[PATCH 0/2] LAZY_WAKE misuse fixes

[PATCH 0/2] LAZY_WAKE misuse fixes

[Pavel Begunkov]

LAZY_WAKE can't be used with requests for which users expect >1 CQE,
otherwise wait(nr=2) may get stuck waiting for a tw that will never
arrive. Fix up invalid use of LAZY_WAKE with sendzc.

Pavel Begunkov (2):
  io_uring/net: fix sendzc lazy wake polling
  io_uring/notif: disable LAZY_WAKE for linked notifs

 io_uring/net.c   | 1 +
 io_uring/notif.c | 5 ++++-
 2 files changed, 5 insertions(+), 1 deletion(-)

-- 
2.44.0

[PATCH 1/2] io_uring/net: fix sendzc lazy wake polling

[Pavel Begunkov]

SEND[MSG]_ZC produces multiple CQEs via notifications, LAZY_WAKE doesn't
handle it and so disable LAZY_WAKE for sendzc polling. It should be
fine, sends are not likely to be polled in the first place.

Fixes: 6ce4a93dbb5b ("io_uring/poll: use IOU_F_TWQ_LAZY_WAKE for wakeups")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 io_uring/net.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/io_uring/net.c b/io_uring/net.c
index 51c41d771c50..503debecad32 100644
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -1198,6 +1198,7 @@ int io_send_zc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	struct io_kiocb *notif;
 
 	zc->done_io = 0;
+	req->flags |= REQ_F_POLL_NO_LAZY;
 
 	if (unlikely(READ_ONCE(sqe->__pad2[0]) || READ_ONCE(sqe->addr3)))
 		return -EINVAL;
-- 
2.44.0

[PATCH 2/2] io_uring/notif: disable LAZY_WAKE for linked notifs

[Pavel Begunkov]

Notifications may now be linked and thus a single tw can post multiple
CQEs, it's not safe to use LAZY_WAKE with them. Disable LAZY_WAKE for
now, if that'd prove to be a problem we can count them and pass the
expected number of CQEs into __io_req_task_work_add().

Fixes: 6fe4220912d19 ("io_uring/notif: implement notification stacking")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 io_uring/notif.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/io_uring/notif.c b/io_uring/notif.c
index d58cdc01e691..28859ae3ee6e 100644
--- a/io_uring/notif.c
+++ b/io_uring/notif.c
@@ -38,6 +38,7 @@ void io_tx_ubuf_complete(struct sk_buff *skb, struct ubuf_info *uarg,
 {
 	struct io_notif_data *nd = container_of(uarg, struct io_notif_data, uarg);
 	struct io_kiocb *notif = cmd_to_io_kiocb(nd);
+	unsigned tw_flags;
 
 	if (nd->zc_report) {
 		if (success && !nd->zc_used && skb)
@@ -53,8 +54,10 @@ void io_tx_ubuf_complete(struct sk_buff *skb, struct ubuf_info *uarg,
 		io_tx_ubuf_complete(skb, &nd->head->uarg, success);
 		return;
 	}
+
+	tw_flags = nd->next ? 0 : IOU_F_TWQ_LAZY_WAKE;
 	notif->io_task_work.func = io_notif_tw_complete;
-	__io_req_task_work_add(notif, IOU_F_TWQ_LAZY_WAKE);
+	__io_req_task_work_add(notif, tw_flags);
 }
 
 static int io_link_skb(struct sk_buff *skb, struct ubuf_info *uarg)
-- 
2.44.0

Re: [PATCH 0/2] LAZY_WAKE misuse fixes

[Jens Axboe]

On Tue, 30 Apr 2024 16:42:29 +0100, Pavel Begunkov wrote:
> LAZY_WAKE can't be used with requests for which users expect >1 CQE,
> otherwise wait(nr=2) may get stuck waiting for a tw that will never
> arrive. Fix up invalid use of LAZY_WAKE with sendzc.
> 
> Pavel Begunkov (2):
>   io_uring/net: fix sendzc lazy wake polling
>   io_uring/notif: disable LAZY_WAKE for linked notifs
> 
> [...]

Applied, thanks!

[1/2] io_uring/net: fix sendzc lazy wake polling
      commit: ef42b85a5609cd822ca0a68dd2bef2b12b5d1ca3
[2/2] io_uring/notif: disable LAZY_WAKE for linked notifs
      commit: 19352a1d395424b5f8c03289a85fbd6622d6601a

Best regards,
-- 
Jens Axboe