Message-ID: <ZLGCz00r9T0LfEbC@boqun-archlinux>
Date:   Fri, 14 Jul 2023 10:15:59 -0700
From:   Boqun Feng <boqun.feng@...il.com>
To:     Asahi Lina <lina@...hilina.net>
Cc:     Miguel Ojeda <ojeda@...nel.org>,
        Alex Gaynor <alex.gaynor@...il.com>,
        Wedson Almeida Filho <wedsonaf@...il.com>,
        Gary Guo <gary@...yguo.net>,
        Björn Roy Baron <bjorn3_gh@...tonmail.com>,
        John Stultz <jstultz@...gle.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Stephen Boyd <sboyd@...nel.org>,
        Josh Stone <jistone@...hat.com>,
        Gaelan Steele <gbs@...ishe.com>,
        Heghedus Razvan <heghedus.razvan@...tonmail.com>,
        linux-kernel@...r.kernel.org, rust-for-linux@...r.kernel.org,
        asahi@...ts.linux.dev
Subject: Re: [PATCH v2] rust: time: New module for timekeeping functions

On Fri, Jul 14, 2023 at 04:55:01PM +0900, Asahi Lina wrote:
> This module is intended to contain functions related to kernel
> timekeeping and time.
> 
> Initially, this implements an abstraction for a time Instant (analogous
> to the Rust std::time::Instant) that represents an opaque instant in
> time. Unlike the std Instant, this is a generic type that is bound to a
> specific clock source, so that only Instants from the same clock source
> can be subtracted/compared.
> 
> Then we implement the relevant clocks available to the kernel:
> KernelTime (CLOCK_MONOTONIC), BootTime (CLOCK_BOOTTIME),
> RealTime (CLOCK_REALTIME), and TaiTime.
> 
> Co-developed-by: Heghedus Razvan <heghedus.razvan@...tonmail.com>
> Signed-off-by: Asahi Lina <lina@...hilina.net>
> ---
> 
> Based on the feedback to v1, now we have proper type checking for kernel
> time. I decided to implement marker traits for monotonic vs. wallclock
> time sources, since it's useful to be able to safely implement different
> semantics conditional on that, but that left me with a name conflict of
> the Monotonic trait with the CLOCK_MONOTONIC / "default ktime" clock
> source. I ended up calling it KernelTime since it's the most fundamental
> kernel timesource, but suggestions welcome!
> 
> Heghedus: I think I need a signoff on this since this is based on the
> playground demo you wrote in the feedback to v1. Can you provide that? I
> can fold it into v3 (if there is one, otherwise Miguel can probably just
> add it when he applies it). Thanks!
> ---
>  rust/bindings/bindings_helper.h |   2 +
>  rust/helpers.c                  |  16 +++++
>  rust/kernel/lib.rs              |   1 +
>  rust/kernel/time.rs             | 150 ++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 169 insertions(+)
> 
> diff --git a/rust/bindings/bindings_helper.h b/rust/bindings/bindings_helper.h
> index 3e601ce2548d..eddfdf887364 100644
> --- a/rust/bindings/bindings_helper.h
> +++ b/rust/bindings/bindings_helper.h
> @@ -8,9 +8,11 @@
>  
>  #include <linux/errname.h>
>  #include <linux/slab.h>
> +#include <linux/ktime.h>
>  #include <linux/refcount.h>
>  #include <linux/wait.h>
>  #include <linux/sched.h>
> +#include <linux/timekeeping.h>
>  
>  /* `bindgen` gets confused at certain things. */
>  const gfp_t BINDINGS_GFP_KERNEL = GFP_KERNEL;
> diff --git a/rust/helpers.c b/rust/helpers.c
> index bb594da56137..eff092302e23 100644
> --- a/rust/helpers.c
> +++ b/rust/helpers.c
> @@ -26,6 +26,7 @@
>  #include <linux/mutex.h>
>  #include <linux/spinlock.h>
>  #include <linux/sched/signal.h>
> +#include <linux/timekeeping.h>
>  #include <linux/wait.h>
>  
>  __noreturn void rust_helper_BUG(void)
> @@ -135,6 +136,21 @@ void rust_helper_put_task_struct(struct task_struct *t)
>  }
>  EXPORT_SYMBOL_GPL(rust_helper_put_task_struct);
>  
> +ktime_t rust_helper_ktime_get_real(void) {
> +	return ktime_get_real();
> +}
> +EXPORT_SYMBOL_GPL(rust_helper_ktime_get_real);
> +
> +ktime_t rust_helper_ktime_get_boottime(void) {
> +	return ktime_get_boottime();
> +}
> +EXPORT_SYMBOL_GPL(rust_helper_ktime_get_boottime);
> +
> +ktime_t rust_helper_ktime_get_clocktai(void) {
> +	return ktime_get_clocktai();
> +}
> +EXPORT_SYMBOL_GPL(rust_helper_ktime_get_clocktai);
> +
>  /*
>   * We use `bindgen`'s `--size_t-is-usize` option to bind the C `size_t` type
>   * as the Rust `usize` type, so we can use it in contexts where Rust
> diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
> index 85b261209977..52c91484c5d8 100644
> --- a/rust/kernel/lib.rs
> +++ b/rust/kernel/lib.rs
> @@ -42,6 +42,7 @@
>  pub mod str;
>  pub mod sync;
>  pub mod task;
> +pub mod time;
>  pub mod types;
>  
>  #[doc(hidden)]
> diff --git a/rust/kernel/time.rs b/rust/kernel/time.rs
> new file mode 100644
> index 000000000000..f3bfeed20145
> --- /dev/null
> +++ b/rust/kernel/time.rs
> @@ -0,0 +1,150 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +//! Timekeeping functions.
> +//!
> +//! C header: [`include/linux/ktime.h`](../../../../include/linux/ktime.h)
> +//! C header: [`include/linux/timekeeping.h`](../../../../include/linux/timekeeping.h)
> +
> +use crate::{bindings, pr_err};
> +use core::marker::PhantomData;
> +use core::time::Duration;
> +
> +/// Represents a clock, that is, a unique time source.
> +pub trait Clock: Sized {}
> +
> +/// A time source that can be queried for the current time.
> +pub trait Now: Clock {

