]> git.proxmox.com Git - mirror_iproute2.git/blob - misc/ss.c
ab206b2011ec92b899709d2c78ce7310e88ec80e
[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 #include "rt_names.h"
39
40 #include <linux/tcp.h>
41 #include <linux/sock_diag.h>
42 #include <linux/inet_diag.h>
43 #include <linux/unix_diag.h>
44 #include <linux/netdevice.h> /* for MAX_ADDR_LEN */
45 #include <linux/filter.h>
46 #include <linux/xdp_diag.h>
47 #include <linux/packet_diag.h>
48 #include <linux/netlink_diag.h>
49 #include <linux/sctp.h>
50 #include <linux/vm_sockets_diag.h>
51 #include <linux/net.h>
52 #include <linux/tipc.h>
53 #include <linux/tipc_netlink.h>
54 #include <linux/tipc_sockets_diag.h>
55 #include <linux/tls.h>
56
57 /* AF_VSOCK/PF_VSOCK is only provided since glibc 2.18 */
58 #ifndef PF_VSOCK
59 #define PF_VSOCK 40
60 #endif
61 #ifndef AF_VSOCK
62 #define AF_VSOCK PF_VSOCK
63 #endif
64
65 #define MAGIC_SEQ 123456
66 #define BUF_CHUNK (1024 * 1024) /* Buffer chunk allocation size */
67 #define BUF_CHUNKS_MAX 5 /* Maximum number of allocated buffer chunks */
68 #define LEN_ALIGN(x) (((x) + 1) & ~1)
69
70 #define DIAG_REQUEST(_req, _r) \
71 struct { \
72 struct nlmsghdr nlh; \
73 _r; \
74 } _req = { \
75 .nlh = { \
76 .nlmsg_type = SOCK_DIAG_BY_FAMILY, \
77 .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,\
78 .nlmsg_seq = MAGIC_SEQ, \
79 .nlmsg_len = sizeof(_req), \
80 }, \
81 }
82
83 #if HAVE_SELINUX
84 #include <selinux/selinux.h>
85 #else
86 /* Stubs for SELinux functions */
87 static int is_selinux_enabled(void)
88 {
89 return -1;
90 }
91
92 static int getpidcon(pid_t pid, char **context)
93 {
94 *context = NULL;
95 return -1;
96 }
97
98 static int getfilecon(char *path, char **context)
99 {
100 *context = NULL;
101 return -1;
102 }
103
104 static int security_get_initial_context(char *name, char **context)
105 {
106 *context = NULL;
107 return -1;
108 }
109 #endif
110
111 int preferred_family = AF_UNSPEC;
112 static int show_options;
113 int show_details;
114 static int show_users;
115 static int show_mem;
116 static int show_tcpinfo;
117 static int show_bpf;
118 static int show_proc_ctx;
119 static int show_sock_ctx;
120 static int show_header = 1;
121 static int follow_events;
122 static int sctp_ino;
123 static int show_tipcinfo;
124 static int show_tos;
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_PROC,
138 COL_MAX
139 };
140
141 enum col_align {
142 ALIGN_LEFT,
143 ALIGN_CENTER,
144 ALIGN_RIGHT
145 };
146
147 struct column {
148 const enum col_align align;
149 const char *header;
150 const char *ldelim;
151 int disabled;
152 int width; /* Calculated, including additional layout spacing */
153 int max_len; /* Measured maximum field length in this column */
154 };
155
156 static struct column columns[] = {
157 { ALIGN_LEFT, "Netid", "", 0, 0, 0 },
158 { ALIGN_LEFT, "State", " ", 0, 0, 0 },
159 { ALIGN_LEFT, "Recv-Q", " ", 0, 0, 0 },
160 { ALIGN_LEFT, "Send-Q", " ", 0, 0, 0 },
161 { ALIGN_RIGHT, "Local Address:", " ", 0, 0, 0 },
162 { ALIGN_LEFT, "Port", "", 0, 0, 0 },
163 { ALIGN_RIGHT, "Peer Address:", " ", 0, 0, 0 },
164 { ALIGN_LEFT, "Port", "", 0, 0, 0 },
165 { ALIGN_LEFT, "Process", "", 0, 0, 0 },
166 { ALIGN_LEFT, "", "", 0, 0, 0 },
167 };
168
169 static struct column *current_field = columns;
170
171 /* Output buffer: chained chunks of BUF_CHUNK bytes. Each field is written to
172 * the buffer as a variable size token. A token consists of a 16 bits length
173 * field, followed by a string which is not NULL-terminated.
174 *
175 * A new chunk is allocated and linked when the current chunk doesn't have
176 * enough room to store the current token as a whole.
177 */
178 struct buf_chunk {
179 struct buf_chunk *next; /* Next chained chunk */
180 char *end; /* Current end of content */
181 char data[0];
182 };
183
184 struct buf_token {
185 uint16_t len; /* Data length, excluding length descriptor */
186 char data[0];
187 };
188
189 static struct {
190 struct buf_token *cur; /* Position of current token in chunk */
191 struct buf_chunk *head; /* First chunk */
192 struct buf_chunk *tail; /* Current chunk */
193 int chunks; /* Number of allocated chunks */
194 } buffer;
195
196 static const char *TCP_PROTO = "tcp";
197 static const char *SCTP_PROTO = "sctp";
198 static const char *UDP_PROTO = "udp";
199 static const char *RAW_PROTO = "raw";
200 static const char *dg_proto;
201
202 enum {
203 TCP_DB,
204 DCCP_DB,
205 UDP_DB,
206 RAW_DB,
207 UNIX_DG_DB,
208 UNIX_ST_DB,
209 UNIX_SQ_DB,
210 PACKET_DG_DB,
211 PACKET_R_DB,
212 NETLINK_DB,
213 SCTP_DB,
214 VSOCK_ST_DB,
215 VSOCK_DG_DB,
216 TIPC_DB,
217 XDP_DB,
218 MAX_DB
219 };
220
221 #define PACKET_DBM ((1<<PACKET_DG_DB)|(1<<PACKET_R_DB))
222 #define UNIX_DBM ((1<<UNIX_DG_DB)|(1<<UNIX_ST_DB)|(1<<UNIX_SQ_DB))
223 #define ALL_DB ((1<<MAX_DB)-1)
224 #define INET_L4_DBM ((1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB)|(1<<SCTP_DB))
225 #define INET_DBM (INET_L4_DBM | (1<<RAW_DB))
226 #define VSOCK_DBM ((1<<VSOCK_ST_DB)|(1<<VSOCK_DG_DB))
227
228 enum {
229 SS_UNKNOWN,
230 SS_ESTABLISHED,
231 SS_SYN_SENT,
232 SS_SYN_RECV,
233 SS_FIN_WAIT1,
234 SS_FIN_WAIT2,
235 SS_TIME_WAIT,
236 SS_CLOSE,
237 SS_CLOSE_WAIT,
238 SS_LAST_ACK,
239 SS_LISTEN,
240 SS_CLOSING,
241 SS_MAX
242 };
243
244 enum {
245 SCTP_STATE_CLOSED = 0,
246 SCTP_STATE_COOKIE_WAIT = 1,
247 SCTP_STATE_COOKIE_ECHOED = 2,
248 SCTP_STATE_ESTABLISHED = 3,
249 SCTP_STATE_SHUTDOWN_PENDING = 4,
250 SCTP_STATE_SHUTDOWN_SENT = 5,
251 SCTP_STATE_SHUTDOWN_RECEIVED = 6,
252 SCTP_STATE_SHUTDOWN_ACK_SENT = 7,
253 };
254
255 #define SS_ALL ((1 << SS_MAX) - 1)
256 #define SS_CONN (SS_ALL & ~((1<<SS_LISTEN)|(1<<SS_CLOSE)|(1<<SS_TIME_WAIT)|(1<<SS_SYN_RECV)))
257 #define TIPC_SS_CONN ((1<<SS_ESTABLISHED)|(1<<SS_LISTEN)|(1<<SS_CLOSE))
258
259 #include "ssfilter.h"
260
261 struct filter {
262 int dbs;
263 int states;
264 uint64_t families;
265 struct ssfilter *f;
266 bool kill;
267 struct rtnl_handle *rth_for_killing;
268 };
269
270 #define FAMILY_MASK(family) ((uint64_t)1 << (family))
271
272 static const struct filter default_dbs[MAX_DB] = {
273 [TCP_DB] = {
274 .states = SS_CONN,
275 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
276 },
277 [DCCP_DB] = {
278 .states = SS_CONN,
279 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
280 },
281 [UDP_DB] = {
282 .states = (1 << SS_ESTABLISHED),
283 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
284 },
285 [RAW_DB] = {
286 .states = (1 << SS_ESTABLISHED),
287 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
288 },
289 [UNIX_DG_DB] = {
290 .states = (1 << SS_CLOSE),
291 .families = FAMILY_MASK(AF_UNIX),
292 },
293 [UNIX_ST_DB] = {
294 .states = SS_CONN,
295 .families = FAMILY_MASK(AF_UNIX),
296 },
297 [UNIX_SQ_DB] = {
298 .states = SS_CONN,
299 .families = FAMILY_MASK(AF_UNIX),
300 },
301 [PACKET_DG_DB] = {
302 .states = (1 << SS_CLOSE),
303 .families = FAMILY_MASK(AF_PACKET),
304 },
305 [PACKET_R_DB] = {
306 .states = (1 << SS_CLOSE),
307 .families = FAMILY_MASK(AF_PACKET),
308 },
309 [NETLINK_DB] = {
310 .states = (1 << SS_CLOSE),
311 .families = FAMILY_MASK(AF_NETLINK),
312 },
313 [SCTP_DB] = {
314 .states = SS_CONN,
315 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
316 },
317 [VSOCK_ST_DB] = {
318 .states = SS_CONN,
319 .families = FAMILY_MASK(AF_VSOCK),
320 },
321 [VSOCK_DG_DB] = {
322 .states = SS_CONN,
323 .families = FAMILY_MASK(AF_VSOCK),
324 },
325 [TIPC_DB] = {
326 .states = TIPC_SS_CONN,
327 .families = FAMILY_MASK(AF_TIPC),
328 },
329 [XDP_DB] = {
330 .states = (1 << SS_CLOSE),
331 .families = FAMILY_MASK(AF_XDP),
332 },
333 };
334
335 static const struct filter default_afs[AF_MAX] = {
336 [AF_INET] = {
337 .dbs = INET_DBM,
338 .states = SS_CONN,
339 },
340 [AF_INET6] = {
341 .dbs = INET_DBM,
342 .states = SS_CONN,
343 },
344 [AF_UNIX] = {
345 .dbs = UNIX_DBM,
346 .states = SS_CONN,
347 },
348 [AF_PACKET] = {
349 .dbs = PACKET_DBM,
350 .states = (1 << SS_CLOSE),
351 },
352 [AF_NETLINK] = {
353 .dbs = (1 << NETLINK_DB),
354 .states = (1 << SS_CLOSE),
355 },
356 [AF_VSOCK] = {
357 .dbs = VSOCK_DBM,
358 .states = SS_CONN,
359 },
360 [AF_TIPC] = {
361 .dbs = (1 << TIPC_DB),
362 .states = TIPC_SS_CONN,
363 },
364 [AF_XDP] = {
365 .dbs = (1 << XDP_DB),
366 .states = (1 << SS_CLOSE),
367 },
368 };
369
370 static int do_default = 1;
371 static struct filter current_filter;
372
373 static void filter_db_set(struct filter *f, int db, bool enable)
374 {
375 if (enable) {
376 f->states |= default_dbs[db].states;
377 f->dbs |= 1 << db;
378 } else {
379 f->dbs &= ~(1 << db);
380 }
381 do_default = 0;
382 }
383
384 static int filter_db_parse(struct filter *f, const char *s)
385 {
386 const struct {
387 const char *name;
388 int dbs[MAX_DB + 1];
389 } db_name_tbl[] = {
390 #define ENTRY(name, ...) { #name, { __VA_ARGS__, MAX_DB } }
391 ENTRY(all, UDP_DB, DCCP_DB, TCP_DB, RAW_DB,
392 UNIX_ST_DB, UNIX_DG_DB, UNIX_SQ_DB,
393 PACKET_R_DB, PACKET_DG_DB, NETLINK_DB,
394 SCTP_DB, VSOCK_ST_DB, VSOCK_DG_DB, XDP_DB),
395 ENTRY(inet, UDP_DB, DCCP_DB, TCP_DB, SCTP_DB, RAW_DB),
396 ENTRY(udp, UDP_DB),
397 ENTRY(dccp, DCCP_DB),
398 ENTRY(tcp, TCP_DB),
399 ENTRY(sctp, SCTP_DB),
400 ENTRY(raw, RAW_DB),
401 ENTRY(unix, UNIX_ST_DB, UNIX_DG_DB, UNIX_SQ_DB),
402 ENTRY(unix_stream, UNIX_ST_DB),
403 ENTRY(u_str, UNIX_ST_DB), /* alias for unix_stream */
404 ENTRY(unix_dgram, UNIX_DG_DB),
405 ENTRY(u_dgr, UNIX_DG_DB), /* alias for unix_dgram */
406 ENTRY(unix_seqpacket, UNIX_SQ_DB),
407 ENTRY(u_seq, UNIX_SQ_DB), /* alias for unix_seqpacket */
408 ENTRY(packet, PACKET_R_DB, PACKET_DG_DB),
409 ENTRY(packet_raw, PACKET_R_DB),
410 ENTRY(p_raw, PACKET_R_DB), /* alias for packet_raw */
411 ENTRY(packet_dgram, PACKET_DG_DB),
412 ENTRY(p_dgr, PACKET_DG_DB), /* alias for packet_dgram */
413 ENTRY(netlink, NETLINK_DB),
414 ENTRY(vsock, VSOCK_ST_DB, VSOCK_DG_DB),
415 ENTRY(vsock_stream, VSOCK_ST_DB),
416 ENTRY(v_str, VSOCK_ST_DB), /* alias for vsock_stream */
417 ENTRY(vsock_dgram, VSOCK_DG_DB),
418 ENTRY(v_dgr, VSOCK_DG_DB), /* alias for vsock_dgram */
419 ENTRY(xdp, XDP_DB),
420 #undef ENTRY
421 };
422 bool enable = true;
423 unsigned int i;
424 const int *dbp;
425
426 if (s[0] == '!') {
427 enable = false;
428 s++;
429 }
430 for (i = 0; i < ARRAY_SIZE(db_name_tbl); i++) {
431 if (strcmp(s, db_name_tbl[i].name))
432 continue;
433 for (dbp = db_name_tbl[i].dbs; *dbp != MAX_DB; dbp++)
434 filter_db_set(f, *dbp, enable);
435 return 0;
436 }
437 return -1;
438 }
439
440 static void filter_af_set(struct filter *f, int af)
441 {
442 f->states |= default_afs[af].states;
443 f->families |= FAMILY_MASK(af);
444 do_default = 0;
445 preferred_family = af;
446 }
447
448 static int filter_af_get(struct filter *f, int af)
449 {
450 return !!(f->families & FAMILY_MASK(af));
451 }
452
453 static void filter_states_set(struct filter *f, int states)
454 {
455 if (states)
456 f->states = states;
457 }
458
459 static void filter_merge_defaults(struct filter *f)
460 {
461 int db;
462 int af;
463
464 for (db = 0; db < MAX_DB; db++) {
465 if (!(f->dbs & (1 << db)))
466 continue;
467
468 if (!(default_dbs[db].families & f->families))
469 f->families |= default_dbs[db].families;
470 }
471 for (af = 0; af < AF_MAX; af++) {
472 if (!(f->families & FAMILY_MASK(af)))
473 continue;
474
475 if (!(default_afs[af].dbs & f->dbs))
476 f->dbs |= default_afs[af].dbs;
477 }
478 }
479
480 static FILE *generic_proc_open(const char *env, const char *name)
481 {
482 const char *p = getenv(env);
483 char store[128];
484
485 if (!p) {
486 p = getenv("PROC_ROOT") ? : "/proc";
487 snprintf(store, sizeof(store)-1, "%s/%s", p, name);
488 p = store;
489 }
490
491 return fopen(p, "r");
492 }
493 #define net_tcp_open() generic_proc_open("PROC_NET_TCP", "net/tcp")
494 #define net_tcp6_open() generic_proc_open("PROC_NET_TCP6", "net/tcp6")
495 #define net_udp_open() generic_proc_open("PROC_NET_UDP", "net/udp")
496 #define net_udp6_open() generic_proc_open("PROC_NET_UDP6", "net/udp6")
497 #define net_raw_open() generic_proc_open("PROC_NET_RAW", "net/raw")
498 #define net_raw6_open() generic_proc_open("PROC_NET_RAW6", "net/raw6")
499 #define net_unix_open() generic_proc_open("PROC_NET_UNIX", "net/unix")
500 #define net_packet_open() generic_proc_open("PROC_NET_PACKET", \
501 "net/packet")
502 #define net_netlink_open() generic_proc_open("PROC_NET_NETLINK", \
503 "net/netlink")
504 #define net_sockstat_open() generic_proc_open("PROC_NET_SOCKSTAT", \
505 "net/sockstat")
506 #define net_sockstat6_open() generic_proc_open("PROC_NET_SOCKSTAT6", \
507 "net/sockstat6")
508 #define net_snmp_open() generic_proc_open("PROC_NET_SNMP", "net/snmp")
509 #define ephemeral_ports_open() generic_proc_open("PROC_IP_LOCAL_PORT_RANGE", \
510 "sys/net/ipv4/ip_local_port_range")
511
512 struct user_ent {
513 struct user_ent *next;
514 unsigned int ino;
515 int pid;
516 int fd;
517 char *process;
518 char *process_ctx;
519 char *socket_ctx;
520 };
521
522 #define USER_ENT_HASH_SIZE 256
523 static struct user_ent *user_ent_hash[USER_ENT_HASH_SIZE];
524
525 static int user_ent_hashfn(unsigned int ino)
526 {
527 int val = (ino >> 24) ^ (ino >> 16) ^ (ino >> 8) ^ ino;
528
529 return val & (USER_ENT_HASH_SIZE - 1);
530 }
531
532 static void user_ent_add(unsigned int ino, char *process,
533 int pid, int fd,
534 char *proc_ctx,
535 char *sock_ctx)
536 {
537 struct user_ent *p, **pp;
538
539 p = malloc(sizeof(struct user_ent));
540 if (!p) {
541 fprintf(stderr, "ss: failed to malloc buffer\n");
542 abort();
543 }
544 p->next = NULL;
545 p->ino = ino;
546 p->pid = pid;
547 p->fd = fd;
548 p->process = strdup(process);
549 p->process_ctx = strdup(proc_ctx);
550 p->socket_ctx = strdup(sock_ctx);
551
552 pp = &user_ent_hash[user_ent_hashfn(ino)];
553 p->next = *pp;
554 *pp = p;
555 }
556
557 static void user_ent_destroy(void)
558 {
559 struct user_ent *p, *p_next;
560 int cnt = 0;
561
562 while (cnt != USER_ENT_HASH_SIZE) {
563 p = user_ent_hash[cnt];
564 while (p) {
565 free(p->process);
566 free(p->process_ctx);
567 free(p->socket_ctx);
568 p_next = p->next;
569 free(p);
570 p = p_next;
571 }
572 cnt++;
573 }
574 }
575
576 static void user_ent_hash_build(void)
577 {
578 const char *root = getenv("PROC_ROOT") ? : "/proc/";
579 struct dirent *d;
580 char name[1024];
581 int nameoff;
582 DIR *dir;
583 char *pid_context;
584 char *sock_context;
585 const char *no_ctx = "unavailable";
586 static int user_ent_hash_build_init;
587
588 /* If show_users & show_proc_ctx set only do this once */
589 if (user_ent_hash_build_init != 0)
590 return;
591
592 user_ent_hash_build_init = 1;
593
594 strlcpy(name, root, sizeof(name));
595
596 if (strlen(name) == 0 || name[strlen(name)-1] != '/')
597 strcat(name, "/");
598
599 nameoff = strlen(name);
600
601 dir = opendir(name);
602 if (!dir)
603 return;
604
605 while ((d = readdir(dir)) != NULL) {
606 struct dirent *d1;
607 char process[16];
608 char *p;
609 int pid, pos;
610 DIR *dir1;
611 char crap;
612
613 if (sscanf(d->d_name, "%d%c", &pid, &crap) != 1)
614 continue;
615
616 if (getpidcon(pid, &pid_context) != 0)
617 pid_context = strdup(no_ctx);
618
619 snprintf(name + nameoff, sizeof(name) - nameoff, "%d/fd/", pid);
620 pos = strlen(name);
621 if ((dir1 = opendir(name)) == NULL) {
622 free(pid_context);
623 continue;
624 }
625
626 process[0] = '\0';
627 p = process;
628
629 while ((d1 = readdir(dir1)) != NULL) {
630 const char *pattern = "socket:[";
631 unsigned int ino;
632 char lnk[64];
633 int fd;
634 ssize_t link_len;
635 char tmp[1024];
636
637 if (sscanf(d1->d_name, "%d%c", &fd, &crap) != 1)
638 continue;
639
640 snprintf(name+pos, sizeof(name) - pos, "%d", fd);
641
642 link_len = readlink(name, lnk, sizeof(lnk)-1);
643 if (link_len == -1)
644 continue;
645 lnk[link_len] = '\0';
646
647 if (strncmp(lnk, pattern, strlen(pattern)))
648 continue;
649
650 sscanf(lnk, "socket:[%u]", &ino);
651
652 snprintf(tmp, sizeof(tmp), "%s/%d/fd/%s",
653 root, pid, d1->d_name);
654
655 if (getfilecon(tmp, &sock_context) <= 0)
656 sock_context = strdup(no_ctx);
657
658 if (*p == '\0') {
659 FILE *fp;
660
661 snprintf(tmp, sizeof(tmp), "%s/%d/stat",
662 root, pid);
663 if ((fp = fopen(tmp, "r")) != NULL) {
664 if (fscanf(fp, "%*d (%[^)])", p) < 1)
