common.c 40.3 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/export.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
#include <linux/sched/mm.h>
11
#include <linux/sched/clock.h>
12
#include <linux/sched/task.h>
13
#include <linux/init.h>
14
#include <linux/kprobes.h>
15
#include <linux/kgdb.h>
L
Linus Torvalds 已提交
16
#include <linux/smp.h>
17
#include <linux/io.h>
18
#include <linux/syscore_ops.h>
19 20

#include <asm/stackprotector.h>
21
#include <asm/perf_event.h>
L
Linus Torvalds 已提交
22
#include <asm/mmu_context.h>
23
#include <asm/archrandom.h>
24 25
#include <asm/hypervisor.h>
#include <asm/processor.h>
26
#include <asm/tlbflush.h>
27
#include <asm/debugreg.h>
28
#include <asm/sections.h>
29
#include <asm/vsyscall.h>
A
Alan Cox 已提交
30 31
#include <linux/topology.h>
#include <linux/cpumask.h>
32
#include <asm/pgtable.h>
A
Arun Sharma 已提交
33
#include <linux/atomic.h>
34 35 36 37
#include <asm/proto.h>
#include <asm/setup.h>
#include <asm/apic.h>
#include <asm/desc.h>
38
#include <asm/fpu/internal.h>
39
#include <asm/mtrr.h>
40
#include <asm/hwcap2.h>
A
Alan Cox 已提交
41
#include <linux/numa.h>
42
#include <asm/asm.h>
43
#include <asm/bugs.h>
44
#include <asm/cpu.h>
45
#include <asm/mce.h>
46
#include <asm/msr.h>
47
#include <asm/pat.h>
48 49
#include <asm/microcode.h>
#include <asm/microcode_intel.h>
50 51

#ifdef CONFIG_X86_LOCAL_APIC
T
Tejun Heo 已提交
52
#include <asm/uv/uv.h>
L
Linus Torvalds 已提交
53 54 55 56
#endif

#include "cpu.h"

57 58
u32 elf_hwcap2 __read_mostly;

59 60
/* all of these masks are initialized in setup_cpu_local_masks() */
cpumask_var_t cpu_initialized_mask;
61 62
cpumask_var_t cpu_callout_mask;
cpumask_var_t cpu_callin_mask;
63 64 65 66

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

B
Brian Gerst 已提交
67
/* correctly size the local cpu masks */
68
void __init setup_cpu_local_masks(void)
B
Brian Gerst 已提交
69 70 71 72 73 74 75
{
	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);
}

76
static void default_init(struct cpuinfo_x86 *c)
77 78
{
#ifdef CONFIG_X86_64
79
	cpu_detect_cache_sizes(c);
80 81 82 83 84 85 86 87 88 89 90 91 92
#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
}

93
static const struct cpu_dev default_cpu = {
94 95 96 97 98
	.c_init		= default_init,
	.c_vendor	= "Unknown",
	.c_x86_vendor	= X86_VENDOR_UNKNOWN,
};

99
static const struct cpu_dev *this_cpu = &default_cpu;
100

B
Brian Gerst 已提交
101
DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = {
Y
Yinghai Lu 已提交
102
#ifdef CONFIG_X86_64
B
Brian Gerst 已提交
103 104 105 106 107
	/*
	 * 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
	 *
108
	 * TLS descriptors are currently at a different place compared to i386.
B
Brian Gerst 已提交
109 110
	 * Hopefully nobody expects them at a fixed place (Wine?)
	 */
A
Akinobu Mita 已提交
111 112 113 114 115 116
	[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
117
#else
A
Akinobu Mita 已提交
118 119 120 121
	[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),
122 123 124 125 126
	/*
	 * 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.
	 */
127
	/* 32-bit code */
A
Akinobu Mita 已提交
128
	[GDT_ENTRY_PNPBIOS_CS32]	= GDT_ENTRY_INIT(0x409a, 0, 0xffff),
129
	/* 16-bit code */
A
Akinobu Mita 已提交
130
	[GDT_ENTRY_PNPBIOS_CS16]	= GDT_ENTRY_INIT(0x009a, 0, 0xffff),
131
	/* 16-bit data */
A
Akinobu Mita 已提交
132
	[GDT_ENTRY_PNPBIOS_DS]		= GDT_ENTRY_INIT(0x0092, 0, 0xffff),
133
	/* 16-bit data */
A
Akinobu Mita 已提交
134
	[GDT_ENTRY_PNPBIOS_TS1]		= GDT_ENTRY_INIT(0x0092, 0, 0),
135
	/* 16-bit data */
A
Akinobu Mita 已提交
136
	[GDT_ENTRY_PNPBIOS_TS2]		= GDT_ENTRY_INIT(0x0092, 0, 0),
137 138 139 140
	/*
	 * The APM segments have byte granularity and their bases
	 * are set at run time.  All have 64k limits.
	 */
141
	/* 32-bit code */
A
Akinobu Mita 已提交
142
	[GDT_ENTRY_APMBIOS_BASE]	= GDT_ENTRY_INIT(0x409a, 0, 0xffff),
143
	/* 16-bit code */
A
Akinobu Mita 已提交
144
	[GDT_ENTRY_APMBIOS_BASE+1]	= GDT_ENTRY_INIT(0x009a, 0, 0xffff),
145
	/* data */
146
	[GDT_ENTRY_APMBIOS_BASE+2]	= GDT_ENTRY_INIT(0x4092, 0, 0xffff),
147

A
Akinobu Mita 已提交
148 149
	[GDT_ENTRY_ESPFIX_SS]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
	[GDT_ENTRY_PERCPU]		= GDT_ENTRY_INIT(0xc092, 0, 0xfffff),
150
	GDT_STACK_CANARY_INIT
Y
Yinghai Lu 已提交
151
#endif
B
Brian Gerst 已提交
152
} };
153
EXPORT_PER_CPU_SYMBOL_GPL(gdt_page);
154

155
static int __init x86_mpx_setup(char *s)
156
{
157
	/* require an exact match without trailing characters */
158 159
	if (strlen(s))
		return 0;
160

161 162 163
	/* do not emit a message if the feature is not present */
	if (!boot_cpu_has(X86_FEATURE_MPX))
		return 1;
164

165 166
	setup_clear_cpu_cap(X86_FEATURE_MPX);
	pr_info("nompx: Intel Memory Protection Extensions (MPX) disabled\n");
167 168
	return 1;
}
169
__setup("nompx", x86_mpx_setup);
170

171
#ifdef CONFIG_X86_64
172
static int __init x86_nopcid_setup(char *s)
173
{
174 175 176
	/* nopcid doesn't accept parameters */
	if (s)
		return -EINVAL;
177 178 179

	/* do not emit a message if the feature is not present */
	if (!boot_cpu_has(X86_FEATURE_PCID))
180
		return 0;
181 182 183

	setup_clear_cpu_cap(X86_FEATURE_PCID);
	pr_info("nopcid: PCID feature disabled\n");
184
	return 0;
185
}
186
early_param("nopcid", x86_nopcid_setup);
187 188
#endif

