Version:  2.0.40 2.2.26 2.4.37 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10

Linux/include/net/sock.h

  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/page_counter.h>
 58 #include <linux/memcontrol.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 #include <linux/net_tstamp.h>
 71 
 72 struct cgroup;
 73 struct cgroup_subsys;
 74 #ifdef CONFIG_NET
 75 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
 76 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
 77 #else
 78 static inline
 79 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
 80 {
 81         return 0;
 82 }
 83 static inline
 84 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
 85 {
 86 }
 87 #endif
 88 /*
 89  * This structure really needs to be cleaned up.
 90  * Most of it is for TCP, and not used by any of
 91  * the other protocols.
 92  */
 93 
 94 /* Define this to get the SOCK_DBG debugging facility. */
 95 #define SOCK_DEBUGGING
 96 #ifdef SOCK_DEBUGGING
 97 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
 98                                         printk(KERN_DEBUG msg); } while (0)
 99 #else
100 /* Validate arguments and do nothing */
101 static inline __printf(2, 3)
102 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
103 {
104 }
105 #endif
106 
107 /* This is the per-socket lock.  The spinlock provides a synchronization
108  * between user contexts and software interrupt processing, whereas the
109  * mini-semaphore synchronizes multiple users amongst themselves.
110  */
111 typedef struct {
112         spinlock_t              slock;
113         int                     owned;
114         wait_queue_head_t       wq;
115         /*
116          * We express the mutex-alike socket_lock semantics
117          * to the lock validator by explicitly managing
118          * the slock as a lock variant (in addition to
119          * the slock itself):
120          */
121 #ifdef CONFIG_DEBUG_LOCK_ALLOC
122         struct lockdep_map dep_map;
123 #endif
124 } socket_lock_t;
125 
126 struct sock;
127 struct proto;
128 struct net;
129 
130 typedef __u32 __bitwise __portpair;
131 typedef __u64 __bitwise __addrpair;
132 
133 /**
134  *      struct sock_common - minimal network layer representation of sockets
135  *      @skc_daddr: Foreign IPv4 addr
136  *      @skc_rcv_saddr: Bound local IPv4 addr
137  *      @skc_hash: hash value used with various protocol lookup tables
138  *      @skc_u16hashes: two u16 hash values used by UDP lookup tables
139  *      @skc_dport: placeholder for inet_dport/tw_dport
140  *      @skc_num: placeholder for inet_num/tw_num
141  *      @skc_family: network address family
142  *      @skc_state: Connection state
143  *      @skc_reuse: %SO_REUSEADDR setting
144  *      @skc_reuseport: %SO_REUSEPORT setting
145  *      @skc_bound_dev_if: bound device index if != 0
146  *      @skc_bind_node: bind hash linkage for various protocol lookup tables
147  *      @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
148  *      @skc_prot: protocol handlers inside a network family
149  *      @skc_net: reference to the network namespace of this socket
150  *      @skc_node: main hash linkage for various protocol lookup tables
151  *      @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
152  *      @skc_tx_queue_mapping: tx queue number for this connection
153  *      @skc_refcnt: reference count
154  *
155  *      This is the minimal network layer representation of sockets, the header
156  *      for struct sock and struct inet_timewait_sock.
157  */
158 struct sock_common {
159         /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
160          * address on 64bit arches : cf INET_MATCH()
161          */
162         union {
163                 __addrpair      skc_addrpair;
164                 struct {
165                         __be32  skc_daddr;
166                         __be32  skc_rcv_saddr;
167                 };
168         };
169         union  {
170                 unsigned int    skc_hash;
171                 __u16           skc_u16hashes[2];
172         };
173         /* skc_dport && skc_num must be grouped as well */
174         union {
175                 __portpair      skc_portpair;
176                 struct {
177                         __be16  skc_dport;
178                         __u16   skc_num;
179                 };
180         };
181 
182         unsigned short          skc_family;
183         volatile unsigned char  skc_state;
184         unsigned char           skc_reuse:4;
185         unsigned char           skc_reuseport:1;
186         unsigned char           skc_ipv6only:1;
187         int                     skc_bound_dev_if;
188         union {
189                 struct hlist_node       skc_bind_node;
190                 struct hlist_nulls_node skc_portaddr_node;
191         };
192         struct proto            *skc_prot;
193 #ifdef CONFIG_NET_NS
194         struct net              *skc_net;
195 #endif
196 
197 #if IS_ENABLED(CONFIG_IPV6)
198         struct in6_addr         skc_v6_daddr;
199         struct in6_addr         skc_v6_rcv_saddr;
200 #endif
201 
202         /*
203          * fields between dontcopy_begin/dontcopy_end
204          * are not copied in sock_copy()
205          */
206         /* private: */
207         int                     skc_dontcopy_begin[0];
208         /* public: */
209         union {
210                 struct hlist_node       skc_node;
211                 struct hlist_nulls_node skc_nulls_node;
212         };
213         int                     skc_tx_queue_mapping;
214         atomic_t                skc_refcnt;
215         /* private: */
216         int                     skc_dontcopy_end[0];
217         /* public: */
218 };
219 
220 struct cg_proto;
221 /**
222   *     struct sock - network layer representation of sockets
223   *     @__sk_common: shared layout with inet_timewait_sock
224   *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
225   *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
226   *     @sk_lock:       synchronizer
227   *     @sk_rcvbuf: size of receive buffer in bytes
228   *     @sk_wq: sock wait queue and async head
229   *     @sk_rx_dst: receive input route used by early demux
230   *     @sk_dst_cache: destination cache
231   *     @sk_dst_lock: destination cache lock
232   *     @sk_policy: flow policy
233   *     @sk_receive_queue: incoming packets
234   *     @sk_wmem_alloc: transmit queue bytes committed
235   *     @sk_write_queue: Packet sending queue
236   *     @sk_omem_alloc: "o" is "option" or "other"
237   *     @sk_wmem_queued: persistent queue size
238   *     @sk_forward_alloc: space allocated forward
239   *     @sk_napi_id: id of the last napi context to receive data for sk
240   *     @sk_ll_usec: usecs to busypoll when there is no data
241   *     @sk_allocation: allocation mode
242   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
243   *     @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
244   *     @sk_sndbuf: size of send buffer in bytes
245   *     @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
246   *                %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
247   *     @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
248   *     @sk_no_check_rx: allow zero checksum in RX packets
249   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
250   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
251   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
252   *     @sk_gso_max_size: Maximum GSO segment size to build
253   *     @sk_gso_max_segs: Maximum number of GSO segments
254   *     @sk_lingertime: %SO_LINGER l_linger setting
255   *     @sk_backlog: always used with the per-socket spinlock held
256   *     @sk_callback_lock: used with the callbacks in the end of this struct
257   *     @sk_error_queue: rarely used
258   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
259   *                       IPV6_ADDRFORM for instance)
260   *     @sk_err: last error
261   *     @sk_err_soft: errors that don't cause failure but are the cause of a
262   *                   persistent failure not just 'timed out'
263   *     @sk_drops: raw/udp drops counter
264   *     @sk_ack_backlog: current listen backlog
265   *     @sk_max_ack_backlog: listen backlog set in listen()
266   *     @sk_priority: %SO_PRIORITY setting
267   *     @sk_cgrp_prioidx: socket group's priority map index
268   *     @sk_type: socket type (%SOCK_STREAM, etc)
269   *     @sk_protocol: which protocol this socket belongs in this network family
270   *     @sk_peer_pid: &struct pid for this socket's peer
271   *     @sk_peer_cred: %SO_PEERCRED setting
272   *     @sk_rcvlowat: %SO_RCVLOWAT setting
273   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
274   *     @sk_sndtimeo: %SO_SNDTIMEO setting
275   *     @sk_rxhash: flow hash received from netif layer
276   *     @sk_incoming_cpu: record cpu processing incoming packets
277   *     @sk_txhash: computed flow hash for use on transmit
278   *     @sk_filter: socket filtering instructions
279   *     @sk_protinfo: private area, net family specific, when not using slab
280   *     @sk_timer: sock cleanup timer
281   *     @sk_stamp: time stamp of last packet received
282   *     @sk_tsflags: SO_TIMESTAMPING socket options
283   *     @sk_tskey: counter to disambiguate concurrent tstamp requests
284   *     @sk_socket: Identd and reporting IO signals
285   *     @sk_user_data: RPC layer private data
286   *     @sk_frag: cached page frag
287   *     @sk_peek_off: current peek_offset value
288   *     @sk_send_head: front of stuff to transmit
289   *     @sk_security: used by security modules
290   *     @sk_mark: generic packet mark
291   *     @sk_classid: this socket's cgroup classid
292   *     @sk_cgrp: this socket's cgroup-specific proto data
293   *     @sk_write_pending: a write to stream socket waits to start
294   *     @sk_state_change: callback to indicate change in the state of the sock
295   *     @sk_data_ready: callback to indicate there is data to be processed
296   *     @sk_write_space: callback to indicate there is bf sending space available
297   *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
298   *     @sk_backlog_rcv: callback to process the backlog
299   *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
300  */
301 struct sock {
302         /*
303          * Now struct inet_timewait_sock also uses sock_common, so please just
304          * don't add nothing before this first member (__sk_common) --acme
305          */
306         struct sock_common      __sk_common;
307 #define sk_node                 __sk_common.skc_node
308 #define sk_nulls_node           __sk_common.skc_nulls_node
309 #define sk_refcnt               __sk_common.skc_refcnt
310 #define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
311 
312 #define sk_dontcopy_begin       __sk_common.skc_dontcopy_begin
313 #define sk_dontcopy_end         __sk_common.skc_dontcopy_end
314 #define sk_hash                 __sk_common.skc_hash
315 #define sk_portpair             __sk_common.skc_portpair
316 #define sk_num                  __sk_common.skc_num
317 #define sk_dport                __sk_common.skc_dport
318 #define sk_addrpair             __sk_common.skc_addrpair
319 #define sk_daddr                __sk_common.skc_daddr
320 #define sk_rcv_saddr            __sk_common.skc_rcv_saddr
321 #define sk_family               __sk_common.skc_family
322 #define sk_state                __sk_common.skc_state
323 #define sk_reuse                __sk_common.skc_reuse
324 #define sk_reuseport            __sk_common.skc_reuseport
325 #define sk_ipv6only             __sk_common.skc_ipv6only
326 #define sk_bound_dev_if         __sk_common.skc_bound_dev_if
327 #define sk_bind_node            __sk_common.skc_bind_node
328 #define sk_prot                 __sk_common.skc_prot
329 #define sk_net                  __sk_common.skc_net
330 #define sk_v6_daddr             __sk_common.skc_v6_daddr
331 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
332 
333         socket_lock_t           sk_lock;
334         struct sk_buff_head     sk_receive_queue;
335         /*
336          * The backlog queue is special, it is always used with
337          * the per-socket spinlock held and requires low latency
338          * access. Therefore we special case it's implementation.
