Re: [PATCH v4 09/16] rust: workqueue: refactor to use `&raw [const|mut]`

[Boqun Feng <boqun.feng@gmail.com>]

On Sun, Mar 16, 2025 at 09:21:00PM -0700, Boqun Feng wrote:
> On Sun, Mar 16, 2025 at 02:14:18AM -0400, Antonio Hickey wrote:
> > Replacing all occurrences of `addr_of!(place)` and `addr_of_mut!(place)`
> > with `&raw const place` and `&raw mut place` respectively.
> > 
> > This will allow us to reduce macro complexity, and improve consistency
> > with existing reference syntax as `&raw const`, `&raw mut` are similar
> > to `&`, `&mut` making it fit more naturally with other existing code.
> > 
> > Suggested-by: Benno Lossin <benno.lossin@proton.me>
> > Link: https://github.com/Rust-for-Linux/linux/issues/1148
> > Signed-off-by: Antonio Hickey <contact@antoniohickey.com>
> > ---
> >  rust/kernel/workqueue.rs | 9 +++++----
> >  1 file changed, 5 insertions(+), 4 deletions(-)
> > 
> > diff --git a/rust/kernel/workqueue.rs b/rust/kernel/workqueue.rs
> > index 0cd100d2aefb..4e27df324d26 100644
> > --- a/rust/kernel/workqueue.rs
> > +++ b/rust/kernel/workqueue.rs
> > @@ -401,9 +401,10 @@ pub fn new(name: &'static CStr, key: &'static LockClassKey) -> impl PinInit<Self
> >      pub unsafe fn raw_get(ptr: *const Self) -> *mut bindings::work_struct {
> >          // SAFETY: The caller promises that the pointer is aligned and not dangling.
> >          //
> > -        // A pointer cast would also be ok due to `#[repr(transparent)]`. We use `addr_of!` so that
> > -        // the compiler does not complain that the `work` field is unused.
> > -        unsafe { Opaque::raw_get(core::ptr::addr_of!((*ptr).work)) }
> > +        // A pointer cast would also be ok due to `#[repr(transparent)]`. We use
> > +        // `&raw const (*ptr).work` so that the compiler does not complain that the
> > +        // `work` field is unused.
> 
> I think we can actually use the pointer casting here. Because now we use
> pin-init for Work initialization, so `work` field is always used.
> 
> Could you replace this with a:
> 
> 	// CAST: `Work` is transparent to `bindings::work_struct`.
> 	ptr.cast_mut().cast()
> 
> in a separate patch?
> 

Hmm.. think more about this, this raw_get() function can be a safe
function because it's just a pointer casting? I've checked the
callsites, it doesn't seems that they need the pointer provenance
from Opaque::raw_get().

Alice, any objection? Am I missing something subtle here? We can also
remove this function at all, and use cast() in __enqueue()s.

Regards,
Boqun

> Thanks!
> 
> Regards,
> Boqun
> 
> > +        unsafe { Opaque::raw_get(&raw const (*ptr).work) }
> >      }
> >  }
> >  
> > @@ -510,7 +511,7 @@ macro_rules! impl_has_work {
> >              unsafe fn raw_get_work(ptr: *mut Self) -> *mut $crate::workqueue::Work<$work_type $(, $id)?> {
> >                  // SAFETY: The caller promises that the pointer is not dangling.
> >                  unsafe {
> > -                    ::core::ptr::addr_of_mut!((*ptr).$field)
> > +                    &raw mut (*ptr).$field
> >                  }
> >              }
> >          }
> > -- 
> > 2.48.1
> >