-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Wednesday 2007-10-24 at 09:59 -0400, Michael Kershaw wrote:

On 10/24/07, Ciro Iriarte <> wrote:

eh?

The email is there, complete; if you can't read it complain to gmail. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFHH1Y/tTMYHG2NR9URAv+kAJ40MbSljvEcHgu2xmnDjfzsCGLE5QCfcB4N 8LLpFarHehl/ZXtVjY1kKhw= =sUnD -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Rikard Johnels wrote:

On Wednesday 24 October 2007 15:48, Ciro Iriarte wrote:

...

Hi, just found a weird behavior on 10.3 (didn't happen on 10.1), when i have little space it directly tells me that there's no space left.... i'm using reiserfs on those fs...

mainwks:~/download> df -h /home/ /srv/ftp/ S.ficheros Tamaño Usado Disp Uso% Montado en /dev/mapper/system-home 32G 32G 130M 100% /home /dev/mapper/system-ftp 15G 15G 236M 99% /srv/ftp

<snip for trim>

I seem to recall the system reserving a certain amount of space to enable root to login in case of a filled system. Or was that only on a ext2 filesystem?

That's on ALL Unix and Linux systems that I've ever used. Once disk usage goes beyond a threshold (set individually in each filesystem layout on each partition at filesystem creation time), only root can write to the filesystem. Any filesystem (ext3, xfs, reiserfs, etc) which doesn't have this capability cannot be a general purpose Unix or Linux filesystem because it cannot be used on whatever filesystem(s) (i.e partition) hold, for example, /tmp, /var/log, /var/tmp, and wherever root's home directory happens to be.

Also, what are you trying to do? Make a small (asy 4k) file, or something bigger? How about inodes? Are you out of those? Do a 'df -i' to check.

-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Ciro Iriarte wrote:

2007/10/24, Aaron Kulkis :

...

Rikard Johnels wrote:

...

On Wednesday 24 October 2007 15:48, Ciro Iriarte wrote:

...

Hi, just found a weird behavior on 10.3 (didn't happen on 10.1), when i have little space it directly tells me that there's no space left.... i'm using reiserfs on those fs...

mainwks:~/download> df -h /home/ /srv/ftp/ S.ficheros Tamaño Usado Disp Uso% Montado en /dev/mapper/system-home 32G 32G 130M 100% /home /dev/mapper/system-ftp 15G 15G 236M 99% /srv/ftp

<snip for trim>

I seem to recall the system reserving a certain amount of space to enable root to login in case of a filled system. Or was that only on a ext2 filesystem? That's on ALL Unix and Linux systems that I've ever used. Once disk usage goes beyond a threshold (set individually in each filesystem layout on each partition at filesystem creation time), only root can write to the filesystem.

Any filesystem (ext3, xfs, reiserfs, etc) which doesn't have this capability cannot be a general purpose Unix or Linux filesystem because it cannot be used on whatever filesystem(s) (i.e partition) hold, for example, /tmp, /var/log, /var/tmp, and wherever root's home directory happens to be.

...

Also, what are you trying to do? Make a small (asy 4k) file, or something bigger? How about inodes? Are you out of those? Do a 'df -i' to check.

That doesn't apply to reiserfs, it does to ext3 and ufs for example but you can set the reserved percentage to 0 (with tune2fs on ext3). And about the inodes, the total quantity is not defined at the fs creation time.

Ok, I just subscribed to this list, so this is the first message I have in my inbox on this topic. I've read the rest on the archive. I do regularly read opensuse-kernel, though. First to start with the questions from the bottom up: ReiserFS doesn't reserve space for anything other than the journal. ReiserFS doesn't have inodes. It has items that are referenced by keys. The only thing there is a "shortage" of is objectids, which are regular 32-bit integers. Unless you've managed to create ~ 4 billion files, you're not running out of them. This bug is likely caused by a patch I put into 10.3 that started using the first_zero_hint value that is calculated in the bitmap code. The calculations have been there for ages, but they haven't been used until the 10.3 kernel. It short circuits the bitmap scanning code to skip ranges it knows are already used. After the 10.3 release, I decided this was a dubious optimization and was likely responsible for problems just like this one (bug 331814). I pushed a patch to mainline that rips the first_zero_hint code out entirely. To test this hypothesis, please download and test a kernel from: ftp://ftp.suse.com/pub/people/jeffm/suse/testpkgs/331814 I'm very much interested in feedback to ensure that this solves the problem. Thanks. - -Jeff - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org iD8DBQFHH8A6LPWxlyuTD7IRAh4QAJoChfifpSZjMSBvP1k9/Xkur04TIQCeJ5qt 6RFgH+hF+kGPkXPqymqm8xA= =/E71 -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Ciro Iriarte wrote:

2007/10/24, Jeff Mahoney :

...

Ok, I just subscribed to this list, so this is the first message I have in my inbox on this topic. I've read the rest on the archive. I do regularly read opensuse-kernel, though.

First to start with the questions from the bottom up:

ReiserFS doesn't reserve space for anything other than the journal.

