internal.h 16.1 KB
Newer Older
1
/* SPDX-License-Identifier: GPL-2.0 */
2 3 4 5 6 7 8 9 10
/*
 * Copyright (C) 1994 Linus Torvalds
 *
 * Pentium III FXSR, SSE support
 * General FPU state handling cleanups
 *	Gareth Hughes <gareth@valinux.com>, May 2000
 * x86-64 work by Andi Kleen 2002
 */

11 12
#ifndef _ASM_X86_FPU_INTERNAL_H
#define _ASM_X86_FPU_INTERNAL_H
13

14
#include <linux/compat.h>
15
#include <linux/sched.h>
16
#include <linux/slab.h>
17

18
#include <asm/user.h>
19
#include <asm/fpu/api.h>
20
#include <asm/fpu/xstate.h>
21
#include <asm/cpufeature.h>
22
#include <asm/trace/fpu.h>
23

24 25 26
/*
 * High level FPU state handling functions:
 */
27
extern void fpu__initialize(struct fpu *fpu);
28 29
extern void fpu__prepare_read(struct fpu *fpu);
extern void fpu__prepare_write(struct fpu *fpu);
30
extern void fpu__save(struct fpu *fpu);
31
extern void fpu__restore(struct fpu *fpu);
32
extern int  fpu__restore_sig(void __user *buf, int ia32_frame);
33 34
extern void fpu__drop(struct fpu *fpu);
extern int  fpu__copy(struct fpu *dst_fpu, struct fpu *src_fpu);
35
extern void fpu__clear(struct fpu *fpu);
I
Ingo Molnar 已提交
36 37
extern int  fpu__exception_code(struct fpu *fpu, int trap_nr);
extern int  dump_fpu(struct pt_regs *ptregs, struct user_i387_struct *fpstate);
38

I
Ingo Molnar 已提交
39 40 41 42 43 44 45
/*
 * Boot time FPU initialization functions:
 */
extern void fpu__init_cpu(void);
extern void fpu__init_system_xstate(void);
extern void fpu__init_cpu_xstate(void);
extern void fpu__init_system(struct cpuinfo_x86 *c);
46 47
extern void fpu__init_check_bugs(void);
extern void fpu__resume_cpu(void);
48
extern u64 fpu__get_supported_xfeatures_mask(void);
49

50 51 52 53 54 55
/*
 * Debugging facility:
 */
#ifdef CONFIG_X86_DEBUG_FPU
# define WARN_ON_FPU(x) WARN_ON_ONCE(x)
#else
56
# define WARN_ON_FPU(x) ({ (void)(x); 0; })
57 58
#endif

59
/*
I
Ingo Molnar 已提交
60
 * FPU related CPU feature flag helper routines:
61
 */
62 63
static __always_inline __pure bool use_xsaveopt(void)
{
64
	return static_cpu_has(X86_FEATURE_XSAVEOPT);
65 66 67 68
}

static __always_inline __pure bool use_xsave(void)
{
69
	return static_cpu_has(X86_FEATURE_XSAVE);
70 71 72 73
}

static __always_inline __pure bool use_fxsr(void)
{
74
	return static_cpu_has(X86_FEATURE_FXSR);
75 76
}

I
Ingo Molnar 已提交
77 78 79 80 81 82 83 84 85 86 87 88
/*
 * fpstate handling functions:
 */

extern union fpregs_state init_fpstate;

extern void fpstate_init(union fpregs_state *state);
#ifdef CONFIG_MATH_EMULATION
extern void fpstate_init_soft(struct swregs_state *soft);
#else
static inline void fpstate_init_soft(struct swregs_state *soft) {}
#endif
89 90 91 92 93 94 95 96 97 98

static inline void fpstate_init_xstate(struct xregs_state *xsave)
{
	/*
	 * XRSTORS requires these bits set in xcomp_bv, or it will
	 * trigger #GP:
	 */
	xsave->header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT | xfeatures_mask;
}

