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))
194 if (urange[1] < urange[0] || gid_lt(high, low)) {
195 low = make_kgid(&init_user_ns, 1);
196 high = make_kgid(&init_user_ns, 0);
198 set_ping_group_range(table, low, high);
204 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
206 size_t *lenp, loff_t *ppos)
211 net = container_of(table->data, struct net,
212 ipv4.sysctl_ip_fwd_update_priority);
213 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
214 if (write && ret == 0)
215 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
221 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
222 void __user *buffer, size_t *lenp, loff_t *ppos)
224 struct net *net = container_of(ctl->data, struct net,
225 ipv4.tcp_congestion_control);
226 char val[TCP_CA_NAME_MAX];
227 struct ctl_table tbl = {
229 .maxlen = TCP_CA_NAME_MAX,
233 tcp_get_default_congestion_control(net, val);
235 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
236 if (write && ret == 0)
237 ret = tcp_set_default_congestion_control(net, val);
241 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
243 void __user *buffer, size_t *lenp,
246 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
249 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
252 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
253 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
258 static int proc_allowed_congestion_control(struct ctl_table *ctl,
260 void __user *buffer, size_t *lenp,
263 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
266 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
270 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
271 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
272 if (write && ret == 0)
273 ret = tcp_set_allowed_congestion_control(tbl.data);
278 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
279 void __user *buffer, size_t *lenp,
282 struct net *net = container_of(table->data, struct net,
283 ipv4.sysctl_tcp_fastopen);
284 struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
285 struct tcp_fastopen_context *ctxt;
286 u32 user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
290 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
295 ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
297 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
299 memset(key, 0, sizeof(key));
302 for (i = 0; i < ARRAY_SIZE(key); i++)
303 user_key[i] = le32_to_cpu(key[i]);
305 snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
306 user_key[0], user_key[1], user_key[2], user_key[3]);
307 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
309 if (write && ret == 0) {
310 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
311 user_key + 2, user_key + 3) != 4) {
316 for (i = 0; i < ARRAY_SIZE(user_key); i++)
317 key[i] = cpu_to_le32(user_key[i]);
319 tcp_fastopen_reset_cipher(net, NULL, key,
320 TCP_FASTOPEN_KEY_LENGTH);
324 pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
325 user_key[0], user_key[1], user_key[2], user_key[3],
326 (char *)tbl.data, ret);
331 static void proc_configure_early_demux(int enabled, int protocol)
333 struct net_protocol *ipprot;
334 #if IS_ENABLED(CONFIG_IPV6)
335 struct inet6_protocol *ip6prot;
340 ipprot = rcu_dereference(inet_protos[protocol]);
342 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
345 #if IS_ENABLED(CONFIG_IPV6)
346 ip6prot = rcu_dereference(inet6_protos[protocol]);
348 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
354 static int proc_tcp_early_demux(struct ctl_table *table, int write,
355 void __user *buffer, size_t *lenp, loff_t *ppos)
359 ret = proc_dointvec(table, write, buffer, lenp, ppos);
362 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
364 proc_configure_early_demux(enabled, IPPROTO_TCP);
370 static int proc_udp_early_demux(struct ctl_table *table, int write,
371 void __user *buffer, size_t *lenp, loff_t *ppos)
375 ret = proc_dointvec(table, write, buffer, lenp, ppos);
378 int enabled = init_net.ipv4.sysctl_udp_early_demux;
380 proc_configure_early_demux(enabled, IPPROTO_UDP);
386 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
389 size_t *lenp, loff_t *ppos)
391 struct net *net = container_of(table->data, struct net,
392 ipv4.sysctl_tcp_fastopen_blackhole_timeout);
395 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
396 if (write && ret == 0)
397 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
402 static int proc_tcp_available_ulp(struct ctl_table *ctl,
404 void __user *buffer, size_t *lenp,
407 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
410 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
413 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
414 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
420 #ifdef CONFIG_IP_ROUTE_MULTIPATH
421 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
422 void __user *buffer, size_t *lenp,
425 struct net *net = container_of(table->data, struct net,
426 ipv4.sysctl_fib_multipath_hash_policy);
429 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
430 if (write && ret == 0)
431 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
437 static struct ctl_table ipv4_table[] = {
439 .procname = "tcp_max_orphans",
440 .data = &sysctl_tcp_max_orphans,
441 .maxlen = sizeof(int),
443 .proc_handler = proc_dointvec
446 .procname = "inet_peer_threshold",
447 .data = &inet_peer_threshold,
448 .maxlen = sizeof(int),
450 .proc_handler = proc_dointvec
453 .procname = "inet_peer_minttl",
454 .data = &inet_peer_minttl,
455 .maxlen = sizeof(int),
457 .proc_handler = proc_dointvec_jiffies,
460 .procname = "inet_peer_maxttl",
461 .data = &inet_peer_maxttl,
462 .maxlen = sizeof(int),
464 .proc_handler = proc_dointvec_jiffies,
467 .procname = "tcp_mem",
468 .maxlen = sizeof(sysctl_tcp_mem),
469 .data = &sysctl_tcp_mem,
471 .proc_handler = proc_doulongvec_minmax,
474 .procname = "tcp_low_latency",
475 .data = &sysctl_tcp_low_latency,
476 .maxlen = sizeof(int),
478 .proc_handler = proc_dointvec
480 #ifdef CONFIG_NETLABEL
482 .procname = "cipso_cache_enable",
483 .data = &cipso_v4_cache_enabled,
484 .maxlen = sizeof(int),
486 .proc_handler = proc_dointvec,
489 .procname = "cipso_cache_bucket_size",
490 .data = &cipso_v4_cache_bucketsize,
491 .maxlen = sizeof(int),
493 .proc_handler = proc_dointvec,
496 .procname = "cipso_rbm_optfmt",
497 .data = &cipso_v4_rbm_optfmt,
498 .maxlen = sizeof(int),
500 .proc_handler = proc_dointvec,
503 .procname = "cipso_rbm_strictvalid",
504 .data = &cipso_v4_rbm_strictvalid,
505 .maxlen = sizeof(int),
507 .proc_handler = proc_dointvec,
509 #endif /* CONFIG_NETLABEL */
511 .procname = "tcp_available_congestion_control",
512 .maxlen = TCP_CA_BUF_MAX,
514 .proc_handler = proc_tcp_available_congestion_control,
517 .procname = "tcp_allowed_congestion_control",
518 .maxlen = TCP_CA_BUF_MAX,
520 .proc_handler = proc_allowed_congestion_control,
523 .procname = "tcp_available_ulp",
524 .maxlen = TCP_ULP_BUF_MAX,
526 .proc_handler = proc_tcp_available_ulp,
529 .procname = "icmp_msgs_per_sec",
530 .data = &sysctl_icmp_msgs_per_sec,
531 .maxlen = sizeof(int),
533 .proc_handler = proc_dointvec_minmax,
537 .procname = "icmp_msgs_burst",
538 .data = &sysctl_icmp_msgs_burst,
539 .maxlen = sizeof(int),
541 .proc_handler = proc_dointvec_minmax,
545 .procname = "udp_mem",
546 .data = &sysctl_udp_mem,
547 .maxlen = sizeof(sysctl_udp_mem),
549 .proc_handler = proc_doulongvec_minmax,
554 static struct ctl_table ipv4_net_table[] = {
556 .procname = "icmp_echo_ignore_all",
557 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
558 .maxlen = sizeof(int),
560 .proc_handler = proc_dointvec
563 .procname = "icmp_echo_ignore_broadcasts",
564 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
565 .maxlen = sizeof(int),
567 .proc_handler = proc_dointvec
570 .procname = "icmp_ignore_bogus_error_responses",
571 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
572 .maxlen = sizeof(int),
574 .proc_handler = proc_dointvec
577 .procname = "icmp_errors_use_inbound_ifaddr",
578 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
579 .maxlen = sizeof(int),
581 .proc_handler = proc_dointvec
584 .procname = "icmp_ratelimit",
585 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
586 .maxlen = sizeof(int),
588 .proc_handler = proc_dointvec_ms_jiffies,
591 .procname = "icmp_ratemask",
592 .data = &init_net.ipv4.sysctl_icmp_ratemask,
593 .maxlen = sizeof(int),
595 .proc_handler = proc_dointvec
598 .procname = "ping_group_range",
599 .data = &init_net.ipv4.ping_group_range.range,
600 .maxlen = sizeof(gid_t)*2,
602 .proc_handler = ipv4_ping_group_range,
605 .procname = "tcp_ecn",
606 .data = &init_net.ipv4.sysctl_tcp_ecn,
607 .maxlen = sizeof(int),
609 .proc_handler = proc_dointvec
612 .procname = "tcp_ecn_fallback",
613 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
614 .maxlen = sizeof(int),
616 .proc_handler = proc_dointvec
619 .procname = "ip_dynaddr",
620 .data = &init_net.ipv4.sysctl_ip_dynaddr,
621 .maxlen = sizeof(int),
623 .proc_handler = proc_dointvec
626 .procname = "ip_early_demux",
627 .data = &init_net.ipv4.sysctl_ip_early_demux,
628 .maxlen = sizeof(int),
630 .proc_handler = proc_dointvec
633 .procname = "udp_early_demux",
634 .data = &init_net.ipv4.sysctl_udp_early_demux,
635 .maxlen = sizeof(int),
637 .proc_handler = proc_udp_early_demux
640 .procname = "tcp_early_demux",
641 .data = &init_net.ipv4.sysctl_tcp_early_demux,
642 .maxlen = sizeof(int),
644 .proc_handler = proc_tcp_early_demux
647 .procname = "ip_default_ttl",
648 .data = &init_net.ipv4.sysctl_ip_default_ttl,
649 .maxlen = sizeof(int),
651 .proc_handler = proc_dointvec_minmax,
652 .extra1 = &ip_ttl_min,
653 .extra2 = &ip_ttl_max,
656 .procname = "ip_local_port_range",
657 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
658 .data = &init_net.ipv4.ip_local_ports.range,
660 .proc_handler = ipv4_local_port_range,
663 .procname = "ip_local_reserved_ports",
664 .data = &init_net.ipv4.sysctl_local_reserved_ports,
667 .proc_handler = proc_do_large_bitmap,
670 .procname = "ip_no_pmtu_disc",
671 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
672 .maxlen = sizeof(int),
674 .proc_handler = proc_dointvec
677 .procname = "ip_forward_use_pmtu",
678 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
679 .maxlen = sizeof(int),
681 .proc_handler = proc_dointvec,
684 .procname = "ip_forward_update_priority",
685 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority,
686 .maxlen = sizeof(int),
688 .proc_handler = ipv4_fwd_update_priority,
693 .procname = "ip_nonlocal_bind",
694 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
695 .maxlen = sizeof(int),
697 .proc_handler = proc_dointvec
700 .procname = "fwmark_reflect",
701 .data = &init_net.ipv4.sysctl_fwmark_reflect,
702 .maxlen = sizeof(int),
704 .proc_handler = proc_dointvec,
707 .procname = "tcp_fwmark_accept",
708 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept,
709 .maxlen = sizeof(int),
711 .proc_handler = proc_dointvec,
713 #ifdef CONFIG_NET_L3_MASTER_DEV
715 .procname = "tcp_l3mdev_accept",
716 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
717 .maxlen = sizeof(int),
719 .proc_handler = proc_dointvec_minmax,
725 .procname = "tcp_mtu_probing",
726 .data = &init_net.ipv4.sysctl_tcp_mtu_probing,
727 .maxlen = sizeof(int),
729 .proc_handler = proc_dointvec,
732 .procname = "tcp_base_mss",
733 .data = &init_net.ipv4.sysctl_tcp_base_mss,
734 .maxlen = sizeof(int),
736 .proc_handler = proc_dointvec,
739 .procname = "tcp_probe_threshold",
740 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
741 .maxlen = sizeof(int),
743 .proc_handler = proc_dointvec,
746 .procname = "tcp_probe_interval",
747 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
748 .maxlen = sizeof(int),
750 .proc_handler = proc_dointvec,
753 .procname = "igmp_link_local_mcast_reports",
754 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
755 .maxlen = sizeof(int),
757 .proc_handler = proc_dointvec
760 .procname = "igmp_max_memberships",
761 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
762 .maxlen = sizeof(int),
764 .proc_handler = proc_dointvec
767 .procname = "igmp_max_msf",
768 .data = &init_net.ipv4.sysctl_igmp_max_msf,
769 .maxlen = sizeof(int),
771 .proc_handler = proc_dointvec
773 #ifdef CONFIG_IP_MULTICAST
775 .procname = "igmp_qrv",
776 .data = &init_net.ipv4.sysctl_igmp_qrv,
777 .maxlen = sizeof(int),
779 .proc_handler = proc_dointvec_minmax,
784 .procname = "tcp_congestion_control",
785 .data = &init_net.ipv4.tcp_congestion_control,
787 .maxlen = TCP_CA_NAME_MAX,
788 .proc_handler = proc_tcp_congestion_control,
791 .procname = "tcp_keepalive_time",
792 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
793 .maxlen = sizeof(int),
795 .proc_handler = proc_dointvec_jiffies,
798 .procname = "tcp_keepalive_probes",
799 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
800 .maxlen = sizeof(int),
802 .proc_handler = proc_dointvec
805 .procname = "tcp_keepalive_intvl",
806 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
807 .maxlen = sizeof(int),
809 .proc_handler = proc_dointvec_jiffies,
812 .procname = "tcp_syn_retries",
813 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
814 .maxlen = sizeof(int),
816 .proc_handler = proc_dointvec_minmax,
817 .extra1 = &tcp_syn_retries_min,
818 .extra2 = &tcp_syn_retries_max
821 .procname = "tcp_synack_retries",
822 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
823 .maxlen = sizeof(int),
825 .proc_handler = proc_dointvec
827 #ifdef CONFIG_SYN_COOKIES
829 .procname = "tcp_syncookies",
830 .data = &init_net.ipv4.sysctl_tcp_syncookies,
831 .maxlen = sizeof(int),
833 .proc_handler = proc_dointvec
837 .procname = "tcp_reordering",
838 .data = &init_net.ipv4.sysctl_tcp_reordering,
839 .maxlen = sizeof(int),
841 .proc_handler = proc_dointvec
844 .procname = "tcp_retries1",
845 .data = &init_net.ipv4.sysctl_tcp_retries1,
846 .maxlen = sizeof(int),
848 .proc_handler = proc_dointvec_minmax,
849 .extra2 = &tcp_retr1_max
852 .procname = "tcp_retries2",
853 .data = &init_net.ipv4.sysctl_tcp_retries2,
854 .maxlen = sizeof(int),
856 .proc_handler = proc_dointvec
859 .procname = "tcp_orphan_retries",
860 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
861 .maxlen = sizeof(int),
863 .proc_handler = proc_dointvec
866 .procname = "tcp_fin_timeout",
867 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
868 .maxlen = sizeof(int),
870 .proc_handler = proc_dointvec_jiffies,
873 .procname = "tcp_notsent_lowat",
874 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
875 .maxlen = sizeof(unsigned int),
877 .proc_handler = proc_douintvec,
880 .procname = "tcp_tw_reuse",
881 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
882 .maxlen = sizeof(int),
884 .proc_handler = proc_dointvec_minmax,
889 .procname = "tcp_max_tw_buckets",
890 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
891 .maxlen = sizeof(int),
893 .proc_handler = proc_dointvec
896 .procname = "tcp_max_syn_backlog",
897 .data = &init_net.ipv4.sysctl_max_syn_backlog,
898 .maxlen = sizeof(int),
900 .proc_handler = proc_dointvec
903 .procname = "tcp_fastopen",
904 .data = &init_net.ipv4.sysctl_tcp_fastopen,
905 .maxlen = sizeof(int),
907 .proc_handler = proc_dointvec,
910 .procname = "tcp_fastopen_key",
912 .data = &init_net.ipv4.sysctl_tcp_fastopen,
913 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
914 .proc_handler = proc_tcp_fastopen_key,
917 .procname = "tcp_fastopen_blackhole_timeout_sec",
918 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
919 .maxlen = sizeof(int),
921 .proc_handler = proc_tfo_blackhole_detect_timeout,
924 #ifdef CONFIG_IP_ROUTE_MULTIPATH
926 .procname = "fib_multipath_use_neigh",
