bpf: Add __sk_buff->delivery_time_type and bpf_skb_set_skb_delivery_time()

* __sk_buff->delivery_time_type:
This patch adds __sk_buff->delivery_time_type.  It tells if the
delivery_time is stored in __sk_buff->tstamp or not.

It will be most useful for ingress to tell if the __sk_buff->tstamp
has the (rcv) timestamp or delivery_time.  If delivery_time_type
is 0 (BPF_SKB_DELIVERY_TIME_NONE), it has the (rcv) timestamp.

Two non-zero types are defined for the delivery_time_type,
BPF_SKB_DELIVERY_TIME_MONO and BPF_SKB_DELIVERY_TIME_UNSPEC.  For UNSPEC,
it can only happen in egress because only mono delivery_time can be
forwarded to ingress now.  The clock of UNSPEC delivery_time
can be deduced from the skb->sk->sk_clockid which is how
the sch_etf doing it also.

* Provide forwarded delivery_time to tc-bpf@ingress:
With the help of the new delivery_time_type, the tc-bpf has a way
to tell if the __sk_buff->tstamp has the (rcv) timestamp or
the delivery_time.  During bpf load time, the verifier will learn if
the bpf prog has accessed the new __sk_buff->delivery_time_type.
If it does, it means the tc-bpf@ingress is expecting the
skb->tstamp could have the delivery_time.  The kernel will then
read the skb->tstamp as-is during bpf insn rewrite without
checking the skb->mono_delivery_time.  This is done by adding a
new prog->delivery_time_access bit.  The same goes for
writing skb->tstamp.

* bpf_skb_set_delivery_time():
The bpf_skb_set_delivery_time() helper is added to allow setting both
delivery_time and the delivery_time_type at the same time.  If the
tc-bpf does not need to change the delivery_time_type, it can directly
write to the __sk_buff->tstamp as the existing tc-bpf has already been
doing.  It will be most useful at ingress to change the
__sk_buff->tstamp from the (rcv) timestamp to
a mono delivery_time and then bpf_redirect_*().

bpf only has mono clock helper (bpf_ktime_get_ns), and
the current known use case is the mono EDT for fq, and
only mono delivery time can be kept during forward now,
so bpf_skb_set_delivery_time() only supports setting
BPF_SKB_DELIVERY_TIME_MONO.  It can be extended later when use cases
come up and the forwarding path also supports other clock bases.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/filter.h

@@ -572,7 +572,8 @@ struct bpf_prog {
 				has_callchain_buf:1, /* callchain buffer allocated? */
 				enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
 				call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
-				call_get_func_ip:1; /* Do we call get_func_ip() */
+				call_get_func_ip:1, /* Do we call get_func_ip() */
+				delivery_time_access:1; /* Accessed __sk_buff->delivery_time_type */
 	enum bpf_prog_type	type;		/* Type of BPF program */
 	enum bpf_attach_type	expected_attach_type; /* For some prog types */
 	u32			len;		/* Number of filter blocks */

include/uapi/linux/bpf.h

@@ -5086,6 +5086,37 @@ union bpf_attr {
  *	Return
  *		0 on success, or a negative error in case of failure. On error
  *		*dst* buffer is zeroed out.
+ *
+ * long bpf_skb_set_delivery_time(struct sk_buff *skb, u64 dtime, u32 dtime_type)
+ *	Description
+ *		Set a *dtime* (delivery time) to the __sk_buff->tstamp and also
+ *		change the __sk_buff->delivery_time_type to *dtime_type*.
+ *
+ *		When setting a delivery time (non zero *dtime*) to
+ *		__sk_buff->tstamp, only BPF_SKB_DELIVERY_TIME_MONO *dtime_type*
+ *		is supported.  It is the only delivery_time_type that will be
+ *		kept after bpf_redirect_*().
+ *
+ *		If there is no need to change the __sk_buff->delivery_time_type,
+ *		the delivery time can be directly written to __sk_buff->tstamp
+ *		instead.
+ *
+ *		*dtime* 0 and *dtime_type* BPF_SKB_DELIVERY_TIME_NONE
+ *		can be used to clear any delivery time stored in
+ *		__sk_buff->tstamp.
+ *
+ *		Only IPv4 and IPv6 skb->protocol are supported.
+ *
+ *		This function is most useful when it needs to set a
+ *		mono delivery time to __sk_buff->tstamp and then
+ *		bpf_redirect_*() to the egress of an iface.  For example,
+ *		changing the (rcv) timestamp in __sk_buff->tstamp at
+ *		ingress to a mono delivery time and then bpf_redirect_*()
+ *		to sch_fq@phy-dev.
+ *	Return
+ *		0 on success.
+ *		**-EINVAL** for invalid input
+ *		**-EOPNOTSUPP** for unsupported delivery_time_type and protocol
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -5280,6 +5311,7 @@ union bpf_attr {
 	FN(xdp_load_bytes),		\
 	FN(xdp_store_bytes),		\
 	FN(copy_from_user_task),	\
+	FN(skb_set_delivery_time),      \
 	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
@@ -5469,6 +5501,12 @@ union {					\
 	__u64 :64;			\
 } __attribute__((aligned(8)))
 
+enum {
+	BPF_SKB_DELIVERY_TIME_NONE,
+	BPF_SKB_DELIVERY_TIME_UNSPEC,
+	BPF_SKB_DELIVERY_TIME_MONO,
+};
+
 /* user accessible mirror of in-kernel sk_buff.
  * new fields can only be added to the end of this structure
  */
@@ -5509,7 +5547,8 @@ struct __sk_buff {
 	__u32 gso_segs;
 	__bpf_md_ptr(struct bpf_sock *, sk);
 	__u32 gso_size;
-	__u32 :32;		/* Padding, future use. */
+	__u8  delivery_time_type;
+	__u32 :24;		/* Padding, future use. */
 	__u64 hwtstamp;
 };
 

net/core/filter.c

@@ -7388,6 +7388,43 @@ static const struct bpf_func_proto bpf_sock_ops_reserve_hdr_opt_proto = {
 	.arg3_type	= ARG_ANYTHING,
 };
 
+BPF_CALL_3(bpf_skb_set_delivery_time, struct sk_buff *, skb,
+	   u64, dtime, u32, dtime_type)
+{
+	/* skb_clear_delivery_time() is done for inet protocol */
+	if (skb->protocol != htons(ETH_P_IP) &&
+	    skb->protocol != htons(ETH_P_IPV6))
+		return -EOPNOTSUPP;
+
+	switch (dtime_type) {
+	case BPF_SKB_DELIVERY_TIME_MONO:
+		if (!dtime)
+			return -EINVAL;
+		skb->tstamp = dtime;
+		skb->mono_delivery_time = 1;
+		break;
+	case BPF_SKB_DELIVERY_TIME_NONE:
+		if (dtime)
+			return -EINVAL;
+		skb->tstamp = 0;
+		skb->mono_delivery_time = 0;
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	return 0;
+}
+
+static const struct bpf_func_proto bpf_skb_set_delivery_time_proto = {
+	.func           = bpf_skb_set_delivery_time,
+	.gpl_only       = false,
+	.ret_type       = RET_INTEGER,
+	.arg1_type      = ARG_PTR_TO_CTX,
+	.arg2_type      = ARG_ANYTHING,
+	.arg3_type      = ARG_ANYTHING,
+};
+
 #endif /* CONFIG_INET */
 
