internal.h 14.8 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>
R
Rik van Riel 已提交
17
#include <linux/mm.h>
18

19
#include <asm/user.h>
20
#include <asm/fpu/api.h>
21
#include <asm/fpu/xstate.h>
22
#include <asm/fpu/xcr.h>
23
#include <asm/cpufeature.h>
24
#include <asm/trace/fpu.h>
25

26 27 28 29
/*
 * High level FPU state handling functions:
 */
extern void fpu__save(struct fpu *fpu);
30
extern int  fpu__restore_sig(void __user *buf, int ia32_frame);
31
extern void fpu__drop(struct fpu *fpu);
32
extern int  fpu__copy(struct task_struct *dst, struct task_struct *src);
33 34
extern void fpu__clear_user_states(struct fpu *fpu);
extern void fpu__clear_all(struct fpu *fpu);
I
Ingo Molnar 已提交
35
extern int  fpu__exception_code(struct fpu *fpu, int trap_nr);
36

I
Ingo Molnar 已提交
37 38 39 40 41 42 43
/*
 * 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);
44 45 46
extern void fpu__init_check_bugs(void);
extern void fpu__resume_cpu(void);

47 48 49 50 51 52
/*
 * Debugging facility:
 */
#ifdef CONFIG_X86_DEBUG_FPU
# define WARN_ON_FPU(x) WARN_ON_ONCE(x)
#else
53
# define WARN_ON_FPU(x) ({ (void)(x); 0; })
54 55
#endif

56
/*
I
Ingo Molnar 已提交
57
 * FPU related CPU feature flag helper routines:
58
 */
59 60
static __always_inline __pure bool use_xsaveopt(void)
{
61
	return static_cpu_has(X86_FEATURE_XSAVEOPT);
62 63 64 65
}

static __always_inline __pure bool use_xsave(void)
{
66
	return static_cpu_has(X86_FEATURE_XSAVE);
67 68 69 70
}

static __always_inline __pure bool use_fxsr(void)
{
71
	return static_cpu_has(X86_FEATURE_FXSR);
72 73
}

I
Ingo Molnar 已提交
74 75 76 77 78 79 80 81 82 83 84 85
/*
 * 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
86

87 88 89
#define user_insn(insn, output, input...)				\
({									\
	int err;							\
90 91 92
									\
	might_fault();							\
									\
93 94 95 96 97 98 99 100 101 102 103 104 105
	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;								\
})

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
#define kernel_insn_err(insn, output, input...)				\
({									\
	int err;							\
	asm volatile("1:" #insn "\n\t"					\
		     "2:\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;								\
})

121
#define kernel_insn(insn, output, input...)				\
122 123
	asm volatile("1:" #insn "\n\t"					\
		     "2:\n"						\
124 125
		     _ASM_EXTABLE_HANDLE(1b, 2b, ex_handler_fprestore)	\
		     : output : input)
126

127
static inline int fnsave_to_user_sigframe(struct fregs_state __user *fx)
128
{
129
	return user_insn(fnsave %[fx]; fwait,  [fx] "=m" (*fx), "m" (*fx));
130 131
}

132
static inline int fxsave_to_user_sigframe(struct fxregs_state __user *fx)
133
{
134
	if (IS_ENABLED(CONFIG_X86_32))
135
		return user_insn(fxsave %[fx], [fx] "=m" (*fx), "m" (*fx));
136
	else
137
		return user_insn(fxsaveq %[fx], [fx] "=m" (*fx), "m" (*fx));
138 139 140

}

141
static inline void fxrstor(struct fxregs_state *fx)
142
{
143
	if (IS_ENABLED(CONFIG_X86_32))
144
		kernel_insn(fxrstor %[fx], "=m" (*fx), [fx] "m" (*fx));
145 146
	else
		kernel_insn(fxrstorq %[fx], "=m" (*fx), [fx] "m" (*fx));
147 148
}

149
static inline int fxrstor_safe(struct fxregs_state *fx)
150 151 152 153 154 155 156
{
	if (IS_ENABLED(CONFIG_X86_32))
		return kernel_insn_err(fxrstor %[fx], "=m" (*fx), [fx] "m" (*fx));
	else
		return kernel_insn_err(fxrstorq %[fx], "=m" (*fx), [fx] "m" (*fx));
}

