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,
74 void *buffer, size_t *lenp, loff_t *ppos)
77 container_of(table->data, struct net, ipv4.ip_local_ports.range);
80 struct ctl_table tmp = {
82 .maxlen = sizeof(range),
84 .extra1 = &ip_local_port_range_min,
85 .extra2 = &ip_local_port_range_max,
88 inet_get_local_port_range(net, &range[0], &range[1]);
90 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
92 if (write && ret == 0) {
93 /* Ensure that the upper limit is not smaller than the lower,
94 * and that the lower does not encroach upon the privileged
97 if ((range[1] < range[0]) ||
98 (range[0] < net->ipv4.sysctl_ip_prot_sock))
101 set_local_port_range(net, range);
107 /* Validate changes from /proc interface. */
108 static int ipv4_privileged_ports(struct ctl_table *table, int write,
109 void *buffer, size_t *lenp, loff_t *ppos)
111 struct net *net = container_of(table->data, struct net,
112 ipv4.sysctl_ip_prot_sock);
116 struct ctl_table tmp = {
118 .maxlen = sizeof(pports),
120 .extra1 = &ip_privileged_port_min,
121 .extra2 = &ip_privileged_port_max,
124 pports = net->ipv4.sysctl_ip_prot_sock;
126 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
128 if (write && ret == 0) {
129 inet_get_local_port_range(net, &range[0], &range[1]);
130 /* Ensure that the local port range doesn't overlap with the
131 * privileged port range.
133 if (range[0] < pports)
136 net->ipv4.sysctl_ip_prot_sock = pports;
142 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
144 kgid_t *data = table->data;
146 container_of(table->data, struct net, ipv4.ping_group_range.range);
149 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
153 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
156 /* Update system visible IP port range */
157 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
159 kgid_t *data = table->data;
161 container_of(table->data, struct net, ipv4.ping_group_range.range);
162 write_seqlock(&net->ipv4.ping_group_range.lock);
165 write_sequnlock(&net->ipv4.ping_group_range.lock);
168 /* Validate changes from /proc interface. */
169 static int ipv4_ping_group_range(struct ctl_table *table, int write,
170 void *buffer, 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,
205 void *buffer, size_t *lenp, loff_t *ppos)
210 net = container_of(table->data, struct net,
211 ipv4.sysctl_ip_fwd_update_priority);
212 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
213 if (write && ret == 0)
214 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
220 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
221 void *buffer, size_t *lenp, loff_t *ppos)
223 struct net *net = container_of(ctl->data, struct net,
224 ipv4.tcp_congestion_control);
225 char val[TCP_CA_NAME_MAX];
226 struct ctl_table tbl = {
228 .maxlen = TCP_CA_NAME_MAX,
232 tcp_get_default_congestion_control(net, val);
234 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
235 if (write && ret == 0)
236 ret = tcp_set_default_congestion_control(net, val);
240 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
241 int write, void *buffer,
242 size_t *lenp, loff_t *ppos)
244 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
247 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
250 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
251 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
256 static int proc_allowed_congestion_control(struct ctl_table *ctl,
257 int write, void *buffer,
258 size_t *lenp, loff_t *ppos)
260 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
263 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
267 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
268 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
269 if (write && ret == 0)
270 ret = tcp_set_allowed_congestion_control(tbl.data);
275 static int sscanf_key(char *buf, __le32 *key)
280 if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
281 user_key + 2, user_key + 3) != 4) {
284 for (i = 0; i < ARRAY_SIZE(user_key); i++)
285 key[i] = cpu_to_le32(user_key[i]);
287 pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
288 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
293 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
294 void *buffer, size_t *lenp, loff_t *ppos)
296 struct net *net = container_of(table->data, struct net,
297 ipv4.sysctl_tcp_fastopen);
298 /* maxlen to print the list of keys in hex (*2), with dashes
299 * separating doublewords and a comma in between keys.
