1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844
|
// Copyright 2012 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifdef UNSAFE_BUFFERS_BUILD
// TODO(crbug.com/390223051): Remove C-library calls to fix the errors.
#pragma allow_unsafe_libc_calls
#endif
#include "gpu/ipc/service/gpu_watchdog_thread.h"
#include <memory>
#include <string>
#include <utility>
#include "base/atomicops.h"
#include "base/bit_cast.h"
#include "base/command_line.h"
#include "base/debug/alias.h"
#include "base/debug/dump_without_crashing.h"
#include "base/files/file_path.h"
#include "base/files/file_util.h"
#include "base/functional/bind.h"
#include "base/functional/callback_helpers.h"
#include "base/logging.h"
#include "base/memory/ptr_util.h"
#include "base/metrics/histogram_functions.h"
#include "base/metrics/histogram_macros.h"
#include "base/native_library.h"
#include "base/numerics/safe_conversions.h"
#include "base/power_monitor/power_monitor.h"
#include "base/process/process.h"
#include "base/strings/string_number_conversions.h"
#include "base/system/sys_info.h"
#include "base/task/current_thread.h"
#include "base/threading/platform_thread.h"
#include "base/time/time.h"
#include "build/build_config.h"
#include "gpu/config/gpu_crash_keys.h"
#include "gpu/config/gpu_switches.h"
#include "gpu/ipc/common/result_codes.h"
#if BUILDFLAG(IS_WIN)
#include <windows.h>
#endif
namespace gpu {
base::TimeDelta GetGpuWatchdogTimeout(bool software_rendering) {
std::string timeout_str =
base::CommandLine::ForCurrentProcess()->GetSwitchValueASCII(
switches::kGpuWatchdogTimeoutSeconds);
if (!timeout_str.empty()) {
size_t timeout_seconds;
if (base::StringToSizeT(timeout_str, &timeout_seconds))
return base::Seconds(timeout_seconds);
LOG(WARNING) << "Invalid --" << switches::kGpuWatchdogTimeoutSeconds << ": "
<< timeout_str;
}
base::TimeDelta timeout = kGpuWatchdogTimeout;
#if BUILDFLAG(IS_WIN)
int num_of_processors = base::SysInfo::NumberOfProcessors();
if (num_of_processors > 8) {
timeout -= base::Seconds(10);
} else if (num_of_processors <= 4) {
timeout += base::Seconds(5);
}
#endif
if (software_rendering) {
timeout *= kSoftwareRenderingFactor;
}
return timeout;
}
GpuWatchdogThread::GpuWatchdogThread(base::TimeDelta timeout,
int restart_factor,
bool is_test_mode,
const std::string& thread_name)
: base::Thread(thread_name),
watchdog_timeout_(timeout),
watchdog_restart_factor_(restart_factor),
is_test_mode_(is_test_mode) {
base::CurrentThread::Get()->AddTaskObserver(this);
// DO NOT CHANGE |watched_thread_name_str_uma_|. It's used for UMA and crash
// report.
if (thread_name == "GpuWatchdog_Compositor")
watched_thread_name_str_uma_ = ".compositor";
else
watched_thread_name_str_uma_ = ".main";
watched_thread_id_str_ =
base::NumberToString(base::PlatformThread::CurrentId().raw());
#if BUILDFLAG(IS_WIN)
// GetCurrentThread returns a pseudo-handle that cannot be used by one thread
// to identify another. DuplicateHandle creates a "real" handle that can be
// used for this purpose.
if (!::DuplicateHandle(::GetCurrentProcess(), ::GetCurrentThread(),
::GetCurrentProcess(), &watched_thread_handle_,
THREAD_QUERY_INFORMATION, FALSE, 0)) {
watched_thread_handle_ = nullptr;
}
#endif
#if BUILDFLAG(IS_LINUX) && !BUILDFLAG(IS_CASTOS)
tty_file_.reset(base::OpenFile(
base::FilePath(FILE_PATH_LITERAL("/sys/class/tty/tty0/active")), "r"));
UpdateActiveTTY();
host_tty_ = active_tty_;
#endif
Arm();
}
GpuWatchdogThread::~GpuWatchdogThread() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
// Stop() might take too long and the watchdog timeout is triggered.
