]>
Commit | Line | Data |
---|---|---|
296af7c9 BS |
1 | /* |
2 | * QEMU System Emulator | |
3 | * | |
4 | * Copyright (c) 2003-2008 Fabrice Bellard | |
5 | * | |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
24 | ||
25 | /* Needed early for CONFIG_BSD etc. */ | |
7b31bbc2 | 26 | #include "qemu/osdep.h" |
33c11879 PB |
27 | #include "qemu-common.h" |
28 | #include "cpu.h" | |
83c9089e | 29 | #include "monitor/monitor.h" |
a4e15de9 | 30 | #include "qapi/qmp/qerror.h" |
d49b6836 | 31 | #include "qemu/error-report.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
da31d594 | 33 | #include "sysemu/block-backend.h" |
022c62cb | 34 | #include "exec/gdbstub.h" |
9c17d615 PB |
35 | #include "sysemu/dma.h" |
36 | #include "sysemu/kvm.h" | |
de0b36b6 | 37 | #include "qmp-commands.h" |
63c91552 | 38 | #include "exec/exec-all.h" |
296af7c9 | 39 | |
1de7afc9 | 40 | #include "qemu/thread.h" |
9c17d615 PB |
41 | #include "sysemu/cpus.h" |
42 | #include "sysemu/qtest.h" | |
1de7afc9 PB |
43 | #include "qemu/main-loop.h" |
44 | #include "qemu/bitmap.h" | |
cb365646 | 45 | #include "qemu/seqlock.h" |
a4e15de9 | 46 | #include "qapi-event.h" |
9cb805fd | 47 | #include "hw/nmi.h" |
8b427044 | 48 | #include "sysemu/replay.h" |
0ff0fc19 JK |
49 | |
50 | #ifndef _WIN32 | |
1de7afc9 | 51 | #include "qemu/compatfd.h" |
0ff0fc19 | 52 | #endif |
296af7c9 | 53 | |
6d9cb73c JK |
54 | #ifdef CONFIG_LINUX |
55 | ||
56 | #include <sys/prctl.h> | |
57 | ||
c0532a76 MT |
58 | #ifndef PR_MCE_KILL |
59 | #define PR_MCE_KILL 33 | |
60 | #endif | |
61 | ||
6d9cb73c JK |
62 | #ifndef PR_MCE_KILL_SET |
63 | #define PR_MCE_KILL_SET 1 | |
64 | #endif | |
65 | ||
66 | #ifndef PR_MCE_KILL_EARLY | |
67 | #define PR_MCE_KILL_EARLY 1 | |
68 | #endif | |
69 | ||
70 | #endif /* CONFIG_LINUX */ | |
71 | ||
27498bef ST |
72 | int64_t max_delay; |
73 | int64_t max_advance; | |
296af7c9 | 74 | |
2adcc85d JH |
75 | /* vcpu throttling controls */ |
76 | static QEMUTimer *throttle_timer; | |
77 | static unsigned int throttle_percentage; | |
78 | ||
79 | #define CPU_THROTTLE_PCT_MIN 1 | |
80 | #define CPU_THROTTLE_PCT_MAX 99 | |
81 | #define CPU_THROTTLE_TIMESLICE_NS 10000000 | |
82 | ||
321bc0b2 TC |
83 | bool cpu_is_stopped(CPUState *cpu) |
84 | { | |
85 | return cpu->stopped || !runstate_is_running(); | |
86 | } | |
87 | ||
a98ae1d8 | 88 | static bool cpu_thread_is_idle(CPUState *cpu) |
ac873f1e | 89 | { |
c64ca814 | 90 | if (cpu->stop || cpu->queued_work_first) { |
ac873f1e PM |
91 | return false; |
92 | } | |
321bc0b2 | 93 | if (cpu_is_stopped(cpu)) { |
ac873f1e PM |
94 | return true; |
95 | } | |
8c2e1b00 | 96 | if (!cpu->halted || cpu_has_work(cpu) || |
215e79c0 | 97 | kvm_halt_in_kernel()) { |
ac873f1e PM |
98 | return false; |
99 | } | |
100 | return true; | |
101 | } | |
102 | ||
103 | static bool all_cpu_threads_idle(void) | |
104 | { | |
182735ef | 105 | CPUState *cpu; |
ac873f1e | 106 | |
bdc44640 | 107 | CPU_FOREACH(cpu) { |
182735ef | 108 | if (!cpu_thread_is_idle(cpu)) { |
ac873f1e PM |
109 | return false; |
110 | } | |
111 | } | |
112 | return true; | |
113 | } | |
114 | ||
946fb27c PB |
115 | /***********************************************************/ |
116 | /* guest cycle counter */ | |
117 | ||
a3270e19 PB |
118 | /* Protected by TimersState seqlock */ |
119 | ||
5045e9d9 | 120 | static bool icount_sleep = true; |
71468395 | 121 | static int64_t vm_clock_warp_start = -1; |
946fb27c PB |
122 | /* Conversion factor from emulated instructions to virtual clock ticks. */ |
123 | static int icount_time_shift; | |
124 | /* Arbitrarily pick 1MIPS as the minimum allowable speed. */ | |
125 | #define MAX_ICOUNT_SHIFT 10 | |
a3270e19 | 126 | |
946fb27c PB |
127 | static QEMUTimer *icount_rt_timer; |
128 | static QEMUTimer *icount_vm_timer; | |
129 | static QEMUTimer *icount_warp_timer; | |
946fb27c PB |
130 | |
131 | typedef struct TimersState { | |
cb365646 | 132 | /* Protected by BQL. */ |
946fb27c PB |
133 | int64_t cpu_ticks_prev; |
134 | int64_t cpu_ticks_offset; | |
cb365646 LPF |
135 | |
136 | /* cpu_clock_offset can be read out of BQL, so protect it with | |
137 | * this lock. | |
138 | */ | |
139 | QemuSeqLock vm_clock_seqlock; | |
946fb27c PB |
140 | int64_t cpu_clock_offset; |
141 | int32_t cpu_ticks_enabled; | |
142 | int64_t dummy; | |
c96778bb FK |
143 | |
144 | /* Compensate for varying guest execution speed. */ | |
145 | int64_t qemu_icount_bias; | |
146 | /* Only written by TCG thread */ | |
147 | int64_t qemu_icount; | |
946fb27c PB |
148 | } TimersState; |
149 | ||
d9cd4007 | 150 | static TimersState timers_state; |
946fb27c | 151 | |
2a62914b | 152 | int64_t cpu_get_icount_raw(void) |
946fb27c PB |
153 | { |
154 | int64_t icount; | |
4917cf44 | 155 | CPUState *cpu = current_cpu; |
946fb27c | 156 | |
c96778bb | 157 | icount = timers_state.qemu_icount; |
4917cf44 | 158 | if (cpu) { |
414b15c9 | 159 | if (!cpu->can_do_io) { |
2a62914b PD |
160 | fprintf(stderr, "Bad icount read\n"); |
161 | exit(1); | |
946fb27c | 162 | } |
28ecfd7a | 163 | icount -= (cpu->icount_decr.u16.low + cpu->icount_extra); |
946fb27c | 164 | } |
2a62914b PD |
165 | return icount; |
166 | } | |
167 | ||
168 | /* Return the virtual CPU time, based on the instruction counter. */ | |
169 | static int64_t cpu_get_icount_locked(void) | |
170 | { | |
171 | int64_t icount = cpu_get_icount_raw(); | |
3f031313 | 172 | return timers_state.qemu_icount_bias + cpu_icount_to_ns(icount); |
946fb27c PB |
173 | } |
174 | ||
17a15f1b PB |
175 | int64_t cpu_get_icount(void) |
176 | { | |
177 | int64_t icount; | |
178 | unsigned start; | |
179 | ||
180 | do { | |
181 | start = seqlock_read_begin(&timers_state.vm_clock_seqlock); | |
182 | icount = cpu_get_icount_locked(); | |
183 | } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start)); | |
184 | ||
185 | return icount; | |
186 | } | |
187 | ||
3f031313 FK |
188 | int64_t cpu_icount_to_ns(int64_t icount) |
189 | { | |
190 | return icount << icount_time_shift; | |
191 | } | |
192 | ||
d90f3cca C |
193 | /* return the time elapsed in VM between vm_start and vm_stop. Unless |
194 | * icount is active, cpu_get_ticks() uses units of the host CPU cycle | |
195 | * counter. | |
196 | * | |
197 | * Caller must hold the BQL | |
198 | */ | |
946fb27c PB |
199 | int64_t cpu_get_ticks(void) |
200 | { | |
5f3e3101 PB |
201 | int64_t ticks; |
202 | ||
946fb27c PB |
203 | if (use_icount) { |
204 | return cpu_get_icount(); | |
205 | } | |
5f3e3101 PB |
206 | |
207 | ticks = timers_state.cpu_ticks_offset; | |
208 | if (timers_state.cpu_ticks_enabled) { | |
4a7428c5 | 209 | ticks += cpu_get_host_ticks(); |
5f3e3101 PB |
210 | } |
211 | ||
212 | if (timers_state.cpu_ticks_prev > ticks) { | |
213 | /* Note: non increasing ticks may happen if the host uses | |
214 | software suspend */ | |
215 | timers_state.cpu_ticks_offset += timers_state.cpu_ticks_prev - ticks; | |
216 | ticks = timers_state.cpu_ticks_prev; | |
946fb27c | 217 | } |
5f3e3101 PB |
218 | |
219 | timers_state.cpu_ticks_prev = ticks; | |
220 | return ticks; | |
946fb27c PB |
221 | } |
222 | ||
cb365646 | 223 | static int64_t cpu_get_clock_locked(void) |
946fb27c | 224 | { |
1d45cea5 | 225 | int64_t time; |
cb365646 | 226 | |
1d45cea5 | 227 | time = timers_state.cpu_clock_offset; |
5f3e3101 | 228 | if (timers_state.cpu_ticks_enabled) { |
1d45cea5 | 229 | time += get_clock(); |
946fb27c | 230 | } |
cb365646 | 231 | |
1d45cea5 | 232 | return time; |
cb365646 LPF |
233 | } |
234 | ||
d90f3cca | 235 | /* Return the monotonic time elapsed in VM, i.e., |
8212ff86 PM |
236 | * the time between vm_start and vm_stop |