301 struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
302 2 * TCP_FASTOPEN_KEY_MAX) +
303 (TCP_FASTOPEN_KEY_MAX * 5)) };
304 struct tcp_fastopen_context *ctx;
305 u32 user_key[TCP_FASTOPEN_KEY_MAX * 4];
306 __le32 key[TCP_FASTOPEN_KEY_MAX * 4];
308 int ret, i = 0, off = 0, n_keys = 0;
310 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
315 ctx = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
317 n_keys = tcp_fastopen_context_len(ctx);
318 memcpy(&key[0], &ctx->key[0], TCP_FASTOPEN_KEY_LENGTH * n_keys);
323 memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
327 for (i = 0; i < n_keys * 4; i++)
328 user_key[i] = le32_to_cpu(key[i]);
330 for (i = 0; i < n_keys; i++) {
331 off += snprintf(tbl.data + off, tbl.maxlen - off,
332 "%08x-%08x-%08x-%08x",
336 user_key[i * 4 + 3]);
338 if (WARN_ON_ONCE(off >= tbl.maxlen - 1))
342 off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
345 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
347 if (write && ret == 0) {
348 backup_data = strchr(tbl.data, ',');
353 if (sscanf_key(tbl.data, key)) {
358 if (sscanf_key(backup_data, key + 4)) {
363 tcp_fastopen_reset_cipher(net, NULL, key,
364 backup_data ? key + 4 : NULL);
372 static void proc_configure_early_demux(int enabled, int protocol)
374 struct net_protocol *ipprot;
375 #if IS_ENABLED(CONFIG_IPV6)
376 struct inet6_protocol *ip6prot;
381 ipprot = rcu_dereference(inet_protos[protocol]);
383 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
386 #if IS_ENABLED(CONFIG_IPV6)
387 ip6prot = rcu_dereference(inet6_protos[protocol]);
389 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
395 static int proc_tcp_early_demux(struct ctl_table *table, int write,
396 void *buffer, size_t *lenp, loff_t *ppos)
400 ret = proc_dointvec(table, write, buffer, lenp, ppos);
403 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
405 proc_configure_early_demux(enabled, IPPROTO_TCP);
411 static int proc_udp_early_demux(struct ctl_table *table, int write,
412 void *buffer, size_t *lenp, loff_t *ppos)
416 ret = proc_dointvec(table, write, buffer, lenp, ppos);
419 int enabled = init_net.ipv4.sysctl_udp_early_demux;
421 proc_configure_early_demux(enabled, IPPROTO_UDP);
427 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
428 int write, void *buffer,
429 size_t *lenp, loff_t *ppos)
431 struct net *net = container_of(table->data, struct net,
432 ipv4.sysctl_tcp_fastopen_blackhole_timeout);
435 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
436 if (write && ret == 0)
437 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
442 static int proc_tcp_available_ulp(struct ctl_table *ctl,
443 int write, void *buffer, size_t *lenp,
446 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
449 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
452 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
453 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
459 #ifdef CONFIG_IP_ROUTE_MULTIPATH
460 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
461 void *buffer, size_t *lenp,
464 struct net *net = container_of(table->data, struct net,
465 ipv4.sysctl_fib_multipath_hash_policy);
468 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
469 if (write && ret == 0)
470 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
476 static struct ctl_table ipv4_table[] = {
478 .procname = "tcp_max_orphans",
479 .data = &sysctl_tcp_max_orphans,
480 .maxlen = sizeof(int),
482 .proc_handler = proc_dointvec
485 .procname = "inet_peer_threshold",
486 .data = &inet_peer_threshold,
487 .maxlen = sizeof(int),
489 .proc_handler = proc_dointvec
492 .procname = "inet_peer_minttl",
