~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/net/sock.h

Version: ~ [ linux-5.11-rc3 ] ~ [ linux-5.10.7 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.89 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.167 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.215 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.251 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.251 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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  *              Definitions for the AF_INET socket handler.
  7  *
  8  * Version:     @(#)sock.h      1.0.4   05/13/93
  9  *
 10  * Authors:     Ross Biro
 11  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 12  *              Corey Minyard <wf-rch!minyard@relay.EU.net>
 13  *              Florian La Roche <flla@stud.uni-sb.de>
 14  *
 15  * Fixes:
 16  *              Alan Cox        :       Volatiles in skbuff pointers. See
 17  *                                      skbuff comments. May be overdone,
 18  *                                      better to prove they can be removed
 19  *                                      than the reverse.
 20  *              Alan Cox        :       Added a zapped field for tcp to note
 21  *                                      a socket is reset and must stay shut up
 22  *              Alan Cox        :       New fields for options
 23  *      Pauline Middelink       :       identd support
 24  *              Alan Cox        :       Eliminate low level recv/recvfrom
 25  *              David S. Miller :       New socket lookup architecture.
 26  *              Steve Whitehouse:       Default routines for sock_ops
 27  *              Arnaldo C. Melo :       removed net_pinfo, tp_pinfo and made
 28  *                                      protinfo be just a void pointer, as the
 29  *                                      protocol specific parts were moved to
 30  *                                      respective headers and ipv4/v6, etc now
 31  *                                      use private slabcaches for its socks
 32  *              Pedro Hortas    :       New flags field for socket options
 33  *
 34  *
 35  *              This program is free software; you can redistribute it and/or
 36  *              modify it under the terms of the GNU General Public License
 37  *              as published by the Free Software Foundation; either version
 38  *              2 of the License, or (at your option) any later version.
 39  */
 40 #ifndef _SOCK_H
 41 #define _SOCK_H
 42 
 43 #include <linux/hardirq.h>
 44 #include <linux/kernel.h>
 45 #include <linux/list.h>
 46 #include <linux/list_nulls.h>
 47 #include <linux/timer.h>
 48 #include <linux/cache.h>
 49 #include <linux/bitops.h>
 50 #include <linux/lockdep.h>
 51 #include <linux/netdevice.h>
 52 #include <linux/skbuff.h>       /* struct sk_buff */
 53 #include <linux/mm.h>
 54 #include <linux/security.h>
 55 #include <linux/slab.h>
 56 #include <linux/uaccess.h>
 57 #include <linux/memcontrol.h>
 58 #include <linux/res_counter.h>
 59 #include <linux/static_key.h>
 60 #include <linux/aio.h>
 61 #include <linux/sched.h>
 62 
 63 #include <linux/filter.h>
 64 #include <linux/rculist_nulls.h>
 65 #include <linux/poll.h>
 66 
 67 #include <linux/atomic.h>
 68 #include <net/dst.h>
 69 #include <net/checksum.h>
 70 
 71 struct cgroup;
 72 struct cgroup_subsys;
 73 #ifdef CONFIG_NET
 74 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
 75 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
 76 #else
 77 static inline
 78 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
 79 {
 80         return 0;
 81 }
 82 static inline
 83 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
 84 {
 85 }
 86 #endif
 87 /*
 88  * This structure really needs to be cleaned up.
 89  * Most of it is for TCP, and not used by any of
 90  * the other protocols.
 91  */
 92 
 93 /* Define this to get the SOCK_DBG debugging facility. */
 94 #define SOCK_DEBUGGING
 95 #ifdef SOCK_DEBUGGING
 96 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
 97                                         printk(KERN_DEBUG msg); } while (0)
 98 #else
 99 /* Validate arguments and do nothing */
100 static inline __printf(2, 3)
101 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
102 {
103 }
104 #endif
105 
106 /* This is the per-socket lock.  The spinlock provides a synchronization
107  * between user contexts and software interrupt processing, whereas the
108  * mini-semaphore synchronizes multiple users amongst themselves.
109  */
110 typedef struct {
111         spinlock_t              slock;
112         int                     owned;
113         wait_queue_head_t       wq;
114         /*
115          * We express the mutex-alike socket_lock semantics
116          * to the lock validator by explicitly managing
117          * the slock as a lock variant (in addition to
118          * the slock itself):
119          */
120 #ifdef CONFIG_DEBUG_LOCK_ALLOC
121         struct lockdep_map dep_map;
122 #endif
123 } socket_lock_t;
124 
125 struct sock;
126 struct proto;
127 struct net;
128 
129 typedef __u32 __bitwise __portpair;
130 typedef __u64 __bitwise __addrpair;
131 
132 /**
133  *      struct sock_common - minimal network layer representation of sockets
134  *      @skc_daddr: Foreign IPv4 addr
135  *      @skc_rcv_saddr: Bound local IPv4 addr
136  *      @skc_hash: hash value used with various protocol lookup tables
137  *      @skc_u16hashes: two u16 hash values used by UDP lookup tables
138  *      @skc_dport: placeholder for inet_dport/tw_dport
139  *      @skc_num: placeholder for inet_num/tw_num
140  *      @skc_family: network address family
141  *      @skc_state: Connection state
142  *      @skc_reuse: %SO_REUSEADDR setting
143  *      @skc_reuseport: %SO_REUSEPORT setting
144  *      @skc_bound_dev_if: bound device index if != 0
145  *      @skc_bind_node: bind hash linkage for various protocol lookup tables
146  *      @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
147  *      @skc_prot: protocol handlers inside a network family
148  *      @skc_net: reference to the network namespace of this socket
149  *      @skc_node: main hash linkage for various protocol lookup tables
150  *      @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
151  *      @skc_tx_queue_mapping: tx queue number for this connection
152  *      @skc_refcnt: reference count
153  *
154  *      This is the minimal network layer representation of sockets, the header
155  *      for struct sock and struct inet_timewait_sock.
156  */
157 struct sock_common {
158         /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
159          * address on 64bit arches : cf INET_MATCH()
160          */
161         union {
162                 __addrpair      skc_addrpair;
163                 struct {
164                         __be32  skc_daddr;
165                         __be32  skc_rcv_saddr;
166                 };
167         };
168         union  {
169                 unsigned int    skc_hash;
170                 __u16           skc_u16hashes[2];
171         };
172         /* skc_dport && skc_num must be grouped as well */
173         union {
174                 __portpair      skc_portpair;
175                 struct {
176                         __be16  skc_dport;
177                         __u16   skc_num;
178                 };
179         };
180 
181         unsigned short          skc_family;
182         volatile unsigned char  skc_state;
183         unsigned char           skc_reuse:4;
184         unsigned char           skc_reuseport:4;
185         int                     skc_bound_dev_if;
186         union {
187                 struct hlist_node       skc_bind_node;
188                 struct hlist_nulls_node skc_portaddr_node;
189         };
190         struct proto            *skc_prot;
191 #ifdef CONFIG_NET_NS
192         struct net              *skc_net;
193 #endif
194 
195 #if IS_ENABLED(CONFIG_IPV6)
196         struct in6_addr         skc_v6_daddr;
197         struct in6_addr         skc_v6_rcv_saddr;
198 #endif
199 
200         /*
201          * fields between dontcopy_begin/dontcopy_end
202          * are not copied in sock_copy()
203          */
204         /* private: */
205         int                     skc_dontcopy_begin[0];
206         /* public: */
207         union {
208                 struct hlist_node       skc_node;
209                 struct hlist_nulls_node skc_nulls_node;
210         };
211         int                     skc_tx_queue_mapping;
212         atomic_t                skc_refcnt;
213         /* private: */
214         int                     skc_dontcopy_end[0];
215         /* public: */
216 };
217 
218 struct cg_proto;
219 /**
220   *     struct sock - network layer representation of sockets
221   *     @__sk_common: shared layout with inet_timewait_sock
222   *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
223   *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
224   *     @sk_lock:       synchronizer
225   *     @sk_rcvbuf: size of receive buffer in bytes
226   *     @sk_wq: sock wait queue and async head
227   *     @sk_rx_dst: receive input route used by early demux
228   *     @sk_dst_cache: destination cache
229   *     @sk_dst_lock: destination cache lock
230   *     @sk_policy: flow policy
231   *     @sk_receive_queue: incoming packets
232   *     @sk_wmem_alloc: transmit queue bytes committed
233   *     @sk_write_queue: Packet sending queue
234   *     @sk_async_wait_queue: DMA copied packets
235   *     @sk_omem_alloc: "o" is "option" or "other"
236   *     @sk_wmem_queued: persistent queue size
237   *     @sk_forward_alloc: space allocated forward
238   *     @sk_napi_id: id of the last napi context to receive data for sk
239   *     @sk_ll_usec: usecs to busypoll when there is no data
240   *     @sk_allocation: allocation mode
241   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
242   *     @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
243   *     @sk_sndbuf: size of send buffer in bytes
244   *     @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
245   *                %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
246   *     @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
247   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
248   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
249   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
250   *     @sk_gso_max_size: Maximum GSO segment size to build
251   *     @sk_gso_max_segs: Maximum number of GSO segments
252   *     @sk_lingertime: %SO_LINGER l_linger setting
253   *     @sk_backlog: always used with the per-socket spinlock held
254   *     @sk_callback_lock: used with the callbacks in the end of this struct
