]> git.proxmox.com Git - mirror_iproute2.git/blob - misc/ss.c
projects / mirror_iproute2.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'master' into next
[mirror_iproute2.git] / misc / ss.c
1 /*
2 * ss.c "sockstat", socket statistics
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 */
11
12 #include <stdio.h>
13 #include <stdlib.h>
14 #include <unistd.h>
15 #include <fcntl.h>
16 #include <sys/ioctl.h>
17 #include <sys/socket.h>
18 #include <sys/uio.h>
19 #include <sys/sysmacros.h>
20 #include <netinet/in.h>
21 #include <string.h>
22 #include <errno.h>
23 #include <netdb.h>
24 #include <arpa/inet.h>
25 #include <dirent.h>
26 #include <fnmatch.h>
27 #include <getopt.h>
28 #include <stdbool.h>
29 #include <limits.h>
30 #include <stdarg.h>
31
32 #include "utils.h"
33 #include "rt_names.h"
34 #include "ll_map.h"
35 #include "libnetlink.h"
36 #include "namespace.h"
37 #include "SNAPSHOT.h"
38
39 #include <linux/tcp.h>
40 #include <linux/sock_diag.h>
41 #include <linux/inet_diag.h>
42 #include <linux/unix_diag.h>
43 #include <linux/netdevice.h> /* for MAX_ADDR_LEN */
44 #include <linux/filter.h>
45 #include <linux/xdp_diag.h>
46 #include <linux/packet_diag.h>
47 #include <linux/netlink_diag.h>
48 #include <linux/sctp.h>
49 #include <linux/vm_sockets_diag.h>
50 #include <linux/net.h>
51 #include <linux/tipc.h>
52 #include <linux/tipc_netlink.h>
53 #include <linux/tipc_sockets_diag.h>
54
55 /* AF_VSOCK/PF_VSOCK is only provided since glibc 2.18 */
56 #ifndef PF_VSOCK
57 #define PF_VSOCK 40
58 #endif
59 #ifndef AF_VSOCK
60 #define AF_VSOCK PF_VSOCK
61 #endif
62
63 #define MAGIC_SEQ 123456
64 #define BUF_CHUNK (1024 * 1024) /* Buffer chunk allocation size */
65 #define BUF_CHUNKS_MAX 5 /* Maximum number of allocated buffer chunks */
66 #define LEN_ALIGN(x) (((x) + 1) & ~1)
67
68 #define DIAG_REQUEST(_req, _r) \
69 struct { \
70 struct nlmsghdr nlh; \
71 _r; \
72 } _req = { \
73 .nlh = { \
74 .nlmsg_type = SOCK_DIAG_BY_FAMILY, \
75 .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,\
76 .nlmsg_seq = MAGIC_SEQ, \
77 .nlmsg_len = sizeof(_req), \
78 }, \
79 }
80
81 #if HAVE_SELINUX
82 #include <selinux/selinux.h>
83 #else
84 /* Stubs for SELinux functions */
85 static int is_selinux_enabled(void)
86 {
87 return -1;
88 }
89
90 static int getpidcon(pid_t pid, char **context)
91 {
92 *context = NULL;
93 return -1;
94 }
95
96 static int getfilecon(char *path, char **context)
97 {
98 *context = NULL;
99 return -1;
100 }
101
102 static int security_get_initial_context(char *name, char **context)
103 {
104 *context = NULL;
105 return -1;
106 }
107 #endif
108
109 int preferred_family = AF_UNSPEC;
110 static int show_options;
111 int show_details;
112 static int show_users;
113 static int show_mem;
114 static int show_tcpinfo;
115 static int show_bpf;
116 static int show_proc_ctx;
117 static int show_sock_ctx;
118 static int show_header = 1;
119 static int follow_events;
120 static int sctp_ino;
121 static int show_tipcinfo;
122 static int show_tos;
123 int numeric;
124 int oneline;
125
126 enum col_id {
127 COL_NETID,
128 COL_STATE,
129 COL_RECVQ,
130 COL_SENDQ,
131 COL_ADDR,
132 COL_SERV,
133 COL_RADDR,
134 COL_RSERV,
135 COL_EXT,
136 COL_MAX
137 };
138
139 enum col_align {
140 ALIGN_LEFT,
141 ALIGN_CENTER,
142 ALIGN_RIGHT
143 };
144
145 struct column {
146 const enum col_align align;
147 const char *header;
148 const char *ldelim;
149 int disabled;
150 int width; /* Calculated, including additional layout spacing */
151 int max_len; /* Measured maximum field length in this column */
152 };
153
154 static struct column columns[] = {
155 { ALIGN_LEFT, "Netid", "", 0, 0, 0 },
156 { ALIGN_LEFT, "State", " ", 0, 0, 0 },
157 { ALIGN_LEFT, "Recv-Q", " ", 0, 0, 0 },
158 { ALIGN_LEFT, "Send-Q", " ", 0, 0, 0 },
159 { ALIGN_RIGHT, "Local Address:", " ", 0, 0, 0 },
160 { ALIGN_LEFT, "Port", "", 0, 0, 0 },
161 { ALIGN_RIGHT, "Peer Address:", " ", 0, 0, 0 },
162 { ALIGN_LEFT, "Port", "", 0, 0, 0 },
163 { ALIGN_LEFT, "", "", 0, 0, 0 },
164 };
165
166 static struct column *current_field = columns;
167
168 /* Output buffer: chained chunks of BUF_CHUNK bytes. Each field is written to
169 * the buffer as a variable size token. A token consists of a 16 bits length
170 * field, followed by a string which is not NULL-terminated.
171 *
172 * A new chunk is allocated and linked when the current chunk doesn't have
173 * enough room to store the current token as a whole.
174 */
175 struct buf_chunk {
176 struct buf_chunk *next; /* Next chained chunk */
177 char *end; /* Current end of content */
178 char data[0];
179 };
180
181 struct buf_token {
182 uint16_t len; /* Data length, excluding length descriptor */
183 char data[0];
184 };
185
186 static struct {
187 struct buf_token *cur; /* Position of current token in chunk */
188 struct buf_chunk *head; /* First chunk */
189 struct buf_chunk *tail; /* Current chunk */
190 int chunks; /* Number of allocated chunks */
191 } buffer;
192
193 static const char *TCP_PROTO = "tcp";
194 static const char *SCTP_PROTO = "sctp";
195 static const char *UDP_PROTO = "udp";
196 static const char *RAW_PROTO = "raw";
197 static const char *dg_proto;
198
199 enum {
200 TCP_DB,
201 DCCP_DB,
202 UDP_DB,
203 RAW_DB,
204 UNIX_DG_DB,
205 UNIX_ST_DB,
206 UNIX_SQ_DB,
207 PACKET_DG_DB,
208 PACKET_R_DB,
209 NETLINK_DB,
210 SCTP_DB,
211 VSOCK_ST_DB,
212 VSOCK_DG_DB,
213 TIPC_DB,
214 XDP_DB,
215 MAX_DB
216 };
217
218 #define PACKET_DBM ((1<<PACKET_DG_DB)|(1<<PACKET_R_DB))
219 #define UNIX_DBM ((1<<UNIX_DG_DB)|(1<<UNIX_ST_DB)|(1<<UNIX_SQ_DB))
220 #define ALL_DB ((1<<MAX_DB)-1)
221 #define INET_L4_DBM ((1<<TCP_DB)|(1<<UDP_DB)|(1<<DCCP_DB)|(1<<SCTP_DB))
222 #define INET_DBM (INET_L4_DBM | (1<<RAW_DB))
223 #define VSOCK_DBM ((1<<VSOCK_ST_DB)|(1<<VSOCK_DG_DB))
224
225 enum {
226 SS_UNKNOWN,
227 SS_ESTABLISHED,
228 SS_SYN_SENT,
229 SS_SYN_RECV,
230 SS_FIN_WAIT1,
231 SS_FIN_WAIT2,
232 SS_TIME_WAIT,
233 SS_CLOSE,
234 SS_CLOSE_WAIT,
235 SS_LAST_ACK,
236 SS_LISTEN,
237 SS_CLOSING,
238 SS_MAX
239 };
240
241 enum {
242 SCTP_STATE_CLOSED = 0,
243 SCTP_STATE_COOKIE_WAIT = 1,
244 SCTP_STATE_COOKIE_ECHOED = 2,
245 SCTP_STATE_ESTABLISHED = 3,
246 SCTP_STATE_SHUTDOWN_PENDING = 4,
247 SCTP_STATE_SHUTDOWN_SENT = 5,
248 SCTP_STATE_SHUTDOWN_RECEIVED = 6,
249 SCTP_STATE_SHUTDOWN_ACK_SENT = 7,
250 };
251
252 #define SS_ALL ((1 << SS_MAX) - 1)
253 #define SS_CONN (SS_ALL & ~((1<<SS_LISTEN)|(1<<SS_CLOSE)|(1<<SS_TIME_WAIT)|(1<<SS_SYN_RECV)))
254 #define TIPC_SS_CONN ((1<<SS_ESTABLISHED)|(1<<SS_LISTEN)|(1<<SS_CLOSE))
255
256 #include "ssfilter.h"
257
258 struct filter {
259 int dbs;
260 int states;
261 uint64_t families;
262 struct ssfilter *f;
263 bool kill;
264 struct rtnl_handle *rth_for_killing;
265 };
266
267 #define FAMILY_MASK(family) ((uint64_t)1 << (family))
268
269 static const struct filter default_dbs[MAX_DB] = {
270 [TCP_DB] = {
271 .states = SS_CONN,
272 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
273 },
274 [DCCP_DB] = {
275 .states = SS_CONN,
276 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
277 },
278 [UDP_DB] = {
279 .states = (1 << SS_ESTABLISHED),
280 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
281 },
282 [RAW_DB] = {
283 .states = (1 << SS_ESTABLISHED),
284 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
285 },
286 [UNIX_DG_DB] = {
287 .states = (1 << SS_CLOSE),
288 .families = FAMILY_MASK(AF_UNIX),
289 },
290 [UNIX_ST_DB] = {
291 .states = SS_CONN,
292 .families = FAMILY_MASK(AF_UNIX),
293 },
294 [UNIX_SQ_DB] = {
295 .states = SS_CONN,
296 .families = FAMILY_MASK(AF_UNIX),
297 },
298 [PACKET_DG_DB] = {
299 .states = (1 << SS_CLOSE),
300 .families = FAMILY_MASK(AF_PACKET),
301 },
302 [PACKET_R_DB] = {
303 .states = (1 << SS_CLOSE),
304 .families = FAMILY_MASK(AF_PACKET),
305 },
306 [NETLINK_DB] = {
307 .states = (1 << SS_CLOSE),
308 .families = FAMILY_MASK(AF_NETLINK),
309 },
310 [SCTP_DB] = {
311 .states = SS_CONN,
312 .families = FAMILY_MASK(AF_INET) | FAMILY_MASK(AF_INET6),
313 },
314 [VSOCK_ST_DB] = {
315 .states = SS_CONN,
316 .families = FAMILY_MASK(AF_VSOCK),
317 },
318 [VSOCK_DG_DB] = {
319 .states = SS_CONN,
320 .families = FAMILY_MASK(AF_VSOCK),
321 },
322 [TIPC_DB] = {
323 .states = TIPC_SS_CONN,
324 .families = FAMILY_MASK(AF_TIPC),
325 },
326 [XDP_DB] = {
327 .states = (1 << SS_CLOSE),
328 .families = FAMILY_MASK(AF_XDP),
329 },
330 };
331
332 static const struct filter default_afs[AF_MAX] = {
333 [AF_INET] = {
334 .dbs = INET_DBM,
335 .states = SS_CONN,
336 },
337 [AF_INET6] = {
338 .dbs = INET_DBM,
339 .states = SS_CONN,
340 },
341 [AF_UNIX] = {
342 .dbs = UNIX_DBM,
343 .states = SS_CONN,
344 },
345 [AF_PACKET] = {
346 .dbs = PACKET_DBM,
347 .states = (1 << SS_CLOSE),
348 },
349 [AF_NETLINK] = {
350 .dbs = (1 << NETLINK_DB),
351 .states = (1 << SS_CLOSE),
352 },
353 [AF_VSOCK] = {
354 .dbs = VSOCK_DBM,
355 .states = SS_CONN,
356 },
357 [AF_TIPC] = {
358 .dbs = (1 << TIPC_DB),
359 .states = TIPC_SS_CONN,
360 },
361 [AF_XDP] = {
362 .dbs = (1 << XDP_DB),
363 .states = (1 << SS_CLOSE),
364 },
365 };
366
367 static int do_default = 1;
368 static struct filter current_filter;
369
370 static void filter_db_set(struct filter *f, int db, bool enable)
371 {
372 if (enable) {
373 f->states |= default_dbs[db].states;
374 f->dbs |= 1 << db;
375 } else {
376 f->dbs &= ~(1 << db);
377 }
378 do_default = 0;
379 }
380
381 static int filter_db_parse(struct filter *f, const char *s)
382 {
383 const struct {
384 const char *name;
385 int dbs[MAX_DB + 1];
386 } db_name_tbl[] = {
387 #define ENTRY(name, ...) { #name, { __VA_ARGS__, MAX_DB } }
388 ENTRY(all, UDP_DB, DCCP_DB, TCP_DB, RAW_DB,
389 UNIX_ST_DB, UNIX_DG_DB, UNIX_SQ_DB,
390 PACKET_R_DB, PACKET_DG_DB, NETLINK_DB,
391 SCTP_DB, VSOCK_ST_DB, VSOCK_DG_DB, XDP_DB),
392 ENTRY(inet, UDP_DB, DCCP_DB, TCP_DB, SCTP_DB, RAW_DB),
393 ENTRY(udp, UDP_DB),
394 ENTRY(dccp, DCCP_DB),
395 ENTRY(tcp, TCP_DB),
396 ENTRY(sctp, SCTP_DB),
397 ENTRY(raw, RAW_DB),
398 ENTRY(unix, UNIX_ST_DB, UNIX_DG_DB, UNIX_SQ_DB),
399 ENTRY(unix_stream, UNIX_ST_DB),
400 ENTRY(u_str, UNIX_ST_DB), /* alias for unix_stream */
401 ENTRY(unix_dgram, UNIX_DG_DB),
402 ENTRY(u_dgr, UNIX_DG_DB), /* alias for unix_dgram */
403 ENTRY(unix_seqpacket, UNIX_SQ_DB),
404 ENTRY(u_seq, UNIX_SQ_DB), /* alias for unix_seqpacket */
405 ENTRY(packet, PACKET_R_DB, PACKET_DG_DB),
406 ENTRY(packet_raw, PACKET_R_DB),
407 ENTRY(p_raw, PACKET_R_DB), /* alias for packet_raw */
408 ENTRY(packet_dgram, PACKET_DG_DB),
409 ENTRY(p_dgr, PACKET_DG_DB), /* alias for packet_dgram */
410 ENTRY(netlink, NETLINK_DB),
411 ENTRY(vsock, VSOCK_ST_DB, VSOCK_DG_DB),
412 ENTRY(vsock_stream, VSOCK_ST_DB),
413 ENTRY(v_str, VSOCK_ST_DB), /* alias for vsock_stream */
414 ENTRY(vsock_dgram, VSOCK_DG_DB),
415 ENTRY(v_dgr, VSOCK_DG_DB), /* alias for vsock_dgram */
416 ENTRY(xdp, XDP_DB),
417 #undef ENTRY
418 };
419 bool enable = true;
420 unsigned int i;
421 const int *dbp;
422
423 if (s[0] == '!') {
424 enable = false;
425 s++;
426 }
427 for (i = 0; i < ARRAY_SIZE(db_name_tbl); i++) {
428 if (strcmp(s, db_name_tbl[i].name))
429 continue;
430 for (dbp = db_name_tbl[i].dbs; *dbp != MAX_DB; dbp++)
431 filter_db_set(f, *dbp, enable);
432 return 0;
433 }
434 return -1;
435 }
436
437 static void filter_af_set(struct filter *f, int af)
438 {
439 f->states |= default_afs[af].states;
440 f->families |= FAMILY_MASK(af);
441 do_default = 0;
442 preferred_family = af;
443 }
444
445 static int filter_af_get(struct filter *f, int af)
446 {
447 return !!(f->families & FAMILY_MASK(af));
448 }
449
450 static void filter_states_set(struct filter *f, int states)
451 {
452 if (states)
453 f->states = states;
454 }
455
456 static void filter_merge_defaults(struct filter *f)
457 {
458 int db;
459 int af;
460
461 for (db = 0; db < MAX_DB; db++) {
462 if (!(f->dbs & (1 << db)))
463 continue;
464
465 if (!(default_dbs[db].families & f->families))
466 f->families |= default_dbs[db].families;
467 }
468 for (af = 0; af < AF_MAX; af++) {
469 if (!(f->families & FAMILY_MASK(af)))
470 continue;
471
472 if (!(default_afs[af].dbs & f->dbs))
473 f->dbs |= default_afs[af].dbs;
474 }
475 }
476
477 static FILE *generic_proc_open(const char *env, const char *name)
478 {
479 const char *p = getenv(env);
480 char store[128];
481
482 if (!p) {
483 p = getenv("PROC_ROOT") ? : "/proc";
484 snprintf(store, sizeof(store)-1, "%s/%s", p, name);
485 p = store;
486 }
487
488 return fopen(p, "r");
489 }
490 #define net_tcp_open() generic_proc_open("PROC_NET_TCP", "net/tcp")
491 #define net_tcp6_open() generic_proc_open("PROC_NET_TCP6", "net/tcp6")
492 #define net_udp_open() generic_proc_open("PROC_NET_UDP", "net/udp")
493 #define net_udp6_open() generic_proc_open("PROC_NET_UDP6", "net/udp6")
494 #define net_raw_open() generic_proc_open("PROC_NET_RAW", "net/raw")
495 #define net_raw6_open() generic_proc_open("PROC_NET_RAW6", "net/raw6")
496 #define net_unix_open() generic_proc_open("PROC_NET_UNIX", "net/unix")
497 #define net_packet_open() generic_proc_open("PROC_NET_PACKET", \
498 "net/packet")
499 #define net_netlink_open() generic_proc_open("PROC_NET_NETLINK", \
500 "net/netlink")
501 #define net_sockstat_open() generic_proc_open("PROC_NET_SOCKSTAT", \
502 "net/sockstat")
503 #define net_sockstat6_open() generic_proc_open("PROC_NET_SOCKSTAT6", \
504 "net/sockstat6")
505 #define net_snmp_open() generic_proc_open("PROC_NET_SNMP", "net/snmp")
506 #define ephemeral_ports_open() generic_proc_open("PROC_IP_LOCAL_PORT_RANGE", \
507 "sys/net/ipv4/ip_local_port_range")
508
509 struct user_ent {
510 struct user_ent *next;
511 unsigned int ino;
512 int pid;
513 int fd;
514 char *process;
515 char *process_ctx;
516 char *socket_ctx;
517 };
518
519 #define USER_ENT_HASH_SIZE 256
520 static struct user_ent *user_ent_hash[USER_ENT_HASH_SIZE];
521
522 static int user_ent_hashfn(unsigned int ino)
523 {
524 int val = (ino >> 24) ^ (ino >> 16) ^ (ino >> 8) ^ ino;
525
526 return val & (USER_ENT_HASH_SIZE - 1);
527 }
528
529 static void user_ent_add(unsigned int ino, char *process,
530 int pid, int fd,
531 char *proc_ctx,
532 char *sock_ctx)
533 {
534 struct user_ent *p, **pp;
535
536 p = malloc(sizeof(struct user_ent));
537 if (!p) {
538 fprintf(stderr, "ss: failed to malloc buffer\n");
539 abort();
540 }
541 p->next = NULL;
542 p->ino = ino;
543 p->pid = pid;
544 p->fd = fd;
545 p->process = strdup(process);
546 p->process_ctx = strdup(proc_ctx);
547 p->socket_ctx = strdup(sock_ctx);
548
549 pp = &user_ent_hash[user_ent_hashfn(ino)];
550 p->next = *pp;
551 *pp = p;
552 }
553
554 static void user_ent_destroy(void)
555 {
556 struct user_ent *p, *p_next;
557 int cnt = 0;
558
559 while (cnt != USER_ENT_HASH_SIZE) {
560 p = user_ent_hash[cnt];
561 while (p) {
562 free(p->process);
563 free(p->process_ctx);
564 free(p->socket_ctx);
565 p_next = p->next;
566 free(p);
567 p = p_next;
568 }
569 cnt++;
570 }
571 }
572
573 static void user_ent_hash_build(void)
574 {
575 const char *root = getenv("PROC_ROOT") ? : "/proc/";
576 struct dirent *d;
577 char name[1024];
578 int nameoff;
579 DIR *dir;
580 char *pid_context;
581 char *sock_context;
582 const char *no_ctx = "unavailable";
583 static int user_ent_hash_build_init;
584
585 /* If show_users & show_proc_ctx set only do this once */
586 if (user_ent_hash_build_init != 0)
587 return;
588
589 user_ent_hash_build_init = 1;
590
591 strlcpy(name, root, sizeof(name));
592
593 if (strlen(name) == 0 || name[strlen(name)-1] != '/')
594 strcat(name, "/");
595
596 nameoff = strlen(name);
597
598 dir = opendir(name);
599 if (!dir)
600 return;
601
602 while ((d = readdir(dir)) != NULL) {
603 struct dirent *d1;
604 char process[16];
605 char *p;
606 int pid, pos;
607 DIR *dir1;
608 char crap;
609
610 if (sscanf(d->d_name, "%d%c", &pid, &crap) != 1)
611 continue;
612
613 if (getpidcon(pid, &pid_context) != 0)
614 pid_context = strdup(no_ctx);
615
616 snprintf(name + nameoff, sizeof(name) - nameoff, "%d/fd/", pid);
617 pos = strlen(name);
618 if ((dir1 = opendir(name)) == NULL) {
619 free(pid_context);
620 continue;
621 }
622
623 process[0] = '\0';
624 p = process;
625
626 while ((d1 = readdir(dir1)) != NULL) {
627 const char *pattern = "socket:[";
628 unsigned int ino;
629 char lnk[64];
630 int fd;
631 ssize_t link_len;
632 char tmp[1024];
633
634 if (sscanf(d1->d_name, "%d%c", &fd, &crap) != 1)
635 continue;
636
637 snprintf(name+pos, sizeof(name) - pos, "%d", fd);
638
639 link_len = readlink(name, lnk, sizeof(lnk)-1);
640 if (link_len == -1)
641 continue;
642 lnk[link_len] = '\0';
643
644 if (strncmp(lnk, pattern, strlen(pattern)))
645 continue;
646
647 sscanf(lnk, "socket:[%u]", &ino);
648
649 snprintf(tmp, sizeof(tmp), "%s/%d/fd/%s",
650 root, pid, d1->d_name);
651
652 if (getfilecon(tmp, &sock_context) <= 0)
653 sock_context = strdup(no_ctx);
654
655 if (*p == '\0') {
656 FILE *fp;
657
658 snprintf(tmp, sizeof(tmp), "%s/%d/stat",
659 root, pid);
660 if ((fp = fopen(tmp, "r")) != NULL) {
661 if (fscanf(fp, "%*d (%[^)])", p) < 1)
662 ; /* ignore */
663 fclose(fp);
664 }
665 }
666 user_ent_add(ino, p, pid, fd,
667 pid_context, sock_context);
668 free(sock_context);
669 }
670 free(pid_context);
671 closedir(dir1);
672 }
673 closedir(dir);
674 }
675
676 enum entry_types {
677 USERS,
678 PROC_CTX,
679 PROC_SOCK_CTX
680 };
681
682 #define ENTRY_BUF_SIZE 512
683 static int find_entry(unsigned int ino, char **buf, int type)
684 {
685 struct user_ent *p;
686 int cnt = 0;
687 char *ptr;
688 char *new_buf;
689 int len, new_buf_len;
690 int buf_used = 0;
691 int buf_len = 0;
692
693 if (!ino)
694 return 0;
695
696 p = user_ent_hash[user_ent_hashfn(ino)];
697 ptr = *buf = NULL;
698 while (p) {
699 if (p->ino != ino)
700 goto next;
701
702 while (1) {
703 ptr = *buf + buf_used;
704 switch (type) {
705 case USERS:
706 len = snprintf(ptr, buf_len - buf_used,
707 "(\"%s\",pid=%d,fd=%d),",
708 p->process, p->pid, p->fd);
709 break;
710 case PROC_CTX:
711 len = snprintf(ptr, buf_len - buf_used,
712 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d),",
713 p->process, p->pid,
714 p->process_ctx, p->fd);
715 break;
716 case PROC_SOCK_CTX:
717 len = snprintf(ptr, buf_len - buf_used,
718 "(\"%s\",pid=%d,proc_ctx=%s,fd=%d,sock_ctx=%s),",
719 p->process, p->pid,
720 p->process_ctx, p->fd,
721 p->socket_ctx);
722 break;
723 default:
724 fprintf(stderr, "ss: invalid type: %d\n", type);
725 abort();
726 }
727
728 if (len < 0 || len >= buf_len - buf_used) {
729 new_buf_len = buf_len + ENTRY_BUF_SIZE;
730 new_buf = realloc(*buf, new_buf_len);
731 if (!new_buf) {
732 fprintf(stderr, "ss: failed to malloc buffer\n");
733 abort();
734 }
735 *buf = new_buf;
736 buf_len = new_buf_len;
737 continue;
738 } else {
739 buf_used += len;
740 break;
741 }
742 }
743 cnt++;
744 next:
745 p = p->next;
746 }
747 if (buf_used) {
748 ptr = *buf + buf_used;
