Commit 25d4dae1 authored by David S. Miller's avatar David S. Miller
Browse files

Merge branch 'XDP-redirect-tracepoints' 



Jesper Dangaard Brouer says:

====================
XDP redirect tracepoints

I feel this is as far as I can take the tracepoint infrastructure to
assist XDP monitoring.

Tracepoints comes with a base overhead of 25 nanosec for an attached
bpf_prog, and 48 nanosec for using a full perf record. This is
problematic for the XDP use-case, but it is very convenient to use the
existing perf infrastructure.

From a performance perspective, the real solution would be to attach
another bpf_prog (that understand xdp_buff), but I'm not sure we want
to introduce yet another bpf attach API for this.

One thing left is to standardize the possible err return codes, to a
limited set, to allow easier (and faster) mapping into a bpf map.
====================

Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents d0fcece7 3ffab546

include/trace/events/xdp.h

+82 −18
Original line number Diff line number Diff line
@@ -31,55 +31,119 @@ TRACE_EVENT(xdp_exception, 
	TP_ARGS(dev, xdp, act),



	TP_STRUCT__entry(

		__array(u8, prog_tag, 8)

		__field(int, prog_id)

		__field(u32, act)

		__field(int, ifindex)

	),



	TP_fast_assign(

		BUILD_BUG_ON(sizeof(__entry->prog_tag) != sizeof(xdp->tag));

		memcpy(__entry->prog_tag, xdp->tag, sizeof(xdp->tag));

		__entry->prog_id	= xdp->aux->id;

		__entry->act		= act;

		__entry->ifindex	= dev->ifindex;

	),



	TP_printk("prog=%s action=%s ifindex=%d",

		  __print_hex_str(__entry->prog_tag, 8),

	TP_printk("prog_id=%d action=%s ifindex=%d",

		  __entry->prog_id,

		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),

		  __entry->ifindex)

);



TRACE_EVENT(xdp_redirect,

DECLARE_EVENT_CLASS(xdp_redirect_template,



	TP_PROTO(const struct net_device *dev,

		 const struct bpf_prog *xdp, u32 act,

		 int to_index, int err),

		 const struct bpf_prog *xdp,

		 int to_ifindex, int err,

		 const struct bpf_map *map, u32 map_index),



	TP_ARGS(dev, xdp, act, to_index, err),

	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),



	TP_STRUCT__entry(

		__array(u8, prog_tag, 8)

		__field(int, prog_id)

		__field(u32, act)

		__field(int, ifindex)

		__field(int, to_index)

		__field(int, err)

		__field(int, to_ifindex)

		__field(u32, map_id)

		__field(int, map_index)

	),



	TP_fast_assign(

		BUILD_BUG_ON(sizeof(__entry->prog_tag) != sizeof(xdp->tag));

		memcpy(__entry->prog_tag, xdp->tag, sizeof(xdp->tag));

		__entry->act		= act;

		__entry->prog_id	= xdp->aux->id;

		__entry->act		= XDP_REDIRECT;

		__entry->ifindex	= dev->ifindex;

		__entry->to_index	= to_index;

		__entry->err		= err;

		__entry->to_ifindex	= to_ifindex;

		__entry->map_id		= map ? map->id : 0;

		__entry->map_index	= map_index;

	),



	TP_printk("prog=%s action=%s ifindex=%d to_index=%d err=%d",

		  __print_hex_str(__entry->prog_tag, 8),

	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d",

		  __entry->prog_id,

		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),

		  __entry->ifindex, __entry->to_index,

		  __entry->ifindex, __entry->to_ifindex,

		  __entry->err)

);



DEFINE_EVENT(xdp_redirect_template, xdp_redirect,

	TP_PROTO(const struct net_device *dev,

		 const struct bpf_prog *xdp,

		 int to_ifindex, int err,

		 const struct bpf_map *map, u32 map_index),

	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index)

);



DEFINE_EVENT(xdp_redirect_template, xdp_redirect_err,

	TP_PROTO(const struct net_device *dev,

		 const struct bpf_prog *xdp,

		 int to_ifindex, int err,

		 const struct bpf_map *map, u32 map_index),

	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index)

);



#define _trace_xdp_redirect(dev, xdp, to)		\

	 trace_xdp_redirect(dev, xdp, to, 0, NULL, 0);



#define _trace_xdp_redirect_err(dev, xdp, to, err)	\

	 trace_xdp_redirect_err(dev, xdp, to, err, NULL, 0);



