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] < READ_ONCE(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 = READ_ONCE(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 WRITE_ONCE(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);
469 static int proc_fib_multipath_hash_fields(struct ctl_table *table, int write,
470 void *buffer, size_t *lenp,
476 net = container_of(table->data, struct net,
477 ipv4.sysctl_fib_multipath_hash_fields);
478 ret = proc_douintvec_minmax(table, write, buffer, lenp, ppos);
479 if (write && ret == 0)
480 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
486 static struct ctl_table ipv4_table[] = {
488 .procname = "tcp_max_orphans",
489 .data = &sysctl_tcp_max_orphans,
490 .maxlen = sizeof(int),
492 .proc_handler = proc_dointvec
495 .procname = "inet_peer_threshold",
496 .data = &inet_peer_threshold,
497 .maxlen = sizeof(int),
499 .proc_handler = proc_dointvec
502 .procname = "inet_peer_minttl",
503 .data = &inet_peer_minttl,
504 .maxlen = sizeof(int),
506 .proc_handler = proc_dointvec_jiffies,
509 .procname = "inet_peer_maxttl",
510 .data = &inet_peer_maxttl,
511 .maxlen = sizeof(int),
513 .proc_handler = proc_dointvec_jiffies,
516 .procname = "tcp_mem",
517 .maxlen = sizeof(sysctl_tcp_mem),
518 .data = &sysctl_tcp_mem,
520 .proc_handler = proc_doulongvec_minmax,
523 .procname = "tcp_low_latency",
524 .data = &sysctl_tcp_low_latency,
525 .maxlen = sizeof(int),
527 .proc_handler = proc_dointvec
529 #ifdef CONFIG_NETLABEL
531 .procname = "cipso_cache_enable",
532 .data = &cipso_v4_cache_enabled,
533 .maxlen = sizeof(int),
535 .proc_handler = proc_dointvec,
538 .procname = "cipso_cache_bucket_size",
539 .data = &cipso_v4_cache_bucketsize,
540 .maxlen = sizeof(int),
542 .proc_handler = proc_dointvec,
545 .procname = "cipso_rbm_optfmt",
546 .data = &cipso_v4_rbm_optfmt,
547 .maxlen = sizeof(int),
549 .proc_handler = proc_dointvec,
552 .procname = "cipso_rbm_strictvalid",
553 .data = &cipso_v4_rbm_strictvalid,
554 .maxlen = sizeof(int),
556 .proc_handler = proc_dointvec,
558 #endif /* CONFIG_NETLABEL */
560 .procname = "tcp_available_ulp",
561 .maxlen = TCP_ULP_BUF_MAX,
563 .proc_handler = proc_tcp_available_ulp,
566 .procname = "icmp_msgs_per_sec",
567 .data = &sysctl_icmp_msgs_per_sec,
568 .maxlen = sizeof(int),
570 .proc_handler = proc_dointvec_minmax,
571 .extra1 = SYSCTL_ZERO,
574 .procname = "icmp_msgs_burst",
575 .data = &sysctl_icmp_msgs_burst,
576 .maxlen = sizeof(int),
578 .proc_handler = proc_dointvec_minmax,
579 .extra1 = SYSCTL_ZERO,
582 .procname = "udp_mem",
583 .data = &sysctl_udp_mem,
584 .maxlen = sizeof(sysctl_udp_mem),
586 .proc_handler = proc_doulongvec_minmax,
589 .procname = "fib_sync_mem",
590 .data = &sysctl_fib_sync_mem,
591 .maxlen = sizeof(sysctl_fib_sync_mem),
593 .proc_handler = proc_douintvec_minmax,
594 .extra1 = &sysctl_fib_sync_mem_min,
595 .extra2 = &sysctl_fib_sync_mem_max,
598 .procname = "tcp_rx_skb_cache",
599 .data = &tcp_rx_skb_cache_key.key,
601 .proc_handler = proc_do_static_key,
604 .procname = "tcp_tx_skb_cache",
605 .data = &tcp_tx_skb_cache_key.key,
607 .proc_handler = proc_do_static_key,
612 static struct ctl_table ipv4_net_table[] = {
614 .procname = "icmp_echo_ignore_all",
615 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
616 .maxlen = sizeof(u8),
618 .proc_handler = proc_dou8vec_minmax,
