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

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

Version: ~ [ linux-5.5-rc7 ] ~ [ linux-5.4.13 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.97 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.166 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.210 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.210 ] ~ [ 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-3.16.81 ] ~ [ 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-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ 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 #include <linux/rbtree.h>
 64 #include <linux/filter.h>
 65 #include <linux/rculist_nulls.h>
 66 #include <linux/poll.h>
 67 
 68 #include <linux/atomic.h>
 69 #include <linux/refcount.h>
 70 #include <net/dst.h>
 71 #include <net/checksum.h>
 72 #include <net/tcp_states.h>
 73 #include <linux/net_tstamp.h>
 74 #include <net/smc.h>
 75 #include <net/l3mdev.h>
 76 
 77 /*
 78  * This structure really needs to be cleaned up.
 79  * Most of it is for TCP, and not used by any of
 80  * the other protocols.
 81  */
 82 
 83 /* Define this to get the SOCK_DBG debugging facility. */
 84 #define SOCK_DEBUGGING
 85 #ifdef SOCK_DEBUGGING
 86 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
 87                                         printk(KERN_DEBUG msg); } while (0)
 88 #else
 89 /* Validate arguments and do nothing */
 90 static inline __printf(2, 3)
 91 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
 92 {
 93 }
 94 #endif
 95 
 96 /* This is the per-socket lock.  The spinlock provides a synchronization
 97  * between user contexts and software interrupt processing, whereas the
 98  * mini-semaphore synchronizes multiple users amongst themselves.
 99  */
100 typedef struct {
101         spinlock_t              slock;
102         int                     owned;
103         wait_queue_head_t       wq;
104         /*
105          * We express the mutex-alike socket_lock semantics
106          * to the lock validator by explicitly managing
107          * the slock as a lock variant (in addition to
108          * the slock itself):
109          */
110 #ifdef CONFIG_DEBUG_LOCK_ALLOC
111         struct lockdep_map dep_map;
112 #endif
113 } socket_lock_t;
114 
115 struct sock;
116 struct proto;
117 struct net;
118 
119 typedef __u32 __bitwise __portpair;
120 typedef __u64 __bitwise __addrpair;
121 
122 /**
123  *      struct sock_common - minimal network layer representation of sockets
124  *      @skc_daddr: Foreign IPv4 addr
125  *      @skc_rcv_saddr: Bound local IPv4 addr
126  *      @skc_hash: hash value used with various protocol lookup tables
127  *      @skc_u16hashes: two u16 hash values used by UDP lookup tables
128  *      @skc_dport: placeholder for inet_dport/tw_dport
129  *      @skc_num: placeholder for inet_num/tw_num
130  *      @skc_family: network address family
131  *      @skc_state: Connection state
132  *      @skc_reuse: %SO_REUSEADDR setting
133  *      @skc_reuseport: %SO_REUSEPORT setting
134  *      @skc_bound_dev_if: bound device index if != 0
135  *      @skc_bind_node: bind hash linkage for various protocol lookup tables
136  *      @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
137  *      @skc_prot: protocol handlers inside a network family
138  *      @skc_net: reference to the network namespace of this socket
139  *      @skc_node: main hash linkage for various protocol lookup tables
140  *      @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
141  *      @skc_tx_queue_mapping: tx queue number for this connection
142  *      @skc_rx_queue_mapping: rx queue number for this connection
143  *      @skc_flags: place holder for sk_flags
144  *              %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
145  *              %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
146  *      @skc_incoming_cpu: record/match cpu processing incoming packets
147  *      @skc_refcnt: reference count
148  *
149  *      This is the minimal network layer representation of sockets, the header
150  *      for struct sock and struct inet_timewait_sock.
151  */
152 struct sock_common {
153         /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
154          * address on 64bit arches : cf INET_MATCH()
155          */
156         union {
157                 __addrpair      skc_addrpair;
158                 struct {
159                         __be32  skc_daddr;
160                         __be32  skc_rcv_saddr;
161                 };
162         };
163         union  {
164                 unsigned int    skc_hash;
165                 __u16           skc_u16hashes[2];
166         };
167         /* skc_dport && skc_num must be grouped as well */
168         union {
169                 __portpair      skc_portpair;
170                 struct {
171                         __be16  skc_dport;
172                         __u16   skc_num;
173                 };
174         };
175 
176         unsigned short          skc_family;
177         volatile unsigned char  skc_state;
178         unsigned char           skc_reuse:4;
179         unsigned char           skc_reuseport:1;
180         unsigned char           skc_ipv6only:1;
181         unsigned char           skc_net_refcnt:1;
182         int                     skc_bound_dev_if;
183         union {
184                 struct hlist_node       skc_bind_node;
185                 struct hlist_node       skc_portaddr_node;
186         };
187         struct proto            *skc_prot;
188         possible_net_t          skc_net;
189 
190 #if IS_ENABLED(CONFIG_IPV6)
191         struct in6_addr         skc_v6_daddr;
192         struct in6_addr         skc_v6_rcv_saddr;
193 #endif
194 
195         atomic64_t              skc_cookie;
196 
197         /* following fields are padding to force
198          * offset(struct sock, sk_refcnt) == 128 on 64bit arches
199          * assuming IPV6 is enabled. We use this padding differently
200          * for different kind of 'sockets'
201          */
202         union {
203                 unsigned long   skc_flags;
204                 struct sock     *skc_listener; /* request_sock */
205                 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
206         };
207         /*
208          * fields between dontcopy_begin/dontcopy_end
209          * are not copied in sock_copy()
210          */
211         /* private: */
212         int                     skc_dontcopy_begin[0];
213         /* public: */
214         union {
215                 struct hlist_node       skc_node;
216                 struct hlist_nulls_node skc_nulls_node;
217         };
218         unsigned short          skc_tx_queue_mapping;
219 #ifdef CONFIG_XPS
220         unsigned short          skc_rx_queue_mapping;
221 #endif
222         union {
223                 int             skc_incoming_cpu;
224                 u32             skc_rcv_wnd;
225                 u32             skc_tw_rcv_nxt; /* struct tcp_timewait_sock  */
226         };
227 
228         refcount_t              skc_refcnt;
229         /* private: */
230         int                     skc_dontcopy_end[0];
231         union {
232                 u32             skc_rxhash;
233                 u32             skc_window_clamp;
234                 u32             skc_tw_snd_nxt; /* struct tcp_timewait_sock */
235         };
236         /* public: */
237 };
238 
239 /**
240   *     struct sock - network layer representation of sockets
241   *     @__sk_common: shared layout with inet_timewait_sock
242   *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
243   *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
244   *     @sk_lock:       synchronizer
245   *     @sk_kern_sock: True if sock is using kernel lock classes
246   *     @sk_rcvbuf: size of receive buffer in bytes
247   *     @sk_wq: sock wait queue and async head
248   *     @sk_rx_dst: receive input route used by early demux
249   *     @sk_dst_cache: destination cache
250   *     @sk_dst_pending_confirm: need to confirm neighbour
251   *     @sk_policy: flow policy
252   *     @sk_receive_queue: incoming packets
253   *     @sk_wmem_alloc: transmit queue bytes committed
254   *     @sk_tsq_flags: TCP Small Queues flags
255   *     @sk_write_queue: Packet sending queue
256   *     @sk_omem_alloc: "o" is "option" or "other"
257   *     @sk_wmem_queued: persistent queue size
258   *     @sk_forward_alloc: space allocated forward
259   *     @sk_napi_id: id of the last napi context to receive data for sk
260   *     @sk_ll_usec: usecs to busypoll when there is no data
261   *     @sk_allocation: allocation mode
262   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
263   *     @sk_pacing_status: Pacing status (requested, handled by sch_fq)
264   *     @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
265   *     @sk_sndbuf: size of send buffer in bytes
266   *     @__sk_flags_offset: empty field used to determine location of bitfield
267   *     @sk_padding: unused element for alignment
268   *     @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
269   *     @sk_no_check_rx: allow zero checksum in RX packets
270   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
271   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
272   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
273   *     @sk_gso_max_size: Maximum GSO segment size to build
274   *     @sk_gso_max_segs: Maximum number of GSO segments
275   *     @sk_pacing_shift: scaling factor for TCP Small Queues
276   *     @sk_lingertime: %SO_LINGER l_linger setting
277   *     @sk_backlog: always used with the per-socket spinlock held
278   *     @sk_callback_lock: used with the callbacks in the end of this struct
279   *     @sk_error_queue: rarely used
280   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
281   *                       IPV6_ADDRFORM for instance)
282   *     @sk_err: last error
283   *     @sk_err_soft: errors that don't cause failure but are the cause of a
284   *                   persistent failure not just 'timed out'
285   *     @sk_drops: raw/udp drops counter
286   *     @sk_ack_backlog: current listen backlog
287   *     @sk_max_ack_backlog: listen backlog set in listen()
288   *     @sk_uid: user id of owner
289   *     @sk_priority: %SO_PRIORITY setting
290   *     @sk_type: socket type (%SOCK_STREAM, etc)
291   *     @sk_protocol: which protocol this socket belongs in this network family
292   *     @sk_peer_pid: &struct pid for this socket's peer
293   *     @sk_peer_cred: %SO_PEERCRED setting
294   *     @sk_rcvlowat: %SO_RCVLOWAT setting
295   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
296   *     @sk_sndtimeo: %SO_SNDTIMEO setting
297   *     @sk_txhash: computed flow hash for use on transmit
298   *     @sk_filter: socket filtering instructions