 bool bpf_helper_changes_pkt_data(void *func)
@@ -7749,6 +7786,8 @@ tc_cls_act_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_tcp_gen_syncookie_proto;
 	case BPF_FUNC_sk_assign:
 		return &bpf_sk_assign_proto;
+	case BPF_FUNC_skb_set_delivery_time:
+		return &bpf_skb_set_delivery_time_proto;
 #endif
 	default:
 		return bpf_sk_base_func_proto(func_id);
@@ -8088,7 +8127,9 @@ static bool bpf_skb_is_valid_access(int off, int size, enum bpf_access_type type
 			return false;
 		info->reg_type = PTR_TO_SOCK_COMMON_OR_NULL;
 		break;
-	case offsetofend(struct __sk_buff, gso_size) ... offsetof(struct __sk_buff, hwtstamp) - 1:
+	case offsetof(struct __sk_buff, delivery_time_type):
+		return false;
+	case offsetofend(struct __sk_buff, delivery_time_type) ... offsetof(struct __sk_buff, hwtstamp) - 1:
 		/* Explicitly prohibit access to padding in __sk_buff. */
 		return false;
 	default:
@@ -8443,6 +8484,15 @@ static bool tc_cls_act_is_valid_access(int off, int size,
 		break;
 	case bpf_ctx_range_till(struct __sk_buff, family, local_port):
 		return false;
+	case offsetof(struct __sk_buff, delivery_time_type):
+		/* The convert_ctx_access() on reading and writing
+		 * __sk_buff->tstamp depends on whether the bpf prog
+		 * has used __sk_buff->delivery_time_type or not.
+		 * Thus, we need to set prog->delivery_time_access
+		 * earlier during is_valid_access() here.
+		 */
+		((struct bpf_prog *)prog)->delivery_time_access = 1;
+		return size == sizeof(__u8);
 	}
 
 	return bpf_skb_is_valid_access(off, size, type, prog, info);
@@ -8838,6 +8888,45 @@ static u32 flow_dissector_convert_ctx_access(enum bpf_access_type type,
 	return insn - insn_buf;
 }
 
+static struct bpf_insn *bpf_convert_dtime_type_read(const struct bpf_insn *si,
+						    struct bpf_insn *insn)
+{
+	__u8 value_reg = si->dst_reg;
+	__u8 skb_reg = si->src_reg;
+	__u8 tmp_reg = BPF_REG_AX;
+
+	*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg,
+			      SKB_MONO_DELIVERY_TIME_OFFSET);
+	*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg,
+				SKB_MONO_DELIVERY_TIME_MASK);
+	*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 2);
+	/* value_reg = BPF_SKB_DELIVERY_TIME_MONO */
+	*insn++ = BPF_MOV32_IMM(value_reg, BPF_SKB_DELIVERY_TIME_MONO);
+	*insn++ = BPF_JMP_A(IS_ENABLED(CONFIG_NET_CLS_ACT) ? 10 : 5);
+
+	*insn++ = BPF_LDX_MEM(BPF_DW, tmp_reg, skb_reg,
+			      offsetof(struct sk_buff, tstamp));
+	*insn++ = BPF_JMP_IMM(BPF_JNE, tmp_reg, 0, 2);
+	/* value_reg = BPF_SKB_DELIVERY_TIME_NONE */
+	*insn++ = BPF_MOV32_IMM(value_reg, BPF_SKB_DELIVERY_TIME_NONE);
+	*insn++ = BPF_JMP_A(IS_ENABLED(CONFIG_NET_CLS_ACT) ? 6 : 1);
+
+#ifdef CONFIG_NET_CLS_ACT
+	*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg, TC_AT_INGRESS_OFFSET);
+	*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg, TC_AT_INGRESS_MASK);
+	*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 2);
+	/* At ingress, value_reg = 0 */
+	*insn++ = BPF_MOV32_IMM(value_reg, 0);
+	*insn++ = BPF_JMP_A(1);
+#endif
+
+	/* value_reg = BPF_SKB_DELIVERYT_TIME_UNSPEC */
+	*insn++ = BPF_MOV32_IMM(value_reg, BPF_SKB_DELIVERY_TIME_UNSPEC);
+
+	/* 15 insns with CONFIG_NET_CLS_ACT */
+	return insn;
+}
+
 static struct bpf_insn *bpf_convert_shinfo_access(const struct bpf_insn *si,
 						  struct bpf_insn *insn)
 {
@@ -8859,29 +8948,32 @@ static struct bpf_insn *bpf_convert_shinfo_access(const struct bpf_insn *si,
 	return insn;
 }
 
-static struct bpf_insn *bpf_convert_tstamp_read(const struct bpf_insn *si,
+static struct bpf_insn *bpf_convert_tstamp_read(const struct bpf_prog *prog,
+						const struct bpf_insn *si,
 						struct bpf_insn *insn)
 {
 	__u8 value_reg = si->dst_reg;
 	__u8 skb_reg = si->src_reg;
 
 #ifdef CONFIG_NET_CLS_ACT
-	__u8 tmp_reg = BPF_REG_AX;
-
-	*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg, TC_AT_INGRESS_OFFSET);
-	*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg, TC_AT_INGRESS_MASK);
-	*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 5);
-	/* @ingress, read __sk_buff->tstamp as the (rcv) timestamp,
-	 * so check the skb->mono_delivery_time.
-	 */
-	*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg,
-			      SKB_MONO_DELIVERY_TIME_OFFSET);
-	*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg,
-				SKB_MONO_DELIVERY_TIME_MASK);
-	*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 2);
-	/* skb->mono_delivery_time is set, read 0 as the (rcv) timestamp. */
-	*insn++ = BPF_MOV64_IMM(value_reg, 0);
-	*insn++ = BPF_JMP_A(1);
+	if (!prog->delivery_time_access) {
+		__u8 tmp_reg = BPF_REG_AX;
+
+		*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg, TC_AT_INGRESS_OFFSET);
+		*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg, TC_AT_INGRESS_MASK);
+		*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 5);
+		/* @ingress, read __sk_buff->tstamp as the (rcv) timestamp,
+		 * so check the skb->mono_delivery_time.
+		 */
+		*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg,
+				      SKB_MONO_DELIVERY_TIME_OFFSET);
+		*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg,
+					SKB_MONO_DELIVERY_TIME_MASK);
+		*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 2);
+		/* skb->mono_delivery_time is set, read 0 as the (rcv) timestamp. */
+		*insn++ = BPF_MOV64_IMM(value_reg, 0);
+		*insn++ = BPF_JMP_A(1);
+	}
 #endif
 
 	*insn++ = BPF_LDX_MEM(BPF_DW, value_reg, skb_reg,
@@ -8889,27 +8981,30 @@ static struct bpf_insn *bpf_convert_tstamp_read(const struct bpf_insn *si,
 	return insn;
 }
 
-static struct bpf_insn *bpf_convert_tstamp_write(const struct bpf_insn *si,
+static struct bpf_insn *bpf_convert_tstamp_write(const struct bpf_prog *prog,
+						 const struct bpf_insn *si,
 						 struct bpf_insn *insn)
 {
 	__u8 value_reg = si->src_reg;
 	__u8 skb_reg = si->dst_reg;
 
 #ifdef CONFIG_NET_CLS_ACT
-	__u8 tmp_reg = BPF_REG_AX;
-
-	*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg, TC_AT_INGRESS_OFFSET);
-	*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg, TC_AT_INGRESS_MASK);
-	*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 3);
-	/* Writing __sk_buff->tstamp at ingress as the (rcv) timestamp.
-	 * Clear the skb->mono_delivery_time.
-	 */
-	*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg,
-			      SKB_MONO_DELIVERY_TIME_OFFSET);
-	*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg,
-				~SKB_MONO_DELIVERY_TIME_MASK);
-	*insn++ = BPF_STX_MEM(BPF_B, skb_reg, tmp_reg,
-			      SKB_MONO_DELIVERY_TIME_OFFSET);
+	if (!prog->delivery_time_access) {
+		__u8 tmp_reg = BPF_REG_AX;
+
+		*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg, TC_AT_INGRESS_OFFSET);
+		*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg, TC_AT_INGRESS_MASK);
+		*insn++ = BPF_JMP32_IMM(BPF_JEQ, tmp_reg, 0, 3);
+		/* Writing __sk_buff->tstamp at ingress as the (rcv) timestamp.
+		 * Clear the skb->mono_delivery_time.
+		 */
+		*insn++ = BPF_LDX_MEM(BPF_B, tmp_reg, skb_reg,
+				      SKB_MONO_DELIVERY_TIME_OFFSET);
+		*insn++ = BPF_ALU32_IMM(BPF_AND, tmp_reg,
+					~SKB_MONO_DELIVERY_TIME_MASK);
+		*insn++ = BPF_STX_MEM(BPF_B, skb_reg, tmp_reg,
+				      SKB_MONO_DELIVERY_TIME_OFFSET);
+	}
 #endif
 
