common.c 35.7 KB
Newer Older
1
#include <linux/bootmem.h>
2
#include <linux/linkage.h>
3
#include <linux/bitops.h>
4
#include <linux/kernel.h>
5
#include <linux/module.h>
6 7
#include <linux/percpu.h>
#include <linux/string.h>
8
#include <linux/ctype.h>
L
Linus Torvalds 已提交
9
#include <linux/delay.h>
10 11
#include <linux/sched.h>
#include <linux/init.h>
12
#include <linux/kprobes.h>
13
#include <linux/kgdb.h>
L
Linus Torvalds 已提交
14
#include <linux/smp.h>
15
#include <linux/io.h>
16
#include <linux/syscore_ops.h>
17 18

#include <asm/stackprotector.h>
19
#include <asm/perf_event.h>
L
Linus Torvalds 已提交
20
#include <asm/mmu_context.h>
21
#include <asm/archrandom.h>
22 23
#include <asm/hypervisor.h>
#include <asm/processor.h>
24
#include <asm/tlbflush.h>
25
#include <asm/debugreg.h>
26
#include <asm/sections.h>
27
#include <asm/vsyscall.h>
A
Alan Cox 已提交
28 29
#include <linux/topology.h>
#include <linux/cpumask.h>
30
#include <asm/pgtable.h>
A
Arun Sharma 已提交
31
#include <linux/atomic.h>
32 33 34 35
#include <asm/proto.h>
#include <asm/setup.h>
#include <asm/apic.h>
#include <asm/desc.h>
36
#include <asm/fpu/internal.h>
37
#include <asm/mtrr.h>
A
Alan Cox 已提交
38
#include <linux/numa.h>
39 40
#include <asm/asm.h>
#include <asm/cpu.h>
41
#include <asm/mce.h>
42
#include <asm/msr.h>
43
#include <asm/pat.h>
44 45
#include <asm/microcode.h>
#include <asm/microcode_intel.h>
46 47

#ifdef CONFIG_X86_LOCAL_APIC
T
Tejun Heo 已提交
48
#include <asm/uv/uv.h>
L
Linus Torvalds 已提交
49 50 51 52
#endif

#include "cpu.h"

53 54
/* all of these masks are initialized in setup_cpu_local_masks() */
cpumask_var_t cpu_initialized_mask;
55 56
cpumask_var_t cpu_callout_mask;
cpumask_var_t cpu_callin_mask;
57 58 59 60

/* representing cpus for which sibling maps can be computed */
cpumask_var_t cpu_sibling_setup_mask;

B
Brian Gerst 已提交
61
/* correctly size the local cpu masks */
62
void __init setup_cpu_local_masks(void)
B
Brian Gerst 已提交
63 64 65 66 67 68 69
{
	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);
}

70
static void default_init(struct cpuinfo_x86 *c)
71 72
{
#ifdef CONFIG_X86_64
73
	cpu_detect_cache_sizes(c);
74 75 76 77 78 79 80 81 82 83 84 85 86
#else
	/* Not much we can do here... */
	/* Check if at least it has cpuid */
	if (c->cpuid_level == -1) {
		/* No cpuid. It must be an ancient CPU */
		if (c->x86 == 4)
			strcpy(c->x86_model_id, "486");
		else if (c->x86 == 3)
			strcpy(c->x86_model_id, "386");
	}
#endif
}

87
static const struct cpu_dev default_cpu = {
88 89 90 91 92
	.c_init		= default_init,
	.c_vendor	= "Unknown",
	.c_x86_vendor	= X86_VENDOR_UNKNOWN,
};

93
static const struct cpu_dev *this_cpu = &default_cpu;
94

B
Brian Gerst 已提交
95
DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = {
Y
Yinghai Lu 已提交
96
#ifdef CONFIG_X86_64
B
Brian Gerst 已提交
97 98 99 100 101
	/*
	 * 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
	 *
102
	 * TLS descriptors are currently at a different place compared to i386.
B
Brian Gerst 已提交
103 104
	 * Hopefully nobody expects them at a fixed place (Wine?)
	 */
A
Akinobu Mita 已提交
105 106 107 108 109 110
	[GDT_ENTRY_KERNEL32_CS]		= GDT_ENTRY_INIT(0xc09b, 0, 0xfffff),
	[GDT_ENTRY_KERNEL_CS]		= GDT_ENTRY_INIT(0xa09b, 0, 0xfffff),
	[GDT_ENTRY_KERNEL_DS]		= GDT_ENTRY_INIT(0xc093, 0, 0xfffff),
	[GDT_ENTRY_DEFAULT_USER32_CS]	= GDT_ENTRY_INIT(0xc0fb, 0, 0xfffff),
	[GDT_ENTRY_DEFAULT_USER_DS]	= GDT_ENTRY_INIT(0xc0f3, 0, 0xfffff),
	[GDT_ENTRY_DEFAULT_USER_CS]	= GDT_ENTRY_INIT(0xa0fb, 0, 0xfffff),
Y
Yinghai Lu 已提交
111
#else
A
Akinobu Mita 已提交
112 113 114 115
	[GDT_ENTRY_KERNEL_CS]		= GDT_ENTRY_INIT(0xc09a, 0, 0xfffff),
	[GDT_ENTRY_KERNEL_DS]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
	[GDT_ENTRY_DEFAULT_USER_CS]	= GDT_ENTRY_INIT(0xc0fa, 0, 0xfffff),
	[GDT_ENTRY_DEFAULT_USER_DS]	= GDT_ENTRY_INIT(0xc0f2, 0, 0xfffff),
116 117 118 119 120
	/*
	 * Segments used for calling PnP BIOS have byte granularity.
	 * They code segments and data segments have fixed 64k limits,
	 * the transfer segment sizes are set at run time.
	 */
121
	/* 32-bit code */
A
Akinobu Mita 已提交
122
	[GDT_ENTRY_PNPBIOS_CS32]	= GDT_ENTRY_INIT(0x409a, 0, 0xffff),
123
	/* 16-bit code */
A
Akinobu Mita 已提交
124
	[GDT_ENTRY_PNPBIOS_CS16]	= GDT_ENTRY_INIT(0x009a, 0, 0xffff),
125
	/* 16-bit data */
A
Akinobu Mita 已提交
126
	[GDT_ENTRY_PNPBIOS_DS]		= GDT_ENTRY_INIT(0x0092, 0, 0xffff),
127
	/* 16-bit data */
A
Akinobu Mita 已提交
128
	[GDT_ENTRY_PNPBIOS_TS1]		= GDT_ENTRY_INIT(0x0092, 0, 0),
129
	/* 16-bit data */
A
Akinobu Mita 已提交
130
	[GDT_ENTRY_PNPBIOS_TS2]		= GDT_ENTRY_INIT(0x0092, 0, 0),
131 132 133 134
	/*
	 * The APM segments have byte granularity and their bases
	 * are set at run time.  All have 64k limits.
	 */
135
	/* 32-bit code */
A
Akinobu Mita 已提交
136
	[GDT_ENTRY_APMBIOS_BASE]	= GDT_ENTRY_INIT(0x409a, 0, 0xffff),
137
	/* 16-bit code */
A
Akinobu Mita 已提交
138
	[GDT_ENTRY_APMBIOS_BASE+1]	= GDT_ENTRY_INIT(0x009a, 0, 0xffff),
139
	/* data */
140
	[GDT_ENTRY_APMBIOS_BASE+2]	= GDT_ENTRY_INIT(0x4092, 0, 0xffff),
141

A
Akinobu Mita 已提交
142 143
	[GDT_ENTRY_ESPFIX_SS]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
	[GDT_ENTRY_PERCPU]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
144
	GDT_STACK_CANARY_INIT
Y
Yinghai Lu 已提交
145
#endif
B
Brian Gerst 已提交
146
} };
147
EXPORT_PER_CPU_SYMBOL_GPL(gdt_page);
148

