Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * Generic socket support routines. Memory allocators, socket lock/release | |
7 | * handler for protocols to use and generic option handler. | |
8 | * | |
9 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * Florian La Roche, <flla@stud.uni-sb.de> | |
13 | * Alan Cox, <A.Cox@swansea.ac.uk> | |
14 | * | |
15 | * Fixes: | |
16 | * Alan Cox : Numerous verify_area() problems | |
17 | * Alan Cox : Connecting on a connecting socket | |
18 | * now returns an error for tcp. | |
19 | * Alan Cox : sock->protocol is set correctly. | |
20 | * and is not sometimes left as 0. | |
21 | * Alan Cox : connect handles icmp errors on a | |
22 | * connect properly. Unfortunately there | |
23 | * is a restart syscall nasty there. I | |
24 | * can't match BSD without hacking the C | |
25 | * library. Ideas urgently sought! | |
26 | * Alan Cox : Disallow bind() to addresses that are | |
27 | * not ours - especially broadcast ones!! | |
28 | * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost) | |
29 | * Alan Cox : sock_wfree/sock_rfree don't destroy sockets, | |
30 | * instead they leave that for the DESTROY timer. | |
31 | * Alan Cox : Clean up error flag in accept | |
32 | * Alan Cox : TCP ack handling is buggy, the DESTROY timer | |
33 | * was buggy. Put a remove_sock() in the handler | |
34 | * for memory when we hit 0. Also altered the timer | |
4ec93edb | 35 | * code. The ACK stuff can wait and needs major |
1da177e4 LT |
36 | * TCP layer surgery. |
37 | * Alan Cox : Fixed TCP ack bug, removed remove sock | |
38 | * and fixed timer/inet_bh race. | |
39 | * Alan Cox : Added zapped flag for TCP | |
40 | * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code | |
41 | * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb | |
42 | * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources | |
43 | * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing. | |
44 | * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so... | |
45 | * Rick Sladkey : Relaxed UDP rules for matching packets. | |
46 | * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support | |
47 | * Pauline Middelink : identd support | |
48 | * Alan Cox : Fixed connect() taking signals I think. | |
49 | * Alan Cox : SO_LINGER supported | |
50 | * Alan Cox : Error reporting fixes | |
51 | * Anonymous : inet_create tidied up (sk->reuse setting) | |
52 | * Alan Cox : inet sockets don't set sk->type! | |
53 | * Alan Cox : Split socket option code | |
54 | * Alan Cox : Callbacks | |
55 | * Alan Cox : Nagle flag for Charles & Johannes stuff | |
56 | * Alex : Removed restriction on inet fioctl | |
57 | * Alan Cox : Splitting INET from NET core | |
58 | * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt() | |
59 | * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code | |
60 | * Alan Cox : Split IP from generic code | |
61 | * Alan Cox : New kfree_skbmem() | |
62 | * Alan Cox : Make SO_DEBUG superuser only. | |
63 | * Alan Cox : Allow anyone to clear SO_DEBUG | |
64 | * (compatibility fix) | |
65 | * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput. | |
66 | * Alan Cox : Allocator for a socket is settable. | |
67 | * Alan Cox : SO_ERROR includes soft errors. | |
68 | * Alan Cox : Allow NULL arguments on some SO_ opts | |
69 | * Alan Cox : Generic socket allocation to make hooks | |
70 | * easier (suggested by Craig Metz). | |
71 | * Michael Pall : SO_ERROR returns positive errno again | |
72 | * Steve Whitehouse: Added default destructor to free | |
73 | * protocol private data. | |
74 | * Steve Whitehouse: Added various other default routines | |
75 | * common to several socket families. | |
76 | * Chris Evans : Call suser() check last on F_SETOWN | |
77 | * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER. | |
78 | * Andi Kleen : Add sock_kmalloc()/sock_kfree_s() | |
79 | * Andi Kleen : Fix write_space callback | |
80 | * Chris Evans : Security fixes - signedness again | |
81 | * Arnaldo C. Melo : cleanups, use skb_queue_purge | |
82 | * | |
83 | * To Fix: | |
84 | * | |
85 | * | |
86 | * This program is free software; you can redistribute it and/or | |
87 | * modify it under the terms of the GNU General Public License | |
88 | * as published by the Free Software Foundation; either version | |
89 | * 2 of the License, or (at your option) any later version. | |
90 | */ | |
91 | ||
e005d193 JP |
92 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
93 | ||
4fc268d2 | 94 | #include <linux/capability.h> |
1da177e4 LT |
95 | #include <linux/errno.h> |
96 | #include <linux/types.h> | |
97 | #include <linux/socket.h> | |
98 | #include <linux/in.h> | |
99 | #include <linux/kernel.h> | |
1da177e4 LT |
100 | #include <linux/module.h> |
101 | #include <linux/proc_fs.h> | |
102 | #include <linux/seq_file.h> | |
103 | #include <linux/sched.h> | |
104 | #include <linux/timer.h> | |
105 | #include <linux/string.h> | |
106 | #include <linux/sockios.h> | |
107 | #include <linux/net.h> | |
108 | #include <linux/mm.h> | |
109 | #include <linux/slab.h> | |
110 | #include <linux/interrupt.h> | |
111 | #include <linux/poll.h> | |
112 | #include <linux/tcp.h> | |
113 | #include <linux/init.h> | |
a1f8e7f7 | 114 | #include <linux/highmem.h> |
3f551f94 | 115 | #include <linux/user_namespace.h> |
c5905afb | 116 | #include <linux/static_key.h> |
3969eb38 | 117 | #include <linux/memcontrol.h> |
8c1ae10d | 118 | #include <linux/prefetch.h> |
1da177e4 LT |
119 | |
120 | #include <asm/uaccess.h> | |
1da177e4 LT |
121 | |
122 | #include <linux/netdevice.h> | |
123 | #include <net/protocol.h> | |
124 | #include <linux/skbuff.h> | |
457c4cbc | 125 | #include <net/net_namespace.h> |
2e6599cb | 126 | #include <net/request_sock.h> |
1da177e4 | 127 | #include <net/sock.h> |
20d49473 | 128 | #include <linux/net_tstamp.h> |
1da177e4 LT |
129 | #include <net/xfrm.h> |
130 | #include <linux/ipsec.h> | |
f8451725 | 131 | #include <net/cls_cgroup.h> |
5bc1421e | 132 | #include <net/netprio_cgroup.h> |
1da177e4 LT |
133 | |
134 | #include <linux/filter.h> | |
135 | ||
3847ce32 SM |
136 | #include <trace/events/sock.h> |
137 | ||
6fa3eb70 S |
138 | #include <net/af_unix.h> |
139 | ||
140 | ||
1da177e4 LT |
141 | #ifdef CONFIG_INET |
142 | #include <net/tcp.h> | |
143 | #endif | |
6fa3eb70 | 144 | #include <linux/xlog.h> |
1da177e4 | 145 | |
36b77a52 | 146 | static DEFINE_MUTEX(proto_list_mutex); |
d1a4c0b3 GC |
147 | static LIST_HEAD(proto_list); |
148 | ||
c35b4e28 EB |
149 | /** |
150 | * sk_ns_capable - General socket capability test | |
151 | * @sk: Socket to use a capability on or through | |
152 | * @user_ns: The user namespace of the capability to use | |
153 | * @cap: The capability to use | |
154 | * | |
155 | * Test to see if the opener of the socket had when the socket was | |
156 | * created and the current process has the capability @cap in the user | |
157 | * namespace @user_ns. | |
158 | */ | |
159 | bool sk_ns_capable(const struct sock *sk, | |
160 | struct user_namespace *user_ns, int cap) | |
161 | { | |
162 | return file_ns_capable(sk->sk_socket->file, user_ns, cap) && | |
163 | ns_capable(user_ns, cap); | |
164 | } | |
165 | EXPORT_SYMBOL(sk_ns_capable); | |
166 | ||
167 | /** | |
168 | * sk_capable - Socket global capability test | |
169 | * @sk: Socket to use a capability on or through | |
170 | * @cap: The global capbility to use | |
171 | * | |
172 | * Test to see if the opener of the socket had when the socket was | |
173 | * created and the current process has the capability @cap in all user | |
174 | * namespaces. | |
175 | */ | |
176 | bool sk_capable(const struct sock *sk, int cap) | |
177 | { | |
178 | return sk_ns_capable(sk, &init_user_ns, cap); | |
179 | } | |
180 | EXPORT_SYMBOL(sk_capable); | |
181 | ||
182 | /** | |
183 | * sk_net_capable - Network namespace socket capability test | |
184 | * @sk: Socket to use a capability on or through | |
185 | * @cap: The capability to use | |
186 | * | |
187 | * Test to see if the opener of the socket had when the socke was created | |
188 | * and the current process has the capability @cap over the network namespace | |
189 | * the socket is a member of. | |
190 | */ | |
191 | bool sk_net_capable(const struct sock *sk, int cap) | |
192 | { | |
193 | return sk_ns_capable(sk, sock_net(sk)->user_ns, cap); | |
194 | } | |
195 | EXPORT_SYMBOL(sk_net_capable); | |
196 | ||
197 | ||
c255a458 | 198 | #ifdef CONFIG_MEMCG_KMEM |
1d62e436 | 199 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss) |
d1a4c0b3 GC |
200 | { |
201 | struct proto *proto; | |
202 | int ret = 0; | |
203 | ||
36b77a52 | 204 | mutex_lock(&proto_list_mutex); |
d1a4c0b3 GC |
205 | list_for_each_entry(proto, &proto_list, node) { |
206 | if (proto->init_cgroup) { | |
1d62e436 | 207 | ret = proto->init_cgroup(memcg, ss); |
d1a4c0b3 GC |
208 | if (ret) |
209 | goto out; | |
210 | } | |
211 | } | |
212 | ||
36b77a52 | 213 | mutex_unlock(&proto_list_mutex); |
d1a4c0b3 GC |
214 | return ret; |
215 | out: | |
216 | list_for_each_entry_continue_reverse(proto, &proto_list, node) | |
217 | if (proto->destroy_cgroup) | |
1d62e436 | 218 | proto->destroy_cgroup(memcg); |
36b77a52 | 219 | mutex_unlock(&proto_list_mutex); |
d1a4c0b3 GC |
220 | return ret; |
221 | } | |
222 | ||
1d62e436 | 223 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg) |
d1a4c0b3 GC |
224 | { |
225 | struct proto *proto; | |
226 | ||
36b77a52 | 227 | mutex_lock(&proto_list_mutex); |
d1a4c0b3 GC |
228 | list_for_each_entry_reverse(proto, &proto_list, node) |
229 | if (proto->destroy_cgroup) | |
1d62e436 | 230 | proto->destroy_cgroup(memcg); |
36b77a52 | 231 | mutex_unlock(&proto_list_mutex); |
d1a4c0b3 GC |
232 | } |
233 | #endif | |
234 | ||
da21f24d IM |
235 | /* |
236 | * Each address family might have different locking rules, so we have | |
237 | * one slock key per address family: | |
238 | */ | |
a5b5bb9a IM |
239 | static struct lock_class_key af_family_keys[AF_MAX]; |
240 | static struct lock_class_key af_family_slock_keys[AF_MAX]; | |
241 | ||
cbda4eaf | 242 | #if defined(CONFIG_MEMCG_KMEM) |
c5905afb | 243 | struct static_key memcg_socket_limit_enabled; |
e1aab161 | 244 | EXPORT_SYMBOL(memcg_socket_limit_enabled); |
cbda4eaf | 245 | #endif |
e1aab161 | 246 | |
a5b5bb9a IM |
247 | /* |
248 | * Make lock validator output more readable. (we pre-construct these | |
249 | * strings build-time, so that runtime initialization of socket | |
250 | * locks is fast): | |
251 | */ | |
36cbd3dc | 252 | static const char *const af_family_key_strings[AF_MAX+1] = { |
a5b5bb9a IM |
253 | "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" , |
254 | "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK", | |
255 | "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" , | |
256 | "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" , | |
257 | "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" , | |
258 | "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" , | |
259 | "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" , | |
cbd151bf | 260 | "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" , |
a5b5bb9a | 261 | "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" , |
cd05acfe | 262 | "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" , |
17926a79 | 263 | "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" , |
bce7b154 | 264 | "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" , |
6f107b58 | 265 | "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" , |
456db6a4 | 266 | "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX" |
a5b5bb9a | 267 | }; |
36cbd3dc | 268 | static const char *const af_family_slock_key_strings[AF_MAX+1] = { |
a5b5bb9a IM |
269 | "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" , |
270 | "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK", | |
271 | "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" , | |
272 | "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" , | |
273 | "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" , | |
274 | "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" , | |
275 | "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" , | |
cbd151bf | 276 | "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" , |
a5b5bb9a | 277 | "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" , |
cd05acfe | 278 | "slock-27" , "slock-28" , "slock-AF_CAN" , |
17926a79 | 279 | "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" , |
bce7b154 | 280 | "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" , |
6f107b58 | 281 | "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" , |
456db6a4 | 282 | "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX" |
a5b5bb9a | 283 | }; |
36cbd3dc | 284 | static const char *const af_family_clock_key_strings[AF_MAX+1] = { |
443aef0e PZ |
285 | "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" , |
286 | "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK", | |
287 | "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" , | |
288 | "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" , | |
289 | "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" , | |
290 | "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" , | |
291 | "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" , | |
cbd151bf | 292 | "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" , |
443aef0e | 293 | "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" , |
b4942af6 | 294 | "clock-27" , "clock-28" , "clock-AF_CAN" , |
e51f802b | 295 | "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" , |
bce7b154 | 296 | "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" , |
6f107b58 | 297 | "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" , |
456db6a4 | 298 | "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX" |
443aef0e | 299 | }; |
da21f24d IM |
300 | |
301 | /* | |
302 | * sk_callback_lock locking rules are per-address-family, | |
303 | * so split the lock classes by using a per-AF key: | |
304 | */ | |
305 | static struct lock_class_key af_callback_keys[AF_MAX]; | |
306 | ||
1da177e4 LT |
307 | /* Take into consideration the size of the struct sk_buff overhead in the |
308 | * determination of these values, since that is non-constant across | |
309 | * platforms. This makes socket queueing behavior and performance | |
310 | * not depend upon such differences. | |
311 | */ | |
312 | #define _SK_MEM_PACKETS 256 | |
87fb4b7b | 313 | #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256) |
1da177e4 LT |
314 | #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) |
315 | #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
316 | ||
317 | /* Run time adjustable parameters. */ | |
ab32ea5d | 318 | __u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX; |
6d8ebc8a | 319 | EXPORT_SYMBOL(sysctl_wmem_max); |
6fa3eb70 | 320 | __u32 sysctl_rmem_max __read_mostly = (SK_RMEM_MAX*8); |
6d8ebc8a | 321 | EXPORT_SYMBOL(sysctl_rmem_max); |
ab32ea5d BH |
322 | __u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX; |
323 | __u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX; | |
1da177e4 | 324 | |
25985edc | 325 | /* Maximal space eaten by iovec or ancillary data plus some space */ |
ab32ea5d | 326 | int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512); |
2a91525c | 327 | EXPORT_SYMBOL(sysctl_optmem_max); |
1da177e4 | 328 | |
c93bdd0e MG |
329 | struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE; |
330 | EXPORT_SYMBOL_GPL(memalloc_socks); | |
331 | ||
7cb02404 MG |
332 | /** |
333 | * sk_set_memalloc - sets %SOCK_MEMALLOC | |
334 | * @sk: socket to set it on | |
335 | * | |
336 | * Set %SOCK_MEMALLOC on a socket for access to emergency reserves. | |
337 | * It's the responsibility of the admin to adjust min_free_kbytes | |
338 | * to meet the requirements | |
339 | */ | |
340 | void sk_set_memalloc(struct sock *sk) | |
341 | { | |
342 | sock_set_flag(sk, SOCK_MEMALLOC); | |
343 | sk->sk_allocation |= __GFP_MEMALLOC; | |
c93bdd0e | 344 | static_key_slow_inc(&memalloc_socks); |
7cb02404 MG |
345 | } |
346 | EXPORT_SYMBOL_GPL(sk_set_memalloc); | |
347 | ||
348 | void sk_clear_memalloc(struct sock *sk) | |
349 | { | |
350 | sock_reset_flag(sk, SOCK_MEMALLOC); | |
351 | sk->sk_allocation &= ~__GFP_MEMALLOC; | |
c93bdd0e | 352 | static_key_slow_dec(&memalloc_socks); |
c76562b6 MG |
353 | |
354 | /* | |
355 | * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward | |
356 | * progress of swapping. However, if SOCK_MEMALLOC is cleared while | |
357 | * it has rmem allocations there is a risk that the user of the | |
358 | * socket cannot make forward progress due to exceeding the rmem | |
359 | * limits. By rights, sk_clear_memalloc() should only be called | |
360 | * on sockets being torn down but warn and reset the accounting if | |
361 | * that assumption breaks. | |
362 | */ | |
363 | if (WARN_ON(sk->sk_forward_alloc)) | |
364 | sk_mem_reclaim(sk); | |
7cb02404 MG |
365 | } |
366 | EXPORT_SYMBOL_GPL(sk_clear_memalloc); | |
367 | ||
b4b9e355 MG |
368 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
369 | { | |
370 | int ret; | |
371 | unsigned long pflags = current->flags; | |
372 | ||
373 | /* these should have been dropped before queueing */ | |
374 | BUG_ON(!sock_flag(sk, SOCK_MEMALLOC)); | |
375 | ||
376 | current->flags |= PF_MEMALLOC; | |
377 | ret = sk->sk_backlog_rcv(sk, skb); | |
378 | tsk_restore_flags(current, pflags, PF_MEMALLOC); | |
379 | ||
380 | return ret; | |
381 | } | |
382 | EXPORT_SYMBOL(__sk_backlog_rcv); | |
383 | ||
1da177e4 LT |
384 | static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen) |
385 | { | |
386 | struct timeval tv; | |
387 | ||
388 | if (optlen < sizeof(tv)) | |
389 | return -EINVAL; | |
390 | if (copy_from_user(&tv, optval, sizeof(tv))) | |
391 | return -EFAULT; | |
ba78073e VA |
392 | if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC) |
393 | return -EDOM; | |
1da177e4 | 394 | |
ba78073e | 395 | if (tv.tv_sec < 0) { |
6f11df83 AM |
396 | static int warned __read_mostly; |
397 | ||
ba78073e | 398 | *timeo_p = 0; |
50aab54f | 399 | if (warned < 10 && net_ratelimit()) { |
ba78073e | 400 | warned++; |
e005d193 JP |
401 | pr_info("%s: `%s' (pid %d) tries to set negative timeout\n", |
402 | __func__, current->comm, task_pid_nr(current)); | |
50aab54f | 403 | } |
ba78073e VA |
404 | return 0; |
405 | } | |
1da177e4 LT |
406 | *timeo_p = MAX_SCHEDULE_TIMEOUT; |
407 | if (tv.tv_sec == 0 && tv.tv_usec == 0) | |
408 | return 0; | |
409 | if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1)) | |
410 | *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ); | |
411 | return 0; | |
412 | } | |
413 | ||
414 | static void sock_warn_obsolete_bsdism(const char *name) | |
415 | { | |
416 | static int warned; | |
417 | static char warncomm[TASK_COMM_LEN]; | |
4ec93edb YH |
418 | if (strcmp(warncomm, current->comm) && warned < 5) { |
419 | strcpy(warncomm, current->comm); | |
e005d193 JP |
420 | pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n", |
421 | warncomm, name); | |
1da177e4 LT |
422 | warned++; |
423 | } | |
424 | } | |
425 | ||
08e29af3 | 426 | static void sock_disable_timestamp(struct sock *sk, unsigned long flags) |
4ec93edb | 427 | { |
08e29af3 ED |
428 | if (sk->sk_flags & flags) { |
429 | sk->sk_flags &= ~flags; | |
430 | if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP)) | |
20d49473 | 431 | net_disable_timestamp(); |
1da177e4 LT |
432 | } |
433 | } | |
434 | ||
435 | ||
f0088a50 DV |
436 | int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
437 | { | |
766e9037 | 438 | int err; |
f0088a50 | 439 | int skb_len; |
3b885787 NH |
440 | unsigned long flags; |
441 | struct sk_buff_head *list = &sk->sk_receive_queue; | |
f0088a50 | 442 | |
0fd7bac6 | 443 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) { |
766e9037 | 444 | atomic_inc(&sk->sk_drops); |
3847ce32 | 445 | trace_sock_rcvqueue_full(sk, skb); |
766e9037 | 446 | return -ENOMEM; |
f0088a50 DV |
447 | } |
448 | ||
fda9ef5d | 449 | err = sk_filter(sk, skb); |
f0088a50 | 450 | if (err) |
766e9037 | 451 | return err; |
f0088a50 | 452 | |
c76562b6 | 453 | if (!sk_rmem_schedule(sk, skb, skb->truesize)) { |
766e9037 ED |
454 | atomic_inc(&sk->sk_drops); |
455 | return -ENOBUFS; | |
3ab224be HA |
456 | } |
457 | ||
f0088a50 DV |
458 | skb->dev = NULL; |
459 | skb_set_owner_r(skb, sk); | |
49ad9599 | 460 | |
f0088a50 DV |
461 | /* Cache the SKB length before we tack it onto the receive |
462 | * queue. Once it is added it no longer belongs to us and | |
463 | * may be freed by other threads of control pulling packets | |
464 | * from the queue. | |
465 | */ | |
466 | skb_len = skb->len; | |
467 | ||
7fee226a ED |
468 | /* we escape from rcu protected region, make sure we dont leak |
469 | * a norefcounted dst | |
470 | */ | |
471 | skb_dst_force(skb); | |
472 | ||
3b885787 NH |
473 | spin_lock_irqsave(&list->lock, flags); |
474 | skb->dropcount = atomic_read(&sk->sk_drops); | |
475 | __skb_queue_tail(list, skb); | |
476 | spin_unlock_irqrestore(&list->lock, flags); | |
f0088a50 DV |
477 | |
478 | if (!sock_flag(sk, SOCK_DEAD)) | |
479 | sk->sk_data_ready(sk, skb_len); | |
766e9037 | 480 | return 0; |
f0088a50 DV |
481 | } |
482 | EXPORT_SYMBOL(sock_queue_rcv_skb); | |
483 | ||
58a5a7b9 | 484 | int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested) |
f0088a50 DV |
485 | { |
486 | int rc = NET_RX_SUCCESS; | |
487 | ||
fda9ef5d | 488 | if (sk_filter(sk, skb)) |
f0088a50 DV |
489 | goto discard_and_relse; |
490 | ||
491 | skb->dev = NULL; | |
492 | ||
f545a38f | 493 | if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) { |
c377411f ED |
494 | atomic_inc(&sk->sk_drops); |
495 | goto discard_and_relse; | |
496 | } | |
58a5a7b9 ACM |
497 | if (nested) |
498 | bh_lock_sock_nested(sk); | |
499 | else | |
500 | bh_lock_sock(sk); | |
a5b5bb9a IM |
501 | if (!sock_owned_by_user(sk)) { |
502 | /* | |
503 | * trylock + unlock semantics: | |
504 | */ | |
505 | mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_); | |
506 | ||
c57943a1 | 507 | rc = sk_backlog_rcv(sk, skb); |
a5b5bb9a IM |
508 | |
509 | mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_); | |
f545a38f | 510 | } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
8eae939f ZY |
511 | bh_unlock_sock(sk); |
512 | atomic_inc(&sk->sk_drops); | |
513 | goto discard_and_relse; | |
514 | } | |
515 | ||
f0088a50 DV |
516 | bh_unlock_sock(sk); |
517 | out: | |
518 | sock_put(sk); | |
519 | return rc; | |
520 | discard_and_relse: | |
521 | kfree_skb(skb); | |
522 | goto out; | |
523 | } | |
524 | EXPORT_SYMBOL(sk_receive_skb); | |
525 | ||
ea94ff3b KK |
526 | void sk_reset_txq(struct sock *sk) |
527 | { | |
528 | sk_tx_queue_clear(sk); | |
529 | } | |
530 | EXPORT_SYMBOL(sk_reset_txq); | |
531 | ||
f0088a50 DV |
532 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie) |
533 | { | |
b6c6712a | 534 | struct dst_entry *dst = __sk_dst_get(sk); |
f0088a50 DV |
535 | |
536 | if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { | |
e022f0b4 | 537 | sk_tx_queue_clear(sk); |
a9b3cd7f | 538 | RCU_INIT_POINTER(sk->sk_dst_cache, NULL); |
f0088a50 DV |
539 | dst_release(dst); |
540 | return NULL; | |
541 | } | |
542 | ||
543 | return dst; | |
544 | } | |
545 | EXPORT_SYMBOL(__sk_dst_check); | |
546 | ||
547 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie) | |
548 | { | |
549 | struct dst_entry *dst = sk_dst_get(sk); | |
550 | ||
551 | if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { | |
552 | sk_dst_reset(sk); | |
553 | dst_release(dst); | |
554 | return NULL; | |
555 | } | |
556 | ||
557 | return dst; | |
558 | } | |
559 | EXPORT_SYMBOL(sk_dst_check); | |
560 | ||
c91f6df2 BH |
561 | static int sock_setbindtodevice(struct sock *sk, char __user *optval, |
562 | int optlen) | |
4878809f DM |
563 | { |
564 | int ret = -ENOPROTOOPT; | |
565 | #ifdef CONFIG_NETDEVICES | |
3b1e0a65 | 566 | struct net *net = sock_net(sk); |
4878809f DM |
567 | char devname[IFNAMSIZ]; |
568 | int index; | |
569 | ||
570 | /* Sorry... */ | |
571 | ret = -EPERM; | |
5e1fccc0 | 572 | if (!ns_capable(net->user_ns, CAP_NET_RAW)) |
4878809f DM |
573 | goto out; |
574 | ||
575 | ret = -EINVAL; | |
576 | if (optlen < 0) | |
577 | goto out; | |
578 | ||
579 | /* Bind this socket to a particular device like "eth0", | |
580 | * as specified in the passed interface name. If the | |
581 | * name is "" or the option length is zero the socket | |
582 | * is not bound. | |
583 | */ | |
584 | if (optlen > IFNAMSIZ - 1) | |
585 | optlen = IFNAMSIZ - 1; | |
586 | memset(devname, 0, sizeof(devname)); | |
587 | ||
588 | ret = -EFAULT; | |
589 | if (copy_from_user(devname, optval, optlen)) | |
590 | goto out; | |
591 | ||
000ba2e4 DM |
592 | index = 0; |
593 | if (devname[0] != '\0') { | |
bf8e56bf | 594 | struct net_device *dev; |
4878809f | 595 | |
bf8e56bf ED |
596 | rcu_read_lock(); |
597 | dev = dev_get_by_name_rcu(net, devname); | |
598 | if (dev) | |
599 | index = dev->ifindex; | |
600 | rcu_read_unlock(); | |
4878809f DM |
601 | ret = -ENODEV; |
602 | if (!dev) | |
603 | goto out; | |
4878809f DM |
604 | } |
605 | ||
606 | lock_sock(sk); | |
607 | sk->sk_bound_dev_if = index; | |
608 | sk_dst_reset(sk); | |
609 | release_sock(sk); | |
610 | ||
611 | ret = 0; | |
612 | ||
613 | out: | |
614 | #endif | |
615 | ||
616 | return ret; | |
617 | } | |
618 | ||
c91f6df2 BH |
619 | static int sock_getbindtodevice(struct sock *sk, char __user *optval, |
620 | int __user *optlen, int len) | |
621 | { | |
622 | int ret = -ENOPROTOOPT; | |
623 | #ifdef CONFIG_NETDEVICES | |
624 | struct net *net = sock_net(sk); | |
c91f6df2 | 625 | char devname[IFNAMSIZ]; |
c91f6df2 BH |
626 | |
627 | if (sk->sk_bound_dev_if == 0) { | |
628 | len = 0; | |
629 | goto zero; | |
630 | } | |
631 | ||
632 | ret = -EINVAL; | |
633 | if (len < IFNAMSIZ) | |
634 | goto out; | |
635 | ||
5dbe7c17 NS |
636 | ret = netdev_get_name(net, devname, sk->sk_bound_dev_if); |
637 | if (ret) | |
c91f6df2 | 638 | goto out; |
c91f6df2 BH |
639 | |
640 | len = strlen(devname) + 1; | |
641 | ||
642 | ret = -EFAULT; | |
643 | if (copy_to_user(optval, devname, len)) | |
644 | goto out; | |
645 | ||
646 | zero: | |
647 | ret = -EFAULT; | |
648 | if (put_user(len, optlen)) | |
649 | goto out; | |
650 | ||
651 | ret = 0; | |
652 | ||
653 | out: | |
654 | #endif | |
655 | ||
656 | return ret; | |
657 | } | |
658 | ||
c0ef877b PE |
659 | static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool) |
660 | { | |
661 | if (valbool) | |
662 | sock_set_flag(sk, bit); | |
663 | else | |
664 | sock_reset_flag(sk, bit); | |
665 | } | |
666 | ||
1da177e4 LT |
667 | /* |
668 | * This is meant for all protocols to use and covers goings on | |
669 | * at the socket level. Everything here is generic. | |
670 | */ | |
671 | ||
672 | int sock_setsockopt(struct socket *sock, int level, int optname, | |
b7058842 | 673 | char __user *optval, unsigned int optlen) |
1da177e4 | 674 | { |
2a91525c | 675 | struct sock *sk = sock->sk; |
1da177e4 LT |
676 | int val; |
677 | int valbool; | |
678 | struct linger ling; | |
679 | int ret = 0; | |
4ec93edb | 680 | |
1da177e4 LT |
681 | /* |
682 | * Options without arguments | |
683 | */ | |
684 | ||
4878809f | 685 | if (optname == SO_BINDTODEVICE) |
c91f6df2 | 686 | return sock_setbindtodevice(sk, optval, optlen); |
4878809f | 687 | |
e71a4783 SH |
688 | if (optlen < sizeof(int)) |
689 | return -EINVAL; | |
4ec93edb | 690 | |
1da177e4 LT |
691 | if (get_user(val, (int __user *)optval)) |
692 | return -EFAULT; | |
4ec93edb | 693 | |
2a91525c | 694 | valbool = val ? 1 : 0; |
1da177e4 LT |
695 | |
696 | lock_sock(sk); | |
697 | ||
2a91525c | 698 | switch (optname) { |
e71a4783 | 699 | case SO_DEBUG: |
2a91525c | 700 | if (val && !capable(CAP_NET_ADMIN)) |
e71a4783 | 701 | ret = -EACCES; |
2a91525c | 702 | else |
c0ef877b | 703 | sock_valbool_flag(sk, SOCK_DBG, valbool); |
e71a4783 SH |
704 | break; |
705 | case SO_REUSEADDR: | |
4a17fd52 | 706 | sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE); |
e71a4783 | 707 | break; |
055dc21a TH |
708 | case SO_REUSEPORT: |
709 | sk->sk_reuseport = valbool; | |
710 | break; | |
e71a4783 | 711 | case SO_TYPE: |
49c794e9 | 712 | case SO_PROTOCOL: |
0d6038ee | 713 | case SO_DOMAIN: |
e71a4783 SH |
714 | case SO_ERROR: |
715 | ret = -ENOPROTOOPT; | |
716 | break; | |
717 | case SO_DONTROUTE: | |
c0ef877b | 718 | sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool); |