299   *     @sk_timer: sock cleanup timer
300   *     @sk_stamp: time stamp of last packet received
301   *     @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
302   *     @sk_tsflags: SO_TIMESTAMPING socket options
303   *     @sk_tskey: counter to disambiguate concurrent tstamp requests
304   *     @sk_zckey: counter to order MSG_ZEROCOPY notifications
305   *     @sk_socket: Identd and reporting IO signals
306   *     @sk_user_data: RPC layer private data
307   *     @sk_frag: cached page frag
308   *     @sk_peek_off: current peek_offset value
309   *     @sk_send_head: front of stuff to transmit
310   *     @sk_security: used by security modules
311   *     @sk_mark: generic packet mark
312   *     @sk_cgrp_data: cgroup data for this cgroup
313   *     @sk_memcg: this socket's memory cgroup association
314   *     @sk_write_pending: a write to stream socket waits to start
315   *     @sk_state_change: callback to indicate change in the state of the sock
316   *     @sk_data_ready: callback to indicate there is data to be processed
317   *     @sk_write_space: callback to indicate there is bf sending space available
318   *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
319   *     @sk_backlog_rcv: callback to process the backlog
320   *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
321   *     @sk_reuseport_cb: reuseport group container
322   *     @sk_rcu: used during RCU grace period
323   *     @sk_clockid: clockid used by time-based scheduling (SO_TXTIME)
324   *     @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME
325   *     @sk_txtime_unused: unused txtime flags
326   */
327 struct sock {
328         /*
329          * Now struct inet_timewait_sock also uses sock_common, so please just
330          * don't add nothing before this first member (__sk_common) --acme
331          */
332         struct sock_common      __sk_common;
333 #define sk_node                 __sk_common.skc_node
334 #define sk_nulls_node           __sk_common.skc_nulls_node
335 #define sk_refcnt               __sk_common.skc_refcnt
336 #define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
337 #ifdef CONFIG_XPS
338 #define sk_rx_queue_mapping     __sk_common.skc_rx_queue_mapping
339 #endif
340 
341 #define sk_dontcopy_begin       __sk_common.skc_dontcopy_begin
342 #define sk_dontcopy_end         __sk_common.skc_dontcopy_end
343 #define sk_hash                 __sk_common.skc_hash
344 #define sk_portpair             __sk_common.skc_portpair
345 #define sk_num                  __sk_common.skc_num
346 #define sk_dport                __sk_common.skc_dport
347 #define sk_addrpair             __sk_common.skc_addrpair
348 #define sk_daddr                __sk_common.skc_daddr
349 #define sk_rcv_saddr            __sk_common.skc_rcv_saddr
350 #define sk_family               __sk_common.skc_family
351 #define sk_state                __sk_common.skc_state
352 #define sk_reuse                __sk_common.skc_reuse
353 #define sk_reuseport            __sk_common.skc_reuseport
354 #define sk_ipv6only             __sk_common.skc_ipv6only
355 #define sk_net_refcnt           __sk_common.skc_net_refcnt
356 #define sk_bound_dev_if         __sk_common.skc_bound_dev_if
357 #define sk_bind_node            __sk_common.skc_bind_node
358 #define sk_prot                 __sk_common.skc_prot
359 #define sk_net                  __sk_common.skc_net
360 #define sk_v6_daddr             __sk_common.skc_v6_daddr
361 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
362 #define sk_cookie               __sk_common.skc_cookie
363 #define sk_incoming_cpu         __sk_common.skc_incoming_cpu
364 #define sk_flags                __sk_common.skc_flags
365 #define sk_rxhash               __sk_common.skc_rxhash
366 
367         socket_lock_t           sk_lock;
368         atomic_t                sk_drops;
369         int                     sk_rcvlowat;
370         struct sk_buff_head     sk_error_queue;
371         struct sk_buff_head     sk_receive_queue;
372         /*
373          * The backlog queue is special, it is always used with
374          * the per-socket spinlock held and requires low latency
375          * access. Therefore we special case it's implementation.
376          * Note : rmem_alloc is in this structure to fill a hole
377          * on 64bit arches, not because its logically part of
378          * backlog.
379          */
380         struct {
381                 atomic_t        rmem_alloc;
382                 int             len;
383                 struct sk_buff  *head;
384                 struct sk_buff  *tail;
385         } sk_backlog;
386 #define sk_rmem_alloc sk_backlog.rmem_alloc
387 
388         int                     sk_forward_alloc;
389 #ifdef CONFIG_NET_RX_BUSY_POLL
390         unsigned int            sk_ll_usec;
391         /* ===== mostly read cache line ===== */
392         unsigned int            sk_napi_id;
393 #endif
394         int                     sk_rcvbuf;
395 
396         struct sk_filter __rcu  *sk_filter;
397         union {
398                 struct socket_wq __rcu  *sk_wq;
399                 struct socket_wq        *sk_wq_raw;
400         };
401 #ifdef CONFIG_XFRM
402         struct xfrm_policy __rcu *sk_policy[2];
403 #endif
404         struct dst_entry        *sk_rx_dst;
405         struct dst_entry __rcu  *sk_dst_cache;
406         atomic_t                sk_omem_alloc;
407         int                     sk_sndbuf;
408 
409         /* ===== cache line for TX ===== */
410         int                     sk_wmem_queued;
411         refcount_t              sk_wmem_alloc;
412         unsigned long           sk_tsq_flags;
413         union {
414                 struct sk_buff  *sk_send_head;
415                 struct rb_root  tcp_rtx_queue;
416         };
417         struct sk_buff_head     sk_write_queue;
418         __s32                   sk_peek_off;
419         int                     sk_write_pending;
420         __u32                   sk_dst_pending_confirm;
421         u32                     sk_pacing_status; /* see enum sk_pacing */
422         long                    sk_sndtimeo;
423         struct timer_list       sk_timer;
424         __u32                   sk_priority;
425         __u32                   sk_mark;
426         unsigned long           sk_pacing_rate; /* bytes per second */
427         unsigned long           sk_max_pacing_rate;
428         struct page_frag        sk_frag;
429         netdev_features_t       sk_route_caps;
430         netdev_features_t       sk_route_nocaps;
431         netdev_features_t       sk_route_forced_caps;
432         int                     sk_gso_type;
433         unsigned int            sk_gso_max_size;
434         gfp_t                   sk_allocation;
435         __u32                   sk_txhash;
436 
437         /*
438          * Because of non atomicity rules, all
439          * changes are protected by socket lock.
440          */
441         unsigned int            __sk_flags_offset[0];
442 #ifdef __BIG_ENDIAN_BITFIELD
443 #define SK_FL_PROTO_SHIFT  16
444 #define SK_FL_PROTO_MASK   0x00ff0000
445 
446 #define SK_FL_TYPE_SHIFT   0
447 #define SK_FL_TYPE_MASK    0x0000ffff
448 #else
449 #define SK_FL_PROTO_SHIFT  8
450 #define SK_FL_PROTO_MASK   0x0000ff00
451 
452 #define SK_FL_TYPE_SHIFT   16
453 #define SK_FL_TYPE_MASK    0xffff0000
454 #endif
455 
456         unsigned int            sk_padding : 1,
457                                 sk_kern_sock : 1,
458                                 sk_no_check_tx : 1,
459                                 sk_no_check_rx : 1,
460                                 sk_userlocks : 4,
461                                 sk_protocol  : 8,
462                                 sk_type      : 16;
463 #define SK_PROTOCOL_MAX U8_MAX
464         u16                     sk_gso_max_segs;
465         u8                      sk_pacing_shift;
466         unsigned long           sk_lingertime;
467         struct proto            *sk_prot_creator;
468         rwlock_t                sk_callback_lock;
469         int                     sk_err,
470                                 sk_err_soft;
471         u32                     sk_ack_backlog;
472         u32                     sk_max_ack_backlog;
473         kuid_t                  sk_uid;
474         struct pid              *sk_peer_pid;
475         const struct cred       *sk_peer_cred;
476         long                    sk_rcvtimeo;
477         ktime_t                 sk_stamp;
478 #if BITS_PER_LONG==32
479         seqlock_t               sk_stamp_seq;
480 #endif
481         u16                     sk_tsflags;
482         u8                      sk_shutdown;
483         u32                     sk_tskey;
484         atomic_t                sk_zckey;
485 
486         u8                      sk_clockid;
487         u8                      sk_txtime_deadline_mode : 1,
488                                 sk_txtime_report_errors : 1,
489                                 sk_txtime_unused : 6;
490 
491         struct socket           *sk_socket;
492         void                    *sk_user_data;
493 #ifdef CONFIG_SECURITY
494         void                    *sk_security;
495 #endif
496         struct sock_cgroup_data sk_cgrp_data;
497         struct mem_cgroup       *sk_memcg;
498         void                    (*sk_state_change)(struct sock *sk);
499         void                    (*sk_data_ready)(struct sock *sk);
500         void                    (*sk_write_space)(struct sock *sk);
501         void                    (*sk_error_report)(struct sock *sk);
502         int                     (*sk_backlog_rcv)(struct sock *sk,
503                                                   struct sk_buff *skb);
504 #ifdef CONFIG_SOCK_VALIDATE_XMIT
505         struct sk_buff*         (*sk_validate_xmit_skb)(struct sock *sk,
506                                                         struct net_device *dev,
507                                                         struct sk_buff *skb);
508 #endif
509         void                    (*sk_destruct)(struct sock *sk);
510         struct sock_reuseport __rcu     *sk_reuseport_cb;
511         struct rcu_head         sk_rcu;
512 };
513 
514 enum sk_pacing {
515         SK_PACING_NONE          = 0,
516         SK_PACING_NEEDED        = 1,
517         SK_PACING_FQ            = 2,
518 };
519 
520 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
521 
522 #define rcu_dereference_sk_user_data(sk)        rcu_dereference(__sk_user_data((sk)))
523 #define rcu_assign_sk_user_data(sk, ptr)        rcu_assign_pointer(__sk_user_data((sk)), ptr)
524 
525 /*
526  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
527  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
528  * on a socket means that the socket will reuse everybody else's port
529  * without looking at the other's sk_reuse value.
