Re: [PATCH v3 1/2] ftrace: disable preemption between ftrace_test_recursion_trylock/unlock()

[王贇 <yun.wang@linux.alibaba.com>]

On 2021/10/15 上午11:13, 王贇 wrote:
[snip]
>>  # define do_ftrace_record_recursion(ip, pip)	do { } while (0)
>>  #endif
>>  
>> +/*
>> + * trace_test_and_set_recursion() is called on several layers
>> + * of the ftrace code when handling the same ftrace entry.
>> + * These calls might be nested/recursive.
>> + *
>> + * It uses TRACE_LIST_*BITs to distinguish between this
>> + * internal recursion and recursion caused by calling
>> + * the traced function by the ftrace code.
>> + *
>> + * Returns: > 0 when no recursion
>> + *          0 when called recursively internally (safe)
> 
> The 0 can also happened when ftrace handler recursively called trylock()
> under the same context, or not?
> 

Never mind... you're right about this.

Regards,
Michael Wang

> Regards,
> Michael Wang
> 
>> + *	    -1 when the traced function was called recursively from
>> + *             the ftrace handler (unsafe)
>> + */
>>  static __always_inline int trace_test_and_set_recursion(unsigned long ip, unsigned long pip,
>>  							int start, int max)
>>  {
>>  	unsigned int val = READ_ONCE(current->trace_recursion);
>>  	int bit;
>>  
>> -	/* A previous recursion check was made */
>> +	/* Called recursively internally by different ftrace code layers? */
>>  	if ((val & TRACE_CONTEXT_MASK) > max)
>>  		return 0;
> 
>>  
>>