string: factorize skip_spaces and export it to be generally available
[safe/jmp/linux-2.6] / include / linux / tcp.h
index f2bb239..7fee8a4 100644 (file)
 
 #include <linux/types.h>
 #include <asm/byteorder.h>
+#include <linux/socket.h>
 
 struct tcphdr {
-       __u16   source;
-       __u16   dest;
-       __u32   seq;
-       __u32   ack_seq;
+       __be16  source;
+       __be16  dest;
+       __be32  seq;
+       __be32  ack_seq;
 #if defined(__LITTLE_ENDIAN_BITFIELD)
        __u16   res1:4,
                doff:4,
@@ -50,9 +51,9 @@ struct tcphdr {
 #else
 #error "Adjust your <asm/byteorder.h> defines"
 #endif 
-       __u16   window;
-       __u16   check;
-       __u16   urg_ptr;
+       __be16  window;
+       __sum16 check;
+       __be16  urg_ptr;
 };
 
 /*
@@ -62,24 +63,30 @@ struct tcphdr {
  */
 union tcp_word_hdr { 
        struct tcphdr hdr;
-       __u32             words[5];
+       __be32            words[5];
 }; 
 
 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3]) 
 
 enum { 
-       TCP_FLAG_CWR = __constant_htonl(0x00800000), 
-       TCP_FLAG_ECE = __constant_htonl(0x00400000), 
-       TCP_FLAG_URG = __constant_htonl(0x00200000), 
-       TCP_FLAG_ACK = __constant_htonl(0x00100000), 
-       TCP_FLAG_PSH = __constant_htonl(0x00080000), 
-       TCP_FLAG_RST = __constant_htonl(0x00040000), 
-       TCP_FLAG_SYN = __constant_htonl(0x00020000), 
-       TCP_FLAG_FIN = __constant_htonl(0x00010000),
-       TCP_RESERVED_BITS = __constant_htonl(0x0F000000),
-       TCP_DATA_OFFSET = __constant_htonl(0xF0000000)
+       TCP_FLAG_CWR = __cpu_to_be32(0x00800000),
+       TCP_FLAG_ECE = __cpu_to_be32(0x00400000),
+       TCP_FLAG_URG = __cpu_to_be32(0x00200000),
+       TCP_FLAG_ACK = __cpu_to_be32(0x00100000),
+       TCP_FLAG_PSH = __cpu_to_be32(0x00080000),
+       TCP_FLAG_RST = __cpu_to_be32(0x00040000),
+       TCP_FLAG_SYN = __cpu_to_be32(0x00020000),
+       TCP_FLAG_FIN = __cpu_to_be32(0x00010000),
+       TCP_RESERVED_BITS = __cpu_to_be32(0x0F000000),
+       TCP_DATA_OFFSET = __cpu_to_be32(0xF0000000)
 }; 
 
+/*
+ * TCP general constants
+ */
+#define TCP_MSS_DEFAULT                 536U   /* IPv4 (RFC1122, RFC2581) */
+#define TCP_MSS_DESIRED                1220U   /* IPv6 (tunneled), EDNS0 (RFC3226) */
+
 /* TCP socket options */
 #define TCP_NODELAY            1       /* Turn off Nagle's algorithm. */
 #define TCP_MAXSEG             2       /* Limit MSS */
@@ -94,14 +101,16 @@ enum {
 #define TCP_INFO               11      /* Information about this connection. */
 #define TCP_QUICKACK           12      /* Block/reenable quick acks */
 #define TCP_CONGESTION         13      /* Congestion control algorithm */
+#define TCP_MD5SIG             14      /* TCP MD5 Signature (RFC2385) */
+#define TCP_COOKIE_TRANSACTIONS        15      /* TCP Cookie Transactions */
 
+/* for TCP_INFO socket option */
 #define TCPI_OPT_TIMESTAMPS    1
 #define TCPI_OPT_SACK          2
 #define TCPI_OPT_WSCALE                4
 #define TCPI_OPT_ECN           8
 
