Re: [PATCH 10/13] tracing/uprobes: Fetch args before reserving a ring buffer

[Namhyung Kim <namhyung@kernel.org>]

Hi Oleg,

On Thu, 31 Oct 2013 19:16:54 +0100, Oleg Nesterov wrote:
> On 10/29, Namhyung Kim wrote:
>>
>> @@ -630,6 +653,19 @@ probe_event_enable(struct trace_uprobe *tu, int flag, filter_func_t filter)
>>  	if (trace_probe_is_enabled(&tu->p))
>>  		return -EINTR;
>>  
>> +	if (atomic_inc_return(&uprobe_buffer_ref) == 1) {
>> +		int cpu;
>> +
>> +		uprobe_cpu_buffer = __alloc_percpu(PAGE_SIZE, PAGE_SIZE);
>> +		if (uprobe_cpu_buffer == NULL) {
>> +			atomic_dec(&uprobe_buffer_ref);
>> +			return -ENOMEM;
>> +		}
>> +
>> +		for_each_possible_cpu(cpu)
>> +			mutex_init(&per_cpu(uprobe_cpu_mutex, cpu));
>> +	}
>> +
>>  	WARN_ON(!uprobe_filter_is_empty(&tu->filter));
>>  
>>  	tu->p.flags |= flag;
>> @@ -646,6 +682,11 @@ static void probe_event_disable(struct trace_uprobe *tu, int flag)
>>  	if (!trace_probe_is_enabled(&tu->p))
>>  		return;
>>  
>> +	if (atomic_dec_and_test(&uprobe_buffer_ref)) {
>> +		free_percpu(uprobe_cpu_buffer);
>> +		uprobe_cpu_buffer = NULL;
>> +	}
>> +
>>  	WARN_ON(!uprobe_filter_is_empty(&tu->filter));
>
> Do we really need atomic_t? probe_event_enable/disable is called under
> event_mutex and we rely on this fact anyway.

Looking at the code, it seems probe_event_enable/disable() is called
without event_mutex when it called from sys_perf_event_open().  So we
still need to protect refcount from concurrent accesses.

>
> Otherwise this logic looks racy even with atomic_t, another thread could
> use the uninitialized uprobe_cpu_buffer/mutex if it registers another probe
> and the handler runs before we complete the initialization, no?

But yeah, this is indeed a problem.  Thanks for pointing it out.  I'll
put a mutex to prevent such cases.

Thanks,
Namhyung