157
static inline int fxrstor_from_user_sigframe(struct fxregs_state __user *fx)
158
{
159
	if (IS_ENABLED(CONFIG_X86_32))
160
		return user_insn(fxrstor %[fx], "=m" (*fx), [fx] "m" (*fx));
161
	else
162 163 164
		return user_insn(fxrstorq %[fx], "=m" (*fx), [fx] "m" (*fx));
}

165
static inline void frstor(struct fregs_state *fx)
166
{
167
	kernel_insn(frstor %[fx], "=m" (*fx), [fx] "m" (*fx));
168 169
}

170
static inline int frstor_safe(struct fregs_state *fx)
171 172 173 174
{
	return kernel_insn_err(frstor %[fx], "=m" (*fx), [fx] "m" (*fx));
}

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

180 181 182 183 184 185 186 187
static inline void fxsave(struct fxregs_state *fx)
{
	if (IS_ENABLED(CONFIG_X86_32))
		asm volatile( "fxsave %[fx]" : [fx] "=m" (*fx));
	else
		asm volatile("fxsaveq %[fx]" : [fx] "=m" (*fx));
}

188 189 190 191 192 193 194
/* 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"

195 196 197 198 199 200 201 202
#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"					\
203
		     _ASM_EXTABLE(1b, 3b)				\
204 205 206 207
		     : [err] "=r" (err)					\
		     : "D" (st), "m" (*st), "a" (lmask), "d" (hmask)	\
		     : "memory")

208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
/*
 * 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.
 */
242
#define XSTATE_XRESTORE(st, lmask, hmask)				\
243 244 245 246
	asm volatile(ALTERNATIVE(XRSTOR,				\
				 XRSTORS, X86_FEATURE_XSAVES)		\
		     "\n"						\
		     "3:\n"						\
247 248
		     _ASM_EXTABLE_HANDLE(661b, 3b, ex_handler_fprestore)\
		     :							\
249 250
		     : "D" (st), "m" (*st), "a" (lmask), "d" (hmask)	\
		     : "memory")
251

252 253 254 255
/*
 * This function is called only during boot time when x86 caps are not set
 * up and alternative can not be used yet.
 */
256
static inline void os_xrstor_booting(struct xregs_state *xstate)
257
{
258
	u64 mask = -1;
259 260
	u32 lmask = mask;
	u32 hmask = mask >> 32;
261
	int err;
262 263 264

	WARN_ON(system_state != SYSTEM_BOOTING);

265
	if (boot_cpu_has(X86_FEATURE_XSAVES))
266
		XSTATE_OP(XRSTORS, xstate, lmask, hmask, err);
267
	else
268
		XSTATE_OP(XRSTOR, xstate, lmask, hmask, err);
269

270 271 272 273
	/*
	 * We should never fault when copying from a kernel buffer, and the FPU
	 * state we set at boot time should be valid.
	 */
274
	WARN_ON_FPU(err);
275 276 277 278
}

/*
 * Save processor xstate to xsave area.
279 280 281
 *
 * Uses either XSAVE or XSAVEOPT or XSAVES depending on the CPU features
 * and command line options. The choice is permanent until the next reboot.
282
 */
283
static inline void os_xsave(struct xregs_state *xstate)
284
{
285
	u64 mask = xfeatures_mask_all;
286 287
	u32 lmask = mask;
	u32 hmask = mask >> 32;
288
	int err;
289

290
	WARN_ON_FPU(!alternatives_patched);
291

292
	XSTATE_XSAVE(xstate, lmask, hmask, err);
293

294 295
	/* We should never fault when copying to a kernel buffer: */
	WARN_ON_FPU(err);
296 297 298 299
}

/*
 * Restore processor xstate from xsave area.
300 301
 *
 * Uses XRSTORS when XSAVES is used, XRSTOR otherwise.
302
 */
303
static inline void os_xrstor(struct xregs_state *xstate, u64 mask)
304 305 306 307
{
	u32 lmask = mask;
	u32 hmask = mask >> 32;

