Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Alexei Starovoitov says: ==================== pull-request: bpf 2020-08-21 The following pull-request contains BPF updates for your *net* tree. We've added 11 non-merge commits during the last 5 day(s) which contain a total of 12 files changed, 78 insertions(+), 24 deletions(-). The main changes are: 1) three fixes in BPF task iterator logic, from Yonghong. 2) fix for compressed dwarf sections in vmlinux, from Jiri. 3) fix xdp attach regression, from Andrii. ====================

David S. Miller committed 6 years ago

Commit 4af7b322 parents 774d977 b16fc0912 files · +78 −24Browse files

Unified · Split

12 files

include/uapi/linux/bpf.h +5 −5
kernel/bpf/bpf_iter.c +14 −1
kernel/bpf/task_iter.c +4 −2
net/core/dev.c +8 −6
tools/bpf/bpftool/pids.c +2 −0
tools/bpf/resolve_btfids/main.c +36 −0
tools/include/uapi/linux/bpf.h +5 −5
tools/lib/bpf/btf_dump.c +1 −1
tools/lib/bpf/libbpf.c +1 −1
tools/testing/selftests/bpf/.gitignore +0 −1
tools/testing/selftests/bpf/Makefile +1 −1
tools/testing/selftests/bpf/prog_tests/test_global_funcs.c +1 −1

include/uapi/linux/bpf.h+5−5

@@ -767,7 +767,7 @@ union bpf_attr { · 766 unmodified lines
767	767	*
768	768	* Also, note that bpf_trace_printk\ () is slow, and should
769	769	* only be used for debugging purposes. For this reason, a notice
770		− * bloc (spanning several lines) is printed to kernel logs and
	770	+ * block (spanning several lines) is printed to kernel logs and
771	771	* states that the helper should not be used "for production use"
772	772	* the first time this helper is used (or more precisely, when
773	773	* trace_printk\ () buffers are allocated). For passing values
@@ -1033,14 +1033,14 @@ union bpf_attr { · 259 unmodified lines
1033	1033	*
1034	1034	* int ret;
1035	1035	* struct bpf_tunnel_key key = {};
1036		− *
	1036	+ *
1037	1037	* ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1038	1038	* if (ret < 0)
1039	1039	* return TC_ACT_SHOT; // drop packet
1040		− *
	1040	+ *
1041	1041	* if (key.remote_ipv4 != 0x0a000001)
1042	1042	* return TC_ACT_SHOT; // drop packet
1043		− *
	1043	+ *
1044	1044	* return TC_ACT_OK; // accept packet
1045	1045	*
1046	1046	* This interface can also be used with all encapsulation devices
@@ -1147,7 +1147,7 @@ union bpf_attr { · 100 unmodified lines
1147	1147	* Description
1148	1148	* Retrieve the realm or the route, that is to say the
1149	1149	* tclassid field of the destination for the skb. The
1150		− * indentifier retrieved is a user-provided tag, similar to the
	1150	+ * identifier retrieved is a user-provided tag, similar to the
1151	1151	* one used with the net_cls cgroup (see description for
1152	1152	* bpf_get_cgroup_classid\ () helper), but here this tag is
1153	1153	* held by a route (a destination entry), not by a task.

kernel/bpf/bpf_iter.c+14−1

@@ -67,6 +67,9 @@ static void bpf_iter_done_stop(struct seq_file *seq) · 66 unmodified lines
67	67	iter_priv->done_stop = true;
68	68	}
69	69
	70	+/* maximum visited objects before bailing out */
	71	+#define MAX_ITER_OBJECTS 1000000
	72	+
70	73	/* bpf_seq_read, a customized and simpler version for bpf iterator.
71	74	* no_llseek is assumed for this file.
72	75	* The following are differences from seq_read():
@@ -79,7 +82,7 @@ static ssize_t bpf_seq_read(struct file file, char __user buf, size_t size, · 6 unmodified lines
79	82	{
80	83	struct seq_file *seq = file->private_data;
81	84	size_t n, offs, copied = 0;
82		− int err = 0;
	85	+ int err = 0, num_objs = 0;
83	86	void *p;
84	87
85	88	mutex_lock(&seq->lock);
@@ -135,6 +138,7 @@ static ssize_t bpf_seq_read(struct file file, char __user buf, size_t size, · 49 unmodified lines
135	138	while (1) {
136	139	loff_t pos = seq->index;
137	140
	141	+ num_objs++;
138	142	offs = seq->count;
139	143	p = seq->op->next(seq, p, &seq->index);
140	144	if (pos == seq->index) {
@@ -153,6 +157,15 @@ static ssize_t bpf_seq_read(struct file file, char __user buf, size_t size, · 12 unmodified lines
153	157	if (seq->count >= size)
154	158	break;
155	159
	160	+ if (num_objs >= MAX_ITER_OBJECTS) {
	161	+ if (offs == 0) {
	162	+ err = -EAGAIN;
	163	+ seq->op->stop(seq, p);
	164	+ goto done;
	165	+ }
	166	+ break;
	167	+ }
	168	+
156	169	err = seq->op->show(seq, p);
157	170	if (err > 0) {
158	171	bpf_iter_dec_seq_num(seq);

