOR_VADDR - 16,
+		  SIZEOF(.UserExceptionVector.text),
+		  .UserExceptionVector.text)
+  SECTION_VECTOR (_DoubleExceptionVector_text,
+		  .DoubleExceptionVector.text,
+		  XCHAL_DOUBLEEXC_VECTOR_VADDR,
+		  32,
+		  .DoubleExceptionVector.literal)
+
+  . = (LOADADDR( .DoubleExceptionVector.text ) + SIZEOF( .DoubleExceptionVector.text ) + 3) & ~ 3;
+  . = ALIGN(1<<XCHAL_MMU_MIN_PTE_PAGE_SIZE);
+
+  __init_end = .;
+
+  . = ALIGN(8192);
+
+  /* BSS section */
+  _bss_start = .;
+  .sbss : { *(.sbss) *(.scommon) }
+  .bss : { *(COMMON) *(.bss) }
+  _bss_end = .;
+  _end = .;
+
+  /* only used by the boot loader  */
+
+  . = ALIGN(0x10);
+  .bootstrap : { *(.bootstrap.literal .bootstrap.text .bootstrap.data) }
+
+  . = ALIGN(0x1000);
+  __initrd_start = .;
+  .initrd : { *(.initrd) }
+  __initrd_end = .;
+
+  .ResetVector.text XCHAL_RESET_VECTOR_VADDR :
+  {
+    *(.ResetVector.text)
+  }
+
+
+  /* Sections to be discarded */
+  /DISCARD/ :
+  {
+        *(.text.exit)
+	*(.text.exit.literal)
+        *(.data.exit)
+        *(.exitcall.exit)
+  }
+
+
+  .debug  0 :  { *(.debug) }
+  .line  0 :  { *(.line) }
+  .debug_srcinfo  0 :  { *(.debug_srcinfo) }
+  .debug_sfnames  0 :  { *(.debug_sfnames) }
+  .debug_aranges  0 :  { *(.debug_aranges) }
+  .debug_pubnames  0 :  { *(.debug_pubnames) }
+  .debug_info  0 :  { *(.debug_info) }
+  .debug_abbrev  0 :  { *(.debug_abbrev) }
+  .debug_line  0 :  { *(.debug_line) }
+  .debug_frame  0 :  { *(.debug_frame) }
+  .debug_str  0 :  { *(.debug_str) }
+  .debug_loc  0 :  { *(.debug_loc) }
+  .debug_macinfo  0 :  { *(.debug_macinfo) }
+  .debug_weaknames  0 :  { *(.debug_weaknames) }
+  .debug_funcnames  0 :  { *(.debug_funcnames) }
+  .debug_typenames  0 :  { *(.debug_typenames) }
+  .debug_varnames  0 :  { *(.debug_varnames) }
+
+  .xt.insn 0 :
+  {
+    *(.xt.insn)
+    *(.gnu.linkonce.x*)
+  }
+
+  .xt.lit 0 :
+  {
+    *(.xt.lit)
+    *(.gnu.linkonce.p*)
+  }
+}
diff --git a/arch/xtensa/kernel/xtensa_ksyms.c b/arch/xtensa/kernel/xtensa_ksyms.c
new file mode 100644
index 000000000000..efae56a51475
--- /dev/null
+++ b/arch/xtensa/kernel/xtensa_ksyms.c
@@ -0,0 +1,123 @@
+/*
+ * arch/xtensa/kernel/xtensa_ksyms.c
+ *
+ * Export Xtensa-specific functions for loadable modules.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005  Tensilica Inc.
+ *
+ * Joe Taylor <joe@tensilica.com>
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <asm/irq.h>
+#include <linux/in6.h>
+#include <linux/pci.h>
+#include <linux/ide.h>
+
+#include <asm/uaccess.h>
+#include <asm/checksum.h>
+#include <asm/dma.h>
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/semaphore.h>
+#ifdef CONFIG_BLK_DEV_FD
+#include <asm/floppy.h>
+#endif
+#ifdef CONFIG_NET
+#include <net/checksum.h>
+#endif /* CONFIG_NET */
+
+
+/*
+ * String functions
+ */
+EXPORT_SYMBOL(memcmp);
+EXPORT_SYMBOL(memset);
+EXPORT_SYMBOL(memcpy);
+EXPORT_SYMBOL(memmove);
+EXPORT_SYMBOL(memchr);
+EXPORT_SYMBOL(strcat);
+EXPORT_SYMBOL(strchr);
+EXPORT_SYMBOL(strlen);
+EXPORT_SYMBOL(strpbrk);
+EXPORT_SYMBOL(strncat);
+EXPORT_SYMBOL(strnlen);
+EXPORT_SYMBOL(strrchr);
+EXPORT_SYMBOL(strstr);
+
+EXPORT_SYMBOL(enable_irq);
+EXPORT_SYMBOL(disable_irq);
+EXPORT_SYMBOL(kernel_thread);
+
+/*
+ * gcc internal math functions
+ */
+extern long long __ashrdi3(long long, int);
+extern long long __ashldi3(long long, int);
+extern long long __lshrdi3(long long, int);
+extern int __divsi3(int, int);
+extern int __modsi3(int, int);
+extern long long __muldi3(long long, long long);
+extern int __mulsi3(int, int);
+extern unsigned int __udivsi3(unsigned int, unsigned int);
+extern unsigned int __umodsi3(unsigned int, unsigned int);
+extern unsigned long long __umoddi3(unsigned long long, unsigned long long);
+extern unsigned long long __udivdi3(unsigned long long, unsigned long long);
+
+EXPORT_SYMBOL(__ashldi3);
+EXPORT_SYMBOL(__ashrdi3);
+EXPORT_SYMBOL(__lshrdi3);
+EXPORT_SYMBOL(__divsi3);
+EXPORT_SYMBOL(__modsi3);
+EXPORT_SYMBOL(__muldi3);
+EXPORT_SYMBOL(__mulsi3);
+EXPORT_SYMBOL(__udivsi3);
+EXPORT_SYMBOL(__umodsi3);
+EXPORT_SYMBOL(__udivdi3);
+EXPORT_SYMBOL(__umoddi3);
+
+/*
+ * Semaphore operations
+ */
+EXPORT_SYMBOL(__down);
+EXPORT_SYMBOL(__down_interruptible);
+EXPORT_SYMBOL(__down_trylock);
+EXPORT_SYMBOL(__up);
+
+#ifdef CONFIG_NET
+/*
+ * Networking support
+ */
+EXPORT_SYMBOL(csum_partial_copy_generic);
+#endif /* CONFIG_NET */
+
+/*
+ * Architecture-specific symbols
+ */
+EXPORT_SYMBOL(__xtensa_copy_user);
+
+/*
+ * Kernel hacking ...
+ */
+
+#if defined(CONFIG_VGA_CONSOLE) || defined(CONFIG_DUMMY_CONSOLE)
+// FIXME EXPORT_SYMBOL(screen_info);
+#endif
+
+EXPORT_SYMBOL(get_wchan);
+
+EXPORT_SYMBOL(outsb);
+EXPORT_SYMBOL(outsw);
+EXPORT_SYMBOL(outsl);
+EXPORT_SYMBOL(insb);
+EXPORT_SYMBOL(insw);
+EXPORT_SYMBOL(insl);
diff --git a/arch/xtensa/lib/Makefile b/arch/xtensa/lib/Makefile
new file mode 100644
index 000000000000..ed935b58e8a4
--- /dev/null
+++ b/arch/xtensa/lib/Makefile
@@ -0,0 +1,7 @@
+#
+# Makefile for Xtensa-specific library files.
+#
+
+lib-y	+= memcopy.o memset.o checksum.o strcasecmp.o \
+	   usercopy.o strncpy_user.o strnlen_user.o
+lib-$(CONFIG_PCI) += pci-auto.o
diff --git a/arch/xtensa/lib/checksum.S b/arch/xtensa/lib/checksum.S
new file mode 100644
index 000000000000..e2d64dfd530c
--- /dev/null
+++ b/arch/xtensa/lib/checksum.S
@@ -0,0 +1,410 @@
+/*
+ * INET		An implementation of the TCP/IP protocol suite for the LINUX
+ *		operating system.  INET is implemented using the  BSD Socket
+ *		interface as the means of communication with the user level.
+ *
+ *		IP/TCP/UDP checksumming routines
+ *
+ * Xtensa version:  Copyright (C) 2001 Tensilica, Inc. by Kevin Chea
+ *                  Optimized by Joe Taylor
+ *
+ *		This program is free software; you can redistribute it and/or
+ *		modify it under the terms of the GNU General Public License
+ *		as published by the Free Software Foundation; either version
+ *		2 of the License, or (at your option) any later version.
+ */
+
+#include <asm/errno.h>
+#include <linux/linkage.h>
+#define _ASMLANGUAGE
+#include <xtensa/config/core.h>
+
+/*
+ * computes a partial checksum, e.g. for TCP/UDP fragments
+ */
+
+/*
+ * unsigned int csum_partial(const unsigned char *buf, int len,
+ *                           unsigned int sum);
+ *    a2 = buf
+ *    a3 = len
+ *    a4 = sum
+ *
+ * This function assumes 2- or 4-byte alignment.  Other alignments will fail!
+ */
+
+/* ONES_ADD converts twos-complement math to ones-complement. */
+#define ONES_ADD(sum, val)	  \
+	add	sum, sum, val	; \
+	bgeu	sum, val, 99f	; \
+	addi	sum, sum, 1	; \
+99:				;
+
+.text
+ENTRY(csum_partial)
+	  /*
+	   * Experiments with Ethernet and SLIP connections show that buf
+	   * is aligned on either a 2-byte or 4-byte boundary.
+	   */
+	entry	sp, 32
+	extui	a5, a2, 0, 2
+	bnez	a5, 8f		/* branch if 2-byte aligned */
+	/* Fall-through on common case, 4-byte alignment */
+1:
+	srli	a5, a3, 5	/* 32-byte chunks */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a5, 2f
+#else
+	beqz	a5, 2f
+	slli	a5, a5, 5
+	add	a5, a5, a2	/* a5 = end of last 32-byte chunk */
+.Loop1:
+#endif
+	l32i	a6, a2, 0
+	l32i	a7, a2, 4
+	ONES_ADD(a4, a6)
+	ONES_ADD(a4, a7)
+	l32i	a6, a2, 8
+	l32i	a7, a2, 12
+	ONES_ADD(a4, a6)
+	ONES_ADD(a4, a7)
+	l32i	a6, a2, 16
+	l32i	a7, a2, 20
+	ONES_ADD(a4, a6)
+	ONES_ADD(a4, a7)
+	l32i	a6, a2, 24
+	l32i	a7, a2, 28
+	ONES_ADD(a4, a6)
+	ONES_ADD(a4, a7)
+	addi	a2, a2, 4*8
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a5, .Loop1
+#endif
+2:
+	extui	a5, a3, 2, 3	/* remaining 4-byte chunks */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a5, 3f
+#else
+	beqz	a5, 3f
+	slli	a5, a5, 2
+	add	a5, a5, a2	/* a5 = end of last 4-byte chunk */
+.Loop2:
+#endif
+	l32i	a6, a2, 0
+	ONES_ADD(a4, a6)
+	addi	a2, a2, 4
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a5, .Loop2
+#endif
+3:
+	_bbci.l	a3, 1, 5f	/* remaining 2-byte chunk */
+	l16ui	a6, a2, 0
+	ONES_ADD(a4, a6)
+	addi	a2, a2, 2
+5:
+	_bbci.l	a3, 0, 7f	/* remaining 1-byte chunk */
+6:	l8ui	a6, a2, 0
+#ifdef __XTENSA_EB__
+	slli	a6, a6, 8	/* load byte into bits 8..15 */
+#endif
+	ONES_ADD(a4, a6)
+7:
+	mov	a2, a4
+	retw
+
+	/* uncommon case, buf is 2-byte aligned */
+8:
+	beqz	a3, 7b		/* branch if len == 0 */
+	beqi	a3, 1, 6b	/* branch if len == 1 */
+
+	extui	a5, a2, 0, 1
+	bnez	a5, 8f		/* branch if 1-byte aligned */
+
+	l16ui	a6, a2, 0	/* common case, len >= 2 */
+	ONES_ADD(a4, a6)
+	addi	a2, a2, 2	/* adjust buf */
+	addi	a3, a3, -2	/* adjust len */
+	j	1b		/* now buf is 4-byte aligned */
+
+	/* case: odd-byte aligned, len > 1
+	 * This case is dog slow, so don't give us an odd address.
+	 * (I don't think this ever happens, but just in case.)
+	 */
+8:
+	srli	a5, a3, 2	/* 4-byte chunks */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a5, 2f
+#else
+	beqz	a5, 2f
+	slli	a5, a5, 2
+	add	a5, a5, a2	/* a5 = end of last 4-byte chunk */
+.Loop3:
+#endif
+	l8ui	a6, a2, 0	/* bits 24..31 */
+	l16ui	a7, a2, 1	/* bits  8..23 */
+	l8ui	a8, a2, 3	/* bits  0.. 8 */
+#ifdef	__XTENSA_EB__
+	slli	a6, a6, 24
+#else
+	slli	a8, a8, 24
+#endif
+	slli	a7, a7, 8
+	or	a7, a7, a6
+	or	a7, a7, a8
+	ONES_ADD(a4, a7)
+	addi	a2, a2, 4
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a5, .Loop3
+#endif
+2:
+	_bbci.l	a3, 1, 3f	/* remaining 2-byte chunk, still odd addr */
+	l8ui	a6, a2, 0
+	l8ui	a7, a2, 1
+#ifdef	__XTENSA_EB__
+	slli	a6, a6, 8
+#else
+	slli	a7, a7, 8
+#endif
+	or	a7, a7, a6
+	ONES_ADD(a4, a7)
+	addi	a2, a2, 2
+3:
+	j	5b		/* branch to handle the remaining byte */
+
+
+
+/*
+ * Copy from ds while checksumming, otherwise like csum_partial
+ *
+ * The macros SRC and DST specify the type of access for the instruction.
+ * thus we can call a custom exception handler for each access type.
+ */
+
+#define SRC(y...)			\
+	9999: y;			\
+	.section __ex_table, "a";	\
+	.long 9999b, 6001f	;	\
+	.previous
+
+#define DST(y...)			\
+	9999: y;			\
+	.section __ex_table, "a";	\
+	.long 9999b, 6002f	;	\
+	.previous
+
+/*
+unsigned int csum_partial_copy_generic (const char *src, char *dst, int len,
+					int sum, int *src_err_ptr, int *dst_err_ptr)
+	a2  = src
+	a3  = dst
+	a4  = len
+	a5  = sum
+	a6  = src_err_ptr
+	a7  = dst_err_ptr
+	a8  = temp
+	a9  = temp
+	a10 = temp
+	a11 = original len for exception handling
+	a12 = original dst for exception handling
+
+    This function is optimized for 4-byte aligned addresses.  Other
+    alignments work, but not nearly as efficiently.
+ */
+
+ENTRY(csum_partial_copy_generic)
+	entry	sp, 32
+	mov	a12, a3
+	mov	a11, a4
+	or	a10, a2, a3
+
+	/* We optimize the following alignment tests for the 4-byte
+	aligned case.  Two bbsi.l instructions might seem more optimal
+	(commented out below).  However, both labels 5: and 3: are out
+	of the imm8 range, so the assembler relaxes them into
+	equivalent bbci.l, j combinations, which is actually
+	slower. */
+
+	extui	a9, a10, 0, 2
+	beqz	a9, 1f		/* branch if both are 4-byte aligned */
+	bbsi.l	a10, 0, 5f	/* branch if one address is odd */
+	j	3f		/* one address is 2-byte aligned */
+
+/*	_bbsi.l	a10, 0, 5f */	/* branch if odd address */
+/*	_bbsi.l	a10, 1, 3f */	/* branch if 2-byte-aligned address */
+
+1:
+	/* src and dst are both 4-byte aligned */
+	srli	a10, a4, 5	/* 32-byte chunks */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a10, 2f
+#else
+	beqz	a10, 2f
+	slli	a10, a10, 5
+	add	a10, a10, a2	/* a10 = end of last 32-byte src chunk */
+.Loop5:
+#endif
+SRC(	l32i	a9, a2, 0	)
+SRC(	l32i	a8, a2, 4	)
+DST(	s32i	a9, a3, 0	)
+DST(	s32i	a8, a3, 4	)
+	ONES_ADD(a5, a9)
+	ONES_ADD(a5, a8)
+SRC(	l32i	a9, a2, 8	)
+SRC(	l32i	a8, a2, 12	)
+DST(	s32i	a9, a3, 8	)
+DST(	s32i	a8, a3, 12	)
+	ONES_ADD(a5, a9)
+	ONES_ADD(a5, a8)
+SRC(	l32i	a9, a2, 16	)
+SRC(	l32i	a8, a2, 20	)
+DST(	s32i	a9, a3, 16	)
+DST(	s32i	a8, a3, 20	)
+	ONES_ADD(a5, a9)
+	ONES_ADD(a5, a8)
+SRC(	l32i	a9, a2, 24	)
+SRC(	l32i	a8, a2, 28	)
+DST(	s32i	a9, a3, 24	)
+DST(	s32i	a8, a3, 28	)
+	ONES_ADD(a5, a9)
+	ONES_ADD(a5, a8)
+	addi	a2, a2, 32
+	addi	a3, a3, 32
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a10, .Loop5
+#endif
+2:
+	extui	a10, a4, 2, 3	/* remaining 4-byte chunks */
+	extui	a4, a4, 0, 2	/* reset len for general-case, 2-byte chunks */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a10, 3f
+#else
+	beqz	a10, 3f
+	slli	a10, a10, 2
+	add	a10, a10, a2	/* a10 = end of last 4-byte src chunk */
+.Loop6:
+#endif
+SRC(	l32i	a9, a2, 0	)
+DST(	s32i	a9, a3, 0	)
+	ONES_ADD(a5, a9)
+	addi	a2, a2, 4
+	addi	a3, a3, 4
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a10, .Loop6
+#endif
+3:
+	/*
+	Control comes to here in two cases: (1) It may fall through
+	to here from the 4-byte alignment case to process, at most,
+	one 2-byte chunk.  (2) It branches to here from above if
+	either src or dst is 2-byte aligned, and we process all bytes
+	here, except for perhaps a trailing odd byte.  It's
+	inefficient, so align your addresses to 4-byte boundaries.
+
+	a2 = src
+	a3 = dst
+	a4 = len
+	a5 = sum
+	*/
+	srli	a10, a4, 1	/* 2-byte chunks */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a10, 4f
+#else
+	beqz	a10, 4f
+	slli	a10, a10, 1
+	add	a10, a10, a2	/* a10 = end of last 2-byte src chunk */
+.Loop7:
+#endif
+SRC(	l16ui	a9, a2, 0	)
+DST(	s16i	a9, a3, 0	)
+	ONES_ADD(a5, a9)
+	addi	a2, a2, 2
+	addi	a3, a3, 2
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a10, .Loop7
+#endif
+4:
+	/* This section processes a possible trailing odd byte. */
+	_bbci.l	a4, 0, 8f	/* 1-byte chunk */
+SRC(	l8ui	a9, a2, 0	)
+DST(	s8i	a9, a3, 0	)
+#ifdef __XTENSA_EB__
+	slli	a9, a9, 8	/* shift byte to bits 8..15 */
+#endif
+	ONES_ADD(a5, a9)
+8:
+	mov	a2, a5
+	retw
+
+5:
+	/* Control branch to here when either src or dst is odd.  We
+	process all bytes using 8-bit accesses.  Grossly inefficient,
+	so don't feed us an odd address. */
+
+	srli	a10, a4, 1	/* handle in pairs for 16-bit csum */
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a10, 6f
+#else
+	beqz	a10, 6f
+	slli	a10, a10, 1
+	add	a10, a10, a2	/* a10 = end of last odd-aligned, 2-byte src chunk */
+.Loop8:
+#endif
+SRC(	l8ui	a9, a2, 0	)
+SRC(	l8ui	a8, a2, 1	)
+DST(	s8i	a9, a3, 0	)
+DST(	s8i	a8, a3, 1	)
+#ifdef __XTENSA_EB__
+	slli	a9, a9, 8	/* combine into a single 16-bit value */
+#else				/* for checksum computation */
+	slli	a8, a8, 8
+#endif
+	or	a9, a9, a8
+	ONES_ADD(a5, a9)
+	addi	a2, a2, 2
+	addi	a3, a3, 2
+#if !XCHAL_HAVE_LOOPS
+	blt	a2, a10, .Loop8
+#endif
+6:
+	j	4b		/* process the possible trailing odd byte */
+
+
+# Exception handler:
+.section .fixup, "ax"
+/*
+	a6  = src_err_ptr
+	a7  = dst_err_ptr
+	a11 = original len for exception handling
+	a12 = original dst for exception handling
+*/
+
+6001:
+	_movi	a2, -EFAULT
+	s32i	a2, a6, 0	/* src_err_ptr */
+
+	# clear the complete destination - computing the rest
+	# is too much work
+	movi	a2, 0
+#if XCHAL_HAVE_LOOPS
+	loopgtz	a11, 2f
+#else
+	beqz	a11, 2f
+	add	a11, a11, a12	/* a11 = ending address */
+.Leloop:
+#endif
+	s8i	a2, a12, 0
+	addi	a12, a12, 1
+#if !XCHAL_HAVE_LOOPS
+	blt	a12, a11, .Leloop
+#endif
+2:
+	retw
+
+6002:
+	movi	a2, -EFAULT
+	s32i	a2, a7, 0	/* dst_err_ptr */
+	movi	a2, 0
+	retw
+
+.previous
+
diff --git a/arch/xtensa/lib/memcopy.S b/arch/xtensa/lib/memcopy.S
new file mode 100644
index 000000000000..e8f6d7eb7222
--- /dev/null
+++ b/arch/xtensa/lib/memcopy.S
@@ -0,0 +1,315 @@
+/*
+ * arch/xtensa/lib/hal/memcopy.S -- Core HAL library functions
+ * xthal_memcpy and xthal_bcopy
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2002 - 2005 Tensilica Inc.
