]> git.itanic.dy.fi Git - linux-stable/commit
io_uring: wait interruptibly for request completions on exit
authorJens Axboe <axboe@kernel.dk>
Mon, 12 Jun 2023 03:14:09 +0000 (21:14 -0600)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 19 Jul 2023 14:22:09 +0000 (16:22 +0200)
commitb50d6e06cca7b67a3d73ca660dda27662b76e6ea
tree6fab7e244ae2902a4b29083381fe78a22c776a83
parent9440b24fbcb969a71c65703bd9cc0b3ecd5c756c
io_uring: wait interruptibly for request completions on exit

commit 4826c59453b3b4677d6bf72814e7ababdea86949 upstream.

WHen the ring exits, cleanup is done and the final cancelation and
waiting on completions is done by io_ring_exit_work. That function is
invoked by kworker, which doesn't take any signals. Because of that, it
doesn't really matter if we wait for completions in TASK_INTERRUPTIBLE
or TASK_UNINTERRUPTIBLE state. However, it does matter to the hung task
detection checker!

Normally we expect cancelations and completions to happen rather
quickly. Some test cases, however, will exit the ring and park the
owning task stopped (eg via SIGSTOP). If the owning task needs to run
task_work to complete requests, then io_ring_exit_work won't make any
progress until the task is runnable again. Hence io_ring_exit_work can
trigger the hung task detection, which is particularly problematic if
panic-on-hung-task is enabled.

As the ring exit doesn't take signals to begin with, have it wait
interruptibly rather than uninterruptibly. io_uring has a separate
stuck-exit warning that triggers independently anyway, so we're not
really missing anything by making this switch.

Cc: stable@vger.kernel.org # 5.10+
Link: https://lore.kernel.org/r/b0e4aaef-7088-56ce-244c-976edeac0e66@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
io_uring/io_uring.c