[CRIU] [PATCH 2/2] compel: Add missing 32s support

[Dmitry Safonov]

On 04/19/2016 03:03 PM, Cyrill Gorcunov wrote:
> On Tue, Apr 19, 2016 at 03:00:00PM +0300, Dmitry Safonov wrote:
>> On 04/19/2016 02:52 PM, Cyrill Gorcunov wrote:
>>> On Tue, Apr 19, 2016 at 02:29:49PM +0300, Dmitry Safonov wrote:
>>>> To be correct - after this patch from the second part, which
>>>> adds this kind of relocation:
>>>> https://github.com/0x7f454c46/criu/commit/7f1eb57e9e376dacc83cb9f96ebe55c02768713a
>>>> Before I didn't get this relocation on compile before it.
>>> mov will produce absolute address which should be addressed
>>> during relocation resolving, but I would rather stick with
>>> lea and figure out why it get broken earlier?
>> That's because parasite_service declared as:
>> int __used __parasite_entry parasite_service(unsigned int cmd, void *args)
>>
>> So the first parameter is number, not a pointer to number.
>> And for the x86_64 we're doing the same:
>>>     movl    __export_parasite_cmd(%rip), %edi
>>>     leaq    __export_parasite_args(%rip), %rsi
>>>     call    parasite_service
>> That's omitting broken offset (label 1/2 change) :)
> Heh, I managed to forget the statement ;)
> fwiw lea get used pretty common for basic
> arithmetic as well

Hehe, yeah. I remember, it was quite fun to debug :)

-- 
Regards,
Dmitry Safonov