From 4c7f8900f1d8a0e464e7092f132a7e93f7c20f2f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sat, 23 Feb 2008 07:06:55 +0100 Subject: x86: stackprotector & PARAVIRT fix on paravirt enabled 64-bit kernels the paravirt ops do function calls themselves - which is bad with the stackprotector - for example pda_init() loads 0 into %gs and then does MSR_GS_BASE write (which modifies gs.base) - but that MSR write is a function call on paravirt, which with stackprotector tries to read the stack canary from the PDA ... crashing the bootup. the solution was suggested by Arjan van de Ven: to exclude paravirt.c from stackprotector, too many lowlevel functionality is in it. It's not like we'll have paravirt functions with character arrays on their stack anyway... Signed-off-by: Arjan van de Ven Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/kernel/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 5e618c3b472..8161e5a91ca 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -14,6 +14,7 @@ nostackp := $(call cc-option, -fno-stack-protector) CFLAGS_vsyscall_64.o := $(PROFILING) -g0 $(nostackp) CFLAGS_hpet.o := $(nostackp) CFLAGS_tsc_64.o := $(nostackp) +CFLAGS_paravirt.o := $(nostackp) obj-y := process_$(BITS).o signal_$(BITS).o entry_$(BITS).o obj-y += traps_$(BITS).o irq_$(BITS).o -- cgit v1.2.3 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 --- arch/x86/kernel/process_64.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index e2319f39988..d8640388039 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -640,7 +640,6 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) write_pda(kernelstack, (unsigned long)task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); #ifdef CONFIG_CC_STACKPROTECTOR - write_pda(stack_canary, next_p->stack_canary); /* * Build time only check to make sure the stack_canary is at * offset 40 in the pda; this is a gcc ABI requirement -- cgit v1.2.3 From ce22bd92cba0958e052cb1ce0f89f1d3a02b60a7 Mon Sep 17 00:00:00 2001 From: Arjan van de Ven Date: Mon, 12 May 2008 15:44:31 +0200 Subject: x86: setup stack canary for the idle threads The idle threads for non-boot CPUs are a bit special in how they are created; the result is that these don't have the stack canary set up properly in their PDA. Easiest fix is to just always set the PDA up correctly when entering the idle thread; this is a NOP for the boot cpu. Signed-off-by: Arjan van de Ven Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/kernel/process_64.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index d8640388039..9e69e223023 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -146,6 +146,15 @@ static inline void play_dead(void) 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. This is as good a place as any for + * doing that. + */ + write_pda(stack_canary, current->stack_canary); +#endif /* endless idle loop with no priority at all */ while (1) { tick_nohz_stop_sched_tick(); -- cgit v1.2.3 From 7e09b2a02dae4616a5a26000169964b32f86cd35 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 14 Feb 2008 09:22:34 +0100 Subject: x86: fix canary of the boot CPU's idle task the boot CPU's idle task has a zero stackprotector canary value. this is a special task that is never forked, so the fork code does not randomize its canary. Do it when we hit cpu_idle(). Academic sidenote: this means that the early init code runs with a zero canary and hence the canary becomes predictable for this short, boot-only amount of time. Although attack vectors against early init code are very rare, it might make sense to move this initialization to an earlier point. (to one of the early init functions that never return - such as start_kernel()) Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/kernel/process_64.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 9e69e223023..d4c7ac7aa43 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -150,9 +150,13 @@ void cpu_idle(void) #ifdef CONFIG_CC_STACKPROTECTOR /* * If we're the non-boot CPU, nothing set the PDA stack - * canary up for us. This is as good a place as any for - * doing that. + * 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 /* endless idle loop with no priority at all */ -- 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 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index d4c7ac7aa43..5107cb214c7 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(); -- 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 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 5107cb214c7..cce47f7fbf2 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -16,6 +16,7 @@ #include +#include #include #include #include -- cgit v1.2.3 From 6e5385d44b2df05e50a8d07ba0e14d3e32685237 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Wed, 7 Jan 2009 18:11:35 +0530 Subject: x86: smp.h move prefill_possible_map declartion to cpu.h Impact: cleanup, moving NON-SMP stuff from smp.h Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/setup.c | 2 +- arch/x86/kernel/smpboot.c | 1 - 2 files changed, 1 insertion(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index ae0d8042cf6..f41c4486c27 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -89,7 +89,7 @@ #include #include -#include +#include #include #include #include diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 07576bee03e..f8c885bed18 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -53,7 +53,6 @@ #include #include #include -#include #include #include #include -- cgit v1.2.3 From f472cdba849cc3d838f3788469316e8572463a8c Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Wed, 7 Jan 2009 21:34:25 +0530 Subject: x86: smp.h move stack_processor_id declartion to cpu.h Impact: cleanup Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 3f95a40f718..f7619a2eaff 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -21,6 +21,7 @@ #include #include #include +#include #ifdef CONFIG_X86_LOCAL_APIC #include #include -- cgit v1.2.3 From 96b89dc6598a50e3aac8e2c6d826ae3795b7d030 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Wed, 7 Jan 2009 21:35:48 +0530 Subject: x86: smp.h move safe_smp_processor_id declartion to cpu.h Impact: cleanup Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/crash.c | 2 +- arch/x86/kernel/reboot.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/crash.c b/arch/x86/kernel/crash.c index c689d19e35a..11b93cabdf7 100644 --- a/arch/x86/kernel/crash.c +++ b/arch/x86/kernel/crash.c @@ -24,7 +24,7 @@ #include #include #include -#include +#include #include #include diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c index 2b46eb41643..f8536fee5c1 100644 --- a/arch/x86/kernel/reboot.c +++ b/arch/x86/kernel/reboot.c @@ -14,6 +14,7 @@ #include #include #include +#include #ifdef CONFIG_X86_32 # include -- cgit v1.2.3 From 6d652ea1d056390a0c33db92b44ed219284b71af Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Wed, 7 Jan 2009 21:38:59 +0530 Subject: x86: smp.h move boot_cpu_id declartion to cpu.h Impact: cleanup Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/io_apic.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 1c4a1302536..109c91db202 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -46,6 +46,7 @@ #include #include #include +#include #include #include #include -- cgit v1.2.3 From 41401db698cbb5d1869776bf336881db267e7d19 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Thu, 8 Jan 2009 15:42:46 +0530 Subject: x86: rename intel_mp_floating to mpf_intel Impact: cleanup, solve 80 columns wrap problems intel_mp_floating should be renamed to mpf_intel. The reason: the 'f' in MPF already means 'floating' which means MP Floating pointer structure - no need to repeat that in the type name. Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/mpparse.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c index c0601c2848a..6cea941c4db 100644 --- a/arch/x86/kernel/mpparse.c +++ b/arch/x86/kernel/mpparse.c @@ -569,14 +569,14 @@ static inline void __init construct_default_ISA_mptable(int mpc_default_type) } } -static struct intel_mp_floating *mpf_found; +static struct mpf_intel *mpf_found; /* * Scan the memory blocks for an SMP configuration block. */ static void __init __get_smp_config(unsigned int early) { - struct intel_mp_floating *mpf = mpf_found; + struct mpf_intel *mpf = mpf_found; if (!mpf) return; @@ -687,14 +687,14 @@ static int __init smp_scan_config(unsigned long base, unsigned long length, unsigned reserve) { unsigned int *bp = phys_to_virt(base); - struct intel_mp_floating *mpf; + struct mpf_intel *mpf; apic_printk(APIC_VERBOSE, "Scan SMP from %p for %ld bytes.\n", bp, length); BUILD_BUG_ON(sizeof(*mpf) != 16); while (length > 0) { - mpf = (struct intel_mp_floating *)bp; + mpf = (struct mpf_intel *)bp; if ((*bp == SMP_MAGIC_IDENT) && (mpf->mpf_length == 1) && !mpf_checksum((unsigned char *)bp, 16) && @@ -1000,7 +1000,7 @@ static int __init update_mp_table(void) { char str[16]; char oem[10]; - struct intel_mp_floating *mpf; + struct mpf_intel *mpf; struct mpc_table *mpc, *mpc_new; if (!enable_update_mptable) @@ -1052,7 +1052,7 @@ static int __init update_mp_table(void) mpc = mpc_new; /* check if we can modify that */ if (mpc_new_phys - mpf->mpf_physptr) { - struct intel_mp_floating *mpf_new; + struct mpf_intel *mpf_new; /* steal 16 bytes from [0, 1k) */ printk(KERN_INFO "mpf new: %x\n", 0x400 - 16); mpf_new = phys_to_virt(0x400 - 16); -- cgit v1.2.3 From 1eb1b3b65dc3e3ffcc6a60e115c085c0c11c1077 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Thu, 8 Jan 2009 15:43:26 +0530 Subject: x86: rename all fields of mpf_intel mpf_X to X Impact: cleanup, solve 80 columns wrap problems It would be cleaner to rename all the mpf->mpf_X fields to mpf->X - that alone would give 4 characters per usage site. (we already know that it's an 'mpf' entity - no need to duplicate that in the field too) Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/mpparse.c | 50 +++++++++++++++++++++++------------------------ 1 file changed, 25 insertions(+), 25 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c index 6cea941c4db..8385d4e7e15 100644 --- a/arch/x86/kernel/mpparse.c +++ b/arch/x86/kernel/mpparse.c @@ -597,9 +597,9 @@ static void __init __get_smp_config(unsigned int early) } printk(KERN_INFO "Intel MultiProcessor Specification v1.%d\n", - mpf->mpf_specification); + mpf->specification); #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86_32) - if (mpf->mpf_feature2 & (1 << 7)) { + if (mpf->feature2 & (1 << 7)) { printk(KERN_INFO " IMCR and PIC compatibility mode.\n"); pic_mode = 1; } else { @@ -610,7 +610,7 @@ static void __init __get_smp_config(unsigned int early) /* * Now see if we need to read further. */ - if (mpf->mpf_feature1 != 0) { + if (mpf->feature1 != 0) { if (early) { /* * local APIC has default address @@ -620,16 +620,16 @@ static void __init __get_smp_config(unsigned int early) } printk(KERN_INFO "Default MP configuration #%d\n", - mpf->mpf_feature1); - construct_default_ISA_mptable(mpf->mpf_feature1); + mpf->feature1); + construct_default_ISA_mptable(mpf->feature1); - } else if (mpf->mpf_physptr) { + } else if (mpf->physptr) { /* * Read the physical hardware table. Anything here will * override the defaults. */ - if (!smp_read_mpc(phys_to_virt(mpf->mpf_physptr), early)) { + if (!smp_read_mpc(phys_to_virt(mpf->physptr), early)) { #ifdef CONFIG_X86_LOCAL_APIC smp_found_config = 0; #endif @@ -696,10 +696,10 @@ static int __init smp_scan_config(unsigned long base, unsigned long length, while (length > 0) { mpf = (struct mpf_intel *)bp; if ((*bp == SMP_MAGIC_IDENT) && - (mpf->mpf_length == 1) && + (mpf->length == 1) && !mpf_checksum((unsigned char *)bp, 16) && - ((mpf->mpf_specification == 1) - || (mpf->mpf_specification == 4))) { + ((mpf->specification == 1) + || (mpf->specification == 4))) { #ifdef CONFIG_X86_LOCAL_APIC smp_found_config = 1; #endif @@ -712,7 +712,7 @@ static int __init smp_scan_config(unsigned long base, unsigned long length, return 1; reserve_bootmem_generic(virt_to_phys(mpf), PAGE_SIZE, BOOTMEM_DEFAULT); - if (mpf->mpf_physptr) { + if (mpf->physptr) { unsigned long size = PAGE_SIZE; #ifdef CONFIG_X86_32 /* @@ -721,14 +721,14 @@ static int __init smp_scan_config(unsigned long base, unsigned long length, * the bottom is mapped now. * PC-9800's MPC table places on the very last * of physical memory; so that simply reserving - * PAGE_SIZE from mpg->mpf_physptr yields BUG() + * PAGE_SIZE from mpf->physptr yields BUG() * in reserve_bootmem. */ unsigned long end = max_low_pfn * PAGE_SIZE; - if (mpf->mpf_physptr + size > end) - size = end - mpf->mpf_physptr; + if (mpf->physptr + size > end) + size = end - mpf->physptr; #endif - reserve_bootmem_generic(mpf->mpf_physptr, size, + reserve_bootmem_generic(mpf->physptr, size, BOOTMEM_DEFAULT); } @@ -1013,19 +1013,19 @@ static int __init update_mp_table(void) /* * Now see if we need to go further. */ - if (mpf->mpf_feature1 != 0) + if (mpf->feature1 != 0) return 0; - if (!mpf->mpf_physptr) + if (!mpf->physptr) return 0; - mpc = phys_to_virt(mpf->mpf_physptr); + mpc = phys_to_virt(mpf->physptr); if (!smp_check_mpc(mpc, oem, str)) return 0; printk(KERN_INFO "mpf: %lx\n", virt_to_phys(mpf)); - printk(KERN_INFO "mpf_physptr: %x\n", mpf->mpf_physptr); + printk(KERN_INFO "physptr: %x\n", mpf->physptr); if (mpc_new_phys && mpc->length > mpc_new_length) { mpc_new_phys = 0; @@ -1046,23 +1046,23 @@ static int __init update_mp_table(void) } printk(KERN_INFO "use in-positon replacing\n"); } else { - mpf->mpf_physptr = mpc_new_phys; + mpf->physptr = mpc_new_phys; mpc_new = phys_to_virt(mpc_new_phys); memcpy(mpc_new, mpc, mpc->length); mpc = mpc_new; /* check if we can modify that */ - if (mpc_new_phys - mpf->mpf_physptr) { + if (mpc_new_phys - mpf->physptr) { struct mpf_intel *mpf_new; /* steal 16 bytes from [0, 1k) */ printk(KERN_INFO "mpf new: %x\n", 0x400 - 16); mpf_new = phys_to_virt(0x400 - 16); memcpy(mpf_new, mpf, 16); mpf = mpf_new; - mpf->mpf_physptr = mpc_new_phys; + mpf->physptr = mpc_new_phys; } - mpf->mpf_checksum = 0; - mpf->mpf_checksum -= mpf_checksum((unsigned char *)mpf, 16); - printk(KERN_INFO "mpf_physptr new: %x\n", mpf->mpf_physptr); + mpf->checksum = 0; + mpf->checksum -= mpf_checksum((unsigned char *)mpf, 16); + printk(KERN_INFO "physptr new: %x\n", mpf->physptr); } /* -- cgit v1.2.3 From 068790334cececc3d2d945617ccc585477da2e38 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Sat, 10 Jan 2009 12:17:37 +0530 Subject: x86: smp.h move cpu_callin_mask and cpu_callin_map declartion to cpumask.h Impact: cleanup Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common.c | 1 + arch/x86/kernel/setup_percpu.c | 1 + arch/x86/kernel/smpboot.c | 1 + 3 files changed, 3 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 14e543b6fd4..f0025846244 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -22,6 +22,7 @@ #include #include #include +#include #ifdef CONFIG_X86_LOCAL_APIC #include #include diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 55c46074eba..bf63de72b64 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -13,6 +13,7 @@ #include #include #include +#include #ifdef CONFIG_X86_LOCAL_APIC unsigned int num_processors; diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 6c2b8444b83..84ac1cf46d8 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -55,6 +55,7 @@ #include #include #include +#include #include #include #include -- cgit v1.2.3 From fb8fd077fbf0de6662acfd240e8e6b25cf3202ca Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Sat, 10 Jan 2009 12:20:24 +0530 Subject: x86: smp.h move cpu_callout_mask and cpu_callout_map declartion to cpumask.h Impact: cleanup Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/smpboot.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 84ac1cf46d8..6c2b8444b83 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -55,7 +55,6 @@ #include #include #include -#include #include #include #include -- 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 +- 3 files changed, 12 insertions(+), 12 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 1c4a1302536..1337eab60ec 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 74b9ff7341e..e0f29be8ab0 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 63c88e6ec02..0b21cb1ea11 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); -- cgit v1.2.3 From 4595f9620cda8a1e973588e743cf5f8436dd20c6 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Sat, 10 Jan 2009 21:58:09 -0800 Subject: x86: change flush_tlb_others to take a const struct cpumask Impact: reduce stack usage, use new cpumask API. This is made a little more tricky by uv_flush_tlb_others which actually alters its argument, for an IPI to be sent to the remaining cpus in the mask. I solve this by allocating a cpumask_var_t for this case and falling back to IPI should this fail. To eliminate temporaries in the caller, all flush_tlb_others implementations now do the this-cpu-elimination step themselves. Note also the curious "cpus_or(f->flush_cpumask, cpumask, f->flush_cpumask)" which has been there since pre-git and yet f->flush_cpumask is always zero at this point. Signed-off-by: Rusty Russell Signed-off-by: Mike Travis --- arch/x86/kernel/tlb_32.c | 67 +++++++++++++++++++++--------------------------- arch/x86/kernel/tlb_64.c | 61 +++++++++++++++++++++++-------------------- arch/x86/kernel/tlb_uv.c | 16 ++++++------ 3 files changed, 70 insertions(+), 74 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index ce505464224..ec53818f4e3 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -20,7 +20,7 @@ DEFINE_PER_CPU(struct tlb_state, cpu_tlbstate) * Optimizations Manfred Spraul */ -static cpumask_t flush_cpumask; +static cpumask_var_t flush_cpumask; static struct mm_struct *flush_mm; static unsigned long flush_va; static DEFINE_SPINLOCK(tlbstate_lock); @@ -92,7 +92,7 @@ void smp_invalidate_interrupt(struct pt_regs *regs) cpu = get_cpu(); - if (!cpu_isset(cpu, flush_cpumask)) + if (!cpumask_test_cpu(cpu, flush_cpumask)) goto out; /* * This was a BUG() but until someone can quote me the @@ -114,35 +114,22 @@ void smp_invalidate_interrupt(struct pt_regs *regs) } ack_APIC_irq(); smp_mb__before_clear_bit(); - cpu_clear(cpu, flush_cpumask); + cpumask_clear_cpu(cpu, flush_cpumask); smp_mb__after_clear_bit(); out: put_cpu_no_resched(); inc_irq_stat(irq_tlb_count); } -void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, - unsigned long va) +void native_flush_tlb_others(const struct cpumask *cpumask, + struct mm_struct *mm, unsigned long va) { - cpumask_t cpumask = *cpumaskp; - /* - * A couple of (to be removed) sanity checks: - * - * - current CPU must not be in mask * - mask must exist :) */ - BUG_ON(cpus_empty(cpumask)); - BUG_ON(cpu_isset(smp_processor_id(), cpumask)); + BUG_ON(cpumask_empty(cpumask)); BUG_ON(!mm); -#ifdef CONFIG_HOTPLUG_CPU - /* If a CPU which we ran on has gone down, OK. */ - cpus_and(cpumask, cpumask, cpu_online_map); - if (unlikely(cpus_empty(cpumask))) - return; -#endif - /* * i'm not happy about this global shared spinlock in the * MM hot path, but we'll see how contended it is. @@ -150,9 +137,17 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, */ spin_lock(&tlbstate_lock); + cpumask_andnot(flush_cpumask, cpumask, cpumask_of(smp_processor_id())); +#ifdef CONFIG_HOTPLUG_CPU + /* If a CPU which we ran on has gone down, OK. */ + cpumask_and(flush_cpumask, flush_cpumask, cpu_online_mask); + if (unlikely(cpumask_empty(flush_cpumask))) { + spin_unlock(&tlbstate_lock); + return; + } +#endif flush_mm = mm; flush_va = va; - cpus_or(flush_cpumask, cpumask, flush_cpumask); /* * Make the above memory operations globally visible before @@ -163,9 +158,9 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, * We have to send the IPI only to * CPUs affected. */ - send_IPI_mask(&cpumask, INVALIDATE_TLB_VECTOR); + send_IPI_mask(flush_cpumask, INVALIDATE_TLB_VECTOR); - while (!cpus_empty(flush_cpumask)) + while (!cpumask_empty(flush_cpumask)) /* nothing. lockup detection does not belong here */ cpu_relax(); @@ -177,25 +172,19 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, void flush_tlb_current_task(void) { struct mm_struct *mm = current->mm; - cpumask_t cpu_mask; preempt_disable(); - cpu_mask = mm->cpu_vm_mask; - cpu_clear(smp_processor_id(), cpu_mask); local_flush_tlb(); - if (!cpus_empty(cpu_mask)) - flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL); + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); preempt_enable(); } void flush_tlb_mm(struct mm_struct *mm) { - cpumask_t cpu_mask; preempt_disable(); - cpu_mask = mm->cpu_vm_mask; - cpu_clear(smp_processor_id(), cpu_mask); if (current->active_mm == mm) { if (current->mm) @@ -203,8 +192,8 @@ void flush_tlb_mm(struct mm_struct *mm) else leave_mm(smp_processor_id()); } - if (!cpus_empty(cpu_mask)) - flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL); + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); preempt_enable(); } @@ -212,11 +201,8 @@ void flush_tlb_mm(struct mm_struct *mm) void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) { struct mm_struct *mm = vma->vm_mm; - cpumask_t cpu_mask; preempt_disable(); - cpu_mask = mm->cpu_vm_mask; - cpu_clear(smp_processor_id(), cpu_mask); if (current->active_mm == mm) { if (current->mm) @@ -225,9 +211,8 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) leave_mm(smp_processor_id()); } - if (!cpus_empty(cpu_mask)) - flush_tlb_others(cpu_mask, mm, va); - + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, va); preempt_enable(); } EXPORT_SYMBOL(flush_tlb_page); @@ -254,3 +239,9 @@ void reset_lazy_tlbstate(void) per_cpu(cpu_tlbstate, cpu).active_mm = &init_mm; } +static int init_flush_cpumask(void) +{ + alloc_cpumask_var(&flush_cpumask, GFP_KERNEL); + return 0; +} +early_initcall(init_flush_cpumask); diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index f8be6f1d2e4..38836aef51b 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -43,10 +43,10 @@ union smp_flush_state { struct { - cpumask_t flush_cpumask; struct mm_struct *flush_mm; unsigned long flush_va; spinlock_t tlbstate_lock; + DECLARE_BITMAP(flush_cpumask, NR_CPUS); }; char pad[SMP_CACHE_BYTES]; } ____cacheline_aligned; @@ -131,7 +131,7 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START; f = &per_cpu(flush_state, sender); - if (!cpu_isset(cpu, f->flush_cpumask)) + if (!cpumask_test_cpu(cpu, to_cpumask(f->flush_cpumask))) goto out; /* * This was a BUG() but until someone can quote me the @@ -153,19 +153,15 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) } out: ack_APIC_irq(); - cpu_clear(cpu, f->flush_cpumask); + cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask)); inc_irq_stat(irq_tlb_count); } -void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, - unsigned long va) +static void flush_tlb_others_ipi(const struct cpumask *cpumask, + struct mm_struct *mm, unsigned long va) { int sender; union smp_flush_state *f; - cpumask_t cpumask = *cpumaskp; - - if (is_uv_system() && uv_flush_tlb_others(&cpumask, mm, va)) - return; /* Caller has disabled preemption */ sender = smp_processor_id() % NUM_INVALIDATE_TLB_VECTORS; @@ -180,7 +176,8 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, f->flush_mm = mm; f->flush_va = va; - cpus_or(f->flush_cpumask, cpumask, f->flush_cpumask); + cpumask_andnot(to_cpumask(f->flush_cpumask), + cpumask, cpumask_of(smp_processor_id())); /* * Make the above memory operations globally visible before @@ -191,9 +188,9 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, * We have to send the IPI only to * CPUs affected. */ - send_IPI_mask(&cpumask, INVALIDATE_TLB_VECTOR_START + sender); + send_IPI_mask(cpumask, INVALIDATE_TLB_VECTOR_START + sender); - while (!cpus_empty(f->flush_cpumask)) + while (!cpumask_empty(to_cpumask(f->flush_cpumask))) cpu_relax(); f->flush_mm = NULL; @@ -201,6 +198,24 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, spin_unlock(&f->tlbstate_lock); } +void native_flush_tlb_others(const struct cpumask *cpumask, + struct mm_struct *mm, unsigned long va) +{ + if (is_uv_system()) { + cpumask_var_t after_uv_flush; + + if (alloc_cpumask_var(&after_uv_flush, GFP_ATOMIC)) { + cpumask_andnot(after_uv_flush, + cpumask, cpumask_of(smp_processor_id())); + if (!uv_flush_tlb_others(after_uv_flush, mm, va)) + flush_tlb_others_ipi(after_uv_flush, mm, va); + free_cpumask_var(after_uv_flush); + return; + } + } + flush_tlb_others_ipi(cpumask, mm, va); +} + static int __cpuinit init_smp_flush(void) { int i; @@ -215,25 +230,18 @@ core_initcall(init_smp_flush); void flush_tlb_current_task(void) { struct mm_struct *mm = current->mm; - cpumask_t cpu_mask; preempt_disable(); - cpu_mask = mm->cpu_vm_mask; - cpu_clear(smp_processor_id(), cpu_mask); local_flush_tlb(); - if (!cpus_empty(cpu_mask)) - flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL); + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); preempt_enable(); } void flush_tlb_mm(struct mm_struct *mm) { - cpumask_t cpu_mask; - preempt_disable(); - cpu_mask = mm->cpu_vm_mask; - cpu_clear(smp_processor_id(), cpu_mask); if (current->active_mm == mm) { if (current->mm) @@ -241,8 +249,8 @@ void flush_tlb_mm(struct mm_struct *mm) else leave_mm(smp_processor_id()); } - if (!cpus_empty(cpu_mask)) - flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL); + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); preempt_enable(); } @@ -250,11 +258,8 @@ void flush_tlb_mm(struct mm_struct *mm) void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) { struct mm_struct *mm = vma->vm_mm; - cpumask_t cpu_mask; preempt_disable(); - cpu_mask = mm->cpu_vm_mask; - cpu_clear(smp_processor_id(), cpu_mask); if (current->active_mm == mm) { if (current->mm) @@ -263,8 +268,8 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) leave_mm(smp_processor_id()); } - if (!cpus_empty(cpu_mask)) - flush_tlb_others(cpu_mask, mm, va); + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, va); preempt_enable(); } diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c index f885023167e..690dcf1a27d 100644 --- a/arch/x86/kernel/tlb_uv.c +++ b/arch/x86/kernel/tlb_uv.c @@ -212,11 +212,11 @@ static int uv_wait_completion(struct bau_desc *bau_desc, * The cpumaskp mask contains the cpus the broadcast was sent to. * * Returns 1 if all remote flushing was done. The mask is zeroed. - * Returns 0 if some remote flushing remains to be done. The mask is left - * unchanged. + * Returns 0 if some remote flushing remains to be done. The mask will have + * some bits still set. */ int uv_flush_send_and_wait(int cpu, int this_blade, struct bau_desc *bau_desc, - cpumask_t *cpumaskp) + struct cpumask *cpumaskp) { int completion_status = 0; int right_shift; @@ -263,13 +263,13 @@ int uv_flush_send_and_wait(int cpu, int this_blade, struct bau_desc *bau_desc, * Success, so clear the remote cpu's from the mask so we don't * use the IPI method of shootdown on them. */ - for_each_cpu_mask(bit, *cpumaskp) { + for_each_cpu(bit, cpumaskp) { blade = uv_cpu_to_blade_id(bit); if (blade == this_blade) continue; - cpu_clear(bit, *cpumaskp); + cpumask_clear_cpu(bit, cpumaskp); } - if (!cpus_empty(*cpumaskp)) + if (!cpumask_empty(cpumaskp)) return 0; return 1; } @@ -296,7 +296,7 @@ int uv_flush_send_and_wait(int cpu, int this_blade, struct bau_desc *bau_desc, * Returns 1 if all remote flushing was done. * Returns 0 if some remote flushing remains to be done. */ -int uv_flush_tlb_others(cpumask_t *cpumaskp, struct mm_struct *mm, +int uv_flush_tlb_others(struct cpumask *cpumaskp, struct mm_struct *mm, unsigned long va) { int i; @@ -315,7 +315,7 @@ int uv_flush_tlb_others(cpumask_t *cpumaskp, struct mm_struct *mm, bau_nodes_clear(&bau_desc->distribution, UV_DISTRIBUTION_SIZE); i = 0; - for_each_cpu_mask(bit, *cpumaskp) { + for_each_cpu(bit, cpumaskp) { blade = uv_cpu_to_blade_id(bit); BUG_ON(blade > (UV_DISTRIBUTION_SIZE - 1)); if (blade == this_blade) { -- cgit v1.2.3 From 0e21990ae7ee11af94f44f240b06e520cf1505d4 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:10 -0800 Subject: SGI UV cpumask: use static temp cpumask in flush_tlb Impact: Improve tlb flush performance for UV Calling alloc_cpumask_var a zillion times a second does affect performance. Replace with static cpumask. Note: when CONFIG_X86_UV is defined, this extra PER_CPU memory will be optimized out for non-UV configs as is_uv_system() will then return a constant 0. Signed-off-by: Mike Travis --- arch/x86/kernel/tlb_64.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index 38836aef51b..7a3f9891302 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -202,16 +202,17 @@ void native_flush_tlb_others(const struct cpumask *cpumask, struct mm_struct *mm, unsigned long va) { if (is_uv_system()) { - cpumask_var_t after_uv_flush; - - if (alloc_cpumask_var(&after_uv_flush, GFP_ATOMIC)) { - cpumask_andnot(after_uv_flush, - cpumask, cpumask_of(smp_processor_id())); - if (!uv_flush_tlb_others(after_uv_flush, mm, va)) - flush_tlb_others_ipi(after_uv_flush, mm, va); - free_cpumask_var(after_uv_flush); - return; - } + /* FIXME: could be an percpu_alloc'd thing */ + static DEFINE_PER_CPU(cpumask_t, flush_tlb_mask); + struct cpumask *after_uv_flush = &get_cpu_var(flush_tlb_mask); + + cpumask_andnot(after_uv_flush, cpumask, + cpumask_of(smp_processor_id())); + if (!uv_flush_tlb_others(after_uv_flush, mm, va)) + flush_tlb_others_ipi(after_uv_flush, mm, va); + + put_cpu_var(flush_tlb_uv_cpumask); + return; } flush_tlb_others_ipi(cpumask, mm, va); } -- cgit v1.2.3 From a1c33bbeb7061f3ed39103c385844474eaa8f921 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:10 -0800 Subject: x86: cleanup remaining cpumask_t code in mce_amd_64.c Impact: Reduce memory usage, use new cpumask API. Use cpumask_var_t for 'cpus' cpumask in struct threshold_bank and update remaining old cpumask_t functions to new cpumask API. Signed-off-by: Mike Travis --- arch/x86/kernel/cpu/mcheck/mce_amd_64.