2GB limit won't go away

2GB limit won't go away

[Chris Worley]

I'm Running Linux 2.4.10-4GB on SuSE 7.3 with reiserfsprogs-3.x.0j-1
(using 3.5.x disk format...ReiserFS version 3.6.25) atop
lvm-1.0.0.2_rc2-6.  I believe I upgraded reiserfs after a disk crash
last xmas (with a few remnants of that catastrophe still in the system).

I try to "mount -o remount,conv" and put "conv" in the options section
for the root partition in the /etcfstab file.

But,  I can't create a file greater than 2GB.

What am I doing wrong?  I'd really like to accomplish this without
changing lvm or my kernel.

Thanks,

Chris

Re: 2GB limit won't go away

[Chris Worley]

I do have the line:

	/dev/worley/newsys  / reiserfs  conv,defaults 1  1

in my /etc/fstab (and reboot); it doesn't help.  Running:

	# dd if=/dev/zero of=./temp bs=1024 count=5242880

returns:

	dd: writing `./temp': File too large
	2097153+0 records in
	2097152+0 records out

Note that I am using SuSE's hybrid kernel that's shipped w/ 7.3.  I've
seen some messages say this might be the cause of the problem.

Thanks,

Chris
On Sun, 2002-06-09 at 05:32, Rafael E. Herrera wrote:
> Chris Worley wrote:
> > I'm Running Linux 2.4.10-4GB on SuSE 7.3 with reiserfsprogs-3.x.0j-1
> > (using 3.5.x disk format...ReiserFS version 3.6.25) atop
> > lvm-1.0.0.2_rc2-6.  I believe I upgraded reiserfs after a disk crash
> > last xmas (with a few remnants of that catastrophe still in the system).
> > 
> > I try to "mount -o remount,conv" and put "conv" in the options section
> > for the root partition in the /etcfstab file.
> > 
> > But,  I can't create a file greater than 2GB.
> > 
> > What am I doing wrong?  I'd really like to accomplish this without
> > changing lvm or my kernel.
> 
> Use a line similar to this:
> 
> /dev/hdc5       /       reiserfs        defaults,conv 1 2
> 
> Or simply, boot in rescue mode and mount your partition with "mount -o 
> conv".
> 
> -- 
>       Rafael
> 

Re: 2GB limit won't go away

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 392 bytes --]

On Mon, 10 Jun 2002 05:28:53 MDT, Chris Worley <chrisw@lnxi.com>  said:

> But,  I can't create a file greater than 2GB.

Two quick guesses:

1) Old/busticated/crufty libc that doesn't have large-file support
1a) other userspace tools that don't have it (dd, cat, etc)

2) a 'ulimit -f' value for max filesize.
-- 
				Valdis Kletnieks
				Computer Systems Senior Engineer
				Virginia Tech


[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: 2GB limit won't go away

[Chris Worley]

I'm using glibc-2.2.4-21; would that be the problem?

My ulimits:

#  ulimit -a
core file size (blocks)     0
data seg size (kbytes)      unlimited
file size (blocks)          unlimited
max locked memory (kbytes)  unlimited
max memory size (kbytes)    unlimited
open files                  1024
pipe size (512 bytes)       8
stack size (kbytes)         unlimited
cpu time (seconds)          unlimited
max user processes          8192
virtual memory (kbytes)     unlimited

Thanks,

Chris

On Sun, 2002-06-09 at 10:08, Valdis.Kletnieks@vt.edu wrote:
> On Mon, 10 Jun 2002 05:28:53 MDT, Chris Worley <chrisw@lnxi.com>  said:
> 
> > But,  I can't create a file greater than 2GB.
> 
> Two quick guesses:
> 
> 1) Old/busticated/crufty libc that doesn't have large-file support
> 1a) other userspace tools that don't have it (dd, cat, etc)
> 
> 2) a 'ulimit -f' value for max filesize.
> -- 
> 				Valdis Kletnieks
> 				Computer Systems Senior Engineer
> 				Virginia Tech
> 

Re: 2GB limit won't go away

[Andreas Dilger]

On Jun 10, 2002  05:28 -0600, Chris Worley wrote:
> I'm Running Linux 2.4.10-4GB on SuSE 7.3 with reiserfsprogs-3.x.0j-1
> (using 3.5.x disk format...ReiserFS version 3.6.25) atop
   ^^^^^^^^^^^^^^^^^^^^^^^

It is my understanding that you cannot use >2GB files with 3.5 disk
format.  The only way to "upgrade" is to backup restore.  If you
mount with the "conv" mount option, you apparently will be able to
create _new_ files in the 3.6 format (i.e. they can be > 2GB), but
existing files cannot grow past 2GB.

