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

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

Version: ~ [ linux-5.7 ] ~ [ linux-5.6.15 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.43 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.125 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.182 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.225 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.225 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.84 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
  3  *              operating system.  INET is implemented using the  BSD Socket
  4  *              interface as the means of communication with the user level.
  5  *
  6  *              Definitions for the AF_INET socket handler.
  7  *
  8  * Version:     @(#)sock.h      1.0.4   05/13/93
  9  *
 10  * Authors:     Ross Biro
 11  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 12  *              Corey Minyard <wf-rch!minyard@relay.EU.net>
 13  *              Florian La Roche <flla@stud.uni-sb.de>
 14  *
 15  * Fixes:
 16  *              Alan Cox        :       Volatiles in skbuff pointers. See
 17  *                                      skbuff comments. May be overdone,
 18  *                                      better to prove they can be removed
 19  *                                      than the reverse.
 20  *              Alan Cox        :       Added a zapped field for tcp to note
 21  *                                      a socket is reset and must stay shut up
 22  *              Alan Cox        :       New fields for options
 23  *      Pauline Middelink       :       identd support
 24  *              Alan Cox        :       Eliminate low level recv/recvfrom
 25  *              David S. Miller :       New socket lookup architecture.
 26  *              Steve Whitehouse:       Default routines for sock_ops
 27  *              Arnaldo C. Melo :       removed net_pinfo, tp_pinfo and made
 28  *                                      protinfo be just a void pointer, as the
 29  *                                      protocol specific parts were moved to
 30  *                                      respective headers and ipv4/v6, etc now
 31  *                                      use private slabcaches for its socks
 32  *              Pedro Hortas    :       New flags field for socket options
 33  *
 34  *
 35  *              This program is free software; you can redistribute it and/or
 36  *              modify it under the terms of the GNU General Public License
 37  *              as published by the Free Software Foundation; either version
 38  *              2 of the License, or (at your option) any later version.
 39  */
 40 #ifndef _SOCK_H
 41 #define _SOCK_H
 42 
 43 #include <linux/hardirq.h>
 44 #include <linux/kernel.h>
 45 #include <linux/list.h>
 46 #include <linux/list_nulls.h>
 47 #include <linux/timer.h>
 48 #include <linux/cache.h>
 49 #include <linux/bitops.h>
 50 #include <linux/lockdep.h>
 51 #include <linux/netdevice.h>
 52 #include <linux/skbuff.h>       /* struct sk_buff */
 53 #include <linux/mm.h>
 54 #include <linux/security.h>
 55 #include <linux/slab.h>
 56 #include <linux/uaccess.h>
 57 #include <linux/memcontrol.h>
 58 #include <linux/res_counter.h>
 59 #include <linux/static_key.h>
 60 #include <linux/aio.h>
 61 #include <linux/sched.h>
 62 
 63 #include <linux/filter.h>
 64 #include <linux/rculist_nulls.h>
 65 #include <linux/poll.h>
 66 
 67 #include <linux/atomic.h>
 68 #include <net/dst.h>
 69 #include <net/checksum.h>
 70 #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_async_wait_queue: DMA copied packets
237   *     @sk_omem_alloc: "o" is "option" or "other"
238   *     @sk_wmem_queued: persistent queue size
239   *     @sk_forward_alloc: space allocated forward
240   *     @sk_napi_id: id of the last napi context to receive data for sk
241   *     @sk_ll_usec: usecs to busypoll when there is no data
242   *     @sk_allocation: allocation mode
243   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
244   *     @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
245   *     @sk_sndbuf: size of send buffer in bytes
246   *     @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
247   *                %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
248   *     @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
249   *     @sk_no_check_rx: allow zero checksum in RX packets
250   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
251   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
252   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
253   *     @sk_gso_max_size: Maximum GSO segment size to build
254   *     @sk_gso_max_segs: Maximum number of GSO segments
255   *     @sk_lingertime: %SO_LINGER l_linger setting
256   *     @sk_backlog: always used with the per-socket spinlock held
257   *     @sk_callback_lock: used with the callbacks in the end of this struct
258   *     @sk_error_queue: rarely used
259   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
260   *                       IPV6_ADDRFORM for instance)
261   *     @sk_err: last error
262   *     @sk_err_soft: errors that don't cause failure but are the cause of a
263   *                   persistent failure not just 'timed out'
264   *     @sk_drops: raw/udp drops counter
265   *     @sk_ack_backlog: current listen backlog
266   *     @sk_max_ack_backlog: listen backlog set in listen()
267   *     @sk_priority: %SO_PRIORITY setting
268   *     @sk_cgrp_prioidx: socket group's priority map index
269   *     @sk_type: socket type (%SOCK_STREAM, etc)
270   *     @sk_protocol: which protocol this socket belongs in this network family
271   *     @sk_peer_pid: &struct pid for this socket's peer
272   *     @sk_peer_cred: %SO_PEERCRED setting
273   *     @sk_rcvlowat: %SO_RCVLOWAT setting
274   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
275   *     @sk_sndtimeo: %SO_SNDTIMEO setting
276   *     @sk_rxhash: flow hash received from netif layer
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         __u32                   sk_txhash;
355 #ifdef CONFIG_NET_RX_BUSY_POLL
356         unsigned int            sk_napi_id;
357         unsigned int            sk_ll_usec;
358 #endif
359         atomic_t                sk_drops;
360         int                     sk_rcvbuf;
361 
362         struct sk_filter __rcu  *sk_filter;
363         struct socket_wq __rcu  *sk_wq;
364 
365 #ifdef CONFIG_NET_DMA
366         struct sk_buff_head     sk_async_wait_queue;
367 #endif
368 
369 #ifdef CONFIG_XFRM
370         struct xfrm_policy      *sk_policy[2];
371 #endif
372         unsigned long           sk_flags;
373         struct dst_entry        *sk_rx_dst;
374         struct dst_entry __rcu  *sk_dst_cache;
375         spinlock_t              sk_dst_lock;
376         atomic_t                sk_wmem_alloc;
377         atomic_t                sk_omem_alloc;
378         int                     sk_sndbuf;
379         struct sk_buff_head     sk_write_queue;
380         kmemcheck_bitfield_begin(flags);
381         unsigned int            sk_shutdown  : 2,
382                                 sk_no_check_tx : 1,
383                                 sk_no_check_rx : 1,
384                                 sk_userlocks : 4,
385                                 sk_protocol  : 8,
386                                 sk_type      : 16;
387         kmemcheck_bitfield_end(flags);
388         int                     sk_wmem_queued;
389         gfp_t                   sk_allocation;
390         u32                     sk_pacing_rate; /* bytes per second */
391         u32                     sk_max_pacing_rate;
392         netdev_features_t       sk_route_caps;
393         netdev_features_t       sk_route_nocaps;
394         int                     sk_gso_type;
395         unsigned int            sk_gso_max_size;
396         u16                     sk_gso_max_segs;
397         int                     sk_rcvlowat;
398         unsigned long           sk_lingertime;
399         struct sk_buff_head     sk_error_queue;
400         struct proto            *sk_prot_creator;
401         rwlock_t                sk_callback_lock;
402         int                     sk_err,
403                                 sk_err_soft;
404         unsigned short          sk_ack_backlog;
405         unsigned short          sk_max_ack_backlog;
406         __u32                   sk_priority;
407 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
408         __u32                   sk_cgrp_prioidx;
409 #endif
410         struct pid              *sk_peer_pid;
411         const struct cred       *sk_peer_cred;
412         long                    sk_rcvtimeo;
413         long                    sk_sndtimeo;
414         void                    *sk_protinfo;
415         struct timer_list       sk_timer;
416         ktime_t                 sk_stamp;
417         u16                     sk_tsflags;
418         u32                     sk_tskey;
419         struct socket           *sk_socket;
420         void                    *sk_user_data;
421         struct page_frag        sk_frag;
422         struct sk_buff          *sk_send_head;
423         __s32                   sk_peek_off;
424         int                     sk_write_pending;
425 #ifdef CONFIG_SECURITY
426         void                    *sk_security;
427 #endif
428         __u32                   sk_mark;
429         u32                     sk_classid;
430         struct cg_proto         *sk_cgrp;
431         void                    (*sk_state_change)(struct sock *sk);
432         void                    (*sk_data_ready)(struct sock *sk);
433         void                    (*sk_write_space)(struct sock *sk);
434         void                    (*sk_error_report)(struct sock *sk);
435         int                     (*sk_backlog_rcv)(struct sock *sk,
436                                                   struct sk_buff *skb);
437         void                    (*sk_destruct)(struct sock *sk);
438 };
439 
440 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
441 
442 #define rcu_dereference_sk_user_data(sk)        rcu_dereference(__sk_user_data((sk)))
443 #define rcu_assign_sk_user_data(sk, ptr)        rcu_assign_pointer(__sk_user_data((sk)), ptr)
444 
445 /*
446  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
447  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
448  * on a socket means that the socket will reuse everybody else's port
449  * without looking at the other's sk_reuse value.
450  */
451 
452 #define SK_NO_REUSE     0
453 #define SK_CAN_REUSE    1
454 #define SK_FORCE_REUSE  2
455 
456 static inline int sk_peek_offset(struct sock *sk, int flags)
457 {
458         if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
459                 return sk->sk_peek_off;
460         else
461                 return 0;
462 }
463 
464 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
465 {
466         if (sk->sk_peek_off >= 0) {
467                 if (sk->sk_peek_off >= val)
468                         sk->sk_peek_off -= val;
469                 else
470                         sk->sk_peek_off = 0;
471         }
472 }
473 
474 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
475 {
476         if (sk->sk_peek_off >= 0)
477                 sk->sk_peek_off += val;
478 }
479 
480 /*
481  * Hashed lists helper routines
482  */
483 static inline struct sock *sk_entry(const struct hlist_node *node)
484 {
485         return hlist_entry(node, struct sock, sk_node);
486 }
487 
488 static inline struct sock *__sk_head(const struct hlist_head *head)
489 {
490         return hlist_entry(head->first, struct sock, sk_node);
491 }
492 
493 static inline struct sock *sk_head(const struct hlist_head *head)
494 {
495         return hlist_empty(head) ? NULL : __sk_head(head);
496 }
497 
498 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
499 {
500         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
501 }
502 
503 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
504 {
505         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
506 }
507 
508 static inline struct sock *sk_next(const struct sock *sk)
509 {
510         return sk->sk_node.next ?
511                 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
512 }
513 
514 static inline struct sock *sk_nulls_next(const struct sock *sk)
515 {
516         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
517                 hlist_nulls_entry(sk->sk_nulls_node.next,
518                                   struct sock, sk_nulls_node) :
519                 NULL;
520 }
521 
522 static inline bool sk_unhashed(const struct sock *sk)
523 {
524         return hlist_unhashed(&sk->sk_node);
525 }
526 
527 static inline bool sk_hashed(const struct sock *sk)
528 {
529         return !sk_unhashed(sk);
530 }
531 
532 static inline void sk_node_init(struct hlist_node *node)
533 {
534         node->pprev = NULL;
535 }
536 
537 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
538 {
539         node->pprev = NULL;
540 }
541 
542 static inline void __sk_del_node(struct sock *sk)
543 {
544         __hlist_del(&sk->sk_node);
545 }
546 
547 /* NB: equivalent to hlist_del_init_rcu */
548 static inline bool __sk_del_node_init(struct sock *sk)
549 {
550         if (sk_hashed(sk)) {
551                 __sk_del_node(sk);
552                 sk_node_init(&sk->sk_node);
553                 return true;
554         }
555         return false;
556 }
557 
558 /* Grab socket reference count. This operation is valid only
559    when sk is ALREADY grabbed f.e. it is found in hash table
560    or a list and the lookup is made under lock preventing hash table
561    modifications.
562  */
563 
564 static inline void sock_hold(struct sock *sk)
565 {
566         atomic_inc(&sk->sk_refcnt);
567 }
568 
569 /* Ungrab socket in the context, which assumes that socket refcnt
570    cannot hit zero, f.e. it is true in context of any socketcall.
571  */
572 static inline void __sock_put(struct sock *sk)
573 {
574         atomic_dec(&sk->sk_refcnt);
575 }
576 
577 static inline bool sk_del_node_init(struct sock *sk)
578 {
579         bool rc = __sk_del_node_init(sk);
580 
581         if (rc) {
582                 /* paranoid for a while -acme */
583                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
584                 __sock_put(sk);
585         }
586         return rc;
587 }
588 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
589 
590 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
591 {
592         if (sk_hashed(sk)) {
593                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
594                 return true;
595         }
596         return false;
597 }
598 
599 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
600 {
601         bool rc = __sk_nulls_del_node_init_rcu(sk);
602 
603         if (rc) {
604                 /* paranoid for a while -acme */
605                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
606                 __sock_put(sk);
607         }
608         return rc;
609 }
610 
611 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
612 {
613         hlist_add_head(&sk->sk_node, list);
614 }
615 
616 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
617 {
618         sock_hold(sk);
619         __sk_add_node(sk, list);
620 }
621 
622 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
623 {
624         sock_hold(sk);
625         hlist_add_head_rcu(&sk->sk_node, list);
626 }
627 
628 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
629 {
630         hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
631 }
632 
633 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
634 {
635         sock_hold(sk);
636         __sk_nulls_add_node_rcu(sk, list);
637 }
638 
639 static inline void __sk_del_bind_node(struct sock *sk)
640 {
641         __hlist_del(&sk->sk_bind_node);
642 }
643 
644 static inline void sk_add_bind_node(struct sock *sk,
645                                         struct hlist_head *list)
646 {
647         hlist_add_head(&sk->sk_bind_node, list);
648 }
649 
650 #define sk_for_each(__sk, list) \
651         hlist_for_each_entry(__sk, list, sk_node)
652 #define sk_for_each_rcu(__sk, list) \
653         hlist_for_each_entry_rcu(__sk, list, sk_node)
654 #define sk_nulls_for_each(__sk, node, list) \
655         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
656 #define sk_nulls_for_each_rcu(__sk, node, list) \
657         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
658 #define sk_for_each_from(__sk) \
659         hlist_for_each_entry_from(__sk, sk_node)
660 #define sk_nulls_for_each_from(__sk, node) \
661         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
662                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
663 #define sk_for_each_safe(__sk, tmp, list) \
664         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
665 #define sk_for_each_bound(__sk, list) \
666         hlist_for_each_entry(__sk, list, sk_bind_node)
667 
668 /**
669  * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
670  * @tpos:       the type * to use as a loop cursor.
671  * @pos:        the &struct hlist_node to use as a loop cursor.
672  * @head:       the head for your list.
673  * @offset:     offset of hlist_node within the struct.
674  *
675  */
676 #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset)                \
677         for (pos = (head)->first;                                              \
678              (!is_a_nulls(pos)) &&                                             \
679                 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;});       \
680              pos = pos->next)
681 
682 static inline struct user_namespace *sk_user_ns(struct sock *sk)
683 {
684         /* Careful only use this in a context where these parameters
685          * can not change and must all be valid, such as recvmsg from
686          * userspace.
687          */
688         return sk->sk_socket->file->f_cred->user_ns;
689 }
690 
691 /* Sock flags */
692 enum sock_flags {
693         SOCK_DEAD,
694         SOCK_DONE,
695         SOCK_URGINLINE,
696         SOCK_KEEPOPEN,
697         SOCK_LINGER,
698         SOCK_DESTROY,
699         SOCK_BROADCAST,
700         SOCK_TIMESTAMP,
701         SOCK_ZAPPED,
702         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
703         SOCK_DBG, /* %SO_DEBUG setting */
704         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
705         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
706         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
707         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
708         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
709         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
710         SOCK_FASYNC, /* fasync() active */
711         SOCK_RXQ_OVFL,
712         SOCK_ZEROCOPY, /* buffers from userspace */
713         SOCK_WIFI_STATUS, /* push wifi status to userspace */
714         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
715                      * Will use last 4 bytes of packet sent from
716                      * user-space instead.
