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>
33 static int thousand = 1000;
34 static int gso_max_segs = GSO_MAX_SEGS;
35 static int tcp_retr1_max = 255;
36 static int ip_local_port_range_min[] = { 1, 1 };
37 static int ip_local_port_range_max[] = { 65535, 65535 };
38 static int tcp_adv_win_scale_min = -31;
39 static int tcp_adv_win_scale_max = 31;
40 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
41 static int tcp_min_snd_mss_max = 65535;
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;
51 static u32 u32_max_div_HZ = UINT_MAX / HZ;
52 static int one_day_secs = 24 * 3600;
55 static int sysctl_tcp_low_latency __read_mostly;
57 /* Update system visible IP port range */
58 static void set_local_port_range(struct net *net, int range[2])
60 bool same_parity = !((range[0] ^ range[1]) & 1);
62 write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
63 if (same_parity && !net->ipv4.ip_local_ports.warned) {
64 net->ipv4.ip_local_ports.warned = true;
65 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
67 net->ipv4.ip_local_ports.range[0] = range[0];
68 net->ipv4.ip_local_ports.range[1] = range[1];
69 write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
72 /* Validate changes from /proc interface. */
73 static int ipv4_local_port_range(struct ctl_table *table, int write,
75 size_t *lenp, loff_t *ppos)
78 container_of(table->data, struct net, ipv4.ip_local_ports.range);
81 struct ctl_table tmp = {
83 .maxlen = sizeof(range),
85 .extra1 = &ip_local_port_range_min,
86 .extra2 = &ip_local_port_range_max,
89 inet_get_local_port_range(net, &range[0], &range[1]);
91 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
93 if (write && ret == 0) {
94 /* Ensure that the upper limit is not smaller than the lower,
95 * and that the lower does not encroach upon the privileged
98 if ((range[1] < range[0]) ||
99 (range[0] < net->ipv4.sysctl_ip_prot_sock))
102 set_local_port_range(net, range);
108 /* Validate changes from /proc interface. */
109 static int ipv4_privileged_ports(struct ctl_table *table, int write,
110 void __user *buffer, size_t *lenp, loff_t *ppos)
112 struct net *net = container_of(table->data, struct net,
113 ipv4.sysctl_ip_prot_sock);
117 struct ctl_table tmp = {
119 .maxlen = sizeof(pports),
121 .extra1 = &ip_privileged_port_min,
122 .extra2 = &ip_privileged_port_max,
125 pports = net->ipv4.sysctl_ip_prot_sock;
127 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
129 if (write && ret == 0) {
130 inet_get_local_port_range(net, &range[0], &range[1]);
131 /* Ensure that the local port range doesn't overlap with the
132 * privileged port range.
134 if (range[0] < pports)
137 net->ipv4.sysctl_ip_prot_sock = pports;
143 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
145 kgid_t *data = table->data;
147 container_of(table->data, struct net, ipv4.ping_group_range.range);
150 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
154 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
157 /* Update system visible IP port range */
158 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
160 kgid_t *data = table->data;
162 container_of(table->data, struct net, ipv4.ping_group_range.range);
163 write_seqlock(&net->ipv4.ping_group_range.lock);
166 write_sequnlock(&net->ipv4.ping_group_range.lock);
169 /* Validate changes from /proc interface. */
170 static int ipv4_ping_group_range(struct ctl_table *table, int write,
172 size_t *lenp, loff_t *ppos)
174 struct user_namespace *user_ns = current_user_ns();
178 struct ctl_table tmp = {
180 .maxlen = sizeof(urange),
182 .extra1 = &ip_ping_group_range_min,
183 .extra2 = &ip_ping_group_range_max,
186 inet_get_ping_group_range_table(table, &low, &high);
187 urange[0] = from_kgid_munged(user_ns, low);
188 urange[1] = from_kgid_munged(user_ns, high);
