On Mon, Apr 14, 2025 at 10:49:49AM +0000, Benno Lossin wrote:
> On Sun Apr 13, 2025 at 7:37 PM CEST, Danilo Krummrich wrote:
> > The Bound device context indicates that a device is bound to a driver.
> > It must be used for APIs that require the device to be bound, such as
> > Devres or dma::CoherentAllocation.
> >
> > Implement Bound and add the corresponding Deref hierarchy, as well as the
> > corresponding ARef conversion for this device context.
> >
> > Signed-off-by: Danilo Krummrich <dakr@xxxxxxxxxx>
> > ---
> > rust/kernel/device.rs | 16 +++++++++++++++-
> > 1 file changed, 15 insertions(+), 1 deletion(-)
> >
> > diff --git a/rust/kernel/device.rs b/rust/kernel/device.rs
> > index 487211842f77..585a3fcfeea3 100644
> > --- a/rust/kernel/device.rs
> > +++ b/rust/kernel/device.rs
> > @@ -232,13 +232,19 @@ pub trait DeviceContext: private::Sealed {}
> > /// any of the bus callbacks, such as `probe()`.
> > pub struct Core;
> >
> > +/// The [`Bound`] context is the context of a bus specific device reference when it is guranteed to
> > +/// be bound for the duration of its lifetime.
> > +pub struct Bound;
>
> One question about this: is it possible for me to
> 1. have access to a `ARef<Device<Bound>>` (or `Core`) via some callback,
> 2. store a clone of the `ARef` in some datastructure,
> 3. wait for the device to become unbound,
> 4. use a `Bound`-only context function and blow something up?
You can never get an ARef<Device> that has a different device context than
Normal.
A device must only ever implement AlwaysRefCounted for Device (i.e.
Device<Normal>).
This is why patch 2 ("rust: device: implement impl_device_context_into_aref!")
implements conversions from Device<Ctx> to ARef<Device>.