+ */
+
+#include <xtensa/coreasm.h>
+
+	.macro	src_b	r, w0, w1
+#ifdef __XTENSA_EB__
+	src	\r, \w0, \w1
+#else
+	src	\r, \w1, \w0
+#endif
+	.endm
+
+	.macro	ssa8	r
+#ifdef __XTENSA_EB__
+	ssa8b	\r
+#else
+	ssa8l	\r
+#endif
+	.endm
+
+
+/*
+ * void *memcpy(void *dst, const void *src, size_t len);
+ * void *memmove(void *dst, const void *src, size_t len);
+ * void *bcopy(const void *src, void *dst, size_t len);
+ *
+ * This function is intended to do the same thing as the standard
+ * library function memcpy() (or bcopy()) for most cases.
+ * However, where the source and/or destination references
+ * an instruction RAM or ROM or a data RAM or ROM, that
+ * source and/or destination will always be accessed with
+ * 32-bit load and store instructions (as required for these
+ * types of devices).
+ *
+ * !!!!!!!  XTFIXME:
+ * !!!!!!!  Handling of IRAM/IROM has not yet
+ * !!!!!!!  been implemented.
+ *
+ * The bcopy version is provided here to avoid the overhead
+ * of an extra call, for callers that require this convention.
+ *
+ * The (general case) algorithm is as follows:
+ *   If destination is unaligned, align it by conditionally
+ *     copying 1 and 2 bytes.
+ *   If source is aligned,
+ *     do 16 bytes with a loop, and then finish up with
+ *     8, 4, 2, and 1 byte copies conditional on the length;
+ *   else (if source is unaligned),
+ *     do the same, but use SRC to align the source data.
+ *   This code tries to use fall-through branches for the common
+ *     case of aligned source and destination and multiple
+ *     of 4 (or 8) length.
+ *
+ * Register use:
+ *	a0/ return address
+ *	a1/ stack pointer
+ *	a2/ return value
+ *	a3/ src
+ *	a4/ length
+ *	a5/ dst
+ *	a6/ tmp
+ *	a7/ tmp
+ *	a8/ tmp
+ *	a9/ tmp
+ *	a10/ tmp
+ *	a11/ tmp
+ */
+
+	.text
+	.align	4
+	.global	bcopy
+	.type   bcopy,@function
+bcopy:
+	entry	sp, 16		# minimal stack frame
+	# a2=src, a3=dst, a4=len
+	mov	a5, a3		# copy dst so that a2 is return value
+	mov	a3, a2
+	mov	a2, a5
+	j	.Lcommon	# go to common code for memcpy+bcopy
+
+
+/*
+ * Byte by byte copy
+ */
+	.align	4
+	.byte	0		# 1 mod 4 alignment for LOOPNEZ
+				# (0 mod 4 alignment for LBEG)
+.Lbytecopy:
+#if XCHAL_HAVE_LOOPS
+	loopnez	a4, .Lbytecopydone
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a4, .Lbytecopydone
+	add	a7, a3, a4	# a7 = end address for source
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lnextbyte:
+	l8ui	a6, a3, 0
+	addi	a3, a3, 1
+	s8i	a6, a5, 0
+	addi	a5, a5, 1
+#if !XCHAL_HAVE_LOOPS
+	blt	a3, a7, .Lnextbyte
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lbytecopydone:
+	retw
+
+/*
+ * Destination is unaligned
+ */
+
+	.align	4
+.Ldst1mod2:	# dst is only byte aligned
+	_bltui	a4, 7, .Lbytecopy	# do short copies byte by byte
+
+	# copy 1 byte
+	l8ui	a6, a3,  0
+	addi	a3, a3,  1
+	addi	a4, a4, -1
+	s8i	a6, a5,  0
+	addi	a5, a5,  1
+	_bbci.l	a5, 1, .Ldstaligned	# if dst is now aligned, then
+					# return to main algorithm
+.Ldst2mod4:	# dst 16-bit aligned
+	# copy 2 bytes
+	_bltui	a4, 6, .Lbytecopy	# do short copies byte by byte
+	l8ui	a6, a3,  0
+	l8ui	a7, a3,  1
+	addi	a3, a3,  2
+	addi	a4, a4, -2
+	s8i	a6, a5,  0
+	s8i	a7, a5,  1
+	addi	a5, a5,  2
+	j	.Ldstaligned	# dst is now aligned, return to main algorithm
+
+	.align	4
+	.global	memcpy
+	.type   memcpy,@function
+memcpy:
+	.global	memmove
+	.type   memmove,@function
+memmove:
+
+	entry	sp, 16		# minimal stack frame
+	# a2/ dst, a3/ src, a4/ len
+	mov	a5, a2		# copy dst so that a2 is return value
+.Lcommon:
+	_bbsi.l	a2, 0, .Ldst1mod2	# if dst is 1 mod 2
+	_bbsi.l	a2, 1, .Ldst2mod4	# if dst is 2 mod 4
+.Ldstaligned:	# return here from .Ldst?mod? once dst is aligned
+	srli	a7, a4, 4	# number of loop iterations with 16B
+				# per iteration
+	movi	a8, 3		# if source is not aligned,
+	_bany	a3, a8, .Lsrcunaligned	# then use shifting copy
+	/*
+	 * Destination and source are word-aligned, use word copy.
+	 */
+	# copy 16 bytes per iteration for word-aligned dst and word-aligned src
+#if XCHAL_HAVE_LOOPS
+	loopnez	a7, .Loop1done
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a7, .Loop1done
+	slli	a8, a7, 4
+	add	a8, a8, a3	# a8 = end of last 16B source chunk
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop1:
+	l32i	a6, a3,  0
+	l32i	a7, a3,  4
+	s32i	a6, a5,  0
+	l32i	a6, a3,  8
+	s32i	a7, a5,  4
+	l32i	a7, a3, 12
+	s32i	a6, a5,  8
+	addi	a3, a3, 16
+	s32i	a7, a5, 12
+	addi	a5, a5, 16
+#if !XCHAL_HAVE_LOOPS
+	blt	a3, a8, .Loop1
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop1done:
+	bbci.l	a4, 3, .L2
+	# copy 8 bytes
+	l32i	a6, a3,  0
+	l32i	a7, a3,  4
+	addi	a3, a3,  8
+	s32i	a6, a5,  0
+	s32i	a7, a5,  4
+	addi	a5, a5,  8
+.L2:
+	bbsi.l	a4, 2, .L3
+	bbsi.l	a4, 1, .L4
+	bbsi.l	a4, 0, .L5
+	retw
+.L3:
+	# copy 4 bytes
+	l32i	a6, a3,  0
+	addi	a3, a3,  4
+	s32i	a6, a5,  0
+	addi	a5, a5,  4
+	bbsi.l	a4, 1, .L4
+	bbsi.l	a4, 0, .L5
+	retw
+.L4:
+	# copy 2 bytes
+	l16ui	a6, a3,  0
+	addi	a3, a3,  2
+	s16i	a6, a5,  0
+	addi	a5, a5,  2
+	bbsi.l	a4, 0, .L5
+	retw
+.L5:
+	# copy 1 byte
+	l8ui	a6, a3,  0
+	s8i	a6, a5,  0
+	retw
+
+/*
+ * Destination is aligned, Source is unaligned
+ */
+
+	.align	4
+.Lsrcunaligned:
+	_beqz	a4, .Ldone	# avoid loading anything for zero-length copies
+	# copy 16 bytes per iteration for word-aligned dst and unaligned src
+	ssa8	a3		# set shift amount from byte offset
+#define SIM_CHECKS_ALIGNMENT	1	/* set to 1 when running on ISS (simulator) with the
+					   lint or ferret client, or 0 to save a few cycles */
+#if XCHAL_UNALIGNED_LOAD_EXCEPTION || SIM_CHECKS_ALIGNMENT
+	and	a11, a3, a8	# save unalignment offset for below
+	sub	a3, a3, a11	# align a3
+#endif
+	l32i	a6, a3, 0	# load first word
+#if XCHAL_HAVE_LOOPS
+	loopnez	a7, .Loop2done
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a7, .Loop2done
+	slli	a10, a7, 4
+	add	a10, a10, a3	# a10 = end of last 16B source chunk
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop2:
+	l32i	a7, a3,  4
+	l32i	a8, a3,  8
+	src_b	a6, a6, a7
+	s32i	a6, a5,  0
+	l32i	a9, a3, 12
+	src_b	a7, a7, a8
+	s32i	a7, a5,  4
+	l32i	a6, a3, 16
+	src_b	a8, a8, a9
+	s32i	a8, a5,  8
+	addi	a3, a3, 16
+	src_b	a9, a9, a6
+	s32i	a9, a5, 12
+	addi	a5, a5, 16
+#if !XCHAL_HAVE_LOOPS
+	blt	a3, a10, .Loop2
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop2done:
+	bbci.l	a4, 3, .L12
+	# copy 8 bytes
+	l32i	a7, a3,  4
+	l32i	a8, a3,  8
+	src_b	a6, a6, a7
+	s32i	a6, a5,  0
+	addi	a3, a3,  8
+	src_b	a7, a7, a8
+	s32i	a7, a5,  4
+	addi	a5, a5,  8
+	mov	a6, a8
+.L12:
+	bbci.l	a4, 2, .L13
+	# copy 4 bytes
+	l32i	a7, a3,  4
+	addi	a3, a3,  4
+	src_b	a6, a6, a7
+	s32i	a6, a5,  0
+	addi	a5, a5,  4
+	mov	a6, a7
+.L13:
+#if XCHAL_UNALIGNED_LOAD_EXCEPTION || SIM_CHECKS_ALIGNMENT
+	add	a3, a3, a11	# readjust a3 with correct misalignment
+#endif
+	bbsi.l	a4, 1, .L14
+	bbsi.l	a4, 0, .L15
+.Ldone:	retw
+.L14:
+	# copy 2 bytes
+	l8ui	a6, a3,  0
+	l8ui	a7, a3,  1
+	addi	a3, a3,  2
+	s8i	a6, a5,  0
+	s8i	a7, a5,  1
+	addi	a5, a5,  2
+	bbsi.l	a4, 0, .L15
+	retw
+.L15:
+	# copy 1 byte
+	l8ui	a6, a3,  0
+	s8i	a6, a5,  0
+	retw
+
+/*
+ * Local Variables:
+ * mode:fundamental
+ * comment-start: "# "
+ * comment-start-skip: "# *"
+ * End:
+ */
diff --git a/arch/xtensa/lib/memset.S b/arch/xtensa/lib/memset.S
new file mode 100644
index 000000000000..4de25134bc62
--- /dev/null
+++ b/arch/xtensa/lib/memset.S
@@ -0,0 +1,160 @@
+/*
+ *  arch/xtensa/lib/memset.S
+ *
+ *  ANSI C standard library function memset
+ *  (Well, almost.  .fixup code might return zero.)
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file "COPYING" in the main directory of
+ *  this archive for more details.
+ *
+ *  Copyright (C) 2002 Tensilica Inc.
+ */
+
+#include <xtensa/coreasm.h>
+
+/*
+ * void *memset(void *dst, int c, size_t length)
+ *
+ * The algorithm is as follows:
+ *   Create a word with c in all byte positions
+ *   If the destination is aligned,
+ *     do 16B chucks with a loop, and then finish up with
+ *     8B, 4B, 2B, and 1B stores conditional on the length.
+ *   If destination is unaligned, align it by conditionally
+ *     setting 1B and 2B and then go to aligned case.
+ *   This code tries to use fall-through branches for the common
+ *     case of an aligned destination (except for the branches to
+ *     the alignment labels).
+ */
+
+/* Load or store instructions that may cause exceptions use the EX macro. */
+
+#define EX(insn,reg1,reg2,offset,handler)	\
+9:	insn	reg1, reg2, offset;		\
+	.section __ex_table, "a";		\
+	.word	9b, handler;			\
+	.previous
+
+
+.text
+.align	4
+.global	memset
+.type	memset,@function
+memset:
+	entry	sp, 16		# minimal stack frame
+	# a2/ dst, a3/ c, a4/ length
+	extui	a3, a3, 0, 8	# mask to just 8 bits
+	slli	a7, a3, 8	# duplicate character in all bytes of word
+	or	a3, a3, a7	# ...
+	slli	a7, a3, 16	# ...
+	or	a3, a3, a7	# ...
+	mov	a5, a2		# copy dst so that a2 is return value
+	movi	a6, 3		# for alignment tests
+	bany	a2, a6, .Ldstunaligned # if dst is unaligned
+.L0:	# return here from .Ldstunaligned when dst is aligned
+	srli	a7, a4, 4	# number of loop iterations with 16B
+				# per iteration
+	bnez	a4, .Laligned
+	retw
+
+/*
+ * Destination is word-aligned.
+ */
+	# set 16 bytes per iteration for word-aligned dst
+	.align	4		# 1 mod 4 alignment for LOOPNEZ
+	.byte	0		# (0 mod 4 alignment for LBEG)
+.Laligned:
+#if XCHAL_HAVE_LOOPS
+	loopnez	a7, .Loop1done
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a7, .Loop1done
+	slli	a6, a7, 4
+	add	a6, a6, a5	# a6 = end of last 16B chunk
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop1:
+	EX(s32i, a3, a5,  0, memset_fixup)
+	EX(s32i, a3, a5,  4, memset_fixup)
+	EX(s32i, a3, a5,  8, memset_fixup)
+	EX(s32i, a3, a5, 12, memset_fixup)
+	addi	a5, a5, 16
+#if !XCHAL_HAVE_LOOPS
+	blt	a5, a6, .Loop1
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop1done:
+	bbci.l	a4, 3, .L2
+	# set 8 bytes
+	EX(s32i, a3, a5,  0, memset_fixup)
+	EX(s32i, a3, a5,  4, memset_fixup)
+	addi	a5, a5,  8
+.L2:
+	bbci.l	a4, 2, .L3
+	# set 4 bytes
+	EX(s32i, a3, a5,  0, memset_fixup)
+	addi	a5, a5,  4
+.L3:
+	bbci.l	a4, 1, .L4
+	# set 2 bytes
+	EX(s16i, a3, a5,  0, memset_fixup)
+	addi	a5, a5,  2
+.L4:
+	bbci.l	a4, 0, .L5
+	# set 1 byte
+	EX(s8i, a3, a5,  0, memset_fixup)
+.L5:
+.Lret1:
+	retw
+
+/*
+ * Destination is unaligned
+ */
+
+.Ldstunaligned:
+	bltui	a4, 8, .Lbyteset	# do short copies byte by byte
+	bbci.l	a5, 0, .L20		# branch if dst alignment half-aligned
+	# dst is only byte aligned
+	# set 1 byte
+	EX(s8i, a3, a5,  0, memset_fixup)
+	addi	a5, a5,  1
+	addi	a4, a4, -1
+	# now retest if dst aligned
+	bbci.l	a5, 1, .L0	# if now aligned, return to main algorithm
+.L20:
+	# dst half-aligned
+	# set 2 bytes
+	EX(s16i, a3, a5,  0, memset_fixup)
+	addi	a5, a5,  2
+	addi	a4, a4, -2
+	j	.L0		# dst is now aligned, return to main algorithm
+
+/*
+ * Byte by byte set
+ */
+	.align	4
+	.byte	0		# 1 mod 4 alignment for LOOPNEZ
+				# (0 mod 4 alignment for LBEG)
+.Lbyteset:
+#if XCHAL_HAVE_LOOPS
+	loopnez	a4, .Lbytesetdone
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a4, .Lbytesetdone
+	add	a6, a5, a4	# a6 = ending address
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lbyteloop:
+	EX(s8i, a3, a5, 0, memset_fixup)
+	addi	a5, a5, 1
+#if !XCHAL_HAVE_LOOPS
+	blt	a5, a6, .Lbyteloop
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lbytesetdone:
+	retw
+
+
+	.section .fixup, "ax"
+	.align	4
+
+/* We return zero if a failure occurred. */
+
+memset_fixup:
+	movi	a2, 0
+	retw
diff --git a/arch/xtensa/lib/pci-auto.c b/arch/xtensa/lib/pci-auto.c
new file mode 100644
index 000000000000..90c790f6123b
--- /dev/null
+++ b/arch/xtensa/lib/pci-auto.c
@@ -0,0 +1,352 @@
+/*
+ * arch/xtensa/kernel/pci-auto.c
+ *
+ * PCI autoconfiguration library
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <zankel@tensilica.com, cez@zankel.net>
+ *
+ * Based on work from Matt Porter <mporter@mvista.com>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+
+#include <asm/pci-bridge.h>
+
+
+/*
+ *
+ * Setting up a PCI
+ *
+ * pci_ctrl->first_busno = <first bus number (0)>
+ * pci_ctrl->last_busno = <last bus number (0xff)>
+ * pci_ctrl->ops = <PCI config operations>
+ * pci_ctrl->map_irq = <function to return the interrupt number for a device>
+ *
+ * pci_ctrl->io_space.start = <IO space start address (PCI view)>
+ * pci_ctrl->io_space.end = <IO space end address (PCI view)>
+ * pci_ctrl->io_space.base = <IO space offset: address 0 from CPU space>
+ * pci_ctrl->mem_space.start = <MEM space start address (PCI view)>
+ * pci_ctrl->mem_space.end = <MEM space end address (PCI view)>
+ * pci_ctrl->mem_space.base = <MEM space offset: address 0 from CPU space>
+ *
+ * pcibios_init_resource(&pci_ctrl->io_resource, <IO space start>,
+ * 			 <IO space end>, IORESOURCE_IO, "PCI host bridge");
+ * pcibios_init_resource(&pci_ctrl->mem_resources[0], <MEM space start>,
+ * 			 <MEM space end>, IORESOURCE_MEM, "PCI host bridge");
+ *
+ * pci_ctrl->last_busno = pciauto_bus_scan(pci_ctrl,pci_ctrl->first_busno);
+ *
+ * int __init pciauto_bus_scan(struct pci_controller *pci_ctrl, int current_bus)
+ *
+ */
+
+
+/* define DEBUG to print some debugging messages. */
+
+#undef DEBUG
+
+#ifdef DEBUG
+# define DBG(x...) printk(x)
+#else
+# define DBG(x...)
+#endif
+
+static int pciauto_upper_iospc;
+static int pciauto_upper_memspc;
+
+static struct pci_dev pciauto_dev;
+static struct pci_bus pciauto_bus;
+
+/*
+ * Helper functions
+ */
+
+/* Initialize the bars of a PCI device.  */
+
+static void __init
+pciauto_setup_bars(struct pci_dev *dev, int bar_limit)
+{
+	int bar_size;
+	int bar, bar_nr;
+	int *upper_limit;
+	int found_mem64 = 0;
+
+	for (bar = PCI_BASE_ADDRESS_0, bar_nr = 0;
+	     bar <= bar_limit;
+	     bar+=4, bar_nr++)
+	{
+		/* Tickle the BAR and get the size */
+		pci_write_config_dword(dev, bar, 0xffffffff);
+		pci_read_config_dword(dev, bar, &bar_size);
+
+		/* If BAR is not implemented go to the next BAR */
+		if (!bar_size)
+			continue;
+
+		/* Check the BAR type and set our address mask */
+		if (bar_size & PCI_BASE_ADDRESS_SPACE_IO)
+		{
+			bar_size &= PCI_BASE_ADDRESS_IO_MASK;
+			upper_limit = &pciauto_upper_iospc;
+			DBG("PCI Autoconfig: BAR %d, I/O, ", bar_nr);
+		}
+		else
+		{
+			if ((bar_size & PCI_BASE_ADDRESS_MEM_TYPE_MASK) ==
+			    PCI_BASE_ADDRESS_MEM_TYPE_64)
+				found_mem64 = 1;
+
+			bar_size &= PCI_BASE_ADDRESS_MEM_MASK;
+			upper_limit = &pciauto_upper_memspc;
+			DBG("PCI Autoconfig: BAR %d, Mem, ", bar_nr);
+		}
+
+		/* Allocate a base address (bar_size is negative!) */
+		*upper_limit = (*upper_limit + bar_size) & bar_size;
+
+		/* Write it out and update our limit */
+		pci_write_config_dword(dev, bar, *upper_limit);
+
+		/*
+		 * If we are a 64-bit decoder then increment to the
+		 * upper 32 bits of the bar and force it to locate
+		 * in the lower 4GB of memory.