189 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
191 if (write && ret == 0) {
192 low = make_kgid(user_ns, urange[0]);
193 high = make_kgid(user_ns, urange[1]);
194 if (!gid_valid(low) || !gid_valid(high))
196 if (urange[1] < urange[0] || gid_lt(high, low)) {
197 low = make_kgid(&init_user_ns, 1);
198 high = make_kgid(&init_user_ns, 0);
200 set_ping_group_range(table, low, high);
206 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
208 size_t *lenp, loff_t *ppos)
213 net = container_of(table->data, struct net,
214 ipv4.sysctl_ip_fwd_update_priority);
215 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
216 if (write && ret == 0)
217 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
223 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
224 void __user *buffer, size_t *lenp, loff_t *ppos)
226 struct net *net = container_of(ctl->data, struct net,
227 ipv4.tcp_congestion_control);
228 char val[TCP_CA_NAME_MAX];
229 struct ctl_table tbl = {
231 .maxlen = TCP_CA_NAME_MAX,
235 tcp_get_default_congestion_control(net, val);
237 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
238 if (write && ret == 0)
239 ret = tcp_set_default_congestion_control(net, val);
243 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
245 void __user *buffer, size_t *lenp,
248 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
251 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
254 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
255 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
260 static int proc_allowed_congestion_control(struct ctl_table *ctl,
262 void __user *buffer, size_t *lenp,
265 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
268 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
272 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
273 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
274 if (write && ret == 0)
275 ret = tcp_set_allowed_congestion_control(tbl.data);
280 static int sscanf_key(char *buf, __le32 *key)
285 if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
286 user_key + 2, user_key + 3) != 4) {
289 for (i = 0; i < ARRAY_SIZE(user_key); i++)
290 key[i] = cpu_to_le32(user_key[i]);
292 pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
293 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
298 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
299 void __user *buffer, size_t *lenp,
302 struct net *net = container_of(table->data, struct net,
303 ipv4.sysctl_tcp_fastopen);
304 /* maxlen to print the list of keys in hex (*2), with dashes
305 * separating doublewords and a comma in between keys.
307 struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
308 2 * TCP_FASTOPEN_KEY_MAX) +
309 (TCP_FASTOPEN_KEY_MAX * 5)) };
310 struct tcp_fastopen_context *ctx;
311 u32 user_key[TCP_FASTOPEN_KEY_MAX * 4];
312 __le32 key[TCP_FASTOPEN_KEY_MAX * 4];
314 int ret, i = 0, off = 0, n_keys = 0;
316 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
321 ctx = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
323 n_keys = tcp_fastopen_context_len(ctx);
324 memcpy(&key[0], &ctx->key[0], TCP_FASTOPEN_KEY_LENGTH * n_keys);
329 memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
333 for (i = 0; i < n_keys * 4; i++)
334 user_key[i] = le32_to_cpu(key[i]);
336 for (i = 0; i < n_keys; i++) {
337 off += snprintf(tbl.data + off, tbl.maxlen - off,
338 "%08x-%08x-%08x-%08x",
342 user_key[i * 4 + 3]);
344 off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
347 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
349 if (write && ret == 0) {
350 backup_data = strchr(tbl.data, ',');
355 if (sscanf_key(tbl.data, key)) {
360 if (sscanf_key(backup_data, key + 4)) {
365 tcp_fastopen_reset_cipher(net, NULL, key,
366 backup_data ? key + 4 : NULL);