DEFINE_EVENT_PRINT(xdp_redirect_template, xdp_redirect_map,

	TP_PROTO(const struct net_device *dev,

		 const struct bpf_prog *xdp,

		 int to_ifindex, int err,

		 const struct bpf_map *map, u32 map_index),

	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),

	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d"

		  " map_id=%d map_index=%d",

		  __entry->prog_id,

		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),

		  __entry->ifindex, __entry->to_ifindex,

		  __entry->err,

		  __entry->map_id, __entry->map_index)

);



DEFINE_EVENT_PRINT(xdp_redirect_template, xdp_redirect_map_err,

	TP_PROTO(const struct net_device *dev,

		 const struct bpf_prog *xdp,

		 int to_ifindex, int err,

		 const struct bpf_map *map, u32 map_index),

	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),

	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d"

		  " map_id=%d map_index=%d",

		  __entry->prog_id,

		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),

		  __entry->ifindex, __entry->to_ifindex,

		  __entry->err,

		  __entry->map_id, __entry->map_index)

);



#define _trace_xdp_redirect_map(dev, xdp, fwd, map, idx)		\

	 trace_xdp_redirect_map(dev, xdp, fwd ? fwd->ifindex : 0,	\

				0, map, idx);



#define _trace_xdp_redirect_map_err(dev, xdp, fwd, map, idx, err)	\

	 trace_xdp_redirect_map_err(dev, xdp, fwd ? fwd->ifindex : 0,	\

				    err, map, idx);



#endif /* _TRACE_XDP_H */



#include <trace/define_trace.h>

net/core/filter.c

+24 −13
Original line number Diff line number Diff line
@@ -2515,16 +2515,20 @@ static int xdp_do_redirect_map(struct net_device *dev, struct xdp_buff *xdp, 
	fwd = __dev_map_lookup_elem(map, index);

	if (!fwd) {

		err = -EINVAL;

		goto out;

		goto err;

	}

	if (ri->map_to_flush && ri->map_to_flush != map)

		xdp_do_flush_map();



	err = __bpf_tx_xdp(fwd, map, xdp, index);

	if (likely(!err))

	if (unlikely(err))

		goto err;



	ri->map_to_flush = map;

out:

	trace_xdp_redirect(dev, xdp_prog, XDP_REDIRECT, index, err);

	_trace_xdp_redirect_map(dev, xdp_prog, fwd, map, index);

	return 0;

err:

	_trace_xdp_redirect_map_err(dev, xdp_prog, fwd, map, index, err);

	return err;

}
@@ -2543,12 +2547,17 @@ int xdp_do_redirect(struct net_device *dev, struct xdp_buff *xdp, 
	ri->ifindex = 0;

	if (unlikely(!fwd)) {

		err = -EINVAL;

		goto out;

		goto err;

	}



	err = __bpf_tx_xdp(fwd, NULL, xdp, 0);

out:

	trace_xdp_redirect(dev, xdp_prog, XDP_REDIRECT, index, err);

	if (unlikely(err))

		goto err;



	_trace_xdp_redirect(dev, xdp_prog, index);

	return 0;

err:

	_trace_xdp_redirect_err(dev, xdp_prog, index, err);

	return err;

}

EXPORT_SYMBOL_GPL(xdp_do_redirect);
@@ -2566,23 +2575,25 @@ int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb, 
	ri->ifindex = 0;

	if (unlikely(!fwd)) {

		err = -EINVAL;

		goto out;

		goto err;

	}



	if (unlikely(!(fwd->flags & IFF_UP))) {

		err = -ENETDOWN;

		goto out;

		goto err;

	}



	len = fwd->mtu + fwd->hard_header_len + VLAN_HLEN;

	if (skb->len > len) {

		err = -EMSGSIZE;

		goto out;

		goto err;

	}



	skb->dev = fwd;

out:

	trace_xdp_redirect(dev, xdp_prog, XDP_REDIRECT, index, err);

	_trace_xdp_redirect(dev, xdp_prog, index);

	return 0;

err:

	_trace_xdp_redirect_err(dev, xdp_prog, index, err);

	return err;