493 .data = &inet_peer_minttl,
494 .maxlen = sizeof(int),
496 .proc_handler = proc_dointvec_jiffies,
499 .procname = "inet_peer_maxttl",
500 .data = &inet_peer_maxttl,
501 .maxlen = sizeof(int),
503 .proc_handler = proc_dointvec_jiffies,
506 .procname = "tcp_mem",
507 .maxlen = sizeof(sysctl_tcp_mem),
508 .data = &sysctl_tcp_mem,
510 .proc_handler = proc_doulongvec_minmax,
513 .procname = "tcp_low_latency",
514 .data = &sysctl_tcp_low_latency,
515 .maxlen = sizeof(int),
517 .proc_handler = proc_dointvec
519 #ifdef CONFIG_NETLABEL
521 .procname = "cipso_cache_enable",
522 .data = &cipso_v4_cache_enabled,
523 .maxlen = sizeof(int),
525 .proc_handler = proc_dointvec,
528 .procname = "cipso_cache_bucket_size",
529 .data = &cipso_v4_cache_bucketsize,
530 .maxlen = sizeof(int),
532 .proc_handler = proc_dointvec,
535 .procname = "cipso_rbm_optfmt",
536 .data = &cipso_v4_rbm_optfmt,
537 .maxlen = sizeof(int),
539 .proc_handler = proc_dointvec,
542 .procname = "cipso_rbm_strictvalid",
543 .data = &cipso_v4_rbm_strictvalid,
544 .maxlen = sizeof(int),
546 .proc_handler = proc_dointvec,
548 #endif /* CONFIG_NETLABEL */
550 .procname = "tcp_available_ulp",
551 .maxlen = TCP_ULP_BUF_MAX,
553 .proc_handler = proc_tcp_available_ulp,
556 .procname = "icmp_msgs_per_sec",
557 .data = &sysctl_icmp_msgs_per_sec,
558 .maxlen = sizeof(int),
560 .proc_handler = proc_dointvec_minmax,
561 .extra1 = SYSCTL_ZERO,
564 .procname = "icmp_msgs_burst",
565 .data = &sysctl_icmp_msgs_burst,
566 .maxlen = sizeof(int),
568 .proc_handler = proc_dointvec_minmax,
569 .extra1 = SYSCTL_ZERO,
572 .procname = "udp_mem",
573 .data = &sysctl_udp_mem,
574 .maxlen = sizeof(sysctl_udp_mem),
576 .proc_handler = proc_doulongvec_minmax,
579 .procname = "fib_sync_mem",
580 .data = &sysctl_fib_sync_mem,
581 .maxlen = sizeof(sysctl_fib_sync_mem),
583 .proc_handler = proc_douintvec_minmax,
584 .extra1 = &sysctl_fib_sync_mem_min,
585 .extra2 = &sysctl_fib_sync_mem_max,
588 .procname = "tcp_rx_skb_cache",
589 .data = &tcp_rx_skb_cache_key.key,
591 .proc_handler = proc_do_static_key,
594 .procname = "tcp_tx_skb_cache",
595 .data = &tcp_tx_skb_cache_key.key,
597 .proc_handler = proc_do_static_key,
602 static struct ctl_table ipv4_net_table[] = {
604 .procname = "icmp_echo_ignore_all",
605 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
606 .maxlen = sizeof(int),
608 .proc_handler = proc_dointvec
611 .procname = "icmp_echo_ignore_broadcasts",
612 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
613 .maxlen = sizeof(int),
615 .proc_handler = proc_dointvec
618 .procname = "icmp_ignore_bogus_error_responses",
619 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
620 .maxlen = sizeof(int),
622 .proc_handler = proc_dointvec
625 .procname = "icmp_errors_use_inbound_ifaddr",
626 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
627 .maxlen = sizeof(int),
629 .proc_handler = proc_dointvec
632 .procname = "icmp_ratelimit",
633 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
634 .maxlen = sizeof(int),
636 .proc_handler = proc_dointvec_ms_jiffies,
639 .procname = "icmp_ratemask",
640 .data = &init_net.ipv4.sysctl_icmp_ratemask,
641 .maxlen = sizeof(int),
643 .proc_handler = proc_dointvec
646 .procname = "ping_group_range",
647 .data = &init_net.ipv4.ping_group_range.range,
648 .maxlen = sizeof(gid_t)*2,
650 .proc_handler = ipv4_ping_group_range,
652 #ifdef CONFIG_NET_L3_MASTER_DEV
654 .procname = "raw_l3mdev_accept",
655 .data = &init_net.ipv4.sysctl_raw_l3mdev_accept,
656 .maxlen = sizeof(int),
658 .proc_handler = proc_dointvec_minmax,
659 .extra1 = SYSCTL_ZERO,
660 .extra2 = SYSCTL_ONE,
