If usb_add_hcd() fails in vhci_hcd_probe() (i.e., a probe failure),
the error path calls usb_remove_hcd() and also sets 
pdev->dev.driver_data to NULL.

Consequently, any subsequent call to platform_get_drvdata(pdev) 
(which returns pdev->dev.driver_data) may yield NULL, causing a 
crash if that pointer is dereferenced.

Fix this by adding a sanity check to ensure "hcd" is non-NULL
before proceeding with further operations.

Reported-by: syzbot <syzbot+83976e47ec1ef91e66f1@syzkaller.appspotmail.com>
Closes: https://syzkaller.appspot.com/bug?extid=83976e47ec1ef91e66f1
Tested-by: syzbot <syzbot+83976e47ec1ef91e66f1@syzkaller.appspotmail.com>
Fixes: 03cd00d538a6 ("usbip: vhci-hcd: Set the vhci structure up to work")
Signed-off-by: Qasim Ijaz <qasdev00@gmail.com>
---
v3:
- Added this section to comply with patch format requirements
- Removed the extra newline after `hcd = platform_get_drvdata(pdev);`

 drivers/usb/usbip/vhci_sysfs.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/usb/usbip/vhci_sysfs.c b/drivers/usb/usbip/vhci_sysfs.c
index d5865460e82d..d4a1aa6d06b2 100644
--- a/drivers/usb/usbip/vhci_sysfs.c
+++ b/drivers/usb/usbip/vhci_sysfs.c
@@ -76,6 +76,9 @@ static ssize_t status_show_vhci(int pdev_nr, char *out)
 	}
 
 	hcd = platform_get_drvdata(pdev);
+	if (!hcd)
+		return 0;
+
 	vhci_hcd = hcd_to_vhci_hcd(hcd);
 	vhci = vhci_hcd->vhci;
 
-- 
2.39.5

On Wed, Jan 22, 2025 at 12:58:57PM +0000, Qasim Ijaz wrote:
> If usb_add_hcd() fails in vhci_hcd_probe() (i.e., a probe failure),
> the error path calls usb_remove_hcd() and also sets 
> pdev->dev.driver_data to NULL.
> 
> Consequently, any subsequent call to platform_get_drvdata(pdev) 
> (which returns pdev->dev.driver_data) may yield NULL, causing a 
> crash if that pointer is dereferenced.

If usb_add_hcd() fails during probe then it should not be possible to 
have any subsequent calls to platform_get_drvdata(pdev).

> Fix this by adding a sanity check to ensure "hcd" is non-NULL
> before proceeding with further operations.
> 
> Reported-by: syzbot <syzbot+83976e47ec1ef91e66f1@syzkaller.appspotmail.com>
> Closes: https://syzkaller.appspot.com/bug?extid=83976e47ec1ef91e66f1
> Tested-by: syzbot <syzbot+83976e47ec1ef91e66f1@syzkaller.appspotmail.com>
> Fixes: 03cd00d538a6 ("usbip: vhci-hcd: Set the vhci structure up to work")
> Signed-off-by: Qasim Ijaz <qasdev00@gmail.com>
> ---
> v3:
> - Added this section to comply with patch format requirements
> - Removed the extra newline after `hcd = platform_get_drvdata(pdev);`
> 
>  drivers/usb/usbip/vhci_sysfs.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/usb/usbip/vhci_sysfs.c b/drivers/usb/usbip/vhci_sysfs.c
> index d5865460e82d..d4a1aa6d06b2 100644
> --- a/drivers/usb/usbip/vhci_sysfs.c
> +++ b/drivers/usb/usbip/vhci_sysfs.c
> @@ -76,6 +76,9 @@ static ssize_t status_show_vhci(int pdev_nr, char *out)
>  	}
>  
>  	hcd = platform_get_drvdata(pdev);
> +	if (!hcd)
> +		return 0;

In this case, for example, the status sysfs attribute file should never 
have been created in the first place if probing failed.

This seems more like a synchronization or logic error, not something to 
be solved with a simple check for a NULL pointer

Alan Stern

> +
>  	vhci_hcd = hcd_to_vhci_hcd(hcd);
>  	vhci = vhci_hcd->vhci;
>  
> -- 
> 2.39.5
> 
>