Skip to contentSkip to conversation

Make check_func_arg type checks table driven

PATCH RESEND Applied Lorenz Bauer 2 participants 17 messages Sep 21, 2020 – Sep 23, 2020

Subsystems (8)
lore
Share & tools
lore
#0 lore raw

v4 cover Make check_func_arg type checks table driven

I'm not sure why, but I missed sending the patchset to netdev@ last week. I guess that is why it's slipped through the cracks.

Changes in v4:
- Output the desired type on BTF ID mismatch (Martin)

Changes in v3:
- Fix BTF_ID_LIST_SINGLE if BTF is disabled (Martin)
- Drop incorrect arg_btf_id in bpf_sk_storage.c (Martin)
- Check for arg_btf_id in check_func_proto (Martin)
- Drop incorrect PTR_TO_BTF_ID from fullsock_types (Martin)
- Introduce btf_seq_file_ids in bpf_trace.c to reduce duplication

Changes in v2:
- Make the series stand alone (Martin)
- Drop incorrect BTF_SET_START fix (Andrii)
- Only support a single BTF ID per argument (Martin)
- Introduce BTF_ID_LIST_SINGLE macro (Andrii)
- Skip check_ctx_reg iff register is NULL
- Change output of check_reg_type slightly, to avoid touching tests

Original cover letter:

Currently, check_func_arg has this pretty gnarly if statement that compares the valid arg_type with the actualy reg_type. Sprinkled in-between are checks for register_is_null, to short circuit these tests if we're dealing with a nullable arg_type. There is also some code for later bounds / access checking hidden away in there.

This series of patches refactors the function into something like this:

   if (reg_is_null && arg_type_is_nullable)
     skip type checking

do type checking, including BTF validation

do bounds / access checking

The type checking is now table driven, which makes it easy to extend the acceptable types. Maybe more importantly, using a table makes it easy to provide more helpful verifier output (see the last patch).

Lorenz Bauer (11):
btf: make btf_set_contains take a const pointer
bpf: check scalar or invalid register in check_helper_mem_access
btf: Add BTF_ID_LIST_SINGLE macro
bpf: allow specifying a BTF ID per argument in function protos
bpf: make BTF pointer type checking generic
bpf: make reference tracking generic
bpf: make context access check generic
bpf: set meta->raw_mode for pointers close to use
bpf: check ARG_PTR_TO_SPINLOCK register type in check_func_arg
bpf: hoist type checking for nullable arg types
bpf: use a table to drive helper arg type checks

 include/linux/bpf.h            |  21 +-
 include/linux/btf_ids.h        |   8 +
 kernel/bpf/bpf_inode_storage.c |   8 +-
 kernel/bpf/btf.c               |  15 +-
 kernel/bpf/stackmap.c          |   5 +-
 kernel/bpf/verifier.c          | 338 ++++++++++++++++++---------------
 kernel/trace/bpf_trace.c       |  15 +-
 net/core/bpf_sk_storage.c      |   8 +-
 net/core/filter.c              |  31 +--
 net/ipv4/bpf_tcp_ca.c          |  19 +-
 tools/include/linux/btf_ids.h  |   8 +
 11 files changed, 239 insertions(+), 237 deletions(-)
Signature · 1 line
2.25.1
#1 lore raw

PATCH v4 1/11 btf: make btf_set_contains take a const pointer

bsearch doesn't modify the contents of the array, so we can take a const pointer.

---
 include/linux/bpf.h | 2 +-
 kernel/bpf/btf.c    | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
include/linux/bpf.h +1 -1
index 5dcce0364634..4cbf92f5ecdb 100644
@@ -1901,6 +1901,6 @@ int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
 		       void *addr1, void *addr2);
 
 struct btf_id_set;
-bool btf_id_set_contains(struct btf_id_set *set, u32 id);
+bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
 
 #endif /* _LINUX_BPF_H */
kernel/bpf/btf.c +1 -1
index f9ac6935ab3c..a2330f6fe2e6 100644
@@ -4772,7 +4772,7 @@ static int btf_id_cmp_func(const void *a, const void *b)
 	return *pa - *pb;
 }
 
-bool btf_id_set_contains(struct btf_id_set *set, u32 id)
+bool btf_id_set_contains(const struct btf_id_set *set, u32 id)
 {
 	return bsearch(&id, set->ids, set->cnt, sizeof(u32), btf_id_cmp_func) != NULL;
 }
Signature · 1 line
2.25.1
#2 lore raw

PATCH v4 2/11 bpf: check scalar or invalid register in check_helper_mem_access

In reply to Lorenz Bauer

Move the check for a NULL or zero register to check_helper_mem_access. This makes check_stack_boundary easier to understand.

---
 kernel/bpf/verifier.c | 24 +++++++++++-------------
 1 file changed, 11 insertions(+), 13 deletions(-)
kernel/bpf/verifier.c +11 -13
index 814bc6c1ad16..c997f81c500b 100644
@@ -3594,18 +3594,6 @@ static int check_stack_boundary(struct bpf_verifier_env *env, int regno,
 	struct bpf_func_state *state = func(env, reg);
 	int err, min_off, max_off, i, j, slot, spi;
 
-	if (reg->type != PTR_TO_STACK) {
-		/* Allow zero-byte read from NULL, regardless of pointer type */
-		if (zero_size_allowed && access_size == 0 &&
-		    register_is_null(reg))
-			return 0;
-
-		verbose(env, "R%d type=%s expected=%s\n", regno,
-			reg_type_str[reg->type],
-			reg_type_str[PTR_TO_STACK]);
-		return -EACCES;
-	}
-
 	if (tnum_is_const(reg->var_off)) {
 		min_off = max_off = reg->var_off.value + reg->off;
 		err = __check_stack_boundary(env, regno, min_off, access_size,
@@ -3750,9 +3738,19 @@ static int check_helper_mem_access(struct bpf_verifier_env *env, int regno,
 					   access_size, zero_size_allowed,
 					   "rdwr",
 					   &env->prog->aux->max_rdwr_access);
-	default: /* scalar_value|ptr_to_stack or invalid ptr */
+	case PTR_TO_STACK:
 		return check_stack_boundary(env, regno, access_size,
 					    zero_size_allowed, meta);
+	default: /* scalar_value or invalid ptr */
+		/* Allow zero-byte read from NULL, regardless of pointer type */
+		if (zero_size_allowed && access_size == 0 &&
+		    register_is_null(reg))
+			return 0;
+
+		verbose(env, "R%d type=%s expected=%s\n", regno,
+			reg_type_str[reg->type],
+			reg_type_str[PTR_TO_STACK]);
+		return -EACCES;
 	}
 }
 
Signature · 1 line
2.25.1
#3 lore raw

PATCH v4 3/11 btf: Add BTF_ID_LIST_SINGLE macro

In reply to Lorenz Bauer

Add a convenience macro that allows defining a BTF ID list with a single item. This lets us cut down on repetitive macros.

---
 include/linux/btf_ids.h       | 8 ++++++++
 tools/include/linux/btf_ids.h | 8 ++++++++
 2 files changed, 16 insertions(+)
include/linux/btf_ids.h +8 -0
index 210b086188a3..57890b357f85 100644
@@ -76,6 +76,13 @@ extern u32 name[];
 #define BTF_ID_LIST_GLOBAL(name)			\
 __BTF_ID_LIST(name, globl)
 
+/* The BTF_ID_LIST_SINGLE macro defines a BTF_ID_LIST with
+ * a single entry.
+ */
+#define BTF_ID_LIST_SINGLE(name, prefix, typename)	\
+	BTF_ID_LIST(name) \
+	BTF_ID(prefix, typename)
+
 /*
  * The BTF_ID_UNUSED macro defines 4 zero bytes.
  * It's used when we want to define 'unused' entry
@@ -140,6 +147,7 @@ extern struct btf_id_set name;
 #define BTF_ID(prefix, name)
 #define BTF_ID_UNUSED
 #define BTF_ID_LIST_GLOBAL(name) u32 name[1];
+#define BTF_ID_LIST_SINGLE(name, prefix, typename) static u32 name[1];
 #define BTF_SET_START(name) static struct btf_id_set name = { 0 };
 #define BTF_SET_START_GLOBAL(name) static struct btf_id_set name = { 0 };
 #define BTF_SET_END(name)
tools/include/linux/btf_ids.h +8 -0
index 210b086188a3..57890b357f85 100644
@@ -76,6 +76,13 @@ extern u32 name[];
 #define BTF_ID_LIST_GLOBAL(name)			\
 __BTF_ID_LIST(name, globl)
 
+/* The BTF_ID_LIST_SINGLE macro defines a BTF_ID_LIST with
+ * a single entry.
+ */
+#define BTF_ID_LIST_SINGLE(name, prefix, typename)	\
+	BTF_ID_LIST(name) \
+	BTF_ID(prefix, typename)
+
 /*
  * The BTF_ID_UNUSED macro defines 4 zero bytes.
  * It's used when we want to define 'unused' entry
@@ -140,6 +147,7 @@ extern struct btf_id_set name;
 #define BTF_ID(prefix, name)
 #define BTF_ID_UNUSED
 #define BTF_ID_LIST_GLOBAL(name) u32 name[1];
+#define BTF_ID_LIST_SINGLE(name, prefix, typename) static u32 name[1];
 #define BTF_SET_START(name) static struct btf_id_set name = { 0 };
 #define BTF_SET_START_GLOBAL(name) static struct btf_id_set name = { 0 };
 #define BTF_SET_END(name)
Signature · 1 line
2.25.1
#4 lore raw

PATCH v4 4/11 bpf: allow specifying a BTF ID per argument in function protos

In reply to Lorenz Bauer

Function prototypes using ARG_PTR_TO_BTF_ID currently use two ways to signal which BTF IDs are acceptable. First, bpf_func_proto.btf_id is an array of IDs, one for each argument. This array is only accessed up to the highest numbered argument that uses ARG_PTR_TO_BTF_ID and may therefore be less than five arguments long. It usually points at a BTF_ID_LIST. Second, check_btf_id is a function pointer that is called by the verifier if present. It gets the actual BTF ID of the register, and the argument number we're currently checking. It turns out that the only user check_arg_btf_id ignores the argument, and is simply used to check whether the BTF ID has a struct sock_common at it's start.