664 .procname = "tcp_ecn",
665 .data = &init_net.ipv4.sysctl_tcp_ecn,
666 .maxlen = sizeof(int),
668 .proc_handler = proc_dointvec
671 .procname = "tcp_ecn_fallback",
672 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
673 .maxlen = sizeof(int),
675 .proc_handler = proc_dointvec
678 .procname = "ip_dynaddr",
679 .data = &init_net.ipv4.sysctl_ip_dynaddr,
680 .maxlen = sizeof(int),
682 .proc_handler = proc_dointvec
685 .procname = "ip_early_demux",
686 .data = &init_net.ipv4.sysctl_ip_early_demux,
687 .maxlen = sizeof(int),
689 .proc_handler = proc_dointvec
692 .procname = "udp_early_demux",
693 .data = &init_net.ipv4.sysctl_udp_early_demux,
694 .maxlen = sizeof(int),
696 .proc_handler = proc_udp_early_demux
699 .procname = "tcp_early_demux",
700 .data = &init_net.ipv4.sysctl_tcp_early_demux,
701 .maxlen = sizeof(int),
703 .proc_handler = proc_tcp_early_demux
706 .procname = "nexthop_compat_mode",
707 .data = &init_net.ipv4.sysctl_nexthop_compat_mode,
708 .maxlen = sizeof(int),
710 .proc_handler = proc_dointvec_minmax,
711 .extra1 = SYSCTL_ZERO,
712 .extra2 = SYSCTL_ONE,
715 .procname = "ip_default_ttl",
716 .data = &init_net.ipv4.sysctl_ip_default_ttl,
717 .maxlen = sizeof(int),
719 .proc_handler = proc_dointvec_minmax,
720 .extra1 = &ip_ttl_min,
721 .extra2 = &ip_ttl_max,
724 .procname = "ip_local_port_range",
725 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
726 .data = &init_net.ipv4.ip_local_ports.range,
728 .proc_handler = ipv4_local_port_range,
731 .procname = "ip_local_reserved_ports",
732 .data = &init_net.ipv4.sysctl_local_reserved_ports,
735 .proc_handler = proc_do_large_bitmap,
738 .procname = "ip_no_pmtu_disc",
739 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
740 .maxlen = sizeof(int),
742 .proc_handler = proc_dointvec
745 .procname = "ip_forward_use_pmtu",
746 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
747 .maxlen = sizeof(int),
749 .proc_handler = proc_dointvec,
752 .procname = "ip_forward_update_priority",
753 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority,
754 .maxlen = sizeof(int),
756 .proc_handler = ipv4_fwd_update_priority,
757 .extra1 = SYSCTL_ZERO,
758 .extra2 = SYSCTL_ONE,
761 .procname = "ip_nonlocal_bind",
762 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
763 .maxlen = sizeof(int),
765 .proc_handler = proc_dointvec
768 .procname = "ip_autobind_reuse",
769 .data = &init_net.ipv4.sysctl_ip_autobind_reuse,
770 .maxlen = sizeof(int),
772 .proc_handler = proc_dointvec_minmax,
773 .extra1 = SYSCTL_ZERO,
774 .extra2 = SYSCTL_ONE,
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,
797 .extra1 = SYSCTL_ZERO,
798 .extra2 = SYSCTL_ONE,
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_mtu_probe_floor",
826 .data = &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
827 .maxlen = sizeof(int),
829 .proc_handler = proc_dointvec_minmax,
830 .extra1 = &tcp_min_snd_mss_min,
831 .extra2 = &tcp_min_snd_mss_max,
834 .procname = "tcp_probe_threshold",
835 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
836 .maxlen = sizeof(int),
838 .proc_handler = proc_dointvec,
841 .procname = "tcp_probe_interval",
842 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
843 .maxlen = sizeof(u32),
845 .proc_handler = proc_douintvec_minmax,
846 .extra2 = &u32_max_div_HZ,
849 .procname = "igmp_link_local_mcast_reports",
850 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
851 .maxlen = sizeof(int),
853 .proc_handler = proc_dointvec
856 .procname = "igmp_max_memberships",
857 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
858 .maxlen = sizeof(int),
860 .proc_handler = proc_dointvec
863 .procname = "igmp_max_msf",
