]> 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 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 <sys/sysmacros.h>
20 #include <netinet/in.h>
21 #include <string.h>
22 #include <errno.h>
23 #include <netdb.h>
24 #include <arpa/inet.h>
25 #include <dirent.h>
26 #include <fnmatch.h>
27 #include <getopt.h>
28 #include <stdbool.h>
29 #include <limits.h>
30 #include <stdarg.h>
31
32 #include "utils.h"
33 #include "rt_names.h"
34 #include "ll_map.h"
35 #include "libnetlink.h"
36 #include "namespace.h"
37 #include "SNAPSHOT.h"
38
39 #include <linux/tcp.h>
40 #include <linux/sock_diag.h>
41 #include <linux/inet_diag.h>
42 #include <linux/unix_diag.h>
43 #include <linux/netdevice.h> /* for MAX_ADDR_LEN */
44 #include <linux/filter.h>
45 #include <linux/xdp_diag.h>
46 #include <linux/packet_diag.h>
47 #include <linux/netlink_diag.h>
48 #include <linux/sctp.h>
49 #include <linux/vm_sockets_diag.h>
50 #include <linux/net.h>
51 #include <linux/tipc.h>
52 #include <linux/tipc_netlink.h>
53 #include <linux/tipc_sockets_diag.h>
54 #include <linux/tls.h>
55
56 /* AF_VSOCK/PF_VSOCK is only provided since glibc 2.18 */
57 #ifndef PF_VSOCK
58 #define PF_VSOCK 40
59 #endif
60 #ifndef AF_VSOCK
61 #define AF_VSOCK PF_VSOCK
62 #endif
63
64 #define MAGIC_SEQ 123456
65 #define BUF_CHUNK (1024 * 1024) /* Buffer chunk allocation size */
66 #define BUF_CHUNKS_MAX 5 /* Maximum number of allocated buffer chunks */
67 #define LEN_ALIGN(x) (((x) + 1) & ~1)
68
69 #define DIAG_REQUEST(_req, _r) \
70 struct { \
71 struct nlmsghdr nlh; \
72 _r; \
73 } _req = { \
74 .nlh = { \
75 .nlmsg_type = SOCK_DIAG_BY_FAMILY, \
76 .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,\
77 .nlmsg_seq = MAGIC_SEQ, \
78 .nlmsg_len = sizeof(_req), \
79 }, \
80 }
81
82 #if HAVE_SELINUX
83 #include <selinux/selinux.h>
84 #else
85 /* Stubs for SELinux functions */
86 static int is_selinux_enabled(void)
87 {
88 return -1;
89 }
90
91 static int getpidcon(pid_t pid, char **context)
92 {
93 *context = NULL;
94 return -1;
95 }
96
97 static int getfilecon(char *path, char **context)
98 {
99 *context = NULL;
100 return -1;
101 }
102
103 static int security_get_initial_context(char *name, char **context)
104 {
105 *context = NULL;
106 return -1;
107 }
108 #endif
109
110 int preferred_family = AF_UNSPEC;
111 static int show_options;
112 int show_details;
113 static int show_users;
114 static int show_mem;
115 static int show_tcpinfo;
116 static int show_bpf;
117 static int show_proc_ctx;
118 static int show_sock_ctx;
119 static int show_header = 1;
120 static int follow_events;
121 static int sctp_ino;
122 static int show_tipcinfo;
123 static int show_tos;
124 int numeric;
125 int oneline;
126
127 enum col_id {
128 COL_NETID,
129 COL_STATE,
130 COL_RECVQ,
131 COL_SENDQ,
132 COL_ADDR,
133 COL_SERV,
134 COL_RADDR,
135 COL_RSERV,
136 COL_EXT,
137 COL_MAX
138 };
139
140 enum col_align {
141 ALIGN_LEFT,
142 ALIGN_CENTER,
143 ALIGN_RIGHT
144 };
145
146 struct column {
147 const enum col_align align;
148 const char *header;
149 const char *ldelim;
150 int disabled;
151 int width; /* Calculated, including additional layout spacing */
152 int max_len; /* Measured maximum field length in this column */
153 };
154
155 static struct column columns[] = {
156 { ALIGN_LEFT, "Netid", "", 0, 0, 0 },
157 { ALIGN_LEFT, "State", " ", 0, 0, 0 },
158 { ALIGN_LEFT, "Recv-Q", " ", 0, 0, 0 },
159 { ALIGN_LEFT, "Send-Q", " ", 0, 0, 0 },
160 { ALIGN_RIGHT, "Local Address:", " ", 0, 0, 0 },
161 { ALIGN_LEFT, "Port", "", 0, 0, 0 },
162 { ALIGN_RIGHT, "Peer Address:", " ", 0, 0, 0 },
163 { ALIGN_LEFT, "Port", "", 0, 0, 0 },
164 { ALIGN_LEFT, "", "", 0, 0, 0 },
165 };
166
167 static struct column *current_field = columns;
168
169 /* Output buffer: chained chunks of BUF_CHUNK bytes. Each field is written to
170 * the buffer as a variable size token. A token consists of a 16 bits length
171 * field, followed by a string which is not NULL-terminated.
172 *
173 * A new chunk is allocated and linked when the current chunk doesn't have
174 * enough room to store the current token as a whole.
175 */
176 struct buf_chunk {
177 struct buf_chunk *next; /* Next chained chunk */
178 char *end; /* Current end of content */
179 char data[0];
180 };
181
182 struct buf_token {
183 uint16_t len; /* Data length, excluding length descriptor */
184 char data[0];
185 };
186
187 static struct {
188 struct buf_token *cur; /* Position of current token in chunk */
189 struct buf_chunk *head; /* First chunk */
190 struct buf_chunk *tail; /* Current chunk */
191 int chunks; /* Number of allocated chunks */
192 } buffer;
193
194 static const char *TCP_PROTO = "tcp";
195 static const char *SCTP_PROTO = "sctp";
196 static const char *UDP_PROTO = "udp";
197 static const char *RAW_PROTO = "raw";
198 static const char *dg_proto;
199
200 enum {
201 TCP_DB,
202 DCCP_DB,
203 UDP_DB,
204 RAW_DB,
205 UNIX_DG_DB,
206 UNIX_ST_DB,
207 UNIX_SQ_DB,
208 PACKET_DG_DB,
209 PACKET_R_DB,
210 NETLINK_DB,
211 SCTP_DB,
212 VSOCK_ST_DB,
213 VSOCK_DG_DB,
214 TIPC_DB,
215 XDP_DB,
216 MAX_DB
217 };
218
219 #define PACKET_DBM ((1<<PACKET_DG_DB)|(1<<PACKET_R_DB))
220 #define UNIX_DBM ((1<<UNIX_DG_DB)|(1<<UNIX_ST_DB)|(1<<UNIX_SQ_DB))
221 #define ALL_DB ((1<<MAX_DB)-1)
222 #define INET_L4_DBM ((1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB)|(1<<SCTP_DB))
223 #define INET_DBM (INET_L4_DBM | (1<<RAW_DB))
224 #define VSOCK_DBM ((1<<VSOCK_ST_DB)|(1<<VSOCK_DG_DB))
225
226 enum {
227 SS_UNKNOWN,
228 SS_ESTABLISHED,
229 SS_SYN_SENT,
230 SS_SYN_RECV,
231 SS_FIN_WAIT1,
232 SS_FIN_WAIT2,
233 SS_TIME_WAIT,
234 SS_CLOSE,
235 SS_CLOSE_WAIT,
236 SS_LAST_ACK,
237 SS_LISTEN,
238 SS_CLOSING,
239 SS_MAX
240 };
241
242 enum {
243 SCTP_STATE_CLOSED = 0,
244 SCTP_STATE_COOKIE_WAIT = 1,
245 SCTP_STATE_COOKIE_ECHOED = 2,
246 SCTP_STATE_ESTABLISHED = 3,
247 SCTP_STATE_SHUTDOWN_PENDING = 4,
248 SCTP_STATE_SHUTDOWN_SENT = 5,
249 SCTP_STATE_SHUTDOWN_RECEIVED = 6,
250 SCTP_STATE_SHUTDOWN_ACK_SENT = 7,
251 };
252
253 #define SS_ALL ((1 << SS_MAX) - 1)
254 #define SS_CONN (SS_ALL & ~((1<<SS_LISTEN)|(1<<SS_CLOSE)|(1<<SS_TIME_WAIT)|(1<<SS_SYN_RECV)))
255 #define TIPC_SS_CONN ((1<<SS_ESTABLISHED)|(1<<SS_LISTEN)|(1<<SS_CLOSE))
256
257 #include "ssfilter.h"
258
259 struct filter {
260 int dbs;
261 int states;
262 uint64_t families;
263 struct ssfilter *f;
264 bool kill;
265 struct rtnl_handle *rth_for_killing;
266 };
267
268 #define FAMILY_MASK(family) ((uint64_t)1 << (family))
269
270 static const struct filter default_dbs[MAX_DB] = {
271 [TCP_DB] = {
272 .states = SS_CONN,
273 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
274 },
275 [DCCP_DB] = {
276 .states = SS_CONN,
277 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
278 },
279 [UDP_DB] = {
280 .states = (1 << SS_ESTABLISHED),
281 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
282 },
283 [RAW_DB] = {
284 .states = (1 << SS_ESTABLISHED),
285 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
286 },
287 [UNIX_DG_DB] = {
288 .states = (1 << SS_CLOSE),
289 .families = FAMILY_MASK(AF_UNIX),
290 },
291 [UNIX_ST_DB] = {
292 .states = SS_CONN,
293 .families = FAMILY_MASK(AF_UNIX),
294 },
295 [UNIX_SQ_DB] = {
296 .states = SS_CONN,
297 .families = FAMILY_MASK(AF_UNIX),
298 },
299 [PACKET_DG_DB] = {
300 .states = (1 << SS_CLOSE),
301 .families = FAMILY_MASK(AF_PACKET),
302 },
303 [PACKET_R_DB] = {
304 .states = (1 << SS_CLOSE),
305 .families = FAMILY_MASK(AF_PACKET),
306 },
307 [NETLINK_DB] = {
308 .states = (1 << SS_CLOSE),
309 .families = FAMILY_MASK(AF_NETLINK),
310 },
311 [SCTP_DB] = {
312 .states = SS_CONN,
313 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
314 },
315 [VSOCK_ST_DB] = {
316 .states = SS_CONN,
317 .families = FAMILY_MASK(AF_VSOCK),
318 },
319 [VSOCK_DG_DB] = {
320 .states = SS_CONN,
321 .families = FAMILY_MASK(AF_VSOCK),
322 },
323 [TIPC_DB] = {
324 .states = TIPC_SS_CONN,
325 .families = FAMILY_MASK(AF_TIPC),
326 },
327 [XDP_DB] = {
328 .states = (1 << SS_CLOSE),
329 .families = FAMILY_MASK(AF_XDP),
330 },
331 };
332
333 static const struct filter default_afs[AF_MAX] = {
334 [AF_INET] = {
335 .dbs = INET_DBM,
336 .states = SS_CONN,
337 },
338 [AF_INET6] = {
339 .dbs = INET_DBM,
340 .states = SS_CONN,
341 },
342 [AF_UNIX] = {
343 .dbs = UNIX_DBM,
344 .states = SS_CONN,
345 },
346 [AF_PACKET] = {
347 .dbs = PACKET_DBM,
348 .states = (1 << SS_CLOSE),
349 },
350 [AF_NETLINK] = {
351 .dbs = (1 << NETLINK_DB),
352 .states = (1 << SS_CLOSE),
353 },
354 [AF_VSOCK] = {
355 .dbs = VSOCK_DBM,
356 .states = SS_CONN,
357 },
358 [AF_TIPC] = {
359 .dbs = (1 << TIPC_DB),
360 .states = TIPC_SS_CONN,
361 },
362 [AF_XDP] = {
363 .dbs = (1 << XDP_DB),
364 .states = (1 << SS_CLOSE),
365 },
366 };
367
368 static int do_default = 1;
369 static struct filter current_filter;
370
371 static void filter_db_set(struct filter *f, int db, bool enable)
372 {
373 if (enable) {
374 f->states |= default_dbs[db].states;
375 f->dbs |= 1 << db;
376 } else {
377 f->dbs &= ~(1 << db);
378 }
379 do_default = 0;
380 }
381
382 static int filter_db_parse(struct filter *f, const char *s)
383 {
384 const struct {
385 const char *name;
386 int dbs[MAX_DB + 1];
387 } db_name_tbl[] = {
388 #define ENTRY(name, ...) { #name, { __VA_ARGS__, MAX_DB } }
389 ENTRY(all, UDP_DB, DCCP_DB, TCP_DB, RAW_DB,
390 UNIX_ST_DB, UNIX_DG_DB, UNIX_SQ_DB,
391 PACKET_R_DB, PACKET_DG_DB, NETLINK_DB,
392 SCTP_DB, VSOCK_ST_DB, VSOCK_DG_DB, XDP_DB),
393 ENTRY(inet, UDP_DB, DCCP_DB, TCP_DB, SCTP_DB, RAW_DB),
394 ENTRY(udp, UDP_DB),
395 ENTRY(dccp, DCCP_DB),
396 ENTRY(tcp, TCP_DB),
397 ENTRY(sctp, SCTP_DB),
398 ENTRY(raw, RAW_DB),
399 ENTRY(unix, UNIX_ST_DB, UNIX_DG_DB, UNIX_SQ_DB),
400 ENTRY(unix_stream, UNIX_ST_DB),
401 ENTRY(u_str, UNIX_ST_DB), /* alias for unix_stream */
402 ENTRY(unix_dgram, UNIX_DG_DB),
403 ENTRY(u_dgr, UNIX_DG_DB), /* alias for unix_dgram */
404 ENTRY(unix_seqpacket, UNIX_SQ_DB),
405 ENTRY(u_seq, UNIX_SQ_DB), /* alias for unix_seqpacket */
406 ENTRY(packet, PACKET_R_DB, PACKET_DG_DB),
407 ENTRY(packet_raw, PACKET_R_DB),
408 ENTRY(p_raw, PACKET_R_DB), /* alias for packet_raw */
409 ENTRY(packet_dgram, PACKET_DG_DB),
410 ENTRY(p_dgr, PACKET_DG_DB), /* alias for packet_dgram */
411 ENTRY(netlink, NETLINK_DB),
412 ENTRY(vsock, VSOCK_ST_DB, VSOCK_DG_DB),
413 ENTRY(vsock_stream, VSOCK_ST_DB),
414 ENTRY(v_str, VSOCK_ST_DB), /* alias for vsock_stream */
415 ENTRY(vsock_dgram, VSOCK_DG_DB),
416 ENTRY(v_dgr, VSOCK_DG_DB), /* alias for vsock_dgram */
417 ENTRY(xdp, XDP_DB),
418 #undef ENTRY
419 };
420 bool enable = true;
421 unsigned int i;
422 const int *dbp;
423
424 if (s[0] == '!') {
425 enable = false;
426 s++;
427 }
428 for (i = 0; i < ARRAY_SIZE(db_name_tbl); i++) {
429 if (strcmp(s, db_name_tbl[i].name))
430 continue;
431 for (dbp = db_name_tbl[i].dbs; *dbp != MAX_DB; dbp++)
432 filter_db_set(f, *dbp, enable);
433 return 0;
434 }
435 return -1;
436 }
437
438 static void filter_af_set(struct filter *f, int af)
439 {
440 f->states |= default_afs[af].states;
441 f->families |= FAMILY_MASK(af);
442 do_default = 0;
443 preferred_family = af;
444 }
445
446 static int filter_af_get(struct filter *f, int af)
447 {
448 return !!(f->families & FAMILY_MASK(af));
449 }
450
451 static void filter_states_set(struct filter *f, int states)
452 {
453 if (states)
454 f->states = states;
455 }
456
457 static void filter_merge_defaults(struct filter *f)
458 {
459 int db;
460 int af;
461
462 for (db = 0; db < MAX_DB; db++) {
463 if (!(f->dbs & (1 << db)))
464 continue;
465
466 if (!(default_dbs[db].families & f->families))
467 f->families |= default_dbs[db].families;
468 }
469 for (af = 0; af < AF_MAX; af++) {
470 if (!(f->families & FAMILY_MASK(af)))
471 continue;
472
473 if (!(default_afs[af].dbs & f->dbs))
474 f->dbs |= default_afs[af].dbs;
475 }
476 }
477
478 static FILE *generic_proc_open(const char *env, const char *name)
479 {
480 const char *p = getenv(env);
481 char store[128];
482
483 if (!p) {
484 p = getenv("PROC_ROOT") ? : "/proc";
485 snprintf(store, sizeof(store)-1, "%s/%s", p, name);
486 p = store;
487 }
488
489 return fopen(p, "r");
490 }
491 #define net_tcp_open() generic_proc_open("PROC_NET_TCP", "net/tcp")
492 #define net_tcp6_open() generic_proc_open("PROC_NET_TCP6", "net/tcp6")
493 #define net_udp_open() generic_proc_open("PROC_NET_UDP", "net/udp")
494 #define net_udp6_open() generic_proc_open("PROC_NET_UDP6", "net/udp6")
495 #define net_raw_open() generic_proc_open("PROC_NET_RAW", "net/raw")
496 #define net_raw6_open() generic_proc_open("PROC_NET_RAW6", "net/raw6")
497 #define net_unix_open() generic_proc_open("PROC_NET_UNIX", "net/unix")
498 #define net_packet_open() generic_proc_open("PROC_NET_PACKET", \
499 "net/packet")
500 #define net_netlink_open() generic_proc_open("PROC_NET_NETLINK", \
501 "net/netlink")
502 #define net_sockstat_open() generic_proc_open("PROC_NET_SOCKSTAT", \
503 "net/sockstat")
504 #define net_sockstat6_open() generic_proc_open("PROC_NET_SOCKSTAT6", \
505 "net/sockstat6")
506 #define net_snmp_open() generic_proc_open("PROC_NET_SNMP", "net/snmp")
507 #define ephemeral_ports_open() generic_proc_open("PROC_IP_LOCAL_PORT_RANGE", \
508 "sys/net/ipv4/ip_local_port_range")
509
510 struct user_ent {
511 struct user_ent *next;
512 unsigned int ino;
513 int pid;
514 int fd;
515 char *process;
516 char *process_ctx;
517 char *socket_ctx;
518 };
519
520 #define USER_ENT_HASH_SIZE 256
521 static struct user_ent *user_ent_hash[USER_ENT_HASH_SIZE];
522
523 static int user_ent_hashfn(unsigned int ino)
524 {
525 int val = (ino >> 24) ^ (ino >> 16) ^ (ino >> 8) ^ ino;
526
527 return val & (USER_ENT_HASH_SIZE - 1);
528 }
529
530 static void user_ent_add(unsigned int ino, char *process,
531 int pid, int fd,
532 char *proc_ctx,
533 char *sock_ctx)
534 {
535 struct user_ent *p, **pp;
536
537 p = malloc(sizeof(struct user_ent));
538 if (!p) {
539 fprintf(stderr, "ss: failed to malloc buffer\n");
540 abort();
541 }
542 p->next = NULL;
543 p->ino = ino;
544 p->pid = pid;
545 p->fd = fd;
546 p->process = strdup(process);
547 p->process_ctx = strdup(proc_ctx);
548 p->socket_ctx = strdup(sock_ctx);
549
550 pp = &user_ent_hash[user_ent_hashfn(ino)];
551 p->next = *pp;
552 *pp = p;
553 }
554
555 static void user_ent_destroy(void)
556 {
557 struct user_ent *p, *p_next;
558 int cnt = 0;
559
560 while (cnt != USER_ENT_HASH_SIZE) {
561 p = user_ent_hash[cnt];
562 while (p) {
563 free(p->process);
564 free(p->process_ctx);
565 free(p->socket_ctx);
566 p_next = p->next;
567 free(p);
568 p = p_next;
569 }
570 cnt++;
571 }
572 }
573
574 static void user_ent_hash_build(void)
575 {
576 const char *root = getenv("PROC_ROOT") ? : "/proc/";
577 struct dirent *d;
578 char name[1024];
579 int nameoff;
580 DIR *dir;
581 char *pid_context;
582 char *sock_context;
583 const char *no_ctx = "unavailable";
584 static int user_ent_hash_build_init;
585
586 /* If show_users & show_proc_ctx set only do this once */
587 if (user_ent_hash_build_init != 0)
588 return;
589
590 user_ent_hash_build_init = 1;
591
592 strlcpy(name, root, sizeof(name));
593
594 if (strlen(name) == 0 || name[strlen(name)-1] != '/')
595 strcat(name, "/");
596
597 nameoff = strlen(name);
598
599 dir = opendir(name);
600 if (!dir)
601 return;
602
603 while ((d = readdir(dir)) != NULL) {
604 struct dirent *d1;
605 char process[16];
606 char *p;
607 int pid, pos;
608 DIR *dir1;
609 char crap;
610
611 if (sscanf(d->d_name, "%d%c", &pid, &crap) != 1)
612 continue;
613
614 if (getpidcon(pid, &pid_context) != 0)
615 pid_context = strdup(no_ctx);
616
617 snprintf(name + nameoff, sizeof(name) - nameoff, "%d/fd/", pid);
618 pos = strlen(name);
619 if ((dir1 = opendir(name)) == NULL) {
620 free(pid_context);
621 continue;
622 }
623
624 process[0] = '\0';
625 p = process;
626
627 while ((d1 = readdir(dir1)) != NULL) {
628 const char *pattern = "socket:[";
629 unsigned int ino;
630 char lnk[64];
631 int fd;
632 ssize_t link_len;
633 char tmp[1024];
634
635 if (sscanf(d1->d_name, "%d%c", &fd, &crap) != 1)
636 continue;
637
638 snprintf(name+pos, sizeof(name) - pos, "%d", fd);
639
640 link_len = readlink(name, lnk, sizeof(lnk)-1);
641 if (link_len == -1)
642 continue;
643 lnk[link_len] = '\0';
644
645 if (strncmp(lnk, pattern, strlen(pattern)))
