Add read_poll_timeout function which poll periodically until a
condition is met or a timeout is reached.

The C's read_poll_timeout (include/linux/iopoll.h) is a complicated
macro and a simple wrapper for Rust doesn't work. So this implements
the same functionality in Rust.

The C version uses usleep_range() while the Rust version uses
fsleep(), which uses the best sleep method so it works with spans that
usleep_range() doesn't work nicely with.

The sleep_before_read argument isn't supported since there is no user
for now. It's rarely used in the C version.

Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Reviewed-by: Fiona Behrens <me@kloenk.dev>
Tested-by: Daniel Almeida <daniel.almeida@collabora.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
---
 rust/kernel/io.rs      |  1 +
 rust/kernel/io/poll.rs | 96 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 97 insertions(+)
 create mode 100644 rust/kernel/io/poll.rs

diff --git a/rust/kernel/io.rs b/rust/kernel/io.rs
index 03b467722b86..ee182b0b5452 100644
--- a/rust/kernel/io.rs
+++ b/rust/kernel/io.rs
@@ -8,6 +8,7 @@
 use crate::{bindings, build_assert, ffi::c_void};
 
 pub mod mem;
+pub mod poll;
 pub mod resource;
 
 pub use resource::Resource;
diff --git a/rust/kernel/io/poll.rs b/rust/kernel/io/poll.rs
new file mode 100644
index 000000000000..e6325725d5a3
--- /dev/null
+++ b/rust/kernel/io/poll.rs
@@ -0,0 +1,96 @@
+// SPDX-License-Identifier: GPL-2.0
+
+//! IO polling.
+//!
+//! C header: [`include/linux/iopoll.h`](srctree/include/linux/iopoll.h).
+
+use crate::{
+    error::{code::*, Result},
+    processor::cpu_relax,
+    task::might_sleep,
+    time::{delay::fsleep, Delta, Instant, Monotonic},
+};
+
+/// Polls periodically until a condition is met or a timeout is reached.
+///
+/// The function repeatedly executes the given operation `op` closure and
+/// checks its result using the condition closure `cond`.
+///
+/// If `cond` returns `true`, the function returns successfully with the result of `op`.
+/// Otherwise, it waits for a duration specified by `sleep_delta`
+/// before executing `op` again.
+///
+/// This process continues until either `cond` returns `true` or the timeout,
+/// specified by `timeout_delta`, is reached. If `timeout_delta` is `None`,
+/// polling continues indefinitely until `cond` evaluates to `true` or an error occurs.
+///
+/// This function can only be used in a nonatomic context.
+///
+/// # Examples
+///
+/// ```no_run
+/// use kernel::io::{Io, poll::read_poll_timeout};
+/// use kernel::time::Delta;
+///
+/// const HW_READY: u16 = 0x01;
+///
+/// fn wait_for_hardware<const SIZE: usize>(io: &Io<SIZE>) -> Result<()> {
+///     match read_poll_timeout(
+///         // The `op` closure reads the value of a specific status register.
+///         || io.try_read16(0x1000),
+///         // The `cond` closure takes a reference to the value returned by `op`
+///         // and checks whether the hardware is ready.
+///         |val: &u16| *val == HW_READY,
+///         Delta::from_millis(50),
+///         Delta::from_secs(3),
+///     ) {
+///         Ok(_) => {
+///             // The hardware is ready. The returned value of the `op` closure
+///             // isn't used.
+///             Ok(())
+///         }
+///         Err(e) => Err(e),
+///     }
+/// }
+/// ```
+#[track_caller]
+pub fn read_poll_timeout<Op, Cond, T>(
+    mut op: Op,
+    mut cond: Cond,
+    sleep_delta: Delta,
+    timeout_delta: Delta,
+) -> Result<T>
+where
+    Op: FnMut() -> Result<T>,
+    Cond: FnMut(&T) -> bool,
+{
+    let start: Instant<Monotonic> = Instant::now();
+
+    // Unlike the C version, we always call `might_sleep()` unconditionally,
+    // as conditional calls are error-prone. We clearly separate
+    // `read_poll_timeout()` and `read_poll_timeout_atomic()` to aid
+    // tools like klint.
+    might_sleep();
+
+    loop {
+        let val = op()?;
+        if cond(&val) {
+            // Unlike the C version, we immediately return.
+            // We know the condition is met so we don't need to check again.
+            return Ok(val);
+        }
+
+        if start.elapsed() > timeout_delta {
+            // Unlike the C version, we immediately return.
+            // We have just called `op()` so we don't need to call it again.
+            return Err(ETIMEDOUT);
+        }
+
+        if !sleep_delta.is_zero() {
+            fsleep(sleep_delta);
+        }
+
+        // fsleep() could be busy-wait loop so we always call cpu_relax().
+        cpu_relax();
+    }
+}
-- 
2.43.0