255   *     @sk_error_queue: rarely used
256   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
257   *                       IPV6_ADDRFORM for instance)
258   *     @sk_err: last error
259   *     @sk_err_soft: errors that don't cause failure but are the cause of a
260   *                   persistent failure not just 'timed out'
261   *     @sk_drops: raw/udp drops counter
262   *     @sk_ack_backlog: current listen backlog
263   *     @sk_max_ack_backlog: listen backlog set in listen()
264   *     @sk_priority: %SO_PRIORITY setting
265   *     @sk_cgrp_prioidx: socket group's priority map index
266   *     @sk_type: socket type (%SOCK_STREAM, etc)
267   *     @sk_protocol: which protocol this socket belongs in this network family
268   *     @sk_peer_pid: &struct pid for this socket's peer
269   *     @sk_peer_cred: %SO_PEERCRED setting
270   *     @sk_rcvlowat: %SO_RCVLOWAT setting
271   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
272   *     @sk_sndtimeo: %SO_SNDTIMEO setting
273   *     @sk_rxhash: flow hash received from netif layer
274   *     @sk_filter: socket filtering instructions
275   *     @sk_protinfo: private area, net family specific, when not using slab
276   *     @sk_timer: sock cleanup timer
277   *     @sk_stamp: time stamp of last packet received
278   *     @sk_socket: Identd and reporting IO signals
279   *     @sk_user_data: RPC layer private data
280   *     @sk_frag: cached page frag
281   *     @sk_peek_off: current peek_offset value
282   *     @sk_send_head: front of stuff to transmit
283   *     @sk_security: used by security modules
284   *     @sk_mark: generic packet mark
285   *     @sk_classid: this socket's cgroup classid
286   *     @sk_cgrp: this socket's cgroup-specific proto data
287   *     @sk_write_pending: a write to stream socket waits to start
288   *     @sk_state_change: callback to indicate change in the state of the sock
289   *     @sk_data_ready: callback to indicate there is data to be processed
290   *     @sk_write_space: callback to indicate there is bf sending space available
291   *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
292   *     @sk_backlog_rcv: callback to process the backlog
293   *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
294  */
295 struct sock {
296         /*
297          * Now struct inet_timewait_sock also uses sock_common, so please just
298          * don't add nothing before this first member (__sk_common) --acme
299          */
300         struct sock_common      __sk_common;
301 #define sk_node                 __sk_common.skc_node
302 #define sk_nulls_node           __sk_common.skc_nulls_node
303 #define sk_refcnt               __sk_common.skc_refcnt
304 #define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
305 
306 #define sk_dontcopy_begin       __sk_common.skc_dontcopy_begin
307 #define sk_dontcopy_end         __sk_common.skc_dontcopy_end
308 #define sk_hash                 __sk_common.skc_hash
309 #define sk_portpair             __sk_common.skc_portpair
310 #define sk_num                  __sk_common.skc_num
311 #define sk_dport                __sk_common.skc_dport
312 #define sk_addrpair             __sk_common.skc_addrpair
313 #define sk_daddr                __sk_common.skc_daddr
314 #define sk_rcv_saddr            __sk_common.skc_rcv_saddr
315 #define sk_family               __sk_common.skc_family
316 #define sk_state                __sk_common.skc_state
317 #define sk_reuse                __sk_common.skc_reuse
318 #define sk_reuseport            __sk_common.skc_reuseport
319 #define sk_bound_dev_if         __sk_common.skc_bound_dev_if
320 #define sk_bind_node            __sk_common.skc_bind_node
321 #define sk_prot                 __sk_common.skc_prot
322 #define sk_net                  __sk_common.skc_net
323 #define sk_v6_daddr             __sk_common.skc_v6_daddr
324 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
325 
326         socket_lock_t           sk_lock;
327         struct sk_buff_head     sk_receive_queue;
328         /*
329          * The backlog queue is special, it is always used with
330          * the per-socket spinlock held and requires low latency
331          * access. Therefore we special case it's implementation.
332          * Note : rmem_alloc is in this structure to fill a hole
333          * on 64bit arches, not because its logically part of
334          * backlog.
335          */
336         struct {
337                 atomic_t        rmem_alloc;
338                 int             len;
339                 struct sk_buff  *head;
340                 struct sk_buff  *tail;
341         } sk_backlog;
342 #define sk_rmem_alloc sk_backlog.rmem_alloc
343         int                     sk_forward_alloc;
344 #ifdef CONFIG_RPS
345         __u32                   sk_rxhash;
346 #endif
347 #ifdef CONFIG_NET_RX_BUSY_POLL
348         unsigned int            sk_napi_id;
349         unsigned int            sk_ll_usec;
350 #endif
351         atomic_t                sk_drops;
352         int                     sk_rcvbuf;
353 
354         struct sk_filter __rcu  *sk_filter;
355         struct socket_wq __rcu  *sk_wq;
356 
357 #ifdef CONFIG_NET_DMA
358         struct sk_buff_head     sk_async_wait_queue;
359 #endif
360 
361 #ifdef CONFIG_XFRM
362         struct xfrm_policy      *sk_policy[2];
363 #endif
364         unsigned long           sk_flags;
365         struct dst_entry        *sk_rx_dst;
366         struct dst_entry __rcu  *sk_dst_cache;
367         spinlock_t              sk_dst_lock;
368         atomic_t                sk_wmem_alloc;
369         atomic_t                sk_omem_alloc;
370         int                     sk_sndbuf;
371         struct sk_buff_head     sk_write_queue;
372         kmemcheck_bitfield_begin(flags);
373         unsigned int            sk_shutdown  : 2,
374                                 sk_no_check  : 2,
375                                 sk_userlocks : 4,
376                                 sk_protocol  : 8,
377                                 sk_type      : 16;
378         kmemcheck_bitfield_end(flags);
379         int                     sk_wmem_queued;
380         gfp_t                   sk_allocation;
381         u32                     sk_pacing_rate; /* bytes per second */
382         u32                     sk_max_pacing_rate;
383         netdev_features_t       sk_route_caps;
384         netdev_features_t       sk_route_nocaps;
385         int                     sk_gso_type;
386         unsigned int            sk_gso_max_size;
387         u16                     sk_gso_max_segs;
388         int                     sk_rcvlowat;
389         unsigned long           sk_lingertime;
390         struct sk_buff_head     sk_error_queue;
391         struct proto            *sk_prot_creator;
392         rwlock_t                sk_callback_lock;
393         int                     sk_err,
394                                 sk_err_soft;
395         unsigned short          sk_ack_backlog;
396         unsigned short          sk_max_ack_backlog;
397         __u32                   sk_priority;
398 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
399         __u32                   sk_cgrp_prioidx;
400 #endif
401         struct pid              *sk_peer_pid;
402         const struct cred       *sk_peer_cred;
403         long                    sk_rcvtimeo;
404         long                    sk_sndtimeo;
405         void                    *sk_protinfo;
406         struct timer_list       sk_timer;
407         ktime_t                 sk_stamp;
408         struct socket           *sk_socket;
409         void                    *sk_user_data;
410         struct page_frag        sk_frag;
411         struct sk_buff          *sk_send_head;
412         __s32                   sk_peek_off;
413         int                     sk_write_pending;
414 #ifdef CONFIG_SECURITY
415         void                    *sk_security;
416 #endif
417         __u32                   sk_mark;
418         u32                     sk_classid;
419         struct cg_proto         *sk_cgrp;
420         void                    (*sk_state_change)(struct sock *sk);
421         void                    (*sk_data_ready)(struct sock *sk, int bytes);
422         void                    (*sk_write_space)(struct sock *sk);
423         void                    (*sk_error_report)(struct sock *sk);
424         int                     (*sk_backlog_rcv)(struct sock *sk,
425                                                   struct sk_buff *skb);
426         void                    (*sk_destruct)(struct sock *sk);
427 };
428 
429 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
430 
431 #define rcu_dereference_sk_user_data(sk)        rcu_dereference(__sk_user_data((sk)))
432 #define rcu_assign_sk_user_data(sk, ptr)        rcu_assign_pointer(__sk_user_data((sk)), ptr)
433 
434 /*
435  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
436  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
437  * on a socket means that the socket will reuse everybody else's port
438  * without looking at the other's sk_reuse value.
439  */
440 
441 #define SK_NO_REUSE     0
442 #define SK_CAN_REUSE    1
443 #define SK_FORCE_REUSE  2
444 
445 static inline int sk_peek_offset(struct sock *sk, int flags)
446 {
447         if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
448                 return sk->sk_peek_off;
449         else
450                 return 0;
451 }
452 
453 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
454 {
455         if (sk->sk_peek_off >= 0) {
456                 if (sk->sk_peek_off >= val)
457                         sk->sk_peek_off -= val;
458                 else
459                         sk->sk_peek_off = 0;
460         }
461 }
462 
463 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
464 {
465         if (sk->sk_peek_off >= 0)
466                 sk->sk_peek_off += val;
467 }
468 
469 /*
470  * Hashed lists helper routines
471  */
472 static inline struct sock *sk_entry(const struct hlist_node *node)
473 {
474         return hlist_entry(node, struct sock, sk_node);
475 }
476 
477 static inline struct sock *__sk_head(const struct hlist_head *head)
478 {
479         return hlist_entry(head->first, struct sock, sk_node);
480 }
481 
482 static inline struct sock *sk_head(const struct hlist_head *head)
483 {
484         return hlist_empty(head) ? NULL : __sk_head(head);
485 }
486 
487 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
488 {
489         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
490 }
491 
492 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
493 {
494         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
495 }
496 
497 static inline struct sock *sk_next(const struct sock *sk)
498 {
499         return sk->sk_node.next ?