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c index 8ae8c4ff094..4772e91e824 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c +++ b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c @@ -67,7 +67,7 @@ static struct threshold_block threshold_defaults = { struct threshold_bank { struct kobject *kobj; struct threshold_block *blocks; - cpumask_t cpus; + cpumask_var_t cpus; }; static DEFINE_PER_CPU(struct threshold_bank *, threshold_banks[NR_BANKS]); @@ -481,7 +481,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) #ifdef CONFIG_SMP if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */ - i = first_cpu(per_cpu(cpu_core_map, cpu)); + i = cpumask_first(&per_cpu(cpu_core_map, cpu)); /* first core not up yet */ if (cpu_data(i).cpu_core_id) @@ -501,7 +501,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) if (err) goto out; - b->cpus = per_cpu(cpu_core_map, cpu); + cpumask_copy(b->cpus, &per_cpu(cpu_core_map, cpu)); per_cpu(threshold_banks, cpu)[bank] = b; goto out; } @@ -512,15 +512,20 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) err = -ENOMEM; goto out; } + if (!alloc_cpumask_var(&b->cpus, GFP_KERNEL)) { + kfree(b); + err = -ENOMEM; + goto out; + } b->kobj = kobject_create_and_add(name, &per_cpu(device_mce, cpu).kobj); if (!b->kobj) goto out_free; #ifndef CONFIG_SMP - b->cpus = CPU_MASK_ALL; + cpumask_setall(b->cpus); #else - b->cpus = per_cpu(cpu_core_map, cpu); + cpumask_copy(b->cpus, &per_cpu(cpu_core_map, cpu)); #endif per_cpu(threshold_banks, cpu)[bank] = b; @@ -529,7 +534,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) if (err) goto out_free; - for_each_cpu_mask_nr(i, b->cpus) { + for_each_cpu(i, b->cpus) { if (i == cpu) continue; @@ -545,6 +550,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) out_free: per_cpu(threshold_banks, cpu)[bank] = NULL; + free_cpumask_var(b->cpus); kfree(b); out: return err; @@ -619,7 +625,7 @@ static void threshold_remove_bank(unsigned int cpu, int bank) #endif /* remove all sibling symlinks before unregistering */ - for_each_cpu_mask_nr(i, b->cpus) { + for_each_cpu(i, b->cpus) { if (i == cpu) continue; @@ -632,6 +638,7 @@ static void threshold_remove_bank(unsigned int cpu, int bank) free_out: kobject_del(b->kobj); kobject_put(b->kobj); + free_cpumask_var(b->cpus); kfree(b); per_cpu(threshold_banks, cpu)[bank] = NULL; } -- cgit v1.2.3 From f9b90566cd46e19f670a1e60a717ff243f060a8a Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 21:58:10 -0800 Subject: x86: reduce stack usage in init_intel_cacheinfo Impact: reduce stack usage. init_intel_cacheinfo() does not use the cpumask so define a subset of struct _cpuid4_info (_cpuid4_info_regs) that can be used instead. Signed-off-by: Mike Travis --- arch/x86/kernel/cpu/intel_cacheinfo.c | 63 ++++++++++++++++++++++++----------- 1 file changed, 44 insertions(+), 19 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c index 48533d77be7..58527a9fc40 100644 --- a/arch/x86/kernel/cpu/intel_cacheinfo.c +++ b/arch/x86/kernel/cpu/intel_cacheinfo.c @@ -132,7 +132,16 @@ struct _cpuid4_info { union _cpuid4_leaf_ecx ecx; unsigned long size; unsigned long can_disable; - cpumask_t shared_cpu_map; /* future?: only cpus/node is needed */ + DECLARE_BITMAP(shared_cpu_map, NR_CPUS); +}; + +/* subset of above _cpuid4_info w/o shared_cpu_map */ +struct _cpuid4_info_regs { + union _cpuid4_leaf_eax eax; + union _cpuid4_leaf_ebx ebx; + union _cpuid4_leaf_ecx ecx; + unsigned long size; + unsigned long can_disable; }; #ifdef CONFIG_PCI @@ -263,7 +272,7 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax, } static void __cpuinit -amd_check_l3_disable(int index, struct _cpuid4_info *this_leaf) +amd_check_l3_disable(int index, struct _cpuid4_info_regs *this_leaf) { if (index < 3) return; @@ -271,7 +280,8 @@ amd_check_l3_disable(int index, struct _cpuid4_info *this_leaf) } static int -__cpuinit cpuid4_cache_lookup(int index, struct _cpuid4_info *this_leaf) +__cpuinit cpuid4_cache_lookup_regs(int index, + struct _cpuid4_info_regs *this_leaf) { union _cpuid4_leaf_eax eax; union _cpuid4_leaf_ebx ebx; @@ -299,6 +309,15 @@ __cpuinit cpuid4_cache_lookup(int index, struct _cpuid4_info *this_leaf) return 0; } +static int +__cpuinit cpuid4_cache_lookup(int index, struct _cpuid4_info *this_leaf) +{ + struct _cpuid4_info_regs *leaf_regs = + (struct _cpuid4_info_regs *)this_leaf; + + return cpuid4_cache_lookup_regs(index, leaf_regs); +} + static int __cpuinit find_num_cache_leaves(void) { unsigned int eax, ebx, ecx, edx; @@ -338,11 +357,10 @@ unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c) * parameters cpuid leaf to find the cache details */ for (i = 0; i < num_cache_leaves; i++) { - struct _cpuid4_info this_leaf; - + struct _cpuid4_info_regs this_leaf; int retval; - retval = cpuid4_cache_lookup(i, &this_leaf); + retval = cpuid4_cache_lookup_regs(i, &this_leaf); if (retval >= 0) { switch(this_leaf.eax.split.level) { case 1: @@ -491,17 +509,20 @@ static void __cpuinit cache_shared_cpu_map_setup(unsigned int cpu, int index) num_threads_sharing = 1 + this_leaf->eax.split.num_threads_sharing; if (num_threads_sharing == 1) - cpu_set(cpu, this_leaf->shared_cpu_map); + cpumask_set_cpu(cpu, to_cpumask(this_leaf->shared_cpu_map)); else { index_msb = get_count_order(num_threads_sharing); for_each_online_cpu(i) { if (cpu_data(i).apicid >> index_msb == c->apicid >> index_msb) { - cpu_set(i, this_leaf->shared_cpu_map); + cpumask_set_cpu(i, + to_cpumask(this_leaf->shared_cpu_map)); if (i != cpu && per_cpu(cpuid4_info, i)) { - sibling_leaf = CPUID4_INFO_IDX(i, index); - cpu_set(cpu, sibling_leaf->shared_cpu_map); + sibling_leaf = + CPUID4_INFO_IDX(i, index); + cpumask_set_cpu(cpu, to_cpumask( + sibling_leaf->shared_cpu_map)); } } } @@ -513,9 +534,10 @@ static void __cpuinit cache_remove_shared_cpu_map(unsigned int cpu, int index) int sibling; this_leaf = CPUID4_INFO_IDX(cpu, index); - for_each_cpu_mask_nr(sibling, this_leaf->shared_cpu_map) { + for_each_cpu(sibling, to_cpumask(this_leaf->shared_cpu_map)) { sibling_leaf = CPUID4_INFO_IDX(sibling, index); - cpu_clear(cpu, sibling_leaf->shared_cpu_map); + cpumask_clear_cpu(cpu, + to_cpumask(sibling_leaf->shared_cpu_map)); } } #else @@ -620,8 +642,9 @@ static ssize_t show_shared_cpu_map_func(struct _cpuid4_info *this_leaf, int n = 0; if (len > 1) { - cpumask_t *mask = &this_leaf->shared_cpu_map; + const struct cpumask *mask; + mask = to_cpumask(this_leaf->shared_cpu_map); n = type? cpulist_scnprintf(buf, len-2, mask) : cpumask_scnprintf(buf, len-2, mask); @@ -684,7 +707,8 @@ static struct pci_dev *get_k8_northbridge(int node) static ssize_t show_cache_disable(struct _cpuid4_info *this_leaf, char *buf) { - int node = cpu_to_node(first_cpu(this_leaf->shared_cpu_map)); + const struct cpumask *mask = to_cpumask(this_leaf->shared_cpu_map); + int node = cpu_to_node(cpumask_first(mask)); struct pci_dev *dev = NULL; ssize_t ret = 0; int i; @@ -718,7 +742,8 @@ static ssize_t store_cache_disable(struct _cpuid4_info *this_leaf, const char *buf, size_t count) { - int node = cpu_to_node(first_cpu(this_leaf->shared_cpu_map)); + const struct cpumask *mask = to_cpumask(this_leaf->shared_cpu_map); + int node = cpu_to_node(cpumask_first(mask)); struct pci_dev *dev = NULL; unsigned int ret, index, val; @@ -863,7 +888,7 @@ err_out: return -ENOMEM; } -static cpumask_t cache_dev_map = CPU_MASK_NONE; +static DECLARE_BITMAP(cache_dev_map, NR_CPUS); /* Add/Remove cache interface for CPU device */ static int __cpuinit cache_add_dev(struct sys_device * sys_dev) @@ -903,7 +928,7 @@ static int __cpuinit cache_add_dev(struct sys_device * sys_dev) } kobject_uevent(&(this_object->kobj), KOBJ_ADD); } - cpu_set(cpu, cache_dev_map); + cpumask_set_cpu(cpu, to_cpumask(cache_dev_map)); kobject_uevent(per_cpu(cache_kobject, cpu), KOBJ_ADD); return 0; @@ -916,9 +941,9 @@ static void __cpuinit cache_remove_dev(struct sys_device * sys_dev) if (per_cpu(cpuid4_info, cpu) == NULL) return; - if (!cpu_isset(cpu, cache_dev_map)) + if (!cpumask_test_cpu(cpu, to_cpumask(cache_dev_map))) return; - cpu_clear(cpu, cache_dev_map); + cpumask_clear_cpu(cpu, to_cpumask(cache_dev_map)); for (i = 0; i < num_cache_leaves; i++) kobject_put(&(INDEX_KOBJECT_PTR(cpu,i)->kobj)); -- cgit v1.2.3 From 9594949b060efe86ecaa1a66839232a3b9800bc9 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Sat, 10 Jan 2009 22:24:06 -0800 Subject: irq: change references from NR_IRQS to nr_irqs Impact: preparation, cleanup, add KERN_INFO printk Modify references from NR_IRQS to nr_irqs as the later will become variable-sized based on nr_cpu_ids when CONFIG_SPARSE_IRQS=y. Signed-off-by: Mike Travis --- arch/x86/kernel/io_apic.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 1337eab60ec..ae80638012d 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -3183,7 +3183,7 @@ unsigned int create_irq_nr(unsigned int irq_want) irq = 0; spin_lock_irqsave(&vector_lock, flags); - for (new = irq_want; new < NR_IRQS; new++) { + for (new = irq_want; new < nr_irqs; new++) { if (platform_legacy_irq(new)) continue; -- cgit v1.2.3 From dd3feda7748b4c2739de47daaaa387fb01926c15 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Mon, 12 Jan 2009 14:44:29 +0530 Subject: x86: microcode_intel.c fix style problems Impact: cleanup Fix: WARNING: Use #include instead of ERROR: trailing whitespace ERROR: "(foo*)" should be "(foo *)" total: 3 errors, 1 warnings Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/microcode_intel.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/microcode_intel.c b/arch/x86/kernel/microcode_intel.c index b7f4c929e61..5e9f4fc5138 100644 --- a/arch/x86/kernel/microcode_intel.c +++ b/arch/x86/kernel/microcode_intel.c @@ -87,9 +87,9 @@ #include #include #include +#include #include -#include #include #include @@ -196,7 +196,7 @@ static inline int update_match_cpu(struct cpu_signature *csig, int sig, int pf) return (!sigmatch(sig, csig->sig, pf, csig->pf)) ? 0 : 1; } -static inline int +static inline int update_match_revision(struct microcode_header_intel *mc_header, int rev) { return (mc_header->rev <= rev) ? 0 : 1; @@ -442,8 +442,8 @@ static int request_microcode_fw(int cpu, struct device *device) return ret; } - ret = generic_load_microcode(cpu, (void*)firmware->data, firmware->size, - &get_ucode_fw); + ret = generic_load_microcode(cpu, (void *)firmware->data, + firmware->size, &get_ucode_fw); release_firmware(firmware); @@ -460,7 +460,7 @@ static int request_microcode_user(int cpu, const void __user *buf, size_t size) /* We should bind the task to the CPU */ BUG_ON(cpu != raw_smp_processor_id()); - return generic_load_microcode(cpu, (void*)buf, size, &get_ucode_user); + return generic_load_microcode(cpu, (void *)buf, size, &get_ucode_user); } static void microcode_fini_cpu(int cpu) -- cgit v1.2.3 From 448dd2fa3ec915ad4325868ef8bb9b9490d9f6a9 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Mon, 12 Jan 2009 14:45:14 +0530 Subject: x86: msr.c fix style problems Impact: cleanup Fix: WARNING: Use #include instead of total: 0 errors, 1 warnings Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/msr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/msr.c b/arch/x86/kernel/msr.c index 726266695b2..3cf3413ec62 100644 --- a/arch/x86/kernel/msr.c +++ b/arch/x86/kernel/msr.c @@ -35,10 +35,10 @@ #include #include #include +#include #include #include -#include #include static struct class *msr_class; -- cgit v1.2.3 From e17029ad69c7b1518413c00f793d89bb77b8527b Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Mon, 12 Jan 2009 14:46:03 +0530 Subject: x86: module_32.c fix style problems Impact: cleanup Fix: ERROR: code indent should use tabs where possible ERROR: trailing whitespace ERROR: spaces required around that '=' (ctx:VxW) total: 3 errors, 0 warnings Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/module_32.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/module_32.c b/arch/x86/kernel/module_32.c index 3db0a5442eb..0edd819050e 100644 --- a/arch/x86/kernel/module_32.c +++ b/arch/x86/kernel/module_32.c @@ -42,7 +42,7 @@ void module_free(struct module *mod, void *module_region) { vfree(module_region); /* FIXME: If module_region == mod->init_region, trim exception - table entries. */ + table entries. */ } /* We don't need anything special. */ @@ -113,13 +113,13 @@ int module_finalize(const Elf_Ehdr *hdr, *para = NULL; char *secstrings = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset; - for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) { + for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) { if (!strcmp(".text", secstrings + s->sh_name)) text = s; if (!strcmp(".altinstructions", secstrings + s->sh_name)) alt = s; if (!strcmp(".smp_locks", secstrings + s->sh_name)) - locks= s; + locks = s; if (!strcmp(".parainstructions", secstrings + s->sh_name)) para = s; } -- cgit v1.2.3 From 3b9dc9f2f123286aaade54c45fef6723d587c664 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Mon, 12 Jan 2009 14:46:48 +0530 Subject: x86: module_64.c fix style problems Impact: cleanup Fix: ERROR: trailing whitespace ERROR: code indent should use tabs where possible WARNING: %Ld/%Lu are not-standard C, use %lld/%llu WARNING: printk() should include KERN_ facility level ERROR: spaces required around that '=' (ctx:VxW) total: 13 errors, 2 warnings Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/module_64.c | 32 ++++++++++++++++---------------- 1 file changed, 16 insertions(+), 16 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/module_64.c b/arch/x86/kernel/module_64.c index 6ba87830d4b..c23880b90b5 100644 --- a/arch/x86/kernel/module_64.c +++ b/arch/x86/kernel/module_64.c @@ -30,14 +30,14 @@ #include #include -#define DEBUGP(fmt...) +#define DEBUGP(fmt...) #ifndef CONFIG_UML void module_free(struct module *mod, void *module_region) { vfree(module_region); /* FIXME: If module_region == mod->init_region, trim exception - table entries. */ + table entries. */ } void *module_alloc(unsigned long size) @@ -77,7 +77,7 @@ int apply_relocate_add(Elf64_Shdr *sechdrs, Elf64_Rela *rel = (void *)sechdrs[relsec].sh_addr; Elf64_Sym *sym; void *loc; - u64 val; + u64 val; DEBUGP("Applying relocate section %u to %u\n", relsec, sechdrs[relsec].sh_info); @@ -91,11 +91,11 @@ int apply_relocate_add(Elf64_Shdr *sechdrs, sym = (Elf64_Sym *)sechdrs[symindex].sh_addr + ELF64_R_SYM(rel[i].r_info); - DEBUGP("type %d st_value %Lx r_addend %Lx loc %Lx\n", - (int)ELF64_R_TYPE(rel[i].r_info), - sym->st_value, rel[i].r_addend, (u64)loc); + DEBUGP("type %d st_value %Lx r_addend %Lx loc %Lx\n", + (int)ELF64_R_TYPE(rel[i].r_info), + sym->st_value, rel[i].r_addend, (u64)loc); - val = sym->st_value + rel[i].r_addend; + val = sym->st_value + rel[i].r_addend; switch (ELF64_R_TYPE(rel[i].r_info)) { case R_X86_64_NONE: @@ -113,16 +113,16 @@ int apply_relocate_add(Elf64_Shdr *sechdrs, if ((s64)val != *(s32 *)loc) goto overflow; break; - case R_X86_64_PC32: + case R_X86_64_PC32: val -= (u64)loc; *(u32 *)loc = val; #if 0 if ((s64)val != *(s32 *)loc) - goto overflow; + goto overflow; #endif break; default: - printk(KERN_ERR "module %s: Unknown rela relocation: %Lu\n", + printk(KERN_ERR "module %s: Unknown rela relocation: %llu\n", me->name, ELF64_R_TYPE(rel[i].r_info)); return -ENOEXEC; } @@ -130,7 +130,7 @@ int apply_relocate_add(Elf64_Shdr *sechdrs, return 0; overflow: - printk(KERN_ERR "overflow in relocation type %d val %Lx\n", + printk(KERN_ERR "overflow in relocation type %d val %Lx\n", (int)ELF64_R_TYPE(rel[i].r_info), val); printk(KERN_ERR "`%s' likely not compiled with -mcmodel=kernel\n", me->name); @@ -143,13 +143,13 @@ int apply_relocate(Elf_Shdr *sechdrs, unsigned int relsec, struct module *me) { - printk("non add relocation not supported\n"); + printk(KERN_ERR "non add relocation not supported\n"); return -ENOSYS; -} +} int module_finalize(const Elf_Ehdr *hdr, - const Elf_Shdr *sechdrs, - struct module *me) + const Elf_Shdr *sechdrs, + struct module *me) { const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL, *para = NULL; @@ -161,7 +161,7 @@ int module_finalize(const Elf_Ehdr *hdr, if (!strcmp(".altinstructions", secstrings + s->sh_name)) alt = s; if (!strcmp(".smp_locks", secstrings + s->sh_name)) - locks= s; + locks = s; if (!strcmp(".parainstructions", secstrings + s->sh_name)) para = s; } -- 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/kernel/io_apic.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index ae80638012d..157986916cd 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 -------------------------------------------------------------------------- */ -- cgit v1.2.3 From a4a0acf8e17e3d08e28b721ceceb898fbc959ceb Mon Sep 17 00:00:00 2001 From: Suresh Siddha Date: Tue, 13 Jan 2009 18:43:03 -0800 Subject: x86: fix broken flush_tlb_others_ipi() This commit broke flush_tlb_others_ipi() causing boot hangs on a 16 logical cpu system: > commit 4595f9620cda8a1e973588e743cf5f8436dd20c6 > Author: Rusty Russell > Date: Sat Jan 10 21:58:09 2009 -0800 > > x86: change flush_tlb_others to take a const struct cpumask This change resulted in sending the invalidate tlb vector to the sender itself causing the hang. flush_tlb_others_ipi() should exclude the sender itself from the destination list. Signed-off-by: Suresh Siddha Signed-off-by: Ingo Molnar --- arch/x86/kernel/tlb_64.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index 7a3f9891302..54ee2ecb5e2 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -188,7 +188,7 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask, * We have to send the IPI only to * CPUs affected. */ - send_IPI_mask(cpumask, INVALIDATE_TLB_VECTOR_START + sender); + send_IPI_mask(f->flush_cpumask, INVALIDATE_TLB_VECTOR_START + sender); while (!cpumask_empty(to_cpumask(f->flush_cpumask))) cpu_relax(); -- cgit v1.2.3 From b5ba7e6d1e7e2ac808afd21be1e56dc34caf20e6 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Mon, 12 Jan 2009 17:46:17 +0530 Subject: x86: replacing mp_config_ioapic with mpc_ioapic Impact: cleanup, solve 80 columns wrap problems Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/acpi/boot.c | 28 ++++++++++----------- arch/x86/kernel/io_apic.c | 60 ++++++++++++++++++++++----------------------- arch/x86/kernel/mpparse.c | 12 ++++----- 3 files changed, 49 insertions(+), 51 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c index d37593c2f43..2b27019e64f 100644 --- a/arch/x86/kernel/acpi/boot.c +++ b/arch/x86/kernel/acpi/boot.c @@ -912,8 +912,8 @@ static u8 __init uniq_ioapic_id(u8 id) DECLARE_BITMAP(used, 256); bitmap_zero(used, 256); for (i = 0; i < nr_ioapics; i++) { - struct mp_config_ioapic *ia = &mp_ioapics[i]; - __set_bit(ia->mp_apicid, used); + struct mpc_ioapic *ia = &mp_ioapics[i]; + __set_bit(ia->apicid, used); } if (!test_bit(id, used)) return id; @@ -945,29 +945,29 @@ void __init mp_register_ioapic(int id, u32 address, u32 gsi_base) idx = nr_ioapics; - mp_ioapics[idx].mp_type = MP_IOAPIC; - mp_ioapics[idx].mp_flags = MPC_APIC_USABLE; - mp_ioapics[idx].mp_apicaddr = address; + mp_ioapics[idx].type = MP_IOAPIC; + mp_ioapics[idx].flags = MPC_APIC_USABLE; + mp_ioapics[idx].apicaddr = address; set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address); - mp_ioapics[idx].mp_apicid = uniq_ioapic_id(id); + mp_ioapics[idx].apicid = uniq_ioapic_id(id); #ifdef CONFIG_X86_32 - mp_ioapics[idx].mp_apicver = io_apic_get_version(idx); + mp_ioapics[idx].apicver = io_apic_get_version(idx); #else - mp_ioapics[idx].mp_apicver = 0; + mp_ioapics[idx].apicver = 0; #endif /* * Build basic GSI lookup table to facilitate gsi->io_apic lookups * and to prevent reprogramming of IOAPIC pins (PCI GSIs). */ - mp_ioapic_routing[idx].apic_id = mp_ioapics[idx].mp_apicid; + mp_ioapic_routing[idx].apic_id = mp_ioapics[idx].apicid; mp_ioapic_routing[idx].gsi_base = gsi_base; mp_ioapic_routing[idx].gsi_end = gsi_base + io_apic_get_redir_entries(idx); - printk(KERN_INFO "IOAPIC[%d]: apic_id %d, version %d, address 0x%lx, " - "GSI %d-%d\n", idx, mp_ioapics[idx].mp_apicid, - mp_ioapics[idx].mp_apicver, mp_ioapics[idx].mp_apicaddr, + printk(KERN_INFO "IOAPIC[%d]: apic_id %d, version %d, address 0x%x, " + "GSI %d-%d\n", idx, mp_ioapics[idx].apicid, + mp_ioapics[idx].apicver, mp_ioapics[idx].apicaddr, mp_ioapic_routing[idx].gsi_base, mp_ioapic_routing[idx].gsi_end); nr_ioapics++; @@ -1026,7 +1026,7 @@ void __init mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi) mp_irq.mp_irqflag = (trigger << 2) | polarity; mp_irq.mp_srcbus = MP_ISA_BUS; mp_irq.mp_srcbusirq = bus_irq; /* IRQ */ - mp_irq.mp_dstapic = mp_ioapics[ioapic].mp_apicid; /* APIC ID */ + mp_irq.mp_dstapic = mp_ioapics[ioapic].apicid; /* APIC ID */ mp_irq.mp_dstirq = pin; /* INTIN# */ save_mp_irq(&mp_irq); @@ -1062,7 +1062,7 @@ void __init mp_config_acpi_legacy_irqs(void) ioapic = mp_find_ioapic(0); if (ioapic < 0) return; - dstapic = mp_ioapics[ioapic].mp_apicid; + dstapic = mp_ioapics[ioapic].apicid; /* * Use the default configuration for the IRQs 0-15. Unless diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 109c91db202..6c51ecdfbf4 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -83,7 +83,7 @@ static DEFINE_SPINLOCK(vector_lock); int nr_ioapic_registers[MAX_IO_APICS]; /* I/O APIC entries */ -struct mp_config_ioapic mp_ioapics[MAX_IO_APICS]; +struct mpc_ioapic mp_ioapics[MAX_IO_APICS]; int nr_ioapics; /* MP IRQ source entries */ @@ -387,7 +387,7 @@ struct io_apic { static __attribute_const__ struct io_apic __iomem *io_apic_base(int idx) { return (void __iomem *) __fix_to_virt(FIX_IO_APIC_BASE_0 + idx) - + (mp_ioapics[idx].mp_apicaddr & ~PAGE_MASK); + + (mp_ioapics[idx].apicaddr & ~PAGE_MASK); } static inline unsigned int io_apic_read(unsigned int apic, unsigned int reg) @@ -946,7 +946,7 @@ static int find_irq_entry(int apic, int pin, int type) for (i = 0; i < mp_irq_entries; i++) if (mp_irqs[i].mp_irqtype == type && - (mp_irqs[i].mp_dstapic == mp_ioapics[apic].mp_apicid || + (mp_irqs[i].mp_dstapic == mp_ioapics[apic].apicid || mp_irqs[i].mp_dstapic == MP_APIC_ALL) && mp_irqs[i].mp_dstirq == pin) return i; @@ -988,7 +988,7 @@ static int __init find_isa_irq_apic(int irq, int type) if (i < mp_irq_entries) { int apic; for(apic = 0; apic < nr_ioapics; apic++) { - if (mp_ioapics[apic].mp_apicid == mp_irqs[i].mp_dstapic) + if (mp_ioapics[apic].apicid == mp_irqs[i].mp_dstapic) return apic; } } @@ -1016,7 +1016,7 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin) int lbus = mp_irqs[i].mp_srcbus; for (apic = 0; apic < nr_ioapics; apic++) - if (mp_ioapics[apic].mp_apicid == mp_irqs[i].mp_dstapic || + if (mp_ioapics[apic].apicid == mp_irqs[i].mp_dstapic || mp_irqs[i].mp_dstapic == MP_APIC_ALL) break; @@ -1567,14 +1567,14 @@ static void setup_IO_APIC_irq(int apic, int pin, unsigned int irq, struct irq_de apic_printk(APIC_VERBOSE,KERN_DEBUG "IOAPIC[%d]: Set routing entry (%d-%d -> 0x%x -> " "IRQ %d Mode:%i Active:%i)\n", - apic, mp_ioapics[apic].mp_apicid, pin, cfg->vector, + apic, mp_ioapics[apic].apicid, pin, cfg->vector, irq, trigger, polarity); - if (setup_ioapic_entry(mp_ioapics[apic].mp_apicid, irq, &entry, + if (setup_ioapic_entry(mp_ioapics[apic].apicid, irq, &entry, dest, trigger, polarity, cfg->vector)) { printk("Failed to setup ioapic entry for ioapic %d, pin %d\n", - mp_ioapics[apic].mp_apicid, pin); + mp_ioapics[apic].apicid, pin); __clear_irq_vector(irq, cfg); return; } @@ -1605,12 +1605,10 @@ static void __init setup_IO_APIC_irqs(void) notcon = 1; apic_printk(APIC_VERBOSE, KERN_DEBUG " %d-%d", - mp_ioapics[apic].mp_apicid, - pin); + mp_ioapics[apic].apicid, pin); } else apic_printk(APIC_VERBOSE, " %d-%d", - mp_ioapics[apic].mp_apicid, - pin); + mp_ioapics[apic].apicid, pin); continue; } if (notcon) { @@ -1700,7 +1698,7 @@ __apicdebuginit(void) print_IO_APIC(void) printk(KERN_DEBUG "number of MP IRQ sources: %d.\n", mp_irq_entries); for (i = 0; i < nr_ioapics; i++) printk(KERN_DEBUG "number of IO-APIC #%d registers: %d.\n", - mp_ioapics[i].mp_apicid, nr_ioapic_registers[i]); + mp_ioapics[i].apicid, nr_ioapic_registers[i]); /* * We are a bit conservative about what we expect. We have to @@ -1720,7 +1718,7 @@ __apicdebuginit(void) print_IO_APIC(void) spin_unlock_irqrestore(&ioapic_lock, flags); printk("\n"); - printk(KERN_DEBUG "IO APIC #%d......\n", mp_ioapics[apic].mp_apicid); + printk(KERN_DEBUG "IO APIC #%d......\n", mp_ioapics[apic].apicid); printk(KERN_DEBUG ".... register #00: %08X\n", reg_00.raw); printk(KERN_DEBUG "....... : physical APIC id: %02X\n", reg_00.bits.ID); printk(KERN_DEBUG "....... : Delivery Type: %X\n", reg_00.bits.delivery_type); @@ -2122,14 +2120,14 @@ static void __init setup_ioapic_ids_from_mpc(void) reg_00.raw = io_apic_read(apic, 0); spin_unlock_irqrestore(&ioapic_lock, flags); - old_id = mp_ioapics[apic].mp_apicid; + old_id = mp_ioapics[apic].apicid; - if (mp_ioapics[apic].mp_apicid >= get_physical_broadcast()) { + if (mp_ioapics[apic].apicid >= get_physical_broadcast()) { printk(KERN_ERR "BIOS bug, IO-APIC#%d ID is %d in the MPC table!...\n", - apic, mp_ioapics[apic].mp_apicid); + apic, mp_ioapics[apic].apicid); printk(KERN_ERR "... fixing up to %d. (tell your hw vendor)\n", reg_00.bits.ID); - mp_ioapics[apic].mp_apicid = reg_00.bits.ID; + mp_ioapics[apic].apicid = reg_00.bits.ID; } /* @@ -2138,9 +2136,9 @@ static void __init setup_ioapic_ids_from_mpc(void) * 'stuck on smp_invalidate_needed IPI wait' messages. */ if (check_apicid_used(phys_id_present_map, - mp_ioapics[apic].mp_apicid)) { + mp_ioapics[apic].apicid)) { printk(KERN_ERR "BIOS bug, IO-APIC#%d ID %d is already used!...\n", - apic, mp_ioapics[apic].mp_apicid); + apic, mp_ioapics[apic].apicid); for (i = 0; i < get_physical_broadcast(); i++) if (!physid_isset(i, phys_id_present_map)) break; @@ -2149,13 +2147,13 @@ static void __init setup_ioapic_ids_from_mpc(void) printk(KERN_ERR "... fixing up to %d. (tell your hw vendor)\n", i); physid_set(i, phys_id_present_map); - mp_ioapics[apic].mp_apicid = i; + mp_ioapics[apic].apicid = i; } else { physid_mask_t tmp; - tmp = apicid_to_cpu_present(mp_ioapics[apic].mp_apicid); + tmp = apicid_to_cpu_present(mp_ioapics[apic].apicid); apic_printk(APIC_VERBOSE, "Setting %d in the " "phys_id_present_map\n", - mp_ioapics[apic].mp_apicid); + mp_ioapics[apic].apicid); physids_or(phys_id_present_map, phys_id_present_map, tmp); } @@ -2164,11 +2162,11 @@ static void __init setup_ioapic_ids_from_mpc(void) * We need to adjust the IRQ routing table * if the ID changed. */ - if (old_id != mp_ioapics[apic].mp_apicid) + if (old_id != mp_ioapics[apic].apicid) for (i = 0; i < mp_irq_entries; i++) if (mp_irqs[i].mp_dstapic == old_id) mp_irqs[i].mp_dstapic - = mp_ioapics[apic].mp_apicid; + = mp_ioapics[apic].apicid; /* * Read the right value from the MPC table and @@ -2176,9 +2174,9 @@ static void __init setup_ioapic_ids_from_mpc(void) */ apic_printk(APIC_VERBOSE, KERN_INFO "...changing IO-APIC physical APIC ID to %d ...", - mp_ioapics[apic].mp_apicid); + mp_ioapics[apic].apicid); - reg_00.bits.ID = mp_ioapics[apic].mp_apicid; + reg_00.bits.ID = mp_ioapics[apic].apicid; spin_lock_irqsave(&ioapic_lock, flags); io_apic_write(apic, 0, reg_00.raw); spin_unlock_irqrestore(&ioapic_lock, flags); @@ -2189,7 +2187,7 @@ static void __init setup_ioapic_ids_from_mpc(void) spin_lock_irqsave(&ioapic_lock, flags); reg_00.raw = io_apic_read(apic, 0); spin_unlock_irqrestore(&ioapic_lock, flags); - if (reg_00.bits.ID != mp_ioapics[apic].mp_apicid) + if (reg_00.bits.ID != mp_ioapics[apic].apicid) printk("could not set ID!\n"); else apic_printk(APIC_VERBOSE, " ok.\n"); @@ -3118,8 +3116,8 @@ static int ioapic_resume(struct sys_device *dev) spin_lock_irqsave(&ioapic_lock, flags); reg_00.raw = io_apic_read(dev->id, 0); - if (reg_00.bits.ID != mp_ioapics[dev->id].mp_apicid) { - reg_00.bits.ID = mp_ioapics[dev->id].mp_apicid; + if (reg_00.bits.ID != mp_ioapics[dev->id].apicid) { + reg_00.bits.ID = mp_ioapics[dev->id].apicid; io_apic_write(dev->id, 0, reg_00.raw); } spin_unlock_irqrestore(&ioapic_lock, flags); @@ -4101,7 +4099,7 @@ void __init ioapic_init_mappings(void) ioapic_res = ioapic_setup_resources(); for (i = 0; i < nr_ioapics; i++) { if (smp_found_config) { - ioapic_phys = mp_ioapics[i].mp_apicaddr; + ioapic_phys = mp_ioapics[i].apicaddr; #ifdef CONFIG_X86_32 if (!ioapic_phys) { printk(KERN_ERR diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c index 8385d4e7e15..a86a6553743 100644 --- a/arch/x86/kernel/mpparse.c +++ b/arch/x86/kernel/mpparse.c @@ -143,11 +143,11 @@ static void __init MP_ioapic_info(struct mpc_ioapic *m) if (bad_ioapic(m->apicaddr)) return; - mp_ioapics[nr_ioapics].mp_apicaddr = m->apicaddr; - mp_ioapics[nr_ioapics].mp_apicid = m->apicid; - mp_ioapics[nr_ioapics].mp_type = m->type; - mp_ioapics[nr_ioapics].mp_apicver = m->apicver; - mp_ioapics[nr_ioapics].mp_flags = m->flags; + mp_ioapics[nr_ioapics].apicaddr = m->apicaddr; + mp_ioapics[nr_ioapics].apicid = m->apicid; + mp_ioapics[nr_ioapics].type = m->type; + mp_ioapics[nr_ioapics].apicver = m->apicver; + mp_ioapics[nr_ioapics].flags = m->flags; nr_ioapics++; } @@ -416,7 +416,7 @@ static void __init construct_default_ioirq_mptable(int mpc_default_type) intsrc.type = MP_INTSRC; intsrc.irqflag = 0; /* conforming */ intsrc.srcbus = 0; - intsrc.dstapic = mp_ioapics[0].mp_apicid; + intsrc.dstapic = mp_ioapics[0].apicid; intsrc.irqtype = mp_INT; -- cgit v1.2.3 From c2c21745ecba23c74690a124bcd371f83bd71e45 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Mon, 12 Jan 2009 17:47:22 +0530 Subject: x86: replacing mp_config_intsrc with mpc_intsrc Impact: cleanup, solve 80 columns wrap problems Signed-off-by: Jaswinder Singh Rajput Signed-off-by: Ingo Molnar --- arch/x86/kernel/acpi/boot.c | 70 ++++++++++++++++++++++----------------------- arch/x86/kernel/io_apic.c | 64 ++++++++++++++++++++--------------------- arch/x86/kernel/mpparse.c | 68 +++++++++++++++++++++---------------------- 3 files changed, 100 insertions(+), 102 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c index 2b27019e64f..4cb5964f149 100644 --- a/arch/x86/kernel/acpi/boot.c +++ b/arch/x86/kernel/acpi/boot.c @@ -973,19 +973,19 @@ void __init mp_register_ioapic(int id, u32 address, u32 gsi_base) nr_ioapics++; } -static void assign_to_mp_irq(struct mp_config_intsrc *m, - struct mp_config_intsrc *mp_irq) +static void assign_to_mp_irq(struct mpc_intsrc *m, + struct mpc_intsrc *mp_irq) { - memcpy(mp_irq, m, sizeof(struct mp_config_intsrc)); + memcpy(mp_irq, m, sizeof(struct mpc_intsrc)); } -static int mp_irq_cmp(struct mp_config_intsrc *mp_irq, - struct mp_config_intsrc *m) +static int mp_irq_cmp(struct mpc_intsrc *mp_irq, + struct mpc_intsrc *m) { - return memcmp(mp_irq, m, sizeof(struct mp_config_intsrc)); + return memcmp(mp_irq, m, sizeof(struct mpc_intsrc)); } -static void save_mp_irq(struct mp_config_intsrc *m) +static void save_mp_irq(struct mpc_intsrc *m) { int i; @@ -1003,7 +1003,7 @@ void __init mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi) { int ioapic; int pin; - struct mp_config_intsrc mp_irq; + struct mpc_intsrc mp_irq; /* * Convert 'gsi' to 'ioapic.pin'. @@ -1021,13 +1021,13 @@ void __init mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi) if ((bus_irq == 0) && (trigger == 3)) trigger = 1; - mp_irq.mp_type = MP_INTSRC; - mp_irq.mp_irqtype = mp_INT; - mp_irq.mp_irqflag = (trigger << 2) | polarity; - mp_irq.mp_srcbus = MP_ISA_BUS; - mp_irq.mp_srcbusirq = bus_irq; /* IRQ */ - mp_irq.mp_dstapic = mp_ioapics[ioapic].apicid; /* APIC ID */ - mp_irq.mp_dstirq = pin; /* INTIN# */ + mp_irq.type = MP_INTSRC; + mp_irq.irqtype = mp_INT; + mp_irq.irqflag = (trigger << 2) | polarity; + mp_irq.srcbus = MP_ISA_BUS; + mp_irq.srcbusirq = bus_irq; /* IRQ */ + mp_irq.dstapic = mp_ioapics[ioapic].apicid; /* APIC ID */ + mp_irq.dstirq = pin; /* INTIN# */ save_mp_irq(&mp_irq); } @@ -1037,7 +1037,7 @@ void __init mp_config_acpi_legacy_irqs(void) int i; int ioapic; unsigned int dstapic; - struct mp_config_intsrc mp_irq; + struct mpc_intsrc mp_irq; #if defined (CONFIG_MCA) || defined (CONFIG_EISA) /* @@ -1072,16 +1072,14 @@ void __init mp_config_acpi_legacy_irqs(void) int idx; for (idx = 0; idx < mp_irq_entries; idx++) { - struct mp_config_intsrc *irq = mp_irqs + idx; + struct mpc_intsrc *irq = mp_irqs + idx; /* Do we already have a mapping for this ISA IRQ? */ - if (irq->mp_srcbus == MP_ISA_BUS - && irq->mp_srcbusirq == i) + if (irq->srcbus == MP_ISA_BUS && irq->srcbusirq == i) break; /* Do we already have a mapping for this IOAPIC pin */ - if (irq->mp_dstapic == dstapic && - irq->mp_dstirq == i) + if (irq->dstapic == dstapic && irq->dstirq == i) break; } @@ -1090,13 +1088,13 @@ void __init mp_config_acpi_legacy_irqs(void) continue; /* IRQ already used */ } - mp_irq.mp_type = MP_INTSRC; - mp_irq.mp_irqflag = 0; /* Conforming */ - mp_irq.mp_srcbus = MP_ISA_BUS; - mp_irq.mp_dstapic = dstapic; - mp_irq.mp_irqtype = mp_INT; - mp_irq.mp_srcbusirq = i; /* Identity mapped */ - mp_irq.mp_dstirq = i; + mp_irq.type = MP_INTSRC; + mp_irq.irqflag = 0; /* Conforming */ + mp_irq.srcbus = MP_ISA_BUS; + mp_irq.dstapic = dstapic; + mp_irq.irqtype = mp_INT; + mp_irq.srcbusirq = i; /* Identity mapped */ + mp_irq.dstirq = i; save_mp_irq(&mp_irq); } @@ -1207,22 +1205,22 @@ int mp_config_acpi_gsi(unsigned char number, unsigned int devfn, u8 pin, u32 gsi, int triggering, int polarity) { #ifdef CONFIG_X86_MPPARSE - struct mp_config_intsrc mp_irq; + struct mpc_intsrc mp_irq; int ioapic; if (!acpi_ioapic) return 0; /* print the entry should happen on mptable identically */ - mp_irq.mp_type = MP_INTSRC; - mp_irq.mp_irqtype = mp_INT; - mp_irq.mp_irqflag = (triggering == ACPI_EDGE_SENSITIVE ? 4 : 0x0c) | + mp_irq.type = MP_INTSRC; + mp_irq.irqtype = mp_INT; + mp_irq.irqflag = (triggering == ACPI_EDGE_SENSITIVE ? 4 : 0x0c) | (polarity == ACPI_ACTIVE_HIGH ? 1 : 3); - mp_irq.mp_srcbus = number; - mp_irq.mp_srcbusirq = (((devfn >> 3) & 0x1f) << 2) | ((pin - 1) & 3); + mp_irq.srcbus = number; + mp_irq.srcbusirq = (((devfn >> 3) & 0x1f) << 2) | ((pin - 1) & 3); ioapic = mp_find_ioapic(gsi); - mp_irq.mp_dstapic = mp_ioapic_routing[ioapic].apic_id; - mp_irq.mp_dstirq = gsi - mp_ioapic_routing[ioapic].gsi_base; + mp_irq.dstapic = mp_ioapic_routing[ioapic].apic_id; + mp_irq.dstirq = gsi - mp_ioapic_routing[ioapic].gsi_base; save_mp_irq(&mp_irq); #endif diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 6c51ecdfbf4..79b8c0c72d3 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -87,7 +87,7 @@ struct mpc_ioapic mp_ioapics[MAX_IO_APICS]; int nr_ioapics; /* MP IRQ source entries */ -struct mp_config_intsrc mp_irqs[MAX_IRQ_SOURCES]; +struct mpc_intsrc mp_irqs[MAX_IRQ_SOURCES]; /* # of MP IRQ source entries */ int mp_irq_entries; @@ -945,10 +945,10 @@ static int find_irq_entry(int apic, int pin, int type) int i; for (i = 0; i < mp_irq_entries; i++) - if (mp_irqs[i].mp_irqtype == type && - (mp_irqs[i].mp_dstapic == mp_ioapics[apic].apicid || - mp_irqs[i].mp_dstapic == MP_APIC_ALL) && - mp_irqs[i].mp_dstirq == pin) + if (mp_irqs[i].irqtype == type && + (mp_irqs[i].dstapic == mp_ioapics[apic].apicid || + mp_irqs[i].dstapic == MP_APIC_ALL) && + mp_irqs[i].dstirq == pin) return i; return -1; @@ -962,13 +962,13 @@ static int __init find_isa_irq_pin(int irq, int type) int i; for (i = 0; i < mp_irq_entries; i++) { - int lbus = mp_irqs[i].mp_srcbus; + int lbus = mp_irqs[i].srcbus; if (test_bit(lbus, mp_bus_not_pci) && - (mp_irqs[i].mp_irqtype == type) && - (mp_irqs[i].mp_srcbusirq == irq)) + (mp_irqs[i].irqtype == type) && + (mp_irqs[i].srcbusirq == irq)) - return mp_irqs[i].mp_dstirq; + return mp_irqs[i].dstirq; } return -1; } @@ -978,17 +978,17 @@ static int __init find_isa_irq_apic(int irq, int type) int i; for (i = 0; i < mp_irq_entries; i++) { - int lbus = mp_irqs[i].mp_srcbus; + int lbus = mp_irqs[i].srcbus; if (test_bit(lbus, mp_bus_not_pci) && - (mp_irqs[i].mp_irqtype == type) && - (mp_irqs[i].mp_srcbusirq == irq)) + (mp_irqs[i].irqtype == type) && + (mp_irqs[i].srcbusirq == irq)) break; } if (i < mp_irq_entries) { int apic; for(apic = 0; apic < nr_ioapics; apic++) { - if (mp_ioapics[apic].apicid == mp_irqs[i].mp_dstapic) + if (mp_ioapics[apic].apicid == mp_irqs[i].dstapic) return apic; } } @@ -1013,23 +1013,23 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin) return -1; } for (i = 0; i < mp_irq_entries; i++) { - int lbus = mp_irqs[i].mp_srcbus; + int lbus = mp_irqs[i].srcbus; for (apic = 0; apic < nr_ioapics; apic++) - if (mp_ioapics[apic].apicid == mp_irqs[i].mp_dstapic || - mp_irqs[i].mp_dstapic == MP_APIC_ALL) + if (mp_ioapics[apic].apicid == mp_irqs[i].dstapic || + mp_irqs[i].dstapic == MP_APIC_ALL) break; if (!test_bit(lbus, mp_bus_not_pci) && - !mp_irqs[i].mp_irqtype && + !mp_irqs[i].irqtype && (bus == lbus) && - (slot == ((mp_irqs[i].mp_srcbusirq >> 2) & 0x1f))) { - int irq = pin_2_irq(i,apic,mp_irqs[i].mp_dstirq); + (slot == ((mp_irqs[i].srcbusirq >> 2) & 0x1f))) { + int irq = pin_2_irq(i, apic, mp_irqs[i].dstirq); if (!