[PATCH v1 0/1] libbpf: perfbuf custom event reader

[PATCH v1 0/1] libbpf: perfbuf custom event reader

[Jon Doron]

From: Jon Doron <jond@wiz.io>

Add support for writing a custom event reader, by exposing the ring
buffer state, and allowing to set it's tail.

Few simple examples where this type of needed:
1. perf_event_read_simple is allocating using malloc, perhaps you want
   to handle the wrap-around in some other way.
2. Since perf buf is per-cpu then the order of the events is not
   guarnteed, for example:
   Given 3 events where each event has a timestamp t0 < t1 < t2,
   and the events are spread on more than 1 CPU, then we can end
   up with the following state in the ring buf:
   CPU[0] => [t0, t2]
   CPU[1] => [t1]
   When you consume the events from CPU[0], you could know there is
   a t1 missing, (assuming there are no drops, and your event data
   contains a sequential index).
   So now one can simply do the following, for CPU[0], you can store
   the address of t0 and t2 in an array (without moving the tail, so
   there data is not perished) then move on the CPU[1] and set the
   address of t1 in the same array.
   So you end up with something like:
   void **arr[] = [&t0, &t1, &t2], now you can consume it orderely
   and move the tails as you process in order.

Jon Doron (1):
  libbpf: perfbuf: allow raw access to buffers

 tools/lib/bpf/libbpf.c | 40 ++++++++++++++++++++++++++++++++++++++++
 tools/lib/bpf/libbpf.h |  6 ++++++
 2 files changed, 46 insertions(+)

-- 
2.36.1

[PATCH v1 1/1] libbpf: perfbuf: allow raw access to buffers

[Jon Doron]

From: Jon Doron <jond@wiz.io>

Add API for perfbuf to support writing a custom event reader.

Signed-off-by: Jon Doron <jond@wiz.io>
---
 tools/lib/bpf/libbpf.c | 40 ++++++++++++++++++++++++++++++++++++++++
 tools/lib/bpf/libbpf.h |  6 ++++++
 2 files changed, 46 insertions(+)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index e89cc9c885b3..37299aa05185 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -12433,6 +12433,46 @@ static int perf_buffer__process_records(struct perf_buffer *pb,
 	return 0;
 }
 
+int perf_buffer__raw_ring_buf(const struct perf_buffer *pb, size_t buf_idx,
+			      void **base, size_t *buf_size, __u64 *head,
+			      __u64 *tail)
+{
+	struct perf_cpu_buf *cpu_buf;
+	struct perf_event_mmap_page *header;
+
+	if (buf_idx >= pb->cpu_cnt)
+		return libbpf_err(-EINVAL);
+
+	cpu_buf = pb->cpu_bufs[buf_idx];
+	if (!cpu_buf)
+		return libbpf_err(-ENOENT);
+
+	header = cpu_buf->base;
+	*head = ring_buffer_read_head(header);
+	*tail = header->data_tail;
+	*base = ((__u8 *)header) + pb->page_size;
+	*buf_size = pb->mmap_size;
+	return 0;
+}
+
+int perf_buffer__set_ring_buf_tail(const struct perf_buffer *pb, size_t buf_idx,
+				   __u64 tail)
+{
+	struct perf_cpu_buf *cpu_buf;
+	struct perf_event_mmap_page *header;
+
+	if (buf_idx >= pb->cpu_cnt)
+		return libbpf_err(-EINVAL);
+
+	cpu_buf = pb->cpu_bufs[buf_idx];
+	if (!cpu_buf)
+		return libbpf_err(-ENOENT);
+
+	header = cpu_buf->base;
+	ring_buffer_write_tail(header, tail);
+	return 0;
+}
+
 int perf_buffer__epoll_fd(const struct perf_buffer *pb)
 {
 	return pb->epoll_fd;
diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index 9e9a3fd3edd8..b6f6b6a12d70 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -1381,6 +1381,12 @@ LIBBPF_API int perf_buffer__consume(struct perf_buffer *pb);
 LIBBPF_API int perf_buffer__consume_buffer(struct perf_buffer *pb, size_t buf_idx);
 LIBBPF_API size_t perf_buffer__buffer_cnt(const struct perf_buffer *pb);
 LIBBPF_API int perf_buffer__buffer_fd(const struct perf_buffer *pb, size_t buf_idx);
+LIBBPF_API int perf_buffer__raw_ring_buf(const struct perf_buffer *pb,
+					 size_t buf_idx, void **base,
+					 size_t *buf_size, __u64 *head,
+					 __u64 *tail);
+LIBBPF_API int perf_buffer__set_ring_buf_tail(const struct perf_buffer *pb,
+					      size_t buf_idx, __u64 tail);
 
 typedef enum bpf_perf_event_ret
 	(*bpf_perf_event_print_t)(struct perf_event_header *hdr,
-- 
2.36.1

