A curious tale of Samba 3.6.6, 3.6.12, and W2K08R2 when an OpLock break occurs on a compound request

[Richard Sharpe]

On Wed, Mar 13, 2013 at 10:10 AM, Richard Sharpe
<realrichardsharpe at gmail.com> wrote:
> On Tue, Mar 12, 2013 at 7:52 PM, Richard Sharpe
> <realrichardsharpe at gmail.com> wrote:
>> On Tue, Mar 12, 2013 at 6:10 PM, Richard Sharpe
>> <realrichardsharpe at gmail.com> wrote:
>>> Hi folks,
>>>
>>> I have been investigating a bug that was interesting. An AFP proxy
>>> running on Windows was causing smbd to crash. This was with 3.6.6+.
>>>
>>> After I got the log files it became obvious that the problem was with
>>> the handling of OpLock breaks in a compound request.
>>>
>>> So, I wrote a little the smbtorture program to test this with the
>>> following result:
>>>
>>> Samba 3.6.6 crashes, as expect.
>>> Samba 3.6.12 returns STATUS_CANCELLED and then STATUS_FILE_CLOSED (but
>>> does not crash.)
>>> W2K08 Returns STATUS_PENDING and then drops the connection.
>>>
>>> I wonder how the proxy deals with the W2K08 behavior?
>>
>> A further observation. Samba 3..6.12 actually sends out an OpLock
>> break, but then responds with CANCELLED because it cannot handle an
>> OpLock break in the middle of a compound request. I suspect Windows
>> drops the connection because it cannot handle the OpLock break
>> processing in the middle of a compound request.
>>
>> I wonder what W2K12 does. I guess I will try it tomorrow.
>
> This is getting more and more curious. If I open the files with OPEN
> (OPEN_EXISTING) W2K08 works fine and sends an OpLock Break. Samba
> 3.6.6 still crashes and Samba 3.6.12 returns STATUS_CANCELLED, which
> is surely wrong.
>
> However, if I open the files with OPEN_IF (OPEN_ALWAYS) W2K08 resets
> the connection.

It turns out not be so curious. I was misinterpreting where the RST
was coming from. It was the initial test program that I wrote that was
faulty.

-- 
Regards,
Richard Sharpe
(何以解憂？唯有杜康。--曹操)