linux-toradex.git/arch, branch v3.2.27 Linux kernel for Apalis and Colibri modules m68k: Make sys_atomic_cmpxchg_32 work on classic m68k 2012-08-09T23:25:13+00:00 Andreas Schwab schwab@linux-m68k.org 2012-07-27T22:20:34+00:00 56c5dc342e25deddf6fdfbed82a1b9d9bc118113 commit 9e2760d18b3cf179534bbc27692c84879c61b97c upstream. User space access must always go through uaccess accessors, since on classic m68k user space and kernel space are completely separate. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> Tested-by: Thorsten Glaser <tg@debian.org> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 9e2760d18b3cf179534bbc27692c84879c61b97c upstream.

User space access must always go through uaccess accessors, since on
classic m68k user space and kernel space are completely separate.

Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
Tested-by: Thorsten Glaser <tg@debian.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ARM: Fix undefined instruction exception handling 2012-08-09T23:25:06+00:00 Russell King rmk+kernel@arm.linux.org.uk 2012-07-30T18:42:10+00:00 83358b9b1d3438293794224b14787a806031cf98 commit 15ac49b65024f55c4371a53214879a9c77c4fbf9 upstream. While trying to get a v3.5 kernel booted on the cubox, I noticed that VFP does not work correctly with VFP bounce handling. This is because of the confusion over 16-bit vs 32-bit instructions, and where PC is supposed to point to. The rule is that FP handlers are entered with regs->ARM_pc pointing at the _next_ instruction to be executed. However, if the exception is not handled, regs->ARM_pc points at the faulting instruction. This is easy for ARM mode, because we know that the next instruction and previous instructions are separated by four bytes. This is not true of Thumb2 though. Since all FP instructions are 32-bit in Thumb2, it makes things easy. We just need to select the appropriate adjustment. Do this by moving the adjustment out of do_undefinstr() into the assembly code, as only the assembly code knows whether it's dealing with a 32-bit or 16-bit instruction. Acked-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 15ac49b65024f55c4371a53214879a9c77c4fbf9 upstream.

While trying to get a v3.5 kernel booted on the cubox, I noticed that
VFP does not work correctly with VFP bounce handling.  This is because
of the confusion over 16-bit vs 32-bit instructions, and where PC is
supposed to point to.

The rule is that FP handlers are entered with regs->ARM_pc pointing at
the _next_ instruction to be executed.  However, if the exception is
not handled, regs->ARM_pc points at the faulting instruction.

This is easy for ARM mode, because we know that the next instruction and
previous instructions are separated by four bytes.  This is not true of
Thumb2 though.

Since all FP instructions are 32-bit in Thumb2, it makes things easy.
We just need to select the appropriate adjustment.  Do this by moving
the adjustment out of do_undefinstr() into the assembly code, as only
the assembly code knows whether it's dealing with a 32-bit or 16-bit
instruction.

Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ARM: 7478/1: errata: extend workaround for erratum #720789 2012-08-09T23:25:06+00:00 Will Deacon will.deacon@arm.com 2012-07-20T17:24:55+00:00 1b48195822f358b7b5e44369cfa34463f0b9e09f commit 5a783cbc48367cfc7b65afc75430953dfe60098f upstream. Commit cdf357f1 ("ARM: 6299/1: errata: TLBIASIDIS and TLBIMVAIS operations can broadcast a faulty ASID") replaced by-ASID TLB flushing operations with all-ASID variants to workaround A9 erratum #720789. This patch extends the workaround to include the tlb_range operations, which were overlooked by the original patch. Tested-by: Steve Capper <steve.capper@arm.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 5a783cbc48367cfc7b65afc75430953dfe60098f upstream.

Commit cdf357f1 ("ARM: 6299/1: errata: TLBIASIDIS and TLBIMVAIS
operations can broadcast a faulty ASID") replaced by-ASID TLB flushing
operations with all-ASID variants to workaround A9 erratum #720789.

This patch extends the workaround to include the tlb_range operations,
which were overlooked by the original patch.

