common.c 35.8 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>
L
Linus Torvalds 已提交
8
#include <linux/delay.h>
9 10
#include <linux/sched.h>
#include <linux/init.h>
11
#include <linux/kprobes.h>
12
#include <linux/kgdb.h>
L
Linus Torvalds 已提交
13
#include <linux/smp.h>
14 15 16
#include <linux/io.h>

#include <asm/stackprotector.h>
17
#include <asm/perf_event.h>
L
Linus Torvalds 已提交
18
#include <asm/mmu_context.h>
19
#include <asm/archrandom.h>
20 21
#include <asm/hypervisor.h>
#include <asm/processor.h>
22
#include <asm/tlbflush.h>
23
#include <asm/debugreg.h>
24
#include <asm/sections.h>
25
#include <asm/vsyscall.h>
A
Alan Cox 已提交
26 27
#include <linux/topology.h>
#include <linux/cpumask.h>
28
#include <asm/pgtable.h>
A
Arun Sharma 已提交
29
#include <linux/atomic.h>
30 31 32 33
#include <asm/proto.h>
#include <asm/setup.h>
#include <asm/apic.h>
#include <asm/desc.h>
34
#include <asm/fpu-internal.h>
35
#include <asm/mtrr.h>
A
Alan Cox 已提交
36
#include <linux/numa.h>
37 38
#include <asm/asm.h>
#include <asm/cpu.h>
39
#include <asm/mce.h>
40
#include <asm/msr.h>
41
#include <asm/pat.h>
42 43
#include <asm/microcode.h>
#include <asm/microcode_intel.h>
44 45

#ifdef CONFIG_X86_LOCAL_APIC
T
Tejun Heo 已提交
46
#include <asm/uv/uv.h>
L
Linus Torvalds 已提交
47 48 49 50
#endif

#include "cpu.h"

51 52
/* all of these masks are initialized in setup_cpu_local_masks() */
cpumask_var_t cpu_initialized_mask;
53 54
cpumask_var_t cpu_callout_mask;
cpumask_var_t cpu_callin_mask;
55 56 57 58

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

B
Brian Gerst 已提交
59
/* correctly size the local cpu masks */
60
void __init setup_cpu_local_masks(void)
B
Brian Gerst 已提交
61 62 63 64 65 66 67
{
	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);
}

68
static void default_init(struct cpuinfo_x86 *c)
69 70
{
#ifdef CONFIG_X86_64
71
	cpu_detect_cache_sizes(c);
72 73 74 75 76 77 78 79 80 81 82 83 84
#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
}

85
static const struct cpu_dev default_cpu = {
86 87 88 89 90
	.c_init		= default_init,
	.c_vendor	= "Unknown",
	.c_x86_vendor	= X86_VENDOR_UNKNOWN,
};

91
static const struct cpu_dev *this_cpu = &default_cpu;
92

B
Brian Gerst 已提交
93
DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = {
Y
Yinghai Lu 已提交
94
#ifdef CONFIG_X86_64
B
Brian Gerst 已提交
95 96 97 98 99
	/*
	 * 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
	 *
100
	 * TLS descriptors are currently at a different place compared to i386.
B
Brian Gerst 已提交
101 102
	 * Hopefully nobody expects them at a fixed place (Wine?)
	 */
A
Akinobu Mita 已提交
103 104 105 106 107 108
	[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
109
#else
A
Akinobu Mita 已提交
110 111 112 113
	[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),
114 115 116 117 118
	/*
	 * 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.
	 */
119
	/* 32-bit code */
A
Akinobu Mita 已提交
120
	[GDT_ENTRY_PNPBIOS_CS32]	= GDT_ENTRY_INIT(0x409a, 0, 0xffff),
121
	/* 16-bit code */
A
Akinobu Mita 已提交
122
	[GDT_ENTRY_PNPBIOS_CS16]	= GDT_ENTRY_INIT(0x009a, 0, 0xffff),
123
	/* 16-bit data */
A
Akinobu Mita 已提交
124
	[GDT_ENTRY_PNPBIOS_DS]		= GDT_ENTRY_INIT(0x0092, 0, 0xffff),
125
	/* 16-bit data */
A
Akinobu Mita 已提交
126
	[GDT_ENTRY_PNPBIOS_TS1]		= GDT_ENTRY_INIT(0x0092, 0, 0),
127
	/* 16-bit data */
A
Akinobu Mita 已提交
128
	[GDT_ENTRY_PNPBIOS_TS2]		= GDT_ENTRY_INIT(0x0092, 0, 0),
129 130 131 132
	/*
	 * The APM segments have byte granularity and their bases
	 * are set at run time.  All have 64k limits.
	 */
133
	/* 32-bit code */
A
Akinobu Mita 已提交
134
	[GDT_ENTRY_APMBIOS_BASE]	= GDT_ENTRY_INIT(0x409a, 0, 0xffff),
135
	/* 16-bit code */
A
Akinobu Mita 已提交
136
	[GDT_ENTRY_APMBIOS_BASE+1]	= GDT_ENTRY_INIT(0x009a, 0, 0xffff),
137
	/* data */
138
	[GDT_ENTRY_APMBIOS_BASE+2]	= GDT_ENTRY_INIT(0x4092, 0, 0xffff),
139

A
Akinobu Mita 已提交
140 141
	[GDT_ENTRY_ESPFIX_SS]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
	[GDT_ENTRY_PERCPU]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
142
	GDT_STACK_CANARY_INIT
Y
Yinghai Lu 已提交
143
#endif
B
Brian Gerst 已提交
144
} };
145
EXPORT_PER_CPU_SYMBOL_GPL(gdt_page);
146

