Re: [PATCH v2] workqueue: rust: add delayed work items

[Alice Ryhl <aliceryhl@google.com>]

On Fri, Jun 27, 2025 at 10:43:52AM -0700, Boqun Feng wrote:
> On Fri, Jun 27, 2025 at 09:38:42AM +0000, Alice Ryhl wrote:
> [...]
> > +    /// Get a pointer to the inner `delayed_work`.
> > +    ///
> > +    /// # Safety
> > +    ///
> > +    /// The provided pointer must not be dangling and must be properly aligned. (But the memory
> > +    /// need not be initialized.)
> > +    #[inline]
> > +    pub unsafe fn raw_as_work(ptr: *const Self) -> *mut Work<T, ID> {
> > +        // SAFETY: The caller promises that the pointer is aligned and not dangling.
> > +        let dw: *mut bindings::delayed_work =
> > +            unsafe { Opaque::raw_get(core::ptr::addr_of!((*ptr).dwork)) };
> 
> This would conflict with your Opaque::cast_into() series ;-) So I
> suggest that we rebase this one onto:
> 
> 	https://lore.kernel.org/rust-for-linux/20250624-opaque-from-raw-v2-0-e4da40bdc59c@google.com/
> 
> and send them together. But it's up to you and maintainers, and even we
> do that let's wait for some reviews on this patch to save extra
> versions.

Will rebase.

> I'm going to review it, just figure this better be pointed out.
> 
> Btw, as a review comment, I believe we are in favor of `&raw mut` now,
> so this should be:
> 
> 
>     unsafe { Opaque::cast_from(&raw const (*ptr).dwork) };
> 
> > +        // SAFETY: The caller promises that the pointer is aligned and not dangling.
> > +        let wrk: *mut bindings::work_struct = unsafe { core::ptr::addr_of_mut!((*dw).work) };
> 
> Ditto.

The rest of the file still uses addr_of!. I think we should stay
consistent. We can update the entire file in one go.

Alice