Tested-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ARM: 7477/1: vfp: Always save VFP state in vfp_pm_suspend on UP 2012-08-09T23:25:05+00:00 Colin Cross ccross@android.com 2012-07-20T01:03:42+00:00 cdab6eed39be615703f582831765e14fb0452842 commit 24b35521b8ddf088531258f06f681bb7b227bf47 upstream. vfp_pm_suspend should save the VFP state in suspend after any lazy context switch. If it only saves when the VFP is enabled, the state can get lost when, on a UP system: Thread 1 uses the VFP Context switch occurs to thread 2, VFP is disabled but the VFP context is not saved Thread 2 initiates suspend vfp_pm_suspend is called with the VFP disabled, and the unsaved VFP context of Thread 1 in the registers Modify vfp_pm_suspend to save the VFP context whenever vfp_current_hw_state is not NULL. Includes a fix from Ido Yariv <ido@wizery.com>, who pointed out that on SMP systems, the state pointer can be pointing to a freed task struct if a task exited on another cpu, fixed by using #ifndef CONFIG_SMP in the new if clause. Cc: Barry Song <bs14@csr.com> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Ido Yariv <ido@wizery.com> Cc: Daniel Drake <dsd@laptop.org> Cc: Will Deacon <will.deacon@arm.com> Signed-off-by: Colin Cross <ccross@android.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 24b35521b8ddf088531258f06f681bb7b227bf47 upstream.

vfp_pm_suspend should save the VFP state in suspend after
any lazy context switch.  If it only saves when the VFP is enabled,
the state can get lost when, on a UP system:
  Thread 1 uses the VFP
  Context switch occurs to thread 2, VFP is disabled but the
     VFP context is not saved
  Thread 2 initiates suspend
  vfp_pm_suspend is called with the VFP disabled, and the unsaved
     VFP context of Thread 1 in the registers

Modify vfp_pm_suspend to save the VFP context whenever
vfp_current_hw_state is not NULL.

Includes a fix from Ido Yariv <ido@wizery.com>, who pointed out that on
SMP systems, the state pointer can be pointing to a freed task struct if
a task exited on another cpu, fixed by using #ifndef CONFIG_SMP in the
new if clause.

Cc: Barry Song <bs14@csr.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Ido Yariv <ido@wizery.com>
Cc: Daniel Drake <dsd@laptop.org>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Colin Cross <ccross@android.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ARM: 7476/1: vfp: only clear vfp state for current cpu in vfp_pm_suspend 2012-08-09T23:25:04+00:00 Colin Cross ccross@android.com 2012-07-20T01:03:43+00:00 fc257bc543a3b08545e2332574a12893f0e54c85 commit a84b895a2348f0dbff31b71ddf954f70a6cde368 upstream. vfp_pm_suspend runs on each cpu, only clear the hardware state pointer for the current cpu. Prevents a possible crash if one cpu clears the hw state pointer when another cpu has already checked if it is valid. Signed-off-by: Colin Cross <ccross@android.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit a84b895a2348f0dbff31b71ddf954f70a6cde368 upstream.

vfp_pm_suspend runs on each cpu, only clear the hardware state
pointer for the current cpu.  Prevents a possible crash if one
cpu clears the hw state pointer when another cpu has already
checked if it is valid.

Signed-off-by: Colin Cross <ccross@android.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ARM: 7467/1: mutex: use generic xchg-based implementation for ARMv6+ 2012-08-09T23:25:04+00:00 Will Deacon will.deacon@arm.com 2012-07-13T18:15:40+00:00 c51cf2421d129a98183c2e9fd2209fa96297f118 commit a76d7bd96d65fa5119adba97e1b58d95f2e78829 upstream. The open-coded mutex implementation for ARMv6+ cores suffers from a severe lack of barriers, so in the uncontended case we don't actually protect any accesses performed during the critical section. Furthermore, the code is largely a duplication of the ARMv6+ atomic_dec code but optimised to remove a branch instruction, as the mutex fastpath was previously inlined. Now that this is executed out-of-line, we can reuse the atomic access code for the locking (in fact, we use the xchg code as this produces shorter critical sections). This patch uses the generic xchg based implementation for mutexes on ARMv6+, which introduces barriers to the lock/unlock operations and also has the benefit of removing a fair amount of inline assembly code. Acked-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Nicolas Pitre <nico@linaro.org> Reported-by: Shan Kang <kangshan0910@gmail.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit a76d7bd96d65fa5119adba97e1b58d95f2e78829 upstream.