147 148
static int __init x86_xsave_setup(char *s)
{
149 150
	if (strlen(s))
		return 0;
151
	setup_clear_cpu_cap(X86_FEATURE_XSAVE);
152
	setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT);
153
	setup_clear_cpu_cap(X86_FEATURE_XSAVES);
154 155
	setup_clear_cpu_cap(X86_FEATURE_AVX);
	setup_clear_cpu_cap(X86_FEATURE_AVX2);
156 157 158 159
	return 1;
}
__setup("noxsave", x86_xsave_setup);

160 161 162 163 164 165 166
static int __init x86_xsaveopt_setup(char *s)
{
	setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT);
	return 1;
}
__setup("noxsaveopt", x86_xsaveopt_setup);

167 168 169 170 171 172 173
static int __init x86_xsaves_setup(char *s)
{
	setup_clear_cpu_cap(X86_FEATURE_XSAVES);
	return 1;
}
__setup("noxsaves", x86_xsaves_setup);

174
#ifdef CONFIG_X86_32
175 176
static int cachesize_override = -1;
static int disable_x86_serial_nr = 1;
L
Linus Torvalds 已提交
177

178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
static int __init cachesize_setup(char *str)
{
	get_option(&str, &cachesize_override);
	return 1;
}
__setup("cachesize=", cachesize_setup);

static int __init x86_fxsr_setup(char *s)
{
	setup_clear_cpu_cap(X86_FEATURE_FXSR);
	setup_clear_cpu_cap(X86_FEATURE_XMM);
	return 1;
}
__setup("nofxsr", x86_fxsr_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;

205 206 207 208 209 210 211
	/*
	 * 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
212 213 214 215 216 217 218 219 220 221 222
	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"

223 224
		      : "=&r" (f1), "=&r" (f2)
		      : "ir" (flag));
225 226 227 228 229

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

/* Probe for the CPUID instruction */
230
int have_cpuid_p(void)
231 232 233 234
{
	return flag_is_changeable_p(X86_EFLAGS_ID);
}

235
static void squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
236
{
I
Ingo Molnar 已提交
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
	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);
253 254 255 256 257 258 259 260
}

static int __init x86_serial_nr_setup(char *s)
{
	disable_x86_serial_nr = 0;
	return 1;
}
__setup("serialnumber", x86_serial_nr_setup);
261
#else
262 263 264 265 266 267 268
static inline int flag_is_changeable_p(u32 flag)
{
	return 1;
}
static inline void squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
{
}
269
#endif
270

271 272
static __init int setup_disable_smep(char *arg)
{
273
	setup_clear_cpu_cap(X86_FEATURE_SMEP);
274 275 276 277
	return 1;
}
__setup("nosmep", setup_disable_smep);

278
static __always_inline void setup_smep(struct cpuinfo_x86 *c)
279
{
280
	if (cpu_has(c, X86_FEATURE_SMEP))
A
Andy Lutomirski 已提交
281
		cr4_set_bits(X86_CR4_SMEP);
282 283
}

284 285
static __init int setup_disable_smap(char *arg)
{
286
	setup_clear_cpu_cap(X86_FEATURE_SMAP);
287 288 289 290
	return 1;
}
__setup("nosmap", setup_disable_smap);

291 292 293 294 295 296 297 298
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);

299 300
	if (cpu_has(c, X86_FEATURE_SMAP)) {
#ifdef CONFIG_X86_SMAP
A
Andy Lutomirski 已提交
301
		cr4_set_bits(X86_CR4_SMAP);
302
#else
A
Andy Lutomirski 已提交
303
		cr4_clear_bits(X86_CR4_SMAP);
304 305
#endif
	}
306 307
}

308 309 310 311 312 313 314 315 316
/*
 * 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
317

318
static const struct cpuid_dependent_feature
319 320 321 322 323 324 325
cpuid_dependent_features[] = {
	{ X86_FEATURE_MWAIT,		0x00000005 },
	{ X86_FEATURE_DCA,		0x00000009 },
	{ X86_FEATURE_XSAVE,		0x0000000d },
	{ 0, 0 }
};

326
static void filter_cpuid_features(struct cpuinfo_x86 *c, bool warn)
327 328
{
	const struct cpuid_dependent_feature *df;
329

330
	for (df = cpuid_dependent_features; df->feature; df++) {
I
Ingo Molnar 已提交
331 332 333

		if (!cpu_has(c, df->feature))
			continue;
334 335 336 337 338 339 340
		/*
		 * 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
341
		if (!((s32)df->level < 0 ?
342
		     (u32)df->level > (u32)c->extended_cpuid_level :
I
Ingo Molnar 已提交
343 344 345 346 347 348 349 350
		     (s32)df->level > (s32)c->cpuid_level))
			continue;

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

		printk(KERN_WARNING
351 352
		       "CPU: CPU feature " X86_CAP_FMT " disabled, no CPUID level 0x%x\n",
				x86_cap_flag(df->feature), df->level);
353
	}
354
}
355

356 357 358
/*
 * 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
359 360
 * in particular, if CPUID levels 0x80000002..4 are supported, this
 * isn't used
361 362 363
 */

/* Look up CPU names by table lookup. */
364
static const char *table_lookup_model(struct cpuinfo_x86 *c)
365
{
366 367
#ifdef CONFIG_X86_32
	const struct legacy_cpu_model_info *info;
368 369 370 371 372 373 374

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

	if (!this_cpu)
		return NULL;

375
	info = this_cpu->legacy_models;
376

377
	while (info->family) {
378 379 380 381
		if (info->family == c->x86)
			return info->model_names[c->x86_model];
		info++;
	}
382
#endif
383 384 385
	return NULL;		/* Not found */
}

386 387
__u32 cpu_caps_cleared[NCAPINTS];
__u32 cpu_caps_set[NCAPINTS];
388

389 390 391 392 393 394 395 396
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
397
	load_stack_canary_segment();
398 399
}

I
Ingo Molnar 已提交
400 401 402 403
/*
 * Current gdt points %fs at the "master" per-cpu area: after this,
 * it's on the real one.
 */
404
void switch_to_new_gdt(int cpu)
405 406 407
{
	struct desc_ptr gdt_descr;