(apic || IO_APIC_IRQ(irq))) continue; - if (pin == (mp_irqs[i].mp_srcbusirq & 3)) + if (pin == (mp_irqs[i].srcbusirq & 3)) return irq; /* * Use the first all-but-pin matching entry as a @@ -1072,7 +1072,7 @@ static int EISA_ELCR(unsigned int irq) * EISA conforming in the MP table, that means its trigger type must * be read in from the ELCR */ -#define default_EISA_trigger(idx) (EISA_ELCR(mp_irqs[idx].mp_srcbusirq)) +#define default_EISA_trigger(idx) (EISA_ELCR(mp_irqs[idx].srcbusirq)) #define default_EISA_polarity(idx) default_ISA_polarity(idx) /* PCI interrupts are always polarity one level triggered, @@ -1089,13 +1089,13 @@ static int EISA_ELCR(unsigned int irq) static int MPBIOS_polarity(int idx) { - int bus = mp_irqs[idx].mp_srcbus; + int bus = mp_irqs[idx].srcbus; int polarity; /* * Determine IRQ line polarity (high active or low active): */ - switch (mp_irqs[idx].mp_irqflag & 3) + switch (mp_irqs[idx].irqflag & 3) { case 0: /* conforms, ie. bus-type dependent polarity */ if (test_bit(bus, mp_bus_not_pci)) @@ -1131,13 +1131,13 @@ static int MPBIOS_polarity(int idx) static int MPBIOS_trigger(int idx) { - int bus = mp_irqs[idx].mp_srcbus; + int bus = mp_irqs[idx].srcbus; int trigger; /* * Determine IRQ trigger mode (edge or level sensitive): */ - switch ((mp_irqs[idx].mp_irqflag>>2) & 3) + switch ((mp_irqs[idx].irqflag>>2) & 3) { case 0: /* conforms, ie. bus-type dependent */ if (test_bit(bus, mp_bus_not_pci)) @@ -1215,16 +1215,16 @@ int (*ioapic_renumber_irq)(int ioapic, int irq); static int pin_2_irq(int idx, int apic, int pin) { int irq, i; - int bus = mp_irqs[idx].mp_srcbus; + int bus = mp_irqs[idx].srcbus; /* * Debugging check, we are in big trouble if this message pops up! */ - if (mp_irqs[idx].mp_dstirq != pin) + if (mp_irqs[idx].dstirq != pin) printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n"); if (test_bit(bus, mp_bus_not_pci)) { - irq = mp_irqs[idx].mp_srcbusirq; + irq = mp_irqs[idx].srcbusirq; } else { /* * PCI IRQs are mapped in order @@ -2164,8 +2164,8 @@ static void __init setup_ioapic_ids_from_mpc(void) */ if (old_id != mp_ioapics[apic].apicid) for (i = 0; i < mp_irq_entries; i++) - if (mp_irqs[i].mp_dstapic == old_id) - mp_irqs[i].mp_dstapic + if (mp_irqs[i].dstapic == old_id) + mp_irqs[i].dstapic = mp_ioapics[apic].apicid; /* @@ -3983,8 +3983,8 @@ int acpi_get_override_irq(int bus_irq, int *trigger, int *polarity) return -1; for (i = 0; i < mp_irq_entries; i++) - if (mp_irqs[i].mp_irqtype == mp_INT && - mp_irqs[i].mp_srcbusirq == bus_irq) + if (mp_irqs[i].irqtype == mp_INT && + mp_irqs[i].srcbusirq == bus_irq) break; if (i >= mp_irq_entries) return -1; diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c index a86a6553743..ad36377dc93 100644 --- a/arch/x86/kernel/mpparse.c +++ b/arch/x86/kernel/mpparse.c @@ -159,55 +159,55 @@ static void print_MP_intsrc_info(struct mpc_intsrc *m) m->srcbusirq, m->dstapic, m->dstirq); } -static void __init print_mp_irq_info(struct mp_config_intsrc *mp_irq) +static void __init print_mp_irq_info(struct mpc_intsrc *mp_irq) { apic_printk(APIC_VERBOSE, "Int: type %d, pol %d, trig %d, bus %02x," " IRQ %02x, APIC ID %x, APIC INT %02x\n", - mp_irq->mp_irqtype, mp_irq->mp_irqflag & 3, - (mp_irq->mp_irqflag >> 2) & 3, mp_irq->mp_srcbus, - mp_irq->mp_srcbusirq, mp_irq->mp_dstapic, mp_irq->mp_dstirq); + mp_irq->irqtype, mp_irq->irqflag & 3, + (mp_irq->irqflag >> 2) & 3, mp_irq->srcbus, + mp_irq->srcbusirq, mp_irq->dstapic, mp_irq->dstirq); } static void __init assign_to_mp_irq(struct mpc_intsrc *m, - struct mp_config_intsrc *mp_irq) + struct mpc_intsrc *mp_irq) { - mp_irq->mp_dstapic = m->dstapic; - mp_irq->mp_type = m->type; - mp_irq->mp_irqtype = m->irqtype; - mp_irq->mp_irqflag = m->irqflag; - mp_irq->mp_srcbus = m->srcbus; - mp_irq->mp_srcbusirq = m->srcbusirq; - mp_irq->mp_dstirq = m->dstirq; + mp_irq->dstapic = m->dstapic; + mp_irq->type = m->type; + mp_irq->irqtype = m->irqtype; + mp_irq->irqflag = m->irqflag; + mp_irq->srcbus = m->srcbus; + mp_irq->srcbusirq = m->srcbusirq; + mp_irq->dstirq = m->dstirq; } -static void __init assign_to_mpc_intsrc(struct mp_config_intsrc *mp_irq, +static void __init assign_to_mpc_intsrc(struct mpc_intsrc *mp_irq, struct mpc_intsrc *m) { - m->dstapic = mp_irq->mp_dstapic; - m->type = mp_irq->mp_type; - m->irqtype = mp_irq->mp_irqtype; - m->irqflag = mp_irq->mp_irqflag; - m->srcbus = mp_irq->mp_srcbus; - m->srcbusirq = mp_irq->mp_srcbusirq; - m->dstirq = mp_irq->mp_dstirq; + m->dstapic = mp_irq->dstapic; + m->type = mp_irq->type; + m->irqtype = mp_irq->irqtype; + m->irqflag = mp_irq->irqflag; + m->srcbus = mp_irq->srcbus; + m->srcbusirq = mp_irq->srcbusirq; + m->dstirq = mp_irq->dstirq; } -static int __init mp_irq_mpc_intsrc_cmp(struct mp_config_intsrc *mp_irq, +static int __init mp_irq_mpc_intsrc_cmp(struct mpc_intsrc *mp_irq, struct mpc_intsrc *m) { - if (mp_irq->mp_dstapic != m->dstapic) + if (mp_irq->dstapic != m->dstapic) return 1; - if (mp_irq->mp_type != m->type) + if (mp_irq->type != m->type) return 2; - if (mp_irq->mp_irqtype != m->irqtype) + if (mp_irq->irqtype != m->irqtype) return 3; - if (mp_irq->mp_irqflag != m->irqflag) + if (mp_irq->irqflag != m->irqflag) return 4; - if (mp_irq->mp_srcbus != m->srcbus) + if (mp_irq->srcbus != m->srcbus) return 5; - if (mp_irq->mp_srcbusirq != m->srcbusirq) + if (mp_irq->srcbusirq != m->srcbusirq) return 6; - if (mp_irq->mp_dstirq != m->dstirq) + if (mp_irq->dstirq != m->dstirq) return 7; return 0; @@ -808,15 +808,15 @@ static int __init get_MP_intsrc_index(struct mpc_intsrc *m) /* not legacy */ for (i = 0; i < mp_irq_entries; i++) { - if (mp_irqs[i].mp_irqtype != mp_INT) + if (mp_irqs[i].irqtype != mp_INT) continue; - if (mp_irqs[i].mp_irqflag != 0x0f) + if (mp_irqs[i].irqflag != 0x0f) continue; - if (mp_irqs[i].mp_srcbus != m->srcbus) + if (mp_irqs[i].srcbus != m->srcbus) continue; - if (mp_irqs[i].mp_srcbusirq != m->srcbusirq) + if (mp_irqs[i].srcbusirq != m->srcbusirq) continue; if (irq_used[i]) { /* already claimed */ @@ -921,10 +921,10 @@ static int __init replace_intsrc_all(struct mpc_table *mpc, if (irq_used[i]) continue; - if (mp_irqs[i].mp_irqtype != mp_INT) + if (mp_irqs[i].irqtype != mp_INT) continue; - if (mp_irqs[i].mp_irqflag != 0x0f) + if (mp_irqs[i].irqflag != 0x0f) continue; if (nr_m_spare > 0) { -- cgit v1.2.3 From 09b3ec7315a18d885127544204f1e389d41058d0 Mon Sep 17 00:00:00 2001 From: Frederik Deweerdt Date: Mon, 12 Jan 2009 22:35:42 +0100 Subject: x86, tlb flush_data: replace per_cpu with an array Impact: micro-optimization, memory reduction On x86_64 flush tlb data is stored in per_cpu variables. This is unnecessary because only the first NUM_INVALIDATE_TLB_VECTORS entries are accessed. This patch aims at making the code less confusing (there's nothing really "per_cpu") by using a plain array. It also would save some memory on most distros out there (Ubuntu x86_64 has NR_CPUS=64 by default). [ Ravikiran G Thirumalai also pointed out that the correct alignment is ____cacheline_internodealigned_in_smp, so that there's no bouncing on vsmp. ] Signed-off-by: Frederik Deweerdt Acked-by: Ravikiran Thirumalai Signed-off-by: Ingo Molnar --- arch/x86/kernel/tlb_64.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index f8be6f1d2e4..8cfea5d1451 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -33,7 +33,7 @@ * To avoid global state use 8 different call vectors. * Each CPU uses a specific vector to trigger flushes on other * CPUs. Depending on the received vector the target CPUs look into - * the right per cpu variable for the flush data. + * the right array slot for the flush data. * * With more than 8 CPUs they are hashed to the 8 available * vectors. The limited global vector space forces us to this right now. @@ -48,13 +48,13 @@ union smp_flush_state { unsigned long flush_va; spinlock_t tlbstate_lock; }; - char pad[SMP_CACHE_BYTES]; -} ____cacheline_aligned; + char pad[CONFIG_X86_INTERNODE_CACHE_BYTES]; +} ____cacheline_internodealigned_in_smp; /* State is put into the per CPU data section, but padded to a full cache line because other CPUs can access it and we don't want false sharing in the per cpu data segment. */ -static DEFINE_PER_CPU(union smp_flush_state, flush_state); +static union smp_flush_state flush_state[NUM_INVALIDATE_TLB_VECTORS]; /* * We cannot call mmdrop() because we are in interrupt context, @@ -129,7 +129,7 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) * Use that to determine where the sender put the data. */ sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START; - f = &per_cpu(flush_state, sender); + f = &flush_state[sender]; if (!cpu_isset(cpu, f->flush_cpumask)) goto out; @@ -169,7 +169,7 @@ void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm, /* Caller has disabled preemption */ sender = smp_processor_id() % NUM_INVALIDATE_TLB_VECTORS; - f = &per_cpu(flush_state, sender); + f = &flush_state[sender]; /* * Could avoid this lock when @@ -205,8 +205,8 @@ static int __cpuinit init_smp_flush(void) { int i; - for_each_possible_cpu(i) - spin_lock_init(&per_cpu(flush_state, i).tlbstate_lock); + for (i = 0; i < ARRAY_SIZE(flush_state); i++) + spin_lock_init(&flush_state[i].tlbstate_lock); return 0; } -- cgit v1.2.3 From f11826385b63566d98c02d35f592232ee77cd791 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Wed, 14 Jan 2009 12:27:35 +0000 Subject: x86: fully honor "nolapic" Impact: widen the effect of the 'nolapic' boot parameter "nolapic" should not only suppress SMP and use of the LAPIC, but it also ought to have the effect of disabling all IO-APIC related activity as well as PCI MSI and HT-IRQs. Signed-off-by: Jan Beulich Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic.c | 7 ++++++- arch/x86/kernel/io_apic.c | 6 ++++++ arch/x86/kernel/smpboot.c | 1 + 3 files changed, 13 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index 0f830e4f567..c3dd64fabcf 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -1126,6 +1126,11 @@ void __cpuinit setup_local_APIC(void) unsigned int value; int i, j; + if (disable_apic) { + disable_ioapic_setup(); + return; + } + #ifdef CONFIG_X86_32 /* Pound the ESR really hard over the head with a big hammer - mbligh */ if (lapic_is_integrated() && esr_disable) { @@ -1566,11 +1571,11 @@ int apic_version[MAX_APICS]; int __init APIC_init_uniprocessor(void) { -#ifdef CONFIG_X86_64 if (disable_apic) { pr_info("Apic disabled\n"); return -1; } +#ifdef CONFIG_X86_64 if (!cpu_has_apic) { disable_apic = 1; pr_info("Apic disabled by BIOS\n"); diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index 1c4a1302536..40747e58f30 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -3258,6 +3258,9 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_ms int err; unsigned dest; + if (disable_apic) + return -ENXIO; + cfg = irq_cfg(irq); err = assign_irq_vector(irq, cfg, TARGET_CPUS); if (err) @@ -3726,6 +3729,9 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev) struct irq_cfg *cfg; int err; + if (disable_apic) + return -ENXIO; + cfg = irq_cfg(irq); err = assign_irq_vector(irq, cfg, TARGET_CPUS); if (!err) { diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index bb1a3b1fc87..31f99ec2e0f 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -1125,6 +1125,7 @@ static int __init smp_sanity_check(unsigned max_cpus) printk(KERN_ERR "... forcing use of dummy APIC emulation." "(tell your hw vendor)\n"); smpboot_clear_io_apic(); + disable_ioapic_setup(); return -1; } -- cgit v1.2.3 From a08c4743ed5b861c4fa3d75be00da7106c926296 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Wed, 14 Jan 2009 12:28:51 +0000 Subject: x86: avoid early crash in disable_local_APIC() E.g. when called due to an early panic. Signed-off-by: Jan Beulich Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index c3dd64fabcf..38d6aab2358 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -895,6 +895,10 @@ void disable_local_APIC(void) { unsigned int value; + /* APIC hasn't been mapped yet */ + if (!apic_phys) + return; + clear_local_APIC(); /* -- cgit v1.2.3 From 54da5b3d44238eeb7417bacf792fb416d473bf4d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 15 Jan 2009 13:04:58 +0100 Subject: x86: fix broken flush_tlb_others_ipi(), fix Impact: cleanup Use the proper type. Signed-off-by: Ingo Molnar --- arch/x86/kernel/tlb_64.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index 54ee2ecb5e2..7f4141d3b66 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -188,7 +188,8 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask, * We have to send the IPI only to * CPUs affected. */ - send_IPI_mask(f->flush_cpumask, INVALIDATE_TLB_VECTOR_START + sender); + send_IPI_mask(to_cpumask(f->flush_cpumask), + INVALIDATE_TLB_VECTOR_START + sender); while (!cpumask_empty(to_cpumask(f->flush_cpumask))) cpu_relax(); -- cgit v1.2.3 From 5cd7376200be7b8bab085557ff5876b04bd84191 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 15 Jan 2009 15:46:08 +0100 Subject: fix: crash: IP: __bitmap_intersects+0x48/0x73 -tip testing found this crash: > [ 35.258515] calling acpi_cpufreq_init+0x0/0x127 @ 1 > [ 35.264127] BUG: unable to handle kernel NULL pointer dereference at (null) > [ 35.267554] IP: [] __bitmap_intersects+0x48/0x73 > [ 35.267554] PGD 0 > [ 35.267554] Oops: 0000 [#1] SMP DEBUG_PAGEALLOC arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c is still broken: there's no allocation of the variable mask, so we pass in an uninitialized cmd.mask field to drv_read(), which then passes it to the scheduler which then crashes ... Switch it over to the much simpler constant-cpumask-pointers approach. Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c index 6f11e029e8c..019276717a7 100644 --- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c @@ -145,7 +145,7 @@ typedef union { struct drv_cmd { unsigned int type; - cpumask_var_t mask; + const struct cpumask *mask; drv_addr_union addr; u32 val; }; @@ -235,8 +235,7 @@ static u32 get_cur_val(const struct cpumask *mask) return 0; } - cpumask_copy(cmd.mask, mask); - + cmd.mask = mask; drv_read(&cmd); dprintk("get_cur_val = %u\n", cmd.val); @@ -403,9 +402,6 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, return -ENODEV; } - if (unlikely(!alloc_cpumask_var(&cmd.mask, GFP_KERNEL))) - return -ENOMEM; - perf = data->acpi_data; result = cpufreq_frequency_table_target(policy, data->freq_table, @@ -450,9 +446,9 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, /* cpufreq holds the hotplug lock, so we are safe from here on */ if (policy->shared_type != CPUFREQ_SHARED_TYPE_ANY) - cpumask_and(cmd.mask, cpu_online_mask, policy->cpus); + cmd.mask = policy->cpus; else - cpumask_copy(cmd.mask, cpumask_of(policy->cpu)); + cmd.mask = cpumask_of(policy->cpu); freqs.old = perf->states[perf->state].core_frequency * 1000; freqs.new = data->freq_table[next_state].frequency; @@ -479,7 +475,6 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, perf->state = next_perf_state; out: - free_cpumask_var(cmd.mask); return result; } -- cgit v1.2.3 From f10fcd47120e80f66665567dbe17f5071c7aef52 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:34 +0900 Subject: x86: make early_per_cpu() a lvalue and use it Make early_per_cpu() a lvalue as per_cpu() is and use it where applicable. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index 38d6aab2358..48578795583 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -1877,17 +1877,8 @@ void __cpuinit generic_processor_info(int apicid, int version) #endif #if defined(CONFIG_X86_SMP) || defined(CONFIG_X86_64) - /* are we being called early in kernel startup? */ - if (early_per_cpu_ptr(x86_cpu_to_apicid)) { - u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid); - u16 *bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid); - - cpu_to_apicid[cpu] = apicid; - bios_cpu_apicid[cpu] = apicid; - } else { - per_cpu(x86_cpu_to_apicid, cpu) = apicid; - per_cpu(x86_bios_cpu_apicid, cpu) = apicid; - } + early_per_cpu(x86_cpu_to_apicid, cpu) = apicid; + early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid; #endif set_cpu_possible(cpu, true); -- cgit v1.2.3 From c90aa894f0240084f2c6e42e2333b211d6cfe2b2 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Tue, 13 Jan 2009 20:41:34 +0900 Subject: x86: cleanup early setup_percpu references [ Based on original patch from Christoph Lameter and Mike Travis. ] * Ruggedize some calls in setup_percpu.c to prevent mishaps in early calls, particularly for non-critical functions. * Cleanup DEBUG_PER_CPU_MAPS usages and some comments. Signed-off-by: Mike Travis Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/setup_percpu.c | 56 ++++++++++++++++++++++++++++++------------ 1 file changed, 40 insertions(+), 16 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index bf63de72b64..56c63ac62b1 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -15,6 +15,12 @@ #include #include +#ifdef CONFIG_DEBUG_PER_CPU_MAPS +# define DBG(x...) printk(KERN_DEBUG x) +#else +# define DBG(x...) +#endif + #ifdef CONFIG_X86_LOCAL_APIC unsigned int num_processors; unsigned disabled_cpus __cpuinitdata; @@ -27,31 +33,39 @@ unsigned int max_physical_apicid; physid_mask_t phys_cpu_present_map; #endif -/* map cpu index to physical APIC ID */ +/* + * Map cpu index to physical APIC ID + */ DEFINE_EARLY_PER_CPU(u16, x86_cpu_to_apicid, BAD_APICID); DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID); EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid); EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); #if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) -#define X86_64_NUMA 1 +#define X86_64_NUMA 1 /* (used later) */ -/* map cpu index to node index */ +/* + * Map cpu index to node index + */ DEFINE_EARLY_PER_CPU(int, x86_cpu_to_node_map, NUMA_NO_NODE); EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map); -/* which logical CPUs are on which nodes */ +/* + * Which logical CPUs are on which nodes + */ cpumask_t *node_to_cpumask_map; EXPORT_SYMBOL(node_to_cpumask_map); -/* setup node_to_cpumask_map */ +/* + * Setup node_to_cpumask_map + */ static void __init setup_node_to_cpumask_map(void); #else static inline void setup_node_to_cpumask_map(void) { } #endif -#if defined(CONFIG_HAVE_SETUP_PER_CPU_AREA) && defined(CONFIG_X86_SMP) +#ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA /* * Copy data used in early init routines from the initial arrays to the * per cpu data areas. These arrays then become expendable and the @@ -200,6 +214,8 @@ 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); + + DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } /* Setup percpu data maps */ @@ -221,6 +237,7 @@ void __init setup_per_cpu_areas(void) * Requires node_possible_map to be valid. * * Note: node_to_cpumask() is not valid until after this is done. + * (Use CONFIG_DEBUG_PER_CPU_MAPS to check this.) */ static void __init setup_node_to_cpumask_map(void) { @@ -236,6 +253,7 @@ static void __init setup_node_to_cpumask_map(void) /* allocate the map */ map = alloc_bootmem_low(nr_node_ids * sizeof(cpumask_t)); + DBG("node_to_cpumask_map at %p for %d nodes\n", map, nr_node_ids); pr_debug("Node to cpumask map at %p for %d nodes\n", map, nr_node_ids); @@ -248,17 +266,23 @@ void __cpuinit numa_set_node(int cpu, int node) { int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); - if (cpu_pda(cpu) && node != NUMA_NO_NODE) - cpu_pda(cpu)->nodenumber = node; - - if (cpu_to_node_map) + /* early setting, no percpu area yet */ + if (cpu_to_node_map) { cpu_to_node_map[cpu] = node; + return; + } - else if (per_cpu_offset(cpu)) - per_cpu(x86_cpu_to_node_map, cpu) = node; +#ifdef CONFIG_DEBUG_PER_CPU_MAPS + if (cpu >= nr_cpu_ids || !per_cpu_offset(cpu)) { + printk(KERN_ERR "numa_set_node: invalid cpu# (%d)\n", cpu); + dump_stack(); + return; + } +#endif + per_cpu(x86_cpu_to_node_map, cpu) = node; - else - pr_debug("Setting node for non-present cpu %d\n", cpu); + if (node != NUMA_NO_NODE) + cpu_pda(cpu)->nodenumber = node; } void __cpuinit numa_clear_node(int cpu) @@ -275,7 +299,7 @@ void __cpuinit numa_add_cpu(int cpu) void __cpuinit numa_remove_cpu(int cpu) { - cpu_clear(cpu, node_to_cpumask_map[cpu_to_node(cpu)]); + cpu_clear(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]); } #else /* CONFIG_DEBUG_PER_CPU_MAPS */ @@ -285,7 +309,7 @@ void __cpuinit numa_remove_cpu(int cpu) */ static void __cpuinit numa_set_cpumask(int cpu, int enable) { - int node = cpu_to_node(cpu); + int node = early_cpu_to_node(cpu); cpumask_t *mask; char buf[64]; -- cgit v1.2.3 From a698c823e15149941b0f0281527d0c0d1daf2639 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: make vmlinux_32.lds.S use PERCPU() macro Make vmlinux_32.lds.S use the generic PERCPU() macro instead of open coding it. This will ease future changes. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/vmlinux_32.lds.S | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/vmlinux_32.lds.S b/arch/x86/kernel/vmlinux_32.lds.S index 82c67559dde..3eba7f7bac0 100644 --- a/arch/x86/kernel/vmlinux_32.lds.S +++ b/arch/x86/kernel/vmlinux_32.lds.S @@ -178,14 +178,7 @@ SECTIONS __initramfs_end = .; } #endif - . = ALIGN(PAGE_SIZE); - .data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) { - __per_cpu_start = .; - *(.data.percpu.page_aligned) - *(.data.percpu) - *(.data.percpu.shared_aligned) - __per_cpu_end = .; - } + PERCPU(PAGE_SIZE) . = ALIGN(PAGE_SIZE); /* freed after init ends here */ -- 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 ++++++++++++++++- 4 files changed, 42 insertions(+), 3 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index b9a4d8c4b93..bc2900ca82c 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 0e275d49556..7ee0363871e 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 56c63ac62b1..44845842e72 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 1a614c0e6be..f50280db0df 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 = .; -- cgit v1.2.3 From f32ff5388d86518c0375ccdb330d3b459b9c405e Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: load pointer to pda into %gs while brining up a CPU [ Based on original patch from Christoph Lameter and Mike Travis. ] CPU startup code in head_64.S loaded address of a zero page into %gs for temporary use till pda is loaded but address to the actual pda is available at the point. Load the real address directly instead. This will help unifying percpu and pda handling later on. This patch is mostly taken from Mike Travis' "x86_64: Fold pda into per cpu area" patch. Signed-off-by: Tejun Heo --- arch/x86/kernel/acpi/sleep.c | 1 + arch/x86/kernel/head64.c | 4 ++-- arch/x86/kernel/head_64.S | 15 ++++++++++----- arch/x86/kernel/smpboot.c | 1 + 4 files changed, 14 insertions(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c index 707c1f6f95f..9ff67f8dc2c 100644 --- a/arch/x86/kernel/acpi/sleep.c +++ b/arch/x86/kernel/acpi/sleep.c @@ -101,6 +101,7 @@ int acpi_save_state_mem(void) stack_start.sp = temp_stack + sizeof(temp_stack); early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(smp_processor_id()); + initial_gs = (unsigned long)cpu_pda(smp_processor_id()); #endif initial_code = (unsigned long)wakeup_long64; saved_magic = 0x123456789abcdef0; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index bc2900ca82c..76ffba2aa66 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -26,8 +26,8 @@ #include #include -/* boot cpu pda */ -static struct x8664_pda _boot_cpu_pda; +/* boot cpu pda, referenced by head_64.S to initialize %gs for boot CPU */ +struct x8664_pda _boot_cpu_pda; #ifdef CONFIG_SMP /* diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 7ee0363871e..2f0ab008988 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -243,12 +243,15 @@ ENTRY(secondary_startup_64) movl %eax,%fs movl %eax,%gs - /* - * Setup up a dummy PDA. this is just for some early bootup code - * that does in_interrupt() - */ + /* Set up %gs. + * + * %gs should point to the pda. For initial boot, make %gs point + * to the _boot_cpu_pda in data section. For a secondary CPU, + * initial_gs should be set to its pda address before the CPU runs + * this code. + */ movl $MSR_GS_BASE,%ecx - movq $empty_zero_page,%rax + movq initial_gs(%rip),%rax movq %rax,%rdx shrq $32,%rdx wrmsr @@ -274,6 +277,8 @@ ENTRY(secondary_startup_64) .align 8 ENTRY(initial_code) .quad x86_64_start_kernel + ENTRY(initial_gs) + .quad _boot_cpu_pda __FINITDATA ENTRY(stack_start) diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 1a712da1dfa..70d846628bb 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -854,6 +854,7 @@ do_rest: #else cpu_pda(cpu)->pcurrent = c_idle.idle; clear_tsk_thread_flag(c_idle.idle, TIF_FORK); + initial_gs = (unsigned long)cpu_pda(cpu); #endif early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); initial_code = (unsigned long)start_secondary; -- cgit v1.2.3 From c8f3329a0ddd751241e96b4100df7eda14b2cbc6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: use static _cpu_pda array _cpu_pda array first uses statically allocated storage in data.init and then switches to allocated bootmem to conserve space. However, after folding pda area into percpu area, _cpu_pda array will be removed completely. Drop the reallocation part to simplify the code for soon-to-follow changes. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common.c | 2 +- arch/x86/kernel/head64.c | 12 ------------ arch/x86/kernel/setup_percpu.c | 14 +++----------- 3 files changed, 4 insertions(+), 24 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index f0025846244..c116c599326 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -879,7 +879,7 @@ static __init int setup_disablecpuid(char *arg) __setup("clearcpuid=", setup_disablecpuid); #ifdef CONFIG_X86_64 -struct x8664_pda **_cpu_pda __read_mostly; +struct x8664_pda *_cpu_pda[NR_CPUS] __read_mostly; EXPORT_SYMBOL(_cpu_pda); struct desc_ptr idt_descr = { 256 * 16 - 1, (unsigned long) idt_table }; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index 76ffba2aa66..462d0beccb6 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -29,20 +29,8 @@ /* boot cpu pda, referenced by head_64.S to initialize %gs for boot CPU */ struct x8664_pda _boot_cpu_pda; -#ifdef CONFIG_SMP -/* - * We install an empty cpu_pda pointer table to indicate to early users - * (numa_set_node) that the cpu_pda pointer table for cpus other than - * the boot cpu is not yet setup. - */ -static struct x8664_pda *__cpu_pda[NR_CPUS] __initdata; -#else -static struct x8664_pda *__cpu_pda[NR_CPUS] __read_mostly; -#endif - 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); diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 44845842e72..73ab01b297c 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -114,7 +114,6 @@ static inline void setup_cpu_pda_map(void) { } static void __init setup_cpu_pda_map(void) { char *pda; - struct x8664_pda **new_cpu_pda; unsigned long size; int cpu; @@ -122,28 +121,21 @@ static void __init setup_cpu_pda_map(void) /* allocate cpu_pda array and pointer table */ { - unsigned long tsize = nr_cpu_ids * sizeof(void *); unsigned long asize = size * (nr_cpu_ids - 1); - tsize = roundup(tsize, cache_line_size()); - new_cpu_pda = alloc_bootmem(tsize + asize); - pda = (char *)new_cpu_pda + tsize; + 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 */ - new_cpu_pda[0] = cpu_pda(0); continue; } - new_cpu_pda[cpu] = (struct x8664_pda *)pda; - new_cpu_pda[cpu]->in_bootmem = 1; + cpu_pda(cpu) = (struct x8664_pda *)pda; + cpu_pda(cpu)->in_bootmem = 1; pda += size; } - - /* point to new pointer table */ - _cpu_pda = new_cpu_pda; } #endif /* CONFIG_SMP && CONFIG_X86_64 */ -- cgit v1.2.3 From 1a51e3a0aed18767cf2762e95456ecfeb0bca5e6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: fold pda into percpu area on SMP [ Based on original patch from Christoph Lameter and Mike Travis. ] Currently pdas and percpu areas are allocated separately. %gs points to local pda and percpu area can be reached using pda->data_offset. This patch folds pda into percpu area. Due to strange gcc requirement, pda needs to be at the beginning of the percpu area so that pda->stack_canary is at %gs:40. To achieve this, a new percpu output section macro - PERCPU_VADDR_PREALLOC() - is added and used to reserve pda sized chunk at the start of the percpu area. After this change, for boot cpu, %gs first points to pda in the data.init area and later during setup_per_cpu_areas() gets updated to point to the actual pda. This means that setup_per_cpu_areas() need to reload %gs for CPU0 while clearing pda area for other cpus as cpu0 already has modified it when control reaches setup_per_cpu_areas(). This patch also removes now unnecessary get_local_pda() and its call sites. A lot of this patch is taken from Mike Travis' "x86_64: Fold pda into per cpu area" patch. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/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 ++- 7 files changed, 72 insertions(+), 131 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 1d41d3f1edb..f8d1b047ef4 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 c116c599326..7041acdf557 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 462d0beccb6..1a311293f73 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 2f0ab008988..7a995d0e9f7 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 73ab01b297c..63d46280227 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 70d846628bb..f2f77ca494d 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 f50280db0df..962f21f1d4d 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 -- cgit v1.2.3 From 9939ddaff52787b2a7c1adf1b2afc95421aa0884 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: merge 64 and 32 SMP percpu handling Now that pda is allocated as part of percpu, percpu doesn't need to be accessed through pda. Unify x86_64 SMP percpu access with x86_32 SMP one. Other than the segment register, operand size and the base of percpu symbols, they behave identical now. This patch replaces now unnecessary pda->data_offset with a dummy field which is necessary to keep stack_canary at its place. This patch also moves per_cpu_offset initialization out of init_gdt() into setup_per_cpu_areas(). Note that this change also necessitates explicit per_cpu_offset initializations in voyager_smp.c. With this change, x86_OP_percpu()'s are as efficient on x86_64 as on x86_32 and also x86_64 can use assembly PER_CPU macros. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/entry_64.S | 7 ++++--- arch/x86/kernel/head64.c | 2 -- arch/x86/kernel/setup_percpu.c | 15 ++++++++------- arch/x86/kernel/smpcommon.c | 3 +-- 5 files changed, 13 insertions(+), 15 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index f8d1b047ef4..f4cc81bfbf8 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -55,7 +55,6 @@ int main(void) ENTRY(irqcount); ENTRY(cpunumber); ENTRY(irqstackptr); - ENTRY(data_offset); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); #undef ENTRY diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index e28c7a98779..4833f3a1965 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -52,6 +52,7 @@ #include #include #include +#include /* Avoid __ASSEMBLER__'ifying just for this. */ #include @@ -1072,10 +1073,10 @@ ENTRY(\sym) TRACE_IRQS_OFF movq %rsp,%rdi /* pt_regs pointer */ xorl %esi,%esi /* no error code */ - movq %gs:pda_data_offset, %rbp - subq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp) + PER_CPU(init_tss, %rbp) + subq $EXCEPTION_STKSZ, TSS_ist + (\ist - 1) * 8(%rbp) call \do_sym - addq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp) + addq $EXCEPTION_STKSZ, TSS_ist + (\ist - 1) * 8(%rbp) jmp paranoid_exit /* %ebx: no swapgs flag */ CFI_ENDPROC END(\sym) diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index 1a311293f73..e99b661a97f 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -38,8 +38,6 @@ void __init x86_64_init_pda(void) #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/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 63d46280227..be1ff34db11 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -125,14 +125,14 @@ static void __init setup_per_cpu_maps(void) #endif } -#ifdef CONFIG_X86_32 -/* - * Great future not-so-futuristic plan: make i386 and x86_64 do it - * the same way - */ +#ifdef CONFIG_X86_64 +unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { + [0] = (unsigned long)__per_cpu_load, +}; +#else unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; -EXPORT_SYMBOL(__per_cpu_offset); #endif +EXPORT_SYMBOL(__per_cpu_offset); /* * Great future plan: @@ -178,6 +178,7 @@ void __init setup_per_cpu_areas(void) #endif memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); + per_cpu_offset(cpu) = ptr - __per_cpu_start; #ifdef CONFIG_X86_64 cpu_pda(cpu) = (void *)ptr; @@ -190,7 +191,7 @@ void __init setup_per_cpu_areas(void) else memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu))); #endif - per_cpu_offset(cpu) = ptr - __per_cpu_start; + per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c index 397e309839d..84395fabc41 100644 --- a/arch/x86/kernel/smpcommon.c +++ b/arch/x86/kernel/smpcommon.c @@ -4,10 +4,10 @@ #include #include -#ifdef CONFIG_X86_32 DEFINE_PER_CPU(unsigned long, this_cpu_off); EXPORT_PER_CPU_SYMBOL(this_cpu_off); +#ifdef CONFIG_X86_32 /* * Initialize the CPU's GDT. This is either the boot CPU doing itself * (still using the master per-cpu area), or a CPU doing it for a @@ -24,7 +24,6 @@ __cpuinit void init_gdt(int cpu) write_gdt_entry(get_cpu_gdt_table(cpu), GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S); - per_cpu(this_cpu_off, cpu) = __per_cpu_offset[cpu]; per_cpu(cpu_number, cpu) = cpu; } #endif -- cgit v1.2.3 From b12d8db8fbfaed1e8222a15333a3645599636854 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: make pda a percpu variable [ Based on original patch from Christoph Lameter and Mike Travis. ] As pda is now allocated in percpu area, it can easily be made a proper percpu variable. Make it so by defining per cpu symbol from linker script and declaring it in C code for SMP and simply defining it for UP. This change cleans up code and brings SMP and UP closer a bit. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common.c | 3 --- arch/x86/kernel/head64.c | 10 ---------- arch/x86/kernel/head_64.S | 5 +++-- arch/x86/kernel/setup_percpu.c | 16 ++++++++++++++-- arch/x86/kernel/vmlinux_64.lds.S | 4 +++- 5 files changed, 20 insertions(+), 18 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 7041acdf557..c49498d4083 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -879,9 +879,6 @@ static __init int setup_disablecpuid(char *arg) __setup("clearcpuid=", setup_disablecpuid); #ifdef CONFIG_X86_64 -struct x8664_pda *_cpu_pda[NR_CPUS] __read_mostly; -EXPORT_SYMBOL(_cpu_pda); - struct desc_ptr idt_descr = { 256 * 16 - 1, (unsigned long) idt_table }; static char boot_cpu_stack[IRQSTACKSIZE] __page_aligned_bss; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index e99b661a97f..71b6f6ec96a 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -26,18 +26,8 @@ #include #include -#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 pda_init(0); } diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 7a995d0e9f7..c8ace880661 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -19,6 +19,7 @@ #include #include #include +#include #ifdef CONFIG_PARAVIRT #include @@ -250,7 +251,7 @@ ENTRY(secondary_startup_64) * secondary CPU,initial_gs should be set to its pda address * before the CPU runs this code. * - * On UP, initial_gs points to _boot_cpu_pda and doesn't + * On UP, initial_gs points to PER_CPU_VAR(__pda) and doesn't * change. */ movl $MSR_GS_BASE,%ecx @@ -284,7 +285,7 @@ ENTRY(secondary_startup_64) #ifdef CONFIG_SMP .quad __per_cpu_load #else - .quad _boot_cpu_pda + .quad PER_CPU_VAR(__pda) #endif __FINITDATA diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index be1ff34db11..daeedf82c15 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -66,6 +66,16 @@ static void __init setup_node_to_cpumask_map(void); static inline void setup_node_to_cpumask_map(void) { } #endif +/* + * Define load_pda_offset() and per-cpu __pda for x86_64. + * load_pda_offset() is responsible for loading the offset of pda into + * %gs. + * + * On SMP, pda offset also duals as percpu base address and thus it + * should be at the start of per-cpu area. To achieve this, it's + * preallocated in vmlinux_64.lds.S directly instead of using + * DEFINE_PER_CPU(). + */ #ifdef CONFIG_X86_64 void __cpuinit load_pda_offset(int cpu) { @@ -74,6 +84,10 @@ void __cpuinit load_pda_offset(int cpu) wrmsrl(MSR_GS_BASE, cpu_pda(cpu)); mb(); } +#ifndef CONFIG_SMP +DEFINE_PER_CPU(struct x8664_pda, __pda); +EXPORT_PER_CPU_SYMBOL(__pda); +#endif #endif /* CONFIG_SMP && CONFIG_X86_64 */ @@ -180,8 +194,6 @@ void __init setup_per_cpu_areas(void) memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); per_cpu_offset(cpu) = ptr - __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. diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S index 962f21f1d4d..d2a0baa87d1 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -217,10 +217,12 @@ 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. Also, pda should be at the head of - * percpu area. Preallocate it. + * percpu area. Preallocate it and define the percpu offset symbol + * so that it can be accessed as a percpu variable. */ . = ALIGN(PAGE_SIZE); PERCPU_VADDR_PREALLOC(0, :percpu, pda_size) + per_cpu____pda = __per_cpu_start; #else PERCPU(PAGE_SIZE) #endif -- cgit v1.2.3 From 49357d19e4fb31e28796eaff83499e7584c26878 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: convert pda ops to wrappers around x86 percpu accessors pda is now a percpu variable and there's no reason it can't use plain x86 percpu accessors. Add x86_test_and_clear_bit_percpu() and replace pda op implementations with wrappers around x86 percpu accessors. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/vmlinux_64.lds.S | 1 - arch/x86/kernel/x8664_ksyms_64.c | 2 -- 2 files changed, 3 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S index d2a0baa87d1..a09abb8fb97 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -14,7 +14,6 @@ OUTPUT_FORMAT("elf64-x86-64", "elf64-x86-64", "elf64-x86-64") OUTPUT_ARCH(i386:x86-64) ENTRY(phys_startup_64) jiffies_64 = jiffies; -_proxy_pda = 1; PHDRS { text PT_LOAD FLAGS(5); /* R_E */ data PT_LOAD FLAGS(7); /* RWE */ diff --git a/arch/x86/kernel/x8664_ksyms_64.c b/arch/x86/kernel/x8664_ksyms_64.c index 695e426aa35..3909e3ba5ce 100644 --- a/arch/x86/kernel/x8664_ksyms_64.c +++ b/arch/x86/kernel/x8664_ksyms_64.c @@ -58,5 +58,3 @@ EXPORT_SYMBOL(__memcpy); EXPORT_SYMBOL(empty_zero_page); EXPORT_SYMBOL(init_level4_pgt); EXPORT_SYMBOL(load_gs_index); - -EXPORT_SYMBOL(_proxy_pda); -- cgit v1.2.3 From 004aa322f855a765741d9437a98dd8fe2e4f32a6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: misc clean up after the percpu update Do the following cleanups: * kill x86_64_init_pda() which now is equivalent to pda_init() * use per_cpu_offset() instead of cpu_pda() when initializing initial_gs Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/kernel/acpi/sleep.c | 2 +- arch/x86/kernel/head64.c | 7 +------ arch/x86/kernel/smpboot.c | 2 +- 3 files changed, 3 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c index 9ff67f8dc2c..4abff454c55 100644 --- a/arch/x86/kernel/acpi/sleep.c +++ b/arch/x86/kernel/acpi/sleep.c @@ -101,7 +101,7 @@ int acpi_save_state_mem(void) stack_start.sp = temp_stack + sizeof(temp_stack); early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(smp_processor_id()); - initial_gs = (unsigned long)cpu_pda(smp_processor_id()); + initial_gs = per_cpu_offset(smp_processor_id()); #endif initial_code = (unsigned long)wakeup_long64; saved_magic = 0x123456789abcdef0; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index 71b6f6ec96a..af67d3227ea 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -26,11 +26,6 @@ #include #include -void __init x86_64_init_pda(void) -{ - pda_init(0); -} - static void __init zap_identity_mappings(void) { pgd_t *pgd = pgd_offset_k(0UL); @@ -96,7 +91,7 @@ void __init x86_64_start_kernel(char * real_mode_data) if (console_loglevel == 10) early_printk("Kernel alive\n"); - x86_64_init_pda(); + pda_init(0); x86_64_start_reservations(real_mode_data); } diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index f2f77ca494d..2f0e0f1090f 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -798,7 +798,7 @@ do_rest: #else cpu_pda(cpu)->pcurrent = c_idle.idle; clear_tsk_thread_flag(c_idle.idle, TIF_FORK); - initial_gs = (unsigned long)cpu_pda(cpu); + initial_gs = per_cpu_offset(cpu); #endif early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); initial_code = (unsigned long)start_secondary; -- 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/kernel/process_32.c | 2 +- arch/x86/kernel/tlb_32.c | 10 +++++----- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index a546f55c77b..77d546817d9 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 ec53818f4e3..e65449d0f7d 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); } -- cgit v1.2.3 From a338af2c648f5e07c582154745a6c60cd2d8bf12 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 16 Jan 2009 11:19:03 +0900 Subject: x86: fix build bug introduced during merge EXPORT_PER_CPU_SYMBOL() got misplaced during merge leading to build failure. Fix it. Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index daeedf82c15..b5c35af2011 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -86,9 +86,8 @@ void __cpuinit load_pda_offset(int cpu) } #ifndef CONFIG_SMP DEFINE_PER_CPU(struct x8664_pda, __pda); -EXPORT_PER_CPU_SYMBOL(__pda); #endif - +EXPORT_PER_CPU_SYMBOL(__pda); #endif /* CONFIG_SMP && CONFIG_X86_64 */ #ifdef CONFIG_X86_64 -- cgit v1.2.3 From cd3adf52309867955d6e2175246b526442235805 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 16 Jan 2009 12:11:43 +0900 Subject: x86_64: initialize this_cpu_off to __per_cpu_load On x86_64, if get_per_cpu_var() is used before per cpu area is setup (if lockdep is turned on, it happens), it needs this_cpu_off to point to __per_cpu_load. Initialize accordingly. Signed-off-by: Tejun Heo --- arch/x86/kernel/smpcommon.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c index 84395fabc41..7e157810062 100644 --- a/arch/x86/kernel/smpcommon.c +++ b/arch/x86/kernel/smpcommon.c @@ -3,8 +3,13 @@ */ #include #include +#include +#ifdef CONFIG_X86_64 +DEFINE_PER_CPU(unsigned long, this_cpu_off) = (unsigned long)__per_cpu_load; +#else DEFINE_PER_CPU(unsigned long, this_cpu_off); +#endif EXPORT_PER_CPU_SYMBOL(this_cpu_off); #ifdef CONFIG_X86_32 -- cgit v1.2.3 From 6eb714c63ed5bd663627f7dda8c4d5258f3b64ef Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Fri, 16 Jan 2009 15:31:15 -0800 Subject: cpufreq: use work_on_cpu in acpi-cpufreq.c for drv_read and drv_write Impact: use new work_on_cpu function to reduce stack usage Replace the saving of current->cpus_allowed and set_cpus_allowed_ptr() with a work_on_cpu function for drv_read() and drv_write(). Basically converts do_drv_{read,write} into "work_on_cpu" functions that are now called by drv_read and drv_write. Note: This patch basically reverts 50c668d6 which reverted 7503bfba, now that the work_on_cpu() function is more stable. Signed-off-by: Mike Travis Acked-by: Rusty Russell Tested-by: Dieter Ries Tested-by: Maciej Rutecki Cc: Dave Jones Cc: --- arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c index 019276717a7..4b1c319d30c 100644 --- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c @@ -150,8 +150,9 @@ struct drv_cmd { u32 val; }; -static void do_drv_read(struct drv_cmd *cmd) +static long do_drv_read(void *_cmd) { + struct drv_cmd *cmd = _cmd; u32 h; switch (cmd->type) { @@ -166,10 +167,12 @@ static void do_drv_read(struct drv_cmd *cmd) default: break; } + return 0; } -static void do_drv_write(struct drv_cmd *cmd) +static long do_drv_write(void *_cmd) { + struct drv_cmd *cmd = _cmd; u32 lo, hi; switch (cmd->type) { @@ -186,30 +189,23 @@ static void do_drv_write(struct drv_cmd *cmd) default: break; } + return 0; } static void drv_read(struct drv_cmd *cmd) { - cpumask_t saved_mask = current->cpus_allowed; cmd->val = 0; - set_cpus_allowed_ptr(current, cmd->mask); - do_drv_read(cmd); - set_cpus_allowed_ptr(current, &saved_mask); + work_on_cpu(cpumask_any(cmd->mask), do_drv_read, cmd); } static void drv_write(struct drv_cmd *cmd) { - cpumask_t saved_mask = current->cpus_allowed; unsigned int i; for_each_cpu(i, cmd->mask) { - set_cpus_allowed_ptr(current, cpumask_of(i)); - do_drv_write(cmd); + work_on_cpu(i, do_drv_write, cmd); } - - set_cpus_allowed_ptr(current, &saved_mask); - return; } static u32 get_cur_val(const struct cpumask *mask) @@ -367,7 +363,7 @@ static unsigned int get_cur_freq_on_cpu(unsigned int cpu) return freq; } -static unsigned int check_freqs(const cpumask_t *mask, unsigned int freq, +static unsigned int check_freqs(const struct cpumask *mask, unsigned int freq, struct acpi_cpufreq_data *data) { unsigned int cur_freq; -- cgit v1.2.3 From cef30b3a84e1c7cbd49987d83c5863c001445842 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Fri, 16 Jan 2009 15:58:13 -0800 Subject: x86: put trigger in to detect mismatched apic versions. Fire off one message if two apic's discovered with different apic versions. Signed-off-by: Mike Travis --- arch/x86/kernel/apic.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index 0f830e4f567..db0998641c5 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -1833,6 +1833,11 @@ void __cpuinit generic_processor_info(int apicid, int version) num_processors++; cpu = cpumask_next_zero(-1, cpu_present_mask); + if (version != apic_version[boot_cpu_physical_apicid]) + WARN_ONCE(1, + "ACPI: apic version mismatch, bootcpu: %x cpu %d: %x\n", + apic_version[boot_cpu_physical_apicid], cpu, version); + physid_set(apicid, phys_cpu_present_map); if (apicid == boot_cpu_physical_apicid) { /* -- cgit v1.2.3 From 1b437c8c73a36daa471dd54a63c426d72af5723d Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:57 +0900 Subject: x86-64: Move irq stats from PDA to per-cpu and consolidate with 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/irq.c | 6 +----- arch/x86/kernel/irq_64.c | 3 +++ arch/x86/kernel/nmi.c | 10 +--------- 3 files changed, 5 insertions(+), 14 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c index 3973e2df7f8..8b30d0c2512 100644 --- a/arch/x86/kernel/irq.c +++ b/arch/x86/kernel/irq.c @@ -36,11 +36,7 @@ void ack_bad_irq(unsigned int irq) #endif } -#ifdef CONFIG_X86_32 -# define irq_stats(x) (&per_cpu(irq_stat, x)) -#else -# define irq_stats(x) cpu_pda(x) -#endif +#define irq_stats(x) (&per_cpu(irq_stat, x)) /* * /proc/interrupts printing: */ diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c index 0b21cb1ea11..1db05247b47 100644 --- a/arch/x86/kernel/irq_64.c +++ b/arch/x86/kernel/irq_64.c @@ -19,6 +19,9 @@ #include #include +DEFINE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat); +EXPORT_PER_CPU_SYMBOL(irq_stat); + /* * Probabilistic stack overflow check: * diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 7228979f1e7..23b6d9e6e4f 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -61,11 +61,7 @@ static int endflag __initdata; static inline unsigned int get_nmi_count(int cpu) { -#ifdef CONFIG_X86_64 - return cpu_pda(cpu)->__nmi_count; -#else - return nmi_count(cpu); -#endif + return per_cpu(irq_stat, cpu).__nmi_count; } static inline int mce_in_progress(void) @@ -82,12 +78,8 @@ static inline int mce_in_progress(void) */ static inline unsigned int get_timer_irqs(int cpu) { -#ifdef CONFIG_X86_64 - return read_pda(apic_timer_irqs) + read_pda(irq0_irqs); -#else return per_cpu(irq_stat, cpu).apic_timer_irqs + per_cpu(irq_stat, cpu).irq0_irqs; -#endif } #ifdef CONFIG_SMP -- cgit v1.2.3 From 9eb912d1aa6b8106e06a73ea6702ec3dab0d6a1a Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:57 +0900 Subject: x86-64: Move TLB state from PDA to per-cpu and consolidate with 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 2 -- arch/x86/kernel/tlb_32.c | 12 ++---------- arch/x86/kernel/tlb_64.c | 13 ++++++++----- 3 files changed, 10 insertions(+), 17 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index c49498d4083..3d0cc6f1711 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -897,8 +897,6 @@ void __cpuinit pda_init(int cpu) pda->irqcount = -1; pda->kernelstack = (unsigned long)stack_thread_info() - PDA_STACKOFFSET + THREAD_SIZE; - pda->active_mm = &init_mm; - pda->mmu_state = 0; if (cpu == 0) { /* others are initialized in smpboot.c */ diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index e65449d0f7d..abf0808d6fc 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -4,8 +4,8 @@ #include -DEFINE_PER_CPU(struct tlb_state, cpu_tlbstate) - ____cacheline_aligned = { &init_mm, 0, }; +DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) + = { &init_mm, 0, }; /* must come after the send_IPI functions above for inlining */ #include @@ -231,14 +231,6 @@ void flush_tlb_all(void) on_each_cpu(do_flush_tlb_all, NULL, 1); } -void reset_lazy_tlbstate(void) -{ - int cpu = raw_smp_processor_id(); - - per_cpu(cpu_tlbstate, cpu).state = 0; - per_cpu(cpu_tlbstate, cpu).active_mm = &init_mm; -} - static int init_flush_cpumask(void) { alloc_cpumask_var(&flush_cpumask, GFP_KERNEL); diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index 7f4141d3b66..e64a32c4882 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -18,6 +18,9 @@ #include #include +DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) + = { &init_mm, 0, }; + #include /* * Smarter SMP flushing macros. @@ -62,9 +65,9 @@ static DEFINE_PER_CPU(union smp_flush_state, flush_state); */ void leave_mm(int cpu) { - if (read_pda(mmu_state) == TLBSTATE_OK) + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) BUG(); - cpu_clear(cpu, read_pda(active_mm)->cpu_vm_mask); + cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); load_cr3(swapper_pg_dir); } EXPORT_SYMBOL_GPL(leave_mm); @@ -142,8 +145,8 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) * BUG(); */ - if (f->flush_mm == read_pda(active_mm)) { - if (read_pda(mmu_state) == TLBSTATE_OK) { + if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) { + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { if (f->flush_va == TLB_FLUSH_ALL) local_flush_tlb(); else @@ -281,7 +284,7 @@ static void do_flush_tlb_all(void *info) unsigned long cpu = smp_processor_id(); __flush_tlb_all(); - if (read_pda(mmu_state) == TLBSTATE_LAZY) + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) leave_mm(cpu); } -- cgit v1.2.3 From 26f80bd6a9ab17bc8a60b6092e7c0d05c5927ce5 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Convert irqstacks to per-cpu Move the irqstackptr variable from the PDA to per-cpu. Make the stacks themselves per-cpu, removing some specific allocation code. Add a seperate flag (is_boot_cpu) to simplify the per-cpu boot adjustments. tj: * sprinkle some underbars around. * irq_stack_ptr is not used till traps_init(), no reason to initialize it early. On SMP, just leaving it NULL till proper initialization in setup_per_cpu_areas() works. Dropped is_boot_cpu and early irq_stack_ptr initialization. * do DECLARE/DEFINE_PER_CPU(char[IRQ_STACK_SIZE], irq_stack) instead of (char, irq_stack[IRQ_STACK_SIZE]). Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 19 +++++++------------ arch/x86/kernel/dumpstack_64.c | 33 +++++++++++++++++---------------- arch/x86/kernel/entry_64.S | 6 +++--- arch/x86/kernel/setup_percpu.c | 4 +++- 5 files changed, 30 insertions(+), 33 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index f4cc81bfbf8..5b821fbdaf7 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -54,7 +54,6 @@ int main(void) ENTRY(pcurrent); ENTRY(irqcount); ENTRY(cpunumber); - ENTRY(irqstackptr); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); #undef ENTRY diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 3d0cc6f1711..496f0a01919 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -881,7 +881,13 @@ __setup("clearcpuid=", setup_disablecpuid); #ifdef CONFIG_X86_64 struct desc_ptr idt_descr = { 256 * 16 - 1, (unsigned long) idt_table }; -static char boot_cpu_stack[IRQSTACKSIZE] __page_aligned_bss; +DEFINE_PER_CPU_PAGE_ALIGNED(char[IRQ_STACK_SIZE], irq_stack); +#ifdef CONFIG_SMP +DEFINE_PER_CPU(char *, irq_stack_ptr); /* will be set during per cpu init */ +#else +DEFINE_PER_CPU(char *, irq_stack_ptr) = + per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64; +#endif void __cpuinit pda_init(int cpu) { @@ -901,18 +907,7 @@ void __cpuinit pda_init(int cpu) if (cpu == 0) { /* others are initialized in smpboot.c */ pda->pcurrent = &init_task; - pda->irqstackptr = boot_cpu_stack; - pda->irqstackptr += IRQSTACKSIZE - 64; } else { - if (!pda->irqstackptr) { - pda->irqstackptr = (char *) - __get_free_pages(GFP_ATOMIC, IRQSTACK_ORDER); - if (!pda->irqstackptr) - panic("cannot allocate irqstack for cpu %d", - cpu); - pda->irqstackptr += IRQSTACKSIZE - 64; - } - if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) pda->nodenumber = cpu_to_node(cpu); } diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c index c302d070704..28e26a4315d 100644 --- a/arch/x86/kernel/dumpstack_64.c +++ b/arch/x86/kernel/dumpstack_64.c @@ -106,7 +106,8 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs, const struct stacktrace_ops *ops, void *data) { const unsigned cpu = get_cpu(); - unsigned long *irqstack_end = (unsigned long *)cpu_pda(cpu)->irqstackptr; + unsigned long *irq_stack_end = + (unsigned long *)per_cpu(irq_stack_ptr, cpu); unsigned used = 0; struct thread_info *tinfo; int graph = 0; @@ -160,23 +161,23 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs, stack = (unsigned long *) estack_end[-2]; continue; } - if (irqstack_end) { - unsigned long *irqstack; - irqstack = irqstack_end - - (IRQSTACKSIZE - 64) / sizeof(*irqstack); + if (irq_stack_end) { + unsigned long *irq_stack; + irq_stack = irq_stack_end - + (IRQ_STACK_SIZE - 64) / sizeof(*irq_stack); - if (stack >= irqstack && stack < irqstack_end) { + if (stack >= irq_stack && stack < irq_stack_end) { if (ops->stack(data, "IRQ") < 0) break; bp = print_context_stack(tinfo, stack, bp, - ops, data, irqstack_end, &graph); + ops, data, irq_stack_end, &graph); /* * We link to the next stack (which would be * the process stack normally) the last * pointer (index -1 to end) in the IRQ stack: */ - stack = (unsigned long *) (irqstack_end[-1]); - irqstack_end = NULL; + stack = (unsigned long *) (irq_stack_end[-1]); + irq_stack_end = NULL; ops->stack(data, "EOI"); continue; } @@ -199,10 +200,10 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, unsigned long *stack; int i; const int cpu = smp_processor_id(); - unsigned long *irqstack_end = - (unsigned long *) (cpu_pda(cpu)->irqstackptr); - unsigned long *irqstack = - (unsigned long *) (cpu_pda(cpu)->irqstackptr - IRQSTACKSIZE); + unsigned long *irq_stack_end = + (unsigned long *)(per_cpu(irq_stack_ptr, cpu)); + unsigned long *irq_stack = + (unsigned long *)(per_cpu(irq_stack_ptr, cpu) - IRQ_STACK_SIZE); /* * debugging aid: "show_stack(NULL, NULL);" prints the @@ -218,9 +219,9 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, stack = sp; for (i = 0; i < kstack_depth_to_print; i++) { - if (stack >= irqstack && stack <= irqstack_end) { - if (stack == irqstack_end) { - stack = (unsigned long *) (irqstack_end[-1]); + if (stack >= irq_stack && stack <= irq_stack_end) { + if (stack == irq_stack_end) { + stack = (unsigned long *) (irq_stack_end[-1]); printk(" "); } } else { diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 4833f3a1965..d22677a6643 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -345,7 +345,7 @@ ENTRY(save_args) 1: incl %gs:pda_irqcount jne 2f popq_cfi %rax /* move return address... */ - mov %gs:pda_irqstackptr,%rsp + mov PER_CPU_VAR(irq_stack_ptr),%rsp EMPTY_FRAME 0 pushq_cfi %rax /* ... to the new stack */ /* @@ -1261,7 +1261,7 @@ ENTRY(call_softirq) mov %rsp,%rbp CFI_DEF_CFA_REGISTER rbp incl %gs:pda_irqcount - cmove %gs:pda_irqstackptr,%rsp + cmove PER_CPU_VAR(irq_stack_ptr),%rsp push %rbp # backlink for old unwinder call __do_softirq leaveq @@ -1300,7 +1300,7 @@ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) 11: incl %gs:pda_irqcount movq %rsp,%rbp CFI_DEF_CFA_REGISTER rbp - cmovzq %gs:pda_irqstackptr,%rsp + cmovzq PER_CPU_VAR(irq_stack_ptr),%rsp pushq %rbp # backlink for old unwinder call xen_evtchn_do_upcall popq %rsp diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index b5c35af2011..8b53ef83c61 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -192,7 +192,10 @@ void __init setup_per_cpu_areas(void) memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); per_cpu_offset(cpu) = ptr - __per_cpu_start; + per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); #ifdef CONFIG_X86_64 + per_cpu(irq_stack_ptr, cpu) = + (char *)per_cpu(irq_stack, cpu) + IRQ_STACK_SIZE - 64; /* * CPU0 modified pda in the init data area, reload pda * offset for CPU0 and clear the area for others. @@ -202,7 +205,6 @@ void __init setup_per_cpu_areas(void) else memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu))); #endif - per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } -- cgit v1.2.3 From 92d65b2371d86d40807e1dbfdccadc4d501edcde Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Convert exception stacks to per-cpu Move the exception stacks to per-cpu, removing specific allocation code. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 23 ++++++++--------------- 1 file changed, 8 insertions(+), 15 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 496f0a01919..b6d7eec0be7 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -913,8 +913,9 @@ void __cpuinit pda_init(int cpu) } } -static char boot_exception_stacks[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + - DEBUG_STKSZ] __page_aligned_bss; +static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks + [(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + DEBUG_STKSZ]) + __aligned(PAGE_SIZE); extern asmlinkage void ignore_sysret(void); @@ -972,15 +973,12 @@ void __cpuinit cpu_init(void) struct tss_struct *t = &per_cpu(init_tss, cpu); struct orig_ist *orig_ist = &per_cpu(orig_ist, cpu); unsigned long v; - char *estacks = NULL; struct task_struct *me; int i; /* CPU 0 is initialised in head64.c */ if (cpu != 0) pda_init(cpu); - else - estacks = boot_exception_stacks; me = current; @@ -1014,18 +1012,13 @@ void __cpuinit cpu_init(void) * set up and load the per-CPU TSS */ if (!orig_ist->ist[0]) { - static const unsigned int order[N_EXCEPTION_STACKS] = { - [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER, - [DEBUG_STACK - 1] = DEBUG_STACK_ORDER + static const unsigned int sizes[N_EXCEPTION_STACKS] = { + [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STKSZ, + [DEBUG_STACK - 1] = DEBUG_STKSZ }; + char *estacks = per_cpu(exception_stacks, cpu); for (v = 0; v < N_EXCEPTION_STACKS; v++) { - if (cpu) { - estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]); - if (!estacks) - panic("Cannot allocate exception " - "stack %ld %d\n", v, cpu); - } - estacks += PAGE_SIZE << order[v]; + estacks += sizes[v]; orig_ist->ist[v] = t->x86_tss.ist[v] = (unsigned long)estacks; } -- cgit v1.2.3 From ea9279066de44053d0c20ea855bc9f4706652d84 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move cpu number from PDA to per-cpu and consolidate with 32-bit. tj: moved cpu_number definition out of CONFIG_HAVE_SETUP_PER_CPU_AREA for voyager. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 1 - arch/x86/kernel/process_32.c | 3 --- arch/x86/kernel/setup_percpu.c | 10 ++++++++++ arch/x86/kernel/smpcommon.c | 2 -- 5 files changed, 10 insertions(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 5b821fbdaf7..cae6697c099 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -53,7 +53,6 @@ int main(void) ENTRY(oldrsp); ENTRY(pcurrent); ENTRY(irqcount); - ENTRY(cpunumber); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); #undef ENTRY diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index b6d7eec0be7..4221e920886 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -899,7 +899,6 @@ void __cpuinit pda_init(int cpu) load_pda_offset(cpu); - pda->cpunumber = cpu; pda->irqcount = -1; pda->kernelstack = (unsigned long)stack_thread_info() - PDA_STACKOFFSET + THREAD_SIZE; diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index 77d546817d9..2c00a57ccb9 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c @@ -66,9 +66,6 @@ asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; EXPORT_PER_CPU_SYMBOL(current_task); -DEFINE_PER_CPU(int, cpu_number); -EXPORT_PER_CPU_SYMBOL(cpu_number); - /* * Return saved PC of a blocked thread. */ diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 8b53ef83c61..258497f93f4 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -22,6 +22,15 @@ # define DBG(x...) #endif +/* + * Could be inside CONFIG_HAVE_SETUP_PER_CPU_AREA with other stuff but + * voyager wants cpu_number too. + */ +#ifdef CONFIG_SMP +DEFINE_PER_CPU(int, cpu_number); +EXPORT_PER_CPU_SYMBOL(cpu_number); +#endif + #ifdef CONFIG_X86_LOCAL_APIC unsigned int num_processors; unsigned disabled_cpus __cpuinitdata; @@ -193,6 +202,7 @@ void __init setup_per_cpu_areas(void) memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); per_cpu_offset(cpu) = ptr - __per_cpu_start; per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); + per_cpu(cpu_number, cpu) = cpu; #ifdef CONFIG_X86_64 per_cpu(irq_stack_ptr, cpu) = (char *)per_cpu(irq_stack, cpu) + IRQ_STACK_SIZE - 64; diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c index 7e157810062..add36b4e37c 100644 --- a/arch/x86/kernel/smpcommon.c +++ b/arch/x86/kernel/smpcommon.c @@ -28,7 +28,5 @@ __cpuinit void init_gdt(int cpu) write_gdt_entry(get_cpu_gdt_table(cpu), GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S); - - per_cpu(cpu_number, cpu) = cpu; } #endif -- cgit v1.2.3 From c6f5e0acd5d12ee23f701f15889872e67b47caa6 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move current task from PDA to per-cpu and consolidate with 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 5 +---- arch/x86/kernel/dumpstack_64.c | 2 +- arch/x86/kernel/process_64.c | 5 ++++- arch/x86/kernel/smpboot.c | 3 +-- 5 files changed, 7 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index cae6697c099..4f7a210e1e5 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -51,7 +51,6 @@ int main(void) #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) ENTRY(kernelstack); ENTRY(oldrsp); - ENTRY(pcurrent); ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 4221e920886..b50e38d1688 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -903,10 +903,7 @@ void __cpuinit pda_init(int cpu) pda->kernelstack = (unsigned long)stack_thread_info() - PDA_STACKOFFSET + THREAD_SIZE; - if (cpu == 0) { - /* others are initialized in smpboot.c */ - pda->pcurrent = &init_task; - } else { + if (cpu != 0) { if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) pda->nodenumber = cpu_to_node(cpu); } diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c index 28e26a4315d..d35db5993fd 100644 --- a/arch/x86/kernel/dumpstack_64.c +++ b/arch/x86/kernel/dumpstack_64.c @@ -242,7 +242,7 @@ void show_registers(struct pt_regs *regs) int i; unsigned long sp; const int cpu = smp_processor_id(); - struct task_struct *cur = cpu_pda(cpu)->pcurrent; + struct task_struct *cur = current; sp = regs->sp; printk("CPU %d ", cpu); diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 416fb9282f4..e00c31a4b3c 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -57,6 +57,9 @@ asmlinkage extern void ret_from_fork(void); +DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; +EXPORT_PER_CPU_SYMBOL(current_task); + unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; static ATOMIC_NOTIFIER_HEAD(idle_notifier); @@ -615,7 +618,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) */ prev->usersp = read_pda(oldrsp); write_pda(oldrsp, next->usersp); - write_pda(pcurrent, next_p); + percpu_write(current_task, next_p); write_pda(kernelstack, (unsigned long)task_stack_page(next_p) + diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 2f0e0f1090f..5854be0fb80 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -790,13 +790,12 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu) set_idle_for_cpu(cpu, c_idle.idle); do_rest: -#ifdef CONFIG_X86_32 per_cpu(current_task, cpu) = c_idle.idle; +#ifdef CONFIG_X86_32 init_gdt(cpu); /* Stack for startup_32 can be just as for start_secondary onwards */ irq_ctx_init(cpu); #else - cpu_pda(cpu)->pcurrent = c_idle.idle; clear_tsk_thread_flag(c_idle.idle, TIF_FORK); initial_gs = per_cpu_offset(cpu); #endif -- cgit v1.2.3 From 9af45651f1f7c89942e016a1a00a7ebddfa727f8 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move kernelstack from PDA to per-cpu. Also clean up PER_CPU_VAR usage in xen-asm_64.S tj: * remove now unused stack_thread_info() * s/kernelstack/kernel_stack/ * added FIXME comment in xen-asm_64.S Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 6 ++++-- arch/x86/kernel/entry_64.S | 4 ++-- arch/x86/kernel/process_64.c | 4 ++-- arch/x86/kernel/smpboot.c | 3 +++ 5 files changed, 11 insertions(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 4f7a210e1e5..cafff5f4a03 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -49,7 +49,6 @@ int main(void) BLANK(); #undef ENTRY #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) - ENTRY(kernelstack); ENTRY(oldrsp); ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index b50e38d1688..06b6290088f 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -889,6 +889,10 @@ DEFINE_PER_CPU(char *, irq_stack_ptr) = per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64; #endif +DEFINE_PER_CPU(unsigned long, kernel_stack) = + (unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE; +EXPORT_PER_CPU_SYMBOL(kernel_stack); + void __cpuinit pda_init(int cpu) { struct x8664_pda *pda = cpu_pda(cpu); @@ -900,8 +904,6 @@ void __cpuinit pda_init(int cpu) load_pda_offset(cpu); pda->irqcount = -1; - pda->kernelstack = (unsigned long)stack_thread_info() - - PDA_STACKOFFSET + THREAD_SIZE; if (cpu != 0) { if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index d22677a6643..0dd45859a7a 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -468,7 +468,7 @@ END(ret_from_fork) ENTRY(system_call) CFI_STARTPROC simple CFI_SIGNAL_FRAME - CFI_DEF_CFA rsp,PDA_STACKOFFSET + CFI_DEF_CFA rsp,KERNEL_STACK_OFFSET CFI_REGISTER rip,rcx /*CFI_REGISTER rflags,r11*/ SWAPGS_UNSAFE_STACK @@ -480,7 +480,7 @@ ENTRY(system_call) ENTRY(system_call_after_swapgs) movq %rsp,%gs:pda_oldrsp - movq %gs:pda_kernelstack,%rsp + movq PER_CPU_VAR(kernel_stack),%rsp /* * No need to follow this irqs off/on section - it's straight * and short: diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index e00c31a4b3c..6c5f5760210 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -620,9 +620,9 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) write_pda(oldrsp, next->usersp); percpu_write(current_task, next_p); - write_pda(kernelstack, + percpu_write(kernel_stack, (unsigned long)task_stack_page(next_p) + - THREAD_SIZE - PDA_STACKOFFSET); + THREAD_SIZE - KERNEL_STACK_OFFSET); #ifdef CONFIG_CC_STACKPROTECTOR write_pda(stack_canary, next_p->stack_canary); /* diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 5854be0fb80..869b98840fd 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -798,6 +798,9 @@ do_rest: #else clear_tsk_thread_flag(c_idle.idle, TIF_FORK); initial_gs = per_cpu_offset(cpu); + per_cpu(kernel_stack, cpu) = + (unsigned long)task_stack_page(c_idle.idle) - + KERNEL_STACK_OFFSET + THREAD_SIZE; #endif early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); initial_code = (unsigned long)start_secondary; -- cgit v1.2.3 From 3d1e42a7cf945e289d6ba26159aa0e2b0645401b Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move oldrsp from PDA to per-cpu. tj: * in asm-offsets_64.c, pda.h inclusion shouldn't be removed as pda is still referenced in the file * s/oldrsp/old_rsp/ Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/entry_64.S | 10 +++++----- arch/x86/kernel/process_64.c | 8 +++++--- 3 files changed, 10 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index cafff5f4a03..afda6deb851 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -49,7 +49,6 @@ int main(void) BLANK(); #undef ENTRY #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) - ENTRY(oldrsp); ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 0dd45859a7a..7c27da407da 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -210,7 +210,7 @@ ENTRY(native_usergs_sysret64) /* %rsp:at FRAMEEND */ .macro FIXUP_TOP_OF_STACK tmp offset=0 - movq %gs:pda_oldrsp,\tmp + movq PER_CPU_VAR(old_rsp),\tmp movq \tmp,RSP+\offset(%rsp) movq $__USER_DS,SS+\offset(%rsp) movq $__USER_CS,CS+\offset(%rsp) @@ -221,7 +221,7 @@ ENTRY(native_usergs_sysret64) .macro RESTORE_TOP_OF_STACK tmp offset=0 movq RSP+\offset(%rsp),\tmp - movq \tmp,%gs:pda_oldrsp + movq \tmp,PER_CPU_VAR(old_rsp) movq EFLAGS+\offset(%rsp),\tmp movq \tmp,R11+\offset(%rsp) .endm @@ -479,7 +479,7 @@ ENTRY(system_call) */ ENTRY(system_call_after_swapgs) - movq %rsp,%gs:pda_oldrsp + movq %rsp,PER_CPU_VAR(old_rsp) movq PER_CPU_VAR(kernel_stack),%rsp /* * No need to follow this irqs off/on section - it's straight @@ -523,7 +523,7 @@ sysret_check: CFI_REGISTER rip,rcx RESTORE_ARGS 0,-ARG_SKIP,1 /*CFI_REGISTER rflags,r11*/ - movq %gs:pda_oldrsp, %rsp + movq PER_CPU_VAR(old_rsp), %rsp USERGS_SYSRET64 CFI_RESTORE_STATE @@ -833,7 +833,7 @@ common_interrupt: XCPT_FRAME addq $-0x80,(%rsp) /* Adjust vector to [-256,-1] range */ interrupt do_IRQ - /* 0(%rsp): oldrsp-ARGOFFSET */ + /* 0(%rsp): old_rsp-ARGOFFSET */ ret_from_intr: DISABLE_INTERRUPTS(CLBR_NONE) TRACE_IRQS_OFF diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 6c5f5760210..48012891892 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -60,6 +60,8 @@ asmlinkage extern void ret_from_fork(void); DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; EXPORT_PER_CPU_SYMBOL(current_task); +DEFINE_PER_CPU(unsigned long, old_rsp); + unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; static ATOMIC_NOTIFIER_HEAD(idle_notifier); @@ -395,7 +397,7 @@ start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) load_gs_index(0); regs->ip = new_ip; regs->sp = new_sp; - write_pda(oldrsp, new_sp); + percpu_write(old_rsp, new_sp); regs->cs = __USER_CS; regs->ss = __USER_DS; regs->flags = 0x200; @@ -616,8 +618,8 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) /* * Switch the PDA and FPU contexts. */ - prev->usersp = read_pda(oldrsp); - write_pda(oldrsp, next->usersp); + prev->usersp = percpu_read(old_rsp); + percpu_write(old_rsp, next->usersp); percpu_write(current_task, next_p); percpu_write(kernel_stack, -- cgit v1.2.3 From 5689553076c4a67b83426b076082c63085b7567a Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move irqcount from PDA to per-cpu. tj: s/irqcount/irq_count/ Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 4 ++-- arch/x86/kernel/entry_64.S | 14 +++++++------- 3 files changed, 9 insertions(+), 10 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index afda6deb851..64c834a39aa 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -49,7 +49,6 @@ int main(void) BLANK(); #undef ENTRY #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) - ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); #undef ENTRY diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 06b6290088f..e2323ecce1d 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -893,6 +893,8 @@ DEFINE_PER_CPU(unsigned long, kernel_stack) = (unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE; EXPORT_PER_CPU_SYMBOL(kernel_stack); +DEFINE_PER_CPU(unsigned int, irq_count) = -1; + void __cpuinit pda_init(int cpu) { struct x8664_pda *pda = cpu_pda(cpu); @@ -903,8 +905,6 @@ void __cpuinit pda_init(int cpu) load_pda_offset(cpu); - pda->irqcount = -1; - if (cpu != 0) { if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) pda->nodenumber = cpu_to_node(cpu); diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 7c27da407da..c52b6091916 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -337,12 +337,12 @@ ENTRY(save_args) je 1f SWAPGS /* - * irqcount is used to check if a CPU is already on an interrupt stack + * irq_count is used to check if a CPU is already on an interrupt stack * or not. While this is essentially redundant with preempt_count it is * a little cheaper to use a separate counter in the PDA (short of * moving irq_enter into assembly, which would be too much work) */ -1: incl %gs:pda_irqcount +1: incl PER_CPU_VAR(irq_count) jne 2f popq_cfi %rax /* move return address... */ mov PER_CPU_VAR(irq_stack_ptr),%rsp @@ -837,7 +837,7 @@ common_interrupt: ret_from_intr: DISABLE_INTERRUPTS(CLBR_NONE) TRACE_IRQS_OFF - decl %gs:pda_irqcount + decl PER_CPU_VAR(irq_count) leaveq CFI_DEF_CFA_REGISTER rsp CFI_ADJUST_CFA_OFFSET -8 @@ -1260,14 +1260,14 @@ ENTRY(call_softirq) CFI_REL_OFFSET rbp,0 mov %rsp,%rbp CFI_DEF_CFA_REGISTER rbp - incl %gs:pda_irqcount + incl PER_CPU_VAR(irq_count) cmove PER_CPU_VAR(irq_stack_ptr),%rsp push %rbp # backlink for old unwinder call __do_softirq leaveq CFI_DEF_CFA_REGISTER rsp CFI_ADJUST_CFA_OFFSET -8 - decl %gs:pda_irqcount + decl PER_CPU_VAR(irq_count) ret CFI_ENDPROC END(call_softirq) @@ -1297,7 +1297,7 @@ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) movq %rdi, %rsp # we don't return, adjust the stack frame CFI_ENDPROC DEFAULT_FRAME -11: incl %gs:pda_irqcount +11: incl PER_CPU_VAR(irq_count) movq %rsp,%rbp CFI_DEF_CFA_REGISTER rbp cmovzq PER_CPU_VAR(irq_stack_ptr),%rsp @@ -1305,7 +1305,7 @@ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) call xen_evtchn_do_upcall popq %rsp CFI_DEF_CFA_REGISTER rsp - decl %gs:pda_irqcount + decl PER_CPU_VAR(irq_count) jmp error_exit CFI_ENDPROC END(do_hypervisor_callback) -- cgit v1.2.3 From e7a22c1ebcc1caa8178df1819d05128bb5b45ab9 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:59 +0900 Subject: x86-64: Move nodenumber from PDA to per-cpu. tj: * s/nodenumber/node_number/ * removed now unused pda variable from pda_init() Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 13 ++++++------- arch/x86/kernel/setup_percpu.c | 4 +++- 2 files changed, 9 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index e2323ecce1d..7976a6a0f65 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -897,18 +897,11 @@ DEFINE_PER_CPU(unsigned int, irq_count) = -1; void __cpuinit pda_init(int cpu) { - struct x8664_pda *pda = cpu_pda(cpu); - /* Setup up data that may be needed in __get_free_pages early */ loadsegment(fs, 0); loadsegment(gs, 0); load_pda_offset(cpu); - - if (cpu != 0) { - if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) - pda->nodenumber = cpu_to_node(cpu); - } } static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks @@ -978,6 +971,12 @@ void __cpuinit cpu_init(void) if (cpu != 0) pda_init(cpu); +#ifdef CONFIG_NUMA + if (cpu != 0 && percpu_read(node_number) == 0 && + cpu_to_node(cpu) != NUMA_NO_NODE) + percpu_write(node_number, cpu_to_node(cpu)); +#endif + me = current; if (cpumask_test_and_set_cpu(cpu, cpu_initialized_mask)) diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 258497f93f4..efbafbbff58 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -53,6 +53,8 @@ EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); #if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) #define X86_64_NUMA 1 /* (used later) */ +DEFINE_PER_CPU(int, node_number) = 0; +EXPORT_PER_CPU_SYMBOL(node_number); /* * Map cpu index to node index @@ -283,7 +285,7 @@ void __cpuinit numa_set_node(int cpu, int node) per_cpu(x86_cpu_to_node_map, cpu) = node; if (node != NUMA_NO_NODE) - cpu_pda(cpu)->nodenumber = node; + per_cpu(node_number, cpu) = node; } void __cpuinit numa_clear_node(int cpu) -- cgit v1.2.3 From c2558e0eba66b49993e619da66c95a50a97830a3 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:59 +0900 Subject: x86-64: Move isidle from PDA to per-cpu. tj: s/isidle/is_idle/ Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/process_64.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 48012891892..4523ff88a69 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -61,6 +61,7 @@ DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; EXPORT_PER_CPU_SYMBOL(current_task); DEFINE_PER_CPU(unsigned long, old_rsp); +static DEFINE_PER_CPU(unsigned char, is_idle); unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; @@ -80,13 +81,13 @@ EXPORT_SYMBOL_GPL(idle_notifier_unregister); void enter_idle(void) { - write_pda(isidle, 1); + percpu_write(is_idle, 1); atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); } static void __exit_idle(void) { - if (test_and_clear_bit_pda(0, isidle) == 0) + if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) return; atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); } -- cgit v1.2.3 From 5cdc5e9e69d4dc3a3630ae1fa666401b2a8dcde6 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 19 Jan 2009 20:49:37 +0100 Subject: x86: fully honor "nolapic", fix Impact: build fix Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index 48578795583..9ca12af6c87 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -1131,7 +1131,9 @@ void __cpuinit setup_local_APIC(void) int i, j; if (disable_apic) { +#ifdef CONFIG_X86_IO_APIC disable_ioapic_setup(); +#endif return; } -- cgit v1.2.3 From c6e50f93db5bd0895ec7c7d1b6f3886c6e1f11b6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 20 Jan 2009 12:29:19 +0900 Subject: x86: cleanup stack protector Impact: cleanup Make the following cleanups. * remove duplicate comment from boot_init_stack_canary() which fits better in the other place - cpu_idle(). * move stack_canary offset check from __switch_to() to boot_init_stack_canary(). Signed-off-by: Tejun Heo --- arch/x86/kernel/process_64.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index aa89eabf09e..088bc9a0f82 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -638,13 +638,6 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) percpu_write(kernel_stack, (unsigned long)task_stack_page(next_p) + THREAD_SIZE - KERNEL_STACK_OFFSET); -#ifdef CONFIG_CC_STACKPROTECTOR - /* - * Build time only check to make sure the stack_canary is at - * offset 40 in the pda; this is a gcc ABI requirement - */ - BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); -#endif /* * Now maybe reload the debug registers and handle I/O bitmaps -- cgit v1.2.3 From 8ce031972b40da58c268caba8c5ea3c0856d7131 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:27 +0900 Subject: x86: remove pda_init() Impact: cleanup Copy the code to cpu_init() to satisfy the requirement that the cpu be reinitialized. Remove all other calls, since the segments are already initialized in head_64.S. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 15 +++------------ arch/x86/kernel/head64.c | 2 -- 2 files changed, 3 insertions(+), 14 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 7976a6a0f65..f83a4d6160f 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -895,15 +895,6 @@ EXPORT_PER_CPU_SYMBOL(kernel_stack); DEFINE_PER_CPU(unsigned int, irq_count) = -1; -void __cpuinit pda_init(int cpu) -{ - /* Setup up data that may be needed in __get_free_pages early */ - loadsegment(fs, 0); - loadsegment(gs, 0); - - load_pda_offset(cpu); -} - static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks [(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + DEBUG_STKSZ]) __aligned(PAGE_SIZE); @@ -967,9 +958,9 @@ void __cpuinit cpu_init(void) struct task_struct *me; int i; - /* CPU 0 is initialised in head64.c */ - if (cpu != 0) - pda_init(cpu); + loadsegment(fs, 0); + loadsegment(gs, 0); + load_pda_offset(cpu); #ifdef CONFIG_NUMA if (cpu != 0 && percpu_read(node_number) == 0 && diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index af67d3227ea..f5b27224769 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -91,8 +91,6 @@ void __init x86_64_start_kernel(char * real_mode_data) if (console_loglevel == 10) early_printk("Kernel alive\n"); - pda_init(0); - x86_64_start_reservations(real_mode_data); } -- cgit v1.2.3 From 8c7e58e690ae60ab4215b025f433ed4af261e103 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:28 +0900 Subject: x86: rework __per_cpu_load adjustments Impact: cleanup Use cpu_number to determine if the adjustment is necessary. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/head_64.S | 24 ++++++++---------------- 1 file changed, 8 insertions(+), 16 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index c8ace880661..98ea26a2fca 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -207,19 +207,15 @@ ENTRY(secondary_startup_64) #ifdef CONFIG_SMP /* - * early_gdt_base should point to the gdt_page in static percpu init - * data area. Computing this requires two symbols - __per_cpu_load - * and per_cpu__gdt_page. As linker can't do no such relocation, do - * it by hand. As early_gdt_descr is manipulated by C code for - * secondary CPUs, this should be done only once for the boot CPU - * when early_gdt_descr_base contains zero. + * Fix up static pointers that need __per_cpu_load added. The assembler + * is unable to do this directly. This is only needed for the boot cpu. + * These values are set up with the correct base addresses by C code for + * secondary cpus. */ - movq early_gdt_descr_base(%rip), %rax - testq %rax, %rax - jnz 1f - movq $__per_cpu_load, %rax - addq $per_cpu__gdt_page, %rax - movq %rax, early_gdt_descr_base(%rip) + movq initial_gs(%rip), %rax + cmpl $0, per_cpu__cpu_number(%rax) + jne 1f + addq %rax, early_gdt_descr_base(%rip) 1: #endif /* @@ -431,12 +427,8 @@ NEXT_PAGE(level2_spare_pgt) .globl early_gdt_descr early_gdt_descr: .word GDT_ENTRIES*8-1 -#ifdef CONFIG_SMP early_gdt_descr_base: - .quad 0x0000000000000000 -#else .quad per_cpu__gdt_page -#endif ENTRY(phys_base) /* This must match the first entry in level2_kernel_pgt */ -- cgit v1.2.3 From 947e76cdc34c782fc947313d4331380686eebbad Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:28 +0900 Subject: x86: move stack_canary into irq_stack Impact: x86_64 percpu area layout change, irq_stack now at the beginning Now that the PDA is empty except for the stack canary, it can be removed. The irqstack is moved to the start of the per-cpu section. If the stack protector is enabled, the canary overlaps the bottom 48 bytes of the irqstack. tj: * updated subject * dropped asm relocation of irq_stack_ptr * updated comments a bit * rebased on top of stack canary changes Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/asm-offsets_64.c | 4 ---- arch/x86/kernel/cpu/common.c | 7 ++++--- arch/x86/kernel/head_64.S | 13 +++++-------- arch/x86/kernel/setup_percpu.c | 34 ++++------------------------------ arch/x86/kernel/vmlinux_64.lds.S | 8 ++++++-- 5 files changed, 19 insertions(+), 47 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 64c834a39aa..94f9c8b39d2 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -48,10 +48,6 @@ int main(void) #endif BLANK(); #undef ENTRY -#define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) - DEFINE(pda_size, sizeof(struct x8664_pda)); - BLANK(); -#undef ENTRY #ifdef CONFIG_PARAVIRT BLANK(); OFFSET(PARAVIRT_enabled, pv_info, paravirt_enabled); diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index f83a4d6160f..098934e72a1 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -881,12 +881,13 @@ __setup("clearcpuid=", setup_disablecpuid); #ifdef CONFIG_X86_64 struct desc_ptr idt_descr = { 256 * 16 - 1, (unsigned long) idt_table }; -DEFINE_PER_CPU_PAGE_ALIGNED(char[IRQ_STACK_SIZE], irq_stack); +DEFINE_PER_CPU_FIRST(union irq_stack_union, + irq_stack_union) __aligned(PAGE_SIZE); #ifdef CONFIG_SMP DEFINE_PER_CPU(char *, irq_stack_ptr); /* will be set during per cpu init */ #else DEFINE_PER_CPU(char *, irq_stack_ptr) = - per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64; + per_cpu_var(irq_stack_union.irq_stack) + IRQ_STACK_SIZE - 64; #endif DEFINE_PER_CPU(unsigned long, kernel_stack) = @@ -960,7 +961,7 @@ void __cpuinit cpu_init(void) loadsegment(fs, 0); loadsegment(gs, 0); - load_pda_offset(cpu); + load_gs_base(cpu); #ifdef CONFIG_NUMA if (cpu != 0 && percpu_read(node_number) == 0 && diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 98ea26a2fca..a0a2b5ca9b7 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -242,13 +242,10 @@ ENTRY(secondary_startup_64) /* Set up %gs. * - * On SMP, %gs should point to the per-cpu area. For initial - * boot, make %gs point to the init data section. For a - * secondary CPU,initial_gs should be set to its pda address - * before the CPU runs this code. - * - * On UP, initial_gs points to PER_CPU_VAR(__pda) and doesn't - * change. + * The base of %gs always points to the bottom of the irqstack + * union. If the stack protector canary is enabled, it is + * located at %gs:40. Note that, on SMP, the boot cpu uses + * init data section till per cpu areas are set up. */ movl $MSR_GS_BASE,%ecx movq initial_gs(%rip),%rax @@ -281,7 +278,7 @@ ENTRY(secondary_startup_64) #ifdef CONFIG_SMP .quad __per_cpu_load #else - .quad PER_CPU_VAR(__pda) + .quad PER_CPU_VAR(irq_stack_union) #endif __FINITDATA diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index efbafbbff58..90b8e154bb5 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -77,30 +77,6 @@ static void __init setup_node_to_cpumask_map(void); static inline void setup_node_to_cpumask_map(void) { } #endif -/* - * Define load_pda_offset() and per-cpu __pda for x86_64. - * load_pda_offset() is responsible for loading the offset of pda into - * %gs. - * - * On SMP, pda offset also duals as percpu base address and thus it - * should be at the start of per-cpu area. To achieve this, it's - * preallocated in vmlinux_64.lds.S directly instead of using - * DEFINE_PER_CPU(). - */ -#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(); -} -#ifndef CONFIG_SMP -DEFINE_PER_CPU(struct x8664_pda, __pda); -#endif -EXPORT_PER_CPU_SYMBOL(__pda); -#endif /* CONFIG_SMP && CONFIG_X86_64 */ - #ifdef CONFIG_X86_64 /* correctly size the local cpu masks */ @@ -207,15 +183,13 @@ void __init setup_per_cpu_areas(void) per_cpu(cpu_number, cpu) = cpu; #ifdef CONFIG_X86_64 per_cpu(irq_stack_ptr, cpu) = - (char *)per_cpu(irq_stack, cpu) + IRQ_STACK_SIZE - 64; + per_cpu(irq_stack_union.irq_stack, cpu) + IRQ_STACK_SIZE - 64; /* - * CPU0 modified pda in the init data area, reload pda - * offset for CPU0 and clear the area for others. + * Up to this point, CPU0 has been using .data.init + * area. Reload %gs offset for CPU0. */ if (cpu == 0) - load_pda_offset(0); - else - memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu))); + load_gs_base(cpu); #endif 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 a09abb8fb97..c9740996430 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -220,8 +220,7 @@ SECTIONS * so that it can be accessed as a percpu variable. */ . = ALIGN(PAGE_SIZE); - PERCPU_VADDR_PREALLOC(0, :percpu, pda_size) - per_cpu____pda = __per_cpu_start; + PERCPU_VADDR(0, :percpu) #else PERCPU(PAGE_SIZE) #endif @@ -262,3 +261,8 @@ SECTIONS */ ASSERT((_end - _text <= KERNEL_IMAGE_SIZE), "kernel image bigger than KERNEL_IMAGE_SIZE") + +#ifdef CONFIG_SMP +ASSERT((per_cpu__irq_stack_union == 0), + "irq_stack_union is not at start of per-cpu area"); +#endif -- cgit v1.2.3 From 0d974d4592708f85044751817da4b7016e1b0602 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Sun, 18 Jan 2009 19:52:25 -0500 Subject: x86: remove pda.h Impact: cleanup Signed-off-by: Brian Gerst --- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 1 - arch/x86/kernel/process_64.c | 1 - arch/x86/kernel/traps.c | 1 - 4 files changed, 4 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 94f9c8b39d2..8793ab33e2c 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -11,7 +11,6 @@ #include #include #include -#include #include #include #include diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 098934e72a1..3887fcf6e51 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -30,7 +30,6 @@ #include #endif -#include #include #include #include diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 088bc9a0f82..c422eebb0c5 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -47,7 +47,6 @@ #include #include #include -#include #include #include #include diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index 98c2d055284..ed5aee5f3fc 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c @@ -59,7 +59,6 @@ #ifdef CONFIG_X86_64 #include #include -#include #else #include #include -- cgit v1.2.3 From 5766b842b23c6b40935a5f3bd435b2bcdaff2143 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 20 Jan 2009 09:13:15 +0100 Subject: x86, cpumask: fix tlb flush race Impact: fix bootup crash The cpumask is now passed in as a reference to mm->cpu_vm_mask, not on the stack - hence it is not constant anymore during the TLB flush. That way it could race and some static sanity checks would trigger: [ 238.154287] ------------[ cut here ]------------ [ 238.156039] kernel BUG at arch/x86/kernel/tlb_32.c:130! [ 238.156039] invalid opcode: 0000 [#1] SMP [ 238.156039] last sysfs file: /sys/class/net/eth2/address [ 238.156039] Modules linked in: [ 238.156039] [ 238.156039] Pid: 6493, comm: ifup-eth Not tainted (2.6.29-rc2-tip #1) P4DC6 [ 238.156039] EIP: 0060:[] EFLAGS: 00010202 CPU: 2 [ 238.156039] EIP is at native_flush_tlb_others+0x35/0x158 [ 238.156039] EAX: c0ef972c EBX: f6143301 ECX: 00000000 EDX: 00000000 [ 238.156039] ESI: f61433a8 EDI: f6143200 EBP: f34f3e00 ESP: f34f3df0 [ 238.156039] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 [ 238.156039] Process ifup-eth (pid: 6493, ti=f34f2000 task=f399ab00 task.ti=f34f2000) [ 238.156039] Stack: [ 238.156039] ffffffff f61433a8 ffffffff f6143200 f34f3e18 c0118e9c 00000000 f6143200 [ 238.156039] f61433a8 f5bec738 f34f3e28 c0119435 c2b5b830 f6143200 f34f3e34 c01c2dc3 [ 238.156039] bffd9000 f34f3e60 c01c3051 00000000 ffffffff f34f3e4c 00000000 00000071 [ 238.156039] Call Trace: [ 238.156039] [] ? flush_tlb_others+0x52/0x5b [ 238.156039] [] ? flush_tlb_mm+0x7f/0x8b [ 238.156039] [] ? tlb_finish_mmu+0x2d/0x55 [ 238.156039] [] ? exit_mmap+0x124/0x170 [ 238.156039] [] ? mmput+0x40/0xf5 [ 238.156039] [] ? flush_old_exec+0x640/0x94b [ 238.156039] [] ? fsnotify_access+0x37/0x39 [ 238.156039] [] ? kernel_read+0x39/0x4b [ 238.156039] [] ? load_elf_binary+0x4a1/0x11bb [ 238.156039] [] ? might_fault+0x51/0x9c [ 238.156039] [] ? paravirt_read_tsc+0x20/0x4f [ 238.156039] [] ? native_sched_clock+0x5d/0x60 [ 238.156039] [] ? search_binary_handler+0xab/0x2c4 [ 238.156039] [] ? load_elf_binary+0x0/0x11bb [ 238.156039] [] ? _raw_read_unlock+0x21/0x46 [ 238.156039] [] ? load_elf_binary+0x0/0x11bb [ 238.156039] [] ? search_binary_handler+0xb2/0x2c4 [ 238.156039] [] ? do_execve+0x21c/0x2ee [ 238.156039] [] ? sys_execve+0x51/0x8c [ 238.156039] [] ? sysenter_do_call+0x12/0x43 Fix it by not assuming that the cpumask is constant. Signed-off-by: Ingo Molnar --- arch/x86/kernel/tlb_32.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index ec53818f4e3..d37bbfcb813 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -125,9 +125,8 @@ void native_flush_tlb_others(const struct cpumask *cpumask, struct mm_struct *mm, unsigned long va) { /* - * - mask must exist :) + * mm must exist :) */ - BUG_ON(cpumask_empty(cpumask)); BUG_ON(!mm); /* @@ -138,14 +137,18 @@ void native_flush_tlb_others(const struct cpumask *cpumask, spin_lock(&tlbstate_lock); cpumask_andnot(flush_cpumask, cpumask, cpumask_of(smp_processor_id())); -#ifdef CONFIG_HOTPLUG_CPU - /* If a CPU which we ran on has gone down, OK. */ cpumask_and(flush_cpumask, flush_cpumask, cpu_online_mask); + + /* + * If a task whose mm mask we are looking at has descheduled and + * has cleared its presence from the mask, or if a CPU which we ran + * on has gone down then there might be no flush work left: + */ if (unlikely(cpumask_empty(flush_cpumask))) { spin_unlock(&tlbstate_lock); return; } -#endif + flush_mm = mm; flush_va = va; -- cgit v1.2.3 From 06deef892c7327992434917fb6592c233430803d Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Wed, 21 Jan 2009 17:26:05 +0900 Subject: x86: clean up gdt_page definition Impact: cleanup && more compact percpu area layout with future changes Move 64-bit GDT to page-aligned section and clean up comment formatting. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 3887fcf6e51..a8f0dedcb0c 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -63,23 +63,23 @@ cpumask_t cpu_sibling_setup_map; static struct cpu_dev *this_cpu __cpuinitdata; +DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { #ifdef CONFIG_X86_64 -/* We need valid kernel segments for data and code in long mode too - * IRET will check the segment types kkeil 2000/10/28 - * Also sysret mandates a special GDT layout - */ -/* The TLS descriptors are currently at a different place compared to i386. - Hopefully nobody expects them at a fixed place (Wine?) */ -DEFINE_PER_CPU(struct gdt_page, gdt_page) = { .gdt = { + /* + * We need valid kernel segments for data and code in long mode too + * IRET will check the segment types kkeil 2000/10/28 + * Also sysret mandates a special GDT layout + * + * The TLS descriptors are currently at a different place compared to i386. + * Hopefully nobody expects them at a fixed place (Wine?) + */ [GDT_ENTRY_KERNEL32_CS] = { { { 0x0000ffff, 0x00cf9b00 } } }, [GDT_ENTRY_KERNEL_CS] = { { { 0x0000ffff, 0x00af9b00 } } }, [GDT_ENTRY_KERNEL_DS] = { { { 0x0000ffff, 0x00cf9300 } } }, [GDT_ENTRY_DEFAULT_USER32_CS] = { { { 0x0000ffff, 0x00cffb00 } } }, [GDT_ENTRY_DEFAULT_USER_DS] = { { { 0x0000ffff, 0x00cff300 } } }, [GDT_ENTRY_DEFAULT_USER_CS] = { { { 0x0000ffff, 0x00affb00 } } }, -} }; #else -DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { [GDT_ENTRY_KERNEL_CS] = { { { 0x0000ffff, 0x00cf9a00 } } }, [GDT_ENTRY_KERNEL_DS] = { { { 0x0000ffff, 0x00cf9200 } } }, [GDT_ENTRY_DEFAULT_USER_CS] = { { { 0x0000ffff, 0x00cffa00 } } }, @@ -112,8 +112,8 @@ DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { [GDT_ENTRY_ESPFIX_SS] = { { { 0x00000000, 0x00c09200 } } }, [GDT_ENTRY_PERCPU] = { { { 0x00000000, 0x00000000 } } }, -} }; #endif +} }; EXPORT_PER_CPU_SYMBOL_GPL(gdt_page); #ifdef CONFIG_X86_32 -- cgit v1.2.3 From 0dd76d736eeb3e0ef86c5b103b47ae0e15edebad Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Wed, 21 Jan 2009 17:26:05 +0900 Subject: x86: set %fs to __KERNEL_PERCPU unconditionally for x86_32 Impact: cleanup %fs is currently set to __KERNEL_DS at boot, and conditionally switched to __KERNEL_PERCPU for secondary cpus. Instead, initialize GDT_ENTRY_PERCPU to the same attributes as GDT_ENTRY_KERNEL_DS and set %fs to __KERNEL_PERCPU unconditionally. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 2 +- arch/x86/kernel/head_32.S | 6 +++--- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index a8f0dedcb0c..fbebbcec001 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -111,7 +111,7 @@ DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { [GDT_ENTRY_APMBIOS_BASE+2] = { { { 0x0000ffff, 0x00409200 } } }, [GDT_ENTRY_ESPFIX_SS] = { { { 0x00000000, 0x00c09200 } } }, - [GDT_ENTRY_PERCPU] = { { { 0x00000000, 0x00000000 } } }, + [GDT_ENTRY_PERCPU] = { { { 0x0000ffff, 0x00cf9200 } } }, #endif } }; EXPORT_PER_CPU_SYMBOL_GPL(gdt_page); diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S index e835b4eea70..24c0e5cd71e 100644 --- a/arch/x86/kernel/head_32.S +++ b/arch/x86/kernel/head_32.S @@ -429,12 +429,14 @@ is386: movl $2,%ecx # set MP ljmp $(__KERNEL_CS),$1f 1: movl $(__KERNEL_DS),%eax # reload all the segment registers movl %eax,%ss # after changing gdt. - movl %eax,%fs # gets reset once there's real percpu movl $(__USER_DS),%eax # DS/ES contains default USER segment movl %eax,%ds movl %eax,%es + movl $(__KERNEL_PERCPU), %eax + movl %eax,%fs # set this cpu's percpu + xorl %eax,%eax # Clear GS and LDT movl %eax,%gs lldt %ax @@ -446,8 +448,6 @@ is386: movl $2,%ecx # set MP movb $1, ready cmpb $0,%cl # the first CPU calls start_kernel je 1f - movl $(__KERNEL_PERCPU), %eax - movl %eax,%fs # set this cpu's percpu movl (stack_start), %esp 1: #endif /* CONFIG_SMP */ -- cgit v1.2.3 From d650a5148593b65a3c3f9a344f46b91b7dfe7713 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Wed, 21 Jan 2009 17:26:06 +0900 Subject: x86: merge irq_regs.h Impact: cleanup, better irq_regs code generation for x86_64 Make 64-bit use the same optimizations as 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/irq_64.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c index 1db05247b47..0b254de8408 100644 --- a/arch/x86/kernel/irq_64.c +++ b/arch/x86/kernel/irq_64.c @@ -22,6 +22,9 @@ DEFINE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat); EXPORT_PER_CPU_SYMBOL(irq_stat); +DEFINE_PER_CPU(struct pt_regs *, irq_regs); +EXPORT_PER_CPU_SYMBOL(irq_regs); + /* * Probabilistic stack overflow check: * -- cgit v1.2.3 From bdbcdd48883940bbd8d17eb01172d58a261a413a Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 21 Jan 2009 17:26:06 +0900 Subject: x86: uv cleanup Impact: cleanup Make the following uv related cleanups. * collect visible uv related definitions and interfaces into uv/uv.h and use it. this cleans up the messy situation where on 64bit, uv is defined properly, on 32bit generic it's dummy and on the rest undefined. after this clean up, uv is defined on 64 and dummy on 32. * update uv_flush_tlb_others() such that it takes cpumask of to-be-flushed cpus as argument, instead of that minus self, and returns yet-to-be-flushed cpumask, instead of modifying the passed in parameter. this interface change will ease dummy implementation of uv_flush_tlb_others() and makes uv tlb flush related stuff defined in tlb_uv proper. Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 1 + arch/x86/kernel/genx2apic_uv_x.c | 1 + arch/x86/kernel/smpboot.c | 1 + arch/x86/kernel/tlb_64.c | 18 +++++------ arch/x86/kernel/tlb_uv.c | 68 +++++++++++++++++++++++----------------- 5 files changed, 50 insertions(+), 39 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index fbebbcec001..99904f288d6 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -28,6 +28,7 @@ #include #include #include +#include #endif #include diff --git a/arch/x86/kernel/genx2apic_uv_x.c b/arch/x86/kernel/genx2apic_uv_x.c index b193e082f6c..bfe36249145 100644 --- a/arch/x86/kernel/genx2apic_uv_x.c +++ b/arch/x86/kernel/genx2apic_uv_x.