[Bug-tar] "unexpected field value in snapshot file" error (bug #184870)

[Nathan Stratton Treadway]

Since upgrading to GNU tar 1.16, we've been getting "Unexpected field
value in snapshot file" errors from our backup scripts on a few of our
machines.

After some investigation I found bug 18487 in the Savannah bug tracker,
which seems to apply to our situation:
  http://savannah.gnu.org/bugs/?18487

I searched the web and gnu-tar list archives for any further discussion
of this problem but didn't find any (and don't see anything applicable
mentioned in the tar 1.18 NEWS file), so I am providing some addition
information in hopes that it is helpful.

I was able to reproduce the problem on one of the affected machines
using a simple test case:

# mkdir temp temp/proc temp/testdir
# mount -tproc proc temp/proc/
# tar --create --file ~/tar_test.tar --one-file-system \
  --listed-incremental ~/tar_incremental.snar  temp
# tar --create --file ~/tar_test-2.tar  --one-file-system \
  --listed-incremental ~/tar_incremental.snar  temp
tar: Unexpected field value in snapshot file
tar: Error is not recoverable: exiting now

As described in the original bug report, the problem appears to be in
the timestamp field for the "proc" entry in the snapshot file.  (If I
unmount temp/proc, delete the snapshot file, and repeat the tar
commands, the second run does not produce any error.)

Here is what the snapshot file looks like (as displayed by "less"):

# less ~/tar_incremental.snar
GNU tar-1.16-2
address@hidden@address@hidden@address@hidden@address@hidden@address@hidden
address@hidden@address@hidden@address@hidden@address@hidden@temp/address@hidden@address@hidden@1186159126^@
address@hidden@address@hidden/address@hidden@^@

(Note the "18446744071852258736" value which appears in the temp/proc
entry.)

I noticed that if I run the "stat" command on the contents of temp, the
proc entry's timestamps (the decimal portion of the seconds) look wrong,
too:

# stat temp/*
  File: emp/proc'
  Size: 0               Blocks: 0          IO Block: 1024   directory
Device: 3h/3d   Inode: 1           Links: 144
Access: (0555/dr-xr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2007-05-01 10:15:14.-1857292880 
Modify: 2007-05-01 10:15:14.-1857292880 
Change: 2007-05-01 10:15:14.-1857292880 
  File: emp/testdir'
  Size: 48              Blocks: 0          IO Block: 131072 directory
Device: 902h/2306d      Inode: 23900       Links: 2
Access: (0755/drwxr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2007-08-03 13:49:54.011572437 -0400
Modify: 2007-08-03 12:38:46.952905198 -0400
Change: 2007-08-03 12:38:46.952905198 -0400


So it would seem that there is something wrong with the "proc"
filesystem itself.  I'm running linux 2.6.8 (Debian 2.6.8-2-386
2.6.8-16sarge1 package) on this particular machine, but I am not sure if
that's (only) trigger for this problem.  (We've noticed this problem on
two other machines that are also running some version of 2.6.8, but also
have machines running 2.6.8 that aren't showing this problem.)

However, it seems like even though the timestamp on "proc" isn't
correct, tar shouldn't be creating a snapshot file which it then
considers to be corrupt when reading it.  Could it be updated (either
the routines creating the file or those reading it) to handle this
situation more gracefully?

Let me know if I can provide any further information on this issue.

Thanks.

                                                Nathan


----------------------------------------------------------------------------
Nathan Stratton Treadway  -  address@hidden  -  Mid-Atlantic region
Ray Ontko & Co.  -  Software consulting services  -   http://www.ontko.com/
 GPG Key: http://www.ontko.com/~nathanst/gpg_key.txt   ID: 1023D/ECFB6239
 Key fingerprint = 6AD8 485E 20B9 5C71 231C  0C32 15F3 ADCD ECFB 6239

signature.asc
Description: Digital signature