Return to BSD News archive
Newsgroups: comp.os.386bsd.bugs Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!agate!howland.reston.ans.net!europa.eng.gtefsd.com!MathWorks.Com!news.duke.edu!concert!sas!mozart.unx.sas.com!torpid.unx.sas.com!sastdr From: sastdr@torpid.unx.sas.com (Thomas David Rivers) Subject: Re: serious fsck bug Sender: news@unx.sas.com (Noter of Newsworthy Events) Message-ID: <Cqo6K2.Hr0@unx.sas.com> Date: Tue, 31 May 1994 13:52:50 GMT References: <NILS.94May28022525@guru.stgt.sub.org> <J88NWK-.dysonj@delphi.com> <NILS.94May29133633@guru.stgt.sub.org> Nntp-Posting-Host: torpid.unx.sas.com Organization: SAS Institute Inc. Lines: 43 In article <NILS.94May29133633@guru.stgt.sub.org> nils@ims.uni-stuttgart.de writes: >In article <J88NWK-.dysonj@delphi.com> John Dyson <dysonj@delphi.com> writes: > >> Are you getting I/O errors sometimes during the operation of your system. >> The UFS filesystem can get really messed up if the metadata is not updated >> properly. BTW, after making your backup, have you tried to remake the >> the offending filesystem????? > >I do not get any error while using the drive. And yes, after backing >up the data I remade the filesystems with different numbers of inodes >and minfree (5 or 10%). I backed up and restored the data using >dump/restore and tar. Again: all operations complete without any >error. But nothing helped so far, fsck still complains. > >After spending one day to repair my filesystems I managed to heal one >of them. Lets concentrate on the other: fsck tells me the same things >all over again: unknown file types, partially allocated inodes, bad >indirect blocks (looks like random huge postitive or negative >numbers), duplicated blocks, bad link counts of files (the link count >should be negative!! could there be a sanity check in fsck, please?) >and never ever cleans/removes/reconnects them. The next run of fsck >results in the same errors. > >After writing a small 'unlink' utility I detected the clri command >which supposedly zeros out a given inode. Using this didn't help, the >offending inodes still contained their meaningless data. > >So, is this a problem of my disk, the kernel, fsck? Every time I've encountered this problem (even with SCSI drives) there has been a hardware fault lurking around. I always begin with checking the cabling; then, if the disk isn't SCSI or IDE; I set up a new bad144 entry. Also, a low-level format (even for SCSI drives) can really help with this problem (for SCSI, the bad sectors will be remapp'd during the format.) If this doesn't work; I go out and buy a new disk. - Dave Rivers - (sastdr@unx.sas.com) -- Imagine Whirled Peas.