fsck on current not always prompting a 2nd run when it should

[Kirk McKusick]

> To: fs at freebsd.org
> Subject: fsck on current not always prompting a 2nd run when it should
> From: "Julian H. Stacey" <jhs at berklix.com>
> Organization: http://berklix.eu BSD Linux Unix Consultants, Munich Germany
> Date: Wed, 03 Feb 2016 02:34:43 +0100
> Cc: "Julian H. Stacey" <jhs at berklix.com>
> 
> Hi fs@ people
> fsck is sometimes (not always) not prompting for a 2nd run of fsck
> when FS is still bad.  (After a few crashes I've become cautious
> enough to give it a 2nd run even after it first detects no errors
> left to fix.  There's a bug in fsck somewhere, it should ask for a
> re-run more than it does.
> 
> fsck running on uname -a
>   FreeBSD lapr.js.berklix.net 11.0-CURRENT FreeBSD 11.0-CURRENT #12182:
>   Mon Oct 19 23:57:08 CEST 2015
>   jhs at lapr.js.berklix.net:/usr/src/sys/amd64/compile/LAPR.small  amd64
> 
> Background:
>    partition was corrupted running 10.2-RELEASE, but fsck is running on current.
> 
>    On 10.2-RELEASE I've been doing intensive IO for days, 
>    cd /usr/ports  ; make reinstall-recursive
>   	# http://berklix.com/~jhs/src/bsd/fixes/FreeBSD/ports/gen/Mk/
>    with a selection of */Makefile.inc that result in my 1000+ favourites
>    being shown by
>   	pkg info | wc -l
>    This has repeatedly reinstalled some common dependency ports/, hence heavy IO.
>    + rdist mirroring,
>    + devd usb ufs stick mounts, mdconfig, gbde, shells to umount & 
>    maybe mdconfig -u after.
>    Plenty of activity might have caused the corruption, but ...
> 
> I'm not querying corruption on 10.2-RELEASE I'm just concerned fsck on
> current fails to tell me to re-run fsck.
> 
> I didnt save previous fsck reports, but have this one.
> Current fstab:
> 	/dev/ada0s4f   /data   ufs  rw,noauto    1  5
> 
> rc.conf:	background_fsck="NO"
> 
> dumpfs -m /dev/ada0s4f
> newfs -O 2 -U -a 4 -b 32768 -d 32768 -e 4096 -f 4096 -g 16384 -h 64 -i 8192 -j -k 6408 -m 8 -o time -s 1793576920 /dev/ada0s4f 
> 
> fsck -y /data
> 	** /dev/ada0s4f
> 
> 	USE JOURNAL? yes
> 
> 	** SU+J Recovering /dev/ada0s4f
> 	** Reading 33554432 byte journal from inode 70.
> 
> 	RECOVER? yes
> 
> 	** Building recovery table.
> 	** Resolving unreferenced inode list.
> 	** Processing journal entries.
> 
> 	WRITE CHANGES? yes
> 
> 	** 1974 journal records in 105472 bytes for 59.89% utilization
> 	** Freed 0 inodes (0 dirs) 0 blocks, and 0 frags.
> 
> 	***** FILE SYSTEM MARKED CLEAN *****
> That ran very suspiciously quickly, so I ran it again, 
> took a long time (as about 900 gig)
> 
> fsck -y /data
> 	** /dev/ada0s4f
> 
> 	USE JOURNAL? yes
> 
> 	** SU+J Recovering /dev/ada0s4f
> 	Journal timestamp does not match fs mount time
> 	** Skipping journal, falling through to full fsck
> 
> 	** Last Mounted on /s4/data
> 	** Phase 1 - Check Blocks and Sizes
> 	** Phase 2 - Check Pathnames
> 	UNALLOCATED  I=19604676  OWNER=root MODE=0
> 	SIZE=0 MTIME=Feb  3 00:52 2016 
> 	NAME=/release/10.2-RELEASE/usr/ports/x11/pixman/work/.PLIST.setuid
> 
> 	UNEXPECTED SOFT UPDATE INCONSISTENCY
> 
> 	REMOVE? yes
> 
> 	UNALLOCATED  I=19604677  OWNER=root MODE=0
> 	SIZE=0 MTIME=Feb  3 00:52 2016 
> 	NAME=/release/10.2-RELEASE/usr/ports/x11/pixman/work/.PLIST.writable
> 
> 	UNEXPECTED SOFT UPDATE INCONSISTENCY
> 
> 	REMOVE? yes
> 
> 	UNALLOCATED  I=19604678  OWNER=root MODE=0
> 	SIZE=0 MTIME=Feb  3 00:52 2016 
> 	NAME=/release/10.2-RELEASE/usr/ports/x11/pixman/work/.PLIST.objdump
> 
> 	UNEXPECTED SOFT UPDATE INCONSISTENCY
> 
> 	REMOVE? yes
> 
> 	UNALLOCATED  I=19604679  OWNER=root MODE=0
> 	SIZE=0 MTIME=Feb  3 00:40 2016 
> 	NAME=/release/10.2-RELEASE/usr/ports/x11/pixman/work/.install_done.pixman._usr_local
> 
> 	UNEXPECTED SOFT UPDATE INCONSISTENCY
> 
> 	REMOVE? yes
> 	** Phase 3 - Check Connectivity
> 	** Phase 4 - Check Reference Counts
> 	** Phase 5 - Check Cyl groups
> 	FREE BLK COUNT(S) WRONG IN SUPERBLK
> 	SALVAGE? yes
> 
> 	SUMMARY INFORMATION BAD
> 	SALVAGE? yes
> 
> 	BLK(S) MISSING IN BIT MAPS
> 	SALVAGE? yes
> 
> 	10707223 files, 158068913 used, 59088404 free (1210332 frags, 7234759 blocks, 0.6% fragmentation)
> 
> 	***** FILE SYSTEM MARKED DIRTY *****
> 
> 	***** FILE SYSTEM WAS MODIFIED *****
> 
> 	***** PLEASE RERUN FSCK *****
> 
> fsck -y /data
> 	** /dev/ada0s4f
> 
> 	USE JOURNAL? yes
> 
> 	** SU+J Recovering /dev/ada0s4f
> 	Journal timestamp does not match fs mount time
> 	** Skipping journal, falling through to full fsck
> 
> 	** Last Mounted on /s4/data
> 	** Phase 1 - Check Blocks and Sizes
> 	** Phase 2 - Check Pathnames
> 	** Phase 2 - Check Pathnames
> 	** Phase 3 - Check Connectivity
> 	** Phase 4 - Check Reference Counts
> 	** Phase 5 - Check Cyl groups
> 	10707223 files, 158068913 used, 59088404 free (1210332 frags, 7234759 blocks, 0.6% fragmentation)
> 
> 	***** FILE SYSTEM MARKED CLEAN *****
> 
> 	***** FILE SYSTEM WAS MODIFIED *****
> 
> Cheers,
> Julian
> -- 
> Julian Stacey,  BSD Linux Unix Sys. Eng. Consultant Munich http://berklix.eu
>  Mail plain text,  No quoted-printable, HTML, base64, MS.doc.
>  Prefix old lines '> '  Reply below old, like play script.  Break lines by 80.

If you tell fsck to use the journal, it assumes that the filesystem is
basically in good shape and it just needs to take care of the
transactions in the journal.  That way the the reboot is much quicker
because it doesn't have to wait for a full fsck.

If media errors or other unexpected problems arise, then they are
not known by the journal and will only be found by running a full
fsck (which is what is what you got when you did a second run above).

	Kirk McKusick