Commit | Line | Data |
---|---|---|
c767a54b JP |
1 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
2 | ||
61c4628b SS |
3 | #include <linux/errno.h> |
4 | #include <linux/kernel.h> | |
5 | #include <linux/mm.h> | |
6 | #include <linux/smp.h> | |
389d1fb1 | 7 | #include <linux/prctl.h> |
61c4628b SS |
8 | #include <linux/slab.h> |
9 | #include <linux/sched.h> | |
7f424a8b PZ |
10 | #include <linux/module.h> |
11 | #include <linux/pm.h> | |
aa276e1c | 12 | #include <linux/clockchips.h> |
9d62dcdf | 13 | #include <linux/random.h> |
7c68af6e | 14 | #include <linux/user-return-notifier.h> |
814e2c84 AI |
15 | #include <linux/dmi.h> |
16 | #include <linux/utsname.h> | |
90e24014 RW |
17 | #include <linux/stackprotector.h> |
18 | #include <linux/tick.h> | |
19 | #include <linux/cpuidle.h> | |
61613521 | 20 | #include <trace/events/power.h> |
24f1e32c | 21 | #include <linux/hw_breakpoint.h> |
93789b32 | 22 | #include <asm/cpu.h> |
d3ec5cae | 23 | #include <asm/apic.h> |
2c1b284e | 24 | #include <asm/syscalls.h> |
389d1fb1 JF |
25 | #include <asm/idle.h> |
26 | #include <asm/uaccess.h> | |
27 | #include <asm/i387.h> | |
1361b83a | 28 | #include <asm/fpu-internal.h> |
66cb5917 | 29 | #include <asm/debugreg.h> |
90e24014 | 30 | #include <asm/nmi.h> |
375074cc | 31 | #include <asm/tlbflush.h> |
90e24014 | 32 | |
45046892 TG |
33 | /* |
34 | * per-CPU TSS segments. Threads are completely 'soft' on Linux, | |
35 | * no more per-task TSS's. The TSS size is kept cacheline-aligned | |
36 | * so they are allowed to end up in the .data..cacheline_aligned | |
37 | * section. Since TSS's are completely CPU-local, we want them | |
38 | * on exact cacheline boundaries, to eliminate cacheline ping-pong. | |
39 | */ | |
24933b82 AL |
40 | __visible DEFINE_PER_CPU_SHARED_ALIGNED(struct tss_struct, cpu_tss) = INIT_TSS; |
41 | EXPORT_PER_CPU_SYMBOL_GPL(cpu_tss); | |
45046892 | 42 | |
90e24014 RW |
43 | #ifdef CONFIG_X86_64 |
44 | static DEFINE_PER_CPU(unsigned char, is_idle); | |
45 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); | |
46 | ||
47 | void idle_notifier_register(struct notifier_block *n) | |
48 | { | |
49 | atomic_notifier_chain_register(&idle_notifier, n); | |
50 | } | |
51 | EXPORT_SYMBOL_GPL(idle_notifier_register); | |
52 | ||
53 | void idle_notifier_unregister(struct notifier_block *n) | |
54 | { | |
55 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
56 | } | |
57 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
58 | #endif | |
c1e3b377 | 59 | |
aa283f49 | 60 | struct kmem_cache *task_xstate_cachep; |
5ee481da | 61 | EXPORT_SYMBOL_GPL(task_xstate_cachep); |
61c4628b | 62 | |
55ccf3fe SS |
63 | /* |
64 | * this gets called so that we can store lazy state into memory and copy the | |
65 | * current task into the new thread. | |
66 | */ | |
61c4628b SS |
67 | int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) |
68 | { | |
69 | *dst = *src; | |
f1853505 | 70 | |
dc56c0f9 | 71 | dst->thread.fpu_counter = 0; |
5e23fee2 ON |
72 | dst->thread.fpu.has_fpu = 0; |
73 | dst->thread.fpu.last_cpu = ~0; | |
74 | dst->thread.fpu.state = NULL; | |
f1853505 ON |
75 | if (tsk_used_math(src)) { |
76 | int err = fpu_alloc(&dst->thread.fpu); | |
77 | if (err) | |
78 | return err; | |
304bceda | 79 | fpu_copy(dst, src); |
aa283f49 | 80 | } |
61c4628b SS |
81 | return 0; |
