From e00320875d0cc5f8099a7227b2f25fbb3231268d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 14 Feb 2008 08:48:23 +0100 Subject: x86: fix stackprotector canary updates during context switches fix a bug noticed and fixed by pageexec@freemail.hu. if built with -fstack-protector-all then we'll have canary checks built into the __switch_to() function. That does not work well with the canary-switching code there: while we already use the %rsp of the new task, we still call __switch_to() whith the previous task's canary value in the PDA, hence the __switch_to() ssp prologue instructions will store the previous canary. Then we update the PDA and upon return from __switch_to() the canary check triggers and we panic. so update the canary after we have called __switch_to(), where we are at the same stackframe level as the last stackframe of the next (and now freshly current) task. Note: this means that we call __switch_to() [and its sub-functions] still with the old canary, but that is not a problem, both the previous and the next task has a high-quality canary. The only (mostly academic) disadvantage is that the canary of one task may leak onto the stack of another task, increasing the risk of information leaks, were an attacker able to read the stack of specific tasks (but not that of others). To solve this we'll have to reorganize the way we switch tasks, and move the PDA setting into the switch_to() assembly code. That will happen in another patch. Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- include/asm-x86/pda.h | 2 -- include/asm-x86/system.h | 6 +++++- include/linux/sched.h | 3 +-- 3 files changed, 6 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/include/asm-x86/pda.h b/include/asm-x86/pda.h index 101fb9e11954..62b734986a44 100644 --- a/include/asm-x86/pda.h +++ b/include/asm-x86/pda.h @@ -16,11 +16,9 @@ struct x8664_pda { unsigned long oldrsp; /* 24 user rsp for system call */ int irqcount; /* 32 Irq nesting counter. Starts -1 */ unsigned int cpunumber; /* 36 Logical CPU number */ -#ifdef CONFIG_CC_STACKPROTECTOR unsigned long stack_canary; /* 40 stack canary value */ /* gcc-ABI: this canary MUST be at offset 40!!! */ -#endif char *irqstackptr; unsigned int __softirq_pending; unsigned int __nmi_count; /* number of NMI on this CPUs */ diff --git a/include/asm-x86/system.h b/include/asm-x86/system.h index a2f04cd79b29..172f54185093 100644 --- a/include/asm-x86/system.h +++ b/include/asm-x86/system.h @@ -92,6 +92,8 @@ do { \ ".globl thread_return\n" \ "thread_return:\n\t" \ "movq %%gs:%P[pda_pcurrent],%%rsi\n\t" \ + "movq %P[task_canary](%%rsi),%%r8\n\t" \ + "movq %%r8,%%gs:%P[pda_canary]\n\t" \ "movq %P[thread_info](%%rsi),%%r8\n\t" \ LOCK_PREFIX "btr %[tif_fork],%P[ti_flags](%%r8)\n\t" \ "movq %%rax,%%rdi\n\t" \ @@ -103,7 +105,9 @@ do { \ [ti_flags] "i" (offsetof(struct thread_info, flags)), \ [tif_fork] "i" (TIF_FORK), \ [thread_info] "i" (offsetof(struct task_struct, stack)), \ - [pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)) \ + [task_canary] "i" (offsetof(struct task_struct, stack_canary)),\ + [pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)), \ + [pda_canary] "i" (offsetof(struct x8664_pda, stack_canary))\ : "memory", "cc" __EXTRA_CLOBBER) #endif diff --git a/include/linux/sched.h b/include/linux/sched.h index 5395a6176f4b..d6a515158783 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1096,10 +1096,9 @@ struct task_struct { pid_t pid; pid_t tgid; -#ifdef CONFIG_CC_STACKPROTECTOR /* Canary value for the -fstack-protector gcc feature */ unsigned long stack_canary; -#endif + /* * pointers to (original) parent process, youngest child, younger sibling, * older sibling, respectively. (p->father can be replaced with -- cgit v1.2.3 From 9b5609fd773e6ac0b1d6d6e1bf68f32cca64e06b Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 14 Feb 2008 09:41:09 +0100 Subject: stackprotector: include files create for core kernel files to include. Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- include/asm-x86/stackprotector.h | 4 ++++ include/linux/stackprotector.h | 8 ++++++++ init/main.c | 1 + 3 files changed, 13 insertions(+) create mode 100644 include/asm-x86/stackprotector.h create mode 100644 include/linux/stackprotector.h (limited to 'include') diff --git a/include/asm-x86/stackprotector.h b/include/asm-x86/stackprotector.h new file mode 100644 index 000000000000..dcac7a6bdba2 --- /dev/null +++ b/include/asm-x86/stackprotector.h @@ -0,0 +1,4 @@ +#ifndef _ASM_STACKPROTECTOR_H +#define _ASM_STACKPROTECTOR_H 1 + +#endif diff --git a/include/linux/stackprotector.h b/include/linux/stackprotector.h new file mode 100644 index 000000000000..d3e8bbe602f8 --- /dev/null +++ b/include/linux/stackprotector.h @@ -0,0 +1,8 @@ +#ifndef _LINUX_STACKPROTECTOR_H +#define _LINUX_STACKPROTECTOR_H 1 + +#ifdef CONFIG_CC_STACKPROTECTOR +# include +#endif + +#endif diff --git a/init/main.c b/init/main.c index f7fb20021d48..a84322ca64a2 100644 --- a/init/main.c +++ b/init/main.c @@ -14,6 +14,7 @@ #include #include #include +#include #include #include #include -- cgit v1.2.3 From 18aa8bb12dcb10adc3d7c9d69714d53667c0ab7f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 14 Feb 2008 09:42:02 +0100 Subject: stackprotector: add boot_init_stack_canary() add the boot_init_stack_canary() and make the secondary idle threads use it. Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/kernel/process_64.c | 6 ++---- include/asm-x86/stackprotector.h | 20 ++++++++++++++++++++ include/linux/stackprotector.h | 4 ++++ 3 files changed, 26 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index d4c7ac7aa430..5107cb214c7b 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -147,7 +147,6 @@ void cpu_idle(void) { current_thread_info()->status |= TS_POLLING; -#ifdef CONFIG_CC_STACKPROTECTOR /* * If we're the non-boot CPU, nothing set the PDA stack * canary up for us - and if we are the boot CPU we have @@ -156,9 +155,8 @@ void cpu_idle(void) * invalid canaries already on the stack wont ever * trigger): */ - current->stack_canary = get_random_int(); - write_pda(stack_canary, current->stack_canary); -#endif + boot_init_stack_canary(); + /* endless idle loop with no priority at all */ while (1) { tick_nohz_stop_sched_tick(); diff --git a/include/asm-x86/stackprotector.h b/include/asm-x86/stackprotector.h index dcac7a6bdba2..0f91f7a2688c 100644 --- a/include/asm-x86/stackprotector.h +++ b/include/asm-x86/stackprotector.h @@ -1,4 +1,24 @@ #ifndef _ASM_STACKPROTECTOR_H #define _ASM_STACKPROTECTOR_H 1 +/* + * Initialize the stackprotector canary value. + * + * NOTE: this must only be called from functions that never return, + * and it must always be inlined. + */ +static __always_inline void boot_init_stack_canary(void) +{ + /* + * If we're the non-boot CPU, nothing set the PDA stack + * canary up for us - and if we are the boot CPU we have + * a 0 stack canary. This is a good place for updating + * it, as we wont ever return from this function (so the + * invalid canaries already on the stack wont ever + * trigger): + */ + current->stack_canary = get_random_int(); + write_pda(stack_canary, current->stack_canary); +} + #endif diff --git a/include/linux/stackprotector.h b/include/linux/stackprotector.h index d3e8bbe602f8..422e71aafd0b 100644 --- a/include/linux/stackprotector.h +++ b/include/linux/stackprotector.h @@ -3,6 +3,10 @@ #ifdef CONFIG_CC_STACKPROTECTOR # include +#else +static inline void boot_init_stack_canary(void) +{ +} #endif #endif -- cgit v1.2.3 From 420594296838fdc9a674470d710cda7d1487f9f4 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 14 Feb 2008 09:44:08 +0100 Subject: x86: fix the stackprotector canary of the boot CPU Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/kernel/process_64.c | 1 + include/linux/stackprotector.h | 4 ++++ init/main.c | 6 ++++++ 3 files changed, 11 insertions(+) (limited to 'include') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 5107cb214c7b..cce47f7fbf22 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -16,6 +16,7 @@ #include +#include #include #include #include diff --git a/include/linux/stackprotector.h b/include/linux/stackprotector.h index 422e71aafd0b..6f3e54c704c0 100644 --- a/include/linux/stackprotector.h +++ b/include/linux/stackprotector.h @@ -1,6 +1,10 @@ #ifndef _LINUX_STACKPROTECTOR_H #define _LINUX_STACKPROTECTOR_H 1 +#include +#include +#include + #ifdef CONFIG_CC_STACKPROTECTOR # include #else diff --git a/init/main.c b/init/main.c index a84322ca64a2..b44e4eb0f5e3 100644 --- a/init/main.c +++ b/init/main.c @@ -546,6 +546,12 @@ asmlinkage void __init start_kernel(void) unwind_init(); lockdep_init(); debug_objects_early_init(); + + /* + * Set up the the initial canary ASAP: + */ + boot_init_stack_canary(); + cgroup_init_early(); local_irq_disable(); -- cgit v1.2.3 From 960a672bd9f1ec06e8f197cf81a50fd07ea02e7f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 14 Feb 2008 09:56:04 +0100 Subject: x86: stackprotector: mix TSC to the boot canary mix the TSC to the boot canary. Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- include/asm-x86/stackprotector.h | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/include/asm-x86/stackprotector.h b/include/asm-x86/stackprotector.h index 0f91f7a2688c..3baf7ad89be1 100644 --- a/include/asm-x86/stackprotector.h +++ b/include/asm-x86/stackprotector.h @@ -1,6 +1,8 @@ #ifndef _ASM_STACKPROTECTOR_H #define _ASM_STACKPROTECTOR_H 1 +#include + /* * Initialize the stackprotector canary value. * @@ -9,16 +11,28 @@ */ static __always_inline void boot_init_stack_canary(void) { + u64 canary; + u64 tsc; + /* * If we're the non-boot CPU, nothing set the PDA stack * canary up for us - and if we are the boot CPU we have * a 0 stack canary. This is a good place for updating * it, as we wont ever return from this function (so the * invalid canaries already on the stack wont ever - * trigger): + * trigger). + * + * We both use the random pool and the current TSC as a source + * of randomness. The TSC only matters for very early init, + * there it already has some randomness on most systems. Later + * on during the bootup the random pool has true entropy too. */ - current->stack_canary = get_random_int(); - write_pda(stack_canary, current->stack_canary); + get_random_bytes(&canary, sizeof(canary)); + tsc = __native_read_tsc(); + canary += tsc + (tsc << 32UL); + + current->stack_canary = canary; + write_pda(stack_canary, canary); } #endif -- cgit v1.2.3 From 7c9f8861e6c9c839f913e49b98c3854daca18f27 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Tue, 22 Apr 2008 16:38:23 -0500 Subject: stackprotector: use canary at end of stack to indicate overruns at oops time (Updated with a common max-stack-used checker that knows about the canary, as suggested by Joe Perches) Use a canary at the end of the stack to clearly indicate at oops time whether the stack has ever overflowed. This is a very simple implementation with a couple of drawbacks: 1) a thread may legitimately use exactly up to the last word on the stack -- but the chances of doing this and then oopsing later seem slim 2) it's possible that the stack usage isn't dense enough that the canary location could get skipped over -- but the worst that happens is that we don't flag the overrun -- though this happens fairly often in my testing :( With the code in place, an intentionally-bloated stack oops might do: BUG: unable to handle kernel paging request at ffff8103f84cc680 IP: [] update_curr+0x9a/0xa8 PGD 8063 PUD 0 Thread overran stack or stack corrupted Oops: 0000 [1] SMP CPU 0 ... ... unless the stack overrun is so bad that it corrupts some other thread. Signed-off-by: Eric Sandeen Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/mm/fault.c | 7 +++++++ include/linux/magic.h | 1 + include/linux/sched.h | 13 +++++++++++++ kernel/exit.c | 5 +---- kernel/fork.c | 5 +++++ kernel/sched.c | 7 +------ 6 files changed, 28 insertions(+), 10 deletions(-) (limited to 'include') diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c index fd7e1798c75a..1f524df68b96 100644 --- a/arch/x86/mm/fault.c +++ b/arch/x86/mm/fault.