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]
Message-ID: <9bbf6989-f41f-4533-a7c8-b274744663cd@arm.com>
Date: Mon, 5 Aug 2024 15:53:35 +0100
From: Christian Loehle <christian.loehle@....com>
To: "Rafael J. Wysocki" <rafael@...nel.org>, stable@...r.kernel.org
Cc: linux-pm@...r.kernel.org, linux-kernel@...r.kernel.org,
 vincent.guittot@...aro.org, qyousef@...alina.io, peterz@...radead.org,
 daniel.lezcano@...aro.org, ulf.hansson@...aro.org, anna-maria@...utronix.de,
 dsmythies@...us.net, kajetan.puchalski@....com, lukasz.luba@....com,
 dietmar.eggemann@....com
Subject: Re: [PATCHv3 0/3] cpuidle: teo: Fixing utilization and intercept
 logic

On 6/28/24 20:06, Rafael J. Wysocki wrote:
> On Fri, Jun 28, 2024 at 12:02 PM Christian Loehle
> <christian.loehle@....com> wrote:
>>
>> Hi all,
>> so my investigation into teo lead to the following fixes.
>>
>> 1/3:
>> As discussed the utilization threshold is too high while
>> there are benefits in certain workloads, there are quite a few
>> regressions, too. Revert the Util-awareness patch.
>> This in itself leads to regressions, but part of it can be offset
>> by the later patches.
>> See
>> https://lore.kernel.org/lkml/CAKfTPtA6ZzRR-zMN7sodOW+N_P+GqwNv4tGR+aMB5VXRT2b5bg@mail.gmail.com/
>> 2/3:
>> Remove the 'recent' intercept logic, see my findings in:
>> https://lore.kernel.org/lkml/0ce2d536-1125-4df8-9a5b-0d5e389cd8af@arm.com/
>> I haven't found a way to salvage this properly, so I removed it.
>> The regular intercept seems to decay fast enough to not need this, but
>> we could change it if that turns out that we need this to be faster in
>> ramp-up and decaying.
>> 3/3:
>> The rest of the intercept logic had issues, too.
>> See the commit.
>>
>> Happy for anyone to take a look and test as well.
>>
>> Some numbers for context, comparing:
>> - IO workload (intercept heavy).
>> - Timer workload very low utilization (check for deepest state)
>> - hackbench (high utilization)
>> - Geekbench 5 on Pixel6 (high utilization)
>> Tests 1 to 3 are on RK3399 with CONFIG_HZ=100.
>> target_residencies: 1, 900, 2000
>>
>> 1. IO workload, 5 runs, results sorted, in read IOPS.
>> fio --minimal --time_based --name=fiotest --filename=/dev/nvme0n1 --runtime=30 --rw=randread --bs=4k --ioengine=psync --iodepth=1 --direct=1 | cut -d \; -f 8;
>>
>> teo fixed v2:
>> /dev/nvme0n1
>> [4599, 4658, 4692, 4694, 4720]
>> /dev/mmcblk2
>> [5700, 5730, 5735, 5747, 5977]
>> /dev/mmcblk1
>> [2052, 2054, 2066, 2067, 2073]
>>
>> teo mainline:
>> /dev/nvme0n1
>> [3793, 3825, 3846, 3865, 3964]
>> /dev/mmcblk2
>> [3831, 4110, 4154, 4203, 4228]
>> /dev/mmcblk1
>> [1559, 1564, 1596, 1611, 1618]
>>
>> menu:
>> /dev/nvme0n1
>> [2571, 2630, 2804, 2813, 2917]
>> /dev/mmcblk2
>> [4181, 4260, 5062, 5260, 5329]
>> /dev/mmcblk1
>> [1567, 1581, 1585, 1603, 1769]
>>
>>
>> 2. Timer workload (through IO for my convenience 😉 )
>> Results in read IOPS, fio same as above.
>> echo "0 2097152 zero" | dmsetup create dm-zeros
>> echo "0 2097152 delay /dev/mapper/dm-zeros 0 50" | dmsetup create dm-slow
>> (Each IO is delayed by timer of 50ms, should be mostly in state2, for 5s total)