The open-coded mutex implementation for ARMv6+ cores suffers from a
severe lack of barriers, so in the uncontended case we don't actually
protect any accesses performed during the critical section.

Furthermore, the code is largely a duplication of the ARMv6+ atomic_dec
code but optimised to remove a branch instruction, as the mutex fastpath
was previously inlined. Now that this is executed out-of-line, we can
reuse the atomic access code for the locking (in fact, we use the xchg
code as this produces shorter critical sections).

This patch uses the generic xchg based implementation for mutexes on
ARMv6+, which introduces barriers to the lock/unlock operations and also
has the benefit of removing a fair amount of inline assembly code.

Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Nicolas Pitre <nico@linaro.org>
Reported-by: Shan Kang <kangshan0910@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ARM: 7466/1: disable interrupt before spinning endlessly 2012-08-09T23:25:03+00:00 Shawn Guo shawn.guo@linaro.org 2012-07-13T07:19:34+00:00 2bd8f381a4382eaeef391e0a551230610bece057 commit 98bd8b96b26db3399a48202318dca4aaa2515355 upstream. The CPU will endlessly spin at the end of machine_halt and machine_restart calls. However, this will lead to a soft lockup warning after about 20 seconds, if CONFIG_LOCKUP_DETECTOR is enabled, as system timer is still alive. Disable interrupt before going to spin endlessly, so that the lockup warning will never be seen. Reported-by: Marek Vasut <marex@denx.de> Signed-off-by: Shawn Guo <shawn.guo@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> [bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 98bd8b96b26db3399a48202318dca4aaa2515355 upstream.

The CPU will endlessly spin at the end of machine_halt and
machine_restart calls.  However, this will lead to a soft lockup
warning after about 20 seconds, if CONFIG_LOCKUP_DETECTOR is enabled,
as system timer is still alive.

Disable interrupt before going to spin endlessly, so that the lockup
warning will never be seen.

Reported-by: Marek Vasut <marex@denx.de>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
s390/mm: fix fault handling for page table walk case 2012-08-09T23:24:58+00:00 Heiko Carstens heiko.carstens@de.ibm.com 2012-07-27T07:45:39+00:00 c929e9bc8cadd3dd1e914498e3905d3d4cc43f85 commit 008c2e8f247f0a8db1e8e26139da12f3a3abcda0 upstream. Make sure the kernel does not incorrectly create a SIGBUS signal during user space accesses: For user space accesses in the switched addressing mode case the kernel may walk page tables and access user address space via the kernel mapping. If a page table entry is invalid the function __handle_fault() gets called in order to emulate a page fault and trigger all the usual actions like paging in a missing page etc. by calling handle_mm_fault(). If handle_mm_fault() returns with an error fixup handling is necessary. For the switched addressing mode case all errors need to be mapped to -EFAULT, so that the calling uaccess function can return -EFAULT to user space. Unfortunately the __handle_fault() incorrectly calls do_sigbus() if VM_FAULT_SIGBUS is set. This however should only happen if a page fault was triggered by a user space instruction. For kernel mode uaccesses the correct action is to only return -EFAULT. So user space may incorrectly see SIGBUS signals because of this bug. For current machines this would only be possible for the switched addressing mode case in conjunction with futex operations. Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [bwh: Backported to 3.2: do_exception() and do_sigbus() parameters differ] Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 008c2e8f247f0a8db1e8e26139da12f3a3abcda0 upstream.

Make sure the kernel does not incorrectly create a SIGBUS signal during
user space accesses:

For user space accesses in the switched addressing mode case the kernel
may walk page tables and access user address space via the kernel
mapping. If a page table entry is invalid the function __handle_fault()
gets called in order to emulate a page fault and trigger all the usual
actions like paging in a missing page etc. by calling handle_mm_fault().

