Re: [syzbot] [kernel?] upstream test error: KMSAN: uninit-value in irqentry_exit_to_kernel_mode_preempt

[Peter Zijlstra]

On Wed, May 13, 2026 at 02:36:10AM +0200, Thomas Gleixner wrote:

> But now four years down the road that obviously does not solve the
> problem because nothing happened and we have to deal with existing
> compilers and the only thing we can do until compiler people get their
> act together is work around this insanity in kernel code again.

> Thoughts?

KMSAN is debug cruft. I don't see why we cannot demand they fix it and
make KMSAN depend whatever clang will have the thing fixed. There is no
reason we need to support existing compilers what so ever for debug
code.