Merge branch 'Enhance and rework logging controls in libbpf'

Andrii Nakryiko says:

====================

Add new open options and per-program setters to control BTF and program
loading log verboseness and allow providing custom log buffers to capture logs
of interest. Note how custom log_buf and log_level are orthogonal, which
matches previous (alas less customizable) behavior of libbpf, even though it
sort of worked by accident: if someone specified log_level = 1 in
bpf_object__load_xattr(), first attempt to load any BPF program resulted in
wasted bpf() syscall with -EINVAL due to !!log_buf != !!log_level. Then on
retry libbpf would allocated log_buffer and try again, after which prog
loading would succeed and libbpf would print verbose program loading log
through its print callback.

This behavior is now documented and made more efficient, not wasting
unnecessary syscall. But additionally, log_level can be controlled globally on
a per-bpf_object level through bpf_object_open_opts, as well as on
a per-program basis with bpf_program__set_log_buf() and
bpf_program__set_log_level() APIs.

Now that we have a more future-proof way to set log_level, deprecate
bpf_object__load_xattr().

v2->v3:
  - added log_buf selftests for bpf_prog_load() and bpf_btf_load();
  - fix !log_buf in bpf_prog_load (John);
  - fix log_level==0 in bpf_btf_load (thanks selftest!);

v1->v2:
  - fix log_level == 0 handling of bpf_prog_load, add as patch #1 (Alexei);
  - add comments explaining log_buf_size overflow prevention (Alexei).
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>

