bpf: Add bpf_xdp_output() helper

Introduce new helper that reuses existing xdp perf_event output
implementation, but can be called from raw_tracepoint programs
that receive 'struct xdp_buff *' as a tracepoint argument.

Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/bpf/158348514556.2239.11050972434793741444.stgit@xdp-tutorial

include/uapi/linux/bpf.h

@@ -2927,6 +2927,29 @@ union bpf_attr {
  *
  *		**-ENOENT** if pidns does not exists for the current task.
  *
+ * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
+ *	Description
+ *		Write raw *data* blob into a special BPF perf event held by
+ *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
+ *		event must have the following attributes: **PERF_SAMPLE_RAW**
+ *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
+ *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
+ *
+ *		The *flags* are used to indicate the index in *map* for which
+ *		the value must be put, masked with **BPF_F_INDEX_MASK**.
+ *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
+ *		to indicate that the index of the current CPU core should be
+ *		used.
+ *
+ *		The value to write, of *size*, is passed through eBPF stack and
+ *		pointed by *data*.
+ *
+ *		*ctx* is a pointer to in-kernel struct xdp_buff.
+ *
+ *		This helper is similar to **bpf_perf_eventoutput**\ () but
+ *		restricted to raw_tracepoint bpf programs.
+ *	Return
+ *		0 on success, or a negative error in case of failure.
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -3049,7 +3072,8 @@ union bpf_attr {
 	FN(send_signal_thread),		\
 	FN(jiffies64),			\
 	FN(read_branch_records),	\
-	FN(get_ns_current_pid_tgid),
+	FN(get_ns_current_pid_tgid),	\
+	FN(xdp_output),
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  * function eBPF program intends to call

kernel/bpf/verifier.c

@@ -3650,7 +3650,8 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env,
 		if (func_id != BPF_FUNC_perf_event_read &&
 		    func_id != BPF_FUNC_perf_event_output &&
 		    func_id != BPF_FUNC_skb_output &&
-		    func_id != BPF_FUNC_perf_event_read_value)
+		    func_id != BPF_FUNC_perf_event_read_value &&
+		    func_id != BPF_FUNC_xdp_output)
 			goto error;
 		break;
 	case BPF_MAP_TYPE_STACK_TRACE:
@@ -3740,6 +3741,7 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env,
 	case BPF_FUNC_perf_event_output:
 	case BPF_FUNC_perf_event_read_value:
 	case BPF_FUNC_skb_output:
+	case BPF_FUNC_xdp_output:
 		if (map->map_type != BPF_MAP_TYPE_PERF_EVENT_ARRAY)
 			goto error;
 		break;

kernel/trace/bpf_trace.c

@@ -1145,6 +1145,7 @@ static const struct bpf_func_proto bpf_perf_event_output_proto_raw_tp = {
 };
 
 extern const struct bpf_func_proto bpf_skb_output_proto;
+extern const struct bpf_func_proto bpf_xdp_output_proto;
 
 BPF_CALL_3(bpf_get_stackid_raw_tp, struct bpf_raw_tracepoint_args *, args,
 	   struct bpf_map *, map, u64, flags)
@@ -1220,6 +1221,8 @@ tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 #ifdef CONFIG_NET
 	case BPF_FUNC_skb_output:
 		return &bpf_skb_output_proto;
+	case BPF_FUNC_xdp_output:
+		return &bpf_xdp_output_proto;
 #endif
 	default:
 		return raw_tp_prog_func_proto(func_id, prog);

net/core/filter.c

@@ -4061,7 +4061,8 @@ BPF_CALL_5(bpf_xdp_event_output, struct xdp_buff *, xdp, struct bpf_map *, map,
 
 	if (unlikely(flags & ~(BPF_F_CTXLEN_MASK | BPF_F_INDEX_MASK)))
 		return -EINVAL;
-	if (unlikely(xdp_size > (unsigned long)(xdp->data_end - xdp->data)))
+	if (unlikely(!xdp ||
+		     xdp_size > (unsigned long)(xdp->data_end - xdp->data)))
 		return -EFAULT;
 
 	return bpf_event_output(map, flags, meta, meta_size, xdp->data,
@@ -4079,6 +4080,19 @@ static const struct bpf_func_proto bpf_xdp_event_output_proto = {
 	.arg5_type	= ARG_CONST_SIZE_OR_ZERO,
 };
 
+static int bpf_xdp_output_btf_ids[5];
+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,
+	.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)
 {
 	return skb->sk ? sock_gen_cookie(skb->sk) : 0;

tools/include/uapi/linux/bpf.h

@@ -2927,6 +2927,29 @@ union bpf_attr {
  *
  *		**-ENOENT** if pidns does not exists for the current task.
  *
+ * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
+ *	Description
+ *		Write raw *data* blob into a special BPF perf event held by
+ *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
+ *		event must have the following attributes: **PERF_SAMPLE_RAW**
+ *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
+ *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
+ *
+ *		The *flags* are used to indicate the index in *map* for which
+ *		the value must be put, masked with **BPF_F_INDEX_MASK**.
+ *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
+ *		to indicate that the index of the current CPU core should be
+ *		used.
+ *
+ *		The value to write, of *size*, is passed through eBPF stack and
+ *		pointed by *data*.
+ *
+ *		*ctx* is a pointer to in-kernel struct xdp_buff.
+ *
+ *		This helper is similar to **bpf_perf_eventoutput**\ () but
+ *		restricted to raw_tracepoint bpf programs.
+ *	Return
+ *		0 on success, or a negative error in case of failure.
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -3049,7 +3072,8 @@ union bpf_attr {
 	FN(send_signal_thread),		\
 	FN(jiffies64),			\
 	FN(read_branch_records),	\
-	FN(get_ns_current_pid_tgid),
+	FN(get_ns_current_pid_tgid),	\
+	FN(xdp_output),
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  * function eBPF program intends to call

tools/testing/selftests/bpf/prog_tests/xdp_bpf2bpf.c

@@ -4,17 +4,51 @@
 #include "test_xdp.skel.h"
 #include "test_xdp_bpf2bpf.skel.h"
 
