Re: [PATCH v2] x86: disable non-instrumented version of copy_mc when KMSAN is enabled

[Tetsuo Handa]

On 2024/03/06 18:16, Ingo Molnar wrote:
> 
> * Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx> wrote:
> 
>> Ping?
>>
>> This is current top crasher.
>> I hope this patch is applied before the merge window opens.

I posted this patch on Sun, 25 Feb 2024 13:40:59 +0900 but did not get response.
Therefore, I reposted on Sat, 2 Mar 2024 07:52:23 +0900, and Linus responded
that this patch is wrong.

> 
> 1) A false positive is not a 'crasher', it's a bug in instrumentation.
> 
> 2) A false positive in intrusive instrumentation that top distributions do 
>    not enable in their kernels has no immediate relevance to the timing of 
>    the merge window.

Not fixing a bug prevents us from finding and fixing other bugs. A refcount bug at
unknown location is preventing linux-next.git from finding other bugs for 20 days
( https://syzkaller.appspot.com/bug?id=8e4e66dfe299a2a00204ad220c641daaf1486a00 ).
Failure to fix bugs as many as possible in linux-next results in more bug reports
when failed-to-find-in-linux-next.git-bugs arrive at linux.git. I want to make it
possible to bisect linux.git cleanly before that refcount bug arrives at linux.git
by fixing bugs which we can fix now.