Re: PROBLEM: ARM-dma-mapping-fix-for-speculative-prefetching cause OOPS

[Dan Williams <dan.j.williams@intel.com>]

On Thu, Jun 30, 2011 at 4:28 AM, Russell King <rmk@arm.linux.org.uk> wrote:
> On Thu, Jun 30, 2011 at 07:16:24PM +0800, Qin Dehua wrote:
>> Commit 2ffe2da3e follows v2.6.32, the message is from kernel build on
>> commit 2ffe2da3e.
>>
>> The config has CONFIG_BUG=y and CONFIG_DEBUG_BUGVERBOSE=y, but the
>> message is Oops, not BUG() macro, so they don't have line number.
>
> In that case, the raid5 code contains an explicit NULL pointer
> dereference which isn't a BUG() - the code line disassembles to:
>
>   0:   ebfff1bc        bl      0xffffc6f8
>   4:   e28dd044        add     sp, sp, #68     ; 0x44
>   8:   e8bd8ff0        pop     {r4, r5, r6, r7, r8, r9, sl, fp, pc}
>   c:   e3a03000        mov     r3, #0  ; 0x0
>  10:   e5833000        str     r3, [r3] <=== faulting instruction
>
> So, if you're saying that's not a BUG(), then I don't know what it is
> and I'm afraid I can't help because the oops doesn't make any sense
> to me.
>

QinDehua,

Can you rebuild with CONFIG_DEBUG_INFO=y, reproduce the crash and then
send the output of:

$ gdb drivers/md/raid5.o
(gdb) li *(raid5d+0x580)
(gdb) li *(__release_stripe+0x1e4)
etc...

...those offsets might change so just grab whatever "PC is at "
reports in the oops.

--
Dan