smb2_file_rename_information must check that the file has been opened with DELETE Access ...

[Jeremy Allison]

On Thu, Nov 15, 2012 at 10:40:02PM +0100, Christian Ambach wrote:
> Hi Jeremy,
> 
> On 11/14/2012 11:54 PM, Jeremy Allison wrote:
> 
> >>Do we have an existing piece in the smb2 torture tests that shows that
> >>this necessary? If there is a test in place, I think the patch is valid.
> >
> >Yep, we already have such a test. Check out smb2.rename - and we have:
> >
> >selftest/knownfail:^samba3.smb2.rename.*.simple_nodelete
> >selftest/knownfail:^samba3.smb2.rename.*.no_share_delete_no_delete_access
> 
> >
> >So I think we can scratch some of these knownfails off the list with this
> >fix :-).
> 
> So you decided to put it in the sharemode check that you considered to
> be unnecessary in Bug 8412? Good to hear that :)

Not quite :-). It's still unnecessary for 99.9999% of cases :-).
This change isn't that, but the one for the second error will be :-).

> I was the one who wrote most of these tests to show that the fix was not
> complete (but also make sure we don't fail with MS Word again).

Thanks - it's much appreciated. The real issue is that the
tests open the containing directory with access masks of
FILE_READ_DATA|FILE_WRITE_DATA, which cause share modes
to be checked - something I don't think *any* windows
clients ever do against a directory handle (the MS Word
test you wrote certainly doesn't) as any requests for
reads and writes on the raw directory handle will be
rejected anyway.

Jeremy.