500                 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
501 }
502 
503 static inline struct sock *sk_nulls_next(const struct sock *sk)
504 {
505         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
506                 hlist_nulls_entry(sk->sk_nulls_node.next,
507                                   struct sock, sk_nulls_node) :
508                 NULL;
509 }
510 
511 static inline bool sk_unhashed(const struct sock *sk)
512 {
513         return hlist_unhashed(&sk->sk_node);
514 }
515 
516 static inline bool sk_hashed(const struct sock *sk)
517 {
518         return !sk_unhashed(sk);
519 }
520 
521 static inline void sk_node_init(struct hlist_node *node)
522 {
523         node->pprev = NULL;
524 }
525 
526 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
527 {
528         node->pprev = NULL;
529 }
530 
531 static inline void __sk_del_node(struct sock *sk)
532 {
533         __hlist_del(&sk->sk_node);
534 }
535 
536 /* NB: equivalent to hlist_del_init_rcu */
537 static inline bool __sk_del_node_init(struct sock *sk)
538 {
539         if (sk_hashed(sk)) {
540                 __sk_del_node(sk);
541                 sk_node_init(&sk->sk_node);
542                 return true;
543         }
544         return false;
545 }
546 
547 /* Grab socket reference count. This operation is valid only
548    when sk is ALREADY grabbed f.e. it is found in hash table
549    or a list and the lookup is made under lock preventing hash table
550    modifications.
551  */
552 
553 static inline void sock_hold(struct sock *sk)
554 {
555         atomic_inc(&sk->sk_refcnt);
556 }
557 
558 /* Ungrab socket in the context, which assumes that socket refcnt
559    cannot hit zero, f.e. it is true in context of any socketcall.
560  */
561 static inline void __sock_put(struct sock *sk)
562 {
563         atomic_dec(&sk->sk_refcnt);
564 }
565 
566 static inline bool sk_del_node_init(struct sock *sk)
567 {
568         bool rc = __sk_del_node_init(sk);
569 
570         if (rc) {
571                 /* paranoid for a while -acme */
572                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
573                 __sock_put(sk);
574         }
575         return rc;
576 }
577 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
578 
579 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
580 {
581         if (sk_hashed(sk)) {
582                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
583                 return true;
584         }
585         return false;
586 }
587 
588 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
589 {
590         bool rc = __sk_nulls_del_node_init_rcu(sk);
591 
592         if (rc) {
593                 /* paranoid for a while -acme */
594                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
595                 __sock_put(sk);
596         }
597         return rc;
598 }
599 
600 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
601 {
602         hlist_add_head(&sk->sk_node, list);
603 }
604 
605 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
606 {
607         sock_hold(sk);
608         __sk_add_node(sk, list);
609 }
610 
611 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
612 {
613         sock_hold(sk);
614         hlist_add_head_rcu(&sk->sk_node, list);
615 }
616 
617 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
618 {
619         hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
620 }
621 
622 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
623 {
624         sock_hold(sk);
625         __sk_nulls_add_node_rcu(sk, list);
626 }
627 
628 static inline void __sk_del_bind_node(struct sock *sk)
629 {
630         __hlist_del(&sk->sk_bind_node);
631 }
632 
633 static inline void sk_add_bind_node(struct sock *sk,
634                                         struct hlist_head *list)
635 {
636         hlist_add_head(&sk->sk_bind_node, list);
637 }
638 
639 #define sk_for_each(__sk, list) \
640         hlist_for_each_entry(__sk, list, sk_node)
641 #define sk_for_each_rcu(__sk, list) \
642         hlist_for_each_entry_rcu(__sk, list, sk_node)
643 #define sk_nulls_for_each(__sk, node, list) \
644         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
645 #define sk_nulls_for_each_rcu(__sk, node, list) \
646         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
647 #define sk_for_each_from(__sk) \
648         hlist_for_each_entry_from(__sk, sk_node)
649 #define sk_nulls_for_each_from(__sk, node) \
650         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
651                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
652 #define sk_for_each_safe(__sk, tmp, list) \
653         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
654 #define sk_for_each_bound(__sk, list) \
655         hlist_for_each_entry(__sk, list, sk_bind_node)
656 
657 static inline struct user_namespace *sk_user_ns(struct sock *sk)
658 {
659         /* Careful only use this in a context where these parameters
660          * can not change and must all be valid, such as recvmsg from
661          * userspace.
662          */
663         return sk->sk_socket->file->f_cred->user_ns;
664 }
665 
666 /* Sock flags */
667 enum sock_flags {
668         SOCK_DEAD,
669         SOCK_DONE,
670         SOCK_URGINLINE,
671         SOCK_KEEPOPEN,
672         SOCK_LINGER,
673         SOCK_DESTROY,
674         SOCK_BROADCAST,
675         SOCK_TIMESTAMP,
676         SOCK_ZAPPED,
677         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
678         SOCK_DBG, /* %SO_DEBUG setting */
679         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
680         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
681         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
682         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
683         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
684         SOCK_TIMESTAMPING_TX_HARDWARE,  /* %SOF_TIMESTAMPING_TX_HARDWARE */
685         SOCK_TIMESTAMPING_TX_SOFTWARE,  /* %SOF_TIMESTAMPING_TX_SOFTWARE */
686         SOCK_TIMESTAMPING_RX_HARDWARE,  /* %SOF_TIMESTAMPING_RX_HARDWARE */
687         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
688         SOCK_TIMESTAMPING_SOFTWARE,     /* %SOF_TIMESTAMPING_SOFTWARE */
689         SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
690         SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
691         SOCK_FASYNC, /* fasync() active */
692         SOCK_RXQ_OVFL,
693         SOCK_ZEROCOPY, /* buffers from userspace */
694         SOCK_WIFI_STATUS, /* push wifi status to userspace */
695         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
696                      * Will use last 4 bytes of packet sent from
697                      * user-space instead.
698                      */
699         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
700         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
701 };
702 
703 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
704 {
705         nsk->sk_flags = osk->sk_flags;
706 }
707 
708 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
709 {
710         __set_bit(flag, &sk->sk_flags);
711 }
712 
713 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
714 {
715         __clear_bit(flag, &sk->sk_flags);
716 }
717 
718 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
719 {
720         return test_bit(flag, &sk->sk_flags);
721 }
722 
723 #ifdef CONFIG_NET
724 extern struct static_key memalloc_socks;
725 static inline int sk_memalloc_socks(void)
726 {
727         return static_key_false(&memalloc_socks);
728 }
729 #else
730 
731 static inline int sk_memalloc_socks(void)
732 {
733         return 0;
734 }
735 
736 #endif
737 
738 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
739 {
740         return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
741 }
742 
743 static inline void sk_acceptq_removed(struct sock *sk)
744 {
745         sk->sk_ack_backlog--;
746 }
747 
748 static inline void sk_acceptq_added(struct sock *sk)
749 {
750         sk->sk_ack_backlog++;
751 }
752 
753 static inline bool sk_acceptq_is_full(const struct sock *sk)
754 {
755         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
756 }
757 
758 /*
759  * Compute minimal free write space needed to queue new packets.
760  */
761 static inline int sk_stream_min_wspace(const struct sock *sk)
762 {
763         return sk->sk_wmem_queued >> 1;
764 }
765 
766 static inline int sk_stream_wspace(const struct sock *sk)
767 {
768         return sk->sk_sndbuf - sk->sk_wmem_queued;
769 }
770 
771 void sk_stream_write_space(struct sock *sk);
772 
773 /* OOB backlog add */
774 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
775 {
776         /* dont let skb dst not refcounted, we are going to leave rcu lock */
777         skb_dst_force(skb);
778 
779         if (!sk->sk_backlog.tail)
780                 sk->sk_backlog.head = skb;
781         else
782                 sk->sk_backlog.tail->next = skb;
783 
784         sk->sk_backlog.tail = skb;
785         skb->next = NULL;
786 }
787 
788 /*
789  * Take into account size of receive queue and backlog queue
790  * Do not take into account this skb truesize,
791  * to allow even a single big packet to come.