149
static int __init x86_mpx_setup(char *s)
150
{
151
	/* require an exact match without trailing characters */
152 153
	if (strlen(s))
		return 0;
154

155 156 157
	/* do not emit a message if the feature is not present */
	if (!boot_cpu_has(X86_FEATURE_MPX))
		return 1;
158

159 160
	setup_clear_cpu_cap(X86_FEATURE_MPX);
	pr_info("nompx: Intel Memory Protection Extensions (MPX) disabled\n");
161 162
	return 1;
}
163
__setup("nompx", x86_mpx_setup);
164

165
#ifdef CONFIG_X86_32
166 167
static int cachesize_override = -1;
static int disable_x86_serial_nr = 1;
L
Linus Torvalds 已提交
168

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
static int __init cachesize_setup(char *str)
{
	get_option(&str, &cachesize_override);
	return 1;
}
__setup("cachesize=", cachesize_setup);

static int __init x86_sep_setup(char *s)
{
	setup_clear_cpu_cap(X86_FEATURE_SEP);
	return 1;
}
__setup("nosep", x86_sep_setup);

/* Standard macro to see if a specific flag is changeable */
static inline int flag_is_changeable_p(u32 flag)
{
	u32 f1, f2;

188 189 190 191 192 193 194
	/*
	 * Cyrix and IDT cpus allow disabling of CPUID
	 * so the code below may return different results
	 * when it is executed before and after enabling
	 * the CPUID. Add "volatile" to not allow gcc to
	 * optimize the subsequent calls to this function.
	 */
I
Ingo Molnar 已提交
195 196 197 198 199 200 201 202 203 204 205
	asm volatile ("pushfl		\n\t"
		      "pushfl		\n\t"
		      "popl %0		\n\t"
		      "movl %0, %1	\n\t"
		      "xorl %2, %0	\n\t"
		      "pushl %0		\n\t"
		      "popfl		\n\t"
		      "pushfl		\n\t"
		      "popl %0		\n\t"
		      "popfl		\n\t"

206 207
		      : "=&r" (f1), "=&r" (f2)
		      : "ir" (flag));
208 209 210 211 212

	return ((f1^f2) & flag) != 0;
}

/* Probe for the CPUID instruction */
213
int have_cpuid_p(void)
214 215 216 217
{
	return flag_is_changeable_p(X86_EFLAGS_ID);
}

218
static void squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
219
{
I
Ingo Molnar 已提交
220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
	unsigned long lo, hi;

	if (!cpu_has(c, X86_FEATURE_PN) || !disable_x86_serial_nr)
		return;

	/* Disable processor serial number: */

	rdmsr(MSR_IA32_BBL_CR_CTL, lo, hi);
	lo |= 0x200000;
	wrmsr(MSR_IA32_BBL_CR_CTL, lo, hi);

	printk(KERN_NOTICE "CPU serial number disabled.\n");
	clear_cpu_cap(c, X86_FEATURE_PN);

	/* Disabling the serial number may affect the cpuid level */
	c->cpuid_level = cpuid_eax(0);
236 237 238 239 240 241 242 243
}

static int __init x86_serial_nr_setup(char *s)
{
	disable_x86_serial_nr = 0;
	return 1;
}
__setup("serialnumber", x86_serial_nr_setup);
244
#else
245 246 247 248 249 250 251
static inline int flag_is_changeable_p(u32 flag)
{
	return 1;
}
static inline void squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
{
}
252
#endif
253

254 255
static __init int setup_disable_smep(char *arg)
{
256
	setup_clear_cpu_cap(X86_FEATURE_SMEP);
257 258 259 260
	return 1;
}
__setup("nosmep", setup_disable_smep);

261
static __always_inline void setup_smep(struct cpuinfo_x86 *c)
262
{
263
	if (cpu_has(c, X86_FEATURE_SMEP))
A
Andy Lutomirski 已提交
264
		cr4_set_bits(X86_CR4_SMEP);
265 266
}

267 268
static __init int setup_disable_smap(char *arg)
{
269
	setup_clear_cpu_cap(X86_FEATURE_SMAP);
270 271 272 273
	return 1;
}
__setup("nosmap", setup_disable_smap);

274 275 276 277 278 279 280 281
static __always_inline void setup_smap(struct cpuinfo_x86 *c)
{
	unsigned long eflags;

	/* This should have been cleared long ago */
	raw_local_save_flags(eflags);
	BUG_ON(eflags & X86_EFLAGS_AC);

282 283
	if (cpu_has(c, X86_FEATURE_SMAP)) {
#ifdef CONFIG_X86_SMAP
A
Andy Lutomirski 已提交
284
		cr4_set_bits(X86_CR4_SMAP);
285
#else
A
Andy Lutomirski 已提交
286
		cr4_clear_bits(X86_CR4_SMAP);
287 288
#endif
	}
289 290
}

291 292 293 294 295 296 297 298 299
/*
 * Some CPU features depend on higher CPUID levels, which may not always
 * be available due to CPUID level capping or broken virtualization
 * software.  Add those features to this table to auto-disable them.
 */
struct cpuid_dependent_feature {
	u32 feature;
	u32 level;
};
I
Ingo Molnar 已提交
300

301
static const struct cpuid_dependent_feature
302 303 304 305 306 307 308
cpuid_dependent_features[] = {
	{ X86_FEATURE_MWAIT,		0x00000005 },
	{ X86_FEATURE_DCA,		0x00000009 },
	{ X86_FEATURE_XSAVE,		0x0000000d },
	{ 0, 0 }
};

309
static void filter_cpuid_features(struct cpuinfo_x86 *c, bool warn)
310 311
{
	const struct cpuid_dependent_feature *df;
312

313
	for (df = cpuid_dependent_features; df->feature; df++) {
I
Ingo Molnar 已提交
314 315 316

		if (!cpu_has(c, df->feature))
			continue;
317 318 319 320 321 322 323
		/*
		 * Note: cpuid_level is set to -1 if unavailable, but
		 * extended_extended_level is set to 0 if unavailable
		 * and the legitimate extended levels are all negative
		 * when signed; hence the weird messing around with
		 * signs here...
		 */
I
Ingo Molnar 已提交
324
		if (!((s32)df->level < 0 ?
325
		     (u32)df->level > (u32)c->extended_cpuid_level :
I
Ingo Molnar 已提交
326 327 328 329 330 331 332 333
		     (s32)df->level > (s32)c->cpuid_level))
			continue;

		clear_cpu_cap(c, df->feature);
		if (!warn)
			continue;

		printk(KERN_WARNING
334 335
		       "CPU: CPU feature " X86_CAP_FMT " disabled, no CPUID level 0x%x\n",
				x86_cap_flag(df->feature), df->level);
336
	}
337
}
338

339 340 341
/*
 * Naming convention should be: <Name> [(<Codename>)]
 * This table only is used unless init_<vendor>() below doesn't set it;
I
Ingo Molnar 已提交
342 343
 * in particular, if CPUID levels 0x80000002..4 are supported, this
 * isn't used
344 345 346
 */

/* Look up CPU names by table lookup. */
347
static const char *table_lookup_model(struct cpuinfo_x86 *c)
348
{
349 350
#ifdef CONFIG_X86_32
	const struct legacy_cpu_model_info *info;
351 352 353 354 355 356 357

	if (c->x86_model >= 16)
		return NULL;	/* Range check */

	if (!this_cpu)
		return NULL;

358
	info = this_cpu->legacy_models;
359

360
	while (info->family) {
361 362 363 364
		if (info->family == c->x86)
			return info->model_names[c->x86_model];
		info++;
	}
365
#endif
366 367 368
	return NULL;		/* Not found */
}

369 370
__u32 cpu_caps_cleared[NCAPINTS];
__u32 cpu_caps_set[NCAPINTS];
371

372 373 374 375 376 377 378 379
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
380
	load_stack_canary_segment();
381 382
}