408
	gdt_descr.address = (long)get_cpu_gdt_table(cpu);
409 410
	gdt_descr.size = GDT_SIZE - 1;
	load_gdt(&gdt_descr);
411
	/* Reload the per-cpu base */
412 413

	load_percpu_segment(cpu);
414 415
}

416
static const struct cpu_dev *cpu_devs[X86_VENDOR_NUM] = {};
L
Linus Torvalds 已提交
417

418
static void get_model_name(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
419 420 421 422
{
	unsigned int *v;
	char *p, *q;

423
	if (c->extended_cpuid_level < 0x80000004)
424
		return;
L
Linus Torvalds 已提交
425

I
Ingo Molnar 已提交
426
	v = (unsigned int *)c->x86_model_id;
L
Linus Torvalds 已提交
427 428 429 430 431
	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;

I
Ingo Molnar 已提交
432 433 434 435
	/*
	 * Intel chips right-justify this string for some dumb reason;
	 * undo that brain damage:
	 */
L
Linus Torvalds 已提交
436
	p = q = &c->x86_model_id[0];
437
	while (*p == ' ')
438
		p++;
439
	if (p != q) {
440 441 442 443
		while (*p)
			*q++ = *p++;
		while (q <= &c->x86_model_id[48])
			*q++ = '\0';	/* Zero-pad the rest */
L
Linus Torvalds 已提交
444 445 446
	}
}

447
void cpu_detect_cache_sizes(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
448
{
449
	unsigned int n, dummy, ebx, ecx, edx, l2size;
L
Linus Torvalds 已提交
450

451
	n = c->extended_cpuid_level;
L
Linus Torvalds 已提交
452 453

	if (n >= 0x80000005) {
454 455
		cpuid(0x80000005, &dummy, &ebx, &ecx, &edx);
		c->x86_cache_size = (ecx>>24) + (edx>>24);
456 457 458 459
#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
460 461 462 463 464
	}

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

465
	cpuid(0x80000006, &dummy, &ebx, &ecx, &edx);
L
Linus Torvalds 已提交
466
	l2size = ecx >> 16;
467

468 469 470
#ifdef CONFIG_X86_64
	c->x86_tlbsize += ((ebx >> 16) & 0xfff) + (ebx & 0xfff);
#else
L
Linus Torvalds 已提交
471
	/* do processor-specific cache resizing */
472 473
	if (this_cpu->legacy_cache_size)
		l2size = this_cpu->legacy_cache_size(c, l2size);
L
Linus Torvalds 已提交
474 475 476 477 478

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

479
	if (l2size == 0)
L
Linus Torvalds 已提交
480
		return;		/* Again, no L2 cache is possible */
481
#endif
L
Linus Torvalds 已提交
482 483 484 485

	c->x86_cache_size = l2size;
}

486 487 488 489 490 491
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];
492
u16 __read_mostly tlb_lld_1g[NR_INFO];
493

494
static void cpu_detect_tlb(struct cpuinfo_x86 *c)
495 496 497 498
{
	if (this_cpu->c_detect_tlb)
		this_cpu->c_detect_tlb(c);

499
	pr_info("Last level iTLB entries: 4KB %d, 2MB %d, 4MB %d\n",
500
		tlb_lli_4k[ENTRIES], tlb_lli_2m[ENTRIES],
501 502 503 504 505
		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]);
506 507
}

508
void detect_ht(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
509
{
510
#ifdef CONFIG_X86_HT
511 512
	u32 eax, ebx, ecx, edx;
	int index_msb, core_bits;
513
	static bool printed;
L
Linus Torvalds 已提交
514

515
	if (!cpu_has(c, X86_FEATURE_HT))
516
		return;
L
Linus Torvalds 已提交
517

518 519
	if (cpu_has(c, X86_FEATURE_CMP_LEGACY))
		goto out;
L
Linus Torvalds 已提交
520

521 522
	if (cpu_has(c, X86_FEATURE_XTOPOLOGY))
		return;
L
Linus Torvalds 已提交
523

524
	cpuid(1, &eax, &ebx, &ecx, &edx);
L
Linus Torvalds 已提交
525

526 527 528
	smp_num_siblings = (ebx & 0xff0000) >> 16;

	if (smp_num_siblings == 1) {
529
		printk_once(KERN_INFO "CPU0: Hyper-Threading is disabled\n");
I
Ingo Molnar 已提交
530 531
		goto out;
	}
532

I
Ingo Molnar 已提交
533 534
	if (smp_num_siblings <= 1)
		goto out;
535

I
Ingo Molnar 已提交
536 537
	index_msb = get_count_order(smp_num_siblings);
	c->phys_proc_id = apic->phys_pkg_id(c->initial_apicid, index_msb);
538

I
Ingo Molnar 已提交
539
	smp_num_siblings = smp_num_siblings / c->x86_max_cores;
540

I
Ingo Molnar 已提交
541
	index_msb = get_count_order(smp_num_siblings);
542

I
Ingo Molnar 已提交
543
	core_bits = get_count_order(c->x86_max_cores);
544

I
Ingo Molnar 已提交
545 546
	c->cpu_core_id = apic->phys_pkg_id(c->initial_apicid, index_msb) &
				       ((1 << core_bits) - 1);
L
Linus Torvalds 已提交
547

548
out:
549
	if (!printed && (c->x86_max_cores * smp_num_siblings) > 1) {
550 551 552 553
		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);
554
		printed = 1;
555 556
	}
#endif
557
}
L
Linus Torvalds 已提交
558

559
static void get_cpu_vendor(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
560 561
{
	char *v = c->x86_vendor_id;
I
Ingo Molnar 已提交
562
	int i;
L
Linus Torvalds 已提交
563 564

	for (i = 0; i < X86_VENDOR_NUM; i++) {
Y
Yinghai Lu 已提交
565 566 567 568 569 570
		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
571

Y
Yinghai Lu 已提交
572 573 574
			this_cpu = cpu_devs[i];
			c->x86_vendor = this_cpu->c_x86_vendor;
			return;
L
Linus Torvalds 已提交
575 576
		}
	}
Y
Yinghai Lu 已提交
577

578 579 580
	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
581

582 583
	c->x86_vendor = X86_VENDOR_UNKNOWN;
	this_cpu = &default_cpu;
L
Linus Torvalds 已提交
584 585
}

586
void cpu_detect(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
587 588
{
	/* Get vendor name */
589 590 591 592
	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
593 594

	c->x86 = 4;
595
	/* Intel-defined flags: level 0x00000001 */
L
Linus Torvalds 已提交
596 597
	if (c->cpuid_level >= 0x00000001) {
		u32 junk, tfms, cap0, misc;
I
Ingo Molnar 已提交
598

L
Linus Torvalds 已提交
599
		cpuid(0x00000001, &tfms, &misc, &junk, &cap0);
600 601 602
		c->x86 = (tfms >> 8) & 0xf;
		c->x86_model = (tfms >> 4) & 0xf;
		c->x86_mask = tfms & 0xf;
I
Ingo Molnar 已提交
603

604
		if (c->x86 == 0xf)
L
Linus Torvalds 已提交
605
			c->x86 += (tfms >> 20) & 0xff;
606
		if (c->x86 >= 0x6)
607
			c->x86_model += ((tfms >> 16) & 0xf) << 4;
I
Ingo Molnar 已提交
608

H
Huang, Ying 已提交
609 610
		if (cap0 & (1<<19)) {
			c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
611
			c->x86_cache_alignment = c->x86_clflush_size;
H
Huang, Ying 已提交
612
		}
L
Linus Torvalds 已提交
613 614
	}
}
615

616
void get_cpu_cap(struct cpuinfo_x86 *c)
617 618
{
	u32 tfms, xlvl;
619
	u32 ebx;
620

621 622 623
	/* Intel-defined flags: level 0x00000001 */
	if (c->cpuid_level >= 0x00000001) {
		u32 capability, excap;
I
Ingo Molnar 已提交
624

625 626 627 628
		cpuid(0x00000001, &tfms, &ebx, &excap, &capability);
		c->x86_capability[0] = capability;
		c->x86_capability[4] = excap;
	}
629

630 631 632 633 634 635
	/* Additional Intel-defined flags: level 0x00000007 */
	if (c->cpuid_level >= 0x00000007) {
		u32 eax, ebx, ecx, edx;

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

636
		c->x86_capability[9] = ebx;
637 638
	}

639 640 641 642 643 644 645 646 647
	/* 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;
	}

648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
	/* 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;
		}
	}

672 673 674
	/* AMD-defined flags: level 0x80000001 */
	xlvl = cpuid_eax(0x80000000);
	c->extended_cpuid_level = xlvl;
I
Ingo Molnar 已提交
675

676 677 678 679
	if ((xlvl & 0xffff0000) == 0x80000000) {
		if (xlvl >= 0x80000001) {
			c->x86_capability[1] = cpuid_edx(0x80000001);
			c->x86_capability[6] = cpuid_ecx(0x80000001);
680 681 682
		}
	}

683 684 685 686 687
	if (c->extended_cpuid_level >= 0x80000008) {
		u32 eax = cpuid_eax(0x80000008);

		c->x86_virt_bits = (eax >> 8) & 0xff;
		c->x86_phys_bits = eax & 0xff;
688
	}
689 690 691
#ifdef CONFIG_X86_32
	else if (cpu_has(c, X86_FEATURE_PAE) || cpu_has(c, X86_FEATURE_PSE36))
		c->x86_phys_bits = 36;
692
#endif
693 694 695

	if (c->extended_cpuid_level >= 0x80000007)
		c->x86_power = cpuid_edx(0x80000007);
696

697
	init_scattered_cpuid_features(c);
698
}
L
Linus Torvalds 已提交
699

700
static void identify_cpu_without_cpuid(struct cpuinfo_x86 *c)
Y
Yinghai Lu 已提交
701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
{
#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
}

726 727 728 729 730 731 732 733 734
/*
 * 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.
 */
735
static void __init early_identify_cpu(struct cpuinfo_x86 *c)
736
{
737 738
#ifdef CONFIG_X86_64
	c->x86_clflush_size = 64;
739 740
	c->x86_phys_bits = 36;
	c->x86_virt_bits = 48;
741
#else
H
Huang, Ying 已提交
742
	c->x86_clflush_size = 32;
743 744
	c->x86_phys_bits = 32;
	c->x86_virt_bits = 32;
745
#endif
746
	c->x86_cache_alignment = c->x86_clflush_size;
747

748
	memset(&c->x86_capability, 0, sizeof c->x86_capability);
749
	c->extended_cpuid_level = 0;
750

Y
Yinghai Lu 已提交
751 752 753 754
	if (!have_cpuid_p())
		identify_cpu_without_cpuid(c);

	/* cyrix could have cpuid enabled via c_identify()*/
755 756 757 758
	if (!have_cpuid_p())
		return;

	cpu_detect(c);
759 760
	get_cpu_vendor(c);
	get_cpu_cap(c);
761
	fpu__detect(c);
762

Y
Yinghai Lu 已提交
763 764
	if (this_cpu->c_early_init)
		this_cpu->c_early_init(c);
765

766
	c->cpu_index = 0;
767
	filter_cpuid_features(c, false);
768

B
Borislav Petkov 已提交
769 770
	if (this_cpu->c_bsp_init)
		this_cpu->c_bsp_init(c);
771 772

	setup_force_cpu_cap(X86_FEATURE_ALWAYS);
773 774
}

775 776
void __init early_cpu_init(void)
{
777
	const struct cpu_dev *const *cdev;
Y
Yinghai Lu 已提交
778 779
	int count = 0;