// Disarm first before calling Stop() to avoid a crash.
if (IsArmed())
Disarm();
PauseWatchdog();
Stop(); // stop the watchdog thread
base::CurrentThread::Get()->RemoveTaskObserver(this);
base::PowerMonitor::GetInstance()->RemovePowerSuspendObserver(this);
GpuWatchdogThreadEventHistogram(GpuWatchdogThreadEvent::kGpuWatchdogEnd);
#if BUILDFLAG(IS_WIN)
if (watched_thread_handle_)
CloseHandle(watched_thread_handle_);
#endif
}
// static
std::unique_ptr<GpuWatchdogThread> GpuWatchdogThread::Create(
bool start_backgrounded,
base::TimeDelta timeout,
int restart_factor,
bool is_test_mode,
const std::string& thread_name) {
auto watchdog_thread = base::WrapUnique(new GpuWatchdogThread(
timeout, restart_factor, is_test_mode, thread_name));
watchdog_thread->Start();
if (start_backgrounded)
watchdog_thread->OnBackgrounded();
return watchdog_thread;
}
// static
std::unique_ptr<GpuWatchdogThread> GpuWatchdogThread::Create(
bool start_backgrounded,
bool software_rendering,
const std::string& thread_name) {
return Create(start_backgrounded, GetGpuWatchdogTimeout(software_rendering),
kRestartFactor, /*test_mode=*/false, thread_name);
}
// static
std::unique_ptr<GpuWatchdogThread> GpuWatchdogThread::Create(
bool start_backgrounded,
const GpuWatchdogThread* existing_watchdog,
const std::string& thread_name) {
DCHECK(existing_watchdog);
return Create(start_backgrounded, existing_watchdog->watchdog_timeout_,
existing_watchdog->watchdog_restart_factor_,
/*test_mode=*/false, thread_name);
}
// Android Chrome goes to the background. Called from the gpu io thread.
void GpuWatchdogThread::OnBackgrounded() {
// Report progress first in case the Watchdog timeout task in the watchdog
// thread is not invalidated soon enough.
InProgress();
task_runner()->PostTask(
FROM_HERE,
base::BindOnce(&GpuWatchdogThread::StopWatchdogTimeoutTask,
base::Unretained(this), kAndroidBackgroundForeground));
}
// Android Chrome goes to the foreground. Called from the gpu io thread.
void GpuWatchdogThread::OnForegrounded() {
task_runner()->PostTask(
FROM_HERE,
base::BindOnce(&GpuWatchdogThread::RestartWatchdogTimeoutTask,
base::Unretained(this), kAndroidBackgroundForeground));
}
// Called from the gpu thread when gpu init has completed.
void GpuWatchdogThread::OnInitComplete() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
task_runner()->PostTask(
FROM_HERE, base::BindOnce(&GpuWatchdogThread::UpdateInitializationFlag,
base::Unretained(this)));
Disarm();
// The PowerMonitorObserver needs to be register on the watchdog thread so the
// notifications are delivered on that thread.
task_runner()->PostTask(FROM_HERE,
base::BindOnce(&GpuWatchdogThread::AddPowerObserver,
base::Unretained(this)));
}
// Called from the gpu thread in viz::GpuServiceImpl::~GpuServiceImpl().
// After this, no Disarm() will be called before the watchdog thread is
// destroyed. If this destruction takes too long, the watchdog timeout
// will be triggered.
void GpuWatchdogThread::OnGpuProcessTearDown() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
in_gpu_process_teardown_ = true;
if (!IsArmed())
Arm();
}
// Called from the watched gpu thread.
void GpuWatchdogThread::PauseWatchdog() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
// Report progress first in case the Watchdog timeout task in the watchdog
// thread is not invalidated soon enough.
