Merge branch 'bpf-bpffs-bpftool-dump-with-btf'

Yonghong Song says:

====================
Commit a26ca7c9 ("bpf: btf: Add pretty print support to the
basic arraymap") and Commit 699c86d6 ("bpf: btf: add pretty print
for hash/lru_hash maps") added bpffs pretty print for array, hash and
lru hash maps. The pretty print gives users a structurally formatted
dump for keys/values which much easy to understand than raw bytes.

This patch set implemented bpffs pretty print support for
percpu arraymap, percpu hashmap and percpu lru hashmap.
For complex key/value types, the pretty print here is even more useful
due to:

  . large volumne of data making it even harder to correlate bytes
    to a particular field in a particular cpu.
  . kernel rounds the value size for each cpu to multiple of 8.
    User has to be aware of this otherwise wrong value may be
    derived from cpu 1/2/...

For example, we may have a bpffs pretty print like below:
   43602: {
        cpu0: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
        cpu1: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
        cpu2: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
        cpu3: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
   }
for a percpu map.

This patch also added percpu formatted print on bpftool. For example,
bpftool may print like below:
    {
        "key": 0,
        "values": [{
                "cpu": 0,
                "value": {
                    "ui32": 0,
                    "ui16": 0,
                }
            },{
                "cpu": 1,
                "value": {
                    "ui32": 1,
                    "ui16": 0,
                }
            },{
                "cpu": 2,
                "value": {
                    "ui32": 2,
                    "ui16": 0,
                }
            },{
                "cpu": 3,
                "value": {
                    "ui32": 3,
                    "ui16": 0,
                }
            }
        ]
    }

Patch #1 implemented bpffs pretty print for percpu arraymap/hash/lru_hash
in kernel. Patch #2 added the test case in tools bpf selftest test_btf.
Patch #3 added percpu map btf based dump.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

kernel/bpf/arraymap.c

@@ -358,6 +358,29 @@ static void array_map_seq_show_elem(struct bpf_map *map, void *key,
 	rcu_read_unlock();
 }
 
+static void percpu_array_map_seq_show_elem(struct bpf_map *map, void *key,
+					   struct seq_file *m)
+{
+	struct bpf_array *array = container_of(map, struct bpf_array, map);
+	u32 index = *(u32 *)key;
+	void __percpu *pptr;
+	int cpu;
+
+	rcu_read_lock();
+
+	seq_printf(m, "%u: {\n", *(u32 *)key);
+	pptr = array->pptrs[index & array->index_mask];
+	for_each_possible_cpu(cpu) {
+		seq_printf(m, "\tcpu%d: ", cpu);
+		btf_type_seq_show(map->btf, map->btf_value_type_id,
+				  per_cpu_ptr(pptr, cpu), m);
+		seq_puts(m, "\n");
+	}
+	seq_puts(m, "}\n");
+
+	rcu_read_unlock();
+}
+
 static int array_map_check_btf(const struct bpf_map *map,
 			       const struct btf_type *key_type,
 			       const struct btf_type *value_type)
@@ -398,6 +421,7 @@ const struct bpf_map_ops percpu_array_map_ops = {
 	.map_lookup_elem = percpu_array_map_lookup_elem,
 	.map_update_elem = array_map_update_elem,
 	.map_delete_elem = array_map_delete_elem,
+	.map_seq_show_elem = percpu_array_map_seq_show_elem,
 	.map_check_btf = array_map_check_btf,
 };
 

kernel/bpf/hashtab.c

@@ -1285,6 +1285,35 @@ int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
 	return ret;
 }
 
+static void htab_percpu_map_seq_show_elem(struct bpf_map *map, void *key,
+					  struct seq_file *m)
+{
+	struct htab_elem *l;
+	void __percpu *pptr;
+	int cpu;
+
+	rcu_read_lock();
+
+	l = __htab_map_lookup_elem(map, key);
+	if (!l) {
+		rcu_read_unlock();
+		return;
+	}
+
+	btf_type_seq_show(map->btf, map->btf_key_type_id, key, m);
+	seq_puts(m, ": {\n");
+	pptr = htab_elem_get_ptr(l, map->key_size);
+	for_each_possible_cpu(cpu) {
+		seq_printf(m, "\tcpu%d: ", cpu);
+		btf_type_seq_show(map->btf, map->btf_value_type_id,
+				  per_cpu_ptr(pptr, cpu), m);
+		seq_puts(m, "\n");
+	}
+	seq_puts(m, "}\n");
+
+	rcu_read_unlock();
+}
+
 const struct bpf_map_ops htab_percpu_map_ops = {
 	.map_alloc_check = htab_map_alloc_check,
 	.map_alloc = htab_map_alloc,
@@ -1293,6 +1322,7 @@ const struct bpf_map_ops htab_percpu_map_ops = {
 	.map_lookup_elem = htab_percpu_map_lookup_elem,
 	.map_update_elem = htab_percpu_map_update_elem,
 	.map_delete_elem = htab_map_delete_elem,
+	.map_seq_show_elem = htab_percpu_map_seq_show_elem,
 };
 
