tcp: allocate tcp_death_row outside of struct netns_ipv4
[linux-block.git] / net / ipv4 / sysctl_net_ipv4.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
4  *
5  * Begun April 1, 1996, Mike Shaver.
6  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
7  */
8
9 #include <linux/sysctl.h>
10 #include <linux/seqlock.h>
11 #include <linux/init.h>
12 #include <linux/slab.h>
13 #include <net/icmp.h>
14 #include <net/ip.h>
15 #include <net/ip_fib.h>
16 #include <net/tcp.h>
17 #include <net/udp.h>
18 #include <net/cipso_ipv4.h>
19 #include <net/ping.h>
20 #include <net/protocol.h>
21 #include <net/netevent.h>
22
23 static int two = 2;
24 static int three __maybe_unused = 3;
25 static int four = 4;
26 static int thousand = 1000;
27 static int tcp_retr1_max = 255;
28 static int ip_local_port_range_min[] = { 1, 1 };
29 static int ip_local_port_range_max[] = { 65535, 65535 };
30 static int tcp_adv_win_scale_min = -31;
31 static int tcp_adv_win_scale_max = 31;
32 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
33 static int tcp_min_snd_mss_max = 65535;
34 static int ip_privileged_port_min;
35 static int ip_privileged_port_max = 65535;
36 static int ip_ttl_min = 1;
37 static int ip_ttl_max = 255;
38 static int tcp_syn_retries_min = 1;
39 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
40 static int ip_ping_group_range_min[] = { 0, 0 };
41 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
42 static u32 u32_max_div_HZ = UINT_MAX / HZ;
43 static int one_day_secs = 24 * 3600;
44 static u32 fib_multipath_hash_fields_all_mask __maybe_unused =
45         FIB_MULTIPATH_HASH_FIELD_ALL_MASK;
46
47 /* obsolete */
48 static int sysctl_tcp_low_latency __read_mostly;
49
50 /* Update system visible IP port range */
51 static void set_local_port_range(struct net *net, int range[2])
52 {
53         bool same_parity = !((range[0] ^ range[1]) & 1);
54
55         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
56         if (same_parity && !net->ipv4.ip_local_ports.warned) {
57                 net->ipv4.ip_local_ports.warned = true;
58                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
59         }
60         net->ipv4.ip_local_ports.range[0] = range[0];
61         net->ipv4.ip_local_ports.range[1] = range[1];
62         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
63 }
64
65 /* Validate changes from /proc interface. */
66 static int ipv4_local_port_range(struct ctl_table *table, int write,
67                                  void *buffer, size_t *lenp, loff_t *ppos)
68 {
69         struct net *net =
70                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
71         int ret;
72         int range[2];
73         struct ctl_table tmp = {
74                 .data = &range,
75                 .maxlen = sizeof(range),
76                 .mode = table->mode,
77                 .extra1 = &ip_local_port_range_min,
78                 .extra2 = &ip_local_port_range_max,
79         };
80
81         inet_get_local_port_range(net, &range[0], &range[1]);
82
83         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
84
85         if (write && ret == 0) {
86                 /* Ensure that the upper limit is not smaller than the lower,
87                  * and that the lower does not encroach upon the privileged
88                  * port limit.
89                  */
90                 if ((range[1] < range[0]) ||
91                     (range[0] < net->ipv4.sysctl_ip_prot_sock))
92                         ret = -EINVAL;
93                 else
94                         set_local_port_range(net, range);
95         }
96
97         return ret;
98 }
99
100 /* Validate changes from /proc interface. */
101 static int ipv4_privileged_ports(struct ctl_table *table, int write,
102                                 void *buffer, size_t *lenp, loff_t *ppos)
103 {
104         struct net *net = container_of(table->data, struct net,
105             ipv4.sysctl_ip_prot_sock);
106         int ret;
107         int pports;
108         int range[2];
109         struct ctl_table tmp = {
110                 .data = &pports,
111                 .maxlen = sizeof(pports),
112                 .mode = table->mode,
113                 .extra1 = &ip_privileged_port_min,
114                 .extra2 = &ip_privileged_port_max,
115         };
116
117         pports = net->ipv4.sysctl_ip_prot_sock;
118
119         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
120
121         if (write && ret == 0) {
122                 inet_get_local_port_range(net, &range[0], &range[1]);
123                 /* Ensure that the local port range doesn't overlap with the
124                  * privileged port range.
125                  */
126                 if (range[0] < pports)
127                         ret = -EINVAL;
128                 else
129                         net->ipv4.sysctl_ip_prot_sock = pports;
130         }
131
132         return ret;
133 }
134
135 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
136 {
137         kgid_t *data = table->data;
138         struct net *net =
139                 container_of(table->data, struct net, ipv4.ping_group_range.range);
140         unsigned int seq;
141         do {
142                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
143
144                 *low = data[0];
145                 *high = data[1];
146         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
147 }
148
149 /* Update system visible IP port range */
150 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
151 {
152         kgid_t *data = table->data;
153         struct net *net =
154                 container_of(table->data, struct net, ipv4.ping_group_range.range);
155         write_seqlock(&net->ipv4.ping_group_range.lock);
156         data[0] = low;
157         data[1] = high;
158         write_sequnlock(&net->ipv4.ping_group_range.lock);
159 }
160
161 /* Validate changes from /proc interface. */
162 static int ipv4_ping_group_range(struct ctl_table *table, int write,
163                                  void *buffer, size_t *lenp, loff_t *ppos)
164 {
165         struct user_namespace *user_ns = current_user_ns();
166         int ret;
167         gid_t urange[2];
168         kgid_t low, high;
169         struct ctl_table tmp = {
170                 .data = &urange,
171                 .maxlen = sizeof(urange),
172                 .mode = table->mode,
173                 .extra1 = &ip_ping_group_range_min,
174                 .extra2 = &ip_ping_group_range_max,
175         };
176
177         inet_get_ping_group_range_table(table, &low, &high);
178         urange[0] = from_kgid_munged(user_ns, low);
179         urange[1] = from_kgid_munged(user_ns, high);
180         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
181
182         if (write && ret == 0) {
183                 low = make_kgid(user_ns, urange[0]);
184                 high = make_kgid(user_ns, urange[1]);
185                 if (!gid_valid(low) || !gid_valid(high))
186                         return -EINVAL;
187                 if (urange[1] < urange[0] || gid_lt(high, low)) {
188                         low = make_kgid(&init_user_ns, 1);
189                         high = make_kgid(&init_user_ns, 0);
190                 }
191                 set_ping_group_range(table, low, high);
192         }
193
194         return ret;
195 }
196
197 static int ipv4_fwd_update_priority(struct ctl_table *table, int write,
198                                     void *buffer, size_t *lenp, loff_t *ppos)
199 {
200         struct net *net;
201         int ret;
202
203         net = container_of(table->data, struct net,
204                            ipv4.sysctl_ip_fwd_update_priority);
205         ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
206         if (write && ret == 0)
207                 call_netevent_notifiers(NETEVENT_IPV4_FWD_UPDATE_PRIORITY_UPDATE,
208                                         net);
209
210         return ret;
211 }
212
213 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
214                                        void *buffer, size_t *lenp, loff_t *ppos)
215 {
216         struct net *net = container_of(ctl->data, struct net,
217                                        ipv4.tcp_congestion_control);
218         char val[TCP_CA_NAME_MAX];
219         struct ctl_table tbl = {
220                 .data = val,
221                 .maxlen = TCP_CA_NAME_MAX,
222         };
223         int ret;
224
225         tcp_get_default_congestion_control(net, val);
226
