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

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

Version: ~ [ linux-5.14-rc3 ] ~ [ linux-5.13.5 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.53 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.135 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.198 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.240 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.276 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.276 ] ~ [ 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/page_counter.h>
 58 #include <linux/memcontrol.h>
 59 #include <linux/static_key.h>
 60 #include <linux/sched.h>
 61 #include <linux/wait.h>
 62 #include <linux/cgroup-defs.h>
 63 
 64 #include <linux/filter.h>
 65 #include <linux/rculist_nulls.h>
 66 #include <linux/poll.h>
 67 
 68 #include <linux/atomic.h>
 69 #include <net/dst.h>
 70 #include <net/checksum.h>
 71 #include <net/tcp_states.h>
 72 #include <linux/net_tstamp.h>
 73 
 74 /*
 75  * This structure really needs to be cleaned up.
 76  * Most of it is for TCP, and not used by any of
 77  * the other protocols.
 78  */
 79 
 80 /* Define this to get the SOCK_DBG debugging facility. */
 81 #define SOCK_DEBUGGING
 82 #ifdef SOCK_DEBUGGING
 83 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
 84                                         printk(KERN_DEBUG msg); } while (0)
 85 #else
 86 /* Validate arguments and do nothing */
 87 static inline __printf(2, 3)
 88 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
 89 {
 90 }
 91 #endif
 92 
 93 /* This is the per-socket lock.  The spinlock provides a synchronization
 94  * between user contexts and software interrupt processing, whereas the
 95  * mini-semaphore synchronizes multiple users amongst themselves.
 96  */
 97 typedef struct {
 98         spinlock_t              slock;
 99         int                     owned;
100         wait_queue_head_t       wq;
101         /*
102          * We express the mutex-alike socket_lock semantics
103          * to the lock validator by explicitly managing
104          * the slock as a lock variant (in addition to
105          * the slock itself):
106          */
107 #ifdef CONFIG_DEBUG_LOCK_ALLOC
108         struct lockdep_map dep_map;
109 #endif
110 } socket_lock_t;
111 
112 struct sock;
113 struct proto;
114 struct net;
115 
116 typedef __u32 __bitwise __portpair;
117 typedef __u64 __bitwise __addrpair;
118 
119 /**
120  *      struct sock_common - minimal network layer representation of sockets
121  *      @skc_daddr: Foreign IPv4 addr
122  *      @skc_rcv_saddr: Bound local IPv4 addr
123  *      @skc_hash: hash value used with various protocol lookup tables
124  *      @skc_u16hashes: two u16 hash values used by UDP lookup tables
125  *      @skc_dport: placeholder for inet_dport/tw_dport
126  *      @skc_num: placeholder for inet_num/tw_num
127  *      @skc_family: network address family
128  *      @skc_state: Connection state
129  *      @skc_reuse: %SO_REUSEADDR setting
130  *      @skc_reuseport: %SO_REUSEPORT setting
131  *      @skc_bound_dev_if: bound device index if != 0
132  *      @skc_bind_node: bind hash linkage for various protocol lookup tables
133  *      @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
134  *      @skc_prot: protocol handlers inside a network family
135  *      @skc_net: reference to the network namespace of this socket
136  *      @skc_node: main hash linkage for various protocol lookup tables
137  *      @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
138  *      @skc_tx_queue_mapping: tx queue number for this connection
139  *      @skc_flags: place holder for sk_flags
140  *              %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
141  *              %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
142  *      @skc_incoming_cpu: record/match cpu processing incoming packets
143  *      @skc_refcnt: reference count
144  *
145  *      This is the minimal network layer representation of sockets, the header
146  *      for struct sock and struct inet_timewait_sock.
147  */
148 struct sock_common {
149         /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
150          * address on 64bit arches : cf INET_MATCH()
151          */
152         union {
153                 __addrpair      skc_addrpair;
154                 struct {
155                         __be32  skc_daddr;
156                         __be32  skc_rcv_saddr;
157                 };
158         };
159         union  {
160                 unsigned int    skc_hash;
161                 __u16           skc_u16hashes[2];
162         };
163         /* skc_dport && skc_num must be grouped as well */
164         union {
165                 __portpair      skc_portpair;
166                 struct {
167                         __be16  skc_dport;
168                         __u16   skc_num;
169                 };
170         };
171 
172         unsigned short          skc_family;
173         volatile unsigned char  skc_state;
174         unsigned char           skc_reuse:4;
175         unsigned char           skc_reuseport:1;
176         unsigned char           skc_ipv6only:1;
177         unsigned char           skc_net_refcnt:1;
178         int                     skc_bound_dev_if;
179         union {
180                 struct hlist_node       skc_bind_node;
181                 struct hlist_node       skc_portaddr_node;
182         };
183         struct proto            *skc_prot;
184         possible_net_t          skc_net;
185 
186 #if IS_ENABLED(CONFIG_IPV6)
187         struct in6_addr         skc_v6_daddr;
188         struct in6_addr         skc_v6_rcv_saddr;
189 #endif
190 
191         atomic64_t              skc_cookie;
192 
193         /* following fields are padding to force
194          * offset(struct sock, sk_refcnt) == 128 on 64bit arches
195          * assuming IPV6 is enabled. We use this padding differently
196          * for different kind of 'sockets'
197          */
198         union {
199                 unsigned long   skc_flags;
200                 struct sock     *skc_listener; /* request_sock */
201                 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
202         };
203         /*
204          * fields between dontcopy_begin/dontcopy_end
205          * are not copied in sock_copy()
206          */
207         /* private: */
208         int                     skc_dontcopy_begin[0];
209         /* public: */
210         union {
211                 struct hlist_node       skc_node;
212                 struct hlist_nulls_node skc_nulls_node;
213         };
214         int                     skc_tx_queue_mapping;
215         union {
216                 int             skc_incoming_cpu;
217                 u32             skc_rcv_wnd;
218                 u32             skc_tw_rcv_nxt; /* struct tcp_timewait_sock  */
219         };
220 
221         atomic_t                skc_refcnt;
222         /* private: */
223         int                     skc_dontcopy_end[0];
224         union {
225                 u32             skc_rxhash;
226                 u32             skc_window_clamp;
227                 u32             skc_tw_snd_nxt; /* struct tcp_timewait_sock */
228         };
229         /* public: */
230 };
231 
232 /**
233   *     struct sock - network layer representation of sockets
234   *     @__sk_common: shared layout with inet_timewait_sock
235   *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
236   *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
237   *     @sk_lock:       synchronizer
238   *     @sk_rcvbuf: size of receive buffer in bytes
239   *     @sk_wq: sock wait queue and async head
240   *     @sk_rx_dst: receive input route used by early demux
241   *     @sk_dst_cache: destination cache
242   *     @sk_policy: flow policy
243   *     @sk_receive_queue: incoming packets
244   *     @sk_wmem_alloc: transmit queue bytes committed
245   *     @sk_write_queue: Packet sending queue
246   *     @sk_omem_alloc: "o" is "option" or "other"
247   *     @sk_wmem_queued: persistent queue size
248   *     @sk_forward_alloc: space allocated forward
249   *     @sk_napi_id: id of the last napi context to receive data for sk
250   *     @sk_ll_usec: usecs to busypoll when there is no data
251   *     @sk_allocation: allocation mode
252   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
253   *     @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
254   *     @sk_sndbuf: size of send buffer in bytes
255   *     @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
256   *     @sk_no_check_rx: allow zero checksum in RX packets
257   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
258   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
259   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
260   *     @sk_gso_max_size: Maximum GSO segment size to build
261   *     @sk_gso_max_segs: Maximum number of GSO segments
262   *     @sk_lingertime: %SO_LINGER l_linger setting
263   *     @sk_backlog: always used with the per-socket spinlock held
264   *     @sk_callback_lock: used with the callbacks in the end of this struct
265   *     @sk_error_queue: rarely used
266   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
267   *                       IPV6_ADDRFORM for instance)
268   *     @sk_err: last error
269   *     @sk_err_soft: errors that don't cause failure but are the cause of a
270   *                   persistent failure not just 'timed out'
271   *     @sk_drops: raw/udp drops counter
272   *     @sk_ack_backlog: current listen backlog
273   *     @sk_max_ack_backlog: listen backlog set in listen()
274   *     @sk_priority: %SO_PRIORITY setting
275   *     @sk_type: socket type (%SOCK_STREAM, etc)
276   *     @sk_protocol: which protocol this socket belongs in this network family
277   *     @sk_peer_pid: &struct pid for this socket's peer
278   *     @sk_peer_cred: %SO_PEERCRED setting
279   *     @sk_rcvlowat: %SO_RCVLOWAT setting
280   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
281   *     @sk_sndtimeo: %SO_SNDTIMEO setting
282   *     @sk_txhash: computed flow hash for use on transmit
283   *     @sk_filter: socket filtering instructions
284   *     @sk_timer: sock cleanup timer
285   *     @sk_stamp: time stamp of last packet received