I
Ingo Molnar 已提交
383 384 385 386
/*
 * Current gdt points %fs at the "master" per-cpu area: after this,
 * it's on the real one.
 */
387
void switch_to_new_gdt(int cpu)
388 389 390
{
	struct desc_ptr gdt_descr;

391
	gdt_descr.address = (long)get_cpu_gdt_table(cpu);
392 393
	gdt_descr.size = GDT_SIZE - 1;
	load_gdt(&gdt_descr);
394
	/* Reload the per-cpu base */
395 396

	load_percpu_segment(cpu);
397 398
}

399
static const struct cpu_dev *cpu_devs[X86_VENDOR_NUM] = {};
L
Linus Torvalds 已提交
400

401
static void get_model_name(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
402 403
{
	unsigned int *v;
404
	char *p, *q, *s;
L
Linus Torvalds 已提交
405

406
	if (c->extended_cpuid_level < 0x80000004)
407
		return;
L
Linus Torvalds 已提交
408

I
Ingo Molnar 已提交
409
	v = (unsigned int *)c->x86_model_id;
L
Linus Torvalds 已提交
410 411 412 413 414
	cpuid(0x80000002, &v[0], &v[1], &v[2], &v[3]);
	cpuid(0x80000003, &v[4], &v[5], &v[6], &v[7]);
	cpuid(0x80000004, &v[8], &v[9], &v[10], &v[11]);
	c->x86_model_id[48] = 0;

415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
	/* Trim whitespace */
	p = q = s = &c->x86_model_id[0];

	while (*p == ' ')
		p++;

	while (*p) {
		/* Note the last non-whitespace index */
		if (!isspace(*p))
			s = q;

		*q++ = *p++;
	}

	*(s + 1) = '\0';
L
Linus Torvalds 已提交
430 431
}

432
void cpu_detect_cache_sizes(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
433
{
434
	unsigned int n, dummy, ebx, ecx, edx, l2size;
L
Linus Torvalds 已提交
435

436
	n = c->extended_cpuid_level;
L
Linus Torvalds 已提交
437 438

	if (n >= 0x80000005) {
439 440
		cpuid(0x80000005, &dummy, &ebx, &ecx, &edx);
		c->x86_cache_size = (ecx>>24) + (edx>>24);
441 442 443 444
#ifdef CONFIG_X86_64
		/* On K8 L1 TLB is inclusive, so don't count it */
		c->x86_tlbsize = 0;
#endif
L
Linus Torvalds 已提交
445 446 447 448 449
	}

	if (n < 0x80000006)	/* Some chips just has a large L1. */
		return;

450
	cpuid(0x80000006, &dummy, &ebx, &ecx, &edx);
L
Linus Torvalds 已提交
451
	l2size = ecx >> 16;
452

453 454 455
#ifdef CONFIG_X86_64
	c->x86_tlbsize += ((ebx >> 16) & 0xfff) + (ebx & 0xfff);
#else
L
Linus Torvalds 已提交
456
	/* do processor-specific cache resizing */
457 458
	if (this_cpu->legacy_cache_size)
		l2size = this_cpu->legacy_cache_size(c, l2size);
L
Linus Torvalds 已提交
459 460 461 462 463

	/* Allow user to override all this if necessary. */
	if (cachesize_override != -1)
		l2size = cachesize_override;

464
	if (l2size == 0)
L
Linus Torvalds 已提交
465
		return;		/* Again, no L2 cache is possible */
466
#endif
L
Linus Torvalds 已提交
467 468 469 470

	c->x86_cache_size = l2size;
}

471 472 473 474 475 476
u16 __read_mostly tlb_lli_4k[NR_INFO];
u16 __read_mostly tlb_lli_2m[NR_INFO];
u16 __read_mostly tlb_lli_4m[NR_INFO];
u16 __read_mostly tlb_lld_4k[NR_INFO];
u16 __read_mostly tlb_lld_2m[NR_INFO];
u16 __read_mostly tlb_lld_4m[NR_INFO];
477
u16 __read_mostly tlb_lld_1g[NR_INFO];
478

479
static void cpu_detect_tlb(struct cpuinfo_x86 *c)
480 481 482 483
{
	if (this_cpu->c_detect_tlb)
		this_cpu->c_detect_tlb(c);

484
	pr_info("Last level iTLB entries: 4KB %d, 2MB %d, 4MB %d\n",
485
		tlb_lli_4k[ENTRIES], tlb_lli_2m[ENTRIES],
486 487 488 489 490
		tlb_lli_4m[ENTRIES]);

	pr_info("Last level dTLB entries: 4KB %d, 2MB %d, 4MB %d, 1GB %d\n",
		tlb_lld_4k[ENTRIES], tlb_lld_2m[ENTRIES],
		tlb_lld_4m[ENTRIES], tlb_lld_1g[ENTRIES]);
491 492
}

493
void detect_ht(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
494
{
B
Borislav Petkov 已提交
495
#ifdef CONFIG_SMP
496 497
	u32 eax, ebx, ecx, edx;
	int index_msb, core_bits;
498
	static bool printed;
L
Linus Torvalds 已提交
499

500
	if (!cpu_has(c, X86_FEATURE_HT))
501
		return;
L
Linus Torvalds 已提交
502

503 504
	if (cpu_has(c, X86_FEATURE_CMP_LEGACY))
		goto out;
L
Linus Torvalds 已提交
505

506 507
	if (cpu_has(c, X86_FEATURE_XTOPOLOGY))
		return;
L
Linus Torvalds 已提交
508

509
	cpuid(1, &eax, &ebx, &ecx, &edx);
L
Linus Torvalds 已提交
510

511 512 513
	smp_num_siblings = (ebx & 0xff0000) >> 16;

	if (smp_num_siblings == 1) {
514
		printk_once(KERN_INFO "CPU0: Hyper-Threading is disabled\n");
I
Ingo Molnar 已提交
515 516
		goto out;
	}
517

I
Ingo Molnar 已提交
518 519
	if (smp_num_siblings <= 1)
		goto out;
520

I
Ingo Molnar 已提交
521 522
	index_msb = get_count_order(smp_num_siblings);
	c->phys_proc_id = apic->phys_pkg_id(c->initial_apicid, index_msb);
523

I
Ingo Molnar 已提交
524
	smp_num_siblings = smp_num_siblings / c->x86_max_cores;
525

I
Ingo Molnar 已提交
526
	index_msb = get_count_order(smp_num_siblings);
527

I
Ingo Molnar 已提交
528
	core_bits = get_count_order(c->x86_max_cores);
529

I
Ingo Molnar 已提交
530 531
	c->cpu_core_id = apic->phys_pkg_id(c->initial_apicid, index_msb) &
				       ((1 << core_bits) - 1);
L
Linus Torvalds 已提交
532

533
out:
534
	if (!printed && (c->x86_max_cores * smp_num_siblings) > 1) {
535 536 537 538
		printk(KERN_INFO  "CPU: Physical Processor ID: %d\n",
		       c->phys_proc_id);
		printk(KERN_INFO  "CPU: Processor Core ID: %d\n",
		       c->cpu_core_id);
539
		printed = 1;
540 541
	}
#endif
542
}
L
Linus Torvalds 已提交
543

544
static void get_cpu_vendor(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
545 546
{
	char *v = c->x86_vendor_id;
I
Ingo Molnar 已提交
547
	int i;
L
Linus Torvalds 已提交
548 549

	for (i = 0; i < X86_VENDOR_NUM; i++) {
Y
Yinghai Lu 已提交
550 551 552 553 554 555
		if (!cpu_devs[i])
			break;

		if (!strcmp(v, cpu_devs[i]->c_ident[0]) ||
		    (cpu_devs[i]->c_ident[1] &&
		     !strcmp(v, cpu_devs[i]->c_ident[1]))) {
I
Ingo Molnar 已提交
556

