tools/bpf: selftests: Add bpf_iter selftests
authorYonghong Song <yhs@fb.com>
Sat, 9 May 2020 17:59:23 +0000 (10:59 -0700)
committerAlexei Starovoitov <ast@kernel.org>
Sun, 10 May 2020 00:05:27 +0000 (17:05 -0700)
The added test includes the following subtests:
  - test verifier change for btf_id_or_null
  - test load/create_iter/read for
    ipv6_route/netlink/bpf_map/task/task_file
  - test anon bpf iterator
  - test anon bpf iterator reading one char at a time
  - test file bpf iterator
  - test overflow (single bpf program output not overflow)
  - test overflow (single bpf program output overflows)
  - test bpf prog returning 1

The ipv6_route tests the following verifier change
  - access fields in the variable length array of the structure.

The netlink load tests the following verifier change
  - put a btf_id ptr value in a stack and accessible to
    tracing/iter programs.

The anon bpf iterator also tests link auto attach through skeleton.

  $ test_progs -n 2
  #2/1 btf_id_or_null:OK
  #2/2 ipv6_route:OK
  #2/3 netlink:OK
  #2/4 bpf_map:OK
  #2/5 task:OK
  #2/6 task_file:OK
  #2/7 anon:OK
  #2/8 anon-read-one-char:OK
  #2/9 file:OK
  #2/10 overflow:OK
  #2/11 overflow-e2big:OK
  #2/12 prog-ret-1:OK
  #2 bpf_iter:OK
  Summary: 1/12 PASSED, 0 SKIPPED, 0 FAILED

Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20200509175923.2477637-1-yhs@fb.com
tools/testing/selftests/bpf/prog_tests/bpf_iter.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/bpf_iter_test_kern1.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/bpf_iter_test_kern2.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/bpf_iter_test_kern3.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/bpf_iter_test_kern4.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/bpf_iter_test_kern_common.h [new file with mode: 0644]

diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_iter.c b/tools/testing/selftests/bpf/prog_tests/bpf_iter.c
new file mode 100644 (file)
index 0000000..87c29dd
--- /dev/null
@@ -0,0 +1,409 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include <test_progs.h>
+#include "bpf_iter_ipv6_route.skel.h"
+#include "bpf_iter_netlink.skel.h"
+#include "bpf_iter_bpf_map.skel.h"
+#include "bpf_iter_task.skel.h"
+#include "bpf_iter_task_file.skel.h"
+#include "bpf_iter_test_kern1.skel.h"
+#include "bpf_iter_test_kern2.skel.h"
+#include "bpf_iter_test_kern3.skel.h"
+#include "bpf_iter_test_kern4.skel.h"
+
+static int duration;
+
+static void test_btf_id_or_null(void)
+{
+       struct bpf_iter_test_kern3 *skel;
+
+       skel = bpf_iter_test_kern3__open_and_load();
+       if (CHECK(skel, "bpf_iter_test_kern3__open_and_load",
+                 "skeleton open_and_load unexpectedly succeeded\n")) {
+               bpf_iter_test_kern3__destroy(skel);
+               return;
+       }
+}
+
+static void do_dummy_read(struct bpf_program *prog)
+{
+       struct bpf_link *link;
+       char buf[16] = {};
+       int iter_fd, len;
+
+       link = bpf_program__attach_iter(prog, NULL);
+       if (CHECK(IS_ERR(link), "attach_iter", "attach_iter failed\n"))
+               return;
+
+       iter_fd = bpf_iter_create(bpf_link__fd(link));
+       if (CHECK(iter_fd < 0, "create_iter", "create_iter failed\n"))
+               goto free_link;
+
+       /* not check contents, but ensure read() ends without error */
+       while ((len = read(iter_fd, buf, sizeof(buf))) > 0)
+               ;
+       CHECK(len < 0, "read", "read failed: %s\n", strerror(errno));
+
+       close(iter_fd);
+
+free_link:
+       bpf_link__destroy(link);
+}
+
+static void test_ipv6_route(void)
+{
+       struct bpf_iter_ipv6_route *skel;
+
+       skel = bpf_iter_ipv6_route__open_and_load();
+       if (CHECK(!skel, "bpf_iter_ipv6_route__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       do_dummy_read(skel->progs.dump_ipv6_route);
+
+       bpf_iter_ipv6_route__destroy(skel);
+}
+
+static void test_netlink(void)
+{
+       struct bpf_iter_netlink *skel;
+
+       skel = bpf_iter_netlink__open_and_load();
+       if (CHECK(!skel, "bpf_iter_netlink__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       do_dummy_read(skel->progs.dump_netlink);