286   *     @sk_tsflags: SO_TIMESTAMPING socket options
287   *     @sk_tskey: counter to disambiguate concurrent tstamp requests
288   *     @sk_socket: Identd and reporting IO signals
289   *     @sk_user_data: RPC layer private data
290   *     @sk_frag: cached page frag
291   *     @sk_peek_off: current peek_offset value
292   *     @sk_send_head: front of stuff to transmit
293   *     @sk_security: used by security modules
294   *     @sk_mark: generic packet mark
295   *     @sk_cgrp_data: cgroup data for this cgroup
296   *     @sk_memcg: this socket's memory cgroup association
297   *     @sk_write_pending: a write to stream socket waits to start
298   *     @sk_state_change: callback to indicate change in the state of the sock
299   *     @sk_data_ready: callback to indicate there is data to be processed
300   *     @sk_write_space: callback to indicate there is bf sending space available
301   *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
302   *     @sk_backlog_rcv: callback to process the backlog
303   *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
304   *     @sk_reuseport_cb: reuseport group container
305  */
306 struct sock {
307         /*
308          * Now struct inet_timewait_sock also uses sock_common, so please just
309          * don't add nothing before this first member (__sk_common) --acme
310          */
311         struct sock_common      __sk_common;
312 #define sk_node                 __sk_common.skc_node
313 #define sk_nulls_node           __sk_common.skc_nulls_node
314 #define sk_refcnt               __sk_common.skc_refcnt
315 #define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
316 
317 #define sk_dontcopy_begin       __sk_common.skc_dontcopy_begin
318 #define sk_dontcopy_end         __sk_common.skc_dontcopy_end
319 #define sk_hash                 __sk_common.skc_hash
320 #define sk_portpair             __sk_common.skc_portpair
321 #define sk_num                  __sk_common.skc_num
322 #define sk_dport                __sk_common.skc_dport
323 #define sk_addrpair             __sk_common.skc_addrpair
324 #define sk_daddr                __sk_common.skc_daddr
325 #define sk_rcv_saddr            __sk_common.skc_rcv_saddr
326 #define sk_family               __sk_common.skc_family
327 #define sk_state                __sk_common.skc_state
328 #define sk_reuse                __sk_common.skc_reuse
329 #define sk_reuseport            __sk_common.skc_reuseport
330 #define sk_ipv6only             __sk_common.skc_ipv6only
331 #define sk_net_refcnt           __sk_common.skc_net_refcnt
332 #define sk_bound_dev_if         __sk_common.skc_bound_dev_if
333 #define sk_bind_node            __sk_common.skc_bind_node
334 #define sk_prot                 __sk_common.skc_prot
335 #define sk_net                  __sk_common.skc_net
336 #define sk_v6_daddr             __sk_common.skc_v6_daddr
337 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
338 #define sk_cookie               __sk_common.skc_cookie
339 #define sk_incoming_cpu         __sk_common.skc_incoming_cpu
340 #define sk_flags                __sk_common.skc_flags
341 #define sk_rxhash               __sk_common.skc_rxhash
342 
343         socket_lock_t           sk_lock;
344         struct sk_buff_head     sk_receive_queue;
345         /*
346          * The backlog queue is special, it is always used with
347          * the per-socket spinlock held and requires low latency
348          * access. Therefore we special case it's implementation.
349          * Note : rmem_alloc is in this structure to fill a hole
350          * on 64bit arches, not because its logically part of
351          * backlog.
352          */
353         struct {
354                 atomic_t        rmem_alloc;
355                 int             len;
356                 struct sk_buff  *head;
357                 struct sk_buff  *tail;
358         } sk_backlog;
359 #define sk_rmem_alloc sk_backlog.rmem_alloc
360         int                     sk_forward_alloc;
361 
362         __u32                   sk_txhash;
363 #ifdef CONFIG_NET_RX_BUSY_POLL
364         unsigned int            sk_napi_id;
365         unsigned int            sk_ll_usec;
366 #endif
367         atomic_t                sk_drops;
368         int                     sk_rcvbuf;
369 
370         struct sk_filter __rcu  *sk_filter;
371         union {
372                 struct socket_wq __rcu  *sk_wq;
373                 struct socket_wq        *sk_wq_raw;
374         };
375 #ifdef CONFIG_XFRM
376         struct xfrm_policy __rcu *sk_policy[2];
377 #endif
378         struct dst_entry        *sk_rx_dst;
379         struct dst_entry __rcu  *sk_dst_cache;
380         /* Note: 32bit hole on 64bit arches */
381         atomic_t                sk_wmem_alloc;
382         atomic_t                sk_omem_alloc;
383         int                     sk_sndbuf;
384         struct sk_buff_head     sk_write_queue;
385 
386         /*
387          * Because of non atomicity rules, all
388          * changes are protected by socket lock.
389          */
390         kmemcheck_bitfield_begin(flags);
391         unsigned int            sk_padding : 2,
392                                 sk_no_check_tx : 1,
393                                 sk_no_check_rx : 1,
394                                 sk_userlocks : 4,
395                                 sk_protocol  : 8,
396                                 sk_type      : 16;
397 #define SK_PROTOCOL_MAX U8_MAX
398         kmemcheck_bitfield_end(flags);
399 
400         int                     sk_wmem_queued;
401         gfp_t                   sk_allocation;
402         u32                     sk_pacing_rate; /* bytes per second */
403         u32                     sk_max_pacing_rate;
404         netdev_features_t       sk_route_caps;
405         netdev_features_t       sk_route_nocaps;
406         int                     sk_gso_type;
407         unsigned int            sk_gso_max_size;
408         u16                     sk_gso_max_segs;
409         int                     sk_rcvlowat;
410         unsigned long           sk_lingertime;
411         struct sk_buff_head     sk_error_queue;
412         struct proto            *sk_prot_creator;
413         rwlock_t                sk_callback_lock;
414         int                     sk_err,
415                                 sk_err_soft;
416         u32                     sk_ack_backlog;
417         u32                     sk_max_ack_backlog;
418         __u32                   sk_priority;
419         __u32                   sk_mark;
420         struct pid              *sk_peer_pid;
421         const struct cred       *sk_peer_cred;
422         long                    sk_rcvtimeo;
423         long                    sk_sndtimeo;
424         struct timer_list       sk_timer;
425         ktime_t                 sk_stamp;
426         u16                     sk_tsflags;
427         u8                      sk_shutdown;
428         u32                     sk_tskey;
429         struct socket           *sk_socket;
430         void                    *sk_user_data;
431         struct page_frag        sk_frag;
432         struct sk_buff          *sk_send_head;
433         __s32                   sk_peek_off;
434         int                     sk_write_pending;
435 #ifdef CONFIG_SECURITY
436         void                    *sk_security;
437 #endif
438         struct sock_cgroup_data sk_cgrp_data;
439         struct mem_cgroup       *sk_memcg;
440         void                    (*sk_state_change)(struct sock *sk);
441         void                    (*sk_data_ready)(struct sock *sk);
442         void                    (*sk_write_space)(struct sock *sk);
443         void                    (*sk_error_report)(struct sock *sk);
444         int                     (*sk_backlog_rcv)(struct sock *sk,
445                                                   struct sk_buff *skb);
446         void                    (*sk_destruct)(struct sock *sk);
447         struct sock_reuseport __rcu     *sk_reuseport_cb;
448         struct rcu_head         sk_rcu;
449 };
450 
451 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
452 
453 #define rcu_dereference_sk_user_data(sk)        rcu_dereference(__sk_user_data((sk)))
454 #define rcu_assign_sk_user_data(sk, ptr)        rcu_assign_pointer(__sk_user_data((sk)), ptr)
455 
456 /*
457  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
458  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
459  * on a socket means that the socket will reuse everybody else's port
460  * without looking at the other's sk_reuse value.
461  */
462 
463 #define SK_NO_REUSE     0
464 #define SK_CAN_REUSE    1
465 #define SK_FORCE_REUSE  2
466 
467 int sk_set_peek_off(struct sock *sk, int val);
468 
469 static inline int sk_peek_offset(struct sock *sk, int flags)
470 {
471         if (unlikely(flags & MSG_PEEK)) {
472                 s32 off = READ_ONCE(sk->sk_peek_off);
473                 if (off >= 0)
474                         return off;
475         }
476 
477         return 0;
478 }
479 
480 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
481 {
482         s32 off = READ_ONCE(sk->sk_peek_off);
483 
484         if (unlikely(off >= 0)) {
485                 off = max_t(s32, off - val, 0);
486                 WRITE_ONCE(sk->sk_peek_off, off);
487         }
488 }
489 
490 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
491 {
492         sk_peek_offset_bwd(sk, -val);
493 }
494 
495 /*
496  * Hashed lists helper routines
497  */
498 static inline struct sock *sk_entry(const struct hlist_node *node)
499 {
500         return hlist_entry(node, struct sock, sk_node);
501 }
502 
503 static inline struct sock *__sk_head(const struct hlist_head *head)
504 {
505         return hlist_entry(head->first, struct sock, sk_node);
506 }
507 
508 static inline struct sock *sk_head(const struct hlist_head *head)
509 {
510         return hlist_empty(head) ? NULL : __sk_head(head);
511 }
512 
513 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
514 {
515         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
516 }
517 
518 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
519 {
520         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
521 }
522 
523 static inline struct sock *sk_next(const struct sock *sk)
524 {
525         return sk->sk_node.next ?
