[PATCH] record: Add wrappers for malloc() functions

[Chris Johns]

On 2/9/19 4:11 pm, Sebastian Huber wrote:
> On 02/09/2019 08:00, Chris Johns wrote:
>>>> I do not see any upside adding this library or wrapping specific functions
>>>> this way.
>>> The benefits are:
>>>
>>> * Slightly more compact (the event indicates the function and the data can be
>>> used for the caller address instead, the generic function entry/exit needs two
>>> events for this).
>> Could you please explain this, sorry I am not sure what this means? If the code
>> is what needs to be generated then why not generate?
> 
> The record items is defined like this:
> 
> /**
>  * @brief The record data integer type.
>  *
>  * It is big enough to store 32-bit integers and pointers.
>  */
> typedef unsigned long rtems_record_data;
> 
> /**
>  * @brief The native record item.
>  */
> typedef struct __attribute__((__packed__)) {
>   uint32_t          event;
>   rtems_record_data data;
> } rtems_record_item;
> 
> The event member contains a 22-bit time stamp and a 10-bit event number.
> 
> The malloc entry/exit wrapper with a special event number looks like this:
> 
> void *__wrap_malloc( size_t size )
> {
>   void *ptr;
> 
>   rtems_record_produce_2(
>     RTEMS_RECORD_MALLOC_ENTRY,
>     (rtems_record_data) RTEMS_RETURN_ADDRESS(),
>     RTEMS_RECORD_ARG_0,
>     size
>   )
>   ptr = __real_malloc( size );
>     rtems_record_produce_2(
>     RTEMS_RECORD_MALLOC_EXIT,
>     (rtems_record_data) RTEMS_RETURN_ADDRESS(),
>     RTEMS_RECORD_RETURN_0,
>     (rtems_record_data) ptr
>   )
>   return ptr;
> }
> 
> If you have to encode the function in the event data, then you need two extra
> items:
> 
> void *__wrap_malloc( size_t size )
> {
>   rtems_record_item items[ 3 ];
>   void *ptr;
> 
>   items[ 0 ].event = RTEMS_RECORD_FUNCTION_ENTRY;
>   items[ 1 ].data = (rtems_record_data) __real_malloc;
>   items[ 1 ].event = RTEMS_RECORD_CALLER;
>   items[ 1 ].data = (rtems_record_data) RTEMS_RETURN_ADDRESS();
>   items[ 2 ].event = RTEMS_RECORD_ARG_0;
>   items[ 2 ].data = size;
>   rtems_record_produce_n( items, RTEMS_ARRAY_SIZE( items ) );
>   ptr = __real_malloc( size );
>   items[ 0 ].event = RTEMS_RECORD_FUNCTION_EXIT;
>   items[ 1 ].data = (rtems_record_data) __real_malloc;
>   items[ 1 ].event = RTEMS_RECORD_CALLER;
>   items[ 1 ].data = (rtems_record_data) RTEMS_RETURN_ADDRESS();
>   items[ 2 ].event = RTEMS_RECORD_RETURN_0;
>   items[ 2 ].data = (rtems_record_data) ptr;
>   rtems_record_produce_n( items, RTEMS_ARRAY_SIZE( items ) );
>   return ptr;
> }
> 
> For this simple one argument function, this is a 50% increase.
> 

Why not generate the call to `rtems_record_produce_2`? You should be able to
determine this give you will know there in a single item plus the return address
to record. In fact there is little need for `rtems_record_produce_n` if you have
[1..n] already defined.

Chris