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

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

Version: ~ [ linux-5.15-rc5 ] ~ [ linux-5.14.11 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.72 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.152 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.210 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.250 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.286 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.288 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ 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() and INET_TW_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          * fields between dontcopy_begin/dontcopy_end
196          * are not copied in sock_copy()
197          */
198         /* private: */
199         int                     skc_dontcopy_begin[0];
200         /* public: */
201         union {
202                 struct hlist_node       skc_node;
203                 struct hlist_nulls_node skc_nulls_node;
204         };
205         int                     skc_tx_queue_mapping;
206         atomic_t                skc_refcnt;
207         /* private: */
208         int                     skc_dontcopy_end[0];
209         /* public: */
210 };
211 
212 struct cg_proto;
213 /**
214   *     struct sock - network layer representation of sockets
215   *     @__sk_common: shared layout with inet_timewait_sock
216   *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217   *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218   *     @sk_lock:       synchronizer
219   *     @sk_rcvbuf: size of receive buffer in bytes
220   *     @sk_wq: sock wait queue and async head
221   *     @sk_rx_dst: receive input route used by early tcp demux
222   *     @sk_dst_cache: destination cache
223   *     @sk_dst_lock: destination cache lock
224   *     @sk_policy: flow policy
225   *     @sk_receive_queue: incoming packets
226   *     @sk_wmem_alloc: transmit queue bytes committed
227   *     @sk_write_queue: Packet sending queue
228   *     @sk_async_wait_queue: DMA copied packets
229   *     @sk_omem_alloc: "o" is "option" or "other"
230   *     @sk_wmem_queued: persistent queue size
231   *     @sk_forward_alloc: space allocated forward
232   *     @sk_allocation: allocation mode
233   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
234   *     @sk_sndbuf: size of send buffer in bytes
235   *     @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
236   *                %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
237   *     @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
238   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
239   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
240   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
241   *     @sk_gso_max_size: Maximum GSO segment size to build
242   *     @sk_gso_max_segs: Maximum number of GSO segments
243   *     @sk_lingertime: %SO_LINGER l_linger setting
244   *     @sk_backlog: always used with the per-socket spinlock held
245   *     @sk_callback_lock: used with the callbacks in the end of this struct
246   *     @sk_error_queue: rarely used
247   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
248   *                       IPV6_ADDRFORM for instance)
249   *     @sk_err: last error
250   *     @sk_err_soft: errors that don't cause failure but are the cause of a
251   *                   persistent failure not just 'timed out'
252   *     @sk_drops: raw/udp drops counter
253   *     @sk_ack_backlog: current listen backlog
254   *     @sk_max_ack_backlog: listen backlog set in listen()
255   *     @sk_priority: %SO_PRIORITY setting
256   *     @sk_cgrp_prioidx: socket group's priority map index
257   *     @sk_type: socket type (%SOCK_STREAM, etc)
258   *     @sk_protocol: which protocol this socket belongs in this network family
259   *     @sk_peer_pid: &struct pid for this socket's peer
260   *     @sk_peer_cred: %SO_PEERCRED setting
261   *     @sk_rcvlowat: %SO_RCVLOWAT setting
262   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
263   *     @sk_sndtimeo: %SO_SNDTIMEO setting
264   *     @sk_rxhash: flow hash received from netif layer
265   *     @sk_filter: socket filtering instructions
266   *     @sk_protinfo: private area, net family specific, when not using slab
267   *     @sk_timer: sock cleanup timer
268   *     @sk_stamp: time stamp of last packet received
269   *     @sk_socket: Identd and reporting IO signals
270   *     @sk_user_data: RPC layer private data
271   *     @sk_frag: cached page frag
272   *     @sk_peek_off: current peek_offset value
273   *     @sk_send_head: front of stuff to transmit
274   *     @sk_security: used by security modules
275   *     @sk_mark: generic packet mark
276   *     @sk_classid: this socket's cgroup classid
277   *     @sk_cgrp: this socket's cgroup-specific proto data
278   *     @sk_write_pending: a write to stream socket waits to start
279   *     @sk_state_change: callback to indicate change in the state of the sock
280   *     @sk_data_ready: callback to indicate there is data to be processed
281   *     @sk_write_space: callback to indicate there is bf sending space available
282   *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
283   *     @sk_backlog_rcv: callback to process the backlog
284   *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
285  */
286 struct sock {
287         /*
288          * Now struct inet_timewait_sock also uses sock_common, so please just
289          * don't add nothing before this first member (__sk_common) --acme
290          */
291         struct sock_common      __sk_common;
292 #define sk_node                 __sk_common.skc_node
293 #define sk_nulls_node           __sk_common.skc_nulls_node
294 #define sk_refcnt               __sk_common.skc_refcnt
295 #define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
296 
297 #define sk_dontcopy_begin       __sk_common.skc_dontcopy_begin
298 #define sk_dontcopy_end         __sk_common.skc_dontcopy_end
299 #define sk_hash                 __sk_common.skc_hash
300 #define sk_family               __sk_common.skc_family
301 #define sk_state                __sk_common.skc_state
302 #define sk_reuse                __sk_common.skc_reuse
303 #define sk_reuseport            __sk_common.skc_reuseport
304 #define sk_bound_dev_if         __sk_common.skc_bound_dev_if
305 #define sk_bind_node            __sk_common.skc_bind_node
306 #define sk_prot                 __sk_common.skc_prot
307 #define sk_net                  __sk_common.skc_net
308         socket_lock_t           sk_lock;
309         struct sk_buff_head     sk_receive_queue;
310         /*
311          * The backlog queue is special, it is always used with
312          * the per-socket spinlock held and requires low latency
313          * access. Therefore we special case it's implementation.
314          * Note : rmem_alloc is in this structure to fill a hole
315          * on 64bit arches, not because its logically part of
316          * backlog.
317          */
318         struct {
319                 atomic_t        rmem_alloc;
320                 int             len;
321                 struct sk_buff  *head;
322                 struct sk_buff  *tail;
323         } sk_backlog;
324 #define sk_rmem_alloc sk_backlog.rmem_alloc
325         int                     sk_forward_alloc;
326 #ifdef CONFIG_RPS
327         __u32                   sk_rxhash;
328 #endif
329         atomic_t                sk_drops;
330         int                     sk_rcvbuf;
331 
332         struct sk_filter __rcu  *sk_filter;
333         struct socket_wq __rcu  *sk_wq;
334 
335 #ifdef CONFIG_NET_DMA
336         struct sk_buff_head     sk_async_wait_queue;
337 #endif
338 
339 #ifdef CONFIG_XFRM
340         struct xfrm_policy      *sk_policy[2];
341 #endif
342         unsigned long           sk_flags;
343         struct dst_entry        *sk_rx_dst;
344         struct dst_entry __rcu  *sk_dst_cache;
345         spinlock_t              sk_dst_lock;
346         atomic_t                sk_wmem_alloc;
347         atomic_t                sk_omem_alloc;
348         int                     sk_sndbuf;
349         struct sk_buff_head     sk_write_queue;
350         kmemcheck_bitfield_begin(flags);
351         unsigned int            sk_shutdown  : 2,
352                                 sk_no_check  : 2,
353                                 sk_userlocks : 4,
354                                 sk_protocol  : 8,
355 #define SK_PROTOCOL_MAX U8_MAX
356                                 sk_type      : 16;
357         kmemcheck_bitfield_end(flags);
358         int                     sk_wmem_queued;
359         gfp_t                   sk_allocation;
360         u32                     sk_pacing_rate; /* bytes per second */
361         netdev_features_t       sk_route_caps;
362         netdev_features_t       sk_route_nocaps;
363         int                     sk_gso_type;
364         unsigned int            sk_gso_max_size;
365         u16                     sk_gso_max_segs;
366         int                     sk_rcvlowat;
367         unsigned long           sk_lingertime;
368         struct sk_buff_head     sk_error_queue;
369         struct proto            *sk_prot_creator;
370         rwlock_t                sk_callback_lock;
371         int                     sk_err,
372                                 sk_err_soft;
373         unsigned short          sk_ack_backlog;
374         unsigned short          sk_max_ack_backlog;
375         __u32                   sk_priority;
376 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
377         __u32                   sk_cgrp_prioidx;
378 #endif
379         struct pid              *sk_peer_pid;
380         const struct cred       *sk_peer_cred;
381         long                    sk_rcvtimeo;
382         long                    sk_sndtimeo;
383         void                    *sk_protinfo;
384         struct timer_list       sk_timer;
385         ktime_t                 sk_stamp;
386         struct socket           *sk_socket;
387         void                    *sk_user_data;
388         struct page_frag        sk_frag;
389         struct sk_buff          *sk_send_head;
390         __s32                   sk_peek_off;
391         int                     sk_write_pending;
392 #ifdef CONFIG_SECURITY
393         void                    *sk_security;
394 #endif
395         __u32                   sk_mark;
396         u32                     sk_classid;
397         struct cg_proto         *sk_cgrp;
398         void                    (*sk_state_change)(struct sock *sk);
399         void                    (*sk_data_ready)(struct sock *sk, int bytes);
400         void                    (*sk_write_space)(struct sock *sk);
401         void                    (*sk_error_report)(struct sock *sk);
402         int                     (*sk_backlog_rcv)(struct sock *sk,
403                                                   struct sk_buff *skb);
404         void                    (*sk_destruct)(struct sock *sk);
405 };
406 
407 /*
408  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
409  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
410  * on a socket means that the socket will reuse everybody else's port
411  * without looking at the other's sk_reuse value.