237 | */ | |
cb365646 LPF |
238 | int64_t cpu_get_clock(void) |
239 | { | |
240 | int64_t ti; | |
241 | unsigned start; | |
242 | ||
243 | do { | |
244 | start = seqlock_read_begin(&timers_state.vm_clock_seqlock); | |
245 | ti = cpu_get_clock_locked(); | |
246 | } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, start)); | |
247 | ||
248 | return ti; | |
946fb27c PB |
249 | } |
250 | ||
cb365646 | 251 | /* enable cpu_get_ticks() |
3224e878 | 252 | * Caller must hold BQL which serves as mutex for vm_clock_seqlock. |
cb365646 | 253 | */ |
946fb27c PB |
254 | void cpu_enable_ticks(void) |
255 | { | |
cb365646 | 256 | /* Here, the really thing protected by seqlock is cpu_clock_offset. */ |
03719e44 | 257 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
946fb27c | 258 | if (!timers_state.cpu_ticks_enabled) { |
4a7428c5 | 259 | timers_state.cpu_ticks_offset -= cpu_get_host_ticks(); |
946fb27c PB |
260 | timers_state.cpu_clock_offset -= get_clock(); |
261 | timers_state.cpu_ticks_enabled = 1; | |
262 | } | |
03719e44 | 263 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
946fb27c PB |
264 | } |
265 | ||
266 | /* disable cpu_get_ticks() : the clock is stopped. You must not call | |
cb365646 | 267 | * cpu_get_ticks() after that. |
3224e878 | 268 | * Caller must hold BQL which serves as mutex for vm_clock_seqlock. |
cb365646 | 269 | */ |
946fb27c PB |
270 | void cpu_disable_ticks(void) |
271 | { | |
cb365646 | 272 | /* Here, the really thing protected by seqlock is cpu_clock_offset. */ |
03719e44 | 273 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
946fb27c | 274 | if (timers_state.cpu_ticks_enabled) { |
4a7428c5 | 275 | timers_state.cpu_ticks_offset += cpu_get_host_ticks(); |
cb365646 | 276 | timers_state.cpu_clock_offset = cpu_get_clock_locked(); |
946fb27c PB |
277 | timers_state.cpu_ticks_enabled = 0; |
278 | } | |
03719e44 | 279 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
946fb27c PB |
280 | } |
281 | ||
282 | /* Correlation between real and virtual time is always going to be | |
283 | fairly approximate, so ignore small variation. | |
284 | When the guest is idle real and virtual time will be aligned in | |
285 | the IO wait loop. */ | |
73bcb24d | 286 | #define ICOUNT_WOBBLE (NANOSECONDS_PER_SECOND / 10) |
946fb27c PB |
287 | |
288 | static void icount_adjust(void) | |
289 | { | |
290 | int64_t cur_time; | |
291 | int64_t cur_icount; | |
292 | int64_t delta; | |
a3270e19 PB |
293 | |
294 | /* Protected by TimersState mutex. */ | |
946fb27c | 295 | static int64_t last_delta; |
468cc7cf | 296 | |
946fb27c PB |
297 | /* If the VM is not running, then do nothing. */ |
298 | if (!runstate_is_running()) { | |
299 | return; | |
300 | } | |
468cc7cf | 301 | |
03719e44 | 302 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
17a15f1b PB |
303 | cur_time = cpu_get_clock_locked(); |
304 | cur_icount = cpu_get_icount_locked(); | |
468cc7cf | 305 | |
946fb27c PB |
306 | delta = cur_icount - cur_time; |
307 | /* FIXME: This is a very crude algorithm, somewhat prone to oscillation. */ | |
308 | if (delta > 0 | |
309 | && last_delta + ICOUNT_WOBBLE < delta * 2 | |
310 | && icount_time_shift > 0) { | |
311 | /* The guest is getting too far ahead. Slow time down. */ | |
312 | icount_time_shift--; | |
313 | } | |
314 | if (delta < 0 | |
315 | && last_delta - ICOUNT_WOBBLE > delta * 2 | |
316 | && icount_time_shift < MAX_ICOUNT_SHIFT) { | |
317 | /* The guest is getting too far behind. Speed time up. */ | |
318 | icount_time_shift++; | |
319 | } | |
320 | last_delta = delta; | |
c96778bb FK |
321 | timers_state.qemu_icount_bias = cur_icount |
322 | - (timers_state.qemu_icount << icount_time_shift); | |
03719e44 | 323 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
946fb27c PB |
324 | } |
325 | ||
326 | static void icount_adjust_rt(void *opaque) | |
327 | { | |
40daca54 | 328 | timer_mod(icount_rt_timer, |
1979b908 | 329 | qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL_RT) + 1000); |
946fb27c PB |
330 | icount_adjust(); |
331 | } | |
332 | ||
333 | static void icount_adjust_vm(void *opaque) | |
334 | { | |
40daca54 AB |
335 | timer_mod(icount_vm_timer, |
336 | qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + | |
73bcb24d | 337 | NANOSECONDS_PER_SECOND / 10); |
946fb27c PB |
338 | icount_adjust(); |
339 | } | |
340 | ||
341 | static int64_t qemu_icount_round(int64_t count) | |
342 | { | |
343 | return (count + (1 << icount_time_shift) - 1) >> icount_time_shift; | |
344 | } | |
345 | ||
efab87cf | 346 | static void icount_warp_rt(void) |
946fb27c | 347 | { |
ccffff48 AB |
348 | unsigned seq; |
349 | int64_t warp_start; | |
350 | ||
17a15f1b PB |
351 | /* The icount_warp_timer is rescheduled soon after vm_clock_warp_start |
352 | * changes from -1 to another value, so the race here is okay. | |
353 | */ | |
ccffff48 AB |
354 | do { |
355 | seq = seqlock_read_begin(&timers_state.vm_clock_seqlock); | |
356 | warp_start = vm_clock_warp_start; | |
357 | } while (seqlock_read_retry(&timers_state.vm_clock_seqlock, seq)); | |
358 | ||
359 | if (warp_start == -1) { | |
946fb27c PB |
360 | return; |
361 | } | |
362 | ||
03719e44 | 363 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
946fb27c | 364 | if (runstate_is_running()) { |
8eda206e PD |
365 | int64_t clock = REPLAY_CLOCK(REPLAY_CLOCK_VIRTUAL_RT, |
366 | cpu_get_clock_locked()); | |
8ed961d9 PB |
367 | int64_t warp_delta; |
368 | ||
369 | warp_delta = clock - vm_clock_warp_start; | |
370 | if (use_icount == 2) { | |
946fb27c | 371 | /* |
40daca54 | 372 | * In adaptive mode, do not let QEMU_CLOCK_VIRTUAL run too |
946fb27c PB |
373 | * far ahead of real time. |
374 | */ | |
17a15f1b | 375 | int64_t cur_icount = cpu_get_icount_locked(); |
bf2a7ddb | 376 | int64_t delta = clock - cur_icount; |
8ed961d9 | 377 | warp_delta = MIN(warp_delta, delta); |
946fb27c | 378 | } |
c96778bb | 379 | timers_state.qemu_icount_bias += warp_delta; |
946fb27c PB |
380 | } |
381 | vm_clock_warp_start = -1; | |
03719e44 | 382 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
8ed961d9 PB |
383 | |
384 | if (qemu_clock_expired(QEMU_CLOCK_VIRTUAL)) { | |
385 | qemu_clock_notify(QEMU_CLOCK_VIRTUAL); | |
386 | } | |
946fb27c PB |
387 | } |
388 | ||
e76d1798 | 389 | static void icount_timer_cb(void *opaque) |
efab87cf | 390 | { |
e76d1798 PD |
391 | /* No need for a checkpoint because the timer already synchronizes |
392 | * with CHECKPOINT_CLOCK_VIRTUAL_RT. | |
393 | */ | |
394 | icount_warp_rt(); | |
efab87cf PD |
395 | } |
396 | ||
8156be56 PB |
397 | void qtest_clock_warp(int64_t dest) |
398 | { | |
40daca54 | 399 | int64_t clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL); |
efef88b3 | 400 | AioContext *aio_context; |
8156be56 | 401 | assert(qtest_enabled()); |
efef88b3 | 402 | aio_context = qemu_get_aio_context(); |
8156be56 | 403 | while (clock < dest) { |
40daca54 | 404 | int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL); |
c9299e2f | 405 | int64_t warp = qemu_soonest_timeout(dest - clock, deadline); |
efef88b3 | 406 | |
03719e44 | 407 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
c96778bb | 408 | timers_state.qemu_icount_bias += warp; |
03719e44 | 409 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
17a15f1b | 410 | |
40daca54 | 411 | qemu_clock_run_timers(QEMU_CLOCK_VIRTUAL); |
efef88b3 | 412 | timerlist_run_timers(aio_context->tlg.tl[QEMU_CLOCK_VIRTUAL]); |
40daca54 | 413 | clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL); |
8156be56 | 414 | } |
40daca54 | 415 | qemu_clock_notify(QEMU_CLOCK_VIRTUAL); |