339          * Note : rmem_alloc is in this structure to fill a hole
340          * on 64bit arches, not because its logically part of
341          * backlog.
342          */
343         struct {
344                 atomic_t        rmem_alloc;
345                 int             len;
346                 struct sk_buff  *head;
347                 struct sk_buff  *tail;
348         } sk_backlog;
349 #define sk_rmem_alloc sk_backlog.rmem_alloc
350         int                     sk_forward_alloc;
351 #ifdef CONFIG_RPS
352         __u32                   sk_rxhash;
353 #endif
354         u16                     sk_incoming_cpu;
355         /* 16bit hole
356          * Warned : sk_incoming_cpu can be set from softirq,
357          * Do not use this hole without fully understanding possible issues.
358          */
359 
360         __u32                   sk_txhash;
361 #ifdef CONFIG_NET_RX_BUSY_POLL
362         unsigned int            sk_napi_id;
363         unsigned int            sk_ll_usec;
364 #endif
365         atomic_t                sk_drops;
366         int                     sk_rcvbuf;
367 
368         struct sk_filter __rcu  *sk_filter;
369         struct socket_wq __rcu  *sk_wq;
370 
371 #ifdef CONFIG_XFRM
372         struct xfrm_policy      *sk_policy[2];
373 #endif
374         unsigned long           sk_flags;
375         struct dst_entry        *sk_rx_dst;
376         struct dst_entry __rcu  *sk_dst_cache;
377         spinlock_t              sk_dst_lock;
378         atomic_t                sk_wmem_alloc;
379         atomic_t                sk_omem_alloc;
380         int                     sk_sndbuf;
381         struct sk_buff_head     sk_write_queue;
382         kmemcheck_bitfield_begin(flags);
383         unsigned int            sk_shutdown  : 2,
384                                 sk_no_check_tx : 1,
385                                 sk_no_check_rx : 1,
386                                 sk_userlocks : 4,
387                                 sk_protocol  : 8,
388                                 sk_type      : 16;
389         kmemcheck_bitfield_end(flags);
390         int                     sk_wmem_queued;
391         gfp_t                   sk_allocation;
392         u32                     sk_pacing_rate; /* bytes per second */
393         u32                     sk_max_pacing_rate;
394         netdev_features_t       sk_route_caps;
395         netdev_features_t       sk_route_nocaps;
396         int                     sk_gso_type;
397         unsigned int            sk_gso_max_size;
398         u16                     sk_gso_max_segs;
399         int                     sk_rcvlowat;
400         unsigned long           sk_lingertime;
401         struct sk_buff_head     sk_error_queue;
402         struct proto            *sk_prot_creator;
403         rwlock_t                sk_callback_lock;
404         int                     sk_err,
405                                 sk_err_soft;
406         unsigned short          sk_ack_backlog;
407         unsigned short          sk_max_ack_backlog;
408         __u32                   sk_priority;
409 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
410         __u32                   sk_cgrp_prioidx;
411 #endif
412         struct pid              *sk_peer_pid;
413         const struct cred       *sk_peer_cred;
414         long                    sk_rcvtimeo;
415         long                    sk_sndtimeo;
416         void                    *sk_protinfo;
417         struct timer_list       sk_timer;
418         ktime_t                 sk_stamp;
419         u16                     sk_tsflags;
420         u32                     sk_tskey;
421         struct socket           *sk_socket;
422         void                    *sk_user_data;
423         struct page_frag        sk_frag;
424         struct sk_buff          *sk_send_head;
425         __s32                   sk_peek_off;
426         int                     sk_write_pending;
427 #ifdef CONFIG_SECURITY
428         void                    *sk_security;
429 #endif
430         __u32                   sk_mark;
431         u32                     sk_classid;
432         struct cg_proto         *sk_cgrp;
433         void                    (*sk_state_change)(struct sock *sk);
434         void                    (*sk_data_ready)(struct sock *sk);
435         void                    (*sk_write_space)(struct sock *sk);
436         void                    (*sk_error_report)(struct sock *sk);
437         int                     (*sk_backlog_rcv)(struct sock *sk,
438                                                   struct sk_buff *skb);
439         void                    (*sk_destruct)(struct sock *sk);
440 };
441 
442 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
443 
444 #define rcu_dereference_sk_user_data(sk)        rcu_dereference(__sk_user_data((sk)))
445 #define rcu_assign_sk_user_data(sk, ptr)        rcu_assign_pointer(__sk_user_data((sk)), ptr)
446 
447 /*
448  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
449  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
450  * on a socket means that the socket will reuse everybody else's port
451  * without looking at the other's sk_reuse value.
452  */
453 
454 #define SK_NO_REUSE     0
455 #define SK_CAN_REUSE    1
456 #define SK_FORCE_REUSE  2
457 
458 static inline int sk_peek_offset(struct sock *sk, int flags)
459 {
460         if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
461                 return sk->sk_peek_off;
462         else
463                 return 0;
464 }
465 
466 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
467 {
468         if (sk->sk_peek_off >= 0) {
469                 if (sk->sk_peek_off >= val)
470                         sk->sk_peek_off -= val;
471                 else
472                         sk->sk_peek_off = 0;
473         }
474 }
475 
476 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
477 {
478         if (sk->sk_peek_off >= 0)
479                 sk->sk_peek_off += val;
480 }
481 
482 /*
483  * Hashed lists helper routines
484  */
485 static inline struct sock *sk_entry(const struct hlist_node *node)
486 {
487         return hlist_entry(node, struct sock, sk_node);
488 }
489 
490 static inline struct sock *__sk_head(const struct hlist_head *head)
491 {
492         return hlist_entry(head->first, struct sock, sk_node);
493 }
494 
495 static inline struct sock *sk_head(const struct hlist_head *head)
496 {
497         return hlist_empty(head) ? NULL : __sk_head(head);
498 }
499 
500 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
501 {
502         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
503 }
504 
505 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
506 {
507         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
508 }
509 
510 static inline struct sock *sk_next(const struct sock *sk)
511 {
512         return sk->sk_node.next ?