82 | } | |
83 | ||
aa283f49 | 84 | void free_thread_xstate(struct task_struct *tsk) |
61c4628b | 85 | { |
86603283 | 86 | fpu_free(&tsk->thread.fpu); |
aa283f49 SS |
87 | } |
88 | ||
38e7c572 | 89 | void arch_release_task_struct(struct task_struct *tsk) |
aa283f49 | 90 | { |
38e7c572 | 91 | free_thread_xstate(tsk); |
61c4628b SS |
92 | } |
93 | ||
94 | void arch_task_cache_init(void) | |
95 | { | |
96 | task_xstate_cachep = | |
97 | kmem_cache_create("task_xstate", xstate_size, | |
98 | __alignof__(union thread_xstate), | |
2dff4405 | 99 | SLAB_PANIC | SLAB_NOTRACK, NULL); |
7496d645 | 100 | setup_xstate_comp(); |
61c4628b | 101 | } |
7f424a8b | 102 | |
389d1fb1 JF |
103 | /* |
104 | * Free current thread data structures etc.. | |
105 | */ | |
106 | void exit_thread(void) | |
107 | { | |
108 | struct task_struct *me = current; | |
109 | struct thread_struct *t = &me->thread; | |
250981e6 | 110 | unsigned long *bp = t->io_bitmap_ptr; |
389d1fb1 | 111 | |
250981e6 | 112 | if (bp) { |
24933b82 | 113 | struct tss_struct *tss = &per_cpu(cpu_tss, get_cpu()); |
389d1fb1 | 114 | |
389d1fb1 JF |
115 | t->io_bitmap_ptr = NULL; |
116 | clear_thread_flag(TIF_IO_BITMAP); | |
117 | /* | |
118 | * Careful, clear this in the TSS too: | |
119 | */ | |
120 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
121 | t->io_bitmap_max = 0; | |
122 | put_cpu(); | |
250981e6 | 123 | kfree(bp); |
389d1fb1 | 124 | } |
1dcc8d7b SS |
125 | |
126 | drop_fpu(me); | |
389d1fb1 JF |
127 | } |
128 | ||
129 | void flush_thread(void) | |
130 | { | |
131 | struct task_struct *tsk = current; | |
132 | ||
24f1e32c | 133 | flush_ptrace_hw_breakpoint(tsk); |
389d1fb1 | 134 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
304bceda SS |
135 | drop_init_fpu(tsk); |
136 | /* | |
137 | * Free the FPU state for non xsave platforms. They get reallocated | |
138 | * lazily at the first use. | |
139 | */ | |
5d2bd700 | 140 | if (!use_eager_fpu()) |
304bceda | 141 | free_thread_xstate(tsk); |
389d1fb1 JF |
142 | } |
143 | ||
144 | static void hard_disable_TSC(void) | |
145 | { | |
375074cc | 146 | cr4_set_bits(X86_CR4_TSD); |
389d1fb1 JF |
147 | } |
148 | ||
149 | void disable_TSC(void) | |
150 | { | |
151 | preempt_disable(); | |
152 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
153 | /* | |
154 | * Must flip the CPU state synchronously with | |
155 | * TIF_NOTSC in the current running context. | |
156 | */ | |
157 | hard_disable_TSC(); | |
158 | preempt_enable(); | |
159 | } | |
160 | ||
161 | static void hard_enable_TSC(void) | |
162 | { | |
375074cc | 163 | cr4_clear_bits(X86_CR4_TSD); |
389d1fb1 JF |
164 | } |
165 | ||
166 | static void enable_TSC(void) | |
167 | { | |
168 | preempt_disable(); | |
169 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
170 | /* | |
171 | * Must flip the CPU state synchronously with | |
172 | * TIF_NOTSC in the current running context. | |
173 | */ | |
174 | hard_enable_TSC(); | |
175 | preempt_enable(); | |
176 | } | |
177 | ||
178 | int get_tsc_mode(unsigned long adr) | |
179 | { | |
180 | unsigned int val; | |
181 | ||
182 | if (test_thread_flag(TIF_NOTSC)) | |
183 | val = PR_TSC_SIGSEGV; | |
184 | else | |
185 | val = PR_TSC_ENABLE; | |
186 | ||
187 | return put_user(val, (unsigned int __user *)adr); | |
188 | } | |
189 | ||
190 | int set_tsc_mode(unsigned int val) | |