189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
static int __init x86_noinvpcid_setup(char *s)
{
	/* noinvpcid doesn't accept parameters */
	if (s)
		return -EINVAL;

	/* do not emit a message if the feature is not present */
	if (!boot_cpu_has(X86_FEATURE_INVPCID))
		return 0;

	setup_clear_cpu_cap(X86_FEATURE_INVPCID);
	pr_info("noinvpcid: INVPCID feature disabled\n");
	return 0;
}
early_param("noinvpcid", x86_noinvpcid_setup);

205
#ifdef CONFIG_X86_32
206 207
static int cachesize_override = -1;
static int disable_x86_serial_nr = 1;
L
Linus Torvalds 已提交
208

209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
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;

228 229 230 231 232 233 234
	/*
	 * 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
235 236 237 238 239 240 241 242 243 244 245
	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"

246 247
		      : "=&r" (f1), "=&r" (f2)
		      : "ir" (flag));
248 249 250 251 252

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

/* Probe for the CPUID instruction */
253
int have_cpuid_p(void)
254 255 256 257
{
	return flag_is_changeable_p(X86_EFLAGS_ID);
}

258
static void squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
259
{
I
Ingo Molnar 已提交
260 261 262 263 264 265 266 267 268 269 270
	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);

271
	pr_notice("CPU serial number disabled.\n");
I
Ingo Molnar 已提交
272 273 274 275
	clear_cpu_cap(c, X86_FEATURE_PN);

	/* Disabling the serial number may affect the cpuid level */
	c->cpuid_level = cpuid_eax(0);
276 277 278 279 280 281 282 283
}

static int __init x86_serial_nr_setup(char *s)
{
	disable_x86_serial_nr = 0;
	return 1;
}
__setup("serialnumber", x86_serial_nr_setup);
284
#else
285 286 287 288 289 290 291
static inline int flag_is_changeable_p(u32 flag)
{
	return 1;
}
static inline void squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
{
}
292
#endif
293

294 295
static __init int setup_disable_smep(char *arg)
{
296
	setup_clear_cpu_cap(X86_FEATURE_SMEP);
297 298
	/* Check for things that depend on SMEP being enabled: */
	check_mpx_erratum(&boot_cpu_data);
299 300 301 302
	return 1;
}
__setup("nosmep", setup_disable_smep);

303
static __always_inline void setup_smep(struct cpuinfo_x86 *c)
304
{
305
	if (cpu_has(c, X86_FEATURE_SMEP))
A
Andy Lutomirski 已提交
306
		cr4_set_bits(X86_CR4_SMEP);
307 308
}

309 310
static __init int setup_disable_smap(char *arg)
{
311
	setup_clear_cpu_cap(X86_FEATURE_SMAP);
312 313 314 315
	return 1;
}
__setup("nosmap", setup_disable_smap);

316 317
static __always_inline void setup_smap(struct cpuinfo_x86 *c)
{
318
	unsigned long eflags = native_save_fl();
319 320 321 322

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

323 324
	if (cpu_has(c, X86_FEATURE_SMAP)) {
#ifdef CONFIG_X86_SMAP
A
Andy Lutomirski 已提交
325
		cr4_set_bits(X86_CR4_SMAP);
326
#else
A
Andy Lutomirski 已提交
327
		cr4_clear_bits(X86_CR4_SMAP);
328 329
#endif
	}
330 331
}

332 333 334 335 336 337 338
/*
 * Protection Keys are not available in 32-bit mode.
 */
static bool pku_disabled;

static __always_inline void setup_pku(struct cpuinfo_x86 *c)
{
339 340 341 342
	/* check the boot processor, plus compile options for PKU: */
	if (!cpu_feature_enabled(X86_FEATURE_PKU))
		return;
	/* checks the actual processor's cpuid bits: */
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
	if (!cpu_has(c, X86_FEATURE_PKU))
		return;
	if (pku_disabled)
		return;

	cr4_set_bits(X86_CR4_PKE);
	/*
	 * Seting X86_CR4_PKE will cause the X86_FEATURE_OSPKE
	 * cpuid bit to be set.  We need to ensure that we
	 * update that bit in this CPU's "cpu_info".
	 */
	get_cpu_cap(c);
}

#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
static __init int setup_disable_pku(char *arg)
{
	/*
	 * Do not clear the X86_FEATURE_PKU bit.  All of the
	 * runtime checks are against OSPKE so clearing the
	 * bit does nothing.
	 *
	 * This way, we will see "pku" in cpuinfo, but not
	 * "ospke", which is exactly what we want.  It shows
	 * that the CPU has PKU, but the OS has not enabled it.
	 * This happens to be exactly how a system would look
	 * if we disabled the config option.
	 */
	pr_info("x86: 'nopku' specified, disabling Memory Protection Keys\n");
	pku_disabled = true;
	return 1;
}
__setup("nopku", setup_disable_pku);
#endif /* CONFIG_X86_64 */

378 379 380 381 382 383 384 385 386
/*
 * 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
387

388
static const struct cpuid_dependent_feature
389 390 391 392 393 394 395
cpuid_dependent_features[] = {
	{ X86_FEATURE_MWAIT,		0x00000005 },
	{ X86_FEATURE_DCA,		0x00000009 },
	{ X86_FEATURE_XSAVE,		0x0000000d },
	{ 0, 0 }
};

396
static void filter_cpuid_features(struct cpuinfo_x86 *c, bool warn)
397 398
{
	const struct cpuid_dependent_feature *df;
399

400
	for (df = cpuid_dependent_features; df->feature; df++) {
I
Ingo Molnar 已提交
401 402 403

		if (!cpu_has(c, df->feature))
			continue;
404 405 406 407 408 409 410
		/*
		 * 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
411
		if (!((s32)df->level < 0 ?
412
		     (u32)df->level > (u32)c->extended_cpuid_level :
I
Ingo Molnar 已提交
413 414 415 416 417 418 419
		     (s32)df->level > (s32)c->cpuid_level))
			continue;

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

420 421
		pr_warn("CPU: CPU feature " X86_CAP_FMT " disabled, no CPUID level 0x%x\n",
			x86_cap_flag(df->feature), df->level);
422
	}
423
}
424

425 426 427
/*
 * 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
428 429
 * in particular, if CPUID levels 0x80000002..4 are supported, this
 * isn't used
430 431 432
 */

/* Look up CPU names by table lookup. */
433
static const char *table_lookup_model(struct cpuinfo_x86 *c)
434
{
435 436
#ifdef CONFIG_X86_32
	const struct legacy_cpu_model_info *info;
437 438 439 440 441 442 443

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

	if (!this_cpu)
		return NULL;

444
	info = this_cpu->legacy_models;
445

446
	while (info->family) {
447 448 449 450
		if (info->family == c->x86)
			return info->model_names[c->x86_model];
		info++;
	}
451
#endif
452 453 454
	return NULL;		/* Not found */
}

455 456
__u32 cpu_caps_cleared[NCAPINTS + NBUGINTS];
__u32 cpu_caps_set[NCAPINTS + NBUGINTS];
457