ReiserFS doesn't have inodes. It has items that are referenced by keys. The only thing there is a "shortage" of is objectids, which are regular 32-bit integers. Unless you've managed to create ~ 4 billion files, you're not running out of them.

This bug is likely caused by a patch I put into 10.3 that started using the first_zero_hint value that is calculated in the bitmap code. The calculations have been there for ages, but they haven't been used until the 10.3 kernel. It short circuits the bitmap scanning code to skip ranges it knows are already used. After the 10.3 release, I decided this was a dubious optimization and was likely responsible for problems just like this one (bug 331814). I pushed a patch to mainline that rips the first_zero_hint code out entirely.

To test this hypothesis, please download and test a kernel from: ftp://ftp.suse.com/pub/people/jeffm/suse/testpkgs/331814

I'm very much interested in feedback to ensure that this solves the problem.

Thanks.

- -Jeff

- -- Jeff Mahoney SUSE Labs

Installed the test kernel but all my reiserfs are read only now...

FS: ---------------------------------------------- mainwks:~> df -h S.ficheros Tamaño Usado Disp Uso% Montado en /dev/mapper/system-root 6,0G 1004M 4,7G 18% / udev 1007M 120K 1007M 1% /dev /dev/md0 99M 27M 68M 28% /boot /dev/mapper/system-datos 151G 151G 755M 100% /datos /dev/mapper/system-home 32G 24G 8,1G 75% /home /dev/mapper/system-ftp 15G 15G 306M 99% /srv/ftp /dev/mapper/system-usr 5,0G 3,3G 1,4G 71% /usr /dev/mapper/system-var 2,0G 417M 1,5G 22% /var /dev/mapper/system-vmware 20G 19G 1,8G 92% /var/lib/vmware ----------------------------------------------

Mounted as follow:

---------------------------- mainwks:~> mount /dev/mapper/system-root on / type ext3 (rw,acl,user_xattr) proc on /proc type proc (rw) sysfs on /sys type sysfs (rw) debugfs on /sys/kernel/debug type debugfs (rw) udev on /dev type tmpfs (rw) devpts on /dev/pts type devpts (rw,mode=0620,gid=5) /dev/md0 on /boot type ext3 (rw,acl,user_xattr) /dev/mapper/system-datos on /datos type reiserfs (rw) /dev/mapper/system-home on /home type reiserfs (rw) /dev/mapper/system-ftp on /srv/ftp type reiserfs (rw) /dev/mapper/system-usr on /usr type ext3 (rw,acl,user_xattr) /dev/mapper/system-var on /var type ext3 (rw,acl,user_xattr) /dev/mapper/system-vmware on /var/lib/vmware type reiserfs (rw) nfsd on /proc/fs/nfsd type nfsd (rw) rpc_pipefs on /var/lib/nfs/rpc_pipefs type rpc_pipefs (rw) -----------------------

Read only messages:

--------------------- ciro@mainwks:~> echo "dosaokdsa" > hbla bash: hbla: Sistema de ficheros de sólo lectura ---------------------

No space left messages:

--------------------- ciro@mainwks:~> echo sdad > /datos/peliculas/test bash: echo: write error: No queda espacio en el dispositivo

ciro@mainwks:~> echo "dosaokdsa" > /srv/ftp/update/10.3/sadad bash: echo: write error: No queda espacio en el dispositivo ---------------------

Can you send me the output from dmesg? The file system must have aborted for some reason. - -Jeff - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org iD8DBQFHIiMnLPWxlyuTD7IRAixWAKCSISxSJbuXXsRroWL4V3hBlK/WowCfe+F9 EA/3286Ua1Sqo5Xv8ETGyWQ= =5FhD -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Ciro Iriarte wrote:

...

...

...

2007/10/24, Jeff Mahoney :

...

Ok, I just subscribed to this list, so this is the first message I have in my inbox on this topic. I've read the rest on the archive. I do regularly read opensuse-kernel, though.

First to start with the questions from the bottom up:

ReiserFS doesn't reserve space for anything other than the journal.

ReiserFS doesn't have inodes. It has items that are referenced by keys. The only thing there is a "shortage" of is objectids, which are regular 32-bit integers. Unless you've managed to create ~ 4 billion files, you're not running out of them.

This bug is likely caused by a patch I put into 10.3 that started using the first_zero_hint value that is calculated in the bitmap code. The calculations have been there for ages, but they haven't been used until the 10.3 kernel. It short circuits the bitmap scanning code to skip ranges it knows are already used. After the 10.3 release, I decided this was a dubious optimization and was likely responsible for problems just like this one (bug 331814). I pushed a patch to mainline that rips the first_zero_hint code out entirely.

To test this hypothesis, please download and test a kernel from: ftp://ftp.suse.com/pub/people/jeffm/suse/testpkgs/331814

I'm very much interested in feedback to ensure that this solves the

2007/10/26, Jeff Mahoney : Ciro Iriarte wrote: problem.

...

...

...

...

Thanks.

- -Jeff

- -- Jeff Mahoney SUSE Labs Installed the test kernel but all my reiserfs are read only now...