191 | { | |
192 | if (val == PR_TSC_SIGSEGV) | |
193 | disable_TSC(); | |
194 | else if (val == PR_TSC_ENABLE) | |
195 | enable_TSC(); | |
196 | else | |
197 | return -EINVAL; | |
198 | ||
199 | return 0; | |
200 | } | |
201 | ||
202 | void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |
203 | struct tss_struct *tss) | |
204 | { | |
205 | struct thread_struct *prev, *next; | |
206 | ||
207 | prev = &prev_p->thread; | |
208 | next = &next_p->thread; | |
209 | ||
ea8e61b7 PZ |
210 | if (test_tsk_thread_flag(prev_p, TIF_BLOCKSTEP) ^ |
211 | test_tsk_thread_flag(next_p, TIF_BLOCKSTEP)) { | |
212 | unsigned long debugctl = get_debugctlmsr(); | |
213 | ||
214 | debugctl &= ~DEBUGCTLMSR_BTF; | |
215 | if (test_tsk_thread_flag(next_p, TIF_BLOCKSTEP)) | |
216 | debugctl |= DEBUGCTLMSR_BTF; | |
217 | ||
218 | update_debugctlmsr(debugctl); | |
219 | } | |
389d1fb1 | 220 | |
389d1fb1 JF |
221 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
222 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
223 | /* prev and next are different */ | |
224 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
225 | hard_disable_TSC(); | |
226 | else | |
227 | hard_enable_TSC(); | |
228 | } | |
229 | ||
230 | if (test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { | |
231 | /* | |
232 | * Copy the relevant range of the IO bitmap. | |
233 | * Normally this is 128 bytes or less: | |
234 | */ | |
235 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
236 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
237 | } else if (test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) { | |
238 | /* | |
239 | * Clear any possible leftover bits: | |
240 | */ | |
241 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
242 | } | |
7c68af6e | 243 | propagate_user_return_notify(prev_p, next_p); |
389d1fb1 JF |
244 | } |
245 | ||
00dba564 TG |
246 | /* |
247 | * Idle related variables and functions | |
248 | */ | |
d1896049 | 249 | unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE; |
00dba564 TG |
250 | EXPORT_SYMBOL(boot_option_idle_override); |
251 | ||
a476bda3 | 252 | static void (*x86_idle)(void); |
00dba564 | 253 | |
90e24014 RW |
254 | #ifndef CONFIG_SMP |
255 | static inline void play_dead(void) | |
256 | { | |
257 | BUG(); | |
258 | } | |
259 | #endif | |
260 | ||
261 | #ifdef CONFIG_X86_64 | |
262 | void enter_idle(void) | |
263 | { | |
c6ae41e7 | 264 | this_cpu_write(is_idle, 1); |
90e24014 RW |
265 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
266 | } | |
267 | ||
268 | static void __exit_idle(void) | |
269 | { | |
270 | if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) | |
271 | return; | |
272 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); | |
273 | } | |
274 | ||
275 | /* Called from interrupts to signify idle end */ | |
276 | void exit_idle(void) | |
277 | { | |
278 | /* idle loop has pid 0 */ | |
279 | if (current->pid) | |
280 | return; | |
281 | __exit_idle(); | |
282 | } | |
283 | #endif | |
284 | ||
7d1a9417 TG |
285 | void arch_cpu_idle_enter(void) |
286 | { | |
287 | local_touch_nmi(); | |
288 | enter_idle(); | |
289 | } | |
90e24014 | 290 | |
7d1a9417 TG |
291 | void arch_cpu_idle_exit(void) |
292 | { | |
293 | __exit_idle(); | |
294 | } | |
90e24014 | 295 | |
7d1a9417 TG |
296 | void arch_cpu_idle_dead(void) |