227         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
228         if (write && ret == 0)
229                 ret = tcp_set_default_congestion_control(net, val);
230         return ret;
231 }
232
233 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
234                                                  int write, void *buffer,
235                                                  size_t *lenp, loff_t *ppos)
236 {
237         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
238         int ret;
239
240         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
241         if (!tbl.data)
242                 return -ENOMEM;
243         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
244         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
245         kfree(tbl.data);
246         return ret;
247 }
248
249 static int proc_allowed_congestion_control(struct ctl_table *ctl,
250                                            int write, void *buffer,
251                                            size_t *lenp, loff_t *ppos)
252 {
253         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
254         int ret;
255
256         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
257         if (!tbl.data)
258                 return -ENOMEM;
259
260         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
261         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
262         if (write && ret == 0)
263                 ret = tcp_set_allowed_congestion_control(tbl.data);
264         kfree(tbl.data);
265         return ret;
266 }
267
268 static int sscanf_key(char *buf, __le32 *key)
269 {
270         u32 user_key[4];
271         int i, ret = 0;
272
273         if (sscanf(buf, "%x-%x-%x-%x", user_key, user_key + 1,
274                    user_key + 2, user_key + 3) != 4) {
275                 ret = -EINVAL;
276         } else {
277                 for (i = 0; i < ARRAY_SIZE(user_key); i++)
278                         key[i] = cpu_to_le32(user_key[i]);
279         }
280         pr_debug("proc TFO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
281                  user_key[0], user_key[1], user_key[2], user_key[3], buf, ret);
282
283         return ret;
284 }
285
286 static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
287                                  void *buffer, size_t *lenp, loff_t *ppos)
288 {
289         struct net *net = container_of(table->data, struct net,
290             ipv4.sysctl_tcp_fastopen);
291         /* maxlen to print the list of keys in hex (*2), with dashes
292          * separating doublewords and a comma in between keys.
293          */
294         struct ctl_table tbl = { .maxlen = ((TCP_FASTOPEN_KEY_LENGTH *
295                                             2 * TCP_FASTOPEN_KEY_MAX) +
296                                             (TCP_FASTOPEN_KEY_MAX * 5)) };
297         u32 user_key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(u32)];
298         __le32 key[TCP_FASTOPEN_KEY_BUF_LENGTH / sizeof(__le32)];
299         char *backup_data;
300         int ret, i = 0, off = 0, n_keys;
301
302         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
303         if (!tbl.data)
304                 return -ENOMEM;
305
306         n_keys = tcp_fastopen_get_cipher(net, NULL, (u64 *)key);
307         if (!n_keys) {
308                 memset(&key[0], 0, TCP_FASTOPEN_KEY_LENGTH);
309                 n_keys = 1;
310         }
311
312         for (i = 0; i < n_keys * 4; i++)
313                 user_key[i] = le32_to_cpu(key[i]);
314
315         for (i = 0; i < n_keys; i++) {
316                 off += snprintf(tbl.data + off, tbl.maxlen - off,
317                                 "%08x-%08x-%08x-%08x",
318                                 user_key[i * 4],
319                                 user_key[i * 4 + 1],
320                                 user_key[i * 4 + 2],
321                                 user_key[i * 4 + 3]);
322
323                 if (WARN_ON_ONCE(off >= tbl.maxlen - 1))
324                         break;
325
326                 if (i + 1 < n_keys)
327                         off += snprintf(tbl.data + off, tbl.maxlen - off, ",");
328         }
329
330         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
331
332         if (write && ret == 0) {
333                 backup_data = strchr(tbl.data, ',');
334                 if (backup_data) {
335                         *backup_data = '\0';
336                         backup_data++;
337                 }
338                 if (sscanf_key(tbl.data, key)) {
339                         ret = -EINVAL;
340                         goto bad_key;
341                 }
342                 if (backup_data) {
343                         if (sscanf_key(backup_data, key + 4)) {
344                                 ret = -EINVAL;
345                                 goto bad_key;
346                         }
347                 }
348                 tcp_fastopen_reset_cipher(net, NULL, key,
349                                           backup_data ? key + 4 : NULL);
350         }
351
352 bad_key:
353         kfree(tbl.data);
354         return ret;
355 }
356
357 static void proc_configure_early_demux(int enabled, int protocol)
358 {
359         struct net_protocol *ipprot;
360 #if IS_ENABLED(CONFIG_IPV6)
361         struct inet6_protocol *ip6prot;
362 #endif
363
364         rcu_read_lock();
365
366         ipprot = rcu_dereference(inet_protos[protocol]);
367         if (ipprot)
368                 ipprot->early_demux = enabled ? ipprot->early_demux_handler :
369                                                 NULL;
370
371 #if IS_ENABLED(CONFIG_IPV6)
372         ip6prot = rcu_dereference(inet6_protos[protocol]);
373         if (ip6prot)
374                 ip6prot->early_demux = enabled ? ip6prot->early_demux_handler :
375                                                  NULL;
376 #endif
377         rcu_read_unlock();
378 }
379
380 static int proc_tcp_early_demux(struct ctl_table *table, int write,
381                                 void *buffer, size_t *lenp, loff_t *ppos)
382 {
383         int ret = 0;
384
385         ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
386
387         if (write && !ret) {
388                 int enabled = init_net.ipv4.sysctl_tcp_early_demux;
389
390                 proc_configure_early_demux(enabled, IPPROTO_TCP);
391         }
392
393         return ret;
394 }
395
396 static int proc_udp_early_demux(struct ctl_table *table, int write,
397                                 void *buffer, size_t *lenp, loff_t *ppos)
398 {
399         int ret = 0;
400
401         ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
402
403         if (write && !ret) {
404                 int enabled = init_net.ipv4.sysctl_udp_early_demux;
405
406                 proc_configure_early_demux(enabled, IPPROTO_UDP);
407         }
408
409         return ret;
410 }
411
412 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
413                                              int write, void *buffer,
414                                              size_t *lenp, loff_t *ppos)
415 {
416         struct net *net = container_of(table->data, struct net,
417             ipv4.sysctl_tcp_fastopen_blackhole_timeout);
418         int ret;
419
420         ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
421         if (write && ret == 0)
422                 atomic_set(&net->ipv4.tfo_active_disable_times, 0);
423
424         return ret;
425 }
426
427 static int proc_tcp_available_ulp(struct ctl_table *ctl,
428                                   int write, void *buffer, size_t *lenp,
429                                   loff_t *ppos)
430 {
431         struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
432         int ret;
433
434         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
435         if (!tbl.data)
436                 return -ENOMEM;
437         tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
438         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
439         kfree(tbl.data);
440
441         return ret;
442 }
443
444 #ifdef CONFIG_IP_ROUTE_MULTIPATH
445 static int proc_fib_multipath_hash_policy(struct ctl_table *table, int write,
446                                           void *buffer, size_t *lenp,
447                                           loff_t *ppos)
448 {
449         struct net *net = container_of(table->data, struct net,
450             ipv4.sysctl_fib_multipath_hash_policy);
451         int ret;
452
453         ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
