[PATCH v2] usbip: vudc: fix NULL deref in vep_dequeue()

[PATCH v2] usbip: vudc: fix NULL deref in vep_dequeue()

[Sam Day via B4 Relay]

From: Sam Day <me@samcday.com>

vep_alloc_request() wasn't initializing vrequest->udc, so cancellations
on the FunctionFS AIO path were arriving in vep_dequeue without a valid
UDC reference.

Since vrequest->udc is never actually properly used anywhere, we opt to
remove it, and update vep_dequeue to obtain a reference to the udc with
ep_to_vudc(), consistent with the other vep_ ops.

AFAICT this bug has existed for ~10 years. Seems that nobody has really
stressed the FunctionFS AIO path on usbip's vudc.

I tested this fix in a QEMU aarch64 guest driving FunctionFS endpoints
via AIO. Before the fix, running `usbip attach` from the host would
cause the guest to oops with the following backtrace:

Call trace:
 vep_dequeue+0x1c/0xe4 (P)
 usb_ep_dequeue+0x14/0x20
 ffs_aio_cancel+0x24/0x34
 __arm64_sys_io_cancel+0xb0/0x124
 do_el0_svc+0x68/0x100
 el0_svc+0x18/0x5c
 el0t_64_sync_handler+0x98/0xdc
 el0t_64_sync+0x154/0x158

Assisted-by: opencode:openai/gpt-5.5
Fixes: b6a0ca111867 ("usbip: vudc: Add UDC specific ops")
Signed-off-by: Sam Day <me@samcday.com>
---
Changes in v2:
- Reworked to remove vrequest->udc entirely and a call to ep_to_vudc()
  from vep_deqe
- Link to v1: https://lore.kernel.org/r/20260619-usbip-vudc-deque-fix-v1-1-9021463b1903@samcday.com
---
 drivers/usb/usbip/vudc.h     | 1 -
 drivers/usb/usbip/vudc_dev.c | 2 +-
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/usb/usbip/vudc.h b/drivers/usb/usbip/vudc.h
index faf61c9c6a98..5ef0e7d9b23a 100644
--- a/drivers/usb/usbip/vudc.h
+++ b/drivers/usb/usbip/vudc.h
@@ -38,7 +38,6 @@ struct vep {
 
 struct vrequest {
 	struct usb_request req;
-	struct vudc *udc;
 	struct list_head req_entry; /* Request queue */
 };
 
diff --git a/drivers/usb/usbip/vudc_dev.c b/drivers/usb/usbip/vudc_dev.c
index c5f079c5a1ea..f0a1a44c18e3 100644
--- a/drivers/usb/usbip/vudc_dev.c
+++ b/drivers/usb/usbip/vudc_dev.c
@@ -344,7 +344,7 @@ static int vep_dequeue(struct usb_ep *_ep, struct usb_request *_req)
 
 	ep = to_vep(_ep);
 	req = to_vrequest(_req);
-	udc = req->udc;
+	udc = ep_to_vudc(ep);
 
 	if (!udc->driver)
 		return -ESHUTDOWN;

---
base-commit: 5cd1731cc883a9914d91e3b93d4597317b5b5339
change-id: 20260619-usbip-vudc-deque-fix-d316de3d3f16

Best regards,
-- 
Sam Day <me@samcday.com>

Re: [PATCH v2] usbip: vudc: fix NULL deref in vep_dequeue()

[Igor Kotrasinski/Security (PLT) /SRPOL/Engineer/Samsung Electronics]

