Re: [PATCH] x86/mm: Don't try to change poison pages to uncacheable in a guest

["Luck, Tony" <tony.luck@intel.com>]

There is only one actual machine check. But the VMM simulates a second machine check to the guest when the guest tries to access the poisoned page.

The stack trace was from Jue. I didn’t try to check it. But it looked reasonable that Linux would flush the cache for a page that is transitioning from cacheable to uncacheable.

Sent from my iPhone

> On May 15, 2020, at 23:54, Borislav Petkov <bp@alien8.de> wrote:
> 
> On Tue, May 05, 2020 at 11:46:48AM -0700, Tony Luck wrote:
>> An interesting thing happened when a guest Linux instance took
>> a machine check. The VMM unmapped the bad page from guest physical
>> space and passed the machine check to the guest.
>> 
>> Linux took all the normal actions to offline the page from the process
>> that was using it. But then guest Linux crashed because it said there
>> was a second machine check inside the kernel with this stack trace:
>> 
>> do_memory_failure
>>    set_mce_nospec
>>         set_memory_uc
>>              _set_memory_uc
>>                   change_page_attr_set_clr
>>                        cpa_flush
>>                             clflush_cache_range_opt
> 
> Maybe I don't see it but how can clflush_cache_range_opt() call
> cpa_flush() ?
> 
>> This was odd, because a CLFLUSH instruction shouldn't raise a machine
>> check (it isn't consuming the data). Further investigation showed that
>> the VMM had passed in another machine check because is appeared that the
>> guest was accessing the bad page.
> 
> This is where you lost me - if the VMM unmaps the page during the first
> MCE, how can the guest even attempt to touch it and do this stack trace
> above?
> 
> /me is confused.
> 
> -- 
> Regards/Gruss,
>    Boris.
> 
> https://people.kernel.org/tglx/notes-about-netiquette