412  */
413 
414 #define SK_NO_REUSE     0
415 #define SK_CAN_REUSE    1
416 #define SK_FORCE_REUSE  2
417 
418 static inline int sk_peek_offset(struct sock *sk, int flags)
419 {
420         if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
421                 return sk->sk_peek_off;
422         else
423                 return 0;
424 }
425 
426 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
427 {
428         if (sk->sk_peek_off >= 0) {
429                 if (sk->sk_peek_off >= val)
430                         sk->sk_peek_off -= val;
431                 else
432                         sk->sk_peek_off = 0;
433         }
434 }
435 
436 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
437 {
438         if (sk->sk_peek_off >= 0)
439                 sk->sk_peek_off += val;
440 }
441 
442 /*
443  * Hashed lists helper routines
444  */
445 static inline struct sock *sk_entry(const struct hlist_node *node)
446 {
447         return hlist_entry(node, struct sock, sk_node);
448 }
449 
450 static inline struct sock *__sk_head(const struct hlist_head *head)
451 {
452         return hlist_entry(head->first, struct sock, sk_node);
453 }
454 
455 static inline struct sock *sk_head(const struct hlist_head *head)
456 {
457         return hlist_empty(head) ? NULL : __sk_head(head);
458 }
459 
460 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
461 {
462         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
463 }
464 
465 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
466 {
467         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
468 }
469 
470 static inline struct sock *sk_next(const struct sock *sk)
471 {
472         return sk->sk_node.next ?
473                 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
474 }
475 
476 static inline struct sock *sk_nulls_next(const struct sock *sk)
477 {
478         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
479                 hlist_nulls_entry(sk->sk_nulls_node.next,
480                                   struct sock, sk_nulls_node) :
481                 NULL;
482 }
483 
484 static inline bool sk_unhashed(const struct sock *sk)
485 {
486         return hlist_unhashed(&sk->sk_node);
487 }
488 
489 static inline bool sk_hashed(const struct sock *sk)
490 {
491         return !sk_unhashed(sk);
492 }
493 
494 static inline void sk_node_init(struct hlist_node *node)
495 {
496         node->pprev = NULL;
497 }
498 
499 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
500 {
501         node->pprev = NULL;
502 }
503 
504 static inline void __sk_del_node(struct sock *sk)
505 {
506         __hlist_del(&sk->sk_node);
507 }
508 
509 /* NB: equivalent to hlist_del_init_rcu */
510 static inline bool __sk_del_node_init(struct sock *sk)
511 {
512         if (sk_hashed(sk)) {
513                 __sk_del_node(sk);
514                 sk_node_init(&sk->sk_node);
515                 return true;
516         }
517         return false;
518 }
519 
520 /* Grab socket reference count. This operation is valid only
521    when sk is ALREADY grabbed f.e. it is found in hash table
522    or a list and the lookup is made under lock preventing hash table
523    modifications.
524  */
525 
526 static inline void sock_hold(struct sock *sk)
527 {
528         atomic_inc(&sk->sk_refcnt);
529 }
530 
531 /* Ungrab socket in the context, which assumes that socket refcnt
532    cannot hit zero, f.e. it is true in context of any socketcall.
533  */
534 static inline void __sock_put(struct sock *sk)
535 {
536         atomic_dec(&sk->sk_refcnt);
537 }
538 
539 static inline bool sk_del_node_init(struct sock *sk)
540 {
541         bool rc = __sk_del_node_init(sk);
542 
543         if (rc) {
544                 /* paranoid for a while -acme */
545                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
546                 __sock_put(sk);
547         }
548         return rc;
549 }
550 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
551 
552 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
553 {
554         if (sk_hashed(sk)) {
555                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
556                 return true;
557         }
558         return false;
559 }
560 
561 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
562 {
563         bool rc = __sk_nulls_del_node_init_rcu(sk);
564 
565         if (rc) {
566                 /* paranoid for a while -acme */
567                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
568                 __sock_put(sk);
569         }
570         return rc;
571 }
572 
573 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
574 {
575         hlist_add_head(&sk->sk_node, list);
576 }
577 
578 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
579 {
580         sock_hold(sk);
581         __sk_add_node(sk, list);
582 }
583 
584 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
585 {
586         sock_hold(sk);
587         hlist_add_head_rcu(&sk->sk_node, list);
588 }
589 
590 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
591 {
592         hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
593 }
594 
595 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
596 {
597         sock_hold(sk);
598         __sk_nulls_add_node_rcu(sk, list);
599 }
600 
601 static inline void __sk_del_bind_node(struct sock *sk)
602 {
603         __hlist_del(&sk->sk_bind_node);
604 }
605 
606 static inline void sk_add_bind_node(struct sock *sk,
607                                         struct hlist_head *list)
608 {
609         hlist_add_head(&sk->sk_bind_node, list);
610 }
611 
612 #define sk_for_each(__sk, list) \
613         hlist_for_each_entry(__sk, list, sk_node)
614 #define sk_for_each_rcu(__sk, list) \
615         hlist_for_each_entry_rcu(__sk, list, sk_node)
616 #define sk_nulls_for_each(__sk, node, list) \
617         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
618 #define sk_nulls_for_each_rcu(__sk, node, list) \
619         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
620 #define sk_for_each_from(__sk) \
621         hlist_for_each_entry_from(__sk, sk_node)
622 #define sk_nulls_for_each_from(__sk, node) \
623         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
624                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
625 #define sk_for_each_safe(__sk, tmp, list) \
626         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
627 #define sk_for_each_bound(__sk, list) \
628         hlist_for_each_entry(__sk, list, sk_bind_node)
629 
630 static inline struct user_namespace *sk_user_ns(struct sock *sk)
631 {
632         /* Careful only use this in a context where these parameters
633          * can not change and must all be valid, such as recvmsg from
634          * userspace.
635          */
636         return sk->sk_socket->file->f_cred->user_ns;
637 }
638 
639 /* Sock flags */
640 enum sock_flags {
641         SOCK_DEAD,
642         SOCK_DONE,
643         SOCK_URGINLINE,
644         SOCK_KEEPOPEN,
645         SOCK_LINGER,
646         SOCK_DESTROY,
647         SOCK_BROADCAST,
648         SOCK_TIMESTAMP,
649         SOCK_ZAPPED,
650         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
651         SOCK_DBG, /* %SO_DEBUG setting */
652         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
653         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
654         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
655         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
656         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
657         SOCK_TIMESTAMPING_TX_HARDWARE,  /* %SOF_TIMESTAMPING_TX_HARDWARE */
658         SOCK_TIMESTAMPING_TX_SOFTWARE,  /* %SOF_TIMESTAMPING_TX_SOFTWARE */
659         SOCK_TIMESTAMPING_RX_HARDWARE,  /* %SOF_TIMESTAMPING_RX_HARDWARE */
660         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
661         SOCK_TIMESTAMPING_SOFTWARE,     /* %SOF_TIMESTAMPING_SOFTWARE */
662         SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
663         SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
664         SOCK_FASYNC, /* fasync() active */
665         SOCK_RXQ_OVFL,
666         SOCK_ZEROCOPY, /* buffers from userspace */
667         SOCK_WIFI_STATUS, /* push wifi status to userspace */
668         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
669                      * Will use last 4 bytes of packet sent from
670                      * user-space instead.
