[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Nios2 problems when building tests



Roger Dahlkvist wrote:
> Instead of using the simulator I just created a simple system with the same contents as the nios2_iss bsp, except for an external sdram since the FPGA I'm currently using can't have enough onchip memory. I replaced linkcmd and made the required changes to bsp.h
>
> So far it runs just perfectly on hardware. Currently it's only a single task counting numbers though.
> I just had a brief look in the debugger and as far as I can tell, it looks ok.
>   
I am thrilled this works.

Is this with my linkcmds change that moves the starting stack?

--joel
> //Roger
>  
>
>
> Joel Sherrill wrote:
>   
>> Morten Kristiansen wrote:
>>
>>     
>>> As a shortcut, could you send me the binary image you have, please?
>>>
>>>
>>>
>>>       
>> Grrr.. now it isn't linking for me. Maybe I didn't get as far as I
>> remembered
>> or I have deleted the RTEMS tree. It looks close so i will work on it.
>>
>> Maybe in the morning you will have something if my build finishes.
>>
>>
>>     
> My changes that should be committed now are.  Attached
> is a patch for the Makefile's so we don't use a file list and
> work with the old ar command.
>
> The starting stack is CLEARLY at the end of RAM in the
> start/start.S and startup/linkcmds.  This is used as
> work area so getting clobbered early.
>
> Attached in a private email is my hello if you want to look at
> it but we know the stack is broken.
>   
>> --joel
>>
>>
>>     
>>> /MK
>>>
>>> 2009/3/25 Roger Dahlkvist <ruppe at kth.se>:
>>>
>>>
>>>       
>>>> Hi,
>>>>
>>>> Yes, I defenately think the errors are memory related. I'll look at the hints you gave.
>>>>
>>>> When I for example run hello.exe in the Altera NiosII ISS I get the following errors before the program terminates.
>>>>
>>>> Error! : Failed memory access in component cpu_0 - Reading data 0x0 from uninitialized memory (addr = 0x1f33c)
>>>> Error! : Simulation failed in component cpu_0 at instruction 33372 (PC=0x8eb8 instr=0x2c000017).
>>>>
>>>> //Roger
>>>>
>>>> Roger Dahlkvist wrote:
>>>>
>>>>
>>>>         
>>>>> Thanks, any help is greatly appreciated.
>>>>>
>>>>> The status for me right now is that I have switched to newlib-1.17.0.  I also changed both cpukit/wrapup/Makefile.am and c/src/wrapup/Makefile.am, as suggested.
>>>>> That fixed most errors, now that both librtemscpu.a and librtemsbsp.a are added correctly.
>>>>>
>>>>>
>>>>>
>>>>>           
>>>> Then you are probably close to in sync with what I have
>>>> on my laptop at home.
>>>>
>>>>
>>>>         
>>>>> I've tried myself to solve the remaining issues, by including /shared/bsppretaskinghook.c to libbsp.a, and define WorkAreaBase in linkcmds.
>>>>> However the resulting tests fail in the simulator, so that might not have been the right way to go, or there are other things to consider aswell.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>           
>>>> How do the tests fail?
>>>>
>>>> Usually at this point it is memory layout.
>>>>
>>>> All BSPs now use the same framework for BSP initialization
>>>> and most use exactly the same mechanism for dividing memory
>>>> between the RTEMS Workspace and C program heap.
>>>>
>>>> Can you get debug IO via printk at this point on the simulator?
>>>> If so, make printk work and turn on debugging in libbsp/shared/bootcard.c.
>>>>
>>>> If not, or if it is just as easy, trace through bootcard.c in the debugger
>>>> and see if things look ok.
>>>>
>>>> Also (very common), the starting stack (the one the BSP start code
>>>> runs on) may NOT be properly accounted for in the linkcmds and
>>>> your work memory (workspace and C program heap) may overlap it.
>>>> Then the init is clobbering the stack out from underneath you.
>>>>
>>>> _-joel
>>>>
>>>>
>>>>         
>>>>> Looking forward to your inputs.
>>>>>
>>>>> //Roger
>>>>>
>>>>>
>>>>> ________________________________________
>>>>> Fr?n: millibit at gmail.com [millibit at gmail.com] f&#246;r Morten Kristiansen [millibit+rtems at gmail.com]
>>>>> Skickat: den 24 mars 2009 09:30
>>>>> Till: Joel Sherrill
>>>>> Kopia: Roger Dahlkvist; rtems-users at rtems.com
>>>>> ?mne: Re: Nios2 problems when building tests
>>>>>
>>>>> 2009/3/20 Joel Sherrill <joel.sherrill at oarcorp.com>:
>>>>>
>>>>>
>>>>>
>>>>>           
>>>>>> My full NIOS patches are at home.  Morten can help you
>>>>>> get things going.
>>>>>>
>>>>>>
>>>>>>
>>>>>>             
>>>>> I should think so. I'll take a look at it this evening.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>           
>>>>>> The biggest issue you are facing is that the NIOS II binutils
>>>>>> version Altera uses is TOO old for the feature used in cpukit/wrapup.
>>>>>>
>>>>>>
>>>>>>
>>>>>>             
>>>>> I fetched the relevant versions into Subversion and tried merging. It
>>>>> wasn't straightforward, but as a quick solution I can create a patch
>>>>> that allows it to read options from a file.
>>>>>
>>>>> /MK
>>>>>
>>>>>
>>>>>
>>>>>           
>>>> --
>>>> Joel Sherrill, Ph.D.             Director of Research & Development
>>>> joel.sherrill at OARcorp.com        On-Line Applications Research
>>>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>>>>   Support Available             (256) 722-9985
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>         
>> _______________________________________________
>> rtems-users mailing list
>> rtems-users at rtems.com
>> http://rtems.rtems.org/mailman/listinfo/rtems-users
>>
>>     
>
>   


-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel.sherrill at OARcorp.com        On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985