458 459 460 461 462
void load_percpu_segment(int cpu)
{
#ifdef CONFIG_X86_32
	loadsegment(fs, __KERNEL_PERCPU);
#else
463
	__loadsegment_simple(gs, 0);
464 465
	wrmsrl(MSR_GS_BASE, (unsigned long)per_cpu(irq_stack_union.gs_base, cpu));
#endif
466
	load_stack_canary_segment();
467 468
}

469 470 471 472 473
#ifdef CONFIG_X86_32
/* The 32-bit entry code needs to find cpu_entry_area. */
DEFINE_PER_CPU(struct cpu_entry_area *, cpu_entry_area);
#endif

474 475 476 477 478 479 480 481 482 483 484
#ifdef CONFIG_X86_64
/*
 * 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
};
485
#endif
486

487 488 489 490 491 492 493 494 495 496 497
/* Load the original GDT from the per-cpu structure */
void load_direct_gdt(int cpu)
{
	struct desc_ptr gdt_descr;

	gdt_descr.address = (long)get_cpu_gdt_rw(cpu);
	gdt_descr.size = GDT_SIZE - 1;
	load_gdt(&gdt_descr);
}
EXPORT_SYMBOL_GPL(load_direct_gdt);

498 499 500 501 502 503 504 505 506
/* Load a fixmap remapping of the per-cpu GDT */
void load_fixmap_gdt(int cpu)
{
	struct desc_ptr gdt_descr;

	gdt_descr.address = (long)get_cpu_gdt_ro(cpu);
	gdt_descr.size = GDT_SIZE - 1;
	load_gdt(&gdt_descr);
}
507
EXPORT_SYMBOL_GPL(load_fixmap_gdt);
508

I
Ingo Molnar 已提交
509 510 511 512
/*
 * Current gdt points %fs at the "master" per-cpu area: after this,
 * it's on the real one.
 */
513
void switch_to_new_gdt(int cpu)
514
{
515 516
	/* Load the original GDT */
	load_direct_gdt(cpu);
517
	/* Reload the per-cpu base */
518
	load_percpu_segment(cpu);
519 520
}

521
static const struct cpu_dev *cpu_devs[X86_VENDOR_NUM] = {};
L
Linus Torvalds 已提交
522

523
static void get_model_name(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
524 525
{
	unsigned int *v;
526
	char *p, *q, *s;
L
Linus Torvalds 已提交
527

528
	if (c->extended_cpuid_level < 0x80000004)
529
		return;
L
Linus Torvalds 已提交
530

I
Ingo Molnar 已提交
531
	v = (unsigned int *)c->x86_model_id;
L
Linus Torvalds 已提交
532 533 534 535 536
	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;

537 538 539 540 541 542 543 544 545 546 547 548 549 550 551
	/* 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
552 553
}

554
void cpu_detect_cache_sizes(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
555
{
556
	unsigned int n, dummy, ebx, ecx, edx, l2size;
L
Linus Torvalds 已提交
557

558
	n = c->extended_cpuid_level;
L
Linus Torvalds 已提交
559 560

	if (n >= 0x80000005) {
561 562
		cpuid(0x80000005, &dummy, &ebx, &ecx, &edx);
		c->x86_cache_size = (ecx>>24) + (edx>>24);
563 564 565 566
#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
567 568 569 570 571
	}

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

572
	cpuid(0x80000006, &dummy, &ebx, &ecx, &edx);
L
Linus Torvalds 已提交
573
	l2size = ecx >> 16;
574

575 576 577
#ifdef CONFIG_X86_64
	c->x86_tlbsize += ((ebx >> 16) & 0xfff) + (ebx & 0xfff);
#else
L
Linus Torvalds 已提交
578
	/* do processor-specific cache resizing */
579 580
	if (this_cpu->legacy_cache_size)
		l2size = this_cpu->legacy_cache_size(c, l2size);
L
Linus Torvalds 已提交
581 582 583 584 585

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

586
	if (l2size == 0)
L
Linus Torvalds 已提交
587
		return;		/* Again, no L2 cache is possible */
588
#endif
L
Linus Torvalds 已提交
589 590 591 592

	c->x86_cache_size = l2size;
}

593 594 595 596 597 598
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];
599
u16 __read_mostly tlb_lld_1g[NR_INFO];
600

601
static void cpu_detect_tlb(struct cpuinfo_x86 *c)
602 603 604 605
{
	if (this_cpu->c_detect_tlb)
		this_cpu->c_detect_tlb(c);

606
	pr_info("Last level iTLB entries: 4KB %d, 2MB %d, 4MB %d\n",
607
		tlb_lli_4k[ENTRIES], tlb_lli_2m[ENTRIES],
608 609 610 611 612
		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]);
613 614
}

615
void detect_ht(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
616
{
B
Borislav Petkov 已提交
617
#ifdef CONFIG_SMP
618 619
	u32 eax, ebx, ecx, edx;
	int index_msb, core_bits;
620
	static bool printed;
L
Linus Torvalds 已提交
621

622
	if (!cpu_has(c, X86_FEATURE_HT))
623
		return;
L
Linus Torvalds 已提交
624

625 626
	if (cpu_has(c, X86_FEATURE_CMP_LEGACY))
		goto out;
L
Linus Torvalds 已提交
627

628 629
	if (cpu_has(c, X86_FEATURE_XTOPOLOGY))
		return;
L
Linus Torvalds 已提交
630

631
	cpuid(1, &eax, &ebx, &ecx, &edx);
L
Linus Torvalds 已提交
632

633 634 635
	smp_num_siblings = (ebx & 0xff0000) >> 16;

	if (smp_num_siblings == 1) {
636
		pr_info_once("CPU0: Hyper-Threading is disabled\n");
I
Ingo Molnar 已提交
637 638
		goto out;
	}
639

I
Ingo Molnar 已提交
640 641
	if (smp_num_siblings <= 1)
		goto out;
642

I
Ingo Molnar 已提交
643 644
	index_msb = get_count_order(smp_num_siblings);
	c->phys_proc_id = apic->phys_pkg_id(c->initial_apicid, index_msb);
645

I
Ingo Molnar 已提交
646
	smp_num_siblings = smp_num_siblings / c->x86_max_cores;
647

I
Ingo Molnar 已提交
648
	index_msb = get_count_order(smp_num_siblings);
649

I
Ingo Molnar 已提交
650
	core_bits = get_count_order(c->x86_max_cores);
651

I
Ingo Molnar 已提交
652 653
	c->cpu_core_id = apic->phys_pkg_id(c->initial_apicid, index_msb) &
				       ((1 << core_bits) - 1);
L
Linus Torvalds 已提交
654

655
out:
656
	if (!printed && (c->x86_max_cores * smp_num_siblings) > 1) {
657 658 659 660
		pr_info("CPU: Physical Processor ID: %d\n",
			c->phys_proc_id);
		pr_info("CPU: Processor Core ID: %d\n",
			c->cpu_core_id);
661
		printed = 1;
662 663
	}
#endif
664
}
L
Linus Torvalds 已提交
665

