xref: /dragonfly/usr.bin/systat/netbw.c (revision 9348a738)
1 /*
2  * Copyright (c) 2013 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Matthew Dillon <dillon@backplane.com>
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in
15  *    the documentation and/or other materials provided with the
16  *    distribution.
17  * 3. Neither the name of The DragonFly Project nor the names of its
18  *    contributors may be used to endorse or promote products derived
19  *    from this software without specific, prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 #include <sys/param.h>
35 #include <sys/queue.h>
36 #include <sys/tree.h>
37 #include <sys/socket.h>
38 #include <sys/socketvar.h>
39 #include <sys/protosw.h>
40 #include <sys/sysctl.h>
41 
42 #include <netinet/in.h>
43 #include <arpa/inet.h>
44 #include <net/route.h>
45 #include <netinet/in_systm.h>
46 #include <netinet/ip.h>
47 #ifdef INET6
48 #include <netinet/ip6.h>
49 #endif
50 #include <netinet/in_pcb.h>
51 #include <netinet/ip_icmp.h>
52 #include <netinet/icmp_var.h>
53 #include <netinet/ip_var.h>
54 #include <netinet/tcp.h>
55 #include <netinet/tcpip.h>
56 #include <netinet/tcp_seq.h>
57 #include <netinet/tcp_fsm.h>
58 #include <netinet/tcp_timer.h>
59 #include <netinet/tcp_var.h>
60 #include <netinet/tcp_debug.h>
61 #include <netinet/udp.h>
62 #include <netinet/udp_var.h>
63 
64 #include <err.h>
65 #include <errno.h>
66 #include <netdb.h>
67 #include <stdlib.h>
68 #include <string.h>
69 #include <nlist.h>
70 #include <paths.h>
71 #include "systat.h"
72 #include "extern.h"
73 
74 struct mytcpcb {
75 	RB_ENTRY(mytcpcb)	rb_node;
76 	int			seq;
77 	struct xtcpcb		xtcp;
78 	struct xtcpcb		last_xtcp;
79 };
80 
81 static int
82 mytcpcb_cmp(struct mytcpcb *tcp1, struct mytcpcb *tcp2)
83 {
84 	int r;
85 
86 	/*
87 	 * Low local or foreign port comes first (local has priority).
88 	 */
89 	if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_lport) >= 1024 &&
90 	    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_lport) >= 1024) {
91 		if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_fport) <
92 		    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_fport))
93 			return(-1);
94 		if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_fport) >
95 		    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_fport))
96 			return(1);
97 	}
98 
99 	if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_lport) <
100 	    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_lport))
101 		return(-1);
102 	if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_lport) >
103 	    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_lport))
104 		return(1);
105 	if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_fport) <
106 	    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_fport))
107 		return(-1);
108 	if (ntohs(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_fport) >
109 	    ntohs(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_fport))
110 		return(1);
111 
112 	/*
113 	 * Sort IPV4 vs IPV6 addresses
114 	 */
115 	if (tcp1->xtcp.xt_inp.inp_af < tcp2->xtcp.xt_inp.inp_af)
116 		return(-1);
117 	if (tcp1->xtcp.xt_inp.inp_af > tcp2->xtcp.xt_inp.inp_af)
118 		return(1);
119 
120 	/*
121 	 * Local and foreign addresses
122 	 */
123 	if (INP_ISIPV4(&tcp1->xtcp.xt_inp)) {
124 		if (ntohl(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_laddr.s_addr) <
125 		    ntohl(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_laddr.s_addr))
126 			return(-1);
127 		if (ntohl(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_laddr.s_addr) >
