On 2020-10-22, Mike Galbraith wrote: > Seems this patch can cause hangs. I can reliably reproduce one such > hang by poking sysrq-b, and it goes away if I revert that. I see an RCU warning and then 30 seconds later the software watchdog kicks in showing the ttyS0 irq handler is stuck in the D state. I thought that kmsg_dumps for non-OOPS scenarios were always preemptible. I will re-evaluate the possible non-OOPS kmsg_dump paths. Thanks for reporting. John Ogness