Re: [PATCH v1 1/1] libbpf: perfbuf: allow raw access to buffers

[Song Liu]

On Thu, Jul 7, 2022 at 12:14 AM Jon Doron <arilou@gmail.com> wrote:

Please prefix the patch with the target tree (bpf or bpf-next). For example,
this patch should go via bpf-next.
>
> From: Jon Doron <jond@wiz.io>
>
> Add API for perfbuf to support writing a custom event reader.

This is too brief for such change. It is ok to duplicate text from the cover
letter.

Please also update libbpf.map.

Also, we should add a selftest to use these new APIs.

>
> Signed-off-by: Jon Doron <jond@wiz.io>
> ---
>  tools/lib/bpf/libbpf.c | 40 ++++++++++++++++++++++++++++++++++++++++
>  tools/lib/bpf/libbpf.h |  6 ++++++
>  2 files changed, 46 insertions(+)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index e89cc9c885b3..37299aa05185 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -12433,6 +12433,46 @@ static int perf_buffer__process_records(struct perf_buffer *pb,
>         return 0;
>  }
>
> +int perf_buffer__raw_ring_buf(const struct perf_buffer *pb, size_t buf_idx,
> +                             void **base, size_t *buf_size, __u64 *head,
> +                             __u64 *tail)

Please add comments to each API function.
> +{
> +       struct perf_cpu_buf *cpu_buf;
> +       struct perf_event_mmap_page *header;
> +
> +       if (buf_idx >= pb->cpu_cnt)
> +               return libbpf_err(-EINVAL);
> +
> +       cpu_buf = pb->cpu_bufs[buf_idx];
> +       if (!cpu_buf)
> +               return libbpf_err(-ENOENT);
> +
> +       header = cpu_buf->base;
> +       *head = ring_buffer_read_head(header);
> +       *tail = header->data_tail;
> +       *base = ((__u8 *)header) + pb->page_size;
> +       *buf_size = pb->mmap_size;
> +       return 0;
> +}
> +
> +int perf_buffer__set_ring_buf_tail(const struct perf_buffer *pb, size_t buf_idx,
> +                                  __u64 tail)
> +{
> +       struct perf_cpu_buf *cpu_buf;
> +       struct perf_event_mmap_page *header;
> +
> +       if (buf_idx >= pb->cpu_cnt)
> +               return libbpf_err(-EINVAL);
> +
> +       cpu_buf = pb->cpu_bufs[buf_idx];
> +       if (!cpu_buf)
> +               return libbpf_err(-ENOENT);
> +
> +       header = cpu_buf->base;
> +       ring_buffer_write_tail(header, tail);
> +       return 0;
> +}
> +
>  int perf_buffer__epoll_fd(const struct perf_buffer *pb)
>  {
>         return pb->epoll_fd;
> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
> index 9e9a3fd3edd8..b6f6b6a12d70 100644
> --- a/tools/lib/bpf/libbpf.h
> +++ b/tools/lib/bpf/libbpf.h
> @@ -1381,6 +1381,12 @@ LIBBPF_API int perf_buffer__consume(struct perf_buffer *pb);
>  LIBBPF_API int perf_buffer__consume_buffer(struct perf_buffer *pb, size_t buf_idx);
>  LIBBPF_API size_t perf_buffer__buffer_cnt(const struct perf_buffer *pb);
>  LIBBPF_API int perf_buffer__buffer_fd(const struct perf_buffer *pb, size_t buf_idx);
> +LIBBPF_API int perf_buffer__raw_ring_buf(const struct perf_buffer *pb,
> +                                        size_t buf_idx, void **base,
> +                                        size_t *buf_size, __u64 *head,
> +                                        __u64 *tail);
> +LIBBPF_API int perf_buffer__set_ring_buf_tail(const struct perf_buffer *pb,
> +                                             size_t buf_idx, __u64 tail);
>
>  typedef enum bpf_perf_event_ret
>         (*bpf_perf_event_print_t)(struct perf_event_header *hdr,
> --
> 2.36.1
>

Re: [PATCH v1 1/1] libbpf: perfbuf: allow raw access to buffers

[Jon Doron]

On 07/07/2022, Song Liu wrote:

Hi Song Liu, thank you for the fast reply, I'll add comments on top
of your original email.

I will not be available through out next week, so sorry if I'll have a 
late reply for follow ups.

