[PATCH 0/3] makedumpfile: Reduce the risk of OOM.

[Atsushi Kumagai]

>Hi Atsushi,
>
>Since complaint has been got from our user about the oom issue, and it's
>still a little longer time to get to 1.5.7, could you please merge this
>patchset firstly to devel branch, then we can back port it with a valid
>commit id which will be consistent?

I've done it, here you are:

http://git.code.sf.net/p/makedumpfile/code
    - [PATCH 3/3] Stop maximizing the bitmap buffer to reduce the risk of OOM. (by Atsushi Kumagai) 0b73282
    - [PATCH 2/3] Move counting pfn_memhole for cyclic mode. (by Atsushi Kumagai) 2648a8f
    - [PATCH 1/3] Remove the 1st bitmap buffer from the ELF path in cyclic mode. (by Atsushi Kumagai) 16b94ab


Thanks
Atsushi Kumagai


>Thanks
>Baoquan
>
>On 06/10/14 at 04:51am, Atsushi Kumagai wrote:
>> Hello,
>>
>> makedumpfile tries maximizing the bitmap buffer size to get the
>> best performance, but it occasionally cause OOM because the safe
>> limit is variable and it can be too aggressive depending on the
>> situation.
>>
>> I think reliability is more important than performance for dump
>> tools, so I decided to remove the maximization. I don't worry
>> about performance degradation, please see [PATCH 3/3].
>>
>> If there is no objection, I'll merge this patch into v1.5.7.
>>
>> BTW, an expert user still can try to maximize the bitmap buffer with
>> --cyclic-buffer option.
>>
>> Atsushi Kumagai (3):
>>       Remove the 1st bitmap buffer from the ELF path in cyclic mode.
>>       Move counting pfn_memhole for cyclic mode.
>>       Stop maximizing the bitmap buffer to reduce the risk of OOM.
>>
>>  makedumpfile.8 |   1 -
>>  makedumpfile.c | 117
>++++++++++++++++++++------------------------------------------------------------------------------
>>  print_info.c   |   1 -
>>  3 files changed, 24 insertions(+), 95 deletions(-)
>>
>> _______________________________________________
>> kexec mailing list
>> kexec at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/kexec
>
>_______________________________________________
>kexec mailing list
>kexec at lists.infradead.org
>http://lists.infradead.org/mailman/listinfo/kexec