InProgress();
task_runner()->PostTask(
FROM_HERE, base::BindOnce(&GpuWatchdogThread::StopWatchdogTimeoutTask,
base::Unretained(this), kGeneralGpuFlow));
}
// Called from the watched gpu thread.
void GpuWatchdogThread::ResumeWatchdog() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
task_runner()->PostTask(
FROM_HERE, base::BindOnce(&GpuWatchdogThread::RestartWatchdogTimeoutTask,
base::Unretained(this), kGeneralGpuFlow));
}
// Running on the watchdog thread.
// On Linux, Init() will be called twice for Sandbox Initialization. The
// watchdog is stopped and then restarted in StartSandboxLinux(). Everything
// should be the same and continue after the second init().
void GpuWatchdogThread::Init() {
// Get and Invalidate weak_ptr should be done on the watchdog thread only.
weak_ptr_ = weak_factory_.GetWeakPtr();
task_runner()->PostDelayedTask(
FROM_HERE,
base::BindOnce(&GpuWatchdogThread::OnWatchdogTimeout, weak_ptr_),
watchdog_timeout_);
last_arm_disarm_counter_ = ReadArmDisarmCounter();
watchdog_start_timeticks_ = base::TimeTicks::Now();
last_on_watchdog_timeout_timeticks_ = watchdog_start_timeticks_;
next_on_watchdog_timeout_time_ = base::Time::Now() + watchdog_timeout_;
in_gpu_initialization_ = true;
#if BUILDFLAG(IS_WIN)
if (watched_thread_handle_) {
if (base::ThreadTicks::IsSupported())
base::ThreadTicks::WaitUntilInitialized();
last_on_watchdog_timeout_thread_ticks_ = GetWatchedThreadTime();
remaining_watched_thread_ticks_ = watchdog_timeout_;
}
#endif
}
// Running on the watchdog thread.
void GpuWatchdogThread::CleanUp() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
weak_factory_.InvalidateWeakPtrs();
}
void GpuWatchdogThread::ReportProgress() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
InProgress();
}
void GpuWatchdogThread::WillProcessTask(const base::PendingTask& pending_task,
bool was_blocked_or_low_priority) {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
// The watchdog is armed at the beginning of the gpu process teardown.
// Do not call Arm() during teardown.
if (in_gpu_process_teardown_)
DCHECK(IsArmed());
else
Arm();
}
void GpuWatchdogThread::DidProcessTask(const base::PendingTask& pending_task) {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
// Keep the watchdog armed during tear down.
if (in_gpu_process_teardown_)
InProgress();
else
Disarm();
}
// Power Suspends. Running on the watchdog thread.
void GpuWatchdogThread::OnSuspend() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
InProgress();
StopWatchdogTimeoutTask(kPowerSuspendResume);
}
// Power Resumes. Running on the watchdog thread.
void GpuWatchdogThread::OnResume() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
RestartWatchdogTimeoutTask(kPowerSuspendResume);
}
// Running on the watchdog thread.
// Call AddPowerSuspendObserver on the watchdog thread so that OnSuspend() and
// OnResume() will be called on this thread.
void GpuWatchdogThread::AddPowerObserver() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
// Adding the Observer to the power monitor is safe even if power monitor is
// not yet initialized.
bool is_system_suspended =
base::PowerMonitor::GetInstance()
->AddPowerSuspendObserverAndReturnSuspendedState(this);
if (is_system_suspended)
StopWatchdogTimeoutTask(kPowerSuspendResume);
}
// Running on the watchdog thread.
