Skip to contentSkip to conversation

libbpf: Don't take direct pointers into BTF data from st_ops

PATCH In review David Vernet 3 participants 5 messages Jul 24, 2024 – Jul 25, 2024

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 libbpf: Don't take direct pointers into BTF data from st_ops

In struct bpf_struct_ops, we have take a pointer to a BTF type name, and a struct btf_type. This was presumably done for convenience, but can actually result in subtle and confusing bugs given that BTF data can be invalidated before a program is loaded. For example, in sched_ext, we may sometimes resize a data section after a skeleton has been opened, but before the struct_ops scheduler map has been loaded. This may cause the BTF data to be realloc'd, which can then cause a UAF when loading the program because the struct_ops map has pointers directly into the BTF data.

We're already storing the BTF type_id in struct bpf_struct_ops. Because type_id is stable, we can therefore just update the places where we were looking at those pointers to instead do the lookups we need from the type_id.

Fixes 590a00888250 ("bpf: libbpf: Add STRUCT_OPS support")
---
 tools/lib/bpf/libbpf.c | 23 +++++++++++++----------
 1 file changed, 13 insertions(+), 10 deletions(-)
tools/lib/bpf/libbpf.c +13 -10
index a3be6f8fac09..e55353887439 100644
@@ -496,8 +496,6 @@ struct bpf_program {
 };
 
 struct bpf_struct_ops {
-	const char *tname;
-	const struct btf_type *type;
 	struct bpf_program **progs;
 	__u32 *kern_func_off;
 	/* e.g. struct tcp_congestion_ops in bpf_prog's btf format */
@@ -1083,11 +1081,14 @@ static int bpf_object_adjust_struct_ops_autoload(struct bpf_object *obj)
 			continue;
 
 		for (j = 0; j < obj->nr_maps; ++j) {
+			const struct btf_type *type;
+
 			map = &obj->maps[j];
 			if (!bpf_map__is_struct_ops(map))
 				continue;
 
-			vlen = btf_vlen(map->st_ops->type);
+			type = btf__type_by_id(obj->btf, map->st_ops->type_id);
+			vlen = btf_vlen(type);
 			for (k = 0; k < vlen; ++k) {
 				slot_prog = map->st_ops->progs[k];
 				if (prog != slot_prog)
@@ -1121,8 +1122,8 @@ static int bpf_map__init_kern_struct_ops(struct bpf_map *map)
 	int err;
 
 	st_ops = map->st_ops;
-	type = st_ops->type;
-	tname = st_ops->tname;
+	type = btf__type_by_id(btf, st_ops->type_id);
+	tname = btf__name_by_offset(btf, type->name_off);
 	err = find_struct_ops_kern_types(obj, tname, &mod_btf,
 					 &kern_type, &kern_type_id,
 					 &kern_vtype, &kern_vtype_id,
@@ -1423,8 +1424,6 @@ static int init_struct_ops_maps(struct bpf_object *obj, const char *sec_name,
 		memcpy(st_ops->data,
 		       data->d_buf + vsi->offset,
 		       type->size);
-		st_ops->tname = tname;
-		st_ops->type = type;
 		st_ops->type_id = type_id;
 
 		pr_debug("struct_ops init: struct %s(type_id=%u) %s found at offset %u\n",
@@ -8445,11 +8444,13 @@ static int bpf_object__resolve_externs(struct bpf_object *obj,
 
 static void bpf_map_prepare_vdata(const struct bpf_map *map)
 {
+	const struct btf_type *type;
 	struct bpf_struct_ops *st_ops;
 	__u32 i;
 
 	st_ops = map->st_ops;
-	for (i = 0; i < btf_vlen(st_ops->type); i++) {
+	type = btf__type_by_id(map->obj->btf, st_ops->type_id);
+	for (i = 0; i < btf_vlen(type); i++) {
 		struct bpf_program *prog = st_ops->progs[i];
 		void *kern_data;
 		int prog_fd;
@@ -9712,6 +9713,7 @@ static struct bpf_map *find_struct_ops_map_by_offset(struct bpf_object *obj,
 static int bpf_object__collect_st_ops_relos(struct bpf_object *obj,
 					    Elf64_Shdr *shdr, Elf_Data *data)
 {
+	const struct btf_type *type;
 	const struct btf_member *member;
 	struct bpf_struct_ops *st_ops;
 	struct bpf_program *prog;
@@ -9771,13 +9773,14 @@ static int bpf_object__collect_st_ops_relos(struct bpf_object *obj,
 		}
 		insn_idx = sym->st_value / BPF_INSN_SZ;
 
-		member = find_member_by_offset(st_ops->type, moff * 8);
+		type = btf__type_by_id(btf, st_ops->type_id);
+		member = find_member_by_offset(type, moff * 8);
 		if (!member) {
 			pr_warn("struct_ops reloc %s: cannot find member at moff %u\n",
 				map->name, moff);
 			return -EINVAL;
 		}
-		member_idx = member - btf_members(st_ops->type);
+		member_idx = member - btf_members(type);
 		name = btf__name_by_offset(btf, member->name_off);
 
 		if (!resolve_func_ptr(btf, member->type, NULL)) {
Signature · 1 line
2.45.2
#1 lore raw

PATCH 2/2 selftests/bpf: Add test for resizing data map with struct_ops

Tests that if you resize a map after opening a skel, that it doesn't cause a UAF which causes a struct_ops map to fail to be able to load.

---
 .../bpf/prog_tests/struct_ops_resize.c        | 30 +++++++++++++++++++
 .../selftests/bpf/progs/struct_ops_resize.c   | 24 +++++++++++++++
 2 files changed, 54 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c
 create mode 100644 tools/testing/selftests/bpf/progs/struct_ops_resize.c
tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c new file +30 -0
new file mode 100644
index 000000000000..7584f91c2bd1
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <test_progs.h>
+#include "struct_ops_resize.skel.h"
+
+static void resize_datasec(void)
+{
+	struct struct_ops_resize *skel;
+	int err;
+
+	skel = struct_ops_resize__open();
+	if (!ASSERT_OK_PTR(skel, "struct_ops_resize__open"))
+		return;
+
+	err  = bpf_map__set_value_size(skel->maps.data_resizable, 1 << 15);
+	if (!ASSERT_OK(err, "bpf_map__set_value_size"))
+		goto cleanup;
+
+	err = struct_ops_resize__load(skel);
+	ASSERT_OK(err, "struct_ops_resize__load");
+
+cleanup:
+	struct_ops_resize__destroy(skel);
+}
+
+void test_struct_ops_resize(void)
+{
+	if (test__start_subtest("resize_datasec"))
+		resize_datasec();
+}
tools/testing/selftests/bpf/progs/struct_ops_resize.c new file +24 -0
new file mode 100644
index 000000000000..d0b235f4bbaa
@@ -0,0 +1,24 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+char _license[] SEC("license") = "GPL";
+
+char resizable[1] SEC(".data.resizable");
+
+SEC("struct_ops/test_1")
+int BPF_PROG(test_1)
+{
+	return 0;
+}
+
+struct bpf_testmod_ops {
+	int (*test_1)(void);
+};
+
+SEC(".struct_ops.link")
+struct bpf_testmod_ops testmod = {
+	.test_1 = (void *)test_1
+};
Signature · 1 line
2.45.2
#2 lore raw

Re: [PATCH bpf-next 2/2] selftests/bpf: Add test for resizing data map with struct_ops

Quoted 37 lines · David Vernet

Tests that if you resize a map after opening a skel, that it doesn't cause a UAF which causes a struct_ops map to fail to be able to load.

---
 .../bpf/prog_tests/struct_ops_resize.c        | 30 +++++++++++++++++++
 .../selftests/bpf/progs/struct_ops_resize.c   | 24 +++++++++++++++
 2 files changed, 54 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c
 create mode 100644 tools/testing/selftests/bpf/progs/struct_ops_resize.c
tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c new file +19 -0
new file mode 100644
index 000000000000..7584f91c2bd1
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <test_progs.h>
+#include "struct_ops_resize.skel.h"
+
+static void resize_datasec(void)
+{
+       struct struct_ops_resize *skel;
+       int err;
+
+       skel = struct_ops_resize__open();
+       if (!ASSERT_OK_PTR(skel, "struct_ops_resize__open"))
+               return;
+
+       err  = bpf_map__set_value_size(skel->maps.data_resizable, 1 << 15);
+       if (!ASSERT_OK(err, "bpf_map__set_value_size"))
+               goto cleanup;
+
+       err = struct_ops_resize__load(skel);

tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c line 29

Quoted 10 lines
tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c new file +10 -0
+       ASSERT_OK(err, "struct_ops_resize__load");
+
+cleanup:
+       struct_ops_resize__destroy(skel);
+}
+
+void test_struct_ops_resize(void)
+{
+       if (test__start_subtest("resize_datasec"))
+               resize_datasec();

It seems a bit unnecessary to add an entire new test with a subtest just for this. Would you mind adding this testing logic into the already existing prog_tests/global_map_resize.c set of cases?

I've applied patch #1, as it's obviously correct, so I didn't want to delay the fix. Thanks!

Quoted 34 lines
tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c new file +1 -0
+}
tools/testing/selftests/bpf/progs/struct_ops_resize.c new file +24 -1
new file mode 100644
index 000000000000..d0b235f4bbaa
@@ -0,0 +1,24 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+char _license[] SEC("license") = "GPL";
+
+char resizable[1] SEC(".data.resizable");
+
+SEC("struct_ops/test_1")
+int BPF_PROG(test_1)
+{
+       return 0;
+}
+
+struct bpf_testmod_ops {
+       int (*test_1)(void);
+};
+
+SEC(".struct_ops.link")
+struct bpf_testmod_ops testmod = {
+       .test_1 = (void *)test_1
+};
--

2.45.2

#3 lore raw

Re: [PATCH bpf-next 2/2] selftests/bpf: Add test for resizing data map with struct_ops

Quoted 44 lines · Andrii Nakryiko
Quoted 43 lines · David Vernet

Tests that if you resize a map after opening a skel, that it doesn't cause a UAF which causes a struct_ops map to fail to be able to load.

---
 .../bpf/prog_tests/struct_ops_resize.c        | 30 +++++++++++++++++++
 .../selftests/bpf/progs/struct_ops_resize.c   | 24 +++++++++++++++
 2 files changed, 54 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c
 create mode 100644 tools/testing/selftests/bpf/progs/struct_ops_resize.c
tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c new file +25 -0
new file mode 100644
index 000000000000..7584f91c2bd1
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <test_progs.h>
+#include "struct_ops_resize.skel.h"
+
+static void resize_datasec(void)
+{
+       struct struct_ops_resize *skel;
+       int err;
+
+       skel = struct_ops_resize__open();
+       if (!ASSERT_OK_PTR(skel, "struct_ops_resize__open"))
+               return;
+
+       err  = bpf_map__set_value_size(skel->maps.data_resizable, 1 << 15);
+       if (!ASSERT_OK(err, "bpf_map__set_value_size"))
+               goto cleanup;
+
+       err = struct_ops_resize__load(skel);
+       ASSERT_OK(err, "struct_ops_resize__load");
+
+cleanup:
+       struct_ops_resize__destroy(skel);
+}
+
Quoted 8 lines

tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c line 29

Quoted 4 lines
tools/testing/selftests/bpf/prog_tests/struct_ops_resize.c new file +4 -0
+void test_struct_ops_resize(void)
+{
+       if (test__start_subtest("resize_datasec"))
+               resize_datasec();

It seems a bit unnecessary to add an entire new test with a subtest just for this. Would you mind adding this testing logic into the already existing prog_tests/global_map_resize.c set of cases?

Sure thing, I'll send a subsequent patch that adds the testcase to prog_tests/global_map_resize.c.

Quoted 2 lines

I've applied patch #1, as it's obviously correct, so I didn't want to delay the fix. Thanks!

Thanks!

#4 lore raw
In reply to David Vernet
This series was applied to bpf/bpf-next.git (master) by Andrii Nakryiko <[email protected]>:

Hello:

This series was applied to bpf/bpf-next.git (master) by Andrii Nakryiko <[email protected]>:

Quoted 11 lines · On Wed, 24 Jul 2024 12:14:58 -0500 you wrote:

In struct bpf_struct_ops, we have take a pointer to a BTF type name, and a struct btf_type. This was presumably done for convenience, but can actually result in subtle and confusing bugs given that BTF data can be invalidated before a program is loaded. For example, in sched_ext, we may sometimes resize a data section after a skeleton has been opened, but before the struct_ops scheduler map has been loaded. This may cause the BTF data to be realloc'd, which can then cause a UAF when loading the program because the struct_ops map has pointers directly into the BTF data.

[...]

Here is the summary with links:
- [bpf-next,1/2] libbpf: Don't take direct pointers into BTF data from st_ops

    https://git.kernel.org/bpf/bpf-next/c/7244100e0389

- [bpf-next,2/2] selftests/bpf: Add test for resizing data map with struct_ops (no matching commit)

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html