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 __u32 mark;
742 };
743
744 struct dctcpstat
745 {
746 unsigned int ce_state;
747 unsigned int alpha;
748 unsigned int ab_ecn;
749 unsigned int ab_tot;
750 bool enabled;
751 };
752
753 struct tcpstat
754 {
755 struct sockstat ss;
756 int timer;
757 int timeout;
758 int probes;
759 char cong_alg[16];
760 double rto, ato, rtt, rttvar;
761 int qack, cwnd, ssthresh, backoff;
762 double send_bps;
763 int snd_wscale;
764 int rcv_wscale;
765 int mss;
766 unsigned int lastsnd;
767 unsigned int lastrcv;
768 unsigned int lastack;
769 double pacing_rate;
770 double pacing_rate_max;
771 unsigned long long bytes_acked;
772 unsigned long long bytes_received;
773 unsigned int segs_out;
774 unsigned int segs_in;
775 unsigned int unacked;
776 unsigned int retrans;
777 unsigned int retrans_total;
778 unsigned int lost;
779 unsigned int sacked;
780 unsigned int fackets;
781 unsigned int reordering;
782 double rcv_rtt;
783 int rcv_space;
784 bool has_ts_opt;
785 bool has_sack_opt;
786 bool has_ecn_opt;
787 bool has_ecnseen_opt;
788 bool has_fastopen_opt;
789 bool has_wscale_opt;
790 struct dctcpstat *dctcp;
791 };
792
sock_state_print(struct sockstat * s,const char * sock_name)793 static void sock_state_print(struct sockstat *s, const char *sock_name)
794 {
795 if (netid_width)
796 printf("%-*s ", netid_width, sock_name);
797 if (state_width)
798 printf("%-*s ", state_width, sstate_name[s->state]);
799
800 printf("%-6d %-6d ", s->rq, s->wq);
801 }
802
sock_details_print(struct sockstat * s)803 static void sock_details_print(struct sockstat *s)
804 {
805 if (s->uid)
806 printf(" uid:%u", s->uid);
807
808 printf(" ino:%u", s->ino);
809 printf(" sk:%llx", s->sk);
810
811 if (s->mark)
812 printf(" fwmark:0x%x", s->mark);
813 }
814
sock_addr_print_width(int addr_len,const char * addr,char * delim,int port_len,const char * port,const char * ifname)815 static void sock_addr_print_width(int addr_len, const char *addr, char *delim,
816 int port_len, const char *port, const char *ifname)
817 {
818 if (ifname) {
819 printf("%*s%%%s%s%-*s ", addr_len, addr, ifname, delim,
820 port_len, port);
821 }
822 else {
823 printf("%*s%s%-*s ", addr_len, addr, delim, port_len, port);
824 }
825 }
826
sock_addr_print(const char * addr,char * delim,const char * port,const char * ifname)827 static void sock_addr_print(const char *addr, char *delim, const char *port,
828 const char *ifname)
829 {
830 sock_addr_print_width(addr_width, addr, delim, serv_width, port, ifname);
831 }
832
833 static const char *tmr_name[] = {
834 "off",
835 "on",
836 "keepalive",
837 "timewait",
838 "persist",
839 "unknown"
840 };
841
print_ms_timer(int timeout)842 static const char *print_ms_timer(int timeout)
843 {
844 static char buf[64];
845 int secs, msecs, minutes;
846 if (timeout < 0)
847 timeout = 0;
848 secs = timeout/1000;
849 minutes = secs/60;
850 secs = secs%60;
851 msecs = timeout%1000;
852 buf[0] = 0;
853 if (minutes) {
854 msecs = 0;
855 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
856 if (minutes > 9)
857 secs = 0;
858 }
859 if (secs) {
860 if (secs > 9)
861 msecs = 0;
862 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
863 }
864 if (msecs)
865 sprintf(buf+strlen(buf), "%03dms", msecs);
866 return buf;
867 }
868
869 struct scache {
870 struct scache *next;
871 int port;
872 char *name;
873 const char *proto;
874 };
875
876 struct scache *rlist;
877
init_service_resolver(void)878 static void init_service_resolver(void)
879 {
880 char buf[128];
881 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
882
883 if (!fp)
884 return;
885
886 if (!fgets(buf, sizeof(buf), fp)) {
887 pclose(fp);
888 return;
889 }
890 while (fgets(buf, sizeof(buf), fp) != NULL) {
891 unsigned int progn, port;
892 char proto[128], prog[128] = "rpc.";
893 struct scache *c;
894
895 if (sscanf(buf, "%u %*d %s %u %s",
896 &progn, proto, &port, prog+4) != 4)
897 continue;
898
899 if (!(c = malloc(sizeof(*c))))
900 continue;
901
902 c->port = port;
903 c->name = strdup(prog);
904 if (strcmp(proto, TCP_PROTO) == 0)
905 c->proto = TCP_PROTO;
906 else if (strcmp(proto, UDP_PROTO) == 0)
907 c->proto = UDP_PROTO;
908 else
909 c->proto = NULL;
910 c->next = rlist;
911 rlist = c;
912 }
913 pclose(fp);
914 }
915
916 /* Even do not try default linux ephemeral port ranges:
917 * default /etc/services contains so much of useless crap
918 * wouldbe "allocated" to this area that resolution
919 * is really harmful. I shrug each time when seeing
920 * "socks" or "cfinger" in dumps.
921 */
is_ephemeral(int port)922 static int is_ephemeral(int port)
923 {
924 static int min = 0, max = 0;
925
926 if (!min) {
927 FILE *f = ephemeral_ports_open();
928 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
929 min = 1024;
930 max = 4999;
931 }
932 if (f)
933 fclose(f);
934 }
935 return port >= min && port <= max;
936 }
937
938
__resolve_service(int port)939 static const char *__resolve_service(int port)
940 {
941 struct scache *c;
942
943 for (c = rlist; c; c = c->next) {
944 if (c->port == port && c->proto == dg_proto)
945 return c->name;
946 }
947
948 if (!is_ephemeral(port)) {
949 static int notfirst;
950 struct servent *se;
951 if (!notfirst) {
952 setservent(1);
953 notfirst = 1;
954 }
955 se = getservbyport(htons(port), dg_proto);
956 if (se)
957 return se->s_name;
958 }
959
960 return NULL;
961 }
962
963 #define SCACHE_BUCKETS 1024
964 static struct scache *cache_htab[SCACHE_BUCKETS];
965
resolve_service(int port)966 static const char *resolve_service(int port)
967 {
968 static char buf[128];
969 struct scache *c;
970 const char *res;
971 int hash;
972
973 if (port == 0) {
974 buf[0] = '*';
975 buf[1] = 0;
976 return buf;
977 }
978
979 if (!resolve_services)
980 goto do_numeric;
981
982 if (dg_proto == RAW_PROTO)
983 return inet_proto_n2a(port, buf, sizeof(buf));
984
985
986 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
987
988 for (c = cache_htab[hash]; c; c = c->next) {
989 if (c->port == port && c->proto == dg_proto)
990 goto do_cache;
991 }
992
993 c = malloc(sizeof(*c));
994 if (!c)
995 goto do_numeric;
996 res = __resolve_service(port);
997 c->port = port;
998 c->name = res ? strdup(res) : NULL;
999 c->proto = dg_proto;
1000 c->next = cache_htab[hash];
1001 cache_htab[hash] = c;
1002
1003 do_cache:
1004 if (c->name)
1005 return c->name;
1006
1007 do_numeric:
1008 sprintf(buf, "%u", port);
1009 return buf;
1010 }
1011
inet_addr_print(const inet_prefix * a,int port,unsigned int ifindex)1012 static void inet_addr_print(const inet_prefix *a, int port, unsigned int ifindex)
1013 {
1014 char buf[1024];
1015 const char *ap = buf;
1016 int est_len = addr_width;
1017 const char *ifname = NULL;
1018
1019 if (a->family == AF_INET) {
1020 if (a->data[0] == 0) {
1021 buf[0] = '*';
1022 buf[1] = 0;
1023 } else {
1024 ap = format_host(AF_INET, 4, a->data, buf, sizeof(buf));
1025 }
1026 } else {
1027 ap = format_host(a->family, 16, a->data, buf, sizeof(buf));
1028 est_len = strlen(ap);
1029 if (est_len <= addr_width)
1030 est_len = addr_width;
1031 else
1032 est_len = addr_width + ((est_len-addr_width+3)/4)*4;
1033 }
1034
1035 if (ifindex) {
1036 ifname = ll_index_to_name(ifindex);
1037 est_len -= strlen(ifname) + 1; /* +1 for percent char */
1038 if (est_len < 0)
1039 est_len = 0;
1040 }
1041
1042 sock_addr_print_width(est_len, ap, ":", serv_width, resolve_service(port),
1043 ifname);
1044 }
1045
1046 struct aafilter
1047 {
1048 inet_prefix addr;
1049 int port;
1050 __u32 mark;
1051 __u32 mask;
1052 struct aafilter *next;
1053 };
1054
inet2_addr_match(const inet_prefix * a,const inet_prefix * p,int plen)1055 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1056 int plen)
1057 {
1058 if (!inet_addr_match(a, p, plen))
1059 return 0;
1060
1061 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1062 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1063 * sockets. Fair? */
1064 if (p->family == AF_INET && a->family == AF_INET6) {
1065 if (a->data[0] == 0 && a->data[1] == 0 &&
1066 a->data[2] == htonl(0xffff)) {
1067 inet_prefix tmp = *a;
1068 tmp.data[0] = a->data[3];
1069 return inet_addr_match(&tmp, p, plen);
1070 }
1071 }
1072 return 1;
1073 }
1074
unix_match(const inet_prefix * a,const inet_prefix * p)1075 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1076 {
1077 char *addr, *pattern;
1078 memcpy(&addr, a->data, sizeof(addr));
1079 memcpy(&pattern, p->data, sizeof(pattern));
1080 if (pattern == NULL)
1081 return 1;
1082 if (addr == NULL)
1083 addr = "";
1084 return !fnmatch(pattern, addr, 0);
1085 }
1086
run_ssfilter(struct ssfilter * f,struct sockstat * s)1087 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1088 {
1089 switch (f->type) {
1090 case SSF_S_AUTO:
1091 {
1092 if (s->local.family == AF_UNIX) {
1093 char *p;
1094 memcpy(&p, s->local.data, sizeof(p));
1095 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1096 strspn(p+1, "0123456789abcdef") == 5);
1097 }
1098 if (s->local.family == AF_PACKET)
1099 return s->lport == 0 && s->local.data[0] == 0;
1100 if (s->local.family == AF_NETLINK)
1101 return s->lport < 0;
1102
1103 return is_ephemeral(s->lport);
1104 }
1105 case SSF_DCOND:
1106 {
1107 struct aafilter *a = (void*)f->pred;
1108 if (a->addr.family == AF_UNIX)
1109 return unix_match(&s->remote, &a->addr);
1110 if (a->port != -1 && a->port != s->rport)
1111 return 0;
1112 if (a->addr.bitlen) {
1113 do {
1114 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1115 return 1;
1116 } while ((a = a->next) != NULL);
1117 return 0;
1118 }
1119 return 1;
1120 }
1121 case SSF_SCOND:
1122 {
1123 struct aafilter *a = (void*)f->pred;
1124 if (a->addr.family == AF_UNIX)
1125 return unix_match(&s->local, &a->addr);
1126 if (a->port != -1 && a->port != s->lport)
1127 return 0;
1128 if (a->addr.bitlen) {
1129 do {
1130 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1131 return 1;
1132 } while ((a = a->next) != NULL);
1133 return 0;
1134 }
1135 return 1;
1136 }
1137 case SSF_D_GE:
1138 {
1139 struct aafilter *a = (void*)f->pred;
1140 return s->rport >= a->port;
1141 }
1142 case SSF_D_LE:
1143 {
1144 struct aafilter *a = (void*)f->pred;
1145 return s->rport <= a->port;
1146 }
1147 case SSF_S_GE:
1148 {
1149 struct aafilter *a = (void*)f->pred;
1150 return s->lport >= a->port;
1151 }
1152 case SSF_S_LE:
1153 {
1154 struct aafilter *a = (void*)f->pred;
1155 return s->lport <= a->port;
1156 }
1157 case SSF_MARKMASK:
1158 {
1159 struct aafilter *a = (void *)f->pred;
1160
1161 return (s->mark & a->mask) == a->mark;
1162 }
1163 /* Yup. It is recursion. Sorry. */
1164 case SSF_AND:
1165 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1166 case SSF_OR:
1167 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1168 case SSF_NOT:
1169 return !run_ssfilter(f->pred, s);
1170 default:
1171 abort();
1172 }
1173 }
1174
1175 /* Relocate external jumps by reloc. */
ssfilter_patch(char * a,int len,int reloc)1176 static void ssfilter_patch(char *a, int len, int reloc)
1177 {
1178 while (len > 0) {
1179 struct inet_diag_bc_op *op = (struct inet_diag_bc_op*)a;
1180 if (op->no == len+4)
1181 op->no += reloc;
1182 len -= op->yes;
1183 a += op->yes;
1184 }
1185 if (len < 0)
1186 abort();
1187 }
1188
ssfilter_bytecompile(struct ssfilter * f,char ** bytecode)1189 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1190 {
1191 switch (f->type) {
1192 case SSF_S_AUTO:
1193 {
1194 if (!(*bytecode=malloc(4))) abort();
1195 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1196 return 4;
1197 }
1198 case SSF_DCOND:
1199 case SSF_SCOND:
1200 {
1201 struct aafilter *a = (void*)f->pred;
1202 struct aafilter *b;
1203 char *ptr;
1204 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1205 int len = 0;
1206
1207 for (b=a; b; b=b->next) {
1208 len += 4 + sizeof(struct inet_diag_hostcond);
1209 if (a->addr.family == AF_INET6)
1210 len += 16;
1211 else
1212 len += 4;
1213 if (b->next)
1214 len += 4;
1215 }
1216 if (!(ptr = malloc(len))) abort();
1217 *bytecode = ptr;
1218 for (b=a; b; b=b->next) {
1219 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1220 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1221 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1222 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond*)(ptr+4);
1223
1224 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1225 cond->family = a->addr.family;
1226 cond->port = a->port;
1227 cond->prefix_len = a->addr.bitlen;
1228 memcpy(cond->addr, a->addr.data, alen);
1229 ptr += oplen;
1230 if (b->next) {
1231 op = (struct inet_diag_bc_op *)ptr;
1232 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1233 ptr += 4;
1234 }
1235 }
1236 return ptr - *bytecode;
1237 }
1238 case SSF_D_GE:
1239 {
1240 struct aafilter *x = (void*)f->pred;
1241 if (!(*bytecode=malloc(8))) abort();
1242 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_GE, 8, 12 };
1243 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1244 return 8;
1245 }
1246 case SSF_D_LE:
1247 {
1248 struct aafilter *x = (void*)f->pred;
1249 if (!(*bytecode=malloc(8))) abort();
1250 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_LE, 8, 12 };
1251 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1252 return 8;
1253 }
1254 case SSF_S_GE:
1255 {
1256 struct aafilter *x = (void*)f->pred;
1257 if (!(*bytecode=malloc(8))) abort();
1258 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1259 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1260 return 8;
1261 }
1262 case SSF_S_LE:
1263 {
1264 struct aafilter *x = (void*)f->pred;
1265 if (!(*bytecode=malloc(8))) abort();
1266 ((struct inet_diag_bc_op*)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1267 ((struct inet_diag_bc_op*)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1268 return 8;
1269 }
1270
1271 case SSF_AND:
1272 {
1273 char *a1, *a2, *a;
1274 int l1, l2;
1275 l1 = ssfilter_bytecompile(f->pred, &a1);
1276 l2 = ssfilter_bytecompile(f->post, &a2);
1277 if (!(a = malloc(l1+l2))) abort();
1278 memcpy(a, a1, l1);
1279 memcpy(a+l1, a2, l2);
1280 free(a1); free(a2);
1281 ssfilter_patch(a, l1, l2);
1282 *bytecode = a;
1283 return l1+l2;
1284 }
1285 case SSF_OR:
1286 {
1287 char *a1, *a2, *a;
1288 int l1, l2;
1289 l1 = ssfilter_bytecompile(f->pred, &a1);
1290 l2 = ssfilter_bytecompile(f->post, &a2);
1291 if (!(a = malloc(l1+l2+4))) abort();
1292 memcpy(a, a1, l1);
1293 memcpy(a+l1+4, a2, l2);
1294 free(a1); free(a2);
1295 *(struct inet_diag_bc_op*)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1296 *bytecode = a;
1297 return l1+l2+4;
1298 }
1299 case SSF_NOT:
1300 {
1301 char *a1, *a;
1302 int l1;
1303 l1 = ssfilter_bytecompile(f->pred, &a1);
1304 if (!(a = malloc(l1+4))) abort();
1305 memcpy(a, a1, l1);
1306 free(a1);
1307 *(struct inet_diag_bc_op*)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1308 *bytecode = a;
1309 return l1+4;
1310 }
1311 case SSF_MARKMASK:
1312 {
1313 struct aafilter *a = (void *)f->pred;
1314 struct instr {
1315 struct inet_diag_bc_op op;
1316 struct inet_diag_markcond cond;
1317 };
1318 int inslen = sizeof(struct instr);
1319
1320 if (!(*bytecode = malloc(inslen))) abort();
1321 ((struct instr *)*bytecode)[0] = (struct instr) {
1322 { INET_DIAG_BC_MARK_COND, inslen, inslen + 4 },
1323 { a->mark, a->mask},
1324 };
1325
1326 return inslen;
1327 }
1328 default:
1329 abort();
1330 }
1331 }
1332
remember_he(struct aafilter * a,struct hostent * he)1333 static int remember_he(struct aafilter *a, struct hostent *he)
1334 {
1335 char **ptr = he->h_addr_list;
1336 int cnt = 0;
1337 int len;
1338
1339 if (he->h_addrtype == AF_INET)
1340 len = 4;
1341 else if (he->h_addrtype == AF_INET6)
1342 len = 16;
1343 else
1344 return 0;
1345
1346 while (*ptr) {
1347 struct aafilter *b = a;
1348 if (a->addr.bitlen) {
1349 if ((b = malloc(sizeof(*b))) == NULL)
1350 return cnt;
1351 *b = *a;
1352 b->next = a->next;
1353 a->next = b;
1354 }
1355 memcpy(b->addr.data, *ptr, len);
1356 b->addr.bytelen = len;
1357 b->addr.bitlen = len*8;
1358 b->addr.family = he->h_addrtype;
1359 ptr++;
1360 cnt++;
1361 }
1362 return cnt;
1363 }
1364
get_dns_host(struct aafilter * a,const char * addr,int fam)1365 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
1366 {
1367 static int notfirst;
1368 int cnt = 0;
1369 struct hostent *he;
1370
1371 a->addr.bitlen = 0;
1372 if (!notfirst) {
1373 sethostent(1);
1374 notfirst = 1;
1375 }
1376 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
1377 if (he)
1378 cnt = remember_he(a, he);
1379 if (fam == AF_UNSPEC) {
1380 he = gethostbyname2(addr, AF_INET6);
1381 if (he)
1382 cnt += remember_he(a, he);
1383 }
1384 return !cnt;
1385 }
1386
1387 static int xll_initted = 0;
1388
xll_init(void)1389 static void xll_init(void)
1390 {
1391 struct rtnl_handle rth;
1392 if (rtnl_open(&rth, 0) < 0)
1393 exit(1);
1394
1395 ll_init_map(&rth);
1396 rtnl_close(&rth);
1397 xll_initted = 1;
1398 }
1399
xll_index_to_name(int index)1400 static const char *xll_index_to_name(int index)
1401 {
1402 if (!xll_initted)
1403 xll_init();
1404 return ll_index_to_name(index);
1405 }
1406
xll_name_to_index(const char * dev)1407 static int xll_name_to_index(const char *dev)
1408 {
1409 if (!xll_initted)
1410 xll_init();
1411 return ll_name_to_index(dev);
1412 }
1413
parse_hostcond(char * addr,bool is_port)1414 void *parse_hostcond(char *addr, bool is_port)
1415 {
1416 char *port = NULL;
1417 struct aafilter a = { .port = -1 };
1418 struct aafilter *res;
1419 int fam = preferred_family;
1420 struct filter *f = ¤t_filter;
1421
1422 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
1423 char *p;
1424 a.addr.family = AF_UNIX;
1425 if (strncmp(addr, "unix:", 5) == 0)
1426 addr+=5;
1427 p = strdup(addr);
1428 a.addr.bitlen = 8*strlen(p);
1429 memcpy(a.addr.data, &p, sizeof(p));
1430 fam = AF_UNIX;
1431 goto out;
1432 }
1433
1434 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
1435 a.addr.family = AF_PACKET;
1436 a.addr.bitlen = 0;
1437 if (strncmp(addr, "link:", 5) == 0)
1438 addr+=5;
1439 port = strchr(addr, ':');
1440 if (port) {
1441 *port = 0;
1442 if (port[1] && strcmp(port+1, "*")) {
1443 if (get_integer(&a.port, port+1, 0)) {
1444 if ((a.port = xll_name_to_index(port+1)) <= 0)
1445 return NULL;
1446 }
1447 }
1448 }
1449 if (addr[0] && strcmp(addr, "*")) {
1450 unsigned short tmp;
1451 a.addr.bitlen = 32;
1452 if (ll_proto_a2n(&tmp, addr))
1453 return NULL;
1454 a.addr.data[0] = ntohs(tmp);
1455 }
1456 fam = AF_PACKET;
1457 goto out;
1458 }
1459
1460 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
1461 a.addr.family = AF_NETLINK;
1462 a.addr.bitlen = 0;
1463 if (strncmp(addr, "netlink:", 8) == 0)
1464 addr+=8;
1465 port = strchr(addr, ':');
1466 if (port) {
1467 *port = 0;
1468 if (port[1] && strcmp(port+1, "*")) {
1469 if (get_integer(&a.port, port+1, 0)) {
1470 if (strcmp(port+1, "kernel") == 0)
1471 a.port = 0;
1472 else
1473 return NULL;
1474 }
1475 }
1476 }
1477 if (addr[0] && strcmp(addr, "*")) {
1478 a.addr.bitlen = 32;
1479 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
1480 return NULL;
1481 }
1482 fam = AF_NETLINK;
1483 goto out;
1484 }
1485
1486 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
1487 fam = AF_INET;
1488 if (!strncmp(addr, "inet:", 5))
1489 addr += 5;
1490 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
1491 fam = AF_INET6;
1492 if (!strncmp(addr, "inet6:", 6))
1493 addr += 6;
1494 }
1495
1496 /* URL-like literal [] */
1497 if (addr[0] == '[') {
1498 addr++;
1499 if ((port = strchr(addr, ']')) == NULL)
1500 return NULL;
1501 *port++ = 0;
1502 } else if (addr[0] == '*') {
1503 port = addr+1;
1504 } else {
1505 port = strrchr(strchr(addr, '/') ? : addr, ':');
1506 }
1507
1508 if (is_port)
1509 port = addr;
1510
1511 if (port && *port) {
1512 if (*port == ':')
1513 *port++ = 0;
1514
1515 if (*port && *port != '*') {
1516 if (get_integer(&a.port, port, 0)) {
1517 struct servent *se1 = NULL;
1518 struct servent *se2 = NULL;
1519 if (current_filter.dbs&(1<<UDP_DB))
1520 se1 = getservbyname(port, UDP_PROTO);
1521 if (current_filter.dbs&(1<<TCP_DB))
1522 se2 = getservbyname(port, TCP_PROTO);
1523 if (se1 && se2 && se1->s_port != se2->s_port) {
1524 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
1525 return NULL;
1526 }
1527 if (!se1)
1528 se1 = se2;
1529 if (se1) {
1530 a.port = ntohs(se1->s_port);
1531 } else {
1532 struct scache *s;
1533 for (s = rlist; s; s = s->next) {
1534 if ((s->proto == UDP_PROTO &&
1535 (current_filter.dbs&(1<<UDP_DB))) ||
1536 (s->proto == TCP_PROTO &&
1537 (current_filter.dbs&(1<<TCP_DB)))) {
1538 if (s->name && strcmp(s->name, port) == 0) {
1539 if (a.port > 0 && a.port != s->port) {
1540 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
1541 return NULL;
1542 }
1543 a.port = s->port;
1544 }
1545 }
1546 }
1547 if (a.port <= 0) {
1548 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
1549 return NULL;
1550 }
1551 }
1552 }
1553 }
1554 }
1555 if (!is_port && addr && *addr && *addr != '*') {
1556 if (get_prefix_1(&a.addr, addr, fam)) {
1557 if (get_dns_host(&a, addr, fam)) {
1558 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
1559 return NULL;
1560 }
1561 }
1562 }
1563
1564 out:
1565 if (fam != AF_UNSPEC) {
1566 f->families = 0;
1567 filter_af_set(f, fam);
1568 filter_states_set(f, 0);
1569 }
1570
1571 res = malloc(sizeof(*res));
1572 if (res)
1573 memcpy(res, &a, sizeof(a));
1574 return res;
1575 }
1576
parse_markmask(const char * markmask)1577 void *parse_markmask(const char *markmask)
1578 {
1579 struct aafilter a, *res;
1580
1581 if (strchr(markmask, '/')) {
1582 if (sscanf(markmask, "%i/%i", &a.mark, &a.mask) != 2)
1583 return NULL;
1584 } else {
1585 a.mask = 0xffffffff;
1586 if (sscanf(markmask, "%i", &a.mark) != 1)
1587 return NULL;
1588 }
1589
1590 res = malloc(sizeof(*res));
1591 if (res)
1592 memcpy(res, &a, sizeof(a));
1593 return res;
1594 }
1595
proto_name(int protocol)1596 static char *proto_name(int protocol)
1597 {
1598 switch (protocol) {
1599 case 0:
1600 return "raw";
1601 case IPPROTO_UDP:
1602 return "udp";
1603 case IPPROTO_TCP:
1604 return "tcp";
1605 case IPPROTO_DCCP:
1606 return "dccp";
1607 }
1608
1609 return "???";
1610 }
1611
inet_stats_print(struct sockstat * s,int protocol)1612 static void inet_stats_print(struct sockstat *s, int protocol)
1613 {
1614 char *buf = NULL;
1615
1616 sock_state_print(s, proto_name(protocol));
1617
1618 inet_addr_print(&s->local, s->lport, s->iface);
1619 inet_addr_print(&s->remote, s->rport, 0);
1620
1621 if (show_proc_ctx || show_sock_ctx) {
1622 if (find_entry(s->ino, &buf,
1623 (show_proc_ctx & show_sock_ctx) ?
