Line data    Source code

       1             : #include "source/common/watchdog/abort_action.h"
       2             : 
       3             : #include "envoy/thread/thread.h"
       4             : 
       5             : #include "source/common/common/assert.h"
       6             : #include "source/common/common/fmt.h"
       7             : #include "source/common/common/logger.h"
       8             : #include "source/common/protobuf/utility.h"
       9             : #include "source/common/thread/terminate_thread.h"
      10             : 
      11             : namespace Envoy {
      12             : namespace Watchdog {
      13             : namespace {
      14             : constexpr uint64_t DefaultWaitDurationMs = 5000;
      15             : } // end namespace
      16             : 
      17             : AbortAction::AbortAction(envoy::watchdog::v3::AbortActionConfig& config,
      18             :                          Server::Configuration::GuardDogActionFactoryContext& /*context*/)
      19             :     : wait_duration_(absl::Milliseconds(
      20           0 :           PROTOBUF_GET_MS_OR_DEFAULT(config, wait_duration, DefaultWaitDurationMs))) {}
      21             : 
      22             : void AbortAction::run(
      23             :     envoy::config::bootstrap::v3::Watchdog::WatchdogAction::WatchdogEvent /*event*/,
      24             :     const std::vector<std::pair<Thread::ThreadId, MonotonicTime>>& thread_last_checkin_pairs,
      25           0 :     MonotonicTime /*now*/) {
      26             : 
      27           0 :   if (thread_last_checkin_pairs.empty()) {
      28           0 :     ENVOY_LOG_MISC(warn, "Watchdog AbortAction called without any thread.");
      29           0 :     return;
      30           0 :   }
      31             : 
      32             :   // The following lines of code won't be considered covered by code coverage
      33             :   // tools since they would run in DEATH tests.
      34           0 :   const auto& thread_id = thread_last_checkin_pairs[0].first;
      35           0 :   const std::string tid_string = thread_id.debugString();
      36           0 :   ENVOY_LOG_MISC(error, "Watchdog AbortAction terminating thread with tid {}.", tid_string);
      37             : 
      38           0 :   if (Thread::terminateThread(thread_id)) {
      39             :     // Successfully signaled to thread to terminate, sleep for wait_duration.
      40           0 :     absl::SleepFor(wait_duration_);
      41           0 :   } else {
      42           0 :     ENVOY_LOG_MISC(error, "Failed to terminate tid {}", tid_string);
      43           0 :   }
      44             : 
      45             :   // Abort from the action since the signaled thread hasn't yet crashed the process.
      46             :   // Panicing in the action gives flexibility since it doesn't depend on
      47             :   // external code to kill the process if the signal fails.
      48           0 :   PANIC(fmt::format(
      49           0 :       "Failed to terminate thread with id {}, aborting from Watchdog AbortAction instead.",
      50           0 :       tid_string));
      51           0 : }
      52             : 
      53             : } // namespace Watchdog
      54             : } // namespace Envoy