e71a4783 SH |
719 | break; |
720 | case SO_BROADCAST: | |
721 | sock_valbool_flag(sk, SOCK_BROADCAST, valbool); | |
722 | break; | |
723 | case SO_SNDBUF: | |
724 | /* Don't error on this BSD doesn't and if you think | |
82981930 ED |
725 | * about it this is right. Otherwise apps have to |
726 | * play 'guess the biggest size' games. RCVBUF/SNDBUF | |
727 | * are treated in BSD as hints | |
728 | */ | |
729 | val = min_t(u32, val, sysctl_wmem_max); | |
b0573dea | 730 | set_sndbuf: |
e71a4783 | 731 | sk->sk_userlocks |= SOCK_SNDBUF_LOCK; |
4b9e9796 | 732 | sk->sk_sndbuf = max_t(int, val * 2, SOCK_MIN_SNDBUF); |
82981930 | 733 | /* Wake up sending tasks if we upped the value. */ |
e71a4783 SH |
734 | sk->sk_write_space(sk); |
735 | break; | |
1da177e4 | 736 | |
e71a4783 SH |
737 | case SO_SNDBUFFORCE: |
738 | if (!capable(CAP_NET_ADMIN)) { | |
739 | ret = -EPERM; | |
740 | break; | |
741 | } | |
742 | goto set_sndbuf; | |
b0573dea | 743 | |
e71a4783 SH |
744 | case SO_RCVBUF: |
745 | /* Don't error on this BSD doesn't and if you think | |
82981930 ED |
746 | * about it this is right. Otherwise apps have to |
747 | * play 'guess the biggest size' games. RCVBUF/SNDBUF | |
748 | * are treated in BSD as hints | |
749 | */ | |
750 | val = min_t(u32, val, sysctl_rmem_max); | |
b0573dea | 751 | set_rcvbuf: |
e71a4783 SH |
752 | sk->sk_userlocks |= SOCK_RCVBUF_LOCK; |
753 | /* | |
754 | * We double it on the way in to account for | |
755 | * "struct sk_buff" etc. overhead. Applications | |
756 | * assume that the SO_RCVBUF setting they make will | |
757 | * allow that much actual data to be received on that | |
758 | * socket. | |
759 | * | |
760 | * Applications are unaware that "struct sk_buff" and | |
761 | * other overheads allocate from the receive buffer | |
762 | * during socket buffer allocation. | |
763 | * | |
764 | * And after considering the possible alternatives, | |
765 | * returning the value we actually used in getsockopt | |
766 | * is the most desirable behavior. | |
767 | */ | |
4b9e9796 | 768 | sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF); |
e71a4783 SH |
769 | break; |
770 | ||
771 | case SO_RCVBUFFORCE: | |
772 | if (!capable(CAP_NET_ADMIN)) { | |
773 | ret = -EPERM; | |
1da177e4 | 774 | break; |
e71a4783 SH |
775 | } |
776 | goto set_rcvbuf; | |
1da177e4 | 777 | |
e71a4783 | 778 | case SO_KEEPALIVE: |
1da177e4 | 779 | #ifdef CONFIG_INET |
3e10986d ED |
780 | if (sk->sk_protocol == IPPROTO_TCP && |
781 | sk->sk_type == SOCK_STREAM) | |
e71a4783 | 782 | tcp_set_keepalive(sk, valbool); |
1da177e4 | 783 | #endif |
e71a4783 SH |
784 | sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool); |
785 | break; | |
786 | ||
787 | case SO_OOBINLINE: | |
788 | sock_valbool_flag(sk, SOCK_URGINLINE, valbool); | |
789 | break; | |
790 | ||
791 | case SO_NO_CHECK: | |
792 | sk->sk_no_check = valbool; | |
793 | break; | |
794 | ||
795 | case SO_PRIORITY: | |
5e1fccc0 EB |
796 | if ((val >= 0 && val <= 6) || |
797 | ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) | |
e71a4783 SH |
798 | sk->sk_priority = val; |
799 | else | |
800 | ret = -EPERM; | |
801 | break; | |
802 | ||
803 | case SO_LINGER: | |
804 | if (optlen < sizeof(ling)) { | |
805 | ret = -EINVAL; /* 1003.1g */ | |
1da177e4 | 806 | break; |
e71a4783 | 807 | } |
2a91525c | 808 | if (copy_from_user(&ling, optval, sizeof(ling))) { |
e71a4783 | 809 | ret = -EFAULT; |
1da177e4 | 810 | break; |
e71a4783 SH |
811 | } |
812 | if (!ling.l_onoff) | |
813 | sock_reset_flag(sk, SOCK_LINGER); | |
814 | else { | |
1da177e4 | 815 | #if (BITS_PER_LONG == 32) |
e71a4783 SH |
816 | if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ) |
817 | sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT; | |
1da177e4 | 818 | else |
e71a4783 SH |
819 | #endif |
820 | sk->sk_lingertime = (unsigned int)ling.l_linger * HZ; | |
821 | sock_set_flag(sk, SOCK_LINGER); | |
822 | } | |
823 | break; | |
824 | ||
825 | case SO_BSDCOMPAT: | |
826 | sock_warn_obsolete_bsdism("setsockopt"); | |
827 | break; | |
828 | ||
829 | case SO_PASSCRED: | |
830 | if (valbool) | |
831 | set_bit(SOCK_PASSCRED, &sock->flags); | |
832 | else | |
833 | clear_bit(SOCK_PASSCRED, &sock->flags); | |
834 | break; | |
835 | ||
836 | case SO_TIMESTAMP: | |
92f37fd2 | 837 | case SO_TIMESTAMPNS: |
e71a4783 | 838 | if (valbool) { |
92f37fd2 ED |
839 | if (optname == SO_TIMESTAMP) |
840 | sock_reset_flag(sk, SOCK_RCVTSTAMPNS); | |
841 | else | |
842 | sock_set_flag(sk, SOCK_RCVTSTAMPNS); | |
e71a4783 | 843 | sock_set_flag(sk, SOCK_RCVTSTAMP); |
20d49473 | 844 | sock_enable_timestamp(sk, SOCK_TIMESTAMP); |
92f37fd2 | 845 | } else { |
e71a4783 | 846 | sock_reset_flag(sk, SOCK_RCVTSTAMP); |
92f37fd2 ED |
847 | sock_reset_flag(sk, SOCK_RCVTSTAMPNS); |
848 | } | |
e71a4783 SH |
849 | break; |
850 | ||
20d49473 PO |
851 | case SO_TIMESTAMPING: |
852 | if (val & ~SOF_TIMESTAMPING_MASK) { | |
f249fb78 | 853 | ret = -EINVAL; |
20d49473 PO |
854 | break; |
855 | } | |
856 | sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE, | |
857 | val & SOF_TIMESTAMPING_TX_HARDWARE); | |
858 | sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE, | |
859 | val & SOF_TIMESTAMPING_TX_SOFTWARE); | |
860 | sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE, | |
861 | val & SOF_TIMESTAMPING_RX_HARDWARE); | |
862 | if (val & SOF_TIMESTAMPING_RX_SOFTWARE) | |
863 | sock_enable_timestamp(sk, | |
864 | SOCK_TIMESTAMPING_RX_SOFTWARE); | |
865 | else | |
866 | sock_disable_timestamp(sk, | |
08e29af3 | 867 | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)); |
20d49473 PO |
868 | sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE, |
869 | val & SOF_TIMESTAMPING_SOFTWARE); | |
870 | sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE, | |
871 | val & SOF_TIMESTAMPING_SYS_HARDWARE); | |
872 | sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE, | |
873 | val & SOF_TIMESTAMPING_RAW_HARDWARE); | |
874 | break; | |
875 | ||
e71a4783 SH |
876 | case SO_RCVLOWAT: |
877 | if (val < 0) | |
878 | val = INT_MAX; | |
879 | sk->sk_rcvlowat = val ? : 1; | |
880 | break; | |
881 | ||
882 | case SO_RCVTIMEO: | |
883 | ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen); | |
884 | break; | |
885 | ||
886 | case SO_SNDTIMEO: | |
887 | ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen); | |
888 | break; | |
1da177e4 | 889 | |
e71a4783 SH |
890 | case SO_ATTACH_FILTER: |
891 | ret = -EINVAL; | |
892 | if (optlen == sizeof(struct sock_fprog)) { | |
893 | struct sock_fprog fprog; | |
1da177e4 | 894 | |
e71a4783 SH |
895 | ret = -EFAULT; |
896 | if (copy_from_user(&fprog, optval, sizeof(fprog))) | |
1da177e4 | 897 | break; |
e71a4783 SH |
898 | |
899 | ret = sk_attach_filter(&fprog, sk); | |
900 | } | |
901 | break; | |
902 | ||
903 | case SO_DETACH_FILTER: | |
55b33325 | 904 | ret = sk_detach_filter(sk); |
e71a4783 | 905 | break; |
1da177e4 | 906 | |
d59577b6 VB |
907 | case SO_LOCK_FILTER: |
908 | if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool) | |
909 | ret = -EPERM; | |
910 | else | |
911 | sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool); | |
912 | break; | |
913 | ||
e71a4783 SH |
914 | case SO_PASSSEC: |
915 | if (valbool) | |
916 | set_bit(SOCK_PASSSEC, &sock->flags); | |
917 | else | |
918 | clear_bit(SOCK_PASSSEC, &sock->flags); | |
919 | break; | |
4a19ec58 | 920 | case SO_MARK: |
5e1fccc0 | 921 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) |
4a19ec58 | 922 | ret = -EPERM; |
2a91525c | 923 | else |
4a19ec58 | 924 | sk->sk_mark = val; |
4a19ec58 | 925 | break; |
877ce7c1 | 926 | |
1da177e4 LT |
927 | /* We implement the SO_SNDLOWAT etc to |
928 | not be settable (1003.1g 5.3) */ | |
3b885787 | 929 | case SO_RXQ_OVFL: |
8083f0fc | 930 | sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool); |
3b885787 | 931 | break; |
6e3e939f JB |
932 | |
933 | case SO_WIFI_STATUS: | |
934 | sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool); | |
935 | break; | |
936 | ||
ef64a54f PE |
937 | case SO_PEEK_OFF: |
938 | if (sock->ops->set_peek_off) | |
d90d9ff6 | 939 | ret = sock->ops->set_peek_off(sk, val); |
ef64a54f PE |
940 | else |
941 | ret = -EOPNOTSUPP; | |
942 | break; | |
3bdc0eba BG |
943 | |
944 | case SO_NOFCS: | |
945 | sock_valbool_flag(sk, SOCK_NOFCS, valbool); | |
946 | break; | |
947 | ||
7d4c04fc KJ |
948 | case SO_SELECT_ERR_QUEUE: |
949 | sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool); | |
950 | break; | |
951 | ||
e71a4783 SH |
952 | default: |
953 | ret = -ENOPROTOOPT; | |
954 | break; | |
4ec93edb | 955 | } |
1da177e4 LT |
956 | release_sock(sk); |
957 | return ret; | |
958 | } | |
2a91525c | 959 | EXPORT_SYMBOL(sock_setsockopt); |
1da177e4 LT |
960 | |
961 | ||
3f551f94 EB |
962 | void cred_to_ucred(struct pid *pid, const struct cred *cred, |
963 | struct ucred *ucred) | |
964 | { | |
965 | ucred->pid = pid_vnr(pid); | |
966 | ucred->uid = ucred->gid = -1; | |
967 | if (cred) { | |
968 | struct user_namespace *current_ns = current_user_ns(); | |
969 | ||
b2e4f544 EB |
970 | ucred->uid = from_kuid_munged(current_ns, cred->euid); |
971 | ucred->gid = from_kgid_munged(current_ns, cred->egid); | |
3f551f94 EB |
972 | } |
973 | } | |
3924773a | 974 | EXPORT_SYMBOL_GPL(cred_to_ucred); |
3f551f94 | 975 | |
1da177e4 LT |
976 | int sock_getsockopt(struct socket *sock, int level, int optname, |
977 | char __user *optval, int __user *optlen) | |
978 | { | |
979 | struct sock *sk = sock->sk; | |
4ec93edb | 980 | |
e71a4783 | 981 | union { |
4ec93edb YH |
982 | int val; |
983 | struct linger ling; | |
1da177e4 LT |
984 | struct timeval tm; |
985 | } v; | |
4ec93edb | 986 | |
4d0392be | 987 | int lv = sizeof(int); |
1da177e4 | 988 | int len; |
4ec93edb | 989 | |
e71a4783 | 990 | if (get_user(len, optlen)) |
4ec93edb | 991 | return -EFAULT; |
e71a4783 | 992 | if (len < 0) |
1da177e4 | 993 | return -EINVAL; |
4ec93edb | 994 | |
50fee1de | 995 | memset(&v, 0, sizeof(v)); |
df0bca04 | 996 | |
2a91525c | 997 | switch (optname) { |
e71a4783 SH |
998 | case SO_DEBUG: |
999 | v.val = sock_flag(sk, SOCK_DBG); | |
1000 | break; | |
1001 | ||
1002 | case SO_DONTROUTE: | |
1003 | v.val = sock_flag(sk, SOCK_LOCALROUTE); | |
1004 | break; | |
1005 | ||
1006 | case SO_BROADCAST: | |
1b23a5df | 1007 | v.val = sock_flag(sk, SOCK_BROADCAST); |
e71a4783 SH |
1008 | break; |
1009 | ||
1010 | case SO_SNDBUF: | |
1011 | v.val = sk->sk_sndbuf; | |
1012 | break; | |
1013 | ||
1014 | case SO_RCVBUF: | |
1015 | v.val = sk->sk_rcvbuf; | |
1016 | break; | |
1017 | ||
1018 | case SO_REUSEADDR: | |
1019 | v.val = sk->sk_reuse; | |
1020 | break; | |
1021 | ||
055dc21a TH |
1022 | case SO_REUSEPORT: |
1023 | v.val = sk->sk_reuseport; | |
1024 | break; | |
1025 | ||
e71a4783 | 1026 | case SO_KEEPALIVE: |
1b23a5df | 1027 | v.val = sock_flag(sk, SOCK_KEEPOPEN); |
e71a4783 SH |
1028 | break; |
1029 | ||
1030 | case SO_TYPE: | |
1031 | v.val = sk->sk_type; | |
1032 | break; | |
1033 | ||
49c794e9 JE |
1034 | case SO_PROTOCOL: |
1035 | v.val = sk->sk_protocol; | |
1036 | break; | |
1037 | ||
0d6038ee JE |
1038 | case SO_DOMAIN: |
1039 | v.val = sk->sk_family; | |
1040 | break; | |
1041 | ||
e71a4783 SH |
1042 | case SO_ERROR: |
1043 | v.val = -sock_error(sk); | |
2a91525c | 1044 | if (v.val == 0) |
e71a4783 SH |
1045 | v.val = xchg(&sk->sk_err_soft, 0); |
1046 | break; | |
1047 | ||
1048 | case SO_OOBINLINE: | |
1b23a5df | 1049 | v.val = sock_flag(sk, SOCK_URGINLINE); |
e71a4783 SH |
1050 | break; |
1051 | ||
1052 | case SO_NO_CHECK: | |
1053 | v.val = sk->sk_no_check; | |
1054 | break; | |
1055 | ||
1056 | case SO_PRIORITY: | |
1057 | v.val = sk->sk_priority; | |
1058 | break; | |
1059 | ||
1060 | case SO_LINGER: | |
1061 | lv = sizeof(v.ling); | |
1b23a5df | 1062 | v.ling.l_onoff = sock_flag(sk, SOCK_LINGER); |
e71a4783 SH |
1063 | v.ling.l_linger = sk->sk_lingertime / HZ; |
1064 | break; | |
1065 | ||
1066 | case SO_BSDCOMPAT: | |
1067 | sock_warn_obsolete_bsdism("getsockopt"); | |
1068 | break; | |
1069 | ||
1070 | case SO_TIMESTAMP: | |
92f37fd2 ED |
1071 | v.val = sock_flag(sk, SOCK_RCVTSTAMP) && |
1072 | !sock_flag(sk, SOCK_RCVTSTAMPNS); | |
1073 | break; | |
1074 | ||
1075 | case SO_TIMESTAMPNS: | |
1076 | v.val = sock_flag(sk, SOCK_RCVTSTAMPNS); | |
e71a4783 SH |
1077 | break; |
1078 | ||
20d49473 PO |
1079 | case SO_TIMESTAMPING: |
1080 | v.val = 0; | |
1081 | if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE)) | |
1082 | v.val |= SOF_TIMESTAMPING_TX_HARDWARE; | |
1083 | if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE)) | |
1084 | v.val |= SOF_TIMESTAMPING_TX_SOFTWARE; | |
1085 | if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE)) | |
1086 | v.val |= SOF_TIMESTAMPING_RX_HARDWARE; | |
1087 | if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE)) | |
1088 | v.val |= SOF_TIMESTAMPING_RX_SOFTWARE; | |
1089 | if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) | |
1090 | v.val |= SOF_TIMESTAMPING_SOFTWARE; | |
1091 | if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)) | |
1092 | v.val |= SOF_TIMESTAMPING_SYS_HARDWARE; | |
1093 | if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) | |
1094 | v.val |= SOF_TIMESTAMPING_RAW_HARDWARE; | |
1095 | break; | |
1096 | ||
e71a4783 | 1097 | case SO_RCVTIMEO: |
2a91525c | 1098 | lv = sizeof(struct timeval); |
e71a4783 SH |
1099 | if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) { |
1100 | v.tm.tv_sec = 0; | |
1101 | v.tm.tv_usec = 0; | |
1102 | } else { | |
1103 | v.tm.tv_sec = sk->sk_rcvtimeo / HZ; | |
1104 | v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ; | |
1105 | } | |
1106 | break; | |
1107 | ||
1108 | case SO_SNDTIMEO: | |
2a91525c | 1109 | lv = sizeof(struct timeval); |
e71a4783 SH |
1110 | if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) { |
1111 | v.tm.tv_sec = 0; | |
1112 | v.tm.tv_usec = 0; | |
1113 | } else { | |
1114 | v.tm.tv_sec = sk->sk_sndtimeo / HZ; | |
1115 | v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ; | |
1116 | } | |
1117 | break; | |
1da177e4 | 1118 | |
e71a4783 SH |
1119 | case SO_RCVLOWAT: |
1120 | v.val = sk->sk_rcvlowat; | |
1121 | break; | |
1da177e4 | 1122 | |
e71a4783 | 1123 | case SO_SNDLOWAT: |
2a91525c | 1124 | v.val = 1; |
e71a4783 | 1125 | break; |
1da177e4 | 1126 | |
e71a4783 | 1127 | case SO_PASSCRED: |
82981930 | 1128 | v.val = !!test_bit(SOCK_PASSCRED, &sock->flags); |
e71a4783 | 1129 | break; |
1da177e4 | 1130 | |
e71a4783 | 1131 | case SO_PEERCRED: |
109f6e39 EB |
1132 | { |
1133 | struct ucred peercred; | |
1134 | if (len > sizeof(peercred)) | |
1135 | len = sizeof(peercred); | |
1136 | cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred); | |
1137 | if (copy_to_user(optval, &peercred, len)) | |
e71a4783 SH |
1138 | return -EFAULT; |
1139 | goto lenout; | |
109f6e39 | 1140 | } |
1da177e4 | 1141 | |
e71a4783 SH |
1142 | case SO_PEERNAME: |
1143 | { | |
1144 | char address[128]; | |
1145 | ||
1146 | if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2)) | |
1147 | return -ENOTCONN; | |
1148 | if (lv < len) | |
1149 | return -EINVAL; | |
1150 | if (copy_to_user(optval, address, len)) | |
1151 | return -EFAULT; | |
1152 | goto lenout; | |
1153 | } | |
1da177e4 | 1154 | |
e71a4783 SH |
1155 | /* Dubious BSD thing... Probably nobody even uses it, but |
1156 | * the UNIX standard wants it for whatever reason... -DaveM | |
1157 | */ | |
1158 | case SO_ACCEPTCONN: | |
1159 | v.val = sk->sk_state == TCP_LISTEN; | |
1160 | break; | |
1da177e4 | 1161 | |
e71a4783 | 1162 | case SO_PASSSEC: |
82981930 | 1163 | v.val = !!test_bit(SOCK_PASSSEC, &sock->flags); |
e71a4783 | 1164 | break; |
877ce7c1 | 1165 | |
e71a4783 SH |
1166 | case SO_PEERSEC: |
1167 | return security_socket_getpeersec_stream(sock, optval, optlen, len); | |
1da177e4 | 1168 | |
4a19ec58 LAT |
1169 | case SO_MARK: |
1170 | v.val = sk->sk_mark; | |
1171 | break; | |
1172 | ||
3b885787 | 1173 | case SO_RXQ_OVFL: |
1b23a5df | 1174 | v.val = sock_flag(sk, SOCK_RXQ_OVFL); |
3b885787 NH |
1175 | break; |
1176 | ||
6e3e939f | 1177 | case SO_WIFI_STATUS: |
1b23a5df | 1178 | v.val = sock_flag(sk, SOCK_WIFI_STATUS); |
6e3e939f JB |
1179 | break; |
1180 | ||
ef64a54f PE |
1181 | case SO_PEEK_OFF: |
1182 | if (!sock->ops->set_peek_off) | |
1183 | return -EOPNOTSUPP; | |
1184 | ||
1185 | v.val = sk->sk_peek_off; | |
1186 | break; | |
bc2f7996 | 1187 | case SO_NOFCS: |
1b23a5df | 1188 | v.val = sock_flag(sk, SOCK_NOFCS); |
bc2f7996 | 1189 | break; |
c91f6df2 | 1190 | |
f7b86bfe | 1191 | case SO_BINDTODEVICE: |
c91f6df2 BH |
1192 | return sock_getbindtodevice(sk, optval, optlen, len); |
1193 | ||
a8fc9277 PE |
1194 | case SO_GET_FILTER: |
1195 | len = sk_get_filter(sk, (struct sock_filter __user *)optval, len); | |
1196 | if (len < 0) | |
1197 | return len; | |
1198 | ||
1199 | goto lenout; | |
c91f6df2 | 1200 | |
d59577b6 VB |
1201 | case SO_LOCK_FILTER: |
1202 | v.val = sock_flag(sk, SOCK_FILTER_LOCKED); | |
1203 | break; | |
1204 | ||
7d4c04fc KJ |
1205 | case SO_SELECT_ERR_QUEUE: |
1206 | v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE); | |
1207 | break; | |
1208 | ||
e71a4783 SH |
1209 | default: |
1210 | return -ENOPROTOOPT; | |
1da177e4 | 1211 | } |
e71a4783 | 1212 | |
1da177e4 LT |
1213 | if (len > lv) |
1214 | len = lv; | |
1215 | if (copy_to_user(optval, &v, len)) | |
1216 | return -EFAULT; | |
1217 | lenout: | |
4ec93edb YH |
1218 | if (put_user(len, optlen)) |
1219 | return -EFAULT; | |
1220 | return 0; | |
1da177e4 LT |
1221 | } |
1222 | ||
a5b5bb9a IM |
1223 | /* |
1224 | * Initialize an sk_lock. | |
1225 | * | |
1226 | * (We also register the sk_lock with the lock validator.) | |
1227 | */ | |
b6f99a21 | 1228 | static inline void sock_lock_init(struct sock *sk) |
a5b5bb9a | 1229 | { |
ed07536e PZ |
1230 | sock_lock_init_class_and_name(sk, |
1231 | af_family_slock_key_strings[sk->sk_family], | |
1232 | af_family_slock_keys + sk->sk_family, | |
1233 | af_family_key_strings[sk->sk_family], | |
1234 | af_family_keys + sk->sk_family); | |
a5b5bb9a IM |
1235 | } |
1236 | ||
4dc6dc71 ED |
1237 | /* |
1238 | * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet, | |
1239 | * even temporarly, because of RCU lookups. sk_node should also be left as is. | |
68835aba | 1240 | * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end |
4dc6dc71 | 1241 | */ |
f1a6c4da PE |
1242 | static void sock_copy(struct sock *nsk, const struct sock *osk) |
1243 | { | |
1244 | #ifdef CONFIG_SECURITY_NETWORK | |
1245 | void *sptr = nsk->sk_security; | |
1246 | #endif | |
68835aba ED |
1247 | memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin)); |
1248 | ||
1249 | memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end, | |
1250 | osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end)); | |
1251 | ||
f1a6c4da PE |
1252 | #ifdef CONFIG_SECURITY_NETWORK |
1253 | nsk->sk_security = sptr; | |
1254 | security_sk_clone(osk, nsk); | |
1255 | #endif | |
1256 | } | |
1257 | ||
fcbdf09d OP |
1258 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size) |
1259 | { | |
1260 | unsigned long nulls1, nulls2; | |
1261 | ||
1262 | nulls1 = offsetof(struct sock, __sk_common.skc_node.next); | |
1263 | nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next); | |
1264 | if (nulls1 > nulls2) | |
1265 | swap(nulls1, nulls2); | |
1266 | ||
1267 | if (nulls1 != 0) | |
1268 | memset((char *)sk, 0, nulls1); | |
1269 | memset((char *)sk + nulls1 + sizeof(void *), 0, | |
1270 | nulls2 - nulls1 - sizeof(void *)); | |
1271 | memset((char *)sk + nulls2 + sizeof(void *), 0, | |
1272 | size - nulls2 - sizeof(void *)); | |
1273 | } | |
1274 | EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls); | |
1275 | ||
2e4afe7b PE |
1276 | static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority, |
1277 | int family) | |
c308c1b2 PE |
1278 | { |
1279 | struct sock *sk; | |
1280 | struct kmem_cache *slab; | |
1281 | ||
1282 | slab = prot->slab; | |
e912b114 ED |
1283 | if (slab != NULL) { |
1284 | sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO); | |
1285 | if (!sk) | |
1286 | return sk; | |
1287 | if (priority & __GFP_ZERO) { | |
fcbdf09d OP |
1288 | if (prot->clear_sk) |
1289 | prot->clear_sk(sk, prot->obj_size); | |
1290 | else | |
1291 | sk_prot_clear_nulls(sk, prot->obj_size); | |
e912b114 | 1292 | } |
fcbdf09d | 1293 | } else |
c308c1b2 PE |
1294 | sk = kmalloc(prot->obj_size, priority); |
1295 | ||
2e4afe7b | 1296 | if (sk != NULL) { |
a98b65a3 VN |
1297 | kmemcheck_annotate_bitfield(sk, flags); |
1298 | ||
2e4afe7b PE |
1299 | if (security_sk_alloc(sk, family, priority)) |
1300 | goto out_free; | |
1301 | ||
1302 | if (!try_module_get(prot->owner)) | |
1303 | goto out_free_sec; | |
e022f0b4 | 1304 | sk_tx_queue_clear(sk); |
2e4afe7b PE |
1305 | } |
1306 | ||
c308c1b2 | 1307 | return sk; |
2e4afe7b PE |
1308 | |
1309 | out_free_sec: | |
1310 | security_sk_free(sk); | |
1311 | out_free: | |
1312 | if (slab != NULL) | |
1313 | kmem_cache_free(slab, sk); | |
1314 | else | |
1315 | kfree(sk); | |
1316 | return NULL; | |
c308c1b2 PE |
1317 | } |
1318 | ||
1319 | static void sk_prot_free(struct proto *prot, struct sock *sk) | |
1320 | { | |
1321 | struct kmem_cache *slab; | |
2e4afe7b | 1322 | struct module *owner; |
c308c1b2 | 1323 | |
2e4afe7b | 1324 | owner = prot->owner; |
c308c1b2 | 1325 | slab = prot->slab; |
2e4afe7b PE |
1326 | |
1327 | security_sk_free(sk); | |
c308c1b2 PE |
1328 | if (slab != NULL) |
1329 | kmem_cache_free(slab, sk); | |
1330 | else | |
1331 | kfree(sk); | |
2e4afe7b | 1332 | module_put(owner); |
c308c1b2 PE |
1333 | } |
1334 | ||
8fb974c9 | 1335 | #if IS_ENABLED(CONFIG_NET_CLS_CGROUP) |
211d2f97 | 1336 | void sock_update_classid(struct sock *sk) |
f8451725 | 1337 | { |
1144182a | 1338 | u32 classid; |
f8451725 | 1339 | |
211d2f97 | 1340 | classid = task_cls_classid(current); |
3afa6d00 | 1341 | if (classid != sk->sk_classid) |
f8451725 HX |
1342 | sk->sk_classid = classid; |
1343 | } | |
82862742 | 1344 | EXPORT_SYMBOL(sock_update_classid); |
8fb974c9 | 1345 | #endif |
5bc1421e | 1346 | |
51e4e7fa | 1347 | #if IS_ENABLED(CONFIG_NETPRIO_CGROUP) |
6ffd4641 | 1348 | void sock_update_netprioidx(struct sock *sk) |
5bc1421e | 1349 | { |
5bc1421e NH |
1350 | if (in_interrupt()) |
1351 | return; | |
2b73bc65 | 1352 | |
6ffd4641 | 1353 | sk->sk_cgrp_prioidx = task_netprioidx(current); |
5bc1421e NH |
1354 | } |
1355 | EXPORT_SYMBOL_GPL(sock_update_netprioidx); | |
f8451725 HX |
1356 | #endif |
1357 | ||
1da177e4 LT |
1358 | /** |
1359 | * sk_alloc - All socket objects are allocated here | |
c4ea43c5 | 1360 | * @net: the applicable net namespace |
4dc3b16b PP |
1361 | * @family: protocol family |
1362 | * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc) | |
1363 | * @prot: struct proto associated with this new sock instance | |
1da177e4 | 1364 | */ |
1b8d7ae4 | 1365 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
6257ff21 | 1366 | struct proto *prot) |
1da177e4 | 1367 | { |
c308c1b2 | 1368 | struct sock *sk; |
1da177e4 | 1369 | |
154adbc8 | 1370 | sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family); |
1da177e4 | 1371 | if (sk) { |
154adbc8 PE |
1372 | sk->sk_family = family; |
1373 | /* | |
1374 | * See comment in struct sock definition to understand | |
1375 | * why we need sk_prot_creator -acme | |
1376 | */ | |
1377 | sk->sk_prot = sk->sk_prot_creator = prot; | |
1378 | sock_lock_init(sk); | |
3b1e0a65 | 1379 | sock_net_set(sk, get_net(net)); |
d66ee058 | 1380 | atomic_set(&sk->sk_wmem_alloc, 1); |
f8451725 | 1381 | |
211d2f97 | 1382 | sock_update_classid(sk); |
6ffd4641 | 1383 | sock_update_netprioidx(sk); |
1da177e4 | 1384 | } |
a79af59e | 1385 | |
2e4afe7b | 1386 | return sk; |
1da177e4 | 1387 | } |
2a91525c | 1388 | EXPORT_SYMBOL(sk_alloc); |
1da177e4 | 1389 | |
2b85a34e | 1390 | static void __sk_free(struct sock *sk) |
1da177e4 LT |
1391 | { |
1392 | struct sk_filter *filter; | |
1da177e4 LT |
1393 | |
1394 | if (sk->sk_destruct) | |
1395 | sk->sk_destruct(sk); | |
1396 | ||
a898def2 PM |
1397 | filter = rcu_dereference_check(sk->sk_filter, |
1398 | atomic_read(&sk->sk_wmem_alloc) == 0); | |
1da177e4 | 1399 | if (filter) { |
309dd5fc | 1400 | sk_filter_uncharge(sk, filter); |
a9b3cd7f | 1401 | RCU_INIT_POINTER(sk->sk_filter, NULL); |
1da177e4 LT |
1402 | } |
1403 | ||
08e29af3 | 1404 | sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP); |
1da177e4 LT |
1405 | |
1406 | if (atomic_read(&sk->sk_omem_alloc)) | |
e005d193 JP |
1407 | pr_debug("%s: optmem leakage (%d bytes) detected\n", |
1408 | __func__, atomic_read(&sk->sk_omem_alloc)); | |
1da177e4 | 1409 | |
109f6e39 EB |
1410 | if (sk->sk_peer_cred) |
1411 | put_cred(sk->sk_peer_cred); | |
1412 | put_pid(sk->sk_peer_pid); | |
3b1e0a65 | 1413 | put_net(sock_net(sk)); |
c308c1b2 | 1414 | sk_prot_free(sk->sk_prot_creator, sk); |
1da177e4 | 1415 | } |
2b85a34e ED |
1416 | |
1417 | void sk_free(struct sock *sk) | |
1418 | { | |
1419 | /* | |
25985edc | 1420 | * We subtract one from sk_wmem_alloc and can know if |
2b85a34e ED |
1421 | * some packets are still in some tx queue. |
1422 | * If not null, sock_wfree() will call __sk_free(sk) later | |
1423 | */ | |
1424 | if (atomic_dec_and_test(&sk->sk_wmem_alloc)) | |
1425 | __sk_free(sk); | |
1426 | } | |
2a91525c | 1427 | EXPORT_SYMBOL(sk_free); |
1da177e4 | 1428 | |
edf02087 | 1429 | /* |
25985edc LDM |
1430 | * Last sock_put should drop reference to sk->sk_net. It has already |
1431 | * been dropped in sk_change_net. Taking reference to stopping namespace | |
edf02087 | 1432 | * is not an option. |
25985edc | 1433 | * Take reference to a socket to remove it from hash _alive_ and after that |
edf02087 DL |
1434 | * destroy it in the context of init_net. |
1435 | */ | |
1436 | void sk_release_kernel(struct sock *sk) | |
1437 | { | |
1438 | if (sk == NULL || sk->sk_socket == NULL) | |
1439 | return; | |
1440 | ||
1441 | sock_hold(sk); | |
1442 | sock_release(sk->sk_socket); | |
65a18ec5 | 1443 | release_net(sock_net(sk)); |
3b1e0a65 | 1444 | sock_net_set(sk, get_net(&init_net)); |
edf02087 DL |
1445 | sock_put(sk); |
1446 | } | |
45af1754 | 1447 | EXPORT_SYMBOL(sk_release_kernel); |
edf02087 | 1448 | |
475f1b52 SR |
1449 | static void sk_update_clone(const struct sock *sk, struct sock *newsk) |
1450 | { | |
1451 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) | |
1452 | sock_update_memcg(newsk); | |
1453 | } | |
1454 | ||
e56c57d0 ED |
1455 | /** |
1456 | * sk_clone_lock - clone a socket, and lock its clone | |
1457 | * @sk: the socket to clone | |
1458 | * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc) | |
1459 | * | |
1460 | * Caller must unlock socket even in error path (bh_unlock_sock(newsk)) | |
1461 | */ | |
1462 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority) | |
87d11ceb | 1463 | { |
8fd1d178 | 1464 | struct sock *newsk; |
87d11ceb | 1465 | |
8fd1d178 | 1466 | newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family); |
87d11ceb ACM |
1467 | if (newsk != NULL) { |
1468 | struct sk_filter *filter; | |
1469 | ||
892c141e | 1470 | sock_copy(newsk, sk); |
87d11ceb ACM |
1471 | |
1472 | /* SANITY */ | |
3b1e0a65 | 1473 | get_net(sock_net(newsk)); |
87d11ceb ACM |
1474 | sk_node_init(&newsk->sk_node); |
1475 | sock_lock_init(newsk); | |
1476 | bh_lock_sock(newsk); | |
fa438ccf | 1477 | newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL; |
8eae939f | 1478 | newsk->sk_backlog.len = 0; |
87d11ceb ACM |
1479 | |
1480 | atomic_set(&newsk->sk_rmem_alloc, 0); | |
2b85a34e ED |
1481 | /* |
1482 | * sk_wmem_alloc set to one (see sk_free() and sock_wfree()) | |
1483 | */ | |
1484 | atomic_set(&newsk->sk_wmem_alloc, 1); | |
87d11ceb ACM |
1485 | atomic_set(&newsk->sk_omem_alloc, 0); |
1486 | skb_queue_head_init(&newsk->sk_receive_queue); | |
1487 | skb_queue_head_init(&newsk->sk_write_queue); | |
97fc2f08 CL |
1488 | #ifdef CONFIG_NET_DMA |
1489 | skb_queue_head_init(&newsk->sk_async_wait_queue); | |
1490 | #endif | |
87d11ceb | 1491 | |
b6c6712a | 1492 | spin_lock_init(&newsk->sk_dst_lock); |
87d11ceb | 1493 | rwlock_init(&newsk->sk_callback_lock); |
443aef0e PZ |
1494 | lockdep_set_class_and_name(&newsk->sk_callback_lock, |
1495 | af_callback_keys + newsk->sk_family, | |
1496 | af_family_clock_key_strings[newsk->sk_family]); | |
87d11ceb ACM |
1497 | |
1498 | newsk->sk_dst_cache = NULL; | |
1499 | newsk->sk_wmem_queued = 0; | |
1500 | newsk->sk_forward_alloc = 0; | |
1501 | newsk->sk_send_head = NULL; | |
87d11ceb ACM |
