2017-12-16 02:43:02

by Josef Bacik

Subject: [PATCH] trace: reenable preemption if we modify the ip

From: Josef Bacik <[email protected]>

Things got moved around between the original bpf_override_return patches
and the final version, and now the ftrace kprobe dispatcher assumes if
you modified the ip that you also enabled preemption. Make a comment of
this and enable preemption, this fixes the lockdep splat that happened
when using this feature.

Signed-off-by: Josef Bacik <[email protected]>
---
kernel/trace/trace_kprobe.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index 5db849809a56..91f4b57dab82 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -1322,8 +1322,15 @@ static int kprobe_dispatcher(struct kprobe *kp, struct pt_regs *regs)
if (tk->tp.flags & TP_FLAG_TRACE)
kprobe_trace_func(tk, regs);
#ifdef CONFIG_PERF_EVENTS
- if (tk->tp.flags & TP_FLAG_PROFILE)
+ if (tk->tp.flags & TP_FLAG_PROFILE) {
ret = kprobe_perf_func(tk, regs);
+ /*
+ * The ftrace kprobe handler leaves it up to us to re-enable
+ * preemption here before returning if we've modified the ip.
+ */
+ if (ret)
+ preempt_enable_no_resched();
+ }
#endif
return ret;
}
--
2.7.5

2017-12-17 19:49:33

by Daniel Borkmann

Subject: Re: [PATCH] trace: reenable preemption if we modify the ip

On 12/16/2017 03:42 AM, Josef Bacik wrote:
> From: Josef Bacik <[email protected]>
>
> Things got moved around between the original bpf_override_return patches
> and the final version, and now the ftrace kprobe dispatcher assumes if
> you modified the ip that you also enabled preemption. Make a comment of
> this and enable preemption, this fixes the lockdep splat that happened
> when using this feature.
>
> Signed-off-by: Josef Bacik <[email protected]>

Applied to bpf-next with Fixes tag, thanks Josef.

2017-12-18 07:29:53

by Masami Hiramatsu

Subject: Re: [PATCH] trace: reenable preemption if we modify the ip

On Fri, 15 Dec 2017 21:42:57 -0500
Josef Bacik <[email protected]> wrote:

> From: Josef Bacik <[email protected]>
>
> Things got moved around between the original bpf_override_return patches
> and the final version, and now the ftrace kprobe dispatcher assumes if
> you modified the ip that you also enabled preemption. Make a comment of
> this and enable preemption, this fixes the lockdep splat that happened
> when using this feature.
>
> Signed-off-by: Josef Bacik <[email protected]>
> ---
> kernel/trace/trace_kprobe.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
> index 5db849809a56..91f4b57dab82 100644
> --- a/kernel/trace/trace_kprobe.c
> +++ b/kernel/trace/trace_kprobe.c
> @@ -1322,8 +1322,15 @@ static int kprobe_dispatcher(struct kprobe *kp, struct pt_regs *regs)
> if (tk->tp.flags & TP_FLAG_TRACE)
> kprobe_trace_func(tk, regs);
> #ifdef CONFIG_PERF_EVENTS
> - if (tk->tp.flags & TP_FLAG_PROFILE)
> + if (tk->tp.flags & TP_FLAG_PROFILE) {
> ret = kprobe_perf_func(tk, regs);
> + /*
> + * The ftrace kprobe handler leaves it up to us to re-enable
> + * preemption here before returning if we've modified the ip.
> + */
> + if (ret)
> + preempt_enable_no_resched();

Where is reset_current_kprobe()?
Since kprobes still expects this modification is used by jprobes,
we need to call it in caller-side.

Thank you,

> + }
> #endif
> return ret;
> }
> --
> 2.7.5
>


--
Masami Hiramatsu <[email protected]>