[okl4-developer] Stuck in kmem::free()

[Charles Chiou]

Hi Geoffrey,

I think I may know what's going on. It looks like the content in the BSS are
not cleared. So does the loader need to clear the BSS after loading
image.boot?

I copied the file 'platform/pxa/pistachio/src/head.spp' for my own head.spp.
I suppose something could be added here to clear the sections "kernel.bss"
and "l4test.bss"?

Thanks

Charles


On 7/23/08 12:25 PM, "Geoffrey Lee" <glee at ok-labs.com> wrote:

> On Wed, Jul 23, 2008 at 12:09:41PM +0800, Charles Chiou wrote:
>> Hi Geoffrey,
>> 
>> I'm building with "./tools/build.py machine=a2bub project=l4test" and trying
>> to follow the video "Porting OKL4 to a new SoC".
>> 
>> I've implemented a serial driver that is supposed to run under the OKL4, but
>> since the code hasn't executed that far (the remapping of UART controller's
>> physical address to 0xf9000000 hasn't been run yet).
>> 
>> Do you mean I should write another serial driver based on physical address?
> 
> It may be possible to re-jig the initialization so that the serial
> gets initialized earlier.
> 
> I've just had another look at the assembly dump and it seems to
> correspond to the tracebuffer.
> 
>>>>>> 0xF00043E4 EAFFFFAD    B           0xf00042a0> >>>> 0xF00043E8
>>>>>> E5933010    LDR         R3,[R3,#+0x10]    <--- !!!
>>>>>> 0xF00043EC E3130002    TST         R3,#0x2
>>>>>> 0xF00043F0 0AFFFFA6    BEQ         0xf0004290
>>>>>> 0xF00043F4 E3A0003D    MOV         R0,#0x3d
>>>>>> 0xF00043F8 EB0026AB    BL          _Z12tb_log_eventm ;
> 
> The offset 0x10 would correspond to the log mask, and the tracebuffer
> code looks like so:
> 
> #define TBUF_RECORD_EVENT(tp)
> \
>     if (EXPECT_FALSE(trace_buffer && trace_buffer->log_mask & (1UL <<
> tp##_major
> ))) \
>         tb_log_event(tp)
> 
> It may be worthwhile to further investigate why the tracebuffer
> pointer appears to be bad.
> 
>> 
>> Thanks
>> 
>> Charles
> 
> -gl
> 
>> 
>> On 7/23/08 12:00 PM, "Geoffrey Lee" <glee at ok-labs.com> wrote:
>> 
>>> On Wed, Jul 23, 2008 at 10:06:19AM +0800, Charles Chiou wrote:
>>>> Hi Geoffrey,
>>>> 
>>> 
>>> Hi Charles
>>> 
>>>> I'm loading image.boot to physical 0xc0100000. When it starts execution,
>>>> the
>>>> MMU is turned off.
>>>> 
>>>> I've not modified kernel sources other than platform additions.
>>> 
>>> What is the build command line you use to build your software?
>>> It may be worthwhile writing a simple serial console so that
>>> you can debug the kernel initialization.
>>> 
>>>> Thanks
>>>> 
>>>> Charles
>>> 
>>> -gl
>>> 
>>>> 
>>>> 
>>>> On 7/22/08 11:53 AM, "Geoffrey Lee" <glee at ok-labs.com> wrote:
>>>> 
>>>>> On Sun, Jul 20, 2008 at 05:31:48PM +0800, Charles Chiou wrote:
>>>>>> Hi,
>>>>>> 
>>>>>> I'm in the process of trying to port pistachio to an SoC with an
>>>>>> ARM926EJ-S
>>>>>> core. It seems to run into prefetch abort exception in kmem::free(). I'd
>>>>>> appreciate if somebody could shed some lights on what I'm doing wrong.
>>>>> 
>>>>> Hi Charles
>>>>> 
>>>>> How did you load your binary to the system and which address did you
>>>>> load it to?  Have you made any other modifications to the OKL4 source
>>>>> code other than adding in your platfrom description?
>>>>> 
>>>>> -gl
>>>>> 
>>>>>> 
>>>>>> The place it occurs is (where r3 is 0x10000100):
>>>>>> 
>>>>>> _ZN6kmem_t4freeEPvm:
>>>>>> 0xF0004268 E92D40F0    STMFD       SP!,{R4-R7,LR}
>>>>>> 0xF000426C E59F318C    LDR         R3,[PC,#+0x18c]
>>>>>> 0xF0004270 E5933000    LDR         R3,[R3,#+0]
>>>>>> 0xF0004274 E24DD010    SUB         SP,SP,#0x10
>>>>>> ...
>>>>>> 0xF00043E4 EAFFFFAD    B           0xf00042a0
>>>>>> 0xF00043E8 E5933010    LDR         R3,[R3,#+0x10]    <--- !!!
>>>>>> 0xF00043EC E3130002    TST         R3,#0x2
>>>>>> 0xF00043F0 0AFFFFA6    BEQ         0xf0004290
>>>>>> 0xF00043F4 E3A0003D    MOV         R0,#0x3d
>>>>>> 0xF00043F8 EB0026AB    BL          _Z12tb_log_eventm ;
>>>>>> _Z12tb_log_eventm=0xf000deac
>>>>>> 0xF00043FC EAFFFFA3    B           0xf0004290
>>>>>> 
>>>>>> It looks like kmem:free() is called with (start = 0xf0022590, end =
>>>>>> 0xf0100000).
>>>>>> 
>>>>>> I'm stuck before the console is up, which is makes debugging very
>>>>>> difficult
>>>>>> :( However, I also think the init hasn't run into where I mapped the
>>>>>> physical memory of my uart to 0xf9000000 (writing there doesn't output
>>>>>> anything).
>>>>>> 
>>>>>> This is my machine.py:
>>>>>> 
>>>>>> class a2(arm926ejs):
>>>>>>     platform = "ambarella"
>>>>>> 
>>>>>> class a2bub(a2):
>>>>>>     virtual = False
>>>>>>     memory = a2.memory.copy()
>>>>>>     memory['physical'] =     [Region(0xc0100000L, 0xc1000000L)]
>>>>>>     bootimage = "binary"
>>>>>> 
>>>>>> 
>>>>>> Thanks for the help!
>>>>>> 
>>>>>> Charles
>>>>>> 
>>>>>> 
>>>>>> _______________________________________________
>>>>>> Developer mailing list
>>>>>> Developer at okl4.org
>>>>>> https://lists.okl4.org/mailman/listinfo/developer
>>>>>> 
>>>> 
>>>> 
>> 
>>