+
+       bpf_iter_netlink__destroy(skel);
+}
+
+static void test_bpf_map(void)
+{
+       struct bpf_iter_bpf_map *skel;
+
+       skel = bpf_iter_bpf_map__open_and_load();
+       if (CHECK(!skel, "bpf_iter_bpf_map__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       do_dummy_read(skel->progs.dump_bpf_map);
+
+       bpf_iter_bpf_map__destroy(skel);
+}
+
+static void test_task(void)
+{
+       struct bpf_iter_task *skel;
+
+       skel = bpf_iter_task__open_and_load();
+       if (CHECK(!skel, "bpf_iter_task__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       do_dummy_read(skel->progs.dump_task);
+
+       bpf_iter_task__destroy(skel);
+}
+
+static void test_task_file(void)
+{
+       struct bpf_iter_task_file *skel;
+
+       skel = bpf_iter_task_file__open_and_load();
+       if (CHECK(!skel, "bpf_iter_task_file__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       do_dummy_read(skel->progs.dump_task_file);
+
+       bpf_iter_task_file__destroy(skel);
+}
+
+/* The expected string is less than 16 bytes */
+static int do_read_with_fd(int iter_fd, const char *expected,
+                          bool read_one_char)
+{
+       int err = -1, len, read_buf_len, start;
+       char buf[16] = {};
+
+       read_buf_len = read_one_char ? 1 : 16;
+       start = 0;
+       while ((len = read(iter_fd, buf + start, read_buf_len)) > 0) {
+               start += len;
+               if (CHECK(start >= 16, "read", "read len %d\n", len))
+                       return -1;
+               read_buf_len = read_one_char ? 1 : 16 - start;
+       }
+       if (CHECK(len < 0, "read", "read failed: %s\n", strerror(errno)))
+               return -1;
+
+       err = strcmp(buf, expected);
+       if (CHECK(err, "read", "incorrect read result: buf %s, expected %s\n",
+                 buf, expected))
+               return -1;
+
+       return 0;
+}
+
+static void test_anon_iter(bool read_one_char)
+{
+       struct bpf_iter_test_kern1 *skel;
+       struct bpf_link *link;
+       int iter_fd, err;
+
+       skel = bpf_iter_test_kern1__open_and_load();
+       if (CHECK(!skel, "bpf_iter_test_kern1__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       err = bpf_iter_test_kern1__attach(skel);
+       if (CHECK(err, "bpf_iter_test_kern1__attach",
+                 "skeleton attach failed\n")) {
+               goto out;
+       }
+
+       link = skel->links.dump_task;
+       iter_fd = bpf_iter_create(bpf_link__fd(link));
+       if (CHECK(iter_fd < 0, "create_iter", "create_iter failed\n"))
+               goto out;
+
+       do_read_with_fd(iter_fd, "abcd", read_one_char);
+       close(iter_fd);
+
+out:
+       bpf_iter_test_kern1__destroy(skel);
+}
+
+static int do_read(const char *path, const char *expected)
+{
+       int err, iter_fd;
+
+       iter_fd = open(path, O_RDONLY);
+       if (CHECK(iter_fd < 0, "open", "open %s failed: %s\n",
+                 path, strerror(errno)))
+               return -1;
+
+       err = do_read_with_fd(iter_fd, expected, false);
+       close(iter_fd);
+       return err;
+}
+
+static void test_file_iter(void)
+{
+       const char *path = "/sys/fs/bpf/bpf_iter_test1";
+       struct bpf_iter_test_kern1 *skel1;
+       struct bpf_iter_test_kern2 *skel2;
+       struct bpf_link *link;
+       int err;
+
+       skel1 = bpf_iter_test_kern1__open_and_load();
+       if (CHECK(!skel1, "bpf_iter_test_kern1__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               return;
+
+       link = bpf_program__attach_iter(skel1->progs.dump_task, NULL);
+       if (CHECK(IS_ERR(link), "attach_iter", "attach_iter failed\n"))
+               goto out;
+
+       /* unlink this path if it exists. */
+       unlink(path);
+
+       err = bpf_link__pin(link, path);
+       if (CHECK(err, "pin_iter", "pin_iter to %s failed: %d\n", path, err))
+               goto free_link;
+
+       err = do_read(path, "abcd");
+       if (err)
+               goto unlink_path;
+
+       /* file based iterator seems working fine. Let us a link update
+        * of the underlying link and `cat` the iterator again, its content
+        * should change.