+		 */
+
+		if (found_mem64)
+			pci_write_config_dword(dev, (bar+=4), 0x00000000);
+
+		DBG("size=0x%x, address=0x%x\n", ~bar_size + 1, *upper_limit);
+	}
+}
+
+/* Initialize the interrupt number. */
+
+static void __init
+pciauto_setup_irq(struct pci_controller* pci_ctrl,struct pci_dev *dev,int devfn)
+{
+	u8 pin;
+	int irq = 0;
+
+	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
+
+	/* Fix illegal pin numbers. */
+
+	if (pin == 0 || pin > 4)
+		pin = 1;
+
+	if (pci_ctrl->map_irq)
+		irq = pci_ctrl->map_irq(dev, PCI_SLOT(devfn), pin);
+
+	if (irq == -1)
+		irq = 0;
+
+	DBG("PCI Autoconfig: Interrupt %d, pin %d\n", irq, pin);
+
+	pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq);
+}
+
+
+static void __init
+pciauto_prescan_setup_bridge(struct pci_dev *dev, int current_bus,
+			     int sub_bus, int *iosave, int *memsave)
+{
+	/* Configure bus number registers */
+	pci_write_config_byte(dev, PCI_PRIMARY_BUS, current_bus);
+	pci_write_config_byte(dev, PCI_SECONDARY_BUS, sub_bus + 1);
+	pci_write_config_byte(dev, PCI_SUBORDINATE_BUS,	0xff);
+
+	/* Round memory allocator to 1MB boundary */
+	pciauto_upper_memspc &= ~(0x100000 - 1);
+	*memsave = pciauto_upper_memspc;
+
+	/* Round I/O allocator to 4KB boundary */
+	pciauto_upper_iospc &= ~(0x1000 - 1);
+	*iosave = pciauto_upper_iospc;
+
+	/* Set up memory and I/O filter limits, assume 32-bit I/O space */
+	pci_write_config_word(dev, PCI_MEMORY_LIMIT,
+			      ((pciauto_upper_memspc - 1) & 0xfff00000) >> 16);
+	pci_write_config_byte(dev, PCI_IO_LIMIT,
+			      ((pciauto_upper_iospc - 1) & 0x0000f000) >> 8);
+	pci_write_config_word(dev, PCI_IO_LIMIT_UPPER16,
+			      ((pciauto_upper_iospc - 1) & 0xffff0000) >> 16);
+}
+
+static void __init
+pciauto_postscan_setup_bridge(struct pci_dev *dev, int current_bus, int sub_bus,
+			      int *iosave, int *memsave)
+{
+	int cmdstat;
+
+	/* Configure bus number registers */
+	pci_write_config_byte(dev, PCI_SUBORDINATE_BUS,	sub_bus);
+
+	/*
+	 * Round memory allocator to 1MB boundary.
+	 * If no space used, allocate minimum.
+	 */
+	pciauto_upper_memspc &= ~(0x100000 - 1);
+	if (*memsave == pciauto_upper_memspc)
+		pciauto_upper_memspc -= 0x00100000;
+
+	pci_write_config_word(dev, PCI_MEMORY_BASE, pciauto_upper_memspc >> 16);
+
+	/* Allocate 1MB for pre-fretch */
+	pci_write_config_word(dev, PCI_PREF_MEMORY_LIMIT,
+			      ((pciauto_upper_memspc - 1) & 0xfff00000) >> 16);
+
+	pciauto_upper_memspc -= 0x100000;
+
+	pci_write_config_word(dev, PCI_PREF_MEMORY_BASE,
+			      pciauto_upper_memspc >> 16);
+
+	/* Round I/O allocator to 4KB boundary */
+	pciauto_upper_iospc &= ~(0x1000 - 1);
+	if (*iosave == pciauto_upper_iospc)
+		pciauto_upper_iospc -= 0x1000;
+
+	pci_write_config_byte(dev, PCI_IO_BASE,
+			      (pciauto_upper_iospc & 0x0000f000) >> 8);
+	pci_write_config_word(dev, PCI_IO_BASE_UPPER16,
+			      pciauto_upper_iospc >> 16);
+
+	/* Enable memory and I/O accesses, enable bus master */
+	pci_read_config_dword(dev, PCI_COMMAND, &cmdstat);
+	pci_write_config_dword(dev, PCI_COMMAND,
+			       cmdstat |
+			       PCI_COMMAND_IO |
+			       PCI_COMMAND_MEMORY |
+			       PCI_COMMAND_MASTER);
+}
+
+/*
+ * Scan the current PCI bus.
+ */
+
+
+int __init pciauto_bus_scan(struct pci_controller *pci_ctrl, int current_bus)
+{
+	int sub_bus, pci_devfn, pci_class, cmdstat, found_multi=0;
+	unsigned short vid;
+	unsigned char header_type;
+	struct pci_dev *dev = &pciauto_dev;
+
+        pciauto_dev.bus = &pciauto_bus;
+        pciauto_dev.sysdata = pci_ctrl;
+	pciauto_bus.ops = pci_ctrl->ops;
+
+	/*
+	 * Fetch our I/O and memory space upper boundaries used
+	 * to allocated base addresses on this pci_controller.
+	 */
+
+	if (current_bus == pci_ctrl->first_busno)
+	{
+		pciauto_upper_iospc = pci_ctrl->io_resource.end + 1;
+		pciauto_upper_memspc = pci_ctrl->mem_resources[0].end + 1;
+	}
+
+	sub_bus = current_bus;
+
+	for (pci_devfn = 0; pci_devfn < 0xff; pci_devfn++)
+	{
+		/* Skip our host bridge */
+		if ((current_bus == pci_ctrl->first_busno) && (pci_devfn == 0))
+			continue;
+
+		if (PCI_FUNC(pci_devfn) && !found_multi)
+			continue;
+
+		pciauto_bus.number = current_bus;
+		pciauto_dev.devfn = pci_devfn;
+
+		/* If config space read fails from this device, move on */
+		if (pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type))
+			continue;
+
+		if (!PCI_FUNC(pci_devfn))
+			found_multi = header_type & 0x80;
+		pci_read_config_word(dev, PCI_VENDOR_ID, &vid);
+
+		if (vid == 0xffff || vid == 0x0000) {
+			found_multi = 0;
+			continue;
+		}
+
+		pci_read_config_dword(dev, PCI_CLASS_REVISION, &pci_class);
+
+		if ((pci_class >> 16) == PCI_CLASS_BRIDGE_PCI) {
+
+			int iosave, memsave;
+
+			DBG("PCI Autoconfig: Found P2P bridge, device %d\n",
+			    PCI_SLOT(pci_devfn));
+
+			/* Allocate PCI I/O and/or memory space */
+			pciauto_setup_bars(dev, PCI_BASE_ADDRESS_1);
+
+			pciauto_prescan_setup_bridge(dev, current_bus, sub_bus,
+					&iosave, &memsave);
+			sub_bus = pciauto_bus_scan(pci_ctrl, sub_bus+1);
+			pciauto_postscan_setup_bridge(dev, current_bus, sub_bus,
+					&iosave, &memsave);
+			pciauto_bus.number = current_bus;
+
+			continue;
+
+		}
+
+
+#if 0
+		/* Skip legacy mode IDE controller */
+
+		if ((pci_class >> 16) == PCI_CLASS_STORAGE_IDE) {
+
+			unsigned char prg_iface;
+			pci_read_config_byte(dev, PCI_CLASS_PROG, &prg_iface);
+
+			if (!(prg_iface & PCIAUTO_IDE_MODE_MASK)) {
+				DBG("PCI Autoconfig: Skipping legacy mode "
+				    "IDE controller\n");
+				continue;
+			}
+		}
+#endif
+
+		/*
+		 * Found a peripheral, enable some standard
+		 * settings
+		 */
+
+		pci_read_config_dword(dev, PCI_COMMAND,	&cmdstat);
+		pci_write_config_dword(dev, PCI_COMMAND,
+				cmdstat |
+					PCI_COMMAND_IO |
+					PCI_COMMAND_MEMORY |
+					PCI_COMMAND_MASTER);
+		pci_write_config_byte(dev, PCI_LATENCY_TIMER, 0x80);
+
+		/* Allocate PCI I/O and/or memory space */
+		DBG("PCI Autoconfig: Found Bus %d, Device %d, Function %d\n",
+		    current_bus, PCI_SLOT(pci_devfn), PCI_FUNC(pci_devfn) );
+
+		pciauto_setup_bars(dev, PCI_BASE_ADDRESS_5);
+		pciauto_setup_irq(pci_ctrl, dev, pci_devfn);
+	}
+	return sub_bus;
+}
+
+
+
+
+
diff --git a/arch/xtensa/lib/strcasecmp.c b/arch/xtensa/lib/strcasecmp.c
new file mode 100644
index 000000000000..165b2d6effa5
--- /dev/null
+++ b/arch/xtensa/lib/strcasecmp.c
@@ -0,0 +1,32 @@
+/*
+ *  linux/arch/xtensa/lib/strcasecmp.c
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file "COPYING" in the main directory of
+ *  this archive for more details.
+ *
+ *  Copyright (C) 2002 Tensilica Inc.
+ */
+
+#include <linux/string.h>
+
+
+/* We handle nothing here except the C locale.  Since this is used in
+   only one place, on strings known to contain only 7 bit ASCII, this
+   is ok.  */
+
+int strcasecmp(const char *a, const char *b)
+{
+	int ca, cb;
+
+	do {
+		ca = *a++ & 0xff;
+		cb = *b++ & 0xff;
+		if (ca >= 'A' && ca <= 'Z')
+			ca += 'a' - 'A';
+		if (cb >= 'A' && cb <= 'Z')
+			cb += 'a' - 'A';
+	} while (ca == cb && ca != '\0');
+
+	return ca - cb;
+}
diff --git a/arch/xtensa/lib/strncpy_user.S b/arch/xtensa/lib/strncpy_user.S
new file mode 100644
index 000000000000..71d55df43893
--- /dev/null
+++ b/arch/xtensa/lib/strncpy_user.S
@@ -0,0 +1,224 @@
+/*
+ *  arch/xtensa/lib/strncpy_user.S
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file "COPYING" in the main directory of
+ *  this archive for more details.
+ *
+ *  Returns: -EFAULT if exception before terminator, N if the entire
+ *  buffer filled, else strlen.
+ *
+ *  Copyright (C) 2002 Tensilica Inc.
+ */
+
+#include <xtensa/coreasm.h>
+#include <linux/errno.h>
+
+/* Load or store instructions that may cause exceptions use the EX macro. */
+
+#define EX(insn,reg1,reg2,offset,handler)	\
+9:	insn	reg1, reg2, offset;		\
+	.section __ex_table, "a";		\
+	.word	9b, handler;			\
+	.previous
+
+/*
+ * char *__strncpy_user(char *dst, const char *src, size_t len)
+ */
+.text
+.begin literal
+.align	4
+.Lmask0:
+	.byte	0xff, 0x00, 0x00, 0x00
+.Lmask1:
+	.byte	0x00, 0xff, 0x00, 0x00
+.Lmask2:
+	.byte	0x00, 0x00, 0xff, 0x00
+.Lmask3:
+	.byte	0x00, 0x00, 0x00, 0xff
+.end literal
+
+# Register use
+#   a0/ return address
+#   a1/ stack pointer
+#   a2/ return value
+#   a3/ src
+#   a4/ len
+#   a5/ mask0
+#   a6/ mask1
+#   a7/ mask2
+#   a8/ mask3
+#   a9/ tmp
+#   a10/ tmp
+#   a11/ dst
+#   a12/ tmp
+
+.align	4
+.global	__strncpy_user
+.type	__strncpy_user,@function
+__strncpy_user:
+	entry	sp, 16		# minimal stack frame
+	# a2/ dst, a3/ src, a4/ len
+	mov	a11, a2		# leave dst in return value register
+	beqz	a4, .Lret	# if len is zero
+	l32r	a5, .Lmask0	# mask for byte 0
+	l32r	a6, .Lmask1	# mask for byte 1
+	l32r	a7, .Lmask2	# mask for byte 2
+	l32r	a8, .Lmask3	# mask for byte 3
+	bbsi.l	a3, 0, .Lsrc1mod2 # if only  8-bit aligned
+	bbsi.l	a3, 1, .Lsrc2mod4 # if only 16-bit aligned
+.Lsrcaligned:	# return here when src is word-aligned
+	srli	a12, a4, 2	# number of loop iterations with 4B per loop
+	movi	a9, 3
+	bnone	a11, a9, .Laligned
+	j	.Ldstunaligned
+
+.Lsrc1mod2:	# src address is odd
+	EX(l8ui, a9, a3, 0, fixup_l)	# get byte 0
+	addi	a3, a3, 1		# advance src pointer
+	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
+	beqz	a9, .Lret		# if byte 0 is zero
+	addi	a11, a11, 1		# advance dst pointer
+	addi	a4, a4, -1		# decrement len
+	beqz	a4, .Lret		# if len is zero
+	bbci.l	a3, 1, .Lsrcaligned	# if src is now word-aligned
+
+.Lsrc2mod4:	# src address is 2 mod 4
+	EX(l8ui, a9, a3, 0, fixup_l)	# get byte 0
+	/* 1-cycle interlock */
+	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
+	beqz	a9, .Lret		# if byte 0 is zero
+	addi	a11, a11, 1		# advance dst pointer
+	addi	a4, a4, -1		# decrement len
+	beqz	a4, .Lret		# if len is zero
+	EX(l8ui, a9, a3, 1, fixup_l)	# get byte 0
+	addi	a3, a3, 2		# advance src pointer
+	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
+	beqz	a9, .Lret		# if byte 0 is zero
+	addi	a11, a11, 1		# advance dst pointer
+	addi	a4, a4, -1		# decrement len
+	bnez	a4, .Lsrcaligned	# if len is nonzero
+.Lret:
+	sub	a2, a11, a2		# compute strlen
+	retw
+
+/*
+ * dst is word-aligned, src is word-aligned
+ */
+	.align	4		# 1 mod 4 alignment for LOOPNEZ
+	.byte	0		# (0 mod 4 alignment for LBEG)
+.Laligned:
+#if XCHAL_HAVE_LOOPS
+	loopnez	a12, .Loop1done
+#else
+	beqz	a12, .Loop1done
+	slli	a12, a12, 2
+	add	a12, a12, a11	# a12 = end of last 4B chunck
+#endif
+.Loop1:
+	EX(l32i, a9, a3, 0, fixup_l)	# get word from src
+	addi	a3, a3, 4		# advance src pointer
+	bnone	a9, a5, .Lz0		# if byte 0 is zero
+	bnone	a9, a6, .Lz1		# if byte 1 is zero
+	bnone	a9, a7, .Lz2		# if byte 2 is zero
+	EX(s32i, a9, a11, 0, fixup_s)	# store word to dst
+	bnone	a9, a8, .Lz3		# if byte 3 is zero
+	addi	a11, a11, 4		# advance dst pointer
+#if !XCHAL_HAVE_LOOPS
+	blt	a11, a12, .Loop1
+#endif
+
+.Loop1done:
+	bbci.l	a4, 1, .L100
+	# copy 2 bytes
+	EX(l16ui, a9, a3, 0, fixup_l)
+	addi	a3, a3, 2		# advance src pointer
+#ifdef __XTENSA_EB__
+	bnone	a9, a7, .Lz0		# if byte 2 is zero
+	bnone	a9, a8, .Lz1		# if byte 3 is zero
+#else
+	bnone	a9, a5, .Lz0		# if byte 0 is zero
+	bnone	a9, a6, .Lz1		# if byte 1 is zero
+#endif
+	EX(s16i, a9, a11, 0, fixup_s)
+	addi	a11, a11, 2		# advance dst pointer
+.L100:
+	bbci.l	a4, 0, .Lret
+	EX(l8ui, a9, a3, 0, fixup_l)
+	/* slot */
+	EX(s8i, a9, a11, 0, fixup_s)
+	beqz	a9, .Lret		# if byte is zero
+	addi	a11, a11, 1-3		# advance dst ptr 1, but also cancel
+					# the effect of adding 3 in .Lz3 code
+	/* fall thru to .Lz3 and "retw" */
+
+.Lz3:	# byte 3 is zero
+	addi	a11, a11, 3		# advance dst pointer
+	sub	a2, a11, a2		# compute strlen
+	retw
+.Lz0:	# byte 0 is zero
+#ifdef __XTENSA_EB__
+	movi	a9, 0
+#endif /* __XTENSA_EB__ */
+	EX(s8i, a9, a11, 0, fixup_s)
+	sub	a2, a11, a2		# compute strlen
+	retw
+.Lz1:	# byte 1 is zero
+#ifdef __XTENSA_EB__
+        extui   a9, a9, 16, 16
+#endif /* __XTENSA_EB__ */
+	EX(s16i, a9, a11, 0, fixup_s)
+	addi	a11, a11, 1		# advance dst pointer
+	sub	a2, a11, a2		# compute strlen
+	retw
+.Lz2:	# byte 2 is zero
+#ifdef __XTENSA_EB__
+        extui   a9, a9, 16, 16
+#endif /* __XTENSA_EB__ */
+	EX(s16i, a9, a11, 0, fixup_s)
+	movi	a9, 0
+	EX(s8i, a9, a11, 2, fixup_s)
+	addi	a11, a11, 2		# advance dst pointer
+	sub	a2, a11, a2		# compute strlen
+	retw
+
+	.align	4		# 1 mod 4 alignment for LOOPNEZ
+	.byte	0		# (0 mod 4 alignment for LBEG)
+.Ldstunaligned:
+/*
+ * for now just use byte copy loop
+ */
+#if XCHAL_HAVE_LOOPS
+	loopnez	a4, .Lunalignedend
+#else
+	beqz	a4, .Lunalignedend
+	add	a12, a11, a4		# a12 = ending address
+#endif /* XCHAL_HAVE_LOOPS */
+.Lnextbyte:
+	EX(l8ui, a9, a3, 0, fixup_l)
+	addi	a3, a3, 1
+	EX(s8i, a9, a11, 0, fixup_s)
+	beqz	a9, .Lunalignedend
+	addi	a11, a11, 1
+#if !XCHAL_HAVE_LOOPS
+	blt	a11, a12, .Lnextbyte
+#endif
+
+.Lunalignedend:
+	sub	a2, a11, a2		# compute strlen
+	retw
+
+
+	.section .fixup, "ax"
+	.align	4
+
+	/* For now, just return -EFAULT.  Future implementations might
+	 * like to clear remaining kernel space, like the fixup
+	 * implementation in memset().  Thus, we differentiate between
+	 * load/store fixups. */
+
+fixup_s:
+fixup_l:
+	movi	a2, -EFAULT
+	retw
+
diff --git a/arch/xtensa/lib/strnlen_user.S b/arch/xtensa/lib/strnlen_user.S
new file mode 100644
index 000000000000..cdff4d670f3b
--- /dev/null
+++ b/arch/xtensa/lib/strnlen_user.S
@@ -0,0 +1,147 @@
+/*
+ *  arch/xtensa/lib/strnlen_user.S
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file "COPYING" in the main directory of
+ *  this archive for more details.
+ *
+ *  Returns strnlen, including trailing zero terminator.
+ *  Zero indicates error.
+ *
+ *  Copyright (C) 2002 Tensilica Inc.
+ */
+
+#include <xtensa/coreasm.h>
+
+/* Load or store instructions that may cause exceptions use the EX macro. */
+
+#define EX(insn,reg1,reg2,offset,handler)	\
+9:	insn	reg1, reg2, offset;		\
+	.section __ex_table, "a";		\
+	.word	9b, handler;			\
+	.previous
+
+/*
+ * size_t __strnlen_user(const char *s, size_t len)
+ */
+.text
+.begin literal
+.align	4
+.Lmask0:
+	.byte	0xff, 0x00, 0x00, 0x00
+.Lmask1:
+	.byte	0x00, 0xff, 0x00, 0x00
+.Lmask2:
+	.byte	0x00, 0x00, 0xff, 0x00
+.Lmask3:
+	.byte	0x00, 0x00, 0x00, 0xff
+.end literal
+
+# Register use:
+#   a2/ src
+#   a3/ len
+#   a4/ tmp
+#   a5/ mask0
+#   a6/ mask1
+#   a7/ mask2
+#   a8/ mask3
+#   a9/ tmp
+#   a10/ tmp
+
+.align	4
+.global	__strnlen_user
+.type	__strnlen_user,@function
+__strnlen_user:
+	entry	sp, 16		# minimal stack frame
+	# a2/ s, a3/ len
+	addi	a4, a2, -4	# because we overincrement at the end;
+				# we compensate with load offsets of 4
+	l32r	a5, .Lmask0	# mask for byte 0
+	l32r	a6, .Lmask1	# mask for byte 1
+	l32r	a7, .Lmask2	# mask for byte 2
+	l32r	a8, .Lmask3	# mask for byte 3
+	bbsi.l	a2, 0, .L1mod2	# if only  8-bit aligned
+	bbsi.l	a2, 1, .L2mod4	# if only 16-bit aligned
+
+/*
+ * String is word-aligned.
+ */
+.Laligned:
+	srli	a10, a3, 2	# number of loop iterations with 4B per loop
+#if XCHAL_HAVE_LOOPS
+	loopnez	a10, .Ldone
+#else
+	beqz	a10, .Ldone
+	slli	a10, a10, 2
+	add	a10, a10, a4	# a10 = end of last 4B chunk
+#endif /* XCHAL_HAVE_LOOPS */
+.Loop:
+	EX(l32i, a9, a4, 4, lenfixup)	# get next word of string
+	addi	a4, a4, 4		# advance string pointer
+	bnone	a9, a5, .Lz0		# if byte 0 is zero
+	bnone	a9, a6, .Lz1		# if byte 1 is zero
+	bnone	a9, a7, .Lz2		# if byte 2 is zero
+	bnone	a9, a8, .Lz3		# if byte 3 is zero
+#if !XCHAL_HAVE_LOOPS
+	blt	a4, a10, .Loop
+#endif
+
+.Ldone:
+	EX(l32i, a9, a4, 4, lenfixup)	# load 4 bytes for remaining checks
+
+	bbci.l	a3, 1, .L100
+	# check two more bytes (bytes 0, 1 of word)
+	addi	a4, a4, 2	# advance string pointer
+	bnone	a9, a5, .Lz0	# if byte 0 is zero
+	bnone	a9, a6, .Lz1	# if byte 1 is zero
+.L100:
+	bbci.l	a3, 0, .L101
+	# check one more byte (byte 2 of word)
+	# Actually, we don't need to check.  Zero or nonzero, we'll add one.
+	# Do not add an extra one for the NULL terminator since we have
+	#  exhausted the original len parameter.
+	addi	a4, a4, 1	# advance string pointer
+.L101:
+	sub	a2, a4, a2	# compute length
+	retw
+
+# NOTE that in several places below, we point to the byte just after
+# the zero byte in order to include the NULL terminator in the count.
