[1/2] libbpf: ringbuf: allow to consume up to a certain amount of items

Commit Message

Andrea Righi April 1, 2024, 7:19 a.m. UTC

In some cases, instead of always consuming all items from ring buffers
in a greedy way, we may want to consume up to a certain amount of items,
for example when we need to copy items from the BPF ring buffer to a
limited user buffer.

This change allows to set an upper limit to the amount of items consumed
from one or more ring buffers.

Link: https://lore.kernel.org/lkml/[e-mail]/T
Signed-off-by: Andrea Righi <[e-mail]>
---
 tools/lib/bpf/ringbuf.c | 29 +++++++++++++++++------------
 1 file changed, 17 insertions(+), 12 deletions(-)

Comments

Andrii Nakryiko April 2, 2024, 5:58 p.m. UTC | #1

On Mon, Apr 1, 2024 at 12:32 AM Andrea Righi <[e-mail]> wrote:
>
> In some cases, instead of always consuming all items from ring buffers
> in a greedy way, we may want to consume up to a certain amount of items,
> for example when we need to copy items from the BPF ring buffer to a
> limited user buffer.
>
> This change allows to set an upper limit to the amount of items consumed
> from one or more ring buffers.
>
> Link: https://lore.kernel.org/lkml/[e-mail]/T
> Signed-off-by: Andrea Righi <[e-mail]>
> ---
>  tools/lib/bpf/ringbuf.c | 29 +++++++++++++++++------------
>  1 file changed, 17 insertions(+), 12 deletions(-)
>
> diff --git a/tools/lib/bpf/ringbuf.c b/tools/lib/bpf/ringbuf.c
> index aacb64278a01..81df535040d1 100644
> --- a/tools/lib/bpf/ringbuf.c
> +++ b/tools/lib/bpf/ringbuf.c
> @@ -231,7 +231,7 @@ static inline int roundup_len(__u32 len)
>         return (len + 7) / 8 * 8;
>  }
>
> -static int64_t ringbuf_process_ring(struct ring *r)
> +static int64_t ringbuf_process_ring(struct ring *r, int64_t max_items)
>  {
>         int *len_ptr, len, err;
>         /* 64-bit to avoid overflow in case of extreme application behavior */
> @@ -264,7 +264,14 @@ static int64_t ringbuf_process_ring(struct ring *r)
>                                                           cons_pos);
>                                         return err;
>                                 }
> -                               cnt++;
> +                               if (++cnt >= max_items) {
> +                                       /* update consumer pos and return the
> +                                        * total amount of items consumed.
> +                                        */
> +                                       smp_store_release(r->consumer_pos,
> +                                                         cons_pos);

Does this fit on a single line under 100 characters? If yes, please
keep it as a single line

but actually it seems cleaner to keep cnt++ intact, let
smp_store_release() below happen, and then check the exit condition.
Were you afraid to do unnecessary checks on discarded samples? I
wouldn't worry about that.

> +                                       goto done;
> +                               }
>                         }
>
>                         smp_store_release(r->consumer_pos, cons_pos);
> @@ -281,19 +288,18 @@ static int64_t ringbuf_process_ring(struct ring *r)
>   */
>  int ring_buffer__consume(struct ring_buffer *rb)
>  {
> -       int64_t err, res = 0;
> +       int64_t err, res = 0, max_items = INT_MAX;

I'm wondering if it might be better to have a convention that zero
means "no limit", which might allow the compiler to eliminate the
condition in ringbuf_process_ring altogether due to constant
propagation. WDYT?

>         int i;
>
>         for (i = 0; i < rb->ring_cnt; i++) {
>                 struct ring *ring = rb->rings[i];
>
> -               err = ringbuf_process_ring(ring);
> +               err = ringbuf_process_ring(ring, max_items);
>                 if (err < 0)
>                         return libbpf_err(err);
>                 res += err;
> +               max_items -= err;
>         }
> -       if (res > INT_MAX)
> -               return INT_MAX;
>         return res;
>  }

[...]

Andrea Righi April 2, 2024, 8:37 p.m. UTC | #2

On Tue, Apr 02, 2024 at 10:58:33AM -0700, Andrii Nakryiko wrote:
> On Mon, Apr 1, 2024 at 12:32 AM Andrea Righi <[e-mail]> wrote:
> >
> > In some cases, instead of always consuming all items from ring buffers
> > in a greedy way, we may want to consume up to a certain amount of items,
> > for example when we need to copy items from the BPF ring buffer to a
> > limited user buffer.
> >
> > This change allows to set an upper limit to the amount of items consumed
> > from one or more ring buffers.
> >
> > Link: https://lore.kernel.org/lkml/[e-mail]/T
> > Signed-off-by: Andrea Righi <[e-mail]>
> > ---
> >  tools/lib/bpf/ringbuf.c | 29 +++++++++++++++++------------
> >  1 file changed, 17 insertions(+), 12 deletions(-)
> >
> > diff --git a/tools/lib/bpf/ringbuf.c b/tools/lib/bpf/ringbuf.c
> > index aacb64278a01..81df535040d1 100644
> > --- a/tools/lib/bpf/ringbuf.c
> > +++ b/tools/lib/bpf/ringbuf.c
> > @@ -231,7 +231,7 @@ static inline int roundup_len(__u32 len)
> >         return (len + 7) / 8 * 8;
> >  }
> >
> > -static int64_t ringbuf_process_ring(struct ring *r)
> > +static int64_t ringbuf_process_ring(struct ring *r, int64_t max_items)
> >  {
> >         int *len_ptr, len, err;
> >         /* 64-bit to avoid overflow in case of extreme application behavior */
> > @@ -264,7 +264,14 @@ static int64_t ringbuf_process_ring(struct ring *r)
> >                                                           cons_pos);
> >                                         return err;
> >                                 }
> > -                               cnt++;
> > +                               if (++cnt >= max_items) {
> > +                                       /* update consumer pos and return the
> > +                                        * total amount of items consumed.
> > +                                        */
> > +                                       smp_store_release(r->consumer_pos,
> > +                                                         cons_pos);
> 
> Does this fit on a single line under 100 characters? If yes, please
> keep it as a single line
> 
> but actually it seems cleaner to keep cnt++ intact, let
> smp_store_release() below happen, and then check the exit condition.
> Were you afraid to do unnecessary checks on discarded samples? I
> wouldn't worry about that.

