Re: %openmosix and sata

Gianluca <[email protected]> Thu, 06 Jul 2006 12:30:55 +0200
Newsgroups gmane.linux.cluster.openmosix.general
Message-ID <[email protected]>
Dear Felipe... and all people,
I've just runned some tests... and the results are:
 - If I launch the test.sh locked (migration not allowed) then Success -
NO Segmentation fault
 - If I launch the test.sh from a normal shell with migration allowed
then Failure - YES Segmentation fault
 - If I launch the test.sh with the batch command (via atd daemon) then
Success - NO Segmentation fault
So... The problem arise when a Shell Script is going to be migrated AND
It is launched from a normal command prompt.
But the Shell Script is successfull migrated IF the Shell Script is
launched via atd daemon (batch - detached from normal command prompt).
I think the way to discover the bug is to understand the differences on
process environments between atd and shell.
In fatc, I think there are some differences that lead to openmosix to
migrate incorrectly shell script launched from a normal command prompt.
Please Felipe, could you create a file launch.sh with follow content:

#!/bin/bash
mosrun -l ./test.sh

and try to execute:
batch -f launch.sh NOW

Ciao,
Gianluca


Gianluca ha scritto:
> Dear Felipe,
> I runned your test file hundreds of time on my machines... without problems.
> But when I try to launch the test with this command:
>  mosrun -5 ./test.sh
> I've got the segmentation fault. :-(
> However... I'm waiting the 2.6.x version for a long time... but It is
> still in alpha version :-( and I prefer to remain with my patched kernel
> until the 2.6.x is not released officialy.
> This is the error:
>
> Unable to handle kernel NULL pointer dereference at virtual address 0000001c
>  printing eip:
> c01661b2
> *pde = 00000000
> Oops: 0002
> CPU:    0
> EIP:    0010:[<c01661b2>]    Not tainted
> EFLAGS: 00010286
>
> But I don't understand what means... and where do I have to look for in
> the kernel code.
> In my kernel the bug arise when I launch the 'mosrun -5' command... but
> I don't know where is the corresponding code executed... and how to find
> out the bug :-(
>
> Ciao,
> Gianluca
>
>
> Felipe ha scritto:
>   
>> Hi Gianluca
>>
>> Thanks for your answer
>>
>> I follow have the same problem ....
>>
>> I try to explain the problem:
>>
>> My script test.sh:
>>    for i in 1 2 3 4 5 ; do 
>>       awk 'BEGIN {for(i=0;i<10000;i++)for(j=0;j<10000;j++);}' &
>>
>>    done
>>
>> My cluster have 2 nodes, when i run above the script, the first time its ok,
>> but when i run it more times is when i get segmentation fault. Example:
>>
>> openmosix-1 ~ # ./test.sh
>> openmosix-1 ~ # ./test.sh
>> Segmentation fault
>>
>>
>>
>> I begin to read the kernel-hacking howto... and find it:
>>
>> dmesg return an "Oops:0002" and "EIP:    0010:[<c01675b7>]", but I don't
>> find "Oops:0" and "EIP:  0010:[<c01675b7>]" in System.map... why ???? I
>> dont know
>>
>> If you could help me, I try to modify your patch.
>>
>>
>>
>> Maybe a solution for my problem is:
>>
>> Install om for kernel 2.6; but it is not recomendable for production ...
>> what do you thing about it?
>>
>>
>> Thanks in advance
>>
>> Sorry for my english
>>
>>
>> El mar, 04-07-2006 a las 10:07 +0200, Gianluca escribió:
>>   
>>     
>>> Dear Felipe,
>>> I know that there are some bugs in my kernel patch :-(
>>> but I'm not an expertise of kernel hacking... so I don't know how degub
>>> linux kernel and find where is the bug! :-(
>>> Months ago, I wrote on the mailing-list... saying that I've prepared
>>> this patch and I asked to community to help me in debugging and
>>> stabilizing my patch for 2.4.31 kernel...
>>> ... but I didn't get any help :-( :-(
>>> Seems that people using openmosix are not expertise of kernel
>>> debugging... and obviously the main developers don't spent time for
>>> debugging 2.4.x series :-(
>>> However, my cluster based on my patch runs from several months without
>>> stops!! :-) :-)
>>> For avoid the bug... don't run the mosrun command with the flag for
>>> migrate the job.
>>> For example, if you launch the command:
>>>  mosrun -4 heavyJob
>>> from a node different to 4... then you'll see the Segmentation Fault :-(
>>> So... if you never use that command... all things go well ;-) ... in
>>> fact, the automatic migration works well.
>>>
>>> Ciao,
>>> Gianluca
>>>
>>> Felipe ha scritto:
>>>     
>>>       
>>>> Thanks
>>>>
>>>> I try with it and the sata disk is ok, but when i probe the om with the
>>>> next script:
>>>>
>>>>   for i in 1 2 3 4 5 ; do 
>>>>      awk 'BEGIN {for(i=0;i<10000;i++)for(j=0;j<10000;j++);}' &
>>>>   done
>>>>