297 | { | |
298 | play_dead(); | |
299 | } | |
90e24014 | 300 | |
7d1a9417 TG |
301 | /* |
302 | * Called from the generic idle code. | |
303 | */ | |
304 | void arch_cpu_idle(void) | |
305 | { | |
16f8b05a | 306 | x86_idle(); |
90e24014 RW |
307 | } |
308 | ||
00dba564 | 309 | /* |
7d1a9417 | 310 | * We use this if we don't have any better idle routine.. |
00dba564 TG |
311 | */ |
312 | void default_idle(void) | |
313 | { | |
4d0e42cc | 314 | trace_cpu_idle_rcuidle(1, smp_processor_id()); |
7d1a9417 | 315 | safe_halt(); |
4d0e42cc | 316 | trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id()); |
00dba564 | 317 | } |
60b8b1de | 318 | #ifdef CONFIG_APM_MODULE |
00dba564 TG |
319 | EXPORT_SYMBOL(default_idle); |
320 | #endif | |
321 | ||
6a377ddc LB |
322 | #ifdef CONFIG_XEN |
323 | bool xen_set_default_idle(void) | |
e5fd47bf | 324 | { |
a476bda3 | 325 | bool ret = !!x86_idle; |
e5fd47bf | 326 | |
a476bda3 | 327 | x86_idle = default_idle; |
e5fd47bf KRW |
328 | |
329 | return ret; | |
330 | } | |
6a377ddc | 331 | #endif |
d3ec5cae IV |
332 | void stop_this_cpu(void *dummy) |
333 | { | |
334 | local_irq_disable(); | |
335 | /* | |
336 | * Remove this CPU: | |
337 | */ | |
4f062896 | 338 | set_cpu_online(smp_processor_id(), false); |
d3ec5cae IV |
339 | disable_local_APIC(); |
340 | ||
27be4570 LB |
341 | for (;;) |
342 | halt(); | |
7f424a8b PZ |
343 | } |
344 | ||
02c68a02 LB |
345 | bool amd_e400_c1e_detected; |
346 | EXPORT_SYMBOL(amd_e400_c1e_detected); | |
aa276e1c | 347 | |
02c68a02 | 348 | static cpumask_var_t amd_e400_c1e_mask; |
4faac97d | 349 | |
02c68a02 | 350 | void amd_e400_remove_cpu(int cpu) |
4faac97d | 351 | { |
02c68a02 LB |
352 | if (amd_e400_c1e_mask != NULL) |
353 | cpumask_clear_cpu(cpu, amd_e400_c1e_mask); | |
4faac97d TG |
354 | } |
355 | ||
aa276e1c | 356 | /* |
02c68a02 | 357 | * AMD Erratum 400 aware idle routine. We check for C1E active in the interrupt |
aa276e1c TG |
358 | * pending message MSR. If we detect C1E, then we handle it the same |
359 | * way as C3 power states (local apic timer and TSC stop) | |
360 | */ | |
02c68a02 | 361 | static void amd_e400_idle(void) |
aa276e1c | 362 | { |
02c68a02 | 363 | if (!amd_e400_c1e_detected) { |
aa276e1c TG |
364 | u32 lo, hi; |
365 | ||
366 | rdmsr(MSR_K8_INT_PENDING_MSG, lo, hi); | |
e8c534ec | 367 | |
aa276e1c | 368 | if (lo & K8_INTP_C1E_ACTIVE_MASK) { |
02c68a02 | 369 | amd_e400_c1e_detected = true; |
40fb1715 | 370 | if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) |
09bfeea1 | 371 | mark_tsc_unstable("TSC halt in AMD C1E"); |
c767a54b | 372 | pr_info("System has AMD C1E enabled\n"); |
aa276e1c TG |
373 | } |
374 | } | |
375 | ||
02c68a02 | 376 | if (amd_e400_c1e_detected) { |
aa276e1c TG |
377 | int cpu = smp_processor_id(); |
378 | ||
02c68a02 LB |
379 | if (!cpumask_test_cpu(cpu, amd_e400_c1e_mask)) { |
380 | cpumask_set_cpu(cpu, amd_e400_c1e_mask); | |
0beefa20 | 381 | /* |
f833bab8 | 382 | * Force broadcast so ACPI can not interfere. |
0beefa20 | 383 | */ |
aa276e1c TG |
384 | clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_FORCE, |
385 | &cpu); | |
c767a54b | 386 | pr_info("Switch to broadcast mode on CPU%d\n", cpu); |
aa276e1c TG |
387 | } |
388 | clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu); | |
0beefa20 | 389 | |
aa276e1c | 390 | default_idle(); |