You could "solve" this problem by doing something like:

tar cf - /home | tar xf - -C /tmp
rm -r /home/*
mv /tmp/home/* /home/*
rmdir /tmp/home

Which will re-create all of your files with the 3.6 format "inodes"
and get rid of your old format inodes.  You have to be careful if
you do this, because you run the risk of accidentally deleting all
of your files if you do the wrong thing.  If you have enough free
space in another partition, you may as well just format it with
3.6 format and copy everything over and delete the old partition.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Re: 2GB limit won't go away

[Chris Worley]

It's new files I'm trying to create.  Am I missing something... such as
LVM underneath is screwing something up?  I just want to add to my
existing partition, and not create new partitions.

Thanks,

Chris

On Sun, 2002-06-09 at 12:35, Andreas Dilger wrote:
> On Jun 10, 2002  05:28 -0600, Chris Worley wrote:
> > I'm Running Linux 2.4.10-4GB on SuSE 7.3 with reiserfsprogs-3.x.0j-1
> > (using 3.5.x disk format...ReiserFS version 3.6.25) atop
>    ^^^^^^^^^^^^^^^^^^^^^^^
> 
> It is my understanding that you cannot use >2GB files with 3.5 disk
> format.  The only way to "upgrade" is to backup restore.  If you
> mount with the "conv" mount option, you apparently will be able to
> create _new_ files in the 3.6 format (i.e. they can be > 2GB), but
> existing files cannot grow past 2GB.
> 
> You could "solve" this problem by doing something like:
> 
> tar cf - /home | tar xf - -C /tmp
> rm -r /home/*
> mv /tmp/home/* /home/*
> rmdir /tmp/home
> 
> Which will re-create all of your files with the 3.6 format "inodes"
> and get rid of your old format inodes.  You have to be careful if
> you do this, because you run the risk of accidentally deleting all
> of your files if you do the wrong thing.  If you have enough free
> space in another partition, you may as well just format it with
> 3.6 format and copy everything over and delete the old partition.
> 
> Cheers, Andreas
> --
> Andreas Dilger
> http://www-mddsp.enel.ucalgary.ca/People/adilger/
> http://sourceforge.net/projects/ext2resize/
> 

Re: 2GB limit won't go away

[Oleg Drokin]

Hello!

   The easiest way for you to go is to boot of rescue CD/diskette
   and then mount your root as -o conv
   (btw, SuSE 7.3 does not come with reiserfsprogs-3.x.0j-1)
   Or if you do not use initrd during your normal boot, you can
   issue 'lilo -R "linux rootfsflags=conv"' command and reboot.
   replace linux with your usual lilo target if needed.
   Flags that are passed from /etc/fstab are only applied to rootfs on
   remount and you cannot convert fs on remount.

Bye,
    Oleg   
On Mon, Jun 10, 2002 at 06:10:28PM -0600, Chris Worley wrote:
> It's new files I'm trying to create.  Am I missing something... such as
> LVM underneath is screwing something up?  I just want to add to my
> existing partition, and not create new partitions.
> 
> Thanks,
> 
> Chris
> 
> On Sun, 2002-06-09 at 12:35, Andreas Dilger wrote:
> > On Jun 10, 2002  05:28 -0600, Chris Worley wrote:
> > > I'm Running Linux 2.4.10-4GB on SuSE 7.3 with reiserfsprogs-3.x.0j-1
> > > (using 3.5.x disk format...ReiserFS version 3.6.25) atop
> >    ^^^^^^^^^^^^^^^^^^^^^^^
> > 
> > It is my understanding that you cannot use >2GB files with 3.5 disk
> > format.  The only way to "upgrade" is to backup restore.  If you
> > mount with the "conv" mount option, you apparently will be able to
> > create _new_ files in the 3.6 format (i.e. they can be > 2GB), but
> > existing files cannot grow past 2GB.
> > 
> > You could "solve" this problem by doing something like:
> > 
> > tar cf - /home | tar xf - -C /tmp
> > rm -r /home/*
> > mv /tmp/home/* /home/*
> > rmdir /tmp/home
> > 
> > Which will re-create all of your files with the 3.6 format "inodes"
> > and get rid of your old format inodes.  You have to be careful if
> > you do this, because you run the risk of accidentally deleting all
> > of your files if you do the wrong thing.  If you have enough free
> > space in another partition, you may as well just format it with
> > 3.6 format and copy everything over and delete the old partition.
> > 
> > Cheers, Andreas
> > --
> > Andreas Dilger
> > http://www-mddsp.enel.ucalgary.ca/People/adilger/
> > http://sourceforge.net/projects/ext2resize/
> > 
> 
> 

