[PATCH 2/2] makedumpfile: make the incomplete vmcore generated by ENOSPC error analyzable

[HATAYAMA, Daisuke]

Hello Wang,

(2014/09/18 16:26), Wang, Xiao/Wang Xiao wrote:
> Hello Petr,
>
> Thanks for your reply.
>
> On 2014/9/17 14:47, Petr Tesarik wrote:
>>> Since the incomplete vmcore generated by ENOSPC error can't be anylyzed by
>>> >crash utility, but sometimes this file may contain important information
>> Hello 王萧,
>>
>> first of all, I like your patch. It has always annoyed me that any
>> incomplete dump was utterly useless. Just one small question: why is
>> this limited to ENOSPC? Why not e.g. EIO? EFBIG?
>
> At present, I just consider ENOSPC for this error maybe more common. And I'm
> considering to deal with other situations in the coming future.
>

For EFBIG, it seems to me enough to do the same thing as for ENOSPC fow now.

For EIO, although I'm not perfectly clear about when EIO is returned, but
such situation would include broken disk, and then we no longer trust disk
behaviour so we should stop makedumpfile from running as soon as possible,
and we should bet we can successfully recover the incomplete dump data
from the broken disk.

-- 
Thanks.
HATAYAMA, Daisuke