>>
>> teo fixed v2:
>> idle_state time
>> 2.0     4.807025
>> -1.0    0.219766
>> 0.0     0.072007
>> 1.0     0.169570
>>
>> 3199 cpu_idle total
>> 38 cpu_idle_miss
>> 31 cpu_idle_miss above
>> 7 cpu_idle_miss below
>>
>> teo mainline:
>> idle_state time
>> 1.0     4.897942
>> -1.0    0.095375
>> 0.0     0.253581
>>
>> 3221 cpu_idle total
>> 1269 cpu_idle_miss
>> 22 cpu_idle_miss above
>> 1247 cpu_idle_miss below
>>
>> menu:
>> idle_state time
>> 2.0     4.295546
>> -1.0    0.234164
>> 1.0     0.356344
>> 0.0     0.401507
>>
>> 3421 cpu_idle total
>> 129 cpu_idle_miss
>> 52 cpu_idle_miss above
>> 77 cpu_idle_miss below
>>
>> Residencies:
>> teo mainline isn't in state2 at all, teo fixed is more in state2 than menu, but
>> both are in state2 the vast majority of the time as expected.
>>
>> tldr: overall teo fixed spends more time in state2 while having
>> fewer idle_miss than menu.
>> teo mainline was just way too aggressive at selecting shallow states.
>>
>> 3. Hackbench, 5 runs
>> for i in $(seq 0 4); do hackbench -l 100 -g 100 ; sleep 1; done
>>
>> teo fixed v2:
>> Time: 4.937
>> Time: 4.898
>> Time: 4.871
>> Time: 4.833
>> Time: 4.898
>>
>> teo mainline:
>> Time: 4.945
>> Time: 5.021
>> Time: 4.927
>> Time: 4.923
>> Time: 5.137
>>
>> menu:
>> Time: 4.964
>> Time: 4.847
>> Time: 4.914
>> Time: 4.841
>> Time: 4.800
>>
>> tldr: all comparable, teo mainline slightly worse
>>
>> 4. Geekbench 5 (multi-core) on Pixel6
>> (Device is cooled for each iteration separately)
>> teo mainline:
>> 3113, 3068, 3079
>> mean 3086.66
>>
>> teo revert util-awareness:
>> 2947, 2941, 2952
>> mean 2946.66 (-4.54%)
>>
>> teo fixed v2:
>> 3032, 3066, 3019
>> mean 3039 (-1.54%)
>>
>>
>> Changes since v2:
>> - Reworded commits according to Dietmar's comments
>> - Dropped the KTIME_MAX as hit part from 3/3 according to Dietmar's
>> remark.
>>
>> Changes since v1:
>> - Removed all non-fixes.
>> - Do a full revert of Util-awareness instead of increasing thresholds.
>> - Address Dietmar's comments.
>> https://lore.kernel.org/linux-kernel/20240606090050.327614-2-christian.loehle@arm.com/T/
>>
>> Kind Regards,
>> Christian
>>
>> Christian Loehle (3):
>>   Revert: "cpuidle: teo: Introduce util-awareness"
>>   cpuidle: teo: Remove recent intercepts metric
>>   cpuidle: teo: Don't count non-existent intercepts
>>
>>  drivers/cpuidle/governors/teo.c | 189 +++++---------------------------
>>  1 file changed, 27 insertions(+), 162 deletions(-)
>>
>> --
> 
> Patches [1-2/3] have been applied as 6.11 material.
> 
> Patch [3/3] looks like it may be improved slightly, see my reply to that patch.
> 
> Thanks!

Hi Rafael,
are you fine with this being backported to stable?
@stable
4b20b07ce72f cpuidle: teo: Don't count non-existent intercepts
449914398083 cpuidle: teo: Remove recent intercepts metric
0a2998fa48f0 Revert: "cpuidle: teo: Introduce util-awareness"
apply as-is to
linux-6.10.y
linux-6.6.y
for linux-6.1.y only 449914398083 ("cpuidle: teo: Remove recent intercepts metric")
is relevant, I'll reply with a backport.


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