+
+.Lz3:	# byte 3 is zero
+	addi	a4, a4, 3	# point to zero byte
+.Lz0:	# byte 0 is zero
+	addi	a4, a4, 1	# point just beyond zero byte
+	sub	a2, a4, a2	# subtract to get length
+	retw
+.Lz1:	# byte 1 is zero
+	addi	a4, a4, 1+1	# point just beyond zero byte
+	sub	a2, a4, a2	# subtract to get length
+	retw
+.Lz2:	# byte 2 is zero
+	addi	a4, a4, 2+1	# point just beyond zero byte
+	sub	a2, a4, a2	# subtract to get length
+	retw
+
+.L1mod2:	# address is odd
+	EX(l8ui, a9, a4, 4, lenfixup)	# get byte 0
+	addi	a4, a4, 1		# advance string pointer
+	beqz	a9, .Lz3		# if byte 0 is zero
+	bbci.l	a4, 1, .Laligned	# if string pointer is now word-aligned
+
+.L2mod4:	# address is 2 mod 4
+	addi	a4, a4, 2	# advance ptr for aligned access
+	EX(l32i, a9, a4, 0, lenfixup)	# get word with first two bytes of string
+	bnone	a9, a7, .Lz2	# if byte 2 (of word, not string) is zero
+	bany	a9, a8, .Laligned # if byte 3 (of word, not string) is nonzero
+	# byte 3 is zero
+	addi	a4, a4, 3+1	# point just beyond zero byte
+	sub	a2, a4, a2	# subtract to get length
+	retw
+
+	.section .fixup, "ax"
+	.align	4
+lenfixup:
+	movi	a2, 0
+	retw
+
diff --git a/arch/xtensa/lib/usercopy.S b/arch/xtensa/lib/usercopy.S
new file mode 100644
index 000000000000..265db2693cbd
--- /dev/null
+++ b/arch/xtensa/lib/usercopy.S
@@ -0,0 +1,321 @@
+/*
+ *  arch/xtensa/lib/usercopy.S
+ *
+ *  Copy to/from user space (derived from arch/xtensa/lib/hal/memcopy.S)
+ *
+ *  DO NOT COMBINE this function with <arch/xtensa/lib/hal/memcopy.S>.
+ *  It needs to remain separate and distinct.  The hal files are part
+ *  of the the Xtensa link-time HAL, and those files may differ per
+ *  processor configuration.  Patching the kernel for another
+ *  processor configuration includes replacing the hal files, and we
+ *  could loose the special functionality for accessing user-space
+ *  memory during such a patch.  We sacrifice a little code space here
+ *  in favor to simplify code maintenance.
+ *
+ *  This file is subject to the terms and conditions of the GNU General
+ *  Public License.  See the file "COPYING" in the main directory of
+ *  this archive for more details.
+ *
+ *  Copyright (C) 2002 Tensilica Inc.
+ */
+
+
+/*
+ * size_t __xtensa_copy_user (void *dst, const void *src, size_t len);
+ *
+ * The returned value is the number of bytes not copied.  Implies zero
+ * is success.
+ *
+ * The general case algorithm is as follows:
+ *   If the destination and source are both aligned,
+ *     do 16B chunks with a loop, and then finish up with
+ *     8B, 4B, 2B, and 1B copies conditional on the length.
+ *   If destination is aligned and source unaligned,
+ *     do the same, but use SRC to align the source data.
+ *   If destination is unaligned, align it by conditionally
+ *     copying 1B and 2B and then retest.
+ *   This code tries to use fall-through braches for the common
+ *     case of aligned destinations (except for the branches to
+ *     the alignment label).
+ *
+ * Register use:
+ *	a0/ return address
+ *	a1/ stack pointer
+ *	a2/ return value
+ *	a3/ src
+ *	a4/ length
+ *	a5/ dst
+ *	a6/ tmp
+ *	a7/ tmp
+ *	a8/ tmp
+ *	a9/ tmp
+ *	a10/ tmp
+ *	a11/ original length
+ */
+
+#include <xtensa/coreasm.h>
+
+#ifdef __XTENSA_EB__
+#define ALIGN(R, W0, W1) src	R, W0, W1
+#define SSA8(R)	ssa8b R
+#else
+#define ALIGN(R, W0, W1) src	R, W1, W0
+#define SSA8(R)	ssa8l R
+#endif
+
+/* Load or store instructions that may cause exceptions use the EX macro. */
+
+#define EX(insn,reg1,reg2,offset,handler)	\
+9:	insn	reg1, reg2, offset;		\
+	.section __ex_table, "a";		\
+	.word	9b, handler;			\
+	.previous
+
+
+	.text
+	.align	4
+	.global	__xtensa_copy_user
+	.type	__xtensa_copy_user,@function
+__xtensa_copy_user:
+	entry	sp, 16		# minimal stack frame
+	# a2/ dst, a3/ src, a4/ len
+	mov	a5, a2		# copy dst so that a2 is return value
+	mov	a11, a4		# preserve original len for error case
+.Lcommon:
+	bbsi.l	a2, 0, .Ldst1mod2 # if dst is 1 mod 2
+	bbsi.l	a2, 1, .Ldst2mod4 # if dst is 2 mod 4
+.Ldstaligned:	# return here from .Ldstunaligned when dst is aligned
+	srli	a7, a4, 4	# number of loop iterations with 16B
+				# per iteration
+	movi	a8, 3		  # if source is also aligned,
+	bnone	a3, a8, .Laligned # then use word copy
+	SSA8(	a3)		# set shift amount from byte offset
+	bnez	a4, .Lsrcunaligned
+	movi	a2, 0		# return success for len==0
+	retw
+
+/*
+ * Destination is unaligned
+ */
+
+.Ldst1mod2:	# dst is only byte aligned
+	bltui	a4, 7, .Lbytecopy	# do short copies byte by byte
+
+	# copy 1 byte
+	EX(l8ui, a6, a3, 0, l_fixup)
+	addi	a3, a3,  1
+	EX(s8i, a6, a5,  0, s_fixup)
+	addi	a5, a5,  1
+	addi	a4, a4, -1
+	bbci.l	a5, 1, .Ldstaligned	# if dst is now aligned, then
+					# return to main algorithm
+.Ldst2mod4:	# dst 16-bit aligned
+	# copy 2 bytes
+	bltui	a4, 6, .Lbytecopy	# do short copies byte by byte
+	EX(l8ui, a6, a3, 0, l_fixup)
+	EX(l8ui, a7, a3, 1, l_fixup)
+	addi	a3, a3,  2
+	EX(s8i, a6, a5,  0, s_fixup)
+	EX(s8i, a7, a5,  1, s_fixup)
+	addi	a5, a5,  2
+	addi	a4, a4, -2
+	j	.Ldstaligned	# dst is now aligned, return to main algorithm
+
+/*
+ * Byte by byte copy
+ */
+	.align	4
+	.byte	0		# 1 mod 4 alignment for LOOPNEZ
+				# (0 mod 4 alignment for LBEG)
+.Lbytecopy:
+#if XCHAL_HAVE_LOOPS
+	loopnez	a4, .Lbytecopydone
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a4, .Lbytecopydone
+	add	a7, a3, a4	# a7 = end address for source
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lnextbyte:
+	EX(l8ui, a6, a3, 0, l_fixup)
+	addi	a3, a3, 1
+	EX(s8i, a6, a5, 0, s_fixup)
+	addi	a5, a5, 1
+#if !XCHAL_HAVE_LOOPS
+	blt	a3, a7, .Lnextbyte
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lbytecopydone:
+	movi	a2, 0		# return success for len bytes copied
+	retw
+
+/*
+ * Destination and source are word-aligned.
+ */
+	# copy 16 bytes per iteration for word-aligned dst and word-aligned src
+	.align	4		# 1 mod 4 alignment for LOOPNEZ
+	.byte	0		# (0 mod 4 alignment for LBEG)
+.Laligned:
+#if XCHAL_HAVE_LOOPS
+	loopnez	a7, .Loop1done
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a7, .Loop1done
+	slli	a8, a7, 4
+	add	a8, a8, a3	# a8 = end of last 16B source chunk
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop1:
+	EX(l32i, a6, a3,  0, l_fixup)
+	EX(l32i, a7, a3,  4, l_fixup)
+	EX(s32i, a6, a5,  0, s_fixup)
+	EX(l32i, a6, a3,  8, l_fixup)
+	EX(s32i, a7, a5,  4, s_fixup)
+	EX(l32i, a7, a3, 12, l_fixup)
+	EX(s32i, a6, a5,  8, s_fixup)
+	addi	a3, a3, 16
+	EX(s32i, a7, a5, 12, s_fixup)
+	addi	a5, a5, 16
+#if !XCHAL_HAVE_LOOPS
+	blt	a3, a8, .Loop1
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop1done:
+	bbci.l	a4, 3, .L2
+	# copy 8 bytes
+	EX(l32i, a6, a3,  0, l_fixup)
+	EX(l32i, a7, a3,  4, l_fixup)
+	addi	a3, a3,  8
+	EX(s32i, a6, a5,  0, s_fixup)
+	EX(s32i, a7, a5,  4, s_fixup)
+	addi	a5, a5,  8
+.L2:
+	bbci.l	a4, 2, .L3
+	# copy 4 bytes
+	EX(l32i, a6, a3,  0, l_fixup)
+	addi	a3, a3,  4
+	EX(s32i, a6, a5,  0, s_fixup)
+	addi	a5, a5,  4
+.L3:
+	bbci.l	a4, 1, .L4
+	# copy 2 bytes
+	EX(l16ui, a6, a3,  0, l_fixup)
+	addi	a3, a3,  2
+	EX(s16i,  a6, a5,  0, s_fixup)
+	addi	a5, a5,  2
+.L4:
+	bbci.l	a4, 0, .L5
+	# copy 1 byte
+	EX(l8ui, a6, a3,  0, l_fixup)
+	EX(s8i,  a6, a5,  0, s_fixup)
+.L5:
+	movi	a2, 0		# return success for len bytes copied
+	retw
+
+/*
+ * Destination is aligned, Source is unaligned
+ */
+
+	.align	4
+	.byte	0		# 1 mod 4 alignement for LOOPNEZ
+				# (0 mod 4 alignment for LBEG)
+.Lsrcunaligned:
+	# copy 16 bytes per iteration for word-aligned dst and unaligned src
+	and	a10, a3, a8	# save unalignment offset for below
+	sub	a3, a3, a10	# align a3 (to avoid sim warnings only; not needed for hardware)
+	EX(l32i, a6, a3, 0, l_fixup)	# load first word
+#if XCHAL_HAVE_LOOPS
+	loopnez	a7, .Loop2done
+#else /* !XCHAL_HAVE_LOOPS */
+	beqz	a7, .Loop2done
+	slli	a10, a7, 4
+	add	a10, a10, a3	# a10 = end of last 16B source chunk
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop2:
+	EX(l32i, a7, a3,  4, l_fixup)
+	EX(l32i, a8, a3,  8, l_fixup)
+	ALIGN(	a6, a6, a7)
+	EX(s32i, a6, a5,  0, s_fixup)
+	EX(l32i, a9, a3, 12, l_fixup)
+	ALIGN(	a7, a7, a8)
+	EX(s32i, a7, a5,  4, s_fixup)
+	EX(l32i, a6, a3, 16, l_fixup)
+	ALIGN(	a8, a8, a9)
+	EX(s32i, a8, a5,  8, s_fixup)
+	addi	a3, a3, 16
+	ALIGN(	a9, a9, a6)
+	EX(s32i, a9, a5, 12, s_fixup)
+	addi	a5, a5, 16
+#if !XCHAL_HAVE_LOOPS
+	blt	a3, a10, .Loop2
+#endif /* !XCHAL_HAVE_LOOPS */
+.Loop2done:
+	bbci.l	a4, 3, .L12
+	# copy 8 bytes
+	EX(l32i, a7, a3,  4, l_fixup)
+	EX(l32i, a8, a3,  8, l_fixup)
+	ALIGN(	a6, a6, a7)
+	EX(s32i, a6, a5,  0, s_fixup)
+	addi	a3, a3,  8
+	ALIGN(	a7, a7, a8)
+	EX(s32i, a7, a5,  4, s_fixup)
+	addi	a5, a5,  8
+	mov	a6, a8
+.L12:
+	bbci.l	a4, 2, .L13
+	# copy 4 bytes
+	EX(l32i, a7, a3,  4, l_fixup)
+	addi	a3, a3,  4
+	ALIGN(	a6, a6, a7)
+	EX(s32i, a6, a5,  0, s_fixup)
+	addi	a5, a5,  4
+	mov	a6, a7
+.L13:
+	add	a3, a3, a10	# readjust a3 with correct misalignment
+	bbci.l	a4, 1, .L14
+	# copy 2 bytes
+	EX(l8ui, a6, a3,  0, l_fixup)
+	EX(l8ui, a7, a3,  1, l_fixup)
+	addi	a3, a3,  2
+	EX(s8i, a6, a5,  0, s_fixup)
+	EX(s8i, a7, a5,  1, s_fixup)
+	addi	a5, a5,  2
+.L14:
+	bbci.l	a4, 0, .L15
+	# copy 1 byte
+	EX(l8ui, a6, a3,  0, l_fixup)
+	EX(s8i,  a6, a5,  0, s_fixup)
+.L15:
+	movi	a2, 0		# return success for len bytes copied
+	retw
+
+
+	.section .fixup, "ax"
+	.align	4
+
+/* a2 = original dst; a5 = current dst; a11= original len
+ * bytes_copied = a5 - a2
+ * retval = bytes_not_copied = original len - bytes_copied
+ * retval = a11 - (a5 - a2)
+ *
+ * Clearing the remaining pieces of kernel memory plugs security
+ * holes.  This functionality is the equivalent of the *_zeroing
+ * functions that some architectures provide.
+ */
+
+.Lmemset:
+	.word	memset
+
+s_fixup:
+	sub	a2, a5, a2	/* a2 <-- bytes copied */
+	sub	a2, a11, a2	/* a2 <-- bytes not copied */
+	retw
+
+l_fixup:
+	sub	a2, a5, a2	/* a2 <-- bytes copied */
+	sub	a2, a11, a2	/* a2 <-- bytes not copied == return value */
+
+	/* void *memset(void *s, int c, size_t n); */
+	mov	a6, a5		/* s */
+	movi	a7, 0		/* c */
+	mov	a8, a2		/* n */
+	l32r	a4, .Lmemset
+	callx4	a4
+	/* Ignore memset return value in a6. */
+	/* a2 still contains bytes not copied. */
+	retw
+
diff --git a/arch/xtensa/mm/Makefile b/arch/xtensa/mm/Makefile
new file mode 100644
index 000000000000..a5aed5932d7b
--- /dev/null
+++ b/arch/xtensa/mm/Makefile
@@ -0,0 +1,13 @@
+#
+# Makefile for the Linux/Xtensa-specific parts of the memory manager.
+#
+# Note! Dependencies are done automagically by 'make dep', which also
+# removes any old dependencies. DON'T put your own dependencies here
+# unless it's something special (ie not a .c file).
+#
+# Note 2! The CFLAGS definition is now in the main makefile...
+
+obj-y	 := init.o fault.o tlb.o misc.o
+obj-m	 :=
+obj-n	 :=
+obj-	 :=
diff --git a/arch/xtensa/mm/fault.c b/arch/xtensa/mm/fault.c
new file mode 100644
index 000000000000..a945a33e85a1
--- /dev/null
+++ b/arch/xtensa/mm/fault.c
@@ -0,0 +1,241 @@
+// TODO VM_EXEC flag work-around, cache aliasing
+/*
+ * arch/xtensa/mm/fault.c
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ */
+
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+#include <asm/hardirq.h>
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/pgalloc.h>
+
+unsigned long asid_cache = ASID_FIRST_VERSION;
+void bad_page_fault(struct pt_regs*, unsigned long, int);
+
+/*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ *
+ * Note: does not handle Miss and MultiHit.
+ */
+
+void do_page_fault(struct pt_regs *regs)
+{
+	struct vm_area_struct * vma;
+	struct mm_struct *mm = current->mm;
+	unsigned int exccause = regs->exccause;
+	unsigned int address = regs->excvaddr;
+	siginfo_t info;
+
+	int is_write, is_exec;
+
+	info.si_code = SEGV_MAPERR;
+
+	/* We fault-in kernel-space virtual memory on-demand. The
+	 * 'reference' page table is init_mm.pgd.
+	 */
+	if (address >= TASK_SIZE && !user_mode(regs))
+		goto vmalloc_fault;
+
+	/* If we're in an interrupt or have no user
+	 * context, we must not take the fault..
+	 */
+	if (in_atomic() || !mm) {
+		bad_page_fault(regs, address, SIGSEGV);
+		return;
+	}
+
+	is_write = (exccause == XCHAL_EXCCAUSE_STORE_CACHE_ATTRIBUTE) ? 1 : 0;
+	is_exec =  (exccause == XCHAL_EXCCAUSE_ITLB_PRIVILEGE ||
+		    exccause == XCHAL_EXCCAUSE_ITLB_MISS ||
+		    exccause == XCHAL_EXCCAUSE_FETCH_CACHE_ATTRIBUTE) ? 1 : 0;
+
+#if 0
+	printk("[%s:%d:%08x:%d:%08x:%s%s]\n", current->comm, current->pid,
+	       address, exccause, regs->pc, is_write? "w":"", is_exec? "x":"");
+#endif
+
+	down_read(&mm->mmap_sem);
+	vma = find_vma(mm, address);
+
+	if (!vma)
+		goto bad_area;
+	if (vma->vm_start <= address)
+		goto good_area;
+	if (!(vma->vm_flags & VM_GROWSDOWN))
+		goto bad_area;
+	if (expand_stack(vma, address))
+		goto bad_area;
+
+	/* Ok, we have a good vm_area for this memory access, so
+	 * we can handle it..
+	 */
+
+good_area:
+	info.si_code = SEGV_ACCERR;
+
+	if (is_write) {
+		if (!(vma->vm_flags & VM_WRITE))
+			goto bad_area;
+	} else if (is_exec) {
+		if (!(vma->vm_flags & VM_EXEC))
+			goto bad_area;
+	} else	/* Allow read even from write-only pages. */
+		if (!(vma->vm_flags & (VM_READ | VM_WRITE)))
+			goto bad_area;
+
+	/* If for any reason at all we couldn't handle the fault,
+	 * make sure we exit gracefully rather than endlessly redo
+	 * the fault.
+	 */
+survive:
+	switch (handle_mm_fault(mm, vma, address, is_write)) {
+	case VM_FAULT_MINOR:
+		current->min_flt++;
+		break;
+	case VM_FAULT_MAJOR:
+		current->maj_flt++;
+		break;
+	case VM_FAULT_SIGBUS:
+		goto do_sigbus;
+	case VM_FAULT_OOM:
+		goto out_of_memory;
+	default:
+		BUG();
+	}
+
+	up_read(&mm->mmap_sem);
+	return;
+
+	/* Something tried to access memory that isn't in our memory map..
+	 * Fix it, but check if it's kernel or user first..
+	 */
+bad_area:
+	up_read(&mm->mmap_sem);
+	if (user_mode(regs)) {
+		current->thread.bad_vaddr = address;
+		current->thread.error_code = is_write;
+		info.si_signo = SIGSEGV;
+		info.si_errno = 0;
+		/* info.si_code has been set above */
+		info.si_addr = (void *) address;
+		force_sig_info(SIGSEGV, &info, current);
+		return;
+	}
+	bad_page_fault(regs, address, SIGSEGV);
+	return;
+
+
+	/* We ran out of memory, or some other thing happened to us that made
+	 * us unable to handle the page fault gracefully.
+	 */
+out_of_memory:
+	up_read(&mm->mmap_sem);
+	if (current->pid == 1) {
+		yield();
+		down_read(&mm->mmap_sem);
+		goto survive;
+	}
+	printk("VM: killing process %s\n", current->comm);
+	if (user_mode(regs))
+		do_exit(SIGKILL);
+	bad_page_fault(regs, address, SIGKILL);
+	return;
+
+do_sigbus:
+	up_read(&mm->mmap_sem);
+
+	/* Send a sigbus, regardless of whether we were in kernel
+	 * or user mode.
+	 */
+	current->thread.bad_vaddr = address;
+	info.si_code = SIGBUS;
+	info.si_errno = 0;
+	info.si_code = BUS_ADRERR;
+	info.si_addr = (void *) address;
+	force_sig_info(SIGBUS, &info, current);
+
+	/* Kernel mode? Handle exceptions or die */
+	if (!user_mode(regs))
+		bad_page_fault(regs, address, SIGBUS);
+
+vmalloc_fault:
+	{
+		/* Synchronize this task's top level page-table
+		 * with the 'reference' page table.
+		 */
+		struct mm_struct *act_mm = current->active_mm;
+		int index = pgd_index(address);
+		pgd_t *pgd, *pgd_k;
+		pmd_t *pmd, *pmd_k;
+		pte_t *pte_k;
+
+		if (act_mm == NULL)
+			goto bad_page_fault;
+
+		pgd = act_mm->pgd + index;
+		pgd_k = init_mm.pgd + index;
+
+		if (!pgd_present(*pgd_k))
+			goto bad_page_fault;
+
+		pgd_val(*pgd) = pgd_val(*pgd_k);
+
+		pmd = pmd_offset(pgd, address);
+		pmd_k = pmd_offset(pgd_k, address);
+		if (!pmd_present(*pmd) || !pmd_present(*pmd_k))
+			goto bad_page_fault;
+
+		pmd_val(*pmd) = pmd_val(*pmd_k);
+		pte_k = pte_offset_kernel(pmd_k, address);
+
+		if (!pte_present(*pte_k))
+			goto bad_page_fault;
+		return;
+	}
+bad_page_fault:
+	bad_page_fault(regs, address, SIGKILL);
+	return;
+}
+
+
+void
+bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
+{
+	extern void die(const char*, struct pt_regs*, long);
+	const struct exception_table_entry *entry;
+
+	/* Are we prepared to handle this kernel fault?  */
+	if ((entry = search_exception_tables(regs->pc)) != NULL) {
+#if 1
+		printk(KERN_DEBUG "%s: Exception at pc=%#010lx (%lx)\n",
+				current->comm, regs->pc, entry->fixup);
+#endif
+		current->thread.bad_uaddr = address;
+		regs->pc = entry->fixup;
+		return;
+	}
+
+	/* Oops. The kernel tried to access some bad page. We'll have to
+	 * terminate things with extreme prejudice.