621 .procname = "icmp_echo_enable_probe",
622 .data = &init_net.ipv4.sysctl_icmp_echo_enable_probe,
623 .maxlen = sizeof(u8),
625 .proc_handler = proc_dou8vec_minmax,
626 .extra1 = SYSCTL_ZERO,
630 .procname = "icmp_echo_ignore_broadcasts",
631 .data = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
632 .maxlen = sizeof(u8),
634 .proc_handler = proc_dou8vec_minmax,
637 .procname = "icmp_ignore_bogus_error_responses",
638 .data = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
639 .maxlen = sizeof(u8),
641 .proc_handler = proc_dou8vec_minmax,
642 .extra1 = SYSCTL_ZERO,
646 .procname = "icmp_errors_use_inbound_ifaddr",
647 .data = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
648 .maxlen = sizeof(u8),
650 .proc_handler = proc_dou8vec_minmax,
651 .extra1 = SYSCTL_ZERO,
655 .procname = "icmp_ratelimit",
656 .data = &init_net.ipv4.sysctl_icmp_ratelimit,
657 .maxlen = sizeof(int),
659 .proc_handler = proc_dointvec_ms_jiffies,
662 .procname = "icmp_ratemask",
663 .data = &init_net.ipv4.sysctl_icmp_ratemask,
664 .maxlen = sizeof(int),
666 .proc_handler = proc_dointvec
669 .procname = "ping_group_range",
670 .data = &init_net.ipv4.ping_group_range.range,
671 .maxlen = sizeof(gid_t)*2,
673 .proc_handler = ipv4_ping_group_range,
675 #ifdef CONFIG_NET_L3_MASTER_DEV
677 .procname = "raw_l3mdev_accept",
678 .data = &init_net.ipv4.sysctl_raw_l3mdev_accept,
679 .maxlen = sizeof(u8),
681 .proc_handler = proc_dou8vec_minmax,
682 .extra1 = SYSCTL_ZERO,
683 .extra2 = SYSCTL_ONE,
687 .procname = "tcp_ecn",
688 .data = &init_net.ipv4.sysctl_tcp_ecn,
689 .maxlen = sizeof(u8),
691 .proc_handler = proc_dou8vec_minmax,
692 .extra1 = SYSCTL_ZERO,
693 .extra2 = SYSCTL_TWO,
696 .procname = "tcp_ecn_fallback",
697 .data = &init_net.ipv4.sysctl_tcp_ecn_fallback,
698 .maxlen = sizeof(u8),
700 .proc_handler = proc_dou8vec_minmax,
701 .extra1 = SYSCTL_ZERO,
702 .extra2 = SYSCTL_ONE,
705 .procname = "ip_dynaddr",
706 .data = &init_net.ipv4.sysctl_ip_dynaddr,
707 .maxlen = sizeof(u8),
709 .proc_handler = proc_dou8vec_minmax,
712 .procname = "ip_early_demux",
713 .data = &init_net.ipv4.sysctl_ip_early_demux,
714 .maxlen = sizeof(u8),
716 .proc_handler = proc_dou8vec_minmax,
719 .procname = "udp_early_demux",
720 .data = &init_net.ipv4.sysctl_udp_early_demux,
721 .maxlen = sizeof(u8),
723 .proc_handler = proc_udp_early_demux
726 .procname = "tcp_early_demux",
727 .data = &init_net.ipv4.sysctl_tcp_early_demux,
728 .maxlen = sizeof(u8),
730 .proc_handler = proc_tcp_early_demux
733 .procname = "nexthop_compat_mode",
734 .data = &init_net.ipv4.sysctl_nexthop_compat_mode,
735 .maxlen = sizeof(u8),
737 .proc_handler = proc_dou8vec_minmax,
738 .extra1 = SYSCTL_ZERO,
739 .extra2 = SYSCTL_ONE,
742 .procname = "ip_default_ttl",
743 .data = &init_net.ipv4.sysctl_ip_default_ttl,
744 .maxlen = sizeof(u8),
746 .proc_handler = proc_dou8vec_minmax,
747 .extra1 = &ip_ttl_min,
748 .extra2 = &ip_ttl_max,
751 .procname = "ip_local_port_range",
752 .maxlen = sizeof(init_net.ipv4.ip_local_ports.range),
753 .data = &init_net.ipv4.ip_local_ports.range,
755 .proc_handler = ipv4_local_port_range,
758 .procname = "ip_local_reserved_ports",
759 .data = &init_net.ipv4.sysctl_local_reserved_ports,
762 .proc_handler = proc_do_large_bitmap,
765 .procname = "ip_no_pmtu_disc",
766 .data = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
