Re: Parsing for obsolete mount options

On Mon, Feb 28, 2011 at 12:45 PM, Jeff Layton [off-list ref] wrote:

On Sat, 26 Feb 2011 12:08:23 -0600
Steve French [off-list ref] wrote:

quoted

I was trying to figure out a way to parse for mount options which are
obsolete for smb2 (but valid for cifs), but without an "-t smb2" fs
type it is harder.    Jeff had preferred (rather than a "-t smb2" file
type) that we call smb2 a "version" of cifs, and by analogy with nfs
("nfsvers=2",  "nfsvers=4" etc.) specify a "vers=smb2" or "vers=2" to
indicate mounting with experimental smb2 is desired ... but to be able
to parse mount options which are now obsolete (for smb2) we have to
scan the whole mount option string, looking for "vers=2" (or
equivalent) first, then either call a (new) smb2_parse_mount_options
function, or flag them in the current cifs_parse_mount_options when we
detect a "vers=2"  - by the time we get out of
cifs_parse_mount_options it is too late because we don't know which
mount options the user specified explicitly (and which we got from
defaults).

As a summary of some of the mount differences:
1) sec=ntlm (and sec=lanman) is no longer going to be valid (ntlmv2
and krb5 only)
2) noblocksend and noautotune probably no longer needed
3) posixpaths/noposixpaths no longer needed
4) noserverino not valid
5) sockopt may not be needed (nodelay can be defaulted)

Probably, the best thing is to do this in 2 phases:

1) parse the mount options to the best of your ability

2) check the result and see whether they make sense

Autonegotiation will make the second part interesting though since you
might not know the version until later. In that case, you probably will
just have to ignore options that don't have any effect on the version
negotiated.

I may have to add a new bool (or int sec_flags_requested) in the vol
structure for at least one requested security option, in particular
for "ntlm" requested (since ntlm is the default, we want to
distinguish "use the default security mechanism" from specifying
"sec=ntlm" ie "give me ntlm" (which will only allow for cifs) which we
want to fail)



-- 
Thanks,

Steve