Merge branch 'bpf-perf'

Alexei Starovoitov says:

====================
bpf_perf_event_output helper

Over the last year there were multiple attempts to let eBPF programs
output data into perf events by He Kuang and Wangnan.
The last one was:
https://lkml.org/lkml/2015/7/20/736
It was almost perfect with exception that all bpf programs would sent
data into one global perf_event.
This patch set takes different approach by letting user space
open independent PERF_COUNT_SW_BPF_OUTPUT events, so that program
output won't collide.

Wangnan is working on corresponding perf patches.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

10 files changed, 308 insertions, 6 deletions

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 564f1f091991..2e032426cfb7 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -287,6 +287,17 @@ enum bpf_func_id {
 	 * Return: realm if != 0
 	 */
 	BPF_FUNC_get_route_realm,
+
+	/**
+	 * bpf_perf_event_output(ctx, map, index, data, size) - output perf raw sample
+	 * @ctx: struct pt_regs*
+	 * @map: pointer to perf_event_array map
+	 * @index: index of event in the map
+	 * @data: data on stack to be output as raw data
+	 * @size: size of data
+	 * Return: 0 on success
+	 */
+	BPF_FUNC_perf_event_output,
 	__BPF_FUNC_MAX_ID,
 };
 
diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index 2881145cda86..d3c417615361 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -110,6 +110,7 @@ enum perf_sw_ids {
 	PERF_COUNT_SW_ALIGNMENT_FAULTS		= 7,
 	PERF_COUNT_SW_EMULATION_FAULTS		= 8,
 	PERF_COUNT_SW_DUMMY			= 9,
+	PERF_COUNT_SW_BPF_OUTPUT		= 10,
 
 	PERF_COUNT_SW_MAX,			/* non-ABI */
 };
diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c
index f2d9e698c753..e3cfe46b074f 100644
--- a/kernel/bpf/arraymap.c
+++ b/kernel/bpf/arraymap.c
@@ -295,6 +295,8 @@ static void *perf_event_fd_array_get_ptr(struct bpf_map *map, int fd)
 		return (void *)attr;
 
 	if (attr->type != PERF_TYPE_RAW &&
+	    !(attr->type == PERF_TYPE_SOFTWARE &&
+	      attr->config == PERF_COUNT_SW_BPF_OUTPUT) &&
 	    attr->type != PERF_TYPE_HARDWARE) {
 		perf_event_release_kernel(event);
 		return ERR_PTR(-EINVAL);
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 1d6b97be79e1..b56cf51f8d42 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -245,6 +245,7 @@ static const struct {
 } func_limit[] = {
 	{BPF_MAP_TYPE_PROG_ARRAY, BPF_FUNC_tail_call},
 	{BPF_MAP_TYPE_PERF_EVENT_ARRAY, BPF_FUNC_perf_event_read},
+	{BPF_MAP_TYPE_PERF_EVENT_ARRAY, BPF_FUNC_perf_event_output},
 };
 
 static void print_verifier_state(struct verifier_env *env)
@@ -910,7 +911,7 @@ static int check_map_func_compatibility(struct bpf_map *map, int func_id)
 		 * don't allow any other map type to be passed into
 		 * the special func;
 		 */
-		if (bool_map != bool_func)
+		if (bool_func && bool_map != bool_func)
 			return -EINVAL;
 	}
 
diff --git a/kernel/events/core.c b/kernel/events/core.c
index b11756f9b6dc..64754bfecd70 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -5286,9 +5286,15 @@ void perf_output_sample(struct perf_output_handle *handle,
 
 	if (sample_type & PERF_SAMPLE_RAW) {
 		if (data->raw) {
-			perf_output_put(handle, data->raw->size);
-			__output_copy(handle, data->raw->data,
-					   data->raw->size);
+			u32 raw_size = data->raw->size;
+			u32 real_size = round_up(raw_size + sizeof(u32),
+						 sizeof(u64)) - sizeof(u32);
+			u64 zero = 0;
+
+			perf_output_put(handle, real_size);
+			__output_copy(handle, data->raw->data, raw_size);
+			if (real_size - raw_size)
+				__output_copy(handle, &zero, real_size - raw_size);
 		} else {
 			struct {
 				u32	size;
@@ -5420,8 +5426,7 @@ void perf_prepare_sample(struct perf_event_header *header,
 		else
 			size += sizeof(u32);
 