+	 */
+	printk(KERN_ALERT "Unable to handle kernel paging request at virtual "
+	       "address %08lx\n pc = %08lx, ra = %08lx\n",
+	       address, regs->pc, regs->areg[0]);
+	die("Oops", regs, sig);
+	do_exit(sig);
+}
+
diff --git a/arch/xtensa/mm/init.c b/arch/xtensa/mm/init.c
new file mode 100644
index 000000000000..56aace84aaeb
--- /dev/null
+++ b/arch/xtensa/mm/init.c
@@ -0,0 +1,551 @@
+/*
+ * arch/xtensa/mm/init.c
+ *
+ * Derived from MIPS, PPC.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel	<chris@zankel.net>
+ * Joe Taylor	<joe@tensilica.com, joetylr@yahoo.com>
+ * Marc Gauthier
+ * Kevin Chea
+ */
+
+#include <linux/config.h>
+#include <linux/init.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/bootmem.h>
+#include <linux/swap.h>
+
+#include <asm/pgtable.h>
+#include <asm/bootparam.h>
+#include <asm/mmu_context.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/pgtable.h>
+
+
+#define DEBUG 0
+
+DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+//static DEFINE_SPINLOCK(tlb_lock);
+
+/*
+ * This flag is used to indicate that the page was mapped and modified in
+ * kernel space, so the cache is probably dirty at that address.
+ * If cache aliasing is enabled and the page color mismatches, update_mmu_cache
+ * synchronizes the caches if this bit is set.
+ */
+
+#define PG_cache_clean PG_arch_1
+
+/* References to section boundaries */
+
+extern char _ftext, _etext, _fdata, _edata, _rodata_end;
+extern char __init_begin, __init_end;
+
+/*
+ * mem_reserve(start, end, must_exist)
+ *
+ * Reserve some memory from the memory pool.
+ *
+ * Parameters:
+ *  start	Start of region,
+ *  end		End of region,
+ *  must_exist	Must exist in memory pool.
+ *
+ * Returns:
+ *  0 (memory area couldn't be mapped)
+ * -1 (success)
+ */
+
+int __init mem_reserve(unsigned long start, unsigned long end, int must_exist)
+{
+	int i;
+
+	if (start == end)
+		return 0;
+
+	start = start & PAGE_MASK;
+	end = PAGE_ALIGN(end);
+
+	for (i = 0; i < sysmem.nr_banks; i++)
+		if (start < sysmem.bank[i].end
+		    && end >= sysmem.bank[i].start)
+			break;
+
+	if (i == sysmem.nr_banks) {
+		if (must_exist)
+			printk (KERN_WARNING "mem_reserve: [0x%0lx, 0x%0lx) "
+				"not in any region!\n", start, end);
+		return 0;
+	}
+
+	if (start > sysmem.bank[i].start) {
+		if (end < sysmem.bank[i].end) {
+			/* split entry */
+			if (sysmem.nr_banks >= SYSMEM_BANKS_MAX)
+				panic("meminfo overflow\n");
+			sysmem.bank[sysmem.nr_banks].start = end;
+			sysmem.bank[sysmem.nr_banks].end = sysmem.bank[i].end;
+			sysmem.nr_banks++;
+		}
+		sysmem.bank[i].end = start;
+	} else {
+		if (end < sysmem.bank[i].end)
+			sysmem.bank[i].start = end;
+		else {
+			/* remove entry */
+			sysmem.nr_banks--;
+			sysmem.bank[i].start = sysmem.bank[sysmem.nr_banks].start;
+			sysmem.bank[i].end   = sysmem.bank[sysmem.nr_banks].end;
+		}
+	}
+	return -1;
+}
+
+
+/*
+ * Initialize the bootmem system and give it all the memory we have available.
+ */
+
+void __init bootmem_init(void)
+{
+	unsigned long pfn;
+	unsigned long bootmap_start, bootmap_size;
+	int i;
+
+	max_low_pfn = max_pfn = 0;
+	min_low_pfn = ~0;
+
+	for (i=0; i < sysmem.nr_banks; i++) {
+		pfn = PAGE_ALIGN(sysmem.bank[i].start) >> PAGE_SHIFT;
+		if (pfn < min_low_pfn)
+			min_low_pfn = pfn;
+		pfn = PAGE_ALIGN(sysmem.bank[i].end - 1) >> PAGE_SHIFT;
+		if (pfn > max_pfn)
+			max_pfn = pfn;
+	}
+
+	if (min_low_pfn > max_pfn)
+		panic("No memory found!\n");
+
+	max_low_pfn = max_pfn < MAX_LOW_MEMORY >> PAGE_SHIFT ?
+		max_pfn : MAX_LOW_MEMORY >> PAGE_SHIFT;
+
+	/* Find an area to use for the bootmem bitmap. */
+
+	bootmap_size = bootmem_bootmap_pages(max_low_pfn) << PAGE_SHIFT;
+	bootmap_start = ~0;
+
+	for (i=0; i<sysmem.nr_banks; i++)
+		if (sysmem.bank[i].end - sysmem.bank[i].start >= bootmap_size) {
+			bootmap_start = sysmem.bank[i].start;
+			break;
+		}
+
+	if (bootmap_start == ~0UL)
+		panic("Cannot find %ld bytes for bootmap\n", bootmap_size);
+
+	/* Reserve the bootmem bitmap area */
+
+	mem_reserve(bootmap_start, bootmap_start + bootmap_size, 1);
+	bootmap_size = init_bootmem_node(NODE_DATA(0), min_low_pfn,
+					 bootmap_start >> PAGE_SHIFT,
+					 max_low_pfn);
+
+	/* Add all remaining memory pieces into the bootmem map */
+
+	for (i=0; i<sysmem.nr_banks; i++)
+		free_bootmem(sysmem.bank[i].start,
+			     sysmem.bank[i].end - sysmem.bank[i].start);
+
+}
+
+
+void __init paging_init(void)
+{
+	unsigned long zones_size[MAX_NR_ZONES];
+	int i;
+
+	/* All pages are DMA-able, so we put them all in the DMA zone. */
+
+	zones_size[ZONE_DMA] = max_low_pfn;
+	for (i = 1; i < MAX_NR_ZONES; i++)
+		zones_size[i] = 0;
+
+#ifdef CONFIG_HIGHMEM
+	zones_size[ZONE_HIGHMEM] = max_pfn - max_low_pfn;
+#endif
+
+	/* Initialize the kernel's page tables. */
+
+	memset(swapper_pg_dir, 0, PAGE_SIZE);
+
+	free_area_init(zones_size);
+}
+
+/*
+ * Flush the mmu and reset associated register to default values.
+ */
+
+void __init init_mmu (void)
+{
+	/* Writing zeros to the <t>TLBCFG special registers ensure
+	 * that valid values exist in the register.  For existing
+	 * PGSZID<w> fields, zero selects the first element of the
+	 * page-size array.  For nonexistant PGSZID<w> fields, zero is
+	 * the best value to write.  Also, when changing PGSZID<w>
+	 * fields, the corresponding TLB must be flushed.
+	 */
+	set_itlbcfg_register (0);
+	set_dtlbcfg_register (0);
+	flush_tlb_all ();
+
+	/* Set rasid register to a known value. */
+
+	set_rasid_register (ASID_ALL_RESERVED);
+
+	/* Set PTEVADDR special register to the start of the page
+	 * table, which is in kernel mappable space (ie. not
+	 * statically mapped).  This register's value is undefined on
+	 * reset.
+	 */
+	set_ptevaddr_register (PGTABLE_START);
+}
+
+/*
+ * Initialize memory pages.
+ */
+
+void __init mem_init(void)
+{
+	unsigned long codesize, reservedpages, datasize, initsize;
+	unsigned long highmemsize, tmp, ram;
+
+	max_mapnr = num_physpages = max_low_pfn;
+	high_memory = (void *) __va(max_mapnr << PAGE_SHIFT);
+	highmemsize = 0;
+
+#if CONFIG_HIGHMEM
+#error HIGHGMEM not implemented in init.c
+#endif
+
+	totalram_pages += free_all_bootmem();
+
+	reservedpages = ram = 0;
+	for (tmp = 0; tmp < max_low_pfn; tmp++) {
+		ram++;
+		if (PageReserved(mem_map+tmp))
+			reservedpages++;
+	}
+
+	codesize =  (unsigned long) &_etext - (unsigned long) &_ftext;
+	datasize =  (unsigned long) &_edata - (unsigned long) &_fdata;
+	initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
+
+	printk("Memory: %luk/%luk available (%ldk kernel code, %ldk reserved, "
+	       "%ldk data, %ldk init %ldk highmem)\n",
+	       (unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
+	       ram << (PAGE_SHIFT-10),
+	       codesize >> 10,
+	       reservedpages << (PAGE_SHIFT-10),
+	       datasize >> 10,
+	       initsize >> 10,
+	       highmemsize >> 10);
+}
+
+void
+free_reserved_mem(void *start, void *end)
+{
+	for (; start < end; start += PAGE_SIZE) {
+		ClearPageReserved(virt_to_page(start));
+		set_page_count(virt_to_page(start), 1);
+		free_page((unsigned long)start);
+		totalram_pages++;
+	}
+}
+
+#ifdef CONFIG_BLK_DEV_INITRD
+extern int initrd_is_mapped;
+
+void free_initrd_mem(unsigned long start, unsigned long end)
+{
+	if (initrd_is_mapped) {
+		free_reserved_mem((void*)start, (void*)end);
+		printk ("Freeing initrd memory: %ldk freed\n",(end-start)>>10);
+	}
+}
+#endif
+
+void free_initmem(void)
+{
+	free_reserved_mem(&__init_begin, &__init_end);
+	printk("Freeing unused kernel memory: %dk freed\n",
+	       (&__init_end - &__init_begin) >> 10);
+}
+
+void show_mem(void)
+{
+	int i, free = 0, total = 0, reserved = 0;
+	int shared = 0, cached = 0;
+
+	printk("Mem-info:\n");
+	show_free_areas();
+	printk("Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
+	i = max_mapnr;
+	while (i-- > 0) {
+		total++;
+		if (PageReserved(mem_map+i))
+			reserved++;
+		else if (PageSwapCache(mem_map+i))
+			cached++;
+		else if (!page_count(mem_map + i))
+			free++;
+		else
+			shared += page_count(mem_map + i) - 1;
+	}
+	printk("%d pages of RAM\n", total);
+	printk("%d reserved pages\n", reserved);
+	printk("%d pages shared\n", shared);
+	printk("%d pages swap cached\n",cached);
+	printk("%d free pages\n", free);
+}
+
+/* ------------------------------------------------------------------------- */
+
+#if (DCACHE_WAY_SIZE > PAGE_SIZE)
+
+/*
+ * With cache aliasing, the page color of the page in kernel space and user
+ * space might mismatch. We temporarily map the page to a different virtual
+ * address with the same color and clear the page there.
+ */
+
+void clear_user_page(void *kaddr, unsigned long vaddr, struct page* page)
+{
+
+  	/*  There shouldn't be any entries for this page. */
+
+	__flush_invalidate_dcache_page_phys(__pa(page_address(page)));
+
+	if (!PAGE_COLOR_EQ(vaddr, kaddr)) {
+		unsigned long v, p;
+
+		/* Temporarily map page to DTLB_WAY_DCACHE_ALIAS0. */
+
+		spin_lock(&tlb_lock);
+
+		p = (unsigned long)pte_val((mk_pte(page,PAGE_KERNEL)));
+		kaddr = (void*)PAGE_COLOR_MAP0(vaddr);
+		v = (unsigned long)kaddr | DTLB_WAY_DCACHE_ALIAS0;
+		__asm__ __volatile__("wdtlb %0,%1; dsync" : :"a" (p), "a" (v));
+
+		clear_page(kaddr);
+
+		spin_unlock(&tlb_lock);
+	} else {
+		clear_page(kaddr);
+	}
+
+	/* We need to make sure that i$ and d$ are coherent. */
+
+	clear_bit(PG_cache_clean, &page->flags);
+}
+
+/*
+ * With cache aliasing, we have to make sure that the page color of the page
+ * in kernel space matches that of the virtual user address before we read
+ * the page. If the page color differ, we create a temporary DTLB entry with
+ * the corrent page color and use this 'temporary' address as the source.
+ * We then use the same approach as in clear_user_page and copy the data
+ * to the kernel space and clear the PG_cache_clean bit to synchronize caches
+ * later.
+ *
+ * Note:
+ * Instead of using another 'way' for the temporary DTLB entry, we could
+ * probably use the same entry that points to the kernel address (after
+ * saving the original value and restoring it when we are done).
+ */
+
+void copy_user_page(void* to, void* from, unsigned long vaddr,
+    		    struct page* to_page)
+{
+	/* There shouldn't be any entries for the new page. */
+
+	__flush_invalidate_dcache_page_phys(__pa(page_address(to_page)));
+
+	spin_lock(&tlb_lock);
+
+	if (!PAGE_COLOR_EQ(vaddr, from)) {
+		unsigned long v, p, t;
+
+		__asm__ __volatile__ ("pdtlb %1,%2; rdtlb1 %0,%1"
+				      : "=a"(p), "=a"(t) : "a"(from));
+		from = (void*)PAGE_COLOR_MAP0(vaddr);
+		v = (unsigned long)from | DTLB_WAY_DCACHE_ALIAS0;
+		__asm__ __volatile__ ("wdtlb %0,%1; dsync" ::"a" (p), "a" (v));
+	}
+
+	if (!PAGE_COLOR_EQ(vaddr, to)) {
+		unsigned long v, p;
+
+		p = (unsigned long)pte_val((mk_pte(to_page,PAGE_KERNEL)));
+		to = (void*)PAGE_COLOR_MAP1(vaddr);
+		v = (unsigned long)to | DTLB_WAY_DCACHE_ALIAS1;
+		__asm__ __volatile__ ("wdtlb %0,%1; dsync" ::"a" (p), "a" (v));
+	}
+	copy_page(to, from);
+
+	spin_unlock(&tlb_lock);
+
+	/* We need to make sure that i$ and d$ are coherent. */
+
+	clear_bit(PG_cache_clean, &to_page->flags);
+}
+
+
+
+/*
+ * Any time the kernel writes to a user page cache page, or it is about to
+ * read from a page cache page this routine is called.
+ *
+ * Note:
+ * The kernel currently only provides one architecture bit in the page
+ * flags that we use for I$/D$ coherency. Maybe, in future, we can
+ * use a sepearte bit for deferred dcache aliasing:
+ * If the page is not mapped yet, we only need to set a flag,
+ * if mapped, we need to invalidate the page.
+ */
+// FIXME: we probably need this for WB caches not only for Page Coloring..
+
+void flush_dcache_page(struct page *page)
+{
+	unsigned long addr = __pa(page_address(page));
+	struct address_space *mapping = page_mapping(page);
+
+	__flush_invalidate_dcache_page_phys(addr);
+
+	if (!test_bit(PG_cache_clean, &page->flags))
+		return;
+
+	/* If this page hasn't been mapped, yet, handle I$/D$ coherency later.*/
+#if 0
+	if (mapping && !mapping_mapped(mapping))
+		clear_bit(PG_cache_clean, &page->flags);
+	else
+#endif
+		__invalidate_icache_page_phys(addr);
+}
+
+void flush_cache_range(struct vm_area_struct* vma, unsigned long s,
+		       unsigned long e)
+{
+	__flush_invalidate_cache_all();
+}
+
+void flush_cache_page(struct vm_area_struct* vma, unsigned long address,
+    		      unsigned long pfn)
+{
+	struct page *page = pfn_to_page(pfn);
+
+	/* Remove any entry for the old mapping. */
+
+	if (current->active_mm == vma->vm_mm) {
+		unsigned long addr = __pa(page_address(page));
+		__flush_invalidate_dcache_page_phys(addr);
+		if ((vma->vm_flags & VM_EXEC) != 0)
+			__invalidate_icache_page_phys(addr);
+	} else {
+		BUG();
+	}
+}
+
+#endif	/* (DCACHE_WAY_SIZE > PAGE_SIZE) */
+
+
+pte_t* pte_alloc_one_kernel (struct mm_struct* mm, unsigned long addr)
+{
+	pte_t* pte = (pte_t*)__get_free_pages(GFP_KERNEL|__GFP_REPEAT, 0);
+	if (likely(pte)) {
+	       	pte_t* ptep = (pte_t*)(pte_val(*pte) + PAGE_OFFSET);
+		int i;
+		for (i = 0; i < 1024; i++, ptep++)
+			pte_clear(mm, addr, ptep);
+	}
+	return pte;
+}
+
+struct page* pte_alloc_one(struct mm_struct *mm, unsigned long addr)
+{
+	struct page *page;
+
+	page = alloc_pages(GFP_KERNEL | __GFP_REPEAT, 0);
+
+	if (likely(page)) {
+		pte_t* ptep = kmap_atomic(page, KM_USER0);
+		int i;
+
+		for (i = 0; i < 1024; i++, ptep++)
+			pte_clear(mm, addr, ptep);
+
+		kunmap_atomic(ptep, KM_USER0);
+	}
+	return page;
+}
+
+
+/*
+ * Handle D$/I$ coherency.
+ *
+ * Note:
+ * We only have one architecture bit for the page flags, so we cannot handle
+ * cache aliasing, yet.
+ */
+
+void
+update_mmu_cache(struct vm_area_struct * vma, unsigned long addr, pte_t pte)
+{
+	unsigned long pfn = pte_pfn(pte);
+	struct page *page;
+	unsigned long vaddr = addr & PAGE_MASK;
+
+	if (!pfn_valid(pfn))
+		return;
+
+	page = pfn_to_page(pfn);
+
+	invalidate_itlb_mapping(addr);
+	invalidate_dtlb_mapping(addr);
+
+	/* We have a new mapping. Use it. */
+
+	write_dtlb_entry(pte, dtlb_probe(addr));
+
+	/* If the processor can execute from this page, synchronize D$/I$. */
+
+	if ((vma->vm_flags & VM_EXEC) != 0) {
+
+		write_itlb_entry(pte, itlb_probe(addr));
+
+		/* Synchronize caches, if not clean. */
+
+		if (!test_and_set_bit(PG_cache_clean, &page->flags)) {
+			__flush_dcache_page(vaddr);
+			__invalidate_icache_page(vaddr);
+		}
+	}
+}
+
diff --git a/arch/xtensa/mm/misc.S b/arch/xtensa/mm/misc.S
new file mode 100644
index 000000000000..327c0f17187c
--- /dev/null
+++ b/arch/xtensa/mm/misc.S
@@ -0,0 +1,374 @@
+/*
+ * arch/xtensa/mm/misc.S
+ *
+ * Miscellaneous assembly functions.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel	<chris@zankel.net>
+ */
+
+/* Note: we might want to implement some of the loops as zero-overhead-loops,
+ *	 where applicable and if supported by the processor.
