bpf: permits narrower load from bpf program context fields

Currently, verifier will reject a program if it contains an
narrower load from the bpf context structure. For example,
        __u8 h = __sk_buff->hash, or
        __u16 p = __sk_buff->protocol
        __u32 sample_period = bpf_perf_event_data->sample_period
which are narrower loads of 4-byte or 8-byte field.

This patch solves the issue by:
  . Introduce a new parameter ctx_field_size to carry the
    field size of narrower load from prog type
    specific *__is_valid_access validator back to verifier.
  . The non-zero ctx_field_size for a memory access indicates
    (1). underlying prog type specific convert_ctx_accesses
         supporting non-whole-field access
    (2). the current insn is a narrower or whole field access.
  . In verifier, for such loads where load memory size is
    less than ctx_field_size, verifier transforms it
    to a full field load followed by proper masking.
  . Currently, __sk_buff and bpf_perf_event_data->sample_period
    are supporting narrowing loads.
  . Narrower stores are still not allowed as typical ctx stores
    are just normal stores.

Because of this change, some tests in verifier will fail and
these tests are removed. As a bonus, rename some out of bound
__sk_buff->cb access to proper field name and remove two
redundant "skb cb oob" tests.
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/bpf.h

@@ -157,7 +157,7 @@ struct bpf_verifier_ops {
 	 * with 'type' (read or write) is allowed
 	 */
 	bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
-				enum bpf_reg_type *reg_type);
+				enum bpf_reg_type *reg_type, int *ctx_field_size);
 	int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
 			    const struct bpf_prog *prog);
 	u32 (*convert_ctx_access)(enum bpf_access_type type,

include/linux/bpf_verifier.h

@@ -73,6 +73,7 @@ struct bpf_insn_aux_data {
 		enum bpf_reg_type ptr_type;	/* pointer type for load/store insns */
 		struct bpf_map *map_ptr;	/* pointer for call insn into lookup_elem */
 	};
+	int ctx_field_size; /* the ctx field size for load/store insns, maybe 0 */
 };
 
 #define MAX_USED_MAPS 64 /* max number of maps accessed by one eBPF program */

kernel/bpf/verifier.c

@@ -758,15 +758,26 @@ static int check_packet_access(struct bpf_verifier_env *env, u32 regno, int off,
 }
 
 /* check access to 'struct bpf_context' fields */
-static int check_ctx_access(struct bpf_verifier_env *env, int off, int size,
+static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off, int size,
 			    enum bpf_access_type t, enum bpf_reg_type *reg_type)
 {
+	int ctx_field_size = 0;
+
 	/* for analyzer ctx accesses are already validated and converted */
 	if (env->analyzer_ops)
 		return 0;
 
 	if (env->prog->aux->ops->is_valid_access &&
-	    env->prog->aux->ops->is_valid_access(off, size, t, reg_type)) {
+	    env->prog->aux->ops->is_valid_access(off, size, t, reg_type, &ctx_field_size)) {
+		/* a non zero ctx_field_size indicates:
+		 * . For this field, the prog type specific ctx conversion algorithm
+		 *   only supports whole field access.
+		 * . This ctx access is a candiate for later verifier transformation
+		 *   to load the whole field and then apply a mask to get correct result.
+		 */
+		if (ctx_field_size)
+			env->insn_aux_data[insn_idx].ctx_field_size = ctx_field_size;
+
 		/* remember the offset of last byte accessed in ctx */
 		if (env->prog->aux->max_ctx_offset < off + size)
 			env->prog->aux->max_ctx_offset = off + size;
@@ -868,7 +879,7 @@ static int check_ptr_alignment(struct bpf_verifier_env *env,
  * if t==write && value_regno==-1, some unknown value is stored into memory
  * if t==read && value_regno==-1, don't care what we read from memory
  */
-static int check_mem_access(struct bpf_verifier_env *env, u32 regno, int off,
+static int check_mem_access(struct bpf_verifier_env *env, int insn_idx, u32 regno, int off,
 			    int bpf_size, enum bpf_access_type t,
 			    int value_regno)
 {
@@ -911,7 +922,7 @@ static int check_mem_access(struct bpf_verifier_env *env, u32 regno, int off,
 			verbose("R%d leaks addr into ctx\n", value_regno);
 			return -EACCES;
 		}
-		err = check_ctx_access(env, off, size, t, &reg_type);
+		err = check_ctx_access(env, insn_idx, off, size, t, &reg_type);
 		if (!err && t == BPF_READ && value_regno >= 0) {
 			mark_reg_unknown_value_and_range(state->regs,
 							 value_regno);
@@ -972,7 +983,7 @@ static int check_mem_access(struct bpf_verifier_env *env, u32 regno, int off,
 	return err;
 }
 
-static int check_xadd(struct bpf_verifier_env *env, struct bpf_insn *insn)
+static int check_xadd(struct bpf_verifier_env *env, int insn_idx, struct bpf_insn *insn)
 {
 	struct bpf_reg_state *regs = env->cur_state.regs;
 	int err;
@@ -994,13 +1005,13 @@ static int check_xadd(struct bpf_verifier_env *env, struct bpf_insn *insn)
 		return err;
 
 	/* check whether atomic_add can read the memory */
-	err = check_mem_access(env, insn->dst_reg, insn->off,
+	err = check_mem_access(env, insn_idx, insn->dst_reg, insn->off,
 			       BPF_SIZE(insn->code), BPF_READ, -1);
 	if (err)
 		return err;
 
 	/* check whether atomic_add can write into the same memory */
-	return check_mem_access(env, insn->dst_reg, insn->off,
+	return check_mem_access(env, insn_idx, insn->dst_reg, insn->off,
 				BPF_SIZE(insn->code), BPF_WRITE, -1);
 }
 
@@ -1416,7 +1427,7 @@ static int check_call(struct bpf_verifier_env *env, int func_id, int insn_idx)
 	 * is inferred from register state.
 	 */
 	for (i = 0; i < meta.access_size; i++) {
-		err = check_mem_access(env, meta.regno, i, BPF_B, BPF_WRITE, -1);
+		err = check_mem_access(env, insn_idx, meta.regno, i, BPF_B, BPF_WRITE, -1);
 		if (err)
 			return err;
 	}
@@ -2993,18 +3004,12 @@ static int do_check(struct bpf_verifier_env *env)
 			/* check that memory (src_reg + off) is readable,
 			 * the state of dst_reg will be updated by this func
 			 */
-			err = check_mem_access(env, insn->src_reg, insn->off,
+			err = check_mem_access(env, insn_idx, insn->src_reg, insn->off,
 					       BPF_SIZE(insn->code), BPF_READ,
 					       insn->dst_reg);
 			if (err)
 				return err;
 
-			if (BPF_SIZE(insn->code) != BPF_W &&
-			    BPF_SIZE(insn->code) != BPF_DW) {
-				insn_idx++;
-				continue;
-			}
-
 			prev_src_type = &env->insn_aux_data[insn_idx].ptr_type;
 
 			if (*prev_src_type == NOT_INIT) {
@@ -3032,7 +3037,7 @@ static int do_check(struct bpf_verifier_env *env)
 			enum bpf_reg_type *prev_dst_type, dst_reg_type;
 
 			if (BPF_MODE(insn->code) == BPF_XADD) {
-				err = check_xadd(env, insn);
+				err = check_xadd(env, insn_idx, insn);
 				if (err)
 					return err;
 				insn_idx++;
@@ -3051,7 +3056,7 @@ static int do_check(struct bpf_verifier_env *env)
 			dst_reg_type = regs[insn->dst_reg].type;
 
