[PATCH bpf-next v6 1/2] bpf: Extend batch operations for map-in-map bpf-maps

[PATCH bpf-next v6 1/2] bpf: Extend batch operations for map-in-map bpf-maps

[Takshak Chahande]

This patch extends batch operations support for map-in-map map-types:
BPF_MAP_TYPE_HASH_OF_MAPS and BPF_MAP_TYPE_ARRAY_OF_MAPS

A usecase where outer HASH map holds hundred of VIP entries and its
associated reuse-ports per VIP stored in REUSEPORT_SOCKARRAY type
inner map, needs to do batch operation for performance gain.

This patch leverages the exiting generic functions for most of the batch
operations. As map-in-map's value contains the actual reference of the inner map,
for BPF_MAP_TYPE_HASH_OF_MAPS type, it needed an extra step to fetch the
map_id from the reference value.

selftests are added in next patch 2/2.

Signed-off-by: Takshak Chahande <ctakshak@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
---
 kernel/bpf/arraymap.c |  2 ++
 kernel/bpf/hashtab.c  | 13 +++++++++++--
 2 files changed, 13 insertions(+), 2 deletions(-)

v4->v6:
- Changes in selftest/bpf patch 2/2

v3->v4:
- Added blank line between var declaration and actual code block (Yonghong)

v1->v3:
- Changes in selftest/bpf patch 2/2

diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c
index b3bf31fd9458..724613da6576 100644
--- a/kernel/bpf/arraymap.c
+++ b/kernel/bpf/arraymap.c
@@ -1345,6 +1345,8 @@ const struct bpf_map_ops array_of_maps_map_ops = {
 	.map_fd_put_ptr = bpf_map_fd_put_ptr,
 	.map_fd_sys_lookup_elem = bpf_map_fd_sys_lookup_elem,
 	.map_gen_lookup = array_of_map_gen_lookup,
+	.map_lookup_batch = generic_map_lookup_batch,
+	.map_update_batch = generic_map_update_batch,
 	.map_check_btf = map_check_no_btf,
 	.map_btf_id = &array_map_btf_ids[0],
 };
diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c
index 3e00e62b2218..705841279d16 100644
--- a/kernel/bpf/hashtab.c
+++ b/kernel/bpf/hashtab.c
@@ -140,7 +140,7 @@ static inline bool htab_use_raw_lock(const struct bpf_htab *htab)
 
 static void htab_init_buckets(struct bpf_htab *htab)
 {
-	unsigned i;
+	unsigned int i;
 
 	for (i = 0; i < htab->n_buckets; i++) {
 		INIT_HLIST_NULLS_HEAD(&htab->buckets[i].head, i);
@@ -1627,7 +1627,7 @@ __htab_map_lookup_and_delete_batch(struct bpf_map *map,
 	void __user *uvalues = u64_to_user_ptr(attr->batch.values);
 	void __user *ukeys = u64_to_user_ptr(attr->batch.keys);
 	void __user *ubatch = u64_to_user_ptr(attr->batch.in_batch);
-	u32 batch, max_count, size, bucket_size;
+	u32 batch, max_count, size, bucket_size, map_id;
 	struct htab_elem *node_to_free = NULL;
 	u64 elem_map_flags, map_flags;
 	struct hlist_nulls_head *head;
@@ -1752,6 +1752,14 @@ __htab_map_lookup_and_delete_batch(struct bpf_map *map,
 			}
 		} else {
 			value = l->key + roundup_key_size;
+			if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) {
+				struct bpf_map **inner_map = value;
+
+				 /* Actual value is the id of the inner map */
+				map_id = map->ops->map_fd_sys_lookup_elem(*inner_map);
+				value = &map_id;
+			}
+
 			if (elem_map_flags & BPF_F_LOCK)
 				copy_map_value_locked(map, dst_val, value,
 						      true);
@@ -2450,5 +2458,6 @@ const struct bpf_map_ops htab_of_maps_map_ops = {
 	.map_fd_sys_lookup_elem = bpf_map_fd_sys_lookup_elem,
 	.map_gen_lookup = htab_of_map_gen_lookup,
 	.map_check_btf = map_check_no_btf,
+	BATCH_OPS(htab),
 	.map_btf_id = &htab_map_btf_ids[0],
 };
-- 
2.30.2