Y
Yinghai Lu 已提交
557 558 559
			this_cpu = cpu_devs[i];
			c->x86_vendor = this_cpu->c_x86_vendor;
			return;
L
Linus Torvalds 已提交
560 561
		}
	}
Y
Yinghai Lu 已提交
562

563 564 565
	printk_once(KERN_ERR
			"CPU: vendor_id '%s' unknown, using generic init.\n" \
			"CPU: Your system may be unstable.\n", v);
Y
Yinghai Lu 已提交
566

567 568
	c->x86_vendor = X86_VENDOR_UNKNOWN;
	this_cpu = &default_cpu;
L
Linus Torvalds 已提交
569 570
}

571
void cpu_detect(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
572 573
{
	/* Get vendor name */
574 575 576 577
	cpuid(0x00000000, (unsigned int *)&c->cpuid_level,
	      (unsigned int *)&c->x86_vendor_id[0],
	      (unsigned int *)&c->x86_vendor_id[8],
	      (unsigned int *)&c->x86_vendor_id[4]);
L
Linus Torvalds 已提交
578 579

	c->x86 = 4;
580
	/* Intel-defined flags: level 0x00000001 */
L
Linus Torvalds 已提交
581 582
	if (c->cpuid_level >= 0x00000001) {
		u32 junk, tfms, cap0, misc;
I
Ingo Molnar 已提交
583

L
Linus Torvalds 已提交
584
		cpuid(0x00000001, &tfms, &misc, &junk, &cap0);
585 586 587
		c->x86 = (tfms >> 8) & 0xf;
		c->x86_model = (tfms >> 4) & 0xf;
		c->x86_mask = tfms & 0xf;
I
Ingo Molnar 已提交
588

589
		if (c->x86 == 0xf)
L
Linus Torvalds 已提交
590
			c->x86 += (tfms >> 20) & 0xff;
591
		if (c->x86 >= 0x6)
592
			c->x86_model += ((tfms >> 16) & 0xf) << 4;
I
Ingo Molnar 已提交
593

H
Huang, Ying 已提交
594 595
		if (cap0 & (1<<19)) {
			c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
596
			c->x86_cache_alignment = c->x86_clflush_size;
H
Huang, Ying 已提交
597
		}
L
Linus Torvalds 已提交
598 599
	}
}
600

601
void get_cpu_cap(struct cpuinfo_x86 *c)
602 603
{
	u32 tfms, xlvl;
604
	u32 ebx;
605

606 607 608
	/* Intel-defined flags: level 0x00000001 */
	if (c->cpuid_level >= 0x00000001) {
		u32 capability, excap;
I
Ingo Molnar 已提交
609

610 611 612 613
		cpuid(0x00000001, &tfms, &ebx, &excap, &capability);
		c->x86_capability[0] = capability;
		c->x86_capability[4] = excap;
	}
614

615 616 617 618 619 620
	/* Additional Intel-defined flags: level 0x00000007 */
	if (c->cpuid_level >= 0x00000007) {
		u32 eax, ebx, ecx, edx;

		cpuid_count(0x00000007, 0, &eax, &ebx, &ecx, &edx);

621
		c->x86_capability[9] = ebx;
622 623
	}

624 625 626 627 628 629 630 631 632
	/* Extended state features: level 0x0000000d */
	if (c->cpuid_level >= 0x0000000d) {
		u32 eax, ebx, ecx, edx;

		cpuid_count(0x0000000d, 1, &eax, &ebx, &ecx, &edx);

		c->x86_capability[10] = eax;
	}

633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
	/* Additional Intel-defined flags: level 0x0000000F */
	if (c->cpuid_level >= 0x0000000F) {
		u32 eax, ebx, ecx, edx;

		/* QoS sub-leaf, EAX=0Fh, ECX=0 */
		cpuid_count(0x0000000F, 0, &eax, &ebx, &ecx, &edx);
		c->x86_capability[11] = edx;
		if (cpu_has(c, X86_FEATURE_CQM_LLC)) {
			/* will be overridden if occupancy monitoring exists */
			c->x86_cache_max_rmid = ebx;

			/* QoS sub-leaf, EAX=0Fh, ECX=1 */
			cpuid_count(0x0000000F, 1, &eax, &ebx, &ecx, &edx);
			c->x86_capability[12] = edx;
			if (cpu_has(c, X86_FEATURE_CQM_OCCUP_LLC)) {
				c->x86_cache_max_rmid = ecx;
				c->x86_cache_occ_scale = ebx;
			}
		} else {
			c->x86_cache_max_rmid = -1;
			c->x86_cache_occ_scale = -1;
		}
	}

657 658 659
	/* AMD-defined flags: level 0x80000001 */
	xlvl = cpuid_eax(0x80000000);
	c->extended_cpuid_level = xlvl;
I
Ingo Molnar 已提交
660

661 662 663 664
	if ((xlvl & 0xffff0000) == 0x80000000) {
		if (xlvl >= 0x80000001) {
			c->x86_capability[1] = cpuid_edx(0x80000001);
			c->x86_capability[6] = cpuid_ecx(0x80000001);
665 666 667
		}
	}

668 669 670 671 672
	if (c->extended_cpuid_level >= 0x80000008) {
		u32 eax = cpuid_eax(0x80000008);

		c->x86_virt_bits = (eax >> 8) & 0xff;
		c->x86_phys_bits = eax & 0xff;
W
Wan Zongshun 已提交
673
		c->x86_capability[13] = cpuid_ebx(0x80000008);
674
	}
675 676 677
#ifdef CONFIG_X86_32
	else if (cpu_has(c, X86_FEATURE_PAE) || cpu_has(c, X86_FEATURE_PSE36))
		c->x86_phys_bits = 36;
678
#endif
679 680 681

	if (c->extended_cpuid_level >= 0x80000007)
		c->x86_power = cpuid_edx(0x80000007);
682

683
	init_scattered_cpuid_features(c);
684
}
L
Linus Torvalds 已提交
685

686
static void identify_cpu_without_cpuid(struct cpuinfo_x86 *c)
Y
Yinghai Lu 已提交
687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711
{
#ifdef CONFIG_X86_32
	int i;

	/*
	 * First of all, decide if this is a 486 or higher
	 * It's a 486 if we can modify the AC flag
	 */
	if (flag_is_changeable_p(X86_EFLAGS_AC))
		c->x86 = 4;
	else
		c->x86 = 3;

	for (i = 0; i < X86_VENDOR_NUM; i++)
		if (cpu_devs[i] && cpu_devs[i]->c_identify) {
			c->x86_vendor_id[0] = 0;
			cpu_devs[i]->c_identify(c);
			if (c->x86_vendor_id[0]) {
				get_cpu_vendor(c);
				break;
			}
		}
#endif
}

712 713 714 715 716 717 718 719 720
/*
 * Do minimum CPU detection early.
 * Fields really needed: vendor, cpuid_level, family, model, mask,
 * cache alignment.
 * The others are not touched to avoid unwanted side effects.
 *
 * WARNING: this function is only called on the BP.  Don't add code here
 * that is supposed to run on all CPUs.
 */
721
static void __init early_identify_cpu(struct cpuinfo_x86 *c)
722
{
723 724
#ifdef CONFIG_X86_64
	c->x86_clflush_size = 64;
725 726
	c->x86_phys_bits = 36;
	c->x86_virt_bits = 48;
727
#else
H
Huang, Ying 已提交
728
	c->x86_clflush_size = 32;
729 730
	c->x86_phys_bits = 32;
	c->x86_virt_bits = 32;
731
#endif
732
	c->x86_cache_alignment = c->x86_clflush_size;
733

734
	memset(&c->x86_capability, 0, sizeof c->x86_capability);
735
	c->extended_cpuid_level = 0;
736

Y
Yinghai Lu 已提交
737 738 739 740
	if (!have_cpuid_p())
		identify_cpu_without_cpuid(c);