8156be56 PB |
416 | } |
417 | ||
e76d1798 | 418 | void qemu_start_warp_timer(void) |
946fb27c | 419 | { |
ce78d18c | 420 | int64_t clock; |
946fb27c PB |
421 | int64_t deadline; |
422 | ||
e76d1798 | 423 | if (!use_icount) { |
946fb27c PB |
424 | return; |
425 | } | |
426 | ||
8bd7f71d PD |
427 | /* Nothing to do if the VM is stopped: QEMU_CLOCK_VIRTUAL timers |
428 | * do not fire, so computing the deadline does not make sense. | |
429 | */ | |
430 | if (!runstate_is_running()) { | |
431 | return; | |
432 | } | |
433 | ||
434 | /* warp clock deterministically in record/replay mode */ | |
e76d1798 | 435 | if (!replay_checkpoint(CHECKPOINT_CLOCK_WARP_START)) { |
8bd7f71d PD |
436 | return; |
437 | } | |
438 | ||
ce78d18c | 439 | if (!all_cpu_threads_idle()) { |
946fb27c PB |
440 | return; |
441 | } | |
442 | ||
8156be56 PB |
443 | if (qtest_enabled()) { |
444 | /* When testing, qtest commands advance icount. */ | |
e76d1798 | 445 | return; |
8156be56 PB |
446 | } |
447 | ||
ac70aafc | 448 | /* We want to use the earliest deadline from ALL vm_clocks */ |
bf2a7ddb | 449 | clock = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT); |
40daca54 | 450 | deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL); |
ce78d18c | 451 | if (deadline < 0) { |
d7a0f71d VC |
452 | static bool notified; |
453 | if (!icount_sleep && !notified) { | |
454 | error_report("WARNING: icount sleep disabled and no active timers"); | |
455 | notified = true; | |
456 | } | |
ce78d18c | 457 | return; |
ac70aafc AB |
458 | } |
459 | ||
946fb27c PB |
460 | if (deadline > 0) { |
461 | /* | |
40daca54 | 462 | * Ensure QEMU_CLOCK_VIRTUAL proceeds even when the virtual CPU goes to |
946fb27c PB |
463 | * sleep. Otherwise, the CPU might be waiting for a future timer |
464 | * interrupt to wake it up, but the interrupt never comes because | |
465 | * the vCPU isn't running any insns and thus doesn't advance the | |
40daca54 | 466 | * QEMU_CLOCK_VIRTUAL. |
946fb27c | 467 | */ |
5045e9d9 VC |
468 | if (!icount_sleep) { |
469 | /* | |
470 | * We never let VCPUs sleep in no sleep icount mode. | |
471 | * If there is a pending QEMU_CLOCK_VIRTUAL timer we just advance | |
472 | * to the next QEMU_CLOCK_VIRTUAL event and notify it. | |
473 | * It is useful when we want a deterministic execution time, | |
474 | * isolated from host latencies. | |
475 | */ | |
03719e44 | 476 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
5045e9d9 | 477 | timers_state.qemu_icount_bias += deadline; |
03719e44 | 478 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
5045e9d9 VC |
479 | qemu_clock_notify(QEMU_CLOCK_VIRTUAL); |
480 | } else { | |
481 | /* | |
482 | * We do stop VCPUs and only advance QEMU_CLOCK_VIRTUAL after some | |
483 | * "real" time, (related to the time left until the next event) has | |
484 | * passed. The QEMU_CLOCK_VIRTUAL_RT clock will do this. | |
485 | * This avoids that the warps are visible externally; for example, | |
486 | * you will not be sending network packets continuously instead of | |
487 | * every 100ms. | |
488 | */ | |
03719e44 | 489 | seqlock_write_begin(&timers_state.vm_clock_seqlock); |
5045e9d9 VC |
490 | if (vm_clock_warp_start == -1 || vm_clock_warp_start > clock) { |
491 | vm_clock_warp_start = clock; | |
492 | } | |
03719e44 | 493 | seqlock_write_end(&timers_state.vm_clock_seqlock); |
5045e9d9 | 494 | timer_mod_anticipate(icount_warp_timer, clock + deadline); |
ce78d18c | 495 | } |
ac70aafc | 496 | } else if (deadline == 0) { |
40daca54 | 497 | qemu_clock_notify(QEMU_CLOCK_VIRTUAL); |
946fb27c PB |
498 | } |
499 | } | |
500 | ||
e76d1798 PD |
501 | static void qemu_account_warp_timer(void) |
502 | { | |
503 | if (!use_icount || !icount_sleep) { | |
504 | return; | |
505 | } | |
506 | ||
507 | /* Nothing to do if the VM is stopped: QEMU_CLOCK_VIRTUAL timers | |
508 | * do not fire, so computing the deadline does not make sense. | |
509 | */ | |
510 | if (!runstate_is_running()) { | |
511 | return; | |
512 | } | |
513 | ||
514 | /* warp clock deterministically in record/replay mode */ | |
515 | if (!replay_checkpoint(CHECKPOINT_CLOCK_WARP_ACCOUNT)) { | |
516 | return; | |
517 | } | |
518 | ||
519 | timer_del(icount_warp_timer); | |
520 | icount_warp_rt(); | |
521 | } | |
522 | ||
d09eae37 FK |
523 | static bool icount_state_needed(void *opaque) |
524 | { | |
525 | return use_icount; | |
526 | } | |
527 | ||
528 | /* | |
529 | * This is a subsection for icount migration. | |
530 | */ | |
531 | static const VMStateDescription icount_vmstate_timers = { | |
532 | .name = "timer/icount", | |
533 | .version_id = 1, | |
534 | .minimum_version_id = 1, | |
5cd8cada | 535 | .needed = icount_state_needed, |
d09eae37 FK |
536 | .fields = (VMStateField[]) { |
537 | VMSTATE_INT64(qemu_icount_bias, TimersState), | |
538 | VMSTATE_INT64(qemu_icount, TimersState), | |
539 | VMSTATE_END_OF_LIST() | |
540 | } | |
541 | }; | |
542 | ||
946fb27c PB |
543 | static const VMStateDescription vmstate_timers = { |
544 | .name = "timer", | |
545 | .version_id = 2, | |
546 | .minimum_version_id = 1, | |
35d08458 | 547 | .fields = (VMStateField[]) { |
946fb27c PB |
548 | VMSTATE_INT64(cpu_ticks_offset, TimersState), |
549 | VMSTATE_INT64(dummy, TimersState), | |
550 | VMSTATE_INT64_V(cpu_clock_offset, TimersState, 2), | |
551 | VMSTATE_END_OF_LIST() | |
d09eae37 | 552 | }, |
5cd8cada JQ |
553 | .subsections = (const VMStateDescription*[]) { |
554 | &icount_vmstate_timers, | |
555 | NULL | |
946fb27c PB |
556 | } |
557 | }; | |
558 | ||
e0eeb4a2 | 559 | static void cpu_throttle_thread(CPUState *cpu, void *opaque) |
2adcc85d | 560 | { |
2adcc85d JH |
561 | double pct; |
562 | double throttle_ratio; | |
563 | long sleeptime_ns; | |
564 | ||
565 | if (!cpu_throttle_get_percentage()) { | |
566 | return; | |
567 | } | |
568 | ||
569 | pct = (double)cpu_throttle_get_percentage()/100; | |
570 | throttle_ratio = pct / (1 - pct); | |
571 | sleeptime_ns = (long)(throttle_ratio * CPU_THROTTLE_TIMESLICE_NS); | |
572 | ||
573 | qemu_mutex_unlock_iothread(); | |
574 | atomic_set(&cpu->throttle_thread_scheduled, 0); | |
575 | g_usleep(sleeptime_ns / 1000); /* Convert ns to us for usleep call */ | |
576 | qemu_mutex_lock_iothread(); | |
577 | } | |
578 | ||
579 | static void cpu_throttle_timer_tick(void *opaque) | |
580 | { | |
581 | CPUState *cpu; | |
582 | double pct; | |
583 | ||
584 | /* Stop the timer if needed */ | |
585 | if (!cpu_throttle_get_percentage()) { | |
586 | return; | |
587 | } | |
588 | CPU_FOREACH(cpu) { | |
589 | if (!atomic_xchg(&cpu->throttle_thread_scheduled, 1)) { | |
e0eeb4a2 | 590 | async_run_on_cpu(cpu, cpu_throttle_thread, NULL); |
2adcc85d JH |
591 | } |
592 | } | |
593 | ||
594 | pct = (double)cpu_throttle_get_percentage()/100; | |
595 | timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) + | |
596 | CPU_THROTTLE_TIMESLICE_NS / (1-pct)); | |
597 | } | |
598 | ||
599 | void cpu_throttle_set(int new_throttle_pct) | |
600 | { | |
601 | /* Ensure throttle percentage is within valid range */ | |
602 | new_throttle_pct = MIN(new_throttle_pct, CPU_THROTTLE_PCT_MAX); | |
603 | new_throttle_pct = MAX(new_throttle_pct, CPU_THROTTLE_PCT_MIN); | |
604 | ||
605 | atomic_set(&throttle_percentage, new_throttle_pct); | |
606 | ||
607 | timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) + | |
608 | CPU_THROTTLE_TIMESLICE_NS); | |
609 | } | |
610 | ||
611 | void cpu_throttle_stop(void) | |
612 | { | |
613 | atomic_set(&throttle_percentage, 0); | |
614 | } | |
615 | ||
616 | bool cpu_throttle_active(void) | |
617 | { | |
618 | return (cpu_throttle_get_percentage() != 0); | |