717                      */
718         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
719         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
720 };
721 
722 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
723 {
724         nsk->sk_flags = osk->sk_flags;
725 }
726 
727 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
728 {
729         __set_bit(flag, &sk->sk_flags);
730 }
731 
732 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
733 {
734         __clear_bit(flag, &sk->sk_flags);
735 }
736 
737 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
738 {
739         return test_bit(flag, &sk->sk_flags);
740 }
741 
742 #ifdef CONFIG_NET
743 extern struct static_key memalloc_socks;
744 static inline int sk_memalloc_socks(void)
745 {
746         return static_key_false(&memalloc_socks);
747 }
748 #else
749 
750 static inline int sk_memalloc_socks(void)
751 {
752         return 0;
753 }
754 
755 #endif
756 
757 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
758 {
759         return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
760 }
761 
762 static inline void sk_acceptq_removed(struct sock *sk)
763 {
764         sk->sk_ack_backlog--;
765 }
766 
767 static inline void sk_acceptq_added(struct sock *sk)
768 {
769         sk->sk_ack_backlog++;
770 }
771 
772 static inline bool sk_acceptq_is_full(const struct sock *sk)
773 {
774         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
775 }
776 
777 /*
778  * Compute minimal free write space needed to queue new packets.
779  */
780 static inline int sk_stream_min_wspace(const struct sock *sk)
781 {
782         return sk->sk_wmem_queued >> 1;
783 }
784 
785 static inline int sk_stream_wspace(const struct sock *sk)
786 {
787         return sk->sk_sndbuf - sk->sk_wmem_queued;
788 }
789 
790 void sk_stream_write_space(struct sock *sk);
791 
792 /* OOB backlog add */
793 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
794 {
795         /* dont let skb dst not refcounted, we are going to leave rcu lock */
796         skb_dst_force(skb);
797 
798         if (!sk->sk_backlog.tail)
799                 sk->sk_backlog.head = skb;
800         else
801                 sk->sk_backlog.tail->next = skb;
802 
803         sk->sk_backlog.tail = skb;
804         skb->next = NULL;
805 }
806 
807 /*
808  * Take into account size of receive queue and backlog queue
809  * Do not take into account this skb truesize,
810  * to allow even a single big packet to come.
811  */
812 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
813 {
814         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
815 
816         return qsize > limit;
817 }
818 
819 /* The per-socket spinlock must be held here. */
820 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
821                                               unsigned int limit)
822 {
823         if (sk_rcvqueues_full(sk, limit))
824                 return -ENOBUFS;
825 
826         __sk_add_backlog(sk, skb);
827         sk->sk_backlog.len += skb->truesize;
828         return 0;
829 }
830 
831 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
832 
833 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
834 {
835         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
836                 return __sk_backlog_rcv(sk, skb);
837 
838         return sk->sk_backlog_rcv(sk, skb);
839 }
840 
841 static inline void sock_rps_record_flow_hash(__u32 hash)
842 {
843 #ifdef CONFIG_RPS
844         struct rps_sock_flow_table *sock_flow_table;
845 
846         rcu_read_lock();
847         sock_flow_table = rcu_dereference(rps_sock_flow_table);
848         rps_record_sock_flow(sock_flow_table, hash);
849         rcu_read_unlock();
850 #endif
851 }
852 
853 static inline void sock_rps_reset_flow_hash(__u32 hash)
854 {
855 #ifdef CONFIG_RPS
856         struct rps_sock_flow_table *sock_flow_table;
857 
858         rcu_read_lock();
859         sock_flow_table = rcu_dereference(rps_sock_flow_table);
860         rps_reset_sock_flow(sock_flow_table, hash);
861         rcu_read_unlock();
862 #endif
863 }
864 
865 static inline void sock_rps_record_flow(const struct sock *sk)
866 {
867 #ifdef CONFIG_RPS
868         sock_rps_record_flow_hash(sk->sk_rxhash);
869 #endif
870 }
871 
872 static inline void sock_rps_reset_flow(const struct sock *sk)
873 {
874 #ifdef CONFIG_RPS
875         sock_rps_reset_flow_hash(sk->sk_rxhash);
876 #endif
877 }
878 
879 static inline void sock_rps_save_rxhash(struct sock *sk,
880                                         const struct sk_buff *skb)
881 {
882 #ifdef CONFIG_RPS
883         if (unlikely(sk->sk_rxhash != skb->hash)) {
884                 sock_rps_reset_flow(sk);
885                 sk->sk_rxhash = skb->hash;
886         }
887 #endif
888 }
889 
890 static inline void sock_rps_reset_rxhash(struct sock *sk)
891 {
892 #ifdef CONFIG_RPS
893         sock_rps_reset_flow(sk);
894         sk->sk_rxhash = 0;
895 #endif
896 }
897 
898 #define sk_wait_event(__sk, __timeo, __condition)                       \
899         ({      int __rc;                                               \
900                 release_sock(__sk);                                     \
901                 __rc = __condition;                                     \
902                 if (!__rc) {                                            \
903                         *(__timeo) = schedule_timeout(*(__timeo));      \
904                 }                                                       \
905                 lock_sock(__sk);                                        \
906                 __rc = __condition;                                     \
907                 __rc;                                                   \
908         })
909 
910 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
911 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
912 void sk_stream_wait_close(struct sock *sk, long timeo_p);
913 int sk_stream_error(struct sock *sk, int flags, int err);
914 void sk_stream_kill_queues(struct sock *sk);
915 void sk_set_memalloc(struct sock *sk);
916 void sk_clear_memalloc(struct sock *sk);
917 
918 int sk_wait_data(struct sock *sk, long *timeo);
919 
920 struct request_sock_ops;
921 struct timewait_sock_ops;
922 struct inet_hashinfo;
923 struct raw_hashinfo;
924 struct module;
925 
926 /*
927  * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
928  * un-modified. Special care is taken when initializing object to zero.
929  */
930 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
931 {
932         if (offsetof(struct sock, sk_node.next) != 0)
933                 memset(sk, 0, offsetof(struct sock, sk_node.next));
934         memset(&sk->sk_node.pprev, 0,
935                size - offsetof(struct sock, sk_node.pprev));
936 }
937 
938 /* Networking protocol blocks we attach to sockets.