Re: 2GB limit won't go away

[Chris Worley]

On Sun, 2002-06-09 at 13:16, Oleg Drokin wrote:
> Hello!
> 
>    The easiest way for you to go is to boot of rescue CD/diskette
>    and then mount your root as -o conv
>    (btw, SuSE 7.3 does not come with reiserfsprogs-3.x.0j-1)

I upgraded around December (upon the advice of some folks in the LVM
group) when my root fs when nuts and I had to try to repair it.

>    Or if you do not use initrd during your normal boot, you can

I do an initrd, to load lvm modules and register the lvm devices... so I
changed the linuxrc in the initrd to mount my root fs with the "conv"
option from within my Linuxrc:

	/bin/mount -o conv /dev/worley/newsys /mnt
	/bin/umount /dev/worley/newsys

It fed back the message:

	reiserfs: converting 3.5.x filesystem to the new format

Which made me hopeful that I'd found a solution.

But, I'm still not able to create large files:

	# dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
	dd: writing `./test.bigfile': File too large
	2097153+0 records in
	2097152+0 records out
	# ls -l ./test.bigfile
-rw-r--r--    1 root     root     2147483648 Jun 09 01:17 ./test.bigfile

So, I think the file system has been converted, but something is still
wrong in creating large files.  It's not a problem w/ dd or ulimit:

	# dd if=/dev/zero of=/dev/null bs=1024 count=5242880
	5242880+0 records in
	5242880+0 records out

	# ulimit -a
	core file size (blocks)     0
	data seg size (kbytes)      unlimited
	file size (blocks)          unlimited
	max locked memory (kbytes)  unlimited
	max memory size (kbytes)    unlimited
	open files                  1024
	pipe size (512 bytes)       8
	stack size (kbytes)         unlimited
	cpu time (seconds)          unlimited
	max user processes          8192
	virtual memory (kbytes)     unlimited

Thanks,

Chris
>    issue 'lilo -R "linux rootfsflags=conv"' command and reboot.
>    replace linux with your usual lilo target if needed.
>    Flags that are passed from /etc/fstab are only applied to rootfs on
>    remount and you cannot convert fs on remount.
> 
> Bye,
>     Oleg   
> On Mon, Jun 10, 2002 at 06:10:28PM -0600, Chris Worley wrote:
> > It's new files I'm trying to create.  Am I missing something... such as
> > LVM underneath is screwing something up?  I just want to add to my
> > existing partition, and not create new partitions.
> > 
> > Thanks,
> > 
> > Chris
> > 
> > On Sun, 2002-06-09 at 12:35, Andreas Dilger wrote:
> > > On Jun 10, 2002  05:28 -0600, Chris Worley wrote:
> > > > I'm Running Linux 2.4.10-4GB on SuSE 7.3 with reiserfsprogs-3.x.0j-1
> > > > (using 3.5.x disk format...ReiserFS version 3.6.25) atop
> > >    ^^^^^^^^^^^^^^^^^^^^^^^
> > > 
> > > It is my understanding that you cannot use >2GB files with 3.5 disk
> > > format.  The only way to "upgrade" is to backup restore.  If you
> > > mount with the "conv" mount option, you apparently will be able to
> > > create _new_ files in the 3.6 format (i.e. they can be > 2GB), but
> > > existing files cannot grow past 2GB.
> > > 
> > > You could "solve" this problem by doing something like:
> > > 
> > > tar cf - /home | tar xf - -C /tmp
> > > rm -r /home/*
> > > mv /tmp/home/* /home/*
> > > rmdir /tmp/home
> > > 
> > > Which will re-create all of your files with the 3.6 format "inodes"
> > > and get rid of your old format inodes.  You have to be careful if
> > > you do this, because you run the risk of accidentally deleting all
> > > of your files if you do the wrong thing.  If you have enough free
> > > space in another partition, you may as well just format it with
> > > 3.6 format and copy everything over and delete the old partition.
> > > 
> > > Cheers, Andreas
> > > --
> > > Andreas Dilger
> > > http://www-mddsp.enel.ucalgary.ca/People/adilger/
> > > http://sourceforge.net/projects/ext2resize/
> > > 
> > 
> > 
> 

Re: 2GB limit won't go away

[Oleg Drokin]

Hello!

On Tue, Jun 11, 2002 at 01:27:27AM -0600, Chris Worley wrote:

> I do an initrd, to load lvm modules and register the lvm devices... so I
> changed the linuxrc in the initrd to mount my root fs with the "conv"
> option from within my Linuxrc:
> 	/bin/mount -o conv /dev/worley/newsys /mnt
> 	/bin/umount /dev/worley/newsys

That would work.

> It fed back the message:
> 	reiserfs: converting 3.5.x filesystem to the new format

That confirms that everything went fine.

> But, I'm still not able to create large files:
> 	# dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> 	dd: writing `./test.bigfile': File too large
> 	2097153+0 records in
> 	2097152+0 records out
> 	# ls -l ./test.bigfile
> -rw-r--r--    1 root     root     2147483648 Jun 09 01:17 ./test.bigfile

Is this file located on the converted volume (/dev/worley/newsys)?
Would following sequence of commans help?
rm -rf ./test.bigfile ; dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880

Bye,
    Oleg

Re: 2GB limit won't go away

[Andreas Dilger]

On Jun 10, 2002  08:58 +0400, Oleg Drokin wrote:
> Would following sequence of commans help?
> dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880

Just as a tip for LFS testers, it is much easier to test with a sparse
file instead of waiting for 2GB to be written to disk:

dd if=/dev/zero of=./test.bigfile bs=1k seek=1023M

(this puts you just short of the 2GB limit in case your seek is broken).
Note that you should probably strace the failing "dd" command to ensure
it is opening the file with O_LARGEFILE.  If not, it isn't a kernel
problem but a tool problem instead.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Re: 2GB limit won't go away

[Chris Worley]

On Sun, 2002-06-09 at 23:42, Andreas Dilger wrote:
> Note that you should probably strace the failing "dd" command to ensure
> it is opening the file with O_LARGEFILE.  If not, it isn't a kernel
> problem but a tool problem instead.

# strace dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
...
open("./test.bigfile", O_WRONLY|O_CREAT|O_TRUNC|O_LARGEFILE, 0666) = 1
...

But, it was a good suggestion.  The dd/seek does speed testing.

Thanks,

Chris

> On Jun 10, 2002  08:58 +0400, Oleg Drokin wrote:
> > Would following sequence of commans help?
> > dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> 
> Just as a tip for LFS testers, it is much easier to test with a sparse
> file instead of waiting for 2GB to be written to disk:
> 
> dd if=/dev/zero of=./test.bigfile bs=1k seek=1023M
> 
> (this puts you just short of the 2GB limit in case your seek is broken).

Re: 2GB limit won't go away

[Chris Worley]

On Sun, 2002-06-09 at 22:58, Oleg Drokin wrote: 
> > -rw-r--r--    1 root     root     2147483648 Jun 09 01:17 ./test.bigfile
> 
> Is this file located on the converted volume (/dev/worley/newsys)?
> Would following sequence of commans help?
> rm -rf ./test.bigfile ; dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880

Everything but /boot is on that volume (and /boot is mounted ro):

# df
Filesystem           1k-blocks      Used Available Use% Mounted on
/dev/worley/newsys    78763672  73604064   5159608  94% /
/dev/hda1                19976     18114       831  96% /boot
shmfs                   514732         0    514732   0% /dev/shm

And deleting it first doesn't help.

I'm wondering if there is a kernel/reiser tools mismatch.  How could I verify what's in the kernel?

Thanks,

Chris

On Sun, 2002-06-09 at 22:58, Oleg Drokin wrote: 
> Hello!
> 
> On Tue, Jun 11, 2002 at 01:27:27AM -0600, Chris Worley wrote:
> 
> > I do an initrd, to load lvm modules and register the lvm devices... so I
> > changed the linuxrc in the initrd to mount my root fs with the "conv"
> > option from within my Linuxrc:
> > 	/bin/mount -o conv /dev/worley/newsys /mnt
> > 	/bin/umount /dev/worley/newsys
> 
> That would work.
> 
> > It fed back the message:
> > 	reiserfs: converting 3.5.x filesystem to the new format
> 
> That confirms that everything went fine.
> 
> > But, I'm still not able to create large files:
> > 	# dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> > 	dd: writing `./test.bigfile': File too large
> > 	2097153+0 records in
> > 	2097152+0 records out
> > 	# ls -l ./test.bigfile
> > -rw-r--r--    1 root     root     2147483648 Jun 09 01:17 ./test.bigfile
> 
> Is this file located on the converted volume (/dev/worley/newsys)?
> Would following sequence of commans help?
> rm -rf ./test.bigfile ; dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> 
> Bye,
>     Oleg
> 

