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/ip_fib.h>
23 #include <net/route.h>
26 #include <net/cipso_ipv4.h>
27 #include <net/inet_frag.h>
29 #include <net/protocol.h>
30 #include <net/netevent.h>
33 static int three __maybe_unused = 3;
35 static int thousand = 1000;
36 static int tcp_retr1_max = 255;
37 static int ip_local_port_range_min[] = { 1, 1 };
38 static int ip_local_port_range_max[] = { 65535, 65535 };
39 static int tcp_adv_win_scale_min = -31;
40 static int tcp_adv_win_scale_max = 31;
41 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
42 static int tcp_min_snd_mss_max = 65535;
43 static int ip_privileged_port_min;
44 static int ip_privileged_port_max = 65535;
45 static int ip_ttl_min = 1;
46 static int ip_ttl_max = 255;
47 static int tcp_syn_retries_min = 1;
48 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
49 static int ip_ping_group_range_min[] = { 0, 0 };
50 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
51 static u32 u32_max_div_HZ = UINT_MAX / HZ;
52 static int one_day_secs = 24 * 3600;
53 static u32 fib_multipath_hash_fields_all_mask __maybe_unused =
54 FIB_MULTIPATH_HASH_FIELD_ALL_MASK;
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,
76 void *buffer, size_t *lenp, loff_t *ppos)
79 container_of(table->data, struct net, ipv4.ip_local_ports.range);
82 struct ctl_table tmp = {
84 .maxlen = sizeof(range),
86 .extra1 = &ip_local_port_range_min,
87 .extra2 = &ip_local_port_range_max,
90 inet_get_local_port_range(net, &range[0], &range[1]);
92 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
94 if (write && ret == 0) {
95 /* Ensure that the upper limit is not smaller than the lower,
96 * and that the lower does not encroach upon the privileged
99 if ((range[1] < range[0]) ||
100 (range[0] < net->ipv4.sysctl_ip_prot_sock))
103 set_local_port_range(net, range);
109 /* Validate changes from /proc interface. */
110 static int ipv4_privileged_ports(struct ctl_table *table, int write,
111 void *buffer, size_t *lenp, loff_t *ppos)
113 struct net *net = container_of(table->data, struct net,
114 ipv4.sysctl_ip_prot_sock);
118 struct ctl_table tmp = {
120 .maxlen = sizeof(pports),
122 .extra1 = &ip_privileged_port_min,
123 .extra2 = &ip_privileged_port_max,
126 pports = net->ipv4.sysctl_ip_prot_sock;
128 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
130 if (write && ret == 0) {
131 inet_get_local_port_range(net, &range[0], &range[1]);
132 /* Ensure that the local port range doesn't overlap with the
133 * privileged port range.
135 if (range[0] < pports)
138 net->ipv4.sysctl_ip_prot_sock = pports;
144 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
146 kgid_t *data = table->data;
148 container_of(table->data, struct net, ipv4.ping_group_range.range);
151 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
155 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
158 /* Update system visible IP port range */
159 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
161 kgid_t *data = table->data;
163 container_of(table->data, struct net, ipv4.ping_group_range.range);
164 write_seqlock(&net->ipv4.ping_group_range.lock);
167 write_sequnlock(&net->ipv4.ping_group_range.lock);
170 /* Validate changes from /proc interface. */
171 static int ipv4_ping_group_range(struct ctl_table *table, int write,
172 void *buffer, 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,
207 void *buffer, size_t *lenp, loff_t *ppos)
212 net = container_of(table->data, struct net,
213 ipv4.sysctl_ip_fwd_update_priority);
214 ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
215 if (write && ret == 0)
216 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
222 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
223 void *buffer, size_t *lenp, loff_t *ppos)
225 struct net *net = container_of(ctl->data, struct net,
226 ipv4.tcp_congestion_control);
227 char val[TCP_CA_NAME_MAX];
228 struct ctl_table tbl = {
230 .maxlen = TCP_CA_NAME_MAX,
234 tcp_get_default_congestion_control(net, val);
236 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
237 if (write && ret == 0)
238 ret = tcp_set_default_congestion_control(net, val);
242 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
243 int write, void *buffer,
244 size_t *lenp, loff_t *ppos)
246 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
249 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
252 tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
253 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
258 static int proc_allowed_congestion_control(struct ctl_table *ctl,
259 int write, void *buffer,
260 size_t *lenp, loff_t *ppos)
262 struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
265 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
269 tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
270 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
271 if (write && ret == 0)
272 ret = tcp_set_allowed_congestion_control(tbl.data);
277 static int sscanf_key(char *buf, __le32 *key)
282 if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
283 user_key + 2, user_key + 3) != 4) {
286 for (i = 0; i < ARRAY_SIZE(user_key); i++)
287 key[i] = cpu_to_le32(user_key[i]);
289 pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
290 user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
295 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
296 void *buffer, size_t *lenp, loff_t *ppos)
298 struct net *net = container_of(table->data, struct net,
299 ipv4.sysctl_tcp_fastopen);
300 /* maxlen to print the list of keys in hex (*2), with dashes
301 * separating doublewords and a comma in between keys.