671                      */
672         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
673         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
674 };
675 
676 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
677 
678 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
679 {
680         nsk->sk_flags = osk->sk_flags;
681 }
682 
683 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
684 {
685         __set_bit(flag, &sk->sk_flags);
686 }
687 
688 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
689 {
690         __clear_bit(flag, &sk->sk_flags);
691 }
692 
693 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
694 {
695         return test_bit(flag, &sk->sk_flags);
696 }
697 
698 #ifdef CONFIG_NET
699 extern struct static_key memalloc_socks;
700 static inline int sk_memalloc_socks(void)
701 {
702         return static_key_false(&memalloc_socks);
703 }
704 #else
705 
706 static inline int sk_memalloc_socks(void)
707 {
708         return 0;
709 }
710 
711 #endif
712 
713 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
714 {
715         return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
716 }
717 
718 static inline void sk_acceptq_removed(struct sock *sk)
719 {
720         sk->sk_ack_backlog--;
721 }
722 
723 static inline void sk_acceptq_added(struct sock *sk)
724 {
725         sk->sk_ack_backlog++;
726 }
727 
728 static inline bool sk_acceptq_is_full(const struct sock *sk)
729 {
730         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
731 }
732 
733 /*
734  * Compute minimal free write space needed to queue new packets.
735  */
736 static inline int sk_stream_min_wspace(const struct sock *sk)
737 {
738         return sk->sk_wmem_queued >> 1;
739 }
740 
741 static inline int sk_stream_wspace(const struct sock *sk)
742 {
743         return sk->sk_sndbuf - sk->sk_wmem_queued;
744 }
745 
746 extern void sk_stream_write_space(struct sock *sk);
747 
748 static inline bool sk_stream_memory_free(const struct sock *sk)
749 {
750         return sk->sk_wmem_queued < sk->sk_sndbuf;
751 }
752 
753 /* OOB backlog add */
754 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
755 {
756         /* dont let skb dst not refcounted, we are going to leave rcu lock */
757         skb_dst_force(skb);
758 
759         if (!sk->sk_backlog.tail)
760                 sk->sk_backlog.head = skb;
761         else
762                 sk->sk_backlog.tail->next = skb;
763 
764         sk->sk_backlog.tail = skb;
765         skb->next = NULL;
766 }
767 
768 /*
769  * Take into account size of receive queue and backlog queue
770  * Do not take into account this skb truesize,
771  * to allow even a single big packet to come.
772  */
773 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
774                                      unsigned int limit)
775 {
776         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
777 
778         return qsize > limit;
779 }
780 
781 /* The per-socket spinlock must be held here. */
782 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
783                                               unsigned int limit)
784 {
785         if (sk_rcvqueues_full(sk, skb, limit))
786                 return -ENOBUFS;
787 
788         /*
789          * If the skb was allocated from pfmemalloc reserves, only
790          * allow SOCK_MEMALLOC sockets to use it as this socket is
791          * helping free memory
792          */
793         if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
794                 return -ENOMEM;
795 
796         __sk_add_backlog(sk, skb);
797         sk->sk_backlog.len += skb->truesize;
798         return 0;
799 }
800 
801 extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
802 
803 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
804 {
805         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
806                 return __sk_backlog_rcv(sk, skb);
807 
808         return sk->sk_backlog_rcv(sk, skb);
809 }
810 
811 static inline void sock_rps_record_flow(const struct sock *sk)
812 {
813 #ifdef CONFIG_RPS
814         struct rps_sock_flow_table *sock_flow_table;
815 
816         rcu_read_lock();
817         sock_flow_table = rcu_dereference(rps_sock_flow_table);
818         rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
819         rcu_read_unlock();
820 #endif
821 }
822 
823 static inline void sock_rps_reset_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_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
831         rcu_read_unlock();
832 #endif
833 }
834 
835 static inline void sock_rps_save_rxhash(struct sock *sk,
836                                         const struct sk_buff *skb)
837 {
838 #ifdef CONFIG_RPS
839         if (unlikely(sk->sk_rxhash != skb->rxhash)) {
840                 sock_rps_reset_flow(sk);
841                 sk->sk_rxhash = skb->rxhash;
842         }
843 #endif
844 }
845 
846 static inline void sock_rps_reset_rxhash(struct sock *sk)
847 {
848 #ifdef CONFIG_RPS
849         sock_rps_reset_flow(sk);
850         sk->sk_rxhash = 0;
851 #endif
852 }
853 
854 #define sk_wait_event(__sk, __timeo, __condition)                       \
855         ({      int __rc;                                               \
856                 release_sock(__sk);                                     \
857                 __rc = __condition;                                     \
858                 if (!__rc) {                                            \
859                         *(__timeo) = schedule_timeout(*(__timeo));      \
860                 }                                                       \
861                 lock_sock(__sk);                                        \
862                 __rc = __condition;                                     \
863                 __rc;                                                   \
864         })
865 
866 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
867 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
868 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
869 extern int sk_stream_error(struct sock *sk, int flags, int err);
870 extern void sk_stream_kill_queues(struct sock *sk);
871 extern void sk_set_memalloc(struct sock *sk);
872 extern void sk_clear_memalloc(struct sock *sk);
873 
874 extern int sk_wait_data(struct sock *sk, long *timeo);
875 
876 struct request_sock_ops;
877 struct timewait_sock_ops;
878 struct inet_hashinfo;
879 struct raw_hashinfo;
880 struct module;
881 
882 /*
883  * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
884  * un-modified. Special care is taken when initializing object to zero.
885  */
886 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
887 {
888         if (offsetof(struct sock, sk_node.next) != 0)
889                 memset(sk, 0, offsetof(struct sock, sk_node.next));
890         memset(&sk->sk_node.pprev, 0,
891                size - offsetof(struct sock, sk_node.pprev));
892 }
893 
894 /* Networking protocol blocks we attach to sockets.
895  * socket layer -> transport layer interface
896  * transport -> network interface is defined by struct inet_proto
897  */
898 struct proto {
899         void                    (*close)(struct sock *sk,
900                                         long timeout);
901         int                     (*connect)(struct sock *sk,
902                                         struct sockaddr *uaddr,
903                                         int addr_len);
904         int                     (*disconnect)(struct sock *sk, int flags);
905 
906         struct sock *           (*accept)(struct sock *sk, int flags, int *err);
907 
908         int                     (*ioctl)(struct sock *sk, int cmd,
909                                          unsigned long arg);
910         int                     (*init)(struct sock *sk);
911         void                    (*destroy)(struct sock *sk);
912         void                    (*shutdown)(struct sock *sk, int how);
913         int                     (*setsockopt)(struct sock *sk, int level,
914                                         int optname, char __user *optval,
915                                         unsigned int optlen);
916         int                     (*getsockopt)(struct sock *sk, int level,
917                                         int optname, char __user *optval,
918                                         int __user *option);
919 #ifdef CONFIG_COMPAT
920         int                     (*compat_setsockopt)(struct sock *sk,
921                                         int level,
922                                         int optname, char __user *optval,
923                                         unsigned int optlen);
924         int                     (*compat_getsockopt)(struct sock *sk,
925                                         int level,
926                                         int optname, char __user *optval,
927                                         int __user *option);
928         int                     (*compat_ioctl)(struct sock *sk,
929                                         unsigned int cmd, unsigned long arg);
930 #endif
931         int                     (*sendmsg)(struct kiocb *iocb, struct sock *sk,
932                                            struct msghdr *msg, size_t len);
933         int                     (*recvmsg)(struct kiocb *iocb, struct sock *sk,
934                                            struct msghdr *msg,
935                                            size_t len, int noblock, int flags,
936                                            int *addr_len);
937         int                     (*sendpage)(struct sock *sk, struct page *page,
938                                         int offset, size_t size, int flags);
939         int                     (*bind)(struct sock *sk,
940                                         struct sockaddr *uaddr, int addr_len);
941 
942         int                     (*backlog_rcv) (struct sock *sk,
943                                                 struct sk_buff *skb);
944 
945         void            (*release_cb)(struct sock *sk);
946 
947         /* Keeping track of sk's, looking them up, and port selection methods. */
948         void                    (*hash)(struct sock *sk);
949         void                    (*unhash)(struct sock *sk);
950         void                    (*rehash)(struct sock *sk);
951         int                     (*get_port)(struct sock *sk, unsigned short snum);
952         void                    (*clear_sk)(struct sock *sk, int size);
953 
954         /* Keeping track of sockets in use */
955 #ifdef CONFIG_PROC_FS
956         unsigned int            inuse_idx;
957 #endif
958 
959         /* Memory pressure */
960         void                    (*enter_memory_pressure)(struct sock *sk);
961         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
962         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
963         /*
964          * Pressure flag: try to collapse.
965          * Technical note: it is used by multiple contexts non atomically.
966          * All the __sk_mem_schedule() is of this nature: accounting
967          * is strict, actions are advisory and have some latency.
968          */
969         int                     *memory_pressure;
970         long                    *sysctl_mem;
971         int                     *sysctl_wmem;
972         int                     *sysctl_rmem;
973         int                     max_header;
974         bool                    no_autobind;
975 
976         struct kmem_cache       *slab;
977         unsigned int            obj_size;
978         int                     slab_flags;
979 
980         struct percpu_counter   *orphan_count;
981 
982         struct request_sock_ops *rsk_prot;
983         struct timewait_sock_ops *twsk_prot;
984 
985         union {
986                 struct inet_hashinfo    *hashinfo;
987                 struct udp_table        *udp_table;
988                 struct raw_hashinfo     *raw_hash;
989         } h;
990 
991         struct module           *owner;
992 
993         char                    name[32];
994 
995         struct list_head        node;
996 #ifdef SOCK_REFCNT_DEBUG
997         atomic_t                socks;
998 #endif
999 #ifdef CONFIG_MEMCG_KMEM
1000         /*
1001          * cgroup specific init/deinit functions. Called once for all
1002          * protocols that implement it, from cgroups populate function.
1003          * This function has to setup any files the protocol want to
1004          * appear in the kmem cgroup filesystem.
1005          */
1006         int                     (*init_cgroup)(struct mem_cgroup *memcg,
1007                                                struct cgroup_subsys *ss);
1008         void                    (*destroy_cgroup)(struct mem_cgroup *memcg);
1009         struct cg_proto         *(*proto_cgroup)(struct mem_cgroup *memcg);
1010 #endif
1011 };
1012 
1013 /*
1014  * Bits in struct cg_proto.flags
1015  */
1016 enum cg_proto_flags {
1017         /* Currently active and new sockets should be assigned to cgroups */
1018         MEMCG_SOCK_ACTIVE,
1019         /* It was ever activated; we must disarm static keys on destruction */
1020         MEMCG_SOCK_ACTIVATED,
1021 };
1022 
1023 struct cg_proto {
1024         void                    (*enter_memory_pressure)(struct sock *sk);
1025         struct res_counter      *memory_allocated;      /* Current allocated memory. */
1026         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
1027         int                     *memory_pressure;
1028         long                    *sysctl_mem;
1029         unsigned long           flags;
1030         /*
1031          * memcg field is used to find which memcg we belong directly
1032          * Each memcg struct can hold more than one cg_proto, so container_of
1033          * won't really cut.
1034          *
1035          * The elegant solution would be having an inverse function to
1036          * proto_cgroup in struct proto, but that means polluting the structure
1037          * for everybody, instead of just for memcg users.
1038          */
1039         struct mem_cgroup       *memcg;
1040 };
1041 
1042 extern int proto_register(struct proto *prot, int alloc_slab);
1043 extern void proto_unregister(struct proto *prot);
1044 
1045 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1046 {
1047         return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1048 }
1049 
1050 static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1051 {
1052         return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1053 }
1054 
1055 #ifdef SOCK_REFCNT_DEBUG
1056 static inline void sk_refcnt_debug_inc(struct sock *sk)
1057 {
1058         atomic_inc(&sk->sk_prot->socks);
1059 }
1060 
1061 static inline void sk_refcnt_debug_dec(struct sock *sk)
1062 {
1063         atomic_dec(&sk->sk_prot->socks);
1064         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1065                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1066 }
1067 
1068 static inline void sk_refcnt_debug_release(const struct sock *sk)
1069 {
1070         if (atomic_read(&sk->sk_refcnt) != 1)
1071                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1072                        sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1073 }
1074 #else /* SOCK_REFCNT_DEBUG */
1075 #define sk_refcnt_debug_inc(sk) do { } while (0)
1076 #define sk_refcnt_debug_dec(sk) do { } while (0)
1077 #define sk_refcnt_debug_release(sk) do { } while (0)
1078 #endif /* SOCK_REFCNT_DEBUG */
1079 
1080 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1081 extern struct static_key memcg_socket_limit_enabled;
1082 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1083                                                struct cg_proto *cg_proto)
1084 {
1085         return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1086 }
1087 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1088 #else
1089 #define mem_cgroup_sockets_enabled 0
1090 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1091                                                struct cg_proto *cg_proto)
1092 {
1093         return NULL;
1094 }
1095 #endif
1096 
1097 
1098 static inline bool sk_has_memory_pressure(const struct sock *sk)
1099 {
1100         return sk->sk_prot->memory_pressure != NULL;
1101 }
1102 
1103 static inline bool sk_under_memory_pressure(const struct sock *sk)
1104 {
1105         if (!sk->sk_prot->memory_pressure)
1106                 return false;
1107 
1108         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1109                 return !!*sk->sk_cgrp->memory_pressure;
1110 
1111         return !!*sk->sk_prot->memory_pressure;
1112 }
1113 
1114 static inline void sk_leave_memory_pressure(struct sock *sk)
1115 {
1116         int *memory_pressure = sk->sk_prot->memory_pressure;
1117 
1118         if (!memory_pressure)
1119                 return;
1120 
1121         if (*memory_pressure)
1122                 *memory_pressure = 0;
1123 
1124         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1125                 struct cg_proto *cg_proto = sk->sk_cgrp;
1126                 struct proto *prot = sk->sk_prot;
1127 
1128                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1129                         if (*cg_proto->memory_pressure)
1130                                 *cg_proto->memory_pressure = 0;
1131         }
1132 
1133 }
1134 
1135 static inline void sk_enter_memory_pressure(struct sock *sk)
1136 {
1137         if (!sk->sk_prot->enter_memory_pressure)
1138                 return;
1139 
1140         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1141                 struct cg_proto *cg_proto = sk->sk_cgrp;
1142                 struct proto *prot = sk->sk_prot;
1143 
1144                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1145                         cg_proto->enter_memory_pressure(sk);
1146         }
1147 
1148         sk->sk_prot->enter_memory_pressure(sk);
1149 }
1150 
1151 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1152 {
1153         long *prot = sk->sk_prot->sysctl_mem;
1154         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1155                 prot = sk->sk_cgrp->sysctl_mem;
1156         return prot[index];
1157 }
1158 
1159 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1160                                               unsigned long amt,
1161                                               int *parent_status)
1162 {
1163         struct res_counter *fail;
1164         int ret;
1165 
1166         ret = res_counter_charge_nofail(prot->memory_allocated,
1167                                         amt << PAGE_SHIFT, &fail);
1168         if (ret < 0)
1169                 *parent_status = OVER_LIMIT;
1170 }
1171 
1172 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1173                                               unsigned long amt)
1174 {
1175         res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1176 }
1177 
1178 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1179 {
1180         u64 ret;
1181         ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1182         return ret >> PAGE_SHIFT;
1183 }
1184 
1185 static inline long
1186 sk_memory_allocated(const struct sock *sk)
1187 {
1188         struct proto *prot = sk->sk_prot;
1189         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1190                 return memcg_memory_allocated_read(sk->sk_cgrp);
1191 
1192         return atomic_long_read(prot->memory_allocated);
1193 }
1194 
1195 static inline long
1196 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1197 {
1198         struct proto *prot = sk->sk_prot;
1199 
1200         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1201                 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1202                 /* update the root cgroup regardless */
1203                 atomic_long_add_return(amt, prot->memory_allocated);
1204                 return memcg_memory_allocated_read(sk->sk_cgrp);
1205         }
1206 
1207         return atomic_long_add_return(amt, prot->memory_allocated);
1208 }
1209 
1210 static inline void
1211 sk_memory_allocated_sub(struct sock *sk, int amt)
1212 {
1213         struct proto *prot = sk->sk_prot;
1214 
1215         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1216                 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1217 
1218         atomic_long_sub(amt, prot->memory_allocated);
1219 }
1220 
1221 static inline void sk_sockets_allocated_dec(struct sock *sk)
1222 {
1223         struct proto *prot = sk->sk_prot;
1224 
1225         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1226                 struct cg_proto *cg_proto = sk->sk_cgrp;
1227 
1228                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1229                         percpu_counter_dec(cg_proto->sockets_allocated);
1230         }
1231 
1232         percpu_counter_dec(prot->sockets_allocated);
1233 }
1234 
1235 static inline void sk_sockets_allocated_inc(struct sock *sk)
1236 {
1237         struct proto *prot = sk->sk_prot;
1238 
1239         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1240                 struct cg_proto *cg_proto = sk->sk_cgrp;
1241 
1242                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1243                         percpu_counter_inc(cg_proto->sockets_allocated);
1244         }
1245 
1246         percpu_counter_inc(prot->sockets_allocated);
1247 }
1248 
1249 static inline int
1250 sk_sockets_allocated_read_positive(struct sock *sk)
1251 {
1252         struct proto *prot = sk->sk_prot;
1253 
1254         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1255                 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
1256 
1257         return percpu_counter_read_positive(prot->sockets_allocated);
1258 }
1259 
1260 static inline int
1261 proto_sockets_allocated_sum_positive(struct proto *prot)
1262 {
1263         return percpu_counter_sum_positive(prot->sockets_allocated);
1264 }
1265 
1266 static inline long
1267 proto_memory_allocated(struct proto *prot)
1268 {
1269         return atomic_long_read(prot->memory_allocated);
1270 }
1271 
1272 static inline bool
1273 proto_memory_pressure(struct proto *prot)
1274 {
1275         if (!prot->memory_pressure)
1276                 return false;
1277         return !!*prot->memory_pressure;
1278 }
1279 
1280 
1281 #ifdef CONFIG_PROC_FS
1282 /* Called with local bh disabled */
1283 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1284 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1285 #else
1286 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1287                 int inc)
1288 {
1289 }
1290 #endif
1291 
1292 
1293 /* With per-bucket locks this operation is not-atomic, so that
1294  * this version is not worse.
