I ran blktests with v6.6-rc1 kernel and observed the test case nbd/002 failed
with the message below. The failure is recreated 100% by running the test case.
---
nbd/002 (tests on partition handling for an nbd device) [failed]
runtime 1.620s ... 0.369s
--- tests/nbd/002.out 2023-04-06 10:11:07.923670528 +0900
+++ /home/shin/Blktests/blktests/results/nodev/nbd/002.out.bad 2023-09-11 12:03:30.901246261 +0900
@@ -1,4 +1,3 @@
Running nbd/002
Testing IOCTL path
-Testing the netlink path
-Test complete
+Didn't have partition on ioctl path
---
I checked nbd changes in v6.6-rc1 and found the commit 0c1c9a27ce90 ("nbd: call
blk_mark_disk_dead in nbd_clear_sock_ioctl") [1] is the trigger. I think the
test case expects partitions on nbd devices are kept after nbd disconnect and
reconnect. On the other hand, the trigger commit looks removing the partitions
after nbd disconnect and reconnect. I'm not sure whether of the test case or the
kernel should be fixed.
[1] https://lore.kernel.org/linux-block/20230811100828.1897174-8-hch@xxxxxx/