void GpuWatchdogThread::RestartWatchdogTimeoutTask(
PauseResumeSource source_of_request) {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
base::TimeDelta timeout;
switch (source_of_request) {
case kAndroidBackgroundForeground:
if (!is_backgrounded_)
return;
is_backgrounded_ = false;
timeout = watchdog_timeout_ * watchdog_restart_factor_;
foregrounded_timeticks_ = base::TimeTicks::Now();
foregrounded_event_ = true;
num_of_timeout_after_foregrounded_ = 0;
break;
case kPowerSuspendResume:
if (!in_power_suspension_)
return;
in_power_suspension_ = false;
timeout = watchdog_timeout_ * watchdog_restart_factor_;
power_resume_timeticks_ = base::TimeTicks::Now();
power_resumed_event_ = true;
num_of_timeout_after_power_resume_ = 0;
break;
case kGeneralGpuFlow:
if (!is_paused_)
return;
is_paused_ = false;
timeout = watchdog_timeout_;
watchdog_resume_timeticks_ = base::TimeTicks::Now();
break;
}
if (!is_backgrounded_ && !in_power_suspension_ && !is_paused_) {
weak_ptr_ = weak_factory_.GetWeakPtr();
task_runner()->PostDelayedTask(
FROM_HERE,
base::BindOnce(&GpuWatchdogThread::OnWatchdogTimeout, weak_ptr_),
timeout);
last_on_watchdog_timeout_timeticks_ = base::TimeTicks::Now();
next_on_watchdog_timeout_time_ = base::Time::Now() + timeout;
last_arm_disarm_counter_ = ReadArmDisarmCounter();
#if BUILDFLAG(IS_WIN)
if (watched_thread_handle_) {
last_on_watchdog_timeout_thread_ticks_ = GetWatchedThreadTime();
remaining_watched_thread_ticks_ = timeout;
}
#endif
}
}
void GpuWatchdogThread::StopWatchdogTimeoutTask(
PauseResumeSource source_of_request) {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
switch (source_of_request) {
case kAndroidBackgroundForeground:
if (is_backgrounded_)
return;
is_backgrounded_ = true;
backgrounded_timeticks_ = base::TimeTicks::Now();
foregrounded_event_ = false;
break;
case kPowerSuspendResume:
if (in_power_suspension_)
return;
in_power_suspension_ = true;
power_suspend_timeticks_ = base::TimeTicks::Now();
power_resumed_event_ = false;
break;
case kGeneralGpuFlow:
if (is_paused_)
return;
is_paused_ = true;
watchdog_pause_timeticks_ = base::TimeTicks::Now();
break;
}
// Revoke any pending watchdog timeout task
weak_factory_.InvalidateWeakPtrs();
}
// On the watchdog thread only.
void GpuWatchdogThread::UpdateInitializationFlag() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
in_gpu_initialization_ = false;
}
// Called from the watched gpu thread.
// The watchdog is armed only in these three functions -
// GpuWatchdogThread(), WillProcessTask(), and OnGpuProcessTearDown()
void GpuWatchdogThread::Arm() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
base::subtle::NoBarrier_AtomicIncrement(&arm_disarm_counter_, 1);
// Arm/Disarm are always called in sequence. Now it's an odd number.
DCHECK(IsArmed());
}
void GpuWatchdogThread::Disarm() {
DCHECK_CALLED_ON_VALID_SEQUENCE(watched_thread_sequence_checker_);
base::subtle::NoBarrier_AtomicIncrement(&arm_disarm_counter_, 1);
// Arm/Disarm are always called in sequence. Now it's an even number.
DCHECK(!IsArmed());
}
void GpuWatchdogThread::InProgress() {
// Increment by 2. This is equivalent to Disarm() + Arm().
// If Watchdog is already disarmed, it stays in the same disarmed status.
base::subtle::NoBarrier_AtomicIncrement(&arm_disarm_counter_, 2);
}
bool GpuWatchdogThread::IsArmed() {
// It's an odd number.
return base::subtle::NoBarrier_Load(&arm_disarm_counter_) & 1;
}
base::subtle::Atomic32 GpuWatchdogThread::ReadArmDisarmCounter() {
return base::subtle::NoBarrier_Load(&arm_disarm_counter_);
}
// Running on the watchdog thread.
void GpuWatchdogThread::OnWatchdogTimeout() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
DCHECK(!is_backgrounded_);
DCHECK(!in_power_suspension_);
DCHECK(!is_paused_);
// If this metric is added too early (eg. watchdog creation time), it cannot
// be persistent. The histogram data will be lost after crash or browser exit.