526                 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
527 }
528 
529 static inline struct sock *sk_nulls_next(const struct sock *sk)
530 {
531         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
532                 hlist_nulls_entry(sk->sk_nulls_node.next,
533                                   struct sock, sk_nulls_node) :
534                 NULL;
535 }
536 
537 static inline bool sk_unhashed(const struct sock *sk)
538 {
539         return hlist_unhashed(&sk->sk_node);
540 }
541 
542 static inline bool sk_hashed(const struct sock *sk)
543 {
544         return !sk_unhashed(sk);
545 }
546 
547 static inline void sk_node_init(struct hlist_node *node)
548 {
549         node->pprev = NULL;
550 }
551 
552 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
553 {
554         node->pprev = NULL;
555 }
556 
557 static inline void __sk_del_node(struct sock *sk)
558 {
559         __hlist_del(&sk->sk_node);
560 }
561 
562 /* NB: equivalent to hlist_del_init_rcu */
563 static inline bool __sk_del_node_init(struct sock *sk)
564 {
565         if (sk_hashed(sk)) {
566                 __sk_del_node(sk);
567                 sk_node_init(&sk->sk_node);
568                 return true;
569         }
570         return false;
571 }
572 
573 /* Grab socket reference count. This operation is valid only
574    when sk is ALREADY grabbed f.e. it is found in hash table
575    or a list and the lookup is made under lock preventing hash table
576    modifications.
577  */
578 
579 static __always_inline void sock_hold(struct sock *sk)
580 {
581         atomic_inc(&sk->sk_refcnt);
582 }
583 
584 /* Ungrab socket in the context, which assumes that socket refcnt
585    cannot hit zero, f.e. it is true in context of any socketcall.
586  */
587 static __always_inline void __sock_put(struct sock *sk)
588 {
589         atomic_dec(&sk->sk_refcnt);
590 }
591 
592 static inline bool sk_del_node_init(struct sock *sk)
593 {
594         bool rc = __sk_del_node_init(sk);
595 
596         if (rc) {
597                 /* paranoid for a while -acme */
598                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
599                 __sock_put(sk);
600         }
601         return rc;
602 }
603 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
604 
605 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
606 {
607         if (sk_hashed(sk)) {
608                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
609                 return true;
610         }
611         return false;
612 }
613 
614 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
615 {
616         bool rc = __sk_nulls_del_node_init_rcu(sk);
617 
618         if (rc) {
619                 /* paranoid for a while -acme */
620                 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
621                 __sock_put(sk);
622         }
623         return rc;
624 }
625 
626 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
627 {
628         hlist_add_head(&sk->sk_node, list);
629 }
630 
631 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
632 {
633         sock_hold(sk);
634         __sk_add_node(sk, list);
635 }
636 
637 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
638 {
639         sock_hold(sk);
640         if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
641             sk->sk_family == AF_INET6)
642                 hlist_add_tail_rcu(&sk->sk_node, list);
643         else
644                 hlist_add_head_rcu(&sk->sk_node, list);
645 }
646 
647 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
648 {
649         if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
650             sk->sk_family == AF_INET6)
651                 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
652         else
653                 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
654 }
655 
656 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
657 {
658         sock_hold(sk);
659         __sk_nulls_add_node_rcu(sk, list);
660 }
661 
662 static inline void __sk_del_bind_node(struct sock *sk)
663 {
664         __hlist_del(&sk->sk_bind_node);
665 }
666 
667 static inline void sk_add_bind_node(struct sock *sk,
668                                         struct hlist_head *list)
669 {
670         hlist_add_head(&sk->sk_bind_node, list);
671 }
672 
673 #define sk_for_each(__sk, list) \
674         hlist_for_each_entry(__sk, list, sk_node)
675 #define sk_for_each_rcu(__sk, list) \
676         hlist_for_each_entry_rcu(__sk, list, sk_node)
677 #define sk_nulls_for_each(__sk, node, list) \
678         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
679 #define sk_nulls_for_each_rcu(__sk, node, list) \
680         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
681 #define sk_for_each_from(__sk) \
682         hlist_for_each_entry_from(__sk, sk_node)
683 #define sk_nulls_for_each_from(__sk, node) \
684         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
685                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
686 #define sk_for_each_safe(__sk, tmp, list) \
687         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
688 #define sk_for_each_bound(__sk, list) \
689         hlist_for_each_entry(__sk, list, sk_bind_node)
690 
691 /**
692  * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
693  * @tpos:       the type * to use as a loop cursor.
694  * @pos:        the &struct hlist_node to use as a loop cursor.
695  * @head:       the head for your list.
696  * @offset:     offset of hlist_node within the struct.
697  *
698  */
699 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset)                  \
700         for (pos = rcu_dereference((head)->first);                             \
701              pos != NULL &&                                                    \
702                 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;});       \
703              pos = rcu_dereference(pos->next))
704 
705 static inline struct user_namespace *sk_user_ns(struct sock *sk)
706 {
707         /* Careful only use this in a context where these parameters
708          * can not change and must all be valid, such as recvmsg from
709          * userspace.
710          */
711         return sk->sk_socket->file->f_cred->user_ns;
712 }
713 
714 /* Sock flags */
715 enum sock_flags {
716         SOCK_DEAD,
717         SOCK_DONE,
718         SOCK_URGINLINE,
719         SOCK_KEEPOPEN,
720         SOCK_LINGER,
721         SOCK_DESTROY,
722         SOCK_BROADCAST,
723         SOCK_TIMESTAMP,
724         SOCK_ZAPPED,
725         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
726         SOCK_DBG, /* %SO_DEBUG setting */
727         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
728         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
729         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
730         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
731         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
732         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
733         SOCK_FASYNC, /* fasync() active */
734         SOCK_RXQ_OVFL,
735         SOCK_ZEROCOPY, /* buffers from userspace */
736         SOCK_WIFI_STATUS, /* push wifi status to userspace */
737         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
738                      * Will use last 4 bytes of packet sent from
739                      * user-space instead.
740                      */
741         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
742         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
743         SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
744 };
745 
746 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
747 
748 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
749 {
750         nsk->sk_flags = osk->sk_flags;
751 }
752 
753 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
754 {
755         __set_bit(flag, &sk->sk_flags);
756 }
757 
758 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
759 {
760         __clear_bit(flag, &sk->sk_flags);
761 }
762 
763 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
764 {
765         return test_bit(flag, &sk->sk_flags);
766 }
767 
768 #ifdef CONFIG_NET
769 extern struct static_key memalloc_socks;
770 static inline int sk_memalloc_socks(void)
771 {
772         return static_key_false(&memalloc_socks);
773 }
774 #else
775 
776 static inline int sk_memalloc_socks(void)
777 {
778         return 0;
779 }
780 
781 #endif
782 
783 static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
784 {
785         return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
786 }
787 
788 static inline void sk_acceptq_removed(struct sock *sk)
789 {
790         sk->sk_ack_backlog--;
791 }
792 
793 static inline void sk_acceptq_added(struct sock *sk)
794 {
795         sk->sk_ack_backlog++;
796 }
797 
798 static inline bool sk_acceptq_is_full(const struct sock *sk)
799 {
800         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
801 }
802 
803 /*
804  * Compute minimal free write space needed to queue new packets.
805  */
806 static inline int sk_stream_min_wspace(const struct sock *sk)
807 {
808         return sk->sk_wmem_queued >> 1;
809 }
810 
811 static inline int sk_stream_wspace(const struct sock *sk)
812 {
813         return sk->sk_sndbuf - sk->sk_wmem_queued;
814 }
815 
816 void sk_stream_write_space(struct sock *sk);
817 
818 /* OOB backlog add */
819 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
820 {
821         /* dont let skb dst not refcounted, we are going to leave rcu lock */
822         skb_dst_force_safe(skb);
823 
824         if (!sk->sk_backlog.tail)
825                 sk->sk_backlog.head = skb;
826         else
827                 sk->sk_backlog.tail->next = skb;
828 
829         sk->sk_backlog.tail = skb;
830         skb->next = NULL;
831 }
832 
833 /*
834  * Take into account size of receive queue and backlog queue
835  * Do not take into account this skb truesize,
836  * to allow even a single big packet to come.
