Re: PATCH: Support binding to a local IPv4 address when mounting a server.

[Ben Greear <greearb@candelatech.com>]

Chuck Lever wrote:
> A handful of generic comments.
>
> 1.  This needs to be broken into smaller patches before submission; 
> preferably before you submit another version for review.  Take a look 
> at the linux-nfs@vger.kernel.org archives to see how we handle large 
> changes like this.
>
> 2.  You should support local addresses only in the text-based path 
> (utils/mount/stropts.c) and not in the legacy paths 
> (utils/mount/nfs[4]mount.c).  I don't think we're ever going to allow 
> a version 7 of the mount data structure.
I could remove the version 7 data field, but what about the other code 
that creates
sockets for 'pinging' nfs daemons and such?   Is that code deprecated now?

If nothing else, it looks like "probe_bothports" needs a client-addr to 
pass on
to methods it calls, and so forth.  That means that other code that 
calls those
methods needs to be updated, and thus my huge repetitive patch.

> 3.  There are some umount-related changes coming up for IPv6 that will 
> touch the umount paths here; that may require some changes in your 
> modifications.
That should be fine.   I tried to make my current patch friendly for 
IPv6 and want to support local
IPv6 binding as well.
>
> 4.  This needs to have support for a new mount option in the kernel 
> (not a command-line option to the mount.nfs command, as you have 
> implemented) to handle passing the address to the kernel.
I think I already have this ready in the kernel, but I've been using the 
version 7 field to pass
in data using mount.nfs.    Can you give an example of a user-space 
command that would use the
new mount option as you suggest?  Do you do /etc/fstab any different for 
instance?

Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com> 
Candela Technologies Inc  http://www.candelatech.com