| Message ID | 1419932967-2694-1-git-send-email-songjun.wu@atmel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, On Tue, Dec 30, 2014 at 05:49:27PM +0800, Songjun Wu wrote: > Executing the 'insmod g_hid.ko', then executing the > 'rmmod g_hid.ko', the NULL pointer oops will be triggered. > > When unloading the module 'g_hid.ko', the urb request will be > dequeued and the completion routine will be excuted. If no urb > packet, the urb request will not be added to the endpoint queue > and the completion routine pointer in urb request is NULL. > Accessing to the NULL function pointer will cause the oops issue. > Add the code to check the urb request is in the endpoint queue > or not.If the urb request is not in the endpoint queue, a negative > error code will be returned. > > This bug was introduced since the file 'atmel_usba_udc.c' was > initialized. Fixes: 914a3f3b3754 (USB: add atmel_usba_udc driver) > Cc: stable@vger.kernel.org # always been there... this is not the way you write this. There are tons of examples in the very git tree you're using of how to do this. Just run: $ git log --grep "Fixes:" Also have a read at Documentation/SubmittingPatches and search for "Fixes" in that file. You might also want to look at Documentation/stable-kernel-rules.txt to figure out how to properly Cc stable.
diff --git a/drivers/usb/gadget/udc/atmel_usba_udc.c b/drivers/usb/gadget/udc/atmel_usba_udc.c index ce88237..48629cc 100644 --- a/drivers/usb/gadget/udc/atmel_usba_udc.c +++ b/drivers/usb/gadget/udc/atmel_usba_udc.c @@ -828,7 +828,7 @@ static int usba_ep_dequeue(struct usb_ep *_ep, struct usb_request *_req) { struct usba_ep *ep = to_usba_ep(_ep); struct usba_udc *udc = ep->udc; - struct usba_request *req = to_usba_req(_req); + struct usba_request *req; unsigned long flags; u32 status; @@ -837,6 +837,16 @@ static int usba_ep_dequeue(struct usb_ep *_ep, struct usb_request *_req) spin_lock_irqsave(&udc->lock, flags); + list_for_each_entry(req, &ep->queue, queue) { + if (&req->req == _req) + break; + } + + if (&req->req != _req) { + spin_unlock_irqrestore(&udc->lock, flags); + return -EINVAL; + } + if (req->using_dma) { /* * If this request is currently being transferred,
Executing the 'insmod g_hid.ko', then executing the 'rmmod g_hid.ko', the NULL pointer oops will be triggered. When unloading the module 'g_hid.ko', the urb request will be dequeued and the completion routine will be excuted. If no urb packet, the urb request will not be added to the endpoint queue and the completion routine pointer in urb request is NULL. Accessing to the NULL function pointer will cause the oops issue. Add the code to check the urb request is in the endpoint queue or not.If the urb request is not in the endpoint queue, a negative error code will be returned. This bug was introduced since the file 'atmel_usba_udc.c' was initialized. Fixes: 914a3f3b3754 (USB: add atmel_usba_udc driver) Cc: stable@vger.kernel.org # always been there... oops dump log is shown in the following. Unable to handle kernel NULL pointer dereference at virtual address 00000000 pgd = dedf0000 [00000000] *pgd=3ede5831, *pte=00000000, *ppte=00000000 Internal error: Oops: 80000007 [#1] ARM Modules linked in: g_hid(-) usb_f_hid libcomposite Signed-off-by: Songjun Wu <songjun.wu@atmel.com> --- drivers/usb/gadget/udc/atmel_usba_udc.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)