792  */
793 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
794                                      unsigned int limit)
795 {
796         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
797 
798         return qsize > limit;
799 }
800 
801 /* The per-socket spinlock must be held here. */
802 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
803                                               unsigned int limit)
804 {
805         if (sk_rcvqueues_full(sk, skb, limit))
806                 return -ENOBUFS;
807 
808         __sk_add_backlog(sk, skb);
809         sk->sk_backlog.len += skb->truesize;
810         return 0;
811 }
812 
813 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
814 
815 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
816 {
817         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
818                 return __sk_backlog_rcv(sk, skb);
819 
820         return sk->sk_backlog_rcv(sk, skb);
821 }
822 
823 static inline void sock_rps_record_flow(const struct sock *sk)
824 {
825 #ifdef CONFIG_RPS
826         struct rps_sock_flow_table *sock_flow_table;
827 
828         rcu_read_lock();
829         sock_flow_table = rcu_dereference(rps_sock_flow_table);
830         rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
831         rcu_read_unlock();
832 #endif
833 }
834 
835 static inline void sock_rps_reset_flow(const struct sock *sk)
836 {
837 #ifdef CONFIG_RPS
838         struct rps_sock_flow_table *sock_flow_table;
839 
840         rcu_read_lock();
841         sock_flow_table = rcu_dereference(rps_sock_flow_table);
842         rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
843         rcu_read_unlock();
844 #endif
845 }
846 
847 static inline void sock_rps_save_rxhash(struct sock *sk,
848                                         const struct sk_buff *skb)
849 {
850 #ifdef CONFIG_RPS
851         if (unlikely(sk->sk_rxhash != skb->rxhash)) {
852                 sock_rps_reset_flow(sk);
853                 sk->sk_rxhash = skb->rxhash;
854         }
855 #endif
856 }
857 
858 static inline void sock_rps_reset_rxhash(struct sock *sk)
859 {
860 #ifdef CONFIG_RPS
861         sock_rps_reset_flow(sk);
862         sk->sk_rxhash = 0;
863 #endif
864 }
865 
866 #define sk_wait_event(__sk, __timeo, __condition)                       \
867         ({      int __rc;                                               \
868                 release_sock(__sk);                                     \
869                 __rc = __condition;                                     \
870                 if (!__rc) {                                            \
871                         *(__timeo) = schedule_timeout(*(__timeo));      \
872                 }                                                       \
873                 lock_sock(__sk);                                        \
874                 __rc = __condition;                                     \
875                 __rc;                                                   \
876         })
877 
878 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
879 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
880 void sk_stream_wait_close(struct sock *sk, long timeo_p);
881 int sk_stream_error(struct sock *sk, int flags, int err);
882 void sk_stream_kill_queues(struct sock *sk);
883 void sk_set_memalloc(struct sock *sk);
884 void sk_clear_memalloc(struct sock *sk);
885 
886 int sk_wait_data(struct sock *sk, long *timeo);
887 
888 struct request_sock_ops;
889 struct timewait_sock_ops;
890 struct inet_hashinfo;
891 struct raw_hashinfo;
892 struct module;
893 
894 /*
895  * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
896  * un-modified. Special care is taken when initializing object to zero.
897  */
898 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
899 {
900         if (offsetof(struct sock, sk_node.next) != 0)
901                 memset(sk, 0, offsetof(struct sock, sk_node.next));
902         memset(&sk->sk_node.pprev, 0,
903                size - offsetof(struct sock, sk_node.pprev));
904 }
905 
906 /* Networking protocol blocks we attach to sockets.
907  * socket layer -> transport layer interface
908  * transport -> network interface is defined by struct inet_proto
909  */
910 struct proto {
911         void                    (*close)(struct sock *sk,
912                                         long timeout);
913         int                     (*connect)(struct sock *sk,
914                                         struct sockaddr *uaddr,
915                                         int addr_len);
916         int                     (*disconnect)(struct sock *sk, int flags);
917 
918         struct sock *           (*accept)(struct sock *sk, int flags, int *err);
919 
920         int                     (*ioctl)(struct sock *sk, int cmd,
921                                          unsigned long arg);
922         int                     (*init)(struct sock *sk);
923         void                    (*destroy)(struct sock *sk);
924         void                    (*shutdown)(struct sock *sk, int how);
925         int                     (*setsockopt)(struct sock *sk, int level,
926                                         int optname, char __user *optval,
927                                         unsigned int optlen);
928         int                     (*getsockopt)(struct sock *sk, int level,
929                                         int optname, char __user *optval,
930                                         int __user *option);
931 #ifdef CONFIG_COMPAT
932         int                     (*compat_setsockopt)(struct sock *sk,
933                                         int level,
934                                         int optname, char __user *optval,
935                                         unsigned int optlen);
936         int                     (*compat_getsockopt)(struct sock *sk,
937                                         int level,
938                                         int optname, char __user *optval,
939                                         int __user *option);
940         int                     (*compat_ioctl)(struct sock *sk,
941                                         unsigned int cmd, unsigned long arg);
942 #endif
943         int                     (*sendmsg)(struct kiocb *iocb, struct sock *sk,
944                                            struct msghdr *msg, size_t len);
945         int                     (*recvmsg)(struct kiocb *iocb, struct sock *sk,
946                                            struct msghdr *msg,
947                                            size_t len, int noblock, int flags,
948                                            int *addr_len);
949         int                     (*sendpage)(struct sock *sk, struct page *page,
950                                         int offset, size_t size, int flags);
951         int                     (*bind)(struct sock *sk,
952                                         struct sockaddr *uaddr, int addr_len);
953 
954         int                     (*backlog_rcv) (struct sock *sk,
955                                                 struct sk_buff *skb);
956 
957         void            (*release_cb)(struct sock *sk);
958         void            (*mtu_reduced)(struct sock *sk);
959 
960         /* Keeping track of sk's, looking them up, and port selection methods. */
961         void                    (*hash)(struct sock *sk);
962         void                    (*unhash)(struct sock *sk);
963         void                    (*rehash)(struct sock *sk);
964         int                     (*get_port)(struct sock *sk, unsigned short snum);
965         void                    (*clear_sk)(struct sock *sk, int size);
966 
967         /* Keeping track of sockets in use */
968 #ifdef CONFIG_PROC_FS
969         unsigned int            inuse_idx;
970 #endif
971 
972         bool                    (*stream_memory_free)(const struct sock *sk);
973         /* Memory pressure */
974         void                    (*enter_memory_pressure)(struct sock *sk);
975         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
976         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
977         /*
978          * Pressure flag: try to collapse.
979          * Technical note: it is used by multiple contexts non atomically.
980          * All the __sk_mem_schedule() is of this nature: accounting
981          * is strict, actions are advisory and have some latency.
982          */
983         int                     *memory_pressure;
984         long                    *sysctl_mem;
985         int                     *sysctl_wmem;
986         int                     *sysctl_rmem;
987         int                     max_header;
988         bool                    no_autobind;
989 
990         struct kmem_cache       *slab;
991         unsigned int            obj_size;
992         int                     slab_flags;
993 
994         struct percpu_counter   *orphan_count;
995 
996         struct request_sock_ops *rsk_prot;
997         struct timewait_sock_ops *twsk_prot;
998 
999         union {
1000                 struct inet_hashinfo    *hashinfo;
1001                 struct udp_table        *udp_table;
1002                 struct raw_hashinfo     *raw_hash;
1003         } h;
1004 
1005         struct module           *owner;
1006 
1007         char                    name[32];
1008 
1009         struct list_head        node;
1010 #ifdef SOCK_REFCNT_DEBUG
1011         atomic_t                socks;
1012 #endif
1013 #ifdef CONFIG_MEMCG_KMEM
1014         /*
1015          * cgroup specific init/deinit functions. Called once for all
1016          * protocols that implement it, from cgroups populate function.
1017          * This function has to setup any files the protocol want to
1018          * appear in the kmem cgroup filesystem.
1019          */
1020         int                     (*init_cgroup)(struct mem_cgroup *memcg,
1021                                                struct cgroup_subsys *ss);
1022         void                    (*destroy_cgroup)(struct mem_cgroup *memcg);
1023         struct cg_proto         *(*proto_cgroup)(struct mem_cgroup *memcg);
1024 #endif
1025 };
1026 
1027 /*
1028  * Bits in struct cg_proto.flags
1029  */
1030 enum cg_proto_flags {
1031         /* Currently active and new sockets should be assigned to cgroups */
1032         MEMCG_SOCK_ACTIVE,
1033         /* It was ever activated; we must disarm static keys on destruction */
1034         MEMCG_SOCK_ACTIVATED,
1035 };
1036 
1037 struct cg_proto {
1038         struct res_counter      memory_allocated;       /* Current allocated memory. */
1039         struct percpu_counter   sockets_allocated;      /* Current number of sockets. */
1040         int                     memory_pressure;
1041         long                    sysctl_mem[3];
1042         unsigned long           flags;
1043         /*
1044          * memcg field is used to find which memcg we belong directly
1045          * Each memcg struct can hold more than one cg_proto, so container_of
1046          * won't really cut.
1047          *
1048          * The elegant solution would be having an inverse function to
1049          * proto_cgroup in struct proto, but that means polluting the structure
1050          * for everybody, instead of just for memcg users.