1 files changed, 276 insertions, 0 deletions

diff --git a/tools/testing/selftests/bpf/prog_tests/log_buf.c b/tools/testing/selftests/bpf/prog_tests/log_buf.c
new file mode 100644
index 000000000000..e469b023962b
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/log_buf.c
@@ -0,0 +1,276 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+
+#include <test_progs.h>
+#include <bpf/btf.h>
+
+#include "test_log_buf.skel.h"
+
+static size_t libbpf_log_pos;
+static char libbpf_log_buf[1024 * 1024];
+static bool libbpf_log_error;
+
+static int libbpf_print_cb(enum libbpf_print_level level, const char *fmt, va_list args)
+{
+	int emitted_cnt;
+	size_t left_cnt;
+
+	left_cnt = sizeof(libbpf_log_buf) - libbpf_log_pos;
+	emitted_cnt = vsnprintf(libbpf_log_buf + libbpf_log_pos, left_cnt, fmt, args);
+
+	if (emitted_cnt < 0 || emitted_cnt + 1 > left_cnt) {
+		libbpf_log_error = true;
+		return 0;
+	}
+
+	libbpf_log_pos += emitted_cnt;
+	return 0;
+}
+
+static void obj_load_log_buf(void)
+{
+	libbpf_print_fn_t old_print_cb = libbpf_set_print(libbpf_print_cb);
+	LIBBPF_OPTS(bpf_object_open_opts, opts);
+	const size_t log_buf_sz = 1024 * 1024;
+	struct test_log_buf* skel;
+	char *obj_log_buf, *good_log_buf, *bad_log_buf;
+	int err;
+
+	obj_log_buf = malloc(3 * log_buf_sz);
+	if (!ASSERT_OK_PTR(obj_log_buf, "obj_log_buf"))
+		return;
+
+	good_log_buf = obj_log_buf + log_buf_sz;
+	bad_log_buf = obj_log_buf + 2 * log_buf_sz;
+	obj_log_buf[0] = good_log_buf[0] = bad_log_buf[0] = '\0';
+
+	opts.kernel_log_buf = obj_log_buf;
+	opts.kernel_log_size = log_buf_sz;
+	opts.kernel_log_level = 4; /* for BTF this will turn into 1 */
+
+	/* In the first round every prog has its own log_buf, so libbpf logs
+	 * don't have program failure logs
+	 */
+	skel = test_log_buf__open_opts(&opts);
+	if (!ASSERT_OK_PTR(skel, "skel_open"))
+		goto cleanup;
+
+	/* set very verbose level for good_prog so we always get detailed logs */
+	bpf_program__set_log_buf(skel->progs.good_prog, good_log_buf, log_buf_sz);
+	bpf_program__set_log_level(skel->progs.good_prog, 2);
+
+	bpf_program__set_log_buf(skel->progs.bad_prog, bad_log_buf, log_buf_sz);
+	/* log_level 0 with custom log_buf means that verbose logs are not
+	 * requested if program load is successful, but libbpf should retry
+	 * with log_level 1 on error and put program's verbose load log into
+	 * custom log_buf
+	 */
+	bpf_program__set_log_level(skel->progs.bad_prog, 0);
+
+	err = test_log_buf__load(skel);
+	if (!ASSERT_ERR(err, "unexpected_load_success"))
+		goto cleanup;
+
+	ASSERT_FALSE(libbpf_log_error, "libbpf_log_error");
+
+	/* there should be no prog loading log because we specified per-prog log buf */
+	ASSERT_NULL(strstr(libbpf_log_buf, "-- BEGIN PROG LOAD LOG --"), "unexp_libbpf_log");
+	ASSERT_OK_PTR(strstr(libbpf_log_buf, "prog 'bad_prog': BPF program load failed"),
+		      "libbpf_log_not_empty");
+	ASSERT_OK_PTR(strstr(obj_log_buf, "DATASEC license"), "obj_log_not_empty");
+	ASSERT_OK_PTR(strstr(good_log_buf, "0: R1=ctx(id=0,off=0,imm=0) R10=fp0"),
+		      "good_log_verbose");
+	ASSERT_OK_PTR(strstr(bad_log_buf, "invalid access to map value, value_size=16 off=16000 size=4"),
+		      "bad_log_not_empty");
+
+	if (env.verbosity > VERBOSE_NONE) {
+		printf("LIBBPF LOG:   \n=================\n%s=================\n", libbpf_log_buf);
+		printf("OBJ LOG:      \n=================\n%s=================\n", obj_log_buf);
+		printf("GOOD_PROG LOG:\n=================\n%s=================\n", good_log_buf);
+		printf("BAD_PROG  LOG:\n=================\n%s=================\n", bad_log_buf);
+	}
+
+	/* reset everything */
+	test_log_buf__destroy(skel);
+	obj_log_buf[0] = good_log_buf[0] = bad_log_buf[0] = '\0';
+	libbpf_log_buf[0] = '\0';
+	libbpf_log_pos = 0;
+	libbpf_log_error = false;
+
+	/* In the second round we let bad_prog's failure be logged through print callback */
+	opts.kernel_log_buf = NULL; /* let everything through into print callback */
+	opts.kernel_log_size = 0;
+	opts.kernel_log_level = 1;
+
+	skel = test_log_buf__open_opts(&opts);
+	if (!ASSERT_OK_PTR(skel, "skel_open"))
+		goto cleanup;
+
+	/* set normal verbose level for good_prog to check log_level is taken into account */
+	bpf_program__set_log_buf(skel->progs.good_prog, good_log_buf, log_buf_sz);
+	bpf_program__set_log_level(skel->progs.good_prog, 1);
+
+	err = test_log_buf__load(skel);
+	if (!ASSERT_ERR(err, "unexpected_load_success"))
+		goto cleanup;
+
+	ASSERT_FALSE(libbpf_log_error, "libbpf_log_error");
+
+	/* this time prog loading error should be logged through print callback */
+	ASSERT_OK_PTR(strstr(libbpf_log_buf, "libbpf: prog 'bad_prog': -- BEGIN PROG LOAD LOG --"),
+		      "libbpf_log_correct");
