Re: NFS export ... again

[Trond Myklebust]

On Fri, 2004-04-16 at 00:05, fabian.frederick@xxxxxxxxxxxxx wrote:
> Trond, Chris,
> 
>      AFAICS, nfsd/export code already does some nfs cache poll through RPC to generate /proc/nfs/exports .... But content appears broken to me as it relates progressive mounting points but never umounts.We only have to fix this in my opinion ... or maybe I'm lured by multi-versioning there ... 

/proc/nfs/exports is a server-side thing, and just reflects the kernel's
cache of currently exported filesystems. It is controlled by mountd by
means of a syscall...

I don't see what you are getting at here. There is no such thing as a
reliable way of ensuring that the server has a fully uptodate list of
all clients that are currently mounting a filesystem: Clients can and do
crash. When they do, they will fail to send a "umount" call to the
server. When that happens, the reference count in /var/lib/nfs/rmtab
will be wrong, and so all these lists from /proc/fs/nfs/exports to
"showmount" will show clients that are no longer there.

Just don't rely on that information: it will never be completely
correct.

Cheers,
  Trond
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/