[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <E4ABEE53CC34664FA3F0BD8AEAF50A1941474EDB@szxeml511-mbs.china.huawei.com>
Date: Mon, 21 Oct 2013 06:15:22 +0000
From: Liuyongan <liuyongan@...wei.com>
To: Masami Hiramatsu <masami.hiramatsu.pt@...achi.com>
CC: "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
Qianhuibin <qianhuibin@...wei.com>
Subject: RE: kprobe pre_handler change return IP
> -----Original Message-----
> From: Masami Hiramatsu [mailto:masami.hiramatsu.pt@...achi.com]
> Sent: Sunday, October 20, 2013 12:00 AM
> To: Liuyongan
> Cc: linux-kernel@...r.kernel.org; Qianhuibin
> Subject: Re: kprobe pre_handler change return IP
>
> (2013/10/17 21:57), Liuyongan wrote:
> > I use kprobe to probe a function suppose:
> > int is_winter(int num) { ... }
> > int replace_is_winter(int num) { ...}
> > I want to replace is_winter() with replace_is_winter(), so when we call
> is_winter, replace_is_winter will be called.
> >
> > so:
> > int my_pre_handler(struct kprobe *p, struct pt_regs *regs)
> > {
> > regs->ip = (unsigned long)&replace_is_winter;
> > return 1;
> > }
> >
> > and echo 0 > /proc/sys/debug/kprobes-optimization so that jump instruction
> will not be used.
> >
> > I got a exception in fault_handler, and trap number is 14.
> >
> > fault_handler: p->addr = 0xffffffffa08e201a, ip = ffffffff8021c59d, trap
> #14n
> >
> > Anyone here can help me ?
>
> If you want to replace something with kprobes, the pre_handler must clean
> current_kprobe up.
> Actually the same thing has been done in setup_detour_execution(). So, what
> you need to do is
Great! It works.
As my_pre_handler() in another modules, I should export current_kprobe using
EXPORT_PER_CPU_SYMBOL(current_kprobe);
in arch/x86/kernel/kprobes.c right after current_kprobe's definition.
>
> > int my_pre_handler(struct kprobe *p, struct pt_regs *regs)
> > {
> > regs->ip = (unsigned long)&replace_is_winter;
>
> reset_current_kprobe();
> preempt_enable_no_resched();
>
> > return 1;
> > }
>
> Happy hacking! ;)
>
>
> Thank you,
>
> --
> Masami HIRAMATSU
> IT Management Research Dept. Linux Technology Center
> Hitachi, Ltd., Yokohama Research Laboratory
> E-mail: masami.hiramatsu.pt@...achi.com
>
Powered by blists - more mailing lists