-enum tcp_ca_state
-{
+enum tcp_ca_state {
        TCP_CA_Open = 0,
 #define TCPF_CA_Open   (1<<TCP_CA_Open)
        TCP_CA_Disorder = 1,
@@ -114,8 +123,7 @@ enum tcp_ca_state
 #define TCPF_CA_Loss   (1<<TCP_CA_Loss)
 };
 
-struct tcp_info
-{
+struct tcp_info {
        __u8    tcpi_state;
        __u8    tcpi_ca_state;
        __u8    tcpi_retransmits;
@@ -157,44 +165,120 @@ struct tcp_info
        __u32   tcpi_total_retrans;
 };
 
+/* for TCP_MD5SIG socket option */
+#define TCP_MD5SIG_MAXKEYLEN   80
+
+struct tcp_md5sig {
+       struct __kernel_sockaddr_storage tcpm_addr;     /* address associated */
+       __u16   __tcpm_pad1;                            /* zero */
+       __u16   tcpm_keylen;                            /* key length */
+       __u32   __tcpm_pad2;                            /* zero */
+       __u8    tcpm_key[TCP_MD5SIG_MAXKEYLEN];         /* key (binary) */
+};
+
+/* for TCP_COOKIE_TRANSACTIONS (TCPCT) socket option */
+#define TCP_COOKIE_MIN          8              /*  64-bits */
+#define TCP_COOKIE_MAX         16              /* 128-bits */
+#define TCP_COOKIE_PAIR_SIZE   (2*TCP_COOKIE_MAX)
+
+/* Flags for both getsockopt and setsockopt */
+#define TCP_COOKIE_IN_ALWAYS   (1 << 0)        /* Discard SYN without cookie */
+#define TCP_COOKIE_OUT_NEVER   (1 << 1)        /* Prohibit outgoing cookies,
+                                                * supercedes everything. */
+
+/* Flags for getsockopt */
+#define TCP_S_DATA_IN          (1 << 2)        /* Was data received? */
+#define TCP_S_DATA_OUT         (1 << 3)        /* Was data sent? */
+
+/* TCP_COOKIE_TRANSACTIONS data */
+struct tcp_cookie_transactions {
+       __u16   tcpct_flags;                    /* see above */
+       __u8    __tcpct_pad1;                   /* zero */
+       __u8    tcpct_cookie_desired;           /* bytes */
+       __u16   tcpct_s_data_desired;           /* bytes of variable data */
+       __u16   tcpct_used;                     /* bytes in value */
+       __u8    tcpct_value[TCP_MSS_DEFAULT];
+};
+
 #ifdef __KERNEL__
 
-#include <linux/config.h>
 #include <linux/skbuff.h>
+#include <linux/dmaengine.h>
 #include <net/sock.h>
 #include <net/inet_connection_sock.h>
 #include <net/inet_timewait_sock.h>
 
+static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
+{
+       return (struct tcphdr *)skb_transport_header(skb);
+}
+
+static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
+{
+       return tcp_hdr(skb)->doff * 4;
+}
+
+static inline unsigned int tcp_optlen(const struct sk_buff *skb)
+{
+       return (tcp_hdr(skb)->doff - 5) * 4;
+}
+
 /* This defines a selective acknowledgement block. */
+struct tcp_sack_block_wire {
+       __be32  start_seq;
+       __be32  end_seq;
+};
+
 struct tcp_sack_block {
-       __u32   start_seq;
-       __u32   end_seq;
+       u32     start_seq;
+       u32     end_seq;
 };
 