837  */
838 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
839 {
840         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
841 
842         return qsize > limit;
843 }
844 
845 /* The per-socket spinlock must be held here. */
846 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
847                                               unsigned int limit)
848 {
849         if (sk_rcvqueues_full(sk, limit))
850                 return -ENOBUFS;
851 
852         /*
853          * If the skb was allocated from pfmemalloc reserves, only
854          * allow SOCK_MEMALLOC sockets to use it as this socket is
855          * helping free memory
856          */
857         if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
858                 return -ENOMEM;
859 
860         __sk_add_backlog(sk, skb);
861         sk->sk_backlog.len += skb->truesize;
862         return 0;
863 }
864 
865 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
866 
867 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
868 {
869         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
870                 return __sk_backlog_rcv(sk, skb);
871 
872         return sk->sk_backlog_rcv(sk, skb);
873 }
874 
875 static inline void sk_incoming_cpu_update(struct sock *sk)
876 {
877         sk->sk_incoming_cpu = raw_smp_processor_id();
878 }
879 
880 static inline void sock_rps_record_flow_hash(__u32 hash)
881 {
882 #ifdef CONFIG_RPS
883         struct rps_sock_flow_table *sock_flow_table;
884 
885         rcu_read_lock();
886         sock_flow_table = rcu_dereference(rps_sock_flow_table);
887         rps_record_sock_flow(sock_flow_table, hash);
888         rcu_read_unlock();
889 #endif
890 }
891 
892 static inline void sock_rps_record_flow(const struct sock *sk)
893 {
894 #ifdef CONFIG_RPS
895         sock_rps_record_flow_hash(sk->sk_rxhash);
896 #endif
897 }
898 
899 static inline void sock_rps_save_rxhash(struct sock *sk,
900                                         const struct sk_buff *skb)
901 {
902 #ifdef CONFIG_RPS
903         if (unlikely(sk->sk_rxhash != skb->hash))
904                 sk->sk_rxhash = skb->hash;
905 #endif
906 }
907 
908 static inline void sock_rps_reset_rxhash(struct sock *sk)
909 {
910 #ifdef CONFIG_RPS
911         sk->sk_rxhash = 0;
912 #endif
913 }
914 
915 #define sk_wait_event(__sk, __timeo, __condition)                       \
916         ({      int __rc;                                               \
917                 release_sock(__sk);                                     \
918                 __rc = __condition;                                     \
919                 if (!__rc) {                                            \
920                         *(__timeo) = schedule_timeout(*(__timeo));      \
921                 }                                                       \
922                 sched_annotate_sleep();                                         \
923                 lock_sock(__sk);                                        \
924                 __rc = __condition;                                     \
925                 __rc;                                                   \
926         })
927 
928 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
929 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
930 void sk_stream_wait_close(struct sock *sk, long timeo_p);
931 int sk_stream_error(struct sock *sk, int flags, int err);
932 void sk_stream_kill_queues(struct sock *sk);
933 void sk_set_memalloc(struct sock *sk);
934 void sk_clear_memalloc(struct sock *sk);
935 
936 void __sk_flush_backlog(struct sock *sk);
937 
938 static inline bool sk_flush_backlog(struct sock *sk)
939 {
940         if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
941                 __sk_flush_backlog(sk);
942                 return true;
943         }
944         return false;
945 }
946 
947 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
948 
949 struct request_sock_ops;
950 struct timewait_sock_ops;
951 struct inet_hashinfo;
952 struct raw_hashinfo;
953 struct module;
954 
955 /*
956  * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
957  * un-modified. Special care is taken when initializing object to zero.
958  */
959 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
960 {
961         if (offsetof(struct sock, sk_node.next) != 0)
962                 memset(sk, 0, offsetof(struct sock, sk_node.next));
963         memset(&sk->sk_node.pprev, 0,
964                size - offsetof(struct sock, sk_node.pprev));
965 }
966 
967 /* Networking protocol blocks we attach to sockets.
968  * socket layer -> transport layer interface
969  */
970 struct proto {
971         void                    (*close)(struct sock *sk,
972                                         long timeout);
973         int                     (*connect)(struct sock *sk,
974                                         struct sockaddr *uaddr,
975                                         int addr_len);
976         int                     (*disconnect)(struct sock *sk, int flags);
977 
978         struct sock *           (*accept)(struct sock *sk, int flags, int *err);
979 
980         int                     (*ioctl)(struct sock *sk, int cmd,
981                                          unsigned long arg);
982         int                     (*init)(struct sock *sk);
983         void                    (*destroy)(struct sock *sk);
984         void                    (*shutdown)(struct sock *sk, int how);
985         int                     (*setsockopt)(struct sock *sk, int level,
986                                         int optname, char __user *optval,
987                                         unsigned int optlen);
988         int                     (*getsockopt)(struct sock *sk, int level,
989                                         int optname, char __user *optval,
990                                         int __user *option);
991 #ifdef CONFIG_COMPAT
992         int                     (*compat_setsockopt)(struct sock *sk,
993                                         int level,
994                                         int optname, char __user *optval,
995                                         unsigned int optlen);
996         int                     (*compat_getsockopt)(struct sock *sk,
997                                         int level,
998                                         int optname, char __user *optval,
999                                         int __user *option);
1000         int                     (*compat_ioctl)(struct sock *sk,
1001                                         unsigned int cmd, unsigned long arg);
1002 #endif
1003         int                     (*sendmsg)(struct sock *sk, struct msghdr *msg,
1004                                            size_t len);
1005         int                     (*recvmsg)(struct sock *sk, struct msghdr *msg,
1006                                            size_t len, int noblock, int flags,
1007                                            int *addr_len);
1008         int                     (*sendpage)(struct sock *sk, struct page *page,
1009                                         int offset, size_t size, int flags);
1010         int                     (*bind)(struct sock *sk,
1011                                         struct sockaddr *uaddr, int addr_len);
1012 
1013         int                     (*backlog_rcv) (struct sock *sk,
1014                                                 struct sk_buff *skb);
1015 
1016         void            (*release_cb)(struct sock *sk);
1017 
1018         /* Keeping track of sk's, looking them up, and port selection methods. */
1019         int                     (*hash)(struct sock *sk);
1020         void                    (*unhash)(struct sock *sk);
1021         void                    (*rehash)(struct sock *sk);
1022         int                     (*get_port)(struct sock *sk, unsigned short snum);
1023         void                    (*clear_sk)(struct sock *sk, int size);
1024 
1025         /* Keeping track of sockets in use */
1026 #ifdef CONFIG_PROC_FS
1027         unsigned int            inuse_idx;
1028 #endif
1029 
1030         bool                    (*stream_memory_free)(const struct sock *sk);
1031         /* Memory pressure */
1032         void                    (*enter_memory_pressure)(struct sock *sk);
1033         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
1034         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
1035         /*
1036          * Pressure flag: try to collapse.
1037          * Technical note: it is used by multiple contexts non atomically.
1038          * All the __sk_mem_schedule() is of this nature: accounting
1039          * is strict, actions are advisory and have some latency.
1040          */
1041         int                     *memory_pressure;
1042         long                    *sysctl_mem;
1043         int                     *sysctl_wmem;
1044         int                     *sysctl_rmem;
1045         int                     max_header;
1046         bool                    no_autobind;
1047 
1048         struct kmem_cache       *slab;
1049         unsigned int            obj_size;
1050         int                     slab_flags;
1051 
1052         struct percpu_counter   *orphan_count;
1053 
1054         struct request_sock_ops *rsk_prot;
1055         struct timewait_sock_ops *twsk_prot;
1056 
1057         union {
1058                 struct inet_hashinfo    *hashinfo;
1059                 struct udp_table        *udp_table;
1060                 struct raw_hashinfo     *raw_hash;
1061         } h;
1062 
1063         struct module           *owner;
1064 
1065         char                    name[32];
1066 
1067         struct list_head        node;
1068 #ifdef SOCK_REFCNT_DEBUG
1069         atomic_t                socks;
1070 #endif
1071         int                     (*diag_destroy)(struct sock *sk, int err);
1072 };
1073 
1074 int proto_register(struct proto *prot, int alloc_slab);
1075 void proto_unregister(struct proto *prot);
1076 
1077 #ifdef SOCK_REFCNT_DEBUG
1078 static inline void sk_refcnt_debug_inc(struct sock *sk)
1079 {
1080         atomic_inc(&sk->sk_prot->socks);
1081 }
1082 
1083 static inline void sk_refcnt_debug_dec(struct sock *sk)
1084 {
1085         atomic_dec(&sk->sk_prot->socks);
1086         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1087                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1088 }
1089 
1090 static inline void sk_refcnt_debug_release(const struct sock *sk)
1091 {
1092         if (atomic_read(&sk->sk_refcnt) != 1)
1093                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1094                        sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1095 }
1096 #else /* SOCK_REFCNT_DEBUG */
1097 #define sk_refcnt_debug_inc(sk) do { } while (0)
1098 #define sk_refcnt_debug_dec(sk) do { } while (0)
1099 #define sk_refcnt_debug_release(sk) do { } while (0)
1100 #endif /* SOCK_REFCNT_DEBUG */
1101 
1102 static inline bool sk_stream_memory_free(const struct sock *sk)
1103 {
1104         if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1105                 return false;
1106 
1107         return sk->sk_prot->stream_memory_free ?
