linux-kernel - Re: [RFC v4 0/2] CPU-Idle latency selftest framework

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 13 Apr 2021 07:52:14 -0700
From:   Doug Smythies <dsmythies@...us.net>
To:     Pratik Rajesh Sampat <psampat@...ux.ibm.com>
Cc:     rjw@...ysocki.net, Daniel Lezcano <daniel.lezcano@...aro.org>,
        shuah@...nel.org, ego@...ux.vnet.ibm.com, svaidy@...ux.ibm.com,
        Linux PM list <linux-pm@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-kselftest@...r.kernel.org, pratik.r.sampat@...il.com,
        dsmythies <dsmythies@...us.net>
Subject: Re: [RFC v4 0/2] CPU-Idle latency selftest framework

Hi Pratik,

V4 seems fine. Thank you.

On Mon, Apr 12, 2021 at 12:43 AM Pratik Rajesh Sampat
<psampat@...ux.ibm.com> wrote:
>
> Changelog v3-->v4
> Based on review comments by Doug Smythies,
> 1. Parsing the thread_siblings_list for CPU topology information to
>    correctly identify the cores the test should run on in
>    default(quick) mode.
> 2. The source CPU to source CPU interaction in the IPI test will always
>    result in a lower latency and cause a bias in the average, hence
>    avoid adding the latency to be averaged for same cpu IPIs. The
>    latency will still be displayed in the detailed logs.
>
> RFC v3: https://lkml.org/lkml/2021/4/4/31

Example output for an Intel i5-10600K, HWP active, performance mode.
System very idle:

$ sudo ./cpuidle.sh -v -i
Inserting /lib/modules/5.12.0-rc7-prs-v4/kernel/drivers/cpuidle/test-cpuidle_latency.ko
module

--IPI Latency Test---
Baseline Avg IPI latency(ns): 686
Observed Avg IPI latency(ns) - State 0: 468
Observed Avg IPI latency(ns) - State 1: 956
Observed Avg IPI latency(ns) - State 2: 17936
Observed Avg IPI latency(ns) - State 3: 17968

--Timeout Latency Test--
Baseline Avg timeout diff(ns): 445
Observed Avg timeout diff(ns) - State 0: 377
Observed Avg timeout diff(ns) - State 1: 630
Observed Avg timeout diff(ns) - State 2: 322812
Observed Avg timeout diff(ns) - State 3: 306067
Removing /lib/modules/5.12.0-rc7-prs-v4/kernel/drivers/cpuidle/test-cpuidle_latency.ko
module
Full Output logged at: cpuidle.log

$ grep . /sys/devices/system/cpu/cpu7/cpuidle/state*/residency
/sys/devices/system/cpu/cpu7/cpuidle/state0/residency:0
/sys/devices/system/cpu/cpu7/cpuidle/state1/residency:1
/sys/devices/system/cpu/cpu7/cpuidle/state2/residency:360
/sys/devices/system/cpu/cpu7/cpuidle/state3/residency:3102
$ grep . /sys/devices/system/cpu/cpu7/cpuidle/state*/latency
/sys/devices/system/cpu/cpu7/cpuidle/state0/latency:0
/sys/devices/system/cpu/cpu7/cpuidle/state1/latency:1
/sys/devices/system/cpu/cpu7/cpuidle/state2/latency:120
/sys/devices/system/cpu/cpu7/cpuidle/state3/latency:1034

... Doug