374 static void proc_configure_early_demux(int enabled, int protocol)
376 struct net_protocol *ipprot;
377 #if IS_ENABLED(CONFIG_IPV6)
378 struct inet6_protocol *ip6prot;
383 ipprot = rcu_dereference(inet_protos[protocol]);
385 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
388 #if IS_ENABLED(CONFIG_IPV6)
389 ip6prot = rcu_dereference(inet6_protos[protocol]);
391 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
397 static int proc_tcp_early_demux(struct ctl_table *table, int write,
398 void __user *buffer, size_t *lenp, loff_t *ppos)
402 ret = proc_dointvec(table, write, buffer, lenp, ppos);
405 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
407 proc_configure_early_demux(enabled, IPPROTO_TCP);
413 static int proc_udp_early_demux(struct ctl_table *table, int write,
414 void __user *buffer, size_t *lenp, loff_t *ppos)
418 ret = proc_dointvec(table, write, buffer, lenp, ppos);
421 int enabled = init_net.ipv4.sysctl_udp_early_demux;
423 proc_configure_early_demux(enabled, IPPROTO_UDP);
429 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
432 size_t *lenp, loff_t *ppos)
434 struct net *net = container_of(table->data, struct net,
435 ipv4.sysctl_tcp_fastopen_blackhole_timeout);
438 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
439 if (write && ret == 0)
440 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
445 static int proc_tcp_available_ulp(struct ctl_table *ctl,
447 void __user *buffer, size_t *lenp,
450 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
453 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
456 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
457 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
463 #ifdef CONFIG_IP_ROUTE_MULTIPATH
464 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
465 void __user *buffer, size_t *lenp,
468 struct net *net = container_of(table->data, struct net,
469 ipv4.sysctl_fib_multipath_hash_policy);
472 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
473 if (write && ret == 0)
474 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
480 static struct ctl_table ipv4_table[] = {
482 .procname = "tcp_max_orphans",
483 .data = &sysctl_tcp_max_orphans,
484 .maxlen = sizeof(int),
486 .proc_handler = proc_dointvec
489 .procname = "inet_peer_threshold",
490 .data = &inet_peer_threshold,
491 .maxlen = sizeof(int),
493 .proc_handler = proc_dointvec
496 .procname = "inet_peer_minttl",
497 .data = &inet_peer_minttl,
498 .maxlen = sizeof(int),
500 .proc_handler = proc_dointvec_jiffies,
503 .procname = "inet_peer_maxttl",
504 .data = &inet_peer_maxttl,
505 .maxlen = sizeof(int),
507 .proc_handler = proc_dointvec_jiffies,
510 .procname = "tcp_mem",
511 .maxlen = sizeof(sysctl_tcp_mem),
512 .data = &sysctl_tcp_mem,
514 .proc_handler = proc_doulongvec_minmax,
517 .procname = "tcp_low_latency",
518 .data = &sysctl_tcp_low_latency,
519 .maxlen = sizeof(int),
521 .proc_handler = proc_dointvec
523 #ifdef CONFIG_NETLABEL
525 .procname = "cipso_cache_enable",
526 .data = &cipso_v4_cache_enabled,
527 .maxlen = sizeof(int),
529 .proc_handler = proc_dointvec,
532 .procname = "cipso_cache_bucket_size",
533 .data = &cipso_v4_cache_bucketsize,
534 .maxlen = sizeof(int),
536 .proc_handler = proc_dointvec,
539 .procname = "cipso_rbm_optfmt",
540 .data = &cipso_v4_rbm_optfmt,
541 .maxlen = sizeof(int),
543 .proc_handler = proc_dointvec,
546 .procname = "cipso_rbm_strictvalid",
547 .data = &cipso_v4_rbm_strictvalid,
548 .maxlen = sizeof(int),
550 .proc_handler = proc_dointvec,
552 #endif /* CONFIG_NETLABEL */
554 .procname = "tcp_available_congestion_control",
555 .maxlen = TCP_CA_BUF_MAX,
557 .proc_handler = proc_tcp_available_congestion_control,
560 .procname = "tcp_allowed_congestion_control",
561 .maxlen = TCP_CA_BUF_MAX,
