[PATCH v4] rust: serdev: Mitigate race conditions

[PATCH v4] rust: serdev: Mitigate race conditions

[Markus Probst]

There are currently 2 race conditions:
- in probe if `Driver::probe` returns Err
- in unbind
. In those cases the driver data will be set to NULL before the serdev
device was closed. If data is received while the driver data is dropped,
the `receive_buf_callback` might try to access the `active` mutex on a
null pointer. The race conditions can only occur if `Driver::receive` is
implemented.

The issue cannot be cleanly fixed without rewriting some logic, which
might introduce new regressions.

Temporarily disable the use of `Driver::receive`, until fixed in the
next release.

Fixes: 99f59aa82341 ("rust: add basic serial device bus abstractions")
Reported-by: Sashiko Bot <sashiko-bot@kernel.org>
Closes: https://lore.kernel.org/linux-serial/20260905000836.C8FC91F00A3D@smtp.kernel.org/
Closes: https://lore.kernel.org/linux-serial/20260903222159.70A911F000E9@smtp.kernel.org/
Signed-off-by: Markus Probst <markus.probst@posteo.de>
---
I will submit a patch (for the next merge cycle) soon, which will
address this issue and make the probe and unbind code less convoluted.
---
Changes in v4:
- remove comments
- Link to v3: https://patch.msgid.link/20260905-rust_serdev_fix-v3-1-b86056c3f7a4@posteo.de

Changes in v3:
- mitigate it
- Link to v2: https://patch.msgid.link/20260905-rust_serdev_fix-v2-0-35dfcd06ef2e@posteo.de

Changes in v2:
- also fix race condition on unbind
- Link to v1: https://patch.msgid.link/20260905-rust_serdev_fix-v1-1-2ea92b154a6b@posteo.de
---
 rust/kernel/serdev.rs              |  6 +++---
 samples/rust/rust_driver_serdev.rs | 13 +------------
 2 files changed, 4 insertions(+), 15 deletions(-)

diff --git a/rust/kernel/serdev.rs b/rust/kernel/serdev.rs
index 17ca504b7f8d..f4c0f43b5273 100644
--- a/rust/kernel/serdev.rs
+++ b/rust/kernel/serdev.rs
@@ -148,9 +148,8 @@ fn drop(self: Pin<&mut Self>) {
 
 impl<T: Driver> Adapter<T> {
     const OPS: &'static bindings::serdev_device_ops = &bindings::serdev_device_ops {
-        receive_buf: if T::HAS_RECEIVE {
-            Some(Self::receive_buf_callback)
-        } else {
+        receive_buf: {
+            const_assert!(!T::HAS_RECEIVE);
             None
         },
         write_wakeup: Some(bindings::serdev_device_write_wakeup),
@@ -233,6 +232,7 @@ extern "C" fn remove_callback(sdev: *mut bindings::serdev_device) {
         T::unbind(sdev, data_pinned);
     }
 
+    #[expect(dead_code)]
     extern "C" fn receive_buf_callback(
         sdev: *mut bindings::serdev_device,
         buf: *const u8,
diff --git a/samples/rust/rust_driver_serdev.rs b/samples/rust/rust_driver_serdev.rs
index 51b4898cd855..3fd3eef15371 100644
--- a/samples/rust/rust_driver_serdev.rs
+++ b/samples/rust/rust_driver_serdev.rs
@@ -4,10 +4,7 @@
 
 use kernel::{
     acpi,
-    device::{
-        Bound,
-        Core, //
-    },
+    device::Core,
     of,
     prelude::*,
     serdev,
@@ -60,14 +57,6 @@ fn probe<'bound>(
 
         Ok(Self { sdev: sdev.into() })
     }
-
-    fn receive<'bound>(
-        sdev: &'bound serdev::Device<Bound>,
-        _this: Pin<&Self>,
-        data: &[u8],
-    ) -> usize {
-        sdev.write(data).unwrap_or_default() as usize
-    }
 }
 
 impl Drop for SampleDriver {

---
base-commit: e5e04726cdd043e309677071ab1b65a4b18f422b
change-id: 20260904-rust_serdev_fix-be3ff9c8a5e8

Re: [PATCH v4] rust: serdev: Mitigate race conditions

[sashiko-bot]

> There are currently 2 race conditions:
> - in probe if `Driver::probe` returns Err
> - in unbind
> . In those cases the driver data will be set to NULL before the serdev
> device was closed. If data is received while the driver data is dropped,
> the `receive_buf_callback` might try to access the `active` mutex on a
> null pointer. The race conditions can only occur if `Driver::receive` is
> implemented.
> 
> The issue cannot be cleanly fixed without rewriting some logic, which
> might introduce new regressions.
> 
> Temporarily disable the use of `Driver::receive`, until fixed in the
> next release.
> 
> Fixes: 99f59aa82341 ("rust: add basic serial device bus abstractions")
> Reported-by: Sashiko Bot <sashiko-bot@kernel.org>
> Closes: https://lore.kernel.org/linux-serial/20260905000836.C8FC91F00A3D@smtp.kernel.org/
> Closes: https://lore.kernel.org/linux-serial/20260903222159.70A911F000E9@smtp.kernel.org/
> Signed-off-by: Markus Probst <markus.probst@posteo.de>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260905-rust_serdev_fix-v4-1-31e1bbcc0c00@posteo.de?part=1

Re: [PATCH v4] rust: serdev: Mitigate race conditions

[Gary Guo]

On Sat Sep 5, 2026 at 7:47 PM BST, Markus Probst wrote:
> There are currently 2 race conditions:
> - in probe if `Driver::probe` returns Err
> - in unbind
> . In those cases the driver data will be set to NULL before the serdev
> device was closed. If data is received while the driver data is dropped,
> the `receive_buf_callback` might try to access the `active` mutex on a
> null pointer. The race conditions can only occur if `Driver::receive` is
> implemented.
>
> The issue cannot be cleanly fixed without rewriting some logic, which
> might introduce new regressions.
>
> Temporarily disable the use of `Driver::receive`, until fixed in the
> next release.

Isn't this basically undoing all the usefulness of having the serdev abstraction
in the first place?

As this is for unbound only, I don't think we need to rush to fix this for rc
anyway. Many C drivers don't get unbind correct (and our misc device). There's
no in-tree user that will hit this API anyway.

I think we'd better just find a proper fix for the next cycle and leave the
current code as is.

