Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
c767a54b JP |
2 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
3 | ||
61c4628b SS |
4 | #include <linux/errno.h> |
5 | #include <linux/kernel.h> | |
6 | #include <linux/mm.h> | |
7 | #include <linux/smp.h> | |
b4c108d7 | 8 | #include <linux/cpu.h> |
389d1fb1 | 9 | #include <linux/prctl.h> |
61c4628b SS |
10 | #include <linux/slab.h> |
11 | #include <linux/sched.h> | |
4c822698 | 12 | #include <linux/sched/idle.h> |
b17b0153 | 13 | #include <linux/sched/debug.h> |
29930025 | 14 | #include <linux/sched/task.h> |
68db0cf1 | 15 | #include <linux/sched/task_stack.h> |
186f4360 PG |
16 | #include <linux/init.h> |
17 | #include <linux/export.h> | |
7f424a8b | 18 | #include <linux/pm.h> |
162a688e | 19 | #include <linux/tick.h> |
9d62dcdf | 20 | #include <linux/random.h> |
7c68af6e | 21 | #include <linux/user-return-notifier.h> |
814e2c84 AI |
22 | #include <linux/dmi.h> |
23 | #include <linux/utsname.h> | |
90e24014 | 24 | #include <linux/stackprotector.h> |
90e24014 | 25 | #include <linux/cpuidle.h> |
89f579ce YW |
26 | #include <linux/acpi.h> |
27 | #include <linux/elf-randomize.h> | |
aaa3896b | 28 | #include <linux/static_call.h> |
61613521 | 29 | #include <trace/events/power.h> |
24f1e32c | 30 | #include <linux/hw_breakpoint.h> |
93789b32 | 31 | #include <asm/cpu.h> |
d3ec5cae | 32 | #include <asm/apic.h> |
7c0f6ba6 | 33 | #include <linux/uaccess.h> |
b253149b | 34 | #include <asm/mwait.h> |
db8268df | 35 | #include <asm/fpu/api.h> |
63e81807 | 36 | #include <asm/fpu/sched.h> |
500afbf6 | 37 | #include <asm/fpu/xstate.h> |
66cb5917 | 38 | #include <asm/debugreg.h> |
90e24014 | 39 | #include <asm/nmi.h> |
375074cc | 40 | #include <asm/tlbflush.h> |
8838eb6c | 41 | #include <asm/mce.h> |
9fda6a06 | 42 | #include <asm/vm86.h> |
7b32aead | 43 | #include <asm/switch_to.h> |
b7ffc44d | 44 | #include <asm/desc.h> |
e9ea1e7f | 45 | #include <asm/prctl.h> |
885f82bf | 46 | #include <asm/spec-ctrl.h> |
577d5cd7 | 47 | #include <asm/io_bitmap.h> |
89f579ce | 48 | #include <asm/proto.h> |
6f9885a3 | 49 | #include <asm/frame.h> |
5d1ceb39 | 50 | #include <asm/unwind.h> |
bfe6ed0c | 51 | #include <asm/tdx.h> |
90e24014 | 52 | |
ff16701a TG |
53 | #include "process.h" |
54 | ||
45046892 TG |
55 | /* |
56 | * per-CPU TSS segments. Threads are completely 'soft' on Linux, | |
57 | * no more per-task TSS's. The TSS size is kept cacheline-aligned | |
58 | * so they are allowed to end up in the .data..cacheline_aligned | |
59 | * section. Since TSS's are completely CPU-local, we want them | |
60 | * on exact cacheline boundaries, to eliminate cacheline ping-pong. | |
61 | */ | |
2fd9c41a | 62 | __visible DEFINE_PER_CPU_PAGE_ALIGNED(struct tss_struct, cpu_tss_rw) = { |
d0a0de21 | 63 | .x86_tss = { |
20bb8344 AL |
64 | /* |
65 | * .sp0 is only used when entering ring 0 from a lower | |
66 | * privilege level. Since the init task never runs anything | |
67 | * but ring 0 code, there is no need for a valid value here. | |
68 | * Poison it. | |
69 | */ | |
70 | .sp0 = (1UL << (BITS_PER_LONG-1)) + 1, | |
9aaefe7b | 71 | |
1591584e | 72 | #ifdef CONFIG_X86_32 |
9aaefe7b | 73 | .sp1 = TOP_OF_INIT_STACK, |
9aaefe7b | 74 | |
d0a0de21 AL |
75 | .ss0 = __KERNEL_DS, |
76 | .ss1 = __KERNEL_CS, | |
d0a0de21 | 77 | #endif |
ecc7e37d | 78 | .io_bitmap_base = IO_BITMAP_OFFSET_INVALID, |
d0a0de21 | 79 | }, |
d0a0de21 | 80 | }; |
c482feef | 81 | EXPORT_PER_CPU_SYMBOL(cpu_tss_rw); |
45046892 | 82 | |
b7ceaec1 AL |
83 | DEFINE_PER_CPU(bool, __tss_limit_invalid); |
84 | EXPORT_PER_CPU_SYMBOL_GPL(__tss_limit_invalid); | |
b7ffc44d | 85 | |
55ccf3fe SS |
86 | /* |
87 | * this gets called so that we can store lazy state into memory and copy the | |
88 | * current task into the new thread. | |
89 | */ | |
61c4628b SS |
90 | int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) |
91 | { | |
5aaeb5c0 | 92 | memcpy(dst, src, arch_task_struct_size); |
2459ee86 AL |
93 | #ifdef CONFIG_VM86 |
94 | dst->thread.vm86 = NULL; | |
95 | #endif | |
f0cbc8b3 TG |
96 | /* Drop the copied pointer to current's fpstate */ |
97 | dst->thread.fpu.fpstate = NULL; | |
500afbf6 | 98 | |
2d16a187 | 99 | return 0; |
61c4628b | 100 | } |
7f424a8b | 101 | |
500afbf6 CB |
102 | #ifdef CONFIG_X86_64 |
103 | void arch_release_task_struct(struct task_struct *tsk) | |
104 | { | |
105 | if (fpu_state_size_dynamic()) | |
106 | fpstate_free(&tsk->thread.fpu); | |
61c4628b | 107 | } |
500afbf6 | 108 | #endif |
7f424a8b | 109 | |
389d1fb1 | 110 | /* |
4bfe6cce | 111 | * Free thread data structures etc.. |
389d1fb1 | 112 | */ |
e6464694 | 113 | void exit_thread(struct task_struct *tsk) |
389d1fb1 | 114 | { |
e6464694 | 115 | struct thread_struct *t = &tsk->thread; |
ca6787ba | 116 | struct fpu *fpu = &t->fpu; |
ea5f1cd7 TG |
117 | |
