RFC smbd smb2 quota support (+ patches to get rid of manual marshall/unmarshall by ndr-ising the existing code)

Noel Power nopower at suse.com
Thu Apr 6 09:26:25 UTC 2017


Gentle reminder :-) please let me know if there anything else I need to
do here

Noel 
On 23/03/17 15:56, Noel Power wrote:
> Hi Jeremy
> On 15/03/17 19:15, Noel Power wrote:
>> On 15/03/17 18:52, Noel Power wrote:
>>> On 13/03/17 22:08, Noel Power wrote:
>>>> Hi Jeremy
>>>> On 13/03/17 09:15, Noel Power wrote:
>>>>> Hi Jeremy
>>>>> On 10/03/17 16:50, Jeremy Allison wrote:
>>>> [...]
>>>>>>> I don't get it :-( afaics whatever numbers (fake or otherwise) you pump
>>>>>>> into sid_list_length, start_sid_length or start_sid_offset you are going
>>>>>>> to end up with an effective array len of 0 -> 4294967295, if it is too
>>>>>>> big the ndr code will detect that, if it is too small then we will find
>>>>>>> out about that later.
>>>>>> I'm sure you're right, and I'll believe you. I still
>>>>>> don't want to see *ANY* unchecked arithmetic in values
>>>>>> read from the wire.
>>>>> that's fine, now I get where you are coming from, I will post new
>>>>> patches as soon as I rework the series
>>>>>
>>>>> Noel
>>>> please see latest patches, hopefully they cover the issues above,
>>>> additionally I added 2 more patches, one to enable smb2 for the existing
>>>> quota check and the other the follow-up patch to combine
>>>> build_user_quota_buffer & fill_quota_buffer into a single common func.
>>> bah, I have to say please ignore these for  a moment, I spotted 2
>>> problems (1 the patches all don't build individually anymore and a small
>>> bug for smbcquota -C) I will post new patches later
>>>
>>> Noel
>> hopefully better now, please see attached, basically a couple of hunks
>> needed to be moved from Patch 11 to ensure Patch 8 would build.
>> Additionally Patch 11 was missing a test to see if max_data was
>> specified before checking if its value was exceeded
>>
>> thanks
>>
>> Noel
>>
> well one thing that I missed was an easy way to run list/get/set each
> time I changed some detail so I've updated the patch set to include a
> basic torture test for smbcquotas exercising get/set/list of quotas
> using smb1 & smb2.
>
> Also patch 14 makes a change to smbcquota to all admin users rather that
> root actually set quoatas (currently only enabled if --enable-selftest)
> I have to wonder though about the requirement for the user to actually
> be root in order to set quotas, is that a mistake or intentional ?
>
> thanks,
>
> Noel
>




More information about the samba-technical mailing list