767 .maxlen = sizeof(u8),
769 .proc_handler = proc_dou8vec_minmax,
772 .procname = "ip_forward_use_pmtu",
773 .data = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
774 .maxlen = sizeof(u8),
776 .proc_handler = proc_dou8vec_minmax,
779 .procname = "ip_forward_update_priority",
780 .data = &init_net.ipv4.sysctl_ip_fwd_update_priority,
781 .maxlen = sizeof(u8),
783 .proc_handler = ipv4_fwd_update_priority,
784 .extra1 = SYSCTL_ZERO,
785 .extra2 = SYSCTL_ONE,
788 .procname = "ip_nonlocal_bind",
789 .data = &init_net.ipv4.sysctl_ip_nonlocal_bind,
790 .maxlen = sizeof(u8),
792 .proc_handler = proc_dou8vec_minmax,
795 .procname = "ip_autobind_reuse",
796 .data = &init_net.ipv4.sysctl_ip_autobind_reuse,
797 .maxlen = sizeof(u8),
799 .proc_handler = proc_dou8vec_minmax,
800 .extra1 = SYSCTL_ZERO,
801 .extra2 = SYSCTL_ONE,
804 .procname = "fwmark_reflect",
805 .data = &init_net.ipv4.sysctl_fwmark_reflect,
806 .maxlen = sizeof(u8),
808 .proc_handler = proc_dou8vec_minmax,
811 .procname = "tcp_fwmark_accept",
812 .data = &init_net.ipv4.sysctl_tcp_fwmark_accept,
813 .maxlen = sizeof(u8),
815 .proc_handler = proc_dou8vec_minmax,
817 #ifdef CONFIG_NET_L3_MASTER_DEV
819 .procname = "tcp_l3mdev_accept",
820 .data = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
821 .maxlen = sizeof(u8),
823 .proc_handler = proc_dou8vec_minmax,
824 .extra1 = SYSCTL_ZERO,
825 .extra2 = SYSCTL_ONE,
829 .procname = "tcp_mtu_probing",
830 .data = &init_net.ipv4.sysctl_tcp_mtu_probing,
831 .maxlen = sizeof(u8),
833 .proc_handler = proc_dou8vec_minmax,
836 .procname = "tcp_base_mss",
837 .data = &init_net.ipv4.sysctl_tcp_base_mss,
838 .maxlen = sizeof(int),
840 .proc_handler = proc_dointvec,
843 .procname = "tcp_min_snd_mss",
844 .data = &init_net.ipv4.sysctl_tcp_min_snd_mss,
845 .maxlen = sizeof(int),
847 .proc_handler = proc_dointvec_minmax,
848 .extra1 = &tcp_min_snd_mss_min,
849 .extra2 = &tcp_min_snd_mss_max,
852 .procname = "tcp_mtu_probe_floor",
853 .data = &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
854 .maxlen = sizeof(int),
856 .proc_handler = proc_dointvec_minmax,
857 .extra1 = &tcp_min_snd_mss_min,
858 .extra2 = &tcp_min_snd_mss_max,
861 .procname = "tcp_probe_threshold",
862 .data = &init_net.ipv4.sysctl_tcp_probe_threshold,
863 .maxlen = sizeof(int),
865 .proc_handler = proc_dointvec,
868 .procname = "tcp_probe_interval",
869 .data = &init_net.ipv4.sysctl_tcp_probe_interval,
870 .maxlen = sizeof(u32),
872 .proc_handler = proc_douintvec_minmax,
873 .extra2 = &u32_max_div_HZ,
876 .procname = "igmp_link_local_mcast_reports",
877 .data = &init_net.ipv4.sysctl_igmp_llm_reports,
878 .maxlen = sizeof(u8),
880 .proc_handler = proc_dou8vec_minmax,
883 .procname = "igmp_max_memberships",
884 .data = &init_net.ipv4.sysctl_igmp_max_memberships,
885 .maxlen = sizeof(int),
887 .proc_handler = proc_dointvec
890 .procname = "igmp_max_msf",
891 .data = &init_net.ipv4.sysctl_igmp_max_msf,
892 .maxlen = sizeof(int),
894 .proc_handler = proc_dointvec
896 #ifdef CONFIG_IP_MULTICAST
898 .procname = "igmp_qrv",
899 .data = &init_net.ipv4.sysctl_igmp_qrv,
900 .maxlen = sizeof(int),
902 .proc_handler = proc_dointvec_minmax,
907 .procname = "tcp_congestion_control",
908 .data = &init_net.ipv4.tcp_congestion_control,
910 .maxlen = TCP_CA_NAME_MAX,
911 .proc_handler = proc_tcp_congestion_control,