118 | if (test_thread_flag(TIF_IO_BITMAP)) | |
4bfe6cce | 119 | io_bitmap_exit(tsk); |
1dcc8d7b | 120 | |
9fda6a06 BG |
121 | free_vm86(t); |
122 | ||
50338615 | 123 | fpu__drop(fpu); |
389d1fb1 JF |
124 | } |
125 | ||
2fff071d TG |
126 | static int set_new_tls(struct task_struct *p, unsigned long tls) |
127 | { | |
128 | struct user_desc __user *utls = (struct user_desc __user *)tls; | |
129 | ||
130 | if (in_ia32_syscall()) | |
131 | return do_set_thread_area(p, -1, utls, 0); | |
132 | else | |
133 | return do_set_thread_area_64(p, ARCH_SET_FS, tls); | |
134 | } | |
135 | ||
c5febea0 | 136 | int copy_thread(struct task_struct *p, const struct kernel_clone_args *args) |
2fff071d | 137 | { |
c5febea0 EB |
138 | unsigned long clone_flags = args->flags; |
139 | unsigned long sp = args->stack; | |
c5febea0 | 140 | unsigned long tls = args->tls; |
2fff071d TG |
141 | struct inactive_task_frame *frame; |
142 | struct fork_frame *fork_frame; | |
143 | struct pt_regs *childregs; | |
4804e382 | 144 | int ret = 0; |
2fff071d TG |
145 | |
146 | childregs = task_pt_regs(p); | |
147 | fork_frame = container_of(childregs, struct fork_frame, regs); | |
148 | frame = &fork_frame->frame; | |
149 | ||
6f9885a3 | 150 | frame->bp = encode_frame_pointer(childregs); |
2fff071d TG |
151 | frame->ret_addr = (unsigned long) ret_from_fork; |
152 | p->thread.sp = (unsigned long) fork_frame; | |
577d5cd7 | 153 | p->thread.io_bitmap = NULL; |
b968e84b | 154 | p->thread.iopl_warn = 0; |
2fff071d TG |
155 | memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps)); |
156 | ||
157 | #ifdef CONFIG_X86_64 | |
005f141e CB |
158 | current_save_fsgs(); |
159 | p->thread.fsindex = current->thread.fsindex; | |
160 | p->thread.fsbase = current->thread.fsbase; | |
161 | p->thread.gsindex = current->thread.gsindex; | |
162 | p->thread.gsbase = current->thread.gsbase; | |
163 | ||
2fff071d TG |
164 | savesegment(es, p->thread.es); |
165 | savesegment(ds, p->thread.ds); | |
166 | #else | |
167 | p->thread.sp0 = (unsigned long) (childregs + 1); | |
3a24a608 | 168 | savesegment(gs, p->thread.gs); |
2fff071d TG |
169 | /* |
170 | * Clear all status flags including IF and set fixed bit. 64bit | |
171 | * does not have this initialization as the frame does not contain | |
172 | * flags. The flags consistency (especially vs. AC) is there | |
173 | * ensured via objtool, which lacks 32bit support. | |
174 | */ | |
175 | frame->flags = X86_EFLAGS_FIXED; | |
176 | #endif | |
177 | ||
5bd2e97c | 178 | fpu_clone(p, clone_flags, args->fn); |
2d16a187 | 179 | |
2fff071d | 180 | /* Kernel thread ? */ |
50b7b6f2 | 181 | if (unlikely(p->flags & PF_KTHREAD)) { |
9782a712 | 182 | p->thread.pkru = pkru_get_init_value(); |
2fff071d | 183 | memset(childregs, 0, sizeof(struct pt_regs)); |
5bd2e97c | 184 | kthread_frame_init(frame, args->fn, args->fn_arg); |
2fff071d TG |
185 | return 0; |
186 | } | |
187 | ||
9782a712 DH |
188 | /* |
189 | * Clone current's PKRU value from hardware. tsk->thread.pkru | |
190 | * is only valid when scheduled out. | |
191 | */ | |
192 | p->thread.pkru = read_pkru(); | |
193 | ||
2fff071d TG |
194 | frame->bx = 0; |
195 | *childregs = *current_pt_regs(); | |
196 | childregs->ax = 0; | |
197 | if (sp) | |
198 | childregs->sp = sp; | |
199 | ||
5bd2e97c | 200 | if (unlikely(args->fn)) { |
50b7b6f2 | 201 | /* |
5bd2e97c EB |
202 | * A user space thread, but it doesn't return to |
203 | * ret_after_fork(). | |
50b7b6f2 SM |
204 | * |
205 | * In order to indicate that to tools like gdb, | |
206 | * we reset the stack and instruction pointers. | |
207 | * | |
208 | * It does the same kernel frame setup to return to a kernel | |
209 | * function that a kernel thread does. | |
210 | */ | |
211 | childregs->sp = 0; | |
212 | childregs->ip = 0; | |
5bd2e97c | 213 | kthread_frame_init(frame, args->fn, args->fn_arg); |
50b7b6f2 SM |
214 | return 0; |
215 | } | |
216 | ||
2fff071d | 217 | /* Set a new TLS for the child thread? */ |
4804e382 | 218 | if (clone_flags & CLONE_SETTLS) |
2fff071d | 219 | ret = set_new_tls(p, tls); |
4804e382 TG |
220 | |
221 | if (!ret && unlikely(test_tsk_thread_flag(current, TIF_IO_BITMAP))) | |
222 | io_bitmap_share(p); | |
223 | ||
2fff071d TG |
224 | return ret; |
225 | } | |
226 | ||
33344368 AL |
227 | static void pkru_flush_thread(void) |
228 | { | |
229 | /* | |
230 | * If PKRU is enabled the default PKRU value has to be loaded into | |
231 | * the hardware right here (similar to context switch). | |
232 | */ | |
233 | pkru_write_default(); | |
234 | } | |
235 | ||
389d1fb1 JF |
236 | void flush_thread(void) |
237 | { | |
238 | struct task_struct *tsk = current; | |
239 | ||
24f1e32c | 240 | flush_ptrace_hw_breakpoint(tsk); |
389d1fb1 | 241 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
110d7f75 | 242 | |
