[PATCH v2 2/2] rust: drm: Add GPUVM abstraction

[Daniel Almeida]

> 
> The lock might be held already by the driver or by TTM when things are called
> from TTM callbacks.
> 
> This is why GPUVM never takes locks by itself, but asserts that the correct lock
> is held.
> 
> I think we really want to get proof by the driver by providing lock guard
> references.
> 
>> See my comment about “[..] a new type for GEM objects which are known to be locked"
>> below.
> 
> <snip>
> 
>>>> +
>>>> +    /// Obtains the [`GpuVmBo`] object that connects `obj` to this VM.
>>>> +    ///
>>>> +    /// This connection is unique, so an instane of [`GpuVmBo`] will be
>>>> +    /// allocated for `obj` once, and that instance will be returned from that
>>>> +    /// point forward.
>>>> +    pub fn obtain_bo(&mut self, obj: &DriverObject<T>) -> Result<ARef<GpuVmBo<T>>> {
>>>> +        // SAFETY: LockedGpuVm implies the right locks are held.
>>> 
>>> No, this must be locked by the dma-resv or the GEM gpuva lock, not by the
>>> GPUVM lock that protects the interval tree.
>> 
>> By “GEM gpuva lock” you’re referring to the custom lock which we
>> currently do not support, right?
> 
> Yes.
> 
>> This series currently rely on manual calls to dma_resv_{lock,unlock}, I wonder
>> if we should ditch that in favor of something written in Rust directly. This
>> would let us introduce a new type for GEM objects which are known to have
>> `resv` locked. WDYT?
> 
> Not all functions that require the dma-resv lock to be held are called with a
> GEM object parameter, it could also be a struct drm_gpuvm_bo, struct drm_gpuva
> or struct drm_gpuvm, since they all carry GEM object pointers.
> 
> For reference, you can look for "_held" in drivers/gpu/drm/drm_gpuvm.c.
> 

Looking at Lyude’s (excellent) KMS series, one thing that comes to mind is
using Lock::from_raw() on the dma-resv lock. This will build a rust Mutex that
we can then assert to be locked (or fail with an Error otherwise).

See [0] for the specific patch whose idea I want to copy.

Can I get a +1 on this idea before pursuing it?

-- Daniel

[0] https://lore.kernel.org/rust-for-linux/20250305230406.567126-10-lyude@redhat.com/