[PATCH bpf-next v6 2/2] selftests/bpf: handle batch operations for map-in-map bpf-maps

[Takshak Chahande]

This patch adds up test cases that handles 4 combinations:
 a) outer map: BPF_MAP_TYPE_ARRAY_OF_MAPS
    inner maps: BPF_MAP_TYPE_ARRAY and BPF_MAP_TYPE_HASH
 b) outer map: BPF_MAP_TYPE_HASH_OF_MAPS
    inner maps: BPF_MAP_TYPE_ARRAY and BPF_MAP_TYPE_HASH

Signed-off-by: Takshak Chahande <ctakshak@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
---
 .../bpf/map_tests/map_in_map_batch_ops.c      | 252 ++++++++++++++++++
 1 file changed, 252 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/map_tests/map_in_map_batch_ops.c

v5->v6:
- Fixed all the other map fd leaks 

v4->v5:
- close all (inner and outer) map fds (Martin)

v3->v4:
- Addressed nits; kept this map test together in map_tests/  (Yonghong, Andrii)

v2->v3:
- Handled transient ENOSPC correctly, bug was found in BPF CI (Daniel)

v1->v2:
- Fixed no format arguments error (Andrii)

diff --git a/tools/testing/selftests/bpf/map_tests/map_in_map_batch_ops.c b/tools/testing/selftests/bpf/map_tests/map_in_map_batch_ops.c
new file mode 100644
index 000000000000..f472d28ad11a
--- /dev/null
+++ b/tools/testing/selftests/bpf/map_tests/map_in_map_batch_ops.c
@@ -0,0 +1,252 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <unistd.h>
+
+#include <bpf/bpf.h>
+#include <bpf/libbpf.h>
+
+#include <test_maps.h>
+
+#define OUTER_MAP_ENTRIES 10
+
+static __u32 get_map_id_from_fd(int map_fd)
+{
+	struct bpf_map_info map_info = {};
+	uint32_t info_len = sizeof(map_info);
+	int ret;
+
+	ret = bpf_obj_get_info_by_fd(map_fd, &map_info, &info_len);
+	CHECK(ret < 0, "Finding map info failed", "error:%s\n",
+	      strerror(errno));
+
+	return map_info.id;
+}
+
+/* This creates number of OUTER_MAP_ENTRIES maps that will be stored
+ * in outer map and return the created map_fds
+ */
+static void create_inner_maps(enum bpf_map_type map_type,
+			      __u32 *inner_map_fds)
+{
+	int map_fd, map_index, ret;
+	__u32 map_key = 0, map_id;
+	char map_name[15];
+
+	for (map_index = 0; map_index < OUTER_MAP_ENTRIES; map_index++) {
+		memset(map_name, 0, sizeof(map_name));
+		sprintf(map_name, "inner_map_fd_%d", map_index);
+		map_fd = bpf_map_create(map_type, map_name, sizeof(__u32),
+					sizeof(__u32), 1, NULL);
+		CHECK(map_fd < 0,
+		      "inner bpf_map_create() failed",
+		      "map_type=(%d) map_name(%s), error:%s\n",
+		      map_type, map_name, strerror(errno));
+
+		/* keep track of the inner map fd as it is required
+		 * to add records in outer map
+		 */
+		inner_map_fds[map_index] = map_fd;
+
+		/* Add entry into this created map
+		 * eg: map1 key = 0, value = map1's map id
+		 *     map2 key = 0, value = map2's map id
+		 */
+		map_id = get_map_id_from_fd(map_fd);
+		ret = bpf_map_update_elem(map_fd, &map_key, &map_id, 0);
+		CHECK(ret != 0,
+		      "bpf_map_update_elem failed",
+		      "map_type=(%d) map_name(%s), error:%s\n",
+		      map_type, map_name, strerror(errno));
+	}
+}
+
+static int create_outer_map(enum bpf_map_type map_type, __u32 inner_map_fd)
+{
+	int outer_map_fd;
+	LIBBPF_OPTS(bpf_map_create_opts, attr);
+
+	attr.inner_map_fd = inner_map_fd;
+	outer_map_fd = bpf_map_create(map_type, "outer_map", sizeof(__u32),
+				      sizeof(__u32), OUTER_MAP_ENTRIES,
+				      &attr);
+	CHECK(outer_map_fd < 0,
+	      "outer bpf_map_create()",
