On Sat, May 17, 2025 at 10:28:04PM +0900, FUJITA Tomonori wrote:
On Sat, 17 May 2025 10:09:53 +0200
"Benno Lossin" [off-list ref] wrote:
quoted
On Fri May 16, 2025 at 11:23 PM CEST, Christian Marangi wrote:
quoted
diff --git a/rust/kernel/net/phy.rs b/rust/kernel/net/phy.rs
index a59469c785e3..079a0f884887 100644
--- a/rust/kernel/net/phy.rs
+++ b/rust/kernel/net/phy.rs
@@ -418,15 +418,18 @@ impl<T: Driver> Adapter<T> {
/// # Safety
///
- /// `phydev` must be passed by the corresponding callback in `phy_driver`.
+ /// `phydev` and `phydrv` must be passed by the corresponding callback in
+ // `phy_driver`.
unsafe extern "C" fn match_phy_device_callback(
phydev: *mut bindings::phy_device,
+ phydrv: *const bindings::phy_driver,
) -> crate::ffi::c_int {
// SAFETY: This callback is called only in contexts
// where we hold `phy_device->lock`, so the accessors on
// `Device` are okay to call.
let dev = unsafe { Device::from_raw(phydev) };
- T::match_phy_device(dev) as i32
+ let drv = unsafe { T::from_raw(phydrv) };
+ T::match_phy_device(dev, drv) as i32
}
/// # Safety@@ -574,6 +577,19 @@ pub const fn create_phy_driver<T: Driver>() -> DriverVTable {
/// This trait is used to create a [`DriverVTable`].
#[vtable]
pub trait Driver {
+ /// # Safety
+ ///
+ /// For the duration of `'a`, the pointer must point at a valid
+ /// `phy_driver`, and the caller must be in a context where all
+ /// methods defined on this struct are safe to call.
+ unsafe fn from_raw<'a>(ptr: *const bindings::phy_driver) -> &'a DriverVTable {
+ // CAST: `DriverVTable` is a `repr(transparent)` wrapper around `bindings::phy_driver`.
+ let ptr = ptr.cast::<DriverVTable>();
+ // SAFETY: by the function requirements the pointer is const and is
+ // always valid to access for the duration of `'a`.
+ unsafe { &*ptr }
+ }
If we go the way of supplying a `&DriverVTable` in the
`match_phy_device` function, then this should be a function in the impl
block of `DriverVTable` and not in `Driver`.
Yeah.
quoted
See my reply to Fujita on the previous version, I don't think that we
need to add the `DriverVTable` to the `match_phy_device` function if we
don't provide accessor methods. Currently that isn't needed, so you only
need the hunks above this one. (I'd wait for Fujita's reply though).
Agreed, to make DriverVTable actually useful in match_phy_device(),
further changes would be needed. I think it's sufficient to simply
make the Rust code compile, as shown in the patch below.
I can take care of making sure Rust uses DriverVTable correctly later.
Thanks a lot, I'm sending new revision with the below change as
suggested.
quoted hunk ↗ jump to hunk
diff --git a/rust/kernel/net/phy.rs b/rust/kernel/net/phy.rs
index a59469c785e3..32ea43ece646 100644
--- a/rust/kernel/net/phy.rs
+++ b/rust/kernel/net/phy.rs
@@ -421,6 +421,7 @@ impl<T: Driver> Adapter<T> {
/// `phydev` must be passed by the corresponding callback in `phy_driver`.
unsafe extern "C" fn match_phy_device_callback(
phydev: *mut bindings::phy_device,
+ _phydrv: *const bindings::phy_driver,
) -> crate::ffi::c_int {
// SAFETY: This callback is called only in contexts
// where we hold `phy_device->lock`, so the accessors on
--
Ansuel