Re: [PATCH 02b/10] perf, trace: Fix probe unregister race

[Frederic Weisbecker]

On Fri, May 21, 2010 at 12:41:16PM +0200, Peter Zijlstra wrote:
> tracepoint_probe_unregister() does not synchronize against the probe callbacks,
> so do that explicitly. This properly serializes the callbacks and the free of
> the data used therein.
> 
> Also, use this_cpu_ptr() where possible.
> 
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>


Acked-by: Frederic Weisbecker <fweisbec@xxxxxxxxx>

(and you previous patches too)



> ---
>  include/trace/ftrace.h          |    2 +-
>  kernel/trace/trace_event_perf.c |   10 ++++++++--
>  kernel/trace/trace_kprobe.c     |    4 ++--
>  kernel/trace/trace_syscalls.c   |    4 ++--
>  4 files changed, 13 insertions(+), 7 deletions(-)
> 
> Index: linux-2.6/include/trace/ftrace.h
> ===================================================================
> --- linux-2.6.orig/include/trace/ftrace.h
> +++ linux-2.6/include/trace/ftrace.h
> @@ -791,7 +791,7 @@ perf_trace_templ_##call(struct ftrace_ev
>  									\
>  	{ assign; }							\
>  									\
> -	head = per_cpu_ptr(event_call->perf_events, smp_processor_id());\
> +	head = this_cpu_ptr(event_call->perf_events);			\
>  	perf_trace_buf_submit(entry, __entry_size, rctx, __addr,	\
>  		__count, __regs, head);					\
>  }
> Index: linux-2.6/kernel/trace/trace_event_perf.c
> ===================================================================
> --- linux-2.6.orig/kernel/trace/trace_event_perf.c
> +++ linux-2.6/kernel/trace/trace_event_perf.c
> @@ -109,7 +109,7 @@ int perf_trace_enable(struct perf_event 
>  	if (WARN_ON_ONCE(!list))
>  		return -EINVAL;
>  
> -	list = per_cpu_ptr(list, smp_processor_id());
> +	list = this_cpu_ptr(list);
>  	hlist_add_head_rcu(&p_event->hlist_entry, list);
>  
>  	return 0;
> @@ -130,6 +130,12 @@ void perf_trace_destroy(struct perf_even
>  
>  	tp_event->perf_event_disable(tp_event);
>  
> +	/*
> +	 * Ensure our callback won't be called anymore. See
> +	 * tracepoint_probe_unregister() and __DO_TRACE().
> +	 */
> +	synchronize_sched();
> +
>  	free_percpu(tp_event->perf_events);
>  	tp_event->perf_events = NULL;
>  
> @@ -156,7 +162,7 @@ __kprobes void *perf_trace_buf_prepare(i
>  	if (*rctxp < 0)
>  		return NULL;
>  
> -	raw_data = per_cpu_ptr(perf_trace_buf[*rctxp], smp_processor_id());
> +	raw_data = this_cpu_ptr(perf_trace_buf[*rctxp]);
>  
>  	/* zero the dead bytes from align to not leak stack to user */
>  	memset(&raw_data[size - sizeof(u64)], 0, sizeof(u64));
> Index: linux-2.6/kernel/trace/trace_kprobe.c
> ===================================================================
> --- linux-2.6.orig/kernel/trace/trace_kprobe.c
> +++ linux-2.6/kernel/trace/trace_kprobe.c
> @@ -1362,7 +1362,7 @@ static __kprobes void kprobe_perf_func(s
>  	for (i = 0; i < tp->nr_args; i++)
>  		call_fetch(&tp->args[i].fetch, regs, data + tp->args[i].offset);
>  
> -	head = per_cpu_ptr(call->perf_events, smp_processor_id());
> +	head = this_cpu_ptr(call->perf_events);
>  	perf_trace_buf_submit(entry, size, rctx, entry->ip, 1, regs, head);
>  }
>  
> @@ -1395,7 +1395,7 @@ static __kprobes void kretprobe_perf_fun
>  	for (i = 0; i < tp->nr_args; i++)
>  		call_fetch(&tp->args[i].fetch, regs, data + tp->args[i].offset);
>  
> -	head = per_cpu_ptr(call->perf_events, smp_processor_id());
> +	head = this_cpu_ptr(call->perf_events);
>  	perf_trace_buf_submit(entry, size, rctx, entry->ret_ip, 1, regs, head);
>  }
>  
> Index: linux-2.6/kernel/trace/trace_syscalls.c
> ===================================================================
> --- linux-2.6.orig/kernel/trace/trace_syscalls.c
> +++ linux-2.6/kernel/trace/trace_syscalls.c
> @@ -469,7 +469,7 @@ static void perf_syscall_enter(struct pt
>  	syscall_get_arguments(current, regs, 0, sys_data->nb_args,
>  			       (unsigned long *)&rec->args);
>  
> -	head = per_cpu_ptr(sys_data->enter_event->perf_events, smp_processor_id());
> +	head = this_cpu_ptr(sys_data->enter_event->perf_events);
>  	perf_trace_buf_submit(rec, size, rctx, 0, 1, regs, head);
>  }
>  
> @@ -545,7 +545,7 @@ static void perf_syscall_exit(struct pt_
>  	rec->nr = syscall_nr;
>  	rec->ret = syscall_get_return_value(current, regs);
>  
> -	head = per_cpu_ptr(sys_data->exit_event->perf_events, smp_processor_id());
> +	head = this_cpu_ptr(sys_data->exit_event->perf_events);
>  	perf_trace_buf_submit(rec, size, rctx, 0, 1, regs, head);
>  }
>  
> 
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/