749 ptr[-1] = '\0';
750 }
751 return cnt;
752 }
753
754 static unsigned long long cookie_sk_get(const uint32_t *cookie)
755 {
756 return (((unsigned long long)cookie[1] << 31) << 1) | cookie[0];
757 }
758
759 static const char *sctp_sstate_name[] = {
760 [SCTP_STATE_CLOSED] = "CLOSED",
761 [SCTP_STATE_COOKIE_WAIT] = "COOKIE_WAIT",
762 [SCTP_STATE_COOKIE_ECHOED] = "COOKIE_ECHOED",
763 [SCTP_STATE_ESTABLISHED] = "ESTAB",
764 [SCTP_STATE_SHUTDOWN_PENDING] = "SHUTDOWN_PENDING",
765 [SCTP_STATE_SHUTDOWN_SENT] = "SHUTDOWN_SENT",
766 [SCTP_STATE_SHUTDOWN_RECEIVED] = "SHUTDOWN_RECEIVED",
767 [SCTP_STATE_SHUTDOWN_ACK_SENT] = "ACK_SENT",
768 };
769
770 static const char * const stype_nameg[] = {
771 "UNKNOWN",
772 [SOCK_STREAM] = "STREAM",
773 [SOCK_DGRAM] = "DGRAM",
774 [SOCK_RDM] = "RDM",
775 [SOCK_SEQPACKET] = "SEQPACKET",
776 };
777
778 struct sockstat {
779 struct sockstat *next;
780 unsigned int type;
781 uint16_t prot;
782 uint16_t raw_prot;
783 inet_prefix local;
784 inet_prefix remote;
785 int lport;
786 int rport;
787 int state;
788 int rq, wq;
789 unsigned int ino;
790 unsigned int uid;
791 int refcnt;
792 unsigned int iface;
793 unsigned long long sk;
794 char *name;
795 char *peer_name;
796 __u32 mark;
797 };
798
799 struct dctcpstat {
800 unsigned int ce_state;
801 unsigned int alpha;
802 unsigned int ab_ecn;
803 unsigned int ab_tot;
804 bool enabled;
805 };
806
807 struct tcpstat {
808 struct sockstat ss;
809 unsigned int timer;
810 unsigned int timeout;
811 int probes;
812 char cong_alg[16];
813 double rto, ato, rtt, rttvar;
814 int qack, ssthresh, backoff;
815 double send_bps;
816 int snd_wscale;
817 int rcv_wscale;
818 int mss;
819 int rcv_mss;
820 int advmss;
821 unsigned int pmtu;
822 unsigned int cwnd;
823 unsigned int lastsnd;
824 unsigned int lastrcv;
825 unsigned int lastack;
826 double pacing_rate;
827 double pacing_rate_max;
828 double delivery_rate;
829 unsigned long long bytes_acked;
830 unsigned long long bytes_received;
831 unsigned int segs_out;
832 unsigned int segs_in;
833 unsigned int data_segs_out;
834 unsigned int data_segs_in;
835 unsigned int unacked;
836 unsigned int retrans;
837 unsigned int retrans_total;
838 unsigned int lost;
839 unsigned int sacked;
840 unsigned int fackets;
841 unsigned int reordering;
842 unsigned int not_sent;
843 unsigned int delivered;
844 unsigned int delivered_ce;
845 unsigned int dsack_dups;
846 unsigned int reord_seen;
847 double rcv_rtt;
848 double min_rtt;
849 int rcv_space;
850 unsigned int rcv_ssthresh;
851 unsigned long long busy_time;
852 unsigned long long rwnd_limited;
853 unsigned long long sndbuf_limited;
854 unsigned long long bytes_sent;
855 unsigned long long bytes_retrans;
856 bool has_ts_opt;
857 bool has_sack_opt;
858 bool has_ecn_opt;
859 bool has_ecnseen_opt;
860 bool has_fastopen_opt;
861 bool has_wscale_opt;
862 bool app_limited;
863 struct dctcpstat *dctcp;
864 struct tcp_bbr_info *bbr_info;
865 };
866
867 /* SCTP assocs share the same inode number with their parent endpoint. So if we
868 * have seen the inode number before, it must be an assoc instead of the next
869 * endpoint. */
870 static bool is_sctp_assoc(struct sockstat *s, const char *sock_name)
871 {
872 if (strcmp(sock_name, "sctp"))
873 return false;
874 if (!sctp_ino || sctp_ino != s->ino)
875 return false;
876 return true;
877 }
878
879 static const char *unix_netid_name(int type)
880 {
881 switch (type) {
882 case SOCK_STREAM:
883 return "u_str";
884 case SOCK_SEQPACKET:
885 return "u_seq";
886 case SOCK_DGRAM:
887 default:
888 return "u_dgr";
889 }
890 }
891
892 static const char *proto_name(int protocol)
893 {
894 switch (protocol) {
895 case 0:
896 return "raw";
897 case IPPROTO_UDP:
898 return "udp";
899 case IPPROTO_TCP:
900 return "tcp";
901 case IPPROTO_SCTP:
902 return "sctp";
903 case IPPROTO_DCCP:
904 return "dccp";
905 case IPPROTO_ICMPV6:
906 return "icmp6";
907 }
908
909 return "???";
910 }
911
912 static const char *vsock_netid_name(int type)
913 {
914 switch (type) {
915 case SOCK_STREAM:
916 return "v_str";
917 case SOCK_DGRAM:
918 return "v_dgr";
919 default:
920 return "???";
921 }
922 }
923
924 static const char *tipc_netid_name(int type)
925 {
926 switch (type) {
927 case SOCK_STREAM:
928 return "ti_st";
929 case SOCK_DGRAM:
930 return "ti_dg";
931 case SOCK_RDM:
932 return "ti_rd";
933 case SOCK_SEQPACKET:
934 return "ti_sq";
935 default:
936 return "???";
937 }
938 }
939
940 /* Allocate and initialize a new buffer chunk */
941 static struct buf_chunk *buf_chunk_new(void)
942 {
943 struct buf_chunk *new = malloc(BUF_CHUNK);
944
945 if (!new)
946 abort();
947
948 new->next = NULL;
949
950 /* This is also the last block */
951 buffer.tail = new;
952
953 /* Next token will be stored at the beginning of chunk data area, and
954 * its initial length is zero.
955 */
956 buffer.cur = (struct buf_token *)new->data;
957 buffer.cur->len = 0;
958
959 new->end = buffer.cur->data;
960
961 buffer.chunks++;
962
963 return new;
964 }
965
966 /* Return available tail room in given chunk */
967 static int buf_chunk_avail(struct buf_chunk *chunk)
968 {
969 return BUF_CHUNK - offsetof(struct buf_chunk, data) -
970 (chunk->end - chunk->data);
971 }
972
973 /* Update end pointer and token length, link new chunk if we hit the end of the
974 * current one. Return -EAGAIN if we got a new chunk, caller has to print again.
975 */
976 static int buf_update(int len)
977 {
978 struct buf_chunk *chunk = buffer.tail;
979 struct buf_token *t = buffer.cur;
980
981 /* Claim success if new content fits in the current chunk, and anyway
982 * if this is the first token in the chunk: in the latter case,
983 * allocating a new chunk won't help, so we'll just cut the output.
984 */
985 if ((len < buf_chunk_avail(chunk) && len != -1 /* glibc < 2.0.6 */) ||
986 t == (struct buf_token *)chunk->data) {
987 len = min(len, buf_chunk_avail(chunk));
988
989 /* Total field length can't exceed 2^16 bytes, cut as needed */
990 len = min(len, USHRT_MAX - t->len);
991
992 chunk->end += len;
993 t->len += len;
994 return 0;
995 }
996
997 /* Content truncated, time to allocate more */
998 chunk->next = buf_chunk_new();
999
1000 /* Copy current token over to new chunk, including length descriptor */
1001 memcpy(chunk->next->data, t, sizeof(t->len) + t->len);
1002 chunk->next->end += t->len;
1003
1004 /* Discard partially written field in old chunk */
1005 chunk->end -= t->len + sizeof(t->len);
1006
1007 return -EAGAIN;
1008 }
1009
1010 /* Append content to buffer as part of the current field */
1011 __attribute__((format(printf, 1, 2)))
1012 static void out(const char *fmt, ...)
1013 {
1014 struct column *f = current_field;
1015 va_list args;
1016 char *pos;
1017 int len;
1018
1019 if (f->disabled)
1020 return;
1021
1022 if (!buffer.head)
1023 buffer.head = buf_chunk_new();
1024
1025 again: /* Append to buffer: if we have a new chunk, print again */
1026
1027 pos = buffer.cur->data + buffer.cur->len;
1028 va_start(args, fmt);
1029
1030 /* Limit to tail room. If we hit the limit, buf_update() will tell us */
1031 len = vsnprintf(pos, buf_chunk_avail(buffer.tail), fmt, args);
1032 va_end(args);
1033
1034 if (buf_update(len))
1035 goto again;
1036 }
1037
1038 static int print_left_spacing(struct column *f, int stored, int printed)
1039 {
1040 int s;
1041
1042 if (!f->width || f->align == ALIGN_LEFT)
1043 return 0;
1044
1045 s = f->width - stored - printed;
1046 if (f->align == ALIGN_CENTER)
1047 /* If count of total spacing is odd, shift right by one */
1048 s = (s + 1) / 2;
1049
1050 if (s > 0)
1051 return printf("%*c", s, ' ');
1052
1053 return 0;
1054 }
1055
1056 static void print_right_spacing(struct column *f, int printed)
1057 {
1058 int s;
1059
1060 if (!f->width || f->align == ALIGN_RIGHT)
1061 return;
1062
1063 s = f->width - printed;
1064 if (f->align == ALIGN_CENTER)
1065 s /= 2;
1066
1067 if (s > 0)
1068 printf("%*c", s, ' ');
1069 }
1070
1071 /* Done with field: update buffer pointer, start new token after current one */
1072 static void field_flush(struct column *f)
1073 {
1074 struct buf_chunk *chunk;
1075 unsigned int pad;
1076
1077 if (f->disabled)
1078 return;
1079
1080 chunk = buffer.tail;
1081 pad = buffer.cur->len % 2;
1082
1083 if (buffer.cur->len > f->max_len)
1084 f->max_len = buffer.cur->len;
1085
1086 /* We need a new chunk if we can't store the next length descriptor.
1087 * Mind the gap between end of previous token and next aligned position
1088 * for length descriptor.
1089 */
1090 if (buf_chunk_avail(chunk) - pad < sizeof(buffer.cur->len)) {
1091 chunk->end += pad;
1092 chunk->next = buf_chunk_new();
1093 return;
1094 }
1095
1096 buffer.cur = (struct buf_token *)(buffer.cur->data +
1097 LEN_ALIGN(buffer.cur->len));
1098 buffer.cur->len = 0;
1099 buffer.tail->end = buffer.cur->data;
1100 }
1101
1102 static int field_is_last(struct column *f)
1103 {
1104 return f - columns == COL_MAX - 1;
1105 }
1106
1107 /* Get the next available token in the buffer starting from the current token */
1108 static struct buf_token *buf_token_next(struct buf_token *cur)
1109 {
1110 struct buf_chunk *chunk = buffer.tail;
1111
1112 /* If we reached the end of chunk contents, get token from next chunk */
1113 if (cur->data + LEN_ALIGN(cur->len) == chunk->end) {
1114 buffer.tail = chunk = chunk->next;
1115 return chunk ? (struct buf_token *)chunk->data : NULL;
1116 }
1117
1118 return (struct buf_token *)(cur->data + LEN_ALIGN(cur->len));
1119 }
1120
1121 /* Free up all allocated buffer chunks */
1122 static void buf_free_all(void)
1123 {
1124 struct buf_chunk *tmp;
1125
1126 for (buffer.tail = buffer.head; buffer.tail; ) {
1127 tmp = buffer.tail;
1128 buffer.tail = buffer.tail->next;
1129 free(tmp);
1130 }
1131 buffer.head = NULL;
1132 buffer.chunks = 0;
1133 }
1134
1135 /* Get current screen width, default to 80 columns if TIOCGWINSZ fails */
1136 static int render_screen_width(void)
1137 {
1138 int width = 80;
1139
1140 if (isatty(STDOUT_FILENO)) {
1141 struct winsize w;
1142
1143 if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) != -1) {
1144 if (w.ws_col > 0)
1145 width = w.ws_col;
1146 }
1147 }
1148
1149 return width;
1150 }
1151
1152 /* Calculate column width from contents length. If columns don't fit on one
1153 * line, break them into the least possible amount of lines and keep them
1154 * aligned across lines. Available screen space is equally spread between fields
1155 * as additional spacing.
1156 */
1157 static void render_calc_width(void)
1158 {
1159 int screen_width = render_screen_width();
1160 struct column *c, *eol = columns - 1;
1161 int first, len = 0, linecols = 0;
1162
1163 /* First pass: set width for each column to measured content length */
1164 for (first = 1, c = columns; c - columns < COL_MAX; c++) {
1165 if (c->disabled)
1166 continue;
1167
1168 if (!first && c->max_len)
1169 c->width = c->max_len + strlen(c->ldelim);
1170 else
1171 c->width = c->max_len;
1172
1173 /* But don't exceed screen size. If we exceed the screen size
1174 * for even a single field, it will just start on a line of its
1175 * own and then naturally wrap.
1176 */
1177 c->width = min(c->width, screen_width);
1178
1179 if (c->width)
1180 first = 0;
1181 }
1182
1183 /* Second pass: find out newlines and distribute available spacing */
1184 for (c = columns; c - columns < COL_MAX; c++) {
1185 int pad, spacing, rem, last;
1186 struct column *tmp;
1187
1188 if (!c->width)
1189 continue;
1190
1191 linecols++;
1192 len += c->width;
1193
1194 for (last = 1, tmp = c + 1; tmp - columns < COL_MAX; tmp++) {
1195 if (tmp->width) {
1196 last = 0;
1197 break;
1198 }
1199 }
1200
1201 if (!last && len < screen_width) {
1202 /* Columns fit on screen so far, nothing to do yet */
1203 continue;
1204 }
1205
1206 if (len == screen_width) {
1207 /* Exact fit, just start with new line */
1208 goto newline;
1209 }
1210
1211 if (len > screen_width) {
1212 /* Screen width exceeded: go back one column */
1213 len -= c->width;
1214 c--;
1215 linecols--;
1216 }
1217
1218 /* Distribute remaining space to columns on this line */
1219 pad = screen_width - len;
1220 spacing = pad / linecols;
1221 rem = pad % linecols;
1222 for (tmp = c; tmp > eol; tmp--) {
1223 if (!tmp->width)
1224 continue;
1225
1226 tmp->width += spacing;
1227 if (rem) {
1228 tmp->width++;
1229 rem--;
1230 }
1231 }
1232
1233 newline:
1234 /* Line break: reset line counters, mark end-of-line */
1235 eol = c;
1236 len = 0;
1237 linecols = 0;
1238 }
1239 }
1240
1241 /* Render buffered output with spacing and delimiters, then free up buffers */
1242 static void render(void)
1243 {
1244 struct buf_token *token;
1245 int printed, line_started = 0;
1246 struct column *f;
1247
1248 if (!buffer.head)
1249 return;
1250
1251 token = (struct buf_token *)buffer.head->data;
1252
1253 /* Ensure end alignment of last token, it wasn't necessarily flushed */
1254 buffer.tail->end += buffer.cur->len % 2;
1255
1256 render_calc_width();
1257
1258 /* Rewind and replay */
1259 buffer.tail = buffer.head;
1260
1261 f = columns;
1262 while (!f->width)
1263 f++;
1264
1265 while (token) {
1266 /* Print left delimiter only if we already started a line */
1267 if (line_started++)
1268 printed = printf("%s", f->ldelim);
1269 else
1270 printed = 0;
1271
1272 /* Print field content from token data with spacing */
1273 printed += print_left_spacing(f, token->len, printed);
1274 printed += fwrite(token->data, 1, token->len, stdout);
1275 print_right_spacing(f, printed);
1276
1277 /* Go to next non-empty field, deal with end-of-line */
1278 do {
1279 if (field_is_last(f)) {
1280 printf("\n");
1281 f = columns;
1282 line_started = 0;
1283 } else {
1284 f++;
1285 }
1286 } while (f->disabled);
1287
1288 token = buf_token_next(token);
1289 }
1290
1291 buf_free_all();
1292 current_field = columns;
1293 }
1294
1295 /* Move to next field, and render buffer if we reached the maximum number of
1296 * chunks, at the last field in a line.
1297 */
1298 static void field_next(void)
1299 {
1300 if (field_is_last(current_field) && buffer.chunks >= BUF_CHUNKS_MAX) {
1301 render();
1302 return;
1303 }
1304
1305 field_flush(current_field);
1306 if (field_is_last(current_field))
1307 current_field = columns;
1308 else
1309 current_field++;
1310 }
1311
1312 /* Walk through fields and flush them until we reach the desired one */
1313 static void field_set(enum col_id id)
1314 {
1315 while (id != current_field - columns)
1316 field_next();
1317 }
1318
1319 /* Print header for all non-empty columns */
1320 static void print_header(void)
1321 {
1322 while (!field_is_last(current_field)) {
1323 if (!current_field->disabled)
1324 out("%s", current_field->header);
1325 field_next();
1326 }
1327 }
1328
1329 static void sock_state_print(struct sockstat *s)
1330 {
1331 const char *sock_name;
1332 static const char * const sstate_name[] = {
1333 "UNKNOWN",
1334 [SS_ESTABLISHED] = "ESTAB",
1335 [SS_SYN_SENT] = "SYN-SENT",
1336 [SS_SYN_RECV] = "SYN-RECV",
1337 [SS_FIN_WAIT1] = "FIN-WAIT-1",
1338 [SS_FIN_WAIT2] = "FIN-WAIT-2",
1339 [SS_TIME_WAIT] = "TIME-WAIT",
1340 [SS_CLOSE] = "UNCONN",
1341 [SS_CLOSE_WAIT] = "CLOSE-WAIT",
1342 [SS_LAST_ACK] = "LAST-ACK",
1343 [SS_LISTEN] = "LISTEN",
1344 [SS_CLOSING] = "CLOSING",
1345 };
1346
1347 switch (s->local.family) {
1348 case AF_UNIX:
1349 sock_name = unix_netid_name(s->type);
1350 break;
1351 case AF_INET:
1352 case AF_INET6:
1353 sock_name = proto_name(s->type);
1354 break;
1355 case AF_PACKET:
1356 sock_name = s->type == SOCK_RAW ? "p_raw" : "p_dgr";
1357 break;
1358 case AF_NETLINK:
1359 sock_name = "nl";
1360 break;
1361 case AF_TIPC:
1362 sock_name = tipc_netid_name(s->type);
1363 break;
1364 case AF_VSOCK:
1365 sock_name = vsock_netid_name(s->type);
1366 break;
1367 case AF_XDP:
1368 sock_name = "xdp";
1369 break;
1370 default:
1371 sock_name = "unknown";
1372 }
1373
1374 if (is_sctp_assoc(s, sock_name)) {
1375 field_set(COL_STATE); /* Empty Netid field */
1376 out("`- %s", sctp_sstate_name[s->state]);
1377 } else {
1378 field_set(COL_NETID);
1379 out("%s", sock_name);
1380 field_set(COL_STATE);
1381 out("%s", sstate_name[s->state]);
1382 }
1383
1384 field_set(COL_RECVQ);
1385 out("%-6d", s->rq);
1386 field_set(COL_SENDQ);
1387 out("%-6d", s->wq);
1388 field_set(COL_ADDR);
1389 }
1390
1391 static void sock_details_print(struct sockstat *s)
1392 {
1393 if (s->uid)
1394 out(" uid:%u", s->uid);
1395
1396 out(" ino:%u", s->ino);
1397 out(" sk:%llx", s->sk);
1398
1399 if (s->mark)
1400 out(" fwmark:0x%x", s->mark);
1401 }
1402
1403 static void sock_addr_print(const char *addr, char *delim, const char *port,
1404 const char *ifname)
1405 {
1406 if (ifname)
1407 out("%s" "%%" "%s%s", addr, ifname, delim);
1408 else
1409 out("%s%s", addr, delim);
1410
1411 field_next();
1412 out("%s", port);
1413 field_next();
1414 }
1415
1416 static const char *print_ms_timer(unsigned int timeout)
1417 {
1418 static char buf[64];
1419 int secs, msecs, minutes;
1420
1421 secs = timeout/1000;
1422 minutes = secs/60;
1423 secs = secs%60;
1424 msecs = timeout%1000;
1425 buf[0] = 0;
1426 if (minutes) {
1427 msecs = 0;
1428 snprintf(buf, sizeof(buf)-16, "%dmin", minutes);
1429 if (minutes > 9)
1430 secs = 0;
1431 }
1432 if (secs) {
1433 if (secs > 9)
1434 msecs = 0;
1435 sprintf(buf+strlen(buf), "%d%s", secs, msecs ? "." : "sec");
1436 }
1437 if (msecs)
1438 sprintf(buf+strlen(buf), "%03dms", msecs);
1439 return buf;
1440 }
1441
1442 struct scache {
1443 struct scache *next;
1444 int port;
1445 char *name;
1446 const char *proto;
1447 };
1448
1449 static struct scache *rlist;
1450
1451 static void init_service_resolver(void)
1452 {
1453 char buf[128];
1454 FILE *fp = popen("/usr/sbin/rpcinfo -p 2>/dev/null", "r");
1455
1456 if (!fp)
1457 return;
1458
1459 if (!fgets(buf, sizeof(buf), fp)) {
1460 pclose(fp);
1461 return;
1462 }
1463 while (fgets(buf, sizeof(buf), fp) != NULL) {
1464 unsigned int progn, port;
1465 char proto[128], prog[128] = "rpc.";
1466 struct scache *c;
1467
1468 if (sscanf(buf, "%u %*d %s %u %s",
1469 &progn, proto, &port, prog+4) != 4)
1470 continue;
1471
1472 if (!(c = malloc(sizeof(*c))))
1473 continue;
1474
1475 c->port = port;
1476 c->name = strdup(prog);
1477 if (strcmp(proto, TCP_PROTO) == 0)
1478 c->proto = TCP_PROTO;
1479 else if (strcmp(proto, UDP_PROTO) == 0)
1480 c->proto = UDP_PROTO;
1481 else if (strcmp(proto, SCTP_PROTO) == 0)
1482 c->proto = SCTP_PROTO;
1483 else
1484 c->proto = NULL;
1485 c->next = rlist;
1486 rlist = c;
1487 }
1488 pclose(fp);
1489 }
1490
1491 /* Even do not try default linux ephemeral port ranges:
1492 * default /etc/services contains so much of useless crap
1493 * wouldbe "allocated" to this area that resolution
1494 * is really harmful. I shrug each time when seeing
1495 * "socks" or "cfinger" in dumps.