530  */
531 
532 #define SK_NO_REUSE     0
533 #define SK_CAN_REUSE    1
534 #define SK_FORCE_REUSE  2
535 
536 int sk_set_peek_off(struct sock *sk, int val);
537 
538 static inline int sk_peek_offset(struct sock *sk, int flags)
539 {
540         if (unlikely(flags & MSG_PEEK)) {
541                 return READ_ONCE(sk->sk_peek_off);
542         }
543 
544         return 0;
545 }
546 
547 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
548 {
549         s32 off = READ_ONCE(sk->sk_peek_off);
550 
551         if (unlikely(off >= 0)) {
552                 off = max_t(s32, off - val, 0);
553                 WRITE_ONCE(sk->sk_peek_off, off);
554         }
555 }
556 
557 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
558 {
559         sk_peek_offset_bwd(sk, -val);
560 }
561 
562 /*
563  * Hashed lists helper routines
564  */
565 static inline struct sock *sk_entry(const struct hlist_node *node)
566 {
567         return hlist_entry(node, struct sock, sk_node);
568 }
569 
570 static inline struct sock *__sk_head(const struct hlist_head *head)
571 {
572         return hlist_entry(head->first, struct sock, sk_node);
573 }
574 
575 static inline struct sock *sk_head(const struct hlist_head *head)
576 {
577         return hlist_empty(head) ? NULL : __sk_head(head);
578 }
579 
580 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
581 {
582         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
583 }
584 
585 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
586 {
587         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
588 }
589 
590 static inline struct sock *sk_next(const struct sock *sk)
591 {
592         return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
593 }
594 
595 static inline struct sock *sk_nulls_next(const struct sock *sk)
596 {
597         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
598                 hlist_nulls_entry(sk->sk_nulls_node.next,
599                                   struct sock, sk_nulls_node) :
600                 NULL;
601 }
602 
603 static inline bool sk_unhashed(const struct sock *sk)
604 {
605         return hlist_unhashed(&sk->sk_node);
606 }
607 
608 static inline bool sk_hashed(const struct sock *sk)
609 {
610         return !sk_unhashed(sk);
611 }
612 
613 static inline void sk_node_init(struct hlist_node *node)
614 {
615         node->pprev = NULL;
616 }
617 
618 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
619 {
620         node->pprev = NULL;
621 }
622 
623 static inline void __sk_del_node(struct sock *sk)
624 {
625         __hlist_del(&sk->sk_node);
626 }
627 
628 /* NB: equivalent to hlist_del_init_rcu */
629 static inline bool __sk_del_node_init(struct sock *sk)
630 {
631         if (sk_hashed(sk)) {
632                 __sk_del_node(sk);
633                 sk_node_init(&sk->sk_node);
634                 return true;
635         }
636         return false;
637 }
638 
639 /* Grab socket reference count. This operation is valid only
640    when sk is ALREADY grabbed f.e. it is found in hash table
641    or a list and the lookup is made under lock preventing hash table
642    modifications.
643  */
644 
645 static __always_inline void sock_hold(struct sock *sk)
646 {
647         refcount_inc(&sk->sk_refcnt);
648 }
649 
650 /* Ungrab socket in the context, which assumes that socket refcnt
651    cannot hit zero, f.e. it is true in context of any socketcall.
652  */
653 static __always_inline void __sock_put(struct sock *sk)
654 {
655         refcount_dec(&sk->sk_refcnt);
656 }
657 
658 static inline bool sk_del_node_init(struct sock *sk)
659 {
660         bool rc = __sk_del_node_init(sk);
661 
662         if (rc) {
663                 /* paranoid for a while -acme */
664                 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
665                 __sock_put(sk);
666         }
667         return rc;
668 }
669 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
670 
671 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
672 {
673         if (sk_hashed(sk)) {
674                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
675                 return true;
676         }
677         return false;
678 }
679 
680 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
681 {
682         bool rc = __sk_nulls_del_node_init_rcu(sk);
683 
684         if (rc) {
685                 /* paranoid for a while -acme */
686                 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
687                 __sock_put(sk);
688         }
689         return rc;
690 }
691 
692 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
693 {
694         hlist_add_head(&sk->sk_node, list);
695 }
696 
697 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
698 {
699         sock_hold(sk);
700         __sk_add_node(sk, list);
701 }
702 
703 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
704 {
705         sock_hold(sk);
706         if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
707             sk->sk_family == AF_INET6)
708                 hlist_add_tail_rcu(&sk->sk_node, list);
709         else
710                 hlist_add_head_rcu(&sk->sk_node, list);
711 }
712 
713 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
714 {
715         hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
716 }
717 
718 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
719 {
720         sock_hold(sk);
721         __sk_nulls_add_node_rcu(sk, list);
722 }
723 
724 static inline void __sk_del_bind_node(struct sock *sk)
725 {
726         __hlist_del(&sk->sk_bind_node);
727 }
728 
729 static inline void sk_add_bind_node(struct sock *sk,
730                                         struct hlist_head *list)
731 {
732         hlist_add_head(&sk->sk_bind_node, list);
733 }
734 
735 #define sk_for_each(__sk, list) \
736         hlist_for_each_entry(__sk, list, sk_node)
737 #define sk_for_each_rcu(__sk, list) \
738         hlist_for_each_entry_rcu(__sk, list, sk_node)
739 #define sk_nulls_for_each(__sk, node, list) \
740         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
741 #define sk_nulls_for_each_rcu(__sk, node, list) \
742         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
743 #define sk_for_each_from(__sk) \
744         hlist_for_each_entry_from(__sk, sk_node)
745 #define sk_nulls_for_each_from(__sk, node) \
746         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
747                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
748 #define sk_for_each_safe(__sk, tmp, list) \
749         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
750 #define sk_for_each_bound(__sk, list) \
751         hlist_for_each_entry(__sk, list, sk_bind_node)
752 
753 /**
754  * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
755  * @tpos:       the type * to use as a loop cursor.
756  * @pos:        the &struct hlist_node to use as a loop cursor.
757  * @head:       the head for your list.
758  * @offset:     offset of hlist_node within the struct.
759  *
760  */
761 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset)                  \
762         for (pos = rcu_dereference(hlist_first_rcu(head));                     \
763              pos != NULL &&                                                    \
764                 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;});       \
765              pos = rcu_dereference(hlist_next_rcu(pos)))
766 
767 static inline struct user_namespace *sk_user_ns(struct sock *sk)
768 {
769         /* Careful only use this in a context where these parameters
770          * can not change and must all be valid, such as recvmsg from
771          * userspace.
772          */
773         return sk->sk_socket->file->f_cred->user_ns;
774 }
775 
776 /* Sock flags */
777 enum sock_flags {
778         SOCK_DEAD,
779         SOCK_DONE,
780         SOCK_URGINLINE,
781         SOCK_KEEPOPEN,
782         SOCK_LINGER,
783         SOCK_DESTROY,
784         SOCK_BROADCAST,
785         SOCK_TIMESTAMP,
786         SOCK_ZAPPED,
787         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
788         SOCK_DBG, /* %SO_DEBUG setting */
789         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
790         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
791         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
792         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
793         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
794         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
795         SOCK_FASYNC, /* fasync() active */
796         SOCK_RXQ_OVFL,
797         SOCK_ZEROCOPY, /* buffers from userspace */
798         SOCK_WIFI_STATUS, /* push wifi status to userspace */
799         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
800                      * Will use last 4 bytes of packet sent from
801                      * user-space instead.
802                      */
803         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
804         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
805         SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
806         SOCK_TXTIME,
807         SOCK_XDP, /* XDP is attached */
808 };
809 
810 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
811 
812 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
813 {
814         nsk->sk_flags = osk->sk_flags;
815 }
816 
817 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
818 {
819         __set_bit(flag, &sk->sk_flags);
820 }
821 
822 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
823 {
824         __clear_bit(flag, &sk->sk_flags);
825 }
826 
827 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
828 {
829         return test_bit(flag, &sk->sk_flags);
830 }
831 
832 #ifdef CONFIG_NET
833 DECLARE_STATIC_KEY_FALSE(memalloc_socks_key);
834 static inline int sk_memalloc_socks(void)
835 {
836         return static_branch_unlikely(&memalloc_socks_key);
837 }
838 #else
839 
840 static inline int sk_memalloc_socks(void)
841 {
842         return 0;
843 }
844 
845 #endif
846 
847 static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
848 {
849         return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
850 }
851 
852 static inline void sk_acceptq_removed(struct sock *sk)
853 {
854         sk->sk_ack_backlog--;
855 }
856 
857 static inline void sk_acceptq_added(struct sock *sk)
858 {
859         sk->sk_ack_backlog++;
860 }
861 
862 static inline bool sk_acceptq_is_full(const struct sock *sk)
863 {
864         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
865 }
866 
867 /*
868  * Compute minimal free write space needed to queue new packets.
869  */
870 static inline int sk_stream_min_wspace(const struct sock *sk)
871 {
872         return sk->sk_wmem_queued >> 1;
873 }
874 
875 static inline int sk_stream_wspace(const struct sock *sk)
876 {
877         return sk->sk_sndbuf - sk->sk_wmem_queued;
878 }
879 
880 void sk_stream_write_space(struct sock *sk);
881 
882 /* OOB backlog add */
883 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
884 {
885         /* dont let skb dst not refcounted, we are going to leave rcu lock */
886         skb_dst_force(skb);
887 
888         if (!sk->sk_backlog.tail)
889                 sk->sk_backlog.head = skb;
890         else
891                 sk->sk_backlog.tail->next = skb;
892 
893         sk->sk_backlog.tail = skb;
894         skb->next = NULL;
895 }
896 
897 /*
898  * Take into account size of receive queue and backlog queue
899  * Do not take into account this skb truesize,
900  * to allow even a single big packet to come.
901  */
902 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
903 {
904         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
905 
906         return qsize > limit;
907 }
908 
909 /* The per-socket spinlock must be held here. */
910 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
911                                               unsigned int limit)
912 {
913         if (sk_rcvqueues_full(sk, limit))
914                 return -ENOBUFS;
915 
916         /*
917          * If the skb was allocated from pfmemalloc reserves, only
918          * allow SOCK_MEMALLOC sockets to use it as this socket is
919          * helping free memory
920          */
921         if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
922                 return -ENOMEM;
923 
924         __sk_add_backlog(sk, skb);
925         sk->sk_backlog.len += skb->truesize;
926         return 0;
927 }
928 
929 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
930 
931 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
932 {
933         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
934                 return __sk_backlog_rcv(sk, skb);
935 
936         return sk->sk_backlog_rcv(sk, skb);
937 }
938 
939 static inline void sk_incoming_cpu_update(struct sock *sk)
940 {
941         int cpu = raw_smp_processor_id();
942 
943         if (unlikely(sk->sk_incoming_cpu != cpu))
944                 sk->sk_incoming_cpu = cpu;
945 }
946 
947 static inline void sock_rps_record_flow_hash(__u32 hash)
948 {
949 #ifdef CONFIG_RPS
950         struct rps_sock_flow_table *sock_flow_table;
951 
952         rcu_read_lock();
953         sock_flow_table = rcu_dereference(rps_sock_flow_table);
954         rps_record_sock_flow(sock_flow_table, hash);
955         rcu_read_unlock();
956 #endif
957 }
958 
959 static inline void sock_rps_record_flow(const struct sock *sk)
960 {
961 #ifdef CONFIG_RPS
962         if (static_key_false(&rfs_needed)) {
963                 /* Reading sk->sk_rxhash might incur an expensive cache line
964                  * miss.
