Re: [PATCH] perf_counter: extensible perf_counter_attr

[Corey Ashford <cjashfor@linux.vnet.ibm.com>]

Hi Peter,

Peter Zijlstra wrote:
> Allow extending the perf_counter_attr structure by linking extended
> structures to it.
> 
> Also, should we grow the directly reserved space in the structure a
> little more?
> 
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> ---
>  include/linux/perf_counter.h |    7 +++++++
>  kernel/perf_counter.c        |    1 +
>  2 files changed, 8 insertions(+), 0 deletions(-)
> 
> diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
> index 3586df8..781d8ce 100644
> --- a/include/linux/perf_counter.h
> +++ b/include/linux/perf_counter.h
> @@ -175,6 +175,13 @@ struct perf_counter_attr {
>  	__u32			__reserved_3;
> 
>  	__u64			__reserved_4;
> +
> +	struct perf_counter_attr_ext *ext_attrs;
> +};
> +
> +struct perf_counter_attr_ext {
> +	struct perf_counter_attr_ext 	*next;
> +	__u64				perf_attr_ext_type;
>  };

Let's say I want to extend the attributes by four 64-bit quantities... from the 
above definition, I'd need four additional records chained together, right?  How 
about something like this instead:

struct perf_counter_attr_ext {
     __u32	num_perf_attrs;
     __u64 	*perf_attr_ext;
};

Another alternative would be to place these two fields into perf_counter_attr 
and so eliminate the extra level of indirection.


Regards,

- Corey

Corey Ashford
Software Engineer
IBM Linux Technology Center, Linux Toolchain
cjashfor@us.ibm.com