619 | } | |
620 | ||
621 | int cpu_throttle_get_percentage(void) | |
622 | { | |
623 | return atomic_read(&throttle_percentage); | |
624 | } | |
625 | ||
4603ea01 PD |
626 | void cpu_ticks_init(void) |
627 | { | |
ccdb3c1f | 628 | seqlock_init(&timers_state.vm_clock_seqlock); |
4603ea01 | 629 | vmstate_register(NULL, 0, &vmstate_timers, &timers_state); |
2adcc85d JH |
630 | throttle_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL_RT, |
631 | cpu_throttle_timer_tick, NULL); | |
4603ea01 PD |
632 | } |
633 | ||
1ad9580b | 634 | void configure_icount(QemuOpts *opts, Error **errp) |
946fb27c | 635 | { |
1ad9580b | 636 | const char *option; |
a8bfac37 | 637 | char *rem_str = NULL; |
1ad9580b | 638 | |
1ad9580b | 639 | option = qemu_opt_get(opts, "shift"); |
946fb27c | 640 | if (!option) { |
a8bfac37 ST |
641 | if (qemu_opt_get(opts, "align") != NULL) { |
642 | error_setg(errp, "Please specify shift option when using align"); | |
643 | } | |
946fb27c PB |
644 | return; |
645 | } | |
f1f4b57e VC |
646 | |
647 | icount_sleep = qemu_opt_get_bool(opts, "sleep", true); | |
5045e9d9 VC |
648 | if (icount_sleep) { |
649 | icount_warp_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL_RT, | |
e76d1798 | 650 | icount_timer_cb, NULL); |
5045e9d9 | 651 | } |
f1f4b57e | 652 | |
a8bfac37 | 653 | icount_align_option = qemu_opt_get_bool(opts, "align", false); |
f1f4b57e VC |
654 | |
655 | if (icount_align_option && !icount_sleep) { | |
778d9f9b | 656 | error_setg(errp, "align=on and sleep=off are incompatible"); |
f1f4b57e | 657 | } |
946fb27c | 658 | if (strcmp(option, "auto") != 0) { |
a8bfac37 ST |
659 | errno = 0; |
660 | icount_time_shift = strtol(option, &rem_str, 0); | |
661 | if (errno != 0 || *rem_str != '\0' || !strlen(option)) { | |
662 | error_setg(errp, "icount: Invalid shift value"); | |
663 | } | |
946fb27c PB |
664 | use_icount = 1; |
665 | return; | |
a8bfac37 ST |
666 | } else if (icount_align_option) { |
667 | error_setg(errp, "shift=auto and align=on are incompatible"); | |
f1f4b57e | 668 | } else if (!icount_sleep) { |
778d9f9b | 669 | error_setg(errp, "shift=auto and sleep=off are incompatible"); |
946fb27c PB |
670 | } |
671 | ||
672 | use_icount = 2; | |
673 | ||
674 | /* 125MIPS seems a reasonable initial guess at the guest speed. | |
675 | It will be corrected fairly quickly anyway. */ | |
676 | icount_time_shift = 3; | |
677 | ||
678 | /* Have both realtime and virtual time triggers for speed adjustment. | |
679 | The realtime trigger catches emulated time passing too slowly, | |
680 | the virtual time trigger catches emulated time passing too fast. | |
681 | Realtime triggers occur even when idle, so use them less frequently | |
682 | than VM triggers. */ | |
bf2a7ddb PD |
683 | icount_rt_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL_RT, |
684 | icount_adjust_rt, NULL); | |
40daca54 | 685 | timer_mod(icount_rt_timer, |
bf2a7ddb | 686 | qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL_RT) + 1000); |
40daca54 AB |
687 | icount_vm_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, |
688 | icount_adjust_vm, NULL); | |
689 | timer_mod(icount_vm_timer, | |
690 | qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + | |
73bcb24d | 691 | NANOSECONDS_PER_SECOND / 10); |
946fb27c PB |
692 | } |
693 | ||
296af7c9 BS |
694 | /***********************************************************/ |
695 | void hw_error(const char *fmt, ...) | |
696 | { | |
697 | va_list ap; | |
55e5c285 | 698 | CPUState *cpu; |
296af7c9 BS |
699 | |
700 | va_start(ap, fmt); | |
701 | fprintf(stderr, "qemu: hardware error: "); | |
702 | vfprintf(stderr, fmt, ap); | |
703 | fprintf(stderr, "\n"); | |
bdc44640 | 704 | CPU_FOREACH(cpu) { |
55e5c285 | 705 | fprintf(stderr, "CPU #%d:\n", cpu->cpu_index); |
878096ee | 706 | cpu_dump_state(cpu, stderr, fprintf, CPU_DUMP_FPU); |
296af7c9 BS |
707 | } |
708 | va_end(ap); | |
709 | abort(); | |
710 | } | |
711 | ||
712 | void cpu_synchronize_all_states(void) | |
713 | { | |
182735ef | 714 | CPUState *cpu; |
296af7c9 | 715 | |
bdc44640 | 716 | CPU_FOREACH(cpu) { |
182735ef | 717 | cpu_synchronize_state(cpu); |
296af7c9 BS |
718 | } |
719 | } | |
720 | ||
721 | void cpu_synchronize_all_post_reset(void) | |
722 | { | |
182735ef | 723 | CPUState *cpu; |
296af7c9 | 724 | |
bdc44640 | 725 | CPU_FOREACH(cpu) { |
182735ef | 726 | cpu_synchronize_post_reset(cpu); |
296af7c9 BS |
727 | } |
728 | } | |
729 | ||
730 | void cpu_synchronize_all_post_init(void) | |
731 | { | |
182735ef | 732 | CPUState *cpu; |
296af7c9 | 733 | |
bdc44640 | 734 | CPU_FOREACH(cpu) { |
182735ef | 735 | cpu_synchronize_post_init(cpu); |
296af7c9 BS |
736 | } |
737 | } | |
738 | ||
56983463 | 739 | static int do_vm_stop(RunState state) |
296af7c9 | 740 | { |
56983463 KW |
741 | int ret = 0; |
742 | ||
1354869c | 743 | if (runstate_is_running()) { |
296af7c9 | 744 | cpu_disable_ticks(); |
296af7c9 | 745 | pause_all_vcpus(); |
f5bbfba1 | 746 | runstate_set(state); |
1dfb4dd9 | 747 | vm_state_notify(0, state); |
a4e15de9 | 748 | qapi_event_send_stop(&error_abort); |
296af7c9 | 749 | } |
56983463 | 750 | |
594a45ce | 751 | bdrv_drain_all(); |
6d0ceb80 | 752 | replay_disable_events(); |
22af08ea | 753 | ret = bdrv_flush_all(); |
594a45ce | 754 | |
56983463 | 755 | return ret; |
296af7c9 BS |
756 | } |
757 | ||
a1fcaa73 | 758 | static bool cpu_can_run(CPUState *cpu) |
296af7c9 | 759 | { |
4fdeee7c | 760 | if (cpu->stop) { |
a1fcaa73 | 761 | return false; |
0ab07c62 | 762 | } |
321bc0b2 | 763 | if (cpu_is_stopped(cpu)) { |
a1fcaa73 | 764 | return false; |
0ab07c62 | 765 | } |
a1fcaa73 | 766 | return true; |
296af7c9 BS |
767 | } |
768 | ||
91325046 | 769 | static void cpu_handle_guest_debug(CPUState *cpu) |
83f338f7 | 770 | { |
64f6b346 | 771 | gdb_set_stop_cpu(cpu); |
8cf71710 | 772 | qemu_system_debug_request(); |
f324e766 | 773 | cpu->stopped = true; |
3c638d06 JK |
774 | } |
775 | ||
6d9cb73c JK |
776 | #ifdef CONFIG_LINUX |
777 | static void sigbus_reraise(void) | |
778 | { | |
779 | sigset_t set; | |
780 | struct sigaction action; | |
781 | ||
782 | memset(&action, 0, sizeof(action)); | |
783 | action.sa_handler = SIG_DFL; | |
784 | if (!sigaction(SIGBUS, &action, NULL)) { | |
785 | raise(SIGBUS); | |
786 | sigemptyset(&set); | |
787 | sigaddset(&set, SIGBUS); | |
a2d1761d | 788 | pthread_sigmask(SIG_UNBLOCK, &set, NULL); |
6d9cb73c JK |
789 | } |
790 | perror("Failed to re-raise SIGBUS!\n"); | |
791 | abort(); | |
792 | } | |
793 | ||
794 | static void sigbus_handler(int n, struct qemu_signalfd_siginfo *siginfo, | |
795 | void *ctx) | |
796 | { | |
797 | if (kvm_on_sigbus(siginfo->ssi_code, | |
798 | (void *)(intptr_t)siginfo->ssi_addr)) { | |
799 | sigbus_reraise(); | |
800 | } | |
801 | } | |
802 | ||
803 | static void qemu_init_sigbus(void) | |
804 | { | |
805 | struct sigaction action; | |
806 | ||
807 | memset(&action, 0, sizeof(action)); | |
808 | action.sa_flags = SA_SIGINFO; | |
809 | action.sa_sigaction = (void (*)(int, siginfo_t*, void*))sigbus_handler; | |
810 | sigaction(SIGBUS, &action, NULL); | |
811 | ||
812 | prctl(PR_MCE_KILL, PR_MCE_KILL_SET, PR_MCE_KILL_EARLY, 0, 0); | |
813 | } | |
814 | ||
290adf38 | 815 | static void qemu_kvm_eat_signals(CPUState *cpu) |
1ab3c6c0 JK |
816 | { |
817 | struct timespec ts = { 0, 0 }; | |
818 | siginfo_t siginfo; | |
819 | sigset_t waitset; | |
820 | sigset_t chkset; | |
821 | int r; | |
822 | ||
823 | sigemptyset(&waitset); | |
824 | sigaddset(&waitset, SIG_IPI); | |
825 | sigaddset(&waitset, SIGBUS); | |
826 | ||
827 | do { | |