 			/* check that memory (dst_reg + off) is writeable */
-			err = check_mem_access(env, insn->dst_reg, insn->off,
+			err = check_mem_access(env, insn_idx, insn->dst_reg, insn->off,
 					       BPF_SIZE(insn->code), BPF_WRITE,
 					       insn->src_reg);
 			if (err)
@@ -3080,7 +3085,7 @@ static int do_check(struct bpf_verifier_env *env)
 				return err;
 
 			/* check that memory (dst_reg + off) is writeable */
-			err = check_mem_access(env, insn->dst_reg, insn->off,
+			err = check_mem_access(env, insn_idx, insn->dst_reg, insn->off,
 					       BPF_SIZE(insn->code), BPF_WRITE,
 					       -1);
 			if (err)
@@ -3383,7 +3388,7 @@ static int convert_ctx_accesses(struct bpf_verifier_env *env)
 	struct bpf_insn insn_buf[16], *insn;
 	struct bpf_prog *new_prog;
 	enum bpf_access_type type;
-	int i, cnt, delta = 0;
+	int i, cnt, off, size, ctx_field_size, is_narrower_load, delta = 0;
 
 	if (ops->gen_prologue) {
 		cnt = ops->gen_prologue(insn_buf, env->seen_direct_write,
@@ -3423,11 +3428,39 @@ static int convert_ctx_accesses(struct bpf_verifier_env *env)
 		if (env->insn_aux_data[i + delta].ptr_type != PTR_TO_CTX)
 			continue;
 
+		off = insn->off;
+		size = bpf_size_to_bytes(BPF_SIZE(insn->code));
+		ctx_field_size = env->insn_aux_data[i + delta].ctx_field_size;
+		is_narrower_load = (type == BPF_READ && size < ctx_field_size);
+
+		/* If the read access is a narrower load of the field,
+		 * convert to a 4/8-byte load, to minimum program type specific
+		 * convert_ctx_access changes. If conversion is successful,
+		 * we will apply proper mask to the result.
+		 */
+		if (is_narrower_load) {
+			int size_code = BPF_H;
+
+			if (ctx_field_size == 4)
+				size_code = BPF_W;
+			else if (ctx_field_size == 8)
+				size_code = BPF_DW;
+			insn->off = off & ~(ctx_field_size - 1);
+			insn->code = BPF_LDX | BPF_MEM | size_code;
+		}
 		cnt = ops->convert_ctx_access(type, insn, insn_buf, env->prog);
 		if (cnt == 0 || cnt >= ARRAY_SIZE(insn_buf)) {
 			verbose("bpf verifier is misconfigured\n");
 			return -EINVAL;
 		}
+		if (is_narrower_load) {
+			if (ctx_field_size <= 4)
+				insn_buf[cnt++] = BPF_ALU32_IMM(BPF_AND, insn->dst_reg,
+							(1 << size * 8) - 1);
+			else
+				insn_buf[cnt++] = BPF_ALU64_IMM(BPF_AND, insn->dst_reg,
+							(1 << size * 8) - 1);
+		}
 
 		new_prog = bpf_patch_insn_data(env, i + delta, insn_buf, cnt);
 		if (!new_prog)

kernel/trace/bpf_trace.c

@@ -479,7 +479,7 @@ static const struct bpf_func_proto *kprobe_prog_func_proto(enum bpf_func_id func
 
 /* bpf+kprobe programs can access fields of 'struct pt_regs' */
 static bool kprobe_prog_is_valid_access(int off, int size, enum bpf_access_type type,
-					enum bpf_reg_type *reg_type)
+					enum bpf_reg_type *reg_type, int *ctx_field_size)
 {
 	if (off < 0 || off >= sizeof(struct pt_regs))
 		return false;
@@ -562,7 +562,7 @@ static const struct bpf_func_proto *tp_prog_func_proto(enum bpf_func_id func_id)
 }
 