Replace both of these mechanisms with an explicit BTF ID for each argument in a function proto. Thanks to btf_struct_ids_match this is very flexible:
check_arg_btf_id can be replaced by requiring struct sock_common.

---
 include/linux/bpf.h            | 18 +++++++-------
 kernel/bpf/bpf_inode_storage.c |  8 +++----
 kernel/bpf/btf.c               | 13 ----------
 kernel/bpf/stackmap.c          |  5 ++--
 kernel/bpf/verifier.c          | 44 +++++++++++++++++-----------------
 kernel/trace/bpf_trace.c       | 15 ++++--------
 net/core/bpf_sk_storage.c      |  8 ++-----
 net/core/filter.c              | 31 +++++++-----------------
 net/ipv4/bpf_tcp_ca.c          | 19 ++++-----------
 9 files changed, 58 insertions(+), 103 deletions(-)
include/linux/bpf.h +10 -8
index 4cbf92f5ecdb..732e2f144b6d 100644
@@ -326,12 +326,16 @@ struct bpf_func_proto {
 		};
 		enum bpf_arg_type arg_type[5];
 	};
-	int *btf_id; /* BTF ids of arguments */
-	bool (*check_btf_id)(u32 btf_id, u32 arg); /* if the argument btf_id is
-						    * valid. Often used if more
-						    * than one btf id is permitted
-						    * for this argument.
-						    */
+	union {
+		struct {
+			u32 *arg1_btf_id;
+			u32 *arg2_btf_id;
+			u32 *arg3_btf_id;
+			u32 *arg4_btf_id;
+			u32 *arg5_btf_id;
+		};
+		u32 *arg_btf_id[5];
+	};
 	int *ret_btf_id; /* return value btf_id */
 	bool (*allowed)(const struct bpf_prog *prog);
 };
@@ -1381,8 +1385,6 @@ int btf_struct_access(struct bpf_verifier_log *log,
 		      u32 *next_btf_id);
 bool btf_struct_ids_match(struct bpf_verifier_log *log,
 			  int off, u32 id, u32 need_type_id);
-int btf_resolve_helper_id(struct bpf_verifier_log *log,
-			  const struct bpf_func_proto *fn, int);
 
 int btf_distill_func_proto(struct bpf_verifier_log *log,
 			   struct btf *btf,
kernel/bpf/bpf_inode_storage.c +3 -5
index 75be02799c0f..6edff97ad594 100644
@@ -249,9 +249,7 @@ const struct bpf_map_ops inode_storage_map_ops = {
 	.map_owner_storage_ptr = inode_storage_ptr,
 };
 
-BTF_ID_LIST(bpf_inode_storage_btf_ids)
-BTF_ID_UNUSED
-BTF_ID(struct, inode)
+BTF_ID_LIST_SINGLE(bpf_inode_storage_btf_ids, struct, inode)
 
 const struct bpf_func_proto bpf_inode_storage_get_proto = {
 	.func		= bpf_inode_storage_get,
@@ -259,9 +257,9 @@ const struct bpf_func_proto bpf_inode_storage_get_proto = {
 	.ret_type	= RET_PTR_TO_MAP_VALUE_OR_NULL,
 	.arg1_type	= ARG_CONST_MAP_PTR,
 	.arg2_type	= ARG_PTR_TO_BTF_ID,
+	.arg2_btf_id	= &bpf_inode_storage_btf_ids[0],
 	.arg3_type	= ARG_PTR_TO_MAP_VALUE_OR_NULL,
 	.arg4_type	= ARG_ANYTHING,
-	.btf_id		= bpf_inode_storage_btf_ids,
 };
 
 const struct bpf_func_proto bpf_inode_storage_delete_proto = {
@@ -270,5 +268,5 @@ const struct bpf_func_proto bpf_inode_storage_delete_proto = {
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_CONST_MAP_PTR,
 	.arg2_type	= ARG_PTR_TO_BTF_ID,
-	.btf_id		= bpf_inode_storage_btf_ids,
+	.arg2_btf_id	= &bpf_inode_storage_btf_ids[0],
 };
kernel/bpf/btf.c +0 -13
index a2330f6fe2e6..5d3c36e13139 100644
@@ -4193,19 +4193,6 @@ bool btf_struct_ids_match(struct bpf_verifier_log *log,
 	return true;
 }
 
-int btf_resolve_helper_id(struct bpf_verifier_log *log,
-			  const struct bpf_func_proto *fn, int arg)
-{
-	int id;
-
-	if (fn->arg_type[arg] != ARG_PTR_TO_BTF_ID || !btf_vmlinux)
-		return -EINVAL;
-	id = fn->btf_id[arg];
-	if (!id || id > btf_vmlinux->nr_types)
-		return -EINVAL;
-	return id;
-}
-
 static int __get_type_size(struct btf *btf, u32 btf_id,
 			   const struct btf_type **bad_type)
 {
kernel/bpf/stackmap.c +2 -3
index a2fa006f430e..06065fa27124 100644
@@ -665,18 +665,17 @@ BPF_CALL_4(bpf_get_task_stack, struct task_struct *, task, void *, buf,
 	return __bpf_get_stack(regs, task, NULL, buf, size, flags);
 }
 
-BTF_ID_LIST(bpf_get_task_stack_btf_ids)
-BTF_ID(struct, task_struct)
+BTF_ID_LIST_SINGLE(bpf_get_task_stack_btf_ids, struct, task_struct)
 
 const struct bpf_func_proto bpf_get_task_stack_proto = {
 	.func		= bpf_get_task_stack,
 	.gpl_only	= false,
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &bpf_get_task_stack_btf_ids[0],
 	.arg2_type	= ARG_PTR_TO_UNINIT_MEM,
 	.arg3_type	= ARG_CONST_SIZE_OR_ZERO,
 	.arg4_type	= ARG_ANYTHING,
-	.btf_id		= bpf_get_task_stack_btf_ids,
 };
 
 BPF_CALL_4(bpf_get_stack_pe, struct bpf_perf_event_data_kern *, ctx,
kernel/bpf/verifier.c +22 -22

79 lines not on this page · raw message

kernel/trace/bpf_trace.c +5 -10
index b2a5380eb187..ebf9be4d0d6a 100644
@@ -743,19 +743,18 @@ BPF_CALL_5(bpf_seq_printf, struct seq_file *, m, char *, fmt, u32, fmt_size,
 	return err;
 }
 
-BTF_ID_LIST(bpf_seq_printf_btf_ids)
-BTF_ID(struct, seq_file)
+BTF_ID_LIST_SINGLE(btf_seq_file_ids, struct, seq_file)
 
 static const struct bpf_func_proto bpf_seq_printf_proto = {
 	.func		= bpf_seq_printf,
 	.gpl_only	= true,
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &btf_seq_file_ids[0],
 	.arg2_type	= ARG_PTR_TO_MEM,
 	.arg3_type	= ARG_CONST_SIZE,
 	.arg4_type      = ARG_PTR_TO_MEM_OR_NULL,
 	.arg5_type      = ARG_CONST_SIZE_OR_ZERO,
-	.btf_id		= bpf_seq_printf_btf_ids,
 };
 
 BPF_CALL_3(bpf_seq_write, struct seq_file *, m, const void *, data, u32, len)
@@ -763,17 +762,14 @@ BPF_CALL_3(bpf_seq_write, struct seq_file *, m, const void *, data, u32, len)
 	return seq_write(m, data, len) ? -EOVERFLOW : 0;
 }
 
-BTF_ID_LIST(bpf_seq_write_btf_ids)
-BTF_ID(struct, seq_file)
-
 static const struct bpf_func_proto bpf_seq_write_proto = {
 	.func		= bpf_seq_write,
 	.gpl_only	= true,
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &btf_seq_file_ids[0],
 	.arg2_type	= ARG_PTR_TO_MEM,
 	.arg3_type	= ARG_CONST_SIZE_OR_ZERO,
-	.btf_id		= bpf_seq_write_btf_ids,
 };
 
 static __always_inline int
@@ -1130,17 +1126,16 @@ static bool bpf_d_path_allowed(const struct bpf_prog *prog)
 	return btf_id_set_contains(&btf_allowlist_d_path, prog->aux->attach_btf_id);
 }
 
