tar: 200MB vfat system allegedly changed mtime in one second!

[Chris Karakas]

Hello,

I admit that the headline looks like coming from the yellow press ;-)

I think I have bumped onto a tar bug. I did this while doing AMANDA
backups on a vfat filesystem mounted as /dos/d. I have hacked AMANDA a
very little bit to let her pass the parameters --incremental
--newer-mtime to tar, since --listed-incremental does not work on vfat
due to non-constant inodes since kernel version 2.2.5 (I think). This is
SuSE Linux 6.4, kernel 2.2.14, tar 1.13.18, AMANDA 2.4.1p1-something. I
am sending this to both amanda-users and bug-tar lists. 

Usually, passing --incremental --newer-mtime to tar seems to work well
for my case and I was about to send a report to amanda-users. But look
at this, that I discovered today:

Suddenly, on Jan 6, 2001 between  06:28:14 GMT and 06:28:15 GMT
(i.e. in only one second!) 236MB of /dos/d (which is 253MB total 
in size, i.e. almost all of /dos/d !) changed its modification time!
Look at the following output :

bacchus:~ # /bin/gtar --create --directory /dos/d --incremental
--newer-mtime "2001-01-06  06:28:14 GMT" --sparse --one-file-system
--ignore-failed-read --totals --file /dev/null .
Total bytes written: 247214080 (236MB, 39MB/s)

bacchus:~ # /bin/gtar --create --directory /dos/d --incremental
--newer-mtime "2001-01-06  06:28:15 GMT" --sparse --one-file-system
--ignore-failed-read --totals --file /dev/null .
Total bytes written: 286720 (280kB, 93kB/s)

Of course, AMANDA has correctly backed up a level 1 of 236+ MB...
But why should this one filesystem change mtime almost as a whole
in just one second? I am clueless. 

I have checked the mtimes with "ls -l" for some of the files that tar
would archive in the 236MB file. They had _not_ changed. I checked the
ctime with "ls -lc" and I found that _a_lot_ of files had a ctime in the
future! But in the 236MB archive there were files with ctime in the
past,
as well as in the future - and anyway this is correct, since I gave the
option --newer-mtime to tar.

I checked the logs. The only logged operation within 3 minutes before 
06:28:14 GMT is a mke2fs on /dev/sdd (MO disk, 2048 bytes sector media)
with the necessary "syncing disks". But we're talking about /dos/d,
which is
on /dev/hda... 

As I said, I consider this to be a bug.

Any ideas?

-- 
Regards

Chris Karakas

don't waste your cpu time, search for an OGR: http://www.distributed.net