666
static void get_cpu_vendor(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
667 668
{
	char *v = c->x86_vendor_id;
I
Ingo Molnar 已提交
669
	int i;
L
Linus Torvalds 已提交
670 671

	for (i = 0; i < X86_VENDOR_NUM; i++) {
Y
Yinghai Lu 已提交
672 673 674 675 676 677
		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
678

Y
Yinghai Lu 已提交
679 680 681
			this_cpu = cpu_devs[i];
			c->x86_vendor = this_cpu->c_x86_vendor;
			return;
L
Linus Torvalds 已提交
682 683
		}
	}
Y
Yinghai Lu 已提交
684

685 686
	pr_err_once("CPU: vendor_id '%s' unknown, using generic init.\n" \
		    "CPU: Your system may be unstable.\n", v);
Y
Yinghai Lu 已提交
687

688 689
	c->x86_vendor = X86_VENDOR_UNKNOWN;
	this_cpu = &default_cpu;
L
Linus Torvalds 已提交
690 691
}

692
void cpu_detect(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
693 694
{
	/* Get vendor name */
695 696 697 698
	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
699 700

	c->x86 = 4;
701
	/* Intel-defined flags: level 0x00000001 */
L
Linus Torvalds 已提交
702 703
	if (c->cpuid_level >= 0x00000001) {
		u32 junk, tfms, cap0, misc;
I
Ingo Molnar 已提交
704

L
Linus Torvalds 已提交
705
		cpuid(0x00000001, &tfms, &misc, &junk, &cap0);
706 707 708
		c->x86		= x86_family(tfms);
		c->x86_model	= x86_model(tfms);
		c->x86_mask	= x86_stepping(tfms);
I
Ingo Molnar 已提交
709

H
Huang, Ying 已提交
710 711
		if (cap0 & (1<<19)) {
			c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
712
			c->x86_cache_alignment = c->x86_clflush_size;
H
Huang, Ying 已提交
713
		}
L
Linus Torvalds 已提交
714 715
	}
}
716

717 718 719 720
static void apply_forced_caps(struct cpuinfo_x86 *c)
{
	int i;

721
	for (i = 0; i < NCAPINTS + NBUGINTS; i++) {
722 723 724 725 726
		c->x86_capability[i] &= ~cpu_caps_cleared[i];
		c->x86_capability[i] |= cpu_caps_set[i];
	}
}

727
void get_cpu_cap(struct cpuinfo_x86 *c)
728
{
729
	u32 eax, ebx, ecx, edx;
730

731 732
	/* Intel-defined flags: level 0x00000001 */
	if (c->cpuid_level >= 0x00000001) {
733
		cpuid(0x00000001, &eax, &ebx, &ecx, &edx);
I
Ingo Molnar 已提交
734

735 736
		c->x86_capability[CPUID_1_ECX] = ecx;
		c->x86_capability[CPUID_1_EDX] = edx;
737
	}
738

739 740 741 742
	/* Thermal and Power Management Leaf: level 0x00000006 (eax) */
	if (c->cpuid_level >= 0x00000006)
		c->x86_capability[CPUID_6_EAX] = cpuid_eax(0x00000006);

743 744 745
	/* Additional Intel-defined flags: level 0x00000007 */
	if (c->cpuid_level >= 0x00000007) {
		cpuid_count(0x00000007, 0, &eax, &ebx, &ecx, &edx);
746
		c->x86_capability[CPUID_7_0_EBX] = ebx;
747
		c->x86_capability[CPUID_7_ECX] = ecx;
748 749
	}

750 751 752 753
	/* Extended state features: level 0x0000000d */
	if (c->cpuid_level >= 0x0000000d) {
		cpuid_count(0x0000000d, 1, &eax, &ebx, &ecx, &edx);

754
		c->x86_capability[CPUID_D_1_EAX] = eax;
755 756
	}

757 758 759 760 761
	/* Additional Intel-defined flags: level 0x0000000F */
	if (c->cpuid_level >= 0x0000000F) {

		/* QoS sub-leaf, EAX=0Fh, ECX=0 */
		cpuid_count(0x0000000F, 0, &eax, &ebx, &ecx, &edx);
762 763
		c->x86_capability[CPUID_F_0_EDX] = edx;

764 765 766 767 768 769
		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);
770 771
			c->x86_capability[CPUID_F_1_EDX] = edx;

772 773 774
			if ((cpu_has(c, X86_FEATURE_CQM_OCCUP_LLC)) ||
			      ((cpu_has(c, X86_FEATURE_CQM_MBM_TOTAL)) ||
			       (cpu_has(c, X86_FEATURE_CQM_MBM_LOCAL)))) {
775 776 777 778 779 780 781 782 783
				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;
		}
	}

784
	/* AMD-defined flags: level 0x80000001 */
785 786 787 788 789 790
	eax = cpuid_eax(0x80000000);
	c->extended_cpuid_level = eax;

	if ((eax & 0xffff0000) == 0x80000000) {
		if (eax >= 0x80000001) {
			cpuid(0x80000001, &eax, &ebx, &ecx, &edx);
I
Ingo Molnar 已提交
791

792 793
			c->x86_capability[CPUID_8000_0001_ECX] = ecx;
			c->x86_capability[CPUID_8000_0001_EDX] = edx;
794 795 796
		}
	}

797 798 799 800 801 802 803
	if (c->extended_cpuid_level >= 0x80000007) {
		cpuid(0x80000007, &eax, &ebx, &ecx, &edx);

		c->x86_capability[CPUID_8000_0007_EBX] = ebx;
		c->x86_power = edx;
	}

804
	if (c->extended_cpuid_level >= 0x80000008) {
805
		cpuid(0x80000008, &eax, &ebx, &ecx, &edx);
806 807 808

		c->x86_virt_bits = (eax >> 8) & 0xff;
		c->x86_phys_bits = eax & 0xff;
809
		c->x86_capability[CPUID_8000_0008_EBX] = ebx;
810
	}
811 812 813
#ifdef CONFIG_X86_32
	else if (cpu_has(c, X86_FEATURE_PAE) || cpu_has(c, X86_FEATURE_PSE36))
		c->x86_phys_bits = 36;
814
#endif
815

816
	if (c->extended_cpuid_level >= 0x8000000a)
817
		c->x86_capability[CPUID_8000_000A_EDX] = cpuid_edx(0x8000000a);
818

819
	init_scattered_cpuid_features(c);
820 821 822 823 824 825 826

	/*
	 * Clear/Set all flags overridden by options, after probe.
	 * This needs to happen each time we re-probe, which may happen
	 * several times during CPU initialization.
	 */
	apply_forced_caps(c);
827
}
L
Linus Torvalds 已提交
828

829
static void identify_cpu_without_cpuid(struct cpuinfo_x86 *c)
Y
Yinghai Lu 已提交
830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854
{
#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
}