-BTF_ID_LIST(bpf_d_path_btf_ids)
-BTF_ID(struct, path)
+BTF_ID_LIST_SINGLE(bpf_d_path_btf_ids, struct, path)
 
 static const struct bpf_func_proto bpf_d_path_proto = {
 	.func		= bpf_d_path,
 	.gpl_only	= false,
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &bpf_d_path_btf_ids[0],
 	.arg2_type	= ARG_PTR_TO_MEM,
 	.arg3_type	= ARG_CONST_SIZE_OR_ZERO,
-	.btf_id		= bpf_d_path_btf_ids,
 	.allowed	= bpf_d_path_allowed,
 };
 
net/core/bpf_sk_storage.c +2 -6
index 4a86ea34f29e..d653a583dbc9 100644
@@ -378,19 +378,15 @@ const struct bpf_func_proto bpf_sk_storage_delete_proto = {
 	.arg2_type	= ARG_PTR_TO_SOCKET,
 };
 
-BTF_ID_LIST(sk_storage_btf_ids)
-BTF_ID_UNUSED
-BTF_ID(struct, sock)
-
 const struct bpf_func_proto sk_storage_get_btf_proto = {
 	.func		= bpf_sk_storage_get,
 	.gpl_only	= false,
 	.ret_type	= RET_PTR_TO_MAP_VALUE_OR_NULL,
 	.arg1_type	= ARG_CONST_MAP_PTR,
 	.arg2_type	= ARG_PTR_TO_BTF_ID,
+	.arg2_btf_id	= &btf_sock_ids[BTF_SOCK_TYPE_SOCK],
 	.arg3_type	= ARG_PTR_TO_MAP_VALUE_OR_NULL,
 	.arg4_type	= ARG_ANYTHING,
-	.btf_id		= sk_storage_btf_ids,
 };
 
 const struct bpf_func_proto sk_storage_delete_btf_proto = {
@@ -399,7 +395,7 @@ const struct bpf_func_proto sk_storage_delete_btf_proto = {
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_CONST_MAP_PTR,
 	.arg2_type	= ARG_PTR_TO_BTF_ID,
-	.btf_id		= sk_storage_btf_ids,
+	.arg2_btf_id	= &btf_sock_ids[BTF_SOCK_TYPE_SOCK],
 };
 
 struct bpf_sk_storage_diag {
net/core/filter.c +9 -22
index d266c6941967..6014e5f40c58 100644
@@ -3803,19 +3803,18 @@ static const struct bpf_func_proto bpf_skb_event_output_proto = {
 	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
 };
 
-BTF_ID_LIST(bpf_skb_output_btf_ids)
-BTF_ID(struct, sk_buff)
+BTF_ID_LIST_SINGLE(bpf_skb_output_btf_ids, struct, sk_buff)
 
 const struct bpf_func_proto bpf_skb_output_proto = {
 	.func		= bpf_skb_event_output,
 	.gpl_only	= true,
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &bpf_skb_output_btf_ids[0],
 	.arg2_type	= ARG_CONST_MAP_PTR,
 	.arg3_type	= ARG_ANYTHING,
 	.arg4_type	= ARG_PTR_TO_MEM,
 	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
-	.btf_id		= bpf_skb_output_btf_ids,
 };
 
 static unsigned short bpf_tunnel_key_af(u64 flags)
@@ -4199,19 +4198,18 @@ static const struct bpf_func_proto bpf_xdp_event_output_proto = {
 	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
 };
 
-BTF_ID_LIST(bpf_xdp_output_btf_ids)
-BTF_ID(struct, xdp_buff)
+BTF_ID_LIST_SINGLE(bpf_xdp_output_btf_ids, struct, xdp_buff)
 
 const struct bpf_func_proto bpf_xdp_output_proto = {
 	.func		= bpf_xdp_event_output,
 	.gpl_only	= true,
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &bpf_xdp_output_btf_ids[0],
 	.arg2_type	= ARG_CONST_MAP_PTR,
 	.arg3_type	= ARG_ANYTHING,
 	.arg4_type	= ARG_PTR_TO_MEM,
 	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
-	.btf_id		= bpf_xdp_output_btf_ids,
 };
 
 BPF_CALL_1(bpf_get_socket_cookie, struct sk_buff *, skb)
@@ -9897,17 +9895,6 @@ BTF_SOCK_TYPE_xxx
 u32 btf_sock_ids[MAX_BTF_SOCK_TYPE];
 #endif
 
-static bool check_arg_btf_id(u32 btf_id, u32 arg)
-{
-	int i;
-
-	/* only one argument, no need to check arg */
-	for (i = 0; i < MAX_BTF_SOCK_TYPE; i++)
-		if (btf_sock_ids[i] == btf_id)
-			return true;
-	return false;
-}
-
 BPF_CALL_1(bpf_skc_to_tcp6_sock, struct sock *, sk)
 {
 	/* tcp6_sock type is not generated in dwarf and hence btf,
@@ -9926,7 +9913,7 @@ const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto = {
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
 	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.check_btf_id		= check_arg_btf_id,
+	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP6],
 };
 
@@ -9943,7 +9930,7 @@ const struct bpf_func_proto bpf_skc_to_tcp_sock_proto = {
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
 	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.check_btf_id		= check_arg_btf_id,
+	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP],
 };
 
@@ -9967,7 +9954,7 @@ const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto = {
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
 	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.check_btf_id		= check_arg_btf_id,
+	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP_TW],
 };
 
@@ -9991,7 +9978,7 @@ const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto = {
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
 	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.check_btf_id		= check_arg_btf_id,
+	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP_REQ],
 };
 
@@ -10013,6 +10000,6 @@ const struct bpf_func_proto bpf_skc_to_udp6_sock_proto = {
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
 	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.check_btf_id		= check_arg_btf_id,
+	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_UDP6],
 };
net/ipv4/bpf_tcp_ca.c +5 -14
index e3939f76b024..74a2ef598c31 100644
@@ -28,23 +28,18 @@ static u32 unsupported_ops[] = {
 static const struct btf_type *tcp_sock_type;
 static u32 tcp_sock_id, sock_id;
 
-static int btf_sk_storage_get_ids[5];
 static struct bpf_func_proto btf_sk_storage_get_proto __read_mostly;
-
-static int btf_sk_storage_delete_ids[5];
 static struct bpf_func_proto btf_sk_storage_delete_proto __read_mostly;
 
-static void convert_sk_func_proto(struct bpf_func_proto *to, int *to_btf_ids,
-				  const struct bpf_func_proto *from)
+static void convert_sk_func_proto(struct bpf_func_proto *to, const struct bpf_func_proto *from)
 {
 	int i;
 
 	*to = *from;
-	to->btf_id = to_btf_ids;
 	for (i = 0; i < ARRAY_SIZE(to->arg_type); i++) {
 		if (to->arg_type[i] == ARG_PTR_TO_SOCKET) {
 			to->arg_type[i] = ARG_PTR_TO_BTF_ID;
-			to->btf_id[i] = tcp_sock_id;
+			to->arg_btf_id[i] = &tcp_sock_id;
 		}
 	}
 }
@@ -64,12 +59,8 @@ static int bpf_tcp_ca_init(struct btf *btf)
 	tcp_sock_id = type_id;
 	tcp_sock_type = btf_type_by_id(btf, tcp_sock_id);
 
-	convert_sk_func_proto(&btf_sk_storage_get_proto,
-			      btf_sk_storage_get_ids,
-			      &bpf_sk_storage_get_proto);
-	convert_sk_func_proto(&btf_sk_storage_delete_proto,
-			      btf_sk_storage_delete_ids,
-			      &bpf_sk_storage_delete_proto);
+	convert_sk_func_proto(&btf_sk_storage_get_proto, &bpf_sk_storage_get_proto);
+	convert_sk_func_proto(&btf_sk_storage_delete_proto, &bpf_sk_storage_delete_proto);
 
 	return 0;
 }
@@ -185,8 +176,8 @@ static const struct bpf_func_proto bpf_tcp_send_ack_proto = {
 	/* In case we want to report error later */
 	.ret_type	= RET_INTEGER,
 	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &tcp_sock_id,
 	.arg2_type	= ARG_ANYTHING,
-	.btf_id		= &tcp_sock_id,
 };
 
 static const struct bpf_func_proto *
Signature · 1 line
2.25.1
#5 lore raw

PATCH v4 5/11 bpf: make BTF pointer type checking generic

In reply to Lorenz Bauer

Perform BTF type checks if the register we're working on contains a BTF pointer, rather than if the argument is for a BTF pointer. This is easier to understand, and allows removing the code from the arg_type checking section of the function.