303 struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
304 2 * TCP_FASTOPEN_KEY_MAX) +
305 (TCP_FASTOPEN_KEY_MAX * 5)) };
306 u32 user_key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(u32)];
307 __le32 key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(__le32)];
309 int ret, i = 0, off = 0, n_keys;
311 tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
315 n_keys = tcp_fastopen_get_cipher(net, NULL, (u64 *)key);
317 memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
321 for (i = 0; i < n_keys * 4; i++)
322 user_key[i] = le32_to_cpu(key[i]);
324 for (i = 0; i < n_keys; i++) {
325 off += snprintf(tbl.data + off, tbl.maxlen - off,
326 "%08x-%08x-%08x-%08x",
330 user_key[i * 4 + 3]);
332 if (WARN_ON_ONCE(off >= tbl.maxlen - 1))
336 off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
339 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
341 if (write && ret == 0) {
342 backup_data = strchr(tbl.data, ',');
347 if (sscanf_key(tbl.data, key)) {
352 if (sscanf_key(backup_data, key + 4)) {
357 tcp_fastopen_reset_cipher(net, NULL, key,
358 backup_data ? key + 4 : NULL);
366 static void proc_configure_early_demux(int enabled, int protocol)
368 struct net_protocol *ipprot;
369 #if IS_ENABLED(CONFIG_IPV6)
370 struct inet6_protocol *ip6prot;
375 ipprot = rcu_dereference(inet_protos[protocol]);
377 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
380 #if IS_ENABLED(CONFIG_IPV6)
381 ip6prot = rcu_dereference(inet6_protos[protocol]);
383 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
389 static int proc_tcp_early_demux(struct ctl_table *table, int write,
390 void *buffer, size_t *lenp, loff_t *ppos)
394 ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
397 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
399 proc_configure_early_demux(enabled, IPPROTO_TCP);
405 static int proc_udp_early_demux(struct ctl_table *table, int write,
406 void *buffer, size_t *lenp, loff_t *ppos)
410 ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
413 int enabled = init_net.ipv4.sysctl_udp_early_demux;
415 proc_configure_early_demux(enabled, IPPROTO_UDP);
421 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
422 int write, void *buffer,
423 size_t *lenp, loff_t *ppos)
425 struct net *net = container_of(table->data, struct net,
426 ipv4.sysctl_tcp_fastopen_blackhole_timeout);
429 ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
430 if (write && ret == 0)
431 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
436 static int proc_tcp_available_ulp(struct ctl_table *ctl,
437 int write, void *buffer, size_t *lenp,
440 struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
443 tbl.data = kmalloc(tbl.maxlen, GFP_USER);
446 tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
447 ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
453 #ifdef CONFIG_IP_ROUTE_MULTIPATH
454 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
455 void *buffer, size_t *lenp,
458 struct net *net = container_of(table->data, struct net,
459 ipv4.sysctl_fib_multipath_hash_policy);
462 ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
463 if (write && ret == 0)
464 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
470 static struct ctl_table ipv4_table[] = {
472 .procname = "tcp_max_orphans",
473 .data = &sysctl_tcp_max_orphans,
474 .maxlen = sizeof(int),
476 .proc_handler = proc_dointvec
479 .procname = "inet_peer_threshold",
480 .data = &inet_peer_threshold,
481 .maxlen = sizeof(int),
483 .proc_handler = proc_dointvec
486 .procname = "inet_peer_minttl",
487 .data = &inet_peer_minttl,
488 .maxlen = sizeof(int),
490 .proc_handler = proc_dointvec_jiffies,
493 .procname = "inet_peer_maxttl",