965                  *
966                  * TCP_ESTABLISHED does cover almost all states where RFS
967                  * might be useful, and is cheaper [1] than testing :
968                  *      IPv4: inet_sk(sk)->inet_daddr
969                  *      IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
970                  * OR   an additional socket flag
971                  * [1] : sk_state and sk_prot are in the same cache line.
972                  */
973                 if (sk->sk_state == TCP_ESTABLISHED)
974                         sock_rps_record_flow_hash(sk->sk_rxhash);
975         }
976 #endif
977 }
978 
979 static inline void sock_rps_save_rxhash(struct sock *sk,
980                                         const struct sk_buff *skb)
981 {
982 #ifdef CONFIG_RPS
983         if (unlikely(sk->sk_rxhash != skb->hash))
984                 sk->sk_rxhash = skb->hash;
985 #endif
986 }
987 
988 static inline void sock_rps_reset_rxhash(struct sock *sk)
989 {
990 #ifdef CONFIG_RPS
991         sk->sk_rxhash = 0;
992 #endif
993 }
994 
995 #define sk_wait_event(__sk, __timeo, __condition, __wait)               \
996         ({      int __rc;                                               \
997                 release_sock(__sk);                                     \
998                 __rc = __condition;                                     \
999                 if (!__rc) {                                            \
1000                         *(__timeo) = wait_woken(__wait,                 \
1001                                                 TASK_INTERRUPTIBLE,     \
1002                                                 *(__timeo));            \
1003                 }                                                       \
1004                 sched_annotate_sleep();                                 \
1005                 lock_sock(__sk);                                        \
1006                 __rc = __condition;                                     \
1007                 __rc;                                                   \
1008         })
1009 
1010 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
1011 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
1012 void sk_stream_wait_close(struct sock *sk, long timeo_p);
1013 int sk_stream_error(struct sock *sk, int flags, int err);
1014 void sk_stream_kill_queues(struct sock *sk);
1015 void sk_set_memalloc(struct sock *sk);
1016 void sk_clear_memalloc(struct sock *sk);
1017 
1018 void __sk_flush_backlog(struct sock *sk);
1019 
1020 static inline bool sk_flush_backlog(struct sock *sk)
1021 {
1022         if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
1023                 __sk_flush_backlog(sk);
1024                 return true;
1025         }
1026         return false;
1027 }
1028 
1029 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1030 
1031 struct request_sock_ops;
1032 struct timewait_sock_ops;
1033 struct inet_hashinfo;
1034 struct raw_hashinfo;
1035 struct smc_hashinfo;
1036 struct module;
1037 
1038 /*
1039  * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1040  * un-modified. Special care is taken when initializing object to zero.
1041  */
1042 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1043 {
1044         if (offsetof(struct sock, sk_node.next) != 0)
1045                 memset(sk, 0, offsetof(struct sock, sk_node.next));
1046         memset(&sk->sk_node.pprev, 0,
1047                size - offsetof(struct sock, sk_node.pprev));
1048 }
1049 
1050 /* Networking protocol blocks we attach to sockets.
1051  * socket layer -> transport layer interface
1052  */
1053 struct proto {
1054         void                    (*close)(struct sock *sk,
1055                                         long timeout);
1056         int                     (*pre_connect)(struct sock *sk,
1057                                         struct sockaddr *uaddr,
1058                                         int addr_len);
1059         int                     (*connect)(struct sock *sk,
1060                                         struct sockaddr *uaddr,
1061                                         int addr_len);
1062         int                     (*disconnect)(struct sock *sk, int flags);
1063 
1064         struct sock *           (*accept)(struct sock *sk, int flags, int *err,
1065                                           bool kern);
1066 
1067         int                     (*ioctl)(struct sock *sk, int cmd,
1068                                          unsigned long arg);
1069         int                     (*init)(struct sock *sk);
1070         void                    (*destroy)(struct sock *sk);
1071         void                    (*shutdown)(struct sock *sk, int how);
1072         int                     (*setsockopt)(struct sock *sk, int level,
1073                                         int optname, char __user *optval,
1074                                         unsigned int optlen);
1075         int                     (*getsockopt)(struct sock *sk, int level,
1076                                         int optname, char __user *optval,
1077                                         int __user *option);
1078         void                    (*keepalive)(struct sock *sk, int valbool);
1079 #ifdef CONFIG_COMPAT
1080         int                     (*compat_setsockopt)(struct sock *sk,
1081                                         int level,
1082                                         int optname, char __user *optval,
1083                                         unsigned int optlen);
1084         int                     (*compat_getsockopt)(struct sock *sk,
1085                                         int level,
1086                                         int optname, char __user *optval,
1087                                         int __user *option);
1088         int                     (*compat_ioctl)(struct sock *sk,
1089                                         unsigned int cmd, unsigned long arg);
1090 #endif
1091         int                     (*sendmsg)(struct sock *sk, struct msghdr *msg,
1092                                            size_t len);
1093         int                     (*recvmsg)(struct sock *sk, struct msghdr *msg,
1094                                            size_t len, int noblock, int flags,
1095                                            int *addr_len);
1096         int                     (*sendpage)(struct sock *sk, struct page *page,
1097                                         int offset, size_t size, int flags);
1098         int                     (*bind)(struct sock *sk,
1099                                         struct sockaddr *uaddr, int addr_len);
1100 
1101         int                     (*backlog_rcv) (struct sock *sk,
1102                                                 struct sk_buff *skb);
1103 
1104         void            (*release_cb)(struct sock *sk);
1105 
1106         /* Keeping track of sk's, looking them up, and port selection methods. */
1107         int                     (*hash)(struct sock *sk);
1108         void                    (*unhash)(struct sock *sk);
1109         void                    (*rehash)(struct sock *sk);
1110         int                     (*get_port)(struct sock *sk, unsigned short snum);
1111 
1112         /* Keeping track of sockets in use */
1113 #ifdef CONFIG_PROC_FS
1114         unsigned int            inuse_idx;
1115 #endif
1116 
1117         bool                    (*stream_memory_free)(const struct sock *sk);
1118         bool                    (*stream_memory_read)(const struct sock *sk);
1119         /* Memory pressure */
1120         void                    (*enter_memory_pressure)(struct sock *sk);
1121         void                    (*leave_memory_pressure)(struct sock *sk);
1122         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
1123         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
1124         /*
1125          * Pressure flag: try to collapse.
1126          * Technical note: it is used by multiple contexts non atomically.
1127          * All the __sk_mem_schedule() is of this nature: accounting
1128          * is strict, actions are advisory and have some latency.
1129          */
1130         unsigned long           *memory_pressure;
1131         long                    *sysctl_mem;
1132 
1133         int                     *sysctl_wmem;
1134         int                     *sysctl_rmem;
1135         u32                     sysctl_wmem_offset;
1136         u32                     sysctl_rmem_offset;
1137 
1138         int                     max_header;
1139         bool                    no_autobind;
1140 
1141         struct kmem_cache       *slab;
1142         unsigned int            obj_size;
1143         slab_flags_t            slab_flags;
1144         unsigned int            useroffset;     /* Usercopy region offset */
1145         unsigned int            usersize;       /* Usercopy region size */
1146 
1147         struct percpu_counter   *orphan_count;
1148 
1149         struct request_sock_ops *rsk_prot;
1150         struct timewait_sock_ops *twsk_prot;
1151 
1152         union {
1153                 struct inet_hashinfo    *hashinfo;
1154                 struct udp_table        *udp_table;
1155                 struct raw_hashinfo     *raw_hash;
1156                 struct smc_hashinfo     *smc_hash;
1157         } h;
1158 
1159         struct module           *owner;
1160 
1161         char                    name[32];
1162 
1163         struct list_head        node;
1164 #ifdef SOCK_REFCNT_DEBUG
1165         atomic_t                socks;
1166 #endif
1167         int                     (*diag_destroy)(struct sock *sk, int err);
1168 } __randomize_layout;
1169 
1170 int proto_register(struct proto *prot, int alloc_slab);
1171 void proto_unregister(struct proto *prot);
1172 int sock_load_diag_module(int family, int protocol);
1173 
1174 #ifdef SOCK_REFCNT_DEBUG
1175 static inline void sk_refcnt_debug_inc(struct sock *sk)
1176 {
1177         atomic_inc(&sk->sk_prot->socks);
1178 }
1179 
1180 static inline void sk_refcnt_debug_dec(struct sock *sk)
1181 {
1182         atomic_dec(&sk->sk_prot->socks);
1183         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1184                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1185 }
1186 
1187 static inline void sk_refcnt_debug_release(const struct sock *sk)
1188 {
1189         if (refcount_read(&sk->sk_refcnt) != 1)
1190                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1191                        sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt));
1192 }
1193 #else /* SOCK_REFCNT_DEBUG */
1194 #define sk_refcnt_debug_inc(sk) do { } while (0)
1195 #define sk_refcnt_debug_dec(sk) do { } while (0)
1196 #define sk_refcnt_debug_release(sk) do { } while (0)
1197 #endif /* SOCK_REFCNT_DEBUG */
1198 
1199 static inline bool sk_stream_memory_free(const struct sock *sk)
1200 {
1201         if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1202                 return false;
1203 
1204         return sk->sk_prot->stream_memory_free ?
