SymbX/systemd
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

cgroup: whenever we migrate a PID to a unit, explicitly drop unit from empty notification queue

Browse Source 
A unit might be pending in the empty queue still when we add a PID to
the cgroup. At that point, let's explicitly remove the unit from that
queue.

Fixes: #36781
This commit is contained in:
Lennart Poettering
2025-03-19 21:40:33 +01:00
parent 0015502168
commit bb160976b0
1 changed files with 9 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/core/cgroup.c
+9 -1
View File
@@ -54,6 +54,8 @@
* out specific attributes from us. */
#define LOG_LEVEL_CGROUP_WRITE(r) (IN_SET(abs(r), ENOENT, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING)
static void unit_remove_from_cgroup_empty_queue(Unit *u);
uint64_t cgroup_tasks_max_resolve(const CGroupTasksMax *tasks_max) {
if (tasks_max->scale == 0)
return tasks_max->value;
@@ -3072,6 +3074,10 @@ int unit_attach_pids_to_cgroup(Unit *u, Set *pids, const char *suffix_path) {
else {
if (ret >= 0)
ret++; /* Count successful additions */
/* the cgroup is definitely not empty now, in case the unit was in
* the cgroup empty queue, drop it from there */
unit_remove_from_cgroup_empty_queue(u);
continue; /* When the bus thing worked via the bus we are fully done for this PID. */
}
}
@@ -3080,8 +3086,10 @@ int unit_attach_pids_to_cgroup(Unit *u, Set *pids, const char *suffix_path) {
ret = r; /* Remember first error */
continue;
} else if (ret >= 0)
} else if (ret >= 0) {
unit_remove_from_cgroup_empty_queue(u);
ret++; /* Count successful additions */
}
r = cg_all_unified();
if (r < 0)