454         if (write && ret == 0)
455                 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
456
457         return ret;
458 }
459
460 static int proc_fib_multipath_hash_fields(struct ctl_table *table, int write,
461                                           void *buffer, size_t *lenp,
462                                           loff_t *ppos)
463 {
464         struct net *net;
465         int ret;
466
467         net = container_of(table->data, struct net,
468                            ipv4.sysctl_fib_multipath_hash_fields);
469         ret = proc_douintvec_minmax(table, write, buffer, lenp, ppos);
470         if (write && ret == 0)
471                 call_netevent_notifiers(NETEVENT_IPV4_MPATH_HASH_UPDATE, net);
472
473         return ret;
474 }
475 #endif
476
477 static struct ctl_table ipv4_table[] = {
478         {
479                 .procname       = "tcp_max_orphans",
480                 .data           = &sysctl_tcp_max_orphans,
481                 .maxlen         = sizeof(int),
482                 .mode           = 0644,
483                 .proc_handler   = proc_dointvec
484         },
485         {
486                 .procname       = "inet_peer_threshold",
487                 .data           = &inet_peer_threshold,
488                 .maxlen         = sizeof(int),
489                 .mode           = 0644,
490                 .proc_handler   = proc_dointvec
491         },
492         {
493                 .procname       = "inet_peer_minttl",
494                 .data           = &inet_peer_minttl,
495                 .maxlen         = sizeof(int),
496                 .mode           = 0644,
497                 .proc_handler   = proc_dointvec_jiffies,
498         },
499         {
500                 .procname       = "inet_peer_maxttl",
501                 .data           = &inet_peer_maxttl,
502                 .maxlen         = sizeof(int),
503                 .mode           = 0644,
504                 .proc_handler   = proc_dointvec_jiffies,
505         },
506         {
507                 .procname       = "tcp_mem",
508                 .maxlen         = sizeof(sysctl_tcp_mem),
509                 .data           = &sysctl_tcp_mem,
510                 .mode           = 0644,
511                 .proc_handler   = proc_doulongvec_minmax,
512         },
513         {
514                 .procname       = "tcp_low_latency",
515                 .data           = &sysctl_tcp_low_latency,
516                 .maxlen         = sizeof(int),
517                 .mode           = 0644,
518                 .proc_handler   = proc_dointvec
519         },
520 #ifdef CONFIG_NETLABEL
521         {
522                 .procname       = "cipso_cache_enable",
523                 .data           = &cipso_v4_cache_enabled,
524                 .maxlen         = sizeof(int),
525                 .mode           = 0644,
526                 .proc_handler   = proc_dointvec,
527         },
528         {
529                 .procname       = "cipso_cache_bucket_size",
530                 .data           = &cipso_v4_cache_bucketsize,
531                 .maxlen         = sizeof(int),
532                 .mode           = 0644,
533                 .proc_handler   = proc_dointvec,
534         },
535         {
536                 .procname       = "cipso_rbm_optfmt",
537                 .data           = &cipso_v4_rbm_optfmt,
538                 .maxlen         = sizeof(int),
539                 .mode           = 0644,
540                 .proc_handler   = proc_dointvec,
541         },
542         {
543                 .procname       = "cipso_rbm_strictvalid",
544                 .data           = &cipso_v4_rbm_strictvalid,
545                 .maxlen         = sizeof(int),
546                 .mode           = 0644,
547                 .proc_handler   = proc_dointvec,
548         },
549 #endif /* CONFIG_NETLABEL */
550         {
551                 .procname       = "tcp_available_ulp",
552                 .maxlen         = TCP_ULP_BUF_MAX,
553                 .mode           = 0444,
554                 .proc_handler   = proc_tcp_available_ulp,
555         },
556         {
557                 .procname       = "icmp_msgs_per_sec",
558                 .data           = &sysctl_icmp_msgs_per_sec,
559                 .maxlen         = sizeof(int),
560                 .mode           = 0644,
561                 .proc_handler   = proc_dointvec_minmax,
562                 .extra1         = SYSCTL_ZERO,
563         },
564         {
565                 .procname       = "icmp_msgs_burst",
566                 .data           = &sysctl_icmp_msgs_burst,
567                 .maxlen         = sizeof(int),
568                 .mode           = 0644,
569                 .proc_handler   = proc_dointvec_minmax,
570                 .extra1         = SYSCTL_ZERO,
571         },
572         {
573                 .procname       = "udp_mem",
574                 .data           = &sysctl_udp_mem,
575                 .maxlen         = sizeof(sysctl_udp_mem),
576                 .mode           = 0644,
577                 .proc_handler   = proc_doulongvec_minmax,
578         },
579         {
580                 .procname       = "fib_sync_mem",
581                 .data           = &sysctl_fib_sync_mem,
582                 .maxlen         = sizeof(sysctl_fib_sync_mem),
583                 .mode           = 0644,
584                 .proc_handler   = proc_douintvec_minmax,
585                 .extra1         = &sysctl_fib_sync_mem_min,
586                 .extra2         = &sysctl_fib_sync_mem_max,
587         },
588         { }
589 };
590
591 static struct ctl_table ipv4_net_table[] = {
592         /* tcp_max_tw_buckets must be first in this table. */
593         {
594                 .procname       = "tcp_max_tw_buckets",
595 /*              .data           = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets, */
596                 .maxlen         = sizeof(int),
597                 .mode           = 0644,
598                 .proc_handler   = proc_dointvec
599         },
600         {
601                 .procname       = "icmp_echo_ignore_all",
602                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
603                 .maxlen         = sizeof(u8),
604                 .mode           = 0644,
605                 .proc_handler   = proc_dou8vec_minmax,
606         },
607         {
608                 .procname       = "icmp_echo_enable_probe",
609                 .data           = &init_net.ipv4.sysctl_icmp_echo_enable_probe,
610                 .maxlen         = sizeof(u8),
611                 .mode           = 0644,
612                 .proc_handler   = proc_dou8vec_minmax,
613                 .extra1         = SYSCTL_ZERO,
614                 .extra2         = SYSCTL_ONE
615         },
616         {
617                 .procname       = "icmp_echo_ignore_broadcasts",
618                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
619                 .maxlen         = sizeof(u8),
620                 .mode           = 0644,
621                 .proc_handler   = proc_dou8vec_minmax,
622         },
623         {
624                 .procname       = "icmp_ignore_bogus_error_responses",
625                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
626                 .maxlen         = sizeof(u8),
627                 .mode           = 0644,
628                 .proc_handler   = proc_dou8vec_minmax,
629         },
630         {
631                 .procname       = "icmp_errors_use_inbound_ifaddr",
632                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
633                 .maxlen         = sizeof(u8),
634                 .mode           = 0644,
635                 .proc_handler   = proc_dou8vec_minmax,
636         },
637         {
638                 .procname       = "icmp_ratelimit",
639                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
640                 .maxlen         = sizeof(int),
641                 .mode           = 0644,
642                 .proc_handler   = proc_dointvec_ms_jiffies,
643         },
644         {
645                 .procname       = "icmp_ratemask",
646                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
647                 .maxlen         = sizeof(int),
648                 .mode           = 0644,
649                 .proc_handler   = proc_dointvec
650         },
651         {
652                 .procname       = "ping_group_range",
653                 .data           = &init_net.ipv4.ping_group_range.range,