1295  */
1296 static inline void __sk_prot_rehash(struct sock *sk)
1297 {
1298         sk->sk_prot->unhash(sk);
1299         sk->sk_prot->hash(sk);
1300 }
1301 
1302 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1303 
1304 /* About 10 seconds */
1305 #define SOCK_DESTROY_TIME (10*HZ)
1306 
1307 /* Sockets 0-1023 can't be bound to unless you are superuser */
1308 #define PROT_SOCK       1024
1309 
1310 #define SHUTDOWN_MASK   3
1311 #define RCV_SHUTDOWN    1
1312 #define SEND_SHUTDOWN   2
1313 
1314 #define SOCK_SNDBUF_LOCK        1
1315 #define SOCK_RCVBUF_LOCK        2
1316 #define SOCK_BINDADDR_LOCK      4
1317 #define SOCK_BINDPORT_LOCK      8
1318 
1319 /* sock_iocb: used to kick off async processing of socket ios */
1320 struct sock_iocb {
1321         struct list_head        list;
1322 
1323         int                     flags;
1324         int                     size;
1325         struct socket           *sock;
1326         struct sock             *sk;
1327         struct scm_cookie       *scm;
1328         struct msghdr           *msg, async_msg;
1329         struct kiocb            *kiocb;
1330 };
1331 
1332 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1333 {
1334         return (struct sock_iocb *)iocb->private;
1335 }
1336 
1337 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1338 {
1339         return si->kiocb;
1340 }
1341 
1342 struct socket_alloc {
1343         struct socket socket;
1344         struct inode vfs_inode;
1345 };
1346 
1347 static inline struct socket *SOCKET_I(struct inode *inode)
1348 {
1349         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1350 }
1351 
1352 static inline struct inode *SOCK_INODE(struct socket *socket)
1353 {
1354         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1355 }
1356 
1357 /*
1358  * Functions for memory accounting
1359  */
1360 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1361 void __sk_mem_reclaim(struct sock *sk, int amount);
1362 
1363 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1364 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1365 #define SK_MEM_SEND     0
1366 #define SK_MEM_RECV     1
1367 
1368 static inline int sk_mem_pages(int amt)
1369 {
1370         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1371 }
1372 
1373 static inline bool sk_has_account(struct sock *sk)
1374 {
1375         /* return true if protocol supports memory accounting */
1376         return !!sk->sk_prot->memory_allocated;
1377 }
1378 
1379 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1380 {
1381         if (!sk_has_account(sk))
1382                 return true;
1383         return size <= sk->sk_forward_alloc ||
1384                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1385 }
1386 
1387 static inline bool
1388 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1389 {
1390         if (!sk_has_account(sk))
1391                 return true;
1392         return size<= sk->sk_forward_alloc ||
1393                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1394                 skb_pfmemalloc(skb);
1395 }
1396 
1397 static inline void sk_mem_reclaim(struct sock *sk)
1398 {
1399         if (!sk_has_account(sk))
1400                 return;
1401         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1402                 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
1403 }
1404 
1405 static inline void sk_mem_reclaim_partial(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, sk->sk_forward_alloc - 1);
1411 }
1412 
1413 static inline void sk_mem_charge(struct sock *sk, int size)
1414 {
1415         if (!sk_has_account(sk))
1416                 return;
1417         sk->sk_forward_alloc -= size;
1418 }
1419 
1420 static inline void sk_mem_uncharge(struct sock *sk, int size)
1421 {
1422         if (!sk_has_account(sk))
1423                 return;
1424         sk->sk_forward_alloc += size;
1425 
1426         /* Avoid a possible overflow.
1427          * TCP send queues can make this happen, if sk_mem_reclaim()
1428          * is not called and more than 2 GBytes are released at once.
1429          *
1430          * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1431          * no need to hold that much forward allocation anyway.
1432          */
1433         if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1434                 __sk_mem_reclaim(sk, 1 << 20);
1435 }
1436 
1437 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1438 {
1439         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1440         sk->sk_wmem_queued -= skb->truesize;
1441         sk_mem_uncharge(sk, skb->truesize);
1442         __kfree_skb(skb);
1443 }
1444 
1445 /* Used by processes to "lock" a socket state, so that
1446  * interrupts and bottom half handlers won't change it
1447  * from under us. It essentially blocks any incoming
1448  * packets, so that we won't get any new data or any
1449  * packets that change the state of the socket.
1450  *
1451  * While locked, BH processing will add new packets to
1452  * the backlog queue.  This queue is processed by the
1453  * owner of the socket lock right before it is released.
1454  *
1455  * Since ~2.3.5 it is also exclusive sleep lock serializing
1456  * accesses from user process context.
1457  */
1458 #define sock_owned_by_user(sk)  ((sk)->sk_lock.owned)
1459 
1460 static inline void sock_release_ownership(struct sock *sk)
1461 {
1462         sk->sk_lock.owned = 0;
1463 }
1464 
1465 /*
1466  * Macro so as to not evaluate some arguments when
1467  * lockdep is not enabled.
1468  *
1469  * Mark both the sk_lock and the sk_lock.slock as a
1470  * per-address-family lock class.
1471  */
1472 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1473 do {                                                                    \
1474         sk->sk_lock.owned = 0;                                          \
1475         init_waitqueue_head(&sk->sk_lock.wq);                           \
1476         spin_lock_init(&(sk)->sk_lock.slock);                           \
1477         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1478                         sizeof((sk)->sk_lock));                         \
1479         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1480                                 (skey), (sname));                               \
1481         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1482 } while (0)
1483 
1484 extern void lock_sock_nested(struct sock *sk, int subclass);
1485 
1486 static inline void lock_sock(struct sock *sk)
1487 {
1488         lock_sock_nested(sk, 0);
1489 }
1490 
1491 extern void release_sock(struct sock *sk);
1492 
1493 /* BH context may only use the following locking interface. */
1494 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1495 #define bh_lock_sock_nested(__sk) \
1496                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1497                                 SINGLE_DEPTH_NESTING)
1498 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1499 
1500 extern bool lock_sock_fast(struct sock *sk);
1501 /**
1502  * unlock_sock_fast - complement of lock_sock_fast
1503  * @sk: socket
1504  * @slow: slow mode
1505  *
1506  * fast unlock socket for user context.