> From: Sam Day <me@samcday.com>
> 
> vep_alloc_request() wasn't initializing vrequest->udc, so cancellations
> on the FunctionFS AIO path were arriving in vep_dequeue without a valid
> UDC reference.
> 
> Since vrequest->udc is never actually properly used anywhere, we opt to
> remove it, and update vep_dequeue to obtain a reference to the udc with
> ep_to_vudc(), consistent with the other vep_ ops.
> 
> AFAICT this bug has existed for ~10 years. Seems that nobody has really
> stressed the FunctionFS AIO path on usbip's vudc.
> 
> I tested this fix in a QEMU aarch64 guest driving FunctionFS endpoints
> via AIO. Before the fix, running `usbip attach` from the host would
> cause the guest to oops with the following backtrace:
> 
> Call trace:
>   vep_dequeue+0x1c/0xe4 (P)
>   usb_ep_dequeue+0x14/0x20
>   ffs_aio_cancel+0x24/0x34
>   __arm64_sys_io_cancel+0xb0/0x124
>   do_el0_svc+0x68/0x100
>   el0_svc+0x18/0x5c
>   el0t_64_sync_handler+0x98/0xdc
>   el0t_64_sync+0x154/0x158
> 
> Assisted-by: opencode:openai/gpt-5.5
> Fixes: b6a0ca111867 ("usbip: vudc: Add UDC specific ops")
> Signed-off-by: Sam Day <me@samcday.com>
> ---
> Changes in v2:
> - Reworked to remove vrequest->udc entirely and a call to ep_to_vudc()
>    from vep_deqe
> - Link to v1: https://lore.kernel.org/r/20260619-usbip-vudc-deque-fix-v1-1-9021463b1903@samcday.com
> ---
>   drivers/usb/usbip/vudc.h     | 1 -
>   drivers/usb/usbip/vudc_dev.c | 2 +-
>   2 files changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/usb/usbip/vudc.h b/drivers/usb/usbip/vudc.h
> index faf61c9c6a98..5ef0e7d9b23a 100644
> --- a/drivers/usb/usbip/vudc.h
> +++ b/drivers/usb/usbip/vudc.h
> @@ -38,7 +38,6 @@ struct vep {
>   
>   struct vrequest {
>   	struct usb_request req;
> -	struct vudc *udc;
>   	struct list_head req_entry; /* Request queue */
>   };
>   
> diff --git a/drivers/usb/usbip/vudc_dev.c b/drivers/usb/usbip/vudc_dev.c
> index c5f079c5a1ea..f0a1a44c18e3 100644
> --- a/drivers/usb/usbip/vudc_dev.c
> +++ b/drivers/usb/usbip/vudc_dev.c
> @@ -344,7 +344,7 @@ static int vep_dequeue(struct usb_ep *_ep, struct usb_request *_req)
>   
>   	ep = to_vep(_ep);
>   	req = to_vrequest(_req);
> -	udc = req->udc;
> +	udc = ep_to_vudc(ep);
>   
>   	if (!udc->driver)
>   		return -ESHUTDOWN;
> 
> ---
> base-commit: 5cd1731cc883a9914d91e3b93d4597317b5b5339
> change-id: 20260619-usbip-vudc-deque-fix-d316de3d3f16
> 
> Best regards,

Looks good to me.
Reviewed-by: Igor Kotrasinski <i.kotrasinsk@samsung.com>

BRs,

Igor Kotrasiński

Re: [PATCH v2] usbip: vudc: fix NULL deref in vep_dequeue()

[Greg Kroah-Hartman]

On Sat, Jun 20, 2026 at 10:18:09AM +1000, Sam Day via B4 Relay wrote:
> From: Sam Day <me@samcday.com>
> 
> vep_alloc_request() wasn't initializing vrequest->udc, so cancellations
> on the FunctionFS AIO path were arriving in vep_dequeue without a valid
> UDC reference.
> 
> Since vrequest->udc is never actually properly used anywhere, we opt to
> remove it, and update vep_dequeue to obtain a reference to the udc with
> ep_to_vudc(), consistent with the other vep_ ops.
> 
> AFAICT this bug has existed for ~10 years. Seems that nobody has really
> stressed the FunctionFS AIO path on usbip's vudc.
> 
> I tested this fix in a QEMU aarch64 guest driving FunctionFS endpoints
> via AIO. Before the fix, running `usbip attach` from the host would
> cause the guest to oops with the following backtrace:
> 
> Call trace:
>  vep_dequeue+0x1c/0xe4 (P)
>  usb_ep_dequeue+0x14/0x20
>  ffs_aio_cancel+0x24/0x34
>  __arm64_sys_io_cancel+0xb0/0x124
>  do_el0_svc+0x68/0x100
>  el0_svc+0x18/0x5c
>  el0t_64_sync_handler+0x98/0xdc
>  el0t_64_sync+0x154/0x158
> 
> Assisted-by: opencode:openai/gpt-5.5
> Fixes: b6a0ca111867 ("usbip: vudc: Add UDC specific ops")
> Signed-off-by: Sam Day <me@samcday.com>
> ---
> Changes in v2:
> - Reworked to remove vrequest->udc entirely and a call to ep_to_vudc()
>   from vep_deqe
> - Link to v1: https://lore.kernel.org/r/20260619-usbip-vudc-deque-fix-v1-1-9021463b1903@samcday.com
> ---
>  drivers/usb/usbip/vudc.h     | 1 -
>  drivers/usb/usbip/vudc_dev.c | 2 +-
>  2 files changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/usb/usbip/vudc.h b/drivers/usb/usbip/vudc.h
> index faf61c9c6a98..5ef0e7d9b23a 100644
> --- a/drivers/usb/usbip/vudc.h
> +++ b/drivers/usb/usbip/vudc.h
> @@ -38,7 +38,6 @@ struct vep {
>  
>  struct vrequest {
>  	struct usb_request req;
> -	struct vudc *udc;
>  	struct list_head req_entry; /* Request queue */
>  };
>  
> diff --git a/drivers/usb/usbip/vudc_dev.c b/drivers/usb/usbip/vudc_dev.c
> index c5f079c5a1ea..f0a1a44c18e3 100644
> --- a/drivers/usb/usbip/vudc_dev.c
> +++ b/drivers/usb/usbip/vudc_dev.c
> @@ -344,7 +344,7 @@ static int vep_dequeue(struct usb_ep *_ep, struct usb_request *_req)
>  
>  	ep = to_vep(_ep);
>  	req = to_vrequest(_req);
> -	udc = req->udc;
> +	udc = ep_to_vudc(ep);

Why is "req" still needed now?  Shouldn't that variable be removed also?

thanks,

greg k-h