e7ecad17 | 243 | fpu_flush_thread(); |
33344368 | 244 | pkru_flush_thread(); |
389d1fb1 JF |
245 | } |
246 | ||
389d1fb1 JF |
247 | void disable_TSC(void) |
248 | { | |
249 | preempt_disable(); | |
250 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
251 | /* | |
252 | * Must flip the CPU state synchronously with | |
253 | * TIF_NOTSC in the current running context. | |
254 | */ | |
5a920155 | 255 | cr4_set_bits(X86_CR4_TSD); |
389d1fb1 JF |
256 | preempt_enable(); |
257 | } | |
258 | ||
389d1fb1 JF |
259 | static void enable_TSC(void) |
260 | { | |
261 | preempt_disable(); | |
262 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
263 | /* | |
264 | * Must flip the CPU state synchronously with | |
265 | * TIF_NOTSC in the current running context. | |
266 | */ | |
5a920155 | 267 | cr4_clear_bits(X86_CR4_TSD); |
389d1fb1 JF |
268 | preempt_enable(); |
269 | } | |
270 | ||
271 | int get_tsc_mode(unsigned long adr) | |
272 | { | |
273 | unsigned int val; | |
274 | ||
275 | if (test_thread_flag(TIF_NOTSC)) | |
276 | val = PR_TSC_SIGSEGV; | |
277 | else | |
278 | val = PR_TSC_ENABLE; | |
279 | ||
280 | return put_user(val, (unsigned int __user *)adr); | |
281 | } | |
282 | ||
283 | int set_tsc_mode(unsigned int val) | |
284 | { | |
285 | if (val == PR_TSC_SIGSEGV) | |
286 | disable_TSC(); | |
287 | else if (val == PR_TSC_ENABLE) | |
288 | enable_TSC(); | |
289 | else | |
290 | return -EINVAL; | |
291 | ||
292 | return 0; | |
293 | } | |
294 | ||
e9ea1e7f KH |
295 | DEFINE_PER_CPU(u64, msr_misc_features_shadow); |
296 | ||
297 | static void set_cpuid_faulting(bool on) | |
298 | { | |
299 | u64 msrval; | |
300 | ||
301 | msrval = this_cpu_read(msr_misc_features_shadow); | |
302 | msrval &= ~MSR_MISC_FEATURES_ENABLES_CPUID_FAULT; | |
303 | msrval |= (on << MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT); | |
304 | this_cpu_write(msr_misc_features_shadow, msrval); | |
305 | wrmsrl(MSR_MISC_FEATURES_ENABLES, msrval); | |
306 | } | |
307 | ||
308 | static void disable_cpuid(void) | |
309 | { | |
310 | preempt_disable(); | |
311 | if (!test_and_set_thread_flag(TIF_NOCPUID)) { | |
312 | /* | |
313 | * Must flip the CPU state synchronously with | |
314 | * TIF_NOCPUID in the current running context. | |
315 | */ | |
316 | set_cpuid_faulting(true); | |
317 | } | |
318 | preempt_enable(); | |
319 | } | |
320 | ||
321 | static void enable_cpuid(void) | |
322 | { | |
323 | preempt_disable(); | |
324 | if (test_and_clear_thread_flag(TIF_NOCPUID)) { | |
325 | /* | |
326 | * Must flip the CPU state synchronously with | |
327 | * TIF_NOCPUID in the current running context. | |
328 | */ | |
329 | set_cpuid_faulting(false); | |
330 | } | |
331 | preempt_enable(); | |
332 | } | |
333 | ||
334 | static int get_cpuid_mode(void) | |
335 | { | |
336 | return !test_thread_flag(TIF_NOCPUID); | |
337 | } | |
338 | ||
f5c0b4f3 | 339 | static int set_cpuid_mode(unsigned long cpuid_enabled) |
e9ea1e7f | 340 | { |
67e87d43 | 341 | if (!boot_cpu_has(X86_FEATURE_CPUID_FAULT)) |
e9ea1e7f KH |
342 | return -ENODEV; |
343 | ||
344 | if (cpuid_enabled) | |
345 | enable_cpuid(); | |
346 | else | |
347 | disable_cpuid(); | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
352 | /* | |
353 | * Called immediately after a successful exec. | |
354 | */ | |
355 | void arch_setup_new_exec(void) | |
356 | { | |
357 | /* If cpuid was previously disabled for this task, re-enable it. */ | |
358 | if (test_thread_flag(TIF_NOCPUID)) | |
359 | enable_cpuid(); | |
71368af9 WL |
360 | |
361 | /* | |
362 | * Don't inherit TIF_SSBD across exec boundary when | |
363 | * PR_SPEC_DISABLE_NOEXEC is used. | |
364 | */ | |
365 | if (test_thread_flag(TIF_SSBD) && | |
366 | task_spec_ssb_noexec(current)) { | |
367 | clear_thread_flag(TIF_SSBD); | |
368 | task_clear_spec_ssb_disable(current); | |
369 | task_clear_spec_ssb_noexec(current); | |
dca99fb6 | 370 | speculation_ctrl_update(read_thread_flags()); |
71368af9 | 371 | } |
e9ea1e7f KH |
372 | } |
373 | ||
111e7b15 | 374 | #ifdef CONFIG_X86_IOPL_IOPERM |
22fe5b04 TG |
375 | static inline void switch_to_bitmap(unsigned long tifp) |
376 | { | |
377 | /* | |
378 | * Invalidate I/O bitmap if the previous task used it. This prevents | |
379 | * any possible leakage of an active I/O bitmap. | |
380 | * | |
381 | * If the next task has an I/O bitmap it will handle it on exit to | |
382 | * user mode. | |
383 | */ | |
384 | if (tifp & _TIF_IO_BITMAP) | |
cadfad87 | 385 | tss_invalidate_io_bitmap(); |
22fe5b04 TG |
386 | } |
387 | ||
388 | static void tss_copy_io_bitmap(struct tss_struct *tss, struct io_bitmap *iobm) | |
060aa16f TG |
389 | { |
390 | /* | |
391 | * Copy at least the byte range of the incoming tasks bitmap which | |
392 | * covers the permitted I/O ports. | |
393 | * | |