 static bool tp_prog_is_valid_access(int off, int size, enum bpf_access_type type,
-				    enum bpf_reg_type *reg_type)
+				    enum bpf_reg_type *reg_type, int *ctx_field_size)
 {
 	if (off < sizeof(void *) || off >= PERF_MAX_TRACE_SIZE)
 		return false;
@@ -581,17 +581,26 @@ const struct bpf_verifier_ops tracepoint_prog_ops = {
 };
 
 static bool pe_prog_is_valid_access(int off, int size, enum bpf_access_type type,
-				    enum bpf_reg_type *reg_type)
+				    enum bpf_reg_type *reg_type, int *ctx_field_size)
 {
+	int sample_period_off;
+
 	if (off < 0 || off >= sizeof(struct bpf_perf_event_data))
 		return false;
 	if (type != BPF_READ)
 		return false;
 	if (off % size != 0)
 		return false;
-	if (off == offsetof(struct bpf_perf_event_data, sample_period)) {
-		if (size != sizeof(u64))
-			return false;
+
+	/* permit 1, 2, 4 byte narrower and 8 normal read access to sample_period */
+	sample_period_off = offsetof(struct bpf_perf_event_data, sample_period);
+	if (off >= sample_period_off && off < sample_period_off + sizeof(__u64)) {
+		*ctx_field_size = 8;
+#ifdef __LITTLE_ENDIAN
+		return (off & 0x7) == 0 && size <= 8 && (size & (size - 1)) == 0;
+#else
+		return ((off & 0x7) + size) == 8 && size <= 8 && (size & (size - 1)) == 0;
+#endif
 	} else {
 		if (size != sizeof(long))
 			return false;

net/core/filter.c

@@ -2856,7 +2856,8 @@ lwt_xmit_func_proto(enum bpf_func_id func_id)
 	}
 }
 
-static bool __is_valid_access(int off, int size)
+static bool __is_valid_access(int off, int size, enum bpf_access_type type,
+			      int *ctx_field_size)
 {
 	if (off < 0 || off >= sizeof(struct __sk_buff))
 		return false;
@@ -2872,9 +2873,27 @@ static bool __is_valid_access(int off, int size)
 		    offsetof(struct __sk_buff, cb[4]) + sizeof(__u32))
 			return false;
 		break;
+	case offsetof(struct __sk_buff, data) ...
+	     offsetof(struct __sk_buff, data) + sizeof(__u32) - 1:
+	case offsetof(struct __sk_buff, data_end) ...
+	     offsetof(struct __sk_buff, data_end) + sizeof(__u32) - 1:
+		if (size != sizeof(__u32))
+			return false;
+		break;
 	default:
+		/* permit narrower load for not cb/data/data_end fields */
+		*ctx_field_size = 4;
+		if (type == BPF_WRITE) {
 			if (size != sizeof(__u32))
 				return false;
+		} else {
+			if (size != sizeof(__u32))
+#ifdef __LITTLE_ENDIAN
+				return (off & 0x3) == 0 && (size == 1 || size == 2);
+#else
+				return (off & 0x3) + size == 4 && (size == 1 || size == 2);
+#endif
+		}
 	}
 
 	return true;
@@ -2882,12 +2901,16 @@ static bool __is_valid_access(int off, int size)
 
 static bool sk_filter_is_valid_access(int off, int size,
 				      enum bpf_access_type type,
-				      enum bpf_reg_type *reg_type)
+				      enum bpf_reg_type *reg_type,
+				      int *ctx_field_size)
 {
 	switch (off) {
-	case offsetof(struct __sk_buff, tc_classid):
-	case offsetof(struct __sk_buff, data):
-	case offsetof(struct __sk_buff, data_end):
+	case offsetof(struct __sk_buff, tc_classid) ...
+	     offsetof(struct __sk_buff, tc_classid) + sizeof(__u32) - 1:
+	case offsetof(struct __sk_buff, data) ...
+	     offsetof(struct __sk_buff, data) + sizeof(__u32) - 1:
+	case offsetof(struct __sk_buff, data_end) ...
+	     offsetof(struct __sk_buff, data_end) + sizeof(__u32) - 1:
 		return false;
 	}
 
@@ -2901,15 +2924,17 @@ static bool sk_filter_is_valid_access(int off, int size,
 		}
 	}
 
-	return __is_valid_access(off, size);
+	return __is_valid_access(off, size, type, ctx_field_size);
 }
 
 static bool lwt_is_valid_access(int off, int size,
 				enum bpf_access_type type,
-				enum bpf_reg_type *reg_type)
+				enum bpf_reg_type *reg_type,
+				int *ctx_field_size)
 {
 	switch (off) {
-	case offsetof(struct __sk_buff, tc_classid):
+	case offsetof(struct __sk_buff, tc_classid) ...
+	     offsetof(struct __sk_buff, tc_classid) + sizeof(__u32) - 1:
 		return false;
 	}
 
@@ -2934,12 +2959,13 @@ static bool lwt_is_valid_access(int off, int size,
 		break;
 	}
 
-	return __is_valid_access(off, size);
+	return __is_valid_access(off, size, type, ctx_field_size);
 }
 
 static bool sock_filter_is_valid_access(int off, int size,
 					enum bpf_access_type type,
-					enum bpf_reg_type *reg_type)
+					enum bpf_reg_type *reg_type,
+					int *ctx_field_size)
 {
 	if (type == BPF_WRITE) {
 		switch (off) {
@@ -3002,7 +3028,8 @@ static int tc_cls_act_prologue(struct bpf_insn *insn_buf, bool direct_write,
 
 static bool tc_cls_act_is_valid_access(int off, int size,
 				       enum bpf_access_type type,
-				       enum bpf_reg_type *reg_type)
+				       enum bpf_reg_type *reg_type,
+				       int *ctx_field_size)
 {
 	if (type == BPF_WRITE) {
 		switch (off) {
@@ -3027,7 +3054,7 @@ static bool tc_cls_act_is_valid_access(int off, int size,
 		break;
 	}
 
-	return __is_valid_access(off, size);
+	return __is_valid_access(off, size, type, ctx_field_size);
 }
 
 static bool __is_valid_xdp_access(int off, int size)
@@ -3044,7 +3071,8 @@ static bool __is_valid_xdp_access(int off, int size)
 
 static bool xdp_is_valid_access(int off, int size,
 				enum bpf_access_type type,
-				enum bpf_reg_type *reg_type)
+				enum bpf_reg_type *reg_type,
+				int *ctx_field_size)
 {
 	if (type == BPF_WRITE)
 		return false;

tools/testing/selftests/bpf/test_verifier.c

@@ -1073,44 +1073,22 @@ static struct bpf_test tests[] = {
 		.result = ACCEPT,
 	},
 	{
-		"check cb access: byte, oob 1",
+		"__sk_buff->hash, offset 0, byte store not permitted",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
 			BPF_STX_MEM(BPF_B, BPF_REG_1, BPF_REG_0,
-				    offsetof(struct __sk_buff, cb[4]) + 4),
+				    offsetof(struct __sk_buff, hash)),
 			BPF_EXIT_INSN(),
 		},
 		.errstr = "invalid bpf_context access",
 		.result = REJECT,
 	},
 	{
-		"check cb access: byte, oob 2",
+		"__sk_buff->tc_index, offset 3, byte store not permitted",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
 			BPF_STX_MEM(BPF_B, BPF_REG_1, BPF_REG_0,
-				    offsetof(struct __sk_buff, cb[0]) - 1),
-			BPF_EXIT_INSN(),
-		},
-		.errstr = "invalid bpf_context access",
-		.result = REJECT,
-	},
-	{
-		"check cb access: byte, oob 3",
-		.insns = {
-			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_1,
-				    offsetof(struct __sk_buff, cb[4]) + 4),
-			BPF_EXIT_INSN(),
-		},
-		.errstr = "invalid bpf_context access",
-		.result = REJECT,
-	},
-	{
-		"check cb access: byte, oob 4",
-		.insns = {
-			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_1,
-				    offsetof(struct __sk_buff, cb[0]) - 1),
+				    offsetof(struct __sk_buff, tc_index) + 3),
 			BPF_EXIT_INSN(),
 		},
 		.errstr = "invalid bpf_context access",
@@ -1188,44 +1166,22 @@ static struct bpf_test tests[] = {
 		.result = REJECT,
 	},
 	{
-		"check cb access: half, oob 1",
+		"check __sk_buff->hash, offset 0, half store not permitted",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
 			BPF_STX_MEM(BPF_H, BPF_REG_1, BPF_REG_0,
-				    offsetof(struct __sk_buff, cb[4]) + 4),
+				    offsetof(struct __sk_buff, hash)),
 			BPF_EXIT_INSN(),
 		},
 		.errstr = "invalid bpf_context access",
 		.result = REJECT,
 	},
 	{
-		"check cb access: half, oob 2",
+		"check __sk_buff->tc_index, offset 2, half store not permitted",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
 			BPF_STX_MEM(BPF_H, BPF_REG_1, BPF_REG_0,
-				    offsetof(struct __sk_buff, cb[0]) - 2),
-			BPF_EXIT_INSN(),
-		},
-		.errstr = "invalid bpf_context access",
-		.result = REJECT,
-	},
-	{
-		"check cb access: half, oob 3",
-		.insns = {
-			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_LDX_MEM(BPF_H, BPF_REG_0, BPF_REG_1,
-				    offsetof(struct __sk_buff, cb[4]) + 4),
-			BPF_EXIT_INSN(),
-		},
-		.errstr = "invalid bpf_context access",
-		.result = REJECT,
-	},
-	{
-		"check cb access: half, oob 4",
-		.insns = {
-			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_LDX_MEM(BPF_H, BPF_REG_0, BPF_REG_1,
-				    offsetof(struct __sk_buff, cb[0]) - 2),
+				    offsetof(struct __sk_buff, tc_index) + 2),
 			BPF_EXIT_INSN(),
 		},
 		.errstr = "invalid bpf_context access",
@@ -1366,28 +1322,6 @@ static struct bpf_test tests[] = {
 	},
 	{
 		"check cb access: double, oob 2",
-		.insns = {
-			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_0,
-				    offsetof(struct __sk_buff, cb[4]) + 8),
-			BPF_EXIT_INSN(),
-		},
-		.errstr = "invalid bpf_context access",
-		.result = REJECT,
-	},
-	{
-		"check cb access: double, oob 3",
-		.insns = {
-			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_0,
-				    offsetof(struct __sk_buff, cb[0]) - 8),
-			BPF_EXIT_INSN(),
-		},
-		.errstr = "invalid bpf_context access",
-		.result = REJECT,
-	},
-	{
-		"check cb access: double, oob 4",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
 			BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_1,
@@ -1398,22 +1332,22 @@ static struct bpf_test tests[] = {
 		.result = REJECT,
 	},
 	{
-		"check cb access: double, oob 5",
+		"check __sk_buff->ifindex dw store not permitted",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
-			BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_1,
-				    offsetof(struct __sk_buff, cb[4]) + 8),
+			BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_0,
+				    offsetof(struct __sk_buff, ifindex)),
 			BPF_EXIT_INSN(),
 		},
 		.errstr = "invalid bpf_context access",
 		.result = REJECT,
 	},
 	{
-		"check cb access: double, oob 6",
+		"check __sk_buff->ifindex dw load not permitted",
 		.insns = {
 			BPF_MOV64_IMM(BPF_REG_0, 0),
 			BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_1,
-				    offsetof(struct __sk_buff, cb[0]) - 8),
+				    offsetof(struct __sk_buff, ifindex)),
 			BPF_EXIT_INSN(),
 		},
 		.errstr = "invalid bpf_context access",