665 ; /* ignore */
666 fclose(fp);
667 }
668 }
669 user_ent_add(ino, p, pid, fd,
670 pid_context, sock_context);
671 free(sock_context);
672 }
673 free(pid_context);
674 closedir(dir1);
675 }
676 closedir(dir);
677 }
678
679 enum entry_types {
680 USERS,
681 PROC_CTX,
682 PROC_SOCK_CTX
683 };
684
685 #define ENTRY_BUF_SIZE 512
686 static int find_entry(unsigned int ino, char **buf, int type)
687 {
688 struct user_ent *p;
689 int cnt = 0;
690 char *ptr;
691 char *new_buf;
692 int len, new_buf_len;
693 int buf_used = 0;
694 int buf_len = 0;
695
696 if (!ino)
697 return 0;
698
699 p = user_ent_hash[user_ent_hashfn(ino)];
700 ptr = *buf = NULL;
701 while (p) {
702 if (p->ino != ino)
703 goto next;
704
705 while (1) {
706 ptr = *buf + buf_used;
707 switch (type) {
708 case USERS:
709 len = snprintf(ptr, buf_len - buf_used,
710 "(\"%s\",pid=%d,fd=%d),",
711 p->process, p->pid, p->fd);
712 break;
713 case PROC_CTX:
714 len = snprintf(ptr, buf_len - buf_used,
715 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d),",
716 p->process, p->pid,
717 p->process_ctx, p->fd);
718 break;
719 case PROC_SOCK_CTX:
720 len = snprintf(ptr, buf_len - buf_used,
721 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d,sock_ctx=%s),",
722 p->process, p->pid,
723 p->process_ctx, p->fd,
724 p->socket_ctx);
725 break;
726 default:
727 fprintf(stderr, "ss: invalid type: %d\n", type);
728 abort();
729 }
730
731 if (len < 0 || len >= buf_len - buf_used) {
732 new_buf_len = buf_len + ENTRY_BUF_SIZE;
733 new_buf = realloc(*buf, new_buf_len);
734 if (!new_buf) {
735 fprintf(stderr, "ss: failed to malloc buffer\n");
736 abort();
737 }
738 *buf = new_buf;
739 buf_len = new_buf_len;
740 continue;
741 } else {
742 buf_used += len;
743 break;
744 }
745 }
746 cnt++;
747 next:
748 p = p->next;
749 }
750 if (buf_used) {
751 ptr = *buf + buf_used;
752 ptr[-1] = '\0';
753 }
754 return cnt;
755 }
756
757 static unsigned long long cookie_sk_get(const uint32_t *cookie)
758 {
759 return (((unsigned long long)cookie[1] << 31) << 1) | cookie[0];
760 }
761
762 static const char *sctp_sstate_name[] = {
763 [SCTP_STATE_CLOSED] = "CLOSED",
764 [SCTP_STATE_COOKIE_WAIT] = "COOKIE_WAIT",
765 [SCTP_STATE_COOKIE_ECHOED] = "COOKIE_ECHOED",
766 [SCTP_STATE_ESTABLISHED] = "ESTAB",
767 [SCTP_STATE_SHUTDOWN_PENDING] = "SHUTDOWN_PENDING",
768 [SCTP_STATE_SHUTDOWN_SENT] = "SHUTDOWN_SENT",
769 [SCTP_STATE_SHUTDOWN_RECEIVED] = "SHUTDOWN_RECEIVED",
770 [SCTP_STATE_SHUTDOWN_ACK_SENT] = "ACK_SENT",
771 };
772
773 static const char * const stype_nameg[] = {
774 "UNKNOWN",
775 [SOCK_STREAM] = "STREAM",
776 [SOCK_DGRAM] = "DGRAM",
777 [SOCK_RDM] = "RDM",
778 [SOCK_SEQPACKET] = "SEQPACKET",
779 };
780
781 struct sockstat {
782 struct sockstat *next;
783 unsigned int type;
784 uint16_t prot;
785 uint16_t raw_prot;
786 inet_prefix local;
787 inet_prefix remote;
788 int lport;
789 int rport;
790 int state;
791 int rq, wq;
792 unsigned int ino;
793 unsigned int uid;
794 int refcnt;
795 unsigned int iface;
796 unsigned long long sk;
797 char *name;
798 char *peer_name;
799 __u32 mark;
800 };
801
802 struct dctcpstat {
803 unsigned int ce_state;
804 unsigned int alpha;
805 unsigned int ab_ecn;
806 unsigned int ab_tot;
807 bool enabled;
808 };
809
810 struct tcpstat {
811 struct sockstat ss;
812 unsigned int timer;
813 unsigned int timeout;
814 int probes;
815 char cong_alg[16];
816 double rto, ato, rtt, rttvar;
817 int qack, ssthresh, backoff;
818 double send_bps;
819 int snd_wscale;
820 int rcv_wscale;
821 int mss;
822 int rcv_mss;
823 int advmss;
824 unsigned int pmtu;
825 unsigned int cwnd;
826 unsigned int lastsnd;
827 unsigned int lastrcv;
828 unsigned int lastack;
829 double pacing_rate;
830 double pacing_rate_max;
831 double delivery_rate;
832 unsigned long long bytes_acked;
833 unsigned long long bytes_received;
834 unsigned int segs_out;
835 unsigned int segs_in;
836 unsigned int data_segs_out;
837 unsigned int data_segs_in;
838 unsigned int unacked;
839 unsigned int retrans;
840 unsigned int retrans_total;
841 unsigned int lost;
842 unsigned int sacked;
843 unsigned int fackets;
844 unsigned int reordering;
845 unsigned int not_sent;
846 unsigned int delivered;
847 unsigned int delivered_ce;
848 unsigned int dsack_dups;
849 unsigned int reord_seen;
850 double rcv_rtt;
851 double min_rtt;
852 int rcv_space;
853 unsigned int rcv_ssthresh;
854 unsigned long long busy_time;
855 unsigned long long rwnd_limited;
856 unsigned long long sndbuf_limited;
857 unsigned long long bytes_sent;
858 unsigned long long bytes_retrans;
859 bool has_ts_opt;
860 bool has_sack_opt;
861 bool has_ecn_opt;
862 bool has_ecnseen_opt;
863 bool has_fastopen_opt;
864 bool has_wscale_opt;
865 bool app_limited;
866 struct dctcpstat *dctcp;
867 struct tcp_bbr_info *bbr_info;
868 };
869
870 /* SCTP assocs share the same inode number with their parent endpoint. So if we
871 * have seen the inode number before, it must be an assoc instead of the next
872 * endpoint. */
873 static bool is_sctp_assoc(struct sockstat *s, const char *sock_name)
874 {
875 if (strcmp(sock_name, "sctp"))
876 return false;
877 if (!sctp_ino || sctp_ino != s->ino)
878 return false;
879 return true;
880 }
881
882 static const char *unix_netid_name(int type)
883 {
884 switch (type) {
885 case SOCK_STREAM:
886 return "u_str";
887 case SOCK_SEQPACKET:
888 return "u_seq";
889 case SOCK_DGRAM:
890 default:
891 return "u_dgr";
892 }
893 }
894
895 static const char *proto_name(int protocol)
896 {
897 switch (protocol) {
898 case 0:
899 return "raw";
900 case IPPROTO_UDP:
901 return "udp";
902 case IPPROTO_TCP:
903 return "tcp";
904 case IPPROTO_SCTP:
905 return "sctp";
906 case IPPROTO_DCCP:
907 return "dccp";
908 case IPPROTO_ICMPV6:
909 return "icmp6";
910 }
911
912 return "???";
913 }
914
915 static const char *vsock_netid_name(int type)
916 {
917 switch (type) {
918 case SOCK_STREAM:
919 return "v_str";
920 case SOCK_DGRAM:
921 return "v_dgr";
922 default:
923 return "???";
924 }
925 }
926
927 static const char *tipc_netid_name(int type)
928 {
929 switch (type) {
930 case SOCK_STREAM:
931 return "ti_st";
932 case SOCK_DGRAM:
933 return "ti_dg";
934 case SOCK_RDM:
935 return "ti_rd";
936 case SOCK_SEQPACKET:
937 return "ti_sq";
938 default:
939 return "???";
940 }
941 }
942
943 /* Allocate and initialize a new buffer chunk */
944 static struct buf_chunk *buf_chunk_new(void)
945 {
946 struct buf_chunk *new = malloc(BUF_CHUNK);
947
948 if (!new)
949 abort();
950
951 new->next = NULL;
952
953 /* This is also the last block */
954 buffer.tail = new;
955
956 /* Next token will be stored at the beginning of chunk data area, and
957 * its initial length is zero.
958 */
959 buffer.cur = (struct buf_token *)new->data;
960 buffer.cur->len = 0;
961
962 new->end = buffer.cur->data;
963
964 buffer.chunks++;
965
966 return new;
967 }
968
969 /* Return available tail room in given chunk */
970 static int buf_chunk_avail(struct buf_chunk *chunk)
971 {
972 return BUF_CHUNK - offsetof(struct buf_chunk, data) -
973 (chunk->end - chunk->data);
974 }
975
976 /* Update end pointer and token length, link new chunk if we hit the end of the
977 * current one. Return -EAGAIN if we got a new chunk, caller has to print again.
978 */
979 static int buf_update(int len)
980 {
981 struct buf_chunk *chunk = buffer.tail;
982 struct buf_token *t = buffer.cur;
983
984 /* Claim success if new content fits in the current chunk, and anyway
985 * if this is the first token in the chunk: in the latter case,
986 * allocating a new chunk won't help, so we'll just cut the output.
987 */
988 if ((len < buf_chunk_avail(chunk) && len != -1 /* glibc < 2.0.6 */) ||
989 t == (struct buf_token *)chunk->data) {
990 len = min(len, buf_chunk_avail(chunk));
991
992 /* Total field length can't exceed 2^16 bytes, cut as needed */
993 len = min(len, USHRT_MAX - t->len);
994
995 chunk->end += len;
996 t->len += len;
997 return 0;
998 }
999
1000 /* Content truncated, time to allocate more */
1001 chunk->next = buf_chunk_new();
1002
1003 /* Copy current token over to new chunk, including length descriptor */
1004 memcpy(chunk->next->data, t, sizeof(t->len) + t->len);
1005 chunk->next->end += t->len;
1006
1007 /* Discard partially written field in old chunk */
1008 chunk->end -= t->len + sizeof(t->len);
1009
1010 return -EAGAIN;
1011 }
1012
1013 /* Append content to buffer as part of the current field */
1014 __attribute__((format(printf, 1, 2)))
1015 static void out(const char *fmt, ...)
1016 {
1017 struct column *f = current_field;
1018 va_list args;
1019 char *pos;
1020 int len;
1021
1022 if (f->disabled)
1023 return;
1024
1025 if (!buffer.head)
1026 buffer.head = buf_chunk_new();
1027
1028 again: /* Append to buffer: if we have a new chunk, print again */
1029
1030 pos = buffer.cur->data + buffer.cur->len;
1031 va_start(args, fmt);
1032
1033 /* Limit to tail room. If we hit the limit, buf_update() will tell us */
1034 len = vsnprintf(pos, buf_chunk_avail(buffer.tail), fmt, args);
1035 va_end(args);
1036
1037 if (buf_update(len))
1038 goto again;
1039 }
1040
1041 static int print_left_spacing(struct column *f, int stored, int printed)
1042 {
1043 int s;
1044
1045 if (!f->width || f->align == ALIGN_LEFT)
1046 return 0;
1047
1048 s = f->width - stored - printed;
1049 if (f->align == ALIGN_CENTER)
1050 /* If count of total spacing is odd, shift right by one */
1051 s = (s + 1) / 2;
1052
1053 if (s > 0)
1054 return printf("%*c", s, ' ');
1055
1056 return 0;
1057 }
1058
1059 static void print_right_spacing(struct column *f, int printed)
1060 {
1061 int s;
1062
1063 if (!f->width || f->align == ALIGN_RIGHT)
1064 return;
1065
1066 s = f->width - printed;
1067 if (f->align == ALIGN_CENTER)
1068 s /= 2;
1069
1070 if (s > 0)
1071 printf("%*c", s, ' ');
1072 }
1073
1074 /* Done with field: update buffer pointer, start new token after current one */
1075 static void field_flush(struct column *f)
1076 {
1077 struct buf_chunk *chunk;
1078 unsigned int pad;
1079
1080 if (f->disabled)
1081 return;
1082
1083 chunk = buffer.tail;
1084 pad = buffer.cur->len % 2;
1085
1086 if (buffer.cur->len > f->max_len)
1087 f->max_len = buffer.cur->len;
1088
1089 /* We need a new chunk if we can't store the next length descriptor.
1090 * Mind the gap between end of previous token and next aligned position
1091 * for length descriptor.
1092 */
1093 if (buf_chunk_avail(chunk) - pad < sizeof(buffer.cur->len)) {
1094 chunk->end += pad;
1095 chunk->next = buf_chunk_new();
1096 return;
1097 }
1098
1099 buffer.cur = (struct buf_token *)(buffer.cur->data +
1100 LEN_ALIGN(buffer.cur->len));
1101 buffer.cur->len = 0;
1102 buffer.tail->end = buffer.cur->data;
1103 }
1104
1105 static int field_is_last(struct column *f)
1106 {
1107 return f - columns == COL_MAX - 1;
1108 }
1109
1110 /* Get the next available token in the buffer starting from the current token */
1111 static struct buf_token *buf_token_next(struct buf_token *cur)
1112 {
1113 struct buf_chunk *chunk = buffer.tail;
1114
1115 /* If we reached the end of chunk contents, get token from next chunk */
1116 if (cur->data + LEN_ALIGN(cur->len) == chunk->end) {
1117 buffer.tail = chunk = chunk->next;
1118 return chunk ? (struct buf_token *)chunk->data : NULL;
1119 }
1120
1121 return (struct buf_token *)(cur->data + LEN_ALIGN(cur->len));
1122 }
1123
1124 /* Free up all allocated buffer chunks */
1125 static void buf_free_all(void)
1126 {
1127 struct buf_chunk *tmp;
1128
1129 for (buffer.tail = buffer.head; buffer.tail; ) {
1130 tmp = buffer.tail;
1131 buffer.tail = buffer.tail->next;
1132 free(tmp);
1133 }
1134 buffer.head = NULL;
1135 buffer.chunks = 0;
1136 }
1137
1138 /* Get current screen width, returns -1 if TIOCGWINSZ fails */
1139 static int render_screen_width(void)
1140 {
1141 int width = -1;
1142
1143 if (isatty(STDOUT_FILENO)) {
1144 struct winsize w;
1145
1146 if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) != -1) {
1147 if (w.ws_col > 0)
1148 width = w.ws_col;
1149 }
1150 }
1151
1152 return width;
1153 }
1154
1155 /* Calculate column width from contents length. If columns don't fit on one
1156 * line, break them into the least possible amount of lines and keep them
1157 * aligned across lines. Available screen space is equally spread between fields
1158 * as additional spacing.
1159 */
1160 static void render_calc_width(void)
1161 {
1162 int screen_width, first, len = 0, linecols = 0;
1163 struct column *c, *eol = columns - 1;
1164 bool compact_output = false;
1165
1166 screen_width = render_screen_width();
1167 if (screen_width == -1) {
1168 screen_width = INT_MAX;
1169 compact_output = true;
1170 }
1171
1172 /* First pass: set width for each column to measured content length */
1173 for (first = 1, c = columns; c - columns < COL_MAX; c++) {
1174 if (c->disabled)
1175 continue;
1176
1177 if (!first && c->max_len)
1178 c->width = c->max_len + strlen(c->ldelim);
1179 else
1180 c->width = c->max_len;
1181
1182 /* But don't exceed screen size. If we exceed the screen size
1183 * for even a single field, it will just start on a line of its
1184 * own and then naturally wrap.
1185 */
1186 c->width = min(c->width, screen_width);
1187
1188 if (c->width)
1189 first = 0;
1190 }
1191
1192 if (compact_output) {
1193 /* Compact output, skip extending columns. */
1194 return;
1195 }
1196
1197 /* Second pass: find out newlines and distribute available spacing */
1198 for (c = columns; c - columns < COL_MAX; c++) {
1199 int pad, spacing, rem, last;
1200 struct column *tmp;
1201
1202 if (!c->width)
1203 continue;
1204
1205 linecols++;
1206 len += c->width;
1207
1208 for (last = 1, tmp = c + 1; tmp - columns < COL_MAX; tmp++) {
1209 if (tmp->width) {
1210 last = 0;
1211 break;
1212 }
1213 }
1214
1215 if (!last && len < screen_width) {
1216 /* Columns fit on screen so far, nothing to do yet */
1217 continue;
1218 }
1219
1220 if (len == screen_width) {
1221 /* Exact fit, just start with new line */
1222 goto newline;
1223 }
1224
1225 if (len > screen_width) {
1226 /* Screen width exceeded: go back one column */
1227 len -= c->width;
1228 c--;
1229 linecols--;
1230 }
1231
1232 /* Distribute remaining space to columns on this line */
1233 pad = screen_width - len;
1234 spacing = pad / linecols;
1235 rem = pad % linecols;
1236 for (tmp = c; tmp > eol; tmp--) {
1237 if (!tmp->width)
1238 continue;
1239
1240 tmp->width += spacing;
1241 if (rem) {
1242 tmp->width++;
1243 rem--;
1244 }
1245 }
1246
1247 newline:
1248 /* Line break: reset line counters, mark end-of-line */
1249 eol = c;
1250 len = 0;
1251 linecols = 0;
1252 }
1253 }
1254
1255 /* Render buffered output with spacing and delimiters, then free up buffers */
1256 static void render(void)
1257 {
1258 struct buf_token *token;
1259 int printed, line_started = 0;
1260 struct column *f;
1261
1262 if (!buffer.head)
1263 return;
1264
1265 token = (struct buf_token *)buffer.head->data;
1266
1267 /* Ensure end alignment of last token, it wasn't necessarily flushed */
1268 buffer.tail->end += buffer.cur->len % 2;
1269
1270 render_calc_width();
1271
1272 /* Rewind and replay */
1273 buffer.tail = buffer.head;
1274
1275 f = columns;
1276 while (!f->width)
1277 f++;
1278
1279 while (token) {
1280 /* Print left delimiter only if we already started a line */
1281 if (line_started++)
1282 printed = printf("%s", f->ldelim);
1283 else
1284 printed = 0;
1285
1286 /* Print field content from token data with spacing */
1287 printed += print_left_spacing(f, token->len, printed);
1288 printed += fwrite(token->data, 1, token->len, stdout);
1289 print_right_spacing(f, printed);
1290
1291 /* Go to next non-empty field, deal with end-of-line */
1292 do {
1293 if (field_is_last(f)) {
1294 printf("\n");
1295 f = columns;
1296 line_started = 0;
1297 } else {
1298 f++;
1299 }
1300 } while (f->disabled);
1301
1302 token = buf_token_next(token);
1303 }
1304 /* Deal with final end-of-line when the last non-empty field printed
1305 * is not the last field.
1306 */
1307 if (line_started)
1308 printf("\n");
1309
1310 buf_free_all();
1311 current_field = columns;
1312 }
1313
1314 /* Move to next field, and render buffer if we reached the maximum number of
1315 * chunks, at the last field in a line.
1316 */
1317 static void field_next(void)
1318 {
1319 if (field_is_last(current_field) && buffer.chunks >= BUF_CHUNKS_MAX) {
1320 render();
1321 return;
1322 }
1323
1324 field_flush(current_field);
1325 if (field_is_last(current_field))
1326 current_field = columns;
1327 else
1328 current_field++;
1329 }
1330
1331 /* Walk through fields and flush them until we reach the desired one */
1332 static void field_set(enum col_id id)
1333 {
1334 while (id != current_field - columns)
1335 field_next();
1336 }
1337
1338 /* Print header for all non-empty columns */
1339 static void print_header(void)
1340 {
1341 while (!field_is_last(current_field)) {
1342 if (!current_field->disabled)
1343 out("%s", current_field->header);
1344 field_next();
1345 }
1346 }
1347
1348 static void sock_state_print(struct sockstat *s)
1349 {
1350 const char *sock_name;
1351 static const char * const sstate_name[] = {
1352 "UNKNOWN",
1353 [SS_ESTABLISHED] = "ESTAB",
1354 [SS_SYN_SENT] = "SYN-SENT",
1355 [SS_SYN_RECV] = "SYN-RECV",
1356 [SS_FIN_WAIT1] = "FIN-WAIT-1",
1357 [SS_FIN_WAIT2] = "FIN-WAIT-2",
1358 [SS_TIME_WAIT] = "TIME-WAIT",
1359 [SS_CLOSE] = "UNCONN",
1360 [SS_CLOSE_WAIT] = "CLOSE-WAIT",
1361 [SS_LAST_ACK] = "LAST-ACK",
1362 [SS_LISTEN] = "LISTEN",
1363 [SS_CLOSING] = "CLOSING",
1364 };
1365
1366 switch (s->local.family) {
1367 case AF_UNIX:
1368 sock_name = unix_netid_name(s->type);
1369 break;
1370 case AF_INET:
1371 case AF_INET6:
1372 sock_name = proto_name(s->type);
1373 break;
1374 case AF_PACKET:
1375 sock_name = s->type == SOCK_RAW ? "p_raw" : "p_dgr";
1376 break;
1377 case AF_NETLINK:
1378 sock_name = "nl";
1379 break;
1380 case AF_TIPC:
1381 sock_name = tipc_netid_name(s->type);
1382 break;
1383 case AF_VSOCK:
1384 sock_name = vsock_netid_name(s->type);
1385 break;
1386 case AF_XDP:
1387 sock_name = "xdp";
1388 break;
1389 default:
1390 sock_name = "unknown";
1391 }
1392
1393 if (is_sctp_assoc(s, sock_name)) {
1394 field_set(COL_STATE); /* Empty Netid field */
1395 out("`- %s", sctp_sstate_name[s->state]);
1396 } else {
1397 field_set(COL_NETID);
1398 out("%s", sock_name);
1399 field_set(COL_STATE);
1400 out("%s", sstate_name[s->state]);
1401 }
1402
1403 field_set(COL_RECVQ);
1404 out("%-6d", s->rq);
1405 field_set(COL_SENDQ);
1406 out("%-6d", s->wq);
1407 field_set(COL_ADDR);
1408 }
1409
1410 static void sock_details_print(struct sockstat *s)
1411 {
1412 if (s->uid)
1413 out(" uid:%u", s->uid);
1414
1415 out(" ino:%u", s->ino);
1416 out(" sk:%llx", s->sk);
1417
1418 if (s->mark)
1419 out(" fwmark:0x%x", s->mark);
1420 }
1421
1422 static void sock_addr_print(const char *addr, char *delim, const char *port,
1423 const char *ifname)
1424 {
1425 if (ifname)
1426 out("%s" "%%" "%s%s", addr, ifname, delim);
1427 else
1428 out("%s%s", addr, delim);
1429
1430 field_next();
1431 out("%s", port);
1432 field_next();
1433 }
1434
1435 static const char *print_ms_timer(unsigned int timeout)
1436 {
1437 static char buf[64];
1438 int secs, msecs, minutes;
1439
1440 secs = timeout/1000;
1441 minutes = secs/60;
1442 secs = secs%60;
1443 msecs = timeout%1000;
1444 buf[0] = 0;
1445 if (minutes) {
1446 msecs = 0;
1447 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
1448 if (minutes > 9)
1449 secs = 0;
1450 }
1451 if (secs) {
1452 if (secs > 9)
1453 msecs = 0;
1454 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
1455 }
1456 if (msecs)
1457 sprintf(buf+strlen(buf), "%03dms", msecs);
1458 return buf;
1459 }
1460
1461 struct scache {
1462 struct scache *next;
1463 int port;
1464 char *name;
1465 const char *proto;
1466 };
1467
1468 static struct scache *rlist;
1469
1470 static void init_service_resolver(void)
1471 {
1472 char buf[128];
1473 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
1474
1475 if (!fp)
1476 return;
1477
1478 if (!fgets(buf, sizeof(buf), fp)) {
1479 pclose(fp);
1480 return;
1481 }
1482 while (fgets(buf, sizeof(buf), fp) != NULL) {
1483 unsigned int progn, port;
1484 char proto[128], prog[128] = "rpc.";
1485 struct scache *c;
1486
1487 if (sscanf(buf, "%u %*d %s %u %s",
1488 &progn, proto, &port, prog+4) != 4)
1489 continue;
1490
1491 if (!(c = malloc(sizeof(*c))))
1492 continue;
1493
1494 c->port = port;
1495 c->name = strdup(prog);
1496 if (strcmp(proto, TCP_PROTO) == 0)
1497 c->proto = TCP_PROTO;
1498 else if (strcmp(proto, UDP_PROTO) == 0)
1499 c->proto = UDP_PROTO;
1500 else if (strcmp(proto, SCTP_PROTO) == 0)
1501 c->proto = SCTP_PROTO;
1502 else
1503 c->proto = NULL;
1504 c->next = rlist;
1505 rlist = c;
1506 }
1507 pclose(fp);
1508 }
1509
1510 /* Even do not try default linux ephemeral port ranges:
1511 * default /etc/services contains so much of useless crap
1512 * wouldbe "allocated" to this area that resolution
1513 * is really harmful. I shrug each time when seeing
1514 * "socks" or "cfinger" in dumps.
1515 */
1516 static int is_ephemeral(int port)
1517 {
1518 static int min = 0, max;
1519
1520 if (!min) {
1521 FILE *f = ephemeral_ports_open();
1522
1523 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
1524 min = 1024;
1525 max = 4999;
1526 }
1527 if (f)
1528 fclose(f);
1529 }
1530 return port >= min && port <= max;
1531 }
1532
1533
1534 static const char *__resolve_service(int port)
1535 {
1536 struct scache *c;
1537
1538 for (c = rlist; c; c = c->next) {
1539 if (c->port == port && c->proto == dg_proto)
1540 return c->name;
1541 }
1542
1543 if (!is_ephemeral(port)) {
1544 static int notfirst;
1545 struct servent *se;
1546
1547 if (!notfirst) {
1548 setservent(1);
1549 notfirst = 1;
1550 }
1551 se = getservbyport(htons(port), dg_proto);
1552 if (se)
1553 return se->s_name;
1554 }
1555
1556 return NULL;
1557 }
1558
1559 #define SCACHE_BUCKETS 1024
1560 static struct scache *cache_htab[SCACHE_BUCKETS];
1561
1562 static const char *resolve_service(int port)
1563 {
1564 static char buf[128];
1565 struct scache *c;
1566 const char *res;
1567 int hash;
1568
1569 if (port == 0) {
1570 buf[0] = '*';
1571 buf[1] = 0;
1572 return buf;
1573 }
1574
1575 if (numeric)
1576 goto do_numeric;
1577
1578 if (dg_proto == RAW_PROTO)
1579 return inet_proto_n2a(port, buf, sizeof(buf));
1580
1581
1582 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
1583
1584 for (c = cache_htab[hash]; c; c = c->next) {
1585 if (c->port == port && c->proto == dg_proto)
1586 goto do_cache;
1587 }
1588
1589 c = malloc(sizeof(*c));
1590 if (!c)
1591 goto do_numeric;
1592 res = __resolve_service(port);
1593 c->port = port;
1594 c->name = res ? strdup(res) : NULL;
1595 c->proto = dg_proto;
1596 c->next = cache_htab[hash];
1597 cache_htab[hash] = c;
1598
1599 do_cache:
1600 if (c->name)
1601 return c->name;
1602
1603 do_numeric:
1604 sprintf(buf, "%u", port);
1605 return buf;
1606 }
1607
1608 static void inet_addr_print(const inet_prefix *a, int port,
1609 unsigned int ifindex, bool v6only)
1610 {
1611 char buf[1024];
1612 const char *ap = buf;
1613 const char *ifname = NULL;
1614
1615 if (a->family == AF_INET) {
1616 ap = format_host(AF_INET, 4, a->data);
1617 } else {
1618 if (!v6only &&
1619 !memcmp(a->data, &in6addr_any, sizeof(in6addr_any))) {
1620 buf[0] = '*';
1621 buf[1] = 0;
1622 } else {
1623 ap = format_host(a->family, 16, a->data);
1624
1625 /* Numeric IPv6 addresses should be bracketed */
1626 if (strchr(ap, ':')) {
1627 snprintf(buf, sizeof(buf),
1628 "[%s]", ap);
1629 ap = buf;
1630 }
1631 }
1632 }
1633
1634 if (ifindex)
1635 ifname = ll_index_to_name(ifindex);
1636
1637 sock_addr_print(ap, ":", resolve_service(port), ifname);
1638 }
1639
1640 struct aafilter {
1641 inet_prefix addr;
1642 int port;
1643 unsigned int iface;
1644 __u32 mark;
1645 __u32 mask;
1646 struct aafilter *next;
1647 };
1648
1649 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1650 int plen)
1651 {
1652 if (!inet_addr_match(a, p, plen))
1653 return 0;
1654
1655 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1656 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1657 * sockets. Fair? */
1658 if (p->family == AF_INET && a->family == AF_INET6) {
1659 if (a->data[0] == 0 && a->data[1] == 0 &&
1660 a->data[2] == htonl(0xffff)) {
1661 inet_prefix tmp = *a;
1662
1663 tmp.data[0] = a->data[3];
1664 return inet_addr_match(&tmp, p, plen);
1665 }
1666 }
1667 return 1;
1668 }
1669
1670 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1671 {
1672 char *addr, *pattern;
1673
1674 memcpy(&addr, a->data, sizeof(addr));
1675 memcpy(&pattern, p->data, sizeof(pattern));
1676 if (pattern == NULL)
1677 return 1;
1678 if (addr == NULL)
1679 addr = "";
1680 return !fnmatch(pattern, addr, 0);
1681 }
1682
1683 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1684 {
1685 switch (f->type) {
1686 case SSF_S_AUTO:
1687 {
1688 if (s->local.family == AF_UNIX) {
1689 char *p;
1690
1691 memcpy(&p, s->local.data, sizeof(p));
1692 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1693 strspn(p+1, "0123456789abcdef") == 5);
1694 }
1695 if (s->local.family == AF_PACKET)
1696 return s->lport == 0 && s->local.data[0] == 0;
1697 if (s->local.family == AF_NETLINK)
1698 return s->lport < 0;
1699 if (s->local.family == AF_VSOCK)
1700 return s->lport > 1023;
1701
1702 return is_ephemeral(s->lport);
1703 }
1704 case SSF_DCOND:
1705 {
1706 struct aafilter *a = (void *)f->pred;
1707
1708 if (a->addr.family == AF_UNIX)
1709 return unix_match(&s->remote, &a->addr);
1710 if (a->port != -1 && a->port != s->rport)
1711 return 0;
1712 if (a->addr.bitlen) {
1713 do {
1714 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1715 return 1;
1716 } while ((a = a->next) != NULL);
1717 return 0;
1718 }
1719 return 1;
1720 }
1721 case SSF_SCOND:
1722 {
1723 struct aafilter *a = (void *)f->pred;
1724
1725 if (a->addr.family == AF_UNIX)
1726 return unix_match(&s->local, &a->addr);
1727 if (a->port != -1 && a->port != s->lport)
1728 return 0;
1729 if (a->addr.bitlen) {
1730 do {
1731 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1732 return 1;
1733 } while ((a = a->next) != NULL);
1734 return 0;
1735 }
1736 return 1;
1737 }
1738 case SSF_D_GE:
1739 {
1740 struct aafilter *a = (void *)f->pred;
1741
1742 return s->rport >= a->port;
1743 }
1744 case SSF_D_LE:
1745 {
1746 struct aafilter *a = (void *)f->pred;
1747
1748 return s->rport <= a->port;
1749 }
1750 case SSF_S_GE:
1751 {
1752 struct aafilter *a = (void *)f->pred;
1753
1754 return s->lport >= a->port;
1755 }
1756 case SSF_S_LE:
1757 {
1758 struct aafilter *a = (void *)f->pred;
1759
1760 return s->lport <= a->port;
1761 }
1762 case SSF_DEVCOND:
1763 {
1764 struct aafilter *a = (void *)f->pred;
1765
1766 return s->iface == a->iface;
1767 }
1768 case SSF_MARKMASK:
1769 {
1770 struct aafilter *a = (void *)f->pred;
1771
1772 return (s->mark & a->mask) == a->mark;
1773 }
1774 /* Yup. It is recursion. Sorry. */
1775 case SSF_AND:
1776 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1777 case SSF_OR:
1778 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1779 case SSF_NOT:
1780 return !run_ssfilter(f->pred, s);
1781 default:
1782 abort();
1783 }
1784 }
1785
1786 /* Relocate external jumps by reloc. */
1787 static void ssfilter_patch(char *a, int len, int reloc)
1788 {
1789 while (len > 0) {
1790 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)a;
1791
1792 if (op->no == len+4)
1793 op->no += reloc;
1794 len -= op->yes;
1795 a += op->yes;
1796 }
1797 if (len < 0)
1798 abort();
1799 }
1800
1801 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1802 {
1803 switch (f->type) {
1804 case SSF_S_AUTO:
1805 {
1806 if (!(*bytecode = malloc(4))) abort();
1807 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1808 return 4;
1809 }
1810 case SSF_DCOND:
1811 case SSF_SCOND:
1812 {
1813 struct aafilter *a = (void *)f->pred;
1814 struct aafilter *b;
1815 char *ptr;
1816 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1817 int len = 0;
1818
1819 for (b = a; b; b = b->next) {
1820 len += 4 + sizeof(struct inet_diag_hostcond);
1821 if (a->addr.family == AF_INET6)
1822 len += 16;
1823 else
1824 len += 4;
1825 if (b->next)
1826 len += 4;
1827 }
1828 if (!(ptr = malloc(len))) abort();
1829 *bytecode = ptr;
1830 for (b = a; b; b = b->next) {
1831 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1832 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1833 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1834 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond *)(ptr+4);
1835
1836 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1837 cond->family = a->addr.family;
1838 cond->port = a->port;
1839 cond->prefix_len = a->addr.bitlen;
1840 memcpy(cond->addr, a->addr.data, alen);
1841 ptr += oplen;
1842 if (b->next) {
1843 op = (struct inet_diag_bc_op *)ptr;
1844 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1845 ptr += 4;
1846 }
1847 }
1848 return ptr - *bytecode;
1849 }
1850 case SSF_D_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_D_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_D_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_D_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 case SSF_S_GE:
1869 {
1870 struct aafilter *x = (void *)f->pred;
1871
1872 if (!(*bytecode = malloc(8))) abort();
1873 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1874 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1875 return 8;
1876 }
1877 case SSF_S_LE:
1878 {
1879 struct aafilter *x = (void *)f->pred;
1880
1881 if (!(*bytecode = malloc(8))) abort();
1882 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1883 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1884 return 8;
1885 }
1886
1887 case SSF_AND:
1888 {
1889 char *a1 = NULL, *a2 = NULL, *a;
1890 int l1, l2;
1891
1892 l1 = ssfilter_bytecompile(f->pred, &a1);
1893 l2 = ssfilter_bytecompile(f->post, &a2);
1894 if (!l1 || !l2) {
1895 free(a1);
1896 free(a2);
1897 return 0;
1898 }
1899 if (!(a = malloc(l1+l2))) abort();
1900 memcpy(a, a1, l1);
1901 memcpy(a+l1, a2, l2);
1902 free(a1); free(a2);
1903 ssfilter_patch(a, l1, l2);
1904 *bytecode = a;
1905 return l1+l2;
1906 }
1907 case SSF_OR:
1908 {
1909 char *a1 = NULL, *a2 = NULL, *a;
1910 int l1, l2;
1911
1912 l1 = ssfilter_bytecompile(f->pred, &a1);
1913 l2 = ssfilter_bytecompile(f->post, &a2);
1914 if (!l1 || !l2) {
1915 free(a1);
1916 free(a2);
1917 return 0;
1918 }
1919 if (!(a = malloc(l1+l2+4))) abort();
1920 memcpy(a, a1, l1);
1921 memcpy(a+l1+4, a2, l2);
1922 free(a1); free(a2);
1923 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1924 *bytecode = a;
1925 return l1+l2+4;
1926 }
1927 case SSF_NOT:
1928 {
1929 char *a1 = NULL, *a;
1930 int l1;
1931
1932 l1 = ssfilter_bytecompile(f->pred, &a1);
1933 if (!l1) {
1934 free(a1);
1935 return 0;
1936 }
1937 if (!(a = malloc(l1+4))) abort();
1938 memcpy(a, a1, l1);
1939 free(a1);
1940 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1941 *bytecode = a;
1942 return l1+4;
1943 }
1944 case SSF_DEVCOND:
1945 {
1946 /* bytecompile for SSF_DEVCOND not supported yet */
1947 return 0;
1948 }
1949 case SSF_MARKMASK:
1950 {
1951 struct aafilter *a = (void *)f->pred;
1952 struct instr {
1953 struct inet_diag_bc_op op;
1954 struct inet_diag_markcond cond;
1955 };
1956 int inslen = sizeof(struct instr);
1957
1958 if (!(*bytecode = malloc(inslen))) abort();
1959 ((struct instr *)*bytecode)[0] = (struct instr) {
1960 { INET_DIAG_BC_MARK_COND, inslen, inslen + 4 },
1961 { a->mark, a->mask},
1962 };
1963
1964 return inslen;
1965 }
1966 default:
1967 abort();
1968 }
1969 }
1970
1971 static int remember_he(struct aafilter *a, struct hostent *he)
1972 {
1973 char **ptr = he->h_addr_list;
1974 int cnt = 0;
1975 int len;
1976
1977 if (he->h_addrtype == AF_INET)
1978 len = 4;
1979 else if (he->h_addrtype == AF_INET6)
1980 len = 16;
1981 else
1982 return 0;
1983
1984 while (*ptr) {
1985 struct aafilter *b = a;
1986
1987 if (a->addr.bitlen) {
1988 if ((b = malloc(sizeof(*b))) == NULL)
1989 return cnt;
1990 *b = *a;
1991 a->next = b;
1992 }
1993 memcpy(b->addr.data, *ptr, len);
1994 b->addr.bytelen = len;
1995 b->addr.bitlen = len*8;
1996 b->addr.family = he->h_addrtype;
1997 ptr++;
1998 cnt++;
1999 }
2000 return cnt;
2001 }
2002
2003 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
2004 {
2005 static int notfirst;
2006 int cnt = 0;
2007 struct hostent *he;
2008
2009 a->addr.bitlen = 0;
2010 if (!notfirst) {
2011 sethostent(1);
2012 notfirst = 1;
2013 }
2014 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
2015 if (he)
2016 cnt = remember_he(a, he);
2017 if (fam == AF_UNSPEC) {
2018 he = gethostbyname2(addr, AF_INET6);
2019 if (he)
2020 cnt += remember_he(a, he);
2021 }
2022 return !cnt;
2023 }
2024
2025 static int xll_initted;
2026
2027 static void xll_init(void)
2028 {
2029 struct rtnl_handle rth;
2030
2031 if (rtnl_open(&rth, 0) < 0)
2032 exit(1);
2033
2034 ll_init_map(&rth);
2035 rtnl_close(&rth);
2036 xll_initted = 1;
2037 }
2038
2039 static const char *xll_index_to_name(int index)
2040 {
2041 if (!xll_initted)
2042 xll_init();
2043 return ll_index_to_name(index);
2044 }
2045
2046 static int xll_name_to_index(const char *dev)
2047 {
2048 if (!xll_initted)
2049 xll_init();
2050 return ll_name_to_index(dev);
2051 }
2052
2053 void *parse_devcond(char *name)
2054 {
2055 struct aafilter a = { .iface = 0 };
2056 struct aafilter *res;
2057
2058 a.iface = xll_name_to_index(name);
2059 if (a.iface == 0) {
2060 char *end;
2061 unsigned long n;
2062
2063 n = strtoul(name, &end, 0);
2064 if (!end || end == name || *end || n > UINT_MAX)
2065 return NULL;
2066
2067 a.iface = n;
2068 }
2069
2070 res = malloc(sizeof(*res));
2071 *res = a;
2072
2073 return res;
2074 }
2075
2076 static void vsock_set_inet_prefix(inet_prefix *a, __u32 cid)
2077 {
2078 *a = (inet_prefix){
2079 .bytelen = sizeof(cid),
2080 .family = AF_VSOCK,
2081 };
2082 memcpy(a->data, &cid, sizeof(cid));
2083 }
2084
2085 void *parse_hostcond(char *addr, bool is_port)
2086 {
2087 char *port = NULL;
2088 struct aafilter a = { .port = -1 };
2089 struct aafilter *res;
2090 int fam = preferred_family;
2091 struct filter *f = &current_filter;
2092
2093 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
2094 char *p;
2095
2096 a.addr.family = AF_UNIX;
2097 if (strncmp(addr, "unix:", 5) == 0)
2098 addr += 5;
2099 p = strdup(addr);
2100 a.addr.bitlen = 8*strlen(p);
2101 memcpy(a.addr.data, &p, sizeof(p));
2102 fam = AF_UNIX;
2103 goto out;
2104 }
2105
2106 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
2107 a.addr.family = AF_PACKET;
2108 a.addr.bitlen = 0;
2109 if (strncmp(addr, "link:", 5) == 0)
2110 addr += 5;
2111 port = strchr(addr, ':');
2112 if (port) {
2113 *port = 0;
2114 if (port[1] && strcmp(port+1, "*")) {
2115 if (get_integer(&a.port, port+1, 0)) {
2116 if ((a.port = xll_name_to_index(port+1)) <= 0)
2117 return NULL;
2118 }
2119 }
2120 }
2121 if (addr[0] && strcmp(addr, "*")) {
2122 unsigned short tmp;
2123
2124 a.addr.bitlen = 32;
2125 if (ll_proto_a2n(&tmp, addr))
2126 return NULL;
2127 a.addr.data[0] = ntohs(tmp);
2128 }
2129 fam = AF_PACKET;
2130 goto out;
2131 }
2132
2133 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
2134 a.addr.family = AF_NETLINK;
2135 a.addr.bitlen = 0;
2136 if (strncmp(addr, "netlink:", 8) == 0)
2137 addr += 8;
2138 port = strchr(addr, ':');
2139 if (port) {
2140 *port = 0;
2141 if (port[1] && strcmp(port+1, "*")) {
2142 if (get_integer(&a.port, port+1, 0)) {
2143 if (strcmp(port+1, "kernel") == 0)
2144 a.port = 0;
2145 else
2146 return NULL;
2147 }
2148 }
2149 }
2150 if (addr[0] && strcmp(addr, "*")) {
2151 a.addr.bitlen = 32;
2152 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
2153 return NULL;
2154 }
2155 fam = AF_NETLINK;
2156 goto out;
2157 }
2158
2159 if (fam == AF_VSOCK || strncmp(addr, "vsock:", 6) == 0) {
2160 __u32 cid = ~(__u32)0;
2161
2162 a.addr.family = AF_VSOCK;
2163 if (strncmp(addr, "vsock:", 6) == 0)
2164 addr += 6;
2165
2166 if (is_port)
2167 port = addr;
2168 else {
2169 port = strchr(addr, ':');
2170 if (port) {
2171 *port = '\0';
2172 port++;
2173 }
2174 }
2175
2176 if (port && strcmp(port, "*") &&
2177 get_u32((__u32 *)&a.port, port, 0))
2178 return NULL;
2179
2180 if (addr[0] && strcmp(addr, "*")) {
2181 a.addr.bitlen = 32;
2182 if (get_u32(&cid, addr, 0))
2183 return NULL;
2184 }
2185 vsock_set_inet_prefix(&a.addr, cid);
2186 fam = AF_VSOCK;
2187 goto out;
2188 }
2189
2190 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
2191 fam = AF_INET;
2192 if (!strncmp(addr, "inet:", 5))
2193 addr += 5;
2194 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
2195 fam = AF_INET6;
2196 if (!strncmp(addr, "inet6:", 6))
2197 addr += 6;
2198 }
2199
2200 /* URL-like literal [] */
2201 if (addr[0] == '[') {
2202 addr++;
2203 if ((port = strchr(addr, ']')) == NULL)
2204 return NULL;
2205 *port++ = 0;
2206 } else if (addr[0] == '*') {
2207 port = addr+1;
2208 } else {
2209 port = strrchr(strchr(addr, '/') ? : addr, ':');
2210 }
2211
2212 if (is_port)
2213 port = addr;
2214
2215 if (port && *port) {
2216 if (*port == ':')
2217 *port++ = 0;
2218
2219 if (*port && *port != '*') {
2220 if (get_integer(&a.port, port, 0)) {
2221 struct servent *se1 = NULL;
2222 struct servent *se2 = NULL;
2223
2224 if (current_filter.dbs&(1<<UDP_DB))
2225 se1 = getservbyname(port, UDP_PROTO);
2226 if (current_filter.dbs&(1<<TCP_DB))
2227 se2 = getservbyname(port, TCP_PROTO);
2228 if (se1 && se2 && se1->s_port != se2->s_port) {
2229 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2230 return NULL;
2231 }
2232 if (!se1)
2233 se1 = se2;
2234 if (se1) {
2235 a.port = ntohs(se1->s_port);
2236 } else {
2237 struct scache *s;
2238
2239 for (s = rlist; s; s = s->next) {
2240 if ((s->proto == UDP_PROTO &&
2241 (current_filter.dbs&(1<<UDP_DB))) ||
2242 (s->proto == TCP_PROTO &&
2243 (current_filter.dbs&(1<<TCP_DB)))) {
2244 if (s->name && strcmp(s->name, port) == 0) {
2245 if (a.port > 0 && a.port != s->port) {
2246 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2247 return NULL;
2248 }
2249 a.port = s->port;
2250 }
2251 }
2252 }
2253 if (a.port <= 0) {
2254 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
2255 return NULL;
2256 }
2257 }
2258 }
2259 }
2260 }
2261 if (!is_port && *addr && *addr != '*') {
2262 if (get_prefix_1(&a.addr, addr, fam)) {
2263 if (get_dns_host(&a, addr, fam)) {
2264 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
2265 return NULL;
2266 }
2267 }
2268 }
2269
2270 out:
2271 if (fam != AF_UNSPEC) {
2272 int states = f->states;
2273 f->families = 0;
2274 filter_af_set(f, fam);
2275 filter_states_set(f, states);
2276 }
2277
2278 res = malloc(sizeof(*res));
2279 if (res)
2280 memcpy(res, &a, sizeof(a));
2281 return res;
2282 }
2283
2284 void *parse_markmask(const char *markmask)
2285 {
2286 struct aafilter a, *res;
2287
2288 if (strchr(markmask, '/')) {
2289 if (sscanf(markmask, "%i/%i", &a.mark, &a.mask) != 2)
2290 return NULL;
2291 } else {
2292 a.mask = 0xffffffff;
2293 if (sscanf(markmask, "%i", &a.mark) != 1)
2294 return NULL;
2295 }
2296
2297 res = malloc(sizeof(*res));
2298 if (res)
2299 memcpy(res, &a, sizeof(a));
2300 return res;
2301 }
2302
2303 static void proc_ctx_print(struct sockstat *s)
2304 {
2305 char *buf;
2306
2307 if (show_proc_ctx || show_sock_ctx) {
2308 if (find_entry(s->ino, &buf,
2309 (show_proc_ctx & show_sock_ctx) ?
2310 PROC_SOCK_CTX : PROC_CTX) > 0) {
2311 out(" users:(%s)", buf);
2312 free(buf);
2313 }
2314 } else if (show_users) {
2315 if (find_entry(s->ino, &buf, USERS) > 0) {
2316 out(" users:(%s)", buf);
2317 free(buf);
2318 }
2319 }
2320 }
2321
2322 static void inet_stats_print(struct sockstat *s, bool v6only)
2323 {
2324 sock_state_print(s);
2325
2326 inet_addr_print(&s->local, s->lport, s->iface, v6only);
2327 inet_addr_print(&s->remote, s->rport, 0, v6only);
2328
2329 proc_ctx_print(s);
2330 }
2331
2332 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
2333 sockstat * s)
2334 {
2335 s->local.family = s->remote.family = family;
2336 if (family == AF_INET) {
2337 sscanf(loc, "%x:%x", s->local.data, (unsigned *)&s->lport);
2338 sscanf(rem, "%x:%x", s->remote.data, (unsigned *)&s->rport);
2339 s->local.bytelen = s->remote.bytelen = 4;
2340 return 0;
2341 } else {
2342 sscanf(loc, "%08x%08x%08x%08x:%x",
2343 s->local.data,
2344 s->local.data + 1,
2345 s->local.data + 2,
2346 s->local.data + 3,
2347 &s->lport);
2348 sscanf(rem, "%08x%08x%08x%08x:%x",
2349 s->remote.data,
2350 s->remote.data + 1,
2351 s->remote.data + 2,
2352 s->remote.data + 3,
2353 &s->rport);
2354 s->local.bytelen = s->remote.bytelen = 16;
2355 return 0;
2356 }
2357 return -1;
2358 }
2359
2360 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
2361 {
2362 char *p;
2363
2364 if ((p = strchr(line, ':')) == NULL)
2365 return -1;
2366
2367 *loc = p+2;
2368 if ((p = strchr(*loc, ':')) == NULL)
2369 return -1;
2370
2371 p[5] = 0;
2372 *rem = p+6;
2373 if ((p = strchr(*rem, ':')) == NULL)
2374 return -1;
2375
2376 p[5] = 0;
2377 *data = p+6;
2378 return 0;
2379 }
2380
2381 static char *sprint_bw(char *buf, double bw)
2382 {
2383 if (numeric)
2384 sprintf(buf, "%.0f", bw);
2385 else if (bw >= 1e9)
2386 sprintf(buf, "%.3gG", bw / 1e9);
2387 else if (bw >= 1e6)
2388 sprintf(buf, "%.3gM", bw / 1e6);
2389 else if (bw >= 1e3)
2390 sprintf(buf, "%.3gK", bw / 1e3);
2391 else
2392 sprintf(buf, "%g", bw);
2393
2394 return buf;
2395 }
2396
2397 static void sctp_stats_print(struct sctp_info *s)
2398 {
2399 if (s->sctpi_tag)
2400 out(" tag:%x", s->sctpi_tag);
2401 if (s->sctpi_state)
2402 out(" state:%s", sctp_sstate_name[s->sctpi_state]);
2403 if (s->sctpi_rwnd)
2404 out(" rwnd:%d", s->sctpi_rwnd);
2405 if (s->sctpi_unackdata)
2406 out(" unackdata:%d", s->sctpi_unackdata);
2407 if (s->sctpi_penddata)
2408 out(" penddata:%d", s->sctpi_penddata);
2409 if (s->sctpi_instrms)
2410 out(" instrms:%d", s->sctpi_instrms);
2411 if (s->sctpi_outstrms)
2412 out(" outstrms:%d", s->sctpi_outstrms);
2413 if (s->sctpi_inqueue)
2414 out(" inqueue:%d", s->sctpi_inqueue);
2415 if (s->sctpi_outqueue)
2416 out(" outqueue:%d", s->sctpi_outqueue);
2417 if (s->sctpi_overall_error)
2418 out(" overerr:%d", s->sctpi_overall_error);
2419 if (s->sctpi_max_burst)
2420 out(" maxburst:%d", s->sctpi_max_burst);
2421 if (s->sctpi_maxseg)
2422 out(" maxseg:%d", s->sctpi_maxseg);
2423 if (s->sctpi_peer_rwnd)
2424 out(" prwnd:%d", s->sctpi_peer_rwnd);
2425 if (s->sctpi_peer_tag)
2426 out(" ptag:%x", s->sctpi_peer_tag);
2427 if (s->sctpi_peer_capable)
2428 out(" pcapable:%d", s->sctpi_peer_capable);
2429 if (s->sctpi_peer_sack)
2430 out(" psack:%d", s->sctpi_peer_sack);
2431 if (s->sctpi_s_autoclose)
2432 out(" autoclose:%d", s->sctpi_s_autoclose);
2433 if (s->sctpi_s_adaptation_ind)
2434 out(" adapind:%d", s->sctpi_s_adaptation_ind);
2435 if (s->sctpi_s_pd_point)
2436 out(" pdpoint:%d", s->sctpi_s_pd_point);
2437 if (s->sctpi_s_nodelay)
2438 out(" nodelay:%d", s->sctpi_s_nodelay);
2439 if (s->sctpi_s_disable_fragments)
2440 out(" nofrag:%d", s->sctpi_s_disable_fragments);
2441 if (s->sctpi_s_v4mapped)
2442 out(" v4mapped:%d", s->sctpi_s_v4mapped);
2443 if (s->sctpi_s_frag_interleave)
2444 out(" fraginl:%d", s->sctpi_s_frag_interleave);
2445 }
2446
2447 static void tcp_stats_print(struct tcpstat *s)
2448 {
2449 char b1[64];
2450
2451 if (s->has_ts_opt)
2452 out(" ts");
2453 if (s->has_sack_opt)
2454 out(" sack");
2455 if (s->has_ecn_opt)
2456 out(" ecn");
2457 if (s->has_ecnseen_opt)
2458 out(" ecnseen");
2459 if (s->has_fastopen_opt)
2460 out(" fastopen");
2461 if (s->cong_alg[0])
2462 out(" %s", s->cong_alg);
2463 if (s->has_wscale_opt)
2464 out(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
2465 if (s->rto)
2466 out(" rto:%g", s->rto);
2467 if (s->backoff)
2468 out(" backoff:%u", s->backoff);
2469 if (s->rtt)
2470 out(" rtt:%g/%g", s->rtt, s->rttvar);
2471 if (s->ato)
2472 out(" ato:%g", s->ato);
2473
2474 if (s->qack)
2475 out(" qack:%d", s->qack);
2476 if (s->qack & 1)
2477 out(" bidir");
2478
2479 if (s->mss)
2480 out(" mss:%d", s->mss);
2481 if (s->pmtu)
2482 out(" pmtu:%u", s->pmtu);
2483 if (s->rcv_mss)
2484 out(" rcvmss:%d", s->rcv_mss);
2485 if (s->advmss)
2486 out(" advmss:%d", s->advmss);
2487 if (s->cwnd)
2488 out(" cwnd:%u", s->cwnd);
2489 if (s->ssthresh)
2490 out(" ssthresh:%d", s->ssthresh);
2491
2492 if (s->bytes_sent)
2493 out(" bytes_sent:%llu", s->bytes_sent);
2494 if (s->bytes_retrans)
2495 out(" bytes_retrans:%llu", s->bytes_retrans);
2496 if (s->bytes_acked)
2497 out(" bytes_acked:%llu", s->bytes_acked);
2498 if (s->bytes_received)
2499 out(" bytes_received:%llu", s->bytes_received);
2500 if (s->segs_out)
2501 out(" segs_out:%u", s->segs_out);
2502 if (s->segs_in)
2503 out(" segs_in:%u", s->segs_in);
2504 if (s->data_segs_out)
2505 out(" data_segs_out:%u", s->data_segs_out);
2506 if (s->data_segs_in)
2507 out(" data_segs_in:%u", s->data_segs_in);
2508
2509 if (s->dctcp && s->dctcp->enabled) {
2510 struct dctcpstat *dctcp = s->dctcp;
2511
2512 out(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
2513 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
2514 dctcp->ab_tot);
2515 } else if (s->dctcp) {
2516 out(" dctcp:fallback_mode");
2517 }
2518
2519 if (s->bbr_info) {
2520 __u64 bw;
2521
2522 bw = s->bbr_info->bbr_bw_hi;
2523 bw <<= 32;
2524 bw |= s->bbr_info->bbr_bw_lo;
2525
2526 out(" bbr:(bw:%sbps,mrtt:%g",
2527 sprint_bw(b1, bw * 8.0),
2528 (double)s->bbr_info->bbr_min_rtt / 1000.0);
2529 if (s->bbr_info->bbr_pacing_gain)
2530 out(",pacing_gain:%g",
2531 (double)s->bbr_info->bbr_pacing_gain / 256.0);
2532 if (s->bbr_info->bbr_cwnd_gain)
2533 out(",cwnd_gain:%g",
2534 (double)s->bbr_info->bbr_cwnd_gain / 256.0);
2535 out(")");
2536 }
2537
2538 if (s->send_bps)
2539 out(" send %sbps", sprint_bw(b1, s->send_bps));
2540 if (s->lastsnd)
2541 out(" lastsnd:%u", s->lastsnd);
2542 if (s->lastrcv)
2543 out(" lastrcv:%u", s->lastrcv);
2544 if (s->lastack)
2545 out(" lastack:%u", s->lastack);
2546
2547 if (s->pacing_rate) {
2548 out(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
2549 if (s->pacing_rate_max)
2550 out("/%sbps", sprint_bw(b1, s->pacing_rate_max));
2551 }
2552
2553 if (s->delivery_rate)
2554 out(" delivery_rate %sbps", sprint_bw(b1, s->delivery_rate));
2555 if (s->delivered)
2556 out(" delivered:%u", s->delivered);
2557 if (s->delivered_ce)
2558 out(" delivered_ce:%u", s->delivered_ce);
2559 if (s->app_limited)
2560 out(" app_limited");
2561
2562 if (s->busy_time) {
2563 out(" busy:%llums", s->busy_time / 1000);
2564 if (s->rwnd_limited)
2565 out(" rwnd_limited:%llums(%.1f%%)",
2566 s->rwnd_limited / 1000,
2567 100.0 * s->rwnd_limited / s->busy_time);
2568 if (s->sndbuf_limited)
2569 out(" sndbuf_limited:%llums(%.1f%%)",
2570 s->sndbuf_limited / 1000,
2571 100.0 * s->sndbuf_limited / s->busy_time);
2572 }
2573
2574 if (s->unacked)
2575 out(" unacked:%u", s->unacked);
2576 if (s->retrans || s->retrans_total)
2577 out(" retrans:%u/%u", s->retrans, s->retrans_total);
2578 if (s->lost)
2579 out(" lost:%u", s->lost);
2580 if (s->sacked && s->ss.state != SS_LISTEN)
2581 out(" sacked:%u", s->sacked);
2582 if (s->dsack_dups)
2583 out(" dsack_dups:%u", s->dsack_dups);
2584 if (s->fackets)
2585 out(" fackets:%u", s->fackets);
2586 if (s->reordering != 3)
2587 out(" reordering:%d", s->reordering);
2588 if (s->reord_seen)
2589 out(" reord_seen:%d", s->reord_seen);
2590 if (s->rcv_rtt)
2591 out(" rcv_rtt:%g", s->rcv_rtt);
2592 if (s->rcv_space)
2593 out(" rcv_space:%d", s->rcv_space);
2594 if (s->rcv_ssthresh)
2595 out(" rcv_ssthresh:%u", s->rcv_ssthresh);
2596 if (s->not_sent)
2597 out(" notsent:%u", s->not_sent);
2598 if (s->min_rtt)
2599 out(" minrtt:%g", s->min_rtt);
2600 }
2601
2602 static void tcp_timer_print(struct tcpstat *s)
2603 {
2604 static const char * const tmr_name[] = {
2605 "off",
2606 "on",
2607 "keepalive",
2608 "timewait",
2609 "persist",
2610 "unknown"
2611 };
2612
2613 if (s->timer) {
2614 if (s->timer > 4)
2615 s->timer = 5;
2616 out(" timer:(%s,%s,%d)",
2617 tmr_name[s->timer],
2618 print_ms_timer(s->timeout),
2619 s->retrans);
2620 }
2621 }
2622
2623 static void sctp_timer_print(struct tcpstat *s)
2624 {
2625 if (s->timer)
2626 out(" timer:(T3_RTX,%s,%d)",
2627 print_ms_timer(s->timeout), s->retrans);
2628 }
2629
2630 static int tcp_show_line(char *line, const struct filter *f, int family)
2631 {
2632 int rto = 0, ato = 0;
2633 struct tcpstat s = {};
2634 char *loc, *rem, *data;
2635 char opt[256];
2636 int n;
2637 int hz = get_user_hz();
2638
2639 if (proc_inet_split_line(line, &loc, &rem, &data))
2640 return -1;
2641
2642 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2643
2644 if (!(f->states & (1 << state)))
2645 return 0;
2646
2647 proc_parse_inet_addr(loc, rem, family, &s.ss);
2648
2649 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
2650 return 0;
2651
2652 opt[0] = 0;
2653 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %u %d %[^\n]\n",
2654 &s.ss.state, &s.ss.wq, &s.ss.rq,
2655 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
2656 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
2657 &s.ssthresh, opt);
2658
2659 if (n < 17)
2660 opt[0] = 0;
2661
2662 if (n < 12) {
2663 rto = 0;
2664 s.cwnd = 2;
2665 s.ssthresh = -1;
2666 ato = s.qack = 0;
2667 }
2668
2669 s.retrans = s.timer != 1 ? s.probes : s.retrans;
2670 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
2671 s.ato = (double)ato / hz;
2672 s.qack /= 2;
2673 s.rto = (double)rto;
2674 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
2675 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
2676 s.ss.type = IPPROTO_TCP;
2677
2678 inet_stats_print(&s.ss, false);
2679
2680 if (show_options)
2681 tcp_timer_print(&s);
2682
2683 if (show_details) {
2684 sock_details_print(&s.ss);
2685 if (opt[0])
2686 out(" opt:\"%s\"", opt);
2687 }
2688
2689 if (show_tcpinfo)
2690 tcp_stats_print(&s);
2691
2692 return 0;
2693 }
2694
2695 static int generic_record_read(FILE *fp,
2696 int (*worker)(char*, const struct filter *, int),
2697 const struct filter *f, int fam)
2698 {
2699 char line[256];
2700
2701 /* skip header */
2702 if (fgets(line, sizeof(line), fp) == NULL)
2703 goto outerr;
2704
2705 while (fgets(line, sizeof(line), fp) != NULL) {
2706 int n = strlen(line);
2707
2708 if (n == 0 || line[n-1] != '\n') {
2709 errno = -EINVAL;
2710 return -1;
2711 }
2712 line[n-1] = 0;
2713
2714 if (worker(line, f, fam) < 0)
2715 return 0;
2716 }
2717 outerr:
2718
2719 return ferror(fp) ? -1 : 0;
2720 }
2721
2722 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
2723 {
2724 const __u32 *skmeminfo;
2725
2726 if (!tb[attrtype]) {
2727 if (attrtype == INET_DIAG_SKMEMINFO) {
2728 if (!tb[INET_DIAG_MEMINFO])
2729 return;
2730
2731 const struct inet_diag_meminfo *minfo =
2732 RTA_DATA(tb[INET_DIAG_MEMINFO]);
2733
2734 out(" mem:(r%u,w%u,f%u,t%u)",
2735 minfo->idiag_rmem,
2736 minfo->idiag_wmem,
2737 minfo->idiag_fmem,
2738 minfo->idiag_tmem);
2739 }
2740 return;
2741 }
2742
2743 skmeminfo = RTA_DATA(tb[attrtype]);
2744
2745 out(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
2746 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
2747 skmeminfo[SK_MEMINFO_RCVBUF],
2748 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
2749 skmeminfo[SK_MEMINFO_SNDBUF],
2750 skmeminfo[SK_MEMINFO_FWD_ALLOC],
2751 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
2752 skmeminfo[SK_MEMINFO_OPTMEM]);
2753
2754 if (RTA_PAYLOAD(tb[attrtype]) >=
2755 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
2756 out(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
2757
2758 if (RTA_PAYLOAD(tb[attrtype]) >=
2759 (SK_MEMINFO_DROPS + 1) * sizeof(__u32))
2760 out(",d%u", skmeminfo[SK_MEMINFO_DROPS]);
2761
2762 out(")");
2763 }
2764
2765 static void print_md5sig(struct tcp_diag_md5sig *sig)
2766 {
2767 out("%s/%d=",
2768 format_host(sig->tcpm_family,
2769 sig->tcpm_family == AF_INET6 ? 16 : 4,
2770 &sig->tcpm_addr),
2771 sig->tcpm_prefixlen);
2772 print_escape_buf(sig->tcpm_key, sig->tcpm_keylen, " ,");
2773 }
2774
2775 static void tcp_tls_version(struct rtattr *attr)
2776 {
2777 u_int16_t val;
2778
2779 if (!attr)
2780 return;
2781 val = rta_getattr_u16(attr);
2782
2783 switch (val) {
2784 case TLS_1_2_VERSION:
2785 out(" version: 1.2");
2786 break;
2787 case TLS_1_3_VERSION:
2788 out(" version: 1.3");
2789 break;
2790 default:
2791 out(" version: unknown(%hu)", val);
2792 break;
2793 }
2794 }
2795
2796 static void tcp_tls_cipher(struct rtattr *attr)
2797 {
2798 u_int16_t val;
2799
2800 if (!attr)
2801 return;
2802 val = rta_getattr_u16(attr);
2803
2804 switch (val) {
2805 case TLS_CIPHER_AES_GCM_128:
2806 out(" cipher: aes-gcm-128");
2807 break;
2808 case TLS_CIPHER_AES_GCM_256:
2809 out(" cipher: aes-gcm-256");
2810 break;
2811 }
2812 }
2813
2814 static void tcp_tls_conf(const char *name, struct rtattr *attr)
2815 {
2816 u_int16_t val;
2817
2818 if (!attr)
2819 return;
2820 val = rta_getattr_u16(attr);
2821
2822 switch (val) {
2823 case TLS_CONF_BASE:
2824 out(" %s: none", name);
2825 break;
2826 case TLS_CONF_SW:
2827 out(" %s: sw", name);
2828 break;
2829 case TLS_CONF_HW:
2830 out(" %s: hw", name);
2831 break;
2832 case TLS_CONF_HW_RECORD:
2833 out(" %s: hw-record", name);
2834 break;
2835 default:
2836 out(" %s: unknown(%hu)", name, val);
2837 break;
2838 }
2839 }
2840
2841 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
2842
2843 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2844 struct rtattr *tb[])
2845 {
2846 double rtt = 0;
2847 struct tcpstat s = {};
2848
2849 s.ss.state = r->idiag_state;
2850
2851 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2852
2853 if (tb[INET_DIAG_INFO]) {
2854 struct tcp_info *info;
2855 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2856
2857 /* workaround for older kernels with less fields */
2858 if (len < sizeof(*info)) {
2859 info = alloca(sizeof(*info));
2860 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2861 memset((char *)info + len, 0, sizeof(*info) - len);
2862 } else
2863 info = RTA_DATA(tb[INET_DIAG_INFO]);
2864
2865 if (show_options) {
2866 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
2867 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
2868 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
2869 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
2870 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
2871 }
2872
2873 if (tb[INET_DIAG_CONG])
2874 strncpy(s.cong_alg,
2875 rta_getattr_str(tb[INET_DIAG_CONG]),
2876 sizeof(s.cong_alg) - 1);
2877
2878 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
2879 s.has_wscale_opt = true;
2880 s.snd_wscale = info->tcpi_snd_wscale;
2881 s.rcv_wscale = info->tcpi_rcv_wscale;
2882 }
2883
2884 if (info->tcpi_rto && info->tcpi_rto != 3000000)
2885 s.rto = (double)info->tcpi_rto / 1000;
2886
2887 s.backoff = info->tcpi_backoff;
2888 s.rtt = (double)info->tcpi_rtt / 1000;
2889 s.rttvar = (double)info->tcpi_rttvar / 1000;
2890 s.ato = (double)info->tcpi_ato / 1000;
2891 s.mss = info->tcpi_snd_mss;
2892 s.rcv_mss = info->tcpi_rcv_mss;
2893 s.advmss = info->tcpi_advmss;
2894 s.rcv_space = info->tcpi_rcv_space;
2895 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
2896 s.lastsnd = info->tcpi_last_data_sent;
2897 s.lastrcv = info->tcpi_last_data_recv;
2898 s.lastack = info->tcpi_last_ack_recv;
2899 s.unacked = info->tcpi_unacked;
2900 s.retrans = info->tcpi_retrans;
2901 s.retrans_total = info->tcpi_total_retrans;
2902 s.lost = info->tcpi_lost;
2903 s.sacked = info->tcpi_sacked;
2904 s.fackets = info->tcpi_fackets;
2905 s.reordering = info->tcpi_reordering;
2906 s.rcv_ssthresh = info->tcpi_rcv_ssthresh;
2907 s.cwnd = info->tcpi_snd_cwnd;
2908 s.pmtu = info->tcpi_pmtu;
2909
2910 if (info->tcpi_snd_ssthresh < 0xFFFF)
2911 s.ssthresh = info->tcpi_snd_ssthresh;
2912
2913 rtt = (double) info->tcpi_rtt;
2914 if (tb[INET_DIAG_VEGASINFO]) {
2915 const struct tcpvegas_info *vinfo
2916 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
2917
2918 if (vinfo->tcpv_enabled &&
2919 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
2920 rtt = vinfo->tcpv_rtt;
2921 }
2922
2923 if (tb[INET_DIAG_DCTCPINFO]) {
2924 struct dctcpstat *dctcp = malloc(sizeof(struct
2925 dctcpstat));
2926
2927 const struct tcp_dctcp_info *dinfo
2928 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
2929
2930 dctcp->enabled = !!dinfo->dctcp_enabled;
2931 dctcp->ce_state = dinfo->dctcp_ce_state;
2932 dctcp->alpha = dinfo->dctcp_alpha;
2933 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
2934 dctcp->ab_tot = dinfo->dctcp_ab_tot;
2935 s.dctcp = dctcp;
2936 }
2937
2938 if (tb[INET_DIAG_BBRINFO]) {
2939 const void *bbr_info = RTA_DATA(tb[INET_DIAG_BBRINFO]);
2940 int len = min(RTA_PAYLOAD(tb[INET_DIAG_BBRINFO]),
2941 sizeof(*s.bbr_info));
2942
2943 s.bbr_info = calloc(1, sizeof(*s.bbr_info));
2944 if (s.bbr_info && bbr_info)
2945 memcpy(s.bbr_info, bbr_info, len);
2946 }
2947
2948 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
2949 s.send_bps = (double) info->tcpi_snd_cwnd *
2950 (double)info->tcpi_snd_mss * 8000000. / rtt;
2951 }
2952
2953 if (info->tcpi_pacing_rate &&
2954 info->tcpi_pacing_rate != ~0ULL) {
2955 s.pacing_rate = info->tcpi_pacing_rate * 8.;
2956
2957 if (info->tcpi_max_pacing_rate &&
2958 info->tcpi_max_pacing_rate != ~0ULL)
2959 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
2960 }
2961 s.bytes_acked = info->tcpi_bytes_acked;
2962 s.bytes_received = info->tcpi_bytes_received;
2963 s.segs_out = info->tcpi_segs_out;
2964 s.segs_in = info->tcpi_segs_in;
2965 s.data_segs_out = info->tcpi_data_segs_out;
2966 s.data_segs_in = info->tcpi_data_segs_in;
2967 s.not_sent = info->tcpi_notsent_bytes;
2968 if (info->tcpi_min_rtt && info->tcpi_min_rtt != ~0U)
2969 s.min_rtt = (double) info->tcpi_min_rtt / 1000;
2970 s.delivery_rate = info->tcpi_delivery_rate * 8.;
2971 s.app_limited = info->tcpi_delivery_rate_app_limited;
2972 s.busy_time = info->tcpi_busy_time;
2973 s.rwnd_limited = info->tcpi_rwnd_limited;
2974 s.sndbuf_limited = info->tcpi_sndbuf_limited;
2975 s.delivered = info->tcpi_delivered;
2976 s.delivered_ce = info->tcpi_delivered_ce;
2977 s.dsack_dups = info->tcpi_dsack_dups;
2978 s.reord_seen = info->tcpi_reord_seen;
2979 s.bytes_sent = info->tcpi_bytes_sent;
2980 s.bytes_retrans = info->tcpi_bytes_retrans;
2981 tcp_stats_print(&s);
2982 free(s.dctcp);
2983 free(s.bbr_info);
2984 }
2985 if (tb[INET_DIAG_MD5SIG]) {
2986 struct tcp_diag_md5sig *sig = RTA_DATA(tb[INET_DIAG_MD5SIG]);
2987 int len = RTA_PAYLOAD(tb[INET_DIAG_MD5SIG]);
2988
2989 out(" md5keys:");
2990 print_md5sig(sig++);
2991 for (len -= sizeof(*sig); len > 0; len -= sizeof(*sig)) {
2992 out(",");
2993 print_md5sig(sig++);
2994 }
2995 }
2996 if (tb[INET_DIAG_ULP_INFO]) {
2997 struct rtattr *ulpinfo[INET_ULP_INFO_MAX + 1] = { 0 };
2998
2999 parse_rtattr_nested(ulpinfo, INET_ULP_INFO_MAX,
3000 tb[INET_DIAG_ULP_INFO]);
3001
3002 if (ulpinfo[INET_ULP_INFO_NAME])
3003 out(" tcp-ulp-%s",
3004 rta_getattr_str(ulpinfo[INET_ULP_INFO_NAME]));
3005
3006 if (ulpinfo[INET_ULP_INFO_TLS]) {
3007 struct rtattr *tlsinfo[TLS_INFO_MAX + 1] = { 0 };
3008
3009 parse_rtattr_nested(tlsinfo, TLS_INFO_MAX,
3010 ulpinfo[INET_ULP_INFO_TLS]);
3011
3012 tcp_tls_version(tlsinfo[TLS_INFO_VERSION]);
3013 tcp_tls_cipher(tlsinfo[TLS_INFO_CIPHER]);
3014 tcp_tls_conf("rxconf", tlsinfo[TLS_INFO_RXCONF]);
3015 tcp_tls_conf("txconf", tlsinfo[TLS_INFO_TXCONF]);
3016 }
3017 }
3018 }
3019
3020 static const char *format_host_sa(struct sockaddr_storage *sa)
3021 {
3022 union {
3023 struct sockaddr_in sin;
3024 struct sockaddr_in6 sin6;
3025 } *saddr = (void *)sa;
3026
3027 switch (sa->ss_family) {
3028 case AF_INET:
3029 return format_host(AF_INET, 4, &saddr->sin.sin_addr);
3030 case AF_INET6:
3031 return format_host(AF_INET6, 16, &saddr->sin6.sin6_addr);
3032 default:
3033 return "";
3034 }
3035 }
3036
3037 static void sctp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
3038 struct rtattr *tb[])
3039 {
3040 struct sockaddr_storage *sa;
3041 int len;
3042
3043 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
3044
3045 if (tb[INET_DIAG_LOCALS]) {
3046 len = RTA_PAYLOAD(tb[INET_DIAG_LOCALS]);
3047 sa = RTA_DATA(tb[INET_DIAG_LOCALS]);
3048
3049 out(" locals:%s", format_host_sa(sa));
3050 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
3051 out(",%s", format_host_sa(sa));
3052
3053 }
3054 if (tb[INET_DIAG_PEERS]) {
3055 len = RTA_PAYLOAD(tb[INET_DIAG_PEERS]);
3056 sa = RTA_DATA(tb[INET_DIAG_PEERS]);
3057
3058 out(" peers:%s", format_host_sa(sa));
3059 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
3060 out(",%s", format_host_sa(sa));
3061 }
3062 if (tb[INET_DIAG_INFO]) {
3063 struct sctp_info *info;
3064 len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
3065
3066 /* workaround for older kernels with less fields */
3067 if (len < sizeof(*info)) {
3068 info = alloca(sizeof(*info));
3069 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
3070 memset((char *)info + len, 0, sizeof(*info) - len);
3071 } else
3072 info = RTA_DATA(tb[INET_DIAG_INFO]);
3073
3074 sctp_stats_print(info);
3075 }
3076 }
3077
3078 static void parse_diag_msg(struct nlmsghdr *nlh, struct sockstat *s)
3079 {
3080 struct rtattr *tb[INET_DIAG_MAX+1];
3081 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3082
3083 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3084 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3085
3086 s->state = r->idiag_state;
3087 s->local.family = s->remote.family = r->idiag_family;
3088 s->lport = ntohs(r->id.idiag_sport);
3089 s->rport = ntohs(r->id.idiag_dport);
3090 s->wq = r->idiag_wqueue;
3091 s->rq = r->idiag_rqueue;
3092 s->ino = r->idiag_inode;
3093 s->uid = r->idiag_uid;
3094 s->iface = r->id.idiag_if;
3095 s->sk = cookie_sk_get(&r->id.idiag_cookie[0]);
3096
3097 s->mark = 0;
3098 if (tb[INET_DIAG_MARK])
3099 s->mark = rta_getattr_u32(tb[INET_DIAG_MARK]);
3100 if (tb[INET_DIAG_PROTOCOL])
3101 s->raw_prot = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3102 else
3103 s->raw_prot = 0;
3104
3105 if (s->local.family == AF_INET)
3106 s->local.bytelen = s->remote.bytelen = 4;
3107 else
3108 s->local.bytelen = s->remote.bytelen = 16;
3109
3110 memcpy(s->local.data, r->id.idiag_src, s->local.bytelen);
3111 memcpy(s->remote.data, r->id.idiag_dst, s->local.bytelen);
3112 }
3113
3114 static int inet_show_sock(struct nlmsghdr *nlh,
3115 struct sockstat *s)
3116 {
3117 struct rtattr *tb[INET_DIAG_MAX+1];
3118 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3119 unsigned char v6only = 0;
3120
3121 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3122 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3123
3124 if (tb[INET_DIAG_PROTOCOL])
3125 s->type = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3126
3127 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3128 v6only = rta_getattr_u8(tb[INET_DIAG_SKV6ONLY]);
3129
3130 inet_stats_print(s, v6only);
3131
3132 if (show_options) {
3133 struct tcpstat t = {};
3134
3135 t.timer = r->idiag_timer;
3136 t.timeout = r->idiag_expires;
3137 t.retrans = r->idiag_retrans;
3138 if (s->type == IPPROTO_SCTP)
3139 sctp_timer_print(&t);
3140 else
3141 tcp_timer_print(&t);
3142 }
3143
3144 if (show_details) {
3145 sock_details_print(s);
3146 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3147 out(" v6only:%u", v6only);
3148
3149 if (tb[INET_DIAG_SHUTDOWN]) {
3150 unsigned char mask;
3151
3152 mask = rta_getattr_u8(tb[INET_DIAG_SHUTDOWN]);
3153 out(" %c-%c",
3154 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3155 }
3156 }
3157
3158 if (show_tos) {
3159 if (tb[INET_DIAG_TOS])
3160 out(" tos:%#x", rta_getattr_u8(tb[INET_DIAG_TOS]));
3161 if (tb[INET_DIAG_TCLASS])
3162 out(" tclass:%#x", rta_getattr_u8(tb[INET_DIAG_TCLASS]));
3163 if (tb[INET_DIAG_CLASS_ID])
3164 out(" class_id:%#x", rta_getattr_u32(tb[INET_DIAG_CLASS_ID]));
3165 }
3166
3167 if (show_mem || (show_tcpinfo && s->type != IPPROTO_UDP)) {
3168 if (!oneline)
3169 out("\n\t");
3170 if (s->type == IPPROTO_SCTP)
3171 sctp_show_info(nlh, r, tb);
3172 else
3173 tcp_show_info(nlh, r, tb);
3174 }
3175 sctp_ino = s->ino;
3176
3177 return 0;
3178 }
3179
3180 static int tcpdiag_send(int fd, int protocol, struct filter *f)
3181 {
3182 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3183 struct {
3184 struct nlmsghdr nlh;
3185 struct inet_diag_req r;
3186 } req = {
3187 .nlh.nlmsg_len = sizeof(req),
3188 .nlh.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST,
3189 .nlh.nlmsg_seq = MAGIC_SEQ,
3190 .r.idiag_family = AF_INET,
3191 .r.idiag_states = f->states,
3192 };
3193 char *bc = NULL;
3194 int bclen;
3195 struct msghdr msg;
3196 struct rtattr rta;
3197 struct iovec iov[3];
3198 int iovlen = 1;
3199
3200 if (protocol == IPPROTO_UDP)
3201 return -1;
3202
3203 if (protocol == IPPROTO_TCP)
3204 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
3205 else
3206 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
3207 if (show_mem) {
3208 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3209 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3210 }
3211
3212 if (show_tcpinfo) {
3213 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3214 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3215 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3216 }
3217
3218 if (show_tos) {
3219 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3220 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3221 }
3222
3223 iov[0] = (struct iovec){
3224 .iov_base = &req,
3225 .iov_len = sizeof(req)
3226 };
3227 if (f->f) {
3228 bclen = ssfilter_bytecompile(f->f, &bc);
3229 if (bclen) {
3230 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3231 rta.rta_len = RTA_LENGTH(bclen);
3232 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3233 iov[2] = (struct iovec){ bc, bclen };
3234 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3235 iovlen = 3;
3236 }
3237 }
3238
3239 msg = (struct msghdr) {
3240 .msg_name = (void *)&nladdr,
3241 .msg_namelen = sizeof(nladdr),
3242 .msg_iov = iov,
3243 .msg_iovlen = iovlen,
3244 };
3245
3246 if (sendmsg(fd, &msg, 0) < 0) {
3247 close(fd);
3248 return -1;
3249 }
3250
3251 return 0;
3252 }
3253
3254 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
3255 {
3256 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3257 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3258 char *bc = NULL;
3259 int bclen;
3260 struct msghdr msg;
3261 struct rtattr rta;
3262 struct iovec iov[3];
3263 int iovlen = 1;
3264
3265 if (family == PF_UNSPEC)
3266 return tcpdiag_send(fd, protocol, f);
3267
3268 memset(&req.r, 0, sizeof(req.r));
3269 req.r.sdiag_family = family;
3270 req.r.sdiag_protocol = protocol;
3271 req.r.idiag_states = f->states;
3272 if (show_mem) {
3273 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3274 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3275 }
3276
3277 if (show_tcpinfo) {
3278 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3279 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3280 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3281 }
3282
3283 if (show_tos) {
3284 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3285 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3286 }
3287
3288 iov[0] = (struct iovec){
3289 .iov_base = &req,
3290 .iov_len = sizeof(req)
3291 };
3292 if (f->f) {
3293 bclen = ssfilter_bytecompile(f->f, &bc);
3294 if (bclen) {
3295 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3296 rta.rta_len = RTA_LENGTH(bclen);
3297 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3298 iov[2] = (struct iovec){ bc, bclen };
3299 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3300 iovlen = 3;
3301 }
3302 }
3303
3304 msg = (struct msghdr) {
3305 .msg_name = (void *)&nladdr,
3306 .msg_namelen = sizeof(nladdr),
3307 .msg_iov = iov,
3308 .msg_iovlen = iovlen,
3309 };
3310
3311 if (sendmsg(fd, &msg, 0) < 0) {
3312 close(fd);
3313 return -1;
3314 }
3315
3316 return 0;
3317 }
3318
3319 struct inet_diag_arg {
3320 struct filter *f;
3321 int protocol;
3322 struct rtnl_handle *rth;
3323 };
3324
3325 static int kill_inet_sock(struct nlmsghdr *h, void *arg, struct sockstat *s)
3326 {
3327 struct inet_diag_msg *d = NLMSG_DATA(h);
3328 struct inet_diag_arg *diag_arg = arg;
3329 struct rtnl_handle *rth = diag_arg->rth;
3330
3331 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3332
3333 req.nlh.nlmsg_type = SOCK_DESTROY;
3334 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
3335 req.nlh.nlmsg_seq = ++rth->seq;
3336 req.r.sdiag_family = d->idiag_family;
3337 req.r.sdiag_protocol = diag_arg->protocol;
3338 req.r.id = d->id;
3339
3340 if (diag_arg->protocol == IPPROTO_RAW) {
3341 struct inet_diag_req_raw *raw = (void *)&req.r;
3342
3343 BUILD_BUG_ON(sizeof(req.r) != sizeof(*raw));
3344 raw->sdiag_raw_protocol = s->raw_prot;
3345 }
3346
3347 return rtnl_talk(rth, &req.nlh, NULL);
3348 }
3349
3350 static int show_one_inet_sock(struct nlmsghdr *h, void *arg)
3351 {
3352 int err;
3353 struct inet_diag_arg *diag_arg = arg;
3354 struct inet_diag_msg *r = NLMSG_DATA(h);
3355 struct sockstat s = {};
3356
3357 if (!(diag_arg->f->families & FAMILY_MASK(r->idiag_family)))
3358 return 0;
3359
3360 parse_diag_msg(h, &s);
3361 s.type = diag_arg->protocol;
3362
3363 if (diag_arg->f->f && run_ssfilter(diag_arg->f->f, &s) == 0)
3364 return 0;
3365
3366 if (diag_arg->f->kill && kill_inet_sock(h, arg, &s) != 0) {
3367 if (errno == EOPNOTSUPP || errno == ENOENT) {
3368 /* Socket can't be closed, or is already closed. */
3369 return 0;
3370 } else {
3371 perror("SOCK_DESTROY answers");
3372 return -1;
3373 }
3374 }
3375
3376 err = inet_show_sock(h, &s);
3377 if (err < 0)
3378 return err;
3379
3380 return 0;
3381 }
3382
3383 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
3384 {
3385 int err = 0;
3386 struct rtnl_handle rth, rth2;
3387 int family = PF_INET;
3388 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
3389
3390 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3391 return -1;
3392
3393 if (f->kill) {
3394 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
3395 rtnl_close(&rth);
3396 return -1;
3397 }
3398 arg.rth = &rth2;
3399 }
3400
3401 rth.dump = MAGIC_SEQ;
3402 rth.dump_fp = dump_fp;
3403 if (preferred_family == PF_INET6)
3404 family = PF_INET6;
3405
3406 again:
3407 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
3408 goto Exit;
3409
3410 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
3411 if (family != PF_UNSPEC) {
3412 family = PF_UNSPEC;
3413 goto again;
3414 }
3415 goto Exit;
3416 }
3417 if (family == PF_INET && preferred_family != PF_INET) {
3418 family = PF_INET6;
3419 goto again;
3420 }
3421
3422 Exit:
3423 rtnl_close(&rth);
3424 if (arg.rth)
3425 rtnl_close(arg.rth);
3426 return err;
3427 }
3428
3429 static int tcp_show_netlink_file(struct filter *f)
3430 {
3431 FILE *fp;
3432 char buf[16384];
3433 int err = -1;
3434
3435 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
3436 perror("fopen($TCPDIAG_FILE)");
3437 return err;
3438 }
3439
3440 while (1) {
3441 int err2;
3442 size_t status, nitems;
3443 struct nlmsghdr *h = (struct nlmsghdr *)buf;
3444 struct sockstat s = {};
3445
3446 status = fread(buf, 1, sizeof(*h), fp);
3447 if (status != sizeof(*h)) {
3448 if (ferror(fp))
3449 perror("Reading header from $TCPDIAG_FILE");
3450 if (feof(fp))
3451 fprintf(stderr, "Unexpected EOF reading $TCPDIAG_FILE");
3452 break;
3453 }
3454
3455 nitems = NLMSG_ALIGN(h->nlmsg_len - sizeof(*h));
3456 status = fread(h+1, 1, nitems, fp);
3457
3458 if (status != nitems) {
3459 if (ferror(fp))
3460 perror("Reading $TCPDIAG_FILE");
3461 if (feof(fp))
3462 fprintf(stderr, "Unexpected EOF reading $TCPDIAG_FILE");
3463 break;
3464 }
3465
3466 /* The only legal exit point */
3467 if (h->nlmsg_type == NLMSG_DONE) {
3468 err = 0;
3469 break;
3470 }
3471
3472 if (h->nlmsg_type == NLMSG_ERROR) {
3473 struct nlmsgerr *err = (struct nlmsgerr *)NLMSG_DATA(h);
3474
3475 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
3476 fprintf(stderr, "ERROR truncated\n");
3477 } else {
3478 errno = -err->error;
3479 perror("TCPDIAG answered");
3480 }
3481 break;
3482 }
3483
3484 parse_diag_msg(h, &s);
3485 s.type = IPPROTO_TCP;
3486
3487 if (f && f->f && run_ssfilter(f->f, &s) == 0)
3488 continue;
3489
3490 err2 = inet_show_sock(h, &s);
3491 if (err2 < 0) {
3492 err = err2;
3493 break;
3494 }
3495 }
3496
3497 fclose(fp);
3498 return err;
3499 }
3500
3501 static int tcp_show(struct filter *f)
3502 {
3503 FILE *fp = NULL;
3504 char *buf = NULL;
3505 int bufsize = 1024*1024;
3506
3507 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3508 return 0;
3509
3510 dg_proto = TCP_PROTO;
3511
3512 if (getenv("TCPDIAG_FILE"))
3513 return tcp_show_netlink_file(f);
3514
3515 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
3516 && inet_show_netlink(f, NULL, IPPROTO_TCP) == 0)
3517 return 0;
3518
3519 /* Sigh... We have to parse /proc/net/tcp... */
3520 while (bufsize >= 64*1024) {
3521 if ((buf = malloc(bufsize)) != NULL)
3522 break;
3523 bufsize /= 2;
3524 }
3525 if (buf == NULL) {
3526 errno = ENOMEM;
3527 return -1;
3528 }
3529
3530 if (f->families & FAMILY_MASK(AF_INET)) {
3531 if ((fp = net_tcp_open()) == NULL)
3532 goto outerr;
3533
3534 setbuffer(fp, buf, bufsize);
3535 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
3536 goto outerr;
3537 fclose(fp);
3538 }
3539
3540 if ((f->families & FAMILY_MASK(AF_INET6)) &&
3541 (fp = net_tcp6_open()) != NULL) {
3542 setbuffer(fp, buf, bufsize);
3543 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
3544 goto outerr;
3545 fclose(fp);
3546 }
3547
3548 free(buf);
3549 return 0;
3550
3551 outerr:
3552 do {
3553 int saved_errno = errno;
3554
3555 free(buf);
3556 if (fp)
3557 fclose(fp);
3558 errno = saved_errno;
3559 return -1;
3560 } while (0);
3561 }
3562
3563 static int dccp_show(struct filter *f)
3564 {
3565 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3566 return 0;
3567
3568 if (!getenv("PROC_NET_DCCP") && !getenv("PROC_ROOT")
3569 && inet_show_netlink(f, NULL, IPPROTO_DCCP) == 0)
3570 return 0;
3571
3572 return 0;
3573 }
3574
3575 static int sctp_show(struct filter *f)
3576 {
3577 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3578 return 0;
3579
3580 if (!getenv("PROC_NET_SCTP") && !getenv("PROC_ROOT")
3581 && inet_show_netlink(f, NULL, IPPROTO_SCTP) == 0)
3582 return 0;
3583
3584 return 0;
3585 }
3586
3587 static int dgram_show_line(char *line, const struct filter *f, int family)
3588 {
3589 struct sockstat s = {};
3590 char *loc, *rem, *data;
3591 char opt[256];
3592 int n;
3593
3594 if (proc_inet_split_line(line, &loc, &rem, &data))
3595 return -1;
3596
3597 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
3598
3599 if (!(f->states & (1 << state)))
3600 return 0;
3601
3602 proc_parse_inet_addr(loc, rem, family, &s);
3603
3604 if (f->f && run_ssfilter(f->f, &s) == 0)
3605 return 0;
3606
3607 opt[0] = 0;
3608 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
3609 &s.state, &s.wq, &s.rq,
3610 &s.uid, &s.ino,
3611 &s.refcnt, &s.sk, opt);
3612
3613 if (n < 9)
3614 opt[0] = 0;
3615
3616 s.type = dg_proto == UDP_PROTO ? IPPROTO_UDP : 0;
3617 inet_stats_print(&s, false);
3618
3619 if (show_details && opt[0])
3620 out(" opt:\"%s\"", opt);
3621
3622 return 0;
3623 }
3624
3625 static int udp_show(struct filter *f)
3626 {
3627 FILE *fp = NULL;
3628
3629 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3630 return 0;
3631
3632 dg_proto = UDP_PROTO;
3633
3634 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
3635 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
3636 return 0;
3637
3638 if (f->families&FAMILY_MASK(AF_INET)) {
3639 if ((fp = net_udp_open()) == NULL)
3640 goto outerr;
3641 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3642 goto outerr;
3643 fclose(fp);
3644 }
3645
3646 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3647 (fp = net_udp6_open()) != NULL) {
3648 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3649 goto outerr;
3650 fclose(fp);
3651 }
3652 return 0;
3653
3654 outerr:
3655 do {
3656 int saved_errno = errno;
3657
3658 if (fp)
3659 fclose(fp);
3660 errno = saved_errno;
3661 return -1;
3662 } while (0);
3663 }
3664
3665 static int raw_show(struct filter *f)
3666 {
3667 FILE *fp = NULL;
3668
3669 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3670 return 0;
3671
3672 dg_proto = RAW_PROTO;
3673
3674 if (!getenv("PROC_NET_RAW") && !getenv("PROC_ROOT") &&
3675 inet_show_netlink(f, NULL, IPPROTO_RAW) == 0)
3676 return 0;
3677
3678 if (f->families&FAMILY_MASK(AF_INET)) {
3679 if ((fp = net_raw_open()) == NULL)
3680 goto outerr;
3681 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3682 goto outerr;
3683 fclose(fp);
3684 }
3685
3686 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3687 (fp = net_raw6_open()) != NULL) {
3688 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3689 goto outerr;
3690 fclose(fp);
3691 }
3692 return 0;
3693
3694 outerr:
3695 do {
3696 int saved_errno = errno;
3697
3698 if (fp)
3699 fclose(fp);
3700 errno = saved_errno;
3701 return -1;
3702 } while (0);
3703 }
3704
3705 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
3706
3707 static void unix_list_drop_first(struct sockstat **list)
3708 {
3709 struct sockstat *s = *list;
3710
3711 (*list) = (*list)->next;
3712 free(s->name);
3713 free(s);
3714 }
3715
3716 static bool unix_type_skip(struct sockstat *s, struct filter *f)
3717 {
3718 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
3719 return true;
3720 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
3721 return true;
3722 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
3723 return true;
3724 return false;
3725 }
3726
3727 static void unix_stats_print(struct sockstat *s, struct filter *f)
3728 {
3729 char port_name[30] = {};
3730
3731 sock_state_print(s);
3732
3733 sock_addr_print(s->name ?: "*", " ",
3734 int_to_str(s->lport, port_name), NULL);
3735 sock_addr_print(s->peer_name ?: "*", " ",
3736 int_to_str(s->rport, port_name), NULL);
3737
3738 proc_ctx_print(s);
3739 }
3740
3741 static int unix_show_sock(struct nlmsghdr *nlh, void *arg)
3742 {
3743 struct filter *f = (struct filter *)arg;
3744 struct unix_diag_msg *r = NLMSG_DATA(nlh);
3745 struct rtattr *tb[UNIX_DIAG_MAX+1];
3746 char name[128];
3747 struct sockstat stat = { .name = "*", .peer_name = "*" };
3748
3749 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr *)(r+1),
3750 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3751
3752 stat.type = r->udiag_type;
3753 stat.state = r->udiag_state;
3754 stat.ino = stat.lport = r->udiag_ino;
3755 stat.local.family = stat.remote.family = AF_UNIX;
3756
3757 if (unix_type_skip(&stat, f))
3758 return 0;
3759
3760 if (tb[UNIX_DIAG_RQLEN]) {
3761 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
3762
3763 stat.rq = rql->udiag_rqueue;
3764 stat.wq = rql->udiag_wqueue;
3765 }
3766 if (tb[UNIX_DIAG_NAME]) {
3767 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
3768
3769 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
3770 name[len] = '\0';
3771 if (name[0] == '\0') {
3772 int i;
3773 for (i = 0; i < len; i++)
3774 if (name[i] == '\0')
3775 name[i] = '@';
3776 }
3777 stat.name = &name[0];
3778 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
3779 }
3780 if (tb[UNIX_DIAG_PEER])
3781 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
3782
3783 if (f->f && run_ssfilter(f->f, &stat) == 0)
3784 return 0;
3785
3786 unix_stats_print(&stat, f);
3787
3788 if (show_mem)
3789 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
3790 if (show_details) {
3791 if (tb[UNIX_DIAG_SHUTDOWN]) {
3792 unsigned char mask;
3793
3794 mask = rta_getattr_u8(tb[UNIX_DIAG_SHUTDOWN]);
3795 out(" %c-%c",
3796 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3797 }
3798 if (tb[UNIX_DIAG_VFS]) {
3799 struct unix_diag_vfs *uv = RTA_DATA(tb[UNIX_DIAG_VFS]);
3800
3801 out(" ino:%u dev:%u/%u", uv->udiag_vfs_ino, major(uv->udiag_vfs_dev),
3802 minor(uv->udiag_vfs_dev));
3803 }
3804 if (tb[UNIX_DIAG_ICONS]) {
3805 int len = RTA_PAYLOAD(tb[UNIX_DIAG_ICONS]);
3806 __u32 *peers = RTA_DATA(tb[UNIX_DIAG_ICONS]);
3807 int i;
3808
3809 out(" peers:");
3810 for (i = 0; i < len / sizeof(__u32); i++)
3811 out(" %u", peers[i]);
3812 }
3813 }
3814
3815 return 0;
3816 }
3817
3818 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
3819 size_t size, rtnl_filter_t show_one_sock)
3820 {
3821 int ret = -1;
3822 struct rtnl_handle rth;
3823
3824 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3825 return -1;
3826
3827 rth.dump = MAGIC_SEQ;
3828
3829 if (rtnl_send(&rth, req, size) < 0)
3830 goto Exit;
3831
3832 if (rtnl_dump_filter(&rth, show_one_sock, f))
3833 goto Exit;
3834
3835 ret = 0;
3836 Exit:
3837 rtnl_close(&rth);
3838 return ret;
3839 }
3840
3841 static int unix_show_netlink(struct filter *f)
3842 {
3843 DIAG_REQUEST(req, struct unix_diag_req r);
3844
3845 req.r.sdiag_family = AF_UNIX;
3846 req.r.udiag_states = f->states;
3847 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
3848 if (show_mem)
3849 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
3850 if (show_details)
3851 req.r.udiag_show |= UDIAG_SHOW_VFS | UDIAG_SHOW_ICONS;
3852
3853 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
3854 }
3855
3856 static int unix_show(struct filter *f)
3857 {
3858 FILE *fp;
3859 char buf[256];
3860 char name[128];
3861 int newformat = 0;
3862 int cnt;
3863 struct sockstat *list = NULL;
3864 const int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
3865 SS_ESTABLISHED, SS_CLOSING };
3866
3867 if (!filter_af_get(f, AF_UNIX))
3868 return 0;
3869
3870 if (!getenv("PROC_NET_UNIX") && !getenv("PROC_ROOT")
3871 && unix_show_netlink(f) == 0)
3872 return 0;
3873
3874 if ((fp = net_unix_open()) == NULL)
3875 return -1;
3876 if (!fgets(buf, sizeof(buf), fp)) {
3877 fclose(fp);
3878 return -1;
3879 }
3880
3881 if (memcmp(buf, "Peer", 4) == 0)
3882 newformat = 1;
3883 cnt = 0;
3884
3885 while (fgets(buf, sizeof(buf), fp)) {
3886 struct sockstat *u, **insp;
3887 int flags;
3888
3889 if (!(u = calloc(1, sizeof(*u))))
3890 break;
3891
3892 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
3893 &u->rport, &u->rq, &u->wq, &flags, &u->type,
3894 &u->state, &u->ino, name) < 8)
3895 name[0] = 0;
3896
3897 u->lport = u->ino;
3898 u->local.family = u->remote.family = AF_UNIX;
3899
3900 if (flags & (1 << 16)) {
3901 u->state = SS_LISTEN;
3902 } else if (u->state > 0 &&
3903 u->state <= ARRAY_SIZE(unix_state_map)) {
3904 u->state = unix_state_map[u->state-1];
3905 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
3906 u->state = SS_ESTABLISHED;
3907 }
3908 if (unix_type_skip(u, f) ||
3909 !(f->states & (1 << u->state))) {
3910 free(u);
3911 continue;
3912 }
3913
3914 if (!newformat) {
3915 u->rport = 0;
3916 u->rq = 0;
3917 u->wq = 0;
3918 }
3919
3920 if (name[0]) {
3921 u->name = strdup(name);
3922 if (!u->name) {
3923 free(u);
3924 break;
3925 }
3926 }
3927
3928 if (u->rport) {
3929 struct sockstat *p;
3930
3931 for (p = list; p; p = p->next) {
3932 if (u->rport == p->lport)
3933 break;
3934 }
3935 if (!p)
3936 u->peer_name = "?";
3937 else
3938 u->peer_name = p->name ? : "*";
3939 }
3940
3941 if (f->f) {
3942 struct sockstat st = {
3943 .local.family = AF_UNIX,
3944 .remote.family = AF_UNIX,
3945 };
3946
3947 memcpy(st.local.data, &u->name, sizeof(u->name));
3948 /* when parsing the old format rport is set to 0 and
3949 * therefore peer_name remains NULL
3950 */
3951 if (u->peer_name && strcmp(u->peer_name, "*"))
3952 memcpy(st.remote.data, &u->peer_name,
3953 sizeof(u->peer_name));
3954 if (run_ssfilter(f->f, &st) == 0) {
3955 free(u->name);
3956 free(u);
3957 continue;
3958 }
3959 }
3960
3961 insp = &list;
3962 while (*insp) {
3963 if (u->type < (*insp)->type ||
3964 (u->type == (*insp)->type &&
3965 u->ino < (*insp)->ino))
3966 break;
3967 insp = &(*insp)->next;
3968 }
3969 u->next = *insp;
3970 *insp = u;
3971
3972 if (++cnt > MAX_UNIX_REMEMBER) {
3973 while (list) {
3974 unix_stats_print(list, f);
3975 unix_list_drop_first(&list);
3976 }
3977 cnt = 0;
3978 }
3979 }
3980 fclose(fp);
3981 while (list) {
3982 unix_stats_print(list, f);
3983 unix_list_drop_first(&list);
3984 }
3985
3986 return 0;
3987 }
3988
3989 static int packet_stats_print(struct sockstat *s, const struct filter *f)
3990 {
3991 const char *addr, *port;
3992 char ll_name[16];
3993
3994 s->local.family = s->remote.family = AF_PACKET;
3995
3996 if (f->f) {
3997 s->local.data[0] = s->prot;
3998 if (run_ssfilter(f->f, s) == 0)
3999 return 1;
4000 }
4001
4002 sock_state_print(s);
4003
4004 if (s->prot == 3)
4005 addr = "*";
4006 else
4007 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
4008
4009 if (s->iface == 0)
4010 port = "*";
4011 else
4012 port = xll_index_to_name(s->iface);
4013
4014 sock_addr_print(addr, ":", port, NULL);
4015 sock_addr_print("", "*", "", NULL);
4016
4017 proc_ctx_print(s);
4018
4019 if (show_details)
4020 sock_details_print(s);
4021
4022 return 0;
4023 }
4024
4025 static void packet_show_ring(struct packet_diag_ring *ring)
4026 {
4027 out("blk_size:%d", ring->pdr_block_size);
4028 out(",blk_nr:%d", ring->pdr_block_nr);
4029 out(",frm_size:%d", ring->pdr_frame_size);
4030 out(",frm_nr:%d", ring->pdr_frame_nr);
4031 out(",tmo:%d", ring->pdr_retire_tmo);
4032 out(",features:0x%x", ring->pdr_features);
4033 }
4034
4035 static int packet_show_sock(struct nlmsghdr *nlh, void *arg)
4036 {
4037 const struct filter *f = arg;
4038 struct packet_diag_msg *r = NLMSG_DATA(nlh);
4039 struct packet_diag_info *pinfo = NULL;
4040 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
4041 struct rtattr *tb[PACKET_DIAG_MAX+1];
4042 struct sockstat stat = {};
4043 uint32_t fanout = 0;
4044 bool has_fanout = false;
4045
4046 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr *)(r+1),
4047 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4048
4049 /* use /proc/net/packet if all info are not available */
4050 if (!tb[PACKET_DIAG_MEMINFO])
4051 return -1;
4052
4053 stat.type = r->pdiag_type;
4054 stat.prot = r->pdiag_num;
4055 stat.ino = r->pdiag_ino;
4056 stat.state = SS_CLOSE;
4057 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
4058
4059 if (tb[PACKET_DIAG_MEMINFO]) {
4060 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
4061
4062 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4063 }
4064
4065 if (tb[PACKET_DIAG_INFO]) {
4066 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
4067 stat.lport = stat.iface = pinfo->pdi_index;
4068 }
4069
4070 if (tb[PACKET_DIAG_UID])
4071 stat.uid = rta_getattr_u32(tb[PACKET_DIAG_UID]);
4072
4073 if (tb[PACKET_DIAG_RX_RING])
4074 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
4075
4076 if (tb[PACKET_DIAG_TX_RING])
4077 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
4078
4079 if (tb[PACKET_DIAG_FANOUT]) {
4080 has_fanout = true;
4081 fanout = rta_getattr_u32(tb[PACKET_DIAG_FANOUT]);
4082 }
4083
4084 if (packet_stats_print(&stat, f))
4085 return 0;
4086
4087 if (show_details) {
4088 if (pinfo) {
4089 if (oneline)
4090 out(" ver:%d", pinfo->pdi_version);
4091 else
4092 out("\n\tver:%d", pinfo->pdi_version);
4093 out(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
4094 out(" flags( ");
4095 if (pinfo->pdi_flags & PDI_RUNNING)
4096 out("running");
4097 if (pinfo->pdi_flags & PDI_AUXDATA)
4098 out(" auxdata");
4099 if (pinfo->pdi_flags & PDI_ORIGDEV)
4100 out(" origdev");
4101 if (pinfo->pdi_flags & PDI_VNETHDR)
4102 out(" vnethdr");
4103 if (pinfo->pdi_flags & PDI_LOSS)
4104 out(" loss");
4105 if (!pinfo->pdi_flags)
4106 out("0");
4107 out(" )");
4108 }
4109 if (ring_rx) {
4110 if (oneline)
4111 out(" ring_rx(");
4112 else
4113 out("\n\tring_rx(");
4114 packet_show_ring(ring_rx);
4115 out(")");
4116 }
4117 if (ring_tx) {
4118 if (oneline)
4119 out(" ring_tx(");
4120 else
4121 out("\n\tring_tx(");
4122 packet_show_ring(ring_tx);
4123 out(")");
4124 }
4125 if (has_fanout) {
4126 uint16_t type = (fanout >> 16) & 0xffff;
4127
4128 if (oneline)
4129 out(" fanout(");
4130 else
4131 out("\n\tfanout(");
4132 out("id:%d,", fanout & 0xffff);
4133 out("type:");
4134
4135 if (type == 0)
4136 out("hash");
4137 else if (type == 1)
4138 out("lb");
4139 else if (type == 2)
4140 out("cpu");
4141 else if (type == 3)
4142 out("roll");
4143 else if (type == 4)
4144 out("random");
4145 else if (type == 5)
4146 out("qm");
4147 else
4148 out("0x%x", type);
4149
4150 out(")");
4151 }
4152 }
4153
4154 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
4155 struct sock_filter *fil =
4156 RTA_DATA(tb[PACKET_DIAG_FILTER]);
4157 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
4158 sizeof(struct sock_filter);
4159
4160 if (oneline)
4161 out(" bpf filter (%d): ", num);
4162 else
4163 out("\n\tbpf filter (%d): ", num);
4164 while (num) {
4165 out(" 0x%02x %u %u %u,",
4166 fil->code, fil->jt, fil->jf, fil->k);
4167 num--;
4168 fil++;
4169 }
4170 }
4171
4172 if (show_mem)
4173 print_skmeminfo(tb, PACKET_DIAG_MEMINFO);
4174 return 0;
4175 }
4176
4177 static int packet_show_netlink(struct filter *f)
4178 {
4179 DIAG_REQUEST(req, struct packet_diag_req r);
4180
4181 req.r.sdiag_family = AF_PACKET;
4182 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
4183 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
4184
4185 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
4186 }
4187
4188 static int packet_show_line(char *buf, const struct filter *f, int fam)
4189 {
4190 unsigned long long sk;
4191 struct sockstat stat = {};
4192 int type, prot, iface, state, rq, uid, ino;
4193
4194 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
4195 &sk,
4196 &type, &prot, &iface, &state,
4197 &rq, &uid, &ino);
4198
4199 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
4200 return 0;
4201 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
4202 return 0;
4203
4204 stat.type = type;
4205 stat.prot = prot;
4206 stat.lport = stat.iface = iface;
4207 stat.state = state;
4208 stat.rq = rq;
4209 stat.uid = uid;
4210 stat.ino = ino;
4211 stat.state = SS_CLOSE;
4212
4213 if (packet_stats_print(&stat, f))
4214 return 0;
4215
4216 return 0;
4217 }
4218
4219 static int packet_show(struct filter *f)
4220 {
4221 FILE *fp;
4222 int rc = 0;
4223
4224 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
4225 return 0;
4226
4227 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
4228 packet_show_netlink(f) == 0)
4229 return 0;
4230
4231 if ((fp = net_packet_open()) == NULL)
4232 return -1;
4233 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
4234 rc = -1;
4235
4236 fclose(fp);
4237 return rc;
4238 }
4239
4240 static int xdp_stats_print(struct sockstat *s, const struct filter *f)
4241 {
4242 const char *addr, *port;
4243 char q_str[16];
4244
4245 s->local.family = s->remote.family = AF_XDP;
4246
4247 if (f->f) {
4248 if (run_ssfilter(f->f, s) == 0)
4249 return 1;
4250 }
4251
4252 sock_state_print(s);
4253
4254 if (s->iface) {
4255 addr = xll_index_to_name(s->iface);
4256 snprintf(q_str, sizeof(q_str), "q%d", s->lport);
4257 port = q_str;
4258 sock_addr_print(addr, ":", port, NULL);
4259 } else {
4260 sock_addr_print("", "*", "", NULL);
4261 }
4262
4263 sock_addr_print("", "*", "", NULL);
4264
4265 proc_ctx_print(s);
4266
4267 if (show_details)
4268 sock_details_print(s);
4269
4270 return 0;
4271 }
4272
4273 static void xdp_show_ring(const char *name, struct xdp_diag_ring *ring)
4274 {
4275 if (oneline)
4276 out(" %s(", name);
4277 else
4278 out("\n\t%s(", name);
4279 out("entries:%u", ring->entries);
4280 out(")");
4281 }
4282
4283 static void xdp_show_umem(struct xdp_diag_umem *umem, struct xdp_diag_ring *fr,
4284 struct xdp_diag_ring *cr)
4285 {
4286 if (oneline)
4287 out(" tumem(");
4288 else
4289 out("\n\tumem(");
4290 out("id:%u", umem->id);
4291 out(",size:%llu", umem->size);
4292 out(",num_pages:%u", umem->num_pages);
4293 out(",chunk_size:%u", umem->chunk_size);
4294 out(",headroom:%u", umem->headroom);
4295 out(",ifindex:%u", umem->ifindex);
4296 out(",qid:%u", umem->queue_id);
4297 out(",zc:%u", umem->flags & XDP_DU_F_ZEROCOPY);
4298 out(",refs:%u", umem->refs);
4299 out(")");
4300
4301 if (fr)
4302 xdp_show_ring("fr", fr);
4303 if (cr)
4304 xdp_show_ring("cr", cr);
4305 }
4306
4307 static int xdp_show_sock(struct nlmsghdr *nlh, void *arg)
4308 {
4309 struct xdp_diag_ring *rx = NULL, *tx = NULL, *fr = NULL, *cr = NULL;
4310 struct xdp_diag_msg *msg = NLMSG_DATA(nlh);
4311 struct rtattr *tb[XDP_DIAG_MAX + 1];
4312 struct xdp_diag_info *info = NULL;
4313 struct xdp_diag_umem *umem = NULL;
4314 const struct filter *f = arg;
4315 struct sockstat stat = {};
4316
4317 parse_rtattr(tb, XDP_DIAG_MAX, (struct rtattr *)(msg + 1),
4318 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*msg)));
4319
4320 stat.type = msg->xdiag_type;
4321 stat.ino = msg->xdiag_ino;
4322 stat.state = SS_CLOSE;
4323 stat.sk = cookie_sk_get(&msg->xdiag_cookie[0]);
4324
4325 if (tb[XDP_DIAG_INFO]) {
4326 info = RTA_DATA(tb[XDP_DIAG_INFO]);
4327 stat.iface = info->ifindex;
4328 stat.lport = info->queue_id;
4329 }
4330
4331 if (tb[XDP_DIAG_UID])
4332 stat.uid = rta_getattr_u32(tb[XDP_DIAG_UID]);
4333 if (tb[XDP_DIAG_RX_RING])
4334 rx = RTA_DATA(tb[XDP_DIAG_RX_RING]);
4335 if (tb[XDP_DIAG_TX_RING])
4336 tx = RTA_DATA(tb[XDP_DIAG_TX_RING]);
4337 if (tb[XDP_DIAG_UMEM])
4338 umem = RTA_DATA(tb[XDP_DIAG_UMEM]);
4339 if (tb[XDP_DIAG_UMEM_FILL_RING])
4340 fr = RTA_DATA(tb[XDP_DIAG_UMEM_FILL_RING]);
4341 if (tb[XDP_DIAG_UMEM_COMPLETION_RING])
4342 cr = RTA_DATA(tb[XDP_DIAG_UMEM_COMPLETION_RING]);
4343 if (tb[XDP_DIAG_MEMINFO]) {
4344 __u32 *skmeminfo = RTA_DATA(tb[XDP_DIAG_MEMINFO]);
4345
4346 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4347 }
4348
4349 if (xdp_stats_print(&stat, f))
4350 return 0;
4351
4352 if (show_details) {
4353 if (rx)
4354 xdp_show_ring("rx", rx);
4355 if (tx)
4356 xdp_show_ring("tx", tx);
4357 if (umem)
4358 xdp_show_umem(umem, fr, cr);
4359 }
4360
4361 if (show_mem)
4362 print_skmeminfo(tb, XDP_DIAG_MEMINFO); // really?
4363
4364
4365 return 0;
4366 }
4367
4368 static int xdp_show(struct filter *f)
4369 {
4370 DIAG_REQUEST(req, struct xdp_diag_req r);
4371
4372 if (!filter_af_get(f, AF_XDP) || !(f->states & (1 << SS_CLOSE)))
4373 return 0;
4374
4375 req.r.sdiag_family = AF_XDP;
4376 req.r.xdiag_show = XDP_SHOW_INFO | XDP_SHOW_RING_CFG | XDP_SHOW_UMEM |
4377 XDP_SHOW_MEMINFO;
4378
4379 return handle_netlink_request(f, &req.nlh, sizeof(req), xdp_show_sock);
4380 }
4381
4382 static int netlink_show_one(struct filter *f,
4383 int prot, int pid, unsigned int groups,
4384 int state, int dst_pid, unsigned int dst_group,
4385 int rq, int wq,
4386 unsigned long long sk, unsigned long long cb)
4387 {
4388 struct sockstat st = {
4389 .state = SS_CLOSE,
4390 .rq = rq,
4391 .wq = wq,
4392 .local.family = AF_NETLINK,
4393 .remote.family = AF_NETLINK,
4394 };
4395
4396 SPRINT_BUF(prot_buf) = {};
4397 const char *prot_name;
4398 char procname[64] = {};
4399
4400 if (f->f) {
4401 st.rport = -1;
4402 st.lport = pid;
4403 st.local.data[0] = prot;
4404 if (run_ssfilter(f->f, &st) == 0)
4405 return 1;
4406 }
4407
4408 sock_state_print(&st);
4409
4410 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
4411
4412 if (pid == -1) {
4413 procname[0] = '*';
4414 } else if (!numeric) {
4415 int done = 0;
4416
4417 if (!pid) {
4418 done = 1;
4419 strncpy(procname, "kernel", 7);
4420 } else if (pid > 0) {
4421 FILE *fp;
4422
4423 snprintf(procname, sizeof(procname), "%s/%d/stat",
4424 getenv("PROC_ROOT") ? : "/proc", pid);
4425 if ((fp = fopen(procname, "r")) != NULL) {
4426 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
4427 snprintf(procname+strlen(procname),
4428 sizeof(procname)-strlen(procname),
4429 "/%d", pid);
4430 done = 1;
4431 }
4432 fclose(fp);
4433 }
4434 }
4435 if (!done)
4436 int_to_str(pid, procname);
4437 } else {
4438 int_to_str(pid, procname);
4439 }
4440
4441 sock_addr_print(prot_name, ":", procname, NULL);
4442
4443 if (state == NETLINK_CONNECTED) {
4444 char dst_group_buf[30];
4445 char dst_pid_buf[30];
4446
4447 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
4448 int_to_str(dst_pid, dst_pid_buf), NULL);
4449 } else {
4450 sock_addr_print("", "*", "", NULL);
4451 }
4452
4453 char *pid_context = NULL;
4454
4455 if (show_proc_ctx) {
4456 /* The pid value will either be:
4457 * 0 if destination kernel - show kernel initial context.
4458 * A valid process pid - use getpidcon.
4459 * A unique value allocated by the kernel or netlink user
4460 * to the process - show context as "not available".
4461 */
4462 if (!pid)
4463 security_get_initial_context("kernel", &pid_context);
4464 else if (pid > 0)
4465 getpidcon(pid, &pid_context);
4466
4467 out(" proc_ctx=%s", pid_context ? : "unavailable");
4468 free(pid_context);
4469 }
4470
4471 if (show_details) {
4472 out(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
4473 }
4474
4475 return 0;
4476 }
4477
4478 static int netlink_show_sock(struct nlmsghdr *nlh, void *arg)
4479 {
4480 struct filter *f = (struct filter *)arg;
4481 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
4482 struct rtattr *tb[NETLINK_DIAG_MAX+1];
4483 int rq = 0, wq = 0;
4484 unsigned long groups = 0;
4485
4486 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr *)(r+1),
4487 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4488
4489 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
4490 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
4491
4492 if (tb[NETLINK_DIAG_MEMINFO]) {
4493 const __u32 *skmeminfo;
4494
4495 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
4496
4497 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4498 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
4499 }
4500
4501 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
4502 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
4503 rq, wq, 0, 0)) {
4504 return 0;
4505 }
4506
4507 if (show_mem) {
4508 out("\t");
4509 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
4510 }
4511
4512 return 0;
4513 }
4514
4515 static int netlink_show_netlink(struct filter *f)
4516 {
4517 DIAG_REQUEST(req, struct netlink_diag_req r);
4518
4519 req.r.sdiag_family = AF_NETLINK;
4520 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
4521 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
4522
4523 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
4524 }
4525
4526 static int netlink_show(struct filter *f)
4527 {
4528 FILE *fp;
4529 char buf[256];
4530 int prot, pid;
4531 unsigned int groups;
4532 int rq, wq, rc;
4533 unsigned long long sk, cb;
4534
4535 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
4536 return 0;
4537
4538 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
4539 netlink_show_netlink(f) == 0)
4540 return 0;
4541
4542 if ((fp = net_netlink_open()) == NULL)
4543 return -1;
4544 if (!fgets(buf, sizeof(buf), fp)) {
4545 fclose(fp);
4546 return -1;
4547 }
4548
4549 while (fgets(buf, sizeof(buf), fp)) {
4550 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
4551 &sk,
4552 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
4553
4554 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
4555 }
4556
4557 fclose(fp);
4558 return 0;
4559 }
4560
4561 static bool vsock_type_skip(struct sockstat *s, struct filter *f)
4562 {
4563 if (s->type == SOCK_STREAM && !(f->dbs & (1 << VSOCK_ST_DB)))
4564 return true;
4565 if (s->type == SOCK_DGRAM && !(f->dbs & (1 << VSOCK_DG_DB)))
4566 return true;
4567 return false;
4568 }
4569
4570 static void vsock_addr_print(inet_prefix *a, __u32 port)
4571 {
4572 char cid_str[sizeof("4294967295")];
4573 char port_str[sizeof("4294967295")];
4574 __u32 cid;
4575
4576 memcpy(&cid, a->data, sizeof(cid));
4577
4578 if (cid == ~(__u32)0)
4579 snprintf(cid_str, sizeof(cid_str), "*");
4580 else
4581 snprintf(cid_str, sizeof(cid_str), "%u", cid);
4582
4583 if (port == ~(__u32)0)
4584 snprintf(port_str, sizeof(port_str), "*");
4585 else
4586 snprintf(port_str, sizeof(port_str), "%u", port);
4587
4588 sock_addr_print(cid_str, ":", port_str, NULL);
4589 }
4590
4591 static void vsock_stats_print(struct sockstat *s, struct filter *f)
4592 {
4593 sock_state_print(s);
4594
4595 vsock_addr_print(&s->local, s->lport);
4596 vsock_addr_print(&s->remote, s->rport);
4597
4598 proc_ctx_print(s);
4599 }
4600
4601 static int vsock_show_sock(struct nlmsghdr *nlh, void *arg)
4602 {
4603 struct filter *f = (struct filter *)arg;
4604 struct vsock_diag_msg *r = NLMSG_DATA(nlh);
4605 struct sockstat stat = {
4606 .type = r->vdiag_type,
4607 .lport = r->vdiag_src_port,
4608 .rport = r->vdiag_dst_port,
4609 .state = r->vdiag_state,
4610 .ino = r->vdiag_ino,
4611 };
4612
4613 vsock_set_inet_prefix(&stat.local, r->vdiag_src_cid);
4614 vsock_set_inet_prefix(&stat.remote, r->vdiag_dst_cid);
4615
4616 if (vsock_type_skip(&stat, f))
4617 return 0;
4618
4619 if (f->f && run_ssfilter(f->f, &stat) == 0)
4620 return 0;
4621
4622 vsock_stats_print(&stat, f);
4623
4624 return 0;
4625 }
4626
4627 static int vsock_show(struct filter *f)
4628 {
4629 DIAG_REQUEST(req, struct vsock_diag_req r);
4630
4631 if (!filter_af_get(f, AF_VSOCK))
4632 return 0;
4633
4634 req.r.sdiag_family = AF_VSOCK;
4635 req.r.vdiag_states = f->states;
4636
4637 return handle_netlink_request(f, &req.nlh, sizeof(req), vsock_show_sock);
4638 }
4639
4640 static void tipc_sock_addr_print(struct rtattr *net_addr, struct rtattr *id)
4641 {
4642 uint32_t node = rta_getattr_u32(net_addr);
4643 uint32_t identity = rta_getattr_u32(id);
4644
4645 SPRINT_BUF(addr) = {};
4646 SPRINT_BUF(port) = {};
4647
4648 sprintf(addr, "%u", node);
4649 sprintf(port, "%u", identity);
4650 sock_addr_print(addr, ":", port, NULL);
4651
4652 }
4653
4654 static int tipc_show_sock(struct nlmsghdr *nlh, void *arg)
4655 {
4656 struct rtattr *stat[TIPC_NLA_SOCK_STAT_MAX + 1] = {};
4657 struct rtattr *attrs[TIPC_NLA_SOCK_MAX + 1] = {};
4658 struct rtattr *con[TIPC_NLA_CON_MAX + 1] = {};
4659 struct rtattr *info[TIPC_NLA_MAX + 1] = {};
4660 struct rtattr *msg_ref;
4661 struct sockstat ss = {};
4662
4663 parse_rtattr(info, TIPC_NLA_MAX, NLMSG_DATA(nlh),
4664 NLMSG_PAYLOAD(nlh, 0));
4665
4666 if (!info[TIPC_NLA_SOCK])
4667 return 0;
4668
4669 msg_ref = info[TIPC_NLA_SOCK];
4670 parse_rtattr(attrs, TIPC_NLA_SOCK_MAX, RTA_DATA(msg_ref),
4671 RTA_PAYLOAD(msg_ref));
4672
4673 msg_ref = attrs[TIPC_NLA_SOCK_STAT];
4674 parse_rtattr(stat, TIPC_NLA_SOCK_STAT_MAX,
4675 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4676
4677
4678 ss.local.family = AF_TIPC;
4679 ss.type = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TYPE]);
4680 ss.state = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TIPC_STATE]);
4681 ss.uid = rta_getattr_u32(attrs[TIPC_NLA_SOCK_UID]);
4682 ss.ino = rta_getattr_u32(attrs[TIPC_NLA_SOCK_INO]);
4683 ss.rq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_RCVQ]);
4684 ss.wq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_SENDQ]);
4685 ss.sk = rta_getattr_u64(attrs[TIPC_NLA_SOCK_COOKIE]);
4686
4687 sock_state_print (&ss);
4688
4689 tipc_sock_addr_print(attrs[TIPC_NLA_SOCK_ADDR],
4690 attrs[TIPC_NLA_SOCK_REF]);
4691
4692 msg_ref = attrs[TIPC_NLA_SOCK_CON];
4693 if (msg_ref) {
4694 parse_rtattr(con, TIPC_NLA_CON_MAX,
4695 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4696
4697 tipc_sock_addr_print(con[TIPC_NLA_CON_NODE],
4698 con[TIPC_NLA_CON_SOCK]);
4699 } else
4700 sock_addr_print("", "-", "", NULL);
4701
4702 if (show_details)
4703 sock_details_print(&ss);
4704
4705 proc_ctx_print(&ss);
4706
4707 if (show_tipcinfo) {
4708 if (oneline)
4709 out(" type:%s", stype_nameg[ss.type]);
4710 else
4711 out("\n type:%s", stype_nameg[ss.type]);
4712 out(" cong:%s ",
4713 stat[TIPC_NLA_SOCK_STAT_LINK_CONG] ? "link" :
4714 stat[TIPC_NLA_SOCK_STAT_CONN_CONG] ? "conn" : "none");
4715 out(" drop:%d ",
4716 rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_DROP]));
4717
4718 if (attrs[TIPC_NLA_SOCK_HAS_PUBL])
4719 out(" publ");
4720
4721 if (con[TIPC_NLA_CON_FLAG])
4722 out(" via {%u,%u} ",
4723 rta_getattr_u32(con[TIPC_NLA_CON_TYPE]),
4724 rta_getattr_u32(con[TIPC_NLA_CON_INST]));
4725 }
4726
4727 return 0;
4728 }
4729
4730 static int tipc_show(struct filter *f)
4731 {
4732 DIAG_REQUEST(req, struct tipc_sock_diag_req r);
4733
4734 memset(&req.r, 0, sizeof(req.r));
4735 req.r.sdiag_family = AF_TIPC;
4736 req.r.tidiag_states = f->states;
4737
4738 return handle_netlink_request(f, &req.nlh, sizeof(req), tipc_show_sock);
4739 }
4740
4741 struct sock_diag_msg {
4742 __u8 sdiag_family;
4743 };
4744
4745 static int generic_show_sock(struct nlmsghdr *nlh, void *arg)
4746 {
4747 struct sock_diag_msg *r = NLMSG_DATA(nlh);
4748 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
4749 int ret;
4750
4751 switch (r->sdiag_family) {
4752 case AF_INET:
4753 case AF_INET6:
4754 inet_arg.rth = inet_arg.f->rth_for_killing;
4755 ret = show_one_inet_sock(nlh, &inet_arg);
4756 break;
4757 case AF_UNIX:
4758 ret = unix_show_sock(nlh, arg);
4759 break;
4760 case AF_PACKET:
4761 ret = packet_show_sock(nlh, arg);
4762 break;
4763 case AF_NETLINK:
4764 ret = netlink_show_sock(nlh, arg);
4765 break;
4766 case AF_VSOCK:
4767 ret = vsock_show_sock(nlh, arg);
4768 break;
4769 case AF_XDP:
4770 ret = xdp_show_sock(nlh, arg);
4771 break;
4772 default:
4773 ret = -1;
4774 }
4775
4776 render();
4777
4778 return ret;
4779 }
4780
4781 static int handle_follow_request(struct filter *f)
4782 {
4783 int ret = 0;
4784 int groups = 0;
4785 struct rtnl_handle rth, rth2;
4786
4787 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << TCP_DB))
4788 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
4789 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << UDP_DB))
4790 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
4791 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << TCP_DB))
4792 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
4793 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << UDP_DB))
4794 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
4795
4796 if (groups == 0)
4797 return -1;
4798
4799 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
4800 return -1;
4801
4802 rth.dump = 0;
4803 rth.local.nl_pid = 0;
4804
4805 if (f->kill) {
4806 if (rtnl_open_byproto(&rth2, groups, NETLINK_SOCK_DIAG)) {
4807 rtnl_close(&rth);
4808 return -1;
4809 }
4810 f->rth_for_killing = &rth2;
4811 }
4812
4813 if (rtnl_dump_filter(&rth, generic_show_sock, f))
4814 ret = -1;
4815
4816 rtnl_close(&rth);
4817 if (f->rth_for_killing)
4818 rtnl_close(f->rth_for_killing);
4819 return ret;
4820 }
4821
4822 static int get_snmp_int(char *proto, char *key, int *result)
4823 {
4824 char buf[1024];
4825 FILE *fp;
4826 int protolen = strlen(proto);
4827 int keylen = strlen(key);
4828
4829 *result = 0;
4830
4831 if ((fp = net_snmp_open()) == NULL)
4832 return -1;
4833
4834 while (fgets(buf, sizeof(buf), fp) != NULL) {
4835 char *p = buf;
4836 int pos = 0;
4837
4838 if (memcmp(buf, proto, protolen))
4839 continue;
4840 while ((p = strchr(p, ' ')) != NULL) {
4841 pos++;
4842 p++;
4843 if (memcmp(p, key, keylen) == 0 &&
4844 (p[keylen] == ' ' || p[keylen] == '\n'))
4845 break;
4846 }
4847 if (fgets(buf, sizeof(buf), fp) == NULL)
4848 break;
4849 if (memcmp(buf, proto, protolen))
4850 break;
4851 p = buf;
4852 while ((p = strchr(p, ' ')) != NULL) {
4853 p++;
4854 if (--pos == 0) {
4855 sscanf(p, "%d", result);
4856 fclose(fp);
4857 return 0;
4858 }
4859 }
4860 }
4861
4862 fclose(fp);
4863 errno = ESRCH;
4864 return -1;
4865 }
4866
4867
4868 /* Get stats from sockstat */
4869
4870 struct ssummary {
4871 int socks;
4872 int tcp_mem;
4873 int tcp_total;
4874 int tcp_orphans;
4875 int tcp_tws;
4876 int tcp4_hashed;
4877 int udp4;
4878 int raw4;
4879 int frag4;
4880 int frag4_mem;
4881 int tcp6_hashed;
4882 int udp6;
4883 int raw6;
4884 int frag6;
4885 int frag6_mem;
4886 };
4887
4888 static void get_sockstat_line(char *line, struct ssummary *s)
4889 {
4890 char id[256], rem[256];
4891
4892 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
4893 return;
4894
4895 if (strcmp(id, "sockets:") == 0)
4896 sscanf(rem, "%*s%d", &s->socks);
4897 else if (strcmp(id, "UDP:") == 0)
4898 sscanf(rem, "%*s%d", &s->udp4);
4899 else if (strcmp(id, "UDP6:") == 0)
4900 sscanf(rem, "%*s%d", &s->udp6);
4901 else if (strcmp(id, "RAW:") == 0)
4902 sscanf(rem, "%*s%d", &s->raw4);
4903 else if (strcmp(id, "RAW6:") == 0)
4904 sscanf(rem, "%*s%d", &s->raw6);
4905 else if (strcmp(id, "TCP6:") == 0)
4906 sscanf(rem, "%*s%d", &s->tcp6_hashed);
4907 else if (strcmp(id, "FRAG:") == 0)
4908 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
4909 else if (strcmp(id, "FRAG6:") == 0)
4910 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
4911 else if (strcmp(id, "TCP:") == 0)
4912 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
4913 &s->tcp4_hashed,
4914 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
4915 }
4916
4917 static int get_sockstat(struct ssummary *s)
4918 {
4919 char buf[256];
4920 FILE *fp;
4921
4922 memset(s, 0, sizeof(*s));
4923
4924 if ((fp = net_sockstat_open()) == NULL)
4925 return -1;
4926 while (fgets(buf, sizeof(buf), fp) != NULL)
4927 get_sockstat_line(buf, s);
4928 fclose(fp);
4929
4930 if ((fp = net_sockstat6_open()) == NULL)
4931 return 0;
4932 while (fgets(buf, sizeof(buf), fp) != NULL)
4933 get_sockstat_line(buf, s);
4934 fclose(fp);
4935
4936 return 0;
4937 }
4938
4939 static int print_summary(void)
4940 {
4941 struct ssummary s;
4942 int tcp_estab;
4943
4944 if (get_sockstat(&s) < 0)
4945 perror("ss: get_sockstat");
4946 if (get_snmp_int("Tcp:", "CurrEstab", &tcp_estab) < 0)
4947 perror("ss: get_snmpstat");
4948
4949 printf("Total: %d\n", s.socks);
4950
4951 printf("TCP: %d (estab %d, closed %d, orphaned %d, timewait %d)\n",
4952 s.tcp_total + s.tcp_tws, tcp_estab,
4953 s.tcp_total - (s.tcp4_hashed + s.tcp6_hashed - s.tcp_tws),
4954 s.tcp_orphans, s.tcp_tws);
4955
4956 printf("\n");
4957 printf("Transport Total IP IPv6\n");
4958 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
4959 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
4960 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
4961 printf("INET %-9d %-9d %-9d\n",
4962 s.raw4+s.udp4+s.tcp4_hashed+
4963 s.raw6+s.udp6+s.tcp6_hashed,
4964 s.raw4+s.udp4+s.tcp4_hashed,
4965 s.raw6+s.udp6+s.tcp6_hashed);
4966 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
4967
4968 printf("\n");
4969
4970 return 0;
4971 }
4972
4973 static void _usage(FILE *dest)
4974 {
4975 fprintf(dest,
4976 "Usage: ss [ OPTIONS ]\n"
4977 " ss [ OPTIONS ] [ FILTER ]\n"
4978 " -h, --help this message\n"
4979 " -V, --version output version information\n"
4980 " -n, --numeric don't resolve service names\n"
4981 " -r, --resolve resolve host names\n"
4982 " -a, --all display all sockets\n"
4983 " -l, --listening display listening sockets\n"
4984 " -o, --options show timer information\n"
4985 " -e, --extended show detailed socket information\n"
4986 " -m, --memory show socket memory usage\n"
4987 " -p, --processes show process using socket\n"
4988 " -i, --info show internal TCP information\n"
4989 " --tipcinfo show internal tipc socket information\n"
4990 " -s, --summary show socket usage summary\n"
4991 " --tos show tos and priority information\n"
4992 " -b, --bpf show bpf filter socket information\n"
4993 " -E, --events continually display sockets as they are destroyed\n"
4994 " -Z, --context display process SELinux security contexts\n"
4995 " -z, --contexts display process and socket SELinux security contexts\n"
4996 " -N, --net switch to the specified network namespace name\n"
4997 "\n"
4998 " -4, --ipv4 display only IP version 4 sockets\n"
4999 " -6, --ipv6 display only IP version 6 sockets\n"
5000 " -0, --packet display PACKET sockets\n"
5001 " -t, --tcp display only TCP sockets\n"
5002 " -S, --sctp display only SCTP sockets\n"
5003 " -u, --udp display only UDP sockets\n"
5004 " -d, --dccp display only DCCP sockets\n"
5005 " -w, --raw display only RAW sockets\n"
5006 " -x, --unix display only Unix domain sockets\n"
5007 " --tipc display only TIPC sockets\n"
5008 " --vsock display only vsock sockets\n"
5009 " -f, --family=FAMILY display sockets of type FAMILY\n"
5010 " FAMILY := {inet|inet6|link|unix|netlink|vsock|tipc|xdp|help}\n"
5011 "\n"
5012 " -K, --kill forcibly close sockets, display what was closed\n"
5013 " -H, --no-header Suppress header line\n"
5014 " -O, --oneline socket's data printed on a single line\n"
5015 "\n"
5016 " -A, --query=QUERY, --socket=QUERY\n"
5017 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink|vsock_stream|vsock_dgram|tipc}[,QUERY]\n"
5018 "\n"
5019 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
5020 " -F, --filter=FILE read filter information from FILE\n"
5021 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
5022 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
5023 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listening|closing}\n"
5024 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
5025 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
5026 " bucket := {syn-recv|time-wait}\n"
5027 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listening|closing}\n"
5028 );
5029 }
5030
5031 static void help(void) __attribute__((noreturn));
5032 static void help(void)
5033 {
5034 _usage(stdout);
5035 exit(0);
5036 }
5037
5038 static void usage(void) __attribute__((noreturn));
5039 static void usage(void)
5040 {
5041 _usage(stderr);
5042 exit(-1);
5043 }
5044
5045
5046 static int scan_state(const char *state)
5047 {
5048 static const char * const sstate_namel[] = {
5049 "UNKNOWN",
5050 [SS_ESTABLISHED] = "established",
5051 [SS_SYN_SENT] = "syn-sent",
5052 [SS_SYN_RECV] = "syn-recv",
5053 [SS_FIN_WAIT1] = "fin-wait-1",
5054 [SS_FIN_WAIT2] = "fin-wait-2",
5055 [SS_TIME_WAIT] = "time-wait",
5056 [SS_CLOSE] = "unconnected",
5057 [SS_CLOSE_WAIT] = "close-wait",
5058 [SS_LAST_ACK] = "last-ack",
5059 [SS_LISTEN] = "listening",
5060 [SS_CLOSING] = "closing",
5061 };
5062 int i;
5063
5064 if (strcasecmp(state, "close") == 0 ||
5065 strcasecmp(state, "closed") == 0)
5066 return (1<<SS_CLOSE);
5067 if (strcasecmp(state, "syn-rcv") == 0)
5068 return (1<<SS_SYN_RECV);
5069 if (strcasecmp(state, "established") == 0)
5070 return (1<<SS_ESTABLISHED);
5071 if (strcasecmp(state, "all") == 0)
5072 return SS_ALL;
5073 if (strcasecmp(state, "connected") == 0)
5074 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
5075 if (strcasecmp(state, "synchronized") == 0)
5076 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
5077 if (strcasecmp(state, "bucket") == 0)
5078 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
5079 if (strcasecmp(state, "big") == 0)
5080 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
5081 for (i = 0; i < SS_MAX; i++) {
5082 if (strcasecmp(state, sstate_namel[i]) == 0)
5083 return (1<<i);
5084 }
5085
5086 fprintf(stderr, "ss: wrong state name: %s\n", state);
5087 exit(-1);
5088 }
5089
5090 /* Values 'v' and 'V' are already used so a non-character is used */
5091 #define OPT_VSOCK 256
5092
5093 /* Values of 't' are already used so a non-character is used */
5094 #define OPT_TIPCSOCK 257
5095 #define OPT_TIPCINFO 258
5096
5097 #define OPT_TOS 259
5098
5099 /* Values of 'x' are already used so a non-character is used */
5100 #define OPT_XDPSOCK 260
5101
5102 static const struct option long_opts[] = {
5103 { "numeric", 0, 0, 'n' },
5104 { "resolve", 0, 0, 'r' },
5105 { "options", 0, 0, 'o' },
5106 { "extended", 0, 0, 'e' },
5107 { "memory", 0, 0, 'm' },
5108 { "info", 0, 0, 'i' },
5109 { "processes", 0, 0, 'p' },
5110 { "bpf", 0, 0, 'b' },
5111 { "events", 0, 0, 'E' },
5112 { "dccp", 0, 0, 'd' },
5113 { "tcp", 0, 0, 't' },
5114 { "sctp", 0, 0, 'S' },
5115 { "udp", 0, 0, 'u' },
5116 { "raw", 0, 0, 'w' },
5117 { "unix", 0, 0, 'x' },
5118 { "tipc", 0, 0, OPT_TIPCSOCK},
5119 { "vsock", 0, 0, OPT_VSOCK },
5120 { "all", 0, 0, 'a' },
5121 { "listening", 0, 0, 'l' },
5122 { "ipv4", 0, 0, '4' },
5123 { "ipv6", 0, 0, '6' },
5124 { "packet", 0, 0, '0' },
5125 { "family", 1, 0, 'f' },
5126 { "socket", 1, 0, 'A' },
5127 { "query", 1, 0, 'A' },
5128 { "summary", 0, 0, 's' },
5129 { "diag", 1, 0, 'D' },
5130 { "filter", 1, 0, 'F' },
5131 { "version", 0, 0, 'V' },
5132 { "help", 0, 0, 'h' },
5133 { "context", 0, 0, 'Z' },
5134 { "contexts", 0, 0, 'z' },
5135 { "net", 1, 0, 'N' },
5136 { "tipcinfo", 0, 0, OPT_TIPCINFO},
5137 { "tos", 0, 0, OPT_TOS },
5138 { "kill", 0, 0, 'K' },
5139 { "no-header", 0, 0, 'H' },
5140 { "xdp", 0, 0, OPT_XDPSOCK},
5141 { "oneline", 0, 0, 'O' },
5142 { 0 }
5143
5144 };
5145
5146 int main(int argc, char *argv[])
5147 {
5148 int saw_states = 0;
5149 int saw_query = 0;
5150 int do_summary = 0;
5151 const char *dump_tcpdiag = NULL;
5152 FILE *filter_fp = NULL;
5153 int ch;
5154 int state_filter = 0;
5155
5156 while ((ch = getopt_long(argc, argv,
5157 "dhaletuwxnro460spbEf:miA:D:F:vVzZN:KHSO",
5158 long_opts, NULL)) != EOF) {
5159 switch (ch) {
5160 case 'n':
5161 numeric = 1;
5162 break;
5163 case 'r':
5164 resolve_hosts = 1;
5165 break;
5166 case 'o':
5167 show_options = 1;
5168 break;
5169 case 'e':
5170 show_options = 1;
5171 show_details++;
5172 break;
5173 case 'm':
5174 show_mem = 1;
5175 break;
5176 case 'i':
5177 show_tcpinfo = 1;
5178 break;
5179 case 'p':
5180 show_users++;
5181 user_ent_hash_build();
5182 break;
5183 case 'b':
5184 show_options = 1;
5185 show_bpf++;
5186 break;
5187 case 'E':
5188 follow_events = 1;
5189 break;
5190 case 'd':
5191 filter_db_set(&current_filter, DCCP_DB, true);
5192 break;
5193 case 't':
5194 filter_db_set(&current_filter, TCP_DB, true);
5195 break;
5196 case 'S':
5197 filter_db_set(&current_filter, SCTP_DB, true);
5198 break;
5199 case 'u':
5200 filter_db_set(&current_filter, UDP_DB, true);
5201 break;
5202 case 'w':
5203 filter_db_set(&current_filter, RAW_DB, true);
5204 break;
5205 case 'x':
5206 filter_af_set(&current_filter, AF_UNIX);
5207 break;
5208 case OPT_VSOCK:
5209 filter_af_set(&current_filter, AF_VSOCK);
5210 break;
5211 case OPT_TIPCSOCK:
5212 filter_af_set(&current_filter, AF_TIPC);
5213 break;
5214 case 'a':
5215 state_filter = SS_ALL;
5216 break;
5217 case 'l':
5218 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
5219 break;
5220 case '4':
5221 filter_af_set(&current_filter, AF_INET);
5222 break;
5223 case '6':
5224 filter_af_set(&current_filter, AF_INET6);
5225 break;
5226 case '0':
5227 filter_af_set(&current_filter, AF_PACKET);
5228 break;
5229 case OPT_XDPSOCK:
5230 filter_af_set(&current_filter, AF_XDP);
5231 break;
5232 case 'f':
5233 if (strcmp(optarg, "inet") == 0)
5234 filter_af_set(&current_filter, AF_INET);
5235 else if (strcmp(optarg, "inet6") == 0)
5236 filter_af_set(&current_filter, AF_INET6);
5237 else if (strcmp(optarg, "link") == 0)
5238 filter_af_set(&current_filter, AF_PACKET);
5239 else if (strcmp(optarg, "unix") == 0)
5240 filter_af_set(&current_filter, AF_UNIX);
5241 else if (strcmp(optarg, "netlink") == 0)
5242 filter_af_set(&current_filter, AF_NETLINK);
5243 else if (strcmp(optarg, "tipc") == 0)
5244 filter_af_set(&current_filter, AF_TIPC);
5245 else if (strcmp(optarg, "vsock") == 0)
5246 filter_af_set(&current_filter, AF_VSOCK);
5247 else if (strcmp(optarg, "xdp") == 0)
5248 filter_af_set(&current_filter, AF_XDP);
5249 else if (strcmp(optarg, "help") == 0)
5250 help();
5251 else {
5252 fprintf(stderr, "ss: \"%s\" is invalid family\n",
5253 optarg);
5254 usage();
5255 }
5256 break;
5257 case 'A':
5258 {
5259 char *p, *p1;
5260
5261 if (!saw_query) {
5262 current_filter.dbs = 0;
5263 state_filter = state_filter ?
5264 state_filter : SS_CONN;
5265 saw_query = 1;
5266 do_default = 0;
5267 }
5268 p = p1 = optarg;
5269 do {
5270 if ((p1 = strchr(p, ',')) != NULL)
5271 *p1 = 0;
5272 if (filter_db_parse(&current_filter, p)) {
5273 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
5274 usage();
5275 }
5276 p = p1 + 1;
5277 } while (p1);
5278 break;
5279 }
5280 case 's':
5281 do_summary = 1;
5282 break;
5283 case 'D':
5284 dump_tcpdiag = optarg;
5285 break;
5286 case 'F':
5287 if (filter_fp) {
5288 fprintf(stderr, "More than one filter file\n");
5289 exit(-1);
5290 }
5291 if (optarg[0] == '-')
5292 filter_fp = stdin;
5293 else
5294 filter_fp = fopen(optarg, "r");
5295 if (!filter_fp) {
5296 perror("fopen filter file");
5297 exit(-1);
5298 }
5299 break;
5300 case 'v':
5301 case 'V':
5302 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
5303 exit(0);
5304 case 'z':
5305 show_sock_ctx++;
5306 /* fall through */
5307 case 'Z':
5308 if (is_selinux_enabled() <= 0) {
5309 fprintf(stderr, "ss: SELinux is not enabled.\n");
5310 exit(1);
5311 }
5312 show_proc_ctx++;
5313 user_ent_hash_build();
5314 break;
5315 case 'N':
5316 if (netns_switch(optarg))
5317 exit(1);
5318 break;
5319 case OPT_TIPCINFO:
5320 show_tipcinfo = 1;
5321 break;
5322 case OPT_TOS:
5323 show_tos = 1;
5324 break;
5325 case 'K':
5326 current_filter.kill = 1;
5327 break;
5328 case 'H':
5329 show_header = 0;
5330 break;
5331 case 'O':
5332 oneline = 1;
5333 break;
5334 case 'h':
5335 help();
5336 case '?':
5337 default:
5338 usage();
5339 }
5340 }
5341
5342 argc -= optind;
5343 argv += optind;
5344
5345 if (do_summary) {
5346 print_summary();
5347 if (do_default && argc == 0)
5348 exit(0);
5349 }
5350
5351 while (argc > 0) {
5352 if (strcmp(*argv, "state") == 0) {
5353 NEXT_ARG();
5354 if (!saw_states)
5355 state_filter = 0;
5356 state_filter |= scan_state(*argv);
5357 saw_states = 1;
5358 } else if (strcmp(*argv, "exclude") == 0 ||
5359 strcmp(*argv, "excl") == 0) {
5360 NEXT_ARG();
5361 if (!saw_states)
5362 state_filter = SS_ALL;
5363 state_filter &= ~scan_state(*argv);
5364 saw_states = 1;
5365 } else {
5366 break;
5367 }
5368 argc--; argv++;
5369 }
5370
5371 if (do_default) {
5372 state_filter = state_filter ? state_filter : SS_CONN;
5373 filter_db_parse(&current_filter, "all");
5374 }
5375
5376 filter_states_set(&current_filter, state_filter);
5377 filter_merge_defaults(&current_filter);
5378
5379 if (!numeric && resolve_hosts &&
5380 (current_filter.dbs & (UNIX_DBM|INET_L4_DBM)))
5381 init_service_resolver();
5382
5383 if (current_filter.dbs == 0) {
5384 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
5385 exit(0);
5386 }
5387 if (current_filter.families == 0) {
5388 fprintf(stderr, "ss: no families to show with such filter.\n");
5389 exit(0);
5390 }
5391 if (current_filter.states == 0) {
5392 fprintf(stderr, "ss: no socket states to show with such filter.\n");
5393 exit(0);
5394 }
5395
5396 if (dump_tcpdiag) {
5397 FILE *dump_fp = stdout;
5398
5399 if (!(current_filter.dbs & (1<<TCP_DB))) {
5400 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
5401 exit(0);
5402 }
5403 if (dump_tcpdiag[0] != '-') {
5404 dump_fp = fopen(dump_tcpdiag, "w");
5405 if (!dump_tcpdiag) {
5406 perror("fopen dump file");
5407 exit(-1);
5408 }
5409 }
5410 inet_show_netlink(&current_filter, dump_fp, IPPROTO_TCP);
5411 fflush(dump_fp);
5412 exit(0);
5413 }
5414
5415 if (ssfilter_parse(&current_filter.f, argc, argv, filter_fp))
5416 usage();
5417
5418 if (!(current_filter.dbs & (current_filter.dbs - 1)))
5419 columns[COL_NETID].disabled = 1;
5420
5421 if (!(current_filter.states & (current_filter.states - 1)))
5422 columns[COL_STATE].disabled = 1;
5423
5424 if (show_header)
5425 print_header();
5426
5427 fflush(stdout);
5428
5429 if (follow_events)
5430 exit(handle_follow_request(&current_filter));
5431
5432 if (current_filter.dbs & (1<<NETLINK_DB))
5433 netlink_show(&current_filter);
5434 if (current_filter.dbs & PACKET_DBM)
5435 packet_show(&current_filter);
5436 if (current_filter.dbs & UNIX_DBM)
5437 unix_show(&current_filter);
5438 if (current_filter.dbs & (1<<RAW_DB))
5439 raw_show(&current_filter);
5440 if (current_filter.dbs & (1<<UDP_DB))
5441 udp_show(&current_filter);
5442 if (current_filter.dbs & (1<<TCP_DB))
5443 tcp_show(&current_filter);
5444 if (current_filter.dbs & (1<<DCCP_DB))
5445 dccp_show(&current_filter);
5446 if (current_filter.dbs & (1<<SCTP_DB))
5447 sctp_show(&current_filter);
5448 if (current_filter.dbs & VSOCK_DBM)
5449 vsock_show(&current_filter);
5450 if (current_filter.dbs & (1<<TIPC_DB))
5451 tipc_show(&current_filter);
5452 if (current_filter.dbs & (1<<XDP_DB))
5453 xdp_show(&current_filter);
5454
5455 if (show_users || show_proc_ctx || show_sock_ctx)
5456 user_ent_destroy();
5457
5458 render();
5459
5460 return 0;
5461 }