Re: [PATCH v4] xmalloc: add support for checking the pool integrity

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 16/12/14 23:37, Julien Grall wrote:
>
>
> On 16/12/2014 23:26, Andrew Cooper wrote:
>> On 16/12/2014 23:06, Julien Grall wrote:
>>> Hi,
>>>
>>> On 16/12/2014 20:28, Andrew Cooper wrote:
>>>> I suspect you also would be better, and certainly more brief, with
>>>> "run_in_exception_handler(show_stack)" instead, which will just
>>>> print a
>>>> stack trace, but nothing more.
>>>
>>> FIY, run_in_exception_handler doesn't exists on ARM.
>>>
>>> Regards,
>>>
>>
>> In which case it even more lucky that __bug() and __warn() are orphaned
>> functions, as they don't work correctly on arm.  Even more reason to
>> remove them.
>>
>> It doesn't look like run_in_exception_handler() would be hard to add to
>> arm at all.  It already has broadly similar bug/warn/assert
>> infrastructure.
>
> It's more difficult than you think. For ARM we had to use a different
> infrastructure than x86 to setup the BUG_FRAME. This is because %c is
> not correctly support on most ARM compiler today.
>
> So I don't think it worth to spend time on it just for one call.
>
> How about introducing dump_stack() which would call
> run_in_exception_handler() on x86 and something different (maybe a new
> BUG_FRAME type) on ARM?

Introducing a new bugframe is precicely what I meant by "this doesn't
look hard".  x86 currently has one more bugframe than arm, being
BUGFRAME_run_fn.

At this point, run_in_exception_handler() effectively becomes common.

~Andrew