927 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
928 .maxlen = sizeof(int),
930 .proc_handler = proc_dointvec_minmax,
935 .procname = "fib_multipath_hash_policy",
936 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
937 .maxlen = sizeof(int),
939 .proc_handler = proc_fib_multipath_hash_policy,
945 .procname = "ip_unprivileged_port_start",
946 .maxlen = sizeof(int),
947 .data = &init_net.ipv4.sysctl_ip_prot_sock,
949 .proc_handler = ipv4_privileged_ports,
951 #ifdef CONFIG_NET_L3_MASTER_DEV
953 .procname = "udp_l3mdev_accept",
954 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
955 .maxlen = sizeof(int),
957 .proc_handler = proc_dointvec_minmax,
963 .procname = "tcp_sack",
964 .data = &init_net.ipv4.sysctl_tcp_sack,
965 .maxlen = sizeof(int),
967 .proc_handler = proc_dointvec
970 .procname = "tcp_window_scaling",
971 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
972 .maxlen = sizeof(int),
974 .proc_handler = proc_dointvec
977 .procname = "tcp_timestamps",
978 .data = &init_net.ipv4.sysctl_tcp_timestamps,
979 .maxlen = sizeof(int),
981 .proc_handler = proc_dointvec
984 .procname = "tcp_early_retrans",
985 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
986 .maxlen = sizeof(int),
988 .proc_handler = proc_dointvec_minmax,
993 .procname = "tcp_recovery",
994 .data = &init_net.ipv4.sysctl_tcp_recovery,
995 .maxlen = sizeof(int),
997 .proc_handler = proc_dointvec,
1000 .procname = "tcp_thin_linear_timeouts",
1001 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1002 .maxlen = sizeof(int),
1004 .proc_handler = proc_dointvec
1007 .procname = "tcp_slow_start_after_idle",
1008 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1009 .maxlen = sizeof(int),
1011 .proc_handler = proc_dointvec
1014 .procname = "tcp_retrans_collapse",
1015 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1016 .maxlen = sizeof(int),
1018 .proc_handler = proc_dointvec
1021 .procname = "tcp_stdurg",
1022 .data = &init_net.ipv4.sysctl_tcp_stdurg,
1023 .maxlen = sizeof(int),
1025 .proc_handler = proc_dointvec
1028 .procname = "tcp_rfc1337",
1029 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
1030 .maxlen = sizeof(int),
1032 .proc_handler = proc_dointvec
1035 .procname = "tcp_abort_on_overflow",
1036 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1037 .maxlen = sizeof(int),
1039 .proc_handler = proc_dointvec
1042 .procname = "tcp_fack",
1043 .data = &init_net.ipv4.sysctl_tcp_fack,
1044 .maxlen = sizeof(int),
1046 .proc_handler = proc_dointvec
1049 .procname = "tcp_max_reordering",
1050 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1051 .maxlen = sizeof(int),
1053 .proc_handler = proc_dointvec
1056 .procname = "tcp_dsack",
1057 .data = &init_net.ipv4.sysctl_tcp_dsack,
1058 .maxlen = sizeof(int),
1060 .proc_handler = proc_dointvec
1063 .procname = "tcp_app_win",
1064 .data = &init_net.ipv4.sysctl_tcp_app_win,
1065 .maxlen = sizeof(int),
1067 .proc_handler = proc_dointvec
1070 .procname = "tcp_adv_win_scale",
1071 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1072 .maxlen = sizeof(int),
1074 .proc_handler = proc_dointvec_minmax,
1075 .extra1 = &tcp_adv_win_scale_min,
1076 .extra2 = &tcp_adv_win_scale_max,
1079 .procname = "tcp_frto",
1080 .data = &init_net.ipv4.sysctl_tcp_frto,
1081 .maxlen = sizeof(int),
1083 .proc_handler = proc_dointvec
1086 .procname = "tcp_no_metrics_save",
1087 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1088 .maxlen = sizeof(int),
1090 .proc_handler = proc_dointvec,
1093 .procname = "tcp_moderate_rcvbuf",
1094 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1095 .maxlen = sizeof(int),
1097 .proc_handler = proc_dointvec,
1100 .procname = "tcp_tso_win_divisor",
1101 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1102 .maxlen = sizeof(int),
1104 .proc_handler = proc_dointvec,
1107 .procname = "tcp_workaround_signed_windows",
1108 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1109 .maxlen = sizeof(int),
1111 .proc_handler = proc_dointvec
1114 .procname = "tcp_limit_output_bytes",
1115 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1116 .maxlen = sizeof(int),
1118 .proc_handler = proc_dointvec
1121 .procname = "tcp_challenge_ack_limit",
1122 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1123 .maxlen = sizeof(int),
1125 .proc_handler = proc_dointvec
1128 .procname = "tcp_min_tso_segs",
1129 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1130 .maxlen = sizeof(int),
1132 .proc_handler = proc_dointvec_minmax,
1134 .extra2 = &gso_max_segs,
1137 .procname = "tcp_min_rtt_wlen",
1138 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1139 .maxlen = sizeof(int),
1141 .proc_handler = proc_dointvec
1144 .procname = "tcp_autocorking",
1145 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1146 .maxlen = sizeof(int),
1148 .proc_handler = proc_dointvec_minmax,
1153 .procname = "tcp_invalid_ratelimit",
1154 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1155 .maxlen = sizeof(int),
1157 .proc_handler = proc_dointvec_ms_jiffies,
1160 .procname = "tcp_pacing_ss_ratio",
1161 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1162 .maxlen = sizeof(int),
1164 .proc_handler = proc_dointvec_minmax,
1166 .extra2 = &thousand,
1169 .procname = "tcp_pacing_ca_ratio",
1170 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1171 .maxlen = sizeof(int),
1173 .proc_handler = proc_dointvec_minmax,
1175 .extra2 = &thousand,
1178 .procname = "tcp_wmem",
1179 .data = &init_net.ipv4.sysctl_tcp_wmem,
1180 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1182 .proc_handler = proc_dointvec_minmax,
1186 .procname = "tcp_rmem",
1187 .data = &init_net.ipv4.sysctl_tcp_rmem,
1188 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1190 .proc_handler = proc_dointvec_minmax,
1194 .procname = "tcp_comp_sack_delay_ns",
1195 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1196 .maxlen = sizeof(unsigned long),
1198 .proc_handler = proc_doulongvec_minmax,
1201 .procname = "tcp_comp_sack_nr",
1202 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1203 .maxlen = sizeof(int),
1205 .proc_handler = proc_dointvec_minmax,
1207 .extra2 = &comp_sack_nr_max,
1210 .procname = "udp_rmem_min",
1211 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1212 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1214 .proc_handler = proc_dointvec_minmax,
1218 .procname = "udp_wmem_min",
1219 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1220 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1222 .proc_handler = proc_dointvec_minmax,
1228 static __net_init int ipv4_sysctl_init_net(struct net *net)
1230 struct ctl_table *table;
1232 table = ipv4_net_table;
1233 if (!net_eq(net, &init_net)) {
1236 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1240 /* Update the variables to point into the current struct net */
1241 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1242 table[i].data += (void *)net - (void *)&init_net;
1245 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1246 if (!net->ipv4.ipv4_hdr)
1249 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1250 if (!net->ipv4.sysctl_local_reserved_ports)
1256 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1258 if (!net_eq(net, &init_net))
1264 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1266 struct ctl_table *table;
1268 kfree(net->ipv4.sysctl_local_reserved_ports);
1269 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1270 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1274 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1275 .init = ipv4_sysctl_init_net,
1276 .exit = ipv4_sysctl_exit_net,
1279 static __init int sysctl_ipv4_init(void)
1281 struct ctl_table_header *hdr;
1283 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1287 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1288 unregister_net_sysctl_table(hdr);
1295 __initcall(sysctl_ipv4_init);