563 .proc_handler = proc_allowed_congestion_control,
566 .procname = "tcp_available_ulp",
567 .maxlen = TCP_ULP_BUF_MAX,
569 .proc_handler = proc_tcp_available_ulp,
572 .procname = "icmp_msgs_per_sec",
573 .data = &sysctl_icmp_msgs_per_sec,
574 .maxlen = sizeof(int),
576 .proc_handler = proc_dointvec_minmax,
577 .extra1 = SYSCTL_ZERO,
580 .procname = "icmp_msgs_burst",
581 .data = &sysctl_icmp_msgs_burst,
582 .maxlen = sizeof(int),
584 .proc_handler = proc_dointvec_minmax,
585 .extra1 = SYSCTL_ZERO,
588 .procname = "udp_mem",
589 .data = &sysctl_udp_mem,
590 .maxlen = sizeof(sysctl_udp_mem),
592 .proc_handler = proc_doulongvec_minmax,
595 .procname = "fib_sync_mem",
596 .data = &sysctl_fib_sync_mem,
597 .maxlen = sizeof(sysctl_fib_sync_mem),
599 .proc_handler = proc_douintvec_minmax,
600 .extra1 = &sysctl_fib_sync_mem_min,
601 .extra2 = &sysctl_fib_sync_mem_max,
604 .procname = "tcp_rx_skb_cache",
605 .data = &tcp_rx_skb_cache_key.key,
607 .proc_handler = proc_do_static_key,
610 .procname = "tcp_tx_skb_cache",
611 .data = &tcp_tx_skb_cache_key.key,
613 .proc_handler = proc_do_static_key,
618 static struct ctl_table ipv4_net_table[] = {
620 .procname = "icmp_echo_ignore_all",
621 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
622 .maxlen = sizeof(int),
624 .proc_handler = proc_dointvec
627 .procname = "icmp_echo_ignore_broadcasts",
628 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
629 .maxlen = sizeof(int),
631 .proc_handler = proc_dointvec
634 .procname = "icmp_ignore_bogus_error_responses",
635 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
636 .maxlen = sizeof(int),
638 .proc_handler = proc_dointvec
641 .procname = "icmp_errors_use_inbound_ifaddr",
642 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
643 .maxlen = sizeof(int),
645 .proc_handler = proc_dointvec
648 .procname = "icmp_ratelimit",
649 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
650 .maxlen = sizeof(int),
652 .proc_handler = proc_dointvec_ms_jiffies,
655 .procname = "icmp_ratemask",
656 .data = &init_net.ipv4.sysctl_icmp_ratemask,
657 .maxlen = sizeof(int),
659 .proc_handler = proc_dointvec
662 .procname = "ping_group_range",
663 .data = &init_net.ipv4.ping_group_range.range,
664 .maxlen = sizeof(gid_t)*2,
666 .proc_handler = ipv4_ping_group_range,
668 #ifdef CONFIG_NET_L3_MASTER_DEV
670 .procname = "raw_l3mdev_accept",
671 .data = &init_net.ipv4.sysctl_raw_l3mdev_accept,
672 .maxlen = sizeof(int),
674 .proc_handler = proc_dointvec_minmax,
675 .extra1 = SYSCTL_ZERO,
676 .extra2 = SYSCTL_ONE,
680 .procname = "tcp_ecn",
681 .data = &init_net.ipv4.sysctl_tcp_ecn,
682 .maxlen = sizeof(int),
684 .proc_handler = proc_dointvec
687 .procname = "tcp_ecn_fallback",
688 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
689 .maxlen = sizeof(int),
691 .proc_handler = proc_dointvec
694 .procname = "ip_dynaddr",
695 .data = &init_net.ipv4.sysctl_ip_dynaddr,
696 .maxlen = sizeof(int),
698 .proc_handler = proc_dointvec
701 .procname = "ip_early_demux",
702 .data = &init_net.ipv4.sysctl_ip_early_demux,
703 .maxlen = sizeof(int),
705 .proc_handler = proc_dointvec
708 .procname = "udp_early_demux",
709 .data = &init_net.ipv4.sysctl_udp_early_demux,
710 .maxlen = sizeof(int),
712 .proc_handler = proc_udp_early_demux
715 .procname = "tcp_early_demux",
716 .data = &init_net.ipv4.sysctl_tcp_early_demux,
717 .maxlen = sizeof(int),
719 .proc_handler = proc_tcp_early_demux
722 .procname = "ip_default_ttl",
723 .data = &init_net.ipv4.sysctl_ip_default_ttl,
724 .maxlen = sizeof(int),
726 .proc_handler = proc_dointvec_minmax,
727 .extra1 = &ip_ttl_min,
728 .extra2 = &ip_ttl_max,
731 .procname = "ip_local_port_range",
732 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
733 .data = &init_net.ipv4.ip_local_ports.range,
735 .proc_handler = ipv4_local_port_range,
