quota-3.06-11 kernel-2.4.20-20.9 """ Jan 4 04:23:04 mail kernel: VFS: Quota for id 14101 referenced but not present. Jan 4 04:23:04 mail kernel: VFS: Can't read quota structure for id 14101. Jan 4 04:24:18 mail kernel: VFS: find_free_dqentry(): Data block full but it sh ouldn't. Jan 4 04:24:18 mail kernel: VFS: Error -5 occured while creating quota. Jan 4 04:24:21 mail kernel: VFS: Inserting already present quota entry (block 3 20). """ These are some of the messages we get pasted across one of our servers screen. Quotas seem to be totally horked... the only solution I have found is to blow away the quota files. I have read that if `repquota -uga` segfaults, this is an indication of corruption. It does segfault. There is a fix, however, what I am wondering is why this happens. We are running EXT3 filesystem on a server with 3000+ users on. It is not convient to blast quotas and recreate them every week.
Problem also seen on another system. kernel-2.4.7-10smp quota-3.06-9.7 Messages says same thing. Here is an example from the repquota: #179 +- 750780417 179 179 none 179 179 179 none #179 +- 754974721 180 180 none 0 179 179 Segmentation fault Note those two huge blocksizes there. Bumping to high priority as it is repro'd on multiple systems.
This does not appear on later releases