855 856 857 858 859 860 861 862 863
/*
 * 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.
 */
864
static void __init early_identify_cpu(struct cpuinfo_x86 *c)
865
{
866 867
#ifdef CONFIG_X86_64
	c->x86_clflush_size = 64;
868 869
	c->x86_phys_bits = 36;
	c->x86_virt_bits = 48;
870
#else
H
Huang, Ying 已提交
871
	c->x86_clflush_size = 32;
872 873
	c->x86_phys_bits = 32;
	c->x86_virt_bits = 32;
874
#endif
875
	c->x86_cache_alignment = c->x86_clflush_size;
876

877
	memset(&c->x86_capability, 0, sizeof c->x86_capability);
878
	c->extended_cpuid_level = 0;
879

Y
Yinghai Lu 已提交
880
	/* cyrix could have cpuid enabled via c_identify()*/
881 882 883 884
	if (have_cpuid_p()) {
		cpu_detect(c);
		get_cpu_vendor(c);
		get_cpu_cap(c);
B
Borislav Petkov 已提交
885
		setup_force_cpu_cap(X86_FEATURE_CPUID);
886

887 888
		if (this_cpu->c_early_init)
			this_cpu->c_early_init(c);
889

890 891
		c->cpu_index = 0;
		filter_cpuid_features(c, false);
892

893 894
		if (this_cpu->c_bsp_init)
			this_cpu->c_bsp_init(c);
B
Borislav Petkov 已提交
895 896 897
	} else {
		identify_cpu_without_cpuid(c);
		setup_clear_cpu_cap(X86_FEATURE_CPUID);
898
	}
899 900

	setup_force_cpu_cap(X86_FEATURE_ALWAYS);
901 902 903 904

	/* Assume for now that ALL x86 CPUs are insecure */
	setup_force_cpu_bug(X86_BUG_CPU_INSECURE);

905
	fpu__init_system(c);
906 907 908 909 910 911 912 913

#ifdef CONFIG_X86_32
	/*
	 * Regardless of whether PCID is enumerated, the SDM says
	 * that it can't be enabled in 32-bit mode.
	 */
	setup_clear_cpu_cap(X86_FEATURE_PCID);
#endif
914 915
}

916 917
void __init early_cpu_init(void)
{
918
	const struct cpu_dev *const *cdev;
Y
Yinghai Lu 已提交
919 920
	int count = 0;

921
#ifdef CONFIG_PROCESSOR_SELECT
922
	pr_info("KERNEL supported cpus:\n");
923 924
#endif

Y
Yinghai Lu 已提交
925
	for (cdev = __x86_cpu_dev_start; cdev < __x86_cpu_dev_end; cdev++) {
926
		const struct cpu_dev *cpudev = *cdev;
927

Y
Yinghai Lu 已提交
928 929 930 931 932
		if (count >= X86_VENDOR_NUM)
			break;
		cpu_devs[count] = cpudev;
		count++;

933
#ifdef CONFIG_PROCESSOR_SELECT
934 935 936 937 938 939
		{
			unsigned int j;

			for (j = 0; j < 2; j++) {
				if (!cpudev->c_ident[j])
					continue;
940
				pr_info("  %s %s\n", cpudev->c_vendor,
941 942
					cpudev->c_ident[j]);
			}
Y
Yinghai Lu 已提交
943
		}
944
#endif
Y
Yinghai Lu 已提交
945
	}
946
	early_identify_cpu(&boot_cpu_data);
947
}
948

949
/*
B
Borislav Petkov 已提交
950 951 952 953 954
 * 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
955
 * unless we can find a reliable way to detect all the broken cases.
B
Borislav Petkov 已提交
956
 * Enable it explicitly on 64-bit for non-constant inputs of cpu_has().
957
 */
958
static void detect_nopl(struct cpuinfo_x86 *c)
959
{
B
Borislav Petkov 已提交
960
#ifdef CONFIG_X86_32
961
	clear_cpu_cap(c, X86_FEATURE_NOPL);
B
Borislav Petkov 已提交
962 963
#else
	set_cpu_cap(c, X86_FEATURE_NOPL);
964
#endif
965
}
966

967 968 969
static void detect_null_seg_behavior(struct cpuinfo_x86 *c)
{
#ifdef CONFIG_X86_64
970
	/*
971 972 973 974 975
	 * Empirically, writing zero to a segment selector on AMD does
	 * not clear the base, whereas writing zero to a segment
	 * selector on Intel does clear the base.  Intel's behavior
	 * allows slightly faster context switches in the common case
	 * where GS is unused by the prev and next threads.
976
	 *
977 978 979 980 981 982
	 * Since neither vendor documents this anywhere that I can see,
	 * detect it directly instead of hardcoding the choice by
	 * vendor.
	 *
	 * I've designated AMD's behavior as the "bug" because it's
	 * counterintuitive and less friendly.
983
	 */
984 985 986 987 988 989 990 991 992

	unsigned long old_base, tmp;
	rdmsrl(MSR_FS_BASE, old_base);
	wrmsrl(MSR_FS_BASE, 1);
	loadsegment(fs, 0);
	rdmsrl(MSR_FS_BASE, tmp);
	if (tmp != 0)
		set_cpu_bug(c, X86_BUG_NULL_SEG);
	wrmsrl(MSR_FS_BASE, old_base);
B
Borislav Petkov 已提交
993
#endif
994 995
}

996
static void generic_identify(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
997
{
Y
Yinghai Lu 已提交
998
	c->extended_cpuid_level = 0;
L
Linus Torvalds 已提交
999

1000
	if (!have_cpuid_p())
Y
Yinghai Lu 已提交
1001
		identify_cpu_without_cpuid(c);
1002

Y
Yinghai Lu 已提交
1003
	/* cyrix could have cpuid enabled via c_identify()*/
I
Ingo Molnar 已提交
1004
	if (!have_cpuid_p())
Y
Yinghai Lu 已提交
1005
		return;
L
Linus Torvalds 已提交
1006

1007
	cpu_detect(c);
L
Linus Torvalds 已提交
1008

1009
	get_cpu_vendor(c);
L
Linus Torvalds 已提交
1010

1011
	get_cpu_cap(c);
L
Linus Torvalds 已提交
1012

1013 1014
	if (c->cpuid_level >= 0x00000001) {
		c->initial_apicid = (cpuid_ebx(1) >> 24) & 0xFF;
1015
#ifdef CONFIG_X86_32
B
Borislav Petkov 已提交
1016
# ifdef CONFIG_SMP
1017
		c->apicid = apic->phys_pkg_id(c->initial_apicid, 0);
1018
# else
1019
		c->apicid = c->initial_apicid;
1020 1021 1022
# endif
#endif
		c->phys_proc_id = c->initial_apicid;
1023
	}
L
Linus Torvalds 已提交
1024

1025
	get_model_name(c); /* Default name */
L
Linus Torvalds 已提交
1026

1027
	detect_nopl(c);
1028 1029

	detect_null_seg_behavior(c);
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054