1108                 sk->sk_prot->stream_memory_free(sk) : true;
1109 }
1110 
1111 static inline bool sk_stream_is_writeable(const struct sock *sk)
1112 {
1113         return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1114                sk_stream_memory_free(sk);
1115 }
1116 
1117 
1118 static inline bool sk_has_memory_pressure(const struct sock *sk)
1119 {
1120         return sk->sk_prot->memory_pressure != NULL;
1121 }
1122 
1123 static inline bool sk_under_memory_pressure(const struct sock *sk)
1124 {
1125         if (!sk->sk_prot->memory_pressure)
1126                 return false;
1127 
1128         if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1129             mem_cgroup_under_socket_pressure(sk->sk_memcg))
1130                 return true;
1131 
1132         return !!*sk->sk_prot->memory_pressure;
1133 }
1134 
1135 static inline void sk_leave_memory_pressure(struct sock *sk)
1136 {
1137         int *memory_pressure = sk->sk_prot->memory_pressure;
1138 
1139         if (!memory_pressure)
1140                 return;
1141 
1142         if (*memory_pressure)
1143                 *memory_pressure = 0;
1144 }
1145 
1146 static inline void sk_enter_memory_pressure(struct sock *sk)
1147 {
1148         if (!sk->sk_prot->enter_memory_pressure)
1149                 return;
1150 
1151         sk->sk_prot->enter_memory_pressure(sk);
1152 }
1153 
1154 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1155 {
1156         return sk->sk_prot->sysctl_mem[index];
1157 }
1158 
1159 static inline long
1160 sk_memory_allocated(const struct sock *sk)
1161 {
1162         return atomic_long_read(sk->sk_prot->memory_allocated);
1163 }
1164 
1165 static inline long
1166 sk_memory_allocated_add(struct sock *sk, int amt)
1167 {
1168         return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
1169 }
1170 
1171 static inline void
1172 sk_memory_allocated_sub(struct sock *sk, int amt)
1173 {
1174         atomic_long_sub(amt, sk->sk_prot->memory_allocated);
1175 }
1176 
1177 static inline void sk_sockets_allocated_dec(struct sock *sk)
1178 {
1179         percpu_counter_dec(sk->sk_prot->sockets_allocated);
1180 }
1181 
1182 static inline void sk_sockets_allocated_inc(struct sock *sk)
1183 {
1184         percpu_counter_inc(sk->sk_prot->sockets_allocated);
1185 }
1186 
1187 static inline int
1188 sk_sockets_allocated_read_positive(struct sock *sk)
1189 {
1190         return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
1191 }
1192 
1193 static inline int
1194 proto_sockets_allocated_sum_positive(struct proto *prot)
1195 {
1196         return percpu_counter_sum_positive(prot->sockets_allocated);
1197 }
1198 
1199 static inline long
1200 proto_memory_allocated(struct proto *prot)
1201 {
1202         return atomic_long_read(prot->memory_allocated);
1203 }
1204 
1205 static inline bool
1206 proto_memory_pressure(struct proto *prot)
1207 {
1208         if (!prot->memory_pressure)
1209                 return false;
1210         return !!*prot->memory_pressure;
1211 }
1212 
1213 
1214 #ifdef CONFIG_PROC_FS
1215 /* Called with local bh disabled */
1216 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1217 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1218 #else
1219 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1220                 int inc)
1221 {
1222 }
1223 #endif
1224 
1225 
1226 /* With per-bucket locks this operation is not-atomic, so that
1227  * this version is not worse.
1228  */
1229 static inline int __sk_prot_rehash(struct sock *sk)
1230 {
1231         sk->sk_prot->unhash(sk);
1232         return sk->sk_prot->hash(sk);
1233 }
1234 
1235 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1236 
1237 /* About 10 seconds */
1238 #define SOCK_DESTROY_TIME (10*HZ)
1239 
1240 /* Sockets 0-1023 can't be bound to unless you are superuser */
1241 #define PROT_SOCK       1024
1242 
1243 #define SHUTDOWN_MASK   3
1244 #define RCV_SHUTDOWN    1
1245 #define SEND_SHUTDOWN   2
1246 
1247 #define SOCK_SNDBUF_LOCK        1
1248 #define SOCK_RCVBUF_LOCK        2
1249 #define SOCK_BINDADDR_LOCK      4
1250 #define SOCK_BINDPORT_LOCK      8
1251 
1252 struct socket_alloc {
1253         struct socket socket;
1254         struct inode vfs_inode;
1255 };
1256 
1257 static inline struct socket *SOCKET_I(struct inode *inode)
1258 {
1259         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1260 }
1261 
1262 static inline struct inode *SOCK_INODE(struct socket *socket)
1263 {
1264         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1265 }
1266 
1267 /*
1268  * Functions for memory accounting
1269  */
1270 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1271 void __sk_mem_reclaim(struct sock *sk, int amount);
1272 
1273 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1274 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1275 #define SK_MEM_SEND     0
1276 #define SK_MEM_RECV     1
1277 
1278 static inline int sk_mem_pages(int amt)
1279 {
1280         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1281 }
1282 
1283 static inline bool sk_has_account(struct sock *sk)
1284 {
1285         /* return true if protocol supports memory accounting */
1286         return !!sk->sk_prot->memory_allocated;
1287 }
1288 
1289 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1290 {
1291         if (!sk_has_account(sk))
1292                 return true;
1293         return size <= sk->sk_forward_alloc ||
1294                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1295 }
1296 
1297 static inline bool
1298 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1299 {
1300         if (!sk_has_account(sk))
1301                 return true;
1302         return size<= sk->sk_forward_alloc ||
1303                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1304                 skb_pfmemalloc(skb);
1305 }
1306 
1307 static inline void sk_mem_reclaim(struct sock *sk)
1308 {
1309         if (!sk_has_account(sk))
1310                 return;
1311         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1312                 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
1313 }
1314 
1315 static inline void sk_mem_reclaim_partial(struct sock *sk)
1316 {
1317         if (!sk_has_account(sk))
1318                 return;
1319         if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1320                 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
1321 }
1322 
1323 static inline void sk_mem_charge(struct sock *sk, int size)
1324 {
1325         if (!sk_has_account(sk))
1326                 return;
1327         sk->sk_forward_alloc -= size;
1328 }
1329 
1330 static inline void sk_mem_uncharge(struct sock *sk, int size)
1331 {
1332         if (!sk_has_account(sk))
1333                 return;
1334         sk->sk_forward_alloc += size;
1335 }
1336 
1337 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1338 {
1339         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1340         sk->sk_wmem_queued -= skb->truesize;
1341         sk_mem_uncharge(sk, skb->truesize);
1342         __kfree_skb(skb);
1343 }
1344 
1345 static inline void sock_release_ownership(struct sock *sk)
1346 {
1347         if (sk->sk_lock.owned) {
1348                 sk->sk_lock.owned = 0;
1349 
1350                 /* The sk_lock has mutex_unlock() semantics: */
1351                 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1352         }
1353 }
1354 
1355 /*
1356  * Macro so as to not evaluate some arguments when
1357  * lockdep is not enabled.
1358  *
1359  * Mark both the sk_lock and the sk_lock.slock as a
1360  * per-address-family lock class.
1361  */
1362 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1363 do {                                                                    \
1364         sk->sk_lock.owned = 0;                                          \
1365         init_waitqueue_head(&sk->sk_lock.wq);                           \
1366         spin_lock_init(&(sk)->sk_lock.slock);                           \
1367         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1368                         sizeof((sk)->sk_lock));                         \
1369         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1370                                 (skey), (sname));                               \
1371         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1372 } while (0)
1373 
1374 #ifdef CONFIG_LOCKDEP
1375 static inline bool lockdep_sock_is_held(const struct sock *csk)
1376 {
1377         struct sock *sk = (struct sock *)csk;
1378 
1379         return lockdep_is_held(&sk->sk_lock) ||
1380                lockdep_is_held(&sk->sk_lock.slock);
1381 }
1382 #endif
1383 
1384 void lock_sock_nested(struct sock *sk, int subclass);
1385 
1386 static inline void lock_sock(struct sock *sk)
1387 {
1388         lock_sock_nested(sk, 0);
1389 }
1390 
1391 void release_sock(struct sock *sk);
1392 
1393 /* BH context may only use the following locking interface. */
1394 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1395 #define bh_lock_sock_nested(__sk) \
1396                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1397                                 SINGLE_DEPTH_NESTING)
1398 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1399 
1400 bool lock_sock_fast(struct sock *sk);
1401 /**
1402  * unlock_sock_fast - complement of lock_sock_fast
1403  * @sk: socket
1404  * @slow: slow mode
1405  *
1406  * fast unlock socket for user context.
