[RFC PATCH v2 0/5] Make BPF ring buffer overwritable

[RFC PATCH v2 0/5] Make BPF ring buffer overwritable

[Francis Laniel]

Hi.


First, I hope you are fine and the same for your relatives.

Normally, when BPF ring buffer are full, producers cannot write anymore and
need to wait for consumer to get some data.
As a consequence, calling bpf_ringbuf_reserve() from eBPF code returns NULL.

This contribution adds a new flag to make BPF ring buffer overwritable.
Perf ring buffers already implement an option to be overwritable. In order to
avoid data corruption, the data is written backward, see
commit 9ecda41acb97 ("perf/core: Add ::write_backward attribute to perf event").
This patch series re-uses the same idea from perf ring buffers but in BPF ring
buffers.
So, calling bpf_ringbuf_reserve() on an overwritable BPF ring buffer never
returns NULL.
As a consequence, oldest data will be overwritten by the newest so consumer will
loose data.

Overwritable ring buffers are useful in BPF programs that are permanently
enabled but rarely read, only on-demand, for example in case of a user request
to investigate problems. We would like to use this in the Traceloop project [1].

The self test added in this series was tested and validated in a VM:
you@vm# ./share/linux/tools/testing/selftests/bpf/test_progs -t ringbuf_over
Can't find bpf_testmod.ko kernel module: -2
WARNING! Selftests relying on bpf_testmod.ko will be skipped.
#135     ringbuf_over_writable:OK
Summary: 1/0 PASSED, 0 SKIPPED, 0 FAILED

You can also test the libbpf implementation by using the last patch of this
series which should be applied to iovisor/bcc:
you@home$ cd /path/to/iovisor/bcc
you@home$ git am -3 v2-0005-for-test-purpose-only-Add-toy-to-play-with-BPF-ri.patch
you@home$ cd /path/to/linux/tools/lib/bpf
you@home$ make -j$(nproc)
you@home$ cp libbpf.a /path/to/iovisor/bcc/libbpf-tools/.output
you@home$ cd /path/to/iovisor/bcc/libbpf-tools/
you@home$ make -j toy
# Start your VM and copy toy executable inside it.
root@vm-amd64:~# ./share/toy &
[1] 287
root@vm-amd64:~# for i in {1..16}; do ls > /dev/null; done
16
15
14
13
12
11
10
9
root@vm-amd64:~# ls > /dev/null && ls > /dev/null
18
17

As you can see, the first eight events are overwritten.

If you see any way to improve this contribution, feel free to share.

Changes since:
 v1:
  * Made producers write backward like perf ring buffer, so it permits avoiding
  memory corruption.
  * Added libbpf implementation to consume all events available.
  * Added selftest.
  * Added documentation.

Francis Laniel (5):
  bpf: Make ring buffer overwritable.
  selftests: Add BPF overwritable ring buffer self tests.
  docs/bpf: Add documentation for overwritable ring buffer.
  libbpf: Add implementation to consume overwritable BPF ring buffer.
  for test purpose only: Add toy to play with BPF ring.

 ...-only-Add-toy-to-play-with-BPF-ring-.patch | 147 ++++++++++++++++
 Documentation/bpf/ringbuf.rst                 |  18 +-
 include/uapi/linux/bpf.h                      |   3 +
 kernel/bpf/ringbuf.c                          |  43 +++--
 tools/include/uapi/linux/bpf.h                |   3 +
 tools/lib/bpf/ringbuf.c                       | 106 ++++++++++++
 tools/testing/selftests/bpf/Makefile          |   5 +-
 .../bpf/prog_tests/ringbuf_overwritable.c     | 158 ++++++++++++++++++
 .../bpf/progs/test_ringbuf_overwritable.c     |  61 +++++++
 9 files changed, 531 insertions(+), 13 deletions(-)
 create mode 100644 0001-for-test-purpose-only-Add-toy-to-play-with-BPF-ring-.patch
 create mode 100644 tools/testing/selftests/bpf/prog_tests/ringbuf_overwritable.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_ringbuf_overwritable.c


Best regards and thank you in advance.
---
[1] https://github.com/kinvolk/traceloop
Traceloop was presented at LPC 2020 (https://lpc.events/event/7/contributions/667/)
--
2.25.1

[RFC PATCH v2 1/5] bpf: Make ring buffer overwritable.

[Francis Laniel]

By default, BPF ring buffer are size bounded, when producers already filled the
buffer, they need to wait for the consumer to get those data before adding new
ones.
In terms of API, bpf_ringbuf_reserve() returns NULL if the buffer is full.

This patch permits making BPF ring buffer overwritable.
When producers already wrote as many data as the buffer size, they will begin to
over write existing data, so the oldest will be replaced.
As a result, bpf_ringbuf_reserve() never returns NULL.

To avoid memory consumption, this patch writes data backward like overwritable
perf ring buffer added in
commit 9ecda41acb97 ("perf/core: Add ::write_backward attribute to perf event").