	/*
	 * ESPFIX is a strange bug.  All real CPUs have it.  Paravirt
	 * systems that run Linux at CPL > 0 may or may not have the
	 * issue, but, even if they have the issue, there's absolutely
	 * nothing we can do about it because we can't use the real IRET
	 * instruction.
	 *
	 * NB: For the time being, only 32-bit kernels support
	 * X86_BUG_ESPFIX as such.  64-bit kernels directly choose
	 * whether to apply espfix using paravirt hooks.  If any
	 * non-paravirt system ever shows up that does *not* have the
	 * ESPFIX issue, we can change this.
	 */
#ifdef CONFIG_X86_32
# ifdef CONFIG_PARAVIRT
	do {
		extern void native_iret(void);
		if (pv_cpu_ops.iret == native_iret)
			set_cpu_bug(c, X86_BUG_ESPFIX);
	} while (0);
# else
	set_cpu_bug(c, X86_BUG_ESPFIX);
# endif
#endif
L
Linus Torvalds 已提交
1055 1056
}

1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
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);
	}
}

1071
/*
1072 1073
 * Validate that ACPI/mptables have the same information about the
 * effective APIC id and update the package map.
1074
 */
1075
static void validate_apic_and_package_id(struct cpuinfo_x86 *c)
1076 1077
{
#ifdef CONFIG_SMP
1078
	unsigned int apicid, cpu = smp_processor_id();
1079 1080 1081

	apicid = apic->cpu_present_to_apicid(cpu);

1082 1083
	if (apicid != c->apicid) {
		pr_err(FW_BUG "CPU%u: APIC id mismatch. Firmware: %x APIC: %x\n",
1084 1085
		       cpu, apicid, c->initial_apicid);
	}
1086
	BUG_ON(topology_update_package_map(c->phys_proc_id, cpu));
1087 1088 1089 1090 1091
#else
	c->logical_proc_id = 0;
#endif
}

L
Linus Torvalds 已提交
1092 1093 1094
/*
 * This does the hard work of actually picking apart the CPU stuff...
 */
1095
static void identify_cpu(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
1096 1097 1098 1099 1100 1101 1102 1103 1104
{
	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 */
1105
	c->x86_max_cores = 1;
1106
	c->x86_coreid_bits = 0;
1107
	c->cu_id = 0xff;
1108
#ifdef CONFIG_X86_64
1109
	c->x86_clflush_size = 64;
1110 1111
	c->x86_phys_bits = 36;
	c->x86_virt_bits = 48;
1112 1113
#else
	c->cpuid_level = -1;	/* CPUID not detected */
1114
	c->x86_clflush_size = 32;
1115 1116
	c->x86_phys_bits = 32;
	c->x86_virt_bits = 32;
1117 1118
#endif
	c->x86_cache_alignment = c->x86_clflush_size;
L
Linus Torvalds 已提交
1119 1120 1121 1122
	memset(&c->x86_capability, 0, sizeof c->x86_capability);

	generic_identify(c);

1123
	if (this_cpu->c_identify)
L
Linus Torvalds 已提交
1124 1125
		this_cpu->c_identify(c);

1126
	/* Clear/Set all flags overridden by options, after probe */
1127
	apply_forced_caps(c);
1128

1129
#ifdef CONFIG_X86_64
1130
	c->apicid = apic->phys_pkg_id(c->initial_apicid, 0);
1131 1132
#endif

L
Linus Torvalds 已提交
1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
	/*
	 * 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);

1149 1150 1151 1152
	/* Set up SMEP/SMAP */
	setup_smep(c);
	setup_smap(c);

L
Linus Torvalds 已提交
1153
	/*
I
Ingo Molnar 已提交
1154 1155
	 * The vendor-specific functions might have changed features.
	 * Now we do "generic changes."
L
Linus Torvalds 已提交
1156 1157
	 */

1158 1159 1160
	/* Filter out anything that depends on CPUID levels we don't have */
	filter_cpuid_features(c, true);

L
Linus Torvalds 已提交
1161
	/* If the model name is still unset, do table lookup. */
1162
	if (!c->x86_model_id[0]) {
1163
		const char *p;
L
Linus Torvalds 已提交
1164
		p = table_lookup_model(c);
1165
		if (p)
L
Linus Torvalds 已提交
1166 1167 1168 1169
			strcpy(c->x86_model_id, p);
		else
			/* Last resort... */
			sprintf(c->x86_model_id, "%02x/%02x",
1170
				c->x86, c->x86_model);
L
Linus Torvalds 已提交
1171 1172
	}

1173 1174 1175 1176
#ifdef CONFIG_X86_64
	detect_ht(c);
#endif

1177
	x86_init_rdrand(c);
1178
	x86_init_cache_qos(c);
1179
	setup_pku(c);
1180 1181

	/*
1182
	 * Clear/Set all flags overridden by options, need do it
1183 1184
	 * before following smp all cpus cap AND.
	 */
1185
	apply_forced_caps(c);
1186

L
Linus Torvalds 已提交
1187 1188 1189 1190 1191 1192
	/*
	 * 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.
	 */
1193
	if (c != &boot_cpu_data) {
L
Linus Torvalds 已提交
1194
		/* AND the already accumulated flags with these */
1195
		for (i = 0; i < NCAPINTS; i++)
L
Linus Torvalds 已提交
1196
			boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
1197 1198 1199 1200

		/* 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
1201 1202 1203
	}

	/* Init Machine Check Exception if available. */
1204
	mcheck_cpu_init(c);
1205 1206

	select_idle_routine(c);
1207

1208
#ifdef CONFIG_NUMA
1209 1210
	numa_add_cpu(smp_processor_id());
#endif
1211
}
S
Shaohua Li 已提交
1212

1213 1214 1215 1216
/*
 * Set up the CPU state needed to execute SYSENTER/SYSEXIT instructions
 * on 32-bit kernels:
 */
1217 1218 1219
#ifdef CONFIG_X86_32
void enable_sep_cpu(void)
{
1220 1221
	struct tss_struct *tss;
	int cpu;
1222

1223 1224 1225
	if (!boot_cpu_has(X86_FEATURE_SEP))
		return;

1226
	cpu = get_cpu();
1227
	tss = &per_cpu(cpu_tss_rw, cpu);
1228 1229

	/*
1230 1231
	 * 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.
1232
	 */
1233 1234

	tss->x86_tss.ss1 = __KERNEL_CS;
1235
	wrmsr(MSR_IA32_SYSENTER_CS, tss->x86_tss.ss1, 0);
1236
	wrmsr(MSR_IA32_SYSENTER_ESP, (unsigned long)(cpu_entry_stack(cpu) + 1), 0);
1237
	wrmsr(MSR_IA32_SYSENTER_EIP, (unsigned long)entry_SYSENTER_32, 0);
1238

1239 1240
	put_cpu();
}
1241 1242
#endif

1243 1244 1245
void __init identify_boot_cpu(void)
{
	identify_cpu(&boot_cpu_data);
1246
#ifdef CONFIG_X86_32
1247
	sysenter_setup();
L
Li Shaohua 已提交
1248
	enable_sep_cpu();
1249
#endif
1250
	cpu_detect_tlb(&boot_cpu_data);
1251
}
S
Shaohua Li 已提交
1252