513                 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
514 }
515 
516 static inline struct sock *sk_nulls_next(const struct sock *sk)
517 {
518         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
519                 hlist_nulls_entry(sk->sk_nulls_node.next,
520                                   struct sock, sk_nulls_node) :
521                 NULL;
522 }
523 
524 static inline bool sk_unhashed(const struct sock *sk)
525 {
526         return hlist_unhashed(&sk->sk_node);
527 }
528 
529 static inline bool sk_hashed(const struct sock *sk)
530 {
531         return !sk_unhashed(sk);
532 }
533 
534 static inline void sk_node_init(struct hlist_node *node)
535 {
536         node->pprev = NULL;
537 }
538 
539 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
540 {
541         node->pprev = NULL;
542 }
543 
544 static inline void __sk_del_node(struct sock *sk)
545 {
546         __hlist_del(&sk->sk_node);
547 }
548 
549 /* NB: equivalent to hlist_del_init_rcu */
550 static inline bool __sk_del_node_init(struct sock *sk)
551 {
552         if (sk_hashed(sk)) {
553                 __sk_del_node(sk);
554                 sk_node_init(&sk->sk_node);
555                 return true;
556         }
557         return false;
558 }
559 
560 /* Grab socket reference count. This operation is valid only
561    when sk is ALREADY grabbed f.e. it is found in hash table
562    or a list and the lookup is made under lock preventing hash table
563    modifications.
564  */
565 
566 static inline void sock_hold(struct sock *sk)
567 {
568         atomic_inc(&sk->sk_refcnt);
569 }
570 
571 /* Ungrab socket in the context, which assumes that socket refcnt
572    cannot hit zero, f.e. it is true in context of any socketcall.
573  */
574 static inline void __sock_put(struct sock *sk)
575 {
576         atomic_dec(&sk->sk_refcnt);
577 }
578 
579 static inline bool sk_del_node_init(struct sock *sk)
580 {
581         bool rc = __sk_del_node_init(sk);
582 
583         if (rc) {
584                 /* paranoid for a while -acme */
585                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
586                 __sock_put(sk);
587         }
588         return rc;
589 }
590 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
591 
592 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
593 {
594         if (sk_hashed(sk)) {
595                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
596                 return true;
597         }
598         return false;
599 }
600 
601 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
602 {
603         bool rc = __sk_nulls_del_node_init_rcu(sk);
604 
605         if (rc) {
606                 /* paranoid for a while -acme */
607                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
608                 __sock_put(sk);
609         }
610         return rc;
611 }
612 
613 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
614 {
615         hlist_add_head(&sk->sk_node, list);
616 }
617 
618 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
619 {
620         sock_hold(sk);
621         __sk_add_node(sk, list);
622 }
623 
624 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
625 {
626         sock_hold(sk);
627         hlist_add_head_rcu(&sk->sk_node, list);
628 }
629 
630 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
631 {
632         hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
633 }
634 
635 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
636 {
637         sock_hold(sk);
638         __sk_nulls_add_node_rcu(sk, list);
639 }
640 
641 static inline void __sk_del_bind_node(struct sock *sk)
642 {
643         __hlist_del(&sk->sk_bind_node);
644 }
645 
646 static inline void sk_add_bind_node(struct sock *sk,
647                                         struct hlist_head *list)
648 {
649         hlist_add_head(&sk->sk_bind_node, list);
650 }
651 
652 #define sk_for_each(__sk, list) \
653         hlist_for_each_entry(__sk, list, sk_node)
654 #define sk_for_each_rcu(__sk, list) \
655         hlist_for_each_entry_rcu(__sk, list, sk_node)
656 #define sk_nulls_for_each(__sk, node, list) \
657         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
658 #define sk_nulls_for_each_rcu(__sk, node, list) \
659         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
660 #define sk_for_each_from(__sk) \
661         hlist_for_each_entry_from(__sk, sk_node)
662 #define sk_nulls_for_each_from(__sk, node) \
663         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
664                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
665 #define sk_for_each_safe(__sk, tmp, list) \
666         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
667 #define sk_for_each_bound(__sk, list) \
668         hlist_for_each_entry(__sk, list, sk_bind_node)
669 
670 /**
671  * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
672  * @tpos:       the type * to use as a loop cursor.
673  * @pos:        the &struct hlist_node to use as a loop cursor.
674  * @head:       the head for your list.
675  * @offset:     offset of hlist_node within the struct.
676  *
677  */
678 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset)                \
679         for (pos = (head)->first;                                              \
680              (!is_a_nulls(pos)) &&                                             \
681                 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;});       \
682              pos = pos->next)
683 
684 static inline struct user_namespace *sk_user_ns(struct sock *sk)
685 {
686         /* Careful only use this in a context where these parameters
687          * can not change and must all be valid, such as recvmsg from
688          * userspace.
689          */
690         return sk->sk_socket->file->f_cred->user_ns;
691 }
692 
693 /* Sock flags */
694 enum sock_flags {
695         SOCK_DEAD,
696         SOCK_DONE,
697         SOCK_URGINLINE,
698         SOCK_KEEPOPEN,
699         SOCK_LINGER,
700         SOCK_DESTROY,
701         SOCK_BROADCAST,
702         SOCK_TIMESTAMP,
703         SOCK_ZAPPED,
704         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
705         SOCK_DBG, /* %SO_DEBUG setting */
706         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
707         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
708         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
709         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
710         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
711         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
712         SOCK_FASYNC, /* fasync() active */
713         SOCK_RXQ_OVFL,
714         SOCK_ZEROCOPY, /* buffers from userspace */
715         SOCK_WIFI_STATUS, /* push wifi status to userspace */
716         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
717                      * Will use last 4 bytes of packet sent from
718                      * user-space instead.
719                      */
720         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
721         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
722 };
723 
724 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
725 {
726         nsk->sk_flags = osk->sk_flags;
727 }
728 
729 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
730 {
731         __set_bit(flag, &sk->sk_flags);
732 }
733 
734 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
735 {
736         __clear_bit(flag, &sk->sk_flags);
737 }
738 
739 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
740 {
741         return test_bit(flag, &sk->sk_flags);
742 }
743 
744 #ifdef CONFIG_NET
745 extern struct static_key memalloc_socks;
746 static inline int sk_memalloc_socks(void)
747 {
748         return static_key_false(&memalloc_socks);
749 }
750 #else
751 
752 static inline int sk_memalloc_socks(void)
753 {
754         return 0;
755 }
756 
757 #endif
758 
759 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
760 {
761         return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
762 }
763 
764 static inline void sk_acceptq_removed(struct sock *sk)
765 {
766         sk->sk_ack_backlog--;
767 }
768 
769 static inline void sk_acceptq_added(struct sock *sk)
770 {
771         sk->sk_ack_backlog++;
772 }
773 
774 static inline bool sk_acceptq_is_full(const struct sock *sk)
775 {
776         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
777 }
778 
779 /*
780  * Compute minimal free write space needed to queue new packets.
781  */
782 static inline int sk_stream_min_wspace(const struct sock *sk)
783 {
784         return sk->sk_wmem_queued >> 1;
785 }
786 
787 static inline int sk_stream_wspace(const struct sock *sk)
788 {
789         return sk->sk_sndbuf - sk->sk_wmem_queued;
790 }
791 
792 void sk_stream_write_space(struct sock *sk);
793 
794 /* OOB backlog add */
795 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
796 {
797         /* dont let skb dst not refcounted, we are going to leave rcu lock */
798         skb_dst_force(skb);
799 
800         if (!sk->sk_backlog.tail)
801                 sk->sk_backlog.head = skb;
802         else
803                 sk->sk_backlog.tail->next = skb;
804 
805         sk->sk_backlog.tail = skb;
806         skb->next = NULL;
807 }
808 
809 /*
810  * Take into account size of receive queue and backlog queue
811  * Do not take into account this skb truesize,
812  * to allow even a single big packet to come.