308
	XSTATE_XRESTORE(xstate, lmask, hmask);
309 310 311 312 313 314 315 316 317 318 319 320
}

/*
 * 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.
 */
321
static inline int xsave_to_user_sigframe(struct xregs_state __user *buf)
322
{
323 324 325
	u64 mask = xfeatures_mask_user();
	u32 lmask = mask;
	u32 hmask = mask >> 32;
326 327 328 329 330 331 332 333 334 335
	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;

336
	stac();
337
	XSTATE_OP(XSAVE, buf, lmask, hmask, err);
338 339
	clac();

340 341 342 343 344 345
	return err;
}

/*
 * Restore xstate from user space xsave area.
 */
346
static inline int xrstor_from_user_sigframe(struct xregs_state __user *buf, u64 mask)
347 348 349 350
{
	struct xregs_state *xstate = ((__force struct xregs_state *)buf);
	u32 lmask = mask;
	u32 hmask = mask >> 32;
351 352 353 354 355
	int err;

	stac();
	XSTATE_OP(XRSTOR, xstate, lmask, hmask, err);
	clac();
356 357 358 359

	return err;
}

360 361 362 363
/*
 * Restore xstate from kernel space xsave area, return an error code instead of
 * an exception.
 */
364
static inline int os_xrstor_safe(struct xregs_state *xstate, u64 mask)
365 366 367 368 369
{
	u32 lmask = mask;
	u32 hmask = mask >> 32;
	int err;

370
	if (cpu_feature_enabled(X86_FEATURE_XSAVES))
371 372 373
		XSTATE_OP(XRSTORS, xstate, lmask, hmask, err);
	else
		XSTATE_OP(XRSTOR, xstate, lmask, hmask, err);
374 375 376 377

	return err;
}

378
extern int save_fpregs_to_fpstate(struct fpu *fpu);
379

380
static inline void __copy_kernel_to_fpregs(union fpregs_state *fpstate, u64 mask)
381
{
382
	if (use_xsave()) {
383
		os_xrstor(&fpstate->xsave, mask);
384 385
	} else {
		if (use_fxsr())
386
			fxrstor(&fpstate->fxsave);
387
		else
388
			frstor(&fpstate->fsave);
389
	}
390 391
}

392
static inline void copy_kernel_to_fpregs(union fpregs_state *fpstate)
393
{
394 395 396 397 398
	/*
	 * 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.
	 */
399
	if (unlikely(static_cpu_has_bug(X86_BUG_FXSAVE_LEAK))) {
400 401 402 403
		asm volatile(
			"fnclex\n\t"
			"emms\n\t"
			"fildl %P[addr]"	/* set F?P to defined value */
404
			: : [addr] "m" (fpstate));
405
	}
406

407
	__copy_kernel_to_fpregs(fpstate, -1);
408 409
}

410
extern int copy_fpstate_to_sigframe(void __user *buf, void __user *fp, int size);
I
Ingo Molnar 已提交
411 412 413 414 415 416 417 418

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

DECLARE_PER_CPU(struct fpu *, fpu_fpregs_owner_ctx);

/*
419 420 421 422 423 424 425 426 427
 * 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
428
 * FPU registers are no longer valid for this task.
429
 *
430 431 432 433
 * 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
434
 */
435
static inline void __cpu_invalidate_fpregs_state(void)
I
Ingo Molnar 已提交
436
{
437
	__this_cpu_write(fpu_fpregs_owner_ctx, NULL);
I
Ingo Molnar 已提交
438 439
}

440 441 442 443 444 445
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
446
{
447
	return fpu == this_cpu_read(fpu_fpregs_owner_ctx) && cpu == fpu->last_cpu;
I
Ingo Molnar 已提交
448 449
}

450 451 452 453 454
/*
 * These generally need preemption protection to work,
 * do try to avoid using these on their own:
 */
static inline void fpregs_deactivate(struct fpu *fpu)
455
{
456
	this_cpu_write(fpu_fpregs_owner_ctx, NULL);
457
	trace_x86_fpu_regs_deactivated(fpu);
458 459
}

460
static inline void fpregs_activate(struct fpu *fpu)
461
{
462
	this_cpu_write(fpu_fpregs_owner_ctx, fpu);
463
	trace_x86_fpu_regs_activated(fpu);
464 465
}