Signed-off-by: Francis Laniel <flaniel@linux.microsoft.com>
---
 include/uapi/linux/bpf.h |  3 +++
 kernel/bpf/ringbuf.c     | 43 ++++++++++++++++++++++++++++++----------
 2 files changed, 36 insertions(+), 10 deletions(-)

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 59a217ca2dfd..c87a667649ab 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -1227,6 +1227,9 @@ enum {
 
 /* Create a map that is suitable to be an inner map with dynamic max entries */
 	BPF_F_INNER_MAP		= (1U << 12),
+
+/* Create an overwritable BPF_RINGBUF */
+	BFP_F_RB_OVERWRITABLE	= (1U << 13),
 };
 
 /* Flags for BPF_PROG_QUERY. */
diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
index ded4faeca192..369c61cfe8aa 100644
--- a/kernel/bpf/ringbuf.c
+++ b/kernel/bpf/ringbuf.c
@@ -12,7 +12,7 @@
 #include <uapi/linux/btf.h>
 #include <linux/btf_ids.h>
 
-#define RINGBUF_CREATE_FLAG_MASK (BPF_F_NUMA_NODE)
+#define RINGBUF_CREATE_FLAG_MASK (BPF_F_NUMA_NODE | BFP_F_RB_OVERWRITABLE)
 
 /* non-mmap()'able part of bpf_ringbuf (everything up to consumer page) */
 #define RINGBUF_PGOFF \
@@ -37,6 +37,8 @@ struct bpf_ringbuf {
 	u64 mask;
 	struct page **pages;
 	int nr_pages;
+	__u8 overwritable: 1,
+	     __reserved:    7;
 	spinlock_t spinlock ____cacheline_aligned_in_smp;
 	/* Consumer and producer counters are put into separate pages to allow
 	 * mapping consumer page as r/w, but restrict producer page to r/o.
@@ -127,7 +129,12 @@ static void bpf_ringbuf_notify(struct irq_work *work)
 	wake_up_all(&rb->waitq);
 }
 
-static struct bpf_ringbuf *bpf_ringbuf_alloc(size_t data_sz, int numa_node)
+static inline bool is_overwritable(struct bpf_ringbuf *rb)
+{
+	return !!rb->overwritable;
+}
+
+static struct bpf_ringbuf *bpf_ringbuf_alloc(size_t data_sz, int numa_node, __u32 flags)
 {
 	struct bpf_ringbuf *rb;
 
@@ -142,6 +149,7 @@ static struct bpf_ringbuf *bpf_ringbuf_alloc(size_t data_sz, int numa_node)
 	rb->mask = data_sz - 1;
 	rb->consumer_pos = 0;
 	rb->producer_pos = 0;
+	rb->overwritable = !!(flags & BFP_F_RB_OVERWRITABLE);
 
 	return rb;
 }
@@ -170,7 +178,7 @@ static struct bpf_map *ringbuf_map_alloc(union bpf_attr *attr)
 
 	bpf_map_init_from_attr(&rb_map->map, attr);
 
-	rb_map->rb = bpf_ringbuf_alloc(attr->max_entries, rb_map->map.numa_node);
+	rb_map->rb = bpf_ringbuf_alloc(attr->max_entries, rb_map->map.numa_node, attr->map_flags);
 	if (!rb_map->rb) {
 		kfree(rb_map);
 		return ERR_PTR(-ENOMEM);
@@ -248,6 +256,7 @@ static unsigned long ringbuf_avail_data_sz(struct bpf_ringbuf *rb)
 
 	cons_pos = smp_load_acquire(&rb->consumer_pos);
 	prod_pos = smp_load_acquire(&rb->producer_pos);
+
 	return prod_pos - cons_pos;
 }
 
@@ -325,14 +334,24 @@ static void *__bpf_ringbuf_reserve(struct bpf_ringbuf *rb, u64 size)
 	}
 
 	prod_pos = rb->producer_pos;
-	new_prod_pos = prod_pos + len;
 
-	/* check for out of ringbuf space by ensuring producer position
-	 * doesn't advance more than (ringbuf_size - 1) ahead
-	 */
-	if (new_prod_pos - cons_pos > rb->mask) {
-		spin_unlock_irqrestore(&rb->spinlock, flags);
-		return NULL;
+	if (!is_overwritable(rb)) {
+		new_prod_pos = prod_pos + len;
+
+		/* check for out of ringbuf space by ensuring producer position
+		 * doesn't advance more than (ringbuf_size - 1) ahead
+		 */
+		if (new_prod_pos - cons_pos > rb->mask) {
+			spin_unlock_irqrestore(&rb->spinlock, flags);
+			return NULL;
+		}
+	} else {
+		/*
+		 * With overwritable ring buffer we go from the end toward the
+		 * beginning.
+		 */
+		prod_pos -= len;
+		new_prod_pos = prod_pos;
 	}
 
 	hdr = (void *)rb->data + (prod_pos & rb->mask);
@@ -457,10 +476,14 @@ BPF_CALL_2(bpf_ringbuf_query, struct bpf_map *, map, u64, flags)
 
 	switch (flags) {
 	case BPF_RB_AVAIL_DATA:
+		if (is_overwritable(rb))
+			return -EINVAL;
 		return ringbuf_avail_data_sz(rb);
 	case BPF_RB_RING_SIZE:
 		return rb->mask + 1;
 	case BPF_RB_CONS_POS:
+		if (is_overwritable(rb))
+			return -EINVAL;
 		return smp_load_acquire(&rb->consumer_pos);
 	case BPF_RB_PROD_POS:
 		return smp_load_acquire(&rb->producer_pos);
-- 
2.25.1

[RFC PATCH v2 2/5] selftests: Add BPF overwritable ring buffer self tests.