813  */
814 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
815 {
816         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
817 
818         return qsize > limit;
819 }
820 
821 /* The per-socket spinlock must be held here. */
822 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
823                                               unsigned int limit)
824 {
825         if (sk_rcvqueues_full(sk, limit))
826                 return -ENOBUFS;
827 
828         __sk_add_backlog(sk, skb);
829         sk->sk_backlog.len += skb->truesize;
830         return 0;
831 }
832 
833 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
834 
835 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
836 {
837         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
838                 return __sk_backlog_rcv(sk, skb);
839 
840         return sk->sk_backlog_rcv(sk, skb);
841 }
842 
843 static inline void sk_incoming_cpu_update(struct sock *sk)
844 {
845         sk->sk_incoming_cpu = raw_smp_processor_id();
846 }
847 
848 static inline void sock_rps_record_flow_hash(__u32 hash)
849 {
850 #ifdef CONFIG_RPS
851         struct rps_sock_flow_table *sock_flow_table;
852 
853         rcu_read_lock();
854         sock_flow_table = rcu_dereference(rps_sock_flow_table);
855         rps_record_sock_flow(sock_flow_table, hash);
856         rcu_read_unlock();
857 #endif
858 }
859 
860 static inline void sock_rps_record_flow(const struct sock *sk)
861 {
862 #ifdef CONFIG_RPS
863         sock_rps_record_flow_hash(sk->sk_rxhash);
864 #endif
865 }
866 
867 static inline void sock_rps_save_rxhash(struct sock *sk,
868                                         const struct sk_buff *skb)
869 {
870 #ifdef CONFIG_RPS
871         if (unlikely(sk->sk_rxhash != skb->hash))
872                 sk->sk_rxhash = skb->hash;
873 #endif
874 }
875 
876 static inline void sock_rps_reset_rxhash(struct sock *sk)
877 {
878 #ifdef CONFIG_RPS
879         sk->sk_rxhash = 0;
880 #endif
881 }
882 
883 #define sk_wait_event(__sk, __timeo, __condition)                       \
884         ({      int __rc;                                               \
885                 release_sock(__sk);                                     \
886                 __rc = __condition;                                     \
887                 if (!__rc) {                                            \
888                         *(__timeo) = schedule_timeout(*(__timeo));      \
889                 }                                                       \
890                 sched_annotate_sleep();                                         \
891                 lock_sock(__sk);                                        \
892                 __rc = __condition;                                     \
893                 __rc;                                                   \
894         })
895 
896 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
897 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
898 void sk_stream_wait_close(struct sock *sk, long timeo_p);
899 int sk_stream_error(struct sock *sk, int flags, int err);
900 void sk_stream_kill_queues(struct sock *sk);
901 void sk_set_memalloc(struct sock *sk);
902 void sk_clear_memalloc(struct sock *sk);
903 
904 int sk_wait_data(struct sock *sk, long *timeo);
905 
906 struct request_sock_ops;
907 struct timewait_sock_ops;
908 struct inet_hashinfo;
909 struct raw_hashinfo;
910 struct module;
911 
912 /*
913  * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
914  * un-modified. Special care is taken when initializing object to zero.
915  */
916 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
917 {
918         if (offsetof(struct sock, sk_node.next) != 0)
919                 memset(sk, 0, offsetof(struct sock, sk_node.next));
920         memset(&sk->sk_node.pprev, 0,
921                size - offsetof(struct sock, sk_node.pprev));
922 }
923 
924 /* Networking protocol blocks we attach to sockets.
925  * socket layer -> transport layer interface
926  * transport -> network interface is defined by struct inet_proto
927  */
928 struct proto {
929         void                    (*close)(struct sock *sk,
930                                         long timeout);
931         int                     (*connect)(struct sock *sk,
932                                         struct sockaddr *uaddr,
933                                         int addr_len);
934         int                     (*disconnect)(struct sock *sk, int flags);
935 
936         struct sock *           (*accept)(struct sock *sk, int flags, int *err);
937 
938         int                     (*ioctl)(struct sock *sk, int cmd,
939                                          unsigned long arg);
940         int                     (*init)(struct sock *sk);
941         void                    (*destroy)(struct sock *sk);
942         void                    (*shutdown)(struct sock *sk, int how);
943         int                     (*setsockopt)(struct sock *sk, int level,
944                                         int optname, char __user *optval,
945                                         unsigned int optlen);
946         int                     (*getsockopt)(struct sock *sk, int level,
947                                         int optname, char __user *optval,
948                                         int __user *option);
949 #ifdef CONFIG_COMPAT
950         int                     (*compat_setsockopt)(struct sock *sk,
951                                         int level,
952                                         int optname, char __user *optval,
953                                         unsigned int optlen);
954         int                     (*compat_getsockopt)(struct sock *sk,
955                                         int level,
956                                         int optname, char __user *optval,
957                                         int __user *option);
958         int                     (*compat_ioctl)(struct sock *sk,
959                                         unsigned int cmd, unsigned long arg);
960 #endif
961         int                     (*sendmsg)(struct kiocb *iocb, struct sock *sk,
962                                            struct msghdr *msg, size_t len);
963         int                     (*recvmsg)(struct kiocb *iocb, struct sock *sk,
964                                            struct msghdr *msg,
965                                            size_t len, int noblock, int flags,
966                                            int *addr_len);
967         int                     (*sendpage)(struct sock *sk, struct page *page,
968                                         int offset, size_t size, int flags);
969         int                     (*bind)(struct sock *sk,
970                                         struct sockaddr *uaddr, int addr_len);
971 
972         int                     (*backlog_rcv) (struct sock *sk,
973                                                 struct sk_buff *skb);
974 
975         void            (*release_cb)(struct sock *sk);
976 
977         /* Keeping track of sk's, looking them up, and port selection methods. */
978         void                    (*hash)(struct sock *sk);
979         void                    (*unhash)(struct sock *sk);
980         void                    (*rehash)(struct sock *sk);
981         int                     (*get_port)(struct sock *sk, unsigned short snum);
982         void                    (*clear_sk)(struct sock *sk, int size);
983 
984         /* Keeping track of sockets in use */
985 #ifdef CONFIG_PROC_FS
986         unsigned int            inuse_idx;
987 #endif
988 
989         bool                    (*stream_memory_free)(const struct sock *sk);
990         /* Memory pressure */
991         void                    (*enter_memory_pressure)(struct sock *sk);
992         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
993         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
994         /*
995          * Pressure flag: try to collapse.
996          * Technical note: it is used by multiple contexts non atomically.
997          * All the __sk_mem_schedule() is of this nature: accounting
998          * is strict, actions are advisory and have some latency.
999          */
1000         int                     *memory_pressure;
1001         long                    *sysctl_mem;
1002         int                     *sysctl_wmem;
1003         int                     *sysctl_rmem;
1004         int                     max_header;
1005         bool                    no_autobind;
1006 
1007         struct kmem_cache       *slab;
1008         unsigned int            obj_size;
1009         int                     slab_flags;
1010 
1011         struct percpu_counter   *orphan_count;
1012 
1013         struct request_sock_ops *rsk_prot;
1014         struct timewait_sock_ops *twsk_prot;
1015 
1016         union {
1017                 struct inet_hashinfo    *hashinfo;
1018                 struct udp_table        *udp_table;
1019                 struct raw_hashinfo     *raw_hash;
1020         } h;
1021 
1022         struct module           *owner;
1023 
1024         char                    name[32];
1025 
1026         struct list_head        node;
1027 #ifdef SOCK_REFCNT_DEBUG
1028         atomic_t                socks;
1029 #endif
1030 #ifdef CONFIG_MEMCG_KMEM
1031         /*
1032          * cgroup specific init/deinit functions. Called once for all
1033          * protocols that implement it, from cgroups populate function.
1034          * This function has to setup any files the protocol want to
1035          * appear in the kmem cgroup filesystem.
1036          */
1037         int                     (*init_cgroup)(struct mem_cgroup *memcg,
1038                                                struct cgroup_subsys *ss);
1039         void                    (*destroy_cgroup)(struct mem_cgroup *memcg);
1040         struct cg_proto         *(*proto_cgroup)(struct mem_cgroup *memcg);
1041 #endif
1042 };
1043 
1044 /*
1045  * Bits in struct cg_proto.flags
1046  */
1047 enum cg_proto_flags {
1048         /* Currently active and new sockets should be assigned to cgroups */
1049         MEMCG_SOCK_ACTIVE,
1050         /* It was ever activated; we must disarm static keys on destruction */
1051         MEMCG_SOCK_ACTIVATED,
1052 };
1053 
1054 struct cg_proto {
1055         struct page_counter     memory_allocated;       /* Current allocated memory. */
1056         struct percpu_counter   sockets_allocated;      /* Current number of sockets. */
1057         int                     memory_pressure;
1058         long                    sysctl_mem[3];
1059         unsigned long           flags;
1060         /*
1061          * memcg field is used to find which memcg we belong directly
1062          * Each memcg struct can hold more than one cg_proto, so container_of
1063          * won't really cut.
1064          *
1065          * The elegant solution would be having an inverse function to
1066          * proto_cgroup in struct proto, but that means polluting the structure
1067          * for everybody, instead of just for memcg users.