I
Ingo Molnar 已提交
99 100 101 102 103
static inline void fpstate_init_fxstate(struct fxregs_state *fx)
{
	fx->cwd = 0x37f;
	fx->mxcsr = MXCSR_DEFAULT;
}
104
extern void fpstate_sanitize_xstate(struct fpu *fpu);
105

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
#define user_insn(insn, output, input...)				\
({									\
	int err;							\
	asm volatile(ASM_STAC "\n"					\
		     "1:" #insn "\n\t"					\
		     "2: " ASM_CLAC "\n"				\
		     ".section .fixup,\"ax\"\n"				\
		     "3:  movl $-1,%[err]\n"				\
		     "    jmp  2b\n"					\
		     ".previous\n"					\
		     _ASM_EXTABLE(1b, 3b)				\
		     : [err] "=r" (err), output				\
		     : "0"(0), input);					\
	err;								\
})

122
#define kernel_insn(insn, output, input...)				\
123 124
	asm volatile("1:" #insn "\n\t"					\
		     "2:\n"						\
125 126
		     _ASM_EXTABLE_HANDLE(1b, 2b, ex_handler_fprestore)	\
		     : output : input)
127

128
static inline int copy_fregs_to_user(struct fregs_state __user *fx)
129
{
130
	return user_insn(fnsave %[fx]; fwait,  [fx] "=m" (*fx), "m" (*fx));
131 132
}

133
static inline int copy_fxregs_to_user(struct fxregs_state __user *fx)
134
{
135
	if (IS_ENABLED(CONFIG_X86_32))
136
		return user_insn(fxsave %[fx], [fx] "=m" (*fx), "m" (*fx));
137
	else if (IS_ENABLED(CONFIG_AS_FXSAVEQ))
138
		return user_insn(fxsaveq %[fx], [fx] "=m" (*fx), "m" (*fx));
139

140
	/* See comment in copy_fxregs_to_kernel() below. */
141
	return user_insn(rex64/fxsave (%[fx]), "=m" (*fx), [fx] "R" (fx));
142 143
}

144
static inline void copy_kernel_to_fxregs(struct fxregs_state *fx)
145
{
146
	if (IS_ENABLED(CONFIG_X86_32)) {
147
		kernel_insn(fxrstor %[fx], "=m" (*fx), [fx] "m" (*fx));
148
	} else {
149
		if (IS_ENABLED(CONFIG_AS_FXSAVEQ)) {
150
			kernel_insn(fxrstorq %[fx], "=m" (*fx), [fx] "m" (*fx));
151 152
		} else {
			/* See comment in copy_fxregs_to_kernel() below. */
153
			kernel_insn(rex64/fxrstor (%[fx]), "=m" (*fx), [fx] "R" (fx), "m" (*fx));
154 155
		}
	}
156 157
}

158
static inline int copy_user_to_fxregs(struct fxregs_state __user *fx)
159
{
160
	if (IS_ENABLED(CONFIG_X86_32))
161
		return user_insn(fxrstor %[fx], "=m" (*fx), [fx] "m" (*fx));
162
	else if (IS_ENABLED(CONFIG_AS_FXSAVEQ))
163 164
		return user_insn(fxrstorq %[fx], "=m" (*fx), [fx] "m" (*fx));

165
	/* See comment in copy_fxregs_to_kernel() below. */
166 167 168 169
	return user_insn(rex64/fxrstor (%[fx]), "=m" (*fx), [fx] "R" (fx),
			  "m" (*fx));
}

170
static inline void copy_kernel_to_fregs(struct fregs_state *fx)
171
{
172
	kernel_insn(frstor %[fx], "=m" (*fx), [fx] "m" (*fx));
173 174
}

175
static inline int copy_user_to_fregs(struct fregs_state __user *fx)
176 177
{
	return user_insn(frstor %[fx], "=m" (*fx), [fx] "m" (*fx));
178 179
}

