io_uring: fix possible deadlock in io_register_iowq_max_workers()

The io_register_iowq_max_workers() function calls io_put_sq_data(),
which acquires the sqd->lock without releasing the uring_lock.
Similar to the commit 009ad9f0c6ee ("io_uring: drop ctx->uring_lock
before acquiring sqd->lock"), this can lead to a potential deadlock
situation.

To resolve this issue, the uring_lock is released before calling
io_put_sq_data(), and then it is re-acquired after the function call.

This change ensures that the locks are acquired in the correct
order, preventing the possibility of a deadlock.

Suggested-by: Maximilian Heyne <mheyne@amazon.de>
Signed-off-by: Hagar Hemdan <hagarhem@amazon.com>
Link: https://lore.kernel.org/r/20240604130527.3597-1-hagarhem@amazon.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/io_uring/register.c b/io_uring/register.c
index ef8c908346a4ef9ed5d4a3a8015e38612d0bd75d..c0010a66a6f2c2e72c795b10e6ad3773716200ac 100644 (file)
--- a/io_uring/register.c
+++ b/io_uring/register.c
@@ -355,8 +355,10 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx,
        }
 
        if (sqd) {
+               mutex_unlock(&ctx->uring_lock);
                mutex_unlock(&sqd->lock);
                io_put_sq_data(sqd);
+               mutex_lock(&ctx->uring_lock);
        }
 
        if (copy_to_user(arg, new_count, sizeof(new_count)))
@@ -380,8 +382,10 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx,
        return 0;
 err:
        if (sqd) {
+               mutex_unlock(&ctx->uring_lock);
                mutex_unlock(&sqd->lock);
                io_put_sq_data(sqd);
+               mutex_lock(&ctx->uring_lock);
        }
        return ret;
 }