From 3e5d8f978435bb9ba4dfe3f4514e65e7885db1a9 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: make percpu symbols zerobased on SMP [ Based on original patch from Christoph Lameter and Mike Travis. ] This patch makes percpu symbols zerobased on x86_64 SMP by adding PERCPU_VADDR() to vmlinux.lds.h which helps setting explicit vaddr on the percpu output section and using it in vmlinux_64.lds.S. A new PHDR is added as existing ones cannot contain sections near address zero. PERCPU_VADDR() also adds a new symbol __per_cpu_load which always points to the vaddr of the loaded percpu data.init region. The following adjustments have been made to accomodate the address change. * code to locate percpu gdt_page in head_64.S is updated to add the load address to the gdt_page offset. * __per_cpu_load is used in places where access to the init data area is necessary. * pda->data_offset is initialized soon after C code is entered as zero value doesn't work anymore. This patch is mostly taken from Mike Travis' "x86_64: Base percpu variables at zero" patch. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 0e275d495563..7ee0363871e8 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -204,6 +204,23 @@ ENTRY(secondary_startup_64) pushq $0 popfq +#ifdef CONFIG_SMP + /* + * early_gdt_base should point to the gdt_page in static percpu init + * data area. Computing this requires two symbols - __per_cpu_load + * and per_cpu__gdt_page. As linker can't do no such relocation, do + * it by hand. As early_gdt_descr is manipulated by C code for + * secondary CPUs, this should be done only once for the boot CPU + * when early_gdt_descr_base contains zero. + */ + movq early_gdt_descr_base(%rip), %rax + testq %rax, %rax + jnz 1f + movq $__per_cpu_load, %rax + addq $per_cpu__gdt_page, %rax + movq %rax, early_gdt_descr_base(%rip) +1: +#endif /* * We must switch to a new descriptor in kernel space for the GDT * because soon the kernel won't have access anymore to the userspace @@ -401,7 +418,12 @@ NEXT_PAGE(level2_spare_pgt) .globl early_gdt_descr early_gdt_descr: .word GDT_ENTRIES*8-1 - .quad per_cpu__gdt_page +#ifdef CONFIG_SMP +early_gdt_descr_base: + .quad 0x0000000000000000 +#else + .quad per_cpu__gdt_page +#endif ENTRY(phys_base) /* This must match the first entry in level2_kernel_pgt */ -- cgit v1.2.3 From f32ff5388d86518c0375ccdb330d3b459b9c405e Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: load pointer to pda into %gs while brining up a CPU [ Based on original patch from Christoph Lameter and Mike Travis. ] CPU startup code in head_64.S loaded address of a zero page into %gs for temporary use till pda is loaded but address to the actual pda is available at the point. Load the real address directly instead. This will help unifying percpu and pda handling later on. This patch is mostly taken from Mike Travis' "x86_64: Fold pda into per cpu area" patch. Signed-off-by: Tejun Heo --- arch/x86/kernel/head_64.S | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 7ee0363871e8..2f0ab0089883 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -243,12 +243,15 @@ ENTRY(secondary_startup_64) movl %eax,%fs movl %eax,%gs - /* - * Setup up a dummy PDA. this is just for some early bootup code - * that does in_interrupt() - */ + /* Set up %gs. + * + * %gs should point to the pda. For initial boot, make %gs point + * to the _boot_cpu_pda in data section. For a secondary CPU, + * initial_gs should be set to its pda address before the CPU runs + * this code. + */ movl $MSR_GS_BASE,%ecx - movq $empty_zero_page,%rax + movq initial_gs(%rip),%rax movq %rax,%rdx shrq $32,%rdx wrmsr @@ -274,6 +277,8 @@ ENTRY(secondary_startup_64) .align 8 ENTRY(initial_code) .quad x86_64_start_kernel + ENTRY(initial_gs) + .quad _boot_cpu_pda __FINITDATA ENTRY(stack_start) -- cgit v1.2.3 From 1a51e3a0aed18767cf2762e95456ecfeb0bca5e6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: fold pda into percpu area on SMP [ Based on original patch from Christoph Lameter and Mike Travis. ] Currently pdas and percpu areas are allocated separately. %gs points to local pda and percpu area can be reached using pda->data_offset. This patch folds pda into percpu area. Due to strange gcc requirement, pda needs to be at the beginning of the percpu area so that pda->stack_canary is at %gs:40. To achieve this, a new percpu output section macro - PERCPU_VADDR_PREALLOC() - is added and used to reserve pda sized chunk at the start of the percpu area. After this change, for boot cpu, %gs first points to pda in the data.init area and later during setup_per_cpu_areas() gets updated to point to the actual pda. This means that setup_per_cpu_areas() need to reload %gs for CPU0 while clearing pda area for other cpus as cpu0 already has modified it when control reaches setup_per_cpu_areas(). This patch also removes now unnecessary get_local_pda() and its call sites. A lot of this patch is taken from Mike Travis' "x86_64: Fold pda into per cpu area" patch. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 2f0ab0089883..7a995d0e9f78 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -245,10 +245,13 @@ ENTRY(secondary_startup_64) /* Set up %gs. * - * %gs should point to the pda. For initial boot, make %gs point - * to the _boot_cpu_pda in data section. For a secondary CPU, - * initial_gs should be set to its pda address before the CPU runs - * this code. + * On SMP, %gs should point to the per-cpu area. For initial + * boot, make %gs point to the init data section. For a + * secondary CPU,initial_gs should be set to its pda address + * before the CPU runs this code. + * + * On UP, initial_gs points to _boot_cpu_pda and doesn't + * change. */ movl $MSR_GS_BASE,%ecx movq initial_gs(%rip),%rax @@ -278,7 +281,11 @@ ENTRY(secondary_startup_64) ENTRY(initial_code) .quad x86_64_start_kernel ENTRY(initial_gs) +#ifdef CONFIG_SMP + .quad __per_cpu_load +#else .quad _boot_cpu_pda +#endif __FINITDATA ENTRY(stack_start) -- cgit v1.2.3 From b12d8db8fbfaed1e8222a15333a3645599636854 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: make pda a percpu variable [ Based on original patch from Christoph Lameter and Mike Travis. ] As pda is now allocated in percpu area, it can easily be made a proper percpu variable. Make it so by defining per cpu symbol from linker script and declaring it in C code for SMP and simply defining it for UP. This change cleans up code and brings SMP and UP closer a bit. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 7a995d0e9f78..c8ace880661b 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -19,6 +19,7 @@ #include #include #include +#include #ifdef CONFIG_PARAVIRT #include @@ -250,7 +251,7 @@ ENTRY(secondary_startup_64) * secondary CPU,initial_gs should be set to its pda address * before the CPU runs this code. * - * On UP, initial_gs points to _boot_cpu_pda and doesn't + * On UP, initial_gs points to PER_CPU_VAR(__pda) and doesn't * change. */ movl $MSR_GS_BASE,%ecx @@ -284,7 +285,7 @@ ENTRY(secondary_startup_64) #ifdef CONFIG_SMP .quad __per_cpu_load #else - .quad _boot_cpu_pda + .quad PER_CPU_VAR(__pda) #endif __FINITDATA -- cgit v1.2.3 From 8c7e58e690ae60ab4215b025f433ed4af261e103 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:28 +0900 Subject: x86: rework __per_cpu_load adjustments Impact: cleanup Use cpu_number to determine if the adjustment is necessary. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/head_64.S | 24 ++++++++---------------- 1 file changed, 8 insertions(+), 16 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index c8ace880661b..98ea26a2fca1 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -207,19 +207,15 @@ ENTRY(secondary_startup_64) #ifdef CONFIG_SMP /* - * early_gdt_base should point to the gdt_page in static percpu init - * data area. Computing this requires two symbols - __per_cpu_load - * and per_cpu__gdt_page. As linker can't do no such relocation, do - * it by hand. As early_gdt_descr is manipulated by C code for - * secondary CPUs, this should be done only once for the boot CPU - * when early_gdt_descr_base contains zero. + * Fix up static pointers that need __per_cpu_load added. The assembler + * is unable to do this directly. This is only needed for the boot cpu. + * These values are set up with the correct base addresses by C code for + * secondary cpus. */ - movq early_gdt_descr_base(%rip), %rax - testq %rax, %rax - jnz 1f - movq $__per_cpu_load, %rax - addq $per_cpu__gdt_page, %rax - movq %rax, early_gdt_descr_base(%rip) + movq initial_gs(%rip), %rax + cmpl $0, per_cpu__cpu_number(%rax) + jne 1f + addq %rax, early_gdt_descr_base(%rip) 1: #endif /* @@ -431,12 +427,8 @@ NEXT_PAGE(level2_spare_pgt) .globl early_gdt_descr early_gdt_descr: .word GDT_ENTRIES*8-1 -#ifdef CONFIG_SMP early_gdt_descr_base: - .quad 0x0000000000000000 -#else .quad per_cpu__gdt_page -#endif ENTRY(phys_base) /* This must match the first entry in level2_kernel_pgt */ -- cgit v1.2.3 From 947e76cdc34c782fc947313d4331380686eebbad Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:28 +0900 Subject: x86: move stack_canary into irq_stack Impact: x86_64 percpu area layout change, irq_stack now at the beginning Now that the PDA is empty except for the stack canary, it can be removed. The irqstack is