If the worker finishes right after we `waitpid` but before we add it to
`WORKER_PIDS` the `worker_reaper` won't `waitpid` it cause it iterates
over `WORKER_PIDS`. So
 - the clean-up triggered by the SIGCHLD won't catch it cause it needs it to
   be in `WORKER_PIDS`
 - and, `register_worker` won't because it was still running when it
   `waitpid`'ed it

Moving the insertion into `WORKER_PIDS` before the `waitpid` solves
this by making sure it is
 - always in the var for `worker_reaper`
 - and, if SIGCHILD should trigger `worker_reaper` before we add it to
   `WORKER_PIDS`, the `waitpid` in `register_worker` itself will catch
   it

Signed-off-by: Hannes Laimer <[email protected]>
---
 src/PVE/RESTEnvironment.pm | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/src/PVE/RESTEnvironment.pm b/src/PVE/RESTEnvironment.pm
index 4ed5c05..4677687 100644
--- a/src/PVE/RESTEnvironment.pm
+++ b/src/PVE/RESTEnvironment.pm
@@ -99,17 +99,18 @@ my $register_worker = sub {
 
     return if !$pid;
 
-    # do not register if already finished
+    $WORKER_PIDS->{$pid} = {
+        user => $user,
+        upid => $upid,
+    };
+
+    # remove immediately if already finished
     my $waitpid = waitpid($pid, WNOHANG);
     if (defined($waitpid) && ($waitpid == $pid)) {
         delete($WORKER_PIDS->{$pid});
         return;
     }
 
-    $WORKER_PIDS->{$pid} = {
-        user => $user,
-        upid => $upid,
-    };
 };
 
 # initialize environment - must be called once at program startup
-- 
2.47.3




Reply via email to