Re: [PATCH v2 0/3] optimize kdump service start time in case of no initramfs rebuild
On 05/03/2017 at 05:21 PM, Xunlei Pang wrote:
On 05/03/2017 at 05:09 PM, Dave Young wrote:
> Hi, Xunlei
> On 05/03/17 at 01:10pm, Xunlei Pang wrote:
>> kdump service takes a few seconds to start even no need to rebuild initramfs,
>> which is a little longer compared to other systemd services, it may affects
>> the system boot time a little, we better optimize it as possible as we can.
>>
>> The main root cause of this is the use of "lsinitrd" which is
time-consuming.
>>
>> This series is actually reducing the "lsinitrd" uses.
>>
>> Tested on Intel(R) Core(TM) i7-5600U CPU @ 2.60GHz.
>> Before this patchset:
>> $ time kdumpctl start
>> kexec: loaded kdump kernel
>> Starting kdump: [OK]
>>
>> real 0m5.156s
>> user 0m5.169s
>> sys 0m1.269s
>>
>> After this patchset:
>> $ time kdumpctl start (when kdump lsinitrd tmp file exists)
>> kexec: loaded kdump kernel
>> Starting kdump: [OK]
>>
>> real 0m0.854s
>> user 0m0.389s
>> sys 0m0.582s
>>
>> Xunlei Pang (3):
>> kdumpctl: remove is_mode_switched()
>> kdumpctl: improve check_wdt_modified()
>> introduce kdump lsinitrd support
> Nice work, read all the 4 patches, overall they looks good, but for
> the third one kdump saved initrd, I'm still not sure if we should add
> these, if we add in the future we could add more. Also where to add the
> tmp file is also a problem.
>
> With the patch 4, even we do not use patch 3, it will still within 1
> second?
It depends on the CPU capacity, on my laptop(4-core 2.60GHz) without patch3
("introduce kdump lsinitrd support"):
kexec: loaded kdump kernel
Starting kdump: [OK]
real 0m1.158s
user 0m1.039s
sys 0m0.305s
One single "lsinitrd -f" consumes around 0.6 second on my machine.
sorry, should be "s/0.6 second/0.35 second".
>> dracut-module-setup.sh | 5 +++
>> kdump-lib.sh | 26 +++++++++++++++
>> kdumpctl | 88 +++++++++++++++++++++++---------------------------
>> mkdumprd | 7 ++++
>> 4 files changed, 78 insertions(+), 48 deletions(-)
>>
>> --
>> 1.8.3.1
>>
> Thanks
> Dave