dump: skip POSIX timers targeting dead threads by pojntfx · Pull Request #2996 · checkpoint-restore/criu

pojntfx · 2026-04-06T07:46:30Z

What:

When a POSIX timer uses SIGEV_THREAD_ID and the target thread has exited, dump fails in PID namespaces (Unable to convert the notify thread id) and restore fails without PID namespaces (EINVAL from timer_create()). We noticed this when an application switched from Java 8 to Java 21.

How:

To fix this, this PR detects dead notify threads during dump and skips the timer instead of failing. The kernel already silently drops signals from these timers, so skipping them during dump is safe.

Tests:

zdtm/static/dead_thread_posix_timer
~~- zdtm/static/dead_thread_posix_timer_h~~

I added dead_thread_posix_timer_h since I wanted to specifically trigger the restore failure case, which needs host mode. I followed the convention that pthread_timers.c and pthread_timers_h.c uses.

…t/dead threads from checkpoint Signed-off-by: Felicitas Pojtinger <felicitas@pojtinger.com>

Copilot

Pull request overview

This PR adjusts CRIU’s POSIX timer dump logic to tolerate SIGEV_THREAD_ID timers whose notify thread has already exited, avoiding dump/restore failures seen in PID-namespace vs host-mode scenarios, and adds ZDTM coverage for both modes.

Changes:

Skip POSIX timers during dump when their SIGEV_THREAD_ID notify thread is detected as dead (instead of failing the dump).
Update the POSIX timer dump loop to compact the encoded timer list after skips (tte->n_posix reflects only kept timers).
Add two ZDTM static tests (dead_thread_posix_timer and _h host-flavor variant) and wire them into the static test Makefile.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
`criu/timer.c`	Detects dead notify threads and skips affected timers; compacts dumped timer entries.
`test/zdtm/static/dead_thread_posix_timer.c`	New ZDTM test creating a `SIGEV_THREAD_ID` timer targeting a thread that exits.
`test/zdtm/static/dead_thread_posix_timer_h.c`	Host-flavor variant of the same test scenario.
`test/zdtm/static/dead_thread_posix_timer.desc`	Runs the test with `ns_pid` feature enabled.
`test/zdtm/static/dead_thread_posix_timer_h.desc`	Runs the test in host flavor (`{'flavor': 'h'}`).
`test/zdtm/static/Makefile`	Registers the new tests and links them with `-lrt -pthread`.

Copilot · 2026-04-16T15:23:24Z

 	for (i = 0; i < item->nr_threads; i++) {
 		if (item->threads[i].real != rtid)
 			continue;

 		vtid = item->threads[i].ns[0].virt;
 		break;
 	}

 	if (vtid == 0) {
-		pr_err("Unable to convert the notify thread id %d\n", rtid);
-		return -1;
+		pr_warn("Skipping timer because notify thread %d is dead\n", rtid);
+		return 1;
 	}


In the pid-namespace path, vtid==0 is treated as “notify thread is dead” and the timer is skipped. That also covers the case where the thread is found but its ns virt tid is unexpectedly 0 (e.g., a regression in thread nsid collection), which would silently drop a valid timer and hide the underlying bug. Consider tracking whether the thread was found separately: only skip when the notify thread isn’t present in item->threads, and keep vtid==0 as a hard error if the thread was found but conversion failed.

Copilot · 2026-04-16T15:23:25Z

 static int encode_notify_thread_id(pid_t rtid, struct pstree_item *item, PosixTimerEntry *pte)
 {
 	pid_t vtid = 0;
 	int i;

 	if (rtid == 0)
 		return 0;

 	if (!(root_ns_mask & CLONE_NEWPID)) {
 		/* Non-pid-namespace case */
+		for (i = 0; i < item->nr_threads; i++) {
+			if (item->threads[i].real == rtid)
+				break;
+		}
+
+		if (i == item->nr_threads) {
+			pr_warn("Skipping timer because notify thread %d is dead\n", rtid);
+			return 1;
+		}


encode_notify_thread_id now uses return value 1 to signal “skip this timer”, while -1 still means error. Since this is a new non-obvious contract, it’s easy for future callers to mishandle. Consider introducing a named constant/enum for the skip return code and documenting it in the function comment (or using a more conventional negative errno for “skip”).

avagin · 2026-04-16T15:27:51Z

 		pthread_timers_h		\
 		leader_posix_timer		\
 		dead_thread_posix_timer		\
+		dead_thread_posix_timer_nons		\


It isn't required. The previous test will be executed in non-ns, ns and uns envs.

I think this was a review on an older commit pushed as part of this PR? From the PR description:

I added dead_thread_posix_timer_h since I wanted to specifically trigger the restore failure case, which needs host mode. I followed the convention that pthread_timers.c and pthread_timers_h.c uses.