Is there a `Clock` that we cannot get the current time? ;-)

> +    /// Returns the current time for this clock.
> +    fn now() -> Instant<Self>;
> +}
> +
> +/// Marker trait for clock sources that are guaranteed to be monotonic.
> +pub trait Monotonic {}
> +
> +/// Marker trait for clock sources that represent a calendar (wall clock)
> +/// relative to the UNIX epoch.
> +pub trait WallTime {}
> +
> +/// An instant in time associated with a given clock source.
> +#[derive(Debug)]
> +pub struct Instant<T: Clock> {
> +    nanoseconds: i64,
> +    _type: PhantomData<T>,
> +}
> +
> +impl<T: Clock> Clone for Instant<T> {
> +    fn clone(&self) -> Self {
> +        *self
> +    }
> +}
> +
> +impl<T: Clock> Copy for Instant<T> {}
> +
> +impl<T: Clock> Instant<T> {
> +    fn new(nanoseconds: i64) -> Self {
> +        Instant {
> +            nanoseconds,
> +            _type: PhantomData,
> +        }
> +    }
> +
> +    /// Returns the time elapsed since an earlier Instant<t>, or
> +    /// None if the argument is a later Instant.
> +    pub fn since(&self, earlier: Instant<T>) -> Option<Duration> {
> +        if earlier.nanoseconds > self.nanoseconds {
> +            None
> +        } else {
> +            // Casting to u64 and subtracting is guaranteed to give the right
> +            // result for all inputs, as long as the condition we checked above
> +            // holds.
> +            Some(Duration::from_nanos(
> +                self.nanoseconds as u64 - earlier.nanoseconds as u64,
> +            ))
> +        }
> +    }
> +}
> +
> +impl<T: Clock + Now + Monotonic> Instant<T> {
> +    /// Returns the time elapsed since this Instant<T>.
> +    ///
> +    /// This is guaranteed to return a positive result, since
> +    /// it is only implemented for monotonic clocks.
> +    pub fn elapsed(&self) -> Duration {
> +        T::now().since(*self).unwrap_or_else(|| {
> +            pr_err!(
> +                "Monotonic clock {} went backwards!",
> +                core::any::type_name::<T>()
> +            );
> +            Duration::ZERO
> +        })
> +    }
> +}
> +
> +/// Contains the various clock source types available to the kernel.
> +pub mod clock {
> +    use super::*;
> +
> +    /// A clock representing the default kernel time source.
> +    ///
> +    /// This is `CLOCK_MONOTONIC` (though it is not the only
> +    /// monotonic clock) and also the default clock used by
> +    /// `ktime_get()` in the C API.
> +    ///
> +    /// This is like `BootTime`, but does not include time
> +    /// spent sleeping.
> +