1205                 sk->sk_prot->stream_memory_free(sk) : true;
1206 }
1207 
1208 static inline bool sk_stream_is_writeable(const struct sock *sk)
1209 {
1210         return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1211                sk_stream_memory_free(sk);
1212 }
1213 
1214 static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1215                                             struct cgroup *ancestor)
1216 {
1217 #ifdef CONFIG_SOCK_CGROUP_DATA
1218         return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1219                                     ancestor);
1220 #else
1221         return -ENOTSUPP;
1222 #endif
1223 }
1224 
1225 static inline bool sk_has_memory_pressure(const struct sock *sk)
1226 {
1227         return sk->sk_prot->memory_pressure != NULL;
1228 }
1229 
1230 static inline bool sk_under_memory_pressure(const struct sock *sk)
1231 {
1232         if (!sk->sk_prot->memory_pressure)
1233                 return false;
1234 
1235         if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1236             mem_cgroup_under_socket_pressure(sk->sk_memcg))
1237                 return true;
1238 
1239         return !!*sk->sk_prot->memory_pressure;
1240 }
1241 
1242 static inline long
1243 sk_memory_allocated(const struct sock *sk)
1244 {
1245         return atomic_long_read(sk->sk_prot->memory_allocated);
1246 }
1247 
1248 static inline long
1249 sk_memory_allocated_add(struct sock *sk, int amt)
1250 {
1251         return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
1252 }
1253 
1254 static inline void
1255 sk_memory_allocated_sub(struct sock *sk, int amt)
1256 {
1257         atomic_long_sub(amt, sk->sk_prot->memory_allocated);
1258 }
1259 
1260 static inline void sk_sockets_allocated_dec(struct sock *sk)
1261 {
1262         percpu_counter_dec(sk->sk_prot->sockets_allocated);
1263 }
1264 
1265 static inline void sk_sockets_allocated_inc(struct sock *sk)
1266 {
1267         percpu_counter_inc(sk->sk_prot->sockets_allocated);
1268 }
1269 
1270 static inline int
1271 sk_sockets_allocated_read_positive(struct sock *sk)
1272 {
1273         return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
1274 }
1275 
1276 static inline int
1277 proto_sockets_allocated_sum_positive(struct proto *prot)
1278 {
1279         return percpu_counter_sum_positive(prot->sockets_allocated);
1280 }
1281 
1282 static inline long
1283 proto_memory_allocated(struct proto *prot)
1284 {
1285         return atomic_long_read(prot->memory_allocated);
1286 }
1287 
1288 static inline bool
1289 proto_memory_pressure(struct proto *prot)
1290 {
1291         if (!prot->memory_pressure)
1292                 return false;
1293         return !!*prot->memory_pressure;
1294 }
1295 
1296 
1297 #ifdef CONFIG_PROC_FS
1298 /* Called with local bh disabled */
1299 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1300 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1301 int sock_inuse_get(struct net *net);
1302 #else
1303 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1304                 int inc)
1305 {
1306 }
1307 #endif
1308 
1309 
1310 /* With per-bucket locks this operation is not-atomic, so that
1311  * this version is not worse.
1312  */
1313 static inline int __sk_prot_rehash(struct sock *sk)
1314 {
1315         sk->sk_prot->unhash(sk);
1316         return sk->sk_prot->hash(sk);
1317 }
1318 
1319 /* About 10 seconds */
1320 #define SOCK_DESTROY_TIME (10*HZ)
1321 
1322 /* Sockets 0-1023 can't be bound to unless you are superuser */
1323 #define PROT_SOCK       1024
1324 
1325 #define SHUTDOWN_MASK   3
1326 #define RCV_SHUTDOWN    1
1327 #define SEND_SHUTDOWN   2
1328 
1329 #define SOCK_SNDBUF_LOCK        1
1330 #define SOCK_RCVBUF_LOCK        2
1331 #define SOCK_BINDADDR_LOCK      4
1332 #define SOCK_BINDPORT_LOCK      8
1333 
1334 struct socket_alloc {
1335         struct socket socket;
1336         struct inode vfs_inode;
1337 };
1338 
1339 static inline struct socket *SOCKET_I(struct inode *inode)
1340 {
1341         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1342 }
1343 
1344 static inline struct inode *SOCK_INODE(struct socket *socket)
1345 {
1346         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1347 }
1348 
1349 /*
1350  * Functions for memory accounting
1351  */
1352 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
1353 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1354 void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1355 void __sk_mem_reclaim(struct sock *sk, int amount);
1356 
1357 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1358  * do not necessarily have 16x time more memory than 4KB ones.
1359  */
1360 #define SK_MEM_QUANTUM 4096
1361 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1362 #define SK_MEM_SEND     0
1363 #define SK_MEM_RECV     1
1364 
1365 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1366 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1367 {
1368         long val = sk->sk_prot->sysctl_mem[index];
1369 
1370 #if PAGE_SIZE > SK_MEM_QUANTUM
1371         val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;
1372 #elif PAGE_SIZE < SK_MEM_QUANTUM
1373         val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT;
1374 #endif
1375         return val;
1376 }
1377 
1378 static inline int sk_mem_pages(int amt)
1379 {
1380         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1381 }
1382 
1383 static inline bool sk_has_account(struct sock *sk)
1384 {
1385         /* return true if protocol supports memory accounting */
1386         return !!sk->sk_prot->memory_allocated;
1387 }
1388 
1389 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1390 {
1391         if (!sk_has_account(sk))
1392                 return true;
1393         return size <= sk->sk_forward_alloc ||
1394                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1395 }
1396 
1397 static inline bool
1398 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1399 {
1400         if (!sk_has_account(sk))
1401                 return true;
1402         return size<= sk->sk_forward_alloc ||
1403                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1404                 skb_pfmemalloc(skb);
1405 }
1406 
1407 static inline void sk_mem_reclaim(struct sock *sk)
1408 {
1409         if (!sk_has_account(sk))
1410                 return;
1411         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1412                 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
1413 }
1414 
1415 static inline void sk_mem_reclaim_partial(struct sock *sk)
1416 {
1417         if (!sk_has_account(sk))
1418                 return;
1419         if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1420                 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
1421 }
1422 
1423 static inline void sk_mem_charge(struct sock *sk, int size)
1424 {
1425         if (!sk_has_account(sk))
1426                 return;
1427         sk->sk_forward_alloc -= size;
1428 }
1429 
1430 static inline void sk_mem_uncharge(struct sock *sk, int size)
1431 {
1432         if (!sk_has_account(sk))
1433                 return;
1434         sk->sk_forward_alloc += size;
1435 
1436         /* Avoid a possible overflow.
1437          * TCP send queues can make this happen, if sk_mem_reclaim()
1438          * is not called and more than 2 GBytes are released at once.
1439          *
1440          * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1441          * no need to hold that much forward allocation anyway.
1442          */
1443         if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1444                 __sk_mem_reclaim(sk, 1 << 20);
1445 }
1446 
1447 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1448 {
1449         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1450         sk->sk_wmem_queued -= skb->truesize;
1451         sk_mem_uncharge(sk, skb->truesize);
1452         __kfree_skb(skb);
1453 }
1454 
1455 static inline void sock_release_ownership(struct sock *sk)
1456 {
1457         if (sk->sk_lock.owned) {
1458                 sk->sk_lock.owned = 0;
1459 
1460                 /* The sk_lock has mutex_unlock() semantics: */
1461                 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1462         }
1463 }
1464 
1465 /*
1466  * Macro so as to not evaluate some arguments when
1467  * lockdep is not enabled.
1468  *
1469  * Mark both the sk_lock and the sk_lock.slock as a
1470  * per-address-family lock class.
1471  */
1472 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1473 do {                                                                    \
1474         sk->sk_lock.owned = 0;                                          \
1475         init_waitqueue_head(&sk->sk_lock.wq);                           \
1476         spin_lock_init(&(sk)->sk_lock.slock);                           \
1477         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1478                         sizeof((sk)->sk_lock));                         \
1479         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1480                                 (skey), (sname));                               \
1481         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1482 } while (0)
1483 
1484 #ifdef CONFIG_LOCKDEP
1485 static inline bool lockdep_sock_is_held(const struct sock *sk)
1486 {
1487         return lockdep_is_held(&sk->sk_lock) ||
1488                lockdep_is_held(&sk->sk_lock.slock);
1489 }
1490 #endif
1491 
1492 void lock_sock_nested(struct sock *sk, int subclass);
1493 
1494 static inline void lock_sock(struct sock *sk)
1495 {
1496         lock_sock_nested(sk, 0);
1497 }
1498 
1499 void __release_sock(struct sock *sk);
1500 void release_sock(struct sock *sk);
1501 
1502 /* BH context may only use the following locking interface. */
1503 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1504 #define bh_lock_sock_nested(__sk) \
1505                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1506                                 SINGLE_DEPTH_NESTING)
1507 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1508 
1509 bool lock_sock_fast(struct sock *sk);
1510 /**
1511  * unlock_sock_fast - complement of lock_sock_fast
1512  * @sk: socket
1513  * @slow: slow mode
1514  *
1515  * fast unlock socket for user context.
1516  * If slow mode is on, we call regular release_sock()
1517  */
1518 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1519 {
1520         if (slow)
1521                 release_sock(sk);
1522         else
1523                 spin_unlock_bh(&sk->sk_lock.slock);
1524 }
1525 
1526 /* Used by processes to "lock" a socket state, so that
1527  * interrupts and bottom half handlers won't change it
1528  * from under us. It essentially blocks any incoming
1529  * packets, so that we won't get any new data or any
1530  * packets that change the state of the socket.
1531  *
1532  * While locked, BH processing will add new packets to
1533  * the backlog queue.  This queue is processed by the
1534  * owner of the socket lock right before it is released.
1535  *
1536  * Since ~2.3.5 it is also exclusive sleep lock serializing
1537  * accesses from user process context.