-		WARN_ON_ONCE(size & (sizeof(u64)-1));
-		header->size += size;
+		header->size += round_up(size, sizeof(u64));
 	}
 
 	if (sample_type & PERF_SAMPLE_BRANCH_STACK) {
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 0fe96c7c8803..47febbe7998e 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -215,6 +215,50 @@ const struct bpf_func_proto bpf_perf_event_read_proto = {
 	.arg2_type	= ARG_ANYTHING,
 };
 
+static u64 bpf_perf_event_output(u64 r1, u64 r2, u64 index, u64 r4, u64 size)
+{
+	struct pt_regs *regs = (struct pt_regs *) (long) r1;
+	struct bpf_map *map = (struct bpf_map *) (long) r2;
+	struct bpf_array *array = container_of(map, struct bpf_array, map);
+	void *data = (void *) (long) r4;
+	struct perf_sample_data sample_data;
+	struct perf_event *event;
+	struct perf_raw_record raw = {
+		.size = size,
+		.data = data,
+	};
+
+	if (unlikely(index >= array->map.max_entries))
+		return -E2BIG;
+
+	event = (struct perf_event *)array->ptrs[index];
+	if (unlikely(!event))
+		return -ENOENT;
+
+	if (unlikely(event->attr.type != PERF_TYPE_SOFTWARE ||
+		     event->attr.config != PERF_COUNT_SW_BPF_OUTPUT))
+		return -EINVAL;
+
+	if (unlikely(event->oncpu != smp_processor_id()))
+		return -EOPNOTSUPP;
+
+	perf_sample_data_init(&sample_data, 0, 0);
+	sample_data.raw = &raw;
+	perf_event_output(event, &sample_data, regs);
+	return 0;
+}
+
+static const struct bpf_func_proto bpf_perf_event_output_proto = {
+	.func		= bpf_perf_event_output,
+	.gpl_only	= false,
+	.ret_type	= RET_INTEGER,
+	.arg1_type	= ARG_PTR_TO_CTX,
+	.arg2_type	= ARG_CONST_MAP_PTR,
+	.arg3_type	= ARG_ANYTHING,
+	.arg4_type	= ARG_PTR_TO_STACK,
+	.arg5_type	= ARG_CONST_STACK_SIZE,
+};
+
 static const struct bpf_func_proto *kprobe_prog_func_proto(enum bpf_func_id func_id)
 {
 	switch (func_id) {
@@ -242,6 +286,8 @@ static const struct bpf_func_proto *kprobe_prog_func_proto(enum bpf_func_id func
 		return &bpf_get_smp_processor_id_proto;
 	case BPF_FUNC_perf_event_read:
 		return &bpf_perf_event_read_proto;
+	case BPF_FUNC_perf_event_output:
+		return &bpf_perf_event_output_proto;
 	default:
 		return NULL;
 	}
diff --git a/samples/bpf/Makefile b/samples/bpf/Makefile
index 63e7d50e6a4f..b30514514e37 100644
--- a/samples/bpf/Makefile
+++ b/samples/bpf/Makefile
@@ -13,6 +13,7 @@ hostprogs-y += tracex3
 hostprogs-y += tracex4
 hostprogs-y += tracex5
 hostprogs-y += tracex6
+hostprogs-y += trace_output
 hostprogs-y += lathist
 
 test_verifier-objs := test_verifier.o libbpf.o
@@ -27,6 +28,7 @@ tracex3-objs := bpf_load.o libbpf.o tracex3_user.o
 tracex4-objs := bpf_load.o libbpf.o tracex4_user.o
 tracex5-objs := bpf_load.o libbpf.o tracex5_user.o
 tracex6-objs := bpf_load.o libbpf.o tracex6_user.o
+trace_output-objs := bpf_load.o libbpf.o trace_output_user.o
 lathist-objs := bpf_load.o libbpf.o lathist_user.o
 
 # Tell kbuild to always build the programs
@@ -40,6 +42,7 @@ always += tracex3_kern.o
 always += tracex4_kern.o
 always += tracex5_kern.o
 always += tracex6_kern.o
+always += trace_output_kern.o
 always += tcbpf1_kern.o
 always += lathist_kern.o
 
@@ -55,6 +58,7 @@ HOSTLOADLIBES_tracex3 += -lelf
 HOSTLOADLIBES_tracex4 += -lelf -lrt
 HOSTLOADLIBES_tracex5 += -lelf
 HOSTLOADLIBES_tracex6 += -lelf
+HOSTLOADLIBES_trace_output += -lelf -lrt
 HOSTLOADLIBES_lathist += -lelf
 
 # point this to your LLVM backend with bpf support
@@ -64,3 +68,6 @@ $(obj)/%.o: $(src)/%.c
 	clang $(NOSTDINC_FLAGS) $(LINUXINCLUDE) $(EXTRA_CFLAGS) \
 		-D__KERNEL__ -Wno-unused-value -Wno-pointer-sign \
 		-O2 -emit-llvm -c $< -o -| $(LLC) -march=bpf -filetype=obj -o $@
+	clang $(NOSTDINC_FLAGS) $(LINUXINCLUDE) $(EXTRA_CFLAGS) \
+		-D__KERNEL__ -Wno-unused-value -Wno-pointer-sign \
+		-O2 -emit-llvm -c $< -o -| $(LLC) -march=bpf -filetype=asm -o $@.s
diff --git a/samples/bpf/bpf_helpers.h b/samples/bpf/bpf_helpers.h
index 21aa1b44c30c..b35c21e0b43f 100644
--- a/samples/bpf/bpf_helpers.h
+++ b/samples/bpf/bpf_helpers.h
@@ -37,6 +37,8 @@ static int (*bpf_clone_redirect)(void *ctx, int ifindex, int flags) =
 	(void *) BPF_FUNC_clone_redirect;
 static int (*bpf_redirect)(int ifindex, int flags) =
 	(void *) BPF_FUNC_redirect;
+static int (*bpf_perf_event_output)(void *ctx, void *map, int index, void *data, int size) =
+	(void *) BPF_FUNC_perf_event_output;
 
