[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5640fbf1-7b8d-4537-9f1a-b401a7a4934b@igalia.com>
Date: Fri, 24 Oct 2025 13:40:03 +0100
From: Tvrtko Ursulin <tvrtko.ursulin@...lia.com>
To: Danilo Krummrich <dakr@...nel.org>
Cc: phasta@...nel.org, Sumit Semwal <sumit.semwal@...aro.org>,
Gustavo Padovan <gustavo@...ovan.org>,
Christian König <christian.koenig@....com>,
linux-media@...r.kernel.org, dri-devel@...ts.freedesktop.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] dma-fence: Correct return of dma_fence_driver_name()
On 24/10/2025 13:12, Danilo Krummrich wrote:
> On 10/24/25 1:31 PM, Tvrtko Ursulin wrote:
>> Also, the short name can be reduced from a verbose starting point similar to yours:
>>
>> "unknown-driver-is-detached-from-the-signaled-fence"
>> "driver-detached-from-the-fence"
>> "driver-detached"
>>
>> Or keep "detached-driver" as good enough. Mea culpa for typing it up transposed. :)
>
> "detached-driver" is misleading, just because the fence is signaled it doesn't
> mean that the driver is detached.
You trim too much of the quote making it unclear if you read the whole
story.
If the driver isn't detached from the signalled fence then it is
vulnerable to use after free.
Current state is that name and timeline helpers _are_ detached from the
driver once the fence is signalled.
Fence ops and lock are still not and that is work in progress.
The names under discussion are printed together in debugfs (intended
audience kernel developers). Currently, per fence, it should be along
the lines of:
"kernel fence: detached-driver signaled-timeline seq 1234 signalled"
You can propose a better name but make it not too verbose and/or
redundant considering the above.
Regards,
Tvrtko
Powered by blists - more mailing lists