828 | r = sigtimedwait(&waitset, &siginfo, &ts); | |
829 | if (r == -1 && !(errno == EAGAIN || errno == EINTR)) { | |
830 | perror("sigtimedwait"); | |
831 | exit(1); | |
832 | } | |
833 | ||
834 | switch (r) { | |
835 | case SIGBUS: | |
290adf38 | 836 | if (kvm_on_sigbus_vcpu(cpu, siginfo.si_code, siginfo.si_addr)) { |
1ab3c6c0 JK |
837 | sigbus_reraise(); |
838 | } | |
839 | break; | |
840 | default: | |
841 | break; | |
842 | } | |
843 | ||
844 | r = sigpending(&chkset); | |
845 | if (r == -1) { | |
846 | perror("sigpending"); | |
847 | exit(1); | |
848 | } | |
849 | } while (sigismember(&chkset, SIG_IPI) || sigismember(&chkset, SIGBUS)); | |
1ab3c6c0 JK |
850 | } |
851 | ||
6d9cb73c JK |
852 | #else /* !CONFIG_LINUX */ |
853 | ||
854 | static void qemu_init_sigbus(void) | |
855 | { | |
856 | } | |
1ab3c6c0 | 857 | |
290adf38 | 858 | static void qemu_kvm_eat_signals(CPUState *cpu) |
1ab3c6c0 JK |
859 | { |
860 | } | |
6d9cb73c JK |
861 | #endif /* !CONFIG_LINUX */ |
862 | ||
296af7c9 | 863 | #ifndef _WIN32 |
55f8d6ac JK |
864 | static void dummy_signal(int sig) |
865 | { | |
866 | } | |
55f8d6ac | 867 | |
13618e05 | 868 | static void qemu_kvm_init_cpu_signals(CPUState *cpu) |
714bd040 PB |
869 | { |
870 | int r; | |
871 | sigset_t set; | |
872 | struct sigaction sigact; | |
873 | ||
874 | memset(&sigact, 0, sizeof(sigact)); | |
875 | sigact.sa_handler = dummy_signal; | |
876 | sigaction(SIG_IPI, &sigact, NULL); | |
877 | ||
714bd040 PB |
878 | pthread_sigmask(SIG_BLOCK, NULL, &set); |
879 | sigdelset(&set, SIG_IPI); | |
714bd040 | 880 | sigdelset(&set, SIGBUS); |
491d6e80 | 881 | r = kvm_set_signal_mask(cpu, &set); |
714bd040 PB |
882 | if (r) { |
883 | fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r)); | |
884 | exit(1); | |
885 | } | |
886 | } | |
887 | ||
55f8d6ac | 888 | #else /* _WIN32 */ |
13618e05 | 889 | static void qemu_kvm_init_cpu_signals(CPUState *cpu) |
ff48eb5f | 890 | { |
714bd040 PB |
891 | abort(); |
892 | } | |
714bd040 | 893 | #endif /* _WIN32 */ |
ff48eb5f | 894 | |
b2532d88 | 895 | static QemuMutex qemu_global_mutex; |
46daff13 | 896 | static QemuCond qemu_io_proceeded_cond; |
6b49809c | 897 | static unsigned iothread_requesting_mutex; |
296af7c9 BS |
898 | |
899 | static QemuThread io_thread; | |
900 | ||
296af7c9 BS |
901 | /* cpu creation */ |
902 | static QemuCond qemu_cpu_cond; | |
903 | /* system init */ | |
296af7c9 BS |
904 | static QemuCond qemu_pause_cond; |
905 | ||
d3b12f5d | 906 | void qemu_init_cpu_loop(void) |
296af7c9 | 907 | { |
6d9cb73c | 908 | qemu_init_sigbus(); |
ed94592b | 909 | qemu_cond_init(&qemu_cpu_cond); |
ed94592b | 910 | qemu_cond_init(&qemu_pause_cond); |
46daff13 | 911 | qemu_cond_init(&qemu_io_proceeded_cond); |
296af7c9 | 912 | qemu_mutex_init(&qemu_global_mutex); |
296af7c9 | 913 | |
b7680cb6 | 914 | qemu_thread_get_self(&io_thread); |
296af7c9 BS |
915 | } |
916 | ||
e0eeb4a2 | 917 | void run_on_cpu(CPUState *cpu, run_on_cpu_func func, void *data) |
e82bcec2 | 918 | { |
d148d90e | 919 | do_run_on_cpu(cpu, func, data, &qemu_global_mutex); |
3c02270d CV |
920 | } |
921 | ||
4c055ab5 GZ |
922 | static void qemu_kvm_destroy_vcpu(CPUState *cpu) |
923 | { | |
924 | if (kvm_destroy_vcpu(cpu) < 0) { | |
925 | error_report("kvm_destroy_vcpu failed"); | |
926 | exit(EXIT_FAILURE); | |
927 | } | |
928 | } | |
929 | ||
930 | static void qemu_tcg_destroy_vcpu(CPUState *cpu) | |
931 | { | |
932 | } | |
933 | ||
509a0d78 | 934 | static void qemu_wait_io_event_common(CPUState *cpu) |
296af7c9 | 935 | { |
4fdeee7c AF |
936 | if (cpu->stop) { |
937 | cpu->stop = false; | |
f324e766 | 938 | cpu->stopped = true; |
96bce683 | 939 | qemu_cond_broadcast(&qemu_pause_cond); |
296af7c9 | 940 | } |
a5403c69 | 941 | process_queued_cpu_work(cpu); |
216fc9a4 | 942 | cpu->thread_kicked = false; |
296af7c9 BS |
943 | } |
944 | ||
d5f8d613 | 945 | static void qemu_tcg_wait_io_event(CPUState *cpu) |
296af7c9 | 946 | { |
16400322 | 947 | while (all_cpu_threads_idle()) { |
d5f8d613 | 948 | qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex); |
16400322 | 949 | } |
296af7c9 | 950 | |
46daff13 PB |
951 | while (iothread_requesting_mutex) { |
952 | qemu_cond_wait(&qemu_io_proceeded_cond, &qemu_global_mutex); | |
953 | } | |
6cabe1f3 | 954 | |
bdc44640 | 955 | CPU_FOREACH(cpu) { |
182735ef | 956 | qemu_wait_io_event_common(cpu); |
6cabe1f3 | 957 | } |
296af7c9 BS |
958 | } |
959 | ||
fd529e8f | 960 | static void qemu_kvm_wait_io_event(CPUState *cpu) |
296af7c9 | 961 | { |
a98ae1d8 | 962 | while (cpu_thread_is_idle(cpu)) { |
f5c121b8 | 963 | qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex); |
16400322 | 964 | } |
296af7c9 | 965 | |
290adf38 | 966 | qemu_kvm_eat_signals(cpu); |
509a0d78 | 967 | qemu_wait_io_event_common(cpu); |
296af7c9 BS |
968 | } |
969 | ||
7e97cd88 | 970 | static void *qemu_kvm_cpu_thread_fn(void *arg) |
296af7c9 | 971 | { |
48a106bd | 972 | CPUState *cpu = arg; |
84b4915d | 973 | int r; |
296af7c9 | 974 | |
ab28bd23 PB |
975 | rcu_register_thread(); |
976 | ||
2e7f7a3c | 977 | qemu_mutex_lock_iothread(); |
814e612e | 978 | qemu_thread_get_self(cpu->thread); |
9f09e18a | 979 | cpu->thread_id = qemu_get_thread_id(); |
626cf8f4 | 980 | cpu->can_do_io = 1; |
4917cf44 | 981 | current_cpu = cpu; |
296af7c9 | 982 | |
504134d2 | 983 | r = kvm_init_vcpu(cpu); |
84b4915d JK |
984 | if (r < 0) { |
985 | fprintf(stderr, "kvm_init_vcpu failed: %s\n", strerror(-r)); | |
986 | exit(1); | |
987 | } | |
296af7c9 | 988 | |
13618e05 | 989 | qemu_kvm_init_cpu_signals(cpu); |
296af7c9 BS |
990 | |
991 | /* signal CPU creation */ | |
61a46217 | 992 | cpu->created = true; |
296af7c9 BS |
993 | qemu_cond_signal(&qemu_cpu_cond); |
994 | ||
4c055ab5 | 995 | do { |
a1fcaa73 | 996 | if (cpu_can_run(cpu)) { |
1458c363 | 997 | r = kvm_cpu_exec(cpu); |
83f338f7 | 998 | if (r == EXCP_DEBUG) { |
91325046 | 999 | cpu_handle_guest_debug(cpu); |
83f338f7 | 1000 | } |
0ab07c62 | 1001 | } |
fd529e8f | 1002 | qemu_kvm_wait_io_event(cpu); |
4c055ab5 | 1003 | } while (!cpu->unplug || cpu_can_run(cpu)); |
296af7c9 | 1004 | |
4c055ab5 | 1005 | qemu_kvm_destroy_vcpu(cpu); |
2c579042 BR |
1006 | cpu->created = false; |
1007 | qemu_cond_signal(&qemu_cpu_cond); | |
4c055ab5 | 1008 | qemu_mutex_unlock_iothread(); |
296af7c9 BS |
1009 | return NULL; |
1010 | } | |
1011 | ||
c7f0f3b1 AL |
1012 | static void *qemu_dummy_cpu_thread_fn(void *arg) |
1013 | { | |
1014 | #ifdef _WIN32 | |
1015 | fprintf(stderr, "qtest is not supported under Windows\n"); | |
1016 | exit(1); | |
1017 | #else | |
10a9021d | 1018 | CPUState *cpu = arg; |
c7f0f3b1 AL |
1019 | sigset_t waitset; |
1020 | int r; | |
1021 | ||
ab28bd23 PB |
1022 | rcu_register_thread(); |
1023 | ||
c7f0f3b1 | 1024 | qemu_mutex_lock_iothread(); |
814e612e | 1025 | qemu_thread_get_self(cpu->thread); |
9f09e18a | 1026 | cpu->thread_id = qemu_get_thread_id(); |
626cf8f4 | 1027 | cpu->can_do_io = 1; |
c7f0f3b1 AL |
1028 | |
1029 | sigemptyset(&waitset); | |
1030 | sigaddset(&waitset, SIG_IPI); | |
1031 | ||
1032 | /* signal CPU creation */ | |
61a46217 | 1033 | cpu->created = true; |
c7f0f3b1 AL |
1034 | qemu_cond_signal(&qemu_cpu_cond); |
1035 | ||
4917cf44 | 1036 | current_cpu = cpu; |
c7f0f3b1 | 1037 | while (1) { |
4917cf44 | 1038 | current_cpu = NULL; |
c7f0f3b1 AL |
1039 | qemu_mutex_unlock_iothread(); |
1040 | do { | |
1041 | int sig; | |
1042 | r = sigwait(&waitset, &sig); | |
1043 | } while (r == -1 && (errno == EAGAIN || errno == EINTR)); | |
1044 | if (r == -1) { | |
1045 | perror("sigwait"); | |
1046 | exit(1); | |
1047 | } | |
1048 | qemu_mutex_lock_iothread(); | |