Re: 2GB limit won't go away

[Chris Worley]

On Mon, 2002-06-10 at 08:05, Chris Worley wrote:
> On Sun, 2002-06-09 at 22:58, Oleg Drokin wrote: 
> > > -rw-r--r--    1 root     root     2147483648 Jun 09 01:17 ./test.bigfile
> > 
> > Is this file located on the converted volume (/dev/worley/newsys)?
> > Would following sequence of commans help?
> > rm -rf ./test.bigfile ; dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> 

I forgot to add, the output is:

# rm test.bigfile ;dd if=/dev/zero of=./test.bigfile bs=1k seek=1023M
dd: advancing past 1098437885952 bytes in output file `./test.bigfile':
File too large

> Everything but /boot is on that volume (and /boot is mounted ro):
> 
> # df
> Filesystem           1k-blocks      Used Available Use% Mounted on
> /dev/worley/newsys    78763672  73604064   5159608  94% /
> /dev/hda1                19976     18114       831  96% /boot
> shmfs                   514732         0    514732   0% /dev/shm
> 
> And deleting it first doesn't help.
> 
> I'm wondering if there is a kernel/reiser tools mismatch.  How could I verify what's in the kernel?
> 
> Thanks,
> 
> Chris
> 
> On Sun, 2002-06-09 at 22:58, Oleg Drokin wrote: 
> > Hello!
> > 
> > On Tue, Jun 11, 2002 at 01:27:27AM -0600, Chris Worley wrote:
> > 
> > > I do an initrd, to load lvm modules and register the lvm devices... so I
> > > changed the linuxrc in the initrd to mount my root fs with the "conv"
> > > option from within my Linuxrc:
> > > 	/bin/mount -o conv /dev/worley/newsys /mnt
> > > 	/bin/umount /dev/worley/newsys
> > 
> > That would work.
> > 
> > > It fed back the message:
> > > 	reiserfs: converting 3.5.x filesystem to the new format
> > 
> > That confirms that everything went fine.
> > 
> > > But, I'm still not able to create large files:
> > > 	# dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> > > 	dd: writing `./test.bigfile': File too large
> > > 	2097153+0 records in
> > > 	2097152+0 records out
> > > 	# ls -l ./test.bigfile
> > > -rw-r--r--    1 root     root     2147483648 Jun 09 01:17 ./test.bigfile
> > 
> > Is this file located on the converted volume (/dev/worley/newsys)?
> > Would following sequence of commans help?
> > rm -rf ./test.bigfile ; dd if=/dev/zero of=./test.bigfile bs=1024 count=5242880
> > 
> > Bye,
> >     Oleg
> > 
> 

Re: 2GB limit won't go away

[Andreas Dilger]

On Jun 10, 2002  08:09 -0600, Chris Worley wrote:
> # rm test.bigfile ;dd if=/dev/zero of=./test.bigfile bs=1k seek=1023M
> dd: advancing past 1098437885952 bytes in output file `./test.bigfile':
> File too large

OK, my bad.  This should have been:
dd if=/dev/zero of=./test.bigfile bs=1k seek=2047k

Even so, this probably won't fix your problem.  I would suggest creating
a test 3.6 format filesystem (loopback or such) and see if that works.
With the seek= argument you don't need a very big filesystem.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Re: 2GB limit won't go away

[Oleg Drokin]

Hello!

On Mon, Jun 10, 2002 at 08:05:54AM -0600, Chris Worley wrote:

> I'm wondering if there is a kernel/reiser tools mismatch.  How could I verify what's in the kernel?

Do a dmesg and see throught the output
if you see somethink like this:
reiserfs: checking transaction log (device 03:02) ...
Using r5 hash to sort names
reiserfs: using 3.5.x disk format
ReiserFS version 3.6.25
VFS: Mounted root (reiserfs filesystem) readonly.

Then yuo have old format.
If you do not see "reiserfs: using 3.5.x disk format" part,
you are using 3.6 format.

Bye,
    Oleg