494 .data = &inet_peer_maxttl,
495 .maxlen = sizeof(int),
497 .proc_handler = proc_dointvec_jiffies,
500 .procname = "tcp_mem",
501 .maxlen = sizeof(sysctl_tcp_mem),
502 .data = &sysctl_tcp_mem,
504 .proc_handler = proc_doulongvec_minmax,
507 .procname = "tcp_low_latency",
508 .data = &sysctl_tcp_low_latency,
509 .maxlen = sizeof(int),
511 .proc_handler = proc_dointvec
513 #ifdef CONFIG_NETLABEL
515 .procname = "cipso_cache_enable",
516 .data = &cipso_v4_cache_enabled,
517 .maxlen = sizeof(int),
519 .proc_handler = proc_dointvec,
522 .procname = "cipso_cache_bucket_size",
523 .data = &cipso_v4_cache_bucketsize,
524 .maxlen = sizeof(int),
526 .proc_handler = proc_dointvec,
529 .procname = "cipso_rbm_optfmt",
530 .data = &cipso_v4_rbm_optfmt,
531 .maxlen = sizeof(int),
533 .proc_handler = proc_dointvec,
536 .procname = "cipso_rbm_strictvalid",
537 .data = &cipso_v4_rbm_strictvalid,
538 .maxlen = sizeof(int),
540 .proc_handler = proc_dointvec,
542 #endif /* CONFIG_NETLABEL */
544 .procname = "tcp_available_ulp",
545 .maxlen = TCP_ULP_BUF_MAX,
547 .proc_handler = proc_tcp_available_ulp,
550 .procname = "icmp_msgs_per_sec",
551 .data = &sysctl_icmp_msgs_per_sec,
552 .maxlen = sizeof(int),
554 .proc_handler = proc_dointvec_minmax,
555 .extra1 = SYSCTL_ZERO,
558 .procname = "icmp_msgs_burst",
559 .data = &sysctl_icmp_msgs_burst,
560 .maxlen = sizeof(int),
562 .proc_handler = proc_dointvec_minmax,
563 .extra1 = SYSCTL_ZERO,
566 .procname = "udp_mem",
567 .data = &sysctl_udp_mem,
568 .maxlen = sizeof(sysctl_udp_mem),
570 .proc_handler = proc_doulongvec_minmax,
573 .procname = "fib_sync_mem",
574 .data = &sysctl_fib_sync_mem,
575 .maxlen = sizeof(sysctl_fib_sync_mem),
577 .proc_handler = proc_douintvec_minmax,
578 .extra1 = &sysctl_fib_sync_mem_min,
579 .extra2 = &sysctl_fib_sync_mem_max,
582 .procname = "tcp_rx_skb_cache",
583 .data = &tcp_rx_skb_cache_key.key,
585 .proc_handler = proc_do_static_key,
588 .procname = "tcp_tx_skb_cache",
589 .data = &tcp_tx_skb_cache_key.key,
591 .proc_handler = proc_do_static_key,
596 static struct ctl_table ipv4_net_table[] = {
598 .procname = "icmp_echo_ignore_all",
599 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
600 .maxlen = sizeof(u8),
602 .proc_handler = proc_dou8vec_minmax,
605 .procname = "icmp_echo_enable_probe",
606 .data = &init_net.ipv4.sysctl_icmp_echo_enable_probe,
607 .maxlen = sizeof(u8),
609 .proc_handler = proc_dou8vec_minmax,
610 .extra1 = SYSCTL_ZERO,
614 .procname = "icmp_echo_ignore_broadcasts",
615 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
616 .maxlen = sizeof(u8),
618 .proc_handler = proc_dou8vec_minmax,
621 .procname = "icmp_ignore_bogus_error_responses",
622 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
623 .maxlen = sizeof(u8),
625 .proc_handler = proc_dou8vec_minmax,
628 .procname = "icmp_errors_use_inbound_ifaddr",
629 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
630 .maxlen = sizeof(u8),
632 .proc_handler = proc_dou8vec_minmax,
635 .procname = "icmp_ratelimit",
636 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
637 .maxlen = sizeof(int),
639 .proc_handler = proc_dointvec_ms_jiffies,
642 .procname = "icmp_ratemask",
643 .data = &init_net.ipv4.sysctl_icmp_ratemask,
644 .maxlen = sizeof(int),
646 .proc_handler = proc_dointvec
649 .procname = "ping_group_range",
650 .data = &init_net.ipv4.ping_group_range.range,
651 .maxlen = sizeof(gid_t)*2,
653 .proc_handler = ipv4_ping_group_range,
655 #ifdef CONFIG_NET_L3_MASTER_DEV
657 .procname = "raw_l3mdev_accept",
658 .data = &init_net.ipv4.sysctl_raw_l3mdev_accept,
659 .maxlen = sizeof(u8),
661 .proc_handler = proc_dou8vec_minmax,
662 .extra1 = SYSCTL_ZERO,
663 .extra2 = SYSCTL_ONE,
667 .procname = "tcp_ecn",
668 .data = &init_net.ipv4.sysctl_tcp_ecn,
669 .maxlen = sizeof(u8),
671 .proc_handler = proc_dou8vec_minmax,
674 .procname = "tcp_ecn_fallback",
675 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
676 .maxlen = sizeof(u8),
678 .proc_handler = proc_dou8vec_minmax,
681 .procname = "ip_dynaddr",
682 .data = &init_net.ipv4.sysctl_ip_dynaddr,
683 .maxlen = sizeof(u8),
685 .proc_handler = proc_dou8vec_minmax,
688 .procname = "ip_early_demux",
689 .data = &init_net.ipv4.sysctl_ip_early_demux,
690 .maxlen = sizeof(u8),
692 .proc_handler = proc_dou8vec_minmax,
695 .procname = "udp_early_demux",
696 .data = &init_net.ipv4.sysctl_udp_early_demux,
697 .maxlen = sizeof(u8),
699 .proc_handler = proc_udp_early_demux
702 .procname = "tcp_early_demux",
703 .data = &init_net.ipv4.sysctl_tcp_early_demux,
704 .maxlen = sizeof(u8),
706 .proc_handler = proc_tcp_early_demux
709 .procname = "nexthop_compat_mode",
710 .data = &init_net.ipv4.sysctl_nexthop_compat_mode,
711 .maxlen = sizeof(u8),
713 .proc_handler = proc_dou8vec_minmax,
714 .extra1 = SYSCTL_ZERO,
715 .extra2 = SYSCTL_ONE,
718 .procname = "ip_default_ttl",
719 .data = &init_net.ipv4.sysctl_ip_default_ttl,
720 .maxlen = sizeof(u8),
722 .proc_handler = proc_dou8vec_minmax,
723 .extra1 = &ip_ttl_min,
724 .extra2 = &ip_ttl_max,
727 .procname = "ip_local_port_range",
728 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
729 .data = &init_net.ipv4.ip_local_ports.range,
731 .proc_handler = ipv4_local_port_range,
734 .procname = "ip_local_reserved_ports",
735 .data = &init_net.ipv4.sysctl_local_reserved_ports,
738 .proc_handler = proc_do_large_bitmap,
741 .procname = "ip_no_pmtu_disc",
742 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
743 .maxlen = sizeof(u8),
745 .proc_handler = proc_dou8vec_minmax,
748 .procname = "ip_forward_use_pmtu",
749 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
750 .maxlen = sizeof(u8),
752 .proc_handler = proc_dou8vec_minmax,
755 .procname = "ip_forward_update_priority",
756 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority,
757 .maxlen = sizeof(u8),
759 .proc_handler = ipv4_fwd_update_priority,
760 .extra1 = SYSCTL_ZERO,
761 .extra2 = SYSCTL_ONE,
764 .procname = "ip_nonlocal_bind",
765 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
766 .maxlen = sizeof(u8),
768 .proc_handler = proc_dou8vec_minmax,
771 .procname = "ip_autobind_reuse",
772 .data = &init_net.ipv4.sysctl_ip_autobind_reuse,
773 .maxlen = sizeof(u8),
775 .proc_handler = proc_dou8vec_minmax,
776 .extra1 = SYSCTL_ZERO,
777 .extra2 = SYSCTL_ONE,
780 .procname = "fwmark_reflect",
781 .data = &init_net.ipv4.sysctl_fwmark_reflect,
782 .maxlen = sizeof(u8),
784 .proc_handler = proc_dou8vec_minmax,
787 .procname = "tcp_fwmark_accept",
788 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept,
789 .maxlen = sizeof(u8),
791 .proc_handler = proc_dou8vec_minmax,
793 #ifdef CONFIG_NET_L3_MASTER_DEV
795 .procname = "tcp_l3mdev_accept",
796 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
797 .maxlen = sizeof(u8),
799 .proc_handler = proc_dou8vec_minmax,
800 .extra1 = SYSCTL_ZERO,
801 .extra2 = SYSCTL_ONE,
805 .procname = "tcp_mtu_probing",
806 .data = &init_net.ipv4.sysctl_tcp_mtu_probing,
807 .maxlen = sizeof(u8),
809 .proc_handler = proc_dou8vec_minmax,
812 .procname = "tcp_base_mss",
813 .data = &init_net.ipv4.sysctl_tcp_base_mss,
814 .maxlen = sizeof(int),
816 .proc_handler = proc_dointvec,
819 .procname = "tcp_min_snd_mss",
820 .data = &init_net.ipv4.sysctl_tcp_min_snd_mss,
821 .maxlen = sizeof(int),
823 .proc_handler = proc_dointvec_minmax,
824 .extra1 = &tcp_min_snd_mss_min,
825 .extra2 = &tcp_min_snd_mss_max,
828 .procname = "tcp_mtu_probe_floor",
