Re: [PATCH] KVM: nVMX: Fix IRQs inject to L2 which belong to L1 since race

[Jan Kiszka <jan.kiszka@siemens.com>]

On 2014-07-04 12:52, Jan Kiszka wrote:
> On 2014-07-04 11:38, Paolo Bonzini wrote:
>> Il 04/07/2014 11:33, Jan Kiszka ha scritto:
>>>
>>> The compiler is not aware of the fact that push/pop exists in this
>>> function and, thus, places the vmcs parameter on the stack without
>>> reserving the space. So the pushfq will overwrite the vmcs pointer and
>>> let the function fail.
>>
>> Is that just a missing "memory" clobber?  push/pop clobbers memory.
> 
> Nope, we would needs some clobber like "stack". I wonder what is
> required to use push in inline assembly safely?

My colleague just found the answer: -mno-red-zone is required for 64-bit
in order to play freely with the stack (or you need to stay off that
zone, apparently some 128 bytes below the stack pointer). The kernel
sets that switch, our unit tests do not.

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SES-DE
Corporate Competence Center Embedded Linux