1502 | newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK; |
1503 | ||
1504 | sock_reset_flag(newsk, SOCK_DONE); | |
1505 | skb_queue_head_init(&newsk->sk_error_queue); | |
1506 | ||
0d7da9dd | 1507 | filter = rcu_dereference_protected(newsk->sk_filter, 1); |
87d11ceb ACM |
1508 | if (filter != NULL) |
1509 | sk_filter_charge(newsk, filter); | |
1510 | ||
1511 | if (unlikely(xfrm_sk_clone_policy(newsk))) { | |
1512 | /* It is still raw copy of parent, so invalidate | |
1513 | * destructor and make plain sk_free() */ | |
1514 | newsk->sk_destruct = NULL; | |
b0691c8e | 1515 | bh_unlock_sock(newsk); |
87d11ceb ACM |
1516 | sk_free(newsk); |
1517 | newsk = NULL; | |
1518 | goto out; | |
1519 | } | |
1520 | ||
1521 | newsk->sk_err = 0; | |
9385be2e | 1522 | newsk->sk_err_soft = 0; |
87d11ceb | 1523 | newsk->sk_priority = 0; |
4dc6dc71 ED |
1524 | /* |
1525 | * Before updating sk_refcnt, we must commit prior changes to memory | |
1526 | * (Documentation/RCU/rculist_nulls.txt for details) | |
1527 | */ | |
1528 | smp_wmb(); | |
87d11ceb ACM |
1529 | atomic_set(&newsk->sk_refcnt, 2); |
1530 | ||
1531 | /* | |
1532 | * Increment the counter in the same struct proto as the master | |
1533 | * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that | |
1534 | * is the same as sk->sk_prot->socks, as this field was copied | |
1535 | * with memcpy). | |
1536 | * | |
1537 | * This _changes_ the previous behaviour, where | |
1538 | * tcp_create_openreq_child always was incrementing the | |
1539 | * equivalent to tcp_prot->socks (inet_sock_nr), so this have | |
1540 | * to be taken into account in all callers. -acme | |
1541 | */ | |
1542 | sk_refcnt_debug_inc(newsk); | |
972692e0 | 1543 | sk_set_socket(newsk, NULL); |
43815482 | 1544 | newsk->sk_wq = NULL; |
87d11ceb | 1545 | |
f3f511e1 GC |
1546 | sk_update_clone(sk, newsk); |
1547 | ||
87d11ceb | 1548 | if (newsk->sk_prot->sockets_allocated) |
180d8cd9 | 1549 | sk_sockets_allocated_inc(newsk); |
704da560 | 1550 | |
08e29af3 | 1551 | if (newsk->sk_flags & SK_FLAGS_TIMESTAMP) |
704da560 | 1552 | net_enable_timestamp(); |
87d11ceb ACM |
1553 | } |
1554 | out: | |
1555 | return newsk; | |
1556 | } | |
e56c57d0 | 1557 | EXPORT_SYMBOL_GPL(sk_clone_lock); |
87d11ceb | 1558 | |
9958089a AK |
1559 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst) |
1560 | { | |
1561 | __sk_dst_set(sk, dst); | |
1562 | sk->sk_route_caps = dst->dev->features; | |
1563 | if (sk->sk_route_caps & NETIF_F_GSO) | |
4fcd6b99 | 1564 | sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE; |
a465419b | 1565 | sk->sk_route_caps &= ~sk->sk_route_nocaps; |
9958089a | 1566 | if (sk_can_gso(sk)) { |
82cc1a7a | 1567 | if (dst->header_len) { |
9958089a | 1568 | sk->sk_route_caps &= ~NETIF_F_GSO_MASK; |
82cc1a7a | 1569 | } else { |
9958089a | 1570 | sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM; |
82cc1a7a | 1571 | sk->sk_gso_max_size = dst->dev->gso_max_size; |
1485348d | 1572 | sk->sk_gso_max_segs = dst->dev->gso_max_segs; |
82cc1a7a | 1573 | } |
9958089a AK |
1574 | } |
1575 | } | |
1576 | EXPORT_SYMBOL_GPL(sk_setup_caps); | |
1577 | ||
1da177e4 LT |
1578 | /* |
1579 | * Simple resource managers for sockets. | |
1580 | */ | |
1581 | ||
1582 | ||
4ec93edb YH |
1583 | /* |
1584 | * Write buffer destructor automatically called from kfree_skb. | |
1da177e4 LT |
1585 | */ |
1586 | void sock_wfree(struct sk_buff *skb) | |
1587 | { | |
1588 | struct sock *sk = skb->sk; | |
d99927f4 | 1589 | unsigned int len = skb->truesize; |
1da177e4 | 1590 | |
d99927f4 ED |
1591 | if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) { |
1592 | /* | |
1593 | * Keep a reference on sk_wmem_alloc, this will be released | |
1594 | * after sk_write_space() call | |
1595 | */ | |
1596 | atomic_sub(len - 1, &sk->sk_wmem_alloc); | |
1da177e4 | 1597 | sk->sk_write_space(sk); |
d99927f4 ED |
1598 | len = 1; |
1599 | } | |
2b85a34e | 1600 | /* |
d99927f4 ED |
1601 | * if sk_wmem_alloc reaches 0, we must finish what sk_free() |
1602 | * could not do because of in-flight packets | |
2b85a34e | 1603 | */ |
d99927f4 | 1604 | if (atomic_sub_and_test(len, &sk->sk_wmem_alloc)) |
2b85a34e | 1605 | __sk_free(sk); |
1da177e4 | 1606 | } |
2a91525c | 1607 | EXPORT_SYMBOL(sock_wfree); |
1da177e4 | 1608 | |
4ec93edb YH |
1609 | /* |
1610 | * Read buffer destructor automatically called from kfree_skb. | |
1da177e4 LT |
1611 | */ |
1612 | void sock_rfree(struct sk_buff *skb) | |
1613 | { | |
1614 | struct sock *sk = skb->sk; | |
d361fd59 | 1615 | unsigned int len = skb->truesize; |
1da177e4 | 1616 | |
d361fd59 ED |
1617 | atomic_sub(len, &sk->sk_rmem_alloc); |
1618 | sk_mem_uncharge(sk, len); | |
1da177e4 | 1619 | } |
2a91525c | 1620 | EXPORT_SYMBOL(sock_rfree); |
1da177e4 | 1621 | |
41063e9d DM |
1622 | void sock_edemux(struct sk_buff *skb) |
1623 | { | |
e812347c ED |
1624 | struct sock *sk = skb->sk; |
1625 | ||
1c463e57 | 1626 | #ifdef CONFIG_INET |
e812347c ED |
1627 | if (sk->sk_state == TCP_TIME_WAIT) |
1628 | inet_twsk_put(inet_twsk(sk)); | |
1629 | else | |
1c463e57 | 1630 | #endif |
e812347c | 1631 | sock_put(sk); |
41063e9d DM |
1632 | } |
1633 | EXPORT_SYMBOL(sock_edemux); | |
1da177e4 | 1634 | |
976d0201 | 1635 | kuid_t sock_i_uid(struct sock *sk) |
1da177e4 | 1636 | { |
976d0201 | 1637 | kuid_t uid; |
6fa3eb70 S |
1638 | |
1639 | /*mtk_net: fix kernel bug*/ | |
1640 | if (!sk) { | |
1641 | pr_info("sk == NULL for sock_i_uid\n"); | |
1642 | return GLOBAL_ROOT_UID; | |
1643 | } | |
1644 | ||
f064af1e | 1645 | read_lock_bh(&sk->sk_callback_lock); |
976d0201 | 1646 | uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID; |
f064af1e | 1647 | read_unlock_bh(&sk->sk_callback_lock); |
1da177e4 LT |
1648 | return uid; |
1649 | } | |
2a91525c | 1650 | EXPORT_SYMBOL(sock_i_uid); |
1da177e4 LT |
1651 | |
1652 | unsigned long sock_i_ino(struct sock *sk) | |
1653 | { | |
1654 | unsigned long ino; | |
1655 | ||
f064af1e | 1656 | read_lock_bh(&sk->sk_callback_lock); |
1da177e4 | 1657 | ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0; |
f064af1e | 1658 | read_unlock_bh(&sk->sk_callback_lock); |
1da177e4 LT |
1659 | return ino; |
1660 | } | |
2a91525c | 1661 | EXPORT_SYMBOL(sock_i_ino); |
1da177e4 LT |
1662 | |
1663 | /* | |
1664 | * Allocate a skb from the socket's send buffer. | |
1665 | */ | |
86a76caf | 1666 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, |
dd0fc66f | 1667 | gfp_t priority) |
1da177e4 LT |
1668 | { |
1669 | if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) { | |
2a91525c | 1670 | struct sk_buff *skb = alloc_skb(size, priority); |
1da177e4 LT |
1671 | if (skb) { |
1672 | skb_set_owner_w(skb, sk); | |
1673 | return skb; | |
1674 | } | |
1675 | } | |
1676 | return NULL; | |
1677 | } | |
2a91525c | 1678 | EXPORT_SYMBOL(sock_wmalloc); |
1da177e4 LT |
1679 | |
1680 | /* | |
1681 | * Allocate a skb from the socket's receive buffer. | |
4ec93edb | 1682 | */ |
86a76caf | 1683 | struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force, |
dd0fc66f | 1684 | gfp_t priority) |
1da177e4 LT |
1685 | { |
1686 | if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) { | |
1687 | struct sk_buff *skb = alloc_skb(size, priority); | |
1688 | if (skb) { | |
1689 | skb_set_owner_r(skb, sk); | |
1690 | return skb; | |
1691 | } | |
1692 | } | |
1693 | return NULL; | |
1694 | } | |
1695 | ||
4ec93edb | 1696 | /* |
1da177e4 | 1697 | * Allocate a memory block from the socket's option memory buffer. |
4ec93edb | 1698 | */ |
dd0fc66f | 1699 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority) |
1da177e4 | 1700 | { |
95c96174 | 1701 | if ((unsigned int)size <= sysctl_optmem_max && |
1da177e4 LT |
1702 | atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) { |
1703 | void *mem; | |
1704 | /* First do the add, to avoid the race if kmalloc | |
4ec93edb | 1705 | * might sleep. |
1da177e4 LT |
1706 | */ |
1707 | atomic_add(size, &sk->sk_omem_alloc); | |
1708 | mem = kmalloc(size, priority); | |
1709 | if (mem) | |
1710 | return mem; | |
1711 | atomic_sub(size, &sk->sk_omem_alloc); | |
1712 | } | |
1713 | return NULL; | |
1714 | } | |
2a91525c | 1715 | EXPORT_SYMBOL(sock_kmalloc); |
1da177e4 LT |
1716 | |
1717 | /* | |
1718 | * Free an option memory block. | |
1719 | */ | |
1720 | void sock_kfree_s(struct sock *sk, void *mem, int size) | |
1721 | { | |
1722 | kfree(mem); | |
1723 | atomic_sub(size, &sk->sk_omem_alloc); | |
1724 | } | |
2a91525c | 1725 | EXPORT_SYMBOL(sock_kfree_s); |
1da177e4 LT |
1726 | |
1727 | /* It is almost wait_for_tcp_memory minus release_sock/lock_sock. | |
1728 | I think, these locks should be removed for datagram sockets. | |
1729 | */ | |
2a91525c | 1730 | static long sock_wait_for_wmem(struct sock *sk, long timeo) |
1da177e4 LT |
1731 | { |
1732 | DEFINE_WAIT(wait); | |
1733 | ||
1734 | clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); | |
1735 | for (;;) { | |
1736 | if (!timeo) | |
1737 | break; | |
1738 | if (signal_pending(current)) | |
1739 | break; | |
1740 | set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); | |
aa395145 | 1741 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 LT |
1742 | if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) |
1743 | break; | |
1744 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1745 | break; | |
1746 | if (sk->sk_err) | |
1747 | break; | |
1748 | timeo = schedule_timeout(timeo); | |
1749 | } | |
aa395145 | 1750 | finish_wait(sk_sleep(sk), &wait); |
1da177e4 LT |
1751 | return timeo; |
1752 | } | |
1753 | ||
1754 | ||
6fa3eb70 S |
1755 | //debug funcion |
1756 | ||
1757 | static int sock_dump_info(struct sock *sk) | |
1758 | { | |
1759 | //dump receiver queue 128 bytes | |
1760 | //struct sk_buff *skb; | |
1761 | //char skbmsg[128]; | |
1762 | //dump receiver queue 128 bytes end | |
1763 | ||
1764 | if(sk->sk_family == AF_UNIX) | |
1765 | { | |
1766 | struct unix_sock *u = unix_sk(sk); | |
1767 | struct sock *other = NULL; | |
1768 | if( (u->path.dentry !=NULL)&&(u->path.dentry->d_iname!=NULL)) | |
1769 | //if( (u->dentry !=NULL)&&(u->dentry->d_iname!=NULL)) | |
1770 | { | |
1771 | #ifdef CONFIG_MTK_NET_LOGGING | |
1772 | printk(KERN_INFO "[mtk_net][sock]sockdbg: socket-Name:%s \n",u->path.dentry->d_iname); | |
1773 | #endif | |
1774 | } | |
1775 | else | |
1776 | { | |
1777 | #ifdef CONFIG_MTK_NET_LOGGING | |
1778 | printk(KERN_INFO "[mtk_net][sock]sockdbg:socket Name (NULL)\n" ); | |
1779 | #endif | |
1780 | } | |
1781 | ||
1782 | if(sk->sk_socket && SOCK_INODE(sk->sk_socket)) | |
1783 | { | |
1784 | #ifdef CONFIG_MTK_NET_LOGGING | |
1785 | printk(KERN_INFO "[mtk_net][sock]sockdbg:socket Inode[%lu]\n" ,SOCK_INODE(sk->sk_socket)->i_ino); | |
1786 | #endif | |
1787 | } | |
1788 | ||
1789 | other = unix_sk(sk)->peer ; | |
1790 | if (!other) | |
1791 | { | |
1792 | #ifdef CONFIG_MTK_NET_LOGGING | |
1793 | printk(KERN_INFO "[mtk_net][sock]sockdbg:peer is (NULL) \n"); | |
1794 | #endif | |
1795 | } else{ | |
1796 | ||
1797 | if ((((struct unix_sock *)other)->path.dentry != NULL)&&(((struct unix_sock *)other)->path.dentry->d_iname != NULL)) | |
1798 | //if ((((struct unix_sock *)other)->dentry != NULL)&&(((struct unix_sock *)other)->dentry->d_iname != NULL)) | |
1799 | { | |
1800 | #ifdef CONFIG_MTK_NET_LOGGING | |
1801 | printk(KERN_INFO "[mtk_net][sock]sockdbg: Peer Name:%s \n",((struct unix_sock *)other)->path.dentry->d_iname); | |
1802 | #endif | |
1803 | } | |
1804 | else | |
1805 | { | |
1806 | #ifdef CONFIG_MTK_NET_LOGGING | |
1807 | printk(KERN_INFO "[mtk_net][sock]sockdbg: Peer Name (NULL) \n"); | |
1808 | #endif | |
1809 | } | |
1810 | ||
1811 | if(other->sk_socket && SOCK_INODE(other->sk_socket)) | |
1812 | { | |
1813 | #ifdef CONFIG_MTK_NET_LOGGING | |
1814 | printk(KERN_INFO "[mtk_net][sock]sockdbg: Peer Inode [%lu] \n", SOCK_INODE(other->sk_socket)->i_ino); | |
1815 | #endif | |
1816 | } | |
1817 | #ifdef CONFIG_MTK_NET_LOGGING | |
1818 | printk(KERN_INFO "[mtk_net][sock]sockdbg: Peer Recieve Queue len:%d \n",other->sk_receive_queue.qlen); | |
1819 | #endif | |
1820 | //dump receiver queue 128 bytes | |
1821 | /* if ((skb = skb_peek_tail(&other->sk_receive_queue)) == NULL) { | |
1822 | ||
1823 | printk(KERN_INFO "sockdbg: Peer Recieve Queue is null (warning) \n"); | |
1824 | }else{ | |
1825 | int i =0 ,len=0; | |
1826 | if((skb->len !=0) && (skb->data != NULL)){ | |
1827 | ||
1828 | if(skb->len >= 127){ | |
1829 | len = 127 ; | |
1830 | }else | |
1831 | { | |
1832 | len = skb->len ; | |
1833 | } | |
1834 | for (i=0;i<len;i++) | |
1835 | sprintf(skbmsg+i, "%x", skb->data[i]); | |
1836 | ||
1837 | skbmsg[len]= '\0' ; | |
1838 | ||
1839 | printk(KERN_INFO "sockdbg: Peer Recieve Queue dump(%d bytes):%s\n", len, skbmsg); | |
1840 | ||
1841 | ||
1842 | }else{ | |
1843 | printk(KERN_INFO "sockdbg: Peer Recieve skb error \n"); | |
1844 | }*/ | |
1845 | //dump receiver queue 128 bytes end | |
1846 | ||
1847 | //} | |
1848 | //dump receiver queue 128 bytes end | |
1849 | ||
1850 | } | |
1851 | } | |
1852 | ||
1853 | return 0 ; | |
1854 | ||
1855 | ||
1856 | } | |
1857 | ||
1858 | ||
1859 | ||
1da177e4 LT |
1860 | /* |
1861 | * Generic send/receive buffer handlers | |
1862 | */ | |
1863 | ||
4cc7f68d HX |
1864 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, |
1865 | unsigned long data_len, int noblock, | |
1866 | int *errcode) | |
1da177e4 LT |
1867 | { |
1868 | struct sk_buff *skb; | |
7d877f3b | 1869 | gfp_t gfp_mask; |
1da177e4 LT |
1870 | long timeo; |
1871 | int err; | |
cc9b17ad JW |
1872 | int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT; |
1873 | ||
1874 | err = -EMSGSIZE; | |
1875 | if (npages > MAX_SKB_FRAGS) | |
1876 | goto failure; | |
1da177e4 LT |
1877 | |
1878 | gfp_mask = sk->sk_allocation; | |
1879 | if (gfp_mask & __GFP_WAIT) | |
1880 | gfp_mask |= __GFP_REPEAT; | |
1881 | ||
1882 | timeo = sock_sndtimeo(sk, noblock); | |
1883 | while (1) { | |
1884 | err = sock_error(sk); | |
1885 | if (err != 0) | |
1886 | goto failure; | |
1887 | ||
1888 | err = -EPIPE; | |
1889 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1890 | goto failure; | |
1891 | ||
1892 | if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) { | |
db38c179 | 1893 | skb = alloc_skb(header_len, gfp_mask); |
1da177e4 | 1894 | if (skb) { |
1da177e4 LT |
1895 | int i; |
1896 | ||
1897 | /* No pages, we're done... */ | |
1898 | if (!data_len) | |
1899 | break; | |
1900 | ||
1da177e4 LT |
1901 | skb->truesize += data_len; |
1902 | skb_shinfo(skb)->nr_frags = npages; | |
1903 | for (i = 0; i < npages; i++) { | |
1904 | struct page *page; | |
1da177e4 LT |
1905 | |
1906 | page = alloc_pages(sk->sk_allocation, 0); | |
1907 | if (!page) { | |
1908 | err = -ENOBUFS; | |
1909 | skb_shinfo(skb)->nr_frags = i; | |
1910 | kfree_skb(skb); | |
1911 | goto failure; | |
1912 | } | |
1913 | ||
ea2ab693 IC |
1914 | __skb_fill_page_desc(skb, i, |
1915 | page, 0, | |
1916 | (data_len >= PAGE_SIZE ? | |
1917 | PAGE_SIZE : | |
1918 | data_len)); | |
1da177e4 LT |
1919 | data_len -= PAGE_SIZE; |
1920 | } | |
1921 | ||
1922 | /* Full success... */ | |
1923 | break; | |
1924 | } | |
1925 | err = -ENOBUFS; | |
1926 | goto failure; | |
1927 | } | |
1928 | set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); | |
1929 | set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); | |
1930 | err = -EAGAIN; | |
1931 | if (!timeo) | |
1932 | goto failure; | |
1933 | if (signal_pending(current)) | |
1934 | goto interrupted; | |
6fa3eb70 S |
1935 | |
1936 | sock_dump_info(sk); | |
1937 | #ifdef CONFIG_MTK_NET_LOGGING | |
1938 | printk(KERN_INFO "[mtk_net][sock]sockdbg: wait_for_wmem, timeo =%ld, wmem =%d, snd buf =%d \n", | |
1939 | timeo, atomic_read(&sk->sk_wmem_alloc), sk->sk_sndbuf); | |
1940 | #endif | |
1da177e4 | 1941 | timeo = sock_wait_for_wmem(sk, timeo); |
6fa3eb70 S |
1942 | #ifdef CONFIG_MTK_NET_LOGGING |
1943 | printk(KERN_INFO "[mtk_net][sock]sockdbg: wait_for_wmem done, header_len=0x%lx, data_len=0x%lx,timeo =%ld \n", | |
1944 | header_len, data_len ,timeo); | |
1945 | #endif | |
1da177e4 LT |
1946 | } |
1947 | ||
1948 | skb_set_owner_w(skb, sk); | |
1949 | return skb; | |
1950 | ||
1951 | interrupted: | |
1952 | err = sock_intr_errno(timeo); | |
1953 | failure: | |
1954 | *errcode = err; | |
1955 | return NULL; | |
1956 | } | |
4cc7f68d | 1957 | EXPORT_SYMBOL(sock_alloc_send_pskb); |
1da177e4 | 1958 | |
4ec93edb | 1959 | struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, |
1da177e4 LT |
1960 | int noblock, int *errcode) |
1961 | { | |
1962 | return sock_alloc_send_pskb(sk, size, 0, noblock, errcode); | |
1963 | } | |
2a91525c | 1964 | EXPORT_SYMBOL(sock_alloc_send_skb); |
1da177e4 | 1965 | |
5640f768 ED |
1966 | /* On 32bit arches, an skb frag is limited to 2^15 */ |
1967 | #define SKB_FRAG_PAGE_ORDER get_order(32768) | |
1968 | ||
1969 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag) | |
1970 | { | |
1971 | int order; | |
1972 | ||
1973 | if (pfrag->page) { | |
1974 | if (atomic_read(&pfrag->page->_count) == 1) { | |
1975 | pfrag->offset = 0; | |
1976 | return true; | |
1977 | } | |
1978 | if (pfrag->offset < pfrag->size) | |
1979 | return true; | |
1980 | put_page(pfrag->page); | |
1981 | } | |
1982 | ||
1983 | /* We restrict high order allocations to users that can afford to wait */ | |
1984 | order = (sk->sk_allocation & __GFP_WAIT) ? SKB_FRAG_PAGE_ORDER : 0; | |
1985 | ||
1986 | do { | |
1987 | gfp_t gfp = sk->sk_allocation; | |
1988 | ||
1989 | if (order) | |
a9e3d789 | 1990 | gfp |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY; |
5640f768 ED |
1991 | pfrag->page = alloc_pages(gfp, order); |
1992 | if (likely(pfrag->page)) { | |
1993 | pfrag->offset = 0; | |
1994 | pfrag->size = PAGE_SIZE << order; | |
1995 | return true; | |
1996 | } | |
1997 | } while (--order >= 0); | |
1998 | ||
1999 | sk_enter_memory_pressure(sk); | |
2000 | sk_stream_moderate_sndbuf(sk); | |
2001 | return false; | |
2002 | } | |
2003 | EXPORT_SYMBOL(sk_page_frag_refill); | |
2004 | ||
1da177e4 | 2005 | static void __lock_sock(struct sock *sk) |
f39234d6 NK |
2006 | __releases(&sk->sk_lock.slock) |
2007 | __acquires(&sk->sk_lock.slock) | |
1da177e4 LT |
2008 | { |
2009 | DEFINE_WAIT(wait); | |
2010 | ||
e71a4783 | 2011 | for (;;) { |
1da177e4 LT |
2012 | prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait, |
2013 | TASK_UNINTERRUPTIBLE); | |
2014 | spin_unlock_bh(&sk->sk_lock.slock); | |
2015 | schedule(); | |
2016 | spin_lock_bh(&sk->sk_lock.slock); | |
e71a4783 | 2017 | if (!sock_owned_by_user(sk)) |
1da177e4 LT |
2018 | break; |
2019 | } | |
2020 | finish_wait(&sk->sk_lock.wq, &wait); | |
2021 | } | |
2022 | ||
2023 | static void __release_sock(struct sock *sk) | |
f39234d6 NK |
2024 | __releases(&sk->sk_lock.slock) |
2025 | __acquires(&sk->sk_lock.slock) | |
1da177e4 LT |
2026 | { |
2027 | struct sk_buff *skb = sk->sk_backlog.head; | |
2028 | ||
2029 | do { | |
2030 | sk->sk_backlog.head = sk->sk_backlog.tail = NULL; | |
2031 | bh_unlock_sock(sk); | |
2032 | ||
2033 | do { | |
2034 | struct sk_buff *next = skb->next; | |
2035 | ||
e4cbb02a | 2036 | prefetch(next); |
7fee226a | 2037 | WARN_ON_ONCE(skb_dst_is_noref(skb)); |
1da177e4 | 2038 | skb->next = NULL; |
c57943a1 | 2039 | sk_backlog_rcv(sk, skb); |
1da177e4 LT |
2040 | |
2041 | /* | |
2042 | * We are in process context here with softirqs | |
2043 | * disabled, use cond_resched_softirq() to preempt. | |
2044 | * This is safe to do because we've taken the backlog | |
2045 | * queue private: | |
2046 | */ | |
2047 | cond_resched_softirq(); | |
2048 | ||
2049 | skb = next; | |
2050 | } while (skb != NULL); | |
2051 | ||
2052 | bh_lock_sock(sk); | |
e71a4783 | 2053 | } while ((skb = sk->sk_backlog.head) != NULL); |
8eae939f ZY |
2054 | |
2055 | /* | |
2056 | * Doing the zeroing here guarantee we can not loop forever | |
2057 | * while a wild producer attempts to flood us. | |
2058 | */ | |
2059 | sk->sk_backlog.len = 0; | |
1da177e4 LT |
2060 | } |
2061 | ||
2062 | /** | |
2063 | * sk_wait_data - wait for data to arrive at sk_receive_queue | |
4dc3b16b PP |
2064 | * @sk: sock to wait on |
2065 | * @timeo: for how long | |
1da177e4 LT |
2066 | * |
2067 | * Now socket state including sk->sk_err is changed only under lock, | |
2068 | * hence we may omit checks after joining wait queue. | |
2069 | * We check receive queue before schedule() only as optimization; | |
2070 | * it is very likely that release_sock() added new data. | |
2071 | */ | |
2072 | int sk_wait_data(struct sock *sk, long *timeo) | |
2073 | { | |
2074 | int rc; | |
2075 | DEFINE_WAIT(wait); | |
2076 | ||
aa395145 | 2077 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 LT |
2078 | set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); |
2079 | rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue)); | |
2080 | clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); | |
aa395145 | 2081 | finish_wait(sk_sleep(sk), &wait); |
1da177e4 LT |
2082 | return rc; |
2083 | } | |
1da177e4 LT |
2084 | EXPORT_SYMBOL(sk_wait_data); |
2085 | ||
3ab224be HA |
2086 | /** |
2087 | * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated | |
2088 | * @sk: socket | |
2089 | * @size: memory size to allocate | |
2090 | * @kind: allocation type | |
2091 | * | |
2092 | * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means | |
2093 | * rmem allocation. This function assumes that protocols which have | |
2094 | * memory_pressure use sk_wmem_queued as write buffer accounting. | |
2095 | */ | |
2096 | int __sk_mem_schedule(struct sock *sk, int size, int kind) | |
2097 | { | |
2098 | struct proto *prot = sk->sk_prot; | |
2099 | int amt = sk_mem_pages(size); | |
8d987e5c | 2100 | long allocated; |
e1aab161 | 2101 | int parent_status = UNDER_LIMIT; |
3ab224be HA |
2102 | |
2103 | sk->sk_forward_alloc += amt * SK_MEM_QUANTUM; | |
180d8cd9 | 2104 | |
e1aab161 | 2105 | allocated = sk_memory_allocated_add(sk, amt, &parent_status); |
3ab224be HA |
2106 | |
2107 | /* Under limit. */ | |
e1aab161 GC |
2108 | if (parent_status == UNDER_LIMIT && |
2109 | allocated <= sk_prot_mem_limits(sk, 0)) { | |
180d8cd9 | 2110 | sk_leave_memory_pressure(sk); |
3ab224be HA |
2111 | return 1; |
2112 | } | |
2113 | ||
e1aab161 GC |
2114 | /* Under pressure. (we or our parents) */ |
2115 | if ((parent_status > SOFT_LIMIT) || | |
2116 | allocated > sk_prot_mem_limits(sk, 1)) | |
180d8cd9 | 2117 | sk_enter_memory_pressure(sk); |
3ab224be | 2118 | |
e1aab161 GC |
2119 | /* Over hard limit (we or our parents) */ |
2120 | if ((parent_status == OVER_LIMIT) || | |
2121 | (allocated > sk_prot_mem_limits(sk, 2))) | |
3ab224be HA |
2122 | goto suppress_allocation; |
2123 | ||
2124 | /* guarantee minimum buffer size under pressure */ | |
2125 | if (kind == SK_MEM_RECV) { | |
2126 | if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0]) | |
2127 | return 1; | |
180d8cd9 | 2128 | |
3ab224be HA |
2129 | } else { /* SK_MEM_SEND */ |
2130 | if (sk->sk_type == SOCK_STREAM) { | |
2131 | if (sk->sk_wmem_queued < prot->sysctl_wmem[0]) | |
2132 | return 1; | |
2133 | } else if (atomic_read(&sk->sk_wmem_alloc) < | |
2134 | prot->sysctl_wmem[0]) | |
2135 | return 1; | |
2136 | } | |
2137 | ||
180d8cd9 | 2138 | if (sk_has_memory_pressure(sk)) { |
1748376b ED |
2139 | int alloc; |
2140 | ||
180d8cd9 | 2141 | if (!sk_under_memory_pressure(sk)) |
1748376b | 2142 | return 1; |
180d8cd9 GC |
2143 | alloc = sk_sockets_allocated_read_positive(sk); |
2144 | if (sk_prot_mem_limits(sk, 2) > alloc * | |
3ab224be HA |
2145 | sk_mem_pages(sk->sk_wmem_queued + |
2146 | atomic_read(&sk->sk_rmem_alloc) + | |
2147 | sk->sk_forward_alloc)) | |
2148 | return 1; | |
2149 | } | |
2150 | ||
2151 | suppress_allocation: | |
2152 | ||
2153 | if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) { | |
2154 | sk_stream_moderate_sndbuf(sk); | |
2155 | ||
2156 | /* Fail only if socket is _under_ its sndbuf. | |
2157 | * In this case we cannot block, so that we have to fail. | |
2158 | */ | |
2159 | if (sk->sk_wmem_queued + size >= sk->sk_sndbuf) | |
2160 | return 1; | |
2161 | } | |
2162 | ||
3847ce32 SM |
2163 | trace_sock_exceed_buf_limit(sk, prot, allocated); |
2164 | ||
3ab224be HA |
2165 | /* Alas. Undo changes. */ |
2166 | sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM; | |
180d8cd9 | 2167 | |
0e90b31f | 2168 | sk_memory_allocated_sub(sk, amt); |
180d8cd9 | 2169 | |
3ab224be HA |
2170 | return 0; |
2171 | } | |
3ab224be HA |
2172 | EXPORT_SYMBOL(__sk_mem_schedule); |
2173 | ||
2174 | /** | |
2175 | * __sk_reclaim - reclaim memory_allocated | |
2176 | * @sk: socket | |
a5b79829 | 2177 | * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple) |
3ab224be | 2178 | */ |
a5b79829 | 2179 | void __sk_mem_reclaim(struct sock *sk, int amount) |
3ab224be | 2180 | { |
a5b79829 ED |
2181 | amount >>= SK_MEM_QUANTUM_SHIFT; |
2182 | sk_memory_allocated_sub(sk, amount); | |
2183 | sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT; | |
3ab224be | 2184 | |
180d8cd9 GC |
2185 | if (sk_under_memory_pressure(sk) && |
2186 | (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0))) | |
2187 | sk_leave_memory_pressure(sk); | |
3ab224be | 2188 | } |
3ab224be HA |
2189 | EXPORT_SYMBOL(__sk_mem_reclaim); |
2190 | ||
2191 | ||
1da177e4 LT |
2192 | /* |
2193 | * Set of default routines for initialising struct proto_ops when | |
2194 | * the protocol does not support a particular function. In certain | |
2195 | * cases where it makes no sense for a protocol to have a "do nothing" | |
2196 | * function, some default processing is provided. | |
2197 | */ | |
2198 | ||
2199 | int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len) | |
2200 | { | |
2201 | return -EOPNOTSUPP; | |
2202 | } | |
2a91525c | 2203 | EXPORT_SYMBOL(sock_no_bind); |
1da177e4 | 2204 | |
4ec93edb | 2205 | int sock_no_connect(struct socket *sock, struct sockaddr *saddr, |
1da177e4 LT |
2206 | int len, int flags) |
2207 | { | |
2208 | return -EOPNOTSUPP; | |
2209 | } | |
2a91525c | 2210 | EXPORT_SYMBOL(sock_no_connect); |
1da177e4 LT |
2211 | |
2212 | int sock_no_socketpair(struct socket *sock1, struct socket *sock2) | |
2213 | { | |
2214 | return -EOPNOTSUPP; | |
2215 | } | |
2a91525c | 2216 | EXPORT_SYMBOL(sock_no_socketpair); |
1da177e4 LT |
2217 | |
2218 | int sock_no_accept(struct socket *sock, struct socket *newsock, int flags) | |
2219 | { | |
2220 | return -EOPNOTSUPP; | |
2221 | } | |
2a91525c | 2222 | EXPORT_SYMBOL(sock_no_accept); |
1da177e4 | 2223 | |
4ec93edb | 2224 | int sock_no_getname(struct socket *sock, struct sockaddr *saddr, |
1da177e4 LT |
2225 | int *len, int peer) |
2226 | { | |
2227 | return -EOPNOTSUPP; | |
2228 | } | |
2a91525c | 2229 | EXPORT_SYMBOL(sock_no_getname); |
1da177e4 | 2230 | |
2a91525c | 2231 | unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt) |
1da177e4 LT |
2232 | { |
2233 | return 0; | |
2234 | } | |
2a91525c | 2235 | EXPORT_SYMBOL(sock_no_poll); |
1da177e4 LT |
2236 | |
2237 | int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
2238 | { | |
2239 | return -EOPNOTSUPP; | |
2240 | } | |
2a91525c | 2241 | EXPORT_SYMBOL(sock_no_ioctl); |
1da177e4 LT |
2242 | |
2243 | int sock_no_listen(struct socket *sock, int backlog) | |
2244 | { | |
2245 | return -EOPNOTSUPP; | |
2246 | } | |
2a91525c | 2247 | EXPORT_SYMBOL(sock_no_listen); |
1da177e4 LT |
2248 | |
2249 | int sock_no_shutdown(struct socket *sock, int how) | |
2250 | { | |
2251 | return -EOPNOTSUPP; | |
2252 | } | |
2a91525c | 2253 | EXPORT_SYMBOL(sock_no_shutdown); |
1da177e4 LT |
2254 | |
2255 | int sock_no_setsockopt(struct socket *sock, int level, int optname, | |
b7058842 | 2256 | char __user *optval, unsigned int optlen) |
1da177e4 LT |
2257 | { |
2258 | return -EOPNOTSUPP; | |
2259 | } | |
2a91525c | 2260 | EXPORT_SYMBOL(sock_no_setsockopt); |
1da177e4 LT |
2261 | |
2262 | int sock_no_getsockopt(struct socket *sock, int level, int optname, | |
2263 | char __user *optval, int __user *optlen) | |
2264 | { | |
2265 | return -EOPNOTSUPP; | |
2266 | } | |
2a91525c | 2267 | EXPORT_SYMBOL(sock_no_getsockopt); |
1da177e4 LT |
2268 | |
2269 | int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m, | |
2270 | size_t len) | |
2271 | { | |
2272 | return -EOPNOTSUPP; | |
2273 | } | |
2a91525c | 2274 | EXPORT_SYMBOL(sock_no_sendmsg); |
1da177e4 LT |
2275 | |
2276 | int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m, | |
2277 | size_t len, int flags) | |
2278 | { | |
2279 | return -EOPNOTSUPP; | |
2280 | } | |
2a91525c | 2281 | EXPORT_SYMBOL(sock_no_recvmsg); |
1da177e4 LT |
2282 | |