180
static inline void copy_fxregs_to_kernel(struct fpu *fpu)
181
{
182
	if (IS_ENABLED(CONFIG_X86_32))
183
		asm volatile( "fxsave %[fx]" : [fx] "=m" (fpu->state.fxsave));
184
	else if (IS_ENABLED(CONFIG_AS_FXSAVEQ))
185
		asm volatile("fxsaveq %[fx]" : [fx] "=m" (fpu->state.fxsave));
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
	else {
		/* Using "rex64; fxsave %0" is broken because, if the memory
		 * operand uses any extended registers for addressing, a second
		 * REX prefix will be generated (to the assembler, rex64
		 * followed by semicolon is a separate instruction), and hence
		 * the 64-bitness is lost.
		 *
		 * Using "fxsaveq %0" would be the ideal choice, but is only
		 * supported starting with gas 2.16.
		 *
		 * Using, as a workaround, the properly prefixed form below
		 * isn't accepted by any binutils version so far released,
		 * complaining that the same type of prefix is used twice if
		 * an extended register is needed for addressing (fix submitted
		 * to mainline 2005-11-21).
		 *
202
		 *  asm volatile("rex64/fxsave %0" : "=m" (fpu->state.fxsave));
203 204 205 206 207 208
		 *
		 * This, however, we can work around by forcing the compiler to
		 * select an addressing mode that doesn't require extended
		 * registers.
		 */
		asm volatile( "rex64/fxsave (%[fx])"
209 210
			     : "=m" (fpu->state.fxsave)
			     : [fx] "R" (&fpu->state.fxsave));
211
	}
212 213
}

214 215 216 217 218 219 220
/* These macros all use (%edi)/(%rdi) as the single memory argument. */
#define XSAVE		".byte " REX_PREFIX "0x0f,0xae,0x27"
#define XSAVEOPT	".byte " REX_PREFIX "0x0f,0xae,0x37"
#define XSAVES		".byte " REX_PREFIX "0x0f,0xc7,0x2f"
#define XRSTOR		".byte " REX_PREFIX "0x0f,0xae,0x2f"
#define XRSTORS		".byte " REX_PREFIX "0x0f,0xc7,0x1f"

221 222 223 224 225 226 227 228
#define XSTATE_OP(op, st, lmask, hmask, err)				\
	asm volatile("1:" op "\n\t"					\
		     "xor %[err], %[err]\n"				\
		     "2:\n\t"						\
		     ".pushsection .fixup,\"ax\"\n\t"			\
		     "3: movl $-2,%[err]\n\t"				\
		     "jmp 2b\n\t"					\
		     ".popsection\n\t"					\
229
		     _ASM_EXTABLE_UA(1b, 3b)				\
230 231 232 233
		     : [err] "=r" (err)					\
		     : "D" (st), "m" (*st), "a" (lmask), "d" (hmask)	\
		     : "memory")

234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
/*
 * If XSAVES is enabled, it replaces XSAVEOPT because it supports a compact
 * format and supervisor states in addition to modified optimization in
 * XSAVEOPT.
 *
 * Otherwise, if XSAVEOPT is enabled, XSAVEOPT replaces XSAVE because XSAVEOPT
 * supports modified optimization which is not supported by XSAVE.
 *
 * We use XSAVE as a fallback.
 *
 * The 661 label is defined in the ALTERNATIVE* macros as the address of the
 * original instruction which gets replaced. We need to use it here as the
 * address of the instruction where we might get an exception at.
 */
#define XSTATE_XSAVE(st, lmask, hmask, err)				\
	asm volatile(ALTERNATIVE_2(XSAVE,				\
				   XSAVEOPT, X86_FEATURE_XSAVEOPT,	\
				   XSAVES,   X86_FEATURE_XSAVES)	\
		     "\n"						\
		     "xor %[err], %[err]\n"				\
		     "3:\n"						\
		     ".pushsection .fixup,\"ax\"\n"			\
		     "4: movl $-2, %[err]\n"				\
		     "jmp 3b\n"						\
		     ".popsection\n"					\
		     _ASM_EXTABLE(661b, 4b)				\
		     : [err] "=r" (err)					\
		     : "D" (st), "m" (*st), "a" (lmask), "d" (hmask)	\
		     : "memory")

