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

TOMOYO Linux Cross Reference
Linux/include/linux/tcp.h

Version: ~ [ linux-5.12 ] ~ [ linux-5.11.16 ] ~ [ linux-5.10.32 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.114 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.188 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.231 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.267 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.267 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
  3  *              operating system.  INET is implemented using the  BSD Socket
  4  *              interface as the means of communication with the user level.
  5  *
  6  *              Definitions for the TCP protocol.
  7  *
  8  * Version:     @(#)tcp.h       1.0.2   04/28/93
  9  *
 10  * Author:      Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 11  *
 12  *              This program is free software; you can redistribute it and/or
 13  *              modify it under the terms of the GNU General Public License
 14  *              as published by the Free Software Foundation; either version
 15  *              2 of the License, or (at your option) any later version.
 16  */
 17 #ifndef _LINUX_TCP_H
 18 #define _LINUX_TCP_H
 19 
 20 
 21 #include <linux/skbuff.h>
 22 #include <linux/win_minmax.h>
 23 #include <net/sock.h>
 24 #include <net/inet_connection_sock.h>
 25 #include <net/inet_timewait_sock.h>
 26 #include <uapi/linux/tcp.h>
 27 
 28 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
 29 {
 30         return (struct tcphdr *)skb_transport_header(skb);
 31 }
 32 
 33 static inline unsigned int __tcp_hdrlen(const struct tcphdr *th)
 34 {
 35         return th->doff * 4;
 36 }
 37 
 38 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
 39 {
 40         return __tcp_hdrlen(tcp_hdr(skb));
 41 }
 42 
 43 static inline struct tcphdr *inner_tcp_hdr(const struct sk_buff *skb)
 44 {
 45         return (struct tcphdr *)skb_inner_transport_header(skb);
 46 }
 47 
 48 static inline unsigned int inner_tcp_hdrlen(const struct sk_buff *skb)
 49 {
 50         return inner_tcp_hdr(skb)->doff * 4;
 51 }
 52 
 53 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
 54 {
 55         return (tcp_hdr(skb)->doff - 5) * 4;
 56 }
 57 
 58 /* TCP Fast Open */
 59 #define TCP_FASTOPEN_COOKIE_MIN 4       /* Min Fast Open Cookie size in bytes */
 60 #define TCP_FASTOPEN_COOKIE_MAX 16      /* Max Fast Open Cookie size in bytes */
 61 #define TCP_FASTOPEN_COOKIE_SIZE 8      /* the size employed by this impl. */
 62 
 63 /* TCP Fast Open Cookie as stored in memory */
 64 struct tcp_fastopen_cookie {
 65         union {
 66                 u8      val[TCP_FASTOPEN_COOKIE_MAX];
 67 #if IS_ENABLED(CONFIG_IPV6)
 68                 struct in6_addr addr;
 69 #endif
 70         };
 71         s8      len;
 72         bool    exp;    /* In RFC6994 experimental option format */
 73 };
 74 
 75 /* This defines a selective acknowledgement block. */
 76 struct tcp_sack_block_wire {
 77         __be32  start_seq;
 78         __be32  end_seq;
 79 };
 80 
 81 struct tcp_sack_block {
 82         u32     start_seq;
 83         u32     end_seq;
 84 };
 85 
 86 /*These are used to set the sack_ok field in struct tcp_options_received */
 87 #define TCP_SACK_SEEN     (1 << 0)   /*1 = peer is SACK capable, */
 88 #define TCP_FACK_ENABLED  (1 << 1)   /*1 = FACK is enabled locally*/
 89 #define TCP_DSACK_SEEN    (1 << 2)   /*1 = DSACK was received from peer*/
 90 
 91 struct tcp_options_received {
 92 /*      PAWS/RTTM data  */
 93         long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
 94         u32     ts_recent;      /* Time stamp to echo next              */
 95         u32     rcv_tsval;      /* Time stamp value                     */
 96         u32     rcv_tsecr;      /* Time stamp echo reply                */
 97         u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
 98                 tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
 99                 dsack : 1,      /* D-SACK is scheduled                  */
100                 wscale_ok : 1,  /* Wscale seen on SYN packet            */
