bin/73019: fsck_ufs: cannot alloc 607016868 bytes for inoinfo

[wynkoop at wynn.com]

The following reply was made to PR bin/73019; it has been noted by GNATS.

From: wynkoop at wynn.com
To: mi at corbulon.video-collage.com (Mikhail Teterin)
Cc: wynkoop at wynn.com, freebsd-gnats-submit at FreeBSD.org
Subject: Re: bin/73019: fsck_ufs: cannot alloc 607016868 bytes for inoinfo
Date: Sun, 12 Dec 2004 21:36:35 +0000 (GMT)

 > 
 > > Greeting-
 > >      I just had the exact same problem in FreeBSD 5.3.  In my case it is 
 > > my var partation on a just installed system.  It will not FSCK using any
 > > superblock and I can not even mount the FS read only. 
 > >      I had softupdates off for the file system.  Before the failure
 > > happened
 > > I was getting DMA write failure errors on the console for the ad0.
 > 
 > Are you using Silicon Image's SATA controller by any chance? See
 > 
 > 	http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/72451
 > 
 > As for recovering your /var, there are, apparently, a lot more superblock
 > copies, than fsck tries automatically.
 > 
 > To get the actual number, try newfs-ing the partition with exactly the
 > same parameters as before, but with ``-N'' flag. This will print out the
 > addresses of the superblocks. See if giving one of them to fsck_ufs with
 > ``-b'' will help you recover stuff. I have not tried this personally, but
 > it seems like you've lost your /var anyway...
 > 
 > 	-mi
 > 
 
 Greeting-
      I tried about 15 alternate superblocks to no good end.  Also there is
 a bug in fsck(8) where it claims that the first alternate superblock is at
 32, and that is where it tells fsck_ufs(8) to look, but at least on my
 5.3 system the first alternate superblock is at 160.  
      There is also another bug in fsck(8) in that it will not pass the 
 -b SUPERBLOCKNUM on to fsck_ufs, instead it returns an error stating that
 -b is an illegal option.  Talk about a way to confuse someone that is new
 to BSD!  Glad I have 20 years working on BSD systems!
 
      I have reinstalled from CD rom and am in the process of reinstalling
 the needed ports.  No user data was lost as I preserved that partation, but
 this problem makes me a little shy to bring the 5.3 system on line.  It was
 to replace a solaris 2.4 running on a 486dx66, but for now the solaris 
 will have to stay.
 
      The raid controler is an IDE unit that requires no special software.
 To the system it looks just like an IDE disk, and you connect 2 same size
 IDE disks to the control unit.  The one I am using in the 5.3 system is an
 older unit called the ARaid99-300.  If you were to want to get one
 today I would instead suggest the units that are made by ARCO in florida.
 They can be found at http://www.arcoide.com/.  I have several of those
 units doing a fine job at various sites.
      BTW the systems shows the raid controler as this:
 
 ad0: 38166MB <ASI ARAID99/Rev1.14> [77545/16/63] at ata0-master UDMA33
 
 at boot time.
 
 -Brett
 
 -- 
 DRM is theft!  We are the stakeholders!  http://www.nyfairuse.org/