[PATCH 3/3] arm64: kprobes: Return DBG_HOOK_ERROR if kprobes can not handle a BRK

[Masami Hiramatsu (Google)]

On Thu, 1 Dec 2022 15:08:52 +0000
Mark Rutland <mark.rutland at arm.com> wrote:

> On Thu, Dec 01, 2022 at 11:39:21PM +0900, Masami Hiramatsu (Google) wrote:
> > From: Masami Hiramatsu (Google) <mhiramat at kernel.org>
> > 
> > Return DBG_HOOK_ERROR if kprobes can not handle a BRK because it
> > fails to find a kprobe corresponding to the address.
> > 
> > Since arm64 kprobes uses stop_machine based text patching for removing
> > BRK, it ensures all running kprobe_break_handler() is done at that point.
> > And after removing the BRK, it removes the kprobe from its hash list.
> > Thus, if the kprobe_break_handler() fails to find kprobe from hash list,
> > there is a bug.
> 
> IIUC this relies on BRK handling not being preemptible, which is something
> we've repeatedly considered changing along with a bunch of other debug
> exception handling.

Interesting idea... and it also need many changes in kprobe itself.

> 
> In case we do try to change that in future, it would be good to have a comment
> somewhere to that effect.

Hmm, it would fundamentally change the assumptions that kprobes relies on,
and would require a lot of thought again. (e.g. current running kprobe is
stored in per-cpu variable, it should be per-task. etc.)

> 
> I think there are other ways we could synchronise against that (e.g. using RCU
> tasks rude) if we ever do that, and this patch looks good to me.
> 
> > 
> > Signed-off-by: Masami Hiramatsu (Google) <mhiramat at kernel.org>
> > ---
> >  arch/arm64/kernel/probes/kprobes.c |   79 +++++++++++++++++-------------------
> >  1 file changed, 37 insertions(+), 42 deletions(-)
> > 
> > diff --git a/arch/arm64/kernel/probes/kprobes.c b/arch/arm64/kernel/probes/kprobes.c
> > index d2ae37f89774..ea56b22d4da8 100644
> > --- a/arch/arm64/kernel/probes/kprobes.c
> > +++ b/arch/arm64/kernel/probes/kprobes.c
> > @@ -298,7 +298,8 @@ int __kprobes kprobe_fault_handler(struct pt_regs *regs, unsigned int fsr)
> >  	return 0;
> >  }
> >  
> > -static void __kprobes kprobe_handler(struct pt_regs *regs)
> > +static int __kprobes
> > +kprobe_breakpoint_handler(struct pt_regs *regs, unsigned long esr)
> >  {
> >  	struct kprobe *p, *cur_kprobe;
> >  	struct kprobe_ctlblk *kcb;
> > @@ -308,39 +309,45 @@ static void __kprobes kprobe_handler(struct pt_regs *regs)
> >  	cur_kprobe = kprobe_running();
> >  
> >  	p = get_kprobe((kprobe_opcode_t *) addr);
> > +	if (WARN_ON_ONCE(!p)) {
> > +		/*
> > +		 * Something went wrong. This must be put by kprobe, but we
> > +		 * could not find corresponding kprobes. Let the kernel handle
> > +		 * this error case.
> > +		 */
> 
> Could we make this:
> 
> 		/*
> 		 * Something went wrong. This BRK used an immediate reserved
> 		 * for kprobes, but we couldn't find any corresponding probe.
> 		 */

OK.

> 
> > +		return DBG_HOOK_ERROR;
> > +	}
> >  
> > -	if (p) {
> > -		if (cur_kprobe) {
> > -			if (reenter_kprobe(p, regs, kcb))
> > -				return;
> > -		} else {
> > -			/* Probe hit */
> > -			set_current_kprobe(p);
> > -			kcb->kprobe_status = KPROBE_HIT_ACTIVE;
> > -
> > -			/*
> > -			 * If we have no pre-handler or it returned 0, we
> > -			 * continue with normal processing.  If we have a
> > -			 * pre-handler and it returned non-zero, it will
> > -			 * modify the execution path and no need to single
> > -			 * stepping. Let's just reset current kprobe and exit.
> > -			 */
> > -			if (!p->pre_handler || !p->pre_handler(p, regs)) {
> > -				setup_singlestep(p, regs, kcb, 0);
> > -			} else
> > -				reset_current_kprobe();
> > -		}
> > +	if (cur_kprobe) {
> > +		/* Hit a kprobe inside another kprobe */
> > +		if (!reenter_kprobe(p, regs, kcb))
> > +			return DBG_HOOK_ERROR;
> > +	} else {
> > +		/* Probe hit */
> > +		set_current_kprobe(p);
> > +		kcb->kprobe_status = KPROBE_HIT_ACTIVE;
> > +
> > +		/*
> > +		 * If we have no pre-handler or it returned 0, we
> > +		 * continue with normal processing.  If we have a
> > +		 * pre-handler and it returned non-zero, it will
> > +		 * modify the execution path and no need to single
> > +		 * stepping. Let's just reset current kprobe and exit.
> > +		 */
> 
> Minor wording nit: could we replace:
> 
> 	no need to single stepping.
> 
> With:
> 	
> 	not need to single-step.

OK, I'll update both in v2.

Thank you!

> 
> Thanks,
> Mark.
> 
> > +		if (!p->pre_handler || !p->pre_handler(p, regs))
> > +			setup_singlestep(p, regs, kcb, 0);
> > +		else
> > +			reset_current_kprobe();
> >  	}
> > -	/*
> > -	 * The breakpoint instruction was removed right
> > -	 * after we hit it.  Another cpu has removed
> > -	 * either a probepoint or a debugger breakpoint
> > -	 * at this address.  In either case, no further
> > -	 * handling of this interrupt is appropriate.
> > -	 * Return back to original instruction, and continue.
> > -	 */
> > +
> > +	return DBG_HOOK_HANDLED;
> >  }
> >  
> > +static struct break_hook kprobes_break_hook = {
> > +	.imm = KPROBES_BRK_IMM,
> > +	.fn = kprobe_breakpoint_handler,
> > +};
> > +
> >  static int __kprobes
> >  kprobe_breakpoint_ss_handler(struct pt_regs *regs, unsigned long esr)
> >  {
> > @@ -365,18 +372,6 @@ static struct break_hook kprobes_break_ss_hook = {
> >  	.fn = kprobe_breakpoint_ss_handler,
> >  };
> >  
> > -static int __kprobes
> > -kprobe_breakpoint_handler(struct pt_regs *regs, unsigned long esr)
> > -{
> > -	kprobe_handler(regs);
> > -	return DBG_HOOK_HANDLED;
> > -}
> > -
> > -static struct break_hook kprobes_break_hook = {
> > -	.imm = KPROBES_BRK_IMM,
> > -	.fn = kprobe_breakpoint_handler,
> > -};
> > -
> >  /*
> >   * Provide a blacklist of symbols identifying ranges which cannot be kprobed.
> >   * This blacklist is exposed to userspace via debugfs (kprobes/blacklist).
> > 


-- 
Masami Hiramatsu (Google) <mhiramat at kernel.org>