I'll remove it regardless, it was mostly used as a reproducer initially. Resolved via c245960

codecov-commenter · 2026-04-16T16:21:24Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 57.23%. Comparing base (f943d4f) to head (17b2913).
⚠️ Report is 52 commits behind head on criu-dev.

Additional details and impacted files

@@             Coverage Diff              @@
##           criu-dev    #2996      +/-   ##
============================================
+ Coverage     57.21%   57.23%   +0.01%     
============================================
  Files           154      154              
  Lines         40400    40413      +13     
  Branches       8857     8859       +2     
============================================
+ Hits          23115    23130      +15     
+ Misses        17021    17019       -2     
  Partials        264      264

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

rst0git · 2026-04-17T19:58:51Z

@pojntfx Would it be possible to add commit messages?

pojntfx · 2026-05-11T23:46:17Z

@rst0git @avagin Thank you for your reviews. I collapsed it into a single commit and rebased on the latest criu-dev. I already added commit messages to the original commits:

When collapsing them into a single new commit I also added a body based on the PR description. Let me know if I should make any further changes :)

3idey · 2026-05-12T14:36:50Z

When collapsing them into a single new commit I also added a body based on the PR description. Let me know if I should make any further changes :)

@pojntfx Thank you!

Could you please separate this into two commits? The PR currently squashes the core fix and the ZDTM test into a single commit. According to the Contribution Guidelines, each logical change should be separated into a distinct patch.

When a POSIX timer uses SIGEV_THREAD_ID and the target thread has exited, dump fails in PID namespaces ("Unable to convert the notify thread id") and restore fails without PID namespaces (EINVAL from timer_create()). We noticed this when an application switched from Java 8 to Java 21. Detect dead notify threads during dump and skip the timer instead of failing. The kernel already silently drops signals from these timers, so skipping them during dump is safe. Assisted-by: Claude:claude-opus-4-7 Signed-off-by: Felicitas Pojtinger <felicitas@pojtinger.com>

Add a test that creates a SIGEV_THREAD_ID POSIX timer pointing at a thread which then exits before checkpoint. Dump and restore must succeed by skipping the timer rather than failing. Assisted-by: Claude:claude-opus-4-7 Signed-off-by: Felicitas Pojtinger <felicitas@pojtinger.com>

pojntfx · 2026-05-13T22:31:36Z

@3idey Done!

pojntfx referenced this pull request in loopholelabs/criu Apr 7, 2026

test: Add ZDTM test for excluding POSIX timers targetting non-existen…

9e14e01

…t/dead threads from checkpoint Signed-off-by: Felicitas Pojtinger <felicitas@pojtinger.com>

avagin requested a review from Copilot April 16, 2026 15:18

Copilot started reviewing on behalf of avagin April 16, 2026 15:19 View session

Copilot AI reviewed Apr 16, 2026

View reviewed changes

avagin reviewed Apr 16, 2026

View reviewed changes

rst0git reviewed Apr 17, 2026

View reviewed changes

Comment thread criu/timer.c

pojntfx force-pushed the pojntfx/arch-951-upstream-criu-changes branch 2 times, most recently from 3526179 to f4e3b56 Compare May 11, 2026 23:42

pojntfx added 2 commits May 13, 2026 15:30

pojntfx force-pushed the pojntfx/arch-951-upstream-criu-changes branch from f4e3b56 to 5b30f6d Compare May 13, 2026 22:31

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

dump: skip POSIX timers targeting dead threads#2996

dump: skip POSIX timers targeting dead threads#2996
pojntfx wants to merge 2 commits into
checkpoint-restore:criu-devfrom
loopholelabs:pojntfx/arch-951-upstream-criu-changes

pojntfx commented Apr 6, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

avagin Apr 16, 2026

Uh oh!

pojntfx May 11, 2026

Uh oh!

codecov-commenter commented Apr 16, 2026 •

edited

Loading

Uh oh!

rst0git commented Apr 17, 2026

Uh oh!

Uh oh!

pojntfx commented May 11, 2026

Uh oh!

3idey commented May 12, 2026

Uh oh!

pojntfx commented May 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Conversation

pojntfx commented Apr 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

avagin Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

pojntfx May 11, 2026

Choose a reason for hiding this comment

Uh oh!

codecov-commenter commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

rst0git commented Apr 17, 2026

Uh oh!

Uh oh!

pojntfx commented May 11, 2026

Uh oh!

3idey commented May 12, 2026

Uh oh!

pojntfx commented May 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

pojntfx commented Apr 6, 2026 •

edited

Loading

codecov-commenter commented Apr 16, 2026 •

edited

Loading