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 869b98840fd..def770b57b5 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -62,6 +62,7 @@ #include #include #include +#include #include #include diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index e64a32c4882..b8bed841ad6 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -15,8 +15,7 @@ #include #include #include -#include -#include +#include DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = { &init_mm, 0, }; @@ -206,16 +205,13 @@ void native_flush_tlb_others(const struct cpumask *cpumask, struct mm_struct *mm, unsigned long va) { if (is_uv_system()) { - /* FIXME: could be an percpu_alloc'd thing */ - static DEFINE_PER_CPU(cpumask_t, flush_tlb_mask); - struct cpumask *after_uv_flush = &get_cpu_var(flush_tlb_mask); + unsigned int cpu; - cpumask_andnot(after_uv_flush, cpumask, - cpumask_of(smp_processor_id())); - if (!uv_flush_tlb_others(after_uv_flush, mm, va)) - flush_tlb_others_ipi(after_uv_flush, mm, va); - - put_cpu_var(flush_tlb_uv_cpumask); + cpu = get_cpu(); + cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu); + if (cpumask) + flush_tlb_others_ipi(cpumask, mm, va); + put_cpu(); return; } flush_tlb_others_ipi(cpumask, mm, va); diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c index 690dcf1a27d..aae15dd7260 100644 --- a/arch/x86/kernel/tlb_uv.c +++ b/arch/x86/kernel/tlb_uv.c @@ -11,6 +11,7 @@ #include #include +#include #include #include #include @@ -209,14 +210,15 @@ static int uv_wait_completion(struct bau_desc *bau_desc, * * Send a broadcast and wait for a broadcast message to complete. * - * The cpumaskp mask contains the cpus the broadcast was sent to. + * The flush_mask contains the cpus the broadcast was sent to. * - * Returns 1 if all remote flushing was done. The mask is zeroed. - * Returns 0 if some remote flushing remains to be done. The mask will have - * some bits still set. + * Returns NULL if all remote flushing was done. The mask is zeroed. + * Returns @flush_mask if some remote flushing remains to be done. The + * mask will have some bits still set. */ -int uv_flush_send_and_wait(int cpu, int this_blade, struct bau_desc *bau_desc, - struct cpumask *cpumaskp) +const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade, + struct bau_desc *bau_desc, + struct cpumask *flush_mask) { int completion_status = 0; int right_shift; @@ -263,59 +265,69 @@ int uv_flush_send_and_wait(int cpu, int this_blade, struct bau_desc *bau_desc, * Success, so clear the remote cpu's from the mask so we don't * use the IPI method of shootdown on them. */ - for_each_cpu(bit, cpumaskp) { + for_each_cpu(bit, flush_mask) { blade = uv_cpu_to_blade_id(bit); if (blade == this_blade) continue; - cpumask_clear_cpu(bit, cpumaskp); + cpumask_clear_cpu(bit, flush_mask); } - if (!cpumask_empty(cpumaskp)) - return 0; - return 1; + if (!cpumask_empty(flush_mask)) + return flush_mask; + return NULL; } /** * uv_flush_tlb_others - globally purge translation cache of a virtual * address or all TLB's - * @cpumaskp: mask of all cpu's in which the address is to be removed + * @cpumask: mask of all cpu's in which the address is to be removed * @mm: mm_struct containing virtual address range * @va: virtual address to be removed (or TLB_FLUSH_ALL for all TLB's on cpu) + * @cpu: the current cpu * * This is the entry point for initiating any UV global TLB shootdown. * * Purges the translation caches of all specified processors of the given * virtual address, or purges all TLB's on specified processors. * - * The caller has derived the cpumaskp from the mm_struct and has subtracted - * the local cpu from the mask. This function is called only if there - * are bits set in the mask. (e.g. flush_tlb_page()) + * The caller has derived the cpumask from the mm_struct. This function + * is called only if there are bits set in the mask. (e.g. flush_tlb_page()) * - * The cpumaskp is converted into a nodemask of the nodes containing + * The cpumask is converted into a nodemask of the nodes containing * the cpus. * - * Returns 1 if all remote flushing was done. - * Returns 0 if some remote flushing remains to be done. + * Note that this function should be called with preemption disabled. + * + * Returns NULL if all remote flushing was done. + * Returns pointer to cpumask if some remote flushing remains to be + * done. The returned pointer is valid till preemption is re-enabled. */ -int uv_flush_tlb_others(struct cpumask *cpumaskp, struct mm_struct *mm, - unsigned long va) +const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask, + struct mm_struct *mm, + unsigned long va, unsigned int cpu) { + static DEFINE_PER_CPU(cpumask_t, flush_tlb_mask); + struct cpumask *flush_mask = &__get_cpu_var(flush_tlb_mask); int i; int bit; int blade; - int cpu; + int uv_cpu; int this_blade; int locals = 0; struct bau_desc *bau_desc; - cpu = uv_blade_processor_id(); + WARN_ON(!in_atomic()); + + cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu)); + + uv_cpu = uv_blade_processor_id(); this_blade = uv_numa_blade_id(); bau_desc = __get_cpu_var(bau_control).descriptor_base; - bau_desc += UV_ITEMS_PER_DESCRIPTOR * cpu; + bau_desc += UV_ITEMS_PER_DESCRIPTOR * uv_cpu; bau_nodes_clear(&bau_desc->distribution, UV_DISTRIBUTION_SIZE); i = 0; - for_each_cpu(bit, cpumaskp) { + for_each_cpu(bit, flush_mask) { blade = uv_cpu_to_blade_id(bit); BUG_ON(blade > (UV_DISTRIBUTION_SIZE - 1)); if (blade == this_blade) { @@ -330,17 +342,17 @@ int uv_flush_tlb_others(struct cpumask *cpumaskp, struct mm_struct *mm, * no off_node flushing; return status for local node */ if (locals) - return 0; + return flush_mask; else - return 1; + return NULL; } __get_cpu_var(ptcstats).requestor++; __get_cpu_var(ptcstats).ntargeted += i; bau_desc->payload.address = va; - bau_desc->payload.sending_cpu = smp_processor_id(); + bau_desc->payload.sending_cpu = cpu; - return uv_flush_send_and_wait(cpu, this_blade, bau_desc, cpumaskp); + return uv_flush_send_and_wait(uv_cpu, this_blade, bau_desc, flush_mask); } /* -- cgit v1.2.3 From 6dd01bedee6c3191643db303a1dc530bad56ec55 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 21 Jan 2009 17:26:06 +0900 Subject: x86: prepare for tlb merge Impact: clean up, ipi vector number reordering for x86_32 Make the following changes to prepare for tlb merge. * reorder x86_32 ip vectors * adjust tlb_32.c and tlb_64.c such that their logics coincide exactly - on spurious invalidate ipi, tlb_32 acks the irq - tlb_64 now has proper memory barriers around clearing flush_cpumask (no change in generated code) * unexport flush_tlb_page from tlb_32.c, there's no user * use unsigned int for cpu id * drop unnecessary includes from tlb_64.c Signed-off-by: Tejun Heo --- arch/x86/kernel/tlb_32.c | 10 +++++----- arch/x86/kernel/tlb_64.c | 18 +++++++----------- 2 files changed, 12 insertions(+), 16 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index abf0808d6fc..93fcb05c7d4 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -84,13 +84,15 @@ EXPORT_SYMBOL_GPL(leave_mm); * * 1) Flush the tlb entries if the cpu uses the mm that's being flushed. * 2) Leave the mm if we are in the lazy tlb mode. + * + * Interrupts are disabled. */ void smp_invalidate_interrupt(struct pt_regs *regs) { - unsigned long cpu; + unsigned int cpu; - cpu = get_cpu(); + cpu = smp_processor_id(); if (!cpumask_test_cpu(cpu, flush_cpumask)) goto out; @@ -112,12 +114,11 @@ void smp_invalidate_interrupt(struct pt_regs *regs) } else leave_mm(cpu); } +out: ack_APIC_irq(); smp_mb__before_clear_bit(); cpumask_clear_cpu(cpu, flush_cpumask); smp_mb__after_clear_bit(); -out: - put_cpu_no_resched(); inc_irq_stat(irq_tlb_count); } @@ -215,7 +216,6 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) flush_tlb_others(&mm->cpu_vm_mask, mm, va); preempt_enable(); } -EXPORT_SYMBOL(flush_tlb_page); static void do_flush_tlb_all(void *info) { diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index b8bed841ad6..19ac661422f 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -1,20 +1,14 @@ #include #include -#include #include #include -#include -#include #include +#include -#include -#include #include #include -#include -#include -#include +#include #include DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) @@ -121,8 +115,8 @@ EXPORT_SYMBOL_GPL(leave_mm); asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) { - int cpu; - int sender; + unsigned int cpu; + unsigned int sender; union smp_flush_state *f; cpu = smp_processor_id(); @@ -155,14 +149,16 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) } out: ack_APIC_irq(); + smp_mb__before_clear_bit(); cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask)); + smp_mb__after_clear_bit(); inc_irq_stat(irq_tlb_count); } static void flush_tlb_others_ipi(const struct cpumask *cpumask, struct mm_struct *mm, unsigned long va) { - int sender; + unsigned int sender; union smp_flush_state *f; /* Caller has disabled preemption */ -- cgit v1.2.3 From 02cf94c370e0dc9bf408fe45eb86fe9ad58eaf7f Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 21 Jan 2009 17:26:06 +0900 Subject: x86: make x86_32 use tlb_64.c Impact: less contention when issuing invalidate IPI, cleanup Make x86_32 use the same tlb code as 64bit. The 64bit code uses multiple IPI vectors for tlb shootdown to reduce contention. This patch makes x86_32 allocate the same 8 IPIs as x86_64 and share the code paths. Note that the usage of asmlinkage is inconsistent for x86_32 and 64 and calls for further cleanup. This has been noted with a FIXME comment in tlb_64.c. Signed-off-by: Tejun Heo --- arch/x86/kernel/Makefile | 2 +- arch/x86/kernel/entry_32.S | 6 +- arch/x86/kernel/irqinit_32.c | 11 +- arch/x86/kernel/tlb_32.c | 239 ------------------------------------------- arch/x86/kernel/tlb_64.c | 12 ++- 5 files changed, 25 insertions(+), 245 deletions(-) delete mode 100644 arch/x86/kernel/tlb_32.c (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index eb074530c7d..a62a15c2222 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -58,7 +58,7 @@ obj-$(CONFIG_PCI) += early-quirks.o apm-y := apm_32.o obj-$(CONFIG_APM) += apm.o obj-$(CONFIG_X86_SMP) += smp.o -obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o tlb_$(BITS).o +obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o tlb_64.o obj-$(CONFIG_X86_32_SMP) += smpcommon.o obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S index 46469029e9d..a0b91aac72a 100644 --- a/arch/x86/kernel/entry_32.S +++ b/arch/x86/kernel/entry_32.S @@ -672,7 +672,7 @@ common_interrupt: ENDPROC(common_interrupt) CFI_ENDPROC -#define BUILD_INTERRUPT(name, nr) \ +#define BUILD_INTERRUPT3(name, nr, fn) \ ENTRY(name) \ RING0_INT_FRAME; \ pushl $~(nr); \ @@ -680,11 +680,13 @@ ENTRY(name) \ SAVE_ALL; \ TRACE_IRQS_OFF \ movl %esp,%eax; \ - call smp_##name; \ + call fn; \ jmp ret_from_intr; \ CFI_ENDPROC; \ ENDPROC(name) +#define BUILD_INTERRUPT(name, nr) BUILD_INTERRUPT3(name, nr, smp_##name) + /* The include is where all of the SMP etc. interrupts come from */ #include "entry_arch.h" diff --git a/arch/x86/kernel/irqinit_32.c b/arch/x86/kernel/irqinit_32.c index 1507ad4e674..bf629cadec1 100644 --- a/arch/x86/kernel/irqinit_32.c +++ b/arch/x86/kernel/irqinit_32.c @@ -149,8 +149,15 @@ void __init native_init_IRQ(void) */ alloc_intr_gate(RESCHEDULE_VECTOR, reschedule_interrupt); - /* IPI for invalidation */ - alloc_intr_gate(INVALIDATE_TLB_VECTOR, invalidate_interrupt); + /* IPIs for invalidation */ + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+0, invalidate_interrupt0); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+1, invalidate_interrupt1); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+2, invalidate_interrupt2); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+3, invalidate_interrupt3); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+4, invalidate_interrupt4); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+5, invalidate_interrupt5); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+6, invalidate_interrupt6); + alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+7, invalidate_interrupt7); /* IPI for generic function call */ alloc_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt); diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c deleted file mode 100644 index 93fcb05c7d4..00000000000 --- a/arch/x86/kernel/tlb_32.c +++ /dev/null @@ -1,239 +0,0 @@ -#include -#include -#include - -#include - -DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) - = { &init_mm, 0, }; - -/* must come after the send_IPI functions above for inlining */ -#include - -/* - * Smarter SMP flushing macros. - * c/o Linus Torvalds. - * - * These mean you can really definitely utterly forget about - * writing to user space from interrupts. (Its not allowed anyway). - * - * Optimizations Manfred Spraul - */ - -static cpumask_var_t flush_cpumask; -static struct mm_struct *flush_mm; -static unsigned long flush_va; -static DEFINE_SPINLOCK(tlbstate_lock); - -/* - * We cannot call mmdrop() because we are in interrupt context, - * instead update mm->cpu_vm_mask. - * - * We need to reload %cr3 since the page tables may be going - * away from under us.. - */ -void leave_mm(int cpu) -{ - 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); - -/* - * - * The flush IPI assumes that a thread switch happens in this order: - * [cpu0: the cpu that switches] - * 1) switch_mm() either 1a) or 1b) - * 1a) thread switch to a different mm - * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask); - * Stop ipi delivery for the old mm. This is not synchronized with - * the other cpus, but smp_invalidate_interrupt ignore flush ipis - * for the wrong mm, and in the worst case we perform a superfluous - * tlb flush. - * 1a2) set cpu_tlbstate to TLBSTATE_OK - * Now the smp_invalidate_interrupt won't call leave_mm if cpu0 - * was in lazy tlb mode. - * 1a3) update cpu_tlbstate[].active_mm - * Now cpu0 accepts tlb flushes for the new mm. - * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask); - * Now the other cpus will send tlb flush ipis. - * 1a4) change cr3. - * 1b) thread switch without mm change - * cpu_tlbstate[].active_mm is correct, cpu0 already handles - * flush ipis. - * 1b1) set cpu_tlbstate to TLBSTATE_OK - * 1b2) test_and_set the cpu bit in cpu_vm_mask. - * Atomically set the bit [other cpus will start sending flush ipis], - * and test the bit. - * 1b3) if the bit was 0: leave_mm was called, flush the tlb. - * 2) switch %%esp, ie current - * - * The interrupt must handle 2 special cases: - * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm. - * - the cpu performs speculative tlb reads, i.e. even if the cpu only - * runs in kernel space, the cpu could load tlb entries for user space - * pages. - * - * The good news is that cpu_tlbstate is local to each cpu, no - * write/read ordering problems. - */ - -/* - * TLB flush IPI: - * - * 1) Flush the tlb entries if the cpu uses the mm that's being flushed. - * 2) Leave the mm if we are in the lazy tlb mode. - * - * Interrupts are disabled. - */ - -void smp_invalidate_interrupt(struct pt_regs *regs) -{ - unsigned int cpu; - - cpu = smp_processor_id(); - - if (!cpumask_test_cpu(cpu, flush_cpumask)) - goto out; - /* - * This was a BUG() but until someone can quote me the - * line from the intel manual that guarantees an IPI to - * multiple CPUs is retried _only_ on the erroring CPUs - * its staying as a return - * - * BUG(); - */ - - 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 - __flush_tlb_one(flush_va); - } else - leave_mm(cpu); - } -out: - ack_APIC_irq(); - smp_mb__before_clear_bit(); - cpumask_clear_cpu(cpu, flush_cpumask); - smp_mb__after_clear_bit(); - inc_irq_stat(irq_tlb_count); -} - -void native_flush_tlb_others(const struct cpumask *cpumask, - struct mm_struct *mm, unsigned long va) -{ - /* - * - mask must exist :) - */ - BUG_ON(cpumask_empty(cpumask)); - BUG_ON(!mm); - - /* - * i'm not happy about this global shared spinlock in the - * MM hot path, but we'll see how contended it is. - * AK: x86-64 has a faster method that could be ported. - */ - spin_lock(&tlbstate_lock); - - cpumask_andnot(flush_cpumask, cpumask, cpumask_of(smp_processor_id())); -#ifdef CONFIG_HOTPLUG_CPU - /* If a CPU which we ran on has gone down, OK. */ - cpumask_and(flush_cpumask, flush_cpumask, cpu_online_mask); - if (unlikely(cpumask_empty(flush_cpumask))) { - spin_unlock(&tlbstate_lock); - return; - } -#endif - flush_mm = mm; - flush_va = va; - - /* - * Make the above memory operations globally visible before - * sending the IPI. - */ - smp_mb(); - /* - * We have to send the IPI only to - * CPUs affected. - */ - send_IPI_mask(flush_cpumask, INVALIDATE_TLB_VECTOR); - - while (!cpumask_empty(flush_cpumask)) - /* nothing. lockup detection does not belong here */ - cpu_relax(); - - flush_mm = NULL; - flush_va = 0; - spin_unlock(&tlbstate_lock); -} - -void flush_tlb_current_task(void) -{ - struct mm_struct *mm = current->mm; - - preempt_disable(); - - local_flush_tlb(); - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); - preempt_enable(); -} - -void flush_tlb_mm(struct mm_struct *mm) -{ - - preempt_disable(); - - if (current->active_mm == mm) { - if (current->mm) - local_flush_tlb(); - else - leave_mm(smp_processor_id()); - } - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); - - preempt_enable(); -} - -void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) -{ - struct mm_struct *mm = vma->vm_mm; - - preempt_disable(); - - if (current->active_mm == mm) { - if (current->mm) - __flush_tlb_one(va); - else - leave_mm(smp_processor_id()); - } - - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, va); - preempt_enable(); -} - -static void do_flush_tlb_all(void *info) -{ - unsigned long cpu = smp_processor_id(); - - __flush_tlb_all(); - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) - leave_mm(cpu); -} - -void flush_tlb_all(void) -{ - on_each_cpu(do_flush_tlb_all, NULL, 1); -} - -static int init_flush_cpumask(void) -{ - alloc_cpumask_var(&flush_cpumask, GFP_KERNEL); - return 0; -} -early_initcall(init_flush_cpumask); diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index 19ac661422f..b3ca1b94065 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -113,7 +113,17 @@ EXPORT_SYMBOL_GPL(leave_mm); * Interrupts are disabled. */ -asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) +/* + * FIXME: use of asmlinkage is not consistent. On x86_64 it's noop + * but still used for documentation purpose but the usage is slightly + * inconsistent. On x86_32, asmlinkage is regparm(0) but interrupt + * entry calls in with the first parameter in %eax. Maybe define + * intrlinkage? + */ +#ifdef CONFIG_X86_64 +asmlinkage +#endif +void smp_invalidate_interrupt(struct pt_regs *regs) { unsigned int cpu; unsigned int sender; -- cgit v1.2.3 From 16c2d3f895a3bc8d8e4c76c2646a6b750c181299 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 21 Jan 2009 17:26:06 +0900 Subject: x86: rename tlb_64.c to tlb.c Impact: file rename tlb_64.c is now the tlb code for both 32 and 64. Rename it to tlb.c. Signed-off-by: Tejun Heo --- arch/x86/kernel/Makefile | 2 +- arch/x86/kernel/tlb.c | 296 +++++++++++++++++++++++++++++++++++++++++++++++ arch/x86/kernel/tlb_64.c | 296 ----------------------------------------------- 3 files changed, 297 insertions(+), 297 deletions(-) create mode 100644 arch/x86/kernel/tlb.c delete mode 100644 arch/x86/kernel/tlb_64.c (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index a62a15c2222..0626a88fbb4 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -58,7 +58,7 @@ obj-$(CONFIG_PCI) += early-quirks.o apm-y := apm_32.o obj-$(CONFIG_APM) += apm.o obj-$(CONFIG_X86_SMP) += smp.o -obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o tlb_64.o +obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o tlb.o obj-$(CONFIG_X86_32_SMP) += smpcommon.o obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o diff --git a/arch/x86/kernel/tlb.c b/arch/x86/kernel/tlb.c new file mode 100644 index 00000000000..b3ca1b94065 --- /dev/null +++ b/arch/x86/kernel/tlb.c @@ -0,0 +1,296 @@ +#include + +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) + = { &init_mm, 0, }; + +#include +/* + * Smarter SMP flushing macros. + * c/o Linus Torvalds. + * + * These mean you can really definitely utterly forget about + * writing to user space from interrupts. (Its not allowed anyway). + * + * Optimizations Manfred Spraul + * + * More scalable flush, from Andi Kleen + * + * To avoid global state use 8 different call vectors. + * Each CPU uses a specific vector to trigger flushes on other + * CPUs. Depending on the received vector the target CPUs look into + * the right per cpu variable for the flush data. + * + * With more than 8 CPUs they are hashed to the 8 available + * vectors. The limited global vector space forces us to this right now. + * In future when interrupts are split into per CPU domains this could be + * fixed, at the cost of triggering multiple IPIs in some cases. + */ + +union smp_flush_state { + struct { + struct mm_struct *flush_mm; + unsigned long flush_va; + spinlock_t tlbstate_lock; + DECLARE_BITMAP(flush_cpumask, NR_CPUS); + }; + char pad[SMP_CACHE_BYTES]; +} ____cacheline_aligned; + +/* State is put into the per CPU data section, but padded + to a full cache line because other CPUs can access it and we don't + want false sharing in the per cpu data segment. */ +static DEFINE_PER_CPU(union smp_flush_state, flush_state); + +/* + * We cannot call mmdrop() because we are in interrupt context, + * instead update mm->cpu_vm_mask. + */ +void leave_mm(int cpu) +{ + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) + BUG(); + cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); + load_cr3(swapper_pg_dir); +} +EXPORT_SYMBOL_GPL(leave_mm); + +/* + * + * The flush IPI assumes that a thread switch happens in this order: + * [cpu0: the cpu that switches] + * 1) switch_mm() either 1a) or 1b) + * 1a) thread switch to a different mm + * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask); + * Stop ipi delivery for the old mm. This is not synchronized with + * the other cpus, but smp_invalidate_interrupt ignore flush ipis + * for the wrong mm, and in the worst case we perform a superfluous + * tlb flush. + * 1a2) set cpu mmu_state to TLBSTATE_OK + * Now the smp_invalidate_interrupt won't call leave_mm if cpu0 + * was in lazy tlb mode. + * 1a3) update cpu active_mm + * Now cpu0 accepts tlb flushes for the new mm. + * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask); + * Now the other cpus will send tlb flush ipis. + * 1a4) change cr3. + * 1b) thread switch without mm change + * cpu active_mm is correct, cpu0 already handles + * flush ipis. + * 1b1) set cpu mmu_state to TLBSTATE_OK + * 1b2) test_and_set the cpu bit in cpu_vm_mask. + * Atomically set the bit [other cpus will start sending flush ipis], + * and test the bit. + * 1b3) if the bit was 0: leave_mm was called, flush the tlb. + * 2) switch %%esp, ie current + * + * The interrupt must handle 2 special cases: + * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm. + * - the cpu performs speculative tlb reads, i.e. even if the cpu only + * runs in kernel space, the cpu could load tlb entries for user space + * pages. + * + * The good news is that cpu mmu_state is local to each cpu, no + * write/read ordering problems. + */ + +/* + * TLB flush IPI: + * + * 1) Flush the tlb entries if the cpu uses the mm that's being flushed. + * 2) Leave the mm if we are in the lazy tlb mode. + * + * Interrupts are disabled. + */ + +/* + * FIXME: use of asmlinkage is not consistent. On x86_64 it's noop + * but still used for documentation purpose but the usage is slightly + * inconsistent. On x86_32, asmlinkage is regparm(0) but interrupt + * entry calls in with the first parameter in %eax. Maybe define + * intrlinkage? + */ +#ifdef CONFIG_X86_64 +asmlinkage +#endif +void smp_invalidate_interrupt(struct pt_regs *regs) +{ + unsigned int cpu; + unsigned int sender; + union smp_flush_state *f; + + cpu = smp_processor_id(); + /* + * orig_rax contains the negated interrupt vector. + * Use that to determine where the sender put the data. + */ + sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START; + f = &per_cpu(flush_state, sender); + + if (!cpumask_test_cpu(cpu, to_cpumask(f->flush_cpumask))) + goto out; + /* + * This was a BUG() but until someone can quote me the + * line from the intel manual that guarantees an IPI to + * multiple CPUs is retried _only_ on the erroring CPUs + * its staying as a return + * + * BUG(); + */ + + if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) { + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { + if (f->flush_va == TLB_FLUSH_ALL) + local_flush_tlb(); + else + __flush_tlb_one(f->flush_va); + } else + leave_mm(cpu); + } +out: + ack_APIC_irq(); + smp_mb__before_clear_bit(); + cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask)); + smp_mb__after_clear_bit(); + inc_irq_stat(irq_tlb_count); +} + +static void flush_tlb_others_ipi(const struct cpumask *cpumask, + struct mm_struct *mm, unsigned long va) +{ + unsigned int sender; + union smp_flush_state *f; + + /* Caller has disabled preemption */ + sender = smp_processor_id() % NUM_INVALIDATE_TLB_VECTORS; + f = &per_cpu(flush_state, sender); + + /* + * Could avoid this lock when + * num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is + * probably not worth checking this for a cache-hot lock. + */ + spin_lock(&f->tlbstate_lock); + + f->flush_mm = mm; + f->flush_va = va; + cpumask_andnot(to_cpumask(f->flush_cpumask), + cpumask, cpumask_of(smp_processor_id())); + + /* + * Make the above memory operations globally visible before + * sending the IPI. + */ + smp_mb(); + /* + * We have to send the IPI only to + * CPUs affected. + */ + send_IPI_mask(to_cpumask(f->flush_cpumask), + INVALIDATE_TLB_VECTOR_START + sender); + + while (!cpumask_empty(to_cpumask(f->flush_cpumask))) + cpu_relax(); + + f->flush_mm = NULL; + f->flush_va = 0; + spin_unlock(&f->tlbstate_lock); +} + +void native_flush_tlb_others(const struct cpumask *cpumask, + struct mm_struct *mm, unsigned long va) +{ + if (is_uv_system()) { + unsigned int cpu; + + cpu = get_cpu(); + cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu); + if (cpumask) + flush_tlb_others_ipi(cpumask, mm, va); + put_cpu(); + return; + } + flush_tlb_others_ipi(cpumask, mm, va); +} + +static int __cpuinit init_smp_flush(void) +{ + int i; + + for_each_possible_cpu(i) + spin_lock_init(&per_cpu(flush_state, i).tlbstate_lock); + + return 0; +} +core_initcall(init_smp_flush); + +void flush_tlb_current_task(void) +{ + struct mm_struct *mm = current->mm; + + preempt_disable(); + + local_flush_tlb(); + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); + preempt_enable(); +} + +void flush_tlb_mm(struct mm_struct *mm) +{ + preempt_disable(); + + if (current->active_mm == mm) { + if (current->mm) + local_flush_tlb(); + else + leave_mm(smp_processor_id()); + } + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); + + preempt_enable(); +} + +void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) +{ + struct mm_struct *mm = vma->vm_mm; + + preempt_disable(); + + if (current->active_mm == mm) { + if (current->mm) + __flush_tlb_one(va); + else + leave_mm(smp_processor_id()); + } + + if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) + flush_tlb_others(&mm->cpu_vm_mask, mm, va); + + preempt_enable(); +} + +static void do_flush_tlb_all(void *info) +{ + unsigned long cpu = smp_processor_id(); + + __flush_tlb_all(); + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) + leave_mm(cpu); +} + +void flush_tlb_all(void) +{ + on_each_cpu(do_flush_tlb_all, NULL, 1); +} diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c deleted file mode 100644 index b3ca1b94065..00000000000 --- a/arch/x86/kernel/tlb_64.c +++ /dev/null @@ -1,296 +0,0 @@ -#include - -#include -#include -#include -#include -#include - -#include -#include -#include -#include - -DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) - = { &init_mm, 0, }; - -#include -/* - * Smarter SMP flushing macros. - * c/o Linus Torvalds. - * - * These mean you can really definitely utterly forget about - * writing to user space from interrupts. (Its not allowed anyway). - * - * Optimizations Manfred Spraul - * - * More scalable flush, from Andi Kleen - * - * To avoid global state use 8 different call vectors. - * Each CPU uses a specific vector to trigger flushes on other - * CPUs. Depending on the received vector the target CPUs look into - * the right per cpu variable for the flush data. - * - * With more than 8 CPUs they are hashed to the 8 available - * vectors. The limited global vector space forces us to this right now. - * In future when interrupts are split into per CPU domains this could be - * fixed, at the cost of triggering multiple IPIs in some cases. - */ - -union smp_flush_state { - struct { - struct mm_struct *flush_mm; - unsigned long flush_va; - spinlock_t tlbstate_lock; - DECLARE_BITMAP(flush_cpumask, NR_CPUS); - }; - char pad[SMP_CACHE_BYTES]; -} ____cacheline_aligned; - -/* State is put into the per CPU data section, but padded - to a full cache line because other CPUs can access it and we don't - want false sharing in the per cpu data segment. */ -static DEFINE_PER_CPU(union smp_flush_state, flush_state); - -/* - * We cannot call mmdrop() because we are in interrupt context, - * instead update mm->cpu_vm_mask. - */ -void leave_mm(int cpu) -{ - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) - BUG(); - cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); - load_cr3(swapper_pg_dir); -} -EXPORT_SYMBOL_GPL(leave_mm); - -/* - * - * The flush IPI assumes that a thread switch happens in this order: - * [cpu0: the cpu that switches] - * 1) switch_mm() either 1a) or 1b) - * 1a) thread switch to a different mm - * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask); - * Stop ipi delivery for the old mm. This is not synchronized with - * the other cpus, but smp_invalidate_interrupt ignore flush ipis - * for the wrong mm, and in the worst case we perform a superfluous - * tlb flush. - * 1a2) set cpu mmu_state to TLBSTATE_OK - * Now the smp_invalidate_interrupt won't call leave_mm if cpu0 - * was in lazy tlb mode. - * 1a3) update cpu active_mm - * Now cpu0 accepts tlb flushes for the new mm. - * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask); - * Now the other cpus will send tlb flush ipis. - * 1a4) change cr3. - * 1b) thread switch without mm change - * cpu active_mm is correct, cpu0 already handles - * flush ipis. - * 1b1) set cpu mmu_state to TLBSTATE_OK - * 1b2) test_and_set the cpu bit in cpu_vm_mask. - * Atomically set the bit [other cpus will start sending flush ipis], - * and test the bit. - * 1b3) if the bit was 0: leave_mm was called, flush the tlb. - * 2) switch %%esp, ie current - * - * The interrupt must handle 2 special cases: - * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm. - * - the cpu performs speculative tlb reads, i.e. even if the cpu only - * runs in kernel space, the cpu could load tlb entries for user space - * pages. - * - * The good news is that cpu mmu_state is local to each cpu, no - * write/read ordering problems. - */ - -/* - * TLB flush IPI: - * - * 1) Flush the tlb entries if the cpu uses the mm that's being flushed. - * 2) Leave the mm if we are in the lazy tlb mode. - * - * Interrupts are disabled. - */ - -/* - * FIXME: use of asmlinkage is not consistent. On x86_64 it's noop - * but still used for documentation purpose but the usage is slightly - * inconsistent. On x86_32, asmlinkage is regparm(0) but interrupt - * entry calls in with the first parameter in %eax. Maybe define - * intrlinkage? - */ -#ifdef CONFIG_X86_64 -asmlinkage -#endif -void smp_invalidate_interrupt(struct pt_regs *regs) -{ - unsigned int cpu; - unsigned int sender; - union smp_flush_state *f; - - cpu = smp_processor_id(); - /* - * orig_rax contains the negated interrupt vector. - * Use that to determine where the sender put the data. - */ - sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START; - f = &per_cpu(flush_state, sender); - - if (!cpumask_test_cpu(cpu, to_cpumask(f->flush_cpumask))) - goto out; - /* - * This was a BUG() but until someone can quote me the - * line from the intel manual that guarantees an IPI to - * multiple CPUs is retried _only_ on the erroring CPUs - * its staying as a return - * - * BUG(); - */ - - if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) { - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { - if (f->flush_va == TLB_FLUSH_ALL) - local_flush_tlb(); - else - __flush_tlb_one(f->flush_va); - } else - leave_mm(cpu); - } -out: - ack_APIC_irq(); - smp_mb__before_clear_bit(); - cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask)); - smp_mb__after_clear_bit(); - inc_irq_stat(irq_tlb_count); -} - -static void flush_tlb_others_ipi(const struct cpumask *cpumask, - struct mm_struct *mm, unsigned long va) -{ - unsigned int sender; - union smp_flush_state *f; - - /* Caller has disabled preemption */ - sender = smp_processor_id() % NUM_INVALIDATE_TLB_VECTORS; - f = &per_cpu(flush_state, sender); - - /* - * Could avoid this lock when - * num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is - * probably not worth checking this for a cache-hot lock. - */ - spin_lock(&f->tlbstate_lock); - - f->flush_mm = mm; - f->flush_va = va; - cpumask_andnot(to_cpumask(f->flush_cpumask), - cpumask, cpumask_of(smp_processor_id())); - - /* - * Make the above memory operations globally visible before - * sending the IPI. - */ - smp_mb(); - /* - * We have to send the IPI only to - * CPUs affected. - */ - send_IPI_mask(to_cpumask(f->flush_cpumask), - INVALIDATE_TLB_VECTOR_START + sender); - - while (!cpumask_empty(to_cpumask(f->flush_cpumask))) - cpu_relax(); - - f->flush_mm = NULL; - f->flush_va = 0; - spin_unlock(&f->tlbstate_lock); -} - -void native_flush_tlb_others(const struct cpumask *cpumask, - struct mm_struct *mm, unsigned long va) -{ - if (is_uv_system()) { - unsigned int cpu; - - cpu = get_cpu(); - cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu); - if (cpumask) - flush_tlb_others_ipi(cpumask, mm, va); - put_cpu(); - return; - } - flush_tlb_others_ipi(cpumask, mm, va); -} - -static int __cpuinit init_smp_flush(void) -{ - int i; - - for_each_possible_cpu(i) - spin_lock_init(&per_cpu(flush_state, i).tlbstate_lock); - - return 0; -} -core_initcall(init_smp_flush); - -void flush_tlb_current_task(void) -{ - struct mm_struct *mm = current->mm; - - preempt_disable(); - - local_flush_tlb(); - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); - preempt_enable(); -} - -void flush_tlb_mm(struct mm_struct *mm) -{ - preempt_disable(); - - if (current->active_mm == mm) { - if (current->mm) - local_flush_tlb(); - else - leave_mm(smp_processor_id()); - } - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); - - preempt_enable(); -} - -void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) -{ - struct mm_struct *mm = vma->vm_mm; - - preempt_disable(); - - if (current->active_mm == mm) { - if (current->mm) - __flush_tlb_one(va); - else - leave_mm(smp_processor_id()); - } - - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, va); - - preempt_enable(); -} - -static void do_flush_tlb_all(void *info) -{ - unsigned long cpu = smp_processor_id(); - - __flush_tlb_all(); - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) - leave_mm(cpu); -} - -void flush_tlb_all(void) -{ - on_each_cpu(do_flush_tlb_all, NULL, 1); -} -- cgit v1.2.3 From 55f4949f5765e7a29863b6d17a774601810732f5 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 21 Jan 2009 10:08:53 +0100 Subject: x86, mm: move tlb.c to arch/x86/mm/ Impact: cleanup Now that it's unified, move the (SMP) TLB flushing code from arch/x86/kernel/ to arch/x86/mm/, where it belongs logically. Signed-off-by: Ingo Molnar --- arch/x86/kernel/Makefile | 2 +- arch/x86/kernel/tlb.c | 296 ----------------------------------------------- 2 files changed, 1 insertion(+), 297 deletions(-) delete mode 100644 arch/x86/kernel/tlb.c (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 0626a88fbb4..0b3272f58bd 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -58,7 +58,7 @@ obj-$(CONFIG_PCI) += early-quirks.o apm-y := apm_32.o obj-$(CONFIG_APM) += apm.o obj-$(CONFIG_X86_SMP) += smp.o -obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o tlb.o +obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o obj-$(CONFIG_X86_32_SMP) += smpcommon.o obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o diff --git a/arch/x86/kernel/tlb.c b/arch/x86/kernel/tlb.c deleted file mode 100644 index b3ca1b94065..00000000000 --- a/arch/x86/kernel/tlb.c +++ /dev/null @@ -1,296 +0,0 @@ -#include - -#include -#include -#include -#include -#include - -#include -#include -#include -#include - -DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) - = { &init_mm, 0, }; - -#include -/* - * Smarter SMP flushing macros. - * c/o Linus Torvalds. - * - * These mean you can really definitely utterly forget about - * writing to user space from interrupts. (Its not allowed anyway). - * - * Optimizations Manfred Spraul - * - * More scalable flush, from Andi Kleen - * - * To avoid global state use 8 different call vectors. - * Each CPU uses a specific vector to trigger flushes on other - * CPUs. Depending on the received vector the target CPUs look into - * the right per cpu variable for the flush data. - * - * With more than 8 CPUs they are hashed to the 8 available - * vectors. The limited global vector space forces us to this right now. - * In future when interrupts are split into per CPU domains this could be - * fixed, at the cost of triggering multiple IPIs in some cases. - */ - -union smp_flush_state { - struct { - struct