394 | * If the previous task which used an I/O bitmap had more bits | |
395 | * permitted, then the copy needs to cover those as well so they | |
396 | * get turned off. | |
397 | */ | |
398 | memcpy(tss->io_bitmap.bitmap, iobm->bitmap, | |
399 | max(tss->io_bitmap.prev_max, iobm->max)); | |
400 | ||
401 | /* | |
402 | * Store the new max and the sequence number of this bitmap | |
403 | * and a pointer to the bitmap itself. | |
404 | */ | |
405 | tss->io_bitmap.prev_max = iobm->max; | |
406 | tss->io_bitmap.prev_sequence = iobm->sequence; | |
407 | } | |
408 | ||
22fe5b04 | 409 | /** |
dbb5ab6d | 410 | * native_tss_update_io_bitmap - Update I/O bitmap before exiting to user mode |
22fe5b04 | 411 | */ |
99bcd4a6 | 412 | void native_tss_update_io_bitmap(void) |
af8b3cd3 | 413 | { |
ff16701a | 414 | struct tss_struct *tss = this_cpu_ptr(&cpu_tss_rw); |
7b0b8cfd | 415 | struct thread_struct *t = ¤t->thread; |
c8137ace | 416 | u16 *base = &tss->x86_tss.io_bitmap_base; |
ff16701a | 417 | |
7b0b8cfd | 418 | if (!test_thread_flag(TIF_IO_BITMAP)) { |
cadfad87 | 419 | native_tss_invalidate_io_bitmap(); |
7b0b8cfd BP |
420 | return; |
421 | } | |
422 | ||
423 | if (IS_ENABLED(CONFIG_X86_IOPL_IOPERM) && t->iopl_emul == 3) { | |
424 | *base = IO_BITMAP_OFFSET_VALID_ALL; | |
425 | } else { | |
426 | struct io_bitmap *iobm = t->io_bitmap; | |
427 | ||
af8b3cd3 | 428 | /* |
7b0b8cfd BP |
429 | * Only copy bitmap data when the sequence number differs. The |
430 | * update time is accounted to the incoming task. | |
af8b3cd3 | 431 | */ |
7b0b8cfd BP |
432 | if (tss->io_bitmap.prev_sequence != iobm->sequence) |
433 | tss_copy_io_bitmap(tss, iobm); | |
434 | ||
435 | /* Enable the bitmap */ | |
436 | *base = IO_BITMAP_OFFSET_VALID_MAP; | |
af8b3cd3 | 437 | } |
7b0b8cfd BP |
438 | |
439 | /* | |
440 | * Make sure that the TSS limit is covering the IO bitmap. It might have | |
441 | * been cut down by a VMEXIT to 0x67 which would cause a subsequent I/O | |
442 | * access from user space to trigger a #GP because tbe bitmap is outside | |
443 | * the TSS limit. | |
444 | */ | |
445 | refresh_tss_limit(); | |
af8b3cd3 | 446 | } |
111e7b15 TG |
447 | #else /* CONFIG_X86_IOPL_IOPERM */ |
448 | static inline void switch_to_bitmap(unsigned long tifp) { } | |
449 | #endif | |
af8b3cd3 | 450 | |
1f50ddb4 TG |
451 | #ifdef CONFIG_SMP |
452 | ||
453 | struct ssb_state { | |
454 | struct ssb_state *shared_state; | |
455 | raw_spinlock_t lock; | |
456 | unsigned int disable_state; | |
457 | unsigned long local_state; | |
458 | }; | |
459 | ||
460 | #define LSTATE_SSB 0 | |
461 | ||
462 | static DEFINE_PER_CPU(struct ssb_state, ssb_state); | |
463 | ||
464 | void speculative_store_bypass_ht_init(void) | |
885f82bf | 465 | { |
1f50ddb4 TG |
466 | struct ssb_state *st = this_cpu_ptr(&ssb_state); |
467 | unsigned int this_cpu = smp_processor_id(); | |
468 | unsigned int cpu; | |
469 | ||
470 | st->local_state = 0; | |
471 | ||
472 | /* | |
473 | * Shared state setup happens once on the first bringup | |
474 | * of the CPU. It's not destroyed on CPU hotunplug. | |
475 | */ | |
476 | if (st->shared_state) | |
477 | return; | |
478 | ||
479 | raw_spin_lock_init(&st->lock); | |
480 | ||
481 | /* | |
482 | * Go over HT siblings and check whether one of them has set up the | |
483 | * shared state pointer already. | |
484 | */ | |
485 | for_each_cpu(cpu, topology_sibling_cpumask(this_cpu)) { | |
486 | if (cpu == this_cpu) | |
487 | continue; | |
488 | ||
489 | if (!per_cpu(ssb_state, cpu).shared_state) | |
490 | continue; | |
491 | ||
492 | /* Link it to the state of the sibling: */ | |
493 | st->shared_state = per_cpu(ssb_state, cpu).shared_state; | |
494 | return; | |
495 | } | |
496 | ||
497 | /* | |
498 | * First HT sibling to come up on the core. Link shared state of | |
499 | * the first HT sibling to itself. The siblings on the same core | |
500 | * which come up later will see the shared state pointer and link | |
d9f6e12f | 501 | * themselves to the state of this CPU. |
1f50ddb4 TG |
502 | */ |
503 | st->shared_state = st; | |
504 | } | |
885f82bf | 505 | |
1f50ddb4 TG |
506 | /* |
507 | * Logic is: First HT sibling enables SSBD for both siblings in the core | |
508 | * and last sibling to disable it, disables it for the whole core. This how | |
509 | * MSR_SPEC_CTRL works in "hardware": | |
510 | * | |
511 | * CORE_SPEC_CTRL = THREAD0_SPEC_CTRL | THREAD1_SPEC_CTRL | |
512 | */ | |
513 | static __always_inline void amd_set_core_ssb_state(unsigned long tifn) | |
514 | { | |
515 | struct ssb_state *st = this_cpu_ptr(&ssb_state); | |
516 | u64 msr = x86_amd_ls_cfg_base; | |
517 | ||
518 | if (!static_cpu_has(X86_FEATURE_ZEN)) { | |
519 | msr |= ssbd_tif_to_amd_ls_cfg(tifn); | |
885f82bf | 520 | wrmsrl(MSR_AMD64_LS_CFG, msr); |
1f50ddb4 TG |
521 | return; |
522 | } | |