829 .data = &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
830 .maxlen = sizeof(int),
832 .proc_handler = proc_dointvec_minmax,
833 .extra1 = &tcp_min_snd_mss_min,
834 .extra2 = &tcp_min_snd_mss_max,
837 .procname = "tcp_probe_threshold",
838 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
839 .maxlen = sizeof(int),
841 .proc_handler = proc_dointvec,
844 .procname = "tcp_probe_interval",
845 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
846 .maxlen = sizeof(u32),
848 .proc_handler = proc_douintvec_minmax,
849 .extra2 = &u32_max_div_HZ,
852 .procname = "igmp_link_local_mcast_reports",
853 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
854 .maxlen = sizeof(u8),
856 .proc_handler = proc_dou8vec_minmax,
859 .procname = "igmp_max_memberships",
860 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
861 .maxlen = sizeof(int),
863 .proc_handler = proc_dointvec
866 .procname = "igmp_max_msf",
867 .data = &init_net.ipv4.sysctl_igmp_max_msf,
868 .maxlen = sizeof(int),
870 .proc_handler = proc_dointvec
872 #ifdef CONFIG_IP_MULTICAST
874 .procname = "igmp_qrv",
875 .data = &init_net.ipv4.sysctl_igmp_qrv,
876 .maxlen = sizeof(int),
878 .proc_handler = proc_dointvec_minmax,
883 .procname = "tcp_congestion_control",
884 .data = &init_net.ipv4.tcp_congestion_control,
886 .maxlen = TCP_CA_NAME_MAX,
887 .proc_handler = proc_tcp_congestion_control,
890 .procname = "tcp_available_congestion_control",
891 .maxlen = TCP_CA_BUF_MAX,
893 .proc_handler = proc_tcp_available_congestion_control,
896 .procname = "tcp_allowed_congestion_control",
897 .maxlen = TCP_CA_BUF_MAX,
899 .proc_handler = proc_allowed_congestion_control,
902 .procname = "tcp_keepalive_time",
903 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
904 .maxlen = sizeof(int),
906 .proc_handler = proc_dointvec_jiffies,
909 .procname = "tcp_keepalive_probes",
910 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
911 .maxlen = sizeof(u8),
913 .proc_handler = proc_dou8vec_minmax,
916 .procname = "tcp_keepalive_intvl",
917 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
918 .maxlen = sizeof(int),
920 .proc_handler = proc_dointvec_jiffies,
923 .procname = "tcp_syn_retries",
924 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
925 .maxlen = sizeof(u8),
927 .proc_handler = proc_dou8vec_minmax,
928 .extra1 = &tcp_syn_retries_min,
929 .extra2 = &tcp_syn_retries_max
932 .procname = "tcp_synack_retries",
933 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
934 .maxlen = sizeof(u8),
936 .proc_handler = proc_dou8vec_minmax,
938 #ifdef CONFIG_SYN_COOKIES
940 .procname = "tcp_syncookies",
941 .data = &init_net.ipv4.sysctl_tcp_syncookies,
942 .maxlen = sizeof(u8),
944 .proc_handler = proc_dou8vec_minmax,
948 .procname = "tcp_reordering",
949 .data = &init_net.ipv4.sysctl_tcp_reordering,
950 .maxlen = sizeof(int),
952 .proc_handler = proc_dointvec
955 .procname = "tcp_retries1",
956 .data = &init_net.ipv4.sysctl_tcp_retries1,
957 .maxlen = sizeof(u8),
959 .proc_handler = proc_dou8vec_minmax,
960 .extra2 = &tcp_retr1_max
963 .procname = "tcp_retries2",
964 .data = &init_net.ipv4.sysctl_tcp_retries2,
965 .maxlen = sizeof(u8),
967 .proc_handler = proc_dou8vec_minmax,
970 .procname = "tcp_orphan_retries",
971 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
972 .maxlen = sizeof(u8),
974 .proc_handler = proc_dou8vec_minmax,
977 .procname = "tcp_fin_timeout",
978 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
979 .maxlen = sizeof(int),
981 .proc_handler = proc_dointvec_jiffies,
984 .procname = "tcp_notsent_lowat",
985 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
986 .maxlen = sizeof(unsigned int),
988 .proc_handler = proc_douintvec,
