Archive-only list for patches
 help / color / mirror / Atom feed
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: stable@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	patches@lists.linux.dev, Jens Axboe <axboe@kernel.dk>,
	Oleg Nesterov <oleg@redhat.com>,
	"Christian Brauner (Amutable)" <brauner@kernel.org>
Subject: [PATCH 7.2 16/88] io-wq: order the exit bit against worker creation task work
Date: Wed, 28 Oct 2026 17:59:24 +0100	[thread overview]
Message-ID: <20261028165942.497675794@linuxfoundation.org> (raw)
In-Reply-To: <20261028165942.107695352@linuxfoundation.org>

7.2-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Christian Brauner <brauner@kernel.org>

Commit 0e5663856b69219ef2f02e04bac9f5c5be19718a upstream.

io_wq_exit_workers() cancels task work for queued worker creation with
task_work_cancel_match(). That has a plain load of task->task_works.

io_queue_worker_create() adds the new entry and tests IO_WQ_BIT_EXIT to
cancel it if the workqueue is already on its way out.

That must be ordered. task_work_add() has a full barrier via cmpxchg().
But io_wq_exit_start() sets the bit with set_bit() which doesn't have
any memory ordering.

So afaict, on weakly ordered architectures the exiting task may load
task_works before the store of the bit is visible. The other side tests
the bit before the store is visible as well. So the entry remains queued
with worker_refs and the exiting task waits on worker_done indefinitely.

If the task still has rings then io_uring_del_tctx_node() provides the
barrier via test_and_set_bit() in io_wq_set_exit_on_idle(). When it
has closed all rings though that barrier is gone.

Add the barrier after the set_bit().

Fixes: 71a85387546e ("io-wq: check for wq exit after adding new worker task_work")
Cc: stable@vger.kernel.org
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Link: https://patch.msgid.link/20260921-work-coredump-fixes-v3-4-8e4adb1619e6@kernel.org
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 io_uring/io-wq.c |    2 ++
 1 file changed, 2 insertions(+)

--- a/io_uring/io-wq.c
+++ b/io_uring/io-wq.c
@@ -1319,6 +1319,8 @@ static bool io_task_work_match(struct ca
 void io_wq_exit_start(struct io_wq *wq)
 {
 	set_bit(IO_WQ_BIT_EXIT, &wq->state);
+	/* Pairs with task_work_add() in io_queue_worker_create(). */
+	smp_mb__after_atomic();
 }
 
 static void io_wq_cancel_tw_create(struct io_wq *wq)



  parent reply	other threads:[~2026-10-28 17:06 UTC|newest]