1507  * If slow mode is on, we call regular release_sock()
1508  */
1509 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1510 {
1511         if (slow)
1512                 release_sock(sk);
1513         else
1514                 spin_unlock_bh(&sk->sk_lock.slock);
1515 }
1516 
1517 
1518 extern struct sock              *sk_alloc(struct net *net, int family,
1519                                           gfp_t priority,
1520                                           struct proto *prot);
1521 extern void                     sk_free(struct sock *sk);
1522 extern void                     sk_release_kernel(struct sock *sk);
1523 extern struct sock              *sk_clone_lock(const struct sock *sk,
1524                                                const gfp_t priority);
1525 
1526 extern struct sk_buff           *sock_wmalloc(struct sock *sk,
1527                                               unsigned long size, int force,
1528                                               gfp_t priority);
1529 extern struct sk_buff           *sock_rmalloc(struct sock *sk,
1530                                               unsigned long size, int force,
1531                                               gfp_t priority);
1532 extern void                     sock_wfree(struct sk_buff *skb);
1533 extern void                     sock_rfree(struct sk_buff *skb);
1534 extern void                     sock_edemux(struct sk_buff *skb);
1535 
1536 extern int                      sock_setsockopt(struct socket *sock, int level,
1537                                                 int op, char __user *optval,
1538                                                 unsigned int optlen);
1539 
1540 extern int                      sock_getsockopt(struct socket *sock, int level,
1541                                                 int op, char __user *optval,
1542                                                 int __user *optlen);
1543 extern struct sk_buff           *sock_alloc_send_skb(struct sock *sk,
1544                                                      unsigned long size,
1545                                                      int noblock,
1546                                                      int *errcode);
1547 extern struct sk_buff           *sock_alloc_send_pskb(struct sock *sk,
1548                                                       unsigned long header_len,
1549                                                       unsigned long data_len,
1550                                                       int noblock,
1551                                                       int *errcode);
1552 extern void *sock_kmalloc(struct sock *sk, int size,
1553                           gfp_t priority);
1554 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1555 extern void sk_send_sigurg(struct sock *sk);
1556 
1557 /*
1558  * Functions to fill in entries in struct proto_ops when a protocol
1559  * does not implement a particular function.
1560  */
1561 extern int                      sock_no_bind(struct socket *,
1562                                              struct sockaddr *, int);
1563 extern int                      sock_no_connect(struct socket *,
1564                                                 struct sockaddr *, int, int);
1565 extern int                      sock_no_socketpair(struct socket *,
1566                                                    struct socket *);
1567 extern int                      sock_no_accept(struct socket *,
1568                                                struct socket *, int);
1569 extern int                      sock_no_getname(struct socket *,
1570                                                 struct sockaddr *, int *, int);
1571 extern unsigned int             sock_no_poll(struct file *, struct socket *,
1572                                              struct poll_table_struct *);
1573 extern int                      sock_no_ioctl(struct socket *, unsigned int,
1574                                               unsigned long);
1575 extern int                      sock_no_listen(struct socket *, int);
1576 extern int                      sock_no_shutdown(struct socket *, int);
1577 extern int                      sock_no_getsockopt(struct socket *, int , int,
1578                                                    char __user *, int __user *);
1579 extern int                      sock_no_setsockopt(struct socket *, int, int,
1580                                                    char __user *, unsigned int);
1581 extern int                      sock_no_sendmsg(struct kiocb *, struct socket *,
1582                                                 struct msghdr *, size_t);
1583 extern int                      sock_no_recvmsg(struct kiocb *, struct socket *,
1584                                                 struct msghdr *, size_t, int);
1585 extern int                      sock_no_mmap(struct file *file,
1586                                              struct socket *sock,
1587                                              struct vm_area_struct *vma);
1588 extern ssize_t                  sock_no_sendpage(struct socket *sock,
1589                                                 struct page *page,
1590                                                 int offset, size_t size,
1591                                                 int flags);
1592 
1593 /*
1594  * Functions to fill in entries in struct proto_ops when a protocol
1595  * uses the inet style.
1596  */
1597 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1598                                   char __user *optval, int __user *optlen);
1599 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1600                                struct msghdr *msg, size_t size, int flags);
1601 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1602                                   char __user *optval, unsigned int optlen);
1603 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1604                 int optname, char __user *optval, int __user *optlen);
1605 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1606                 int optname, char __user *optval, unsigned int optlen);
1607 
1608 extern void sk_common_release(struct sock *sk);
1609 
1610 /*
1611  *      Default socket callbacks and setup code
1612  */
1613 
1614 /* Initialise core socket variables */
1615 extern void sock_init_data(struct socket *sock, struct sock *sk);
1616 
1617 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1618 
1619 /**
1620  *      sk_filter_release - release a socket filter
1621  *      @fp: filter to remove
1622  *
1623  *      Remove a filter from a socket and release its resources.
1624  */
1625 
1626 static inline void sk_filter_release(struct sk_filter *fp)
1627 {
1628         if (atomic_dec_and_test(&fp->refcnt))
1629                 call_rcu(&fp->rcu, sk_filter_release_rcu);
1630 }
1631 
1632 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1633 {
1634         unsigned int size = sk_filter_len(fp);
1635 
1636         atomic_sub(size, &sk->sk_omem_alloc);
1637         sk_filter_release(fp);
1638 }
1639 
1640 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1641 {
1642         atomic_inc(&fp->refcnt);
1643         atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1644 }
1645 
1646 /*
1647  * Socket reference counting postulates.
1648  *
1649  * * Each user of socket SHOULD hold a reference count.
1650  * * Each access point to socket (an hash table bucket, reference from a list,
1651  *   running timer, skb in flight MUST hold a reference count.
1652  * * When reference count hits 0, it means it will never increase back.
1653  * * When reference count hits 0, it means that no references from
1654  *   outside exist to this socket and current process on current CPU
1655  *   is last user and may/should destroy this socket.
1656  * * sk_free is called from any context: process, BH, IRQ. When
1657  *   it is called, socket has no references from outside -> sk_free
1658  *   may release descendant resources allocated by the socket, but
1659  *   to the time when it is called, socket is NOT referenced by any
1660  *   hash tables, lists etc.
1661  * * Packets, delivered from outside (from network or from another process)
1662  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1663  *   when they sit in queue. Otherwise, packets will leak to hole, when
1664  *   socket is looked up by one cpu and unhasing is made by another CPU.
1665  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1666  *   (leak to backlog). Packet socket does all the processing inside
1667  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1668  *   use separate SMP lock, so that they are prone too.
1669  */
1670 
1671 /* Ungrab socket and destroy it, if it was the last reference. */
1672 static inline void sock_put(struct sock *sk)
1673 {
1674         if (atomic_dec_and_test(&sk->sk_refcnt))
1675                 sk_free(sk);
1676 }
1677 
1678 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1679                           const int nested);
1680 
1681 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1682 {
1683         sk->sk_tx_queue_mapping = tx_queue;
1684 }
1685 
1686 static inline void sk_tx_queue_clear(struct sock *sk)
1687 {
1688         sk->sk_tx_queue_mapping = -1;
1689 }
1690 
1691 static inline int sk_tx_queue_get(const struct sock *sk)
1692 {
1693         return sk ? sk->sk_tx_queue_mapping : -1;
1694 }
1695 
1696 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1697 {
1698         sk_tx_queue_clear(sk);
1699         sk->sk_socket = sock;
1700 }
1701 
1702 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1703 {
1704         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1705         return &rcu_dereference_raw(sk->sk_wq)->wait;
1706 }
1707 /* Detach socket from process context.
1708  * Announce socket dead, detach it from wait queue and inode.
1709  * Note that parent inode held reference count on this struct sock,
1710  * we do not release it in this function, because protocol
1711  * probably wants some additional cleanups or even continuing
1712  * to work with this socket (TCP).
