Re: [PATCH 4/5] KVM: x86: Wrong register masking in 64-bit mode

[Nadav Amit]

On 5/7/14, 6:52 PM, Paolo Bonzini wrote:
> Il 07/05/2014 14:32, Nadav Amit ha scritto:
> > 32-bit operations are zero extended in 64-bit mode. Currently, the
> > code does
> > not handle them correctly and keeps the high bits. In 16-bit mode, the
> > high
> > 32-bits are kept intact.
> >
> > In addition, although it is not well-documented, when address override
> > prefix
> > is used with REP-string instruction, RCX high half is zeroed even if
> > ECX was
> > zero on the first iteration (as if an assignment was performed to ECX).
>
> Is this true even for REPZ and ZF=0 or REPNZ and ZF=1?
>
> Paolo
The REPZ and REPNZ condition is checked on the end of an iteration (see 
the REP instruction description on the SDM), so it does not matter. So 
even REPZ/RENZ would zero RCX high half.

This "feature" is not well-documented, but can be observed. Here is a 
small code you can try.

---
#include <stdio.h>
unsigned long src, dst;
int main()
{
	unsigned long long rsi, rdi, rcx;
	rcx = 0xffffffff00000000ull;
	rsi = (unsigned long long)&src | 0xffffffff00000000ull;
	rdi = (unsigned long long)&dst | 0xffffffff00000000ull;
	printf("before: rsi %llx rdi %llx rcx %llx\n", rsi, rdi ,rcx);
	asm volatile (  ".byte 0x67\n\t"
			"repne cmpsd\n\t"
			: "+S" (rsi), "+D" (rdi), "+c" (rcx)
			: : "memory", "cc" );
	printf("after: rsi %llx rdi %llx rcx %llx\n", rsi, rdi ,rcx);
	return 0;
}
---

Nadav

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/