[Francis Laniel]

Add tests to confirm behavior of overwritable BPF ring buffer, particularly the
oldest data being overwritten by newest ones.

Signed-off-by: Francis Laniel <flaniel@linux.microsoft.com>
---
 tools/testing/selftests/bpf/Makefile          |   5 +-
 .../bpf/prog_tests/ringbuf_overwritable.c     | 158 ++++++++++++++++++
 .../bpf/progs/test_ringbuf_overwritable.c     |  61 +++++++
 3 files changed, 222 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/ringbuf_overwritable.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_ringbuf_overwritable.c

diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
index 8d59ec7f4c2d..96e95dcfc982 100644
--- a/tools/testing/selftests/bpf/Makefile
+++ b/tools/testing/selftests/bpf/Makefile
@@ -351,8 +351,9 @@ LINKED_SKELS := test_static_linked.skel.h linked_funcs.skel.h		\
 		test_usdt.skel.h
 
 LSKELS := kfunc_call_test.c fentry_test.c fexit_test.c fexit_sleep.c \
-	test_ringbuf.c atomics.c trace_printk.c trace_vprintk.c \
-	map_ptr_kern.c core_kern.c core_kern_overflow.c
+	test_ringbuf.c test_ringbuf_overwritable.c atomics.c \
+	trace_printk.c trace_vprintk.c map_ptr_kern.c \
+	core_kern.c core_kern_overflow.c
 # Generate both light skeleton and libbpf skeleton for these
 LSKELS_EXTRA := test_ksyms_module.c test_ksyms_weak.c kfunc_call_test_subprog.c
 SKEL_BLACKLIST += $$(LSKELS)
diff --git a/tools/testing/selftests/bpf/prog_tests/ringbuf_overwritable.c b/tools/testing/selftests/bpf/prog_tests/ringbuf_overwritable.c
new file mode 100644
index 000000000000..b5ec1e62f761
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/ringbuf_overwritable.c
@@ -0,0 +1,158 @@
+// SPDX-License-Identifier: GPL-2.0
+#define _GNU_SOURCE
+#include <linux/compiler.h>
+#include <asm/barrier.h>
+#include <test_progs.h>
+#include <sys/mman.h>
+#include <sys/epoll.h>
+#include <time.h>
+#include <sched.h>
+#include <signal.h>
+#include <pthread.h>
+#include <sys/sysinfo.h>
+#include <linux/perf_event.h>
+#include <linux/ring_buffer.h>
+#include "test_ringbuf_overwritable.lskel.h"
+
+struct sample {
+	int count;
+	/*
+	 * filler size will be computed to have 8 samples in a 4096 bytes long
+	 * buffer.
+	 */
+	char filler[4096 / 8 - sizeof(int) - 8];
+};
+
+struct ring {
+	ring_buffer_sample_fn sample_cb;
+	__u8 overwritable: 1,
+	     __reserved:   7;
+	void *ctx;
+	void *data;
+	unsigned long *consumer_pos;
+	unsigned long *producer_pos;
+	unsigned long mask;
+	int map_fd;
+};
+
+struct ring_buffer {
+	struct epoll_event *events;
+	struct ring *rings;
+	size_t page_size;
+	int epoll_fd;
+	int ring_cnt;
+};
+
+static int duration;
+static struct test_ringbuf_overwritable_lskel *skel;
+
+void test_ringbuf_overwritable(void)
+{
+	const size_t rec_sz = BPF_RINGBUF_HDR_SZ + sizeof(struct sample);
+	int page_size = getpagesize();
+	int sample_cnt = 0, sample_read = 0;
+	unsigned long mask = page_size - 1;
+	struct ring_buffer *ringbuf;
+	int err, *len_ptr, len;
+	struct sample *sample;
+	long read_pos;
+	void *data_ptr;
+
+	skel = test_ringbuf_overwritable_lskel__open();
+	if (CHECK(!skel, "skel_open", "skeleton open failed\n"))
+		return;
+
+	skel->maps.ringbuf.max_entries = page_size;
+
+	err = test_ringbuf_overwritable_lskel__load(skel);
+	if (CHECK(err != 0, "skel_load", "skeleton load failed\n"))
+		goto cleanup;
+
+	/* only trigger BPF program for current process */
+	skel->bss->pid = getpid();
+
+	ringbuf = ring_buffer__new(skel->maps.ringbuf.map_fd, NULL, NULL, NULL);
+	if (CHECK(!ringbuf, "ringbuf_create", "failed to create ringbuf\n"))
+		goto cleanup;
+
+	/* There is only one ring in this ringbuf. */
+	data_ptr = ringbuf->rings[0].data;
+
+	err = test_ringbuf_overwritable_lskel__attach(skel);
+	if (CHECK(err, "skel_attach", "skeleton attachment failed: %d\n", err))
+		goto cleanup;
+
+	/* Trigger one sample. */
+	syscall(__NR_getpgid);
+	sample_cnt++;
+
+	CHECK(skel->bss->avail_data != -EINVAL,
+	      "err_avail_size", "exp %d, got %ld\n",
+	      -EINVAL, skel->bss->avail_data);
+	CHECK(skel->bss->ring_size != page_size,
+	      "err_ring_size", "exp %ld, got %ld\n",
+	      (long)page_size, skel->bss->ring_size);
+	CHECK(skel->bss->cons_pos != -EINVAL,
+	      "err_cons_pos", "exp %d, got %ld\n",
+	      -EINVAL, skel->bss->cons_pos);
+	CHECK(skel->bss->prod_pos != sample_cnt * -rec_sz,
+	      "err_prod_pos", "exp %ld, got %ld\n",
+	      sample_cnt * -rec_sz, skel->bss->prod_pos);
+
+	len_ptr = data_ptr + (skel->bss->prod_pos & mask);
+	len = smp_load_acquire(len_ptr);
+
+	CHECK(len != sizeof(struct sample),
+	      "err_sample_len", "exp %ld, got %d\n",
+	      sizeof(struct sample), len);
+
+	sample = (void *)len_ptr + BPF_RINGBUF_HDR_SZ;
+
+	CHECK(sample->count != sample_cnt,
+	      "err_sample_cnt", "exp %d, got %d",
+	      sample_cnt, sample->count);
+
+	/* Trigger many samples, so we overwrite data */
+	for (int i = 0; i < 16; i++) {
+		syscall(__NR_getpgid);
+		sample_cnt++;
+	}
+
+	CHECK(skel->bss->avail_data != -EINVAL,
+	      "err_avail_size", "exp %d, got %ld\n",
+	      -EINVAL, skel->bss->avail_data);
+	CHECK(skel->bss->ring_size != page_size,
+	      "err_ring_size", "exp %ld, got %ld\n",
+	      (long)page_size, skel->bss->ring_size);
+	CHECK(skel->bss->cons_pos != -EINVAL,
+	      "err_cons_pos", "exp %d, got %ld\n",
+	      -EINVAL, skel->bss->cons_pos);
+	CHECK(skel->bss->prod_pos != sample_cnt * -rec_sz,
+	      "err_prod_pos", "exp %ld, got %ld\n",
+	      sample_cnt * -rec_sz, skel->bss->prod_pos);
+
+	read_pos = skel->bss->prod_pos;
+	sample_read = 0;
+	while (read_pos - skel->bss->prod_pos < mask) {
+		len_ptr = data_ptr + (read_pos & mask);
+		len = smp_load_acquire(len_ptr);
+
+		sample = (void *)len_ptr + BPF_RINGBUF_HDR_SZ;
+
+		CHECK(sample->count != sample_cnt - sample_read,
+		      "err_sample_cnt", "exp %d, got %d",
+		      sample_cnt - sample_read, sample->count);
+
+		sample_read++;
+		read_pos += round_up(len + BPF_RINGBUF_HDR_SZ, 8);
+	}
+
+	CHECK(sample_read != page_size / rec_sz,
+	      "err_sample_read", "exp %ld, got %d",
+	      page_size / rec_sz, sample_read);
+
+	test_ringbuf_overwritable_lskel__detach(skel);
+cleanup:
+	ring_buffer__free(ringbuf);
+	test_ringbuf_overwritable_lskel__destroy(skel);
+}
diff --git a/tools/testing/selftests/bpf/progs/test_ringbuf_overwritable.c b/tools/testing/selftests/bpf/progs/test_ringbuf_overwritable.c
new file mode 100644
index 000000000000..e28be35059b7
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/test_ringbuf_overwritable.c
@@ -0,0 +1,61 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (c) 2020 Facebook
+
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include "bpf_misc.h"
+
+char _license[] SEC("license") = "GPL";
+
+struct sample {
+	int count;
+	/*
+	 * filler size will be computed to have 8 samples in a 4096 bytes long
+	 * buffer.
+	 */
+	char filler[4096 / 8 - sizeof(int) - BPF_RINGBUF_HDR_SZ];
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_RINGBUF);
+	__uint(map_flags, BFP_F_RB_OVERWRITABLE);
+} ringbuf SEC(".maps");
+
+/* inputs */
+int pid = 0;
+
+/* outputs */
+long avail_data = 0;
+long ring_size = 0;
+long cons_pos = 0;
+long prod_pos = 0;
+
+static int count;
+
+SEC("fentry/" SYS_PREFIX "sys_getpgid")
+int test_ringbuf_overwritable(void *ctx)
+{
+	int cur_pid = bpf_get_current_pid_tgid() >> 32;
+	struct sample *sample;
+
+	if (cur_pid != pid)
+		return 0;
+
+	sample = bpf_ringbuf_reserve(&ringbuf, sizeof(*sample), 0);
+	if (!sample)
+		return 0;
+
+	__sync_fetch_and_add(&count, 1);
+	sample->count = count;
+
+	bpf_printk("count: %d\n", count);
+
+	bpf_ringbuf_submit(sample, 0);
+
+	avail_data = bpf_ringbuf_query(&ringbuf, BPF_RB_AVAIL_DATA);
+	ring_size = bpf_ringbuf_query(&ringbuf, BPF_RB_RING_SIZE);
+	cons_pos = bpf_ringbuf_query(&ringbuf, BPF_RB_CONS_POS);
+	prod_pos = bpf_ringbuf_query(&ringbuf, BPF_RB_PROD_POS);
+
+	return 0;
+}
-- 
2.25.1

[RFC PATCH v2 3/5] docs/bpf: Add documentation for overwritable ring buffer.

[Francis Laniel]

Add documentation to precise behavior of overwritable BPF ring buffer compared
to conventionnal ones.

Signed-off-by: Francis Laniel <flaniel@linux.microsoft.com>
---
 Documentation/bpf/ringbuf.rst | 18 +++++++++++++++++-
 1 file changed, 17 insertions(+), 1 deletion(-)

diff --git a/Documentation/bpf/ringbuf.rst b/Documentation/bpf/ringbuf.rst
index 6a615cd62bda..e062381ff604 100644
--- a/Documentation/bpf/ringbuf.rst
+++ b/Documentation/bpf/ringbuf.rst
@@ -124,7 +124,7 @@ buffer.  Currently 4 are supported:
 
 - ``BPF_RB_AVAIL_DATA`` returns amount of unconsumed data in ring buffer;
 - ``BPF_RB_RING_SIZE`` returns the size of ring buffer;
-- ``BPF_RB_CONS_POS``/``BPF_RB_PROD_POS`` returns current logical possition
+- ``BPF_RB_CONS_POS``/``BPF_RB_PROD_POS`` returns current logical position
   of consumer/producer, respectively.
 
 Returned values are momentarily snapshots of ring buffer state and could be
@@ -204,3 +204,19 @@ buffer. For extreme cases, when BPF program wants more manual control of
 notifications, commit/discard/output helpers accept ``BPF_RB_NO_WAKEUP`` and
 ``BPF_RB_FORCE_WAKEUP`` flags, which give full control over notifications of
 data availability, but require extra caution and diligence in using this API.
+
+Specific case of overwritable ring buffer
+-----------------------------------------
+
+Using ``BFP_F_RB_OVERWRITABLE`` when creating the ring buffer will make it
+overwritable.
+As a consequence, the producers will never be stopped from writing data, *i.e.*
+in this mode ``bpf_ringbuf_reserve()`` never blocks and returns NULL, but oldest
+events will be replaced by newest ones.
+
+In terms of implementation, this feature uses the same logic than overwritable
+perf ring buffer.
+The ring buffer is written backward, while it should be read forward from the
+producer position.
+As a consequence, in this mode, the consumer position has no meaning and can be
+used freely by userspace implementation.
-- 
2.25.1

[RFC PATCH v2 4/5] libbpf: Add implementation to consume overwritable BPF ring buffer.

[Francis Laniel]

If the BPF ring buffer is overwritable, ringbuf_process_overwritable_ring() will
be called to handle the data consumption.
All the available data will be consumed but some checks will be performed:
* check we do not read data we already read, if there is no new data, nothing
happens.
* check we do not read more than the buffer size.
* check we do not read invalid data by checking they fit the buffer size.

Signed-off-by: Francis Laniel <flaniel@linux.microsoft.com>
---
 tools/include/uapi/linux/bpf.h |   3 +
 tools/lib/bpf/ringbuf.c        | 106 +++++++++++++++++++++++++++++++++
 2 files changed, 109 insertions(+)

diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index 59a217ca2dfd..cd73a89e8ead 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -1227,6 +1227,9 @@ enum {
 
 /* Create a map that is suitable to be an inner map with dynamic max entries */
 	BPF_F_INNER_MAP		= (1U << 12),
+
+/* Create an over writable BPF_RINGBUF */
+	BFP_F_RB_OVERWRITABLE	= (1U << 13),
 };
 
 /* Flags for BPF_PROG_QUERY. */
diff --git a/tools/lib/bpf/ringbuf.c b/tools/lib/bpf/ringbuf.c
index 8bc117bcc7bc..2362a6280fc5 100644
--- a/tools/lib/bpf/ringbuf.c
+++ b/tools/lib/bpf/ringbuf.c
@@ -23,6 +23,8 @@
 
 struct ring {
 	ring_buffer_sample_fn sample_cb;
+	__u8 overwritable: 1,
+	     __reserved:   7;
 	void *ctx;
 	void *data;
 	unsigned long *consumer_pos;
@@ -51,6 +53,11 @@ static void ringbuf_unmap_ring(struct ring_buffer *rb, struct ring *r)
 	}
 }
 
+static inline bool is_overwritable(struct ring *r)
+{
+	return !!r->overwritable;
+}
+
 /* Add extra RINGBUF maps to this ring buffer manager */
 int ring_buffer__add(struct ring_buffer *rb, int map_fd,
 		     ring_buffer_sample_fn sample_cb, void *ctx)
@@ -95,6 +102,7 @@ int ring_buffer__add(struct ring_buffer *rb, int map_fd,
 	r->sample_cb = sample_cb;
 	r->ctx = ctx;
 	r->mask = info.max_entries - 1;
+	r->overwritable = !!(info.map_flags & BFP_F_RB_OVERWRITABLE);
 