 struct tcp_options_received {
 /*     PAWS/RTTM data  */
        long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
-       __u32   ts_recent;      /* Time stamp to echo next              */
-       __u32   rcv_tsval;      /* Time stamp value                     */
-       __u32   rcv_tsecr;      /* Time stamp echo reply                */
-       __u16   saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
+       u32     ts_recent;      /* Time stamp to echo next              */
+       u32     rcv_tsval;      /* Time stamp value                     */
+       u32     rcv_tsecr;      /* Time stamp echo reply                */
+       u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
                tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
                dsack : 1,      /* D-SACK is scheduled                  */
                wscale_ok : 1,  /* Wscale seen on SYN packet            */
                sack_ok : 4,    /* SACK seen on SYN packet              */
                snd_wscale : 4, /* Window scaling received from sender  */
                rcv_wscale : 4; /* Window scaling to send to receiver   */
-/*     SACKs data      */
-       __u8    eff_sacks;      /* Size of SACK array to send with next packet */
-       __u8    num_sacks;      /* Number of SACK blocks                */
-       __u16   user_mss;       /* mss requested by user in ioctl */
-       __u16   mss_clamp;      /* Maximal mss, negotiated at connection setup */
+       u8      cookie_plus:6,  /* bytes in authenticator/cookie option */
+               cookie_out_never:1,
+               cookie_in_always:1;
+       u8      num_sacks;      /* Number of SACK blocks                */
+       u16     user_mss;       /* mss requested by user in ioctl       */
+       u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
 };
 
+static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
+{
+       rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
+       rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
+       rx_opt->cookie_plus = 0;
+}
+
+/* This is the max number of SACKS that we'll generate and process. It's safe
+ * to increase this, although since:
+ *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
+ * only four options will fit in a standard TCP header */
+#define TCP_NUM_SACKS 4
+
+struct tcp_cookie_values;
+struct tcp_request_sock_ops;
+
 struct tcp_request_sock {
-       struct inet_request_sock req;
-       __u32                    rcv_isn;
-       __u32                    snt_isn;
+       struct inet_request_sock        req;
+#ifdef CONFIG_TCP_MD5SIG
+       /* Only used by TCP MD5 Signature so far. */
+       const struct tcp_request_sock_ops *af_specific;
+#endif
+       u32                             rcv_isn;
+       u32                             snt_isn;
 };
 
 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
@@ -205,26 +289,29 @@ static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 struct tcp_sock {
        /* inet_connection_sock has to be the first member of tcp_sock */
        struct inet_connection_sock     inet_conn;
-       int     tcp_header_len; /* Bytes of tcp header to send          */
+       u16     tcp_header_len; /* Bytes of tcp header to send          */
+       u16     xmit_size_goal_segs; /* Goal for segmenting output packets */
 
 /*
  *     Header prediction flags
  *     0x5?10 << 16 + snd_wnd in net byte order
  */
-       __u32   pred_flags;
+       __be32  pred_flags;
 
 /*
  *     RFC793 variables by their proper names. This means you can
  *     read the code and the spec side by side (and laugh ...)
  *     See RFC793 and RFC1122. The RFC writes these in capitals.
  */
-       __u32   rcv_nxt;        /* What we want to receive next         */
-       __u32   snd_nxt;        /* Next sequence we send                */
+       u32     rcv_nxt;        /* What we want to receive next         */
+       u32     copied_seq;     /* Head of yet unread data              */
+       u32     rcv_wup;        /* rcv_nxt on last window update sent   */
+       u32     snd_nxt;        /* Next sequence we send                */
 
-       __u32   snd_una;        /* First byte we want an ack for        */
-       __u32   snd_sml;        /* Last byte of the most recently transmitted small packet */
-       __u32   rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
-       __u32   lsndtime;       /* timestamp of last sent data packet (for restart window) */
+       u32     snd_una;        /* First byte we want an ack for        */
+       u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
+       u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
+       u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 
        /* Data for direct copy to user */
        struct {
@@ -233,34 +320,44 @@ struct tcp_sock {
                struct iovec            *iov;
                int                     memory;
                int                     len;
+#ifdef CONFIG_NET_DMA
+               /* members for async copy */
+               struct dma_chan         *dma_chan;
+               int                     wakeup;
+               struct dma_pinned_list  *pinned_list;
+               dma_cookie_t            dma_cookie;
+#endif
        } ucopy;
 