0beefa20 TG |
391 | |
392 | /* | |
393 | * The switch back from broadcast mode needs to be | |
394 | * called with interrupts disabled. | |
395 | */ | |
ea811747 PZ |
396 | local_irq_disable(); |
397 | clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu); | |
398 | local_irq_enable(); | |
aa276e1c TG |
399 | } else |
400 | default_idle(); | |
401 | } | |
402 | ||
148f9bb8 | 403 | void select_idle_routine(const struct cpuinfo_x86 *c) |
7f424a8b | 404 | { |
3e5095d1 | 405 | #ifdef CONFIG_SMP |
7d1a9417 | 406 | if (boot_option_idle_override == IDLE_POLL && smp_num_siblings > 1) |
c767a54b | 407 | pr_warn_once("WARNING: polling idle and HT enabled, performance may degrade\n"); |
7f424a8b | 408 | #endif |
7d1a9417 | 409 | if (x86_idle || boot_option_idle_override == IDLE_POLL) |
6ddd2a27 TG |
410 | return; |
411 | ||
7d7dc116 | 412 | if (cpu_has_bug(c, X86_BUG_AMD_APIC_C1E)) { |
9d8888c2 | 413 | /* E400: APIC timer interrupt does not wake up CPU from C1e */ |
c767a54b | 414 | pr_info("using AMD E400 aware idle routine\n"); |
a476bda3 | 415 | x86_idle = amd_e400_idle; |
6ddd2a27 | 416 | } else |
a476bda3 | 417 | x86_idle = default_idle; |
7f424a8b PZ |
418 | } |
419 | ||
02c68a02 | 420 | void __init init_amd_e400_c1e_mask(void) |
30e1e6d1 | 421 | { |
02c68a02 | 422 | /* If we're using amd_e400_idle, we need to allocate amd_e400_c1e_mask. */ |
a476bda3 | 423 | if (x86_idle == amd_e400_idle) |
02c68a02 | 424 | zalloc_cpumask_var(&amd_e400_c1e_mask, GFP_KERNEL); |
30e1e6d1 RR |
425 | } |
426 | ||
7f424a8b PZ |
427 | static int __init idle_setup(char *str) |
428 | { | |
ab6bc3e3 CG |
429 | if (!str) |
430 | return -EINVAL; | |
431 | ||
7f424a8b | 432 | if (!strcmp(str, "poll")) { |
c767a54b | 433 | pr_info("using polling idle threads\n"); |
d1896049 | 434 | boot_option_idle_override = IDLE_POLL; |
7d1a9417 | 435 | cpu_idle_poll_ctrl(true); |
d1896049 | 436 | } else if (!strcmp(str, "halt")) { |
c1e3b377 ZY |
437 | /* |
438 | * When the boot option of idle=halt is added, halt is | |
439 | * forced to be used for CPU idle. In such case CPU C2/C3 | |
440 | * won't be used again. | |
441 | * To continue to load the CPU idle driver, don't touch | |
442 | * the boot_option_idle_override. | |
443 | */ | |
a476bda3 | 444 | x86_idle = default_idle; |
d1896049 | 445 | boot_option_idle_override = IDLE_HALT; |
da5e09a1 ZY |
446 | } else if (!strcmp(str, "nomwait")) { |
447 | /* | |
448 | * If the boot option of "idle=nomwait" is added, | |
449 | * it means that mwait will be disabled for CPU C2/C3 | |
450 | * states. In such case it won't touch the variable | |
451 | * of boot_option_idle_override. | |
452 | */ | |
d1896049 | 453 | boot_option_idle_override = IDLE_NOMWAIT; |
c1e3b377 | 454 | } else |
7f424a8b PZ |
455 | return -1; |
456 | ||
7f424a8b PZ |
457 | return 0; |
458 | } | |
459 | early_param("idle", idle_setup); | |
460 | ||
9d62dcdf AW |
461 | unsigned long arch_align_stack(unsigned long sp) |
462 | { | |
463 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) | |
464 | sp -= get_random_int() % 8192; | |
465 | return sp & ~0xf; | |
466 | } | |
467 | ||
468 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
469 | { | |
470 | unsigned long range_end = mm->brk + 0x02000000; | |
471 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
472 | } | |
473 |