FS: ---------------------------------------------- mainwks:~> df -h S.ficheros Tamaño Usado Disp Uso% Montado en /dev/mapper/system-root 6,0G 1004M 4,7G 18% / udev 1007M 120K 1007M 1% /dev /dev/md0 99M 27M 68M 28% /boot /dev/mapper/system-datos 151G 151G 755M 100% /datos /dev/mapper/system-home 32G 24G 8,1G 75% /home /dev/mapper/system-ftp 15G 15G 306M 99% /srv/ftp /dev/mapper/system-usr 5,0G 3,3G 1,4G 71% /usr /dev/mapper/system-var 2,0G 417M 1,5G 22% /var /dev/mapper/system-vmware 20G 19G 1,8G 92% /var/lib/vmware ----------------------------------------------

Mounted as follow:

---------------------------- mainwks:~> mount /dev/mapper/system-root on / type ext3 (rw,acl,user_xattr) proc on /proc type proc (rw) sysfs on /sys type sysfs (rw) debugfs on /sys/kernel/debug type debugfs (rw) udev on /dev type tmpfs (rw) devpts on /dev/pts type devpts (rw,mode=0620,gid=5) /dev/md0 on /boot type ext3 (rw,acl,user_xattr) /dev/mapper/system-datos on /datos type reiserfs (rw) /dev/mapper/system-home on /home type reiserfs (rw) /dev/mapper/system-ftp on /srv/ftp type reiserfs (rw) /dev/mapper/system-usr on /usr type ext3 (rw,acl,user_xattr) /dev/mapper/system-var on /var type ext3 (rw,acl,user_xattr) /dev/mapper/system-vmware on /var/lib/vmware type reiserfs (rw) nfsd on /proc/fs/nfsd type nfsd (rw) rpc_pipefs on /var/lib/nfs/rpc_pipefs type rpc_pipefs (rw) -----------------------

Read only messages:

--------------------- ciro@mainwks:~> echo "dosaokdsa" > hbla bash: hbla: Sistema de ficheros de sólo lectura ---------------------

No space left messages:

--------------------- ciro@mainwks:~> echo sdad > /datos/peliculas/test bash: echo: write error: No queda espacio en el dispositivo

ciro@mainwks:~> echo "dosaokdsa" > /srv/ftp/update/10.3/sadad bash: echo: write error: No queda espacio en el dispositivo --------------------- Can you send me the output from dmesg? The file system must have aborted for some reason.

Ok, I think I really have it this time. That RPM had an old version of the patch that is in the HEAD kernel. It was missing a memset 0 -> 0xff change that would cause the problems you're describing. I'm building an update RPM now, and it'll be posted shortly. - -Jeff - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org iD8DBQFHJgUhLPWxlyuTD7IRAg8KAJsHMudvBUZ5Qy38K64N5Bs184MeWwCgj7yR 4nagiovRHJpxNGexxzs7pjE= =K82I -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Friday 2007-10-26 at 14:20 -0300, Ciro Iriarte wrote:

Read only messages:

--------------------- ciro@mainwks:~> echo "dosaokdsa" > hbla bash: hbla: Sistema de ficheros de sólo lectura ---------------------

Ciro, just a detail. Just to help others understand the problems, you know you can temporarily tell the command line to output messages in English instead of Spanish like this: LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8 command In fact, I use a similar trick to get them in Spanish: LANG=es_ES.UTF-8 LC_ALL=es_ES.UTF-8 command - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFHIjnNtTMYHG2NR9URAjmlAJ4kgnb/hYpINYua0aTsiTUjtJl3VgCcCCfM iGKcrS0YXnrb3Z5vXbPI5/0= =Nn4x -----END PGP SIGNATURE-----

On 10/24/07, Aaron Kulkis wrote:

Rikard Johnels wrote:

...

On Wednesday 24 October 2007 15:48, Ciro Iriarte wrote:

...

Hi, just found a weird behavior on 10.3 (didn't happen on 10.1), when i have little space it directly tells me that there's no space left.... i'm using reiserfs on those fs...

mainwks:~/download> df -h /home/ /srv/ftp/ S.ficheros Tamaño Usado Disp Uso% Montado en /dev/mapper/system-home 32G 32G 130M 100% /home /dev/mapper/system-ftp 15G 15G 236M 99% /srv/ftp

<snip for trim>

I seem to recall the system reserving a certain amount of space to enable root to login in case of a filled system. Or was that only on a ext2 filesystem?

That's on ALL Unix and Linux systems that I've ever used. Once disk usage goes beyond a threshold (set individually in each filesystem layout on each partition at filesystem creation time), only root can write to the filesystem.

Any filesystem (ext3, xfs, reiserfs, etc) which doesn't have this capability cannot be a general purpose Unix or Linux filesystem because it cannot be used on whatever filesystem(s) (i.e partition) hold, for example, /tmp, /var/log, /var/tmp, and wherever root's home directory happens to be.

I don't think it is anywhere near that common. UFS (traditional Unix File System) never had that, and I don't think XFS has it today. Greg -- Greg Freemyer Litigation Triage Solutions Specialist http://www.linkedin.com/in/gregfreemyer The Norcross Group The Intersection of Evidence & Technology http://www.norcrossgroup.com -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org