523 | ||
524 | if (tifn & _TIF_SSBD) { | |
525 | /* | |
526 | * Since this can race with prctl(), block reentry on the | |
527 | * same CPU. | |
528 | */ | |
529 | if (__test_and_set_bit(LSTATE_SSB, &st->local_state)) | |
530 | return; | |
531 | ||
532 | msr |= x86_amd_ls_cfg_ssbd_mask; | |
533 | ||
534 | raw_spin_lock(&st->shared_state->lock); | |
535 | /* First sibling enables SSBD: */ | |
536 | if (!st->shared_state->disable_state) | |
537 | wrmsrl(MSR_AMD64_LS_CFG, msr); | |
538 | st->shared_state->disable_state++; | |
539 | raw_spin_unlock(&st->shared_state->lock); | |
885f82bf | 540 | } else { |
1f50ddb4 TG |
541 | if (!__test_and_clear_bit(LSTATE_SSB, &st->local_state)) |
542 | return; | |
543 | ||
544 | raw_spin_lock(&st->shared_state->lock); | |
545 | st->shared_state->disable_state--; | |
546 | if (!st->shared_state->disable_state) | |
547 | wrmsrl(MSR_AMD64_LS_CFG, msr); | |
548 | raw_spin_unlock(&st->shared_state->lock); | |
885f82bf TG |
549 | } |
550 | } | |
1f50ddb4 TG |
551 | #else |
552 | static __always_inline void amd_set_core_ssb_state(unsigned long tifn) | |
553 | { | |
554 | u64 msr = x86_amd_ls_cfg_base | ssbd_tif_to_amd_ls_cfg(tifn); | |
555 | ||
556 | wrmsrl(MSR_AMD64_LS_CFG, msr); | |
557 | } | |
558 | #endif | |
559 | ||
11fb0683 TL |
560 | static __always_inline void amd_set_ssb_virt_state(unsigned long tifn) |
561 | { | |
562 | /* | |
563 | * SSBD has the same definition in SPEC_CTRL and VIRT_SPEC_CTRL, | |
564 | * so ssbd_tif_to_spec_ctrl() just works. | |
565 | */ | |
566 | wrmsrl(MSR_AMD64_VIRT_SPEC_CTRL, ssbd_tif_to_spec_ctrl(tifn)); | |
567 | } | |
568 | ||
01daf568 TC |
569 | /* |
570 | * Update the MSRs managing speculation control, during context switch. | |
571 | * | |
572 | * tifp: Previous task's thread flags | |
573 | * tifn: Next task's thread flags | |
574 | */ | |
575 | static __always_inline void __speculation_ctrl_update(unsigned long tifp, | |
576 | unsigned long tifn) | |
1f50ddb4 | 577 | { |
5bfbe3ad | 578 | unsigned long tif_diff = tifp ^ tifn; |
01daf568 TC |
579 | u64 msr = x86_spec_ctrl_base; |
580 | bool updmsr = false; | |
581 | ||
2f5fb193 TG |
582 | lockdep_assert_irqs_disabled(); |
583 | ||
dbbe2ad0 AS |
584 | /* Handle change of TIF_SSBD depending on the mitigation method. */ |
585 | if (static_cpu_has(X86_FEATURE_VIRT_SSBD)) { | |
586 | if (tif_diff & _TIF_SSBD) | |
01daf568 | 587 | amd_set_ssb_virt_state(tifn); |
dbbe2ad0 AS |
588 | } else if (static_cpu_has(X86_FEATURE_LS_CFG_SSBD)) { |
589 | if (tif_diff & _TIF_SSBD) | |
01daf568 | 590 | amd_set_core_ssb_state(tifn); |
dbbe2ad0 AS |
591 | } else if (static_cpu_has(X86_FEATURE_SPEC_CTRL_SSBD) || |
592 | static_cpu_has(X86_FEATURE_AMD_SSBD)) { | |
593 | updmsr |= !!(tif_diff & _TIF_SSBD); | |
594 | msr |= ssbd_tif_to_spec_ctrl(tifn); | |
01daf568 | 595 | } |
1f50ddb4 | 596 | |
dbbe2ad0 | 597 | /* Only evaluate TIF_SPEC_IB if conditional STIBP is enabled. */ |
5bfbe3ad TC |
598 | if (IS_ENABLED(CONFIG_SMP) && |
599 | static_branch_unlikely(&switch_to_cond_stibp)) { | |
600 | updmsr |= !!(tif_diff & _TIF_SPEC_IB); | |
601 | msr |= stibp_tif_to_spec_ctrl(tifn); | |
602 | } | |
603 | ||
01daf568 | 604 | if (updmsr) |
66065157 | 605 | update_spec_ctrl_cond(msr); |
1f50ddb4 TG |
606 | } |
607 | ||
6d991ba5 | 608 | static unsigned long speculation_ctrl_update_tif(struct task_struct *tsk) |
1f50ddb4 | 609 | { |
6d991ba5 TG |
610 | if (test_and_clear_tsk_thread_flag(tsk, TIF_SPEC_FORCE_UPDATE)) { |
611 | if (task_spec_ssb_disable(tsk)) | |
612 | set_tsk_thread_flag(tsk, TIF_SSBD); | |
613 | else | |
614 | clear_tsk_thread_flag(tsk, TIF_SSBD); | |
9137bb27 TG |
615 | |
616 | if (task_spec_ib_disable(tsk)) | |
617 | set_tsk_thread_flag(tsk, TIF_SPEC_IB); | |
618 | else | |
619 | clear_tsk_thread_flag(tsk, TIF_SPEC_IB); | |
6d991ba5 TG |
620 | } |
621 | /* Return the updated threadinfo flags*/ | |
dca99fb6 | 622 | return read_task_thread_flags(tsk); |
1f50ddb4 | 623 | } |
885f82bf | 624 | |
26c4d75b | 625 | void speculation_ctrl_update(unsigned long tif) |
885f82bf | 626 | { |
2f5fb193 TG |
627 | unsigned long flags; |
628 | ||
01daf568 | 629 | /* Forced update. Make sure all relevant TIF flags are different */ |
2f5fb193 | 630 | local_irq_save(flags); |
01daf568 | 631 | __speculation_ctrl_update(~tif, tif); |
2f5fb193 | 632 | local_irq_restore(flags); |
885f82bf TG |
633 | } |
634 | ||
6d991ba5 TG |
635 | /* Called from seccomp/prctl update */ |
636 | void speculation_ctrl_update_current(void) | |
637 | { | |
638 | preempt_disable(); | |
639 | speculation_ctrl_update(speculation_ctrl_update_tif(current)); | |
640 | preempt_enable(); | |
641 | } | |
642 | ||
d8f0b353 TG |
643 | static inline void cr4_toggle_bits_irqsoff(unsigned long mask) |
644 | { | |
645 | unsigned long newval, cr4 = this_cpu_read(cpu_tlbstate.cr4); | |