1068          */
1069         struct mem_cgroup       *memcg;
1070 };
1071 
1072 int proto_register(struct proto *prot, int alloc_slab);
1073 void proto_unregister(struct proto *prot);
1074 
1075 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1076 {
1077         return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1078 }
1079 
1080 #ifdef SOCK_REFCNT_DEBUG
1081 static inline void sk_refcnt_debug_inc(struct sock *sk)
1082 {
1083         atomic_inc(&sk->sk_prot->socks);
1084 }
1085 
1086 static inline void sk_refcnt_debug_dec(struct sock *sk)
1087 {
1088         atomic_dec(&sk->sk_prot->socks);
1089         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1090                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1091 }
1092 
1093 static inline void sk_refcnt_debug_release(const struct sock *sk)
1094 {
1095         if (atomic_read(&sk->sk_refcnt) != 1)
1096                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1097                        sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1098 }
1099 #else /* SOCK_REFCNT_DEBUG */
1100 #define sk_refcnt_debug_inc(sk) do { } while (0)
1101 #define sk_refcnt_debug_dec(sk) do { } while (0)
1102 #define sk_refcnt_debug_release(sk) do { } while (0)
1103 #endif /* SOCK_REFCNT_DEBUG */
1104 
1105 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1106 extern struct static_key memcg_socket_limit_enabled;
1107 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1108                                                struct cg_proto *cg_proto)
1109 {
1110         return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1111 }
1112 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1113 #else
1114 #define mem_cgroup_sockets_enabled 0
1115 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1116                                                struct cg_proto *cg_proto)
1117 {
1118         return NULL;
1119 }
1120 #endif
1121 
1122 static inline bool sk_stream_memory_free(const struct sock *sk)
1123 {
1124         if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1125                 return false;
1126 
1127         return sk->sk_prot->stream_memory_free ?
1128                 sk->sk_prot->stream_memory_free(sk) : true;
1129 }
1130 
1131 static inline bool sk_stream_is_writeable(const struct sock *sk)
1132 {
1133         return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1134                sk_stream_memory_free(sk);
1135 }
1136 
1137 
1138 static inline bool sk_has_memory_pressure(const struct sock *sk)
1139 {
1140         return sk->sk_prot->memory_pressure != NULL;
1141 }
1142 
1143 static inline bool sk_under_memory_pressure(const struct sock *sk)
1144 {
1145         if (!sk->sk_prot->memory_pressure)
1146                 return false;
1147 
1148         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1149                 return !!sk->sk_cgrp->memory_pressure;
1150 
1151         return !!*sk->sk_prot->memory_pressure;
1152 }
1153 
1154 static inline void sk_leave_memory_pressure(struct sock *sk)
1155 {
1156         int *memory_pressure = sk->sk_prot->memory_pressure;
1157 
1158         if (!memory_pressure)
1159                 return;
1160 
1161         if (*memory_pressure)
1162                 *memory_pressure = 0;
1163 
1164         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1165                 struct cg_proto *cg_proto = sk->sk_cgrp;
1166                 struct proto *prot = sk->sk_prot;
1167 
1168                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1169                         cg_proto->memory_pressure = 0;
1170         }
1171 
1172 }
1173 
1174 static inline void sk_enter_memory_pressure(struct sock *sk)
1175 {
1176         if (!sk->sk_prot->enter_memory_pressure)
1177                 return;
1178 
1179         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1180                 struct cg_proto *cg_proto = sk->sk_cgrp;
1181                 struct proto *prot = sk->sk_prot;
1182 
1183                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1184                         cg_proto->memory_pressure = 1;
1185         }
1186 
1187         sk->sk_prot->enter_memory_pressure(sk);
1188 }
1189 
1190 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1191 {
1192         long *prot = sk->sk_prot->sysctl_mem;
1193         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1194                 prot = sk->sk_cgrp->sysctl_mem;
1195         return prot[index];
1196 }
1197 
1198 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1199                                               unsigned long amt,
1200                                               int *parent_status)
1201 {
1202         page_counter_charge(&prot->memory_allocated, amt);
1203 
1204         if (page_counter_read(&prot->memory_allocated) >
1205             prot->memory_allocated.limit)
1206                 *parent_status = OVER_LIMIT;
1207 }
1208 
1209 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1210                                               unsigned long amt)
1211 {
1212         page_counter_uncharge(&prot->memory_allocated, amt);
1213 }
1214 
1215 static inline long
1216 sk_memory_allocated(const struct sock *sk)
1217 {
1218         struct proto *prot = sk->sk_prot;
1219 
1220         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1221                 return page_counter_read(&sk->sk_cgrp->memory_allocated);
1222 
1223         return atomic_long_read(prot->memory_allocated);
1224 }
1225 
1226 static inline long
1227 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1228 {
1229         struct proto *prot = sk->sk_prot;
1230 
1231         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1232                 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1233                 /* update the root cgroup regardless */
1234                 atomic_long_add_return(amt, prot->memory_allocated);
1235                 return page_counter_read(&sk->sk_cgrp->memory_allocated);
1236         }
1237 
1238         return atomic_long_add_return(amt, prot->memory_allocated);
1239 }
1240 
1241 static inline void
1242 sk_memory_allocated_sub(struct sock *sk, int amt)
1243 {
1244         struct proto *prot = sk->sk_prot;
1245 
1246         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1247                 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1248 
1249         atomic_long_sub(amt, prot->memory_allocated);
1250 }
1251 
1252 static inline void sk_sockets_allocated_dec(struct sock *sk)
1253 {
1254         struct proto *prot = sk->sk_prot;
1255 
1256         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1257                 struct cg_proto *cg_proto = sk->sk_cgrp;
1258 
1259                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1260                         percpu_counter_dec(&cg_proto->sockets_allocated);
1261         }
1262 
1263         percpu_counter_dec(prot->sockets_allocated);
1264 }
1265 
1266 static inline void sk_sockets_allocated_inc(struct sock *sk)
1267 {
1268         struct proto *prot = sk->sk_prot;
1269 
1270         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1271                 struct cg_proto *cg_proto = sk->sk_cgrp;
1272 
1273                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1274                         percpu_counter_inc(&cg_proto->sockets_allocated);
1275         }
1276 
1277         percpu_counter_inc(prot->sockets_allocated);
1278 }
1279 
1280 static inline int
1281 sk_sockets_allocated_read_positive(struct sock *sk)
1282 {
1283         struct proto *prot = sk->sk_prot;
1284 
1285         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1286                 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
1287 
1288         return percpu_counter_read_positive(prot->sockets_allocated);
1289 }
1290 
1291 static inline int
1292 proto_sockets_allocated_sum_positive(struct proto *prot)
1293 {
1294         return percpu_counter_sum_positive(prot->sockets_allocated);
1295 }
1296 
1297 static inline long
1298 proto_memory_allocated(struct proto *prot)
1299 {
1300         return atomic_long_read(prot->memory_allocated);
1301 }
1302 
1303 static inline bool
1304 proto_memory_pressure(struct proto *prot)
1305 {
1306         if (!prot->memory_pressure)
1307                 return false;
1308         return !!*prot->memory_pressure;
1309 }
1310 
1311 
1312 #ifdef CONFIG_PROC_FS
1313 /* Called with local bh disabled */
1314 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1315 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1316 #else
1317 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1318                 int inc)
1319 {
1320 }
1321 #endif
1322 
1323 
1324 /* With per-bucket locks this operation is not-atomic, so that
1325  * this version is not worse.