991 .procname = "tcp_tw_reuse",
992 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
993 .maxlen = sizeof(u8),
995 .proc_handler = proc_dou8vec_minmax,
996 .extra1 = SYSCTL_ZERO,
1000 .procname = "tcp_max_tw_buckets",
1001 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
1002 .maxlen = sizeof(int),
1004 .proc_handler = proc_dointvec
1007 .procname = "tcp_max_syn_backlog",
1008 .data = &init_net.ipv4.sysctl_max_syn_backlog,
1009 .maxlen = sizeof(int),
1011 .proc_handler = proc_dointvec
1014 .procname = "tcp_fastopen",
1015 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1016 .maxlen = sizeof(int),
1018 .proc_handler = proc_dointvec,
1021 .procname = "tcp_fastopen_key",
1023 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1024 /* maxlen to print the list of keys in hex (*2), with dashes
1025 * separating doublewords and a comma in between keys.
1027 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
1028 2 * TCP_FASTOPEN_KEY_MAX) +
1029 (TCP_FASTOPEN_KEY_MAX * 5)),
1030 .proc_handler = proc_tcp_fastopen_key,
1033 .procname = "tcp_fastopen_blackhole_timeout_sec",
1034 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1035 .maxlen = sizeof(int),
1037 .proc_handler = proc_tfo_blackhole_detect_timeout,
1038 .extra1 = SYSCTL_ZERO,
1040 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1042 .procname = "fib_multipath_use_neigh",
1043 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1044 .maxlen = sizeof(u8),
1046 .proc_handler = proc_dou8vec_minmax,
1047 .extra1 = SYSCTL_ZERO,
1048 .extra2 = SYSCTL_ONE,
1051 .procname = "fib_multipath_hash_policy",
1052 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1053 .maxlen = sizeof(u8),
1055 .proc_handler = proc_fib_multipath_hash_policy,
1056 .extra1 = SYSCTL_ZERO,
1060 .procname = "fib_multipath_hash_fields",
1061 .data = &init_net.ipv4.sysctl_fib_multipath_hash_fields,
1062 .maxlen = sizeof(u32),
1064 .proc_handler = proc_douintvec_minmax,
1065 .extra1 = SYSCTL_ONE,
1066 .extra2 = &fib_multipath_hash_fields_all_mask,
1070 .procname = "ip_unprivileged_port_start",
1071 .maxlen = sizeof(int),
1072 .data = &init_net.ipv4.sysctl_ip_prot_sock,
1074 .proc_handler = ipv4_privileged_ports,
1076 #ifdef CONFIG_NET_L3_MASTER_DEV
1078 .procname = "udp_l3mdev_accept",
1079 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1080 .maxlen = sizeof(u8),
1082 .proc_handler = proc_dou8vec_minmax,
1083 .extra1 = SYSCTL_ZERO,
1084 .extra2 = SYSCTL_ONE,
1088 .procname = "tcp_sack",
1089 .data = &init_net.ipv4.sysctl_tcp_sack,
1090 .maxlen = sizeof(u8),
1092 .proc_handler = proc_dou8vec_minmax,
1095 .procname = "tcp_window_scaling",
1096 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
1097 .maxlen = sizeof(u8),
1099 .proc_handler = proc_dou8vec_minmax,
1102 .procname = "tcp_timestamps",
1103 .data = &init_net.ipv4.sysctl_tcp_timestamps,
1104 .maxlen = sizeof(u8),
1106 .proc_handler = proc_dou8vec_minmax,
1109 .procname = "tcp_early_retrans",
1110 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
1111 .maxlen = sizeof(u8),
1113 .proc_handler = proc_dou8vec_minmax,
1114 .extra1 = SYSCTL_ZERO,
1118 .procname = "tcp_recovery",
1119 .data = &init_net.ipv4.sysctl_tcp_recovery,
1120 .maxlen = sizeof(u8),
1122 .proc_handler = proc_dou8vec_minmax,
1125 .procname = "tcp_thin_linear_timeouts",
1126 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1127 .maxlen = sizeof(u8),
1129 .proc_handler = proc_dou8vec_minmax,
1132 .procname = "tcp_slow_start_after_idle",
1133 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1134 .maxlen = sizeof(u8),
1136 .proc_handler = proc_dou8vec_minmax,
1139 .procname = "tcp_retrans_collapse",