1407  * If slow mode is on, we call regular release_sock()
1408  */
1409 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1410 {
1411         if (slow)
1412                 release_sock(sk);
1413         else
1414                 spin_unlock_bh(&sk->sk_lock.slock);
1415 }
1416 
1417 /* Used by processes to "lock" a socket state, so that
1418  * interrupts and bottom half handlers won't change it
1419  * from under us. It essentially blocks any incoming
1420  * packets, so that we won't get any new data or any
1421  * packets that change the state of the socket.
1422  *
1423  * While locked, BH processing will add new packets to
1424  * the backlog queue.  This queue is processed by the
1425  * owner of the socket lock right before it is released.
1426  *
1427  * Since ~2.3.5 it is also exclusive sleep lock serializing
1428  * accesses from user process context.
1429  */
1430 
1431 static inline void sock_owned_by_me(const struct sock *sk)
1432 {
1433 #ifdef CONFIG_LOCKDEP
1434         WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
1435 #endif
1436 }
1437 
1438 static inline bool sock_owned_by_user(const struct sock *sk)
1439 {
1440         sock_owned_by_me(sk);
1441         return sk->sk_lock.owned;
1442 }
1443 
1444 /* no reclassification while locks are held */
1445 static inline bool sock_allow_reclassification(const struct sock *csk)
1446 {
1447         struct sock *sk = (struct sock *)csk;
1448 
1449         return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1450 }
1451 
1452 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1453                       struct proto *prot, int kern);
1454 void sk_free(struct sock *sk);
1455 void sk_destruct(struct sock *sk);
1456 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1457 
1458 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1459                              gfp_t priority);
1460 void __sock_wfree(struct sk_buff *skb);
1461 void sock_wfree(struct sk_buff *skb);
1462 void skb_orphan_partial(struct sk_buff *skb);
1463 void sock_rfree(struct sk_buff *skb);
1464 void sock_efree(struct sk_buff *skb);
1465 #ifdef CONFIG_INET
1466 void sock_edemux(struct sk_buff *skb);
1467 #else
1468 #define sock_edemux(skb) sock_efree(skb)
1469 #endif
1470 
1471 int sock_setsockopt(struct socket *sock, int level, int op,
1472                     char __user *optval, unsigned int optlen);
1473 
1474 int sock_getsockopt(struct socket *sock, int level, int op,
1475                     char __user *optval, int __user *optlen);
1476 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1477                                     int noblock, int *errcode);
1478 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1479                                      unsigned long data_len, int noblock,
1480                                      int *errcode, int max_page_order);
1481 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1482 void sock_kfree_s(struct sock *sk, void *mem, int size);
1483 void sock_kzfree_s(struct sock *sk, void *mem, int size);
1484 void sk_send_sigurg(struct sock *sk);
1485 
1486 struct sockcm_cookie {
1487         u32 mark;
1488         u16 tsflags;
1489 };
1490 
1491 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1492                      struct sockcm_cookie *sockc);
1493 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1494                    struct sockcm_cookie *sockc);
1495 
1496 /*
1497  * Functions to fill in entries in struct proto_ops when a protocol
1498  * does not implement a particular function.
1499  */
1500 int sock_no_bind(struct socket *, struct sockaddr *, int);
1501 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1502 int sock_no_socketpair(struct socket *, struct socket *);
1503 int sock_no_accept(struct socket *, struct socket *, int);
1504 int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1505 unsigned int sock_no_poll(struct file *, struct socket *,
1506                           struct poll_table_struct *);
1507 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1508 int sock_no_listen(struct socket *, int);
1509 int sock_no_shutdown(struct socket *, int);
1510 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1511 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1512 int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1513 int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
1514 int sock_no_mmap(struct file *file, struct socket *sock,
1515                  struct vm_area_struct *vma);
1516 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1517                          size_t size, int flags);
1518 
1519 /*
1520  * Functions to fill in entries in struct proto_ops when a protocol
1521  * uses the inet style.
1522  */
1523 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1524                                   char __user *optval, int __user *optlen);
1525 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1526                         int flags);
1527 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1528                                   char __user *optval, unsigned int optlen);
1529 int compat_sock_common_getsockopt(struct socket *sock, int level,
1530                 int optname, char __user *optval, int __user *optlen);
1531 int compat_sock_common_setsockopt(struct socket *sock, int level,
1532                 int optname, char __user *optval, unsigned int optlen);
1533 
1534 void sk_common_release(struct sock *sk);
1535 
1536 /*
1537  *      Default socket callbacks and setup code
1538  */
1539 
1540 /* Initialise core socket variables */
1541 void sock_init_data(struct socket *sock, struct sock *sk);
1542 
1543 /*
1544  * Socket reference counting postulates.
1545  *
1546  * * Each user of socket SHOULD hold a reference count.
1547  * * Each access point to socket (an hash table bucket, reference from a list,
1548  *   running timer, skb in flight MUST hold a reference count.
1549  * * When reference count hits 0, it means it will never increase back.
1550  * * When reference count hits 0, it means that no references from
1551  *   outside exist to this socket and current process on current CPU
1552  *   is last user and may/should destroy this socket.
1553  * * sk_free is called from any context: process, BH, IRQ. When
1554  *   it is called, socket has no references from outside -> sk_free
1555  *   may release descendant resources allocated by the socket, but
1556  *   to the time when it is called, socket is NOT referenced by any
1557  *   hash tables, lists etc.
1558  * * Packets, delivered from outside (from network or from another process)
1559  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1560  *   when they sit in queue. Otherwise, packets will leak to hole, when
1561  *   socket is looked up by one cpu and unhasing is made by another CPU.
1562  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1563  *   (leak to backlog). Packet socket does all the processing inside
1564  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1565  *   use separate SMP lock, so that they are prone too.
1566  */
1567 
1568 /* Ungrab socket and destroy it, if it was the last reference. */
1569 static inline void sock_put(struct sock *sk)
1570 {
1571         if (atomic_dec_and_test(&sk->sk_refcnt))
1572                 sk_free(sk);
1573 }
1574 /* Generic version of sock_put(), dealing with all sockets
1575  * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1576  */
1577 void sock_gen_put(struct sock *sk);
1578 
1579 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
1580                      unsigned int trim_cap);
1581 static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1582                                  const int nested)
1583 {
1584         return __sk_receive_skb(sk, skb, nested, 1);
1585 }
1586 
1587 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1588 {
1589         sk->sk_tx_queue_mapping = tx_queue;
1590 }
1591 
1592 static inline void sk_tx_queue_clear(struct sock *sk)
1593 {
1594         sk->sk_tx_queue_mapping = -1;
1595 }
1596 
1597 static inline int sk_tx_queue_get(const struct sock *sk)
1598 {
1599         return sk ? sk->sk_tx_queue_mapping : -1;
1600 }
1601 
1602 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1603 {
1604         sk_tx_queue_clear(sk);
1605         sk->sk_socket = sock;
1606 }
1607 
1608 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1609 {
1610         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1611         return &rcu_dereference_raw(sk->sk_wq)->wait;
1612 }
1613 /* Detach socket from process context.
1614  * Announce socket dead, detach it from wait queue and inode.
1615  * Note that parent inode held reference count on this struct sock,
1616  * we do not release it in this function, because protocol
1617  * probably wants some additional cleanups or even continuing
1618  * to work with this socket (TCP).
