[PATCH 6.1] io_uring: fix possible deadlock in io_register_iowq_max_workers()

[PATCH 6.1] io_uring: fix possible deadlock in io_register_iowq_max_workers()

[Hagar Hemdan]

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>
[Hagar: Modified to apply on v6.1]
Signed-off-by: Hagar Hemdan <hagarhem@amazon.com>
---
 io_uring/io_uring.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 92c1aa8f3501..4f0ae938b146 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3921,8 +3921,10 @@ 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)))
@@ -3947,8 +3949,11 @@ 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;
 }
-- 
2.40.1

Re: [PATCH 6.1] io_uring: fix possible deadlock in io_register_iowq_max_workers()

[Greg KH]

On Tue, Nov 12, 2024 at 08:30:06AM +0000, Hagar Hemdan wrote:
> 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>
> [Hagar: Modified to apply on v6.1]
> Signed-off-by: Hagar Hemdan <hagarhem@amazon.com>
> ---
>  io_uring/io_uring.c | 5 +++++
>  1 file changed, 5 insertions(+)

What about 6.6.y?  We can't just take patches for older branches and not
newer ones, you know this :)

thanks,

greg k-h

[PATCH 6.1] io_uring: fix possible deadlock in io_register_iowq_max_workers()

[Hagar Hemdan]

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>
[Hagar: Modified to apply on v6.1]
Signed-off-by: Hagar Hemdan <hagarhem@amazon.com>
---
 io_uring/io_uring.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 92c1aa8f3501..4f0ae938b146 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3921,8 +3921,10 @@ 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)))
@@ -3947,8 +3949,11 @@ 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;
 }
-- 
2.40.1

Re: [PATCH 6.1] io_uring: fix possible deadlock in io_register_iowq_max_workers()

[Jens Axboe]

On 11/12/24 1:39 AM, Greg KH wrote:
> On Tue, Nov 12, 2024 at 08:30:06AM +0000, Hagar Hemdan wrote:
>> 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>
>> [Hagar: Modified to apply on v6.1]
>> Signed-off-by: Hagar Hemdan <hagarhem@amazon.com>
>> ---
>>  io_uring/io_uring.c | 5 +++++
>>  1 file changed, 5 insertions(+)
> 
> What about 6.6.y?  We can't just take patches for older branches and not
> newer ones, you know this :)

Hagar, thanks for doing the other ones too. Greg, they look fine to me.

-- 
Jens Axboe

Re: [PATCH 6.1] io_uring: fix possible deadlock in io_register_iowq_max_workers()

[Greg KH]

On Tue, Nov 12, 2024 at 07:56:45AM -0700, Jens Axboe wrote:
> On 11/12/24 1:39 AM, Greg KH wrote:
> > On Tue, Nov 12, 2024 at 08:30:06AM +0000, Hagar Hemdan wrote:
> >> 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>
> >> [Hagar: Modified to apply on v6.1]
> >> Signed-off-by: Hagar Hemdan <hagarhem@amazon.com>
> >> ---
> >>  io_uring/io_uring.c | 5 +++++
> >>  1 file changed, 5 insertions(+)
> > 
> > What about 6.6.y?  We can't just take patches for older branches and not
> > newer ones, you know this :)
> 
> Hagar, thanks for doing the other ones too. Greg, they look fine to me.

Thanks, all now queued up.

greg k-h