On Fri, 21 Feb 2025 15:23:11 +0200, Mathias Nyman wrote: > update_ring_for_set_deq_completion() isn't needed anymore. > Niklas already wrote a patch to remove it. > > It's sitting in my for-usb-next branch > > https://git.kernel.org/pub/scm/linux/kernel/git/mnyman/xhci.git/commit/?h=for-usb-next&id=ee7dab196a7dfc48a1608274329323cb10b4340d I know that it's pure paranoia, but this patch removes a quiet debug message about an obviously abnormal and likely harmful condition. My patch turned it into a nice red error. This will be something to keep in mind, for example somebody could write a patch which reclaims an empty transfer ring segment without chceking if there is a Set TR Dequeue pending to this segemnt. That could lead to very interesting outcomes if all TRBs in the segment are No-Op'ed and the next page is a transfer ring of other endpoint. Other than that, I suppose there is currently no problem. The loop for finding new dequeue position doesn't seem prone to jumping off the ring as long as the ring itself isn't corrupted. Michal
