| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Wed, 25 Dec 2013 14:32:07 +0800
From: Alex Shi <alex.shi@...aro.org>
To: Fengguang Wu <fengguang.wu@...el.com>
CC: Yuanhan Liu <yuanhan.liu@...el.com>,
"Huang, Ying" <ying.huang@...el.com>, lkp@...org,
LKML <linux-kernel@...r.kernel.org>
Subject: Re: [sched] perf-stat.cpu-migrations changes in dd write test
On 12/25/2013 11:56 AM, Fengguang Wu wrote:
> On Wed, Dec 25, 2013 at 11:30:27AM +0800, Alex Shi wrote:
>> On 12/23/2013 01:24 PM, Fengguang Wu wrote:
>>> On Mon, Dec 23, 2013 at 01:19:04PM +0800, Alex Shi wrote:
>>>> On 12/22/2013 08:30 AM, fengguang.wu@...el.com wrote:
>>>>> Alex,
>>>>>
>>>>> We noticed some perf-stat.cpu-migrations changes by your commit
>>>>
>>>> Thanks for your wonder testing, Fengguang!
>>>> How many CPU in your system and how many dd task running?
>>>
>>> For the case
>>>
>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>
>>> It's a KVM with 4 CPUs, 16G memory. There are 4 virtual disks, on top
>>> of each disk there runs 10 dd tasks.
>>
>> I don't know much of kvm. Where the data get from, VM guest or VM host?
>
> The data are collected in the VM guest.
do you have similar testing on native linux?
>
> Thanks,
> Fengguang
>
>> Looks like any guest cpu usage will be counted as user usage in vmstat.
>> and Does the CS and cpu migration reducing cause dd performance increase?
>>>
>>> Thanks,
>>> Fengguang
>>>
>>>>>
>>>>> 6eb8b571283e64e96ca1a91aad085f9e08f857f0 is the first bad commit
>>>>> commit 6eb8b571283e64e96ca1a91aad085f9e08f857f0
>>>>> Author: Alex Shi <alex.shi@...aro.org>
>>>>> Date: Tue Nov 19 20:29:51 2013 +0800
>>>>>
>>>>> sched: remove rq->cpu_load[load_idx] array
>>>>>
>>>>> Since load_idx effect removed in load balance, we don't need the
>>>>> load_idx decays in scheduler. that will save some process in sched_tick
>>>>> and others places.
>>>>>
>>>>> Signed-off-by: Alex Shi <alex.shi@...aro.org>
>>>>>
>>>>>
>>>>> Here are the changed stats comparing to its parent commit:
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 37779 ~40% +115.2% 81288 ~50% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 37779 +115.2% 81288 TOTAL proc-vmstat.pgscan_direct_normal
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: proc-vmstat.pgscan_direct_normal
>>>>>
>>>>> 160000 ++----------------------------------------------------------------+
>>>>> | |
>>>>> 140000 ++ O
>>>>> | |
>>>>> 120000 ++ |
>>>>> | O |
>>>>> 100000 ++ |
>>>>> | |
>>>>> 80000 ++ O |
>>>>> | ..*
>>>>> 60000 ++ .... |
>>>>> | .... |
>>>>> 40000 *+............... O .... |
>>>>> O *............... ........*. |
>>>>> 20000 ++-------------------------------*--------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 121579 ~ 6% -50.8% 59852 ~13% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 121579 -50.8% 59852 TOTAL softirqs.SCHED
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: softirqs.SCHED
>>>>>
>>>>> 140000 ++----------------------------------------------------------------+
>>>>> 130000 ++ ....*........... ....*....... |
>>>>> | ........ .... ........ ...... |
>>>>> 120000 *+... *.... ..|
>>>>> 110000 ++ *
>>>>> | |
>>>>> 100000 ++ |
>>>>> 90000 ++ |
>>>>> 80000 ++ |
>>>>> | |
>>>>> 70000 ++ O O |
>>>>> 60000 ++ O |
>>>>> | O
>>>>> 50000 O+ |
>>>>> 40000 ++----------------------------------------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 5636 ~ 4% -22.1% 4392 ~17% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 5636 -22.1% 4392 TOTAL proc-vmstat.nr_tlb_remote_flush
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: proc-vmstat.nr_tlb_remote_flush
>>>>>
>>>>> 6500 ++------------------------------------------------------------------+
>>>>> | |
>>>>> 6000 ++ ........*............ |