 	/* Map writable consumer page */
 	tmp = mmap(NULL, rb->page_size, PROT_READ | PROT_WRITE, MAP_SHARED,
@@ -202,6 +210,101 @@ static inline int roundup_len(__u32 len)
 	return (len + 7) / 8 * 8;
 }
 
+
+static int64_t ringbuf_process_overwritable_ring(struct ring *r)
+{
+	/* 64-bit to avoid overflow in case of extreme application behavior */
+	int64_t cnt = 0;
+	unsigned long read_pos, prod_pos, previous_prod_pos;
+
+	prod_pos = smp_load_acquire(r->producer_pos);
+	previous_prod_pos = smp_load_acquire(r->consumer_pos);
+
+	/*
+	 * For overwritable ring buffer, we use consumer_pos as the previous
+	 * producer_pos.
+	 * So, if between two calls to this function, the prod_pos did not move,
+	 * it means there is no new data, so we can return right now rather than
+	 * dealing with data we already proceeded.
+	 * NOTE the kernel space does not care about consumer_pos to reserve()
+	 * in overwritable ring buffers, hence we can hijack this field.
+	 */
+	if (previous_prod_pos == prod_pos)
+		return 0;
+
+	/*
+	 * BPF ring buffer is over writable, we start reading from
+	 * producer position.
+	 */
+	read_pos = prod_pos;
+	while (read_pos - prod_pos < r->mask) {
+		int *len_ptr, len;
+
+		len_ptr = r->data + (read_pos & r->mask);
+		len = smp_load_acquire(len_ptr);
+
+		/* sample not committed yet, bail out for now */
+		if (len & BPF_RINGBUF_BUSY_BIT)
+			break;
+
+		/*
+		 * If len is 0, it means we read all the data
+		 * available in the buffer and jump on 0 data:
+		 *
+		 * prod_pos                         read_pos
+		 *     |                                |
+		 *     V                                V
+		 * +---+------+----------+-------+------+
+		 * |   |D....D|C........C|B.....B|A....A|
+		 * +---+------+----------+-------+------+
+		 */
+		if (!len)
+			break;
+
+		/*
+		 * If adding the event len to the current
+		 * consumer position makes us wrap the buffer,
+		 * it means we already did "one loop" around the
+		 * buffer.
+		 * So, the pointed data would not be usable:
+		 *
+		 *                               prod_pos
+		 *                   read_pos----+   |
+		 *                               |   |
+		 *                               V   V
+		 * +---+------+----------+-------+---+--+
+		 * |..E|D....D|C........C|B.....B|A..|E.|
+		 * +---+------+----------+-------+---+--+
+		 */
+		if (read_pos - prod_pos + len > r->mask)
+			break;
+
+		read_pos += roundup_len(len);
+
+		if ((len & BPF_RINGBUF_DISCARD_BIT) == 0) {
+			void *sample;
+			int err;
+
+			sample = (void *)len_ptr + BPF_RINGBUF_HDR_SZ;
+			err = r->sample_cb(r->ctx, sample, len);
+			if (err < 0) {
+				/* update consumer pos and bail out */
+				smp_store_release(r->consumer_pos,
+						  prod_pos);
+				return err;
+			}
+			cnt++;
+		}
+
+		/* This prevents reading data we already processed. */
+		if (previous_prod_pos && read_pos >= previous_prod_pos)
+			break;
+	}
+
+	smp_store_release(r->consumer_pos, prod_pos);
+	return cnt;
+}
+
 static int64_t ringbuf_process_ring(struct ring* r)
 {
 	int *len_ptr, len, err;
@@ -211,6 +314,9 @@ static int64_t ringbuf_process_ring(struct ring* r)
 	bool got_new_data;
 	void *sample;
 
+	if (is_overwritable(r))
+		return ringbuf_process_overwritable_ring(r);
+
 	cons_pos = smp_load_acquire(r->consumer_pos);
 	do {
 		got_new_data = false;
-- 
2.25.1

[RFC PATCH v2 5/5] for test purpose only: Add toy to play with BPF ring.

[Francis Laniel]

This patch should be applied on iovisor/bcc.

Signed-off-by: Francis Laniel <flaniel@linux.microsoft.com>
---
 ...-only-Add-toy-to-play-with-BPF-ring-.patch | 147 ++++++++++++++++++
 1 file changed, 147 insertions(+)
 create mode 100644 0001-for-test-purpose-only-Add-toy-to-play-with-BPF-ring-.patch