+	      "map_type=(%d), error:%s\n",
+	      map_type, strerror(errno));
+
+	return outer_map_fd;
+}
+
+static void validate_fetch_results(int outer_map_fd,
+				   __u32 *fetched_keys, __u32 *fetched_values,
+				   __u32 max_entries_fetched)
+{
+	__u32 inner_map_key, inner_map_value;
+	int inner_map_fd, entry, err;
+	__u32 outer_map_value;
+
+	for (entry = 0; entry < max_entries_fetched; ++entry) {
+		outer_map_value = fetched_values[entry];
+		inner_map_fd = bpf_map_get_fd_by_id(outer_map_value);
+		CHECK(inner_map_fd < 0,
+		      "Failed to get inner map fd",
+		      "from id(%d), error=%s\n",
+		      outer_map_value, strerror(errno));
+		err = bpf_map_get_next_key(inner_map_fd, NULL, &inner_map_key);
+		CHECK(err != 0,
+		      "Failed to get inner map key",
+		      "error=%s\n", strerror(errno));
+
+		err = bpf_map_lookup_elem(inner_map_fd, &inner_map_key,
+					  &inner_map_value);
+
+		close(inner_map_fd);
+
+		CHECK(err != 0,
+		      "Failed to get inner map value",
+		      "for key(%d), error=%s\n",
+		      inner_map_key, strerror(errno));
+
+		/* Actual value validation */
+		CHECK(outer_map_value != inner_map_value,
+		      "Failed to validate inner map value",
+		      "fetched(%d) and lookedup(%d)!\n",
+		      outer_map_value, inner_map_value);
+	}
+}
+
+static void fetch_and_validate(int outer_map_fd,
+			       struct bpf_map_batch_opts *opts,
+			       __u32 batch_size, bool delete_entries)
+{
+	__u32 *fetched_keys, *fetched_values, total_fetched = 0;
+	__u32 batch_key = 0, fetch_count, step_size;
+	int err, max_entries = OUTER_MAP_ENTRIES;
+	__u32 value_size = sizeof(__u32);
+
+	/* Total entries needs to be fetched */
+	fetched_keys = calloc(max_entries, value_size);
+	fetched_values = calloc(max_entries, value_size);
+	CHECK((!fetched_keys || !fetched_values),
+	      "Memory allocation failed for fetched_keys or fetched_values",
+	      "error=%s\n", strerror(errno));
+
+	for (step_size = batch_size;
+	     step_size <= max_entries;
+	     step_size += batch_size) {
+		fetch_count = step_size;
+		err = delete_entries
+		      ? bpf_map_lookup_and_delete_batch(outer_map_fd,
+				      total_fetched ? &batch_key : NULL,
+				      &batch_key,
+				      fetched_keys + total_fetched,
+				      fetched_values + total_fetched,
+				      &fetch_count, opts)
+		      : bpf_map_lookup_batch(outer_map_fd,
+				      total_fetched ? &batch_key : NULL,
+				      &batch_key,
+				      fetched_keys + total_fetched,
+				      fetched_values + total_fetched,
+				      &fetch_count, opts);
+
+		if (err && errno == ENOSPC) {
+			/* Fetch again with higher batch size */
+			total_fetched = 0;
+			continue;
+		}
+
+		CHECK((err < 0 && (errno != ENOENT)),
+		      "lookup with steps failed",
+		      "error: %s\n", strerror(errno));
+
+		/* Update the total fetched number */
+		total_fetched += fetch_count;
+		if (err)
+			break;
+	}
+
+	CHECK((total_fetched != max_entries),
+	      "Unable to fetch expected entries !",
+	      "total_fetched(%d) and max_entries(%d) error: (%d):%s\n",
+	      total_fetched, max_entries, errno, strerror(errno));
+
+	/* validate the fetched entries */
+	validate_fetch_results(outer_map_fd, fetched_keys,
+			       fetched_values, total_fetched);
+	printf("batch_op(%s) is successful with batch_size(%d)\n",
+	       delete_entries ? "LOOKUP_AND_DELETE" : "LOOKUP", batch_size);
+