1326  */
1327 static inline void __sk_prot_rehash(struct sock *sk)
1328 {
1329         sk->sk_prot->unhash(sk);
1330         sk->sk_prot->hash(sk);
1331 }
1332 
1333 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1334 
1335 /* About 10 seconds */
1336 #define SOCK_DESTROY_TIME (10*HZ)
1337 
1338 /* Sockets 0-1023 can't be bound to unless you are superuser */
1339 #define PROT_SOCK       1024
1340 
1341 #define SHUTDOWN_MASK   3
1342 #define RCV_SHUTDOWN    1
1343 #define SEND_SHUTDOWN   2
1344 
1345 #define SOCK_SNDBUF_LOCK        1
1346 #define SOCK_RCVBUF_LOCK        2
1347 #define SOCK_BINDADDR_LOCK      4
1348 #define SOCK_BINDPORT_LOCK      8
1349 
1350 struct socket_alloc {
1351         struct socket socket;
1352         struct inode vfs_inode;
1353 };
1354 
1355 static inline struct socket *SOCKET_I(struct inode *inode)
1356 {
1357         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1358 }
1359 
1360 static inline struct inode *SOCK_INODE(struct socket *socket)
1361 {
1362         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1363 }
1364 
1365 /*
1366  * Functions for memory accounting
1367  */
1368 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1369 void __sk_mem_reclaim(struct sock *sk);
1370 
1371 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1372 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1373 #define SK_MEM_SEND     0
1374 #define SK_MEM_RECV     1
1375 
1376 static inline int sk_mem_pages(int amt)
1377 {
1378         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1379 }
1380 
1381 static inline bool sk_has_account(struct sock *sk)
1382 {
1383         /* return true if protocol supports memory accounting */
1384         return !!sk->sk_prot->memory_allocated;
1385 }
1386 
1387 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1388 {
1389         if (!sk_has_account(sk))
1390                 return true;
1391         return size <= sk->sk_forward_alloc ||
1392                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1393 }
1394 
1395 static inline bool
1396 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1397 {
1398         if (!sk_has_account(sk))
1399                 return true;
1400         return size<= sk->sk_forward_alloc ||
1401                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1402                 skb_pfmemalloc(skb);
1403 }
1404 
1405 static inline void sk_mem_reclaim(struct sock *sk)
1406 {
1407         if (!sk_has_account(sk))
1408                 return;
1409         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1410                 __sk_mem_reclaim(sk);
1411 }
1412 
1413 static inline void sk_mem_reclaim_partial(struct sock *sk)
1414 {
1415         if (!sk_has_account(sk))
1416                 return;
1417         if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1418                 __sk_mem_reclaim(sk);
1419 }
1420 
1421 static inline void sk_mem_charge(struct sock *sk, int size)
1422 {
1423         if (!sk_has_account(sk))
1424                 return;
1425         sk->sk_forward_alloc -= size;
1426 }
1427 
1428 static inline void sk_mem_uncharge(struct sock *sk, int size)
1429 {
1430         if (!sk_has_account(sk))
1431                 return;
1432         sk->sk_forward_alloc += size;
1433 }
1434 
1435 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1436 {
1437         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1438         sk->sk_wmem_queued -= skb->truesize;
1439         sk_mem_uncharge(sk, skb->truesize);
1440         __kfree_skb(skb);
1441 }
1442 
1443 /* Used by processes to "lock" a socket state, so that
1444  * interrupts and bottom half handlers won't change it
1445  * from under us. It essentially blocks any incoming
1446  * packets, so that we won't get any new data or any
1447  * packets that change the state of the socket.
1448  *
1449  * While locked, BH processing will add new packets to
1450  * the backlog queue.  This queue is processed by the
1451  * owner of the socket lock right before it is released.
1452  *
1453  * Since ~2.3.5 it is also exclusive sleep lock serializing
1454  * accesses from user process context.
1455  */
1456 #define sock_owned_by_user(sk)  ((sk)->sk_lock.owned)
1457 
1458 static inline void sock_release_ownership(struct sock *sk)
1459 {
1460         sk->sk_lock.owned = 0;
1461 }
1462 
1463 /*
1464  * Macro so as to not evaluate some arguments when
1465  * lockdep is not enabled.
1466  *
1467  * Mark both the sk_lock and the sk_lock.slock as a
1468  * per-address-family lock class.
1469  */
1470 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1471 do {                                                                    \
1472         sk->sk_lock.owned = 0;                                          \
1473         init_waitqueue_head(&sk->sk_lock.wq);                           \
1474         spin_lock_init(&(sk)->sk_lock.slock);                           \
1475         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1476                         sizeof((sk)->sk_lock));                         \
1477         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1478                                 (skey), (sname));                               \
1479         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1480 } while (0)
1481 
1482 void lock_sock_nested(struct sock *sk, int subclass);
1483 
1484 static inline void lock_sock(struct sock *sk)
1485 {
1486         lock_sock_nested(sk, 0);
1487 }
1488 
1489 void release_sock(struct sock *sk);
1490 
1491 /* BH context may only use the following locking interface. */
1492 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1493 #define bh_lock_sock_nested(__sk) \
1494                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1495                                 SINGLE_DEPTH_NESTING)
1496 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1497 
1498 bool lock_sock_fast(struct sock *sk);
1499 /**
1500  * unlock_sock_fast - complement of lock_sock_fast
1501  * @sk: socket
1502  * @slow: slow mode
1503  *
1504  * fast unlock socket for user context.
1505  * If slow mode is on, we call regular release_sock()
1506  */
1507 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1508 {
1509         if (slow)
1510                 release_sock(sk);
1511         else
1512                 spin_unlock_bh(&sk->sk_lock.slock);
1513 }
1514 
1515 
1516 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1517                       struct proto *prot);
1518 void sk_free(struct sock *sk);
1519 void sk_release_kernel(struct sock *sk);
1520 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1521 
1522 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1523                              gfp_t priority);
1524 void sock_wfree(struct sk_buff *skb);
1525 void skb_orphan_partial(struct sk_buff *skb);
1526 void sock_rfree(struct sk_buff *skb);
1527 void sock_efree(struct sk_buff *skb);
1528 #ifdef CONFIG_INET
1529 void sock_edemux(struct sk_buff *skb);
1530 #else
1531 #define sock_edemux(skb) sock_efree(skb)
1532 #endif
1533 
1534 int sock_setsockopt(struct socket *sock, int level, int op,
1535                     char __user *optval, unsigned int optlen);
1536 
1537 int sock_getsockopt(struct socket *sock, int level, int op,
1538                     char __user *optval, int __user *optlen);
1539 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1540                                     int noblock, int *errcode);
1541 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1542                                      unsigned long data_len, int noblock,
1543                                      int *errcode, int max_page_order);
1544 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1545 void sock_kfree_s(struct sock *sk, void *mem, int size);
1546 void sock_kzfree_s(struct sock *sk, void *mem, int size);
1547 void sk_send_sigurg(struct sock *sk);
1548 
1549 /*
1550  * Functions to fill in entries in struct proto_ops when a protocol
1551  * does not implement a particular function.
1552  */
1553 int sock_no_bind(struct socket *, struct sockaddr *, int);
1554 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1555 int sock_no_socketpair(struct socket *, struct socket *);
1556 int sock_no_accept(struct socket *, struct socket *, int);
1557 int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1558 unsigned int sock_no_poll(struct file *, struct socket *,
1559                           struct poll_table_struct *);
1560 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1561 int sock_no_listen(struct socket *, int);
1562 int sock_no_shutdown(struct socket *, int);
1563 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1564 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1565 int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1566 int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1567                     int);
1568 int sock_no_mmap(struct file *file, struct socket *sock,
1569                  struct vm_area_struct *vma);
1570 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1571                          size_t size, int flags);
1572 
1573 /*
1574  * Functions to fill in entries in struct proto_ops when a protocol
1575  * uses the inet style.
1576  */
1577 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1578                                   char __user *optval, int __user *optlen);
1579 int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1580                                struct msghdr *msg, size_t size, int flags);
1581 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1582                                   char __user *optval, unsigned int optlen);
1583 int compat_sock_common_getsockopt(struct socket *sock, int level,
1584                 int optname, char __user *optval, int __user *optlen);
1585 int compat_sock_common_setsockopt(struct socket *sock, int level,
1586                 int optname, char __user *optval, unsigned int optlen);
1587 
1588 void sk_common_release(struct sock *sk);
1589 
1590 /*
1591  *      Default socket callbacks and setup code
1592  */
1593 
1594 /* Initialise core socket variables */
1595 void sock_init_data(struct socket *sock, struct sock *sk);
1596 
1597 /*
1598  * Socket reference counting postulates.
1599  *
1600  * * Each user of socket SHOULD hold a reference count.
1601  * * Each access point to socket (an hash table bucket, reference from a list,
1602  *   running timer, skb in flight MUST hold a reference count.
1603  * * When reference count hits 0, it means it will never increase back.
1604  * * When reference count hits 0, it means that no references from
1605  *   outside exist to this socket and current process on current CPU
1606  *   is last user and may/should destroy this socket.
1607  * * sk_free is called from any context: process, BH, IRQ. When
1608  *   it is called, socket has no references from outside -> sk_free
1609  *   may release descendant resources allocated by the socket, but
1610  *   to the time when it is called, socket is NOT referenced by any
1611  *   hash tables, lists etc.
1612  * * Packets, delivered from outside (from network or from another process)
1613  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1614  *   when they sit in queue. Otherwise, packets will leak to hole, when
1615  *   socket is looked up by one cpu and unhasing is made by another CPU.
1616  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1617  *   (leak to backlog). Packet socket does all the processing inside
1618  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1619  *   use separate SMP lock, so that they are prone too.
1620  */
1621 
1622 /* Ungrab socket and destroy it, if it was the last reference. */
1623 static inline void sock_put(struct sock *sk)
1624 {
1625         if (atomic_dec_and_test(&sk->sk_refcnt))
1626                 sk_free(sk);
1627 }
1628 /* Generic version of sock_put(), dealing with all sockets
1629  * (TCP_TIMEWAIT, ESTABLISHED...)
