Skip to contentSkip to conversation

mtab corruption

DISCUSSION Dieter Stüken 4 participants 6 messages Jan 3, 2007 – Jan 9, 2007 Kernel automounter

lore
Share & tools
lore
#0 lore raw

I frequently end up with a broken mtab. This results in a message like:
mount(nfs): warning: /home/stueken is already mounted

I'm using a SuSE-10.1 x86_64-smp system with a 2.6.16.21-0.25-smp kernel. After I read about new patches to solve some mtab locking problems I upgraded from autofs-4.1.4-23 to autofs-4.1.4-37, but the mtab corruption still occurs.

Seems the problem became even worse since a mount request sometimes takes about 5 seconds (due to some LDAP configuration problems) while a bunch of other mount requests are coming in.
In addition this is an SMP system.

Should I consider to update my kernel, too?

Unfortunately I could not convince my sysads to do so, as they are afraid of even more problems with newer kernels...

I noticed the upcoming autofs5. Should I wait for it?

thanks for advice,

- Dieter.

maintainer
#1 lore raw
Quoted 7 lines · Dieter Stüken

I frequently end up with a broken mtab. This results in a message like:
mount(nfs): warning: /home/stueken is already mounted

I'm using a SuSE-10.1 x86_64-smp system with a 2.6.16.21-0.25-smp kernel. After I read about new patches to solve some mtab locking problems I upgraded from autofs-4.1.4-23 to autofs-4.1.4-37, but the mtab corruption still occurs.

I don't know what this means, I don't use SuSE.

Quoted 5 lines

Seems the problem became even worse since a mount request sometimes takes about 5 seconds (due to some LDAP configuration problems) while a bunch of other mount requests are coming in.
In addition this is an SMP system.

This corruption is not actually an autofs problem although verious things have been tried over the years to reduce it. mount(8) is where the problem lies.

Does your system use mount.nfs from nfs-utils?
There is an mtab locking error I spotted in nfs-utils recently.

Quoted 2 lines

Should I consider to update my kernel, too?

Waste of time, nothing to do with the kernel.
/etc/mtab is managed entirely by user space code.

Quoted 5 lines

Unfortunately I could not convince my sysads to do so, as they are afraid of even more problems with newer kernels...

I noticed the upcoming autofs5. Should I wait for it?

At the moment v5 still expects /etc/mtab to be correct.
This can be fairly easily changed but I would prefer to be able to rely on mount(8) because, for large direct mount maps (at least) we will have a performance penalty.

Ian

_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs

#2 lore raw
Quoted 3 lines · Ian Kent
Quoted 2 lines · Dieter Stüken

I frequently end up with a broken mtab. This results in a message like:
mount(nfs): warning: /home/stueken is already mounted

Sorry, I don't have anything useful to add on the mtab corruption issue. We do a ton of NFS automounting and don't see this issue.

Quoted 4 lines
Quoted 2 lines

I'm using a SuSE-10.1 x86_64-smp system with a 2.6.16.21-0.25-smp kernel. I upgraded from autofs-4.1.4-23 to autofs-4.1.4-37, but the mtab

I don't know what this means, I don't use SuSE.

The trailing numbers differentiate variants of the "same" software version, e.g. security patches. SuSE has a policy of backporting patches to old versions, if feasible, in contrast to Debian, which I think would send out the latest version including the latest patch. The idea is to minimize feature changes except at distro version transitions.

Quoted 2 lines

Does your system use mount.nfs from nfs-utils?
There is an mtab locking error I spotted in nfs-utils recently.

SuSE 10.1 uses /bin/mount from util-linux-2.12r-35 which does NFS intrinsically, and not mount.nfs from nfs-utils-1.0.7, which is installed but minus mount.nfs.

