We currently report EINVAL whenever a struct pid has no tasked attached
anymore thereby conflating two concepts:
(1) The task has already been reaped.
(2) The caller requested a pidfd for a thread-group leader but the pid
actually references a struct pid that isn't used as a thread-group
leader.
This is causing issues for non-threaded workloads as in [1].
This patch tries to allow userspace to distinguish between (1) and (2).
This is racy of course but that shouldn't matter.
Link: https://github.com/systemd/systemd/pull/36982 [1]
Signed-off-by: Christian Brauner <brauner@xxxxxxxxxx>
---
kernel/fork.c | 31 ++++++++++++++++++++++++++++---
1 file changed, 28 insertions(+), 3 deletions(-)
diff --git a/kernel/fork.c b/kernel/fork.c
index 182ec2e9087d..0fe54fcd11b3 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -2108,10 +2108,35 @@ static int __pidfd_prepare(struct pid *pid, unsigned int flags, struct file **re
*/
int pidfd_prepare(struct pid *pid, unsigned int flags, struct file **ret)
{
- bool thread = flags & PIDFD_THREAD;
+ int err = 0;
- if (!pid_has_task(pid, thread ? PIDTYPE_PID : PIDTYPE_TGID))
- return -EINVAL;
+ if (!(flags & PIDFD_THREAD)) {
+ /*
+ * If this is struct pid isn't used as a thread-group
+ * leader pid but the caller requested to create a
+ * thread-group leader pidfd then report ENOENT to the
+ * caller as a hint.
+ */
+ if (!pid_has_task(pid, PIDTYPE_TGID))
+ err = -ENOENT;
+ }
+
+ /*
+ * If this wasn't a thread-group leader struct pid or the task
+ * got reaped in the meantime report -ESRCH to userspace.
+ *
+ * This is racy of course. This could've not been a thread-group
+ * leader struct pid and we set ENOENT above but in the meantime
+ * the task got reaped. Or there was a multi-threaded-exec by a
+ * subthread and we were a thread-group leader but now got
+ * killed. All of that doesn't matter since the task has already
+ * been reaped that distinction is meaningless to userspace so
+ * just report ESRCH.
+ */
+ if (!pid_has_task(pid, PIDTYPE_PID))
+ err = -ESRCH;
+ if (err)
+ return err;
return __pidfd_prepare(pid, flags, ret);
}
--
2.47.2