1630  */
1631 void sock_gen_put(struct sock *sk);
1632 
1633 int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
1634 
1635 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1636 {
1637         sk->sk_tx_queue_mapping = tx_queue;
1638 }
1639 
1640 static inline void sk_tx_queue_clear(struct sock *sk)
1641 {
1642         sk->sk_tx_queue_mapping = -1;
1643 }
1644 
1645 static inline int sk_tx_queue_get(const struct sock *sk)
1646 {
1647         return sk ? sk->sk_tx_queue_mapping : -1;
1648 }
1649 
1650 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1651 {
1652         sk_tx_queue_clear(sk);
1653         sk->sk_socket = sock;
1654 }
1655 
1656 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1657 {
1658         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1659         return &rcu_dereference_raw(sk->sk_wq)->wait;
1660 }
1661 /* Detach socket from process context.
1662  * Announce socket dead, detach it from wait queue and inode.
1663  * Note that parent inode held reference count on this struct sock,
1664  * we do not release it in this function, because protocol
1665  * probably wants some additional cleanups or even continuing
1666  * to work with this socket (TCP).
1667  */
1668 static inline void sock_orphan(struct sock *sk)
1669 {
1670         write_lock_bh(&sk->sk_callback_lock);
1671         sock_set_flag(sk, SOCK_DEAD);
1672         sk_set_socket(sk, NULL);
1673         sk->sk_wq  = NULL;
1674         write_unlock_bh(&sk->sk_callback_lock);
1675 }
1676 
1677 static inline void sock_graft(struct sock *sk, struct socket *parent)
1678 {
1679         write_lock_bh(&sk->sk_callback_lock);
1680         sk->sk_wq = parent->wq;
1681         parent->sk = sk;
1682         sk_set_socket(sk, parent);
1683         security_sock_graft(sk, parent);
1684         write_unlock_bh(&sk->sk_callback_lock);
1685 }
1686 
1687 kuid_t sock_i_uid(struct sock *sk);
1688 unsigned long sock_i_ino(struct sock *sk);
1689 
1690 static inline struct dst_entry *
1691 __sk_dst_get(struct sock *sk)
1692 {
1693         return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1694                                                        lockdep_is_held(&sk->sk_lock.slock));
1695 }
1696 
1697 static inline struct dst_entry *
1698 sk_dst_get(struct sock *sk)
1699 {
1700         struct dst_entry *dst;
1701 
1702         rcu_read_lock();
1703         dst = rcu_dereference(sk->sk_dst_cache);
1704         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1705                 dst = NULL;
1706         rcu_read_unlock();
1707         return dst;
1708 }
1709 
1710 static inline void dst_negative_advice(struct sock *sk)
1711 {
1712         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1713 
1714         if (dst && dst->ops->negative_advice) {
1715                 ndst = dst->ops->negative_advice(dst);
1716 
1717                 if (ndst != dst) {
1718                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1719                         sk_tx_queue_clear(sk);
1720                 }
1721         }
1722 }
1723 
1724 static inline void
1725 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1726 {
1727         struct dst_entry *old_dst;
1728 
1729         sk_tx_queue_clear(sk);
1730         /*
1731          * This can be called while sk is owned by the caller only,
1732          * with no state that can be checked in a rcu_dereference_check() cond
1733          */
1734         old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1735         rcu_assign_pointer(sk->sk_dst_cache, dst);
1736         dst_release(old_dst);
1737 }
1738 
1739 static inline void
1740 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1741 {
1742         struct dst_entry *old_dst;
1743 
1744         sk_tx_queue_clear(sk);
1745         old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1746         dst_release(old_dst);
1747 }
1748 
1749 static inline void
1750 __sk_dst_reset(struct sock *sk)
1751 {
1752         __sk_dst_set(sk, NULL);
1753 }
1754 
1755 static inline void
1756 sk_dst_reset(struct sock *sk)
1757 {
1758         sk_dst_set(sk, NULL);
1759 }
1760 
1761 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1762 
1763 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1764 
1765 bool sk_mc_loop(struct sock *sk);
1766 
1767 static inline bool sk_can_gso(const struct sock *sk)
1768 {
1769         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1770 }
1771 
1772 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1773 
1774 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1775 {
1776         sk->sk_route_nocaps |= flags;
1777         sk->sk_route_caps &= ~flags;
1778 }
1779 
1780 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1781                                            struct iov_iter *from, char *to,
1782                                            int copy, int offset)
1783 {
1784         if (skb->ip_summed == CHECKSUM_NONE) {
1785                 __wsum csum = 0;
1786                 if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1787                         return -EFAULT;
1788                 skb->csum = csum_block_add(skb->csum, csum, offset);
1789         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1790                 if (copy_from_iter_nocache(to, copy, from) != copy)
1791                         return -EFAULT;
1792         } else if (copy_from_iter(to, copy, from) != copy)
1793                 return -EFAULT;
1794 
1795         return 0;
1796 }
1797 
1798 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1799                                        struct iov_iter *from, int copy)
1800 {
1801         int err, offset = skb->len;
1802 
1803         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1804                                        copy, offset);
1805         if (err)
1806                 __skb_trim(skb, offset);
1807 
1808         return err;
1809 }
1810 
1811 static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
1812                                            struct sk_buff *skb,
1813                                            struct page *page,
1814                                            int off, int copy)
1815 {
1816         int err;
1817 
1818         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1819                                        copy, skb->len);
1820         if (err)
1821                 return err;
1822 
1823         skb->len             += copy;
1824         skb->data_len        += copy;
1825         skb->truesize        += copy;
1826         sk->sk_wmem_queued   += copy;
1827         sk_mem_charge(sk, copy);
1828         return 0;
1829 }
1830 
1831 /**
1832  * sk_wmem_alloc_get - returns write allocations
1833  * @sk: socket
1834  *
1835  * Returns sk_wmem_alloc minus initial offset of one
1836  */
1837 static inline int sk_wmem_alloc_get(const struct sock *sk)
1838 {
1839         return atomic_read(&sk->sk_wmem_alloc) - 1;
1840 }
1841 
1842 /**
1843  * sk_rmem_alloc_get - returns read allocations
1844  * @sk: socket
1845  *
1846  * Returns sk_rmem_alloc
1847  */
1848 static inline int sk_rmem_alloc_get(const struct sock *sk)
1849 {
1850         return atomic_read(&sk->sk_rmem_alloc);
1851 }
1852 
1853 /**
1854  * sk_has_allocations - check if allocations are outstanding
1855  * @sk: socket
1856  *
1857  * Returns true if socket has write or read allocations
1858  */
1859 static inline bool sk_has_allocations(const struct sock *sk)
1860 {
1861         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1862 }
1863 
1864 /**
1865  * wq_has_sleeper - check if there are any waiting processes
1866  * @wq: struct socket_wq
1867  *
1868  * Returns true if socket_wq has waiting processes
1869  *
1870  * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1871  * barrier call. They were added due to the race found within the tcp code.
1872  *
1873  * Consider following tcp code paths:
1874  *
1875  * CPU1                  CPU2
1876  *
1877  * sys_select            receive packet
1878  *   ...                 ...
1879  *   __add_wait_queue    update tp->rcv_nxt
1880  *   ...                 ...
1881  *   tp->rcv_nxt check   sock_def_readable
1882  *   ...                 {
1883  *   schedule               rcu_read_lock();
1884  *                          wq = rcu_dereference(sk->sk_wq);
1885  *                          if (wq && waitqueue_active(&wq->wait))
1886  *                              wake_up_interruptible(&wq->wait)
1887  *                          ...
1888  *                       }
1889  *
1890  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1891  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1892  * could then endup calling schedule and sleep forever if there are no more
1893  * data on the socket.
1894  *
1895  */
1896 static inline bool wq_has_sleeper(struct socket_wq *wq)
1897 {
1898         /* We need to be sure we are in sync with the
1899          * add_wait_queue modifications to the wait queue.
1900          *
1901          * This memory barrier is paired in the sock_poll_wait.
1902          */
1903         smp_mb();
1904         return wq && waitqueue_active(&wq->wait);
1905 }
1906 
1907 /**
1908  * sock_poll_wait - place memory barrier behind the poll_wait call.
1909  * @filp:           file
1910  * @wait_address:   socket wait queue
1911  * @p:              poll_table
1912  *
1913  * See the comments in the wq_has_sleeper function.