kernel/bpf/task_iter.c+4−2

@@ -29,8 +29,9 @@ static struct task_struct task_seq_get_next(struct pid_namespace ns, · 28 unmodified lines
29	29
30	30	rcu_read_lock();
31	31	retry:
32		− pid = idr_get_next(&ns->idr, tid);
	32	+ pid = find_ge_pid(*tid, ns);
33	33	if (pid) {
	34	+ *tid = pid_nr_ns(pid, ns);
34	35	task = get_pid_task(pid, PIDTYPE_PID);
35	36	if (!task) {
36	37	++*tid;
@@ -178,10 +179,11 @@ again: · 141 unmodified lines
178	179	f = fcheck_files(curr_files, curr_fd);
179	180	if (!f)
180	181	continue;
	182	+ if (!get_file_rcu(f))
	183	+ continue;
181	184
182	185	/* set info->fd */
183	186	info->fd = curr_fd;
184		− get_file(f);
185	187	rcu_read_unlock();
186	188	return f;
187	189	}

net/core/dev.c+8−6

@@ -8742,13 +8742,15 @@ struct bpf_xdp_link { · 8741 unmodified lines
8742	8742	int flags;
8743	8743	};
8744	8744
8745		−static enum bpf_xdp_mode dev_xdp_mode(u32 flags)
	8745	+static enum bpf_xdp_mode dev_xdp_mode(struct net_device *dev, u32 flags)
8746	8746	{
8747	8747	if (flags & XDP_FLAGS_HW_MODE)
8748	8748	return XDP_MODE_HW;
8749	8749	if (flags & XDP_FLAGS_DRV_MODE)
8750	8750	return XDP_MODE_DRV;
8751		− return XDP_MODE_SKB;
	8751	+ if (flags & XDP_FLAGS_SKB_MODE)
	8752	+ return XDP_MODE_SKB;
	8753	+ return dev->netdev_ops->ndo_bpf ? XDP_MODE_DRV : XDP_MODE_SKB;
8752	8754	}
8753	8755
8754	8756	static bpf_op_t dev_xdp_bpf_op(struct net_device *dev, enum bpf_xdp_mode mode)
@@ -8896,7 +8898,7 @@ static int dev_xdp_attach(struct net_device dev, struct netlink_ext_ack extack · 141 unmodified lines
8896	8898	return -EINVAL;
8897	8899	}
8898	8900
8899		− mode = dev_xdp_mode(flags);
	8901	+ mode = dev_xdp_mode(dev, flags);
8900	8902	/* can't replace attached link */
8901	8903	if (dev_xdp_link(dev, mode)) {
8902	8904	NL_SET_ERR_MSG(extack, "Can't replace active BPF XDP link");
@@ -8984,7 +8986,7 @@ static int dev_xdp_detach_link(struct net_device *dev, · 81 unmodified lines
8984	8986
8985	8987	ASSERT_RTNL();
8986	8988
8987		− mode = dev_xdp_mode(link->flags);
	8989	+ mode = dev_xdp_mode(dev, link->flags);
8988	8990	if (dev_xdp_link(dev, mode) != link)
8989	8991	return -EINVAL;
8990	8992
@@ -9080,7 +9082,7 @@ static int bpf_xdp_link_update(struct bpf_link link, struct bpf_prog new_prog, · 89 unmodified lines
9080	9082	goto out_unlock;
9081	9083	}
9082	9084
9083		− mode = dev_xdp_mode(xdp_link->flags);
	9085	+ mode = dev_xdp_mode(xdp_link->dev, xdp_link->flags);
9084	9086	bpf_op = dev_xdp_bpf_op(xdp_link->dev, mode);
9085	9087	err = dev_xdp_install(xdp_link->dev, mode, bpf_op, NULL,
9086	9088	xdp_link->flags, new_prog);
@@ -9164,7 +9166,7 @@ out_put_dev: · 77 unmodified lines
9164	9166	int dev_change_xdp_fd(struct net_device dev, struct netlink_ext_ack extack,
9165	9167	int fd, int expected_fd, u32 flags)
9166	9168	{
9167		− enum bpf_xdp_mode mode = dev_xdp_mode(flags);
	9169	+ enum bpf_xdp_mode mode = dev_xdp_mode(dev, flags);
9168	9170	struct bpf_prog new_prog = NULL, old_prog = NULL;
9169	9171	int err;
9170	9172