+struct meta {
+	int ifindex;
+	int pkt_len;
+};
+
+static void on_sample(void *ctx, int cpu, void *data, __u32 size)
+{
+	int duration = 0;
+	struct meta *meta = (struct meta *)data;
+	struct ipv4_packet *trace_pkt_v4 = data + sizeof(*meta);
+
+	if (CHECK(size < sizeof(pkt_v4) + sizeof(*meta),
+		  "check_size", "size %u < %zu\n",
+		  size, sizeof(pkt_v4) + sizeof(*meta)))
+		return;
+
+	if (CHECK(meta->ifindex != if_nametoindex("lo"), "check_meta_ifindex",
+		  "meta->ifindex = %d\n", meta->ifindex))
+		return;
+
+	if (CHECK(meta->pkt_len != sizeof(pkt_v4), "check_meta_pkt_len",
+		  "meta->pkt_len = %zd\n", sizeof(pkt_v4)))
+		return;
+
+	if (CHECK(memcmp(trace_pkt_v4, &pkt_v4, sizeof(pkt_v4)),
+		  "check_packet_content", "content not the same\n"))
+		return;
+
+	*(bool *)ctx = true;
+}
+
 void test_xdp_bpf2bpf(void)
 {
 	__u32 duration = 0, retval, size;
 	char buf[128];
 	int err, pkt_fd, map_fd;
+	bool passed = false;
 	struct iphdr *iph = (void *)buf + sizeof(struct ethhdr);
 	struct iptnl_info value4 = {.family = AF_INET};
 	struct test_xdp *pkt_skel = NULL;
 	struct test_xdp_bpf2bpf *ftrace_skel = NULL;
 	struct vip key4 = {.protocol = 6, .family = AF_INET};
 	struct bpf_program *prog;
+	struct perf_buffer *pb = NULL;
+	struct perf_buffer_opts pb_opts = {};
 
 	/* Load XDP program to introspect */
 	pkt_skel = test_xdp__open_and_load();
@@ -50,6 +84,14 @@ void test_xdp_bpf2bpf(void)
 	if (CHECK(err, "ftrace_attach", "ftrace attach failed: %d\n", err))
 		goto out;
 
+	/* Set up perf buffer */
+	pb_opts.sample_cb = on_sample;
+	pb_opts.ctx = &passed;
+	pb = perf_buffer__new(bpf_map__fd(ftrace_skel->maps.perf_buf_map),
+			      1, &pb_opts);
+	if (CHECK(IS_ERR(pb), "perf_buf__new", "err %ld\n", PTR_ERR(pb)))
+		goto out;
+
 	/* Run test program */
 	err = bpf_prog_test_run(pkt_fd, 1, &pkt_v4, sizeof(pkt_v4),
 				buf, &size, &retval, &duration);
@@ -60,6 +102,15 @@ void test_xdp_bpf2bpf(void)
 		  err, errno, retval, size))
 		goto out;
 
+	/* Make sure bpf_xdp_output() was triggered and it sent the expected
+	 * data to the perf ring buffer.
+	 */
+	err = perf_buffer__poll(pb, 100);
+	if (CHECK(err < 0, "perf_buffer__poll", "err %d\n", err))
+		goto out;
+
+	CHECK_FAIL(!passed);
+
 	/* Verify test results */
 	if (CHECK(ftrace_skel->bss->test_result_fentry != if_nametoindex("lo"),
 		  "result", "fentry failed err %llu\n",
@@ -70,6 +121,8 @@ void test_xdp_bpf2bpf(void)
 	      "fexit failed err %llu\n", ftrace_skel->bss->test_result_fexit);
 
 out:
+	if (pb)
+		perf_buffer__free(pb);
 	test_xdp__destroy(pkt_skel);
 	test_xdp_bpf2bpf__destroy(ftrace_skel);
 }

tools/testing/selftests/bpf/progs/test_xdp_bpf2bpf.c

@@ -3,6 +3,8 @@
 #include <bpf/bpf_tracing.h>
 #include <bpf/bpf_helpers.h>
 
+char _license[] SEC("license") = "GPL";
+
 struct net_device {
 	/* Structure does not need to contain all entries,
 	 * as "preserve_access_index" will use BTF to fix this...
@@ -27,10 +29,32 @@ struct xdp_buff {
 	struct xdp_rxq_info *rxq;
 } __attribute__((preserve_access_index));
 
+struct meta {
+	int ifindex;
+	int pkt_len;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERF_EVENT_ARRAY);
+	__uint(key_size, sizeof(int));
+	__uint(value_size, sizeof(int));
+} perf_buf_map SEC(".maps");
+
 __u64 test_result_fentry = 0;
 SEC("fentry/FUNC")
 int BPF_PROG(trace_on_entry, struct xdp_buff *xdp)
 {
+	struct meta meta;
+	void *data_end = (void *)(long)xdp->data_end;
+	void *data = (void *)(long)xdp->data;
+
+	meta.ifindex = xdp->rxq->dev->ifindex;
+	meta.pkt_len = data_end - data;
+	bpf_xdp_output(xdp, &perf_buf_map,
+		       ((__u64) meta.pkt_len << 32) |
+		       BPF_F_CURRENT_CPU,
+		       &meta, sizeof(meta));
+
 	test_result_fentry = xdp->rxq->dev->ifindex;
 	return 0;
 }