780
#ifdef CONFIG_PROCESSOR_SELECT
781
	printk(KERN_INFO "KERNEL supported cpus:\n");
782 783
#endif

Y
Yinghai Lu 已提交
784
	for (cdev = __x86_cpu_dev_start; cdev < __x86_cpu_dev_end; cdev++) {
785
		const struct cpu_dev *cpudev = *cdev;
786

Y
Yinghai Lu 已提交
787 788 789 790 791
		if (count >= X86_VENDOR_NUM)
			break;
		cpu_devs[count] = cpudev;
		count++;

792
#ifdef CONFIG_PROCESSOR_SELECT
793 794 795 796 797 798 799 800 801
		{
			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
802
		}
803
#endif
Y
Yinghai Lu 已提交
804
	}
805
	early_identify_cpu(&boot_cpu_data);
806
}
807

808
/*
B
Borislav Petkov 已提交
809 810 811 812 813
 * 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
814
 * unless we can find a reliable way to detect all the broken cases.
B
Borislav Petkov 已提交
815
 * Enable it explicitly on 64-bit for non-constant inputs of cpu_has().
816
 */
817
static void detect_nopl(struct cpuinfo_x86 *c)
818
{
B
Borislav Petkov 已提交
819
#ifdef CONFIG_X86_32
820
	clear_cpu_cap(c, X86_FEATURE_NOPL);
B
Borislav Petkov 已提交
821 822 823
#else
	set_cpu_cap(c, X86_FEATURE_NOPL);
#endif
824 825
}

826
static void generic_identify(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
827
{
Y
Yinghai Lu 已提交
828
	c->extended_cpuid_level = 0;
L
Linus Torvalds 已提交
829

830
	if (!have_cpuid_p())
Y
Yinghai Lu 已提交
831
		identify_cpu_without_cpuid(c);
832

Y
Yinghai Lu 已提交
833
	/* cyrix could have cpuid enabled via c_identify()*/
I
Ingo Molnar 已提交
834
	if (!have_cpuid_p())
Y
Yinghai Lu 已提交
835
		return;
L
Linus Torvalds 已提交
836

837
	cpu_detect(c);
L
Linus Torvalds 已提交
838

839
	get_cpu_vendor(c);
L
Linus Torvalds 已提交
840

841
	get_cpu_cap(c);
L
Linus Torvalds 已提交
842

843 844
	if (c->cpuid_level >= 0x00000001) {
		c->initial_apicid = (cpuid_ebx(1) >> 24) & 0xFF;
845 846
#ifdef CONFIG_X86_32
# ifdef CONFIG_X86_HT
847
		c->apicid = apic->phys_pkg_id(c->initial_apicid, 0);
848
# else
849
		c->apicid = c->initial_apicid;
850 851 852
# endif
#endif
		c->phys_proc_id = c->initial_apicid;
853
	}
L
Linus Torvalds 已提交
854

855
	get_model_name(c); /* Default name */
L
Linus Torvalds 已提交
856

857
	detect_nopl(c);
L
Linus Torvalds 已提交
858 859
}

860 861 862 863 864 865 866 867 868 869 870 871 872 873
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
874 875 876
/*
 * This does the hard work of actually picking apart the CPU stuff...
 */
877
static void identify_cpu(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
878 879 880 881 882 883 884 885 886
{
	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 */
887
	c->x86_max_cores = 1;
888
	c->x86_coreid_bits = 0;
889
#ifdef CONFIG_X86_64
890
	c->x86_clflush_size = 64;
891 892
	c->x86_phys_bits = 36;
	c->x86_virt_bits = 48;
893 894
#else
	c->cpuid_level = -1;	/* CPUID not detected */
895
	c->x86_clflush_size = 32;
896 897
	c->x86_phys_bits = 32;
	c->x86_virt_bits = 32;
898 899
#endif
	c->x86_cache_alignment = c->x86_clflush_size;
L
Linus Torvalds 已提交
900 901 902 903
	memset(&c->x86_capability, 0, sizeof c->x86_capability);

	generic_identify(c);

904
	if (this_cpu->c_identify)
L
Linus Torvalds 已提交
905 906
		this_cpu->c_identify(c);

907 908 909 910 911 912
	/* 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];
	}

913
#ifdef CONFIG_X86_64
914
	c->apicid = apic->phys_pkg_id(c->initial_apicid, 0);
915 916
#endif

L
Linus Torvalds 已提交
917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932
	/*
	 * 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);

933 934 935 936
	/* Set up SMEP/SMAP */
	setup_smep(c);
	setup_smap(c);

L
Linus Torvalds 已提交
937
	/*
I
Ingo Molnar 已提交
938 939
	 * The vendor-specific functions might have changed features.
	 * Now we do "generic changes."
L
Linus Torvalds 已提交
940 941
	 */

942 943 944
	/* Filter out anything that depends on CPUID levels we don't have */
	filter_cpuid_features(c, true);

L
Linus Torvalds 已提交
945
	/* If the model name is still unset, do table lookup. */
946
	if (!c->x86_model_id[0]) {
947
		const char *p;
L
Linus Torvalds 已提交
948
		p = table_lookup_model(c);
949
		if (p)
L
Linus Torvalds 已提交
950 951 952 953
			strcpy(c->x86_model_id, p);
		else
			/* Last resort... */
			sprintf(c->x86_model_id, "%02x/%02x",
954
				c->x86, c->x86_model);
L
Linus Torvalds 已提交
955 956
	}

957 958 959 960
#ifdef CONFIG_X86_64
	detect_ht(c);
#endif

961
	init_hypervisor(c);
962
	x86_init_rdrand(c);
963
	x86_init_cache_qos(c);
964 965 966 967 968 969 970 971 972 973