1713  */
1714 static inline void sock_orphan(struct sock *sk)
1715 {
1716         write_lock_bh(&sk->sk_callback_lock);
1717         sock_set_flag(sk, SOCK_DEAD);
1718         sk_set_socket(sk, NULL);
1719         sk->sk_wq  = NULL;
1720         write_unlock_bh(&sk->sk_callback_lock);
1721 }
1722 
1723 static inline void sock_graft(struct sock *sk, struct socket *parent)
1724 {
1725         write_lock_bh(&sk->sk_callback_lock);
1726         sk->sk_wq = parent->wq;
1727         parent->sk = sk;
1728         sk_set_socket(sk, parent);
1729         security_sock_graft(sk, parent);
1730         write_unlock_bh(&sk->sk_callback_lock);
1731 }
1732 
1733 extern kuid_t sock_i_uid(struct sock *sk);
1734 extern unsigned long sock_i_ino(struct sock *sk);
1735 
1736 static inline struct dst_entry *
1737 __sk_dst_get(struct sock *sk)
1738 {
1739         return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1740                                                        lockdep_is_held(&sk->sk_lock.slock));
1741 }
1742 
1743 static inline struct dst_entry *
1744 sk_dst_get(struct sock *sk)
1745 {
1746         struct dst_entry *dst;
1747 
1748         rcu_read_lock();
1749         dst = rcu_dereference(sk->sk_dst_cache);
1750         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1751                 dst = NULL;
1752         rcu_read_unlock();
1753         return dst;
1754 }
1755 
1756 extern void sk_reset_txq(struct sock *sk);
1757 
1758 static inline void dst_negative_advice(struct sock *sk)
1759 {
1760         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1761 
1762         if (dst && dst->ops->negative_advice) {
1763                 ndst = dst->ops->negative_advice(dst);
1764 
1765                 if (ndst != dst) {
1766                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1767                         sk_reset_txq(sk);
1768                 }
1769         }
1770 }
1771 
1772 static inline void
1773 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1774 {
1775         struct dst_entry *old_dst;
1776 
1777         sk_tx_queue_clear(sk);
1778         /*
1779          * This can be called while sk is owned by the caller only,
1780          * with no state that can be checked in a rcu_dereference_check() cond
1781          */
1782         old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1783         rcu_assign_pointer(sk->sk_dst_cache, dst);
1784         dst_release(old_dst);
1785 }
1786 
1787 static inline void
1788 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1789 {
1790         struct dst_entry *old_dst;
1791 
1792         sk_tx_queue_clear(sk);
1793         old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1794         dst_release(old_dst);
1795 }
1796 
1797 static inline void
1798 __sk_dst_reset(struct sock *sk)
1799 {
1800         __sk_dst_set(sk, NULL);
1801 }
1802 
1803 static inline void
1804 sk_dst_reset(struct sock *sk)
1805 {
1806         sk_dst_set(sk, NULL);
1807 }
1808 
1809 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1810 
1811 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1812 
1813 static inline bool sk_can_gso(const struct sock *sk)
1814 {
1815         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1816 }
1817 
1818 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1819 
1820 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1821 {
1822         sk->sk_route_nocaps |= flags;
1823         sk->sk_route_caps &= ~flags;
1824 }
1825 
1826 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1827                                            char __user *from, char *to,
1828                                            int copy, int offset)
1829 {
1830         if (skb->ip_summed == CHECKSUM_NONE) {
1831                 int err = 0;
1832                 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1833                 if (err)
1834                         return err;
1835                 skb->csum = csum_block_add(skb->csum, csum, offset);
1836         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1837                 if (!access_ok(VERIFY_READ, from, copy) ||
1838                     __copy_from_user_nocache(to, from, copy))
1839                         return -EFAULT;
1840         } else if (copy_from_user(to, from, copy))
1841                 return -EFAULT;
1842 
1843         return 0;
1844 }
1845 
1846 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1847                                        char __user *from, int copy)
1848 {
1849         int err, offset = skb->len;
1850 
1851         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1852                                        copy, offset);
1853         if (err)
1854                 __skb_trim(skb, offset);
1855 
1856         return err;
1857 }
1858 
1859 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1860                                            struct sk_buff *skb,
1861                                            struct page *page,
1862                                            int off, int copy)
1863 {
1864         int err;
1865 
1866         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1867                                        copy, skb->len);
1868         if (err)
1869                 return err;
1870 
1871         skb->len             += copy;
1872         skb->data_len        += copy;
1873         skb->truesize        += copy;
1874         sk->sk_wmem_queued   += copy;
1875         sk_mem_charge(sk, copy);
1876         return 0;
1877 }
1878 
1879 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1880                                    struct sk_buff *skb, struct page *page,
1881                                    int off, int copy)
1882 {
1883         if (skb->ip_summed == CHECKSUM_NONE) {
1884                 int err = 0;
1885                 __wsum csum = csum_and_copy_from_user(from,
1886                                                      page_address(page) + off,
1887                                                             copy, 0, &err);
1888                 if (err)
1889                         return err;
1890                 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1891         } else if (copy_from_user(page_address(page) + off, from, copy))
1892                 return -EFAULT;
1893 
1894         skb->len             += copy;
1895         skb->data_len        += copy;
1896         skb->truesize        += copy;
1897         sk->sk_wmem_queued   += copy;
1898         sk_mem_charge(sk, copy);
1899         return 0;
1900 }
1901 
1902 /**
1903  * sk_wmem_alloc_get - returns write allocations
1904  * @sk: socket
1905  *
1906  * Returns sk_wmem_alloc minus initial offset of one
1907  */
1908 static inline int sk_wmem_alloc_get(const struct sock *sk)
1909 {
1910         return atomic_read(&sk->sk_wmem_alloc) - 1;
1911 }
1912 
1913 /**
1914  * sk_rmem_alloc_get - returns read allocations
1915  * @sk: socket
1916  *
1917  * Returns sk_rmem_alloc
1918  */
1919 static inline int sk_rmem_alloc_get(const struct sock *sk)
1920 {
1921         return atomic_read(&sk->sk_rmem_alloc);
1922 }
1923 
1924 /**
1925  * sk_has_allocations - check if allocations are outstanding
1926  * @sk: socket
1927  *
1928  * Returns true if socket has write or read allocations
1929  */
1930 static inline bool sk_has_allocations(const struct sock *sk)
1931 {
1932         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1933 }
1934 
1935 /**
1936  * wq_has_sleeper - check if there are any waiting processes
1937  * @wq: struct socket_wq
1938  *
1939  * Returns true if socket_wq has waiting processes
1940  *
1941  * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1942  * barrier call. They were added due to the race found within the tcp code.
1943  *
1944  * Consider following tcp code paths:
1945  *
1946  * CPU1                  CPU2
1947  *
1948  * sys_select            receive packet
1949  *   ...                 ...
1950  *   __add_wait_queue    update tp->rcv_nxt
1951  *   ...                 ...
1952  *   tp->rcv_nxt check   sock_def_readable
1953  *   ...                 {
1954  *   schedule               rcu_read_lock();
1955  *                          wq = rcu_dereference(sk->sk_wq);
1956  *                          if (wq && waitqueue_active(&wq->wait))
1957  *                              wake_up_interruptible(&wq->wait)
1958  *                          ...
1959  *                       }
1960  *
1961  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1962  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1963  * could then endup calling schedule and sleep forever if there are no more
1964  * data on the socket.
1965  *
1966  */
1967 static inline bool wq_has_sleeper(struct socket_wq *wq)
1968 {
1969         /* We need to be sure we are in sync with the
1970          * add_wait_queue modifications to the wait queue.
1971          *
1972          * This memory barrier is paired in the sock_poll_wait.
1973          */
1974         smp_mb();
1975         return wq && waitqueue_active(&wq->wait);
1976 }
1977 
1978 /**
1979  * sock_poll_wait - place memory barrier behind the poll_wait call.
1980  * @filp:           file
1981  * @wait_address:   socket wait queue
1982  * @p:              poll_table
1983  *
1984  * See the comments in the wq_has_sleeper function.
1985  */
1986 static inline void sock_poll_wait(struct file *filp,
1987                 wait_queue_head_t *wait_address, poll_table *p)
1988 {
1989         if (!poll_does_not_wait(p) && wait_address) {
1990                 poll_wait(filp, wait_address, p);
1991                 /* We need to be sure we are in sync with the
1992                  * socket flags modification.
1993                  *
1994                  * This memory barrier is paired in the wq_has_sleeper.