939  * socket layer -> transport layer interface
940  * transport -> network interface is defined by struct inet_proto
941  */
942 struct proto {
943         void                    (*close)(struct sock *sk,
944                                         long timeout);
945         int                     (*connect)(struct sock *sk,
946                                         struct sockaddr *uaddr,
947                                         int addr_len);
948         int                     (*disconnect)(struct sock *sk, int flags);
949 
950         struct sock *           (*accept)(struct sock *sk, int flags, int *err);
951 
952         int                     (*ioctl)(struct sock *sk, int cmd,
953                                          unsigned long arg);
954         int                     (*init)(struct sock *sk);
955         void                    (*destroy)(struct sock *sk);
956         void                    (*shutdown)(struct sock *sk, int how);
957         int                     (*setsockopt)(struct sock *sk, int level,
958                                         int optname, char __user *optval,
959                                         unsigned int optlen);
960         int                     (*getsockopt)(struct sock *sk, int level,
961                                         int optname, char __user *optval,
962                                         int __user *option);
963 #ifdef CONFIG_COMPAT
964         int                     (*compat_setsockopt)(struct sock *sk,
965                                         int level,
966                                         int optname, char __user *optval,
967                                         unsigned int optlen);
968         int                     (*compat_getsockopt)(struct sock *sk,
969                                         int level,
970                                         int optname, char __user *optval,
971                                         int __user *option);
972         int                     (*compat_ioctl)(struct sock *sk,
973                                         unsigned int cmd, unsigned long arg);
974 #endif
975         int                     (*sendmsg)(struct kiocb *iocb, struct sock *sk,
976                                            struct msghdr *msg, size_t len);
977         int                     (*recvmsg)(struct kiocb *iocb, struct sock *sk,
978                                            struct msghdr *msg,
979                                            size_t len, int noblock, int flags,
980                                            int *addr_len);
981         int                     (*sendpage)(struct sock *sk, struct page *page,
982                                         int offset, size_t size, int flags);
983         int                     (*bind)(struct sock *sk,
984                                         struct sockaddr *uaddr, int addr_len);
985 
986         int                     (*backlog_rcv) (struct sock *sk,
987                                                 struct sk_buff *skb);
988 
989         void            (*release_cb)(struct sock *sk);
990 
991         /* Keeping track of sk's, looking them up, and port selection methods. */
992         void                    (*hash)(struct sock *sk);
993         void                    (*unhash)(struct sock *sk);
994         void                    (*rehash)(struct sock *sk);
995         int                     (*get_port)(struct sock *sk, unsigned short snum);
996         void                    (*clear_sk)(struct sock *sk, int size);
997 
998         /* Keeping track of sockets in use */
999 #ifdef CONFIG_PROC_FS
1000         unsigned int            inuse_idx;
1001 #endif
1002 
1003         bool                    (*stream_memory_free)(const struct sock *sk);
1004         /* Memory pressure */
1005         void                    (*enter_memory_pressure)(struct sock *sk);
1006         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
1007         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
1008         /*
1009          * Pressure flag: try to collapse.
1010          * Technical note: it is used by multiple contexts non atomically.
1011          * All the __sk_mem_schedule() is of this nature: accounting
1012          * is strict, actions are advisory and have some latency.
1013          */
1014         int                     *memory_pressure;
1015         long                    *sysctl_mem;
1016         int                     *sysctl_wmem;
1017         int                     *sysctl_rmem;
1018         int                     max_header;
1019         bool                    no_autobind;
1020 
1021         struct kmem_cache       *slab;
1022         unsigned int            obj_size;
1023         int                     slab_flags;
1024 
1025         struct percpu_counter   *orphan_count;
1026 
1027         struct request_sock_ops *rsk_prot;
1028         struct timewait_sock_ops *twsk_prot;
1029 
1030         union {
1031                 struct inet_hashinfo    *hashinfo;
1032                 struct udp_table        *udp_table;
1033                 struct raw_hashinfo     *raw_hash;
1034         } h;
1035 
1036         struct module           *owner;
1037 
1038         char                    name[32];
1039 
1040         struct list_head        node;
1041 #ifdef SOCK_REFCNT_DEBUG
1042         atomic_t                socks;
1043 #endif
1044 #ifdef CONFIG_MEMCG_KMEM
1045         /*
1046          * cgroup specific init/deinit functions. Called once for all
1047          * protocols that implement it, from cgroups populate function.
1048          * This function has to setup any files the protocol want to
1049          * appear in the kmem cgroup filesystem.
1050          */
1051         int                     (*init_cgroup)(struct mem_cgroup *memcg,
1052                                                struct cgroup_subsys *ss);
1053         void                    (*destroy_cgroup)(struct mem_cgroup *memcg);
1054         struct cg_proto         *(*proto_cgroup)(struct mem_cgroup *memcg);
1055 #endif
1056 };
1057 
1058 /*
1059  * Bits in struct cg_proto.flags
1060  */
1061 enum cg_proto_flags {
1062         /* Currently active and new sockets should be assigned to cgroups */
1063         MEMCG_SOCK_ACTIVE,
1064         /* It was ever activated; we must disarm static keys on destruction */
1065         MEMCG_SOCK_ACTIVATED,
1066 };
1067 
1068 struct cg_proto {
1069         struct res_counter      memory_allocated;       /* Current allocated memory. */
1070         struct percpu_counter   sockets_allocated;      /* Current number of sockets. */
1071         int                     memory_pressure;
1072         long                    sysctl_mem[3];
1073         unsigned long           flags;
1074         /*
1075          * memcg field is used to find which memcg we belong directly
1076          * Each memcg struct can hold more than one cg_proto, so container_of
1077          * won't really cut.
1078          *
1079          * The elegant solution would be having an inverse function to
1080          * proto_cgroup in struct proto, but that means polluting the structure
1081          * for everybody, instead of just for memcg users.
1082          */
1083         struct mem_cgroup       *memcg;
1084 };
1085 
1086 int proto_register(struct proto *prot, int alloc_slab);
1087 void proto_unregister(struct proto *prot);
1088 
1089 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1090 {
1091         return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1092 }
1093 
1094 static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1095 {
1096         return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1097 }
1098 
1099 #ifdef SOCK_REFCNT_DEBUG
1100 static inline void sk_refcnt_debug_inc(struct sock *sk)
1101 {
1102         atomic_inc(&sk->sk_prot->socks);
1103 }
1104 
1105 static inline void sk_refcnt_debug_dec(struct sock *sk)
1106 {
1107         atomic_dec(&sk->sk_prot->socks);
1108         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1109                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1110 }
1111 
1112 static inline void sk_refcnt_debug_release(const struct sock *sk)
1113 {
1114         if (atomic_read(&sk->sk_refcnt) != 1)
1115                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1116                        sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1117 }
1118 #else /* SOCK_REFCNT_DEBUG */
1119 #define sk_refcnt_debug_inc(sk) do { } while (0)
1120 #define sk_refcnt_debug_dec(sk) do { } while (0)
1121 #define sk_refcnt_debug_release(sk) do { } while (0)
1122 #endif /* SOCK_REFCNT_DEBUG */
1123 
1124 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1125 extern struct static_key memcg_socket_limit_enabled;
1126 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1127                                                struct cg_proto *cg_proto)
1128 {
1129         return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1130 }
1131 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1132 #else
1133 #define mem_cgroup_sockets_enabled 0
1134 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1135                                                struct cg_proto *cg_proto)
1136 {
1137         return NULL;
1138 }
1139 #endif
1140 
1141 static inline bool sk_stream_memory_free(const struct sock *sk)
1142 {
1143         if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1144                 return false;
1145 
1146         return sk->sk_prot->stream_memory_free ?