	/*
	 * 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
974 975 976 977 978 979
	/*
	 * 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.
	 */
980
	if (c != &boot_cpu_data) {
L
Linus Torvalds 已提交
981
		/* AND the already accumulated flags with these */
982
		for (i = 0; i < NCAPINTS; i++)
L
Linus Torvalds 已提交
983
			boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
984 985 986 987

		/* 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
988 989 990
	}

	/* Init Machine Check Exception if available. */
991
	mcheck_cpu_init(c);
992 993

	select_idle_routine(c);
994

995
#ifdef CONFIG_NUMA
996 997
	numa_add_cpu(smp_processor_id());
#endif
998
}
S
Shaohua Li 已提交
999

1000 1001 1002 1003
/*
 * Set up the CPU state needed to execute SYSENTER/SYSEXIT instructions
 * on 32-bit kernels:
 */
1004 1005 1006
#ifdef CONFIG_X86_32
void enable_sep_cpu(void)
{
1007 1008
	struct tss_struct *tss;
	int cpu;
1009

1010 1011 1012 1013 1014 1015 1016
	cpu = get_cpu();
	tss = &per_cpu(cpu_tss, cpu);

	if (!boot_cpu_has(X86_FEATURE_SEP))
		goto out;

	/*
1017 1018
	 * 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.
1019
	 */
1020 1021

	tss->x86_tss.ss1 = __KERNEL_CS;
1022 1023
	wrmsr(MSR_IA32_SYSENTER_CS, tss->x86_tss.ss1, 0);

1024 1025 1026
	wrmsr(MSR_IA32_SYSENTER_ESP,
	      (unsigned long)tss + offsetofend(struct tss_struct, SYSENTER_stack),
	      0);
1027 1028 1029 1030

	wrmsr(MSR_IA32_SYSENTER_EIP, (unsigned long)ia32_sysenter_target, 0);

out:
1031 1032
	put_cpu();
}
1033 1034
#endif

1035 1036 1037
void __init identify_boot_cpu(void)
{
	identify_cpu(&boot_cpu_data);
1038
	init_amd_e400_c1e_mask();
1039
#ifdef CONFIG_X86_32
1040
	sysenter_setup();
L
Li Shaohua 已提交
1041
	enable_sep_cpu();
1042
#endif
1043
	cpu_detect_tlb(&boot_cpu_data);
1044
}
S
Shaohua Li 已提交
1045

1046
void identify_secondary_cpu(struct cpuinfo_x86 *c)
1047 1048 1049
{
	BUG_ON(c == &boot_cpu_data);
	identify_cpu(c);
1050
#ifdef CONFIG_X86_32
1051
	enable_sep_cpu();
1052
#endif
1053
	mtrr_ap_init();
L
Linus Torvalds 已提交
1054 1055
}

1056
struct msr_range {
I
Ingo Molnar 已提交
1057 1058
	unsigned	min;
	unsigned	max;
1059
};
L
Linus Torvalds 已提交
1060

1061
static const struct msr_range msr_range_array[] = {
1062 1063 1064 1065 1066
	{ 0x00000000, 0x00000418},
	{ 0xc0000000, 0xc000040b},
	{ 0xc0010000, 0xc0010142},
	{ 0xc0011000, 0xc001103b},
};
L
Linus Torvalds 已提交
1067

1068
static void __print_cpu_msr(void)
1069
{
I
Ingo Molnar 已提交
1070
	unsigned index_min, index_max;
1071 1072 1073 1074 1075 1076 1077
	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
1078

1079
		for (index = index_min; index < index_max; index++) {
1080
			if (rdmsrl_safe(index, &val))
1081 1082
				continue;
			printk(KERN_INFO " MSR%08x: %016llx\n", index, val);
L
Linus Torvalds 已提交
1083
		}
1084 1085
	}
}
1086

1087
static int show_msr;
I
Ingo Molnar 已提交
1088

1089 1090 1091
static __init int setup_show_msr(char *arg)
{
	int num;
1092

1093
	get_option(&arg, &num);
1094

1095 1096 1097
	if (num > 0)
		show_msr = num;
	return 1;
L
Linus Torvalds 已提交
1098
}
1099
__setup("show_msr=", setup_show_msr);
L
Linus Torvalds 已提交
1100

A
Andi Kleen 已提交
1101 1102
static __init int setup_noclflush(char *arg)
{
1103
	setup_clear_cpu_cap(X86_FEATURE_CLFLUSH);
1104
	setup_clear_cpu_cap(X86_FEATURE_CLFLUSHOPT);
A
Andi Kleen 已提交
1105 1106 1107 1108
	return 1;
}
__setup("noclflush", setup_noclflush);

1109
void print_cpu_info(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
1110
{
1111
	const char *vendor = NULL;
L
Linus Torvalds 已提交
1112

I
Ingo Molnar 已提交
1113
	if (c->x86_vendor < X86_VENDOR_NUM) {
L
Linus Torvalds 已提交
1114
		vendor = this_cpu->c_vendor;
I
Ingo Molnar 已提交
1115 1116 1117 1118
	} else {
		if (c->cpuid_level >= 0)
			vendor = c->x86_vendor_id;
	}
L
Linus Torvalds 已提交
1119

1120
	if (vendor && !strstr(c->x86_model_id, vendor))
1121
		printk(KERN_CONT "%s ", vendor);
L
Linus Torvalds 已提交
1122

1123
	if (c->x86_model_id[0])
1124
		printk(KERN_CONT "%s", strim(c->x86_model_id));
L
Linus Torvalds 已提交
1125
	else
1126
		printk(KERN_CONT "%d86", c->x86);
L
Linus Torvalds 已提交
1127

1128 1129
	printk(KERN_CONT " (fam: %02x, model: %02x", c->x86, c->x86_model);