If handle_mm_fault() returns with an error fixup handling is necessary.
For the switched addressing mode case all errors need to be mapped to
-EFAULT, so that the calling uaccess function can return -EFAULT to
user space.

Unfortunately the __handle_fault() incorrectly calls do_sigbus() if
VM_FAULT_SIGBUS is set. This however should only happen if a page fault
was triggered by a user space instruction. For kernel mode uaccesses
the correct action is to only return -EFAULT.
So user space may incorrectly see SIGBUS signals because of this bug.

For current machines this would only be possible for the switched
addressing mode case in conjunction with futex operations.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
[bwh: Backported to 3.2: do_exception() and do_sigbus() parameters differ]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Redefine ATOMIC_INIT and ATOMIC64_INIT to drop the casts 2012-08-09T23:24:54+00:00 Tony Luck tony.luck@intel.com 2012-07-26T17:55:26+00:00 d0192ce751ef75b4f40f5b2d4264f2406080163b commit a119365586b0130dfea06457f584953e0ff6481d upstream. The following build error occured during a ia64 build with swap-over-NFS patches applied. net/core/sock.c:274:36: error: initializer element is not constant net/core/sock.c:274:36: error: (near initialization for 'memalloc_socks') net/core/sock.c:274:36: error: initializer element is not constant This is identical to a parisc build error. Fengguang Wu, Mel Gorman and James Bottomley did all the legwork to track the root cause of the problem. This fix and entire commit log is shamelessly copied from them with one extra detail to change a dubious runtime use of ATOMIC_INIT() to atomic_set() in drivers/char/mspec.c Dave Anglin says: > Here is the line in sock.i: > > struct static_key memalloc_socks = ((struct static_key) { .enabled = > ((atomic_t) { (0) }) }); The above line contains two compound literals. It also uses a designated initializer to initialize the field enabled. A compound literal is not a constant expression. The location of the above statement isn't fully clear, but if a compound literal occurs outside the body of a function, the initializer list must consist of constant expressions. Signed-off-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit a119365586b0130dfea06457f584953e0ff6481d upstream.

The following build error occured during a ia64 build with
swap-over-NFS patches applied.

net/core/sock.c:274:36: error: initializer element is not constant
net/core/sock.c:274:36: error: (near initialization for 'memalloc_socks')
net/core/sock.c:274:36: error: initializer element is not constant

This is identical to a parisc build error. Fengguang Wu, Mel Gorman
and James Bottomley did all the legwork to track the root cause of
the problem. This fix and entire commit log is shamelessly copied
from them with one extra detail to change a dubious runtime use of
ATOMIC_INIT() to atomic_set() in drivers/char/mspec.c

Dave Anglin says:
> Here is the line in sock.i:
>
> struct static_key memalloc_socks = ((struct static_key) { .enabled =
> ((atomic_t) { (0) }) });

The above line contains two compound literals.  It also uses a designated
initializer to initialize the field enabled.  A compound literal is not a
constant expression.

The location of the above statement isn't fully clear, but if a compound
literal occurs outside the body of a function, the initializer list must
consist of constant expressions.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
s390/mm: downgrade page table after fork of a 31 bit process 2012-08-09T23:24:54+00:00 Martin Schwidefsky schwidefsky@de.ibm.com 2012-07-26T06:53:06+00:00 f22027d0e8398e377a42a62241137b77a35eda97 commit 0f6f281b731d20bfe75c13f85d33f3f05b440222 upstream. The downgrade of the 4 level page table created by init_new_context is currently done only in start_thread31. If a 31 bit process forks the new mm uses a 4 level page table, including the task size of 2<<42 that goes along with it. This is incorrect as now a 31 bit process can map memory beyond 2GB. Define arch_dup_mmap to do the downgrade after fork. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk> 
commit 0f6f281b731d20bfe75c13f85d33f3f05b440222 upstream.

The downgrade of the 4 level page table created by init_new_context is
currently done only in start_thread31. If a 31 bit process forks the
new mm uses a 4 level page table, including the task size of 2<<42
that goes along with it. This is incorrect as now a 31 bit process
can map memory beyond 2GB. Define arch_dup_mmap to do the downgrade
after fork.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>