4917cf44 | 1049 | current_cpu = cpu; |
509a0d78 | 1050 | qemu_wait_io_event_common(cpu); |
c7f0f3b1 AL |
1051 | } |
1052 | ||
1053 | return NULL; | |
1054 | #endif | |
1055 | } | |
1056 | ||
1be7fcb8 AB |
1057 | static int64_t tcg_get_icount_limit(void) |
1058 | { | |
1059 | int64_t deadline; | |
1060 | ||
1061 | if (replay_mode != REPLAY_MODE_PLAY) { | |
1062 | deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL); | |
1063 | ||
1064 | /* Maintain prior (possibly buggy) behaviour where if no deadline | |
1065 | * was set (as there is no QEMU_CLOCK_VIRTUAL timer) or it is more than | |
1066 | * INT32_MAX nanoseconds ahead, we still use INT32_MAX | |
1067 | * nanoseconds. | |
1068 | */ | |
1069 | if ((deadline < 0) || (deadline > INT32_MAX)) { | |
1070 | deadline = INT32_MAX; | |
1071 | } | |
1072 | ||
1073 | return qemu_icount_round(deadline); | |
1074 | } else { | |
1075 | return replay_get_instructions(); | |
1076 | } | |
1077 | } | |
1078 | ||
1079 | static int tcg_cpu_exec(CPUState *cpu) | |
1080 | { | |
1081 | int ret; | |
1082 | #ifdef CONFIG_PROFILER | |
1083 | int64_t ti; | |
1084 | #endif | |
1085 | ||
1086 | #ifdef CONFIG_PROFILER | |
1087 | ti = profile_getclock(); | |
1088 | #endif | |
1089 | if (use_icount) { | |
1090 | int64_t count; | |
1091 | int decr; | |
1092 | timers_state.qemu_icount -= (cpu->icount_decr.u16.low | |
1093 | + cpu->icount_extra); | |
1094 | cpu->icount_decr.u16.low = 0; | |
1095 | cpu->icount_extra = 0; | |
1096 | count = tcg_get_icount_limit(); | |
1097 | timers_state.qemu_icount += count; | |
1098 | decr = (count > 0xffff) ? 0xffff : count; | |
1099 | count -= decr; | |
1100 | cpu->icount_decr.u16.low = decr; | |
1101 | cpu->icount_extra = count; | |
1102 | } | |
1103 | cpu_exec_start(cpu); | |
1104 | ret = cpu_exec(cpu); | |
1105 | cpu_exec_end(cpu); | |
1106 | #ifdef CONFIG_PROFILER | |
1107 | tcg_time += profile_getclock() - ti; | |
1108 | #endif | |
1109 | if (use_icount) { | |
1110 | /* Fold pending instructions back into the | |
1111 | instruction counter, and clear the interrupt flag. */ | |
1112 | timers_state.qemu_icount -= (cpu->icount_decr.u16.low | |
1113 | + cpu->icount_extra); | |
1114 | cpu->icount_decr.u32 = 0; | |
1115 | cpu->icount_extra = 0; | |
1116 | replay_account_executed_instructions(); | |
1117 | } | |
1118 | return ret; | |
1119 | } | |
1120 | ||
c93bbbef AB |
1121 | /* Destroy any remaining vCPUs which have been unplugged and have |
1122 | * finished running | |
1123 | */ | |
1124 | static void deal_with_unplugged_cpus(void) | |
1be7fcb8 | 1125 | { |
c93bbbef | 1126 | CPUState *cpu; |
1be7fcb8 | 1127 | |
c93bbbef AB |
1128 | CPU_FOREACH(cpu) { |
1129 | if (cpu->unplug && !cpu_can_run(cpu)) { | |
1130 | qemu_tcg_destroy_vcpu(cpu); | |
1131 | cpu->created = false; | |
1132 | qemu_cond_signal(&qemu_cpu_cond); | |
1be7fcb8 AB |
1133 | break; |
1134 | } | |
1135 | } | |
1be7fcb8 | 1136 | } |
bdb7ca67 | 1137 | |
7e97cd88 | 1138 | static void *qemu_tcg_cpu_thread_fn(void *arg) |
296af7c9 | 1139 | { |
c3586ba7 | 1140 | CPUState *cpu = arg; |
296af7c9 | 1141 | |
ab28bd23 PB |
1142 | rcu_register_thread(); |
1143 | ||
2e7f7a3c | 1144 | qemu_mutex_lock_iothread(); |
814e612e | 1145 | qemu_thread_get_self(cpu->thread); |
296af7c9 | 1146 | |
38fcbd3f AF |
1147 | CPU_FOREACH(cpu) { |
1148 | cpu->thread_id = qemu_get_thread_id(); | |
1149 | cpu->created = true; | |
626cf8f4 | 1150 | cpu->can_do_io = 1; |
38fcbd3f | 1151 | } |
296af7c9 BS |
1152 | qemu_cond_signal(&qemu_cpu_cond); |
1153 | ||
fa7d1867 | 1154 | /* wait for initial kick-off after machine start */ |
c28e399c | 1155 | while (first_cpu->stopped) { |
d5f8d613 | 1156 | qemu_cond_wait(first_cpu->halt_cond, &qemu_global_mutex); |
8e564b4e JK |
1157 | |
1158 | /* process any pending work */ | |
bdc44640 | 1159 | CPU_FOREACH(cpu) { |
182735ef | 1160 | qemu_wait_io_event_common(cpu); |
8e564b4e | 1161 | } |
0ab07c62 | 1162 | } |
296af7c9 | 1163 | |
21618b3e | 1164 | /* process any pending work */ |
aed807c8 | 1165 | atomic_mb_set(&exit_request, 1); |
21618b3e | 1166 | |
c93bbbef AB |
1167 | cpu = first_cpu; |
1168 | ||
296af7c9 | 1169 | while (1) { |
c93bbbef AB |
1170 | /* Account partial waits to QEMU_CLOCK_VIRTUAL. */ |
1171 | qemu_account_warp_timer(); | |
1172 | ||
1173 | if (!cpu) { | |
1174 | cpu = first_cpu; | |
1175 | } | |
1176 | ||
1177 | for (; cpu != NULL && !exit_request; cpu = CPU_NEXT(cpu)) { | |
1178 | ||
1179 | qemu_clock_enable(QEMU_CLOCK_VIRTUAL, | |
1180 | (cpu->singlestep_enabled & SSTEP_NOTIMER) == 0); | |
1181 | ||
1182 | if (cpu_can_run(cpu)) { | |
1183 | int r; | |
1184 | r = tcg_cpu_exec(cpu); | |
1185 | if (r == EXCP_DEBUG) { | |
1186 | cpu_handle_guest_debug(cpu); | |
1187 | break; | |
1188 | } | |
1189 | } else if (cpu->stop || cpu->stopped) { | |
1190 | if (cpu->unplug) { | |
1191 | cpu = CPU_NEXT(cpu); | |
1192 | } | |
1193 | break; | |
1194 | } | |
1195 | ||
1196 | } /* for cpu.. */ | |
1197 | ||
1198 | /* Pairs with smp_wmb in qemu_cpu_kick. */ | |
1199 | atomic_mb_set(&exit_request, 0); | |
ac70aafc AB |
1200 | |
1201 | if (use_icount) { | |
40daca54 | 1202 | int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL); |
ac70aafc AB |
1203 | |
1204 | if (deadline == 0) { | |
40daca54 | 1205 | qemu_clock_notify(QEMU_CLOCK_VIRTUAL); |
ac70aafc | 1206 | } |
3b2319a3 | 1207 | } |
d5f8d613 | 1208 | qemu_tcg_wait_io_event(QTAILQ_FIRST(&cpus)); |
c93bbbef | 1209 | deal_with_unplugged_cpus(); |
296af7c9 BS |
1210 | } |
1211 | ||
1212 | return NULL; | |
1213 | } | |
1214 | ||
2ff09a40 | 1215 | static void qemu_cpu_kick_thread(CPUState *cpu) |
cc015e9a PB |
1216 | { |
1217 | #ifndef _WIN32 | |
1218 | int err; | |
1219 | ||
e0c38211 PB |
1220 | if (cpu->thread_kicked) { |
1221 | return; | |
9102deda | 1222 | } |
e0c38211 | 1223 | cpu->thread_kicked = true; |
814e612e | 1224 | err = pthread_kill(cpu->thread->thread, SIG_IPI); |
cc015e9a PB |
1225 | if (err) { |
1226 | fprintf(stderr, "qemu:%s: %s", __func__, strerror(err)); | |
1227 | exit(1); | |
1228 | } | |
1229 | #else /* _WIN32 */ | |
e0c38211 PB |
1230 | abort(); |
1231 | #endif | |
1232 | } | |
ed9164a3 | 1233 | |
e0c38211 PB |
1234 | static void qemu_cpu_kick_no_halt(void) |
1235 | { | |
1236 | CPUState *cpu; | |
1237 | /* Ensure whatever caused the exit has reached the CPU threads before | |
1238 | * writing exit_request. | |
1239 | */ | |
1240 | atomic_mb_set(&exit_request, 1); | |
1241 | cpu = atomic_mb_read(&tcg_current_cpu); | |
1242 | if (cpu) { | |
1243 | cpu_exit(cpu); | |
cc015e9a | 1244 | } |
cc015e9a PB |
1245 | } |
1246 | ||
c08d7424 | 1247 | void qemu_cpu_kick(CPUState *cpu) |
296af7c9 | 1248 | { |
f5c121b8 | 1249 | qemu_cond_broadcast(cpu->halt_cond); |
e0c38211 PB |
1250 | if (tcg_enabled()) { |
1251 | qemu_cpu_kick_no_halt(); | |
1252 | } else { | |
1253 | qemu_cpu_kick_thread(cpu); | |
1254 | } | |
296af7c9 BS |
1255 | } |
1256 | ||
46d62fac | 1257 | void qemu_cpu_kick_self(void) |
296af7c9 | 1258 | { |
4917cf44 | 1259 | assert(current_cpu); |
9102deda | 1260 | qemu_cpu_kick_thread(current_cpu); |
296af7c9 BS |
1261 | } |
1262 | ||
60e82579 | 1263 | bool qemu_cpu_is_self(CPUState *cpu) |
296af7c9 | 1264 | { |
814e612e | 1265 | return qemu_thread_is_self(cpu->thread); |
296af7c9 BS |
1266 | } |
1267 | ||
79e2b9ae | 1268 | bool qemu_in_vcpu_thread(void) |
aa723c23 | 1269 | { |
4917cf44 | 1270 | return current_cpu && qemu_cpu_is_self(current_cpu); |
aa723c23 JQ |
1271 | } |
1272 | ||
afbe7053 PB |
1273 | static __thread bool iothread_locked = false; |
1274 | ||
1275 | bool qemu_mutex_iothread_locked(void) | |
1276 | { | |