+        */
+       skel2 = bpf_iter_test_kern2__open_and_load();
+       if (CHECK(!skel2, "bpf_iter_test_kern2__open_and_load",
+                 "skeleton open_and_load failed\n"))
+               goto unlink_path;
+
+       err = bpf_link__update_program(link, skel2->progs.dump_task);
+       if (CHECK(err, "update_prog", "update_prog failed\n"))
+               goto destroy_skel2;
+
+       do_read(path, "ABCD");
+
+destroy_skel2:
+       bpf_iter_test_kern2__destroy(skel2);
+unlink_path:
+       unlink(path);
+free_link:
+       bpf_link__destroy(link);
+out:
+       bpf_iter_test_kern1__destroy(skel1);
+}
+
+static void test_overflow(bool test_e2big_overflow, bool ret1)
+{
+       __u32 map_info_len, total_read_len, expected_read_len;
+       int err, iter_fd, map1_fd, map2_fd, len;
+       struct bpf_map_info map_info = {};
+       struct bpf_iter_test_kern4 *skel;
+       struct bpf_link *link;
+       __u32 page_size;
+       char *buf;
+
+       skel = bpf_iter_test_kern4__open();
+       if (CHECK(!skel, "bpf_iter_test_kern4__open",
+                 "skeleton open failed\n"))
+               return;
+
+       /* create two maps: bpf program will only do bpf_seq_write
+        * for these two maps. The goal is one map output almost
+        * fills seq_file buffer and then the other will trigger
+        * overflow and needs restart.
+        */
+       map1_fd = bpf_create_map(BPF_MAP_TYPE_ARRAY, 4, 8, 1, 0);
+       if (CHECK(map1_fd < 0, "bpf_create_map",
+                 "map_creation failed: %s\n", strerror(errno)))
+               goto out;
+       map2_fd = bpf_create_map(BPF_MAP_TYPE_ARRAY, 4, 8, 1, 0);
+       if (CHECK(map2_fd < 0, "bpf_create_map",
+                 "map_creation failed: %s\n", strerror(errno)))
+               goto free_map1;
+
+       /* bpf_seq_printf kernel buffer is one page, so one map
+        * bpf_seq_write will mostly fill it, and the other map
+        * will partially fill and then trigger overflow and need
+        * bpf_seq_read restart.