// Delay the recording of kGpuWatchdogStart until the firs
// OnWatchdogTimeout() to ensure this metric is created in the persistent
// memory.
if (!is_watchdog_start_histogram_recorded_) {
is_watchdog_start_histogram_recorded_ = true;
GpuWatchdogThreadEventHistogram(GpuWatchdogThreadEvent::kGpuWatchdogStart);
}
GpuWatchdogTimeoutHistogram(GpuWatchdogTimeoutEvent::kTimeout);
if (power_resumed_event_)
num_of_timeout_after_power_resume_++;
if (foregrounded_event_)
num_of_timeout_after_foregrounded_++;
#if BUILDFLAG(IS_LINUX) && !BUILDFLAG(IS_CASTOS)
UpdateActiveTTY();
#endif
// Collect all needed info for gpu hang detection.
auto arm_disarm_counter = ReadArmDisarmCounter();
bool disarmed = arm_disarm_counter % 2 == 0; // even number
bool gpu_makes_progress = arm_disarm_counter != last_arm_disarm_counter_;
bool no_gpu_hang = disarmed || gpu_makes_progress || SlowWatchdogThread();
bool watched_thread_needs_more_time =
WatchedThreadNeedsMoreThreadTime(no_gpu_hang);
no_gpu_hang = no_gpu_hang || watched_thread_needs_more_time ||
ContinueOnNonHostX11ServerTty();
// No gpu hang. Continue with another OnWatchdogTimeout task.
if (no_gpu_hang) {
ContinueWithNextWatchdogTimeoutTask();
return;
}
// If the watched thread makes a progress after crash dump, the GPU process
// will not be killed and every thing continues after this function.
// Otherwise, this is the end of the GPU process.
DeliberatelyTerminateToRecoverFromHang();
}
void GpuWatchdogThread::ContinueWithNextWatchdogTimeoutTask() {
last_on_watchdog_timeout_timeticks_ = base::TimeTicks::Now();
next_on_watchdog_timeout_time_ = base::Time::Now() + watchdog_timeout_;
last_arm_disarm_counter_ = ReadArmDisarmCounter();
task_runner()->PostDelayedTask(
FROM_HERE,
base::BindOnce(&GpuWatchdogThread::OnWatchdogTimeout, weak_ptr_),
watchdog_timeout_);
}
bool GpuWatchdogThread::SlowWatchdogThread() {
// If it takes 15 more seconds than the expected time between two
// OnWatchdogTimeout() calls, the system is considered slow and it's not a GPU
// hang.
bool slow_watchdog_thread =
(base::Time::Now() - next_on_watchdog_timeout_time_) >=
kUnreasonableTimeoutDelay;
// Record this case only when a GPU hang is detected and the thread is slow.
if (slow_watchdog_thread)
GpuWatchdogTimeoutHistogram(GpuWatchdogTimeoutEvent::kSlowWatchdogThread);
return slow_watchdog_thread;
}
bool GpuWatchdogThread::WatchedThreadNeedsMoreThreadTime(
bool no_gpu_hang_detected) {
#if BUILDFLAG(IS_WIN)
if (!watched_thread_handle_)
return false;
WatchedThreadNeedsMoreThreadTimeHistogram(
no_gpu_hang_detected,
/*start_of_more_thread_time*/ false);
if (!no_gpu_hang_detected && count_of_more_gpu_thread_time_allowed_ >=
kMaxCountOfMoreGpuThreadTimeAllowed) {
less_than_full_thread_time_after_capped_ = true;
} else {
less_than_full_thread_time_after_capped_ = false;
}
// Calculate how many thread ticks the watched thread spent doing the work.
base::ThreadTicks now = GetWatchedThreadTime();
base::TimeDelta thread_time_elapsed =
now - last_on_watchdog_timeout_thread_ticks_;
last_on_watchdog_timeout_thread_ticks_ = now;
remaining_watched_thread_ticks_ -= thread_time_elapsed;
if (no_gpu_hang_detected ||
count_of_more_gpu_thread_time_allowed_ >=
kMaxCountOfMoreGpuThreadTimeAllowed ||
thread_time_elapsed.is_negative() /* bogus data */ ||
remaining_watched_thread_ticks_ <= base::TimeDelta()) {
// Reset the remaining thread ticks.
