The reaping of generator processes run as part of a daemon-reload should not call waitid(PID_ALL). The waitid() call in execute_directory() is intended only to reap the executed processes, but if a service process exits at about the same time as a daemon-reload, then that service process is reaped as well, preventing it from being reaped in the proper place in manager_dispatch_sigchld().
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=43625 --- src/util.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) diff --git a/src/util.c b/src/util.c index 58a0aeb..11f77ab 100644 --- a/src/util.c +++ b/src/util.c @@ -4623,11 +4623,12 @@ void execute_directory(const char *directory, DIR *d, char *argv[]) { } while (!hashmap_isempty(pids)) { + pid_t pid = PTR_TO_UINT(hashmap_first_key(pids)); siginfo_t si; char *path; zero(si); - if (waitid(P_ALL, 0, &si, WEXITED) < 0) { + if (waitid(P_PID, pid, &si, WEXITED) < 0) { if (errno == EINTR) continue; -- 1.7.7.6 _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel