RE: [PATCH] mce: fix missing stack-dumping in mce_panic()

From: Luck, Tony
Date: Sun Jan 08 2023 - 23:45:36 EST


>> Maybe then I can look at some way to suppress the oops dump for the "useless"
>> cases? Or perhaps at least add an extra message that stack dumps from arbitrary
>> machine checks are rarely useful.
>
> IMHO the user will always expect a stack dump when system panic. I'm afraid that if these
> "useless" cases are eliminated, it will cause user confused and trying to find what eats
> the stack dump?

What about the "extra message" option? Always print the full oops. but for ones from
mce_panic() arrange somehow to add an extra bit of helpful text. Something like this:

1) For poison consumption in kernel code:

"The reason for this panic is consumption of poisoned data by the kernel"

2) For the general machine check case:

"The reason for this panic is a hardware detected error reported as a machine check.
Note that the stack and register dumps are likely useless to diagnose the root cause
of this panic."

-Tony