1619  */
1620 static inline void sock_orphan(struct sock *sk)
1621 {
1622         write_lock_bh(&sk->sk_callback_lock);
1623         sock_set_flag(sk, SOCK_DEAD);
1624         sk_set_socket(sk, NULL);
1625         sk->sk_wq  = NULL;
1626         write_unlock_bh(&sk->sk_callback_lock);
1627 }
1628 
1629 static inline void sock_graft(struct sock *sk, struct socket *parent)
1630 {
1631         write_lock_bh(&sk->sk_callback_lock);
1632         sk->sk_wq = parent->wq;
1633         parent->sk = sk;
1634         sk_set_socket(sk, parent);
1635         security_sock_graft(sk, parent);
1636         write_unlock_bh(&sk->sk_callback_lock);
1637 }
1638 
1639 kuid_t sock_i_uid(struct sock *sk);
1640 unsigned long sock_i_ino(struct sock *sk);
1641 
1642 static inline u32 net_tx_rndhash(void)
1643 {
1644         u32 v = prandom_u32();
1645 
1646         return v ?: 1;
1647 }
1648 
1649 static inline void sk_set_txhash(struct sock *sk)
1650 {
1651         sk->sk_txhash = net_tx_rndhash();
1652 }
1653 
1654 static inline void sk_rethink_txhash(struct sock *sk)
1655 {
1656         if (sk->sk_txhash)
1657                 sk_set_txhash(sk);
1658 }
1659 
1660 static inline struct dst_entry *
1661 __sk_dst_get(struct sock *sk)
1662 {
1663         return rcu_dereference_check(sk->sk_dst_cache,
1664                                      lockdep_sock_is_held(sk));
1665 }
1666 
1667 static inline struct dst_entry *
1668 sk_dst_get(struct sock *sk)
1669 {
1670         struct dst_entry *dst;
1671 
1672         rcu_read_lock();
1673         dst = rcu_dereference(sk->sk_dst_cache);
1674         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1675                 dst = NULL;
1676         rcu_read_unlock();
1677         return dst;
1678 }
1679 
1680 static inline void dst_negative_advice(struct sock *sk)
1681 {
1682         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1683 
1684         sk_rethink_txhash(sk);
1685 
1686         if (dst && dst->ops->negative_advice) {
1687                 ndst = dst->ops->negative_advice(dst);
1688 
1689                 if (ndst != dst) {
1690                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1691                         sk_tx_queue_clear(sk);
1692                 }
1693         }
1694 }
1695 
1696 static inline void
1697 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1698 {
1699         struct dst_entry *old_dst;
1700 
1701         sk_tx_queue_clear(sk);
1702         /*
1703          * This can be called while sk is owned by the caller only,
1704          * with no state that can be checked in a rcu_dereference_check() cond
1705          */
1706         old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1707         rcu_assign_pointer(sk->sk_dst_cache, dst);
1708         dst_release(old_dst);
1709 }
1710 
1711 static inline void
1712 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1713 {
1714         struct dst_entry *old_dst;
1715 
1716         sk_tx_queue_clear(sk);
1717         old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1718         dst_release(old_dst);
1719 }
1720 
1721 static inline void
1722 __sk_dst_reset(struct sock *sk)
1723 {
1724         __sk_dst_set(sk, NULL);
1725 }
1726 
1727 static inline void
1728 sk_dst_reset(struct sock *sk)
1729 {
1730         sk_dst_set(sk, NULL);
1731 }
1732 
1733 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1734 
1735 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1736 
1737 bool sk_mc_loop(struct sock *sk);
1738 
1739 static inline bool sk_can_gso(const struct sock *sk)
1740 {
1741         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1742 }
1743 
1744 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1745 
1746 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1747 {
1748         sk->sk_route_nocaps |= flags;
1749         sk->sk_route_caps &= ~flags;
1750 }
1751 
1752 static inline bool sk_check_csum_caps(struct sock *sk)
1753 {
1754         return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1755                (sk->sk_family == PF_INET &&
1756                 (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1757                (sk->sk_family == PF_INET6 &&
1758                 (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1759 }
1760 
1761 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1762                                            struct iov_iter *from, char *to,
1763                                            int copy, int offset)
1764 {
1765         if (skb->ip_summed == CHECKSUM_NONE) {
1766                 __wsum csum = 0;
1767                 if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1768                         return -EFAULT;
1769                 skb->csum = csum_block_add(skb->csum, csum, offset);
1770         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1771                 if (copy_from_iter_nocache(to, copy, from) != copy)
1772                         return -EFAULT;
1773         } else if (copy_from_iter(to, copy, from) != copy)
1774                 return -EFAULT;
1775 
1776         return 0;
1777 }
1778 
1779 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1780                                        struct iov_iter *from, int copy)
1781 {
1782         int err, offset = skb->len;
1783 
1784         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1785                                        copy, offset);
1786         if (err)
1787                 __skb_trim(skb, offset);
1788 
1789         return err;
1790 }
1791 
1792 static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
1793                                            struct sk_buff *skb,
1794                                            struct page *page,
1795                                            int off, int copy)
1796 {
1797         int err;
1798 
1799         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1800                                        copy, skb->len);
1801         if (err)
1802                 return err;
1803 
1804         skb->len             += copy;
1805         skb->data_len        += copy;
1806         skb->truesize        += copy;
1807         sk->sk_wmem_queued   += copy;
1808         sk_mem_charge(sk, copy);
1809         return 0;
1810 }
1811 
1812 /**
1813  * sk_wmem_alloc_get - returns write allocations
1814  * @sk: socket
1815  *
1816  * Returns sk_wmem_alloc minus initial offset of one
1817  */
1818 static inline int sk_wmem_alloc_get(const struct sock *sk)
1819 {
1820         return atomic_read(&sk->sk_wmem_alloc) - 1;
1821 }
1822 
1823 /**
1824  * sk_rmem_alloc_get - returns read allocations
1825  * @sk: socket
1826  *
1827  * Returns sk_rmem_alloc
1828  */
1829 static inline int sk_rmem_alloc_get(const struct sock *sk)
1830 {
1831         return atomic_read(&sk->sk_rmem_alloc);
1832 }
1833 
1834 /**
1835  * sk_has_allocations - check if allocations are outstanding
1836  * @sk: socket
1837  *
1838  * Returns true if socket has write or read allocations
1839  */
1840 static inline bool sk_has_allocations(const struct sock *sk)
1841 {
1842         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1843 }
1844 
1845 /**
1846  * skwq_has_sleeper - check if there are any waiting processes
1847  * @wq: struct socket_wq
1848  *
1849  * Returns true if socket_wq has waiting processes
1850  *
1851  * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1852  * barrier call. They were added due to the race found within the tcp code.
1853  *
1854  * Consider following tcp code paths:
1855  *
1856  * CPU1                  CPU2
1857  *
1858  * sys_select            receive packet
1859  *   ...                 ...
1860  *   __add_wait_queue    update tp->rcv_nxt
1861  *   ...                 ...
1862  *   tp->rcv_nxt check   sock_def_readable
1863  *   ...                 {
1864  *   schedule               rcu_read_lock();
1865  *                          wq = rcu_dereference(sk->sk_wq);
1866  *                          if (wq && waitqueue_active(&wq->wait))
1867  *                              wake_up_interruptible(&wq->wait)
1868  *                          ...
1869  *                       }
1870  *
1871  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1872  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1873  * could then endup calling schedule and sleep forever if there are no more
1874  * data on the socket.
1875  *
1876  */
1877 static inline bool skwq_has_sleeper(struct socket_wq *wq)
1878 {
1879         return wq && wq_has_sleeper(&wq->wait);
1880 }
1881 
1882 /**
1883  * sock_poll_wait - place memory barrier behind the poll_wait call.
1884  * @filp:           file
1885  * @wait_address:   socket wait queue
1886  * @p:              poll_table
1887  *
1888  * See the comments in the wq_has_sleeper function.
1889  */
1890 static inline void sock_poll_wait(struct file *filp,
1891                 wait_queue_head_t *wait_address, poll_table *p)
1892 {
1893         if (!poll_does_not_wait(p) && wait_address) {
1894                 poll_wait(filp, wait_address, p);
1895                 /* We need to be sure we are in sync with the
1896                  * socket flags modification.
1897                  *
1898                  * This memory barrier is paired in the wq_has_sleeper.
1899                  */
1900                 smp_mb();
1901         }
1902 }
1903 
1904 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1905 {
1906         if (sk->sk_txhash) {
1907                 skb->l4_hash = 1;
1908                 skb->hash = sk->sk_txhash;
1909         }
1910 }
1911 
1912 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
1913 
1914 /*
1915  *      Queue a received datagram if it will fit. Stream and sequenced
1916  *      protocols can't normally use this as they need to fit buffers in
1917  *      and play with them.
1918  *
1919  *      Inlined as it's very short and called for pretty much every
1920  *      packet ever received.
1921  */
1922 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1923 {
1924         skb_orphan(skb);
1925         skb->sk = sk;
1926         skb->destructor = sock_rfree;
1927         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1928         sk_mem_charge(sk, skb->truesize);
1929 }
1930 
1931 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1932                     unsigned long expires);
1933 
1934 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1935 
1936 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1937 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1938 
1939 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1940 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1941 
1942 /*
1943  *      Recover an error report and clear atomically
1944  */
1945 
1946 static inline int sock_error(struct sock *sk)
1947 {
1948         int err;
1949         if (likely(!sk->sk_err))
1950                 return 0;
1951         err = xchg(&sk->sk_err, 0);
1952         return -err;
1953 }
1954 
1955 static inline unsigned long sock_wspace(struct sock *sk)
1956 {
1957         int amt = 0;
1958 
1959         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1960                 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1961                 if (amt < 0)
1962                         amt = 0;
1963         }
1964         return amt;
1965 }
1966 
1967 /* Note:
1968  *  We use sk->sk_wq_raw, from contexts knowing this
1969  *  pointer is not NULL and cannot disappear/change.
