[okl4-developer] tests failing in l4test and test_libs=all; is it a problem?

[Nelson Tam]

Hi Gabi,

On 30/04/2008, at 3:34 PM, Gabi Voiculescu wrote:
>
> I can run both project=iguana test_libs=all and project=l4test with  
> the caches turned of.
>
> I get two failures:
>
> - in test_libs=all
>
> - in l4test:
>
> Are the two tests an indication about something wrong in my port?
>
> The PD2200 test can cause problems for MIPS, and I did not try very  
> hard to find out why it fails.

You can safely ignore the warning in PD2200, it does not indicate any  
problems with the port.

> From what I understand, the ARM_FASS01 test uses a FASS feature that  
> is v5 specific, not v6 as in my arm1176.

That's correct.  You can safely ignore this one also.

> Here are the most impotant ones that affected my port (you may  
> consider to revise for the new release):
>
> had to change arch/arm/pistachio/cpu arm1176jz folder, by copying  
> the content from arm1136js,  adding  CONFIGURE_ARM_TRUSTZONE macro  
> and comment out CONFIGURE_ARM_VFP. Comment out   
> CONFIG_ARM_THREAD_REGISTERS. Also commented out the L2 cache macros  
> in syscon.h.
>
> The original arch/arm/.../include/cache.h had some bugs in arm1176  
> original folder so I had to start with a fresh copy from arm1136 to  
> get it working.
>
> Had to change arch/arm/tools/machines.py to have the class  
> arm1176jzf search for the cpu/arm1176jz (so changed the cpu line  
> into cpu ="arm1176jz")
>
> Next, change arch/arm/.../v6/src/space.cc and tcb.h to have the line:
> #include <kernel/ver/utcb.h>
> before the line where USER_UTCB_REF is on the left side of  
> attribution.
>
> Next, ensure I have in platform/..../tools/machines.py zero_bss =  
> True.
> Otherwise I get weird roottask page fault errors on global variables  
> that were or were not initialized before. In my case l4test.bss  
> section has 2kB initialized with 0xffffffff and the following 2k  
> with 0x0 and so on repeated. The 0xffffffff affected the  
> initialization of freep in malloc, bootinfo fed into the roottask,  
> content of pointer to pointer given by USER_UTCB_REF and a lot more.
>
> Personally I need to convert the elf.nobits file into a binary to  
> load onto my board.
> So I use arm-linux-objdump -O binary to the elf.nobits file not to  
> the elf file. the elf.nobits is created by calling elfveawer with  
> argument --nobits on the .elf file, but the zero_bss=True makes that  
> automatic.
>
> Of course, all it remains to do is the port of the registers,  
> interrupt numbers and physical addresses.

Thanks for sharing this with us!
--
(nt)

Nelson Tam

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1560 bytes
Desc: not available
Url : http://lists.okl4.org/pipermail/developer/attachments/20080502/b5c0b269/attachment.bin