101                 sack_ok : 4,    /* SACK seen on SYN packet              */
102                 snd_wscale : 4, /* Window scaling received from sender  */
103                 rcv_wscale : 4; /* Window scaling to send to receiver   */
104         u8      num_sacks;      /* Number of SACK blocks                */
105         u16     user_mss;       /* mss requested by user in ioctl       */
106         u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
107 };
108 
109 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
110 {
111         rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
112         rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
113 }
114 
115 /* This is the max number of SACKS that we'll generate and process. It's safe
116  * to increase this, although since:
117  *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
118  * only four options will fit in a standard TCP header */
119 #define TCP_NUM_SACKS 4
120 
121 struct tcp_request_sock_ops;
122 
123 struct tcp_request_sock {
124         struct inet_request_sock        req;
125         const struct tcp_request_sock_ops *af_specific;
126         struct skb_mstamp               snt_synack; /* first SYNACK sent time */
127         bool                            tfo_listener;
128         u32                             txhash;
129         u32                             rcv_isn;
130         u32                             snt_isn;
131         u32                             ts_off;
132         u32                             last_oow_ack_time; /* last SYNACK */
133         u32                             rcv_nxt; /* the ack # by SYNACK. For
134                                                   * FastOpen it's the seq#
135                                                   * after data-in-SYN.
136                                                   */
137 };
138 
139 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
140 {
141         return (struct tcp_request_sock *)req;
142 }
143 
144 struct tcp_sock {
145         /* inet_connection_sock has to be the first member of tcp_sock */
146         struct inet_connection_sock     inet_conn;
147         u16     tcp_header_len; /* Bytes of tcp header to send          */
148         u16     gso_segs;       /* Max number of segs per GSO packet    */
149 
150 /*
151  *      Header prediction flags
152  *      0x5?10 << 16 + snd_wnd in net byte order
153  */
154         __be32  pred_flags;
155 
156 /*
157  *      RFC793 variables by their proper names. This means you can
158  *      read the code and the spec side by side (and laugh ...)
159  *      See RFC793 and RFC1122. The RFC writes these in capitals.
160  */
161         u64     bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
162                                  * sum(delta(rcv_nxt)), or how many bytes
163                                  * were acked.
164                                  */
165         u32     segs_in;        /* RFC4898 tcpEStatsPerfSegsIn
166                                  * total number of segments in.
167                                  */
168         u32     data_segs_in;   /* RFC4898 tcpEStatsPerfDataSegsIn
169                                  * total number of data segments in.
170                                  */
171         u32     rcv_nxt;        /* What we want to receive next         */
172         u32     copied_seq;     /* Head of yet unread data              */
173         u32     rcv_wup;        /* rcv_nxt on last window update sent   */
174         u32     snd_nxt;        /* Next sequence we send                */
175         u32     segs_out;       /* RFC4898 tcpEStatsPerfSegsOut
176                                  * The total number of segments sent.
177                                  */
178         u32     data_segs_out;  /* RFC4898 tcpEStatsPerfDataSegsOut
179                                  * total number of data segments sent.
180                                  */
181         u64     bytes_acked;    /* RFC4898 tcpEStatsAppHCThruOctetsAcked
182                                  * sum(delta(snd_una)), or how many bytes
183                                  * were acked.
184                                  */
185         u32     snd_una;        /* First byte we want an ack for        */
186         u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
187         u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
188         u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
189         u32     last_oow_ack_time;  /* timestamp of last out-of-window ACK */
190 
191         u32     tsoffset;       /* timestamp offset */
192 
193         struct list_head tsq_node; /* anchor in tsq_tasklet.head list */
194 
195         /* Data for direct copy to user */
196         struct {
197                 struct sk_buff_head     prequeue;
198                 struct task_struct      *task;
199                 struct msghdr           *msg;
200                 int                     memory;
201                 int                     len;
202         } ucopy;
203 
204         u32     snd_wl1;        /* Sequence for window update           */
205         u32     snd_wnd;        /* The window we expect to receive      */
206         u32     max_window;     /* Maximal window ever seen from peer   */
207         u32     mss_cache;      /* Cached effective mss, not including SACKS */
208 
209         u32     window_clamp;   /* Maximal window to advertise          */
210         u32     rcv_ssthresh;   /* Current window clamp                 */
211 
212         /* Information of the most recently (s)acked skb */
213         struct tcp_rack {
214                 struct skb_mstamp mstamp; /* (Re)sent time of the skb */
215                 u32 rtt_us;  /* Associated RTT */
216                 u32 end_seq; /* Ending TCP sequence of the skb */
217                 u8 advanced; /* mstamp advanced since last lost marking */
218                 u8 reord;    /* reordering detected */
219         } rack;
220         u16     advmss;         /* Advertised MSS                       */
221         u32     chrono_start;   /* Start time in jiffies of a TCP chrono */
222         u32     chrono_stat[3]; /* Time in jiffies for chrono_stat stats */
223         u8      chrono_type:2,  /* current chronograph type */
224                 rate_app_limited:1,  /* rate_{delivered,interval_us} limited? */
225                 fastopen_connect:1, /* FASTOPEN_CONNECT sockopt */
226                 unused:4;
227         u8      nonagle     : 4,/* Disable Nagle algorithm?             */
228                 thin_lto    : 1,/* Use linear timeouts for thin streams */
229                 unused1     : 1,
230                 repair      : 1,
231                 frto        : 1;/* F-RTO (RFC5682) activated in CA_Loss */
232         u8      repair_queue;
233         u8      syn_data:1,     /* SYN includes data */
234                 syn_fastopen:1, /* SYN includes Fast Open option */
235                 syn_fastopen_exp:1,/* SYN includes Fast Open exp. option */
236                 syn_fastopen_ch:1, /* Active TFO re-enabling probe */
237                 syn_data_acked:1,/* data in SYN is acked by SYN-ACK */
238                 save_syn:1,     /* Save headers of SYN packet */
239                 is_cwnd_limited:1;/* forward progress limited by snd_cwnd? */
240         u32     tlp_high_seq;   /* snd_nxt at the time of TLP retransmit. */
241 
242 /* RTT measurement */
243         struct skb_mstamp tcp_mstamp; /* most recent packet received/sent */
244         u32     srtt_us;        /* smoothed round trip time << 3 in usecs */
245         u32     mdev_us;        /* medium deviation                     */
246         u32     mdev_max_us;    /* maximal mdev for the last rtt period */
247         u32     rttvar_us;      /* smoothed mdev_max                    */
248         u32     rtt_seq;        /* sequence number to update rttvar     */
249         struct  minmax rtt_min;
250 
251         u32     packets_out;    /* Packets which are "in flight"        */
252         u32     retrans_out;    /* Retransmitted packets out            */
253         u32     max_packets_out;  /* max packets_out in last window */
254         u32     max_packets_seq;  /* right edge of max_packets_out flight */
255 
256         u16     urg_data;       /* Saved octet of OOB data and control flags */
257         u8      ecn_flags;      /* ECN status bits.                     */
258         u8      keepalive_probes; /* num of allowed keep alive probes   */
259         u32     reordering;     /* Packet reordering metric.            */
260         u32     snd_up;         /* Urgent pointer               */
261 
262 /*
263  *      Options received (usually on last packet, some only on SYN packets).