1147                 sk->sk_prot->stream_memory_free(sk) : true;
1148 }
1149 
1150 static inline bool sk_stream_is_writeable(const struct sock *sk)
1151 {
1152         return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1153                sk_stream_memory_free(sk);
1154 }
1155 
1156 
1157 static inline bool sk_has_memory_pressure(const struct sock *sk)
1158 {
1159         return sk->sk_prot->memory_pressure != NULL;
1160 }
1161 
1162 static inline bool sk_under_memory_pressure(const struct sock *sk)
1163 {
1164         if (!sk->sk_prot->memory_pressure)
1165                 return false;
1166 
1167         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1168                 return !!sk->sk_cgrp->memory_pressure;
1169 
1170         return !!*sk->sk_prot->memory_pressure;
1171 }
1172 
1173 static inline void sk_leave_memory_pressure(struct sock *sk)
1174 {
1175         int *memory_pressure = sk->sk_prot->memory_pressure;
1176 
1177         if (!memory_pressure)
1178                 return;
1179 
1180         if (*memory_pressure)
1181                 *memory_pressure = 0;
1182 
1183         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1184                 struct cg_proto *cg_proto = sk->sk_cgrp;
1185                 struct proto *prot = sk->sk_prot;
1186 
1187                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1188                         cg_proto->memory_pressure = 0;
1189         }
1190 
1191 }
1192 
1193 static inline void sk_enter_memory_pressure(struct sock *sk)
1194 {
1195         if (!sk->sk_prot->enter_memory_pressure)
1196                 return;
1197 
1198         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1199                 struct cg_proto *cg_proto = sk->sk_cgrp;
1200                 struct proto *prot = sk->sk_prot;
1201 
1202                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1203                         cg_proto->memory_pressure = 1;
1204         }
1205 
1206         sk->sk_prot->enter_memory_pressure(sk);
1207 }
1208 
1209 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1210 {
1211         long *prot = sk->sk_prot->sysctl_mem;
1212         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1213                 prot = sk->sk_cgrp->sysctl_mem;
1214         return prot[index];
1215 }
1216 
1217 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1218                                               unsigned long amt,
1219                                               int *parent_status)
1220 {
1221         struct res_counter *fail;
1222         int ret;
1223 
1224         ret = res_counter_charge_nofail(&prot->memory_allocated,
1225                                         amt << PAGE_SHIFT, &fail);
1226         if (ret < 0)
1227                 *parent_status = OVER_LIMIT;
1228 }
1229 
1230 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1231                                               unsigned long amt)
1232 {
1233         res_counter_uncharge(&prot->memory_allocated, amt << PAGE_SHIFT);
1234 }
1235 
1236 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1237 {
1238         u64 ret;
1239         ret = res_counter_read_u64(&prot->memory_allocated, RES_USAGE);
1240         return ret >> PAGE_SHIFT;
1241 }
1242 
1243 static inline long
1244 sk_memory_allocated(const struct sock *sk)
1245 {
1246         struct proto *prot = sk->sk_prot;
1247         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1248                 return memcg_memory_allocated_read(sk->sk_cgrp);
1249 
1250         return atomic_long_read(prot->memory_allocated);
1251 }
1252 
1253 static inline long
1254 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1255 {
1256         struct proto *prot = sk->sk_prot;
1257 
1258         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1259                 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1260                 /* update the root cgroup regardless */
1261                 atomic_long_add_return(amt, prot->memory_allocated);
1262                 return memcg_memory_allocated_read(sk->sk_cgrp);
1263         }
1264 
1265         return atomic_long_add_return(amt, prot->memory_allocated);
1266 }
1267 
1268 static inline void
1269 sk_memory_allocated_sub(struct sock *sk, int amt)
1270 {
1271         struct proto *prot = sk->sk_prot;
1272 
1273         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1274                 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1275 
1276         atomic_long_sub(amt, prot->memory_allocated);
1277 }
1278 
1279 static inline void sk_sockets_allocated_dec(struct sock *sk)
1280 {
1281         struct proto *prot = sk->sk_prot;
1282 
1283         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1284                 struct cg_proto *cg_proto = sk->sk_cgrp;
1285 
1286                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1287                         percpu_counter_dec(&cg_proto->sockets_allocated);
1288         }
1289 
1290         percpu_counter_dec(prot->sockets_allocated);
1291 }
1292 
1293 static inline void sk_sockets_allocated_inc(struct sock *sk)
1294 {
1295         struct proto *prot = sk->sk_prot;
1296 
1297         if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1298                 struct cg_proto *cg_proto = sk->sk_cgrp;
1299 
1300                 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1301                         percpu_counter_inc(&cg_proto->sockets_allocated);
1302         }
1303 
1304         percpu_counter_inc(prot->sockets_allocated);
1305 }
1306 
1307 static inline int
1308 sk_sockets_allocated_read_positive(struct sock *sk)
1309 {
1310         struct proto *prot = sk->sk_prot;
1311 
1312         if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1313                 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
1314 
1315         return percpu_counter_read_positive(prot->sockets_allocated);
1316 }
1317 
1318 static inline int
1319 proto_sockets_allocated_sum_positive(struct proto *prot)
1320 {
1321         return percpu_counter_sum_positive(prot->sockets_allocated);
1322 }
1323 
1324 static inline long
1325 proto_memory_allocated(struct proto *prot)
1326 {
1327         return atomic_long_read(prot->memory_allocated);
1328 }
1329 
1330 static inline bool
1331 proto_memory_pressure(struct proto *prot)
1332 {
1333         if (!prot->memory_pressure)
1334                 return false;
1335         return !!*prot->memory_pressure;
1336 }
1337 
1338 
1339 #ifdef CONFIG_PROC_FS
1340 /* Called with local bh disabled */
1341 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1342 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1343 #else
1344 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1345                 int inc)
1346 {
1347 }
1348 #endif
1349 
1350 
1351 /* With per-bucket locks this operation is not-atomic, so that
1352  * this version is not worse.
1353  */
1354 static inline void __sk_prot_rehash(struct sock *sk)
1355 {
1356         sk->sk_prot->unhash(sk);
1357         sk->sk_prot->hash(sk);
1358 }
1359 
1360 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1361 
1362 /* About 10 seconds */
1363 #define SOCK_DESTROY_TIME (10*HZ)
1364 
1365 /* Sockets 0-1023 can't be bound to unless you are superuser */
1366 #define PROT_SOCK       1024
1367 
1368 #define SHUTDOWN_MASK   3
1369 #define RCV_SHUTDOWN    1
1370 #define SEND_SHUTDOWN   2
1371 
1372 #define SOCK_SNDBUF_LOCK        1
1373 #define SOCK_RCVBUF_LOCK        2
1374 #define SOCK_BINDADDR_LOCK      4
1375 #define SOCK_BINDPORT_LOCK      8
1376 
1377 /* sock_iocb: used to kick off async processing of socket ios */
1378 struct sock_iocb {
1379         struct list_head        list;
1380 
1381         int                     flags;
1382         int                     size;
1383         struct socket           *sock;
1384         struct sock             *sk;
1385         struct scm_cookie       *scm;
1386         struct msghdr           *msg, async_msg;
1387         struct kiocb            *kiocb;
1388 };
1389 
1390 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1391 {
1392         return (struct sock_iocb *)iocb->private;
1393 }
1394 
1395 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1396 {
1397         return si->kiocb;
1398 }
1399 
1400 struct socket_alloc {
1401         struct socket socket;
1402         struct inode vfs_inode;
1403 };
1404 
1405 static inline struct socket *SOCKET_I(struct inode *inode)
1406 {
1407         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1408 }
1409 
1410 static inline struct inode *SOCK_INODE(struct socket *socket)
1411 {
1412         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1413 }
1414 
1415 /*
1416  * Functions for memory accounting
1417  */
1418 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1419 void __sk_mem_reclaim(struct sock *sk);
1420 
1421 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1422 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1423 #define SK_MEM_SEND     0
1424 #define SK_MEM_RECV     1
1425 
1426 static inline int sk_mem_pages(int amt)
1427 {
1428         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1429 }
1430 
1431 static inline bool sk_has_account(struct sock *sk)
1432 {
1433         /* return true if protocol supports memory accounting */
1434         return !!sk->sk_prot->memory_allocated;
1435 }
1436 
1437 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1438 {
1439         if (!sk_has_account(sk))
1440                 return true;
1441         return size <= sk->sk_forward_alloc ||
1442                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1443 }
1444 
1445 static inline bool
1446 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1447 {
1448         if (!sk_has_account(sk))
1449                 return true;
1450         return size<= sk->sk_forward_alloc ||
1451                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1452                 skb_pfmemalloc(skb);
1453 }
1454 
1455 static inline void sk_mem_reclaim(struct sock *sk)
1456 {
1457         if (!sk_has_account(sk))
1458                 return;
1459         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1460                 __sk_mem_reclaim(sk);
1461 }
1462 
1463 static inline void sk_mem_reclaim_partial(struct sock *sk)
1464 {
1465         if (!sk_has_account(sk))
1466                 return;
1467         if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1468                 __sk_mem_reclaim(sk);
1469 }
1470 
1471 static inline void sk_mem_charge(struct sock *sk, int size)
1472 {
1473         if (!sk_has_account(sk))
1474                 return;
1475         sk->sk_forward_alloc -= size;
1476 }
1477 
1478 static inline void sk_mem_uncharge(struct sock *sk, int size)
1479 {
1480         if (!sk_has_account(sk))
1481                 return;
1482         sk->sk_forward_alloc += size;
1483 }
1484 
1485 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1486 {
1487         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1488         sk->sk_wmem_queued -= skb->truesize;
1489         sk_mem_uncharge(sk, skb->truesize);
1490         __kfree_skb(skb);
1491 }
1492 
1493 /* Used by processes to "lock" a socket state, so that
1494  * interrupts and bottom half handlers won't change it
1495  * from under us. It essentially blocks any incoming
1496  * packets, so that we won't get any new data or any
1497  * packets that change the state of the socket.
1498  *
1499  * While locked, BH processing will add new packets to
1500  * the backlog queue.  This queue is processed by the
1501  * owner of the socket lock right before it is released.
1502  *
1503  * Since ~2.3.5 it is also exclusive sleep lock serializing
1504  * accesses from user process context.
1505  */
1506 #define sock_owned_by_user(sk)  ((sk)->sk_lock.owned)
1507 
1508 static inline void sock_release_ownership(struct sock *sk)
1509 {
1510         sk->sk_lock.owned = 0;
1511 }
1512 
1513 /*
1514  * Macro so as to not evaluate some arguments when
1515  * lockdep is not enabled.
1516  *
1517  * Mark both the sk_lock and the sk_lock.slock as a
1518  * per-address-family lock class.
1519  */
1520 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1521 do {                                                                    \
1522         sk->sk_lock.owned = 0;                                          \
1523         init_waitqueue_head(&sk->sk_lock.wq);                           \
1524         spin_lock_init(&(sk)->sk_lock.slock);                           \
1525         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1526                         sizeof((sk)->sk_lock));                         \
1527         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1528                                 (skey), (sname));                               \
1529         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1530 } while (0)
1531 
1532 void lock_sock_nested(struct sock *sk, int subclass);
1533 
1534 static inline void lock_sock(struct sock *sk)
1535 {
1536         lock_sock_nested(sk, 0);
1537 }
1538 
1539 void release_sock(struct sock *sk);
1540 
1541 /* BH context may only use the following locking interface. */
1542 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1543 #define bh_lock_sock_nested(__sk) \
1544                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1545                                 SINGLE_DEPTH_NESTING)
1546 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1547 
1548 bool lock_sock_fast(struct sock *sk);
1549 /**
1550  * unlock_sock_fast - complement of lock_sock_fast
1551  * @sk: socket
1552  * @slow: slow mode
1553  *
1554  * fast unlock socket for user context.
1555  * If slow mode is on, we call regular release_sock()
1556  */
1557 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1558 {
1559         if (slow)
1560                 release_sock(sk);
1561         else
1562                 spin_unlock_bh(&sk->sk_lock.slock);
1563 }
1564 
1565 
1566 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1567                       struct proto *prot);
1568 void sk_free(struct sock *sk);
1569 void sk_release_kernel(struct sock *sk);
1570 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1571 
1572 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1573                              gfp_t priority);
1574 void sock_wfree(struct sk_buff *skb);
1575 void skb_orphan_partial(struct sk_buff *skb);
1576 void sock_rfree(struct sk_buff *skb);
1577 void sock_edemux(struct sk_buff *skb);
1578 
1579 int sock_setsockopt(struct socket *sock, int level, int op,
1580                     char __user *optval, unsigned int optlen);
1581 
1582 int sock_getsockopt(struct socket *sock, int level, int op,
1583                     char __user *optval, int __user *optlen);
1584 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1585                                     int noblock, int *errcode);
1586 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1587                                      unsigned long data_len, int noblock,
1588                                      int *errcode, int max_page_order);
1589 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1590 void sock_kfree_s(struct sock *sk, void *mem, int size);
1591 void sk_send_sigurg(struct sock *sk);
1592 
1593 /*
1594  * Functions to fill in entries in struct proto_ops when a protocol
1595  * does not implement a particular function.
1596  */
1597 int sock_no_bind(struct socket *, struct sockaddr *, int);
1598 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1599 int sock_no_socketpair(struct socket *, struct socket *);
1600 int sock_no_accept(struct socket *, struct socket *, int);
1601 int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1602 unsigned int sock_no_poll(struct file *, struct socket *,
1603                           struct poll_table_struct *);
1604 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1605 int sock_no_listen(struct socket *, int);
1606 int sock_no_shutdown(struct socket *, int);
1607 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1608 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1609 int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1610 int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1611                     int);
1612 int sock_no_mmap(struct file *file, struct socket *sock,
1613                  struct vm_area_struct *vma);
1614 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1615                          size_t size, int flags);
1616 
1617 /*
1618  * Functions to fill in entries in struct proto_ops when a protocol
1619  * uses the inet style.
1620  */
1621 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1622                                   char __user *optval, int __user *optlen);
1623 int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1624                                struct msghdr *msg, size_t size, int flags);
1625 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1626                                   char __user *optval, unsigned int optlen);
1627 int compat_sock_common_getsockopt(struct socket *sock, int level,
1628                 int optname, char __user *optval, int __user *optlen);
1629 int compat_sock_common_setsockopt(struct socket *sock, int level,
1630                 int optname, char __user *optval, unsigned int optlen);
1631 
1632 void sk_common_release(struct sock *sk);
1633 
1634 /*
1635  *      Default socket callbacks and setup code
1636  */
1637 
1638 /* Initialise core socket variables */
1639 void sock_init_data(struct socket *sock, struct sock *sk);
1640 
1641 /*
1642  * Socket reference counting postulates.
1643  *
1644  * * Each user of socket SHOULD hold a reference count.
1645  * * Each access point to socket (an hash table bucket, reference from a list,
1646  *   running timer, skb in flight MUST hold a reference count.
1647  * * When reference count hits 0, it means it will never increase back.
1648  * * When reference count hits 0, it means that no references from
1649  *   outside exist to this socket and current process on current CPU
1650  *   is last user and may/should destroy this socket.
1651  * * sk_free is called from any context: process, BH, IRQ. When
1652  *   it is called, socket has no references from outside -> sk_free
1653  *   may release descendant resources allocated by the socket, but
1654  *   to the time when it is called, socket is NOT referenced by any
1655  *   hash tables, lists etc.
1656  * * Packets, delivered from outside (from network or from another process)
1657  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1658  *   when they sit in queue. Otherwise, packets will leak to hole, when
1659  *   socket is looked up by one cpu and unhasing is made by another CPU.
1660  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1661  *   (leak to backlog). Packet socket does all the processing inside
1662  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1663  *   use separate SMP lock, so that they are prone too.
1664  */
1665 
1666 /* Ungrab socket and destroy it, if it was the last reference. */
1667 static inline void sock_put(struct sock *sk)
1668 {
1669         if (atomic_dec_and_test(&sk->sk_refcnt))
1670                 sk_free(sk);
1671 }
1672 /* Generic version of sock_put(), dealing with all sockets
1673  * (TCP_TIMEWAIT, ESTABLISHED...)
1674  */
1675 void sock_gen_put(struct sock *sk);
1676 
1677 int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
1678 
1679 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1680 {
1681         sk->sk_tx_queue_mapping = tx_queue;
1682 }
1683 
1684 static inline void sk_tx_queue_clear(struct sock *sk)
1685 {
1686         sk->sk_tx_queue_mapping = -1;
1687 }
1688 
1689 static inline int sk_tx_queue_get(const struct sock *sk)
1690 {
1691         return sk ? sk->sk_tx_queue_mapping : -1;
1692 }
1693 
1694 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1695 {
1696         sk_tx_queue_clear(sk);
1697         sk->sk_socket = sock;
1698 }
1699 
1700 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1701 {
1702         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1703         return &rcu_dereference_raw(sk->sk_wq)->wait;
1704 }
1705 /* Detach socket from process context.
