fsck_ufs: cannot alloc 647172276 bytes for inoinfo

[Don Lewis]

On 18 May, Scott Long wrote:
> 
> Einstein Oliveira wrote:
>> Hi,
>> 
>> I'm using FreeBSD 5.4-RELEASE-p1 and I found some messages about a 
>> problem like this in freebsd-current a few months ago.
>> 
>> The fact is that this problem has just ocurred (on 5.4-RELEASE-p1), 
>> probably because of a power outage (I didn't find anything in logs that 
>> would cause a forced reboot).
>> 
>> Here is some information:
>> 
>> [fsck]
>> 
>> # fsck /usr
>> ** /dev/ad0s1f
>> ** Last Mounted on /usr
>> ** Phase 1 - Check Blocks and Sizes
>> fsck_ufs: cannot alloc 647172276 bytes for inoinfo
>> 
>> [dumpfs]
>> 
>> # dumpfs /usr
>> ....
>> cg 11:
>> magic   806e54b3        tell    7e584000        time    Fri Nov 26 
>> 01:12:38 1943
>> cgx     -1312777034     ndblk   -931553057      niblk   -2017999697 
>> initiblk -1985690579
>> nbfree  2120952272      ndir    831365510       nifree  93179446 nffree  
>> -1130132161
>> rotor   -541929111      irotor  1807500773      frotor  -1630357508
>> frsum   237144368       -1927321463     -1872608999     -481058689 
>> 811654083       830922798       -91646688
>> sum of frsum: 1537687372
>> Segmentation fault (core dumped)
>> 
>> 
>> I found in the discussion mentioned above
>> 
>> http://www.freebsd.org/cgi/getmsg.cgi?fetch=195794+200562+/usr/local/www/db/text/2005/freebsd-current/20050417.freebsd-current 
>> 
>> 
>> this possible solution:
>> 
>>> At line 92 in src/sbin/fsck_ffs/pass1.c, you should see the following
>>> block of code:
>>>
>>>         for (c = 0; c < sblock.fs_ncg; c++) {
>>>                 inumber = c * sblock.fs_ipg;
>>>                 setinodebuf(inumber);
>>>                 getblk(&cgblk, cgtod(&sblock, c), sblock.fs_cgsize);
>>>                 if (sblock.fs_magic == FS_UFS2_MAGIC)
>>>                         inosused = cgrp.cg_initediblk;
>>>                 else
>>>                         inosused = sblock.fs_ipg;
>>>
>>> Try changing
>>>     inosused = cgrp.cg_initediblk;
>>> to
>>>     inosused = (cgrp.cg_initediblk <= sblock.fs_ipg) ?
>>>         cgrp.cg_initediblk : sblock.fs_ipg;
>> 
>> 
>> but it doesn't solve the problem.
>> 
>> Any ideas ?
>> 
> 
> Hi,
> 
> This unfortunately looks to be a case of filesystem damage that fsck 
> can't cope with.  Power failures are especially brutal on ATA disks
> because of their write cache and they way they optimize writes.  You
> might try something like ffsinfo to locate a alternate superblock that
> is more sane.

In this case, the problem isn't the superblock, but one of the cylinder
group blocks.  UFS2 records the number of inodes that have been
initialized in each cylinder group in the cylinder group blocks so that
newfs doesn't have to initialize all of them when the file system is
created.  Fsck scans all the cylinder group blocks and allocates space
for all the inodes that might be valid.

The cylinder group block in question is pretty clearly hosed, but it
should be possible to get past this problem.  Loss of some or all of the
contents of that cylinder group, data corruption, and lots of files
getting moved to lost+found are all likely, but it should still be
possible to salvage a lot of the contents of the file system.