[Crash-utility] crash failed to get backtrace in module on powerpc target(e500v2)

[Dave Anderson]

----- Original Message -----
>   
> Hi Experts:
> 
> I met a issue when try to analysis vmcore which was triggered by a error in module on powerpc target. 
> 
> BG:
> Crash version is 6.1.4, I cross build it with powerpc toolchain, and run it on p2020(e500v2) target directly.
> 
> 
> It show the following error:
> /*************************************************************************/
> crash> bt 
> PID: 588    TASK: eed6e880  CPU: 1   COMMAND: "insmod" 
> R0:  c0097cfc   R1:  eed81e80   R2:  eed6e880   R3:  eed81db8    
> R4:  00021000   R5:  c07955df   R6:  000025d1   R7:  c0793bb0    
> R8:  00001a2f   R9:  00000000   R10: 00000020   R11: eed80000    
> R12: 24000044   R13: 100a73a0   R14: 00000000   R15: 00000000    
> R16: 00000000   R17: 00000000   R18: 00000000   R19: 00000000    
> R20: 00000000   R21: 00000000   R22: 00000000   R23: 00000000    
> R24: 00000000   R25: 00000000   R26: 00028b70   R27: f23ea000    
> R28: 00000000   R29: eed81db8   R30: c0753178   R31: c0790000    
> NIP: c05c7e9c   MSR: 00021000   OR3: eed81e48   CTR: 00000001 
> LR:  c05c7e9c   XER: 20000000   CCR: 24000042   MQ:  c0007fdc 
> DAR: 00000000 DSISR: 00000000        Syscall Result: 00000000 
>  NIP [00000000c05c7e9c] panic 
>  LR  [00000000c05c7e9c] panic 
>  #0 [eed81e80] panic at c05c7e9c 
> bt: ppc_back_trace hit unknown symbol (f23e70b4). 
> /*************************************************************************/
> 
> I doubt if the crash can not worked well on e500v2 target for the moment?
> 
> BTW, In my module, module_init will call a function, that function will call panic and cause kdump, but from the log when issue happen, looks it did not show the right call trace:
> /*************************************************************************/
> [eed81e80] [c05c7e94] panic+0x12c/0x1ac                                        
> [eed81ed0] [f23e70b4] dummy_cleanup_module+0x0/0x44 [dummy]                    
> [eed81ee0] [f23ea030] dummy_init_module+0x30/0x68 [dummy]                      
> [eed81ef0] [c0002124] do_one_initcall+0x3c/0x1e0                               
> [eed81f20] [c008e698] sys_init_module+0xf8/0x21c 
> /*************************************************************************/
> Do you have some comments on this kind of secne?
> 
> Thanks a lot for your valuable time!

Hello James,

I believe that this error is due to the fact that the dummy module's
symbols have not been registered with the kernel, because the module 
forced the panic too early.  If you do a "sym -M", you probably will
not see the "dummy" module's symbols, and therefore crash cannot resolve
the f23e70b4 address.

After do_one_initcall() is called normally, there is still a lot of further
module initialization to be accomplished.  In this case, it is not being done,
and the dummy module's "module" structure has not been fully initialization. 

Dave


 
> Best Regards,
> James
> 
> Liu Jianbo | WIND RIVER | Senior Engineer - Technical Support
> Tel 86 28 65318098 | Cell 86 13558641588 | Fax 86 28 65319983  
>