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include @@ -581,6 +582,8 @@ void __kprobes do_page_fault(struct pt_regs *regs, unsigned long error_code) unsigned long address; int write, si_code; int fault; + unsigned long *stackend; + #ifdef CONFIG_X86_64 unsigned long flags; #endif @@ -850,6 +853,10 @@ no_context: show_fault_oops(regs, error_code, address); + stackend = end_of_stack(tsk); + if (*stackend != STACK_END_MAGIC) + printk(KERN_ALERT "Thread overran stack, or stack corrupted\n"); + tsk->thread.cr2 = address; tsk->thread.trap_no = 14; tsk->thread.error_code = error_code; diff --git a/include/linux/magic.h b/include/linux/magic.h index 1fa0c2ce4dec..74e68e201166 100644 --- a/include/linux/magic.h +++ b/include/linux/magic.h @@ -42,4 +42,5 @@ #define FUTEXFS_SUPER_MAGIC 0xBAD1DEA #define INOTIFYFS_SUPER_MAGIC 0x2BAD1DEA +#define STACK_END_MAGIC 0x57AC6E9D #endif /* __LINUX_MAGIC_H__ */ diff --git a/include/linux/sched.h b/include/linux/sched.h index d6a515158783..c5181e77f305 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1969,6 +1969,19 @@ static inline unsigned long *end_of_stack(struct task_struct *p) extern void thread_info_cache_init(void); +#ifdef CONFIG_DEBUG_STACK_USAGE +static inline unsigned long stack_not_used(struct task_struct *p) +{ + unsigned long *n = end_of_stack(p); + + do { /* Skip over canary */ + n++; + } while (!*n); + + return (unsigned long)n - (unsigned long)end_of_stack(p); +} +#endif + /* set thread flags in other task's structures * - see asm/thread_info.h for TIF_xxxx flags available */ diff --git a/kernel/exit.c b/kernel/exit.c index 8f6185e69b69..fb8de6cbf2c7 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -899,12 +899,9 @@ static void check_stack_usage(void) { static DEFINE_SPINLOCK(low_water_lock); static int lowest_to_date = THREAD_SIZE; - unsigned long *n = end_of_stack(current); unsigned long free; - while (*n == 0) - n++; - free = (unsigned long)n - (unsigned long)end_of_stack(current); + free = stack_not_used(current); if (free >= lowest_to_date) return; diff --git a/kernel/fork.c b/kernel/fork.c index 19908b26cf80..d428336e7aa1 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -54,6 +54,7 @@ #include #include #include +#include #include #include @@ -186,6 +187,8 @@ static struct task_struct *dup_task_struct(struct task_struct *orig) { struct task_struct *tsk; struct thread_info *ti; + unsigned long *stackend; + int err; prepare_to_copy(orig); @@ -211,6 +214,8 @@ static struct task_struct *dup_task_struct(struct task_struct *orig) goto out; setup_thread_stack(tsk, orig); + stackend = end_of_stack(tsk); + *stackend = STACK_END_MAGIC; /* for overflow detection */ #ifdef CONFIG_CC_STACKPROTECTOR tsk->stack_canary = get_random_int(); diff --git a/kernel/sched.c b/kernel/sched.c index cfa222a91539..a964ed945094 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5748,12 +5748,7 @@ void sched_show_task(struct task_struct *p) printk(KERN_CONT " %016lx ", thread_saved_pc(p)); #endif #ifdef CONFIG_DEBUG_STACK_USAGE - { - unsigned long *n = end_of_stack(p); - while (!*n) - n++; - free = (unsigned long)n - (unsigned long)end_of_stack(p); - } + free = stack_not_used(p); #endif printk(KERN_CONT "%5lu %5d %6d\n", free, task_pid_nr(p), task_pid_nr(p->real_parent)); -- cgit v1.2.3 From af9ff7868f0f76d3364351b1641b9dfa99588e77 Mon Sep 17 00:00:00 2001 From: Arjan van de Ven Date: Sat, 12 Jul 2008 09:36:38 -0700 Subject: x86: simplify stackprotector self-check Clean up the code by removing no longer needed code; make sure the pda is updated and kept in sync Signed-off-by: Arjan van de Ven Signed-off-by: Ingo Molnar --- include/asm-x86/pda.h | 1 + kernel/panic.c | 29 +++++++---------------------- 2 files changed, 8 insertions(+), 22 deletions(-) (limited to 'include') diff --git a/include/asm-x86/pda.h b/include/asm-x86/pda.h index 62b734986a44..a5ff5bb76299 100644 --- a/include/asm-x86/pda.h +++ b/include/asm-x86/pda.h @@ -131,4 +131,5 @@ do { \ #define PDA_STACKOFFSET (5*8) +#define refresh_stack_canary() write_pda(stack_canary, current->stack_canary) #endif diff --git a/kernel/panic.c b/kernel/panic.c index 28153aec7100..87445a894c3a 100644 --- a/kernel/panic.c +++ b/kernel/panic.c @@ -328,37 +328,21 @@ EXPORT_SYMBOL(warn_on_slowpath); #ifndef GCC_HAS_SP #warning You have selected the CONFIG_CC_STACKPROTECTOR option, but the gcc used does not support this. #endif + static unsigned long __stack_check_testing; + /* * Self test function for the stack-protector feature. * This test requires that the local variable absolutely has - * a stack slot, hence the barrier()s. + * a stack slot. */ static noinline void __stack_chk_test_func(void) { - unsigned long foo; - barrier(); - /* - * we need to make sure we're not about to clobber the return address, - * while real exploits do this, it's unhealthy on a running system. - * Besides, if we would, the test is already failed anyway so - * time to pull the emergency brake on it. - */ - if ((unsigned long)__builtin_return_address(0) == - *(((unsigned long *)&foo)+1)) { - printk(KERN_ERR "No -fstack-protector-stack-frame!\n"); - } -#ifdef CONFIG_FRAME_POINTER - /* We also don't want to clobber the frame pointer */ - if ((unsigned long)__builtin_return_address(0) == - *(((unsigned long *)&foo)+2)) { - printk(KERN_ERR "No -fstack-protector-stack-frame!\n"); - } -#endif - if (current->stack_canary != *(((unsigned long *)&foo)+1)) - printk(KERN_ERR "No -fstack-protector canary found\n"); + unsigned long dummy_buffer[64]; /* force gcc to use the canary */ current->stack_canary = ~current->stack_canary; + refresh_stack_canary(); + dummy_buffer[3] = 1; /* fool gcc into keeping the variable */ } static int __stack_chk_test(void) @@ -371,6 +355,7 @@ static int __stack_chk_test(void) WARN_ON(1); }; current->stack_canary = ~current->stack_canary; + refresh_stack_canary(); return 0; } /* -- cgit v1.2.3 From 7f7ace0cda64c99599c23785f8979a072e118058 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:08 -0800 Subject: cpumask: update irq_desc to use cpumask_var_t Impact: reduce memory usage, use new cpumask API. Replace the affinity and pending_masks with cpumask_var_t's. This adds to the significant size reduction done with the SPARSE_IRQS changes. The added functions (init_alloc_desc_masks & init_copy_desc_masks) are in the include file so they can be inlined (and optimized out for the !CONFIG_CPUMASKS_OFFSTACK case.) [Naming chosen to be consistent with the other init*irq functions, as well as the backwards arg declaration of "from, to" instead of the more common "to, from" standard.] Includes a slight change to the declaration of struct irq_desc to embed the pending_mask within ifdef(CONFIG_SMP) to be consistent with other references, and some small changes to Xen. Tested: sparse/non-sparse/cpumask_offstack/non-cpumask_offstack/nonuma/nosmp on x86_64 Signed-off-by: Mike Travis Cc: Chris Wright Cc: Jeremy Fitzhardinge Cc: KOSAKI Motohiro Cc: Venkatesh Pallipadi Cc: virtualization@lists.osdl.org Cc: xen-devel@lists.xensource.com Cc: Yinghai Lu --- arch/x86/kernel/io_apic.c | 20 ++++++------ arch/x86/kernel/irq_32.c | 2 +- arch/x86/kernel/irq_64.c | 2 +- drivers/xen/events.c | 4 +-- include/linux/irq.h | 81 +++++++++++++++++++++++++++++++++++++++++++++-- kernel/irq/chip.c | 5 ++- kernel/irq/handle.c | 26 ++++++++------- kernel/irq/manage.c | 12 +++---- kernel/irq/migration.c | 12 +++---- kernel/irq/numa_migrate.c | 12 ++++++- kernel/irq/proc.c | 4 +-- 11 files changed, 135 insertions(+), 45 deletions(-) (limited to 'include') diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 1c4a1302536c..1337eab60ecc 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -356,7 +356,7 @@ set_extra_move_desc(struct irq_desc *desc, const struct cpumask *mask) if (!cfg->move_in_progress) { /* it means that domain is not changed */ - if (!cpumask_intersects(&desc->affinity, mask)) + if (!cpumask_intersects(desc->affinity, mask)) cfg->move_desc_pending = 1; } } @@ -579,9 +579,9 @@ set_desc_affinity(struct irq_desc *desc, const struct cpumask *mask) if (assign_irq_vector(irq, cfg, mask)) return BAD_APICID; - cpumask_and(&desc->affinity, cfg->domain, mask); + cpumask_and(desc->affinity, cfg->domain, mask); set_extra_move_desc(desc, mask); - return cpu_mask_to_apicid_and(&desc->affinity, cpu_online_mask); + return cpu_mask_to_apicid_and(desc->affinity, cpu_online_mask); } static void @@ -2383,7 +2383,7 @@ migrate_ioapic_irq_desc(struct irq_desc *desc, const struct cpumask *mask) if (cfg->move_in_progress) send_cleanup_vector(cfg); - cpumask_copy(&desc->affinity, mask); + cpumask_copy(desc->affinity, mask); } static int migrate_irq_remapped_level_desc(struct irq_desc *desc) @@ -2405,11 +2405,11 @@ static int migrate_irq_remapped_level_desc(struct irq_desc *desc) } /* everthing is clear. we have right of way */ - migrate_ioapic_irq_desc(desc, &desc->pending_mask); + migrate_ioapic_irq_desc(desc, desc->pending_mask); ret = 0; desc->status &= ~IRQ_MOVE_PENDING; - cpumask_clear(&desc->pending_mask); + cpumask_clear(desc->pending_mask); unmask: unmask_IO_APIC_irq_desc(desc); @@ -2434,7 +2434,7 @@ static void ir_irq_migration(struct work_struct *work) continue; } - desc->chip->set_affinity(irq, &desc->pending_mask); + desc->chip->set_affinity(irq, desc->pending_mask); spin_unlock_irqrestore(&desc->lock, flags); } } @@ -2448,7 +2448,7 @@ static void set_ir_ioapic_affinity_irq_desc(struct irq_desc *desc, { if (desc->status & IRQ_LEVEL) { desc->status |= IRQ_MOVE_PENDING; - cpumask_copy(&desc->pending_mask, mask); + cpumask_copy(desc->pending_mask, mask); migrate_irq_remapped_level_desc(desc); return; } @@ -2516,7 +2516,7 @@ static void irq_complete_move(struct irq_desc **descp) /* domain has not changed, but affinity did */ me = smp_processor_id(); - if (cpu_isset(me, desc->affinity)) { + if (cpumask_test_cpu(me, desc->affinity)) { *descp = desc = move_irq_desc(desc, me); /* get the new one */ cfg = desc->chip_data; @@ -4039,7 +4039,7 @@ void __init setup_ioapic_dest(void) */ if (desc->status & (IRQ_NO_BALANCING | IRQ_AFFINITY_SET)) - mask = &desc->affinity; + mask = desc->affinity; else mask = TARGET_CPUS; diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c index 74b9ff7341e9..e0f29be8ab0b 100644 --- a/arch/x86/kernel/irq_32.c +++ b/arch/x86/kernel/irq_32.c @@ -248,7 +248,7 @@ void fixup_irqs(void) if (irq == 2) continue; - affinity = &desc->affinity; + affinity = desc->affinity; if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids) { printk("Breaking affinity for irq %i\n", irq); affinity = cpu_all_mask; diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c index 63c88e6ec025..0b21cb1ea11f 100644 --- a/arch/x86/kernel/irq_64.c +++ b/arch/x86/kernel/irq_64.c @@ -100,7 +100,7 @@ void fixup_irqs(void) /* interrupt's are disabled at this point */ spin_lock(&desc->lock); - affinity = &desc->affinity; + affinity = desc->affinity; if (!irq_has_action(irq) || cpumask_equal(affinity, cpu_online_mask)) { spin_unlock(&desc->lock); diff --git a/drivers/xen/events.c b/drivers/xen/events.c index eb0dfdeaa949..e0767ff35d6c 100644 --- a/drivers/xen/events.c +++ b/drivers/xen/events.c @@ -125,7 +125,7 @@ static void bind_evtchn_to_cpu(unsigned int chn, unsigned int cpu) BUG_ON(irq == -1); #ifdef CONFIG_SMP - irq_to_desc(irq)->affinity = cpumask_of_cpu(cpu); + cpumask_copy(irq_to_desc(irq)->affinity, cpumask_of(cpu)); #endif __clear_bit(chn, cpu_evtchn_mask[cpu_evtchn[chn]]); @@ -142,7 +142,7 @@ static void init_evtchn_cpu_bindings(void) /* By default all event channels notify CPU#0. */ for_each_irq_desc(i, desc) { - desc->affinity = cpumask_of_cpu(0); + cpumask_copy(desc->affinity, cpumask_of(0)); } #endif diff --git a/include/linux/irq.h b/include/linux/irq.h index f899b502f186..fa27210f1dfd 100644 --- a/include/linux/irq.h +++ b/include/linux/irq.h @@ -182,11 +182,11 @@ struct irq_desc { unsigned int irqs_unhandled; spinlock_t lock; #ifdef CONFIG_SMP - cpumask_t affinity; + cpumask_var_t affinity; unsigned int cpu; -#endif #ifdef CONFIG_GENERIC_PENDING_IRQ - cpumask_t pending_mask; + cpumask_var_t pending_mask; +#endif #endif #ifdef CONFIG_PROC_FS struct proc_dir_entry *dir; @@ -422,4 +422,79 @@ extern int set_irq_msi(unsigned int irq, struct msi_desc *entry); #endif /* !CONFIG_S390 */ +#ifdef CONFIG_SMP +/** + * init_alloc_desc_masks - allocate cpumasks for irq_desc + * @desc: pointer to irq_desc struct + * @boot: true if need bootmem + * + * Allocates affinity and pending_mask cpumask if required. + * Returns true if successful (or not required). + * Side effect: affinity has all bits set, pending_mask has all bits clear. + */ +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, + bool boot) +{ + if (boot) { + alloc_bootmem_cpumask_var(&desc->affinity); + cpumask_setall(desc->affinity); + +#ifdef CONFIG_GENERIC_PENDING_IRQ + alloc_bootmem_cpumask_var(&desc->pending_mask); + cpumask_clear(desc->pending_mask); +#endif + return true; + } + + if (!alloc_cpumask_var_node(&desc->affinity, GFP_ATOMIC, node)) + return false; + cpumask_setall(desc->affinity); + +#ifdef CONFIG_GENERIC_PENDING_IRQ + if (!alloc_cpumask_var_node(&desc->pending_mask, GFP_ATOMIC, node)) { + free_cpumask_var(desc->affinity); + return false; + } + cpumask_clear(desc->pending_mask); +#endif + return true; +} + +/** + * init_copy_desc_masks - copy cpumasks for irq_desc + * @old_desc: pointer to old irq_desc struct + * @new_desc: pointer to new irq_desc struct + * + * Insures affinity and pending_masks are copied to new irq_desc. + * If !CONFIG_CPUMASKS_OFFSTACK the cpumasks are embedded in the + * irq_desc struct so the copy is redundant. + */ + +static inline void init_copy_desc_masks(struct irq_desc *old_desc, + struct irq_desc *new_desc) +{ +#ifdef CONFIG_CPUMASKS_OFFSTACK + cpumask_copy(new_desc->affinity, old_desc->affinity); + +#ifdef CONFIG_GENERIC_PENDING_IRQ + cpumask_copy(new_desc->pending_mask, old_desc->pending_mask); +#endif +#endif +} + +#else /* !CONFIG_SMP */ + +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, + bool boot) +{ + return true; +} + +static inline void init_copy_desc_masks(struct irq_desc *old_desc, + struct irq_desc *new_desc) +{ +} + +#endif /* CONFIG_SMP */ + #endif /* _LINUX_IRQ_H */ diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index f63c706d25e1..c248eba98b43 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -46,7 +46,10 @@ void dynamic_irq_init(unsigned int irq) desc->irq_count = 0; desc->irqs_unhandled = 0; #ifdef CONFIG_SMP - cpumask_setall(&desc->affinity); + cpumask_setall(desc->affinity); +#ifdef CONFIG_GENERIC_PENDING_IRQ + cpumask_clear(desc->pending_mask); +#endif #endif spin_unlock_irqrestore(&desc->lock, flags); } diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c index c20db0be9173..b8fa1354f01c 100644 --- a/kernel/irq/handle.c +++ b/kernel/irq/handle.c @@ -64,9 +64,6 @@ static struct irq_desc irq_desc_init = { .handle_irq = handle_bad_irq, .depth = 1, .lock = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock), -#ifdef CONFIG_SMP - .affinity = CPU_MASK_ALL -#endif }; void init_kstat_irqs(struct irq_desc *desc, int cpu, int nr) @@ -88,6 +85,8 @@ void init_kstat_irqs(struct irq_desc *desc, int cpu, int nr) static void init_one_irq_desc(int irq, struct irq_desc *desc, int cpu) { + int node = cpu_to_node(cpu); + memcpy(desc, &irq_desc_init, sizeof(struct irq_desc)); spin_lock_init(&desc->lock); @@ -101,6 +100,10 @@ static void init_one_irq_desc(int irq, struct irq_desc *desc, int cpu) printk(KERN_ERR "can not alloc kstat_irqs\n"); BUG_ON(1); } + if (!init_alloc_desc_masks(desc, node, false)) { + printk(KERN_ERR "can not alloc irq_desc cpumasks\n"); + BUG_ON(1); + } arch_init_chip_data(desc, cpu); } @@ -119,9 +122,6 @@ static struct irq_desc irq_desc_legacy[NR_IRQS_LEGACY] __cacheline_aligned_in_sm .handle_irq = handle_bad_irq, .depth = 1, .lock = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock), -#ifdef CONFIG_SMP - .affinity = CPU_MASK_ALL -#endif } }; @@ -141,7 +141,7 @@ int __init early_irq_init(void) desc[i].irq = i; desc[i].kstat_irqs = kstat_irqs_legacy[i]; lockdep_set_class(&desc[i].lock, &irq_desc_lock_class); - + init_alloc_desc_masks(&desc[i], 0, true); irq_desc_ptrs[i] = desc + i; } @@ -188,6 +188,10 @@ struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu) printk(KERN_ERR "can not alloc irq_desc\n"); BUG_ON(1); } + if (!init_alloc_desc_masks(desc, node, false)) { + printk(KERN_ERR "can not alloc irq_desc cpumasks\n"); + BUG_ON(1); + } init_one_irq_desc(irq, desc, cpu); irq_desc_ptrs[irq] = desc; @@ -207,9 +211,6 @@ struct irq_desc irq_desc[NR_IRQS] __cacheline_aligned_in_smp = { .handle_irq = handle_bad_irq, .depth = 1, .lock = __SPIN_LOCK_UNLOCKED(irq_desc->lock), -#ifdef CONFIG_SMP - .affinity = CPU_MASK_ALL -#endif } }; @@ -222,9 +223,10 @@ int __init early_irq_init(void) desc = irq_desc; count = ARRAY_SIZE(irq_desc); - for (i = 0; i < count; i++) + for (i = 0; i < count; i++) { desc[i].irq = i; - + init_alloc_desc_masks(&desc[i], 0, true); + } return arch_early_irq_init(); } diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index cd0cd8dcb345..b98739af4558 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -98,14 +98,14 @@ int irq_set_affinity(unsigned int irq, const struct cpumask *cpumask) #ifdef CONFIG_GENERIC_PENDING_IRQ if (desc->status & IRQ_MOVE_PCNTXT || desc->status & IRQ_DISABLED) { - cpumask_copy(&desc->affinity, cpumask); + cpumask_copy(desc->affinity, cpumask); desc->chip->set_affinity(irq, cpumask); } else { desc->status |= IRQ_MOVE_PENDING; - cpumask_copy(&desc->pending_mask, cpumask); + cpumask_copy(desc->pending_mask, cpumask); } #else - cpumask_copy(&desc->affinity, cpumask); + cpumask_copy(desc->affinity, cpumask); desc->chip->set_affinity(irq, cpumask); #endif desc->status |= IRQ_AFFINITY_SET; @@ -127,16 +127,16 @@ int do_irq_select_affinity(unsigned int irq, struct irq_desc *desc) * one of the targets is online. */ if (desc->status & (IRQ_AFFINITY_SET | IRQ_NO_BALANCING)) { - if (cpumask_any_and(&desc->affinity, cpu_online_mask) + if (cpumask_any_and(desc->affinity, cpu_online_mask) < nr_cpu_ids) goto set_affinity; else desc->status &= ~IRQ_AFFINITY_SET; } - cpumask_and(&desc->affinity, cpu_online_mask, irq_default_affinity); + cpumask_and(desc->affinity, cpu_online_mask, irq_default_affinity); set_affinity: - desc->chip->set_affinity(irq, &desc->affinity); + desc->chip->set_affinity(irq, desc->affinity); return 0; } diff --git a/kernel/irq/migration.c b/kernel/irq/migration.c index bd72329e630c..e05ad9be43b7 100644 --- a/kernel/irq/migration.c +++ b/kernel/irq/migration.c @@ -18,7 +18,7 @@ void move_masked_irq(int irq) desc->status &= ~IRQ_MOVE_PENDING; - if (unlikely(cpumask_empty(&desc->pending_mask))) + if (unlikely(cpumask_empty(desc->pending_mask))) return; if (!desc->chip->set_affinity) @@ -38,13 +38,13 @@ void move_masked_irq(int irq) * For correct operation this depends on the caller * masking the irqs. */ - if (likely(cpumask_any_and(&desc->pending_mask, cpu_online_mask) + if (likely(cpumask_any_and(desc->pending_mask, cpu_online_mask) < nr_cpu_ids)) { - cpumask_and(&desc->affinity, - &desc->pending_mask, cpu_online_mask); - desc->chip->set_affinity(irq, &desc->affinity); + cpumask_and(desc->affinity, + desc->pending_mask, cpu_online_mask); + desc->chip->set_affinity(irq, desc->affinity); } - cpumask_clear(&desc->pending_mask); + cpumask_clear(desc->pending_mask); } void move_native_irq(int irq) diff --git a/kernel/irq/numa_migrate.c b/kernel/irq/numa_migrate.c index ecf765c6a77a..f001a4ea6414 100644 --- a/kernel/irq/numa_migrate.c +++ b/kernel/irq/numa_migrate.c @@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int irq, struct irq_desc *old_desc, desc->cpu = cpu; lockdep_set_class(&desc->lock, &irq_desc_lock_class); init_copy_kstat_irqs(old_desc, desc, cpu, nr_cpu_ids); + init_copy_desc_masks(old_desc, desc); arch_init_copy_chip_data(old_desc, desc, cpu); } @@ -76,11 +77,20 @@ static struct irq_desc *__real_move_irq_desc(struct irq_desc *old_desc, node = cpu_to_node(cpu); desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node); if (!desc) { - printk(KERN_ERR "irq %d: can not get new irq_desc for migration.\n", irq); + printk(KERN_ERR "irq %d: can not get new irq_desc " + "for migration.\n", irq); /* still use old one */ desc = old_desc; goto out_unlock; } + if (!init_alloc_desc_masks(desc, node, false)) { + printk(KERN_ERR "irq %d: can not get new irq_desc cpumask " + "for migration.\n", irq); + /* still use old one */ + kfree(desc); + desc = old_desc; + goto out_unlock; + } init_copy_one_irq_desc(irq, old_desc, desc, cpu); irq_desc_ptrs[irq] = desc; diff --git a/kernel/irq/proc.c b/kernel/irq/proc.c index aae3f742bcec..692363dd591f 100644 --- a/kernel/irq/proc.c +++ b/kernel/irq/proc.c @@ -20,11 +20,11 @@ static struct proc_dir_entry *root_irq_dir; static int irq_affinity_proc_show(struct seq_file *m, void *v) { struct irq_desc *desc = irq_to_desc((long)m->private); - const struct cpumask *mask = &desc->affinity; + const struct cpumask *mask = desc->affinity; #ifdef CONFIG_GENERIC_PENDING_IRQ if (desc->status & IRQ_MOVE_PENDING) - mask = &desc->pending_mask; + mask = desc->pending_mask; #endif seq_cpumask(m, mask); seq_putc(m, '\n'); -- cgit v1.2.3 From fbd59a8d1f7cf325fdb6828659f1fb76631e87b3 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Sat, 10 Jan 2009 21:58:08 -0800 Subject: cpumask: Use topology_core_cpumask()/topology_thread_cpumask() Impact: reduce stack usage, use new cpumask API. This actually uses topology_core_cpumask() and topology_thread_cpumask(), removing the only users of topology_core_siblings() and topology_thread_siblings() Signed-off-by: Rusty Russell Signed-off-by: Mike Travis Cc: linux-net-drivers@solarflare.com --- Documentation/cputopology.txt | 6 +++--- drivers/base/topology.c | 33 ++++++++++++++++----------------- drivers/net/sfc/efx.c | 4 ++-- include/linux/topology.h | 6 ++++++ 4 files changed, 27 insertions(+), 22 deletions(-) (limited to 'include') diff --git a/Documentation/cputopology.txt b/Documentation/cputopology.txt index 45932ec21cee..b41f3e58aefa 100644 --- a/Documentation/cputopology.txt +++ b/Documentation/cputopology.txt @@ -18,11 +18,11 @@ For an architecture to support this feature, it must define some of these macros in include/asm-XXX/topology.h: #define topology_physical_package_id(cpu) #define topology_core_id(cpu) -#define topology_thread_siblings(cpu) -#define topology_core_siblings(cpu) +#define topology_thread_cpumask(cpu) +#define topology_core_cpumask(cpu) The type of **_id is int. -The type of siblings is cpumask_t. +The type of siblings is (const) struct cpumask *. To be consistent on all architectures, include/linux/topology.h provides default definitions for any of the above macros that are diff --git a/drivers/base/topology.c b/drivers/base/topology.c index a778fb52b11f..bf6b13206d00 100644 --- a/drivers/base/topology.c +++ b/drivers/base/topology.c @@ -31,7 +31,10 @@ #include #include -#define define_one_ro(_name) \ +#define define_one_ro_named(_name, _func) \ +static