1 /*
2 * ss.c "sockstat", socket statistics
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 */
11
12 #include <stdio.h>
13 #include <stdlib.h>
14 #include <unistd.h>
15 #include <syslog.h>
16 #include <fcntl.h>
17 #include <sys/ioctl.h>
18 #include <sys/socket.h>
19 #include <sys/uio.h>
20 #include <netinet/in.h>
21 #include <string.h>
22 #include <errno.h>
23 #include <netdb.h>
24 #include <arpa/inet.h>
25 #include <dirent.h>
26 #include <fnmatch.h>
27 #include <getopt.h>
28 #include <stdbool.h>
29
30 #include "utils.h"
31 #include "rt_names.h"
32 #include "ll_map.h"
33 #include "libnetlink.h"
34 #include "namespace.h"
35 #include "SNAPSHOT.h"
36
37 #include <linux/tcp.h>
38 #include <linux/sock_diag.h>
39 #include <linux/inet_diag.h>
40 #include <linux/unix_diag.h>
41 #include <linux/netdevice.h> /* for MAX_ADDR_LEN */
42 #include <linux/filter.h>
43 #include <linux/packet_diag.h>
44 #include <linux/netlink_diag.h>
45
46 #define MAGIC_SEQ 123456
47
48 #define DIAG_REQUEST(_req, _r) \
49 struct { \
50 struct nlmsghdr nlh; \
51 _r; \
52 } _req = { \
53 .nlh = { \
54 .nlmsg_type = SOCK_DIAG_BY_FAMILY, \
55 .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,\
56 .nlmsg_seq = MAGIC_SEQ, \
57 .nlmsg_len = sizeof(_req), \
58 }, \
59 }
60
61 #if HAVE_SELINUX
62 #include <selinux/selinux.h>
63 #else
64 /* Stubs for SELinux functions */
is_selinux_enabled(void)65 static int is_selinux_enabled(void)
66 {
67 return -1;
68 }
69
getpidcon(pid_t pid,char ** context)70 static int getpidcon(pid_t pid, char **context)
71 {
72 *context = NULL;
73 return -1;
74 }
75
getfilecon(char * path,char ** context)76 static int getfilecon(char *path, char **context)
77 {
78 *context = NULL;
79 return -1;
80 }
81
security_get_initial_context(char * name,char ** context)82 static int security_get_initial_context(char *name, char **context)
83 {
84 *context = NULL;
85 return -1;
86 }
87 #endif
88
89 int resolve_hosts = 0;
90 int resolve_services = 1;
91 int preferred_family = AF_UNSPEC;
92 int show_options = 0;
93 int show_details = 0;
94 int show_users = 0;
95 int show_mem = 0;
96 int show_tcpinfo = 0;
97 int show_bpf = 0;
98 int show_proc_ctx = 0;
99 int show_sock_ctx = 0;
100 /* If show_users & show_proc_ctx only do user_ent_hash_build() once */
101 int user_ent_hash_build_init = 0;
102 int follow_events = 0;
103
104 int netid_width;
105 int state_width;
106 int addrp_width;
107 int addr_width;
108 int serv_width;
109 int screen_width;
110
111 static const char *TCP_PROTO = "tcp";
112 static const char *UDP_PROTO = "udp";
113 static const char *RAW_PROTO = "raw";
114 static const char *dg_proto = NULL;
115
116 enum
117 {
118 TCP_DB,
119 DCCP_DB,
120 UDP_DB,
121 RAW_DB,
122 UNIX_DG_DB,
123 UNIX_ST_DB,
124 UNIX_SQ_DB,
125 PACKET_DG_DB,
126 PACKET_R_DB,
127 NETLINK_DB,
128 MAX_DB
129 };
130
131 #define PACKET_DBM ((1<<PACKET_DG_DB)|(1<<PACKET_R_DB))
132 #define UNIX_DBM ((1<<UNIX_DG_DB)|(1<<UNIX_ST_DB)|(1<<UNIX_SQ_DB))
133 #define ALL_DB ((1<<MAX_DB)-1)
134 #define INET_DBM ((1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB)|(1<<RAW_DB))
135
136 enum {
137 SS_UNKNOWN,
138 SS_ESTABLISHED,
139 SS_SYN_SENT,
140 SS_SYN_RECV,
141 SS_FIN_WAIT1,
142 SS_FIN_WAIT2,
143 SS_TIME_WAIT,
144 SS_CLOSE,
145 SS_CLOSE_WAIT,
146 SS_LAST_ACK,
147 SS_LISTEN,
148 SS_CLOSING,
149 SS_MAX
150 };
151
152 #define SS_ALL ((1 << SS_MAX) - 1)
153 #define SS_CONN (SS_ALL & ~((1<<SS_LISTEN)|(1<<SS_CLOSE)|(1<<SS_TIME_WAIT)|(1<<SS_SYN_RECV)))
154
155 #include "ssfilter.h"
156
157 struct filter
158 {
159 int dbs;
160 int states;
161 int families;
162 struct ssfilter *f;
163 bool kill;
164 };
165
166 static const struct filter default_dbs[MAX_DB] = {
167 [TCP_DB] = {
168 .states = SS_CONN,
169 .families = (1 << AF_INET) | (1 << AF_INET6),
170 },
171 [DCCP_DB] = {
172 .states = SS_CONN,
173 .families = (1 << AF_INET) | (1 << AF_INET6),
174 },
175 [UDP_DB] = {
176 .states = (1 << SS_ESTABLISHED),
177 .families = (1 << AF_INET) | (1 << AF_INET6),
178 },
179 [RAW_DB] = {
180 .states = (1 << SS_ESTABLISHED),
181 .families = (1 << AF_INET) | (1 << AF_INET6),
182 },
183 [UNIX_DG_DB] = {
184 .states = (1 << SS_CLOSE),
185 .families = (1 << AF_UNIX),
186 },
187 [UNIX_ST_DB] = {
188 .states = SS_CONN,
189 .families = (1 << AF_UNIX),
190 },
191 [UNIX_SQ_DB] = {
192 .states = SS_CONN,
193 .families = (1 << AF_UNIX),
194 },
195 [PACKET_DG_DB] = {
196 .states = (1 << SS_CLOSE),
197 .families = (1 << AF_PACKET),
198 },
199 [PACKET_R_DB] = {
200 .states = (1 << SS_CLOSE),
201 .families = (1 << AF_PACKET),
202 },
203 [NETLINK_DB] = {
204 .states = (1 << SS_CLOSE),
205 .families = (1 << AF_NETLINK),
206 },
207 };
208
209 static const struct filter default_afs[AF_MAX] = {
210 [AF_INET] = {
211 .dbs = INET_DBM,
212 .states = SS_CONN,
213 },
214 [AF_INET6] = {
215 .dbs = INET_DBM,
216 .states = SS_CONN,
217 },
218 [AF_UNIX] = {
219 .dbs = UNIX_DBM,
220 .states = SS_CONN,
221 },
222 [AF_PACKET] = {
223 .dbs = PACKET_DBM,
224 .states = (1 << SS_CLOSE),
225 },
226 [AF_NETLINK] = {
227 .dbs = (1 << NETLINK_DB),
228 .states = (1 << SS_CLOSE),
229 },
230 };
231
232 static int do_default = 1;
233 static struct filter current_filter;
234
filter_db_set(struct filter * f,int db)235 static void filter_db_set(struct filter *f, int db)
236 {
237 f->states |= default_dbs[db].states;
238 f->dbs |= 1 << db;
239 do_default = 0;
240 }
241
filter_af_set(struct filter * f,int af)242 static void filter_af_set(struct filter *f, int af)
243 {
244 f->states |= default_afs[af].states;
245 f->families |= 1 << af;
246 do_default = 0;
247 preferred_family = af;
248 }
249
filter_af_get(struct filter * f,int af)250 static int filter_af_get(struct filter *f, int af)
251 {
252 return f->families & (1 << af);
253 }
254
filter_default_dbs(struct filter * f)255 static void filter_default_dbs(struct filter *f)
256 {
257 filter_db_set(f, UDP_DB);
258 filter_db_set(f, DCCP_DB);
259 filter_db_set(f, TCP_DB);
260 filter_db_set(f, RAW_DB);
261 filter_db_set(f, UNIX_ST_DB);
262 filter_db_set(f, UNIX_DG_DB);
263 filter_db_set(f, UNIX_SQ_DB);
264 filter_db_set(f, PACKET_R_DB);
265 filter_db_set(f, PACKET_DG_DB);
266 filter_db_set(f, NETLINK_DB);
267 }
268
filter_states_set(struct filter * f,int states)269 static void filter_states_set(struct filter *f, int states)
270 {
271 if (states)
272 f->states = (f->states | states) & states;
273 }
274
filter_merge_defaults(struct filter * f)275 static void filter_merge_defaults(struct filter *f)
276 {
277 int db;
278 int af;
279
280 for (db = 0; db < MAX_DB; db++) {
281 if (!(f->dbs & (1 << db)))
282 continue;
283
284 if (!(default_dbs[db].families & f->families))
285 f->families |= default_dbs[db].families;
286 }
287 for (af = 0; af < AF_MAX; af++) {
288 if (!(f->families & (1 << af)))
289 continue;
290
291 if (!(default_afs[af].dbs & f->dbs))
292 f->dbs |= default_afs[af].dbs;
293 }
294 }
295
generic_proc_open(const char * env,const char * name)296 static FILE *generic_proc_open(const char *env, const char *name)
297 {
298 const char *p = getenv(env);
299 char store[128];
300
301 if (!p) {
302 p = getenv("PROC_ROOT") ? : "/proc";
303 snprintf(store, sizeof(store)-1, "%s/%s", p, name);
304 p = store;
305 }
306
307 return fopen(p, "r");
308 }
309
net_tcp_open(void)310 static FILE *net_tcp_open(void)
311 {
312 return generic_proc_open("PROC_NET_TCP", "net/tcp");
313 }
314
net_tcp6_open(void)315 static FILE *net_tcp6_open(void)
316 {
317 return generic_proc_open("PROC_NET_TCP6", "net/tcp6");
318 }
319
net_udp_open(void)320 static FILE *net_udp_open(void)
321 {
322 return generic_proc_open("PROC_NET_UDP", "net/udp");
323 }
324
net_udp6_open(void)325 static FILE *net_udp6_open(void)
326 {
327 return generic_proc_open("PROC_NET_UDP6", "net/udp6");
328 }
329
net_raw_open(void)330 static FILE *net_raw_open(void)
331 {
332 return generic_proc_open("PROC_NET_RAW", "net/raw");
333 }
334
net_raw6_open(void)335 static FILE *net_raw6_open(void)
336 {
337 return generic_proc_open("PROC_NET_RAW6", "net/raw6");
338 }
339
net_unix_open(void)340 static FILE *net_unix_open(void)
341 {
342 return generic_proc_open("PROC_NET_UNIX", "net/unix");
343 }
344
net_packet_open(void)345 static FILE *net_packet_open(void)
346 {
347 return generic_proc_open("PROC_NET_PACKET", "net/packet");
348 }
349
net_netlink_open(void)350 static FILE *net_netlink_open(void)
351 {
352 return generic_proc_open("PROC_NET_NETLINK", "net/netlink");
353 }
354
slabinfo_open(void)355 static FILE *slabinfo_open(void)
356 {
357 return generic_proc_open("PROC_SLABINFO", "slabinfo");
358 }
359
net_sockstat_open(void)360 static FILE *net_sockstat_open(void)
361 {
362 return generic_proc_open("PROC_NET_SOCKSTAT", "net/sockstat");
363 }
364
net_sockstat6_open(void)365 static FILE *net_sockstat6_open(void)
366 {
367 return generic_proc_open("PROC_NET_SOCKSTAT6", "net/sockstat6");
368 }
369
net_snmp_open(void)370 static FILE *net_snmp_open(void)
371 {
372 return generic_proc_open("PROC_NET_SNMP", "net/snmp");
373 }
374
ephemeral_ports_open(void)375 static FILE *ephemeral_ports_open(void)
376 {
377 return generic_proc_open("PROC_IP_LOCAL_PORT_RANGE", "sys/net/ipv4/ip_local_port_range");
378 }
379
380 struct user_ent {
381 struct user_ent *next;
382 unsigned int ino;
383 int pid;
384 int fd;
385 char *process;
386 char *process_ctx;
387 char *socket_ctx;
388 };
389
390 #define USER_ENT_HASH_SIZE 256
391 struct user_ent *user_ent_hash[USER_ENT_HASH_SIZE];
392
user_ent_hashfn(unsigned int ino)393 static int user_ent_hashfn(unsigned int ino)
394 {
395 int val = (ino >> 24) ^ (ino >> 16) ^ (ino >> 8) ^ ino;
396
397 return val & (USER_ENT_HASH_SIZE - 1);
398 }
399
user_ent_add(unsigned int ino,char * process,int pid,int fd,char * proc_ctx,char * sock_ctx)400 static void user_ent_add(unsigned int ino, char *process,
401 int pid, int fd,
402 char *proc_ctx,
403 char *sock_ctx)
404 {
405 struct user_ent *p, **pp;
406
407 p = malloc(sizeof(struct user_ent));
408 if (!p) {
409 fprintf(stderr, "ss: failed to malloc buffer\n");
410 abort();
411 }
412 p->next = NULL;
413 p->ino = ino;
414 p->pid = pid;
415 p->fd = fd;
416 p->process = strdup(process);
417 p->process_ctx = strdup(proc_ctx);
418 p->socket_ctx = strdup(sock_ctx);
419
420 pp = &user_ent_hash[user_ent_hashfn(ino)];
421 p->next = *pp;
422 *pp = p;
423 }
424
user_ent_destroy(void)425 static void user_ent_destroy(void)
426 {
427 struct user_ent *p, *p_next;
428 int cnt = 0;
429
430 while (cnt != USER_ENT_HASH_SIZE) {
431 p = user_ent_hash[cnt];
432 while (p) {
433 free(p->process);
434 free(p->process_ctx);
435 free(p->socket_ctx);
436 p_next = p->next;
437 free(p);
438 p = p_next;
439 }
440 cnt++;
441 }
442 }
443
user_ent_hash_build(void)444 static void user_ent_hash_build(void)
445 {
446 const char *root = getenv("PROC_ROOT") ? : "/proc/";
447 struct dirent *d;
448 char name[1024];
449 int nameoff;
450 DIR *dir;
451 char *pid_context;
452 char *sock_context;
453 const char *no_ctx = "unavailable";
454
455 /* If show_users & show_proc_ctx set only do this once */
456 if (user_ent_hash_build_init != 0)
457 return;
458
459 user_ent_hash_build_init = 1;
460
461 strncpy(name, root, sizeof(name)-1);
462 name[sizeof(name)-1] = 0;
463
464 if (strlen(name) == 0 || name[strlen(name)-1] != '/')
465 strcat(name, "/");
466
467 nameoff = strlen(name);
468
469 dir = opendir(name);
470 if (!dir)
471 return;
472
473 while ((d = readdir(dir)) != NULL) {
474 struct dirent *d1;
475 char process[16];
476 char *p;
477 int pid, pos;
478 DIR *dir1;
479 char crap;
480
481 if (sscanf(d->d_name, "%d%c", &pid, &crap) != 1)
482 continue;
483
484 if (getpidcon(pid, &pid_context) != 0)
485 pid_context = strdup(no_ctx);
486
487 snprintf(name + nameoff, sizeof(name) - nameoff, "%d/fd/", pid);
488 pos = strlen(name);
489 if ((dir1 = opendir(name)) == NULL) {
490 free(pid_context);
491 continue;
492 }
493
494 process[0] = '\0';
495 p = process;
496
497 while ((d1 = readdir(dir1)) != NULL) {
498 const char *pattern = "socket:[";
499 unsigned int ino;
500 char lnk[64];
501 int fd;
502 ssize_t link_len;
503 char tmp[1024];
504
505 if (sscanf(d1->d_name, "%d%c", &fd, &crap) != 1)
506 continue;
507
508 snprintf(name+pos, sizeof(name) - pos, "%d", fd);
509
510 link_len = readlink(name, lnk, sizeof(lnk)-1);
511 if (link_len == -1)
512 continue;
513 lnk[link_len] = '\0';
514
515 if (strncmp(lnk, pattern, strlen(pattern)))
516 continue;
517
518 sscanf(lnk, "socket:[%u]", &ino);
519
520 snprintf(tmp, sizeof(tmp), "%s/%d/fd/%s",
521 root, pid, d1->d_name);
522
523 if (getfilecon(tmp, &sock_context) <= 0)
524 sock_context = strdup(no_ctx);
525
526 if (*p == '\0') {
527 FILE *fp;
528
529 snprintf(tmp, sizeof(tmp), "%s/%d/stat",
530 root, pid);
531 if ((fp = fopen(tmp, "r")) != NULL) {
532 if (fscanf(fp, "%*d (%[^)])", p) < 1)
533 ; /* ignore */
534 fclose(fp);
535 }
536 }
537 user_ent_add(ino, p, pid, fd,
538 pid_context, sock_context);
539 free(sock_context);
540 }
541 free(pid_context);
542 closedir(dir1);
543 }
544 closedir(dir);
545 }
546
547 enum entry_types {
548 USERS,
549 PROC_CTX,
550 PROC_SOCK_CTX
551 };
552
553 #define ENTRY_BUF_SIZE 512
find_entry(unsigned ino,char ** buf,int type)554 static int find_entry(unsigned ino, char **buf, int type)
555 {
556 struct user_ent *p;
557 int cnt = 0;
558 char *ptr;
559 char *new_buf;
560 int len, new_buf_len;
561 int buf_used = 0;
562 int buf_len = 0;
563
564 if (!ino)
565 return 0;
566
567 p = user_ent_hash[user_ent_hashfn(ino)];
568 ptr = *buf = NULL;
569 while (p) {
570 if (p->ino != ino)
571 goto next;
572
573 while (1) {
574 ptr = *buf + buf_used;
575 switch (type) {
576 case USERS:
577 len = snprintf(ptr, buf_len - buf_used,
578 "(\"%s\",pid=%d,fd=%d),",
579 p->process, p->pid, p->fd);
580 break;
581 case PROC_CTX:
582 len = snprintf(ptr, buf_len - buf_used,
583 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d),",
584 p->process, p->pid,
585 p->process_ctx, p->fd);
586 break;
587 case PROC_SOCK_CTX:
588 len = snprintf(ptr, buf_len - buf_used,
589 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d,sock_ctx=%s),",
590 p->process, p->pid,
591 p->process_ctx, p->fd,
592 p->socket_ctx);
593 break;
594 default:
595 fprintf(stderr, "ss: invalid type: %d\n", type);
596 abort();
597 }
598
599 if (len < 0 || len >= buf_len - buf_used) {
600 new_buf_len = buf_len + ENTRY_BUF_SIZE;
601 new_buf = realloc(*buf, new_buf_len);
602 if (!new_buf) {
603 fprintf(stderr, "ss: failed to malloc buffer\n");
604 abort();
605 }
606 *buf = new_buf;
607 buf_len = new_buf_len;
608 continue;
609 } else {
610 buf_used += len;
611 break;
612 }
613 }
614 cnt++;
615 next:
616 p = p->next;
617 }
618 if (buf_used) {
619 ptr = *buf + buf_used;
620 ptr[-1] = '\0';
621 }
622 return cnt;
623 }
624
625 /* Get stats from slab */
626
627 struct slabstat
628 {
629 int socks;
630 int tcp_ports;
631 int tcp_tws;
632 int tcp_syns;
633 int skbs;
634 };
635
636 static struct slabstat slabstat;
637
638 static const char *slabstat_ids[] =
639 {
640 "sock",
641 "tcp_bind_bucket",
642 "tcp_tw_bucket",
643 "tcp_open_request",
644 "skbuff_head_cache",
645 };
646
get_slabstat(struct slabstat * s)647 static int get_slabstat(struct slabstat *s)
648 {
649 char buf[256];
650 FILE *fp;
651 int cnt;
652 static int slabstat_valid;
653
654 if (slabstat_valid)
655 return 0;
656
657 memset(s, 0, sizeof(*s));
658
659 fp = slabinfo_open();
660 if (!fp)
661 return -1;
662
663 cnt = sizeof(*s)/sizeof(int);
664
665 if (!fgets(buf, sizeof(buf), fp)) {
666 fclose(fp);
667 return -1;
668 }
669 while(fgets(buf, sizeof(buf), fp) != NULL) {
670 int i;
671 for (i=0; i<sizeof(slabstat_ids)/sizeof(slabstat_ids[0]); i++) {
672 if (memcmp(buf, slabstat_ids[i], strlen(slabstat_ids[i])) == 0) {
673 sscanf(buf, "%*s%d", ((int *)s) + i);
674 cnt--;
675 break;
676 }
677 }
678 if (cnt <= 0)
679 break;
680 }
681
682 slabstat_valid = 1;
683
684 fclose(fp);
685 return 0;
686 }
687
cookie_sk_get(const uint32_t * cookie)688 static unsigned long long cookie_sk_get(const uint32_t *cookie)
689 {
690 return (((unsigned long long)cookie[1] << 31) << 1) | cookie[0];
691 }
692
693 static const char *sstate_name[] = {
694 "UNKNOWN",
695 [SS_ESTABLISHED] = "ESTAB",
696 [SS_SYN_SENT] = "SYN-SENT",
697 [SS_SYN_RECV] = "SYN-RECV",
698 [SS_FIN_WAIT1] = "FIN-WAIT-1",
699 [SS_FIN_WAIT2] = "FIN-WAIT-2",
700 [SS_TIME_WAIT] = "TIME-WAIT",
701 [SS_CLOSE] = "UNCONN",
702 [SS_CLOSE_WAIT] = "CLOSE-WAIT",
703 [SS_LAST_ACK] = "LAST-ACK",
704 [SS_LISTEN] = "LISTEN",
705 [SS_CLOSING] = "CLOSING",
706 };
707
708 static const char *sstate_namel[] = {
709 "UNKNOWN",
710 [SS_ESTABLISHED] = "established",
711 [SS_SYN_SENT] = "syn-sent",
712 [SS_SYN_RECV] = "syn-recv",
713 [SS_FIN_WAIT1] = "fin-wait-1",
714 [SS_FIN_WAIT2] = "fin-wait-2",
715 [SS_TIME_WAIT] = "time-wait",
716 [SS_CLOSE] = "unconnected",
717 [SS_CLOSE_WAIT] = "close-wait",
718 [SS_LAST_ACK] = "last-ack",
719 [SS_LISTEN] = "listening",
720 [SS_CLOSING] = "closing",
721 };
722
723 struct sockstat
724 {
725 struct sockstat *next;
726 unsigned int type;
727 uint16_t prot;
728 inet_prefix local;
729 inet_prefix remote;
730 int lport;
731 int rport;
732 int state;
733 int rq, wq;
734 unsigned ino;
735 unsigned uid;
736 int refcnt;
737 unsigned int iface;
738 unsigned long long sk;
739 char *name;
740 char *peer_name;
741 };
742
743 struct dctcpstat
744 {
745 unsigned int ce_state;
746 unsigned int alpha;
747 unsigned int ab_ecn;
748 unsigned int ab_tot;
749 bool enabled;
750 };
751
752 struct tcpstat
753 {
754 struct sockstat ss;
755 int timer;
756 int timeout;
757 int probes;
758 char cong_alg[16];
759 double rto, ato, rtt, rttvar;
760 int qack, cwnd, ssthresh, backoff;
761 double send_bps;
762 int snd_wscale;
763 int rcv_wscale;
764 int mss;
765 unsigned int lastsnd;
766 unsigned int lastrcv;
767 unsigned int lastack;
768 double pacing_rate;
769 double pacing_rate_max;
770 unsigned long long bytes_acked;
771 unsigned long long bytes_received;
772 unsigned int segs_out;
773 unsigned int segs_in;
774 unsigned int unacked;
775 unsigned int retrans;
776 unsigned int retrans_total;
777 unsigned int lost;
778 unsigned int sacked;
779 unsigned int fackets;
780 unsigned int reordering;
781 double rcv_rtt;
782 int rcv_space;
783 bool has_ts_opt;
784 bool has_sack_opt;
785 bool has_ecn_opt;
786 bool has_ecnseen_opt;
787 bool has_fastopen_opt;
788 bool has_wscale_opt;
789 struct dctcpstat *dctcp;
790 };
791
sock_state_print(struct sockstat * s,const char * sock_name)792 static void sock_state_print(struct sockstat *s, const char *sock_name)
793 {
794 if (netid_width)
795 printf("%-*s ", netid_width, sock_name);
796 if (state_width)
797 printf("%-*s ", state_width, sstate_name[s->state]);
798
799 printf("%-6d %-6d ", s->rq, s->wq);
800 }
801
sock_details_print(struct sockstat * s)802 static void sock_details_print(struct sockstat *s)
803 {
804 if (s->uid)
805 printf(" uid:%u", s->uid);
806
807 printf(" ino:%u", s->ino);
808 printf(" sk:%llx", s->sk);
809 }
810
sock_addr_print_width(int addr_len,const char * addr,char * delim,int port_len,const char * port,const char * ifname)811 static void sock_addr_print_width(int addr_len, const char *addr, char *delim,
812 int port_len, const char *port, const char *ifname)
813 {
814 if (ifname) {
815 printf("%*s%%%s%s%-*s ", addr_len, addr, ifname, delim,
816 port_len, port);
817 }
818 else {
819 printf("%*s%s%-*s ", addr_len, addr, delim, port_len, port);
820 }
821 }
822
sock_addr_print(const char * addr,char * delim,const char * port,const char * ifname)823 static void sock_addr_print(const char *addr, char *delim, const char *port,
824 const char *ifname)
825 {
826 sock_addr_print_width(addr_width, addr, delim, serv_width, port, ifname);
827 }
828
829 static const char *tmr_name[] = {
830 "off",
831 "on",
832 "keepalive",
833 "timewait",
834 "persist",
835 "unknown"
836 };
837
print_ms_timer(int timeout)838 static const char *print_ms_timer(int timeout)
839 {
840 static char buf[64];
841 int secs, msecs, minutes;
842 if (timeout < 0)
843 timeout = 0;
844 secs = timeout/1000;
845 minutes = secs/60;
846 secs = secs%60;
847 msecs = timeout%1000;
848 buf[0] = 0;
849 if (minutes) {
850 msecs = 0;
851 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
852 if (minutes > 9)
853 secs = 0;
854 }
855 if (secs) {
856 if (secs > 9)
857 msecs = 0;
858 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
859 }
860 if (msecs)
861 sprintf(buf+strlen(buf), "%03dms", msecs);
862 return buf;
863 }
864
865 struct scache {
866 struct scache *next;
867 int port;
868 char *name;
869 const char *proto;
870 };
871
872 struct scache *rlist;
873
init_service_resolver(void)874 static void init_service_resolver(void)
875 {
876 char buf[128];
877 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
878
879 if (!fp)
880 return;
881
882 if (!fgets(buf, sizeof(buf), fp)) {
883 pclose(fp);
884 return;
885 }
886 while (fgets(buf, sizeof(buf), fp) != NULL) {
887 unsigned int progn, port;
888 char proto[128], prog[128] = "rpc.";
889 struct scache *c;
890
891 if (sscanf(buf, "%u %*d %s %u %s",
892 &progn, proto, &port, prog+4) != 4)
893 continue;
894
895 if (!(c = malloc(sizeof(*c))))
896 continue;
897
898 c->port = port;
899 c->name = strdup(prog);
900 if (strcmp(proto, TCP_PROTO) == 0)
901 c->proto = TCP_PROTO;
902 else if (strcmp(proto, UDP_PROTO) == 0)
903 c->proto = UDP_PROTO;
904 else
905 c->proto = NULL;
906 c->next = rlist;
907 rlist = c;
908 }
909 pclose(fp);
910 }
911
912 /* Even do not try default linux ephemeral port ranges:
913 * default /etc/services contains so much of useless crap
914 * wouldbe "allocated" to this area that resolution
915 * is really harmful. I shrug each time when seeing
916 * "socks" or "cfinger" in dumps.
917 */
is_ephemeral(int port)918 static int is_ephemeral(int port)
919 {
920 static int min = 0, max = 0;
921
922 if (!min) {
923 FILE *f = ephemeral_ports_open();
924 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
925 min = 1024;
926 max = 4999;
927 }
928 if (f)
929 fclose(f);
930 }
931 return port >= min && port <= max;
932 }
933
934
__resolve_service(int port)935 static const char *__resolve_service(int port)
936 {
937 struct scache *c;
938
939 for (c = rlist; c; c = c->next) {
940 if (c->port == port && c->proto == dg_proto)
941 return c->name;
942 }
943
944 if (!is_ephemeral(port)) {
945 static int notfirst;
946 struct servent *se;
947 if (!notfirst) {
948 setservent(1);
949 notfirst = 1;
950 }
951 se = getservbyport(htons(port), dg_proto);
952 if (se)
953 return se->s_name;
954 }
955
956 return NULL;
957 }
958
959 #define SCACHE_BUCKETS 1024
960 static struct scache *cache_htab[SCACHE_BUCKETS];
961
resolve_service(int port)962 static const char *resolve_service(int port)
963 {
964 static char buf[128];
965 struct scache *c;
966 const char *res;
967 int hash;
968
969 if (port == 0) {
970 buf[0] = '*';
971 buf[1] = 0;
972 return buf;
973 }
974
975 if (!resolve_services)
976 goto do_numeric;
977
978 if (dg_proto == RAW_PROTO)
979 return inet_proto_n2a(port, buf, sizeof(buf));
980
981
982 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
983
984 for (c = cache_htab[hash]; c; c = c->next) {
985 if (c->port == port && c->proto == dg_proto)
986 goto do_cache;
987 }
988
989 c = malloc(sizeof(*c));
990 if (!c)
991 goto do_numeric;
992 res = __resolve_service(port);
993 c->port = port;
994 c->name = res ? strdup(res) : NULL;
995 c->proto = dg_proto;
996 c->next = cache_htab[hash];
997 cache_htab[hash] = c;
998
999 do_cache:
1000 if (c->name)
1001 return c->name;
1002
1003 do_numeric:
1004 sprintf(buf, "%u", port);
1005 return buf;
1006 }
1007
inet_addr_print(const inet_prefix * a,int port,unsigned int ifindex)1008 static void inet_addr_print(const inet_prefix *a, int port, unsigned int ifindex)
1009 {
1010 char buf[1024];
1011 const char *ap = buf;
1012 int est_len = addr_width;
1013 const char *ifname = NULL;
1014
1015 if (a->family == AF_INET) {
1016 if (a->data[0] == 0) {
1017 buf[0] = '*';
1018 buf[1] = 0;
1019 } else {
1020 ap = format_host(AF_INET, 4, a->data, buf, sizeof(buf));
1021 }
1022 } else {
1023 ap = format_host(a->family, 16, a->data, buf, sizeof(buf));
1024 est_len = strlen(ap);
1025 if (est_len <= addr_width)
1026 est_len = addr_width;
1027 else
1028 est_len = addr_width + ((est_len-addr_width+3)/4)*4;
1029 }
1030
1031 if (ifindex) {
1032 ifname = ll_index_to_name(ifindex);
1033 est_len -= strlen(ifname) + 1; /* +1 for percent char */
1034 if (est_len < 0)
1035 est_len = 0;
1036 }
1037
1038 sock_addr_print_width(est_len, ap, ":", serv_width, resolve_service(port),
1039 ifname);
1040 }
1041
1042 struct aafilter
1043 {
1044 inet_prefix addr;
1045 int port;
1046 struct aafilter *next;
1047 };
1048
inet2_addr_match(const inet_prefix * a,const inet_prefix * p,int plen)1049 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1050 int plen)
1051 {
1052 if (!inet_addr_match(a, p, plen))
1053 return 0;
1054
1055 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1056 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1057 * sockets. Fair? */
1058 if (p->family == AF_INET && a->family == AF_INET6) {
1059 if (a->data[0] == 0 && a->data[1] == 0 &&
1060 a->data[2] == htonl(0xffff)) {
1061 inet_prefix tmp = *a;
1062 tmp.data[0] = a->data[3];
1063 return inet_addr_match(&tmp, p, plen);
1064 }
1065 }
1066 return 1;
1067 }
1068
unix_match(const inet_prefix * a,const inet_prefix * p)1069 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1070 {
1071 char *addr, *pattern;
1072 memcpy(&addr, a->data, sizeof(addr));
1073 memcpy(&pattern, p->data, sizeof(pattern));
1074 if (pattern == NULL)
1075 return 1;
1076 if (addr == NULL)
1077 addr = "";
1078 return !fnmatch(pattern, addr, 0);
1079 }
1080
run_ssfilter(struct ssfilter * f,struct sockstat * s)1081 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1082 {
1083 switch (f->type) {
1084 case SSF_S_AUTO:
1085 {
1086 if (s->local.family == AF_UNIX) {
1087 char *p;
1088 memcpy(&p, s->local.data, sizeof(p));
1089 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1090 strspn(p+1, "0123456789abcdef") == 5);
1091 }
1092 if (s->local.family == AF_PACKET)
1093 return s->lport == 0 && s->local.data[0] == 0;
1094 if (s->local.family == AF_NETLINK)
1095 return s->lport < 0;
1096
1097 return is_ephemeral(s->lport);
1098 }
1099 case SSF_DCOND:
1100 {
1101 struct aafilter *a = (void*)f->pred;
1102 if (a->addr.family == AF_UNIX)
1103 return unix_match(&s->remote, &a->addr);
1104 if (a->port != -1 && a->port != s->rport)
1105 return 0;
1106 if (a->addr.bitlen) {
1107 do {
1108 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1109 return 1;
1110 } while ((a = a->next) != NULL);
1111 return 0;
1112 }
1113 return 1;
1114 }
1115 case SSF_SCOND:
1116 {
1117 struct aafilter *a = (void*)f->pred;
1118 if (a->addr.family == AF_UNIX)
1119 return unix_match(&s->local, &a->addr);
1120 if (a->port != -1 && a->port != s->lport)
1121 return 0;
1122 if (a->addr.bitlen) {
1123 do {
1124 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1125 return 1;
1126 } while ((a = a->next) != NULL);
1127 return 0;
1128 }
1129 return 1;
1130 }
1131 case SSF_D_GE:
1132 {
1133 struct aafilter *a = (void*)f->pred;
1134 return s->rport >= a->port;
1135 }
1136 case SSF_D_LE:
1137 {
1138 struct aafilter *a = (void*)f->pred;
1139 return s->rport <= a->port;
1140 }
1141 case SSF_S_GE:
1142 {
1143 struct aafilter *a = (void*)f->pred;
1144 return s->lport >= a->port;
1145 }
1146 case SSF_S_LE:
1147 {
1148 struct aafilter *a = (void*)f->pred;
1149 return s->lport <= a->port;
1150 }
1151
1152 /* Yup. It is recursion. Sorry. */
1153 case SSF_AND:
1154 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1155 case SSF_OR:
1156 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1157 case SSF_NOT:
1158 return !run_ssfilter(f->pred, s);
1159 default:
1160 abort();
1161 }
1162 }
1163
1164 /* Relocate external jumps by reloc. */
ssfilter_patch(char * a,int len,int reloc)1165 static void ssfilter_patch(char *a, int len, int reloc)
1166 {
1167 while (len > 0) {
1168 struct inet_diag_bc_op *op = (struct inet_diag_bc_op*)a;
1169 if (op->no == len+4)
1170 op->no += reloc;
1171 len -= op->yes;
1172 a += op->yes;
1173 }
1174 if (len < 0)
1175 abort();
1176 }
1177
ssfilter_bytecompile(struct ssfilter * f,char ** bytecode)1178 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1179 {
1180 switch (f->type) {
1181 case SSF_S_AUTO:
1182 {
1183 if (!(*bytecode=malloc(4))) abort();
1184 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1185 return 4;
1186 }
1187 case SSF_DCOND:
1188 case SSF_SCOND:
1189 {
1190 struct aafilter *a = (void*)f->pred;
1191 struct aafilter *b;
1192 char *ptr;
1193 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1194 int len = 0;
1195
1196 for (b=a; b; b=b->next) {
1197 len += 4 + sizeof(struct inet_diag_hostcond);
1198 if (a->addr.family == AF_INET6)
1199 len += 16;
1200 else
1201 len += 4;
1202 if (b->next)
1203 len += 4;
1204 }
1205 if (!(ptr = malloc(len))) abort();
1206 *bytecode = ptr;
1207 for (b=a; b; b=b->next) {
1208 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1209 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1210 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1211 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond*)(ptr+4);
1212
1213 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1214 cond->family = a->addr.family;
1215 cond->port = a->port;
1216 cond->prefix_len = a->addr.bitlen;
1217 memcpy(cond->addr, a->addr.data, alen);
1218 ptr += oplen;
1219 if (b->next) {
1220 op = (struct inet_diag_bc_op *)ptr;
1221 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1222 ptr += 4;
1223 }
1224 }
1225 return ptr - *bytecode;
1226 }
1227 case SSF_D_GE:
1228 {
1229 struct aafilter *x = (void*)f->pred;
1230 if (!(*bytecode=malloc(8))) abort();
1231 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_GE, 8, 12 };
1232 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1233 return 8;
1234 }
1235 case SSF_D_LE:
1236 {
1237 struct aafilter *x = (void*)f->pred;
1238 if (!(*bytecode=malloc(8))) abort();
1239 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_LE, 8, 12 };
1240 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1241 return 8;
1242 }
1243 case SSF_S_GE:
1244 {
1245 struct aafilter *x = (void*)f->pred;
1246 if (!(*bytecode=malloc(8))) abort();
1247 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1248 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1249 return 8;
1250 }
1251 case SSF_S_LE:
1252 {
1253 struct aafilter *x = (void*)f->pred;
1254 if (!(*bytecode=malloc(8))) abort();
1255 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1256 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1257 return 8;
1258 }
1259
1260 case SSF_AND:
1261 {
1262 char *a1, *a2, *a;
1263 int l1, l2;
1264 l1 = ssfilter_bytecompile(f->pred, &a1);
1265 l2 = ssfilter_bytecompile(f->post, &a2);
1266 if (!(a = malloc(l1+l2))) abort();
1267 memcpy(a, a1, l1);
1268 memcpy(a+l1, a2, l2);
1269 free(a1); free(a2);
1270 ssfilter_patch(a, l1, l2);
1271 *bytecode = a;
1272 return l1+l2;
1273 }
1274 case SSF_OR:
1275 {
1276 char *a1, *a2, *a;
1277 int l1, l2;
1278 l1 = ssfilter_bytecompile(f->pred, &a1);
1279 l2 = ssfilter_bytecompile(f->post, &a2);
1280 if (!(a = malloc(l1+l2+4))) abort();
1281 memcpy(a, a1, l1);
1282 memcpy(a+l1+4, a2, l2);
1283 free(a1); free(a2);
1284 *(struct inet_diag_bc_op*)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1285 *bytecode = a;
1286 return l1+l2+4;
1287 }
1288 case SSF_NOT:
1289 {
1290 char *a1, *a;
1291 int l1;
1292 l1 = ssfilter_bytecompile(f->pred, &a1);
1293 if (!(a = malloc(l1+4))) abort();
1294 memcpy(a, a1, l1);
1295 free(a1);
1296 *(struct inet_diag_bc_op*)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1297 *bytecode = a;
1298 return l1+4;
1299 }
1300 default:
1301 abort();
1302 }
1303 }
1304
remember_he(struct aafilter * a,struct hostent * he)1305 static int remember_he(struct aafilter *a, struct hostent *he)
1306 {
1307 char **ptr = he->h_addr_list;
1308 int cnt = 0;
1309 int len;
1310
1311 if (he->h_addrtype == AF_INET)
1312 len = 4;
1313 else if (he->h_addrtype == AF_INET6)
1314 len = 16;
1315 else
1316 return 0;
1317
1318 while (*ptr) {
1319 struct aafilter *b = a;
1320 if (a->addr.bitlen) {
1321 if ((b = malloc(sizeof(*b))) == NULL)
1322 return cnt;
1323 *b = *a;
1324 b->next = a->next;
1325 a->next = b;
1326 }
1327 memcpy(b->addr.data, *ptr, len);
1328 b->addr.bytelen = len;
1329 b->addr.bitlen = len*8;
1330 b->addr.family = he->h_addrtype;
1331 ptr++;
1332 cnt++;
1333 }
1334 return cnt;
1335 }
1336
get_dns_host(struct aafilter * a,const char * addr,int fam)1337 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
1338 {
1339 static int notfirst;
1340 int cnt = 0;
1341 struct hostent *he;
1342
1343 a->addr.bitlen = 0;
1344 if (!notfirst) {
1345 sethostent(1);
1346 notfirst = 1;
1347 }
1348 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
1349 if (he)
1350 cnt = remember_he(a, he);
1351 if (fam == AF_UNSPEC) {
1352 he = gethostbyname2(addr, AF_INET6);
1353 if (he)
1354 cnt += remember_he(a, he);
1355 }
1356 return !cnt;
1357 }
1358
1359 static int xll_initted = 0;
1360
xll_init(void)1361 static void xll_init(void)
1362 {
1363 struct rtnl_handle rth;
1364 if (rtnl_open(&rth, 0) < 0)
1365 exit(1);
1366
1367 ll_init_map(&rth);
1368 rtnl_close(&rth);
1369 xll_initted = 1;
1370 }
1371
xll_index_to_name(int index)1372 static const char *xll_index_to_name(int index)
1373 {
1374 if (!xll_initted)
1375 xll_init();
1376 return ll_index_to_name(index);
1377 }
1378
xll_name_to_index(const char * dev)1379 static int xll_name_to_index(const char *dev)
1380 {
1381 if (!xll_initted)
1382 xll_init();
1383 return ll_name_to_index(dev);
1384 }
1385
parse_hostcond(char * addr,bool is_port)1386 void *parse_hostcond(char *addr, bool is_port)
1387 {
1388 char *port = NULL;
1389 struct aafilter a = { .port = -1 };
1390 struct aafilter *res;
1391 int fam = preferred_family;
1392 struct filter *f = ¤t_filter;
1393
1394 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
1395 char *p;
1396 a.addr.family = AF_UNIX;
1397 if (strncmp(addr, "unix:", 5) == 0)
1398 addr+=5;
1399 p = strdup(addr);
1400 a.addr.bitlen = 8*strlen(p);
1401 memcpy(a.addr.data, &p, sizeof(p));
1402 fam = AF_UNIX;
1403 goto out;
1404 }
1405
1406 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
1407 a.addr.family = AF_PACKET;
1408 a.addr.bitlen = 0;
1409 if (strncmp(addr, "link:", 5) == 0)
1410 addr+=5;
1411 port = strchr(addr, ':');
1412 if (port) {
1413 *port = 0;
1414 if (port[1] && strcmp(port+1, "*")) {
1415 if (get_integer(&a.port, port+1, 0)) {
1416 if ((a.port = xll_name_to_index(port+1)) <= 0)
1417 return NULL;
1418 }
1419 }
1420 }
1421 if (addr[0] && strcmp(addr, "*")) {
1422 unsigned short tmp;
1423 a.addr.bitlen = 32;
1424 if (ll_proto_a2n(&tmp, addr))
1425 return NULL;
1426 a.addr.data[0] = ntohs(tmp);
1427 }
1428 fam = AF_PACKET;
1429 goto out;
1430 }
1431
1432 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
1433 a.addr.family = AF_NETLINK;
1434 a.addr.bitlen = 0;
1435 if (strncmp(addr, "netlink:", 8) == 0)
1436 addr+=8;
1437 port = strchr(addr, ':');
1438 if (port) {
1439 *port = 0;
1440 if (port[1] && strcmp(port+1, "*")) {
1441 if (get_integer(&a.port, port+1, 0)) {
1442 if (strcmp(port+1, "kernel") == 0)
1443 a.port = 0;
1444 else
1445 return NULL;
1446 }
1447 }
1448 }
1449 if (addr[0] && strcmp(addr, "*")) {
1450 a.addr.bitlen = 32;
1451 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
1452 return NULL;
1453 }
1454 fam = AF_NETLINK;
1455 goto out;
1456 }
1457
1458 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
1459 fam = AF_INET;
1460 if (!strncmp(addr, "inet:", 5))
1461 addr += 5;
1462 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
1463 fam = AF_INET6;
1464 if (!strncmp(addr, "inet6:", 6))
1465 addr += 6;
1466 }
1467
1468 /* URL-like literal [] */
1469 if (addr[0] == '[') {
1470 addr++;
1471 if ((port = strchr(addr, ']')) == NULL)
1472 return NULL;
1473 *port++ = 0;
1474 } else if (addr[0] == '*') {
1475 port = addr+1;
1476 } else {
1477 port = strrchr(strchr(addr, '/') ? : addr, ':');
1478 }
1479
1480 if (is_port)
1481 port = addr;
1482
1483 if (port && *port) {
1484 if (*port == ':')
1485 *port++ = 0;
1486
1487 if (*port && *port != '*') {
1488 if (get_integer(&a.port, port, 0)) {
1489 struct servent *se1 = NULL;
1490 struct servent *se2 = NULL;
1491 if (current_filter.dbs&(1<<UDP_DB))
1492 se1 = getservbyname(port, UDP_PROTO);
1493 if (current_filter.dbs&(1<<TCP_DB))
1494 se2 = getservbyname(port, TCP_PROTO);
1495 if (se1 && se2 && se1->s_port != se2->s_port) {
1496 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
1497 return NULL;
1498 }
1499 if (!se1)
1500 se1 = se2;
1501 if (se1) {
1502 a.port = ntohs(se1->s_port);
1503 } else {
1504 struct scache *s;
1505 for (s = rlist; s; s = s->next) {
1506 if ((s->proto == UDP_PROTO &&
1507 (current_filter.dbs&(1<<UDP_DB))) ||
1508 (s->proto == TCP_PROTO &&
1509 (current_filter.dbs&(1<<TCP_DB)))) {
1510 if (s->name && strcmp(s->name, port) == 0) {
1511 if (a.port > 0 && a.port != s->port) {
1512 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
1513 return NULL;
1514 }
1515 a.port = s->port;
1516 }
1517 }
1518 }
1519 if (a.port <= 0) {
1520 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
1521 return NULL;
1522 }
1523 }
1524 }
1525 }
1526 }
1527 if (!is_port && addr && *addr && *addr != '*') {
1528 if (get_prefix_1(&a.addr, addr, fam)) {
1529 if (get_dns_host(&a, addr, fam)) {
1530 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
1531 return NULL;
1532 }
1533 }
1534 }
1535
1536 out:
1537 if (fam != AF_UNSPEC) {
1538 f->families = 0;
1539 filter_af_set(f, fam);
1540 filter_states_set(f, 0);
1541 }
1542
1543 res = malloc(sizeof(*res));
1544 if (res)
1545 memcpy(res, &a, sizeof(a));
1546 return res;
1547 }
1548
proto_name(int protocol)1549 static char *proto_name(int protocol)
1550 {
1551 switch (protocol) {
1552 case 0:
1553 return "raw";
1554 case IPPROTO_UDP:
1555 return "udp";
1556 case IPPROTO_TCP:
1557 return "tcp";
1558 case IPPROTO_DCCP:
1559 return "dccp";
1560 }
1561
1562 return "???";
1563 }
1564
inet_stats_print(struct sockstat * s,int protocol)1565 static void inet_stats_print(struct sockstat *s, int protocol)
1566 {
1567 char *buf = NULL;
1568
1569 sock_state_print(s, proto_name(protocol));
1570
1571 inet_addr_print(&s->local, s->lport, s->iface);
1572 inet_addr_print(&s->remote, s->rport, 0);
1573
1574 if (show_proc_ctx || show_sock_ctx) {
1575 if (find_entry(s->ino, &buf,
1576 (show_proc_ctx & show_sock_ctx) ?
1577 PROC_SOCK_CTX : PROC_CTX) > 0) {
1578 printf(" users:(%s)", buf);
1579 free(buf);
1580 }
1581 } else if (show_users) {
1582 if (find_entry(s->ino, &buf, USERS) > 0) {
1583 printf(" users:(%s)", buf);
1584 free(buf);
1585 }
1586 }
1587 }
1588
proc_parse_inet_addr(char * loc,char * rem,int family,struct sockstat * s)1589 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
1590 sockstat *s)
1591 {
1592 s->local.family = s->remote.family = family;
1593 if (family == AF_INET) {
1594 sscanf(loc, "%x:%x", s->local.data, (unsigned*)&s->lport);
1595 sscanf(rem, "%x:%x", s->remote.data, (unsigned*)&s->rport);
1596 s->local.bytelen = s->remote.bytelen = 4;
1597 return 0;
1598 } else {
1599 sscanf(loc, "%08x%08x%08x%08x:%x",
1600 s->local.data,
1601 s->local.data + 1,
1602 s->local.data + 2,
1603 s->local.data + 3,
1604 &s->lport);
1605 sscanf(rem, "%08x%08x%08x%08x:%x",
1606 s->remote.data,
1607 s->remote.data + 1,
1608 s->remote.data + 2,
1609 s->remote.data + 3,
1610 &s->rport);
1611 s->local.bytelen = s->remote.bytelen = 16;
1612 return 0;
1613 }
1614 return -1;
1615 }
1616
proc_inet_split_line(char * line,char ** loc,char ** rem,char ** data)1617 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
1618 {
1619 char *p;
1620
1621 if ((p = strchr(line, ':')) == NULL)
1622 return -1;
1623
1624 *loc = p+2;
1625 if ((p = strchr(*loc, ':')) == NULL)
1626 return -1;
1627
1628 p[5] = 0;
1629 *rem = p+6;
1630 if ((p = strchr(*rem, ':')) == NULL)
1631 return -1;
1632
1633 p[5] = 0;
1634 *data = p+6;
1635 return 0;
1636 }
1637
sprint_bw(char * buf,double bw)1638 static char *sprint_bw(char *buf, double bw)
1639 {
1640 if (bw > 1000000.)
1641 sprintf(buf,"%.1fM", bw / 1000000.);
1642 else if (bw > 1000.)
1643 sprintf(buf,"%.1fK", bw / 1000.);
1644 else
1645 sprintf(buf, "%g", bw);
1646
1647 return buf;
1648 }
1649
tcp_stats_print(struct tcpstat * s)1650 static void tcp_stats_print(struct tcpstat *s)
1651 {
1652 char b1[64];
1653
1654 if (s->has_ts_opt)
1655 printf(" ts");
1656 if (s->has_sack_opt)
1657 printf(" sack");
1658 if (s->has_ecn_opt)
1659 printf(" ecn");
1660 if (s->has_ecnseen_opt)
1661 printf(" ecnseen");
1662 if (s->has_fastopen_opt)
1663 printf(" fastopen");
1664 if (s->cong_alg[0])
1665 printf(" %s", s->cong_alg);
1666 if (s->has_wscale_opt)
1667 printf(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
1668 if (s->rto)
1669 printf(" rto:%g", s->rto);
1670 if (s->backoff)
1671 printf(" backoff:%u", s->backoff);
1672 if (s->rtt)
1673 printf(" rtt:%g/%g", s->rtt, s->rttvar);
1674 if (s->ato)
1675 printf(" ato:%g", s->ato);
1676
1677 if (s->qack)
1678 printf(" qack:%d", s->qack);
1679 if (s->qack & 1)
1680 printf(" bidir");
1681
1682 if (s->mss)
1683 printf(" mss:%d", s->mss);
1684 if (s->cwnd)
1685 printf(" cwnd:%d", s->cwnd);
1686 if (s->ssthresh)
1687 printf(" ssthresh:%d", s->ssthresh);
1688
1689 if (s->bytes_acked)
1690 printf(" bytes_acked:%llu", s->bytes_acked);
1691 if (s->bytes_received)
1692 printf(" bytes_received:%llu", s->bytes_received);
1693 if (s->segs_out)
1694 printf(" segs_out:%u", s->segs_out);
1695 if (s->segs_in)
1696 printf(" segs_in:%u", s->segs_in);
1697
1698 if (s->dctcp && s->dctcp->enabled) {
1699 struct dctcpstat *dctcp = s->dctcp;
1700
1701 printf(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
1702 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
1703 dctcp->ab_tot);
1704 } else if (s->dctcp) {
1705 printf(" dctcp:fallback_mode");
1706 }
1707
1708 if (s->send_bps)
1709 printf(" send %sbps", sprint_bw(b1, s->send_bps));
1710 if (s->lastsnd)
1711 printf(" lastsnd:%u", s->lastsnd);
1712 if (s->lastrcv)
1713 printf(" lastrcv:%u", s->lastrcv);
1714 if (s->lastack)
1715 printf(" lastack:%u", s->lastack);
1716
1717 if (s->pacing_rate) {
1718 printf(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
1719 if (s->pacing_rate_max)
1720 printf("/%sbps", sprint_bw(b1,
1721 s->pacing_rate_max));
1722 }
1723
1724 if (s->unacked)
1725 printf(" unacked:%u", s->unacked);
1726 if (s->retrans || s->retrans_total)
1727 printf(" retrans:%u/%u", s->retrans, s->retrans_total);
1728 if (s->lost)
1729 printf(" lost:%u", s->lost);
1730 if (s->sacked && s->ss.state != SS_LISTEN)
1731 printf(" sacked:%u", s->sacked);
1732 if (s->fackets)
1733 printf(" fackets:%u", s->fackets);
1734 if (s->reordering != 3)
1735 printf(" reordering:%d", s->reordering);
1736 if (s->rcv_rtt)
1737 printf(" rcv_rtt:%g", s->rcv_rtt);
1738 if (s->rcv_space)
1739 printf(" rcv_space:%d", s->rcv_space);
1740 }
1741
tcp_timer_print(struct tcpstat * s)1742 static void tcp_timer_print(struct tcpstat *s)
1743 {
1744 if (s->timer) {
1745 if (s->timer > 4)
1746 s->timer = 5;
1747 printf(" timer:(%s,%s,%d)",
1748 tmr_name[s->timer],
1749 print_ms_timer(s->timeout),
1750 s->retrans);
1751 }
1752 }
1753
tcp_show_line(char * line,const struct filter * f,int family)1754 static int tcp_show_line(char *line, const struct filter *f, int family)
1755 {
1756 int rto = 0, ato = 0;
1757 struct tcpstat s = {};
1758 char *loc, *rem, *data;
1759 char opt[256];
1760 int n;
1761 int hz = get_user_hz();
1762
1763 if (proc_inet_split_line(line, &loc, &rem, &data))
1764 return -1;
1765
1766 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
1767 if (!(f->states & (1 << state)))
1768 return 0;
1769
1770 proc_parse_inet_addr(loc, rem, family, &s.ss);
1771
1772 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
1773 return 0;
1774
1775 opt[0] = 0;
1776 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %d %d %[^\n]\n",
1777 &s.ss.state, &s.ss.wq, &s.ss.rq,
1778 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
1779 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
1780 &s.ssthresh, opt);
1781
1782 if (n < 17)
1783 opt[0] = 0;
1784
1785 if (n < 12) {
1786 rto = 0;
1787 s.cwnd = 2;
1788 s.ssthresh = -1;
1789 ato = s.qack = 0;
1790 }
1791
1792 s.retrans = s.timer != 1 ? s.probes : s.retrans;
1793 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
1794 s.ato = (double)ato / hz;
1795 s.qack /= 2;
1796 s.rto = (double)rto;
1797 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
1798 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
1799
1800 inet_stats_print(&s.ss, IPPROTO_TCP);
1801
1802 if (show_options)
1803 tcp_timer_print(&s);
1804
1805 if (show_details) {
1806 sock_details_print(&s.ss);
1807 if (opt[0])
1808 printf(" opt:\"%s\"", opt);
1809 }
1810
1811 if (show_tcpinfo)
1812 tcp_stats_print(&s);
1813
1814 printf("\n");
1815 return 0;
1816 }
1817
generic_record_read(FILE * fp,int (* worker)(char *,const struct filter *,int),const struct filter * f,int fam)1818 static int generic_record_read(FILE *fp,
1819 int (*worker)(char*, const struct filter *, int),
1820 const struct filter *f, int fam)
1821 {
1822 char line[256];
1823
1824 /* skip header */
1825 if (fgets(line, sizeof(line), fp) == NULL)
1826 goto outerr;
1827
1828 while (fgets(line, sizeof(line), fp) != NULL) {
1829 int n = strlen(line);
1830 if (n == 0 || line[n-1] != '\n') {
1831 errno = -EINVAL;
1832 return -1;
1833 }
1834 line[n-1] = 0;
1835
1836 if (worker(line, f, fam) < 0)
1837 return 0;
1838 }
1839 outerr:
1840
1841 return ferror(fp) ? -1 : 0;
1842 }
1843
print_skmeminfo(struct rtattr * tb[],int attrtype)1844 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
1845 {
1846 const __u32 *skmeminfo;
1847
1848 if (!tb[attrtype]) {
1849 if (attrtype == INET_DIAG_SKMEMINFO) {
1850 if (!tb[INET_DIAG_MEMINFO])
1851 return;
1852
1853 const struct inet_diag_meminfo *minfo =
1854 RTA_DATA(tb[INET_DIAG_MEMINFO]);
1855
1856 printf(" mem:(r%u,w%u,f%u,t%u)",
1857 minfo->idiag_rmem,
1858 minfo->idiag_wmem,
1859 minfo->idiag_fmem,
1860 minfo->idiag_tmem);
1861 }
1862 return;
1863 }
1864
1865 skmeminfo = RTA_DATA(tb[attrtype]);
1866
1867 printf(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
1868 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
1869 skmeminfo[SK_MEMINFO_RCVBUF],
1870 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
1871 skmeminfo[SK_MEMINFO_SNDBUF],
1872 skmeminfo[SK_MEMINFO_FWD_ALLOC],
1873 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
1874 skmeminfo[SK_MEMINFO_OPTMEM]);
1875
1876 if (RTA_PAYLOAD(tb[attrtype]) >=
1877 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
1878 printf(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
1879
1880 printf(")");
1881 }
1882
1883 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
1884
tcp_show_info(const struct nlmsghdr * nlh,struct inet_diag_msg * r,struct rtattr * tb[])1885 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
1886 struct rtattr *tb[])
1887 {
1888 double rtt = 0;
1889 struct tcpstat s = {};
1890
1891 s.ss.state = r->idiag_state;
1892
1893 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
1894
1895 if (tb[INET_DIAG_INFO]) {
1896 struct tcp_info *info;
1897 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
1898
1899 /* workaround for older kernels with less fields */
1900 if (len < sizeof(*info)) {
1901 info = alloca(sizeof(*info));
1902 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
1903 memset((char *)info + len, 0, sizeof(*info) - len);
1904 } else
1905 info = RTA_DATA(tb[INET_DIAG_INFO]);
1906
1907 if (show_options) {
1908 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
1909 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
1910 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
1911 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
1912 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
1913 }
1914
1915 if (tb[INET_DIAG_CONG])
1916 strncpy(s.cong_alg,
1917 rta_getattr_str(tb[INET_DIAG_CONG]),
1918 sizeof(s.cong_alg) - 1);
1919
1920 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
1921 s.has_wscale_opt = true;
1922 s.snd_wscale = info->tcpi_snd_wscale;
1923 s.rcv_wscale = info->tcpi_rcv_wscale;
1924 }
1925
1926 if (info->tcpi_rto && info->tcpi_rto != 3000000)
1927 s.rto = (double)info->tcpi_rto / 1000;
1928
1929 s.backoff = info->tcpi_backoff;
1930 s.rtt = (double)info->tcpi_rtt / 1000;
1931 s.rttvar = (double)info->tcpi_rttvar / 1000;
1932 s.ato = (double)info->tcpi_ato / 1000;
1933 s.mss = info->tcpi_snd_mss;
1934 s.rcv_space = info->tcpi_rcv_space;
1935 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
1936 s.lastsnd = info->tcpi_last_data_sent;
1937 s.lastrcv = info->tcpi_last_data_recv;
1938 s.lastack = info->tcpi_last_ack_recv;
1939 s.unacked = info->tcpi_unacked;
1940 s.retrans = info->tcpi_retrans;
1941 s.retrans_total = info->tcpi_total_retrans;
1942 s.lost = info->tcpi_lost;
1943 s.sacked = info->tcpi_sacked;
1944 s.reordering = info->tcpi_reordering;
1945 s.rcv_space = info->tcpi_rcv_space;
1946 s.cwnd = info->tcpi_snd_cwnd;
1947
1948 if (info->tcpi_snd_ssthresh < 0xFFFF)
1949 s.ssthresh = info->tcpi_snd_ssthresh;
1950
1951 rtt = (double) info->tcpi_rtt;
1952 if (tb[INET_DIAG_VEGASINFO]) {
1953 const struct tcpvegas_info *vinfo
1954 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
1955
1956 if (vinfo->tcpv_enabled &&
1957 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
1958 rtt = vinfo->tcpv_rtt;
1959 }
1960
1961 if (tb[INET_DIAG_DCTCPINFO]) {
1962 struct dctcpstat *dctcp = malloc(sizeof(struct
1963 dctcpstat));
1964
1965 const struct tcp_dctcp_info *dinfo
1966 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
1967
1968 dctcp->enabled = !!dinfo->dctcp_enabled;
1969 dctcp->ce_state = dinfo->dctcp_ce_state;
1970 dctcp->alpha = dinfo->dctcp_alpha;
1971 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
1972 dctcp->ab_tot = dinfo->dctcp_ab_tot;
1973 s.dctcp = dctcp;
1974 }
1975
1976 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
1977 s.send_bps = (double) info->tcpi_snd_cwnd *
1978 (double)info->tcpi_snd_mss * 8000000. / rtt;
1979 }
1980
1981 if (info->tcpi_pacing_rate &&
1982 info->tcpi_pacing_rate != ~0ULL) {
1983 s.pacing_rate = info->tcpi_pacing_rate * 8.;
1984
1985 if (info->tcpi_max_pacing_rate &&
1986 info->tcpi_max_pacing_rate != ~0ULL)
1987 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
1988 }
1989 s.bytes_acked = info->tcpi_bytes_acked;
1990 s.bytes_received = info->tcpi_bytes_received;
1991 s.segs_out = info->tcpi_segs_out;
1992 s.segs_in = info->tcpi_segs_in;
1993 tcp_stats_print(&s);
1994 free(s.dctcp);
1995 }
1996 }
1997
inet_show_sock(struct nlmsghdr * nlh,struct filter * f,int protocol)1998 static int inet_show_sock(struct nlmsghdr *nlh, struct filter *f, int protocol)
1999 {
2000 struct rtattr * tb[INET_DIAG_MAX+1];
2001 struct inet_diag_msg *r = NLMSG_DATA(nlh);
2002 struct sockstat s = {};
2003
2004 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr*)(r+1),
2005 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2006
2007 s.state = r->idiag_state;
2008 s.local.family = s.remote.family = r->idiag_family;
2009 s.lport = ntohs(r->id.idiag_sport);
2010 s.rport = ntohs(r->id.idiag_dport);
2011 s.wq = r->idiag_wqueue;
2012 s.rq = r->idiag_rqueue;
2013 s.ino = r->idiag_inode;
2014 s.uid = r->idiag_uid;
2015 s.iface = r->id.idiag_if;
2016 s.sk = cookie_sk_get(&r->id.idiag_cookie[0]);
2017
2018 if (s.local.family == AF_INET) {
2019 s.local.bytelen = s.remote.bytelen = 4;
2020 } else {
2021 s.local.bytelen = s.remote.bytelen = 16;
2022 }
2023
2024 memcpy(s.local.data, r->id.idiag_src, s.local.bytelen);
2025 memcpy(s.remote.data, r->id.idiag_dst, s.local.bytelen);
2026
2027 if (f && f->f && run_ssfilter(f->f, &s) == 0)
2028 return 0;
2029
2030 if (tb[INET_DIAG_PROTOCOL])
2031 protocol = *(__u8 *)RTA_DATA(tb[INET_DIAG_PROTOCOL]);
2032
2033 inet_stats_print(&s, protocol);
2034
2035 if (show_options) {
2036 struct tcpstat t = {};
2037
2038 t.timer = r->idiag_timer;
2039 t.timeout = r->idiag_expires;
2040 t.retrans = r->idiag_retrans;
2041 tcp_timer_print(&t);
2042 }
2043
2044 if (show_details) {
2045 sock_details_print(&s);
2046 if (s.local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY]) {
2047 unsigned char v6only;
2048 v6only = *(__u8 *)RTA_DATA(tb[INET_DIAG_SKV6ONLY]);
2049 printf(" v6only:%u", v6only);
2050 }
2051 if (tb[INET_DIAG_SHUTDOWN]) {
2052 unsigned char mask;
2053 mask = *(__u8 *)RTA_DATA(tb[INET_DIAG_SHUTDOWN]);
2054 printf(" %c-%c", mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
2055 }
2056 }
2057
2058 if (show_mem || show_tcpinfo) {
2059 printf("\n\t");
2060 tcp_show_info(nlh, r, tb);
2061 }
2062
2063 printf("\n");
2064 return 0;
2065 }
2066
tcpdiag_send(int fd,int protocol,struct filter * f)2067 static int tcpdiag_send(int fd, int protocol, struct filter *f)
2068 {
2069 struct sockaddr_nl nladdr;
2070 struct {
2071 struct nlmsghdr nlh;
2072 struct inet_diag_req r;
2073 } req;
2074 char *bc = NULL;
2075 int bclen;
2076 struct msghdr msg;
2077 struct rtattr rta;
2078 struct iovec iov[3];
2079
2080 if (protocol == IPPROTO_UDP)
2081 return -1;
2082
2083 memset(&nladdr, 0, sizeof(nladdr));
2084 nladdr.nl_family = AF_NETLINK;
2085
2086 req.nlh.nlmsg_len = sizeof(req);
2087 if (protocol == IPPROTO_TCP)
2088 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
2089 else
2090 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
2091 req.nlh.nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST;
2092 req.nlh.nlmsg_pid = 0;
2093 req.nlh.nlmsg_seq = MAGIC_SEQ;
2094 memset(&req.r, 0, sizeof(req.r));
2095 req.r.idiag_family = AF_INET;
2096 req.r.idiag_states = f->states;
2097 if (show_mem) {
2098 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
2099 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
2100 }
2101
2102 if (show_tcpinfo) {
2103 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
2104 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
2105 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
2106 }
2107
2108 iov[0] = (struct iovec){
2109 .iov_base = &req,
2110 .iov_len = sizeof(req)
2111 };
2112 if (f->f) {
2113 bclen = ssfilter_bytecompile(f->f, &bc);
2114 rta.rta_type = INET_DIAG_REQ_BYTECODE;
2115 rta.rta_len = RTA_LENGTH(bclen);
2116 iov[1] = (struct iovec){ &rta, sizeof(rta) };
2117 iov[2] = (struct iovec){ bc, bclen };
2118 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
2119 }
2120
2121 msg = (struct msghdr) {
2122 .msg_name = (void*)&nladdr,
2123 .msg_namelen = sizeof(nladdr),
2124 .msg_iov = iov,
2125 .msg_iovlen = f->f ? 3 : 1,
2126 };
2127
2128 if (sendmsg(fd, &msg, 0) < 0) {
2129 close(fd);
2130 return -1;
2131 }
2132
2133 return 0;
2134 }
2135
sockdiag_send(int family,int fd,int protocol,struct filter * f)2136 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
2137 {
2138 struct sockaddr_nl nladdr;
2139 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
2140 char *bc = NULL;
2141 int bclen;
2142 struct msghdr msg;
2143 struct rtattr rta;
2144 struct iovec iov[3];
2145
2146 if (family == PF_UNSPEC)
2147 return tcpdiag_send(fd, protocol, f);
2148
2149 memset(&nladdr, 0, sizeof(nladdr));
2150 nladdr.nl_family = AF_NETLINK;
2151
2152 memset(&req.r, 0, sizeof(req.r));
2153 req.r.sdiag_family = family;
2154 req.r.sdiag_protocol = protocol;
2155 req.r.idiag_states = f->states;
2156 if (show_mem) {
2157 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
2158 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
2159 }
2160
2161 if (show_tcpinfo) {
2162 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
2163 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
2164 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
2165 }
2166
2167 iov[0] = (struct iovec){
2168 .iov_base = &req,
2169 .iov_len = sizeof(req)
2170 };
2171 if (f->f) {
2172 bclen = ssfilter_bytecompile(f->f, &bc);
2173 rta.rta_type = INET_DIAG_REQ_BYTECODE;
2174 rta.rta_len = RTA_LENGTH(bclen);
2175 iov[1] = (struct iovec){ &rta, sizeof(rta) };
2176 iov[2] = (struct iovec){ bc, bclen };
2177 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
2178 }
2179
2180 msg = (struct msghdr) {
2181 .msg_name = (void*)&nladdr,
2182 .msg_namelen = sizeof(nladdr),
2183 .msg_iov = iov,
2184 .msg_iovlen = f->f ? 3 : 1,
2185 };
2186
2187 if (sendmsg(fd, &msg, 0) < 0) {
2188 close(fd);
2189 return -1;
2190 }
2191
2192 return 0;
2193 }
2194
2195 struct inet_diag_arg {
2196 struct filter *f;
2197 int protocol;
2198 struct rtnl_handle *rth;
2199 };
2200
kill_inet_sock(const struct sockaddr_nl * addr,struct nlmsghdr * h,void * arg)2201 static int kill_inet_sock(const struct sockaddr_nl *addr,
2202 struct nlmsghdr *h, void *arg)
2203 {
2204 struct inet_diag_msg *d = NLMSG_DATA(h);
2205 struct inet_diag_arg *diag_arg = arg;
2206 struct rtnl_handle *rth = diag_arg->rth;
2207 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
2208
2209 req.nlh.nlmsg_type = SOCK_DESTROY;
2210 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
2211 req.nlh.nlmsg_seq = ++rth->seq;
2212 req.r.sdiag_family = d->idiag_family;
2213 req.r.sdiag_protocol = diag_arg->protocol;
2214 req.r.id = d->id;
2215
2216 return rtnl_talk(rth, &req.nlh, NULL, 0);
2217 }
2218
show_one_inet_sock(const struct sockaddr_nl * addr,struct nlmsghdr * h,void * arg)2219 static int show_one_inet_sock(const struct sockaddr_nl *addr,
2220 struct nlmsghdr *h, void *arg)
2221 {
2222 int err;
2223 struct inet_diag_arg *diag_arg = arg;
2224 struct inet_diag_msg *r = NLMSG_DATA(h);
2225
2226 if (!(diag_arg->f->families & (1 << r->idiag_family)))
2227 return 0;
2228 if (diag_arg->f->kill && kill_inet_sock(addr, h, arg) != 0) {
2229 if (errno == EOPNOTSUPP || errno == ENOENT) {
2230 /* Socket can't be closed, or is already closed. */
2231 return 0;
2232 } else {
2233 perror("SOCK_DESTROY answers");
2234 return -1;
2235 }
2236 }
2237 if ((err = inet_show_sock(h, diag_arg->f, diag_arg->protocol)) < 0)
2238 return err;
2239
2240 return 0;
2241 }
2242
inet_show_netlink(struct filter * f,FILE * dump_fp,int protocol)2243 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
2244 {
2245 int err = 0;
2246 struct rtnl_handle rth, rth2;
2247 int family = PF_INET;
2248 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
2249
2250 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
2251 return -1;
2252
2253 if (f->kill) {
2254 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
2255 rtnl_close(&rth);
2256 return -1;
2257 }
2258 arg.rth = &rth2;
2259 }
2260
2261 rth.dump = MAGIC_SEQ;
2262 rth.dump_fp = dump_fp;
2263 if (preferred_family == PF_INET6)
2264 family = PF_INET6;
2265
2266 again:
2267 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
2268 goto Exit;
2269
2270 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
2271 if (family != PF_UNSPEC) {
2272 family = PF_UNSPEC;
2273 goto again;
2274 }
2275 goto Exit;
2276 }
2277 if (family == PF_INET && preferred_family != PF_INET) {
2278 family = PF_INET6;
2279 goto again;
2280 }
2281
2282 Exit:
2283 rtnl_close(&rth);
2284 if (arg.rth)
2285 rtnl_close(arg.rth);
2286 return err;
2287 }
2288
tcp_show_netlink_file(struct filter * f)2289 static int tcp_show_netlink_file(struct filter *f)
2290 {
2291 FILE *fp;
2292 char buf[16384];
2293
2294 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
2295 perror("fopen($TCPDIAG_FILE)");
2296 return -1;
2297 }
2298
2299 while (1) {
2300 int status, err;
2301 struct nlmsghdr *h = (struct nlmsghdr*)buf;
2302
2303 status = fread(buf, 1, sizeof(*h), fp);
2304 if (status < 0) {
2305 perror("Reading header from $TCPDIAG_FILE");
2306 return -1;
2307 }
2308 if (status != sizeof(*h)) {
2309 perror("Unexpected EOF reading $TCPDIAG_FILE");
2310 return -1;
2311 }
2312
2313 status = fread(h+1, 1, NLMSG_ALIGN(h->nlmsg_len-sizeof(*h)), fp);
2314
2315 if (status < 0) {
2316 perror("Reading $TCPDIAG_FILE");
2317 return -1;
2318 }
2319 if (status + sizeof(*h) < h->nlmsg_len) {
2320 perror("Unexpected EOF reading $TCPDIAG_FILE");
2321 return -1;
2322 }
2323
2324 /* The only legal exit point */
2325 if (h->nlmsg_type == NLMSG_DONE)
2326 return 0;
2327
2328 if (h->nlmsg_type == NLMSG_ERROR) {
2329 struct nlmsgerr *err = (struct nlmsgerr*)NLMSG_DATA(h);
2330 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
2331 fprintf(stderr, "ERROR truncated\n");
2332 } else {
2333 errno = -err->error;
2334 perror("TCPDIAG answered");
2335 }
2336 return -1;
2337 }
2338
2339 err = inet_show_sock(h, f, IPPROTO_TCP);
2340 if (err < 0)
2341 return err;
2342 }
2343 }
2344
tcp_show(struct filter * f,int socktype)2345 static int tcp_show(struct filter *f, int socktype)
2346 {
2347 FILE *fp = NULL;
2348 char *buf = NULL;
2349 int bufsize = 64*1024;
2350
2351 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2352 return 0;
2353
2354 dg_proto = TCP_PROTO;
2355
2356 if (getenv("TCPDIAG_FILE"))
2357 return tcp_show_netlink_file(f);
2358
2359 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
2360 && inet_show_netlink(f, NULL, socktype) == 0)
2361 return 0;
2362
2363 /* Sigh... We have to parse /proc/net/tcp... */
2364
2365
2366 /* Estimate amount of sockets and try to allocate
2367 * huge buffer to read all the table at one read.
2368 * Limit it by 16MB though. The assumption is: as soon as
2369 * kernel was able to hold information about N connections,
2370 * it is able to give us some memory for snapshot.
2371 */
2372 if (1) {
2373 get_slabstat(&slabstat);
2374
2375 int guess = slabstat.socks+slabstat.tcp_syns;
2376 if (f->states&(1<<SS_TIME_WAIT))
2377 guess += slabstat.tcp_tws;
2378 if (guess > (16*1024*1024)/128)
2379 guess = (16*1024*1024)/128;
2380 guess *= 128;
2381 if (guess > bufsize)
2382 bufsize = guess;
2383 }
2384 while (bufsize >= 64*1024) {
2385 if ((buf = malloc(bufsize)) != NULL)
2386 break;
2387 bufsize /= 2;
2388 }
2389 if (buf == NULL) {
2390 errno = ENOMEM;
2391 return -1;
2392 }
2393
2394 if (f->families & (1<<AF_INET)) {
2395 if ((fp = net_tcp_open()) == NULL)
2396 goto outerr;
2397
2398 setbuffer(fp, buf, bufsize);
2399 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
2400 goto outerr;
2401 fclose(fp);
2402 }
2403
2404 if ((f->families & (1<<AF_INET6)) &&
2405 (fp = net_tcp6_open()) != NULL) {
2406 setbuffer(fp, buf, bufsize);
2407 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
2408 goto outerr;
2409 fclose(fp);
2410 }
2411
2412 free(buf);
2413 return 0;
2414
2415 outerr:
2416 do {
2417 int saved_errno = errno;
2418 free(buf);
2419 if (fp)
2420 fclose(fp);
2421 errno = saved_errno;
2422 return -1;
2423 } while (0);
2424 }
2425
2426
dgram_show_line(char * line,const struct filter * f,int family)2427 static int dgram_show_line(char *line, const struct filter *f, int family)
2428 {
2429 struct sockstat s = {};
2430 char *loc, *rem, *data;
2431 char opt[256];
2432 int n;
2433
2434 if (proc_inet_split_line(line, &loc, &rem, &data))
2435 return -1;
2436
2437 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2438 if (!(f->states & (1 << state)))
2439 return 0;
2440
2441 proc_parse_inet_addr(loc, rem, family, &s);
2442
2443 if (f->f && run_ssfilter(f->f, &s) == 0)
2444 return 0;
2445
2446 opt[0] = 0;
2447 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
2448 &s.state, &s.wq, &s.rq,
2449 &s.uid, &s.ino,
2450 &s.refcnt, &s.sk, opt);
2451
2452 if (n < 9)
2453 opt[0] = 0;
2454
2455 inet_stats_print(&s, dg_proto == UDP_PROTO ? IPPROTO_UDP : 0);
2456
2457 if (show_details && opt[0])
2458 printf(" opt:\"%s\"", opt);
2459
2460 printf("\n");
2461 return 0;
2462 }
2463
udp_show(struct filter * f)2464 static int udp_show(struct filter *f)
2465 {
2466 FILE *fp = NULL;
2467
2468 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2469 return 0;
2470
2471 dg_proto = UDP_PROTO;
2472
2473 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
2474 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
2475 return 0;
2476
2477 if (f->families&(1<<AF_INET)) {
2478 if ((fp = net_udp_open()) == NULL)
2479 goto outerr;
2480 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
2481 goto outerr;
2482 fclose(fp);
2483 }
2484
2485 if ((f->families&(1<<AF_INET6)) &&
2486 (fp = net_udp6_open()) != NULL) {
2487 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
2488 goto outerr;
2489 fclose(fp);
2490 }
2491 return 0;
2492
2493 outerr:
2494 do {
2495 int saved_errno = errno;
2496 if (fp)
2497 fclose(fp);
2498 errno = saved_errno;
2499 return -1;
2500 } while (0);
2501 }
2502
raw_show(struct filter * f)2503 static int raw_show(struct filter *f)
2504 {
2505 FILE *fp = NULL;
2506
2507 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2508 return 0;
2509
2510 dg_proto = RAW_PROTO;
2511
2512 if (f->families&(1<<AF_INET)) {
2513 if ((fp = net_raw_open()) == NULL)
2514 goto outerr;
2515 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
2516 goto outerr;
2517 fclose(fp);
2518 }
2519
2520 if ((f->families&(1<<AF_INET6)) &&
2521 (fp = net_raw6_open()) != NULL) {
2522 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
2523 goto outerr;
2524 fclose(fp);
2525 }
2526 return 0;
2527
2528 outerr:
2529 do {
2530 int saved_errno = errno;
2531 if (fp)
2532 fclose(fp);
2533 errno = saved_errno;
2534 return -1;
2535 } while (0);
2536 }
2537
2538 int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
2539 SS_ESTABLISHED, SS_CLOSING };
2540
2541 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
2542
unix_list_free(struct sockstat * list)2543 static void unix_list_free(struct sockstat *list)
2544 {
2545 while (list) {
2546 struct sockstat *s = list;
2547
2548 list = list->next;
2549 free(s->name);
2550 free(s);
2551 }
2552 }
2553
unix_netid_name(int type)2554 static const char *unix_netid_name(int type)
2555 {
2556 const char *netid;
2557
2558 switch (type) {
2559 case SOCK_STREAM:
2560 netid = "u_str";
2561 break;
2562 case SOCK_SEQPACKET:
2563 netid = "u_seq";
2564 break;
2565 case SOCK_DGRAM:
2566 default:
2567 netid = "u_dgr";
2568 break;
2569 }
2570 return netid;
2571 }
2572
unix_type_skip(struct sockstat * s,struct filter * f)2573 static bool unix_type_skip(struct sockstat *s, struct filter *f)
2574 {
2575 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
2576 return true;
2577 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
2578 return true;
2579 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
2580 return true;
2581 return false;
2582 }
2583
unix_use_proc(void)2584 static bool unix_use_proc(void)
2585 {
2586 return getenv("PROC_NET_UNIX") || getenv("PROC_ROOT");
2587 }
2588
unix_stats_print(struct sockstat * list,struct filter * f)2589 static void unix_stats_print(struct sockstat *list, struct filter *f)
2590 {
2591 struct sockstat *s;
2592 char *peer;
2593 char *ctx_buf = NULL;
2594 bool use_proc = unix_use_proc();
2595 char port_name[30] = {};
2596
2597 for (s = list; s; s = s->next) {
2598 if (!(f->states & (1 << s->state)))
2599 continue;
2600 if (unix_type_skip(s, f))
2601 continue;
2602
2603 peer = "*";
2604 if (s->peer_name)
2605 peer = s->peer_name;
2606
2607 if (s->rport && use_proc) {
2608 struct sockstat *p;
2609
2610 for (p = list; p; p = p->next) {
2611 if (s->rport == p->lport)
2612 break;
2613 }
2614
2615 if (!p) {
2616 peer = "?";
2617 } else {
2618 peer = p->name ? : "*";
2619 }
2620 }
2621
2622 if (use_proc && f->f) {
2623 struct sockstat st;
2624 st.local.family = AF_UNIX;
2625 st.remote.family = AF_UNIX;
2626 memcpy(st.local.data, &s->name, sizeof(s->name));
2627 if (strcmp(peer, "*") == 0)
2628 memset(st.remote.data, 0, sizeof(peer));
2629 else
2630 memcpy(st.remote.data, &peer, sizeof(peer));
2631 if (run_ssfilter(f->f, &st) == 0)
2632 continue;
2633 }
2634
2635 sock_state_print(s, unix_netid_name(s->type));
2636
2637 sock_addr_print(s->name ?: "*", " ",
2638 int_to_str(s->lport, port_name), NULL);
2639 sock_addr_print(peer, " ", int_to_str(s->rport, port_name),
2640 NULL);
2641
2642 if (show_proc_ctx || show_sock_ctx) {
2643 if (find_entry(s->ino, &ctx_buf,
2644 (show_proc_ctx & show_sock_ctx) ?
2645 PROC_SOCK_CTX : PROC_CTX) > 0) {
2646 printf(" users:(%s)", ctx_buf);
2647 free(ctx_buf);
2648 }
2649 } else if (show_users) {
2650 if (find_entry(s->ino, &ctx_buf, USERS) > 0) {
2651 printf(" users:(%s)", ctx_buf);
2652 free(ctx_buf);
2653 }
2654 }
2655 printf("\n");
2656 }
2657 }
2658
unix_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)2659 static int unix_show_sock(const struct sockaddr_nl *addr, struct nlmsghdr *nlh,
2660 void *arg)
2661 {
2662 struct filter *f = (struct filter *)arg;
2663 struct unix_diag_msg *r = NLMSG_DATA(nlh);
2664 struct rtattr *tb[UNIX_DIAG_MAX+1];
2665 char name[128];
2666 struct sockstat stat = { .name = "*", .peer_name = "*" };
2667
2668 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr*)(r+1),
2669 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2670
2671 stat.type = r->udiag_type;
2672 stat.state = r->udiag_state;
2673 stat.ino = stat.lport = r->udiag_ino;
2674 stat.local.family = stat.remote.family = AF_UNIX;
2675
2676 if (unix_type_skip(&stat, f))
2677 return 0;
2678
2679 if (tb[UNIX_DIAG_RQLEN]) {
2680 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
2681 stat.rq = rql->udiag_rqueue;
2682 stat.wq = rql->udiag_wqueue;
2683 }
2684 if (tb[UNIX_DIAG_NAME]) {
2685 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
2686
2687 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
2688 name[len] = '\0';
2689 if (name[0] == '\0')
2690 name[0] = '@';
2691 stat.name = &name[0];
2692 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
2693 }
2694 if (tb[UNIX_DIAG_PEER])
2695 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
2696
2697 if (f->f && run_ssfilter(f->f, &stat) == 0)
2698 return 0;
2699
2700 unix_stats_print(&stat, f);
2701
2702 if (show_mem) {
2703 printf("\t");
2704 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
2705 }
2706 if (show_details) {
2707 if (tb[UNIX_DIAG_SHUTDOWN]) {
2708 unsigned char mask;
2709 mask = *(__u8 *)RTA_DATA(tb[UNIX_DIAG_SHUTDOWN]);
2710 printf(" %c-%c", mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
2711 }
2712 }
2713 if (show_mem || show_details)
2714 printf("\n");
2715
2716 return 0;
2717 }
2718
handle_netlink_request(struct filter * f,struct nlmsghdr * req,size_t size,rtnl_filter_t show_one_sock)2719 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
2720 size_t size, rtnl_filter_t show_one_sock)
2721 {
2722 int ret = -1;
2723 struct rtnl_handle rth;
2724
2725 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
2726 return -1;
2727
2728 rth.dump = MAGIC_SEQ;
2729
2730 if (rtnl_send(&rth, req, size) < 0)
2731 goto Exit;
2732
2733 if (rtnl_dump_filter(&rth, show_one_sock, f))
2734 goto Exit;
2735
2736 ret = 0;
2737 Exit:
2738 rtnl_close(&rth);
2739 return ret;
2740 }
2741
unix_show_netlink(struct filter * f)2742 static int unix_show_netlink(struct filter *f)
2743 {
2744 DIAG_REQUEST(req, struct unix_diag_req r);
2745
2746 req.r.sdiag_family = AF_UNIX;
2747 req.r.udiag_states = f->states;
2748 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
2749 if (show_mem)
2750 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
2751
2752 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
2753 }
2754
unix_show(struct filter * f)2755 static int unix_show(struct filter *f)
2756 {
2757 FILE *fp;
2758 char buf[256];
2759 char name[128];
2760 int newformat = 0;
2761 int cnt;
2762 struct sockstat *list = NULL;
2763
2764 if (!filter_af_get(f, AF_UNIX))
2765 return 0;
2766
2767 if (!unix_use_proc() && unix_show_netlink(f) == 0)
2768 return 0;
2769
2770 if ((fp = net_unix_open()) == NULL)
2771 return -1;
2772 if (!fgets(buf, sizeof(buf), fp)) {
2773 fclose(fp);
2774 return -1;
2775 }
2776
2777 if (memcmp(buf, "Peer", 4) == 0)
2778 newformat = 1;
2779 cnt = 0;
2780
2781 while (fgets(buf, sizeof(buf), fp)) {
2782 struct sockstat *u, **insp;
2783 int flags;
2784
2785 if (!(u = calloc(1, sizeof(*u))))
2786 break;
2787 u->name = NULL;
2788 u->peer_name = NULL;
2789
2790 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
2791 &u->rport, &u->rq, &u->wq, &flags, &u->type,
2792 &u->state, &u->ino, name) < 8)
2793 name[0] = 0;
2794
2795 u->lport = u->ino;
2796 u->local.family = u->remote.family = AF_UNIX;
2797
2798 if (flags & (1 << 16)) {
2799 u->state = SS_LISTEN;
2800 } else {
2801 u->state = unix_state_map[u->state-1];
2802 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
2803 u->state = SS_ESTABLISHED;
2804 }
2805
2806 if (!newformat) {
2807 u->rport = 0;
2808 u->rq = 0;
2809 u->wq = 0;
2810 }
2811
2812 insp = &list;
2813 while (*insp) {
2814 if (u->type < (*insp)->type ||
2815 (u->type == (*insp)->type &&
2816 u->ino < (*insp)->ino))
2817 break;
2818 insp = &(*insp)->next;
2819 }
2820 u->next = *insp;
2821 *insp = u;
2822
2823 if (name[0]) {
2824 if ((u->name = malloc(strlen(name)+1)) == NULL)
2825 break;
2826 strcpy(u->name, name);
2827 }
2828 if (++cnt > MAX_UNIX_REMEMBER) {
2829 unix_stats_print(list, f);
2830 unix_list_free(list);
2831 list = NULL;
2832 cnt = 0;
2833 }
2834 }
2835 fclose(fp);
2836 if (list) {
2837 unix_stats_print(list, f);
2838 unix_list_free(list);
2839 list = NULL;
2840 cnt = 0;
2841 }
2842
2843 return 0;
2844 }
2845
packet_stats_print(struct sockstat * s,const struct filter * f)2846 static int packet_stats_print(struct sockstat *s, const struct filter *f)
2847 {
2848 char *buf = NULL;
2849 const char *addr, *port;
2850 char ll_name[16];
2851
2852 if (f->f) {
2853 s->local.family = AF_PACKET;
2854 s->remote.family = AF_PACKET;
2855 s->local.data[0] = s->prot;
2856 if (run_ssfilter(f->f, s) == 0)
2857 return 1;
2858 }
2859
2860 sock_state_print(s, s->type == SOCK_RAW ? "p_raw" : "p_dgr");
2861
2862 if (s->prot == 3)
2863 addr = "*";
2864 else
2865 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
2866
2867 if (s->iface == 0)
2868 port = "*";
2869 else
2870 port = xll_index_to_name(s->iface);
2871
2872 sock_addr_print(addr, ":", port, NULL);
2873 sock_addr_print("", "*", "", NULL);
2874
2875 if (show_proc_ctx || show_sock_ctx) {
2876 if (find_entry(s->ino, &buf,
2877 (show_proc_ctx & show_sock_ctx) ?
2878 PROC_SOCK_CTX : PROC_CTX) > 0) {
2879 printf(" users:(%s)", buf);
2880 free(buf);
2881 }
2882 } else if (show_users) {
2883 if (find_entry(s->ino, &buf, USERS) > 0) {
2884 printf(" users:(%s)", buf);
2885 free(buf);
2886 }
2887 }
2888
2889 if (show_details)
2890 sock_details_print(s);
2891
2892 return 0;
2893 }
2894
packet_show_ring(struct packet_diag_ring * ring)2895 static void packet_show_ring(struct packet_diag_ring *ring)
2896 {
2897 printf("blk_size:%d", ring->pdr_block_size);
2898 printf(",blk_nr:%d", ring->pdr_block_nr);
2899 printf(",frm_size:%d", ring->pdr_frame_size);
2900 printf(",frm_nr:%d", ring->pdr_frame_nr);
2901 printf(",tmo:%d", ring->pdr_retire_tmo);
2902 printf(",features:0x%x", ring->pdr_features);
2903 }
2904
packet_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)2905 static int packet_show_sock(const struct sockaddr_nl *addr,
2906 struct nlmsghdr *nlh, void *arg)
2907 {
2908 const struct filter *f = arg;
2909 struct packet_diag_msg *r = NLMSG_DATA(nlh);
2910 struct packet_diag_info *pinfo = NULL;
2911 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
2912 struct rtattr *tb[PACKET_DIAG_MAX+1];
2913 struct sockstat stat = {};
2914 uint32_t fanout = 0;
2915 bool has_fanout = false;
2916
2917 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr*)(r+1),
2918 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2919
2920 /* use /proc/net/packet if all info are not available */
2921 if (!tb[PACKET_DIAG_MEMINFO])
2922 return -1;
2923
2924 stat.type = r->pdiag_type;
2925 stat.prot = r->pdiag_num;
2926 stat.ino = r->pdiag_ino;
2927 stat.state = SS_CLOSE;
2928 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
2929
2930 if (tb[PACKET_DIAG_MEMINFO]) {
2931 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
2932 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
2933 }
2934
2935 if (tb[PACKET_DIAG_INFO]) {
2936 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
2937 stat.lport = stat.iface = pinfo->pdi_index;
2938 }
2939
2940 if (tb[PACKET_DIAG_UID])
2941 stat.uid = *(__u32 *)RTA_DATA(tb[PACKET_DIAG_UID]);
2942
2943 if (tb[PACKET_DIAG_RX_RING])
2944 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
2945
2946 if (tb[PACKET_DIAG_TX_RING])
2947 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
2948
2949 if (tb[PACKET_DIAG_FANOUT]) {
2950 has_fanout = true;
2951 fanout = *(uint32_t *)RTA_DATA(tb[PACKET_DIAG_FANOUT]);
2952 }
2953
2954 if (packet_stats_print(&stat, f))
2955 return 0;
2956
2957 if (show_details) {
2958 if (pinfo) {
2959 printf("\n\tver:%d", pinfo->pdi_version);
2960 printf(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
2961 printf(" flags( ");
2962 if (pinfo->pdi_flags & PDI_RUNNING)
2963 printf("running");
2964 if (pinfo->pdi_flags & PDI_AUXDATA)
2965 printf(" auxdata");
2966 if (pinfo->pdi_flags & PDI_ORIGDEV)
2967 printf(" origdev");
2968 if (pinfo->pdi_flags & PDI_VNETHDR)
2969 printf(" vnethdr");
2970 if (pinfo->pdi_flags & PDI_LOSS)
2971 printf(" loss");
2972 if (!pinfo->pdi_flags)
2973 printf("0");
2974 printf(" )");
2975 }
2976 if (ring_rx) {
2977 printf("\n\tring_rx(");
2978 packet_show_ring(ring_rx);
2979 printf(")");
2980 }
2981 if (ring_tx) {
2982 printf("\n\tring_tx(");
2983 packet_show_ring(ring_tx);
2984 printf(")");
2985 }
2986 if (has_fanout) {
2987 uint16_t type = (fanout >> 16) & 0xffff;
2988
2989 printf("\n\tfanout(");
2990 printf("id:%d,", fanout & 0xffff);
2991 printf("type:");
2992
2993 if (type == 0)
2994 printf("hash");
2995 else if (type == 1)
2996 printf("lb");
2997 else if (type == 2)
2998 printf("cpu");
2999 else if (type == 3)
3000 printf("roll");
3001 else if (type == 4)
3002 printf("random");
3003 else if (type == 5)
3004 printf("qm");
3005 else
3006 printf("0x%x", type);
3007
3008 printf(")");
3009 }
3010 }
3011
3012 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
3013 struct sock_filter *fil =
3014 RTA_DATA(tb[PACKET_DIAG_FILTER]);
3015 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
3016 sizeof(struct sock_filter);
3017
3018 printf("\n\tbpf filter (%d): ", num);
3019 while (num) {
3020 printf(" 0x%02x %u %u %u,",
3021 fil->code, fil->jt, fil->jf, fil->k);
3022 num--;
3023 fil++;
3024 }
3025 }
3026 printf("\n");
3027 return 0;
3028 }
3029
packet_show_netlink(struct filter * f)3030 static int packet_show_netlink(struct filter *f)
3031 {
3032 DIAG_REQUEST(req, struct packet_diag_req r);
3033
3034 req.r.sdiag_family = AF_PACKET;
3035 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
3036 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
3037
3038 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
3039 }
3040
packet_show_line(char * buf,const struct filter * f,int fam)3041 static int packet_show_line(char *buf, const struct filter *f, int fam)
3042 {
3043 unsigned long long sk;
3044 struct sockstat stat = {};
3045 int type, prot, iface, state, rq, uid, ino;
3046
3047 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
3048 &sk,
3049 &type, &prot, &iface, &state,
3050 &rq, &uid, &ino);
3051
3052 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
3053 return 0;
3054 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
3055 return 0;
3056
3057 stat.type = type;
3058 stat.prot = prot;
3059 stat.lport = stat.iface = iface;
3060 stat.state = state;
3061 stat.rq = rq;
3062 stat.uid = uid;
3063 stat.ino = ino;
3064 stat.state = SS_CLOSE;
3065
3066 if (packet_stats_print(&stat, f))
3067 return 0;
3068
3069 printf("\n");
3070 return 0;
3071 }
3072
packet_show(struct filter * f)3073 static int packet_show(struct filter *f)
3074 {
3075 FILE *fp;
3076 int rc = 0;
3077
3078 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
3079 return 0;
3080
3081 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
3082 packet_show_netlink(f) == 0)
3083 return 0;
3084
3085 if ((fp = net_packet_open()) == NULL)
3086 return -1;
3087 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
3088 rc = -1;
3089
3090 fclose(fp);
3091 return rc;
3092 }
3093
netlink_show_one(struct filter * f,int prot,int pid,unsigned groups,int state,int dst_pid,unsigned dst_group,int rq,int wq,unsigned long long sk,unsigned long long cb)3094 static int netlink_show_one(struct filter *f,
3095 int prot, int pid, unsigned groups,
3096 int state, int dst_pid, unsigned dst_group,
3097 int rq, int wq,
3098 unsigned long long sk, unsigned long long cb)
3099 {
3100 struct sockstat st;
3101 SPRINT_BUF(prot_buf) = {};
3102 const char *prot_name;
3103 char procname[64] = {};
3104
3105 st.state = SS_CLOSE;
3106 st.rq = rq;
3107 st.wq = wq;
3108
3109 if (f->f) {
3110 st.local.family = AF_NETLINK;
3111 st.remote.family = AF_NETLINK;
3112 st.rport = -1;
3113 st.lport = pid;
3114 st.local.data[0] = prot;
3115 if (run_ssfilter(f->f, &st) == 0)
3116 return 1;
3117 }
3118
3119 sock_state_print(&st, "nl");
3120
3121 if (resolve_services)
3122 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
3123 else
3124 prot_name = int_to_str(prot, prot_buf);
3125
3126 if (pid == -1) {
3127 procname[0] = '*';
3128 } else if (resolve_services) {
3129 int done = 0;
3130 if (!pid) {
3131 done = 1;
3132 strncpy(procname, "kernel", 6);
3133 } else if (pid > 0) {
3134 FILE *fp;
3135 snprintf(procname, sizeof(procname), "%s/%d/stat",
3136 getenv("PROC_ROOT") ? : "/proc", pid);
3137 if ((fp = fopen(procname, "r")) != NULL) {
3138 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
3139 snprintf(procname+strlen(procname),
3140 sizeof(procname)-strlen(procname),
3141 "/%d", pid);
3142 done = 1;
3143 }
3144 fclose(fp);
3145 }
3146 }
3147 if (!done)
3148 int_to_str(pid, procname);
3149 } else {
3150 int_to_str(pid, procname);
3151 }
3152
3153 sock_addr_print(prot_name, ":", procname, NULL);
3154
3155 if (state == NETLINK_CONNECTED) {
3156 char dst_group_buf[30];
3157 char dst_pid_buf[30];
3158 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
3159 int_to_str(dst_pid, dst_pid_buf), NULL);
3160 } else {
3161 sock_addr_print("", "*", "", NULL);
3162 }
3163
3164 char *pid_context = NULL;
3165 if (show_proc_ctx) {
3166 /* The pid value will either be:
3167 * 0 if destination kernel - show kernel initial context.
3168 * A valid process pid - use getpidcon.
3169 * A unique value allocated by the kernel or netlink user
3170 * to the process - show context as "not available".
3171 */
3172 if (!pid)
3173 security_get_initial_context("kernel", &pid_context);
3174 else if (pid > 0)
3175 getpidcon(pid, &pid_context);
3176
3177 if (pid_context != NULL) {
3178 printf("proc_ctx=%-*s ", serv_width, pid_context);
3179 free(pid_context);
3180 } else {
3181 printf("proc_ctx=%-*s ", serv_width, "unavailable");
3182 }
3183 }
3184
3185 if (show_details) {
3186 printf(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
3187 }
3188 printf("\n");
3189
3190 return 0;
3191 }
3192
netlink_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)3193 static int netlink_show_sock(const struct sockaddr_nl *addr,
3194 struct nlmsghdr *nlh, void *arg)
3195 {
3196 struct filter *f = (struct filter *)arg;
3197 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
3198 struct rtattr *tb[NETLINK_DIAG_MAX+1];
3199 int rq = 0, wq = 0;
3200 unsigned long groups = 0;
3201
3202 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr*)(r+1),
3203 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3204
3205 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
3206 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
3207
3208 if (tb[NETLINK_DIAG_MEMINFO]) {
3209 const __u32 *skmeminfo;
3210 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
3211
3212 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
3213 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
3214 }
3215
3216 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
3217 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
3218 rq, wq, 0, 0)) {
3219 return 0;
3220 }
3221
3222 if (show_mem) {
3223 printf("\t");
3224 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
3225 printf("\n");
3226 }
3227
3228 return 0;
3229 }
3230
netlink_show_netlink(struct filter * f)3231 static int netlink_show_netlink(struct filter *f)
3232 {
3233 DIAG_REQUEST(req, struct netlink_diag_req r);
3234
3235 req.r.sdiag_family = AF_NETLINK;
3236 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
3237 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
3238
3239 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
3240 }
3241
netlink_show(struct filter * f)3242 static int netlink_show(struct filter *f)
3243 {
3244 FILE *fp;
3245 char buf[256];
3246 int prot, pid;
3247 unsigned groups;
3248 int rq, wq, rc;
3249 unsigned long long sk, cb;
3250
3251 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
3252 return 0;
3253
3254 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
3255 netlink_show_netlink(f) == 0)
3256 return 0;
3257
3258 if ((fp = net_netlink_open()) == NULL)
3259 return -1;
3260 if (!fgets(buf, sizeof(buf), fp)) {
3261 fclose(fp);
3262 return -1;
3263 }
3264
3265 while (fgets(buf, sizeof(buf), fp)) {
3266 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
3267 &sk,
3268 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
3269
3270 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
3271 }
3272
3273 fclose(fp);
3274 return 0;
3275 }
3276
3277 struct sock_diag_msg {
3278 __u8 sdiag_family;
3279 };
3280
generic_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)3281 static int generic_show_sock(const struct sockaddr_nl *addr,
3282 struct nlmsghdr *nlh, void *arg)
3283 {
3284 struct sock_diag_msg *r = NLMSG_DATA(nlh);
3285 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
3286
3287 switch (r->sdiag_family) {
3288 case AF_INET:
3289 case AF_INET6:
3290 return show_one_inet_sock(addr, nlh, &inet_arg);
3291 case AF_UNIX:
3292 return unix_show_sock(addr, nlh, arg);
3293 case AF_PACKET:
3294 return packet_show_sock(addr, nlh, arg);
3295 case AF_NETLINK:
3296 return netlink_show_sock(addr, nlh, arg);
3297 default:
3298 return -1;
3299 }
3300 }
3301
handle_follow_request(struct filter * f)3302 static int handle_follow_request(struct filter *f)
3303 {
3304 int ret = -1;
3305 int groups = 0;
3306 struct rtnl_handle rth;
3307
3308 if (f->families & (1 << AF_INET) && f->dbs & (1 << TCP_DB))
3309 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
3310 if (f->families & (1 << AF_INET) && f->dbs & (1 << UDP_DB))
3311 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
3312 if (f->families & (1 << AF_INET6) && f->dbs & (1 << TCP_DB))
3313 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
3314 if (f->families & (1 << AF_INET6) && f->dbs & (1 << UDP_DB))
3315 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
3316
3317 if (groups == 0)
3318 return -1;
3319
3320 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
3321 return -1;
3322
3323 rth.dump = 0;
3324 rth.local.nl_pid = 0;
3325
3326 if (rtnl_dump_filter(&rth, generic_show_sock, f))
3327 goto Exit;
3328
3329 ret = 0;
3330 Exit:
3331 rtnl_close(&rth);
3332 return ret;
3333 }
3334
3335 struct snmpstat
3336 {
3337 int tcp_estab;
3338 };
3339
get_snmp_int(char * proto,char * key,int * result)3340 static int get_snmp_int(char *proto, char *key, int *result)
3341 {
3342 char buf[1024];
3343 FILE *fp;
3344 int protolen = strlen(proto);
3345 int keylen = strlen(key);
3346
3347 *result = 0;
3348
3349 if ((fp = net_snmp_open()) == NULL)
3350 return -1;
3351
3352 while (fgets(buf, sizeof(buf), fp) != NULL) {
3353 char *p = buf;
3354 int pos = 0;
3355 if (memcmp(buf, proto, protolen))
3356 continue;
3357 while ((p = strchr(p, ' ')) != NULL) {
3358 pos++;
3359 p++;
3360 if (memcmp(p, key, keylen) == 0 &&
3361 (p[keylen] == ' ' || p[keylen] == '\n'))
3362 break;
3363 }
3364 if (fgets(buf, sizeof(buf), fp) == NULL)
3365 break;
3366 if (memcmp(buf, proto, protolen))
3367 break;
3368 p = buf;
3369 while ((p = strchr(p, ' ')) != NULL) {
3370 p++;
3371 if (--pos == 0) {
3372 sscanf(p, "%d", result);
3373 fclose(fp);
3374 return 0;
3375 }
3376 }
3377 }
3378
3379 fclose(fp);
3380 errno = ESRCH;
3381 return -1;
3382 }
3383
3384
3385 /* Get stats from sockstat */
3386
3387 struct ssummary
3388 {
3389 int socks;
3390 int tcp_mem;
3391 int tcp_total;
3392 int tcp_orphans;
3393 int tcp_tws;
3394 int tcp4_hashed;
3395 int udp4;
3396 int raw4;
3397 int frag4;
3398 int frag4_mem;
3399 int tcp6_hashed;
3400 int udp6;
3401 int raw6;
3402 int frag6;
3403 int frag6_mem;
3404 };
3405
get_sockstat_line(char * line,struct ssummary * s)3406 static void get_sockstat_line(char *line, struct ssummary *s)
3407 {
3408 char id[256], rem[256];
3409
3410 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
3411 return;
3412
3413 if (strcmp(id, "sockets:") == 0)
3414 sscanf(rem, "%*s%d", &s->socks);
3415 else if (strcmp(id, "UDP:") == 0)
3416 sscanf(rem, "%*s%d", &s->udp4);
3417 else if (strcmp(id, "UDP6:") == 0)
3418 sscanf(rem, "%*s%d", &s->udp6);
3419 else if (strcmp(id, "RAW:") == 0)
3420 sscanf(rem, "%*s%d", &s->raw4);
3421 else if (strcmp(id, "RAW6:") == 0)
3422 sscanf(rem, "%*s%d", &s->raw6);
3423 else if (strcmp(id, "TCP6:") == 0)
3424 sscanf(rem, "%*s%d", &s->tcp6_hashed);
3425 else if (strcmp(id, "FRAG:") == 0)
3426 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
3427 else if (strcmp(id, "FRAG6:") == 0)
3428 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
3429 else if (strcmp(id, "TCP:") == 0)
3430 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
3431 &s->tcp4_hashed,
3432 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
3433 }
3434
get_sockstat(struct ssummary * s)3435 static int get_sockstat(struct ssummary *s)
3436 {
3437 char buf[256];
3438 FILE *fp;
3439
3440 memset(s, 0, sizeof(*s));
3441
3442 if ((fp = net_sockstat_open()) == NULL)
3443 return -1;
3444 while(fgets(buf, sizeof(buf), fp) != NULL)
3445 get_sockstat_line(buf, s);
3446 fclose(fp);
3447
3448 if ((fp = net_sockstat6_open()) == NULL)
3449 return 0;
3450 while(fgets(buf, sizeof(buf), fp) != NULL)
3451 get_sockstat_line(buf, s);
3452 fclose(fp);
3453
3454 return 0;
3455 }
3456
print_summary(void)3457 static int print_summary(void)
3458 {
3459 struct ssummary s;
3460 struct snmpstat sn;
3461
3462 if (get_sockstat(&s) < 0)
3463 perror("ss: get_sockstat");
3464 if (get_snmp_int("Tcp:", "CurrEstab", &sn.tcp_estab) < 0)
3465 perror("ss: get_snmpstat");
3466
3467 get_slabstat(&slabstat);
3468
3469 printf("Total: %d (kernel %d)\n", s.socks, slabstat.socks);
3470
3471 printf("TCP: %d (estab %d, closed %d, orphaned %d, synrecv %d, timewait %d/%d), ports %d\n",
3472 s.tcp_total + slabstat.tcp_syns + s.tcp_tws,
3473 sn.tcp_estab,
3474 s.tcp_total - (s.tcp4_hashed+s.tcp6_hashed-s.tcp_tws),
3475 s.tcp_orphans,
3476 slabstat.tcp_syns,
3477 s.tcp_tws, slabstat.tcp_tws,
3478 slabstat.tcp_ports
3479 );
3480
3481 printf("\n");
3482 printf("Transport Total IP IPv6\n");
3483 printf("* %-9d %-9s %-9s\n", slabstat.socks, "-", "-");
3484 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
3485 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
3486 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
3487 printf("INET %-9d %-9d %-9d\n",
3488 s.raw4+s.udp4+s.tcp4_hashed+
3489 s.raw6+s.udp6+s.tcp6_hashed,
3490 s.raw4+s.udp4+s.tcp4_hashed,
3491 s.raw6+s.udp6+s.tcp6_hashed);
3492 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
3493
3494 printf("\n");
3495
3496 return 0;
3497 }
3498
_usage(FILE * dest)3499 static void _usage(FILE *dest)
3500 {
3501 fprintf(dest,
3502 "Usage: ss [ OPTIONS ]\n"
3503 " ss [ OPTIONS ] [ FILTER ]\n"
3504 " -h, --help this message\n"
3505 " -V, --version output version information\n"
3506 " -n, --numeric don't resolve service names\n"
3507 " -r, --resolve resolve host names\n"
3508 " -a, --all display all sockets\n"
3509 " -l, --listening display listening sockets\n"
3510 " -o, --options show timer information\n"
3511 " -e, --extended show detailed socket information\n"
3512 " -m, --memory show socket memory usage\n"
3513 " -p, --processes show process using socket\n"
3514 " -i, --info show internal TCP information\n"
3515 " -s, --summary show socket usage summary\n"
3516 " -b, --bpf show bpf filter socket information\n"
3517 " -E, --events continually display sockets as they are destroyed\n"
3518 " -Z, --context display process SELinux security contexts\n"
3519 " -z, --contexts display process and socket SELinux security contexts\n"
3520 " -N, --net switch to the specified network namespace name\n"
3521 "\n"
3522 " -4, --ipv4 display only IP version 4 sockets\n"
3523 " -6, --ipv6 display only IP version 6 sockets\n"
3524 " -0, --packet display PACKET sockets\n"
3525 " -t, --tcp display only TCP sockets\n"
3526 " -u, --udp display only UDP sockets\n"
3527 " -d, --dccp display only DCCP sockets\n"
3528 " -w, --raw display only RAW sockets\n"
3529 " -x, --unix display only Unix domain sockets\n"
3530 " -f, --family=FAMILY display sockets of type FAMILY\n"
3531 "\n"
3532 " -K, --kill forcibly close sockets, display what was closed\n"
3533 "\n"
3534 " -A, --query=QUERY, --socket=QUERY\n"
3535 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink}[,QUERY]\n"
3536 "\n"
3537 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
3538 " -F, --filter=FILE read filter information from FILE\n"
3539 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
3540 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
3541 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listen|closing}\n"
3542 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
3543 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
3544 " bucket := {syn-recv|time-wait}\n"
3545 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listen|closing}\n"
3546 );
3547 }
3548
3549 static void help(void) __attribute__((noreturn));
help(void)3550 static void help(void)
3551 {
3552 _usage(stdout);
3553 exit(0);
3554 }
3555
3556 static void usage(void) __attribute__((noreturn));
usage(void)3557 static void usage(void)
3558 {
3559 _usage(stderr);
3560 exit(-1);
3561 }
3562
3563
scan_state(const char * state)3564 static int scan_state(const char *state)
3565 {
3566 int i;
3567 if (strcasecmp(state, "close") == 0 ||
3568 strcasecmp(state, "closed") == 0)
3569 return (1<<SS_CLOSE);
3570 if (strcasecmp(state, "syn-rcv") == 0)
3571 return (1<<SS_SYN_RECV);
3572 if (strcasecmp(state, "established") == 0)
3573 return (1<<SS_ESTABLISHED);
3574 if (strcasecmp(state, "all") == 0)
3575 return SS_ALL;
3576 if (strcasecmp(state, "connected") == 0)
3577 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
3578 if (strcasecmp(state, "synchronized") == 0)
3579 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
3580 if (strcasecmp(state, "bucket") == 0)
3581 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
3582 if (strcasecmp(state, "big") == 0)
3583 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
3584 for (i=0; i<SS_MAX; i++) {
3585 if (strcasecmp(state, sstate_namel[i]) == 0)
3586 return (1<<i);
3587 }
3588
3589 fprintf(stderr, "ss: wrong state name: %s\n", state);
3590 exit(-1);
3591 }
3592
3593 static const struct option long_opts[] = {
3594 { "numeric", 0, 0, 'n' },
3595 { "resolve", 0, 0, 'r' },
3596 { "options", 0, 0, 'o' },
3597 { "extended", 0, 0, 'e' },
3598 { "memory", 0, 0, 'm' },
3599 { "info", 0, 0, 'i' },
3600 { "processes", 0, 0, 'p' },
3601 { "bpf", 0, 0, 'b' },
3602 { "events", 0, 0, 'E' },
3603 { "dccp", 0, 0, 'd' },
3604 { "tcp", 0, 0, 't' },
3605 { "udp", 0, 0, 'u' },
3606 { "raw", 0, 0, 'w' },
3607 { "unix", 0, 0, 'x' },
3608 { "all", 0, 0, 'a' },
3609 { "listening", 0, 0, 'l' },
3610 { "ipv4", 0, 0, '4' },
3611 { "ipv6", 0, 0, '6' },
3612 { "packet", 0, 0, '0' },
3613 { "family", 1, 0, 'f' },
3614 { "socket", 1, 0, 'A' },
3615 { "query", 1, 0, 'A' },
3616 { "summary", 0, 0, 's' },
3617 { "diag", 1, 0, 'D' },
3618 { "filter", 1, 0, 'F' },
3619 { "version", 0, 0, 'V' },
3620 { "help", 0, 0, 'h' },
3621 { "context", 0, 0, 'Z' },
3622 { "contexts", 0, 0, 'z' },
3623 { "net", 1, 0, 'N' },
3624 { "kill", 0, 0, 'K' },
3625 { 0 }
3626
3627 };
3628
main(int argc,char * argv[])3629 int main(int argc, char *argv[])
3630 {
3631 int saw_states = 0;
3632 int saw_query = 0;
3633 int do_summary = 0;
3634 const char *dump_tcpdiag = NULL;
3635 FILE *filter_fp = NULL;
3636 int ch;
3637 int state_filter = 0;
3638
3639 while ((ch = getopt_long(argc, argv, "dhaletuwxnro460spbEf:miA:D:F:vVzZN:K",
3640 long_opts, NULL)) != EOF) {
3641 switch(ch) {
3642 case 'n':
3643 resolve_services = 0;
3644 break;
3645 case 'r':
3646 resolve_hosts = 1;
3647 break;
3648 case 'o':
3649 show_options = 1;
3650 break;
3651 case 'e':
3652 show_options = 1;
3653 show_details++;
3654 break;
3655 case 'm':
3656 show_mem = 1;
3657 break;
3658 case 'i':
3659 show_tcpinfo = 1;
3660 break;
3661 case 'p':
3662 show_users++;
3663 user_ent_hash_build();
3664 break;
3665 case 'b':
3666 show_options = 1;
3667 show_bpf++;
3668 break;
3669 case 'E':
3670 follow_events = 1;
3671 break;
3672 case 'd':
3673 filter_db_set(¤t_filter, DCCP_DB);
3674 break;
3675 case 't':
3676 filter_db_set(¤t_filter, TCP_DB);
3677 break;
3678 case 'u':
3679 filter_db_set(¤t_filter, UDP_DB);
3680 break;
3681 case 'w':
3682 filter_db_set(¤t_filter, RAW_DB);
3683 break;
3684 case 'x':
3685 filter_af_set(¤t_filter, AF_UNIX);
3686 break;
3687 case 'a':
3688 state_filter = SS_ALL;
3689 break;
3690 case 'l':
3691 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
3692 break;
3693 case '4':
3694 filter_af_set(¤t_filter, AF_INET);
3695 break;
3696 case '6':
3697 filter_af_set(¤t_filter, AF_INET6);
3698 break;
3699 case '0':
3700 filter_af_set(¤t_filter, AF_PACKET);
3701 break;
3702 case 'f':
3703 if (strcmp(optarg, "inet") == 0)
3704 filter_af_set(¤t_filter, AF_INET);
3705 else if (strcmp(optarg, "inet6") == 0)
3706 filter_af_set(¤t_filter, AF_INET6);
3707 else if (strcmp(optarg, "link") == 0)
3708 filter_af_set(¤t_filter, AF_PACKET);
3709 else if (strcmp(optarg, "unix") == 0)
3710 filter_af_set(¤t_filter, AF_UNIX);
3711 else if (strcmp(optarg, "netlink") == 0)
3712 filter_af_set(¤t_filter, AF_NETLINK);
3713 else if (strcmp(optarg, "help") == 0)
3714 help();
3715 else {
3716 fprintf(stderr, "ss: \"%s\" is invalid family\n",
3717 optarg);
3718 usage();
3719 }
3720 break;
3721 case 'A':
3722 {
3723 char *p, *p1;
3724 if (!saw_query) {
3725 current_filter.dbs = 0;
3726 state_filter = state_filter ?
3727 state_filter : SS_CONN;
3728 saw_query = 1;
3729 do_default = 0;
3730 }
3731 p = p1 = optarg;
3732 do {
3733 if ((p1 = strchr(p, ',')) != NULL)
3734 *p1 = 0;
3735 if (strcmp(p, "all") == 0) {
3736 filter_default_dbs(¤t_filter);
3737 } else if (strcmp(p, "inet") == 0) {
3738 filter_db_set(¤t_filter, UDP_DB);
3739 filter_db_set(¤t_filter, DCCP_DB);
3740 filter_db_set(¤t_filter, TCP_DB);
3741 filter_db_set(¤t_filter, RAW_DB);
3742 } else if (strcmp(p, "udp") == 0) {
3743 filter_db_set(¤t_filter, UDP_DB);
3744 } else if (strcmp(p, "dccp") == 0) {
3745 filter_db_set(¤t_filter, DCCP_DB);
3746 } else if (strcmp(p, "tcp") == 0) {
3747 filter_db_set(¤t_filter, TCP_DB);
3748 } else if (strcmp(p, "raw") == 0) {
3749 filter_db_set(¤t_filter, RAW_DB);
3750 } else if (strcmp(p, "unix") == 0) {
3751 filter_db_set(¤t_filter, UNIX_ST_DB);
3752 filter_db_set(¤t_filter, UNIX_DG_DB);
3753 filter_db_set(¤t_filter, UNIX_SQ_DB);
3754 } else if (strcasecmp(p, "unix_stream") == 0 ||
3755 strcmp(p, "u_str") == 0) {
3756 filter_db_set(¤t_filter, UNIX_ST_DB);
3757 } else if (strcasecmp(p, "unix_dgram") == 0 ||
3758 strcmp(p, "u_dgr") == 0) {
3759 filter_db_set(¤t_filter, UNIX_DG_DB);
3760 } else if (strcasecmp(p, "unix_seqpacket") == 0 ||
3761 strcmp(p, "u_seq") == 0) {
3762 filter_db_set(¤t_filter, UNIX_SQ_DB);
3763 } else if (strcmp(p, "packet") == 0) {
3764 filter_db_set(¤t_filter, PACKET_R_DB);
3765 filter_db_set(¤t_filter, PACKET_DG_DB);
3766 } else if (strcmp(p, "packet_raw") == 0 ||
3767 strcmp(p, "p_raw") == 0) {
3768 filter_db_set(¤t_filter, PACKET_R_DB);
3769 } else if (strcmp(p, "packet_dgram") == 0 ||
3770 strcmp(p, "p_dgr") == 0) {
3771 filter_db_set(¤t_filter, PACKET_DG_DB);
3772 } else if (strcmp(p, "netlink") == 0) {
3773 filter_db_set(¤t_filter, NETLINK_DB);
3774 } else {
3775 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
3776 usage();
3777 }
3778 p = p1 + 1;
3779 } while (p1);
3780 break;
3781 }
3782 case 's':
3783 do_summary = 1;
3784 break;
3785 case 'D':
3786 dump_tcpdiag = optarg;
3787 break;
3788 case 'F':
3789 if (filter_fp) {
3790 fprintf(stderr, "More than one filter file\n");
3791 exit(-1);
3792 }
3793 if (optarg[0] == '-')
3794 filter_fp = stdin;
3795 else
3796 filter_fp = fopen(optarg, "r");
3797 if (!filter_fp) {
3798 perror("fopen filter file");
3799 exit(-1);
3800 }
3801 break;
3802 case 'v':
3803 case 'V':
3804 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
3805 exit(0);
3806 case 'z':
3807 show_sock_ctx++;
3808 case 'Z':
3809 if (is_selinux_enabled() <= 0) {
3810 fprintf(stderr, "ss: SELinux is not enabled.\n");
3811 exit(1);
3812 }
3813 show_proc_ctx++;
3814 user_ent_hash_build();
3815 break;
3816 case 'N':
3817 if (netns_switch(optarg))
3818 exit(1);
3819 break;
3820 case 'K':
3821 current_filter.kill = 1;
3822 break;
3823 case 'h':
3824 help();
3825 case '?':
3826 default:
3827 usage();
3828 }
3829 }
3830
3831 argc -= optind;
3832 argv += optind;
3833
3834 if (do_summary) {
3835 print_summary();
3836 if (do_default && argc == 0)
3837 exit(0);
3838 }
3839
3840 while (argc > 0) {
3841 if (strcmp(*argv, "state") == 0) {
3842 NEXT_ARG();
3843 if (!saw_states)
3844 state_filter = 0;
3845 state_filter |= scan_state(*argv);
3846 saw_states = 1;
3847 } else if (strcmp(*argv, "exclude") == 0 ||
3848 strcmp(*argv, "excl") == 0) {
3849 NEXT_ARG();
3850 if (!saw_states)
3851 state_filter = SS_ALL;
3852 state_filter &= ~scan_state(*argv);
3853 saw_states = 1;
3854 } else {
3855 break;
3856 }
3857 argc--; argv++;
3858 }
3859
3860 if (do_default) {
3861 state_filter = state_filter ? state_filter : SS_CONN;
3862 filter_default_dbs(¤t_filter);
3863 }
3864
3865 filter_states_set(¤t_filter, state_filter);
3866 filter_merge_defaults(¤t_filter);
3867
3868 if (resolve_services && resolve_hosts &&
3869 (current_filter.dbs&(UNIX_DBM|(1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB))))
3870 init_service_resolver();
3871
3872
3873 if (current_filter.dbs == 0) {
3874 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
3875 exit(0);
3876 }
3877 if (current_filter.families == 0) {
3878 fprintf(stderr, "ss: no families to show with such filter.\n");
3879 exit(0);
3880 }
3881 if (current_filter.states == 0) {
3882 fprintf(stderr, "ss: no socket states to show with such filter.\n");
3883 exit(0);
3884 }
3885
3886 if (dump_tcpdiag) {
3887 FILE *dump_fp = stdout;
3888 if (!(current_filter.dbs & (1<<TCP_DB))) {
3889 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
3890 exit(0);
3891 }
3892 if (dump_tcpdiag[0] != '-') {
3893 dump_fp = fopen(dump_tcpdiag, "w");
3894 if (!dump_tcpdiag) {
3895 perror("fopen dump file");
3896 exit(-1);
3897 }
3898 }
3899 inet_show_netlink(¤t_filter, dump_fp, IPPROTO_TCP);
3900 fflush(dump_fp);
3901 exit(0);
3902 }
3903
3904 if (ssfilter_parse(¤t_filter.f, argc, argv, filter_fp))
3905 usage();
3906
3907 netid_width = 0;
3908 if (current_filter.dbs&(current_filter.dbs-1))
3909 netid_width = 5;
3910
3911 state_width = 0;
3912 if (current_filter.states&(current_filter.states-1))
3913 state_width = 10;
3914
3915 screen_width = 80;
3916 if (isatty(STDOUT_FILENO)) {
3917 struct winsize w;
3918
3919 if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) != -1) {
3920 if (w.ws_col > 0)
3921 screen_width = w.ws_col;
3922 }
3923 }
3924
3925 addrp_width = screen_width;
3926 addrp_width -= netid_width+1;
3927 addrp_width -= state_width+1;
3928 addrp_width -= 14;
3929
3930 if (addrp_width&1) {
3931 if (netid_width)
3932 netid_width++;
3933 else if (state_width)
3934 state_width++;
3935 }
3936
3937 addrp_width /= 2;
3938 addrp_width--;
3939
3940 serv_width = resolve_services ? 7 : 5;
3941
3942 if (addrp_width < 15+serv_width+1)
3943 addrp_width = 15+serv_width+1;
3944
3945 addr_width = addrp_width - serv_width - 1;
3946
3947 if (netid_width)
3948 printf("%-*s ", netid_width, "Netid");
3949 if (state_width)
3950 printf("%-*s ", state_width, "State");
3951 printf("%-6s %-6s ", "Recv-Q", "Send-Q");
3952
3953 /* Make enough space for the local/remote port field */
3954 addr_width -= 13;
3955 serv_width += 13;
3956
3957 printf("%*s:%-*s %*s:%-*s\n",
3958 addr_width, "Local Address", serv_width, "Port",
3959 addr_width, "Peer Address", serv_width, "Port");
3960
3961 fflush(stdout);
3962
3963 if (follow_events)
3964 exit(handle_follow_request(¤t_filter));
3965
3966 if (current_filter.dbs & (1<<NETLINK_DB))
3967 netlink_show(¤t_filter);
3968 if (current_filter.dbs & PACKET_DBM)
3969 packet_show(¤t_filter);
3970 if (current_filter.dbs & UNIX_DBM)
3971 unix_show(¤t_filter);
3972 if (current_filter.dbs & (1<<RAW_DB))
3973 raw_show(¤t_filter);
3974 if (current_filter.dbs & (1<<UDP_DB))
3975 udp_show(¤t_filter);
3976 if (current_filter.dbs & (1<<TCP_DB))
3977 tcp_show(¤t_filter, IPPROTO_TCP);
3978 if (current_filter.dbs & (1<<DCCP_DB))
3979 tcp_show(¤t_filter, IPPROTO_DCCP);
3980
3981 if (show_users || show_proc_ctx || show_sock_ctx)
3982 user_ent_destroy();
3983
3984 return 0;
3985 }
3986