+	ASSERT_STREQ(obj_log_buf, "", "obj_log__empty");
+	ASSERT_STREQ(good_log_buf, "processed 4 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0\n",
+		     "good_log_ok");
+	ASSERT_STREQ(bad_log_buf, "", "bad_log_empty");
+
+	if (env.verbosity > VERBOSE_NONE) {
+		printf("LIBBPF LOG:   \n=================\n%s=================\n", libbpf_log_buf);
+		printf("OBJ LOG:      \n=================\n%s=================\n", obj_log_buf);
+		printf("GOOD_PROG LOG:\n=================\n%s=================\n", good_log_buf);
+		printf("BAD_PROG  LOG:\n=================\n%s=================\n", bad_log_buf);
+	}
+
+cleanup:
+	free(obj_log_buf);
+	test_log_buf__destroy(skel);
+	libbpf_set_print(old_print_cb);
+}
+
+static void bpf_prog_load_log_buf(void)
+{
+	const struct bpf_insn good_prog_insns[] = {
+		BPF_MOV64_IMM(BPF_REG_0, 0),
+		BPF_EXIT_INSN(),
+	};
+	const size_t good_prog_insn_cnt = sizeof(good_prog_insns) / sizeof(struct bpf_insn);
+	const struct bpf_insn bad_prog_insns[] = {
+		BPF_EXIT_INSN(),
+	};
+	size_t bad_prog_insn_cnt = sizeof(bad_prog_insns) / sizeof(struct bpf_insn);
+	LIBBPF_OPTS(bpf_prog_load_opts, opts);
+	const size_t log_buf_sz = 1024 * 1024;
+	char *log_buf;
+	int fd = -1;
+
+	log_buf = malloc(log_buf_sz);
+	if (!ASSERT_OK_PTR(log_buf, "log_buf_alloc"))
+		return;
+	opts.log_buf = log_buf;
+	opts.log_size = log_buf_sz;
+
+	/* with log_level == 0 log_buf shoud stay empty for good prog */
+	log_buf[0] = '\0';
+	opts.log_level = 0;
+	fd = bpf_prog_load(BPF_PROG_TYPE_SOCKET_FILTER, "good_prog", "GPL",
+			   good_prog_insns, good_prog_insn_cnt, &opts);
+	ASSERT_STREQ(log_buf, "", "good_log_0");
+	ASSERT_GE(fd, 0, "good_fd1");
+	if (fd >= 0)
+		close(fd);
+	fd = -1;
+
+	/* log_level == 2 should always fill log_buf, even for good prog */
+	log_buf[0] = '\0';
+	opts.log_level = 2;
+	fd = bpf_prog_load(BPF_PROG_TYPE_SOCKET_FILTER, "good_prog", "GPL",
+			   good_prog_insns, good_prog_insn_cnt, &opts);
+	ASSERT_OK_PTR(strstr(log_buf, "0: R1=ctx(id=0,off=0,imm=0) R10=fp0"), "good_log_2");
+	ASSERT_GE(fd, 0, "good_fd2");
+	if (fd >= 0)
+		close(fd);
+	fd = -1;
+
+	/* log_level == 0 should fill log_buf for bad prog */
+	log_buf[0] = '\0';
+	opts.log_level = 0;
+	fd = bpf_prog_load(BPF_PROG_TYPE_SOCKET_FILTER, "bad_prog", "GPL",
+			   bad_prog_insns, bad_prog_insn_cnt, &opts);
+	ASSERT_OK_PTR(strstr(log_buf, "R0 !read_ok"), "bad_log_0");
+	ASSERT_LT(fd, 0, "bad_fd");
+	if (fd >= 0)
+		close(fd);
+	fd = -1;
+
+	free(log_buf);
+}
+
+static void bpf_btf_load_log_buf(void)
+{
+	LIBBPF_OPTS(bpf_btf_load_opts, opts);
+	const size_t log_buf_sz = 1024 * 1024;
+	const void *raw_btf_data;
+	__u32 raw_btf_size;
+	struct btf *btf;
+	char *log_buf;
+	int fd = -1;
+
+	btf = btf__new_empty();
+	if (!ASSERT_OK_PTR(btf, "empty_btf"))
+		return;
+
+	ASSERT_GT(btf__add_int(btf, "int", 4, 0), 0, "int_type");
+
+	raw_btf_data = btf__raw_data(btf, &raw_btf_size);
+	if (!ASSERT_OK_PTR(raw_btf_data, "raw_btf_data_good"))
+		goto cleanup;
+
+	log_buf = malloc(log_buf_sz);
+	if (!ASSERT_OK_PTR(log_buf, "log_buf_alloc"))
+		goto cleanup;
+	opts.log_buf = log_buf;
+	opts.log_size = log_buf_sz;
+
+	/* with log_level == 0 log_buf shoud stay empty for good BTF */
+	log_buf[0] = '\0';
+	opts.log_level = 0;
+	fd = bpf_btf_load(raw_btf_data, raw_btf_size, &opts);
+	ASSERT_STREQ(log_buf, "", "good_log_0");
+	ASSERT_GE(fd, 0, "good_fd1");
+	if (fd >= 0)
+		close(fd);
+	fd = -1;
+
+	/* log_level == 2 should always fill log_buf, even for good BTF */
+	log_buf[0] = '\0';
+	opts.log_level = 2;
+	fd = bpf_btf_load(raw_btf_data, raw_btf_size, &opts);
+	printf("LOG_BUF: %s\n", log_buf);
+	ASSERT_OK_PTR(strstr(log_buf, "magic: 0xeb9f"), "good_log_2");
+	ASSERT_GE(fd, 0, "good_fd2");
+	if (fd >= 0)
+		close(fd);
+	fd = -1;
+
+	/* make BTF bad, add pointer pointing to non-existing type */
+	ASSERT_GT(btf__add_ptr(btf, 100), 0, "bad_ptr_type");
+
+	raw_btf_data = btf__raw_data(btf, &raw_btf_size);
+	if (!ASSERT_OK_PTR(raw_btf_data, "raw_btf_data_bad"))
+		goto cleanup;
+
+	/* log_level == 0 should fill log_buf for bad BTF */
+	log_buf[0] = '\0';
+	opts.log_level = 0;
+	fd = bpf_btf_load(raw_btf_data, raw_btf_size, &opts);
+	printf("LOG_BUF: %s\n", log_buf);
+	ASSERT_OK_PTR(strstr(log_buf, "[2] PTR (anon) type_id=100 Invalid type_id"), "bad_log_0");
+	ASSERT_LT(fd, 0, "bad_fd");
+	if (fd >= 0)
+		close(fd);
+	fd = -1;
+
+cleanup:
+	free(log_buf);
+	btf__free(btf);
+}
+
+void test_log_buf(void)
+{
+	if (test__start_subtest("obj_load_log_buf"))
+		obj_load_log_buf();
+	if (test__start_subtest("bpf_prog_load_log_buf"))
+		bpf_prog_load_log_buf();
+	if (test__start_subtest("bpf_btf_load_log_buf"))
+		bpf_btf_load_log_buf();
+}