1140 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1141 .maxlen = sizeof(u8),
1143 .proc_handler = proc_dou8vec_minmax,
1146 .procname = "tcp_stdurg",
1147 .data = &init_net.ipv4.sysctl_tcp_stdurg,
1148 .maxlen = sizeof(u8),
1150 .proc_handler = proc_dou8vec_minmax,
1153 .procname = "tcp_rfc1337",
1154 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
1155 .maxlen = sizeof(u8),
1157 .proc_handler = proc_dou8vec_minmax,
1160 .procname = "tcp_abort_on_overflow",
1161 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1162 .maxlen = sizeof(u8),
1164 .proc_handler = proc_dou8vec_minmax,
1167 .procname = "tcp_fack",
1168 .data = &init_net.ipv4.sysctl_tcp_fack,
1169 .maxlen = sizeof(u8),
1171 .proc_handler = proc_dou8vec_minmax,
1174 .procname = "tcp_max_reordering",
1175 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1176 .maxlen = sizeof(int),
1178 .proc_handler = proc_dointvec
1181 .procname = "tcp_dsack",
1182 .data = &init_net.ipv4.sysctl_tcp_dsack,
1183 .maxlen = sizeof(u8),
1185 .proc_handler = proc_dou8vec_minmax,
1188 .procname = "tcp_app_win",
1189 .data = &init_net.ipv4.sysctl_tcp_app_win,
1190 .maxlen = sizeof(u8),
1192 .proc_handler = proc_dou8vec_minmax,
1195 .procname = "tcp_adv_win_scale",
1196 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1197 .maxlen = sizeof(int),
1199 .proc_handler = proc_dointvec_minmax,
1200 .extra1 = &tcp_adv_win_scale_min,
1201 .extra2 = &tcp_adv_win_scale_max,
1204 .procname = "tcp_frto",
1205 .data = &init_net.ipv4.sysctl_tcp_frto,
1206 .maxlen = sizeof(u8),
1208 .proc_handler = proc_dou8vec_minmax,
1211 .procname = "tcp_no_metrics_save",
1212 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1213 .maxlen = sizeof(u8),
1215 .proc_handler = proc_dou8vec_minmax,
1218 .procname = "tcp_no_ssthresh_metrics_save",
1219 .data = &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1220 .maxlen = sizeof(u8),
1222 .proc_handler = proc_dou8vec_minmax,
1223 .extra1 = SYSCTL_ZERO,
1224 .extra2 = SYSCTL_ONE,
1227 .procname = "tcp_moderate_rcvbuf",
1228 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1229 .maxlen = sizeof(u8),
1231 .proc_handler = proc_dou8vec_minmax,
1234 .procname = "tcp_tso_win_divisor",
1235 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1236 .maxlen = sizeof(u8),
1238 .proc_handler = proc_dou8vec_minmax,
1241 .procname = "tcp_workaround_signed_windows",
1242 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1243 .maxlen = sizeof(u8),
1245 .proc_handler = proc_dou8vec_minmax,
1248 .procname = "tcp_limit_output_bytes",
1249 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1250 .maxlen = sizeof(int),
1252 .proc_handler = proc_dointvec
1255 .procname = "tcp_challenge_ack_limit",
1256 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1257 .maxlen = sizeof(int),
1259 .proc_handler = proc_dointvec
1262 .procname = "tcp_min_tso_segs",
1263 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1264 .maxlen = sizeof(u8),
1266 .proc_handler = proc_dou8vec_minmax,
1267 .extra1 = SYSCTL_ONE,
1270 .procname = "tcp_min_rtt_wlen",
1271 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1272 .maxlen = sizeof(int),
1274 .proc_handler = proc_dointvec_minmax,
1275 .extra1 = SYSCTL_ZERO,
1276 .extra2 = &one_day_secs
1279 .procname = "tcp_autocorking",
1280 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1281 .maxlen = sizeof(u8),
1283 .proc_handler = proc_dou8vec_minmax,
1284 .extra1 = SYSCTL_ZERO,
1285 .extra2 = SYSCTL_ONE,
1288 .procname = "tcp_invalid_ratelimit",
1289 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1290 .maxlen = sizeof(int),
1292 .proc_handler = proc_dointvec_ms_jiffies,
1295 .procname = "tcp_pacing_ss_ratio",
1296 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1297 .maxlen = sizeof(int),