264  */
265         struct tcp_options_received rx_opt;
266 
267 /*
268  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
269  */
270         u32     snd_ssthresh;   /* Slow start size threshold            */
271         u32     snd_cwnd;       /* Sending congestion window            */
272         u32     snd_cwnd_cnt;   /* Linear increase counter              */
273         u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
274         u32     snd_cwnd_used;
275         u32     snd_cwnd_stamp;
276         u32     prior_cwnd;     /* Congestion window at start of Recovery. */
277         u32     prr_delivered;  /* Number of newly delivered packets to
278                                  * receiver in Recovery. */
279         u32     prr_out;        /* Total number of pkts sent during Recovery. */
280         u32     delivered;      /* Total data packets delivered incl. rexmits */
281         u32     lost;           /* Total data packets lost incl. rexmits */
282         u32     app_limited;    /* limited until "delivered" reaches this val */
283         struct skb_mstamp first_tx_mstamp;  /* start of window send phase */
284         struct skb_mstamp delivered_mstamp; /* time we reached "delivered" */
285         u32     rate_delivered;    /* saved rate sample: packets delivered */
286         u32     rate_interval_us;  /* saved rate sample: time elapsed */
287 
288         u32     rcv_wnd;        /* Current receiver window              */
289         u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
290         u32     notsent_lowat;  /* TCP_NOTSENT_LOWAT */
291         u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
292         u32     lost_out;       /* Lost packets                 */
293         u32     sacked_out;     /* SACK'd packets                       */
294         u32     fackets_out;    /* FACK'd packets                       */
295 
296         /* from STCP, retrans queue hinting */
297         struct sk_buff* lost_skb_hint;
298         struct sk_buff *retransmit_skb_hint;
299 
300         /* OOO segments go in this rbtree. Socket lock must be held. */
301         struct rb_root  out_of_order_queue;
302         struct sk_buff  *ooo_last_skb; /* cache rb_last(out_of_order_queue) */
303 
304         /* SACKs data, these 2 need to be together (see tcp_options_write) */
305         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
306         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
307 
308         struct tcp_sack_block recv_sack_cache[4];
309 
310         struct sk_buff *highest_sack;   /* skb just after the highest
311                                          * skb with SACKed bit set
312                                          * (validity guaranteed only if
313                                          * sacked_out > 0)
314                                          */
315 
316         int     lost_cnt_hint;
317 
318         u32     prior_ssthresh; /* ssthresh saved at recovery start     */
319         u32     high_seq;       /* snd_nxt at onset of congestion       */
320 
321         u32     retrans_stamp;  /* Timestamp of the last retransmit,
322                                  * also used in SYN-SENT to remember stamp of
323                                  * the first SYN. */
324         u32     undo_marker;    /* snd_una upon a new recovery episode. */
325         int     undo_retrans;   /* number of undoable retransmissions. */
326         u32     total_retrans;  /* Total retransmits for entire connection */
327 
328         u32     urg_seq;        /* Seq of received urgent pointer */
329         unsigned int            keepalive_time;   /* time before keep alive takes place */
330         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
331 
332         int                     linger2;
333 
334 /* Receiver side RTT estimation */
335         struct {
336                 u32             rtt_us;
337                 u32             seq;
338                 struct skb_mstamp time;
339         } rcv_rtt_est;
340 
341 /* Receiver queue space */
342         struct {
343                 int             space;
344                 u32             seq;
345                 struct skb_mstamp time;
346         } rcvq_space;
347 
348 /* TCP-specific MTU probe information. */
349         struct {
350                 u32               probe_seq_start;
351                 u32               probe_seq_end;
352         } mtu_probe;
353         u32     mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
354                            * while socket was owned by user.
355                            */
356 
357 #ifdef CONFIG_TCP_MD5SIG
358 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
359         const struct tcp_sock_af_ops    *af_specific;
360 
361 /* TCP MD5 Signature Option information */
362         struct tcp_md5sig_info  __rcu *md5sig_info;
363 #endif
364 
365 /* TCP fastopen related information */
366         struct tcp_fastopen_request *fastopen_req;
367         /* fastopen_rsk points to request_sock that resulted in this big
368          * socket. Used to retransmit SYNACKs etc.
