diff options
| author | Pavel Begunkov <asml.silence@gmail.com> | 2021-03-15 17:23:08 +0300 |
|---|---|---|
| committer | Jens Axboe <axboe@kernel.dk> | 2021-03-15 18:32:40 +0300 |
| commit | b7f5a0bfe2061b2c7b2164de06fa4072d7373a45 (patch) | |
| tree | 2ad3c494d8f86a89c2d3ec30db5852df025c1cd8 /fs/io_uring.c | |
| parent | 9b46571142e47503ed4f3ae3be5ed3968d8cb9cc (diff) | |
| download | linux-b7f5a0bfe2061b2c7b2164de06fa4072d7373a45.tar.xz | |
io_uring: fix sqpoll cancellation via task_work
Running sqpoll cancellations via task_work_run() is a bad idea because
it depends on other task works to be run, but those may be locked in
currently running task_work_run() because of how it's (splicing the list
in batches).
Enqueue and run them through a separate callback head, namely
struct io_sq_data::park_task_work. As a nice bonus we now precisely
control where it's run, that's much safer than guessing where it can
happen as it was before.
Reported-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Diffstat (limited to 'fs/io_uring.c')
| -rw-r--r-- | fs/io_uring.c | 6 |
1 files changed, 5 insertions, 1 deletions
diff --git a/fs/io_uring.c b/fs/io_uring.c index 5c2de43b99f5..5f4e312111ea 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -274,6 +274,7 @@ struct io_sq_data { unsigned long state; struct completion exited; + struct callback_head *park_task_work; }; #define IO_IOPOLL_BATCH 8 @@ -6727,6 +6728,7 @@ static int io_sq_thread(void *data) cond_resched(); mutex_lock(&sqd->lock); io_run_task_work(); + io_run_task_work_head(&sqd->park_task_work); timeout = jiffies + sqd->sq_thread_idle; continue; } @@ -6781,6 +6783,7 @@ static int io_sq_thread(void *data) } finish_wait(&sqd->wait, &wait); + io_run_task_work_head(&sqd->park_task_work); timeout = jiffies + sqd->sq_thread_idle; } @@ -6792,6 +6795,7 @@ static int io_sq_thread(void *data) mutex_unlock(&sqd->lock); io_run_task_work(); + io_run_task_work_head(&sqd->park_task_work); complete(&sqd->exited); do_exit(0); } @@ -8890,7 +8894,7 @@ static void io_sqpoll_cancel_sync(struct io_ring_ctx *ctx) if (task) { init_completion(&work.completion); init_task_work(&work.task_work, io_sqpoll_cancel_cb); - WARN_ON_ONCE(task_work_add(task, &work.task_work, TWA_SIGNAL)); + io_task_work_add_head(&sqd->park_task_work, &work.task_work); wake_up_process(task); } io_sq_thread_unpark(sqd); |