1299 .proc_handler = proc_dointvec_minmax,
1300 .extra1 = SYSCTL_ZERO,
1301 .extra2 = &thousand,
1304 .procname = "tcp_pacing_ca_ratio",
1305 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1306 .maxlen = sizeof(int),
1308 .proc_handler = proc_dointvec_minmax,
1309 .extra1 = SYSCTL_ZERO,
1310 .extra2 = &thousand,
1313 .procname = "tcp_wmem",
1314 .data = &init_net.ipv4.sysctl_tcp_wmem,
1315 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1317 .proc_handler = proc_dointvec_minmax,
1318 .extra1 = SYSCTL_ONE,
1321 .procname = "tcp_rmem",
1322 .data = &init_net.ipv4.sysctl_tcp_rmem,
1323 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1325 .proc_handler = proc_dointvec_minmax,
1326 .extra1 = SYSCTL_ONE,
1329 .procname = "tcp_comp_sack_delay_ns",
1330 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1331 .maxlen = sizeof(unsigned long),
1333 .proc_handler = proc_doulongvec_minmax,
1336 .procname = "tcp_comp_sack_slack_ns",
1337 .data = &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1338 .maxlen = sizeof(unsigned long),
1340 .proc_handler = proc_doulongvec_minmax,
1343 .procname = "tcp_comp_sack_nr",
1344 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1345 .maxlen = sizeof(u8),
1347 .proc_handler = proc_dou8vec_minmax,
1348 .extra1 = SYSCTL_ZERO,
1351 .procname = "tcp_reflect_tos",
1352 .data = &init_net.ipv4.sysctl_tcp_reflect_tos,
1353 .maxlen = sizeof(u8),
1355 .proc_handler = proc_dou8vec_minmax,
1356 .extra1 = SYSCTL_ZERO,
1357 .extra2 = SYSCTL_ONE,
1360 .procname = "udp_rmem_min",
1361 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1362 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1364 .proc_handler = proc_dointvec_minmax,
1365 .extra1 = SYSCTL_ONE
1368 .procname = "udp_wmem_min",
1369 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1370 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1372 .proc_handler = proc_dointvec_minmax,
1373 .extra1 = SYSCTL_ONE
1376 .procname = "fib_notify_on_flag_change",
1377 .data = &init_net.ipv4.sysctl_fib_notify_on_flag_change,
1378 .maxlen = sizeof(u8),
1380 .proc_handler = proc_dou8vec_minmax,
1381 .extra1 = SYSCTL_ZERO,
1387 static __net_init int ipv4_sysctl_init_net(struct net *net)
1389 struct ctl_table *table;
1391 table = ipv4_net_table;
1392 if (!net_eq(net, &init_net)) {
1395 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1399 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++) {
1400 if (table[i].data) {
1401 /* Update the variables to point into
1402 * the current struct net
1404 table[i].data += (void *)net - (void *)&init_net;
1406 /* Entries without data pointer are global;
1407 * Make them read-only in non-init_net ns
1409 table[i].mode &= ~0222;
1414 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1415 if (!net->ipv4.ipv4_hdr)
1418 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1419 if (!net->ipv4.sysctl_local_reserved_ports)
1425 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1427 if (!net_eq(net, &init_net))
1433 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1435 struct ctl_table *table;
1437 kfree(net->ipv4.sysctl_local_reserved_ports);
1438 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1439 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1443 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1444 .init = ipv4_sysctl_init_net,
1445 .exit = ipv4_sysctl_exit_net,
1448 static __init int sysctl_ipv4_init(void)
1450 struct ctl_table_header *hdr;
1452 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1456 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1457 unregister_net_sysctl_table(hdr);
1464 __initcall(sysctl_ipv4_init);