[PATCHSET] honour output buffer length for SMB2 GetInfo

[Jeremy Allison]

On Mon, Aug 05, 2013 at 10:24:58AM +0200, Ralph Wuerthner wrote:
> On 03/08/13 19:22, Volker Lendecke wrote:
> >On Fri, Aug 02, 2013 at 05:09:15PM -0700, Jeremy Allison wrote:
> >>On Mon, Jul 29, 2013 at 10:01:06AM +0200, Ralph Wuerthner wrote:
> >>>Hi,
> >>>
> >>>please see attached patchset to honour output buffer length set by
> >>>the client for SMB2 GetInfo requests. The patchset also enhances
> >>>GetInfo for file system information classes SMB_QUERY_FS_VOLUME_INFO
> >>>and SMB_QUERY_FS_ATTRIBUTE_INFO to return partial, but valid data in
> >>>case output buffer length is too small. This behaviour is documented
> >>>in MS-FSCC.
> >>>
> >>>We had a customer defect where a client submitted a GetInfo
> >>>SMB_QUERY_FS_VOLUME_INFO request with a output buffer length of 24
> >>>bytes. Samba responded with a full length response of 40 bytes which
> >>>resulted in a client error.
> >>
> >>Sorry for the delay. I'm happy with this patchset and will
> >>add my:
> >>
> >>Reviewed-by: Jeremy Allison <jra at samba.org>
> >>
> >>tag. Can I get a second Team member +1 on this and I'll
> >>push to master and get a bug logged to get this into the
> >>next release ?
> >
> >+1 from me.
> >
> >One question: It seems we don't have a torture test for
> >this. Can we get one?
> 
> For testing my patchset I wrote a simple smbtorture test. Initially
> I thought about adding this to smb2.getinfo, but smb2.getinfo is on
> the known fail list because of missing streams support in Samba in
> standard configuration and an issue with file timestamps.
> 
> Can we add this to smb2.getinfo regardless of the issues mentioned
> above or what would be the preferred solution?

Yes, you can add it into smb2.getinfo. At least we'll
have the test there we can run, even if it's not being
passed on ever check-in.

I'll push your fix and be happy to review the test.

Jeremy.