864 .data = &init_net.ipv4.sysctl_igmp_max_msf,
865 .maxlen = sizeof(int),
867 .proc_handler = proc_dointvec
869 #ifdef CONFIG_IP_MULTICAST
871 .procname = "igmp_qrv",
872 .data = &init_net.ipv4.sysctl_igmp_qrv,
873 .maxlen = sizeof(int),
875 .proc_handler = proc_dointvec_minmax,
880 .procname = "tcp_congestion_control",
881 .data = &init_net.ipv4.tcp_congestion_control,
883 .maxlen = TCP_CA_NAME_MAX,
884 .proc_handler = proc_tcp_congestion_control,
887 .procname = "tcp_available_congestion_control",
888 .maxlen = TCP_CA_BUF_MAX,
890 .proc_handler = proc_tcp_available_congestion_control,
893 .procname = "tcp_allowed_congestion_control",
894 .maxlen = TCP_CA_BUF_MAX,
896 .proc_handler = proc_allowed_congestion_control,
899 .procname = "tcp_keepalive_time",
900 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
901 .maxlen = sizeof(int),
903 .proc_handler = proc_dointvec_jiffies,
906 .procname = "tcp_keepalive_probes",
907 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
908 .maxlen = sizeof(int),
910 .proc_handler = proc_dointvec
913 .procname = "tcp_keepalive_intvl",
914 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
915 .maxlen = sizeof(int),
917 .proc_handler = proc_dointvec_jiffies,
920 .procname = "tcp_syn_retries",
921 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
922 .maxlen = sizeof(int),
924 .proc_handler = proc_dointvec_minmax,
925 .extra1 = &tcp_syn_retries_min,
926 .extra2 = &tcp_syn_retries_max
929 .procname = "tcp_synack_retries",
930 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
931 .maxlen = sizeof(int),
933 .proc_handler = proc_dointvec
935 #ifdef CONFIG_SYN_COOKIES
937 .procname = "tcp_syncookies",
938 .data = &init_net.ipv4.sysctl_tcp_syncookies,
939 .maxlen = sizeof(int),
941 .proc_handler = proc_dointvec
945 .procname = "tcp_reordering",
946 .data = &init_net.ipv4.sysctl_tcp_reordering,
947 .maxlen = sizeof(int),
949 .proc_handler = proc_dointvec
952 .procname = "tcp_retries1",
953 .data = &init_net.ipv4.sysctl_tcp_retries1,
954 .maxlen = sizeof(int),
956 .proc_handler = proc_dointvec_minmax,
957 .extra2 = &tcp_retr1_max
960 .procname = "tcp_retries2",
961 .data = &init_net.ipv4.sysctl_tcp_retries2,
962 .maxlen = sizeof(int),
964 .proc_handler = proc_dointvec
967 .procname = "tcp_orphan_retries",
968 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
969 .maxlen = sizeof(int),
971 .proc_handler = proc_dointvec
974 .procname = "tcp_fin_timeout",
975 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
976 .maxlen = sizeof(int),
978 .proc_handler = proc_dointvec_jiffies,
981 .procname = "tcp_notsent_lowat",
982 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
983 .maxlen = sizeof(unsigned int),
985 .proc_handler = proc_douintvec,
988 .procname = "tcp_tw_reuse",
989 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
990 .maxlen = sizeof(int),
992 .proc_handler = proc_dointvec_minmax,
993 .extra1 = SYSCTL_ZERO,
997 .procname = "tcp_max_tw_buckets",
998 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
999 .maxlen = sizeof(int),
1001 .proc_handler = proc_dointvec
1004 .procname = "tcp_max_syn_backlog",
1005 .data = &init_net.ipv4.sysctl_max_syn_backlog,
1006 .maxlen = sizeof(int),
1008 .proc_handler = proc_dointvec
1011 .procname = "tcp_fastopen",
1012 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1013 .maxlen = sizeof(int),
1015 .proc_handler = proc_dointvec,
1018 .procname = "tcp_fastopen_key",
1020 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1021 /* maxlen to print the list of keys in hex (*2), with dashes
1022 * separating doublewords and a comma in between keys.
