]> git.proxmox.com Git - mirror_iproute2.git/blob - misc/ss.c
projects / mirror_iproute2.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ss: fix end-of-line printing in misc/ss.c
[mirror_iproute2.git] / misc / ss.c
1 /*
2 * ss.c "sockstat", socket statistics
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 */
11
12 #include <stdio.h>
13 #include <stdlib.h>
14 #include <unistd.h>
15 #include <fcntl.h>
16 #include <sys/ioctl.h>
17 #include <sys/socket.h>
18 #include <sys/uio.h>
19 #include <sys/sysmacros.h>
20 #include <netinet/in.h>
21 #include <string.h>
22 #include <errno.h>
23 #include <netdb.h>
24 #include <arpa/inet.h>
25 #include <dirent.h>
26 #include <fnmatch.h>
27 #include <getopt.h>
28 #include <stdbool.h>
29 #include <limits.h>
30 #include <stdarg.h>
31
32 #include "utils.h"
33 #include "rt_names.h"
34 #include "ll_map.h"
35 #include "libnetlink.h"
36 #include "namespace.h"
37 #include "SNAPSHOT.h"
38
39 #include <linux/tcp.h>
40 #include <linux/sock_diag.h>
41 #include <linux/inet_diag.h>
42 #include <linux/unix_diag.h>
43 #include <linux/netdevice.h> /* for MAX_ADDR_LEN */
44 #include <linux/filter.h>
45 #include <linux/xdp_diag.h>
46 #include <linux/packet_diag.h>
47 #include <linux/netlink_diag.h>
48 #include <linux/sctp.h>
49 #include <linux/vm_sockets_diag.h>
50 #include <linux/net.h>
51 #include <linux/tipc.h>
52 #include <linux/tipc_netlink.h>
53 #include <linux/tipc_sockets_diag.h>
54 #include <linux/tls.h>
55
56 /* AF_VSOCK/PF_VSOCK is only provided since glibc 2.18 */
57 #ifndef PF_VSOCK
58 #define PF_VSOCK 40
59 #endif
60 #ifndef AF_VSOCK
61 #define AF_VSOCK PF_VSOCK
62 #endif
63
64 #define MAGIC_SEQ 123456
65 #define BUF_CHUNK (1024 * 1024) /* Buffer chunk allocation size */
66 #define BUF_CHUNKS_MAX 5 /* Maximum number of allocated buffer chunks */
67 #define LEN_ALIGN(x) (((x) + 1) & ~1)
68
69 #define DIAG_REQUEST(_req, _r) \
70 struct { \
71 struct nlmsghdr nlh; \
72 _r; \
73 } _req = { \
74 .nlh = { \
75 .nlmsg_type = SOCK_DIAG_BY_FAMILY, \
76 .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,\
77 .nlmsg_seq = MAGIC_SEQ, \
78 .nlmsg_len = sizeof(_req), \
79 }, \
80 }
81
82 #if HAVE_SELINUX
83 #include <selinux/selinux.h>
84 #else
85 /* Stubs for SELinux functions */
86 static int is_selinux_enabled(void)
87 {
88 return -1;
89 }
90
91 static int getpidcon(pid_t pid, char **context)
92 {
93 *context = NULL;
94 return -1;
95 }
96
97 static int getfilecon(char *path, char **context)
98 {
99 *context = NULL;
100 return -1;
101 }
102
103 static int security_get_initial_context(char *name, char **context)
104 {
105 *context = NULL;
106 return -1;
107 }
108 #endif
109
110 int preferred_family = AF_UNSPEC;
111 static int show_options;
112 int show_details;
113 static int show_users;
114 static int show_mem;
115 static int show_tcpinfo;
116 static int show_bpf;
117 static int show_proc_ctx;
118 static int show_sock_ctx;
119 static int show_header = 1;
120 static int follow_events;
121 static int sctp_ino;
122 static int show_tipcinfo;
123 static int show_tos;
124 int numeric;
125 int oneline;
126
127 enum col_id {
128 COL_NETID,
129 COL_STATE,
130 COL_RECVQ,
131 COL_SENDQ,
132 COL_ADDR,
133 COL_SERV,
134 COL_RADDR,
135 COL_RSERV,
136 COL_EXT,
137 COL_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, default to 80 columns if TIOCGWINSZ fails */
1139 static int render_screen_width(void)
1140 {
1141 int width = 80;
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 = render_screen_width();
1163 struct column *c, *eol = columns - 1;
1164 int first, len = 0, linecols = 0;
1165
1166 /* First pass: set width for each column to measured content length */
1167 for (first = 1, c = columns; c - columns < COL_MAX; c++) {
1168 if (c->disabled)
1169 continue;
1170
1171 if (!first && c->max_len)
1172 c->width = c->max_len + strlen(c->ldelim);
1173 else
1174 c->width = c->max_len;
1175
1176 /* But don't exceed screen size. If we exceed the screen size
1177 * for even a single field, it will just start on a line of its
1178 * own and then naturally wrap.
1179 */
1180 c->width = min(c->width, screen_width);
1181
1182 if (c->width)
1183 first = 0;
1184 }
1185
1186 /* Second pass: find out newlines and distribute available spacing */
1187 for (c = columns; c - columns < COL_MAX; c++) {
1188 int pad, spacing, rem, last;
1189 struct column *tmp;
1190
1191 if (!c->width)
1192 continue;
1193
1194 linecols++;
1195 len += c->width;
1196
1197 for (last = 1, tmp = c + 1; tmp - columns < COL_MAX; tmp++) {
1198 if (tmp->width) {
1199 last = 0;
1200 break;
1201 }
1202 }
1203
1204 if (!last && len < screen_width) {
1205 /* Columns fit on screen so far, nothing to do yet */
1206 continue;
1207 }
1208
1209 if (len == screen_width) {
1210 /* Exact fit, just start with new line */
1211 goto newline;
1212 }
1213
1214 if (len > screen_width) {
1215 /* Screen width exceeded: go back one column */
1216 len -= c->width;
1217 c--;
1218 linecols--;
1219 }
1220
1221 /* Distribute remaining space to columns on this line */
1222 pad = screen_width - len;
1223 spacing = pad / linecols;
1224 rem = pad % linecols;
1225 for (tmp = c; tmp > eol; tmp--) {
1226 if (!tmp->width)
1227 continue;
1228
1229 tmp->width += spacing;
1230 if (rem) {
1231 tmp->width++;
1232 rem--;
1233 }
1234 }
1235
1236 newline:
1237 /* Line break: reset line counters, mark end-of-line */
1238 eol = c;
1239 len = 0;
1240 linecols = 0;
1241 }
1242 }
1243
1244 /* Render buffered output with spacing and delimiters, then free up buffers */
1245 static void render(void)
1246 {
1247 struct buf_token *token;
1248 int printed, line_started = 0;
1249 struct column *f;
1250
1251 if (!buffer.head)
1252 return;
1253
1254 token = (struct buf_token *)buffer.head->data;
1255
1256 /* Ensure end alignment of last token, it wasn't necessarily flushed */
1257 buffer.tail->end += buffer.cur->len % 2;
1258
1259 render_calc_width();
1260
1261 /* Rewind and replay */
1262 buffer.tail = buffer.head;
1263
1264 f = columns;
1265 while (!f->width)
1266 f++;
1267
1268 while (token) {
1269 /* Print left delimiter only if we already started a line */
1270 if (line_started++)
1271 printed = printf("%s", f->ldelim);
1272 else
1273 printed = 0;
1274
1275 /* Print field content from token data with spacing */
1276 printed += print_left_spacing(f, token->len, printed);
1277 printed += fwrite(token->data, 1, token->len, stdout);
1278 print_right_spacing(f, printed);
1279
1280 /* Go to next non-empty field, deal with end-of-line */
1281 do {
1282 if (field_is_last(f)) {
1283 printf("\n");
1284 f = columns;
1285 line_started = 0;
1286 } else {
1287 f++;
1288 }
1289 } while (f->disabled);
1290
1291 token = buf_token_next(token);
1292 }
1293 /* Deal with final end-of-line when the last non-empty field printed
1294 * is not the last field.
1295 */
1296 if (line_started)
1297 printf("\n");
1298
1299 buf_free_all();
1300 current_field = columns;
1301 }
1302
1303 /* Move to next field, and render buffer if we reached the maximum number of
1304 * chunks, at the last field in a line.
1305 */
1306 static void field_next(void)
1307 {
1308 if (field_is_last(current_field) && buffer.chunks >= BUF_CHUNKS_MAX) {
1309 render();
1310 return;
1311 }
1312
1313 field_flush(current_field);
1314 if (field_is_last(current_field))
1315 current_field = columns;
1316 else
1317 current_field++;
1318 }
1319
1320 /* Walk through fields and flush them until we reach the desired one */
1321 static void field_set(enum col_id id)
1322 {
1323 while (id != current_field - columns)
1324 field_next();
1325 }
1326
1327 /* Print header for all non-empty columns */
1328 static void print_header(void)
1329 {
1330 while (!field_is_last(current_field)) {
1331 if (!current_field->disabled)
1332 out("%s", current_field->header);
1333 field_next();
1334 }
1335 }
1336
1337 static void sock_state_print(struct sockstat *s)
1338 {
1339 const char *sock_name;
1340 static const char * const sstate_name[] = {
1341 "UNKNOWN",
1342 [SS_ESTABLISHED] = "ESTAB",
1343 [SS_SYN_SENT] = "SYN-SENT",
1344 [SS_SYN_RECV] = "SYN-RECV",
1345 [SS_FIN_WAIT1] = "FIN-WAIT-1",
1346 [SS_FIN_WAIT2] = "FIN-WAIT-2",
1347 [SS_TIME_WAIT] = "TIME-WAIT",
1348 [SS_CLOSE] = "UNCONN",
1349 [SS_CLOSE_WAIT] = "CLOSE-WAIT",
1350 [SS_LAST_ACK] = "LAST-ACK",
1351 [SS_LISTEN] = "LISTEN",
1352 [SS_CLOSING] = "CLOSING",
1353 };
1354
1355 switch (s->local.family) {
1356 case AF_UNIX:
1357 sock_name = unix_netid_name(s->type);
1358 break;
1359 case AF_INET:
1360 case AF_INET6:
1361 sock_name = proto_name(s->type);
1362 break;
1363 case AF_PACKET:
1364 sock_name = s->type == SOCK_RAW ? "p_raw" : "p_dgr";
1365 break;
1366 case AF_NETLINK:
1367 sock_name = "nl";
1368 break;
1369 case AF_TIPC:
1370 sock_name = tipc_netid_name(s->type);
1371 break;
1372 case AF_VSOCK:
1373 sock_name = vsock_netid_name(s->type);
1374 break;
1375 case AF_XDP:
1376 sock_name = "xdp";
1377 break;
1378 default:
1379 sock_name = "unknown";
1380 }
1381
1382 if (is_sctp_assoc(s, sock_name)) {
1383 field_set(COL_STATE); /* Empty Netid field */
1384 out("`- %s", sctp_sstate_name[s->state]);
1385 } else {
1386 field_set(COL_NETID);
1387 out("%s", sock_name);
1388 field_set(COL_STATE);
1389 out("%s", sstate_name[s->state]);
1390 }
1391
1392 field_set(COL_RECVQ);
1393 out("%-6d", s->rq);
1394 field_set(COL_SENDQ);
1395 out("%-6d", s->wq);
1396 field_set(COL_ADDR);
1397 }
1398
1399 static void sock_details_print(struct sockstat *s)
1400 {
1401 if (s->uid)
1402 out(" uid:%u", s->uid);
1403
1404 out(" ino:%u", s->ino);
1405 out(" sk:%llx", s->sk);
1406
1407 if (s->mark)
1408 out(" fwmark:0x%x", s->mark);
1409 }
1410
1411 static void sock_addr_print(const char *addr, char *delim, const char *port,
1412 const char *ifname)
1413 {
1414 if (ifname)
1415 out("%s" "%%" "%s%s", addr, ifname, delim);
1416 else
1417 out("%s%s", addr, delim);
1418
1419 field_next();
1420 out("%s", port);
1421 field_next();
1422 }
1423
1424 static const char *print_ms_timer(unsigned int timeout)
1425 {
1426 static char buf[64];
1427 int secs, msecs, minutes;
1428
1429 secs = timeout/1000;
1430 minutes = secs/60;
1431 secs = secs%60;
1432 msecs = timeout%1000;
1433 buf[0] = 0;
1434 if (minutes) {
1435 msecs = 0;
1436 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
1437 if (minutes > 9)
1438 secs = 0;
1439 }
1440 if (secs) {
1441 if (secs > 9)
1442 msecs = 0;
1443 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
1444 }
1445 if (msecs)
1446 sprintf(buf+strlen(buf), "%03dms", msecs);
1447 return buf;
1448 }
1449
1450 struct scache {
1451 struct scache *next;
1452 int port;
1453 char *name;
1454 const char *proto;
1455 };
1456
1457 static struct scache *rlist;
1458
1459 static void init_service_resolver(void)
1460 {
1461 char buf[128];
1462 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
1463
1464 if (!fp)
1465 return;
1466
1467 if (!fgets(buf, sizeof(buf), fp)) {
1468 pclose(fp);
1469 return;
1470 }
1471 while (fgets(buf, sizeof(buf), fp) != NULL) {
1472 unsigned int progn, port;
1473 char proto[128], prog[128] = "rpc.";
1474 struct scache *c;
1475
1476 if (sscanf(buf, "%u %*d %s %u %s",
1477 &progn, proto, &port, prog+4) != 4)
1478 continue;
1479
1480 if (!(c = malloc(sizeof(*c))))
1481 continue;
1482
1483 c->port = port;
1484 c->name = strdup(prog);
1485 if (strcmp(proto, TCP_PROTO) == 0)
1486 c->proto = TCP_PROTO;
1487 else if (strcmp(proto, UDP_PROTO) == 0)
1488 c->proto = UDP_PROTO;
1489 else if (strcmp(proto, SCTP_PROTO) == 0)
1490 c->proto = SCTP_PROTO;
1491 else
1492 c->proto = NULL;
1493 c->next = rlist;
1494 rlist = c;
1495 }
1496 pclose(fp);
1497 }
1498
1499 /* Even do not try default linux ephemeral port ranges:
1500 * default /etc/services contains so much of useless crap
1501 * wouldbe "allocated" to this area that resolution
1502 * is really harmful. I shrug each time when seeing
1503 * "socks" or "cfinger" in dumps.