646 | ||
647 | newval = cr4 ^ mask; | |
648 | if (newval != cr4) { | |
649 | this_cpu_write(cpu_tlbstate.cr4, newval); | |
650 | __write_cr4(newval); | |
651 | } | |
652 | } | |
653 | ||
ff16701a | 654 | void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p) |
389d1fb1 | 655 | { |
af8b3cd3 | 656 | unsigned long tifp, tifn; |
389d1fb1 | 657 | |
dca99fb6 MR |
658 | tifn = read_task_thread_flags(next_p); |
659 | tifp = read_task_thread_flags(prev_p); | |
22fe5b04 TG |
660 | |
661 | switch_to_bitmap(tifp); | |
af8b3cd3 KH |
662 | |
663 | propagate_user_return_notify(prev_p, next_p); | |
664 | ||
b9894a2f KH |
665 | if ((tifp & _TIF_BLOCKSTEP || tifn & _TIF_BLOCKSTEP) && |
666 | arch_has_block_step()) { | |
667 | unsigned long debugctl, msk; | |
ea8e61b7 | 668 | |
b9894a2f | 669 | rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); |
ea8e61b7 | 670 | debugctl &= ~DEBUGCTLMSR_BTF; |
b9894a2f KH |
671 | msk = tifn & _TIF_BLOCKSTEP; |
672 | debugctl |= (msk >> TIF_BLOCKSTEP) << DEBUGCTLMSR_BTF_SHIFT; | |
673 | wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); | |
ea8e61b7 | 674 | } |
389d1fb1 | 675 | |
5a920155 | 676 | if ((tifp ^ tifn) & _TIF_NOTSC) |
9d0b6232 | 677 | cr4_toggle_bits_irqsoff(X86_CR4_TSD); |
e9ea1e7f KH |
678 | |
679 | if ((tifp ^ tifn) & _TIF_NOCPUID) | |
680 | set_cpuid_faulting(!!(tifn & _TIF_NOCPUID)); | |
885f82bf | 681 | |
6d991ba5 TG |
682 | if (likely(!((tifp | tifn) & _TIF_SPEC_FORCE_UPDATE))) { |
683 | __speculation_ctrl_update(tifp, tifn); | |
684 | } else { | |
685 | speculation_ctrl_update_tif(prev_p); | |
686 | tifn = speculation_ctrl_update_tif(next_p); | |
687 | ||
688 | /* Enforce MSR update to ensure consistent state */ | |
689 | __speculation_ctrl_update(~tifn, tifn); | |
690 | } | |
389d1fb1 JF |
691 | } |
692 | ||
00dba564 TG |
693 | /* |
694 | * Idle related variables and functions | |
695 | */ | |
d1896049 | 696 | unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE; |
00dba564 TG |
697 | EXPORT_SYMBOL(boot_option_idle_override); |
698 | ||
aaa3896b PZ |
699 | /* |
700 | * We use this if we don't have any better idle routine.. | |
701 | */ | |
702 | void __cpuidle default_idle(void) | |
703 | { | |
704 | raw_safe_halt(); | |
89b30987 | 705 | raw_local_irq_disable(); |
aaa3896b PZ |
706 | } |
707 | #if defined(CONFIG_APM_MODULE) || defined(CONFIG_HALTPOLL_CPUIDLE_MODULE) | |
708 | EXPORT_SYMBOL(default_idle); | |
709 | #endif | |
710 | ||
711 | DEFINE_STATIC_CALL_NULL(x86_idle, default_idle); | |
712 | ||
713 | static bool x86_idle_set(void) | |
714 | { | |
715 | return !!static_call_query(x86_idle); | |
716 | } | |
00dba564 | 717 | |
90e24014 | 718 | #ifndef CONFIG_SMP |
eab89405 | 719 | static inline void __noreturn play_dead(void) |
90e24014 RW |
720 | { |
721 | BUG(); | |
722 | } | |
723 | #endif | |
724 | ||
7d1a9417 TG |
725 | void arch_cpu_idle_enter(void) |
726 | { | |
6a369583 | 727 | tsc_verify_tsc_adjust(false); |
7d1a9417 | 728 | local_touch_nmi(); |
7d1a9417 | 729 | } |
90e24014 | 730 | |
071c44e4 | 731 | void __noreturn arch_cpu_idle_dead(void) |
7d1a9417 TG |
732 | { |
733 | play_dead(); | |
734 | } | |
90e24014 | 735 | |
7d1a9417 TG |
736 | /* |
737 | * Called from the generic idle code. | |
738 | */ | |
aaa3896b | 739 | void __cpuidle arch_cpu_idle(void) |
00dba564 | 740 | { |
aaa3896b | 741 | static_call(x86_idle)(); |
00dba564 | 742 | } |
716ff71a | 743 | EXPORT_SYMBOL_GPL(arch_cpu_idle); |
00dba564 | 744 | |
6a377ddc LB |
745 | #ifdef CONFIG_XEN |
746 | bool xen_set_default_idle(void) | |
e5fd47bf | 747 | { |
aaa3896b | 748 | bool ret = x86_idle_set(); |
e5fd47bf | 749 | |
aaa3896b | 750 | static_call_update(x86_idle, default_idle); |
e5fd47bf KRW |
751 | |
752 | return ret; | |
753 | } | |
6a377ddc | 754 | #endif |
bba4ed01 | 755 | |
f9cdf7ca | 756 | void __noreturn stop_this_cpu(void *dummy) |
d3ec5cae IV |
757 | { |
758 | local_irq_disable(); | |
759 | /* | |
760 | * Remove this CPU: | |
761 | */ | |
4f062896 | 762 | set_cpu_online(smp_processor_id(), false); |
d3ec5cae | 763 | disable_local_APIC(); |
8838eb6c | 764 | mcheck_cpu_clear(this_cpu_ptr(&cpu_info)); |
d3ec5cae | 765 | |
f23d74f6 TL |
766 | /* |
767 | * Use wbinvd on processors that support SME. This provides support | |
768 | * for performing a successful kexec when going from SME inactive | |
769 | * to SME active (or vice-versa). The cache must be cleared so that | |
770 | * if there are entries with the same physical address, both with and | |
771 | * without the encryption bit, they don't race each other when flushed | |
772 | * and potentially end up with the wrong entry being committed to | |
773 | * memory. | |
08f253ec ML |
774 | * |
775 | * Test the CPUID bit directly because the machine might've cleared | |
776 | * X86_FEATURE_SME due to cmdline options. | |