1538  */
1539 
1540 static inline void sock_owned_by_me(const struct sock *sk)
1541 {
1542 #ifdef CONFIG_LOCKDEP
1543         WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
1544 #endif
1545 }
1546 
1547 static inline bool sock_owned_by_user(const struct sock *sk)
1548 {
1549         sock_owned_by_me(sk);
1550         return sk->sk_lock.owned;
1551 }
1552 
1553 static inline bool sock_owned_by_user_nocheck(const struct sock *sk)
1554 {
1555         return sk->sk_lock.owned;
1556 }
1557 
1558 /* no reclassification while locks are held */
1559 static inline bool sock_allow_reclassification(const struct sock *csk)
1560 {
1561         struct sock *sk = (struct sock *)csk;
1562 
1563         return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1564 }
1565 
1566 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1567                       struct proto *prot, int kern);
1568 void sk_free(struct sock *sk);
1569 void sk_destruct(struct sock *sk);
1570 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1571 void sk_free_unlock_clone(struct sock *sk);
1572 
1573 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1574                              gfp_t priority);
1575 void __sock_wfree(struct sk_buff *skb);
1576 void sock_wfree(struct sk_buff *skb);
1577 struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1578                              gfp_t priority);
1579 void skb_orphan_partial(struct sk_buff *skb);
1580 void sock_rfree(struct sk_buff *skb);
1581 void sock_efree(struct sk_buff *skb);
1582 #ifdef CONFIG_INET
1583 void sock_edemux(struct sk_buff *skb);
1584 #else
1585 #define sock_edemux sock_efree
1586 #endif
1587 
1588 int sock_setsockopt(struct socket *sock, int level, int op,
1589                     char __user *optval, unsigned int optlen);
1590 
1591 int sock_getsockopt(struct socket *sock, int level, int op,
1592                     char __user *optval, int __user *optlen);
1593 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1594                                     int noblock, int *errcode);
1595 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1596                                      unsigned long data_len, int noblock,
1597                                      int *errcode, int max_page_order);
1598 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1599 void sock_kfree_s(struct sock *sk, void *mem, int size);
1600 void sock_kzfree_s(struct sock *sk, void *mem, int size);
1601 void sk_send_sigurg(struct sock *sk);
1602 
1603 struct sockcm_cookie {
1604         u64 transmit_time;
1605         u32 mark;
1606         u16 tsflags;
1607 };
1608 
1609 static inline void sockcm_init(struct sockcm_cookie *sockc,
1610                                const struct sock *sk)
1611 {
1612         *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags };
1613 }
1614 
1615 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1616                      struct sockcm_cookie *sockc);
1617 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1618                    struct sockcm_cookie *sockc);
1619 
1620 /*
1621  * Functions to fill in entries in struct proto_ops when a protocol
1622  * does not implement a particular function.
1623  */
1624 int sock_no_bind(struct socket *, struct sockaddr *, int);
1625 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1626 int sock_no_socketpair(struct socket *, struct socket *);
1627 int sock_no_accept(struct socket *, struct socket *, int, bool);
1628 int sock_no_getname(struct socket *, struct sockaddr *, int);
1629 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1630 int sock_no_listen(struct socket *, int);
1631 int sock_no_shutdown(struct socket *, int);
1632 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1633 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1634 int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1635 int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len);
1636 int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
1637 int sock_no_mmap(struct file *file, struct socket *sock,
1638                  struct vm_area_struct *vma);
1639 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1640                          size_t size, int flags);
1641 ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
1642                                 int offset, size_t size, int flags);
1643 
1644 /*
1645  * Functions to fill in entries in struct proto_ops when a protocol
1646  * uses the inet style.
1647  */
1648 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1649                                   char __user *optval, int __user *optlen);
1650 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1651                         int flags);
1652 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1653                                   char __user *optval, unsigned int optlen);
1654 int compat_sock_common_getsockopt(struct socket *sock, int level,
1655                 int optname, char __user *optval, int __user *optlen);
1656 int compat_sock_common_setsockopt(struct socket *sock, int level,
1657                 int optname, char __user *optval, unsigned int optlen);
1658 
1659 void sk_common_release(struct sock *sk);
1660 
1661 /*
1662  *      Default socket callbacks and setup code
1663  */
1664 
1665 /* Initialise core socket variables */
1666 void sock_init_data(struct socket *sock, struct sock *sk);
1667 
1668 /*
1669  * Socket reference counting postulates.
1670  *
1671  * * Each user of socket SHOULD hold a reference count.
1672  * * Each access point to socket (an hash table bucket, reference from a list,
1673  *   running timer, skb in flight MUST hold a reference count.
1674  * * When reference count hits 0, it means it will never increase back.
1675  * * When reference count hits 0, it means that no references from
1676  *   outside exist to this socket and current process on current CPU
1677  *   is last user and may/should destroy this socket.
1678  * * sk_free is called from any context: process, BH, IRQ. When
1679  *   it is called, socket has no references from outside -> sk_free
1680  *   may release descendant resources allocated by the socket, but
1681  *   to the time when it is called, socket is NOT referenced by any
1682  *   hash tables, lists etc.
1683  * * Packets, delivered from outside (from network or from another process)
1684  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1685  *   when they sit in queue. Otherwise, packets will leak to hole, when
1686  *   socket is looked up by one cpu and unhasing is made by another CPU.
1687  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1688  *   (leak to backlog). Packet socket does all the processing inside
1689  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1690  *   use separate SMP lock, so that they are prone too.
1691  */
1692 
1693 /* Ungrab socket and destroy it, if it was the last reference. */
1694 static inline void sock_put(struct sock *sk)
1695 {
1696         if (refcount_dec_and_test(&sk->sk_refcnt))
1697                 sk_free(sk);
1698 }
1699 /* Generic version of sock_put(), dealing with all sockets
1700  * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1701  */
1702 void sock_gen_put(struct sock *sk);
1703 
1704 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
1705                      unsigned int trim_cap, bool refcounted);
1706 static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1707                                  const int nested)
1708 {
1709         return __sk_receive_skb(sk, skb, nested, 1, true);
1710 }
1711 
1712 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1713 {
1714         /* sk_tx_queue_mapping accept only upto a 16-bit value */
1715         if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX))
1716                 return;
1717         sk->sk_tx_queue_mapping = tx_queue;
1718 }
1719 
1720 #define NO_QUEUE_MAPPING        USHRT_MAX
1721 
1722 static inline void sk_tx_queue_clear(struct sock *sk)
1723 {
1724         sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING;
1725 }
1726 
1727 static inline int sk_tx_queue_get(const struct sock *sk)
1728 {
1729         if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING)
1730                 return sk->sk_tx_queue_mapping;
1731 
1732         return -1;
1733 }
1734 
1735 static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb)
1736 {
1737 #ifdef CONFIG_XPS
1738         if (skb_rx_queue_recorded(skb)) {
1739                 u16 rx_queue = skb_get_rx_queue(skb);
1740 
1741                 if (WARN_ON_ONCE(rx_queue == NO_QUEUE_MAPPING))
1742                         return;
1743 
1744                 sk->sk_rx_queue_mapping = rx_queue;
1745         }
1746 #endif
1747 }
1748 
1749 static inline void sk_rx_queue_clear(struct sock *sk)
1750 {
1751 #ifdef CONFIG_XPS
1752         sk->sk_rx_queue_mapping = NO_QUEUE_MAPPING;
1753 #endif
1754 }
1755 
1756 #ifdef CONFIG_XPS
1757 static inline int sk_rx_queue_get(const struct sock *sk)
1758 {
1759         if (sk && sk->sk_rx_queue_mapping != NO_QUEUE_MAPPING)
1760                 return sk->sk_rx_queue_mapping;
1761 
1762         return -1;
1763 }
1764 #endif
1765 
1766 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1767 {
1768         sk_tx_queue_clear(sk);
1769         sk->sk_socket = sock;
1770 }
1771 
1772 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1773 {
1774         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1775         return &rcu_dereference_raw(sk->sk_wq)->wait;
1776 }
1777 /* Detach socket from process context.
1778  * Announce socket dead, detach it from wait queue and inode.
1779  * Note that parent inode held reference count on this struct sock,
1780  * we do not release it in this function, because protocol
1781  * probably wants some additional cleanups or even continuing
1782  * to work with this socket (TCP).