diff --git a/0001-for-test-purpose-only-Add-toy-to-play-with-BPF-ring-.patch b/0001-for-test-purpose-only-Add-toy-to-play-with-BPF-ring-.patch
new file mode 100644
index 000000000000..37d08cc08a88
--- /dev/null
+++ b/0001-for-test-purpose-only-Add-toy-to-play-with-BPF-ring-.patch
@@ -0,0 +1,147 @@
+From e4b95b1f9625f62d0978173973070dce38bd7210 Mon Sep 17 00:00:00 2001
+From: Francis Laniel <flaniel@linux.microsoft.com>
+Date: Tue, 9 Aug 2022 18:18:53 +0200
+Subject: [PATCH] for test purpose only: Add toy to play with BPF ring buffer.
+
+Signed-off-by: Francis Laniel <flaniel@linux.microsoft.com>
+---
+ libbpf-tools/Makefile  |  1 +
+ libbpf-tools/toy.bpf.c | 29 +++++++++++++++++++
+ libbpf-tools/toy.c     | 65 ++++++++++++++++++++++++++++++++++++++++++
+ libbpf-tools/toy.h     |  4 +++
+ 4 files changed, 99 insertions(+)
+ create mode 100644 libbpf-tools/toy.bpf.c
+ create mode 100644 libbpf-tools/toy.c
+ create mode 100644 libbpf-tools/toy.h
+
+diff --git a/libbpf-tools/Makefile b/libbpf-tools/Makefile
+index 3e40f6e5..0d81d3b7 100644
+--- a/libbpf-tools/Makefile
++++ b/libbpf-tools/Makefile
+@@ -68,6 +68,7 @@ APPS = \
+ 	tcplife \
+ 	tcprtt \
+ 	tcpsynbl \
++	toy \
+ 	vfsstat \
+ 	#
+ 
+diff --git a/libbpf-tools/toy.bpf.c b/libbpf-tools/toy.bpf.c
+new file mode 100644
+index 00000000..3c28a20b
+--- /dev/null
++++ b/libbpf-tools/toy.bpf.c
+@@ -0,0 +1,29 @@
++#include <linux/types.h>
++#include <bpf/bpf_helpers.h>
++#include <linux/bpf.h>
++#include "toy.h"
++
++
++struct {
++	__uint(type, BPF_MAP_TYPE_RINGBUF);
++	__uint(max_entries, 4096);
++	__uint(map_flags, 1U << 13);
++} buffer SEC(".maps");
++
++static __u32 count = 0;
++
++SEC("tracepoint/syscalls/sys_enter_execve")
++int sys_enter_execve(void) {
++	count++;
++	struct event *event = bpf_ringbuf_reserve(&buffer, sizeof(struct event), 0);
++	if (!event) {
++		return 1;
++	}
++
++	event->count = count;
++	bpf_ringbuf_submit(event, 0);
++
++	return 0;
++}
++
++char _license[] SEC("license") = "GPL";
+diff --git a/libbpf-tools/toy.c b/libbpf-tools/toy.c
+new file mode 100644
+index 00000000..4cd8b588
+--- /dev/null
++++ b/libbpf-tools/toy.c
+@@ -0,0 +1,65 @@
++#include <bpf/libbpf.h>
++#include <stdio.h>
++#include <unistd.h>
++#include "toy.h"
++#include "toy.skel.h"
++#include "btf_helpers.h"
++
++
++static int buf_process_sample(void *ctx, void *data, size_t len) {
++	struct event *evt = (struct event *)data;
++
++	printf("%d\n", evt->count);
++
++	return 0;
++}
++
++int main(void) {
++	LIBBPF_OPTS(bpf_object_open_opts, open_opts);
++	int buffer_map_fd = -1;
++	struct toy_bpf *obj;
++	int err;
++
++	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
++
++	err = ensure_core_btf(&open_opts);
++	if (err) {
++		fprintf(stderr, "failed to fetch necessary BTF for CO-RE: %s\n", strerror(-err));
++		return 1;
++	}
++
++	obj = toy_bpf__open_opts(&open_opts);
++	if (!obj) {
++		fprintf(stderr, "failed to open BPF object\n");
++		return 1;
++	}
++
++	err = toy_bpf__load(obj);
++	if (err) {
++		fprintf(stderr, "failed to load BPF object: %d\n", err);
++		return 1;
++	}
++
++	struct ring_buffer *ring_buffer;
++
++	buffer_map_fd = bpf_object__find_map_fd_by_name(obj->obj, "buffer");
++	ring_buffer = ring_buffer__new(buffer_map_fd, buf_process_sample, NULL, NULL);
++
++	if(!ring_buffer) {
++		fprintf(stderr, "failed to create ring buffer\n");
++		return 1;
++	}
++
++	err = toy_bpf__attach(obj);
++	if (err) {
++		fprintf(stderr, "failed to attach BPF programs\n");
++		return 1;
++	}
++
++	for (;;) {
++		ring_buffer__consume(ring_buffer);
++		sleep(1);
++	}
++
++	return 0;
++}
+diff --git a/libbpf-tools/toy.h b/libbpf-tools/toy.h
+new file mode 100644
+index 00000000..ebfedf06
+--- /dev/null
++++ b/libbpf-tools/toy.h
+@@ -0,0 +1,4 @@
++struct event {
++	__u32 count;
++	char filler[4096 / 8 - sizeof(__u32) - 8];
++};
+-- 
+2.25.1
+
-- 
2.25.1

Re: [RFC PATCH v2 0/5] Make BPF ring buffer overwritable

[Andrii Nakryiko]

On Tue, Sep 6, 2022 at 12:58 PM Francis Laniel
<flaniel@linux.microsoft.com> wrote:
>
> Hi.
>
>
> First, I hope you are fine and the same for your relatives.
>
> Normally, when BPF ring buffer are full, producers cannot write anymore and
> need to wait for consumer to get some data.
> As a consequence, calling bpf_ringbuf_reserve() from eBPF code returns NULL.
>
> This contribution adds a new flag to make BPF ring buffer overwritable.
> Perf ring buffers already implement an option to be overwritable. In order to
> avoid data corruption, the data is written backward, see

No, you are not avoiding data corruption. This patch set doesn't apply
cleanly, so I can't try this locally, but try the following example:

1. Allocate very small ringbuf (4KB).
2. Write BPF program that does multiple reservations within single
run. Enough reservations to overfill entire ringbuf and wrap over. For
each reserved sample fill it completely with garbage.
3. Try to submit all (successful) reservations. My theory is you'll
observe a crash.