/*
 * Use XRSTORS to restore context if it is enabled. XRSTORS supports compact
 * XSAVE area format.
 */
268
#define XSTATE_XRESTORE(st, lmask, hmask)				\
269 270 271 272
	asm volatile(ALTERNATIVE(XRSTOR,				\
				 XRSTORS, X86_FEATURE_XSAVES)		\
		     "\n"						\
		     "3:\n"						\
273 274
		     _ASM_EXTABLE_HANDLE(661b, 3b, ex_handler_fprestore)\
		     :							\
275 276
		     : "D" (st), "m" (*st), "a" (lmask), "d" (hmask)	\
		     : "memory")
277

278 279 280 281
/*
 * This function is called only during boot time when x86 caps are not set
 * up and alternative can not be used yet.
 */
282
static inline void copy_xregs_to_kernel_booting(struct xregs_state *xstate)
283 284 285 286
{
	u64 mask = -1;
	u32 lmask = mask;
	u32 hmask = mask >> 32;
287
	int err;
288 289 290

	WARN_ON(system_state != SYSTEM_BOOTING);

291
	if (static_cpu_has(X86_FEATURE_XSAVES))
292
		XSTATE_OP(XSAVES, xstate, lmask, hmask, err);
293
	else
294
		XSTATE_OP(XSAVE, xstate, lmask, hmask, err);
295 296 297

	/* We should never fault when copying to a kernel buffer: */
	WARN_ON_FPU(err);
298 299 300 301 302 303
}

/*
 * This function is called only during boot time when x86 caps are not set
 * up and alternative can not be used yet.
 */
304
static inline void copy_kernel_to_xregs_booting(struct xregs_state *xstate)
305
{
306
	u64 mask = -1;
307 308
	u32 lmask = mask;
	u32 hmask = mask >> 32;
309
	int err;
310 311 312

	WARN_ON(system_state != SYSTEM_BOOTING);

313
	if (static_cpu_has(X86_FEATURE_XSAVES))
314
		XSTATE_OP(XRSTORS, xstate, lmask, hmask, err);
315
	else
316
		XSTATE_OP(XRSTOR, xstate, lmask, hmask, err);
317

318 319 320 321
	/*
	 * We should never fault when copying from a kernel buffer, and the FPU
	 * state we set at boot time should be valid.
	 */
322
	WARN_ON_FPU(err);
323 324 325 326 327
}

/*
 * Save processor xstate to xsave area.
 */
328
static inline void copy_xregs_to_kernel(struct xregs_state *xstate)
329 330 331 332
{
	u64 mask = -1;
	u32 lmask = mask;
	u32 hmask = mask >> 32;
333
	int err;
334

335
	WARN_ON_FPU(!alternatives_patched);
336

337
	XSTATE_XSAVE(xstate, lmask, hmask, err);
338

339 340
	/* We should never fault when copying to a kernel buffer: */
	WARN_ON_FPU(err);
341 342 343 344 345
}

/*
 * Restore processor xstate from xsave area.
 */
346
static inline void copy_kernel_to_xregs(struct xregs_state *xstate, u64 mask)
347 348 349 350
{
	u32 lmask = mask;
	u32 hmask = mask >> 32;

351
	XSTATE_XRESTORE(xstate, lmask, hmask);
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375
}

/*
 * Save xstate to user space xsave area.
 *
 * We don't use modified optimization because xrstor/xrstors might track
 * a different application.
 *
 * We don't use compacted format xsave area for
 * backward compatibility for old applications which don't understand
 * compacted format of xsave area.
 */
static inline int copy_xregs_to_user(struct xregs_state __user *buf)
{
	int err;

	/*
	 * Clear the xsave header first, so that reserved fields are
	 * initialized to zero.
	 */
	err = __clear_user(&buf->header, sizeof(buf->header));
	if (unlikely(err))
		return -EFAULT;

376 377 378 379
	stac();
	XSTATE_OP(XSAVE, buf, -1, -1, err);
	clac();

