On Wed, May 24, 2023 at 11:07:41AM -0700, Bart Van Assche wrote: > On 5/23/23 22:53, Christoph Hellwig wrote: >> I turns out this causes some kind of hang I haven't been able to >> debug yet in blktests' hotplug test. Can you drop this and the >> subsequent patches for now? > > Hi Christoph, > > I haven't seen this hang in my tests. If you can tell me how to run > blktests I can help with root-causing this issue. This is how I run > blktests: This is a simple ./check run with this config, and most importantly modular scsi_debug (which is not my usual config, othewise I would have noticed it earlier): ----- snip ----- TEST_DEVS=(/dev/vdb) nvme_trtype=tcp ----- snip ----- It hangs in block/001 when probing scsi_debug: [ 242.790601] INFO: task modprobe:3702 blocked for more than 120 seconds. [ 242.791572] Not tainted 6.4.0-rc2+ #1179 [ 242.792201] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 242.793387] task:modprobe state:D stack:0 pid:3702 ppid:3686 flags:0x00004002 [ 242.794724] Call Trace: [ 242.795121] <TASK> [ 242.795465] __schedule+0x307/0x840 [ 242.796053] ? call_usermodehelper_exec+0xee/0x180 [ 242.796812] schedule+0x57/0xa0 [ 242.797316] async_synchronize_full+0xa0/0x130 [ 242.798029] ? destroy_sched_domains_rcu+0x20/0x20 [ 242.798803] do_init_module+0x19f/0x200 [ 242.799657] __do_sys_finit_module+0x9e/0xf0 [ 242.800324] do_syscall_64+0x34/0x80 [ 242.800879] entry_SYSCALL_64_after_hwframe+0x63/0xcd