654                 .maxlen         = sizeof(gid_t)*2,
655                 .mode           = 0644,
656                 .proc_handler   = ipv4_ping_group_range,
657         },
658 #ifdef CONFIG_NET_L3_MASTER_DEV
659         {
660                 .procname       = "raw_l3mdev_accept",
661                 .data           = &init_net.ipv4.sysctl_raw_l3mdev_accept,
662                 .maxlen         = sizeof(u8),
663                 .mode           = 0644,
664                 .proc_handler   = proc_dou8vec_minmax,
665                 .extra1         = SYSCTL_ZERO,
666                 .extra2         = SYSCTL_ONE,
667         },
668 #endif
669         {
670                 .procname       = "tcp_ecn",
671                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
672                 .maxlen         = sizeof(u8),
673                 .mode           = 0644,
674                 .proc_handler   = proc_dou8vec_minmax,
675         },
676         {
677                 .procname       = "tcp_ecn_fallback",
678                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
679                 .maxlen         = sizeof(u8),
680                 .mode           = 0644,
681                 .proc_handler   = proc_dou8vec_minmax,
682         },
683         {
684                 .procname       = "ip_dynaddr",
685                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
686                 .maxlen         = sizeof(u8),
687                 .mode           = 0644,
688                 .proc_handler   = proc_dou8vec_minmax,
689         },
690         {
691                 .procname       = "ip_early_demux",
692                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
693                 .maxlen         = sizeof(u8),
694                 .mode           = 0644,
695                 .proc_handler   = proc_dou8vec_minmax,
696         },
697         {
698                 .procname       = "udp_early_demux",
699                 .data           = &init_net.ipv4.sysctl_udp_early_demux,
700                 .maxlen         = sizeof(u8),
701                 .mode           = 0644,
702                 .proc_handler   = proc_udp_early_demux
703         },
704         {
705                 .procname       = "tcp_early_demux",
706                 .data           = &init_net.ipv4.sysctl_tcp_early_demux,
707                 .maxlen         = sizeof(u8),
708                 .mode           = 0644,
709                 .proc_handler   = proc_tcp_early_demux
710         },
711         {
712                 .procname       = "nexthop_compat_mode",
713                 .data           = &init_net.ipv4.sysctl_nexthop_compat_mode,
714                 .maxlen         = sizeof(u8),
715                 .mode           = 0644,
716                 .proc_handler   = proc_dou8vec_minmax,
717                 .extra1         = SYSCTL_ZERO,
718                 .extra2         = SYSCTL_ONE,
719         },
720         {
721                 .procname       = "ip_default_ttl",
722                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
723                 .maxlen         = sizeof(u8),
724                 .mode           = 0644,
725                 .proc_handler   = proc_dou8vec_minmax,
726                 .extra1         = &ip_ttl_min,
727                 .extra2         = &ip_ttl_max,
728         },
729         {
730                 .procname       = "ip_local_port_range",
731                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
732                 .data           = &init_net.ipv4.ip_local_ports.range,
733                 .mode           = 0644,
734                 .proc_handler   = ipv4_local_port_range,
735         },
736         {
737                 .procname       = "ip_local_reserved_ports",
738                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
739                 .maxlen         = 65536,
740                 .mode           = 0644,
741                 .proc_handler   = proc_do_large_bitmap,
742         },
743         {
744                 .procname       = "ip_no_pmtu_disc",
745                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
746                 .maxlen         = sizeof(u8),
747                 .mode           = 0644,
748                 .proc_handler   = proc_dou8vec_minmax,
749         },
750         {
751                 .procname       = "ip_forward_use_pmtu",
752                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
753                 .maxlen         = sizeof(u8),
754                 .mode           = 0644,
755                 .proc_handler   = proc_dou8vec_minmax,
756         },
757         {
758                 .procname       = "ip_forward_update_priority",
759                 .data           = &init_net.ipv4.sysctl_ip_fwd_update_priority,
760                 .maxlen         = sizeof(u8),
761                 .mode           = 0644,
762                 .proc_handler   = ipv4_fwd_update_priority,
763                 .extra1         = SYSCTL_ZERO,
764                 .extra2         = SYSCTL_ONE,
765         },
766         {
767                 .procname       = "ip_nonlocal_bind",
768                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
769                 .maxlen         = sizeof(u8),
770                 .mode           = 0644,
771                 .proc_handler   = proc_dou8vec_minmax,
772         },
773         {
774                 .procname       = "ip_autobind_reuse",
775                 .data           = &init_net.ipv4.sysctl_ip_autobind_reuse,
776                 .maxlen         = sizeof(u8),
777                 .mode           = 0644,
778                 .proc_handler   = proc_dou8vec_minmax,
779                 .extra1         = SYSCTL_ZERO,
780                 .extra2         = SYSCTL_ONE,
781         },
782         {
783                 .procname       = "fwmark_reflect",
784                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
785                 .maxlen         = sizeof(u8),
786                 .mode           = 0644,
787                 .proc_handler   = proc_dou8vec_minmax,
788         },
789         {
790                 .procname       = "tcp_fwmark_accept",
791                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
792                 .maxlen         = sizeof(u8),
793                 .mode           = 0644,
794                 .proc_handler   = proc_dou8vec_minmax,
795         },
796 #ifdef CONFIG_NET_L3_MASTER_DEV
797         {
798                 .procname       = "tcp_l3mdev_accept",
799                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
800                 .maxlen         = sizeof(u8),
801                 .mode           = 0644,
802                 .proc_handler   = proc_dou8vec_minmax,
803                 .extra1         = SYSCTL_ZERO,
804                 .extra2         = SYSCTL_ONE,
805         },
806 #endif
807         {
808                 .procname       = "tcp_mtu_probing",
809                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
810                 .maxlen         = sizeof(u8),
811                 .mode           = 0644,
812                 .proc_handler   = proc_dou8vec_minmax,
813         },
814         {
815                 .procname       = "tcp_base_mss",
816                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
817                 .maxlen         = sizeof(int),
818                 .mode           = 0644,
819                 .proc_handler   = proc_dointvec,
820         },
821         {
822                 .procname       = "tcp_min_snd_mss",
823                 .data           = &init_net.ipv4.sysctl_tcp_min_snd_mss,
824                 .maxlen         = sizeof(int),
825                 .mode           = 0644,
826                 .proc_handler   = proc_dointvec_minmax,
827                 .extra1         = &tcp_min_snd_mss_min,
828                 .extra2         = &tcp_min_snd_mss_max,
829         },
830         {
831                 .procname       = "tcp_mtu_probe_floor",
832                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probe_floor,
833                 .maxlen         = sizeof(int),
834                 .mode           = 0644,
835                 .proc_handler   = proc_dointvec_minmax,
836                 .extra1         = &tcp_min_snd_mss_min,
837                 .extra2         = &tcp_min_snd_mss_max,
838         },
839         {
840                 .procname       = "tcp_probe_threshold",
841                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
842                 .maxlen         = sizeof(int),
843                 .mode           = 0644,
844                 .proc_handler   = proc_dointvec,
845         },
846         {
847                 .procname       = "tcp_probe_interval",
848                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
849                 .maxlen         = sizeof(u32),
850                 .mode           = 0644,
851                 .proc_handler   = proc_douintvec_minmax,
852                 .extra2         = &u32_max_div_HZ,
853         },
854         {
855                 .procname       = "igmp_link_local_mcast_reports",
856                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
857                 .maxlen         = sizeof(u8),
858                 .mode           = 0644,
859                 .proc_handler   = proc_dou8vec_minmax,
860         },
861         {
862                 .procname       = "igmp_max_memberships",
863                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
864                 .maxlen         = sizeof(int),
865                 .mode           = 0644,
866                 .proc_handler   = proc_dointvec
867         },
868         {
869                 .procname       = "igmp_max_msf",
870                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
871                 .maxlen         = sizeof(int),
872                 .mode           = 0644,
873                 .proc_handler   = proc_dointvec
874         },
875 #ifdef CONFIG_IP_MULTICAST
876         {
877                 .procname       = "igmp_qrv",
878                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
879                 .maxlen         = sizeof(int),
880                 .mode           = 0644,
881                 .proc_handler   = proc_dointvec_minmax,
882                 .extra1         = SYSCTL_ONE
883         },
884 #endif
885         {
886                 .procname       = "tcp_congestion_control",
887                 .data           = &init_net.ipv4.tcp_congestion_control,
888                 .mode           = 0644,
889                 .maxlen         = TCP_CA_NAME_MAX,
890                 .proc_handler   = proc_tcp_congestion_control,
891         },
892         {
893                 .procname       = "tcp_available_congestion_control",
894                 .maxlen         = TCP_CA_BUF_MAX,
895                 .mode           = 0444,
896                 .proc_handler   = proc_tcp_available_congestion_control,
897         },
898         {
899                 .procname       = "tcp_allowed_congestion_control",
900                 .maxlen         = TCP_CA_BUF_MAX,
901                 .mode           = 0644,
902                 .proc_handler   = proc_allowed_congestion_control,
903         },
904         {
905                 .procname       = "tcp_keepalive_time",
906                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
907                 .maxlen         = sizeof(int),
908                 .mode           = 0644,
909                 .proc_handler   = proc_dointvec_jiffies,
910         },
911         {
912                 .procname       = "tcp_keepalive_probes",
913                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
914                 .maxlen         = sizeof(u8),
915                 .mode           = 0644,
916                 .proc_handler   = proc_dou8vec_minmax,
917         },
918         {
919                 .procname       = "tcp_keepalive_intvl",
920                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
921                 .maxlen         = sizeof(int),
922                 .mode           = 0644,
923                 .proc_handler   = proc_dointvec_jiffies,
924         },
925         {
926                 .procname       = "tcp_syn_retries",
927                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
928                 .maxlen         = sizeof(u8),
929                 .mode           = 0644,
930                 .proc_handler   = proc_dou8vec_minmax,
931                 .extra1         = &tcp_syn_retries_min,
932                 .extra2         = &tcp_syn_retries_max
933         },
934         {
935                 .procname       = "tcp_synack_retries",
936                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
937                 .maxlen         = sizeof(u8),
938                 .mode           = 0644,
939                 .proc_handler   = proc_dou8vec_minmax,
940         },
941 #ifdef CONFIG_SYN_COOKIES
942         {
943                 .procname       = "tcp_syncookies",
944                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
945                 .maxlen         = sizeof(u8),
946                 .mode           = 0644,
947                 .proc_handler   = proc_dou8vec_minmax,
948         },
949 #endif
950         {
951                 .procname       = "tcp_migrate_req",
952                 .data           = &init_net.ipv4.sysctl_tcp_migrate_req,
953                 .maxlen         = sizeof(u8),
954                 .mode           = 0644,
955                 .proc_handler   = proc_dou8vec_minmax,
956                 .extra1         = SYSCTL_ZERO,
957                 .extra2         = SYSCTL_ONE
958         },
959         {
960                 .procname       = "tcp_reordering",
961                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
962                 .maxlen         = sizeof(int),
963                 .mode           = 0644,
964                 .proc_handler   = proc_dointvec
965         },
966         {
967                 .procname       = "tcp_retries1",
968                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
969                 .maxlen         = sizeof(u8),
970                 .mode           = 0644,
971                 .proc_handler   = proc_dou8vec_minmax,
972                 .extra2         = &tcp_retr1_max
973         },
974         {
975                 .procname       = "tcp_retries2",
976                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
977                 .maxlen         = sizeof(u8),
978                 .mode           = 0644,
979                 .proc_handler   = proc_dou8vec_minmax,
980         },
981         {
982                 .procname       = "tcp_orphan_retries",
983                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
984                 .maxlen         = sizeof(u8),
985                 .mode           = 0644,
986                 .proc_handler   = proc_dou8vec_minmax,
987         },
988         {
989                 .procname       = "tcp_fin_timeout",
990                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
991                 .maxlen         = sizeof(int),
992                 .mode           = 0644,
993                 .proc_handler   = proc_dointvec_jiffies,
994         },
995         {
996                 .procname       = "tcp_notsent_lowat",
997                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
998                 .maxlen         = sizeof(unsigned int),
999                 .mode           = 0644,
1000                 .proc_handler   = proc_douintvec,
1001         },
1002         {
1003                 .procname       = "tcp_tw_reuse",
1004                 .data           = &init_net.ipv4.sysctl_tcp_tw_reuse,
1005                 .maxlen         = sizeof(u8),
1006                 .mode           = 0644,
1007                 .proc_handler   = proc_dou8vec_minmax,
1008                 .extra1         = SYSCTL_ZERO,
1009                 .extra2         = &two,
1010         },
1011         {
1012                 .procname       = "tcp_max_syn_backlog",
1013                 .data           = &init_net.ipv4.sysctl_max_syn_backlog,
1014                 .maxlen         = sizeof(int),
1015                 .mode           = 0644,
1016                 .proc_handler   = proc_dointvec
1017         },
1018         {
1019                 .procname       = "tcp_fastopen",
1020                 .data           = &init_net.ipv4.sysctl_tcp_fastopen,
1021                 .maxlen         = sizeof(int),
1022                 .mode           = 0644,
1023                 .proc_handler   = proc_dointvec,
1024         },
1025         {
1026                 .procname       = "tcp_fastopen_key",
1027                 .mode           = 0600,
1028                 .data           = &init_net.ipv4.sysctl_tcp_fastopen,
1029                 /* maxlen to print the list of keys in hex (*2), with dashes
1030                  * separating doublewords and a comma in between keys.