SYSDEV_ATTR(_name, 0444, _func, NULL) + +#define define_one_ro(_name) \ static SYSDEV_ATTR(_name, 0444, show_##_name, NULL) #define define_id_show_func(name) \ @@ -42,8 +45,8 @@ static ssize_t show_##name(struct sys_device *dev, \ return sprintf(buf, "%d\n", topology_##name(cpu)); \ } -#if defined(topology_thread_siblings) || defined(topology_core_siblings) -static ssize_t show_cpumap(int type, cpumask_t *mask, char *buf) +#if defined(topology_thread_cpumask) || defined(topology_core_cpumask) +static ssize_t show_cpumap(int type, const struct cpumask *mask, char *buf) { ptrdiff_t len = PTR_ALIGN(buf + PAGE_SIZE - 1, PAGE_SIZE) - buf; int n = 0; @@ -65,7 +68,7 @@ static ssize_t show_##name(struct sys_device *dev, \ struct sysdev_attribute *attr, char *buf) \ { \ unsigned int cpu = dev->id; \ - return show_cpumap(0, &(topology_##name(cpu)), buf); \ + return show_cpumap(0, topology_##name(cpu), buf); \ } #define define_siblings_show_list(name) \ @@ -74,7 +77,7 @@ static ssize_t show_##name##_list(struct sys_device *dev, \ char *buf) \ { \ unsigned int cpu = dev->id; \ - return show_cpumap(1, &(topology_##name(cpu)), buf); \ + return show_cpumap(1, topology_##name(cpu), buf); \ } #else @@ -82,9 +85,7 @@ static ssize_t show_##name##_list(struct sys_device *dev, \ static ssize_t show_##name(struct sys_device *dev, \ struct sysdev_attribute *attr, char *buf) \ { \ - unsigned int cpu = dev->id; \ - cpumask_t mask = topology_##name(cpu); \ - return show_cpumap(0, &mask, buf); \ + return show_cpumap(0, topology_##name(dev->id), buf); \ } #define define_siblings_show_list(name) \ @@ -92,9 +93,7 @@ static ssize_t show_##name##_list(struct sys_device *dev, \ struct sysdev_attribute *attr, \ char *buf) \ { \ - unsigned int cpu = dev->id; \ - cpumask_t mask = topology_##name(cpu); \ - return show_cpumap(1, &mask, buf); \ + return show_cpumap(1, topology_##name(dev->id), buf); \ } #endif @@ -107,13 +106,13 @@ define_one_ro(physical_package_id); define_id_show_func(core_id); define_one_ro(core_id); -define_siblings_show_func(thread_siblings); -define_one_ro(thread_siblings); -define_one_ro(thread_siblings_list); +define_siblings_show_func(thread_cpumask); +define_one_ro_named(thread_siblings, show_thread_cpumask); +define_one_ro_named(thread_siblings_list, show_thread_cpumask_list); -define_siblings_show_func(core_siblings); -define_one_ro(core_siblings); -define_one_ro(core_siblings_list); +define_siblings_show_func(core_cpumask); +define_one_ro_named(core_siblings, show_core_cpumask); +define_one_ro_named(core_siblings_list, show_core_cpumask_list); static struct attribute *default_attrs[] = { &attr_physical_package_id.attr, diff --git a/drivers/net/sfc/efx.c b/drivers/net/sfc/efx.c index 7673fd92eaf5..f2e56ceee0ea 100644 --- a/drivers/net/sfc/efx.c +++ b/drivers/net/sfc/efx.c @@ -863,8 +863,8 @@ static int efx_wanted_rx_queues(void) for_each_online_cpu(cpu) { if (!cpu_isset(cpu, core_mask)) { ++count; - cpus_or(core_mask, core_mask, - topology_core_siblings(cpu)); + cpumask_or(&core_mask, &core_mask, + topology_core_cpumask(cpu)); } } diff --git a/include/linux/topology.h b/include/linux/topology.h index e632d29f0544..a16b9e06f2e5 100644 --- a/include/linux/topology.h +++ b/include/linux/topology.h @@ -193,5 +193,11 @@ int arch_update_cpu_topology(void); #ifndef topology_core_siblings #define topology_core_siblings(cpu) cpumask_of_cpu(cpu) #endif +#ifndef topology_thread_cpumask +#define topology_thread_cpumask(cpu) cpumask_of(cpu) +#endif +#ifndef topology_core_cpumask +#define topology_core_cpumask(cpu) cpumask_of(cpu) +#endif #endif /* _LINUX_TOPOLOGY_H */ -- cgit v1.2.3 From 802bf931f2688ad125b73db597ce63cc842fb27a Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:09 -0800 Subject: cpumask: fix bug in use cpumask_var_t in irq_desc Impact: fix bug where new irq_desc uses old cpumask pointers which are freed. As Yinghai pointed out, init_copy_one_irq_desc() copies the old desc to the new desc overwriting the cpumask pointers. Since the old_desc and the cpumask pointers are freed, then memory corruption will occur if these old pointers are used. Move the allocation of these pointers to after the copy. Signed-off-by: Mike Travis Cc: Yinghai Lu --- include/linux/irq.h | 9 +++++++-- kernel/irq/handle.c | 8 +------- kernel/irq/numa_migrate.c | 13 ++++++++----- 3 files changed, 16 insertions(+), 14 deletions(-) (limited to 'include') diff --git a/include/linux/irq.h b/include/linux/irq.h index fa27210f1dfd..27a67536511e 100644 --- a/include/linux/irq.h +++ b/include/linux/irq.h @@ -426,15 +426,18 @@ extern int set_irq_msi(unsigned int irq, struct msi_desc *entry); /** * init_alloc_desc_masks - allocate cpumasks for irq_desc * @desc: pointer to irq_desc struct + * @cpu: cpu which will be handling the cpumasks * @boot: true if need bootmem * * Allocates affinity and pending_mask cpumask if required. * Returns true if successful (or not required). * Side effect: affinity has all bits set, pending_mask has all bits clear. */ -static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int cpu, bool boot) { + int node; + if (boot) { alloc_bootmem_cpumask_var(&desc->affinity); cpumask_setall(desc->affinity); @@ -446,6 +449,8 @@ static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, return true; } + node = cpu_to_node(cpu); + if (!alloc_cpumask_var_node(&desc->affinity, GFP_ATOMIC, node)) return false; cpumask_setall(desc->affinity); @@ -484,7 +489,7 @@ static inline void init_copy_desc_masks(struct irq_desc *old_desc, #else /* !CONFIG_SMP */ -static inline bool init_alloc_desc_masks(struct irq_desc *desc, int node, +static inline bool init_alloc_desc_masks(struct irq_desc *desc, int cpu, bool boot) { return true; diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c index b8fa1354f01c..f01c0a30cb42 100644 --- a/kernel/irq/handle.c +++ b/kernel/irq/handle.c @@ -85,8 +85,6 @@ void init_kstat_irqs(struct irq_desc *desc, int cpu, int nr) static void init_one_irq_desc(int irq, struct irq_desc *desc, int cpu) { - int node = cpu_to_node(cpu); - memcpy(desc, &irq_desc_init, sizeof(struct irq_desc)); spin_lock_init(&desc->lock); @@ -100,7 +98,7 @@ static void init_one_irq_desc(int irq, struct irq_desc *desc, int cpu) printk(KERN_ERR "can not alloc kstat_irqs\n"); BUG_ON(1); } - if (!init_alloc_desc_masks(desc, node, false)) { + if (!init_alloc_desc_masks(desc, cpu, false)) { printk(KERN_ERR "can not alloc irq_desc cpumasks\n"); BUG_ON(1); } @@ -188,10 +186,6 @@ struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu) printk(KERN_ERR "can not alloc irq_desc\n"); BUG_ON(1); } - if (!init_alloc_desc_masks(desc, node, false)) { - printk(KERN_ERR "can not alloc irq_desc cpumasks\n"); - BUG_ON(1); - } init_one_irq_desc(irq, desc, cpu); irq_desc_ptrs[irq] = desc; diff --git a/kernel/irq/numa_migrate.c b/kernel/irq/numa_migrate.c index f001a4ea6414..666260e4c065 100644 --- a/kernel/irq/numa_migrate.c +++ b/kernel/irq/numa_migrate.c @@ -38,16 +38,22 @@ static void free_kstat_irqs(struct irq_desc *old_desc, struct irq_desc *desc) old_desc->kstat_irqs = NULL; } -static void init_copy_one_irq_desc(int irq, struct irq_desc *old_desc, +static bool init_copy_one_irq_desc(int irq, struct irq_desc *old_desc, struct irq_desc *desc, int cpu) { memcpy(desc, old_desc, sizeof(struct irq_desc)); + if (!init_alloc_desc_masks(desc, cpu, false)) { + printk(KERN_ERR "irq %d: can not get new irq_desc cpumask " + "for migration.\n", irq); + return false; + } spin_lock_init(&desc->lock); desc->cpu = cpu; lockdep_set_class(&desc->lock, &irq_desc_lock_class); init_copy_kstat_irqs(old_desc, desc, cpu, nr_cpu_ids); init_copy_desc_masks(old_desc, desc); arch_init_copy_chip_data(old_desc, desc, cpu); + return true; } static void free_one_irq_desc(struct irq_desc *old_desc, struct irq_desc *desc) @@ -83,15 +89,12 @@ static struct irq_desc *__real_move_irq_desc(struct irq_desc *old_desc, desc = old_desc; goto out_unlock; } - if (!init_alloc_desc_masks(desc, node, false)) { - printk(KERN_ERR "irq %d: can not get new irq_desc cpumask " - "for migration.\n", irq); + if (!init_copy_one_irq_desc(irq, old_desc, desc, cpu)) { /* still use old one */ kfree(desc); desc = old_desc; goto out_unlock; } - init_copy_one_irq_desc(irq, old_desc, desc, cpu); irq_desc_ptrs[irq] = desc; -- cgit v1.2.3 From 9332fccdedf8e09448f3b69b624211ae879f6c45 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 22:24:07 -0800 Subject: irq: initialize nr_irqs based on nr_cpu_ids Impact: Reduce memory usage. This is the second half of the changes to make the irq_desc_ptrs be variable sized based on nr_cpu_ids. This is done by adding a new "max_nr_irqs" macro to irq_vectors.h (and a dummy in irqnr.h) to return a max NR_IRQS value based on NR_CPUS or nr_cpu_ids. This necessitated moving the define of MAX_IO_APICS to a separate file (asm/apicnum.h) so it could be included without the baggage of the other asm/apicdef.h declarations. Signed-off-by: Mike Travis --- arch/x86/include/asm/apicdef.h | 8 ++------ arch/x86/include/asm/apicnum.h | 12 ++++++++++++ arch/x86/include/asm/irq_vectors.h | 16 +++++++++++----- include/linux/irqnr.h | 7 +++++++ kernel/irq/handle.c | 3 +++ 5 files changed, 35 insertions(+), 11 deletions(-) create mode 100644 arch/x86/include/asm/apicnum.h (limited to 'include') diff --git a/arch/x86/include/asm/apicdef.h b/arch/x86/include/asm/apicdef.h index 63134e31e8b9..1a6454ef7f6c 100644 --- a/arch/x86/include/asm/apicdef.h +++ b/arch/x86/include/asm/apicdef.h @@ -132,12 +132,8 @@ #define APIC_BASE_MSR 0x800 #define X2APIC_ENABLE (1UL << 10) -#ifdef CONFIG_X86_32 -# define MAX_IO_APICS 64 -#else -# define MAX_IO_APICS 128 -# define MAX_LOCAL_APIC 32768 -#endif +/* get MAX_IO_APICS */ +#include /* * All x86-64 systems are xAPIC compatible. diff --git a/arch/x86/include/asm/apicnum.h b/arch/x86/include/asm/apicnum.h new file mode 100644 index 000000000000..82f613c607ce --- /dev/null +++ b/arch/x86/include/asm/apicnum.h @@ -0,0 +1,12 @@ +#ifndef _ASM_X86_APICNUM_H +#define _ASM_X86_APICNUM_H + +/* define MAX_IO_APICS */ +#ifdef CONFIG_X86_32 +# define MAX_IO_APICS 64 +#else +# define MAX_IO_APICS 128 +# define MAX_LOCAL_APIC 32768 +#endif + +#endif /* _ASM_X86_APICNUM_H */ diff --git a/arch/x86/include/asm/irq_vectors.h b/arch/x86/include/asm/irq_vectors.h index f7ff65032b9d..602361ad0e74 100644 --- a/arch/x86/include/asm/irq_vectors.h +++ b/arch/x86/include/asm/irq_vectors.h @@ -105,6 +105,8 @@ #if defined(CONFIG_X86_IO_APIC) && !defined(CONFIG_X86_VOYAGER) +#include /* need MAX_IO_APICS */ + #ifndef CONFIG_SPARSE_IRQ # if NR_CPUS < MAX_IO_APICS # define NR_IRQS (NR_VECTORS + (32 * NR_CPUS)) @@ -112,11 +114,15 @@ # define NR_IRQS (NR_VECTORS + (32 * MAX_IO_APICS)) # endif #else -# if (8 * NR_CPUS) > (32 * MAX_IO_APICS) -# define NR_IRQS (NR_VECTORS + (8 * NR_CPUS)) -# else -# define NR_IRQS (NR_VECTORS + (32 * MAX_IO_APICS)) -# endif + +/* defined as a macro so nr_irqs = max_nr_irqs(nr_cpu_ids) can be used */ +# define max_nr_irqs(nr_cpus) \ + ((8 * nr_cpus) > (32 * MAX_IO_APICS) ? \ + (NR_VECTORS + (8 * NR_CPUS)) : \ + (NR_VECTORS + (32 * MAX_IO_APICS))) \ + +# define NR_IRQS max_nr_irqs(NR_CPUS) + #endif #elif defined(CONFIG_X86_VOYAGER) diff --git a/include/linux/irqnr.h b/include/linux/irqnr.h index 86af92e9e84c..de66e4e10406 100644 --- a/include/linux/irqnr.h +++ b/include/linux/irqnr.h @@ -20,11 +20,18 @@ # define for_each_irq_desc_reverse(irq, desc) \ for (irq = nr_irqs - 1; irq >= 0; irq--) + #else /* CONFIG_GENERIC_HARDIRQS */ +#include /* need possible max_nr_irqs() */ + extern int nr_irqs; extern struct irq_desc *irq_to_desc(unsigned int irq); +# ifndef max_nr_irqs +# define max_nr_irqs(nr_cpus) NR_IRQS +# endif + # define for_each_irq_desc(irq, desc) \ for (irq = 0, desc = irq_to_desc(irq); irq < nr_irqs; \ irq++, desc = irq_to_desc(irq)) \ diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c index d0b8f7e72790..ebba7a116f14 100644 --- a/kernel/irq/handle.c +++ b/kernel/irq/handle.c @@ -133,6 +133,9 @@ int __init early_irq_init(void) int legacy_count; int i; + /* initialize nr_irqs based on nr_cpu_ids */ + nr_irqs = max_nr_irqs(nr_cpu_ids); + printk(KERN_INFO "NR_IRQS:%d nr_irqs:%d\n", NR_IRQS, nr_irqs); desc = irq_desc_legacy; -- cgit v1.2.3 From 92296c6d6e908c35fca287a21af27be814af9c75 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sun, 11 Jan 2009 09:22:58 -0800 Subject: cpumask, irq: non-x86 build failures Ingo Molnar wrote: > All non-x86 architectures fail to build: > > In file included from /home/mingo/tip/include/linux/random.h:11, > from /home/mingo/tip/include/linux/stackprotector.h:6, > from /home/mingo/tip/init/main.c:17: > /home/mingo/tip/include/linux/irqnr.h:26:63: error: asm/irq_vectors.h: No such file or directory Do not include asm/irq_vectors.h in generic code - it's not available on all architectures. Signed-off-by: Ingo Molnar --- arch/x86/include/asm/apicdef.h | 8 ++++++-- include/linux/irqnr.h | 6 ------ kernel/irq/handle.c | 5 +++++ 3 files changed, 11 insertions(+), 8 deletions(-) (limited to 'include') diff --git a/arch/x86/include/asm/apicdef.h b/arch/x86/include/asm/apicdef.h index 1a6454ef7f6c..63134e31e8b9 100644 --- a/arch/x86/include/asm/apicdef.h +++ b/arch/x86/include/asm/apicdef.h @@ -132,8 +132,12 @@ #define APIC_BASE_MSR 0x800 #define X2APIC_ENABLE (1UL << 10) -/* get MAX_IO_APICS */ -#include +#ifdef CONFIG_X86_32 +# define MAX_IO_APICS 64 +#else +# define MAX_IO_APICS 128 +# define MAX_LOCAL_APIC 32768 +#endif /* * All x86-64 systems are xAPIC compatible. diff --git a/include/linux/irqnr.h b/include/linux/irqnr.h index de66e4e10406..887477bc2ab0 100644 --- a/include/linux/irqnr.h +++ b/include/linux/irqnr.h @@ -23,15 +23,9 @@ #else /* CONFIG_GENERIC_HARDIRQS */ -#include /* need possible max_nr_irqs() */ - extern int nr_irqs; extern struct irq_desc *irq_to_desc(unsigned int irq); -# ifndef max_nr_irqs -# define max_nr_irqs(nr_cpus) NR_IRQS -# endif - # define for_each_irq_desc(irq, desc) \ for (irq = 0, desc = irq_to_desc(irq); irq < nr_irqs; \ irq++, desc = irq_to_desc(irq)) \ diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c index b39f32ac8f80..04d3e46031e5 100644 --- a/kernel/irq/handle.c +++ b/kernel/irq/handle.c @@ -58,6 +58,11 @@ int nr_irqs = NR_IRQS; EXPORT_SYMBOL_GPL(nr_irqs); #ifdef CONFIG_SPARSE_IRQ + +#ifndef max_nr_irqs +#define max_nr_irqs(nr_cpus) NR_IRQS +#endif + static struct irq_desc irq_desc_init = { .irq = -1, .status = IRQ_DISABLED, -- cgit v1.2.3 From 4a046d1754ee6ebb6f399696805ed61ea0444d4c Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 12 Jan 2009 17:39:24 -0800 Subject: x86: arch_probe_nr_irqs Impact: save RAM with large NR_CPUS, get smaller nr_irqs Signed-off-by: Yinghai Lu Signed-off-by: Mike Travis --- arch/x86/include/asm/irq_vectors.h | 7 ++----- arch/x86/kernel/io_apic.c | 16 ++++++++++++++++ include/linux/interrupt.h | 1 + kernel/irq/handle.c | 9 ++------- kernel/softirq.c | 5 +++++ 5 files changed, 26 insertions(+), 12 deletions(-) (limited to 'include') diff --git a/arch/x86/include/asm/irq_vectors.h b/arch/x86/include/asm/irq_vectors.h index 602361ad0e74..a16a2ab2b429 100644 --- a/arch/x86/include/asm/irq_vectors.h +++ b/arch/x86/include/asm/irq_vectors.h @@ -115,14 +115,11 @@ # endif #else -/* defined as a macro so nr_irqs = max_nr_irqs(nr_cpu_ids) can be used */ -# define max_nr_irqs(nr_cpus) \ - ((8 * nr_cpus) > (32 * MAX_IO_APICS) ? \ +# define NR_IRQS \ + ((8 * NR_CPUS) > (32 * MAX_IO_APICS) ? \ (NR_VECTORS + (8 * NR_CPUS)) : \ (NR_VECTORS + (32 * MAX_IO_APICS))) \ -# define NR_IRQS max_nr_irqs(NR_CPUS) - #endif #elif defined(CONFIG_X86_VOYAGER) diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index ae80638012de..157986916cd1 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -3850,6 +3850,22 @@ void __init probe_nr_irqs_gsi(void) nr_irqs_gsi = nr; } +#ifdef CONFIG_SPARSE_IRQ +int __init arch_probe_nr_irqs(void) +{ + int nr; + + nr = ((8 * nr_cpu_ids) > (32 * nr_ioapics) ? + (NR_VECTORS + (8 * nr_cpu_ids)) : + (NR_VECTORS + (32 * nr_ioapics))); + + if (nr < nr_irqs && nr > nr_irqs_gsi) + nr_irqs = nr; + + return 0; +} +#endif + /* -------------------------------------------------------------------------- ACPI-based IOAPIC Configuration -------------------------------------------------------------------------- */ diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h index 9127f6b51a39..472f11765f60 100644 --- a/include/linux/interrupt.h +++ b/include/linux/interrupt.h @@ -467,6 +467,7 @@ int show_interrupts(struct seq_file *p, void *v); struct irq_desc; extern int early_irq_init(void); +extern int arch_probe_nr_irqs(void); extern int arch_early_irq_init(void); extern int arch_init_chip_data(struct irq_desc *desc, int cpu); diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c index 04d3e46031e5..375d68cd5bf0 100644 --- a/kernel/irq/handle.c +++ b/kernel/irq/handle.c @@ -59,10 +59,6 @@ EXPORT_SYMBOL_GPL(nr_irqs); #ifdef CONFIG_SPARSE_IRQ -#ifndef max_nr_irqs -#define max_nr_irqs(nr_cpus) NR_IRQS -#endif - static struct irq_desc irq_desc_init = { .irq = -1, .status = IRQ_DISABLED, @@ -137,9 +133,8 @@ int __init early_irq_init(void) int legacy_count; int i; - /* initialize nr_irqs based on nr_cpu_ids */ - nr_irqs = max_nr_irqs(nr_cpu_ids); - + /* initialize nr_irqs based on nr_cpu_ids */ + arch_probe_nr_irqs(); printk(KERN_INFO "NR_IRQS:%d nr_irqs:%d\n", NR_IRQS, nr_irqs); desc = irq_desc_legacy; diff --git a/kernel/softirq.c b/kernel/softirq.c index bdbe9de9cd8d..0365b4899a3d 100644 --- a/kernel/softirq.c +++ b/kernel/softirq.c @@ -795,6 +795,11 @@ int __init __weak early_irq_init(void) return 0; } +int __init __weak arch_probe_nr_irqs(void) +{ + return 0; +} + int __init __weak arch_early_irq_init(void) { return 0; -- cgit v1.2.3 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/head64.c | 2 ++ arch/x86/kernel/head_64.S | 24 +++++++++++++++++- arch/x86/kernel/setup_percpu.c | 2 +- arch/x86/kernel/vmlinux_64.lds.S | 17 ++++++++++++- include/asm-generic/sections.h | 2 +- include/asm-generic/vmlinux.lds.h | 51 ++++++++++++++++++++++++++++++++++----- 6 files changed, 88 insertions(+), 10 deletions(-) (limited to 'include') diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index b9a4d8c4b935..bc2900ca82c7 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -44,6 +44,8 @@ void __init x86_64_init_pda(void) { _cpu_pda = __cpu_pda; cpu_pda(0) = &_boot_cpu_pda; + cpu_pda(0)->data_offset = + (unsigned long)(__per_cpu_load - __per_cpu_start); pda_init(0); } 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 */ diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 56c63ac62b10..44845842e722 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -213,7 +213,7 @@ void __init setup_per_cpu_areas(void) } #endif per_cpu_offset(cpu) = ptr - __per_cpu_start; - memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start); + memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S index 1a614c0e6bef..f50280db0dfe 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -19,6 +19,9 @@ PHDRS { data PT_LOAD FLAGS(7); /* RWE */ user PT_LOAD FLAGS(7); /* RWE */ data.init PT_LOAD FLAGS(7); /* RWE */ +#ifdef CONFIG_SMP + percpu PT_LOAD FLAGS(7); /* RWE */ +#endif note PT_NOTE FLAGS(0); /* ___ */ } SECTIONS @@ -208,14 +211,26 @@ SECTIONS __initramfs_end = .; #endif +#ifdef CONFIG_SMP + /* + * percpu offsets are zero-based on SMP. PERCPU_VADDR() changes the + * output PHDR, so the next output section - __data_nosave - should + * switch it back to data.init. + */ + . = ALIGN(PAGE_SIZE); + PERCPU_VADDR(0, :percpu) +#else PERCPU(PAGE_SIZE) +#endif . = ALIGN(PAGE_SIZE); __init_end = .; . = ALIGN(PAGE_SIZE); __nosave_begin = .; - .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) { *(.data.nosave) } + .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) { + *(.data.nosave) + } :data.init /* switch back to data.init, see PERCPU_VADDR() above */ . = ALIGN(PAGE_SIZE); __nosave_end = .; diff --git a/include/asm-generic/sections.h b/include/asm-generic/sections.h index 79a7ff925bf8..4ce48e878530 100644 --- a/include/asm-generic/sections.h +++ b/include/asm-generic/sections.h @@ -9,7 +9,7 @@ extern char __bss_start[], __bss_stop[]; extern char __init_begin[], __init_end[]; extern char _sinittext[], _einittext[]; extern char _end[]; -extern char __per_cpu_start[], __per_cpu_end[]; +extern char __per_cpu_load[], __per_cpu_start[], __per_cpu_end[]; extern char __kprobes_text_start[], __kprobes_text_end[]; extern char __initdata_begin[], __initdata_end[]; extern char __start_rodata[], __end_rodata[]; diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index c61fab1dd2f8..fc2f55f2dcd6 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -430,12 +430,51 @@ *(.initcall7.init) \ *(.initcall7s.init) -#define PERCPU(align) \ - . = ALIGN(align); \ - VMLINUX_SYMBOL(__per_cpu_start) = .; \ - .data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) { \ +#define PERCPU_PROLOG(vaddr) \ + VMLINUX_SYMBOL(__per_cpu_load) = .; \ + .data.percpu vaddr : AT(__per_cpu_load - LOAD_OFFSET) { \ + VMLINUX_SYMBOL(__per_cpu_start) = .; + +#define PERCPU_EPILOG(phdr) \ + VMLINUX_SYMBOL(__per_cpu_end) = .; \ + } phdr \ + . = __per_cpu_load + SIZEOF(.data.percpu); + +/** + * PERCPU_VADDR - define output section for percpu area + * @vaddr: explicit base address (optional) + * @phdr: destination PHDR (optional) + * + * Macro which expands to output section for percpu area. If @vaddr + * is not blank, it specifies explicit base address and all percpu + * symbols will be offset from the given address. If blank, @vaddr + * always equals @laddr + LOAD_OFFSET. + * + * @phdr defines the output PHDR to use if not blank. Be warned that + * output PHDR is sticky. If @phdr is specified, the next output + * section in the linker script will go there too. @phdr should have + * a leading colon. + * + * This macro defines three symbols, __per_cpu_load, __per_cpu_start + * and __per_cpu_end. The first one is the vaddr of loaded percpu + * init data. __per_cpu_start equals @vaddr and __per_cpu_end is the + * end offset. + */ +#define PERCPU_VADDR(vaddr, phdr) \ + PERCPU_PROLOG(vaddr) \ *(.data.percpu.page_aligned) \ *(.data.percpu) \ *(.data.percpu.shared_aligned) \ - } \ - VMLINUX_SYMBOL(__per_cpu_end) = .; + PERCPU_EPILOG(phdr) + +/** + * PERCPU - define output section for percpu area, simple version + * @align: required alignment + * + * Align to @align and outputs output section for percpu area. This + * macro doesn't maniuplate @vaddr or @phdr and __per_cpu_load and + * __per_cpu_start will be identical. + */ +#define PERCPU(align) \ + . = ALIGN(align); \ + PERCPU_VADDR( , ) -- 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/include/asm/percpu.h | 8 +++ arch/x86/include/asm/smp.h | 2 - arch/x86/kernel/asm-offsets_64.c | 1 + arch/x86/kernel/cpu/common.c | 6 +-- arch/x86/kernel/head64.c | 8 ++- arch/x86/kernel/head_64.S | 15 ++++-- arch/x86/kernel/setup_percpu.c | 107 ++++++++++++++++---------------------- arch/x86/kernel/smpboot.c | 60 +-------------------- arch/x86/kernel/vmlinux_64.lds.S | 6 ++- arch/x86/xen/smp.c | 10 ---- include/asm-generic/vmlinux.lds.h | 25 ++++++++- 11 files changed, 104 insertions(+), 144 deletions(-) (limited to 'include') diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h index df644f3e53e6..0ed77cf33f76 100644 --- a/arch/x86/include/asm/percpu.h +++ b/arch/x86/include/asm/percpu.h @@ -1,6 +1,14 @@ #ifndef _ASM_X86_PERCPU_H #define _ASM_X86_PERCPU_H +#ifndef __ASSEMBLY__ +#ifdef CONFIG_X86_64 +extern void load_pda_offset(int cpu); +#else +static inline void load_pda_offset(int cpu) { } +#endif +#endif + #ifdef CONFIG_X86_64 #include diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h index a8cea7b09434..127415402ea1 100644 --- a/arch/x86/include/asm/smp.h +++ b/arch/x86/include/asm/smp.h @@ -19,8 +19,6 @@ #include #include -extern int __cpuinit get_local_pda(int cpu); - extern int smp_num_siblings; extern unsigned int num_processors; diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 1d41d3f1edbc..f8d1b047ef4f 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -56,6 +56,7 @@ int main(void) ENTRY(cpunumber); ENTRY(irqstackptr); ENTRY(data_offset); + DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); #undef ENTRY #ifdef CONFIG_PARAVIRT diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index c116c599326e..7041acdf5579 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -893,10 +893,8 @@ void __cpuinit pda_init(int cpu) /* Setup up data that may be needed in __get_free_pages early */ loadsegment(fs, 0); loadsegment(gs, 0); - /* Memory clobbers used to order PDA accessed */ - mb(); - wrmsrl(MSR_GS_BASE, pda); - mb(); + + load_pda_offset(cpu); pda->cpunumber = cpu; pda->irqcount = -1; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index 462d0beccb6b..1a311293f733 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -26,12 +26,18 @@ #include #include -/* boot cpu pda, referenced by head_64.S to initialize %gs for boot CPU */ +#ifndef CONFIG_SMP +/* boot cpu pda, referenced by head_64.S to initialize %gs on UP */ struct x8664_pda _boot_cpu_pda; +#endif void __init x86_64_init_pda(void) { +#ifdef CONFIG_SMP + cpu_pda(0) = (void *)__per_cpu_load; +#else cpu_pda(0) = &_boot_cpu_pda; +#endif cpu_pda(0)->data_offset = (unsigned long)(__per_cpu_load - __per_cpu_start); pda_init(0); 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) diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 73ab01b297c5..63d462802272 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -13,6 +13,7 @@ #include #include #include +#include #include #ifdef CONFIG_DEBUG_PER_CPU_MAPS @@ -65,6 +66,36 @@ static void __init setup_node_to_cpumask_map(void); static inline void setup_node_to_cpumask_map(void) { } #endif +#ifdef CONFIG_X86_64 +void __cpuinit load_pda_offset(int cpu) +{ + /* Memory clobbers used to order pda/percpu accesses */ + mb(); + wrmsrl(MSR_GS_BASE, cpu_pda(cpu)); + mb(); +} + +#endif /* CONFIG_SMP && CONFIG_X86_64 */ + +#ifdef CONFIG_X86_64 + +/* correctly size the local cpu masks */ +static void setup_cpu_local_masks(void) +{ + alloc_bootmem_cpumask_var(&cpu_initialized_mask); + alloc_bootmem_cpumask_var(&cpu_callin_mask); + alloc_bootmem_cpumask_var(&cpu_callout_mask); + alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); +} + +#else /* CONFIG_X86_32 */ + +static inline void setup_cpu_local_masks(void) +{ +} + +#endif /* CONFIG_X86_32 */ + #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA /* * Copy data used in early init routines from the initial arrays to the @@ -101,63 +132,7 @@ static void __init setup_per_cpu_maps(void) */ unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; EXPORT_SYMBOL(__per_cpu_offset); -static inline void setup_cpu_pda_map(void) { } - -#elif !defined(CONFIG_SMP) -static inline void setup_cpu_pda_map(void) { } - -#else /* CONFIG_SMP && CONFIG_X86_64 */ - -/* - * Allocate cpu_pda pointer table and array via alloc_bootmem. - */ -static void __init setup_cpu_pda_map(void) -{ - char *pda; - unsigned long size; - int cpu; - - size = roundup(sizeof(struct x8664_pda), cache_line_size()); - - /* allocate cpu_pda array and pointer table */ - { - unsigned long asize = size * (nr_cpu_ids - 1); - - pda = alloc_bootmem(asize); - } - - /* initialize pointer table to static pda's */ - for_each_possible_cpu(cpu) { - if (cpu == 0) { - /* leave boot cpu pda in place */ - continue; - } - cpu_pda(cpu) = (struct x8664_pda *)pda; - cpu_pda(cpu)->in_bootmem = 1; - pda += size; - } -} - -#endif /* CONFIG_SMP && CONFIG_X86_64 */ - -#ifdef CONFIG_X86_64 - -/* correctly size the local cpu masks */ -static void setup_cpu_local_masks(void) -{ - alloc_bootmem_cpumask_var(&cpu_initialized_mask); - alloc_bootmem_cpumask_var(&cpu_callin_mask); - alloc_bootmem_cpumask_var(&cpu_callout_mask); - alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); -} - -#else /* CONFIG_X86_32 */ - -static inline void setup_cpu_local_masks(void) -{ -} - -#endif /* CONFIG_X86_32 */ +#endif /* * Great future plan: @@ -171,9 +146,6 @@ void __init setup_per_cpu_areas(void) int cpu; unsigned long align = 1; - /* Setup cpu_pda map */ - setup_cpu_pda_map(); - /* Copy section for each CPU (we discard the original) */ old_size = PERCPU_ENOUGH_ROOM; align = max_t(unsigned long, PAGE_SIZE, align); @@ -204,8 +176,21 @@ void __init setup_per_cpu_areas(void) cpu, node, __pa(ptr)); } #endif - per_cpu_offset(cpu) = ptr - __per_cpu_start; + memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); +#ifdef CONFIG_X86_64 + cpu_pda(cpu) = (void *)ptr; + + /* + * CPU0 modified pda in the init data area, reload pda + * offset for CPU0 and clear the area for others. + */ + if (cpu == 0) + load_pda_offset(0); + else + memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu))); +#endif + per_cpu_offset(cpu) = ptr - __per_cpu_start; DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 70d846628bbf..f2f77ca494d4 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -744,52 +744,6 @@ static void __cpuinit do_fork_idle(struct work_struct *work) complete(&c_idle->done); } -#ifdef CONFIG_X86_64 - -/* __ref because it's safe to call free_bootmem when after_bootmem == 0. */ -static void __ref free_bootmem_pda(struct x8664_pda *oldpda) -{ - if (!after_bootmem) - free_bootmem((unsigned long)oldpda, sizeof(*oldpda)); -} - -/* - * Allocate node local memory for the AP pda. - * - * Must be called after the _cpu_pda pointer table is initialized. - */ -int __cpuinit get_local_pda(int cpu) -{ - struct x8664_pda *oldpda, *newpda; - unsigned long size = sizeof(struct x8664_pda); - int node = cpu_to_node(cpu); - - if (cpu_pda(cpu) && !cpu_pda(cpu)->in_bootmem) - return 0; - - oldpda = cpu_pda(cpu); - newpda = kmalloc_node(size, GFP_ATOMIC, node); - if (!newpda) { - printk(KERN_ERR "Could not allocate node local PDA " - "for CPU %d on node %d\n", cpu, node); - - if (oldpda) - return 0; /* have a usable pda */ - else - return -1; - } - - if (oldpda) { - memcpy(newpda, oldpda, size); - free_bootmem_pda(oldpda); - } - - newpda->in_bootmem = 0; - cpu_pda(cpu) = newpda; - return 0; -} -#endif /* CONFIG_X86_64 */ - static int __cpuinit do_boot_cpu(int apicid, int cpu) /* * NOTE - on most systems this is a PHYSICAL apic ID, but on multiquad @@ -807,16 +761,6 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu) }; INIT_WORK(&c_idle.work, do_fork_idle); -#ifdef CONFIG_X86_64 - /* Allocate node local memory for AP pdas */ - if (cpu > 0) { - boot_error = get_local_pda(cpu); - if (boot_error) - goto restore_state; - /* if can't get pda memory, can't start cpu */ - } -#endif - alternatives_smp_switch(1); c_idle.idle = get_idle_for_cpu(cpu); @@ -931,9 +875,7 @@ do_rest: inquire_remote_apic(apicid); } } -#ifdef CONFIG_X86_64 -restore_state: -#endif + if (boot_error) { /* Try to put things back the way they were before ... */ numa_remove_cpu(cpu); /* was set by numa_add_cpu */ diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S index f50280db0dfe..962f21f1d4d7 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -5,6 +5,7 @@ #define LOAD_OFFSET __START_KERNEL_map #include +#include #include #undef i386 /* in case the preprocessor is a 32bit one */ @@ -215,10 +216,11 @@ SECTIONS /* * percpu offsets are zero-based on SMP. PERCPU_VADDR() changes the * output PHDR, so the next output section - __data_nosave - should - * switch it back to data.init. + * switch it back to data.init. Also, pda should be at the head of + * percpu area. Preallocate it. */ . = ALIGN(PAGE_SIZE); - PERCPU_VADDR(0, :percpu) + PERCPU_VADDR_PREALLOC(0, :percpu, pda_size) #else PERCPU(PAGE_SIZE) #endif diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index c44e2069c7c7..83fa4236477d 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -283,16 +283,6 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) struct task_struct *idle = idle_task(cpu); int rc; -#ifdef CONFIG_X86_64 - /* Allocate node local memory for AP pdas */ - WARN_ON(cpu == 0); - if (cpu > 0) { - rc = get_local_pda(cpu); - if (rc) - return rc; - } -#endif - #ifdef CONFIG_X86_32 init_gdt(cpu); per_cpu(current_task, cpu) = idle; diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index fc2f55f2dcd6..e53319cf29cb 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -441,9 +441,10 @@ . = __per_cpu_load + SIZEOF(.data.percpu); /** - * PERCPU_VADDR - define output section for percpu area + * PERCPU_VADDR_PREALLOC - define output section for percpu area with prealloc * @vaddr: explicit base address (optional) * @phdr: destination PHDR (optional) + * @prealloc: the size of prealloc area * * Macro which expands to output section for percpu area. If @vaddr * is not blank, it specifies explicit base address and all percpu @@ -455,11 +456,33 @@ * section in the linker script will go there too. @phdr should have * a leading colon. * + * If @prealloc is non-zero, the specified number of bytes will be + * reserved at the start of percpu area. As the prealloc area is + * likely to break alignment, this macro puts areas in increasing + * alignment order. + * * This macro defines three symbols, __per_cpu_load, __per_cpu_start * and __per_cpu_end. The first one is the vaddr of loaded percpu * init data. __per_cpu_start equals @vaddr and __per_cpu_end is the * end offset. */ +#define PERCPU_VADDR_PREALLOC(vaddr, segment, prealloc) \ + PERCPU_PROLOG(vaddr) \ + . += prealloc; \ + *(.data.percpu) \ + *(.data.percpu.shared_aligned) \ + *(.data.percpu.page_aligned) \ + PERCPU_EPILOG(segment) + +/** + * PERCPU_VADDR - define output section for percpu area + * @vaddr: explicit base address (optional) + * @phdr: destination PHDR (optional) + * + * Macro which expands to output section for percpu area. Mostly + * identical to PERCPU_VADDR_PREALLOC(@vaddr, @phdr, 0) other than + * using slighly different layout. + */ #define PERCPU_VADDR(vaddr, phdr) \ PERCPU_PROLOG(vaddr) \ *(.data.percpu.page_aligned) \ -- cgit v1.2.3 From 6dbde3530850d4d8bfc1b6bd4006d92786a2787f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 15 Jan 2009 22:15:53 +0900 Subject: percpu: add optimized generic percpu accessors It is an optimization and a cleanup, and adds the following new generic percpu methods: percpu_read() percpu_write() percpu_add() percpu_sub() percpu_and() percpu_or() percpu_xor() and implements support for them on x86. (other architectures will fall back to a default implementation) The advantage is that for example to read a local percpu variable, instead of this sequence: return __get_cpu_var(var); ffffffff8102ca2b: 48 8b 14 fd 80 09 74 mov -0x7e8bf680(,%rdi,8),%rdx ffffffff8102ca32: 81 ffffffff8102ca33: 48 c7 c0 d8 59 00 00 mov $0x59d8,%rax ffffffff8102ca3a: 48 8b 04 10 mov (%rax,%rdx,1),%rax We can get a single instruction by using the optimized variants: return percpu_read(var); ffffffff8102ca3f: 65 48 8b 05 91 8f fd mov %gs:0x7efd8f91(%rip),%rax I also cleaned up the x86-specific APIs and made the x86 code use these new generic percpu primitives. tj: * fixed generic percpu_sub() definition as Roel Kluin pointed out * added percpu_and() for completeness's sake * made generic percpu ops atomic against preemption Signed-off-by: Ingo Molnar Signed-off-by: Tejun Heo --- arch/x86/include/asm/current.h | 2 +- arch/x86/include/asm/irq_regs_32.h | 4 +-- arch/x86/include/asm/mmu_context_32.h | 12 ++++---- arch/x86/include/asm/pda.h | 10 +++---- arch/x86/include/asm/percpu.h | 24 ++++++++-------- arch/x86/include/asm/smp.h | 2 +- arch/x86/kernel/process_32.c | 2 +- arch/x86/kernel/tlb_32.c | 10 +++---- arch/x86/mach-voyager/voyager_smp.c | 4 +-- arch/x86/xen/enlighten.c | 14 +++++----- arch/x86/xen/irq.c | 8 +++--- arch/x86/xen/mmu.c | 2 +- arch/x86/xen/multicalls.h | 2 +- arch/x86/xen/smp.c | 2 +- include/asm-generic/percpu.h | 52 +++++++++++++++++++++++++++++++++++ 15 files changed, 102 insertions(+), 48 deletions(-) (limited to 'include') diff --git a/arch/x86/include/asm/current.h b/arch/x86/include/asm/current.h index 0930b4f8d672..0728480f5c56 100644 --- a/arch/x86/include/asm/current.h +++ b/arch/x86/include/asm/current.h @@ -10,7 +10,7 @@ struct task_struct; DECLARE_PER_CPU(struct task_struct *, current_task); static __always_inline struct task_struct *get_current(void) { - return x86_read_percpu(current_task); + return percpu_read(current_task); } #else /* X86_32 */ diff --git a/arch/x86/include/asm/irq_regs_32.h b/arch/x86/include/asm/irq_regs_32.h index 86afd7473457..d7ed33ee94e9 100644 --- a/arch/x86/include/asm/irq_regs_32.h +++ b/arch/x86/include/asm/irq_regs_32.h @@ -15,7 +15,7 @@ DECLARE_PER_CPU(struct pt_regs *, irq_regs); static inline struct pt_regs *get_irq_regs(void) { - return x86_read_percpu(irq_regs); + return percpu_read(irq_regs); } static inline struct pt_regs *set_irq_regs(struct pt_regs *new_regs) @@ -23,7 +23,7 @@ static inline struct pt_regs *set_irq_regs(struct pt_regs *new_regs) struct pt_regs *old_regs; old_regs = get_irq_regs(); - x86_write_percpu(irq_regs, new_regs); + percpu_write(irq_regs, new_regs); return old_regs; } diff --git a/arch/x86/include/asm/mmu_context_32.h b/arch/x86/include/asm/mmu_context_32.h index 7e98ce1d2c0e..08b53454f831 100644 --- a/arch/x86/include/asm/mmu_context_32.h +++ b/arch/x86/include/asm/mmu_context_32.h @@ -4,8 +4,8 @@ static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk) { #ifdef CONFIG_SMP - if (x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_OK) - x86_write_percpu(cpu_tlbstate.state, TLBSTATE_LAZY); + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) + percpu_write(cpu_tlbstate.state, TLBSTATE_LAZY); #endif } @@ -19,8 +19,8 @@ static inline void switch_mm(struct mm_struct *prev, /* stop flush ipis for the previous mm */ cpu_clear(cpu, prev->cpu_vm_mask); #ifdef CONFIG_SMP - x86_write_percpu(cpu_tlbstate.state, TLBSTATE_OK); - x86_write_percpu(cpu_tlbstate.active_mm, next); + percpu_write(cpu_tlbstate.state, TLBSTATE_OK); + percpu_write(cpu_tlbstate.active_mm, next); #endif cpu_set(cpu, next->cpu_vm_mask); @@ -35,8 +35,8 @@ static inline void switch_mm(struct mm_struct *prev, } #ifdef CONFIG_SMP else { - x86_write_percpu(cpu_tlbstate.state, TLBSTATE_OK); - BUG_ON(x86_read_percpu(cpu_tlbstate.active_mm) != next); + percpu_write(cpu_tlbstate.state, TLBSTATE_OK); + BUG_ON(percpu_read(cpu_tlbstate.active_mm) != next); if (!cpu_test_and_set(cpu, next->cpu_vm_mask)) { /* We were in lazy tlb mode and leave_mm disabled diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index e3d3a081d798..47f274fe6953 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -45,11 +45,11 @@ extern void pda_init(int); #define cpu_pda(cpu) (&per_cpu(__pda, cpu)) -#define read_pda(field) x86_read_percpu(__pda.field) -#define write_pda(field, val) x86_write_percpu(__pda.field, val) -#define add_pda(field, val) x86_add_percpu(__pda.field, val) -#define sub_pda(field, val) x86_sub_percpu(__pda.field, val) -#define or_pda(field, val) x86_or_percpu(__pda.field, val) +#define read_pda(field) percpu_read(__pda.field) +#define write_pda(field, val) percpu_write(__pda.field, val) +#define add_pda(field, val) percpu_add(__pda.field, val) +#define sub_pda(field, val) percpu_sub(__pda.field, val) +#define or_pda(field, val) percpu_or(__pda.field, val) /* This is not atomic against other CPUs -- CPU preemption needs to be off */ #define test_and_clear_bit_pda(bit, field) \ diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h index 328b31a429d7..03aa4b00a1c3 100644 --- a/arch/x86/include/asm/percpu.h +++ b/arch/x86/include/asm/percpu.h @@ -40,16 +40,11 @@ #ifdef CONFIG_SMP #define __percpu_seg_str "%%"__stringify(__percpu_seg)":" -#define __my_cpu_offset x86_read_percpu(this_cpu_off) +#define __my_cpu_offset percpu_read(this_cpu_off) #else #define __percpu_seg_str #endif -#include - -/* We can use this directly for local CPU (faster). */ -DECLARE_PER_CPU(unsigned long, this_cpu_off); - /* For arch-specific code, we can use direct single-insn ops (they * don't give an lvalue though). */ extern void __bad_percpu_size(void); @@ -115,11 +110,13 @@ do { \ ret__; \ }) -#define x86_read_percpu(var) percpu_from_op("mov", per_cpu__##var) -#define x86_write_percpu(var, val) percpu_to_op("mov", per_cpu__##var, val) -#define x86_add_percpu(var, val) percpu_to_op("add", per_cpu__##var, val) -#define x86_sub_percpu(var, val) percpu_to_op("sub", per_cpu__##var, val) -#define x86_or_percpu(var, val) percpu_to_op("or", per_cpu__##var, val) +#define percpu_read(var) percpu_from_op("mov", per_cpu__##var) +#define percpu_write(var, val) percpu_to_op("mov", per_cpu__##var, val) +#define percpu_add(var, val) percpu_to_op("add", per_cpu__##var, val) +#define percpu_sub(var, val) percpu_to_op("sub", per_cpu__##var, val) +#define percpu_and(var, val) percpu_to_op("and", per_cpu__##var, val) +#define percpu_or(var, val) percpu_to_op("or", per_cpu__##var, val) +#define percpu_xor(var, val) percpu_to_op("xor", per_cpu__##var, val) /* This is not atomic against other CPUs -- CPU preemption needs to be off */ #define x86_test_and_clear_bit_percpu(bit, var) \ @@ -131,6 +128,11 @@ do { \ old__; \ }) +#include + +/* We can use this directly for local CPU (faster). */ +DECLARE_PER_CPU(unsigned long, this_cpu_off); + #ifdef CONFIG_X86_64 extern void load_pda_offset(int cpu); #else diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h index 127415402ea1..c7bbbbe65d3f 100644 --- a/arch/x86/include/asm/smp.h +++ b/arch/x86/include/asm/smp.h @@ -160,7 +160,7 @@ extern unsigned disabled_cpus __cpuinitdata; * from the initial startup. We map APIC_BASE very early in page_setup(), * so this is correct in the x86 case. */ -#define raw_smp_processor_id() (x86_read_percpu(cpu_number)) +#define raw_smp_processor_id() (percpu_read(cpu_number)) extern int safe_smp_processor_id(void); #elif defined(CONFIG_X86_64_SMP) diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index a546f55c77b4..77d546817d94 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c @@ -591,7 +591,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) if (prev->gs | next->gs) loadsegment(gs, next->gs); - x86_write_percpu(current_task, next_p); + percpu_write(current_task, next_p); return prev_p; } diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index ec53818f4e38..e65449d0f7d9 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -34,8 +34,8 @@ static DEFINE_SPINLOCK(tlbstate_lock); */ void leave_mm(int cpu) { - BUG_ON(x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_OK); - cpu_clear(cpu, x86_read_percpu(cpu_tlbstate.active_mm)->cpu_vm_mask); + BUG_ON(percpu_read(cpu_tlbstate.state) == TLBSTATE_OK); + cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); load_cr3(swapper_pg_dir); } EXPORT_SYMBOL_GPL(leave_mm); @@ -103,8 +103,8 @@ void smp_invalidate_interrupt(struct pt_regs *regs) * BUG(); */ - if (flush_mm == x86_read_percpu(cpu_tlbstate.active_mm)) { - if (x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_OK) { + if (flush_mm == percpu_read(cpu_tlbstate.active_mm)) { + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { if (flush_va == TLB_FLUSH_ALL) local_flush_tlb(); else @@ -222,7 +222,7 @@ static void do_flush_tlb_all(void *info) unsigned long cpu = smp_processor_id(); __flush_tlb_all(); - if (x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_LAZY) + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) leave_mm(cpu); } diff --git a/arch/x86/mach-voyager/voyager_smp.c b/arch/x86/mach-voyager/voyager_smp.c index 1a48368acb09..96f15b09a4c5 100644 --- a/arch/x86/mach-voyager/voyager_smp.c +++ b/arch/x86/mach-voyager/voyager_smp.c @@ -402,7 +402,7 @@ void __init find_smp_config(void) VOYAGER_SUS_IN_CONTROL_PORT); current_thread_info()->cpu = boot_cpu_id; - x86_write_percpu(cpu_number, boot_cpu_id); + percpu_write(cpu_number, boot_cpu_id); } /* @@ -1782,7 +1782,7 @@ static void __init voyager_smp_cpus_done(unsigned int max_cpus) void __init smp_setup_processor_id(void) { current_thread_info()->cpu = hard_smp_processor_id(); - x86_write_percpu(cpu_number, hard_smp_processor_id()); + percpu_write(cpu_number, hard_smp_processor_id()); } static void voyager_send_call_func(cpumask_t callmask) diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 312414ef9365..75b94139e1f2 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -695,17 +695,17 @@ static void xen_write_cr0(unsigned long cr0) static void xen_write_cr2(unsigned long cr2) { - x86_read_percpu(xen_vcpu)->arch.cr2 = cr2; + percpu_read(xen_vcpu)->arch.cr2 = cr2; } static unsigned long xen_read_cr2(void) { - return x86_read_percpu(xen_vcpu)->arch.cr2; + return percpu_read(xen_vcpu)->arch.cr2; } static unsigned long xen_read_cr2_direct(void) { - return x86_read_percpu(xen_vcpu_info.arch.cr2); + return percpu_read(xen_vcpu_info.arch.cr2); } static void xen_write_cr4(unsigned long cr4) @@ -718,12 +718,12 @@ static void xen_write_cr4(unsigned long cr4) static unsigned long xen_read_cr3(void) { - return x86_read_percpu(xen_cr3); + return percpu_read(xen_cr3); } static void set_current_cr3(void *v) { - x86_write_percpu(xen_current_cr3, (unsigned long)v); + percpu_write(xen_current_cr3, (unsigned long)v); } static void __xen_write_cr3(bool kernel, unsigned long cr3) @@ -748,7 +748,7 @@ static void __xen_write_cr3(bool kernel, unsigned long cr3) MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); if (kernel) { - x86_write_percpu(xen_cr3, cr3); + percpu_write(xen_cr3, cr3); /* Update xen_current_cr3 once the batch has actually been submitted. */ @@ -764,7 +764,7 @@ static void xen_write_cr3(unsigned long cr3) /* Update while interrupts are disabled, so its atomic with respect to ipis */ - x86_write_percpu(xen_cr3, cr3); + percpu_write(xen_cr3, cr3); __xen_write_cr3(true, cr3); diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c index bb042608c602..2e8271431e1a 100644 --- a/arch/x86/xen/irq.c +++ b/arch/x86/xen/irq.c @@ -39,7 +39,7 @@ static unsigned long xen_save_fl(void) struct vcpu_info *vcpu; unsigned long flags; - vcpu = x86_read_percpu(xen_vcpu); + vcpu = percpu_read(xen_vcpu); /* flag has opposite sense of mask */ flags = !vcpu->evtchn_upcall_mask; @@ -62,7 +62,7 @@ static void xen_restore_fl(unsigned long flags) make sure we're don't switch CPUs between getting the vcpu pointer and updating the mask. */ preempt_disable(); - vcpu = x86_read_percpu(xen_vcpu); + vcpu = percpu_read(xen_vcpu); vcpu->evtchn_upcall_mask = flags; preempt_enable_no_resched(); @@ -83,7 +83,7 @@ static void xen_irq_disable(void) make sure we're don't switch CPUs between getting the vcpu pointer and updating the mask. */ preempt_disable(); - x86_read_percpu(xen_vcpu)->evtchn_upcall_mask = 1; + percpu_read(xen_vcpu)->evtchn_upcall_mask = 1; preempt_enable_no_resched(); } @@ -96,7 +96,7 @@ static void xen_irq_enable(void) the caller is confused and is trying to re-enable interrupts on an indeterminate processor. */ - vcpu = x86_read_percpu(xen_vcpu); + vcpu = percpu_read(xen_vcpu); vcpu->evtchn_upcall_mask = 0; /* Doesn't matter