PosixACL: lost default rules after reboot

[ppppggg]

I'm using latest stable ZoL on CentOS 7 (up to date).

The problem is that POSIX ACLs which I set using setfacl are lost after reboots.

[root@localhost ~]# rpm -qa | grep ^kernel
kernel-3.10.0-327.13.1.el7.x86_64
kernel-headers-3.10.0-327.13.1.el7.x86_64
kernel-devel-3.10.0-327.13.1.el7.x86_64
kernel-tools-3.10.0-327.13.1.el7.x86_64
kernel-3.10.0-327.el7.x86_64
kernel-tools-libs-3.10.0-327.13.1.el7.x86_64

[root@localhost ~]# dmesg | grep -E 'SPL:|ZFS:'
[    6.210253] SPL: Loaded module v0.6.5.6-1
[    6.763326] ZFS: Loaded module v0.6.5.6-1, ZFS pool version 5000, ZFS filesystem version 5
[   11.014731] SPL: using hostid 0x00000000

zpool details:

[root@localhost ~]# zpool history
History for 'data':
2016-04-13.14:12:42 zpool create -f -m /mnt/data data /root/disk1
2016-04-13.14:15:08 zfs set acltype=posixacl data
2016-04-13.14:15:20 zfs set aclinherit=passthrough data
2016-04-13.14:15:43 zfs set compression=lz4 data
2016-04-13.14:15:58 zfs set atime=off data
2016-04-13.14:16:05 zfs set relatime=off data

Steps to reproduce:

[root@localhost ~]# cd /mnt/data/
[root@localhost data]# mkdir test_dir
[root@localhost data]# setfacl -R --mask -m u:uadm:rwX test_dir/
[root@localhost data]# setfacl -R -d --mask -m u:uadm:rwX test_dir/
[root@localhost data]# getfacl test_dir/
# file: test_dir/
# owner: root
# group: root
user::rwx
user:uadm:rwx
group::r-x
mask::rwx
other::r-x
default:user::rwx
default:user:uadm:rwx
default:group::r-x
default:mask::rwx
default:other::r-x

[root@localhost data]# reboot

After reboot I login into test system again and get this ACLs:

[root@localhost ~]# cd /mnt/data/
[root@localhost data]# getfacl test_dir/
# file: test_dir/
# owner: root
# group: root
user::rwx
user:uadm:rwx
group::r-x
mask::rwx
other::r-x

No anymore default rules after reboot.

[ppppggg]

zfs umount -a
zfs mount -a

is enought to reproduce the reported problem.

[ppppggg]

The problem didn't reproduced on versions below:

[    5.892572] SPL: Loaded module v0.6.3-1.3
[    6.229796] ZFS: Loaded module v0.6.3-1.3, ZFS pool version 5000, ZFS filesystem version 5
[    8.531342] SPL: using hostid 0x00000000

[ppppggg]

Problem was introduced between this two versions:

[    0.895993] SPL: Loaded module v0.6.5.4-1
[    0.948350] ZFS: Loaded module v0.6.5.4-1, ZFS pool version 5000, ZFS filesystem version 5

and

[    0.895052] SPL: Loaded module v0.6.5.5-1
[    0.943625] ZFS: Loaded module v0.6.5.5-1, ZFS pool version 5000, ZFS filesystem version 5

I've updated ZFS step-by-step from 0.6.3 to 0.6.5.5 to find this.

[nedbass]

@ppppggg thanks for bisecting to the release that introduced the bug. I'll look into it.

[behlendorf]

@ppppggg thanks for so clearly reporting this issue. @nedbass identified the root cause and has a fix which has been merged to master. We'll get this fix in to the next point release but you can cherry-pick it now if you need it.

98f0369 Fix ZPL miswrite of default POSIX ACL