1624 PROC_SOCK_CTX : PROC_CTX) > 0) {
1625 printf(" users:(%s)", buf);
1626 free(buf);
1627 }
1628 } else if (show_users) {
1629 if (find_entry(s->ino, &buf, USERS) > 0) {
1630 printf(" users:(%s)", buf);
1631 free(buf);
1632 }
1633 }
1634 }
1635
proc_parse_inet_addr(char * loc,char * rem,int family,struct sockstat * s)1636 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
1637 sockstat *s)
1638 {
1639 s->local.family = s->remote.family = family;
1640 if (family == AF_INET) {
1641 sscanf(loc, "%x:%x", s->local.data, (unsigned*)&s->lport);
1642 sscanf(rem, "%x:%x", s->remote.data, (unsigned*)&s->rport);
1643 s->local.bytelen = s->remote.bytelen = 4;
1644 return 0;
1645 } else {
1646 sscanf(loc, "%08x%08x%08x%08x:%x",
1647 s->local.data,
1648 s->local.data + 1,
1649 s->local.data + 2,
1650 s->local.data + 3,
1651 &s->lport);
1652 sscanf(rem, "%08x%08x%08x%08x:%x",
1653 s->remote.data,
1654 s->remote.data + 1,
1655 s->remote.data + 2,
1656 s->remote.data + 3,
1657 &s->rport);
1658 s->local.bytelen = s->remote.bytelen = 16;
1659 return 0;
1660 }
1661 return -1;
1662 }
1663
proc_inet_split_line(char * line,char ** loc,char ** rem,char ** data)1664 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
1665 {
1666 char *p;
1667
1668 if ((p = strchr(line, ':')) == NULL)
1669 return -1;
1670
1671 *loc = p+2;
1672 if ((p = strchr(*loc, ':')) == NULL)
1673 return -1;
1674
1675 p[5] = 0;
1676 *rem = p+6;
1677 if ((p = strchr(*rem, ':')) == NULL)
1678 return -1;
1679
1680 p[5] = 0;
1681 *data = p+6;
1682 return 0;
1683 }
1684
sprint_bw(char * buf,double bw)1685 static char *sprint_bw(char *buf, double bw)
1686 {
1687 if (bw > 1000000.)
1688 sprintf(buf,"%.1fM", bw / 1000000.);
1689 else if (bw > 1000.)
1690 sprintf(buf,"%.1fK", bw / 1000.);
1691 else
1692 sprintf(buf, "%g", bw);
1693
1694 return buf;
1695 }
1696
tcp_stats_print(struct tcpstat * s)1697 static void tcp_stats_print(struct tcpstat *s)
1698 {
1699 char b1[64];
1700
1701 if (s->has_ts_opt)
1702 printf(" ts");
1703 if (s->has_sack_opt)
1704 printf(" sack");
1705 if (s->has_ecn_opt)
1706 printf(" ecn");
1707 if (s->has_ecnseen_opt)
1708 printf(" ecnseen");
1709 if (s->has_fastopen_opt)
1710 printf(" fastopen");
1711 if (s->cong_alg[0])
1712 printf(" %s", s->cong_alg);
1713 if (s->has_wscale_opt)
1714 printf(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
1715 if (s->rto)
1716 printf(" rto:%g", s->rto);
1717 if (s->backoff)
1718 printf(" backoff:%u", s->backoff);
1719 if (s->rtt)
1720 printf(" rtt:%g/%g", s->rtt, s->rttvar);
1721 if (s->ato)
1722 printf(" ato:%g", s->ato);
1723
1724 if (s->qack)
1725 printf(" qack:%d", s->qack);
1726 if (s->qack & 1)
1727 printf(" bidir");
1728
1729 if (s->mss)
1730 printf(" mss:%d", s->mss);
1731 if (s->cwnd)
1732 printf(" cwnd:%d", s->cwnd);
1733 if (s->ssthresh)
1734 printf(" ssthresh:%d", s->ssthresh);
1735
1736 if (s->bytes_acked)
1737 printf(" bytes_acked:%llu", s->bytes_acked);
1738 if (s->bytes_received)
1739 printf(" bytes_received:%llu", s->bytes_received);
1740 if (s->segs_out)
1741 printf(" segs_out:%u", s->segs_out);
1742 if (s->segs_in)
1743 printf(" segs_in:%u", s->segs_in);
1744
1745 if (s->dctcp && s->dctcp->enabled) {
1746 struct dctcpstat *dctcp = s->dctcp;
1747
1748 printf(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
1749 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
1750 dctcp->ab_tot);
1751 } else if (s->dctcp) {
1752 printf(" dctcp:fallback_mode");
1753 }
1754
1755 if (s->send_bps)
1756 printf(" send %sbps", sprint_bw(b1, s->send_bps));
1757 if (s->lastsnd)
1758 printf(" lastsnd:%u", s->lastsnd);
1759 if (s->lastrcv)
1760 printf(" lastrcv:%u", s->lastrcv);
1761 if (s->lastack)
1762 printf(" lastack:%u", s->lastack);
1763
1764 if (s->pacing_rate) {
1765 printf(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
1766 if (s->pacing_rate_max)
1767 printf("/%sbps", sprint_bw(b1,
1768 s->pacing_rate_max));
1769 }
1770
1771 if (s->unacked)
1772 printf(" unacked:%u", s->unacked);
1773 if (s->retrans || s->retrans_total)
1774 printf(" retrans:%u/%u", s->retrans, s->retrans_total);
1775 if (s->lost)
1776 printf(" lost:%u", s->lost);
1777 if (s->sacked && s->ss.state != SS_LISTEN)
1778 printf(" sacked:%u", s->sacked);
1779 if (s->fackets)
1780 printf(" fackets:%u", s->fackets);
1781 if (s->reordering != 3)
1782 printf(" reordering:%d", s->reordering);
1783 if (s->rcv_rtt)
1784 printf(" rcv_rtt:%g", s->rcv_rtt);
1785 if (s->rcv_space)
1786 printf(" rcv_space:%d", s->rcv_space);
1787 }
1788
tcp_timer_print(struct tcpstat * s)1789 static void tcp_timer_print(struct tcpstat *s)
1790 {
1791 if (s->timer) {
1792 if (s->timer > 4)
1793 s->timer = 5;
1794 printf(" timer:(%s,%s,%d)",
1795 tmr_name[s->timer],
1796 print_ms_timer(s->timeout),
1797 s->retrans);
1798 }
1799 }
1800
tcp_show_line(char * line,const struct filter * f,int family)1801 static int tcp_show_line(char *line, const struct filter *f, int family)
1802 {
1803 int rto = 0, ato = 0;
1804 struct tcpstat s = {};
1805 char *loc, *rem, *data;
1806 char opt[256];
1807 int n;
1808 int hz = get_user_hz();
1809
1810 if (proc_inet_split_line(line, &loc, &rem, &data))
1811 return -1;
1812
1813 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
1814 if (!(f->states & (1 << state)))
1815 return 0;
1816
1817 proc_parse_inet_addr(loc, rem, family, &s.ss);
1818
1819 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
1820 return 0;
1821
1822 opt[0] = 0;
1823 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %d %d %[^\n]\n",
1824 &s.ss.state, &s.ss.wq, &s.ss.rq,
1825 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
1826 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
1827 &s.ssthresh, opt);
1828
1829 if (n < 17)
1830 opt[0] = 0;
1831
1832 if (n < 12) {
1833 rto = 0;
1834 s.cwnd = 2;
1835 s.ssthresh = -1;
1836 ato = s.qack = 0;
1837 }
1838
1839 s.retrans = s.timer != 1 ? s.probes : s.retrans;
1840 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
1841 s.ato = (double)ato / hz;
1842 s.qack /= 2;
1843 s.rto = (double)rto;
1844 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
1845 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
1846
1847 inet_stats_print(&s.ss, IPPROTO_TCP);
1848
1849 if (show_options)
1850 tcp_timer_print(&s);
1851
1852 if (show_details) {
1853 sock_details_print(&s.ss);
1854 if (opt[0])
1855 printf(" opt:\"%s\"", opt);
1856 }
1857
1858 if (show_tcpinfo)
1859 tcp_stats_print(&s);
1860
1861 printf("\n");
1862 return 0;
1863 }
1864
generic_record_read(FILE * fp,int (* worker)(char *,const struct filter *,int),const struct filter * f,int fam)1865 static int generic_record_read(FILE *fp,
1866 int (*worker)(char*, const struct filter *, int),
1867 const struct filter *f, int fam)
1868 {
1869 char line[256];
1870
1871 /* skip header */
1872 if (fgets(line, sizeof(line), fp) == NULL)
1873 goto outerr;
1874
1875 while (fgets(line, sizeof(line), fp) != NULL) {
1876 int n = strlen(line);
1877 if (n == 0 || line[n-1] != '\n') {
1878 errno = -EINVAL;
1879 return -1;
1880 }
1881 line[n-1] = 0;
1882
1883 if (worker(line, f, fam) < 0)
1884 return 0;
1885 }
1886 outerr:
1887
1888 return ferror(fp) ? -1 : 0;
1889 }
1890
print_skmeminfo(struct rtattr * tb[],int attrtype)1891 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
1892 {
1893 const __u32 *skmeminfo;
1894
1895 if (!tb[attrtype]) {
1896 if (attrtype == INET_DIAG_SKMEMINFO) {
1897 if (!tb[INET_DIAG_MEMINFO])
1898 return;
1899
1900 const struct inet_diag_meminfo *minfo =
1901 RTA_DATA(tb[INET_DIAG_MEMINFO]);
1902
1903 printf(" mem:(r%u,w%u,f%u,t%u)",
1904 minfo->idiag_rmem,
1905 minfo->idiag_wmem,
1906 minfo->idiag_fmem,
1907 minfo->idiag_tmem);
1908 }
1909 return;
1910 }
1911
1912 skmeminfo = RTA_DATA(tb[attrtype]);
1913
1914 printf(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
1915 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
1916 skmeminfo[SK_MEMINFO_RCVBUF],
1917 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
1918 skmeminfo[SK_MEMINFO_SNDBUF],
1919 skmeminfo[SK_MEMINFO_FWD_ALLOC],
1920 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
1921 skmeminfo[SK_MEMINFO_OPTMEM]);
1922
1923 if (RTA_PAYLOAD(tb[attrtype]) >=
1924 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
1925 printf(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
1926
1927 printf(")");
1928 }
1929
1930 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
1931
tcp_show_info(const struct nlmsghdr * nlh,struct inet_diag_msg * r,struct rtattr * tb[])1932 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
1933 struct rtattr *tb[])
1934 {
1935 double rtt = 0;
1936 struct tcpstat s = {};
1937
1938 s.ss.state = r->idiag_state;
1939
1940 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
1941
1942 if (tb[INET_DIAG_INFO]) {
1943 struct tcp_info *info;
1944 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
1945
1946 /* workaround for older kernels with less fields */
1947 if (len < sizeof(*info)) {
1948 info = alloca(sizeof(*info));
1949 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
1950 memset((char *)info + len, 0, sizeof(*info) - len);
1951 } else
1952 info = RTA_DATA(tb[INET_DIAG_INFO]);
1953
1954 if (show_options) {
1955 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
1956 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
1957 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
1958 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
1959 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
1960 }
1961
1962 if (tb[INET_DIAG_CONG])
1963 strncpy(s.cong_alg,
1964 rta_getattr_str(tb[INET_DIAG_CONG]),
1965 sizeof(s.cong_alg) - 1);
1966
1967 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
1968 s.has_wscale_opt = true;
1969 s.snd_wscale = info->tcpi_snd_wscale;
1970 s.rcv_wscale = info->tcpi_rcv_wscale;
1971 }
1972
1973 if (info->tcpi_rto && info->tcpi_rto != 3000000)
1974 s.rto = (double)info->tcpi_rto / 1000;
1975
1976 s.backoff = info->tcpi_backoff;
1977 s.rtt = (double)info->tcpi_rtt / 1000;
1978 s.rttvar = (double)info->tcpi_rttvar / 1000;
1979 s.ato = (double)info->tcpi_ato / 1000;
1980 s.mss = info->tcpi_snd_mss;
1981 s.rcv_space = info->tcpi_rcv_space;
1982 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
1983 s.lastsnd = info->tcpi_last_data_sent;
1984 s.lastrcv = info->tcpi_last_data_recv;
1985 s.lastack = info->tcpi_last_ack_recv;
1986 s.unacked = info->tcpi_unacked;
1987 s.retrans = info->tcpi_retrans;
1988 s.retrans_total = info->tcpi_total_retrans;
1989 s.lost = info->tcpi_lost;
1990 s.sacked = info->tcpi_sacked;
1991 s.reordering = info->tcpi_reordering;
1992 s.rcv_space = info->tcpi_rcv_space;
1993 s.cwnd = info->tcpi_snd_cwnd;
1994
1995 if (info->tcpi_snd_ssthresh < 0xFFFF)
1996 s.ssthresh = info->tcpi_snd_ssthresh;
1997
1998 rtt = (double) info->tcpi_rtt;
1999 if (tb[INET_DIAG_VEGASINFO]) {
2000 const struct tcpvegas_info *vinfo
2001 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
2002
2003 if (vinfo->tcpv_enabled &&
2004 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
2005 rtt = vinfo->tcpv_rtt;
2006 }
2007
2008 if (tb[INET_DIAG_DCTCPINFO]) {
2009 struct dctcpstat *dctcp = malloc(sizeof(struct
2010 dctcpstat));
2011
2012 const struct tcp_dctcp_info *dinfo
2013 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
2014
2015 dctcp->enabled = !!dinfo->dctcp_enabled;
2016 dctcp->ce_state = dinfo->dctcp_ce_state;
2017 dctcp->alpha = dinfo->dctcp_alpha;
2018 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
2019 dctcp->ab_tot = dinfo->dctcp_ab_tot;
2020 s.dctcp = dctcp;
2021 }
2022
2023 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
2024 s.send_bps = (double) info->tcpi_snd_cwnd *
2025 (double)info->tcpi_snd_mss * 8000000. / rtt;
2026 }
2027
2028 if (info->tcpi_pacing_rate &&
2029 info->tcpi_pacing_rate != ~0ULL) {
2030 s.pacing_rate = info->tcpi_pacing_rate * 8.;
2031
2032 if (info->tcpi_max_pacing_rate &&
2033 info->tcpi_max_pacing_rate != ~0ULL)
2034 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
2035 }
2036 s.bytes_acked = info->tcpi_bytes_acked;
2037 s.bytes_received = info->tcpi_bytes_received;
2038 s.segs_out = info->tcpi_segs_out;
2039 s.segs_in = info->tcpi_segs_in;
2040 tcp_stats_print(&s);
2041 free(s.dctcp);
2042 }
2043 }
2044
inet_show_sock(struct nlmsghdr * nlh,struct filter * f,int protocol)2045 static int inet_show_sock(struct nlmsghdr *nlh, struct filter *f, int protocol)
2046 {
2047 struct rtattr * tb[INET_DIAG_MAX+1];
2048 struct inet_diag_msg *r = NLMSG_DATA(nlh);
2049 struct sockstat s = {};
2050
2051 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr*)(r+1),
2052 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2053
2054 s.state = r->idiag_state;
2055 s.local.family = s.remote.family = r->idiag_family;
2056 s.lport = ntohs(r->id.idiag_sport);
2057 s.rport = ntohs(r->id.idiag_dport);
2058 s.wq = r->idiag_wqueue;
2059 s.rq = r->idiag_rqueue;
2060 s.ino = r->idiag_inode;
2061 s.uid = r->idiag_uid;
2062 s.iface = r->id.idiag_if;
2063 s.sk = cookie_sk_get(&r->id.idiag_cookie[0]);
2064
2065 s.mark = 0;
2066 if (tb[INET_DIAG_MARK])
2067 s.mark = *(__u32 *) RTA_DATA(tb[INET_DIAG_MARK]);
2068
2069 if (s.local.family == AF_INET) {
2070 s.local.bytelen = s.remote.bytelen = 4;
2071 } else {
2072 s.local.bytelen = s.remote.bytelen = 16;
2073 }
2074
2075 memcpy(s.local.data, r->id.idiag_src, s.local.bytelen);
2076 memcpy(s.remote.data, r->id.idiag_dst, s.local.bytelen);
2077
2078 if (f && f->f && run_ssfilter(f->f, &s) == 0)
2079 return 0;
2080
2081 if (tb[INET_DIAG_PROTOCOL])
2082 protocol = *(__u8 *)RTA_DATA(tb[INET_DIAG_PROTOCOL]);
2083
2084 inet_stats_print(&s, protocol);
2085
2086 if (show_options) {
2087 struct tcpstat t = {};
2088
2089 t.timer = r->idiag_timer;
2090 t.timeout = r->idiag_expires;
2091 t.retrans = r->idiag_retrans;
2092 tcp_timer_print(&t);
2093 }
2094
2095 if (show_details) {
2096 sock_details_print(&s);
2097 if (s.local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY]) {
2098 unsigned char v6only;
2099 v6only = *(__u8 *)RTA_DATA(tb[INET_DIAG_SKV6ONLY]);
2100 printf(" v6only:%u", v6only);
2101 }
2102 if (tb[INET_DIAG_SHUTDOWN]) {
2103 unsigned char mask;
2104 mask = *(__u8 *)RTA_DATA(tb[INET_DIAG_SHUTDOWN]);
2105 printf(" %c-%c", mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
2106 }
2107 }
2108
2109 if (show_mem || show_tcpinfo) {
2110 printf("\n\t");
2111 tcp_show_info(nlh, r, tb);
2112 }
2113
2114 printf("\n");
2115 return 0;
2116 }
2117
tcpdiag_send(int fd,int protocol,struct filter * f)2118 static int tcpdiag_send(int fd, int protocol, struct filter *f)
2119 {
2120 struct sockaddr_nl nladdr;
2121 struct {
2122 struct nlmsghdr nlh;
2123 struct inet_diag_req r;
2124 } req;
2125 char *bc = NULL;
2126 int bclen;
2127 struct msghdr msg;
2128 struct rtattr rta;
2129 struct iovec iov[3];
2130
2131 if (protocol == IPPROTO_UDP)
2132 return -1;
2133
2134 memset(&nladdr, 0, sizeof(nladdr));
2135 nladdr.nl_family = AF_NETLINK;
2136
2137 req.nlh.nlmsg_len = sizeof(req);
2138 if (protocol == IPPROTO_TCP)
2139 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
2140 else
2141 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
2142 req.nlh.nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST;
2143 req.nlh.nlmsg_pid = 0;
2144 req.nlh.nlmsg_seq = MAGIC_SEQ;
2145 memset(&req.r, 0, sizeof(req.r));
2146 req.r.idiag_family = AF_INET;
2147 req.r.idiag_states = f->states;
2148 if (show_mem) {
2149 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
2150 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
2151 }
2152
2153 if (show_tcpinfo) {
2154 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
2155 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
2156 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
2157 }
2158
2159 iov[0] = (struct iovec){
2160 .iov_base = &req,
2161 .iov_len = sizeof(req)
2162 };
2163 if (f->f) {
2164 bclen = ssfilter_bytecompile(f->f, &bc);
2165 rta.rta_type = INET_DIAG_REQ_BYTECODE;
2166 rta.rta_len = RTA_LENGTH(bclen);
2167 iov[1] = (struct iovec){ &rta, sizeof(rta) };
2168 iov[2] = (struct iovec){ bc, bclen };
2169 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
2170 }
2171
2172 msg = (struct msghdr) {
2173 .msg_name = (void*)&nladdr,
2174 .msg_namelen = sizeof(nladdr),
2175 .msg_iov = iov,
2176 .msg_iovlen = f->f ? 3 : 1,
2177 };
2178
2179 if (sendmsg(fd, &msg, 0) < 0) {
2180 close(fd);
2181 return -1;
2182 }
2183
2184 return 0;
2185 }
2186
sockdiag_send(int family,int fd,int protocol,struct filter * f)2187 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
2188 {
2189 struct sockaddr_nl nladdr;
2190 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
2191 char *bc = NULL;
2192 int bclen;
2193 struct msghdr msg;
2194 struct rtattr rta;
2195 struct iovec iov[3];
2196
2197 if (family == PF_UNSPEC)
2198 return tcpdiag_send(fd, protocol, f);
2199
2200 memset(&nladdr, 0, sizeof(nladdr));
2201 nladdr.nl_family = AF_NETLINK;
2202
2203 memset(&req.r, 0, sizeof(req.r));
2204 req.r.sdiag_family = family;
2205 req.r.sdiag_protocol = protocol;
2206 req.r.idiag_states = f->states;
2207 if (show_mem) {
2208 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
2209 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
2210 }
2211
2212 if (show_tcpinfo) {
2213 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
2214 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
2215 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
2216 }
2217
2218 iov[0] = (struct iovec){
2219 .iov_base = &req,
2220 .iov_len = sizeof(req)
2221 };
2222 if (f->f) {
2223 bclen = ssfilter_bytecompile(f->f, &bc);
2224 rta.rta_type = INET_DIAG_REQ_BYTECODE;
2225 rta.rta_len = RTA_LENGTH(bclen);
2226 iov[1] = (struct iovec){ &rta, sizeof(rta) };
2227 iov[2] = (struct iovec){ bc, bclen };
2228 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
2229 }
2230
2231 msg = (struct msghdr) {
2232 .msg_name = (void*)&nladdr,
2233 .msg_namelen = sizeof(nladdr),
2234 .msg_iov = iov,
2235 .msg_iovlen = f->f ? 3 : 1,
2236 };
2237
2238 if (sendmsg(fd, &msg, 0) < 0) {
2239 close(fd);
2240 return -1;
2241 }
2242
2243 return 0;
2244 }
2245
2246 struct inet_diag_arg {
2247 struct filter *f;
2248 int protocol;
2249 struct rtnl_handle *rth;
2250 };
2251
kill_inet_sock(const struct sockaddr_nl * addr,struct nlmsghdr * h,void * arg)2252 static int kill_inet_sock(const struct sockaddr_nl *addr,
2253 struct nlmsghdr *h, void *arg)
2254 {
2255 struct inet_diag_msg *d = NLMSG_DATA(h);
2256 struct inet_diag_arg *diag_arg = arg;
2257 struct rtnl_handle *rth = diag_arg->rth;
2258 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
2259
2260 req.nlh.nlmsg_type = SOCK_DESTROY;
2261 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
2262 req.nlh.nlmsg_seq = ++rth->seq;
2263 req.r.sdiag_family = d->idiag_family;
2264 req.r.sdiag_protocol = diag_arg->protocol;
2265 req.r.id = d->id;
2266
2267 return rtnl_talk(rth, &req.nlh, NULL, 0);
2268 }
2269
show_one_inet_sock(const struct sockaddr_nl * addr,struct nlmsghdr * h,void * arg)2270 static int show_one_inet_sock(const struct sockaddr_nl *addr,
2271 struct nlmsghdr *h, void *arg)
2272 {
2273 int err;
2274 struct inet_diag_arg *diag_arg = arg;
2275 struct inet_diag_msg *r = NLMSG_DATA(h);
2276
2277 if (!(diag_arg->f->families & (1 << r->idiag_family)))
2278 return 0;
2279 if (diag_arg->f->kill && kill_inet_sock(addr, h, arg) != 0) {
2280 if (errno == EOPNOTSUPP || errno == ENOENT) {
2281 /* Socket can't be closed, or is already closed. */
2282 return 0;
2283 } else {
2284 perror("SOCK_DESTROY answers");
2285 return -1;
2286 }
2287 }
2288 if ((err = inet_show_sock(h, diag_arg->f, diag_arg->protocol)) < 0)
2289 return err;
2290
2291 return 0;
2292 }
2293
inet_show_netlink(struct filter * f,FILE * dump_fp,int protocol)2294 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
2295 {
2296 int err = 0;
2297 struct rtnl_handle rth, rth2;
2298 int family = PF_INET;
2299 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
2300
2301 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
2302 return -1;
2303
2304 if (f->kill) {
2305 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
2306 rtnl_close(&rth);
2307 return -1;
2308 }
2309 arg.rth = &rth2;
2310 }
2311
2312 rth.dump = MAGIC_SEQ;
2313 rth.dump_fp = dump_fp;
2314 if (preferred_family == PF_INET6)
2315 family = PF_INET6;
2316
2317 again:
2318 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
2319 goto Exit;
2320
2321 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
2322 if (family != PF_UNSPEC) {
2323 family = PF_UNSPEC;
2324 goto again;
2325 }
2326 goto Exit;
2327 }
2328 if (family == PF_INET && preferred_family != PF_INET) {
2329 family = PF_INET6;
2330 goto again;
2331 }
2332
2333 Exit:
2334 rtnl_close(&rth);
2335 if (arg.rth)
2336 rtnl_close(arg.rth);
2337 return err;
2338 }
2339
tcp_show_netlink_file(struct filter * f)2340 static int tcp_show_netlink_file(struct filter *f)
2341 {
2342 FILE *fp;
2343 char buf[16384];
2344
2345 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
2346 perror("fopen($TCPDIAG_FILE)");
2347 return -1;
2348 }
2349
2350 while (1) {
2351 int status, err;
2352 struct nlmsghdr *h = (struct nlmsghdr*)buf;
2353
2354 status = fread(buf, 1, sizeof(*h), fp);
2355 if (status < 0) {
2356 perror("Reading header from $TCPDIAG_FILE");
2357 return -1;
2358 }
2359 if (status != sizeof(*h)) {
2360 perror("Unexpected EOF reading $TCPDIAG_FILE");
2361 return -1;
2362 }
2363
2364 status = fread(h+1, 1, NLMSG_ALIGN(h->nlmsg_len-sizeof(*h)), fp);
2365
2366 if (status < 0) {
2367 perror("Reading $TCPDIAG_FILE");
2368 return -1;
2369 }
2370 if (status + sizeof(*h) < h->nlmsg_len) {
2371 perror("Unexpected EOF reading $TCPDIAG_FILE");
2372 return -1;
2373 }
2374
2375 /* The only legal exit point */
2376 if (h->nlmsg_type == NLMSG_DONE)
2377 return 0;
2378
2379 if (h->nlmsg_type == NLMSG_ERROR) {
2380 struct nlmsgerr *err = (struct nlmsgerr*)NLMSG_DATA(h);
2381 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
2382 fprintf(stderr, "ERROR truncated\n");
2383 } else {
2384 errno = -err->error;
2385 perror("TCPDIAG answered");
2386 }
2387 return -1;
2388 }
2389
2390 err = inet_show_sock(h, f, IPPROTO_TCP);
2391 if (err < 0)
2392 return err;
2393 }
2394 }
2395
tcp_show(struct filter * f,int socktype)2396 static int tcp_show(struct filter *f, int socktype)
2397 {
2398 FILE *fp = NULL;
2399 char *buf = NULL;
2400 int bufsize = 64*1024;
2401
2402 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2403 return 0;
2404
2405 dg_proto = TCP_PROTO;
2406
2407 if (getenv("TCPDIAG_FILE"))
2408 return tcp_show_netlink_file(f);
2409
2410 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
2411 && inet_show_netlink(f, NULL, socktype) == 0)
2412 return 0;
2413
2414 /* Sigh... We have to parse /proc/net/tcp... */
2415
2416
2417 /* Estimate amount of sockets and try to allocate
2418 * huge buffer to read all the table at one read.
2419 * Limit it by 16MB though. The assumption is: as soon as
2420 * kernel was able to hold information about N connections,
2421 * it is able to give us some memory for snapshot.
2422 */
2423 if (1) {
2424 get_slabstat(&slabstat);
2425
2426 int guess = slabstat.socks+slabstat.tcp_syns;
2427 if (f->states&(1<<SS_TIME_WAIT))
2428 guess += slabstat.tcp_tws;
2429 if (guess > (16*1024*1024)/128)
2430 guess = (16*1024*1024)/128;
2431 guess *= 128;
2432 if (guess > bufsize)
2433 bufsize = guess;
2434 }
2435 while (bufsize >= 64*1024) {
2436 if ((buf = malloc(bufsize)) != NULL)
2437 break;
2438 bufsize /= 2;
2439 }
2440 if (buf == NULL) {
2441 errno = ENOMEM;
2442 return -1;
2443 }
2444
2445 if (f->families & (1<<AF_INET)) {
2446 if ((fp = net_tcp_open()) == NULL)
2447 goto outerr;
2448
2449 setbuffer(fp, buf, bufsize);
2450 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
2451 goto outerr;
2452 fclose(fp);
2453 }
2454
2455 if ((f->families & (1<<AF_INET6)) &&
2456 (fp = net_tcp6_open()) != NULL) {
2457 setbuffer(fp, buf, bufsize);
2458 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
2459 goto outerr;
2460 fclose(fp);
2461 }
2462
2463 free(buf);
2464 return 0;
2465
2466 outerr:
2467 do {
2468 int saved_errno = errno;
2469 free(buf);
2470 if (fp)
2471 fclose(fp);
2472 errno = saved_errno;
2473 return -1;
2474 } while (0);
2475 }
2476
2477
dgram_show_line(char * line,const struct filter * f,int family)2478 static int dgram_show_line(char *line, const struct filter *f, int family)
2479 {
2480 struct sockstat s = {};
2481 char *loc, *rem, *data;
2482 char opt[256];
2483 int n;
2484
2485 if (proc_inet_split_line(line, &loc, &rem, &data))
2486 return -1;
2487
2488 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2489 if (!(f->states & (1 << state)))
2490 return 0;
2491
2492 proc_parse_inet_addr(loc, rem, family, &s);
2493
2494 if (f->f && run_ssfilter(f->f, &s) == 0)
2495 return 0;
2496
2497 opt[0] = 0;
2498 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
2499 &s.state, &s.wq, &s.rq,
2500 &s.uid, &s.ino,
2501 &s.refcnt, &s.sk, opt);
2502
2503 if (n < 9)
2504 opt[0] = 0;
2505
2506 inet_stats_print(&s, dg_proto == UDP_PROTO ? IPPROTO_UDP : 0);
2507
2508 if (show_details && opt[0])
2509 printf(" opt:\"%s\"", opt);
2510
2511 printf("\n");
2512 return 0;
2513 }
2514
udp_show(struct filter * f)2515 static int udp_show(struct filter *f)
2516 {
2517 FILE *fp = NULL;
2518
2519 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2520 return 0;
2521
2522 dg_proto = UDP_PROTO;
2523
2524 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
2525 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
2526 return 0;
2527
2528 if (f->families&(1<<AF_INET)) {
2529 if ((fp = net_udp_open()) == NULL)
2530 goto outerr;
2531 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
2532 goto outerr;
2533 fclose(fp);
2534 }
2535
2536 if ((f->families&(1<<AF_INET6)) &&
2537 (fp = net_udp6_open()) != NULL) {
2538 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
2539 goto outerr;
2540 fclose(fp);
2541 }
2542 return 0;
2543
2544 outerr:
2545 do {
2546 int saved_errno = errno;
2547 if (fp)
2548 fclose(fp);
2549 errno = saved_errno;
2550 return -1;
2551 } while (0);
2552 }
2553
raw_show(struct filter * f)2554 static int raw_show(struct filter *f)
2555 {
2556 FILE *fp = NULL;
2557
2558 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2559 return 0;
2560
2561 dg_proto = RAW_PROTO;
2562
2563 if (f->families&(1<<AF_INET)) {
2564 if ((fp = net_raw_open()) == NULL)
2565 goto outerr;
2566 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
2567 goto outerr;
2568 fclose(fp);
2569 }
2570
2571 if ((f->families&(1<<AF_INET6)) &&
2572 (fp = net_raw6_open()) != NULL) {
2573 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
2574 goto outerr;
2575 fclose(fp);
2576 }
2577 return 0;
2578
2579 outerr:
2580 do {
2581 int saved_errno = errno;
2582 if (fp)
2583 fclose(fp);
2584 errno = saved_errno;
2585 return -1;
2586 } while (0);
2587 }
2588
2589 int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
2590 SS_ESTABLISHED, SS_CLOSING };
2591
2592 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
2593
unix_list_free(struct sockstat * list)2594 static void unix_list_free(struct sockstat *list)
2595 {
2596 while (list) {
2597 struct sockstat *s = list;
2598
2599 list = list->next;
2600 free(s->name);
2601 free(s);
2602 }
2603 }
2604
unix_netid_name(int type)2605 static const char *unix_netid_name(int type)
2606 {
2607 const char *netid;
2608
2609 switch (type) {
2610 case SOCK_STREAM:
2611 netid = "u_str";
2612 break;
2613 case SOCK_SEQPACKET:
2614 netid = "u_seq";
2615 break;
2616 case SOCK_DGRAM:
2617 default:
2618 netid = "u_dgr";
2619 break;
2620 }
2621 return netid;
2622 }
2623
unix_type_skip(struct sockstat * s,struct filter * f)2624 static bool unix_type_skip(struct sockstat *s, struct filter *f)
2625 {
2626 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
2627 return true;
2628 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
2629 return true;
2630 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
2631 return true;
2632 return false;
2633 }
2634
unix_use_proc(void)2635 static bool unix_use_proc(void)
2636 {
2637 return getenv("PROC_NET_UNIX") || getenv("PROC_ROOT");
2638 }
2639
unix_stats_print(struct sockstat * list,struct filter * f)2640 static void unix_stats_print(struct sockstat *list, struct filter *f)
2641 {
2642 struct sockstat *s;
2643 char *peer;
2644 char *ctx_buf = NULL;
2645 bool use_proc = unix_use_proc();
2646 char port_name[30] = {};
2647
2648 for (s = list; s; s = s->next) {
2649 if (!(f->states & (1 << s->state)))
2650 continue;
2651 if (unix_type_skip(s, f))
2652 continue;
2653
2654 peer = "*";
2655 if (s->peer_name)
2656 peer = s->peer_name;
2657
2658 if (s->rport && use_proc) {
2659 struct sockstat *p;
2660
2661 for (p = list; p; p = p->next) {
2662 if (s->rport == p->lport)
2663 break;
2664 }
2665
2666 if (!p) {
2667 peer = "?";
2668 } else {
2669 peer = p->name ? : "*";
2670 }
2671 }
2672
2673 if (use_proc && f->f) {
2674 struct sockstat st;
2675 st.local.family = AF_UNIX;
2676 st.remote.family = AF_UNIX;
2677 memcpy(st.local.data, &s->name, sizeof(s->name));
2678 if (strcmp(peer, "*") == 0)
2679 memset(st.remote.data, 0, sizeof(peer));
2680 else
2681 memcpy(st.remote.data, &peer, sizeof(peer));
2682 if (run_ssfilter(f->f, &st) == 0)
2683 continue;
2684 }
2685
2686 sock_state_print(s, unix_netid_name(s->type));
2687
2688 sock_addr_print(s->name ?: "*", " ",
2689 int_to_str(s->lport, port_name), NULL);
2690 sock_addr_print(peer, " ", int_to_str(s->rport, port_name),
2691 NULL);
2692
2693 if (show_proc_ctx || show_sock_ctx) {
2694 if (find_entry(s->ino, &ctx_buf,
2695 (show_proc_ctx & show_sock_ctx) ?
2696 PROC_SOCK_CTX : PROC_CTX) > 0) {
2697 printf(" users:(%s)", ctx_buf);
2698 free(ctx_buf);
2699 }
2700 } else if (show_users) {
2701 if (find_entry(s->ino, &ctx_buf, USERS) > 0) {
2702 printf(" users:(%s)", ctx_buf);
2703 free(ctx_buf);
2704 }
2705 }
2706 printf("\n");
2707 }
2708 }
2709
unix_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)2710 static int unix_show_sock(const struct sockaddr_nl *addr, struct nlmsghdr *nlh,
2711 void *arg)
2712 {
2713 struct filter *f = (struct filter *)arg;
2714 struct unix_diag_msg *r = NLMSG_DATA(nlh);
2715 struct rtattr *tb[UNIX_DIAG_MAX+1];
2716 char name[128];
2717 struct sockstat stat = { .name = "*", .peer_name = "*" };
2718
2719 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr*)(r+1),
2720 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2721
2722 stat.type = r->udiag_type;
2723 stat.state = r->udiag_state;
2724 stat.ino = stat.lport = r->udiag_ino;
2725 stat.local.family = stat.remote.family = AF_UNIX;
2726
2727 if (unix_type_skip(&stat, f))
2728 return 0;
2729
2730 if (tb[UNIX_DIAG_RQLEN]) {
2731 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
2732 stat.rq = rql->udiag_rqueue;
2733 stat.wq = rql->udiag_wqueue;
2734 }
2735 if (tb[UNIX_DIAG_NAME]) {
2736 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
2737
2738 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
2739 name[len] = '\0';
2740 if (name[0] == '\0')
2741 name[0] = '@';
2742 stat.name = &name[0];
2743 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
2744 }
2745 if (tb[UNIX_DIAG_PEER])
2746 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
2747
2748 if (f->f && run_ssfilter(f->f, &stat) == 0)
2749 return 0;
2750
2751 unix_stats_print(&stat, f);
2752
2753 if (show_mem) {
2754 printf("\t");
2755 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
2756 }
2757 if (show_details) {
2758 if (tb[UNIX_DIAG_SHUTDOWN]) {
2759 unsigned char mask;
2760 mask = *(__u8 *)RTA_DATA(tb[UNIX_DIAG_SHUTDOWN]);
2761 printf(" %c-%c", mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
2762 }
2763 }
2764 if (show_mem || show_details)
2765 printf("\n");
2766
2767 return 0;
2768 }
2769
handle_netlink_request(struct filter * f,struct nlmsghdr * req,size_t size,rtnl_filter_t show_one_sock)2770 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
2771 size_t size, rtnl_filter_t show_one_sock)
2772 {
2773 int ret = -1;
2774 struct rtnl_handle rth;
2775
2776 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
2777 return -1;
2778
2779 rth.dump = MAGIC_SEQ;
2780
2781 if (rtnl_send(&rth, req, size) < 0)
2782 goto Exit;
2783
2784 if (rtnl_dump_filter(&rth, show_one_sock, f))
2785 goto Exit;
2786
2787 ret = 0;
2788 Exit:
2789 rtnl_close(&rth);
2790 return ret;
2791 }
2792
unix_show_netlink(struct filter * f)2793 static int unix_show_netlink(struct filter *f)
2794 {
2795 DIAG_REQUEST(req, struct unix_diag_req r);
2796
2797 req.r.sdiag_family = AF_UNIX;
2798 req.r.udiag_states = f->states;
2799 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
2800 if (show_mem)
2801 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
2802
2803 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
2804 }
2805
unix_show(struct filter * f)2806 static int unix_show(struct filter *f)
2807 {
2808 FILE *fp;
2809 char buf[256];
2810 char name[128];
2811 int newformat = 0;
2812 int cnt;
2813 struct sockstat *list = NULL;
2814
2815 if (!filter_af_get(f, AF_UNIX))
2816 return 0;
2817
2818 if (!unix_use_proc() && unix_show_netlink(f) == 0)
2819 return 0;
2820
2821 if ((fp = net_unix_open()) == NULL)
2822 return -1;
2823 if (!fgets(buf, sizeof(buf), fp)) {
2824 fclose(fp);
2825 return -1;
2826 }
2827
2828 if (memcmp(buf, "Peer", 4) == 0)
2829 newformat = 1;
2830 cnt = 0;
2831
2832 while (fgets(buf, sizeof(buf), fp)) {
2833 struct sockstat *u, **insp;
2834 int flags;
2835
2836 if (!(u = calloc(1, sizeof(*u))))
2837 break;
2838 u->name = NULL;
2839 u->peer_name = NULL;
2840
2841 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
2842 &u->rport, &u->rq, &u->wq, &flags, &u->type,
2843 &u->state, &u->ino, name) < 8)
2844 name[0] = 0;
2845
2846 u->lport = u->ino;
2847 u->local.family = u->remote.family = AF_UNIX;
2848
2849 if (flags & (1 << 16)) {
2850 u->state = SS_LISTEN;
2851 } else {
2852 u->state = unix_state_map[u->state-1];
2853 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
2854 u->state = SS_ESTABLISHED;
2855 }
2856
2857 if (!newformat) {
2858 u->rport = 0;
2859 u->rq = 0;
2860 u->wq = 0;
2861 }
2862
2863 insp = &list;
2864 while (*insp) {
2865 if (u->type < (*insp)->type ||
2866 (u->type == (*insp)->type &&
2867 u->ino < (*insp)->ino))
2868 break;
2869 insp = &(*insp)->next;
2870 }
2871 u->next = *insp;
2872 *insp = u;
2873
2874 if (name[0]) {
2875 if ((u->name = malloc(strlen(name)+1)) == NULL)
2876 break;
2877 strcpy(u->name, name);
2878 }
2879 if (++cnt > MAX_UNIX_REMEMBER) {
2880 unix_stats_print(list, f);
2881 unix_list_free(list);
2882 list = NULL;
2883 cnt = 0;
2884 }
2885 }
2886 fclose(fp);
2887 if (list) {
2888 unix_stats_print(list, f);
2889 unix_list_free(list);
2890 list = NULL;
2891 cnt = 0;
2892 }
2893
2894 return 0;
2895 }
2896
packet_stats_print(struct sockstat * s,const struct filter * f)2897 static int packet_stats_print(struct sockstat *s, const struct filter *f)
2898 {
2899 char *buf = NULL;
2900 const char *addr, *port;
2901 char ll_name[16];
2902
2903 if (f->f) {
2904 s->local.family = AF_PACKET;
2905 s->remote.family = AF_PACKET;
2906 s->local.data[0] = s->prot;
2907 if (run_ssfilter(f->f, s) == 0)
2908 return 1;
2909 }
2910
2911 sock_state_print(s, s->type == SOCK_RAW ? "p_raw" : "p_dgr");
2912
2913 if (s->prot == 3)
2914 addr = "*";
2915 else
2916 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
2917
2918 if (s->iface == 0)
2919 port = "*";
2920 else
2921 port = xll_index_to_name(s->iface);
2922
2923 sock_addr_print(addr, ":", port, NULL);
2924 sock_addr_print("", "*", "", NULL);
2925
2926 if (show_proc_ctx || show_sock_ctx) {
2927 if (find_entry(s->ino, &buf,
2928 (show_proc_ctx & show_sock_ctx) ?
2929 PROC_SOCK_CTX : PROC_CTX) > 0) {
2930 printf(" users:(%s)", buf);
2931 free(buf);
2932 }
2933 } else if (show_users) {
2934 if (find_entry(s->ino, &buf, USERS) > 0) {
2935 printf(" users:(%s)", buf);
2936 free(buf);
2937 }
2938 }
2939
2940 if (show_details)
2941 sock_details_print(s);
2942
2943 return 0;
2944 }
2945
packet_show_ring(struct packet_diag_ring * ring)2946 static void packet_show_ring(struct packet_diag_ring *ring)
2947 {
2948 printf("blk_size:%d", ring->pdr_block_size);
2949 printf(",blk_nr:%d", ring->pdr_block_nr);
2950 printf(",frm_size:%d", ring->pdr_frame_size);
2951 printf(",frm_nr:%d", ring->pdr_frame_nr);
2952 printf(",tmo:%d", ring->pdr_retire_tmo);
2953 printf(",features:0x%x", ring->pdr_features);
2954 }
2955
packet_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)2956 static int packet_show_sock(const struct sockaddr_nl *addr,
2957 struct nlmsghdr *nlh, void *arg)
2958 {
2959 const struct filter *f = arg;
2960 struct packet_diag_msg *r = NLMSG_DATA(nlh);
2961 struct packet_diag_info *pinfo = NULL;
2962 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
2963 struct rtattr *tb[PACKET_DIAG_MAX+1];
2964 struct sockstat stat = {};
2965 uint32_t fanout = 0;
2966 bool has_fanout = false;
2967
2968 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr*)(r+1),
2969 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2970
2971 /* use /proc/net/packet if all info are not available */
2972 if (!tb[PACKET_DIAG_MEMINFO])
2973 return -1;
2974
2975 stat.type = r->pdiag_type;
2976 stat.prot = r->pdiag_num;
2977 stat.ino = r->pdiag_ino;
2978 stat.state = SS_CLOSE;
2979 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
2980
2981 if (tb[PACKET_DIAG_MEMINFO]) {
2982 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
2983 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
2984 }
2985
2986 if (tb[PACKET_DIAG_INFO]) {
2987 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
2988 stat.lport = stat.iface = pinfo->pdi_index;
2989 }
2990
2991 if (tb[PACKET_DIAG_UID])
2992 stat.uid = *(__u32 *)RTA_DATA(tb[PACKET_DIAG_UID]);
2993
2994 if (tb[PACKET_DIAG_RX_RING])
2995 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
2996
2997 if (tb[PACKET_DIAG_TX_RING])
2998 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
2999
3000 if (tb[PACKET_DIAG_FANOUT]) {
3001 has_fanout = true;
3002 fanout = *(uint32_t *)RTA_DATA(tb[PACKET_DIAG_FANOUT]);
3003 }
3004
3005 if (packet_stats_print(&stat, f))
3006 return 0;
3007
3008 if (show_details) {
3009 if (pinfo) {
3010 printf("\n\tver:%d", pinfo->pdi_version);
3011 printf(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
3012 printf(" flags( ");
3013 if (pinfo->pdi_flags & PDI_RUNNING)
3014 printf("running");
3015 if (pinfo->pdi_flags & PDI_AUXDATA)
3016 printf(" auxdata");
3017 if (pinfo->pdi_flags & PDI_ORIGDEV)
3018 printf(" origdev");
3019 if (pinfo->pdi_flags & PDI_VNETHDR)
3020 printf(" vnethdr");
3021 if (pinfo->pdi_flags & PDI_LOSS)
3022 printf(" loss");
3023 if (!pinfo->pdi_flags)
3024 printf("0");
3025 printf(" )");
3026 }
3027 if (ring_rx) {
3028 printf("\n\tring_rx(");
3029 packet_show_ring(ring_rx);
3030 printf(")");
3031 }
3032 if (ring_tx) {
3033 printf("\n\tring_tx(");
3034 packet_show_ring(ring_tx);
3035 printf(")");
3036 }
3037 if (has_fanout) {
3038 uint16_t type = (fanout >> 16) & 0xffff;
3039
3040 printf("\n\tfanout(");
3041 printf("id:%d,", fanout & 0xffff);
3042 printf("type:");
3043
3044 if (type == 0)
3045 printf("hash");
3046 else if (type == 1)
3047 printf("lb");
3048 else if (type == 2)
3049 printf("cpu");
3050 else if (type == 3)
3051 printf("roll");
3052 else if (type == 4)
3053 printf("random");
3054 else if (type == 5)
3055 printf("qm");
3056 else
3057 printf("0x%x", type);
3058
3059 printf(")");
3060 }
3061 }
3062
3063 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
3064 struct sock_filter *fil =
3065 RTA_DATA(tb[PACKET_DIAG_FILTER]);
3066 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
3067 sizeof(struct sock_filter);
3068
3069 printf("\n\tbpf filter (%d): ", num);
3070 while (num) {
3071 printf(" 0x%02x %u %u %u,",
3072 fil->code, fil->jt, fil->jf, fil->k);
3073 num--;
3074 fil++;
3075 }
3076 }
3077 printf("\n");
3078 return 0;
3079 }
3080
packet_show_netlink(struct filter * f)3081 static int packet_show_netlink(struct filter *f)
3082 {
3083 DIAG_REQUEST(req, struct packet_diag_req r);
3084
3085 req.r.sdiag_family = AF_PACKET;
3086 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
3087 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
3088
3089 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
3090 }
3091
packet_show_line(char * buf,const struct filter * f,int fam)3092 static int packet_show_line(char *buf, const struct filter *f, int fam)
3093 {
3094 unsigned long long sk;
3095 struct sockstat stat = {};
3096 int type, prot, iface, state, rq, uid, ino;
3097
3098 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
3099 &sk,
3100 &type, &prot, &iface, &state,
3101 &rq, &uid, &ino);
3102
3103 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
3104 return 0;
3105 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
3106 return 0;
3107
3108 stat.type = type;
3109 stat.prot = prot;
3110 stat.lport = stat.iface = iface;
3111 stat.state = state;
3112 stat.rq = rq;
3113 stat.uid = uid;
3114 stat.ino = ino;
3115 stat.state = SS_CLOSE;
3116
3117 if (packet_stats_print(&stat, f))
3118 return 0;
3119
3120 printf("\n");
3121 return 0;
3122 }
3123
packet_show(struct filter * f)3124 static int packet_show(struct filter *f)
3125 {
3126 FILE *fp;
3127 int rc = 0;
3128
3129 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
3130 return 0;
3131
3132 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
3133 packet_show_netlink(f) == 0)
3134 return 0;
3135
3136 if ((fp = net_packet_open()) == NULL)
3137 return -1;
3138 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
3139 rc = -1;
3140
3141 fclose(fp);
3142 return rc;
3143 }
3144
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)3145 static int netlink_show_one(struct filter *f,
3146 int prot, int pid, unsigned groups,
3147 int state, int dst_pid, unsigned dst_group,
3148 int rq, int wq,
3149 unsigned long long sk, unsigned long long cb)
3150 {
3151 struct sockstat st;
3152 SPRINT_BUF(prot_buf) = {};
3153 const char *prot_name;
3154 char procname[64] = {};
3155
3156 st.state = SS_CLOSE;
3157 st.rq = rq;
3158 st.wq = wq;
3159
3160 if (f->f) {
3161 st.local.family = AF_NETLINK;
3162 st.remote.family = AF_NETLINK;
3163 st.rport = -1;
3164 st.lport = pid;
3165 st.local.data[0] = prot;
3166 if (run_ssfilter(f->f, &st) == 0)
3167 return 1;
3168 }
3169
3170 sock_state_print(&st, "nl");
3171
3172 if (resolve_services)
3173 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
3174 else
3175 prot_name = int_to_str(prot, prot_buf);
3176
3177 if (pid == -1) {
3178 procname[0] = '*';
3179 } else if (resolve_services) {
3180 int done = 0;
3181 if (!pid) {
3182 done = 1;
3183 strncpy(procname, "kernel", 6);
3184 } else if (pid > 0) {
3185 FILE *fp;
3186 snprintf(procname, sizeof(procname), "%s/%d/stat",
3187 getenv("PROC_ROOT") ? : "/proc", pid);
3188 if ((fp = fopen(procname, "r")) != NULL) {
3189 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
3190 snprintf(procname+strlen(procname),
3191 sizeof(procname)-strlen(procname),
3192 "/%d", pid);
3193 done = 1;
3194 }
3195 fclose(fp);
3196 }
3197 }
3198 if (!done)
3199 int_to_str(pid, procname);
3200 } else {
3201 int_to_str(pid, procname);
3202 }
3203
3204 sock_addr_print(prot_name, ":", procname, NULL);
3205
3206 if (state == NETLINK_CONNECTED) {
3207 char dst_group_buf[30];
3208 char dst_pid_buf[30];
3209 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
3210 int_to_str(dst_pid, dst_pid_buf), NULL);
3211 } else {
3212 sock_addr_print("", "*", "", NULL);
3213 }
3214
3215 char *pid_context = NULL;
3216 if (show_proc_ctx) {
3217 /* The pid value will either be:
3218 * 0 if destination kernel - show kernel initial context.
3219 * A valid process pid - use getpidcon.
3220 * A unique value allocated by the kernel or netlink user
3221 * to the process - show context as "not available".
3222 */
3223 if (!pid)
3224 security_get_initial_context("kernel", &pid_context);
3225 else if (pid > 0)
3226 getpidcon(pid, &pid_context);
3227
3228 if (pid_context != NULL) {
3229 printf("proc_ctx=%-*s ", serv_width, pid_context);
3230 free(pid_context);
3231 } else {
3232 printf("proc_ctx=%-*s ", serv_width, "unavailable");
3233 }
3234 }
3235
3236 if (show_details) {
3237 printf(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
3238 }
3239 printf("\n");
3240
3241 return 0;
3242 }
3243
netlink_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)3244 static int netlink_show_sock(const struct sockaddr_nl *addr,
3245 struct nlmsghdr *nlh, void *arg)
3246 {
3247 struct filter *f = (struct filter *)arg;
3248 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
3249 struct rtattr *tb[NETLINK_DIAG_MAX+1];
3250 int rq = 0, wq = 0;
3251 unsigned long groups = 0;
3252
3253 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr*)(r+1),
3254 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3255
3256 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
3257 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
3258
3259 if (tb[NETLINK_DIAG_MEMINFO]) {
3260 const __u32 *skmeminfo;
3261 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
3262
3263 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
3264 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
3265 }
3266
3267 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
3268 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
3269 rq, wq, 0, 0)) {
3270 return 0;
3271 }
3272
3273 if (show_mem) {
3274 printf("\t");
3275 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
3276 printf("\n");
3277 }
3278
3279 return 0;
3280 }
3281
netlink_show_netlink(struct filter * f)3282 static int netlink_show_netlink(struct filter *f)
3283 {
3284 DIAG_REQUEST(req, struct netlink_diag_req r);
3285
3286 req.r.sdiag_family = AF_NETLINK;
3287 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
3288 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
3289
3290 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
3291 }
3292
netlink_show(struct filter * f)3293 static int netlink_show(struct filter *f)
3294 {
3295 FILE *fp;
3296 char buf[256];
3297 int prot, pid;
3298 unsigned groups;
3299 int rq, wq, rc;
3300 unsigned long long sk, cb;
3301
3302 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
3303 return 0;
3304
3305 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
3306 netlink_show_netlink(f) == 0)
3307 return 0;
3308
3309 if ((fp = net_netlink_open()) == NULL)
3310 return -1;
3311 if (!fgets(buf, sizeof(buf), fp)) {
3312 fclose(fp);
3313 return -1;
3314 }
3315
3316 while (fgets(buf, sizeof(buf), fp)) {
3317 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
3318 &sk,
3319 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
3320
3321 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
3322 }
3323
3324 fclose(fp);
3325 return 0;
3326 }
3327
3328 struct sock_diag_msg {
3329 __u8 sdiag_family;
3330 };
3331
generic_show_sock(const struct sockaddr_nl * addr,struct nlmsghdr * nlh,void * arg)3332 static int generic_show_sock(const struct sockaddr_nl *addr,
3333 struct nlmsghdr *nlh, void *arg)
3334 {
3335 struct sock_diag_msg *r = NLMSG_DATA(nlh);
3336 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
3337
3338 switch (r->sdiag_family) {
3339 case AF_INET:
3340 case AF_INET6:
3341 return show_one_inet_sock(addr, nlh, &inet_arg);
3342 case AF_UNIX:
3343 return unix_show_sock(addr, nlh, arg);
3344 case AF_PACKET:
3345 return packet_show_sock(addr, nlh, arg);
3346 case AF_NETLINK:
3347 return netlink_show_sock(addr, nlh, arg);
3348 default:
3349 return -1;
3350 }
3351 }
3352
handle_follow_request(struct filter * f)3353 static int handle_follow_request(struct filter *f)
3354 {
3355 int ret = -1;
3356 int groups = 0;
3357 struct rtnl_handle rth;
3358
3359 if (f->families & (1 << AF_INET) && f->dbs & (1 << TCP_DB))
3360 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
3361 if (f->families & (1 << AF_INET) && f->dbs & (1 << UDP_DB))
3362 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
3363 if (f->families & (1 << AF_INET6) && f->dbs & (1 << TCP_DB))
3364 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
3365 if (f->families & (1 << AF_INET6) && f->dbs & (1 << UDP_DB))
3366 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
3367
3368 if (groups == 0)
3369 return -1;
3370
3371 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
3372 return -1;
3373
3374 rth.dump = 0;
3375 rth.local.nl_pid = 0;
3376
3377 if (rtnl_dump_filter(&rth, generic_show_sock, f))
3378 goto Exit;
3379
3380 ret = 0;
3381 Exit:
3382 rtnl_close(&rth);
3383 return ret;
3384 }
3385
3386 struct snmpstat
3387 {
3388 int tcp_estab;
3389 };
3390
get_snmp_int(char * proto,char * key,int * result)3391 static int get_snmp_int(char *proto, char *key, int *result)
3392 {
3393 char buf[1024];
3394 FILE *fp;
3395 int protolen = strlen(proto);
3396 int keylen = strlen(key);
3397
3398 *result = 0;
3399
3400 if ((fp = net_snmp_open()) == NULL)
3401 return -1;
3402
3403 while (fgets(buf, sizeof(buf), fp) != NULL) {
3404 char *p = buf;
3405 int pos = 0;
3406 if (memcmp(buf, proto, protolen))
3407 continue;
3408 while ((p = strchr(p, ' ')) != NULL) {
3409 pos++;
3410 p++;
3411 if (memcmp(p, key, keylen) == 0 &&
3412 (p[keylen] == ' ' || p[keylen] == '\n'))
3413 break;
3414 }
3415 if (fgets(buf, sizeof(buf), fp) == NULL)
3416 break;
3417 if (memcmp(buf, proto, protolen))
3418 break;
3419 p = buf;
3420 while ((p = strchr(p, ' ')) != NULL) {
3421 p++;
3422 if (--pos == 0) {
3423 sscanf(p, "%d", result);
3424 fclose(fp);
3425 return 0;
3426 }
3427 }
3428 }
3429
3430 fclose(fp);
3431 errno = ESRCH;
3432 return -1;
3433 }
3434
3435
3436 /* Get stats from sockstat */
3437
3438 struct ssummary
3439 {
3440 int socks;
3441 int tcp_mem;
3442 int tcp_total;
3443 int tcp_orphans;
3444 int tcp_tws;
3445 int tcp4_hashed;
3446 int udp4;
3447 int raw4;
3448 int frag4;
3449 int frag4_mem;
3450 int tcp6_hashed;
3451 int udp6;
3452 int raw6;
3453 int frag6;
3454 int frag6_mem;
3455 };
3456
get_sockstat_line(char * line,struct ssummary * s)3457 static void get_sockstat_line(char *line, struct ssummary *s)
3458 {
3459 char id[256], rem[256];
3460
3461 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
3462 return;
3463
3464 if (strcmp(id, "sockets:") == 0)
3465 sscanf(rem, "%*s%d", &s->socks);
3466 else if (strcmp(id, "UDP:") == 0)
3467 sscanf(rem, "%*s%d", &s->udp4);
3468 else if (strcmp(id, "UDP6:") == 0)
3469 sscanf(rem, "%*s%d", &s->udp6);
3470 else if (strcmp(id, "RAW:") == 0)
3471 sscanf(rem, "%*s%d", &s->raw4);
3472 else if (strcmp(id, "RAW6:") == 0)
3473 sscanf(rem, "%*s%d", &s->raw6);
3474 else if (strcmp(id, "TCP6:") == 0)
3475 sscanf(rem, "%*s%d", &s->tcp6_hashed);
3476 else if (strcmp(id, "FRAG:") == 0)
3477 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
3478 else if (strcmp(id, "FRAG6:") == 0)
3479 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
3480 else if (strcmp(id, "TCP:") == 0)
3481 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
3482 &s->tcp4_hashed,
3483 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
3484 }
3485
get_sockstat(struct ssummary * s)3486 static int get_sockstat(struct ssummary *s)
3487 {
3488 char buf[256];
3489 FILE *fp;
3490
3491 memset(s, 0, sizeof(*s));
3492
3493 if ((fp = net_sockstat_open()) == NULL)
3494 return -1;
3495 while(fgets(buf, sizeof(buf), fp) != NULL)
3496 get_sockstat_line(buf, s);
3497 fclose(fp);
3498
3499 if ((fp = net_sockstat6_open()) == NULL)
3500 return 0;
3501 while(fgets(buf, sizeof(buf), fp) != NULL)
3502 get_sockstat_line(buf, s);
3503 fclose(fp);
3504
3505 return 0;
3506 }
3507
print_summary(void)3508 static int print_summary(void)
3509 {
3510 struct ssummary s;
3511 struct snmpstat sn;
3512
3513 if (get_sockstat(&s) < 0)
3514 perror("ss: get_sockstat");
3515 if (get_snmp_int("Tcp:", "CurrEstab", &sn.tcp_estab) < 0)
3516 perror("ss: get_snmpstat");
3517
3518 get_slabstat(&slabstat);
3519
3520 printf("Total: %d (kernel %d)\n", s.socks, slabstat.socks);
3521
3522 printf("TCP: %d (estab %d, closed %d, orphaned %d, synrecv %d, timewait %d/%d), ports %d\n",
3523 s.tcp_total + slabstat.tcp_syns + s.tcp_tws,
3524 sn.tcp_estab,
3525 s.tcp_total - (s.tcp4_hashed+s.tcp6_hashed-s.tcp_tws),
3526 s.tcp_orphans,
3527 slabstat.tcp_syns,
3528 s.tcp_tws, slabstat.tcp_tws,
3529 slabstat.tcp_ports
3530 );
3531
3532 printf("\n");
3533 printf("Transport Total IP IPv6\n");
3534 printf("* %-9d %-9s %-9s\n", slabstat.socks, "-", "-");
3535 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
3536 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
3537 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
3538 printf("INET %-9d %-9d %-9d\n",
3539 s.raw4+s.udp4+s.tcp4_hashed+
3540 s.raw6+s.udp6+s.tcp6_hashed,
3541 s.raw4+s.udp4+s.tcp4_hashed,
3542 s.raw6+s.udp6+s.tcp6_hashed);
3543 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
3544
3545 printf("\n");
3546
3547 return 0;
3548 }
3549
_usage(FILE * dest)3550 static void _usage(FILE *dest)
3551 {
3552 fprintf(dest,
3553 "Usage: ss [ OPTIONS ]\n"
3554 " ss [ OPTIONS ] [ FILTER ]\n"
3555 " -h, --help this message\n"
3556 " -V, --version output version information\n"
3557 " -n, --numeric don't resolve service names\n"
3558 " -r, --resolve resolve host names\n"
3559 " -a, --all display all sockets\n"
3560 " -l, --listening display listening sockets\n"
3561 " -o, --options show timer information\n"
3562 " -e, --extended show detailed socket information\n"
3563 " -m, --memory show socket memory usage\n"
3564 " -p, --processes show process using socket\n"
3565 " -i, --info show internal TCP information\n"
3566 " -s, --summary show socket usage summary\n"
3567 " -b, --bpf show bpf filter socket information\n"
3568 " -E, --events continually display sockets as they are destroyed\n"
3569 " -Z, --context display process SELinux security contexts\n"
3570 " -z, --contexts display process and socket SELinux security contexts\n"
3571 " -N, --net switch to the specified network namespace name\n"
3572 "\n"
3573 " -4, --ipv4 display only IP version 4 sockets\n"
3574 " -6, --ipv6 display only IP version 6 sockets\n"
3575 " -0, --packet display PACKET sockets\n"
3576 " -t, --tcp display only TCP sockets\n"
3577 " -u, --udp display only UDP sockets\n"
3578 " -d, --dccp display only DCCP sockets\n"
3579 " -w, --raw display only RAW sockets\n"
3580 " -x, --unix display only Unix domain sockets\n"
3581 " -f, --family=FAMILY display sockets of type FAMILY\n"
3582 "\n"
3583 " -K, --kill forcibly close sockets, display what was closed\n"
3584 "\n"
3585 " -A, --query=QUERY, --socket=QUERY\n"
3586 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink}[,QUERY]\n"
3587 "\n"
3588 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
3589 " -F, --filter=FILE read filter information from FILE\n"
3590 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
3591 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
3592 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listen|closing}\n"
3593 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
3594 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
3595 " bucket := {syn-recv|time-wait}\n"
3596 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listen|closing}\n"
3597 );
3598 }
3599
3600 static void help(void) __attribute__((noreturn));
help(void)3601 static void help(void)
3602 {
3603 _usage(stdout);
3604 exit(0);
3605 }
3606
3607 static void usage(void) __attribute__((noreturn));
usage(void)3608 static void usage(void)
3609 {
3610 _usage(stderr);
3611 exit(-1);
3612 }
3613
3614
scan_state(const char * state)3615 static int scan_state(const char *state)
3616 {
3617 int i;
3618 if (strcasecmp(state, "close") == 0 ||
3619 strcasecmp(state, "closed") == 0)
3620 return (1<<SS_CLOSE);
3621 if (strcasecmp(state, "syn-rcv") == 0)
3622 return (1<<SS_SYN_RECV);
3623 if (strcasecmp(state, "established") == 0)
3624 return (1<<SS_ESTABLISHED);
3625 if (strcasecmp(state, "all") == 0)
3626 return SS_ALL;
3627 if (strcasecmp(state, "connected") == 0)
3628 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
3629 if (strcasecmp(state, "synchronized") == 0)
3630 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
3631 if (strcasecmp(state, "bucket") == 0)
3632 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
3633 if (strcasecmp(state, "big") == 0)
3634 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
3635 for (i=0; i<SS_MAX; i++) {
3636 if (strcasecmp(state, sstate_namel[i]) == 0)
3637 return (1<<i);
3638 }
3639
3640 fprintf(stderr, "ss: wrong state name: %s\n", state);
3641 exit(-1);
3642 }
3643
3644 static const struct option long_opts[] = {
3645 { "numeric", 0, 0, 'n' },
3646 { "resolve", 0, 0, 'r' },
3647 { "options", 0, 0, 'o' },
3648 { "extended", 0, 0, 'e' },
3649 { "memory", 0, 0, 'm' },
3650 { "info", 0, 0, 'i' },
3651 { "processes", 0, 0, 'p' },
3652 { "bpf", 0, 0, 'b' },
3653 { "events", 0, 0, 'E' },
3654 { "dccp", 0, 0, 'd' },
3655 { "tcp", 0, 0, 't' },
3656 { "udp", 0, 0, 'u' },
3657 { "raw", 0, 0, 'w' },
3658 { "unix", 0, 0, 'x' },
3659 { "all", 0, 0, 'a' },
3660 { "listening", 0, 0, 'l' },
3661 { "ipv4", 0, 0, '4' },
3662 { "ipv6", 0, 0, '6' },
3663 { "packet", 0, 0, '0' },
3664 { "family", 1, 0, 'f' },
3665 { "socket", 1, 0, 'A' },
3666 { "query", 1, 0, 'A' },
3667 { "summary", 0, 0, 's' },
3668 { "diag", 1, 0, 'D' },
3669 { "filter", 1, 0, 'F' },
3670 { "version", 0, 0, 'V' },
3671 { "help", 0, 0, 'h' },
3672 { "context", 0, 0, 'Z' },
3673 { "contexts", 0, 0, 'z' },
3674 { "net", 1, 0, 'N' },
3675 { "kill", 0, 0, 'K' },
3676 { 0 }
3677
3678 };
3679
main(int argc,char * argv[])3680 int main(int argc, char *argv[])
3681 {
3682 int saw_states = 0;
3683 int saw_query = 0;
3684 int do_summary = 0;
3685 const char *dump_tcpdiag = NULL;
3686 FILE *filter_fp = NULL;
3687 int ch;
3688 int state_filter = 0;
3689
3690 while ((ch = getopt_long(argc, argv, "dhaletuwxnro460spbEf:miA:D:F:vVzZN:K",
3691 long_opts, NULL)) != EOF) {
3692 switch(ch) {
3693 case 'n':
3694 resolve_services = 0;
3695 break;
3696 case 'r':
3697 resolve_hosts = 1;
3698 break;
3699 case 'o':
3700 show_options = 1;
3701 break;
3702 case 'e':
3703 show_options = 1;
3704 show_details++;
3705 break;
3706 case 'm':
3707 show_mem = 1;
3708 break;
3709 case 'i':
3710 show_tcpinfo = 1;
3711 break;
3712 case 'p':
3713 show_users++;
3714 user_ent_hash_build();
3715 break;
3716 case 'b':
3717 show_options = 1;
3718 show_bpf++;
3719 break;
3720 case 'E':
3721 follow_events = 1;
3722 break;
3723 case 'd':
3724 filter_db_set(¤t_filter, DCCP_DB);
3725 break;
3726 case 't':
3727 filter_db_set(¤t_filter, TCP_DB);
3728 break;
3729 case 'u':
3730 filter_db_set(¤t_filter, UDP_DB);
3731 break;
3732 case 'w':
3733 filter_db_set(¤t_filter, RAW_DB);
3734 break;
3735 case 'x':
3736 filter_af_set(¤t_filter, AF_UNIX);
3737 break;
3738 case 'a':
3739 state_filter = SS_ALL;
3740 break;
3741 case 'l':
3742 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
3743 break;
3744 case '4':
3745 filter_af_set(¤t_filter, AF_INET);
3746 break;
3747 case '6':
3748 filter_af_set(¤t_filter, AF_INET6);
3749 break;
3750 case '0':
3751 filter_af_set(¤t_filter, AF_PACKET);
3752 break;
3753 case 'f':
3754 if (strcmp(optarg, "inet") == 0)
3755 filter_af_set(¤t_filter, AF_INET);
3756 else if (strcmp(optarg, "inet6") == 0)
3757 filter_af_set(¤t_filter, AF_INET6);
3758 else if (strcmp(optarg, "link") == 0)
3759 filter_af_set(¤t_filter, AF_PACKET);
3760 else if (strcmp(optarg, "unix") == 0)
3761 filter_af_set(¤t_filter, AF_UNIX);
3762 else if (strcmp(optarg, "netlink") == 0)
3763 filter_af_set(¤t_filter, AF_NETLINK);
3764 else if (strcmp(optarg, "help") == 0)
3765 help();
3766 else {
3767 fprintf(stderr, "ss: \"%s\" is invalid family\n",
3768 optarg);
3769 usage();
3770 }
3771 break;
3772 case 'A':
3773 {
3774 char *p, *p1;
3775 if (!saw_query) {
3776 current_filter.dbs = 0;
3777 state_filter = state_filter ?
3778 state_filter : SS_CONN;
3779 saw_query = 1;
3780 do_default = 0;
3781 }
3782 p = p1 = optarg;
3783 do {
3784 if ((p1 = strchr(p, ',')) != NULL)
3785 *p1 = 0;
3786 if (strcmp(p, "all") == 0) {
3787 filter_default_dbs(¤t_filter);
3788 } else if (strcmp(p, "inet") == 0) {
3789 filter_db_set(¤t_filter, UDP_DB);
3790 filter_db_set(¤t_filter, DCCP_DB);
3791 filter_db_set(¤t_filter, TCP_DB);
3792 filter_db_set(¤t_filter, RAW_DB);
3793 } else if (strcmp(p, "udp") == 0) {
3794 filter_db_set(¤t_filter, UDP_DB);
3795 } else if (strcmp(p, "dccp") == 0) {
3796 filter_db_set(¤t_filter, DCCP_DB);
3797 } else if (strcmp(p, "tcp") == 0) {
3798 filter_db_set(¤t_filter, TCP_DB);
3799 } else if (strcmp(p, "raw") == 0) {
3800 filter_db_set(¤t_filter, RAW_DB);
3801 } else if (strcmp(p, "unix") == 0) {
3802 filter_db_set(¤t_filter, UNIX_ST_DB);
3803 filter_db_set(¤t_filter, UNIX_DG_DB);
3804 filter_db_set(¤t_filter, UNIX_SQ_DB);
3805 } else if (strcasecmp(p, "unix_stream") == 0 ||
3806 strcmp(p, "u_str") == 0) {
3807 filter_db_set(¤t_filter, UNIX_ST_DB);
3808 } else if (strcasecmp(p, "unix_dgram") == 0 ||
3809 strcmp(p, "u_dgr") == 0) {
3810 filter_db_set(¤t_filter, UNIX_DG_DB);
3811 } else if (strcasecmp(p, "unix_seqpacket") == 0 ||
3812 strcmp(p, "u_seq") == 0) {
3813 filter_db_set(¤t_filter, UNIX_SQ_DB);
3814 } else if (strcmp(p, "packet") == 0) {
3815 filter_db_set(¤t_filter, PACKET_R_DB);
3816 filter_db_set(¤t_filter, PACKET_DG_DB);
3817 } else if (strcmp(p, "packet_raw") == 0 ||
3818 strcmp(p, "p_raw") == 0) {
3819 filter_db_set(¤t_filter, PACKET_R_DB);
3820 } else if (strcmp(p, "packet_dgram") == 0 ||
3821 strcmp(p, "p_dgr") == 0) {
3822 filter_db_set(¤t_filter, PACKET_DG_DB);
3823 } else if (strcmp(p, "netlink") == 0) {
3824 filter_db_set(¤t_filter, NETLINK_DB);
3825 } else {
3826 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
3827 usage();
3828 }
3829 p = p1 + 1;
3830 } while (p1);
3831 break;
3832 }
3833 case 's':
3834 do_summary = 1;
3835 break;
3836 case 'D':
3837 dump_tcpdiag = optarg;
3838 break;
3839 case 'F':
3840 if (filter_fp) {
3841 fprintf(stderr, "More than one filter file\n");
3842 exit(-1);
3843 }
3844 if (optarg[0] == '-')
3845 filter_fp = stdin;
3846 else
3847 filter_fp = fopen(optarg, "r");
3848 if (!filter_fp) {
3849 perror("fopen filter file");
3850 exit(-1);
3851 }
3852 break;
3853 case 'v':
3854 case 'V':
3855 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
3856 exit(0);
3857 case 'z':
3858 show_sock_ctx++;
3859 case 'Z':
3860 if (is_selinux_enabled() <= 0) {
3861 fprintf(stderr, "ss: SELinux is not enabled.\n");
3862 exit(1);
3863 }
3864 show_proc_ctx++;
3865 user_ent_hash_build();
3866 break;
3867 case 'N':
3868 if (netns_switch(optarg))
3869 exit(1);
3870 break;
3871 case 'K':
3872 current_filter.kill = 1;
3873 break;
3874 case 'h':
3875 help();
3876 case '?':
3877 default:
3878 usage();
3879 }
3880 }
3881
3882 argc -= optind;
3883 argv += optind;
3884
3885 if (do_summary) {
3886 print_summary();
3887 if (do_default && argc == 0)
3888 exit(0);
3889 }
3890
3891 while (argc > 0) {
3892 if (strcmp(*argv, "state") == 0) {
3893 NEXT_ARG();
3894 if (!saw_states)
3895 state_filter = 0;
3896 state_filter |= scan_state(*argv);
3897 saw_states = 1;
3898 } else if (strcmp(*argv, "exclude") == 0 ||
3899 strcmp(*argv, "excl") == 0) {
3900 NEXT_ARG();
3901 if (!saw_states)
3902 state_filter = SS_ALL;
3903 state_filter &= ~scan_state(*argv);
3904 saw_states = 1;
3905 } else {
3906 break;
3907 }
3908 argc--; argv++;
3909 }
3910
3911 if (do_default) {
3912 state_filter = state_filter ? state_filter : SS_CONN;
3913 filter_default_dbs(¤t_filter);
3914 }
3915
3916 filter_states_set(¤t_filter, state_filter);
3917 filter_merge_defaults(¤t_filter);
3918
3919 if (resolve_services && resolve_hosts &&
3920 (current_filter.dbs&(UNIX_DBM|(1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB))))
3921 init_service_resolver();
3922
3923
3924 if (current_filter.dbs == 0) {
3925 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
3926 exit(0);
3927 }
3928 if (current_filter.families == 0) {
3929 fprintf(stderr, "ss: no families to show with such filter.\n");
3930 exit(0);
3931 }
3932 if (current_filter.states == 0) {
3933 fprintf(stderr, "ss: no socket states to show with such filter.\n");
3934 exit(0);
3935 }
3936
3937 if (dump_tcpdiag) {
3938 FILE *dump_fp = stdout;
3939 if (!(current_filter.dbs & (1<<TCP_DB))) {
3940 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
3941 exit(0);
3942 }
3943 if (dump_tcpdiag[0] != '-') {
3944 dump_fp = fopen(dump_tcpdiag, "w");
3945 if (!dump_tcpdiag) {
3946 perror("fopen dump file");
3947 exit(-1);
3948 }
3949 }
3950 inet_show_netlink(¤t_filter, dump_fp, IPPROTO_TCP);
3951 fflush(dump_fp);
3952 exit(0);
3953 }
3954
3955 if (ssfilter_parse(¤t_filter.f, argc, argv, filter_fp))
3956 usage();
3957
3958 netid_width = 0;
3959 if (current_filter.dbs&(current_filter.dbs-1))
3960 netid_width = 5;
3961
3962 state_width = 0;
3963 if (current_filter.states&(current_filter.states-1))
3964 state_width = 10;
3965
3966 screen_width = 80;
3967 if (isatty(STDOUT_FILENO)) {
3968 struct winsize w;
3969
3970 if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) != -1) {
3971 if (w.ws_col > 0)
3972 screen_width = w.ws_col;
3973 }
3974 }
3975
3976 addrp_width = screen_width;
3977 addrp_width -= netid_width+1;
3978 addrp_width -= state_width+1;
3979 addrp_width -= 14;
3980
3981 if (addrp_width&1) {
3982 if (netid_width)
3983 netid_width++;
3984 else if (state_width)
3985 state_width++;
3986 }
3987
3988 addrp_width /= 2;
3989 addrp_width--;
3990
3991 serv_width = resolve_services ? 7 : 5;
3992
3993 if (addrp_width < 15+serv_width+1)
3994 addrp_width = 15+serv_width+1;
3995
3996 addr_width = addrp_width - serv_width - 1;
3997
3998 if (netid_width)
3999 printf("%-*s ", netid_width, "Netid");
4000 if (state_width)
4001 printf("%-*s ", state_width, "State");
4002 printf("%-6s %-6s ", "Recv-Q", "Send-Q");
4003
4004 /* Make enough space for the local/remote port field */
4005 addr_width -= 13;
4006 serv_width += 13;
4007
4008 printf("%*s:%-*s %*s:%-*s\n",
4009 addr_width, "Local Address", serv_width, "Port",
4010 addr_width, "Peer Address", serv_width, "Port");
4011
4012 fflush(stdout);
4013
4014 if (follow_events)
4015 exit(handle_follow_request(¤t_filter));
4016
4017 if (current_filter.dbs & (1<<NETLINK_DB))
4018 netlink_show(¤t_filter);
4019 if (current_filter.dbs & PACKET_DBM)
4020 packet_show(¤t_filter);
4021 if (current_filter.dbs & UNIX_DBM)
4022 unix_show(¤t_filter);
4023 if (current_filter.dbs & (1<<RAW_DB))
4024 raw_show(¤t_filter);
4025 if (current_filter.dbs & (1<<UDP_DB))
4026 udp_show(¤t_filter);
4027 if (current_filter.dbs & (1<<TCP_DB))
4028 tcp_show(¤t_filter, IPPROTO_TCP);
4029 if (current_filter.dbs & (1<<DCCP_DB))
4030 tcp_show(¤t_filter, IPPROTO_DCCP);
4031
4032 if (show_users || show_proc_ctx || show_sock_ctx)
4033 user_ent_destroy();
4034
4035 return 0;
4036 }
4037