]> git.proxmox.com Git - mirror_iproute2.git/blob - misc/ss.c
projects / mirror_iproute2.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'master' into net-next
[mirror_iproute2.git] / misc / ss.c
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 <fcntl.h>
16 #include <sys/ioctl.h>
17 #include <sys/socket.h>
18 #include <sys/uio.h>
19 #include <netinet/in.h>
20 #include <string.h>
21 #include <errno.h>
22 #include <netdb.h>
23 #include <arpa/inet.h>
24 #include <dirent.h>
25 #include <fnmatch.h>
26 #include <getopt.h>
27 #include <stdbool.h>
28 #include <limits.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 #include <linux/sctp.h>
46 #include <linux/vm_sockets_diag.h>
47
48 #define MAGIC_SEQ 123456
49
50 #define DIAG_REQUEST(_req, _r) \
51 struct { \
52 struct nlmsghdr nlh; \
53 _r; \
54 } _req = { \
55 .nlh = { \
56 .nlmsg_type = SOCK_DIAG_BY_FAMILY, \
57 .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,\
58 .nlmsg_seq = MAGIC_SEQ, \
59 .nlmsg_len = sizeof(_req), \
60 }, \
61 }
62
63 #if HAVE_SELINUX
64 #include <selinux/selinux.h>
65 #else
66 /* Stubs for SELinux functions */
67 static int is_selinux_enabled(void)
68 {
69 return -1;
70 }
71
72 static int getpidcon(pid_t pid, char **context)
73 {
74 *context = NULL;
75 return -1;
76 }
77
78 static int getfilecon(char *path, char **context)
79 {
80 *context = NULL;
81 return -1;
82 }
83
84 static int security_get_initial_context(char *name, char **context)
85 {
86 *context = NULL;
87 return -1;
88 }
89 #endif
90
91 int resolve_services = 1;
92 int preferred_family = AF_UNSPEC;
93 int show_options;
94 int show_details;
95 int show_users;
96 int show_mem;
97 int show_tcpinfo;
98 int show_bpf;
99 int show_proc_ctx;
100 int show_sock_ctx;
101 int show_header = 1;
102 int follow_events;
103 int sctp_ino;
104
105 int netid_width;
106 int state_width;
107 int addr_width;
108 int serv_width;
109 char *odd_width_pad = "";
110
111 static const char *TCP_PROTO = "tcp";
112 static const char *SCTP_PROTO = "sctp";
113 static const char *UDP_PROTO = "udp";
114 static const char *RAW_PROTO = "raw";
115 static const char *dg_proto;
116
117 enum {
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 SCTP_DB,
129 VSOCK_ST_DB,
130 VSOCK_DG_DB,
131 MAX_DB
132 };
133
134 #define PACKET_DBM ((1<<PACKET_DG_DB)|(1<<PACKET_R_DB))
135 #define UNIX_DBM ((1<<UNIX_DG_DB)|(1<<UNIX_ST_DB)|(1<<UNIX_SQ_DB))
136 #define ALL_DB ((1<<MAX_DB)-1)
137 #define INET_L4_DBM ((1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB)|(1<<SCTP_DB))
138 #define INET_DBM (INET_L4_DBM | (1<<RAW_DB))
139 #define VSOCK_DBM ((1<<VSOCK_ST_DB)|(1<<VSOCK_DG_DB))
140
141 enum {
142 SS_UNKNOWN,
143 SS_ESTABLISHED,
144 SS_SYN_SENT,
145 SS_SYN_RECV,
146 SS_FIN_WAIT1,
147 SS_FIN_WAIT2,
148 SS_TIME_WAIT,
149 SS_CLOSE,
150 SS_CLOSE_WAIT,
151 SS_LAST_ACK,
152 SS_LISTEN,
153 SS_CLOSING,
154 SS_MAX
155 };
156
157 enum {
158 SCTP_STATE_CLOSED = 0,
159 SCTP_STATE_COOKIE_WAIT = 1,
160 SCTP_STATE_COOKIE_ECHOED = 2,
161 SCTP_STATE_ESTABLISHED = 3,
162 SCTP_STATE_SHUTDOWN_PENDING = 4,
163 SCTP_STATE_SHUTDOWN_SENT = 5,
164 SCTP_STATE_SHUTDOWN_RECEIVED = 6,
165 SCTP_STATE_SHUTDOWN_ACK_SENT = 7,
166 };
167
168 #define SS_ALL ((1 << SS_MAX) - 1)
169 #define SS_CONN (SS_ALL & ~((1<<SS_LISTEN)|(1<<SS_CLOSE)|(1<<SS_TIME_WAIT)|(1<<SS_SYN_RECV)))
170
171 #include "ssfilter.h"
172
173 struct filter {
174 int dbs;
175 int states;
176 uint64_t families;
177 struct ssfilter *f;
178 bool kill;
179 };
180
181 #define FAMILY_MASK(family) ((uint64_t)1 << (family))
182
183 static const struct filter default_dbs[MAX_DB] = {
184 [TCP_DB] = {
185 .states = SS_CONN,
186 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
187 },
188 [DCCP_DB] = {
189 .states = SS_CONN,
190 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
191 },
192 [UDP_DB] = {
193 .states = (1 << SS_ESTABLISHED),
194 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
195 },
196 [RAW_DB] = {
197 .states = (1 << SS_ESTABLISHED),
198 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
199 },
200 [UNIX_DG_DB] = {
201 .states = (1 << SS_CLOSE),
202 .families = FAMILY_MASK(AF_UNIX),
203 },
204 [UNIX_ST_DB] = {
205 .states = SS_CONN,
206 .families = FAMILY_MASK(AF_UNIX),
207 },
208 [UNIX_SQ_DB] = {
209 .states = SS_CONN,
210 .families = FAMILY_MASK(AF_UNIX),
211 },
212 [PACKET_DG_DB] = {
213 .states = (1 << SS_CLOSE),
214 .families = FAMILY_MASK(AF_PACKET),
215 },
216 [PACKET_R_DB] = {
217 .states = (1 << SS_CLOSE),
218 .families = FAMILY_MASK(AF_PACKET),
219 },
220 [NETLINK_DB] = {
221 .states = (1 << SS_CLOSE),
222 .families = FAMILY_MASK(AF_NETLINK),
223 },
224 [SCTP_DB] = {
225 .states = SS_CONN,
226 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
227 },
228 [VSOCK_ST_DB] = {
229 .states = SS_CONN,
230 .families = FAMILY_MASK(AF_VSOCK),
231 },
232 [VSOCK_DG_DB] = {
233 .states = SS_CONN,
234 .families = FAMILY_MASK(AF_VSOCK),
235 },
236 };
237
238 static const struct filter default_afs[AF_MAX] = {
239 [AF_INET] = {
240 .dbs = INET_DBM,
241 .states = SS_CONN,
242 },
243 [AF_INET6] = {
244 .dbs = INET_DBM,
245 .states = SS_CONN,
246 },
247 [AF_UNIX] = {
248 .dbs = UNIX_DBM,
249 .states = SS_CONN,
250 },
251 [AF_PACKET] = {
252 .dbs = PACKET_DBM,
253 .states = (1 << SS_CLOSE),
254 },
255 [AF_NETLINK] = {
256 .dbs = (1 << NETLINK_DB),
257 .states = (1 << SS_CLOSE),
258 },
259 [AF_VSOCK] = {
260 .dbs = VSOCK_DBM,
261 .states = SS_CONN,
262 },
263 };
264
265 static int do_default = 1;
266 static struct filter current_filter;
267
268 static void filter_db_set(struct filter *f, int db)
269 {
270 f->states |= default_dbs[db].states;
271 f->dbs |= 1 << db;
272 do_default = 0;
273 }
274
275 static void filter_af_set(struct filter *f, int af)
276 {
277 f->states |= default_afs[af].states;
278 f->families |= FAMILY_MASK(af);
279 do_default = 0;
280 preferred_family = af;
281 }
282
283 static int filter_af_get(struct filter *f, int af)
284 {
285 return !!(f->families & FAMILY_MASK(af));
286 }
287
288 static void filter_default_dbs(struct filter *f)
289 {
290 filter_db_set(f, UDP_DB);
291 filter_db_set(f, DCCP_DB);
292 filter_db_set(f, TCP_DB);
293 filter_db_set(f, RAW_DB);
294 filter_db_set(f, UNIX_ST_DB);
295 filter_db_set(f, UNIX_DG_DB);
296 filter_db_set(f, UNIX_SQ_DB);
297 filter_db_set(f, PACKET_R_DB);
298 filter_db_set(f, PACKET_DG_DB);
299 filter_db_set(f, NETLINK_DB);
300 filter_db_set(f, SCTP_DB);
301 filter_db_set(f, VSOCK_ST_DB);
302 filter_db_set(f, VSOCK_DG_DB);
303 }
304
305 static void filter_states_set(struct filter *f, int states)
306 {
307 if (states)
308 f->states = states;
309 }
310
311 static void filter_merge_defaults(struct filter *f)
312 {
313 int db;
314 int af;
315
316 for (db = 0; db < MAX_DB; db++) {
317 if (!(f->dbs & (1 << db)))
318 continue;
319
320 if (!(default_dbs[db].families & f->families))
321 f->families |= default_dbs[db].families;
322 }
323 for (af = 0; af < AF_MAX; af++) {
324 if (!(f->families & FAMILY_MASK(af)))
325 continue;
326
327 if (!(default_afs[af].dbs & f->dbs))
328 f->dbs |= default_afs[af].dbs;
329 }
330 }
331
332 static FILE *generic_proc_open(const char *env, const char *name)
333 {
334 const char *p = getenv(env);
335 char store[128];
336
337 if (!p) {
338 p = getenv("PROC_ROOT") ? : "/proc";
339 snprintf(store, sizeof(store)-1, "%s/%s", p, name);
340 p = store;
341 }
342
343 return fopen(p, "r");
344 }
345 #define net_tcp_open() generic_proc_open("PROC_NET_TCP", "net/tcp")
346 #define net_tcp6_open() generic_proc_open("PROC_NET_TCP6", "net/tcp6")
347 #define net_udp_open() generic_proc_open("PROC_NET_UDP", "net/udp")
348 #define net_udp6_open() generic_proc_open("PROC_NET_UDP6", "net/udp6")
349 #define net_raw_open() generic_proc_open("PROC_NET_RAW", "net/raw")
350 #define net_raw6_open() generic_proc_open("PROC_NET_RAW6", "net/raw6")
351 #define net_unix_open() generic_proc_open("PROC_NET_UNIX", "net/unix")
352 #define net_packet_open() generic_proc_open("PROC_NET_PACKET", \
353 "net/packet")
354 #define net_netlink_open() generic_proc_open("PROC_NET_NETLINK", \
355 "net/netlink")
356 #define slabinfo_open() generic_proc_open("PROC_SLABINFO", "slabinfo")
357 #define net_sockstat_open() generic_proc_open("PROC_NET_SOCKSTAT", \
358 "net/sockstat")
359 #define net_sockstat6_open() generic_proc_open("PROC_NET_SOCKSTAT6", \
360 "net/sockstat6")
361 #define net_snmp_open() generic_proc_open("PROC_NET_SNMP", "net/snmp")
362 #define ephemeral_ports_open() generic_proc_open("PROC_IP_LOCAL_PORT_RANGE", \
363 "sys/net/ipv4/ip_local_port_range")
364
365 struct user_ent {
366 struct user_ent *next;
367 unsigned int ino;
368 int pid;
369 int fd;
370 char *process;
371 char *process_ctx;
372 char *socket_ctx;
373 };
374
375 #define USER_ENT_HASH_SIZE 256
376 struct user_ent *user_ent_hash[USER_ENT_HASH_SIZE];
377
378 static int user_ent_hashfn(unsigned int ino)
379 {
380 int val = (ino >> 24) ^ (ino >> 16) ^ (ino >> 8) ^ ino;
381
382 return val & (USER_ENT_HASH_SIZE - 1);
383 }
384
385 static void user_ent_add(unsigned int ino, char *process,
386 int pid, int fd,
387 char *proc_ctx,
388 char *sock_ctx)
389 {
390 struct user_ent *p, **pp;
391
392 p = malloc(sizeof(struct user_ent));
393 if (!p) {
394 fprintf(stderr, "ss: failed to malloc buffer\n");
395 abort();
396 }
397 p->next = NULL;
398 p->ino = ino;
399 p->pid = pid;
400 p->fd = fd;
401 p->process = strdup(process);
402 p->process_ctx = strdup(proc_ctx);
403 p->socket_ctx = strdup(sock_ctx);
404
405 pp = &user_ent_hash[user_ent_hashfn(ino)];
406 p->next = *pp;
407 *pp = p;
408 }
409
410 static void user_ent_destroy(void)
411 {
412 struct user_ent *p, *p_next;
413 int cnt = 0;
414
415 while (cnt != USER_ENT_HASH_SIZE) {
416 p = user_ent_hash[cnt];
417 while (p) {
418 free(p->process);
419 free(p->process_ctx);
420 free(p->socket_ctx);
421 p_next = p->next;
422 free(p);
423 p = p_next;
424 }
425 cnt++;
426 }
427 }
428
429 static void user_ent_hash_build(void)
430 {
431 const char *root = getenv("PROC_ROOT") ? : "/proc/";
432 struct dirent *d;
433 char name[1024];
434 int nameoff;
435 DIR *dir;
436 char *pid_context;
437 char *sock_context;
438 const char *no_ctx = "unavailable";
439 static int user_ent_hash_build_init;
440
441 /* If show_users & show_proc_ctx set only do this once */
442 if (user_ent_hash_build_init != 0)
443 return;
444
445 user_ent_hash_build_init = 1;
446
447 strlcpy(name, root, sizeof(name));
448
449 if (strlen(name) == 0 || name[strlen(name)-1] != '/')
450 strcat(name, "/");
451
452 nameoff = strlen(name);
453
454 dir = opendir(name);
455 if (!dir)
456 return;
457
458 while ((d = readdir(dir)) != NULL) {
459 struct dirent *d1;
460 char process[16];
461 char *p;
462 int pid, pos;
463 DIR *dir1;
464 char crap;
465
466 if (sscanf(d->d_name, "%d%c", &pid, &crap) != 1)
467 continue;
468
469 if (getpidcon(pid, &pid_context) != 0)
470 pid_context = strdup(no_ctx);
471
472 snprintf(name + nameoff, sizeof(name) - nameoff, "%d/fd/", pid);
473 pos = strlen(name);
474 if ((dir1 = opendir(name)) == NULL) {
475 free(pid_context);
476 continue;
477 }
478
479 process[0] = '\0';
480 p = process;
481
482 while ((d1 = readdir(dir1)) != NULL) {
483 const char *pattern = "socket:[";
484 unsigned int ino;
485 char lnk[64];
486 int fd;
487 ssize_t link_len;
488 char tmp[1024];
489
490 if (sscanf(d1->d_name, "%d%c", &fd, &crap) != 1)
491 continue;
492
493 snprintf(name+pos, sizeof(name) - pos, "%d", fd);
494
495 link_len = readlink(name, lnk, sizeof(lnk)-1);
496 if (link_len == -1)
497 continue;
498 lnk[link_len] = '\0';
499
500 if (strncmp(lnk, pattern, strlen(pattern)))
501 continue;
502
503 sscanf(lnk, "socket:[%u]", &ino);
504
505 snprintf(tmp, sizeof(tmp), "%s/%d/fd/%s",
506 root, pid, d1->d_name);
507
508 if (getfilecon(tmp, &sock_context) <= 0)
509 sock_context = strdup(no_ctx);
510
511 if (*p == '\0') {
512 FILE *fp;
513
514 snprintf(tmp, sizeof(tmp), "%s/%d/stat",
515 root, pid);
516 if ((fp = fopen(tmp, "r")) != NULL) {
517 if (fscanf(fp, "%*d (%[^)])", p) < 1)
518 ; /* ignore */
519 fclose(fp);
520 }
521 }
522 user_ent_add(ino, p, pid, fd,
523 pid_context, sock_context);
524 free(sock_context);
525 }
526 free(pid_context);
527 closedir(dir1);
528 }
529 closedir(dir);
530 }
531
532 enum entry_types {
533 USERS,
534 PROC_CTX,
535 PROC_SOCK_CTX
536 };
537
538 #define ENTRY_BUF_SIZE 512
539 static int find_entry(unsigned int ino, char **buf, int type)
540 {
541 struct user_ent *p;
542 int cnt = 0;
543 char *ptr;
544 char *new_buf;
545 int len, new_buf_len;
546 int buf_used = 0;
547 int buf_len = 0;
548
549 if (!ino)
550 return 0;
551
552 p = user_ent_hash[user_ent_hashfn(ino)];
553 ptr = *buf = NULL;
554 while (p) {
555 if (p->ino != ino)
556 goto next;
557
558 while (1) {
559 ptr = *buf + buf_used;
560 switch (type) {
561 case USERS:
562 len = snprintf(ptr, buf_len - buf_used,
563 "(\"%s\",pid=%d,fd=%d),",
564 p->process, p->pid, p->fd);
565 break;
566 case PROC_CTX:
567 len = snprintf(ptr, buf_len - buf_used,
568 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d),",
569 p->process, p->pid,
570 p->process_ctx, p->fd);
571 break;
572 case PROC_SOCK_CTX:
573 len = snprintf(ptr, buf_len - buf_used,
574 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d,sock_ctx=%s),",
575 p->process, p->pid,
576 p->process_ctx, p->fd,
577 p->socket_ctx);
578 break;
579 default:
580 fprintf(stderr, "ss: invalid type: %d\n", type);
581 abort();
582 }
583
584 if (len < 0 || len >= buf_len - buf_used) {
585 new_buf_len = buf_len + ENTRY_BUF_SIZE;
586 new_buf = realloc(*buf, new_buf_len);
587 if (!new_buf) {
588 fprintf(stderr, "ss: failed to malloc buffer\n");
589 abort();
590 }
591 *buf = new_buf;
592 buf_len = new_buf_len;
593 continue;
594 } else {
595 buf_used += len;
596 break;
597 }
598 }
599 cnt++;
600 next:
601 p = p->next;
602 }
603 if (buf_used) {
604 ptr = *buf + buf_used;
605 ptr[-1] = '\0';
606 }
607 return cnt;
608 }
609
610 /* Get stats from slab */
611
612 struct slabstat {
613 int socks;
614 int tcp_ports;
615 int tcp_tws;
616 int tcp_syns;
617 int skbs;
618 };
619
620 static struct slabstat slabstat;
621
622 static int get_slabstat(struct slabstat *s)
623 {
624 char buf[256];
625 FILE *fp;
626 int cnt;
627 static int slabstat_valid;
628 static const char * const slabstat_ids[] = {
629 "sock",
630 "tcp_bind_bucket",
631 "tcp_tw_bucket",
632 "tcp_open_request",
633 "skbuff_head_cache",
634 };
635
636 if (slabstat_valid)
637 return 0;
638
639 memset(s, 0, sizeof(*s));
640
641 fp = slabinfo_open();
642 if (!fp)
643 return -1;
644
645 cnt = sizeof(*s)/sizeof(int);
646
647 if (!fgets(buf, sizeof(buf), fp)) {
648 fclose(fp);
649 return -1;
650 }
651 while (fgets(buf, sizeof(buf), fp) != NULL) {
652 int i;
653
654 for (i = 0; i < ARRAY_SIZE(slabstat_ids); i++) {
655 if (memcmp(buf, slabstat_ids[i], strlen(slabstat_ids[i])) == 0) {
656 sscanf(buf, "%*s%d", ((int *)s) + i);
657 cnt--;
658 break;
659 }
660 }
661 if (cnt <= 0)
662 break;
663 }
664
665 slabstat_valid = 1;
666
667 fclose(fp);
668 return 0;
669 }
670
671 static unsigned long long cookie_sk_get(const uint32_t *cookie)
672 {
673 return (((unsigned long long)cookie[1] << 31) << 1) | cookie[0];
674 }
675
676 static const char *sctp_sstate_name[] = {
677 [SCTP_STATE_CLOSED] = "CLOSED",
678 [SCTP_STATE_COOKIE_WAIT] = "COOKIE_WAIT",
679 [SCTP_STATE_COOKIE_ECHOED] = "COOKIE_ECHOED",
680 [SCTP_STATE_ESTABLISHED] = "ESTAB",
681 [SCTP_STATE_SHUTDOWN_PENDING] = "SHUTDOWN_PENDING",
682 [SCTP_STATE_SHUTDOWN_SENT] = "SHUTDOWN_SENT",
683 [SCTP_STATE_SHUTDOWN_RECEIVED] = "SHUTDOWN_RECEIVED",
684 [SCTP_STATE_SHUTDOWN_ACK_SENT] = "ACK_SENT",
685 };
686
687 struct sockstat {
688 struct sockstat *next;
689 unsigned int type;
690 uint16_t prot;
691 uint16_t raw_prot;
692 inet_prefix local;
693 inet_prefix remote;
694 int lport;
695 int rport;
696 int state;
697 int rq, wq;
698 unsigned int ino;
699 unsigned int uid;
700 int refcnt;
701 unsigned int iface;
702 unsigned long long sk;
703 char *name;
704 char *peer_name;
705 __u32 mark;
706 };
707
708 struct dctcpstat {
709 unsigned int ce_state;
710 unsigned int alpha;
711 unsigned int ab_ecn;
712 unsigned int ab_tot;
713 bool enabled;
714 };
715
716 struct tcpstat {
717 struct sockstat ss;
718 unsigned int timer;
719 unsigned int timeout;
720 int probes;
721 char cong_alg[16];
722 double rto, ato, rtt, rttvar;
723 int qack, ssthresh, backoff;
724 double send_bps;
725 int snd_wscale;
726 int rcv_wscale;
727 int mss;
728 int rcv_mss;
729 int advmss;
730 unsigned int cwnd;
731 unsigned int lastsnd;
732 unsigned int lastrcv;
733 unsigned int lastack;
734 double pacing_rate;
735 double pacing_rate_max;
736 double delivery_rate;
737 unsigned long long bytes_acked;
738 unsigned long long bytes_received;
739 unsigned int segs_out;
740 unsigned int segs_in;
741 unsigned int data_segs_out;
742 unsigned int data_segs_in;
743 unsigned int unacked;
744 unsigned int retrans;
745 unsigned int retrans_total;
746 unsigned int lost;
747 unsigned int sacked;
748 unsigned int fackets;
749 unsigned int reordering;
750 unsigned int not_sent;
751 double rcv_rtt;
752 double min_rtt;
753 int rcv_space;
754 unsigned int rcv_ssthresh;
755 unsigned long long busy_time;
756 unsigned long long rwnd_limited;
757 unsigned long long sndbuf_limited;
758 bool has_ts_opt;
759 bool has_sack_opt;
760 bool has_ecn_opt;
761 bool has_ecnseen_opt;
762 bool has_fastopen_opt;
763 bool has_wscale_opt;
764 bool app_limited;
765 struct dctcpstat *dctcp;
766 struct tcp_bbr_info *bbr_info;
767 };
768
769 /* SCTP assocs share the same inode number with their parent endpoint. So if we
770 * have seen the inode number before, it must be an assoc instead of the next
771 * endpoint. */
772 static bool is_sctp_assoc(struct sockstat *s, const char *sock_name)
773 {
774 if (strcmp(sock_name, "sctp"))
775 return false;
776 if (!sctp_ino || sctp_ino != s->ino)
777 return false;
778 return true;
779 }
780
781 static const char *unix_netid_name(int type)
782 {
783 switch (type) {
784 case SOCK_STREAM:
785 return "u_str";
786 case SOCK_SEQPACKET:
787 return "u_seq";
788 case SOCK_DGRAM:
789 default:
790 return "u_dgr";
791 }
792 }
793
794 static const char *proto_name(int protocol)
795 {
796 switch (protocol) {
797 case 0:
798 return "raw";
799 case IPPROTO_UDP:
800 return "udp";
801 case IPPROTO_TCP:
802 return "tcp";
803 case IPPROTO_SCTP:
804 return "sctp";
805 case IPPROTO_DCCP:
806 return "dccp";
807 case IPPROTO_ICMPV6:
808 return "icmp6";
809 }
810
811 return "???";
812 }
813
814 static const char *vsock_netid_name(int type)
815 {
816 switch (type) {
817 case SOCK_STREAM:
818 return "v_str";
819 case SOCK_DGRAM:
820 return "v_dgr";
821 default:
822 return "???";
823 }
824 }
825
826 static void sock_state_print(struct sockstat *s)
827 {
828 const char *sock_name;
829 static const char * const sstate_name[] = {
830 "UNKNOWN",
831 [SS_ESTABLISHED] = "ESTAB",
832 [SS_SYN_SENT] = "SYN-SENT",
833 [SS_SYN_RECV] = "SYN-RECV",
834 [SS_FIN_WAIT1] = "FIN-WAIT-1",
835 [SS_FIN_WAIT2] = "FIN-WAIT-2",
836 [SS_TIME_WAIT] = "TIME-WAIT",
837 [SS_CLOSE] = "UNCONN",
838 [SS_CLOSE_WAIT] = "CLOSE-WAIT",
839 [SS_LAST_ACK] = "LAST-ACK",
840 [SS_LISTEN] = "LISTEN",
841 [SS_CLOSING] = "CLOSING",
842 };
843
844 switch (s->local.family) {
845 case AF_UNIX:
846 sock_name = unix_netid_name(s->type);
847 break;
848 case AF_INET:
849 case AF_INET6:
850 sock_name = proto_name(s->type);
851 break;
852 case AF_PACKET:
853 sock_name = s->type == SOCK_RAW ? "p_raw" : "p_dgr";
854 break;
855 case AF_NETLINK:
856 sock_name = "nl";
857 break;
858 case AF_VSOCK:
859 sock_name = vsock_netid_name(s->type);
860 break;
861 default:
862 sock_name = "unknown";
863 }
864
865 if (netid_width)
866 printf("%-*s ", netid_width,
867 is_sctp_assoc(s, sock_name) ? "" : sock_name);
868 if (state_width) {
869 if (is_sctp_assoc(s, sock_name))
870 printf("`- %-*s ", state_width - 3,
871 sctp_sstate_name[s->state]);
872 else
873 printf("%-*s ", state_width, sstate_name[s->state]);
874 }
875
876 printf("%-6d %-6d %s", s->rq, s->wq, odd_width_pad);
877 }
878
879 static void sock_details_print(struct sockstat *s)
880 {
881 if (s->uid)
882 printf(" uid:%u", s->uid);
883
884 printf(" ino:%u", s->ino);
885 printf(" sk:%llx", s->sk);
886
887 if (s->mark)
888 printf(" fwmark:0x%x", s->mark);
889 }
890
891 static void sock_addr_print_width(int addr_len, const char *addr, char *delim,
892 int port_len, const char *port, const char *ifname)
893 {
894 if (ifname) {
895 printf("%*s%%%s%s%-*s ", addr_len, addr, ifname, delim,
896 port_len, port);
897 } else {
898 printf("%*s%s%-*s ", addr_len, addr, delim, port_len, port);
899 }
900 }
901
902 static void sock_addr_print(const char *addr, char *delim, const char *port,
903 const char *ifname)
904 {
905 sock_addr_print_width(addr_width, addr, delim, serv_width, port, ifname);
906 }
907
908 static const char *print_ms_timer(unsigned int timeout)
909 {
910 static char buf[64];
911 int secs, msecs, minutes;
912
913 secs = timeout/1000;
914 minutes = secs/60;
915 secs = secs%60;
916 msecs = timeout%1000;
917 buf[0] = 0;
918 if (minutes) {
919 msecs = 0;
920 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
921 if (minutes > 9)
922 secs = 0;
923 }
924 if (secs) {
925 if (secs > 9)
926 msecs = 0;
927 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
928 }
929 if (msecs)
930 sprintf(buf+strlen(buf), "%03dms", msecs);
931 return buf;
932 }
933
934 struct scache {
935 struct scache *next;
936 int port;
937 char *name;
938 const char *proto;
939 };
940
941 struct scache *rlist;
942
943 static void init_service_resolver(void)
944 {
945 char buf[128];
946 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
947
948 if (!fp)
949 return;
950
951 if (!fgets(buf, sizeof(buf), fp)) {
952 pclose(fp);
953 return;
954 }
955 while (fgets(buf, sizeof(buf), fp) != NULL) {
956 unsigned int progn, port;
957 char proto[128], prog[128] = "rpc.";
958 struct scache *c;
959
960 if (sscanf(buf, "%u %*d %s %u %s",
961 &progn, proto, &port, prog+4) != 4)
962 continue;
963
964 if (!(c = malloc(sizeof(*c))))
965 continue;
966
967 c->port = port;
968 c->name = strdup(prog);
969 if (strcmp(proto, TCP_PROTO) == 0)
970 c->proto = TCP_PROTO;
971 else if (strcmp(proto, UDP_PROTO) == 0)
972 c->proto = UDP_PROTO;
973 else if (strcmp(proto, SCTP_PROTO) == 0)
974 c->proto = SCTP_PROTO;
975 else
976 c->proto = NULL;
977 c->next = rlist;
978 rlist = c;
979 }
980 pclose(fp);
981 }
982
983 /* Even do not try default linux ephemeral port ranges:
984 * default /etc/services contains so much of useless crap
985 * wouldbe "allocated" to this area that resolution
986 * is really harmful. I shrug each time when seeing
987 * "socks" or "cfinger" in dumps.
988 */
989 static int is_ephemeral(int port)
990 {
991 static int min = 0, max;
992
993 if (!min) {
994 FILE *f = ephemeral_ports_open();
995
996 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
997 min = 1024;
998 max = 4999;
999 }
1000 if (f)
1001 fclose(f);
1002 }
1003 return port >= min && port <= max;
1004 }
1005
1006
1007 static const char *__resolve_service(int port)
1008 {
1009 struct scache *c;
1010
1011 for (c = rlist; c; c = c->next) {
1012 if (c->port == port && c->proto == dg_proto)
1013 return c->name;
1014 }
1015
1016 if (!is_ephemeral(port)) {
1017 static int notfirst;
1018 struct servent *se;
1019
1020 if (!notfirst) {
1021 setservent(1);
1022 notfirst = 1;
1023 }
1024 se = getservbyport(htons(port), dg_proto);
1025 if (se)
1026 return se->s_name;
1027 }
1028
1029 return NULL;
1030 }
1031
1032 #define SCACHE_BUCKETS 1024
1033 static struct scache *cache_htab[SCACHE_BUCKETS];
1034
1035 static const char *resolve_service(int port)
1036 {
1037 static char buf[128];
1038 struct scache *c;
1039 const char *res;
1040 int hash;
1041
1042 if (port == 0) {
1043 buf[0] = '*';
1044 buf[1] = 0;
1045 return buf;
1046 }
1047
1048 if (!resolve_services)
1049 goto do_numeric;
1050
1051 if (dg_proto == RAW_PROTO)
1052 return inet_proto_n2a(port, buf, sizeof(buf));
1053
1054
1055 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
1056
1057 for (c = cache_htab[hash]; c; c = c->next) {
1058 if (c->port == port && c->proto == dg_proto)
1059 goto do_cache;
1060 }
1061
1062 c = malloc(sizeof(*c));
1063 if (!c)
1064 goto do_numeric;
1065 res = __resolve_service(port);
1066 c->port = port;
1067 c->name = res ? strdup(res) : NULL;
1068 c->proto = dg_proto;
1069 c->next = cache_htab[hash];
1070 cache_htab[hash] = c;
1071
1072 do_cache:
1073 if (c->name)
1074 return c->name;
1075
1076 do_numeric:
1077 sprintf(buf, "%u", port);
1078 return buf;
1079 }
1080
1081 static void inet_addr_print(const inet_prefix *a, int port,
1082 unsigned int ifindex, bool v6only)
1083 {
1084 char buf[1024];
1085 const char *ap = buf;
1086 int est_len = addr_width;
1087 const char *ifname = NULL;
1088
1089 if (a->family == AF_INET) {
1090 ap = format_host(AF_INET, 4, a->data);
1091 } else {
1092 if (!v6only &&
1093 !memcmp(a->data, &in6addr_any, sizeof(in6addr_any))) {
1094 buf[0] = '*';
1095 buf[1] = 0;
1096 } else {
1097 ap = format_host(a->family, 16, a->data);
1098
1099 /* Numeric IPv6 addresses should be bracketed */
1100 if (strchr(ap, ':')) {
1101 snprintf(buf, sizeof(buf),
1102 "[%s]", ap);
1103 ap = buf;
1104 }
1105
1106 est_len = strlen(ap);
1107 if (est_len <= addr_width)
1108 est_len = addr_width;
1109 else
1110 est_len = addr_width + ((est_len-addr_width+3)/4)*4;
1111 }
1112 }
1113
1114 if (ifindex) {
1115 ifname = ll_index_to_name(ifindex);
1116 est_len -= strlen(ifname) + 1; /* +1 for percent char */
1117 if (est_len < 0)
1118 est_len = 0;
1119 }
1120
1121 sock_addr_print_width(est_len, ap, ":", serv_width, resolve_service(port),
1122 ifname);
1123 }
1124
1125 struct aafilter {
1126 inet_prefix addr;
1127 int port;
1128 unsigned int iface;
1129 __u32 mark;
1130 __u32 mask;
1131 struct aafilter *next;
1132 };
1133
1134 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1135 int plen)
1136 {
1137 if (!inet_addr_match(a, p, plen))
1138 return 0;
1139
1140 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1141 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1142 * sockets. Fair? */
1143 if (p->family == AF_INET && a->family == AF_INET6) {
1144 if (a->data[0] == 0 && a->data[1] == 0 &&
1145 a->data[2] == htonl(0xffff)) {
1146 inet_prefix tmp = *a;
1147
1148 tmp.data[0] = a->data[3];
1149 return inet_addr_match(&tmp, p, plen);
1150 }
1151 }
1152 return 1;
1153 }
1154
1155 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1156 {
1157 char *addr, *pattern;
1158
1159 memcpy(&addr, a->data, sizeof(addr));
1160 memcpy(&pattern, p->data, sizeof(pattern));
1161 if (pattern == NULL)
1162 return 1;
1163 if (addr == NULL)
1164 addr = "";
1165 return !fnmatch(pattern, addr, 0);
1166 }
1167
1168 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1169 {
1170 switch (f->type) {
1171 case SSF_S_AUTO:
1172 {
1173 if (s->local.family == AF_UNIX) {
1174 char *p;
1175
1176 memcpy(&p, s->local.data, sizeof(p));
1177 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1178 strspn(p+1, "0123456789abcdef") == 5);
1179 }
1180 if (s->local.family == AF_PACKET)
1181 return s->lport == 0 && s->local.data[0] == 0;
1182 if (s->local.family == AF_NETLINK)
1183 return s->lport < 0;
1184 if (s->local.family == AF_VSOCK)
1185 return s->lport > 1023;
1186
1187 return is_ephemeral(s->lport);
1188 }
1189 case SSF_DCOND:
1190 {
1191 struct aafilter *a = (void *)f->pred;
1192
1193 if (a->addr.family == AF_UNIX)
1194 return unix_match(&s->remote, &a->addr);
1195 if (a->port != -1 && a->port != s->rport)
1196 return 0;
1197 if (a->addr.bitlen) {
1198 do {
1199 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1200 return 1;
1201 } while ((a = a->next) != NULL);
1202 return 0;
1203 }
1204 return 1;
1205 }
1206 case SSF_SCOND:
1207 {
1208 struct aafilter *a = (void *)f->pred;
1209
1210 if (a->addr.family == AF_UNIX)
1211 return unix_match(&s->local, &a->addr);
1212 if (a->port != -1 && a->port != s->lport)
1213 return 0;
1214 if (a->addr.bitlen) {
1215 do {
1216 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1217 return 1;
1218 } while ((a = a->next) != NULL);
1219 return 0;
1220 }
1221 return 1;
1222 }
1223 case SSF_D_GE:
1224 {
1225 struct aafilter *a = (void *)f->pred;
1226
1227 return s->rport >= a->port;
1228 }
1229 case SSF_D_LE:
1230 {
1231 struct aafilter *a = (void *)f->pred;
1232
1233 return s->rport <= a->port;
1234 }
1235 case SSF_S_GE:
1236 {
1237 struct aafilter *a = (void *)f->pred;
1238
1239 return s->lport >= a->port;
1240 }
1241 case SSF_S_LE:
1242 {
1243 struct aafilter *a = (void *)f->pred;
1244
1245 return s->lport <= a->port;
1246 }
1247 case SSF_DEVCOND:
1248 {
1249 struct aafilter *a = (void *)f->pred;
1250
1251 return s->iface == a->iface;
1252 }
1253 case SSF_MARKMASK:
1254 {
1255 struct aafilter *a = (void *)f->pred;
1256
1257 return (s->mark & a->mask) == a->mark;
1258 }
1259 /* Yup. It is recursion. Sorry. */
1260 case SSF_AND:
1261 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1262 case SSF_OR:
1263 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1264 case SSF_NOT:
1265 return !run_ssfilter(f->pred, s);
1266 default:
1267 abort();
1268 }
1269 }
1270
1271 /* Relocate external jumps by reloc. */
1272 static void ssfilter_patch(char *a, int len, int reloc)
1273 {
1274 while (len > 0) {
1275 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)a;
1276
1277 if (op->no == len+4)
1278 op->no += reloc;
1279 len -= op->yes;
1280 a += op->yes;
1281 }
1282 if (len < 0)
1283 abort();
1284 }
1285
1286 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1287 {
1288 switch (f->type) {
1289 case SSF_S_AUTO:
1290 {
1291 if (!(*bytecode = malloc(4))) abort();
1292 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1293 return 4;
1294 }
1295 case SSF_DCOND:
1296 case SSF_SCOND:
1297 {
1298 struct aafilter *a = (void *)f->pred;
1299 struct aafilter *b;
1300 char *ptr;
1301 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1302 int len = 0;
1303
1304 for (b = a; b; b = b->next) {
1305 len += 4 + sizeof(struct inet_diag_hostcond);
1306 if (a->addr.family == AF_INET6)
1307 len += 16;
1308 else
1309 len += 4;
1310 if (b->next)
1311 len += 4;
1312 }
1313 if (!(ptr = malloc(len))) abort();
1314 *bytecode = ptr;
1315 for (b = a; b; b = b->next) {
1316 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1317 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1318 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1319 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond *)(ptr+4);
1320
1321 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1322 cond->family = a->addr.family;
1323 cond->port = a->port;
1324 cond->prefix_len = a->addr.bitlen;
1325 memcpy(cond->addr, a->addr.data, alen);
1326 ptr += oplen;
1327 if (b->next) {
1328 op = (struct inet_diag_bc_op *)ptr;
1329 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1330 ptr += 4;
1331 }
1332 }
1333 return ptr - *bytecode;
1334 }
1335 case SSF_D_GE:
1336 {
1337 struct aafilter *x = (void *)f->pred;
1338
1339 if (!(*bytecode = malloc(8))) abort();
1340 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_GE, 8, 12 };
1341 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1342 return 8;
1343 }
1344 case SSF_D_LE:
1345 {
1346 struct aafilter *x = (void *)f->pred;
1347
1348 if (!(*bytecode = malloc(8))) abort();
1349 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_LE, 8, 12 };
1350 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1351 return 8;
1352 }
1353 case SSF_S_GE:
1354 {
1355 struct aafilter *x = (void *)f->pred;
1356
1357 if (!(*bytecode = malloc(8))) abort();
1358 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1359 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1360 return 8;
1361 }
1362 case SSF_S_LE:
1363 {
1364 struct aafilter *x = (void *)f->pred;
1365
1366 if (!(*bytecode = malloc(8))) abort();
1367 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1368 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1369 return 8;
1370 }
1371
1372 case SSF_AND:
1373 {
1374 char *a1 = NULL, *a2 = NULL, *a;
1375 int l1, l2;
1376
1377 l1 = ssfilter_bytecompile(f->pred, &a1);
1378 l2 = ssfilter_bytecompile(f->post, &a2);
1379 if (!l1 || !l2) {
1380 free(a1);
1381 free(a2);
1382 return 0;
1383 }
1384 if (!(a = malloc(l1+l2))) abort();
1385 memcpy(a, a1, l1);
1386 memcpy(a+l1, a2, l2);
1387 free(a1); free(a2);
1388 ssfilter_patch(a, l1, l2);
1389 *bytecode = a;
1390 return l1+l2;
1391 }
1392 case SSF_OR:
1393 {
1394 char *a1 = NULL, *a2 = NULL, *a;
1395 int l1, l2;
1396
1397 l1 = ssfilter_bytecompile(f->pred, &a1);
1398 l2 = ssfilter_bytecompile(f->post, &a2);
1399 if (!l1 || !l2) {
1400 free(a1);
1401 free(a2);
1402 return 0;
1403 }
1404 if (!(a = malloc(l1+l2+4))) abort();
1405 memcpy(a, a1, l1);
1406 memcpy(a+l1+4, a2, l2);
1407 free(a1); free(a2);
1408 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1409 *bytecode = a;
1410 return l1+l2+4;
1411 }
1412 case SSF_NOT:
1413 {
1414 char *a1 = NULL, *a;
1415 int l1;
1416
1417 l1 = ssfilter_bytecompile(f->pred, &a1);
1418 if (!l1) {
1419 free(a1);
1420 return 0;
1421 }
1422 if (!(a = malloc(l1+4))) abort();
1423 memcpy(a, a1, l1);
1424 free(a1);
1425 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1426 *bytecode = a;
1427 return l1+4;
1428 }
1429 case SSF_DEVCOND:
1430 {
1431 /* bytecompile for SSF_DEVCOND not supported yet */
1432 return 0;
1433 }
1434 case SSF_MARKMASK:
1435 {
1436 struct aafilter *a = (void *)f->pred;
1437 struct instr {
1438 struct inet_diag_bc_op op;
1439 struct inet_diag_markcond cond;
1440 };
1441 int inslen = sizeof(struct instr);
1442
1443 if (!(*bytecode = malloc(inslen))) abort();
1444 ((struct instr *)*bytecode)[0] = (struct instr) {
1445 { INET_DIAG_BC_MARK_COND, inslen, inslen + 4 },
1446 { a->mark, a->mask},
1447 };
1448
1449 return inslen;
1450 }
1451 default:
1452 abort();
1453 }
1454 }
1455
1456 static int remember_he(struct aafilter *a, struct hostent *he)
1457 {
1458 char **ptr = he->h_addr_list;
1459 int cnt = 0;
1460 int len;
1461
1462 if (he->h_addrtype == AF_INET)
1463 len = 4;
1464 else if (he->h_addrtype == AF_INET6)
1465 len = 16;
1466 else
1467 return 0;
1468
1469 while (*ptr) {
1470 struct aafilter *b = a;
1471
1472 if (a->addr.bitlen) {
1473 if ((b = malloc(sizeof(*b))) == NULL)
1474 return cnt;
1475 *b = *a;
1476 a->next = b;
1477 }
1478 memcpy(b->addr.data, *ptr, len);
1479 b->addr.bytelen = len;
1480 b->addr.bitlen = len*8;
1481 b->addr.family = he->h_addrtype;
1482 ptr++;
1483 cnt++;
1484 }
1485 return cnt;
1486 }
1487
1488 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
1489 {
1490 static int notfirst;
1491 int cnt = 0;
1492 struct hostent *he;
1493
1494 a->addr.bitlen = 0;
1495 if (!notfirst) {
1496 sethostent(1);
1497 notfirst = 1;
1498 }
1499 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
1500 if (he)
1501 cnt = remember_he(a, he);
1502 if (fam == AF_UNSPEC) {
1503 he = gethostbyname2(addr, AF_INET6);
1504 if (he)
1505 cnt += remember_he(a, he);
1506 }
1507 return !cnt;
1508 }
1509
1510 static int xll_initted;
1511
1512 static void xll_init(void)
1513 {
1514 struct rtnl_handle rth;
1515
1516 if (rtnl_open(&rth, 0) < 0)
1517 exit(1);
1518
1519 ll_init_map(&rth);
1520 rtnl_close(&rth);
1521 xll_initted = 1;
1522 }
1523
1524 static const char *xll_index_to_name(int index)
1525 {
1526 if (!xll_initted)
1527 xll_init();
1528 return ll_index_to_name(index);
1529 }
1530
1531 static int xll_name_to_index(const char *dev)
1532 {
1533 if (!xll_initted)
1534 xll_init();
1535 return ll_name_to_index(dev);
1536 }
1537
1538 void *parse_devcond(char *name)
1539 {
1540 struct aafilter a = { .iface = 0 };
1541 struct aafilter *res;
1542
1543 a.iface = xll_name_to_index(name);
1544 if (a.iface == 0) {
1545 char *end;
1546 unsigned long n;
1547
1548 n = strtoul(name, &end, 0);
1549 if (!end || end == name || *end || n > UINT_MAX)
1550 return NULL;
1551
1552 a.iface = n;
1553 }
1554
1555 res = malloc(sizeof(*res));
1556 *res = a;
1557
1558 return res;
1559 }
1560
1561 static void vsock_set_inet_prefix(inet_prefix *a, __u32 cid)
1562 {
1563 *a = (inet_prefix){
1564 .bytelen = sizeof(cid),
1565 .family = AF_VSOCK,
1566 };
1567 memcpy(a->data, &cid, sizeof(cid));
1568 }
1569
1570 void *parse_hostcond(char *addr, bool is_port)
1571 {
1572 char *port = NULL;
1573 struct aafilter a = { .port = -1 };
1574 struct aafilter *res;
1575 int fam = preferred_family;
1576 struct filter *f = &current_filter;
1577
1578 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
1579 char *p;
1580
1581 a.addr.family = AF_UNIX;
1582 if (strncmp(addr, "unix:", 5) == 0)
1583 addr += 5;
1584 p = strdup(addr);
1585 a.addr.bitlen = 8*strlen(p);
1586 memcpy(a.addr.data, &p, sizeof(p));
1587 fam = AF_UNIX;
1588 goto out;
1589 }
1590
1591 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
1592 a.addr.family = AF_PACKET;
1593 a.addr.bitlen = 0;
1594 if (strncmp(addr, "link:", 5) == 0)
1595 addr += 5;
1596 port = strchr(addr, ':');
1597 if (port) {
1598 *port = 0;
1599 if (port[1] && strcmp(port+1, "*")) {
1600 if (get_integer(&a.port, port+1, 0)) {
1601 if ((a.port = xll_name_to_index(port+1)) <= 0)
1602 return NULL;
1603 }
1604 }
1605 }
1606 if (addr[0] && strcmp(addr, "*")) {
1607 unsigned short tmp;
1608
1609 a.addr.bitlen = 32;
1610 if (ll_proto_a2n(&tmp, addr))
1611 return NULL;
1612 a.addr.data[0] = ntohs(tmp);
1613 }
1614 fam = AF_PACKET;
1615 goto out;
1616 }
1617
1618 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
1619 a.addr.family = AF_NETLINK;
1620 a.addr.bitlen = 0;
1621 if (strncmp(addr, "netlink:", 8) == 0)
1622 addr += 8;
1623 port = strchr(addr, ':');
1624 if (port) {
1625 *port = 0;
1626 if (port[1] && strcmp(port+1, "*")) {
1627 if (get_integer(&a.port, port+1, 0)) {
1628 if (strcmp(port+1, "kernel") == 0)
1629 a.port = 0;
1630 else
1631 return NULL;
1632 }
1633 }
1634 }
1635 if (addr[0] && strcmp(addr, "*")) {
1636 a.addr.bitlen = 32;
1637 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
1638 return NULL;
1639 }
1640 fam = AF_NETLINK;
1641 goto out;
1642 }
1643
1644 if (fam == AF_VSOCK || strncmp(addr, "vsock:", 6) == 0) {
1645 __u32 cid = ~(__u32)0;
1646
1647 a.addr.family = AF_VSOCK;
1648 if (strncmp(addr, "vsock:", 6) == 0)
1649 addr += 6;
1650
1651 if (is_port)
1652 port = addr;
1653 else {
1654 port = strchr(addr, ':');
1655 if (port) {
1656 *port = '\0';
1657 port++;
1658 }
1659 }
1660
1661 if (port && strcmp(port, "*") &&
1662 get_u32((__u32 *)&a.port, port, 0))
1663 return NULL;
1664
1665 if (addr[0] && strcmp(addr, "*")) {
1666 a.addr.bitlen = 32;
1667 if (get_u32(&cid, addr, 0))
1668 return NULL;
1669 }
1670 vsock_set_inet_prefix(&a.addr, cid);
1671 fam = AF_VSOCK;
1672 goto out;
1673 }
1674
1675 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
1676 fam = AF_INET;
1677 if (!strncmp(addr, "inet:", 5))
1678 addr += 5;
1679 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
1680 fam = AF_INET6;
1681 if (!strncmp(addr, "inet6:", 6))
1682 addr += 6;
1683 }
1684
1685 /* URL-like literal [] */
1686 if (addr[0] == '[') {
1687 addr++;
1688 if ((port = strchr(addr, ']')) == NULL)
1689 return NULL;
1690 *port++ = 0;
1691 } else if (addr[0] == '*') {
1692 port = addr+1;
1693 } else {
1694 port = strrchr(strchr(addr, '/') ? : addr, ':');
1695 }
1696
1697 if (is_port)
1698 port = addr;
1699
1700 if (port && *port) {
1701 if (*port == ':')
1702 *port++ = 0;
1703
1704 if (*port && *port != '*') {
1705 if (get_integer(&a.port, port, 0)) {
1706 struct servent *se1 = NULL;
1707 struct servent *se2 = NULL;
1708
1709 if (current_filter.dbs&(1<<UDP_DB))
1710 se1 = getservbyname(port, UDP_PROTO);
1711 if (current_filter.dbs&(1<<TCP_DB))
1712 se2 = getservbyname(port, TCP_PROTO);
1713 if (se1 && se2 && se1->s_port != se2->s_port) {
1714 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
1715 return NULL;
1716 }
1717 if (!se1)
1718 se1 = se2;
1719 if (se1) {
1720 a.port = ntohs(se1->s_port);
1721 } else {
1722 struct scache *s;
1723
1724 for (s = rlist; s; s = s->next) {
1725 if ((s->proto == UDP_PROTO &&
1726 (current_filter.dbs&(1<<UDP_DB))) ||
1727 (s->proto == TCP_PROTO &&
1728 (current_filter.dbs&(1<<TCP_DB)))) {
1729 if (s->name && strcmp(s->name, port) == 0) {
1730 if (a.port > 0 && a.port != s->port) {
1731 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
1732 return NULL;
1733 }
1734 a.port = s->port;
1735 }
1736 }
1737 }
1738 if (a.port <= 0) {
1739 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
1740 return NULL;
1741 }
1742 }
1743 }
1744 }
1745 }
1746 if (!is_port && *addr && *addr != '*') {
1747 if (get_prefix_1(&a.addr, addr, fam)) {
1748 if (get_dns_host(&a, addr, fam)) {
1749 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
1750 return NULL;
1751 }
1752 }
1753 }
1754
1755 out:
1756 if (fam != AF_UNSPEC) {
1757 int states = f->states;
1758 f->families = 0;
1759 filter_af_set(f, fam);
1760 filter_states_set(f, states);
1761 }
1762
1763 res = malloc(sizeof(*res));
1764 if (res)
1765 memcpy(res, &a, sizeof(a));
1766 return res;
1767 }
1768
1769 void *parse_markmask(const char *markmask)
1770 {
1771 struct aafilter a, *res;
1772
1773 if (strchr(markmask, '/')) {
1774 if (sscanf(markmask, "%i/%i", &a.mark, &a.mask) != 2)
1775 return NULL;
1776 } else {
1777 a.mask = 0xffffffff;
1778 if (sscanf(markmask, "%i", &a.mark) != 1)
1779 return NULL;
1780 }
1781
1782 res = malloc(sizeof(*res));
1783 if (res)
1784 memcpy(res, &a, sizeof(a));
1785 return res;
1786 }
1787
1788 static void proc_ctx_print(struct sockstat *s)
1789 {
1790 char *buf;
1791
1792 if (show_proc_ctx || show_sock_ctx) {
1793 if (find_entry(s->ino, &buf,
1794 (show_proc_ctx & show_sock_ctx) ?
1795 PROC_SOCK_CTX : PROC_CTX) > 0) {
1796 printf(" users:(%s)", buf);
1797 free(buf);
1798 }
1799 } else if (show_users) {
1800 if (find_entry(s->ino, &buf, USERS) > 0) {
1801 printf(" users:(%s)", buf);
1802 free(buf);
1803 }
1804 }
1805 }
1806
1807 static void inet_stats_print(struct sockstat *s, bool v6only)
1808 {
1809 sock_state_print(s);
1810
1811 inet_addr_print(&s->local, s->lport, s->iface, v6only);
1812 inet_addr_print(&s->remote, s->rport, 0, v6only);
1813
1814 proc_ctx_print(s);
1815 }
1816
1817 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
1818 sockstat * s)
1819 {
1820 s->local.family = s->remote.family = family;
1821 if (family == AF_INET) {
1822 sscanf(loc, "%x:%x", s->local.data, (unsigned *)&s->lport);
1823 sscanf(rem, "%x:%x", s->remote.data, (unsigned *)&s->rport);
1824 s->local.bytelen = s->remote.bytelen = 4;
1825 return 0;
1826 } else {
1827 sscanf(loc, "%08x%08x%08x%08x:%x",
1828 s->local.data,
1829 s->local.data + 1,
1830 s->local.data + 2,
1831 s->local.data + 3,
1832 &s->lport);
1833 sscanf(rem, "%08x%08x%08x%08x:%x",
1834 s->remote.data,
1835 s->remote.data + 1,
1836 s->remote.data + 2,
1837 s->remote.data + 3,
1838 &s->rport);
1839 s->local.bytelen = s->remote.bytelen = 16;
1840 return 0;
1841 }
1842 return -1;
1843 }
1844
1845 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
1846 {
1847 char *p;
1848
1849 if ((p = strchr(line, ':')) == NULL)
1850 return -1;
1851
1852 *loc = p+2;
1853 if ((p = strchr(*loc, ':')) == NULL)
1854 return -1;
1855
1856 p[5] = 0;
1857 *rem = p+6;
1858 if ((p = strchr(*rem, ':')) == NULL)
1859 return -1;
1860
1861 p[5] = 0;
1862 *data = p+6;
1863 return 0;
1864 }
1865
1866 static char *sprint_bw(char *buf, double bw)
1867 {
1868 if (bw > 1000000.)
1869 sprintf(buf, "%.1fM", bw / 1000000.);
1870 else if (bw > 1000.)
1871 sprintf(buf, "%.1fK", bw / 1000.);
1872 else
1873 sprintf(buf, "%g", bw);
1874
1875 return buf;
1876 }
1877
1878 static void sctp_stats_print(struct sctp_info *s)
1879 {
1880 if (s->sctpi_tag)
1881 printf(" tag:%x", s->sctpi_tag);
1882 if (s->sctpi_state)
1883 printf(" state:%s", sctp_sstate_name[s->sctpi_state]);
1884 if (s->sctpi_rwnd)
1885 printf(" rwnd:%d", s->sctpi_rwnd);
1886 if (s->sctpi_unackdata)
1887 printf(" unackdata:%d", s->sctpi_unackdata);
1888 if (s->sctpi_penddata)
1889 printf(" penddata:%d", s->sctpi_penddata);
1890 if (s->sctpi_instrms)
1891 printf(" instrms:%d", s->sctpi_instrms);
1892 if (s->sctpi_outstrms)
1893 printf(" outstrms:%d", s->sctpi_outstrms);
1894 if (s->sctpi_inqueue)
1895 printf(" inqueue:%d", s->sctpi_inqueue);
1896 if (s->sctpi_outqueue)
1897 printf(" outqueue:%d", s->sctpi_outqueue);
1898 if (s->sctpi_overall_error)
1899 printf(" overerr:%d", s->sctpi_overall_error);
1900 if (s->sctpi_max_burst)
1901 printf(" maxburst:%d", s->sctpi_max_burst);
1902 if (s->sctpi_maxseg)
1903 printf(" maxseg:%d", s->sctpi_maxseg);
1904 if (s->sctpi_peer_rwnd)
1905 printf(" prwnd:%d", s->sctpi_peer_rwnd);
1906 if (s->sctpi_peer_tag)
1907 printf(" ptag:%x", s->sctpi_peer_tag);
1908 if (s->sctpi_peer_capable)
1909 printf(" pcapable:%d", s->sctpi_peer_capable);
1910 if (s->sctpi_peer_sack)
1911 printf(" psack:%d", s->sctpi_peer_sack);
1912 if (s->sctpi_s_autoclose)
1913 printf(" autoclose:%d", s->sctpi_s_autoclose);
1914 if (s->sctpi_s_adaptation_ind)
1915 printf(" adapind:%d", s->sctpi_s_adaptation_ind);
1916 if (s->sctpi_s_pd_point)
1917 printf(" pdpoint:%d", s->sctpi_s_pd_point);
1918 if (s->sctpi_s_nodelay)
1919 printf(" nodealy:%d", s->sctpi_s_nodelay);
1920 if (s->sctpi_s_disable_fragments)
1921 printf(" nofrag:%d", s->sctpi_s_disable_fragments);
1922 if (s->sctpi_s_v4mapped)
1923 printf(" v4mapped:%d", s->sctpi_s_v4mapped);
1924 if (s->sctpi_s_frag_interleave)
1925 printf(" fraginl:%d", s->sctpi_s_frag_interleave);
1926 }
1927
1928 static void tcp_stats_print(struct tcpstat *s)
1929 {
1930 char b1[64];
1931
1932 if (s->has_ts_opt)
1933 printf(" ts");
1934 if (s->has_sack_opt)
1935 printf(" sack");
1936 if (s->has_ecn_opt)
1937 printf(" ecn");
1938 if (s->has_ecnseen_opt)
1939 printf(" ecnseen");
1940 if (s->has_fastopen_opt)
1941 printf(" fastopen");
1942 if (s->cong_alg[0])
1943 printf(" %s", s->cong_alg);
1944 if (s->has_wscale_opt)
1945 printf(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
1946 if (s->rto)
1947 printf(" rto:%g", s->rto);
1948 if (s->backoff)
1949 printf(" backoff:%u", s->backoff);
1950 if (s->rtt)
1951 printf(" rtt:%g/%g", s->rtt, s->rttvar);
1952 if (s->ato)
1953 printf(" ato:%g", s->ato);
1954
1955 if (s->qack)
1956 printf(" qack:%d", s->qack);
1957 if (s->qack & 1)
1958 printf(" bidir");
1959
1960 if (s->mss)
1961 printf(" mss:%d", s->mss);
1962 if (s->rcv_mss)
1963 printf(" rcvmss:%d", s->rcv_mss);
1964 if (s->advmss)
1965 printf(" advmss:%d", s->advmss);
1966 if (s->cwnd)
1967 printf(" cwnd:%u", s->cwnd);
1968 if (s->ssthresh)
1969 printf(" ssthresh:%d", s->ssthresh);
1970
1971 if (s->bytes_acked)
1972 printf(" bytes_acked:%llu", s->bytes_acked);
1973 if (s->bytes_received)
1974 printf(" bytes_received:%llu", s->bytes_received);
1975 if (s->segs_out)
1976 printf(" segs_out:%u", s->segs_out);
1977 if (s->segs_in)
1978 printf(" segs_in:%u", s->segs_in);
1979 if (s->data_segs_out)
1980 printf(" data_segs_out:%u", s->data_segs_out);
1981 if (s->data_segs_in)
1982 printf(" data_segs_in:%u", s->data_segs_in);
1983
1984 if (s->dctcp && s->dctcp->enabled) {
1985 struct dctcpstat *dctcp = s->dctcp;
1986
1987 printf(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
1988 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
1989 dctcp->ab_tot);
1990 } else if (s->dctcp) {
1991 printf(" dctcp:fallback_mode");
1992 }
1993
1994 if (s->bbr_info) {
1995 __u64 bw;
1996
1997 bw = s->bbr_info->bbr_bw_hi;
1998 bw <<= 32;
1999 bw |= s->bbr_info->bbr_bw_lo;
2000
2001 printf(" bbr:(bw:%sbps,mrtt:%g",
2002 sprint_bw(b1, bw * 8.0),
2003 (double)s->bbr_info->bbr_min_rtt / 1000.0);
2004 if (s->bbr_info->bbr_pacing_gain)
2005 printf(",pacing_gain:%g",
2006 (double)s->bbr_info->bbr_pacing_gain / 256.0);
2007 if (s->bbr_info->bbr_cwnd_gain)
2008 printf(",cwnd_gain:%g",
2009 (double)s->bbr_info->bbr_cwnd_gain / 256.0);
2010 printf(")");
2011 }
2012
2013 if (s->send_bps)
2014 printf(" send %sbps", sprint_bw(b1, s->send_bps));
2015 if (s->lastsnd)
2016 printf(" lastsnd:%u", s->lastsnd);
2017 if (s->lastrcv)
2018 printf(" lastrcv:%u", s->lastrcv);
2019 if (s->lastack)
2020 printf(" lastack:%u", s->lastack);
2021
2022 if (s->pacing_rate) {
2023 printf(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
2024 if (s->pacing_rate_max)
2025 printf("/%sbps", sprint_bw(b1,
2026 s->pacing_rate_max));
2027 }
2028
2029 if (s->delivery_rate)
2030 printf(" delivery_rate %sbps", sprint_bw(b1, s->delivery_rate));
2031 if (s->app_limited)
2032 printf(" app_limited");
2033
2034 if (s->busy_time) {
2035 printf(" busy:%llums", s->busy_time / 1000);
2036 if (s->rwnd_limited)
2037 printf(" rwnd_limited:%llums(%.1f%%)",
2038 s->rwnd_limited / 1000,
2039 100.0 * s->rwnd_limited / s->busy_time);
2040 if (s->sndbuf_limited)
2041 printf(" sndbuf_limited:%llums(%.1f%%)",
2042 s->sndbuf_limited / 1000,
2043 100.0 * s->sndbuf_limited / s->busy_time);
2044 }
2045
2046 if (s->unacked)
2047 printf(" unacked:%u", s->unacked);
2048 if (s->retrans || s->retrans_total)
2049 printf(" retrans:%u/%u", s->retrans, s->retrans_total);
2050 if (s->lost)
2051 printf(" lost:%u", s->lost);
2052 if (s->sacked && s->ss.state != SS_LISTEN)
2053 printf(" sacked:%u", s->sacked);
2054 if (s->fackets)
2055 printf(" fackets:%u", s->fackets);
2056 if (s->reordering != 3)
2057 printf(" reordering:%d", s->reordering);
2058 if (s->rcv_rtt)
2059 printf(" rcv_rtt:%g", s->rcv_rtt);
2060 if (s->rcv_space)
2061 printf(" rcv_space:%d", s->rcv_space);
2062 if (s->rcv_ssthresh)
2063 printf(" rcv_ssthresh:%u", s->rcv_ssthresh);
2064 if (s->not_sent)
2065 printf(" notsent:%u", s->not_sent);
2066 if (s->min_rtt)
2067 printf(" minrtt:%g", s->min_rtt);
2068 }
2069
2070 static void tcp_timer_print(struct tcpstat *s)
2071 {
2072 static const char * const tmr_name[] = {
2073 "off",
2074 "on",
2075 "keepalive",
2076 "timewait",
2077 "persist",
2078 "unknown"
2079 };
2080
2081 if (s->timer) {
2082 if (s->timer > 4)
2083 s->timer = 5;
2084 printf(" timer:(%s,%s,%d)",
2085 tmr_name[s->timer],
2086 print_ms_timer(s->timeout),
2087 s->retrans);
2088 }
2089 }
2090
2091 static void sctp_timer_print(struct tcpstat *s)
2092 {
2093 if (s->timer)
2094 printf(" timer:(T3_RTX,%s,%d)",
2095 print_ms_timer(s->timeout), s->retrans);
2096 }
2097
2098 static int tcp_show_line(char *line, const struct filter *f, int family)
2099 {
2100 int rto = 0, ato = 0;
2101 struct tcpstat s = {};
2102 char *loc, *rem, *data;
2103 char opt[256];
2104 int n;
2105 int hz = get_user_hz();
2106
2107 if (proc_inet_split_line(line, &loc, &rem, &data))
2108 return -1;
2109
2110 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2111
2112 if (!(f->states & (1 << state)))
2113 return 0;
2114
2115 proc_parse_inet_addr(loc, rem, family, &s.ss);
2116
2117 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
2118 return 0;
2119
2120 opt[0] = 0;
2121 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %u %d %[^\n]\n",
2122 &s.ss.state, &s.ss.wq, &s.ss.rq,
2123 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
2124 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
2125 &s.ssthresh, opt);
2126
2127 if (n < 17)
2128 opt[0] = 0;
2129
2130 if (n < 12) {
2131 rto = 0;
2132 s.cwnd = 2;
2133 s.ssthresh = -1;
2134 ato = s.qack = 0;
2135 }
2136
2137 s.retrans = s.timer != 1 ? s.probes : s.retrans;
2138 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
2139 s.ato = (double)ato / hz;
2140 s.qack /= 2;
2141 s.rto = (double)rto;
2142 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
2143 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
2144 s.ss.type = IPPROTO_TCP;
2145
2146 inet_stats_print(&s.ss, false);
2147
2148 if (show_options)
2149 tcp_timer_print(&s);
2150
2151 if (show_details) {
2152 sock_details_print(&s.ss);
2153 if (opt[0])
2154 printf(" opt:\"%s\"", opt);
2155 }
2156
2157 if (show_tcpinfo)
2158 tcp_stats_print(&s);
2159
2160 printf("\n");
2161 return 0;
2162 }
2163
2164 static int generic_record_read(FILE *fp,
2165 int (*worker)(char*, const struct filter *, int),
2166 const struct filter *f, int fam)
2167 {
2168 char line[256];
2169
2170 /* skip header */
2171 if (fgets(line, sizeof(line), fp) == NULL)
2172 goto outerr;
2173
2174 while (fgets(line, sizeof(line), fp) != NULL) {
2175 int n = strlen(line);
2176
2177 if (n == 0 || line[n-1] != '\n') {
2178 errno = -EINVAL;
2179 return -1;
2180 }
2181 line[n-1] = 0;
2182
2183 if (worker(line, f, fam) < 0)
2184 return 0;
2185 }
2186 outerr:
2187
2188 return ferror(fp) ? -1 : 0;
2189 }
2190
2191 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
2192 {
2193 const __u32 *skmeminfo;
2194
2195 if (!tb[attrtype]) {
2196 if (attrtype == INET_DIAG_SKMEMINFO) {
2197 if (!tb[INET_DIAG_MEMINFO])
2198 return;
2199
2200 const struct inet_diag_meminfo *minfo =
2201 RTA_DATA(tb[INET_DIAG_MEMINFO]);
2202
2203 printf(" mem:(r%u,w%u,f%u,t%u)",
2204 minfo->idiag_rmem,
2205 minfo->idiag_wmem,
2206 minfo->idiag_fmem,
2207 minfo->idiag_tmem);
2208 }
2209 return;
2210 }
2211
2212 skmeminfo = RTA_DATA(tb[attrtype]);
2213
2214 printf(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
2215 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
2216 skmeminfo[SK_MEMINFO_RCVBUF],
2217 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
2218 skmeminfo[SK_MEMINFO_SNDBUF],
2219 skmeminfo[SK_MEMINFO_FWD_ALLOC],
2220 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
2221 skmeminfo[SK_MEMINFO_OPTMEM]);
2222
2223 if (RTA_PAYLOAD(tb[attrtype]) >=
2224 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
2225 printf(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
2226
2227 if (RTA_PAYLOAD(tb[attrtype]) >=
2228 (SK_MEMINFO_DROPS + 1) * sizeof(__u32))
2229 printf(",d%u", skmeminfo[SK_MEMINFO_DROPS]);
2230
2231 printf(")");
2232 }
2233
2234 static void print_md5sig(struct tcp_diag_md5sig *sig)
2235 {
2236 printf("%s/%d=",
2237 format_host(sig->tcpm_family,
2238 sig->tcpm_family == AF_INET6 ? 16 : 4,
2239 &sig->tcpm_addr),
2240 sig->tcpm_prefixlen);
2241 print_escape_buf(sig->tcpm_key, sig->tcpm_keylen, " ,");
2242 }
2243
2244 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
2245
2246 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2247 struct rtattr *tb[])
2248 {
2249 double rtt = 0;
2250 struct tcpstat s = {};
2251
2252 s.ss.state = r->idiag_state;
2253
2254 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2255
2256 if (tb[INET_DIAG_INFO]) {
2257 struct tcp_info *info;
2258 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2259
2260 /* workaround for older kernels with less fields */
2261 if (len < sizeof(*info)) {
2262 info = alloca(sizeof(*info));
2263 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2264 memset((char *)info + len, 0, sizeof(*info) - len);
2265 } else
2266 info = RTA_DATA(tb[INET_DIAG_INFO]);
2267
2268 if (show_options) {
2269 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
2270 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
2271 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
2272 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
2273 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
2274 }
2275
2276 if (tb[INET_DIAG_CONG])
2277 strncpy(s.cong_alg,
2278 rta_getattr_str(tb[INET_DIAG_CONG]),
2279 sizeof(s.cong_alg) - 1);
2280
2281 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
2282 s.has_wscale_opt = true;
2283 s.snd_wscale = info->tcpi_snd_wscale;
2284 s.rcv_wscale = info->tcpi_rcv_wscale;
2285 }
2286
2287 if (info->tcpi_rto && info->tcpi_rto != 3000000)
2288 s.rto = (double)info->tcpi_rto / 1000;
2289
2290 s.backoff = info->tcpi_backoff;
2291 s.rtt = (double)info->tcpi_rtt / 1000;
2292 s.rttvar = (double)info->tcpi_rttvar / 1000;
2293 s.ato = (double)info->tcpi_ato / 1000;
2294 s.mss = info->tcpi_snd_mss;
2295 s.rcv_mss = info->tcpi_rcv_mss;
2296 s.advmss = info->tcpi_advmss;
2297 s.rcv_space = info->tcpi_rcv_space;
2298 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
2299 s.lastsnd = info->tcpi_last_data_sent;
2300 s.lastrcv = info->tcpi_last_data_recv;
2301 s.lastack = info->tcpi_last_ack_recv;
2302 s.unacked = info->tcpi_unacked;
2303 s.retrans = info->tcpi_retrans;
2304 s.retrans_total = info->tcpi_total_retrans;
2305 s.lost = info->tcpi_lost;
2306 s.sacked = info->tcpi_sacked;
2307 s.fackets = info->tcpi_fackets;
2308 s.reordering = info->tcpi_reordering;
2309 s.rcv_space = info->tcpi_rcv_space;
2310 s.rcv_ssthresh = info->tcpi_rcv_ssthresh;
2311 s.cwnd = info->tcpi_snd_cwnd;
2312
2313 if (info->tcpi_snd_ssthresh < 0xFFFF)
2314 s.ssthresh = info->tcpi_snd_ssthresh;
2315
2316 rtt = (double) info->tcpi_rtt;
2317 if (tb[INET_DIAG_VEGASINFO]) {
2318 const struct tcpvegas_info *vinfo
2319 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
2320
2321 if (vinfo->tcpv_enabled &&
2322 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
2323 rtt = vinfo->tcpv_rtt;
2324 }
2325
2326 if (tb[INET_DIAG_DCTCPINFO]) {
2327 struct dctcpstat *dctcp = malloc(sizeof(struct
2328 dctcpstat));
2329
2330 const struct tcp_dctcp_info *dinfo
2331 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
2332
2333 dctcp->enabled = !!dinfo->dctcp_enabled;
2334 dctcp->ce_state = dinfo->dctcp_ce_state;
2335 dctcp->alpha = dinfo->dctcp_alpha;
2336 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
2337 dctcp->ab_tot = dinfo->dctcp_ab_tot;
2338 s.dctcp = dctcp;
2339 }
2340
2341 if (tb[INET_DIAG_BBRINFO]) {
2342 const void *bbr_info = RTA_DATA(tb[INET_DIAG_BBRINFO]);
2343 int len = min(RTA_PAYLOAD(tb[INET_DIAG_BBRINFO]),
2344 sizeof(*s.bbr_info));
2345
2346 s.bbr_info = calloc(1, sizeof(*s.bbr_info));
2347 if (s.bbr_info && bbr_info)
2348 memcpy(s.bbr_info, bbr_info, len);
2349 }
2350
2351 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
2352 s.send_bps = (double) info->tcpi_snd_cwnd *
2353 (double)info->tcpi_snd_mss * 8000000. / rtt;
2354 }
2355
2356 if (info->tcpi_pacing_rate &&
2357 info->tcpi_pacing_rate != ~0ULL) {
2358 s.pacing_rate = info->tcpi_pacing_rate * 8.;
2359
2360 if (info->tcpi_max_pacing_rate &&
2361 info->tcpi_max_pacing_rate != ~0ULL)
2362 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
2363 }
2364 s.bytes_acked = info->tcpi_bytes_acked;
2365 s.bytes_received = info->tcpi_bytes_received;
2366 s.segs_out = info->tcpi_segs_out;
2367 s.segs_in = info->tcpi_segs_in;
2368 s.data_segs_out = info->tcpi_data_segs_out;
2369 s.data_segs_in = info->tcpi_data_segs_in;
2370 s.not_sent = info->tcpi_notsent_bytes;
2371 if (info->tcpi_min_rtt && info->tcpi_min_rtt != ~0U)
2372 s.min_rtt = (double) info->tcpi_min_rtt / 1000;
2373 s.delivery_rate = info->tcpi_delivery_rate * 8.;
2374 s.app_limited = info->tcpi_delivery_rate_app_limited;
2375 s.busy_time = info->tcpi_busy_time;
2376 s.rwnd_limited = info->tcpi_rwnd_limited;
2377 s.sndbuf_limited = info->tcpi_sndbuf_limited;
2378 tcp_stats_print(&s);
2379 free(s.dctcp);
2380 free(s.bbr_info);
2381 }
2382 if (tb[INET_DIAG_MD5SIG]) {
2383 struct tcp_diag_md5sig *sig = RTA_DATA(tb[INET_DIAG_MD5SIG]);
2384 int len = RTA_PAYLOAD(tb[INET_DIAG_MD5SIG]);
2385
2386 printf(" md5keys:");
2387 print_md5sig(sig++);
2388 for (len -= sizeof(*sig); len > 0; len -= sizeof(*sig)) {
2389 printf(",");
2390 print_md5sig(sig++);
2391 }
2392 }
2393 }
2394
2395 static const char *format_host_sa(struct sockaddr_storage *sa)
2396 {
2397 union {
2398 struct sockaddr_in sin;
2399 struct sockaddr_in6 sin6;
2400 } *saddr = (void *)sa;
2401
2402 switch (sa->ss_family) {
2403 case AF_INET:
2404 return format_host(AF_INET, 4, &saddr->sin.sin_addr);
2405 case AF_INET6:
2406 return format_host(AF_INET6, 16, &saddr->sin6.sin6_addr);
2407 default:
2408 return "";
2409 }
2410 }
2411
2412 static void sctp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2413 struct rtattr *tb[])
2414 {
2415 struct sockaddr_storage *sa;
2416 int len;
2417
2418 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2419
2420 if (tb[INET_DIAG_LOCALS]) {
2421 len = RTA_PAYLOAD(tb[INET_DIAG_LOCALS]);
2422 sa = RTA_DATA(tb[INET_DIAG_LOCALS]);
2423
2424 printf("locals:%s", format_host_sa(sa));
2425 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
2426 printf(",%s", format_host_sa(sa));
2427
2428 }
2429 if (tb[INET_DIAG_PEERS]) {
2430 len = RTA_PAYLOAD(tb[INET_DIAG_PEERS]);
2431 sa = RTA_DATA(tb[INET_DIAG_PEERS]);
2432
2433 printf(" peers:%s", format_host_sa(sa));
2434 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
2435 printf(",%s", format_host_sa(sa));
2436 }
2437 if (tb[INET_DIAG_INFO]) {
2438 struct sctp_info *info;
2439 len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2440
2441 /* workaround for older kernels with less fields */
2442 if (len < sizeof(*info)) {
2443 info = alloca(sizeof(*info));
2444 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2445 memset((char *)info + len, 0, sizeof(*info) - len);
2446 } else
2447 info = RTA_DATA(tb[INET_DIAG_INFO]);
2448
2449 sctp_stats_print(info);
2450 }
2451 }
2452
2453 static void parse_diag_msg(struct nlmsghdr *nlh, struct sockstat *s)
2454 {
2455 struct rtattr *tb[INET_DIAG_MAX+1];
2456 struct inet_diag_msg *r = NLMSG_DATA(nlh);
2457
2458 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
2459 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2460
2461 s->state = r->idiag_state;
2462 s->local.family = s->remote.family = r->idiag_family;
2463 s->lport = ntohs(r->id.idiag_sport);
2464 s->rport = ntohs(r->id.idiag_dport);
2465 s->wq = r->idiag_wqueue;
2466 s->rq = r->idiag_rqueue;
2467 s->ino = r->idiag_inode;
2468 s->uid = r->idiag_uid;
2469 s->iface = r->id.idiag_if;
2470 s->sk = cookie_sk_get(&r->id.idiag_cookie[0]);
2471
2472 s->mark = 0;
2473 if (tb[INET_DIAG_MARK])
2474 s->mark = rta_getattr_u32(tb[INET_DIAG_MARK]);
2475 if (tb[INET_DIAG_PROTOCOL])
2476 s->raw_prot = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
2477 else
2478 s->raw_prot = 0;
2479
2480 if (s->local.family == AF_INET)
2481 s->local.bytelen = s->remote.bytelen = 4;
2482 else
2483 s->local.bytelen = s->remote.bytelen = 16;
2484
2485 memcpy(s->local.data, r->id.idiag_src, s->local.bytelen);
2486 memcpy(s->remote.data, r->id.idiag_dst, s->local.bytelen);
2487 }
2488
2489 static int inet_show_sock(struct nlmsghdr *nlh,
2490 struct sockstat *s)
2491 {
2492 struct rtattr *tb[INET_DIAG_MAX+1];
2493 struct inet_diag_msg *r = NLMSG_DATA(nlh);
2494 unsigned char v6only = 0;
2495
2496 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
2497 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2498
2499 if (tb[INET_DIAG_PROTOCOL])
2500 s->type = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
2501
2502 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
2503 v6only = rta_getattr_u8(tb[INET_DIAG_SKV6ONLY]);
2504
2505 inet_stats_print(s, v6only);
2506
2507 if (show_options) {
2508 struct tcpstat t = {};
2509
2510 t.timer = r->idiag_timer;
2511 t.timeout = r->idiag_expires;
2512 t.retrans = r->idiag_retrans;
2513 if (s->type == IPPROTO_SCTP)
2514 sctp_timer_print(&t);
2515 else
2516 tcp_timer_print(&t);
2517 }
2518
2519 if (show_details) {
2520 sock_details_print(s);
2521 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
2522 printf(" v6only:%u", v6only);
2523
2524 if (tb[INET_DIAG_SHUTDOWN]) {
2525 unsigned char mask;
2526
2527 mask = rta_getattr_u8(tb[INET_DIAG_SHUTDOWN]);
2528 printf(" %c-%c", mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
2529 }
2530 }
2531
2532 if (show_mem || (show_tcpinfo && s->type != IPPROTO_UDP)) {
2533 printf("\n\t");
2534 if (s->type == IPPROTO_SCTP)
2535 sctp_show_info(nlh, r, tb);
2536 else
2537 tcp_show_info(nlh, r, tb);
2538 }
2539 sctp_ino = s->ino;
2540
2541 printf("\n");
2542 return 0;
2543 }
2544
2545 static int tcpdiag_send(int fd, int protocol, struct filter *f)
2546 {
2547 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
2548 struct {
2549 struct nlmsghdr nlh;
2550 struct inet_diag_req r;
2551 } req = {
2552 .nlh.nlmsg_len = sizeof(req),
2553 .nlh.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST,
2554 .nlh.nlmsg_seq = MAGIC_SEQ,
2555 .r.idiag_family = AF_INET,
2556 .r.idiag_states = f->states,
2557 };
2558 char *bc = NULL;
2559 int bclen;
2560 struct msghdr msg;
2561 struct rtattr rta;
2562 struct iovec iov[3];
2563 int iovlen = 1;
2564
2565 if (protocol == IPPROTO_UDP)
2566 return -1;
2567
2568 if (protocol == IPPROTO_TCP)
2569 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
2570 else
2571 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
2572 if (show_mem) {
2573 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
2574 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
2575 }
2576
2577 if (show_tcpinfo) {
2578 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
2579 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
2580 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
2581 }
2582
2583 iov[0] = (struct iovec){
2584 .iov_base = &req,
2585 .iov_len = sizeof(req)
2586 };
2587 if (f->f) {
2588 bclen = ssfilter_bytecompile(f->f, &bc);
2589 if (bclen) {
2590 rta.rta_type = INET_DIAG_REQ_BYTECODE;
2591 rta.rta_len = RTA_LENGTH(bclen);
2592 iov[1] = (struct iovec){ &rta, sizeof(rta) };
2593 iov[2] = (struct iovec){ bc, bclen };
2594 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
2595 iovlen = 3;
2596 }
2597 }
2598
2599 msg = (struct msghdr) {
2600 .msg_name = (void *)&nladdr,
2601 .msg_namelen = sizeof(nladdr),
2602 .msg_iov = iov,
2603 .msg_iovlen = iovlen,
2604 };
2605
2606 if (sendmsg(fd, &msg, 0) < 0) {
2607 close(fd);
2608 return -1;
2609 }
2610
2611 return 0;
2612 }
2613
2614 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
2615 {
2616 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
2617 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
2618 char *bc = NULL;
2619 int bclen;
2620 struct msghdr msg;
2621 struct rtattr rta;
2622 struct iovec iov[3];
2623 int iovlen = 1;
2624
2625 if (family == PF_UNSPEC)
2626 return tcpdiag_send(fd, protocol, f);
2627
2628 memset(&req.r, 0, sizeof(req.r));
2629 req.r.sdiag_family = family;
2630 req.r.sdiag_protocol = protocol;
2631 req.r.idiag_states = f->states;
2632 if (show_mem) {
2633 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
2634 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
2635 }
2636
2637 if (show_tcpinfo) {
2638 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
2639 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
2640 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
2641 }
2642
2643 iov[0] = (struct iovec){
2644 .iov_base = &req,
2645 .iov_len = sizeof(req)
2646 };
2647 if (f->f) {
2648 bclen = ssfilter_bytecompile(f->f, &bc);
2649 if (bclen) {
2650 rta.rta_type = INET_DIAG_REQ_BYTECODE;
2651 rta.rta_len = RTA_LENGTH(bclen);
2652 iov[1] = (struct iovec){ &rta, sizeof(rta) };
2653 iov[2] = (struct iovec){ bc, bclen };
2654 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
2655 iovlen = 3;
2656 }
2657 }
2658
2659 msg = (struct msghdr) {
2660 .msg_name = (void *)&nladdr,
2661 .msg_namelen = sizeof(nladdr),
2662 .msg_iov = iov,
2663 .msg_iovlen = iovlen,
2664 };
2665
2666 if (sendmsg(fd, &msg, 0) < 0) {
2667 close(fd);
2668 return -1;
2669 }
2670
2671 return 0;
2672 }
2673
2674 struct inet_diag_arg {
2675 struct filter *f;
2676 int protocol;
2677 struct rtnl_handle *rth;
2678 };
2679
2680 static int kill_inet_sock(struct nlmsghdr *h, void *arg, struct sockstat *s)
2681 {
2682 struct inet_diag_msg *d = NLMSG_DATA(h);
2683 struct inet_diag_arg *diag_arg = arg;
2684 struct rtnl_handle *rth = diag_arg->rth;
2685
2686 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
2687
2688 req.nlh.nlmsg_type = SOCK_DESTROY;
2689 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
2690 req.nlh.nlmsg_seq = ++rth->seq;
2691 req.r.sdiag_family = d->idiag_family;
2692 req.r.sdiag_protocol = diag_arg->protocol;
2693 req.r.id = d->id;
2694
2695 if (diag_arg->protocol == IPPROTO_RAW) {
2696 struct inet_diag_req_raw *raw = (void *)&req.r;
2697
2698 BUILD_BUG_ON(sizeof(req.r) != sizeof(*raw));
2699 raw->sdiag_raw_protocol = s->raw_prot;
2700 }
2701
2702 return rtnl_talk(rth, &req.nlh, NULL);
2703 }
2704
2705 static int show_one_inet_sock(const struct sockaddr_nl *addr,
2706 struct nlmsghdr *h, void *arg)
2707 {
2708 int err;
2709 struct inet_diag_arg *diag_arg = arg;
2710 struct inet_diag_msg *r = NLMSG_DATA(h);
2711 struct sockstat s = {};
2712
2713 if (!(diag_arg->f->families & FAMILY_MASK(r->idiag_family)))
2714 return 0;
2715
2716 parse_diag_msg(h, &s);
2717 s.type = diag_arg->protocol;
2718
2719 if (diag_arg->f->f && run_ssfilter(diag_arg->f->f, &s) == 0)
2720 return 0;
2721
2722 if (diag_arg->f->kill && kill_inet_sock(h, arg, &s) != 0) {
2723 if (errno == EOPNOTSUPP || errno == ENOENT) {
2724 /* Socket can't be closed, or is already closed. */
2725 return 0;
2726 } else {
2727 perror("SOCK_DESTROY answers");
2728 return -1;
2729 }
2730 }
2731
2732 err = inet_show_sock(h, &s);
2733 if (err < 0)
2734 return err;
2735
2736 return 0;
2737 }
2738
2739 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
2740 {
2741 int err = 0;
2742 struct rtnl_handle rth, rth2;
2743 int family = PF_INET;
2744 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
2745
2746 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
2747 return -1;
2748
2749 if (f->kill) {
2750 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
2751 rtnl_close(&rth);
2752 return -1;
2753 }
2754 arg.rth = &rth2;
2755 }
2756
2757 rth.dump = MAGIC_SEQ;
2758 rth.dump_fp = dump_fp;
2759 if (preferred_family == PF_INET6)
2760 family = PF_INET6;
2761
2762 again:
2763 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
2764 goto Exit;
2765
2766 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
2767 if (family != PF_UNSPEC) {
2768 family = PF_UNSPEC;
2769 goto again;
2770 }
2771 goto Exit;
2772 }
2773 if (family == PF_INET && preferred_family != PF_INET) {
2774 family = PF_INET6;
2775 goto again;
2776 }
2777
2778 Exit:
2779 rtnl_close(&rth);
2780 if (arg.rth)
2781 rtnl_close(arg.rth);
2782 return err;
2783 }
2784
2785 static int tcp_show_netlink_file(struct filter *f)
2786 {
2787 FILE *fp;
2788 char buf[16384];
2789 int err = -1;
2790
2791 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
2792 perror("fopen($TCPDIAG_FILE)");
2793 return err;
2794 }
2795
2796 while (1) {
2797 int status, err2;
2798 struct nlmsghdr *h = (struct nlmsghdr *)buf;
2799 struct sockstat s = {};
2800
2801 status = fread(buf, 1, sizeof(*h), fp);
2802 if (status < 0) {
2803 perror("Reading header from $TCPDIAG_FILE");
2804 break;
2805 }
2806 if (status != sizeof(*h)) {
2807 perror("Unexpected EOF reading $TCPDIAG_FILE");
2808 break;
2809 }
2810
2811 status = fread(h+1, 1, NLMSG_ALIGN(h->nlmsg_len-sizeof(*h)), fp);
2812
2813 if (status < 0) {
2814 perror("Reading $TCPDIAG_FILE");
2815 break;
2816 }
2817 if (status + sizeof(*h) < h->nlmsg_len) {
2818 perror("Unexpected EOF reading $TCPDIAG_FILE");
2819 break;
2820 }
2821
2822 /* The only legal exit point */
2823 if (h->nlmsg_type == NLMSG_DONE) {
2824 err = 0;
2825 break;
2826 }
2827
2828 if (h->nlmsg_type == NLMSG_ERROR) {
2829 struct nlmsgerr *err = (struct nlmsgerr *)NLMSG_DATA(h);
2830
2831 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
2832 fprintf(stderr, "ERROR truncated\n");
2833 } else {
2834 errno = -err->error;
2835 perror("TCPDIAG answered");
2836 }
2837 break;
2838 }
2839
2840 parse_diag_msg(h, &s);
2841 s.type = IPPROTO_TCP;
2842
2843 if (f && f->f && run_ssfilter(f->f, &s) == 0)
2844 continue;
2845
2846 err2 = inet_show_sock(h, &s);
2847 if (err2 < 0) {
2848 err = err2;
2849 break;
2850 }
2851 }
2852
2853 fclose(fp);
2854 return err;
2855 }
2856
2857 static int tcp_show(struct filter *f)
2858 {
2859 FILE *fp = NULL;
2860 char *buf = NULL;
2861 int bufsize = 64*1024;
2862
2863 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2864 return 0;
2865
2866 dg_proto = TCP_PROTO;
2867
2868 if (getenv("TCPDIAG_FILE"))
2869 return tcp_show_netlink_file(f);
2870
2871 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
2872 && inet_show_netlink(f, NULL, IPPROTO_TCP) == 0)
2873 return 0;
2874
2875 /* Sigh... We have to parse /proc/net/tcp... */
2876
2877
2878 /* Estimate amount of sockets and try to allocate
2879 * huge buffer to read all the table at one read.
2880 * Limit it by 16MB though. The assumption is: as soon as
2881 * kernel was able to hold information about N connections,
2882 * it is able to give us some memory for snapshot.
2883 */
2884 if (1) {
2885 get_slabstat(&slabstat);
2886
2887 int guess = slabstat.socks+slabstat.tcp_syns;
2888
2889 if (f->states&(1<<SS_TIME_WAIT))
2890 guess += slabstat.tcp_tws;
2891 if (guess > (16*1024*1024)/128)
2892 guess = (16*1024*1024)/128;
2893 guess *= 128;
2894 if (guess > bufsize)
2895 bufsize = guess;
2896 }
2897 while (bufsize >= 64*1024) {
2898 if ((buf = malloc(bufsize)) != NULL)
2899 break;
2900 bufsize /= 2;
2901 }
2902 if (buf == NULL) {
2903 errno = ENOMEM;
2904 return -1;
2905 }
2906
2907 if (f->families & FAMILY_MASK(AF_INET)) {
2908 if ((fp = net_tcp_open()) == NULL)
2909 goto outerr;
2910
2911 setbuffer(fp, buf, bufsize);
2912 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
2913 goto outerr;
2914 fclose(fp);
2915 }
2916
2917 if ((f->families & FAMILY_MASK(AF_INET6)) &&
2918 (fp = net_tcp6_open()) != NULL) {
2919 setbuffer(fp, buf, bufsize);
2920 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
2921 goto outerr;
2922 fclose(fp);
2923 }
2924
2925 free(buf);
2926 return 0;
2927
2928 outerr:
2929 do {
2930 int saved_errno = errno;
2931
2932 free(buf);
2933 if (fp)
2934 fclose(fp);
2935 errno = saved_errno;
2936 return -1;
2937 } while (0);
2938 }
2939
2940 static int dccp_show(struct filter *f)
2941 {
2942 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2943 return 0;
2944
2945 if (!getenv("PROC_NET_DCCP") && !getenv("PROC_ROOT")
2946 && inet_show_netlink(f, NULL, IPPROTO_DCCP) == 0)
2947 return 0;
2948
2949 return 0;
2950 }
2951
2952 static int sctp_show(struct filter *f)
2953 {
2954 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
2955 return 0;
2956
2957 if (!getenv("PROC_NET_SCTP") && !getenv("PROC_ROOT")
2958 && inet_show_netlink(f, NULL, IPPROTO_SCTP) == 0)
2959 return 0;
2960
2961 return 0;
2962 }
2963
2964 static int dgram_show_line(char *line, const struct filter *f, int family)
2965 {
2966 struct sockstat s = {};
2967 char *loc, *rem, *data;
2968 char opt[256];
2969 int n;
2970
2971 if (proc_inet_split_line(line, &loc, &rem, &data))
2972 return -1;
2973
2974 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2975
2976 if (!(f->states & (1 << state)))
2977 return 0;
2978
2979 proc_parse_inet_addr(loc, rem, family, &s);
2980
2981 if (f->f && run_ssfilter(f->f, &s) == 0)
2982 return 0;
2983
2984 opt[0] = 0;
2985 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
2986 &s.state, &s.wq, &s.rq,
2987 &s.uid, &s.ino,
2988 &s.refcnt, &s.sk, opt);
2989
2990 if (n < 9)
2991 opt[0] = 0;
2992
2993 s.type = dg_proto == UDP_PROTO ? IPPROTO_UDP : 0;
2994 inet_stats_print(&s, false);
2995
2996 if (show_details && opt[0])
2997 printf(" opt:\"%s\"", opt);
2998
2999 printf("\n");
3000 return 0;
3001 }
3002
3003 static int udp_show(struct filter *f)
3004 {
3005 FILE *fp = NULL;
3006
3007 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3008 return 0;
3009
3010 dg_proto = UDP_PROTO;
3011
3012 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
3013 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
3014 return 0;
3015
3016 if (f->families&FAMILY_MASK(AF_INET)) {
3017 if ((fp = net_udp_open()) == NULL)
3018 goto outerr;
3019 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3020 goto outerr;
3021 fclose(fp);
3022 }
3023
3024 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3025 (fp = net_udp6_open()) != NULL) {
3026 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3027 goto outerr;
3028 fclose(fp);
3029 }
3030 return 0;
3031
3032 outerr:
3033 do {
3034 int saved_errno = errno;
3035
3036 if (fp)
3037 fclose(fp);
3038 errno = saved_errno;
3039 return -1;
3040 } while (0);
3041 }
3042
3043 static int raw_show(struct filter *f)
3044 {
3045 FILE *fp = NULL;
3046
3047 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3048 return 0;
3049
3050 dg_proto = RAW_PROTO;
3051
3052 if (!getenv("PROC_NET_RAW") && !getenv("PROC_ROOT") &&
3053 inet_show_netlink(f, NULL, IPPROTO_RAW) == 0)
3054 return 0;
3055
3056 if (f->families&FAMILY_MASK(AF_INET)) {
3057 if ((fp = net_raw_open()) == NULL)
3058 goto outerr;
3059 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3060 goto outerr;
3061 fclose(fp);
3062 }
3063
3064 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3065 (fp = net_raw6_open()) != NULL) {
3066 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3067 goto outerr;
3068 fclose(fp);
3069 }
3070 return 0;
3071
3072 outerr:
3073 do {
3074 int saved_errno = errno;
3075
3076 if (fp)
3077 fclose(fp);
3078 errno = saved_errno;
3079 return -1;
3080 } while (0);
3081 }
3082
3083 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
3084
3085 static void unix_list_drop_first(struct sockstat **list)
3086 {
3087 struct sockstat *s = *list;
3088
3089 (*list) = (*list)->next;
3090 free(s->name);
3091 free(s);
3092 }
3093
3094 static bool unix_type_skip(struct sockstat *s, struct filter *f)
3095 {
3096 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
3097 return true;
3098 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
3099 return true;
3100 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
3101 return true;
3102 return false;
3103 }
3104
3105 static void unix_stats_print(struct sockstat *s, struct filter *f)
3106 {
3107 char port_name[30] = {};
3108
3109 sock_state_print(s);
3110
3111 sock_addr_print(s->name ?: "*", " ",
3112 int_to_str(s->lport, port_name), NULL);
3113 sock_addr_print(s->peer_name ?: "*", " ",
3114 int_to_str(s->rport, port_name), NULL);
3115
3116 proc_ctx_print(s);
3117 }
3118
3119 static int unix_show_sock(const struct sockaddr_nl *addr, struct nlmsghdr *nlh,
3120 void *arg)
3121 {
3122 struct filter *f = (struct filter *)arg;
3123 struct unix_diag_msg *r = NLMSG_DATA(nlh);
3124 struct rtattr *tb[UNIX_DIAG_MAX+1];
3125 char name[128];
3126 struct sockstat stat = { .name = "*", .peer_name = "*" };
3127
3128 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr *)(r+1),
3129 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3130
3131 stat.type = r->udiag_type;
3132 stat.state = r->udiag_state;
3133 stat.ino = stat.lport = r->udiag_ino;
3134 stat.local.family = stat.remote.family = AF_UNIX;
3135
3136 if (unix_type_skip(&stat, f))
3137 return 0;
3138
3139 if (tb[UNIX_DIAG_RQLEN]) {
3140 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
3141
3142 stat.rq = rql->udiag_rqueue;
3143 stat.wq = rql->udiag_wqueue;
3144 }
3145 if (tb[UNIX_DIAG_NAME]) {
3146 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
3147
3148 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
3149 name[len] = '\0';
3150 if (name[0] == '\0') {
3151 int i;
3152 for (i = 0; i < len; i++)
3153 if (name[i] == '\0')
3154 name[i] = '@';
3155 }
3156 stat.name = &name[0];
3157 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
3158 }
3159 if (tb[UNIX_DIAG_PEER])
3160 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
3161
3162 if (f->f && run_ssfilter(f->f, &stat) == 0)
3163 return 0;
3164
3165 unix_stats_print(&stat, f);
3166
3167 if (show_mem)
3168 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
3169 if (show_details) {
3170 if (tb[UNIX_DIAG_SHUTDOWN]) {
3171 unsigned char mask;
3172
3173 mask = rta_getattr_u8(tb[UNIX_DIAG_SHUTDOWN]);
3174 printf(" %c-%c", mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3175 }
3176 }
3177 printf("\n");
3178
3179 return 0;
3180 }
3181
3182 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
3183 size_t size, rtnl_filter_t show_one_sock)
3184 {
3185 int ret = -1;
3186 struct rtnl_handle rth;
3187
3188 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3189 return -1;
3190
3191 rth.dump = MAGIC_SEQ;
3192
3193 if (rtnl_send(&rth, req, size) < 0)
3194 goto Exit;
3195
3196 if (rtnl_dump_filter(&rth, show_one_sock, f))
3197 goto Exit;
3198
3199 ret = 0;
3200 Exit:
3201 rtnl_close(&rth);
3202 return ret;
3203 }
3204
3205 static int unix_show_netlink(struct filter *f)
3206 {
3207 DIAG_REQUEST(req, struct unix_diag_req r);
3208
3209 req.r.sdiag_family = AF_UNIX;
3210 req.r.udiag_states = f->states;
3211 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
3212 if (show_mem)
3213 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
3214
3215 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
3216 }
3217
3218 static int unix_show(struct filter *f)
3219 {
3220 FILE *fp;
3221 char buf[256];
3222 char name[128];
3223 int newformat = 0;
3224 int cnt;
3225 struct sockstat *list = NULL;
3226 const int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
3227 SS_ESTABLISHED, SS_CLOSING };
3228
3229 if (!filter_af_get(f, AF_UNIX))
3230 return 0;
3231
3232 if (!getenv("PROC_NET_UNIX") && !getenv("PROC_ROOT")
3233 && unix_show_netlink(f) == 0)
3234 return 0;
3235
3236 if ((fp = net_unix_open()) == NULL)
3237 return -1;
3238 if (!fgets(buf, sizeof(buf), fp)) {
3239 fclose(fp);
3240 return -1;
3241 }
3242
3243 if (memcmp(buf, "Peer", 4) == 0)
3244 newformat = 1;
3245 cnt = 0;
3246
3247 while (fgets(buf, sizeof(buf), fp)) {
3248 struct sockstat *u, **insp;
3249 int flags;
3250
3251 if (!(u = calloc(1, sizeof(*u))))
3252 break;
3253
3254 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
3255 &u->rport, &u->rq, &u->wq, &flags, &u->type,
3256 &u->state, &u->ino, name) < 8)
3257 name[0] = 0;
3258
3259 u->lport = u->ino;
3260 u->local.family = u->remote.family = AF_UNIX;
3261
3262 if (flags & (1 << 16)) {
3263 u->state = SS_LISTEN;
3264 } else if (u->state > 0 &&
3265 u->state <= ARRAY_SIZE(unix_state_map)) {
3266 u->state = unix_state_map[u->state-1];
3267 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
3268 u->state = SS_ESTABLISHED;
3269 }
3270 if (unix_type_skip(u, f) ||
3271 !(f->states & (1 << u->state))) {
3272 free(u);
3273 continue;
3274 }
3275
3276 if (!newformat) {
3277 u->rport = 0;
3278 u->rq = 0;
3279 u->wq = 0;
3280 }
3281
3282 if (name[0]) {
3283 u->name = strdup(name);
3284 if (!u->name) {
3285 free(u);
3286 break;
3287 }
3288 }
3289
3290 if (u->rport) {
3291 struct sockstat *p;
3292
3293 for (p = list; p; p = p->next) {
3294 if (u->rport == p->lport)
3295 break;
3296 }
3297 if (!p)
3298 u->peer_name = "?";
3299 else
3300 u->peer_name = p->name ? : "*";
3301 }
3302
3303 if (f->f) {
3304 struct sockstat st = {
3305 .local.family = AF_UNIX,
3306 .remote.family = AF_UNIX,
3307 };
3308
3309 memcpy(st.local.data, &u->name, sizeof(u->name));
3310 if (strcmp(u->peer_name, "*"))
3311 memcpy(st.remote.data, &u->peer_name,
3312 sizeof(u->peer_name));
3313 if (run_ssfilter(f->f, &st) == 0) {
3314 free(u->name);
3315 free(u);
3316 continue;
3317 }
3318 }
3319
3320 insp = &list;
3321 while (*insp) {
3322 if (u->type < (*insp)->type ||
3323 (u->type == (*insp)->type &&
3324 u->ino < (*insp)->ino))
3325 break;
3326 insp = &(*insp)->next;
3327 }
3328 u->next = *insp;
3329 *insp = u;
3330
3331 if (++cnt > MAX_UNIX_REMEMBER) {
3332 while (list) {
3333 unix_stats_print(list, f);
3334 printf("\n");
3335
3336 unix_list_drop_first(&list);
3337 }
3338 cnt = 0;
3339 }
3340 }
3341 fclose(fp);
3342 while (list) {
3343 unix_stats_print(list, f);
3344 printf("\n");
3345
3346 unix_list_drop_first(&list);
3347 }
3348
3349 return 0;
3350 }
3351
3352 static int packet_stats_print(struct sockstat *s, const struct filter *f)
3353 {
3354 const char *addr, *port;
3355 char ll_name[16];
3356
3357 s->local.family = s->remote.family = AF_PACKET;
3358
3359 if (f->f) {
3360 s->local.data[0] = s->prot;
3361 if (run_ssfilter(f->f, s) == 0)
3362 return 1;
3363 }
3364
3365 sock_state_print(s);
3366
3367 if (s->prot == 3)
3368 addr = "*";
3369 else
3370 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
3371
3372 if (s->iface == 0)
3373 port = "*";
3374 else
3375 port = xll_index_to_name(s->iface);
3376
3377 sock_addr_print(addr, ":", port, NULL);
3378 sock_addr_print("", "*", "", NULL);
3379
3380 proc_ctx_print(s);
3381
3382 if (show_details)
3383 sock_details_print(s);
3384
3385 return 0;
3386 }
3387
3388 static void packet_show_ring(struct packet_diag_ring *ring)
3389 {
3390 printf("blk_size:%d", ring->pdr_block_size);
3391 printf(",blk_nr:%d", ring->pdr_block_nr);
3392 printf(",frm_size:%d", ring->pdr_frame_size);
3393 printf(",frm_nr:%d", ring->pdr_frame_nr);
3394 printf(",tmo:%d", ring->pdr_retire_tmo);
3395 printf(",features:0x%x", ring->pdr_features);
3396 }
3397
3398 static int packet_show_sock(const struct sockaddr_nl *addr,
3399 struct nlmsghdr *nlh, void *arg)
3400 {
3401 const struct filter *f = arg;
3402 struct packet_diag_msg *r = NLMSG_DATA(nlh);
3403 struct packet_diag_info *pinfo = NULL;
3404 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
3405 struct rtattr *tb[PACKET_DIAG_MAX+1];
3406 struct sockstat stat = {};
3407 uint32_t fanout = 0;
3408 bool has_fanout = false;
3409
3410 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr *)(r+1),
3411 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3412
3413 /* use /proc/net/packet if all info are not available */
3414 if (!tb[PACKET_DIAG_MEMINFO])
3415 return -1;
3416
3417 stat.type = r->pdiag_type;
3418 stat.prot = r->pdiag_num;
3419 stat.ino = r->pdiag_ino;
3420 stat.state = SS_CLOSE;
3421 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
3422
3423 if (tb[PACKET_DIAG_MEMINFO]) {
3424 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
3425
3426 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
3427 }
3428
3429 if (tb[PACKET_DIAG_INFO]) {
3430 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
3431 stat.lport = stat.iface = pinfo->pdi_index;
3432 }
3433
3434 if (tb[PACKET_DIAG_UID])
3435 stat.uid = rta_getattr_u32(tb[PACKET_DIAG_UID]);
3436
3437 if (tb[PACKET_DIAG_RX_RING])
3438 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
3439
3440 if (tb[PACKET_DIAG_TX_RING])
3441 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
3442
3443 if (tb[PACKET_DIAG_FANOUT]) {
3444 has_fanout = true;
3445 fanout = rta_getattr_u32(tb[PACKET_DIAG_FANOUT]);
3446 }
3447
3448 if (packet_stats_print(&stat, f))
3449 return 0;
3450
3451 if (show_details) {
3452 if (pinfo) {
3453 printf("\n\tver:%d", pinfo->pdi_version);
3454 printf(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
3455 printf(" flags( ");
3456 if (pinfo->pdi_flags & PDI_RUNNING)
3457 printf("running");
3458 if (pinfo->pdi_flags & PDI_AUXDATA)
3459 printf(" auxdata");
3460 if (pinfo->pdi_flags & PDI_ORIGDEV)
3461 printf(" origdev");
3462 if (pinfo->pdi_flags & PDI_VNETHDR)
3463 printf(" vnethdr");
3464 if (pinfo->pdi_flags & PDI_LOSS)
3465 printf(" loss");
3466 if (!pinfo->pdi_flags)
3467 printf("0");
3468 printf(" )");
3469 }
3470 if (ring_rx) {
3471 printf("\n\tring_rx(");
3472 packet_show_ring(ring_rx);
3473 printf(")");
3474 }
3475 if (ring_tx) {
3476 printf("\n\tring_tx(");
3477 packet_show_ring(ring_tx);
3478 printf(")");
3479 }
3480 if (has_fanout) {
3481 uint16_t type = (fanout >> 16) & 0xffff;
3482
3483 printf("\n\tfanout(");
3484 printf("id:%d,", fanout & 0xffff);
3485 printf("type:");
3486
3487 if (type == 0)
3488 printf("hash");
3489 else if (type == 1)
3490 printf("lb");
3491 else if (type == 2)
3492 printf("cpu");
3493 else if (type == 3)
3494 printf("roll");
3495 else if (type == 4)
3496 printf("random");
3497 else if (type == 5)
3498 printf("qm");
3499 else
3500 printf("0x%x", type);
3501
3502 printf(")");
3503 }
3504 }
3505
3506 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
3507 struct sock_filter *fil =
3508 RTA_DATA(tb[PACKET_DIAG_FILTER]);
3509 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
3510 sizeof(struct sock_filter);
3511
3512 printf("\n\tbpf filter (%d): ", num);
3513 while (num) {
3514 printf(" 0x%02x %u %u %u,",
3515 fil->code, fil->jt, fil->jf, fil->k);
3516 num--;
3517 fil++;
3518 }
3519 }
3520 printf("\n");
3521 return 0;
3522 }
3523
3524 static int packet_show_netlink(struct filter *f)
3525 {
3526 DIAG_REQUEST(req, struct packet_diag_req r);
3527
3528 req.r.sdiag_family = AF_PACKET;
3529 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
3530 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
3531
3532 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
3533 }
3534
3535 static int packet_show_line(char *buf, const struct filter *f, int fam)
3536 {
3537 unsigned long long sk;
3538 struct sockstat stat = {};
3539 int type, prot, iface, state, rq, uid, ino;
3540
3541 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
3542 &sk,
3543 &type, &prot, &iface, &state,
3544 &rq, &uid, &ino);
3545
3546 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
3547 return 0;
3548 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
3549 return 0;
3550
3551 stat.type = type;
3552 stat.prot = prot;
3553 stat.lport = stat.iface = iface;
3554 stat.state = state;
3555 stat.rq = rq;
3556 stat.uid = uid;
3557 stat.ino = ino;
3558 stat.state = SS_CLOSE;
3559
3560 if (packet_stats_print(&stat, f))
3561 return 0;
3562
3563 printf("\n");
3564 return 0;
3565 }
3566
3567 static int packet_show(struct filter *f)
3568 {
3569 FILE *fp;
3570 int rc = 0;
3571
3572 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
3573 return 0;
3574
3575 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
3576 packet_show_netlink(f) == 0)
3577 return 0;
3578
3579 if ((fp = net_packet_open()) == NULL)
3580 return -1;
3581 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
3582 rc = -1;
3583
3584 fclose(fp);
3585 return rc;
3586 }
3587
3588 static int netlink_show_one(struct filter *f,
3589 int prot, int pid, unsigned int groups,
3590 int state, int dst_pid, unsigned int dst_group,
3591 int rq, int wq,
3592 unsigned long long sk, unsigned long long cb)
3593 {
3594 struct sockstat st = {
3595 .state = SS_CLOSE,
3596 .rq = rq,
3597 .wq = wq,
3598 .local.family = AF_NETLINK,
3599 .remote.family = AF_NETLINK,
3600 };
3601
3602 SPRINT_BUF(prot_buf) = {};
3603 const char *prot_name;
3604 char procname[64] = {};
3605
3606 if (f->f) {
3607 st.rport = -1;
3608 st.lport = pid;
3609 st.local.data[0] = prot;
3610 if (run_ssfilter(f->f, &st) == 0)
3611 return 1;
3612 }
3613
3614 sock_state_print(&st);
3615
3616 if (resolve_services)
3617 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
3618 else
3619 prot_name = int_to_str(prot, prot_buf);
3620
3621 if (pid == -1) {
3622 procname[0] = '*';
3623 } else if (resolve_services) {
3624 int done = 0;
3625
3626 if (!pid) {
3627 done = 1;
3628 strncpy(procname, "kernel", 6);
3629 } else if (pid > 0) {
3630 FILE *fp;
3631
3632 snprintf(procname, sizeof(procname), "%s/%d/stat",
3633 getenv("PROC_ROOT") ? : "/proc", pid);
3634 if ((fp = fopen(procname, "r")) != NULL) {
3635 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
3636 snprintf(procname+strlen(procname),
3637 sizeof(procname)-strlen(procname),
3638 "/%d", pid);
3639 done = 1;
3640 }
3641 fclose(fp);
3642 }
3643 }
3644 if (!done)
3645 int_to_str(pid, procname);
3646 } else {
3647 int_to_str(pid, procname);
3648 }
3649
3650 sock_addr_print(prot_name, ":", procname, NULL);
3651
3652 if (state == NETLINK_CONNECTED) {
3653 char dst_group_buf[30];
3654 char dst_pid_buf[30];
3655
3656 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
3657 int_to_str(dst_pid, dst_pid_buf), NULL);
3658 } else {
3659 sock_addr_print("", "*", "", NULL);
3660 }
3661
3662 char *pid_context = NULL;
3663
3664 if (show_proc_ctx) {
3665 /* The pid value will either be:
3666 * 0 if destination kernel - show kernel initial context.
3667 * A valid process pid - use getpidcon.
3668 * A unique value allocated by the kernel or netlink user
3669 * to the process - show context as "not available".
3670 */
3671 if (!pid)
3672 security_get_initial_context("kernel", &pid_context);
3673 else if (pid > 0)
3674 getpidcon(pid, &pid_context);
3675
3676 printf(" proc_ctx=%s", pid_context ? : "unavailable");
3677 free(pid_context);
3678 }
3679
3680 if (show_details) {
3681 printf(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
3682 }
3683 printf("\n");
3684
3685 return 0;
3686 }
3687
3688 static int netlink_show_sock(const struct sockaddr_nl *addr,
3689 struct nlmsghdr *nlh, void *arg)
3690 {
3691 struct filter *f = (struct filter *)arg;
3692 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
3693 struct rtattr *tb[NETLINK_DIAG_MAX+1];
3694 int rq = 0, wq = 0;
3695 unsigned long groups = 0;
3696
3697 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr *)(r+1),
3698 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3699
3700 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
3701 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
3702
3703 if (tb[NETLINK_DIAG_MEMINFO]) {
3704 const __u32 *skmeminfo;
3705
3706 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
3707
3708 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
3709 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
3710 }
3711
3712 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
3713 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
3714 rq, wq, 0, 0)) {
3715 return 0;
3716 }
3717
3718 if (show_mem) {
3719 printf("\t");
3720 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
3721 printf("\n");
3722 }
3723
3724 return 0;
3725 }
3726
3727 static int netlink_show_netlink(struct filter *f)
3728 {
3729 DIAG_REQUEST(req, struct netlink_diag_req r);
3730
3731 req.r.sdiag_family = AF_NETLINK;
3732 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
3733 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
3734
3735 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
3736 }
3737
3738 static int netlink_show(struct filter *f)
3739 {
3740 FILE *fp;
3741 char buf[256];
3742 int prot, pid;
3743 unsigned int groups;
3744 int rq, wq, rc;
3745 unsigned long long sk, cb;
3746
3747 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
3748 return 0;
3749
3750 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
3751 netlink_show_netlink(f) == 0)
3752 return 0;
3753
3754 if ((fp = net_netlink_open()) == NULL)
3755 return -1;
3756 if (!fgets(buf, sizeof(buf), fp)) {
3757 fclose(fp);
3758 return -1;
3759 }
3760
3761 while (fgets(buf, sizeof(buf), fp)) {
3762 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
3763 &sk,
3764 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
3765
3766 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
3767 }
3768
3769 fclose(fp);
3770 return 0;
3771 }
3772
3773 static bool vsock_type_skip(struct sockstat *s, struct filter *f)
3774 {
3775 if (s->type == SOCK_STREAM && !(f->dbs & (1 << VSOCK_ST_DB)))
3776 return true;
3777 if (s->type == SOCK_DGRAM && !(f->dbs & (1 << VSOCK_DG_DB)))
3778 return true;
3779 return false;
3780 }
3781
3782 static void vsock_addr_print(inet_prefix *a, __u32 port)
3783 {
3784 char cid_str[sizeof("4294967295")];
3785 char port_str[sizeof("4294967295")];
3786 __u32 cid;
3787
3788 memcpy(&cid, a->data, sizeof(cid));
3789
3790 if (cid == ~(__u32)0)
3791 snprintf(cid_str, sizeof(cid_str), "*");
3792 else
3793 snprintf(cid_str, sizeof(cid_str), "%u", cid);
3794
3795 if (port == ~(__u32)0)
3796 snprintf(port_str, sizeof(port_str), "*");
3797 else
3798 snprintf(port_str, sizeof(port_str), "%u", port);
3799
3800 sock_addr_print(cid_str, ":", port_str, NULL);
3801 }
3802
3803 static void vsock_stats_print(struct sockstat *s, struct filter *f)
3804 {
3805 sock_state_print(s);
3806
3807 vsock_addr_print(&s->local, s->lport);
3808 vsock_addr_print(&s->remote, s->rport);
3809
3810 proc_ctx_print(s);
3811
3812 printf("\n");
3813 }
3814
3815 static int vsock_show_sock(const struct sockaddr_nl *addr,
3816 struct nlmsghdr *nlh, void *arg)
3817 {
3818 struct filter *f = (struct filter *)arg;
3819 struct vsock_diag_msg *r = NLMSG_DATA(nlh);
3820 struct sockstat stat = {
3821 .type = r->vdiag_type,
3822 .lport = r->vdiag_src_port,
3823 .rport = r->vdiag_dst_port,
3824 .state = r->vdiag_state,
3825 .ino = r->vdiag_ino,
3826 };
3827
3828 vsock_set_inet_prefix(&stat.local, r->vdiag_src_cid);
3829 vsock_set_inet_prefix(&stat.remote, r->vdiag_dst_cid);
3830
3831 if (vsock_type_skip(&stat, f))
3832 return 0;
3833
3834 if (f->f && run_ssfilter(f->f, &stat) == 0)
3835 return 0;
3836
3837 vsock_stats_print(&stat, f);
3838
3839 return 0;
3840 }
3841
3842 static int vsock_show(struct filter *f)
3843 {
3844 DIAG_REQUEST(req, struct vsock_diag_req r);
3845
3846 if (!filter_af_get(f, AF_VSOCK))
3847 return 0;
3848
3849 req.r.sdiag_family = AF_VSOCK;
3850 req.r.vdiag_states = f->states;
3851
3852 return handle_netlink_request(f, &req.nlh, sizeof(req), vsock_show_sock);
3853 }
3854
3855 struct sock_diag_msg {
3856 __u8 sdiag_family;
3857 };
3858
3859 static int generic_show_sock(const struct sockaddr_nl *addr,
3860 struct nlmsghdr *nlh, void *arg)
3861 {
3862 struct sock_diag_msg *r = NLMSG_DATA(nlh);
3863 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
3864
3865 switch (r->sdiag_family) {
3866 case AF_INET:
3867 case AF_INET6:
3868 return show_one_inet_sock(addr, nlh, &inet_arg);
3869 case AF_UNIX:
3870 return unix_show_sock(addr, nlh, arg);
3871 case AF_PACKET:
3872 return packet_show_sock(addr, nlh, arg);
3873 case AF_NETLINK:
3874 return netlink_show_sock(addr, nlh, arg);
3875 case AF_VSOCK:
3876 return vsock_show_sock(addr, nlh, arg);
3877 default:
3878 return -1;
3879 }
3880 }
3881
3882 static int handle_follow_request(struct filter *f)
3883 {
3884 int ret = 0;
3885 int groups = 0;
3886 struct rtnl_handle rth;
3887
3888 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << TCP_DB))
3889 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
3890 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << UDP_DB))
3891 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
3892 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << TCP_DB))
3893 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
3894 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << UDP_DB))
3895 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
3896
3897 if (groups == 0)
3898 return -1;
3899
3900 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
3901 return -1;
3902
3903 rth.dump = 0;
3904 rth.local.nl_pid = 0;
3905
3906 if (rtnl_dump_filter(&rth, generic_show_sock, f))
3907 ret = -1;
3908
3909 rtnl_close(&rth);
3910 return ret;
3911 }
3912
3913 static int get_snmp_int(char *proto, char *key, int *result)
3914 {
3915 char buf[1024];
3916 FILE *fp;
3917 int protolen = strlen(proto);
3918 int keylen = strlen(key);
3919
3920 *result = 0;
3921
3922 if ((fp = net_snmp_open()) == NULL)
3923 return -1;
3924
3925 while (fgets(buf, sizeof(buf), fp) != NULL) {
3926 char *p = buf;
3927 int pos = 0;
3928
3929 if (memcmp(buf, proto, protolen))
3930 continue;
3931 while ((p = strchr(p, ' ')) != NULL) {
3932 pos++;
3933 p++;
3934 if (memcmp(p, key, keylen) == 0 &&
3935 (p[keylen] == ' ' || p[keylen] == '\n'))
3936 break;
3937 }
3938 if (fgets(buf, sizeof(buf), fp) == NULL)
3939 break;
3940 if (memcmp(buf, proto, protolen))
3941 break;
3942 p = buf;
3943 while ((p = strchr(p, ' ')) != NULL) {
3944 p++;
3945 if (--pos == 0) {
3946 sscanf(p, "%d", result);
3947 fclose(fp);
3948 return 0;
3949 }
3950 }
3951 }
3952
3953 fclose(fp);
3954 errno = ESRCH;
3955 return -1;
3956 }
3957
3958
3959 /* Get stats from sockstat */
3960
3961 struct ssummary {
3962 int socks;
3963 int tcp_mem;
3964 int tcp_total;
3965 int tcp_orphans;
3966 int tcp_tws;
3967 int tcp4_hashed;
3968 int udp4;
3969 int raw4;
3970 int frag4;
3971 int frag4_mem;
3972 int tcp6_hashed;
3973 int udp6;
3974 int raw6;
3975 int frag6;
3976 int frag6_mem;
3977 };
3978
3979 static void get_sockstat_line(char *line, struct ssummary *s)
3980 {
3981 char id[256], rem[256];
3982
3983 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
3984 return;
3985
3986 if (strcmp(id, "sockets:") == 0)
3987 sscanf(rem, "%*s%d", &s->socks);
3988 else if (strcmp(id, "UDP:") == 0)
3989 sscanf(rem, "%*s%d", &s->udp4);
3990 else if (strcmp(id, "UDP6:") == 0)
3991 sscanf(rem, "%*s%d", &s->udp6);
3992 else if (strcmp(id, "RAW:") == 0)
3993 sscanf(rem, "%*s%d", &s->raw4);
3994 else if (strcmp(id, "RAW6:") == 0)
3995 sscanf(rem, "%*s%d", &s->raw6);
3996 else if (strcmp(id, "TCP6:") == 0)
3997 sscanf(rem, "%*s%d", &s->tcp6_hashed);
3998 else if (strcmp(id, "FRAG:") == 0)
3999 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
4000 else if (strcmp(id, "FRAG6:") == 0)
4001 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
4002 else if (strcmp(id, "TCP:") == 0)
4003 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
4004 &s->tcp4_hashed,
4005 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
4006 }
4007
4008 static int get_sockstat(struct ssummary *s)
4009 {
4010 char buf[256];
4011 FILE *fp;
4012
4013 memset(s, 0, sizeof(*s));
4014
4015 if ((fp = net_sockstat_open()) == NULL)
4016 return -1;
4017 while (fgets(buf, sizeof(buf), fp) != NULL)
4018 get_sockstat_line(buf, s);
4019 fclose(fp);
4020
4021 if ((fp = net_sockstat6_open()) == NULL)
4022 return 0;
4023 while (fgets(buf, sizeof(buf), fp) != NULL)
4024 get_sockstat_line(buf, s);
4025 fclose(fp);
4026
4027 return 0;
4028 }
4029
4030 static int print_summary(void)
4031 {
4032 struct ssummary s;
4033 int tcp_estab;
4034
4035 if (get_sockstat(&s) < 0)
4036 perror("ss: get_sockstat");
4037 if (get_snmp_int("Tcp:", "CurrEstab", &tcp_estab) < 0)
4038 perror("ss: get_snmpstat");
4039
4040 get_slabstat(&slabstat);
4041
4042 printf("Total: %d (kernel %d)\n", s.socks, slabstat.socks);
4043
4044 printf("TCP: %d (estab %d, closed %d, orphaned %d, synrecv %d, timewait %d/%d), ports %d\n",
4045 s.tcp_total + slabstat.tcp_syns + s.tcp_tws,
4046 tcp_estab,
4047 s.tcp_total - (s.tcp4_hashed+s.tcp6_hashed-s.tcp_tws),
4048 s.tcp_orphans,
4049 slabstat.tcp_syns,
4050 s.tcp_tws, slabstat.tcp_tws,
4051 slabstat.tcp_ports
4052 );
4053
4054 printf("\n");
4055 printf("Transport Total IP IPv6\n");
4056 printf("* %-9d %-9s %-9s\n", slabstat.socks, "-", "-");
4057 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
4058 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
4059 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
4060 printf("INET %-9d %-9d %-9d\n",
4061 s.raw4+s.udp4+s.tcp4_hashed+
4062 s.raw6+s.udp6+s.tcp6_hashed,
4063 s.raw4+s.udp4+s.tcp4_hashed,
4064 s.raw6+s.udp6+s.tcp6_hashed);
4065 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
4066
4067 printf("\n");
4068
4069 return 0;
4070 }
4071
4072 static void _usage(FILE *dest)
4073 {
4074 fprintf(dest,
4075 "Usage: ss [ OPTIONS ]\n"
4076 " ss [ OPTIONS ] [ FILTER ]\n"
4077 " -h, --help this message\n"
4078 " -V, --version output version information\n"
4079 " -n, --numeric don't resolve service names\n"
4080 " -r, --resolve resolve host names\n"
4081 " -a, --all display all sockets\n"
4082 " -l, --listening display listening sockets\n"
4083 " -o, --options show timer information\n"
4084 " -e, --extended show detailed socket information\n"
4085 " -m, --memory show socket memory usage\n"
4086 " -p, --processes show process using socket\n"
4087 " -i, --info show internal TCP information\n"
4088 " -s, --summary show socket usage summary\n"
4089 " -b, --bpf show bpf filter socket information\n"
4090 " -E, --events continually display sockets as they are destroyed\n"
4091 " -Z, --context display process SELinux security contexts\n"
4092 " -z, --contexts display process and socket SELinux security contexts\n"
4093 " -N, --net switch to the specified network namespace name\n"
4094 "\n"
4095 " -4, --ipv4 display only IP version 4 sockets\n"
4096 " -6, --ipv6 display only IP version 6 sockets\n"
4097 " -0, --packet display PACKET sockets\n"
4098 " -t, --tcp display only TCP sockets\n"
4099 " -S, --sctp display only SCTP sockets\n"
4100 " -u, --udp display only UDP sockets\n"
4101 " -d, --dccp display only DCCP sockets\n"
4102 " -w, --raw display only RAW sockets\n"
4103 " -x, --unix display only Unix domain sockets\n"
4104 " --vsock display only vsock sockets\n"
4105 " -f, --family=FAMILY display sockets of type FAMILY\n"
4106 " FAMILY := {inet|inet6|link|unix|netlink|vsock|help}\n"
4107 "\n"
4108 " -K, --kill forcibly close sockets, display what was closed\n"
4109 " -H, --no-header Suppress header line\n"
4110 "\n"
4111 " -A, --query=QUERY, --socket=QUERY\n"
4112 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink|vsock_stream|vsock_dgram}[,QUERY]\n"
4113 "\n"
4114 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
4115 " -F, --filter=FILE read filter information from FILE\n"
4116 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
4117 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
4118 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listening|closing}\n"
4119 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
4120 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
4121 " bucket := {syn-recv|time-wait}\n"
4122 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listening|closing}\n"
4123 );
4124 }
4125
4126 static void help(void) __attribute__((noreturn));
4127 static void help(void)
4128 {
4129 _usage(stdout);
4130 exit(0);
4131 }
4132
4133 static void usage(void) __attribute__((noreturn));
4134 static void usage(void)
4135 {
4136 _usage(stderr);
4137 exit(-1);
4138 }
4139
4140
4141 static int scan_state(const char *state)
4142 {
4143 static const char * const sstate_namel[] = {
4144 "UNKNOWN",
4145 [SS_ESTABLISHED] = "established",
4146 [SS_SYN_SENT] = "syn-sent",
4147 [SS_SYN_RECV] = "syn-recv",
4148 [SS_FIN_WAIT1] = "fin-wait-1",
4149 [SS_FIN_WAIT2] = "fin-wait-2",
4150 [SS_TIME_WAIT] = "time-wait",
4151 [SS_CLOSE] = "unconnected",
4152 [SS_CLOSE_WAIT] = "close-wait",
4153 [SS_LAST_ACK] = "last-ack",
4154 [SS_LISTEN] = "listening",
4155 [SS_CLOSING] = "closing",
4156 };
4157 int i;
4158
4159 if (strcasecmp(state, "close") == 0 ||
4160 strcasecmp(state, "closed") == 0)
4161 return (1<<SS_CLOSE);
4162 if (strcasecmp(state, "syn-rcv") == 0)
4163 return (1<<SS_SYN_RECV);
4164 if (strcasecmp(state, "established") == 0)
4165 return (1<<SS_ESTABLISHED);
4166 if (strcasecmp(state, "all") == 0)
4167 return SS_ALL;
4168 if (strcasecmp(state, "connected") == 0)
4169 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
4170 if (strcasecmp(state, "synchronized") == 0)
4171 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
4172 if (strcasecmp(state, "bucket") == 0)
4173 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
4174 if (strcasecmp(state, "big") == 0)
4175 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
4176 for (i = 0; i < SS_MAX; i++) {
4177 if (strcasecmp(state, sstate_namel[i]) == 0)
4178 return (1<<i);
4179 }
4180
4181 fprintf(stderr, "ss: wrong state name: %s\n", state);
4182 exit(-1);
4183 }
4184
4185 /* Values 'v' and 'V' are already used so a non-character is used */
4186 #define OPT_VSOCK 256
4187
4188 static const struct option long_opts[] = {
4189 { "numeric", 0, 0, 'n' },
4190 { "resolve", 0, 0, 'r' },
4191 { "options", 0, 0, 'o' },
4192 { "extended", 0, 0, 'e' },
4193 { "memory", 0, 0, 'm' },
4194 { "info", 0, 0, 'i' },
4195 { "processes", 0, 0, 'p' },
4196 { "bpf", 0, 0, 'b' },
4197 { "events", 0, 0, 'E' },
4198 { "dccp", 0, 0, 'd' },
4199 { "tcp", 0, 0, 't' },
4200 { "sctp", 0, 0, 'S' },
4201 { "udp", 0, 0, 'u' },
4202 { "raw", 0, 0, 'w' },
4203 { "unix", 0, 0, 'x' },
4204 { "vsock", 0, 0, OPT_VSOCK },
4205 { "all", 0, 0, 'a' },
4206 { "listening", 0, 0, 'l' },
4207 { "ipv4", 0, 0, '4' },
4208 { "ipv6", 0, 0, '6' },
4209 { "packet", 0, 0, '0' },
4210 { "family", 1, 0, 'f' },
4211 { "socket", 1, 0, 'A' },
4212 { "query", 1, 0, 'A' },
4213 { "summary", 0, 0, 's' },
4214 { "diag", 1, 0, 'D' },
4215 { "filter", 1, 0, 'F' },
4216 { "version", 0, 0, 'V' },
4217 { "help", 0, 0, 'h' },
4218 { "context", 0, 0, 'Z' },
4219 { "contexts", 0, 0, 'z' },
4220 { "net", 1, 0, 'N' },
4221 { "kill", 0, 0, 'K' },
4222 { "no-header", 0, 0, 'H' },
4223 { 0 }
4224
4225 };
4226
4227 int main(int argc, char *argv[])
4228 {
4229 int saw_states = 0;
4230 int saw_query = 0;
4231 int do_summary = 0;
4232 const char *dump_tcpdiag = NULL;
4233 FILE *filter_fp = NULL;
4234 int ch;
4235 int state_filter = 0;
4236 int addrp_width, screen_width = 80;
4237
4238 while ((ch = getopt_long(argc, argv,
4239 "dhaletuwxnro460spbEf:miA:D:F:vVzZN:KHS",
4240 long_opts, NULL)) != EOF) {
4241 switch (ch) {
4242 case 'n':
4243 resolve_services = 0;
4244 break;
4245 case 'r':
4246 resolve_hosts = 1;
4247 break;
4248 case 'o':
4249 show_options = 1;
4250 break;
4251 case 'e':
4252 show_options = 1;
4253 show_details++;
4254 break;
4255 case 'm':
4256 show_mem = 1;
4257 break;
4258 case 'i':
4259 show_tcpinfo = 1;
4260 break;
4261 case 'p':
4262 show_users++;
4263 user_ent_hash_build();
4264 break;
4265 case 'b':
4266 show_options = 1;
4267 show_bpf++;
4268 break;
4269 case 'E':
4270 follow_events = 1;
4271 break;
4272 case 'd':
4273 filter_db_set(&current_filter, DCCP_DB);
4274 break;
4275 case 't':
4276 filter_db_set(&current_filter, TCP_DB);
4277 break;
4278 case 'S':
4279 filter_db_set(&current_filter, SCTP_DB);
4280 break;
4281 case 'u':
4282 filter_db_set(&current_filter, UDP_DB);
4283 break;
4284 case 'w':
4285 filter_db_set(&current_filter, RAW_DB);
4286 break;
4287 case 'x':
4288 filter_af_set(&current_filter, AF_UNIX);
4289 break;
4290 case OPT_VSOCK:
4291 filter_af_set(&current_filter, AF_VSOCK);
4292 break;
4293 case 'a':
4294 state_filter = SS_ALL;
4295 break;
4296 case 'l':
4297 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
4298 break;
4299 case '4':
4300 filter_af_set(&current_filter, AF_INET);
4301 break;
4302 case '6':
4303 filter_af_set(&current_filter, AF_INET6);
4304 break;
4305 case '0':
4306 filter_af_set(&current_filter, AF_PACKET);
4307 break;
4308 case 'f':
4309 if (strcmp(optarg, "inet") == 0)
4310 filter_af_set(&current_filter, AF_INET);
4311 else if (strcmp(optarg, "inet6") == 0)
4312 filter_af_set(&current_filter, AF_INET6);
4313 else if (strcmp(optarg, "link") == 0)
4314 filter_af_set(&current_filter, AF_PACKET);
4315 else if (strcmp(optarg, "unix") == 0)
4316 filter_af_set(&current_filter, AF_UNIX);
4317 else if (strcmp(optarg, "netlink") == 0)
4318 filter_af_set(&current_filter, AF_NETLINK);
4319 else if (strcmp(optarg, "vsock") == 0)
4320 filter_af_set(&current_filter, AF_VSOCK);
4321 else if (strcmp(optarg, "help") == 0)
4322 help();
4323 else {
4324 fprintf(stderr, "ss: \"%s\" is invalid family\n",
4325 optarg);
4326 usage();
4327 }
4328 break;
4329 case 'A':
4330 {
4331 char *p, *p1;
4332
4333 if (!saw_query) {
4334 current_filter.dbs = 0;
4335 state_filter = state_filter ?
4336 state_filter : SS_CONN;
4337 saw_query = 1;
4338 do_default = 0;
4339 }
4340 p = p1 = optarg;
4341 do {
4342 if ((p1 = strchr(p, ',')) != NULL)
4343 *p1 = 0;
4344 if (strcmp(p, "all") == 0) {
4345 filter_default_dbs(&current_filter);
4346 } else if (strcmp(p, "inet") == 0) {
4347 filter_db_set(&current_filter, UDP_DB);
4348 filter_db_set(&current_filter, DCCP_DB);
4349 filter_db_set(&current_filter, TCP_DB);
4350 filter_db_set(&current_filter, SCTP_DB);
4351 filter_db_set(&current_filter, RAW_DB);
4352 } else if (strcmp(p, "udp") == 0) {
4353 filter_db_set(&current_filter, UDP_DB);
4354 } else if (strcmp(p, "dccp") == 0) {
4355 filter_db_set(&current_filter, DCCP_DB);
4356 } else if (strcmp(p, "tcp") == 0) {
4357 filter_db_set(&current_filter, TCP_DB);
4358 } else if (strcmp(p, "sctp") == 0) {
4359 filter_db_set(&current_filter, SCTP_DB);
4360 } else if (strcmp(p, "raw") == 0) {
4361 filter_db_set(&current_filter, RAW_DB);
4362 } else if (strcmp(p, "unix") == 0) {
4363 filter_db_set(&current_filter, UNIX_ST_DB);
4364 filter_db_set(&current_filter, UNIX_DG_DB);
4365 filter_db_set(&current_filter, UNIX_SQ_DB);
4366 } else if (strcasecmp(p, "unix_stream") == 0 ||
4367 strcmp(p, "u_str") == 0) {
4368 filter_db_set(&current_filter, UNIX_ST_DB);
4369 } else if (strcasecmp(p, "unix_dgram") == 0 ||
4370 strcmp(p, "u_dgr") == 0) {
4371 filter_db_set(&current_filter, UNIX_DG_DB);
4372 } else if (strcasecmp(p, "unix_seqpacket") == 0 ||
4373 strcmp(p, "u_seq") == 0) {
4374 filter_db_set(&current_filter, UNIX_SQ_DB);
4375 } else if (strcmp(p, "packet") == 0) {
4376 filter_db_set(&current_filter, PACKET_R_DB);
4377 filter_db_set(&current_filter, PACKET_DG_DB);
4378 } else if (strcmp(p, "packet_raw") == 0 ||
4379 strcmp(p, "p_raw") == 0) {
4380 filter_db_set(&current_filter, PACKET_R_DB);
4381 } else if (strcmp(p, "packet_dgram") == 0 ||
4382 strcmp(p, "p_dgr") == 0) {
4383 filter_db_set(&current_filter, PACKET_DG_DB);
4384 } else if (strcmp(p, "netlink") == 0) {
4385 filter_db_set(&current_filter, NETLINK_DB);
4386 } else if (strcmp(p, "vsock") == 0) {
4387 filter_db_set(&current_filter, VSOCK_ST_DB);
4388 filter_db_set(&current_filter, VSOCK_DG_DB);
4389 } else if (strcmp(p, "vsock_stream") == 0 ||
4390 strcmp(p, "v_str") == 0) {
4391 filter_db_set(&current_filter, VSOCK_ST_DB);
4392 } else if (strcmp(p, "vsock_dgram") == 0 ||
4393 strcmp(p, "v_dgr") == 0) {
4394 filter_db_set(&current_filter, VSOCK_DG_DB);
4395 } else {
4396 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
4397 usage();
4398 }
4399 p = p1 + 1;
4400 } while (p1);
4401 break;
4402 }
4403 case 's':
4404 do_summary = 1;
4405 break;
4406 case 'D':
4407 dump_tcpdiag = optarg;
4408 break;
4409 case 'F':
4410 if (filter_fp) {
4411 fprintf(stderr, "More than one filter file\n");
4412 exit(-1);
4413 }
4414 if (optarg[0] == '-')
4415 filter_fp = stdin;
4416 else
4417 filter_fp = fopen(optarg, "r");
4418 if (!filter_fp) {
4419 perror("fopen filter file");
4420 exit(-1);
4421 }
4422 break;
4423 case 'v':
4424 case 'V':
4425 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
4426 exit(0);
4427 case 'z':
4428 show_sock_ctx++;
4429 /* fall through */
4430 case 'Z':
4431 if (is_selinux_enabled() <= 0) {
4432 fprintf(stderr, "ss: SELinux is not enabled.\n");
4433 exit(1);
4434 }
4435 show_proc_ctx++;
4436 user_ent_hash_build();
4437 break;
4438 case 'N':
4439 if (netns_switch(optarg))
4440 exit(1);
4441 break;
4442 case 'K':
4443 current_filter.kill = 1;
4444 break;
4445 case 'H':
4446 show_header = 0;
4447 break;
4448 case 'h':
4449 help();
4450 case '?':
4451 default:
4452 usage();
4453 }
4454 }
4455
4456 argc -= optind;
4457 argv += optind;
4458
4459 if (do_summary) {
4460 print_summary();
4461 if (do_default && argc == 0)
4462 exit(0);
4463 }
4464
4465 while (argc > 0) {
4466 if (strcmp(*argv, "state") == 0) {
4467 NEXT_ARG();
4468 if (!saw_states)
4469 state_filter = 0;
4470 state_filter |= scan_state(*argv);
4471 saw_states = 1;
4472 } else if (strcmp(*argv, "exclude") == 0 ||
4473 strcmp(*argv, "excl") == 0) {
4474 NEXT_ARG();
4475 if (!saw_states)
4476 state_filter = SS_ALL;
4477 state_filter &= ~scan_state(*argv);
4478 saw_states = 1;
4479 } else {
4480 break;
4481 }
4482 argc--; argv++;
4483 }
4484
4485 if (do_default) {
4486 state_filter = state_filter ? state_filter : SS_CONN;
4487 filter_default_dbs(&current_filter);
4488 }
4489
4490 filter_states_set(&current_filter, state_filter);
4491 filter_merge_defaults(&current_filter);
4492
4493 if (resolve_services && resolve_hosts &&
4494 (current_filter.dbs & (UNIX_DBM|INET_L4_DBM)))
4495 init_service_resolver();
4496
4497 if (current_filter.dbs == 0) {
4498 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
4499 exit(0);
4500 }
4501 if (current_filter.families == 0) {
4502 fprintf(stderr, "ss: no families to show with such filter.\n");
4503 exit(0);
4504 }
4505 if (current_filter.states == 0) {
4506 fprintf(stderr, "ss: no socket states to show with such filter.\n");
4507 exit(0);
4508 }
4509
4510 if (dump_tcpdiag) {
4511 FILE *dump_fp = stdout;
4512
4513 if (!(current_filter.dbs & (1<<TCP_DB))) {
4514 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
4515 exit(0);
4516 }
4517 if (dump_tcpdiag[0] != '-') {
4518 dump_fp = fopen(dump_tcpdiag, "w");
4519 if (!dump_tcpdiag) {
4520 perror("fopen dump file");
4521 exit(-1);
4522 }
4523 }
4524 inet_show_netlink(&current_filter, dump_fp, IPPROTO_TCP);
4525 fflush(dump_fp);
4526 exit(0);
4527 }
4528
4529 if (ssfilter_parse(&current_filter.f, argc, argv, filter_fp))
4530 usage();
4531
4532 netid_width = 0;
4533 if (current_filter.dbs&(current_filter.dbs-1))
4534 netid_width = 5;
4535
4536 state_width = 0;
4537 if (current_filter.states&(current_filter.states-1))
4538 state_width = 10;
4539
4540 if (isatty(STDOUT_FILENO)) {
4541 struct winsize w;
4542
4543 if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) != -1) {
4544 if (w.ws_col > 0)
4545 screen_width = w.ws_col;
4546 }
4547 }
4548
4549 addrp_width = screen_width;
4550 if (netid_width)
4551 addrp_width -= netid_width + 1;
4552 if (state_width)
4553 addrp_width -= state_width + 1;
4554 addrp_width -= 14;
4555
4556 if (addrp_width&1) {
4557 if (netid_width)
4558 netid_width++;
4559 else if (state_width)
4560 state_width++;
4561 else
4562 odd_width_pad = " ";
4563 }
4564
4565 addrp_width /= 2;
4566 addrp_width--;
4567
4568 serv_width = resolve_services ? 7 : 5;
4569
4570 if (addrp_width < 15+serv_width+1)
4571 addrp_width = 15+serv_width+1;
4572
4573 addr_width = addrp_width - serv_width - 1;
4574
4575 if (show_header) {
4576 if (netid_width)
4577 printf("%-*s ", netid_width, "Netid");
4578 if (state_width)
4579 printf("%-*s ", state_width, "State");
4580 printf("%-6s %-6s %s", "Recv-Q", "Send-Q", odd_width_pad);
4581 }
4582
4583 /* Make enough space for the local/remote port field */
4584 addr_width -= 13;
4585 serv_width += 13;
4586
4587 if (show_header) {
4588 printf("%*s:%-*s %*s:%-*s\n",
4589 addr_width, "Local Address", serv_width, "Port",
4590 addr_width, "Peer Address", serv_width, "Port");
4591 }
4592
4593 fflush(stdout);
4594
4595 if (follow_events)
4596 exit(handle_follow_request(&current_filter));
4597
4598 if (current_filter.dbs & (1<<NETLINK_DB))
4599 netlink_show(&current_filter);
4600 if (current_filter.dbs & PACKET_DBM)
4601 packet_show(&current_filter);
4602 if (current_filter.dbs & UNIX_DBM)
4603 unix_show(&current_filter);
4604 if (current_filter.dbs & (1<<RAW_DB))
4605 raw_show(&current_filter);
4606 if (current_filter.dbs & (1<<UDP_DB))
4607 udp_show(&current_filter);
4608 if (current_filter.dbs & (1<<TCP_DB))
4609 tcp_show(&current_filter);
4610 if (current_filter.dbs & (1<<DCCP_DB))
4611 dccp_show(&current_filter);
4612 if (current_filter.dbs & (1<<SCTP_DB))
4613 sctp_show(&current_filter);
4614 if (current_filter.dbs & VSOCK_DBM)
4615 vsock_show(&current_filter);
4616
4617 if (show_users || show_proc_ctx || show_sock_ctx)
4618 user_ent_destroy();
4619
4620 return 0;
4621 }
Upstream mirror of iproute2