1253
void identify_secondary_cpu(struct cpuinfo_x86 *c)
1254 1255 1256
{
	BUG_ON(c == &boot_cpu_data);
	identify_cpu(c);
1257
#ifdef CONFIG_X86_32
1258
	enable_sep_cpu();
1259
#endif
1260
	mtrr_ap_init();
1261
	validate_apic_and_package_id(c);
L
Linus Torvalds 已提交
1262 1263
}

A
Andi Kleen 已提交
1264 1265
static __init int setup_noclflush(char *arg)
{
1266
	setup_clear_cpu_cap(X86_FEATURE_CLFLUSH);
1267
	setup_clear_cpu_cap(X86_FEATURE_CLFLUSHOPT);
A
Andi Kleen 已提交
1268 1269 1270 1271
	return 1;
}
__setup("noclflush", setup_noclflush);

1272
void print_cpu_info(struct cpuinfo_x86 *c)
L
Linus Torvalds 已提交
1273
{
1274
	const char *vendor = NULL;
L
Linus Torvalds 已提交
1275

I
Ingo Molnar 已提交
1276
	if (c->x86_vendor < X86_VENDOR_NUM) {
L
Linus Torvalds 已提交
1277
		vendor = this_cpu->c_vendor;
I
Ingo Molnar 已提交
1278 1279 1280 1281
	} else {
		if (c->cpuid_level >= 0)
			vendor = c->x86_vendor_id;
	}
L
Linus Torvalds 已提交
1282

1283
	if (vendor && !strstr(c->x86_model_id, vendor))
1284
		pr_cont("%s ", vendor);
L
Linus Torvalds 已提交
1285

1286
	if (c->x86_model_id[0])
1287
		pr_cont("%s", c->x86_model_id);
L
Linus Torvalds 已提交
1288
	else
1289
		pr_cont("%d86", c->x86);
L
Linus Torvalds 已提交
1290

1291
	pr_cont(" (family: 0x%x, model: 0x%x", c->x86, c->x86_model);
1292

1293
	if (c->x86_mask || c->cpuid_level >= 0)
1294
		pr_cont(", stepping: 0x%x)\n", c->x86_mask);
L
Linus Torvalds 已提交
1295
	else
1296
		pr_cont(")\n");
L
Linus Torvalds 已提交
1297 1298
}

1299 1300 1301 1302 1303 1304
/*
 * clearcpuid= was already parsed in fpu__init_parse_early_param.
 * But we need to keep a dummy __setup around otherwise it would
 * show up as an environment variable for init.
 */
static __init int setup_clearcpuid(char *arg)
1305 1306 1307
{
	return 1;
}
1308
__setup("clearcpuid=", setup_clearcpuid);
1309

1310
#ifdef CONFIG_X86_64
1311
DEFINE_PER_CPU_FIRST(union irq_stack_union,
1312
		     irq_stack_union) __aligned(PAGE_SIZE) __visible;
I
Ingo Molnar 已提交
1313

1314
/*
1315 1316
 * The following percpu variables are hot.  Align current_task to
 * cacheline size such that they fall in the same cacheline.
1317 1318 1319 1320
 */
DEFINE_PER_CPU(struct task_struct *, current_task) ____cacheline_aligned =
	&init_task;
EXPORT_PER_CPU_SYMBOL(current_task);
1321

1322
DEFINE_PER_CPU(char *, irq_stack_ptr) =
1323
	init_per_cpu_var(irq_stack_union.irq_stack) + IRQ_STACK_SIZE;
1324

1325
DEFINE_PER_CPU(unsigned int, irq_count) __visible = -1;
1326

1327 1328 1329
DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
EXPORT_PER_CPU_SYMBOL(__preempt_count);

1330 1331
/* May not be marked __init: used by software suspend */
void syscall_init(void)
L
Linus Torvalds 已提交
1332
{
1333 1334 1335
	extern char _entry_trampoline[];
	extern char entry_SYSCALL_64_trampoline[];

1336
	int cpu = smp_processor_id();
1337 1338 1339
	unsigned long SYSCALL64_entry_trampoline =
		(unsigned long)get_cpu_entry_area(cpu)->entry_trampoline +
		(entry_SYSCALL_64_trampoline - _entry_trampoline);
1340

1341
	wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
1342
	wrmsrl(MSR_LSTAR, SYSCALL64_entry_trampoline);
1343 1344

#ifdef CONFIG_IA32_EMULATION
1345
	wrmsrl(MSR_CSTAR, (unsigned long)entry_SYSCALL_compat);
1346
	/*
1347 1348 1349 1350
	 * 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).
1351 1352
	 */
	wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS);
1353
	wrmsrl_safe(MSR_IA32_SYSENTER_ESP, (unsigned long)(cpu_entry_stack(cpu) + 1));
1354
	wrmsrl_safe(MSR_IA32_SYSENTER_EIP, (u64)entry_SYSENTER_compat);
1355
#else
1356
	wrmsrl(MSR_CSTAR, (unsigned long)ignore_sysret);
1357
	wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)GDT_ENTRY_INVALID_SEG);
1358 1359
	wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL);
	wrmsrl_safe(MSR_IA32_SYSENTER_EIP, 0ULL);
1360
#endif
1361

1362 1363
	/* Flags to clear on syscall */
	wrmsrl(MSR_SYSCALL_MASK,
1364
	       X86_EFLAGS_TF|X86_EFLAGS_DF|X86_EFLAGS_IF|
1365
	       X86_EFLAGS_IOPL|X86_EFLAGS_AC|X86_EFLAGS_NT);
L
Linus Torvalds 已提交
1366
}
1367

1368 1369 1370 1371 1372 1373
/*
 * 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);

1374
static DEFINE_PER_CPU(unsigned long, debug_stack_addr);
1375
DEFINE_PER_CPU(int, debug_stack_usage);
1376 1377 1378

int is_debug_stack(unsigned long addr)
{
1379 1380 1381
	return __this_cpu_read(debug_stack_usage) ||
		(addr <= __this_cpu_read(debug_stack_addr) &&
		 addr > (__this_cpu_read(debug_stack_addr) - DEBUG_STKSZ));
1382
}
1383
NOKPROBE_SYMBOL(is_debug_stack);
1384

1385
DEFINE_PER_CPU(u32, debug_idt_ctr);
1386

1387 1388
void debug_stack_set_zero(void)
{
1389 1390
	this_cpu_inc(debug_idt_ctr);
	load_current_idt();
1391
}
1392
NOKPROBE_SYMBOL(debug_stack_set_zero);
1393 1394 1395

void debug_stack_reset(void)
{
1396
	if (WARN_ON(!this_cpu_read(debug_idt_ctr)))
1397
		return;
1398 1399
	if (this_cpu_dec_return(debug_idt_ctr) == 0)
		load_current_idt();
1400
}
1401
NOKPROBE_SYMBOL(debug_stack_reset);
1402