 /* llvm builtin functions that eBPF C program may use to
  * emit BPF_LD_ABS and BPF_LD_IND instructions
diff --git a/samples/bpf/trace_output_kern.c b/samples/bpf/trace_output_kern.c
new file mode 100644
index 000000000000..8d8d1ec429eb
--- /dev/null
+++ b/samples/bpf/trace_output_kern.c
@@ -0,0 +1,31 @@
+#include <linux/ptrace.h>
+#include <linux/version.h>
+#include <uapi/linux/bpf.h>
+#include "bpf_helpers.h"
+
+struct bpf_map_def SEC("maps") my_map = {
+	.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
+	.key_size = sizeof(int),
+	.value_size = sizeof(u32),
+	.max_entries = 2,
+};
+
+SEC("kprobe/sys_write")
+int bpf_prog1(struct pt_regs *ctx)
+{
+	struct S {
+		u64 pid;
+		u64 cookie;
+	} data;
+
+	memset(&data, 0, sizeof(data));
+	data.pid = bpf_get_current_pid_tgid();
+	data.cookie = 0x12345678;
+
+	bpf_perf_event_output(ctx, &my_map, 0, &data, sizeof(data));
+
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";
+u32 _version SEC("version") = LINUX_VERSION_CODE;
diff --git a/samples/bpf/trace_output_user.c b/samples/bpf/trace_output_user.c
new file mode 100644
index 000000000000..661a7d052f2c
--- /dev/null
+++ b/samples/bpf/trace_output_user.c
@@ -0,0 +1,196 @@
+/* This program is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ */
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <string.h>
+#include <fcntl.h>
+#include <poll.h>
+#include <sys/ioctl.h>
+#include <linux/perf_event.h>
+#include <linux/bpf.h>
+#include <errno.h>
+#include <assert.h>
+#include <sys/syscall.h>
+#include <sys/ioctl.h>
+#include <sys/mman.h>
+#include <time.h>
+#include <signal.h>
+#include "libbpf.h"
+#include "bpf_load.h"
+
+static int pmu_fd;
+
+int page_size;
+int page_cnt = 8;
+volatile struct perf_event_mmap_page *header;
+
+typedef void (*print_fn)(void *data, int size);
+
+static int perf_event_mmap(int fd)
+{
+	void *base;
+	int mmap_size;
+
+	page_size = getpagesize();
+	mmap_size = page_size * (page_cnt + 1);
+
+	base = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
+	if (base == MAP_FAILED) {
+		printf("mmap err\n");
+		return -1;
+	}
+
+	header = base;
+	return 0;
+}
+
+static int perf_event_poll(int fd)
+{
+	struct pollfd pfd = { .fd = fd, .events = POLLIN };
+
+	return poll(&pfd, 1, 1000);
+}
+
+struct perf_event_sample {
+	struct perf_event_header header;
+	__u32 size;
+	char data[];
+};
+
+void perf_event_read(print_fn fn)
+{
+	__u64 data_tail = header->data_tail;
+	__u64 data_head = header->data_head;
+	__u64 buffer_size = page_cnt * page_size;
+	void *base, *begin, *end;
+	char buf[256];
+
+	asm volatile("" ::: "memory"); /* in real code it should be smp_rmb() */
+	if (data_head == data_tail)
+		return;
+
+	base = ((char *)header) + page_size;
+
+	begin = base + data_tail % buffer_size;
+	end = base + data_head % buffer_size;
+
+	while (begin != end) {
+		struct perf_event_sample *e;
+
+		e = begin;
+		if (begin + e->header.size > base + buffer_size) {
+			long len = base + buffer_size - begin;
+
+			assert(len < e->header.size);
+			memcpy(buf, begin, len);
+			memcpy(buf + len, base, e->header.size - len);
+			e = (void *) buf;
+			begin = base + e->header.size - len;
+		} else if (begin + e->header.size == base + buffer_size) {
+			begin = base;
+		} else {
+			begin += e->header.size;
+		}
+
+		if (e->header.type == PERF_RECORD_SAMPLE) {
+			fn(e->data, e->size);
+		} else if (e->header.type == PERF_RECORD_LOST) {
+			struct {
+				struct perf_event_header header;
+				__u64 id;
+				__u64 lost;
+			} *lost = (void *) e;
+			printf("lost %lld events\n", lost->lost);
+		} else {
+			printf("unknown event type=%d size=%d\n",
+			       e->header.type, e->header.size);
+		}
+	}
+
+	__sync_synchronize(); /* smp_mb() */
+	header->data_tail = data_head;
+}
+
+static __u64 time_get_ns(void)
+{
+	struct timespec ts;
+
+	clock_gettime(CLOCK_MONOTONIC, &ts);
+	return ts.tv_sec * 1000000000ull + ts.tv_nsec;
+}
+
+static __u64 start_time;
+
+#define MAX_CNT 100000ll
+
+static void print_bpf_output(void *data, int size)
+{
+	static __u64 cnt;
+	struct {
+		__u64 pid;
+		__u64 cookie;
+	} *e = data;
+
+	if (e->cookie != 0x12345678) {
+		printf("BUG pid %llx cookie %llx sized %d\n",
+		       e->pid, e->cookie, size);
+		kill(0, SIGINT);
+	}
+
+	cnt++;
+
+	if (cnt == MAX_CNT) {
+		printf("recv %lld events per sec\n",
+		       MAX_CNT * 1000000000ll / (time_get_ns() - start_time));
+		kill(0, SIGINT);
+	}
+}
+
+static void test_bpf_perf_event(void)
+{
+	struct perf_event_attr attr = {
+		.sample_type = PERF_SAMPLE_RAW,
+		.type = PERF_TYPE_SOFTWARE,
+		.config = PERF_COUNT_SW_BPF_OUTPUT,
+	};
+	int key = 0;
+
+	pmu_fd = perf_event_open(&attr, -1/*pid*/, 0/*cpu*/, -1/*group_fd*/, 0);
+
+	assert(pmu_fd >= 0);
+	assert(bpf_update_elem(map_fd[0], &key, &pmu_fd, BPF_ANY) == 0);
+	ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE, 0);
+}
+
+int main(int argc, char **argv)
+{
+	char filename[256];
+	FILE *f;
+
+	snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
+
+	if (load_bpf_file(filename)) {
+		printf("%s", bpf_log_buf);
+		return 1;
+	}
+
+	test_bpf_perf_event();
+
+	if (perf_event_mmap(pmu_fd) < 0)
+		return 1;
+
+	f = popen("taskset 1 dd if=/dev/zero of=/dev/null", "r");
+	(void) f;
+
+	start_time = time_get_ns();
+	for (;;) {
+		perf_event_poll(pmu_fd);
+		perf_event_read(print_bpf_output);
+	}
+
+	return 0;
+}