From Bugzilla Helper: User-Agent: Mozilla/4.78 [de] (X11; U; Linux 2.4.17 i686) Description of problem: quotaon doesn't recognize Version2 quota which quotacheck -F vsfv0 /dev/hda? produces. If you want to enable quotas, quotaon tells you it couldn't find quota.user and doesn't look for aquota.user, so it doesn't start the quotasystem. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Add usrquota in /etc/fstab to selected partition 2. quotacheck -F vsfv0 /dev/hda? (produces aquota.user) 3. edquota $user 4. quotaon /dev/hda? Actual Results: quotaon can't find quota.user (of course, the file doesn't exist), but doesn't look for aquota.user (exists) and there is no -F vfsv0 switch like for quotacheck or edquota. Expected Results: quotaon should start Version2 quotasystem. Additional info: If I use -F vfsold for quotacheck and edquota, it produces quota.user and quotaon is able to start the quotas. But I would like to use v2 quotas. Kernel is a handmade 2.4.17 with enables quota-support
I had this problem too, so checked here and noticed this bug report. Then I tracked down what the problem is -- the problem is that the RedHat kernels use the Alan Cox (ac) patches, which change the way quotas work in the kernel in a fundamental way (changes the command codes to the quotactl system call among other things). Since the quotaon binary is compiled against the ac-patched kernel and you (and I) are running the standard Linux kernel, the commands don't match up and the quotactl call doesn't work right. I haven't tried this yet, but I am going to try to recompile the quota package binaries against the standard 2.4.17 kernel to see if that works. [ Personal rant: I think Red Hat using such a drastically different kernel (a 7.8 Meg patch to the kernel source) is a really bad idea. Sure Alan Cox knows what he's doing, but this is the kind of kernel forking is really a bad thing -- there's nothing wrong with the standard kernel! ]
OK, I tried what I mentioned above, and it seems the quota user-level packages are pretty closely tied to the ac kernel patches, so just recompiling with the standard 2.4.17 definitions did not work. I do have it working now, but not with the 2.4.17 kernel. I patched up to the 2.4.18 kernel, then applied the ac patches to make a 2.4.18-ac3 kernel, and then everything works great. I'm still somewhat annoyed that the quota package doesn't work with a standard kernel, but since it's working now I'm finding it harder to be *seriously* annoyed....
It seems that in some secret hidden place on a filesystem, it is written whether new or old quotas should be used for "quotaon". Try removing aquota.user, then run convertquota, quotacheck and quotaon. Now it should work. Unfortunately it wipes the quota data, so if you want to keep it, you have to find a way to do a backup/restore.
There is a new errata candidate for the 7.x series, quota-3.06) kicking around in rawhide (ftp://ftp.redhat.com:/pub/redhat/linux/rawhide/SRPMS/SRPMS) that should address this issue with v1/v2 quota file formats. Phil =--=