[PATCH bpf-next v5 0/2] add batched ops for percpu array

[PATCH bpf-next v5 0/2] add batched ops for percpu array

[Pedro Tammela]

This patchset introduces batched operations for the per-cpu variant of
the array map.

Also updates the batch ops test for arrays.

v4 -> v5:
- Revert removal of percpu macros

v3 -> v4:
- Prefer 'calloc()' over 'malloc()' on batch ops tests
- Add missing static keyword in a couple of test functions
- 'offset' to 'cpu_offset' as suggested by Martin

v2 -> v3:
- Remove percpu macros as suggested by Andrii
- Update tests that used the per cpu macros

v1 -> v2:
- Amended a more descriptive commit message

Pedro Tammela (2):
  bpf: add batched ops support for percpu array
  bpf: selftests: update array map tests for per-cpu batched ops

 kernel/bpf/arraymap.c                         |   2 +
 .../bpf/map_tests/array_map_batch_ops.c       | 104 +++++++++++++-----
 2 files changed, 77 insertions(+), 29 deletions(-)

-- 
2.25.1

[PATCH bpf-next v5 1/2] bpf: add batched ops support for percpu array

[Pedro Tammela]

Uses the already in-place infrastructure provided by the
'generic_map_*_batch' functions.

No tweak was needed as it transparently handles the percpu variant.

As arrays don't have delete operations, let it return a error to
user space (default behaviour).

Suggested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
---
 kernel/bpf/arraymap.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c
index 463d25e1e67e..3c4105603f9d 100644
--- a/kernel/bpf/arraymap.c
+++ b/kernel/bpf/arraymap.c
@@ -698,6 +698,8 @@ const struct bpf_map_ops percpu_array_map_ops = {
 	.map_delete_elem = array_map_delete_elem,
 	.map_seq_show_elem = percpu_array_map_seq_show_elem,
 	.map_check_btf = array_map_check_btf,
+	.map_lookup_batch = generic_map_lookup_batch,
+	.map_update_batch = generic_map_update_batch,
 	.map_set_for_each_callback_args = map_set_for_each_callback_args,
 	.map_for_each_callback = bpf_for_each_array_elem,
 	.map_btf_name = "bpf_array",
-- 
2.25.1

[PATCH bpf-next v5 2/2] bpf: selftests: update array map tests for per-cpu batched ops

[Pedro Tammela]

Follows the same logic as the hashtable tests.

Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
---
 .../bpf/map_tests/array_map_batch_ops.c       | 104 +++++++++++++-----
 1 file changed, 75 insertions(+), 29 deletions(-)

diff --git a/tools/testing/selftests/bpf/map_tests/array_map_batch_ops.c b/tools/testing/selftests/bpf/map_tests/array_map_batch_ops.c
index e42ea1195d18..f4d870da7684 100644
--- a/tools/testing/selftests/bpf/map_tests/array_map_batch_ops.c
+++ b/tools/testing/selftests/bpf/map_tests/array_map_batch_ops.c
@@ -9,10 +9,13 @@
 
 #include <test_maps.h>
 
+static int nr_cpus;
+
 static void map_batch_update(int map_fd, __u32 max_entries, int *keys,
-			     int *values)
+			     __s64 *values, bool is_pcpu)
 {
-	int i, err;
+	int i, j, err;
+	int cpu_offset = 0;
 	DECLARE_LIBBPF_OPTS(bpf_map_batch_opts, opts,
 		.elem_flags = 0,
 		.flags = 0,
@@ -20,22 +23,41 @@ static void map_batch_update(int map_fd, __u32 max_entries, int *keys,
 
 	for (i = 0; i < max_entries; i++) {
 		keys[i] = i;
-		values[i] = i + 1;
+		if (is_pcpu) {
+			cpu_offset = i * nr_cpus;
+			for (j = 0; j < nr_cpus; j++)
+				(values + cpu_offset)[j] = i + 1 + j;
+		} else {
+			values[i] = i + 1;
+		}
 	}
 
 	err = bpf_map_update_batch(map_fd, keys, values, &max_entries, &opts);
 	CHECK(err, "bpf_map_update_batch()", "error:%s\n", strerror(errno));
 }
 
-static void map_batch_verify(int *visited, __u32 max_entries,
-			     int *keys, int *values)
+static void map_batch_verify(int *visited, __u32 max_entries, int *keys,
+			     __s64 *values, bool is_pcpu)
 {
-	int i;
+	int i, j;
+	int cpu_offset = 0;
 
 	memset(visited, 0, max_entries * sizeof(*visited));
 	for (i = 0; i < max_entries; i++) {
-		CHECK(keys[i] + 1 != values[i], "key/value checking",
-		      "error: i %d key %d value %d\n", i, keys[i], values[i]);
+		if (is_pcpu) {
+			cpu_offset = i * nr_cpus;
+			for (j = 0; j < nr_cpus; j++) {
+				__s64 value = (values + cpu_offset)[j];
+				CHECK(keys[i] + j + 1 != value,
+				      "key/value checking",
+				      "error: i %d j %d key %d value %lld\n", i,
+				      j, keys[i], value);
+			}
+		} else {
+			CHECK(keys[i] + 1 != values[i], "key/value checking",
+			      "error: i %d key %d value %lld\n", i, keys[i],
+			      values[i]);
+		}
 		visited[i] = 1;
 	}
 	for (i = 0; i < max_entries; i++) {
@@ -44,19 +66,21 @@ static void map_batch_verify(int *visited, __u32 max_entries,
 	}
 }
 