738 .procname = "ip_local_reserved_ports",
739 .data = &init_net.ipv4.sysctl_local_reserved_ports,
742 .proc_handler = proc_do_large_bitmap,
745 .procname = "ip_no_pmtu_disc",
746 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
747 .maxlen = sizeof(int),
749 .proc_handler = proc_dointvec
752 .procname = "ip_forward_use_pmtu",
753 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
754 .maxlen = sizeof(int),
756 .proc_handler = proc_dointvec,
759 .procname = "ip_forward_update_priority",
760 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority,
761 .maxlen = sizeof(int),
763 .proc_handler = ipv4_fwd_update_priority,
764 .extra1 = SYSCTL_ZERO,
765 .extra2 = SYSCTL_ONE,
768 .procname = "ip_nonlocal_bind",
769 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
770 .maxlen = sizeof(int),
772 .proc_handler = proc_dointvec
775 .procname = "fwmark_reflect",
776 .data = &init_net.ipv4.sysctl_fwmark_reflect,
777 .maxlen = sizeof(int),
779 .proc_handler = proc_dointvec,
782 .procname = "tcp_fwmark_accept",
783 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept,
784 .maxlen = sizeof(int),
786 .proc_handler = proc_dointvec,
788 #ifdef CONFIG_NET_L3_MASTER_DEV
790 .procname = "tcp_l3mdev_accept",
791 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
792 .maxlen = sizeof(int),
794 .proc_handler = proc_dointvec_minmax,
795 .extra1 = SYSCTL_ZERO,
796 .extra2 = SYSCTL_ONE,
800 .procname = "tcp_mtu_probing",
801 .data = &init_net.ipv4.sysctl_tcp_mtu_probing,
802 .maxlen = sizeof(int),
804 .proc_handler = proc_dointvec,
807 .procname = "tcp_base_mss",
808 .data = &init_net.ipv4.sysctl_tcp_base_mss,
809 .maxlen = sizeof(int),
811 .proc_handler = proc_dointvec,
814 .procname = "tcp_min_snd_mss",
815 .data = &init_net.ipv4.sysctl_tcp_min_snd_mss,
816 .maxlen = sizeof(int),
818 .proc_handler = proc_dointvec_minmax,
819 .extra1 = &tcp_min_snd_mss_min,
820 .extra2 = &tcp_min_snd_mss_max,
823 .procname = "tcp_mtu_probe_floor",
824 .data = &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
825 .maxlen = sizeof(int),
827 .proc_handler = proc_dointvec_minmax,
828 .extra1 = &tcp_min_snd_mss_min,
829 .extra2 = &tcp_min_snd_mss_max,
832 .procname = "tcp_probe_threshold",
833 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
834 .maxlen = sizeof(int),
836 .proc_handler = proc_dointvec,
839 .procname = "tcp_probe_interval",
840 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
841 .maxlen = sizeof(u32),
843 .proc_handler = proc_douintvec_minmax,
844 .extra2 = &u32_max_div_HZ,
847 .procname = "igmp_link_local_mcast_reports",
848 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
849 .maxlen = sizeof(int),
851 .proc_handler = proc_dointvec
854 .procname = "igmp_max_memberships",
855 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
856 .maxlen = sizeof(int),
858 .proc_handler = proc_dointvec
861 .procname = "igmp_max_msf",
862 .data = &init_net.ipv4.sysctl_igmp_max_msf,
863 .maxlen = sizeof(int),
865 .proc_handler = proc_dointvec
867 #ifdef CONFIG_IP_MULTICAST
869 .procname = "igmp_qrv",
870 .data = &init_net.ipv4.sysctl_igmp_qrv,
871 .maxlen = sizeof(int),
873 .proc_handler = proc_dointvec_minmax,
878 .procname = "tcp_congestion_control",
879 .data = &init_net.ipv4.tcp_congestion_control,
881 .maxlen = TCP_CA_NAME_MAX,
882 .proc_handler = proc_tcp_congestion_control,
885 .procname = "tcp_keepalive_time",
886 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
887 .maxlen = sizeof(int),
889 .proc_handler = proc_dointvec_jiffies,
892 .procname = "tcp_keepalive_probes",
893 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
894 .maxlen = sizeof(int),
896 .proc_handler = proc_dointvec
899 .procname = "tcp_keepalive_intvl",
900 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