---
 kernel/bpf/verifier.c | 38 ++++++++++++++++++++------------------
 1 file changed, 20 insertions(+), 18 deletions(-)
kernel/bpf/verifier.c +20 -18
index 9f95d6d55c5f..99c0d7adcb1e 100644
@@ -4001,27 +4001,9 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 				goto err_type;
 		}
 	} else if (arg_type == ARG_PTR_TO_BTF_ID) {
-		const u32 *btf_id = fn->arg_btf_id[arg];
-
 		expected_type = PTR_TO_BTF_ID;
 		if (type != expected_type)
 			goto err_type;
-
-		if (!btf_id) {
-			verbose(env, "verifier internal error: missing BTF ID\n");
-			return -EFAULT;
-		}
-
-		if (!btf_struct_ids_match(&env->log, reg->off, reg->btf_id, *btf_id)) {
-			verbose(env, "R%d is of type %s but %s is expected\n",
-				regno, kernel_type_name(reg->btf_id), kernel_type_name(*btf_id));
-			return -EACCES;
-		}
-		if (!tnum_is_const(reg->var_off) || reg->var_off.value) {
-			verbose(env, "R%d is a pointer to in-kernel struct with non-zero offset\n",
-				regno);
-			return -EACCES;
-		}
 	} else if (arg_type == ARG_PTR_TO_SPIN_LOCK) {
 		if (meta->func_id == BPF_FUNC_spin_lock) {
 			if (process_spin_lock(env, regno, true))
@@ -4076,6 +4058,26 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		return -EFAULT;
 	}
 
+	if (type == PTR_TO_BTF_ID) {
+		const u32 *btf_id = fn->arg_btf_id[arg];
+
+		if (!btf_id) {
+			verbose(env, "verifier internal error: missing BTF ID\n");
+			return -EFAULT;
+		}
+
+		if (!btf_struct_ids_match(&env->log, reg->off, reg->btf_id, *btf_id)) {
+			verbose(env, "R%d is of type %s but %s is expected\n",
+				regno, kernel_type_name(reg->btf_id), kernel_type_name(*btf_id));
+			return -EACCES;
+		}
+		if (!tnum_is_const(reg->var_off) || reg->var_off.value) {
+			verbose(env, "R%d is a pointer to in-kernel struct with non-zero offset\n",
+				regno);
+			return -EACCES;
+		}
+	}
+
 	if (arg_type == ARG_CONST_MAP_PTR) {
 		/* bpf_map_xxx(map_ptr) call: remember that map_ptr */
 		meta->map_ptr = reg->map_ptr;
Signature · 1 line
2.25.1
#6 lore raw

PATCH v4 6/11 bpf: make reference tracking generic

In reply to Lorenz Bauer

Instead of dealing with reg->ref_obj_id individually for every arg type that needs it, rely on the fact that ref_obj_id is zero if the register is not reference tracked.

---
 kernel/bpf/verifier.c | 26 ++++++++++----------------
 1 file changed, 10 insertions(+), 16 deletions(-)
kernel/bpf/verifier.c +10 -16
index 99c0d7adcb1e..0f7a9c65db5c 100644
@@ -3983,15 +3983,6 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		/* Any sk pointer can be ARG_PTR_TO_SOCK_COMMON */
 		if (!type_is_sk_pointer(type))
 			goto err_type;
-		if (reg->ref_obj_id) {
-			if (meta->ref_obj_id) {
-				verbose(env, "verifier internal error: more than one arg with ref_obj_id R%d %u %u\n",
-					regno, reg->ref_obj_id,
-					meta->ref_obj_id);
-				return -EFAULT;
-			}
-			meta->ref_obj_id = reg->ref_obj_id;
-		}
 	} else if (arg_type == ARG_PTR_TO_SOCKET ||
 		   arg_type == ARG_PTR_TO_SOCKET_OR_NULL) {
 		expected_type = PTR_TO_SOCKET;
@@ -4040,13 +4031,6 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 			/* final test in check_stack_boundary() */;
 		else if (type != expected_type)
 			goto err_type;
-		if (meta->ref_obj_id) {
-			verbose(env, "verifier internal error: more than one arg with ref_obj_id R%d %u %u\n",
-				regno, reg->ref_obj_id,
-				meta->ref_obj_id);
-			return -EFAULT;
-		}
-		meta->ref_obj_id = reg->ref_obj_id;
 	} else if (arg_type_is_int_ptr(arg_type)) {
 		expected_type = PTR_TO_STACK;
 		if (!type_is_pkt_pointer(type) &&
@@ -4078,6 +4062,16 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		}
 	}
 
+	if (reg->ref_obj_id) {
+		if (meta->ref_obj_id) {
+			verbose(env, "verifier internal error: more than one arg with ref_obj_id R%d %u %u\n",
+				regno, reg->ref_obj_id,
+				meta->ref_obj_id);
+			return -EFAULT;
+		}
+		meta->ref_obj_id = reg->ref_obj_id;
+	}
+
 	if (arg_type == ARG_CONST_MAP_PTR) {
 		/* bpf_map_xxx(map_ptr) call: remember that map_ptr */
 		meta->map_ptr = reg->map_ptr;
Signature · 1 line
2.25.1
#7 lore raw

PATCH v4 7/11 bpf: make context access check generic

In reply to Lorenz Bauer

Always check context access if the register we're operating on is PTR_TO_CTX, rather than relying on ARG_PTR_TO_CTX. This allows simplifying the arg_type checking section of the function.

---
 kernel/bpf/verifier.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
kernel/bpf/verifier.c +4 -3
index 0f7a9c65db5c..7d0f9ba18916 100644
@@ -3974,9 +3974,6 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		      arg_type == ARG_PTR_TO_CTX_OR_NULL)) {
 			if (type != expected_type)
 				goto err_type;
-			err = check_ctx_reg(env, reg, regno);
-			if (err < 0)
-				return err;
 		}
 	} else if (arg_type == ARG_PTR_TO_SOCK_COMMON) {
 		expected_type = PTR_TO_SOCK_COMMON;
@@ -4060,6 +4057,10 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 				regno);
 			return -EACCES;
 		}
+	} else if (type == PTR_TO_CTX) {
+		err = check_ctx_reg(env, reg, regno);
+		if (err < 0)
+			return err;
 	}
 
 	if (reg->ref_obj_id) {
Signature · 1 line
2.25.1
#8 lore raw

PATCH v4 8/11 bpf: set meta->raw_mode for pointers close to use

In reply to Lorenz Bauer

If we encounter a pointer to memory, we set meta->raw_mode depending on the type of memory we point at. What isn't obvious is that this information is only used when the next memory size argument is encountered.

Move the assignment closer to where it's used, and add a comment that explains what is going on.

---
 kernel/bpf/verifier.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
kernel/bpf/verifier.c +5 -1
index 7d0f9ba18916..e09eedb30117 100644
@@ -4020,7 +4020,6 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 			 type != PTR_TO_RDWR_BUF &&
 			 type != expected_type)
 			goto err_type;
-		meta->raw_mode = arg_type == ARG_PTR_TO_UNINIT_MEM;
 	} else if (arg_type_is_alloc_mem_ptr(arg_type)) {
 		expected_type = PTR_TO_MEM;
 		if (register_is_null(reg) &&
@@ -4109,6 +4108,11 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		err = check_helper_mem_access(env, regno,
 					      meta->map_ptr->value_size, false,
 					      meta);
+	} else if (arg_type_is_mem_ptr(arg_type)) {
+		/* The access to this pointer is only checked when we hit the
+		 * next is_mem_size argument below.
+		 */
+		meta->raw_mode = (arg_type == ARG_PTR_TO_UNINIT_MEM);
 	} else if (arg_type_is_mem_size(arg_type)) {
 		bool zero_size_allowed = (arg_type == ARG_CONST_SIZE_OR_ZERO);
 
Signature · 1 line
2.25.1
#9 lore raw

PATCH v4 9/11 bpf: check ARG_PTR_TO_SPINLOCK register type in check_func_arg

In reply to Lorenz Bauer

Move the check for PTR_TO_MAP_VALUE to check_func_arg, where all other checking is done as well. Move the invocation of process_spin_lock away from the register type checking, to allow a future refactoring.

