Re: [RFC PATCH] kprobes: arm: enable OPTPROBES for arm 32

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

(2014/08/08 10:25), Wang Nan wrote:
> On 2014/8/7 14:59, Masami Hiramatsu wrote:
>> (2014/08/06 15:24), Wang Nan wrote:
>>>>> +
>>>>> +static void
>>>>> +optimized_callback(struct optimized_kprobe *op, struct pt_regs *regs)
>>>>> +{
>>>>> +	unsigned long flags;
>>>>> +
>>>>> +	regs->ARM_pc = (unsigned long)op->kp.addr;
>>>>> +	regs->ARM_ORIG_r0 = ~0UL;
>>>>> +
>>>>> +
>>>>> +	local_irq_save(flags);
>>>>> +	/* 
>>>>> +	 * This is possible if op is under delayed unoptimizing.
>>>>> +	 * We need simulate the replaced instruction.
>>>>> +	 */
>>>>> +	if (kprobe_disabled(&op->kp)) {
>>>>> +		struct kprobe *p = &op->kp;
>>>>> +		op->kp.ainsn.insn_singlestep(p->opcode, &p->ainsn, regs);
>>>>> +	} else {
>>>>> +		kprobe_handler(regs);
>>>>> +	}
>>>>
>>>> You don't need brace "{}" for one statement.
>>>> By the way, why don't you call opt_pre_handler()?
>>>>
>>>
>>> I use kprobe_handler because it handles instruction emulation.
>>>
>>> In addition, I'm not very sure whether skipping the complex checks
>>> in kprobe_handler() is safe or not.
>>
>> That seems to do same thing on x86. Then you should do something like
>> the optimized_callback() on x86 as below.
>>
>> static void
>> optimized_callback(struct optimized_kprobe *op, struct pt_regs *regs)
>> {
>>         struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
>>         unsigned long flags;
>>
>>         local_irq_save(flags);
>>         if (kprobe_running()) {
>>                 kprobes_inc_nmissed_count(&op->kp);
> 
> In this case we still need a singlestep, right?

Ah, right! and if the singlestep requires setting up the regs->ARM_pc,
we also do that before this check. So the right code will be;

static void
optimized_callback(struct optimized_kprobe *op, struct pt_regs *regs)
{
	struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
	unsigned long flags;

	local_irq_save(flags);
	/* Save skipped registers */
	regs->ARM_pc = (unsigned long)op->kp.addr;
	regs->ARM_ORIG_r0 = ~0UL;

        if (kprobe_running())
		kprobes_inc_nmissed_count(&op->kp);
        else {
		__this_cpu_write(current_kprobe, &op->kp);
 		kcb->kprobe_status = KPROBE_HIT_ACTIVE;
		opt_pre_handler(&op->kp, regs);
		__this_cpu_write(current_kprobe, NULL);
	}
	op->kp.ainsn.insn_singlestep(op->kp.opcode, &op->kp.ainsn, regs);
	local_irq_restore(flags);
}

Thank you,

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