1031                  */
1032                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH *
1033                                    2 * TCP_FASTOPEN_KEY_MAX) +
1034                                    (TCP_FASTOPEN_KEY_MAX * 5)),
1035                 .proc_handler   = proc_tcp_fastopen_key,
1036         },
1037         {
1038                 .procname       = "tcp_fastopen_blackhole_timeout_sec",
1039                 .data           = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
1040                 .maxlen         = sizeof(int),
1041                 .mode           = 0644,
1042                 .proc_handler   = proc_tfo_blackhole_detect_timeout,
1043                 .extra1         = SYSCTL_ZERO,
1044         },
1045 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1046         {
1047                 .procname       = "fib_multipath_use_neigh",
1048                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1049                 .maxlen         = sizeof(u8),
1050                 .mode           = 0644,
1051                 .proc_handler   = proc_dou8vec_minmax,
1052                 .extra1         = SYSCTL_ZERO,
1053                 .extra2         = SYSCTL_ONE,
1054         },
1055         {
1056                 .procname       = "fib_multipath_hash_policy",
1057                 .data           = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1058                 .maxlen         = sizeof(u8),
1059                 .mode           = 0644,
1060                 .proc_handler   = proc_fib_multipath_hash_policy,
1061                 .extra1         = SYSCTL_ZERO,
1062                 .extra2         = &three,
1063         },
1064         {
1065                 .procname       = "fib_multipath_hash_fields",
1066                 .data           = &init_net.ipv4.sysctl_fib_multipath_hash_fields,
1067                 .maxlen         = sizeof(u32),
1068                 .mode           = 0644,
1069                 .proc_handler   = proc_fib_multipath_hash_fields,
1070                 .extra1         = SYSCTL_ONE,
1071                 .extra2         = &fib_multipath_hash_fields_all_mask,
1072         },
1073 #endif
1074         {
1075                 .procname       = "ip_unprivileged_port_start",
1076                 .maxlen         = sizeof(int),
1077                 .data           = &init_net.ipv4.sysctl_ip_prot_sock,
1078                 .mode           = 0644,
1079                 .proc_handler   = ipv4_privileged_ports,
1080         },
1081 #ifdef CONFIG_NET_L3_MASTER_DEV
1082         {
1083                 .procname       = "udp_l3mdev_accept",
1084                 .data           = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1085                 .maxlen         = sizeof(u8),
1086                 .mode           = 0644,
1087                 .proc_handler   = proc_dou8vec_minmax,
1088                 .extra1         = SYSCTL_ZERO,
1089                 .extra2         = SYSCTL_ONE,
1090         },
1091 #endif
1092         {
1093                 .procname       = "tcp_sack",
1094                 .data           = &init_net.ipv4.sysctl_tcp_sack,
1095                 .maxlen         = sizeof(u8),
1096                 .mode           = 0644,
1097                 .proc_handler   = proc_dou8vec_minmax,
1098         },
1099         {
1100                 .procname       = "tcp_window_scaling",
1101                 .data           = &init_net.ipv4.sysctl_tcp_window_scaling,
1102                 .maxlen         = sizeof(u8),
1103                 .mode           = 0644,
1104                 .proc_handler   = proc_dou8vec_minmax,
1105         },
1106         {
1107                 .procname       = "tcp_timestamps",
1108                 .data           = &init_net.ipv4.sysctl_tcp_timestamps,
1109                 .maxlen         = sizeof(u8),
1110                 .mode           = 0644,
1111                 .proc_handler   = proc_dou8vec_minmax,
1112         },
1113         {
1114                 .procname       = "tcp_early_retrans",
1115                 .data           = &init_net.ipv4.sysctl_tcp_early_retrans,
1116                 .maxlen         = sizeof(u8),
1117                 .mode           = 0644,
1118                 .proc_handler   = proc_dou8vec_minmax,
1119                 .extra1         = SYSCTL_ZERO,
1120                 .extra2         = &four,
1121         },
1122         {
1123                 .procname       = "tcp_recovery",
1124                 .data           = &init_net.ipv4.sysctl_tcp_recovery,
1125                 .maxlen         = sizeof(u8),
1126                 .mode           = 0644,
1127                 .proc_handler   = proc_dou8vec_minmax,
1128         },
1129         {
1130                 .procname       = "tcp_thin_linear_timeouts",
1131                 .data           = &init_net.ipv4.sysctl_tcp_thin_linear_timeouts,
1132                 .maxlen         = sizeof(u8),
1133                 .mode           = 0644,
1134                 .proc_handler   = proc_dou8vec_minmax,
1135         },
1136         {
1137                 .procname       = "tcp_slow_start_after_idle",
1138                 .data           = &init_net.ipv4.sysctl_tcp_slow_start_after_idle,
1139                 .maxlen         = sizeof(u8),
1140                 .mode           = 0644,
1141                 .proc_handler   = proc_dou8vec_minmax,
1142         },
1143         {
1144                 .procname       = "tcp_retrans_collapse",
1145                 .data           = &init_net.ipv4.sysctl_tcp_retrans_collapse,
1146                 .maxlen         = sizeof(u8),
1147                 .mode           = 0644,
1148                 .proc_handler   = proc_dou8vec_minmax,
1149         },
1150         {
1151                 .procname       = "tcp_stdurg",
1152                 .data           = &init_net.ipv4.sysctl_tcp_stdurg,
1153                 .maxlen         = sizeof(u8),
1154                 .mode           = 0644,
1155                 .proc_handler   = proc_dou8vec_minmax,
1156         },
1157         {
1158                 .procname       = "tcp_rfc1337",
1159                 .data           = &init_net.ipv4.sysctl_tcp_rfc1337,
1160                 .maxlen         = sizeof(u8),
1161                 .mode           = 0644,
1162                 .proc_handler   = proc_dou8vec_minmax,
1163         },
1164         {
1165                 .procname       = "tcp_abort_on_overflow",
1166                 .data           = &init_net.ipv4.sysctl_tcp_abort_on_overflow,
1167                 .maxlen         = sizeof(u8),
1168                 .mode           = 0644,
1169                 .proc_handler   = proc_dou8vec_minmax,
1170         },
1171         {
1172                 .procname       = "tcp_fack",
1173                 .data           = &init_net.ipv4.sysctl_tcp_fack,
1174                 .maxlen         = sizeof(u8),
1175                 .mode           = 0644,
1176                 .proc_handler   = proc_dou8vec_minmax,
1177         },
1178         {
1179                 .procname       = "tcp_max_reordering",
1180                 .data           = &init_net.ipv4.sysctl_tcp_max_reordering,
1181                 .maxlen         = sizeof(int),
1182                 .mode           = 0644,
1183                 .proc_handler   = proc_dointvec
1184         },
1185         {
1186                 .procname       = "tcp_dsack",
1187                 .data           = &init_net.ipv4.sysctl_tcp_dsack,
1188                 .maxlen         = sizeof(u8),
1189                 .mode           = 0644,
1190                 .proc_handler   = proc_dou8vec_minmax,
1191         },
1192         {
1193                 .procname       = "tcp_app_win",
1194                 .data           = &init_net.ipv4.sysctl_tcp_app_win,
1195                 .maxlen         = sizeof(u8),
1196                 .mode           = 0644,
1197                 .proc_handler   = proc_dou8vec_minmax,
1198         },
1199         {
1200                 .procname       = "tcp_adv_win_scale",
1201                 .data           = &init_net.ipv4.sysctl_tcp_adv_win_scale,
1202                 .maxlen         = sizeof(int),
1203                 .mode           = 0644,
1204                 .proc_handler   = proc_dointvec_minmax,
1205                 .extra1         = &tcp_adv_win_scale_min,
1206                 .extra2         = &tcp_adv_win_scale_max,
1207         },
1208         {
1209                 .procname       = "tcp_frto",
1210                 .data           = &init_net.ipv4.sysctl_tcp_frto,
1211                 .maxlen         = sizeof(u8),
1212                 .mode           = 0644,
1213                 .proc_handler   = proc_dou8vec_minmax,
1214         },
1215         {
1216                 .procname       = "tcp_no_metrics_save",
1217                 .data           = &init_net.ipv4.sysctl_tcp_nometrics_save,
1218                 .maxlen         = sizeof(u8),
1219                 .mode           = 0644,
1220                 .proc_handler   = proc_dou8vec_minmax,
1221         },
1222         {
1223                 .procname       = "tcp_no_ssthresh_metrics_save",
1224                 .data           = &init_net.ipv4.sysctl_tcp_no_ssthresh_metrics_save,
1225                 .maxlen         = sizeof(u8),
1226                 .mode           = 0644,
1227                 .proc_handler   = proc_dou8vec_minmax,
1228                 .extra1         = SYSCTL_ZERO,
1229                 .extra2         = SYSCTL_ONE,
1230         },
1231         {
1232                 .procname       = "tcp_moderate_rcvbuf",
1233                 .data           = &init_net.ipv4.sysctl_tcp_moderate_rcvbuf,
1234                 .maxlen         = sizeof(u8),
1235                 .mode           = 0644,
1236                 .proc_handler   = proc_dou8vec_minmax,
1237         },
1238         {
1239                 .procname       = "tcp_tso_win_divisor",
1240                 .data           = &init_net.ipv4.sysctl_tcp_tso_win_divisor,
1241                 .maxlen         = sizeof(u8),
1242                 .mode           = 0644,
1243                 .proc_handler   = proc_dou8vec_minmax,
1244         },
1245         {
1246                 .procname       = "tcp_workaround_signed_windows",
1247                 .data           = &init_net.ipv4.sysctl_tcp_workaround_signed_windows,
1248                 .maxlen         = sizeof(u8),
1249                 .mode           = 0644,
1250                 .proc_handler   = proc_dou8vec_minmax,
1251         },
1252         {
1253                 .procname       = "tcp_limit_output_bytes",
1254                 .data           = &init_net.ipv4.sysctl_tcp_limit_output_bytes,
1255                 .maxlen         = sizeof(int),
1256                 .mode           = 0644,
1257                 .proc_handler   = proc_dointvec
1258         },
1259         {
1260                 .procname       = "tcp_challenge_ack_limit",
1261                 .data           = &init_net.ipv4.sysctl_tcp_challenge_ack_limit,
1262                 .maxlen         = sizeof(int),
1263                 .mode           = 0644,
1264                 .proc_handler   = proc_dointvec
1265         },
1266         {
1267                 .procname       = "tcp_min_tso_segs",
1268                 .data           = &init_net.ipv4.sysctl_tcp_min_tso_segs,
1269                 .maxlen         = sizeof(u8),
1270                 .mode           = 0644,
1271                 .proc_handler   = proc_dou8vec_minmax,
1272                 .extra1         = SYSCTL_ONE,
1273         },
1274         {
1275                 .procname       = "tcp_min_rtt_wlen",
1276                 .data           = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
1277                 .maxlen         = sizeof(int),
1278                 .mode           = 0644,
1279                 .proc_handler   = proc_dointvec_minmax,
1280                 .extra1         = SYSCTL_ZERO,
1281                 .extra2         = &one_day_secs
1282         },
1283         {
1284                 .procname       = "tcp_autocorking",
1285                 .data           = &init_net.ipv4.sysctl_tcp_autocorking,
1286                 .maxlen         = sizeof(u8),
1287                 .mode           = 0644,
1288                 .proc_handler   = proc_dou8vec_minmax,
1289                 .extra1         = SYSCTL_ZERO,
1290                 .extra2         = SYSCTL_ONE,
1291         },
1292         {
1293                 .procname       = "tcp_invalid_ratelimit",
1294                 .data           = &init_net.ipv4.sysctl_tcp_invalid_ratelimit,
1295                 .maxlen         = sizeof(int),
1296                 .mode           = 0644,
1297                 .proc_handler   = proc_dointvec_ms_jiffies,
1298         },
1299         {
1300                 .procname       = "tcp_pacing_ss_ratio",
1301                 .data           = &init_net.ipv4.sysctl_tcp_pacing_ss_ratio,
1302                 .maxlen         = sizeof(int),
1303                 .mode           = 0644,
1304                 .proc_handler   = proc_dointvec_minmax,
1305                 .extra1         = SYSCTL_ZERO,
1306                 .extra2         = &thousand,
1307         },
1308         {
1309                 .procname       = "tcp_pacing_ca_ratio",
1310                 .data           = &init_net.ipv4.sysctl_tcp_pacing_ca_ratio,
