pidfd: block SIGCHLD during tmp process creation #2491
Conversation
This patch blocks SIGCHLD during temporary process creation to prevent a race condition between kill() and waitpid() where sigchld_handler() causes `criu restore` to fail with an error. Fixes: checkpoint-restore#2490 Signed-off-by: Bhavik Sachdev <b.sachdev1904@gmail.com> Signed-off-by: Radostin Stoyanov <rstoyanov@fedoraproject.org>
|
Hi Radostin! Thanks a lot for looking into it! We had a call with Bhavik today and he wants to see if we can use pstree helpers to solve this problem. So, let's wait a bit with merging this. |
No worries, I just wanted to confirm that the pidfd_dead test will pass with these changes in CI.
I'm not sure how easy it would be to do that but the messages in the following commits might help: c09ba04 5f72963 |
| */ | ||
| sigemptyset(&blockmask); | ||
| sigaddset(&blockmask, SIGCHLD); | ||
| if (sigprocmask(SIG_BLOCK, &blockmask, &dead->oldmask) == -1) { |
There was a problem hiding this comment.
sigprocmask should be called right before killing this process. If I remember it right, a process can be created in one process, but killed in another process.
This pull request updates pidfd to block
SIGCHLDduring temporary process creation to prevent a race condition betweenkill()andwaitpid()wheresigchld_handler()causescriu restoreto fail with an error.Fixes: #2490