Re: [PATCH v3 2/5] x86,fs/resctrl: Remove unnecessary non-linear MBA configuration check
From: Chen Yu
Date: Tue Nov 17 2026 - 00:06:45 EST
On Mon, Nov 16, 2026 at 08:20:06PM -0800, Reinette Chatre wrote:
> Hi,
>
> On 11/16/26 7:36 PM, kernel test robot wrote:
> >
> > Hello,
> >
> > kernel test robot noticed a 4.4% regression of stress-ng.sockfd.ops_per_sec on:
> >
> > commit: 12a0f554b3d62eb1270d6d60e9c6ab2c07dbbe35 ("[PATCH v3 2/5] x86,fs/resctrl: Remove unnecessary non-linear MBA configuration check")
> > url: https://github.com/intel-lab-lkp/linux/commits/Reinette-Chatre/x86-fs-resctrl-Remove-unused-non-linear-bandwidth-delay-mapping/20261104-161329
> > patch link: https://lore.kernel.org/all/7e1aeee81a38a33ca889db68e4c645f935e4e6cd.1793834224.git.reinette.chatre@xxxxxxxxx/
> > patch subject: [PATCH v3 2/5] x86,fs/resctrl: Remove unnecessary non-linear MBA configuration check
> >
> > testcase: stress-ng
> > config: x86_64-rhel-9.4
> > compiler: gcc-14
> > test machine: 224 threads 2 sockets Intel(R) Xeon(R) Platinum 8480CTDX (Sapphire Rapids) with 256G memory
> > parameters:
> >
> > nr_threads: 100%
> > testtime: 60s
> > test: sockfd
> > cpufreq_governor: performance
> >
> >
> > If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
> > | Reported-by: kernel test robot <yi1.lai@xxxxxxxxx>
> > | Closes: https://lore.kernel.org/oe-lkp/202611171056.912725cf-lkp@xxxxxxxxx
> >
> >
> > Details are as below:
> > -------------------------------------------------------------------------------------------------->
> >
> >
> > The kernel config and materials to reproduce are available at:
> > https://download.01.org/0day-ci/archive/20261117/202611171056.912725cf-lkp@xxxxxxxxx
> >
> > =========================================================================================
> > compiler/cpufreq_governor/kconfig/nr_threads/rootfs/tbox_group/test/testcase/testtime:
> > gcc-14/performance/x86_64-rhel-9.4/100%/debian-13-x86_64-20250902.cgz/lkp-spr-r02/sockfd/stress-ng/60s
> >
> > commit:
> > 0d4f0ec1db ("x86,fs/resctrl: Remove unused non-linear bandwidth delay mapping")
> > 12a0f554b3 ("[PATCH v3 2/5] x86,fs/resctrl: Remove unnecessary non-linear MBA configuration check")
> >
> > 0d4f0ec1dbca25be 12a0f554b3d62eb1270d6d60e9c
> > ---------------- ---------------------------
> > %stddev %change %stddev
> > \ | \
> > 56484401 -4.4% 54013430 stress-ng.sockfd.ops
> > 940555 -4.4% 899375 stress-ng.sockfd.ops_per_sec
> > 3901263 -1.5% 3841842 stress-ng.time.involuntary_context_switches
> > 2386355 ± 6% -10.4% 2138169 ± 5% stress-ng.time.voluntary_context_switches
> >
> I am unable to see the connection between this test and the patch identified.
>
> The patch identified is a resctrl change and looking at the reproducer script
> resctrl was not mounted during the test. The kernel config does have
> CONFIG_X86_CPU_RESCTRL=y so resctrl will do its enumeration and then just do
> nothing until resctrl is mounted, which never happens.
>
> The patch does remove a struct member so this does change the resctrl memory
> layout. The structure changed, resctrl_membw, can be found in rdt_resources_all[]
> that is a writable global with static storage. I expect it to be located in
> .data. Could this possibly impact the test?
>
I agree. I think this is a data layout change, not a real regression in the code.
rdt_resources_all[] is in .data, and arch/x86 is linked before net/, So all the .data
variables after the shrink of struct resctrl_membw moves ahead, then variables in the
network component share a cache line with different neighbors. I don't think resctrl
should change anything for this, perf c2c on sockfd should show which hot global
variable is sharing a cache line, and that variable should be cache aligned.
thanks,
Chenyu