[fix](executor) Fix rare self-deadlock that can cause the time-sharing task executor to hang. #58273
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
Author
|
run nonConcurrent |
kaka11chen
force-pushed
the
time_sharing_task_executor_timeout_test
branch
from
90c5711 to
4fab545
Compare
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 34076 ms |
ClickBench: Total hot run time: 27.91 s |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
Author
|
run nonConcurrent |
7 similar comments
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run nonConcurrent |
kaka11chen
force-pushed
the
time_sharing_task_executor_timeout_test
branch
from
4fab545 to
d3e7247
Compare
Contributor
Author
|
run buildall |
kaka11chen
force-pushed
the
time_sharing_task_executor_timeout_test
branch
from
d3e7247 to
f3cc40b
Compare
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 35954 ms |
TPC-DS: Total hot run time: 186512 ms |
ClickBench: Total hot run time: 27.2 s |
Contributor
Author
|
run nonConcurrent |
kaka11chen
force-pushed
the
time_sharing_task_executor_timeout_test
branch
from
f3cc40b to
8651d73
Compare
Contributor
Author
|
run buildall |
kaka11chen
force-pushed
the
time_sharing_task_executor_timeout_test
branch
from
8651d73 to
79a2525
Compare
Contributor
Author
|
run buildall |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
Author
|
run nonConcurrent |
kaka11chen
force-pushed
the
time_sharing_task_executor_timeout_test
branch
from
70ed1bd to
e1e8ec2
Compare
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 32094 ms |
TPC-DS: Total hot run time: 172925 ms |
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
kaka11chen
changed the title
morningman
approved these changes
Jan 13, 2026
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
Contributor
|
PR approved by anyone and no changes requested. |
suxiaogang223
approved these changes
Jan 13, 2026
kaka11chen
changed the title
github-actions bot
pushed a commit
that referenced
this pull request
Jan 13, 2026
…g task executor to hang. (#58273) ### What problem does this PR solve? ### Release note #### Summary 1. Fix a self-deadlock that can occur in `_dispatch_thread` when a destructor path attempts to re-acquire the same mutex already held by the thread. The root cause is destructors (triggered while holding `_lock`) performing operations that try to re-acquire the same mutex. A safe fix should ensure destructors that may call `remove_task()` run outside the `_lock` scope or avoid re-locking the same mutex inside destructor paths. 2. Call `_task_executor->wait()` in `TaskExecutorSimplifiedScanScheduler::stop()`. --- #### Details / Reproduction steps 1. `std::shared_ptr<PrioritizedSplitRunner> split = _tokenless->_entries->take();` 3. `l.lock();` — `_dispatch_thread` acquires `_lock`. 4. After the `while` loop finishes, `split` goes out of scope and the `shared_ptr` is destroyed. 5. `PrioritizedSplitRunner` destructor runs → destroys `_split_runner` (`ScannerSplitRunner`). 6. `ScannerSplitRunner::_scan_func` destructor runs → destroys captured `ctx` (`std::shared_ptr<ScannerContext>`). 7. `ScannerContext::~ScannerContext()` calls `remove_task()`. 8. `remove_task()` attempts to acquire `_lock`. 9. Result: **self-deadlock** because `_lock` is already held by `_dispatch_thread`. --- #### Solution We must explicitly release 'split' BEFORE acquiring _lock to avoid self-deadlock. The destructor chain (`PrioritizedSplitRunner` -> `ScannerSplitRunner`-> `_scan_func` lambda -> captured `ScannerContext`) may call `remove_task()` which tries to acquire `_lock`. Since `_lock` is not a recursive mutex, this would deadlock.