Ok, it makes sense, I'll change it.

> 
> > +                                       goto done;
> > +                               }
> >                         }
> >
> >                         smp_store_release(r->consumer_pos, cons_pos);
> > @@ -281,19 +288,18 @@ static int64_t ringbuf_process_ring(struct ring *r)
> >   */
> >  int ring_buffer__consume(struct ring_buffer *rb)
> >  {
> > -       int64_t err, res = 0;
> > +       int64_t err, res = 0, max_items = INT_MAX;
> 
> I'm wondering if it might be better to have a convention that zero
> means "no limit", which might allow the compiler to eliminate the
> condition in ringbuf_process_ring altogether due to constant
> propagation. WDYT?

Indeed, in this way we won't add any potential overhead to the existing
code that doesn't care about max_items. Will add that.

-Andrea

Patch

diff --git a/tools/lib/bpf/ringbuf.c b/tools/lib/bpf/ringbuf.c
index aacb64278a01..81df535040d1 100644
--- a/tools/lib/bpf/ringbuf.c
+++ b/tools/lib/bpf/ringbuf.c
@@ -231,7 +231,7 @@  static inline int roundup_len(__u32 len)
 	return (len + 7) / 8 * 8;
 }
 
-static int64_t ringbuf_process_ring(struct ring *r)
+static int64_t ringbuf_process_ring(struct ring *r, int64_t max_items)
 {
 	int *len_ptr, len, err;
 	/* 64-bit to avoid overflow in case of extreme application behavior */
@@ -264,7 +264,14 @@  static int64_t ringbuf_process_ring(struct ring *r)
 							  cons_pos);
 					return err;
 				}
-				cnt++;
+				if (++cnt >= max_items) {
+					/* update consumer pos and return the
+					 * total amount of items consumed.
+					 */
+					smp_store_release(r->consumer_pos,
+							  cons_pos);
+					goto done;
+				}
 			}
 
 			smp_store_release(r->consumer_pos, cons_pos);
@@ -281,19 +288,18 @@  static int64_t ringbuf_process_ring(struct ring *r)
  */
 int ring_buffer__consume(struct ring_buffer *rb)
 {
-	int64_t err, res = 0;
+	int64_t err, res = 0, max_items = INT_MAX;
 	int i;
 
 	for (i = 0; i < rb->ring_cnt; i++) {
 		struct ring *ring = rb->rings[i];
 
-		err = ringbuf_process_ring(ring);
+		err = ringbuf_process_ring(ring, max_items);
 		if (err < 0)
 			return libbpf_err(err);
 		res += err;
+		max_items -= err;
 	}
-	if (res > INT_MAX)
-		return INT_MAX;
 	return res;
 }
 
@@ -304,7 +310,7 @@  int ring_buffer__consume(struct ring_buffer *rb)
 int ring_buffer__poll(struct ring_buffer *rb, int timeout_ms)
 {
 	int i, cnt;
-	int64_t err, res = 0;
+	int64_t err, res = 0, max_items = INT_MAX;
 
 	cnt = epoll_wait(rb->epoll_fd, rb->events, rb->ring_cnt, timeout_ms);
 	if (cnt < 0)
@@ -314,13 +320,12 @@  int ring_buffer__poll(struct ring_buffer *rb, int timeout_ms)
 		__u32 ring_id = rb->events[i].data.fd;
 		struct ring *ring = rb->rings[ring_id];
 
-		err = ringbuf_process_ring(ring);
+		err = ringbuf_process_ring(ring, max_items);
 		if (err < 0)
 			return libbpf_err(err);
 		res += err;
+		max_items -= err;
 	}
-	if (res > INT_MAX)
-		return INT_MAX;
 	return res;
 }
 
@@ -375,11 +380,11 @@  int ring__consume(struct ring *r)
 {
 	int64_t res;
 
-	res = ringbuf_process_ring(r);
+	res = ringbuf_process_ring(r, INT_MAX);
 	if (res < 0)
 		return libbpf_err(res);
 
-	return res > INT_MAX ? INT_MAX : res;
+	return res;
 }
 
 static void user_ringbuf_unmap_ring(struct user_ring_buffer *rb)