Re: perf: allow command to attach local data to thread/evsel structs

[David Ahern <dsahern@gmail.com>]

On 02/07/2012 02:33 PM, Arnaldo Carvalho de Melo wrote:
>>> I think we either use {thread_conf,evsel_conf} for global configuration
>>> options for these two core data structures or we just provide some
>>> optional, per perf_tool allocator.
>>
>> Meaning conf's that parallel symbol_conf -- allocated and 'owned' by the
>> perf library but exported for the user to set values. In this case
>> handlers are needed for allocation and free as instances come and go?
>> e.g., thread__new invokes thread_conf.allocator if defined and
>> thread__priv returns a pointer to private data.
> 
> Well, the symbol way of doing things is to just allocate as many bytes
> as the tool being used asks for and then at free time nothing special
> has to be done.

Consider the case of events and you want to track time last seen by cpu.
You need an array:

u64 last_seen[]

with CPU as an index, but do not want to assume some value of max_cpus.
So one option is:

struct priv {
    u64 *last_seen; /* time this event was last seen */
    u32 max_cpu;    /* highest cpu slot allocated */
};

last_seen is malloc'ed first time the event is processed and realloc'ed
as max_cpu increases. Attaching that struct to an event means an
embedded allocation that does not follow the static model of symbol_conf.

David