Re: [PATCH] more sanity checks in Dwarf2 unwinder

[Jan Beulich]

>>  	while (unwind(info) == 0 && UNW_PC(info)) {
>>  		n++;
>>  		oad->ops->address(oad->data, UNW_PC(info));
>>  		if (arch_unw_user_mode(info))
>>  			break;
>> +		if ((sp & ~(PAGE_SIZE - 1)) == (UNW_SP(info) & ~(PAGE_SIZE - 1))
>> +		    && sp > UNW_SP(info))
>> +			break;
>
>Hmm, but that wouldn't catch the case when the SP is completely
>corrupted for some reason.
>Maybe it would be better to just run a brute force check here like 
>the old in_exception_stack(). Similar on x86-64.

Correct. Even though I know Linus disagrees here, I'm not sure
I want to do this, as my ultimate goal would be to eliminate the
hand-crafted linking (which we know got broken a few times on
x86-64, because it's so easy to forget about).
Not the least of the reasons for this is that this increases the
chances of stucks.

>> +	if (UNW_PC(frame) % state.codeAlign
>> +	    || UNW_SP(frame) % sleb128abs(state.dataAlign)
>> +	    || (pc == UNW_PC(frame) && sp == UNW_SP(frame)))
>> +		return -EIO;
>
>Would it be possible to add printks for the EIOs? We want to know 
>when dwarf2 is corrupted.

Certainly, will be a follow-up patch.

Jan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/