}

EXPORT_SYMBOL_GPL(xdp_do_generic_redirect);

samples/bpf/Makefile

+4 −0
Original line number Diff line number Diff line
@@ -39,6 +39,7 @@ hostprogs-y += per_socket_stats_example 
hostprogs-y += load_sock_ops

hostprogs-y += xdp_redirect

hostprogs-y += xdp_redirect_map

hostprogs-y += xdp_monitor

hostprogs-y += syscall_tp



# Libbpf dependencies
@@ -83,6 +84,7 @@ test_map_in_map-objs := bpf_load.o $(LIBBPF) test_map_in_map_user.o 
per_socket_stats_example-objs := $(LIBBPF) cookie_uid_helper_example.o

xdp_redirect-objs := bpf_load.o $(LIBBPF) xdp_redirect_user.o

xdp_redirect_map-objs := bpf_load.o $(LIBBPF) xdp_redirect_map_user.o

xdp_monitor-objs := bpf_load.o $(LIBBPF) xdp_monitor_user.o

syscall_tp-objs := bpf_load.o $(LIBBPF) syscall_tp_user.o



# Tell kbuild to always build the programs
@@ -127,6 +129,7 @@ always += tcp_iw_kern.o 
always += tcp_clamp_kern.o

always += xdp_redirect_kern.o

always += xdp_redirect_map_kern.o

always += xdp_monitor_kern.o

always += syscall_tp_kern.o



HOSTCFLAGS += -I$(objtree)/usr/include
@@ -166,6 +169,7 @@ HOSTLOADLIBES_xdp_tx_iptunnel += -lelf 
HOSTLOADLIBES_test_map_in_map += -lelf

HOSTLOADLIBES_xdp_redirect += -lelf

HOSTLOADLIBES_xdp_redirect_map += -lelf

HOSTLOADLIBES_xdp_monitor += -lelf

HOSTLOADLIBES_syscall_tp += -lelf



# Allows pointing LLC/CLANG to a LLVM backend with bpf support, redefine on cmdline:

samples/bpf/xdp_monitor_kern.c

0 → 100644
+88 −0
Original line number Diff line number Diff line 
/* XDP monitor tool, based on tracepoints

 *

 *  Copyright(c) 2017 Jesper Dangaard Brouer, Red Hat Inc.

 */

#include <uapi/linux/bpf.h>

#include "bpf_helpers.h"



struct bpf_map_def SEC("maps") redirect_err_cnt = {

	.type = BPF_MAP_TYPE_PERCPU_ARRAY,

	.key_size = sizeof(u32),

	.value_size = sizeof(u64),

	.max_entries = 2,

	/* TODO: have entries for all possible errno's */

};



/* Tracepoint format: /sys/kernel/debug/tracing/events/xdp/xdp_redirect/format

 * Code in:                kernel/include/trace/events/xdp.h

 */

struct xdp_redirect_ctx {

	unsigned short common_type;	//	offset:0;  size:2; signed:0;

	unsigned char common_flags;	//	offset:2;  size:1; signed:0;

	unsigned char common_preempt_count;//	offset:3;  size:1; signed:0;

	int common_pid;			//	offset:4;  size:4; signed:1;



	int prog_id;			//	offset:8;  size:4; signed:1;

	u32 act;			//	offset:12  size:4; signed:0;

	int ifindex;			//	offset:16  size:4; signed:1;

	int err;			//	offset:20  size:4; signed:1;

	int to_ifindex;			//	offset:24  size:4; signed:1;

	u32 map_id;			//	offset:28  size:4; signed:0;

	int map_index;			//	offset:32  size:4; signed:1;

};					//	offset:36



enum {

	XDP_REDIRECT_SUCCESS = 0,

	XDP_REDIRECT_ERROR = 1

};



static __always_inline

int xdp_redirect_collect_stat(struct xdp_redirect_ctx *ctx)

{

	u32 key = XDP_REDIRECT_ERROR;

	int err = ctx->err;

	u64 *cnt;



	if (!err)

		key = XDP_REDIRECT_SUCCESS;



	cnt  = bpf_map_lookup_elem(&redirect_err_cnt, &key);

	if (!cnt)

		return 0;

	*cnt += 1;



	return 0; /* Indicate event was filtered (no further processing)*/

	/*

	 * Returning 1 here would allow e.g. a perf-record tracepoint

	 * to see and record these events, but it doesn't work well

	 * in-practice as stopping perf-record also unload this

	 * bpf_prog.  Plus, there is additional overhead of doing so.