moved to the start of the per-cpu section. If the stack protector is enabled, the canary overlaps the bottom 48 bytes of the irqstack. tj: * updated subject * dropped asm relocation of irq_stack_ptr * updated comments a bit * rebased on top of stack canary changes Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/head_64.S | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 98ea26a2fca1..a0a2b5ca9b7d 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -242,13 +242,10 @@ ENTRY(secondary_startup_64) /* Set up %gs. * - * On SMP, %gs should point to the per-cpu area. For initial - * boot, make %gs point to the init data section. For a - * secondary CPU,initial_gs should be set to its pda address - * before the CPU runs this code. - * - * On UP, initial_gs points to PER_CPU_VAR(__pda) and doesn't - * change. + * The base of %gs always points to the bottom of the irqstack + * union. If the stack protector canary is enabled, it is + * located at %gs:40. Note that, on SMP, the boot cpu uses + * init data section till per cpu areas are set up. */ movl $MSR_GS_BASE,%ecx movq initial_gs(%rip),%rax @@ -281,7 +278,7 @@ ENTRY(secondary_startup_64) #ifdef CONFIG_SMP .quad __per_cpu_load #else - .quad PER_CPU_VAR(__pda) + .quad PER_CPU_VAR(irq_stack_union) #endif __FINITDATA -- cgit v1.2.3 From 2add8e235cbe0dcd672c33fc322754e15500238c Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Sun, 8 Feb 2009 09:58:39 -0500 Subject: x86: use linker to offset symbols by __per_cpu_load Impact: cleanup and bug fix Use the linker to create symbols for certain per-cpu variables that are offset by __per_cpu_load. This allows the removal of the runtime fixup of the GDT pointer, which fixes a bug with resume reported by Jiri Slaby. Reported-by: Jiri Slaby Signed-off-by: Brian Gerst Acked-by: Jiri Slaby Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 21 ++------------------- 1 file changed, 2 insertions(+), 19 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index a0a2b5ca9b7d..2e648e3a5ea4 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -205,19 +205,6 @@ ENTRY(secondary_startup_64) pushq $0 popfq -#ifdef CONFIG_SMP - /* - * Fix up static pointers that need __per_cpu_load added. The assembler - * is unable to do this directly. This is only needed for the boot cpu. - * These values are set up with the correct base addresses by C code for - * secondary cpus. - */ - movq initial_gs(%rip), %rax - cmpl $0, per_cpu__cpu_number(%rax) - jne 1f - addq %rax, early_gdt_descr_base(%rip) -1: -#endif /* * We must switch to a new descriptor in kernel space for the GDT * because soon the kernel won't have access anymore to the userspace @@ -275,11 +262,7 @@ ENTRY(secondary_startup_64) ENTRY(initial_code) .quad x86_64_start_kernel ENTRY(initial_gs) -#ifdef CONFIG_SMP - .quad __per_cpu_load -#else - .quad PER_CPU_VAR(irq_stack_union) -#endif + .quad INIT_PER_CPU_VAR(irq_stack_union) __FINITDATA ENTRY(stack_start) @@ -425,7 +408,7 @@ NEXT_PAGE(level2_spare_pgt) early_gdt_descr: .word GDT_ENTRIES*8-1 early_gdt_descr_base: - .quad per_cpu__gdt_page + .quad INIT_PER_CPU_VAR(gdt_page) ENTRY(phys_base) /* This must match the first entry in level2_kernel_pgt */ -- cgit v1.2.3 From 2a0b1001116de51f13b2969a9a52bd2bc294644f Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Mon, 23 Feb 2009 22:56:57 +0300 Subject: x86: head_64.S - remove useless balign Impact: cleanup NEXT_PAGE already has 'balign' so no need to keep this redundant one. Signed-off-by: Cyrill Gorcunov Cc: heukelum@fastmail.fm Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 2 -- 1 file changed, 2 deletions(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 2e648e3a5ea4..52c9af429dd8 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -329,8 +329,6 @@ early_idt_ripmsg: #endif /* CONFIG_EARLY_PRINTK */ .previous -.balign PAGE_SIZE - #define NEXT_PAGE(name) \ .balign PAGE_SIZE; \ ENTRY(name) -- cgit v1.2.3 From 5e112ae23b404ccba0a61630b82ec44f0d084880 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Mon, 23 Feb 2009 22:56:58 +0300 Subject: x86: head_64.S - use IDT_ENTRIES instead of hardcoded number Impact: cleanup Signed-off-by: Cyrill Gorcunov Cc: heukelum@fastmail.fm Cc: Cyrill Gorcunov Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel/head_64.S') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 52c9af429dd8..54b29bb24e71 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -417,7 +417,7 @@ ENTRY(phys_base) .section .bss, "aw", @nobits .align L1_CACHE_BYTES ENTRY(idt_table) - .skip 256 * 16 + .skip IDT_ENTRIES * 16 .section .bss.page_aligned, "aw", @nobits .align PAGE_SIZE -- cgit v1.2.3