Thread overview: 95+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-28 16:59 [PATCH 7.2 00/88] 7.2.14-rc1 review Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 01/88] sched_ext: Wait for SCX_OPSS_DISPATCHING before moving tasks off DSQs Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 02/88] ntfs: fix race between fallocate and mmap reads Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 03/88] ntfs: take invalidate_lock in ntfs_filemap_page_mkwrite() Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 04/88] mm/damon/core: error damos_commit_quota_goal() for zero target_value Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 05/88] sched/fair: Avoid overflow in place_entity() Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 06/88] Bluetooth: L2CAP: refuse __l2cap_chan_add if chan already has conn Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 07/88] tools/dma: fix building with O= Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 08/88] tracing: Hold trace_types_lock when updating modules in persistent instances Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 09/88] tracing: Hold trace_types_lock when resetting last boot data from readers Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 10/88] tracing: Reject incomplete event filter expressions Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 11/88] tracing: Save the event file instead of the compatible histogram Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 12/88] tracing: Unregister field variable histograms when the command fails Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 13/88] tracing: Fix 64-bit and signed histogram fields on 32-bit kernels Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 14/88] tracing/histograms: Verify event_call identity when updating synth_event refcount Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 15/88] tracing/syscalls: Use copy_from_user for the faultable syscall buffer Greg Kroah-Hartman
2026-10-28 16:59 ` Greg Kroah-Hartman [this message]
2026-10-28 16:59 ` [PATCH 7.2 17/88] io_uring/waitid: own the request before clearing ->head Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 18/88] io_uring/cmd: dont leave list poison in ->hash_node Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 19/88] io_uring/wait: dont spin on a stuck overflow list Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 20/88] io_uring/io-wq: dont lose the signal that ends work Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 21/88] io_uring/io-wq: dont hide a link from cancellations Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 22/88] io_uring: dont flood a cancelling task with task_work Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 23/88] io_uring: dont lose the DEFER_TASKRUN wake up Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 24/88] io_uring/cancel: rearm local work wake up before sleep Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 25/88] io_uring/mpscq: disable preemption while pushing Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 26/88] io_uring/io-wq: dont go to sleep on pending work Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 27/88] io_uring/io-wq: dont strand work when a worker exits Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 28/88] io_uring/io-wq: dont exit with a creation pending Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 29/88] io_uring/io-wq: let cancel find picked up hashed work Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 30/88] io_uring/poll: dont lose a cancel as io-wq arms poll Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 31/88] io_uring: dont lose a cancel as io-wq parks a request Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 32/88] io_uring/cancel: dont forget that the work was running Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 33/88] io_uring/cancel: recognize "no timeout" on 32-bit Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 34/88] io_uring/cancel: let CANCEL_ALL find parked requests Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 35/88] io_uring: wake ring pollers for local task_work Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 36/88] io_uring/wait: look at all CQ waiters on wake up Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 37/88] io_uring/poll: drop stale double poll entry on rearm Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 38/88] io_uring/futex: allocate private hash before waiting Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 39/88] io_uring: drop the CQE cache if a refill fails Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 40/88] io_uring: dont split 128b SQEs at the SQPOLL cap Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 41/88] io_uring/register: dont wrap 32b CQEs when resizing Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 42/88] io_uring/register: handle index wrap " Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 43/88] io_uring/register: keep 128b SQEs intact on resize Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 44/88] io_uring/fdinfo: fix CQE dump of CQE32 rings Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 45/88] io_uring/timeout: dont skip over reached sequences Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 46/88] io_uring/timeout: keep repeat count on multishot update Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 47/88] io_uring/timeout: use the right clock for ABS updates Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 48/88] io_uring/wait: no early -ETIME if min_wait finds work Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 49/88] io_uring/napi: make stale entries time out after 60s Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 50/88] io_uring/napi: dont busy poll past an expired timeout Greg Kroah-Hartman
2026-10-28 16:59 ` [PATCH 7.2 51/88] io_uring/zcrx: fix buffer offsets in areas above 4GB Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 52/88] io_uring/kbuf: handle index wrap in last buffer check Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 53/88] io_uring/sqpoll: let the thread sleep on disabled rings Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 54/88] io_uring/openclose: keep the lock in io_pipe_fixed() Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 55/88] io_uring: dont pass restart error codes to userspace Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 56/88] io_uring/net: dont retry recv on uncommitted buffers Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 57/88] io_uring/timeout: no ABS update of multishot timeouts Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 58/88] io_uring/timeout: dont end a linked timeout with 1 Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 59/88] io_uring: look up the file before queueing to io-wq Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 60/88] io_uring/poll: avoid retry wakeup loss when dropping ownership Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 61/88] io_uring/timeout: dont put a sequence target behind the last flush Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 62/88] io_uring/timeout: dont flush with the sequence " Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 63/88] io_uring/timeout: dont lose a cancel as io-wq arms a timeout Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 64/88] io_uring/poll: dont loop on HUP with multiple references Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 65/88] io_uring/net: keep recv buffer select length within MAX_RW_COUNT Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 66/88] io_uring/wait: wake real-time waiters from hard interrupt context Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 67/88] io_uring/waitid: prevent cancel reference wrap Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 68/88] io_uring: end the drain when the drain request completes Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 69/88] io_uring: keep the drain active when a deferred drain request is started Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 70/88] keys: Avoid the owner account dereference in named keyring lookup Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 71/88] keys: Check the authorization payload in process keyring searches Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 72/88] keys: Protect the type name while describing a key Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 73/88] keys: make keyring indexes independent of kernel addresses Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 74/88] keys: Protect the read method check against type removal Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 75/88] KEYS: Fix key_reject_and_link() race with keyring restriction Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 76/88] KEYS: trusted: Fix blob allocation size in tpm2_key_decode() Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 77/88] keys/trusted_keys: return immediately after TPM unseal failure Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 78/88] asymmetric_keys: reject trust keys without IDs Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 79/88] ftrace: Initialize modname and offset in print_rec() Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 80/88] coredump: hold RCU while releasing parked threads Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 81/88] signal: only SIGKILL and the freezers interrupt a coredumping task Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 82/88] signal: dont retarget shared signals in a dying thread group Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 83/88] ptrace: refuse to change the signal mask of a user worker Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 84/88] fork: release the files of a failed fork after sched_cancel_fork() Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 85/88] exit: hang up the tty before closing the files Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 86/88] coredump: prevent TIF_NOTIFY_SIGNAL from interrupting coredumps Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 87/88] pid_namespace: prevent TIF_NOTIFY_SIGNAL from interrupting the reaper Greg Kroah-Hartman
2026-10-28 17:00 ` [PATCH 7.2 88/88] signal: allow taks to temporarily block TIF_NOTIFY_SIGNAL Greg Kroah-Hartman
2026-10-28 19:39 ` [PATCH 7.2 00/88] 7.2.14-rc1 review Brett A C Sheffield
2026-10-29  4:09 ` Peter Schneider
2026-10-29 10:08 ` Pavel Machek
2026-10-29 13:49 ` Miguel Ojeda
2026-10-29 14:17 ` Ron Economos
2026-10-29 14:43 ` SJ Park

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261028165942.497675794@linuxfoundation.org \
    --to=gregkh@linuxfoundation.org \
    --cc=axboe@kernel.dk \
    --cc=brauner@kernel.org \
    --cc=oleg@redhat.com \
    --cc=patches@lists.linux.dev \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox