From b6734c35af028f06772c0b2c836c7d579e6d4dad Mon Sep 17 00:00:00 2001 From: "H. Peter Anvin" Date: Mon, 18 Aug 2008 17:39:32 -0700 Subject: x86: add NOPL as a synthetic CPU feature bit The long noops ("NOPL") are supposed to be detected by family >= 6. Unfortunately, several non-Intel x86 implementations, both hardware and software, don't obey this dictum. Instead, probe for NOPL directly by executing a NOPL instruction and see if we get #UD. Signed-off-by: H. Peter Anvin --- arch/x86/kernel/cpu/common.c | 32 +++++++++++++++++++++++++++++++- arch/x86/kernel/cpu/common_64.c | 36 ++++++++++++++++++++++++++++++++++++ arch/x86/kernel/cpu/feature_names.c | 3 ++- 3 files changed, 69 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel/cpu') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 80ab20d4fa3..0785b3c8d04 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -13,6 +13,7 @@ #include #include #include +#include #ifdef CONFIG_X86_LOCAL_APIC #include #include @@ -341,6 +342,35 @@ static void __init early_cpu_detect(void) early_get_cap(c); } +/* + * The NOPL instruction is supposed to exist on all CPUs with + * family >= 6, unfortunately, that's not true in practice because + * of early VIA chips and (more importantly) broken virtualizers that + * are not easy to detect. Hence, probe for it based on first + * principles. + */ +static void __cpuinit detect_nopl(struct cpuinfo_x86 *c) +{ + const u32 nopl_signature = 0x888c53b1; /* Random number */ + u32 has_nopl = nopl_signature; + + clear_cpu_cap(c, X86_FEATURE_NOPL); + if (c->x86 >= 6) { + asm volatile("\n" + "1: .byte 0x0f,0x1f,0xc0\n" /* nopl %eax */ + "2:\n" + " .section .fixup,\"ax\"\n" + "3: xor %0,%0\n" + " jmp 2b\n" + " .previous\n" + _ASM_EXTABLE(1b,3b) + : "+a" (has_nopl)); + + if (has_nopl == nopl_signature) + set_cpu_cap(c, X86_FEATURE_NOPL); + } +} + static void __cpuinit generic_identify(struct cpuinfo_x86 *c) { u32 tfms, xlvl; @@ -395,8 +425,8 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 *c) } init_scattered_cpuid_features(c); + detect_nopl(c); } - } static void __cpuinit squash_the_stupid_serial_number(struct cpuinfo_x86 *c) diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index dd6e3f15017..c3afba5a81a 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -18,6 +18,7 @@ #include #include #include +#include #include #ifdef CONFIG_X86_LOCAL_APIC #include @@ -215,6 +216,39 @@ static void __init early_cpu_support_print(void) } } +/* + * The NOPL instruction is supposed to exist on all CPUs with + * family >= 6, unfortunately, that's not true in practice because + * of early VIA chips and (more importantly) broken virtualizers that + * are not easy to detect. Hence, probe for it based on first + * principles. + * + * Note: no 64-bit chip is known to lack these, but put the code here + * for consistency with 32 bits, and to make it utterly trivial to + * diagnose the problem should it ever surface. + */ +static void __cpuinit detect_nopl(struct cpuinfo_x86 *c) +{ + const u32 nopl_signature = 0x888c53b1; /* Random number */ + u32 has_nopl = nopl_signature; + + clear_cpu_cap(c, X86_FEATURE_NOPL); + if (c->x86 >= 6) { + asm volatile("\n" + "1: .byte 0x0f,0x1f,0xc0\n" /* nopl %eax */ + "2:\n" + " .section .fixup,\"ax\"\n" + "3: xor %0,%0\n" + " jmp 2b\n" + " .previous\n" + _ASM_EXTABLE(1b,3b) + : "+a" (has_nopl)); + + if (has_nopl == nopl_signature) + set_cpu_cap(c, X86_FEATURE_NOPL); + } +} + static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c); void __init early_cpu_init(void) @@ -313,6 +347,8 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c) c->x86_phys_bits = eax & 0xff; } + detect_nopl(c); + if (c->x86_vendor != X86_VENDOR_UNKNOWN && cpu_devs[c->x86_vendor]->c_early_init) cpu_devs[c->x86_vendor]->c_early_init(c); diff --git a/arch/x86/kernel/cpu/feature_names.c b/arch/x86/kernel/cpu/feature_names.c index e43ad4ad4cb..c9017799497 100644 --- a/arch/x86/kernel/cpu/feature_names.c +++ b/arch/x86/kernel/cpu/feature_names.c @@ -39,7 +39,8 @@ const char * const x86_cap_flags[NCAPINTS*32] = { NULL, NULL, NULL, NULL, "constant_tsc", "up", NULL, "arch_perfmon", "pebs", "bts", NULL, NULL, - "rep_good", NULL, NULL, NULL, NULL, NULL, NULL, NULL, + "rep_good", NULL, NULL, NULL, + "nopl", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, /* Intel-defined (#2) */ -- cgit v1.2.3 From 12cf105cd66d95cf32c73cfa847a50bd1b700f23 Mon Sep 17 00:00:00 2001 From: Krzysztof Helt Date: Thu, 4 Sep 2008 21:09:43 +0200 Subject: x86: delay early cpu initialization until cpuid is done Move early cpu initialization after cpu early get cap so the early cpu initialization can fix up cpu caps. Signed-off-by: Krzysztof Helt Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel/cpu') diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 0785b3c8d04..8aab8517642 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -335,11 +335,11 @@ static void __init early_cpu_detect(void) get_cpu_vendor(c, 1); + early_get_cap(c); + if (c->x86_vendor != X86_VENDOR_UNKNOWN && cpu_devs[c->x86_vendor]->c_early_init) cpu_devs[c->x86_vendor]->c_early_init(c); - - early_get_cap(c); } /* -- cgit v1.2.3 From dd786dd12c99634055a9066f25ea957f29991c22 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Thu, 4 Sep 2008 21:09:43 +0200 Subject: x86: move mtrr cpu cap setting early in early_init_xxxx Krzysztof Helt found MTRR is not detected on k6-2 root cause: we moved mtrr_bp_init() early for mtrr trimming, and in early_detect we only read the CPU capability from cpuid, so some cpu doesn't have that bit in cpuid. So we need to add early_init_xxxx to preset those bit before mtrr_bp_init for those earlier cpus. this patch is for v2.6.27 Reported-by: Krzysztof Helt Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/amd.c | 9 +++++---- arch/x86/kernel/cpu/centaur.c | 11 +++++++++++ arch/x86/kernel/cpu/cyrix.c | 32 ++++++++++++++++++++++++++++---- 3 files changed, 44 insertions(+), 8 deletions(-) (limited to 'arch/x86/kernel/cpu') diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index cae9cabc303..18514ed2610 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -31,6 +31,11 @@ static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) if (c->x86_power & (1<<8)) set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); } + + /* Set MTRR capability flag if appropriate */ + if (c->x86_model == 13 || c->x86_model == 9 || + (c->x86_model == 8 && c->x86_mask >= 8)) + set_cpu_cap(c, X86_FEATURE_K6_MTRR); } static void __cpuinit init_amd(struct cpuinfo_x86 *c) @@ -166,10 +171,6 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) mbytes); } - /* Set MTRR capability flag if appropriate */ - if (c->x86_model == 13 || c->x86_model == 9 || - (c->x86_model == 8 && c->x86_mask >= 8)) - set_cpu_cap(c, X86_FEATURE_K6_MTRR); break; } diff --git a/arch/x86/kernel/cpu/centaur.c b/arch/x86/kernel/cpu/centaur.c index e0f45edd6a5..a0534c04d38 100644 --- a/arch/x86/kernel/cpu/centaur.c +++ b/arch/x86/kernel/cpu/centaur.c @@ -314,6 +314,16 @@ enum { EAMD3D = 1<<20, }; +static void __cpuinit early_init_centaur(struct cpuinfo_x86 *c) +{ + switch (c->x86) { + case 5: + /* Emulate MTRRs using Centaur's MCR. */ + set_cpu_cap(c, X86_FEATURE_CENTAUR_MCR); + break; + } +} + static void __cpuinit init_centaur(struct cpuinfo_x86 *c) { @@ -462,6 +472,7 @@ centaur_size_cache(struct cpuinfo_x86 *c, unsigned int size) static struct cpu_dev centaur_cpu_dev __cpuinitdata = { .c_vendor = "Centaur", .c_ident = { "CentaurHauls" }, + .c_early_init = early_init_centaur, .c_init = init_centaur, .c_size_cache = centaur_size_cache, }; diff --git a/arch/x86/kernel/cpu/cyrix.c b/arch/x86/kernel/cpu/cyrix.c index e710a21bb6e..898a5a2002e 100644 --- a/arch/x86/kernel/cpu/cyrix.c +++ b/arch/x86/kernel/cpu/cyrix.c @@ -15,13 +15,11 @@ /* * Read NSC/Cyrix DEVID registers (DIR) to get more detailed info. about the CPU */ -static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) +static void __cpuinit __do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) { unsigned char ccr2, ccr3; - unsigned long flags; /* we test for DEVID by checking whether CCR3 is writable */ - local_irq_save(flags); ccr3 = getCx86(CX86_CCR3); setCx86(CX86_CCR3, ccr3 ^ 0x80); getCx86(0xc0); /* dummy to change bus */ @@ -44,9 +42,16 @@ static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) *dir0 = getCx86(CX86_DIR0); *dir1 = getCx86(CX86_DIR1); } - local_irq_restore(flags); } +static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) +{ + unsigned long flags; + + local_irq_save(flags); + __do_cyrix_devid(dir0, dir1); + local_irq_restore(flags); +} /* * Cx86_dir0_msb is a HACK needed by check_cx686_cpuid/slop in bugs.h in * order to identify the Cyrix CPU model after we're out of setup.c @@ -161,6 +166,24 @@ static void __cpuinit geode_configure(void) local_irq_restore(flags); } +static void __cpuinit early_init_cyrix(struct cpuinfo_x86 *c) +{ + unsigned char dir0, dir0_msn, dir1 = 0; + + __do_cyrix_devid(&dir0, &dir1); + dir0_msn = dir0 >> 4; /* identifies CPU "family" */ + + switch (dir0_msn) { + case 3: /* 6x86/6x86L */ + /* Emulate MTRRs using Cyrix's ARRs. */ + set_cpu_cap(c, X86_FEATURE_CYRIX_ARR); + break; + case 5: /* 6x86MX/M II */ + /* Emulate MTRRs using Cyrix's ARRs. */ + set_cpu_cap(c, X86_FEATURE_CYRIX_ARR); + break; + } +} static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) { @@ -416,6 +439,7 @@ static void __cpuinit cyrix_identify(struct cpuinfo_x86 *c) static struct cpu_dev cyrix_cpu_dev __cpuinitdata = { .c_vendor = "Cyrix", .c_ident = { "CyrixInstead" }, + .c_early_init = early_init_cyrix, .c_init = init_cyrix, .c_identify = cyrix_identify, }; -- cgit v1.2.3 From d04ec773d7ca1bbc05a2768be95c1cebe2b07757 Mon Sep 17 00:00:00 2001 From: Andreas Herrmann Date: Wed, 6 Aug 2008 10:27:30 +0200 Subject: x86: pda_init(): fix memory leak when using CPU hotplug pda->irqstackptr is allocated whenever a CPU is set online. But it is never freed. This results in a memory leak of 16K for each CPU offline/online cycle. Fix is to allocate pda->irqstackptr only once. Signed-off-by: Andreas Herrmann Cc: akpm@linux-foundation.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common_64.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'arch/x86/kernel/cpu') diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index c3afba5a81a..4f2eeb5652e 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -529,17 +529,20 @@ void pda_init(int cpu) /* others are initialized in smpboot.c */ pda->pcurrent = &init_task; pda->irqstackptr = boot_cpu_stack; + pda->irqstackptr += IRQSTACKSIZE - 64; } else { - pda->irqstackptr = (char *) - __get_free_pages(GFP_ATOMIC, IRQSTACK_ORDER); - if (!pda->irqstackptr) - panic("cannot allocate irqstack for cpu %d", cpu); + 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); } - - pda->irqstackptr += IRQSTACKSIZE-64; } char boot_exception_stacks[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + -- cgit v1.2.3 From 23952a96ae738277f3139b63d622e22984589031 Mon Sep 17 00:00:00 2001 From: Andreas Herrmann Date: Wed, 6 Aug 2008 10:29:37 +0200 Subject: x86: cpu_init(): fix memory leak when using CPU hotplug Exception stacks are allocated each time a CPU is set online. But the allocated space is never freed. Thus with one CPU hotplug offline/online cycle there is a memory leak of 24K (6 pages) for a CPU. Fix is to allocate exception stacks only once -- when the CPU is set online for the first time. Signed-off-by: Andreas Herrmann Cc: akpm@linux-foundation.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common_64.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-) (limited to 'arch/x86/kernel/cpu') diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index 4f2eeb5652e..a11f5d4477c 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -640,19 +640,22 @@ void __cpuinit cpu_init(void) /* * set up and load the per-CPU TSS */ - for (v = 0; v < N_EXCEPTION_STACKS; v++) { + 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 + [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER, + [DEBUG_STACK - 1] = DEBUG_STACK_ORDER }; - if (cpu) { - estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]); - if (!estacks) - panic("Cannot allocate exception stack %ld %d\n", - v, 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]; + orig_ist->ist[v] = t->x86_tss.ist[v] = + (unsigned long)estacks; } - estacks += PAGE_SIZE << order[v]; - orig_ist->ist[v] = t->x86_tss.ist[v] = (unsigned long)estacks; } t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap); -- cgit v1.2.3