linux-kernel - Re: [PATCH] uprobes: fix scratch register selection for rip-relative fixups

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Fri, 02 May 2014 17:04:08 +0200
From:	Denys Vlasenko <dvlasenk@...hat.com>
To:	Jim Keniston <jkenisto@...ux.vnet.ibm.com>
CC:	linux-kernel@...r.kernel.org,
	Masami Hiramatsu <masami.hiramatsu.pt@...achi.com>,
	Srikar Dronamraju <srikar@...ux.vnet.ibm.com>,
	Ingo Molnar <mingo@...nel.org>, Oleg Nesterov <oleg@...hat.com>
Subject: Re: [PATCH] uprobes: fix scratch register selection for rip-relative
 fixups

On 05/02/2014 02:48 AM, Jim Keniston wrote:
> On Thu, 2014-05-01 at 19:09 +0200, Denys Vlasenko wrote:
>> +#define VEX2_VVVV(insn)		X86_VEX_V((insn)->vex_prefix.bytes[1])
>> +#define VEX3_VVVV(insn)		X86_VEX_V((insn)->vex_prefix.bytes[2])
> 
> I disclaim any knowledge about VEX* stuff.
...
> skipped this next part...
> 
>> +	/* Similar treatment for VEX3 prefix */
>> +	/* TODO: add XOP/EVEX treatment when insn decoder supports them */
>> +	if (insn->vex_prefix.nbytes == 3) {
>> +		/*
>> +		 * vex2:     c5    rvvvvLpp   (has no b bit)
>> +		 * vex3/xop: c4/8f rxbmmmmm wvvvvLpp
>> +		 * evex:     62    rxbR00mm.wvvvv1pp.zllBVaaa
>> +		 *   (evex will need setting of both b and x since
>> +		 *   in non-sib encoding evex.x is 4th bit of MODRM.rm)
>> +		 * Setting VEX3.b (setting because it has inverted meaning):
>> +		 */
>> +		cursor = auprobe->insn + insn_offset_vex_prefix(insn) + 1;
>> +		*cursor |= 0x20;
>>  	}
> 
> ... resumed reviewing here.

I do realize that most people reading this code won't be
aficionados of memorizing x86 insn encoding quirks.

To help them, I added the comment which spells out bit layout
of vex/evex/xop prefixes (see above). Does it help?

> VEX again.  Covering my ears and humming...
> 
>> +	if (insn->vex_prefix.nbytes == 2) {
>> +		reg2 = VEX2_VVVV(insn) ^ 0xf;
>> +	}
>> +	if (insn->vex_prefix.nbytes == 3) {
>> +		reg2 = VEX3_VVVV(insn) ^ 0xf;
>> +	}
>> +	/* TODO: add XOP, EXEV vvvv reading */
>> +

In fact there is room for improvement in this part:
it's better to ignore most-significant bit of this field
(32-bit mode strikes back).

>> +	reg2 = 3; /* BX */
>> +	if (can_use_regs & UPROBE_FIX_RIP_DI)
>> +		reg2 = 7;
>> +	if (can_use_regs & UPROBE_FIX_RIP_SI)
>> +		reg2 = 6;
> 
> It seems more natural to code this as:
> 	if (can_use_regs & UPROBE_FIX_RIP_SI)
> 		reg2 = 6;
> 	else if (can_use_regs & UPROBE_FIX_RIP_DI)
> 		reg2 = 7;
> 	else
> 		reg2 = 3; /* BX */
> ... which is pretty much how you do it in scratch_reg().

Okay.
I can also get rid of "can_use_regs" variable altogether then!
Sending new version in a jiffy.
-- 
vda


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