-void test_array_map_batch_ops(void)
+static void __test_map_lookup_and_update_batch(bool is_pcpu)
 {
 	struct bpf_create_map_attr xattr = {
 		.name = "array_map",
-		.map_type = BPF_MAP_TYPE_ARRAY,
+		.map_type = is_pcpu ? BPF_MAP_TYPE_PERCPU_ARRAY :
+				      BPF_MAP_TYPE_ARRAY,
 		.key_size = sizeof(int),
-		.value_size = sizeof(int),
+		.value_size = sizeof(__s64),
 	};
-	int map_fd, *keys, *values, *visited;
+	int map_fd, *keys, *visited;
 	__u32 count, total, total_success;
 	const __u32 max_entries = 10;
 	__u64 batch = 0;
-	int err, step;
+	int err, step, value_size;
+	void *values;
 	DECLARE_LIBBPF_OPTS(bpf_map_batch_opts, opts,
 		.elem_flags = 0,
 		.flags = 0,
@@ -67,22 +91,23 @@ void test_array_map_batch_ops(void)
 	CHECK(map_fd == -1,
 	      "bpf_create_map_xattr()", "error:%s\n", strerror(errno));
 
-	keys = malloc(max_entries * sizeof(int));
-	values = malloc(max_entries * sizeof(int));
-	visited = malloc(max_entries * sizeof(int));
+	value_size = sizeof(__s64);
+	if (is_pcpu)
+		value_size *= nr_cpus;
+
+	keys = calloc(max_entries, sizeof(*keys));
+	values = calloc(max_entries, value_size);
+	visited = calloc(max_entries, sizeof(*visited));
 	CHECK(!keys || !values || !visited, "malloc()", "error:%s\n",
 	      strerror(errno));
 
-	/* populate elements to the map */
-	map_batch_update(map_fd, max_entries, keys, values);
-
 	/* test 1: lookup in a loop with various steps. */
 	total_success = 0;
 	for (step = 1; step < max_entries; step++) {
-		map_batch_update(map_fd, max_entries, keys, values);
-		map_batch_verify(visited, max_entries, keys, values);
+		map_batch_update(map_fd, max_entries, keys, values, is_pcpu);
+		map_batch_verify(visited, max_entries, keys, values, is_pcpu);
 		memset(keys, 0, max_entries * sizeof(*keys));
-		memset(values, 0, max_entries * sizeof(*values));
+		memset(values, 0, max_entries * value_size);
 		batch = 0;
 		total = 0;
 		/* iteratively lookup/delete elements with 'step'
@@ -91,10 +116,10 @@ void test_array_map_batch_ops(void)
 		count = step;
 		while (true) {
 			err = bpf_map_lookup_batch(map_fd,
-						total ? &batch : NULL, &batch,
-						keys + total,
-						values + total,
-						&count, &opts);
+						   total ? &batch : NULL,
+						   &batch, keys + total,
+						   values + total * value_size,
+						   &count, &opts);
 
 			CHECK((err && errno != ENOENT), "lookup with steps",
 			      "error: %s\n", strerror(errno));
@@ -108,7 +133,7 @@ void test_array_map_batch_ops(void)
 		CHECK(total != max_entries, "lookup with steps",
 		      "total = %u, max_entries = %u\n", total, max_entries);
 
-		map_batch_verify(visited, max_entries, keys, values);
+		map_batch_verify(visited, max_entries, keys, values, is_pcpu);
 
 		total_success++;
 	}
@@ -116,9 +141,30 @@ void test_array_map_batch_ops(void)
 	CHECK(total_success == 0, "check total_success",
 	      "unexpected failure\n");
 
-	printf("%s:PASS\n", __func__);
-
 	free(keys);
 	free(values);
 	free(visited);
 }
+
+static void array_map_batch_ops(void)
+{
+	__test_map_lookup_and_update_batch(false);
+	printf("test_%s:PASS\n", __func__);
+}
+
+static void array_percpu_map_batch_ops(void)
+{
+	__test_map_lookup_and_update_batch(true);
+	printf("test_%s:PASS\n", __func__);
+}
+
+void test_array_map_batch_ops(void)
+{
+	nr_cpus = libbpf_num_possible_cpus();
+
+	CHECK(nr_cpus < 0, "nr_cpus checking",
+	      "error: get possible cpus failed");
+
+	array_map_batch_ops();
+	array_percpu_map_batch_ops();
+}
-- 
2.25.1

Re: [PATCH bpf-next v5 0/2] add batched ops for percpu array

[patchwork-bot+netdevbpf]

Hello:

This series was applied to bpf/bpf-next.git (refs/heads/master):

On Sat, 24 Apr 2021 18:45:08 -0300 you wrote:
> This patchset introduces batched operations for the per-cpu variant of
> the array map.
> 
> Also updates the batch ops test for arrays.
> 
> v4 -> v5:
> - Revert removal of percpu macros
> 
> [...]

Here is the summary with links:
  - [bpf-next,v5,1/2] bpf: add batched ops support for percpu array
    https://git.kernel.org/bpf/bpf-next/c/f008d732ab18
  - [bpf-next,v5,2/2] bpf: selftests: update array map tests for per-cpu batched ops
    https://git.kernel.org/bpf/bpf-next/c/3733bfbbdd28

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html