380 381 382 383 384 385 386 387 388 389 390
	return err;
}

/*
 * Restore xstate from user space xsave area.
 */
static inline int copy_user_to_xregs(struct xregs_state __user *buf, u64 mask)
{
	struct xregs_state *xstate = ((__force struct xregs_state *)buf);
	u32 lmask = mask;
	u32 hmask = mask >> 32;
391 392 393 394 395
	int err;

	stac();
	XSTATE_OP(XRSTOR, xstate, lmask, hmask, err);
	clac();
396 397 398 399

	return err;
}

400 401
/*
 * These must be called with preempt disabled. Returns
402 403 404 405 406 407
 * 'true' if the FPU state is still intact and we can
 * keep registers active.
 *
 * The legacy FNSAVE instruction cleared all FPU state
 * unconditionally, so registers are essentially destroyed.
 * Modern FPU state can be kept in registers, if there are
408
 * no pending FP exceptions.
409
 */
410
static inline int copy_fpregs_to_fpstate(struct fpu *fpu)
411
{
412
	if (likely(use_xsave())) {
413
		copy_xregs_to_kernel(&fpu->state.xsave);
414 415
		return 1;
	}
416

417
	if (likely(use_fxsr())) {
418
		copy_fxregs_to_kernel(fpu);
419
		return 1;
420 421 422
	}

	/*
423 424
	 * Legacy FPU register saving, FNSAVE always clears FPU registers,
	 * so we have to mark them inactive:
425
	 */
426
	asm volatile("fnsave %[fp]; fwait" : [fp] "=m" (fpu->state.fsave));
427 428

	return 0;
429 430
}

431
static inline void __copy_kernel_to_fpregs(union fpregs_state *fpstate, u64 mask)
432
{
433
	if (use_xsave()) {
434
		copy_kernel_to_xregs(&fpstate->xsave, mask);
435 436
	} else {
		if (use_fxsr())
437
			copy_kernel_to_fxregs(&fpstate->fxsave);
438
		else
439
			copy_kernel_to_fregs(&fpstate->fsave);
440
	}
441 442
}

443
static inline void copy_kernel_to_fpregs(union fpregs_state *fpstate)
444
{
445 446 447 448 449
	/*
	 * AMD K7/K8 CPUs don't save/restore FDP/FIP/FOP unless an exception is
	 * pending. Clear the x87 state here by setting it to fixed values.
	 * "m" is a random variable that should be in L1.
	 */
450
	if (unlikely(static_cpu_has_bug(X86_BUG_FXSAVE_LEAK))) {
451 452 453 454
		asm volatile(
			"fnclex\n\t"
			"emms\n\t"
			"fildl %P[addr]"	/* set F?P to defined value */
455
			: : [addr] "m" (fpstate));
456
	}
457

458
	__copy_kernel_to_fpregs(fpstate, -1);
459 460
}

461
extern int copy_fpstate_to_sigframe(void __user *buf, void __user *fp, int size);
I
Ingo Molnar 已提交
462 463 464 465 466 467 468 469

/*
 * FPU context switch related helper methods:
 */

DECLARE_PER_CPU(struct fpu *, fpu_fpregs_owner_ctx);

/*
470 471 472 473 474 475 476 477 478
 * The in-register FPU state for an FPU context on a CPU is assumed to be
 * valid if the fpu->last_cpu matches the CPU, and the fpu_fpregs_owner_ctx
 * matches the FPU.
 *
 * If the FPU register state is valid, the kernel can skip restoring the
 * FPU state from memory.
 *
 * Any code that clobbers the FPU registers or updates the in-memory
 * FPU state for a task MUST let the rest of the kernel know that the
479
 * FPU registers are no longer valid for this task.
480
 *
481 482 483 484
 * Either one of these invalidation functions is enough. Invalidate
 * a resource you control: CPU if using the CPU for something else
 * (with preemption disabled), FPU for the current task, or a task that
 * is prevented from running by the current task.
I
Ingo Molnar 已提交
485
 */
486
static inline void __cpu_invalidate_fpregs_state(void)
I
Ingo Molnar 已提交
487
{
488
	__this_cpu_write(fpu_fpregs_owner_ctx, NULL);
I
Ingo Molnar 已提交
489 490
}