128 		    ntohl(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_laddr.s_addr))
129 			return(1);
130 		if (ntohl(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_faddr.s_addr) <
131 		    ntohl(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_faddr.s_addr))
132 			return(-1);
133 		if (ntohl(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie_faddr.s_addr) >
134 		    ntohl(tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie_faddr.s_addr))
135 			return(1);
136 	} else if (INP_ISIPV6(&tcp1->xtcp.xt_inp)) {
137 		r = bcmp(&tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie6_faddr,
138 			 &tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie6_faddr,
139 			 sizeof(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie6_faddr));
140 		if (r)
141 			return(r);
142 	} else {
143 		r = bcmp(&tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie6_faddr,
144 			 &tcp2->xtcp.xt_inp.inp_inc.inc_ie.ie6_faddr,
145 			 sizeof(tcp1->xtcp.xt_inp.inp_inc.inc_ie.ie6_faddr));
146 		if (r)
147 			return(r);
148 	}
149 	return(0);
150 }
151 
152 struct mytcpcb_tree;
153 RB_HEAD(mytcpcb_tree, mytcpcb);
154 RB_PROTOTYPE(mytcpcb_tree, mytcpcb, rb_node, mytcpcb_cmp);
155 RB_GENERATE(mytcpcb_tree, mytcpcb, rb_node, mytcpcb_cmp);
156 
157 static struct mytcpcb_tree mytcp_tree;
158 static struct timeval tv_curr;
159 static struct timeval tv_last;
160 static struct tcp_stats tcp_curr;
161 static struct tcp_stats tcp_last;
162 static int tcp_pcb_seq;
163 
164 static const char *numtok(double value);
165 static void netbwline(int row, struct mytcpcb *elm, double delta_time);
166 const char * netaddrstr(u_char vflags, union in_dependaddr *depaddr,
167 			u_int16_t port);
168 static void updatepcb(struct xtcpcb *xtcp);
169 
170 #define DELTARATE(field)	\
171 	((double)(tcp_curr.field - tcp_last.field) / delta_time)
172 
173 #define DELTAELM(field)		\
174 	((double)(tcp_seq_diff_t)(elm->xtcp.field -		\
175 				  elm->last_xtcp.field) /	\
176 	 delta_time)
177 
178 #define DELTAELMSCALE(field, scale)		\
179 	((double)((tcp_seq_diff_t)(elm->xtcp.field -		\
180 				   elm->last_xtcp.field) << scale) / \
181          delta_time)
182 
183 WINDOW *
184 opennetbw(void)
185 {
186 	RB_INIT(&mytcp_tree);
187 	return (subwin(stdscr, LINES-0-1, 0, 0, 0));
188 }
189 
190 void
191 closenetbw(WINDOW *w)
192 {
193 	struct mytcpcb *mytcp;
194 
195 	while ((mytcp = RB_ROOT(&mytcp_tree)) != NULL) {
196 		RB_REMOVE(mytcpcb_tree, &mytcp_tree, mytcp);
197 		free(mytcp);
198 	}
199 
200         if (w != NULL) {
201 		wclear(w);
202 		wrefresh(w);
203 		delwin(w);
204 	}
205 }
206 
207 int
208 initnetbw(void)
209 {
210 	return(1);
211 }
212 
213 void
214 fetchnetbw(void)
215 {
216 	struct tcp_stats tcp_array[SMP_MAXCPU];
217 	struct xtcpcb *tcp_pcbs;
218 	size_t npcbs;
219 	size_t len;
220 	size_t i;
221 	size_t j;
222 	size_t ncpus;
223 
224 	/*
225 	 * Extract PCB list
226 	 */
227 	len = 0;
228 	if (sysctlbyname("net.inet.tcp.pcblist", NULL, &len, NULL, 0) < 0)
229 		return;
230 	len += 128 * sizeof(tcp_pcbs[0]);
231 	tcp_pcbs = malloc(len);
232 	if (sysctlbyname("net.inet.tcp.pcblist", tcp_pcbs, &len, NULL, 0) < 0) {
233 		free(tcp_pcbs);
234 		return;
235 	}
236 	npcbs = len / sizeof(tcp_pcbs[0]);
237 	++tcp_pcb_seq;
238 
239 	for (i = 0; i < npcbs; ++i) {
240 		if (tcp_pcbs[i].xt_len != sizeof(tcp_pcbs[0]))
241 			break;
242 		updatepcb(&tcp_pcbs[i]);
243 	}
244 	free(tcp_pcbs);
245 
246 	/*
247 	 * General stats
248 	 */
249 	len = sizeof(tcp_array);
250 	if (sysctlbyname("net.inet.tcp.stats", tcp_array, &len, NULL, 0) < 0)
251 		return;
252 	ncpus = len / sizeof(tcp_array[0]);
253 	tcp_last = tcp_curr;
254 	tv_last = tv_curr;
255 	bzero(&tcp_curr, sizeof(tcp_curr));
256 	gettimeofday(&tv_curr, NULL);
257 
258 	for (i = 0; i < ncpus; ++i) {
259 		for (j = 0; j < sizeof(tcp_curr) / sizeof(u_long); ++j) {
260 			((u_long *)&tcp_curr)[j] +=
261 				((u_long *)&tcp_array[i])[j];
262 		}
263 	}
264 }
265 
266 void
267 labelnetbw(void)
268 {
269 	wmove(wnd, 0, 0);
270 	wclrtobot(wnd);
271 #if 0
272 	mvwaddstr(wnd, 0, LADDR, "Local Address");
273 	mvwaddstr(wnd, 0, FADDR, "Foreign Address");
274 	mvwaddstr(wnd, 0, PROTO, "Proto");
275 	mvwaddstr(wnd, 0, RCVCC, "Recv-Q");
276 	mvwaddstr(wnd, 0, SNDCC, "Send-Q");
277 	mvwaddstr(wnd, 0, STATE, "(state)");
278 #endif
279 }
280 
281 void
282 shownetbw(void)
283 {
284 	double delta_time;
285 	struct mytcpcb *elm;
286 	struct mytcpcb *delm;
287 	int row;
288 
289 	delta_time = (double)(tv_curr.tv_sec - tv_last.tv_sec) - 1.0 +
290 		     (tv_curr.tv_usec + 1000000 - tv_last.tv_usec) / 1e6;
291 	if (delta_time < 0.1)
292 		return;
293 
294 	mvwprintw(wnd, 0, 0,
295 		  "tcp accepts %s connects %s "
296 		  "         rcv %s snd %s rexmit %s",
297 		  numtok(DELTARATE(tcps_accepts)),
298 		  numtok(DELTARATE(tcps_connects) - DELTARATE(tcps_accepts)),
299 		  numtok(DELTARATE(tcps_rcvbyte)),
300 		  numtok(DELTARATE(tcps_sndbyte)),
301 		  numtok(DELTARATE(tcps_sndrexmitbyte)));
302 
303 	row = 2;
304 	delm = NULL;
305 	RB_FOREACH(elm, mytcpcb_tree, &mytcp_tree) {
306 		if (delm) {
307 			RB_REMOVE(mytcpcb_tree, &mytcp_tree, delm);
308 			free(delm);
309 			delm = NULL;
310 		}
311 		if (elm->seq == tcp_pcb_seq &&
312 		    (elm->xtcp.xt_socket.so_rcv.sb_cc ||
313 		     elm->xtcp.xt_socket.so_snd.sb_cc ||
314 		     DELTAELM(xt_tp.snd_max) ||
315 		     DELTAELM(xt_tp.rcv_nxt)
316 		    )) {
317 			if (row < LINES - 3)
318 				netbwline(row, elm, delta_time);
319 			++row;
320 		} else if (elm->seq != tcp_pcb_seq) {
321 			delm = elm;
322 		}
323 	}
324 	if (delm) {
325 		RB_REMOVE(mytcpcb_tree, &mytcp_tree, delm);
326 		free(delm);
327 		delm = NULL;
328 	}
329 	wmove(wnd, row, 0);
330 	wclrtobot(wnd);
331 	mvwprintw(wnd, LINES-2, 0,
332 		  "Rate/sec, "
333 		  "R=rxpend T=txpend N=nodelay T=tstmp "
334 		  "S=sack X=winscale F=fastrec");
335 }
336 
337 static
338 void
339 netbwline(int row, struct mytcpcb *elm, double delta_time)
340 {
341 	mvwprintw(wnd, row, 0,
342 		  "%s %s "
343 		  /*"rxb %s txb %s "*/
344 		  "rcv %s snd %s "
345 		  "[%c%c%c%c%c%c%c]",
346 		  netaddrstr(
347 		    elm->xtcp.xt_inp.inp_af,
348 		    &elm->xtcp.xt_inp.inp_inc.inc_ie.
349 			ie_dependladdr,
350 		    ntohs(elm->xtcp.xt_inp.inp_inc.inc_ie.ie_lport)),
351 		  netaddrstr(
352 		    elm->xtcp.xt_inp.inp_af,
353 		    &elm->xtcp.xt_inp.inp_inc.inc_ie.
354 			ie_dependfaddr,
355 		    ntohs(elm->xtcp.xt_inp.inp_inc.inc_ie.ie_fport)),
356 		/*
357 		  numtok(elm->xtcp.xt_socket.so_rcv.sb_cc),
358 		  numtok(elm->xtcp.xt_socket.so_snd.sb_cc),
359 		*/
360 		  numtok(DELTAELM(xt_tp.rcv_nxt)),
361 		  numtok(DELTAELM(xt_tp.snd_max)),
362 		  (elm->xtcp.xt_socket.so_rcv.sb_cc > 15000 ?
363 		   'R' : ' '),
364 		  (elm->xtcp.xt_socket.so_snd.sb_cc > 15000 ?
365 		   'T' : ' '),
366 		  ((elm->xtcp.xt_tp.t_flags & TF_NODELAY) ?
367 		   'N' : ' '),
368 		  ((elm->xtcp.xt_tp.t_flags & TF_RCVD_TSTMP) ?
369 		   'T' : ' '),
370 		  ((elm->xtcp.xt_tp.t_flags &
371 		   TF_SACK_PERMITTED) ?
372 		   'S' : ' '),
373 		  ((elm->xtcp.xt_tp.t_flags & TF_RCVD_SCALE) ?