remaining_watched_thread_ticks_ = watchdog_timeout_;
count_of_more_gpu_thread_time_allowed_ = 0;
return false;
} else {
// This is the start of allowing more thread time.
if (count_of_more_gpu_thread_time_allowed_ == 0) {
WatchedThreadNeedsMoreThreadTimeHistogram(
no_gpu_hang_detected, /*start_of_more_thread_time*/ true);
}
count_of_more_gpu_thread_time_allowed_++;
return true;
}
#else
return false;
#endif
}
#if BUILDFLAG(IS_WIN)
base::ThreadTicks GpuWatchdogThread::GetWatchedThreadTime() {
DCHECK(watched_thread_handle_);
if (base::ThreadTicks::IsSupported()) {
// Note: GetForThread() might return bogus results if running on different
// CPUs between two calls.
return base::ThreadTicks::GetForThread(
base::PlatformThreadHandle(watched_thread_handle_));
} else {
FILETIME creation_time;
FILETIME exit_time;
FILETIME kernel_time;
FILETIME user_time;
BOOL result = GetThreadTimes(watched_thread_handle_, &creation_time,
&exit_time, &kernel_time, &user_time);
if (!result)
return base::ThreadTicks();
// Need to bit_cast to fix alignment, then divide by 10 to convert
// 100-nanoseconds to microseconds.
int64_t user_time_us = base::bit_cast<int64_t, FILETIME>(user_time) / 10;
int64_t kernel_time_us =
base::bit_cast<int64_t, FILETIME>(kernel_time) / 10;
return base::ThreadTicks() +
base::Microseconds(user_time_us + kernel_time_us);
}
}
#endif
void GpuWatchdogThread::DeliberatelyTerminateToRecoverFromHang() {
DCHECK(task_runner()->RunsTasksInCurrentSequence());
// If this is for gpu testing, do not terminate the gpu process.
// Just signal and quit.
if (is_test_mode_) {
test_result_timeout_and_gpu_hang_.Set();
return;
}
#if BUILDFLAG(IS_WIN)
if (IsDebuggerPresent())
return;
#endif
// Store variables so they're available in crash dumps to help determine the
// cause of any hang.
base::TimeTicks function_begin_timeticks = base::TimeTicks::Now();
base::debug::Alias(&in_gpu_initialization_);
base::debug::Alias(&num_of_timeout_after_power_resume_);
base::debug::Alias(&num_of_timeout_after_foregrounded_);
base::debug::Alias(&function_begin_timeticks);
base::debug::Alias(&watchdog_start_timeticks_);
base::debug::Alias(&power_suspend_timeticks_);
base::debug::Alias(&power_resume_timeticks_);
base::debug::Alias(&backgrounded_timeticks_);
base::debug::Alias(&foregrounded_timeticks_);
base::debug::Alias(&watchdog_pause_timeticks_);
base::debug::Alias(&watchdog_resume_timeticks_);
base::debug::Alias(&in_power_suspension_);
base::debug::Alias(&in_gpu_process_teardown_);
base::debug::Alias(&is_backgrounded_);
base::debug::Alias(&last_on_watchdog_timeout_timeticks_);
base::TimeDelta timeticks_elapses =
function_begin_timeticks - last_on_watchdog_timeout_timeticks_;
base::debug::Alias(&timeticks_elapses);
#if BUILDFLAG(IS_WIN)
base::debug::Alias(&remaining_watched_thread_ticks_);
base::debug::Alias(&less_than_full_thread_time_after_capped_);
#endif
// The watchdog currently doesn't watch multiple threads. If multiple threads
// are supported, use '|' to separate thread ids in "list_of_hung_threads".