Make sure that sample size isn't an exact multiple of ringbuf size so
that you don't have a nice neat overlap.

Look at bpf_ringbuf_restore_from_rec(). Each successful reservation
contains a very delicate and precise offset that allows
bpf_ringbuf_commit() to find the BPF ringbuf map in memory. By
allowing to overwrite active not-yet-submitted reservation you allow
to corrupt this information in the record header. This will make
kernel code to dereference garbage addresses.

This whole backwards approach won't work with BPF ringbuf. It works
with BPF perfbuf only under some non-enforceable assumptions, from
what I understand. We need something else for an overwritable ringbuf.


> commit 9ecda41acb97 ("perf/core: Add ::write_backward attribute to perf event").
> This patch series re-uses the same idea from perf ring buffers but in BPF ring
> buffers.
> So, calling bpf_ringbuf_reserve() on an overwritable BPF ring buffer never
> returns NULL.
> As a consequence, oldest data will be overwritten by the newest so consumer will
> loose data.
>
> Overwritable ring buffers are useful in BPF programs that are permanently
> enabled but rarely read, only on-demand, for example in case of a user request
> to investigate problems. We would like to use this in the Traceloop project [1].
>
> The self test added in this series was tested and validated in a VM:
> you@vm# ./share/linux/tools/testing/selftests/bpf/test_progs -t ringbuf_over
> Can't find bpf_testmod.ko kernel module: -2
> WARNING! Selftests relying on bpf_testmod.ko will be skipped.
> #135     ringbuf_over_writable:OK
> Summary: 1/0 PASSED, 0 SKIPPED, 0 FAILED
>
> You can also test the libbpf implementation by using the last patch of this
> series which should be applied to iovisor/bcc:
> you@home$ cd /path/to/iovisor/bcc
> you@home$ git am -3 v2-0005-for-test-purpose-only-Add-toy-to-play-with-BPF-ri.patch
> you@home$ cd /path/to/linux/tools/lib/bpf
> you@home$ make -j$(nproc)
> you@home$ cp libbpf.a /path/to/iovisor/bcc/libbpf-tools/.output
> you@home$ cd /path/to/iovisor/bcc/libbpf-tools/
> you@home$ make -j toy
> # Start your VM and copy toy executable inside it.
> root@vm-amd64:~# ./share/toy &
> [1] 287
> root@vm-amd64:~# for i in {1..16}; do ls > /dev/null; done
> 16
> 15
> 14
> 13
> 12
> 11
> 10
> 9
> root@vm-amd64:~# ls > /dev/null && ls > /dev/null
> 18
> 17
>
> As you can see, the first eight events are overwritten.
>
> If you see any way to improve this contribution, feel free to share.
>
> Changes since:
>  v1:
>   * Made producers write backward like perf ring buffer, so it permits avoiding
>   memory corruption.
>   * Added libbpf implementation to consume all events available.
>   * Added selftest.
>   * Added documentation.
>
> Francis Laniel (5):
>   bpf: Make ring buffer overwritable.
>   selftests: Add BPF overwritable ring buffer self tests.
>   docs/bpf: Add documentation for overwritable ring buffer.
>   libbpf: Add implementation to consume overwritable BPF ring buffer.
>   for test purpose only: Add toy to play with BPF ring.
>
>  ...-only-Add-toy-to-play-with-BPF-ring-.patch | 147 ++++++++++++++++
>  Documentation/bpf/ringbuf.rst                 |  18 +-
>  include/uapi/linux/bpf.h                      |   3 +
>  kernel/bpf/ringbuf.c                          |  43 +++--
>  tools/include/uapi/linux/bpf.h                |   3 +
>  tools/lib/bpf/ringbuf.c                       | 106 ++++++++++++
>  tools/testing/selftests/bpf/Makefile          |   5 +-
>  .../bpf/prog_tests/ringbuf_overwritable.c     | 158 ++++++++++++++++++
>  .../bpf/progs/test_ringbuf_overwritable.c     |  61 +++++++
>  9 files changed, 531 insertions(+), 13 deletions(-)
>  create mode 100644 0001-for-test-purpose-only-Add-toy-to-play-with-BPF-ring-.patch
>  create mode 100644 tools/testing/selftests/bpf/prog_tests/ringbuf_overwritable.c
>  create mode 100644 tools/testing/selftests/bpf/progs/test_ringbuf_overwritable.c
>
>
> Best regards and thank you in advance.
> ---
> [1] https://github.com/kinvolk/traceloop
> Traceloop was presented at LPC 2020 (https://lpc.events/event/7/contributions/667/)
> --
> 2.25.1
>