> I just tried this and don't seem to be able to reproduce the
> problem - it works as advertised for me. Are you able to build
> your own kernel? If so, I'd suggest putting some printk's into
> xfs_qm_syscalls.c::xfs_qm_scall_quotaon(); esp. where EINVAL is
> returned and we will get a better picture of where this is going
> I'll try it out on a few more machines tomorrow and see if I can
> uncover the problem. It smells like an uninitialised or perhaps
> only partially inited flags variable, or something like that --
> nothing leaps out at me after reading through the code though.
seems i have just found out the reason, the problem only occured on the
machine with scsi hdd.
i have setup 3 scsi machines:
machine A - adaptec 29160, ibm 18g scsi
machine B - adaptec 29160, seagate 9g scsi
machine C - adaptec 2940UW, seagate 9g scsi
both machines got the "Invalid argument" problem, but i have setup 2 ide hdd
machines last night, and don't have such problem.
> Also, what does "repquota -vug /" on your machine now say?
[root@xfs root]# repquota -vug /
repquota: Not all specified mountpoints are using quota.