Could you add one example for each `*Time` type? Not only they can show
people how to use the API, but also we can use them to generate unit
tests:

	https://lore.kernel.org/rust-for-linux/20230614180837.630180-1-ojeda@kernel.orge 

If you want to test your examples, you can either apply that patchset or
use the rust-dev branch:

	https://github.com/Rust-for-Linux/linux.git rust-dev

I use the following command to run tests:

	./tools/testing/kunit/kunit.py run --make_options LLVM=1 --arch x86_64 --kconfig_add CONFIG_RUST=y

Thanks!

Regards,
Boqun

> +    pub struct KernelTime;
> +
> +    impl Clock for KernelTime {}
> +    impl Monotonic for KernelTime {}
> +    impl Now for KernelTime {
> +        fn now() -> Instant<Self> {
> +            Instant::<Self>::new(unsafe { bindings::ktime_get() })
> +        }
> +    }
> +
> +    /// A clock representing the time elapsed since boot.
> +    ///
> +    /// This is `CLOCK_MONOTONIC` (though it is not the only
> +    /// monotonic clock) and also the default clock used by
> +    /// `ktime_get()` in the C API.
> +    ///
> +    /// This is like `KernelTime`, but does include time
> +    /// spent sleeping.
> +    pub struct BootTime;
> +
> +    impl Clock for BootTime {}
> +    impl Monotonic for BootTime {}
> +    impl Now for BootTime {
> +        fn now() -> Instant<Self> {
> +            Instant::<Self>::new(unsafe { bindings::ktime_get_boottime() })
> +        }
> +    }
> +
> +    /// A clock representing TAI time.
> +    ///
> +    /// This clock is not monotonic and can be changed from userspace.
> +    /// However, it is not affected by leap seconds.
> +    pub struct TaiTime;
> +
> +    impl Clock for TaiTime {}
> +    impl WallTime for TaiTime {}
> +    impl Now for TaiTime {
> +        fn now() -> Instant<Self> {
> +            Instant::<Self>::new(unsafe { bindings::ktime_get_clocktai() })
> +        }
> +    }
> +
> +    /// A clock representing wall clock time.
> +    ///
> +    /// This clock is not monotonic and can be changed from userspace.
> +    pub struct RealTime;
> +
> +    impl Clock for RealTime {}
> +    impl WallTime for RealTime {}
> +    impl Now for RealTime {
> +        fn now() -> Instant<Self> {
> +            Instant::<Self>::new(unsafe { bindings::ktime_get_real() })
> +        }
> +    }
> +}
> 
> ---
> base-commit: 06c2afb862f9da8dc5efa4b6076a0e48c3fbaaa5
> change-id: 20230714-rust-time-10dd9ed25333
> 
> Thank you,
> ~~ Lina
> 

Powered by blists - more mailing lists