tools/bpf/bpftool/pids.c+2−0

@@ -134,6 +134,8 @@ int build_obj_refs_table(struct obj_refs_table *table, enum bpf_obj_type type) · 133 unmodified lines
134	134	while (true) {
135	135	ret = read(fd, buf, sizeof(buf));
136	136	if (ret < 0) {
	137	+ if (errno == EAGAIN)
	138	+ continue;
137	139	err = -errno;
138	140	p_err("failed to read PID iterator output: %d", err);
139	141	goto out;

tools/bpf/resolve_btfids/main.c+36−0

@@ -233,6 +233,39 @@ static struct btf_id add_symbol(struct rb_root root, char *name, size_t size) · 232 unmodified lines
233	233	return btf_id__add(root, id, false);
234	234	}
235	235
	236	+/*
	237	+ * The data of compressed section should be aligned to 4
	238	+ * (for 32bit) or 8 (for 64 bit) bytes. The binutils ld
	239	+ * sets sh_addralign to 1, which makes libelf fail with
	240	+ * misaligned section error during the update:
	241	+ * FAILED elf_update(WRITE): invalid section alignment
	242	+ *
	243	+ * While waiting for ld fix, we fix the compressed sections
	244	+ * sh_addralign value manualy.
	245	+ */
	246	+static int compressed_section_fix(Elf elf, Elf_Scn scn, GElf_Shdr *sh)
	247	+{
	248	+ int expected = gelf_getclass(elf) == ELFCLASS32 ? 4 : 8;
	249	+
	250	+ if (!(sh->sh_flags & SHF_COMPRESSED))
	251	+ return 0;
	252	+
	253	+ if (sh->sh_addralign == expected)
	254	+ return 0;
	255	+
	256	+ pr_debug2(" - fixing wrong alignment sh_addralign %u, expected %u\n",
	257	+ sh->sh_addralign, expected);
	258	+
	259	+ sh->sh_addralign = expected;
	260	+
	261	+ if (gelf_update_shdr(scn, sh) == 0) {
	262	+ printf("FAILED cannot update section header: %s\n",
	263	+ elf_errmsg(-1));
	264	+ return -1;
	265	+ }
	266	+ return 0;
	267	+}
	268	+
236	269	static int elf_collect(struct object *obj)
237	270	{
238	271	Elf_Scn *scn = NULL;
@@ -309,6 +342,9 @@ static int elf_collect(struct object *obj) · 70 unmodified lines
309	342	obj->efile.idlist_shndx = idx;
310	343	obj->efile.idlist_addr = sh.sh_addr;
311	344	}
	345	+
	346	+ if (compressed_section_fix(elf, scn, &sh))
	347	+ return -1;
312	348	}
313	349
314	350	return 0;

tools/include/uapi/linux/bpf.h+5−5

@@ -767,7 +767,7 @@ union bpf_attr { · 766 unmodified lines
767	767	*
768	768	* Also, note that bpf_trace_printk\ () is slow, and should
769	769	* only be used for debugging purposes. For this reason, a notice
770		− * bloc (spanning several lines) is printed to kernel logs and
	770	+ * block (spanning several lines) is printed to kernel logs and
771	771	* states that the helper should not be used "for production use"
772	772	* the first time this helper is used (or more precisely, when
773	773	* trace_printk\ () buffers are allocated). For passing values
@@ -1033,14 +1033,14 @@ union bpf_attr { · 259 unmodified lines
1033	1033	*
1034	1034	* int ret;
1035	1035	* struct bpf_tunnel_key key = {};
1036		− *
	1036	+ *
1037	1037	* ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1038	1038	* if (ret < 0)
1039	1039	* return TC_ACT_SHOT; // drop packet
1040		− *
	1040	+ *
1041	1041	* if (key.remote_ipv4 != 0x0a000001)
1042	1042	* return TC_ACT_SHOT; // drop packet
1043		− *
	1043	+ *
1044	1044	* return TC_ACT_OK; // accept packet
1045	1045	*
1046	1046	* This interface can also be used with all encapsulation devices
@@ -1147,7 +1147,7 @@ union bpf_attr { · 100 unmodified lines
1147	1147	* Description
1148	1148	* Retrieve the realm or the route, that is to say the
1149	1149	* tclassid field of the destination for the skb. The
1150		− * indentifier retrieved is a user-provided tag, similar to the
	1150	+ * identifier retrieved is a user-provided tag, similar to the
1151	1151	* one used with the net_cls cgroup (see description for
1152	1152	* bpf_get_cgroup_classid\ () helper), but here this tag is
1153	1153	* held by a route (a destination entry), not by a task.

tools/lib/bpf/btf_dump.c+1−1

@@ -879,7 +879,7 @@ static void btf_dump_emit_struct_def(struct btf_dump *d, · 878 unmodified lines
879	879	btf_dump_printf(d, ": %d", m_sz);
880	880	off = m_off + m_sz;
881	881	} else {
882		− m_sz = max(0LL, btf__resolve_size(d->btf, m->type));
	882	+ m_sz = max((__s64)0, btf__resolve_size(d->btf, m->type));
883	883	off = m_off + m_sz * 8;
884	884	}
885	885	btf_dump_printf(d, ";");

tools/lib/bpf/libbpf.c+1−1

@@ -2264,7 +2264,7 @@ static int bpf_object__init_user_btf_maps(struct bpf_object *obj, bool strict, · 2263 unmodified lines
2264	2264	data = elf_getdata(scn, NULL);
2265	2265	if (!scn \|\| !data) {
2266	2266	pr_warn("failed to get Elf_Data from map section %d (%s)\n",
2267		− obj->efile.maps_shndx, MAPS_ELF_SEC);
	2267	+ obj->efile.btf_maps_shndx, MAPS_ELF_SEC);
2268	2268	return -EINVAL;
2269	2269	}
2270	2270

tools/testing/selftests/bpf/.gitignore+0−1

@@ -6,7 +6,6 @@ test_lpm_map · 5 unmodified lines
6	6	test_tag
7	7	FEATURE-DUMP.libbpf
8	8	fixdep
9		−test_align
10	9	test_dev_cgroup
11	10	/test_progs*
12	11	test_tcpbpf_user

tools/testing/selftests/bpf/Makefile+1−1

@@ -32,7 +32,7 @@ LDLIBS += -lcap -lelf -lz -lrt -lpthread · 31 unmodified lines
32	32
33	33	# Order correspond to 'make run_tests' order
34	34	TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test_progs \
35		− test_align test_verifier_log test_dev_cgroup test_tcpbpf_user \
	35	+ test_verifier_log test_dev_cgroup test_tcpbpf_user \
36	36	test_sock test_btf test_sockmap get_cgroup_id_user test_socket_cookie \
37	37	test_cgroup_storage \
38	38	test_netcnt test_tcpnotify_user test_sock_fields test_sysctl \

tools/testing/selftests/bpf/prog_tests/test_global_funcs.c+1−1

@@ -19,7 +19,7 @@ static int libbpf_debug_print(enum libbpf_print_level level, · 18 unmodified lines
19	19	log_buf = va_arg(args, char *);
20	20	if (!log_buf)
21	21	goto out;
22		− if (strstr(log_buf, err_str) == 0)
	22	+ if (err_str && strstr(log_buf, err_str) == 0)
23	23	found = true;
24	24	out:
25	25	printf(format, log_buf);