Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1575795 > unrolled thread

[BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible deadlock case in kretprobe

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible deadlock case in kretprobe Masami Hiramatsu <mhiramat@kernel.org> - 2017-02-07 16:20 +0100
    Re: [BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible  deadlock case in kretprobe Will Deacon <will.deacon@arm.com> - 2017-02-08 16:30 +0100
      Re: [BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible  deadlock case in kretprobe Masami Hiramatsu <mhiramat@kernel.org> - 2017-02-08 23:00 +0100

#1575795 — [BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible deadlock case in kretprobe

Similar to x86 kretprobe deadlock issue, arm64 also implements
kretprobe-booster (trampoline code directly call handler.)
So it has same deadlock issue if there are 2 kretprobes on
normal function and the function called from FIQ (or anywhere
which can be invoked when local_irq_disabled).

This fixes the issue as same as we did on x86.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
---
 arch/arm64/kernel/probes/kprobes.c |   12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/kernel/probes/kprobes.c b/arch/arm64/kernel/probes/kprobes.c
index 2a07aae..401f7c9 100644
--- a/arch/arm64/kernel/probes/kprobes.c
+++ b/arch/arm64/kernel/probes/kprobes.c
@@ -561,6 +561,8 @@ bool arch_within_kprobe_blacklist(unsigned long addr)
 	return false;
 }
 
+static struct kprobe dummy_retprobe = {.addr = (void *)&kretprobe_trampoline};
+
 void __kprobes __used *trampoline_probe_handler(struct pt_regs *regs)
 {
 	struct kretprobe_instance *ri = NULL;
@@ -572,6 +574,11 @@ void __kprobes __used *trampoline_probe_handler(struct pt_regs *regs)
 	kprobe_opcode_t *correct_ret_addr = NULL;
 
 	INIT_HLIST_HEAD(&empty_rp);
+
+	/* This prevents kernel to change running cpu while processing */
+	preempt_disable();
+	get_kprobe_ctlblk()->kprobe_status = KPROBE_HIT_ACTIVE;
+	__this_cpu_write(current_kprobe, &dummy_retprobe);
 	kretprobe_hash_lock(current, &head, &flags);
 
 	/*
@@ -614,10 +621,9 @@ void __kprobes __used *trampoline_probe_handler(struct pt_regs *regs)
 		orig_ret_address = (unsigned long)ri->ret_addr;
 		if (ri->rp && ri->rp->handler) {
 			__this_cpu_write(current_kprobe, &ri->rp->kp);
-			get_kprobe_ctlblk()->kprobe_status = KPROBE_HIT_ACTIVE;
 			ri->ret_addr = correct_ret_addr;
 			ri->rp->handler(ri, regs);
-			__this_cpu_write(current_kprobe, NULL);
+			__this_cpu_write(current_kprobe, &dummy_retprobe);
 		}
 
 		recycle_rp_inst(ri, &empty_rp);
@@ -632,6 +638,8 @@ void __kprobes __used *trampoline_probe_handler(struct pt_regs *regs)
 	}
 
 	kretprobe_hash_unlock(current, &flags);
+	__this_cpu_write(current_kprobe, NULL);
+	preempt_enable_no_resched();
 
 	hlist_for_each_entry_safe(ri, tmp, &empty_rp, hlist) {
 		hlist_del(&ri->hlist);

[toc] | [next] | [standalone]

#1576650 — Re: [BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible deadlock case in kretprobe

[adding linux-arm-kernel]

On Wed, Feb 08, 2017 at 12:13:14AM +0900, Masami Hiramatsu wrote:
> Similar to x86 kretprobe deadlock issue, arm64 also implements
> kretprobe-booster (trampoline code directly call handler.)
> So it has same deadlock issue if there are 2 kretprobes on
> normal function and the function called from FIQ (or anywhere
> which can be invoked when local_irq_disabled).

We don't support FIQ on arm64, so I'm not worried about that particular
case. What are the other cases? I can think of debug exceptions, but those
shouldn't be generally kprobe-able, and taking data aborts in things like
get_user/put_user. Are those affected by this bug?

Either way, could you please expand the commit message like you have
for x86? It makes it much easier to understand the change when looking
back at the log in future.

Thanks,

Will

[toc] | [prev] | [next] | [standalone]

#1577132 — Re: [BUGFIX PATCH tip/master 2/3] kprobes/arm64: Fix a possible deadlock case in kretprobe

On Wed, 8 Feb 2017 15:06:55 +0000
Will Deacon <will.deacon@arm.com> wrote:

> [adding linux-arm-kernel]
> 
> On Wed, Feb 08, 2017 at 12:13:14AM +0900, Masami Hiramatsu wrote:
> > Similar to x86 kretprobe deadlock issue, arm64 also implements
> > kretprobe-booster (trampoline code directly call handler.)
> > So it has same deadlock issue if there are 2 kretprobes on
> > normal function and the function called from FIQ (or anywhere
> > which can be invoked when local_irq_disabled).
> 
> We don't support FIQ on arm64, so I'm not worried about that particular
> case. What are the other cases? I can think of debug exceptions, but those
> shouldn't be generally kprobe-able, and taking data aborts in things like
> get_user/put_user. Are those affected by this bug?

Hmm, in that case, this may not needed at this point. Would you have
any plan to support FIQ like as NMI in x86?
If something can interrupt while the critical region between 
spin_lock_irqsave() and spin_unlock_irqrestore(), and it can be
kprobe'd, it is safer to apply this patch.

> Either way, could you please expand the commit message like you have
> for x86? It makes it much easier to understand the change when looking
> back at the log in future.

Ah, sorry. I will update the comment.

Thank you,

> 
> Thanks,
> 
> Will


-- 
Masami Hiramatsu <mhiramat@kernel.org>

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web