James F. Carter Voice 310 825 2897 FAX 310 206 6673
UCLA-Mathnet; 6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA 90095-1555
Email: [email protected] http://www.math.ucla.edu/~jimc (q.v. for PGP key)

_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs

maintainer
#3 lore raw
Quoted 11 lines · Jim Carter
Quoted 3 lines · Ian Kent
Quoted 2 lines · Dieter Stüken

I frequently end up with a broken mtab. This results in a message like:
mount(nfs): warning: /home/stueken is already mounted

Sorry, I don't have anything useful to add on the mtab corruption issue. We do a ton of NFS automounting and don't see this issue.

Quoted 3 lines
Quoted 2 lines

I'm using a SuSE-10.1 x86_64-smp system with a 2.6.16.21-0.25-smp kernel. I upgraded from autofs-4.1.4-23 to autofs-4.1.4-37, but the mtab

Quoted 7 lines
Quoted 1 line

I don't know what this means, I don't use SuSE.

The trailing numbers differentiate variants of the "same" software version, e.g. security patches. SuSE has a policy of backporting patches to old versions, if feasible, in contrast to Debian, which I think would send out the latest version including the latest patch. The idea is to minimize feature changes except at distro version transitions.

Sure, but my point was that I'm not familiar with what patches the SuSE maintainer may have added. Perhaps this would be a good time for the SuSE maintainer to speak up, if present.

Quoted 7 lines
Quoted 3 lines

Does your system use mount.nfs from nfs-utils?
There is an mtab locking error I spotted in nfs-utils recently.

SuSE 10.1 uses /bin/mount from util-linux-2.12r-35 which does NFS intrinsically, and not mount.nfs from nfs-utils-1.0.7, which is installed but minus mount.nfs.

OK. So we still don't know if the locking code in util-linux is broken.

Ian

_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs

#4 lore raw
Quoted 20 lines · Ian Kent
Quoted 11 lines · Jim Carter
Quoted 3 lines · Ian Kent
Quoted 2 lines · Dieter Stüken

I frequently end up with a broken mtab. This results in a message like:
mount(nfs): warning: /home/stueken is already mounted

Sorry, I don't have anything useful to add on the mtab corruption issue. We do a ton of NFS automounting and don't see this issue.

Quoted 3 lines
Quoted 2 lines

I'm using a SuSE-10.1 x86_64-smp system with a 2.6.16.21-0.25-smp kernel. I upgraded from autofs-4.1.4-23 to autofs-4.1.4-37, but the mtab

Quoted 7 lines
Quoted 1 line

I don't know what this means, I don't use SuSE.

The trailing numbers differentiate variants of the "same" software version, e.g. security patches. SuSE has a policy of backporting patches to old versions, if feasible, in contrast to Debian, which I think would send out the latest version including the latest patch. The idea is to minimize feature changes except at distro version transitions.

Quoted 3 lines

Sure, but my point was that I'm not familiar with what patches the SuSE maintainer may have added. Perhaps this would be a good time for the SuSE maintainer to speak up, if present.

Yes, I know about this problem. However I have not had the time yet to do some further debugging.
We have a patch addressing the mtab locking part of fstab.c however this seems not to be enough. According to a recent mail on the util-linux-ng list there is a principal problem with the method used in mount. This needs some more investigation. As soon this problem is fixed I think this is important enough to backport.

Matthias

_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs

maintainer
#5 lore raw
Quoted 5 lines · Matthias Koenig
Quoted 4 lines

Sure, but my point was that I'm not familiar with what patches the SuSE maintainer may have added. Perhaps this would be a good time for the SuSE maintainer to speak up, if present.

Quoted 7 lines

Yes, I know about this problem. However I have not had the time yet to do some further debugging.
We have a patch addressing the mtab locking part of fstab.c however this seems not to be enough. According to a recent mail on the util-linux-ng list there is a principal problem with the method used in mount. This needs some more investigation. As soon this problem is fixed I think this is important enough to backport.

Thanks for the reply Matthias.

I'm also following the discussion on util-linux-ng. We'll see how it develops.

Ian