Messages in this thread | | | Date | Thu, 17 May 2012 19:37:03 -0700 | Subject | Re: [git pull] machine check recovery fix | From | Tony Luck <> |
| |
On Thu, May 17, 2012 at 5:25 PM, Linus Torvalds <torvalds@linux-foundation.org> wrote: > If the machine check happened in kernel space, we currently *are* > returning to the instruction that executed. With or without your > patch. That's my argument.
When we assign the severity for the error we check for kernel vs. user space. If we took the machine check in kernel space it will get assigned MCE_PANIC_SEVERITY severity ... and we won;t try to do any recovery. We only play the games with TIF_MCE_NOTIFY if we see severity MCE_AR_SEVERITY ... which we will only do if the machine check happened in user mode.
So current recovery code only tries to deal with the user case.
Machine checks in kernel space are a future project ... and I agree will be a monster pain because we'll have to figure out everything in machine check context.
-Tony
| |