Bug 6713
Summary: | Write by group not working in any 2.6.16.X | ||
---|---|---|---|
Product: | File System | Reporter: | Dan Ohnesorg (lk) |
Component: | NFS | Assignee: | Trond Myklebust (trondmy) |
Status: | REJECTED UNREPRODUCIBLE | ||
Severity: | normal | CC: | akpm, neilb, protasnb |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.16.any | Subsystem: | |
Regression: | --- | Bisected commit-id: |
Description
Dan Ohnesorg
2006-06-19 03:19:26 UTC
That sounds fairly serious. Some details on how to set up the userspace NFS server might help. Also, why can't you use the in-kernel nfs server? NFS with the default AUTH_SYS authentication is limited to 16 groups. That is a protocol limitation rather than a Linux implementation problem. It used to be (in 2.4.x kernels) that you could somewhat compensate for that by putting the NFS groups that interest you as the first 16 in the list. Unfortunately in 2.6.x kernels someone decided that it was a good idea to sort the groups list, and so we can no longer use that hack. Thank you for reply Ad setup of user space nfs server, we didn't made any special configuration, only apt-get install nfs-user-server, edited /etc/exports to this: # /etc/exports: the access control list for filesystems which may be exported # to NFS clients. See exports(5). /www 192.168.0.2/255.255.255.255(rw,no_root_squash,sync) /mnt/backup 192.168.0.2/255.255.255.255(rw,no_root_squash,async) and restart nfs server, thats all. Ad using kernel space nfs: First we need reexport nfs mounted NAS to clients, but kernel space nfs doesn't know how to do it. Second the NAS is build on OpenNAS and OpenNAS uses only user space nfs. Now I can confirm, that the problem is even in 2.6.17. Neil Brown has contacted me privately, I will send him tcpdump of communication between client and server. Sorry, the confirmation in 2.6.17 was made too early, even if the situation is not perfect, we have got another results. Please wait until we discover exactly what is changed. Any update on this? Thanks. I suspect the problem has gone away, or been resolved. Please reopen if this is not the case. |