From: Alice Ryhl <aliceryhl@google.com>
To: Sami Tolvanen <samitolvanen@google.com>
Cc: "Danilo Krummrich" <dakr@kernel.org>,
"Matthew Brost" <matthew.brost@intel.com>,
"Thomas Hellström" <thomas.hellstrom@linux.intel.com>,
"David Airlie" <airlied@gmail.com>,
"Simona Vetter" <simona@ffwll.ch>,
"Miguel Ojeda" <ojeda@kernel.org>,
"Boqun Feng" <boqun@kernel.org>, "Gary Guo" <gary@garyguo.net>,
"Björn Roy Baron" <bjorn3_gh@protonmail.com>,
"Benno Lossin" <lossin@kernel.org>,
"Andreas Hindborg" <a.hindborg@kernel.org>,
"Trevor Gross" <tmgross@umich.edu>,
"Daniel Almeida" <daniel.almeida@collabora.com>,
"Asahi Lina" <lina+kernel@asahilina.net>,
dri-devel@lists.freedesktop.org, rust-for-linux@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3 2/2] rust: drm: gpuvm: require Send and Sync for GpuVm's shared data
Date: Wed, 10 Jun 2026 07:13:54 +0000 [thread overview]
Message-ID: <aikOstbyTYGuV8xp@google.com> (raw)
In-Reply-To: <CABCJKucVW-FF2sti9gVoKkp-StLMof7kSqisbmUvjyEAaKLuTw@mail.gmail.com>
On Mon, Jun 08, 2026 at 06:54:28PM -0700, Sami Tolvanen wrote:
> On Mon, Jun 8, 2026 at 5:33 PM Sami Tolvanen <samitolvanen@google.com> wrote:
> >
> > GpuVm implements Send and Sync unconditionally, but it shares and drops
> > T::VmBoData and T::Object across threads: obtain() can return ARefs to
> > the same bo on several threads, and deferred_cleanup() drops them on
> > whichever thread calls it. This is unsound unless both are Send + Sync,
> > so bound both impls accordingly.
>
> Sashiko keeps finding more issues here. For the benefit of those not
> subscribed to dri-devel:
>
> > This is a pre-existing issue, but does GpuVm also need Send + Sync bounds
> > for T::VaData here?
> >
> > Since GpuVm<T> explicitly bypasses the lack of Send on VaData,
> > UniqueRefGpuVm<T> automatically derives Send. This allows safe code to map
> > a VA containing !Send data on one thread, send the UniqueRefGpuVm<T> to
> > another thread, and unmap the VA there.
> >
> > Unmapping via sm_unmap() returns a GpuVaRemoved<T> which drops the !Send
> > VaData on the wrong thread, potentially causing data races.
>
> Looking at the code, UniqueRefGpuVm seems like a better place for this
> than GpuVm. However, before I go add another patch to the series, I
> wonder if it would make more sense to just add these bounds to the
> DriverGpuVm trait instead? I'd be happy to hear some non-AI thoughts
> about this too.
Adding bounds on the trait does sound like a good idea for simplicity.
It's not like we're ever going to use GPUVM with non-thread-safe types.
If it's help, note that you can use `where GpuVm<T>: Send` bounds or
similar to "reuse" the implementations elsewhere. For example,
GpuVmBo<T> being Sync requires GpuVm<T> being Sync because of the
GpuVmBo::gpuvm() method, so you can use `where GpuVm<T>: Send` as a
bound to directly express that if you want.
But if we just add bounds to the trait, it may not be necessary.
Alice
prev parent reply other threads:[~2026-06-10 7:13 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20260609003256.1829625-4-samitolvanen@google.com>
2026-06-09 0:32 ` [PATCH v3 1/2] rust: drm: gpuvm: implement Send and Sync for GpuVaAlloc and GpuVmBo Sami Tolvanen
2026-06-09 0:32 ` [PATCH v3 2/2] rust: drm: gpuvm: require Send and Sync for GpuVm's shared data Sami Tolvanen
2026-06-09 1:54 ` Sami Tolvanen
2026-06-10 7:13 ` Alice Ryhl [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=aikOstbyTYGuV8xp@google.com \
--to=aliceryhl@google.com \
--cc=a.hindborg@kernel.org \
--cc=airlied@gmail.com \
--cc=bjorn3_gh@protonmail.com \
--cc=boqun@kernel.org \
--cc=dakr@kernel.org \
--cc=daniel.almeida@collabora.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=gary@garyguo.net \
--cc=lina+kernel@asahilina.net \
--cc=linux-kernel@vger.kernel.org \
--cc=lossin@kernel.org \
--cc=matthew.brost@intel.com \
--cc=ojeda@kernel.org \
--cc=rust-for-linux@vger.kernel.org \
--cc=samitolvanen@google.com \
--cc=simona@ffwll.ch \
--cc=thomas.hellstrom@linux.intel.com \
--cc=tmgross@umich.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox