1 // SPDX-License-Identifier: GPL-2.0
3 * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
5 * Begun April 1, 1996, Mike Shaver.
6 * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
10 #include <linux/module.h>
11 #include <linux/sysctl.h>
12 #include <linux/igmp.h>
13 #include <linux/inetdevice.h>
14 #include <linux/seqlock.h>
15 #include <linux/init.h>
16 #include <linux/slab.h>
17 #include <linux/nsproxy.h>
18 #include <linux/swap.h>
22 #include <net/route.h>
25 #include <net/cipso_ipv4.h>
26 #include <net/inet_frag.h>
28 #include <net/protocol.h>
29 #include <net/netevent.h>
35 static int thousand = 1000;
36 static int gso_max_segs = GSO_MAX_SEGS;
37 static int tcp_retr1_max = 255;
38 static int ip_local_port_range_min[] = { 1, 1 };
39 static int ip_local_port_range_max[] = { 65535, 65535 };
40 static int tcp_adv_win_scale_min = -31;
41 static int tcp_adv_win_scale_max = 31;
42 static int ip_privileged_port_min;
43 static int ip_privileged_port_max = 65535;
44 static int ip_ttl_min = 1;
45 static int ip_ttl_max = 255;
46 static int tcp_syn_retries_min = 1;
47 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
48 static int ip_ping_group_range_min[] = { 0, 0 };
49 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
50 static int comp_sack_nr_max = 255;
53 static int sysctl_tcp_low_latency __read_mostly;
55 /* Update system visible IP port range */
56 static void set_local_port_range(struct net *net, int range[2])
58 bool same_parity = !((range[0] ^ range[1]) & 1);
60 write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
61 if (same_parity && !net->ipv4.ip_local_ports.warned) {
62 net->ipv4.ip_local_ports.warned = true;
63 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
65 net->ipv4.ip_local_ports.range[0] = range[0];
66 net->ipv4.ip_local_ports.range[1] = range[1];
67 write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
70 /* Validate changes from /proc interface. */
71 static int ipv4_local_port_range(struct ctl_table *table, int write,
73 size_t *lenp, loff_t *ppos)
76 container_of(table->data, struct net, ipv4.ip_local_ports.range);
79 struct ctl_table tmp = {
81 .maxlen = sizeof(range),
83 .extra1 = &ip_local_port_range_min,
84 .extra2 = &ip_local_port_range_max,
87 inet_get_local_port_range(net, &range[0], &range[1]);
89 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
91 if (write && ret == 0) {
92 /* Ensure that the upper limit is not smaller than the lower,
93 * and that the lower does not encroach upon the privileged
96 if ((range[1] < range[0]) ||
97 (range[0] < net->ipv4.sysctl_ip_prot_sock))
100 set_local_port_range(net, range);
106 /* Validate changes from /proc interface. */
107 static int ipv4_privileged_ports(struct ctl_table *table, int write,
108 void __user *buffer, size_t *lenp, loff_t *ppos)
110 struct net *net = container_of(table->data, struct net,
111 ipv4.sysctl_ip_prot_sock);
115 struct ctl_table tmp = {
117 .maxlen = sizeof(pports),
119 .extra1 = &ip_privileged_port_min,
120 .extra2 = &ip_privileged_port_max,
123 pports = net->ipv4.sysctl_ip_prot_sock;
125 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
127 if (write && ret == 0) {
128 inet_get_local_port_range(net, &range[0], &range[1]);
129 /* Ensure that the local port range doesn't overlap with the
130 * privileged port range.
132 if (range[0] < pports)
135 net->ipv4.sysctl_ip_prot_sock = pports;
141 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
143 kgid_t *data = table->data;
145 container_of(table->data, struct net, ipv4.ping_group_range.range);
148 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
152 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
155 /* Update system visible IP port range */
156 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
158 kgid_t *data = table->data;
160 container_of(table->data, struct net, ipv4.ping_group_range.range);
161 write_seqlock(&net->ipv4.ping_group_range.lock);
164 write_sequnlock(&net->ipv4.ping_group_range.lock);
167 /* Validate changes from /proc interface. */
168 static int ipv4_ping_group_range(struct ctl_table *table, int write,
170 size_t *lenp, loff_t *ppos)
172 struct user_namespace *user_ns = current_user_ns();
176 struct ctl_table tmp = {
178 .maxlen = sizeof(urange),
180 .extra1 = &ip_ping_group_range_min,
181 .extra2 = &ip_ping_group_range_max,
184 inet_get_ping_group_range_table(table, &low, &high);
185 urange[0] = from_kgid_munged(user_ns, low);
186 urange[1] = from_kgid_munged(user_ns, high);
187 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
189 if (write && ret == 0) {
190 low = make_kgid(user_ns, urange[0]);
191 high = make_kgid(user_ns, urange[1]);
192 if (!gid_valid(low) || !gid_valid(high) ||
193 (urange[1] < urange[0]) || gid_lt(high, low)) {
194 low = make_kgid(&init_user_ns, 1);
195 high = make_kgid(&init_user_ns, 0);
197 set_ping_group_range(table, low, high);
203 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
204 void __user *buffer, size_t *lenp, loff_t *ppos)
206 struct net *net = container_of(ctl->data, struct net,
207 ipv4.tcp_congestion_control);
208 char val[TCP_CA_NAME_MAX];
209 struct ctl_table tbl = {
211 .maxlen = TCP_CA_NAME_MAX,
215 tcp_get_default_congestion_control(net, val);
217 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
218 if (write && ret == 0)
219 ret = tcp_set_default_congestion_control(net, val);
223 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
225 void __user *buffer, size_t *lenp,
228 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
231 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
234 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
235 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
240 static int proc_allowed_congestion_control(struct ctl_table *ctl,
242 void __user *buffer, size_t *lenp,
245 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
248 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
252 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
253 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
254 if (write && ret == 0)
255 ret = tcp_set_allowed_congestion_control(tbl.data);
260 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
261 void __user *buffer, size_t *lenp,
264 struct net *net = container_of(table->data, struct net,
265 ipv4.sysctl_tcp_fastopen);
266 struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
267 struct tcp_fastopen_context *ctxt;
269 u32 user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
271 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
276 ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
278 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
280 memset(user_key, 0, sizeof(user_key));
283 snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
284 user_key[0], user_key[1], user_key[2], user_key[3]);
285 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
287 if (write && ret == 0) {
288 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
289 user_key + 2, user_key + 3) != 4) {
293 tcp_fastopen_reset_cipher(net, NULL, user_key,
294 TCP_FASTOPEN_KEY_LENGTH);
298 pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
299 user_key[0], user_key[1], user_key[2], user_key[3],
300 (char *)tbl.data, ret);
305 static void proc_configure_early_demux(int enabled, int protocol)
307 struct net_protocol *ipprot;
308 #if IS_ENABLED(CONFIG_IPV6)
309 struct inet6_protocol *ip6prot;
314 ipprot = rcu_dereference(inet_protos[protocol]);
316 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
319 #if IS_ENABLED(CONFIG_IPV6)
320 ip6prot = rcu_dereference(inet6_protos[protocol]);
322 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
328 static int proc_tcp_early_demux(struct ctl_table *table, int write,
329 void __user *buffer, size_t *lenp, loff_t *ppos)
333 ret = proc_dointvec(table, write, buffer, lenp, ppos);
336 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
338 proc_configure_early_demux(enabled, IPPROTO_TCP);
344 static int proc_udp_early_demux(struct ctl_table *table, int write,
345 void __user *buffer, size_t *lenp, loff_t *ppos)
349 ret = proc_dointvec(table, write, buffer, lenp, ppos);
352 int enabled = init_net.ipv4.sysctl_udp_early_demux;
354 proc_configure_early_demux(enabled, IPPROTO_UDP);
360 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
363 size_t *lenp, loff_t *ppos)
365 struct net *net = container_of(table->data, struct net,
366 ipv4.sysctl_tcp_fastopen_blackhole_timeout);
369 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
370 if (write && ret == 0)
371 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
376 static int proc_tcp_available_ulp(struct ctl_table *ctl,
378 void __user *buffer, size_t *lenp,
381 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
384 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
387 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
388 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
394 #ifdef CONFIG_IP_ROUTE_MULTIPATH
395 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
396 void __user *buffer, size_t *lenp,
399 struct net *net = container_of(table->data, struct net,
400 ipv4.sysctl_fib_multipath_hash_policy);
403 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
404 if (write && ret == 0)
405 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
411 static struct ctl_table ipv4_table[] = {
413 .procname = "tcp_max_orphans",
414 .data = &sysctl_tcp_max_orphans,
415 .maxlen = sizeof(int),
417 .proc_handler = proc_dointvec
420 .procname = "inet_peer_threshold",
421 .data = &inet_peer_threshold,
422 .maxlen = sizeof(int),
424 .proc_handler = proc_dointvec
427 .procname = "inet_peer_minttl",
428 .data = &inet_peer_minttl,
429 .maxlen = sizeof(int),
431 .proc_handler = proc_dointvec_jiffies,
434 .procname = "inet_peer_maxttl",
435 .data = &inet_peer_maxttl,
436 .maxlen = sizeof(int),
438 .proc_handler = proc_dointvec_jiffies,
441 .procname = "tcp_mem",
442 .maxlen = sizeof(sysctl_tcp_mem),
443 .data = &sysctl_tcp_mem,
445 .proc_handler = proc_doulongvec_minmax,
448 .procname = "tcp_low_latency",
449 .data = &sysctl_tcp_low_latency,
450 .maxlen = sizeof(int),
452 .proc_handler = proc_dointvec
454 #ifdef CONFIG_NETLABEL
456 .procname = "cipso_cache_enable",
457 .data = &cipso_v4_cache_enabled,
458 .maxlen = sizeof(int),
460 .proc_handler = proc_dointvec,
463 .procname = "cipso_cache_bucket_size",
464 .data = &cipso_v4_cache_bucketsize,
465 .maxlen = sizeof(int),
467 .proc_handler = proc_dointvec,
470 .procname = "cipso_rbm_optfmt",
471 .data = &cipso_v4_rbm_optfmt,
472 .maxlen = sizeof(int),
474 .proc_handler = proc_dointvec,
477 .procname = "cipso_rbm_strictvalid",
478 .data = &cipso_v4_rbm_strictvalid,
479 .maxlen = sizeof(int),
481 .proc_handler = proc_dointvec,
483 #endif /* CONFIG_NETLABEL */
485 .procname = "tcp_available_congestion_control",
486 .maxlen = TCP_CA_BUF_MAX,
488 .proc_handler = proc_tcp_available_congestion_control,
491 .procname = "tcp_allowed_congestion_control",
492 .maxlen = TCP_CA_BUF_MAX,
494 .proc_handler = proc_allowed_congestion_control,
497 .procname = "tcp_available_ulp",
498 .maxlen = TCP_ULP_BUF_MAX,
500 .proc_handler = proc_tcp_available_ulp,
503 .procname = "icmp_msgs_per_sec",
504 .data = &sysctl_icmp_msgs_per_sec,
505 .maxlen = sizeof(int),
507 .proc_handler = proc_dointvec_minmax,
511 .procname = "icmp_msgs_burst",
512 .data = &sysctl_icmp_msgs_burst,
513 .maxlen = sizeof(int),
515 .proc_handler = proc_dointvec_minmax,
519 .procname = "udp_mem",
520 .data = &sysctl_udp_mem,
521 .maxlen = sizeof(sysctl_udp_mem),
523 .proc_handler = proc_doulongvec_minmax,
528 static struct ctl_table ipv4_net_table[] = {
530 .procname = "icmp_echo_ignore_all",
531 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
532 .maxlen = sizeof(int),
534 .proc_handler = proc_dointvec
537 .procname = "icmp_echo_ignore_broadcasts",
538 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
539 .maxlen = sizeof(int),
541 .proc_handler = proc_dointvec
544 .procname = "icmp_ignore_bogus_error_responses",
545 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
546 .maxlen = sizeof(int),
548 .proc_handler = proc_dointvec
551 .procname = "icmp_errors_use_inbound_ifaddr",
552 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
553 .maxlen = sizeof(int),
555 .proc_handler = proc_dointvec
558 .procname = "icmp_ratelimit",
559 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
560 .maxlen = sizeof(int),
562 .proc_handler = proc_dointvec_ms_jiffies,
565 .procname = "icmp_ratemask",
566 .data = &init_net.ipv4.sysctl_icmp_ratemask,
567 .maxlen = sizeof(int),
569 .proc_handler = proc_dointvec
572 .procname = "ping_group_range",
573 .data = &init_net.ipv4.ping_group_range.range,
574 .maxlen = sizeof(gid_t)*2,
576 .proc_handler = ipv4_ping_group_range,
579 .procname = "tcp_ecn",
580 .data = &init_net.ipv4.sysctl_tcp_ecn,
581 .maxlen = sizeof(int),
583 .proc_handler = proc_dointvec
586 .procname = "tcp_ecn_fallback",
587 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
588 .maxlen = sizeof(int),
590 .proc_handler = proc_dointvec
593 .procname = "ip_dynaddr",
594 .data = &init_net.ipv4.sysctl_ip_dynaddr,
595 .maxlen = sizeof(int),
597 .proc_handler = proc_dointvec
600 .procname = "ip_early_demux",
601 .data = &init_net.ipv4.sysctl_ip_early_demux,
602 .maxlen = sizeof(int),
604 .proc_handler = proc_dointvec
607 .procname = "udp_early_demux",
608 .data = &init_net.ipv4.sysctl_udp_early_demux,
609 .maxlen = sizeof(int),
611 .proc_handler = proc_udp_early_demux
614 .procname = "tcp_early_demux",
615 .data = &init_net.ipv4.sysctl_tcp_early_demux,
616 .maxlen = sizeof(int),
618 .proc_handler = proc_tcp_early_demux
621 .procname = "ip_default_ttl",
622 .data = &init_net.ipv4.sysctl_ip_default_ttl,
623 .maxlen = sizeof(int),
625 .proc_handler = proc_dointvec_minmax,
626 .extra1 = &ip_ttl_min,
627 .extra2 = &ip_ttl_max,
630 .procname = "ip_local_port_range",
631 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
632 .data = &init_net.ipv4.ip_local_ports.range,
634 .proc_handler = ipv4_local_port_range,
637 .procname = "ip_local_reserved_ports",
638 .data = &init_net.ipv4.sysctl_local_reserved_ports,
641 .proc_handler = proc_do_large_bitmap,
644 .procname = "ip_no_pmtu_disc",
645 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
646 .maxlen = sizeof(int),
648 .proc_handler = proc_dointvec
651 .procname = "ip_forward_use_pmtu",
652 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
653 .maxlen = sizeof(int),
655 .proc_handler = proc_dointvec,
658 .procname = "ip_nonlocal_bind",
659 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
660 .maxlen = sizeof(int),
662 .proc_handler = proc_dointvec
665 .procname = "fwmark_reflect",
666 .data = &init_net.ipv4.sysctl_fwmark_reflect,
667 .maxlen = sizeof(int),
669 .proc_handler = proc_dointvec,
672 .procname = "tcp_fwmark_accept",
673 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept,
674 .maxlen = sizeof(int),
676 .proc_handler = proc_dointvec,
678 #ifdef CONFIG_NET_L3_MASTER_DEV
680 .procname = "tcp_l3mdev_accept",
681 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
682 .maxlen = sizeof(int),
684 .proc_handler = proc_dointvec_minmax,
690 .procname = "tcp_mtu_probing",
691 .data = &init_net.ipv4.sysctl_tcp_mtu_probing,
692 .maxlen = sizeof(int),
694 .proc_handler = proc_dointvec,
697 .procname = "tcp_base_mss",
698 .data = &init_net.ipv4.sysctl_tcp_base_mss,
699 .maxlen = sizeof(int),
701 .proc_handler = proc_dointvec,
704 .procname = "tcp_probe_threshold",
705 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
706 .maxlen = sizeof(int),
708 .proc_handler = proc_dointvec,
711 .procname = "tcp_probe_interval",
712 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
713 .maxlen = sizeof(int),
715 .proc_handler = proc_dointvec,
718 .procname = "igmp_link_local_mcast_reports",
719 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
720 .maxlen = sizeof(int),
722 .proc_handler = proc_dointvec
725 .procname = "igmp_max_memberships",
726 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
727 .maxlen = sizeof(int),
729 .proc_handler = proc_dointvec
732 .procname = "igmp_max_msf",
733 .data = &init_net.ipv4.sysctl_igmp_max_msf,
734 .maxlen = sizeof(int),
736 .proc_handler = proc_dointvec
738 #ifdef CONFIG_IP_MULTICAST
740 .procname = "igmp_qrv",
741 .data = &init_net.ipv4.sysctl_igmp_qrv,
742 .maxlen = sizeof(int),
744 .proc_handler = proc_dointvec_minmax,
749 .procname = "tcp_congestion_control",
750 .data = &init_net.ipv4.tcp_congestion_control,
752 .maxlen = TCP_CA_NAME_MAX,
753 .proc_handler = proc_tcp_congestion_control,
756 .procname = "tcp_keepalive_time",
757 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
758 .maxlen = sizeof(int),
760 .proc_handler = proc_dointvec_jiffies,
763 .procname = "tcp_keepalive_probes",
764 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
765 .maxlen = sizeof(int),
767 .proc_handler = proc_dointvec
770 .procname = "tcp_keepalive_intvl",
771 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
772 .maxlen = sizeof(int),
774 .proc_handler = proc_dointvec_jiffies,
777 .procname = "tcp_syn_retries",
778 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
779 .maxlen = sizeof(int),
781 .proc_handler = proc_dointvec_minmax,
782 .extra1 = &tcp_syn_retries_min,
783 .extra2 = &tcp_syn_retries_max
786 .procname = "tcp_synack_retries",
787 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
788 .maxlen = sizeof(int),
790 .proc_handler = proc_dointvec
792 #ifdef CONFIG_SYN_COOKIES
794 .procname = "tcp_syncookies",
795 .data = &init_net.ipv4.sysctl_tcp_syncookies,
796 .maxlen = sizeof(int),
798 .proc_handler = proc_dointvec
802 .procname = "tcp_reordering",
803 .data = &init_net.ipv4.sysctl_tcp_reordering,
804 .maxlen = sizeof(int),
806 .proc_handler = proc_dointvec
809 .procname = "tcp_retries1",
810 .data = &init_net.ipv4.sysctl_tcp_retries1,
811 .maxlen = sizeof(int),
813 .proc_handler = proc_dointvec_minmax,
814 .extra2 = &tcp_retr1_max
817 .procname = "tcp_retries2",
818 .data = &init_net.ipv4.sysctl_tcp_retries2,
819 .maxlen = sizeof(int),
821 .proc_handler = proc_dointvec
824 .procname = "tcp_orphan_retries",
825 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
826 .maxlen = sizeof(int),
828 .proc_handler = proc_dointvec
831 .procname = "tcp_fin_timeout",
832 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
833 .maxlen = sizeof(int),
835 .proc_handler = proc_dointvec_jiffies,
838 .procname = "tcp_notsent_lowat",
839 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
840 .maxlen = sizeof(unsigned int),
842 .proc_handler = proc_douintvec,
845 .procname = "tcp_tw_reuse",
846 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
847 .maxlen = sizeof(int),
849 .proc_handler = proc_dointvec_minmax,
854 .procname = "tcp_max_tw_buckets",
855 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
856 .maxlen = sizeof(int),
858 .proc_handler = proc_dointvec
861 .procname = "tcp_max_syn_backlog",
862 .data = &init_net.ipv4.sysctl_max_syn_backlog,
863 .maxlen = sizeof(int),
865 .proc_handler = proc_dointvec
868 .procname = "tcp_fastopen",
869 .data = &init_net.ipv4.sysctl_tcp_fastopen,
870 .maxlen = sizeof(int),
872 .proc_handler = proc_dointvec,
875 .procname = "tcp_fastopen_key",
877 .data = &init_net.ipv4.sysctl_tcp_fastopen,
878 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
879 .proc_handler = proc_tcp_fastopen_key,
882 .procname = "tcp_fastopen_blackhole_timeout_sec",
883 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
884 .maxlen = sizeof(int),
886 .proc_handler = proc_tfo_blackhole_detect_timeout,
889 #ifdef CONFIG_IP_ROUTE_MULTIPATH
891 .procname = "fib_multipath_use_neigh",
892 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
893 .maxlen = sizeof(int),
895 .proc_handler = proc_dointvec_minmax,
900 .procname = "fib_multipath_hash_policy",
901 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
902 .maxlen = sizeof(int),
904 .proc_handler = proc_fib_multipath_hash_policy,
910 .procname = "ip_unprivileged_port_start",
911 .maxlen = sizeof(int),
912 .data = &init_net.ipv4.sysctl_ip_prot_sock,
914 .proc_handler = ipv4_privileged_ports,
916 #ifdef CONFIG_NET_L3_MASTER_DEV
918 .procname = "udp_l3mdev_accept",
919 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
920 .maxlen = sizeof(int),
922 .proc_handler = proc_dointvec_minmax,
928 .procname = "tcp_sack",
929 .data = &init_net.ipv4.sysctl_tcp_sack,
930 .maxlen = sizeof(int),
932 .proc_handler = proc_dointvec
935 .procname = "tcp_window_scaling",
936 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
937 .maxlen = sizeof(int),
939 .proc_handler = proc_dointvec
942 .procname = "tcp_timestamps",
943 .data = &init_net.ipv4.sysctl_tcp_timestamps,
944 .maxlen = sizeof(int),
946 .proc_handler = proc_dointvec
949 .procname = "tcp_early_retrans",
950 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
951 .maxlen = sizeof(int),
953 .proc_handler = proc_dointvec_minmax,
958 .procname = "tcp_recovery",
959 .data = &init_net.ipv4.sysctl_tcp_recovery,
960 .maxlen = sizeof(int),
962 .proc_handler = proc_dointvec,
965 .procname = "tcp_thin_linear_timeouts",
966 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
967 .maxlen = sizeof(int),
969 .proc_handler = proc_dointvec
972 .procname = "tcp_slow_start_after_idle",
973 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
974 .maxlen = sizeof(int),
976 .proc_handler = proc_dointvec
979 .procname = "tcp_retrans_collapse",
980 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
981 .maxlen = sizeof(int),
983 .proc_handler = proc_dointvec
986 .procname = "tcp_stdurg",
987 .data = &init_net.ipv4.sysctl_tcp_stdurg,
988 .maxlen = sizeof(int),
990 .proc_handler = proc_dointvec
993 .procname = "tcp_rfc1337",
994 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
995 .maxlen = sizeof(int),
997 .proc_handler = proc_dointvec
1000 .procname = "tcp_abort_on_overflow",
1001 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1002 .maxlen = sizeof(int),
1004 .proc_handler = proc_dointvec
1007 .procname = "tcp_fack",
1008 .data = &init_net.ipv4.sysctl_tcp_fack,
1009 .maxlen = sizeof(int),
1011 .proc_handler = proc_dointvec
1014 .procname = "tcp_max_reordering",
1015 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1016 .maxlen = sizeof(int),
1018 .proc_handler = proc_dointvec
1021 .procname = "tcp_dsack",
1022 .data = &init_net.ipv4.sysctl_tcp_dsack,
1023 .maxlen = sizeof(int),
1025 .proc_handler = proc_dointvec
1028 .procname = "tcp_app_win",
1029 .data = &init_net.ipv4.sysctl_tcp_app_win,
1030 .maxlen = sizeof(int),
1032 .proc_handler = proc_dointvec
1035 .procname = "tcp_adv_win_scale",
1036 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1037 .maxlen = sizeof(int),
1039 .proc_handler = proc_dointvec_minmax,
1040 .extra1 = &tcp_adv_win_scale_min,
1041 .extra2 = &tcp_adv_win_scale_max,
1044 .procname = "tcp_frto",
1045 .data = &init_net.ipv4.sysctl_tcp_frto,
1046 .maxlen = sizeof(int),
1048 .proc_handler = proc_dointvec
1051 .procname = "tcp_no_metrics_save",
1052 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1053 .maxlen = sizeof(int),
1055 .proc_handler = proc_dointvec,
1058 .procname = "tcp_moderate_rcvbuf",
1059 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1060 .maxlen = sizeof(int),
1062 .proc_handler = proc_dointvec,
1065 .procname = "tcp_tso_win_divisor",
1066 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1067 .maxlen = sizeof(int),
1069 .proc_handler = proc_dointvec,
1072 .procname = "tcp_workaround_signed_windows",
1073 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1074 .maxlen = sizeof(int),
1076 .proc_handler = proc_dointvec
1079 .procname = "tcp_limit_output_bytes",
1080 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1081 .maxlen = sizeof(int),
1083 .proc_handler = proc_dointvec
1086 .procname = "tcp_challenge_ack_limit",
1087 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1088 .maxlen = sizeof(int),
1090 .proc_handler = proc_dointvec
1093 .procname = "tcp_min_tso_segs",
1094 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1095 .maxlen = sizeof(int),
1097 .proc_handler = proc_dointvec_minmax,
1099 .extra2 = &gso_max_segs,
1102 .procname = "tcp_min_rtt_wlen",
1103 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1104 .maxlen = sizeof(int),
1106 .proc_handler = proc_dointvec
1109 .procname = "tcp_autocorking",
1110 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1111 .maxlen = sizeof(int),
1113 .proc_handler = proc_dointvec_minmax,
1118 .procname = "tcp_invalid_ratelimit",
1119 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1120 .maxlen = sizeof(int),
1122 .proc_handler = proc_dointvec_ms_jiffies,
1125 .procname = "tcp_pacing_ss_ratio",
1126 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1127 .maxlen = sizeof(int),
1129 .proc_handler = proc_dointvec_minmax,
1131 .extra2 = &thousand,
1134 .procname = "tcp_pacing_ca_ratio",
1135 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1136 .maxlen = sizeof(int),
1138 .proc_handler = proc_dointvec_minmax,
1140 .extra2 = &thousand,
1143 .procname = "tcp_wmem",
1144 .data = &init_net.ipv4.sysctl_tcp_wmem,
1145 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1147 .proc_handler = proc_dointvec_minmax,
1151 .procname = "tcp_rmem",
1152 .data = &init_net.ipv4.sysctl_tcp_rmem,
1153 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1155 .proc_handler = proc_dointvec_minmax,
1159 .procname = "tcp_comp_sack_delay_ns",
1160 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1161 .maxlen = sizeof(unsigned long),
1163 .proc_handler = proc_doulongvec_minmax,
1166 .procname = "tcp_comp_sack_nr",
1167 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1168 .maxlen = sizeof(int),
1170 .proc_handler = proc_dointvec_minmax,
1172 .extra2 = &comp_sack_nr_max,
1175 .procname = "udp_rmem_min",
1176 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1177 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1179 .proc_handler = proc_dointvec_minmax,
1183 .procname = "udp_wmem_min",
1184 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1185 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1187 .proc_handler = proc_dointvec_minmax,
1193 static __net_init int ipv4_sysctl_init_net(struct net *net)
1195 struct ctl_table *table;
1197 table = ipv4_net_table;
1198 if (!net_eq(net, &init_net)) {
1201 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1205 /* Update the variables to point into the current struct net */
1206 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1207 table[i].data += (void *)net - (void *)&init_net;
1210 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1211 if (!net->ipv4.ipv4_hdr)
1214 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1215 if (!net->ipv4.sysctl_local_reserved_ports)
1221 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1223 if (!net_eq(net, &init_net))
1229 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1231 struct ctl_table *table;
1233 kfree(net->ipv4.sysctl_local_reserved_ports);
1234 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1235 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1239 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1240 .init = ipv4_sysctl_init_net,
1241 .exit = ipv4_sysctl_exit_net,
1244 static __init int sysctl_ipv4_init(void)
1246 struct ctl_table_header *hdr;
1248 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1252 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1253 unregister_net_sysctl_table(hdr);
1260 __initcall(sysctl_ipv4_init);