---
 kernel/bpf/verifier.c | 28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)
kernel/bpf/verifier.c +14 -14
index e09eedb30117..eefc8256df1c 100644
@@ -3781,10 +3781,6 @@ static int process_spin_lock(struct bpf_verifier_env *env, int regno,
 	struct bpf_map *map = reg->map_ptr;
 	u64 val = reg->var_off.value;
 
-	if (reg->type != PTR_TO_MAP_VALUE) {
-		verbose(env, "R%d is not a pointer to map_value\n", regno);
-		return -EINVAL;
-	}
 	if (!is_const) {
 		verbose(env,
 			"R%d doesn't have constant offset. bpf_spin_lock has to be at the constant offset\n",
@@ -3993,16 +3989,9 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		if (type != expected_type)
 			goto err_type;
 	} else if (arg_type == ARG_PTR_TO_SPIN_LOCK) {
-		if (meta->func_id == BPF_FUNC_spin_lock) {
-			if (process_spin_lock(env, regno, true))
-				return -EACCES;
-		} else if (meta->func_id == BPF_FUNC_spin_unlock) {
-			if (process_spin_lock(env, regno, false))
-				return -EACCES;
-		} else {
-			verbose(env, "verifier internal error\n");
-			return -EFAULT;
-		}
+		expected_type = PTR_TO_MAP_VALUE;
+		if (type != expected_type)
+			goto err_type;
 	} else if (arg_type_is_mem_ptr(arg_type)) {
 		expected_type = PTR_TO_STACK;
 		/* One exception here. In case function allows for NULL to be
@@ -4108,6 +4097,17 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 		err = check_helper_mem_access(env, regno,
 					      meta->map_ptr->value_size, false,
 					      meta);
+	} else if (arg_type == ARG_PTR_TO_SPIN_LOCK) {
+		if (meta->func_id == BPF_FUNC_spin_lock) {
+			if (process_spin_lock(env, regno, true))
+				return -EACCES;
+		} else if (meta->func_id == BPF_FUNC_spin_unlock) {
+			if (process_spin_lock(env, regno, false))
+				return -EACCES;
+		} else {
+			verbose(env, "verifier internal error\n");
+			return -EFAULT;
+		}
 	} else if (arg_type_is_mem_ptr(arg_type)) {
 		/* The access to this pointer is only checked when we hit the
 		 * next is_mem_size argument below.
Signature · 1 line
2.25.1
#10 lore raw

PATCH v4 10/11 bpf: hoist type checking for nullable arg types

In reply to Lorenz Bauer

check_func_arg has a plethora of weird if statements with empty branches. They work around the fact that *_OR_NULL argument types should accept a SCALAR_VALUE register, as long as it's value is 0. These statements make it difficult to reason about the type checking logic.

Instead, skip more detailed type checking logic iff the register is 0, and the function expects a nullable type. This allows simplifying the type checking itself.

---
 kernel/bpf/verifier.c | 64 ++++++++++++++++++++-----------------------
 1 file changed, 30 insertions(+), 34 deletions(-)
kernel/bpf/verifier.c +30 -34
index eefc8256df1c..d64ac79982ad 100644
@@ -435,6 +435,15 @@ static bool arg_type_may_be_refcounted(enum bpf_arg_type type)
 	return type == ARG_PTR_TO_SOCK_COMMON;
 }
 
+static bool arg_type_may_be_null(enum bpf_arg_type type)
+{
+	return type == ARG_PTR_TO_MAP_VALUE_OR_NULL ||
+	       type == ARG_PTR_TO_MEM_OR_NULL ||
+	       type == ARG_PTR_TO_CTX_OR_NULL ||
+	       type == ARG_PTR_TO_SOCKET_OR_NULL ||
+	       type == ARG_PTR_TO_ALLOC_MEM_OR_NULL;
+}
+
 /* Determine whether the function releases some resources allocated by another
  * function call. The first reference type argument will be assumed to be
  * released by release_reference().
@@ -3941,17 +3950,20 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 			return err;
 	}
 
+	if (register_is_null(reg) && arg_type_may_be_null(arg_type))
+		/* A NULL register has a SCALAR_VALUE type, so skip
+		 * type checking.
+		 */
+		goto skip_type_check;
+
 	if (arg_type == ARG_PTR_TO_MAP_KEY ||
 	    arg_type == ARG_PTR_TO_MAP_VALUE ||
 	    arg_type == ARG_PTR_TO_UNINIT_MAP_VALUE ||
 	    arg_type == ARG_PTR_TO_MAP_VALUE_OR_NULL) {
 		expected_type = PTR_TO_STACK;
-		if (register_is_null(reg) &&
-		    arg_type == ARG_PTR_TO_MAP_VALUE_OR_NULL)
-			/* final test in check_stack_boundary() */;
-		else if (!type_is_pkt_pointer(type) &&
-			 type != PTR_TO_MAP_VALUE &&
-			 type != expected_type)
+		if (!type_is_pkt_pointer(type) &&
+		    type != PTR_TO_MAP_VALUE &&
+		    type != expected_type)
 			goto err_type;
 	} else if (arg_type == ARG_CONST_SIZE ||
 		   arg_type == ARG_CONST_SIZE_OR_ZERO ||
@@ -3966,11 +3978,8 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 	} else if (arg_type == ARG_PTR_TO_CTX ||
 		   arg_type == ARG_PTR_TO_CTX_OR_NULL) {
 		expected_type = PTR_TO_CTX;
-		if (!(register_is_null(reg) &&
-		      arg_type == ARG_PTR_TO_CTX_OR_NULL)) {
-			if (type != expected_type)
-				goto err_type;
-		}
+		if (type != expected_type)
+			goto err_type;
 	} else if (arg_type == ARG_PTR_TO_SOCK_COMMON) {
 		expected_type = PTR_TO_SOCK_COMMON;
 		/* Any sk pointer can be ARG_PTR_TO_SOCK_COMMON */
@@ -3979,11 +3988,8 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 	} else if (arg_type == ARG_PTR_TO_SOCKET ||
 		   arg_type == ARG_PTR_TO_SOCKET_OR_NULL) {
 		expected_type = PTR_TO_SOCKET;
-		if (!(register_is_null(reg) &&
-		      arg_type == ARG_PTR_TO_SOCKET_OR_NULL)) {
-			if (type != expected_type)
-				goto err_type;
-		}
+		if (type != expected_type)
+			goto err_type;
 	} else if (arg_type == ARG_PTR_TO_BTF_ID) {
 		expected_type = PTR_TO_BTF_ID;
 		if (type != expected_type)
@@ -3994,27 +4000,16 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 			goto err_type;
 	} else if (arg_type_is_mem_ptr(arg_type)) {
 		expected_type = PTR_TO_STACK;
-		/* One exception here. In case function allows for NULL to be
-		 * passed in as argument, it's a SCALAR_VALUE type. Final test
-		 * happens during stack boundary checking.
-		 */
-		if (register_is_null(reg) &&
-		    (arg_type == ARG_PTR_TO_MEM_OR_NULL ||
-		     arg_type == ARG_PTR_TO_ALLOC_MEM_OR_NULL))
-			/* final test in check_stack_boundary() */;
-		else if (!type_is_pkt_pointer(type) &&
-			 type != PTR_TO_MAP_VALUE &&
-			 type != PTR_TO_MEM &&
-			 type != PTR_TO_RDONLY_BUF &&
-			 type != PTR_TO_RDWR_BUF &&
-			 type != expected_type)
+		if (!type_is_pkt_pointer(type) &&
+		    type != PTR_TO_MAP_VALUE &&
+		    type != PTR_TO_MEM &&
+		    type != PTR_TO_RDONLY_BUF &&
+		    type != PTR_TO_RDWR_BUF &&
+		    type != expected_type)
 			goto err_type;
 	} else if (arg_type_is_alloc_mem_ptr(arg_type)) {
 		expected_type = PTR_TO_MEM;
-		if (register_is_null(reg) &&
-		    arg_type == ARG_PTR_TO_ALLOC_MEM_OR_NULL)
-			/* final test in check_stack_boundary() */;
-		else if (type != expected_type)
+		if (type != expected_type)
 			goto err_type;
 	} else if (arg_type_is_int_ptr(arg_type)) {
 		expected_type = PTR_TO_STACK;
@@ -4051,6 +4046,7 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 			return err;
 	}
 
+skip_type_check:
 	if (reg->ref_obj_id) {
 		if (meta->ref_obj_id) {
 			verbose(env, "verifier internal error: more than one arg with ref_obj_id R%d %u %u\n",
Signature · 1 line
2.25.1
#11 lore raw

PATCH v4 11/11 bpf: use a table to drive helper arg type checks

In reply to Lorenz Bauer

The mapping between bpf_arg_type and bpf_reg_type is encoded in a big hairy if statement that is hard to follow. The debug output also leaves to be desired: if a reg_type doesn't match we only print one of the options, instead printing all the valid ones.

Convert the if statement into a table which is then used to drive type checking. If none of the reg_types match we print all options, e.g.:

R2 type=rdonly_buf expected=fp, pkt, pkt_meta, map_value

---
 include/linux/bpf.h   |   1 +
 kernel/bpf/verifier.c | 183 +++++++++++++++++++++++++-----------------
 2 files changed, 110 insertions(+), 74 deletions(-)
include/linux/bpf.h +1 -0
index 732e2f144b6d..4ca817a9195c 100644
@@ -292,6 +292,7 @@ enum bpf_arg_type {
 	ARG_PTR_TO_ALLOC_MEM,	/* pointer to dynamically allocated memory */
 	ARG_PTR_TO_ALLOC_MEM_OR_NULL,	/* pointer to dynamically allocated memory or NULL */
 	ARG_CONST_ALLOC_SIZE_OR_ZERO,	/* number of allocated bytes requested */
+	__BPF_ARG_TYPE_MAX,
 };
 
 /* type of values returned from helper functions */
kernel/bpf/verifier.c +109 -74

217 lines not on this page · raw message

Signature · 1 line
2.25.1
#12 lore raw

Re: [PATCH bpf-next v4 11/11] bpf: use a table to drive helper arg type checks

Review

Quoted 3 lines · Lorenz Bauer
(quoted diff) +3 -0
+struct bpf_reg_types {
+	const enum bpf_reg_type types[10];
+};

any idea on how to make it more robust?

Quoted 15 lines
(quoted diff) +15 -0
+
+static const struct bpf_reg_types *compatible_reg_types[] = {
+	[ARG_PTR_TO_MAP_KEY]		= &map_key_value_types,
+	[ARG_PTR_TO_MAP_VALUE]		= &map_key_value_types,
+	[ARG_PTR_TO_UNINIT_MAP_VALUE]	= &map_key_value_types,
+	[ARG_PTR_TO_MAP_VALUE_OR_NULL]	= &map_key_value_types,
+	[ARG_CONST_SIZE]		= &scalar_types,
+	[ARG_CONST_SIZE_OR_ZERO]	= &scalar_types,
+	[ARG_CONST_ALLOC_SIZE_OR_ZERO]	= &scalar_types,
+	[ARG_CONST_MAP_PTR]		= &const_map_ptr_types,
+	[ARG_PTR_TO_CTX]		= &context_types,
+	[ARG_PTR_TO_CTX_OR_NULL]	= &context_types,
+	[ARG_PTR_TO_SOCK_COMMON]	= &sock_types,
+	[ARG_PTR_TO_SOCKET]		= &fullsock_types,
+	[ARG_PTR_TO_SOCKET_OR_NULL]	= &fullsock_types,

Review

Quoted 10 lines
(quoted diff) +10 -0
+	[ARG_PTR_TO_BTF_ID]		= &btf_ptr_types,
+	[ARG_PTR_TO_SPIN_LOCK]		= &spin_lock_types,
+	[ARG_PTR_TO_MEM]		= &mem_types,
+	[ARG_PTR_TO_MEM_OR_NULL]	= &mem_types,
+	[ARG_PTR_TO_UNINIT_MEM]		= &mem_types,
+	[ARG_PTR_TO_ALLOC_MEM]		= &alloc_mem_types,
+	[ARG_PTR_TO_ALLOC_MEM_OR_NULL]	= &alloc_mem_types,
+	[ARG_PTR_TO_INT]		= &int_ptr_types,
+	[ARG_PTR_TO_LONG]		= &int_ptr_types,
+	[__BPF_ARG_TYPE_MAX]		= NULL,

I don't understand what this extra value is for.
I tried:

include/linux/bpf.h +0 -1
index fc5c901c7542..87b0d5dcc1ff 100644
@@ -292,7 +292,6 @@ enum bpf_arg_type {
        ARG_PTR_TO_ALLOC_MEM,   /* pointer to dynamically allocated memory */
        ARG_PTR_TO_ALLOC_MEM_OR_NULL,   /* pointer to dynamically allocated memory or NULL */
        ARG_CONST_ALLOC_SIZE_OR_ZERO,   /* number of allocated bytes requested */
-       __BPF_ARG_TYPE_MAX,
 };
 
 /* type of values returned from helper functions */
kernel/bpf/verifier.c +0 -1
index 15ab889b0a3f..83faa67858b6 100644
@@ -4025,7 +4025,6 @@ static const struct bpf_reg_types *compatible_reg_types[] = {
        [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
        [ARG_PTR_TO_INT]                = &int_ptr_types,
        [ARG_PTR_TO_LONG]               = &int_ptr_types,
-       [__BPF_ARG_TYPE_MAX]            = NULL,
 };
 

and everything is fine as I think it should be.

Review

Quoted 5 lines
(quoted diff) +3 -0
+	compatible = compatible_reg_types[arg_type];
+	if (!compatible) {
+		verbose(env, "verifier internal error: unsupported arg type %d\n", arg_type);
 		return -EFAULT;
 	}

This check will trigger the same way when somebody adds new ARG_* and doesn't add to the table.

Quoted 9 lines
(quoted diff) +4 -0
+	err = check_reg_type(env, regno, compatible);
+	if (err)
+		return err;
+
 	if (type == PTR_TO_BTF_ID) {
 		const u32 *btf_id = fn->arg_btf_id[arg];

Review static int check_func_arg(struct bpf_verifier_env *env, u32 arg, · line 4180

Quoted 7 lines
(quoted diff) +0 -4
@@ -4174,10 +4213,6 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
 	}
 
 	return err;
-err_type:
-	verbose(env, "R%d type=%s expected=%s\n", regno,
-		reg_type_str[type], reg_type_str[expected_type]);
-	return -EACCES;

I'm not a fan of table driven checks. I think one explicit switch statement would have been easier to read, but I guess we can convert back to it later if table becomes too limiting. The improvement in the verifier output is important and justifies this approach.

Applied to bpf-next. Thanks!

#13 lore raw

Re: [PATCH bpf-next v4 11/11] bpf: use a table to drive helper arg type checks

Quoted 7 lines · Alexei Starovoitov

Review

Quoted 3 lines · Lorenz Bauer
(quoted diff) +3 -0
+struct bpf_reg_types {
+     const enum bpf_reg_type types[10];
+};

any idea on how to make it more robust?

I kind of copied this from the bpf_iter context. I prototyped using an enum bpf_reg_type * and then terminating the array with NOT_INIT. Writing this out is more involved, and might need some macro magic to make it palatable. The current approach is a lot simpler, and I figured that the compiler will error out if we ever exceed the 10 items.

Quoted 54 lines
Quoted 16 lines
(quoted diff) +15 -0
+
+static const struct bpf_reg_types *compatible_reg_types[] = {
+     [ARG_PTR_TO_MAP_KEY]            = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE]          = &map_key_value_types,
+     [ARG_PTR_TO_UNINIT_MAP_VALUE]   = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE_OR_NULL]  = &map_key_value_types,
+     [ARG_CONST_SIZE]                = &scalar_types,
+     [ARG_CONST_SIZE_OR_ZERO]        = &scalar_types,
+     [ARG_CONST_ALLOC_SIZE_OR_ZERO]  = &scalar_types,
+     [ARG_CONST_MAP_PTR]             = &const_map_ptr_types,
+     [ARG_PTR_TO_CTX]                = &context_types,
+     [ARG_PTR_TO_CTX_OR_NULL]        = &context_types,
+     [ARG_PTR_TO_SOCK_COMMON]        = &sock_types,
+     [ARG_PTR_TO_SOCKET]             = &fullsock_types,
+     [ARG_PTR_TO_SOCKET_OR_NULL]     = &fullsock_types,

Review

Quoted 10 lines
(quoted diff) +10 -0
+     [ARG_PTR_TO_BTF_ID]             = &btf_ptr_types,
+     [ARG_PTR_TO_SPIN_LOCK]          = &spin_lock_types,
+     [ARG_PTR_TO_MEM]                = &mem_types,
+     [ARG_PTR_TO_MEM_OR_NULL]        = &mem_types,
+     [ARG_PTR_TO_UNINIT_MEM]         = &mem_types,
+     [ARG_PTR_TO_ALLOC_MEM]          = &alloc_mem_types,
+     [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
+     [ARG_PTR_TO_INT]                = &int_ptr_types,
+     [ARG_PTR_TO_LONG]               = &int_ptr_types,
+     [__BPF_ARG_TYPE_MAX]            = NULL,

I don't understand what this extra value is for.
I tried:

include/linux/bpf.h +0 -1
index fc5c901c7542..87b0d5dcc1ff 100644
@@ -292,7 +292,6 @@ enum bpf_arg_type {
        ARG_PTR_TO_ALLOC_MEM,   /* pointer to dynamically allocated memory */
        ARG_PTR_TO_ALLOC_MEM_OR_NULL,   /* pointer to dynamically allocated memory or NULL */
        ARG_CONST_ALLOC_SIZE_OR_ZERO,   /* number of allocated bytes requested */
-       __BPF_ARG_TYPE_MAX,
 };
 
 /* type of values returned from helper functions */
kernel/bpf/verifier.c +0 -1
index 15ab889b0a3f..83faa67858b6 100644
@@ -4025,7 +4025,6 @@ static const struct bpf_reg_types *compatible_reg_types[] = {
        [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
        [ARG_PTR_TO_INT]                = &int_ptr_types,
        [ARG_PTR_TO_LONG]               = &int_ptr_types,
-       [__BPF_ARG_TYPE_MAX]            = NULL,
 };
 

and everything is fine as I think it should be.

Quoted 7 lines

Review

Quoted 5 lines
(quoted diff) +3 -0
+     compatible = compatible_reg_types[arg_type];
+     if (!compatible) {
+             verbose(env, "verifier internal error: unsupported arg type %d\n", arg_type);
              return -EFAULT;
      }

This check will trigger the same way when somebody adds new ARG_* and doesn't add to the table.

I think in that case that value of compatible will be undefined, since it points past the end of compatible_reg_types. Hence the __BPF_ARG_TYPE_MAX to ensure that the array has a NULL slot for new arg types.

Quoted 23 lines
Quoted 10 lines
(quoted diff) +4 -0
+     err = check_reg_type(env, regno, compatible);
+     if (err)
+             return err;
+
      if (type == PTR_TO_BTF_ID) {
              const u32 *btf_id = fn->arg_btf_id[arg];

Review static int check_func_arg(struct bpf_verifier_env *env, u32 arg, · line 4180

Quoted 7 lines
(quoted diff) +0 -4
@@ -4174,10 +4213,6 @@ static int check_func_arg(struct bpf_verifier_env *env, u32 arg,
      }
 
      return err;
-err_type:
-     verbose(env, "R%d type=%s expected=%s\n", regno,
-             reg_type_str[type], reg_type_str[expected_type]);
-     return -EACCES;

I'm not a fan of table driven checks. I think one explicit switch statement would have been easier to read, but I guess we can convert back to it later if table becomes too limiting. The improvement in the verifier output is important and justifies this approach.

Quoted 1 line

Applied to bpf-next. Thanks!

Thank you!

--
Lorenz Bauer | Systems Engineer
6th Floor, County Hall/The Riverside Building, SE1 7PB, UK

www.cloudflare.com

#14 lore raw

Re: [PATCH bpf-next v4 11/11] bpf: use a table to drive helper arg type checks

Quoted 10 lines · Lorenz Bauer
Quoted 7 lines · Alexei Starovoitov

Review

Quoted 3 lines · Lorenz Bauer
(quoted diff) +3 -0
+struct bpf_reg_types {
+     const enum bpf_reg_type types[10];
+};

any idea on how to make it more robust?

Quoted 6 lines

I kind of copied this from the bpf_iter context. I prototyped using an enum bpf_reg_type * and then terminating the array with NOT_INIT. Writing this out is more involved, and might need some macro magic to make it palatable. The current approach is a lot simpler, and I figured that the compiler will error out if we ever exceed the 10 items.

The compiler will be silent if number of types is exactly 10, but at run-time the loop will access out of bounds.

Quoted 60 lines
Quoted 60 lines
Quoted 16 lines
(quoted diff) +15 -0
+
+static const struct bpf_reg_types *compatible_reg_types[] = {
+     [ARG_PTR_TO_MAP_KEY]            = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE]          = &map_key_value_types,
+     [ARG_PTR_TO_UNINIT_MAP_VALUE]   = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE_OR_NULL]  = &map_key_value_types,
+     [ARG_CONST_SIZE]                = &scalar_types,
+     [ARG_CONST_SIZE_OR_ZERO]        = &scalar_types,
+     [ARG_CONST_ALLOC_SIZE_OR_ZERO]  = &scalar_types,
+     [ARG_CONST_MAP_PTR]             = &const_map_ptr_types,
+     [ARG_PTR_TO_CTX]                = &context_types,
+     [ARG_PTR_TO_CTX_OR_NULL]        = &context_types,
+     [ARG_PTR_TO_SOCK_COMMON]        = &sock_types,
+     [ARG_PTR_TO_SOCKET]             = &fullsock_types,
+     [ARG_PTR_TO_SOCKET_OR_NULL]     = &fullsock_types,

Review

Quoted 10 lines
(quoted diff) +10 -0
+     [ARG_PTR_TO_BTF_ID]             = &btf_ptr_types,
+     [ARG_PTR_TO_SPIN_LOCK]          = &spin_lock_types,
+     [ARG_PTR_TO_MEM]                = &mem_types,
+     [ARG_PTR_TO_MEM_OR_NULL]        = &mem_types,
+     [ARG_PTR_TO_UNINIT_MEM]         = &mem_types,
+     [ARG_PTR_TO_ALLOC_MEM]          = &alloc_mem_types,
+     [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
+     [ARG_PTR_TO_INT]                = &int_ptr_types,
+     [ARG_PTR_TO_LONG]               = &int_ptr_types,
+     [__BPF_ARG_TYPE_MAX]            = NULL,

I don't understand what this extra value is for.
I tried:

include/linux/bpf.h +0 -1
index fc5c901c7542..87b0d5dcc1ff 100644
@@ -292,7 +292,6 @@ enum bpf_arg_type {
        ARG_PTR_TO_ALLOC_MEM,   /* pointer to dynamically allocated memory */
        ARG_PTR_TO_ALLOC_MEM_OR_NULL,   /* pointer to dynamically allocated memory or NULL */
        ARG_CONST_ALLOC_SIZE_OR_ZERO,   /* number of allocated bytes requested */
-       __BPF_ARG_TYPE_MAX,
 };
 
 /* type of values returned from helper functions */
kernel/bpf/verifier.c +0 -1
index 15ab889b0a3f..83faa67858b6 100644
@@ -4025,7 +4025,6 @@ static const struct bpf_reg_types *compatible_reg_types[] = {
        [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
        [ARG_PTR_TO_INT]                = &int_ptr_types,
        [ARG_PTR_TO_LONG]               = &int_ptr_types,
-       [__BPF_ARG_TYPE_MAX]            = NULL,
 };
 

and everything is fine as I think it should be.

Quoted 5 lines
(quoted diff) +3 -0
+     compatible = compatible_reg_types[arg_type];
+     if (!compatible) {
+             verbose(env, "verifier internal error: unsupported arg type %d\n", arg_type);
              return -EFAULT;
      }
Quoted 6 lines
Quoted 1 line

This check will trigger the same way when somebody adds new ARG_* and doesn't add to the table.

I think in that case that value of compatible will be undefined, since it points past the end of compatible_reg_types. Hence the __BPF_ARG_TYPE_MAX to ensure that the array has a NULL slot for new arg types.

I still don't see a point.
If anyone adds one more ARG_ to the end (or anywhere else)
the compatible_reg_types array will be zero inited in that place by the compiler. Just like it does already for ARG_ANYTHING and ARG_DONTCARE.
Am I still missing something?
If not please follow up with removal of __BPF_ARG_TYPE_MAX.

#15 lore raw

Re: [PATCH bpf-next v4 11/11] bpf: use a table to drive helper arg type checks

Quoted 19 lines · Alexei Starovoitov
Quoted 10 lines · Lorenz Bauer
Quoted 7 lines · Alexei Starovoitov

Review

Quoted 3 lines · Lorenz Bauer
(quoted diff) +3 -0
+struct bpf_reg_types {
+     const enum bpf_reg_type types[10];
+};

any idea on how to make it more robust?

Quoted 6 lines

I kind of copied this from the bpf_iter context. I prototyped using an enum bpf_reg_type * and then terminating the array with NOT_INIT. Writing this out is more involved, and might need some macro magic to make it palatable. The current approach is a lot simpler, and I figured that the compiler will error out if we ever exceed the 10 items.

Quoted 2 lines

The compiler will be silent if number of types is exactly 10, but at run-time the loop will access out of bounds.

Which loop do you refer to?

The one in check_reg_type shouldn't go out of bounds due to ARRAY_SIZE:

    for (i = 0; i < ARRAY_SIZE(compatible->types); i++) {
        expected = compatible->types[i];
        if (expected == NOT_INIT)
            break;
Quoted 68 lines
Quoted 61 lines
Quoted 61 lines
Quoted 17 lines
(quoted diff) +15 -0
+
+static const struct bpf_reg_types *compatible_reg_types[] = {
+     [ARG_PTR_TO_MAP_KEY]            = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE]          = &map_key_value_types,
+     [ARG_PTR_TO_UNINIT_MAP_VALUE]   = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE_OR_NULL]  = &map_key_value_types,
+     [ARG_CONST_SIZE]                = &scalar_types,
+     [ARG_CONST_SIZE_OR_ZERO]        = &scalar_types,
+     [ARG_CONST_ALLOC_SIZE_OR_ZERO]  = &scalar_types,
+     [ARG_CONST_MAP_PTR]             = &const_map_ptr_types,
+     [ARG_PTR_TO_CTX]                = &context_types,
+     [ARG_PTR_TO_CTX_OR_NULL]        = &context_types,
+     [ARG_PTR_TO_SOCK_COMMON]        = &sock_types,
+     [ARG_PTR_TO_SOCKET]             = &fullsock_types,
+     [ARG_PTR_TO_SOCKET_OR_NULL]     = &fullsock_types,

Review

Quoted 10 lines
(quoted diff) +10 -0
+     [ARG_PTR_TO_BTF_ID]             = &btf_ptr_types,
+     [ARG_PTR_TO_SPIN_LOCK]          = &spin_lock_types,
+     [ARG_PTR_TO_MEM]                = &mem_types,
+     [ARG_PTR_TO_MEM_OR_NULL]        = &mem_types,
+     [ARG_PTR_TO_UNINIT_MEM]         = &mem_types,
+     [ARG_PTR_TO_ALLOC_MEM]          = &alloc_mem_types,
+     [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
+     [ARG_PTR_TO_INT]                = &int_ptr_types,
+     [ARG_PTR_TO_LONG]               = &int_ptr_types,
+     [__BPF_ARG_TYPE_MAX]            = NULL,

I don't understand what this extra value is for.
I tried:

include/linux/bpf.h +0 -1
index fc5c901c7542..87b0d5dcc1ff 100644
@@ -292,7 +292,6 @@ enum bpf_arg_type {
        ARG_PTR_TO_ALLOC_MEM,   /* pointer to dynamically allocated memory */
        ARG_PTR_TO_ALLOC_MEM_OR_NULL,   /* pointer to dynamically allocated memory or NULL */
        ARG_CONST_ALLOC_SIZE_OR_ZERO,   /* number of allocated bytes requested */
-       __BPF_ARG_TYPE_MAX,
 };
 
 /* type of values returned from helper functions */
kernel/bpf/verifier.c +0 -1
index 15ab889b0a3f..83faa67858b6 100644
@@ -4025,7 +4025,6 @@ static const struct bpf_reg_types *compatible_reg_types[] = {
        [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
        [ARG_PTR_TO_INT]                = &int_ptr_types,
        [ARG_PTR_TO_LONG]               = &int_ptr_types,
-       [__BPF_ARG_TYPE_MAX]            = NULL,
 };
 

and everything is fine as I think it should be.

Quoted 5 lines
(quoted diff) +3 -0
+     compatible = compatible_reg_types[arg_type];
+     if (!compatible) {
+             verbose(env, "verifier internal error: unsupported arg type %d\n", arg_type);
              return -EFAULT;
      }
Quoted 6 lines
Quoted 1 line

This check will trigger the same way when somebody adds new ARG_* and doesn't add to the table.

I think in that case that value of compatible will be undefined, since it points past the end of compatible_reg_types. Hence the __BPF_ARG_TYPE_MAX to ensure that the array has a NULL slot for new arg types.

Quoted 4 lines

I still don't see a point.
If anyone adds one more ARG_ to the end (or anywhere else)
the compatible_reg_types array will be zero inited in that place by the compiler. Just like it does already for ARG_ANYTHING and ARG_DONTCARE.

I looked up designated initializers when I wrote this, since I wasn't super familiar with them:
https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html#Designated-Inits

Note that the length of the array is the highest value specified plus one.

So ARG_ANYTHING and ARG_DONTCARE are OK since there is a higher enum value present in the initializer. If someone adds a new item to enum bpf_arg_type I assume they would add it to the end. In that case the highest value of the initializer doesn't change, and then indexing into compatible_reg_types with the new enum value would be out of bounds. Adding __BPF_ARG_TYPE_MAX fixes that.

It's very possible I misunderstood how this whole contraption works, happy to send a patch.

Signature · 4 lines
Lorenz Bauer  |  Systems Engineer
6th Floor, County Hall/The Riverside Building, SE1 7PB, UK

www.cloudflare.com
#16 lore raw

Re: [PATCH bpf-next v4 11/11] bpf: use a table to drive helper arg type checks

Quoted 22 lines · Lorenz Bauer
Quoted 19 lines · Alexei Starovoitov
Quoted 16 lines · Lorenz Bauer
Quoted 7 lines · Alexei Starovoitov

Review

Quoted 3 lines · Lorenz Bauer
(quoted diff) +3 -0
+struct bpf_reg_types {
+     const enum bpf_reg_type types[10];
+};

any idea on how to make it more robust?

I kind of copied this from the bpf_iter context. I prototyped using an enum bpf_reg_type * and then terminating the array with NOT_INIT. Writing this out is more involved, and might need some macro magic to make it palatable. The current approach is a lot simpler, and I figured that the compiler will error out if we ever exceed the 10 items.

Quoted 8 lines
Quoted 2 lines

The compiler will be silent if number of types is exactly 10, but at run-time the loop will access out of bounds.

Which loop do you refer to?

The one in check_reg_type shouldn't go out of bounds due to ARRAY_SIZE:

    for (i = 0; i < ARRAY_SIZE(compatible->types); i++) {

ahh. right. it will always be 10 here. got it.

Quoted 81 lines
        expected = compatible->types[i];
        if (expected == NOT_INIT)
            break;
Quoted 68 lines
Quoted 61 lines
Quoted 61 lines
Quoted 17 lines
(quoted diff) +15 -0
+
+static const struct bpf_reg_types *compatible_reg_types[] = {
+     [ARG_PTR_TO_MAP_KEY]            = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE]          = &map_key_value_types,
+     [ARG_PTR_TO_UNINIT_MAP_VALUE]   = &map_key_value_types,
+     [ARG_PTR_TO_MAP_VALUE_OR_NULL]  = &map_key_value_types,
+     [ARG_CONST_SIZE]                = &scalar_types,
+     [ARG_CONST_SIZE_OR_ZERO]        = &scalar_types,
+     [ARG_CONST_ALLOC_SIZE_OR_ZERO]  = &scalar_types,
+     [ARG_CONST_MAP_PTR]             = &const_map_ptr_types,
+     [ARG_PTR_TO_CTX]                = &context_types,
+     [ARG_PTR_TO_CTX_OR_NULL]        = &context_types,
+     [ARG_PTR_TO_SOCK_COMMON]        = &sock_types,
+     [ARG_PTR_TO_SOCKET]             = &fullsock_types,
+     [ARG_PTR_TO_SOCKET_OR_NULL]     = &fullsock_types,

Review

Quoted 10 lines
(quoted diff) +10 -0
+     [ARG_PTR_TO_BTF_ID]             = &btf_ptr_types,
+     [ARG_PTR_TO_SPIN_LOCK]          = &spin_lock_types,
+     [ARG_PTR_TO_MEM]                = &mem_types,
+     [ARG_PTR_TO_MEM_OR_NULL]        = &mem_types,
+     [ARG_PTR_TO_UNINIT_MEM]         = &mem_types,
+     [ARG_PTR_TO_ALLOC_MEM]          = &alloc_mem_types,
+     [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
+     [ARG_PTR_TO_INT]                = &int_ptr_types,
+     [ARG_PTR_TO_LONG]               = &int_ptr_types,
+     [__BPF_ARG_TYPE_MAX]            = NULL,

I don't understand what this extra value is for.
I tried:

include/linux/bpf.h +0 -1
index fc5c901c7542..87b0d5dcc1ff 100644
@@ -292,7 +292,6 @@ enum bpf_arg_type {
        ARG_PTR_TO_ALLOC_MEM,   /* pointer to dynamically allocated memory */
        ARG_PTR_TO_ALLOC_MEM_OR_NULL,   /* pointer to dynamically allocated memory or NULL */
        ARG_CONST_ALLOC_SIZE_OR_ZERO,   /* number of allocated bytes requested */
-       __BPF_ARG_TYPE_MAX,
 };
 
 /* type of values returned from helper functions */
kernel/bpf/verifier.c +0 -1
index 15ab889b0a3f..83faa67858b6 100644
@@ -4025,7 +4025,6 @@ static const struct bpf_reg_types *compatible_reg_types[] = {
        [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
        [ARG_PTR_TO_INT]                = &int_ptr_types,
        [ARG_PTR_TO_LONG]               = &int_ptr_types,
-       [__BPF_ARG_TYPE_MAX]            = NULL,
 };
 

and everything is fine as I think it should be.

Quoted 5 lines
(quoted diff) +3 -0
+     compatible = compatible_reg_types[arg_type];
+     if (!compatible) {
+             verbose(env, "verifier internal error: unsupported arg type %d\n", arg_type);
              return -EFAULT;
      }
Quoted 6 lines
Quoted 1 line

This check will trigger the same way when somebody adds new ARG_* and doesn't add to the table.

I think in that case that value of compatible will be undefined, since it points past the end of compatible_reg_types. Hence the __BPF_ARG_TYPE_MAX to ensure that the array has a NULL slot for new arg types.

Quoted 4 lines

I still don't see a point.
If anyone adds one more ARG_ to the end (or anywhere else)
the compatible_reg_types array will be zero inited in that place by the compiler. Just like it does already for ARG_ANYTHING and ARG_DONTCARE.

I looked up designated initializers when I wrote this, since I wasn't super familiar with them:
https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html#Designated-Inits

Quoted 8 lines

Note that the length of the array is the highest value specified plus one.

So ARG_ANYTHING and ARG_DONTCARE are OK since there is a higher enum value present in the initializer. If someone adds a new item to enum bpf_arg_type I assume they would add it to the end. In that case the highest value of the initializer doesn't change, and then indexing into compatible_reg_types with the new enum value would be out of bounds. Adding __BPF_ARG_TYPE_MAX fixes that.

I see. Could you do this instead then:
-static const struct bpf_reg_types *compatible_reg_types[] = { +static const struct bpf_reg_types *compatible_reg_types[__BPF_ARG_TYPE_MAX] = {

        [ARG_PTR_TO_MAP_KEY]            = &map_key_value_types,
        [ARG_PTR_TO_MAP_VALUE]          = &map_key_value_types,
        [ARG_PTR_TO_UNINIT_MAP_VALUE]   = &map_key_value_types,
@@ -4025,7 +4025,6 @@ static const struct bpf_reg_types
*compatible_reg_types[] = {
        [ARG_PTR_TO_ALLOC_MEM_OR_NULL]  = &alloc_mem_types,
        [ARG_PTR_TO_INT]                = &int_ptr_types,
        [ARG_PTR_TO_LONG]               = &int_ptr_types,
-       [__BPF_ARG_TYPE_MAX]            = NULL,

};

That way is more obvious.
That =NULL initializer just for the last element and not for ARG_ANYTHING/DONTCARE
bothered me enough to start this whole discussion.