1706  * Announce socket dead, detach it from wait queue and inode.
1707  * Note that parent inode held reference count on this struct sock,
1708  * we do not release it in this function, because protocol
1709  * probably wants some additional cleanups or even continuing
1710  * to work with this socket (TCP).
1711  */
1712 static inline void sock_orphan(struct sock *sk)
1713 {
1714         write_lock_bh(&sk->sk_callback_lock);
1715         sock_set_flag(sk, SOCK_DEAD);
1716         sk_set_socket(sk, NULL);
1717         sk->sk_wq  = NULL;
1718         write_unlock_bh(&sk->sk_callback_lock);
1719 }
1720 
1721 static inline void sock_graft(struct sock *sk, struct socket *parent)
1722 {
1723         write_lock_bh(&sk->sk_callback_lock);
1724         sk->sk_wq = parent->wq;
1725         parent->sk = sk;
1726         sk_set_socket(sk, parent);
1727         security_sock_graft(sk, parent);
1728         write_unlock_bh(&sk->sk_callback_lock);
1729 }
1730 
1731 kuid_t sock_i_uid(struct sock *sk);
1732 unsigned long sock_i_ino(struct sock *sk);
1733 
1734 static inline struct dst_entry *
1735 __sk_dst_get(struct sock *sk)
1736 {
1737         return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1738                                                        lockdep_is_held(&sk->sk_lock.slock));
1739 }
1740 
1741 static inline struct dst_entry *
1742 sk_dst_get(struct sock *sk)
1743 {
1744         struct dst_entry *dst;
1745 
1746         rcu_read_lock();
1747         dst = rcu_dereference(sk->sk_dst_cache);
1748         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1749                 dst = NULL;
1750         rcu_read_unlock();
1751         return dst;
1752 }
1753 
1754 static inline void dst_negative_advice(struct sock *sk)
1755 {
1756         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1757 
1758         if (dst && dst->ops->negative_advice) {
1759                 ndst = dst->ops->negative_advice(dst);
1760 
1761                 if (ndst != dst) {
1762                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1763                         sk_tx_queue_clear(sk);
1764                 }
1765         }
1766 }
1767 
1768 static inline void
1769 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1770 {
1771         struct dst_entry *old_dst;
1772 
1773         sk_tx_queue_clear(sk);
1774         /*
1775          * This can be called while sk is owned by the caller only,
1776          * with no state that can be checked in a rcu_dereference_check() cond
1777          */
1778         old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1779         rcu_assign_pointer(sk->sk_dst_cache, dst);
1780         dst_release(old_dst);
1781 }
1782 
1783 static inline void
1784 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1785 {
1786         struct dst_entry *old_dst;
1787 
1788         sk_tx_queue_clear(sk);
1789         old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1790         dst_release(old_dst);
1791 }
1792 
1793 static inline void
1794 __sk_dst_reset(struct sock *sk)
1795 {
1796         __sk_dst_set(sk, NULL);
1797 }
1798 
1799 static inline void
1800 sk_dst_reset(struct sock *sk)
1801 {
1802         sk_dst_set(sk, NULL);
1803 }
1804 
1805 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1806 
1807 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1808 
1809 static inline bool sk_can_gso(const struct sock *sk)
1810 {
1811         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1812 }
1813 
1814 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1815 
1816 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1817 {
1818         sk->sk_route_nocaps |= flags;
1819         sk->sk_route_caps &= ~flags;
1820 }
1821 
1822 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1823                                            char __user *from, char *to,
1824                                            int copy, int offset)
1825 {
1826         if (skb->ip_summed == CHECKSUM_NONE) {
1827                 int err = 0;
1828                 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1829                 if (err)
1830                         return err;
1831                 skb->csum = csum_block_add(skb->csum, csum, offset);
1832         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1833                 if (!access_ok(VERIFY_READ, from, copy) ||
1834                     __copy_from_user_nocache(to, from, copy))
1835                         return -EFAULT;
1836         } else if (copy_from_user(to, from, copy))
1837                 return -EFAULT;
1838 
1839         return 0;
1840 }
1841 
1842 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1843                                        char __user *from, int copy)
1844 {
1845         int err, offset = skb->len;
1846 
1847         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1848                                        copy, offset);
1849         if (err)
1850                 __skb_trim(skb, offset);
1851 
1852         return err;
1853 }
1854 
1855 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1856                                            struct sk_buff *skb,
1857                                            struct page *page,
1858                                            int off, int copy)
1859 {
1860         int err;
1861 
1862         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1863                                        copy, skb->len);
1864         if (err)
1865                 return err;
1866 
1867         skb->len             += copy;
1868         skb->data_len        += copy;
1869         skb->truesize        += copy;
1870         sk->sk_wmem_queued   += copy;
1871         sk_mem_charge(sk, copy);
1872         return 0;
1873 }
1874 
1875 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1876                                    struct sk_buff *skb, struct page *page,
1877                                    int off, int copy)
1878 {
1879         if (skb->ip_summed == CHECKSUM_NONE) {
1880                 int err = 0;
1881                 __wsum csum = csum_and_copy_from_user(from,
1882                                                      page_address(page) + off,
1883                                                             copy, 0, &err);
1884                 if (err)
1885                         return err;
1886                 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1887         } else if (copy_from_user(page_address(page) + off, from, copy))
1888                 return -EFAULT;
1889 
1890         skb->len             += copy;
1891         skb->data_len        += copy;
1892         skb->truesize        += copy;
1893         sk->sk_wmem_queued   += copy;
1894         sk_mem_charge(sk, copy);
1895         return 0;
1896 }
1897 
1898 /**
1899  * sk_wmem_alloc_get - returns write allocations
1900  * @sk: socket
1901  *
1902  * Returns sk_wmem_alloc minus initial offset of one
1903  */
1904 static inline int sk_wmem_alloc_get(const struct sock *sk)
1905 {
1906         return atomic_read(&sk->sk_wmem_alloc) - 1;
1907 }
1908 
1909 /**
1910  * sk_rmem_alloc_get - returns read allocations
1911  * @sk: socket
1912  *
1913  * Returns sk_rmem_alloc
1914  */
1915 static inline int sk_rmem_alloc_get(const struct sock *sk)
1916 {
1917         return atomic_read(&sk->sk_rmem_alloc);
1918 }
1919 
1920 /**
1921  * sk_has_allocations - check if allocations are outstanding
1922  * @sk: socket
1923  *
1924  * Returns true if socket has write or read allocations
1925  */
1926 static inline bool sk_has_allocations(const struct sock *sk)
1927 {
1928         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1929 }
1930 
1931 /**
1932  * wq_has_sleeper - check if there are any waiting processes
1933  * @wq: struct socket_wq
1934  *
1935  * Returns true if socket_wq has waiting processes
1936  *
1937  * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1938  * barrier call. They were added due to the race found within the tcp code.
1939  *
1940  * Consider following tcp code paths:
1941  *
1942  * CPU1                  CPU2
1943  *
1944  * sys_select            receive packet
1945  *   ...                 ...
1946  *   __add_wait_queue    update tp->rcv_nxt
1947  *   ...                 ...
1948  *   tp->rcv_nxt check   sock_def_readable
1949  *   ...                 {
1950  *   schedule               rcu_read_lock();
1951  *                          wq = rcu_dereference(sk->sk_wq);
1952  *                          if (wq && waitqueue_active(&wq->wait))
1953  *                              wake_up_interruptible(&wq->wait)
1954  *                          ...
1955  *                       }
1956  *
1957  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1958  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1959  * could then endup calling schedule and sleep forever if there are no more
1960  * data on the socket.
1961  *
1962  */
1963 static inline bool wq_has_sleeper(struct socket_wq *wq)
1964 {
1965         /* We need to be sure we are in sync with the
1966          * add_wait_queue modifications to the wait queue.