1496 */
1497 static int is_ephemeral(int port)
1498 {
1499 static int min = 0, max;
1500
1501 if (!min) {
1502 FILE *f = ephemeral_ports_open();
1503
1504 if (!f || fscanf(f, "%d %d", &min, &max) < 2) {
1505 min = 1024;
1506 max = 4999;
1507 }
1508 if (f)
1509 fclose(f);
1510 }
1511 return port >= min && port <= max;
1512 }
1513
1514
1515 static const char *__resolve_service(int port)
1516 {
1517 struct scache *c;
1518
1519 for (c = rlist; c; c = c->next) {
1520 if (c->port == port && c->proto == dg_proto)
1521 return c->name;
1522 }
1523
1524 if (!is_ephemeral(port)) {
1525 static int notfirst;
1526 struct servent *se;
1527
1528 if (!notfirst) {
1529 setservent(1);
1530 notfirst = 1;
1531 }
1532 se = getservbyport(htons(port), dg_proto);
1533 if (se)
1534 return se->s_name;
1535 }
1536
1537 return NULL;
1538 }
1539
1540 #define SCACHE_BUCKETS 1024
1541 static struct scache *cache_htab[SCACHE_BUCKETS];
1542
1543 static const char *resolve_service(int port)
1544 {
1545 static char buf[128];
1546 struct scache *c;
1547 const char *res;
1548 int hash;
1549
1550 if (port == 0) {
1551 buf[0] = '*';
1552 buf[1] = 0;
1553 return buf;
1554 }
1555
1556 if (numeric)
1557 goto do_numeric;
1558
1559 if (dg_proto == RAW_PROTO)
1560 return inet_proto_n2a(port, buf, sizeof(buf));
1561
1562
1563 hash = (port^(((unsigned long)dg_proto)>>2)) % SCACHE_BUCKETS;
1564
1565 for (c = cache_htab[hash]; c; c = c->next) {
1566 if (c->port == port && c->proto == dg_proto)
1567 goto do_cache;
1568 }
1569
1570 c = malloc(sizeof(*c));
1571 if (!c)
1572 goto do_numeric;
1573 res = __resolve_service(port);
1574 c->port = port;
1575 c->name = res ? strdup(res) : NULL;
1576 c->proto = dg_proto;
1577 c->next = cache_htab[hash];
1578 cache_htab[hash] = c;
1579
1580 do_cache:
1581 if (c->name)
1582 return c->name;
1583
1584 do_numeric:
1585 sprintf(buf, "%u", port);
1586 return buf;
1587 }
1588
1589 static void inet_addr_print(const inet_prefix *a, int port,
1590 unsigned int ifindex, bool v6only)
1591 {
1592 char buf[1024];
1593 const char *ap = buf;
1594 const char *ifname = NULL;
1595
1596 if (a->family == AF_INET) {
1597 ap = format_host(AF_INET, 4, a->data);
1598 } else {
1599 if (!v6only &&
1600 !memcmp(a->data, &in6addr_any, sizeof(in6addr_any))) {
1601 buf[0] = '*';
1602 buf[1] = 0;
1603 } else {
1604 ap = format_host(a->family, 16, a->data);
1605
1606 /* Numeric IPv6 addresses should be bracketed */
1607 if (strchr(ap, ':')) {
1608 snprintf(buf, sizeof(buf),
1609 "[%s]", ap);
1610 ap = buf;
1611 }
1612 }
1613 }
1614
1615 if (ifindex)
1616 ifname = ll_index_to_name(ifindex);
1617
1618 sock_addr_print(ap, ":", resolve_service(port), ifname);
1619 }
1620
1621 struct aafilter {
1622 inet_prefix addr;
1623 int port;
1624 unsigned int iface;
1625 __u32 mark;
1626 __u32 mask;
1627 struct aafilter *next;
1628 };
1629
1630 static int inet2_addr_match(const inet_prefix *a, const inet_prefix *p,
1631 int plen)
1632 {
1633 if (!inet_addr_match(a, p, plen))
1634 return 0;
1635
1636 /* Cursed "v4 mapped" addresses: v4 mapped socket matches
1637 * pure IPv4 rule, but v4-mapped rule selects only v4-mapped
1638 * sockets. Fair? */
1639 if (p->family == AF_INET && a->family == AF_INET6) {
1640 if (a->data[0] == 0 && a->data[1] == 0 &&
1641 a->data[2] == htonl(0xffff)) {
1642 inet_prefix tmp = *a;
1643
1644 tmp.data[0] = a->data[3];
1645 return inet_addr_match(&tmp, p, plen);
1646 }
1647 }
1648 return 1;
1649 }
1650
1651 static int unix_match(const inet_prefix *a, const inet_prefix *p)
1652 {
1653 char *addr, *pattern;
1654
1655 memcpy(&addr, a->data, sizeof(addr));
1656 memcpy(&pattern, p->data, sizeof(pattern));
1657 if (pattern == NULL)
1658 return 1;
1659 if (addr == NULL)
1660 addr = "";
1661 return !fnmatch(pattern, addr, 0);
1662 }
1663
1664 static int run_ssfilter(struct ssfilter *f, struct sockstat *s)
1665 {
1666 switch (f->type) {
1667 case SSF_S_AUTO:
1668 {
1669 if (s->local.family == AF_UNIX) {
1670 char *p;
1671
1672 memcpy(&p, s->local.data, sizeof(p));
1673 return p == NULL || (p[0] == '@' && strlen(p) == 6 &&
1674 strspn(p+1, "0123456789abcdef") == 5);
1675 }
1676 if (s->local.family == AF_PACKET)
1677 return s->lport == 0 && s->local.data[0] == 0;
1678 if (s->local.family == AF_NETLINK)
1679 return s->lport < 0;
1680 if (s->local.family == AF_VSOCK)
1681 return s->lport > 1023;
1682
1683 return is_ephemeral(s->lport);
1684 }
1685 case SSF_DCOND:
1686 {
1687 struct aafilter *a = (void *)f->pred;
1688
1689 if (a->addr.family == AF_UNIX)
1690 return unix_match(&s->remote, &a->addr);
1691 if (a->port != -1 && a->port != s->rport)
1692 return 0;
1693 if (a->addr.bitlen) {
1694 do {
1695 if (!inet2_addr_match(&s->remote, &a->addr, a->addr.bitlen))
1696 return 1;
1697 } while ((a = a->next) != NULL);
1698 return 0;
1699 }
1700 return 1;
1701 }
1702 case SSF_SCOND:
1703 {
1704 struct aafilter *a = (void *)f->pred;
1705
1706 if (a->addr.family == AF_UNIX)
1707 return unix_match(&s->local, &a->addr);
1708 if (a->port != -1 && a->port != s->lport)
1709 return 0;
1710 if (a->addr.bitlen) {
1711 do {
1712 if (!inet2_addr_match(&s->local, &a->addr, a->addr.bitlen))
1713 return 1;
1714 } while ((a = a->next) != NULL);
1715 return 0;
1716 }
1717 return 1;
1718 }
1719 case SSF_D_GE:
1720 {
1721 struct aafilter *a = (void *)f->pred;
1722
1723 return s->rport >= a->port;
1724 }
1725 case SSF_D_LE:
1726 {
1727 struct aafilter *a = (void *)f->pred;
1728
1729 return s->rport <= a->port;
1730 }
1731 case SSF_S_GE:
1732 {
1733 struct aafilter *a = (void *)f->pred;
1734
1735 return s->lport >= a->port;
1736 }
1737 case SSF_S_LE:
1738 {
1739 struct aafilter *a = (void *)f->pred;
1740
1741 return s->lport <= a->port;
1742 }
1743 case SSF_DEVCOND:
1744 {
1745 struct aafilter *a = (void *)f->pred;
1746
1747 return s->iface == a->iface;
1748 }
1749 case SSF_MARKMASK:
1750 {
1751 struct aafilter *a = (void *)f->pred;
1752
1753 return (s->mark & a->mask) == a->mark;
1754 }
1755 /* Yup. It is recursion. Sorry. */
1756 case SSF_AND:
1757 return run_ssfilter(f->pred, s) && run_ssfilter(f->post, s);
1758 case SSF_OR:
1759 return run_ssfilter(f->pred, s) || run_ssfilter(f->post, s);
1760 case SSF_NOT:
1761 return !run_ssfilter(f->pred, s);
1762 default:
1763 abort();
1764 }
1765 }
1766
1767 /* Relocate external jumps by reloc. */
1768 static void ssfilter_patch(char *a, int len, int reloc)
1769 {
1770 while (len > 0) {
1771 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)a;
1772
1773 if (op->no == len+4)
1774 op->no += reloc;
1775 len -= op->yes;
1776 a += op->yes;
1777 }
1778 if (len < 0)
1779 abort();
1780 }
1781
1782 static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode)
1783 {
1784 switch (f->type) {
1785 case SSF_S_AUTO:
1786 {
1787 if (!(*bytecode = malloc(4))) abort();
1788 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_AUTO, 4, 8 };
1789 return 4;
1790 }
1791 case SSF_DCOND:
1792 case SSF_SCOND:
1793 {
1794 struct aafilter *a = (void *)f->pred;
1795 struct aafilter *b;
1796 char *ptr;
1797 int code = (f->type == SSF_DCOND ? INET_DIAG_BC_D_COND : INET_DIAG_BC_S_COND);
1798 int len = 0;
1799
1800 for (b = a; b; b = b->next) {
1801 len += 4 + sizeof(struct inet_diag_hostcond);
1802 if (a->addr.family == AF_INET6)
1803 len += 16;
1804 else
1805 len += 4;
1806 if (b->next)
1807 len += 4;
1808 }
1809 if (!(ptr = malloc(len))) abort();
1810 *bytecode = ptr;
1811 for (b = a; b; b = b->next) {
1812 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)ptr;
1813 int alen = (a->addr.family == AF_INET6 ? 16 : 4);
1814 int oplen = alen + 4 + sizeof(struct inet_diag_hostcond);
1815 struct inet_diag_hostcond *cond = (struct inet_diag_hostcond *)(ptr+4);
1816
1817 *op = (struct inet_diag_bc_op){ code, oplen, oplen+4 };
1818 cond->family = a->addr.family;
1819 cond->port = a->port;
1820 cond->prefix_len = a->addr.bitlen;
1821 memcpy(cond->addr, a->addr.data, alen);
1822 ptr += oplen;
1823 if (b->next) {
1824 op = (struct inet_diag_bc_op *)ptr;
1825 *op = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, len - (ptr-*bytecode)};
1826 ptr += 4;
1827 }
1828 }
1829 return ptr - *bytecode;
1830 }
1831 case SSF_D_GE:
1832 {
1833 struct aafilter *x = (void *)f->pred;
1834
1835 if (!(*bytecode = malloc(8))) abort();
1836 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_GE, 8, 12 };
1837 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1838 return 8;
1839 }
1840 case SSF_D_LE:
1841 {
1842 struct aafilter *x = (void *)f->pred;
1843
1844 if (!(*bytecode = malloc(8))) abort();
1845 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_D_LE, 8, 12 };
1846 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1847 return 8;
1848 }
1849 case SSF_S_GE:
1850 {
1851 struct aafilter *x = (void *)f->pred;
1852
1853 if (!(*bytecode = malloc(8))) abort();
1854 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_GE, 8, 12 };
1855 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1856 return 8;
1857 }
1858 case SSF_S_LE:
1859 {
1860 struct aafilter *x = (void *)f->pred;
1861
1862 if (!(*bytecode = malloc(8))) abort();
1863 ((struct inet_diag_bc_op *)*bytecode)[0] = (struct inet_diag_bc_op){ INET_DIAG_BC_S_LE, 8, 12 };
1864 ((struct inet_diag_bc_op *)*bytecode)[1] = (struct inet_diag_bc_op){ 0, 0, x->port };
1865 return 8;
1866 }
1867
1868 case SSF_AND:
1869 {
1870 char *a1 = NULL, *a2 = NULL, *a;
1871 int l1, l2;
1872
1873 l1 = ssfilter_bytecompile(f->pred, &a1);
1874 l2 = ssfilter_bytecompile(f->post, &a2);
1875 if (!l1 || !l2) {
1876 free(a1);
1877 free(a2);
1878 return 0;
1879 }
1880 if (!(a = malloc(l1+l2))) abort();
1881 memcpy(a, a1, l1);
1882 memcpy(a+l1, a2, l2);
1883 free(a1); free(a2);
1884 ssfilter_patch(a, l1, l2);
1885 *bytecode = a;
1886 return l1+l2;
1887 }
1888 case SSF_OR:
1889 {
1890 char *a1 = NULL, *a2 = NULL, *a;
1891 int l1, l2;
1892
1893 l1 = ssfilter_bytecompile(f->pred, &a1);
1894 l2 = ssfilter_bytecompile(f->post, &a2);
1895 if (!l1 || !l2) {
1896 free(a1);
1897 free(a2);
1898 return 0;
1899 }
1900 if (!(a = malloc(l1+l2+4))) abort();
1901 memcpy(a, a1, l1);
1902 memcpy(a+l1+4, a2, l2);
1903 free(a1); free(a2);
1904 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, l2+4 };
1905 *bytecode = a;
1906 return l1+l2+4;
1907 }
1908 case SSF_NOT:
1909 {
1910 char *a1 = NULL, *a;
1911 int l1;
1912
1913 l1 = ssfilter_bytecompile(f->pred, &a1);
1914 if (!l1) {
1915 free(a1);
1916 return 0;
1917 }
1918 if (!(a = malloc(l1+4))) abort();
1919 memcpy(a, a1, l1);
1920 free(a1);
1921 *(struct inet_diag_bc_op *)(a+l1) = (struct inet_diag_bc_op){ INET_DIAG_BC_JMP, 4, 8 };
1922 *bytecode = a;
1923 return l1+4;
1924 }
1925 case SSF_DEVCOND:
1926 {
1927 /* bytecompile for SSF_DEVCOND not supported yet */
1928 return 0;
1929 }
1930 case SSF_MARKMASK:
1931 {
1932 struct aafilter *a = (void *)f->pred;
1933 struct instr {
1934 struct inet_diag_bc_op op;
1935 struct inet_diag_markcond cond;
1936 };
1937 int inslen = sizeof(struct instr);
1938
1939 if (!(*bytecode = malloc(inslen))) abort();
1940 ((struct instr *)*bytecode)[0] = (struct instr) {
1941 { INET_DIAG_BC_MARK_COND, inslen, inslen + 4 },
1942 { a->mark, a->mask},
1943 };
1944
1945 return inslen;
1946 }
1947 default:
1948 abort();
1949 }
1950 }
1951
1952 static int remember_he(struct aafilter *a, struct hostent *he)
1953 {
1954 char **ptr = he->h_addr_list;
1955 int cnt = 0;
1956 int len;
1957
1958 if (he->h_addrtype == AF_INET)
1959 len = 4;
1960 else if (he->h_addrtype == AF_INET6)
1961 len = 16;
1962 else
1963 return 0;
1964
1965 while (*ptr) {
1966 struct aafilter *b = a;
1967
1968 if (a->addr.bitlen) {
1969 if ((b = malloc(sizeof(*b))) == NULL)
1970 return cnt;
1971 *b = *a;
1972 a->next = b;
1973 }
1974 memcpy(b->addr.data, *ptr, len);
1975 b->addr.bytelen = len;
1976 b->addr.bitlen = len*8;
1977 b->addr.family = he->h_addrtype;
1978 ptr++;
1979 cnt++;
1980 }
1981 return cnt;
1982 }
1983
1984 static int get_dns_host(struct aafilter *a, const char *addr, int fam)
1985 {
1986 static int notfirst;
1987 int cnt = 0;
1988 struct hostent *he;
1989
1990 a->addr.bitlen = 0;
1991 if (!notfirst) {
1992 sethostent(1);
1993 notfirst = 1;
1994 }
1995 he = gethostbyname2(addr, fam == AF_UNSPEC ? AF_INET : fam);
1996 if (he)
1997 cnt = remember_he(a, he);
1998 if (fam == AF_UNSPEC) {
1999 he = gethostbyname2(addr, AF_INET6);
2000 if (he)
2001 cnt += remember_he(a, he);
2002 }
2003 return !cnt;
2004 }
2005
2006 static int xll_initted;
2007
2008 static void xll_init(void)
2009 {
2010 struct rtnl_handle rth;
2011
2012 if (rtnl_open(&rth, 0) < 0)
2013 exit(1);
2014
2015 ll_init_map(&rth);
2016 rtnl_close(&rth);
2017 xll_initted = 1;
2018 }
2019
2020 static const char *xll_index_to_name(int index)
2021 {
2022 if (!xll_initted)
2023 xll_init();
2024 return ll_index_to_name(index);
2025 }
2026
2027 static int xll_name_to_index(const char *dev)
2028 {
2029 if (!xll_initted)
2030 xll_init();
2031 return ll_name_to_index(dev);
2032 }
2033
2034 void *parse_devcond(char *name)
2035 {
2036 struct aafilter a = { .iface = 0 };
2037 struct aafilter *res;
2038
2039 a.iface = xll_name_to_index(name);
2040 if (a.iface == 0) {
2041 char *end;
2042 unsigned long n;
2043
2044 n = strtoul(name, &end, 0);
2045 if (!end || end == name || *end || n > UINT_MAX)
2046 return NULL;
2047
2048 a.iface = n;
2049 }
2050
2051 res = malloc(sizeof(*res));
2052 *res = a;
2053
2054 return res;
2055 }
2056
2057 static void vsock_set_inet_prefix(inet_prefix *a, __u32 cid)
2058 {
2059 *a = (inet_prefix){
2060 .bytelen = sizeof(cid),
2061 .family = AF_VSOCK,
2062 };
2063 memcpy(a->data, &cid, sizeof(cid));
2064 }
2065
2066 void *parse_hostcond(char *addr, bool is_port)
2067 {
2068 char *port = NULL;
2069 struct aafilter a = { .port = -1 };
2070 struct aafilter *res;
2071 int fam = preferred_family;
2072 struct filter *f = &current_filter;
2073
2074 if (fam == AF_UNIX || strncmp(addr, "unix:", 5) == 0) {
2075 char *p;
2076
2077 a.addr.family = AF_UNIX;
2078 if (strncmp(addr, "unix:", 5) == 0)
2079 addr += 5;
2080 p = strdup(addr);
2081 a.addr.bitlen = 8*strlen(p);
2082 memcpy(a.addr.data, &p, sizeof(p));
2083 fam = AF_UNIX;
2084 goto out;
2085 }
2086
2087 if (fam == AF_PACKET || strncmp(addr, "link:", 5) == 0) {
2088 a.addr.family = AF_PACKET;
2089 a.addr.bitlen = 0;
2090 if (strncmp(addr, "link:", 5) == 0)
2091 addr += 5;
2092 port = strchr(addr, ':');
2093 if (port) {
2094 *port = 0;
2095 if (port[1] && strcmp(port+1, "*")) {
2096 if (get_integer(&a.port, port+1, 0)) {
2097 if ((a.port = xll_name_to_index(port+1)) <= 0)
2098 return NULL;
2099 }
2100 }
2101 }
2102 if (addr[0] && strcmp(addr, "*")) {
2103 unsigned short tmp;
2104
2105 a.addr.bitlen = 32;
2106 if (ll_proto_a2n(&tmp, addr))
2107 return NULL;
2108 a.addr.data[0] = ntohs(tmp);
2109 }
2110 fam = AF_PACKET;
2111 goto out;
2112 }
2113
2114 if (fam == AF_NETLINK || strncmp(addr, "netlink:", 8) == 0) {
2115 a.addr.family = AF_NETLINK;
2116 a.addr.bitlen = 0;
2117 if (strncmp(addr, "netlink:", 8) == 0)
2118 addr += 8;
2119 port = strchr(addr, ':');
2120 if (port) {
2121 *port = 0;
2122 if (port[1] && strcmp(port+1, "*")) {
2123 if (get_integer(&a.port, port+1, 0)) {
2124 if (strcmp(port+1, "kernel") == 0)
2125 a.port = 0;
2126 else
2127 return NULL;
2128 }
2129 }
2130 }
2131 if (addr[0] && strcmp(addr, "*")) {
2132 a.addr.bitlen = 32;
2133 if (nl_proto_a2n(&a.addr.data[0], addr) == -1)
2134 return NULL;
2135 }
2136 fam = AF_NETLINK;
2137 goto out;
2138 }
2139
2140 if (fam == AF_VSOCK || strncmp(addr, "vsock:", 6) == 0) {
2141 __u32 cid = ~(__u32)0;
2142
2143 a.addr.family = AF_VSOCK;
2144 if (strncmp(addr, "vsock:", 6) == 0)
2145 addr += 6;
2146
2147 if (is_port)
2148 port = addr;
2149 else {
2150 port = strchr(addr, ':');
2151 if (port) {
2152 *port = '\0';
2153 port++;
2154 }
2155 }
2156
2157 if (port && strcmp(port, "*") &&
2158 get_u32((__u32 *)&a.port, port, 0))
2159 return NULL;
2160
2161 if (addr[0] && strcmp(addr, "*")) {
2162 a.addr.bitlen = 32;
2163 if (get_u32(&cid, addr, 0))
2164 return NULL;
2165 }
2166 vsock_set_inet_prefix(&a.addr, cid);
2167 fam = AF_VSOCK;
2168 goto out;
2169 }
2170
2171 if (fam == AF_INET || !strncmp(addr, "inet:", 5)) {
2172 fam = AF_INET;
2173 if (!strncmp(addr, "inet:", 5))
2174 addr += 5;
2175 } else if (fam == AF_INET6 || !strncmp(addr, "inet6:", 6)) {
2176 fam = AF_INET6;
2177 if (!strncmp(addr, "inet6:", 6))
2178 addr += 6;
2179 }
2180
2181 /* URL-like literal [] */
2182 if (addr[0] == '[') {
2183 addr++;
2184 if ((port = strchr(addr, ']')) == NULL)
2185 return NULL;
2186 *port++ = 0;
2187 } else if (addr[0] == '*') {
2188 port = addr+1;
2189 } else {
2190 port = strrchr(strchr(addr, '/') ? : addr, ':');
2191 }
2192
2193 if (is_port)
2194 port = addr;
2195
2196 if (port && *port) {
2197 if (*port == ':')
2198 *port++ = 0;
2199
2200 if (*port && *port != '*') {
2201 if (get_integer(&a.port, port, 0)) {
2202 struct servent *se1 = NULL;
2203 struct servent *se2 = NULL;
2204
2205 if (current_filter.dbs&(1<<UDP_DB))
2206 se1 = getservbyname(port, UDP_PROTO);
2207 if (current_filter.dbs&(1<<TCP_DB))
2208 se2 = getservbyname(port, TCP_PROTO);
2209 if (se1 && se2 && se1->s_port != se2->s_port) {
2210 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2211 return NULL;
2212 }
2213 if (!se1)
2214 se1 = se2;
2215 if (se1) {
2216 a.port = ntohs(se1->s_port);
2217 } else {
2218 struct scache *s;
2219
2220 for (s = rlist; s; s = s->next) {
2221 if ((s->proto == UDP_PROTO &&
2222 (current_filter.dbs&(1<<UDP_DB))) ||
2223 (s->proto == TCP_PROTO &&
2224 (current_filter.dbs&(1<<TCP_DB)))) {
2225 if (s->name && strcmp(s->name, port) == 0) {
2226 if (a.port > 0 && a.port != s->port) {
2227 fprintf(stderr, "Error: ambiguous port \"%s\".\n", port);
2228 return NULL;
2229 }
2230 a.port = s->port;
2231 }
2232 }
2233 }
2234 if (a.port <= 0) {
2235 fprintf(stderr, "Error: \"%s\" does not look like a port.\n", port);
2236 return NULL;
2237 }
2238 }
2239 }
2240 }
2241 }
2242 if (!is_port && *addr && *addr != '*') {
2243 if (get_prefix_1(&a.addr, addr, fam)) {
2244 if (get_dns_host(&a, addr, fam)) {
2245 fprintf(stderr, "Error: an inet prefix is expected rather than \"%s\".\n", addr);
2246 return NULL;
2247 }
2248 }
2249 }
2250
2251 out:
2252 if (fam != AF_UNSPEC) {
2253 int states = f->states;
2254 f->families = 0;
2255 filter_af_set(f, fam);
2256 filter_states_set(f, states);
2257 }
2258
2259 res = malloc(sizeof(*res));
2260 if (res)
2261 memcpy(res, &a, sizeof(a));
2262 return res;
2263 }
2264
2265 void *parse_markmask(const char *markmask)
2266 {
2267 struct aafilter a, *res;
2268
2269 if (strchr(markmask, '/')) {
2270 if (sscanf(markmask, "%i/%i", &a.mark, &a.mask) != 2)
2271 return NULL;
2272 } else {
2273 a.mask = 0xffffffff;
2274 if (sscanf(markmask, "%i", &a.mark) != 1)
2275 return NULL;
2276 }
2277
2278 res = malloc(sizeof(*res));
2279 if (res)
2280 memcpy(res, &a, sizeof(a));
2281 return res;
2282 }
2283
2284 static void proc_ctx_print(struct sockstat *s)
2285 {
2286 char *buf;
2287
2288 if (show_proc_ctx || show_sock_ctx) {
2289 if (find_entry(s->ino, &buf,
2290 (show_proc_ctx & show_sock_ctx) ?
2291 PROC_SOCK_CTX : PROC_CTX) > 0) {
2292 out(" users:(%s)", buf);
2293 free(buf);
2294 }
2295 } else if (show_users) {
2296 if (find_entry(s->ino, &buf, USERS) > 0) {
2297 out(" users:(%s)", buf);
2298 free(buf);
2299 }
2300 }
2301 }
2302
2303 static void inet_stats_print(struct sockstat *s, bool v6only)
2304 {
2305 sock_state_print(s);
2306
2307 inet_addr_print(&s->local, s->lport, s->iface, v6only);
2308 inet_addr_print(&s->remote, s->rport, 0, v6only);
2309
2310 proc_ctx_print(s);
2311 }
2312
2313 static int proc_parse_inet_addr(char *loc, char *rem, int family, struct
2314 sockstat * s)
2315 {
2316 s->local.family = s->remote.family = family;
2317 if (family == AF_INET) {
2318 sscanf(loc, "%x:%x", s->local.data, (unsigned *)&s->lport);
2319 sscanf(rem, "%x:%x", s->remote.data, (unsigned *)&s->rport);
2320 s->local.bytelen = s->remote.bytelen = 4;
2321 return 0;
2322 } else {
2323 sscanf(loc, "%08x%08x%08x%08x:%x",
2324 s->local.data,
2325 s->local.data + 1,
2326 s->local.data + 2,
2327 s->local.data + 3,
2328 &s->lport);
2329 sscanf(rem, "%08x%08x%08x%08x:%x",
2330 s->remote.data,
2331 s->remote.data + 1,
2332 s->remote.data + 2,
2333 s->remote.data + 3,
2334 &s->rport);
2335 s->local.bytelen = s->remote.bytelen = 16;
2336 return 0;
2337 }
2338 return -1;
2339 }
2340
2341 static int proc_inet_split_line(char *line, char **loc, char **rem, char **data)
2342 {
2343 char *p;
2344
2345 if ((p = strchr(line, ':')) == NULL)
2346 return -1;
2347
2348 *loc = p+2;
2349 if ((p = strchr(*loc, ':')) == NULL)
2350 return -1;
2351
2352 p[5] = 0;
2353 *rem = p+6;
2354 if ((p = strchr(*rem, ':')) == NULL)
2355 return -1;
2356
2357 p[5] = 0;
2358 *data = p+6;
2359 return 0;
2360 }
2361
2362 static char *sprint_bw(char *buf, double bw)
2363 {
2364 if (numeric)
2365 sprintf(buf, "%.0f", bw);
2366 else if (bw > 1000000.)
2367 sprintf(buf, "%.1fM", bw / 1000000.);
2368 else if (bw > 1000.)
2369 sprintf(buf, "%.1fK", bw / 1000.);
2370 else
2371 sprintf(buf, "%g", bw);
2372
2373 return buf;
2374 }
2375
2376 static void sctp_stats_print(struct sctp_info *s)
2377 {
2378 if (s->sctpi_tag)
2379 out(" tag:%x", s->sctpi_tag);
2380 if (s->sctpi_state)
2381 out(" state:%s", sctp_sstate_name[s->sctpi_state]);
2382 if (s->sctpi_rwnd)
2383 out(" rwnd:%d", s->sctpi_rwnd);
2384 if (s->sctpi_unackdata)
2385 out(" unackdata:%d", s->sctpi_unackdata);
2386 if (s->sctpi_penddata)
2387 out(" penddata:%d", s->sctpi_penddata);
2388 if (s->sctpi_instrms)
2389 out(" instrms:%d", s->sctpi_instrms);
2390 if (s->sctpi_outstrms)
2391 out(" outstrms:%d", s->sctpi_outstrms);
2392 if (s->sctpi_inqueue)
2393 out(" inqueue:%d", s->sctpi_inqueue);
2394 if (s->sctpi_outqueue)
2395 out(" outqueue:%d", s->sctpi_outqueue);
2396 if (s->sctpi_overall_error)
2397 out(" overerr:%d", s->sctpi_overall_error);
2398 if (s->sctpi_max_burst)
2399 out(" maxburst:%d", s->sctpi_max_burst);
2400 if (s->sctpi_maxseg)
2401 out(" maxseg:%d", s->sctpi_maxseg);
2402 if (s->sctpi_peer_rwnd)
2403 out(" prwnd:%d", s->sctpi_peer_rwnd);
2404 if (s->sctpi_peer_tag)
2405 out(" ptag:%x", s->sctpi_peer_tag);
2406 if (s->sctpi_peer_capable)
2407 out(" pcapable:%d", s->sctpi_peer_capable);
2408 if (s->sctpi_peer_sack)
2409 out(" psack:%d", s->sctpi_peer_sack);
2410 if (s->sctpi_s_autoclose)
2411 out(" autoclose:%d", s->sctpi_s_autoclose);
2412 if (s->sctpi_s_adaptation_ind)
2413 out(" adapind:%d", s->sctpi_s_adaptation_ind);
2414 if (s->sctpi_s_pd_point)
2415 out(" pdpoint:%d", s->sctpi_s_pd_point);
2416 if (s->sctpi_s_nodelay)
2417 out(" nodelay:%d", s->sctpi_s_nodelay);
2418 if (s->sctpi_s_disable_fragments)
2419 out(" nofrag:%d", s->sctpi_s_disable_fragments);
2420 if (s->sctpi_s_v4mapped)
2421 out(" v4mapped:%d", s->sctpi_s_v4mapped);
2422 if (s->sctpi_s_frag_interleave)
2423 out(" fraginl:%d", s->sctpi_s_frag_interleave);
2424 }
2425
2426 static void tcp_stats_print(struct tcpstat *s)
2427 {
2428 char b1[64];
2429
2430 if (s->has_ts_opt)
2431 out(" ts");
2432 if (s->has_sack_opt)
2433 out(" sack");
2434 if (s->has_ecn_opt)
2435 out(" ecn");
2436 if (s->has_ecnseen_opt)
2437 out(" ecnseen");
2438 if (s->has_fastopen_opt)
2439 out(" fastopen");
2440 if (s->cong_alg[0])
2441 out(" %s", s->cong_alg);
2442 if (s->has_wscale_opt)
2443 out(" wscale:%d,%d", s->snd_wscale, s->rcv_wscale);
2444 if (s->rto)
2445 out(" rto:%g", s->rto);
2446 if (s->backoff)
2447 out(" backoff:%u", s->backoff);
2448 if (s->rtt)
2449 out(" rtt:%g/%g", s->rtt, s->rttvar);
2450 if (s->ato)
2451 out(" ato:%g", s->ato);
2452
2453 if (s->qack)
2454 out(" qack:%d", s->qack);
2455 if (s->qack & 1)
2456 out(" bidir");
2457
2458 if (s->mss)
2459 out(" mss:%d", s->mss);
2460 if (s->pmtu)
2461 out(" pmtu:%u", s->pmtu);
2462 if (s->rcv_mss)
2463 out(" rcvmss:%d", s->rcv_mss);
2464 if (s->advmss)
2465 out(" advmss:%d", s->advmss);
2466 if (s->cwnd)
2467 out(" cwnd:%u", s->cwnd);
2468 if (s->ssthresh)
2469 out(" ssthresh:%d", s->ssthresh);
2470
2471 if (s->bytes_sent)
2472 out(" bytes_sent:%llu", s->bytes_sent);
2473 if (s->bytes_retrans)
2474 out(" bytes_retrans:%llu", s->bytes_retrans);
2475 if (s->bytes_acked)
2476 out(" bytes_acked:%llu", s->bytes_acked);
2477 if (s->bytes_received)
2478 out(" bytes_received:%llu", s->bytes_received);
2479 if (s->segs_out)
2480 out(" segs_out:%u", s->segs_out);
2481 if (s->segs_in)
2482 out(" segs_in:%u", s->segs_in);
2483 if (s->data_segs_out)
2484 out(" data_segs_out:%u", s->data_segs_out);
2485 if (s->data_segs_in)
2486 out(" data_segs_in:%u", s->data_segs_in);
2487
2488 if (s->dctcp && s->dctcp->enabled) {
2489 struct dctcpstat *dctcp = s->dctcp;
2490
2491 out(" dctcp:(ce_state:%u,alpha:%u,ab_ecn:%u,ab_tot:%u)",
2492 dctcp->ce_state, dctcp->alpha, dctcp->ab_ecn,
2493 dctcp->ab_tot);
2494 } else if (s->dctcp) {
2495 out(" dctcp:fallback_mode");
2496 }
2497
2498 if (s->bbr_info) {
2499 __u64 bw;
2500
2501 bw = s->bbr_info->bbr_bw_hi;
2502 bw <<= 32;
2503 bw |= s->bbr_info->bbr_bw_lo;
2504
2505 out(" bbr:(bw:%sbps,mrtt:%g",
2506 sprint_bw(b1, bw * 8.0),
2507 (double)s->bbr_info->bbr_min_rtt / 1000.0);
2508 if (s->bbr_info->bbr_pacing_gain)
2509 out(",pacing_gain:%g",
2510 (double)s->bbr_info->bbr_pacing_gain / 256.0);
2511 if (s->bbr_info->bbr_cwnd_gain)
2512 out(",cwnd_gain:%g",
2513 (double)s->bbr_info->bbr_cwnd_gain / 256.0);
2514 out(")");
2515 }
2516
2517 if (s->send_bps)
2518 out(" send %sbps", sprint_bw(b1, s->send_bps));
2519 if (s->lastsnd)
2520 out(" lastsnd:%u", s->lastsnd);
2521 if (s->lastrcv)
2522 out(" lastrcv:%u", s->lastrcv);
2523 if (s->lastack)
2524 out(" lastack:%u", s->lastack);
2525
2526 if (s->pacing_rate) {
2527 out(" pacing_rate %sbps", sprint_bw(b1, s->pacing_rate));
2528 if (s->pacing_rate_max)
2529 out("/%sbps", sprint_bw(b1, s->pacing_rate_max));
2530 }
2531
2532 if (s->delivery_rate)
2533 out(" delivery_rate %sbps", sprint_bw(b1, s->delivery_rate));
2534 if (s->delivered)
2535 out(" delivered:%u", s->delivered);
2536 if (s->delivered_ce)
2537 out(" delivered_ce:%u", s->delivered_ce);
2538 if (s->app_limited)
2539 out(" app_limited");
2540
2541 if (s->busy_time) {
2542 out(" busy:%llums", s->busy_time / 1000);
2543 if (s->rwnd_limited)
2544 out(" rwnd_limited:%llums(%.1f%%)",
2545 s->rwnd_limited / 1000,
2546 100.0 * s->rwnd_limited / s->busy_time);
2547 if (s->sndbuf_limited)
2548 out(" sndbuf_limited:%llums(%.1f%%)",
2549 s->sndbuf_limited / 1000,
2550 100.0 * s->sndbuf_limited / s->busy_time);
2551 }
2552
2553 if (s->unacked)
2554 out(" unacked:%u", s->unacked);
2555 if (s->retrans || s->retrans_total)
2556 out(" retrans:%u/%u", s->retrans, s->retrans_total);
2557 if (s->lost)
2558 out(" lost:%u", s->lost);
2559 if (s->sacked && s->ss.state != SS_LISTEN)
2560 out(" sacked:%u", s->sacked);
2561 if (s->dsack_dups)
2562 out(" dsack_dups:%u", s->dsack_dups);
2563 if (s->fackets)
2564 out(" fackets:%u", s->fackets);
2565 if (s->reordering != 3)
2566 out(" reordering:%d", s->reordering);
2567 if (s->reord_seen)
2568 out(" reord_seen:%d", s->reord_seen);
2569 if (s->rcv_rtt)
2570 out(" rcv_rtt:%g", s->rcv_rtt);
2571 if (s->rcv_space)
2572 out(" rcv_space:%d", s->rcv_space);
2573 if (s->rcv_ssthresh)
2574 out(" rcv_ssthresh:%u", s->rcv_ssthresh);
2575 if (s->not_sent)
2576 out(" notsent:%u", s->not_sent);
2577 if (s->min_rtt)
2578 out(" minrtt:%g", s->min_rtt);
2579 }
2580
2581 static void tcp_timer_print(struct tcpstat *s)
2582 {
2583 static const char * const tmr_name[] = {
2584 "off",
2585 "on",
2586 "keepalive",
2587 "timewait",
2588 "persist",
2589 "unknown"
2590 };
2591
2592 if (s->timer) {
2593 if (s->timer > 4)
2594 s->timer = 5;
2595 out(" timer:(%s,%s,%d)",
2596 tmr_name[s->timer],
2597 print_ms_timer(s->timeout),
2598 s->retrans);
2599 }
2600 }
2601
2602 static void sctp_timer_print(struct tcpstat *s)
2603 {
2604 if (s->timer)
2605 out(" timer:(T3_RTX,%s,%d)",
2606 print_ms_timer(s->timeout), s->retrans);
2607 }
2608
2609 static int tcp_show_line(char *line, const struct filter *f, int family)
2610 {
2611 int rto = 0, ato = 0;
2612 struct tcpstat s = {};
2613 char *loc, *rem, *data;
2614 char opt[256];
2615 int n;
2616 int hz = get_user_hz();
2617
2618 if (proc_inet_split_line(line, &loc, &rem, &data))
2619 return -1;
2620
2621 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
2622
2623 if (!(f->states & (1 << state)))
2624 return 0;
2625
2626 proc_parse_inet_addr(loc, rem, family, &s.ss);
2627
2628 if (f->f && run_ssfilter(f->f, &s.ss) == 0)
2629 return 0;
2630
2631 opt[0] = 0;
2632 n = sscanf(data, "%x %x:%x %x:%x %x %d %d %u %d %llx %d %d %d %u %d %[^\n]\n",
2633 &s.ss.state, &s.ss.wq, &s.ss.rq,
2634 &s.timer, &s.timeout, &s.retrans, &s.ss.uid, &s.probes,
2635 &s.ss.ino, &s.ss.refcnt, &s.ss.sk, &rto, &ato, &s.qack, &s.cwnd,
2636 &s.ssthresh, opt);
2637
2638 if (n < 17)
2639 opt[0] = 0;
2640
2641 if (n < 12) {
2642 rto = 0;
2643 s.cwnd = 2;
2644 s.ssthresh = -1;
2645 ato = s.qack = 0;
2646 }
2647
2648 s.retrans = s.timer != 1 ? s.probes : s.retrans;
2649 s.timeout = (s.timeout * 1000 + hz - 1) / hz;
2650 s.ato = (double)ato / hz;
2651 s.qack /= 2;
2652 s.rto = (double)rto;
2653 s.ssthresh = s.ssthresh == -1 ? 0 : s.ssthresh;
2654 s.rto = s.rto != 3 * hz ? s.rto / hz : 0;
2655 s.ss.type = IPPROTO_TCP;
2656
2657 inet_stats_print(&s.ss, false);
2658
2659 if (show_options)
2660 tcp_timer_print(&s);
2661
2662 if (show_details) {
2663 sock_details_print(&s.ss);
2664 if (opt[0])
2665 out(" opt:\"%s\"", opt);
2666 }
2667
2668 if (show_tcpinfo)
2669 tcp_stats_print(&s);
2670
2671 return 0;
2672 }
2673
2674 static int generic_record_read(FILE *fp,
2675 int (*worker)(char*, const struct filter *, int),
2676 const struct filter *f, int fam)
2677 {
2678 char line[256];
2679
2680 /* skip header */
2681 if (fgets(line, sizeof(line), fp) == NULL)
2682 goto outerr;
2683
2684 while (fgets(line, sizeof(line), fp) != NULL) {
2685 int n = strlen(line);
2686
2687 if (n == 0 || line[n-1] != '\n') {
2688 errno = -EINVAL;
2689 return -1;
2690 }
2691 line[n-1] = 0;
2692
2693 if (worker(line, f, fam) < 0)
2694 return 0;
2695 }
2696 outerr:
2697
2698 return ferror(fp) ? -1 : 0;
2699 }
2700
2701 static void print_skmeminfo(struct rtattr *tb[], int attrtype)
2702 {
2703 const __u32 *skmeminfo;
2704
2705 if (!tb[attrtype]) {
2706 if (attrtype == INET_DIAG_SKMEMINFO) {
2707 if (!tb[INET_DIAG_MEMINFO])
2708 return;
2709
2710 const struct inet_diag_meminfo *minfo =
2711 RTA_DATA(tb[INET_DIAG_MEMINFO]);
2712
2713 out(" mem:(r%u,w%u,f%u,t%u)",
2714 minfo->idiag_rmem,
2715 minfo->idiag_wmem,
2716 minfo->idiag_fmem,
2717 minfo->idiag_tmem);
2718 }
2719 return;
2720 }
2721
2722 skmeminfo = RTA_DATA(tb[attrtype]);
2723
2724 out(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
2725 skmeminfo[SK_MEMINFO_RMEM_ALLOC],
2726 skmeminfo[SK_MEMINFO_RCVBUF],
2727 skmeminfo[SK_MEMINFO_WMEM_ALLOC],
2728 skmeminfo[SK_MEMINFO_SNDBUF],
2729 skmeminfo[SK_MEMINFO_FWD_ALLOC],
2730 skmeminfo[SK_MEMINFO_WMEM_QUEUED],
2731 skmeminfo[SK_MEMINFO_OPTMEM]);
2732
2733 if (RTA_PAYLOAD(tb[attrtype]) >=
2734 (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
2735 out(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
2736
2737 if (RTA_PAYLOAD(tb[attrtype]) >=
2738 (SK_MEMINFO_DROPS + 1) * sizeof(__u32))
2739 out(",d%u", skmeminfo[SK_MEMINFO_DROPS]);
2740
2741 out(")");
2742 }
2743
2744 static void print_md5sig(struct tcp_diag_md5sig *sig)
2745 {
2746 out("%s/%d=",
2747 format_host(sig->tcpm_family,
2748 sig->tcpm_family == AF_INET6 ? 16 : 4,
2749 &sig->tcpm_addr),
2750 sig->tcpm_prefixlen);
2751 print_escape_buf(sig->tcpm_key, sig->tcpm_keylen, " ,");
2752 }
2753
2754 #define TCPI_HAS_OPT(info, opt) !!(info->tcpi_options & (opt))
2755
2756 static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2757 struct rtattr *tb[])
2758 {
2759 double rtt = 0;
2760 struct tcpstat s = {};
2761
2762 s.ss.state = r->idiag_state;
2763
2764 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2765
2766 if (tb[INET_DIAG_INFO]) {
2767 struct tcp_info *info;
2768 int len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2769
2770 /* workaround for older kernels with less fields */
2771 if (len < sizeof(*info)) {
2772 info = alloca(sizeof(*info));
2773 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2774 memset((char *)info + len, 0, sizeof(*info) - len);
2775 } else
2776 info = RTA_DATA(tb[INET_DIAG_INFO]);
2777
2778 if (show_options) {
2779 s.has_ts_opt = TCPI_HAS_OPT(info, TCPI_OPT_TIMESTAMPS);
2780 s.has_sack_opt = TCPI_HAS_OPT(info, TCPI_OPT_SACK);
2781 s.has_ecn_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN);
2782 s.has_ecnseen_opt = TCPI_HAS_OPT(info, TCPI_OPT_ECN_SEEN);
2783 s.has_fastopen_opt = TCPI_HAS_OPT(info, TCPI_OPT_SYN_DATA);
2784 }
2785
2786 if (tb[INET_DIAG_CONG])
2787 strncpy(s.cong_alg,
2788 rta_getattr_str(tb[INET_DIAG_CONG]),
2789 sizeof(s.cong_alg) - 1);
2790
2791 if (TCPI_HAS_OPT(info, TCPI_OPT_WSCALE)) {
2792 s.has_wscale_opt = true;
2793 s.snd_wscale = info->tcpi_snd_wscale;
2794 s.rcv_wscale = info->tcpi_rcv_wscale;
2795 }
2796
2797 if (info->tcpi_rto && info->tcpi_rto != 3000000)
2798 s.rto = (double)info->tcpi_rto / 1000;
2799
2800 s.backoff = info->tcpi_backoff;
2801 s.rtt = (double)info->tcpi_rtt / 1000;
2802 s.rttvar = (double)info->tcpi_rttvar / 1000;
2803 s.ato = (double)info->tcpi_ato / 1000;
2804 s.mss = info->tcpi_snd_mss;
2805 s.rcv_mss = info->tcpi_rcv_mss;
2806 s.advmss = info->tcpi_advmss;
2807 s.rcv_space = info->tcpi_rcv_space;
2808 s.rcv_rtt = (double)info->tcpi_rcv_rtt / 1000;
2809 s.lastsnd = info->tcpi_last_data_sent;
2810 s.lastrcv = info->tcpi_last_data_recv;
2811 s.lastack = info->tcpi_last_ack_recv;
2812 s.unacked = info->tcpi_unacked;
2813 s.retrans = info->tcpi_retrans;
2814 s.retrans_total = info->tcpi_total_retrans;
2815 s.lost = info->tcpi_lost;
2816 s.sacked = info->tcpi_sacked;
2817 s.fackets = info->tcpi_fackets;
2818 s.reordering = info->tcpi_reordering;
2819 s.rcv_ssthresh = info->tcpi_rcv_ssthresh;
2820 s.cwnd = info->tcpi_snd_cwnd;
2821 s.pmtu = info->tcpi_pmtu;
2822
2823 if (info->tcpi_snd_ssthresh < 0xFFFF)
2824 s.ssthresh = info->tcpi_snd_ssthresh;
2825
2826 rtt = (double) info->tcpi_rtt;
2827 if (tb[INET_DIAG_VEGASINFO]) {
2828 const struct tcpvegas_info *vinfo
2829 = RTA_DATA(tb[INET_DIAG_VEGASINFO]);
2830
2831 if (vinfo->tcpv_enabled &&
2832 vinfo->tcpv_rtt && vinfo->tcpv_rtt != 0x7fffffff)
2833 rtt = vinfo->tcpv_rtt;
2834 }
2835
2836 if (tb[INET_DIAG_DCTCPINFO]) {
2837 struct dctcpstat *dctcp = malloc(sizeof(struct
2838 dctcpstat));
2839
2840 const struct tcp_dctcp_info *dinfo
2841 = RTA_DATA(tb[INET_DIAG_DCTCPINFO]);
2842
2843 dctcp->enabled = !!dinfo->dctcp_enabled;
2844 dctcp->ce_state = dinfo->dctcp_ce_state;
2845 dctcp->alpha = dinfo->dctcp_alpha;
2846 dctcp->ab_ecn = dinfo->dctcp_ab_ecn;
2847 dctcp->ab_tot = dinfo->dctcp_ab_tot;
2848 s.dctcp = dctcp;
2849 }
2850
2851 if (tb[INET_DIAG_BBRINFO]) {
2852 const void *bbr_info = RTA_DATA(tb[INET_DIAG_BBRINFO]);
2853 int len = min(RTA_PAYLOAD(tb[INET_DIAG_BBRINFO]),
2854 sizeof(*s.bbr_info));
2855
2856 s.bbr_info = calloc(1, sizeof(*s.bbr_info));
2857 if (s.bbr_info && bbr_info)
2858 memcpy(s.bbr_info, bbr_info, len);
2859 }
2860
2861 if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) {
2862 s.send_bps = (double) info->tcpi_snd_cwnd *
2863 (double)info->tcpi_snd_mss * 8000000. / rtt;
2864 }
2865
2866 if (info->tcpi_pacing_rate &&
2867 info->tcpi_pacing_rate != ~0ULL) {
2868 s.pacing_rate = info->tcpi_pacing_rate * 8.;
2869
2870 if (info->tcpi_max_pacing_rate &&
2871 info->tcpi_max_pacing_rate != ~0ULL)
2872 s.pacing_rate_max = info->tcpi_max_pacing_rate * 8.;
2873 }
2874 s.bytes_acked = info->tcpi_bytes_acked;
2875 s.bytes_received = info->tcpi_bytes_received;
2876 s.segs_out = info->tcpi_segs_out;
2877 s.segs_in = info->tcpi_segs_in;
2878 s.data_segs_out = info->tcpi_data_segs_out;
2879 s.data_segs_in = info->tcpi_data_segs_in;
2880 s.not_sent = info->tcpi_notsent_bytes;
2881 if (info->tcpi_min_rtt && info->tcpi_min_rtt != ~0U)
2882 s.min_rtt = (double) info->tcpi_min_rtt / 1000;
2883 s.delivery_rate = info->tcpi_delivery_rate * 8.;
2884 s.app_limited = info->tcpi_delivery_rate_app_limited;
2885 s.busy_time = info->tcpi_busy_time;
2886 s.rwnd_limited = info->tcpi_rwnd_limited;
2887 s.sndbuf_limited = info->tcpi_sndbuf_limited;
2888 s.delivered = info->tcpi_delivered;
2889 s.delivered_ce = info->tcpi_delivered_ce;
2890 s.dsack_dups = info->tcpi_dsack_dups;
2891 s.reord_seen = info->tcpi_reord_seen;
2892 s.bytes_sent = info->tcpi_bytes_sent;
2893 s.bytes_retrans = info->tcpi_bytes_retrans;
2894 tcp_stats_print(&s);
2895 free(s.dctcp);
2896 free(s.bbr_info);
2897 }
2898 if (tb[INET_DIAG_MD5SIG]) {
2899 struct tcp_diag_md5sig *sig = RTA_DATA(tb[INET_DIAG_MD5SIG]);
2900 int len = RTA_PAYLOAD(tb[INET_DIAG_MD5SIG]);
2901
2902 out(" md5keys:");
2903 print_md5sig(sig++);
2904 for (len -= sizeof(*sig); len > 0; len -= sizeof(*sig)) {
2905 out(",");
2906 print_md5sig(sig++);
2907 }
2908 }
2909 }
2910
2911 static const char *format_host_sa(struct sockaddr_storage *sa)
2912 {
2913 union {
2914 struct sockaddr_in sin;
2915 struct sockaddr_in6 sin6;
2916 } *saddr = (void *)sa;
2917
2918 switch (sa->ss_family) {
2919 case AF_INET:
2920 return format_host(AF_INET, 4, &saddr->sin.sin_addr);
2921 case AF_INET6:
2922 return format_host(AF_INET6, 16, &saddr->sin6.sin6_addr);
2923 default:
2924 return "";
2925 }
2926 }
2927
2928 static void sctp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r,
2929 struct rtattr *tb[])
2930 {
2931 struct sockaddr_storage *sa;
2932 int len;
2933
2934 print_skmeminfo(tb, INET_DIAG_SKMEMINFO);
2935
2936 if (tb[INET_DIAG_LOCALS]) {
2937 len = RTA_PAYLOAD(tb[INET_DIAG_LOCALS]);
2938 sa = RTA_DATA(tb[INET_DIAG_LOCALS]);
2939
2940 out(" locals:%s", format_host_sa(sa));
2941 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
2942 out(",%s", format_host_sa(sa));
2943
2944 }
2945 if (tb[INET_DIAG_PEERS]) {
2946 len = RTA_PAYLOAD(tb[INET_DIAG_PEERS]);
2947 sa = RTA_DATA(tb[INET_DIAG_PEERS]);
2948
2949 out(" peers:%s", format_host_sa(sa));
2950 for (sa++, len -= sizeof(*sa); len > 0; sa++, len -= sizeof(*sa))
2951 out(",%s", format_host_sa(sa));
2952 }
2953 if (tb[INET_DIAG_INFO]) {
2954 struct sctp_info *info;
2955 len = RTA_PAYLOAD(tb[INET_DIAG_INFO]);
2956
2957 /* workaround for older kernels with less fields */
2958 if (len < sizeof(*info)) {
2959 info = alloca(sizeof(*info));
2960 memcpy(info, RTA_DATA(tb[INET_DIAG_INFO]), len);
2961 memset((char *)info + len, 0, sizeof(*info) - len);
2962 } else
2963 info = RTA_DATA(tb[INET_DIAG_INFO]);
2964
2965 sctp_stats_print(info);
2966 }
2967 }
2968
2969 static void parse_diag_msg(struct nlmsghdr *nlh, struct sockstat *s)
2970 {
2971 struct rtattr *tb[INET_DIAG_MAX+1];
2972 struct inet_diag_msg *r = NLMSG_DATA(nlh);
2973
2974 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
2975 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
2976
2977 s->state = r->idiag_state;
2978 s->local.family = s->remote.family = r->idiag_family;
2979 s->lport = ntohs(r->id.idiag_sport);
2980 s->rport = ntohs(r->id.idiag_dport);
2981 s->wq = r->idiag_wqueue;
2982 s->rq = r->idiag_rqueue;
2983 s->ino = r->idiag_inode;
2984 s->uid = r->idiag_uid;
2985 s->iface = r->id.idiag_if;
2986 s->sk = cookie_sk_get(&r->id.idiag_cookie[0]);
2987
2988 s->mark = 0;
2989 if (tb[INET_DIAG_MARK])
2990 s->mark = rta_getattr_u32(tb[INET_DIAG_MARK]);
2991 if (tb[INET_DIAG_PROTOCOL])
2992 s->raw_prot = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
2993 else
2994 s->raw_prot = 0;
2995
2996 if (s->local.family == AF_INET)
2997 s->local.bytelen = s->remote.bytelen = 4;
2998 else
2999 s->local.bytelen = s->remote.bytelen = 16;
3000
3001 memcpy(s->local.data, r->id.idiag_src, s->local.bytelen);
3002 memcpy(s->remote.data, r->id.idiag_dst, s->local.bytelen);
3003 }
3004
3005 static int inet_show_sock(struct nlmsghdr *nlh,
3006 struct sockstat *s)
3007 {
3008 struct rtattr *tb[INET_DIAG_MAX+1];
3009 struct inet_diag_msg *r = NLMSG_DATA(nlh);
3010 unsigned char v6only = 0;
3011
3012 parse_rtattr(tb, INET_DIAG_MAX, (struct rtattr *)(r+1),
3013 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3014
3015 if (tb[INET_DIAG_PROTOCOL])
3016 s->type = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]);
3017
3018 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3019 v6only = rta_getattr_u8(tb[INET_DIAG_SKV6ONLY]);
3020
3021 inet_stats_print(s, v6only);
3022
3023 if (show_options) {
3024 struct tcpstat t = {};
3025
3026 t.timer = r->idiag_timer;
3027 t.timeout = r->idiag_expires;
3028 t.retrans = r->idiag_retrans;
3029 if (s->type == IPPROTO_SCTP)
3030 sctp_timer_print(&t);
3031 else
3032 tcp_timer_print(&t);
3033 }
3034
3035 if (show_details) {
3036 sock_details_print(s);
3037 if (s->local.family == AF_INET6 && tb[INET_DIAG_SKV6ONLY])
3038 out(" v6only:%u", v6only);
3039
3040 if (tb[INET_DIAG_SHUTDOWN]) {
3041 unsigned char mask;
3042
3043 mask = rta_getattr_u8(tb[INET_DIAG_SHUTDOWN]);
3044 out(" %c-%c",
3045 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3046 }
3047 }
3048
3049 if (show_tos) {
3050 if (tb[INET_DIAG_TOS])
3051 out(" tos:%#x", rta_getattr_u8(tb[INET_DIAG_TOS]));
3052 if (tb[INET_DIAG_TCLASS])
3053 out(" tclass:%#x", rta_getattr_u8(tb[INET_DIAG_TCLASS]));
3054 if (tb[INET_DIAG_CLASS_ID])
3055 out(" class_id:%#x", rta_getattr_u32(tb[INET_DIAG_CLASS_ID]));
3056 }
3057
3058 if (show_mem || (show_tcpinfo && s->type != IPPROTO_UDP)) {
3059 if (!oneline)
3060 out("\n\t");
3061 if (s->type == IPPROTO_SCTP)
3062 sctp_show_info(nlh, r, tb);
3063 else
3064 tcp_show_info(nlh, r, tb);
3065 }
3066 sctp_ino = s->ino;
3067
3068 return 0;
3069 }
3070
3071 static int tcpdiag_send(int fd, int protocol, struct filter *f)
3072 {
3073 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3074 struct {
3075 struct nlmsghdr nlh;
3076 struct inet_diag_req r;
3077 } req = {
3078 .nlh.nlmsg_len = sizeof(req),
3079 .nlh.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST,
3080 .nlh.nlmsg_seq = MAGIC_SEQ,
3081 .r.idiag_family = AF_INET,
3082 .r.idiag_states = f->states,
3083 };
3084 char *bc = NULL;
3085 int bclen;
3086 struct msghdr msg;
3087 struct rtattr rta;
3088 struct iovec iov[3];
3089 int iovlen = 1;
3090
3091 if (protocol == IPPROTO_UDP)
3092 return -1;
3093
3094 if (protocol == IPPROTO_TCP)
3095 req.nlh.nlmsg_type = TCPDIAG_GETSOCK;
3096 else
3097 req.nlh.nlmsg_type = DCCPDIAG_GETSOCK;
3098 if (show_mem) {
3099 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3100 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3101 }
3102
3103 if (show_tcpinfo) {
3104 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3105 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3106 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3107 }
3108
3109 if (show_tos) {
3110 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3111 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3112 }
3113
3114 iov[0] = (struct iovec){
3115 .iov_base = &req,
3116 .iov_len = sizeof(req)
3117 };
3118 if (f->f) {
3119 bclen = ssfilter_bytecompile(f->f, &bc);
3120 if (bclen) {
3121 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3122 rta.rta_len = RTA_LENGTH(bclen);
3123 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3124 iov[2] = (struct iovec){ bc, bclen };
3125 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3126 iovlen = 3;
3127 }
3128 }
3129
3130 msg = (struct msghdr) {
3131 .msg_name = (void *)&nladdr,
3132 .msg_namelen = sizeof(nladdr),
3133 .msg_iov = iov,
3134 .msg_iovlen = iovlen,
3135 };
3136
3137 if (sendmsg(fd, &msg, 0) < 0) {
3138 close(fd);
3139 return -1;
3140 }
3141
3142 return 0;
3143 }
3144
3145 static int sockdiag_send(int family, int fd, int protocol, struct filter *f)
3146 {
3147 struct sockaddr_nl nladdr = { .nl_family = AF_NETLINK };
3148 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3149 char *bc = NULL;
3150 int bclen;
3151 struct msghdr msg;
3152 struct rtattr rta;
3153 struct iovec iov[3];
3154 int iovlen = 1;
3155
3156 if (family == PF_UNSPEC)
3157 return tcpdiag_send(fd, protocol, f);
3158
3159 memset(&req.r, 0, sizeof(req.r));
3160 req.r.sdiag_family = family;
3161 req.r.sdiag_protocol = protocol;
3162 req.r.idiag_states = f->states;
3163 if (show_mem) {
3164 req.r.idiag_ext |= (1<<(INET_DIAG_MEMINFO-1));
3165 req.r.idiag_ext |= (1<<(INET_DIAG_SKMEMINFO-1));
3166 }
3167
3168 if (show_tcpinfo) {
3169 req.r.idiag_ext |= (1<<(INET_DIAG_INFO-1));
3170 req.r.idiag_ext |= (1<<(INET_DIAG_VEGASINFO-1));
3171 req.r.idiag_ext |= (1<<(INET_DIAG_CONG-1));
3172 }
3173
3174 if (show_tos) {
3175 req.r.idiag_ext |= (1<<(INET_DIAG_TOS-1));
3176 req.r.idiag_ext |= (1<<(INET_DIAG_TCLASS-1));
3177 }
3178
3179 iov[0] = (struct iovec){
3180 .iov_base = &req,
3181 .iov_len = sizeof(req)
3182 };
3183 if (f->f) {
3184 bclen = ssfilter_bytecompile(f->f, &bc);
3185 if (bclen) {
3186 rta.rta_type = INET_DIAG_REQ_BYTECODE;
3187 rta.rta_len = RTA_LENGTH(bclen);
3188 iov[1] = (struct iovec){ &rta, sizeof(rta) };
3189 iov[2] = (struct iovec){ bc, bclen };
3190 req.nlh.nlmsg_len += RTA_LENGTH(bclen);
3191 iovlen = 3;
3192 }
3193 }
3194
3195 msg = (struct msghdr) {
3196 .msg_name = (void *)&nladdr,
3197 .msg_namelen = sizeof(nladdr),
3198 .msg_iov = iov,
3199 .msg_iovlen = iovlen,
3200 };
3201
3202 if (sendmsg(fd, &msg, 0) < 0) {
3203 close(fd);
3204 return -1;
3205 }
3206
3207 return 0;
3208 }
3209
3210 struct inet_diag_arg {
3211 struct filter *f;
3212 int protocol;
3213 struct rtnl_handle *rth;
3214 };
3215
3216 static int kill_inet_sock(struct nlmsghdr *h, void *arg, struct sockstat *s)
3217 {
3218 struct inet_diag_msg *d = NLMSG_DATA(h);
3219 struct inet_diag_arg *diag_arg = arg;
3220 struct rtnl_handle *rth = diag_arg->rth;
3221
3222 DIAG_REQUEST(req, struct inet_diag_req_v2 r);
3223
3224 req.nlh.nlmsg_type = SOCK_DESTROY;
3225 req.nlh.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
3226 req.nlh.nlmsg_seq = ++rth->seq;
3227 req.r.sdiag_family = d->idiag_family;
3228 req.r.sdiag_protocol = diag_arg->protocol;
3229 req.r.id = d->id;
3230
3231 if (diag_arg->protocol == IPPROTO_RAW) {
3232 struct inet_diag_req_raw *raw = (void *)&req.r;
3233
3234 BUILD_BUG_ON(sizeof(req.r) != sizeof(*raw));
3235 raw->sdiag_raw_protocol = s->raw_prot;
3236 }
3237
3238 return rtnl_talk(rth, &req.nlh, NULL);
3239 }
3240
3241 static int show_one_inet_sock(struct nlmsghdr *h, void *arg)
3242 {
3243 int err;
3244 struct inet_diag_arg *diag_arg = arg;
3245 struct inet_diag_msg *r = NLMSG_DATA(h);
3246 struct sockstat s = {};
3247
3248 if (!(diag_arg->f->families & FAMILY_MASK(r->idiag_family)))
3249 return 0;
3250
3251 parse_diag_msg(h, &s);
3252 s.type = diag_arg->protocol;
3253
3254 if (diag_arg->f->f && run_ssfilter(diag_arg->f->f, &s) == 0)
3255 return 0;
3256
3257 if (diag_arg->f->kill && kill_inet_sock(h, arg, &s) != 0) {
3258 if (errno == EOPNOTSUPP || errno == ENOENT) {
3259 /* Socket can't be closed, or is already closed. */
3260 return 0;
3261 } else {
3262 perror("SOCK_DESTROY answers");
3263 return -1;
3264 }
3265 }
3266
3267 err = inet_show_sock(h, &s);
3268 if (err < 0)
3269 return err;
3270
3271 return 0;
3272 }
3273
3274 static int inet_show_netlink(struct filter *f, FILE *dump_fp, int protocol)
3275 {
3276 int err = 0;
3277 struct rtnl_handle rth, rth2;
3278 int family = PF_INET;
3279 struct inet_diag_arg arg = { .f = f, .protocol = protocol };
3280
3281 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3282 return -1;
3283
3284 if (f->kill) {
3285 if (rtnl_open_byproto(&rth2, 0, NETLINK_SOCK_DIAG)) {
3286 rtnl_close(&rth);
3287 return -1;
3288 }
3289 arg.rth = &rth2;
3290 }
3291
3292 rth.dump = MAGIC_SEQ;
3293 rth.dump_fp = dump_fp;
3294 if (preferred_family == PF_INET6)
3295 family = PF_INET6;
3296
3297 again:
3298 if ((err = sockdiag_send(family, rth.fd, protocol, f)))
3299 goto Exit;
3300
3301 if ((err = rtnl_dump_filter(&rth, show_one_inet_sock, &arg))) {
3302 if (family != PF_UNSPEC) {
3303 family = PF_UNSPEC;
3304 goto again;
3305 }
3306 goto Exit;
3307 }
3308 if (family == PF_INET && preferred_family != PF_INET) {
3309 family = PF_INET6;
3310 goto again;
3311 }
3312
3313 Exit:
3314 rtnl_close(&rth);
3315 if (arg.rth)
3316 rtnl_close(arg.rth);
3317 return err;
3318 }
3319
3320 static int tcp_show_netlink_file(struct filter *f)
3321 {
3322 FILE *fp;
3323 char buf[16384];
3324 int err = -1;
3325
3326 if ((fp = fopen(getenv("TCPDIAG_FILE"), "r")) == NULL) {
3327 perror("fopen($TCPDIAG_FILE)");
3328 return err;
3329 }
3330
3331 while (1) {
3332 int status, err2;
3333 struct nlmsghdr *h = (struct nlmsghdr *)buf;
3334 struct sockstat s = {};
3335
3336 status = fread(buf, 1, sizeof(*h), fp);
3337 if (status < 0) {
3338 perror("Reading header from $TCPDIAG_FILE");
3339 break;
3340 }
3341 if (status != sizeof(*h)) {
3342 perror("Unexpected EOF reading $TCPDIAG_FILE");
3343 break;
3344 }
3345
3346 status = fread(h+1, 1, NLMSG_ALIGN(h->nlmsg_len-sizeof(*h)), fp);
3347
3348 if (status < 0) {
3349 perror("Reading $TCPDIAG_FILE");
3350 break;
3351 }
3352 if (status + sizeof(*h) < h->nlmsg_len) {
3353 perror("Unexpected EOF reading $TCPDIAG_FILE");
3354 break;
3355 }
3356
3357 /* The only legal exit point */
3358 if (h->nlmsg_type == NLMSG_DONE) {
3359 err = 0;
3360 break;
3361 }
3362
3363 if (h->nlmsg_type == NLMSG_ERROR) {
3364 struct nlmsgerr *err = (struct nlmsgerr *)NLMSG_DATA(h);
3365
3366 if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
3367 fprintf(stderr, "ERROR truncated\n");
3368 } else {
3369 errno = -err->error;
3370 perror("TCPDIAG answered");
3371 }
3372 break;
3373 }
3374
3375 parse_diag_msg(h, &s);
3376 s.type = IPPROTO_TCP;
3377
3378 if (f && f->f && run_ssfilter(f->f, &s) == 0)
3379 continue;
3380
3381 err2 = inet_show_sock(h, &s);
3382 if (err2 < 0) {
3383 err = err2;
3384 break;
3385 }
3386 }
3387
3388 fclose(fp);
3389 return err;
3390 }
3391
3392 static int tcp_show(struct filter *f)
3393 {
3394 FILE *fp = NULL;
3395 char *buf = NULL;
3396 int bufsize = 1024*1024;
3397
3398 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3399 return 0;
3400
3401 dg_proto = TCP_PROTO;
3402
3403 if (getenv("TCPDIAG_FILE"))
3404 return tcp_show_netlink_file(f);
3405
3406 if (!getenv("PROC_NET_TCP") && !getenv("PROC_ROOT")
3407 && inet_show_netlink(f, NULL, IPPROTO_TCP) == 0)
3408 return 0;
3409
3410 /* Sigh... We have to parse /proc/net/tcp... */
3411 while (bufsize >= 64*1024) {
3412 if ((buf = malloc(bufsize)) != NULL)
3413 break;
3414 bufsize /= 2;
3415 }
3416 if (buf == NULL) {
3417 errno = ENOMEM;
3418 return -1;
3419 }
3420
3421 if (f->families & FAMILY_MASK(AF_INET)) {
3422 if ((fp = net_tcp_open()) == NULL)
3423 goto outerr;
3424
3425 setbuffer(fp, buf, bufsize);
3426 if (generic_record_read(fp, tcp_show_line, f, AF_INET))
3427 goto outerr;
3428 fclose(fp);
3429 }
3430
3431 if ((f->families & FAMILY_MASK(AF_INET6)) &&
3432 (fp = net_tcp6_open()) != NULL) {
3433 setbuffer(fp, buf, bufsize);
3434 if (generic_record_read(fp, tcp_show_line, f, AF_INET6))
3435 goto outerr;
3436 fclose(fp);
3437 }
3438
3439 free(buf);
3440 return 0;
3441
3442 outerr:
3443 do {
3444 int saved_errno = errno;
3445
3446 free(buf);
3447 if (fp)
3448 fclose(fp);
3449 errno = saved_errno;
3450 return -1;
3451 } while (0);
3452 }
3453
3454 static int dccp_show(struct filter *f)
3455 {
3456 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3457 return 0;
3458
3459 if (!getenv("PROC_NET_DCCP") && !getenv("PROC_ROOT")
3460 && inet_show_netlink(f, NULL, IPPROTO_DCCP) == 0)
3461 return 0;
3462
3463 return 0;
3464 }
3465
3466 static int sctp_show(struct filter *f)
3467 {
3468 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3469 return 0;
3470
3471 if (!getenv("PROC_NET_SCTP") && !getenv("PROC_ROOT")
3472 && inet_show_netlink(f, NULL, IPPROTO_SCTP) == 0)
3473 return 0;
3474
3475 return 0;
3476 }
3477
3478 static int dgram_show_line(char *line, const struct filter *f, int family)
3479 {
3480 struct sockstat s = {};
3481 char *loc, *rem, *data;
3482 char opt[256];
3483 int n;
3484
3485 if (proc_inet_split_line(line, &loc, &rem, &data))
3486 return -1;
3487
3488 int state = (data[1] >= 'A') ? (data[1] - 'A' + 10) : (data[1] - '0');
3489
3490 if (!(f->states & (1 << state)))
3491 return 0;
3492
3493 proc_parse_inet_addr(loc, rem, family, &s);
3494
3495 if (f->f && run_ssfilter(f->f, &s) == 0)
3496 return 0;
3497
3498 opt[0] = 0;
3499 n = sscanf(data, "%x %x:%x %*x:%*x %*x %d %*d %u %d %llx %[^\n]\n",
3500 &s.state, &s.wq, &s.rq,
3501 &s.uid, &s.ino,
3502 &s.refcnt, &s.sk, opt);
3503
3504 if (n < 9)
3505 opt[0] = 0;
3506
3507 s.type = dg_proto == UDP_PROTO ? IPPROTO_UDP : 0;
3508 inet_stats_print(&s, false);
3509
3510 if (show_details && opt[0])
3511 out(" opt:\"%s\"", opt);
3512
3513 return 0;
3514 }
3515
3516 static int udp_show(struct filter *f)
3517 {
3518 FILE *fp = NULL;
3519
3520 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3521 return 0;
3522
3523 dg_proto = UDP_PROTO;
3524
3525 if (!getenv("PROC_NET_UDP") && !getenv("PROC_ROOT")
3526 && inet_show_netlink(f, NULL, IPPROTO_UDP) == 0)
3527 return 0;
3528
3529 if (f->families&FAMILY_MASK(AF_INET)) {
3530 if ((fp = net_udp_open()) == NULL)
3531 goto outerr;
3532 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3533 goto outerr;
3534 fclose(fp);
3535 }
3536
3537 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3538 (fp = net_udp6_open()) != NULL) {
3539 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3540 goto outerr;
3541 fclose(fp);
3542 }
3543 return 0;
3544
3545 outerr:
3546 do {
3547 int saved_errno = errno;
3548
3549 if (fp)
3550 fclose(fp);
3551 errno = saved_errno;
3552 return -1;
3553 } while (0);
3554 }
3555
3556 static int raw_show(struct filter *f)
3557 {
3558 FILE *fp = NULL;
3559
3560 if (!filter_af_get(f, AF_INET) && !filter_af_get(f, AF_INET6))
3561 return 0;
3562
3563 dg_proto = RAW_PROTO;
3564
3565 if (!getenv("PROC_NET_RAW") && !getenv("PROC_ROOT") &&
3566 inet_show_netlink(f, NULL, IPPROTO_RAW) == 0)
3567 return 0;
3568
3569 if (f->families&FAMILY_MASK(AF_INET)) {
3570 if ((fp = net_raw_open()) == NULL)
3571 goto outerr;
3572 if (generic_record_read(fp, dgram_show_line, f, AF_INET))
3573 goto outerr;
3574 fclose(fp);
3575 }
3576
3577 if ((f->families&FAMILY_MASK(AF_INET6)) &&
3578 (fp = net_raw6_open()) != NULL) {
3579 if (generic_record_read(fp, dgram_show_line, f, AF_INET6))
3580 goto outerr;
3581 fclose(fp);
3582 }
3583 return 0;
3584
3585 outerr:
3586 do {
3587 int saved_errno = errno;
3588
3589 if (fp)
3590 fclose(fp);
3591 errno = saved_errno;
3592 return -1;
3593 } while (0);
3594 }
3595
3596 #define MAX_UNIX_REMEMBER (1024*1024/sizeof(struct sockstat))
3597
3598 static void unix_list_drop_first(struct sockstat **list)
3599 {
3600 struct sockstat *s = *list;
3601
3602 (*list) = (*list)->next;
3603 free(s->name);
3604 free(s);
3605 }
3606
3607 static bool unix_type_skip(struct sockstat *s, struct filter *f)
3608 {
3609 if (s->type == SOCK_STREAM && !(f->dbs&(1<<UNIX_ST_DB)))
3610 return true;
3611 if (s->type == SOCK_DGRAM && !(f->dbs&(1<<UNIX_DG_DB)))
3612 return true;
3613 if (s->type == SOCK_SEQPACKET && !(f->dbs&(1<<UNIX_SQ_DB)))
3614 return true;
3615 return false;
3616 }
3617
3618 static void unix_stats_print(struct sockstat *s, struct filter *f)
3619 {
3620 char port_name[30] = {};
3621
3622 sock_state_print(s);
3623
3624 sock_addr_print(s->name ?: "*", " ",
3625 int_to_str(s->lport, port_name), NULL);
3626 sock_addr_print(s->peer_name ?: "*", " ",
3627 int_to_str(s->rport, port_name), NULL);
3628
3629 proc_ctx_print(s);
3630 }
3631
3632 static int unix_show_sock(struct nlmsghdr *nlh, void *arg)
3633 {
3634 struct filter *f = (struct filter *)arg;
3635 struct unix_diag_msg *r = NLMSG_DATA(nlh);
3636 struct rtattr *tb[UNIX_DIAG_MAX+1];
3637 char name[128];
3638 struct sockstat stat = { .name = "*", .peer_name = "*" };
3639
3640 parse_rtattr(tb, UNIX_DIAG_MAX, (struct rtattr *)(r+1),
3641 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3642
3643 stat.type = r->udiag_type;
3644 stat.state = r->udiag_state;
3645 stat.ino = stat.lport = r->udiag_ino;
3646 stat.local.family = stat.remote.family = AF_UNIX;
3647
3648 if (unix_type_skip(&stat, f))
3649 return 0;
3650
3651 if (tb[UNIX_DIAG_RQLEN]) {
3652 struct unix_diag_rqlen *rql = RTA_DATA(tb[UNIX_DIAG_RQLEN]);
3653
3654 stat.rq = rql->udiag_rqueue;
3655 stat.wq = rql->udiag_wqueue;
3656 }
3657 if (tb[UNIX_DIAG_NAME]) {
3658 int len = RTA_PAYLOAD(tb[UNIX_DIAG_NAME]);
3659
3660 memcpy(name, RTA_DATA(tb[UNIX_DIAG_NAME]), len);
3661 name[len] = '\0';
3662 if (name[0] == '\0') {
3663 int i;
3664 for (i = 0; i < len; i++)
3665 if (name[i] == '\0')
3666 name[i] = '@';
3667 }
3668 stat.name = &name[0];
3669 memcpy(stat.local.data, &stat.name, sizeof(stat.name));
3670 }
3671 if (tb[UNIX_DIAG_PEER])
3672 stat.rport = rta_getattr_u32(tb[UNIX_DIAG_PEER]);
3673
3674 if (f->f && run_ssfilter(f->f, &stat) == 0)
3675 return 0;
3676
3677 unix_stats_print(&stat, f);
3678
3679 if (show_mem)
3680 print_skmeminfo(tb, UNIX_DIAG_MEMINFO);
3681 if (show_details) {
3682 if (tb[UNIX_DIAG_SHUTDOWN]) {
3683 unsigned char mask;
3684
3685 mask = rta_getattr_u8(tb[UNIX_DIAG_SHUTDOWN]);
3686 out(" %c-%c",
3687 mask & 1 ? '-' : '<', mask & 2 ? '-' : '>');
3688 }
3689 if (tb[UNIX_DIAG_VFS]) {
3690 struct unix_diag_vfs *uv = RTA_DATA(tb[UNIX_DIAG_VFS]);
3691
3692 out(" ino:%u dev:%u/%u", uv->udiag_vfs_ino, major(uv->udiag_vfs_dev),
3693 minor(uv->udiag_vfs_dev));
3694 }
3695 if (tb[UNIX_DIAG_ICONS]) {
3696 int len = RTA_PAYLOAD(tb[UNIX_DIAG_ICONS]);
3697 __u32 *peers = RTA_DATA(tb[UNIX_DIAG_ICONS]);
3698 int i;
3699
3700 out(" peers:");
3701 for (i = 0; i < len / sizeof(__u32); i++)
3702 out(" %u", peers[i]);
3703 }
3704 }
3705
3706 return 0;
3707 }
3708
3709 static int handle_netlink_request(struct filter *f, struct nlmsghdr *req,
3710 size_t size, rtnl_filter_t show_one_sock)
3711 {
3712 int ret = -1;
3713 struct rtnl_handle rth;
3714
3715 if (rtnl_open_byproto(&rth, 0, NETLINK_SOCK_DIAG))
3716 return -1;
3717
3718 rth.dump = MAGIC_SEQ;
3719
3720 if (rtnl_send(&rth, req, size) < 0)
3721 goto Exit;
3722
3723 if (rtnl_dump_filter(&rth, show_one_sock, f))
3724 goto Exit;
3725
3726 ret = 0;
3727 Exit:
3728 rtnl_close(&rth);
3729 return ret;
3730 }
3731
3732 static int unix_show_netlink(struct filter *f)
3733 {
3734 DIAG_REQUEST(req, struct unix_diag_req r);
3735
3736 req.r.sdiag_family = AF_UNIX;
3737 req.r.udiag_states = f->states;
3738 req.r.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER | UDIAG_SHOW_RQLEN;
3739 if (show_mem)
3740 req.r.udiag_show |= UDIAG_SHOW_MEMINFO;
3741 if (show_details)
3742 req.r.udiag_show |= UDIAG_SHOW_VFS | UDIAG_SHOW_ICONS;
3743
3744 return handle_netlink_request(f, &req.nlh, sizeof(req), unix_show_sock);
3745 }
3746
3747 static int unix_show(struct filter *f)
3748 {
3749 FILE *fp;
3750 char buf[256];
3751 char name[128];
3752 int newformat = 0;
3753 int cnt;
3754 struct sockstat *list = NULL;
3755 const int unix_state_map[] = { SS_CLOSE, SS_SYN_SENT,
3756 SS_ESTABLISHED, SS_CLOSING };
3757
3758 if (!filter_af_get(f, AF_UNIX))
3759 return 0;
3760
3761 if (!getenv("PROC_NET_UNIX") && !getenv("PROC_ROOT")
3762 && unix_show_netlink(f) == 0)
3763 return 0;
3764
3765 if ((fp = net_unix_open()) == NULL)
3766 return -1;
3767 if (!fgets(buf, sizeof(buf), fp)) {
3768 fclose(fp);
3769 return -1;
3770 }
3771
3772 if (memcmp(buf, "Peer", 4) == 0)
3773 newformat = 1;
3774 cnt = 0;
3775
3776 while (fgets(buf, sizeof(buf), fp)) {
3777 struct sockstat *u, **insp;
3778 int flags;
3779
3780 if (!(u = calloc(1, sizeof(*u))))
3781 break;
3782
3783 if (sscanf(buf, "%x: %x %x %x %x %x %d %s",
3784 &u->rport, &u->rq, &u->wq, &flags, &u->type,
3785 &u->state, &u->ino, name) < 8)
3786 name[0] = 0;
3787
3788 u->lport = u->ino;
3789 u->local.family = u->remote.family = AF_UNIX;
3790
3791 if (flags & (1 << 16)) {
3792 u->state = SS_LISTEN;
3793 } else if (u->state > 0 &&
3794 u->state <= ARRAY_SIZE(unix_state_map)) {
3795 u->state = unix_state_map[u->state-1];
3796 if (u->type == SOCK_DGRAM && u->state == SS_CLOSE && u->rport)
3797 u->state = SS_ESTABLISHED;
3798 }
3799 if (unix_type_skip(u, f) ||
3800 !(f->states & (1 << u->state))) {
3801 free(u);
3802 continue;
3803 }
3804
3805 if (!newformat) {
3806 u->rport = 0;
3807 u->rq = 0;
3808 u->wq = 0;
3809 }
3810
3811 if (name[0]) {
3812 u->name = strdup(name);
3813 if (!u->name) {
3814 free(u);
3815 break;
3816 }
3817 }
3818
3819 if (u->rport) {
3820 struct sockstat *p;
3821
3822 for (p = list; p; p = p->next) {
3823 if (u->rport == p->lport)
3824 break;
3825 }
3826 if (!p)
3827 u->peer_name = "?";
3828 else
3829 u->peer_name = p->name ? : "*";
3830 }
3831
3832 if (f->f) {
3833 struct sockstat st = {
3834 .local.family = AF_UNIX,
3835 .remote.family = AF_UNIX,
3836 };
3837
3838 memcpy(st.local.data, &u->name, sizeof(u->name));
3839 /* when parsing the old format rport is set to 0 and
3840 * therefore peer_name remains NULL
3841 */
3842 if (u->peer_name && strcmp(u->peer_name, "*"))
3843 memcpy(st.remote.data, &u->peer_name,
3844 sizeof(u->peer_name));
3845 if (run_ssfilter(f->f, &st) == 0) {
3846 free(u->name);
3847 free(u);
3848 continue;
3849 }
3850 }
3851
3852 insp = &list;
3853 while (*insp) {
3854 if (u->type < (*insp)->type ||
3855 (u->type == (*insp)->type &&
3856 u->ino < (*insp)->ino))
3857 break;
3858 insp = &(*insp)->next;
3859 }
3860 u->next = *insp;
3861 *insp = u;
3862
3863 if (++cnt > MAX_UNIX_REMEMBER) {
3864 while (list) {
3865 unix_stats_print(list, f);
3866 unix_list_drop_first(&list);
3867 }
3868 cnt = 0;
3869 }
3870 }
3871 fclose(fp);
3872 while (list) {
3873 unix_stats_print(list, f);
3874 unix_list_drop_first(&list);
3875 }
3876
3877 return 0;
3878 }
3879
3880 static int packet_stats_print(struct sockstat *s, const struct filter *f)
3881 {
3882 const char *addr, *port;
3883 char ll_name[16];
3884
3885 s->local.family = s->remote.family = AF_PACKET;
3886
3887 if (f->f) {
3888 s->local.data[0] = s->prot;
3889 if (run_ssfilter(f->f, s) == 0)
3890 return 1;
3891 }
3892
3893 sock_state_print(s);
3894
3895 if (s->prot == 3)
3896 addr = "*";
3897 else
3898 addr = ll_proto_n2a(htons(s->prot), ll_name, sizeof(ll_name));
3899
3900 if (s->iface == 0)
3901 port = "*";
3902 else
3903 port = xll_index_to_name(s->iface);
3904
3905 sock_addr_print(addr, ":", port, NULL);
3906 sock_addr_print("", "*", "", NULL);
3907
3908 proc_ctx_print(s);
3909
3910 if (show_details)
3911 sock_details_print(s);
3912
3913 return 0;
3914 }
3915
3916 static void packet_show_ring(struct packet_diag_ring *ring)
3917 {
3918 out("blk_size:%d", ring->pdr_block_size);
3919 out(",blk_nr:%d", ring->pdr_block_nr);
3920 out(",frm_size:%d", ring->pdr_frame_size);
3921 out(",frm_nr:%d", ring->pdr_frame_nr);
3922 out(",tmo:%d", ring->pdr_retire_tmo);
3923 out(",features:0x%x", ring->pdr_features);
3924 }
3925
3926 static int packet_show_sock(struct nlmsghdr *nlh, void *arg)
3927 {
3928 const struct filter *f = arg;
3929 struct packet_diag_msg *r = NLMSG_DATA(nlh);
3930 struct packet_diag_info *pinfo = NULL;
3931 struct packet_diag_ring *ring_rx = NULL, *ring_tx = NULL;
3932 struct rtattr *tb[PACKET_DIAG_MAX+1];
3933 struct sockstat stat = {};
3934 uint32_t fanout = 0;
3935 bool has_fanout = false;
3936
3937 parse_rtattr(tb, PACKET_DIAG_MAX, (struct rtattr *)(r+1),
3938 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
3939
3940 /* use /proc/net/packet if all info are not available */
3941 if (!tb[PACKET_DIAG_MEMINFO])
3942 return -1;
3943
3944 stat.type = r->pdiag_type;
3945 stat.prot = r->pdiag_num;
3946 stat.ino = r->pdiag_ino;
3947 stat.state = SS_CLOSE;
3948 stat.sk = cookie_sk_get(&r->pdiag_cookie[0]);
3949
3950 if (tb[PACKET_DIAG_MEMINFO]) {
3951 __u32 *skmeminfo = RTA_DATA(tb[PACKET_DIAG_MEMINFO]);
3952
3953 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
3954 }
3955
3956 if (tb[PACKET_DIAG_INFO]) {
3957 pinfo = RTA_DATA(tb[PACKET_DIAG_INFO]);
3958 stat.lport = stat.iface = pinfo->pdi_index;
3959 }
3960
3961 if (tb[PACKET_DIAG_UID])
3962 stat.uid = rta_getattr_u32(tb[PACKET_DIAG_UID]);
3963
3964 if (tb[PACKET_DIAG_RX_RING])
3965 ring_rx = RTA_DATA(tb[PACKET_DIAG_RX_RING]);
3966
3967 if (tb[PACKET_DIAG_TX_RING])
3968 ring_tx = RTA_DATA(tb[PACKET_DIAG_TX_RING]);
3969
3970 if (tb[PACKET_DIAG_FANOUT]) {
3971 has_fanout = true;
3972 fanout = rta_getattr_u32(tb[PACKET_DIAG_FANOUT]);
3973 }
3974
3975 if (packet_stats_print(&stat, f))
3976 return 0;
3977
3978 if (show_details) {
3979 if (pinfo) {
3980 if (oneline)
3981 out(" ver:%d", pinfo->pdi_version);
3982 else
3983 out("\n\tver:%d", pinfo->pdi_version);
3984 out(" cpy_thresh:%d", pinfo->pdi_copy_thresh);
3985 out(" flags( ");
3986 if (pinfo->pdi_flags & PDI_RUNNING)
3987 out("running");
3988 if (pinfo->pdi_flags & PDI_AUXDATA)
3989 out(" auxdata");
3990 if (pinfo->pdi_flags & PDI_ORIGDEV)
3991 out(" origdev");
3992 if (pinfo->pdi_flags & PDI_VNETHDR)
3993 out(" vnethdr");
3994 if (pinfo->pdi_flags & PDI_LOSS)
3995 out(" loss");
3996 if (!pinfo->pdi_flags)
3997 out("0");
3998 out(" )");
3999 }
4000 if (ring_rx) {
4001 if (oneline)
4002 out(" ring_rx(");
4003 else
4004 out("\n\tring_rx(");
4005 packet_show_ring(ring_rx);
4006 out(")");
4007 }
4008 if (ring_tx) {
4009 if (oneline)
4010 out(" ring_tx(");
4011 else
4012 out("\n\tring_tx(");
4013 packet_show_ring(ring_tx);
4014 out(")");
4015 }
4016 if (has_fanout) {
4017 uint16_t type = (fanout >> 16) & 0xffff;
4018
4019 if (oneline)
4020 out(" fanout(");
4021 else
4022 out("\n\tfanout(");
4023 out("id:%d,", fanout & 0xffff);
4024 out("type:");
4025
4026 if (type == 0)
4027 out("hash");
4028 else if (type == 1)
4029 out("lb");
4030 else if (type == 2)
4031 out("cpu");
4032 else if (type == 3)
4033 out("roll");
4034 else if (type == 4)
4035 out("random");
4036 else if (type == 5)
4037 out("qm");
4038 else
4039 out("0x%x", type);
4040
4041 out(")");
4042 }
4043 }
4044
4045 if (show_bpf && tb[PACKET_DIAG_FILTER]) {
4046 struct sock_filter *fil =
4047 RTA_DATA(tb[PACKET_DIAG_FILTER]);
4048 int num = RTA_PAYLOAD(tb[PACKET_DIAG_FILTER]) /
4049 sizeof(struct sock_filter);
4050
4051 if (oneline)
4052 out(" bpf filter (%d): ", num);
4053 else
4054 out("\n\tbpf filter (%d): ", num);
4055 while (num) {
4056 out(" 0x%02x %u %u %u,",
4057 fil->code, fil->jt, fil->jf, fil->k);
4058 num--;
4059 fil++;
4060 }
4061 }
4062
4063 if (show_mem)
4064 print_skmeminfo(tb, PACKET_DIAG_MEMINFO);
4065 return 0;
4066 }
4067
4068 static int packet_show_netlink(struct filter *f)
4069 {
4070 DIAG_REQUEST(req, struct packet_diag_req r);
4071
4072 req.r.sdiag_family = AF_PACKET;
4073 req.r.pdiag_show = PACKET_SHOW_INFO | PACKET_SHOW_MEMINFO |
4074 PACKET_SHOW_FILTER | PACKET_SHOW_RING_CFG | PACKET_SHOW_FANOUT;
4075
4076 return handle_netlink_request(f, &req.nlh, sizeof(req), packet_show_sock);
4077 }
4078
4079 static int packet_show_line(char *buf, const struct filter *f, int fam)
4080 {
4081 unsigned long long sk;
4082 struct sockstat stat = {};
4083 int type, prot, iface, state, rq, uid, ino;
4084
4085 sscanf(buf, "%llx %*d %d %x %d %d %u %u %u",
4086 &sk,
4087 &type, &prot, &iface, &state,
4088 &rq, &uid, &ino);
4089
4090 if (stat.type == SOCK_RAW && !(f->dbs&(1<<PACKET_R_DB)))
4091 return 0;
4092 if (stat.type == SOCK_DGRAM && !(f->dbs&(1<<PACKET_DG_DB)))
4093 return 0;
4094
4095 stat.type = type;
4096 stat.prot = prot;
4097 stat.lport = stat.iface = iface;
4098 stat.state = state;
4099 stat.rq = rq;
4100 stat.uid = uid;
4101 stat.ino = ino;
4102 stat.state = SS_CLOSE;
4103
4104 if (packet_stats_print(&stat, f))
4105 return 0;
4106
4107 return 0;
4108 }
4109
4110 static int packet_show(struct filter *f)
4111 {
4112 FILE *fp;
4113 int rc = 0;
4114
4115 if (!filter_af_get(f, AF_PACKET) || !(f->states & (1 << SS_CLOSE)))
4116 return 0;
4117
4118 if (!getenv("PROC_NET_PACKET") && !getenv("PROC_ROOT") &&
4119 packet_show_netlink(f) == 0)
4120 return 0;
4121
4122 if ((fp = net_packet_open()) == NULL)
4123 return -1;
4124 if (generic_record_read(fp, packet_show_line, f, AF_PACKET))
4125 rc = -1;
4126
4127 fclose(fp);
4128 return rc;
4129 }
4130
4131 static int xdp_stats_print(struct sockstat *s, const struct filter *f)
4132 {
4133 const char *addr, *port;
4134 char q_str[16];
4135
4136 s->local.family = s->remote.family = AF_XDP;
4137
4138 if (f->f) {
4139 if (run_ssfilter(f->f, s) == 0)
4140 return 1;
4141 }
4142
4143 sock_state_print(s);
4144
4145 if (s->iface) {
4146 addr = xll_index_to_name(s->iface);
4147 snprintf(q_str, sizeof(q_str), "q%d", s->lport);
4148 port = q_str;
4149 sock_addr_print(addr, ":", port, NULL);
4150 } else {
4151 sock_addr_print("", "*", "", NULL);
4152 }
4153
4154 sock_addr_print("", "*", "", NULL);
4155
4156 proc_ctx_print(s);
4157
4158 if (show_details)
4159 sock_details_print(s);
4160
4161 return 0;
4162 }
4163
4164 static void xdp_show_ring(const char *name, struct xdp_diag_ring *ring)
4165 {
4166 if (oneline)
4167 out(" %s(", name);
4168 else
4169 out("\n\t%s(", name);
4170 out("entries:%u", ring->entries);
4171 out(")");
4172 }
4173
4174 static void xdp_show_umem(struct xdp_diag_umem *umem, struct xdp_diag_ring *fr,
4175 struct xdp_diag_ring *cr)
4176 {
4177 if (oneline)
4178 out(" tumem(");
4179 else
4180 out("\n\tumem(");
4181 out("id:%u", umem->id);
4182 out(",size:%llu", umem->size);
4183 out(",num_pages:%u", umem->num_pages);
4184 out(",chunk_size:%u", umem->chunk_size);
4185 out(",headroom:%u", umem->headroom);
4186 out(",ifindex:%u", umem->ifindex);
4187 out(",qid:%u", umem->queue_id);
4188 out(",zc:%u", umem->flags & XDP_DU_F_ZEROCOPY);
4189 out(",refs:%u", umem->refs);
4190 out(")");
4191
4192 if (fr)
4193 xdp_show_ring("fr", fr);
4194 if (cr)
4195 xdp_show_ring("cr", cr);
4196 }
4197
4198 static int xdp_show_sock(struct nlmsghdr *nlh, void *arg)
4199 {
4200 struct xdp_diag_ring *rx = NULL, *tx = NULL, *fr = NULL, *cr = NULL;
4201 struct xdp_diag_msg *msg = NLMSG_DATA(nlh);
4202 struct rtattr *tb[XDP_DIAG_MAX + 1];
4203 struct xdp_diag_info *info = NULL;
4204 struct xdp_diag_umem *umem = NULL;
4205 const struct filter *f = arg;
4206 struct sockstat stat = {};
4207
4208 parse_rtattr(tb, XDP_DIAG_MAX, (struct rtattr *)(msg + 1),
4209 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*msg)));
4210
4211 stat.type = msg->xdiag_type;
4212 stat.ino = msg->xdiag_ino;
4213 stat.state = SS_CLOSE;
4214 stat.sk = cookie_sk_get(&msg->xdiag_cookie[0]);
4215
4216 if (tb[XDP_DIAG_INFO]) {
4217 info = RTA_DATA(tb[XDP_DIAG_INFO]);
4218 stat.iface = info->ifindex;
4219 stat.lport = info->queue_id;
4220 }
4221
4222 if (tb[XDP_DIAG_UID])
4223 stat.uid = rta_getattr_u32(tb[XDP_DIAG_UID]);
4224 if (tb[XDP_DIAG_RX_RING])
4225 rx = RTA_DATA(tb[XDP_DIAG_RX_RING]);
4226 if (tb[XDP_DIAG_TX_RING])
4227 tx = RTA_DATA(tb[XDP_DIAG_TX_RING]);
4228 if (tb[XDP_DIAG_UMEM])
4229 umem = RTA_DATA(tb[XDP_DIAG_UMEM]);
4230 if (tb[XDP_DIAG_UMEM_FILL_RING])
4231 fr = RTA_DATA(tb[XDP_DIAG_UMEM_FILL_RING]);
4232 if (tb[XDP_DIAG_UMEM_COMPLETION_RING])
4233 cr = RTA_DATA(tb[XDP_DIAG_UMEM_COMPLETION_RING]);
4234 if (tb[XDP_DIAG_MEMINFO]) {
4235 __u32 *skmeminfo = RTA_DATA(tb[XDP_DIAG_MEMINFO]);
4236
4237 stat.rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4238 }
4239
4240 if (xdp_stats_print(&stat, f))
4241 return 0;
4242
4243 if (show_details) {
4244 if (rx)
4245 xdp_show_ring("rx", rx);
4246 if (tx)
4247 xdp_show_ring("tx", tx);
4248 if (umem)
4249 xdp_show_umem(umem, fr, cr);
4250 }
4251
4252 if (show_mem)
4253 print_skmeminfo(tb, XDP_DIAG_MEMINFO); // really?
4254
4255
4256 return 0;
4257 }
4258
4259 static int xdp_show(struct filter *f)
4260 {
4261 DIAG_REQUEST(req, struct xdp_diag_req r);
4262
4263 if (!filter_af_get(f, AF_XDP) || !(f->states & (1 << SS_CLOSE)))
4264 return 0;
4265
4266 req.r.sdiag_family = AF_XDP;
4267 req.r.xdiag_show = XDP_SHOW_INFO | XDP_SHOW_RING_CFG | XDP_SHOW_UMEM |
4268 XDP_SHOW_MEMINFO;
4269
4270 return handle_netlink_request(f, &req.nlh, sizeof(req), xdp_show_sock);
4271 }
4272
4273 static int netlink_show_one(struct filter *f,
4274 int prot, int pid, unsigned int groups,
4275 int state, int dst_pid, unsigned int dst_group,
4276 int rq, int wq,
4277 unsigned long long sk, unsigned long long cb)
4278 {
4279 struct sockstat st = {
4280 .state = SS_CLOSE,
4281 .rq = rq,
4282 .wq = wq,
4283 .local.family = AF_NETLINK,
4284 .remote.family = AF_NETLINK,
4285 };
4286
4287 SPRINT_BUF(prot_buf) = {};
4288 const char *prot_name;
4289 char procname[64] = {};
4290
4291 if (f->f) {
4292 st.rport = -1;
4293 st.lport = pid;
4294 st.local.data[0] = prot;
4295 if (run_ssfilter(f->f, &st) == 0)
4296 return 1;
4297 }
4298
4299 sock_state_print(&st);
4300
4301 prot_name = nl_proto_n2a(prot, prot_buf, sizeof(prot_buf));
4302
4303 if (pid == -1) {
4304 procname[0] = '*';
4305 } else if (!numeric) {
4306 int done = 0;
4307
4308 if (!pid) {
4309 done = 1;
4310 strncpy(procname, "kernel", 7);
4311 } else if (pid > 0) {
4312 FILE *fp;
4313
4314 snprintf(procname, sizeof(procname), "%s/%d/stat",
4315 getenv("PROC_ROOT") ? : "/proc", pid);
4316 if ((fp = fopen(procname, "r")) != NULL) {
4317 if (fscanf(fp, "%*d (%[^)])", procname) == 1) {
4318 snprintf(procname+strlen(procname),
4319 sizeof(procname)-strlen(procname),
4320 "/%d", pid);
4321 done = 1;
4322 }
4323 fclose(fp);
4324 }
4325 }
4326 if (!done)
4327 int_to_str(pid, procname);
4328 } else {
4329 int_to_str(pid, procname);
4330 }
4331
4332 sock_addr_print(prot_name, ":", procname, NULL);
4333
4334 if (state == NETLINK_CONNECTED) {
4335 char dst_group_buf[30];
4336 char dst_pid_buf[30];
4337
4338 sock_addr_print(int_to_str(dst_group, dst_group_buf), ":",
4339 int_to_str(dst_pid, dst_pid_buf), NULL);
4340 } else {
4341 sock_addr_print("", "*", "", NULL);
4342 }
4343
4344 char *pid_context = NULL;
4345
4346 if (show_proc_ctx) {
4347 /* The pid value will either be:
4348 * 0 if destination kernel - show kernel initial context.
4349 * A valid process pid - use getpidcon.
4350 * A unique value allocated by the kernel or netlink user
4351 * to the process - show context as "not available".
4352 */
4353 if (!pid)
4354 security_get_initial_context("kernel", &pid_context);
4355 else if (pid > 0)
4356 getpidcon(pid, &pid_context);
4357
4358 out(" proc_ctx=%s", pid_context ? : "unavailable");
4359 free(pid_context);
4360 }
4361
4362 if (show_details) {
4363 out(" sk=%llx cb=%llx groups=0x%08x", sk, cb, groups);
4364 }
4365
4366 return 0;
4367 }
4368
4369 static int netlink_show_sock(struct nlmsghdr *nlh, void *arg)
4370 {
4371 struct filter *f = (struct filter *)arg;
4372 struct netlink_diag_msg *r = NLMSG_DATA(nlh);
4373 struct rtattr *tb[NETLINK_DIAG_MAX+1];
4374 int rq = 0, wq = 0;
4375 unsigned long groups = 0;
4376
4377 parse_rtattr(tb, NETLINK_DIAG_MAX, (struct rtattr *)(r+1),
4378 nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
4379
4380 if (tb[NETLINK_DIAG_GROUPS] && RTA_PAYLOAD(tb[NETLINK_DIAG_GROUPS]))
4381 groups = *(unsigned long *) RTA_DATA(tb[NETLINK_DIAG_GROUPS]);
4382
4383 if (tb[NETLINK_DIAG_MEMINFO]) {
4384 const __u32 *skmeminfo;
4385
4386 skmeminfo = RTA_DATA(tb[NETLINK_DIAG_MEMINFO]);
4387
4388 rq = skmeminfo[SK_MEMINFO_RMEM_ALLOC];
4389 wq = skmeminfo[SK_MEMINFO_WMEM_ALLOC];
4390 }
4391
4392 if (netlink_show_one(f, r->ndiag_protocol, r->ndiag_portid, groups,
4393 r->ndiag_state, r->ndiag_dst_portid, r->ndiag_dst_group,
4394 rq, wq, 0, 0)) {
4395 return 0;
4396 }
4397
4398 if (show_mem) {
4399 out("\t");
4400 print_skmeminfo(tb, NETLINK_DIAG_MEMINFO);
4401 }
4402
4403 return 0;
4404 }
4405
4406 static int netlink_show_netlink(struct filter *f)
4407 {
4408 DIAG_REQUEST(req, struct netlink_diag_req r);
4409
4410 req.r.sdiag_family = AF_NETLINK;
4411 req.r.sdiag_protocol = NDIAG_PROTO_ALL;
4412 req.r.ndiag_show = NDIAG_SHOW_GROUPS | NDIAG_SHOW_MEMINFO;
4413
4414 return handle_netlink_request(f, &req.nlh, sizeof(req), netlink_show_sock);
4415 }
4416
4417 static int netlink_show(struct filter *f)
4418 {
4419 FILE *fp;
4420 char buf[256];
4421 int prot, pid;
4422 unsigned int groups;
4423 int rq, wq, rc;
4424 unsigned long long sk, cb;
4425
4426 if (!filter_af_get(f, AF_NETLINK) || !(f->states & (1 << SS_CLOSE)))
4427 return 0;
4428
4429 if (!getenv("PROC_NET_NETLINK") && !getenv("PROC_ROOT") &&
4430 netlink_show_netlink(f) == 0)
4431 return 0;
4432
4433 if ((fp = net_netlink_open()) == NULL)
4434 return -1;
4435 if (!fgets(buf, sizeof(buf), fp)) {
4436 fclose(fp);
4437 return -1;
4438 }
4439
4440 while (fgets(buf, sizeof(buf), fp)) {
4441 sscanf(buf, "%llx %d %d %x %d %d %llx %d",
4442 &sk,
4443 &prot, &pid, &groups, &rq, &wq, &cb, &rc);
4444
4445 netlink_show_one(f, prot, pid, groups, 0, 0, 0, rq, wq, sk, cb);
4446 }
4447
4448 fclose(fp);
4449 return 0;
4450 }
4451
4452 static bool vsock_type_skip(struct sockstat *s, struct filter *f)
4453 {
4454 if (s->type == SOCK_STREAM && !(f->dbs & (1 << VSOCK_ST_DB)))
4455 return true;
4456 if (s->type == SOCK_DGRAM && !(f->dbs & (1 << VSOCK_DG_DB)))
4457 return true;
4458 return false;
4459 }
4460
4461 static void vsock_addr_print(inet_prefix *a, __u32 port)
4462 {
4463 char cid_str[sizeof("4294967295")];
4464 char port_str[sizeof("4294967295")];
4465 __u32 cid;
4466
4467 memcpy(&cid, a->data, sizeof(cid));
4468
4469 if (cid == ~(__u32)0)
4470 snprintf(cid_str, sizeof(cid_str), "*");
4471 else
4472 snprintf(cid_str, sizeof(cid_str), "%u", cid);
4473
4474 if (port == ~(__u32)0)
4475 snprintf(port_str, sizeof(port_str), "*");
4476 else
4477 snprintf(port_str, sizeof(port_str), "%u", port);
4478
4479 sock_addr_print(cid_str, ":", port_str, NULL);
4480 }
4481
4482 static void vsock_stats_print(struct sockstat *s, struct filter *f)
4483 {
4484 sock_state_print(s);
4485
4486 vsock_addr_print(&s->local, s->lport);
4487 vsock_addr_print(&s->remote, s->rport);
4488
4489 proc_ctx_print(s);
4490 }
4491
4492 static int vsock_show_sock(struct nlmsghdr *nlh, void *arg)
4493 {
4494 struct filter *f = (struct filter *)arg;
4495 struct vsock_diag_msg *r = NLMSG_DATA(nlh);
4496 struct sockstat stat = {
4497 .type = r->vdiag_type,
4498 .lport = r->vdiag_src_port,
4499 .rport = r->vdiag_dst_port,
4500 .state = r->vdiag_state,
4501 .ino = r->vdiag_ino,
4502 };
4503
4504 vsock_set_inet_prefix(&stat.local, r->vdiag_src_cid);
4505 vsock_set_inet_prefix(&stat.remote, r->vdiag_dst_cid);
4506
4507 if (vsock_type_skip(&stat, f))
4508 return 0;
4509
4510 if (f->f && run_ssfilter(f->f, &stat) == 0)
4511 return 0;
4512
4513 vsock_stats_print(&stat, f);
4514
4515 return 0;
4516 }
4517
4518 static int vsock_show(struct filter *f)
4519 {
4520 DIAG_REQUEST(req, struct vsock_diag_req r);
4521
4522 if (!filter_af_get(f, AF_VSOCK))
4523 return 0;
4524
4525 req.r.sdiag_family = AF_VSOCK;
4526 req.r.vdiag_states = f->states;
4527
4528 return handle_netlink_request(f, &req.nlh, sizeof(req), vsock_show_sock);
4529 }
4530
4531 static void tipc_sock_addr_print(struct rtattr *net_addr, struct rtattr *id)
4532 {
4533 uint32_t node = rta_getattr_u32(net_addr);
4534 uint32_t identity = rta_getattr_u32(id);
4535
4536 SPRINT_BUF(addr) = {};
4537 SPRINT_BUF(port) = {};
4538
4539 sprintf(addr, "%u", node);
4540 sprintf(port, "%u", identity);
4541 sock_addr_print(addr, ":", port, NULL);
4542
4543 }
4544
4545 static int tipc_show_sock(struct nlmsghdr *nlh, void *arg)
4546 {
4547 struct rtattr *stat[TIPC_NLA_SOCK_STAT_MAX + 1] = {};
4548 struct rtattr *attrs[TIPC_NLA_SOCK_MAX + 1] = {};
4549 struct rtattr *con[TIPC_NLA_CON_MAX + 1] = {};
4550 struct rtattr *info[TIPC_NLA_MAX + 1] = {};
4551 struct rtattr *msg_ref;
4552 struct sockstat ss = {};
4553
4554 parse_rtattr(info, TIPC_NLA_MAX, NLMSG_DATA(nlh),
4555 NLMSG_PAYLOAD(nlh, 0));
4556
4557 if (!info[TIPC_NLA_SOCK])
4558 return 0;
4559
4560 msg_ref = info[TIPC_NLA_SOCK];
4561 parse_rtattr(attrs, TIPC_NLA_SOCK_MAX, RTA_DATA(msg_ref),
4562 RTA_PAYLOAD(msg_ref));
4563
4564 msg_ref = attrs[TIPC_NLA_SOCK_STAT];
4565 parse_rtattr(stat, TIPC_NLA_SOCK_STAT_MAX,
4566 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4567
4568
4569 ss.local.family = AF_TIPC;
4570 ss.type = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TYPE]);
4571 ss.state = rta_getattr_u32(attrs[TIPC_NLA_SOCK_TIPC_STATE]);
4572 ss.uid = rta_getattr_u32(attrs[TIPC_NLA_SOCK_UID]);
4573 ss.ino = rta_getattr_u32(attrs[TIPC_NLA_SOCK_INO]);
4574 ss.rq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_RCVQ]);
4575 ss.wq = rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_SENDQ]);
4576 ss.sk = rta_getattr_u64(attrs[TIPC_NLA_SOCK_COOKIE]);
4577
4578 sock_state_print (&ss);
4579
4580 tipc_sock_addr_print(attrs[TIPC_NLA_SOCK_ADDR],
4581 attrs[TIPC_NLA_SOCK_REF]);
4582
4583 msg_ref = attrs[TIPC_NLA_SOCK_CON];
4584 if (msg_ref) {
4585 parse_rtattr(con, TIPC_NLA_CON_MAX,
4586 RTA_DATA(msg_ref), RTA_PAYLOAD(msg_ref));
4587
4588 tipc_sock_addr_print(con[TIPC_NLA_CON_NODE],
4589 con[TIPC_NLA_CON_SOCK]);
4590 } else
4591 sock_addr_print("", "-", "", NULL);
4592
4593 if (show_details)
4594 sock_details_print(&ss);
4595
4596 proc_ctx_print(&ss);
4597
4598 if (show_tipcinfo) {
4599 if (oneline)
4600 out(" type:%s", stype_nameg[ss.type]);
4601 else
4602 out("\n type:%s", stype_nameg[ss.type]);
4603 out(" cong:%s ",
4604 stat[TIPC_NLA_SOCK_STAT_LINK_CONG] ? "link" :
4605 stat[TIPC_NLA_SOCK_STAT_CONN_CONG] ? "conn" : "none");
4606 out(" drop:%d ",
4607 rta_getattr_u32(stat[TIPC_NLA_SOCK_STAT_DROP]));
4608
4609 if (attrs[TIPC_NLA_SOCK_HAS_PUBL])
4610 out(" publ");
4611
4612 if (con[TIPC_NLA_CON_FLAG])
4613 out(" via {%u,%u} ",
4614 rta_getattr_u32(con[TIPC_NLA_CON_TYPE]),
4615 rta_getattr_u32(con[TIPC_NLA_CON_INST]));
4616 }
4617
4618 return 0;
4619 }
4620
4621 static int tipc_show(struct filter *f)
4622 {
4623 DIAG_REQUEST(req, struct tipc_sock_diag_req r);
4624
4625 memset(&req.r, 0, sizeof(req.r));
4626 req.r.sdiag_family = AF_TIPC;
4627 req.r.tidiag_states = f->states;
4628
4629 return handle_netlink_request(f, &req.nlh, sizeof(req), tipc_show_sock);
4630 }
4631
4632 struct sock_diag_msg {
4633 __u8 sdiag_family;
4634 };
4635
4636 static int generic_show_sock(struct nlmsghdr *nlh, void *arg)
4637 {
4638 struct sock_diag_msg *r = NLMSG_DATA(nlh);
4639 struct inet_diag_arg inet_arg = { .f = arg, .protocol = IPPROTO_MAX };
4640 int ret;
4641
4642 switch (r->sdiag_family) {
4643 case AF_INET:
4644 case AF_INET6:
4645 inet_arg.rth = inet_arg.f->rth_for_killing;
4646 ret = show_one_inet_sock(nlh, &inet_arg);
4647 break;
4648 case AF_UNIX:
4649 ret = unix_show_sock(nlh, arg);
4650 break;
4651 case AF_PACKET:
4652 ret = packet_show_sock(nlh, arg);
4653 break;
4654 case AF_NETLINK:
4655 ret = netlink_show_sock(nlh, arg);
4656 break;
4657 case AF_VSOCK:
4658 ret = vsock_show_sock(nlh, arg);
4659 break;
4660 case AF_XDP:
4661 ret = xdp_show_sock(nlh, arg);
4662 break;
4663 default:
4664 ret = -1;
4665 }
4666
4667 render();
4668
4669 return ret;
4670 }
4671
4672 static int handle_follow_request(struct filter *f)
4673 {
4674 int ret = 0;
4675 int groups = 0;
4676 struct rtnl_handle rth, rth2;
4677
4678 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << TCP_DB))
4679 groups |= 1 << (SKNLGRP_INET_TCP_DESTROY - 1);
4680 if (f->families & FAMILY_MASK(AF_INET) && f->dbs & (1 << UDP_DB))
4681 groups |= 1 << (SKNLGRP_INET_UDP_DESTROY - 1);
4682 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << TCP_DB))
4683 groups |= 1 << (SKNLGRP_INET6_TCP_DESTROY - 1);
4684 if (f->families & FAMILY_MASK(AF_INET6) && f->dbs & (1 << UDP_DB))
4685 groups |= 1 << (SKNLGRP_INET6_UDP_DESTROY - 1);
4686
4687 if (groups == 0)
4688 return -1;
4689
4690 if (rtnl_open_byproto(&rth, groups, NETLINK_SOCK_DIAG))
4691 return -1;
4692
4693 rth.dump = 0;
4694 rth.local.nl_pid = 0;
4695
4696 if (f->kill) {
4697 if (rtnl_open_byproto(&rth2, groups, NETLINK_SOCK_DIAG)) {
4698 rtnl_close(&rth);
4699 return -1;
4700 }
4701 f->rth_for_killing = &rth2;
4702 }
4703
4704 if (rtnl_dump_filter(&rth, generic_show_sock, f))
4705 ret = -1;
4706
4707 rtnl_close(&rth);
4708 if (f->rth_for_killing)
4709 rtnl_close(f->rth_for_killing);
4710 return ret;
4711 }
4712
4713 static int get_snmp_int(char *proto, char *key, int *result)
4714 {
4715 char buf[1024];
4716 FILE *fp;
4717 int protolen = strlen(proto);
4718 int keylen = strlen(key);
4719
4720 *result = 0;
4721
4722 if ((fp = net_snmp_open()) == NULL)
4723 return -1;
4724
4725 while (fgets(buf, sizeof(buf), fp) != NULL) {
4726 char *p = buf;
4727 int pos = 0;
4728
4729 if (memcmp(buf, proto, protolen))
4730 continue;
4731 while ((p = strchr(p, ' ')) != NULL) {
4732 pos++;
4733 p++;
4734 if (memcmp(p, key, keylen) == 0 &&
4735 (p[keylen] == ' ' || p[keylen] == '\n'))
4736 break;
4737 }
4738 if (fgets(buf, sizeof(buf), fp) == NULL)
4739 break;
4740 if (memcmp(buf, proto, protolen))
4741 break;
4742 p = buf;
4743 while ((p = strchr(p, ' ')) != NULL) {
4744 p++;
4745 if (--pos == 0) {
4746 sscanf(p, "%d", result);
4747 fclose(fp);
4748 return 0;
4749 }
4750 }
4751 }
4752
4753 fclose(fp);
4754 errno = ESRCH;
4755 return -1;
4756 }
4757
4758
4759 /* Get stats from sockstat */
4760
4761 struct ssummary {
4762 int socks;
4763 int tcp_mem;
4764 int tcp_total;
4765 int tcp_orphans;
4766 int tcp_tws;
4767 int tcp4_hashed;
4768 int udp4;
4769 int raw4;
4770 int frag4;
4771 int frag4_mem;
4772 int tcp6_hashed;
4773 int udp6;
4774 int raw6;
4775 int frag6;
4776 int frag6_mem;
4777 };
4778
4779 static void get_sockstat_line(char *line, struct ssummary *s)
4780 {
4781 char id[256], rem[256];
4782
4783 if (sscanf(line, "%[^ ] %[^\n]\n", id, rem) != 2)
4784 return;
4785
4786 if (strcmp(id, "sockets:") == 0)
4787 sscanf(rem, "%*s%d", &s->socks);
4788 else if (strcmp(id, "UDP:") == 0)
4789 sscanf(rem, "%*s%d", &s->udp4);
4790 else if (strcmp(id, "UDP6:") == 0)
4791 sscanf(rem, "%*s%d", &s->udp6);
4792 else if (strcmp(id, "RAW:") == 0)
4793 sscanf(rem, "%*s%d", &s->raw4);
4794 else if (strcmp(id, "RAW6:") == 0)
4795 sscanf(rem, "%*s%d", &s->raw6);
4796 else if (strcmp(id, "TCP6:") == 0)
4797 sscanf(rem, "%*s%d", &s->tcp6_hashed);
4798 else if (strcmp(id, "FRAG:") == 0)
4799 sscanf(rem, "%*s%d%*s%d", &s->frag4, &s->frag4_mem);
4800 else if (strcmp(id, "FRAG6:") == 0)
4801 sscanf(rem, "%*s%d%*s%d", &s->frag6, &s->frag6_mem);
4802 else if (strcmp(id, "TCP:") == 0)
4803 sscanf(rem, "%*s%d%*s%d%*s%d%*s%d%*s%d",
4804 &s->tcp4_hashed,
4805 &s->tcp_orphans, &s->tcp_tws, &s->tcp_total, &s->tcp_mem);
4806 }
4807
4808 static int get_sockstat(struct ssummary *s)
4809 {
4810 char buf[256];
4811 FILE *fp;
4812
4813 memset(s, 0, sizeof(*s));
4814
4815 if ((fp = net_sockstat_open()) == NULL)
4816 return -1;
4817 while (fgets(buf, sizeof(buf), fp) != NULL)
4818 get_sockstat_line(buf, s);
4819 fclose(fp);
4820
4821 if ((fp = net_sockstat6_open()) == NULL)
4822 return 0;
4823 while (fgets(buf, sizeof(buf), fp) != NULL)
4824 get_sockstat_line(buf, s);
4825 fclose(fp);
4826
4827 return 0;
4828 }
4829
4830 static int print_summary(void)
