gjournal doesn't want to stop after "gjournal stop" command

[Miroslav Lachman]

I am removing gjournal from all of our machines. I did it many times 
befor with FreeBSD 7.x versions without problem, but on two machines 
upgraded to 8.3, there is the same problem on both of them. The gjournal 
does not want to stop.

There is a list of steps I did:

# uname -srmi
FreeBSD 8.3-RELEASE amd64 GENERIC


# df -h -i /vol1/
Filesystem                    Size    Used   Avail Capacity  Mounted on
/dev/mirror/gm0s2g.journal    781G     59G    659G     8%    /vol1


# gjournal list
Geom name: gjournal 357929960
ID: 357929960
Providers:
1. Name: mirror/gm0s2g.journal
    Mediasize: 866111751168 (806G)
    Sectorsize: 512
    Mode: r1w1e1
Consumers:
1. Name: mirror/gm0s2e
    Mediasize: 2621440000 (2.5G)
    Sectorsize: 512
    Stripesize: 0
    Stripeoffset: 2620062720
    Mode: r1w1e1
    Jend: 2621439488
    Jstart: 0
    Role: Journal
2. Name: mirror/gm0s2g
    Mediasize: 866111751680 (806G)
    Sectorsize: 512
    Stripesize: 0
    Stripeoffset: 946535424
    Mode: r1w1e1
    Role: Data


# umount /vol1


# gjournal stop -v mirror/gm0s2g.journal
Done.


And then log/messages is spammed with following error every 5 seconds


# tail -f /var/log/messages
Jan  8 09:54:56 edith su: quip to root on /dev/pts/0
Jan  8 16:58:27 edith kernel: GEOM_JOURNAL: Journal 357929960: 
ufsid/4ac9defaef44694b contains data.
Jan  8 16:58:32 edith kernel: GEOM_JOURNAL: Timeout. Journal gjournal 
357929960 cannot be completed.
Jan  8 16:58:32 edith kernel: GEOM_JOURNAL: Journal 357929960: 
mirror/gm0s2g contains data.
Jan  8 16:58:37 edith kernel: GEOM_JOURNAL: Timeout. Journal gjournal 
357929960 cannot be completed.
Jan  8 16:58:37 edith kernel: GEOM_JOURNAL: Journal 357929960: 
ufsid/4ac9defaef44694b contains data.
Jan  8 16:58:42 edith kernel: GEOM_JOURNAL: Timeout. Journal gjournal 
357929960 cannot be completed.


# kldstat
Id Refs Address            Size     Name
  1   10 0xffffffff80100000 e56d68   kernel
  2    1 0xffffffff80f57000 1e5f0    geom_journal.ko
  3    1 0xffffffff80f76000 210d8    geom_mirror.ko
  4    1 0xffffffff80f98000 2800     coretemp.ko
  5    1 0xffffffff81012000 2bdbb    pf.ko


# kldunload geom_journal
^C^C

kldunload hangs, I can't interupt it by Ctrl+C


Next commands were done in another ssh screen

# gjournal clear -v mirror/gm0s2e mirror/gm0s2g
Metadata cleared on mirror/gm0s2e.
Cannot clear metadata on mirror/gm0s2g: Operation not permitted.
gjournal: Not fully done.


# fsck -t ufs /dev/mirror/gm0s2g
** /dev/mirror/gm0s2g (NO WRITE)
** Last Mounted on /vol1
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
5986 files, 31059582 used, 378538310 free (8030 frags, 47316285 blocks, 
0.0% fragmentation)


This exact steps were previously done on 7.x machines without any problem.
On 8.3 machines, nothing stop gjournal and only reboot without 
geom_journal loaded solved this problem.

Is there any other way to stop it without reboot?


Miroslav Lachman