1967          *
1968          * This memory barrier is paired in the sock_poll_wait.
1969          */
1970         smp_mb();
1971         return wq && waitqueue_active(&wq->wait);
1972 }
1973 
1974 /**
1975  * sock_poll_wait - place memory barrier behind the poll_wait call.
1976  * @filp:           file
1977  * @wait_address:   socket wait queue
1978  * @p:              poll_table
1979  *
1980  * See the comments in the wq_has_sleeper function.
1981  */
1982 static inline void sock_poll_wait(struct file *filp,
1983                 wait_queue_head_t *wait_address, poll_table *p)
1984 {
1985         if (!poll_does_not_wait(p) && wait_address) {
1986                 poll_wait(filp, wait_address, p);
1987                 /* We need to be sure we are in sync with the
1988                  * socket flags modification.
1989                  *
1990                  * This memory barrier is paired in the wq_has_sleeper.
1991                  */
1992                 smp_mb();
1993         }
1994 }
1995 
1996 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1997 {
1998         if (sk->sk_txhash) {
1999                 skb->l4_hash = 1;
2000                 skb->hash = sk->sk_txhash;
2001         }
2002 }
2003 
2004 /*
2005  *      Queue a received datagram if it will fit. Stream and sequenced
2006  *      protocols can't normally use this as they need to fit buffers in
2007  *      and play with them.
2008  *
2009  *      Inlined as it's very short and called for pretty much every
2010  *      packet ever received.
2011  */
2012 
2013 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2014 {
2015         skb_orphan(skb);
2016         skb->sk = sk;
2017         skb->destructor = sock_wfree;
2018         skb_set_hash_from_sk(skb, sk);
2019         /*
2020          * We used to take a refcount on sk, but following operation
2021          * is enough to guarantee sk_free() wont free this sock until
2022          * all in-flight packets are completed
2023          */
2024         atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2025 }
2026 
2027 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2028 {
2029         skb_orphan(skb);
2030         skb->sk = sk;
2031         skb->destructor = sock_rfree;
2032         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2033         sk_mem_charge(sk, skb->truesize);
2034 }
2035 
2036 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2037                     unsigned long expires);
2038 
2039 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2040 
2041 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2042 
2043 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2044 
2045 /*
2046  *      Recover an error report and clear atomically
2047  */
2048 
2049 static inline int sock_error(struct sock *sk)
2050 {
2051         int err;
2052         if (likely(!sk->sk_err))
2053                 return 0;
2054         err = xchg(&sk->sk_err, 0);
2055         return -err;
2056 }
2057 
2058 static inline unsigned long sock_wspace(struct sock *sk)
2059 {
2060         int amt = 0;
2061 
2062         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2063                 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2064                 if (amt < 0)
2065                         amt = 0;
2066         }
2067         return amt;
2068 }
2069 
2070 static inline void sk_wake_async(struct sock *sk, int how, int band)
2071 {
2072         if (sock_flag(sk, SOCK_FASYNC))
2073                 sock_wake_async(sk->sk_socket, how, band);
2074 }
2075 
2076 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2077  * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2078  * Note: for send buffers, TCP works better if we can build two skbs at
2079  * minimum.
2080  */
2081 #define TCP_SKB_MIN_TRUESIZE    (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2082 
2083 #define SOCK_MIN_SNDBUF         (TCP_SKB_MIN_TRUESIZE * 2)
2084 #define SOCK_MIN_RCVBUF          TCP_SKB_MIN_TRUESIZE
2085 
2086 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2087 {
2088         if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2089                 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2090                 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2091         }
2092 }
2093 
2094 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2095 
2096 /**
2097  * sk_page_frag - return an appropriate page_frag
2098  * @sk: socket
2099  *
2100  * If socket allocation mode allows current thread to sleep, it means its
2101  * safe to use the per task page_frag instead of the per socket one.
2102  */
2103 static inline struct page_frag *sk_page_frag(struct sock *sk)
2104 {
2105         if (sk->sk_allocation & __GFP_WAIT)
2106                 return &current->task_frag;
2107 
2108         return &sk->sk_frag;
2109 }
2110 
2111 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2112 
2113 /*
2114  *      Default write policy as shown to user space via poll/select/SIGIO
2115  */
2116 static inline bool sock_writeable(const struct sock *sk)
2117 {
2118         return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2119 }
2120 
2121 static inline gfp_t gfp_any(void)
2122 {
2123         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2124 }
2125 
2126 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2127 {
2128         return noblock ? 0 : sk->sk_rcvtimeo;
2129 }
2130 
2131 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2132 {
2133         return noblock ? 0 : sk->sk_sndtimeo;
2134 }
2135 
2136 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2137 {
2138         return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2139 }
2140 
2141 /* Alas, with timeout socket operations are not restartable.
2142  * Compare this to poll().
2143  */
2144 static inline int sock_intr_errno(long timeo)
2145 {
2146         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2147 }
2148 
2149 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2150                            struct sk_buff *skb);
2151 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2152                              struct sk_buff *skb);
2153 
2154 static inline void
2155 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2156 {
2157         ktime_t kt = skb->tstamp;
2158         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2159 
2160         /*
2161          * generate control messages if
2162          * - receive time stamping in software requested
2163          * - software time stamp available and wanted
2164          * - hardware time stamps available and wanted
2165          */
2166         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2167             (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2168             (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2169             (hwtstamps->hwtstamp.tv64 &&
2170              (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2171                 __sock_recv_timestamp(msg, sk, skb);
2172         else
2173                 sk->sk_stamp = kt;
2174 
2175         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2176                 __sock_recv_wifi_status(msg, sk, skb);
2177 }
2178 
2179 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2180                               struct sk_buff *skb);
2181 
2182 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2183                                           struct sk_buff *skb)
2184 {
2185 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2186                            (1UL << SOCK_RCVTSTAMP))
2187 #define TSFLAGS_ANY       (SOF_TIMESTAMPING_SOFTWARE                    | \
2188                            SOF_TIMESTAMPING_RAW_HARDWARE)
2189 
2190         if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
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:   completed with instructions for time stamping
2200  *
2201  * Note : callers should take care of initial *tx_flags value (usually 0)
2202  */
2203 void sock_tx_timestamp(const 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         struct net *current_net = sock_net(sk);
2252 
2253         if (!net_eq(current_net, net)) {
2254                 put_net(current_net);
2255                 sock_net_set(sk, hold_net(net));
2256         }
2257 }
2258 
2259 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2260 {
2261         if (skb->sk) {
2262                 struct sock *sk = skb->sk;
2263 
2264                 skb->destructor = NULL;
2265                 skb->sk = NULL;
2266                 return sk;
2267         }
2268         return NULL;
2269 }
2270 
2271 void sock_enable_timestamp(struct sock *sk, int flag);
2272 int sock_get_timestamp(struct sock *, struct timeval __user *);
2273 int sock_get_timestampns(struct sock *, struct timespec __user *);
2274 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2275                        int type);
2276 
2277 bool sk_ns_capable(const struct sock *sk,
2278                    struct user_namespace *user_ns, int cap);
2279 bool sk_capable(const struct sock *sk, int cap);
2280 bool sk_net_capable(const struct sock *sk, int cap);
2281 
2282 /*
2283  *      Enable debug/info messages
2284  */
2285 extern int net_msg_warn;
2286 #define NETDEBUG(fmt, args...) \
2287         do { if (net_msg_warn) printk(fmt,##args); } while (0)
2288 
2289 #define LIMIT_NETDEBUG(fmt, args...) \
2290         do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2291 
2292 extern __u32 sysctl_wmem_max;
2293 extern __u32 sysctl_rmem_max;
2294 
2295 extern int sysctl_optmem_max;
2296 
2297 extern __u32 sysctl_wmem_default;
2298 extern __u32 sysctl_rmem_default;
2299 
2300 #endif  /* _SOCK_H */
2301 

~ [ 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