 const struct bpf_map_ops htab_lru_percpu_map_ops = {
@@ -1303,6 +1333,7 @@ const struct bpf_map_ops htab_lru_percpu_map_ops = {
 	.map_lookup_elem = htab_lru_percpu_map_lookup_elem,
 	.map_update_elem = htab_lru_percpu_map_update_elem,
 	.map_delete_elem = htab_lru_map_delete_elem,
+	.map_seq_show_elem = htab_percpu_map_seq_show_elem,
 };
 
 static int fd_htab_map_alloc_check(union bpf_attr *attr)

tools/bpf/bpftool/map.c

@@ -169,9 +169,28 @@ static int do_dump_btf(const struct btf_dumper *d,
 	if (ret)
 		goto err_end_obj;
 
-	jsonw_name(d->jw, "value");
+	if (!map_is_per_cpu(map_info->type)) {
+		jsonw_name(d->jw, "value");
+		ret = btf_dumper_type(d, map_info->btf_value_type_id, value);
+	} else {
+		unsigned int i, n, step;
 
-	ret = btf_dumper_type(d, map_info->btf_value_type_id, value);
+		jsonw_name(d->jw, "values");
+		jsonw_start_array(d->jw);
+		n = get_possible_cpus();
+		step = round_up(map_info->value_size, 8);
+		for (i = 0; i < n; i++) {
+			jsonw_start_object(d->jw);
+			jsonw_int_field(d->jw, "cpu", i);
+			jsonw_name(d->jw, "value");
+			ret = btf_dumper_type(d, map_info->btf_value_type_id,
+					      value + i * step);
+			jsonw_end_object(d->jw);
+			if (ret)
+				break;
+		}
+		jsonw_end_array(d->jw);
+	}
 
 err_end_obj:
 	/* end of key-value pair */
@@ -298,6 +317,16 @@ static void print_entry_json(struct bpf_map_info *info, unsigned char *key,
 			jsonw_end_object(json_wtr);
 		}
 		jsonw_end_array(json_wtr);
+		if (btf) {
+			struct btf_dumper d = {
+				.btf = btf,
+				.jw = json_wtr,
+				.is_plain_text = false,
+			};
+
+			jsonw_name(json_wtr, "formatted");
+			do_dump_btf(&d, info, key, value);
+		}
 	}
 
 	jsonw_end_object(json_wtr);

tools/testing/selftests/bpf/test_btf.c

@@ -4,6 +4,7 @@
 #include <linux/bpf.h>
 #include <linux/btf.h>
 #include <linux/err.h>
+#include <linux/kernel.h>
 #include <bpf/bpf.h>
 #include <sys/resource.h>
 #include <libelf.h>
@@ -45,7 +46,6 @@ static int count_result(int err)
 	return err;
 }
 
-#define min(a, b) ((a) < (b) ? (a) : (b))
 #define __printf(a, b)	__attribute__((format(printf, a, b)))
 
 __printf(1, 2)
@@ -130,6 +130,7 @@ struct btf_raw_test {
 	bool map_create_err;
 	bool ordered_map;
 	bool lossless_map;
+	bool percpu_map;
 	int hdr_len_delta;
 	int type_off_delta;
 	int str_off_delta;
@@ -2157,6 +2158,7 @@ static struct btf_pprint_test_meta {
 	const char *map_name;
 	bool ordered_map;
 	bool lossless_map;
+	bool percpu_map;
 } pprint_tests_meta[] = {
 {
 	.descr = "BTF pretty print array",
@@ -2164,6 +2166,7 @@ static struct btf_pprint_test_meta {
 	.map_name = "pprint_test_array",
 	.ordered_map = true,
 	.lossless_map = true,
+	.percpu_map = false,
 },
 
 {
@@ -2172,6 +2175,7 @@ static struct btf_pprint_test_meta {
 	.map_name = "pprint_test_hash",
 	.ordered_map = false,
 	.lossless_map = true,
+	.percpu_map = false,
 },
 