1783  */
1784 static inline void sock_orphan(struct sock *sk)
1785 {
1786         write_lock_bh(&sk->sk_callback_lock);
1787         sock_set_flag(sk, SOCK_DEAD);
1788         sk_set_socket(sk, NULL);
1789         sk->sk_wq  = NULL;
1790         write_unlock_bh(&sk->sk_callback_lock);
1791 }
1792 
1793 static inline void sock_graft(struct sock *sk, struct socket *parent)
1794 {
1795         WARN_ON(parent->sk);
1796         write_lock_bh(&sk->sk_callback_lock);
1797         rcu_assign_pointer(sk->sk_wq, parent->wq);
1798         parent->sk = sk;
1799         sk_set_socket(sk, parent);
1800         sk->sk_uid = SOCK_INODE(parent)->i_uid;
1801         security_sock_graft(sk, parent);
1802         write_unlock_bh(&sk->sk_callback_lock);
1803 }
1804 
1805 kuid_t sock_i_uid(struct sock *sk);
1806 unsigned long sock_i_ino(struct sock *sk);
1807 
1808 static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
1809 {
1810         return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
1811 }
1812 
1813 static inline u32 net_tx_rndhash(void)
1814 {
1815         u32 v = prandom_u32();
1816 
1817         return v ?: 1;
1818 }
1819 
1820 static inline void sk_set_txhash(struct sock *sk)
1821 {
1822         sk->sk_txhash = net_tx_rndhash();
1823 }
1824 
1825 static inline void sk_rethink_txhash(struct sock *sk)
1826 {
1827         if (sk->sk_txhash)
1828                 sk_set_txhash(sk);
1829 }
1830 
1831 static inline struct dst_entry *
1832 __sk_dst_get(struct sock *sk)
1833 {
1834         return rcu_dereference_check(sk->sk_dst_cache,
1835                                      lockdep_sock_is_held(sk));
1836 }
1837 
1838 static inline struct dst_entry *
1839 sk_dst_get(struct sock *sk)
1840 {
1841         struct dst_entry *dst;
1842 
1843         rcu_read_lock();
1844         dst = rcu_dereference(sk->sk_dst_cache);
1845         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1846                 dst = NULL;
1847         rcu_read_unlock();
1848         return dst;
1849 }
1850 
1851 static inline void dst_negative_advice(struct sock *sk)
1852 {
1853         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1854 
1855         sk_rethink_txhash(sk);
1856 
1857         if (dst && dst->ops->negative_advice) {
1858                 ndst = dst->ops->negative_advice(dst);
1859 
1860                 if (ndst != dst) {
1861                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1862                         sk_tx_queue_clear(sk);
1863                         sk->sk_dst_pending_confirm = 0;
1864                 }
1865         }
1866 }
1867 
1868 static inline void
1869 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1870 {
1871         struct dst_entry *old_dst;
1872 
1873         sk_tx_queue_clear(sk);
1874         sk->sk_dst_pending_confirm = 0;
1875         old_dst = rcu_dereference_protected(sk->sk_dst_cache,
1876                                             lockdep_sock_is_held(sk));
1877         rcu_assign_pointer(sk->sk_dst_cache, dst);
1878         dst_release(old_dst);
1879 }
1880 
1881 static inline void
1882 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1883 {
1884         struct dst_entry *old_dst;
1885 
1886         sk_tx_queue_clear(sk);
1887         sk->sk_dst_pending_confirm = 0;
1888         old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1889         dst_release(old_dst);
1890 }
1891 
1892 static inline void
1893 __sk_dst_reset(struct sock *sk)
1894 {
1895         __sk_dst_set(sk, NULL);
1896 }
1897 
1898 static inline void
1899 sk_dst_reset(struct sock *sk)
1900 {
1901         sk_dst_set(sk, NULL);
1902 }
1903 
1904 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1905 
1906 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1907 
1908 static inline void sk_dst_confirm(struct sock *sk)
1909 {
1910         if (!sk->sk_dst_pending_confirm)
1911                 sk->sk_dst_pending_confirm = 1;
1912 }
1913 
1914 static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
1915 {
1916         if (skb_get_dst_pending_confirm(skb)) {
1917                 struct sock *sk = skb->sk;
1918                 unsigned long now = jiffies;
1919 
1920                 /* avoid dirtying neighbour */
1921                 if (n->confirmed != now)
1922                         n->confirmed = now;
1923                 if (sk && sk->sk_dst_pending_confirm)
1924                         sk->sk_dst_pending_confirm = 0;
1925         }
1926 }
1927 
1928 bool sk_mc_loop(struct sock *sk);
1929 
1930 static inline bool sk_can_gso(const struct sock *sk)
1931 {
1932         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1933 }
1934 
1935 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1936 
1937 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1938 {
1939         sk->sk_route_nocaps |= flags;
1940         sk->sk_route_caps &= ~flags;
1941 }
1942 
1943 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1944                                            struct iov_iter *from, char *to,
1945                                            int copy, int offset)
1946 {
1947         if (skb->ip_summed == CHECKSUM_NONE) {
1948                 __wsum csum = 0;
1949                 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
1950                         return -EFAULT;
1951                 skb->csum = csum_block_add(skb->csum, csum, offset);
1952         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1953                 if (!copy_from_iter_full_nocache(to, copy, from))
1954                         return -EFAULT;
1955         } else if (!copy_from_iter_full(to, copy, from))
1956                 return -EFAULT;
1957 
1958         return 0;
1959 }
1960 
1961 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1962                                        struct iov_iter *from, int copy)
1963 {
1964         int err, offset = skb->len;
1965 
1966         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1967                                        copy, offset);
1968         if (err)
1969                 __skb_trim(skb, offset);
1970 
1971         return err;
1972 }
1973 
1974 static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
1975                                            struct sk_buff *skb,
1976                                            struct page *page,
1977                                            int off, int copy)
1978 {
1979         int err;
1980 
1981         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1982                                        copy, skb->len);
1983         if (err)
1984                 return err;
1985 
1986         skb->len             += copy;
1987         skb->data_len        += copy;
1988         skb->truesize        += copy;
1989         sk->sk_wmem_queued   += copy;
1990         sk_mem_charge(sk, copy);
1991         return 0;
1992 }
1993 
1994 /**
1995  * sk_wmem_alloc_get - returns write allocations
1996  * @sk: socket
1997  *
1998  * Returns sk_wmem_alloc minus initial offset of one
1999  */
2000 static inline int sk_wmem_alloc_get(const struct sock *sk)
2001 {
2002         return refcount_read(&sk->sk_wmem_alloc) - 1;
2003 }
2004 
2005 /**
2006  * sk_rmem_alloc_get - returns read allocations
2007  * @sk: socket
2008  *
2009  * Returns sk_rmem_alloc
2010  */
2011 static inline int sk_rmem_alloc_get(const struct sock *sk)
2012 {
2013         return atomic_read(&sk->sk_rmem_alloc);
2014 }
2015 
2016 /**
2017  * sk_has_allocations - check if allocations are outstanding
2018  * @sk: socket
2019  *
2020  * Returns true if socket has write or read allocations
2021  */
2022 static inline bool sk_has_allocations(const struct sock *sk)
2023 {
2024         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
2025 }
2026 
2027 /**
2028  * skwq_has_sleeper - check if there are any waiting processes
2029  * @wq: struct socket_wq
2030  *
2031  * Returns true if socket_wq has waiting processes
2032  *
2033  * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
2034  * barrier call. They were added due to the race found within the tcp code.
2035  *
2036  * Consider following tcp code paths::
2037  *
2038  *   CPU1                CPU2
2039  *   sys_select          receive packet
2040  *   ...                 ...
2041  *   __add_wait_queue    update tp->rcv_nxt
2042  *   ...                 ...
2043  *   tp->rcv_nxt check   sock_def_readable
2044  *   ...                 {
2045  *   schedule               rcu_read_lock();
2046  *                          wq = rcu_dereference(sk->sk_wq);
2047  *                          if (wq && waitqueue_active(&wq->wait))
2048  *                              wake_up_interruptible(&wq->wait)
2049  *                          ...
2050  *                       }
2051  *
2052  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2053  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
2054  * could then endup calling schedule and sleep forever if there are no more
2055  * data on the socket.
2056  *
2057  */
2058 static inline bool skwq_has_sleeper(struct socket_wq *wq)
2059 {
2060         return wq && wq_has_sleeper(&wq->wait);
2061 }
2062 
2063 /**
2064  * sock_poll_wait - place memory barrier behind the poll_wait call.
2065  * @filp:           file
2066  * @sock:           socket to wait on
2067  * @p:              poll_table
2068  *
2069  * See the comments in the wq_has_sleeper function.
2070  *
2071  * Do not derive sock from filp->private_data here. An SMC socket establishes
2072  * an internal TCP socket that is used in the fallback case. All socket
2073  * operations on the SMC socket are then forwarded to the TCP socket. In case of
2074  * poll, the filp->private_data pointer references the SMC socket because the
2075  * TCP socket has no file assigned.
2076  */
2077 static inline void sock_poll_wait(struct file *filp, struct socket *sock,
2078                                   poll_table *p)
2079 {
2080         if (!poll_does_not_wait(p)) {
2081                 poll_wait(filp, &sock->wq->wait, p);
2082                 /* We need to be sure we are in sync with the
2083                  * socket flags modification.
2084                  *
2085                  * This memory barrier is paired in the wq_has_sleeper.
2086                  */
2087                 smp_mb();
2088         }
2089 }
2090 
2091 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2092 {
2093         if (sk->sk_txhash) {
2094                 skb->l4_hash = 1;
2095                 skb->hash = sk->sk_txhash;
2096         }
2097 }
2098 
2099 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2100 
2101 /*
2102  *      Queue a received datagram if it will fit. Stream and sequenced
2103  *      protocols can't normally use this as they need to fit buffers in
2104  *      and play with them.
2105  *
2106  *      Inlined as it's very short and called for pretty much every
2107  *      packet ever received.
2108  */
2109 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2110 {
2111         skb_orphan(skb);
2112         skb->sk = sk;
2113         skb->destructor = sock_rfree;
2114         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2115         sk_mem_charge(sk, skb->truesize);
2116 }
2117 
2118 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2119                     unsigned long expires);
2120 
2121 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2122 
2123 int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2124                         struct sk_buff *skb, unsigned int flags,
2125                         void (*destructor)(struct sock *sk,
2126                                            struct sk_buff *skb));
2127 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2128 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2129 
2130 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2131 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
2132 
2133 /*
2134  *      Recover an error report and clear atomically
2135  */
2136 
2137 static inline int sock_error(struct sock *sk)
2138 {
2139         int err;
2140         if (likely(!sk->sk_err))
2141                 return 0;
2142         err = xchg(&sk->sk_err, 0);
2143         return -err;
2144 }
2145 
2146 static inline unsigned long sock_wspace(struct sock *sk)
2147 {
2148         int amt = 0;
2149 
2150         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2151                 amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc);
2152                 if (amt < 0)
2153                         amt = 0;
2154         }
2155         return amt;
2156 }
2157 
2158 /* Note:
2159  *  We use sk->sk_wq_raw, from contexts knowing this
2160  *  pointer is not NULL and cannot disappear/change.
2161  */
2162 static inline void sk_set_bit(int nr, struct sock *sk)
2163 {
2164         if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2165             !sock_flag(sk, SOCK_FASYNC))
2166                 return;
2167 
2168         set_bit(nr, &sk->sk_wq_raw->flags);
2169 }
2170 
2171 static inline void sk_clear_bit(int nr, struct sock *sk)
2172 {
2173         if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2174             !sock_flag(sk, SOCK_FASYNC))
2175                 return;
2176 
2177         clear_bit(nr, &sk->sk_wq_raw->flags);
2178 }
2179 
2180 static inline void sk_wake_async(const struct sock *sk, int how, int band)
2181 {
2182         if (sock_flag(sk, SOCK_FASYNC)) {
2183                 rcu_read_lock();
2184                 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2185                 rcu_read_unlock();
2186         }
2187 }
2188 
2189 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2190  * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2191  * Note: for send buffers, TCP works better if we can build two skbs at
2192  * minimum.
2193  */
2194 #define TCP_SKB_MIN_TRUESIZE    (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2195 
2196 #define SOCK_MIN_SNDBUF         (TCP_SKB_MIN_TRUESIZE * 2)
2197 #define SOCK_MIN_RCVBUF          TCP_SKB_MIN_TRUESIZE
2198 
2199 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2200 {
2201         if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2202                 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2203                 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2204         }
2205 }
2206 
2207 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2208                                     bool force_schedule);
2209 
2210 /**
2211  * sk_page_frag - return an appropriate page_frag
2212  * @sk: socket
2213  *
2214  * If socket allocation mode allows current thread to sleep, it means its
2215  * safe to use the per task page_frag instead of the per socket one.