914 .procname = "tcp_available_congestion_control",
915 .maxlen = TCP_CA_BUF_MAX,
917 .proc_handler = proc_tcp_available_congestion_control,
920 .procname = "tcp_allowed_congestion_control",
921 .maxlen = TCP_CA_BUF_MAX,
923 .proc_handler = proc_allowed_congestion_control,
926 .procname = "tcp_keepalive_time",
927 .data = &init_net.ipv4.sysctl_tcp_keepalive_time,
928 .maxlen = sizeof(int),
930 .proc_handler = proc_dointvec_jiffies,
933 .procname = "tcp_keepalive_probes",
934 .data = &init_net.ipv4.sysctl_tcp_keepalive_probes,
935 .maxlen = sizeof(u8),
937 .proc_handler = proc_dou8vec_minmax,
940 .procname = "tcp_keepalive_intvl",
941 .data = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
942 .maxlen = sizeof(int),
944 .proc_handler = proc_dointvec_jiffies,
947 .procname = "tcp_syn_retries",
948 .data = &init_net.ipv4.sysctl_tcp_syn_retries,
949 .maxlen = sizeof(u8),
951 .proc_handler = proc_dou8vec_minmax,
952 .extra1 = &tcp_syn_retries_min,
953 .extra2 = &tcp_syn_retries_max
956 .procname = "tcp_synack_retries",
957 .data = &init_net.ipv4.sysctl_tcp_synack_retries,
958 .maxlen = sizeof(u8),
960 .proc_handler = proc_dou8vec_minmax,
962 #ifdef CONFIG_SYN_COOKIES
964 .procname = "tcp_syncookies",
965 .data = &init_net.ipv4.sysctl_tcp_syncookies,
966 .maxlen = sizeof(u8),
968 .proc_handler = proc_dou8vec_minmax,
972 .procname = "tcp_migrate_req",
973 .data = &init_net.ipv4.sysctl_tcp_migrate_req,
974 .maxlen = sizeof(u8),
976 .proc_handler = proc_dou8vec_minmax,
977 .extra1 = SYSCTL_ZERO,
981 .procname = "tcp_reordering",
982 .data = &init_net.ipv4.sysctl_tcp_reordering,
983 .maxlen = sizeof(int),
985 .proc_handler = proc_dointvec
988 .procname = "tcp_retries1",
989 .data = &init_net.ipv4.sysctl_tcp_retries1,
990 .maxlen = sizeof(u8),
992 .proc_handler = proc_dou8vec_minmax,
993 .extra2 = &tcp_retr1_max
996 .procname = "tcp_retries2",
997 .data = &init_net.ipv4.sysctl_tcp_retries2,
998 .maxlen = sizeof(u8),
1000 .proc_handler = proc_dou8vec_minmax,
1003 .procname = "tcp_orphan_retries",
1004 .data = &init_net.ipv4.sysctl_tcp_orphan_retries,
1005 .maxlen = sizeof(u8),
1007 .proc_handler = proc_dou8vec_minmax,
1010 .procname = "tcp_fin_timeout",
1011 .data = &init_net.ipv4.sysctl_tcp_fin_timeout,
1012 .maxlen = sizeof(int),
1014 .proc_handler = proc_dointvec_jiffies,
1017 .procname = "tcp_notsent_lowat",
1018 .data = &init_net.ipv4.sysctl_tcp_notsent_lowat,
1019 .maxlen = sizeof(unsigned int),
1021 .proc_handler = proc_douintvec,
1024 .procname = "tcp_tw_reuse",
1025 .data = &init_net.ipv4.sysctl_tcp_tw_reuse,
1026 .maxlen = sizeof(u8),
1028 .proc_handler = proc_dou8vec_minmax,
1029 .extra1 = SYSCTL_ZERO,
1033 .procname = "tcp_max_tw_buckets",
1034 .data = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
1035 .maxlen = sizeof(int),
1037 .proc_handler = proc_dointvec
1040 .procname = "tcp_max_syn_backlog",
1041 .data = &init_net.ipv4.sysctl_max_syn_backlog,
1042 .maxlen = sizeof(int),
1044 .proc_handler = proc_dointvec
1047 .procname = "tcp_fastopen",
1048 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1049 .maxlen = sizeof(int),
1051 .proc_handler = proc_dointvec,
1054 .procname = "tcp_fastopen_key",
1056 .data = &init_net.ipv4.sysctl_tcp_fastopen,
1057 /* maxlen to print the list of keys in hex (*2), with dashes
1058 * separating doublewords and a comma in between keys.