491 492 493 494 495 496
static inline void __fpu_invalidate_fpregs_state(struct fpu *fpu)
{
	fpu->last_cpu = -1;
}

static inline int fpregs_state_valid(struct fpu *fpu, unsigned int cpu)
I
Ingo Molnar 已提交
497 498 499 500
{
	return fpu == this_cpu_read_stable(fpu_fpregs_owner_ctx) && cpu == fpu->last_cpu;
}

501 502 503 504 505
/*
 * These generally need preemption protection to work,
 * do try to avoid using these on their own:
 */
static inline void fpregs_deactivate(struct fpu *fpu)
506
{
507
	this_cpu_write(fpu_fpregs_owner_ctx, NULL);
508
	trace_x86_fpu_regs_deactivated(fpu);
509 510
}

511
static inline void fpregs_activate(struct fpu *fpu)
512
{
513
	this_cpu_write(fpu_fpregs_owner_ctx, fpu);
514
	trace_x86_fpu_regs_activated(fpu);
515 516 517 518 519 520 521
}

/*
 * FPU state switching for scheduling.
 *
 * This is a two-stage process:
 *
522 523
 *  - switch_fpu_prepare() saves the old state.
 *    This is done within the context of the old process.
524 525 526 527
 *
 *  - switch_fpu_finish() restores the new state as
 *    necessary.
 */
528 529
static inline void
switch_fpu_prepare(struct fpu *old_fpu, int cpu)
530
{
531
	if (static_cpu_has(X86_FEATURE_FPU) && old_fpu->initialized) {
532
		if (!copy_fpregs_to_fpstate(old_fpu))
533
			old_fpu->last_cpu = -1;
534
		else
535
			old_fpu->last_cpu = cpu;
536

537
		/* But leave fpu_fpregs_owner_ctx! */
538
		trace_x86_fpu_regs_deactivated(old_fpu);
539 540
	} else
		old_fpu->last_cpu = -1;
541 542
}

I
Ingo Molnar 已提交
543 544 545 546
/*
 * Misc helper functions:
 */

547
/*
548 549
 * Set up the userspace FPU context for the new task, if the task
 * has used the FPU.
550
 */
551
static inline void switch_fpu_finish(struct fpu *new_fpu, int cpu)
552
{
553
	bool preload = static_cpu_has(X86_FEATURE_FPU) &&
554
		       new_fpu->initialized;
555 556 557 558 559 560

	if (preload) {
		if (!fpregs_state_valid(new_fpu, cpu))
			copy_kernel_to_fpregs(&new_fpu->state);
		fpregs_activate(new_fpu);
	}
561 562 563
}

/*
564
 * Needs to be preemption-safe.
565
 *
566
 * NOTE! user_fpu_begin() must be used only immediately before restoring
567 568 569
 * the save state. It does not do any saving/restoring on its own. In
 * lazy FPU mode, it is just an optimization to avoid a #NM exception,
 * the task can lose the FPU right after preempt_enable().
570 571 572
 */
static inline void user_fpu_begin(void)
{
573 574
	struct fpu *fpu = &current->thread.fpu;

575
	preempt_disable();
576
	fpregs_activate(fpu);
577 578 579
	preempt_enable();
}

I
Ingo Molnar 已提交
580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606
/*
 * MXCSR and XCR definitions:
 */

extern unsigned int mxcsr_feature_mask;

#define XCR_XFEATURE_ENABLED_MASK	0x00000000

static inline u64 xgetbv(u32 index)
{
	u32 eax, edx;

	asm volatile(".byte 0x0f,0x01,0xd0" /* xgetbv */
		     : "=a" (eax), "=d" (edx)
		     : "c" (index));
	return eax + ((u64)edx << 32);
}

static inline void xsetbv(u32 index, u64 value)
{
	u32 eax = value;
	u32 edx = value >> 32;

	asm volatile(".byte 0x0f,0x01,0xd1" /* xsetbv */
		     : : "a" (eax), "d" (edx), "c" (index));
}

607
#endif /* _ASM_X86_FPU_INTERNAL_H */