1051          */
1052         struct mem_cgroup       *memcg;
1053 };
1054 
1055 int proto_register(struct proto *prot, int alloc_slab);
1056 void proto_unregister(struct proto *prot);
1057 
1058 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1059 {
1060         return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1061 }
1062 
1063 static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1064 {
1065         return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1066 }
1067 
1068 #ifdef SOCK_REFCNT_DEBUG
1069 static inline void sk_refcnt_debug_inc(struct sock *sk)
1070 {
1071         atomic_inc(&sk->sk_prot->socks);
1072 }
1073 
1074 static inline void sk_refcnt_debug_dec(struct sock *sk)
1075 {
1076         atomic_dec(&sk->sk_prot->socks);
1077         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1078                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1079 }
1080 
1081 static inline void sk_refcnt_debug_release(const struct sock *sk)
1082 {
1083         if (atomic_read(&sk->sk_refcnt) != 1)
1084                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1085                        sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1086 }
1087 #else /* SOCK_REFCNT_DEBUG */
1088 #define sk_refcnt_debug_inc(sk) do { } while (0)
1089 #define sk_refcnt_debug_dec(sk) do { } while (0)
1090 #define sk_refcnt_debug_release(sk) do { } while (0)
1091 #endif /* SOCK_REFCNT_DEBUG */
1092 
1093 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1094 extern struct static_key memcg_socket_limit_enabled;
1095 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1096                                                struct cg_proto *cg_proto)
1097 {
1098         return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1099 }
1100 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1101 #else
1102 #define mem_cgroup_sockets_enabled 0
1103 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1104                                                struct cg_proto *cg_proto)
1105 {
1106         return NULL;
1107 }
1108 #endif
1109 
1110 static inline bool sk_stream_memory_free(const struct sock *sk)
1111 {
1112         if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1113                 return false;
1114 
1115         return sk->sk_prot->stream_memory_free ?
1116                 sk->sk_prot->stream_memory_free(sk) : true;
1117 }
1118 
1119 static inline bool sk_stream_is_writeable(const struct sock *sk)
1120 {
1121         return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1122                sk_stream_memory_free(sk);
1123 }
1124 
1125 
1126 static inline bool sk_has_memory_pressure(const struct sock *sk)
1127 {
1128         return sk->sk_prot->memory_pressure != NULL;
1129 }
1130 
1131 static inline bool sk_under_memory_pressure(const struct sock *sk)
1132 {
1133         if (!sk->sk_prot->memory_pressure)
1134                 return false;
1135 
1136         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1137                 return !!sk->sk_cgrp->memory_pressure;
1138 
1139         return !!*sk->sk_prot->memory_pressure;
1140 }
1141 
1142 static inline void sk_leave_memory_pressure(struct sock *sk)
1143 {
1144         int *memory_pressure = sk->sk_prot->memory_pressure;
1145 
1146         if (!memory_pressure)
1147                 return;
1148 
1149         if (*memory_pressure)
1150                 *memory_pressure = 0;
1151 
1152         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1153                 struct cg_proto *cg_proto = sk->sk_cgrp;
1154                 struct proto *prot = sk->sk_prot;
1155 
1156                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1157                         cg_proto->memory_pressure = 0;
1158         }
1159 
1160 }
1161 
1162 static inline void sk_enter_memory_pressure(struct sock *sk)
1163 {
1164         if (!sk->sk_prot->enter_memory_pressure)
1165                 return;
1166 
1167         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1168                 struct cg_proto *cg_proto = sk->sk_cgrp;
1169                 struct proto *prot = sk->sk_prot;
1170 
1171                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1172                         cg_proto->memory_pressure = 1;
1173         }
1174 
1175         sk->sk_prot->enter_memory_pressure(sk);
1176 }
1177 
1178 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1179 {
1180         long *prot = sk->sk_prot->sysctl_mem;
1181         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1182                 prot = sk->sk_cgrp->sysctl_mem;
1183         return prot[index];
1184 }
1185 
1186 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1187                                               unsigned long amt,
1188                                               int *parent_status)
1189 {
1190         struct res_counter *fail;
1191         int ret;
1192 
1193         ret = res_counter_charge_nofail(&prot->memory_allocated,
1194                                         amt << PAGE_SHIFT, &fail);
1195         if (ret < 0)
1196                 *parent_status = OVER_LIMIT;
1197 }
1198 
1199 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1200                                               unsigned long amt)
1201 {
1202         res_counter_uncharge(&prot->memory_allocated, amt << PAGE_SHIFT);
1203 }
1204 
1205 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1206 {
1207         u64 ret;
1208         ret = res_counter_read_u64(&prot->memory_allocated, RES_USAGE);
1209         return ret >> PAGE_SHIFT;
1210 }
1211 
1212 static inline long
1213 sk_memory_allocated(const struct sock *sk)
1214 {
1215         struct proto *prot = sk->sk_prot;
1216         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1217                 return memcg_memory_allocated_read(sk->sk_cgrp);
1218 
1219         return atomic_long_read(prot->memory_allocated);
1220 }
1221 
1222 static inline long
1223 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1224 {
1225         struct proto *prot = sk->sk_prot;
1226 
1227         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1228                 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1229                 /* update the root cgroup regardless */
1230                 atomic_long_add_return(amt, prot->memory_allocated);
1231                 return memcg_memory_allocated_read(sk->sk_cgrp);
1232         }
1233 
1234         return atomic_long_add_return(amt, prot->memory_allocated);
1235 }
1236 
1237 static inline void
1238 sk_memory_allocated_sub(struct sock *sk, int amt)
1239 {
1240         struct proto *prot = sk->sk_prot;
1241 
1242         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1243                 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1244 
1245         atomic_long_sub(amt, prot->memory_allocated);
1246 }
1247 
1248 static inline void sk_sockets_allocated_dec(struct sock *sk)
1249 {
1250         struct proto *prot = sk->sk_prot;
1251 
1252         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1253                 struct cg_proto *cg_proto = sk->sk_cgrp;
1254 
1255                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1256                         percpu_counter_dec(&cg_proto->sockets_allocated);
1257         }
1258 
1259         percpu_counter_dec(prot->sockets_allocated);
1260 }
1261 
1262 static inline void sk_sockets_allocated_inc(struct sock *sk)
1263 {
1264         struct proto *prot = sk->sk_prot;
1265 
1266         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1267                 struct cg_proto *cg_proto = sk->sk_cgrp;
1268 
1269                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1270                         percpu_counter_inc(&cg_proto->sockets_allocated);
1271         }
1272 
1273         percpu_counter_inc(prot->sockets_allocated);
1274 }
1275 
1276 static inline int
1277 sk_sockets_allocated_read_positive(struct sock *sk)
1278 {
1279         struct proto *prot = sk->sk_prot;
1280 
1281         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1282                 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
1283 
1284         return percpu_counter_read_positive(prot->sockets_allocated);
1285 }
1286 
1287 static inline int
1288 proto_sockets_allocated_sum_positive(struct proto *prot)
1289 {
1290         return percpu_counter_sum_positive(prot->sockets_allocated);
1291 }
1292 
1293 static inline long
1294 proto_memory_allocated(struct proto *prot)
1295 {
1296         return atomic_long_read(prot->memory_allocated);
1297 }
1298 
1299 static inline bool
1300 proto_memory_pressure(struct proto *prot)
1301 {
1302         if (!prot->memory_pressure)
1303                 return false;
1304         return !!*prot->memory_pressure;
1305 }
1306 
1307 
1308 #ifdef CONFIG_PROC_FS
1309 /* Called with local bh disabled */
1310 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1311 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1312 #else
1313 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1314                 int inc)
1315 {
1316 }
1317 #endif
1318 
1319 
1320 /* With per-bucket locks this operation is not-atomic, so that
1321  * this version is not worse.