zzzxl1993
pushed a commit
to zzzxl1993/doris
that referenced
this pull request
Jan 13, 2026
…g task executor to hang. (apache#58273) ### What problem does this PR solve? ### Release note #### Summary 1. Fix a self-deadlock that can occur in `_dispatch_thread` when a destructor path attempts to re-acquire the same mutex already held by the thread. The root cause is destructors (triggered while holding `_lock`) performing operations that try to re-acquire the same mutex. A safe fix should ensure destructors that may call `remove_task()` run outside the `_lock` scope or avoid re-locking the same mutex inside destructor paths. 2. Call `_task_executor->wait()` in `TaskExecutorSimplifiedScanScheduler::stop()`. --- #### Details / Reproduction steps 1. `std::shared_ptr<PrioritizedSplitRunner> split = _tokenless->_entries->take();` 3. `l.lock();` — `_dispatch_thread` acquires `_lock`. 4. After the `while` loop finishes, `split` goes out of scope and the `shared_ptr` is destroyed. 5. `PrioritizedSplitRunner` destructor runs → destroys `_split_runner` (`ScannerSplitRunner`). 6. `ScannerSplitRunner::_scan_func` destructor runs → destroys captured `ctx` (`std::shared_ptr<ScannerContext>`). 7. `ScannerContext::~ScannerContext()` calls `remove_task()`. 8. `remove_task()` attempts to acquire `_lock`. 9. Result: **self-deadlock** because `_lock` is already held by `_dispatch_thread`. --- #### Solution We must explicitly release 'split' BEFORE acquiring _lock to avoid self-deadlock. The destructor chain (`PrioritizedSplitRunner` -> `ScannerSplitRunner`-> `_scan_func` lambda -> captured `ScannerContext`) may call `remove_task()` which tries to acquire `_lock`. Since `_lock` is not a recursive mutex, this would deadlock.
kaka11chen
added a commit
to kaka11chen/doris
that referenced
this pull request
Jan 21, 2026
…g task executor to hang. (apache#58273) 1. Fix a self-deadlock that can occur in `_dispatch_thread` when a destructor path attempts to re-acquire the same mutex already held by the thread. The root cause is destructors (triggered while holding `_lock`) performing operations that try to re-acquire the same mutex. A safe fix should ensure destructors that may call `remove_task()` run outside the `_lock` scope or avoid re-locking the same mutex inside destructor paths. 2. Call `_task_executor->wait()` in `TaskExecutorSimplifiedScanScheduler::stop()`. --- 1. `std::shared_ptr<PrioritizedSplitRunner> split = _tokenless->_entries->take();` 3. `l.lock();` — `_dispatch_thread` acquires `_lock`. 4. After the `while` loop finishes, `split` goes out of scope and the `shared_ptr` is destroyed. 5. `PrioritizedSplitRunner` destructor runs → destroys `_split_runner` (`ScannerSplitRunner`). 6. `ScannerSplitRunner::_scan_func` destructor runs → destroys captured `ctx` (`std::shared_ptr<ScannerContext>`). 7. `ScannerContext::~ScannerContext()` calls `remove_task()`. 8. `remove_task()` attempts to acquire `_lock`. 9. Result: **self-deadlock** because `_lock` is already held by `_dispatch_thread`. --- We must explicitly release 'split' BEFORE acquiring _lock to avoid self-deadlock. The destructor chain (`PrioritizedSplitRunner` -> `ScannerSplitRunner`-> `_scan_func` lambda -> captured `ScannerContext`) may call `remove_task()` which tries to acquire `_lock`. Since `_lock` is not a recursive mutex, this would deadlock.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What problem does this PR solve?
Release note
Summary
_dispatch_threadwhen a destructor path attempts to re-acquire the same mutex already held by the thread. The root cause is destructors (triggered while holding_lock) performing operations that try to re-acquire the same mutex. A safe fix should ensure destructors that may callremove_task()run outside the_lockscope or avoid re-locking the same mutex inside destructor paths._task_executor->wait()inTaskExecutorSimplifiedScanScheduler::stop().Details / Reproduction steps
std::shared_ptr<PrioritizedSplitRunner> split = _tokenless->_entries->take();l.lock();—_dispatch_threadacquires_lock.whileloop finishes,splitgoes out of scope and theshared_ptris destroyed.PrioritizedSplitRunnerdestructor runs → destroys_split_runner(ScannerSplitRunner).ScannerSplitRunner::_scan_funcdestructor runs → destroys capturedctx(std::shared_ptr<ScannerContext>).ScannerContext::~ScannerContext()callsremove_task().remove_task()attempts to acquire_lock._lockis already held by_dispatch_thread.Solution
We must explicitly release 'split' BEFORE acquiring _lock to avoid self-deadlock. The destructor chain (
PrioritizedSplitRunner->ScannerSplitRunner->_scan_funclambda -> capturedScannerContext) may callremove_task()which tries to acquire_lock. Since_lockis not a recursive mutex, this would deadlock.Check List (For Author)
Test
Behavior changed:
Does this need documentation?
Check List (For Reviewer who merge this PR)