1060 .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
1061 2 * TCP_FASTOPEN_KEY_MAX) +
1062 (TCP_FASTOPEN_KEY_MAX * 5)),
1063 .proc_handler = proc_tcp_fastopen_key,
1066 .procname = "tcp_fastopen_blackhole_timeout_sec",
1067 .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1068 .maxlen = sizeof(int),
1070 .proc_handler = proc_tfo_blackhole_detect_timeout,
1071 .extra1 = SYSCTL_ZERO,
1073 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1075 .procname = "fib_multipath_use_neigh",
1076 .data = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1077 .maxlen = sizeof(u8),
1079 .proc_handler = proc_dou8vec_minmax,
1080 .extra1 = SYSCTL_ZERO,
1081 .extra2 = SYSCTL_ONE,
1084 .procname = "fib_multipath_hash_policy",
1085 .data = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1086 .maxlen = sizeof(u8),
1088 .proc_handler = proc_fib_multipath_hash_policy,
1089 .extra1 = SYSCTL_ZERO,
1093 .procname = "fib_multipath_hash_fields",
1094 .data = &init_net.ipv4.sysctl_fib_multipath_hash_fields,
1095 .maxlen = sizeof(u32),
1097 .proc_handler = proc_fib_multipath_hash_fields,
1098 .extra1 = SYSCTL_ONE,
1099 .extra2 = &fib_multipath_hash_fields_all_mask,
1103 .procname = "ip_unprivileged_port_start",
1104 .maxlen = sizeof(int),
1105 .data = &init_net.ipv4.sysctl_ip_prot_sock,
1107 .proc_handler = ipv4_privileged_ports,
1109 #ifdef CONFIG_NET_L3_MASTER_DEV
1111 .procname = "udp_l3mdev_accept",
1112 .data = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1113 .maxlen = sizeof(u8),
1115 .proc_handler = proc_dou8vec_minmax,
1116 .extra1 = SYSCTL_ZERO,
1117 .extra2 = SYSCTL_ONE,
1121 .procname = "tcp_sack",
1122 .data = &init_net.ipv4.sysctl_tcp_sack,
1123 .maxlen = sizeof(u8),
1125 .proc_handler = proc_dou8vec_minmax,
1128 .procname = "tcp_window_scaling",
1129 .data = &init_net.ipv4.sysctl_tcp_window_scaling,
1130 .maxlen = sizeof(u8),
1132 .proc_handler = proc_dou8vec_minmax,
1135 .procname = "tcp_timestamps",
1136 .data = &init_net.ipv4.sysctl_tcp_timestamps,
1137 .maxlen = sizeof(u8),
1139 .proc_handler = proc_dou8vec_minmax,
1142 .procname = "tcp_early_retrans",
1143 .data = &init_net.ipv4.sysctl_tcp_early_retrans,
1144 .maxlen = sizeof(u8),
1146 .proc_handler = proc_dou8vec_minmax,
1147 .extra1 = SYSCTL_ZERO,
1151 .procname = "tcp_recovery",
1152 .data = &init_net.ipv4.sysctl_tcp_recovery,
1153 .maxlen = sizeof(u8),
1155 .proc_handler = proc_dou8vec_minmax,
1158 .procname = "tcp_thin_linear_timeouts",
1159 .data = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1160 .maxlen = sizeof(u8),
1162 .proc_handler = proc_dou8vec_minmax,
1165 .procname = "tcp_slow_start_after_idle",
1166 .data = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1167 .maxlen = sizeof(u8),
1169 .proc_handler = proc_dou8vec_minmax,
1172 .procname = "tcp_retrans_collapse",
1173 .data = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1174 .maxlen = sizeof(u8),
1176 .proc_handler = proc_dou8vec_minmax,
1179 .procname = "tcp_stdurg",
1180 .data = &init_net.ipv4.sysctl_tcp_stdurg,
1181 .maxlen = sizeof(u8),
1183 .proc_handler = proc_dou8vec_minmax,
1186 .procname = "tcp_rfc1337",
1187 .data = &init_net.ipv4.sysctl_tcp_rfc1337,
1188 .maxlen = sizeof(u8),
1190 .proc_handler = proc_dou8vec_minmax,
1193 .procname = "tcp_abort_on_overflow",
1194 .data = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1195 .maxlen = sizeof(u8),
1197 .proc_handler = proc_dou8vec_minmax,
1200 .procname = "tcp_fack",