	 */

}



SEC("tracepoint/xdp/xdp_redirect_err")

int trace_xdp_redirect_err(struct xdp_redirect_ctx *ctx)

{

	return xdp_redirect_collect_stat(ctx);

}





SEC("tracepoint/xdp/xdp_redirect_map_err")

int trace_xdp_redirect_map_err(struct xdp_redirect_ctx *ctx)

{

	return xdp_redirect_collect_stat(ctx);

}



/* Likely unloaded when prog starts */

SEC("tracepoint/xdp/xdp_redirect")

int trace_xdp_redirect(struct xdp_redirect_ctx *ctx)

{

	return xdp_redirect_collect_stat(ctx);

}



/* Likely unloaded when prog starts */

SEC("tracepoint/xdp/xdp_redirect_map")

int trace_xdp_redirect_map(struct xdp_redirect_ctx *ctx)

{

	return xdp_redirect_collect_stat(ctx);

}

samples/bpf/xdp_monitor_user.c

0 → 100644
+295 −0
Original line number Diff line number Diff line 
/* Copyright(c) 2017 Jesper Dangaard Brouer, Red Hat, Inc.

 */

static const char *__doc__=

 "XDP monitor tool, based on tracepoints\n"

;



static const char *__doc_err_only__=

 " NOTICE: Only tracking XDP redirect errors\n"

 "         Enable TX success stats via '--stats'\n"

 "         (which comes with a per packet processing overhead)\n"

;



#include <errno.h>

#include <stdio.h>

#include <stdlib.h>

#include <stdbool.h>

#include <stdint.h>

#include <string.h>

#include <ctype.h>

#include <unistd.h>

#include <locale.h>



#include <getopt.h>

#include <net/if.h>

#include <time.h>



#include "libbpf.h"

#include "bpf_load.h"

#include "bpf_util.h"



static int verbose = 1;

static bool debug = false;



static const struct option long_options[] = {

	{"help",	no_argument,		NULL, 'h' },

	{"debug",	no_argument,		NULL, 'D' },

	{"stats",	no_argument,		NULL, 'S' },

	{"sec", 	required_argument,	NULL, 's' },

	{0, 0, NULL,  0 }

};



static void usage(char *argv[])

{

	int i;

	printf("\nDOCUMENTATION:\n%s\n", __doc__);

	printf("\n");

	printf(" Usage: %s (options-see-below)\n",

	       argv[0]);

	printf(" Listing options:\n");

	for (i = 0; long_options[i].name != 0; i++) {

		printf(" --%-15s", long_options[i].name);

		if (long_options[i].flag != NULL)

			printf(" flag (internal value:%d)",

			       *long_options[i].flag);

		else

			printf("(internal short-option: -%c)",

			       long_options[i].val);

		printf("\n");

	}

	printf("\n");

}



#define NANOSEC_PER_SEC 1000000000 /* 10^9 */

__u64 gettime(void)

{

	struct timespec t;

	int res;



	res = clock_gettime(CLOCK_MONOTONIC, &t);

	if (res < 0) {

		fprintf(stderr, "Error with gettimeofday! (%i)\n", res);

		exit(EXIT_FAILURE);

	}

	return (__u64) t.tv_sec * NANOSEC_PER_SEC + t.tv_nsec;

}



enum {

	REDIR_SUCCESS = 0,

	REDIR_ERROR = 1,

};

#define REDIR_RES_MAX 2

static const char *redir_names[REDIR_RES_MAX] = {

	[REDIR_SUCCESS]	= "Success",

	[REDIR_ERROR]	= "Error",

};

static const char *err2str(int err)

{

	if (err < REDIR_RES_MAX)

		return redir_names[err];

	return NULL;

}



struct record {

	__u64 counter;

	__u64 timestamp;

};



struct stats_record {

	struct record xdp_redir[REDIR_RES_MAX];

};



static void stats_print_headers(bool err_only)

{

	if (err_only)

		printf("\n%s\n", __doc_err_only__);



	printf("%-14s %-10s %-18s %-9s\n",

	       "XDP_REDIRECT", "pps ", "pps-human-readable", "measure-period");

}



static void stats_print(struct stats_record *rec,

			struct stats_record *prev,

			bool err_only)

{

	int i = 0;



	if (err_only)

		i = REDIR_ERROR;



	for (; i < REDIR_RES_MAX; i++) {

		struct record *r = &rec->xdp_redir[i];

		struct record *p = &prev->xdp_redir[i];

		__u64 period  = 0;

		__u64 packets = 0;

		double pps = 0;

		double period_ = 0;



		if (p->timestamp) {

			packets = r->counter - p->counter;

			period  = r->timestamp - p->timestamp;

			if (period > 0) {

				period_ = ((double) period / NANOSEC_PER_SEC);

				pps = packets / period_;

			}

		}



		printf("%-14s %-10.0f %'-18.0f %f\n",

		       err2str(i), pps, pps, period_);

	}

}



static __u64 get_key32_value64_percpu(int fd, __u32 key)