+	free(fetched_keys);
+	free(fetched_values);
+}
+
+static void _map_in_map_batch_ops(enum bpf_map_type outer_map_type,
+				  enum bpf_map_type inner_map_type)
+{
+	__u32 *outer_map_keys, *inner_map_fds;
+	__u32 max_entries = OUTER_MAP_ENTRIES;
+	LIBBPF_OPTS(bpf_map_batch_opts, opts);
+	__u32 value_size = sizeof(__u32);
+	int batch_size[2] = {5, 10};
+	__u32 map_index, op_index;
+	int outer_map_fd, ret;
+
+	outer_map_keys = calloc(max_entries, value_size);
+	inner_map_fds = calloc(max_entries, value_size);
+	CHECK((!outer_map_keys || !inner_map_fds),
+	      "Memory allocation failed for outer_map_keys or inner_map_fds",
+	      "error=%s\n", strerror(errno));
+
+	create_inner_maps(inner_map_type, inner_map_fds);
+
+	outer_map_fd = create_outer_map(outer_map_type, *inner_map_fds);
+	/* create outer map keys */
+	for (map_index = 0; map_index < max_entries; map_index++)
+		outer_map_keys[map_index] =
+			((outer_map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS)
+			 ? 9 : 1000) - map_index;
+
+	/* batch operation - map_update */
+	ret = bpf_map_update_batch(outer_map_fd, outer_map_keys,
+				   inner_map_fds, &max_entries, &opts);
+	CHECK(ret != 0,
+	      "Failed to update the outer map batch ops",
+	      "error=%s\n", strerror(errno));
+
+	/* batch operation - map_lookup */
+	for (op_index = 0; op_index < 2; ++op_index)
+		fetch_and_validate(outer_map_fd, &opts,
+				   batch_size[op_index], false);
+
+	/* batch operation - map_lookup_delete */
+	if (outer_map_type == BPF_MAP_TYPE_HASH_OF_MAPS)
+		fetch_and_validate(outer_map_fd, &opts,
+				   max_entries, true /*delete*/);
+
+	/* close all map fds */
+	for (map_index = 0; map_index < max_entries; map_index++)
+		close(inner_map_fds[map_index]);
+	close(outer_map_fd);
+
+	free(inner_map_fds);
+	free(outer_map_keys);
+}
+
+void test_map_in_map_batch_ops_array(void)
+{
+	_map_in_map_batch_ops(BPF_MAP_TYPE_ARRAY_OF_MAPS, BPF_MAP_TYPE_ARRAY);
+	printf("%s:PASS with inner ARRAY map\n", __func__);
+	_map_in_map_batch_ops(BPF_MAP_TYPE_ARRAY_OF_MAPS, BPF_MAP_TYPE_HASH);
+	printf("%s:PASS with inner HASH map\n", __func__);
+}
+
+void test_map_in_map_batch_ops_hash(void)
+{
+	_map_in_map_batch_ops(BPF_MAP_TYPE_HASH_OF_MAPS, BPF_MAP_TYPE_ARRAY);
+	printf("%s:PASS with inner ARRAY map\n", __func__);
+	_map_in_map_batch_ops(BPF_MAP_TYPE_HASH_OF_MAPS, BPF_MAP_TYPE_HASH);
+	printf("%s:PASS with inner HASH map\n", __func__);
+}
-- 
2.30.2

Re: [PATCH bpf-next v6 1/2] bpf: Extend batch operations for map-in-map bpf-maps

[patchwork-bot+netdevbpf]

Hello:

This series was applied to bpf/bpf-next.git (master)
by Alexei Starovoitov <ast@kernel.org>:

On Tue, 10 May 2022 01:22:20 -0700 you wrote:
> This patch extends batch operations support for map-in-map map-types:
> BPF_MAP_TYPE_HASH_OF_MAPS and BPF_MAP_TYPE_ARRAY_OF_MAPS
> 
> A usecase where outer HASH map holds hundred of VIP entries and its
> associated reuse-ports per VIP stored in REUSEPORT_SOCKARRAY type
> inner map, needs to do batch operation for performance gain.
> 
> [...]

Here is the summary with links:
  - [bpf-next,v6,1/2] bpf: Extend batch operations for map-in-map bpf-maps
    https://git.kernel.org/bpf/bpf-next/c/9263dddc7b6f
  - [bpf-next,v6,2/2] selftests/bpf: handle batch operations for map-in-map bpf-maps
    https://git.kernel.org/bpf/bpf-next/c/a82ebb093fc7

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