Re: [PATCH] bpf: hide do_bpf_send_signal when unused

From: Yonghong Song
Date: Mon Jun 24 2019 - 16:38:59 EST




On 6/17/19 5:18 PM, Steven Rostedt wrote:
> On Mon, 17 Jun 2019 16:27:33 -0700
> Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote:
>
>> On Mon, Jun 17, 2019 at 4:13 PM Matt Mullins <mmullins@xxxxxx> wrote:
>>>>
>>>> The bug (really just a warning) reported is exactly here.
>>>
>>> I don't think bpf_send_signal is tied to modules at all;
>>> send_signal_irq_work_init and the corresponding initcall should be
>>> moved outside that #ifdef.
>>
>> right. I guess send_signal_irq_work_init was accidentally placed
>> after bpf_event_init and happened to be within that ifdef.
>> Should definitely be outside.
>
> So Arnd did find a bug. Just the wrong solution ;-)
>
> -- Steve

Hi, Arnd,

The following change can fix the issue.

diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index c102c240bb0b..ca1255d14576 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -1431,6 +1431,20 @@ int bpf_get_perf_event_info(const struct
perf_event *event, u32 *prog_id,
return err;
}

+static int __init send_signal_irq_work_init(void)
+{
+ int cpu;
+ struct send_signal_irq_work *work;
+
+ for_each_possible_cpu(cpu) {
+ work = per_cpu_ptr(&send_signal_work, cpu);
+ init_irq_work(&work->irq_work, do_bpf_send_signal);
+ }
+ return 0;
+}
+
+subsys_initcall(send_signal_irq_work_init);
+
#ifdef CONFIG_MODULES
static int bpf_event_notify(struct notifier_block *nb, unsigned long op,
void *module)
@@ -1478,18 +1492,5 @@ static int __init bpf_event_init(void)
return 0;
}

-static int __init send_signal_irq_work_init(void)
-{
- int cpu;
- struct send_signal_irq_work *work;
-
- for_each_possible_cpu(cpu) {
- work = per_cpu_ptr(&send_signal_work, cpu);
- init_irq_work(&work->irq_work, do_bpf_send_signal);
- }
- return 0;
-}
-
fs_initcall(bpf_event_init);
-subsys_initcall(send_signal_irq_work_init);
#endif /* CONFIG_MODULES */

Could you submit a new revision? Thanks!

Yonghong