f23d74f6 | 777 | */ |
08f253ec | 778 | if (cpuid_eax(0x8000001f) & BIT(0)) |
f23d74f6 | 779 | native_wbinvd(); |
bba4ed01 TL |
780 | for (;;) { |
781 | /* | |
f23d74f6 TL |
782 | * Use native_halt() so that memory contents don't change |
783 | * (stack usage and variables) after possibly issuing the | |
784 | * native_wbinvd() above. | |
bba4ed01 | 785 | */ |
f23d74f6 | 786 | native_halt(); |
bba4ed01 | 787 | } |
7f424a8b PZ |
788 | } |
789 | ||
aa276e1c | 790 | /* |
07c94a38 BP |
791 | * AMD Erratum 400 aware idle routine. We handle it the same way as C3 power |
792 | * states (local apic timer and TSC stop). | |
58c644ba PZ |
793 | * |
794 | * XXX this function is completely buggered vs RCU and tracing. | |
aa276e1c | 795 | */ |
02c68a02 | 796 | static void amd_e400_idle(void) |
aa276e1c | 797 | { |
07c94a38 BP |
798 | /* |
799 | * We cannot use static_cpu_has_bug() here because X86_BUG_AMD_APIC_C1E | |
800 | * gets set after static_cpu_has() places have been converted via | |
801 | * alternatives. | |
802 | */ | |
803 | if (!boot_cpu_has_bug(X86_BUG_AMD_APIC_C1E)) { | |
804 | default_idle(); | |
805 | return; | |
aa276e1c TG |
806 | } |
807 | ||
07c94a38 | 808 | tick_broadcast_enter(); |
aa276e1c | 809 | |
07c94a38 | 810 | default_idle(); |
0beefa20 | 811 | |
07c94a38 | 812 | tick_broadcast_exit(); |
aa276e1c TG |
813 | } |
814 | ||
b253149b | 815 | /* |
aebef63c WK |
816 | * Prefer MWAIT over HALT if MWAIT is supported, MWAIT_CPUID leaf |
817 | * exists and whenever MONITOR/MWAIT extensions are present there is at | |
818 | * least one C1 substate. | |
b253149b | 819 | * |
aebef63c WK |
820 | * Do not prefer MWAIT if MONITOR instruction has a bug or idle=nomwait |
821 | * is passed to kernel commandline parameter. | |
b253149b LB |
822 | */ |
823 | static int prefer_mwait_c1_over_halt(const struct cpuinfo_x86 *c) | |
824 | { | |
aebef63c WK |
825 | u32 eax, ebx, ecx, edx; |
826 | ||
8bcedb4c WK |
827 | /* User has disallowed the use of MWAIT. Fallback to HALT */ |
828 | if (boot_option_idle_override == IDLE_NOMWAIT) | |
b253149b LB |
829 | return 0; |
830 | ||
aebef63c WK |
831 | /* MWAIT is not supported on this platform. Fallback to HALT */ |
832 | if (!cpu_has(c, X86_FEATURE_MWAIT)) | |
b253149b LB |
833 | return 0; |
834 | ||
aebef63c WK |
835 | /* Monitor has a bug. Fallback to HALT */ |
836 | if (boot_cpu_has_bug(X86_BUG_MONITOR)) | |
b253149b LB |
837 | return 0; |
838 | ||
aebef63c WK |
839 | cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx); |
840 | ||
841 | /* | |
842 | * If MWAIT extensions are not available, it is safe to use MWAIT | |
843 | * with EAX=0, ECX=0. | |
844 | */ | |
845 | if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED)) | |
846 | return 1; | |
847 | ||
848 | /* | |
849 | * If MWAIT extensions are available, there should be at least one | |
850 | * MWAIT C1 substate present. | |
851 | */ | |
852 | return (edx & MWAIT_C1_SUBSTATE_MASK); | |
b253149b LB |
853 | } |
854 | ||
855 | /* | |
0fb0328d HR |
856 | * MONITOR/MWAIT with no hints, used for default C1 state. This invokes MWAIT |
857 | * with interrupts enabled and no flags, which is backwards compatible with the | |
858 | * original MWAIT implementation. | |
b253149b | 859 | */ |
6727ad9e | 860 | static __cpuidle void mwait_idle(void) |
b253149b | 861 | { |
f8e617f4 MG |
862 | if (!current_set_polling_and_test()) { |
863 | if (this_cpu_has(X86_BUG_CLFLUSH_MONITOR)) { | |
ca59809f | 864 | mb(); /* quirk */ |
b253149b | 865 | clflush((void *)¤t_thread_info()->flags); |
ca59809f | 866 | mb(); /* quirk */ |
f8e617f4 | 867 | } |
b253149b LB |
868 | |
869 | __monitor((void *)¤t_thread_info()->flags, 0, 0); | |
89b30987 | 870 | if (!need_resched()) { |
b253149b | 871 | __sti_mwait(0, 0); |
89b30987 PZ |
872 | raw_local_irq_disable(); |
873 | } | |
f8e617f4 MG |
874 | } |
875 | __current_clr_polling(); | |
b253149b LB |
876 | } |
877 | ||
148f9bb8 | 878 | void select_idle_routine(const struct cpuinfo_x86 *c) |
7f424a8b | 879 | { |
3e5095d1 | 880 | #ifdef CONFIG_SMP |
7d1a9417 | 881 | if (boot_option_idle_override == IDLE_POLL && smp_num_siblings > 1) |
c767a54b | 882 | pr_warn_once("WARNING: polling idle and HT enabled, performance may degrade\n"); |
7f424a8b | 883 | #endif |
aaa3896b | 884 | if (x86_idle_set() || boot_option_idle_override == IDLE_POLL) |
6ddd2a27 TG |
885 | return; |
886 | ||
3344ed30 | 887 | if (boot_cpu_has_bug(X86_BUG_AMD_E400)) { |
c767a54b | 888 | pr_info("using AMD E400 aware idle routine\n"); |
aaa3896b | 889 | static_call_update(x86_idle, amd_e400_idle); |
b253149b LB |
890 | } else if (prefer_mwait_c1_over_halt(c)) { |
891 | pr_info("using mwait in idle threads\n"); | |
aaa3896b | 892 | static_call_update(x86_idle, mwait_idle); |
bfe6ed0c KS |
893 | } else if (cpu_feature_enabled(X86_FEATURE_TDX_GUEST)) { |