901 .maxlen = sizeof(int),
903 .proc_handler = proc_dointvec_jiffies,
906 .procname = "tcp_syn_retries",
907 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
908 .maxlen = sizeof(int),
910 .proc_handler = proc_dointvec_minmax,
911 .extra1 = &tcp_syn_retries_min,
912 .extra2 = &tcp_syn_retries_max
915 .procname = "tcp_synack_retries",
916 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
917 .maxlen = sizeof(int),
919 .proc_handler = proc_dointvec
921 #ifdef CONFIG_SYN_COOKIES
923 .procname = "tcp_syncookies",
924 .data = &init_net.ipv4.sysctl_tcp_syncookies,
925 .maxlen = sizeof(int),
927 .proc_handler = proc_dointvec
931 .procname = "tcp_reordering",
932 .data = &init_net.ipv4.sysctl_tcp_reordering,
933 .maxlen = sizeof(int),
935 .proc_handler = proc_dointvec
938 .procname = "tcp_retries1",
939 .data = &init_net.ipv4.sysctl_tcp_retries1,
940 .maxlen = sizeof(int),
942 .proc_handler = proc_dointvec_minmax,
943 .extra2 = &tcp_retr1_max
946 .procname = "tcp_retries2",
947 .data = &init_net.ipv4.sysctl_tcp_retries2,
948 .maxlen = sizeof(int),
950 .proc_handler = proc_dointvec
953 .procname = "tcp_orphan_retries",
954 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
955 .maxlen = sizeof(int),
957 .proc_handler = proc_dointvec
960 .procname = "tcp_fin_timeout",
961 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
962 .maxlen = sizeof(int),
964 .proc_handler = proc_dointvec_jiffies,
967 .procname = "tcp_notsent_lowat",
968 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
969 .maxlen = sizeof(unsigned int),
971 .proc_handler = proc_douintvec,
974 .procname = "tcp_tw_reuse",
975 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
976 .maxlen = sizeof(int),
978 .proc_handler = proc_dointvec_minmax,
979 .extra1 = SYSCTL_ZERO,
983 .procname = "tcp_max_tw_buckets",
984 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
985 .maxlen = sizeof(int),
987 .proc_handler = proc_dointvec
990 .procname = "tcp_max_syn_backlog",
991 .data = &init_net.ipv4.sysctl_max_syn_backlog,
992 .maxlen = sizeof(int),
994 .proc_handler = proc_dointvec
997 .procname = "tcp_fastopen",
998 .data = &init_net.ipv4.sysctl_tcp_fastopen,
999 .maxlen = sizeof(int),
1001 .proc_handler = proc_dointvec,
1004 .procname = "tcp_fastopen_key",
1006 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1007 /* maxlen to print the list of keys in hex (*2), with dashes
1008 * separating doublewords and a comma in between keys.
1010 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
1011 2 * TCP_FASTOPEN_KEY_MAX) +
1012 (TCP_FASTOPEN_KEY_MAX * 5)),
1013 .proc_handler = proc_tcp_fastopen_key,
1016 .procname = "tcp_fastopen_blackhole_timeout_sec",
1017 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1018 .maxlen = sizeof(int),
1020 .proc_handler = proc_tfo_blackhole_detect_timeout,
1021 .extra1 = SYSCTL_ZERO,
1023 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1025 .procname = "fib_multipath_use_neigh",
1026 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1027 .maxlen = sizeof(int),
1029 .proc_handler = proc_dointvec_minmax,
1030 .extra1 = SYSCTL_ZERO,
1031 .extra2 = SYSCTL_ONE,
1034 .procname = "fib_multipath_hash_policy",
1035 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1036 .maxlen = sizeof(int),
1038 .proc_handler = proc_fib_multipath_hash_policy,
1039 .extra1 = SYSCTL_ZERO,
1040 .extra2 = SYSCTL_ONE,
1044 .procname = "ip_unprivileged_port_start",
1045 .maxlen = sizeof(int),
1046 .data = &init_net.ipv4.sysctl_ip_prot_sock,
1048 .proc_handler = ipv4_privileged_ports,
1050 #ifdef CONFIG_NET_L3_MASTER_DEV
1052 .procname = "udp_l3mdev_accept",
1053 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1054 .maxlen = sizeof(int),
1056 .proc_handler = proc_dointvec_minmax,
1057 .extra1 = SYSCTL_ZERO,
1058 .extra2 = SYSCTL_ONE,