1322  */
1323 static inline void __sk_prot_rehash(struct sock *sk)
1324 {
1325         sk->sk_prot->unhash(sk);
1326         sk->sk_prot->hash(sk);
1327 }
1328 
1329 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1330 
1331 /* About 10 seconds */
1332 #define SOCK_DESTROY_TIME (10*HZ)
1333 
1334 /* Sockets 0-1023 can't be bound to unless you are superuser */
1335 #define PROT_SOCK       1024
1336 
1337 #define SHUTDOWN_MASK   3
1338 #define RCV_SHUTDOWN    1
1339 #define SEND_SHUTDOWN   2
1340 
1341 #define SOCK_SNDBUF_LOCK        1
1342 #define SOCK_RCVBUF_LOCK        2
1343 #define SOCK_BINDADDR_LOCK      4
1344 #define SOCK_BINDPORT_LOCK      8
1345 
1346 /* sock_iocb: used to kick off async processing of socket ios */
1347 struct sock_iocb {
1348         struct list_head        list;
1349 
1350         int                     flags;
1351         int                     size;
1352         struct socket           *sock;
1353         struct sock             *sk;
1354         struct scm_cookie       *scm;
1355         struct msghdr           *msg, async_msg;
1356         struct kiocb            *kiocb;
1357 };
1358 
1359 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1360 {
1361         return (struct sock_iocb *)iocb->private;
1362 }
1363 
1364 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1365 {
1366         return si->kiocb;
1367 }
1368 
1369 struct socket_alloc {
1370         struct socket socket;
1371         struct inode vfs_inode;
1372 };
1373 
1374 static inline struct socket *SOCKET_I(struct inode *inode)
1375 {
1376         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1377 }
1378 
1379 static inline struct inode *SOCK_INODE(struct socket *socket)
1380 {
1381         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1382 }
1383 
1384 /*
1385  * Functions for memory accounting
1386  */
1387 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1388 void __sk_mem_reclaim(struct sock *sk);
1389 
1390 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1391 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1392 #define SK_MEM_SEND     0
1393 #define SK_MEM_RECV     1
1394 
1395 static inline int sk_mem_pages(int amt)
1396 {
1397         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1398 }
1399 
1400 static inline bool sk_has_account(struct sock *sk)
1401 {
1402         /* return true if protocol supports memory accounting */
1403         return !!sk->sk_prot->memory_allocated;
1404 }
1405 
1406 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1407 {
1408         if (!sk_has_account(sk))
1409                 return true;
1410         return size <= sk->sk_forward_alloc ||
1411                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1412 }
1413 
1414 static inline bool
1415 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1416 {
1417         if (!sk_has_account(sk))
1418                 return true;
1419         return size<= sk->sk_forward_alloc ||
1420                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1421                 skb_pfmemalloc(skb);
1422 }
1423 
1424 static inline void sk_mem_reclaim(struct sock *sk)
1425 {
1426         if (!sk_has_account(sk))
1427                 return;
1428         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1429                 __sk_mem_reclaim(sk);
1430 }
1431 
1432 static inline void sk_mem_reclaim_partial(struct sock *sk)
1433 {
1434         if (!sk_has_account(sk))
1435                 return;
1436         if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1437                 __sk_mem_reclaim(sk);
1438 }
1439 
1440 static inline void sk_mem_charge(struct sock *sk, int size)
1441 {
1442         if (!sk_has_account(sk))
1443                 return;
1444         sk->sk_forward_alloc -= size;
1445 }
1446 
1447 static inline void sk_mem_uncharge(struct sock *sk, int size)
1448 {
1449         if (!sk_has_account(sk))
1450                 return;
1451         sk->sk_forward_alloc += size;
1452 }
1453 
1454 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1455 {
1456         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1457         sk->sk_wmem_queued -= skb->truesize;
1458         sk_mem_uncharge(sk, skb->truesize);
1459         __kfree_skb(skb);
1460 }
1461 
1462 /* Used by processes to "lock" a socket state, so that
1463  * interrupts and bottom half handlers won't change it
1464  * from under us. It essentially blocks any incoming
1465  * packets, so that we won't get any new data or any
1466  * packets that change the state of the socket.
1467  *
1468  * While locked, BH processing will add new packets to
1469  * the backlog queue.  This queue is processed by the
1470  * owner of the socket lock right before it is released.
1471  *
1472  * Since ~2.3.5 it is also exclusive sleep lock serializing
1473  * accesses from user process context.
1474  */
1475 #define sock_owned_by_user(sk)  ((sk)->sk_lock.owned)
1476 
1477 static inline void sock_release_ownership(struct sock *sk)
1478 {
1479         sk->sk_lock.owned = 0;
1480 }
1481 
1482 /*
1483  * Macro so as to not evaluate some arguments when
1484  * lockdep is not enabled.
1485  *
1486  * Mark both the sk_lock and the sk_lock.slock as a
1487  * per-address-family lock class.
1488  */
1489 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1490 do {                                                                    \
1491         sk->sk_lock.owned = 0;                                          \
1492         init_waitqueue_head(&sk->sk_lock.wq);                           \
1493         spin_lock_init(&(sk)->sk_lock.slock);                           \
1494         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1495                         sizeof((sk)->sk_lock));                         \
1496         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1497                                 (skey), (sname));                               \
1498         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1499 } while (0)
1500 
1501 void lock_sock_nested(struct sock *sk, int subclass);
1502 
1503 static inline void lock_sock(struct sock *sk)
1504 {
1505         lock_sock_nested(sk, 0);
1506 }
1507 
1508 void release_sock(struct sock *sk);
1509 
1510 /* BH context may only use the following locking interface. */
1511 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1512 #define bh_lock_sock_nested(__sk) \
1513                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1514                                 SINGLE_DEPTH_NESTING)
1515 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1516 
1517 bool lock_sock_fast(struct sock *sk);
1518 /**
1519  * unlock_sock_fast - complement of lock_sock_fast
1520  * @sk: socket
1521  * @slow: slow mode
1522  *
1523  * fast unlock socket for user context.
1524  * If slow mode is on, we call regular release_sock()
1525  */
1526 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1527 {
1528         if (slow)
1529                 release_sock(sk);
1530         else
1531                 spin_unlock_bh(&sk->sk_lock.slock);
1532 }
1533 
1534 
1535 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1536                       struct proto *prot);
1537 void sk_free(struct sock *sk);
1538 void sk_release_kernel(struct sock *sk);
1539 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1540 
1541 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1542                              gfp_t priority);
1543 struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
1544                              gfp_t priority);
1545 void sock_wfree(struct sk_buff *skb);
1546 void skb_orphan_partial(struct sk_buff *skb);
1547 void sock_rfree(struct sk_buff *skb);
1548 void sock_edemux(struct sk_buff *skb);
1549 
1550 int sock_setsockopt(struct socket *sock, int level, int op,
1551                     char __user *optval, unsigned int optlen);
1552 
1553 int sock_getsockopt(struct socket *sock, int level, int op,
1554                     char __user *optval, int __user *optlen);
1555 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1556                                     int noblock, int *errcode);
1557 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1558                                      unsigned long data_len, int noblock,
1559                                      int *errcode, int max_page_order);
1560 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1561 void sock_kfree_s(struct sock *sk, void *mem, int size);
1562 void sk_send_sigurg(struct sock *sk);
1563 
1564 /*
1565  * Functions to fill in entries in struct proto_ops when a protocol
1566  * does not implement a particular function.
1567  */
1568 int sock_no_bind(struct socket *, struct sockaddr *, int);
1569 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1570 int sock_no_socketpair(struct socket *, struct socket *);
1571 int sock_no_accept(struct socket *, struct socket *, int);
1572 int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1573 unsigned int sock_no_poll(struct file *, struct socket *,
1574                           struct poll_table_struct *);
1575 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1576 int sock_no_listen(struct socket *, int);
1577 int sock_no_shutdown(struct socket *, int);
1578 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1579 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1580 int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1581 int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1582                     int);
1583 int sock_no_mmap(struct file *file, struct socket *sock,
1584                  struct vm_area_struct *vma);
1585 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1586                          size_t size, int flags);
1587 
1588 /*
1589  * Functions to fill in entries in struct proto_ops when a protocol
1590  * uses the inet style.
1591  */
1592 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1593                                   char __user *optval, int __user *optlen);
1594 int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1595                                struct msghdr *msg, size_t size, int flags);
1596 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1597                                   char __user *optval, unsigned int optlen);
1598 int compat_sock_common_getsockopt(struct socket *sock, int level,
1599                 int optname, char __user *optval, int __user *optlen);
1600 int compat_sock_common_setsockopt(struct socket *sock, int level,
1601                 int optname, char __user *optval, unsigned int optlen);
1602 
1603 void sk_common_release(struct sock *sk);
1604 
1605 /*
1606  *      Default socket callbacks and setup code
1607  */
1608 
1609 /* Initialise core socket variables */
1610 void sock_init_data(struct socket *sock, struct sock *sk);
1611 
1612 void sk_filter_release_rcu(struct rcu_head *rcu);
1613 
1614 /**
1615  *      sk_filter_release - release a socket filter
1616  *      @fp: filter to remove
1617  *
1618  *      Remove a filter from a socket and release its resources.
1619  */
1620 
1621 static inline void sk_filter_release(struct sk_filter *fp)
1622 {
1623         if (atomic_dec_and_test(&fp->refcnt))
1624                 call_rcu(&fp->rcu, sk_filter_release_rcu);
1625 }
1626 
1627 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1628 {
1629         atomic_sub(sk_filter_size(fp->len), &sk->sk_omem_alloc);
1630         sk_filter_release(fp);
1631 }
1632 
1633 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1634 {
1635         atomic_inc(&fp->refcnt);
1636         atomic_add(sk_filter_size(fp->len), &sk->sk_omem_alloc);
1637 }
1638 
1639 /*
1640  * Socket reference counting postulates.
1641  *
1642  * * Each user of socket SHOULD hold a reference count.
1643  * * Each access point to socket (an hash table bucket, reference from a list,
1644  *   running timer, skb in flight MUST hold a reference count.
1645  * * When reference count hits 0, it means it will never increase back.
1646  * * When reference count hits 0, it means that no references from
1647  *   outside exist to this socket and current process on current CPU
1648  *   is last user and may/should destroy this socket.
1649  * * sk_free is called from any context: process, BH, IRQ. When
1650  *   it is called, socket has no references from outside -> sk_free
1651  *   may release descendant resources allocated by the socket, but
1652  *   to the time when it is called, socket is NOT referenced by any
1653  *   hash tables, lists etc.
1654  * * Packets, delivered from outside (from network or from another process)
1655  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1656  *   when they sit in queue. Otherwise, packets will leak to hole, when
1657  *   socket is looked up by one cpu and unhasing is made by another CPU.
1658  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1659  *   (leak to backlog). Packet socket does all the processing inside
1660  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1661  *   use separate SMP lock, so that they are prone too.