466 467 468
/*
 * Internal helper, do not use directly. Use switch_fpu_return() instead.
 */
469
static inline void __fpregs_load_activate(void)
470
{
471 472 473
	struct fpu *fpu = &current->thread.fpu;
	int cpu = smp_processor_id();

474
	if (WARN_ON_ONCE(current->flags & PF_KTHREAD))
475 476
		return;

477
	if (!fpregs_state_valid(fpu, cpu)) {
478
		copy_kernel_to_fpregs(&fpu->state);
479
		fpregs_activate(fpu);
480
		fpu->last_cpu = cpu;
481
	}
482
	clear_thread_flag(TIF_NEED_FPU_LOAD);
483 484
}

485 486 487 488 489
/*
 * FPU state switching for scheduling.
 *
 * This is a two-stage process:
 *
490 491
 *  - switch_fpu_prepare() saves the old state.
 *    This is done within the context of the old process.
492
 *
493 494
 *  - switch_fpu_finish() sets TIF_NEED_FPU_LOAD; the floating point state
 *    will get loaded on return to userspace, or when the kernel needs it.
495
 *
496 497 498 499 500 501 502 503
 * If TIF_NEED_FPU_LOAD is cleared then the CPU's FPU registers
 * are saved in the current thread's FPU register state.
 *
 * If TIF_NEED_FPU_LOAD is set then CPU's FPU registers may not
 * hold current()'s FPU registers. It is required to load the
 * registers before returning to userland or using the content
 * otherwise.
 *
504
 * The FPU context is only stored/restored for a user task and
505
 * PF_KTHREAD is used to distinguish between kernel and user threads.
506
 */
507
static inline void switch_fpu_prepare(struct fpu *old_fpu, int cpu)
508
{
509
	if (static_cpu_has(X86_FEATURE_FPU) && !(current->flags & PF_KTHREAD)) {
510
		if (!save_fpregs_to_fpstate(old_fpu))
511
			old_fpu->last_cpu = -1;
512
		else
513
			old_fpu->last_cpu = cpu;
514

515
		/* But leave fpu_fpregs_owner_ctx! */
516
		trace_x86_fpu_regs_deactivated(old_fpu);
517
	}
518 519
}

I
Ingo Molnar 已提交
520 521 522 523
/*
 * Misc helper functions:
 */

524
/*
525 526
 * Load PKRU from the FPU context if available. Delay loading of the
 * complete FPU state until the return to userland.
527
 */
528
static inline void switch_fpu_finish(struct fpu *new_fpu)
529
{
R
Rik van Riel 已提交
530 531
	u32 pkru_val = init_pkru_value;
	struct pkru_state *pk;
532

R
Rik van Riel 已提交
533 534
	if (!static_cpu_has(X86_FEATURE_FPU))
		return;
535

536
	set_thread_flag(TIF_NEED_FPU_LOAD);
R
Rik van Riel 已提交
537 538 539

	if (!cpu_feature_enabled(X86_FEATURE_OSPKE))
		return;
540

R
Rik van Riel 已提交
541 542 543 544
	/*
	 * PKRU state is switched eagerly because it needs to be valid before we
	 * return to userland e.g. for a copy_to_user() operation.
	 */
545
	if (!(current->flags & PF_KTHREAD)) {
546 547 548 549 550 551 552 553
		/*
		 * If the PKRU bit in xsave.header.xfeatures is not set,
		 * then the PKRU component was in init state, which means
		 * XRSTOR will set PKRU to 0. If the bit is not set then
		 * get_xsave_addr() will return NULL because the PKRU value
		 * in memory is not valid. This means pkru_val has to be
		 * set to 0 and not to init_pkru_value.
		 */
R
Rik van Riel 已提交
554
		pk = get_xsave_addr(&new_fpu->state.xsave, XFEATURE_PKRU);
555
		pkru_val = pk ? pk->pkru : 0;
R
Rik van Riel 已提交
556 557
	}
	__write_pkru(pkru_val);
558 559
}

560
#endif /* _ASM_X86_FPU_INTERNAL_H */