Bug 2212314

Summary:	systemd deadlocks waiting for its child forever when receiving SIGQUIT signal
Product:	Red Hat Enterprise Linux 8	Reporter:	Renaud Métrich <rmetrich>
Component:	systemd	Assignee:	Michal Sekletar <msekleta>
Status:	CLOSED MIGRATED	QA Contact:	Frantisek Sumsal <fsumsal>
Severity:	high	Docs Contact:
Priority:	high
Version:	8.8	CC:	dtardon, systemd-maint-list
Target Milestone:	rc	Keywords:	MigratedToJIRA, Reproducer
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2023-09-21 15:15:04 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Renaud Métrich 2023-06-05 09:35:33 UTC

Description of problem:

We have a customer that could observe systemd not reaping any children, hence zombies accumulating.
The vmcore showed that PID 1 was waiting forever for a child it spawned, still named "systemd", the child being itself hanging in writing to a pipe.

We believe some SIGQUIT signal was sent to PID 1, which leads to exact same observation:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
# kill -QUIT 1

# cat /proc/1/stack
[<0>] do_wait+0x165/0x2f0
[<0>] kernel_waitid+0x118/0x180
[<0>] __do_sys_waitid+0x120/0x130
[<0>] do_syscall_64+0x5b/0x1b0
[<0>] entry_SYSCALL_64_after_hwframe+0x61/0xc6

# cat /proc/2037/stack 
[<0>] pipe_wait+0x6c/0xc0
[<0>] pipe_write+0x16c/0x3f0
[<0>] new_sync_write+0x112/0x160
[<0>] __kernel_write+0x4f/0x100
[<0>] dump_emit+0x91/0xd0
[<0>] elf_core_dump+0x890/0xa50
[<0>] do_coredump+0x73f/0xf52
[<0>] get_signal+0x14f/0x870
[<0>] do_signal+0x36/0x690
[<0>] exit_to_usermode_loop+0x89/0x100
[<0>] do_syscall_64+0x19c/0x1b0
[<0>] entry_SYSCALL_64_after_hwframe+0x61/0xc6
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

The root cause for this is systemd, upon receiving SIGQUIT, will dump core.
To do that it spawns a child, then waits for it to terminate.
The child sends itself the SIGQUIT signal.

This triggers the core_pattern handler to execute and kernel to create a pipe for coredumping and send data to the pipe immediately.
The core_pattern handler executes, sends to /run/systemd/coredump Unix socket where the coredump service will read the coredump from (the pipe create by the kernel).

Normally systemd should spawns systemd-coredump instantiable service, which in turn reads from the pipe, but it cannot since it's waiting for its child to terminate.

--> Deadlock

Version-Release number of selected component (if applicable):

systemd-239

How reproducible:

Always

Steps to Reproduce:
1. Send SIGQUIT to PID 1

Actual results:

Deadlock of PID 1 and its child

Expected results:

PID 1 functional

Additional infos:

Because PID 1 is critical for RHEL systems, it has to be super robust.

Comment 1 Renaud Métrich 2023-06-05 09:47:07 UTC

It's very possible other scenarios but SIGQUIT can cause same deadlock.
Looking at the systemd code, I see this in particular:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
 801 void sigkill_wait(pid_t pid) {
 802         assert(pid > 1);
 803 
 804         if (kill(pid, SIGKILL) > 0)
 805                 (void) wait_for_terminate(pid, NULL);
 806 }

 819 void sigterm_wait(pid_t pid) {
 820         assert(pid > 1);
 821 
 822         if (kill_and_sigcont(pid, SIGTERM) > 0)
 823                 (void) wait_for_terminate(pid, NULL);
 824 }
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

sigterm_wait() is called by some DBus thing:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
1067 static void bus_kill_exec(sd_bus *bus) {
1068         if (pid_is_valid(bus->busexec_pid) > 0) {
1069                 sigterm_wait(bus->busexec_pid);
1070                 bus->busexec_pid = 0;
1071         }
1072 }
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

I don't know how to reproduce this code path, but I guess that if PID 1 sends a TERM to that other process, which in turn starts dumping core (because it didn't like receiving the signal), then same deadlock will happen.

Comment 2 David Tardon 2023-06-06 12:07:57 UTC

(In reply to Renaud Métrich from comment #1)
> I don't know how to reproduce this code path, but I guess that if PID 1
> sends a TERM to that other process, which in turn starts dumping core
> (because it didn't like receiving the signal), then same deadlock will
> happen.

This code path cannot be triggered in PID1 (or anywhere else in systemd codebase). It requires direct use of sd-bus API, like this:

int main(void) {
        sd_bus *bus;

        sd_bus_new(&bus);
        sd_bus_set_exec(bus, path, argv); // Sets bus->exec_path
        sd_bus_start(bus); // Starts the command in bus->exec_path and sets bus->busexec_pid accordingly
        sd_bus_close(bus); // Calls bus_kill_exec()
}

Comment 3 RHEL Program Management 2023-09-21 13:54:36 UTC

Issue migration from Bugzilla to Jira is in process at this time. This will be the last message in Jira copied from the Bugzilla bug.

Comment 4 RHEL Program Management 2023-09-21 15:15:04 UTC

This BZ has been automatically migrated to the issues.redhat.com Red Hat Issue Tracker. All future work related to this report will be managed there.

Due to differences in account names between systems, some fields were not replicated.  Be sure to add yourself to Jira issue's "Watchers" field to continue receiving updates and add others to the "Need Info From" field to continue requesting information.

To find the migrated issue, look in the "Links" section for a direct link to the new issue location. The issue key will have an icon of 2 footprints next to it, and begin with "RHEL-" followed by an integer.  You can also find this issue by visiting https://issues.redhat.com/issues/?jql= and searching the "Bugzilla Bug" field for this BZ's number, e.g. a search like:

"Bugzilla Bug" = 1234567

In the event you have trouble locating or viewing this issue, you can file an issue by sending mail to rh-issues. You can also visit https://access.redhat.com/articles/7032570 for general account information.