1662  */
1663 
1664 /* Ungrab socket and destroy it, if it was the last reference. */
1665 static inline void sock_put(struct sock *sk)
1666 {
1667         if (atomic_dec_and_test(&sk->sk_refcnt))
1668                 sk_free(sk);
1669 }
1670 /* Generic version of sock_put(), dealing with all sockets
1671  * (TCP_TIMEWAIT, ESTABLISHED...)
1672  */
1673 void sock_gen_put(struct sock *sk);
1674 
1675 int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
1676 
1677 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1678 {
1679         sk->sk_tx_queue_mapping = tx_queue;
1680 }
1681 
1682 static inline void sk_tx_queue_clear(struct sock *sk)
1683 {
1684         sk->sk_tx_queue_mapping = -1;
1685 }
1686 
1687 static inline int sk_tx_queue_get(const struct sock *sk)
1688 {
1689         return sk ? sk->sk_tx_queue_mapping : -1;
1690 }
1691 
1692 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1693 {
1694         sk_tx_queue_clear(sk);
1695         sk->sk_socket = sock;
1696 }
1697 
1698 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1699 {
1700         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1701         return &rcu_dereference_raw(sk->sk_wq)->wait;
1702 }
1703 /* Detach socket from process context.
1704  * Announce socket dead, detach it from wait queue and inode.
1705  * Note that parent inode held reference count on this struct sock,
1706  * we do not release it in this function, because protocol
1707  * probably wants some additional cleanups or even continuing
1708  * to work with this socket (TCP).
1709  */
1710 static inline void sock_orphan(struct sock *sk)
1711 {
1712         write_lock_bh(&sk->sk_callback_lock);
1713         sock_set_flag(sk, SOCK_DEAD);
1714         sk_set_socket(sk, NULL);
1715         sk->sk_wq  = NULL;
1716         write_unlock_bh(&sk->sk_callback_lock);
1717 }
1718 
1719 static inline void sock_graft(struct sock *sk, struct socket *parent)
1720 {
1721         write_lock_bh(&sk->sk_callback_lock);
1722         sk->sk_wq = parent->wq;
1723         parent->sk = sk;
1724         sk_set_socket(sk, parent);
1725         security_sock_graft(sk, parent);
1726         write_unlock_bh(&sk->sk_callback_lock);
1727 }
1728 
1729 kuid_t sock_i_uid(struct sock *sk);
1730 unsigned long sock_i_ino(struct sock *sk);
1731 
1732 static inline struct dst_entry *
1733 __sk_dst_get(struct sock *sk)
1734 {
1735         return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1736                                                        lockdep_is_held(&sk->sk_lock.slock));
1737 }
1738 
1739 static inline struct dst_entry *
1740 sk_dst_get(struct sock *sk)
1741 {
1742         struct dst_entry *dst;
1743 
1744         rcu_read_lock();
1745         dst = rcu_dereference(sk->sk_dst_cache);
1746         if (dst)
1747                 dst_hold(dst);
1748         rcu_read_unlock();
1749         return dst;
1750 }
1751 
1752 static inline void dst_negative_advice(struct sock *sk)
1753 {
1754         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1755 
1756         if (dst && dst->ops->negative_advice) {
1757                 ndst = dst->ops->negative_advice(dst);
1758 
1759                 if (ndst != dst) {
1760                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1761                         sk_tx_queue_clear(sk);
1762                 }
1763         }
1764 }
1765 
1766 static inline void
1767 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1768 {
1769         struct dst_entry *old_dst;
1770 
1771         sk_tx_queue_clear(sk);
1772         /*
1773          * This can be called while sk is owned by the caller only,
1774          * with no state that can be checked in a rcu_dereference_check() cond
1775          */
1776         old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1777         rcu_assign_pointer(sk->sk_dst_cache, dst);
1778         dst_release(old_dst);
1779 }
1780 
1781 static inline void
1782 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1783 {
1784         spin_lock(&sk->sk_dst_lock);
1785         __sk_dst_set(sk, dst);
1786         spin_unlock(&sk->sk_dst_lock);
1787 }
1788 
1789 static inline void
1790 __sk_dst_reset(struct sock *sk)
1791 {
1792         __sk_dst_set(sk, NULL);
1793 }
1794 
1795 static inline void
1796 sk_dst_reset(struct sock *sk)
1797 {
1798         spin_lock(&sk->sk_dst_lock);
1799         __sk_dst_reset(sk);
1800         spin_unlock(&sk->sk_dst_lock);
1801 }
1802 
1803 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1804 
1805 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1806 
1807 static inline bool sk_can_gso(const struct sock *sk)
1808 {
1809         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1810 }
1811 
1812 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1813 
1814 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1815 {
1816         sk->sk_route_nocaps |= flags;
1817         sk->sk_route_caps &= ~flags;
1818 }
1819 
1820 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1821                                            char __user *from, char *to,
1822                                            int copy, int offset)
1823 {
1824         if (skb->ip_summed == CHECKSUM_NONE) {
1825                 int err = 0;
1826                 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1827                 if (err)
1828                         return err;
1829                 skb->csum = csum_block_add(skb->csum, csum, offset);
1830         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1831                 if (!access_ok(VERIFY_READ, from, copy) ||
1832                     __copy_from_user_nocache(to, from, copy))
1833                         return -EFAULT;
1834         } else if (copy_from_user(to, from, copy))
1835                 return -EFAULT;
1836 
1837         return 0;
1838 }
1839 
1840 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1841                                        char __user *from, int copy)
1842 {
1843         int err, offset = skb->len;
1844 
1845         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1846                                        copy, offset);
1847         if (err)
1848                 __skb_trim(skb, offset);
1849 
1850         return err;
1851 }
1852 
1853 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1854                                            struct sk_buff *skb,
1855                                            struct page *page,
1856                                            int off, int copy)
1857 {
1858         int err;
1859 
1860         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1861                                        copy, skb->len);
1862         if (err)
1863                 return err;
1864 
1865         skb->len             += copy;
1866         skb->data_len        += copy;
1867         skb->truesize        += copy;
1868         sk->sk_wmem_queued   += copy;
1869         sk_mem_charge(sk, copy);
1870         return 0;
1871 }
1872 
1873 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1874                                    struct sk_buff *skb, struct page *page,
1875                                    int off, int copy)
1876 {
1877         if (skb->ip_summed == CHECKSUM_NONE) {
1878                 int err = 0;
1879                 __wsum csum = csum_and_copy_from_user(from,
1880                                                      page_address(page) + off,
1881                                                             copy, 0, &err);
1882                 if (err)
1883                         return err;
1884                 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1885         } else if (copy_from_user(page_address(page) + off, from, copy))
1886                 return -EFAULT;
1887 
1888         skb->len             += copy;
1889         skb->data_len        += copy;
1890         skb->truesize        += copy;
1891         sk->sk_wmem_queued   += copy;
1892         sk_mem_charge(sk, copy);
1893         return 0;
1894 }
1895 
1896 /**
1897  * sk_wmem_alloc_get - returns write allocations
1898  * @sk: socket
1899  *
1900  * Returns sk_wmem_alloc minus initial offset of one
1901  */
1902 static inline int sk_wmem_alloc_get(const struct sock *sk)
1903 {
1904         return atomic_read(&sk->sk_wmem_alloc) - 1;
1905 }
1906 
1907 /**
1908  * sk_rmem_alloc_get - returns read allocations
1909  * @sk: socket
1910  *
1911  * Returns sk_rmem_alloc
1912  */
1913 static inline int sk_rmem_alloc_get(const struct sock *sk)
1914 {
1915         return atomic_read(&sk->sk_rmem_alloc);
1916 }
1917 
1918 /**
1919  * sk_has_allocations - check if allocations are outstanding
1920  * @sk: socket
1921  *
1922  * Returns true if socket has write or read allocations
1923  */
1924 static inline bool sk_has_allocations(const struct sock *sk)
1925 {
1926         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1927 }
1928 
1929 /**
1930  * wq_has_sleeper - check if there are any waiting processes
1931  * @wq: struct socket_wq
1932  *
1933  * Returns true if socket_wq has waiting processes
1934  *
1935  * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1936  * barrier call. They were added due to the race found within the tcp code.
1937  *
1938  * Consider following tcp code paths:
1939  *
1940  * CPU1                  CPU2
1941  *
1942  * sys_select            receive packet
1943  *   ...                 ...
1944  *   __add_wait_queue    update tp->rcv_nxt
1945  *   ...                 ...
1946  *   tp->rcv_nxt check   sock_def_readable
1947  *   ...                 {
1948  *   schedule               rcu_read_lock();
1949  *                          wq = rcu_dereference(sk->sk_wq);
1950  *                          if (wq && waitqueue_active(&wq->wait))
1951  *                              wake_up_interruptible(&wq->wait)
1952  *                          ...
1953  *                       }
1954  *
1955  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1956  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1957  * could then endup calling schedule and sleep forever if there are no more
1958  * data on the socket.
1959  *
1960  */
1961 static inline bool wq_has_sleeper(struct socket_wq *wq)
1962 {
1963         /* We need to be sure we are in sync with the
1964          * add_wait_queue modifications to the wait queue.
1965          *
1966          * This memory barrier is paired in the sock_poll_wait.