2216  */
2217 static inline struct page_frag *sk_page_frag(struct sock *sk)
2218 {
2219         if (gfpflags_allow_blocking(sk->sk_allocation))
2220                 return &current->task_frag;
2221 
2222         return &sk->sk_frag;
2223 }
2224 
2225 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2226 
2227 /*
2228  *      Default write policy as shown to user space via poll/select/SIGIO
2229  */
2230 static inline bool sock_writeable(const struct sock *sk)
2231 {
2232         return refcount_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2233 }
2234 
2235 static inline gfp_t gfp_any(void)
2236 {
2237         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2238 }
2239 
2240 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2241 {
2242         return noblock ? 0 : sk->sk_rcvtimeo;
2243 }
2244 
2245 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2246 {
2247         return noblock ? 0 : sk->sk_sndtimeo;
2248 }
2249 
2250 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2251 {
2252         return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2253 }
2254 
2255 /* Alas, with timeout socket operations are not restartable.
2256  * Compare this to poll().
2257  */
2258 static inline int sock_intr_errno(long timeo)
2259 {
2260         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2261 }
2262 
2263 struct sock_skb_cb {
2264         u32 dropcount;
2265 };
2266 
2267 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2268  * using skb->cb[] would keep using it directly and utilize its
2269  * alignement guarantee.
2270  */
2271 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2272                             sizeof(struct sock_skb_cb)))
2273 
2274 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2275                             SOCK_SKB_CB_OFFSET))
2276 
2277 #define sock_skb_cb_check_size(size) \
2278         BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2279 
2280 static inline void
2281 sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2282 {
2283         SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2284                                                 atomic_read(&sk->sk_drops) : 0;
2285 }
2286 
2287 static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2288 {
2289         int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2290 
2291         atomic_add(segs, &sk->sk_drops);
2292 }
2293 
2294 static inline ktime_t sock_read_timestamp(struct sock *sk)
2295 {
2296 #if BITS_PER_LONG==32
2297         unsigned int seq;
2298         ktime_t kt;
2299 
2300         do {
2301                 seq = read_seqbegin(&sk->sk_stamp_seq);
2302                 kt = sk->sk_stamp;
2303         } while (read_seqretry(&sk->sk_stamp_seq, seq));
2304 
2305         return kt;
2306 #else
2307         return sk->sk_stamp;
2308 #endif
2309 }
2310 
2311 static inline void sock_write_timestamp(struct sock *sk, ktime_t kt)
2312 {
2313 #if BITS_PER_LONG==32
2314         write_seqlock(&sk->sk_stamp_seq);
2315         sk->sk_stamp = kt;
2316         write_sequnlock(&sk->sk_stamp_seq);
2317 #else
2318         sk->sk_stamp = kt;
2319 #endif
2320 }
2321 
2322 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2323                            struct sk_buff *skb);
2324 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2325                              struct sk_buff *skb);
2326 
2327 static inline void
2328 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2329 {
2330         ktime_t kt = skb->tstamp;
2331         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2332 
2333         /*
2334          * generate control messages if
2335          * - receive time stamping in software requested
2336          * - software time stamp available and wanted
2337          * - hardware time stamps available and wanted
2338          */
2339         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2340             (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2341             (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2342             (hwtstamps->hwtstamp &&
2343              (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2344                 __sock_recv_timestamp(msg, sk, skb);
2345         else
2346                 sock_write_timestamp(sk, kt);
2347 
2348         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2349                 __sock_recv_wifi_status(msg, sk, skb);
2350 }
2351 
2352 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2353                               struct sk_buff *skb);
2354 
2355 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2356 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2357                                           struct sk_buff *skb)
2358 {
2359 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2360                            (1UL << SOCK_RCVTSTAMP))
2361 #define TSFLAGS_ANY       (SOF_TIMESTAMPING_SOFTWARE                    | \
2362                            SOF_TIMESTAMPING_RAW_HARDWARE)
2363 
2364         if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2365                 __sock_recv_ts_and_drops(msg, sk, skb);
2366         else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
2367                 sock_write_timestamp(sk, skb->tstamp);
2368         else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP))
2369                 sock_write_timestamp(sk, 0);
2370 }
2371 
2372 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
2373 
2374 /**
2375  * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2376  * @sk:         socket sending this packet
2377  * @tsflags:    timestamping flags to use
2378  * @tx_flags:   completed with instructions for time stamping
2379  * @tskey:      filled in with next sk_tskey (not for TCP, which uses seqno)
2380  *
2381  * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2382  */
2383 static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2384                                       __u8 *tx_flags, __u32 *tskey)
2385 {
2386         if (unlikely(tsflags)) {
2387                 __sock_tx_timestamp(tsflags, tx_flags);
2388                 if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey &&
2389                     tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK)
2390                         *tskey = sk->sk_tskey++;
2391         }
2392         if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2393                 *tx_flags |= SKBTX_WIFI_STATUS;
2394 }
2395 
2396 static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2397                                      __u8 *tx_flags)
2398 {
2399         _sock_tx_timestamp(sk, tsflags, tx_flags, NULL);
2400 }
2401 
2402 static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags)
2403 {
2404         _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags,
2405                            &skb_shinfo(skb)->tskey);
2406 }
2407 
2408 /**
2409  * sk_eat_skb - Release a skb if it is no longer needed
2410  * @sk: socket to eat this skb from
2411  * @skb: socket buffer to eat
2412  *
2413  * This routine must be called with interrupts disabled or with the socket
2414  * locked so that the sk_buff queue operation is ok.
2415 */
2416 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2417 {
2418         __skb_unlink(skb, &sk->sk_receive_queue);
2419         __kfree_skb(skb);
2420 }
2421 
2422 static inline
2423 struct net *sock_net(const struct sock *sk)
2424 {
2425         return read_pnet(&sk->sk_net);
2426 }
2427 
2428 static inline
2429 void sock_net_set(struct sock *sk, struct net *net)
2430 {
2431         write_pnet(&sk->sk_net, net);
2432 }
2433 
2434 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2435 {
2436         if (skb->sk) {
2437                 struct sock *sk = skb->sk;
2438 
2439                 skb->destructor = NULL;
2440                 skb->sk = NULL;
2441                 return sk;
2442         }
2443         return NULL;
2444 }
2445 
2446 /* This helper checks if a socket is a full socket,
2447  * ie _not_ a timewait or request socket.
2448  */
2449 static inline bool sk_fullsock(const struct sock *sk)
2450 {
2451         return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2452 }
2453 
2454 /* Checks if this SKB belongs to an HW offloaded socket
2455  * and whether any SW fallbacks are required based on dev.
2456  */
2457 static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb,
2458                                                    struct net_device *dev)
2459 {
2460 #ifdef CONFIG_SOCK_VALIDATE_XMIT
2461         struct sock *sk = skb->sk;
2462 
2463         if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb)
2464                 skb = sk->sk_validate_xmit_skb(sk, dev, skb);
2465 #endif
2466 
2467         return skb;
2468 }
2469 
2470 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2471  * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2472  */
2473 static inline bool sk_listener(const struct sock *sk)
2474 {
2475         return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2476 }
2477 
2478 void sock_enable_timestamp(struct sock *sk, int flag);
2479 int sock_get_timestamp(struct sock *, struct timeval __user *);
2480 int sock_get_timestampns(struct sock *, struct timespec __user *);
2481 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2482                        int type);
2483 
2484 bool sk_ns_capable(const struct sock *sk,
2485                    struct user_namespace *user_ns, int cap);
2486 bool sk_capable(const struct sock *sk, int cap);
2487 bool sk_net_capable(const struct sock *sk, int cap);
2488 
2489 void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2490 
2491 /* Take into consideration the size of the struct sk_buff overhead in the
2492  * determination of these values, since that is non-constant across
2493  * platforms.  This makes socket queueing behavior and performance
2494  * not depend upon such differences.
2495  */
2496 #define _SK_MEM_PACKETS         256
2497 #define _SK_MEM_OVERHEAD        SKB_TRUESIZE(256)
2498 #define SK_WMEM_MAX             (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2499 #define SK_RMEM_MAX             (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2500 
2501 extern __u32 sysctl_wmem_max;
2502 extern __u32 sysctl_rmem_max;
2503 
2504 extern int sysctl_tstamp_allow_data;
2505 extern int sysctl_optmem_max;
2506 
2507 extern __u32 sysctl_wmem_default;
2508 extern __u32 sysctl_rmem_default;
2509 
2510 static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
2511 {
2512         /* Does this proto have per netns sysctl_wmem ? */
2513         if (proto->sysctl_wmem_offset)
2514                 return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset);
2515 
2516         return *proto->sysctl_wmem;
2517 }
2518 
2519 static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
2520 {
2521         /* Does this proto have per netns sysctl_rmem ? */
2522         if (proto->sysctl_rmem_offset)
2523                 return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset);
2524 
2525         return *proto->sysctl_rmem;
2526 }
2527 
2528 /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2529  * Some wifi drivers need to tweak it to get more chunks.
2530  * They can use this helper from their ndo_start_xmit()
2531  */
2532 static inline void sk_pacing_shift_update(struct sock *sk, int val)
2533 {
2534         if (!sk || !sk_fullsock(sk) || sk->sk_pacing_shift == val)
2535                 return;
2536         sk->sk_pacing_shift = val;
2537 }
2538 
2539 /* if a socket is bound to a device, check that the given device
2540  * index is either the same or that the socket is bound to an L3
2541  * master device and the given device index is also enslaved to
2542  * that L3 master
2543  */
2544 static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
2545 {
2546         int mdif;
2547 
2548         if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif)
2549                 return true;
2550 
2551         mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif);
2552         if (mdif && mdif == sk->sk_bound_dev_if)
2553                 return true;
2554 
2555         return false;
2556 }
2557 
2558 #endif  /* _SOCK_H */
2559 

~ [ 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