1062 .procname = "tcp_sack",
1063 .data = &init_net.ipv4.sysctl_tcp_sack,
1064 .maxlen = sizeof(int),
1066 .proc_handler = proc_dointvec
1069 .procname = "tcp_window_scaling",
1070 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
1071 .maxlen = sizeof(int),
1073 .proc_handler = proc_dointvec
1076 .procname = "tcp_timestamps",
1077 .data = &init_net.ipv4.sysctl_tcp_timestamps,
1078 .maxlen = sizeof(int),
1080 .proc_handler = proc_dointvec
1083 .procname = "tcp_early_retrans",
1084 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
1085 .maxlen = sizeof(int),
1087 .proc_handler = proc_dointvec_minmax,
1088 .extra1 = SYSCTL_ZERO,
1092 .procname = "tcp_recovery",
1093 .data = &init_net.ipv4.sysctl_tcp_recovery,
1094 .maxlen = sizeof(int),
1096 .proc_handler = proc_dointvec,
1099 .procname = "tcp_thin_linear_timeouts",
1100 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1101 .maxlen = sizeof(int),
1103 .proc_handler = proc_dointvec
1106 .procname = "tcp_slow_start_after_idle",
1107 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1108 .maxlen = sizeof(int),
1110 .proc_handler = proc_dointvec
1113 .procname = "tcp_retrans_collapse",
1114 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1115 .maxlen = sizeof(int),
1117 .proc_handler = proc_dointvec
1120 .procname = "tcp_stdurg",
1121 .data = &init_net.ipv4.sysctl_tcp_stdurg,
1122 .maxlen = sizeof(int),
1124 .proc_handler = proc_dointvec
1127 .procname = "tcp_rfc1337",
1128 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
1129 .maxlen = sizeof(int),
1131 .proc_handler = proc_dointvec
1134 .procname = "tcp_abort_on_overflow",
1135 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1136 .maxlen = sizeof(int),
1138 .proc_handler = proc_dointvec
1141 .procname = "tcp_fack",
1142 .data = &init_net.ipv4.sysctl_tcp_fack,
1143 .maxlen = sizeof(int),
1145 .proc_handler = proc_dointvec
1148 .procname = "tcp_max_reordering",
1149 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1150 .maxlen = sizeof(int),
1152 .proc_handler = proc_dointvec
1155 .procname = "tcp_dsack",
1156 .data = &init_net.ipv4.sysctl_tcp_dsack,
1157 .maxlen = sizeof(int),
1159 .proc_handler = proc_dointvec
1162 .procname = "tcp_app_win",
1163 .data = &init_net.ipv4.sysctl_tcp_app_win,
1164 .maxlen = sizeof(int),
1166 .proc_handler = proc_dointvec
1169 .procname = "tcp_adv_win_scale",
1170 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1171 .maxlen = sizeof(int),
1173 .proc_handler = proc_dointvec_minmax,
1174 .extra1 = &tcp_adv_win_scale_min,
1175 .extra2 = &tcp_adv_win_scale_max,
1178 .procname = "tcp_frto",
1179 .data = &init_net.ipv4.sysctl_tcp_frto,
1180 .maxlen = sizeof(int),
1182 .proc_handler = proc_dointvec
1185 .procname = "tcp_no_metrics_save",
1186 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1187 .maxlen = sizeof(int),
1189 .proc_handler = proc_dointvec,
1192 .procname = "tcp_moderate_rcvbuf",
1193 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1194 .maxlen = sizeof(int),
1196 .proc_handler = proc_dointvec,
1199 .procname = "tcp_tso_win_divisor",
1200 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1201 .maxlen = sizeof(int),
1203 .proc_handler = proc_dointvec,
1206 .procname = "tcp_workaround_signed_windows",
1207 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1208 .maxlen = sizeof(int),
1210 .proc_handler = proc_dointvec
1213 .procname = "tcp_limit_output_bytes",
1214 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1215 .maxlen = sizeof(int),
1217 .proc_handler = proc_dointvec
1220 .procname = "tcp_challenge_ack_limit",
1221 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1222 .maxlen = sizeof(int),
1224 .proc_handler = proc_dointvec
1227 .procname = "tcp_min_tso_segs",
1228 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1229 .maxlen = sizeof(int),