crash_keys::list_of_hung_threads.Set(watched_thread_id_str_);
crash_keys::gpu_watchdog_crashed_in_gpu_init.Set(
in_gpu_initialization_ ? "1" : "0");
crash_keys::gpu_watchdog_kill_after_power_resume.Set(
WithinOneMinFromPowerResumed() ? "1" : "0");
const int num_of_processors = base::SysInfo::NumberOfProcessors();
crash_keys::num_of_processors.Set(base::NumberToString(num_of_processors));
crash_keys::gpu_thread.Set(watched_thread_name_str_uma_);
// Check the arm_disarm_counter value one more time.
auto last_arm_disarm_counter = ReadArmDisarmCounter();
base::debug::Alias(&last_arm_disarm_counter);
// Create a crash dump first
base::debug::DumpWithoutCrashing();
// A kKill event is triggered and DumpWithoutCrashing() is called in the
// watchdog timeout routine OnWatchdogTimeout(). If it turns out
// gpu does not hang after the crash dump, another histogram
// kNoKillForGpuProgressDuringCrashDumping will be recorded later.
GpuWatchdogTimeoutHistogram(GpuWatchdogTimeoutEvent::kKill);
// Final check after the crash dump. If the watched thread makes a progress
// (disarmed) during generating crash dump, no need to crash the GPU process.
bool gpu_hang = IsArmed();
if (gpu_hang) {
// Still armed without any progress. The GPU process is now killed.
GpuWatchdogThreadEventHistogram(GpuWatchdogThreadEvent::kGpuWatchdogKill);
#if BUILDFLAG(IS_WIN)
if (less_than_full_thread_time_after_capped_)
GpuWatchdogTimeoutHistogram(
GpuWatchdogTimeoutEvent::kKillOnLessThreadTime);
#endif
// Use RESULT_CODE_HUNG so this crash is separated from other
// EXCEPTION_ACCESS_VIOLATION buckets for UMA analysis.
// TerminateCurrentProcessImmediately itself will not generate a dump.
base::Process::TerminateCurrentProcessImmediately(RESULT_CODE_HUNG);
// The end of the GPU process.
} else {
crash_keys::list_of_hung_threads.Clear();
crash_keys::gpu_watchdog_crashed_in_gpu_init.Clear();
crash_keys::gpu_watchdog_kill_after_power_resume.Clear();
crash_keys::num_of_processors.Clear();
crash_keys::gpu_thread.Clear();
GpuWatchdogTimeoutHistogram(
GpuWatchdogTimeoutEvent::kNoKillForGpuProgressDuringCrashDumping);
#if BUILDFLAG(IS_WIN)
// Reset the counters for WatchedThreadNeedsMoreThreadTime().
remaining_watched_thread_ticks_ = watchdog_timeout_;
count_of_more_gpu_thread_time_allowed_ = 0;
#endif
ContinueWithNextWatchdogTimeoutTask();
}
}
void GpuWatchdogThread::GpuWatchdogThreadEventHistogram(
GpuWatchdogThreadEvent thread_event) {
base::UmaHistogramEnumeration("GPU.WatchdogThread.Event", thread_event);
base::UmaHistogramEnumeration(
"GPU.WatchdogThread.Event" + watched_thread_name_str_uma_, thread_event);
}
void GpuWatchdogThread::GpuWatchdogTimeoutHistogram(
GpuWatchdogTimeoutEvent timeout_event) {
base::UmaHistogramEnumeration("GPU.WatchdogThread.Timeout", timeout_event);
base::UmaHistogramEnumeration(
"GPU.WatchdogThread.Timeout" + watched_thread_name_str_uma_,
timeout_event);
bool recorded = false;
if (in_gpu_initialization_) {
base::UmaHistogramEnumeration("GPU.WatchdogThread.Timeout.Init",
timeout_event);
base::UmaHistogramEnumeration(
"GPU.WatchdogThread.Timeout.Init" + watched_thread_name_str_uma_,
timeout_event);
recorded = true;
}
if (WithinOneMinFromPowerResumed()) {
base::UmaHistogramEnumeration("GPU.WatchdogThread.Timeout.PowerResume",
timeout_event);
base::UmaHistogramEnumeration(
"GPU.WatchdogThread.Timeout.PowerResume" + watched_thread_name_str_uma_,
timeout_event);
recorded = true;
}
if (WithinOneMinFromForegrounded()) {
base::UmaHistogramEnumeration("GPU.WatchdogThread.Timeout.Foregrounded",
timeout_event);
base::UmaHistogramEnumeration("GPU.WatchdogThread.Timeout.Foregrounded" +
watched_thread_name_str_uma_,
timeout_event);
recorded = true;
}
if (!recorded) {
base::UmaHistogramEnumeration("GPU.WatchdogThread.Timeout.Normal",
timeout_event);
base::UmaHistogramEnumeration(
"GPU.WatchdogThread.Timeout.Normal" + watched_thread_name_str_uma_,
timeout_event);
}
}
#if BUILDFLAG(IS_WIN)
void GpuWatchdogThread::WatchedThreadNeedsMoreThreadTimeHistogram(
bool no_gpu_hang_detected,
bool start_of_more_thread_time) {
if (start_of_more_thread_time) {
// This is the start of allowing more thread time. Only record it once for
// all following timeouts on the same detected gpu hang, so we know this
// is equivalent one crash in our crash reports.