I
Ingo Molnar 已提交
1403
#else	/* CONFIG_X86_64 */
1404

1405 1406
DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
EXPORT_PER_CPU_SYMBOL(current_task);
1407 1408
DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
EXPORT_PER_CPU_SYMBOL(__preempt_count);
1409

1410 1411 1412 1413 1414 1415 1416 1417 1418
/*
 * 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);

1419
#ifdef CONFIG_CC_STACKPROTECTOR
1420
DEFINE_PER_CPU_ALIGNED(struct stack_canary, stack_canary);
1421
#endif
1422

I
Ingo Molnar 已提交
1423
#endif	/* CONFIG_X86_64 */
1424

1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
/*
 * 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);
	}
}
1440

1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454
#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 */

1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
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
}

1468 1469 1470 1471 1472
/*
 * 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.
1473
 * A lot of state is already set up in PDA init for 64 bit
1474
 */
1475
#ifdef CONFIG_X86_64
I
Ingo Molnar 已提交
1476

1477
void cpu_init(void)
1478
{
1479
	struct orig_ist *oist;
1480
	struct task_struct *me;
I
Ingo Molnar 已提交
1481 1482
	struct tss_struct *t;
	unsigned long v;
1483
	int cpu = raw_smp_processor_id();
1484 1485
	int i;

1486 1487
	wait_for_master_cpu(cpu);

1488 1489 1490 1491 1492 1493
	/*
	 * Initialize the CR4 shadow before doing anything that could
	 * try to read it.
	 */
	cr4_init_shadow();

1494 1495
	if (cpu)
		load_ucode_ap();
1496

1497
	t = &per_cpu(cpu_tss_rw, cpu);
1498
	oist = &per_cpu(orig_ist, cpu);
I
Ingo Molnar 已提交
1499

1500
#ifdef CONFIG_NUMA
1501
	if (this_cpu_read(numa_node) == 0 &&
1502 1503
	    early_cpu_to_node(cpu) != NUMA_NO_NODE)
		set_numa_node(early_cpu_to_node(cpu));
1504
#endif
1505 1506 1507

	me = current;

1508
	pr_debug("Initializing CPU#%d\n", cpu);
1509

A
Andy Lutomirski 已提交
1510
	cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
1511 1512 1513 1514 1515 1516

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

1517
	switch_to_new_gdt(cpu);
1518 1519
	loadsegment(fs, 0);

1520
	load_current_idt();
1521 1522 1523 1524 1525 1526 1527 1528

	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();

1529
	x86_configure_nx();
1530
	x2apic_setup();
1531 1532 1533 1534

	/*
	 * set up and load the per-CPU TSS
	 */
1535
	if (!oist->ist[0]) {
1536
		char *estacks = get_cpu_entry_area(cpu)->exception_stacks;
I
Ingo Molnar 已提交
1537

1538
		for (v = 0; v < N_EXCEPTION_STACKS; v++) {
I
Ingo Molnar 已提交
1539
			estacks += exception_stack_sizes[v];
1540
			oist->ist[v] = t->x86_tss.ist[v] =
1541
					(unsigned long)estacks;
1542 1543
			if (v == DEBUG_STACK-1)
				per_cpu(debug_stack_addr, cpu) = (unsigned long)estacks;
1544 1545 1546
		}
	}

1547
	t->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET;
I
Ingo Molnar 已提交
1548

1549 1550 1551 1552 1553 1554 1555
	/*
	 * <= 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;

V
Vegard Nossum 已提交
1556
	mmgrab(&init_mm);
1557
	me->active_mm = &init_mm;
S
Stoyan Gaydarov 已提交
1558
	BUG_ON(me->mm);
1559
	initialize_tlbstate_and_flush();
1560 1561
	enter_lazy_tlb(&init_mm, me);

1562
	/*
1563 1564
	 * Initialize the TSS.  sp0 points to the entry trampoline stack
	 * regardless of what task is running.
1565
	 */
1566
	set_tss_desc(cpu, &get_cpu_entry_area(cpu)->tss.x86_tss);
1567
	load_TR_desc();
1568
	load_sp0((unsigned long)(cpu_entry_stack(cpu) + 1));
1569

1570
	load_mm_ldt(&init_mm);
1571

1572 1573
	clear_all_debug_regs();
	dbg_restore_debug_regs();
1574

I
Ingo Molnar 已提交
1575
	fpu__init_cpu();
1576 1577 1578

	if (is_uv_system())
		uv_cpu_init();
1579 1580

	load_fixmap_gdt(cpu);
1581 1582 1583 1584
}

#else

1585
void cpu_init(void)
1586
{
1587 1588
	int cpu = smp_processor_id();
	struct task_struct *curr = current;
1589
	struct tss_struct *t = &per_cpu(cpu_tss_rw, cpu);
1590

1591
	wait_for_master_cpu(cpu);
1592

1593 1594 1595 1596 1597 1598
	/*
	 * Initialize the CR4 shadow before doing anything that could
	 * try to read it.
	 */
	cr4_init_shadow();

1599
	show_ucode_info_early();
1600

1601
	pr_info("Initializing CPU#%d\n", cpu);
1602

1603
	if (cpu_feature_enabled(X86_FEATURE_VME) ||
1604
	    boot_cpu_has(X86_FEATURE_TSC) ||
1605
	    boot_cpu_has(X86_FEATURE_DE))
A
Andy Lutomirski 已提交
1606
		cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
1607

1608
	load_current_idt();
1609
	switch_to_new_gdt(cpu);
L
Linus Torvalds 已提交
1610 1611 1612 1613

	/*
	 * Set up and load the per-CPU TSS and LDT
	 */
V
Vegard Nossum 已提交
1614
	mmgrab(&init_mm);
1615
	curr->active_mm = &init_mm;
S
Stoyan Gaydarov 已提交
1616
	BUG_ON(curr->mm);
1617
	initialize_tlbstate_and_flush();
1618
	enter_lazy_tlb(&init_mm, curr);
L
Linus Torvalds 已提交
1619

1620 1621 1622 1623
	/*
	 * Initialize the TSS.  Don't bother initializing sp0, as the initial
	 * task never enters user mode.
	 */
1624
	set_tss_desc(cpu, &get_cpu_entry_area(cpu)->tss.x86_tss);
L
Linus Torvalds 已提交
1625
	load_TR_desc();
1626

1627
	load_mm_ldt(&init_mm);
L
Linus Torvalds 已提交
1628

1629
	t->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET;
1630

1631
#ifdef CONFIG_DOUBLEFAULT
L
Linus Torvalds 已提交
1632 1633
	/* Set up doublefault TSS pointer in the GDT */
	__set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss);
1634
#endif
L
Linus Torvalds 已提交
1635

1636
	clear_all_debug_regs();
1637
	dbg_restore_debug_regs();
L
Linus Torvalds 已提交
1638

I
Ingo Molnar 已提交
1639
	fpu__init_cpu();
1640 1641

	load_fixmap_gdt(cpu);
L
Linus Torvalds 已提交
1642
}
1643
#endif
1644

1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
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);