Re: [PATCH v2 4/5] bpf: Implement function bpf_perf_event_read() that get the selected hardware PMU conuter

[Alexei Starovoitov <ast@plumgrid.com>]

On 7/22/15 1:09 AM, Kaixu Xia wrote:
> According to the perf_event_map_fd and key, the function
> bpf_perf_event_read() can convert the corresponding map
> value to the pointer to struct perf_event and return the
> Hardware PMU counter value.
>
> The key can't be passed to bpf_perf_event_read() directly
> because the function argument constraint is lacked.

I don't understand above sentence.

> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index 69a1f6b..e3bb181 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -250,6 +250,8 @@ enum bpf_func_id {
>   	 * Return: 0 on success
>   	 */
>   	BPF_FUNC_get_current_comm,
> +
> +	BPF_FUNC_perf_event_read,	/* u64 bpf_perf_event_read(&map, &key) */

no need for extra empty line.

> +
> +static u64 bpf_perf_event_read(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5)
> +{
> +	struct bpf_map *map = (struct bpf_map *) (unsigned long) r1;
> +	void *key = (void *) (unsigned long) r2;
> +	struct perf_event *event;
> +	void *ptr;
> +
> +	if (map->map_type != BPF_MAP_TYPE_PERF_EVENT_ARRAY)
> +		return -EINVAL;

please check this statically in verifier instead of in run-time.

> +
> +	rcu_read_lock();

unnecessary.

> +	ptr = map->ops->map_lookup_elem(map, key);
> +	rcu_read_unlock();
> +	if (!ptr || !(*(unsigned long *)ptr))
> +		return -EBADF;

all these casts can be removed. First cast of 'r1' into
perf_event_array will be enough.

> +const struct bpf_func_proto bpf_perf_event_read_proto = {
> +	.func		= bpf_perf_event_read,
> +	.gpl_only	= false,
> +	.ret_type	= RET_INTEGER,
> +	.arg1_type	= ARG_CONST_MAP_PTR,
> +	.arg2_type	= ARG_PTR_TO_MAP_KEY,

make it arg2_type = ARG_ANYTHING then you'll just index
into array the way prog_array does and similar to bpf_tail_call.