BSoD with bugcheck 0x133 DPC_WATCHDOG_VIOLATION #663
Description
Reported in Npcap 1.71. "The system cumulatively spent an extended period of time at IRQL DISPATCH_LEVEL or above."
Analyzing these dumps, it appears that there is significant resource contention within the Npcap driver when opening a new capture handle while another handle is processing packets. There are packets being sent and received on several different processors at once, each of which has acquired read access to a locked resource. Opening a new handle requires write access to the resource, which means that the thread waits at DPC level for all the read accesses to stop, while each new read access is queued for when the write access is finished. The write operation itself is very fast (2 pointer writes), but because it forces this wait, it appears that the system stays at DPC level for too long, causing the bugcheck 0x133 DPC_WATCHDOG_VIOLATION.
Related: #535