+ */
+
+#include <linux/linkage.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+
+#include <xtensa/cacheasm.h>
+#include <xtensa/cacheattrasm.h>
+
+/* clear_page (page) */
+
+ENTRY(clear_page)
+	entry	a1, 16
+	addi	a4, a2, PAGE_SIZE
+	movi	a3, 0
+
+1:	s32i	a3, a2, 0
+	s32i	a3, a2, 4
+	s32i	a3, a2, 8
+	s32i	a3, a2, 12
+	s32i	a3, a2, 16
+	s32i	a3, a2, 20
+	s32i	a3, a2, 24
+	s32i	a3, a2, 28
+	addi	a2, a2, 32
+	blt	a2, a4, 1b
+
+	retw
+
+/*
+ * copy_page (void *to, void *from)
+ *                  a2        a3
+ */
+
+ENTRY(copy_page)
+	entry	a1, 16
+	addi	a4, a2, PAGE_SIZE
+
+1:	l32i	a5, a3, 0
+	l32i	a6, a3, 4
+	l32i	a7, a3, 8
+	s32i	a5, a2, 0
+	s32i	a6, a2, 4
+	s32i	a7, a2, 8
+	l32i	a5, a3, 12
+	l32i	a6, a3, 16
+	l32i	a7, a3, 20
+	s32i	a5, a2, 12
+	s32i	a6, a2, 16
+	s32i	a7, a2, 20
+	l32i	a5, a3, 24
+	l32i	a6, a3, 28
+	s32i	a5, a2, 24
+	s32i	a6, a2, 28
+	addi	a2, a2, 32
+	addi	a3, a3, 32
+	blt	a2, a4, 1b
+
+	retw
+
+
+/*
+ * void __flush_invalidate_cache_all(void)
+ */
+
+ENTRY(__flush_invalidate_cache_all)
+	entry	sp, 16
+	dcache_writeback_inv_all a2, a3
+	icache_invalidate_all a2, a3
+	retw
+
+/*
+ * void __invalidate_icache_all(void)
+ */
+
+ENTRY(__invalidate_icache_all)
+	entry	sp, 16
+	icache_invalidate_all a2, a3
+	retw
+
+/*
+ * void __flush_invalidate_dcache_all(void)
+ */
+
+ENTRY(__flush_invalidate_dcache_all)
+	entry	sp, 16
+	dcache_writeback_inv_all a2, a3
+	retw
+
+
+/*
+ * void __flush_invalidate_cache_range(ulong start, ulong size)
+ */
+
+ENTRY(__flush_invalidate_cache_range)
+	entry	sp, 16
+	mov	a4, a2
+	mov	a5, a3
+	dcache_writeback_inv_region a4, a5, a6
+	icache_invalidate_region a2, a3, a4
+	retw
+
+/*
+ * void __invalidate_icache_page(ulong start)
+ */
+
+ENTRY(__invalidate_icache_page)
+	entry	sp, 16
+	movi	a3, PAGE_SIZE
+	icache_invalidate_region a2, a3, a4
+	retw
+
+/*
+ * void __invalidate_dcache_page(ulong start)
+ */
+
+ENTRY(__invalidate_dcache_page)
+	entry	sp, 16
+	movi	a3, PAGE_SIZE
+	dcache_invalidate_region a2, a3, a4
+	retw
+
+/*
+ * void __invalidate_icache_range(ulong start, ulong size)
+ */
+
+ENTRY(__invalidate_icache_range)
+	entry	sp, 16
+	icache_invalidate_region a2, a3, a4
+	retw
+
+/*
+ * void __invalidate_dcache_range(ulong start, ulong size)
+ */
+
+ENTRY(__invalidate_dcache_range)
+	entry	sp, 16
+	dcache_invalidate_region a2, a3, a4
+	retw
+
+/*
+ * void __flush_dcache_page(ulong start)
+ */
+
+ENTRY(__flush_dcache_page)
+	entry	sp, 16
+	movi	a3, PAGE_SIZE
+	dcache_writeback_region a2, a3, a4
+	retw
+
+/*
+ * void __flush_invalidate_dcache_page(ulong start)
+ */
+
+ENTRY(__flush_invalidate_dcache_page)
+	entry	sp, 16
+	movi	a3, PAGE_SIZE
+	dcache_writeback_inv_region a2, a3, a4
+	retw
+
+/*
+ * void __flush_invalidate_dcache_range(ulong start, ulong size)
+ */
+
+ENTRY(__flush_invalidate_dcache_range)
+	entry	sp, 16
+	dcache_writeback_inv_region a2, a3, a4
+	retw
+
+/*
+ * void __invalidate_dcache_all(void)
+ */
+
+ENTRY(__invalidate_dcache_all)
+	entry	sp, 16
+	dcache_invalidate_all a2, a3
+	retw
+
+/*
+ * void __flush_invalidate_dcache_page_phys(ulong start)
+ */
+
+ENTRY(__flush_invalidate_dcache_page_phys)
+	entry	sp, 16
+
+	movi	a3, XCHAL_DCACHE_SIZE
+	movi	a4, PAGE_MASK | 1
+	addi	a2, a2, 1
+
+1:	addi	a3, a3, -XCHAL_DCACHE_LINESIZE
+
+	ldct	a6, a3
+	dsync
+	and	a6, a6, a4
+	beq	a6, a2, 2f
+	bgeui	a3, 2, 1b
+	retw
+
+2:	diwbi	a3, 0
+	bgeui	a3, 2, 1b
+	retw
+
+ENTRY(check_dcache_low0)
+	entry	sp, 16
+
+	movi	a3, XCHAL_DCACHE_SIZE / 4
+	movi	a4, PAGE_MASK | 1
+	addi	a2, a2, 1
+
+1:	addi	a3, a3, -XCHAL_DCACHE_LINESIZE
+
+	ldct	a6, a3
+	dsync
+	and	a6, a6, a4
+	beq	a6, a2, 2f
+	bgeui	a3, 2, 1b
+	retw
+
+2:	j 2b
+
+ENTRY(check_dcache_high0)
+	entry	sp, 16
+
+	movi	a5, XCHAL_DCACHE_SIZE / 4
+	movi	a3, XCHAL_DCACHE_SIZE / 2
+	movi	a4, PAGE_MASK | 1
+	addi	a2, a2, 1
+
+1:	addi	a3, a3, -XCHAL_DCACHE_LINESIZE
+	addi	a5, a5, -XCHAL_DCACHE_LINESIZE
+
+	ldct	a6, a3
+	dsync
+	and	a6, a6, a4
+	beq	a6, a2, 2f
+	bgeui	a5, 2, 1b
+	retw
+
+2:	j 2b
+
+ENTRY(check_dcache_low1)
+	entry	sp, 16
+
+	movi	a5, XCHAL_DCACHE_SIZE / 4
+	movi	a3, XCHAL_DCACHE_SIZE * 3 / 4
+	movi	a4, PAGE_MASK | 1
+	addi	a2, a2, 1
+
+1:	addi	a3, a3, -XCHAL_DCACHE_LINESIZE
+	addi	a5, a5, -XCHAL_DCACHE_LINESIZE
+
+	ldct	a6, a3
+	dsync
+	and	a6, a6, a4
+	beq	a6, a2, 2f
+	bgeui	a5, 2, 1b
+	retw
+
+2:	j 2b
+
+ENTRY(check_dcache_high1)
+	entry	sp, 16
+
+	movi	a5, XCHAL_DCACHE_SIZE / 4
+	movi	a3, XCHAL_DCACHE_SIZE
+	movi	a4, PAGE_MASK | 1
+	addi	a2, a2, 1
+
+1:	addi	a3, a3, -XCHAL_DCACHE_LINESIZE
+	addi	a5, a5, -XCHAL_DCACHE_LINESIZE
+
+	ldct	a6, a3
+	dsync
+	and	a6, a6, a4
+	beq	a6, a2, 2f
+	bgeui	a5, 2, 1b
+	retw
+
+2:	j 2b
+
+
+/*
+ * void __invalidate_icache_page_phys(ulong start)
+ */
+
+ENTRY(__invalidate_icache_page_phys)
+	entry	sp, 16
+
+	movi	a3, XCHAL_ICACHE_SIZE
+	movi	a4, PAGE_MASK | 1
+	addi	a2, a2, 1
+
+1:	addi	a3, a3, -XCHAL_ICACHE_LINESIZE
+
+	lict	a6, a3
+	isync
+	and	a6, a6, a4
+	beq	a6, a2, 2f
+	bgeui	a3, 2, 1b
+	retw
+
+2:	iii	a3, 0
+	bgeui	a3, 2, 1b
+	retw
+
+
+#if 0
+
+	movi	a3, XCHAL_DCACHE_WAYS - 1
+	movi	a4, PAGE_SIZE
+
+1:	mov	a5, a2
+	add	a6, a2, a4
+
+2:	diwbi	a5, 0
+	diwbi	a5, XCHAL_DCACHE_LINESIZE
+	diwbi	a5, XCHAL_DCACHE_LINESIZE * 2
+	diwbi	a5, XCHAL_DCACHE_LINESIZE * 3
+
+	addi	a5, a5, XCHAL_DCACHE_LINESIZE * 4
+	blt	a5, a6, 2b
+
+	addi	a3, a3, -1
+	addi	a2, a2, XCHAL_DCACHE_SIZE / XCHAL_DCACHE_WAYS
+	bgez	a3, 1b
+
+	retw
+
+ENTRY(__invalidate_icache_page_index)
+	entry	sp, 16
+
+	movi	a3, XCHAL_ICACHE_WAYS - 1
+	movi	a4, PAGE_SIZE
+
+1:	mov	a5, a2
+	add	a6, a2, a4
+
+2:	iii	a5, 0
+	iii	a5, XCHAL_ICACHE_LINESIZE
+	iii	a5, XCHAL_ICACHE_LINESIZE * 2
+	iii	a5, XCHAL_ICACHE_LINESIZE * 3
+
+	addi	a5, a5, XCHAL_ICACHE_LINESIZE * 4
+	blt	a5, a6, 2b
+
+	addi	a3, a3, -1
+	addi	a2, a2, XCHAL_ICACHE_SIZE / XCHAL_ICACHE_WAYS
+	bgez	a3, 2b
+
+	retw
+
+#endif
+
+
+
+
+
+
diff --git a/arch/xtensa/mm/pgtable.c b/arch/xtensa/mm/pgtable.c
new file mode 100644
index 000000000000..e5e119c820e4
--- /dev/null
+++ b/arch/xtensa/mm/pgtable.c
@@ -0,0 +1,76 @@
+/*
+ * arch/xtensa/mm/fault.c
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ */
+
+#if (DCACHE_SIZE > PAGE_SIZE)
+
+pte_t* pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
+{
+	pte_t *pte, p;
+	int color = ADDR_COLOR(address);
+	int i;
+
+	p = (pte_t*) __get_free_pages(GFP_KERNEL|__GFP_REPEAT, COLOR_ORDER);
+
+	if (likely(p)) {
+		struct page *page;
+
+		for (i = 0; i < COLOR_SIZE; i++, p++) {
+			page = virt_to_page(pte);
+
+			set_page_count(page, 1);
+			ClearPageCompound(page);
+
+			if (ADDR_COLOR(p) == color)
+				pte = p;
+			else
+				free_page(p);
+		}
+		clear_page(pte);
+	}
+	return pte;
+}
+
+#ifdef PROFILING
+
+int mask;
+int hit;
+int flush;
+
+#endif
+
+struct page* pte_alloc_one(struct mm_struct *mm, unsigned long address)
+{
+	struct page *page, p;
+	int color = ADDR_COLOR(address);
+
+	p = alloc_pages(GFP_KERNEL | __GFP_REPEAT, PTE_ORDER);
+
+	if (likely(p)) {
+		for (i = 0; i < PAGE_ORDER; i++) {
+			set_page_count(p, 1);
+			ClearPageCompound(p);
+
+			if (PADDR_COLOR(page_address(pg)) == color)
+				page = p;
+			else
+				free_page(p);
+		}
+		clear_highpage(page);
+	}
+
+	return page;
+}
+
+#endif
+
+
+
diff --git a/arch/xtensa/mm/tlb.c b/arch/xtensa/mm/tlb.c
new file mode 100644
index 000000000000..d3bd3bfc3b3b
--- /dev/null
+++ b/arch/xtensa/mm/tlb.c
@@ -0,0 +1,545 @@
+/*
+ * arch/xtensa/mm/mmu.c
+ *
+ * Logic that manipulates the Xtensa MMU.  Derived from MIPS.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001 - 2003 Tensilica Inc.
+ *
+ * Joe Taylor
+ * Chris Zankel	<chris@zankel.net>
+ * Marc Gauthier
+ */
+
+#include <linux/mm.h>
+#include <asm/processor.h>
+#include <asm/mmu_context.h>
+#include <asm/tlbflush.h>
+#include <asm/system.h>
+#include <asm/cacheflush.h>
+
+
+static inline void __flush_itlb_all (void)
+{
+	int way, index;
+
+	for (way = 0; way < XCHAL_ITLB_ARF_WAYS; way++) {
+		for (index = 0; index < ITLB_ENTRIES_PER_ARF_WAY; index++) {
+			int entry = way + (index << PAGE_SHIFT);
+			invalidate_itlb_entry_no_isync (entry);
+		}
+	}
+	asm volatile ("isync\n");
+}
+
+static inline void __flush_dtlb_all (void)
+{
+	int way, index;
+
+	for (way = 0; way < XCHAL_DTLB_ARF_WAYS; way++) {
+		for (index = 0; index < DTLB_ENTRIES_PER_ARF_WAY; index++) {
+			int entry = way + (index << PAGE_SHIFT);
+			invalidate_dtlb_entry_no_isync (entry);
+		}
+	}
+	asm volatile ("isync\n");
+}
+
+
+void flush_tlb_all (void)
+{
+	__flush_itlb_all();
+	__flush_dtlb_all();
+}
+
+/* If mm is current, we simply assign the current task a new ASID, thus,
+ * invalidating all previous tlb entries. If mm is someone else's user mapping,
+ * wie invalidate the context, thus, when that user mapping is swapped in,
+ * a new context will be assigned to it.
+ */
+
+void flush_tlb_mm(struct mm_struct *mm)
+{
+#if 0
+	printk("[tlbmm<%lx>]\n", (unsigned long)mm->context);
+#endif
+
+	if (mm == current->active_mm) {
+		int flags;
+		local_save_flags(flags);
+		get_new_mmu_context(mm, asid_cache);
+		set_rasid_register(ASID_INSERT(mm->context));
+		local_irq_restore(flags);
+	}
+	else
+		mm->context = 0;
+}
+
+void flush_tlb_range (struct vm_area_struct *vma,
+    		      unsigned long start, unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	unsigned long flags;
+
+	if (mm->context == NO_CONTEXT)
+		return;
+
+#if 0
+	printk("[tlbrange<%02lx,%08lx,%08lx>]\n",
+			(unsigned long)mm->context, start, end);
+#endif
+	local_save_flags(flags);
+
+	if (end-start + (PAGE_SIZE-1) <= SMALLEST_NTLB_ENTRIES << PAGE_SHIFT) {
+		int oldpid = get_rasid_register();
+		set_rasid_register (ASID_INSERT(mm->context));
+		start &= PAGE_MASK;
+ 		if (vma->vm_flags & VM_EXEC)
+			while(start < end) {
+				invalidate_itlb_mapping(start);
+				invalidate_dtlb_mapping(start);
+				start += PAGE_SIZE;
+			}
+		else
+			while(start < end) {
+				invalidate_dtlb_mapping(start);
+				start += PAGE_SIZE;
+			}
+
+		set_rasid_register(oldpid);
+	} else {
+		get_new_mmu_context(mm, asid_cache);
+		if (mm == current->active_mm)
+			set_rasid_register(ASID_INSERT(mm->context));
+	}
+	local_irq_restore(flags);
+}
+
+void flush_tlb_page (struct vm_area_struct *vma, unsigned long page)
+{
+	struct mm_struct* mm = vma->vm_mm;
+	unsigned long flags;
+	int oldpid;
+#if 0
+	printk("[tlbpage<%02lx,%08lx>]\n",
+			(unsigned long)mm->context, page);
+#endif
+
+	if(mm->context == NO_CONTEXT)
+		return;
+
+	local_save_flags(flags);
+
+       	oldpid = get_rasid_register();
+
+	if (vma->vm_flags & VM_EXEC)
+		invalidate_itlb_mapping(page);
+	invalidate_dtlb_mapping(page);
+
+	set_rasid_register(oldpid);
+
+	local_irq_restore(flags);
+
+#if 0
+	flush_tlb_all();
+	return;
+#endif
+}
+
+
+#ifdef DEBUG_TLB
+
+#define USE_ITLB  0
+#define USE_DTLB  1
+
+struct way_config_t {
+	int indicies;
+	int indicies_log2;
+	int pgsz_log2;
+	int arf;
+};
+
+static struct way_config_t itlb[XCHAL_ITLB_WAYS] =
+{
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY0_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY0_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY0_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY0_SET, ARF)
+	},
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY1_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY1_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY1_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY1_SET, ARF)
+	},
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY2_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY2_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY2_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY2_SET, ARF)
+	},
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY3_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY3_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY3_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY3_SET, ARF)
+	},
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY4_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY4_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY4_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY4_SET, ARF)
+	},
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY5_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY5_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY5_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY5_SET, ARF)
+	},
+	{ XCHAL_ITLB_SET(XCHAL_ITLB_WAY6_SET, ENTRIES),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY6_SET, ENTRIES_LOG2),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY6_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_ITLB_SET(XCHAL_ITLB_WAY6_SET, ARF)
+	}
+};
+
+static struct way_config_t dtlb[XCHAL_DTLB_WAYS] =
+{
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY0_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY0_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY0_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY0_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY1_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY1_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY1_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY1_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY2_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY2_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY2_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY2_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY3_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY3_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY3_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY3_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY4_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY4_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY4_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY4_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY5_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY5_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY5_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY5_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY6_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY6_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY6_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY6_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY7_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY7_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY7_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY7_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY8_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY8_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY8_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY8_SET, ARF)
+	},
+	{ XCHAL_DTLB_SET(XCHAL_DTLB_WAY9_SET, ENTRIES),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY9_SET, ENTRIES_LOG2),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY9_SET, PAGESZ_LOG2_MIN),
+	  XCHAL_DTLB_SET(XCHAL_DTLB_WAY9_SET, ARF)
+	}
+};
+
+/*  Total number of entries:  */
+#define ITLB_TOTAL_ENTRIES	\
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY0_SET, ENTRIES) + \
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY1_SET, ENTRIES) + \
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY2_SET, ENTRIES) + \
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY3_SET, ENTRIES) + \
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY4_SET, ENTRIES) + \
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY5_SET, ENTRIES) + \
+		XCHAL_ITLB_SET(XCHAL_ITLB_WAY6_SET, ENTRIES)
+#define DTLB_TOTAL_ENTRIES	\
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY0_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY1_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY2_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY3_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY4_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY5_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY6_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY7_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY8_SET, ENTRIES) + \
+		XCHAL_DTLB_SET(XCHAL_DTLB_WAY9_SET, ENTRIES)
+
+
+typedef struct {
+    unsigned		va;
+    unsigned		pa;
+    unsigned char	asid;
+    unsigned char	ca;
+    unsigned char	way;
+    unsigned char	index;
+    unsigned char	pgsz_log2;	/* 0 .. 32 */
+    unsigned char	type;		/* 0=ITLB 1=DTLB */
+} tlb_dump_entry_t;
+
+/*  Return -1 if a precedes b, +1 if a follows b, 0 if same:  */
+int cmp_tlb_dump_info( tlb_dump_entry_t *a, tlb_dump_entry_t *b )
+{
+    if (a->asid < b->asid) return -1;
+    if (a->asid > b->asid) return  1;
+    if (a->va < b->va) return -1;
+    if (a->va > b->va) return  1;
+    if (a->pa < b->pa) return -1;
+    if (a->pa > b->pa) return  1;
+    if (a->ca < b->ca) return -1;
+    if (a->ca > b->ca) return  1;
+    if (a->way < b->way) return -1;
+    if (a->way > b->way) return  1;
+    if (a->index < b->index) return -1;
+    if (a->index > b->index) return  1;
+    return 0;
+}
+
+void sort_tlb_dump_info( tlb_dump_entry_t *t, int n )
+{
+    int i, j;
+    /*  Simple O(n*n) sort:  */
+    for (i = 0; i < n-1; i++)
+	for (j = i+1; j < n; j++)
+	    if (cmp_tlb_dump_info(t+i, t+j) > 0) {
+		tlb_dump_entry_t tmp = t[i];
+		t[i] = t[j];
+		t[j] = tmp;
+	    }
+}
+
+
+static tlb_dump_entry_t itlb_dump_info[ITLB_TOTAL_ENTRIES];
+static tlb_dump_entry_t dtlb_dump_info[DTLB_TOTAL_ENTRIES];
+
+
+static inline char *way_type (int type)
+{
+	return type ? "autorefill" : "non-autorefill";
+}
+
+void print_entry (struct way_config_t *way_info,
+		  unsigned int way,
+		  unsigned int index,
+		  unsigned int virtual,
+		  unsigned int translation)
+{
+	char valid_chr;
+	unsigned int va, pa, asid, ca;
+
+	va = virtual &
+	  	~((1 << (way_info->pgsz_log2 + way_info->indicies_log2)) - 1);
+	asid = virtual & ((1 << XCHAL_MMU_ASID_BITS) - 1);
+	pa = translation & ~((1 << way_info->pgsz_log2) - 1);
+	ca = translation & ((1 << XCHAL_MMU_CA_BITS) - 1);
+	valid_chr = asid ? 'V' : 'I';
+
+	/* Compute and incorporate the effect of the index bits on the
+	 * va.  It's more useful for kernel debugging, since we always
+	 * want to know the effective va anyway. */
+
+	va += index << way_info->pgsz_log2;
+
+	printk ("\t[%d,%d] (%c) vpn 0x%.8x  ppn 0x%.8x  asid 0x%.2x  am 0x%x\n",
+		way, index, valid_chr, va, pa, asid, ca);
+}
+
+void print_itlb_entry (struct way_config_t *way_info, int way, int index)
+{
+	print_entry (way_info, way, index,
+		     read_itlb_virtual (way + (index << way_info->pgsz_log2)),
+		     read_itlb_translation (way + (index << way_info->pgsz_log2)));
+}
+
+void print_dtlb_entry (struct way_config_t *way_info, int way, int index)
+{
+	print_entry (way_info, way, index,
+		     read_dtlb_virtual (way + (index << way_info->pgsz_log2)),
+		     read_dtlb_translation (way + (index << way_info->pgsz_log2)));
+}
+
+void dump_itlb (void)
+{
+	int way, index;
+
+	printk ("\nITLB: ways = %d\n", XCHAL_ITLB_WAYS);
+
+	for (way = 0; way < XCHAL_ITLB_WAYS; way++) {
+		printk ("\nWay: %d, Entries: %d, MinPageSize: %d, Type: %s\n",
+			way, itlb[way].indicies,
+			itlb[way].pgsz_log2, way_type(itlb[way].arf));
+		for (index = 0; index < itlb[way].indicies; index++) {
+			print_itlb_entry(&itlb[way], way, index);
+		}
+	}
+}
+
+void dump_dtlb (void)
+{
+	int way, index;
+
+	printk ("\nDTLB: ways = %d\n", XCHAL_DTLB_WAYS);
+
+	for (way = 0; way < XCHAL_DTLB_WAYS; way++) {
+		printk ("\nWay: %d, Entries: %d, MinPageSize: %d, Type: %s\n",
+			way, dtlb[way].indicies,
+			dtlb[way].pgsz_log2, way_type(dtlb[way].arf));
+		for (index = 0; index < dtlb[way].indicies; index++) {
+			print_dtlb_entry(&dtlb[way], way, index);
+		}
+	}
+}
+
+void dump_tlb (tlb_dump_entry_t *tinfo, struct way_config_t *config,
+		int entries, int ways, int type, int show_invalid)
+{
+    tlb_dump_entry_t *e = tinfo;
+    int way, i;
+
+    /*  Gather all info:  */
+    for (way = 0; way < ways; way++) {
+	struct way_config_t *cfg = config + way;
+	for (i = 0; i < cfg->indicies; i++) {
+	    unsigned wayindex = way + (i << cfg->pgsz_log2);
+	    unsigned vv = (type ? read_dtlb_virtual (wayindex)
+		    		: read_itlb_virtual (wayindex));
+	    unsigned pp = (type ? read_dtlb_translation (wayindex)
+		    		: read_itlb_translation (wayindex));
+
+	    /* Compute and incorporate the effect of the index bits on the
+	     * va.  It's more useful for kernel debugging, since we always
+	     * want to know the effective va anyway. */
+
+	    e->va = (vv & ~((1 << (cfg->pgsz_log2 + cfg->indicies_log2)) - 1));
+	    e->va += (i << cfg->pgsz_log2);
+	    e->pa = (pp & ~((1 << cfg->pgsz_log2) - 1));
+	    e->asid = (vv & ((1 << XCHAL_MMU_ASID_BITS) - 1));
+	    e->ca = (pp & ((1 << XCHAL_MMU_CA_BITS) - 1));
+	    e->way = way;
+	    e->index = i;
+	    e->pgsz_log2 = cfg->pgsz_log2;
+	    e->type = type;
+	    e++;
+	}
+    }
+#if 1
+    /*  Sort by ASID and VADDR:  */
+    sort_tlb_dump_info (tinfo, entries);
+#endif
+
+    /*  Display all sorted info:  */
+    printk ("\n%cTLB dump:\n", (type ? 'D' : 'I'));
+    for (e = tinfo, i = 0; i < entries; i++, e++) {
+#if 0
+	if (e->asid == 0 && !show_invalid)
+	    continue;
+#endif
+	printk ("%c way=%d i=%d  ASID=%02X V=%08X -> P=%08X CA=%X (%d %cB)\n",
+		(e->type ? 'D' : 'I'), e->way, e->index,
+		e->asid, e->va, e->pa, e->ca,
+		(1 << (e->pgsz_log2 % 10)),
+		" kMG"[e->pgsz_log2 / 10]
+		);
+    }
+}
+
+void dump_tlbs2 (int showinv)
+{
+    dump_tlb (itlb_dump_info, itlb, ITLB_TOTAL_ENTRIES, XCHAL_ITLB_WAYS, 0, showinv);
+    dump_tlb (dtlb_dump_info, dtlb, DTLB_TOTAL_ENTRIES, XCHAL_DTLB_WAYS, 1, showinv);
+}
+
+void dump_all_tlbs (void)
+{
+    dump_tlbs2 (1);
+}
+
+void dump_valid_tlbs (void)
+{
+    dump_tlbs2 (0);
+}
+
+
+void dump_tlbs (void)
+{
+	dump_itlb();
+	dump_dtlb();
+}
+
+void dump_cache_tag(int dcache, int idx)
+{
+	int w, i, s, e;
+	unsigned long tag, index;
+	unsigned long num_lines, num_ways, cache_size, line_size;
+
+	num_ways = dcache ? XCHAL_DCACHE_WAYS : XCHAL_ICACHE_WAYS;
+	cache_size = dcache ? XCHAL_DCACHE_SIZE : XCHAL_ICACHE_SIZE;
+	line_size = dcache ? XCHAL_DCACHE_LINESIZE : XCHAL_ICACHE_LINESIZE;
+
+	num_lines = cache_size / num_ways;
+
+	s = 0; e = num_lines;
+
+	if (idx >= 0)
+		e = (s = idx * line_size) + 1;
+
+	for (i = s; i < e; i+= line_size) {
+		printk("\nline %#08x:", i);
+		for (w = 0; w < num_ways; w++) {
+			index = w * num_lines + i;
+			if (dcache)
+				__asm__ __volatile__("ldct %0, %1\n\t"
+						: "=a"(tag) : "a"(index));
+			else
+				__asm__ __volatile__("lict %0, %1\n\t"
+						: "=a"(tag) : "a"(index));
+
+			printk(" %#010lx", tag);
+		}
+	}
+	printk ("\n");
+}
+
+void dump_icache(int index)
+{
+	unsigned long data, addr;
+	int w, i;
+
+	const unsigned long num_ways = XCHAL_ICACHE_WAYS;
+	const unsigned long cache_size = XCHAL_ICACHE_SIZE;
+	const unsigned long line_size = XCHAL_ICACHE_LINESIZE;
+	const unsigned long num_lines = cache_size / num_ways / line_size;
+
+	for (w = 0; w < num_ways; w++) {
+		printk ("\nWay %d", w);
+
+		for (i = 0; i < line_size; i+= 4) {
+			addr = w * num_lines + index * line_size + i;
+			__asm__ __volatile__("licw %0, %1\n\t"
+					: "=a"(data) : "a"(addr));
+			printk(" %#010lx", data);
+		}
+	}
+	printk ("\n");
+}
+
+void dump_cache_tags(void)
+{
+	printk("Instruction cache\n");
+	dump_cache_tag(0, -1);
+	printk("Data cache\n");
+	dump_cache_tag(1, -1);
+}
+
+#endif
diff --git a/arch/xtensa/platform-iss/Makefile b/arch/xtensa/platform-iss/Makefile
new file mode 100644
index 000000000000..5b394e9620e5
--- /dev/null
+++ b/arch/xtensa/platform-iss/Makefile
@@ -0,0 +1,13 @@
+# $Id: Makefile,v 1.1.1.1 2002/08/28 16:10:14 aroll Exp $
+#
+# Makefile for the Xtensa Instruction Set Simulator (ISS)
+# "prom monitor" library routines under Linux.