>>>>> *........ .... |
>>>>> 5500 ++ O *................*................|
>>>>> | *
>>>>> 5000 ++ O |
>>>>> | |
>>>>> 4500 ++ O |
>>>>> | |
>>>>> 4000 ++ |
>>>>> | O
>>>>> 3500 O+ |
>>>>> | |
>>>>> 3000 ++------------------------------------------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 5679 ~ 4% -21.9% 4437 ~17% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 5679 -21.9% 4437 TOTAL proc-vmstat.nr_tlb_remote_flush_received
>>>>>
>>>>>
>>>>>
>>>>> 6500 ++------------------------------------------------------------------+
>>>>> | |
>>>>> 6000 ++ ........*........ |
>>>>> *........ ...... |
>>>>> | .. ........*................|
>>>>> 5500 ++ O *........ *
>>>>> | |
>>>>> 5000 ++ O |
>>>>> | |
>>>>> 4500 ++ O |
>>>>> | |
>>>>> | |
>>>>> 4000 ++ |
>>>>> | O
>>>>> 3500 O+------------------------------------------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 167 ~11% +19.4% 200 ~ 4% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 167 +19.4% 200 TOTAL uptime.idle
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: uptime.idle
>>>>>
>>>>> 220 ++-------------------------------------------------------------------+
>>>>> | O
>>>>> 210 ++ |
>>>>> 200 O+ O |
>>>>> | ..*..... O |
>>>>> 190 ++ .... O ... |
>>>>> 180 ++ .... .... |
>>>>> | .... .... |
>>>>> 170 *+. . .*
>>>>> 160 ++ *...... .... |
>>>>> | .... ... |
>>>>> 150 ++ .... ... |
>>>>> 140 ++ .. .... |
>>>>> | *. |
>>>>> 130 ++-------------------------------------------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 154781 ~ 5% -66.8% 51449 ~ 8% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 154781 -66.8% 51449 TOTAL perf-stat.cpu-migrations
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: perf-stat.cpu-migrations
>>>>>
>>>>> 180000 ++----------------------------------------------------------------+
>>>>> | ....*........... |
>>>>> 160000 ++ ........ .... ........*...............|
>>>>> *.... *........ *
>>>>> 140000 ++ |
>>>>> | |
>>>>> 120000 ++ |
>>>>> | |
>>>>> 100000 ++ |
>>>>> | |
>>>>> 80000 ++ |
>>>>> | |
>>>>> 60000 ++ O |
>>>>> | O O O
>>>>> 40000 O+----------------------------------------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 1943 ~ 3% -7.8% 1792 ~ 4% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 1943 -7.8% 1792 TOTAL vmstat.system.cs
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: vmstat.system.cs
>>>>>
>>>>> 2050 ++------------------------------------------------------------------+
>>>>> | ........*................*..... |
>>>>> 2000 ++ .*........ ... |
>>>>> 1950 ++ .... ... |
>>>>> | ... .... |
>>>>> 1900 ++ ... .|
>>>>> 1850 ++.... O O *
>>>>> *. |
>>>>> 1800 ++ O O
>>>>> 1750 ++ |
>>>>> | |
>>>>> 1700 ++ |
>>>>> 1650 O+ |
>>>>> | |
>>>>> 1600 ++------------------------------------------------------------------+
>>>>>
>>>>>
>>>>> 1de157d718f6494 6eb8b571283e64e96ca1
>>>>> --------------- -------------------------
>>>>> 1216177 ~ 3% -9.4% 1101605 ~ 7% kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd
>>>>> 1216177 -9.4% 1101605 TOTAL perf-stat.context-switches
>>>>>
>>>>> kbuildx/micro/dd-write/4HDD-JBOD-cfq-ext4-10dd: perf-stat.context-switches
>>>>>
>>>>> 1.3e+06 ++--------------------------------------------------------------+
>>>>> | ..*............... |
>>>>> 1.25e+06 ++ ...... *...............*....... |
>>>>> | ..... ...... |
>>>>> 1.2e+06 *+. O ..|
>>>>> 1.15e+06 ++ O *
>>>>> | O |
>>>>> 1.1e+06 ++ |
>>>>> | O
>>>>> 1.05e+06 ++ |
>>>>> 1e+06 ++ |
>>>>> | |
>>>>> 950000 O+ |
>>>>> | |
>>>>> 900000 ++--------------------------------------------------------------+
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Thanks
>>>> Alex
>>
>>
>> --
>> Thanks
>> Alex
--
Thanks
Alex
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists