[PATCH bpf-next] bpf: force BPF_F_RDONLY_PROG on insn array creation

[PATCH bpf-next] bpf: force BPF_F_RDONLY_PROG on insn array creation

[Anton Protopopov]

The original implementation added a hack to check_mem_access()
to prevent programs from writing into insn arrays. To get rid
of this hack, enforce BPF_F_RDONLY_PROG on map creation.

Also fix the corresponding selftest, as the error message changes
with this patch.

Suggested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Anton Protopopov <a.s.protopopov@gmail.com>
---
 kernel/bpf/bpf_insn_array.c                        |  3 +++
 kernel/bpf/verifier.c                              | 13 ++++++-------
 tools/testing/selftests/bpf/progs/verifier_gotox.c |  2 +-
 3 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/kernel/bpf/bpf_insn_array.c b/kernel/bpf/bpf_insn_array.c
index 61ce52882632..c96630cb75bf 100644
--- a/kernel/bpf/bpf_insn_array.c
+++ b/kernel/bpf/bpf_insn_array.c
@@ -55,6 +55,9 @@ static struct bpf_map *insn_array_alloc(union bpf_attr *attr)
 
 	bpf_map_init_from_attr(&insn_array->map, attr);
 
+	/* BPF programs aren't allowed to write to the map */
+	insn_array->map.map_flags |= BPF_F_RDONLY_PROG;
+
 	return &insn_array->map;
 }
 
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index d865848b789d..58f99557ba38 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -7565,11 +7565,6 @@ static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regn
 			verbose(env, "R%d leaks addr into map\n", value_regno);
 			return -EACCES;
 		}
-		if (t == BPF_WRITE && insn_array) {
-			verbose(env, "writes into insn_array not allowed\n");
-			return -EACCES;
-		}
-
 		err = check_map_access_type(env, regno, off, size, t);
 		if (err)
 			return err;
@@ -7584,10 +7579,14 @@ static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regn
 		} else if (t == BPF_READ && value_regno >= 0) {
 			struct bpf_map *map = reg->map_ptr;
 
-			/* if map is read-only, track its contents as scalars */
+			/*
+			 * If map is read-only, track its contents as scalars,
+			 * unless it is an insn array (see the special case below)
+			 */
 			if (tnum_is_const(reg->var_off) &&
 			    bpf_map_is_rdonly(map) &&
-			    map->ops->map_direct_value_addr) {
+			    map->ops->map_direct_value_addr &&
+			    map->map_type != BPF_MAP_TYPE_INSN_ARRAY) {
 				int map_off = off + reg->var_off.value;
 				u64 val = 0;
 
diff --git a/tools/testing/selftests/bpf/progs/verifier_gotox.c b/tools/testing/selftests/bpf/progs/verifier_gotox.c
index 536c9f3e2170..607dad058ca1 100644
--- a/tools/testing/selftests/bpf/progs/verifier_gotox.c
+++ b/tools/testing/selftests/bpf/progs/verifier_gotox.c
@@ -244,7 +244,7 @@ jt0_%=:								\
 }
 
 SEC("socket")
-__failure __msg("writes into insn_array not allowed")
+__failure __msg("write into map forbidden, value_size=16 off=8 size=8")
 __naked void jump_table_no_writes(void)
 {
 	asm volatile ("						\
-- 
2.34.1

Re: [PATCH bpf-next] bpf: force BPF_F_RDONLY_PROG on insn array creation

[Anton Protopopov]

On 25/11/27 09:06PM, Anton Protopopov wrote:
> The original implementation added a hack to check_mem_access()
> to prevent programs from writing into insn arrays. To get rid
> of this hack, enforce BPF_F_RDONLY_PROG on map creation.
> 
> Also fix the corresponding selftest, as the error message changes
> with this patch.

Will resend it as part of a series, as the following patch depends on this one.
(See https://lore.kernel.org/bpf/20251127210732.3241888-1-a.s.protopopov@gmail.com/T/#u)

> [...]