	/* cyrix could have cpuid enabled via c_identify()*/
741 742 743 744
	if (!have_cpuid_p())
		return;

	cpu_detect(c);
745 746
	get_cpu_vendor(c);
	get_cpu_cap(c);
747

Y
Yinghai Lu 已提交
748 749
	if (this_cpu->c_early_init)
		this_cpu->c_early_init(c);
750

751
	c->cpu_index = 0;
752
	filter_cpuid_features(c, false);
753

B
Borislav Petkov 已提交
754 755
	if (this_cpu->c_bsp_init)
		this_cpu->c_bsp_init(c);
756 757

	setup_force_cpu_cap(X86_FEATURE_ALWAYS);
758
	fpu__init_system(c);
759 760
}

761 762
void __init early_cpu_init(void)
{
763
	const struct cpu_dev *const *cdev;
Y
Yinghai Lu 已提交
764 765
	int count = 0;

766
#ifdef CONFIG_PROCESSOR_SELECT
767
	printk(KERN_INFO "KERNEL supported cpus:\n");
768 769
#endif

Y
Yinghai Lu 已提交
770
	for (cdev = __x86_cpu_dev_start; cdev < __x86_cpu_dev_end; cdev++) {
771
		const struct cpu_dev *cpudev = *cdev;
772

Y
Yinghai Lu 已提交
773 774 775 776 777
		if (count >= X86_VENDOR_NUM)
			break;
		cpu_devs[count] = cpudev;
		count++;

778
#ifdef CONFIG_PROCESSOR_SELECT
779 780 781 782 783 784 785 786 787
		{
			unsigned int j;

			for (j = 0; j < 2; j++) {
				if (!cpudev->c_ident[j])
					continue;
				printk(KERN_INFO "  %s %s\n", cpudev->c_vendor,
					cpudev->c_ident[j]);
			}
Y
Yinghai Lu 已提交
788
		}
789
#endif
Y
Yinghai Lu 已提交
790
	}
791
	early_identify_cpu(&boot_cpu_data);
792
}
793

794
/*
B
Borislav Petkov 已提交
795 796 797 798 799
 * The NOPL instruction is supposed to exist on all CPUs of 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. In the latter case it doesn't even *fail* reliably, so
 * probing for it doesn't even work. Disable it completely on 32-bit
800
 * unless we can find a reliable way to detect all the broken cases.
B
Borislav Petkov 已提交
801
 * Enable it explicitly on 64-bit for non-constant inputs of cpu_has().
802
 */
803
static void detect_nopl(struct cpuinfo_x86 *c)
804
{
B
Borislav Petkov 已提交
805
#ifdef CONFIG_X86_32
806
	clear_cpu_cap(c, X86_FEATURE_NOPL);
B
Borislav Petkov 已提交
807 808 809
#else
	set_cpu_cap(c, X86_FEATURE_NOPL);
#endif
810 811
}

812
static void generic_identify(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
813
{
Y
Yinghai Lu 已提交
814
	c->extended_cpuid_level = 0;
L
Linus Torvalds 已提交
815

816
	if (!have_cpuid_p())
Y
Yinghai Lu 已提交
817
		identify_cpu_without_cpuid(c);
818

Y
Yinghai Lu 已提交
819
	/* cyrix could have cpuid enabled via c_identify()*/
I
Ingo Molnar 已提交
820
	if (!have_cpuid_p())
Y
Yinghai Lu 已提交
821
		return;
L
Linus Torvalds 已提交
822

823
	cpu_detect(c);
L
Linus Torvalds 已提交
824

825
	get_cpu_vendor(c);
L
Linus Torvalds 已提交
826

827
	get_cpu_cap(c);
L
Linus Torvalds 已提交
828

829 830
	if (c->cpuid_level >= 0x00000001) {
		c->initial_apicid = (cpuid_ebx(1) >> 24) & 0xFF;
831
#ifdef CONFIG_X86_32
B
Borislav Petkov 已提交
832
# ifdef CONFIG_SMP
833
		c->apicid = apic->phys_pkg_id(c->initial_apicid, 0);
834
# else
835
		c->apicid = c->initial_apicid;
836 837 838
# endif
#endif
		c->phys_proc_id = c->initial_apicid;
839
	}
L
Linus Torvalds 已提交
840

841
	get_model_name(c); /* Default name */
L
Linus Torvalds 已提交
842

843
	detect_nopl(c);
L
Linus Torvalds 已提交
844 845
}

846 847 848 849 850 851 852 853 854 855 856 857 858 859
static void x86_init_cache_qos(struct cpuinfo_x86 *c)
{
	/*
	 * The heavy lifting of max_rmid and cache_occ_scale are handled
	 * in get_cpu_cap().  Here we just set the max_rmid for the boot_cpu
	 * in case CQM bits really aren't there in this CPU.
	 */
	if (c != &boot_cpu_data) {
		boot_cpu_data.x86_cache_max_rmid =
			min(boot_cpu_data.x86_cache_max_rmid,
			    c->x86_cache_max_rmid);
	}
}

L
Linus Torvalds 已提交
860 861 862
/*
 * This does the hard work of actually picking apart the CPU stuff...
 */
863
static void identify_cpu(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
864 865 866 867 868 869 870 871 872
{
	int i;

	c->loops_per_jiffy = loops_per_jiffy;
	c->x86_cache_size = -1;
	c->x86_vendor = X86_VENDOR_UNKNOWN;
	c->x86_model = c->x86_mask = 0;	/* So far unknown... */
	c->x86_vendor_id[0] = '\0'; /* Unset */
	c->x86_model_id[0] = '\0';  /* Unset */
873
	c->x86_max_cores = 1;
874
	c->x86_coreid_bits = 0;
875
#ifdef CONFIG_X86_64
876
	c->x86_clflush_size = 64;
877 878
	c->x86_phys_bits = 36;
	c->x86_virt_bits = 48;
879 880
#else
	c->cpuid_level = -1;	/* CPUID not detected */
881
	c->x86_clflush_size = 32;
882 883
	c->x86_phys_bits = 32;
	c->x86_virt_bits = 32;
884 885
#endif
	c->x86_cache_alignment = c->x86_clflush_size;
L
Linus Torvalds 已提交
886 887 888 889
	memset(&c->x86_capability, 0, sizeof c->x86_capability);

	generic_identify(c);

890
	if (this_cpu->c_identify)
L
Linus Torvalds 已提交
891 892
		this_cpu->c_identify(c);

893 894 895 896 897 898
	/* Clear/Set all flags overriden by options, after probe */
	for (i = 0; i < NCAPINTS; i++) {
		c->x86_capability[i] &= ~cpu_caps_cleared[i];
		c->x86_capability[i] |= cpu_caps_set[i];
	}

899
#ifdef CONFIG_X86_64
900
	c->apicid = apic->phys_pkg_id(c->initial_apicid, 0);
901 902
#endif

L
Linus Torvalds 已提交
903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918
	/*
	 * Vendor-specific initialization.  In this section we
	 * canonicalize the feature flags, meaning if there are
	 * features a certain CPU supports which CPUID doesn't
	 * tell us, CPUID claiming incorrect flags, or other bugs,
	 * we handle them here.
	 *
	 * At the end of this section, c->x86_capability better
	 * indicate the features this CPU genuinely supports!
	 */
	if (this_cpu->c_init)
		this_cpu->c_init(c);

	/* Disable the PN if appropriate */
	squash_the_stupid_serial_number(c);

919 920 921 922
	/* Set up SMEP/SMAP */
	setup_smep(c);
	setup_smap(c);

L
Linus Torvalds 已提交
923
	/*
I
Ingo Molnar 已提交
924 925
	 * The vendor-specific functions might have changed features.
	 * Now we do "generic changes."
L
Linus Torvalds 已提交
926 927
	 */

928 929 930
	/* Filter out anything that depends on CPUID levels we don't have */
	filter_cpuid_features(c, true);

