2.6.0test-1 error while writing files to loopback UDF filesystem UDF-fs DEBUG fs/udf/balloc.c:192:udf_
Hi,
I created an UDF filesystem (dd of=file if=... && mkudffs file && mount -o loop -t udf /mnt) and then added some files to it (tar cf - * | (cd /mnt ; tar xvpf -)).
That went well for a while, but after aprox 2GB (beware: no file was longer then +/- 1GB), I got these errors in syslog:
Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:192:udf_ bitmap_free_blocks: bit 3125 already set Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:193:udf_ bitmap_free_blocks: byte=20 Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:192:udf_ bitmap_free_blocks: bit 3125 already set Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:193:udf_
bitmap_free_blocks: byte=60
etc.
I then did a compare (cmp -l) and found that the copied file was different from the original one, so it seems something is going wrong while writing to the UDF filesystem.
As I wrote in the subjectline, I'm using 2.6.0-test1.
Folkert van Heusden
+--------------------------------------------------------------------------+ | UNIX sysop? Then give MultiTail ( http://www.vanheusden.com/multitail/ ) | | a try, it brings monitoring logfiles (and such) to a different level! | +---------------------------------------------------= www.vanheusden.com =-+
Re: 2.6.0test-1 error while writing files to loopback UDF filesystem UDF-fs DEBUG fs/udf/balloc.c:192:udf_ still with 2.6.0-test4!
Hi,
Just retested it with 2.6.0-test4.
This time, the kernel did *NOT* log any of the messages as it did with test1,
BUT: if I compare the original with the new file, I get differences all over the place.
So what I did was (just in case):
# create backup image truncate /data2/backup.udf 4294967296 mkudffs /data2/backup.udf
# mount it & copy data mount -o loop -t udf /data2/backup.udf /mnt (cd /data/backup/Backup ; tar cf - *) | (cd /mnt ; tar xvf -) umount /mnt
# verify things mount -o loop -t udf /data2/backup.udf /mnt cd /mnt find . -type f |xargs -n 1 ./docmp
docmp is:
--------
echo $1
cmp -l /mnt/$1 /data/backup/Backup/$1
Quoted 13 lines · Folkert van Heusden
Hi,
I created an UDF filesystem (dd of=file if=... && mkudffs file && mount -o loop -t udf /mnt) and then added some files to it (tar cf - * | (cd /mnt ; tar xvpf -)).
That went well for a while, but after aprox 2GB (beware: no file was longer then +/- 1GB), I got these errors in syslog:Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:192:udf_ bitmap_free_blocks: bit 3125 already set Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:193:udf_ bitmap_free_blocks: byte=20 Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:192:udf_ bitmap_free_blocks: bit 3125 already set Sep 14 00:04:38 boemboem kernel: UDF-fs DEBUG fs/udf/balloc.c:193:udf_
Quoted 6 lines
bitmap_free_blocks: byte=60
etc.
I then did a compare (cmp -l) and found that the copied file was different from the original one, so it seems something is going wrong while writing to the UDF filesystem.
As I wrote in the subjectline, I'm using 2.6.0-test1.
Folkert van Heusden
p.s.: truncate is an utility which is available on most BSD systems (not mac
os x which is a BSD). I wrote a version for linux:
http://vanheusden.com/Linux/truncate-0.1.tgz (64bit safe)
+--------------------------------------------------------------------------+ | UNIX sysop? Then give MultiTail ( http://www.vanheusden.com/multitail/ ) | | a try, it brings monitoring logfiles (and such) to a different level! | +---------------------------------------------------= www.vanheusden.com =-+