Thanks in advance,
-- Jon.

>On Thu, Jul 7, 2022 at 12:14 AM Jon Doron <arilou@gmail.com> wrote:
>
>Please prefix the patch with the target tree (bpf or bpf-next). For example,
>this patch should go via bpf-next.

Done

>>
>> From: Jon Doron <jond@wiz.io>
>>
>> Add API for perfbuf to support writing a custom event reader.
>
>This is too brief for such change. It is ok to duplicate text from the cover
>letter.
>

Done

>Please also update libbpf.map.
>

Done

>Also, we should add a selftest to use these new APIs.
>

I'm not sure I'm following here what were you had in mind, in practice I 
could just change **bpf_perf_event_read_simple** implementation to use
these new APIs to initialize it's variables (data_head, data_tail, 
base, mmap_size) and it would act exactly the same.

>>
>> Signed-off-by: Jon Doron <jond@wiz.io>
>> ---
>>  tools/lib/bpf/libbpf.c | 40 ++++++++++++++++++++++++++++++++++++++++
>>  tools/lib/bpf/libbpf.h |  6 ++++++
>>  2 files changed, 46 insertions(+)
>>
>> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
>> index e89cc9c885b3..37299aa05185 100644
>> --- a/tools/lib/bpf/libbpf.c
>> +++ b/tools/lib/bpf/libbpf.c
>> @@ -12433,6 +12433,46 @@ static int perf_buffer__process_records(struct perf_buffer *pb,
>>         return 0;
>>  }
>>
>> +int perf_buffer__raw_ring_buf(const struct perf_buffer *pb, size_t buf_idx,
>> +                             void **base, size_t *buf_size, __u64 *head,
>> +                             __u64 *tail)
>
>Please add comments to each API function.

Done

>> +{
>> +       struct perf_cpu_buf *cpu_buf;
>> +       struct perf_event_mmap_page *header;
>> +
>> +       if (buf_idx >= pb->cpu_cnt)
>> +               return libbpf_err(-EINVAL);
>> +
>> +       cpu_buf = pb->cpu_bufs[buf_idx];
>> +       if (!cpu_buf)
>> +               return libbpf_err(-ENOENT);
>> +
>> +       header = cpu_buf->base;
>> +       *head = ring_buffer_read_head(header);
>> +       *tail = header->data_tail;
>> +       *base = ((__u8 *)header) + pb->page_size;
>> +       *buf_size = pb->mmap_size;
>> +       return 0;
>> +}
>> +
>> +int perf_buffer__set_ring_buf_tail(const struct perf_buffer *pb, size_t buf_idx,
>> +                                  __u64 tail)
>> +{
>> +       struct perf_cpu_buf *cpu_buf;
>> +       struct perf_event_mmap_page *header;
>> +
>> +       if (buf_idx >= pb->cpu_cnt)
>> +               return libbpf_err(-EINVAL);
>> +
>> +       cpu_buf = pb->cpu_bufs[buf_idx];
>> +       if (!cpu_buf)
>> +               return libbpf_err(-ENOENT);
>> +
>> +       header = cpu_buf->base;
>> +       ring_buffer_write_tail(header, tail);
>> +       return 0;
>> +}
>> +
>>  int perf_buffer__epoll_fd(const struct perf_buffer *pb)
>>  {
>>         return pb->epoll_fd;
>> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
>> index 9e9a3fd3edd8..b6f6b6a12d70 100644
>> --- a/tools/lib/bpf/libbpf.h
>> +++ b/tools/lib/bpf/libbpf.h
>> @@ -1381,6 +1381,12 @@ LIBBPF_API int perf_buffer__consume(struct perf_buffer *pb);
>>  LIBBPF_API int perf_buffer__consume_buffer(struct perf_buffer *pb, size_t buf_idx);
>>  LIBBPF_API size_t perf_buffer__buffer_cnt(const struct perf_buffer *pb);
>>  LIBBPF_API int perf_buffer__buffer_fd(const struct perf_buffer *pb, size_t buf_idx);
>> +LIBBPF_API int perf_buffer__raw_ring_buf(const struct perf_buffer *pb,
>> +                                        size_t buf_idx, void **base,
>> +                                        size_t *buf_size, __u64 *head,
>> +                                        __u64 *tail);
>> +LIBBPF_API int perf_buffer__set_ring_buf_tail(const struct perf_buffer *pb,
>> +                                             size_t buf_idx, __u64 tail);
>>
>>  typedef enum bpf_perf_event_ret
>>         (*bpf_perf_event_print_t)(struct perf_event_header *hdr,
>> --
>> 2.36.1
>>