mm_struct *flush_mm; - unsigned long flush_va; - spinlock_t tlbstate_lock; - DECLARE_BITMAP(flush_cpumask, NR_CPUS); - }; - char pad[SMP_CACHE_BYTES]; -} ____cacheline_aligned; - -/* State is put into the per CPU data section, but padded - to a full cache line because other CPUs can access it and we don't - want false sharing in the per cpu data segment. */ -static DEFINE_PER_CPU(union smp_flush_state, flush_state); - -/* - * We cannot call mmdrop() because we are in interrupt context, - * instead update mm->cpu_vm_mask. - */ -void leave_mm(int cpu) -{ - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) - BUG(); - cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); - load_cr3(swapper_pg_dir); -} -EXPORT_SYMBOL_GPL(leave_mm); - -/* - * - * The flush IPI assumes that a thread switch happens in this order: - * [cpu0: the cpu that switches] - * 1) switch_mm() either 1a) or 1b) - * 1a) thread switch to a different mm - * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask); - * Stop ipi delivery for the old mm. This is not synchronized with - * the other cpus, but smp_invalidate_interrupt ignore flush ipis - * for the wrong mm, and in the worst case we perform a superfluous - * tlb flush. - * 1a2) set cpu mmu_state to TLBSTATE_OK - * Now the smp_invalidate_interrupt won't call leave_mm if cpu0 - * was in lazy tlb mode. - * 1a3) update cpu active_mm - * Now cpu0 accepts tlb flushes for the new mm. - * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask); - * Now the other cpus will send tlb flush ipis. - * 1a4) change cr3. - * 1b) thread switch without mm change - * cpu active_mm is correct, cpu0 already handles - * flush ipis. - * 1b1) set cpu mmu_state to TLBSTATE_OK - * 1b2) test_and_set the cpu bit in cpu_vm_mask. - * Atomically set the bit [other cpus will start sending flush ipis], - * and test the bit. - * 1b3) if the bit was 0: leave_mm was called, flush the tlb. - * 2) switch %%esp, ie current - * - * The interrupt must handle 2 special cases: - * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm. - * - the cpu performs speculative tlb reads, i.e. even if the cpu only - * runs in kernel space, the cpu could load tlb entries for user space - * pages. - * - * The good news is that cpu mmu_state is local to each cpu, no - * write/read ordering problems. - */ - -/* - * TLB flush IPI: - * - * 1) Flush the tlb entries if the cpu uses the mm that's being flushed. - * 2) Leave the mm if we are in the lazy tlb mode. - * - * Interrupts are disabled. - */ - -/* - * FIXME: use of asmlinkage is not consistent. On x86_64 it's noop - * but still used for documentation purpose but the usage is slightly - * inconsistent. On x86_32, asmlinkage is regparm(0) but interrupt - * entry calls in with the first parameter in %eax. Maybe define - * intrlinkage? - */ -#ifdef CONFIG_X86_64 -asmlinkage -#endif -void smp_invalidate_interrupt(struct pt_regs *regs) -{ - unsigned int cpu; - unsigned int sender; - union smp_flush_state *f; - - cpu = smp_processor_id(); - /* - * orig_rax contains the negated interrupt vector. - * Use that to determine where the sender put the data. - */ - sender = ~regs->orig_ax - INVALIDATE_TLB_VECTOR_START; - f = &per_cpu(flush_state, sender); - - if (!cpumask_test_cpu(cpu, to_cpumask(f->flush_cpumask))) - goto out; - /* - * This was a BUG() but until someone can quote me the - * line from the intel manual that guarantees an IPI to - * multiple CPUs is retried _only_ on the erroring CPUs - * its staying as a return - * - * BUG(); - */ - - if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) { - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { - if (f->flush_va == TLB_FLUSH_ALL) - local_flush_tlb(); - else - __flush_tlb_one(f->flush_va); - } else - leave_mm(cpu); - } -out: - ack_APIC_irq(); - smp_mb__before_clear_bit(); - cpumask_clear_cpu(cpu, to_cpumask(f->flush_cpumask)); - smp_mb__after_clear_bit(); - inc_irq_stat(irq_tlb_count); -} - -static void flush_tlb_others_ipi(const struct cpumask *cpumask, - struct mm_struct *mm, unsigned long va) -{ - unsigned int sender; - union smp_flush_state *f; - - /* Caller has disabled preemption */ - sender = smp_processor_id() % NUM_INVALIDATE_TLB_VECTORS; - f = &per_cpu(flush_state, sender); - - /* - * Could avoid this lock when - * num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is - * probably not worth checking this for a cache-hot lock. - */ - spin_lock(&f->tlbstate_lock); - - f->flush_mm = mm; - f->flush_va = va; - cpumask_andnot(to_cpumask(f->flush_cpumask), - cpumask, cpumask_of(smp_processor_id())); - - /* - * Make the above memory operations globally visible before - * sending the IPI. - */ - smp_mb(); - /* - * We have to send the IPI only to - * CPUs affected. - */ - send_IPI_mask(to_cpumask(f->flush_cpumask), - INVALIDATE_TLB_VECTOR_START + sender); - - while (!cpumask_empty(to_cpumask(f->flush_cpumask))) - cpu_relax(); - - f->flush_mm = NULL; - f->flush_va = 0; - spin_unlock(&f->tlbstate_lock); -} - -void native_flush_tlb_others(const struct cpumask *cpumask, - struct mm_struct *mm, unsigned long va) -{ - if (is_uv_system()) { - unsigned int cpu; - - cpu = get_cpu(); - cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu); - if (cpumask) - flush_tlb_others_ipi(cpumask, mm, va); - put_cpu(); - return; - } - flush_tlb_others_ipi(cpumask, mm, va); -} - -static int __cpuinit init_smp_flush(void) -{ - int i; - - for_each_possible_cpu(i) - spin_lock_init(&per_cpu(flush_state, i).tlbstate_lock); - - return 0; -} -core_initcall(init_smp_flush); - -void flush_tlb_current_task(void) -{ - struct mm_struct *mm = current->mm; - - preempt_disable(); - - local_flush_tlb(); - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); - preempt_enable(); -} - -void flush_tlb_mm(struct mm_struct *mm) -{ - preempt_disable(); - - if (current->active_mm == mm) { - if (current->mm) - local_flush_tlb(); - else - leave_mm(smp_processor_id()); - } - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, TLB_FLUSH_ALL); - - preempt_enable(); -} - -void flush_tlb_page(struct vm_area_struct *vma, unsigned long va) -{ - struct mm_struct *mm = vma->vm_mm; - - preempt_disable(); - - if (current->active_mm == mm) { - if (current->mm) - __flush_tlb_one(va); - else - leave_mm(smp_processor_id()); - } - - if (cpumask_any_but(&mm->cpu_vm_mask, smp_processor_id()) < nr_cpu_ids) - flush_tlb_others(&mm->cpu_vm_mask, mm, va); - - preempt_enable(); -} - -static void do_flush_tlb_all(void *info) -{ - unsigned long cpu = smp_processor_id(); - - __flush_tlb_all(); - if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) - leave_mm(cpu); -} - -void flush_tlb_all(void) -{ - on_each_cpu(do_flush_tlb_all, NULL, 1); -} -- cgit v1.2.3 From 03b486322e994dde49e67aedb391867b7cf28822 Mon Sep 17 00:00:00 2001 From: Nick Piggin Date: Tue, 20 Jan 2009 04:36:04 +0100 Subject: x86: make UV support configurable Make X86 SGI Ultraviolet support configurable. Saves about 13K of text size on my modest config. text data bss dec hex filename 6770537 1158680 694356 8623573 8395d5 vmlinux 6757492 1157664 694228 8609384 835e68 vmlinux.nouv Signed-off-by: Nick Piggin Signed-off-by: Ingo Molnar --- arch/x86/kernel/Makefile | 5 +++-- arch/x86/kernel/efi.c | 2 ++ arch/x86/kernel/entry_64.S | 2 ++ arch/x86/kernel/genapic_64.c | 2 ++ arch/x86/kernel/io_apic.c | 2 +- 5 files changed, 10 insertions(+), 3 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 0b3272f58bd..a99437c965c 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -115,10 +115,11 @@ obj-$(CONFIG_SWIOTLB) += pci-swiotlb_64.o # NB rename without _64 ### # 64 bit specific files ifeq ($(CONFIG_X86_64),y) - obj-y += genapic_64.o genapic_flat_64.o genx2apic_uv_x.o tlb_uv.o - obj-y += bios_uv.o uv_irq.o uv_sysfs.o + obj-y += genapic_64.o genapic_flat_64.o obj-y += genx2apic_cluster.o obj-y += genx2apic_phys.o + obj-$(CONFIG_X86_UV) += genx2apic_uv_x.o tlb_uv.o + obj-$(CONFIG_X86_UV) += bios_uv.o uv_irq.o uv_sysfs.o obj-$(CONFIG_X86_PM_TIMER) += pmtimer_64.o obj-$(CONFIG_AUDIT) += audit_64.o diff --git a/arch/x86/kernel/efi.c b/arch/x86/kernel/efi.c index 1119d247fe1..b205272ad39 100644 --- a/arch/x86/kernel/efi.c +++ b/arch/x86/kernel/efi.c @@ -366,10 +366,12 @@ void __init efi_init(void) SMBIOS_TABLE_GUID)) { efi.smbios = config_tables[i].table; printk(" SMBIOS=0x%lx ", config_tables[i].table); +#ifdef CONFIG_X86_UV } else if (!efi_guidcmp(config_tables[i].guid, UV_SYSTEM_TABLE_GUID)) { efi.uv_systab = config_tables[i].table; printk(" UVsystab=0x%lx ", config_tables[i].table); +#endif } else if (!efi_guidcmp(config_tables[i].guid, HCDP_TABLE_GUID)) { efi.hcdp = config_tables[i].table; diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index c52b6091916..a52703864a1 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -982,8 +982,10 @@ apicinterrupt IRQ_MOVE_CLEANUP_VECTOR \ irq_move_cleanup_interrupt smp_irq_move_cleanup_interrupt #endif +#ifdef CONFIG_X86_UV apicinterrupt UV_BAU_MESSAGE \ uv_bau_message_intr1 uv_bau_message_interrupt +#endif apicinterrupt LOCAL_TIMER_VECTOR \ apic_timer_interrupt smp_apic_timer_interrupt diff --git a/arch/x86/kernel/genapic_64.c b/arch/x86/kernel/genapic_64.c index 2bced78b0b8..e656c272115 100644 --- a/arch/x86/kernel/genapic_64.c +++ b/arch/x86/kernel/genapic_64.c @@ -32,7 +32,9 @@ extern struct genapic apic_x2apic_cluster; struct genapic __read_mostly *genapic = &apic_flat; static struct genapic *apic_probe[] __initdata = { +#ifdef CONFIG_X86_UV &apic_x2apic_uv_x, +#endif &apic_x2apic_phys, &apic_x2apic_cluster, &apic_physflat, diff --git a/arch/x86/kernel/io_apic.c b/arch/x86/kernel/io_apic.c index f7966039072..e4d36bd56b6 100644 --- a/arch/x86/kernel/io_apic.c +++ b/arch/x86/kernel/io_apic.c @@ -3765,7 +3765,7 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev) } #endif /* CONFIG_HT_IRQ */ -#ifdef CONFIG_X86_64 +#ifdef CONFIG_X86_UV /* * Re-target the irq to the specified CPU and enable the specified MMR located * on the specified blade to allow the sending of MSIs to the specified CPU. -- cgit v1.2.3 From ab897d2013128f470240a541b31cf5e636984e71 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Thu, 22 Jan 2009 14:24:16 -0800 Subject: x86/pvops: remove pte_flags pvop pte_flags() was introduced as a new pvop in order to extract just the flags portion of a pte, which is a potentially cheaper operation than extracting the page number as well. It turns out this operation is not needed, because simply using a mask to extract the flags from a pte is sufficient for all current users. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/kernel/paravirt.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index e4c8fb60887..202514be592 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -435,7 +435,6 @@ struct pv_mmu_ops pv_mmu_ops = { #endif /* PAGETABLE_LEVELS >= 3 */ .pte_val = native_pte_val, - .pte_flags = native_pte_flags, .pgd_val = native_pgd_val, .make_pte = native_make_pte, -- cgit v1.2.3 From 03d2989df9c1c7df5b33c7a87e0790465461836a Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Fri, 23 Jan 2009 11:03:28 +0900 Subject: x86: remove idle_timestamp from 32bit irq_cpustat_t Impact: bogus irq_cpustat field removed idle_timestamp is left over from the removed irqbalance code. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/process_32.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index 2c00a57ccb9..1a1ae8edc40 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c @@ -108,7 +108,6 @@ void cpu_idle(void) play_dead(); local_irq_disable(); - __get_cpu_var(irq_stat).idle_timestamp = jiffies; /* Don't trace irqs off for idle */ stop_critical_timings(); pm_idle(); -- cgit v1.2.3 From 3819cd489ec5d18a4cbd2f05acdc516473caa105 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Fri, 23 Jan 2009 11:03:29 +0900 Subject: x86: remove include of apic.h from hardirq_64.h Impact: cleanup APIC definitions aren't needed here. Remove the include and fix up the fallout. tj: added include to mce_intel_64.c. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/mcheck/mce_intel_64.c | 1 + arch/x86/kernel/efi_64.c | 1 + arch/x86/kernel/irq_64.c | 1 + 3 files changed, 3 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c index 4b48f251fd3..5e8c79e748a 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c +++ b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c @@ -7,6 +7,7 @@ #include #include #include +#include #include #include #include diff --git a/arch/x86/kernel/efi_64.c b/arch/x86/kernel/efi_64.c index 652c5287215..a4ee29127fd 100644 --- a/arch/x86/kernel/efi_64.c +++ b/arch/x86/kernel/efi_64.c @@ -36,6 +36,7 @@ #include #include #include +#include static pgd_t save_pgd __initdata; static unsigned long efi_flags __initdata; diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c index 0b254de8408..018963aa6ee 100644 --- a/arch/x86/kernel/irq_64.c +++ b/arch/x86/kernel/irq_64.c @@ -18,6 +18,7 @@ #include #include #include +#include DEFINE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat); EXPORT_PER_CPU_SYMBOL(irq_stat); -- cgit v1.2.3 From 0d77e7f04d5da160307f4f5c030a171e004f602b Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:47 +0900 Subject: x86: merge setup_per_cpu_maps() into setup_per_cpu_areas() Impact: minor optimization Eliminates the need for two loops over possible cpus. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 48 +++++++++++++++++------------------------- 1 file changed, 19 insertions(+), 29 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 90b8e154bb5..d0b1476490a 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -97,33 +97,6 @@ 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 - * per cpu data areas. These arrays then become expendable and the - * *_early_ptr's are zeroed indicating that the static arrays are gone. - */ -static void __init setup_per_cpu_maps(void) -{ - int cpu; - - for_each_possible_cpu(cpu) { - per_cpu(x86_cpu_to_apicid, cpu) = - early_per_cpu_map(x86_cpu_to_apicid, cpu); - per_cpu(x86_bios_cpu_apicid, cpu) = - early_per_cpu_map(x86_bios_cpu_apicid, cpu); -#ifdef X86_64_NUMA - per_cpu(x86_cpu_to_node_map, cpu) = - early_per_cpu_map(x86_cpu_to_node_map, cpu); -#endif - } - - /* indicate the early static arrays will soon be gone */ - early_per_cpu_ptr(x86_cpu_to_apicid) = NULL; - early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL; -#ifdef X86_64_NUMA - early_per_cpu_ptr(x86_cpu_to_node_map) = NULL; -#endif -} #ifdef CONFIG_X86_64 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { @@ -181,6 +154,19 @@ void __init setup_per_cpu_areas(void) per_cpu_offset(cpu) = ptr - __per_cpu_start; per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); per_cpu(cpu_number, cpu) = cpu; + /* + * Copy data used in early init routines from the initial arrays to the + * per cpu data areas. These arrays then become expendable and the + * *_early_ptr's are zeroed indicating that the static arrays are gone. + */ + per_cpu(x86_cpu_to_apicid, cpu) = + early_per_cpu_map(x86_cpu_to_apicid, cpu); + per_cpu(x86_bios_cpu_apicid, cpu) = + early_per_cpu_map(x86_bios_cpu_apicid, cpu); +#ifdef X86_64_NUMA + per_cpu(x86_cpu_to_node_map, cpu) = + early_per_cpu_map(x86_cpu_to_node_map, cpu); +#endif #ifdef CONFIG_X86_64 per_cpu(irq_stack_ptr, cpu) = per_cpu(irq_stack_union.irq_stack, cpu) + IRQ_STACK_SIZE - 64; @@ -195,8 +181,12 @@ void __init setup_per_cpu_areas(void) DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } - /* Setup percpu data maps */ - setup_per_cpu_maps(); + /* indicate the early static arrays will soon be gone */ + early_per_cpu_ptr(x86_cpu_to_apicid) = NULL; + early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL; +#ifdef X86_64_NUMA + early_per_cpu_ptr(x86_cpu_to_node_map) = NULL; +#endif /* Setup node to cpumask map */ setup_node_to_cpumask_map(); -- cgit v1.2.3 From 6470aff619fbb9dff8dfe8afa5033084cd55ca20 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:47 +0900 Subject: x86: move 64-bit NUMA code Impact: Code movement, no functional change. Move the 64-bit NUMA code from setup_percpu.c to numa_64.c Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 237 +---------------------------------------- 1 file changed, 5 insertions(+), 232 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index d0b1476490a..cb6d622520b 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -51,32 +51,6 @@ DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID); EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid); EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); -#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) -#define X86_64_NUMA 1 /* (used later) */ -DEFINE_PER_CPU(int, node_number) = 0; -EXPORT_PER_CPU_SYMBOL(node_number); - -/* - * Map cpu index to node index - */ -DEFINE_EARLY_PER_CPU(int, x86_cpu_to_node_map, NUMA_NO_NODE); -EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map); - -/* - * Which logical CPUs are on which nodes - */ -cpumask_t *node_to_cpumask_map; -EXPORT_SYMBOL(node_to_cpumask_map); - -/* - * Setup node_to_cpumask_map - */ -static void __init setup_node_to_cpumask_map(void); - -#else -static inline void setup_node_to_cpumask_map(void) { } -#endif - #ifdef CONFIG_X86_64 /* correctly size the local cpu masks */ @@ -163,13 +137,13 @@ void __init setup_per_cpu_areas(void) early_per_cpu_map(x86_cpu_to_apicid, cpu); per_cpu(x86_bios_cpu_apicid, cpu) = early_per_cpu_map(x86_bios_cpu_apicid, cpu); -#ifdef X86_64_NUMA - per_cpu(x86_cpu_to_node_map, cpu) = - early_per_cpu_map(x86_cpu_to_node_map, cpu); -#endif #ifdef CONFIG_X86_64 per_cpu(irq_stack_ptr, cpu) = per_cpu(irq_stack_union.irq_stack, cpu) + IRQ_STACK_SIZE - 64; +#ifdef CONFIG_NUMA + per_cpu(x86_cpu_to_node_map, cpu) = + early_per_cpu_map(x86_cpu_to_node_map, cpu); +#endif /* * Up to this point, CPU0 has been using .data.init * area. Reload %gs offset for CPU0. @@ -184,7 +158,7 @@ void __init setup_per_cpu_areas(void) /* indicate the early static arrays will soon be gone */ early_per_cpu_ptr(x86_cpu_to_apicid) = NULL; early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL; -#ifdef X86_64_NUMA +#if defined(CONFIG_X86_64) && defined(CONFIG_NUMA) early_per_cpu_ptr(x86_cpu_to_node_map) = NULL; #endif @@ -197,204 +171,3 @@ void __init setup_per_cpu_areas(void) #endif -#ifdef X86_64_NUMA - -/* - * Allocate node_to_cpumask_map based on number of available nodes - * Requires node_possible_map to be valid. - * - * Note: node_to_cpumask() is not valid until after this is done. - * (Use CONFIG_DEBUG_PER_CPU_MAPS to check this.) - */ -static void __init setup_node_to_cpumask_map(void) -{ - unsigned int node, num = 0; - cpumask_t *map; - - /* setup nr_node_ids if not done yet */ - if (nr_node_ids == MAX_NUMNODES) { - for_each_node_mask(node, node_possible_map) - num = node; - nr_node_ids = num + 1; - } - - /* allocate the map */ - map = alloc_bootmem_low(nr_node_ids * sizeof(cpumask_t)); - DBG("node_to_cpumask_map at %p for %d nodes\n", map, nr_node_ids); - - pr_debug("Node to cpumask map at %p for %d nodes\n", - map, nr_node_ids); - - /* node_to_cpumask() will now work */ - node_to_cpumask_map = map; -} - -void __cpuinit numa_set_node(int cpu, int node) -{ - int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); - - /* early setting, no percpu area yet */ - if (cpu_to_node_map) { - cpu_to_node_map[cpu] = node; - return; - } - -#ifdef CONFIG_DEBUG_PER_CPU_MAPS - if (cpu >= nr_cpu_ids || !per_cpu_offset(cpu)) { - printk(KERN_ERR "numa_set_node: invalid cpu# (%d)\n", cpu); - dump_stack(); - return; - } -#endif - per_cpu(x86_cpu_to_node_map, cpu) = node; - - if (node != NUMA_NO_NODE) - per_cpu(node_number, cpu) = node; -} - -void __cpuinit numa_clear_node(int cpu) -{ - numa_set_node(cpu, NUMA_NO_NODE); -} - -#ifndef CONFIG_DEBUG_PER_CPU_MAPS - -void __cpuinit numa_add_cpu(int cpu) -{ - cpu_set(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]); -} - -void __cpuinit numa_remove_cpu(int cpu) -{ - cpu_clear(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]); -} - -#else /* CONFIG_DEBUG_PER_CPU_MAPS */ - -/* - * --------- debug versions of the numa functions --------- - */ -static void __cpuinit numa_set_cpumask(int cpu, int enable) -{ - int node = early_cpu_to_node(cpu); - cpumask_t *mask; - char buf[64]; - - if (node_to_cpumask_map == NULL) { - printk(KERN_ERR "node_to_cpumask_map NULL\n"); - dump_stack(); - return; - } - - mask = &node_to_cpumask_map[node]; - if (enable) - cpu_set(cpu, *mask); - else - cpu_clear(cpu, *mask); - - cpulist_scnprintf(buf, sizeof(buf), mask); - printk(KERN_DEBUG "%s cpu %d node %d: mask now %s\n", - enable ? "numa_add_cpu" : "numa_remove_cpu", cpu, node, buf); -} - -void __cpuinit numa_add_cpu(int cpu) -{ - numa_set_cpumask(cpu, 1); -} - -void __cpuinit numa_remove_cpu(int cpu) -{ - numa_set_cpumask(cpu, 0); -} - -int cpu_to_node(int cpu) -{ - if (early_per_cpu_ptr(x86_cpu_to_node_map)) { - printk(KERN_WARNING - "cpu_to_node(%d): usage too early!\n", cpu); - dump_stack(); - return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu]; - } - return per_cpu(x86_cpu_to_node_map, cpu); -} -EXPORT_SYMBOL(cpu_to_node); - -/* - * Same function as cpu_to_node() but used if called before the - * per_cpu areas are setup. - */ -int early_cpu_to_node(int cpu) -{ - if (early_per_cpu_ptr(x86_cpu_to_node_map)) - return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu]; - - if (!per_cpu_offset(cpu)) { - printk(KERN_WARNING - "early_cpu_to_node(%d): no per_cpu area!\n", cpu); - dump_stack(); - return NUMA_NO_NODE; - } - return per_cpu(x86_cpu_to_node_map, cpu); -} - - -/* empty cpumask */ -static const cpumask_t cpu_mask_none; - -/* - * Returns a pointer to the bitmask of CPUs on Node 'node'. - */ -const cpumask_t *cpumask_of_node(int node) -{ - if (node_to_cpumask_map == NULL) { - printk(KERN_WARNING - "cpumask_of_node(%d): no node_to_cpumask_map!\n", - node); - dump_stack(); - return (const cpumask_t *)&cpu_online_map; - } - if (node >= nr_node_ids) { - printk(KERN_WARNING - "cpumask_of_node(%d): node > nr_node_ids(%d)\n", - node, nr_node_ids); - dump_stack(); - return &cpu_mask_none; - } - return &node_to_cpumask_map[node]; -} -EXPORT_SYMBOL(cpumask_of_node); - -/* - * Returns a bitmask of CPUs on Node 'node'. - * - * Side note: this function creates the returned cpumask on the stack - * so with a high NR_CPUS count, excessive stack space is used. The - * node_to_cpumask_ptr function should be used whenever possible. - */ -cpumask_t node_to_cpumask(int node) -{ - if (node_to_cpumask_map == NULL) { - printk(KERN_WARNING - "node_to_cpumask(%d): no node_to_cpumask_map!\n", node); - dump_stack(); - return cpu_online_map; - } - if (node >= nr_node_ids) { - printk(KERN_WARNING - "node_to_cpumask(%d): node > nr_node_ids(%d)\n", - node, nr_node_ids); - dump_stack(); - return cpu_mask_none; - } - return node_to_cpumask_map[node]; -} -EXPORT_SYMBOL(node_to_cpumask); - -/* - * --------- end of debug versions of the numa functions --------- - */ - -#endif /* CONFIG_DEBUG_PER_CPU_MAPS */ - -#endif /* X86_64_NUMA */ - -- cgit v1.2.3 From 2f2f52bad72f5e1ca5d1b9ad00a7b57a8cbd9159 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:47 +0900 Subject: x86: move setup_cpu_local_masks() Impact: Code movement, no functional change. Move setup_cpu_local_masks() to kernel/cpu/common.c, where the masks are defined. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 9 +++++++++ arch/x86/kernel/setup_percpu.c | 19 ------------------- 2 files changed, 9 insertions(+), 19 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 99904f288d6..67e30c8a282 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -52,6 +52,15 @@ cpumask_var_t cpu_initialized_mask; /* representing cpus for which sibling maps can be computed */ cpumask_var_t cpu_sibling_setup_mask; +/* correctly size the local cpu masks */ +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 */ cpumask_t cpu_callin_map; diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index cb6d622520b..7bebdba8eb8 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -51,25 +51,6 @@ DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID); EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid); EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); -#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 #ifdef CONFIG_X86_64 -- cgit v1.2.3 From 74631a248dc2c2129a96f6b8b706ed54bb5c3d3c Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:47 +0900 Subject: x86: always page-align per-cpu area start and size Impact: cleanup The way the code is written, align is always PAGE_SIZE. Simplify the code by removing the align variable. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 7bebdba8eb8..5d4a4964a8b 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -69,15 +69,12 @@ EXPORT_SYMBOL(__per_cpu_offset); */ void __init setup_per_cpu_areas(void) { - ssize_t size, old_size; + ssize_t size; char *ptr; int cpu; - unsigned long align = 1; /* Copy section for each CPU (we discard the original) */ - old_size = PERCPU_ENOUGH_ROOM; - align = max_t(unsigned long, PAGE_SIZE, align); - size = roundup(old_size, align); + size = roundup(PERCPU_ENOUGH_ROOM, PAGE_SIZE); pr_info("NR_CPUS:%d nr_cpumask_bits:%d nr_cpu_ids:%d nr_node_ids:%d\n", NR_CPUS, nr_cpumask_bits, nr_cpu_ids, nr_node_ids); @@ -86,20 +83,17 @@ void __init setup_per_cpu_areas(void) for_each_possible_cpu(cpu) { #ifndef CONFIG_NEED_MULTIPLE_NODES - ptr = __alloc_bootmem(size, align, - __pa(MAX_DMA_ADDRESS)); + ptr = alloc_bootmem_pages(size); #else int node = early_cpu_to_node(cpu); if (!node_online(node) || !NODE_DATA(node)) { - ptr = __alloc_bootmem(size, align, - __pa(MAX_DMA_ADDRESS)); + ptr = alloc_bootmem_pages(size); pr_info("cpu %d has no node %d or node-local memory\n", cpu, node); pr_debug("per cpu data for cpu%d at %016lx\n", cpu, __pa(ptr)); } else { - ptr = __alloc_bootmem_node(NODE_DATA(node), size, align, - __pa(MAX_DMA_ADDRESS)); + ptr = alloc_bootmem_pages_node(NODE_DATA(node), size); pr_debug("per cpu data for cpu%d on node%d at %016lx\n", cpu, node, __pa(ptr)); } -- cgit v1.2.3 From ec70de8b04bf37213982a5e8f303bc38679f3f8e Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:47 +0900 Subject: x86: move apic variables to apic.c Impact: Code movement Move the variable definitions to apic.c. Ifdef the copying of the two early per-cpu variables, since Voyager doesn't use them. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/apic.c | 18 ++++++++++++++++++ arch/x86/kernel/setup_percpu.c | 22 ++-------------------- 2 files changed, 20 insertions(+), 20 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index 1df341a528a..c6f15647eba 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -60,6 +60,24 @@ # error SPURIOUS_APIC_VECTOR definition error #endif +unsigned int num_processors; +unsigned disabled_cpus __cpuinitdata; +/* Processor that is doing the boot up */ +unsigned int boot_cpu_physical_apicid = -1U; +EXPORT_SYMBOL(boot_cpu_physical_apicid); +unsigned int max_physical_apicid; + +/* Bitmask of physically existing CPUs */ +physid_mask_t phys_cpu_present_map; + +/* + * Map cpu index to physical APIC ID + */ +DEFINE_EARLY_PER_CPU(u16, x86_cpu_to_apicid, BAD_APICID); +DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID); +EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid); +EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); + #ifdef CONFIG_X86_32 /* * Knob to control our willingness to enable the local APIC. diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 5d4a4964a8b..d367996693e 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -31,26 +31,6 @@ DEFINE_PER_CPU(int, cpu_number); EXPORT_PER_CPU_SYMBOL(cpu_number); #endif -#ifdef CONFIG_X86_LOCAL_APIC -unsigned int num_processors; -unsigned disabled_cpus __cpuinitdata; -/* Processor that is doing the boot up */ -unsigned int boot_cpu_physical_apicid = -1U; -EXPORT_SYMBOL(boot_cpu_physical_apicid); -unsigned int max_physical_apicid; - -/* Bitmask of physically existing CPUs */ -physid_mask_t phys_cpu_present_map; -#endif - -/* - * Map cpu index to physical APIC ID - */ -DEFINE_EARLY_PER_CPU(u16, x86_cpu_to_apicid, BAD_APICID); -DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID); -EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid); -EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); - #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA #ifdef CONFIG_X86_64 @@ -108,10 +88,12 @@ void __init setup_per_cpu_areas(void) * per cpu data areas. These arrays then become expendable and the * *_early_ptr's are zeroed indicating that the static arrays are gone. */ +#ifdef CONFIG_X86_LOCAL_APIC per_cpu(x86_cpu_to_apicid, cpu) = early_per_cpu_map(x86_cpu_to_apicid, cpu); per_cpu(x86_bios_cpu_apicid, cpu) = early_per_cpu_map(x86_bios_cpu_apicid, cpu); +#endif #ifdef CONFIG_X86_64 per_cpu(irq_stack_ptr, cpu) = per_cpu(irq_stack_union.irq_stack, cpu) + IRQ_STACK_SIZE - 64; -- cgit v1.2.3 From 996db817e3d1529d711e55b938d72ae4060b39fd Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:47 +0900 Subject: x86: only compile setup_percpu.o on SMP Impact: Minor build optimization Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/Makefile | 3 ++- arch/x86/kernel/setup_percpu.c | 6 ------ 2 files changed, 2 insertions(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index a99437c965c..73de055c29c 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -28,7 +28,7 @@ CFLAGS_paravirt.o := $(nostackp) obj-y := process_$(BITS).o signal.o entry_$(BITS).o obj-y += traps.o irq.o irq_$(BITS).o dumpstack_$(BITS).o obj-y += time_$(BITS).o ioport.o ldt.o dumpstack.o -obj-y += setup.o i8259.o irqinit_$(BITS).o setup_percpu.o +obj-y += setup.o i8259.o irqinit_$(BITS).o obj-$(CONFIG_X86_VISWS) += visws_quirks.o obj-$(CONFIG_X86_32) += probe_roms_32.o obj-$(CONFIG_X86_32) += sys_i386_32.o i386_ksyms_32.o @@ -59,6 +59,7 @@ apm-y := apm_32.o obj-$(CONFIG_APM) += apm.o obj-$(CONFIG_X86_SMP) += smp.o obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o +obj-$(CONFIG_SMP) += setup_percpu.o obj-$(CONFIG_X86_32_SMP) += smpcommon.o obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index d367996693e..f30ff691c34 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -22,14 +22,8 @@ # define DBG(x...) #endif -/* - * Could be inside CONFIG_HAVE_SETUP_PER_CPU_AREA with other stuff but - * voyager wants cpu_number too. - */ -#ifdef CONFIG_SMP DEFINE_PER_CPU(int, cpu_number); EXPORT_PER_CPU_SYMBOL(cpu_number); -#endif #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA -- cgit v1.2.3 From 1688401a0fddba8991aa5c0943b8ae9583998d60 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:48 +0900 Subject: x86: move this_cpu_offset Impact: Small cleanup Define BOOT_PERCPU_OFFSET and use it for this_cpu_offset and __per_cpu_offset initializers. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 15 ++++++++++----- arch/x86/kernel/smpcommon.c | 7 ------- 2 files changed, 10 insertions(+), 12 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index f30ff691c34..36c2e81dfc3 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -25,15 +25,20 @@ DEFINE_PER_CPU(int, cpu_number); EXPORT_PER_CPU_SYMBOL(cpu_number); +#ifdef CONFIG_X86_64 +#define BOOT_PERCPU_OFFSET ((unsigned long)__per_cpu_load) +#else +#define BOOT_PERCPU_OFFSET 0 +#endif + +DEFINE_PER_CPU(unsigned long, this_cpu_off) = BOOT_PERCPU_OFFSET; +EXPORT_PER_CPU_SYMBOL(this_cpu_off); + #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA -#ifdef CONFIG_X86_64 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { - [0] = (unsigned long)__per_cpu_load, + [0] = BOOT_PERCPU_OFFSET, }; -#else -unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; -#endif EXPORT_SYMBOL(__per_cpu_offset); /* diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c index add36b4e37c..5ec29a1a846 100644 --- a/arch/x86/kernel/smpcommon.c +++ b/arch/x86/kernel/smpcommon.c @@ -5,13 +5,6 @@ #include #include -#ifdef CONFIG_X86_64 -DEFINE_PER_CPU(unsigned long, this_cpu_off) = (unsigned long)__per_cpu_load; -#else -DEFINE_PER_CPU(unsigned long, this_cpu_off); -#endif -EXPORT_PER_CPU_SYMBOL(this_cpu_off); - #ifdef CONFIG_X86_32 /* * Initialize the CPU's GDT. This is either the boot CPU doing itself -- cgit v1.2.3 From 34019be1cd2941128b5de6d7c0fbdb51f967d268 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:48 +0900 Subject: x86: don't assume boot cpu is #0 Impact: minor cleanup Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 36c2e81dfc3..be77f1a1231 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -15,6 +15,7 @@ #include #include #include +#include #ifdef CONFIG_DEBUG_PER_CPU_MAPS # define DBG(x...) printk(KERN_DEBUG x) @@ -37,7 +38,7 @@ EXPORT_PER_CPU_SYMBOL(this_cpu_off); #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { - [0] = BOOT_PERCPU_OFFSET, + [0 ... NR_CPUS-1] = BOOT_PERCPU_OFFSET, }; EXPORT_SYMBOL(__per_cpu_offset); @@ -101,10 +102,10 @@ void __init setup_per_cpu_areas(void) early_per_cpu_map(x86_cpu_to_node_map, cpu); #endif /* - * Up to this point, CPU0 has been using .data.init - * area. Reload %gs offset for CPU0. + * Up to this point, the boot CPU has been using .data.init + * area. Reload %gs offset for the boot CPU. */ - if (cpu == 0) + if (cpu == boot_cpu_id) load_gs_base(cpu); #endif -- cgit v1.2.3 From 89c9c4c58ee86e6e8802597271f23679e0c46647 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:48 +0900 Subject: x86: make Voyager use x86 per-cpu setup. Impact: standardize all x86 platforms on same setup code With the preceding changes, Voyager can use the same per-cpu setup code as all the other x86 platforms. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index be77f1a1231..599dc1cc1da 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -35,8 +35,6 @@ EXPORT_PER_CPU_SYMBOL(cpu_number); DEFINE_PER_CPU(unsigned long, this_cpu_off) = BOOT_PERCPU_OFFSET; EXPORT_PER_CPU_SYMBOL(this_cpu_off); -#ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA - unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = BOOT_PERCPU_OFFSET, }; @@ -125,6 +123,3 @@ void __init setup_per_cpu_areas(void) /* Setup cpu initialized, callin, callout masks */ setup_cpu_local_masks(); } - -#endif - -- cgit v1.2.3 From b2d2f4312b117a6cc647c8521e2643a88771f757 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:48 +0900 Subject: x86: initialize per-cpu GDT segment in per-cpu setup Impact: cleanup Rename init_gdt() to setup_percpu_segment(), and move it to setup_percpu.c. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/Makefile | 3 +-- arch/x86/kernel/setup_percpu.c | 14 ++++++++++++++ arch/x86/kernel/smpboot.c | 4 ---- arch/x86/kernel/smpcommon.c | 25 ------------------------- 4 files changed, 15 insertions(+), 31 deletions(-) delete mode 100644 arch/x86/kernel/smpcommon.c (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 73de055c29c..37fa30bada1 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -60,8 +60,7 @@ obj-$(CONFIG_APM) += apm.o obj-$(CONFIG_X86_SMP) += smp.o obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o obj-$(CONFIG_SMP) += setup_percpu.o -obj-$(CONFIG_X86_32_SMP) += smpcommon.o -obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o +obj-$(CONFIG_X86_64_SMP) += tsc_sync.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o obj-$(CONFIG_X86_MPPARSE) += mpparse.o obj-$(CONFIG_X86_LOCAL_APIC) += apic.o nmi.o diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 599dc1cc1da..bcca3a7b374 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -40,6 +40,19 @@ unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { }; EXPORT_SYMBOL(__per_cpu_offset); +static inline void setup_percpu_segment(int cpu) +{ +#ifdef CONFIG_X86_32 + struct desc_struct gdt; + + pack_descriptor(&gdt, per_cpu_offset(cpu), 0xFFFFF, + 0x2 | DESCTYPE_S, 0x8); + gdt.s = 1; + write_gdt_entry(get_cpu_gdt_table(cpu), + GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S); +#endif +} + /* * Great future plan: * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data. @@ -81,6 +94,7 @@ void __init setup_per_cpu_areas(void) per_cpu_offset(cpu) = ptr - __per_cpu_start; per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); per_cpu(cpu_number, cpu) = cpu; + setup_percpu_segment(cpu); /* * Copy data used in early init routines from the initial arrays to the * per cpu data areas. These arrays then become expendable and the diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index def770b57b5..f9dbcff4354 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -793,7 +793,6 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu) do_rest: per_cpu(current_task, cpu) = c_idle.idle; #ifdef CONFIG_X86_32 - init_gdt(cpu); /* Stack for startup_32 can be just as for start_secondary onwards */ irq_ctx_init(cpu); #else @@ -1186,9 +1185,6 @@ out: void __init native_smp_prepare_boot_cpu(void) { int me = smp_processor_id(); -#ifdef CONFIG_X86_32 - init_gdt(me); -#endif switch_to_new_gdt(); /* already set me in cpu_online_mask in boot_cpu_init() */ cpumask_set_cpu(me, cpu_callout_mask); diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c deleted file mode 100644 index 5ec29a1a846..00000000000 --- a/arch/x86/kernel/smpcommon.c +++ /dev/null @@ -1,25 +0,0 @@ -/* - * SMP stuff which is common to all sub-architectures. - */ -#include -#include -#include - -#ifdef CONFIG_X86_32 -/* - * Initialize the CPU's GDT. This is either the boot CPU doing itself - * (still using the master per-cpu area), or a CPU doing it for a - * secondary which will soon come up. - */ -__cpuinit void init_gdt(int cpu) -{ - struct desc_struct gdt; - - pack_descriptor(&gdt, __per_cpu_offset[cpu], 0xFFFFF, - 0x2 | DESCTYPE_S, 0x8); - gdt.s = 1; - - write_gdt_entry(get_cpu_gdt_table(cpu), - GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S); -} -#endif -- cgit v1.2.3 From 2697fbd5faf19c84c17441b1752bdcbdcfd1248c Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:48 +0900 Subject: x86: load new GDT after setting up boot cpu per-cpu area Impact: sync 32 and 64-bit code Merge load_gs_base() into switch_to_new_gdt(). Load the GDT and per-cpu state for the boot cpu when its new area is set up. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 15 +++++++++------ arch/x86/kernel/setup_percpu.c | 6 +++--- 2 files changed, 12 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 67e30c8a282..0c766b80d91 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -258,12 +258,17 @@ __u32 cleared_cpu_caps[NCAPINTS] __cpuinitdata; void switch_to_new_gdt(void) { struct desc_ptr gdt_descr; + int cpu = smp_processor_id(); - gdt_descr.address = (long)get_cpu_gdt_table(smp_processor_id()); + gdt_descr.address = (long)get_cpu_gdt_table(cpu); gdt_descr.size = GDT_SIZE - 1; load_gdt(&gdt_descr); + /* Reload the per-cpu base */ #ifdef CONFIG_X86_32 - asm("mov %0, %%fs" : : "r" (__KERNEL_PERCPU) : "memory"); + loadsegment(fs, __KERNEL_PERCPU); +#else + loadsegment(gs, 0); + wrmsrl(MSR_GS_BASE, (unsigned long)per_cpu(irq_stack_union.gs_base, cpu)); #endif } @@ -968,10 +973,6 @@ void __cpuinit cpu_init(void) struct task_struct *me; int i; - loadsegment(fs, 0); - loadsegment(gs, 0); - load_gs_base(cpu); - #ifdef CONFIG_NUMA if (cpu != 0 && percpu_read(node_number) == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) @@ -993,6 +994,8 @@ void __cpuinit cpu_init(void) */ switch_to_new_gdt(); + loadsegment(fs, 0); + load_idt((const struct desc_ptr *)&idt_descr); memset(me->thread.tls_array, 0, GDT_ENTRY_TLS_ENTRIES * 8); diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index bcca3a7b374..4caa78d7cb1 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -112,14 +112,14 @@ void __init setup_per_cpu_areas(void) #ifdef CONFIG_NUMA per_cpu(x86_cpu_to_node_map, cpu) = early_per_cpu_map(x86_cpu_to_node_map, cpu); +#endif #endif /* * Up to this point, the boot CPU has been using .data.init - * area. Reload %gs offset for the boot CPU. + * area. Reload any changed state for the boot CPU. */ if (cpu == boot_cpu_id) - load_gs_base(cpu); -#endif + switch_to_new_gdt(); DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } -- cgit v1.2.3 From 22f25138c345ec46a13744c93c093ff822cd98d1 Mon Sep 17 00:00:00 2001 From: James Bottomley Date: Tue, 27 Jan 2009 14:21:37 +0900 Subject: x86: fix build breakage on voyage Impact: build fix x86_cpu_to_apicid and x86_bios_cpu_apicid aren't defined for voyage. Earlier patch forgot to conditionalize early percpu clearing. Fix it. Signed-off-by: James Bottomley Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 4caa78d7cb1..c7458ead22d 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -125,8 +125,10 @@ void __init setup_per_cpu_areas(void) } /* indicate the early static arrays will soon be gone */ +#ifdef CONFIG_X86_LOCAL_APIC early_per_cpu_ptr(x86_cpu_to_apicid) = NULL; early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL; +#endif #if defined(CONFIG_X86_64) && defined(CONFIG_NUMA) early_per_cpu_ptr(x86_cpu_to_node_map) = NULL; #endif -- cgit v1.2.3 From cf3997f507624757f149fcc42b76fb03c151fb65 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 27 Jan 2009 14:25:05 +0900 Subject: x86: clean up indentation in setup_per_cpu_areas() Impact: cosmetic cleanup Signed-off-by: Tejun Heo --- arch/x86/kernel/setup_percpu.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index c7458ead22d..0d1e7ac439f 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -96,22 +96,25 @@ void __init setup_per_cpu_areas(void) per_cpu(cpu_number, cpu) = cpu; setup_percpu_segment(cpu); /* - * Copy data used in early init routines from the initial arrays to the - * per cpu data areas. These arrays then become expendable and the - * *_early_ptr's are zeroed indicating that the static arrays are gone. + * Copy data used in early init routines from the + * initial arrays to the per cpu data areas. These + * arrays then become expendable and the *_early_ptr's + * are zeroed indicating that the static arrays are + * gone. */ #ifdef CONFIG_X86_LOCAL_APIC per_cpu(x86_cpu_to_apicid, cpu) = - early_per_cpu_map(x86_cpu_to_apicid, cpu); + early_per_cpu_map(x86_cpu_to_apicid, cpu); per_cpu(x86_bios_cpu_apicid, cpu) = - early_per_cpu_map(x86_bios_cpu_apicid, cpu); + early_per_cpu_map(x86_bios_cpu_apicid, cpu); #endif #ifdef CONFIG_X86_64 per_cpu(irq_stack_ptr, cpu) = - per_cpu(irq_stack_union.irq_stack, cpu) + IRQ_STACK_SIZE - 64; + per_cpu(irq_stack_union.irq_stack, cpu) + + IRQ_STACK_SIZE - 64; #ifdef CONFIG_NUMA per_cpu(x86_cpu_to_node_map, cpu) = - early_per_cpu_map(x86_cpu_to_node_map, cpu); + early_per_cpu_map(x86_cpu_to_node_map, cpu); #endif #endif /* -- cgit v1.2.3 From 41edafdb78feac1d1f8823846209975fde990633 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:02 -0800 Subject: x86/pvops: add a paravirt_ident functions to allow special patching Impact: Optimization Several paravirt ops implementations simply return their arguments, the most obvious being the make_pte/pte_val class of operations on native. On 32-bit, the identity function is literally a no-op, as the calling convention uses the same registers for the first argument and return. On 64-bit, it can be implemented with a single "mov". This patch adds special identity functions for 32 and 64 bit argument, and machinery to recognize them and replace them with either nops or a mov as appropriate. At the moment, the only users for the identity functions are the pagetable entry conversion functions. The result is a measureable improvement on pagetable-heavy benchmarks (2-3%, reducing the pvops overhead from 5 to 2%). Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/paravirt.c | 75 ++++++++++++++++++++++++++++++++----- arch/x86/kernel/paravirt_patch_32.c | 12 ++++++ arch/x86/kernel/paravirt_patch_64.c | 15 ++++++++ 3 files changed, 93 insertions(+), 9 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index 202514be592..dd25e2b1593 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -44,6 +44,17 @@ void _paravirt_nop(void) { } +/* identity function, which can be inlined */ +u32 _paravirt_ident_32(u32 x) +{ + return x; +} + +u64 _paravirt_ident_64(u64 x) +{ + return x; +} + static void __init default_banner(void) { printk(KERN_INFO "Booting paravirtualized kernel on %s\n", @@ -138,9 +149,16 @@ unsigned paravirt_patch_default(u8 type, u16 clobbers, void *insnbuf, if (opfunc == NULL) /* If there's no function, patch it with a ud2a (BUG) */ ret = paravirt_patch_insns(insnbuf, len, ud2a, ud2a+sizeof(ud2a)); - else if (opfunc == paravirt_nop) + else if (opfunc == _paravirt_nop) /* If the operation is a nop, then nop the callsite */ ret = paravirt_patch_nop(); + + /* identity functions just return their single argument */ + else if (opfunc == _paravirt_ident_32) + ret = paravirt_patch_ident_32(insnbuf, len); + else if (opfunc == _paravirt_ident_64) + ret = paravirt_patch_ident_64(insnbuf, len); + else if (type == PARAVIRT_PATCH(pv_cpu_ops.iret) || type == PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit) || type == PARAVIRT_PATCH(pv_cpu_ops.usergs_sysret32) || @@ -373,6 +391,45 @@ struct pv_apic_ops pv_apic_ops = { #endif }; +typedef pte_t make_pte_t(pteval_t); +typedef pmd_t make_pmd_t(pmdval_t); +typedef pud_t make_pud_t(pudval_t); +typedef pgd_t make_pgd_t(pgdval_t); + +typedef pteval_t pte_val_t(pte_t); +typedef pmdval_t pmd_val_t(pmd_t); +typedef pudval_t pud_val_t(pud_t); +typedef pgdval_t pgd_val_t(pgd_t); + + +#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE) +/* 32-bit pagetable entries */ +#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32 +#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32 + +#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32 +#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32 + +#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32 +#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32 + +#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32 +#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32 +#else +/* 64-bit pagetable entries */ +#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64 +#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64 + +#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64 +#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64 + +#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64 +#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64 + +#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64 +#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64 +#endif + struct pv_mmu_ops pv_mmu_ops = { #ifndef CONFIG_X86_64 .pagetable_setup_start = native_pagetable_setup_start, @@ -424,21 +481,21 @@ struct pv_mmu_ops pv_mmu_ops = { .pmd_clear = native_pmd_clear, #endif .set_pud = native_set_pud, - .pmd_val = native_pmd_val, - .make_pmd = native_make_pmd, + .pmd_val = paravirt_native_pmd_val, + .make_pmd = paravirt_native_make_pmd, #if PAGETABLE_LEVELS == 4 - .pud_val = native_pud_val, - .make_pud = native_make_pud, + .pud_val = paravirt_native_pud_val, + .make_pud = paravirt_native_make_pud, .set_pgd = native_set_pgd, #endif #endif /* PAGETABLE_LEVELS >= 3 */ - .pte_val = native_pte_val, - .pgd_val = native_pgd_val, + .pte_val = paravirt_native_pte_val, + .pgd_val = paravirt_native_pgd_val, - .make_pte = native_make_pte, - .make_pgd = native_make_pgd, + .make_pte = paravirt_native_make_pte, + .make_pgd = paravirt_native_make_pgd, .dup_mmap = paravirt_nop, .exit_mmap = paravirt_nop, diff --git a/arch/x86/kernel/paravirt_patch_32.c b/arch/x86/kernel/paravirt_patch_32.c index 9fe644f4861..d9f32e6d6ab 100644 --- a/arch/x86/kernel/paravirt_patch_32.c +++ b/arch/x86/kernel/paravirt_patch_32.c @@ -12,6 +12,18 @@ DEF_NATIVE(pv_mmu_ops, read_cr3, "mov %cr3, %eax"); DEF_NATIVE(pv_cpu_ops, clts, "clts"); DEF_NATIVE(pv_cpu_ops, read_tsc, "rdtsc"); +unsigned paravirt_patch_ident_32(void *insnbuf, unsigned len) +{ + /* arg in %eax, return in %eax */ + return 0; +} + +unsigned paravirt_patch_ident_64(void *insnbuf, unsigned len) +{ + /* arg in %edx:%eax, return in %edx:%eax */ + return 0; +} + unsigned native_patch(u8 type, u16 clobbers, void *ibuf, unsigned long addr, unsigned len) { diff --git a/arch/x86/kernel/paravirt_patch_64.c b/arch/x86/kernel/paravirt_patch_64.c index 061d01df9ae..3f08f34f93e 100644 --- a/arch/x86/kernel/paravirt_patch_64.c +++ b/arch/x86/kernel/paravirt_patch_64.c @@ -19,6 +19,21 @@ DEF_NATIVE(pv_cpu_ops, usergs_sysret64, "swapgs; sysretq"); DEF_NATIVE(pv_cpu_ops, usergs_sysret32, "swapgs; sysretl"); DEF_NATIVE(pv_cpu_ops, swapgs, "swapgs"); +DEF_NATIVE(, mov32, "mov %edi, %eax"); +DEF_NATIVE(, mov64, "mov %rdi, %rax"); + +unsigned paravirt_patch_ident_32(void *insnbuf, unsigned len) +{ + return paravirt_patch_insns(insnbuf, len, + start__mov32, end__mov32); +} + +unsigned paravirt_patch_ident_64(void *insnbuf, unsigned len) +{ + return paravirt_patch_insns(insnbuf, len, + start__mov64, end__mov64); +} + unsigned native_patch(u8 type, u16 clobbers, void *ibuf, unsigned long addr, unsigned len) { -- cgit v1.2.3 From b8aa287f77be943e37a84fa4657e27df95269bfb Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:03 -0800 Subject: x86: fix paravirt clobber in entry_64.S Impact: Fix latent bug The clobber is trying to say that anything except RDI is available for clobbering, but actually clobbers everything. This hasn't mattered because the clobbers were basically ignored, but subsequent patches will rely on them. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/entry_64.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index a52703864a1..e4c9710cae5 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -1140,7 +1140,7 @@ ENTRY(native_load_gs_index) CFI_STARTPROC pushf CFI_ADJUST_CFA_OFFSET 8 - DISABLE_INTERRUPTS(CLBR_ANY | ~(CLBR_RDI)) + DISABLE_INTERRUPTS(CLBR_ANY & ~CLBR_RDI) SWAPGS gs_change: movl %edi,%gs -- cgit v1.2.3 From ecb93d1ccd0aac63f03be2db3cac3fa974716f4c Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:05 -0800 Subject: x86/paravirt: add register-saving thunks to reduce caller register pressure Impact: Optimization One of the problems with inserting a pile of C calls where previously there were none is that the register pressure is greatly increased. The C calling convention says that the caller must expect a certain set of registers may be trashed by the callee, and that the callee can use those registers without restriction. This includes the function argument registers, and several others. This patch seeks to alleviate this pressure by introducing wrapper thunks that will do the register saving/restoring, so that the callsite doesn't need to worry about it, but the callee function can be conventional compiler-generated code. In many cases (particularly performance-sensitive cases) the callee will be in assembler anyway, and need not use the compiler's calling convention. Standard calling convention is: arguments return scratch x86-32 eax edx ecx eax ? x86-64 rdi rsi rdx rcx rax r8 r9 r10 r11 The thunk preserves all argument and scratch registers. The return register is not preserved, and is available as a scratch register for unwrapped callee code (and of course the return value). Wrapped function pointers are themselves wrapped in a struct paravirt_callee_save structure, in order to get some warning from the compiler when functions with mismatched calling conventions are used. The most common paravirt ops, both statically and dynamically, are interrupt enable/disable/save/restore, so handle them first. This is particularly easy since their calls are handled specially anyway. XXX Deal with VMI. What's their calling convention? Signed-off-by: H. Peter Anvin --- arch/x86/kernel/paravirt.c | 8 ++++---- arch/x86/kernel/vsmp_64.c | 12 ++++++++---- 2 files changed, 12 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index dd25e2b1593..8adb6b5aa42 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -310,10 +310,10 @@ struct pv_time_ops pv_time_ops = { struct pv_irq_ops pv_irq_ops = { .init_IRQ = native_init_IRQ, - .save_fl = native_save_fl, - .restore_fl = native_restore_fl, - .irq_disable = native_irq_disable, - .irq_enable = native_irq_enable, + .save_fl = __PV_IS_CALLEE_SAVE(native_save_fl), + .restore_fl = __PV_IS_CALLEE_SAVE(native_restore_fl), + .irq_disable = __PV_IS_CALLEE_SAVE(native_irq_disable), + .irq_enable = __PV_IS_CALLEE_SAVE(native_irq_enable), .safe_halt = native_safe_halt, .halt = native_halt, #ifdef CONFIG_X86_64 diff --git a/arch/x86/kernel/vsmp_64.c b/arch/x86/kernel/vsmp_64.c index a688f3bfaec..c609205df59 100644 --- a/arch/x86/kernel/vsmp_64.c +++ b/arch/x86/kernel/vsmp_64.c @@ -37,6 +37,7 @@ static unsigned long vsmp_save_fl(void) flags &= ~X86_EFLAGS_IF; return flags; } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_save_fl); static void vsmp_restore_fl(unsigned long flags) { @@ -46,6 +47,7 @@ static void vsmp_restore_fl(unsigned long flags) flags |= X86_EFLAGS_AC; native_restore_fl(flags); } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_restore_fl); static void vsmp_irq_disable(void) { @@ -53,6 +55,7 @@ static void vsmp_irq_disable(void) native_restore_fl((flags & ~X86_EFLAGS_IF) | X86_EFLAGS_AC); } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_irq_disable); static void vsmp_irq_enable(void) { @@ -60,6 +63,7 @@ static void vsmp_irq_enable(void) native_restore_fl((flags | X86_EFLAGS_IF) & (~X86_EFLAGS_AC)); } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_irq_enable); static unsigned __init_or_module vsmp_patch(u8 type, u16 clobbers, void *ibuf, unsigned long addr, unsigned len) @@ -90,10 +94,10 @@ static void __init set_vsmp_pv_ops(void) cap, ctl); if (cap & ctl & (1 << 4)) { /* Setup irq ops and turn on vSMP IRQ fastpath handling */ - pv_irq_ops.irq_disable = vsmp_irq_disable; - pv_irq_ops.irq_enable = vsmp_irq_enable; - pv_irq_ops.save_fl = vsmp_save_fl; - pv_irq_ops.restore_fl = vsmp_restore_fl; + pv_irq_ops.irq_disable = PV_CALLEE_SAVE(vsmp_irq_disable); + pv_irq_ops.irq_enable = PV_CALLEE_SAVE(vsmp_irq_enable); + pv_irq_ops.save_fl = PV_CALLEE_SAVE(vsmp_save_fl); + pv_irq_ops.restore_fl = PV_CALLEE_SAVE(vsmp_restore_fl); pv_init_ops.patch = vsmp_patch; ctl &= ~(1 << 4); -- cgit v1.2.3 From da5de7c22eb705be709a57e486e7475a6969b994 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:07 -0800 Subject: x86/paravirt: use callee-saved convention for pte_val/make_pte/etc Impact: Optimization In the native case, pte_val, make_pte, etc are all just identity functions, so there's no need to clobber a lot of registers over them. (This changes the 32-bit callee-save calling convention to return both EAX and EDX so functions can return 64-bit values.) Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/paravirt.c | 53 +++++++++++----------------------------------- 1 file changed, 12 insertions(+), 41 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index 8adb6b5aa42..cea11c8e304 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -391,43 +391,12 @@ struct pv_apic_ops pv_apic_ops = { #endif }; -typedef pte_t make_pte_t(pteval_t); -typedef pmd_t make_pmd_t(pmdval_t); -typedef pud_t make_pud_t(pudval_t); -typedef pgd_t make_pgd_t(pgdval_t); - -typedef pteval_t pte_val_t(pte_t); -typedef pmdval_t pmd_val_t(pmd_t); -typedef pudval_t pud_val_t(pud_t); -typedef pgdval_t pgd_val_t(pgd_t); - - #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE) /* 32-bit pagetable entries */ -#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32 -#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32 - -#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32 -#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32 - -#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32 -#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32 - -#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32 -#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32 +#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32) #else /* 64-bit pagetable entries */ -#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64 -#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64 - -#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64 -#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64 - -#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64 -#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64 - -#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64 -#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64 +#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64) #endif struct pv_mmu_ops pv_mmu_ops = { @@ -481,21 +450,23 @@ struct pv_mmu_ops pv_mmu_ops = { .pmd_clear = native_pmd_clear, #endif .set_pud = native_set_pud, - .pmd_val = paravirt_native_pmd_val, - .make_pmd = paravirt_native_make_pmd, + + .pmd_val = PTE_IDENT, + .make_pmd = PTE_IDENT, #if PAGETABLE_LEVELS == 4 - .pud_val = paravirt_native_pud_val, - .make_pud = paravirt_native_make_pud, + .pud_val = PTE_IDENT, + .make_pud = PTE_IDENT, + .set_pgd = native_set_pgd, #endif #endif /* PAGETABLE_LEVELS >= 3 */ - .pte_val = paravirt_native_pte_val, - .pgd_val = paravirt_native_pgd_val, + .pte_val = PTE_IDENT, + .pgd_val = PTE_IDENT, - .make_pte = paravirt_native_make_pte, - .make_pgd = paravirt_native_make_pgd, + .make_pte = PTE_IDENT, + .make_pgd = PTE_IDENT, .dup_mmap = paravirt_nop, .exit_mmap = paravirt_nop, -- cgit v1.2.3 From 2749ebe320ff9f77548d10fcc0a3464ac21c8e58 Mon Sep 17 00:00:00 2001 From: Cliff Wickman Date: Thu, 29 Jan 2009 15:35:26 -0600 Subject: x86: UV fix uv_flush_send_and_wait() Impact: fix possible tlb mis-flushing on UV uv_flush_send_and_wait() should return a pointer if the broadcast remote tlb shootdown requests fail. That causes the conventional IPI method of shootdown to be used. Signed-off-by: Cliff Wickman Signed-off-by: Tejun Heo --- arch/x86/kernel/tlb_uv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c index 89fce1b6d01..f4b2f27d19b 100644 --- a/arch/x86/kernel/tlb_uv.c +++ b/arch/x86/kernel/tlb_uv.c @@ -259,7 +259,7 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade, * the cpu's, all of which are still in the mask. */ __get_cpu_var(ptcstats).ptc_i++; - return 0; + return flush_mask; } /* -- cgit v1.2.3 From 552be871e67ff577ed36beb2f53d078b42304739 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Fri, 30 Jan 2009 17:47:53 +0900 Subject: x86: pass in cpu number to switch_to_new_gdt() Impact: cleanup, prepare for xen boot fix. Xen needs to call this function very early to setup the GDT and per-cpu segments. Remove the call to smp_processor_id() and just pass in the cpu number. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 7 +++---- arch/x86/kernel/setup_percpu.c | 2 +- arch/x86/kernel/smpboot.c | 2 +- 3 files changed, 5 insertions(+), 6 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 652fdc9a757..6eacd64b602 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -255,10 +255,9 @@ __u32 cleared_cpu_caps[NCAPINTS] __cpuinitdata; /* Current gdt points %fs at the "master" per-cpu area: after this, * it's on the real one. */ -void switch_to_new_gdt(void) +void switch_to_new_gdt(int cpu) { struct desc_ptr gdt_descr; - int cpu = smp_processor_id(); gdt_descr.address = (long)get_cpu_gdt_table(cpu); gdt_descr.size = GDT_SIZE - 1; @@ -993,7 +992,7 @@ void __cpuinit cpu_init(void) * and set up the GDT descriptor: */ - switch_to_new_gdt(); + switch_to_new_gdt(cpu); loadsegment(fs, 0); load_idt((const struct desc_ptr *)&idt_descr); @@ -1098,7 +1097,7 @@ void __cpuinit cpu_init(void) clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE); load_idt(&idt_descr); - switch_to_new_gdt(); + switch_to_new_gdt(cpu); /* * Set up and load the per-CPU TSS and LDT diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 0d1e7ac439f..ef91747bbed 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -122,7 +122,7 @@ void __init setup_per_cpu_areas(void) * area. Reload any changed state for the boot CPU. */ if (cpu == boot_cpu_id) - switch_to_new_gdt(); + switch_to_new_gdt(cpu); DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index f9dbcff4354..612d3c74f6a 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -1185,7 +1185,7 @@ out: void __init native_smp_prepare_boot_cpu(void) { int me = smp_processor_id(); - switch_to_new_gdt(); + switch_to_new_gdt(me); /* already set me in cpu_online_mask in boot_cpu_init() */ cpumask_set_cpu(me, cpu_callout_mask); per_cpu(cpu_state, me) = CPU_ONLINE; -- cgit v1.2.3 From 11e3a840cd5b731cdd8f6f956dfae78a8046d09c Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 30 Jan 2009 17:47:54 +0900 Subject: x86: split loading percpu segments from loading gdt Impact: split out a function, no functional change Xen needs to be able to access percpu data from very early on. For various reasons, it cannot also load the gdt at that time. It does, however, have a pefectly functional gdt at that point, so there's no pressing need to reload the gdt. Split the function to load the segment registers off, so Xen can call it directly. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Tejun Heo --- arch/x86/kernel/cpu/common.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 6eacd64b602..0f73ea42308 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -253,6 +253,16 @@ static char __cpuinit *table_lookup_model(struct cpuinfo_x86 *c) __u32 cleared_cpu_caps[NCAPINTS] __cpuinitdata; +void load_percpu_segment(int cpu) +{ +#ifdef CONFIG_X86_32 + loadsegment(fs, __KERNEL_PERCPU); +#else + loadsegment(gs, 0); + wrmsrl(MSR_GS_BASE, (unsigned long)per_cpu(irq_stack_union.gs_base, cpu)); +#endif +} + /* Current gdt points %fs at the "master" per-cpu area: after this, * it's on the real one. */ void switch_to_new_gdt(int cpu) @@ -263,12 +273,8 @@ void switch_to_new_gdt(int cpu) gdt_descr.size = GDT_SIZE - 1; load_gdt(&gdt_descr); /* Reload the per-cpu base */ -#ifdef CONFIG_X86_32 - loadsegment(fs, __KERNEL_PERCPU); -#else - loadsegment(gs, 0); - wrmsrl(MSR_GS_BASE, (unsigned long)per_cpu(irq_stack_union.gs_base, cpu)); -#endif + + load_percpu_segment(cpu); } static struct cpu_dev *cpu_devs[X86_VENDOR_NUM] = {}; -- cgit v1.2.3 From 664c7954721adfc9bd61de6ec78f89f482f1a802 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 30 Jan 2009 23:18:41 -0800 Subject: x86/vmi: fix interrupt enable/disable/save/restore calling convention. Zach says: > Enable/Disable have no clobbers at all. > Save clobbers only return value, %eax > Restore also clobbers nothing. This is precisely compatible with the calling convention, so we can just call them directly without wrapping. (Compile tested only.) Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/vmi_32.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/vmi_32.c b/arch/x86/kernel/vmi_32.c index 1d3302cc2dd..eb9e7347928 100644 --- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -670,10 +670,11 @@ static inline int __init activate_vmi(void) para_fill(pv_mmu_ops.write_cr2, SetCR2); para_fill(pv_mmu_ops.write_cr3, SetCR3); para_fill(pv_cpu_ops.write_cr4, SetCR4); - para_fill(pv_irq_ops.save_fl, GetInterruptMask); - para_fill(pv_irq_ops.restore_fl, SetInterruptMask); - para_fill(pv_irq_ops.irq_disable, DisableInterrupts); - para_fill(pv_irq_ops.irq_enable, EnableInterrupts); + + para_fill(pv_irq_ops.save_fl.func, GetInterruptMask); + para_fill(pv_irq_ops.restore_fl.func, SetInterruptMask); + para_fill(pv_irq_ops.irq_disable.func, DisableInterrupts); + para_fill(pv_irq_ops.irq_enable.func, EnableInterrupts); para_fill(pv_cpu_ops.wbinvd, WBINVD); para_fill(pv_cpu_ops.read_tsc, RDTSC); -- cgit v1.2.3 From ef3892bd63420380d115f755d351d2071f1f805f Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 2 Feb 2009 18:16:19 -0800 Subject: x86, percpu: fix kexec with vmlinux Impact: fix regression with kexec with vmlinux Split data.init into data.init, percpu, data.init2 sections instead of let data.init wrap percpu secion. Thus kexec loading will be happy, because sections will not overlap. Before the patch we have: Elf file type is EXEC (Executable file) Entry point 0x200000 There are 6 program headers, starting at offset 64 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flags Align LOAD 0x0000000000200000 0xffffffff80200000 0x0000000000200000 0x0000000000ca6000 0x0000000000ca6000 R E 200000 LOAD 0x0000000000ea6000 0xffffffff80ea6000 0x0000000000ea6000 0x000000000014dfe0 0x000000000014dfe0 RWE 200000 LOAD 0x0000000001000000 0xffffffffff600000 0x0000000000ff4000 0x0000000000000888 0x0000000000000888 RWE 200000 LOAD 0x00000000011f6000 0xffffffff80ff6000 0x0000000000ff6000 0x0000000000073086 0x0000000000a2d938 RWE 200000 LOAD 0x0000000001400000 0x0000000000000000 0x000000000106a000 0x00000000001d2ce0 0x00000000001d2ce0 RWE 200000 NOTE 0x00000000009e2c1c 0xffffffff809e2c1c 0x00000000009e2c1c 0x0000000000000024 0x0000000000000024 4 Section to Segment mapping: Segment Sections... 00 .text .notes __ex_table .rodata __bug_table .pci_fixup .builtin_fw __ksymtab __ksymtab_gpl __ksymtab_strings __init_rodata __param 01 .data .init.rodata .data.cacheline_aligned .data.read_mostly 02 .vsyscall_0 .vsyscall_fn .vsyscall_gtod_data .vsyscall_1 .vsyscall_2 .vgetcpu_mode .jiffies 03 .data.init_task .smp_locks .init.text .init.data .init.setup .initcall.init .con_initcall.init .x86_cpu_dev.init .altinstructions .altinstr_replacement .exit.text .init.ramfs .bss 04 .data.percpu 05 .notes After patch we've got: Elf file type is EXEC (Executable file) Entry point 0x200000 There are 7 program headers, starting at offset 64 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flags Align LOAD 0x0000000000200000 0xffffffff80200000 0x0000000000200000 0x0000000000ca6000 0x0000000000ca6000 R E 200000 LOAD 0x0000000000ea6000 0xffffffff80ea6000 0x0000000000ea6000 0x000000000014dfe0 0x000000000014dfe0 RWE 200000 LOAD 0x0000000001000000 0xffffffffff600000 0x0000000000ff4000 0x0000000000000888 0x0000000000000888 RWE 200000 LOAD 0x00000000011f6000 0xffffffff80ff6000 0x0000000000ff6000 0x0000000000073086 0x0000000000073086 RWE 200000 LOAD 0x0000000001400000 0x0000000000000000 0x000000000106a000 0x00000000001d2ce0 0x00000000001d2ce0 RWE 200000 LOAD 0x000000000163d000 0xffffffff8123d000 0x000000000123d000 0x0000000000000000 0x00000000007e6938 RWE 200000 NOTE 0x00000000009e2c1c 0xffffffff809e2c1c 0x00000000009e2c1c 0x0000000000000024 0x0000000000000024 4 Section to Segment mapping: Segment Sections... 00 .text .notes __ex_table .rodata __bug_table .pci_fixup .builtin_fw __ksymtab __ksymtab_gpl __ksymtab_strings __init_rodata __param 01 .data .init.rodata .data.cacheline_aligned .data.read_mostly 02 .vsyscall_0 .vsyscall_fn .vsyscall_gtod_data .vsyscall_1 .vsyscall_2 .vgetcpu_mode .jiffies 03 .data.init_task .smp_locks .init.text .init.data .init.setup .initcall.init .con_initcall.init .x86_cpu_dev.init .altinstructions .altinstr_replacement .exit.text .init.ramfs 04 .data.percpu 05 .bss 06 .notes Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/vmlinux_64.lds.S | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S index c9740996430..07f62d287ff 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -22,6 +22,7 @@ PHDRS { #ifdef CONFIG_SMP percpu PT_LOAD FLAGS(7); /* RWE */ #endif + data.init2 PT_LOAD FLAGS(7); /* RWE */ note PT_NOTE FLAGS(0); /* ___ */ } SECTIONS @@ -215,7 +216,7 @@ 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. Also, pda should be at the head of + * start another section data.init2. Also, pda should be at the head of * percpu area. Preallocate it and define the percpu offset symbol * so that it can be accessed as a percpu variable. */ @@ -232,7 +233,7 @@ SECTIONS __nosave_begin = .; .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) { *(.data.nosave) - } :data.init /* switch back to data.init, see PERCPU_VADDR() above */ + } :data.init2 /* use another section data.init2, see PERCPU_VADDR() above */ . = ALIGN(PAGE_SIZE); __nosave_end = .; -- cgit v1.2.3