1995                  */
1996                 smp_mb();
1997         }
1998 }
1999 
2000 /*
2001  *      Queue a received datagram if it will fit. Stream and sequenced
2002  *      protocols can't normally use this as they need to fit buffers in
2003  *      and play with them.
2004  *
2005  *      Inlined as it's very short and called for pretty much every
2006  *      packet ever received.
2007  */
2008 
2009 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2010 {
2011         skb_orphan(skb);
2012         skb->sk = sk;
2013         skb->destructor = sock_wfree;
2014         /*
2015          * We used to take a refcount on sk, but following operation
2016          * is enough to guarantee sk_free() wont free this sock until
2017          * all in-flight packets are completed
2018          */
2019         atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2020 }
2021 
2022 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2023 {
2024         skb_orphan(skb);
2025         skb->sk = sk;
2026         skb->destructor = sock_rfree;
2027         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2028         sk_mem_charge(sk, skb->truesize);
2029 }
2030 
2031 extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2032                            unsigned long expires);
2033 
2034 extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2035 
2036 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2037 
2038 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2039 
2040 /*
2041  *      Recover an error report and clear atomically
2042  */
2043 
2044 static inline int sock_error(struct sock *sk)
2045 {
2046         int err;
2047         if (likely(!sk->sk_err))
2048                 return 0;
2049         err = xchg(&sk->sk_err, 0);
2050         return -err;
2051 }
2052 
2053 static inline unsigned long sock_wspace(struct sock *sk)
2054 {
2055         int amt = 0;
2056 
2057         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2058                 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2059                 if (amt < 0)
2060                         amt = 0;
2061         }
2062         return amt;
2063 }
2064 
2065 static inline void sk_wake_async(struct sock *sk, int how, int band)
2066 {
2067         if (sock_flag(sk, SOCK_FASYNC))
2068                 sock_wake_async(sk->sk_socket, how, band);
2069 }
2070 
2071 #define SOCK_MIN_SNDBUF 2048
2072 /*
2073  * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2074  * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2075  */
2076 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
2077 
2078 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2079 {
2080         if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2081                 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2082                 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2083         }
2084 }
2085 
2086 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2087 
2088 /**
2089  * sk_page_frag - return an appropriate page_frag
2090  * @sk: socket
2091  *
2092  * If socket allocation mode allows current thread to sleep, it means its
2093  * safe to use the per task page_frag instead of the per socket one.
2094  */
2095 static inline struct page_frag *sk_page_frag(struct sock *sk)
2096 {
2097         if (sk->sk_allocation & __GFP_WAIT)
2098                 return &current->task_frag;
2099 
2100         return &sk->sk_frag;
2101 }
2102 
2103 extern bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2104 
2105 /*
2106  *      Default write policy as shown to user space via poll/select/SIGIO
2107  */
2108 static inline bool sock_writeable(const struct sock *sk)
2109 {
2110         return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2111 }
2112 
2113 static inline gfp_t gfp_any(void)
2114 {
2115         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2116 }
2117 
2118 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2119 {
2120         return noblock ? 0 : sk->sk_rcvtimeo;
2121 }
2122 
2123 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2124 {
2125         return noblock ? 0 : sk->sk_sndtimeo;
2126 }
2127 
2128 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2129 {
2130         return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2131 }
2132 
2133 /* Alas, with timeout socket operations are not restartable.
2134  * Compare this to poll().
2135  */
2136 static inline int sock_intr_errno(long timeo)
2137 {
2138         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2139 }
2140 
2141 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2142         struct sk_buff *skb);
2143 extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2144         struct sk_buff *skb);
2145 
2146 static inline void
2147 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2148 {
2149         ktime_t kt = skb->tstamp;
2150         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2151 
2152         /*
2153          * generate control messages if
2154          * - receive time stamping in software requested (SOCK_RCVTSTAMP
2155          *   or SOCK_TIMESTAMPING_RX_SOFTWARE)
2156          * - software time stamp available and wanted
2157          *   (SOCK_TIMESTAMPING_SOFTWARE)
2158          * - hardware time stamps available and wanted
2159          *   (SOCK_TIMESTAMPING_SYS_HARDWARE or
2160          *   SOCK_TIMESTAMPING_RAW_HARDWARE)
2161          */
2162         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2163             sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2164             (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2165             (hwtstamps->hwtstamp.tv64 &&
2166              sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2167             (hwtstamps->syststamp.tv64 &&
2168              sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2169                 __sock_recv_timestamp(msg, sk, skb);
2170         else
2171                 sk->sk_stamp = kt;
2172 
2173         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2174                 __sock_recv_wifi_status(msg, sk, skb);
2175 }
2176 
2177 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2178                                      struct sk_buff *skb);
2179 
2180 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2181                                           struct sk_buff *skb)
2182 {
2183 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2184                            (1UL << SOCK_RCVTSTAMP)                      | \
2185                            (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)       | \
2186                            (1UL << SOCK_TIMESTAMPING_SOFTWARE)          | \
2187                            (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE)      | \
2188                            (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2189 
2190         if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2191                 __sock_recv_ts_and_drops(msg, sk, skb);
2192         else
2193                 sk->sk_stamp = skb->tstamp;
2194 }
2195 
2196 /**
2197  * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2198  * @sk:         socket sending this packet
2199  * @tx_flags:   filled with instructions for time stamping
2200  *
2201  * Currently only depends on SOCK_TIMESTAMPING* flags.
2202  */
2203 extern void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2204 
2205 /**
2206  * sk_eat_skb - Release a skb if it is no longer needed
2207  * @sk: socket to eat this skb from
2208  * @skb: socket buffer to eat
2209  * @copied_early: flag indicating whether DMA operations copied this data early
2210  *
2211  * This routine must be called with interrupts disabled or with the socket
2212  * locked so that the sk_buff queue operation is ok.
2213 */
2214 #ifdef CONFIG_NET_DMA
2215 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2216 {
2217         __skb_unlink(skb, &sk->sk_receive_queue);
2218         if (!copied_early)
2219                 __kfree_skb(skb);
2220         else
2221                 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2222 }
2223 #else
2224 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2225 {
2226         __skb_unlink(skb, &sk->sk_receive_queue);
2227         __kfree_skb(skb);
2228 }
2229 #endif
2230 
2231 static inline
2232 struct net *sock_net(const struct sock *sk)
2233 {
2234         return read_pnet(&sk->sk_net);
2235 }
2236 
2237 static inline
2238 void sock_net_set(struct sock *sk, struct net *net)
2239 {
2240         write_pnet(&sk->sk_net, net);
2241 }
2242 
2243 /*
2244  * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2245  * They should not hold a reference to a namespace in order to allow
2246  * to stop it.
2247  * Sockets after sk_change_net should be released using sk_release_kernel
2248  */
2249 static inline void sk_change_net(struct sock *sk, struct net *net)
2250 {
2251         put_net(sock_net(sk));
2252         sock_net_set(sk, hold_net(net));
2253 }
2254 
2255 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2256 {
2257         if (skb->sk) {
2258                 struct sock *sk = skb->sk;
2259 
2260                 skb->destructor = NULL;
2261                 skb->sk = NULL;
2262                 return sk;
2263         }
2264         return NULL;
2265 }
2266 
2267 extern void sock_enable_timestamp(struct sock *sk, int flag);
2268 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2269 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2270 
2271 bool sk_ns_capable(const struct sock *sk,
2272                    struct user_namespace *user_ns, int cap);
2273 bool sk_capable(const struct sock *sk, int cap);
2274 bool sk_net_capable(const struct sock *sk, int cap);
2275 
2276 /*
2277  *      Enable debug/info messages
2278  */
2279 extern int net_msg_warn;
2280 #define NETDEBUG(fmt, args...) \
2281         do { if (net_msg_warn) printk(fmt,##args); } while (0)
2282 
2283 #define LIMIT_NETDEBUG(fmt, args...) \
2284         do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2285 
2286 extern __u32 sysctl_wmem_max;
2287 extern __u32 sysctl_rmem_max;
2288 
2289 extern int sysctl_optmem_max;
2290 
2291 extern __u32 sysctl_wmem_default;
2292 extern __u32 sysctl_rmem_default;
2293 
2294 #endif  /* _SOCK_H */
2295 

~ [ 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