On Sun, Aug 17, 2025 at 01:47:23PM +0900, FUJITA Tomonori wrote:
> Add read_poll_timeout function which poll periodically until a
> condition is met or a timeout is reached.
> 
> The C's read_poll_timeout (include/linux/iopoll.h) is a complicated
> macro and a simple wrapper for Rust doesn't work. So this implements
> the same functionality in Rust.
> 
> The C version uses usleep_range() while the Rust version uses
> fsleep(), which uses the best sleep method so it works with spans that
> usleep_range() doesn't work nicely with.
> 
> The sleep_before_read argument isn't supported since there is no user
> for now. It's rarely used in the C version.
> 
> Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
> Reviewed-by: Fiona Behrens <me@kloenk.dev>
> Tested-by: Daniel Almeida <daniel.almeida@collabora.com>
> Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Hi Fujita,

> On 17 Aug 2025, at 01:47, FUJITA Tomonori <fujita.tomonori@gmail.com> wrote:
> 
> Add read_poll_timeout function which poll periodically until a
> condition is met or a timeout is reached.
> 
> The C's read_poll_timeout (include/linux/iopoll.h) is a complicated
> macro and a simple wrapper for Rust doesn't work. So this implements
> the same functionality in Rust.
> 
> The C version uses usleep_range() while the Rust version uses
> fsleep(), which uses the best sleep method so it works with spans that
> usleep_range() doesn't work nicely with.
> 
> The sleep_before_read argument isn't supported since there is no user
> for now. It's rarely used in the C version.
> 
> Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
> Reviewed-by: Fiona Behrens <me@kloenk.dev>
> Tested-by: Daniel Almeida <daniel.almeida@collabora.com>
> Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
> ---
> rust/kernel/io.rs      |  1 +
> rust/kernel/io/poll.rs | 96 ++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 97 insertions(+)
> create mode 100644 rust/kernel/io/poll.rs
> 
> diff --git a/rust/kernel/io.rs b/rust/kernel/io.rs
> index 03b467722b86..ee182b0b5452 100644
> --- a/rust/kernel/io.rs
> +++ b/rust/kernel/io.rs
> @@ -8,6 +8,7 @@
> use crate::{bindings, build_assert, ffi::c_void};
> 
> pub mod mem;
> +pub mod poll;
> pub mod resource;
> 
> pub use resource::Resource;
> diff --git a/rust/kernel/io/poll.rs b/rust/kernel/io/poll.rs
> new file mode 100644
> index 000000000000..e6325725d5a3
> --- /dev/null
> +++ b/rust/kernel/io/poll.rs
> @@ -0,0 +1,96 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +//! IO polling.
> +//!
> +//! C header: [`include/linux/iopoll.h`](srctree/include/linux/iopoll.h).
> +
> +use crate::{
> +    error::{code::*, Result},
> +    processor::cpu_relax,
> +    task::might_sleep,
> +    time::{delay::fsleep, Delta, Instant, Monotonic},
> +};
> +
> +/// Polls periodically until a condition is met or a timeout is reached.
> +///
> +/// The function repeatedly executes the given operation `op` closure and
> +/// checks its result using the condition closure `cond`.
> +///
> +/// If `cond` returns `true`, the function returns successfully with the result of `op`.
> +/// Otherwise, it waits for a duration specified by `sleep_delta`
> +/// before executing `op` again.
> +///
> +/// This process continues until either `cond` returns `true` or the timeout,
> +/// specified by `timeout_delta`, is reached. If `timeout_delta` is `None`,
> +/// polling continues indefinitely until `cond` evaluates to `true` or an error occurs.
> +///
> +/// This function can only be used in a nonatomic context.
> +///
> +/// # Examples
> +///
> +/// ```no_run
> +/// use kernel::io::{Io, poll::read_poll_timeout};
> +/// use kernel::time::Delta;
> +///
> +/// const HW_READY: u16 = 0x01;
> +///
> +/// fn wait_for_hardware<const SIZE: usize>(io: &Io<SIZE>) -> Result<()> {
> +///     match read_poll_timeout(
> +///         // The `op` closure reads the value of a specific status register.
> +///         || io.try_read16(0x1000),
> +///         // The `cond` closure takes a reference to the value returned by `op`
> +///         // and checks whether the hardware is ready.
> +///         |val: &u16| *val == HW_READY,
> +///         Delta::from_millis(50),
> +///         Delta::from_secs(3),
> +///     ) {
> +///         Ok(_) => {
> +///             // The hardware is ready. The returned value of the `op` closure
> +///             // isn't used.
> +///             Ok(())
> +///         }
> +///         Err(e) => Err(e),
> +///     }
> +/// }
> +/// ```
> +#[track_caller]
> +pub fn read_poll_timeout<Op, Cond, T>(
> +    mut op: Op,
> +    mut cond: Cond,
> +    sleep_delta: Delta,
> +    timeout_delta: Delta,
> +) -> Result<T>
> +where
> +    Op: FnMut() -> Result<T>,
> +    Cond: FnMut(&T) -> bool,
> +{
> +    let start: Instant<Monotonic> = Instant::now();
> +
> +    // Unlike the C version, we always call `might_sleep()` unconditionally,
> +    // as conditional calls are error-prone. We clearly separate
> +    // `read_poll_timeout()` and `read_poll_timeout_atomic()` to aid
> +    // tools like klint.
> +    might_sleep();
> +
> +    loop {
> +        let val = op()?;
> +        if cond(&val) {
> +            // Unlike the C version, we immediately return.
> +            // We know the condition is met so we don't need to check again.
> +            return Ok(val);
> +        }
> +
> +        if start.elapsed() > timeout_delta {
> +            // Unlike the C version, we immediately return.
> +            // We have just called `op()` so we don't need to call it again.
> +            return Err(ETIMEDOUT);
> +        }
> +
> +        if !sleep_delta.is_zero() {
> +            fsleep(sleep_delta);
> +        }
> +
> +        // fsleep() could be busy-wait loop so we always call cpu_relax().
> +        cpu_relax();
> +    }
> +}
> -- 
> 2.43.0
> 
> 