{

	/* For percpu maps, userspace gets a value per possible CPU */

	unsigned int nr_cpus = bpf_num_possible_cpus();

	__u64 values[nr_cpus];

	__u64 sum = 0;

	int i;



	if ((bpf_map_lookup_elem(fd, &key, values)) != 0) {

		fprintf(stderr,

			"ERR: bpf_map_lookup_elem failed key:0x%X\n", key);

		return 0;

	}



	/* Sum values from each CPU */

	for (i = 0; i < nr_cpus; i++) {

		sum += values[i];

	}

	return sum;

}



static bool stats_collect(int fd, struct stats_record *rec)

{

	int i;



	/* TODO: Detect if someone unloaded the perf event_fd's, as

	 * this can happen by someone running perf-record -e

	 */



	for (i = 0; i < REDIR_RES_MAX; i++) {

		rec->xdp_redir[i].timestamp = gettime();

		rec->xdp_redir[i].counter = get_key32_value64_percpu(fd, i);

	}

	return true;

}



static void stats_poll(int interval, bool err_only)

{

	struct stats_record rec, prev;

	int map_fd;



	memset(&rec, 0, sizeof(rec));



	/* Trick to pretty printf with thousands separators use %' */

	setlocale(LC_NUMERIC, "en_US");



	/* Header */

	if (verbose)

		printf("\n%s", __doc__);



	/* TODO Need more advanced stats on error types */

	if (verbose)

		printf(" - Stats map: %s\n", map_data[0].name);

	map_fd = map_data[0].fd;



	stats_print_headers(err_only);

	fflush(stdout);



	while (1) {

		memcpy(&prev, &rec, sizeof(rec));

		stats_collect(map_fd, &rec);

		stats_print(&rec, &prev, err_only);

		fflush(stdout);

		sleep(interval);

	}

}



void print_bpf_prog_info(void)

{

	int i;



	/* Prog info */

	printf("Loaded BPF prog have %d bpf program(s)\n", prog_cnt);

	for (i = 0; i < prog_cnt; i++) {

		printf(" - prog_fd[%d] = fd(%d)\n", i, prog_fd[i]);

	}



	/* Maps info */

	printf("Loaded BPF prog have %d map(s)\n", map_data_count);

	for (i = 0; i < map_data_count; i++) {

		char *name = map_data[i].name;

		int fd     = map_data[i].fd;



		printf(" - map_data[%d] = fd(%d) name:%s\n", i, fd, name);

	}



	/* Event info */

	printf("Searching for (max:%d) event file descriptor(s)\n", prog_cnt);

	for (i = 0; i < prog_cnt; i++) {

		if (event_fd[i] != -1)

			printf(" - event_fd[%d] = fd(%d)\n", i, event_fd[i]);

	}

}



int main(int argc, char **argv)

{

	int longindex = 0, opt;

	int ret = EXIT_SUCCESS;

	char bpf_obj_file[256];



	/* Default settings: */

	bool errors_only = true;

	int interval = 2;



	snprintf(bpf_obj_file, sizeof(bpf_obj_file), "%s_kern.o", argv[0]);



	/* Parse commands line args */

	while ((opt = getopt_long(argc, argv, "h",

				  long_options, &longindex)) != -1) {

		switch (opt) {

		case 'D':

			debug = true;

			break;

		case 'S':

			errors_only = false;

			break;

		case 's':

			interval = atoi(optarg);

			break;

		case 'h':

		default:

			usage(argv);

			return EXIT_FAILURE;

		}

	}



	if (load_bpf_file(bpf_obj_file)) {

		printf("ERROR - bpf_log_buf: %s", bpf_log_buf);

		return 1;

	}

	if (!prog_fd[0]) {

		printf("ERROR - load_bpf_file: %s\n", strerror(errno));

		return 1;

	}



	if (debug) {

		print_bpf_prog_info();

	}



	/* Unload/stop tracepoint event by closing fd's */

	if (errors_only) {

		/* The prog_fd[i] and event_fd[i] depend on the

		 * order the functions was defined in _kern.c

		 */

		close(event_fd[2]); /* tracepoint/xdp/xdp_redirect */

		close(prog_fd[2]);  /* func: trace_xdp_redirect */

		close(event_fd[3]); /* tracepoint/xdp/xdp_redirect_map */

		close(prog_fd[3]);  /* func: trace_xdp_redirect_map */

	}



	stats_poll(interval, errors_only);



	return ret;

}

CRA Git | Maintained and supported by SUSTech CRA and CCSE