1914  */
1915 static inline void sock_poll_wait(struct file *filp,
1916                 wait_queue_head_t *wait_address, poll_table *p)
1917 {
1918         if (!poll_does_not_wait(p) && wait_address) {
1919                 poll_wait(filp, wait_address, p);
1920                 /* We need to be sure we are in sync with the
1921                  * socket flags modification.
1922                  *
1923                  * This memory barrier is paired in the wq_has_sleeper.
1924                  */
1925                 smp_mb();
1926         }
1927 }
1928 
1929 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1930 {
1931         if (sk->sk_txhash) {
1932                 skb->l4_hash = 1;
1933                 skb->hash = sk->sk_txhash;
1934         }
1935 }
1936 
1937 /*
1938  *      Queue a received datagram if it will fit. Stream and sequenced
1939  *      protocols can't normally use this as they need to fit buffers in
1940  *      and play with them.
1941  *
1942  *      Inlined as it's very short and called for pretty much every
1943  *      packet ever received.
1944  */
1945 
1946 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1947 {
1948         skb_orphan(skb);
1949         skb->sk = sk;
1950         skb->destructor = sock_wfree;
1951         skb_set_hash_from_sk(skb, sk);
1952         /*
1953          * We used to take a refcount on sk, but following operation
1954          * is enough to guarantee sk_free() wont free this sock until
1955          * all in-flight packets are completed
1956          */
1957         atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1958 }
1959 
1960 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1961 {
1962         skb_orphan(skb);
1963         skb->sk = sk;
1964         skb->destructor = sock_rfree;
1965         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1966         sk_mem_charge(sk, skb->truesize);
1967 }
1968 
1969 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1970                     unsigned long expires);
1971 
1972 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1973 
1974 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1975 
1976 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1977 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1978 
1979 /*
1980  *      Recover an error report and clear atomically
1981  */
1982 
1983 static inline int sock_error(struct sock *sk)
1984 {
1985         int err;
1986         if (likely(!sk->sk_err))
1987                 return 0;
1988         err = xchg(&sk->sk_err, 0);
1989         return -err;
1990 }
1991 
1992 static inline unsigned long sock_wspace(struct sock *sk)
1993 {
1994         int amt = 0;
1995 
1996         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1997                 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1998                 if (amt < 0)
1999                         amt = 0;
2000         }
2001         return amt;
2002 }
2003 
2004 static inline void sk_wake_async(struct sock *sk, int how, int band)
2005 {
2006         if (sock_flag(sk, SOCK_FASYNC))
2007                 sock_wake_async(sk->sk_socket, how, band);
2008 }
2009 
2010 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2011  * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2012  * Note: for send buffers, TCP works better if we can build two skbs at
2013  * minimum.
2014  */
2015 #define TCP_SKB_MIN_TRUESIZE    (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2016 
2017 #define SOCK_MIN_SNDBUF         (TCP_SKB_MIN_TRUESIZE * 2)
2018 #define SOCK_MIN_RCVBUF          TCP_SKB_MIN_TRUESIZE
2019 
2020 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2021 {
2022         if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2023                 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2024                 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2025         }
2026 }
2027 
2028 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2029 
2030 /**
2031  * sk_page_frag - return an appropriate page_frag
2032  * @sk: socket
2033  *
2034  * If socket allocation mode allows current thread to sleep, it means its
2035  * safe to use the per task page_frag instead of the per socket one.
2036  */
2037 static inline struct page_frag *sk_page_frag(struct sock *sk)
2038 {
2039         if (sk->sk_allocation & __GFP_WAIT)
2040                 return &current->task_frag;
2041 
2042         return &sk->sk_frag;
2043 }
2044 
2045 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2046 
2047 /*
2048  *      Default write policy as shown to user space via poll/select/SIGIO
2049  */
2050 static inline bool sock_writeable(const struct sock *sk)
2051 {
2052         return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2053 }
2054 
2055 static inline gfp_t gfp_any(void)
2056 {
2057         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2058 }
2059 
2060 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2061 {
2062         return noblock ? 0 : sk->sk_rcvtimeo;
2063 }
2064 
2065 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2066 {
2067         return noblock ? 0 : sk->sk_sndtimeo;
2068 }
2069 
2070 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2071 {
2072         return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2073 }
2074 
2075 /* Alas, with timeout socket operations are not restartable.
2076  * Compare this to poll().
2077  */
2078 static inline int sock_intr_errno(long timeo)
2079 {
2080         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2081 }
2082 
2083 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2084                            struct sk_buff *skb);
2085 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2086                              struct sk_buff *skb);
2087 
2088 static inline void
2089 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2090 {
2091         ktime_t kt = skb->tstamp;
2092         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2093 
2094         /*
2095          * generate control messages if
2096          * - receive time stamping in software requested
2097          * - software time stamp available and wanted
2098          * - hardware time stamps available and wanted
2099          */
2100         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2101             (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2102             (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2103             (hwtstamps->hwtstamp.tv64 &&
2104              (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2105                 __sock_recv_timestamp(msg, sk, skb);
2106         else
2107                 sk->sk_stamp = kt;
2108 
2109         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2110                 __sock_recv_wifi_status(msg, sk, skb);
2111 }
2112 
2113 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2114                               struct sk_buff *skb);
2115 
2116 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2117                                           struct sk_buff *skb)
2118 {
2119 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2120                            (1UL << SOCK_RCVTSTAMP))
2121 #define TSFLAGS_ANY       (SOF_TIMESTAMPING_SOFTWARE                    | \
2122                            SOF_TIMESTAMPING_RAW_HARDWARE)
2123 
2124         if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2125                 __sock_recv_ts_and_drops(msg, sk, skb);
2126         else
2127                 sk->sk_stamp = skb->tstamp;
2128 }
2129 
2130 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags);
2131 
2132 /**
2133  * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2134  * @sk:         socket sending this packet
2135  * @tx_flags:   completed with instructions for time stamping
2136  *
2137  * Note : callers should take care of initial *tx_flags value (usually 0)
2138  */
2139 static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
2140 {
2141         if (unlikely(sk->sk_tsflags))
2142                 __sock_tx_timestamp(sk, tx_flags);
2143         if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2144                 *tx_flags |= SKBTX_WIFI_STATUS;
2145 }
2146 
2147 /**
2148  * sk_eat_skb - Release a skb if it is no longer needed
2149  * @sk: socket to eat this skb from
2150  * @skb: socket buffer to eat
2151  *
2152  * This routine must be called with interrupts disabled or with the socket
2153  * locked so that the sk_buff queue operation is ok.
2154 */
2155 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2156 {
2157         __skb_unlink(skb, &sk->sk_receive_queue);
2158         __kfree_skb(skb);
2159 }
2160 
2161 static inline
2162 struct net *sock_net(const struct sock *sk)
2163 {
2164         return read_pnet(&sk->sk_net);
2165 }
2166 
2167 static inline
2168 void sock_net_set(struct sock *sk, struct net *net)
2169 {
2170         write_pnet(&sk->sk_net, net);
2171 }
2172 
2173 /*
2174  * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2175  * They should not hold a reference to a namespace in order to allow
2176  * to stop it.
2177  * Sockets after sk_change_net should be released using sk_release_kernel
2178  */
2179 static inline void sk_change_net(struct sock *sk, struct net *net)
2180 {
2181         struct net *current_net = sock_net(sk);
2182 
2183         if (!net_eq(current_net, net)) {
2184                 put_net(current_net);
2185                 sock_net_set(sk, hold_net(net));
2186         }
2187 }
2188 
2189 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2190 {
2191         if (skb->sk) {
2192                 struct sock *sk = skb->sk;
2193 
2194                 skb->destructor = NULL;
2195                 skb->sk = NULL;
2196                 return sk;
2197         }
2198         return NULL;
2199 }
2200 
2201 void sock_enable_timestamp(struct sock *sk, int flag);
2202 int sock_get_timestamp(struct sock *, struct timeval __user *);
2203 int sock_get_timestampns(struct sock *, struct timespec __user *);
2204 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2205                        int type);
2206 
2207 bool sk_ns_capable(const struct sock *sk,
2208                    struct user_namespace *user_ns, int cap);
2209 bool sk_capable(const struct sock *sk, int cap);
2210 bool sk_net_capable(const struct sock *sk, int cap);
2211 
2212 extern __u32 sysctl_wmem_max;
2213 extern __u32 sysctl_rmem_max;
2214 
2215 extern int sysctl_tstamp_allow_data;
2216 extern int sysctl_optmem_max;
2217 
2218 extern __u32 sysctl_wmem_default;
2219 extern __u32 sysctl_rmem_default;
2220 
2221 #endif  /* _SOCK_H */
2222 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us