 	/* skb->tstamp = tstamp */
@@ -9226,9 +9321,13 @@ static u32 bpf_convert_ctx_access(enum bpf_access_type type,
 		BUILD_BUG_ON(sizeof_field(struct sk_buff, tstamp) != 8);
 
 		if (type == BPF_WRITE)
-			insn = bpf_convert_tstamp_write(si, insn);
+			insn = bpf_convert_tstamp_write(prog, si, insn);
 		else
-			insn = bpf_convert_tstamp_read(si, insn);
+			insn = bpf_convert_tstamp_read(prog, si, insn);
+		break;
+
+	case offsetof(struct __sk_buff, delivery_time_type):
+		insn = bpf_convert_dtime_type_read(si, insn);
 		break;
 
 	case offsetof(struct __sk_buff, gso_segs):

tools/include/uapi/linux/bpf.h

@@ -5086,6 +5086,37 @@ union bpf_attr {
  *	Return
  *		0 on success, or a negative error in case of failure. On error
  *		*dst* buffer is zeroed out.
+ *
+ * long bpf_skb_set_delivery_time(struct sk_buff *skb, u64 dtime, u32 dtime_type)
+ *	Description
+ *		Set a *dtime* (delivery time) to the __sk_buff->tstamp and also
+ *		change the __sk_buff->delivery_time_type to *dtime_type*.
+ *
+ *		When setting a delivery time (non zero *dtime*) to
+ *		__sk_buff->tstamp, only BPF_SKB_DELIVERY_TIME_MONO *dtime_type*
+ *		is supported.  It is the only delivery_time_type that will be
+ *		kept after bpf_redirect_*().
+ *
+ *		If there is no need to change the __sk_buff->delivery_time_type,
+ *		the delivery time can be directly written to __sk_buff->tstamp
+ *		instead.
+ *
+ *		*dtime* 0 and *dtime_type* BPF_SKB_DELIVERY_TIME_NONE
+ *		can be used to clear any delivery time stored in
+ *		__sk_buff->tstamp.
+ *
+ *		Only IPv4 and IPv6 skb->protocol are supported.
+ *
+ *		This function is most useful when it needs to set a
+ *		mono delivery time to __sk_buff->tstamp and then
+ *		bpf_redirect_*() to the egress of an iface.  For example,
+ *		changing the (rcv) timestamp in __sk_buff->tstamp at
+ *		ingress to a mono delivery time and then bpf_redirect_*()
+ *		to sch_fq@phy-dev.
+ *	Return
+ *		0 on success.
+ *		**-EINVAL** for invalid input
+ *		**-EOPNOTSUPP** for unsupported delivery_time_type and protocol
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -5280,6 +5311,7 @@ union bpf_attr {
 	FN(xdp_load_bytes),		\
 	FN(xdp_store_bytes),		\
 	FN(copy_from_user_task),	\
+	FN(skb_set_delivery_time),      \
 	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
@@ -5469,6 +5501,12 @@ union {					\
 	__u64 :64;			\
 } __attribute__((aligned(8)))
 
+enum {
+	BPF_SKB_DELIVERY_TIME_NONE,
+	BPF_SKB_DELIVERY_TIME_UNSPEC,
+	BPF_SKB_DELIVERY_TIME_MONO,
+};
+
 /* user accessible mirror of in-kernel sk_buff.
  * new fields can only be added to the end of this structure
  */
@@ -5509,7 +5547,8 @@ struct __sk_buff {
 	__u32 gso_segs;
 	__bpf_md_ptr(struct bpf_sock *, sk);
 	__u32 gso_size;
-	__u32 :32;		/* Padding, future use. */
+	__u8  delivery_time_type;
+	__u32 :24;		/* Padding, future use. */
 	__u64 hwtstamp;
 };