podman image load --input file.tar sometimes hangs infinitely #25184
Labels
kind/bug
Categorizes issue or PR as related to a bug.
Comments
AlekseiNikiforovIBM
added
the
kind/bug
|
@giuseppe PTAL |
|
thanks for the report. Could we replace the diff --git a/pkg/rootless/rootless_linux.c b/pkg/rootless/rootless_linux.c
index 4f71d49e5c..3d74af6a6c 100644
--- a/pkg/rootless/rootless_linux.c
+++ b/pkg/rootless/rootless_linux.c
@@ -658,7 +658,7 @@ create_pause_process (const char *pause_pid_file_path, char **argv)
if (pipe (p) < 0)
return -1;
- pid = fork ();
+ pid = syscall_clone (SIGCHLD, NULL);
if (pid < 0)
{
close (p[0]);
@@ -689,7 +689,7 @@ create_pause_process (const char *pause_pid_file_path, char **argv)
close (p[0]);
setsid ();
- pid = fork ();
+ pid = syscall_clone (SIGCHLD, NULL);
if (pid < 0)
_exit (EXIT_FAILURE);I am not able to reproduce locally on x64_64, are you able to test this patch? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Issue Description
Sometimes "podman image load --input file.tar" hangs indefinitely. It happens inside a privileged podman container on s390x rarely.
Steps to reproduce the issue
Steps to reproduce the issue
Describe the results you received
Sometimes command never starts loading image and never finishes.
Describe the results you expected
Podman should successfully load image in a reasonable amount of time.
podman info output
Podman in a container
Yes
Privileged Or Rootless
Privileged
Upstream Latest Release
No
Additional environment details
Host is RHEL 9.5 s390x in a cloud, inside KVM. It runs podman as root, guest is privileged podman container with Ubuntu 24.04. Issue reproduces in this Ubuntu system.
Additional information
Backtrace of child process hanging:
It looks like podman first does
syscall(clone)[1] and then doesforkglibc function call [2].According to what @iii-i told me, calling
forkglibc function should be avoided after doingsyscall(clone(~CLONE_VM))or similar actions, because glibc internal state may be inconsistent, and that can lead toforkfunction hanging.Either
forkcalls should be replaced bysyscall(clone)reimplementations to avoid going inside glibcforkif they may be called after doingsyscall(clone), orsyscall(clone)could be reimplemented as compination offork+unshare, or even justunshareifforkis not needed.[1]
podman/pkg/rootless/rootless_linux.c
Line 973 in 48f8742
[2]
podman/pkg/rootless/rootless_linux.c
Line 661 in 48f8742
The text was updated successfully, but these errors were encountered: