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 tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
43 static int tcp_min_snd_mss_max = 65535;
44 static int ip_privileged_port_min;
45 static int ip_privileged_port_max = 65535;
46 static int ip_ttl_min = 1;
47 static int ip_ttl_max = 255;
48 static int tcp_syn_retries_min = 1;
49 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
50 static int ip_ping_group_range_min[] = { 0, 0 };
51 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
52 static int comp_sack_nr_max = 255;
53 static u32 u32_max_div_HZ = UINT_MAX / HZ;
54 static int one_day_secs = 24 * 3600;
57 static int sysctl_tcp_low_latency __read_mostly;
59 /* Update system visible IP port range */
60 static void set_local_port_range(struct net *net, int range[2])
62 bool same_parity = !((range[0] ^ range[1]) & 1);
64 write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
65 if (same_parity && !net->ipv4.ip_local_ports.warned) {
66 net->ipv4.ip_local_ports.warned = true;
67 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
69 net->ipv4.ip_local_ports.range[0] = range[0];
70 net->ipv4.ip_local_ports.range[1] = range[1];
71 write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
74 /* Validate changes from /proc interface. */
75 static int ipv4_local_port_range(struct ctl_table *table, int write,
77 size_t *lenp, loff_t *ppos)
80 container_of(table->data, struct net, ipv4.ip_local_ports.range);
83 struct ctl_table tmp = {
85 .maxlen = sizeof(range),
87 .extra1 = &ip_local_port_range_min,
88 .extra2 = &ip_local_port_range_max,
91 inet_get_local_port_range(net, &range[0], &range[1]);
93 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
95 if (write && ret == 0) {
96 /* Ensure that the upper limit is not smaller than the lower,
97 * and that the lower does not encroach upon the privileged
100 if ((range[1] < range[0]) ||
101 (range[0] < net->ipv4.sysctl_ip_prot_sock))
104 set_local_port_range(net, range);
110 /* Validate changes from /proc interface. */
111 static int ipv4_privileged_ports(struct ctl_table *table, int write,
112 void __user *buffer, size_t *lenp, loff_t *ppos)
114 struct net *net = container_of(table->data, struct net,
115 ipv4.sysctl_ip_prot_sock);
119 struct ctl_table tmp = {
121 .maxlen = sizeof(pports),
123 .extra1 = &ip_privileged_port_min,
124 .extra2 = &ip_privileged_port_max,
127 pports = net->ipv4.sysctl_ip_prot_sock;
129 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
131 if (write && ret == 0) {
132 inet_get_local_port_range(net, &range[0], &range[1]);
133 /* Ensure that the local port range doesn't overlap with the
134 * privileged port range.
136 if (range[0] < pports)
139 net->ipv4.sysctl_ip_prot_sock = pports;
145 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
147 kgid_t *data = table->data;
149 container_of(table->data, struct net, ipv4.ping_group_range.range);
152 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
156 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
159 /* Update system visible IP port range */
160 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
162 kgid_t *data = table->data;
164 container_of(table->data, struct net, ipv4.ping_group_range.range);
165 write_seqlock(&net->ipv4.ping_group_range.lock);
168 write_sequnlock(&net->ipv4.ping_group_range.lock);
171 /* Validate changes from /proc interface. */
172 static int ipv4_ping_group_range(struct ctl_table *table, int write,
174 size_t *lenp, loff_t *ppos)
176 struct user_namespace *user_ns = current_user_ns();
180 struct ctl_table tmp = {
182 .maxlen = sizeof(urange),
184 .extra1 = &ip_ping_group_range_min,
185 .extra2 = &ip_ping_group_range_max,
188 inet_get_ping_group_range_table(table, &low, &high);
189 urange[0] = from_kgid_munged(user_ns, low);
190 urange[1] = from_kgid_munged(user_ns, high);
191 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
193 if (write && ret == 0) {
194 low = make_kgid(user_ns, urange[0]);
195 high = make_kgid(user_ns, urange[1]);
196 if (!gid_valid(low) || !gid_valid(high))
198 if (urange[1] < urange[0] || gid_lt(high, low)) {
199 low = make_kgid(&init_user_ns, 1);
200 high = make_kgid(&init_user_ns, 0);
202 set_ping_group_range(table, low, high);
208 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
210 size_t *lenp, loff_t *ppos)
215 net = container_of(table->data, struct net,
216 ipv4.sysctl_ip_fwd_update_priority);
217 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
218 if (write && ret == 0)
219 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
225 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
226 void __user *buffer, size_t *lenp, loff_t *ppos)
228 struct net *net = container_of(ctl->data, struct net,
229 ipv4.tcp_congestion_control);
230 char val[TCP_CA_NAME_MAX];
231 struct ctl_table tbl = {
233 .maxlen = TCP_CA_NAME_MAX,
237 tcp_get_default_congestion_control(net, val);
239 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
240 if (write && ret == 0)
241 ret = tcp_set_default_congestion_control(net, val);
245 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
247 void __user *buffer, size_t *lenp,
250 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
253 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
256 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
257 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
262 static int proc_allowed_congestion_control(struct ctl_table *ctl,
264 void __user *buffer, size_t *lenp,
267 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
270 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
274 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
275 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
276 if (write && ret == 0)
277 ret = tcp_set_allowed_congestion_control(tbl.data);
282 static int sscanf_key(char *buf, __le32 *key)
287 if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
288 user_key + 2, user_key + 3) != 4) {
291 for (i = 0; i < ARRAY_SIZE(user_key); i++)
292 key[i] = cpu_to_le32(user_key[i]);
294 pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
295 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
300 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
301 void __user *buffer, size_t *lenp,
304 struct net *net = container_of(table->data, struct net,
305 ipv4.sysctl_tcp_fastopen);
306 /* maxlen to print the list of keys in hex (*2), with dashes
307 * separating doublewords and a comma in between keys.
309 struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
310 2 * TCP_FASTOPEN_KEY_MAX) +
311 (TCP_FASTOPEN_KEY_MAX * 5)) };
312 struct tcp_fastopen_context *ctx;
313 u32 user_key[TCP_FASTOPEN_KEY_MAX * 4];
314 __le32 key[TCP_FASTOPEN_KEY_MAX * 4];
316 int ret, i = 0, off = 0, n_keys = 0;
318 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
323 ctx = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
325 n_keys = tcp_fastopen_context_len(ctx);
326 memcpy(&key[0], &ctx->key[0], TCP_FASTOPEN_KEY_LENGTH * n_keys);
331 memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
335 for (i = 0; i < n_keys * 4; i++)
336 user_key[i] = le32_to_cpu(key[i]);
338 for (i = 0; i < n_keys; i++) {
339 off += snprintf(tbl.data + off, tbl.maxlen - off,
340 "%08x-%08x-%08x-%08x",
344 user_key[i * 4 + 3]);
346 off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
349 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
351 if (write && ret == 0) {
352 backup_data = strchr(tbl.data, ',');
357 if (sscanf_key(tbl.data, key)) {
362 if (sscanf_key(backup_data, key + 4)) {
367 tcp_fastopen_reset_cipher(net, NULL, key,
368 backup_data ? key + 4 : NULL);
376 static void proc_configure_early_demux(int enabled, int protocol)
378 struct net_protocol *ipprot;
379 #if IS_ENABLED(CONFIG_IPV6)
380 struct inet6_protocol *ip6prot;
385 ipprot = rcu_dereference(inet_protos[protocol]);
387 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
390 #if IS_ENABLED(CONFIG_IPV6)
391 ip6prot = rcu_dereference(inet6_protos[protocol]);
393 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
399 static int proc_tcp_early_demux(struct ctl_table *table, int write,
400 void __user *buffer, size_t *lenp, loff_t *ppos)
404 ret = proc_dointvec(table, write, buffer, lenp, ppos);
407 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
409 proc_configure_early_demux(enabled, IPPROTO_TCP);
415 static int proc_udp_early_demux(struct ctl_table *table, int write,
416 void __user *buffer, size_t *lenp, loff_t *ppos)
420 ret = proc_dointvec(table, write, buffer, lenp, ppos);
423 int enabled = init_net.ipv4.sysctl_udp_early_demux;
425 proc_configure_early_demux(enabled, IPPROTO_UDP);
431 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
434 size_t *lenp, loff_t *ppos)
436 struct net *net = container_of(table->data, struct net,
437 ipv4.sysctl_tcp_fastopen_blackhole_timeout);
440 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
441 if (write && ret == 0)
442 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
447 static int proc_tcp_available_ulp(struct ctl_table *ctl,
449 void __user *buffer, size_t *lenp,
452 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
455 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
458 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
459 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
465 #ifdef CONFIG_IP_ROUTE_MULTIPATH
466 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
467 void __user *buffer, size_t *lenp,
470 struct net *net = container_of(table->data, struct net,
471 ipv4.sysctl_fib_multipath_hash_policy);
474 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
475 if (write && ret == 0)
476 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
482 static struct ctl_table ipv4_table[] = {
484 .procname = "tcp_max_orphans",
485 .data = &sysctl_tcp_max_orphans,
486 .maxlen = sizeof(int),
488 .proc_handler = proc_dointvec
491 .procname = "inet_peer_threshold",
492 .data = &inet_peer_threshold,
493 .maxlen = sizeof(int),
495 .proc_handler = proc_dointvec
498 .procname = "inet_peer_minttl",
499 .data = &inet_peer_minttl,
500 .maxlen = sizeof(int),
502 .proc_handler = proc_dointvec_jiffies,
505 .procname = "inet_peer_maxttl",
506 .data = &inet_peer_maxttl,
507 .maxlen = sizeof(int),
509 .proc_handler = proc_dointvec_jiffies,
512 .procname = "tcp_mem",
513 .maxlen = sizeof(sysctl_tcp_mem),
514 .data = &sysctl_tcp_mem,
516 .proc_handler = proc_doulongvec_minmax,
519 .procname = "tcp_low_latency",
520 .data = &sysctl_tcp_low_latency,
521 .maxlen = sizeof(int),
523 .proc_handler = proc_dointvec
525 #ifdef CONFIG_NETLABEL
527 .procname = "cipso_cache_enable",
528 .data = &cipso_v4_cache_enabled,
529 .maxlen = sizeof(int),
531 .proc_handler = proc_dointvec,
534 .procname = "cipso_cache_bucket_size",
535 .data = &cipso_v4_cache_bucketsize,
536 .maxlen = sizeof(int),
538 .proc_handler = proc_dointvec,
541 .procname = "cipso_rbm_optfmt",
542 .data = &cipso_v4_rbm_optfmt,
543 .maxlen = sizeof(int),
545 .proc_handler = proc_dointvec,
548 .procname = "cipso_rbm_strictvalid",
549 .data = &cipso_v4_rbm_strictvalid,
550 .maxlen = sizeof(int),
552 .proc_handler = proc_dointvec,
554 #endif /* CONFIG_NETLABEL */
556 .procname = "tcp_available_congestion_control",
557 .maxlen = TCP_CA_BUF_MAX,
559 .proc_handler = proc_tcp_available_congestion_control,
562 .procname = "tcp_allowed_congestion_control",
563 .maxlen = TCP_CA_BUF_MAX,
565 .proc_handler = proc_allowed_congestion_control,
568 .procname = "tcp_available_ulp",
569 .maxlen = TCP_ULP_BUF_MAX,
571 .proc_handler = proc_tcp_available_ulp,
574 .procname = "icmp_msgs_per_sec",
575 .data = &sysctl_icmp_msgs_per_sec,
576 .maxlen = sizeof(int),
578 .proc_handler = proc_dointvec_minmax,
582 .procname = "icmp_msgs_burst",
583 .data = &sysctl_icmp_msgs_burst,
584 .maxlen = sizeof(int),
586 .proc_handler = proc_dointvec_minmax,
590 .procname = "udp_mem",
591 .data = &sysctl_udp_mem,
592 .maxlen = sizeof(sysctl_udp_mem),
594 .proc_handler = proc_doulongvec_minmax,
597 .procname = "fib_sync_mem",
598 .data = &sysctl_fib_sync_mem,
599 .maxlen = sizeof(sysctl_fib_sync_mem),
601 .proc_handler = proc_douintvec_minmax,
602 .extra1 = &sysctl_fib_sync_mem_min,
603 .extra2 = &sysctl_fib_sync_mem_max,
606 .procname = "tcp_rx_skb_cache",
607 .data = &tcp_rx_skb_cache_key.key,
609 .proc_handler = proc_do_static_key,
612 .procname = "tcp_tx_skb_cache",
613 .data = &tcp_tx_skb_cache_key.key,
615 .proc_handler = proc_do_static_key,
620 static struct ctl_table ipv4_net_table[] = {
622 .procname = "icmp_echo_ignore_all",
623 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
624 .maxlen = sizeof(int),
626 .proc_handler = proc_dointvec
629 .procname = "icmp_echo_ignore_broadcasts",
630 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
631 .maxlen = sizeof(int),
633 .proc_handler = proc_dointvec
636 .procname = "icmp_ignore_bogus_error_responses",
637 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
638 .maxlen = sizeof(int),
640 .proc_handler = proc_dointvec
643 .procname = "icmp_errors_use_inbound_ifaddr",
644 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
645 .maxlen = sizeof(int),
647 .proc_handler = proc_dointvec
650 .procname = "icmp_ratelimit",
651 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
652 .maxlen = sizeof(int),
654 .proc_handler = proc_dointvec_ms_jiffies,
657 .procname = "icmp_ratemask",
658 .data = &init_net.ipv4.sysctl_icmp_ratemask,
659 .maxlen = sizeof(int),
661 .proc_handler = proc_dointvec
664 .procname = "ping_group_range",
665 .data = &init_net.ipv4.ping_group_range.range,
666 .maxlen = sizeof(gid_t)*2,
668 .proc_handler = ipv4_ping_group_range,
670 #ifdef CONFIG_NET_L3_MASTER_DEV
672 .procname = "raw_l3mdev_accept",
673 .data = &init_net.ipv4.sysctl_raw_l3mdev_accept,
674 .maxlen = sizeof(int),
676 .proc_handler = proc_dointvec_minmax,
682 .procname = "tcp_ecn",
683 .data = &init_net.ipv4.sysctl_tcp_ecn,
684 .maxlen = sizeof(int),
686 .proc_handler = proc_dointvec
689 .procname = "tcp_ecn_fallback",
690 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
691 .maxlen = sizeof(int),
693 .proc_handler = proc_dointvec
696 .procname = "ip_dynaddr",
697 .data = &init_net.ipv4.sysctl_ip_dynaddr,
698 .maxlen = sizeof(int),
700 .proc_handler = proc_dointvec
703 .procname = "ip_early_demux",
704 .data = &init_net.ipv4.sysctl_ip_early_demux,
705 .maxlen = sizeof(int),
707 .proc_handler = proc_dointvec
710 .procname = "udp_early_demux",
711 .data = &init_net.ipv4.sysctl_udp_early_demux,
712 .maxlen = sizeof(int),
714 .proc_handler = proc_udp_early_demux
717 .procname = "tcp_early_demux",
718 .data = &init_net.ipv4.sysctl_tcp_early_demux,
719 .maxlen = sizeof(int),
721 .proc_handler = proc_tcp_early_demux
724 .procname = "ip_default_ttl",
725 .data = &init_net.ipv4.sysctl_ip_default_ttl,
726 .maxlen = sizeof(int),
728 .proc_handler = proc_dointvec_minmax,
729 .extra1 = &ip_ttl_min,
730 .extra2 = &ip_ttl_max,
733 .procname = "ip_local_port_range",
734 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
735 .data = &init_net.ipv4.ip_local_ports.range,
737 .proc_handler = ipv4_local_port_range,
740 .procname = "ip_local_reserved_ports",
741 .data = &init_net.ipv4.sysctl_local_reserved_ports,
744 .proc_handler = proc_do_large_bitmap,
747 .procname = "ip_no_pmtu_disc",
748 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
749 .maxlen = sizeof(int),
751 .proc_handler = proc_dointvec
754 .procname = "ip_forward_use_pmtu",
755 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
756 .maxlen = sizeof(int),
758 .proc_handler = proc_dointvec,
761 .procname = "ip_forward_update_priority",
762 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority,
763 .maxlen = sizeof(int),
765 .proc_handler = ipv4_fwd_update_priority,
770 .procname = "ip_nonlocal_bind",
771 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
772 .maxlen = sizeof(int),
774 .proc_handler = proc_dointvec
777 .procname = "fwmark_reflect",
778 .data = &init_net.ipv4.sysctl_fwmark_reflect,
779 .maxlen = sizeof(int),
781 .proc_handler = proc_dointvec,
784 .procname = "tcp_fwmark_accept",
785 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept,
786 .maxlen = sizeof(int),
788 .proc_handler = proc_dointvec,
790 #ifdef CONFIG_NET_L3_MASTER_DEV
792 .procname = "tcp_l3mdev_accept",
793 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
794 .maxlen = sizeof(int),
796 .proc_handler = proc_dointvec_minmax,
802 .procname = "tcp_mtu_probing",
803 .data = &init_net.ipv4.sysctl_tcp_mtu_probing,
804 .maxlen = sizeof(int),
806 .proc_handler = proc_dointvec,
809 .procname = "tcp_base_mss",
810 .data = &init_net.ipv4.sysctl_tcp_base_mss,
811 .maxlen = sizeof(int),
813 .proc_handler = proc_dointvec,
816 .procname = "tcp_min_snd_mss",
817 .data = &init_net.ipv4.sysctl_tcp_min_snd_mss,
818 .maxlen = sizeof(int),
820 .proc_handler = proc_dointvec_minmax,
821 .extra1 = &tcp_min_snd_mss_min,
822 .extra2 = &tcp_min_snd_mss_max,
825 .procname = "tcp_probe_threshold",
826 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
827 .maxlen = sizeof(int),
829 .proc_handler = proc_dointvec,
832 .procname = "tcp_probe_interval",
833 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
834 .maxlen = sizeof(u32),
836 .proc_handler = proc_douintvec_minmax,
837 .extra2 = &u32_max_div_HZ,
840 .procname = "igmp_link_local_mcast_reports",
841 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
842 .maxlen = sizeof(int),
844 .proc_handler = proc_dointvec
847 .procname = "igmp_max_memberships",
848 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
849 .maxlen = sizeof(int),
851 .proc_handler = proc_dointvec
854 .procname = "igmp_max_msf",
855 .data = &init_net.ipv4.sysctl_igmp_max_msf,
856 .maxlen = sizeof(int),
858 .proc_handler = proc_dointvec
860 #ifdef CONFIG_IP_MULTICAST
862 .procname = "igmp_qrv",
863 .data = &init_net.ipv4.sysctl_igmp_qrv,
864 .maxlen = sizeof(int),
866 .proc_handler = proc_dointvec_minmax,
871 .procname = "tcp_congestion_control",
872 .data = &init_net.ipv4.tcp_congestion_control,
874 .maxlen = TCP_CA_NAME_MAX,
875 .proc_handler = proc_tcp_congestion_control,
878 .procname = "tcp_keepalive_time",
879 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
880 .maxlen = sizeof(int),
882 .proc_handler = proc_dointvec_jiffies,
885 .procname = "tcp_keepalive_probes",
886 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
887 .maxlen = sizeof(int),
889 .proc_handler = proc_dointvec
892 .procname = "tcp_keepalive_intvl",
893 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
894 .maxlen = sizeof(int),
896 .proc_handler = proc_dointvec_jiffies,
899 .procname = "tcp_syn_retries",
900 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
901 .maxlen = sizeof(int),
903 .proc_handler = proc_dointvec_minmax,
904 .extra1 = &tcp_syn_retries_min,
905 .extra2 = &tcp_syn_retries_max
908 .procname = "tcp_synack_retries",
909 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
910 .maxlen = sizeof(int),
912 .proc_handler = proc_dointvec
914 #ifdef CONFIG_SYN_COOKIES
916 .procname = "tcp_syncookies",
917 .data = &init_net.ipv4.sysctl_tcp_syncookies,
918 .maxlen = sizeof(int),
920 .proc_handler = proc_dointvec
924 .procname = "tcp_reordering",
925 .data = &init_net.ipv4.sysctl_tcp_reordering,
926 .maxlen = sizeof(int),
928 .proc_handler = proc_dointvec
931 .procname = "tcp_retries1",
932 .data = &init_net.ipv4.sysctl_tcp_retries1,
933 .maxlen = sizeof(int),
935 .proc_handler = proc_dointvec_minmax,
936 .extra2 = &tcp_retr1_max
939 .procname = "tcp_retries2",
940 .data = &init_net.ipv4.sysctl_tcp_retries2,
941 .maxlen = sizeof(int),
943 .proc_handler = proc_dointvec
946 .procname = "tcp_orphan_retries",
947 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
948 .maxlen = sizeof(int),
950 .proc_handler = proc_dointvec
953 .procname = "tcp_fin_timeout",
954 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
955 .maxlen = sizeof(int),
957 .proc_handler = proc_dointvec_jiffies,
960 .procname = "tcp_notsent_lowat",
961 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
962 .maxlen = sizeof(unsigned int),
964 .proc_handler = proc_douintvec,
967 .procname = "tcp_tw_reuse",
968 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
969 .maxlen = sizeof(int),
971 .proc_handler = proc_dointvec_minmax,
976 .procname = "tcp_max_tw_buckets",
977 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
978 .maxlen = sizeof(int),
980 .proc_handler = proc_dointvec
983 .procname = "tcp_max_syn_backlog",
984 .data = &init_net.ipv4.sysctl_max_syn_backlog,
985 .maxlen = sizeof(int),
987 .proc_handler = proc_dointvec
990 .procname = "tcp_fastopen",
991 .data = &init_net.ipv4.sysctl_tcp_fastopen,
992 .maxlen = sizeof(int),
994 .proc_handler = proc_dointvec,
997 .procname = "tcp_fastopen_key",
999 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1000 /* maxlen to print the list of keys in hex (*2), with dashes
1001 * separating doublewords and a comma in between keys.
1003 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
1004 2 * TCP_FASTOPEN_KEY_MAX) +
1005 (TCP_FASTOPEN_KEY_MAX * 5)),
1006 .proc_handler = proc_tcp_fastopen_key,
1009 .procname = "tcp_fastopen_blackhole_timeout_sec",
1010 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1011 .maxlen = sizeof(int),
1013 .proc_handler = proc_tfo_blackhole_detect_timeout,
1016 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1018 .procname = "fib_multipath_use_neigh",
1019 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1020 .maxlen = sizeof(int),
1022 .proc_handler = proc_dointvec_minmax,
1027 .procname = "fib_multipath_hash_policy",
1028 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1029 .maxlen = sizeof(int),
1031 .proc_handler = proc_fib_multipath_hash_policy,
1037 .procname = "ip_unprivileged_port_start",
1038 .maxlen = sizeof(int),
1039 .data = &init_net.ipv4.sysctl_ip_prot_sock,
1041 .proc_handler = ipv4_privileged_ports,
1043 #ifdef CONFIG_NET_L3_MASTER_DEV
1045 .procname = "udp_l3mdev_accept",
1046 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1047 .maxlen = sizeof(int),
1049 .proc_handler = proc_dointvec_minmax,
1055 .procname = "tcp_sack",
1056 .data = &init_net.ipv4.sysctl_tcp_sack,
1057 .maxlen = sizeof(int),
1059 .proc_handler = proc_dointvec
1062 .procname = "tcp_window_scaling",
1063 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
1064 .maxlen = sizeof(int),
1066 .proc_handler = proc_dointvec
1069 .procname = "tcp_timestamps",
1070 .data = &init_net.ipv4.sysctl_tcp_timestamps,
1071 .maxlen = sizeof(int),
1073 .proc_handler = proc_dointvec
1076 .procname = "tcp_early_retrans",
1077 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
1078 .maxlen = sizeof(int),
1080 .proc_handler = proc_dointvec_minmax,
1085 .procname = "tcp_recovery",
1086 .data = &init_net.ipv4.sysctl_tcp_recovery,
1087 .maxlen = sizeof(int),
1089 .proc_handler = proc_dointvec,
1092 .procname = "tcp_thin_linear_timeouts",
1093 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1094 .maxlen = sizeof(int),
1096 .proc_handler = proc_dointvec
1099 .procname = "tcp_slow_start_after_idle",
1100 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1101 .maxlen = sizeof(int),
1103 .proc_handler = proc_dointvec
1106 .procname = "tcp_retrans_collapse",
1107 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1108 .maxlen = sizeof(int),
1110 .proc_handler = proc_dointvec
1113 .procname = "tcp_stdurg",
1114 .data = &init_net.ipv4.sysctl_tcp_stdurg,
1115 .maxlen = sizeof(int),
1117 .proc_handler = proc_dointvec
1120 .procname = "tcp_rfc1337",
1121 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
1122 .maxlen = sizeof(int),
1124 .proc_handler = proc_dointvec
1127 .procname = "tcp_abort_on_overflow",
1128 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1129 .maxlen = sizeof(int),
1131 .proc_handler = proc_dointvec
1134 .procname = "tcp_fack",
1135 .data = &init_net.ipv4.sysctl_tcp_fack,
1136 .maxlen = sizeof(int),
1138 .proc_handler = proc_dointvec
1141 .procname = "tcp_max_reordering",
1142 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1143 .maxlen = sizeof(int),
1145 .proc_handler = proc_dointvec
1148 .procname = "tcp_dsack",
1149 .data = &init_net.ipv4.sysctl_tcp_dsack,
1150 .maxlen = sizeof(int),
1152 .proc_handler = proc_dointvec
1155 .procname = "tcp_app_win",
1156 .data = &init_net.ipv4.sysctl_tcp_app_win,
1157 .maxlen = sizeof(int),
1159 .proc_handler = proc_dointvec
1162 .procname = "tcp_adv_win_scale",
1163 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1164 .maxlen = sizeof(int),
1166 .proc_handler = proc_dointvec_minmax,
1167 .extra1 = &tcp_adv_win_scale_min,
1168 .extra2 = &tcp_adv_win_scale_max,
1171 .procname = "tcp_frto",
1172 .data = &init_net.ipv4.sysctl_tcp_frto,
1173 .maxlen = sizeof(int),
1175 .proc_handler = proc_dointvec
1178 .procname = "tcp_no_metrics_save",
1179 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1180 .maxlen = sizeof(int),
1182 .proc_handler = proc_dointvec,
1185 .procname = "tcp_moderate_rcvbuf",
1186 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1187 .maxlen = sizeof(int),
1189 .proc_handler = proc_dointvec,
1192 .procname = "tcp_tso_win_divisor",
