Re: [PATCH v5 6/7] rust: hrtimer: Add HrTimer::raw_cb_time()

[Andreas Hindborg <a.hindborg@kernel.org>]

"Lyude Paul" <lyude@redhat.com> writes:

> This is a simple private unsafe wrapper for retrieving the current time
> according to the hrtimer_clock_base struct for a given timer. This will be
> used for implementing functions such as forward_now(), which rely on
> retrieving the current time from the hrtimer's clock base.
>
> Signed-off-by: Lyude Paul <lyude@redhat.com>
>
> ---
> V2:
> - Convert safety comment to invariant comment in from_raw()
> - Add raw_clock_base() and implement clock_base() on HrTimer<T> as well
>
> V4:
> - Drop HrTimerClockBase entirely, reword commit as this is now about adding
>   raw_cb_time()
>
> Signed-off-by: Lyude Paul <lyude@redhat.com>
> ---
>  rust/kernel/time.rs         |  1 -
>  rust/kernel/time/hrtimer.rs | 27 +++++++++++++++++++++++++++
>  2 files changed, 27 insertions(+), 1 deletion(-)
>
> diff --git a/rust/kernel/time.rs b/rust/kernel/time.rs
> index eed77297d58a6..27ee78070d72e 100644
> --- a/rust/kernel/time.rs
> +++ b/rust/kernel/time.rs
> @@ -209,7 +209,6 @@ pub(crate) fn into_nanos(self) -> i64 {
>      /// # Safety
>      ///
>      /// The caller promises that `nanos` is in the range from 0 to `KTIME_MAX`.
> -    #[expect(unused)]
>      #[inline]
>      pub(crate) unsafe fn from_nanos(nanos: i64) -> Self {
>          debug_assert!(nanos >= 0);
> diff --git a/rust/kernel/time/hrtimer.rs b/rust/kernel/time/hrtimer.rs
> index 4a8416fbd187d..79d86e1099a1e 100644
> --- a/rust/kernel/time/hrtimer.rs
> +++ b/rust/kernel/time/hrtimer.rs
> @@ -193,6 +193,33 @@ unsafe fn raw_forward(self_ptr: *mut Self, now: HrTimerInstant<T>, interval: Del
>          }
>      }
>
> +    /// Retrieve the current time according to the `struct hrtimer_clock_base` for `self_ptr`.
> +    ///
> +    /// # Safety
> +    ///
> +    /// - `self_ptr` must point to a valid `Self`.
> +    /// - The caller must ensure that the `hrtimer_clock_base` cannot possibly change in the context
> +    ///   this function is being called in. This means either exclusive access to `self_ptr` is
> +    ///   required, or we must be from within the timer callback context of `self_ptr`.
> +    #[expect(unused)]
> +    unsafe fn raw_cb_time(self_ptr: *const Self) -> HrTimerInstant<T>

Can we call it `raw_clock_base_time`?

> +    where
> +        T: HasHrTimer<T>,
> +    {
> +        // SAFETY: We're guaranteed `self_ptr` points to a valid `Self` by our safety contract.
> +        let clock_base = unsafe { (*Self::raw_get(self_ptr)).base };
> +
> +        // SAFETY: The C API guarantees that `get_time` is initialized to a valid function pointer
> +        // for as long as we expose hrtimers to users.
> +        let get_time_fn = unsafe { (*clock_base).get_time.unwrap_unchecked() };
> +
> +        // SAFETY:
> +        // - get_time_fn() returns a ktime_t, so we're guaranteed its return value is between `0`
> +        //   and `KTIME_MAX`.
> +        // - get_time_fn() itself has no special requirements.
> +        unsafe { Instant::from_nanos(get_time_fn()) }
> +    }
> +

How does this differ from Instant<C>::now()? Could we do this statically
by going through `<<T as HasHrTimer<T>>::TimerMode as HrTimerMode>::Clock`?


Best regards,
Andreas Hindborg