2020-03-06 22:16:50

by Pavel Begunkov

Subject: [PATCH 1/1] io_uring: fix lockup with timeouts

There is a recipe to deadlock the kernel: submit a timeout sqe with a
linked_timeout (e.g. test_single_link_timeout_ception() from liburing),
and SIGKILL the process.

Then, io_kill_timeouts() takes @ctx->completion_lock, but the timeout
isn't flagged with REQ_F_COMP_LOCKED, and will try to double grab it
during io_put_free() to cancel the linked timeout. Probably, the same
can happen with another io_kill_timeout() call site, that is
io_commit_cqring().

Signed-off-by: Pavel Begunkov <[email protected]>
---
fs/io_uring.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 104f76aace29..94eca92d1354 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1089,6 +1089,7 @@ static void io_kill_timeout(struct io_kiocb *req)
if (ret != -1) {
atomic_inc(&req->ctx->cq_timeouts);
list_del_init(&req->list);
+ req->flags |= REQ_F_COMP_LOCKED;
io_cqring_fill_event(req, 0);
io_put_req(req);
}
--
2.24.0

2020-03-07 15:37:31

by Jens Axboe

Subject: Re: [PATCH 1/1] io_uring: fix lockup with timeouts

On 3/6/20 3:15 PM, Pavel Begunkov wrote:
> There is a recipe to deadlock the kernel: submit a timeout sqe with a
> linked_timeout (e.g. test_single_link_timeout_ception() from liburing),
> and SIGKILL the process.
>
> Then, io_kill_timeouts() takes @ctx->completion_lock, but the timeout
> isn't flagged with REQ_F_COMP_LOCKED, and will try to double grab it
> during io_put_free() to cancel the linked timeout. Probably, the same
> can happen with another io_kill_timeout() call site, that is
> io_commit_cqring().

Thanks, looks good, applied.

--
Jens Axboe