L
Linus Torvalds 已提交
931
	/* If the model name is still unset, do table lookup. */
932
	if (!c->x86_model_id[0]) {
933
		const char *p;
L
Linus Torvalds 已提交
934
		p = table_lookup_model(c);
935
		if (p)
L
Linus Torvalds 已提交
936 937 938 939
			strcpy(c->x86_model_id, p);
		else
			/* Last resort... */
			sprintf(c->x86_model_id, "%02x/%02x",
940
				c->x86, c->x86_model);
L
Linus Torvalds 已提交
941 942
	}

943 944 945 946
#ifdef CONFIG_X86_64
	detect_ht(c);
#endif

947
	init_hypervisor(c);
948
	x86_init_rdrand(c);
949
	x86_init_cache_qos(c);
950 951 952 953 954 955 956 957 958 959

	/*
	 * Clear/Set all flags overriden by options, need do it
	 * before following smp all cpus cap AND.
	 */
	for (i = 0; i < NCAPINTS; i++) {
		c->x86_capability[i] &= ~cpu_caps_cleared[i];
		c->x86_capability[i] |= cpu_caps_set[i];
	}

L
Linus Torvalds 已提交
960 961 962 963 964 965
	/*
	 * On SMP, boot_cpu_data holds the common feature set between
	 * all CPUs; so make sure that we indicate which features are
	 * common between the CPUs.  The first time this routine gets
	 * executed, c == &boot_cpu_data.
	 */
966
	if (c != &boot_cpu_data) {
L
Linus Torvalds 已提交
967
		/* AND the already accumulated flags with these */
968
		for (i = 0; i < NCAPINTS; i++)
L
Linus Torvalds 已提交
969
			boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
970 971 972 973

		/* OR, i.e. replicate the bug flags */
		for (i = NCAPINTS; i < NCAPINTS + NBUGINTS; i++)
			c->x86_capability[i] |= boot_cpu_data.x86_capability[i];
L
Linus Torvalds 已提交
974 975 976
	}

	/* Init Machine Check Exception if available. */
977
	mcheck_cpu_init(c);
978 979

	select_idle_routine(c);
980

981
#ifdef CONFIG_NUMA
982 983
	numa_add_cpu(smp_processor_id());
#endif
984
}
S
Shaohua Li 已提交
985

986 987 988 989
/*
 * Set up the CPU state needed to execute SYSENTER/SYSEXIT instructions
 * on 32-bit kernels:
 */
990 991 992
#ifdef CONFIG_X86_32
void enable_sep_cpu(void)
{
993 994
	struct tss_struct *tss;
	int cpu;
995

996 997 998 999 1000 1001 1002
	cpu = get_cpu();
	tss = &per_cpu(cpu_tss, cpu);

	if (!boot_cpu_has(X86_FEATURE_SEP))
		goto out;

	/*
1003 1004
	 * We cache MSR_IA32_SYSENTER_CS's value in the TSS's ss1 field --
	 * see the big comment in struct x86_hw_tss's definition.
1005
	 */
1006 1007

	tss->x86_tss.ss1 = __KERNEL_CS;
1008 1009
	wrmsr(MSR_IA32_SYSENTER_CS, tss->x86_tss.ss1, 0);

1010 1011 1012
	wrmsr(MSR_IA32_SYSENTER_ESP,
	      (unsigned long)tss + offsetofend(struct tss_struct, SYSENTER_stack),
	      0);
1013

1014
	wrmsr(MSR_IA32_SYSENTER_EIP, (unsigned long)entry_SYSENTER_32, 0);
1015 1016

out:
1017 1018
	put_cpu();
}
1019 1020
#endif

1021 1022 1023
void __init identify_boot_cpu(void)
{
	identify_cpu(&boot_cpu_data);
1024
	init_amd_e400_c1e_mask();
1025
#ifdef CONFIG_X86_32
1026
	sysenter_setup();
L
Li Shaohua 已提交
1027
	enable_sep_cpu();
1028
#endif
1029
	cpu_detect_tlb(&boot_cpu_data);
1030
}
S
Shaohua Li 已提交
1031

1032
void identify_secondary_cpu(struct cpuinfo_x86 *c)
1033 1034 1035
{
	BUG_ON(c == &boot_cpu_data);
	identify_cpu(c);
1036
#ifdef CONFIG_X86_32
1037
	enable_sep_cpu();
1038
#endif
1039
	mtrr_ap_init();
L
Linus Torvalds 已提交
1040 1041
}

1042
struct msr_range {
I
Ingo Molnar 已提交
1043 1044
	unsigned	min;
	unsigned	max;
1045
};
L
Linus Torvalds 已提交
1046

1047
static const struct msr_range msr_range_array[] = {
1048 1049 1050 1051 1052
	{ 0x00000000, 0x00000418},
	{ 0xc0000000, 0xc000040b},
	{ 0xc0010000, 0xc0010142},
	{ 0xc0011000, 0xc001103b},
};
L
Linus Torvalds 已提交
1053

1054
static void __print_cpu_msr(void)
1055
{
I
Ingo Molnar 已提交
1056
	unsigned index_min, index_max;
1057 1058 1059 1060 1061 1062 1063
	unsigned index;
	u64 val;
	int i;

	for (i = 0; i < ARRAY_SIZE(msr_range_array); i++) {
		index_min = msr_range_array[i].min;
		index_max = msr_range_array[i].max;
I
Ingo Molnar 已提交
1064

1065
		for (index = index_min; index < index_max; index++) {
1066
			if (rdmsrl_safe(index, &val))
1067 1068
				continue;
			printk(KERN_INFO " MSR%08x: %016llx\n", index, val);
L
Linus Torvalds 已提交
1069
		}
1070 1071
	}
}
1072

1073
static int show_msr;
I
Ingo Molnar 已提交
1074

1075 1076 1077
static __init int setup_show_msr(char *arg)
{
	int num;
1078

1079
	get_option(&arg, &num);
1080

1081 1082 1083
	if (num > 0)
		show_msr = num;
	return 1;
L
Linus Torvalds 已提交
1084
}
1085
__setup("show_msr=", setup_show_msr);
L
Linus Torvalds 已提交
1086

A
Andi Kleen 已提交
1087 1088
static __init int setup_noclflush(char *arg)
{
1089
	setup_clear_cpu_cap(X86_FEATURE_CLFLUSH);
1090
	setup_clear_cpu_cap(X86_FEATURE_CLFLUSHOPT);
A
Andi Kleen 已提交
1091 1092 1093 1094
	return 1;
}
__setup("noclflush", setup_noclflush);

1095
void print_cpu_info(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
1096
{
1097
	const char *vendor = NULL;
L
Linus Torvalds 已提交
1098

I
Ingo Molnar 已提交
1099
	if (c->x86_vendor < X86_VENDOR_NUM) {
L
Linus Torvalds 已提交
1100
		vendor = this_cpu->c_vendor;
I
Ingo Molnar 已提交
1101 1102 1103 1104
	} else {
		if (c->cpuid_level >= 0)
			vendor = c->x86_vendor_id;
	}
L
Linus Torvalds 已提交
1105

1106
	if (vendor && !strstr(c->x86_model_id, vendor))
1107
		printk(KERN_CONT "%s ", vendor);
L
Linus Torvalds 已提交
1108

1109
	if (c->x86_model_id[0])
1110
		printk(KERN_CONT "%s", c->x86_model_id);
L
Linus Torvalds 已提交
1111
	else
1112
		printk(KERN_CONT "%d86", c->x86);
L
Linus Torvalds 已提交
1113

1114
	printk(KERN_CONT " (family: 0x%x, model: 0x%x", c->x86, c->x86_model);
1115

1116
	if (c->x86_mask || c->cpuid_level >= 0)
1117
		printk(KERN_CONT ", stepping: 0x%x)\n", c->x86_mask);
L
Linus Torvalds 已提交
1118
	else
1119
		printk(KERN_CONT ")\n");
1120

1121
	print_cpu_msr(c);
1122 1123
}

1124
void print_cpu_msr(struct cpuinfo_x86 *c)
1125
{
1126
	if (c->cpu_index < show_msr)
1127
		__print_cpu_msr();
L
Linus Torvalds 已提交
1128 1129
}

1130 1131 1132
static __init int setup_disablecpuid(char *arg)
{
	int bit;
I
Ingo Molnar 已提交
1133

1134 1135 1136 1137
	if (get_option(&arg, &bit) && bit < NCAPINTS*32)
		setup_clear_cpu_cap(bit);
	else
		return 0;
I
Ingo Molnar 已提交
1138

1139 1140 1141 1142
	return 1;
}
__setup("clearcpuid=", setup_disablecpuid);

1143
#ifdef CONFIG_X86_64
1144
struct desc_ptr idt_descr = { NR_VECTORS * 16 - 1, (unsigned long) idt_table };
1145 1146
struct desc_ptr debug_idt_descr = { NR_VECTORS * 16 - 1,
				    (unsigned long) debug_idt_table };
1147

1148
DEFINE_PER_CPU_FIRST(union irq_stack_union,
1149
		     irq_stack_union) __aligned(PAGE_SIZE) __visible;
I
Ingo Molnar 已提交
1150

1151
/*
1152 1153
 * The following percpu variables are hot.  Align current_task to
 * cacheline size such that they fall in the same cacheline.
1154 1155 1156 1157
 */
DEFINE_PER_CPU(struct task_struct *, current_task) ____cacheline_aligned =
	&init_task;
EXPORT_PER_CPU_SYMBOL(current_task);
1158

1159 1160 1161
DEFINE_PER_CPU(char *, irq_stack_ptr) =
	init_per_cpu_var(irq_stack_union.irq_stack) + IRQ_STACK_SIZE - 64;

1162
DEFINE_PER_CPU(unsigned int, irq_count) __visible = -1;
1163

1164 1165 1166
DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
EXPORT_PER_CPU_SYMBOL(__preempt_count);

I
Ingo Molnar 已提交
1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
/*
 * Special IST stacks which the CPU switches to when it calls
 * an IST-marked descriptor entry. Up to 7 stacks (hardware
 * limit), all of them are 4K, except the debug stack which
 * is 8K.
 */
static const unsigned int exception_stack_sizes[N_EXCEPTION_STACKS] = {
	  [0 ... N_EXCEPTION_STACKS - 1]	= EXCEPTION_STKSZ,
	  [DEBUG_STACK - 1]			= DEBUG_STKSZ
};

1178
static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks
1179
	[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + DEBUG_STKSZ]);
1180 1181 1182

/* May not be marked __init: used by software suspend */
void syscall_init(void)
L
Linus Torvalds 已提交
1183
{
1184 1185 1186 1187 1188 1189
	/*
	 * LSTAR and STAR live in a bit strange symbiosis.
	 * They both write to the same internal register. STAR allows to
	 * set CS/DS but only a 32bit target. LSTAR sets the 64bit rip.
	 */
	wrmsrl(MSR_STAR,  ((u64)__USER32_CS)<<48  | ((u64)__KERNEL_CS)<<32);
1190
	wrmsrl(MSR_LSTAR, (unsigned long)entry_SYSCALL_64);
1191 1192

#ifdef CONFIG_IA32_EMULATION
1193
	wrmsrl(MSR_CSTAR, (unsigned long)entry_SYSCALL_compat);
1194
	/*
1195 1196 1197 1198
	 * This only works on Intel CPUs.
	 * On AMD CPUs these MSRs are 32-bit, CPU truncates MSR_IA32_SYSENTER_EIP.
	 * This does not cause SYSENTER to jump to the wrong location, because
	 * AMD doesn't allow SYSENTER in long mode (either 32- or 64-bit).
1199 1200 1201
	 */
	wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS);
	wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL);
1202
	wrmsrl_safe(MSR_IA32_SYSENTER_EIP, (u64)entry_SYSENTER_compat);
1203
#else
1204
	wrmsrl(MSR_CSTAR, (unsigned long)ignore_sysret);
1205
	wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)GDT_ENTRY_INVALID_SEG);
1206 1207
	wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL);
	wrmsrl_safe(MSR_IA32_SYSENTER_EIP, 0ULL);
1208
#endif
1209

1210 1211
	/* Flags to clear on syscall */
	wrmsrl(MSR_SYSCALL_MASK,
1212
	       X86_EFLAGS_TF|X86_EFLAGS_DF|X86_EFLAGS_IF|
1213
	       X86_EFLAGS_IOPL|X86_EFLAGS_AC|X86_EFLAGS_NT);
L
Linus Torvalds 已提交
1214
}
1215

1216 1217 1218 1219 1220 1221
/*
 * Copies of the original ist values from the tss are only accessed during
 * debugging, no special alignment required.
 */
DEFINE_PER_CPU(struct orig_ist, orig_ist);

1222
static DEFINE_PER_CPU(unsigned long, debug_stack_addr);
1223
DEFINE_PER_CPU(int, debug_stack_usage);
1224 1225 1226

int is_debug_stack(unsigned long addr)
{
1227 1228 1229
	return __this_cpu_read(debug_stack_usage) ||
		(addr <= __this_cpu_read(debug_stack_addr) &&
		 addr > (__this_cpu_read(debug_stack_addr) - DEBUG_STKSZ));
1230
}
1231
NOKPROBE_SYMBOL(is_debug_stack);
1232

1233
DEFINE_PER_CPU(u32, debug_idt_ctr);
1234

1235 1236
void debug_stack_set_zero(void)
{
1237 1238
	this_cpu_inc(debug_idt_ctr);
	load_current_idt();
1239
}
1240
NOKPROBE_SYMBOL(debug_stack_set_zero);
1241 1242 1243

void debug_stack_reset(void)
{
1244
	if (WARN_ON(!this_cpu_read(debug_idt_ctr)))
1245
		return;
1246 1247
	if (this_cpu_dec_return(debug_idt_ctr) == 0)
		load_current_idt();
1248
}
1249
NOKPROBE_SYMBOL(debug_stack_reset);
1250

I
Ingo Molnar 已提交
1251
#else	/* CONFIG_X86_64 */
1252

1253 1254
DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
EXPORT_PER_CPU_SYMBOL(current_task);
1255 1256
DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
EXPORT_PER_CPU_SYMBOL(__preempt_count);
1257

1258 1259 1260 1261 1262 1263 1264 1265 1266
/*
 * On x86_32, vm86 modifies tss.sp0, so sp0 isn't a reliable way to find
 * the top of the kernel stack.  Use an extra percpu variable to track the
 * top of the kernel stack directly.
 */
DEFINE_PER_CPU(unsigned long, cpu_current_top_of_stack) =
	(unsigned long)&init_thread_union + THREAD_SIZE;
EXPORT_PER_CPU_SYMBOL(cpu_current_top_of_stack);

1267
#ifdef CONFIG_CC_STACKPROTECTOR
1268
DEFINE_PER_CPU_ALIGNED(struct stack_canary, stack_canary);
1269
#endif
1270

I
Ingo Molnar 已提交
1271
#endif	/* CONFIG_X86_64 */
1272

1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
/*
 * Clear all 6 debug registers:
 */
static void clear_all_debug_regs(void)
{
	int i;

	for (i = 0; i < 8; i++) {
		/* Ignore db4, db5 */
		if ((i == 4) || (i == 5))
			continue;

		set_debugreg(0, i);
	}
}
1288

1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
#ifdef CONFIG_KGDB
/*
 * Restore debug regs if using kgdbwait and you have a kernel debugger
 * connection established.
 */
static void dbg_restore_debug_regs(void)
{
	if (unlikely(kgdb_connected && arch_kgdb_ops.correct_hw_break))
		arch_kgdb_ops.correct_hw_break();
}
#else /* ! CONFIG_KGDB */
#define dbg_restore_debug_regs()
#endif /* ! CONFIG_KGDB */

1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
static void wait_for_master_cpu(int cpu)
{
#ifdef CONFIG_SMP
	/*
	 * wait for ACK from master CPU before continuing
	 * with AP initialization
	 */
	WARN_ON(cpumask_test_and_set_cpu(cpu, cpu_initialized_mask));
	while (!cpumask_test_cpu(cpu, cpu_callout_mask))
		cpu_relax();
#endif
}

1316 1317 1318 1319 1320
/*
 * cpu_init() initializes state that is per-CPU. Some data is already
 * initialized (naturally) in the bootstrap process, such as the GDT
 * and IDT. We reload them nevertheless, this function acts as a
 * 'CPU state barrier', nothing should get across.
1321
 * A lot of state is already set up in PDA init for 64 bit
1322
 */
1323
#ifdef CONFIG_X86_64
I
Ingo Molnar 已提交
1324

1325
void cpu_init(void)
1326
{
1327
	struct orig_ist *oist;
1328
	struct task_struct *me;
I
Ingo Molnar 已提交
1329 1330
	struct tss_struct *t;
	unsigned long v;
1331
	int cpu = stack_smp_processor_id();
1332 1333
	int i;

1334 1335
	wait_for_master_cpu(cpu);

1336 1337 1338 1339 1340 1341
	/*
	 * Initialize the CR4 shadow before doing anything that could
	 * try to read it.
	 */
	cr4_init_shadow();

1342 1343 1344 1345 1346 1347
	/*
	 * Load microcode on this cpu if a valid microcode is available.
	 * This is early microcode loading procedure.
	 */
	load_ucode_ap();

1348
	t = &per_cpu(cpu_tss, cpu);
1349
	oist = &per_cpu(orig_ist, cpu);
I
Ingo Molnar 已提交
1350

1351
#ifdef CONFIG_NUMA
1352
	if (this_cpu_read(numa_node) == 0 &&
1353 1354
	    early_cpu_to_node(cpu) != NUMA_NO_NODE)
		set_numa_node(early_cpu_to_node(cpu));
1355
#endif
1356 1357 1358

	me = current;

1359
	pr_debug("Initializing CPU#%d\n", cpu);
1360

A
Andy Lutomirski 已提交
1361
	cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
1362 1363 1364 1365 1366 1367

	/*
	 * Initialize the per-CPU GDT with the boot GDT,
	 * and set up the GDT descriptor:
	 */

1368
	switch_to_new_gdt(cpu);
1369 1370
	loadsegment(fs, 0);

1371
	load_current_idt();
1372 1373 1374 1375 1376 1377 1378 1379

	memset(me->thread.tls_array, 0, GDT_ENTRY_TLS_ENTRIES * 8);
	syscall_init();

	wrmsrl(MSR_FS_BASE, 0);
	wrmsrl(MSR_KERNEL_GS_BASE, 0);
	barrier();

1380
	x86_configure_nx();
1381
	x2apic_setup();
1382 1383 1384 1385

	/*
	 * set up and load the per-CPU TSS
	 */
1386
	if (!oist->ist[0]) {
1387
		char *estacks = per_cpu(exception_stacks, cpu);
I
Ingo Molnar 已提交
1388

1389
		for (v = 0; v < N_EXCEPTION_STACKS; v++) {
I
Ingo Molnar 已提交
1390
			estacks += exception_stack_sizes[v];
1391
			oist->ist[v] = t->x86_tss.ist[v] =
1392
					(unsigned long)estacks;
1393 1394
			if (v == DEBUG_STACK-1)
				per_cpu(debug_stack_addr, cpu) = (unsigned long)estacks;
1395 1396 1397 1398
		}
	}

	t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap);
I
Ingo Molnar 已提交
1399

1400 1401 1402 1403 1404 1405 1406 1407 1408
	/*
	 * <= is required because the CPU will access up to
	 * 8 bits beyond the end of the IO permission bitmap.
	 */
	for (i = 0; i <= IO_BITMAP_LONGS; i++)
		t->io_bitmap[i] = ~0UL;

	atomic_inc(&init_mm.mm_count);
	me->active_mm = &init_mm;
S
Stoyan Gaydarov 已提交
1409
	BUG_ON(me->mm);
1410 1411 1412 1413 1414
	enter_lazy_tlb(&init_mm, me);

	load_sp0(t, &current->thread);
	set_tss_desc(cpu, t);
	load_TR_desc();
1415
	load_mm_ldt(&init_mm);
1416

1417 1418
	clear_all_debug_regs();
	dbg_restore_debug_regs();
1419

I
Ingo Molnar 已提交
1420
	fpu__init_cpu();
1421 1422 1423 1424 1425 1426 1427

	if (is_uv_system())
		uv_cpu_init();
}

#else

1428
void cpu_init(void)
1429
{
1430 1431
	int cpu = smp_processor_id();
	struct task_struct *curr = current;
1432
	struct tss_struct *t = &per_cpu(cpu_tss, cpu);
1433
	struct thread_struct *thread = &curr->thread;
1434

1435
	wait_for_master_cpu(cpu);
1436

1437 1438 1439 1440 1441 1442
	/*
	 * Initialize the CR4 shadow before doing anything that could
	 * try to read it.
	 */
	cr4_init_shadow();

1443
	show_ucode_info_early();
1444 1445 1446

	printk(KERN_INFO "Initializing CPU#%d\n", cpu);

D
Dave Hansen 已提交
1447
	if (cpu_feature_enabled(X86_FEATURE_VME) || cpu_has_tsc || cpu_has_de)
A
Andy Lutomirski 已提交
1448
		cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
1449

1450
	load_current_idt();
1451
	switch_to_new_gdt(cpu);
L
Linus Torvalds 已提交
1452 1453 1454 1455 1456

	/*
	 * Set up and load the per-CPU TSS and LDT
	 */
	atomic_inc(&init_mm.mm_count);
1457
	curr->active_mm = &init_mm;
S
Stoyan Gaydarov 已提交
1458
	BUG_ON(curr->mm);
1459
	enter_lazy_tlb(&init_mm, curr);
L
Linus Torvalds 已提交
1460

1461
	load_sp0(t, thread);
1462
	set_tss_desc(cpu, t);
L
Linus Torvalds 已提交
1463
	load_TR_desc();
1464
	load_mm_ldt(&init_mm);
L
Linus Torvalds 已提交
1465

1466 1467
	t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap);

1468
#ifdef CONFIG_DOUBLEFAULT
L
Linus Torvalds 已提交
1469 1470
	/* Set up doublefault TSS pointer in the GDT */
	__set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss);
1471
#endif
L
Linus Torvalds 已提交
1472

1473
	clear_all_debug_regs();
1474
	dbg_restore_debug_regs();
L
Linus Torvalds 已提交
1475

I
Ingo Molnar 已提交
1476
	fpu__init_cpu();
L
Linus Torvalds 已提交
1477
}
1478
#endif
1479 1480 1481 1482 1483 1484 1485 1486

#ifdef CONFIG_X86_DEBUG_STATIC_CPU_HAS
void warn_pre_alternatives(void)
{
	WARN(1, "You're using static_cpu_has before alternatives have run!\n");
}
EXPORT_SYMBOL_GPL(warn_pre_alternatives);
#endif
1487 1488 1489 1490 1491 1492

inline bool __static_cpu_has_safe(u16 bit)
{
	return boot_cpu_has(bit);
}
EXPORT_SYMBOL_GPL(__static_cpu_has_safe);
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509

static void bsp_resume(void)
{
	if (this_cpu->c_bsp_resume)
		this_cpu->c_bsp_resume(&boot_cpu_data);
}

static struct syscore_ops cpu_syscore_ops = {
	.resume		= bsp_resume,
};

static int __init init_cpu_syscore(void)
{
	register_syscore_ops(&cpu_syscore_ops);
	return 0;
}
core_initcall(init_cpu_syscore);