]> git.hungrycats.org Git - linux/commitdiff
io_uring: fix possible deadlock in io_register_iowq_max_workers()
authorHagar Hemdan <hagarhem@amazon.com>
Tue, 4 Jun 2024 13:05:27 +0000 (13:05 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 17 Nov 2024 13:59:39 +0000 (14:59 +0100)
commit 73254a297c2dd094abec7c9efee32455ae875bdf upstream.

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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
io_uring/io_uring.c

index 57c51e9638753d0d7c137c570aee2ff852366c53..4e86da84f38ab611e96002a1a384f8a6d70a5add 100644 (file)
@@ -10653,8 +10653,10 @@ static 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)))
@@ -10679,8 +10681,11 @@ static 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;
 }