2283 | int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma) | |
2284 | { | |
2285 | /* Mirror missing mmap method error code */ | |
2286 | return -ENODEV; | |
2287 | } | |
2a91525c | 2288 | EXPORT_SYMBOL(sock_no_mmap); |
1da177e4 LT |
2289 | |
2290 | ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags) | |
2291 | { | |
2292 | ssize_t res; | |
2293 | struct msghdr msg = {.msg_flags = flags}; | |
2294 | struct kvec iov; | |
2295 | char *kaddr = kmap(page); | |
2296 | iov.iov_base = kaddr + offset; | |
2297 | iov.iov_len = size; | |
2298 | res = kernel_sendmsg(sock, &msg, &iov, 1, size); | |
2299 | kunmap(page); | |
2300 | return res; | |
2301 | } | |
2a91525c | 2302 | EXPORT_SYMBOL(sock_no_sendpage); |
1da177e4 LT |
2303 | |
2304 | /* | |
2305 | * Default Socket Callbacks | |
2306 | */ | |
2307 | ||
2308 | static void sock_def_wakeup(struct sock *sk) | |
2309 | { | |
43815482 ED |
2310 | struct socket_wq *wq; |
2311 | ||
2312 | rcu_read_lock(); | |
2313 | wq = rcu_dereference(sk->sk_wq); | |
2314 | if (wq_has_sleeper(wq)) | |
2315 | wake_up_interruptible_all(&wq->wait); | |
2316 | rcu_read_unlock(); | |
1da177e4 LT |
2317 | } |
2318 | ||
2319 | static void sock_def_error_report(struct sock *sk) | |
2320 | { | |
43815482 ED |
2321 | struct socket_wq *wq; |
2322 | ||
2323 | rcu_read_lock(); | |
2324 | wq = rcu_dereference(sk->sk_wq); | |
2325 | if (wq_has_sleeper(wq)) | |
2326 | wake_up_interruptible_poll(&wq->wait, POLLERR); | |
8d8ad9d7 | 2327 | sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR); |
43815482 | 2328 | rcu_read_unlock(); |
1da177e4 LT |
2329 | } |
2330 | ||
2331 | static void sock_def_readable(struct sock *sk, int len) | |
2332 | { | |
43815482 ED |
2333 | struct socket_wq *wq; |
2334 | ||
2335 | rcu_read_lock(); | |
2336 | wq = rcu_dereference(sk->sk_wq); | |
2337 | if (wq_has_sleeper(wq)) | |
2c6607c6 | 2338 | wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI | |
37e5540b | 2339 | POLLRDNORM | POLLRDBAND); |
8d8ad9d7 | 2340 | sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN); |
43815482 | 2341 | rcu_read_unlock(); |
1da177e4 LT |
2342 | } |
2343 | ||
2344 | static void sock_def_write_space(struct sock *sk) | |
2345 | { | |
43815482 ED |
2346 | struct socket_wq *wq; |
2347 | ||
2348 | rcu_read_lock(); | |
1da177e4 LT |
2349 | |
2350 | /* Do not wake up a writer until he can make "significant" | |
2351 | * progress. --DaveM | |
2352 | */ | |
e71a4783 | 2353 | if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) { |
43815482 ED |
2354 | wq = rcu_dereference(sk->sk_wq); |
2355 | if (wq_has_sleeper(wq)) | |
2356 | wake_up_interruptible_sync_poll(&wq->wait, POLLOUT | | |
37e5540b | 2357 | POLLWRNORM | POLLWRBAND); |
1da177e4 LT |
2358 | |
2359 | /* Should agree with poll, otherwise some programs break */ | |
2360 | if (sock_writeable(sk)) | |
8d8ad9d7 | 2361 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 LT |
2362 | } |
2363 | ||
43815482 | 2364 | rcu_read_unlock(); |
1da177e4 LT |
2365 | } |
2366 | ||
2367 | static void sock_def_destruct(struct sock *sk) | |
2368 | { | |
a51482bd | 2369 | kfree(sk->sk_protinfo); |
1da177e4 LT |
2370 | } |
2371 | ||
2372 | void sk_send_sigurg(struct sock *sk) | |
2373 | { | |
2374 | if (sk->sk_socket && sk->sk_socket->file) | |
2375 | if (send_sigurg(&sk->sk_socket->file->f_owner)) | |
8d8ad9d7 | 2376 | sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI); |
1da177e4 | 2377 | } |
2a91525c | 2378 | EXPORT_SYMBOL(sk_send_sigurg); |
1da177e4 LT |
2379 | |
2380 | void sk_reset_timer(struct sock *sk, struct timer_list* timer, | |
2381 | unsigned long expires) | |
2382 | { | |
2383 | if (!mod_timer(timer, expires)) | |
2384 | sock_hold(sk); | |
2385 | } | |
1da177e4 LT |
2386 | EXPORT_SYMBOL(sk_reset_timer); |
2387 | ||
2388 | void sk_stop_timer(struct sock *sk, struct timer_list* timer) | |
2389 | { | |
25cc4ae9 | 2390 | if (del_timer(timer)) |
1da177e4 LT |
2391 | __sock_put(sk); |
2392 | } | |
1da177e4 LT |
2393 | EXPORT_SYMBOL(sk_stop_timer); |
2394 | ||
2395 | void sock_init_data(struct socket *sock, struct sock *sk) | |
2396 | { | |
2397 | skb_queue_head_init(&sk->sk_receive_queue); | |
2398 | skb_queue_head_init(&sk->sk_write_queue); | |
2399 | skb_queue_head_init(&sk->sk_error_queue); | |
97fc2f08 CL |
2400 | #ifdef CONFIG_NET_DMA |
2401 | skb_queue_head_init(&sk->sk_async_wait_queue); | |
2402 | #endif | |
1da177e4 LT |
2403 | |
2404 | sk->sk_send_head = NULL; | |
2405 | ||
2406 | init_timer(&sk->sk_timer); | |
4ec93edb | 2407 | |
1da177e4 LT |
2408 | sk->sk_allocation = GFP_KERNEL; |
2409 | sk->sk_rcvbuf = sysctl_rmem_default; | |
2410 | sk->sk_sndbuf = sysctl_wmem_default; | |
2411 | sk->sk_state = TCP_CLOSE; | |
972692e0 | 2412 | sk_set_socket(sk, sock); |
1da177e4 LT |
2413 | |
2414 | sock_set_flag(sk, SOCK_ZAPPED); | |
2415 | ||
e71a4783 | 2416 | if (sock) { |
1da177e4 | 2417 | sk->sk_type = sock->type; |
43815482 | 2418 | sk->sk_wq = sock->wq; |
1da177e4 LT |
2419 | sock->sk = sk; |
2420 | } else | |
43815482 | 2421 | sk->sk_wq = NULL; |
1da177e4 | 2422 | |
b6c6712a | 2423 | spin_lock_init(&sk->sk_dst_lock); |
1da177e4 | 2424 | rwlock_init(&sk->sk_callback_lock); |
443aef0e PZ |
2425 | lockdep_set_class_and_name(&sk->sk_callback_lock, |
2426 | af_callback_keys + sk->sk_family, | |
2427 | af_family_clock_key_strings[sk->sk_family]); | |
1da177e4 LT |
2428 | |
2429 | sk->sk_state_change = sock_def_wakeup; | |
2430 | sk->sk_data_ready = sock_def_readable; | |
2431 | sk->sk_write_space = sock_def_write_space; | |
2432 | sk->sk_error_report = sock_def_error_report; | |
2433 | sk->sk_destruct = sock_def_destruct; | |
2434 | ||
5640f768 ED |
2435 | sk->sk_frag.page = NULL; |
2436 | sk->sk_frag.offset = 0; | |
ef64a54f | 2437 | sk->sk_peek_off = -1; |
1da177e4 | 2438 | |
109f6e39 EB |
2439 | sk->sk_peer_pid = NULL; |
2440 | sk->sk_peer_cred = NULL; | |
1da177e4 LT |
2441 | sk->sk_write_pending = 0; |
2442 | sk->sk_rcvlowat = 1; | |
2443 | sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; | |
2444 | sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT; | |
2445 | ||
f37f0afb | 2446 | sk->sk_stamp = ktime_set(-1L, 0); |
1da177e4 | 2447 | |
5e25ba50 | 2448 | sk->sk_pacing_rate = ~0U; |
4dc6dc71 ED |
2449 | /* |
2450 | * Before updating sk_refcnt, we must commit prior changes to memory | |
2451 | * (Documentation/RCU/rculist_nulls.txt for details) | |
2452 | */ | |
2453 | smp_wmb(); | |
1da177e4 | 2454 | atomic_set(&sk->sk_refcnt, 1); |
33c732c3 | 2455 | atomic_set(&sk->sk_drops, 0); |
1da177e4 | 2456 | } |
2a91525c | 2457 | EXPORT_SYMBOL(sock_init_data); |
1da177e4 | 2458 | |
b5606c2d | 2459 | void lock_sock_nested(struct sock *sk, int subclass) |
1da177e4 LT |
2460 | { |
2461 | might_sleep(); | |
a5b5bb9a | 2462 | spin_lock_bh(&sk->sk_lock.slock); |
d2e9117c | 2463 | if (sk->sk_lock.owned) |
1da177e4 | 2464 | __lock_sock(sk); |
d2e9117c | 2465 | sk->sk_lock.owned = 1; |
a5b5bb9a IM |
2466 | spin_unlock(&sk->sk_lock.slock); |
2467 | /* | |
2468 | * The sk_lock has mutex_lock() semantics here: | |
2469 | */ | |
fcc70d5f | 2470 | mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_); |
a5b5bb9a | 2471 | local_bh_enable(); |
1da177e4 | 2472 | } |
fcc70d5f | 2473 | EXPORT_SYMBOL(lock_sock_nested); |
1da177e4 | 2474 | |
b5606c2d | 2475 | void release_sock(struct sock *sk) |
1da177e4 | 2476 | { |
a5b5bb9a IM |
2477 | /* |
2478 | * The sk_lock has mutex_unlock() semantics: | |
2479 | */ | |
2480 | mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_); | |
2481 | ||
2482 | spin_lock_bh(&sk->sk_lock.slock); | |
1da177e4 LT |
2483 | if (sk->sk_backlog.tail) |
2484 | __release_sock(sk); | |
46d3ceab | 2485 | |
cbbb5a25 ED |
2486 | /* Warning : release_cb() might need to release sk ownership, |
2487 | * ie call sock_release_ownership(sk) before us. | |
2488 | */ | |
46d3ceab ED |
2489 | if (sk->sk_prot->release_cb) |
2490 | sk->sk_prot->release_cb(sk); | |
2491 | ||
cbbb5a25 | 2492 | sock_release_ownership(sk); |
a5b5bb9a IM |
2493 | if (waitqueue_active(&sk->sk_lock.wq)) |
2494 | wake_up(&sk->sk_lock.wq); | |
2495 | spin_unlock_bh(&sk->sk_lock.slock); | |
1da177e4 LT |
2496 | } |
2497 | EXPORT_SYMBOL(release_sock); | |
2498 | ||
8a74ad60 ED |
2499 | /** |
2500 | * lock_sock_fast - fast version of lock_sock | |
2501 | * @sk: socket | |
2502 | * | |
2503 | * This version should be used for very small section, where process wont block | |
2504 | * return false if fast path is taken | |
2505 | * sk_lock.slock locked, owned = 0, BH disabled | |
2506 | * return true if slow path is taken | |
2507 | * sk_lock.slock unlocked, owned = 1, BH enabled | |
2508 | */ | |
2509 | bool lock_sock_fast(struct sock *sk) | |
2510 | { | |
2511 | might_sleep(); | |
2512 | spin_lock_bh(&sk->sk_lock.slock); | |
2513 | ||
2514 | if (!sk->sk_lock.owned) | |
2515 | /* | |
2516 | * Note : We must disable BH | |
2517 | */ | |
2518 | return false; | |
2519 | ||
2520 | __lock_sock(sk); | |
2521 | sk->sk_lock.owned = 1; | |
2522 | spin_unlock(&sk->sk_lock.slock); | |
2523 | /* | |
2524 | * The sk_lock has mutex_lock() semantics here: | |
2525 | */ | |
2526 | mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_); | |
2527 | local_bh_enable(); | |
2528 | return true; | |
2529 | } | |
2530 | EXPORT_SYMBOL(lock_sock_fast); | |
2531 | ||
1da177e4 | 2532 | int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp) |
4ec93edb | 2533 | { |
b7aa0bf7 | 2534 | struct timeval tv; |
1da177e4 | 2535 | if (!sock_flag(sk, SOCK_TIMESTAMP)) |
20d49473 | 2536 | sock_enable_timestamp(sk, SOCK_TIMESTAMP); |
b7aa0bf7 ED |
2537 | tv = ktime_to_timeval(sk->sk_stamp); |
2538 | if (tv.tv_sec == -1) | |
1da177e4 | 2539 | return -ENOENT; |
b7aa0bf7 ED |
2540 | if (tv.tv_sec == 0) { |
2541 | sk->sk_stamp = ktime_get_real(); | |
2542 | tv = ktime_to_timeval(sk->sk_stamp); | |
2543 | } | |
2544 | return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0; | |
4ec93edb | 2545 | } |
1da177e4 LT |
2546 | EXPORT_SYMBOL(sock_get_timestamp); |
2547 | ||
ae40eb1e ED |
2548 | int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp) |
2549 | { | |
2550 | struct timespec ts; | |
2551 | if (!sock_flag(sk, SOCK_TIMESTAMP)) | |
20d49473 | 2552 | sock_enable_timestamp(sk, SOCK_TIMESTAMP); |
ae40eb1e ED |
2553 | ts = ktime_to_timespec(sk->sk_stamp); |
2554 | if (ts.tv_sec == -1) | |
2555 | return -ENOENT; | |
2556 | if (ts.tv_sec == 0) { | |
2557 | sk->sk_stamp = ktime_get_real(); | |
2558 | ts = ktime_to_timespec(sk->sk_stamp); | |
2559 | } | |
2560 | return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0; | |
2561 | } | |
2562 | EXPORT_SYMBOL(sock_get_timestampns); | |
2563 | ||
20d49473 | 2564 | void sock_enable_timestamp(struct sock *sk, int flag) |
4ec93edb | 2565 | { |
20d49473 | 2566 | if (!sock_flag(sk, flag)) { |
08e29af3 ED |
2567 | unsigned long previous_flags = sk->sk_flags; |
2568 | ||
20d49473 PO |
2569 | sock_set_flag(sk, flag); |
2570 | /* | |
2571 | * we just set one of the two flags which require net | |
2572 | * time stamping, but time stamping might have been on | |
2573 | * already because of the other one | |
2574 | */ | |
08e29af3 | 2575 | if (!(previous_flags & SK_FLAGS_TIMESTAMP)) |
20d49473 | 2576 | net_enable_timestamp(); |
1da177e4 LT |
2577 | } |
2578 | } | |
1da177e4 LT |
2579 | |
2580 | /* | |
2581 | * Get a socket option on an socket. | |
2582 | * | |
2583 | * FIX: POSIX 1003.1g is very ambiguous here. It states that | |
2584 | * asynchronous errors should be reported by getsockopt. We assume | |
2585 | * this means if you specify SO_ERROR (otherwise whats the point of it). | |
2586 | */ | |
2587 | int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
2588 | char __user *optval, int __user *optlen) | |
2589 | { | |
2590 | struct sock *sk = sock->sk; | |
2591 | ||
2592 | return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen); | |
2593 | } | |
1da177e4 LT |
2594 | EXPORT_SYMBOL(sock_common_getsockopt); |
2595 | ||
3fdadf7d | 2596 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
2597 | int compat_sock_common_getsockopt(struct socket *sock, int level, int optname, |
2598 | char __user *optval, int __user *optlen) | |
3fdadf7d DM |
2599 | { |
2600 | struct sock *sk = sock->sk; | |
2601 | ||
1e51f951 | 2602 | if (sk->sk_prot->compat_getsockopt != NULL) |
543d9cfe ACM |
2603 | return sk->sk_prot->compat_getsockopt(sk, level, optname, |
2604 | optval, optlen); | |
3fdadf7d DM |
2605 | return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen); |
2606 | } | |
2607 | EXPORT_SYMBOL(compat_sock_common_getsockopt); | |
2608 | #endif | |
2609 | ||
1da177e4 LT |
2610 | int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, |
2611 | struct msghdr *msg, size_t size, int flags) | |
2612 | { | |
2613 | struct sock *sk = sock->sk; | |
2614 | int addr_len = 0; | |
2615 | int err; | |
2616 | ||
2617 | err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT, | |
2618 | flags & ~MSG_DONTWAIT, &addr_len); | |
2619 | if (err >= 0) | |
2620 | msg->msg_namelen = addr_len; | |
2621 | return err; | |
2622 | } | |
1da177e4 LT |
2623 | EXPORT_SYMBOL(sock_common_recvmsg); |
2624 | ||
2625 | /* | |
2626 | * Set socket options on an inet socket. | |
2627 | */ | |
2628 | int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
b7058842 | 2629 | char __user *optval, unsigned int optlen) |
1da177e4 LT |
2630 | { |
2631 | struct sock *sk = sock->sk; | |
2632 | ||
2633 | return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen); | |
2634 | } | |
1da177e4 LT |
2635 | EXPORT_SYMBOL(sock_common_setsockopt); |
2636 | ||
3fdadf7d | 2637 | #ifdef CONFIG_COMPAT |
543d9cfe | 2638 | int compat_sock_common_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2639 | char __user *optval, unsigned int optlen) |
3fdadf7d DM |
2640 | { |
2641 | struct sock *sk = sock->sk; | |
2642 | ||
543d9cfe ACM |
2643 | if (sk->sk_prot->compat_setsockopt != NULL) |
2644 | return sk->sk_prot->compat_setsockopt(sk, level, optname, | |
2645 | optval, optlen); | |
3fdadf7d DM |
2646 | return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen); |
2647 | } | |
2648 | EXPORT_SYMBOL(compat_sock_common_setsockopt); | |
2649 | #endif | |
2650 | ||
1da177e4 LT |
2651 | void sk_common_release(struct sock *sk) |
2652 | { | |
2653 | if (sk->sk_prot->destroy) | |
2654 | sk->sk_prot->destroy(sk); | |
2655 | ||
2656 | /* | |
2657 | * Observation: when sock_common_release is called, processes have | |
2658 | * no access to socket. But net still has. | |
2659 | * Step one, detach it from networking: | |
2660 | * | |
2661 | * A. Remove from hash tables. | |
2662 | */ | |
2663 | ||
2664 | sk->sk_prot->unhash(sk); | |
2665 | ||
2666 | /* | |
2667 | * In this point socket cannot receive new packets, but it is possible | |
2668 | * that some packets are in flight because some CPU runs receiver and | |
2669 | * did hash table lookup before we unhashed socket. They will achieve | |
2670 | * receive queue and will be purged by socket destructor. | |
2671 | * | |
2672 | * Also we still have packets pending on receive queue and probably, | |
2673 | * our own packets waiting in device queues. sock_destroy will drain | |
2674 | * receive queue, but transmitted packets will delay socket destruction | |
2675 | * until the last reference will be released. | |
2676 | */ | |
2677 | ||
2678 | sock_orphan(sk); | |
2679 | ||
2680 | xfrm_sk_free_policy(sk); | |
2681 | ||
e6848976 | 2682 | sk_refcnt_debug_release(sk); |
5640f768 ED |
2683 | |
2684 | if (sk->sk_frag.page) { | |
2685 | put_page(sk->sk_frag.page); | |
2686 | sk->sk_frag.page = NULL; | |
2687 | } | |
2688 | ||
1da177e4 LT |
2689 | sock_put(sk); |
2690 | } | |
1da177e4 LT |
2691 | EXPORT_SYMBOL(sk_common_release); |
2692 | ||
13ff3d6f PE |
2693 | #ifdef CONFIG_PROC_FS |
2694 | #define PROTO_INUSE_NR 64 /* should be enough for the first time */ | |
1338d466 PE |
2695 | struct prot_inuse { |
2696 | int val[PROTO_INUSE_NR]; | |
2697 | }; | |
13ff3d6f PE |
2698 | |
2699 | static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR); | |
70ee1159 PE |
2700 | |
2701 | #ifdef CONFIG_NET_NS | |
2702 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int val) | |
2703 | { | |
d6d9ca0f | 2704 | __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val); |
70ee1159 PE |
2705 | } |
2706 | EXPORT_SYMBOL_GPL(sock_prot_inuse_add); | |
2707 | ||
2708 | int sock_prot_inuse_get(struct net *net, struct proto *prot) | |
2709 | { | |
2710 | int cpu, idx = prot->inuse_idx; | |
2711 | int res = 0; | |
2712 | ||
2713 | for_each_possible_cpu(cpu) | |
2714 | res += per_cpu_ptr(net->core.inuse, cpu)->val[idx]; | |
2715 | ||
2716 | return res >= 0 ? res : 0; | |
2717 | } | |
2718 | EXPORT_SYMBOL_GPL(sock_prot_inuse_get); | |
2719 | ||
2c8c1e72 | 2720 | static int __net_init sock_inuse_init_net(struct net *net) |
70ee1159 PE |
2721 | { |
2722 | net->core.inuse = alloc_percpu(struct prot_inuse); | |
2723 | return net->core.inuse ? 0 : -ENOMEM; | |
2724 | } | |
2725 | ||
2c8c1e72 | 2726 | static void __net_exit sock_inuse_exit_net(struct net *net) |
70ee1159 PE |
2727 | { |
2728 | free_percpu(net->core.inuse); | |
2729 | } | |
2730 | ||
2731 | static struct pernet_operations net_inuse_ops = { | |
2732 | .init = sock_inuse_init_net, | |
2733 | .exit = sock_inuse_exit_net, | |
2734 | }; | |
2735 | ||
2736 | static __init int net_inuse_init(void) | |
2737 | { | |
2738 | if (register_pernet_subsys(&net_inuse_ops)) | |
2739 | panic("Cannot initialize net inuse counters"); | |
2740 | ||
2741 | return 0; | |
2742 | } | |
2743 | ||
2744 | core_initcall(net_inuse_init); | |
2745 | #else | |
1338d466 PE |
2746 | static DEFINE_PER_CPU(struct prot_inuse, prot_inuse); |
2747 | ||
c29a0bc4 | 2748 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int val) |
1338d466 | 2749 | { |
d6d9ca0f | 2750 | __this_cpu_add(prot_inuse.val[prot->inuse_idx], val); |
1338d466 PE |
2751 | } |
2752 | EXPORT_SYMBOL_GPL(sock_prot_inuse_add); | |
2753 | ||
c29a0bc4 | 2754 | int sock_prot_inuse_get(struct net *net, struct proto *prot) |
1338d466 PE |
2755 | { |
2756 | int cpu, idx = prot->inuse_idx; | |
2757 | int res = 0; | |
2758 | ||
2759 | for_each_possible_cpu(cpu) | |
2760 | res += per_cpu(prot_inuse, cpu).val[idx]; | |
2761 | ||
2762 | return res >= 0 ? res : 0; | |
2763 | } | |
2764 | EXPORT_SYMBOL_GPL(sock_prot_inuse_get); | |
70ee1159 | 2765 | #endif |
13ff3d6f PE |
2766 | |
2767 | static void assign_proto_idx(struct proto *prot) | |
2768 | { | |
2769 | prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR); | |
2770 | ||
2771 | if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) { | |
e005d193 | 2772 | pr_err("PROTO_INUSE_NR exhausted\n"); |
13ff3d6f PE |
2773 | return; |
2774 | } | |
2775 | ||
2776 | set_bit(prot->inuse_idx, proto_inuse_idx); | |
2777 | } | |
2778 | ||
2779 | static void release_proto_idx(struct proto *prot) | |
2780 | { | |
2781 | if (prot->inuse_idx != PROTO_INUSE_NR - 1) | |
2782 | clear_bit(prot->inuse_idx, proto_inuse_idx); | |
2783 | } | |
2784 | #else | |
2785 | static inline void assign_proto_idx(struct proto *prot) | |
2786 | { | |
2787 | } | |
2788 | ||
2789 | static inline void release_proto_idx(struct proto *prot) | |
2790 | { | |
2791 | } | |
2792 | #endif | |
2793 | ||
b733c007 PE |
2794 | int proto_register(struct proto *prot, int alloc_slab) |
2795 | { | |
1da177e4 LT |
2796 | if (alloc_slab) { |
2797 | prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0, | |
271b72c7 ED |
2798 | SLAB_HWCACHE_ALIGN | prot->slab_flags, |
2799 | NULL); | |
1da177e4 LT |
2800 | |
2801 | if (prot->slab == NULL) { | |
e005d193 JP |
2802 | pr_crit("%s: Can't create sock SLAB cache!\n", |
2803 | prot->name); | |
60e7663d | 2804 | goto out; |
1da177e4 | 2805 | } |
2e6599cb ACM |
2806 | |
2807 | if (prot->rsk_prot != NULL) { | |
faf23422 | 2808 | prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name); |
7e56b5d6 | 2809 | if (prot->rsk_prot->slab_name == NULL) |
2e6599cb ACM |
2810 | goto out_free_sock_slab; |
2811 | ||
7e56b5d6 | 2812 | prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name, |
2e6599cb | 2813 | prot->rsk_prot->obj_size, 0, |
20c2df83 | 2814 | SLAB_HWCACHE_ALIGN, NULL); |
2e6599cb ACM |
2815 | |
2816 | if (prot->rsk_prot->slab == NULL) { | |
e005d193 JP |
2817 | pr_crit("%s: Can't create request sock SLAB cache!\n", |
2818 | prot->name); | |
2e6599cb ACM |
2819 | goto out_free_request_sock_slab_name; |
2820 | } | |
2821 | } | |
8feaf0c0 | 2822 | |
6d6ee43e | 2823 | if (prot->twsk_prot != NULL) { |
faf23422 | 2824 | prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name); |
8feaf0c0 | 2825 | |
7e56b5d6 | 2826 | if (prot->twsk_prot->twsk_slab_name == NULL) |
8feaf0c0 ACM |
2827 | goto out_free_request_sock_slab; |
2828 | ||
6d6ee43e | 2829 | prot->twsk_prot->twsk_slab = |
7e56b5d6 | 2830 | kmem_cache_create(prot->twsk_prot->twsk_slab_name, |
6d6ee43e | 2831 | prot->twsk_prot->twsk_obj_size, |
3ab5aee7 ED |
2832 | 0, |
2833 | SLAB_HWCACHE_ALIGN | | |
2834 | prot->slab_flags, | |
20c2df83 | 2835 | NULL); |
6d6ee43e | 2836 | if (prot->twsk_prot->twsk_slab == NULL) |
8feaf0c0 ACM |
2837 | goto out_free_timewait_sock_slab_name; |
2838 | } | |
1da177e4 LT |
2839 | } |
2840 | ||
36b77a52 | 2841 | mutex_lock(&proto_list_mutex); |
1da177e4 | 2842 | list_add(&prot->node, &proto_list); |
13ff3d6f | 2843 | assign_proto_idx(prot); |
36b77a52 | 2844 | mutex_unlock(&proto_list_mutex); |
b733c007 PE |
2845 | return 0; |
2846 | ||
8feaf0c0 | 2847 | out_free_timewait_sock_slab_name: |
7e56b5d6 | 2848 | kfree(prot->twsk_prot->twsk_slab_name); |
8feaf0c0 ACM |
2849 | out_free_request_sock_slab: |
2850 | if (prot->rsk_prot && prot->rsk_prot->slab) { | |
2851 | kmem_cache_destroy(prot->rsk_prot->slab); | |
2852 | prot->rsk_prot->slab = NULL; | |
2853 | } | |
2e6599cb | 2854 | out_free_request_sock_slab_name: |
72150e9b DC |
2855 | if (prot->rsk_prot) |
2856 | kfree(prot->rsk_prot->slab_name); | |
2e6599cb ACM |
2857 | out_free_sock_slab: |
2858 | kmem_cache_destroy(prot->slab); | |
2859 | prot->slab = NULL; | |
b733c007 PE |
2860 | out: |
2861 | return -ENOBUFS; | |
1da177e4 | 2862 | } |
1da177e4 LT |
2863 | EXPORT_SYMBOL(proto_register); |
2864 | ||
2865 | void proto_unregister(struct proto *prot) | |
2866 | { | |
36b77a52 | 2867 | mutex_lock(&proto_list_mutex); |
13ff3d6f | 2868 | release_proto_idx(prot); |
0a3f4358 | 2869 | list_del(&prot->node); |
36b77a52 | 2870 | mutex_unlock(&proto_list_mutex); |
1da177e4 LT |
2871 | |
2872 | if (prot->slab != NULL) { | |
2873 | kmem_cache_destroy(prot->slab); | |
2874 | prot->slab = NULL; | |
2875 | } | |
2876 | ||
2e6599cb | 2877 | if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) { |
2e6599cb | 2878 | kmem_cache_destroy(prot->rsk_prot->slab); |
7e56b5d6 | 2879 | kfree(prot->rsk_prot->slab_name); |
2e6599cb ACM |
2880 | prot->rsk_prot->slab = NULL; |
2881 | } | |
2882 | ||
6d6ee43e | 2883 | if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) { |
6d6ee43e | 2884 | kmem_cache_destroy(prot->twsk_prot->twsk_slab); |
7e56b5d6 | 2885 | kfree(prot->twsk_prot->twsk_slab_name); |
6d6ee43e | 2886 | prot->twsk_prot->twsk_slab = NULL; |
8feaf0c0 | 2887 | } |
1da177e4 | 2888 | } |
1da177e4 LT |
2889 | EXPORT_SYMBOL(proto_unregister); |
2890 | ||
2891 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 2892 | static void *proto_seq_start(struct seq_file *seq, loff_t *pos) |
36b77a52 | 2893 | __acquires(proto_list_mutex) |
1da177e4 | 2894 | { |
36b77a52 | 2895 | mutex_lock(&proto_list_mutex); |
60f0438a | 2896 | return seq_list_start_head(&proto_list, *pos); |
1da177e4 LT |
2897 | } |
2898 | ||
2899 | static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2900 | { | |
60f0438a | 2901 | return seq_list_next(v, &proto_list, pos); |
1da177e4 LT |
2902 | } |
2903 | ||
2904 | static void proto_seq_stop(struct seq_file *seq, void *v) | |
36b77a52 | 2905 | __releases(proto_list_mutex) |
1da177e4 | 2906 | { |
36b77a52 | 2907 | mutex_unlock(&proto_list_mutex); |
1da177e4 LT |
2908 | } |
2909 | ||
2910 | static char proto_method_implemented(const void *method) | |
2911 | { | |
2912 | return method == NULL ? 'n' : 'y'; | |
2913 | } | |
180d8cd9 GC |
2914 | static long sock_prot_memory_allocated(struct proto *proto) |
2915 | { | |
cb75a36c | 2916 | return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L; |
180d8cd9 GC |
2917 | } |
2918 | ||
2919 | static char *sock_prot_memory_pressure(struct proto *proto) | |
2920 | { | |
2921 | return proto->memory_pressure != NULL ? | |
2922 | proto_memory_pressure(proto) ? "yes" : "no" : "NI"; | |
2923 | } | |
1da177e4 LT |
2924 | |
2925 | static void proto_seq_printf(struct seq_file *seq, struct proto *proto) | |
2926 | { | |
180d8cd9 | 2927 | |
8d987e5c | 2928 | seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s " |
1da177e4 LT |
2929 | "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n", |
2930 | proto->name, | |
2931 | proto->obj_size, | |
14e943db | 2932 | sock_prot_inuse_get(seq_file_net(seq), proto), |
180d8cd9 GC |
2933 | sock_prot_memory_allocated(proto), |
2934 | sock_prot_memory_pressure(proto), | |
1da177e4 LT |
2935 | proto->max_header, |
2936 | proto->slab == NULL ? "no" : "yes", | |
2937 | module_name(proto->owner), | |
2938 | proto_method_implemented(proto->close), | |
2939 | proto_method_implemented(proto->connect), | |
2940 | proto_method_implemented(proto->disconnect), | |
2941 | proto_method_implemented(proto->accept), | |
2942 | proto_method_implemented(proto->ioctl), | |
2943 | proto_method_implemented(proto->init), | |
2944 | proto_method_implemented(proto->destroy), | |
2945 | proto_method_implemented(proto->shutdown), | |
2946 | proto_method_implemented(proto->setsockopt), | |
2947 | proto_method_implemented(proto->getsockopt), | |
2948 | proto_method_implemented(proto->sendmsg), | |
2949 | proto_method_implemented(proto->recvmsg), | |
2950 | proto_method_implemented(proto->sendpage), | |
2951 | proto_method_implemented(proto->bind), | |
2952 | proto_method_implemented(proto->backlog_rcv), | |
2953 | proto_method_implemented(proto->hash), | |
2954 | proto_method_implemented(proto->unhash), | |
2955 | proto_method_implemented(proto->get_port), | |
2956 | proto_method_implemented(proto->enter_memory_pressure)); | |
2957 | } | |
2958 | ||
2959 | static int proto_seq_show(struct seq_file *seq, void *v) | |
2960 | { | |
60f0438a | 2961 | if (v == &proto_list) |
1da177e4 LT |
2962 | seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s", |
2963 | "protocol", | |
2964 | "size", | |
2965 | "sockets", | |
2966 | "memory", | |
2967 | "press", | |
2968 | "maxhdr", | |
2969 | "slab", | |
2970 | "module", | |
2971 | "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n"); | |
2972 | else | |
60f0438a | 2973 | proto_seq_printf(seq, list_entry(v, struct proto, node)); |
1da177e4 LT |
2974 | return 0; |
2975 | } | |
2976 | ||
f690808e | 2977 | static const struct seq_operations proto_seq_ops = { |
1da177e4 LT |
2978 | .start = proto_seq_start, |
2979 | .next = proto_seq_next, | |
2980 | .stop = proto_seq_stop, | |
2981 | .show = proto_seq_show, | |
2982 | }; | |
2983 | ||
2984 | static int proto_seq_open(struct inode *inode, struct file *file) | |
2985 | { | |
14e943db ED |
2986 | return seq_open_net(inode, file, &proto_seq_ops, |
2987 | sizeof(struct seq_net_private)); | |
1da177e4 LT |
2988 | } |
2989 | ||
9a32144e | 2990 | static const struct file_operations proto_seq_fops = { |
1da177e4 LT |
2991 | .owner = THIS_MODULE, |
2992 | .open = proto_seq_open, | |
2993 | .read = seq_read, | |
2994 | .llseek = seq_lseek, | |
14e943db ED |
2995 | .release = seq_release_net, |
2996 | }; | |
2997 | ||
2998 | static __net_init int proto_init_net(struct net *net) | |
2999 | { | |
d4beaa66 | 3000 | if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops)) |
14e943db ED |
3001 | return -ENOMEM; |
3002 | ||
3003 | return 0; | |
3004 | } | |
3005 | ||
3006 | static __net_exit void proto_exit_net(struct net *net) | |
3007 | { | |
ece31ffd | 3008 | remove_proc_entry("protocols", net->proc_net); |
14e943db ED |
3009 | } |
3010 | ||
3011 | ||
3012 | static __net_initdata struct pernet_operations proto_net_ops = { | |
3013 | .init = proto_init_net, | |
3014 | .exit = proto_exit_net, | |
1da177e4 LT |
3015 | }; |
3016 | ||
3017 | static int __init proto_init(void) | |
3018 | { | |
14e943db | 3019 | return register_pernet_subsys(&proto_net_ops); |
1da177e4 LT |
3020 | } |
3021 | ||
3022 | subsys_initcall(proto_init); | |
3023 | ||
3024 | #endif /* PROC_FS */ |