1130
	if (c->x86_mask || c->cpuid_level >= 0)
1131
		printk(KERN_CONT ", stepping: %02x)\n", c->x86_mask);
L
Linus Torvalds 已提交
1132
	else
1133
		printk(KERN_CONT ")\n");
1134

1135
	print_cpu_msr(c);
1136 1137
}

1138
void print_cpu_msr(struct cpuinfo_x86 *c)
1139
{
1140
	if (c->cpu_index < show_msr)
1141
		__print_cpu_msr();
L
Linus Torvalds 已提交
1142 1143
}

1144 1145 1146
static __init int setup_disablecpuid(char *arg)
{
	int bit;
I
Ingo Molnar 已提交
1147

1148 1149 1150 1151
	if (get_option(&arg, &bit) && bit < NCAPINTS*32)
		setup_clear_cpu_cap(bit);
	else
		return 0;
I
Ingo Molnar 已提交
1152

1153 1154 1155 1156
	return 1;
}
__setup("clearcpuid=", setup_disablecpuid);

1157
DEFINE_PER_CPU(unsigned long, kernel_stack) =
1158
	(unsigned long)&init_thread_union + THREAD_SIZE;
1159 1160
EXPORT_PER_CPU_SYMBOL(kernel_stack);

1161
#ifdef CONFIG_X86_64
1162
struct desc_ptr idt_descr = { NR_VECTORS * 16 - 1, (unsigned long) idt_table };
1163 1164
struct desc_ptr debug_idt_descr = { NR_VECTORS * 16 - 1,
				    (unsigned long) debug_idt_table };
1165

1166
DEFINE_PER_CPU_FIRST(union irq_stack_union,
1167
		     irq_stack_union) __aligned(PAGE_SIZE) __visible;
I
Ingo Molnar 已提交
1168

1169
/*
1170 1171
 * The following percpu variables are hot.  Align current_task to
 * cacheline size such that they fall in the same cacheline.
1172 1173 1174 1175
 */
DEFINE_PER_CPU(struct task_struct *, current_task) ____cacheline_aligned =
	&init_task;
EXPORT_PER_CPU_SYMBOL(current_task);
1176

1177 1178 1179
DEFINE_PER_CPU(char *, irq_stack_ptr) =
	init_per_cpu_var(irq_stack_union.irq_stack) + IRQ_STACK_SIZE - 64;

1180
DEFINE_PER_CPU(unsigned int, irq_count) __visible = -1;
1181

1182 1183 1184
DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
EXPORT_PER_CPU_SYMBOL(__preempt_count);

I
Ingo Molnar 已提交
1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195
/*
 * 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
};

1196
static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks
1197
	[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + DEBUG_STKSZ]);
1198 1199 1200

/* May not be marked __init: used by software suspend */
void syscall_init(void)
L
Linus Torvalds 已提交
1201
{
1202 1203 1204 1205 1206 1207 1208
	/*
	 * 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);
	wrmsrl(MSR_LSTAR, system_call);
1209 1210

#ifdef CONFIG_IA32_EMULATION
1211 1212
	wrmsrl(MSR_CSTAR, ia32_cstar_target);
	/*
1213 1214 1215 1216
	 * 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).
1217 1218 1219 1220
	 */
	wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS);
	wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL);
	wrmsrl_safe(MSR_IA32_SYSENTER_EIP, (u64)ia32_sysenter_target);
1221 1222
#else
	wrmsrl(MSR_CSTAR, ignore_sysret);
1223
	wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)GDT_ENTRY_INVALID_SEG);
1224 1225
	wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL);
	wrmsrl_safe(MSR_IA32_SYSENTER_EIP, 0ULL);
1226
#endif
1227

1228 1229
	/* Flags to clear on syscall */
	wrmsrl(MSR_SYSCALL_MASK,
1230
	       X86_EFLAGS_TF|X86_EFLAGS_DF|X86_EFLAGS_IF|
1231
	       X86_EFLAGS_IOPL|X86_EFLAGS_AC|X86_EFLAGS_NT);
L
Linus Torvalds 已提交
1232
}
1233

1234 1235 1236 1237 1238 1239
/*
 * 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);

1240
static DEFINE_PER_CPU(unsigned long, debug_stack_addr);
1241
DEFINE_PER_CPU(int, debug_stack_usage);
1242 1243 1244

int is_debug_stack(unsigned long addr)
{
1245 1246 1247
	return __this_cpu_read(debug_stack_usage) ||
		(addr <= __this_cpu_read(debug_stack_addr) &&
		 addr > (__this_cpu_read(debug_stack_addr) - DEBUG_STKSZ));
1248
}
1249
NOKPROBE_SYMBOL(is_debug_stack);
1250

1251
DEFINE_PER_CPU(u32, debug_idt_ctr);
1252

1253 1254
void debug_stack_set_zero(void)
{
1255 1256
	this_cpu_inc(debug_idt_ctr);
	load_current_idt();
1257
}
1258
NOKPROBE_SYMBOL(debug_stack_set_zero);
1259 1260 1261

void debug_stack_reset(void)
{
1262
	if (WARN_ON(!this_cpu_read(debug_idt_ctr)))
1263
		return;
1264 1265
	if (this_cpu_dec_return(debug_idt_ctr) == 0)
		load_current_idt();
1266
}
1267
NOKPROBE_SYMBOL(debug_stack_reset);
1268

I
Ingo Molnar 已提交
1269
#else	/* CONFIG_X86_64 */
1270

1271 1272
DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
EXPORT_PER_CPU_SYMBOL(current_task);
1273 1274
DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
EXPORT_PER_CPU_SYMBOL(__preempt_count);
1275

1276 1277 1278 1279 1280 1281 1282 1283 1284
/*
 * 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);

1285
#ifdef CONFIG_CC_STACKPROTECTOR
1286
DEFINE_PER_CPU_ALIGNED(struct stack_canary, stack_canary);
1287
#endif
1288

