Closed stevenjswanson closed 7 years ago
Is this produced by linux-nova::devel and xfstests::hacked-for-gce branches? I couldn't reproduce it and generic/003 looks passing.
generic/003 9s ... 10s Ran: generic/003 Passed all 1 test
You need to run NOVA/001
and then generic/003
.
Try this:
# git clone git@github.com:NVSL/nova-testscripts.git
# cd nova-testscripts/nova-ci
# ./run_tests.sh xfstests NOVA/001 generic/003
The results should end up in results/latest
, including the output of dmesg
for the run. There's i directory for each run named by time. If you don't find it in latest
look in one of the recent ones.
Let me know if it doesn't work.
Reproduced and fixed by https://github.com/NVSL/linux-nova/commit/79c48e403f114bbbffd845f1aa98f3ecb1962297
The cause was inode's checksum was not updated in some gc functions. I also added iput(target_inode)
to nova_seq_gc()
, to avoid VFS: Busy inodes after unmount of pmem1...
, but I'm not sure if it's really appropriate. Please review.