Re: [PATCH 2/5] x86/uprobes: implement x86 specific arch_uprobe_*_step

[Oleg Nesterov <oleg@redhat.com>]

On 08/08, Sebastian Andrzej Siewior wrote:
>
> On 08/08/2012 02:57 PM, Oleg Nesterov wrote:
>>> +static int insn_changes_flags(struct arch_uprobe *auprobe)
>>> +{
>>> +	/* popf reads flags from stack */
>>> +	if (auprobe->insn[0] == 0x9d)
>>> +		return 1;
>>
>> Ah, somehow I didn't think about this before.
>>
>> ->insn[0] doesn't look right, we should skip the prefixes.
>
> Why? I tried 'lock popf' and I got invalid instruction. The same for
> 'rep popf'.

	int main(void)
	{
		asm volatile ("pushf; rep; popf");

		return 0;
	}

objdump:

	00000000040047c <main>:
	  40047c:       55                      push   %rbp
	  40047d:       48 89 e5                mov    %rsp,%rbp
	  400480:       9c                      pushfq
	  400481:       f3 9d                   repz popfq
	  400483:       b8 00 00 00 00          mov    $0x0,%eax
	  400488:       c9                      leaveq
	  400489:       c3                      retq



OK, probably nobody should do this (although the kernel should not
assume this imho), but

	asm volatile ("pushfw; popfw");

doesn't look bad and the code is

	000000000040047c <main>:
	  40047c:       55                      push   %rbp
	  40047d:       48 89 e5                mov    %rsp,%rbp
	  400480:       66 9c                   pushfw
	  400482:       66 9d                   popfw
	  400484:       b8 00 00 00 00          mov    $0x0,%eax
	  400489:       c9                      leaveq
	  40048a:       c3                      retq



And in any case it would be better to re-use auprobe->fixups.

Oleg.