646 continue;
647
648 sscanf(lnk, "socket:[%u]", &ino);
649
650 snprintf(tmp, sizeof(tmp), "%s/%d/fd/%s",
651 root, pid, d1->d_name);
652
653 if (getfilecon(tmp, &sock_context) <= 0)
654 sock_context = strdup(no_ctx);
655
656 if (*p == '\0') {
657 FILE *fp;
658
659 snprintf(tmp, sizeof(tmp), "%s/%d/stat",
660 root, pid);
661 if ((fp = fopen(tmp, "r")) != NULL) {
662 if (fscanf(fp, "%*d (%[^)])", p) < 1)
663 ; /* ignore */
664 fclose(fp);
665 }
666 }
667 user_ent_add(ino, p, pid, fd,
668 pid_context, sock_context);
669 free(sock_context);
670 }
671 free(pid_context);
672 closedir(dir1);
673 }
674 closedir(dir);
675 }
676
677 enum entry_types {
678 USERS,
679 PROC_CTX,
680 PROC_SOCK_CTX
681 };
682
683 #define ENTRY_BUF_SIZE 512
684 static int find_entry(unsigned int ino, char **buf, int type)
685 {
686 struct user_ent *p;
687 int cnt = 0;
688 char *ptr;
689 char *new_buf;
690 int len, new_buf_len;
691 int buf_used = 0;
692 int buf_len = 0;
693
694 if (!ino)
695 return 0;
696
697 p = user_ent_hash[user_ent_hashfn(ino)];
698 ptr = *buf = NULL;
699 while (p) {
700 if (p->ino != ino)
701 goto next;
702
703 while (1) {
704 ptr = *buf + buf_used;
705 switch (type) {
706 case USERS:
707 len = snprintf(ptr, buf_len - buf_used,
708 "(\"%s\",pid=%d,fd=%d),",
709 p->process, p->pid, p->fd);
710 break;
711 case PROC_CTX:
712 len = snprintf(ptr, buf_len - buf_used,
713 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d),",
714 p->process, p->pid,
715 p->process_ctx, p->fd);
716 break;
717 case PROC_SOCK_CTX:
718 len = snprintf(ptr, buf_len - buf_used,
719 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d,sock_ctx=%s),",
720 p->process, p->pid,
721 p->process_ctx, p->fd,
722 p->socket_ctx);
723 break;
724 default:
725 fprintf(stderr, "ss: invalid type: %d\n", type);
726 abort();
727 }
728
729 if (len < 0 || len >= buf_len - buf_used) {
730 new_buf_len = buf_len + ENTRY_BUF_SIZE;
731 new_buf = realloc(*buf, new_buf_len);
732 if (!new_buf) {
733 fprintf(stderr, "ss: failed to malloc buffer\n");
734 abort();
735 }
736 *buf = new_buf;
737 buf_len = new_buf_len;
738 continue;
739 } else {
740 buf_used += len;
741 break;
742 }
743 }
744 cnt++;
745 next:
746 p = p->next;
747 }
748 if (buf_used) {
749 ptr = *buf + buf_used;
750 ptr[-1] = '\0';
751 }
752 return cnt;
753 }
754
755 static unsigned long long cookie_sk_get(const uint32_t *cookie)
756 {
757 return (((unsigned long long)cookie[1] << 31) << 1) | cookie[0];
758 }
759
760 static const char *sctp_sstate_name[] = {
761 [SCTP_STATE_CLOSED] = "CLOSED",
762 [SCTP_STATE_COOKIE_WAIT] = "COOKIE_WAIT",
763 [SCTP_STATE_COOKIE_ECHOED] = "COOKIE_ECHOED",
764 [SCTP_STATE_ESTABLISHED] = "ESTAB",
765 [SCTP_STATE_SHUTDOWN_PENDING] = "SHUTDOWN_PENDING",
766 [SCTP_STATE_SHUTDOWN_SENT] = "SHUTDOWN_SENT",
767 [SCTP_STATE_SHUTDOWN_RECEIVED] = "SHUTDOWN_RECEIVED",
768 [SCTP_STATE_SHUTDOWN_ACK_SENT] = "ACK_SENT",
769 };
770
771 static const char * const stype_nameg[] = {
772 "UNKNOWN",
773 [SOCK_STREAM] = "STREAM",
774 [SOCK_DGRAM] = "DGRAM",
775 [SOCK_RDM] = "RDM",
776 [SOCK_SEQPACKET] = "SEQPACKET",
777 };
778
779 struct sockstat {
780 struct sockstat *next;
781 unsigned int type;
782 uint16_t prot;
783 uint16_t raw_prot;
784 inet_prefix local;
785 inet_prefix remote;
786 int lport;
787 int rport;
788 int state;
789 int rq, wq;
790 unsigned int ino;
791 unsigned int uid;
792 int refcnt;
793 unsigned int iface;
794 unsigned long long sk;
795 char *name;
796 char *peer_name;
797 __u32 mark;
798 };
799
800 struct dctcpstat {
801 unsigned int ce_state;
802 unsigned int alpha;
803 unsigned int ab_ecn;
804 unsigned int ab_tot;
805 bool enabled;
806 };
807
808 struct tcpstat {
809 struct sockstat ss;
810 unsigned int timer;
811 unsigned int timeout;
812 int probes;
813 char cong_alg[16];
814 double rto, ato, rtt, rttvar;
815 int qack, ssthresh, backoff;
816 double send_bps;
817 int snd_wscale;
818 int rcv_wscale;
819 int mss;
820 int rcv_mss;
821 int advmss;
822 unsigned int pmtu;
823 unsigned int cwnd;
824 unsigned int lastsnd;
825 unsigned int lastrcv;
826 unsigned int lastack;
827 double pacing_rate;
828 double pacing_rate_max;
829 double delivery_rate;
830 unsigned long long bytes_acked;
831 unsigned long long bytes_received;
832 unsigned int segs_out;
833 unsigned int segs_in;
834 unsigned int data_segs_out;
835 unsigned int data_segs_in;
836 unsigned int unacked;
837 unsigned int retrans;
838 unsigned int retrans_total;
839 unsigned int lost;
840 unsigned int sacked;
841 unsigned int fackets;
842 unsigned int reordering;
843 unsigned int not_sent;
844 unsigned int delivered;
845 unsigned int delivered_ce;
846 unsigned int dsack_dups;
847 unsigned int reord_seen;
848 double rcv_rtt;
849 double min_rtt;
850 int rcv_space;
851 unsigned int rcv_ssthresh;
852 unsigned long long busy_time;
853 unsigned long long rwnd_limited;
854 unsigned long long sndbuf_limited;
855 unsigned long long bytes_sent;
856 unsigned long long bytes_retrans;
857 bool has_ts_opt;
858 bool has_sack_opt;
859 bool has_ecn_opt;
860 bool has_ecnseen_opt;
861 bool has_fastopen_opt;
862 bool has_wscale_opt;
863 bool app_limited;
864 struct dctcpstat *dctcp;
865 struct tcp_bbr_info *bbr_info;
866 };
867
868 /* SCTP assocs share the same inode number with their parent endpoint. So if we
869 * have seen the inode number before, it must be an assoc instead of the next
870 * endpoint. */
871 static bool is_sctp_assoc(struct sockstat *s, const char *sock_name)
872 {
873 if (strcmp(sock_name, "sctp"))
874 return false;
875 if (!sctp_ino || sctp_ino != s->ino)
876 return false;
877 return true;
878 }
879
880 static const char *unix_netid_name(int type)
881 {
882 switch (type) {
883 case SOCK_STREAM:
884 return "u_str";
885 case SOCK_SEQPACKET:
886 return "u_seq";
887 case SOCK_DGRAM:
888 default:
889 return "u_dgr";
890 }
891 }
892
893 static const char *proto_name(int protocol)
894 {
895 switch (protocol) {
896 case 0:
897 return "raw";
898 case IPPROTO_UDP:
899 return "udp";
900 case IPPROTO_TCP:
901 return "tcp";
902 case IPPROTO_SCTP:
903 return "sctp";
904 case IPPROTO_DCCP:
905 return "dccp";
906 case IPPROTO_ICMPV6:
907 return "icmp6";
908 }
909
910 return "???";
911 }
912
913 static const char *vsock_netid_name(int type)
914 {
915 switch (type) {
916 case SOCK_STREAM:
917 return "v_str";
918 case SOCK_DGRAM:
919 return "v_dgr";
920 default:
921 return "???";
922 }
923 }
924
925 static const char *tipc_netid_name(int type)
926 {
927 switch (type) {
928 case SOCK_STREAM:
929 return "ti_st";
930 case SOCK_DGRAM:
931 return "ti_dg";
932 case SOCK_RDM:
933 return "ti_rd";
934 case SOCK_SEQPACKET:
935 return "ti_sq";
936 default:
937 return "???";
938 }
939 }
940
941 /* Allocate and initialize a new buffer chunk */
942 static struct buf_chunk *buf_chunk_new(void)
943 {
944 struct buf_chunk *new = malloc(BUF_CHUNK);
945
946 if (!new)
947 abort();
948
949 new->next = NULL;
950
951 /* This is also the last block */
952 buffer.tail = new;
953
954 /* Next token will be stored at the beginning of chunk data area, and
955 * its initial length is zero.
956 */
957 buffer.cur = (struct buf_token *)new->data;
958 buffer.cur->len = 0;
959
960 new->end = buffer.cur->data;
961
962 buffer.chunks++;
963
964 return new;
965 }
966
967 /* Return available tail room in given chunk */
968 static int buf_chunk_avail(struct buf_chunk *chunk)
969 {
970 return BUF_CHUNK - offsetof(struct buf_chunk, data) -
971 (chunk->end - chunk->data);
972 }
973
974 /* Update end pointer and token length, link new chunk if we hit the end of the
975 * current one. Return -EAGAIN if we got a new chunk, caller has to print again.
976 */
977 static int buf_update(int len)
978 {
979 struct buf_chunk *chunk = buffer.tail;
980 struct buf_token *t = buffer.cur;
981
982 /* Claim success if new content fits in the current chunk, and anyway
983 * if this is the first token in the chunk: in the latter case,
984 * allocating a new chunk won't help, so we'll just cut the output.
985 */
986 if ((len < buf_chunk_avail(chunk) && len != -1 /* glibc < 2.0.6 */) ||
987 t == (struct buf_token *)chunk->data) {
988 len = min(len, buf_chunk_avail(chunk));
989
990 /* Total field length can't exceed 2^16 bytes, cut as needed */
991 len = min(len, USHRT_MAX - t->len);
992
993 chunk->end += len;
994 t->len += len;
995 return 0;
996 }
997
998 /* Content truncated, time to allocate more */
999 chunk->next = buf_chunk_new();
1000
1001 /* Copy current token over to new chunk, including length descriptor */
1002 memcpy(chunk->next->data, t, sizeof(t->len) + t->len);
1003 chunk->next->end += t->len;
1004
1005 /* Discard partially written field in old chunk */
1006 chunk->end -= t->len + sizeof(t->len);
1007
1008 return -EAGAIN;
1009 }
1010
1011 /* Append content to buffer as part of the current field */
1012 __attribute__((format(printf, 1, 2)))
1013 static void out(const char *fmt, ...)
1014 {
1015 struct column *f = current_field;
1016 va_list args;
1017 char *pos;
1018 int len;
1019
1020 if (f->disabled)
1021 return;
1022
1023 if (!buffer.head)
1024 buffer.head = buf_chunk_new();
1025
1026 again: /* Append to buffer: if we have a new chunk, print again */
1027
1028 pos = buffer.cur->data + buffer.cur->len;
1029 va_start(args, fmt);
1030
1031 /* Limit to tail room. If we hit the limit, buf_update() will tell us */
1032 len = vsnprintf(pos, buf_chunk_avail(buffer.tail), fmt, args);
1033 va_end(args);
1034
1035 if (buf_update(len))
1036 goto again;
1037 }
1038
1039 static int print_left_spacing(struct column *f, int stored, int printed)
1040 {
1041 int s;
1042
1043 if (!f->width || f->align == ALIGN_LEFT)
1044 return 0;
1045
1046 s = f->width - stored - printed;
1047 if (f->align == ALIGN_CENTER)
1048 /* If count of total spacing is odd, shift right by one */
1049 s = (s + 1) / 2;
1050
1051 if (s > 0)
1052 return printf("%*c", s, ' ');
1053
1054 return 0;
1055 }
1056
1057 static void print_right_spacing(struct column *f, int printed)
1058 {
1059 int s;
1060
1061 if (!f->width || f->align == ALIGN_RIGHT)
1062 return;
1063
1064 s = f->width - printed;
1065 if (f->align == ALIGN_CENTER)
1066 s /= 2;
1067
1068 if (s > 0)
1069 printf("%*c", s, ' ');
1070 }
1071
1072 /* Done with field: update buffer pointer, start new token after current one */
1073 static void field_flush(struct column *f)
1074 {
1075 struct buf_chunk *chunk;
1076 unsigned int pad;
1077
1078 if (f->disabled)
1079 return;
1080
1081 chunk = buffer.tail;
1082 pad = buffer.cur->len % 2;
1083
1084 if (buffer.cur->len > f->max_len)
1085 f->max_len = buffer.cur->len;
1086
1087 /* We need a new chunk if we can't store the next length descriptor.
1088 * Mind the gap between end of previous token and next aligned position
1089 * for length descriptor.
1090 */
1091 if (buf_chunk_avail(chunk) - pad < sizeof(buffer.cur->len)) {
1092 chunk->end += pad;
1093 chunk->next = buf_chunk_new();
1094 return;
1095 }
1096
1097 buffer.cur = (struct buf_token *)(buffer.cur->data +
1098 LEN_ALIGN(buffer.cur->len));
1099 buffer.cur->len = 0;
1100 buffer.tail->end = buffer.cur->data;
1101 }
1102
1103 static int field_is_last(struct column *f)
1104 {
1105 return f - columns == COL_MAX - 1;
1106 }
1107
1108 /* Get the next available token in the buffer starting from the current token */
1109 static struct buf_token *buf_token_next(struct buf_token *cur)
1110 {
1111 struct buf_chunk *chunk = buffer.tail;
1112
1113 /* If we reached the end of chunk contents, get token from next chunk */
1114 if (cur->data + LEN_ALIGN(cur->len) == chunk->end) {
1115 buffer.tail = chunk = chunk->next;
1116 return chunk ? (struct buf_token *)chunk->data : NULL;
1117 }
1118
1119 return (struct buf_token *)(cur->data + LEN_ALIGN(cur->len));
1120 }
1121
1122 /* Free up all allocated buffer chunks */
1123 static void buf_free_all(void)
1124 {
1125 struct buf_chunk *tmp;
1126
1127 for (buffer.tail = buffer.head; buffer.tail; ) {
1128 tmp = buffer.tail;
1129 buffer.tail = buffer.tail->next;
1130 free(tmp);
1131 }
1132 buffer.head = NULL;
1133 buffer.chunks = 0;
1134 }
1135
1136 /* Get current screen width, default to 80 columns if TIOCGWINSZ fails */
1137 static int render_screen_width(void)
1138 {
1139 int width = 80;
1140
1141 if (isatty(STDOUT_FILENO)) {
1142 struct winsize w;
1143
1144 if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) != -1) {
1145 if (w.ws_col > 0)
1146 width = w.ws_col;
1147 }
1148 }
1149
1150 return width;
1151 }
1152
1153 /* Calculate column width from contents length. If columns don't fit on one
1154 * line, break them into the least possible amount of lines and keep them
1155 * aligned across lines. Available screen space is equally spread between fields
1156 * as additional spacing.
1157 */
1158 static void render_calc_width(void)
1159 {
1160 int screen_width = render_screen_width();
1161 struct column *c, *eol = columns - 1;
1162 int first, len = 0, linecols = 0;
1163
1164 /* First pass: set width for each column to measured content length */
1165 for (first = 1, c = columns; c - columns < COL_MAX; c++) {
1166 if (c->disabled)
1167 continue;
1168
1169 if (!first && c->max_len)
1170 c->width = c->max_len + strlen(c->ldelim);
1171 else
1172 c->width = c->max_len;
1173
1174 /* But don't exceed screen size. If we exceed the screen size
1175 * for even a single field, it will just start on a line of its
1176 * own and then naturally wrap.
1177 */
1178 c->width = min(c->width, screen_width);
1179
1180 if (c->width)
1181 first = 0;
1182 }
1183
1184 /* Second pass: find out newlines and distribute available spacing */
1185 for (c = columns; c - columns < COL_MAX; c++) {
1186 int pad, spacing, rem, last;
1187 struct column *tmp;
1188
1189 if (!c->width)
1190 continue;
1191
1192 linecols++;
1193 len += c->width;
1194
1195 for (last = 1, tmp = c + 1; tmp - columns < COL_MAX; tmp++) {
1196 if (tmp->width) {
1197 last = 0;
1198 break;
1199 }
1200 }
1201
1202 if (!last && len < screen_width) {
1203 /* Columns fit on screen so far, nothing to do yet */
1204 continue;
1205 }
1206
1207 if (len == screen_width) {
1208 /* Exact fit, just start with new line */
1209 goto newline;
1210 }
1211
1212 if (len > screen_width) {
1213 /* Screen width exceeded: go back one column */
1214 len -= c->width;
1215 c--;
1216 linecols--;
1217 }
1218
1219 /* Distribute remaining space to columns on this line */
1220 pad = screen_width - len;
1221 spacing = pad / linecols;
1222 rem = pad % linecols;
1223 for (tmp = c; tmp > eol; tmp--) {
1224 if (!tmp->width)
1225 continue;
1226
1227 tmp->width += spacing;
1228 if (rem) {
1229 tmp->width++;
1230 rem--;
1231 }
1232 }
1233
1234 newline:
1235 /* Line break: reset line counters, mark end-of-line */
1236 eol = c;
1237 len = 0;
1238 linecols = 0;
1239 }
1240 }
1241
1242 /* Render buffered output with spacing and delimiters, then free up buffers */
1243 static void render(void)
1244 {
1245 struct buf_token *token;
1246 int printed, line_started = 0;
1247 struct column *f;
1248
1249 if (!buffer.head)
1250 return;
1251
1252 token = (struct buf_token *)buffer.head->data;
1253
1254 /* Ensure end alignment of last token, it wasn't necessarily flushed */
1255 buffer.tail->end += buffer.cur->len % 2;
1256
1257 render_calc_width();
1258
1259 /* Rewind and replay */
1260 buffer.tail = buffer.head;
1261
1262 f = columns;
1263 while (!f->width)
1264 f++;
1265
1266 while (token) {
1267 /* Print left delimiter only if we already started a line */
1268 if (line_started++)
1269 printed = printf("%s", f->ldelim);
1270 else
1271 printed = 0;
1272
1273 /* Print field content from token data with spacing */
1274 printed += print_left_spacing(f, token->len, printed);
1275 printed += fwrite(token->data, 1, token->len, stdout);
1276 print_right_spacing(f, printed);
1277
1278 /* Go to next non-empty field, deal with end-of-line */
1279 do {
1280 if (field_is_last(f)) {
1281 printf("\n");
1282 f = columns;
1283 line_started = 0;
1284 } else {
1285 f++;
1286 }
1287 } while (f->disabled);
1288
1289 token = buf_token_next(token);
1290 }
1291
1292 buf_free_all();
1293 current_field = columns;
1294 }
1295
1296 /* Move to next field, and render buffer if we reached the maximum number of
1297 * chunks, at the last field in a line.
1298 */
1299 static void field_next(void)
1300 {
1301 if (field_is_last(current_field) && buffer.chunks >= BUF_CHUNKS_MAX) {
1302 render();
1303 return;
1304 }
1305
1306 field_flush(current_field);
1307 if (field_is_last(current_field))
1308 current_field = columns;
1309 else
1310 current_field++;
1311 }
1312
1313 /* Walk through fields and flush them until we reach the desired one */
1314 static void field_set(enum col_id id)
1315 {
1316 while (id != current_field - columns)
1317 field_next();
1318 }
1319
1320 /* Print header for all non-empty columns */
1321 static void print_header(void)
1322 {
1323 while (!field_is_last(current_field)) {
1324 if (!current_field->disabled)
1325 out("%s", current_field->header);
1326 field_next();
1327 }
1328 }
1329
1330 static void sock_state_print(struct sockstat *s)
1331 {
1332 const char *sock_name;
1333 static const char * const sstate_name[] = {
1334 "UNKNOWN",
1335 [SS_ESTABLISHED] = "ESTAB",
1336 [SS_SYN_SENT] = "SYN-SENT",
1337 [SS_SYN_RECV] = "SYN-RECV",
1338 [SS_FIN_WAIT1] = "FIN-WAIT-1",
1339 [SS_FIN_WAIT2] = "FIN-WAIT-2",
1340 [SS_TIME_WAIT] = "TIME-WAIT",
1341 [SS_CLOSE] = "UNCONN",
1342 [SS_CLOSE_WAIT] = "CLOSE-WAIT",
1343 [SS_LAST_ACK] = "LAST-ACK",
1344 [SS_LISTEN] = "LISTEN",
1345 [SS_CLOSING] = "CLOSING",
1346 };
1347
1348 switch (s->local.family) {
1349 case AF_UNIX:
1350 sock_name = unix_netid_name(s->type);
1351 break;
1352 case AF_INET:
1353 case AF_INET6:
1354 sock_name = proto_name(s->type);
1355 break;
1356 case AF_PACKET:
1357 sock_name = s->type == SOCK_RAW ? "p_raw" : "p_dgr";
1358 break;
1359 case AF_NETLINK:
1360 sock_name = "nl";
1361 break;
1362 case AF_TIPC:
1363 sock_name = tipc_netid_name(s->type);
1364 break;
1365 case AF_VSOCK:
1366 sock_name = vsock_netid_name(s->type);
1367 break;
1368 case AF_XDP:
1369 sock_name = "xdp";
1370 break;
1371 default:
1372 sock_name = "unknown";
1373 }
1374
1375 if (is_sctp_assoc(s, sock_name)) {
1376 field_set(COL_STATE); /* Empty Netid field */
1377 out("`- %s", sctp_sstate_name[s->state]);
1378 } else {
1379 field_set(COL_NETID);
1380 out("%s", sock_name);
1381 field_set(COL_STATE);
1382 out("%s", sstate_name[s->state]);
1383 }
1384
1385 field_set(COL_RECVQ);
1386 out("%-6d", s->rq);
1387 field_set(COL_SENDQ);
1388 out("%-6d", s->wq);
1389 field_set(COL_ADDR);
1390 }
1391
1392 static void sock_details_print(struct sockstat *s)
1393 {
1394 if (s->uid)
1395 out(" uid:%u", s->uid);
1396
1397 out(" ino:%u", s->ino);
1398 out(" sk:%llx", s->sk);
1399
1400 if (s->mark)
1401 out(" fwmark:0x%x", s->mark);
1402 }
1403
1404 static void sock_addr_print(const char *addr, char *delim, const char *port,
1405 const char *ifname)
1406 {
1407 if (ifname)
1408 out("%s" "%%" "%s%s", addr, ifname, delim);
1409 else
1410 out("%s%s", addr, delim);
1411
1412 field_next();
1413 out("%s", port);
1414 field_next();
1415 }
1416
1417 static const char *print_ms_timer(unsigned int timeout)
1418 {
1419 static char buf[64];
1420 int secs, msecs, minutes;
1421
1422 secs = timeout/1000;
1423 minutes = secs/60;
1424 secs = secs%60;
1425 msecs = timeout%1000;
1426 buf[0] = 0;
1427 if (minutes) {
1428 msecs = 0;
1429 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
1430 if (minutes > 9)
1431 secs = 0;
1432 }
1433 if (secs) {
1434 if (secs > 9)
1435 msecs = 0;
1436 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
1437 }
1438 if (msecs)
1439 sprintf(buf+strlen(buf), "%03dms", msecs);
1440 return buf;
1441 }
1442
1443 struct scache {
1444 struct scache *next;
1445 int port;
1446 char *name;
1447 const char *proto;
1448 };
1449
1450 static struct scache *rlist;
1451
1452 static void init_service_resolver(void)
1453 {
1454 char buf[128];
1455 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
1456
1457 if (!fp)
1458 return;
1459
1460 if (!fgets(buf, sizeof(buf), fp)) {
1461 pclose(fp);
1462 return;
1463 }
1464 while (fgets(buf, sizeof(buf), fp) != NULL) {
1465 unsigned int progn, port;
1466 char proto[128], prog[128] = "rpc.";
1467 struct scache *c;
1468
1469 if (sscanf(buf, "%u %*d %s %u %s",
1470 &progn, proto, &port, prog+4) != 4)
1471 continue;
1472
1473 if (!(c = malloc(sizeof(*c))))
1474 continue;
1475
1476 c->port = port;
1477 c->name = strdup(prog);
1478 if (strcmp(proto, TCP_PROTO) == 0)
1479 c->proto = TCP_PROTO;
1480 else if (strcmp(proto, UDP_PROTO) == 0)
1481 c->proto = UDP_PROTO;
1482 else if (strcmp(proto, SCTP_PROTO) == 0)
1483 c->proto = SCTP_PROTO;
1484 else
1485 c->proto = NULL;
1486 c->next = rlist;
1487 rlist = c;
1488 }
1489 pclose(fp);
1490 }
1491
1492 /* Even do not try default linux ephemeral port ranges:
1493 * default /etc/services contains so much of useless crap
1494 * wouldbe "allocated" to this area that resolution
1495 * is really harmful. I shrug each time when seeing
1496 * "socks" or "cfinger" in dumps.
1497 */
1498 static int is_ephemeral(int port)
1499 {
1500 static int min = 0, max;
1501
1502 if (!min) {
1503 FILE *f = ephemeral_ports_open();
1504
1505 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
1506 min = 1024;
1507 max = 4999;
1508 }
1509 if (f)
1510 fclose(f);
1511 }
1512 return port >= min && port <= max;
1513 }
1514
1515
1516 static const char *__resolve_service(int port)
1517 {
1518 struct scache *c;
1519
1520 for (c = rlist; c; c = c->next) {
1521 if (c->port == port && c->proto == dg_proto)
1522 return c->name;
1523 }
1524
1525 if (!is_ephemeral(port)) {
1526 static int notfirst;
1527 struct servent *se;
1528
1529 if (!notfirst) {
1530 setservent(1);
1531 notfirst = 1;
1532 }
1533 se = getservbyport(htons(port), dg_proto);
1534 if (se)
1535 return se->s_name;
1536 }
1537
1538 return NULL;
1539 }
1540
1541 #define SCACHE_BUCKETS 1024
1542 static struct scache *cache_htab[SCACHE_BUCKETS];
1543
1544 static const char *resolve_service(int port)
1545 {
1546 static char buf[128];
1547 struct scache *c;
1548 const char *res;
1549 int hash;
1550
1551 if (port == 0) {
1552 buf[0] = '*';
1553 buf[1] = 0;
1554 return buf;
1555 }
1556
1557 if (numeric)
1558 goto do_numeric;
1559
1560 if (dg_proto == RAW_PROTO)
1561 return inet_proto_n2a(port, buf, sizeof(buf));
1562
1563
1564 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
1565
1566 for (c = cache_htab[hash]; c; c = c->next) {
1567 if (c->port == port && c->proto == dg_proto)
1568 goto do_cache;
1569 }
1570
1571 c = malloc(sizeof(*c));
1572 if (!c)
1573 goto do_numeric;
1574 res = __resolve_service(port);
1575 c->port = port;
1576 c->name = res ? strdup(res) : NULL;
1577 c->proto = dg_proto;
1578 c->next = cache_htab[hash];
1579 cache_htab[hash] = c;
1580
1581 do_cache:
1582 if (c->name)
1583 return c->name;
1584
1585 do_numeric:
1586 sprintf(buf, "%u", port);
1587 return buf;
1588 }
1589
1590 static void inet_addr_print(const inet_prefix *a, int port,
1591 unsigned int ifindex, bool v6only)
1592 {
1593 char buf[1024];
1594 const char *ap = buf;
1595 const char *ifname = NULL;
1596
1597 if (a->family == AF_INET) {
1598 ap = format_host(AF_INET, 4, a->data);
1599 } else {
1600 if (!v6only &&
1601 !memcmp(a->data, &in6addr_any, sizeof(in6addr_any))) {
1602 buf[0] = '*';
1603 buf[1] = 0;
1604 } else {
1605 ap = format_host(a->family, 16, a->data);
1606
1607 /* Numeric IPv6 addresses should be bracketed */
1608 if (strchr(ap, ':')) {
1609 snprintf(buf, sizeof(buf),
1610 "[%s]", ap);
1611 ap = buf;
1612 }
1613 }
1614 }
1615
1616 if (ifindex)
1617 ifname = ll_index_to_name(ifindex);
1618
1619 sock_addr_print(ap, ":", resolve_service(port), ifname);
1620 }
1621
1622 struct aafilter {
1623 inet_prefix addr;
1624 int port;
1625 unsigned int iface;
1626 __u32 mark;
1627 __u32 mask;
1628 struct aafilter *next;
1629 };
1630
1631 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1632 int plen)
1633 {
1634 if (!inet_addr_match(a, p, plen))
1635 return 0;
1636
1637 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1638 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1639 * sockets. Fair? */
1640 if (p->family == AF_INET && a->family == AF_INET6) {
1641 if (a->data[0] == 0 && a->data[1] == 0 &&
1642 a->data[2] == htonl(0xffff)) {
1643 inet_prefix tmp = *a;
1644
1645 tmp.data[0] = a->data[3];
1646 return inet_addr_match(&tmp, p, plen);
1647 }
1648 }
1649 return 1;
1650 }
1651
1652 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1653 {
1654 char *addr, *pattern;
1655
1656 memcpy(&addr, a->data, sizeof(addr));
1657 memcpy(&pattern, p->data, sizeof(pattern));
1658 if (pattern == NULL)
1659 return 1;
1660 if (addr == NULL)
1661 addr = "";
1662 return !fnmatch(pattern, addr, 0);
1663 }
1664
1665 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1666 {
1667 switch (f->type) {
1668 case SSF_S_AUTO:
1669 {
1670 if (s->local.family == AF_UNIX) {
1671 char *p;
1672
1673 memcpy(&p, s->local.data, sizeof(p));
1674 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1675 strspn(p+1, "0123456789abcdef") == 5);
1676 }
1677 if (s->local.family == AF_PACKET)
1678 return s->lport == 0 && s->local.data[0] == 0;
1679 if (s->local.family == AF_NETLINK)
1680 return s->lport < 0;
1681 if (s->local.family == AF_VSOCK)
1682 return s->lport > 1023;
1683
1684 return is_ephemeral(s->lport);
1685 }
1686 case SSF_DCOND:
1687 {
1688 struct aafilter *a = (void *)f->pred;
1689
1690 if (a->addr.family == AF_UNIX)
1691 return unix_match(&s->remote, &a->addr);
1692 if (a->port != -1 && a->port != s->rport)
1693 return 0;
1694 if (a->addr.bitlen) {
1695 do {
1696 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1697 return 1;
1698 } while ((a = a->next) != NULL);
1699 return 0;
1700 }
1701 return 1;
1702 }
1703 case SSF_SCOND:
1704 {
1705 struct aafilter *a = (void *)f->pred;
1706
1707 if (a->addr.family == AF_UNIX)
1708 return unix_match(&s->local, &a->addr);
1709 if (a->port != -1 && a->port != s->lport)
1710 return 0;
1711 if (a->addr.bitlen) {
1712 do {
1713 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1714 return 1;
1715 } while ((a = a->next) != NULL);
1716 return 0;
1717 }
1718 return 1;
1719 }
1720 case SSF_D_GE:
1721 {
1722 struct aafilter *a = (void *)f->pred;
1723
1724 return s->rport >= a->port;
1725 }
1726 case SSF_D_LE:
1727 {
1728 struct aafilter *a = (void *)f->pred;
1729
1730 return s->rport <= a->port;
1731 }
1732 case SSF_S_GE:
1733 {
1734 struct aafilter *a = (void *)f->pred;
1735
1736 return s->lport >= a->port;
1737 }
1738 case SSF_S_LE:
1739 {
1740 struct aafilter *a = (void *)f->pred;
1741
1742 return s->lport <= a->port;
1743 }
1744 case SSF_DEVCOND:
1745 {
1746 struct aafilter *a = (void *)f->pred;
1747
1748 return s->iface == a->iface;
1749 }
1750 case SSF_MARKMASK:
1751 {
1752 struct aafilter *a = (void *)f->pred;
1753
1754 return (s->mark & a->mask) == a->mark;
1755 }
1756 /* Yup. It is recursion. Sorry. */
1757 case SSF_AND:
1758 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1759 case SSF_OR:
1760 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1761 case SSF_NOT:
1762 return !run_ssfilter(f->pred, s);
1763 default:
1764 abort();
1765 }
1766 }
1767
1768 /* Relocate external jumps by reloc. */
1769 static void ssfilter_patch(char *a, int len, int reloc)
1770 {
1771 while (len > 0) {
1772 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)a;
1773
1774 if (op->no == len+4)
1775 op->no += reloc;
1776 len -= op->yes;
1777 a += op->yes;
1778 }
1779 if (len < 0)
1780 abort();
1781 }
1782
1783 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1784 {
1785 switch (f->type) {
1786 case SSF_S_AUTO:
1787 {
1788 if (!(*bytecode = malloc(4))) abort();
1789 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1790 return 4;
1791 }
1792 case SSF_DCOND:
1793 case SSF_SCOND:
1794 {
1795 struct aafilter *a = (void *)f->pred;
1796 struct aafilter *b;
1797 char *ptr;
1798 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1799 int len = 0;
1800
1801 for (b = a; b; b = b->next) {
1802 len += 4 + sizeof(struct inet_diag_hostcond);
1803 if (a->addr.family == AF_INET6)
1804 len += 16;
1805 else
1806 len += 4;
1807 if (b->next)
1808 len += 4;
1809 }
1810 if (!(ptr = malloc(len))) abort();
1811 *bytecode = ptr;
1812 for (b = a; b; b = b->next) {
1813 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1814 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1815 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1816 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond *)(ptr+4);
1817
1818 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1819 cond->family = a->addr.family;
1820 cond->port = a->port;
1821 cond->prefix_len = a->addr.bitlen;
1822 memcpy(cond->addr, a->addr.data, alen);
1823 ptr += oplen;
1824 if (b->next) {
1825 op = (struct inet_diag_bc_op *)ptr;
1826 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1827 ptr += 4;
1828 }
1829 }
1830 return ptr - *bytecode;
1831 }
1832 case SSF_D_GE:
1833 {
1834 struct aafilter *x = (void *)f->pred;
1835
1836 if (!(*bytecode = malloc(8))) abort();
1837 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_GE, 8, 12 };
1838 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1839 return 8;
1840 }
1841 case SSF_D_LE:
1842 {
1843 struct aafilter *x = (void *)f->pred;
1844
1845 if (!(*bytecode = malloc(8))) abort();
1846 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_LE, 8, 12 };
1847 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1848 return 8;
1849 }
1850 case SSF_S_GE:
1851 {
1852 struct aafilter *x = (void *)f->pred;
1853
1854 if (!(*bytecode = malloc(8))) abort();
1855 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1856 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1857 return 8;
1858 }
1859 case SSF_S_LE:
1860 {
1861 struct aafilter *x = (void *)f->pred;
1862
1863 if (!(*bytecode = malloc(8))) abort();
1864 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1865 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1866 return 8;
1867 }
1868
1869 case SSF_AND:
1870 {
1871 char *a1 = NULL, *a2 = NULL, *a;
1872 int l1, l2;
1873
1874 l1 = ssfilter_bytecompile(f->pred, &a1);
1875 l2 = ssfilter_bytecompile(f->post, &a2);
1876 if (!l1 || !l2) {
1877 free(a1);
1878 free(a2);
1879 return 0;
1880 }
1881 if (!(a = malloc(l1+l2))) abort();
1882 memcpy(a, a1, l1);
1883 memcpy(a+l1, a2, l2);
1884 free(a1); free(a2);
1885 ssfilter_patch(a, l1, l2);
1886 *bytecode = a;
1887 return l1+l2;
1888 }
1889 case SSF_OR:
1890 {
1891 char *a1 = NULL, *a2 = NULL, *a;
1892 int l1, l2;
1893
1894 l1 = ssfilter_bytecompile(f->pred, &a1);
1895 l2 = ssfilter_bytecompile(f->post, &a2);
1896 if (!l1 || !l2) {
1897 free(a1);
1898 free(a2);
1899 return 0;
1900 }
1901 if (!(a = malloc(l1+l2+4))) abort();
1902 memcpy(a, a1, l1);
1903 memcpy(a+l1+4, a2, l2);
1904 free(a1); free(a2);
1905 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1906 *bytecode = a;
1907 return l1+l2+4;
1908 }
1909 case SSF_NOT:
1910 {
1911 char *a1 = NULL, *a;
1912 int l1;
1913
1914 l1 = ssfilter_bytecompile(f->pred, &a1);
1915 if (!l1) {
1916 free(a1);
1917 return 0;
1918 }
1919 if (!(a = malloc(l1+4))) abort();
1920 memcpy(a, a1, l1);
1921 free(a1);
1922 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1923 *bytecode = a;
1924 return l1+4;
1925 }
1926 case SSF_DEVCOND:
1927 {
1928 /* bytecompile for SSF_DEVCOND not supported yet */
1929 return 0;
1930 }
1931 case SSF_MARKMASK:
1932 {
1933 struct aafilter *a = (void *)f->pred;
1934 struct instr {
1935 struct inet_diag_bc_op op;
1936 struct inet_diag_markcond cond;
1937 };
1938 int inslen = sizeof(struct instr);
1939
1940 if (!(*bytecode = malloc(inslen))) abort();
1941 ((struct instr *)*bytecode)[0] = (struct instr) {
1942 { INET_DIAG_BC_MARK_COND, inslen, inslen + 4 },
1943 { a->mark, a->mask},
1944 };
1945
1946 return inslen;
1947 }
1948 default:
1949 abort();
1950 }
1951 }
1952
1953 static int remember_he(struct aafilter *a, struct hostent *he)
1954 {
1955 char **ptr = he->h_addr_list;
1956 int cnt = 0;
1957 int len;
1958
1959 if (he->h_addrtype == AF_INET)
1960 len = 4;
1961 else if (he->h_addrtype == AF_INET6)
1962 len = 16;
1963 else
1964 return 0;
1965
1966 while (*ptr) {
1967 struct aafilter *b = a;
1968
1969 if (a->addr.bitlen) {
1970 if ((b = malloc(sizeof(*b))) == NULL)
1971 return cnt;
1972 *b = *a;
1973 a->next = b;
1974 }
1975 memcpy(b->addr.data, *ptr, len);
1976 b->addr.bytelen = len;
1977 b->addr.bitlen = len*8;
1978 b->addr.family = he->h_addrtype;
1979 ptr++;
1980 cnt++;
1981 }
1982 return cnt;
1983 }
1984
1985 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
1986 {
1987 static int notfirst;
1988 int cnt = 0;
1989 struct hostent *he;
1990
1991 a->addr.bitlen = 0;
1992 if (!notfirst) {
1993 sethostent(1);
1994 notfirst = 1;
1995 }
1996 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
1997 if (he)
1998 cnt = remember_he(a, he);
1999 if (fam == AF_UNSPEC) {
2000 he = gethostbyname2(addr, AF_INET6);
2001 if (he)
2002 cnt += remember_he(a, he);
2003 }
2004 return !cnt;
2005 }
2006
2007 static int xll_initted;
2008
2009 static void xll_init(void)
2010 {
2011 struct rtnl_handle rth;
2012
2013 if (rtnl_open(&rth, 0) < 0)
2014 exit(1);
2015
2016 ll_init_map(&rth);
2017 rtnl_close(&rth);
2018 xll_initted = 1;
2019 }
2020
2021 static const char *xll_index_to_name(int index)
2022 {
2023 if (!xll_initted)
2024 xll_init();
2025 return ll_index_to_name(index);
2026 }
2027
2028 static int xll_name_to_index(const char *dev)
2029 {
2030 if (!xll_initted)
2031 xll_init();
2032 return ll_name_to_index(dev);
2033 }
2034
2035 void *parse_devcond(char *name)
2036 {
2037 struct aafilter a = { .iface = 0 };
2038 struct aafilter *res;
2039
2040 a.iface = xll_name_to_index(name);
2041 if (a.iface == 0) {
2042 char *end;
2043 unsigned long n;
2044
2045 n = strtoul(name, &end, 0);
2046 if (!end || end == name || *end || n > UINT_MAX)
2047 return NULL;
2048
2049 a.iface = n;
2050 }
2051
2052 res = malloc(sizeof(*res));
2053 *res = a;
2054
2055 return res;
2056 }
2057
2058 static void vsock_set_inet_prefix(inet_prefix *a, __u32 cid)
2059 {
2060 *a = (inet_prefix){
2061 .bytelen = sizeof(cid),
2062 .family = AF_VSOCK,
2063 };
2064 memcpy(a->data, &cid, sizeof(cid));
2065 }
2066
2067 void *parse_hostcond(char *addr, bool is_port)
2068 {
2069 char *port = NULL;
2070 struct aafilter a = { .port = -1 };
2071 struct aafilter *res;
2072 int fam = preferred_family;
2073 struct filter *f = &current_filter;
2074
2075 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
2076 char *p;
2077
2078 a.addr.family = AF_UNIX;
2079 if (strncmp(addr, "unix:", 5) == 0)
2080 addr += 5;
2081 p = strdup(addr);
2082 a.addr.bitlen = 8*strlen(p);
2083 memcpy(a.addr.data, &p, sizeof(p));
2084 fam = AF_UNIX;
2085 goto out;
2086 }
2087
2088 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
2089 a.addr.family = AF_PACKET;
2090 a.addr.bitlen = 0;
2091 if (strncmp(addr, "link:", 5) == 0)
2092 addr += 5;
2093 port = strchr(addr, ':');
2094 if (port) {
2095 *port = 0;
2096 if (port[1] && strcmp(port+1, "*")) {
2097 if (get_integer(&a.port, port+1, 0)) {
2098 if ((a.port = xll_name_to_index(port+1)) <= 0)
2099 return NULL;
2100 }
2101 }
2102 }
2103 if (addr[0] && strcmp(addr, "*")) {
2104 unsigned short tmp;
2105
2106 a.addr.bitlen = 32;
2107 if (ll_proto_a2n(&tmp, addr))
2108 return NULL;
2109 a.addr.data[0] = ntohs(tmp);
2110 }
2111 fam = AF_PACKET;
2112 goto out;
2113 }
2114
2115 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
2116 a.addr.family = AF_NETLINK;
2117 a.addr.bitlen = 0;
2118 if (strncmp(addr, "netlink:", 8) == 0)
2119 addr += 8;
2120 port = strchr(addr, ':');
2121 if (port) {
2122 *port = 0;
2123 if (port[1] && strcmp(port+1, "*")) {
2124 if (get_integer(&a.port, port+1, 0)) {
2125 if (strcmp(port+1, "kernel") == 0)
2126 a.port = 0;
2127 else
2128 return NULL;
2129 }
2130 }
2131 }
2132 if (addr[0] && strcmp(addr, "*")) {
2133 a.addr.bitlen = 32;
2134 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
2135 return NULL;
2136 }
2137 fam = AF_NETLINK;
2138 goto out;
2139 }
2140
2141 if (fam == AF_VSOCK || strncmp(addr, "vsock:", 6) == 0) {
2142 __u32 cid = ~(__u32)0;
2143
2144 a.addr.family = AF_VSOCK;
2145 if (strncmp(addr, "vsock:", 6) == 0)
2146 addr += 6;
2147
2148 if (is_port)
2149 port = addr;
2150 else {
2151 port = strchr(addr, ':');
2152 if (port) {
2153 *port = '\0';
2154 port++;
2155 }
2156 }
2157
2158 if (port && strcmp(port, "*") &&
2159 get_u32((__u32 *)&a.port, port, 0))
2160 return NULL;
2161
2162 if (addr[0] && strcmp(addr, "*")) {
2163 a.addr.bitlen = 32;
2164 if (get_u32(&cid, addr, 0))
2165 return NULL;
2166 }
2167 vsock_set_inet_prefix(&a.addr, cid);
2168 fam = AF_VSOCK;
2169 goto out;
2170 }
2171
2172 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
2173 fam = AF_INET;
2174 if (!strncmp(addr, "inet:", 5))
2175 addr += 5;
2176 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
2177 fam = AF_INET6;
2178 if (!strncmp(addr, "inet6:", 6))
2179 addr += 6;
2180 }
2181
2182 /* URL-like literal [] */
2183 if (addr[0] == '[') {
2184 addr++;
2185 if ((port = strchr(addr, ']')) == NULL)
2186 return NULL;
2187 *port++ = 0;
2188 } else if (addr[0] == '*') {
2189 port = addr+1;
2190 } else {
2191 port = strrchr(strchr(addr, '/') ? : addr, ':');
2192 }
2193
2194 if (is_port)
2195 port = addr;
2196
2197 if (port && *port) {
2198 if (*port == ':')
2199 *port++ = 0;
2200
2201 if (*port && *port != '*') {
2202 if (get_integer(&a.port, port, 0)) {
2203 struct servent *se1 = NULL;
2204 struct servent *se2 = NULL;
2205
2206 if (current_filter.dbs&(1<<UDP_DB))
2207 se1 = getservbyname(port, UDP_PROTO);
2208 if (current_filter.dbs&(1<<TCP_DB))
2209 se2 = getservbyname(port, TCP_PROTO);
2210 if (se1 && se2 && se1->s_port != se2->s_port) {
2211 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2212 return NULL;
2213 }
2214 if (!se1)
2215 se1 = se2;
2216 if (se1) {
2217 a.port = ntohs(se1->s_port);
2218 } else {
2219 struct scache *s;
2220
2221 for (s = rlist; s; s = s->next) {
2222 if ((s->proto == UDP_PROTO &&
2223 (current_filter.dbs&(1<<UDP_DB))) ||
2224 (s->proto == TCP_PROTO &&
2225 (current_filter.dbs&(1<<TCP_DB)))) {
2226 if (s->name && strcmp(s->name, port) == 0) {
2227 if (a.port > 0 && a.port != s->port) {
2228 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2229 return NULL;
2230 }
2231 a.port = s->port;
2232 }
2233 }
2234 }
2235 if (a.port <= 0) {
2236 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
2237 return NULL;
2238 }
2239 }
2240 }
2241 }
2242 }
2243 if (!is_port && *addr && *addr != '*') {
2244 if (get_prefix_1(&a.addr, addr, fam)) {
2245 if (get_dns_host(&a, addr, fam)) {
2246 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
2247 return NULL;
2248 }
2249 }
2250 }
2251
2252 out:
2253 if (fam != AF_UNSPEC) {
2254 int states = f->states;
2255 f->families = 0;
2256 filter_af_set(f, fam);
2257 filter_states_set(f, states);
2258 }
2259
2260 res = malloc(sizeof(*res));
2261 if (res)
2262 memcpy(res, &a, sizeof(a));
2263 return res;
2264 }
2265
2266 void *parse_markmask(const char *markmask)
2267 {
2268 struct aafilter a, *res;
2269
2270 if (strchr(markmask, '/')) {
2271 if (sscanf(markmask, "%i/%i", &a.mark, &a.mask) != 2)
2272 return NULL;
2273 } else {
2274 a.mask = 0xffffffff;
2275 if (sscanf(markmask, "%i", &a.mark) != 1)
2276 return NULL;
2277 }
2278
2279 res = malloc(sizeof(*res));
2280 if (res)
2281 memcpy(res, &a, sizeof(a));
2282 return res;
2283 }
2284
2285 static void proc_ctx_print(struct sockstat *s)
2286 {
2287 char *buf;
2288
2289 if (show_proc_ctx || show_sock_ctx) {
2290 if (find_entry(s->ino, &buf,
2291 (show_proc_ctx & show_sock_ctx) ?
2292 PROC_SOCK_CTX : PROC_CTX) > 0) {
2293 out(" users:(%s)", buf);
2294 free(buf);
2295 }
2296 } else if (show_users) {
2297 if (find_entry(s->ino, &buf, USERS) > 0) {
2298 out(" users:(%s)", buf);
2299 free(buf);
2300 }
2301 }
2302 }
2303
2304 static void inet_stats_print(struct sockstat *s, bool v6only)
2305 {
2306 sock_state_print(s);
2307
2308 inet_addr_print(&s->local, s->lport, s->iface, v6only);
2309 inet_addr_print(&s->remote, s->rport, 0, v6only);
2310
2311 proc_ctx_print(s);
2312 }
2313
2314 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
2315 sockstat * s)
2316 {
2317 s->local.family = s->remote.family = family;
2318 if (family == AF_INET) {
2319 sscanf(loc, "%x:%x", s->local.data, (unsigned *)&s->lport);
2320 sscanf(rem, "%x:%x", s->remote.data, (unsigned *)&s->rport);
2321 s->local.bytelen = s->remote.bytelen = 4;
2322 return 0;
2323 } else {
2324 sscanf(loc, "%08x%08x%08x%08x:%x",
2325 s->local.data,
2326 s->local.data + 1,
2327 s->local.data + 2,
2328 s->local.data + 3,
2329 &s->lport);
2330 sscanf(rem, "%08x%08x%08x%08x:%x",
2331 s->remote.data,
2332 s->remote.data + 1,
2333 s->remote.data + 2,
2334 s->remote.data + 3,
2335 &s->rport);
2336 s->local.bytelen = s->remote.bytelen = 16;
2337 return 0;
2338 }
2339 return -1;
2340 }
2341
2342 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
2343 {
2344 char *p;
2345
2346 if ((p = strchr(line, ':')) == NULL)
2347 return -1;
2348
2349 *loc = p+2;
2350 if ((p = strchr(*loc, ':')) == NULL)
2351 return -1;
2352
2353 p[5] = 0;
2354 *rem = p+6;
2355 if ((p = strchr(*rem, ':')) == NULL)
2356 return -1;
2357
2358 p[5] = 0;
2359 *data = p+6;
2360 return 0;
2361 }
2362
2363 static char *sprint_bw(char *buf, double bw)
2364 {
2365 if (numeric)
2366 sprintf(buf, "%.0f", bw);
2367 else if (bw > 1000000.)
2368 sprintf(buf, "%.1fM", bw / 1000000.);
2369 else if (bw > 1000.)
2370 sprintf(buf, "%.1fK", bw / 1000.);
2371 else
2372 sprintf(buf, "%g", bw);
2373
2374 return buf;
2375 }
2376
2377 static void sctp_stats_print(struct sctp_info *s)
2378 {
2379 if (s->sctpi_tag)
2380 out(" tag:%x", s->sctpi_tag);
2381 if (s->sctpi_state)
2382 out(" state:%s", sctp_sstate_name[s->sctpi_state]);
2383 if (s->sctpi_rwnd)
2384 out(" rwnd:%d", s->sctpi_rwnd);
2385 if (s->sctpi_unackdata)
2386 out(" unackdata:%d", s->sctpi_unackdata);
2387 if (s->sctpi_penddata)
2388 out(" penddata:%d", s->sctpi_penddata);
2389 if (s->sctpi_instrms)
2390 out(" instrms:%d", s->sctpi_instrms);
2391 if (s->sctpi_outstrms)
2392 out(" outstrms:%d", s->sctpi_outstrms);
2393 if (s->sctpi_inqueue)
2394 out(" inqueue:%d", s->sctpi_inqueue);
2395 if (s->sctpi_outqueue)
2396 out(" outqueue:%d", s->sctpi_outqueue);
2397 if (s->sctpi_overall_error)
2398 out(" overerr:%d", s->sctpi_overall_error);
2399 if (s->sctpi_max_burst)
2400 out(" maxburst:%d", s->sctpi_max_burst);
2401 if (s->sctpi_maxseg)
2402 out(" maxseg:%d", s->sctpi_maxseg);
2403 if (s->sctpi_peer_rwnd)
2404 out(" prwnd:%d", s->sctpi_peer_rwnd);
2405 if (s->sctpi_peer_tag)
2406 out(" ptag:%x", s->sctpi_peer_tag);
2407 if (s->sctpi_peer_capable)
2408 out(" pcapable:%d", s->sctpi_peer_capable);
2409 if (s->sctpi_peer_sack)
2410 out(" psack:%d", s->sctpi_peer_sack);
2411 if (s->sctpi_s_autoclose)
2412 out(" autoclose:%d", s->sctpi_s_autoclose);
2413 if (s->sctpi_s_adaptation_ind)
2414 out(" adapind:%d", s->sctpi_s_adaptation_ind);
2415 if (s->sctpi_s_pd_point)
2416 out(" pdpoint:%d", s->sctpi_s_pd_point);
2417 if (s->sctpi_s_nodelay)
2418 out(" nodelay:%d", s->sctpi_s_nodelay);
2419 if (s->sctpi_s_disable_fragments)
2420 out(" nofrag:%d", s->sctpi_s_disable_fragments);
2421 if (s->sctpi_s_v4mapped)
2422 out(" v4mapped:%d", s->sctpi_s_v4mapped);
2423 if (s->sctpi_s_frag_interleave)
2424 out(" fraginl:%d", s->sctpi_s_frag_interleave);
2425 }
2426
2427 static void tcp_stats_print(struct tcpstat *s)
2428 {
2429 char b1[64];
2430
2431 if (s->has_ts_opt)
2432 out(" ts");
2433 if (s->has_sack_opt)
2434 out(" sack");
2435 if (s->has_ecn_opt)
2436 out(" ecn");
2437 if (s->has_ecnseen_opt)
2438 out(" ecnseen");
2439 if (s->has_fastopen_opt)
2440 out(" fastopen");
2441 if (s->cong_alg[0])
2442 out(" %s", s->cong_alg);
2443 if (s->has_wscale_opt)
2444 out(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
2445 if (s->rto)
2446 out(" rto:%g", s->rto);
2447 if (s->backoff)
2448 out(" backoff:%u", s->backoff);
2449 if (s->rtt)
2450 out(" rtt:%g/%g", s->rtt, s->rttvar);
2451 if (s->ato)
2452 out(" ato:%g", s->ato);
2453
2454 if (s->qack)
2455 out(" qack:%d", s->qack);
2456 if (s->qack & 1)
2457 out(" bidir");
2458
2459 if (s->mss)
2460 out(" mss:%d", s->mss);
2461 if (s->pmtu)
2462 out(" pmtu:%u", s->pmtu);
2463 if (s->rcv_mss)
2464 out(" rcvmss:%d", s->rcv_mss);
2465 if (s->advmss)
2466 out(" advmss:%d", s->advmss);
2467 if (s->cwnd)
2468 out(" cwnd:%u", s->cwnd);
2469 if (s->ssthresh)
2470 out(" ssthresh:%d", s->ssthresh);
2471
2472 if (s->bytes_sent)
2473 out(" bytes_sent:%llu", s->bytes_sent);
2474 if (s->bytes_retrans)
2475 out(" bytes_retrans:%llu", s->bytes_retrans);
2476 if (s->bytes_acked)
2477 out(" bytes_acked:%llu", s->bytes_acked);
2478 if (s->bytes_received)
2479 out(" bytes_received:%llu", s->bytes_received);
2480 if (s->segs_out)
2481 out(" segs_out:%u", s->segs_out);
2482 if (s->segs_in)
2483 out(" segs_in:%u", s->segs_in);
2484 if (s->data_segs_out)
2485 out(" data_segs_out:%u", s->data_segs_out);
2486 if (s->data_segs_in)
2487 out(" data_segs_in:%u", s->data_segs_in);
2488
2489 if (s->dctcp && s->dctcp->enabled) {
2490 struct dctcpstat *dctcp = s->dctcp;
2491
2492 out(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
2493 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
2494 dctcp->ab_tot);
2495 } else if (s->dctcp) {
2496 out(" dctcp:fallback_mode");
2497 }
2498
2499 if (s->bbr_info) {
2500 __u64 bw;
2501
2502 bw = s->bbr_info->bbr_bw_hi;
2503 bw <<= 32;
2504 bw |= s->bbr_info->bbr_bw_lo;
2505
2506 out(" bbr:(bw:%sbps,mrtt:%g",
2507 sprint_bw(b1, bw * 8.0),
2508 (double)s->bbr_info->bbr_min_rtt / 1000.0);
2509 if (s->bbr_info->bbr_pacing_gain)
2510 out(",pacing_gain:%g",
2511 (double)s->bbr_info->bbr_pacing_gain / 256.0);
2512 if (s->bbr_info->bbr_cwnd_gain)
2513 out(",cwnd_gain:%g",
2514 (double)s->bbr_info->bbr_cwnd_gain / 256.0);
2515 out(")");
2516 }
2517
2518 if (s->send_bps)
2519 out(" send %sbps", sprint_bw(b1, s->send_bps));
2520 if (s->lastsnd)
2521 out(" lastsnd:%u", s->lastsnd);
2522 if (s->lastrcv)
2523 out(" lastrcv:%u", s->lastrcv);
2524 if (s->lastack)
2525 out(" lastack:%u", s->lastack);
2526
2527 if (s->pacing_rate) {
2528 out(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
2529 if (s->pacing_rate_max)
2530 out("/%sbps", sprint_bw(b1, s->pacing_rate_max));
2531 }
2532
2533 if (s->delivery_rate)
2534 out(" delivery_rate %sbps", sprint_bw(b1, s->delivery_rate));
2535 if (s->delivered)
2536 out(" delivered:%u", s->delivered);
2537 if (s->delivered_ce)
2538 out(" delivered_ce:%u", s->delivered_ce);
2539 if (s->app_limited)
2540 out(" app_limited");
2541
2542 if (s->busy_time) {
2543 out(" busy:%llums", s->busy_time / 1000);
2544 if (s->rwnd_limited)
2545 out(" rwnd_limited:%llums(%.1f%%)",
2546 s->rwnd_limited / 1000,
2547 100.0 * s->rwnd_limited / s->busy_time);
2548 if (s->sndbuf_limited)
2549 out(" sndbuf_limited:%llums(%.1f%%)",
2550 s->sndbuf_limited / 1000,
2551 100.0 * s->sndbuf_limited / s->busy_time);
2552 }
2553
2554 if (s->unacked)
2555 out(" unacked:%u", s->unacked);
2556 if (s->retrans || s->retrans_total)
2557 out(" retrans:%u/%u", s->retrans, s->retrans_total);
2558 if (s->lost)
2559 out(" lost:%u", s->lost);
2560 if (s->sacked && s->ss.state != SS_LISTEN)
2561 out(" sacked:%u", s->sacked);
2562 if (s->dsack_dups)
2563 out(" dsack_dups:%u", s->dsack_dups);
2564 if (s->fackets)
2565 out(" fackets:%u", s->fackets);
2566 if (s->reordering != 3)
2567 out(" reordering:%d", s->reordering);
2568 if (s->reord_seen)
2569 out(" reord_seen:%d", s->reord_seen);
2570 if (s->rcv_rtt)
2571 out(" rcv_rtt:%g", s->rcv_rtt);
2572 if (s->rcv_space)
2573 out(" rcv_space:%d", s->rcv_space);
2574 if (s->rcv_ssthresh)
2575 out(" rcv_ssthresh:%u", s->rcv_ssthresh);
2576 if (s->not_sent)
2577 out(" notsent:%u", s->not_sent);
2578 if (s->min_rtt)
2579 out(" minrtt:%g", s->min_rtt);
2580 }
2581
2582 static void tcp_timer_print(struct tcpstat *s)
2583 {
2584 static const char * const tmr_name[] = {
2585 "off",
2586 "on",
2587 "keepalive",
2588 "timewait",
2589 "persist",
2590 "unknown"
2591 };
2592
2593 if (s->timer) {
2594 if (s->timer > 4)
2595 s->timer = 5;
2596 out(" timer:(%s,%s,%d)",
2597 tmr_name[s->timer],
2598 print_ms_timer(s->timeout),
2599 s->retrans);
2600 }
2601 }
2602
2603 static void sctp_timer_print(struct tcpstat *s)
2604 {
2605 if (s->timer)
2606 out(" timer:(T3_RTX,%s,%d)",
2607 print_ms_timer(s->timeout), s->retrans);
2608 }
2609
2610 static int tcp_show_line(char *line, const struct filter *f, int family)
2611 {
2612 int rto = 0, ato = 0;
2613 struct tcpstat s = {};
2614 char *loc, *rem, *data;
2615 char opt[256];
2616 int n;
2617 int hz = get_user_hz();
2618
2619 if (proc_inet_split_line(line, &loc, &rem, &data))
2620 return -1;
2621
2622 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2623
2624 if (!(f->states & (1 << state)))
2625 return 0;
2626
2627 proc_parse_inet_addr(loc, rem, family, &s.ss);
2628
2629 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
2630 return 0;
2631
2632 opt[0] = 0;
2633 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %u %d %[^\n]\n",
2634 &s.ss.state, &s.ss.wq, &s.ss.rq,
2635 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
2636 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
2637 &s.ssthresh, opt);
2638
2639 if (n < 17)
2640 opt[0] = 0;
2641
2642 if (n < 12) {
2643 rto = 0;
2644 s.cwnd = 2;
2645 s.ssthresh = -1;
2646 ato = s.qack = 0;
2647 }
2648
2649 s.retrans = s.timer != 1 ? s.probes : s.retrans;
2650 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
2651 s.ato = (double)ato / hz;
2652 s.qack /= 2;
2653 s.rto = (double)rto;
2654 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
2655 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
2656 s.ss.type = IPPROTO_TCP;
2657
2658 inet_stats_print(&s.ss, false);
2659
2660 if (show_options)
2661 tcp_timer_print(&s);
2662
2663 if (show_details) {
2664 sock_details_print(&s.ss);
2665 if (opt[0])
2666 out(" opt:\"%s\"", opt);
2667 }
2668
2669 if (show_tcpinfo)
2670 tcp_stats_print(&s);
2671
2672 return 0;
2673 }
2674
2675 static int generic_record_read(FILE *fp,
2676 int (*worker)(char*, const struct filter *, int),
2677 const struct filter *f, int fam)
2678 {
2679 char line[256];
2680
2681 /* skip header */
2682 if (fgets(line, sizeof(line), fp) == NULL)
2683 goto outerr;
2684
2685 while (fgets(line, sizeof(line), fp) != NULL) {
2686 int n = strlen(line);
2687
2688 if (n == 0 || line[n-1] != '\n') {
2689 errno = -EINVAL;
2690 return -1;
2691 }
2692 line[n-1] = 0;
2693
2694 if (worker(line, f, fam) < 0)
2695 return 0;
2696 }
2697 outerr:
2698
2699 return ferror(fp) ? -1 : 0;
2700 }
2701
2702 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
2703 {
2704 const __u32 *skmeminfo;
2705
2706 if (!tb[attrtype]) {
2707 if (attrtype == INET_DIAG_SKMEMINFO) {
2708 if (!tb[INET_DIAG_MEMINFO])
2709 return;
2710
2711 const struct inet_diag_meminfo *minfo =
2712 RTA_DATA(tb[INET_DIAG_MEMINFO]);
2713
2714 out(" mem:(r%u,w%u,f%u,t%u)",
2715 minfo->idiag_rmem,
2716 minfo->idiag_wmem,
2717 minfo->idiag_fmem,
2718 minfo->idiag_tmem);
2719 }
2720 return;
2721 }
2722
2723 skmeminfo = RTA_DATA(tb[attrtype]);
2724
2725 out(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
2726 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
2727 skmeminfo[SK_MEMINFO_RCVBUF],
2728 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
2729 skmeminfo[SK_MEMINFO_SNDBUF],
2730 skmeminfo[SK_MEMINFO_FWD_ALLOC],
2731 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
2732 skmeminfo[SK_MEMINFO_OPTMEM]);
2733
2734 if (RTA_PAYLOAD(tb[attrtype]) >=
2735 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
2736 out(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
2737
2738 if (RTA_PAYLOAD(tb[attrtype]) >=
2739 (SK_MEMINFO_DROPS + 1) * sizeof(__u32))
2740 out(",d%u", skmeminfo[SK_MEMINFO_DROPS]);
2741
2742 out(")");
2743 }
2744
2745 static void print_md5sig(struct tcp_diag_md5sig *sig)
2746 {
2747 out("%s/%d=",
2748 format_host(sig->tcpm_family,
2749 sig->tcpm_family == AF_INET6 ? 16 : 4,
2750 &sig->tcpm_addr),
2751 sig->tcpm_prefixlen);
2752 print_escape_buf(sig->tcpm_key, sig->tcpm_keylen, " ,");
2753 }
2754
2755 static void tcp_tls_version(struct rtattr *attr)
2756 {
2757 u_int16_t val;
2758
2759 if (!attr)
2760 return;
2761 val = rta_getattr_u16(attr);
2762
2763 switch (val) {
2764 case TLS_1_2_VERSION:
2765 out(" version: 1.2");
2766 break;
2767 case TLS_1_3_VERSION:
2768 out(" version: 1.3");
2769 break;
2770 default:
2771 out(" version: unknown(%hu)", val);
2772 break;
2773 }
2774 }
2775
2776 static void tcp_tls_cipher(struct rtattr *attr)
2777 {
2778 u_int16_t val;
2779
2780 if (!attr)
2781 return;
2782 val = rta_getattr_u16(attr);
2783
2784 switch (val) {
2785 case TLS_CIPHER_AES_GCM_128:
2786 out(" cipher: aes-gcm-128");
2787 break;
2788 case TLS_CIPHER_AES_GCM_256:
2789 out(" cipher: aes-gcm-256");
2790 break;
2791 }
2792 }
2793
2794 static void tcp_tls_conf(const char *name, struct rtattr *attr)
2795 {
2796 u_int16_t val;
2797
2798 if (!attr)
2799 return;
2800 val = rta_getattr_u16(attr);
2801
2802 switch (val) {
2803 case TLS_CONF_BASE:
2804 out(" %s: none", name);
2805 break;
2806 case TLS_CONF_SW:
2807 out(" %s: sw", name);
2808 break;
2809 case TLS_CONF_HW:
2810 out(" %s: hw", name);
2811 break;
2812 case TLS_CONF_HW_RECORD:
2813 out(" %s: hw-record", name);
2814 break;
2815 default:
2816 out(" %s: unknown(%hu)", name, val);
2817 break;
2818 }
2819 }
2820
2821 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
2822
2823 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2824 struct rtattr *tb[])
2825 {
2826 double rtt = 0;
2827 struct tcpstat s = {};
2828
2829 s.ss.state = r->idiag_state;
2830
2831 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2832
2833 if (tb[INET_DIAG_INFO]) {
2834 struct tcp_info *info;
2835 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2836
2837 /* workaround for older kernels with less fields */
2838 if (len < sizeof(*info)) {
2839 info = alloca(sizeof(*info));
2840 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2841 memset((char *)info + len, 0, sizeof(*info) - len);
2842 } else
2843 info = RTA_DATA(tb[INET_DIAG_INFO]);
2844
2845 if (show_options) {
2846 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
2847 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
2848 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
2849 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
2850 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
2851 }
2852
2853 if (tb[INET_DIAG_CONG])
2854 strncpy(s.cong_alg,
2855 rta_getattr_str(tb[INET_DIAG_CONG]),
2856 sizeof(s.cong_alg) - 1);
2857
2858 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
2859 s.has_wscale_opt = true;
2860 s.snd_wscale = info->tcpi_snd_wscale;
2861 s.rcv_wscale = info->tcpi_rcv_wscale;
2862 }
2863
2864 if (info->tcpi_rto && info->tcpi_rto != 3000000)
2865 s.rto = (double)info->tcpi_rto / 1000;
2866
2867 s.backoff = info->tcpi_backoff;
2868 s.rtt = (double)info->tcpi_rtt / 1000;
2869 s.rttvar = (double)info->tcpi_rttvar / 1000;
2870 s.ato = (double)info->tcpi_ato / 1000;
2871 s.mss = info->tcpi_snd_mss;
2872 s.rcv_mss = info->tcpi_rcv_mss;
2873 s.advmss = info->tcpi_advmss;
2874 s.rcv_space = info->tcpi_rcv_space;
2875 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
2876 s.lastsnd = info->tcpi_last_data_sent;
2877 s.lastrcv = info->tcpi_last_data_recv;
2878 s.lastack = info->tcpi_last_ack_recv;
2879 s.unacked = info->tcpi_unacked;
2880 s.retrans = info->tcpi_retrans;
2881 s.retrans_total = info->tcpi_total_retrans;
2882 s.lost = info->tcpi_lost;
2883 s.sacked = info->tcpi_sacked;
2884 s.fackets = info->tcpi_fackets;
2885 s.reordering = info->tcpi_reordering;
2886 s.rcv_ssthresh = info->tcpi_rcv_ssthresh;
2887 s.cwnd = info->tcpi_snd_cwnd;
2888 s.pmtu = info->tcpi_pmtu;
2889
2890 if (info->tcpi_snd_ssthresh < 0xFFFF)
2891 s.ssthresh = info->tcpi_snd_ssthresh;
2892
2893 rtt = (double) info->tcpi_rtt;
2894 if (tb[INET_DIAG_VEGASINFO]) {
2895 const struct tcpvegas_info *vinfo
2896 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
2897
2898 if (vinfo->tcpv_enabled &&
2899 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
2900 rtt = vinfo->tcpv_rtt;
2901 }
2902
2903 if (tb[INET_DIAG_DCTCPINFO]) {
2904 struct dctcpstat *dctcp = malloc(sizeof(struct
2905 dctcpstat));
2906
2907 const struct tcp_dctcp_info *dinfo
2908 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
2909
2910 dctcp->enabled = !!dinfo->dctcp_enabled;
2911 dctcp->ce_state = dinfo->dctcp_ce_state;
2912 dctcp->alpha = dinfo->dctcp_alpha;
2913 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
2914 dctcp->ab_tot = dinfo->dctcp_ab_tot;
2915 s.dctcp = dctcp;
2916 }
2917
2918 if (tb[INET_DIAG_BBRINFO]) {
2919 const void *bbr_info = RTA_DATA(tb[INET_DIAG_BBRINFO]);
2920 int len = min(RTA_PAYLOAD(tb[INET_DIAG_BBRINFO]),
2921 sizeof(*s.bbr_info));
2922
2923 s.bbr_info = calloc(1, sizeof(*s.bbr_info));
2924 if (s.bbr_info && bbr_info)
2925 memcpy(s.bbr_info, bbr_info, len);
2926 }
2927
2928 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
2929 s.send_bps = (double) info->tcpi_snd_cwnd *
2930 (double)info->tcpi_snd_mss * 8000000. / rtt;
2931 }
2932
2933 if (info->tcpi_pacing_rate &&
2934 info->tcpi_pacing_rate != ~0ULL) {
2935 s.pacing_rate = info->tcpi_pacing_rate * 8.;
2936
2937 if (info->tcpi_max_pacing_rate &&
2938 info->tcpi_max_pacing_rate != ~0ULL)
2939 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
2940 }
2941 s.bytes_acked = info->tcpi_bytes_acked;
2942 s.bytes_received = info->tcpi_bytes_received;
2943 s.segs_out = info->tcpi_segs_out;
2944 s.segs_in = info->tcpi_segs_in;
2945 s.data_segs_out = info->tcpi_data_segs_out;
2946 s.data_segs_in = info->tcpi_data_segs_in;
2947 s.not_sent = info->tcpi_notsent_bytes;
2948 if (info->tcpi_min_rtt && info->tcpi_min_rtt != ~0U)
2949 s.min_rtt = (double) info->tcpi_min_rtt / 1000;
2950 s.delivery_rate = info->tcpi_delivery_rate * 8.;
2951 s.app_limited = info->tcpi_delivery_rate_app_limited;
2952 s.busy_time = info->tcpi_busy_time;
2953 s.rwnd_limited = info->tcpi_rwnd_limited;
2954 s.sndbuf_limited = info->tcpi_sndbuf_limited;
2955 s.delivered = info->tcpi_delivered;
2956 s.delivered_ce = info->tcpi_delivered_ce;
2957 s.dsack_dups = info->tcpi_dsack_dups;
2958 s.reord_seen = info->tcpi_reord_seen;
2959 s.bytes_sent = info->tcpi_bytes_sent;
2960 s.bytes_retrans = info->tcpi_bytes_retrans;
2961 tcp_stats_print(&s);
2962 free(s.dctcp);
2963 free(s.bbr_info);
2964 }
2965 if (tb[INET_DIAG_MD5SIG]) {
2966 struct tcp_diag_md5sig *sig = RTA_DATA(tb[INET_DIAG_MD5SIG]);
2967 int len = RTA_PAYLOAD(tb[INET_DIAG_MD5SIG]);
2968
2969 out(" md5keys:");
2970 print_md5sig(sig++);
2971 for (len -= sizeof(*sig); len > 0; len -= sizeof(*sig)) {
2972 out(",");
2973 print_md5sig(sig++);
2974 }
2975 }
2976 if (tb[INET_DIAG_ULP_INFO]) {
2977 struct rtattr *ulpinfo[INET_ULP_INFO_MAX + 1] = { 0 };
2978
2979 parse_rtattr_nested(ulpinfo, INET_ULP_INFO_MAX,
2980 tb[INET_DIAG_ULP_INFO]);
2981
2982 if (ulpinfo[INET_ULP_INFO_NAME])
2983 out(" tcp-ulp-%s",
2984 rta_getattr_str(ulpinfo[INET_ULP_INFO_NAME]));
2985
2986 if (ulpinfo[INET_ULP_INFO_TLS]) {
2987 struct rtattr *tlsinfo[TLS_INFO_MAX + 1] = { 0 };
2988
2989 parse_rtattr_nested(tlsinfo, TLS_INFO_MAX,
2990 ulpinfo[INET_ULP_INFO_TLS]);
2991
2992 tcp_tls_version(tlsinfo[TLS_INFO_VERSION]);
2993 tcp_tls_cipher(tlsinfo[TLS_INFO_CIPHER]);
2994 tcp_tls_conf("rxconf", tlsinfo[TLS_INFO_RXCONF]);
2995 tcp_tls_conf("txconf", tlsinfo[TLS_INFO_TXCONF]);
2996 }
2997 }
2998 }
2999
3000 static const char *format_host_sa(struct sockaddr_storage *sa)
3001 {
3002 union {
3003 struct sockaddr_in sin;
3004 struct sockaddr_in6 sin6;
3005 } *saddr = (void *)sa;
3006
3007 switch (sa->ss_family) {
3008 case AF_INET:
3009 return format_host(AF_INET, 4, &saddr->sin.sin_addr);
3010 case AF_INET6:
3011 return format_host(AF_INET6, 16, &saddr->sin6.sin6_addr);
3012 default:
3013 return "";
3014 }
3015 }
3016
3017 static void sctp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
3018 struct rtattr *tb[])
3019 {
3020 struct sockaddr_storage *sa;
3021 int len;
3022
3023 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
3024
3025 if (tb[INET_DIAG_LOCALS]) {
3026 len = RTA_PAYLOAD(tb[INET_DIAG_LOCALS]);
3027 sa = RTA_DATA(tb[INET_DIAG_LOCALS]);
3028
3029 out(" locals:%s", format_host_sa(sa));
3030 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
3031 out(",%s", format_host_sa(sa));
3032
3033 }
3034 if (tb[INET_DIAG_PEERS]) {
3035 len = RTA_PAYLOAD(tb[INET_DIAG_PEERS]);
3036 sa = RTA_DATA(tb[INET_DIAG_PEERS]);
3037
3038 out(" peers:%s", format_host_sa(sa));
3039 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
3040 out(",%s", format_host_sa(sa));
3041 }
3042 if (tb[INET_DIAG_INFO]) {
3043 struct sctp_info *info;
3044 len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
3045
3046 /* workaround for older kernels with less fields */
3047 if (len < sizeof(*info)) {
3048 info = alloca(sizeof(*info));
3049 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
3050 memset((char *)info + len, 0, sizeof(*info) - len);
3051 } else
3052 info = RTA_DATA(tb[INET_DIAG_INFO]);
3053
3054 sctp_stats_print(info);
3055 }
3056 }
3057
3058 static void parse_diag_msg(struct nlmsghdr *nlh, struct sockstat *s)
3059 {
3060 struct rtattr *tb[INET_DIAG_MAX+1];
3061 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3062
3063 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3064 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3065
3066 s->state = r->idiag_state;
3067 s->local.family = s->remote.family = r->idiag_family;
3068 s->lport = ntohs(r->id.idiag_sport);
3069 s->rport = ntohs(r->id.idiag_dport);
3070 s->wq = r->idiag_wqueue;
3071 s->rq = r->idiag_rqueue;
3072 s->ino = r->idiag_inode;
3073 s->uid = r->idiag_uid;
3074 s->iface = r->id.idiag_if;
3075 s->sk = cookie_sk_get(&r->id.idiag_cookie[0]);
3076
3077 s->mark = 0;
3078 if (tb[INET_DIAG_MARK])
3079 s->mark = rta_getattr_u32(tb[INET_DIAG_MARK]);
3080 if (tb[INET_DIAG_PROTOCOL])
3081 s->raw_prot = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3082 else
3083 s->raw_prot = 0;
3084
3085 if (s->local.family == AF_INET)
3086 s->local.bytelen = s->remote.bytelen = 4;
3087 else
3088 s->local.bytelen = s->remote.bytelen = 16;
3089
3090 memcpy(s->local.data, r->id.idiag_src, s->local.bytelen);
3091 memcpy(s->remote.data, r->id.idiag_dst, s->local.bytelen);
3092 }
3093
3094 static int inet_show_sock(struct nlmsghdr *nlh,
3095 struct sockstat *s)
3096 {
3097 struct rtattr *tb[INET_DIAG_MAX+1];
3098 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3099 unsigned char v6only = 0;
3100
3101 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3102 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3103
3104 if (tb[INET_DIAG_PROTOCOL])
3105 s->type = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3106
3107 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3108 v6only = rta_getattr_u8(tb[INET_DIAG_SKV6ONLY]);
3109
3110 inet_stats_print(s, v6only);
3111
3112 if (show_options) {
3113 struct tcpstat t = {};
3114
3115 t.timer = r->idiag_timer;
3116 t.timeout = r->idiag_expires;
3117 t.retrans = r->idiag_retrans;
3118 if (s->type == IPPROTO_SCTP)
3119 sctp_timer_print(&t);
3120 else
3121 tcp_timer_print(&t);
3122 }
3123
3124 if (show_details) {
3125 sock_details_print(s);
3126 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3127 out(" v6only:%u", v6only);
3128
3129 if (tb[INET_DIAG_SHUTDOWN]) {
3130 unsigned char mask;
3131
3132 mask = rta_getattr_u8(tb[INET_DIAG_SHUTDOWN]);
3133 out(" %c-%c",
3134 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3135 }
3136 }
3137
3138 if (show_tos) {
3139 if (tb[INET_DIAG_TOS])
3140 out(" tos:%#x", rta_getattr_u8(tb[INET_DIAG_TOS]));
3141 if (tb[INET_DIAG_TCLASS])
3142 out(" tclass:%#x", rta_getattr_u8(tb[INET_DIAG_TCLASS]));
3143 if (tb[INET_DIAG_CLASS_ID])
3144 out(" class_id:%#x", rta_getattr_u32(tb[INET_DIAG_CLASS_ID]));
3145 }
3146
3147 if (show_mem || (show_tcpinfo && s->type != IPPROTO_UDP)) {
3148 if (!oneline)
3149 out("\n\t");
3150 if (s->type == IPPROTO_SCTP)
3151 sctp_show_info(nlh, r, tb);
3152 else
3153 tcp_show_info(nlh, r, tb);
3154 }
3155 sctp_ino = s->ino;
3156
3157 return 0;
3158 }
3159
3160 static int tcpdiag_send(int fd, int protocol, struct filter *f)
3161 {
3162 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3163 struct {
3164 struct nlmsghdr nlh;
3165 struct inet_diag_req r;
3166 } req = {
3167 .nlh.nlmsg_len = sizeof(req),
3168 .nlh.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST,
3169 .nlh.nlmsg_seq = MAGIC_SEQ,
3170 .r.idiag_family = AF_INET,
3171 .r.idiag_states = f->states,
3172 };
3173 char *bc = NULL;
3174 int bclen;
3175 struct msghdr msg;
3176 struct rtattr rta;
3177 struct iovec iov[3];
3178 int iovlen = 1;
3179
3180 if (protocol == IPPROTO_UDP)
3181 return -1;
3182
3183 if (protocol == IPPROTO_TCP)
3184 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
3185 else
3186 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
3187 if (show_mem) {
3188 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3189 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3190 }
3191
3192 if (show_tcpinfo) {
3193 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3194 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3195 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3196 }
3197
3198 if (show_tos) {
3199 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3200 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3201 }
3202
3203 iov[0] = (struct iovec){
3204 .iov_base = &req,
3205 .iov_len = sizeof(req)
3206 };
3207 if (f->f) {
3208 bclen = ssfilter_bytecompile(f->f, &bc);
3209 if (bclen) {
3210 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3211 rta.rta_len = RTA_LENGTH(bclen);
3212 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3213 iov[2] = (struct iovec){ bc, bclen };
3214 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3215 iovlen = 3;
3216 }
3217 }
3218
3219 msg = (struct msghdr) {
3220 .msg_name = (void *)&nladdr,
3221 .msg_namelen = sizeof(nladdr),
3222 .msg_iov = iov,
3223 .msg_iovlen = iovlen,
3224 };
3225
3226 if (sendmsg(fd, &msg, 0) < 0) {
3227 close(fd);
3228 return -1;
3229 }
3230
3231 return 0;
3232 }
3233
3234 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
3235 {
3236 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3237 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3238 char *bc = NULL;
3239 int bclen;
3240 struct msghdr msg;
3241 struct rtattr rta;
3242 struct iovec iov[3];
3243 int iovlen = 1;
3244
3245 if (family == PF_UNSPEC)
3246 return tcpdiag_send(fd, protocol, f);
3247
3248 memset(&req.r, 0, sizeof(req.r));
3249 req.r.sdiag_family = family;
3250 req.r.sdiag_protocol = protocol;
3251 req.r.idiag_states = f->states;
3252 if (show_mem) {
3253 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3254 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3255 }
3256
3257 if (show_tcpinfo) {
3258 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3259 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3260 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3261 }
3262
3263 if (show_tos) {
3264 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3265 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3266 }
3267
3268 iov[0] = (struct iovec){
3269 .iov_base = &req,
3270 .iov_len = sizeof(req)
3271 };
3272 if (f->f) {
3273 bclen = ssfilter_bytecompile(f->f, &bc);
3274 if (bclen) {
3275 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3276 rta.rta_len = RTA_LENGTH(bclen);
3277 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3278 iov[2] = (struct iovec){ bc, bclen };
3279 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3280 iovlen = 3;
3281 }
3282 }
3283
3284 msg = (struct msghdr) {
3285 .msg_name = (void *)&nladdr,
3286 .msg_namelen = sizeof(nladdr),
3287 .msg_iov = iov,
3288 .msg_iovlen = iovlen,
3289 };
3290
3291 if (sendmsg(fd, &msg, 0) < 0) {
3292 close(fd);
3293 return -1;
3294 }
3295
3296 return 0;
3297 }
3298
3299 struct inet_diag_arg {
3300 struct filter *f;
3301 int protocol;
3302 struct rtnl_handle *rth;
3303 };
3304
3305 static int kill_inet_sock(struct nlmsghdr *h, void *arg, struct sockstat *s)
3306 {
3307 struct inet_diag_msg *d = NLMSG_DATA(h);
3308 struct inet_diag_arg *diag_arg = arg;
3309 struct rtnl_handle *rth = diag_arg->rth;
3310
3311 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3312
3313 req.nlh.nlmsg_type = SOCK_DESTROY;
3314 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
3315 req.nlh.nlmsg_seq = ++rth->seq;
3316 req.r.sdiag_family = d->idiag_family;
3317 req.r.sdiag_protocol = diag_arg->protocol;
3318 req.r.id = d->id;
3319
3320 if (diag_arg->protocol == IPPROTO_RAW) {
3321 struct inet_diag_req_raw *raw = (void *)&req.r;
3322
3323 BUILD_BUG_ON(sizeof(req.r) != sizeof(*raw));
3324 raw->sdiag_raw_protocol = s->raw_prot;
3325 }
3326
3327 return rtnl_talk(rth, &req.nlh, NULL);
3328 }
3329
3330 static int show_one_inet_sock(struct nlmsghdr *h, void *arg)
3331 {
3332 int err;
3333 struct inet_diag_arg *diag_arg = arg;
3334 struct inet_diag_msg *r = NLMSG_DATA(h);
3335 struct sockstat s = {};
3336
3337 if (!(diag_arg->f->families & FAMILY_MASK(r->idiag_family)))
3338 return 0;
3339
3340 parse_diag_msg(h, &s);
3341 s.type = diag_arg->protocol;
3342
3343 if (diag_arg->f->f && run_ssfilter(diag_arg->f->f, &s) == 0)
3344 return 0;
3345
3346 if (diag_arg->f->kill && kill_inet_sock(h, arg, &s) != 0) {
3347 if (errno == EOPNOTSUPP || errno == ENOENT) {
3348 /* Socket can't be closed, or is already closed. */
3349 return 0;
3350 } else {
3351 perror("SOCK_DESTROY answers");
3352 return -1;
3353 }
3354 }
3355
3356 err = inet_show_sock(h, &s);
3357 if (err < 0)
3358 return err;
3359
3360 return 0;
3361 }
3362
3363 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
3364 {
3365 int err = 0;
3366 struct rtnl_handle rth, rth2;
3367 int family = PF_INET;
3368 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
3369
3370 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3371 return -1;
3372
3373 if (f->kill) {
3374 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
3375 rtnl_close(&rth);
3376 return -1;
3377 }
3378 arg.rth = &rth2;
3379 }
3380
3381 rth.dump = MAGIC_SEQ;
3382 rth.dump_fp = dump_fp;
3383 if (preferred_family == PF_INET6)
3384 family = PF_INET6;
3385
3386 again:
3387 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
3388 goto Exit;
3389
3390 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
3391 if (family != PF_UNSPEC) {
3392 family = PF_UNSPEC;
3393 goto again;
3394 }
3395 goto Exit;
3396 }
3397 if (family == PF_INET && preferred_family != PF_INET) {
3398 family = PF_INET6;
3399 goto again;
3400 }
3401
3402 Exit:
3403 rtnl_close(&rth);
3404 if (arg.rth)
3405 rtnl_close(arg.rth);
3406 return err;
3407 }
3408
3409 static int tcp_show_netlink_file(struct filter *f)
3410 {
3411 FILE *fp;
3412 char buf[16384];
3413 int err = -1;
3414
3415 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
3416 perror("fopen($TCPDIAG_FILE)");
3417 return err;
3418 }
3419
3420 while (1) {
3421 int status, err2;
3422 struct nlmsghdr *h = (struct nlmsghdr *)buf;
3423 struct sockstat s = {};
3424
3425 status = fread(buf, 1, sizeof(*h), fp);
3426 if (status < 0) {
3427 perror("Reading header from $TCPDIAG_FILE");
3428 break;
3429 }
3430 if (status != sizeof(*h)) {
3431 perror("Unexpected EOF reading $TCPDIAG_FILE");
3432 break;
3433 }
3434
3435 status = fread(h+1, 1, NLMSG_ALIGN(h->nlmsg_len-sizeof(*h)), fp);
3436
3437 if (status < 0) {
3438 perror("Reading $TCPDIAG_FILE");
3439 break;
3440 }
3441 if (status + sizeof(*h) < h->nlmsg_len) {
3442 perror("Unexpected EOF reading $TCPDIAG_FILE");
3443 break;
3444 }
3445
3446 /* The only legal exit point */
3447 if (h->nlmsg_type == NLMSG_DONE) {
3448 err = 0;
3449 break;
3450 }
3451
3452 if (h->nlmsg_type == NLMSG_ERROR) {
3453 struct nlmsgerr *err = (struct nlmsgerr *)NLMSG_DATA(h);
3454
3455 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
3456 fprintf(stderr, "ERROR truncated\n");
3457 } else {
3458 errno = -err->error;
3459 perror("TCPDIAG answered");
3460 }
3461 break;
3462 }
3463
3464 parse_diag_msg(h, &s);
3465 s.type = IPPROTO_TCP;
3466
3467 if (f && f->f && run_ssfilter(f->f, &s) == 0)
3468 continue;
3469
3470 err2 = inet_show_sock(h, &s);
3471 if (err2 < 0) {
3472 err = err2;
3473 break;
3474 }
3475 }
3476
3477 fclose(fp);
3478 return err;
3479 }
3480
3481 static int tcp_show(struct filter *f)
3482 {
3483 FILE *fp = NULL;
3484 char *buf = NULL;
3485 int bufsize = 1024*1024;
3486
3487 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3488 return 0;
3489
3490 dg_proto = TCP_PROTO;
3491
3492 if (getenv("TCPDIAG_FILE"))
3493 return tcp_show_netlink_file(f);
3494
3495 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
3496 && inet_show_netlink(f, NULL, IPPROTO_TCP) == 0)
3497 return 0;
3498
3499 /* Sigh... We have to parse /proc/net/tcp... */
3500 while (bufsize >= 64*1024) {
3501 if ((buf = malloc(bufsize)) != NULL)
3502 break;
3503 bufsize /= 2;
3504 }
3505 if (buf == NULL) {
3506 errno = ENOMEM;
3507 return -1;
3508 }
3509
3510 if (f->families & FAMILY_MASK(AF_INET)) {
3511 if ((fp = net_tcp_open()) == NULL)
3512 goto outerr;
3513
3514 setbuffer(fp, buf, bufsize);
3515 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
3516 goto outerr;
3517 fclose(fp);
3518 }
3519
3520 if ((f->families & FAMILY_MASK(AF_INET6)) &&
3521 (fp = net_tcp6_open()) != NULL) {
3522 setbuffer(fp, buf, bufsize);
3523 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
3524 goto outerr;
3525 fclose(fp);
3526 }
3527
3528 free(buf);
3529 return 0;
3530
3531 outerr:
3532 do {
3533 int saved_errno = errno;
3534
3535 free(buf);
3536 if (fp)
3537 fclose(fp);
3538 errno = saved_errno;
3539 return -1;
3540 } while (0);
3541 }
3542
3543 static int dccp_show(struct filter *f)
3544 {
3545 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3546 return 0;
3547
3548 if (!getenv("PROC_NET_DCCP") && !getenv("PROC_ROOT")
3549 && inet_show_netlink(f, NULL, IPPROTO_DCCP) == 0)
3550 return 0;
3551
3552 return 0;
3553 }
3554
3555 static int sctp_show(struct filter *f)
3556 {
3557 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3558 return 0;
3559
3560 if (!getenv("PROC_NET_SCTP") && !getenv("PROC_ROOT")
3561 && inet_show_netlink(f, NULL, IPPROTO_SCTP) == 0)
3562 return 0;
3563
3564 return 0;
3565 }
3566
3567 static int dgram_show_line(char *line, const struct filter *f, int family)
3568 {
3569 struct sockstat s = {};
3570 char *loc, *rem, *data;
3571 char opt[256];
3572 int n;
3573
3574 if (proc_inet_split_line(line, &loc, &rem, &data))
3575 return -1;
3576
3577 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
3578
3579 if (!(f->states & (1 << state)))
3580 return 0;
3581
3582 proc_parse_inet_addr(loc, rem, family, &s);
3583
3584 if (f->f && run_ssfilter(f->f, &s) == 0)
3585 return 0;
3586
3587 opt[0] = 0;
3588 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
3589 &s.state, &s.wq, &s.rq,
3590 &s.uid, &s.ino,
3591 &s.refcnt, &s.sk, opt);
3592
3593 if (n < 9)
3594 opt[0] = 0;
3595
3596 s.type = dg_proto == UDP_PROTO ? IPPROTO_UDP : 0;
3597 inet_stats_print(&s, false);
3598
3599 if (show_details && opt[0])
3600 out(" opt:\"%s\"", opt);
3601
3602 return 0;
3603 }
3604
3605 static int udp_show(struct filter *f)
3606 {
3607 FILE *fp = NULL;
3608
3609 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3610 return 0;
3611
3612 dg_proto = UDP_PROTO;
3613
3614 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
3615 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
3616 return 0;
3617
3618 if (f->families&FAMILY_MASK(AF_INET)) {
3619 if ((fp = net_udp_open()) == NULL)
3620 goto outerr;
3621 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3622 goto outerr;
3623 fclose(fp);
3624 }
3625
3626 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3627 (fp = net_udp6_open()) != NULL) {
3628 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3629 goto outerr;
3630 fclose(fp);
3631 }
3632 return 0;
3633
3634 outerr:
3635 do {
3636 int saved_errno = errno;
3637
3638 if (fp)
3639 fclose(fp);
3640 errno = saved_errno;
3641 return -1;
3642 } while (0);
3643 }
3644
3645 static int raw_show(struct filter *f)
3646 {
3647 FILE *fp = NULL;
3648
3649 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3650 return 0;
3651
3652 dg_proto = RAW_PROTO;
3653
3654 if (!getenv("PROC_NET_RAW") && !getenv("PROC_ROOT") &&
3655 inet_show_netlink(f, NULL, IPPROTO_RAW) == 0)
3656 return 0;
3657
3658 if (f->families&FAMILY_MASK(AF_INET)) {
3659 if ((fp = net_raw_open()) == NULL)
3660 goto outerr;
3661 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3662 goto outerr;
3663 fclose(fp);
3664 }
3665
3666 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3667 (fp = net_raw6_open()) != NULL) {
3668 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3669 goto outerr;
3670 fclose(fp);
3671 }
3672 return 0;
3673
3674 outerr:
3675 do {
3676 int saved_errno = errno;
3677
3678 if (fp)
3679 fclose(fp);
3680 errno = saved_errno;
3681 return -1;
3682 } while (0);
3683 }
3684
3685 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
3686
3687 static void unix_list_drop_first(struct sockstat **list)
3688 {
3689 struct sockstat *s = *list;
3690
3691 (*list) = (*list)->next;
3692 free(s->name);
3693 free(s);
3694 }
3695
3696 static bool unix_type_skip(struct sockstat *s, struct filter *f)
3697 {
3698 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
3699 return true;
3700 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
3701 return true;
3702 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
3703 return true;
3704 return false;
3705 }
3706
3707 static void unix_stats_print(struct sockstat *s, struct filter *f)
3708 {
3709 char port_name[30] = {};
3710
3711 sock_state_print(s);
3712
3713 sock_addr_print(s->name ?: "*", " ",
3714 int_to_str(s->lport, port_name), NULL);
3715 sock_addr_print(s->peer_name ?: "*", " ",
3716 int_to_str(s->rport, port_name), NULL);
3717
3718 proc_ctx_print(s);
3719 }
3720
3721 static int unix_show_sock(struct nlmsghdr *nlh, void *arg)
3722 {
3723 struct filter *f = (struct filter *)arg;
3724 struct unix_diag_msg *r = NLMSG_DATA(nlh);
3725 struct rtattr *tb[UNIX_DIAG_MAX+1];
3726 char name[128];
3727 struct sockstat stat = { .name = "*", .peer_name = "*" };
3728
3729 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr *)(r+1),
3730 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3731
3732 stat.type = r->udiag_type;
3733 stat.state = r->udiag_state;
3734 stat.ino = stat.lport = r->udiag_ino;
3735 stat.local.family = stat.remote.family = AF_UNIX;
3736
3737 if (unix_type_skip(&stat, f))
3738 return 0;
3739
3740 if (tb[UNIX_DIAG_RQLEN]) {
3741 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
3742
3743 stat.rq = rql->udiag_rqueue;
3744 stat.wq = rql->udiag_wqueue;
3745 }
3746 if (tb[UNIX_DIAG_NAME]) {
3747 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
3748
3749 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
3750 name[len] = '\0';
3751 if (name[0] == '\0') {
3752 int i;
3753 for (i = 0; i < len; i++)
3754 if (name[i] == '\0')
3755 name[i] = '@';
3756 }
3757 stat.name = &name[0];
3758 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
3759 }
3760 if (tb[UNIX_DIAG_PEER])
3761 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
3762
3763 if (f->f && run_ssfilter(f->f, &stat) == 0)
3764 return 0;
3765
3766 unix_stats_print(&stat, f);
3767
3768 if (show_mem)
3769 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
3770 if (show_details) {
3771 if (tb[UNIX_DIAG_SHUTDOWN]) {
3772 unsigned char mask;
3773
3774 mask = rta_getattr_u8(tb[UNIX_DIAG_SHUTDOWN]);
3775 out(" %c-%c",
3776 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3777 }
3778 if (tb[UNIX_DIAG_VFS]) {
3779 struct unix_diag_vfs *uv = RTA_DATA(tb[UNIX_DIAG_VFS]);
3780
3781 out(" ino:%u dev:%u/%u", uv->udiag_vfs_ino, major(uv->udiag_vfs_dev),
3782 minor(uv->udiag_vfs_dev));
3783 }
3784 if (tb[UNIX_DIAG_ICONS]) {
3785 int len = RTA_PAYLOAD(tb[UNIX_DIAG_ICONS]);
3786 __u32 *peers = RTA_DATA(tb[UNIX_DIAG_ICONS]);
3787 int i;
3788
3789 out(" peers:");
3790 for (i = 0; i < len / sizeof(__u32); i++)
3791 out(" %u", peers[i]);
3792 }
3793 }
3794
3795 return 0;
3796 }
3797
3798 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
3799 size_t size, rtnl_filter_t show_one_sock)
3800 {
3801 int ret = -1;
3802 struct rtnl_handle rth;
3803
3804 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3805 return -1;
3806
3807 rth.dump = MAGIC_SEQ;
3808
3809 if (rtnl_send(&rth, req, size) < 0)
3810 goto Exit;
3811
3812 if (rtnl_dump_filter(&rth, show_one_sock, f))
3813 goto Exit;
3814
3815 ret = 0;
3816 Exit:
3817 rtnl_close(&rth);
3818 return ret;
3819 }
3820
3821 static int unix_show_netlink(struct filter *f)
3822 {
3823 DIAG_REQUEST(req, struct unix_diag_req r);
3824
3825 req.r.sdiag_family = AF_UNIX;
3826 req.r.udiag_states = f->states;
3827 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
3828 if (show_mem)
3829 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
3830 if (show_details)
3831 req.r.udiag_show |= UDIAG_SHOW_VFS | UDIAG_SHOW_ICONS;
3832
3833 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
3834 }
3835
3836 static int unix_show(struct filter *f)
3837 {
3838 FILE *fp;
3839 char buf[256];
3840 char name[128];
3841 int newformat = 0;
3842 int cnt;
3843 struct sockstat *list = NULL;
3844 const int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
3845 SS_ESTABLISHED, SS_CLOSING };
3846
3847 if (!filter_af_get(f, AF_UNIX))
3848 return 0;
3849
3850 if (!getenv("PROC_NET_UNIX") && !getenv("PROC_ROOT")
3851 && unix_show_netlink(f) == 0)
3852 return 0;
3853
3854 if ((fp = net_unix_open()) == NULL)
3855 return -1;
3856 if (!fgets(buf, sizeof(buf), fp)) {
3857 fclose(fp);
3858 return -1;
3859 }
3860
3861 if (memcmp(buf, "Peer", 4) == 0)
3862 newformat = 1;
3863 cnt = 0;
3864
3865 while (fgets(buf, sizeof(buf), fp)) {
3866 struct sockstat *u, **insp;
3867 int flags;
3868
3869 if (!(u = calloc(1, sizeof(*u))))
3870 break;
3871
3872 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
3873 &u->rport, &u->rq, &u->wq, &flags, &u->type,
3874 &u->state, &u->ino, name) < 8)
3875 name[0] = 0;
3876
3877 u->lport = u->ino;
3878 u->local.family = u->remote.family = AF_UNIX;
3879
3880 if (flags & (1 << 16)) {
3881 u->state = SS_LISTEN;
3882 } else if (u->state > 0 &&
3883 u->state <= ARRAY_SIZE(unix_state_map)) {
3884 u->state = unix_state_map[u->state-1];
3885 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
3886 u->state = SS_ESTABLISHED;
3887 }
3888 if (unix_type_skip(u, f) ||
3889 !(f->states & (1 << u->state))) {
3890 free(u);
3891 continue;
3892 }
3893
3894 if (!newformat) {
3895 u->rport = 0;
3896 u->rq = 0;
3897 u->wq = 0;
3898 }
3899
3900 if (name[0]) {
3901 u->name = strdup(name);
3902 if (!u->name) {
3903 free(u);
3904 break;
3905 }
3906 }
3907
3908 if (u->rport) {
3909 struct sockstat *p;
3910
3911 for (p = list; p; p = p->next) {
3912 if (u->rport == p->lport)
3913 break;
3914 }
3915 if (!p)
3916 u->peer_name = "?";
3917 else
3918 u->peer_name = p->name ? : "*";
3919 }
3920
3921 if (f->f) {
3922 struct sockstat st = {
3923 .local.family = AF_UNIX,
3924 .remote.family = AF_UNIX,
3925 };
3926
3927 memcpy(st.local.data, &u->name, sizeof(u->name));
3928 /* when parsing the old format rport is set to 0 and
3929 * therefore peer_name remains NULL
3930 */
3931 if (u->peer_name && strcmp(u->peer_name, "*"))
3932 memcpy(st.remote.data, &u->peer_name,
3933 sizeof(u->peer_name));
3934 if (run_ssfilter(f->f, &st) == 0) {
3935 free(u->name);
3936 free(u);
3937 continue;
3938 }
3939 }
3940
3941 insp = &list;
3942 while (*insp) {
3943 if (u->type < (*insp)->type ||
3944 (u->type == (*insp)->type &&
3945 u->ino < (*insp)->ino))
3946 break;
3947 insp = &(*insp)->next;
3948 }
3949 u->next = *insp;
3950 *insp = u;
3951
3952 if (++cnt > MAX_UNIX_REMEMBER) {
3953 while (list) {
3954 unix_stats_print(list, f);
3955 unix_list_drop_first(&list);
3956 }
3957 cnt = 0;
3958 }
3959 }
3960 fclose(fp);
3961 while (list) {
3962 unix_stats_print(list, f);
3963 unix_list_drop_first(&list);
3964 }
3965
3966 return 0;
3967 }
3968
3969 static int packet_stats_print(struct sockstat *s, const struct filter *f)
3970 {
3971 const char *addr, *port;
3972 char ll_name[16];
3973
3974 s->local.family = s->remote.family = AF_PACKET;
3975
3976 if (f->f) {
3977 s->local.data[0] = s->prot;
3978 if (run_ssfilter(f->f, s) == 0)
3979 return 1;
3980 }
3981
3982 sock_state_print(s);
3983
3984 if (s->prot == 3)
3985 addr = "*";
3986 else
3987 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
3988
3989 if (s->iface == 0)
3990 port = "*";
3991 else
3992 port = xll_index_to_name(s->iface);
3993
3994 sock_addr_print(addr, ":", port, NULL);
3995 sock_addr_print("", "*", "", NULL);
3996
3997 proc_ctx_print(s);
3998
3999 if (show_details)
4000 sock_details_print(s);
4001
4002 return 0;
4003 }
4004
4005 static void packet_show_ring(struct packet_diag_ring *ring)
4006 {
4007 out("blk_size:%d", ring->pdr_block_size);
4008 out(",blk_nr:%d", ring->pdr_block_nr);
4009 out(",frm_size:%d", ring->pdr_frame_size);
4010 out(",frm_nr:%d", ring->pdr_frame_nr);
4011 out(",tmo:%d", ring->pdr_retire_tmo);
4012 out(",features:0x%x", ring->pdr_features);
4013 }
4014
4015 static int packet_show_sock(struct nlmsghdr *nlh, void *arg)
4016 {
4017 const struct filter *f = arg;
4018 struct packet_diag_msg *r = NLMSG_DATA(nlh);
4019 struct packet_diag_info *pinfo = NULL;
4020 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
4021 struct rtattr *tb[PACKET_DIAG_MAX+1];
4022 struct sockstat stat = {};
4023 uint32_t fanout = 0;
4024 bool has_fanout = false;
4025
4026 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr *)(r+1),
4027 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4028
4029 /* use /proc/net/packet if all info are not available */
4030 if (!tb[PACKET_DIAG_MEMINFO])
4031 return -1;
4032
4033 stat.type = r->pdiag_type;
4034 stat.prot = r->pdiag_num;
4035 stat.ino = r->pdiag_ino;
4036 stat.state = SS_CLOSE;
4037 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
4038
4039 if (tb[PACKET_DIAG_MEMINFO]) {
4040 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
4041
4042 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4043 }
4044
4045 if (tb[PACKET_DIAG_INFO]) {
4046 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
4047 stat.lport = stat.iface = pinfo->pdi_index;
4048 }
4049
4050 if (tb[PACKET_DIAG_UID])
4051 stat.uid = rta_getattr_u32(tb[PACKET_DIAG_UID]);
4052
4053 if (tb[PACKET_DIAG_RX_RING])
4054 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
4055
4056 if (tb[PACKET_DIAG_TX_RING])
4057 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
4058
4059 if (tb[PACKET_DIAG_FANOUT]) {
4060 has_fanout = true;
4061 fanout = rta_getattr_u32(tb[PACKET_DIAG_FANOUT]);
4062 }
4063
4064 if (packet_stats_print(&stat, f))
4065 return 0;
4066
4067 if (show_details) {
4068 if (pinfo) {
4069 if (oneline)
4070 out(" ver:%d", pinfo->pdi_version);
4071 else
4072 out("\n\tver:%d", pinfo->pdi_version);
4073 out(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
4074 out(" flags( ");
4075 if (pinfo->pdi_flags & PDI_RUNNING)
4076 out("running");
4077 if (pinfo->pdi_flags & PDI_AUXDATA)
4078 out(" auxdata");
4079 if (pinfo->pdi_flags & PDI_ORIGDEV)
4080 out(" origdev");
4081 if (pinfo->pdi_flags & PDI_VNETHDR)
4082 out(" vnethdr");
4083 if (pinfo->pdi_flags & PDI_LOSS)
4084 out(" loss");
4085 if (!pinfo->pdi_flags)
4086 out("0");
4087 out(" )");
4088 }
4089 if (ring_rx) {
4090 if (oneline)
4091 out(" ring_rx(");
4092 else
4093 out("\n\tring_rx(");
4094 packet_show_ring(ring_rx);
4095 out(")");
4096 }
4097 if (ring_tx) {
4098 if (oneline)
4099 out(" ring_tx(");
4100 else
4101 out("\n\tring_tx(");
4102 packet_show_ring(ring_tx);
4103 out(")");
4104 }
4105 if (has_fanout) {
4106 uint16_t type = (fanout >> 16) & 0xffff;
4107
4108 if (oneline)
4109 out(" fanout(");
4110 else
4111 out("\n\tfanout(");
4112 out("id:%d,", fanout & 0xffff);
4113 out("type:");
4114
4115 if (type == 0)
4116 out("hash");
4117 else if (type == 1)
4118 out("lb");
4119 else if (type == 2)
4120 out("cpu");
4121 else if (type == 3)
4122 out("roll");
4123 else if (type == 4)
4124 out("random");
4125 else if (type == 5)
4126 out("qm");
4127 else
4128 out("0x%x", type);
4129
4130 out(")");
4131 }
4132 }
4133
4134 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
4135 struct sock_filter *fil =
4136 RTA_DATA(tb[PACKET_DIAG_FILTER]);
4137 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
4138 sizeof(struct sock_filter);
4139
4140 if (oneline)
4141 out(" bpf filter (%d): ", num);
4142 else
4143 out("\n\tbpf filter (%d): ", num);
4144 while (num) {
4145 out(" 0x%02x %u %u %u,",
4146 fil->code, fil->jt, fil->jf, fil->k);
4147 num--;
4148 fil++;
4149 }
4150 }
4151
4152 if (show_mem)
4153 print_skmeminfo(tb, PACKET_DIAG_MEMINFO);
4154 return 0;
4155 }
4156
4157 static int packet_show_netlink(struct filter *f)
4158 {
4159 DIAG_REQUEST(req, struct packet_diag_req r);
4160
4161 req.r.sdiag_family = AF_PACKET;
4162 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
4163 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
4164
4165 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
4166 }
4167
4168 static int packet_show_line(char *buf, const struct filter *f, int fam)
4169 {
4170 unsigned long long sk;
4171 struct sockstat stat = {};
4172 int type, prot, iface, state, rq, uid, ino;
4173
4174 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
4175 &sk,
4176 &type, &prot, &iface, &state,
4177 &rq, &uid, &ino);
4178
4179 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
4180 return 0;
4181 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
4182 return 0;
4183
4184 stat.type = type;
4185 stat.prot = prot;
4186 stat.lport = stat.iface = iface;
4187 stat.state = state;
4188 stat.rq = rq;
4189 stat.uid = uid;
4190 stat.ino = ino;
4191 stat.state = SS_CLOSE;
4192
4193 if (packet_stats_print(&stat, f))
4194 return 0;
4195
4196 return 0;
4197 }
4198
4199 static int packet_show(struct filter *f)
4200 {
4201 FILE *fp;
4202 int rc = 0;
4203
4204 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
4205 return 0;
4206
4207 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
4208 packet_show_netlink(f) == 0)
4209 return 0;
4210
4211 if ((fp = net_packet_open()) == NULL)
4212 return -1;
4213 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
4214 rc = -1;
4215
4216 fclose(fp);
4217 return rc;
4218 }
4219
4220 static int xdp_stats_print(struct sockstat *s, const struct filter *f)
4221 {
4222 const char *addr, *port;
4223 char q_str[16];
4224
4225 s->local.family = s->remote.family = AF_XDP;
4226
4227 if (f->f) {
4228 if (run_ssfilter(f->f, s) == 0)
4229 return 1;
4230 }
4231
4232 sock_state_print(s);
4233
4234 if (s->iface) {
4235 addr = xll_index_to_name(s->iface);
4236 snprintf(q_str, sizeof(q_str), "q%d", s->lport);
4237 port = q_str;
4238 sock_addr_print(addr, ":", port, NULL);
4239 } else {
4240 sock_addr_print("", "*", "", NULL);
4241 }
4242
4243 sock_addr_print("", "*", "", NULL);
4244
4245 proc_ctx_print(s);
4246
4247 if (show_details)
4248 sock_details_print(s);
4249
4250 return 0;
4251 }
4252
4253 static void xdp_show_ring(const char *name, struct xdp_diag_ring *ring)
4254 {
4255 if (oneline)
4256 out(" %s(", name);
4257 else
4258 out("\n\t%s(", name);
4259 out("entries:%u", ring->entries);
4260 out(")");
4261 }
4262
4263 static void xdp_show_umem(struct xdp_diag_umem *umem, struct xdp_diag_ring *fr,
4264 struct xdp_diag_ring *cr)
4265 {
4266 if (oneline)
4267 out(" tumem(");
4268 else
4269 out("\n\tumem(");
4270 out("id:%u", umem->id);
4271 out(",size:%llu", umem->size);
4272 out(",num_pages:%u", umem->num_pages);
4273 out(",chunk_size:%u", umem->chunk_size);
4274 out(",headroom:%u", umem->headroom);
4275 out(",ifindex:%u", umem->ifindex);
4276 out(",qid:%u", umem->queue_id);
4277 out(",zc:%u", umem->flags & XDP_DU_F_ZEROCOPY);
4278 out(",refs:%u", umem->refs);
4279 out(")");
4280
4281 if (fr)
4282 xdp_show_ring("fr", fr);
4283 if (cr)
4284 xdp_show_ring("cr", cr);
4285 }
4286
4287 static int xdp_show_sock(struct nlmsghdr *nlh, void *arg)
4288 {
4289 struct xdp_diag_ring *rx = NULL, *tx = NULL, *fr = NULL, *cr = NULL;
4290 struct xdp_diag_msg *msg = NLMSG_DATA(nlh);
4291 struct rtattr *tb[XDP_DIAG_MAX + 1];
4292 struct xdp_diag_info *info = NULL;
4293 struct xdp_diag_umem *umem = NULL;
4294 const struct filter *f = arg;
4295 struct sockstat stat = {};
4296
4297 parse_rtattr(tb, XDP_DIAG_MAX, (struct rtattr *)(msg + 1),
4298 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*msg)));
4299
4300 stat.type = msg->xdiag_type;
4301 stat.ino = msg->xdiag_ino;
4302 stat.state = SS_CLOSE;
4303 stat.sk = cookie_sk_get(&msg->xdiag_cookie[0]);
4304
4305 if (tb[XDP_DIAG_INFO]) {
4306 info = RTA_DATA(tb[XDP_DIAG_INFO]);
4307 stat.iface = info->ifindex;
4308 stat.lport = info->queue_id;
4309 }
4310
4311 if (tb[XDP_DIAG_UID])
4312 stat.uid = rta_getattr_u32(tb[XDP_DIAG_UID]);
4313 if (tb[XDP_DIAG_RX_RING])
4314 rx = RTA_DATA(tb[XDP_DIAG_RX_RING]);
4315 if (tb[XDP_DIAG_TX_RING])
4316 tx = RTA_DATA(tb[XDP_DIAG_TX_RING]);
4317 if (tb[XDP_DIAG_UMEM])
4318 umem = RTA_DATA(tb[XDP_DIAG_UMEM]);
4319 if (tb[XDP_DIAG_UMEM_FILL_RING])
4320 fr = RTA_DATA(tb[XDP_DIAG_UMEM_FILL_RING]);
4321 if (tb[XDP_DIAG_UMEM_COMPLETION_RING])
4322 cr = RTA_DATA(tb[XDP_DIAG_UMEM_COMPLETION_RING]);
4323 if (tb[XDP_DIAG_MEMINFO]) {
4324 __u32 *skmeminfo = RTA_DATA(tb[XDP_DIAG_MEMINFO]);
4325
4326 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4327 }
4328
4329 if (xdp_stats_print(&stat, f))
4330 return 0;
4331
4332 if (show_details) {
4333 if (rx)
4334 xdp_show_ring("rx", rx);
4335 if (tx)
4336 xdp_show_ring("tx", tx);
4337 if (umem)
4338 xdp_show_umem(umem, fr, cr);
4339 }
4340
4341 if (show_mem)
4342 print_skmeminfo(tb, XDP_DIAG_MEMINFO); // really?
4343
4344
4345 return 0;
4346 }
4347
4348 static int xdp_show(struct filter *f)
4349 {
4350 DIAG_REQUEST(req, struct xdp_diag_req r);
4351
4352 if (!filter_af_get(f, AF_XDP) || !(f->states & (1 << SS_CLOSE)))
4353 return 0;
4354
4355 req.r.sdiag_family = AF_XDP;
4356 req.r.xdiag_show = XDP_SHOW_INFO | XDP_SHOW_RING_CFG | XDP_SHOW_UMEM |
4357 XDP_SHOW_MEMINFO;
4358
4359 return handle_netlink_request(f, &req.nlh, sizeof(req), xdp_show_sock);
4360 }
4361
4362 static int netlink_show_one(struct filter *f,
4363 int prot, int pid, unsigned int groups,
4364 int state, int dst_pid, unsigned int dst_group,
4365 int rq, int wq,
4366 unsigned long long sk, unsigned long long cb)
4367 {
4368 struct sockstat st = {
4369 .state = SS_CLOSE,
4370 .rq = rq,
4371 .wq = wq,
4372 .local.family = AF_NETLINK,
4373 .remote.family = AF_NETLINK,
4374 };
4375
4376 SPRINT_BUF(prot_buf) = {};
4377 const char *prot_name;
4378 char procname[64] = {};
4379
4380 if (f->f) {
4381 st.rport = -1;
4382 st.lport = pid;
4383 st.local.data[0] = prot;
4384 if (run_ssfilter(f->f, &st) == 0)
4385 return 1;
4386 }
4387
4388 sock_state_print(&st);
4389
4390 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
4391
4392 if (pid == -1) {
4393 procname[0] = '*';
4394 } else if (!numeric) {
4395 int done = 0;
4396
4397 if (!pid) {
4398 done = 1;
4399 strncpy(procname, "kernel", 7);
4400 } else if (pid > 0) {
4401 FILE *fp;
4402
4403 snprintf(procname, sizeof(procname), "%s/%d/stat",
4404 getenv("PROC_ROOT") ? : "/proc", pid);
4405 if ((fp = fopen(procname, "r")) != NULL) {
4406 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
4407 snprintf(procname+strlen(procname),
4408 sizeof(procname)-strlen(procname),
4409 "/%d", pid);
4410 done = 1;
4411 }
4412 fclose(fp);
4413 }
4414 }
4415 if (!done)
4416 int_to_str(pid, procname);
4417 } else {
4418 int_to_str(pid, procname);
4419 }
4420
4421 sock_addr_print(prot_name, ":", procname, NULL);
4422
4423 if (state == NETLINK_CONNECTED) {
4424 char dst_group_buf[30];
4425 char dst_pid_buf[30];
4426
4427 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
4428 int_to_str(dst_pid, dst_pid_buf), NULL);
4429 } else {
4430 sock_addr_print("", "*", "", NULL);
4431 }
4432
4433 char *pid_context = NULL;
4434
4435 if (show_proc_ctx) {
4436 /* The pid value will either be:
4437 * 0 if destination kernel - show kernel initial context.
4438 * A valid process pid - use getpidcon.
4439 * A unique value allocated by the kernel or netlink user
4440 * to the process - show context as "not available".
4441 */
4442 if (!pid)
4443 security_get_initial_context("kernel", &pid_context);
4444 else if (pid > 0)
4445 getpidcon(pid, &pid_context);
4446
4447 out(" proc_ctx=%s", pid_context ? : "unavailable");
4448 free(pid_context);
4449 }
4450
4451 if (show_details) {
4452 out(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
4453 }
4454
4455 return 0;
4456 }
4457
4458 static int netlink_show_sock(struct nlmsghdr *nlh, void *arg)
4459 {
4460 struct filter *f = (struct filter *)arg;
4461 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
4462 struct rtattr *tb[NETLINK_DIAG_MAX+1];
4463 int rq = 0, wq = 0;
4464 unsigned long groups = 0;
4465
4466 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr *)(r+1),
4467 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4468
4469 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
4470 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
4471
4472 if (tb[NETLINK_DIAG_MEMINFO]) {
4473 const __u32 *skmeminfo;
4474
4475 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
4476
4477 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4478 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
4479 }
4480
4481 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
4482 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
4483 rq, wq, 0, 0)) {
4484 return 0;
4485 }
4486
4487 if (show_mem) {
4488 out("\t");
4489 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
4490 }
4491
4492 return 0;
4493 }
4494
4495 static int netlink_show_netlink(struct filter *f)
4496 {
4497 DIAG_REQUEST(req, struct netlink_diag_req r);
4498
4499 req.r.sdiag_family = AF_NETLINK;
4500 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
4501 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
4502
4503 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
4504 }
4505
4506 static int netlink_show(struct filter *f)
4507 {
4508 FILE *fp;
4509 char buf[256];
4510 int prot, pid;
4511 unsigned int groups;
4512 int rq, wq, rc;
4513 unsigned long long sk, cb;
4514
4515 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
4516 return 0;
4517
4518 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
4519 netlink_show_netlink(f) == 0)
4520 return 0;
4521
4522 if ((fp = net_netlink_open()) == NULL)
4523 return -1;
4524 if (!fgets(buf, sizeof(buf), fp)) {
4525 fclose(fp);
4526 return -1;
4527 }
4528
4529 while (fgets(buf, sizeof(buf), fp)) {
4530 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
4531 &sk,
4532 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
4533
4534 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
4535 }
4536
4537 fclose(fp);
4538 return 0;
4539 }
4540
4541 static bool vsock_type_skip(struct sockstat *s, struct filter *f)
4542 {
4543 if (s->type == SOCK_STREAM && !(f->dbs & (1 << VSOCK_ST_DB)))
4544 return true;
4545 if (s->type == SOCK_DGRAM && !(f->dbs & (1 << VSOCK_DG_DB)))
4546 return true;
4547 return false;
4548 }
4549
4550 static void vsock_addr_print(inet_prefix *a, __u32 port)
4551 {
4552 char cid_str[sizeof("4294967295")];
4553 char port_str[sizeof("4294967295")];
4554 __u32 cid;
4555
4556 memcpy(&cid, a->data, sizeof(cid));
4557
4558 if (cid == ~(__u32)0)
4559 snprintf(cid_str, sizeof(cid_str), "*");
4560 else
4561 snprintf(cid_str, sizeof(cid_str), "%u", cid);
4562
4563 if (port == ~(__u32)0)
4564 snprintf(port_str, sizeof(port_str), "*");
4565 else
4566 snprintf(port_str, sizeof(port_str), "%u", port);
4567
4568 sock_addr_print(cid_str, ":", port_str, NULL);
4569 }
4570
4571 static void vsock_stats_print(struct sockstat *s, struct filter *f)
4572 {
4573 sock_state_print(s);
4574
4575 vsock_addr_print(&s->local, s->lport);
4576 vsock_addr_print(&s->remote, s->rport);
4577
4578 proc_ctx_print(s);
4579 }
4580
4581 static int vsock_show_sock(struct nlmsghdr *nlh, void *arg)
4582 {
4583 struct filter *f = (struct filter *)arg;
4584 struct vsock_diag_msg *r = NLMSG_DATA(nlh);
4585 struct sockstat stat = {
4586 .type = r->vdiag_type,
4587 .lport = r->vdiag_src_port,
4588 .rport = r->vdiag_dst_port,
4589 .state = r->vdiag_state,
4590 .ino = r->vdiag_ino,
4591 };
4592
4593 vsock_set_inet_prefix(&stat.local, r->vdiag_src_cid);
4594 vsock_set_inet_prefix(&stat.remote, r->vdiag_dst_cid);
4595
4596 if (vsock_type_skip(&stat, f))
4597 return 0;
4598
4599 if (f->f && run_ssfilter(f->f, &stat) == 0)
4600 return 0;
4601
4602 vsock_stats_print(&stat, f);
4603
4604 return 0;
4605 }
4606
4607 static int vsock_show(struct filter *f)
4608 {
4609 DIAG_REQUEST(req, struct vsock_diag_req r);
4610
4611 if (!filter_af_get(f, AF_VSOCK))
4612 return 0;
4613
4614 req.r.sdiag_family = AF_VSOCK;
4615 req.r.vdiag_states = f->states;
4616
4617 return handle_netlink_request(f, &req.nlh, sizeof(req), vsock_show_sock);
4618 }
4619
4620 static void tipc_sock_addr_print(struct rtattr *net_addr, struct rtattr *id)
4621 {
4622 uint32_t node = rta_getattr_u32(net_addr);
4623 uint32_t identity = rta_getattr_u32(id);
4624
4625 SPRINT_BUF(addr) = {};
4626 SPRINT_BUF(port) = {};
4627
4628 sprintf(addr, "%u", node);
4629 sprintf(port, "%u", identity);
4630 sock_addr_print(addr, ":", port, NULL);
4631
4632 }
4633
4634 static int tipc_show_sock(struct nlmsghdr *nlh, void *arg)
4635 {
4636 struct rtattr *stat[TIPC_NLA_SOCK_STAT_MAX + 1] = {};
4637 struct rtattr *attrs[TIPC_NLA_SOCK_MAX + 1] = {};
4638 struct rtattr *con[TIPC_NLA_CON_MAX + 1] = {};
4639 struct rtattr *info[TIPC_NLA_MAX + 1] = {};
4640 struct rtattr *msg_ref;
4641 struct sockstat ss = {};
4642
4643 parse_rtattr(info, TIPC_NLA_MAX, NLMSG_DATA(nlh),
4644 NLMSG_PAYLOAD(nlh, 0));
4645
4646 if (!info[TIPC_NLA_SOCK])
4647 return 0;
4648
4649 msg_ref = info[TIPC_NLA_SOCK];
4650 parse_rtattr(attrs, TIPC_NLA_SOCK_MAX, RTA_DATA(msg_ref),
4651 RTA_PAYLOAD(msg_ref));
4652
4653 msg_ref = attrs[TIPC_NLA_SOCK_STAT];
4654 parse_rtattr(stat, TIPC_NLA_SOCK_STAT_MAX,
4655 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4656
4657
4658 ss.local.family = AF_TIPC;
4659 ss.type = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TYPE]);
4660 ss.state = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TIPC_STATE]);
4661 ss.uid = rta_getattr_u32(attrs[TIPC_NLA_SOCK_UID]);
4662 ss.ino = rta_getattr_u32(attrs[TIPC_NLA_SOCK_INO]);
4663 ss.rq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_RCVQ]);
4664 ss.wq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_SENDQ]);
4665 ss.sk = rta_getattr_u64(attrs[TIPC_NLA_SOCK_COOKIE]);
4666
4667 sock_state_print (&ss);
4668
4669 tipc_sock_addr_print(attrs[TIPC_NLA_SOCK_ADDR],
4670 attrs[TIPC_NLA_SOCK_REF]);
4671
4672 msg_ref = attrs[TIPC_NLA_SOCK_CON];
4673 if (msg_ref) {
4674 parse_rtattr(con, TIPC_NLA_CON_MAX,
4675 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4676
4677 tipc_sock_addr_print(con[TIPC_NLA_CON_NODE],
4678 con[TIPC_NLA_CON_SOCK]);
4679 } else
4680 sock_addr_print("", "-", "", NULL);
4681
4682 if (show_details)
4683 sock_details_print(&ss);
4684
4685 proc_ctx_print(&ss);
4686
4687 if (show_tipcinfo) {
4688 if (oneline)
4689 out(" type:%s", stype_nameg[ss.type]);
4690 else
4691 out("\n type:%s", stype_nameg[ss.type]);
4692 out(" cong:%s ",
4693 stat[TIPC_NLA_SOCK_STAT_LINK_CONG] ? "link" :
4694 stat[TIPC_NLA_SOCK_STAT_CONN_CONG] ? "conn" : "none");
4695 out(" drop:%d ",
4696 rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_DROP]));
4697
4698 if (attrs[TIPC_NLA_SOCK_HAS_PUBL])
4699 out(" publ");
4700
4701 if (con[TIPC_NLA_CON_FLAG])
4702 out(" via {%u,%u} ",
4703 rta_getattr_u32(con[TIPC_NLA_CON_TYPE]),
4704 rta_getattr_u32(con[TIPC_NLA_CON_INST]));
4705 }
4706
4707 return 0;
4708 }
4709
4710 static int tipc_show(struct filter *f)
4711 {
4712 DIAG_REQUEST(req, struct tipc_sock_diag_req r);
4713
4714 memset(&req.r, 0, sizeof(req.r));
4715 req.r.sdiag_family = AF_TIPC;
4716 req.r.tidiag_states = f->states;
4717
4718 return handle_netlink_request(f, &req.nlh, sizeof(req), tipc_show_sock);
4719 }
4720
4721 struct sock_diag_msg {
4722 __u8 sdiag_family;
4723 };
4724
4725 static int generic_show_sock(struct nlmsghdr *nlh, void *arg)
4726 {
4727 struct sock_diag_msg *r = NLMSG_DATA(nlh);
4728 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
4729 int ret;
4730
4731 switch (r->sdiag_family) {
4732 case AF_INET:
4733 case AF_INET6:
4734 inet_arg.rth = inet_arg.f->rth_for_killing;
4735 ret = show_one_inet_sock(nlh, &inet_arg);
4736 break;
4737 case AF_UNIX:
4738 ret = unix_show_sock(nlh, arg);
4739 break;
4740 case AF_PACKET:
4741 ret = packet_show_sock(nlh, arg);
4742 break;
4743 case AF_NETLINK:
4744 ret = netlink_show_sock(nlh, arg);
4745 break;
4746 case AF_VSOCK:
4747 ret = vsock_show_sock(nlh, arg);
4748 break;
4749 case AF_XDP:
4750 ret = xdp_show_sock(nlh, arg);
4751 break;
4752 default:
4753 ret = -1;
4754 }
4755
4756 render();
4757
4758 return ret;
4759 }
4760
4761 static int handle_follow_request(struct filter *f)
4762 {
4763 int ret = 0;
4764 int groups = 0;
4765 struct rtnl_handle rth, rth2;
4766
4767 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << TCP_DB))
4768 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
4769 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << UDP_DB))
4770 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
4771 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << TCP_DB))
4772 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
4773 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << UDP_DB))
4774 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
4775
4776 if (groups == 0)
4777 return -1;
4778
4779 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
4780 return -1;
4781
4782 rth.dump = 0;
4783 rth.local.nl_pid = 0;
4784
4785 if (f->kill) {
4786 if (rtnl_open_byproto(&rth2, groups, NETLINK_SOCK_DIAG)) {
4787 rtnl_close(&rth);
4788 return -1;
4789 }
4790 f->rth_for_killing = &rth2;
4791 }
4792
4793 if (rtnl_dump_filter(&rth, generic_show_sock, f))
4794 ret = -1;
4795
4796 rtnl_close(&rth);
4797 if (f->rth_for_killing)
4798 rtnl_close(f->rth_for_killing);
4799 return ret;
4800 }
4801
4802 static int get_snmp_int(char *proto, char *key, int *result)
4803 {
4804 char buf[1024];
4805 FILE *fp;
4806 int protolen = strlen(proto);
4807 int keylen = strlen(key);
4808
4809 *result = 0;
4810
4811 if ((fp = net_snmp_open()) == NULL)
4812 return -1;
4813
4814 while (fgets(buf, sizeof(buf), fp) != NULL) {
4815 char *p = buf;
4816 int pos = 0;
4817
4818 if (memcmp(buf, proto, protolen))
4819 continue;
4820 while ((p = strchr(p, ' ')) != NULL) {
4821 pos++;
4822 p++;
4823 if (memcmp(p, key, keylen) == 0 &&
4824 (p[keylen] == ' ' || p[keylen] == '\n'))
4825 break;
4826 }
4827 if (fgets(buf, sizeof(buf), fp) == NULL)
4828 break;
4829 if (memcmp(buf, proto, protolen))
4830 break;
4831 p = buf;
4832 while ((p = strchr(p, ' ')) != NULL) {
4833 p++;
4834 if (--pos == 0) {
4835 sscanf(p, "%d", result);
4836 fclose(fp);
4837 return 0;
4838 }
4839 }
4840 }
4841
4842 fclose(fp);
4843 errno = ESRCH;
4844 return -1;
4845 }
4846
4847
4848 /* Get stats from sockstat */
4849
4850 struct ssummary {
4851 int socks;
4852 int tcp_mem;
4853 int tcp_total;
4854 int tcp_orphans;
4855 int tcp_tws;
4856 int tcp4_hashed;
4857 int udp4;
4858 int raw4;
4859 int frag4;
4860 int frag4_mem;
4861 int tcp6_hashed;
4862 int udp6;
4863 int raw6;
4864 int frag6;
4865 int frag6_mem;
4866 };
4867
4868 static void get_sockstat_line(char *line, struct ssummary *s)
4869 {
4870 char id[256], rem[256];
4871
4872 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
4873 return;
4874
4875 if (strcmp(id, "sockets:") == 0)
4876 sscanf(rem, "%*s%d", &s->socks);
4877 else if (strcmp(id, "UDP:") == 0)
4878 sscanf(rem, "%*s%d", &s->udp4);
4879 else if (strcmp(id, "UDP6:") == 0)
4880 sscanf(rem, "%*s%d", &s->udp6);
4881 else if (strcmp(id, "RAW:") == 0)
4882 sscanf(rem, "%*s%d", &s->raw4);
4883 else if (strcmp(id, "RAW6:") == 0)
4884 sscanf(rem, "%*s%d", &s->raw6);
4885 else if (strcmp(id, "TCP6:") == 0)
4886 sscanf(rem, "%*s%d", &s->tcp6_hashed);
4887 else if (strcmp(id, "FRAG:") == 0)
4888 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
4889 else if (strcmp(id, "FRAG6:") == 0)
4890 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
4891 else if (strcmp(id, "TCP:") == 0)
4892 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
4893 &s->tcp4_hashed,
4894 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
4895 }
4896
4897 static int get_sockstat(struct ssummary *s)
4898 {
4899 char buf[256];
4900 FILE *fp;
4901
4902 memset(s, 0, sizeof(*s));
4903
4904 if ((fp = net_sockstat_open()) == NULL)
4905 return -1;
4906 while (fgets(buf, sizeof(buf), fp) != NULL)
4907 get_sockstat_line(buf, s);
4908 fclose(fp);
4909
4910 if ((fp = net_sockstat6_open()) == NULL)
4911 return 0;
4912 while (fgets(buf, sizeof(buf), fp) != NULL)
4913 get_sockstat_line(buf, s);
4914 fclose(fp);
4915
4916 return 0;
4917 }
4918
4919 static int print_summary(void)
4920 {
4921 struct ssummary s;
4922 int tcp_estab;
4923
4924 if (get_sockstat(&s) < 0)
4925 perror("ss: get_sockstat");
4926 if (get_snmp_int("Tcp:", "CurrEstab", &tcp_estab) < 0)
4927 perror("ss: get_snmpstat");
4928
4929 printf("Total: %d\n", s.socks);
4930
4931 printf("TCP: %d (estab %d, closed %d, orphaned %d, timewait %d)\n",
4932 s.tcp_total + s.tcp_tws, tcp_estab,
4933 s.tcp_total - (s.tcp4_hashed + s.tcp6_hashed - s.tcp_tws),
4934 s.tcp_orphans, s.tcp_tws);
4935
4936 printf("\n");
4937 printf("Transport Total IP IPv6\n");
4938 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
4939 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
4940 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
4941 printf("INET %-9d %-9d %-9d\n",
4942 s.raw4+s.udp4+s.tcp4_hashed+
4943 s.raw6+s.udp6+s.tcp6_hashed,
4944 s.raw4+s.udp4+s.tcp4_hashed,
4945 s.raw6+s.udp6+s.tcp6_hashed);
4946 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
4947
4948 printf("\n");
4949
4950 return 0;
4951 }
4952
4953 static void _usage(FILE *dest)
4954 {
4955 fprintf(dest,
4956 "Usage: ss [ OPTIONS ]\n"
4957 " ss [ OPTIONS ] [ FILTER ]\n"
4958 " -h, --help this message\n"
4959 " -V, --version output version information\n"
4960 " -n, --numeric don't resolve service names\n"
4961 " -r, --resolve resolve host names\n"
4962 " -a, --all display all sockets\n"
4963 " -l, --listening display listening sockets\n"
4964 " -o, --options show timer information\n"
4965 " -e, --extended show detailed socket information\n"
4966 " -m, --memory show socket memory usage\n"
4967 " -p, --processes show process using socket\n"
4968 " -i, --info show internal TCP information\n"
4969 " --tipcinfo show internal tipc socket information\n"
4970 " -s, --summary show socket usage summary\n"
4971 " --tos show tos and priority information\n"
4972 " -b, --bpf show bpf filter socket information\n"
4973 " -E, --events continually display sockets as they are destroyed\n"
4974 " -Z, --context display process SELinux security contexts\n"
4975 " -z, --contexts display process and socket SELinux security contexts\n"
4976 " -N, --net switch to the specified network namespace name\n"
4977 "\n"
4978 " -4, --ipv4 display only IP version 4 sockets\n"
4979 " -6, --ipv6 display only IP version 6 sockets\n"
4980 " -0, --packet display PACKET sockets\n"
4981 " -t, --tcp display only TCP sockets\n"
4982 " -S, --sctp display only SCTP sockets\n"
4983 " -u, --udp display only UDP sockets\n"
4984 " -d, --dccp display only DCCP sockets\n"
4985 " -w, --raw display only RAW sockets\n"
4986 " -x, --unix display only Unix domain sockets\n"
4987 " --tipc display only TIPC sockets\n"
4988 " --vsock display only vsock sockets\n"
4989 " -f, --family=FAMILY display sockets of type FAMILY\n"
4990 " FAMILY := {inet|inet6|link|unix|netlink|vsock|tipc|xdp|help}\n"
4991 "\n"
4992 " -K, --kill forcibly close sockets, display what was closed\n"
4993 " -H, --no-header Suppress header line\n"
4994 " -O, --oneline socket's data printed on a single line\n"
4995 "\n"
4996 " -A, --query=QUERY, --socket=QUERY\n"
4997 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink|vsock_stream|vsock_dgram|tipc}[,QUERY]\n"
4998 "\n"
4999 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
5000 " -F, --filter=FILE read filter information from FILE\n"
5001 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
5002 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
5003 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listening|closing}\n"
5004 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
5005 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
5006 " bucket := {syn-recv|time-wait}\n"
5007 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listening|closing}\n"
5008 );
5009 }
5010
5011 static void help(void) __attribute__((noreturn));
5012 static void help(void)
5013 {
5014 _usage(stdout);
5015 exit(0);
5016 }
5017
5018 static void usage(void) __attribute__((noreturn));
5019 static void usage(void)
5020 {
5021 _usage(stderr);
5022 exit(-1);
5023 }
5024
5025
5026 static int scan_state(const char *state)
5027 {
5028 static const char * const sstate_namel[] = {
5029 "UNKNOWN",
5030 [SS_ESTABLISHED] = "established",
5031 [SS_SYN_SENT] = "syn-sent",
5032 [SS_SYN_RECV] = "syn-recv",
5033 [SS_FIN_WAIT1] = "fin-wait-1",
5034 [SS_FIN_WAIT2] = "fin-wait-2",
5035 [SS_TIME_WAIT] = "time-wait",
5036 [SS_CLOSE] = "unconnected",
5037 [SS_CLOSE_WAIT] = "close-wait",
5038 [SS_LAST_ACK] = "last-ack",
5039 [SS_LISTEN] = "listening",
5040 [SS_CLOSING] = "closing",
5041 };
5042 int i;
5043
5044 if (strcasecmp(state, "close") == 0 ||
5045 strcasecmp(state, "closed") == 0)
5046 return (1<<SS_CLOSE);
5047 if (strcasecmp(state, "syn-rcv") == 0)
5048 return (1<<SS_SYN_RECV);
5049 if (strcasecmp(state, "established") == 0)
5050 return (1<<SS_ESTABLISHED);
5051 if (strcasecmp(state, "all") == 0)
5052 return SS_ALL;
5053 if (strcasecmp(state, "connected") == 0)
5054 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
5055 if (strcasecmp(state, "synchronized") == 0)
5056 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
5057 if (strcasecmp(state, "bucket") == 0)
5058 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
5059 if (strcasecmp(state, "big") == 0)
5060 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
5061 for (i = 0; i < SS_MAX; i++) {
5062 if (strcasecmp(state, sstate_namel[i]) == 0)
5063 return (1<<i);
5064 }
5065
5066 fprintf(stderr, "ss: wrong state name: %s\n", state);
5067 exit(-1);
5068 }
5069
5070 /* Values 'v' and 'V' are already used so a non-character is used */
5071 #define OPT_VSOCK 256
5072
5073 /* Values of 't' are already used so a non-character is used */
5074 #define OPT_TIPCSOCK 257
5075 #define OPT_TIPCINFO 258
5076
5077 #define OPT_TOS 259
5078
5079 /* Values of 'x' are already used so a non-character is used */
5080 #define OPT_XDPSOCK 260
5081
5082 static const struct option long_opts[] = {
5083 { "numeric", 0, 0, 'n' },
5084 { "resolve", 0, 0, 'r' },
5085 { "options", 0, 0, 'o' },
5086 { "extended", 0, 0, 'e' },
5087 { "memory", 0, 0, 'm' },
5088 { "info", 0, 0, 'i' },
5089 { "processes", 0, 0, 'p' },
5090 { "bpf", 0, 0, 'b' },
5091 { "events", 0, 0, 'E' },
5092 { "dccp", 0, 0, 'd' },
5093 { "tcp", 0, 0, 't' },
5094 { "sctp", 0, 0, 'S' },
5095 { "udp", 0, 0, 'u' },
5096 { "raw", 0, 0, 'w' },
5097 { "unix", 0, 0, 'x' },
5098 { "tipc", 0, 0, OPT_TIPCSOCK},
5099 { "vsock", 0, 0, OPT_VSOCK },
5100 { "all", 0, 0, 'a' },
5101 { "listening", 0, 0, 'l' },
5102 { "ipv4", 0, 0, '4' },
5103 { "ipv6", 0, 0, '6' },
5104 { "packet", 0, 0, '0' },
5105 { "family", 1, 0, 'f' },
5106 { "socket", 1, 0, 'A' },
5107 { "query", 1, 0, 'A' },
5108 { "summary", 0, 0, 's' },
5109 { "diag", 1, 0, 'D' },
5110 { "filter", 1, 0, 'F' },
5111 { "version", 0, 0, 'V' },
5112 { "help", 0, 0, 'h' },
5113 { "context", 0, 0, 'Z' },
5114 { "contexts", 0, 0, 'z' },
5115 { "net", 1, 0, 'N' },
5116 { "tipcinfo", 0, 0, OPT_TIPCINFO},
5117 { "tos", 0, 0, OPT_TOS },
5118 { "kill", 0, 0, 'K' },
5119 { "no-header", 0, 0, 'H' },
5120 { "xdp", 0, 0, OPT_XDPSOCK},
5121 { "oneline", 0, 0, 'O' },
5122 { 0 }
5123
5124 };
5125
5126 int main(int argc, char *argv[])
5127 {
5128 int saw_states = 0;
5129 int saw_query = 0;
5130 int do_summary = 0;
5131 const char *dump_tcpdiag = NULL;
5132 FILE *filter_fp = NULL;
5133 int ch;
5134 int state_filter = 0;
5135
5136 while ((ch = getopt_long(argc, argv,
5137 "dhaletuwxnro460spbEf:miA:D:F:vVzZN:KHSO",
5138 long_opts, NULL)) != EOF) {
5139 switch (ch) {
5140 case 'n':
5141 numeric = 1;
5142 break;
5143 case 'r':
5144 resolve_hosts = 1;
5145 break;
5146 case 'o':
5147 show_options = 1;
5148 break;
5149 case 'e':
5150 show_options = 1;
5151 show_details++;
5152 break;
5153 case 'm':
5154 show_mem = 1;
5155 break;
5156 case 'i':
5157 show_tcpinfo = 1;
5158 break;
5159 case 'p':
5160 show_users++;
5161 user_ent_hash_build();
5162 break;
5163 case 'b':
5164 show_options = 1;
5165 show_bpf++;
5166 break;
5167 case 'E':
5168 follow_events = 1;
5169 break;
5170 case 'd':
5171 filter_db_set(&current_filter, DCCP_DB, true);
5172 break;
5173 case 't':
5174 filter_db_set(&current_filter, TCP_DB, true);
5175 break;
5176 case 'S':
5177 filter_db_set(&current_filter, SCTP_DB, true);
5178 break;
5179 case 'u':
5180 filter_db_set(&current_filter, UDP_DB, true);
5181 break;
5182 case 'w':
5183 filter_db_set(&current_filter, RAW_DB, true);
5184 break;
5185 case 'x':
5186 filter_af_set(&current_filter, AF_UNIX);
5187 break;
5188 case OPT_VSOCK:
5189 filter_af_set(&current_filter, AF_VSOCK);
5190 break;
5191 case OPT_TIPCSOCK:
5192 filter_af_set(&current_filter, AF_TIPC);
5193 break;
5194 case 'a':
5195 state_filter = SS_ALL;
5196 break;
5197 case 'l':
5198 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
5199 break;
5200 case '4':
5201 filter_af_set(&current_filter, AF_INET);
5202 break;
5203 case '6':
5204 filter_af_set(&current_filter, AF_INET6);
5205 break;
5206 case '0':
5207 filter_af_set(&current_filter, AF_PACKET);
5208 break;
5209 case OPT_XDPSOCK:
5210 filter_af_set(&current_filter, AF_XDP);
5211 break;
5212 case 'f':
5213 if (strcmp(optarg, "inet") == 0)
5214 filter_af_set(&current_filter, AF_INET);
5215 else if (strcmp(optarg, "inet6") == 0)
5216 filter_af_set(&current_filter, AF_INET6);
5217 else if (strcmp(optarg, "link") == 0)
5218 filter_af_set(&current_filter, AF_PACKET);
5219 else if (strcmp(optarg, "unix") == 0)
5220 filter_af_set(&current_filter, AF_UNIX);
5221 else if (strcmp(optarg, "netlink") == 0)
5222 filter_af_set(&current_filter, AF_NETLINK);
5223 else if (strcmp(optarg, "tipc") == 0)
5224 filter_af_set(&current_filter, AF_TIPC);
5225 else if (strcmp(optarg, "vsock") == 0)
5226 filter_af_set(&current_filter, AF_VSOCK);
5227 else if (strcmp(optarg, "xdp") == 0)
5228 filter_af_set(&current_filter, AF_XDP);
5229 else if (strcmp(optarg, "help") == 0)
5230 help();
5231 else {
5232 fprintf(stderr, "ss: \"%s\" is invalid family\n",
5233 optarg);
5234 usage();
5235 }
5236 break;
5237 case 'A':
5238 {
5239 char *p, *p1;
5240
5241 if (!saw_query) {
5242 current_filter.dbs = 0;
5243 state_filter = state_filter ?
5244 state_filter : SS_CONN;
5245 saw_query = 1;
5246 do_default = 0;
5247 }
5248 p = p1 = optarg;
5249 do {
5250 if ((p1 = strchr(p, ',')) != NULL)
5251 *p1 = 0;
5252 if (filter_db_parse(&current_filter, p)) {
5253 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
5254 usage();
5255 }
5256 p = p1 + 1;
5257 } while (p1);
5258 break;
5259 }
5260 case 's':
5261 do_summary = 1;
5262 break;
5263 case 'D':
5264 dump_tcpdiag = optarg;
5265 break;
5266 case 'F':
5267 if (filter_fp) {
5268 fprintf(stderr, "More than one filter file\n");
5269 exit(-1);
5270 }
5271 if (optarg[0] == '-')
5272 filter_fp = stdin;
5273 else
5274 filter_fp = fopen(optarg, "r");
5275 if (!filter_fp) {
5276 perror("fopen filter file");
5277 exit(-1);
5278 }
5279 break;
5280 case 'v':
5281 case 'V':
5282 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
5283 exit(0);
5284 case 'z':
5285 show_sock_ctx++;
5286 /* fall through */
5287 case 'Z':
5288 if (is_selinux_enabled() <= 0) {
5289 fprintf(stderr, "ss: SELinux is not enabled.\n");
5290 exit(1);
5291 }
5292 show_proc_ctx++;
5293 user_ent_hash_build();
5294 break;
5295 case 'N':
5296 if (netns_switch(optarg))
5297 exit(1);
5298 break;
5299 case OPT_TIPCINFO:
5300 show_tipcinfo = 1;
5301 break;
5302 case OPT_TOS:
5303 show_tos = 1;
5304 break;
5305 case 'K':
5306 current_filter.kill = 1;
5307 break;
5308 case 'H':
5309 show_header = 0;
5310 break;
5311 case 'O':
5312 oneline = 1;
5313 break;
5314 case 'h':
5315 help();
5316 case '?':
5317 default:
5318 usage();
5319 }
5320 }
5321
5322 argc -= optind;
5323 argv += optind;
5324
5325 if (do_summary) {
5326 print_summary();
5327 if (do_default && argc == 0)
5328 exit(0);
5329 }
5330
5331 while (argc > 0) {
5332 if (strcmp(*argv, "state") == 0) {
5333 NEXT_ARG();
5334 if (!saw_states)
5335 state_filter = 0;
5336 state_filter |= scan_state(*argv);
5337 saw_states = 1;
5338 } else if (strcmp(*argv, "exclude") == 0 ||
5339 strcmp(*argv, "excl") == 0) {
5340 NEXT_ARG();
5341 if (!saw_states)
5342 state_filter = SS_ALL;
5343 state_filter &= ~scan_state(*argv);
5344 saw_states = 1;
5345 } else {
5346 break;
5347 }
5348 argc--; argv++;
5349 }
5350
5351 if (do_default) {
5352 state_filter = state_filter ? state_filter : SS_CONN;
5353 filter_db_parse(&current_filter, "all");
5354 }
5355
5356 filter_states_set(&current_filter, state_filter);
5357 filter_merge_defaults(&current_filter);
5358
5359 if (!numeric && resolve_hosts &&
5360 (current_filter.dbs & (UNIX_DBM|INET_L4_DBM)))
5361 init_service_resolver();
5362
5363 if (current_filter.dbs == 0) {
5364 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
5365 exit(0);
5366 }
5367 if (current_filter.families == 0) {
5368 fprintf(stderr, "ss: no families to show with such filter.\n");
5369 exit(0);
5370 }
5371 if (current_filter.states == 0) {
5372 fprintf(stderr, "ss: no socket states to show with such filter.\n");
5373 exit(0);
5374 }
5375
5376 if (dump_tcpdiag) {
5377 FILE *dump_fp = stdout;
5378
5379 if (!(current_filter.dbs & (1<<TCP_DB))) {
5380 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
5381 exit(0);
5382 }
5383 if (dump_tcpdiag[0] != '-') {
5384 dump_fp = fopen(dump_tcpdiag, "w");
5385 if (!dump_tcpdiag) {
5386 perror("fopen dump file");
5387 exit(-1);
5388 }
5389 }
5390 inet_show_netlink(&current_filter, dump_fp, IPPROTO_TCP);
5391 fflush(dump_fp);
5392 exit(0);
5393 }
5394
5395 if (ssfilter_parse(&current_filter.f, argc, argv, filter_fp))
5396 usage();
5397
5398 if (!(current_filter.dbs & (current_filter.dbs - 1)))
5399 columns[COL_NETID].disabled = 1;
5400
5401 if (!(current_filter.states & (current_filter.states - 1)))
5402 columns[COL_STATE].disabled = 1;
5403
5404 if (show_header)
5405 print_header();
5406
5407 fflush(stdout);
5408
5409 if (follow_events)
5410 exit(handle_follow_request(&current_filter));
5411
5412 if (current_filter.dbs & (1<<NETLINK_DB))
5413 netlink_show(&current_filter);
5414 if (current_filter.dbs & PACKET_DBM)
5415 packet_show(&current_filter);
5416 if (current_filter.dbs & UNIX_DBM)
5417 unix_show(&current_filter);
5418 if (current_filter.dbs & (1<<RAW_DB))
5419 raw_show(&current_filter);
5420 if (current_filter.dbs & (1<<UDP_DB))
5421 udp_show(&current_filter);
5422 if (current_filter.dbs & (1<<TCP_DB))
5423 tcp_show(&current_filter);
5424 if (current_filter.dbs & (1<<DCCP_DB))
5425 dccp_show(&current_filter);
5426 if (current_filter.dbs & (1<<SCTP_DB))
5427 sctp_show(&current_filter);
5428 if (current_filter.dbs & VSOCK_DBM)
5429 vsock_show(&current_filter);
5430 if (current_filter.dbs & (1<<TIPC_DB))
5431 tipc_show(&current_filter);
5432 if (current_filter.dbs & (1<<XDP_DB))
5433 xdp_show(&current_filter);
5434
5435 if (show_users || show_proc_ctx || show_sock_ctx)
5436 user_ent_destroy();
5437
5438 render();
5439
5440 return 0;
5441 }
Upstream mirror of iproute2