[Mauiusers] Maui bug buffer overrun

[Bas van der Vlies]

It is about a week ago when i asked some questions. It would be nice  
to have some answers ;-) and some clarification what steps are  
necessary to submit patches or update the docs.  So what is the  
clusterresources policy?

Just another question can i distribute a patched maui version?

Regards



On Apr 11, 2008, at 9:03 PM, Bas van der Vlies wrote:

>
> On Apr 11, 2008, at 5:41 PM, Douglas Wightman wrote:
>
>> It's possible I was looking at the wrong patch.  I was looking at the
>> one mentioned here:
>>
>> http://www.clusterresources.com/pipermail/mauiusers/2005-September/ 
>> 001734.html
>>
> Hi Douglas,
>
>    There is a thread here with an explanation:
>     - http://www.clusterresources.com/pipermail/mauiusers/2006- 
> November/002370.html
>
>   And some other where your quote came from.  There is no patch  :-
> (  only a statement that is under investigation of cluster resources
> and that  you encountered the same bug.  After that is was quiet.
>
> Now we have the attention of cluster resources people. I have another
> question. I have made a patch for FairShare groups in combo with
> Torque. Fairshare groups only look at the primary unix group. With
> this patch secondary groups can also be used for a FairShare.  The
> archives are useful :-)
>   * http://www.clusterresources.com/pipermail/mauiusers/2008- 
> January/003066.html
>
> It fixes also 2 other issues
>
> Thanks
>
>
>>
>> On Fri, 2008-04-11 at 11:35 +0200, Bas van der Vlies wrote:
>>> Douglas Wightman wrote:
>>>> Going over the thread you reference it appears as though there  
>>>> was a
>>>> request by Dave that the patch be tweaked to utilize the
>>>> MUSNPrintF()
>>>> function found in Maui.  Has this been done?  We'd love to  
>>>> integrate
>>>> your patch but we'd like to stick to Maui'isms.
>>>>
>>>> Thanks!
>>>>
>>>
>>> Douglas,
>>>
>>>   Thanks for a reaction. But what you write was not the real
>>> problem. Did
>>> you read the thread, because a colleague of yours wrote:
>>> {{{
>>> On Nov 7, 2006, at 4:43 PM, wightman wrote:
>>>
>>>>> The patch looks good.  We will definitely include it in the next
>>>>> patch
>>>>> release.
>>> This was the NODEALLOCMAPS patch.
>>>
>>>>>
>>>>> Also, we are still investigating the segv from MMAX_JOB.  We have
>>>>> reproduced the crash and, as somebody mentioned earlier in this
>>>>> thread,
>>>>> the real issue is that the MJob array is not properly
>>>>> terminated.  We
>>>>> are running Maui under valgrind in hopes of catching when this
>>>>> occurs.
>>>>>
>>> }}}
>>>
>>> Or is this something else?
>>>
>>>
>>> PS) What is the procedure for a patch to be applied. Several people
>>> have
>>> mailes some improvements or bug fixes?
>>>
>>> Regards
>>>
>>>> - Douglas
>>>>
>>>> On Wed, 2008-04-09 at 23:07 +0200, Bas van der Vlies wrote:
>>>>> On Apr 9, 2008, at 5:21 PM, Miguel Ros wrote:
>>>>>> Hi all,
>>>>>>
>>>>>> I had some problems with a Linux x86_64 fedora core 8: Maui
>>>>>> was crashing with SIGSEGV from time to time and showq was
>>>>>> not responding correctly
>>>>>>
>>>>>> There are a lot of buffer overrun problems in Maui that are not
>>>>>> solved
>>>>>> yet in the official release:
>>>>>>
>>>>>> http://www.clusterresources.com/pipermail/mauiusers/2005- 
>>>>>> September/001724.html
>>>>>>
>>>>>> I have attached a patch file for Maui 3.2.6p19 - 3.2.6p20, I hope
>>>>>> that it will be applied to the official release
>>>>>>
>>>>>> Regards,
>>>>>> Miguel
>>>>>> <sprintf_bug.patch><ATT00001.txt>
>>>>>
>>>>> Thanks for the patch, maybe we must setup a repository where we
>>>>> collect the patches. Because there is lately no response when
>>>>> people
>>>>> submit patches for maui.
>>>>>
>>>>>
>>>>> I also reported that maui crashed when i lot of jobs where
>>>>> submitted
>>>>> and i promise that it will be fixed,  see thread:
>>>>> *   [Mauiusers] maui crashes with segv (version 3.2.6.p16) (2006)
>>>>>
>>>>>
>>>>> and i added some explanation (also from the thread):
>>>>> {{{
>>>>> I have adjusted the MAX_MHBUF parameter in msched.h to a higher
>>>>> value.
>>>>> No it did not crash, but it is still possible not to find a free
>>>>> job
>>>>> hash table slot.  IF you allow more jobs then to have also  
>>>>> increase
>>>>> this
>>>>> parameter.
>>>>>
>>>>> /* HvB */
>>>>> /*
>>>>> #define MAX_MHBUF           768
>>>>> */
>>>>> #define MAX_MHBUF          4000
>>>>>
>>>>> }}}
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Bas van der Vlies
>>>>> basv at sara.nl
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> mauiusers mailing list
>>>>> mauiusers at supercluster.org
>>>>> http://www.supercluster.org/mailman/listinfo/mauiusers
>>>>
>>>
>>>
>>
>
> --
> Bas van der Vlies
> basv at sara.nl
>
>
>
> _______________________________________________
> mauiusers mailing list
> mauiusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/mauiusers

--
Bas van der Vlies
basv at sara.nl