Re: [boot crash] Re: [tip:x86/mce3] x86, mce: use 64bit machine check code on 32bit

[Ingo Molnar <mingo@elte.hu>]

* Andi Kleen <ak@linux.intel.com> wrote:

> Ingo Molnar wrote:
>> * Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com> wrote:
>>
>>> One possibility is: if the BIOS doesn't clear status in banks, new  
>>> mce codes will try to log such junks. If the junk is totally junk  
>>> but can be decoded as a valid log with MISCV or ADDRV bit, and if  
>>> the cpu try to access register which is not implemented (e.g.  
>>> IA32_MCi_MISC/ADDR), then such access might cause a general  
>>> protection exception. (ref. ASDM 3A 15.3.2.3)
>>
>> btw., that reminds me: mce_rdmsrl() needs to be fixed to use  
>> rdmsrl_safe() and it should emit a WARN_ONCE() if it ever hits an  
>> error while trying to access registers.
>
> In general systems (like VMs) who don't have MCA MSRs don't 
> declare the capability bits (there are own capability bits for all 
> of this) and then the MSRs are never touched. So far I've not had 
> a single report of this going wrong.
>
> I suspect the problem on your system is something else too we just 
> need to debug properly.

That sidenote was unrelated to this bug.

We obviously dont #GP fault in the MSR access, i'd see that in the 
crash.

We want to use rdmsr_safe()/wrmsr_safe() in general as a defensive 
measure, for all code that does not actually add functionality but 
is diagnostic (like MCE).

	Ingo