Re: [drivers/usb/usbip] BUG: corrupted list in vep_queue

[Shuah Khan <skhan@linuxfoundation.org>]

On 10/11/24 10:21, Chenyuan Yang wrote:
> Hi Greg,
> 
> I'm curious about the best approach to fix this issue. I'm unsure if
> adding a spinlock or mutex around the code would be good to ensure
> that only one thread can modify the list at a time (then fix this
> issue).
> 
> Alternatively, we could check if the entry is already in the list
> within the vep_queue function:
> ```
> if (!list_empty(&new_entry->list)) {
>      // Entry is already in the list, handle it (e.g., log an error or ignore)
>      return -EEXIST; // Or another appropriate error code
> }
> 

vep_queue() function is holding udc lock when it updates
the status and calls list_add_tail(). vep_dequeue() does
the same.

vep_free_request() doesn't see to hold the udc lock. I will
tale a look at this a bit more and get back to you on the
best approach. My guess is that the fix will be in another
function such as vep_free_request() - at the moment - that
routine is the one I suspect.


> // Safe to add the entry to the list
> list_add_tail(&new_entry->list, &queue->list);
> ```
> 
> This approach would avoid unnecessary locking overhead if the entry is
> not already in the list.
> 
> Please let me know your thoughts on these options or if you have any
> other suggestions.
> 
> Best,
> Chenyuan
> 

thanks,
-- Shuah