GpuWatchdogTimeoutHistogram(GpuWatchdogTimeoutEvent::kMoreThreadTime);
} else {
if (count_of_more_gpu_thread_time_allowed_ > 0) {
if (no_gpu_hang_detected) {
// If count_of_more_gpu_thread_time_allowed_ > 0, we know extra time was
// extended in the previous OnWatchdogTimeout(). Now we find gpu makes
// progress. Record this case.
GpuWatchdogTimeoutHistogram(
GpuWatchdogTimeoutEvent::kProgressAfterMoreThreadTime);
} else if (count_of_more_gpu_thread_time_allowed_ >=
kMaxCountOfMoreGpuThreadTimeAllowed) {
GpuWatchdogTimeoutHistogram(
GpuWatchdogTimeoutEvent::kLessThanFullThreadTimeAfterCapped);
}
}
}
}
#endif
bool GpuWatchdogThread::WithinOneMinFromPowerResumed() {
size_t count = base::ClampFloor<size_t>(base::Minutes(1) / watchdog_timeout_);
return power_resumed_event_ && num_of_timeout_after_power_resume_ <= count;
}
bool GpuWatchdogThread::WithinOneMinFromForegrounded() {
size_t count = base::ClampFloor<size_t>(base::Minutes(1) / watchdog_timeout_);
return foregrounded_event_ && num_of_timeout_after_foregrounded_ <= count;
}
#if BUILDFLAG(IS_LINUX) && !BUILDFLAG(IS_CASTOS)
void GpuWatchdogThread::UpdateActiveTTY() {
last_active_tty_ = active_tty_;
active_tty_ = -1;
char tty_string[8] = {};
if (tty_file_ && !fseek(tty_file_.get(), 0, SEEK_SET) &&
fread(tty_string, 1, 7, tty_file_.get())) {
int tty_number;
if (sscanf(tty_string, "tty%d\n", &tty_number) == 1) {
active_tty_ = tty_number;
}
}
}
#endif
bool GpuWatchdogThread::ContinueOnNonHostX11ServerTty() {
#if BUILDFLAG(IS_LINUX) && !BUILDFLAG(IS_CASTOS)
if (host_tty_ == -1 || active_tty_ == -1)
return false;
// Don't crash if we're not on the TTY of our host X11 server.
if (active_tty_ != host_tty_) {
// Only record for the time there is a change on TTY
if (last_active_tty_ == active_tty_) {
GpuWatchdogTimeoutHistogram(
GpuWatchdogTimeoutEvent::kContinueOnNonHostServerTty);
}
return true;
}
#endif
return false;
}
// For gpu testing only. Return whether a GPU hang was detected or not.
bool GpuWatchdogThread::IsGpuHangDetectedForTesting() {
DCHECK(is_test_mode_);
return test_result_timeout_and_gpu_hang_.IsSet();
}
} // namespace gpu
|