+#
+# Note! Dependencies are done automagically by 'make dep', which also
+# removes any old dependencies. DON'T put your own dependencies here
+# unless it's something special (ie not a .c file).
+#
+# Note 2! The CFLAGS definitions are in the main makefile...
+
+obj-y			= io.o console.o setup.o network.o
+
diff --git a/arch/xtensa/platform-iss/console.c b/arch/xtensa/platform-iss/console.c
new file mode 100644
index 000000000000..4fbddf92a921
--- /dev/null
+++ b/arch/xtensa/platform-iss/console.c
@@ -0,0 +1,300 @@
+/*
+ * arch/xtensa/platform-iss/console.c
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001-2005 Tensilica Inc.
+ *   Authors	Christian Zankel, Joe Taylor
+ */
+
+#include <linux/module.h>
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/console.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/major.h>
+#include <linux/param.h>
+#include <linux/serial.h>
+#include <linux/serialP.h>
+#include <linux/console.h>
+
+#include <asm/uaccess.h>
+#include <asm/irq.h>
+
+#include <xtensa/simcall.h>
+
+#include <linux/tty.h>
+#include <linux/tty_flip.h>
+
+#ifdef SERIAL_INLINE
+#define _INLINE_ inline
+#endif
+
+#define SERIAL_MAX_NUM_LINES 1
+#define SERIAL_TIMER_VALUE (20 * HZ)
+
+static struct tty_driver *serial_driver;
+static struct timer_list serial_timer;
+
+static DEFINE_SPINLOCK(timer_lock);
+
+int errno;
+
+static int __simc (int a, int b, int c, int d, int e, int f)
+{
+	int ret;
+	__asm__ __volatile__ ("simcall\n"
+			"mov %0, a2\n"
+			"mov %1, a3\n" : "=a" (ret), "=a" (errno)
+			: : "a2", "a3");
+	return ret;
+}
+
+static char *serial_version = "0.1";
+static char *serial_name = "ISS serial driver";
+
+/*
+ * This routine is called whenever a serial port is opened.  It
+ * enables interrupts for a serial port, linking in its async structure into
+ * the IRQ chain.   It also performs the serial-specific
+ * initialization for the tty structure.
+ */
+
+static void rs_poll(unsigned long);
+
+static int rs_open(struct tty_struct *tty, struct file * filp)
+{
+	int line = tty->index;
+
+	if ((line < 0) || (line >= SERIAL_MAX_NUM_LINES))
+		return -ENODEV;
+
+	spin_lock(&timer_lock);
+
+	if (tty->count == 1) {
+		init_timer(&serial_timer);
+		serial_timer.data = (unsigned long) tty;
+		serial_timer.function = rs_poll;
+		mod_timer(&serial_timer, jiffies + SERIAL_TIMER_VALUE);
+	}
+	spin_unlock(&timer_lock);
+
+	return 0;
+}
+
+
+/*
+ * ------------------------------------------------------------
+ * iss_serial_close()
+ *
+ * This routine is called when the serial port gets closed.  First, we
+ * wait for the last remaining data to be sent.  Then, we unlink its
+ * async structure from the interrupt chain if necessary, and we free
+ * that IRQ if nothing is left in the chain.
+ * ------------------------------------------------------------
+ */
+static void rs_close(struct tty_struct *tty, struct file * filp)
+{
+	spin_lock(&timer_lock);
+	if (tty->count == 1)
+		del_timer_sync(&serial_timer);
+	spin_unlock(&timer_lock);
+}
+
+
+static int rs_write(struct tty_struct * tty,
+		    const unsigned char *buf, int count)
+{
+	/* see drivers/char/serialX.c to reference original version */
+
+	__simc (SYS_write, 1, (unsigned long)buf, count, 0, 0);
+	return count;
+}
+
+static void rs_poll(unsigned long priv)
+{
+	struct tty_struct* tty = (struct tty_struct*) priv;
+
+	struct timeval tv = { .tv_sec = 0, .tv_usec = 0 };
+	int i = 0;
+	unsigned char c;
+
+	spin_lock(&timer_lock);
+
+	while (__simc(SYS_select_one, 0, XTISS_SELECT_ONE_READ, (int)&tv,0,0)){
+		__simc (SYS_read, 0, (unsigned long)&c, 1, 0, 0);
+		tty->flip.count++;
+		*tty->flip.char_buf_ptr++ = c;
+		*tty->flip.flag_buf_ptr++ = TTY_NORMAL;
+		i++;
+	}
+
+	if (i)
+		tty_flip_buffer_push(tty);
+
+
+	mod_timer(&serial_timer, jiffies + SERIAL_TIMER_VALUE);
+	spin_unlock(&timer_lock);
+}
+
+
+static void rs_put_char(struct tty_struct *tty, unsigned char ch)
+{
+	char buf[2];
+
+	if (!tty)
+		return;
+
+	buf[0] = ch;
+	buf[1] = '\0';		/* Is this NULL necessary? */
+	__simc (SYS_write, 1, (unsigned long) buf, 1, 0, 0);
+}
+
+static void rs_flush_chars(struct tty_struct *tty)
+{
+}
+
+static int rs_write_room(struct tty_struct *tty)
+{
+	/* Let's say iss can always accept 2K characters.. */
+	return 2 * 1024;
+}
+
+static int rs_chars_in_buffer(struct tty_struct *tty)
+{
+	/* the iss doesn't buffer characters */
+	return 0;
+}
+
+static void rs_hangup(struct tty_struct *tty)
+{
+	/* Stub, once again.. */
+}
+
+static void rs_wait_until_sent(struct tty_struct *tty, int timeout)
+{
+	/* Stub, once again.. */
+}
+
+static int rs_read_proc(char *page, char **start, off_t off, int count,
+			int *eof, void *data)
+{
+	int len = 0;
+	off_t begin = 0;
+
+	len += sprintf(page, "serinfo:1.0 driver:%s\n", serial_version);
+	*eof = 1;
+
+	if (off >= len + begin)
+		return 0;
+
+	*start = page + (off - begin);
+	return ((count < begin + len - off) ? count : begin + len - off);
+}
+
+
+static struct tty_operations serial_ops = {
+	.open = rs_open,
+	.close = rs_close,
+	.write = rs_write,
+	.put_char = rs_put_char,
+	.flush_chars = rs_flush_chars,
+	.write_room = rs_write_room,
+	.chars_in_buffer = rs_chars_in_buffer,
+	.hangup = rs_hangup,
+	.wait_until_sent = rs_wait_until_sent,
+	.read_proc = rs_read_proc
+};
+
+int __init rs_init(void)
+{
+	serial_driver = alloc_tty_driver(1);
+
+	printk ("%s %s\n", serial_name, serial_version);
+
+	/* Initialize the tty_driver structure */
+
+	serial_driver->owner = THIS_MODULE;
+	serial_driver->driver_name = "iss_serial";
+	serial_driver->name = "ttyS";
+	serial_driver->major = TTY_MAJOR;
+	serial_driver->minor_start = 64;
+	serial_driver->type = TTY_DRIVER_TYPE_SERIAL;
+	serial_driver->subtype = SERIAL_TYPE_NORMAL;
+	serial_driver->init_termios = tty_std_termios;
+	serial_driver->init_termios.c_cflag =
+		B9600 | CS8 | CREAD | HUPCL | CLOCAL;
+	serial_driver->flags = TTY_DRIVER_REAL_RAW;
+
+	tty_set_operations(serial_driver, &serial_ops);
+
+	if (tty_register_driver(serial_driver))
+		panic("Couldn't register serial driver\n");
+	return 0;
+}
+
+
+static __exit void rs_exit(void)
+{
+	int error;
+
+	if ((error = tty_unregister_driver(serial_driver)))
+		printk("ISS_SERIAL: failed to unregister serial driver (%d)\n",
+		       error);
+	put_tty_driver(serial_driver);
+}
+
+
+/* We use `late_initcall' instead of just `__initcall' as a workaround for
+ * the fact that (1) simcons_tty_init can't be called before tty_init,
+ * (2) tty_init is called via `module_init', (3) if statically linked,
+ * module_init == device_init, and (4) there's no ordering of init lists.
+ * We can do this easily because simcons is always statically linked, but
+ * other tty drivers that depend on tty_init and which must use
+ * `module_init' to declare their init routines are likely to be broken.
+ */
+
+late_initcall(rs_init);
+
+
+#ifdef CONFIG_SERIAL_CONSOLE
+
+static void iss_console_write(struct console *co, const char *s, unsigned count)
+{
+	int len = strlen(s);
+
+	if (s != 0 && *s != 0)
+		__simc (SYS_write, 1, (unsigned long)s,
+			count < len ? count : len,0,0);
+}
+
+static struct tty_driver* iss_console_device(struct console *c, int *index)
+{
+	*index = c->index;
+	return serial_driver;
+}
+
+
+static struct console sercons = {
+	.name = "ttyS",
+	.write = iss_console_write,
+	.device = iss_console_device,
+	.flags = CON_PRINTBUFFER,
+	.index = -1
+};
+
+static int __init iss_console_init(void)
+{
+	register_console(&sercons);
+	return 0;
+}
+
+console_initcall(iss_console_init);
+
+#endif /* CONFIG_SERIAL_CONSOLE */
+
diff --git a/arch/xtensa/platform-iss/io.c b/arch/xtensa/platform-iss/io.c
new file mode 100644
index 000000000000..5b161a5cb65f
--- /dev/null
+++ b/arch/xtensa/platform-iss/io.c
@@ -0,0 +1,32 @@
+/* This file isn't really needed right now. */
+
+#if 0
+
+#include <asm/io.h>
+#include <xtensa/simcall.h>
+
+extern int __simc ();
+
+
+char iss_serial_getc()
+{
+  char c;
+  __simc( SYS_read, 0, &c, 1 );
+  return c;
+}
+
+void iss_serial_putc( char c )
+{
+  __simc( SYS_write, 1, &c, 1 );
+}
+
+void iss_serial_puts( char *s )
+{
+  if( s != 0 && *s != 0 )
+    __simc( SYS_write, 1, s, strlen(s) );
+}
+
+/*#error Need I/O ports to specific hardware!*/
+
+#endif
+
diff --git a/arch/xtensa/platform-iss/network.c b/arch/xtensa/platform-iss/network.c
new file mode 100644
index 000000000000..498d7dced1f4
--- /dev/null
+++ b/arch/xtensa/platform-iss/network.c
@@ -0,0 +1,855 @@
+/*
+ *
+ * arch/xtensa/platform-iss/network.c
+ *
+ * Platform specific initialization.
+ *
+ * Authors: Chris Zankel <chris@zankel.net>
+ * Based on work form the UML team.
+ *
+ * Copyright 2005 Tensilica Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+
+#include <linux/config.h>
+#include <linux/list.h>
+#include <linux/irq.h>
+#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <linux/timer.h>
+#include <linux/if_ether.h>
+#include <linux/inetdevice.h>
+#include <linux/init.h>
+#include <linux/if_tun.h>
+#include <linux/etherdevice.h>
+#include <linux/interrupt.h>
+#include <linux/ioctl.h>
+#include <linux/bootmem.h>
+#include <linux/ethtool.h>
+#include <linux/rtnetlink.h>
+#include <linux/timer.h>
+
+#include <xtensa/simcall.h>
+
+#define DRIVER_NAME "iss-netdev"
+#define ETH_MAX_PACKET 1500
+#define ETH_HEADER_OTHER 14
+#define ISS_NET_TIMER_VALUE (2 * HZ)
+
+
+static DEFINE_SPINLOCK(opened_lock);
+static LIST_HEAD(opened);
+
+static DEFINE_SPINLOCK(devices_lock);
+static LIST_HEAD(devices);
+
+/* ------------------------------------------------------------------------- */
+
+/* We currently only support the TUNTAP transport protocol. */
+
+#define TRANSPORT_TUNTAP_NAME "tuntap"
+#define TRANSPORT_TUNTAP_MTU ETH_MAX_PACKET
+
+struct tuntap_info {
+	char dev_name[IFNAMSIZ];
+	int fixed_config;
+	unsigned char gw[ETH_ALEN];
+	int fd;
+};
+
+/* ------------------------------------------------------------------------- */
+
+
+/* This structure contains out private information for the driver. */
+
+struct iss_net_private {
+
+	struct list_head device_list;
+	struct list_head opened_list;
+
+	spinlock_t lock;
+	struct net_device *dev;
+	struct platform_device pdev;
+	struct timer_list tl;
+	struct net_device_stats stats;
+
+	struct timer_list timer;
+	unsigned int timer_val;
+
+	int index;
+	int mtu;
+
+	unsigned char mac[ETH_ALEN];
+	int have_mac;
+
+	struct {
+		union {
+			struct tuntap_info tuntap;
+		} info;
+
+		int (*open)(struct iss_net_private *lp);
+		void (*close)(struct iss_net_private *lp);
+		int (*read)(struct iss_net_private *lp, struct sk_buff **skb);
+		int (*write)(struct iss_net_private *lp, struct sk_buff **skb);
+		unsigned short (*protocol)(struct sk_buff *skb);
+		int (*poll)(struct iss_net_private *lp);
+	} tp;
+
+};
+
+/* ======================= ISS SIMCALL INTERFACE =========================== */
+
+/* Note: __simc must _not_ be declared inline! */
+
+static int errno;
+
+static int __simc (int a, int b, int c, int d, int e, int f)
+{
+	int ret;
+	__asm__ __volatile__ ("simcall\n"
+	    		      "mov %0, a2\n"
+			      "mov %1, a3\n" : "=a" (ret), "=a" (errno)
+			      : : "a2", "a3");
+	return ret;
+}
+
+static int inline simc_open(char *file, int flags, int mode)
+{
+	return __simc(SYS_open, (int) file, flags, mode, 0, 0);
+}
+
+static int inline simc_close(int fd)
+{
+	return __simc(SYS_close, fd, 0, 0, 0, 0);
+}
+
+static int inline simc_ioctl(int fd, int request, void *arg)
+{
+	return __simc(SYS_ioctl, fd, request, (int) arg, 0, 0);
+}
+
+static int inline simc_read(int fd, void *buf, size_t count)
+{
+	return __simc(SYS_read, fd, (int) buf, count, 0, 0);
+}
+
+static int inline simc_write(int fd, void *buf, size_t count)
+{
+	return __simc(SYS_write, fd, (int) buf, count, 0, 0);
+}
+
+static int inline simc_poll(int fd)
+{
+	struct timeval tv = { .tv_sec = 0, .tv_usec = 0 };
+
+	return __simc(SYS_select_one, fd, XTISS_SELECT_ONE_READ, (int)&tv,0,0);
+}
+
+/* ================================ HELPERS ================================ */
+
+
+static char *split_if_spec(char *str, ...)