894 | pr_info("using TDX aware idle routine\n"); | |
aaa3896b | 895 | static_call_update(x86_idle, tdx_safe_halt); |
6ddd2a27 | 896 | } else |
aaa3896b | 897 | static_call_update(x86_idle, default_idle); |
7f424a8b PZ |
898 | } |
899 | ||
07c94a38 | 900 | void amd_e400_c1e_apic_setup(void) |
30e1e6d1 | 901 | { |
07c94a38 BP |
902 | if (boot_cpu_has_bug(X86_BUG_AMD_APIC_C1E)) { |
903 | pr_info("Switch to broadcast mode on CPU%d\n", smp_processor_id()); | |
904 | local_irq_disable(); | |
905 | tick_broadcast_force(); | |
906 | local_irq_enable(); | |
907 | } | |
30e1e6d1 RR |
908 | } |
909 | ||
e7ff3a47 TG |
910 | void __init arch_post_acpi_subsys_init(void) |
911 | { | |
912 | u32 lo, hi; | |
913 | ||
914 | if (!boot_cpu_has_bug(X86_BUG_AMD_E400)) | |
915 | return; | |
916 | ||
917 | /* | |
918 | * AMD E400 detection needs to happen after ACPI has been enabled. If | |
919 | * the machine is affected K8_INTP_C1E_ACTIVE_MASK bits are set in | |
920 | * MSR_K8_INT_PENDING_MSG. | |
921 | */ | |
922 | rdmsr(MSR_K8_INT_PENDING_MSG, lo, hi); | |
923 | if (!(lo & K8_INTP_C1E_ACTIVE_MASK)) | |
924 | return; | |
925 | ||
926 | boot_cpu_set_bug(X86_BUG_AMD_APIC_C1E); | |
927 | ||
928 | if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) | |
929 | mark_tsc_unstable("TSC halt in AMD C1E"); | |
930 | pr_info("System has AMD C1E enabled\n"); | |
931 | } | |
932 | ||
7f424a8b PZ |
933 | static int __init idle_setup(char *str) |
934 | { | |
ab6bc3e3 CG |
935 | if (!str) |
936 | return -EINVAL; | |
937 | ||
7f424a8b | 938 | if (!strcmp(str, "poll")) { |
c767a54b | 939 | pr_info("using polling idle threads\n"); |
d1896049 | 940 | boot_option_idle_override = IDLE_POLL; |
7d1a9417 | 941 | cpu_idle_poll_ctrl(true); |
d1896049 | 942 | } else if (!strcmp(str, "halt")) { |
c1e3b377 ZY |
943 | /* |
944 | * When the boot option of idle=halt is added, halt is | |
945 | * forced to be used for CPU idle. In such case CPU C2/C3 | |
946 | * won't be used again. | |
947 | * To continue to load the CPU idle driver, don't touch | |
948 | * the boot_option_idle_override. | |
949 | */ | |
aaa3896b | 950 | static_call_update(x86_idle, default_idle); |
d1896049 | 951 | boot_option_idle_override = IDLE_HALT; |
da5e09a1 ZY |
952 | } else if (!strcmp(str, "nomwait")) { |
953 | /* | |
954 | * If the boot option of "idle=nomwait" is added, | |
8bcedb4c WK |
955 | * it means that mwait will be disabled for CPU C1/C2/C3 |
956 | * states. | |
da5e09a1 | 957 | */ |
d1896049 | 958 | boot_option_idle_override = IDLE_NOMWAIT; |
c1e3b377 | 959 | } else |
7f424a8b PZ |
960 | return -1; |
961 | ||
7f424a8b PZ |
962 | return 0; |
963 | } | |
964 | early_param("idle", idle_setup); | |
965 | ||
9d62dcdf AW |
966 | unsigned long arch_align_stack(unsigned long sp) |
967 | { | |
968 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) | |
8032bf12 | 969 | sp -= get_random_u32_below(8192); |
9d62dcdf AW |
970 | return sp & ~0xf; |
971 | } | |
972 | ||
973 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
974 | { | |
9c6f0902 | 975 | return randomize_page(mm->brk, 0x02000000); |
9d62dcdf AW |
976 | } |
977 | ||
7ba78053 TG |
978 | /* |
979 | * Called from fs/proc with a reference on @p to find the function | |
980 | * which called into schedule(). This needs to be done carefully | |
981 | * because the task might wake up and we might look at a stack | |
982 | * changing under us. | |
983 | */ | |
42a20f86 | 984 | unsigned long __get_wchan(struct task_struct *p) |
7ba78053 | 985 | { |
5d1ceb39 PZ |
986 | struct unwind_state state; |
987 | unsigned long addr = 0; | |
7ba78053 | 988 | |
0dc636b3 PZ |
989 | if (!try_get_task_stack(p)) |
990 | return 0; | |
991 | ||
5d1ceb39 PZ |
992 | for (unwind_start(&state, p, NULL, NULL); !unwind_done(&state); |
993 | unwind_next_frame(&state)) { | |
994 | addr = unwind_get_return_address(&state); | |
995 | if (!addr) | |
996 | break; | |
997 | if (in_sched_functions(addr)) | |
998 | continue; | |
999 | break; | |
1000 | } | |
1001 | ||
0dc636b3 PZ |
1002 | put_task_stack(p); |
1003 | ||
5d1ceb39 | 1004 | return addr; |
7ba78053 | 1005 | } |
b0b9b014 | 1006 | |
f5c0b4f3 | 1007 | long do_arch_prctl_common(int option, unsigned long arg2) |
b0b9b014 | 1008 | { |
e9ea1e7f KH |
1009 | switch (option) { |
1010 | case ARCH_GET_CPUID: | |
1011 | return get_cpuid_mode(); | |
1012 | case ARCH_SET_CPUID: | |
f5c0b4f3 | 1013 | return set_cpuid_mode(arg2); |
db8268df CB |
1014 | case ARCH_GET_XCOMP_SUPP: |
1015 | case ARCH_GET_XCOMP_PERM: | |
1016 | case ARCH_REQ_XCOMP_PERM: | |
980fe2fd TG |
1017 | case ARCH_GET_XCOMP_GUEST_PERM: |
1018 | case ARCH_REQ_XCOMP_GUEST_PERM: | |
f5c0b4f3 | 1019 | return fpu_xstate_prctl(option, arg2); |
e9ea1e7f KH |
1020 | } |
1021 | ||
b0b9b014 KH |
1022 | return -EINVAL; |
1023 | } |