1311                 .maxlen         = sizeof(int),
1312                 .mode           = 0644,
1313                 .proc_handler   = proc_dointvec_minmax,
1314                 .extra1         = SYSCTL_ZERO,
1315                 .extra2         = &thousand,
1316         },
1317         {
1318                 .procname       = "tcp_wmem",
1319                 .data           = &init_net.ipv4.sysctl_tcp_wmem,
1320                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_wmem),
1321                 .mode           = 0644,
1322                 .proc_handler   = proc_dointvec_minmax,
1323                 .extra1         = SYSCTL_ONE,
1324         },
1325         {
1326                 .procname       = "tcp_rmem",
1327                 .data           = &init_net.ipv4.sysctl_tcp_rmem,
1328                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_rmem),
1329                 .mode           = 0644,
1330                 .proc_handler   = proc_dointvec_minmax,
1331                 .extra1         = SYSCTL_ONE,
1332         },
1333         {
1334                 .procname       = "tcp_comp_sack_delay_ns",
1335                 .data           = &init_net.ipv4.sysctl_tcp_comp_sack_delay_ns,
1336                 .maxlen         = sizeof(unsigned long),
1337                 .mode           = 0644,
1338                 .proc_handler   = proc_doulongvec_minmax,
1339         },
1340         {
1341                 .procname       = "tcp_comp_sack_slack_ns",
1342                 .data           = &init_net.ipv4.sysctl_tcp_comp_sack_slack_ns,
1343                 .maxlen         = sizeof(unsigned long),
1344                 .mode           = 0644,
1345                 .proc_handler   = proc_doulongvec_minmax,
1346         },
1347         {
1348                 .procname       = "tcp_comp_sack_nr",
1349                 .data           = &init_net.ipv4.sysctl_tcp_comp_sack_nr,
1350                 .maxlen         = sizeof(u8),
1351                 .mode           = 0644,
1352                 .proc_handler   = proc_dou8vec_minmax,
1353                 .extra1         = SYSCTL_ZERO,
1354         },
1355         {
1356                 .procname       = "tcp_reflect_tos",
1357                 .data           = &init_net.ipv4.sysctl_tcp_reflect_tos,
1358                 .maxlen         = sizeof(u8),
1359                 .mode           = 0644,
1360                 .proc_handler   = proc_dou8vec_minmax,
1361                 .extra1         = SYSCTL_ZERO,
1362                 .extra2         = SYSCTL_ONE,
1363         },
1364         {
1365                 .procname       = "udp_rmem_min",
1366                 .data           = &init_net.ipv4.sysctl_udp_rmem_min,
1367                 .maxlen         = sizeof(init_net.ipv4.sysctl_udp_rmem_min),
1368                 .mode           = 0644,
1369                 .proc_handler   = proc_dointvec_minmax,
1370                 .extra1         = SYSCTL_ONE
1371         },
1372         {
1373                 .procname       = "udp_wmem_min",
1374                 .data           = &init_net.ipv4.sysctl_udp_wmem_min,
1375                 .maxlen         = sizeof(init_net.ipv4.sysctl_udp_wmem_min),
1376                 .mode           = 0644,
1377                 .proc_handler   = proc_dointvec_minmax,
1378                 .extra1         = SYSCTL_ONE
1379         },
1380         {
1381                 .procname       = "fib_notify_on_flag_change",
1382                 .data           = &init_net.ipv4.sysctl_fib_notify_on_flag_change,
1383                 .maxlen         = sizeof(u8),
1384                 .mode           = 0644,
1385                 .proc_handler   = proc_dou8vec_minmax,
1386                 .extra1         = SYSCTL_ZERO,
1387                 .extra2         = &two,
1388         },
1389         { }
1390 };
1391
1392 static __net_init int ipv4_sysctl_init_net(struct net *net)
1393 {
1394         struct ctl_table *table;
1395
1396         table = ipv4_net_table;
1397         if (!net_eq(net, &init_net)) {
1398                 int i;
1399
1400                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1401                 if (!table)
1402                         goto err_alloc;
1403
1404                 /* skip first entry (sysctl_max_tw_buckets) */
1405                 for (i = 1; i < ARRAY_SIZE(ipv4_net_table) - 1; i++) {
1406                         if (table[i].data) {
1407                                 /* Update the variables to point into
1408                                  * the current struct net
1409                                  */
1410                                 table[i].data += (void *)net - (void *)&init_net;
1411                         } else {
1412                                 /* Entries without data pointer are global;
1413                                  * Make them read-only in non-init_net ns
1414                                  */
1415                                 table[i].mode &= ~0222;
1416                         }
1417                 }
1418         }
1419
1420         table[0].data = &net->ipv4.tcp_death_row->sysctl_max_tw_buckets;
1421
1422         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1423         if (!net->ipv4.ipv4_hdr)
1424                 goto err_reg;
1425
1426         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1427         if (!net->ipv4.sysctl_local_reserved_ports)
1428                 goto err_ports;
1429
1430         return 0;
1431
1432 err_ports:
1433         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1434 err_reg:
1435         if (!net_eq(net, &init_net))
1436                 kfree(table);
1437 err_alloc:
1438         return -ENOMEM;
1439 }
1440
1441 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1442 {
1443         struct ctl_table *table;
1444
1445         kfree(net->ipv4.sysctl_local_reserved_ports);
1446         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1447         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1448         kfree(table);
1449 }
1450
1451 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1452         .init = ipv4_sysctl_init_net,
1453         .exit = ipv4_sysctl_exit_net,
1454 };
1455
1456 static __init int sysctl_ipv4_init(void)
1457 {
1458         struct ctl_table_header *hdr;
1459
1460         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1461         if (!hdr)
1462                 return -ENOMEM;
1463
1464         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1465                 unregister_net_sysctl_table(hdr);
1466                 return -ENOMEM;
1467         }
1468
1469         return 0;
1470 }
1471
1472 __initcall(sysctl_ipv4_init);