4831 {
4832 struct ssummary s;
4833 int tcp_estab;
4834
4835 if (get_sockstat(&s) < 0)
4836 perror("ss: get_sockstat");
4837 if (get_snmp_int("Tcp:", "CurrEstab", &tcp_estab) < 0)
4838 perror("ss: get_snmpstat");
4839
4840 printf("Total: %d\n", s.socks);
4841
4842 printf("TCP: %d (estab %d, closed %d, orphaned %d, timewait %d)\n",
4843 s.tcp_total + s.tcp_tws, tcp_estab,
4844 s.tcp_total - (s.tcp4_hashed + s.tcp6_hashed - s.tcp_tws),
4845 s.tcp_orphans, s.tcp_tws);
4846
4847 printf("\n");
4848 printf("Transport Total IP IPv6\n");
4849 printf("RAW %-9d %-9d %-9d\n", s.raw4+s.raw6, s.raw4, s.raw6);
4850 printf("UDP %-9d %-9d %-9d\n", s.udp4+s.udp6, s.udp4, s.udp6);
4851 printf("TCP %-9d %-9d %-9d\n", s.tcp4_hashed+s.tcp6_hashed, s.tcp4_hashed, s.tcp6_hashed);
4852 printf("INET %-9d %-9d %-9d\n",
4853 s.raw4+s.udp4+s.tcp4_hashed+
4854 s.raw6+s.udp6+s.tcp6_hashed,
4855 s.raw4+s.udp4+s.tcp4_hashed,
4856 s.raw6+s.udp6+s.tcp6_hashed);
4857 printf("FRAG %-9d %-9d %-9d\n", s.frag4+s.frag6, s.frag4, s.frag6);
4858
4859 printf("\n");
4860
4861 return 0;
4862 }
4863
4864 static void _usage(FILE *dest)
4865 {
4866 fprintf(dest,
4867 "Usage: ss [ OPTIONS ]\n"
4868 " ss [ OPTIONS ] [ FILTER ]\n"
4869 " -h, --help this message\n"
4870 " -V, --version output version information\n"
4871 " -n, --numeric don't resolve service names\n"
4872 " -r, --resolve resolve host names\n"
4873 " -a, --all display all sockets\n"
4874 " -l, --listening display listening sockets\n"
4875 " -o, --options show timer information\n"
4876 " -e, --extended show detailed socket information\n"
4877 " -m, --memory show socket memory usage\n"
4878 " -p, --processes show process using socket\n"
4879 " -i, --info show internal TCP information\n"
4880 " --tipcinfo show internal tipc socket information\n"
4881 " -s, --summary show socket usage summary\n"
4882 " --tos show tos and priority information\n"
4883 " -b, --bpf show bpf filter socket information\n"
4884 " -E, --events continually display sockets as they are destroyed\n"
4885 " -Z, --context display process SELinux security contexts\n"
4886 " -z, --contexts display process and socket SELinux security contexts\n"
4887 " -N, --net switch to the specified network namespace name\n"
4888 "\n"
4889 " -4, --ipv4 display only IP version 4 sockets\n"
4890 " -6, --ipv6 display only IP version 6 sockets\n"
4891 " -0, --packet display PACKET sockets\n"
4892 " -t, --tcp display only TCP sockets\n"
4893 " -S, --sctp display only SCTP sockets\n"
4894 " -u, --udp display only UDP sockets\n"
4895 " -d, --dccp display only DCCP sockets\n"
4896 " -w, --raw display only RAW sockets\n"
4897 " -x, --unix display only Unix domain sockets\n"
4898 " --tipc display only TIPC sockets\n"
4899 " --vsock display only vsock sockets\n"
4900 " -f, --family=FAMILY display sockets of type FAMILY\n"
4901 " FAMILY := {inet|inet6|link|unix|netlink|vsock|tipc|xdp|help}\n"
4902 "\n"
4903 " -K, --kill forcibly close sockets, display what was closed\n"
4904 " -H, --no-header Suppress header line\n"
4905 " -O, --oneline socket's data printed on a single line\n"
4906 "\n"
4907 " -A, --query=QUERY, --socket=QUERY\n"
4908 " QUERY := {all|inet|tcp|udp|raw|unix|unix_dgram|unix_stream|unix_seqpacket|packet|netlink|vsock_stream|vsock_dgram|tipc}[,QUERY]\n"
4909 "\n"
4910 " -D, --diag=FILE Dump raw information about TCP sockets to FILE\n"
4911 " -F, --filter=FILE read filter information from FILE\n"
4912 " FILTER := [ state STATE-FILTER ] [ EXPRESSION ]\n"
4913 " STATE-FILTER := {all|connected|synchronized|bucket|big|TCP-STATES}\n"
4914 " TCP-STATES := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|closed|close-wait|last-ack|listening|closing}\n"
4915 " connected := {established|syn-sent|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
4916 " synchronized := {established|syn-recv|fin-wait-{1,2}|time-wait|close-wait|last-ack|closing}\n"
4917 " bucket := {syn-recv|time-wait}\n"
4918 " big := {established|syn-sent|fin-wait-{1,2}|closed|close-wait|last-ack|listening|closing}\n"
4919 );
4920 }
4921
4922 static void help(void) __attribute__((noreturn));
4923 static void help(void)
4924 {
4925 _usage(stdout);
4926 exit(0);
4927 }
4928
4929 static void usage(void) __attribute__((noreturn));
4930 static void usage(void)
4931 {
4932 _usage(stderr);
4933 exit(-1);
4934 }
4935
4936
4937 static int scan_state(const char *state)
4938 {
4939 static const char * const sstate_namel[] = {
4940 "UNKNOWN",
4941 [SS_ESTABLISHED] = "established",
4942 [SS_SYN_SENT] = "syn-sent",
4943 [SS_SYN_RECV] = "syn-recv",
4944 [SS_FIN_WAIT1] = "fin-wait-1",
4945 [SS_FIN_WAIT2] = "fin-wait-2",
4946 [SS_TIME_WAIT] = "time-wait",
4947 [SS_CLOSE] = "unconnected",
4948 [SS_CLOSE_WAIT] = "close-wait",
4949 [SS_LAST_ACK] = "last-ack",
4950 [SS_LISTEN] = "listening",
4951 [SS_CLOSING] = "closing",
4952 };
4953 int i;
4954
4955 if (strcasecmp(state, "close") == 0 ||
4956 strcasecmp(state, "closed") == 0)
4957 return (1<<SS_CLOSE);
4958 if (strcasecmp(state, "syn-rcv") == 0)
4959 return (1<<SS_SYN_RECV);
4960 if (strcasecmp(state, "established") == 0)
4961 return (1<<SS_ESTABLISHED);
4962 if (strcasecmp(state, "all") == 0)
4963 return SS_ALL;
4964 if (strcasecmp(state, "connected") == 0)
4965 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN));
4966 if (strcasecmp(state, "synchronized") == 0)
4967 return SS_ALL & ~((1<<SS_CLOSE)|(1<<SS_LISTEN)|(1<<SS_SYN_SENT));
4968 if (strcasecmp(state, "bucket") == 0)
4969 return (1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT);
4970 if (strcasecmp(state, "big") == 0)
4971 return SS_ALL & ~((1<<SS_SYN_RECV)|(1<<SS_TIME_WAIT));
4972 for (i = 0; i < SS_MAX; i++) {
4973 if (strcasecmp(state, sstate_namel[i]) == 0)
4974 return (1<<i);
4975 }
4976
4977 fprintf(stderr, "ss: wrong state name: %s\n", state);
4978 exit(-1);
4979 }
4980
4981 /* Values 'v' and 'V' are already used so a non-character is used */
4982 #define OPT_VSOCK 256
4983
4984 /* Values of 't' are already used so a non-character is used */
4985 #define OPT_TIPCSOCK 257
4986 #define OPT_TIPCINFO 258
4987
4988 #define OPT_TOS 259
4989
4990 /* Values of 'x' are already used so a non-character is used */
4991 #define OPT_XDPSOCK 260
4992
4993 static const struct option long_opts[] = {
4994 { "numeric", 0, 0, 'n' },
4995 { "resolve", 0, 0, 'r' },
4996 { "options", 0, 0, 'o' },
4997 { "extended", 0, 0, 'e' },
4998 { "memory", 0, 0, 'm' },
4999 { "info", 0, 0, 'i' },
5000 { "processes", 0, 0, 'p' },
5001 { "bpf", 0, 0, 'b' },
5002 { "events", 0, 0, 'E' },
5003 { "dccp", 0, 0, 'd' },
5004 { "tcp", 0, 0, 't' },
5005 { "sctp", 0, 0, 'S' },
5006 { "udp", 0, 0, 'u' },
5007 { "raw", 0, 0, 'w' },
5008 { "unix", 0, 0, 'x' },
5009 { "tipc", 0, 0, OPT_TIPCSOCK},
5010 { "vsock", 0, 0, OPT_VSOCK },
5011 { "all", 0, 0, 'a' },
5012 { "listening", 0, 0, 'l' },
5013 { "ipv4", 0, 0, '4' },
5014 { "ipv6", 0, 0, '6' },
5015 { "packet", 0, 0, '0' },
5016 { "family", 1, 0, 'f' },
5017 { "socket", 1, 0, 'A' },
5018 { "query", 1, 0, 'A' },
5019 { "summary", 0, 0, 's' },
5020 { "diag", 1, 0, 'D' },
5021 { "filter", 1, 0, 'F' },
5022 { "version", 0, 0, 'V' },
5023 { "help", 0, 0, 'h' },
5024 { "context", 0, 0, 'Z' },
5025 { "contexts", 0, 0, 'z' },
5026 { "net", 1, 0, 'N' },
5027 { "tipcinfo", 0, 0, OPT_TIPCINFO},
5028 { "tos", 0, 0, OPT_TOS },
5029 { "kill", 0, 0, 'K' },
5030 { "no-header", 0, 0, 'H' },
5031 { "xdp", 0, 0, OPT_XDPSOCK},
5032 { "oneline", 0, 0, 'O' },
5033 { 0 }
5034
5035 };
5036
5037 int main(int argc, char *argv[])
5038 {
5039 int saw_states = 0;
5040 int saw_query = 0;
5041 int do_summary = 0;
5042 const char *dump_tcpdiag = NULL;
5043 FILE *filter_fp = NULL;
5044 int ch;
5045 int state_filter = 0;
5046
5047 while ((ch = getopt_long(argc, argv,
5048 "dhaletuwxnro460spbEf:miA:D:F:vVzZN:KHSO",
5049 long_opts, NULL)) != EOF) {
5050 switch (ch) {
5051 case 'n':
5052 numeric = 1;
5053 break;
5054 case 'r':
5055 resolve_hosts = 1;
5056 break;
5057 case 'o':
5058 show_options = 1;
5059 break;
5060 case 'e':
5061 show_options = 1;
5062 show_details++;
5063 break;
5064 case 'm':
5065 show_mem = 1;
5066 break;
5067 case 'i':
5068 show_tcpinfo = 1;
5069 break;
5070 case 'p':
5071 show_users++;
5072 user_ent_hash_build();
5073 break;
5074 case 'b':
5075 show_options = 1;
5076 show_bpf++;
5077 break;
5078 case 'E':
5079 follow_events = 1;
5080 break;
5081 case 'd':
5082 filter_db_set(&current_filter, DCCP_DB, true);
5083 break;
5084 case 't':
5085 filter_db_set(&current_filter, TCP_DB, true);
5086 break;
5087 case 'S':
5088 filter_db_set(&current_filter, SCTP_DB, true);
5089 break;
5090 case 'u':
5091 filter_db_set(&current_filter, UDP_DB, true);
5092 break;
5093 case 'w':
5094 filter_db_set(&current_filter, RAW_DB, true);
5095 break;
5096 case 'x':
5097 filter_af_set(&current_filter, AF_UNIX);
5098 break;
5099 case OPT_VSOCK:
5100 filter_af_set(&current_filter, AF_VSOCK);
5101 break;
5102 case OPT_TIPCSOCK:
5103 filter_af_set(&current_filter, AF_TIPC);
5104 break;
5105 case 'a':
5106 state_filter = SS_ALL;
5107 break;
5108 case 'l':
5109 state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
5110 break;
5111 case '4':
5112 filter_af_set(&current_filter, AF_INET);
5113 break;
5114 case '6':
5115 filter_af_set(&current_filter, AF_INET6);
5116 break;
5117 case '0':
5118 filter_af_set(&current_filter, AF_PACKET);
5119 break;
5120 case OPT_XDPSOCK:
5121 filter_af_set(&current_filter, AF_XDP);
5122 break;
5123 case 'f':
5124 if (strcmp(optarg, "inet") == 0)
5125 filter_af_set(&current_filter, AF_INET);
5126 else if (strcmp(optarg, "inet6") == 0)
5127 filter_af_set(&current_filter, AF_INET6);
5128 else if (strcmp(optarg, "link") == 0)
5129 filter_af_set(&current_filter, AF_PACKET);
5130 else if (strcmp(optarg, "unix") == 0)
5131 filter_af_set(&current_filter, AF_UNIX);
5132 else if (strcmp(optarg, "netlink") == 0)
5133 filter_af_set(&current_filter, AF_NETLINK);
5134 else if (strcmp(optarg, "tipc") == 0)
5135 filter_af_set(&current_filter, AF_TIPC);
5136 else if (strcmp(optarg, "vsock") == 0)
5137 filter_af_set(&current_filter, AF_VSOCK);
5138 else if (strcmp(optarg, "xdp") == 0)
5139 filter_af_set(&current_filter, AF_XDP);
5140 else if (strcmp(optarg, "help") == 0)
5141 help();
5142 else {
5143 fprintf(stderr, "ss: \"%s\" is invalid family\n",
5144 optarg);
5145 usage();
5146 }
5147 break;
5148 case 'A':
5149 {
5150 char *p, *p1;
5151
5152 if (!saw_query) {
5153 current_filter.dbs = 0;
5154 state_filter = state_filter ?
5155 state_filter : SS_CONN;
5156 saw_query = 1;
5157 do_default = 0;
5158 }
5159 p = p1 = optarg;
5160 do {
5161 if ((p1 = strchr(p, ',')) != NULL)
5162 *p1 = 0;
5163 if (filter_db_parse(&current_filter, p)) {
5164 fprintf(stderr, "ss: \"%s\" is illegal socket table id\n", p);
5165 usage();
5166 }
5167 p = p1 + 1;
5168 } while (p1);
5169 break;
5170 }
5171 case 's':
5172 do_summary = 1;
5173 break;
5174 case 'D':
5175 dump_tcpdiag = optarg;
5176 break;
5177 case 'F':
5178 if (filter_fp) {
5179 fprintf(stderr, "More than one filter file\n");
5180 exit(-1);
5181 }
5182 if (optarg[0] == '-')
5183 filter_fp = stdin;
5184 else
5185 filter_fp = fopen(optarg, "r");
5186 if (!filter_fp) {
5187 perror("fopen filter file");
5188 exit(-1);
5189 }
5190 break;
5191 case 'v':
5192 case 'V':
5193 printf("ss utility, iproute2-ss%s\n", SNAPSHOT);
5194 exit(0);
5195 case 'z':
5196 show_sock_ctx++;
5197 /* fall through */
5198 case 'Z':
5199 if (is_selinux_enabled() <= 0) {
5200 fprintf(stderr, "ss: SELinux is not enabled.\n");
5201 exit(1);
5202 }
5203 show_proc_ctx++;
5204 user_ent_hash_build();
5205 break;
5206 case 'N':
5207 if (netns_switch(optarg))
5208 exit(1);
5209 break;
5210 case OPT_TIPCINFO:
5211 show_tipcinfo = 1;
5212 break;
5213 case OPT_TOS:
5214 show_tos = 1;
5215 break;
5216 case 'K':
5217 current_filter.kill = 1;
5218 break;
5219 case 'H':
5220 show_header = 0;
5221 break;
5222 case 'O':
5223 oneline = 1;
5224 break;
5225 case 'h':
5226 help();
5227 case '?':
5228 default:
5229 usage();
5230 }
5231 }
5232
5233 argc -= optind;
5234 argv += optind;
5235
5236 if (do_summary) {
5237 print_summary();
5238 if (do_default && argc == 0)
5239 exit(0);
5240 }
5241
5242 while (argc > 0) {
5243 if (strcmp(*argv, "state") == 0) {
5244 NEXT_ARG();
5245 if (!saw_states)
5246 state_filter = 0;
5247 state_filter |= scan_state(*argv);
5248 saw_states = 1;
5249 } else if (strcmp(*argv, "exclude") == 0 ||
5250 strcmp(*argv, "excl") == 0) {
5251 NEXT_ARG();
5252 if (!saw_states)
5253 state_filter = SS_ALL;
5254 state_filter &= ~scan_state(*argv);
5255 saw_states = 1;
5256 } else {
5257 break;
5258 }
5259 argc--; argv++;
5260 }
5261
5262 if (do_default) {
5263 state_filter = state_filter ? state_filter : SS_CONN;
5264 filter_db_parse(&current_filter, "all");
5265 }
5266
5267 filter_states_set(&current_filter, state_filter);
5268 filter_merge_defaults(&current_filter);
5269
5270 if (!numeric && resolve_hosts &&
5271 (current_filter.dbs & (UNIX_DBM|INET_L4_DBM)))
5272 init_service_resolver();
5273
5274 if (current_filter.dbs == 0) {
5275 fprintf(stderr, "ss: no socket tables to show with such filter.\n");
5276 exit(0);
5277 }
5278 if (current_filter.families == 0) {
5279 fprintf(stderr, "ss: no families to show with such filter.\n");
5280 exit(0);
5281 }
5282 if (current_filter.states == 0) {
5283 fprintf(stderr, "ss: no socket states to show with such filter.\n");
5284 exit(0);
5285 }
5286
5287 if (dump_tcpdiag) {
5288 FILE *dump_fp = stdout;
5289
5290 if (!(current_filter.dbs & (1<<TCP_DB))) {
5291 fprintf(stderr, "ss: tcpdiag dump requested and no tcp in filter.\n");
5292 exit(0);
5293 }
5294 if (dump_tcpdiag[0] != '-') {
5295 dump_fp = fopen(dump_tcpdiag, "w");
5296 if (!dump_tcpdiag) {
5297 perror("fopen dump file");
5298 exit(-1);
5299 }
5300 }
5301 inet_show_netlink(&current_filter, dump_fp, IPPROTO_TCP);
5302 fflush(dump_fp);
5303 exit(0);
5304 }
5305
5306 if (ssfilter_parse(&current_filter.f, argc, argv, filter_fp))
5307 usage();
5308
5309 if (!(current_filter.dbs & (current_filter.dbs - 1)))
5310 columns[COL_NETID].disabled = 1;
5311
5312 if (!(current_filter.states & (current_filter.states - 1)))
5313 columns[COL_STATE].disabled = 1;
5314
5315 if (show_header)
5316 print_header();
5317
5318 fflush(stdout);
5319
5320 if (follow_events)
5321 exit(handle_follow_request(&current_filter));
5322
5323 if (current_filter.dbs & (1<<NETLINK_DB))
5324 netlink_show(&current_filter);
5325 if (current_filter.dbs & PACKET_DBM)
5326 packet_show(&current_filter);
5327 if (current_filter.dbs & UNIX_DBM)
5328 unix_show(&current_filter);
5329 if (current_filter.dbs & (1<<RAW_DB))
5330 raw_show(&current_filter);
5331 if (current_filter.dbs & (1<<UDP_DB))
5332 udp_show(&current_filter);
5333 if (current_filter.dbs & (1<<TCP_DB))
5334 tcp_show(&current_filter);
5335 if (current_filter.dbs & (1<<DCCP_DB))
5336 dccp_show(&current_filter);
5337 if (current_filter.dbs & (1<<SCTP_DB))
5338 sctp_show(&current_filter);
5339 if (current_filter.dbs & VSOCK_DBM)
5340 vsock_show(&current_filter);
5341 if (current_filter.dbs & (1<<TIPC_DB))
5342 tipc_show(&current_filter);
5343 if (current_filter.dbs & (1<<XDP_DB))
5344 xdp_show(&current_filter);
5345
5346 if (show_users || show_proc_ctx || show_sock_ctx)
5347 user_ent_destroy();
5348
5349 render();
5350
5351 return 0;
5352 }
Upstream mirror of iproute2