Best,
Gary

>
> Fixes: 99f59aa82341 ("rust: add basic serial device bus abstractions")
> Reported-by: Sashiko Bot <sashiko-bot@kernel.org>
> Closes: https://lore.kernel.org/linux-serial/20260905000836.C8FC91F00A3D@smtp.kernel.org/
> Closes: https://lore.kernel.org/linux-serial/20260903222159.70A911F000E9@smtp.kernel.org/
> Signed-off-by: Markus Probst <markus.probst@posteo.de>
> ---
> I will submit a patch (for the next merge cycle) soon, which will
> address this issue and make the probe and unbind code less convoluted.
> ---
> Changes in v4:
> - remove comments
> - Link to v3: https://patch.msgid.link/20260905-rust_serdev_fix-v3-1-b86056c3f7a4@posteo.de
>
> Changes in v3:
> - mitigate it
> - Link to v2: https://patch.msgid.link/20260905-rust_serdev_fix-v2-0-35dfcd06ef2e@posteo.de
>
> Changes in v2:
> - also fix race condition on unbind
> - Link to v1: https://patch.msgid.link/20260905-rust_serdev_fix-v1-1-2ea92b154a6b@posteo.de
> ---
>  rust/kernel/serdev.rs              |  6 +++---
>  samples/rust/rust_driver_serdev.rs | 13 +------------
>  2 files changed, 4 insertions(+), 15 deletions(-)
>
> diff --git a/rust/kernel/serdev.rs b/rust/kernel/serdev.rs
> index 17ca504b7f8d..f4c0f43b5273 100644
> --- a/rust/kernel/serdev.rs
> +++ b/rust/kernel/serdev.rs
> @@ -148,9 +148,8 @@ fn drop(self: Pin<&mut Self>) {
>  
>  impl<T: Driver> Adapter<T> {
>      const OPS: &'static bindings::serdev_device_ops = &bindings::serdev_device_ops {
> -        receive_buf: if T::HAS_RECEIVE {
> -            Some(Self::receive_buf_callback)
> -        } else {
> +        receive_buf: {
> +            const_assert!(!T::HAS_RECEIVE);
>              None
>          },
>          write_wakeup: Some(bindings::serdev_device_write_wakeup),
> @@ -233,6 +232,7 @@ extern "C" fn remove_callback(sdev: *mut bindings::serdev_device) {
>          T::unbind(sdev, data_pinned);
>      }
>  
> +    #[expect(dead_code)]
>      extern "C" fn receive_buf_callback(
>          sdev: *mut bindings::serdev_device,
>          buf: *const u8,
> diff --git a/samples/rust/rust_driver_serdev.rs b/samples/rust/rust_driver_serdev.rs
> index 51b4898cd855..3fd3eef15371 100644
> --- a/samples/rust/rust_driver_serdev.rs
> +++ b/samples/rust/rust_driver_serdev.rs
> @@ -4,10 +4,7 @@
>  
>  use kernel::{
>      acpi,
> -    device::{
> -        Bound,
> -        Core, //
> -    },
> +    device::Core,
>      of,
>      prelude::*,
>      serdev,
> @@ -60,14 +57,6 @@ fn probe<'bound>(
>  
>          Ok(Self { sdev: sdev.into() })
>      }
> -
> -    fn receive<'bound>(
> -        sdev: &'bound serdev::Device<Bound>,
> -        _this: Pin<&Self>,
> -        data: &[u8],
> -    ) -> usize {
> -        sdev.write(data).unwrap_or_default() as usize
> -    }
>  }
>  
>  impl Drop for SampleDriver {
>
> ---
> base-commit: e5e04726cdd043e309677071ab1b65a4b18f422b
> change-id: 20260904-rust_serdev_fix-be3ff9c8a5e8

Re: [PATCH v4] rust: serdev: Mitigate race conditions

[Markus Probst]

[-- Attachment #1: Type: text/plain, Size: 4834 bytes --]

On Sun, 2026-09-06 at 00:39 +0100, Gary Guo wrote:
> On Sat Sep 5, 2026 at 7:47 PM BST, Markus Probst wrote:
> > There are currently 2 race conditions:
> > - in probe if `Driver::probe` returns Err
> > - in unbind
> > . In those cases the driver data will be set to NULL before the serdev
> > device was closed. If data is received while the driver data is dropped,
> > the `receive_buf_callback` might try to access the `active` mutex on a
> > null pointer. The race conditions can only occur if `Driver::receive` is
> > implemented.
> > 
> > The issue cannot be cleanly fixed without rewriting some logic, which
> > might introduce new regressions.
> > 
> > Temporarily disable the use of `Driver::receive`, until fixed in the
> > next release.
> 
> Isn't this basically undoing all the usefulness of having the serdev abstraction
> in the first place?
It removes 50% of its functionality, so kinda. The initial
"synology_microp" driver [1] in the mailing list for instance currently
doesn't rely on it (it will need it later though).
> 
> As this is for unbound only, I don't think we need to rush to fix this for rc
> anyway. Many C drivers don't get unbind correct (and our misc device). There's
> no in-tree user that will hit this API anyway.
Also on probe failure (i.e. Driver returns an error on probe).
> 
> I think we'd better just find a proper fix for the next cycle and leave the
> current code as is.
Thats also an option I wouldn't mind.

The proper fix is almost ready.

Thanks
- Markus Probst

[1]
https://lore.kernel.org/rust-for-linux/20260724-synology_microp_initial-v18-0-fb2f49f10e77@posteo.de/

> 
> Best,
> Gary
> 
> > 
> > Fixes: 99f59aa82341 ("rust: add basic serial device bus abstractions")
> > Reported-by: Sashiko Bot <sashiko-bot@kernel.org>
> > Closes: https://lore.kernel.org/linux-serial/20260905000836.C8FC91F00A3D@smtp.kernel.org/
> > Closes: https://lore.kernel.org/linux-serial/20260903222159.70A911F000E9@smtp.kernel.org/
> > Signed-off-by: Markus Probst <markus.probst@posteo.de>
> > ---
> > I will submit a patch (for the next merge cycle) soon, which will
> > address this issue and make the probe and unbind code less convoluted.
> > ---
> > Changes in v4:
> > - remove comments
> > - Link to v3: https://patch.msgid.link/20260905-rust_serdev_fix-v3-1-b86056c3f7a4@posteo.de
> > 
> > Changes in v3:
> > - mitigate it
> > - Link to v2: https://patch.msgid.link/20260905-rust_serdev_fix-v2-0-35dfcd06ef2e@posteo.de
> > 
> > Changes in v2:
> > - also fix race condition on unbind
> > - Link to v1: https://patch.msgid.link/20260905-rust_serdev_fix-v1-1-2ea92b154a6b@posteo.de
> > ---
> >  rust/kernel/serdev.rs              |  6 +++---
> >  samples/rust/rust_driver_serdev.rs | 13 +------------
> >  2 files changed, 4 insertions(+), 15 deletions(-)
> > 
> > diff --git a/rust/kernel/serdev.rs b/rust/kernel/serdev.rs
> > index 17ca504b7f8d..f4c0f43b5273 100644
> > --- a/rust/kernel/serdev.rs
> > +++ b/rust/kernel/serdev.rs
> > @@ -148,9 +148,8 @@ fn drop(self: Pin<&mut Self>) {
> >  
> >  impl<T: Driver> Adapter<T> {
> >      const OPS: &'static bindings::serdev_device_ops = &bindings::serdev_device_ops {
> > -        receive_buf: if T::HAS_RECEIVE {
> > -            Some(Self::receive_buf_callback)
> > -        } else {
> > +        receive_buf: {
> > +            const_assert!(!T::HAS_RECEIVE);
> >              None
> >          },
> >          write_wakeup: Some(bindings::serdev_device_write_wakeup),
> > @@ -233,6 +232,7 @@ extern "C" fn remove_callback(sdev: *mut bindings::serdev_device) {
> >          T::unbind(sdev, data_pinned);
> >      }
> >  
> > +    #[expect(dead_code)]
> >      extern "C" fn receive_buf_callback(
> >          sdev: *mut bindings::serdev_device,
> >          buf: *const u8,
> > diff --git a/samples/rust/rust_driver_serdev.rs b/samples/rust/rust_driver_serdev.rs
> > index 51b4898cd855..3fd3eef15371 100644
> > --- a/samples/rust/rust_driver_serdev.rs
> > +++ b/samples/rust/rust_driver_serdev.rs
> > @@ -4,10 +4,7 @@
> >  
> >  use kernel::{
> >      acpi,
> > -    device::{
> > -        Bound,
> > -        Core, //
> > -    },
> > +    device::Core,
> >      of,
> >      prelude::*,
> >      serdev,
> > @@ -60,14 +57,6 @@ fn probe<'bound>(
> >  
> >          Ok(Self { sdev: sdev.into() })
> >      }
> > -
> > -    fn receive<'bound>(
> > -        sdev: &'bound serdev::Device<Bound>,
> > -        _this: Pin<&Self>,
> > -        data: &[u8],
> > -    ) -> usize {
> > -        sdev.write(data).unwrap_or_default() as usize
> > -    }
> >  }
> >  
> >  impl Drop for SampleDriver {
> > 
> > ---
> > base-commit: e5e04726cdd043e309677071ab1b65a4b18f422b
> > change-id: 20260904-rust_serdev_fix-be3ff9c8a5e8
> 

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 870 bytes --]