1504 */
1505 static int is_ephemeral(int port)
1506 {
1507 static int min = 0, max;
1508
1509 if (!min) {
1510 FILE *f = ephemeral_ports_open();
1511
1512 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
1513 min = 1024;
1514 max = 4999;
1515 }
1516 if (f)
1517 fclose(f);
1518 }
1519 return port >= min && port <= max;
1520 }
1521
1522
1523 static const char *__resolve_service(int port)
1524 {
1525 struct scache *c;
1526
1527 for (c = rlist; c; c = c->next) {
1528 if (c->port == port && c->proto == dg_proto)
1529 return c->name;
1530 }
1531
1532 if (!is_ephemeral(port)) {
1533 static int notfirst;
1534 struct servent *se;
1535
1536 if (!notfirst) {
1537 setservent(1);
1538 notfirst = 1;
1539 }
1540 se = getservbyport(htons(port), dg_proto);
1541 if (se)
1542 return se->s_name;
1543 }
1544
1545 return NULL;
1546 }
1547
1548 #define SCACHE_BUCKETS 1024
1549 static struct scache *cache_htab[SCACHE_BUCKETS];
1550
1551 static const char *resolve_service(int port)
1552 {
1553 static char buf[128];
1554 struct scache *c;
1555 const char *res;
1556 int hash;
1557
1558 if (port == 0) {
1559 buf[0] = '*';
1560 buf[1] = 0;
1561 return buf;
1562 }
1563
1564 if (numeric)
1565 goto do_numeric;
1566
1567 if (dg_proto == RAW_PROTO)
1568 return inet_proto_n2a(port, buf, sizeof(buf));
1569
1570
1571 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
1572
1573 for (c = cache_htab[hash]; c; c = c->next) {
1574 if (c->port == port && c->proto == dg_proto)
1575 goto do_cache;
1576 }
1577
1578 c = malloc(sizeof(*c));
1579 if (!c)
1580 goto do_numeric;
1581 res = __resolve_service(port);
1582 c->port = port;
1583 c->name = res ? strdup(res) : NULL;
1584 c->proto = dg_proto;
1585 c->next = cache_htab[hash];
1586 cache_htab[hash] = c;
1587
1588 do_cache:
1589 if (c->name)
1590 return c->name;
1591
1592 do_numeric:
1593 sprintf(buf, "%u", port);
1594 return buf;
1595 }
1596
1597 static void inet_addr_print(const inet_prefix *a, int port,
1598 unsigned int ifindex, bool v6only)
1599 {
1600 char buf[1024];
1601 const char *ap = buf;
1602 const char *ifname = NULL;
1603
1604 if (a->family == AF_INET) {
1605 ap = format_host(AF_INET, 4, a->data);
1606 } else {
1607 if (!v6only &&
1608 !memcmp(a->data, &in6addr_any, sizeof(in6addr_any))) {
1609 buf[0] = '*';
1610 buf[1] = 0;
1611 } else {
1612 ap = format_host(a->family, 16, a->data);
1613
1614 /* Numeric IPv6 addresses should be bracketed */
1615 if (strchr(ap, ':')) {
1616 snprintf(buf, sizeof(buf),
1617 "[%s]", ap);
1618 ap = buf;
1619 }
1620 }
1621 }
1622
1623 if (ifindex)
1624 ifname = ll_index_to_name(ifindex);
1625
1626 sock_addr_print(ap, ":", resolve_service(port), ifname);
1627 }
1628
1629 struct aafilter {
1630 inet_prefix addr;
1631 int port;
1632 unsigned int iface;
1633 __u32 mark;
1634 __u32 mask;
1635 struct aafilter *next;
1636 };
1637
1638 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1639 int plen)
1640 {
1641 if (!inet_addr_match(a, p, plen))
1642 return 0;
1643
1644 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1645 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1646 * sockets. Fair? */
1647 if (p->family == AF_INET && a->family == AF_INET6) {
1648 if (a->data[0] == 0 && a->data[1] == 0 &&
1649 a->data[2] == htonl(0xffff)) {
1650 inet_prefix tmp = *a;
1651
1652 tmp.data[0] = a->data[3];
1653 return inet_addr_match(&tmp, p, plen);
1654 }
1655 }
1656 return 1;
1657 }
1658
1659 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1660 {
1661 char *addr, *pattern;
1662
1663 memcpy(&addr, a->data, sizeof(addr));
1664 memcpy(&pattern, p->data, sizeof(pattern));
1665 if (pattern == NULL)
1666 return 1;
1667 if (addr == NULL)
1668 addr = "";
1669 return !fnmatch(pattern, addr, 0);
1670 }
1671
1672 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1673 {
1674 switch (f->type) {
1675 case SSF_S_AUTO:
1676 {
1677 if (s->local.family == AF_UNIX) {
1678 char *p;
1679
1680 memcpy(&p, s->local.data, sizeof(p));
1681 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1682 strspn(p+1, "0123456789abcdef") == 5);
1683 }
1684 if (s->local.family == AF_PACKET)
1685 return s->lport == 0 && s->local.data[0] == 0;
1686 if (s->local.family == AF_NETLINK)
1687 return s->lport < 0;
1688 if (s->local.family == AF_VSOCK)
1689 return s->lport > 1023;
1690
1691 return is_ephemeral(s->lport);
1692 }
1693 case SSF_DCOND:
1694 {
1695 struct aafilter *a = (void *)f->pred;
1696
1697 if (a->addr.family == AF_UNIX)
1698 return unix_match(&s->remote, &a->addr);
1699 if (a->port != -1 && a->port != s->rport)
1700 return 0;
1701 if (a->addr.bitlen) {
1702 do {
1703 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1704 return 1;
1705 } while ((a = a->next) != NULL);
1706 return 0;
1707 }
1708 return 1;
1709 }
1710 case SSF_SCOND:
1711 {
1712 struct aafilter *a = (void *)f->pred;
1713
1714 if (a->addr.family == AF_UNIX)
1715 return unix_match(&s->local, &a->addr);
1716 if (a->port != -1 && a->port != s->lport)
1717 return 0;
1718 if (a->addr.bitlen) {
1719 do {
1720 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1721 return 1;
1722 } while ((a = a->next) != NULL);
1723 return 0;
1724 }
1725 return 1;
1726 }
1727 case SSF_D_GE:
1728 {
1729 struct aafilter *a = (void *)f->pred;
1730
1731 return s->rport >= a->port;
1732 }
1733 case SSF_D_LE:
1734 {
1735 struct aafilter *a = (void *)f->pred;
1736
1737 return s->rport <= a->port;
1738 }
1739 case SSF_S_GE:
1740 {
1741 struct aafilter *a = (void *)f->pred;
1742
1743 return s->lport >= a->port;
1744 }
1745 case SSF_S_LE:
1746 {
1747 struct aafilter *a = (void *)f->pred;
1748
1749 return s->lport <= a->port;
1750 }
1751 case SSF_DEVCOND:
1752 {
1753 struct aafilter *a = (void *)f->pred;
1754
1755 return s->iface == a->iface;
1756 }
1757 case SSF_MARKMASK:
1758 {
1759 struct aafilter *a = (void *)f->pred;
1760
1761 return (s->mark & a->mask) == a->mark;
1762 }
1763 /* Yup. It is recursion. Sorry. */
1764 case SSF_AND:
1765 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1766 case SSF_OR:
1767 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1768 case SSF_NOT:
1769 return !run_ssfilter(f->pred, s);
1770 default:
1771 abort();
1772 }
1773 }
1774
1775 /* Relocate external jumps by reloc. */
1776 static void ssfilter_patch(char *a, int len, int reloc)
1777 {
1778 while (len > 0) {
1779 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)a;
1780
1781 if (op->no == len+4)
1782 op->no += reloc;
1783 len -= op->yes;
1784 a += op->yes;
1785 }
1786 if (len < 0)
1787 abort();
1788 }
1789
1790 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1791 {
1792 switch (f->type) {
1793 case SSF_S_AUTO:
1794 {
1795 if (!(*bytecode = malloc(4))) abort();
1796 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1797 return 4;
1798 }
1799 case SSF_DCOND:
1800 case SSF_SCOND:
1801 {
1802 struct aafilter *a = (void *)f->pred;
1803 struct aafilter *b;
1804 char *ptr;
1805 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1806 int len = 0;
1807
1808 for (b = a; b; b = b->next) {
1809 len += 4 + sizeof(struct inet_diag_hostcond);
1810 if (a->addr.family == AF_INET6)
1811 len += 16;
1812 else
1813 len += 4;
1814 if (b->next)
1815 len += 4;
1816 }
1817 if (!(ptr = malloc(len))) abort();
1818 *bytecode = ptr;
1819 for (b = a; b; b = b->next) {
1820 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1821 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1822 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1823 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond *)(ptr+4);
1824
1825 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1826 cond->family = a->addr.family;
1827 cond->port = a->port;
1828 cond->prefix_len = a->addr.bitlen;
1829 memcpy(cond->addr, a->addr.data, alen);
1830 ptr += oplen;
1831 if (b->next) {
1832 op = (struct inet_diag_bc_op *)ptr;
1833 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1834 ptr += 4;
1835 }
1836 }
1837 return ptr - *bytecode;
1838 }
1839 case SSF_D_GE:
1840 {
1841 struct aafilter *x = (void *)f->pred;
1842
1843 if (!(*bytecode = malloc(8))) abort();
1844 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_GE, 8, 12 };
1845 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1846 return 8;
1847 }
1848 case SSF_D_LE:
1849 {
1850 struct aafilter *x = (void *)f->pred;
1851
1852 if (!(*bytecode = malloc(8))) abort();
1853 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_LE, 8, 12 };
1854 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1855 return 8;
1856 }
1857 case SSF_S_GE:
1858 {
1859 struct aafilter *x = (void *)f->pred;
1860
1861 if (!(*bytecode = malloc(8))) abort();
1862 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1863 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1864 return 8;
1865 }
1866 case SSF_S_LE:
1867 {
1868 struct aafilter *x = (void *)f->pred;
1869
1870 if (!(*bytecode = malloc(8))) abort();
1871 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1872 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1873 return 8;
1874 }
1875
1876 case SSF_AND:
1877 {
1878 char *a1 = NULL, *a2 = NULL, *a;
1879 int l1, l2;
1880
1881 l1 = ssfilter_bytecompile(f->pred, &a1);
1882 l2 = ssfilter_bytecompile(f->post, &a2);
1883 if (!l1 || !l2) {
1884 free(a1);
1885 free(a2);
1886 return 0;
1887 }
1888 if (!(a = malloc(l1+l2))) abort();
1889 memcpy(a, a1, l1);
1890 memcpy(a+l1, a2, l2);
1891 free(a1); free(a2);
1892 ssfilter_patch(a, l1, l2);
1893 *bytecode = a;
1894 return l1+l2;
1895 }
1896 case SSF_OR:
1897 {
1898 char *a1 = NULL, *a2 = NULL, *a;
1899 int l1, l2;
1900
1901 l1 = ssfilter_bytecompile(f->pred, &a1);
1902 l2 = ssfilter_bytecompile(f->post, &a2);
1903 if (!l1 || !l2) {
1904 free(a1);
1905 free(a2);
1906 return 0;
1907 }
1908 if (!(a = malloc(l1+l2+4))) abort();
1909 memcpy(a, a1, l1);
1910 memcpy(a+l1+4, a2, l2);
1911 free(a1); free(a2);
1912 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1913 *bytecode = a;
1914 return l1+l2+4;
1915 }
1916 case SSF_NOT:
1917 {
1918 char *a1 = NULL, *a;
1919 int l1;
1920
1921 l1 = ssfilter_bytecompile(f->pred, &a1);
1922 if (!l1) {
1923 free(a1);
1924 return 0;
1925 }
1926 if (!(a = malloc(l1+4))) abort();
1927 memcpy(a, a1, l1);
1928 free(a1);
1929 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1930 *bytecode = a;
1931 return l1+4;
1932 }
1933 case SSF_DEVCOND:
1934 {
1935 /* bytecompile for SSF_DEVCOND not supported yet */
1936 return 0;
1937 }
1938 case SSF_MARKMASK:
1939 {
1940 struct aafilter *a = (void *)f->pred;
1941 struct instr {
1942 struct inet_diag_bc_op op;
1943 struct inet_diag_markcond cond;
1944 };
1945 int inslen = sizeof(struct instr);
1946
1947 if (!(*bytecode = malloc(inslen))) abort();
1948 ((struct instr *)*bytecode)[0] = (struct instr) {
1949 { INET_DIAG_BC_MARK_COND, inslen, inslen + 4 },
1950 { a->mark, a->mask},
1951 };
1952
1953 return inslen;
1954 }
1955 default:
1956 abort();
1957 }
1958 }
1959
1960 static int remember_he(struct aafilter *a, struct hostent *he)
1961 {
1962 char **ptr = he->h_addr_list;
1963 int cnt = 0;
1964 int len;
1965
1966 if (he->h_addrtype == AF_INET)
1967 len = 4;
1968 else if (he->h_addrtype == AF_INET6)
1969 len = 16;
1970 else
1971 return 0;
1972
1973 while (*ptr) {
1974 struct aafilter *b = a;
1975
1976 if (a->addr.bitlen) {
1977 if ((b = malloc(sizeof(*b))) == NULL)
1978 return cnt;
1979 *b = *a;
1980 a->next = b;
1981 }
1982 memcpy(b->addr.data, *ptr, len);
1983 b->addr.bytelen = len;
1984 b->addr.bitlen = len*8;
1985 b->addr.family = he->h_addrtype;
1986 ptr++;
1987 cnt++;
1988 }
1989 return cnt;
1990 }
1991
1992 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
1993 {
1994 static int notfirst;
1995 int cnt = 0;
1996 struct hostent *he;
1997
1998 a->addr.bitlen = 0;
1999 if (!notfirst) {
2000 sethostent(1);
2001 notfirst = 1;
2002 }
2003 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
2004 if (he)
2005 cnt = remember_he(a, he);
2006 if (fam == AF_UNSPEC) {
2007 he = gethostbyname2(addr, AF_INET6);
2008 if (he)
2009 cnt += remember_he(a, he);
2010 }
2011 return !cnt;
2012 }
2013
2014 static int xll_initted;
2015
2016 static void xll_init(void)
2017 {
2018 struct rtnl_handle rth;
2019
2020 if (rtnl_open(&rth, 0) < 0)
2021 exit(1);
2022
2023 ll_init_map(&rth);
2024 rtnl_close(&rth);
2025 xll_initted = 1;
2026 }
2027
2028 static const char *xll_index_to_name(int index)
2029 {
2030 if (!xll_initted)
2031 xll_init();
2032 return ll_index_to_name(index);
2033 }
2034
2035 static int xll_name_to_index(const char *dev)
2036 {
2037 if (!xll_initted)
2038 xll_init();
2039 return ll_name_to_index(dev);
2040 }
2041
2042 void *parse_devcond(char *name)
2043 {
2044 struct aafilter a = { .iface = 0 };
2045 struct aafilter *res;
2046
2047 a.iface = xll_name_to_index(name);
2048 if (a.iface == 0) {
2049 char *end;
2050 unsigned long n;
2051
2052 n = strtoul(name, &end, 0);
2053 if (!end || end == name || *end || n > UINT_MAX)
2054 return NULL;
2055
2056 a.iface = n;
2057 }
2058
2059 res = malloc(sizeof(*res));
2060 *res = a;
2061
2062 return res;
2063 }
2064
2065 static void vsock_set_inet_prefix(inet_prefix *a, __u32 cid)
2066 {
2067 *a = (inet_prefix){
2068 .bytelen = sizeof(cid),
2069 .family = AF_VSOCK,
2070 };
2071 memcpy(a->data, &cid, sizeof(cid));
2072 }
2073
2074 void *parse_hostcond(char *addr, bool is_port)
2075 {
2076 char *port = NULL;
2077 struct aafilter a = { .port = -1 };
2078 struct aafilter *res;
2079 int fam = preferred_family;
2080 struct filter *f = &current_filter;
2081
2082 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
2083 char *p;
2084
2085 a.addr.family = AF_UNIX;
2086 if (strncmp(addr, "unix:", 5) == 0)
2087 addr += 5;
2088 p = strdup(addr);
2089 a.addr.bitlen = 8*strlen(p);
2090 memcpy(a.addr.data, &p, sizeof(p));
2091 fam = AF_UNIX;
2092 goto out;
2093 }
2094
2095 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
2096 a.addr.family = AF_PACKET;
2097 a.addr.bitlen = 0;
2098 if (strncmp(addr, "link:", 5) == 0)
2099 addr += 5;
2100 port = strchr(addr, ':');
2101 if (port) {
2102 *port = 0;
2103 if (port[1] && strcmp(port+1, "*")) {
2104 if (get_integer(&a.port, port+1, 0)) {
2105 if ((a.port = xll_name_to_index(port+1)) <= 0)
2106 return NULL;
2107 }
2108 }
2109 }
2110 if (addr[0] && strcmp(addr, "*")) {
2111 unsigned short tmp;
2112
2113 a.addr.bitlen = 32;
2114 if (ll_proto_a2n(&tmp, addr))
2115 return NULL;
2116 a.addr.data[0] = ntohs(tmp);
2117 }
2118 fam = AF_PACKET;
2119 goto out;
2120 }
2121
2122 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
2123 a.addr.family = AF_NETLINK;
2124 a.addr.bitlen = 0;
2125 if (strncmp(addr, "netlink:", 8) == 0)
2126 addr += 8;
2127 port = strchr(addr, ':');
2128 if (port) {
2129 *port = 0;
2130 if (port[1] && strcmp(port+1, "*")) {
2131 if (get_integer(&a.port, port+1, 0)) {
2132 if (strcmp(port+1, "kernel") == 0)
2133 a.port = 0;
2134 else
2135 return NULL;
2136 }
2137 }
2138 }
2139 if (addr[0] && strcmp(addr, "*")) {
2140 a.addr.bitlen = 32;
2141 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
2142 return NULL;
2143 }
2144 fam = AF_NETLINK;
2145 goto out;
2146 }
2147
2148 if (fam == AF_VSOCK || strncmp(addr, "vsock:", 6) == 0) {
2149 __u32 cid = ~(__u32)0;
2150
2151 a.addr.family = AF_VSOCK;
2152 if (strncmp(addr, "vsock:", 6) == 0)
2153 addr += 6;
2154
2155 if (is_port)
2156 port = addr;
2157 else {
2158 port = strchr(addr, ':');
2159 if (port) {
2160 *port = '\0';
2161 port++;
2162 }
2163 }
2164
2165 if (port && strcmp(port, "*") &&
2166 get_u32((__u32 *)&a.port, port, 0))
2167 return NULL;
2168
2169 if (addr[0] && strcmp(addr, "*")) {
2170 a.addr.bitlen = 32;
2171 if (get_u32(&cid, addr, 0))
2172 return NULL;
2173 }
2174 vsock_set_inet_prefix(&a.addr, cid);
2175 fam = AF_VSOCK;
2176 goto out;
2177 }
2178
2179 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
2180 fam = AF_INET;
2181 if (!strncmp(addr, "inet:", 5))
2182 addr += 5;
2183 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
2184 fam = AF_INET6;
2185 if (!strncmp(addr, "inet6:", 6))
2186 addr += 6;
2187 }
2188
2189 /* URL-like literal [] */
2190 if (addr[0] == '[') {
2191 addr++;
2192 if ((port = strchr(addr, ']')) == NULL)
2193 return NULL;
2194 *port++ = 0;
2195 } else if (addr[0] == '*') {
2196 port = addr+1;
2197 } else {
2198 port = strrchr(strchr(addr, '/') ? : addr, ':');
2199 }
2200
2201 if (is_port)
2202 port = addr;
2203
2204 if (port && *port) {
2205 if (*port == ':')
2206 *port++ = 0;
2207
2208 if (*port && *port != '*') {
2209 if (get_integer(&a.port, port, 0)) {
2210 struct servent *se1 = NULL;
2211 struct servent *se2 = NULL;
2212
2213 if (current_filter.dbs&(1<<UDP_DB))
2214 se1 = getservbyname(port, UDP_PROTO);
2215 if (current_filter.dbs&(1<<TCP_DB))
2216 se2 = getservbyname(port, TCP_PROTO);
2217 if (se1 && se2 && se1->s_port != se2->s_port) {
2218 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2219 return NULL;
2220 }
2221 if (!se1)
2222 se1 = se2;
2223 if (se1) {
2224 a.port = ntohs(se1->s_port);
2225 } else {
2226 struct scache *s;
2227
2228 for (s = rlist; s; s = s->next) {
2229 if ((s->proto == UDP_PROTO &&
2230 (current_filter.dbs&(1<<UDP_DB))) ||
2231 (s->proto == TCP_PROTO &&
2232 (current_filter.dbs&(1<<TCP_DB)))) {
2233 if (s->name && strcmp(s->name, port) == 0) {
2234 if (a.port > 0 && a.port != s->port) {
2235 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2236 return NULL;
2237 }
2238 a.port = s->port;
2239 }
2240 }
2241 }
2242 if (a.port <= 0) {
2243 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
2244 return NULL;
2245 }
2246 }
2247 }
2248 }
2249 }
2250 if (!is_port && *addr && *addr != '*') {
2251 if (get_prefix_1(&a.addr, addr, fam)) {
2252 if (get_dns_host(&a, addr, fam)) {
2253 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
2254 return NULL;
2255 }
2256 }
2257 }
2258
2259 out:
2260 if (fam != AF_UNSPEC) {
2261 int states = f->states;
2262 f->families = 0;
2263 filter_af_set(f, fam);
2264 filter_states_set(f, states);
2265 }
2266
2267 res = malloc(sizeof(*res));
2268 if (res)
2269 memcpy(res, &a, sizeof(a));
2270 return res;
2271 }
2272
2273 void *parse_markmask(const char *markmask)
2274 {
2275 struct aafilter a, *res;
2276
2277 if (strchr(markmask, '/')) {
2278 if (sscanf(markmask, "%i/%i", &a.mark, &a.mask) != 2)
2279 return NULL;
2280 } else {
2281 a.mask = 0xffffffff;
2282 if (sscanf(markmask, "%i", &a.mark) != 1)
2283 return NULL;
2284 }
2285
2286 res = malloc(sizeof(*res));
2287 if (res)
2288 memcpy(res, &a, sizeof(a));
2289 return res;
2290 }
2291
2292 static void proc_ctx_print(struct sockstat *s)
2293 {
2294 char *buf;
2295
2296 if (show_proc_ctx || show_sock_ctx) {
2297 if (find_entry(s->ino, &buf,
2298 (show_proc_ctx & show_sock_ctx) ?
2299 PROC_SOCK_CTX : PROC_CTX) > 0) {
2300 out(" users:(%s)", buf);
2301 free(buf);
2302 }
2303 } else if (show_users) {
2304 if (find_entry(s->ino, &buf, USERS) > 0) {
2305 out(" users:(%s)", buf);
2306 free(buf);
2307 }
2308 }
2309 }
2310
2311 static void inet_stats_print(struct sockstat *s, bool v6only)
2312 {
2313 sock_state_print(s);
2314
2315 inet_addr_print(&s->local, s->lport, s->iface, v6only);
2316 inet_addr_print(&s->remote, s->rport, 0, v6only);
2317
2318 proc_ctx_print(s);
2319 }
2320
2321 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
2322 sockstat * s)
2323 {
2324 s->local.family = s->remote.family = family;
2325 if (family == AF_INET) {
2326 sscanf(loc, "%x:%x", s->local.data, (unsigned *)&s->lport);
2327 sscanf(rem, "%x:%x", s->remote.data, (unsigned *)&s->rport);
2328 s->local.bytelen = s->remote.bytelen = 4;
2329 return 0;
2330 } else {
2331 sscanf(loc, "%08x%08x%08x%08x:%x",
2332 s->local.data,
2333 s->local.data + 1,
2334 s->local.data + 2,
2335 s->local.data + 3,
2336 &s->lport);
2337 sscanf(rem, "%08x%08x%08x%08x:%x",
2338 s->remote.data,
2339 s->remote.data + 1,
2340 s->remote.data + 2,
2341 s->remote.data + 3,
2342 &s->rport);
2343 s->local.bytelen = s->remote.bytelen = 16;
2344 return 0;
2345 }
2346 return -1;
2347 }
2348
2349 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
2350 {
2351 char *p;
2352
2353 if ((p = strchr(line, ':')) == NULL)
2354 return -1;
2355
2356 *loc = p+2;
2357 if ((p = strchr(*loc, ':')) == NULL)
2358 return -1;
2359
2360 p[5] = 0;
2361 *rem = p+6;
2362 if ((p = strchr(*rem, ':')) == NULL)
2363 return -1;
2364
2365 p[5] = 0;
2366 *data = p+6;
2367 return 0;
2368 }
2369
2370 static char *sprint_bw(char *buf, double bw)
2371 {
2372 if (numeric)
2373 sprintf(buf, "%.0f", bw);
2374 else if (bw > 1000000.)
2375 sprintf(buf, "%.1fM", bw / 1000000.);
2376 else if (bw > 1000.)
2377 sprintf(buf, "%.1fK", bw / 1000.);
2378 else
2379 sprintf(buf, "%g", bw);
2380
2381 return buf;
2382 }
2383
2384 static void sctp_stats_print(struct sctp_info *s)
2385 {
2386 if (s->sctpi_tag)
2387 out(" tag:%x", s->sctpi_tag);
2388 if (s->sctpi_state)
2389 out(" state:%s", sctp_sstate_name[s->sctpi_state]);
2390 if (s->sctpi_rwnd)
2391 out(" rwnd:%d", s->sctpi_rwnd);
2392 if (s->sctpi_unackdata)
2393 out(" unackdata:%d", s->sctpi_unackdata);
2394 if (s->sctpi_penddata)
2395 out(" penddata:%d", s->sctpi_penddata);
2396 if (s->sctpi_instrms)
2397 out(" instrms:%d", s->sctpi_instrms);
2398 if (s->sctpi_outstrms)
2399 out(" outstrms:%d", s->sctpi_outstrms);
2400 if (s->sctpi_inqueue)
2401 out(" inqueue:%d", s->sctpi_inqueue);
2402 if (s->sctpi_outqueue)
2403 out(" outqueue:%d", s->sctpi_outqueue);
2404 if (s->sctpi_overall_error)
2405 out(" overerr:%d", s->sctpi_overall_error);
2406 if (s->sctpi_max_burst)
2407 out(" maxburst:%d", s->sctpi_max_burst);
2408 if (s->sctpi_maxseg)
2409 out(" maxseg:%d", s->sctpi_maxseg);
2410 if (s->sctpi_peer_rwnd)
2411 out(" prwnd:%d", s->sctpi_peer_rwnd);
2412 if (s->sctpi_peer_tag)
2413 out(" ptag:%x", s->sctpi_peer_tag);
2414 if (s->sctpi_peer_capable)
2415 out(" pcapable:%d", s->sctpi_peer_capable);
2416 if (s->sctpi_peer_sack)
2417 out(" psack:%d", s->sctpi_peer_sack);
2418 if (s->sctpi_s_autoclose)
2419 out(" autoclose:%d", s->sctpi_s_autoclose);
2420 if (s->sctpi_s_adaptation_ind)
2421 out(" adapind:%d", s->sctpi_s_adaptation_ind);
2422 if (s->sctpi_s_pd_point)
2423 out(" pdpoint:%d", s->sctpi_s_pd_point);
2424 if (s->sctpi_s_nodelay)
2425 out(" nodelay:%d", s->sctpi_s_nodelay);
2426 if (s->sctpi_s_disable_fragments)
2427 out(" nofrag:%d", s->sctpi_s_disable_fragments);
2428 if (s->sctpi_s_v4mapped)
2429 out(" v4mapped:%d", s->sctpi_s_v4mapped);
2430 if (s->sctpi_s_frag_interleave)
2431 out(" fraginl:%d", s->sctpi_s_frag_interleave);
2432 }
2433
2434 static void tcp_stats_print(struct tcpstat *s)
2435 {
2436 char b1[64];
2437
2438 if (s->has_ts_opt)
2439 out(" ts");
2440 if (s->has_sack_opt)
2441 out(" sack");
2442 if (s->has_ecn_opt)
2443 out(" ecn");
2444 if (s->has_ecnseen_opt)
2445 out(" ecnseen");
2446 if (s->has_fastopen_opt)
2447 out(" fastopen");
2448 if (s->cong_alg[0])
2449 out(" %s", s->cong_alg);
2450 if (s->has_wscale_opt)
2451 out(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
2452 if (s->rto)
2453 out(" rto:%g", s->rto);
2454 if (s->backoff)
2455 out(" backoff:%u", s->backoff);
2456 if (s->rtt)
2457 out(" rtt:%g/%g", s->rtt, s->rttvar);
2458 if (s->ato)
2459 out(" ato:%g", s->ato);
2460
2461 if (s->qack)
2462 out(" qack:%d", s->qack);
2463 if (s->qack & 1)
2464 out(" bidir");
2465
2466 if (s->mss)
2467 out(" mss:%d", s->mss);
2468 if (s->pmtu)
2469 out(" pmtu:%u", s->pmtu);
2470 if (s->rcv_mss)
2471 out(" rcvmss:%d", s->rcv_mss);
2472 if (s->advmss)
2473 out(" advmss:%d", s->advmss);
2474 if (s->cwnd)
2475 out(" cwnd:%u", s->cwnd);
2476 if (s->ssthresh)
2477 out(" ssthresh:%d", s->ssthresh);
2478
2479 if (s->bytes_sent)
2480 out(" bytes_sent:%llu", s->bytes_sent);
2481 if (s->bytes_retrans)
2482 out(" bytes_retrans:%llu", s->bytes_retrans);
2483 if (s->bytes_acked)
2484 out(" bytes_acked:%llu", s->bytes_acked);
2485 if (s->bytes_received)
2486 out(" bytes_received:%llu", s->bytes_received);
2487 if (s->segs_out)
2488 out(" segs_out:%u", s->segs_out);
2489 if (s->segs_in)
2490 out(" segs_in:%u", s->segs_in);
2491 if (s->data_segs_out)
2492 out(" data_segs_out:%u", s->data_segs_out);
2493 if (s->data_segs_in)
2494 out(" data_segs_in:%u", s->data_segs_in);
2495
2496 if (s->dctcp && s->dctcp->enabled) {
2497 struct dctcpstat *dctcp = s->dctcp;
2498
2499 out(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
2500 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
2501 dctcp->ab_tot);
2502 } else if (s->dctcp) {
2503 out(" dctcp:fallback_mode");
2504 }
2505
2506 if (s->bbr_info) {
2507 __u64 bw;
2508
2509 bw = s->bbr_info->bbr_bw_hi;
2510 bw <<= 32;
2511 bw |= s->bbr_info->bbr_bw_lo;
2512
2513 out(" bbr:(bw:%sbps,mrtt:%g",
2514 sprint_bw(b1, bw * 8.0),
2515 (double)s->bbr_info->bbr_min_rtt / 1000.0);
2516 if (s->bbr_info->bbr_pacing_gain)
2517 out(",pacing_gain:%g",
2518 (double)s->bbr_info->bbr_pacing_gain / 256.0);
2519 if (s->bbr_info->bbr_cwnd_gain)
2520 out(",cwnd_gain:%g",
2521 (double)s->bbr_info->bbr_cwnd_gain / 256.0);
2522 out(")");
2523 }
2524
2525 if (s->send_bps)
2526 out(" send %sbps", sprint_bw(b1, s->send_bps));
2527 if (s->lastsnd)
2528 out(" lastsnd:%u", s->lastsnd);
2529 if (s->lastrcv)
2530 out(" lastrcv:%u", s->lastrcv);
2531 if (s->lastack)
2532 out(" lastack:%u", s->lastack);
2533
2534 if (s->pacing_rate) {
2535 out(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
2536 if (s->pacing_rate_max)
2537 out("/%sbps", sprint_bw(b1, s->pacing_rate_max));
2538 }
2539
2540 if (s->delivery_rate)
2541 out(" delivery_rate %sbps", sprint_bw(b1, s->delivery_rate));
2542 if (s->delivered)
2543 out(" delivered:%u", s->delivered);
2544 if (s->delivered_ce)
2545 out(" delivered_ce:%u", s->delivered_ce);
2546 if (s->app_limited)
2547 out(" app_limited");
2548
2549 if (s->busy_time) {
2550 out(" busy:%llums", s->busy_time / 1000);
2551 if (s->rwnd_limited)
2552 out(" rwnd_limited:%llums(%.1f%%)",
2553 s->rwnd_limited / 1000,
2554 100.0 * s->rwnd_limited / s->busy_time);
2555 if (s->sndbuf_limited)
2556 out(" sndbuf_limited:%llums(%.1f%%)",
2557 s->sndbuf_limited / 1000,
2558 100.0 * s->sndbuf_limited / s->busy_time);
2559 }
2560
2561 if (s->unacked)
2562 out(" unacked:%u", s->unacked);
2563 if (s->retrans || s->retrans_total)
2564 out(" retrans:%u/%u", s->retrans, s->retrans_total);
2565 if (s->lost)
2566 out(" lost:%u", s->lost);
2567 if (s->sacked && s->ss.state != SS_LISTEN)
2568 out(" sacked:%u", s->sacked);
2569 if (s->dsack_dups)
2570 out(" dsack_dups:%u", s->dsack_dups);
2571 if (s->fackets)
2572 out(" fackets:%u", s->fackets);
2573 if (s->reordering != 3)
2574 out(" reordering:%d", s->reordering);
2575 if (s->reord_seen)
2576 out(" reord_seen:%d", s->reord_seen);
2577 if (s->rcv_rtt)
2578 out(" rcv_rtt:%g", s->rcv_rtt);
2579 if (s->rcv_space)
2580 out(" rcv_space:%d", s->rcv_space);
2581 if (s->rcv_ssthresh)
2582 out(" rcv_ssthresh:%u", s->rcv_ssthresh);
2583 if (s->not_sent)
2584 out(" notsent:%u", s->not_sent);
2585 if (s->min_rtt)
2586 out(" minrtt:%g", s->min_rtt);
2587 }
2588
2589 static void tcp_timer_print(struct tcpstat *s)
2590 {
2591 static const char * const tmr_name[] = {
2592 "off",
2593 "on",
2594 "keepalive",
2595 "timewait",
2596 "persist",
2597 "unknown"
2598 };
2599
2600 if (s->timer) {
2601 if (s->timer > 4)
2602 s->timer = 5;
2603 out(" timer:(%s,%s,%d)",
2604 tmr_name[s->timer],
2605 print_ms_timer(s->timeout),
2606 s->retrans);
2607 }
2608 }
2609
2610 static void sctp_timer_print(struct tcpstat *s)
2611 {
2612 if (s->timer)
2613 out(" timer:(T3_RTX,%s,%d)",
2614 print_ms_timer(s->timeout), s->retrans);
2615 }
2616
2617 static int tcp_show_line(char *line, const struct filter *f, int family)
2618 {
2619 int rto = 0, ato = 0;
2620 struct tcpstat s = {};
2621 char *loc, *rem, *data;
2622 char opt[256];
2623 int n;
2624 int hz = get_user_hz();
2625
2626 if (proc_inet_split_line(line, &loc, &rem, &data))
2627 return -1;
2628
2629 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2630
2631 if (!(f->states & (1 << state)))
2632 return 0;
2633
2634 proc_parse_inet_addr(loc, rem, family, &s.ss);
2635
2636 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
2637 return 0;
2638
2639 opt[0] = 0;
2640 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %u %d %[^\n]\n",
2641 &s.ss.state, &s.ss.wq, &s.ss.rq,
2642 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
2643 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
2644 &s.ssthresh, opt);
2645
2646 if (n < 17)
2647 opt[0] = 0;
2648
2649 if (n < 12) {
2650 rto = 0;
2651 s.cwnd = 2;
2652 s.ssthresh = -1;
2653 ato = s.qack = 0;
2654 }
2655
2656 s.retrans = s.timer != 1 ? s.probes : s.retrans;
2657 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
2658 s.ato = (double)ato / hz;
2659 s.qack /= 2;
2660 s.rto = (double)rto;
2661 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
2662 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
2663 s.ss.type = IPPROTO_TCP;
2664
2665 inet_stats_print(&s.ss, false);
2666
2667 if (show_options)
2668 tcp_timer_print(&s);
2669
2670 if (show_details) {
2671 sock_details_print(&s.ss);
2672 if (opt[0])
2673 out(" opt:\"%s\"", opt);
2674 }
2675
2676 if (show_tcpinfo)
2677 tcp_stats_print(&s);
2678
2679 return 0;
2680 }
2681
2682 static int generic_record_read(FILE *fp,
2683 int (*worker)(char*, const struct filter *, int),
2684 const struct filter *f, int fam)
2685 {
2686 char line[256];
2687
2688 /* skip header */
2689 if (fgets(line, sizeof(line), fp) == NULL)
2690 goto outerr;
2691
2692 while (fgets(line, sizeof(line), fp) != NULL) {
2693 int n = strlen(line);
2694
2695 if (n == 0 || line[n-1] != '\n') {
2696 errno = -EINVAL;
2697 return -1;
2698 }
2699 line[n-1] = 0;
2700
2701 if (worker(line, f, fam) < 0)
2702 return 0;
2703 }
2704 outerr:
2705
2706 return ferror(fp) ? -1 : 0;
2707 }
2708
2709 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
2710 {
2711 const __u32 *skmeminfo;
2712
2713 if (!tb[attrtype]) {
2714 if (attrtype == INET_DIAG_SKMEMINFO) {
2715 if (!tb[INET_DIAG_MEMINFO])
2716 return;
2717
2718 const struct inet_diag_meminfo *minfo =
2719 RTA_DATA(tb[INET_DIAG_MEMINFO]);
2720
2721 out(" mem:(r%u,w%u,f%u,t%u)",
2722 minfo->idiag_rmem,
2723 minfo->idiag_wmem,
2724 minfo->idiag_fmem,
2725 minfo->idiag_tmem);
2726 }
2727 return;
2728 }
2729
2730 skmeminfo = RTA_DATA(tb[attrtype]);
2731
2732 out(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
2733 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
2734 skmeminfo[SK_MEMINFO_RCVBUF],
2735 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
2736 skmeminfo[SK_MEMINFO_SNDBUF],
2737 skmeminfo[SK_MEMINFO_FWD_ALLOC],
2738 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
2739 skmeminfo[SK_MEMINFO_OPTMEM]);
2740
2741 if (RTA_PAYLOAD(tb[attrtype]) >=
2742 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
2743 out(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
2744
2745 if (RTA_PAYLOAD(tb[attrtype]) >=
2746 (SK_MEMINFO_DROPS + 1) * sizeof(__u32))
2747 out(",d%u", skmeminfo[SK_MEMINFO_DROPS]);
2748
2749 out(")");
2750 }
2751
2752 static void print_md5sig(struct tcp_diag_md5sig *sig)
2753 {
2754 out("%s/%d=",
2755 format_host(sig->tcpm_family,
2756 sig->tcpm_family == AF_INET6 ? 16 : 4,
2757 &sig->tcpm_addr),
2758 sig->tcpm_prefixlen);
2759 print_escape_buf(sig->tcpm_key, sig->tcpm_keylen, " ,");
2760 }
2761
2762 static void tcp_tls_version(struct rtattr *attr)
2763 {
2764 u_int16_t val;
2765
2766 if (!attr)
2767 return;
2768 val = rta_getattr_u16(attr);
2769
2770 switch (val) {
2771 case TLS_1_2_VERSION:
2772 out(" version: 1.2");
2773 break;
2774 case TLS_1_3_VERSION:
2775 out(" version: 1.3");
2776 break;
2777 default:
2778 out(" version: unknown(%hu)", val);
2779 break;
2780 }
2781 }
2782
2783 static void tcp_tls_cipher(struct rtattr *attr)
2784 {
2785 u_int16_t val;
2786
2787 if (!attr)
2788 return;
2789 val = rta_getattr_u16(attr);
2790
2791 switch (val) {
2792 case TLS_CIPHER_AES_GCM_128:
2793 out(" cipher: aes-gcm-128");
2794 break;
2795 case TLS_CIPHER_AES_GCM_256:
2796 out(" cipher: aes-gcm-256");
2797 break;
2798 }
2799 }
2800
2801 static void tcp_tls_conf(const char *name, struct rtattr *attr)
2802 {
2803 u_int16_t val;
2804
2805 if (!attr)
2806 return;
2807 val = rta_getattr_u16(attr);
2808
2809 switch (val) {
2810 case TLS_CONF_BASE:
2811 out(" %s: none", name);
2812 break;
2813 case TLS_CONF_SW:
2814 out(" %s: sw", name);
2815 break;
2816 case TLS_CONF_HW:
2817 out(" %s: hw", name);
2818 break;
2819 case TLS_CONF_HW_RECORD:
2820 out(" %s: hw-record", name);
2821 break;
2822 default:
2823 out(" %s: unknown(%hu)", name, val);
2824 break;
2825 }
2826 }
2827
2828 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
2829
2830 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2831 struct rtattr *tb[])
2832 {
2833 double rtt = 0;
2834 struct tcpstat s = {};
2835
2836 s.ss.state = r->idiag_state;
2837
2838 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2839
2840 if (tb[INET_DIAG_INFO]) {
2841 struct tcp_info *info;
2842 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2843
2844 /* workaround for older kernels with less fields */
2845 if (len < sizeof(*info)) {
2846 info = alloca(sizeof(*info));
2847 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2848 memset((char *)info + len, 0, sizeof(*info) - len);
2849 } else
2850 info = RTA_DATA(tb[INET_DIAG_INFO]);
2851
2852 if (show_options) {
2853 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
2854 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
2855 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
2856 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
2857 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
2858 }
2859
2860 if (tb[INET_DIAG_CONG])
2861 strncpy(s.cong_alg,
2862 rta_getattr_str(tb[INET_DIAG_CONG]),
2863 sizeof(s.cong_alg) - 1);
2864
2865 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
2866 s.has_wscale_opt = true;
2867 s.snd_wscale = info->tcpi_snd_wscale;
2868 s.rcv_wscale = info->tcpi_rcv_wscale;
2869 }
2870
2871 if (info->tcpi_rto && info->tcpi_rto != 3000000)
2872 s.rto = (double)info->tcpi_rto / 1000;
2873
2874 s.backoff = info->tcpi_backoff;
2875 s.rtt = (double)info->tcpi_rtt / 1000;
2876 s.rttvar = (double)info->tcpi_rttvar / 1000;
2877 s.ato = (double)info->tcpi_ato / 1000;
2878 s.mss = info->tcpi_snd_mss;
2879 s.rcv_mss = info->tcpi_rcv_mss;
2880 s.advmss = info->tcpi_advmss;
2881 s.rcv_space = info->tcpi_rcv_space;
2882 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
2883 s.lastsnd = info->tcpi_last_data_sent;
2884 s.lastrcv = info->tcpi_last_data_recv;
2885 s.lastack = info->tcpi_last_ack_recv;
2886 s.unacked = info->tcpi_unacked;
2887 s.retrans = info->tcpi_retrans;
2888 s.retrans_total = info->tcpi_total_retrans;
2889 s.lost = info->tcpi_lost;
2890 s.sacked = info->tcpi_sacked;
2891 s.fackets = info->tcpi_fackets;
2892 s.reordering = info->tcpi_reordering;
2893 s.rcv_ssthresh = info->tcpi_rcv_ssthresh;
2894 s.cwnd = info->tcpi_snd_cwnd;
2895 s.pmtu = info->tcpi_pmtu;
2896
2897 if (info->tcpi_snd_ssthresh < 0xFFFF)
2898 s.ssthresh = info->tcpi_snd_ssthresh;
2899
2900 rtt = (double) info->tcpi_rtt;
2901 if (tb[INET_DIAG_VEGASINFO]) {
2902 const struct tcpvegas_info *vinfo
2903 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
2904
2905 if (vinfo->tcpv_enabled &&
2906 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
2907 rtt = vinfo->tcpv_rtt;
2908 }
2909
2910 if (tb[INET_DIAG_DCTCPINFO]) {
2911 struct dctcpstat *dctcp = malloc(sizeof(struct
2912 dctcpstat));
2913
2914 const struct tcp_dctcp_info *dinfo
2915 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
2916
2917 dctcp->enabled = !!dinfo->dctcp_enabled;
2918 dctcp->ce_state = dinfo->dctcp_ce_state;
2919 dctcp->alpha = dinfo->dctcp_alpha;
2920 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
2921 dctcp->ab_tot = dinfo->dctcp_ab_tot;
2922 s.dctcp = dctcp;
2923 }
2924
2925 if (tb[INET_DIAG_BBRINFO]) {
2926 const void *bbr_info = RTA_DATA(tb[INET_DIAG_BBRINFO]);
2927 int len = min(RTA_PAYLOAD(tb[INET_DIAG_BBRINFO]),
2928 sizeof(*s.bbr_info));
2929
2930 s.bbr_info = calloc(1, sizeof(*s.bbr_info));
2931 if (s.bbr_info && bbr_info)
2932 memcpy(s.bbr_info, bbr_info, len);
2933 }
2934
2935 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
2936 s.send_bps = (double) info->tcpi_snd_cwnd *
2937 (double)info->tcpi_snd_mss * 8000000. / rtt;
2938 }
2939
2940 if (info->tcpi_pacing_rate &&
2941 info->tcpi_pacing_rate != ~0ULL) {
2942 s.pacing_rate = info->tcpi_pacing_rate * 8.;
2943
2944 if (info->tcpi_max_pacing_rate &&
2945 info->tcpi_max_pacing_rate != ~0ULL)
2946 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
2947 }
2948 s.bytes_acked = info->tcpi_bytes_acked;
2949 s.bytes_received = info->tcpi_bytes_received;
2950 s.segs_out = info->tcpi_segs_out;
2951 s.segs_in = info->tcpi_segs_in;
2952 s.data_segs_out = info->tcpi_data_segs_out;
2953 s.data_segs_in = info->tcpi_data_segs_in;
2954 s.not_sent = info->tcpi_notsent_bytes;
2955 if (info->tcpi_min_rtt && info->tcpi_min_rtt != ~0U)
2956 s.min_rtt = (double) info->tcpi_min_rtt / 1000;
2957 s.delivery_rate = info->tcpi_delivery_rate * 8.;
2958 s.app_limited = info->tcpi_delivery_rate_app_limited;
2959 s.busy_time = info->tcpi_busy_time;
2960 s.rwnd_limited = info->tcpi_rwnd_limited;
2961 s.sndbuf_limited = info->tcpi_sndbuf_limited;
2962 s.delivered = info->tcpi_delivered;
2963 s.delivered_ce = info->tcpi_delivered_ce;
2964 s.dsack_dups = info->tcpi_dsack_dups;
2965 s.reord_seen = info->tcpi_reord_seen;
2966 s.bytes_sent = info->tcpi_bytes_sent;
2967 s.bytes_retrans = info->tcpi_bytes_retrans;
2968 tcp_stats_print(&s);
2969 free(s.dctcp);
2970 free(s.bbr_info);
2971 }
2972 if (tb[INET_DIAG_MD5SIG]) {
2973 struct tcp_diag_md5sig *sig = RTA_DATA(tb[INET_DIAG_MD5SIG]);
2974 int len = RTA_PAYLOAD(tb[INET_DIAG_MD5SIG]);
2975
2976 out(" md5keys:");
2977 print_md5sig(sig++);
2978 for (len -= sizeof(*sig); len > 0; len -= sizeof(*sig)) {
2979 out(",");
2980 print_md5sig(sig++);
2981 }
2982 }
2983 if (tb[INET_DIAG_ULP_INFO]) {
2984 struct rtattr *ulpinfo[INET_ULP_INFO_MAX + 1] = { 0 };
2985
2986 parse_rtattr_nested(ulpinfo, INET_ULP_INFO_MAX,
2987 tb[INET_DIAG_ULP_INFO]);
2988
2989 if (ulpinfo[INET_ULP_INFO_NAME])
2990 out(" tcp-ulp-%s",
2991 rta_getattr_str(ulpinfo[INET_ULP_INFO_NAME]));
2992
2993 if (ulpinfo[INET_ULP_INFO_TLS]) {
2994 struct rtattr *tlsinfo[TLS_INFO_MAX + 1] = { 0 };
2995
2996 parse_rtattr_nested(tlsinfo, TLS_INFO_MAX,
2997 ulpinfo[INET_ULP_INFO_TLS]);
2998
2999 tcp_tls_version(tlsinfo[TLS_INFO_VERSION]);
3000 tcp_tls_cipher(tlsinfo[TLS_INFO_CIPHER]);
3001 tcp_tls_conf("rxconf", tlsinfo[TLS_INFO_RXCONF]);
3002 tcp_tls_conf("txconf", tlsinfo[TLS_INFO_TXCONF]);
3003 }
3004 }
3005 }
3006
3007 static const char *format_host_sa(struct sockaddr_storage *sa)
3008 {
3009 union {
3010 struct sockaddr_in sin;
3011 struct sockaddr_in6 sin6;
3012 } *saddr = (void *)sa;
3013
3014 switch (sa->ss_family) {
3015 case AF_INET:
3016 return format_host(AF_INET, 4, &saddr->sin.sin_addr);
3017 case AF_INET6:
3018 return format_host(AF_INET6, 16, &saddr->sin6.sin6_addr);
3019 default:
3020 return "";
3021 }
3022 }
3023
3024 static void sctp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
3025 struct rtattr *tb[])
3026 {
3027 struct sockaddr_storage *sa;
3028 int len;
3029
3030 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
3031
3032 if (tb[INET_DIAG_LOCALS]) {
3033 len = RTA_PAYLOAD(tb[INET_DIAG_LOCALS]);
3034 sa = RTA_DATA(tb[INET_DIAG_LOCALS]);
3035
3036 out(" locals:%s", format_host_sa(sa));
3037 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
3038 out(",%s", format_host_sa(sa));
3039
3040 }
3041 if (tb[INET_DIAG_PEERS]) {
3042 len = RTA_PAYLOAD(tb[INET_DIAG_PEERS]);
3043 sa = RTA_DATA(tb[INET_DIAG_PEERS]);
3044
3045 out(" peers:%s", format_host_sa(sa));
3046 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
3047 out(",%s", format_host_sa(sa));
3048 }
3049 if (tb[INET_DIAG_INFO]) {
3050 struct sctp_info *info;
3051 len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
3052
3053 /* workaround for older kernels with less fields */
3054 if (len < sizeof(*info)) {
3055 info = alloca(sizeof(*info));
3056 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
3057 memset((char *)info + len, 0, sizeof(*info) - len);
3058 } else
3059 info = RTA_DATA(tb[INET_DIAG_INFO]);
3060
3061 sctp_stats_print(info);
3062 }
3063 }
3064
3065 static void parse_diag_msg(struct nlmsghdr *nlh, struct sockstat *s)
3066 {
3067 struct rtattr *tb[INET_DIAG_MAX+1];
3068 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3069
3070 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3071 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3072
3073 s->state = r->idiag_state;
3074 s->local.family = s->remote.family = r->idiag_family;
3075 s->lport = ntohs(r->id.idiag_sport);
3076 s->rport = ntohs(r->id.idiag_dport);
3077 s->wq = r->idiag_wqueue;
3078 s->rq = r->idiag_rqueue;
3079 s->ino = r->idiag_inode;
3080 s->uid = r->idiag_uid;
3081 s->iface = r->id.idiag_if;
3082 s->sk = cookie_sk_get(&r->id.idiag_cookie[0]);
3083
3084 s->mark = 0;
3085 if (tb[INET_DIAG_MARK])
3086 s->mark = rta_getattr_u32(tb[INET_DIAG_MARK]);
3087 if (tb[INET_DIAG_PROTOCOL])
3088 s->raw_prot = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3089 else
3090 s->raw_prot = 0;
3091
3092 if (s->local.family == AF_INET)
3093 s->local.bytelen = s->remote.bytelen = 4;
3094 else
3095 s->local.bytelen = s->remote.bytelen = 16;
3096
3097 memcpy(s->local.data, r->id.idiag_src, s->local.bytelen);
3098 memcpy(s->remote.data, r->id.idiag_dst, s->local.bytelen);
3099 }
3100
3101 static int inet_show_sock(struct nlmsghdr *nlh,
3102 struct sockstat *s)
3103 {
3104 struct rtattr *tb[INET_DIAG_MAX+1];
3105 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3106 unsigned char v6only = 0;
3107
3108 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3109 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3110
3111 if (tb[INET_DIAG_PROTOCOL])
3112 s->type = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3113
3114 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3115 v6only = rta_getattr_u8(tb[INET_DIAG_SKV6ONLY]);
3116
3117 inet_stats_print(s, v6only);
3118
3119 if (show_options) {
3120 struct tcpstat t = {};
3121
3122 t.timer = r->idiag_timer;
3123 t.timeout = r->idiag_expires;
3124 t.retrans = r->idiag_retrans;
3125 if (s->type == IPPROTO_SCTP)
3126 sctp_timer_print(&t);
3127 else
3128 tcp_timer_print(&t);
3129 }
3130
3131 if (show_details) {
3132 sock_details_print(s);
3133 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3134 out(" v6only:%u", v6only);
3135
3136 if (tb[INET_DIAG_SHUTDOWN]) {
3137 unsigned char mask;
3138
3139 mask = rta_getattr_u8(tb[INET_DIAG_SHUTDOWN]);
3140 out(" %c-%c",
3141 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3142 }
3143 }
3144
3145 if (show_tos) {
3146 if (tb[INET_DIAG_TOS])
3147 out(" tos:%#x", rta_getattr_u8(tb[INET_DIAG_TOS]));
3148 if (tb[INET_DIAG_TCLASS])
3149 out(" tclass:%#x", rta_getattr_u8(tb[INET_DIAG_TCLASS]));
3150 if (tb[INET_DIAG_CLASS_ID])
3151 out(" class_id:%#x", rta_getattr_u32(tb[INET_DIAG_CLASS_ID]));
3152 }
3153
3154 if (show_mem || (show_tcpinfo && s->type != IPPROTO_UDP)) {
3155 if (!oneline)
3156 out("\n\t");
3157 if (s->type == IPPROTO_SCTP)
3158 sctp_show_info(nlh, r, tb);
3159 else
3160 tcp_show_info(nlh, r, tb);
3161 }
3162 sctp_ino = s->ino;
3163
3164 return 0;
3165 }
3166
3167 static int tcpdiag_send(int fd, int protocol, struct filter *f)
3168 {
3169 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3170 struct {
3171 struct nlmsghdr nlh;
3172 struct inet_diag_req r;
3173 } req = {
3174 .nlh.nlmsg_len = sizeof(req),
3175 .nlh.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST,
3176 .nlh.nlmsg_seq = MAGIC_SEQ,
3177 .r.idiag_family = AF_INET,
3178 .r.idiag_states = f->states,
3179 };
3180 char *bc = NULL;
3181 int bclen;
3182 struct msghdr msg;
3183 struct rtattr rta;
3184 struct iovec iov[3];
3185 int iovlen = 1;
3186
3187 if (protocol == IPPROTO_UDP)
3188 return -1;
3189
3190 if (protocol == IPPROTO_TCP)
3191 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
3192 else
3193 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
3194 if (show_mem) {
3195 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3196 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3197 }
3198
3199 if (show_tcpinfo) {
3200 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3201 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3202 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3203 }
3204
3205 if (show_tos) {
3206 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3207 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3208 }
3209
3210 iov[0] = (struct iovec){
3211 .iov_base = &req,
3212 .iov_len = sizeof(req)
3213 };
3214 if (f->f) {
3215 bclen = ssfilter_bytecompile(f->f, &bc);
3216 if (bclen) {
3217 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3218 rta.rta_len = RTA_LENGTH(bclen);
3219 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3220 iov[2] = (struct iovec){ bc, bclen };
3221 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3222 iovlen = 3;
3223 }
3224 }
3225
3226 msg = (struct msghdr) {
3227 .msg_name = (void *)&nladdr,
3228 .msg_namelen = sizeof(nladdr),
3229 .msg_iov = iov,
3230 .msg_iovlen = iovlen,
3231 };
3232
3233 if (sendmsg(fd, &msg, 0) < 0) {
3234 close(fd);
3235 return -1;
3236 }
3237
3238 return 0;
3239 }
3240
3241 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
3242 {
3243 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3244 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3245 char *bc = NULL;
3246 int bclen;
3247 struct msghdr msg;
3248 struct rtattr rta;
3249 struct iovec iov[3];
3250 int iovlen = 1;
3251
3252 if (family == PF_UNSPEC)
3253 return tcpdiag_send(fd, protocol, f);
3254
3255 memset(&req.r, 0, sizeof(req.r));
3256 req.r.sdiag_family = family;
3257 req.r.sdiag_protocol = protocol;
3258 req.r.idiag_states = f->states;
3259 if (show_mem) {
3260 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3261 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3262 }
3263
3264 if (show_tcpinfo) {
3265 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3266 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3267 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3268 }
3269
3270 if (show_tos) {
3271 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3272 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3273 }
3274
3275 iov[0] = (struct iovec){
3276 .iov_base = &req,
3277 .iov_len = sizeof(req)
3278 };
3279 if (f->f) {
3280 bclen = ssfilter_bytecompile(f->f, &bc);
3281 if (bclen) {
3282 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3283 rta.rta_len = RTA_LENGTH(bclen);
3284 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3285 iov[2] = (struct iovec){ bc, bclen };
3286 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3287 iovlen = 3;
3288 }
3289 }
3290
3291 msg = (struct msghdr) {
3292 .msg_name = (void *)&nladdr,
3293 .msg_namelen = sizeof(nladdr),
3294 .msg_iov = iov,
3295 .msg_iovlen = iovlen,
3296 };
3297
3298 if (sendmsg(fd, &msg, 0) < 0) {
3299 close(fd);
3300 return -1;
3301 }
3302
3303 return 0;
3304 }
3305
3306 struct inet_diag_arg {
3307 struct filter *f;
3308 int protocol;
3309 struct rtnl_handle *rth;
3310 };
3311
3312 static int kill_inet_sock(struct nlmsghdr *h, void *arg, struct sockstat *s)
3313 {
3314 struct inet_diag_msg *d = NLMSG_DATA(h);
3315 struct inet_diag_arg *diag_arg = arg;
3316 struct rtnl_handle *rth = diag_arg->rth;
3317
3318 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3319
3320 req.nlh.nlmsg_type = SOCK_DESTROY;
3321 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
3322 req.nlh.nlmsg_seq = ++rth->seq;
3323 req.r.sdiag_family = d->idiag_family;
3324 req.r.sdiag_protocol = diag_arg->protocol;
3325 req.r.id = d->id;
3326
3327 if (diag_arg->protocol == IPPROTO_RAW) {
3328 struct inet_diag_req_raw *raw = (void *)&req.r;
3329
3330 BUILD_BUG_ON(sizeof(req.r) != sizeof(*raw));
3331 raw->sdiag_raw_protocol = s->raw_prot;
3332 }
3333
3334 return rtnl_talk(rth, &req.nlh, NULL);
3335 }
3336
3337 static int show_one_inet_sock(struct nlmsghdr *h, void *arg)
3338 {
3339 int err;
3340 struct inet_diag_arg *diag_arg = arg;
3341 struct inet_diag_msg *r = NLMSG_DATA(h);
3342 struct sockstat s = {};
3343
3344 if (!(diag_arg->f->families & FAMILY_MASK(r->idiag_family)))
3345 return 0;
3346
3347 parse_diag_msg(h, &s);
3348 s.type = diag_arg->protocol;
3349
3350 if (diag_arg->f->f && run_ssfilter(diag_arg->f->f, &s) == 0)
3351 return 0;
3352
3353 if (diag_arg->f->kill && kill_inet_sock(h, arg, &s) != 0) {
3354 if (errno == EOPNOTSUPP || errno == ENOENT) {
3355 /* Socket can't be closed, or is already closed. */
3356 return 0;
3357 } else {
3358 perror("SOCK_DESTROY answers");
3359 return -1;
3360 }
3361 }
3362
3363 err = inet_show_sock(h, &s);
3364 if (err < 0)
3365 return err;
3366
3367 return 0;
3368 }
3369
3370 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
3371 {
3372 int err = 0;
3373 struct rtnl_handle rth, rth2;
3374 int family = PF_INET;
3375 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
3376
3377 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3378 return -1;
3379
3380 if (f->kill) {
3381 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
3382 rtnl_close(&rth);
3383 return -1;
3384 }
3385 arg.rth = &rth2;
3386 }
3387
3388 rth.dump = MAGIC_SEQ;
3389 rth.dump_fp = dump_fp;
3390 if (preferred_family == PF_INET6)
3391 family = PF_INET6;
3392
3393 again:
3394 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
3395 goto Exit;
3396
3397 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
3398 if (family != PF_UNSPEC) {
3399 family = PF_UNSPEC;
3400 goto again;
3401 }
3402 goto Exit;
3403 }
3404 if (family == PF_INET && preferred_family != PF_INET) {
3405 family = PF_INET6;
3406 goto again;
3407 }
3408
3409 Exit:
3410 rtnl_close(&rth);
3411 if (arg.rth)
3412 rtnl_close(arg.rth);
3413 return err;
3414 }
3415
3416 static int tcp_show_netlink_file(struct filter *f)
3417 {
3418 FILE *fp;
3419 char buf[16384];
3420 int err = -1;
3421
3422 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
3423 perror("fopen($TCPDIAG_FILE)");
3424 return err;
3425 }
3426
3427 while (1) {
3428 int err2;
3429 size_t status, nitems;
3430 struct nlmsghdr *h = (struct nlmsghdr *)buf;
3431 struct sockstat s = {};
3432
3433 status = fread(buf, 1, sizeof(*h), fp);
3434 if (status != sizeof(*h)) {
3435 if (ferror(fp))
3436 perror("Reading header from $TCPDIAG_FILE");
3437 if (feof(fp))
3438 fprintf(stderr, "Unexpected EOF reading $TCPDIAG_FILE");
3439 break;
3440 }
3441
3442 nitems = NLMSG_ALIGN(h->nlmsg_len - sizeof(*h));
3443 status = fread(h+1, 1, nitems, fp);
3444
3445 if (status != nitems) {
3446 if (ferror(fp))
3447 perror("Reading $TCPDIAG_FILE");
3448 if (feof(fp))
3449 fprintf(stderr, "Unexpected EOF reading $TCPDIAG_FILE");
3450 break;
3451 }
3452
3453 /* The only legal exit point */
3454 if (h->nlmsg_type == NLMSG_DONE) {
3455 err = 0;
3456 break;
3457 }
3458
3459 if (h->nlmsg_type == NLMSG_ERROR) {
3460 struct nlmsgerr *err = (struct nlmsgerr *)NLMSG_DATA(h);
3461
3462 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
3463 fprintf(stderr, "ERROR truncated\n");
3464 } else {
3465 errno = -err->error;
3466 perror("TCPDIAG answered");
3467 }
3468 break;
3469 }
3470
3471 parse_diag_msg(h, &s);
3472 s.type = IPPROTO_TCP;
3473
3474 if (f && f->f && run_ssfilter(f->f, &s) == 0)
3475 continue;
3476
3477 err2 = inet_show_sock(h, &s);
3478 if (err2 < 0) {
3479 err = err2;
3480 break;
3481 }
3482 }
3483
3484 fclose(fp);
3485 return err;
3486 }
3487
3488 static int tcp_show(struct filter *f)
3489 {
3490 FILE *fp = NULL;
3491 char *buf = NULL;
3492 int bufsize = 1024*1024;
3493
3494 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3495 return 0;
3496
3497 dg_proto = TCP_PROTO;
3498
3499 if (getenv("TCPDIAG_FILE"))
3500 return tcp_show_netlink_file(f);
3501
3502 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
3503 && inet_show_netlink(f, NULL, IPPROTO_TCP) == 0)
3504 return 0;
3505
3506 /* Sigh... We have to parse /proc/net/tcp... */
3507 while (bufsize >= 64*1024) {
3508 if ((buf = malloc(bufsize)) != NULL)
3509 break;
3510 bufsize /= 2;
3511 }
3512 if (buf == NULL) {
3513 errno = ENOMEM;
3514 return -1;
3515 }
3516
3517 if (f->families & FAMILY_MASK(AF_INET)) {
3518 if ((fp = net_tcp_open()) == NULL)
3519 goto outerr;
3520
3521 setbuffer(fp, buf, bufsize);
3522 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
3523 goto outerr;
3524 fclose(fp);
3525 }
3526
3527 if ((f->families & FAMILY_MASK(AF_INET6)) &&
3528 (fp = net_tcp6_open()) != NULL) {
3529 setbuffer(fp, buf, bufsize);
3530 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
3531 goto outerr;
3532 fclose(fp);
3533 }
3534
3535 free(buf);
3536 return 0;
3537
3538 outerr:
3539 do {
3540 int saved_errno = errno;
3541
3542 free(buf);
3543 if (fp)
3544 fclose(fp);
3545 errno = saved_errno;
3546 return -1;
3547 } while (0);
3548 }
3549
3550 static int dccp_show(struct filter *f)
3551 {
3552 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3553 return 0;
3554
3555 if (!getenv("PROC_NET_DCCP") && !getenv("PROC_ROOT")
3556 && inet_show_netlink(f, NULL, IPPROTO_DCCP) == 0)
3557 return 0;
3558
3559 return 0;
3560 }
3561
3562 static int sctp_show(struct filter *f)
3563 {
3564 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3565 return 0;
3566
3567 if (!getenv("PROC_NET_SCTP") && !getenv("PROC_ROOT")
3568 && inet_show_netlink(f, NULL, IPPROTO_SCTP) == 0)
3569 return 0;
3570
3571 return 0;
3572 }
3573
3574 static int dgram_show_line(char *line, const struct filter *f, int family)
3575 {
3576 struct sockstat s = {};
3577 char *loc, *rem, *data;
3578 char opt[256];
3579 int n;
3580
3581 if (proc_inet_split_line(line, &loc, &rem, &data))
3582 return -1;
3583
3584 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
3585
3586 if (!(f->states & (1 << state)))
3587 return 0;
3588
3589 proc_parse_inet_addr(loc, rem, family, &s);
3590
3591 if (f->f && run_ssfilter(f->f, &s) == 0)
3592 return 0;
3593
3594 opt[0] = 0;
3595 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
3596 &s.state, &s.wq, &s.rq,
3597 &s.uid, &s.ino,
3598 &s.refcnt, &s.sk, opt);
3599
3600 if (n < 9)
3601 opt[0] = 0;
3602
3603 s.type = dg_proto == UDP_PROTO ? IPPROTO_UDP : 0;
3604 inet_stats_print(&s, false);
3605
3606 if (show_details && opt[0])
3607 out(" opt:\"%s\"", opt);
3608
3609 return 0;
3610 }
3611
3612 static int udp_show(struct filter *f)
3613 {
3614 FILE *fp = NULL;
3615
3616 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3617 return 0;
3618
3619 dg_proto = UDP_PROTO;
3620
3621 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
3622 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
3623 return 0;
3624
3625 if (f->families&FAMILY_MASK(AF_INET)) {
3626 if ((fp = net_udp_open()) == NULL)
3627 goto outerr;
3628 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3629 goto outerr;
3630 fclose(fp);
3631 }
3632
3633 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3634 (fp = net_udp6_open()) != NULL) {
3635 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3636 goto outerr;
3637 fclose(fp);
3638 }
3639 return 0;
3640
3641 outerr:
3642 do {
3643 int saved_errno = errno;
3644
3645 if (fp)
3646 fclose(fp);
3647 errno = saved_errno;
3648 return -1;
3649 } while (0);
3650 }
3651
3652 static int raw_show(struct filter *f)
3653 {
3654 FILE *fp = NULL;
3655
3656 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3657 return 0;
3658
3659 dg_proto = RAW_PROTO;
3660
3661 if (!getenv("PROC_NET_RAW") && !getenv("PROC_ROOT") &&
3662 inet_show_netlink(f, NULL, IPPROTO_RAW) == 0)
3663 return 0;
3664
3665 if (f->families&FAMILY_MASK(AF_INET)) {
3666 if ((fp = net_raw_open()) == NULL)
3667 goto outerr;
3668 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3669 goto outerr;
3670 fclose(fp);
3671 }
3672
3673 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3674 (fp = net_raw6_open()) != NULL) {
3675 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3676 goto outerr;
3677 fclose(fp);
3678 }
3679 return 0;
3680
3681 outerr:
3682 do {
3683 int saved_errno = errno;
3684
3685 if (fp)
3686 fclose(fp);
3687 errno = saved_errno;
3688 return -1;
3689 } while (0);
3690 }
3691
3692 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
3693
3694 static void unix_list_drop_first(struct sockstat **list)
3695 {
3696 struct sockstat *s = *list;
3697
3698 (*list) = (*list)->next;
3699 free(s->name);
3700 free(s);
3701 }
3702
3703 static bool unix_type_skip(struct sockstat *s, struct filter *f)
3704 {
3705 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
3706 return true;
3707 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
3708 return true;
3709 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
3710 return true;
3711 return false;
3712 }
3713
3714 static void unix_stats_print(struct sockstat *s, struct filter *f)
3715 {
3716 char port_name[30] = {};
3717
3718 sock_state_print(s);
3719
3720 sock_addr_print(s->name ?: "*", " ",
3721 int_to_str(s->lport, port_name), NULL);
3722 sock_addr_print(s->peer_name ?: "*", " ",
3723 int_to_str(s->rport, port_name), NULL);
3724
3725 proc_ctx_print(s);
3726 }
3727
3728 static int unix_show_sock(struct nlmsghdr *nlh, void *arg)
3729 {
3730 struct filter *f = (struct filter *)arg;
3731 struct unix_diag_msg *r = NLMSG_DATA(nlh);
3732 struct rtattr *tb[UNIX_DIAG_MAX+1];
3733 char name[128];
3734 struct sockstat stat = { .name = "*", .peer_name = "*" };
3735
3736 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr *)(r+1),
3737 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3738
3739 stat.type = r->udiag_type;
3740 stat.state = r->udiag_state;
3741 stat.ino = stat.lport = r->udiag_ino;
3742 stat.local.family = stat.remote.family = AF_UNIX;
3743
3744 if (unix_type_skip(&stat, f))
3745 return 0;
3746
3747 if (tb[UNIX_DIAG_RQLEN]) {
3748 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
3749
3750 stat.rq = rql->udiag_rqueue;
3751 stat.wq = rql->udiag_wqueue;
3752 }
3753 if (tb[UNIX_DIAG_NAME]) {
3754 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
3755
3756 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
3757 name[len] = '\0';
3758 if (name[0] == '\0') {
3759 int i;
3760 for (i = 0; i < len; i++)
3761 if (name[i] == '\0')
3762 name[i] = '@';
3763 }
3764 stat.name = &name[0];
3765 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
3766 }
3767 if (tb[UNIX_DIAG_PEER])
3768 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
3769
3770 if (f->f && run_ssfilter(f->f, &stat) == 0)
3771 return 0;
3772
3773 unix_stats_print(&stat, f);
3774
3775 if (show_mem)
3776 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
3777 if (show_details) {
3778 if (tb[UNIX_DIAG_SHUTDOWN]) {
3779 unsigned char mask;
3780
3781 mask = rta_getattr_u8(tb[UNIX_DIAG_SHUTDOWN]);
3782 out(" %c-%c",
3783 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3784 }
3785 if (tb[UNIX_DIAG_VFS]) {
3786 struct unix_diag_vfs *uv = RTA_DATA(tb[UNIX_DIAG_VFS]);
3787
3788 out(" ino:%u dev:%u/%u", uv->udiag_vfs_ino, major(uv->udiag_vfs_dev),
3789 minor(uv->udiag_vfs_dev));
3790 }
3791 if (tb[UNIX_DIAG_ICONS]) {
3792 int len = RTA_PAYLOAD(tb[UNIX_DIAG_ICONS]);
3793 __u32 *peers = RTA_DATA(tb[UNIX_DIAG_ICONS]);
3794 int i;
3795
3796 out(" peers:");
3797 for (i = 0; i < len / sizeof(__u32); i++)
3798 out(" %u", peers[i]);
3799 }
3800 }
3801
3802 return 0;
3803 }
3804
3805 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
3806 size_t size, rtnl_filter_t show_one_sock)
3807 {
3808 int ret = -1;
3809 struct rtnl_handle rth;
3810
3811 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3812 return -1;
3813
3814 rth.dump = MAGIC_SEQ;
3815
3816 if (rtnl_send(&rth, req, size) < 0)
3817 goto Exit;
3818
3819 if (rtnl_dump_filter(&rth, show_one_sock, f))
3820 goto Exit;
3821
3822 ret = 0;
3823 Exit:
3824 rtnl_close(&rth);
3825 return ret;
3826 }
3827
3828 static int unix_show_netlink(struct filter *f)
3829 {
3830 DIAG_REQUEST(req, struct unix_diag_req r);
3831
3832 req.r.sdiag_family = AF_UNIX;
3833 req.r.udiag_states = f->states;
3834 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
3835 if (show_mem)
3836 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
3837 if (show_details)
3838 req.r.udiag_show |= UDIAG_SHOW_VFS | UDIAG_SHOW_ICONS;
3839
3840 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
3841 }
3842
3843 static int unix_show(struct filter *f)
3844 {
3845 FILE *fp;
3846 char buf[256];
3847 char name[128];
3848 int newformat = 0;
3849 int cnt;
3850 struct sockstat *list = NULL;
3851 const int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
3852 SS_ESTABLISHED, SS_CLOSING };
3853
3854 if (!filter_af_get(f, AF_UNIX))
3855 return 0;
3856
3857 if (!getenv("PROC_NET_UNIX") && !getenv("PROC_ROOT")
3858 && unix_show_netlink(f) == 0)
3859 return 0;
3860
3861 if ((fp = net_unix_open()) == NULL)
3862 return -1;
3863 if (!fgets(buf, sizeof(buf), fp)) {
3864 fclose(fp);
3865 return -1;
3866 }
3867
3868 if (memcmp(buf, "Peer", 4) == 0)
3869 newformat = 1;
3870 cnt = 0;
3871
3872 while (fgets(buf, sizeof(buf), fp)) {
3873 struct sockstat *u, **insp;
3874 int flags;
3875
3876 if (!(u = calloc(1, sizeof(*u))))
3877 break;
3878
3879 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
3880 &u->rport, &u->rq, &u->wq, &flags, &u->type,
3881 &u->state, &u->ino, name) < 8)
3882 name[0] = 0;
3883
3884 u->lport = u->ino;
3885 u->local.family = u->remote.family = AF_UNIX;
3886
3887 if (flags & (1 << 16)) {
3888 u->state = SS_LISTEN;
3889 } else if (u->state > 0 &&
3890 u->state <= ARRAY_SIZE(unix_state_map)) {
3891 u->state = unix_state_map[u->state-1];
3892 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
3893 u->state = SS_ESTABLISHED;
3894 }
3895 if (unix_type_skip(u, f) ||
3896 !(f->states & (1 << u->state))) {
3897 free(u);
3898 continue;
3899 }
3900
3901 if (!newformat) {
3902 u->rport = 0;
3903 u->rq = 0;
3904 u->wq = 0;
3905 }
3906
3907 if (name[0]) {
3908 u->name = strdup(name);
3909 if (!u->name) {
3910 free(u);
3911 break;
3912 }
3913 }
3914
3915 if (u->rport) {
3916 struct sockstat *p;
3917
3918 for (p = list; p; p = p->next) {
3919 if (u->rport == p->lport)
3920 break;
3921 }
3922 if (!p)
3923 u->peer_name = "?";
3924 else
3925 u->peer_name = p->name ? : "*";
3926 }
3927
3928 if (f->f) {
3929 struct sockstat st = {
3930 .local.family = AF_UNIX,
3931 .remote.family = AF_UNIX,
3932 };
3933
3934 memcpy(st.local.data, &u->name, sizeof(u->name));
3935 /* when parsing the old format rport is set to 0 and
3936 * therefore peer_name remains NULL
3937 */
3938 if (u->peer_name && strcmp(u->peer_name, "*"))
3939 memcpy(st.remote.data, &u->peer_name,
3940 sizeof(u->peer_name));
3941 if (run_ssfilter(f->f, &st) == 0) {
3942 free(u->name);
3943 free(u);
3944 continue;
3945 }
3946 }
3947
3948 insp = &list;
3949 while (*insp) {
3950 if (u->type < (*insp)->type ||
3951 (u->type == (*insp)->type &&
3952 u->ino < (*insp)->ino))
3953 break;
3954 insp = &(*insp)->next;
3955 }
3956 u->next = *insp;
3957 *insp = u;
3958
3959 if (++cnt > MAX_UNIX_REMEMBER) {
3960 while (list) {
3961 unix_stats_print(list, f);
3962 unix_list_drop_first(&list);
3963 }
3964 cnt = 0;
3965 }
3966 }
3967 fclose(fp);
3968 while (list) {
3969 unix_stats_print(list, f);
3970 unix_list_drop_first(&list);
3971 }
3972
3973 return 0;
3974 }
3975
3976 static int packet_stats_print(struct sockstat *s, const struct filter *f)
3977 {
3978 const char *addr, *port;
3979 char ll_name[16];
3980
3981 s->local.family = s->remote.family = AF_PACKET;
3982
3983 if (f->f) {
3984 s->local.data[0] = s->prot;
3985 if (run_ssfilter(f->f, s) == 0)
3986 return 1;
3987 }
3988
3989 sock_state_print(s);
3990
3991 if (s->prot == 3)
3992 addr = "*";
3993 else
3994 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
3995
3996 if (s->iface == 0)
3997 port = "*";
3998 else
3999 port = xll_index_to_name(s->iface);
4000
4001 sock_addr_print(addr, ":", port, NULL);
4002 sock_addr_print("", "*", "", NULL);
4003
4004 proc_ctx_print(s);
4005
4006 if (show_details)
4007 sock_details_print(s);
4008
4009 return 0;
4010 }
4011
4012 static void packet_show_ring(struct packet_diag_ring *ring)
4013 {
4014 out("blk_size:%d", ring->pdr_block_size);
4015 out(",blk_nr:%d", ring->pdr_block_nr);
4016 out(",frm_size:%d", ring->pdr_frame_size);
4017 out(",frm_nr:%d", ring->pdr_frame_nr);
4018 out(",tmo:%d", ring->pdr_retire_tmo);
4019 out(",features:0x%x", ring->pdr_features);
4020 }
4021
4022 static int packet_show_sock(struct nlmsghdr *nlh, void *arg)
4023 {
4024 const struct filter *f = arg;
4025 struct packet_diag_msg *r = NLMSG_DATA(nlh);
4026 struct packet_diag_info *pinfo = NULL;
4027 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
4028 struct rtattr *tb[PACKET_DIAG_MAX+1];
4029 struct sockstat stat = {};
4030 uint32_t fanout = 0;
4031 bool has_fanout = false;
4032
4033 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr *)(r+1),
4034 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4035
4036 /* use /proc/net/packet if all info are not available */
4037 if (!tb[PACKET_DIAG_MEMINFO])
4038 return -1;
4039
4040 stat.type = r->pdiag_type;
4041 stat.prot = r->pdiag_num;
4042 stat.ino = r->pdiag_ino;
4043 stat.state = SS_CLOSE;
4044 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
4045
4046 if (tb[PACKET_DIAG_MEMINFO]) {
4047 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
4048
4049 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4050 }
4051
4052 if (tb[PACKET_DIAG_INFO]) {
4053 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
4054 stat.lport = stat.iface = pinfo->pdi_index;
4055 }
4056
4057 if (tb[PACKET_DIAG_UID])
4058 stat.uid = rta_getattr_u32(tb[PACKET_DIAG_UID]);
4059
4060 if (tb[PACKET_DIAG_RX_RING])
4061 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
4062
4063 if (tb[PACKET_DIAG_TX_RING])
4064 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
4065
4066 if (tb[PACKET_DIAG_FANOUT]) {
4067 has_fanout = true;
4068 fanout = rta_getattr_u32(tb[PACKET_DIAG_FANOUT]);
4069 }
4070
4071 if (packet_stats_print(&stat, f))
4072 return 0;
4073
4074 if (show_details) {
4075 if (pinfo) {
4076 if (oneline)
4077 out(" ver:%d", pinfo->pdi_version);
4078 else
4079 out("\n\tver:%d", pinfo->pdi_version);
4080 out(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
4081 out(" flags( ");
4082 if (pinfo->pdi_flags & PDI_RUNNING)
4083 out("running");
4084 if (pinfo->pdi_flags & PDI_AUXDATA)
4085 out(" auxdata");
4086 if (pinfo->pdi_flags & PDI_ORIGDEV)
4087 out(" origdev");
4088 if (pinfo->pdi_flags & PDI_VNETHDR)
4089 out(" vnethdr");
4090 if (pinfo->pdi_flags & PDI_LOSS)
4091 out(" loss");
4092 if (!pinfo->pdi_flags)
4093 out("0");
4094 out(" )");
4095 }
4096 if (ring_rx) {
4097 if (oneline)
4098 out(" ring_rx(");
4099 else
4100 out("\n\tring_rx(");
4101 packet_show_ring(ring_rx);
4102 out(")");
4103 }
4104 if (ring_tx) {
4105 if (oneline)
4106 out(" ring_tx(");
4107 else
4108 out("\n\tring_tx(");
4109 packet_show_ring(ring_tx);
4110 out(")");
4111 }
4112 if (has_fanout) {
4113 uint16_t type = (fanout >> 16) & 0xffff;
4114
4115 if (oneline)
4116 out(" fanout(");
4117 else
4118 out("\n\tfanout(");
4119 out("id:%d,", fanout & 0xffff);
4120 out("type:");
4121
4122 if (type == 0)
4123 out("hash");
4124 else if (type == 1)
4125 out("lb");
4126 else if (type == 2)
4127 out("cpu");
4128 else if (type == 3)
4129 out("roll");
4130 else if (type == 4)
4131 out("random");
4132 else if (type == 5)
4133 out("qm");
4134 else
4135 out("0x%x", type);
4136
4137 out(")");
4138 }
4139 }
4140
4141 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
4142 struct sock_filter *fil =
4143 RTA_DATA(tb[PACKET_DIAG_FILTER]);
4144 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
4145 sizeof(struct sock_filter);
4146
4147 if (oneline)
4148 out(" bpf filter (%d): ", num);
4149 else
4150 out("\n\tbpf filter (%d): ", num);
4151 while (num) {
4152 out(" 0x%02x %u %u %u,",
4153 fil->code, fil->jt, fil->jf, fil->k);
4154 num--;
4155 fil++;
4156 }
4157 }
4158
4159 if (show_mem)
4160 print_skmeminfo(tb, PACKET_DIAG_MEMINFO);
4161 return 0;
4162 }
4163
4164 static int packet_show_netlink(struct filter *f)
4165 {
4166 DIAG_REQUEST(req, struct packet_diag_req r);
4167
4168 req.r.sdiag_family = AF_PACKET;
4169 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
4170 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
4171
4172 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
4173 }
4174
4175 static int packet_show_line(char *buf, const struct filter *f, int fam)
4176 {
4177 unsigned long long sk;
4178 struct sockstat stat = {};
4179 int type, prot, iface, state, rq, uid, ino;
4180
4181 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
4182 &sk,
4183 &type, &prot, &iface, &state,
4184 &rq, &uid, &ino);
4185
4186 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
4187 return 0;
4188 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
4189 return 0;
4190
4191 stat.type = type;
4192 stat.prot = prot;
4193 stat.lport = stat.iface = iface;
4194 stat.state = state;
4195 stat.rq = rq;
4196 stat.uid = uid;
4197 stat.ino = ino;
4198 stat.state = SS_CLOSE;
4199
4200 if (packet_stats_print(&stat, f))
4201 return 0;
4202
4203 return 0;
4204 }
4205
4206 static int packet_show(struct filter *f)
4207 {
4208 FILE *fp;
4209 int rc = 0;
4210
4211 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
4212 return 0;
4213
4214 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
4215 packet_show_netlink(f) == 0)
4216 return 0;
4217
4218 if ((fp = net_packet_open()) == NULL)
4219 return -1;
4220 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
4221 rc = -1;
4222
4223 fclose(fp);
4224 return rc;
4225 }
4226
4227 static int xdp_stats_print(struct sockstat *s, const struct filter *f)
4228 {
4229 const char *addr, *port;
4230 char q_str[16];
4231
4232 s->local.family = s->remote.family = AF_XDP;
4233
4234 if (f->f) {
4235 if (run_ssfilter(f->f, s) == 0)
4236 return 1;
4237 }
4238
4239 sock_state_print(s);
4240
4241 if (s->iface) {
4242 addr = xll_index_to_name(s->iface);
4243 snprintf(q_str, sizeof(q_str), "q%d", s->lport);
4244 port = q_str;
4245 sock_addr_print(addr, ":", port, NULL);
4246 } else {
4247 sock_addr_print("", "*", "", NULL);
4248 }
4249
4250 sock_addr_print("", "*", "", NULL);
4251
4252 proc_ctx_print(s);
4253
4254 if (show_details)
4255 sock_details_print(s);
4256
4257 return 0;
4258 }
4259
4260 static void xdp_show_ring(const char *name, struct xdp_diag_ring *ring)
4261 {
4262 if (oneline)
4263 out(" %s(", name);
4264 else
4265 out("\n\t%s(", name);
4266 out("entries:%u", ring->entries);
4267 out(")");
4268 }
4269
4270 static void xdp_show_umem(struct xdp_diag_umem *umem, struct xdp_diag_ring *fr,
4271 struct xdp_diag_ring *cr)
4272 {
4273 if (oneline)
4274 out(" tumem(");
4275 else
4276 out("\n\tumem(");
4277 out("id:%u", umem->id);
4278 out(",size:%llu", umem->size);
4279 out(",num_pages:%u", umem->num_pages);
4280 out(",chunk_size:%u", umem->chunk_size);
4281 out(",headroom:%u", umem->headroom);
4282 out(",ifindex:%u", umem->ifindex);
4283 out(",qid:%u", umem->queue_id);
4284 out(",zc:%u", umem->flags & XDP_DU_F_ZEROCOPY);
4285 out(",refs:%u", umem->refs);
4286 out(")");
4287
4288 if (fr)
4289 xdp_show_ring("fr", fr);
4290 if (cr)
4291 xdp_show_ring("cr", cr);
4292 }
4293
4294 static int xdp_show_sock(struct nlmsghdr *nlh, void *arg)
4295 {
4296 struct xdp_diag_ring *rx = NULL, *tx = NULL, *fr = NULL, *cr = NULL;
4297 struct xdp_diag_msg *msg = NLMSG_DATA(nlh);
4298 struct rtattr *tb[XDP_DIAG_MAX + 1];
4299 struct xdp_diag_info *info = NULL;
4300 struct xdp_diag_umem *umem = NULL;
4301 const struct filter *f = arg;
4302 struct sockstat stat = {};
4303
4304 parse_rtattr(tb, XDP_DIAG_MAX, (struct rtattr *)(msg + 1),
4305 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*msg)));
4306
4307 stat.type = msg->xdiag_type;
4308 stat.ino = msg->xdiag_ino;
4309 stat.state = SS_CLOSE;
4310 stat.sk = cookie_sk_get(&msg->xdiag_cookie[0]);
4311
4312 if (tb[XDP_DIAG_INFO]) {
4313 info = RTA_DATA(tb[XDP_DIAG_INFO]);
4314 stat.iface = info->ifindex;
4315 stat.lport = info->queue_id;
4316 }
4317
4318 if (tb[XDP_DIAG_UID])
4319 stat.uid = rta_getattr_u32(tb[XDP_DIAG_UID]);
4320 if (tb[XDP_DIAG_RX_RING])
4321 rx = RTA_DATA(tb[XDP_DIAG_RX_RING]);
4322 if (tb[XDP_DIAG_TX_RING])
4323 tx = RTA_DATA(tb[XDP_DIAG_TX_RING]);
4324 if (tb[XDP_DIAG_UMEM])
4325 umem = RTA_DATA(tb[XDP_DIAG_UMEM]);
4326 if (tb[XDP_DIAG_UMEM_FILL_RING])
4327 fr = RTA_DATA(tb[XDP_DIAG_UMEM_FILL_RING]);
4328 if (tb[XDP_DIAG_UMEM_COMPLETION_RING])
4329 cr = RTA_DATA(tb[XDP_DIAG_UMEM_COMPLETION_RING]);
4330 if (tb[XDP_DIAG_MEMINFO]) {
4331 __u32 *skmeminfo = RTA_DATA(tb[XDP_DIAG_MEMINFO]);
4332
4333 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4334 }
4335
4336 if (xdp_stats_print(&stat, f))
4337 return 0;
4338
4339 if (show_details) {
4340 if (rx)
4341 xdp_show_ring("rx", rx);
4342 if (tx)
4343 xdp_show_ring("tx", tx);
4344 if (umem)
4345 xdp_show_umem(umem, fr, cr);
4346 }
4347
4348 if (show_mem)
4349 print_skmeminfo(tb, XDP_DIAG_MEMINFO); // really?
4350
4351
4352 return 0;
4353 }
4354
4355 static int xdp_show(struct filter *f)
4356 {
4357 DIAG_REQUEST(req, struct xdp_diag_req r);
4358
4359 if (!filter_af_get(f, AF_XDP) || !(f->states & (1 << SS_CLOSE)))
4360 return 0;
4361
4362 req.r.sdiag_family = AF_XDP;
4363 req.r.xdiag_show = XDP_SHOW_INFO | XDP_SHOW_RING_CFG | XDP_SHOW_UMEM |
4364 XDP_SHOW_MEMINFO;
4365
4366 return handle_netlink_request(f, &req.nlh, sizeof(req), xdp_show_sock);
4367 }
4368
4369 static int netlink_show_one(struct filter *f,
4370 int prot, int pid, unsigned int groups,
4371 int state, int dst_pid, unsigned int dst_group,
4372 int rq, int wq,
4373 unsigned long long sk, unsigned long long cb)
4374 {
4375 struct sockstat st = {
4376 .state = SS_CLOSE,
4377 .rq = rq,
4378 .wq = wq,
4379 .local.family = AF_NETLINK,
4380 .remote.family = AF_NETLINK,
4381 };
4382
4383 SPRINT_BUF(prot_buf) = {};
4384 const char *prot_name;
4385 char procname[64] = {};
4386
4387 if (f->f) {
4388 st.rport = -1;
4389 st.lport = pid;
4390 st.local.data[0] = prot;
4391 if (run_ssfilter(f->f, &st) == 0)
4392 return 1;
4393 }
4394
4395 sock_state_print(&st);
4396
4397 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
4398
4399 if (pid == -1) {
4400 procname[0] = '*';
4401 } else if (!numeric) {
4402 int done = 0;
4403
4404 if (!pid) {
4405 done = 1;
4406 strncpy(procname, "kernel", 7);
4407 } else if (pid > 0) {
4408 FILE *fp;
4409
4410 snprintf(procname, sizeof(procname), "%s/%d/stat",
4411 getenv("PROC_ROOT") ? : "/proc", pid);
4412 if ((fp = fopen(procname, "r")) != NULL) {
4413 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
4414 snprintf(procname+strlen(procname),
4415 sizeof(procname)-strlen(procname),
4416 "/%d", pid);
4417 done = 1;
4418 }
4419 fclose(fp);
4420 }
4421 }
4422 if (!done)
4423 int_to_str(pid, procname);
4424 } else {
4425 int_to_str(pid, procname);
4426 }
4427
4428 sock_addr_print(prot_name, ":", procname, NULL);
4429
4430 if (state == NETLINK_CONNECTED) {
4431 char dst_group_buf[30];
4432 char dst_pid_buf[30];
4433
4434 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
4435 int_to_str(dst_pid, dst_pid_buf), NULL);
4436 } else {
4437 sock_addr_print("", "*", "", NULL);
4438 }
4439
4440 char *pid_context = NULL;
4441
4442 if (show_proc_ctx) {
4443 /* The pid value will either be:
4444 * 0 if destination kernel - show kernel initial context.