 {
@@ -2180,30 +2184,83 @@ static struct btf_pprint_test_meta {
 	.map_name = "pprint_test_lru_hash",
 	.ordered_map = false,
 	.lossless_map = false,
+	.percpu_map = false,
+},
+
+{
+	.descr = "BTF pretty print percpu array",
+	.map_type = BPF_MAP_TYPE_PERCPU_ARRAY,
+	.map_name = "pprint_test_percpu_array",
+	.ordered_map = true,
+	.lossless_map = true,
+	.percpu_map = true,
+},
+
+{
+	.descr = "BTF pretty print percpu hash",
+	.map_type = BPF_MAP_TYPE_PERCPU_HASH,
+	.map_name = "pprint_test_percpu_hash",
+	.ordered_map = false,
+	.lossless_map = true,
+	.percpu_map = true,
+},
+
+{
+	.descr = "BTF pretty print lru percpu hash",
+	.map_type = BPF_MAP_TYPE_LRU_PERCPU_HASH,
+	.map_name = "pprint_test_lru_percpu_hash",
+	.ordered_map = false,
+	.lossless_map = false,
+	.percpu_map = true,
 },
 
 };
 
 
-static void set_pprint_mapv(struct pprint_mapv *v, uint32_t i)
+static void set_pprint_mapv(struct pprint_mapv *v, uint32_t i,
+			    int num_cpus, int rounded_value_size)
 {
-	v->ui32 = i;
-	v->si32 = -i;
-	v->unused_bits2a = 3;
-	v->bits28 = i;
-	v->unused_bits2b = 3;
-	v->ui64 = i;
-	v->aenum = i & 0x03;
+	int cpu;
+
+	for (cpu = 0; cpu < num_cpus; cpu++) {
+		v->ui32 = i + cpu;
+		v->si32 = -i;
+		v->unused_bits2a = 3;
+		v->bits28 = i;
+		v->unused_bits2b = 3;
+		v->ui64 = i;
+		v->aenum = i & 0x03;
+		v = (void *)v + rounded_value_size;
+	}
 }
 
+static int check_line(const char *expected_line, int nexpected_line,
+		      int expected_line_len, const char *line)
+{
+	if (CHECK(nexpected_line == expected_line_len,
+		  "expected_line is too long"))
+		return -1;
+
+	if (strcmp(expected_line, line)) {
+		fprintf(stderr, "unexpected pprint output\n");
+		fprintf(stderr, "expected: %s", expected_line);
+		fprintf(stderr, "    read: %s", line);
+		return -1;
+	}
+
+	return 0;
+}
+
+
 static int do_test_pprint(void)
 {
 	const struct btf_raw_test *test = &pprint_test_template;
 	struct bpf_create_map_attr create_attr = {};
+	bool ordered_map, lossless_map, percpu_map;
+	int err, ret, num_cpus, rounded_value_size;
+	struct pprint_mapv *mapv = NULL;
 	unsigned int key, nr_read_elems;
-	bool ordered_map, lossless_map;
 	int map_fd = -1, btf_fd = -1;
-	struct pprint_mapv mapv = {};
 	unsigned int raw_btf_size;
 	char expected_line[255];
 	FILE *pin_file = NULL;
@@ -2212,7 +2269,6 @@ static int do_test_pprint(void)
 	char *line = NULL;
 	uint8_t *raw_btf;
 	ssize_t nread;
-	int err, ret;
 
 	fprintf(stderr, "%s......", test->descr);
 	raw_btf = btf_raw_create(&hdr_tmpl, test->raw_types,
@@ -2261,9 +2317,18 @@ static int do_test_pprint(void)
 	if (CHECK(err, "bpf_obj_pin(%s): errno:%d.", pin_path, errno))
 		goto done;
 
+	percpu_map = test->percpu_map;
+	num_cpus = percpu_map ? bpf_num_possible_cpus() : 1;
+	rounded_value_size = round_up(sizeof(struct pprint_mapv), 8);
+	mapv = calloc(num_cpus, rounded_value_size);
+	if (CHECK(!mapv, "mapv allocation failure")) {
+		err = -1;
+		goto done;
+	}
+
 	for (key = 0; key < test->max_entries; key++) {
-		set_pprint_mapv(&mapv, key);
-		bpf_map_update_elem(map_fd, &key, &mapv, 0);
+		set_pprint_mapv(mapv, key, num_cpus, rounded_value_size);
+		bpf_map_update_elem(map_fd, &key, mapv, 0);
 	}
 
 	pin_file = fopen(pin_path, "r");
@@ -2286,33 +2351,74 @@ static int do_test_pprint(void)
 	ordered_map = test->ordered_map;
 	lossless_map = test->lossless_map;
 	do {
+		struct pprint_mapv *cmapv;
 		ssize_t nexpected_line;
 		unsigned int next_key;
+		int cpu;
 
 		next_key = ordered_map ? nr_read_elems : atoi(line);
-		set_pprint_mapv(&mapv, next_key);
-		nexpected_line = snprintf(expected_line, sizeof(expected_line),
-					  "%u: {%u,0,%d,0x%x,0x%x,0x%x,{%lu|[%u,%u,%u,%u,%u,%u,%u,%u]},%s}\n",
-					  next_key,
-					  mapv.ui32, mapv.si32,
-					  mapv.unused_bits2a, mapv.bits28, mapv.unused_bits2b,
-					  mapv.ui64,
-					  mapv.ui8a[0], mapv.ui8a[1], mapv.ui8a[2], mapv.ui8a[3],
-					  mapv.ui8a[4], mapv.ui8a[5], mapv.ui8a[6], mapv.ui8a[7],
-					  pprint_enum_str[mapv.aenum]);
-
-		if (CHECK(nexpected_line == sizeof(expected_line),
-			  "expected_line is too long")) {
-			err = -1;
-			goto done;
+		set_pprint_mapv(mapv, next_key, num_cpus, rounded_value_size);
+		cmapv = mapv;
+
+		for (cpu = 0; cpu < num_cpus; cpu++) {
+			if (percpu_map) {
+				/* for percpu map, the format looks like:
+				 * <key>: {
+				 *	cpu0: <value_on_cpu0>
+				 *	cpu1: <value_on_cpu1>
+				 *	...
+				 *	cpun: <value_on_cpun>
+				 * }
+				 *
+				 * let us verify the line containing the key here.
+				 */
+				if (cpu == 0) {
+					nexpected_line = snprintf(expected_line,
+								  sizeof(expected_line),
+								  "%u: {\n",
+								  next_key);
+
+					err = check_line(expected_line, nexpected_line,
+							 sizeof(expected_line), line);
+					if (err == -1)
+						goto done;
+				}
+
+				/* read value@cpu */
+				nread = getline(&line, &line_len, pin_file);
+				if (nread < 0)
+					break;
+			}
+
+			nexpected_line = snprintf(expected_line, sizeof(expected_line),
+						  "%s%u: {%u,0,%d,0x%x,0x%x,0x%x,"
+						  "{%lu|[%u,%u,%u,%u,%u,%u,%u,%u]},%s}\n",
+						  percpu_map ? "\tcpu" : "",
+						  percpu_map ? cpu : next_key,
+						  cmapv->ui32, cmapv->si32,
+						  cmapv->unused_bits2a,
+						  cmapv->bits28,
+						  cmapv->unused_bits2b,
+						  cmapv->ui64,
+						  cmapv->ui8a[0], cmapv->ui8a[1],
+						  cmapv->ui8a[2], cmapv->ui8a[3],
+						  cmapv->ui8a[4], cmapv->ui8a[5],
+						  cmapv->ui8a[6], cmapv->ui8a[7],
+						  pprint_enum_str[cmapv->aenum]);
+
+			err = check_line(expected_line, nexpected_line,
+					 sizeof(expected_line), line);
+			if (err == -1)
+				goto done;
+
+			cmapv = (void *)cmapv + rounded_value_size;
 		}
 
-		if (strcmp(expected_line, line)) {
-			err = -1;
-			fprintf(stderr, "unexpected pprint output\n");
-			fprintf(stderr, "expected: %s", expected_line);
-			fprintf(stderr, "    read: %s", line);
-			goto done;
+		if (percpu_map) {
+			/* skip the last bracket for the percpu map */
+			nread = getline(&line, &line_len, pin_file);
+			if (nread < 0)
+				break;
 		}
 
 		nread = getline(&line, &line_len, pin_file);
@@ -2334,6 +2440,8 @@ static int do_test_pprint(void)
 	err = 0;
 
 done:
+	if (mapv)
+		free(mapv);
 	if (!err)
 		fprintf(stderr, "OK");
 	if (*btf_log_buf && (err || args.always_log))
@@ -2361,6 +2469,7 @@ static int test_pprint(void)
 		pprint_test_template.map_name = pprint_tests_meta[i].map_name;
 		pprint_test_template.ordered_map = pprint_tests_meta[i].ordered_map;
 		pprint_test_template.lossless_map = pprint_tests_meta[i].lossless_map;
+		pprint_test_template.percpu_map = pprint_tests_meta[i].percpu_map;
 
 		err |= count_result(do_test_pprint());
 	}