linux-kernel - Re: [linus:master] [mm] c544a952ba: netperf.Throughput

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] [day] [month] [year] [list]

Message-ID: <20250709092046.10428-1-nikhil.dhama@amd.com>
Date: Wed, 9 Jul 2025 14:50:46 +0530
From: Nikhil Dhama <nikhil.dhama@....com>
To: <oliver.sang@...el.com>
CC: <akpm@...ux-foundation.org>, <bharata@....com>,
	<huang.ying.caritas@...il.com>, <linux-kernel@...r.kernel.org>,
	<linux-mm@...ck.org>, <mgorman@...hsingularity.net>, <nikhil.dhama@....com>,
	<oe-lkp@...ts.linux.dev>, <raghavendra.kt@....com>,
	<ying.huang@...ux.alibaba.com>
Subject: Re: [linus:master] [mm]  c544a952ba:  netperf.Throughput_Mbps 18.9%

Hi,

On 7/3/2025 11:19 AM, kernel test robot wrote:
>
> Hello,
>
>
> we reported
> "[linus:master] [mm]  c544a952ba:  vm-scalability.throughput 33.3% improvement"
> in
> https://lore.kernel.org/all/202506121415.eeceda1d-lkp@intel.com/
>
> now we captured a netperf regression on a Core machine.
>
> we don't have enough knowledge if this is a tradeoff, just make out this report
> FYI.
>
>
> kernel test robot noticed a 18.9% regression of netperf.Throughput_Mbps on:
>

[...]

> The kernel config and materials to reproduce are available at:
> https://download.01.org/0day-ci/archive/20250703/202507030638.d8cff1fb-lkp@intel.com

I tried reproducing this issue using the shared kernel config,
On a 192 cores AMD EPYC server with 768G memory, 
I am getting following score (normalised wrt v6.15 vanilla):

                      netperf UDP_STREAM
                      ------------------       
v6.15 vanilla                 	     100
v6.15 with (c544a952ba)          100.085



> commit: c544a952ba61b1a025455098033c17e0573ab085 ("mm: pcp: increase pcp->free_count threshold to trigger free_high")
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
>
> [still regression on linus/master      66701750d5565c574af42bef0b789ce0203e3071]
> [still regression on linux-next/master 3f804361f3b9af33e00b90ec9cb5afcc96831e60]
>
> testcase: netperf
> config: x86_64-rhel-9.4
> compiler: gcc-12
> test machine: 4 threads Intel(R) Core(TM) i5-6500 CPU @ 3.20GHz (Skylake) with 32G memory
> parameters:
>
>         ip: ipv4
>         runtime: 300s
>         nr_threads: 1
>         cluster: cs-localhost
>         test: UDP_STREAM
>         cpufreq_governor: performance

To match thread count and memory, On the same machine I disabled all cpus
keeping only 4 of them enabled, and limited memory to 32G.

with this configuration, I am getting the following score:

                      netperf UDP_STREAM
                      ------------------
v6.15 vanilla                        100
v6.15 with (c544a952ba)              154

I am not able to repoduce the issue, am I missing something which is required to reproduce?

Thanks,
Nikhil