1201 .data = &init_net.ipv4.sysctl_tcp_fack,
1202 .maxlen = sizeof(u8),
1204 .proc_handler = proc_dou8vec_minmax,
1207 .procname = "tcp_max_reordering",
1208 .data = &init_net.ipv4.sysctl_tcp_max_reordering,
1209 .maxlen = sizeof(int),
1211 .proc_handler = proc_dointvec
1214 .procname = "tcp_dsack",
1215 .data = &init_net.ipv4.sysctl_tcp_dsack,
1216 .maxlen = sizeof(u8),
1218 .proc_handler = proc_dou8vec_minmax,
1221 .procname = "tcp_app_win",
1222 .data = &init_net.ipv4.sysctl_tcp_app_win,
1223 .maxlen = sizeof(u8),
1225 .proc_handler = proc_dou8vec_minmax,
1228 .procname = "tcp_adv_win_scale",
1229 .data = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1230 .maxlen = sizeof(int),
1232 .proc_handler = proc_dointvec_minmax,
1233 .extra1 = &tcp_adv_win_scale_min,
1234 .extra2 = &tcp_adv_win_scale_max,
1237 .procname = "tcp_frto",
1238 .data = &init_net.ipv4.sysctl_tcp_frto,
1239 .maxlen = sizeof(u8),
1241 .proc_handler = proc_dou8vec_minmax,
1244 .procname = "tcp_no_metrics_save",
1245 .data = &init_net.ipv4.sysctl_tcp_nometrics_save,
1246 .maxlen = sizeof(u8),
1248 .proc_handler = proc_dou8vec_minmax,
1251 .procname = "tcp_no_ssthresh_metrics_save",
1252 .data = &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1253 .maxlen = sizeof(u8),
1255 .proc_handler = proc_dou8vec_minmax,
1256 .extra1 = SYSCTL_ZERO,
1257 .extra2 = SYSCTL_ONE,
1260 .procname = "tcp_moderate_rcvbuf",
1261 .data = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1262 .maxlen = sizeof(u8),
1264 .proc_handler = proc_dou8vec_minmax,
1267 .procname = "tcp_tso_win_divisor",
1268 .data = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1269 .maxlen = sizeof(u8),
1271 .proc_handler = proc_dou8vec_minmax,
1274 .procname = "tcp_workaround_signed_windows",
1275 .data = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1276 .maxlen = sizeof(u8),
1278 .proc_handler = proc_dou8vec_minmax,
1281 .procname = "tcp_limit_output_bytes",
1282 .data = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1283 .maxlen = sizeof(int),
1285 .proc_handler = proc_dointvec
1288 .procname = "tcp_challenge_ack_limit",
1289 .data = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1290 .maxlen = sizeof(int),
1292 .proc_handler = proc_dointvec
1295 .procname = "tcp_min_tso_segs",
1296 .data = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1297 .maxlen = sizeof(u8),
1299 .proc_handler = proc_dou8vec_minmax,
1300 .extra1 = SYSCTL_ONE,
1303 .procname = "tcp_min_rtt_wlen",
1304 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1305 .maxlen = sizeof(int),
1307 .proc_handler = proc_dointvec_minmax,
1308 .extra1 = SYSCTL_ZERO,
1309 .extra2 = &one_day_secs
1312 .procname = "tcp_autocorking",
1313 .data = &init_net.ipv4.sysctl_tcp_autocorking,
1314 .maxlen = sizeof(u8),
1316 .proc_handler = proc_dou8vec_minmax,
1317 .extra1 = SYSCTL_ZERO,
1318 .extra2 = SYSCTL_ONE,
1321 .procname = "tcp_invalid_ratelimit",
1322 .data = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1323 .maxlen = sizeof(int),
1325 .proc_handler = proc_dointvec_ms_jiffies,
1328 .procname = "tcp_pacing_ss_ratio",
1329 .data = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1330 .maxlen = sizeof(int),
1332 .proc_handler = proc_dointvec_minmax,
1333 .extra1 = SYSCTL_ZERO,
1334 .extra2 = &thousand,
1337 .procname = "tcp_pacing_ca_ratio",
1338 .data = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1339 .maxlen = sizeof(int),