1277 | return iothread_locked; | |
1278 | } | |
1279 | ||
296af7c9 BS |
1280 | void qemu_mutex_lock_iothread(void) |
1281 | { | |
21618b3e | 1282 | atomic_inc(&iothread_requesting_mutex); |
2e7f7a3c PB |
1283 | /* In the simple case there is no need to bump the VCPU thread out of |
1284 | * TCG code execution. | |
1285 | */ | |
1286 | if (!tcg_enabled() || qemu_in_vcpu_thread() || | |
46036b24 | 1287 | !first_cpu || !first_cpu->created) { |
296af7c9 | 1288 | qemu_mutex_lock(&qemu_global_mutex); |
21618b3e | 1289 | atomic_dec(&iothread_requesting_mutex); |
1a28cac3 | 1290 | } else { |
1a28cac3 | 1291 | if (qemu_mutex_trylock(&qemu_global_mutex)) { |
e0c38211 | 1292 | qemu_cpu_kick_no_halt(); |
1a28cac3 MT |
1293 | qemu_mutex_lock(&qemu_global_mutex); |
1294 | } | |
6b49809c | 1295 | atomic_dec(&iothread_requesting_mutex); |
46daff13 | 1296 | qemu_cond_broadcast(&qemu_io_proceeded_cond); |
1a28cac3 | 1297 | } |
afbe7053 | 1298 | iothread_locked = true; |
296af7c9 BS |
1299 | } |
1300 | ||
1301 | void qemu_mutex_unlock_iothread(void) | |
1302 | { | |
afbe7053 | 1303 | iothread_locked = false; |
296af7c9 BS |
1304 | qemu_mutex_unlock(&qemu_global_mutex); |
1305 | } | |
1306 | ||
e8faee06 | 1307 | static bool all_vcpus_paused(void) |
296af7c9 | 1308 | { |
bdc44640 | 1309 | CPUState *cpu; |
296af7c9 | 1310 | |
bdc44640 | 1311 | CPU_FOREACH(cpu) { |
182735ef | 1312 | if (!cpu->stopped) { |
e8faee06 | 1313 | return false; |
0ab07c62 | 1314 | } |
296af7c9 BS |
1315 | } |
1316 | ||
e8faee06 | 1317 | return true; |
296af7c9 BS |
1318 | } |
1319 | ||
1320 | void pause_all_vcpus(void) | |
1321 | { | |
bdc44640 | 1322 | CPUState *cpu; |
296af7c9 | 1323 | |
40daca54 | 1324 | qemu_clock_enable(QEMU_CLOCK_VIRTUAL, false); |
bdc44640 | 1325 | CPU_FOREACH(cpu) { |
182735ef AF |
1326 | cpu->stop = true; |
1327 | qemu_cpu_kick(cpu); | |
296af7c9 BS |
1328 | } |
1329 | ||
aa723c23 | 1330 | if (qemu_in_vcpu_thread()) { |
d798e974 JK |
1331 | cpu_stop_current(); |
1332 | if (!kvm_enabled()) { | |
bdc44640 | 1333 | CPU_FOREACH(cpu) { |
182735ef AF |
1334 | cpu->stop = false; |
1335 | cpu->stopped = true; | |
d798e974 JK |
1336 | } |
1337 | return; | |
1338 | } | |
1339 | } | |
1340 | ||
296af7c9 | 1341 | while (!all_vcpus_paused()) { |
be7d6c57 | 1342 | qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex); |
bdc44640 | 1343 | CPU_FOREACH(cpu) { |
182735ef | 1344 | qemu_cpu_kick(cpu); |
296af7c9 BS |
1345 | } |
1346 | } | |
1347 | } | |
1348 | ||
2993683b IM |
1349 | void cpu_resume(CPUState *cpu) |
1350 | { | |
1351 | cpu->stop = false; | |
1352 | cpu->stopped = false; | |
1353 | qemu_cpu_kick(cpu); | |
1354 | } | |
1355 | ||
296af7c9 BS |
1356 | void resume_all_vcpus(void) |
1357 | { | |
bdc44640 | 1358 | CPUState *cpu; |
296af7c9 | 1359 | |
40daca54 | 1360 | qemu_clock_enable(QEMU_CLOCK_VIRTUAL, true); |
bdc44640 | 1361 | CPU_FOREACH(cpu) { |
182735ef | 1362 | cpu_resume(cpu); |
296af7c9 BS |
1363 | } |
1364 | } | |
1365 | ||
4c055ab5 GZ |
1366 | void cpu_remove(CPUState *cpu) |
1367 | { | |
1368 | cpu->stop = true; | |
1369 | cpu->unplug = true; | |
1370 | qemu_cpu_kick(cpu); | |
1371 | } | |
1372 | ||
2c579042 BR |
1373 | void cpu_remove_sync(CPUState *cpu) |
1374 | { | |
1375 | cpu_remove(cpu); | |
1376 | while (cpu->created) { | |
1377 | qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex); | |
1378 | } | |
1379 | } | |
1380 | ||
4900116e DDAG |
1381 | /* For temporary buffers for forming a name */ |
1382 | #define VCPU_THREAD_NAME_SIZE 16 | |
1383 | ||
e5ab30a2 | 1384 | static void qemu_tcg_init_vcpu(CPUState *cpu) |
296af7c9 | 1385 | { |
4900116e | 1386 | char thread_name[VCPU_THREAD_NAME_SIZE]; |
d5f8d613 FK |
1387 | static QemuCond *tcg_halt_cond; |
1388 | static QemuThread *tcg_cpu_thread; | |
4900116e | 1389 | |
296af7c9 BS |
1390 | /* share a single thread for all cpus with TCG */ |
1391 | if (!tcg_cpu_thread) { | |
814e612e | 1392 | cpu->thread = g_malloc0(sizeof(QemuThread)); |
f5c121b8 AF |
1393 | cpu->halt_cond = g_malloc0(sizeof(QemuCond)); |
1394 | qemu_cond_init(cpu->halt_cond); | |
1395 | tcg_halt_cond = cpu->halt_cond; | |
4900116e DDAG |
1396 | snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG", |
1397 | cpu->cpu_index); | |
1398 | qemu_thread_create(cpu->thread, thread_name, qemu_tcg_cpu_thread_fn, | |
1399 | cpu, QEMU_THREAD_JOINABLE); | |
1ecf47bf | 1400 | #ifdef _WIN32 |
814e612e | 1401 | cpu->hThread = qemu_thread_get_handle(cpu->thread); |
1ecf47bf | 1402 | #endif |
61a46217 | 1403 | while (!cpu->created) { |
18a85728 | 1404 | qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex); |
0ab07c62 | 1405 | } |
814e612e | 1406 | tcg_cpu_thread = cpu->thread; |
296af7c9 | 1407 | } else { |
814e612e | 1408 | cpu->thread = tcg_cpu_thread; |
f5c121b8 | 1409 | cpu->halt_cond = tcg_halt_cond; |
296af7c9 BS |
1410 | } |
1411 | } | |
1412 | ||
48a106bd | 1413 | static void qemu_kvm_start_vcpu(CPUState *cpu) |
296af7c9 | 1414 | { |
4900116e DDAG |
1415 | char thread_name[VCPU_THREAD_NAME_SIZE]; |
1416 | ||
814e612e | 1417 | cpu->thread = g_malloc0(sizeof(QemuThread)); |
f5c121b8 AF |
1418 | cpu->halt_cond = g_malloc0(sizeof(QemuCond)); |
1419 | qemu_cond_init(cpu->halt_cond); | |
4900116e DDAG |
1420 | snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/KVM", |
1421 | cpu->cpu_index); | |
1422 | qemu_thread_create(cpu->thread, thread_name, qemu_kvm_cpu_thread_fn, | |
1423 | cpu, QEMU_THREAD_JOINABLE); | |
61a46217 | 1424 | while (!cpu->created) { |
18a85728 | 1425 | qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex); |
0ab07c62 | 1426 | } |
296af7c9 BS |
1427 | } |
1428 | ||
10a9021d | 1429 | static void qemu_dummy_start_vcpu(CPUState *cpu) |
c7f0f3b1 | 1430 | { |
4900116e DDAG |
1431 | char thread_name[VCPU_THREAD_NAME_SIZE]; |
1432 | ||
814e612e | 1433 | cpu->thread = g_malloc0(sizeof(QemuThread)); |
f5c121b8 AF |
1434 | cpu->halt_cond = g_malloc0(sizeof(QemuCond)); |
1435 | qemu_cond_init(cpu->halt_cond); | |
4900116e DDAG |
1436 | snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/DUMMY", |
1437 | cpu->cpu_index); | |
1438 | qemu_thread_create(cpu->thread, thread_name, qemu_dummy_cpu_thread_fn, cpu, | |
c7f0f3b1 | 1439 | QEMU_THREAD_JOINABLE); |
61a46217 | 1440 | while (!cpu->created) { |
c7f0f3b1 AL |
1441 | qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex); |
1442 | } | |
1443 | } | |
1444 | ||
c643bed9 | 1445 | void qemu_init_vcpu(CPUState *cpu) |
296af7c9 | 1446 | { |
ce3960eb AF |
1447 | cpu->nr_cores = smp_cores; |
1448 | cpu->nr_threads = smp_threads; | |
f324e766 | 1449 | cpu->stopped = true; |
56943e8c PM |
1450 | |
1451 | if (!cpu->as) { | |
1452 | /* If the target cpu hasn't set up any address spaces itself, | |
1453 | * give it the default one. | |
1454 | */ | |
6731d864 PC |
1455 | AddressSpace *as = address_space_init_shareable(cpu->memory, |
1456 | "cpu-memory"); | |
12ebc9a7 | 1457 | cpu->num_ases = 1; |
6731d864 | 1458 | cpu_address_space_init(cpu, as, 0); |
56943e8c PM |
1459 | } |
1460 | ||
0ab07c62 | 1461 | if (kvm_enabled()) { |
48a106bd | 1462 | qemu_kvm_start_vcpu(cpu); |
c7f0f3b1 | 1463 | } else if (tcg_enabled()) { |
e5ab30a2 | 1464 | qemu_tcg_init_vcpu(cpu); |
c7f0f3b1 | 1465 | } else { |
10a9021d | 1466 | qemu_dummy_start_vcpu(cpu); |
0ab07c62 | 1467 | } |
296af7c9 BS |
1468 | } |
1469 | ||
b4a3d965 | 1470 | void cpu_stop_current(void) |
296af7c9 | 1471 | { |
4917cf44 AF |
1472 | if (current_cpu) { |
1473 | current_cpu->stop = false; | |
1474 | current_cpu->stopped = true; | |
1475 | cpu_exit(current_cpu); | |
96bce683 | 1476 | qemu_cond_broadcast(&qemu_pause_cond); |