1193 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1194 .maxlen = sizeof(int),
1196 .proc_handler = proc_dointvec,
1199 .procname = "tcp_workaround_signed_windows",
1200 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1201 .maxlen = sizeof(int),
1203 .proc_handler = proc_dointvec
1206 .procname = "tcp_limit_output_bytes",
1207 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1208 .maxlen = sizeof(int),
1210 .proc_handler = proc_dointvec
1213 .procname = "tcp_challenge_ack_limit",
1214 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1215 .maxlen = sizeof(int),
1217 .proc_handler = proc_dointvec
1220 .procname = "tcp_min_tso_segs",
1221 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1222 .maxlen = sizeof(int),
1224 .proc_handler = proc_dointvec_minmax,
1226 .extra2 = &gso_max_segs,
1229 .procname = "tcp_min_rtt_wlen",
1230 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1231 .maxlen = sizeof(int),
1233 .proc_handler = proc_dointvec_minmax,
1235 .extra2 = &one_day_secs
1238 .procname = "tcp_autocorking",
1239 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1240 .maxlen = sizeof(int),
1242 .proc_handler = proc_dointvec_minmax,
1247 .procname = "tcp_invalid_ratelimit",
1248 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1249 .maxlen = sizeof(int),
1251 .proc_handler = proc_dointvec_ms_jiffies,
1254 .procname = "tcp_pacing_ss_ratio",
1255 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1256 .maxlen = sizeof(int),
1258 .proc_handler = proc_dointvec_minmax,
1260 .extra2 = &thousand,
1263 .procname = "tcp_pacing_ca_ratio",
1264 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1265 .maxlen = sizeof(int),
1267 .proc_handler = proc_dointvec_minmax,
1269 .extra2 = &thousand,
1272 .procname = "tcp_wmem",
1273 .data = &init_net.ipv4.sysctl_tcp_wmem,
1274 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1276 .proc_handler = proc_dointvec_minmax,
1280 .procname = "tcp_rmem",
1281 .data = &init_net.ipv4.sysctl_tcp_rmem,
1282 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1284 .proc_handler = proc_dointvec_minmax,
1288 .procname = "tcp_comp_sack_delay_ns",
1289 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1290 .maxlen = sizeof(unsigned long),
1292 .proc_handler = proc_doulongvec_minmax,
1295 .procname = "tcp_comp_sack_nr",
1296 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1297 .maxlen = sizeof(int),
1299 .proc_handler = proc_dointvec_minmax,
1301 .extra2 = &comp_sack_nr_max,
1304 .procname = "udp_rmem_min",
1305 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1306 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1308 .proc_handler = proc_dointvec_minmax,
1312 .procname = "udp_wmem_min",
1313 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1314 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1316 .proc_handler = proc_dointvec_minmax,
1322 static __net_init int ipv4_sysctl_init_net(struct net *net)
1324 struct ctl_table *table;
1326 table = ipv4_net_table;
1327 if (!net_eq(net, &init_net)) {
1330 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1334 /* Update the variables to point into the current struct net */
1335 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1336 table[i].data += (void *)net - (void *)&init_net;
1339 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1340 if (!net->ipv4.ipv4_hdr)
1343 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1344 if (!net->ipv4.sysctl_local_reserved_ports)
1350 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1352 if (!net_eq(net, &init_net))
1358 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1360 struct ctl_table *table;
1362 kfree(net->ipv4.sysctl_local_reserved_ports);
1363 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1364 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1368 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1369 .init = ipv4_sysctl_init_net,
1370 .exit = ipv4_sysctl_exit_net,
1373 static __init int sysctl_ipv4_init(void)
1375 struct ctl_table_header *hdr;
1377 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1381 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1382 unregister_net_sysctl_table(hdr);
1389 __initcall(sysctl_ipv4_init);