374 		   'X' : ' '),
375 		  ((elm->xtcp.xt_tp.t_flags & TF_FASTRECOVERY) ?
376 				   'F' : ' ')
377 	);
378 	wclrtoeol(wnd);
379 }
380 
381 #if 0
382 int
383 cmdnetbw(const char *cmd __unused, char *args __unused)
384 {
385 	fetchnetbw();
386 	shownetbw();
387 	refresh();
388 
389 	return (0);
390 }
391 #endif
392 
393 #define MAXINDEXES 8
394 
395 static
396 const char *
397 numtok(double value)
398 {
399 	static char buf[MAXINDEXES][32];
400 	static int nexti;
401 	static const char *suffixes[] = { " ", "K", "M", "G", "T", NULL };
402 	int suffix = 0;
403 	const char *fmt;
404 
405 	while (value >= 1000.0 && suffixes[suffix+1]) {
406 		value /= 1000.0;
407 		++suffix;
408 	}
409 	nexti = (nexti + 1) % MAXINDEXES;
410 	if (value < 0.001) {
411 		fmt = "      ";
412 	} else if (value < 1.0) {
413 		fmt = "%5.3f%s";
414 	} else if (value < 10.0) {
415 		fmt = "%5.3f%s";
416 	} else if (value < 100.0) {
417 		fmt = "%5.2f%s";
418 	} else {
419 		fmt = "%5.1f%s";
420 	}
421 	snprintf(buf[nexti], sizeof(buf[nexti]),
422 		 fmt, value, suffixes[suffix]);
423 	return (buf[nexti]);
424 }
425 
426 const char *
427 netaddrstr(u_char af, union in_dependaddr *depaddr, u_int16_t port)
428 {
429 	static char buf[MAXINDEXES][64];
430 	static int nexta;
431 	char bufip[64];
432 
433 	nexta = (nexta + 1) % MAXINDEXES;
434 
435 	if (af == AF_INET) {
436 		snprintf(bufip, sizeof(bufip),
437 			 "%d.%d.%d.%d",
438 			 (ntohl(depaddr->id46_addr.ia46_addr4.s_addr) >> 24) &
439 			  255,
440 			 (ntohl(depaddr->id46_addr.ia46_addr4.s_addr) >> 16) &
441 			  255,
442 			 (ntohl(depaddr->id46_addr.ia46_addr4.s_addr) >> 8) &
443 			  255,
444 			 (ntohl(depaddr->id46_addr.ia46_addr4.s_addr) >> 0) &
445 			  255);
446 		snprintf(buf[nexta], sizeof(buf[nexta]),
447 			 "%15s:%-5d", bufip, port);
448 	} else if (af == AF_INET6) {
449 		snprintf(bufip, sizeof(bufip),
450 			 "%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x",
451 			 ntohs(depaddr->id6_addr.s6_addr16[0]),
452 			 ntohs(depaddr->id6_addr.s6_addr16[1]),
453 			 ntohs(depaddr->id6_addr.s6_addr16[2]),
454 			 ntohs(depaddr->id6_addr.s6_addr16[3]),
455 			 ntohs(depaddr->id6_addr.s6_addr16[4]),
456 			 ntohs(depaddr->id6_addr.s6_addr16[5]),
457 			 ntohs(depaddr->id6_addr.s6_addr16[6]),
458 			 ntohs(depaddr->id6_addr.s6_addr16[7]));
459 		snprintf(buf[nexta], sizeof(buf[nexta]),
460 			 "%39s:%-5d", bufip, port);
461 	} else {
462 		snprintf(bufip, sizeof(bufip), "<unknown>");
463 		snprintf(buf[nexta], sizeof(buf[nexta]),
464 			 "%15s:%-5d", bufip, port);
465 	}
466 	return (buf[nexta]);
467 }
468 
469 static
470 void
471 updatepcb(struct xtcpcb *xtcp)
472 {
473 	struct mytcpcb dummy;
474 	struct mytcpcb *elm;
475 
476 	dummy.xtcp = *xtcp;
477 	if ((elm = RB_FIND(mytcpcb_tree, &mytcp_tree, &dummy)) == NULL) {
478 		elm = malloc(sizeof(*elm));
479 		bzero(elm, sizeof(*elm));
480 		elm->xtcp = *xtcp;
481 		elm->last_xtcp = *xtcp;
482 		RB_INSERT(mytcpcb_tree, &mytcp_tree, elm);
483 	} else {
484 		elm->last_xtcp = elm->xtcp;
485 		elm->xtcp = *xtcp;
486 	}
487 	elm->seq = tcp_pcb_seq;
488 }
489