Re: Re: [PATCH v10 3.3-rc2 1/9] uprobes: Install and remove breakpoints.

[Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>]

(2012/02/08 18:40), Denys Vlasenko wrote:
> On Tue, Feb 7, 2012 at 6:17 PM, Srikar Dronamraju
> <srikar@linux.vnet.ibm.com> wrote:
>> Changelog: (Since v10): Add code to clear REX.B prefix pointed out by Denys Vlasenko
>> and fix suggested by Masami Hiramatsu.
> ...
>> +       /*
>> +        * Point cursor at the modrm byte.  The next 4 bytes are the
>> +        * displacement.  Beyond the displacement, for some instructions,
>> +        * is the immediate operand.
>> +        */
>> +       cursor = uprobe->insn + insn_offset_modrm(insn);
>> +       insn_get_length(insn);
>> +       if (insn->rex_prefix.nbytes)
>> +               *cursor &= 0xfe;        /* Clearing REX.B bit */
> 
> It looks like cursor points to mod/reg/rm byte, not rex byte.
> Comment above says it too. You seem to be clearing a bit
> in a wrong byte. I think it should be

Right, but...

> 
>         /* Clear REX.b bit (extension of MODRM.rm field):
>          * we want to encode rax/rcx, not r8/r9.
>          */
>         if (insn->rex_prefix.nbytes)
>                 insn->rex_prefix.bytes[0] &= 0xfe;

No, that is a meaningless operation.
As I originally said,

> insn_get_length(insn);
> if (insn->rex_prefix.nbytes) {
> 	cursor = uprobe->insn + insn_offset_rex_prefix(insn);
> 	*cursor &= 0xfe;	/* Clearing REX.B bit */
> }

You have to move the cursor itself, since the .rex_prefix
is just a copied data...

Thanks,

-- 
Masami HIRAMATSU
Software Platform Research Dept. Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@hitachi.com