>>>>
>>>> When i run the script, its ok, but the next run display the next
>>>> message:
>>>>
>>>>         Segmentation fault
>>>>         
>>>> dmesg returns:
>>>>
>>>>         <1>Unable to handle kernel NULL pointer dereference at virtual
>>>>         address 0000001c
>>>>          printing eip:
>>>>         c01675b7
>>>>         *pde = 00000000
>>>>         Oops: 0002
>>>>         CPU:    0
>>>>         EIP:    0010:[<c01675b7>]    Not tainted
>>>>         EFLAGS: 00010286
>>>>         eax: 0000001c   ebx: f756c000   ecx: 00000000   edx: ffff0001
>>>>         esi: 080ed000   edi: 080e8000   ebp: f756dd34   esp: f756da34
>>>>         ds: 0018   es: 0018   ss: 0018
>>>>         Process test.sh (pid: 1705, stackpage=f756d000)
>>>>         Stack: f756da38 00000000 00000000 f59bf920 080e2000 c016c5ee
>>>>         080e7d74 080ec93c
>>>>                00005d74 00000003 00001812 0000009a f760a000 f762003c
>>>>         f6457b58 00000000
>>>>                00011e8c f6457bc0 00000001 060c0705 00000003 c02b5fac
>>>>         f6457b00 f756c000
>>>>         Call Trace:    [<c016c5ee>] [<c02b5fac>] [<c02af770>]
>>>>         [<c016c1c0>] [<c015136d>]
>>>>           [<c0150f6c>] [<c016a33f>] [<c01404eb>] [<c016a130>]
>>>>         [<c015136d>] [<c015153d>]
>>>>           [<c0105ba2>] [<c010a18f>] [<c01d5ffb>] [<c0105b60>]
>>>>         [<c01d4a97>] [<c01da988>]
>>>>           [<c010a06a>]
>>>>         
>>>>         Code: 0f c1 10 85 d2 0f 85 9a 15 00 00 89 3c 24 29 fe 89 74 24
>>>>         04
>>>>
>>>> My cluster have 2 nodes
>>>>
>>>> I compile the kernel with differents optios, but the result is the same.
>>>>
>>>> I have many questions, for example:
>>>>
>>>> om have support for my hardware?
>>>> Maybe the om conigurations is bad
>>>> My filesystem is Reiser 3.x .... is it the problem?
>>>> ?????
>>>>
>>>> any help is good for me 
>>>>
>>>> thanks in advance
>>>>
>>>>
>>>>
>>>>
>>>> El sáb, 01-07-2006 a las 09:54 +0200, Gianluca escribió:
>>>>   
>>>>       
>>>>         
>>>>> I have patched the 2.4.31 kernel... to resolving the same problem.
>>>>> attached you'll find my patch... you can try it ;-)
>>>>>
>>>>> Felipe ha scritto:
>>>>>     
>>>>>         
>>>>>           
>>>>>> I tried that also, but my pc's have the  I/O Controller Hub 7 and the
>>>>>> libata patch for kernel 2.4.26 doesnt have support for it.
>>>>>>
>>>>>> then which is the best versio of om for my pc's?
>>>>>>
>>>>>> Do you now how can i patch the kernel 2.4.26 with the newest libata
>>>>>> patch?
>>>>>>
>>>>>>
>>>>>>
>>>>>> El mié, 28-06-2006 a las 17:09 -0500, Brian English escribió:
>>>>>>   
>>>>>>       
>>>>>>           
>>>>>>             
>>>>>>> I have tried using the Gentoo Openmosix Kernels.  Try your best to
>>>>>>> avoidthese, especially the 2.4.30 one, as they are unstable at best.
>>>>>>> I would get segfaults while running the OpenMosixTest package as well
>>>>>>> as using the cluster to encode wav files to flac files. Your best bet
>>>>>>> is to manually install the latest stable version from the site
>>>>>>> (2.4.26) and patch the kernel for SATA support (I'm not sure on how to
>>>>>>> do that, but I think it can be done).
>>>>>>>     
>>>>>>>         
>>>>>>>             
>>>>>>>               
>>>>>>   
>>>>>>
>>>>>>
>>>>>>
>>>>>> Using Tomcat but need to do more? Need to support web services, security?
>>>>>> Get stuff done quickly with pre-integrated technology to make your job easier
>>>>>> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
>>>>>> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
>>>>>> _______________________________________________
>>>>>> openMosix-general mailing list
>>>>>> openMosix-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>>>>>> https://lists.sourceforge.net/lists/listinfo/openmosix-general
>>>>>>
>>>>>>   
>>>>>>       
>>>>>>           
>>>>>>             
>>>> Using Tomcat but need to do more? Need to support web services, security?
>>>> Get stuff done quickly with pre-integrated technology to make your job easier
>>>> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
>>>> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
>>>> _______________________________________________
>>>> openMosix-general mailing list
>>>> openMosix-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>>>> https://lists.sourceforge.net/lists/listinfo/openmosix-general
>>>>
>>>>   
>>>>       
>>>>         
>>   
>>     
>
>
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> openMosix-general mailing list
> openMosix-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/openmosix-general
>
>   


Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642