369          */
370         struct request_sock *fastopen_rsk;
371         u32     *saved_syn;
372 };
373 
374 enum tsq_enum {
375         TSQ_THROTTLED,
376         TSQ_QUEUED,
377         TCP_TSQ_DEFERRED,          /* tcp_tasklet_func() found socket was owned */
378         TCP_WRITE_TIMER_DEFERRED,  /* tcp_write_timer() found socket was owned */
379         TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */
380         TCP_MTU_REDUCED_DEFERRED,  /* tcp_v{4|6}_err() could not call
381                                     * tcp_v{4|6}_mtu_reduced()
382                                     */
383 };
384 
385 enum tsq_flags {
386         TSQF_THROTTLED                  = (1UL << TSQ_THROTTLED),
387         TSQF_QUEUED                     = (1UL << TSQ_QUEUED),
388         TCPF_TSQ_DEFERRED               = (1UL << TCP_TSQ_DEFERRED),
389         TCPF_WRITE_TIMER_DEFERRED       = (1UL << TCP_WRITE_TIMER_DEFERRED),
390         TCPF_DELACK_TIMER_DEFERRED      = (1UL << TCP_DELACK_TIMER_DEFERRED),
391         TCPF_MTU_REDUCED_DEFERRED       = (1UL << TCP_MTU_REDUCED_DEFERRED),
392 };
393 
394 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
395 {
396         return (struct tcp_sock *)sk;
397 }
398 
399 struct tcp_timewait_sock {
400         struct inet_timewait_sock tw_sk;
401 #define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt
402 #define tw_snd_nxt tw_sk.__tw_common.skc_tw_snd_nxt
403         u32                       tw_rcv_wnd;
404         u32                       tw_ts_offset;
405         u32                       tw_ts_recent;
406 
407         /* The time we sent the last out-of-window ACK: */
408         u32                       tw_last_oow_ack_time;
409 
410         long                      tw_ts_recent_stamp;
411 #ifdef CONFIG_TCP_MD5SIG
412         struct tcp_md5sig_key     *tw_md5_key;
413 #endif
414 };
415 
416 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
417 {
418         return (struct tcp_timewait_sock *)sk;
419 }
420 
421 static inline bool tcp_passive_fastopen(const struct sock *sk)
422 {
423         return (sk->sk_state == TCP_SYN_RECV &&
424                 tcp_sk(sk)->fastopen_rsk != NULL);
425 }
426 
427 static inline void fastopen_queue_tune(struct sock *sk, int backlog)
428 {
429         struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
430         int somaxconn = READ_ONCE(sock_net(sk)->core.sysctl_somaxconn);
431 
432         queue->fastopenq.max_qlen = min_t(unsigned int, backlog, somaxconn);
433 }
434 
435 static inline void tcp_move_syn(struct tcp_sock *tp,
436                                 struct request_sock *req)
437 {
438         tp->saved_syn = req->saved_syn;
439         req->saved_syn = NULL;
440 }
441 
442 static inline void tcp_saved_syn_free(struct tcp_sock *tp)
443 {
444         kfree(tp->saved_syn);
445         tp->saved_syn = NULL;
446 }
447 
448 struct sk_buff *tcp_get_timestamping_opt_stats(const struct sock *sk);
449 
450 static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
451 {
452         /* We use READ_ONCE() here because socket might not be locked.
453          * This happens for listeners.
454          */
455         u16 user_mss = READ_ONCE(tp->rx_opt.user_mss);
456 
457         return (user_mss && user_mss < mss) ? user_mss : mss;
458 }
459 #endif  /* _LINUX_TCP_H */
460 

~ [ 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