1024 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
1025 2 * TCP_FASTOPEN_KEY_MAX) +
1026 (TCP_FASTOPEN_KEY_MAX * 5)),
1027 .proc_handler = proc_tcp_fastopen_key,
1030 .procname = "tcp_fastopen_blackhole_timeout_sec",
1031 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1032 .maxlen = sizeof(int),
1034 .proc_handler = proc_tfo_blackhole_detect_timeout,
1035 .extra1 = SYSCTL_ZERO,
1037 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1039 .procname = "fib_multipath_use_neigh",
1040 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1041 .maxlen = sizeof(int),
1043 .proc_handler = proc_dointvec_minmax,
1044 .extra1 = SYSCTL_ZERO,
1045 .extra2 = SYSCTL_ONE,
1048 .procname = "fib_multipath_hash_policy",
1049 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1050 .maxlen = sizeof(int),
1052 .proc_handler = proc_fib_multipath_hash_policy,
1053 .extra1 = SYSCTL_ZERO,
1058 .procname = "ip_unprivileged_port_start",
1059 .maxlen = sizeof(int),
1060 .data = &init_net.ipv4.sysctl_ip_prot_sock,
1062 .proc_handler = ipv4_privileged_ports,
1064 #ifdef CONFIG_NET_L3_MASTER_DEV
1066 .procname = "udp_l3mdev_accept",
1067 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1068 .maxlen = sizeof(int),
1070 .proc_handler = proc_dointvec_minmax,
1071 .extra1 = SYSCTL_ZERO,
1072 .extra2 = SYSCTL_ONE,
1076 .procname = "tcp_sack",
1077 .data = &init_net.ipv4.sysctl_tcp_sack,
1078 .maxlen = sizeof(int),
1080 .proc_handler = proc_dointvec
1083 .procname = "tcp_window_scaling",
1084 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
1085 .maxlen = sizeof(int),
1087 .proc_handler = proc_dointvec
1090 .procname = "tcp_timestamps",
1091 .data = &init_net.ipv4.sysctl_tcp_timestamps,
1092 .maxlen = sizeof(int),
1094 .proc_handler = proc_dointvec
1097 .procname = "tcp_early_retrans",
1098 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
1099 .maxlen = sizeof(int),
1101 .proc_handler = proc_dointvec_minmax,
1102 .extra1 = SYSCTL_ZERO,
1106 .procname = "tcp_recovery",
1107 .data = &init_net.ipv4.sysctl_tcp_recovery,
1108 .maxlen = sizeof(int),
1110 .proc_handler = proc_dointvec,
1113 .procname = "tcp_thin_linear_timeouts",
1114 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1115 .maxlen = sizeof(int),
1117 .proc_handler = proc_dointvec
1120 .procname = "tcp_slow_start_after_idle",
1121 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1122 .maxlen = sizeof(int),
1124 .proc_handler = proc_dointvec
1127 .procname = "tcp_retrans_collapse",
1128 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1129 .maxlen = sizeof(int),
1131 .proc_handler = proc_dointvec
1134 .procname = "tcp_stdurg",
1135 .data = &init_net.ipv4.sysctl_tcp_stdurg,
1136 .maxlen = sizeof(int),
1138 .proc_handler = proc_dointvec
1141 .procname = "tcp_rfc1337",
1142 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
1143 .maxlen = sizeof(int),
1145 .proc_handler = proc_dointvec
1148 .procname = "tcp_abort_on_overflow",
1149 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1150 .maxlen = sizeof(int),
1152 .proc_handler = proc_dointvec
1155 .procname = "tcp_fack",
1156 .data = &init_net.ipv4.sysctl_tcp_fack,
1157 .maxlen = sizeof(int),
1159 .proc_handler = proc_dointvec
1162 .procname = "tcp_max_reordering",
1163 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1164 .maxlen = sizeof(int),
1166 .proc_handler = proc_dointvec
1169 .procname = "tcp_dsack",
1170 .data = &init_net.ipv4.sysctl_tcp_dsack,
1171 .maxlen = sizeof(int),
1173 .proc_handler = proc_dointvec
1176 .procname = "tcp_app_win",
1177 .data = &init_net.ipv4.sysctl_tcp_app_win,
1178 .maxlen = sizeof(int),
1180 .proc_handler = proc_dointvec
1183 .procname = "tcp_adv_win_scale",
1184 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1185 .maxlen = sizeof(int),
1187 .proc_handler = proc_dointvec_minmax,
1188 .extra1 = &tcp_adv_win_scale_min,
1189 .extra2 = &tcp_adv_win_scale_max,
1192 .procname = "tcp_frto",
1193 .data = &init_net.ipv4.sysctl_tcp_frto,
1194 .maxlen = sizeof(int),
1196 .proc_handler = proc_dointvec
1199 .procname = "tcp_no_metrics_save",
1200 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1201 .maxlen = sizeof(int),
1203 .proc_handler = proc_dointvec,
1206 .procname = "tcp_no_ssthresh_metrics_save",
1207 .data = &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1208 .maxlen = sizeof(int),
1210 .proc_handler = proc_dointvec_minmax,
1211 .extra1 = SYSCTL_ZERO,
1212 .extra2 = SYSCTL_ONE,
1215 .procname = "tcp_moderate_rcvbuf",
1216 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1217 .maxlen = sizeof(int),
1219 .proc_handler = proc_dointvec,
1222 .procname = "tcp_tso_win_divisor",
1223 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1224 .maxlen = sizeof(int),
1226 .proc_handler = proc_dointvec,
1229 .procname = "tcp_workaround_signed_windows",
1230 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1231 .maxlen = sizeof(int),
1233 .proc_handler = proc_dointvec
1236 .procname = "tcp_limit_output_bytes",
1237 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1238 .maxlen = sizeof(int),
1240 .proc_handler = proc_dointvec
1243 .procname = "tcp_challenge_ack_limit",
1244 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1245 .maxlen = sizeof(int),
1247 .proc_handler = proc_dointvec
1250 .procname = "tcp_min_tso_segs",
1251 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1252 .maxlen = sizeof(int),
1254 .proc_handler = proc_dointvec_minmax,
1255 .extra1 = SYSCTL_ONE,
1256 .extra2 = &gso_max_segs,
1259 .procname = "tcp_min_rtt_wlen",
1260 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1261 .maxlen = sizeof(int),
1263 .proc_handler = proc_dointvec_minmax,
1264 .extra1 = SYSCTL_ZERO,
1265 .extra2 = &one_day_secs
1268 .procname = "tcp_autocorking",
1269 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1270 .maxlen = sizeof(int),
1272 .proc_handler = proc_dointvec_minmax,
1273 .extra1 = SYSCTL_ZERO,
1274 .extra2 = SYSCTL_ONE,
1277 .procname = "tcp_invalid_ratelimit",
1278 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1279 .maxlen = sizeof(int),
1281 .proc_handler = proc_dointvec_ms_jiffies,
1284 .procname = "tcp_pacing_ss_ratio",
1285 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1286 .maxlen = sizeof(int),
1288 .proc_handler = proc_dointvec_minmax,
1289 .extra1 = SYSCTL_ZERO,
1290 .extra2 = &thousand,
1293 .procname = "tcp_pacing_ca_ratio",
1294 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1295 .maxlen = sizeof(int),
1297 .proc_handler = proc_dointvec_minmax,
1298 .extra1 = SYSCTL_ZERO,
1299 .extra2 = &thousand,
1302 .procname = "tcp_wmem",
1303 .data = &init_net.ipv4.sysctl_tcp_wmem,
1304 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1306 .proc_handler = proc_dointvec_minmax,
1307 .extra1 = SYSCTL_ONE,
1310 .procname = "tcp_rmem",
1311 .data = &init_net.ipv4.sysctl_tcp_rmem,
1312 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1314 .proc_handler = proc_dointvec_minmax,
1315 .extra1 = SYSCTL_ONE,
1318 .procname = "tcp_comp_sack_delay_ns",
1319 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1320 .maxlen = sizeof(unsigned long),
1322 .proc_handler = proc_doulongvec_minmax,
1325 .procname = "tcp_comp_sack_slack_ns",
1326 .data = &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1327 .maxlen = sizeof(unsigned long),
1329 .proc_handler = proc_doulongvec_minmax,
1332 .procname = "tcp_comp_sack_nr",
1333 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1334 .maxlen = sizeof(int),
1336 .proc_handler = proc_dointvec_minmax,
1337 .extra1 = SYSCTL_ZERO,
1338 .extra2 = &comp_sack_nr_max,
1341 .procname = "udp_rmem_min",
1342 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1343 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1345 .proc_handler = proc_dointvec_minmax,
1346 .extra1 = SYSCTL_ONE
1349 .procname = "udp_wmem_min",
1350 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1351 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1353 .proc_handler = proc_dointvec_minmax,
1354 .extra1 = SYSCTL_ONE
1359 static __net_init int ipv4_sysctl_init_net(struct net *net)
1361 struct ctl_table *table;
1363 table = ipv4_net_table;
1364 if (!net_eq(net, &init_net)) {
1367 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1371 /* Update the variables to point into the current struct net */
1372 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1373 table[i].data += (void *)net - (void *)&init_net;
1376 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1377 if (!net->ipv4.ipv4_hdr)
1380 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1381 if (!net->ipv4.sysctl_local_reserved_ports)
1387 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1389 if (!net_eq(net, &init_net))
1395 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1397 struct ctl_table *table;
1399 kfree(net->ipv4.sysctl_local_reserved_ports);
1400 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1401 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1405 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1406 .init = ipv4_sysctl_init_net,
1407 .exit = ipv4_sysctl_exit_net,
1410 static __init int sysctl_ipv4_init(void)
1412 struct ctl_table_header *hdr;
1414 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1418 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1419 unregister_net_sysctl_table(hdr);
1426 __initcall(sysctl_ipv4_init);