Thanks for working on this. Definitely going to be needed by a lot of drivers.

Reviewed-by: Daniel Almeida <daniel.almeida@collabora.com>

How is the atomic version going to look like? The same, except for
might_sleep() and without the sleep_delta argument?

On Tue, 19 Aug 2025 15:30:51 -0300
Daniel Almeida <daniel.almeida@collabora.com> wrote:

> Thanks for working on this. Definitely going to be needed by a lot of drivers.
> 
> Reviewed-by: Daniel Almeida <daniel.almeida@collabora.com>

Thanks!

> How is the atomic version going to look like? The same, except for
> might_sleep() and without the sleep_delta argument?

If we follow the C implementation, it will be different; C's
read_poll_atomic doesn't use ktime to calculate a timeout.

It would look like the following. I think that the read_poll_timeout
patchset is almost complete so I'll send the
read_poll_timeout_atomic() patchset shortly.

diff --git a/rust/kernel/io/poll.rs b/rust/kernel/io/poll.rs
index e6325725d5a3..dc4f1ecdf31f 100644
--- a/rust/kernel/io/poll.rs
+++ b/rust/kernel/io/poll.rs
@@ -8,7 +8,10 @@
     error::{code::*, Result},
     processor::cpu_relax,
     task::might_sleep,
-    time::{delay::fsleep, Delta, Instant, Monotonic},
+    time::{
+        delay::{fsleep, udelay},
+        Delta, Instant, Monotonic,
+    },
 };
 
 /// Polls periodically until a condition is met or a timeout is reached.
@@ -94,3 +97,86 @@ pub fn read_poll_timeout<Op, Cond, T>(
         cpu_relax();
     }
 }
+
+/// Polls periodically until a condition is met or a timeout is reached.
+///
+/// The function repeatedly executes the given operation `op` closure and
+/// checks its result using the condition closure `cond`.
+///
+/// If `cond` returns `true`, the function returns successfully with the result of `op`.
+/// Otherwise, it performs a busy wait for a duration specified by `delay_delta`
+/// before executing `op` again.
+///
+/// This process continues until either `cond` returns `true` or the timeout,
+/// specified by `timeout_delta`, is reached. If `timeout_delta` is `None`,
+/// polling continues indefinitely until `cond` evaluates to `true` or an error occurs.
+///
+/// # Examples
+///
+/// ```no_run
+/// use kernel::io::{Io, poll::read_poll_timeout_atomic};
+/// use kernel::time::Delta;
+///
+/// const HW_READY: u16 = 0x01;
+///
+/// fn wait_for_hardware<const SIZE: usize>(io: &Io<SIZE>) -> Result<()> {
+///     match read_poll_timeout_atomic(
+///         // The `op` closure reads the value of a specific status register.
+///         || io.try_read16(0x1000),
+///         // The `cond` closure takes a reference to the value returned by `op`
+///         // and checks whether the hardware is ready.
+///         |val: &u16| *val == HW_READY,
+///         Delta::from_micros(50),
+///         Delta::from_micros(300),
+///     ) {
+///         Ok(_) => {
+///             // The hardware is ready. The returned value of the `op` closure
+///             // isn't used.
+///             Ok(())
+///         }
+///         Err(e) => Err(e),
+///     }
+/// }
+/// ```
+pub fn read_poll_timeout_atomic<Op, Cond, T>(
+    mut op: Op,
+    mut cond: Cond,
+    delay_delta: Delta,
+    timeout_delta: Delta,
+) -> Result<T>
+where
+    Op: FnMut() -> Result<T>,
+    Cond: FnMut(&T) -> bool,
+{
+    let mut left_ns = timeout_delta.as_nanos();
+    let delay_ns = delay_delta.as_nanos();
+
+    let timeout_is_zero = timeout_delta.is_zero();
+
+    loop {
+        let val = op()?;
+        if cond(&val) {
+            // Unlike the C version, we immediately return.
+            // We know the condition is met so we don't need to check again.
+            return Ok(val);
+        }
+
+        if !timeout_is_zero && left_ns < 0 {
+            // Unlike the C version, we immediately return.
+            // We have just called `op()` so we don't need to call it again.
+            return Err(ETIMEDOUT);
+        }
+
+        if !delay_delta.is_zero() {
+            udelay(delay_delta);
+            if !timeout_is_zero {
+                left_ns -= delay_ns;
+            }
+        }
+
+        cpu_relax();
+        if !timeout_is_zero {
+            left_ns -= 1;
+        }
+    }
+}
-- 
2.43.0

On Sun Aug 17, 2025 at 1:47 PM JST, FUJITA Tomonori wrote:
> Add read_poll_timeout function which poll periodically until a
> condition is met or a timeout is reached.
>
> The C's read_poll_timeout (include/linux/iopoll.h) is a complicated
> macro and a simple wrapper for Rust doesn't work. So this implements
> the same functionality in Rust.
>
> The C version uses usleep_range() while the Rust version uses
> fsleep(), which uses the best sleep method so it works with spans that
> usleep_range() doesn't work nicely with.
>
> The sleep_before_read argument isn't supported since there is no user
> for now. It's rarely used in the C version.
>
> Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
> Reviewed-by: Fiona Behrens <me@kloenk.dev>
> Tested-by: Daniel Almeida <daniel.almeida@collabora.com>
> Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>

Tested this with nova-core, and it seems to work fine!

Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Tested-by: Alexandre Courbot <acourbot@nvidia.com>

Just one last comment about the documentation below.

<snip>
> +/// Polls periodically until a condition is met or a timeout is reached.
> +///
> +/// The function repeatedly executes the given operation `op` closure and
> +/// checks its result using the condition closure `cond`.
> +///
> +/// If `cond` returns `true`, the function returns successfully with the result of `op`.
> +/// Otherwise, it waits for a duration specified by `sleep_delta`
> +/// before executing `op` again.
> +///
> +/// This process continues until either `cond` returns `true` or the timeout,
> +/// specified by `timeout_delta`, is reached. If `timeout_delta` is `None`,

For precision: "This process continues until either `op` returns and
error, `cond` returns `true`, or the timeout specified by
`timeout_delta` is reached."

> +/// polling continues indefinitely until `cond` evaluates to `true` or an error occurs.
> +///
> +/// This function can only be used in a nonatomic context.

Here I'd add an errors section:

# Errors

If `op` returns an error, then that error is returned directly.

If the timeout specified by `timeout_delta` is reached, then
`Err(ETIMEDOUT)` is returned.

