Hello,

On Mon, Sep 28, 2026 at 01:44:22PM -0500, Danish Khateeb wrote:
> pwm_cdev_fileops has no .compat_ioctl, so every ioctl on /dev/pwmchipN
> fails with -ENOTTY for 32-bit tasks on a 64-bit kernel, e.g. armhf
> userspace on arm64.
> 
> struct pwmchip_waveform has the same layout for 32-bit and 64-bit tasks,
> so compat_ptr_ioctl() is enough. PWM_IOCTL_REQUEST and PWM_IOCTL_FREE
> take an integer rather than a pointer, which is fine too: compat_ptr()
> only changed the value on s390, which dropped compat support in v6.19.
> 
> Fixes: 9c06f26ba5f5 ("pwm: Add support for pwmchip devices for faster and easier userspace access")
> Cc: stable@vger.kernel.org
> Assisted-by: LLM
> Signed-off-by: Danish Khateeb <danishkhateeb03@gmail.com>
> ---
> 
> Notes:
>     Tested on x86_64 in QEMU (pwm/for-next 17dbb6938d3a plus this patch,
>     KASAN and lockdep) with a PCA9685 instantiated on i2c-stub, which gets
>     /dev/pwmchip0. A test program runs all six ioctls, including the error
>     cases (hwpwm out of range or not requested, __pad != 0, use after FREE,
>     SETEXACTWF returning -EDOM):
>     
>     - before: every ioctl from the -m32 build fails with -ENOTTY, while the
>       64-bit build works
>     - after: the -m32 output matches the 64-bit output line for line
>     
>     No splats. W=1 builds are clean for arm64 (COMPAT=y, LLVM=1) and for
>     i386 (no COMPAT).
> 
>  drivers/pwm/core.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/pwm/core.c b/drivers/pwm/core.c
> index 2a050ff4608b..4c6d716e5fde 100644
> --- a/drivers/pwm/core.c
> +++ b/drivers/pwm/core.c
> @@ -2389,6 +2389,7 @@ static const struct file_operations pwm_cdev_fileops = {
>  	.release = pwm_cdev_release,
>  	.owner = THIS_MODULE,
>  	.unlocked_ioctl = pwm_cdev_ioctl,
> +	.compat_ioctl = compat_ptr_ioctl,
>  };

Not entirely sure about the stable backport, but the change looks fine.

Applied to
https://git.kernel.org/pub/scm/linux/kernel/git/ukleinek/linux.git pwm/for-next
but dropped the Cc:.

Thanks
Uwe