I
Ingo Molnar 已提交
1289
#endif	/* CONFIG_X86_64 */
1290

1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
/*
 * 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);
	}
}
1306

1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
#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 */

1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
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
}

1334 1335 1336 1337 1338
/*
 * 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.
1339
 * A lot of state is already set up in PDA init for 64 bit
1340
 */
1341
#ifdef CONFIG_X86_64
I
Ingo Molnar 已提交
1342

1343
void cpu_init(void)
1344
{
1345
	struct orig_ist *oist;
1346
	struct task_struct *me;
I
Ingo Molnar 已提交
1347 1348
	struct tss_struct *t;
	unsigned long v;
1349
	int cpu = stack_smp_processor_id();
1350 1351
	int i;

1352 1353
	wait_for_master_cpu(cpu);

1354 1355 1356 1357 1358 1359
	/*
	 * Initialize the CR4 shadow before doing anything that could
	 * try to read it.
	 */
	cr4_init_shadow();

1360 1361 1362 1363 1364 1365
	/*
	 * Load microcode on this cpu if a valid microcode is available.
	 * This is early microcode loading procedure.
	 */
	load_ucode_ap();

1366
	t = &per_cpu(cpu_tss, cpu);
1367
	oist = &per_cpu(orig_ist, cpu);
I
Ingo Molnar 已提交
1368

1369
#ifdef CONFIG_NUMA
1370
	if (this_cpu_read(numa_node) == 0 &&
1371 1372
	    early_cpu_to_node(cpu) != NUMA_NO_NODE)
		set_numa_node(early_cpu_to_node(cpu));
1373
#endif
1374 1375 1376

	me = current;

1377
	pr_debug("Initializing CPU#%d\n", cpu);
1378

A
Andy Lutomirski 已提交
1379
	cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
1380 1381 1382 1383 1384 1385

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

1386
	switch_to_new_gdt(cpu);
1387 1388
	loadsegment(fs, 0);

1389
	load_current_idt();
1390 1391 1392 1393 1394 1395 1396 1397

	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();

1398
	x86_configure_nx();
1399
	x2apic_setup();
1400 1401 1402 1403

	/*
	 * set up and load the per-CPU TSS
	 */
1404
	if (!oist->ist[0]) {
1405
		char *estacks = per_cpu(exception_stacks, cpu);
I
Ingo Molnar 已提交
1406

1407
		for (v = 0; v < N_EXCEPTION_STACKS; v++) {
I
Ingo Molnar 已提交
1408
			estacks += exception_stack_sizes[v];
1409
			oist->ist[v] = t->x86_tss.ist[v] =
1410
					(unsigned long)estacks;
1411 1412
			if (v == DEBUG_STACK-1)
				per_cpu(debug_stack_addr, cpu) = (unsigned long)estacks;
1413 1414 1415 1416
		}
	}

	t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap);
I
Ingo Molnar 已提交
1417

1418 1419 1420 1421 1422 1423 1424 1425 1426
	/*
	 * <= 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
1427
	BUG_ON(me->mm);
1428 1429 1430 1431 1432 1433 1434
	enter_lazy_tlb(&init_mm, me);

	load_sp0(t, &current->thread);
	set_tss_desc(cpu, t);
	load_TR_desc();
	load_LDT(&init_mm.context);

1435 1436
	clear_all_debug_regs();
	dbg_restore_debug_regs();
1437

1438
	fpu__cpu_init();
1439 1440 1441 1442 1443 1444 1445

	if (is_uv_system())
		uv_cpu_init();
}

#else

1446
void cpu_init(void)
1447
{
1448 1449
	int cpu = smp_processor_id();
	struct task_struct *curr = current;
1450
	struct tss_struct *t = &per_cpu(cpu_tss, cpu);
1451
	struct thread_struct *thread = &curr->thread;
1452

1453
	wait_for_master_cpu(cpu);
1454

1455 1456 1457 1458 1459 1460
	/*
	 * Initialize the CR4 shadow before doing anything that could
	 * try to read it.
	 */
	cr4_init_shadow();

1461
	show_ucode_info_early();
1462 1463 1464

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

D
Dave Hansen 已提交
1465
	if (cpu_feature_enabled(X86_FEATURE_VME) || cpu_has_tsc || cpu_has_de)
A
Andy Lutomirski 已提交
1466
		cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
1467

1468
	load_current_idt();
1469
	switch_to_new_gdt(cpu);
L
Linus Torvalds 已提交
1470 1471 1472 1473 1474

	/*
	 * Set up and load the per-CPU TSS and LDT
	 */
	atomic_inc(&init_mm.mm_count);
1475
	curr->active_mm = &init_mm;
S
Stoyan Gaydarov 已提交
1476
	BUG_ON(curr->mm);
1477
	enter_lazy_tlb(&init_mm, curr);
L
Linus Torvalds 已提交
1478

1479
	load_sp0(t, thread);
1480
	set_tss_desc(cpu, t);
L
Linus Torvalds 已提交
1481 1482 1483
	load_TR_desc();
	load_LDT(&init_mm.context);

1484 1485
	t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap);

1486
#ifdef CONFIG_DOUBLEFAULT
L
Linus Torvalds 已提交
1487 1488
	/* Set up doublefault TSS pointer in the GDT */
	__set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss);
1489
#endif
L
Linus Torvalds 已提交
1490

1491
	clear_all_debug_regs();
1492
	dbg_restore_debug_regs();
L
Linus Torvalds 已提交
1493

1494
	fpu__cpu_init();
L
Linus Torvalds 已提交
1495
}
1496
#endif
1497 1498 1499 1500 1501 1502 1503 1504

#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
1505 1506 1507 1508 1509 1510

inline bool __static_cpu_has_safe(u16 bit)
{
	return boot_cpu_has(bit);
}
EXPORT_SYMBOL_GPL(__static_cpu_has_safe);