lkml.org 
[lkml]   [2022]   [Aug]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] x86/numa: Use cpumask_available instead of hardcoded NULL check
On Tue, 02 Aug 2022 16:37:44 +0530  Ingo Molnar <mingo@kernel.org> wrote:
> Your fix makes sense I suppose, but I'm wondering how testing didn't
> trigger this warning.
>
> Off-stack isn't a rare config option:
>
> kepler:~/tip> make allmodconfig
> #
> # No change to .config
> #
> kepler:~/tip> grep CPUMASK_OFFSTACK .config
> CONFIG_CPUMASK_OFFSTACK=y
> kepler:~/tip>
>
> What am I missing?

Maybe this triggers on certain config options set, or maybe due to new
gcc version? (I'm using gcc-12, I also likely saw while on gcc-11.)
It nevertheless is a helpful warning.

I just now tried `make defconfig` (default configuration based on
'x86_64_defconfig') and compiling with `make -j13 all`, and gcc doesn't
give any warning. (CONFIG_CPUMASK_OFFSTACK isn't even listed in the
.config file produced, grep fails.)

The config on which I can reproduce the warning can be found here:
https://gist.github.com/siddhpant/0197ea2b9873e8719d5d7ef991e2cd89
(It has 8969 lines, thus uploaded as a gist.)

This is a modification of a config found on syzkaller, which I was using
to compile and test some bug. I had noticed the gcc warning earlier while
on similar detours and usually ignored it, but now I finally took a look.

I tested compiling with it 5 times (`make clean` and `make -j13 all`),
and gcc gave the warning in all attempts. I also tried `make -j1 all`,
which also had gcc spitting out the warning, so it cannot be any race.

> > Fixes: c032ef60d1aa ("cpumask: convert node_to_cpumask_map[] to cpumask_var_t")
> > Fixes: de2d9445f162 ("x86: Unify node_to_cpumask_map handling between 32 and 64bit")
>
> These are ancient commits from 2009 & 2011.

Yes, that's where blaming the file leads me to.

Thanks,
Siddh

\
 
 \ /
  Last update: 2022-08-02 18:32    [W:0.107 / U:0.196 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site