b4a3d965 | 1477 | } |
296af7c9 BS |
1478 | } |
1479 | ||
56983463 | 1480 | int vm_stop(RunState state) |
296af7c9 | 1481 | { |
aa723c23 | 1482 | if (qemu_in_vcpu_thread()) { |
74892d24 | 1483 | qemu_system_vmstop_request_prepare(); |
1dfb4dd9 | 1484 | qemu_system_vmstop_request(state); |
296af7c9 BS |
1485 | /* |
1486 | * FIXME: should not return to device code in case | |
1487 | * vm_stop() has been requested. | |
1488 | */ | |
b4a3d965 | 1489 | cpu_stop_current(); |
56983463 | 1490 | return 0; |
296af7c9 | 1491 | } |
56983463 KW |
1492 | |
1493 | return do_vm_stop(state); | |
296af7c9 BS |
1494 | } |
1495 | ||
8a9236f1 LC |
1496 | /* does a state transition even if the VM is already stopped, |
1497 | current state is forgotten forever */ | |
56983463 | 1498 | int vm_stop_force_state(RunState state) |
8a9236f1 LC |
1499 | { |
1500 | if (runstate_is_running()) { | |
56983463 | 1501 | return vm_stop(state); |
8a9236f1 LC |
1502 | } else { |
1503 | runstate_set(state); | |
b2780d32 WC |
1504 | |
1505 | bdrv_drain_all(); | |
594a45ce KW |
1506 | /* Make sure to return an error if the flush in a previous vm_stop() |
1507 | * failed. */ | |
22af08ea | 1508 | return bdrv_flush_all(); |
8a9236f1 LC |
1509 | } |
1510 | } | |
1511 | ||
9a78eead | 1512 | void list_cpus(FILE *f, fprintf_function cpu_fprintf, const char *optarg) |
262353cb BS |
1513 | { |
1514 | /* XXX: implement xxx_cpu_list for targets that still miss it */ | |
e916cbf8 PM |
1515 | #if defined(cpu_list) |
1516 | cpu_list(f, cpu_fprintf); | |
262353cb BS |
1517 | #endif |
1518 | } | |
de0b36b6 LC |
1519 | |
1520 | CpuInfoList *qmp_query_cpus(Error **errp) | |
1521 | { | |
1522 | CpuInfoList *head = NULL, *cur_item = NULL; | |
182735ef | 1523 | CPUState *cpu; |
de0b36b6 | 1524 | |
bdc44640 | 1525 | CPU_FOREACH(cpu) { |
de0b36b6 | 1526 | CpuInfoList *info; |
182735ef AF |
1527 | #if defined(TARGET_I386) |
1528 | X86CPU *x86_cpu = X86_CPU(cpu); | |
1529 | CPUX86State *env = &x86_cpu->env; | |
1530 | #elif defined(TARGET_PPC) | |
1531 | PowerPCCPU *ppc_cpu = POWERPC_CPU(cpu); | |
1532 | CPUPPCState *env = &ppc_cpu->env; | |
1533 | #elif defined(TARGET_SPARC) | |
1534 | SPARCCPU *sparc_cpu = SPARC_CPU(cpu); | |
1535 | CPUSPARCState *env = &sparc_cpu->env; | |
1536 | #elif defined(TARGET_MIPS) | |
1537 | MIPSCPU *mips_cpu = MIPS_CPU(cpu); | |
1538 | CPUMIPSState *env = &mips_cpu->env; | |
48e06fe0 BK |
1539 | #elif defined(TARGET_TRICORE) |
1540 | TriCoreCPU *tricore_cpu = TRICORE_CPU(cpu); | |
1541 | CPUTriCoreState *env = &tricore_cpu->env; | |
182735ef | 1542 | #endif |
de0b36b6 | 1543 | |
cb446eca | 1544 | cpu_synchronize_state(cpu); |
de0b36b6 LC |
1545 | |
1546 | info = g_malloc0(sizeof(*info)); | |
1547 | info->value = g_malloc0(sizeof(*info->value)); | |
55e5c285 | 1548 | info->value->CPU = cpu->cpu_index; |
182735ef | 1549 | info->value->current = (cpu == first_cpu); |
259186a7 | 1550 | info->value->halted = cpu->halted; |
58f88d4b | 1551 | info->value->qom_path = object_get_canonical_path(OBJECT(cpu)); |
9f09e18a | 1552 | info->value->thread_id = cpu->thread_id; |
de0b36b6 | 1553 | #if defined(TARGET_I386) |
86f4b687 | 1554 | info->value->arch = CPU_INFO_ARCH_X86; |
544a3731 | 1555 | info->value->u.x86.pc = env->eip + env->segs[R_CS].base; |
de0b36b6 | 1556 | #elif defined(TARGET_PPC) |
86f4b687 | 1557 | info->value->arch = CPU_INFO_ARCH_PPC; |
544a3731 | 1558 | info->value->u.ppc.nip = env->nip; |
de0b36b6 | 1559 | #elif defined(TARGET_SPARC) |
86f4b687 | 1560 | info->value->arch = CPU_INFO_ARCH_SPARC; |
544a3731 EB |
1561 | info->value->u.q_sparc.pc = env->pc; |
1562 | info->value->u.q_sparc.npc = env->npc; | |
de0b36b6 | 1563 | #elif defined(TARGET_MIPS) |
86f4b687 | 1564 | info->value->arch = CPU_INFO_ARCH_MIPS; |
544a3731 | 1565 | info->value->u.q_mips.PC = env->active_tc.PC; |
48e06fe0 | 1566 | #elif defined(TARGET_TRICORE) |
86f4b687 | 1567 | info->value->arch = CPU_INFO_ARCH_TRICORE; |
544a3731 | 1568 | info->value->u.tricore.PC = env->PC; |
86f4b687 EB |
1569 | #else |
1570 | info->value->arch = CPU_INFO_ARCH_OTHER; | |
de0b36b6 LC |
1571 | #endif |
1572 | ||
1573 | /* XXX: waiting for the qapi to support GSList */ | |
1574 | if (!cur_item) { | |
1575 | head = cur_item = info; | |
1576 | } else { | |
1577 | cur_item->next = info; | |
1578 | cur_item = info; | |
1579 | } | |
1580 | } | |
1581 | ||
1582 | return head; | |
1583 | } | |
0cfd6a9a LC |
1584 | |
1585 | void qmp_memsave(int64_t addr, int64_t size, const char *filename, | |
1586 | bool has_cpu, int64_t cpu_index, Error **errp) | |
1587 | { | |
1588 | FILE *f; | |
1589 | uint32_t l; | |
55e5c285 | 1590 | CPUState *cpu; |
0cfd6a9a | 1591 | uint8_t buf[1024]; |
0dc9daf0 | 1592 | int64_t orig_addr = addr, orig_size = size; |
0cfd6a9a LC |
1593 | |
1594 | if (!has_cpu) { | |
1595 | cpu_index = 0; | |
1596 | } | |
1597 | ||
151d1322 AF |
1598 | cpu = qemu_get_cpu(cpu_index); |
1599 | if (cpu == NULL) { | |
c6bd8c70 MA |
1600 | error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cpu-index", |
1601 | "a CPU number"); | |
0cfd6a9a LC |
1602 | return; |
1603 | } | |
1604 | ||
1605 | f = fopen(filename, "wb"); | |
1606 | if (!f) { | |
618da851 | 1607 | error_setg_file_open(errp, errno, filename); |
0cfd6a9a LC |
1608 | return; |
1609 | } | |
1610 | ||
1611 | while (size != 0) { | |
1612 | l = sizeof(buf); | |
1613 | if (l > size) | |
1614 | l = size; | |
2f4d0f59 | 1615 | if (cpu_memory_rw_debug(cpu, addr, buf, l, 0) != 0) { |
0dc9daf0 BP |
1616 | error_setg(errp, "Invalid addr 0x%016" PRIx64 "/size %" PRId64 |
1617 | " specified", orig_addr, orig_size); | |
2f4d0f59 AK |
1618 | goto exit; |
1619 | } | |
0cfd6a9a | 1620 | if (fwrite(buf, 1, l, f) != l) { |
c6bd8c70 | 1621 | error_setg(errp, QERR_IO_ERROR); |
0cfd6a9a LC |
1622 | goto exit; |
1623 | } | |
1624 | addr += l; | |
1625 | size -= l; | |
1626 | } | |
1627 | ||
1628 | exit: | |
1629 | fclose(f); | |
1630 | } | |
6d3962bf LC |
1631 | |
1632 | void qmp_pmemsave(int64_t addr, int64_t size, const char *filename, | |
1633 | Error **errp) | |
1634 | { | |
1635 | FILE *f; | |
1636 | uint32_t l; | |
1637 | uint8_t buf[1024]; | |
1638 | ||
1639 | f = fopen(filename, "wb"); | |
1640 | if (!f) { | |
618da851 | 1641 | error_setg_file_open(errp, errno, filename); |
6d3962bf LC |
1642 | return; |
1643 | } | |
1644 | ||
1645 | while (size != 0) { | |
1646 | l = sizeof(buf); | |
1647 | if (l > size) | |
1648 | l = size; | |
eb6282f2 | 1649 | cpu_physical_memory_read(addr, buf, l); |
6d3962bf | 1650 | if (fwrite(buf, 1, l, f) != l) { |
c6bd8c70 | 1651 | error_setg(errp, QERR_IO_ERROR); |
6d3962bf LC |
1652 | goto exit; |
1653 | } | |
1654 | addr += l; | |
1655 | size -= l; | |
1656 | } | |
1657 | ||
1658 | exit: | |
1659 | fclose(f); | |
1660 | } | |
ab49ab5c LC |
1661 | |
1662 | void qmp_inject_nmi(Error **errp) | |
1663 | { | |
9cb805fd | 1664 | nmi_monitor_handle(monitor_get_cpu_index(), errp); |
ab49ab5c | 1665 | } |
27498bef ST |
1666 | |
1667 | void dump_drift_info(FILE *f, fprintf_function cpu_fprintf) | |
1668 | { | |
1669 | if (!use_icount) { | |
1670 | return; | |
1671 | } | |
1672 | ||
1673 | cpu_fprintf(f, "Host - Guest clock %"PRIi64" ms\n", | |
1674 | (cpu_get_clock() - cpu_get_icount())/SCALE_MS); | |
1675 | if (icount_align_option) { | |
1676 | cpu_fprintf(f, "Max guest delay %"PRIi64" ms\n", -max_delay/SCALE_MS); | |
1677 | cpu_fprintf(f, "Max guest advance %"PRIi64" ms\n", max_advance/SCALE_MS); | |
1678 | } else { | |
1679 | cpu_fprintf(f, "Max guest delay NA\n"); | |
1680 | cpu_fprintf(f, "Max guest advance NA\n"); | |
1681 | } | |
1682 | } |