Re: [PATCH kvm-unit-tests] x86: always inline functions called after set_exception_return

[David Matlack]

On Wed, Dec 9, 2015 at 7:02 AM, Paolo Bonzini  wrote:
> On 07/12/2015 21:36, David Matlack wrote:
>> set_exception_return forces exceptions handlers to return to a specific
>> address instead of returning to the instruction address pushed by the
>> CPU at the time of the exception. The unit tests apic.c and vmx.c use
>> this functionality to recover from expected exceptions.
>>
>> When using set_exception_return we have to be careful not to modify the
>> stack (such as by doing a function call) as triggering the exception will
>> likely jump us past the instructions which undo the stack manipulation
>> (such as a ret). To accomplish this, declare all functions called after
>> set_exception_return as __always_inline, so that the compiler always
>> inlines them.
>
> set_exception_return is generally not a great idea IMHO---thanks for
> looking at it.

Yup. This is a band-aid just to fix the current implementation.

>
> A couple years ago we discussed adding setjmp/longjmp to libcflat
> (http://www.spinics.net/lists/kvm/msg94159.html which is however missing
> a 32-bit version).  Making the exceptions do a longjmp would be a much
> safer option.

Good idea! I might give this a try, but don't hold your breath :)

>
> Paolo
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH kvm-unit-tests] x86: always inline functions called after set_exception_return

[Paolo Bonzini]

On 07/12/2015 21:36, David Matlack wrote:
> set_exception_return forces exceptions handlers to return to a specific
> address instead of returning to the instruction address pushed by the
> CPU at the time of the exception. The unit tests apic.c and vmx.c use
> this functionality to recover from expected exceptions.
> 
> When using set_exception_return we have to be careful not to modify the
> stack (such as by doing a function call) as triggering the exception will
> likely jump us past the instructions which undo the stack manipulation
> (such as a ret). To accomplish this, declare all functions called after
> set_exception_return as __always_inline, so that the compiler always
> inlines them.

set_exception_return is generally not a great idea IMHO---thanks for
looking at it.

A couple years ago we discussed adding setjmp/longjmp to libcflat
(http://www.spinics.net/lists/kvm/msg94159.html which is however missing
a 32-bit version).  Making the exceptions do a longjmp would be a much
safer option.

Paolo
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html