diff mbox series

usb: core: Add error handling in usb_reset_device for autoresume failure

Message ID 20250220095218.970-1-vulab@iscas.ac.cn (mailing list archive)
State New
Headers show
Series usb: core: Add error handling in usb_reset_device for autoresume failure | expand

Commit Message

Wentao Liang Feb. 20, 2025, 9:52 a.m. UTC
In usb_reset_device(),  the function continues execution and
calls usb_autosuspend_device() after usb_autosuspend_device fails.

To fix this, add error handling for usb_autoresume_device()
and return the error code immediately. This ensures that
usb_autosuspend_device() is not called when usb_autoresume_device()
fails, maintaining device state consistency.

Fixes: 94fcda1f8ab5 ("usbcore: remove unused argument in autosuspend")
Cc: stable@vger.kernel.org # 2.6.20+
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
---
 drivers/usb/core/hub.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Comments

Greg Kroah-Hartman Feb. 20, 2025, 10:15 a.m. UTC | #1
On Thu, Feb 20, 2025 at 05:52:18PM +0800, Wentao Liang wrote:
> In usb_reset_device(),  the function continues execution and
> calls usb_autosuspend_device() after usb_autosuspend_device fails.
> 
> To fix this, add error handling for usb_autoresume_device()
> and return the error code immediately. This ensures that
> usb_autosuspend_device() is not called when usb_autoresume_device()
> fails, maintaining device state consistency.
> 
> Fixes: 94fcda1f8ab5 ("usbcore: remove unused argument in autosuspend")
> Cc: stable@vger.kernel.org # 2.6.20+
> Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
> ---
>  drivers/usb/core/hub.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
> index 21ac9b464696..f2efdbdd1533 100644
> --- a/drivers/usb/core/hub.c
> +++ b/drivers/usb/core/hub.c
> @@ -6292,7 +6292,9 @@ int usb_reset_device(struct usb_device *udev)
>  	noio_flag = memalloc_noio_save();
>  
>  	/* Prevent autosuspend during the reset */
> -	usb_autoresume_device(udev);

Have you ever seen this function fail?

> +	ret = usb_autoresume_device(udev);
> +	if (ret < 0)

if (ret) please.

Also, how was this tested?  How was it found?

thanks,

greg k-h
Oliver Neukum Feb. 20, 2025, 10:22 a.m. UTC | #2
On 20.02.25 10:52, Wentao Liang wrote:
> In usb_reset_device(),  the function continues execution and
> calls usb_autosuspend_device() after usb_autosuspend_device fails.

This can only fail if the device needs to be physically
resumed. In that case you called usb_reset_device() while
you weren't supposed to. The purpose of the call is to keep
the counter elevated in order to disable runtime power management
temporarily.

The code is older than helpers to elevate the count. The correct
fix would be to use them rather than handle an error that cannot
happen.

	Regards
		Oliver
diff mbox series

Patch

diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
index 21ac9b464696..f2efdbdd1533 100644
--- a/drivers/usb/core/hub.c
+++ b/drivers/usb/core/hub.c
@@ -6292,7 +6292,9 @@  int usb_reset_device(struct usb_device *udev)
 	noio_flag = memalloc_noio_save();
 
 	/* Prevent autosuspend during the reset */
-	usb_autoresume_device(udev);
+	ret = usb_autoresume_device(udev);
+	if (ret < 0)
+		goto error_autoresume;
 
 	if (config) {
 		for (i = 0; i < config->desc.bNumInterfaces; ++i) {
@@ -6341,6 +6343,7 @@  int usb_reset_device(struct usb_device *udev)
 	}
 
 	usb_autosuspend_device(udev);
+error_autoresume:
 	memalloc_noio_restore(noio_flag);
 	udev->reset_in_progress = 0;
 	return ret;