1967          */
1968         smp_mb();
1969         return wq && waitqueue_active(&wq->wait);
1970 }
1971 
1972 /**
1973  * sock_poll_wait - place memory barrier behind the poll_wait call.
1974  * @filp:           file
1975  * @wait_address:   socket wait queue
1976  * @p:              poll_table
1977  *
1978  * See the comments in the wq_has_sleeper function.
1979  */
1980 static inline void sock_poll_wait(struct file *filp,
1981                 wait_queue_head_t *wait_address, poll_table *p)
1982 {
1983         if (!poll_does_not_wait(p) && wait_address) {
1984                 poll_wait(filp, wait_address, p);
1985                 /* We need to be sure we are in sync with the
1986                  * socket flags modification.
1987                  *
1988                  * This memory barrier is paired in the wq_has_sleeper.
1989                  */
1990                 smp_mb();
1991         }
1992 }
1993 
1994 /*
1995  *      Queue a received datagram if it will fit. Stream and sequenced
1996  *      protocols can't normally use this as they need to fit buffers in
1997  *      and play with them.
1998  *
1999  *      Inlined as it's very short and called for pretty much every
2000  *      packet ever received.
2001  */
2002 
2003 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2004 {
2005         skb_orphan(skb);
2006         skb->sk = sk;
2007         skb->destructor = sock_wfree;
2008         /*
2009          * We used to take a refcount on sk, but following operation
2010          * is enough to guarantee sk_free() wont free this sock until
2011          * all in-flight packets are completed
2012          */
2013         atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2014 }
2015 
2016 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2017 {
2018         skb_orphan(skb);
2019         skb->sk = sk;
2020         skb->destructor = sock_rfree;
2021         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2022         sk_mem_charge(sk, skb->truesize);
2023 }
2024 
2025 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2026                     unsigned long expires);
2027 
2028 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2029 
2030 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2031 
2032 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2033 
2034 /*
2035  *      Recover an error report and clear atomically
2036  */
2037 
2038 static inline int sock_error(struct sock *sk)
2039 {
2040         int err;
2041         if (likely(!sk->sk_err))
2042                 return 0;
2043         err = xchg(&sk->sk_err, 0);
2044         return -err;
2045 }
2046 
2047 static inline unsigned long sock_wspace(struct sock *sk)
2048 {
2049         int amt = 0;
2050 
2051         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2052                 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2053                 if (amt < 0)
2054                         amt = 0;
2055         }
2056         return amt;
2057 }
2058 
2059 static inline void sk_wake_async(struct sock *sk, int how, int band)
2060 {
2061         if (sock_flag(sk, SOCK_FASYNC))
2062                 sock_wake_async(sk->sk_socket, how, band);
2063 }
2064 
2065 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2066  * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2067  * Note: for send buffers, TCP works better if we can build two skbs at
2068  * minimum.
2069  */
2070 #define TCP_SKB_MIN_TRUESIZE    (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2071 
2072 #define SOCK_MIN_SNDBUF         (TCP_SKB_MIN_TRUESIZE * 2)
2073 #define SOCK_MIN_RCVBUF          TCP_SKB_MIN_TRUESIZE
2074 
2075 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2076 {
2077         if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2078                 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2079                 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2080         }
2081 }
2082 
2083 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2084 
2085 /**
2086  * sk_page_frag - return an appropriate page_frag
2087  * @sk: socket
2088  *
2089  * If socket allocation mode allows current thread to sleep, it means its
2090  * safe to use the per task page_frag instead of the per socket one.
2091  */
2092 static inline struct page_frag *sk_page_frag(struct sock *sk)
2093 {
2094         if (sk->sk_allocation & __GFP_WAIT)
2095                 return &current->task_frag;
2096 
2097         return &sk->sk_frag;
2098 }
2099 
2100 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2101 
2102 /*
2103  *      Default write policy as shown to user space via poll/select/SIGIO
2104  */
2105 static inline bool sock_writeable(const struct sock *sk)
2106 {
2107         return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2108 }
2109 
2110 static inline gfp_t gfp_any(void)
2111 {
2112         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2113 }
2114 
2115 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2116 {
2117         return noblock ? 0 : sk->sk_rcvtimeo;
2118 }
2119 
2120 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2121 {
2122         return noblock ? 0 : sk->sk_sndtimeo;
2123 }
2124 
2125 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2126 {
2127         return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2128 }
2129 
2130 /* Alas, with timeout socket operations are not restartable.
2131  * Compare this to poll().
2132  */
2133 static inline int sock_intr_errno(long timeo)
2134 {
2135         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2136 }
2137 
2138 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2139                            struct sk_buff *skb);
2140 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2141                              struct sk_buff *skb);
2142 
2143 static inline void
2144 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2145 {
2146         ktime_t kt = skb->tstamp;
2147         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2148 
2149         /*
2150          * generate control messages if
2151          * - receive time stamping in software requested (SOCK_RCVTSTAMP
2152          *   or SOCK_TIMESTAMPING_RX_SOFTWARE)
2153          * - software time stamp available and wanted
2154          *   (SOCK_TIMESTAMPING_SOFTWARE)
2155          * - hardware time stamps available and wanted
2156          *   (SOCK_TIMESTAMPING_SYS_HARDWARE or
2157          *   SOCK_TIMESTAMPING_RAW_HARDWARE)
2158          */
2159         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2160             sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2161             (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2162             (hwtstamps->hwtstamp.tv64 &&
2163              sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2164             (hwtstamps->syststamp.tv64 &&
2165              sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2166                 __sock_recv_timestamp(msg, sk, skb);
2167         else
2168                 sk->sk_stamp = kt;
2169 
2170         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2171                 __sock_recv_wifi_status(msg, sk, skb);
2172 }
2173 
2174 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2175                               struct sk_buff *skb);
2176 
2177 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2178                                           struct sk_buff *skb)
2179 {
2180 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2181                            (1UL << SOCK_RCVTSTAMP)                      | \
2182                            (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)       | \
2183                            (1UL << SOCK_TIMESTAMPING_SOFTWARE)          | \
2184                            (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE)      | \
2185                            (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2186 
2187         if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2188                 __sock_recv_ts_and_drops(msg, sk, skb);
2189         else
2190                 sk->sk_stamp = skb->tstamp;
2191 }
2192 
2193 /**
2194  * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2195  * @sk:         socket sending this packet
2196  * @tx_flags:   filled with instructions for time stamping
2197  *
2198  * Currently only depends on SOCK_TIMESTAMPING* flags.
2199  */
2200 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2201 
2202 /**
2203  * sk_eat_skb - Release a skb if it is no longer needed
2204  * @sk: socket to eat this skb from
2205  * @skb: socket buffer to eat
2206  * @copied_early: flag indicating whether DMA operations copied this data early
2207  *
2208  * This routine must be called with interrupts disabled or with the socket
2209  * locked so that the sk_buff queue operation is ok.
2210 */
2211 #ifdef CONFIG_NET_DMA
2212 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2213 {
2214         __skb_unlink(skb, &sk->sk_receive_queue);
2215         if (!copied_early)
2216                 __kfree_skb(skb);
2217         else
2218                 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2219 }
2220 #else
2221 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2222 {
2223         __skb_unlink(skb, &sk->sk_receive_queue);
2224         __kfree_skb(skb);
2225 }
2226 #endif
2227 
2228 static inline
2229 struct net *sock_net(const struct sock *sk)
2230 {
2231         return read_pnet(&sk->sk_net);
2232 }
2233 
2234 static inline
2235 void sock_net_set(struct sock *sk, struct net *net)
2236 {
2237         write_pnet(&sk->sk_net, net);
2238 }
2239 
2240 /*
2241  * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2242  * They should not hold a reference to a namespace in order to allow
2243  * to stop it.
2244  * Sockets after sk_change_net should be released using sk_release_kernel
2245  */
2246 static inline void sk_change_net(struct sock *sk, struct net *net)
2247 {
2248         put_net(sock_net(sk));
2249         sock_net_set(sk, hold_net(net));
2250 }
2251 
2252 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2253 {
2254         if (skb->sk) {
2255                 struct sock *sk = skb->sk;
2256 
2257                 skb->destructor = NULL;
2258                 skb->sk = NULL;
2259                 return sk;
2260         }
2261         return NULL;
2262 }
2263 
2264 void sock_enable_timestamp(struct sock *sk, int flag);
2265 int sock_get_timestamp(struct sock *, struct timeval __user *);
2266 int sock_get_timestampns(struct sock *, struct timespec __user *);
2267 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2268                        int type);
2269 
2270 /*
2271  *      Enable debug/info messages
2272  */
2273 extern int net_msg_warn;
2274 #define NETDEBUG(fmt, args...) \
2275         do { if (net_msg_warn) printk(fmt,##args); } while (0)
2276 
2277 #define LIMIT_NETDEBUG(fmt, args...) \
2278         do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2279 
2280 extern __u32 sysctl_wmem_max;
2281 extern __u32 sysctl_rmem_max;
2282 
2283 extern int sysctl_optmem_max;
2284 
2285 extern __u32 sysctl_wmem_default;
2286 extern __u32 sysctl_rmem_default;
2287 
2288 #endif  /* _SOCK_H */
2289 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp