We had kernel panic with 5.8.11 and 5.8.13. Here's the trace, copied manually from the console: r5l_log_stripe r5l_write_stripe ops_run_io ? async_gen_syndrome ? ops_run_reconstruct6 handle_stripe raid5_do_work ? __switch_to_asm process_one_work worker_thread ? process_scheduled_works kthread ? __kthread_queue_delayed_work ret_from_fork The machine has an array with 14 disks and one with 20, both with it's own journal ssd. It has high I/O flux and is usually very stable. The first crash happened with 5.8.13; I downgraded to 5.8.11 and it happened again. In both cases I had done a check on the arrays (simultaneously) by doing echo check > /sys/block/md<number>/md/sync_action. Note that the crash happened more than 24h after the checks had successfully finished. The only thing I've changed from the usual procedure is to change /sys/block/md<number>/md/group_thread_cnt. I normally use 5 to not limit checks by cpu. In the crash cases I raised group_thread_cnt to 10 soon after the checks started. Any ideas?