1231 .proc_handler = proc_dointvec_minmax,
1232 .extra1 = SYSCTL_ONE,
1233 .extra2 = &gso_max_segs,
1236 .procname = "tcp_min_rtt_wlen",
1237 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1238 .maxlen = sizeof(int),
1240 .proc_handler = proc_dointvec_minmax,
1241 .extra1 = SYSCTL_ZERO,
1242 .extra2 = &one_day_secs
1245 .procname = "tcp_autocorking",
1246 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1247 .maxlen = sizeof(int),
1249 .proc_handler = proc_dointvec_minmax,
1250 .extra1 = SYSCTL_ZERO,
1251 .extra2 = SYSCTL_ONE,
1254 .procname = "tcp_invalid_ratelimit",
1255 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1256 .maxlen = sizeof(int),
1258 .proc_handler = proc_dointvec_ms_jiffies,
1261 .procname = "tcp_pacing_ss_ratio",
1262 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1263 .maxlen = sizeof(int),
1265 .proc_handler = proc_dointvec_minmax,
1266 .extra1 = SYSCTL_ZERO,
1267 .extra2 = &thousand,
1270 .procname = "tcp_pacing_ca_ratio",
1271 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1272 .maxlen = sizeof(int),
1274 .proc_handler = proc_dointvec_minmax,
1275 .extra1 = SYSCTL_ZERO,
1276 .extra2 = &thousand,
1279 .procname = "tcp_wmem",
1280 .data = &init_net.ipv4.sysctl_tcp_wmem,
1281 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1283 .proc_handler = proc_dointvec_minmax,
1284 .extra1 = SYSCTL_ONE,
1287 .procname = "tcp_rmem",
1288 .data = &init_net.ipv4.sysctl_tcp_rmem,
1289 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1291 .proc_handler = proc_dointvec_minmax,
1292 .extra1 = SYSCTL_ONE,
1295 .procname = "tcp_comp_sack_delay_ns",
1296 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1297 .maxlen = sizeof(unsigned long),
1299 .proc_handler = proc_doulongvec_minmax,
1302 .procname = "tcp_comp_sack_nr",
1303 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1304 .maxlen = sizeof(int),
1306 .proc_handler = proc_dointvec_minmax,
1307 .extra1 = SYSCTL_ZERO,
1308 .extra2 = &comp_sack_nr_max,
1311 .procname = "udp_rmem_min",
1312 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1313 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1315 .proc_handler = proc_dointvec_minmax,
1316 .extra1 = SYSCTL_ONE
1319 .procname = "udp_wmem_min",
1320 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1321 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1323 .proc_handler = proc_dointvec_minmax,
1324 .extra1 = SYSCTL_ONE
1329 static __net_init int ipv4_sysctl_init_net(struct net *net)
1331 struct ctl_table *table;
1333 table = ipv4_net_table;
1334 if (!net_eq(net, &init_net)) {
1337 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1341 /* Update the variables to point into the current struct net */
1342 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1343 table[i].data += (void *)net - (void *)&init_net;
1346 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1347 if (!net->ipv4.ipv4_hdr)
1350 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1351 if (!net->ipv4.sysctl_local_reserved_ports)
1357 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1359 if (!net_eq(net, &init_net))
1365 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1367 struct ctl_table *table;
1369 kfree(net->ipv4.sysctl_local_reserved_ports);
1370 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1371 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1375 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1376 .init = ipv4_sysctl_init_net,
1377 .exit = ipv4_sysctl_exit_net,
1380 static __init int sysctl_ipv4_init(void)
1382 struct ctl_table_header *hdr;
1384 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1388 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1389 unregister_net_sysctl_table(hdr);
1396 __initcall(sysctl_ipv4_init);