Re: [PATCH 6/6] x86/MPX: bndmk, bndldx, and bndstx don't allow RIP-relative addressing

["Jan Beulich" <JBeulich at suse dot com>]

>>> On 08.10.13 at 18:13, "H.J. Lu" <hjl.tools@gmail.com> wrote:
> On Tue, Oct 8, 2013 at 7:44 AM, Jan Beulich <JBeulich@suse.com> wrote:
>> gas/
>> 2013-10-08  Jan Beulich <jbeulich@suse.com>
>>
>>         * tc-i386.c (i386_index_check): Reject RIP-relative addressing for
>>         bndmk, bndldx, and bndstx. Warn about register scaling by other
>>         than 1 for bndldx and bndstx.
>>
>> --- 2013-10-07/gas/config/tc-i386.c
>> +++ 2013-10-07/gas/config/tc-i386.c
>> @@ -8360,6 +8375,25 @@ bad_address:
>>                            || i.index_reg->reg_num == RegEiz))
>>                       || !i.index_reg->reg_type.bitfield.baseindex)))
>>             goto bad_address;
>> +
>> +         /* bndmk, bndldx, and bndstx have special restrictions. */
>> +         if (current_templates->start->base_opcode == 0xf30f1b
>> +             || (current_templates->start->base_opcode & ~1) == 0x0f1a)
>> +           {
>> +             /* They cannot use RIP-relative addressing. */
>> +             if (i.base_reg
>> +                 && i.base_reg->reg_num == (addr_mode == CODE_64BIT ? RegRip
>> +                                                                    : RegEip))
>> +               {
>> +                 as_bad (_("`%s' cannot be used here"), operand_string);
>> +                 return 0;
>> +               }
> 
> RegEip should be disallowed much earlier since address size
> prefix doesn't work for MPX.

Only if that diagnostic gets converted back to an error, which I
don't agree to so far.

>> +             /* bndldx and bndstx ignore their scale factor. */
>> +             if (current_templates->start->base_opcode != 0xf30f1b
>> +                 && i.log2_scale_factor)
>> +               as_warn (_("register scaling is being ignored here"));
> 
> Scaling factor is still encoded.  I am not sure if it belongs to assembler.

That's why it's a warning - it's providing a hint to the programmer
that what (s)he wrote makes no sense, but is being accepted. I
wouldn't, however, mind hiding this one when quiet_warnings is set.

Jan