1341 .proc_handler = proc_dointvec_minmax,
1342 .extra1 = SYSCTL_ZERO,
1343 .extra2 = &thousand,
1346 .procname = "tcp_wmem",
1347 .data = &init_net.ipv4.sysctl_tcp_wmem,
1348 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1350 .proc_handler = proc_dointvec_minmax,
1351 .extra1 = SYSCTL_ONE,
1354 .procname = "tcp_rmem",
1355 .data = &init_net.ipv4.sysctl_tcp_rmem,
1356 .maxlen = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1358 .proc_handler = proc_dointvec_minmax,
1359 .extra1 = SYSCTL_ONE,
1362 .procname = "tcp_comp_sack_delay_ns",
1363 .data = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1364 .maxlen = sizeof(unsigned long),
1366 .proc_handler = proc_doulongvec_minmax,
1369 .procname = "tcp_comp_sack_slack_ns",
1370 .data = &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1371 .maxlen = sizeof(unsigned long),
1373 .proc_handler = proc_doulongvec_minmax,
1376 .procname = "tcp_comp_sack_nr",
1377 .data = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1378 .maxlen = sizeof(u8),
1380 .proc_handler = proc_dou8vec_minmax,
1381 .extra1 = SYSCTL_ZERO,
1384 .procname = "tcp_reflect_tos",
1385 .data = &init_net.ipv4.sysctl_tcp_reflect_tos,
1386 .maxlen = sizeof(u8),
1388 .proc_handler = proc_dou8vec_minmax,
1389 .extra1 = SYSCTL_ZERO,
1390 .extra2 = SYSCTL_ONE,
1393 .procname = "udp_rmem_min",
1394 .data = &init_net.ipv4.sysctl_udp_rmem_min,
1395 .maxlen = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1397 .proc_handler = proc_dointvec_minmax,
1398 .extra1 = SYSCTL_ONE
1401 .procname = "udp_wmem_min",
1402 .data = &init_net.ipv4.sysctl_udp_wmem_min,
1403 .maxlen = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1405 .proc_handler = proc_dointvec_minmax,
1406 .extra1 = SYSCTL_ONE
1409 .procname = "fib_notify_on_flag_change",
1410 .data = &init_net.ipv4.sysctl_fib_notify_on_flag_change,
1411 .maxlen = sizeof(u8),
1413 .proc_handler = proc_dou8vec_minmax,
1414 .extra1 = SYSCTL_ZERO,
1420 static __net_init int ipv4_sysctl_init_net(struct net *net)
1422 struct ctl_table *table;
1424 table = ipv4_net_table;
1425 if (!net_eq(net, &init_net)) {
1428 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1432 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++) {
1433 if (table[i].data) {
1434 /* Update the variables to point into
1435 * the current struct net
1437 table[i].data += (void *)net - (void *)&init_net;
1439 /* Entries without data pointer are global;
1440 * Make them read-only in non-init_net ns
1442 table[i].mode &= ~0222;
1447 net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1448 if (!net->ipv4.ipv4_hdr)
1451 net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1452 if (!net->ipv4.sysctl_local_reserved_ports)
1458 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1460 if (!net_eq(net, &init_net))
1466 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1468 struct ctl_table *table;
1470 kfree(net->ipv4.sysctl_local_reserved_ports);
1471 table = net->ipv4.ipv4_hdr->ctl_table_arg;
1472 unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1476 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1477 .init = ipv4_sysctl_init_net,
1478 .exit = ipv4_sysctl_exit_net,
1481 static __init int sysctl_ipv4_init(void)
1483 struct ctl_table_header *hdr;
1485 hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1489 if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1490 unregister_net_sysctl_table(hdr);
1497 __initcall(sysctl_ipv4_init);