if we get preempted here, because any diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 503c240e26c7..7bc7852cc5c4 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -1074,7 +1074,7 @@ static void drop_other_mm_ref(void *info) /* If this cpu still has a stale cr3 reference, then make sure it has been flushed. */ - if (x86_read_percpu(xen_current_cr3) == __pa(mm->pgd)) { + if (percpu_read(xen_current_cr3) == __pa(mm->pgd)) { load_cr3(swapper_pg_dir); arch_flush_lazy_cpu_mode(); } diff --git a/arch/x86/xen/multicalls.h b/arch/x86/xen/multicalls.h index 858938241616..e786fa7f2615 100644 --- a/arch/x86/xen/multicalls.h +++ b/arch/x86/xen/multicalls.h @@ -39,7 +39,7 @@ static inline void xen_mc_issue(unsigned mode) xen_mc_flush(); /* restore flags saved in xen_mc_batch */ - local_irq_restore(x86_read_percpu(xen_mc_irq_flags)); + local_irq_restore(percpu_read(xen_mc_irq_flags)); } /* Set up a callback to be called when the current batch is flushed */ diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 83fa4236477d..3bfd6dd0b47c 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -78,7 +78,7 @@ static __cpuinit void cpu_bringup(void) xen_setup_cpu_clockevents(); cpu_set(cpu, cpu_online_map); - x86_write_percpu(cpu_state, CPU_ONLINE); + percpu_write(cpu_state, CPU_ONLINE); wmb(); /* We can take interrupts now: we're officially "up". */ diff --git a/include/asm-generic/percpu.h b/include/asm-generic/percpu.h index b0e63c672ebd..00f45ff081a6 100644 --- a/include/asm-generic/percpu.h +++ b/include/asm-generic/percpu.h @@ -80,4 +80,56 @@ extern void setup_per_cpu_areas(void); #define DECLARE_PER_CPU(type, name) extern PER_CPU_ATTRIBUTES \ __typeof__(type) per_cpu_var(name) +/* + * Optional methods for optimized non-lvalue per-cpu variable access. + * + * @var can be a percpu variable or a field of it and its size should + * equal char, int or long. percpu_read() evaluates to a lvalue and + * all others to void. + * + * These operations are guaranteed to be atomic w.r.t. preemption. + * The generic versions use plain get/put_cpu_var(). Archs are + * encouraged to implement single-instruction alternatives which don't + * require preemption protection. + */ +#ifndef percpu_read +# define percpu_read(var) \ + ({ \ + typeof(per_cpu_var(var)) __tmp_var__; \ + __tmp_var__ = get_cpu_var(var); \ + put_cpu_var(var); \ + __tmp_var__; \ + }) +#endif + +#define __percpu_generic_to_op(var, val, op) \ +do { \ + get_cpu_var(var) op val; \ + put_cpu_var(var); \ +} while (0) + +#ifndef percpu_write +# define percpu_write(var, val) __percpu_generic_to_op(var, (val), =) +#endif + +#ifndef percpu_add +# define percpu_add(var, val) __percpu_generic_to_op(var, (val), +=) +#endif + +#ifndef percpu_sub +# define percpu_sub(var, val) __percpu_generic_to_op(var, (val), -=) +#endif + +#ifndef percpu_and +# define percpu_and(var, val) __percpu_generic_to_op(var, (val), &=) +#endif + +#ifndef percpu_or +# define percpu_or(var, val) __percpu_generic_to_op(var, (val), |=) +#endif + +#ifndef percpu_xor +# define percpu_xor(var, val) __percpu_generic_to_op(var, (val), ^=) +#endif + #endif /* _ASM_GENERIC_PERCPU_H_ */ -- cgit v1.2.3 From 145cd30bac885dffad9db9d487baad07b68a3d04 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 17 Jan 2009 14:42:50 +0900 Subject: linker script: add missing VMLINUX_SYMBOL The newly added PERCPU_*() macros define and use __per_cpu_load but VMLINUX_SYMBOL() was missing from usages causing build failures on archs where linker visible symbol is different from C symbols (e.g. blackfin). Signed-off-by: Tejun Heo --- include/asm-generic/vmlinux.lds.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index e53319cf29cb..aa6b9b1b30b5 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -432,13 +432,14 @@ #define PERCPU_PROLOG(vaddr) \ VMLINUX_SYMBOL(__per_cpu_load) = .; \ - .data.percpu vaddr : AT(__per_cpu_load - LOAD_OFFSET) { \ + .data.percpu vaddr : AT(VMLINUX_SYMBOL(__per_cpu_load) \ + - LOAD_OFFSET) { \ VMLINUX_SYMBOL(__per_cpu_start) = .; #define PERCPU_EPILOG(phdr) \ VMLINUX_SYMBOL(__per_cpu_end) = .; \ } phdr \ - . = __per_cpu_load + SIZEOF(.data.percpu); + . = VMLINUX_SYMBOL(__per_cpu_load) + SIZEOF(.data.percpu); /** * PERCPU_VADDR_PREALLOC - define output section for percpu area with prealloc -- cgit v1.2.3 From 0bd74fa8e29dcad98f7e8ffe01ec05fb3326abaf Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:27 +0900 Subject: percpu: refactor percpu.h Impact: cleanup Refactor the DEFINE_PER_CPU_* macros and add .data.percpu.first section. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- include/asm-generic/vmlinux.lds.h | 1 + include/linux/percpu.h | 41 ++++++++++++++++++++++----------------- 2 files changed, 24 insertions(+), 18 deletions(-) (limited to 'include') diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index aa6b9b1b30b5..32bbf50d3055 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -486,6 +486,7 @@ */ #define PERCPU_VADDR(vaddr, phdr) \ PERCPU_PROLOG(vaddr) \ + *(.data.percpu.first) \ *(.data.percpu.page_aligned) \ *(.data.percpu) \ *(.data.percpu.shared_aligned) \ diff --git a/include/linux/percpu.h b/include/linux/percpu.h index 9f2a3751873a..0e24202b5a4e 100644 --- a/include/linux/percpu.h +++ b/include/linux/percpu.h @@ -9,34 +9,39 @@ #include #ifdef CONFIG_SMP -#define DEFINE_PER_CPU(type, name) \ - __attribute__((__section__(".data.percpu"))) \ - PER_CPU_ATTRIBUTES __typeof__(type) per_cpu__##name +#define PER_CPU_BASE_SECTION ".data.percpu" #ifdef MODULE -#define SHARED_ALIGNED_SECTION ".data.percpu" +#define PER_CPU_SHARED_ALIGNED_SECTION "" #else -#define SHARED_ALIGNED_SECTION ".data.percpu.shared_aligned" +#define PER_CPU_SHARED_ALIGNED_SECTION ".shared_aligned" #endif +#define PER_CPU_FIRST_SECTION ".first" -#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name) \ - __attribute__((__section__(SHARED_ALIGNED_SECTION))) \ - PER_CPU_ATTRIBUTES __typeof__(type) per_cpu__##name \ - ____cacheline_aligned_in_smp +#else + +#define PER_CPU_BASE_SECTION ".data" +#define PER_CPU_SHARED_ALIGNED_SECTION "" +#define PER_CPU_FIRST_SECTION "" + +#endif -#define DEFINE_PER_CPU_PAGE_ALIGNED(type, name) \ - __attribute__((__section__(".data.percpu.page_aligned"))) \ +#define DEFINE_PER_CPU_SECTION(type, name, section) \ + __attribute__((__section__(PER_CPU_BASE_SECTION section))) \ PER_CPU_ATTRIBUTES __typeof__(type) per_cpu__##name -#else + #define DEFINE_PER_CPU(type, name) \ - PER_CPU_ATTRIBUTES __typeof__(type) per_cpu__##name + DEFINE_PER_CPU_SECTION(type, name, "") -#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name) \ - DEFINE_PER_CPU(type, name) +#define DEFINE_PER_CPU_SHARED_ALIGNED(type, name) \ + DEFINE_PER_CPU_SECTION(type, name, PER_CPU_SHARED_ALIGNED_SECTION) \ + ____cacheline_aligned_in_smp -#define DEFINE_PER_CPU_PAGE_ALIGNED(type, name) \ - DEFINE_PER_CPU(type, name) -#endif +#define DEFINE_PER_CPU_PAGE_ALIGNED(type, name) \ + DEFINE_PER_CPU_SECTION(type, name, ".page_aligned") + +#define DEFINE_PER_CPU_FIRST(type, name) \ + DEFINE_PER_CPU_SECTION(type, name, PER_CPU_FIRST_SECTION) #define EXPORT_PER_CPU_SYMBOL(var) EXPORT_SYMBOL(per_cpu__##var) #define EXPORT_PER_CPU_SYMBOL_GPL(var) EXPORT_SYMBOL_GPL(per_cpu__##var) -- cgit v1.2.3 From 6b7c38d55587f43bcd2cbce3a98b1c0826982090 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 19 Jan 2009 12:21:28 +0900 Subject: linker script: kill PERCPU_VADDR_PREALLOC() Impact: cleanup With .data.percpu.first in place, PERCPU_VADDR_PREALLOC() is no longer necessary. Kill it. Signed-off-by: Tejun Heo --- include/asm-generic/vmlinux.lds.h | 45 +++++++-------------------------------- 1 file changed, 8 insertions(+), 37 deletions(-) (limited to 'include') diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index 32bbf50d3055..53e21f36a802 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -430,22 +430,10 @@ *(.initcall7.init) \ *(.initcall7s.init) -#define PERCPU_PROLOG(vaddr) \ - VMLINUX_SYMBOL(__per_cpu_load) = .; \ - .data.percpu vaddr : AT(VMLINUX_SYMBOL(__per_cpu_load) \ - - LOAD_OFFSET) { \ - VMLINUX_SYMBOL(__per_cpu_start) = .; - -#define PERCPU_EPILOG(phdr) \ - VMLINUX_SYMBOL(__per_cpu_end) = .; \ - } phdr \ - . = VMLINUX_SYMBOL(__per_cpu_load) + SIZEOF(.data.percpu); - /** - * PERCPU_VADDR_PREALLOC - define output section for percpu area with prealloc + * PERCPU_VADDR - define output section for percpu area * @vaddr: explicit base address (optional) * @phdr: destination PHDR (optional) - * @prealloc: the size of prealloc area * * Macro which expands to output section for percpu area. If @vaddr * is not blank, it specifies explicit base address and all percpu @@ -457,40 +445,23 @@ * section in the linker script will go there too. @phdr should have * a leading colon. * - * If @prealloc is non-zero, the specified number of bytes will be - * reserved at the start of percpu area. As the prealloc area is - * likely to break alignment, this macro puts areas in increasing - * alignment order. - * * This macro defines three symbols, __per_cpu_load, __per_cpu_start * and __per_cpu_end. The first one is the vaddr of loaded percpu * init data. __per_cpu_start equals @vaddr and __per_cpu_end is the * end offset. */ -#define PERCPU_VADDR_PREALLOC(vaddr, segment, prealloc) \ - PERCPU_PROLOG(vaddr) \ - . += prealloc; \ - *(.data.percpu) \ - *(.data.percpu.shared_aligned) \ - *(.data.percpu.page_aligned) \ - PERCPU_EPILOG(segment) - -/** - * PERCPU_VADDR - define output section for percpu area - * @vaddr: explicit base address (optional) - * @phdr: destination PHDR (optional) - * - * Macro which expands to output section for percpu area. Mostly - * identical to PERCPU_VADDR_PREALLOC(@vaddr, @phdr, 0) other than - * using slighly different layout. - */ #define PERCPU_VADDR(vaddr, phdr) \ - PERCPU_PROLOG(vaddr) \ + VMLINUX_SYMBOL(__per_cpu_load) = .; \ + .data.percpu vaddr : AT(VMLINUX_SYMBOL(__per_cpu_load) \ + - LOAD_OFFSET) { \ + VMLINUX_SYMBOL(__per_cpu_start) = .; \ *(.data.percpu.first) \ *(.data.percpu.page_aligned) \ *(.data.percpu) \ *(.data.percpu.shared_aligned) \ - PERCPU_EPILOG(phdr) + VMLINUX_SYMBOL(__per_cpu_end) = .; \ + } phdr \ + . = VMLINUX_SYMBOL(__per_cpu_load) + SIZEOF(.data.percpu); /** * PERCPU - define output section for percpu area, simple version -- cgit v1.2.3 From 5a611268b69f05262936dd177205acbce4471358 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 26 Jan 2009 08:44:05 -0500 Subject: generic, x86: fix __per_cpu_load relocation This patch fixes this linker error: WARNING: Absolute relocations present Offset Info Type Sym.Value Sym.Name c0a4e07d 00e78001 R_386_32 c0ab0000 __per_cpu_load Now, __per_cpu_load is a section-relative symbol: c0aa4000 D __per_cpu_load c0aa4000 A __per_cpu_load_abs Signed-off-by: Brian Gerst Signed-off-by: Ingo Molnar --- include/asm-generic/vmlinux.lds.h | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index 53e21f36a802..f3180a85c66a 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -451,17 +451,18 @@ * end offset. */ #define PERCPU_VADDR(vaddr, phdr) \ - VMLINUX_SYMBOL(__per_cpu_load) = .; \ - .data.percpu vaddr : AT(VMLINUX_SYMBOL(__per_cpu_load) \ + VMLINUX_SYMBOL(__per_cpu_load_abs) = .; \ + .data.percpu vaddr : AT(VMLINUX_SYMBOL(__per_cpu_load_abs) \ - LOAD_OFFSET) { \ VMLINUX_SYMBOL(__per_cpu_start) = .; \ + VMLINUX_SYMBOL(__per_cpu_load) = LOADADDR(.data.percpu) + LOAD_OFFSET;\ *(.data.percpu.first) \ *(.data.percpu.page_aligned) \ *(.data.percpu) \ *(.data.percpu.shared_aligned) \ VMLINUX_SYMBOL(__per_cpu_end) = .; \ } phdr \ - . = VMLINUX_SYMBOL(__per_cpu_load) + SIZEOF(.data.percpu); + . = VMLINUX_SYMBOL(__per_cpu_load_abs) + SIZEOF(.data.percpu); /** * PERCPU - define output section for percpu area, simple version -- cgit v1.2.3