4445 * A valid process pid - use getpidcon.
4446 * A unique value allocated by the kernel or netlink user
4447 * to the process - show context as "not available".
4448 */
4449 if (!pid)
4450 security_get_initial_context("kernel", &pid_context);
4451 else if (pid > 0)
4452 getpidcon(pid, &pid_context);
4453
4454 out(" proc_ctx=%s", pid_context ? : "unavailable");
4455 free(pid_context);
4456 }
4457
4458 if (show_details) {
4459 out(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
4460 }
4461
4462 return 0;
4463 }
4464
4465 static int netlink_show_sock(struct nlmsghdr *nlh, void *arg)
4466 {
4467 struct filter *f = (struct filter *)arg;
4468 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
4469 struct rtattr *tb[NETLINK_DIAG_MAX+1];
4470 int rq = 0, wq = 0;
4471 unsigned long groups = 0;
4472
4473 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr *)(r+1),
4474 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4475
4476 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
4477 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
4478
4479 if (tb[NETLINK_DIAG_MEMINFO]) {
4480 const __u32 *skmeminfo;
4481
4482 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
4483
4484 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4485 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
4486 }
4487
4488 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
4489 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
4490 rq, wq, 0, 0)) {
4491 return 0;
4492 }
4493
4494 if (show_mem) {
4495 out("\t");
4496 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
4497 }
4498
4499 return 0;
4500 }
4501
4502 static int netlink_show_netlink(struct filter *f)
4503 {
4504 DIAG_REQUEST(req, struct netlink_diag_req r);
4505
4506 req.r.sdiag_family = AF_NETLINK;
4507 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
4508 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
4509
4510 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
4511 }
4512
4513 static int netlink_show(struct filter *f)
4514 {
4515 FILE *fp;
4516 char buf[256];
4517 int prot, pid;
4518 unsigned int groups;
4519 int rq, wq, rc;
4520 unsigned long long sk, cb;
4521
4522 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
4523 return 0;
4524
4525 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
4526 netlink_show_netlink(f) == 0)
4527 return 0;
4528
4529 if ((fp = net_netlink_open()) == NULL)
4530 return -1;
4531 if (!fgets(buf, sizeof(buf), fp)) {
4532 fclose(fp);
4533 return -1;
4534 }
4535
4536 while (fgets(buf, sizeof(buf), fp)) {
4537 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
4538 &sk,
4539 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
4540
4541 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
4542 }
4543
4544 fclose(fp);
4545 return 0;
4546 }
4547
4548 static bool vsock_type_skip(struct sockstat *s, struct filter *f)
4549 {
4550 if (s->type == SOCK_STREAM && !(f->dbs & (1 << VSOCK_ST_DB)))
4551 return true;
4552 if (s->type == SOCK_DGRAM && !(f->dbs & (1 << VSOCK_DG_DB)))
4553 return true;
4554 return false;
4555 }
4556
4557 static void vsock_addr_print(inet_prefix *a, __u32 port)
4558 {
4559 char cid_str[sizeof("4294967295")];
4560 char port_str[sizeof("4294967295")];
4561 __u32 cid;
4562
4563 memcpy(&cid, a->data, sizeof(cid));
4564
4565 if (cid == ~(__u32)0)
4566 snprintf(cid_str, sizeof(cid_str), "*");
4567 else
4568 snprintf(cid_str, sizeof(cid_str), "%u", cid);
4569
4570 if (port == ~(__u32)0)
4571 snprintf(port_str, sizeof(port_str), "*");
4572 else
4573 snprintf(port_str, sizeof(port_str), "%u", port);
4574
4575 sock_addr_print(cid_str, ":", port_str, NULL);
4576 }
4577
4578 static void vsock_stats_print(struct sockstat *s, struct filter *f)
4579 {
4580 sock_state_print(s);
4581
4582 vsock_addr_print(&s->local, s->lport);
4583 vsock_addr_print(&s->remote, s->rport);
4584
4585 proc_ctx_print(s);
4586 }
4587
4588 static int vsock_show_sock(struct nlmsghdr *nlh, void *arg)
4589 {
4590 struct filter *f = (struct filter *)arg;
4591 struct vsock_diag_msg *r = NLMSG_DATA(nlh);
4592 struct sockstat stat = {
4593 .type = r->vdiag_type,
4594 .lport = r->vdiag_src_port,
4595 .rport = r->vdiag_dst_port,
4596 .state = r->vdiag_state,
4597 .ino = r->vdiag_ino,
4598 };
4599
4600 vsock_set_inet_prefix(&stat.local, r->vdiag_src_cid);
4601 vsock_set_inet_prefix(&stat.remote, r->vdiag_dst_cid);
4602
4603 if (vsock_type_skip(&stat, f))
4604 return 0;
4605
4606 if (f->f && run_ssfilter(f->f, &stat) == 0)
4607 return 0;
4608
4609 vsock_stats_print(&stat, f);
4610
4611 return 0;
4612 }
4613
4614 static int vsock_show(struct filter *f)
4615 {
4616 DIAG_REQUEST(req, struct vsock_diag_req r);
4617
4618 if (!filter_af_get(f, AF_VSOCK))
4619 return 0;
4620
4621 req.r.sdiag_family = AF_VSOCK;
4622 req.r.vdiag_states = f->states;
4623
4624 return handle_netlink_request(f, &req.nlh, sizeof(req), vsock_show_sock);
4625 }
4626
4627 static void tipc_sock_addr_print(struct rtattr *net_addr, struct rtattr *id)
4628 {
4629 uint32_t node = rta_getattr_u32(net_addr);
4630 uint32_t identity = rta_getattr_u32(id);
4631
4632 SPRINT_BUF(addr) = {};
4633 SPRINT_BUF(port) = {};
4634
4635 sprintf(addr, "%u", node);
4636 sprintf(port, "%u", identity);
4637 sock_addr_print(addr, ":", port, NULL);
4638
4639 }
4640
4641 static int tipc_show_sock(struct nlmsghdr *nlh, void *arg)
4642 {
4643 struct rtattr *stat[TIPC_NLA_SOCK_STAT_MAX + 1] = {};
4644 struct rtattr *attrs[TIPC_NLA_SOCK_MAX + 1] = {};
4645 struct rtattr *con[TIPC_NLA_CON_MAX + 1] = {};
4646 struct rtattr *info[TIPC_NLA_MAX + 1] = {};
4647 struct rtattr *msg_ref;
4648 struct sockstat ss = {};
4649
4650 parse_rtattr(info, TIPC_NLA_MAX, NLMSG_DATA(nlh),
4651 NLMSG_PAYLOAD(nlh, 0));
4652
4653 if (!info[TIPC_NLA_SOCK])
4654 return 0;
4655
4656 msg_ref = info[TIPC_NLA_SOCK];
4657 parse_rtattr(attrs, TIPC_NLA_SOCK_MAX, RTA_DATA(msg_ref),
4658 RTA_PAYLOAD(msg_ref));
4659
4660 msg_ref = attrs[TIPC_NLA_SOCK_STAT];
4661 parse_rtattr(stat, TIPC_NLA_SOCK_STAT_MAX,
4662 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4663
4664
4665 ss.local.family = AF_TIPC;
4666 ss.type = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TYPE]);
4667 ss.state = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TIPC_STATE]);
4668 ss.uid = rta_getattr_u32(attrs[TIPC_NLA_SOCK_UID]);
4669 ss.ino = rta_getattr_u32(attrs[TIPC_NLA_SOCK_INO]);
4670 ss.rq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_RCVQ]);
4671 ss.wq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_SENDQ]);
4672 ss.sk = rta_getattr_u64(attrs[TIPC_NLA_SOCK_COOKIE]);
4673
4674 sock_state_print (&ss);
4675
4676 tipc_sock_addr_print(attrs[TIPC_NLA_SOCK_ADDR],
4677 attrs[TIPC_NLA_SOCK_REF]);
4678
4679 msg_ref = attrs[TIPC_NLA_SOCK_CON];
4680 if (msg_ref) {
4681 parse_rtattr(con, TIPC_NLA_CON_MAX,
4682 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4683
4684 tipc_sock_addr_print(con[TIPC_NLA_CON_NODE],
4685 con[TIPC_NLA_CON_SOCK]);
4686 } else
4687 sock_addr_print("", "-", "", NULL);
4688
4689 if (show_details)
4690 sock_details_print(&ss);
4691
4692 proc_ctx_print(&ss);
4693
4694 if (show_tipcinfo) {
4695 if (oneline)
4696 out(" type:%s", stype_nameg[ss.type]);
4697 else
4698 out("\n type:%s", stype_nameg[ss.type]);
4699 out(" cong:%s ",
4700 stat[TIPC_NLA_SOCK_STAT_LINK_CONG] ? "link" :
4701 stat[TIPC_NLA_SOCK_STAT_CONN_CONG] ? "conn" : "none");
4702 out(" drop:%d ",
4703 rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_DROP]));
4704
4705 if (attrs[TIPC_NLA_SOCK_HAS_PUBL])
4706 out(" publ");
4707
4708 if (con[TIPC_NLA_CON_FLAG])
4709 out(" via {%u,%u} ",
4710 rta_getattr_u32(con[TIPC_NLA_CON_TYPE]),
4711 rta_getattr_u32(con[TIPC_NLA_CON_INST]));
4712 }
4713
4714 return 0;
4715 }
4716
4717 static int tipc_show(struct filter *f)
4718 {
4719 DIAG_REQUEST(req, struct tipc_sock_diag_req r);
4720
4721 memset(&req.r, 0, sizeof(req.r));
4722 req.r.sdiag_family = AF_TIPC;
4723 req.r.tidiag_states = f->states;
4724
4725 return handle_netlink_request(f, &req.nlh, sizeof(req), tipc_show_sock);
4726 }
4727
4728 struct sock_diag_msg {
4729 __u8 sdiag_family;
4730 };
4731
4732 static int generic_show_sock(struct nlmsghdr *nlh, void *arg)
4733 {
4734 struct sock_diag_msg *r = NLMSG_DATA(nlh);
4735 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
4736 int ret;
4737
4738 switch (r->sdiag_family) {
4739 case AF_INET:
4740 case AF_INET6:
4741 inet_arg.rth = inet_arg.f->rth_for_killing;
4742 ret = show_one_inet_sock(nlh, &inet_arg);
4743 break;
4744 case AF_UNIX:
4745 ret = unix_show_sock(nlh, arg);
4746 break;
4747 case AF_PACKET:
4748 ret = packet_show_sock(nlh, arg);
4749 break;
4750 case AF_NETLINK:
4751 ret = netlink_show_sock(nlh, arg);
4752 break;
4753 case AF_VSOCK:
4754 ret = vsock_show_sock(nlh, arg);
4755 break;
4756 case AF_XDP:
4757 ret = xdp_show_sock(nlh, arg);
4758 break;
4759 default:
4760 ret = -1;
4761 }
4762
4763 render();
4764
4765 return ret;
4766 }
4767
4768 static int handle_follow_request(struct filter *f)
4769 {
4770 int ret = 0;
4771 int groups = 0;
4772 struct rtnl_handle rth, rth2;
4773
4774 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << TCP_DB))
4775 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
4776 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << UDP_DB))
4777 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
4778 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << TCP_DB))
4779 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
4780 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << UDP_DB))
4781 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
4782
4783 if (groups == 0)
4784 return -1;
4785
4786 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
4787 return -1;
4788
4789 rth.dump = 0;
4790 rth.local.nl_pid = 0;
4791
4792 if (f->kill) {
4793 if (rtnl_open_byproto(&rth2, groups, NETLINK_SOCK_DIAG)) {
4794 rtnl_close(&rth);
4795 return -1;
4796 }
4797 f->rth_for_killing = &rth2;
4798 }
4799
4800 if (rtnl_dump_filter(&rth, generic_show_sock, f))
4801 ret = -1;
4802
4803 rtnl_close(&rth);
4804 if (f->rth_for_killing)
4805 rtnl_close(f->rth_for_killing);
4806 return ret;
4807 }
4808
4809 static int get_snmp_int(char *proto, char *key, int *result)
4810 {
4811 char buf[1024];
4812 FILE *fp;
4813 int protolen = strlen(proto);
4814 int keylen = strlen(key);
4815
4816 *result = 0;
4817
4818 if ((fp = net_snmp_open()) == NULL)
4819 return -1;
4820
4821 while (fgets(buf, sizeof(buf), fp) != NULL) {
4822 char *p = buf;
4823 int pos = 0;
4824
4825 if (memcmp(buf, proto, protolen))
4826 continue;
4827 while ((p = strchr(p, ' ')) != NULL) {
4828 pos++;
4829 p++;
4830 if (memcmp(p, key, keylen) == 0 &&
4831 (p[keylen] == ' ' || p[keylen] == '\n'))
4832 break;
4833 }
4834 if (fgets(buf, sizeof(buf), fp) == NULL)
4835 break;
4836 if (memcmp(buf, proto, protolen))
4837 break;
4838 p = buf;
4839 while ((p = strchr(p, ' ')) != NULL) {
4840 p++;
4841 if (--pos == 0) {
4842 sscanf(p, "%d", result);
4843 fclose(fp);
4844 return 0;
4845 }
4846 }
4847 }
4848
4849 fclose(fp);
4850 errno = ESRCH;
4851 return -1;
4852 }
4853
4854
4855 /* Get stats from sockstat */
4856
4857 struct ssummary {
4858 int socks;
4859 int tcp_mem;
4860 int tcp_total;
4861 int tcp_orphans;
4862 int tcp_tws;
4863 int tcp4_hashed;
4864 int udp4;
4865 int raw4;
4866 int frag4;
4867 int frag4_mem;
4868 int tcp6_hashed;
4869 int udp6;
4870 int raw6;
4871 int frag6;
4872 int frag6_mem;
4873 };
4874
4875 static void get_sockstat_line(char *line, struct ssummary *s)
4876 {
4877 char id[256], rem[256];
4878
4879 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
4880 return;
4881
4882 if (strcmp(id, "sockets:") == 0)
4883 sscanf(rem, "%*s%d", &s->socks);
4884 else if (strcmp(id, "UDP:") == 0)
4885 sscanf(rem, "%*s%d", &s->udp4);
4886 else if (strcmp(id, "UDP6:") == 0)
4887 sscanf(rem, "%*s%d", &s->udp6);
4888 else if (strcmp(id, "RAW:") == 0)
4889 sscanf(rem, "%*s%d", &s->raw4);
4890 else if (strcmp(id, "RAW6:") == 0)
4891 sscanf(rem, "%*s%d", &s->raw6);
4892 else if (strcmp(id, "TCP6:") == 0)
4893 sscanf(rem, "%*s%d", &s->tcp6_hashed);
4894 else if (strcmp(id, "FRAG:") == 0)
4895 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
4896 else if (strcmp(id, "FRAG6:") == 0)
4897 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
4898 else if (strcmp(id, "TCP:") == 0)
4899 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
4900 &s->tcp4_hashed,
4901 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
4902 }
4903
4904 static int get_sockstat(struct ssummary *s)
4905 {
4906 char buf[256];
4907 FILE *fp;
4908
4909 memset(s, 0, sizeof(*s));
4910
4911 if ((fp = net_sockstat_open()) == NULL)
4912 return -1;
4913 while (fgets(buf, sizeof(buf), fp) != NULL)
4914 get_sockstat_line(buf, s);
4915 fclose(fp);
4916
4917 if ((fp = net_sockstat6_open()) == NULL)
4918 return 0;
4919 while (fgets(buf, sizeof(buf), fp) != NULL)
4920 get_sockstat_line(buf, s);
4921 fclose(fp);
4922
4923 return 0;
4924 }
4925
4926 static int print_summary(void)
4927 {
4928 struct ssummary s;
4929 int tcp_estab;
4930
4931 if (get_sockstat(&s) < 0)
4932 perror("ss: get_sockstat");
4933 if (get_snmp_int("Tcp:", "CurrEstab", &tcp_estab) < 0)
4934 perror("ss: get_snmpstat");
4935
4936 printf("Total: %d\n", s.socks);
4937
4938 printf("TCP: %d (estab %d, closed %d, orphaned %d, timewait %d)\n",
4939 s.tcp_total + s.tcp_tws, tcp_estab,
4940 s.tcp_total - (s.tcp4_hashed + s.tcp6_hashed - s.tcp_tws),
4941 s.tcp_orphans, s.tcp_tws);
4942
4943 printf("\n");
4944 printf("Transport Total IP IPv6\n");
4945 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
4946 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
4947 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
4948 printf("INET %-9d %-9d %-9d\n",
4949 s.raw4+s.udp4+s.tcp4_hashed+
4950 s.raw6+s.udp6+s.tcp6_hashed,
4951 s.raw4+s.udp4+s.tcp4_hashed,
4952 s.raw6+s.udp6+s.tcp6_hashed);
4953 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
4954
4955 printf("\n");
4956
4957 return 0;
4958 }
4959
4960 static void _usage(FILE *dest)
4961 {
4962 fprintf(dest,
4963 "Usage: ss [ OPTIONS ]\n"
4964 " ss [ OPTIONS ] [ FILTER ]\n"
4965 " -h, --help this message\n"
4966 " -V, --version output version information\n"
4967 " -n, --numeric don't resolve service names\n"
4968 " -r, --resolve resolve host names\n"
4969 " -a, --all display all sockets\n"
4970 " -l, --listening display listening sockets\n"
4971 " -o, --options show timer information\n"
4972 " -e, --extended show detailed socket information\n"
4973 " -m, --memory show socket memory usage\n"
4974 " -p, --processes show process using socket\n"
4975 " -i, --info show internal TCP information\n"
4976 " --tipcinfo show internal tipc socket information\n"
4977 " -s, --summary show socket usage summary\n"
4978 " --tos show tos and priority information\n"
4979 " -b, --bpf show bpf filter socket information\n"
4980 " -E, --events continually display sockets as they are destroyed\n"
4981 " -Z, --context display process SELinux security contexts\n"
4982 " -z, --contexts display process and socket SELinux security contexts\n"
4983 " -N, --net switch to the specified network namespace name\n"
4984 "\n"
4985 " -4, --ipv4 display only IP version 4 sockets\n"
4986 " -6, --ipv6 display only IP version 6 sockets\n"
4987 " -0, --packet display PACKET sockets\n"
4988 " -t, --tcp display only TCP sockets\n"
4989 " -S, --sctp display only SCTP sockets\n"
4990 " -u, --udp display only UDP sockets\n"
4991 " -d, --dccp display only DCCP sockets\n"
4992 " -w, --raw display only RAW sockets\n"
4993 " -x, --unix display only Unix domain sockets\n"
4994 " --tipc display only TIPC sockets\n"
4995 " --vsock display only vsock sockets\n"
4996 " -f, --family=FAMILY display sockets of type FAMILY\n"
4997 " FAMILY := {inet|inet6|link|unix|netlink|vsock|tipc|xdp|help}\n"
4998 "\n"
4999 " -K, --kill forcibly close sockets, display what was closed\n"
5000 " -H, --no-header Suppress header line\n"
5001 " -O, --oneline socket's data printed on a single line\n"
5002 "\n"
5003 " -A, --query=QUERY, --socket=QUERY\n"
5004 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink|vsock_stream|vsock_dgram|tipc}[,QUERY]\n"
5005 "\n"
5006 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
5007 " -F, --filter=FILE read filter information from FILE\n"
5008 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
5009 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
5010 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listening|closing}\n"
5011 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
5012 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
5013 " bucket := {syn-recv|time-wait}\n"
5014 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listening|closing}\n"
5015 );
5016 }
5017
5018 static void help(void) __attribute__((noreturn));
5019 static void help(void)
5020 {
5021 _usage(stdout);
5022 exit(0);
5023 }
5024
5025 static void usage(void) __attribute__((noreturn));
5026 static void usage(void)
5027 {
5028 _usage(stderr);
5029 exit(-1);
5030 }
5031
5032
5033 static int scan_state(const char *state)
5034 {
5035 static const char * const sstate_namel[] = {
5036 "UNKNOWN",
5037 [SS_ESTABLISHED] = "established",
5038 [SS_SYN_SENT] = "syn-sent",
5039 [SS_SYN_RECV] = "syn-recv",
5040 [SS_FIN_WAIT1] = "fin-wait-1",
5041 [SS_FIN_WAIT2] = "fin-wait-2",
5042 [SS_TIME_WAIT] = "time-wait",
5043 [SS_CLOSE] = "unconnected",
5044 [SS_CLOSE_WAIT] = "close-wait",
5045 [SS_LAST_ACK] = "last-ack",
5046 [SS_LISTEN] = "listening",
5047 [SS_CLOSING] = "closing",
5048 };
5049 int i;
5050
5051 if (strcasecmp(state, "close") == 0 ||
5052 strcasecmp(state, "closed") == 0)
5053 return (1<<SS_CLOSE);
5054 if (strcasecmp(state, "syn-rcv") == 0)
5055 return (1<<SS_SYN_RECV);
5056 if (strcasecmp(state, "established") == 0)
5057 return (1<<SS_ESTABLISHED);
5058 if (strcasecmp(state, "all") == 0)
5059 return SS_ALL;
5060 if (strcasecmp(state, "connected") == 0)
5061 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
5062 if (strcasecmp(state, "synchronized") == 0)
5063 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
5064 if (strcasecmp(state, "bucket") == 0)
5065 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
5066 if (strcasecmp(state, "big") == 0)
5067 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
5068 for (i = 0; i < SS_MAX; i++) {
5069 if (strcasecmp(state, sstate_namel[i]) == 0)
5070 return (1<<i);
5071 }
5072
5073 fprintf(stderr, "ss: wrong state name: %s\n", state);
5074 exit(-1);
5075 }
5076
5077 /* Values 'v' and 'V' are already used so a non-character is used */
5078 #define OPT_VSOCK 256
5079
5080 /* Values of 't' are already used so a non-character is used */
5081 #define OPT_TIPCSOCK 257
5082 #define OPT_TIPCINFO 258
5083
5084 #define OPT_TOS 259
5085
5086 /* Values of 'x' are already used so a non-character is used */
5087 #define OPT_XDPSOCK 260
5088
5089 static const struct option long_opts[] = {
5090 { "numeric", 0, 0, 'n' },
5091 { "resolve", 0, 0, 'r' },
5092 { "options", 0, 0, 'o' },
5093 { "extended", 0, 0, 'e' },
5094 { "memory", 0, 0, 'm' },
5095 { "info", 0, 0, 'i' },
5096 { "processes", 0, 0, 'p' },
5097 { "bpf", 0, 0, 'b' },
5098 { "events", 0, 0, 'E' },
5099 { "dccp", 0, 0, 'd' },
5100 { "tcp", 0, 0, 't' },
5101 { "sctp", 0, 0, 'S' },
5102 { "udp", 0, 0, 'u' },
5103 { "raw", 0, 0, 'w' },
5104 { "unix", 0, 0, 'x' },
5105 { "tipc", 0, 0, OPT_TIPCSOCK},
5106 { "vsock", 0, 0, OPT_VSOCK },
5107 { "all", 0, 0, 'a' },
5108 { "listening", 0, 0, 'l' },
5109 { "ipv4", 0, 0, '4' },
5110 { "ipv6", 0, 0, '6' },
5111 { "packet", 0, 0, '0' },
5112 { "family", 1, 0, 'f' },
5113 { "socket", 1, 0, 'A' },
5114 { "query", 1, 0, 'A' },
5115 { "summary", 0, 0, 's' },
5116 { "diag", 1, 0, 'D' },
5117 { "filter", 1, 0, 'F' },
5118 { "version", 0, 0, 'V' },
5119 { "help", 0, 0, 'h' },
5120 { "context", 0, 0, 'Z' },
5121 { "contexts", 0, 0, 'z' },
5122 { "net", 1, 0, 'N' },
5123 { "tipcinfo", 0, 0, OPT_TIPCINFO},
5124 { "tos", 0, 0, OPT_TOS },
5125 { "kill", 0, 0, 'K' },
5126 { "no-header", 0, 0, 'H' },
5127 { "xdp", 0, 0, OPT_XDPSOCK},
5128 { "oneline", 0, 0, 'O' },
5129 { 0 }
5130
5131 };
5132
5133 int main(int argc, char *argv[])
5134 {
5135 int saw_states = 0;
5136 int saw_query = 0;
5137 int do_summary = 0;
5138 const char *dump_tcpdiag = NULL;
5139 FILE *filter_fp = NULL;
5140 int ch;
5141 int state_filter = 0;
5142
5143 while ((ch = getopt_long(argc, argv,
5144 "dhaletuwxnro460spbEf:miA:D:F:vVzZN:KHSO",
5145 long_opts, NULL)) != EOF) {
5146 switch (ch) {
5147 case 'n':
5148 numeric = 1;
5149 break;
5150 case 'r':
5151 resolve_hosts = 1;
5152 break;
5153 case 'o':
5154 show_options = 1;
5155 break;
5156 case 'e':
5157 show_options = 1;
5158 show_details++;
5159 break;
5160 case 'm':
5161 show_mem = 1;
5162 break;
5163 case 'i':
5164 show_tcpinfo = 1;
5165 break;
5166 case 'p':
5167 show_users++;
5168 user_ent_hash_build();
5169 break;
5170 case 'b':
5171 show_options = 1;
5172 show_bpf++;
5173 break;
5174 case 'E':
5175 follow_events = 1;
5176 break;
5177 case 'd':
5178 filter_db_set(&current_filter, DCCP_DB, true);
5179 break;
5180 case 't':
5181 filter_db_set(&current_filter, TCP_DB, true);
5182 break;
5183 case 'S':
5184 filter_db_set(&current_filter, SCTP_DB, true);
5185 break;
5186 case 'u':
5187 filter_db_set(&current_filter, UDP_DB, true);
5188 break;
5189 case 'w':
5190 filter_db_set(&current_filter, RAW_DB, true);
5191 break;
5192 case 'x':
5193 filter_af_set(&current_filter, AF_UNIX);
5194 break;
5195 case OPT_VSOCK:
5196 filter_af_set(&current_filter, AF_VSOCK);
5197 break;
5198 case OPT_TIPCSOCK:
5199 filter_af_set(&current_filter, AF_TIPC);
5200 break;
5201 case 'a':
5202 state_filter = SS_ALL;
5203 break;
5204 case 'l':
5205 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
5206 break;
5207 case '4':
5208 filter_af_set(&current_filter, AF_INET);
5209 break;
5210 case '6':
5211 filter_af_set(&current_filter, AF_INET6);
5212 break;
5213 case '0':
5214 filter_af_set(&current_filter, AF_PACKET);
5215 break;
5216 case OPT_XDPSOCK:
5217 filter_af_set(&current_filter, AF_XDP);
5218 break;
5219 case 'f':
5220 if (strcmp(optarg, "inet") == 0)
5221 filter_af_set(&current_filter, AF_INET);
5222 else if (strcmp(optarg, "inet6") == 0)
5223 filter_af_set(&current_filter, AF_INET6);
5224 else if (strcmp(optarg, "link") == 0)
5225 filter_af_set(&current_filter, AF_PACKET);
5226 else if (strcmp(optarg, "unix") == 0)
5227 filter_af_set(&current_filter, AF_UNIX);
5228 else if (strcmp(optarg, "netlink") == 0)
5229 filter_af_set(&current_filter, AF_NETLINK);
5230 else if (strcmp(optarg, "tipc") == 0)
5231 filter_af_set(&current_filter, AF_TIPC);
5232 else if (strcmp(optarg, "vsock") == 0)
5233 filter_af_set(&current_filter, AF_VSOCK);
5234 else if (strcmp(optarg, "xdp") == 0)
5235 filter_af_set(&current_filter, AF_XDP);
5236 else if (strcmp(optarg, "help") == 0)
5237 help();
5238 else {
5239 fprintf(stderr, "ss: \"%s\" is invalid family\n",
5240 optarg);
5241 usage();
5242 }
5243 break;
5244 case 'A':
5245 {
5246 char *p, *p1;
5247
5248 if (!saw_query) {
5249 current_filter.dbs = 0;
5250 state_filter = state_filter ?
5251 state_filter : SS_CONN;
5252 saw_query = 1;
5253 do_default = 0;
5254 }
5255 p = p1 = optarg;
5256 do {
5257 if ((p1 = strchr(p, ',')) != NULL)
5258 *p1 = 0;
5259 if (filter_db_parse(&current_filter, p)) {
5260 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
5261 usage();
5262 }
5263 p = p1 + 1;
5264 } while (p1);
5265 break;
5266 }
5267 case 's':
5268 do_summary = 1;
5269 break;
5270 case 'D':
5271 dump_tcpdiag = optarg;
5272 break;
5273 case 'F':
5274 if (filter_fp) {
5275 fprintf(stderr, "More than one filter file\n");
5276 exit(-1);
5277 }
5278 if (optarg[0] == '-')
5279 filter_fp = stdin;
5280 else
5281 filter_fp = fopen(optarg, "r");
5282 if (!filter_fp) {
5283 perror("fopen filter file");
5284 exit(-1);
5285 }
5286 break;
5287 case 'v':
5288 case 'V':
5289 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
5290 exit(0);
5291 case 'z':
5292 show_sock_ctx++;
5293 /* fall through */
5294 case 'Z':
5295 if (is_selinux_enabled() <= 0) {
5296 fprintf(stderr, "ss: SELinux is not enabled.\n");
5297 exit(1);
5298 }
5299 show_proc_ctx++;
5300 user_ent_hash_build();
5301 break;
5302 case 'N':
5303 if (netns_switch(optarg))
5304 exit(1);
5305 break;
5306 case OPT_TIPCINFO:
5307 show_tipcinfo = 1;
5308 break;
5309 case OPT_TOS:
5310 show_tos = 1;
5311 break;
5312 case 'K':
5313 current_filter.kill = 1;
5314 break;
5315 case 'H':
5316 show_header = 0;
5317 break;
5318 case 'O':
5319 oneline = 1;
5320 break;
5321 case 'h':
5322 help();
5323 case '?':
5324 default:
5325 usage();
5326 }
5327 }
5328
5329 argc -= optind;
5330 argv += optind;
5331
5332 if (do_summary) {
5333 print_summary();
5334 if (do_default && argc == 0)
5335 exit(0);
5336 }
5337
5338 while (argc > 0) {
5339 if (strcmp(*argv, "state") == 0) {
5340 NEXT_ARG();
5341 if (!saw_states)
5342 state_filter = 0;
5343 state_filter |= scan_state(*argv);
5344 saw_states = 1;
5345 } else if (strcmp(*argv, "exclude") == 0 ||
5346 strcmp(*argv, "excl") == 0) {
5347 NEXT_ARG();
5348 if (!saw_states)
5349 state_filter = SS_ALL;
5350 state_filter &= ~scan_state(*argv);
5351 saw_states = 1;
5352 } else {
5353 break;
5354 }
5355 argc--; argv++;
5356 }
5357
5358 if (do_default) {
5359 state_filter = state_filter ? state_filter : SS_CONN;
5360 filter_db_parse(&current_filter, "all");
5361 }
5362
5363 filter_states_set(&current_filter, state_filter);
5364 filter_merge_defaults(&current_filter);
5365
5366 if (!numeric && resolve_hosts &&
5367 (current_filter.dbs & (UNIX_DBM|INET_L4_DBM)))
5368 init_service_resolver();
5369
5370 if (current_filter.dbs == 0) {
5371 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
5372 exit(0);
5373 }
5374 if (current_filter.families == 0) {
5375 fprintf(stderr, "ss: no families to show with such filter.\n");
5376 exit(0);
5377 }
5378 if (current_filter.states == 0) {
5379 fprintf(stderr, "ss: no socket states to show with such filter.\n");
5380 exit(0);
5381 }
5382
5383 if (dump_tcpdiag) {
5384 FILE *dump_fp = stdout;
5385
5386 if (!(current_filter.dbs & (1<<TCP_DB))) {
5387 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
5388 exit(0);
5389 }
5390 if (dump_tcpdiag[0] != '-') {
5391 dump_fp = fopen(dump_tcpdiag, "w");
5392 if (!dump_tcpdiag) {
5393 perror("fopen dump file");
5394 exit(-1);
5395 }
5396 }
5397 inet_show_netlink(&current_filter, dump_fp, IPPROTO_TCP);
5398 fflush(dump_fp);
5399 exit(0);
5400 }
5401
5402 if (ssfilter_parse(&current_filter.f, argc, argv, filter_fp))
5403 usage();
5404
5405 if (!(current_filter.dbs & (current_filter.dbs - 1)))
5406 columns[COL_NETID].disabled = 1;
5407
5408 if (!(current_filter.states & (current_filter.states - 1)))
5409 columns[COL_STATE].disabled = 1;
5410
5411 if (show_header)
5412 print_header();
5413
5414 fflush(stdout);
5415
5416 if (follow_events)
5417 exit(handle_follow_request(&current_filter));
5418
5419 if (current_filter.dbs & (1<<NETLINK_DB))
5420 netlink_show(&current_filter);
5421 if (current_filter.dbs & PACKET_DBM)
5422 packet_show(&current_filter);
5423 if (current_filter.dbs & UNIX_DBM)
5424 unix_show(&current_filter);
5425 if (current_filter.dbs & (1<<RAW_DB))
5426 raw_show(&current_filter);
5427 if (current_filter.dbs & (1<<UDP_DB))
5428 udp_show(&current_filter);
5429 if (current_filter.dbs & (1<<TCP_DB))
5430 tcp_show(&current_filter);
5431 if (current_filter.dbs & (1<<DCCP_DB))
5432 dccp_show(&current_filter);
5433 if (current_filter.dbs & (1<<SCTP_DB))
5434 sctp_show(&current_filter);
5435 if (current_filter.dbs & VSOCK_DBM)
5436 vsock_show(&current_filter);
5437 if (current_filter.dbs & (1<<TIPC_DB))
5438 tipc_show(&current_filter);
5439 if (current_filter.dbs & (1<<XDP_DB))
5440 xdp_show(&current_filter);
5441
5442 if (show_users || show_proc_ctx || show_sock_ctx)
5443 user_ent_destroy();
5444
5445 render();
5446
5447 return 0;
5448 }
Upstream mirror of iproute2