Re: [PATCH v9 2/3] x86/mce: Add per-bank CMCI storm mitigation

[Philip Li]

On Wed, Oct 11, 2023 at 11:42:26PM +0800, Feng Tang wrote:
> On Wed, Oct 11, 2023 at 11:16:46PM +0800, Luck, Tony wrote:
> > > kernel test robot noticed a -8.8% regression of stress-ng.clock.ops_per_sec on:
> > >
> > >
> > > commit: 26bff7b04b829cccc6a97726d6398391a62e34ef ("[PATCH v9 2/3] x86/mce: Add per-bank CMCI storm mitigation")
> > > url: https://github.com/intel-lab-lkp/linux/commits/Tony-Luck/x86-mce-Remove-old-CMCI-storm-mitigation-code/20231005-024047
> > > patch link: https://lore.kernel.org/all/20231004183623.17067-3-tony.luck@xxxxxxxxx/
> > > patch subject: [PATCH v9 2/3] x86/mce: Add per-bank CMCI storm mitigation
> > >
> > > testcase: stress-ng
> > > test machine: 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory
> > > parameters:
> > >
> > >       nr_threads: 10%
> > >       disk: 1HDD
> > >       testtime: 60s
> > >       fs: ext4
> > >       class: os
> > >       test: clock
> > >       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 <oliver.sang@xxxxxxxxx>
> > > | Closes: https://lore.kernel.org/oe-lkp/202310111637.dee70328-oliver.sang@xxxxxxxxx
> > 
> > Is the test injecting massive numbers of corrected memory errors? The code in this patch
> > is only executed when handling CMCI interrupts, or polling machine check banks (at most
> > once per second).
> > 
> > I'm guessing this report is just because alignment of some hot path code changed.
> 
> IIRC, CONFIG_DEBUG_FORCE_FUNCTION_ALIGN_64B was enabled in 0Day's
> kernel config for quite a while, to force each funtion's start
> address aligned on 64 bytes. Don't know if this has been changed
> recently.

yes, the kernel is always built with -falign-functions=64 to test
performance now.

> 
> Also I noticed the patch introduce a new per-cpu variable 'storm_desc",
> if the function address is 64B aligned, then per-cpu data alignment
> may be related.
> 
> Thanks,
> Feng
> 
> > 
> > -Tony
>