1970  */
1971 static inline void sk_set_bit(int nr, struct sock *sk)
1972 {
1973         if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
1974             !sock_flag(sk, SOCK_FASYNC))
1975                 return;
1976 
1977         set_bit(nr, &sk->sk_wq_raw->flags);
1978 }
1979 
1980 static inline void sk_clear_bit(int nr, struct sock *sk)
1981 {
1982         if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
1983             !sock_flag(sk, SOCK_FASYNC))
1984                 return;
1985 
1986         clear_bit(nr, &sk->sk_wq_raw->flags);
1987 }
1988 
1989 static inline void sk_wake_async(const struct sock *sk, int how, int band)
1990 {
1991         if (sock_flag(sk, SOCK_FASYNC)) {
1992                 rcu_read_lock();
1993                 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
1994                 rcu_read_unlock();
1995         }
1996 }
1997 
1998 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
1999  * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2000  * Note: for send buffers, TCP works better if we can build two skbs at
2001  * minimum.
2002  */
2003 #define TCP_SKB_MIN_TRUESIZE    (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2004 
2005 #define SOCK_MIN_SNDBUF         (TCP_SKB_MIN_TRUESIZE * 2)
2006 #define SOCK_MIN_RCVBUF          TCP_SKB_MIN_TRUESIZE
2007 
2008 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2009 {
2010         if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2011                 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2012                 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2013         }
2014 }
2015 
2016 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2017                                     bool force_schedule);
2018 
2019 /**
2020  * sk_page_frag - return an appropriate page_frag
2021  * @sk: socket
2022  *
2023  * If socket allocation mode allows current thread to sleep, it means its
2024  * safe to use the per task page_frag instead of the per socket one.
2025  */
2026 static inline struct page_frag *sk_page_frag(struct sock *sk)
2027 {
2028         if (gfpflags_allow_blocking(sk->sk_allocation))
2029                 return &current->task_frag;
2030 
2031         return &sk->sk_frag;
2032 }
2033 
2034 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2035 
2036 /*
2037  *      Default write policy as shown to user space via poll/select/SIGIO
2038  */
2039 static inline bool sock_writeable(const struct sock *sk)
2040 {
2041         return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2042 }
2043 
2044 static inline gfp_t gfp_any(void)
2045 {
2046         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2047 }
2048 
2049 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2050 {
2051         return noblock ? 0 : sk->sk_rcvtimeo;
2052 }
2053 
2054 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2055 {
2056         return noblock ? 0 : sk->sk_sndtimeo;
2057 }
2058 
2059 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2060 {
2061         return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2062 }
2063 
2064 /* Alas, with timeout socket operations are not restartable.
2065  * Compare this to poll().
2066  */
2067 static inline int sock_intr_errno(long timeo)
2068 {
2069         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2070 }
2071 
2072 struct sock_skb_cb {
2073         u32 dropcount;
2074 };
2075 
2076 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2077  * using skb->cb[] would keep using it directly and utilize its
2078  * alignement guarantee.
2079  */
2080 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2081                             sizeof(struct sock_skb_cb)))
2082 
2083 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2084                             SOCK_SKB_CB_OFFSET))
2085 
2086 #define sock_skb_cb_check_size(size) \
2087         BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2088 
2089 static inline void
2090 sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2091 {
2092         SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
2093 }
2094 
2095 static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2096 {
2097         int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2098 
2099         atomic_add(segs, &sk->sk_drops);
2100 }
2101 
2102 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2103                            struct sk_buff *skb);
2104 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2105                              struct sk_buff *skb);
2106 
2107 static inline void
2108 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2109 {
2110         ktime_t kt = skb->tstamp;
2111         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2112 
2113         /*
2114          * generate control messages if
2115          * - receive time stamping in software requested
2116          * - software time stamp available and wanted
2117          * - hardware time stamps available and wanted
2118          */
2119         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2120             (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2121             (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2122             (hwtstamps->hwtstamp.tv64 &&
2123              (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2124                 __sock_recv_timestamp(msg, sk, skb);
2125         else
2126                 sk->sk_stamp = kt;
2127 
2128         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2129                 __sock_recv_wifi_status(msg, sk, skb);
2130 }
2131 
2132 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2133                               struct sk_buff *skb);
2134 
2135 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2136                                           struct sk_buff *skb)
2137 {
2138 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2139                            (1UL << SOCK_RCVTSTAMP))
2140 #define TSFLAGS_ANY       (SOF_TIMESTAMPING_SOFTWARE                    | \
2141                            SOF_TIMESTAMPING_RAW_HARDWARE)
2142 
2143         if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2144                 __sock_recv_ts_and_drops(msg, sk, skb);
2145         else
2146                 sk->sk_stamp = skb->tstamp;
2147 }
2148 
2149 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
2150 
2151 /**
2152  * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2153  * @sk:         socket sending this packet
2154  * @tsflags:    timestamping flags to use
2155  * @tx_flags:   completed with instructions for time stamping
2156  *
2157  * Note : callers should take care of initial *tx_flags value (usually 0)
2158  */
2159 static inline void sock_tx_timestamp(const struct sock *sk, __u16 tsflags,
2160                                      __u8 *tx_flags)
2161 {
2162         if (unlikely(tsflags))
2163                 __sock_tx_timestamp(tsflags, tx_flags);
2164         if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2165                 *tx_flags |= SKBTX_WIFI_STATUS;
2166 }
2167 
2168 /**
2169  * sk_eat_skb - Release a skb if it is no longer needed
2170  * @sk: socket to eat this skb from
2171  * @skb: socket buffer to eat
2172  *
2173  * This routine must be called with interrupts disabled or with the socket
2174  * locked so that the sk_buff queue operation is ok.
2175 */
2176 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2177 {
2178         __skb_unlink(skb, &sk->sk_receive_queue);
2179         __kfree_skb(skb);
2180 }
2181 
2182 static inline
2183 struct net *sock_net(const struct sock *sk)
2184 {
2185         return read_pnet(&sk->sk_net);
2186 }
2187 
2188 static inline
2189 void sock_net_set(struct sock *sk, struct net *net)
2190 {
2191         write_pnet(&sk->sk_net, net);
2192 }
2193 
2194 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2195 {
2196         if (skb->sk) {
2197                 struct sock *sk = skb->sk;
2198 
2199                 skb->destructor = NULL;
2200                 skb->sk = NULL;
2201                 return sk;
2202         }
2203         return NULL;
2204 }
2205 
2206 /* This helper checks if a socket is a full socket,
2207  * ie _not_ a timewait or request socket.
2208  */
2209 static inline bool sk_fullsock(const struct sock *sk)
2210 {
2211         return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2212 }
2213 
2214 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2215  * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2216  */
2217 static inline bool sk_listener(const struct sock *sk)
2218 {
2219         return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2220 }
2221 
2222 /**
2223  * sk_state_load - read sk->sk_state for lockless contexts
2224  * @sk: socket pointer
2225  *
2226  * Paired with sk_state_store(). Used in places we do not hold socket lock :
2227  * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2228  */
2229 static inline int sk_state_load(const struct sock *sk)
2230 {
2231         return smp_load_acquire(&sk->sk_state);
2232 }
2233 
2234 /**
2235  * sk_state_store - update sk->sk_state
2236  * @sk: socket pointer
2237  * @newstate: new state
2238  *
2239  * Paired with sk_state_load(). Should be used in contexts where
2240  * state change might impact lockless readers.
2241  */
2242 static inline void sk_state_store(struct sock *sk, int newstate)
2243 {
2244         smp_store_release(&sk->sk_state, newstate);
2245 }
2246 
2247 void sock_enable_timestamp(struct sock *sk, int flag);
2248 int sock_get_timestamp(struct sock *, struct timeval __user *);
2249 int sock_get_timestampns(struct sock *, struct timespec __user *);
2250 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2251                        int type);
2252 
2253 bool sk_ns_capable(const struct sock *sk,
2254                    struct user_namespace *user_ns, int cap);
2255 bool sk_capable(const struct sock *sk, int cap);
2256 bool sk_net_capable(const struct sock *sk, int cap);
2257 
2258 extern __u32 sysctl_wmem_max;
2259 extern __u32 sysctl_rmem_max;
2260 
2261 extern int sysctl_tstamp_allow_data;
2262 extern int sysctl_optmem_max;
2263 
2264 extern __u32 sysctl_wmem_default;
2265 extern __u32 sysctl_rmem_default;
2266 
2267 #endif  /* _SOCK_H */
2268 

~ [ 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