+        */
+       page_size = sysconf(_SC_PAGE_SIZE);
+
+       if (test_e2big_overflow) {
+               skel->rodata->print_len = (page_size + 8) / 8;
+               expected_read_len = 2 * (page_size + 8);
+       } else if (!ret1) {
+               skel->rodata->print_len = (page_size - 8) / 8;
+               expected_read_len = 2 * (page_size - 8);
+       } else {
+               skel->rodata->print_len = 1;
+               expected_read_len = 2 * 8;
+       }
+       skel->rodata->ret1 = ret1;
+
+       if (CHECK(bpf_iter_test_kern4__load(skel),
+                 "bpf_iter_test_kern4__load", "skeleton load failed\n"))
+               goto free_map2;
+
+       /* setup filtering map_id in bpf program */
+       map_info_len = sizeof(map_info);
+       err = bpf_obj_get_info_by_fd(map1_fd, &map_info, &map_info_len);
+       if (CHECK(err, "get_map_info", "get map info failed: %s\n",
+                 strerror(errno)))
+               goto free_map2;
+       skel->bss->map1_id = map_info.id;
+
+       err = bpf_obj_get_info_by_fd(map2_fd, &map_info, &map_info_len);
+       if (CHECK(err, "get_map_info", "get map info failed: %s\n",
+                 strerror(errno)))
+               goto free_map2;
+       skel->bss->map2_id = map_info.id;
+
+       link = bpf_program__attach_iter(skel->progs.dump_bpf_map, NULL);
+       if (CHECK(IS_ERR(link), "attach_iter", "attach_iter failed\n"))
+               goto free_map2;
+
+       iter_fd = bpf_iter_create(bpf_link__fd(link));
+       if (CHECK(iter_fd < 0, "create_iter", "create_iter failed\n"))
+               goto free_link;
+
+       buf = malloc(expected_read_len);
+       if (!buf)
+               goto close_iter;
+
+       /* do read */
+       total_read_len = 0;
+       if (test_e2big_overflow) {
+               while ((len = read(iter_fd, buf, expected_read_len)) > 0)
+                       total_read_len += len;
+
+               CHECK(len != -1 || errno != E2BIG, "read",
+                     "expected ret -1, errno E2BIG, but get ret %d, error %s\n",
+                         len, strerror(errno));
+               goto free_buf;
+       } else if (!ret1) {
+               while ((len = read(iter_fd, buf, expected_read_len)) > 0)
+                       total_read_len += len;
+
+               if (CHECK(len < 0, "read", "read failed: %s\n",
+                         strerror(errno)))
+                       goto free_buf;
+       } else {
+               do {
+                       len = read(iter_fd, buf, expected_read_len);
+                       if (len > 0)
+                               total_read_len += len;
+               } while (len > 0 || len == -EAGAIN);
+
+               if (CHECK(len < 0, "read", "read failed: %s\n",
+                         strerror(errno)))
+                       goto free_buf;
+       }
+
+       if (CHECK(total_read_len != expected_read_len, "read",
+                 "total len %u, expected len %u\n", total_read_len,
+                 expected_read_len))
+               goto free_buf;
+
+       if (CHECK(skel->bss->map1_accessed != 1, "map1_accessed",
+                 "expected 1 actual %d\n", skel->bss->map1_accessed))
+               goto free_buf;
+
+       if (CHECK(skel->bss->map2_accessed != 2, "map2_accessed",
+                 "expected 2 actual %d\n", skel->bss->map2_accessed))
+               goto free_buf;
+
+       CHECK(skel->bss->map2_seqnum1 != skel->bss->map2_seqnum2,
+             "map2_seqnum", "two different seqnum %lld %lld\n",
+             skel->bss->map2_seqnum1, skel->bss->map2_seqnum2);
+
+free_buf:
+       free(buf);
+close_iter:
+       close(iter_fd);
+free_link:
+       bpf_link__destroy(link);
+free_map2:
+       close(map2_fd);
+free_map1:
+       close(map1_fd);
+out:
+       bpf_iter_test_kern4__destroy(skel);
+}
+
+void test_bpf_iter(void)
+{
+       if (test__start_subtest("btf_id_or_null"))
+               test_btf_id_or_null();
+       if (test__start_subtest("ipv6_route"))
+               test_ipv6_route();
+       if (test__start_subtest("netlink"))
+               test_netlink();
+       if (test__start_subtest("bpf_map"))
+               test_bpf_map();