+{
+	char **arg, *end;
+	va_list ap;
+
+	va_start(ap, str);
+	while ((arg = va_arg(ap, char**)) != NULL) {
+		if (*str == '\0')
+			return NULL;
+		end = strchr(str, ',');
+		if (end != str)
+			*arg = str;
+		if (end == NULL)
+			return NULL;
+		*end ++ = '\0';
+		str = end;
+	}
+	va_end(ap);
+	return str;
+}
+
+
+#if 0
+/* Adjust SKB. */
+
+struct sk_buff *ether_adjust_skb(struct sk_buff *skb, int extra)
+{
+	if ((skb != NULL) && (skb_tailroom(skb) < extra)) {
+		struct sk_buff *skb2;
+
+		skb2 = skb_copy_expand(skb, 0, extra, GFP_ATOMIC);
+		dev_kfree_skb(skb);
+		skb = skb2;
+	}
+	if (skb != NULL)
+		skb_put(skb, extra);
+
+	return skb;
+}
+#endif
+
+/* Return the IP address as a string for a given device. */
+
+static void dev_ip_addr(void *d, char *buf, char *bin_buf)
+{
+	struct net_device *dev = d;
+	struct in_device *ip = dev->ip_ptr;
+	struct in_ifaddr *in;
+	u32 addr;
+
+	if ((ip == NULL) || ((in = ip->ifa_list) == NULL)) {
+		printk(KERN_WARNING "Device not assigned an IP address!\n");
+		return;
+	}
+
+	addr = in->ifa_address;
+	sprintf(buf, "%d.%d.%d.%d", addr & 0xff, (addr >> 8) & 0xff,
+		(addr >> 16) & 0xff, addr >> 24);
+
+	if (bin_buf) {
+		bin_buf[0] = addr & 0xff;
+		bin_buf[1] = (addr >> 8) & 0xff;
+		bin_buf[2] = (addr >> 16) & 0xff;
+		bin_buf[3] = addr >> 24;
+	}
+}
+
+/* Set Ethernet address of the specified device. */
+
+static void inline set_ether_mac(void *d, unsigned char *addr)
+{
+	struct net_device *dev = d;
+	memcpy(dev->dev_addr, addr, ETH_ALEN);
+}
+
+
+/* ======================= TUNTAP TRANSPORT INTERFACE ====================== */
+
+static int tuntap_open(struct iss_net_private *lp)
+{
+	struct ifreq ifr;
+	char *dev_name = lp->tp.info.tuntap.dev_name;
+	int err = -EINVAL;
+	int fd;
+
+	/* We currently only support a fixed configuration. */
+
+	if (!lp->tp.info.tuntap.fixed_config)
+		return -EINVAL;
+
+	if ((fd = simc_open("/dev/net/tun", 02, 0)) < 0) {	/* O_RDWR */
+		printk("Failed to open /dev/net/tun, returned %d "
+		       "(errno = %d)\n", fd, errno);
+		return fd;
+	}
+
+	memset(&ifr, 0, sizeof ifr);
+	ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
+	strlcpy(ifr.ifr_name, dev_name, sizeof ifr.ifr_name - 1);
+
+	if ((err = simc_ioctl(fd, TUNSETIFF, (void*) &ifr)) < 0) {
+		printk("Failed to set interface, returned %d "
+		       "(errno = %d)\n", err, errno);
+		simc_close(fd);
+		return err;
+	}
+
+	lp->tp.info.tuntap.fd = fd;
+	return err;
+}
+
+static void tuntap_close(struct iss_net_private *lp)
+{
+#if 0
+	if (lp->tp.info.tuntap.fixed_config)
+		iter_addresses(lp->tp.info.tuntap.dev, close_addr, lp->host.dev_name);
+#endif
+	simc_close(lp->tp.info.tuntap.fd);
+	lp->tp.info.tuntap.fd = -1;
+}
+
+static int tuntap_read (struct iss_net_private *lp, struct sk_buff **skb)
+{
+#if 0
+	*skb = ether_adjust_skb(*skb, ETH_HEADER_OTHER);
+	if (*skb == NULL)
+		return -ENOMEM;
+#endif
+
+	return simc_read(lp->tp.info.tuntap.fd,
+			(*skb)->data, (*skb)->dev->mtu + ETH_HEADER_OTHER);
+}
+
+static int tuntap_write (struct iss_net_private *lp, struct sk_buff **skb)
+{
+	return simc_write(lp->tp.info.tuntap.fd, (*skb)->data, (*skb)->len);
+}
+
+unsigned short tuntap_protocol(struct sk_buff *skb)
+{
+	return eth_type_trans(skb, skb->dev);
+}
+
+static int tuntap_poll(struct iss_net_private *lp)
+{
+	return simc_poll(lp->tp.info.tuntap.fd);
+}
+
+/*
+ * Currently only a device name is supported.
+ * ethX=tuntap[,[mac address][,[device name]]]
+ */
+
+static int tuntap_probe(struct iss_net_private *lp, int index, char *init)
+{
+	const int len = strlen(TRANSPORT_TUNTAP_NAME);
+	char *dev_name = NULL, *mac_str = NULL, *rem = NULL;
+
+	/* Transport should be 'tuntap': ethX=tuntap,mac,dev_name */
+
+	if (strncmp(init, TRANSPORT_TUNTAP_NAME, len))
+		return 0;
+
+	if (*(init += strlen(TRANSPORT_TUNTAP_NAME)) == ',') {
+		if ((rem=split_if_spec(init+1, &mac_str, &dev_name)) != NULL) {
+			printk("Extra garbage on specification : '%s'\n", rem);
+			return 0;
+		}
+	} else if (*init != '\0') {
+		printk("Invalid argument: %s. Skipping device!\n", init);
+		return 0;
+	}
+
+	if (dev_name) {
+		strncpy(lp->tp.info.tuntap.dev_name, dev_name,
+			 sizeof lp->tp.info.tuntap.dev_name);
+		lp->tp.info.tuntap.fixed_config = 1;
+	} else
+		strcpy(lp->tp.info.tuntap.dev_name, TRANSPORT_TUNTAP_NAME);
+
+
+#if 0
+	if (setup_etheraddr(mac_str, lp->mac))
+		lp->have_mac = 1;
+#endif
+	lp->mtu = TRANSPORT_TUNTAP_MTU;
+
+	//lp->info.tuntap.gate_addr = gate_addr;
+
+	lp->tp.info.tuntap.fd = -1;
+
+	lp->tp.open = tuntap_open;
+	lp->tp.close = tuntap_close;
+	lp->tp.read = tuntap_read;
+	lp->tp.write = tuntap_write;
+	lp->tp.protocol = tuntap_protocol;
+	lp->tp.poll = tuntap_poll;
+
+	printk("TUN/TAP backend - ");
+#if 0
+	if (lp->host.gate_addr != NULL)
+		printk("IP = %s", lp->host.gate_addr);
+#endif
+	printk("\n");
+
+	return 1;
+}
+
+/* ================================ ISS NET ================================ */
+
+static int iss_net_rx(struct net_device *dev)
+{
+	struct iss_net_private *lp = dev->priv;
+	int pkt_len;
+	struct sk_buff *skb;
+
+	/* Check if there is any new data. */
+
+	if (lp->tp.poll(lp) == 0)
+		return 0;
+
+	/* Try to allocate memory, if it fails, try again next round. */
+
+	if ((skb = dev_alloc_skb(dev->mtu + 2 + ETH_HEADER_OTHER)) == NULL) {
+		lp->stats.rx_dropped++;
+		return 0;
+	}
+
+	skb_reserve(skb, 2);
+
+	/* Setup skb */
+
+	skb->dev = dev;
+	skb->mac.raw = skb->data;
+	pkt_len = lp->tp.read(lp, &skb);
+	skb_put(skb, pkt_len);
+
+	if (pkt_len > 0) {
+		skb_trim(skb, pkt_len);
+		skb->protocol = lp->tp.protocol(skb);
+	//	netif_rx(skb);
+		netif_rx_ni(skb);
+
+		lp->stats.rx_bytes += skb->len;
+		lp->stats.rx_packets++;
+		return pkt_len;
+	}
+	kfree_skb(skb);
+	return pkt_len;
+}
+
+static int iss_net_poll(void)
+{
+	struct list_head *ele;
+	int err, ret = 0;
+
+	spin_lock(&opened_lock);
+
+	list_for_each(ele, &opened) {
+		struct iss_net_private *lp;
+
+		lp = list_entry(ele, struct iss_net_private, opened_list);
+
+		if (!netif_running(lp->dev))
+			break;
+
+		spin_lock(&lp->lock);
+
+		while ((err = iss_net_rx(lp->dev)) > 0)
+			ret++;
+
+		spin_unlock(&lp->lock);
+
+		if (err < 0) {
+			printk(KERN_ERR "Device '%s' read returned %d, "
+			       "shutting it down\n", lp->dev->name, err);
+			dev_close(lp->dev);
+		} else {
+			// FIXME reactivate_fd(lp->fd, ISS_ETH_IRQ);
+		}
+	}
+
+	spin_unlock(&opened_lock);
+	return ret;
+}
+
+
+static void iss_net_timer(unsigned long priv)
+{
+	struct iss_net_private* lp = (struct iss_net_private*) priv;
+
+	spin_lock(&lp->lock);
+
+	iss_net_poll();
+
+	mod_timer(&lp->timer, jiffies + lp->timer_val);
+
+	spin_unlock(&lp->lock);
+}
+
+
+static int iss_net_open(struct net_device *dev)
+{
+	struct iss_net_private *lp = dev->priv;
+	char addr[sizeof "255.255.255.255\0"];
+	int err;
+
+	spin_lock(&lp->lock);
+
+	if ((err = lp->tp.open(lp)) < 0)
+		goto out;
+
+	if (!lp->have_mac) {
+		dev_ip_addr(dev, addr, &lp->mac[2]);
+		set_ether_mac(dev, lp->mac);
+	}
+
+	netif_start_queue(dev);
+
+	/* clear buffer - it can happen that the host side of the interface
+	 * is full when we gethere. In this case, new data is never queued,
+	 * SIGIOs never arrive, and the net never works.
+	 */
+	while ((err = iss_net_rx(dev)) > 0)
+		;
+
+	spin_lock(&opened_lock);
+	list_add(&lp->opened_list, &opened);
+	spin_unlock(&opened_lock);
+
+	init_timer(&lp->timer);
+	lp->timer_val = ISS_NET_TIMER_VALUE;
+	lp->timer.data = (unsigned long) lp;
+	lp->timer.function = iss_net_timer;
+	mod_timer(&lp->timer, jiffies + lp->timer_val);
+
+out:
+	spin_unlock(&lp->lock);
+	return err;
+}
+
+static int iss_net_close(struct net_device *dev)
+{
+	struct iss_net_private *lp = dev->priv;
+printk("iss_net_close!\n");
+	netif_stop_queue(dev);
+	spin_lock(&lp->lock);
+
+	spin_lock(&opened_lock);
+	list_del(&opened);
+	spin_unlock(&opened_lock);
+
+	del_timer_sync(&lp->timer);
+
+	lp->tp.close(lp);
+
+	spin_unlock(&lp->lock);
+	return 0;
+}
+
+static int iss_net_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct iss_net_private *lp = dev->priv;
+	unsigned long flags;
+	int len;
+
+	netif_stop_queue(dev);
+	spin_lock_irqsave(&lp->lock, flags);
+
+	len = lp->tp.write(lp, &skb);
+
+	if (len == skb->len) {
+		lp->stats.tx_packets++;
+		lp->stats.tx_bytes += skb->len;
+		dev->trans_start = jiffies;
+		netif_start_queue(dev);
+
+		/* this is normally done in the interrupt when tx finishes */
+		netif_wake_queue(dev);
+
+	} else if (len == 0) {
+		netif_start_queue(dev);
+		lp->stats.tx_dropped++;
+
+	} else {
+		netif_start_queue(dev);
+		printk(KERN_ERR "iss_net_start_xmit: failed(%d)\n", len);
+	}
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	dev_kfree_skb(skb);
+	return 0;
+}
+
+
+static struct net_device_stats *iss_net_get_stats(struct net_device *dev)
+{
+	struct iss_net_private *lp = dev->priv;
+	return &lp->stats;
+}
+
+static void iss_net_set_multicast_list(struct net_device *dev)
+{
+#if 0
+	if (dev->flags & IFF_PROMISC)
+		return;
+	else if (dev->mc_count)
+		dev->flags |= IFF_ALLMULTI;
+	else
+		dev->flags &= ~IFF_ALLMULTI;
+#endif
+}
+
+static void iss_net_tx_timeout(struct net_device *dev)
+{
+#if 0
+	dev->trans_start = jiffies;
+	netif_wake_queue(dev);
+#endif
+}
+
+static int iss_net_set_mac(struct net_device *dev, void *addr)
+{
+#if 0
+	struct iss_net_private *lp = dev->priv;
+	struct sockaddr *hwaddr = addr;
+
+	spin_lock(&lp->lock);
+	memcpy(dev->dev_addr, hwaddr->sa_data, ETH_ALEN);
+	spin_unlock(&lp->lock);
+#endif
+
+	return 0;
+}
+
+static int iss_net_change_mtu(struct net_device *dev, int new_mtu)
+{
+#if 0
+	struct iss_net_private *lp = dev->priv;
+	int err = 0;
+
+	spin_lock(&lp->lock);
+
+	// FIXME not needed new_mtu = transport_set_mtu(new_mtu, &lp->user);
+
+	if (new_mtu < 0)
+		err = new_mtu;
+	else
+		dev->mtu = new_mtu;
+
+	spin_unlock(&lp->lock);
+	return err;
+#endif
+	return -EINVAL;
+}
+
+static int iss_net_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
+{
+#if 0
+	static const struct ethtool_drvinfo info = {
+		.cmd     = ETHTOOL_GDRVINFO,
+		.driver  = DRIVER_NAME,
+		.version = "42",
+	};
+	void *useraddr;
+	u32 ethcmd;
+
+	switch (cmd) {
+	case SIOCETHTOOL:
+		useraddr = ifr->ifr_data;
+		if (copy_from_user(&ethcmd, useraddr, sizeof(ethcmd)))
+			return -EFAULT;
+
+		switch (ethcmd) {
+			case ETHTOOL_GDRVINFO:
+				if (copy_to_user(useraddr, &info, sizeof(info)))
+					return -EFAULT;
+				return 0;
+			default:
+				return -EOPNOTSUPP;
+		}
+	default:
+		return -EINVAL;
+	}
+#endif
+	return -EINVAL;
+}
+
+void iss_net_user_timer_expire(unsigned long _conn)
+{
+}
+
+
+static struct device_driver iss_net_driver = {
+	.name  = DRIVER_NAME,
+	.bus   = &platform_bus_type,
+};
+
+static int driver_registered;
+
+static int iss_net_configure(int index, char *init)
+{
+	struct net_device *dev;
+	struct iss_net_private *lp;
+	int err;
+
+	if ((dev = alloc_etherdev(sizeof *lp)) == NULL) {
+		printk(KERN_ERR "eth_configure: failed to allocate device\n");
+		return 1;
+	}
+
+	/* Initialize private element. */
+
+	lp = dev->priv;
+	*lp = ((struct iss_net_private) {
+		.device_list		= LIST_HEAD_INIT(lp->device_list),
+		.opened_list		= LIST_HEAD_INIT(lp->opened_list),
+		.lock			= SPIN_LOCK_UNLOCKED,
+		.dev			= dev,
+		.index			= index,
+		//.fd                   = -1,
+		.mac			= { 0xfe, 0xfd, 0x0, 0x0, 0x0, 0x0 },
+		.have_mac		= 0,
+		});
+
+	/*
+	 * Try all transport protocols.
+	 * Note: more protocols can be added by adding '&& !X_init(lp, eth)'.
+	 */
+
+	if (!tuntap_probe(lp, index, init)) {
+		printk("Invalid arguments. Skipping device!\n");
+		goto errout;
+	}
+
+	printk(KERN_INFO "Netdevice %d ", index);
+	if (lp->have_mac)
+		printk("(%02x:%02x:%02x:%02x:%02x:%02x) ",
+				lp->mac[0], lp->mac[1],
+				lp->mac[2], lp->mac[3],
+				lp->mac[4], lp->mac[5]);
+	printk(": ");
+
+	/* sysfs register */
+
+	if (!driver_registered) {
+		driver_register(&iss_net_driver);
+		driver_registered = 1;
+	}
+
+	spin_lock(&devices_lock);
+	list_add(&lp->device_list, &devices);
+	spin_unlock(&devices_lock);
+
+	lp->pdev.id = index;
+	lp->pdev.name = DRIVER_NAME;
+	platform_device_register(&lp->pdev);
+	SET_NETDEV_DEV(dev,&lp->pdev.dev);
+
+	/*
+	 * If this name ends up conflicting with an existing registered
+	 * netdevice, that is OK, register_netdev{,ice}() will notice this
+	 * and fail.
+	 */
+	snprintf(dev->name, sizeof dev->name, "eth%d", index);
+
+	dev->mtu = lp->mtu;
+	dev->open = iss_net_open;
+	dev->hard_start_xmit = iss_net_start_xmit;
+	dev->stop = iss_net_close;
+	dev->get_stats = iss_net_get_stats;
+	dev->set_multicast_list = iss_net_set_multicast_list;
+	dev->tx_timeout = iss_net_tx_timeout;
+	dev->set_mac_address = iss_net_set_mac;
+	dev->change_mtu = iss_net_change_mtu;
+	dev->do_ioctl = iss_net_ioctl;
+	dev->watchdog_timeo = (HZ >> 1);
+	dev->irq = -1;
+
+	rtnl_lock();
+	err = register_netdevice(dev);
+	rtnl_unlock();
+
+	if (err) {
+		printk("Error registering net device!\n");
+		/* XXX: should we call ->remove() here? */
+		free_netdev(dev);
+		return 1;
+	}
+
+	init_timer(&lp->tl);
+	lp->tl.function = iss_net_user_timer_expire;
+
+#if 0
+	if (lp->have_mac)
+		set_ether_mac(dev, lp->mac);
+#endif
+	return 0;
+
+errout:
+	// FIXME: unregister; free, etc..
+	return -EIO;
+
+}
+
+/* ------------------------------------------------------------------------- */
+
+/* Filled in during early boot */
+
+struct list_head eth_cmd_line = LIST_HEAD_INIT(eth_cmd_line);
+
+struct iss_net_init {
+	struct list_head list;
+	char *init;		/* init string */
+	int index;
+};
+
+/*
+ * Parse the command line and look for 'ethX=...' fields, and register all
+ * those fields. They will be later initialized in iss_net_init.
+ */
+
+#define ERR KERN_ERR "iss_net_setup: "
+
+static int iss_net_setup(char *str)
+{
+	struct iss_net_private *device = NULL;
+	struct iss_net_init *new;
+	struct list_head *ele;
+	char *end;
+	int n;
+
+	n = simple_strtoul(str, &end, 0);
+	if (end == str) {
+		printk(ERR "Failed to parse '%s'\n", str);
+		return 1;
+	}
+	if (n < 0) {
+		printk(ERR "Device %d is negative\n", n);
+		return 1;
+	}
+	if (*(str = end) != '=') {
+		printk(ERR "Expected '=' after device number\n");
+		return 1;
+	}
+
+	spin_lock(&devices_lock);
+
+	list_for_each(ele, &devices) {
+		device = list_entry(ele, struct iss_net_private, device_list);
+		if (device->index == n)
+			break;
+	}
+
+	spin_unlock(&devices_lock);
+
+	if (device && device->index == n) {
+		printk(ERR "Device %d already configured\n", n);
+		return 1;
+	}
+
+	if ((new = alloc_bootmem(sizeof new)) == NULL) {
+		printk("Alloc_bootmem failed\n");
+		return 1;
+	}
+
+	INIT_LIST_HEAD(&new->list);
+	new->index = n;
+	new->init = str + 1;
+
+	list_add_tail(&new->list, &eth_cmd_line);
+	return 1;
+}
+
+#undef ERR
+
+__setup("eth", iss_net_setup);
+
+/*
+ * Initialize all ISS Ethernet devices previously registered in iss_net_setup.
+ */
+
+static int iss_net_init(void)
+{
+	struct list_head *ele, *next;
+
+	/* Walk through all Ethernet devices specified in the command line. */
+
+	list_for_each_safe(ele, next, &eth_cmd_line) {
+		struct iss_net_init *eth;
+		eth = list_entry(ele, struct iss_net_init, list);
+		iss_net_configure(eth->index, eth->init);
+	}
+
+	return 1;
+}
+
+module_init(iss_net_init);
+
diff --git a/arch/xtensa/platform-iss/setup.c b/arch/xtensa/platform-iss/setup.c
new file mode 100644
index 000000000000..2e6dcbf0cc04
--- /dev/null
+++ b/arch/xtensa/platform-iss/setup.c
@@ -0,0 +1,112 @@
+/*
+ *
+ * arch/xtensa/platform-iss/setup.c
+ *
+ * Platform specific initialization.
+ *
+ * Authors: Chris Zankel <chris@zankel.net>
+ *          Joe Taylor <joe@tensilica.com>
+ *
+ * Copyright 2001 - 2005 Tensilica Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#include <linux/config.h>
+#include <linux/stddef.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/errno.h>
+#include <linux/reboot.h>
+#include <linux/pci.h>
+#include <linux/kdev_t.h>
+#include <linux/types.h>
+#include <linux/major.h>
+#include <linux/blkdev.h>
+#include <linux/console.h>
+#include <linux/delay.h>
+#include <linux/stringify.h>
+#include <linux/notifier.h>
+
+#include <asm/platform.h>
+#include <asm/bootparam.h>
+
+
+void __init platform_init(bp_tag_t* bootparam)
+{
+
+}
+
+void platform_halt(void)
+{
+	printk (" ** Called platform_halt(), looping forever! **\n");
+	while (1);
+}
+
+void platform_power_off(void)
+{
+	printk (" ** Called platform_power_off(), looping forever! **\n");
+	while (1);
+}
+void platform_restart(void)
+{
+	/* Flush and reset the mmu, simulate a processor reset, and
+	 * jump to the reset vector. */
+
+	__asm__ __volatile__("movi	a2, 15\n\t"
+			     "wsr	a2, " __stringify(ICOUNTLEVEL) "\n\t"
+			     "movi	a2, 0\n\t"
+			     "wsr	a2, " __stringify(ICOUNT) "\n\t"
+			     "wsr	a2, " __stringify(IBREAKENABLE) "\n\t"
+			     "wsr	a2, " __stringify(LCOUNT) "\n\t"
+			     "movi	a2, 0x1f\n\t"
+			     "wsr	a2, " __stringify(PS) "\n\t"
+			     "isync\n\t"
+			     "jx	%0\n\t"
+			     :
+			     : "a" (XCHAL_RESET_VECTOR_VADDR)
+			     : "a2");
+
+	/* control never gets here */
+}
+
+extern void iss_net_poll(void);
+
+const char twirl[]="|/-\\|/-\\";
+
+void platform_heartbeat(void)
+{
+#if 0
+	static int i = 0, j = 0;
+
+	if (--i < 0) {
+		i = 99;
+		printk("\r%c\r", twirl[j++]);
+		if (j == 8)
+			j = 0;
+	}
+#endif
+}
+
+
+
+static int
+iss_panic_event(struct notifier_block *this, unsigned long event, void *ptr)
+{
+	__asm__ __volatile__("movi a2, -1; simcall\n");
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block iss_panic_block = {
+	iss_panic_event,
+	NULL,
+	0
+};
+
+void __init platform_setup(char **p_cmdline)
+{
+	notifier_chain_register(&panic_notifier_list, &iss_panic_block);
+}