On 08/04/2010 10:27 AM, Wei Yongjun wrote:
> Group 8 instruction, BT[S|R|C] should be mask as BitOp.
>
> Signed-off-by: Wei Yongjun<[email protected]>
> ---
> arch/x86/kvm/emulate.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
> index d197b46..eba5a67 100644
> --- a/arch/x86/kvm/emulate.c
> +++ b/arch/x86/kvm/emulate.c
> @@ -2261,7 +2261,7 @@ static struct opcode twobyte_table[256] = {
> D(DstReg | SrcMem16 | ModRM | Mov),
> /* 0xB8 - 0xBF */
> N, N,
> - G(0, group8), D(DstMem | SrcReg | ModRM | BitOp | Lock),
> + G(BitOp, group8), D(DstMem | SrcReg | ModRM | BitOp | Lock),
> N, N, D(ByteOp | DstReg | SrcMem | ModRM | Mov),
> D(DstReg | SrcMem16 | ModRM | Mov),
> /* 0xC0 - 0xCF */
This is correct with your 4/4 patch, but incorrect before it. It will
incorrectly cause the emulator to adjust the address of the source
operand. This is documented to happen with a register source but not
with an immediate source. You can test it with this:
#include <stdio.h>
int main()
{
int a[3] = {0, 0, 0};
asm ("btcl $32, %0" :: "m" (a[0]) : "memory");
asm ("btcl $1, %0" :: "m" (a[1]) : "memory");
asm ("btcl %1, %0" :: "m" (a[0]), "r" (66) : "memory");
printf ("%x %x %x\n", a[0], a[1], a[2]);
}
It prints "1 2 4". I'm quite confident that instead it would print "0 3
4" with this patch, and "5 2 0" with this patch + 4/4.
Paolo
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
