[Bug 274271] zfs deadlock while/after writing files
Date: Wed, 04 Oct 2023 23:27:35 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=274271
Bug ID: 274271
Summary: zfs deadlock while/after writing files
Product: Base System
Version: 13.2-RELEASE
Hardware: amd64
OS: Any
Status: New
Severity: Affects Only Me
Priority: ---
Component: kern
Assignee: bugs@FreeBSD.org
Reporter: pmc@citylink.dinoex.sub.org
After building ports for 47+ hours, finally the clone of /usr/ports should be
exported; but then zfs was tired and deadlocked:
UID PID PPID C PRI NI VSZ RSS MWCHAN STAT TT TIME
COMM
0 40329 41971 13 134 0 302596 151632 select IN+ 3 0:05.97 git
-C /var/sysup/mnt/tmp.6.41971 checkout-index -a
--prefix=/var/sysup/mnt/tmp.8.41971/
0 40678 40329 14 134 0 0 0 - ZN+ 3 0:02.08
<defunct>
0 40679 40329 10 134 0 0 0 - ZN+ 3 0:02.06
<defunct>
0 40680 40329 0 134 0 0 0 - ZN+ 3 0:02.08
<defunct>
0 40681 40329 8 134 0 682132 165796 zfsvfs-> DN+ 3 0:00.99
/usr/local/libexec/git-core/git checkout--worker
--prefix=/var/sysup/mnt/tmp.8.41971/
0 40682 40329 11 134 0 0 0 - ZN+ 3 0:02.35
<defunct>
0 40683 40329 0 134 0 0 0 - ZN+ 3 0:02.37
<defunct>
0 40684 40329 8 134 0 0 0 - ZN+ 3 0:03.11
<defunct>
killing of the git process is not possible, umount of the fs is not possible,
reboot will also fail (get stuck).
scrub of the concerned pool is still possible and shows no errors. Neither are
any errors reported from the disks or the ECC memory.
pool: build
state: ONLINE
scan: scrub repaired 0B in 00:15:11 with 0 errors on Wed Oct 4 23:34:26 2023
config:
NAME STATE READ WRITE CKSUM
build ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
da3p1.eli ONLINE 0 0 0
da4p1.eli ONLINE 0 0 0
da5p1.eli ONLINE 0 0 0
special
mirror-1 ONLINE 0 0 0
ada1p9.elip10 ONLINE 0 0 0
ada6p9.elip10 ONLINE 0 0 0
cache
ada3p9.elip6 ONLINE 0 0 0
errors: No known data errors
The last command in zpool history before the deadlock is
2023-10-04.22:30:58 zfs create -o mountpoint=legacy build/tmp.7.41971
This was then legacy-mounted to /var/sysup/mnt/tmp.8.41971/ and the git process
started, which didn't complete.
The fileset has received most/all of the data (the other git workers have
apparently completed):
build/tmp.7.41971 970M 63.0G 970M
legacy
--
You are receiving this mail because:
You are the assignee for the bug.