> +///
> +/// # Examples
> +///
> +/// ```no_run
> +/// use kernel::io::{Io, poll::read_poll_timeout};
> +/// use kernel::time::Delta;
> +///
> +/// const HW_READY: u16 = 0x01;
> +///
> +/// fn wait_for_hardware<const SIZE: usize>(io: &Io<SIZE>) -> Result<()> {
> +///     match read_poll_timeout(
> +///         // The `op` closure reads the value of a specific status register.
> +///         || io.try_read16(0x1000),
> +///         // The `cond` closure takes a reference to the value returned by `op`
> +///         // and checks whether the hardware is ready.
> +///         |val: &u16| *val == HW_READY,
> +///         Delta::from_millis(50),
> +///         Delta::from_secs(3),
> +///     ) {
> +///         Ok(_) => {
> +///             // The hardware is ready. The returned value of the `op` closure
> +///             // isn't used.
> +///             Ok(())
> +///         }
> +///         Err(e) => Err(e),
> +///     }
> +/// }
> +/// ```
> +#[track_caller]
> +pub fn read_poll_timeout<Op, Cond, T>(
> +    mut op: Op,
> +    mut cond: Cond,
> +    sleep_delta: Delta,
> +    timeout_delta: Delta,
> +) -> Result<T>
> +where
> +    Op: FnMut() -> Result<T>,
> +    Cond: FnMut(&T) -> bool,
> +{
> +    let start: Instant<Monotonic> = Instant::now();
> +
> +    // Unlike the C version, we always call `might_sleep()` unconditionally,
> +    // as conditional calls are error-prone. We clearly separate
> +    // `read_poll_timeout()` and `read_poll_timeout_atomic()` to aid
> +    // tools like klint.
> +    might_sleep();
> +
> +    loop {
> +        let val = op()?;
> +        if cond(&val) {
> +            // Unlike the C version, we immediately return.
> +            // We know the condition is met so we don't need to check again.

nit: this comment looks superfluous to me, this is a different
implementation from the C version anyway.

> +            return Ok(val);
> +        }
> +
> +        if start.elapsed() > timeout_delta {
> +            // Unlike the C version, we immediately return.
> +            // We have just called `op()` so we don't need to call it again.

Same here.

On Tue, 19 Aug 2025 09:49:38 +0900
"Alexandre Courbot" <acourbot@nvidia.com> wrote:

> On Sun Aug 17, 2025 at 1:47 PM JST, FUJITA Tomonori wrote:
>> Add read_poll_timeout function which poll periodically until a
>> condition is met or a timeout is reached.
>>
>> The C's read_poll_timeout (include/linux/iopoll.h) is a complicated
>> macro and a simple wrapper for Rust doesn't work. So this implements
>> the same functionality in Rust.
>>
>> The C version uses usleep_range() while the Rust version uses
>> fsleep(), which uses the best sleep method so it works with spans that
>> usleep_range() doesn't work nicely with.
>>
>> The sleep_before_read argument isn't supported since there is no user
>> for now. It's rarely used in the C version.
>>
>> Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
>> Reviewed-by: Fiona Behrens <me@kloenk.dev>
>> Tested-by: Daniel Almeida <daniel.almeida@collabora.com>
>> Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
> 
> Tested this with nova-core, and it seems to work fine!
> 
> Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
> Tested-by: Alexandre Courbot <acourbot@nvidia.com>

Thanks!

> Just one last comment about the documentation below.
> 
> <snip>
>> +/// Polls periodically until a condition is met or a timeout is reached.
>> +///
>> +/// The function repeatedly executes the given operation `op` closure and
>> +/// checks its result using the condition closure `cond`.
>> +///
>> +/// If `cond` returns `true`, the function returns successfully with the result of `op`.
>> +/// Otherwise, it waits for a duration specified by `sleep_delta`
>> +/// before executing `op` again.
>> +///
>> +/// This process continues until either `cond` returns `true` or the timeout,
>> +/// specified by `timeout_delta`, is reached. If `timeout_delta` is `None`,
> 
> For precision: "This process continues until either `op` returns and
> error, `cond` returns `true`, or the timeout specified by
> `timeout_delta` is reached."

Indeed, `op` should be mentioned here.

I think it should be "op returns an error", not "op returns and error",
right?

There are three alternatives, so I’d remove "either". The sentence
would be:

This process continues until either `op` returns an error, `cond`
returns `true`, or the timeout specified by `timeout_delta` is
reached.

I just realized that "If `timeout_delta` is `None`," comment is outdated. I'll update.


>> +/// polling continues indefinitely until `cond` evaluates to `true` or an error occurs.
>> +///
>> +/// This function can only be used in a nonatomic context.
> 
> Here I'd add an errors section:
> 
> # Errors
> 
> If `op` returns an error, then that error is returned directly.
> 
> If the timeout specified by `timeout_delta` is reached, then
> `Err(ETIMEDOUT)` is returned.

Thanks, looks nice. I'll add it.


>> +    loop {
>> +        let val = op()?;
>> +        if cond(&val) {
>> +            // Unlike the C version, we immediately return.
>> +            // We know the condition is met so we don't need to check again.
> 
> nit: this comment looks superfluous to me, this is a different
> implementation from the C version anyway.

A previous review mentioned that, since this function shares the same
name as the C version, it's good to make the different behavior
explicit. It may be a bit redundant, but I'd prefer to keep it.