Re: Caching semi-digested credentials in struct cred

[David Howells <dhowells@redhat.com>]

Trond Myklebust <Trond.Myklebust@netapp.com> wrote:

> You should be using the rpc_auth that is cached in the rpc_client:
> 
>   NFS_CLIENT(inode)->cl_auth
> 
> W.r.t. the nfs_find_open_context(), btw, there is no reason why we
> couldn't change that to take a generic cred as its argument.

I think I've been under a misapprehension as to what rpc_auth is for.  I
assumed it was the credential NFS was using for the operations by the user
that it was created for, but given it's a 'superblock' level thing, I guess
not.

> We have yet to see any numbers comparing the current vs. cred performance.

That's something that concerns me, as does the increased risk of overrunning
the stack.  I'm adding an extra argument to an awful lot of functions, and
it's going to degrade performance, no two ways about it.  The question is by
how much?

> > If you'd rather, I can just drop the credentials into the NFS VFS methods
> > and leave it for you to make work.
> 
> The latter sounds good for the moment. There are several people who are
> working on the rpc auth code, and thus have an interest here, so it will
> take some time to work out a consensus for what needs to be done.

The problem with that is that there are bits in the bottom of the sunrpc stuff
that assume they can use current->fs[ug]id (rpcauth_lookupcred() for example),
which is the reason for passing this pointer all the way down:-/

David