+       if (test__start_subtest("task"))
+               test_task();
+       if (test__start_subtest("task_file"))
+               test_task_file();
+       if (test__start_subtest("anon"))
+               test_anon_iter(false);
+       if (test__start_subtest("anon-read-one-char"))
+               test_anon_iter(true);
+       if (test__start_subtest("file"))
+               test_file_iter();
+       if (test__start_subtest("overflow"))
+               test_overflow(false, false);
+       if (test__start_subtest("overflow-e2big"))
+               test_overflow(true, false);
+       if (test__start_subtest("prog-ret-1"))
+               test_overflow(false, true);
+}
diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_test_kern1.c b/tools/testing/selftests/bpf/progs/bpf_iter_test_kern1.c
new file mode 100644 (file)
index 0000000..c71a7c2
--- /dev/null
@@ -0,0 +1,4 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#define START_CHAR 'a'
+#include "bpf_iter_test_kern_common.h"
diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_test_kern2.c b/tools/testing/selftests/bpf/progs/bpf_iter_test_kern2.c
new file mode 100644 (file)
index 0000000..8bdc8dc
--- /dev/null
@@ -0,0 +1,4 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#define START_CHAR 'A'
+#include "bpf_iter_test_kern_common.h"
diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_test_kern3.c b/tools/testing/selftests/bpf/progs/bpf_iter_test_kern3.c
new file mode 100644 (file)
index 0000000..636a00f
--- /dev/null
@@ -0,0 +1,18 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+
+char _license[] SEC("license") = "GPL";
+
+SEC("iter/task")
+int dump_task(struct bpf_iter__task *ctx)
+{
+       struct seq_file *seq = ctx->meta->seq;
+       struct task_struct *task = ctx->task;
+       int tgid;
+
+       tgid = task->tgid;
+       bpf_seq_write(seq, &tgid, sizeof(tgid));
+       return 0;
+}
diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_test_kern4.c b/tools/testing/selftests/bpf/progs/bpf_iter_test_kern4.c
new file mode 100644 (file)
index 0000000..b18dc04
--- /dev/null
@@ -0,0 +1,52 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+
+char _license[] SEC("license") = "GPL";
+
+__u32 map1_id = 0, map2_id = 0;
+__u32 map1_accessed = 0, map2_accessed = 0;
+__u64 map1_seqnum = 0, map2_seqnum1 = 0, map2_seqnum2 = 0;
+
+static volatile const __u32 print_len;
+static volatile const __u32 ret1;
+
+SEC("iter/bpf_map")
+int dump_bpf_map(struct bpf_iter__bpf_map *ctx)
+{
+       struct seq_file *seq = ctx->meta->seq;
+       struct bpf_map *map = ctx->map;
+       __u64 seq_num;
+       int i, ret = 0;
+
+       if (map == (void *)0)
+               return 0;
+
+       /* only dump map1_id and map2_id */
+       if (map->id != map1_id && map->id != map2_id)
+               return 0;
+
+       seq_num = ctx->meta->seq_num;
+       if (map->id == map1_id) {
+               map1_seqnum = seq_num;
+               map1_accessed++;
+       }
+
+       if (map->id == map2_id) {
+               if (map2_accessed == 0) {
+                       map2_seqnum1 = seq_num;
+                       if (ret1)
+                               ret = 1;
+               } else {
+                       map2_seqnum2 = seq_num;
+               }
+               map2_accessed++;
+       }
+
+       /* fill seq_file buffer */
+       for (i = 0; i < print_len; i++)
+               bpf_seq_write(seq, &seq_num, sizeof(seq_num));
+
+       return ret;
+}
diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_test_kern_common.h b/tools/testing/selftests/bpf/progs/bpf_iter_test_kern_common.h
new file mode 100644 (file)
index 0000000..bdd51cf
--- /dev/null
@@ -0,0 +1,22 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2020 Facebook */
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+
+char _license[] SEC("license") = "GPL";
+int count = 0;
+
+SEC("iter/task")
+int dump_task(struct bpf_iter__task *ctx)
+{
+       struct seq_file *seq = ctx->meta->seq;
+       char c;
+
+       if (count < 4) {
+               c = START_CHAR + count;
+               bpf_seq_write(seq, &c, sizeof(c));
+               count++;
+       }
+
+       return 0;
+}