-       __u32   snd_wl1;        /* Sequence for window update           */
-       __u32   snd_wnd;        /* The window we expect to receive      */
-       __u32   max_window;     /* Maximal window ever seen from peer   */
-       __u32   mss_cache;      /* Cached effective mss, not including SACKS */
-       __u16   xmit_size_goal; /* Goal for segmenting output packets   */
-       /* XXX Two bytes hole, try to pack */
+       u32     snd_wl1;        /* Sequence for window update           */
+       u32     snd_wnd;        /* The window we expect to receive      */
+       u32     max_window;     /* Maximal window ever seen from peer   */
+       u32     mss_cache;      /* Cached effective mss, not including SACKS */
 
-       __u32   window_clamp;   /* Maximal window to advertise          */
-       __u32   rcv_ssthresh;   /* Current window clamp                 */
+       u32     window_clamp;   /* Maximal window to advertise          */
+       u32     rcv_ssthresh;   /* Current window clamp                 */
 
-       __u32   frto_highmark;  /* snd_nxt when RTO occurred */
-       __u8    reordering;     /* Packet reordering metric.            */
-       __u8    frto_counter;   /* Number of new acks after RTO */
-       __u8    nonagle;        /* Disable Nagle algorithm?             */
-       __u8    keepalive_probes; /* num of allowed keep alive probes   */
+       u32     frto_highmark;  /* snd_nxt when RTO occurred */
+       u16     advmss;         /* Advertised MSS                       */
+       u8      frto_counter;   /* Number of new acks after RTO */
+       u8      nonagle;        /* Disable Nagle algorithm?             */
 
 /* RTT measurement */
-       __u32   srtt;           /* smoothed round trip time << 3        */
-       __u32   mdev;           /* medium deviation                     */
-       __u32   mdev_max;       /* maximal mdev for the last rtt period */
-       __u32   rttvar;         /* smoothed mdev_max                    */
-       __u32   rtt_seq;        /* sequence number to update rttvar     */
-
-       __u32   packets_out;    /* Packets which are "in flight"        */
-       __u32   left_out;       /* Packets which leaved network */
-       __u32   retrans_out;    /* Retransmitted packets out            */
+       u32     srtt;           /* smoothed round trip time << 3        */
+       u32     mdev;           /* medium deviation                     */
+       u32     mdev_max;       /* maximal mdev for the last rtt period */
+       u32     rttvar;         /* smoothed mdev_max                    */
+       u32     rtt_seq;        /* sequence number to update rttvar     */
+
+       u32     packets_out;    /* Packets which are "in flight"        */
+       u32     retrans_out;    /* Retransmitted packets out            */
+
+       u16     urg_data;       /* Saved octet of OOB data and control flags */
+       u8      ecn_flags;      /* ECN status bits.                     */
+       u8      reordering;     /* Packet reordering metric.            */
+       u32     snd_up;         /* Urgent pointer               */
+
+       u8      keepalive_probes; /* num of allowed keep alive probes   */
 /*
  *      Options received (usually on last packet, some only on SYN packets).
  */
@@ -269,80 +366,94 @@ struct tcp_sock {
 /*
  *     Slow start and congestion control (see also Nagle, and Karn & Partridge)
  */
-       __u32   snd_ssthresh;   /* Slow start size threshold            */
-       __u32   snd_cwnd;       /* Sending congestion window            */
-       __u16   snd_cwnd_cnt;   /* Linear increase counter              */
-       __u16   snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
-       __u32   snd_cwnd_used;
-       __u32   snd_cwnd_stamp;
+       u32     snd_ssthresh;   /* Slow start size threshold            */
+       u32     snd_cwnd;       /* Sending congestion window            */
+       u32     snd_cwnd_cnt;   /* Linear increase counter              */
+       u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
+       u32     snd_cwnd_used;
+       u32     snd_cwnd_stamp;
+
+       u32     rcv_wnd;        /* Current receiver window              */
+       u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
+       u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
+       u32     lost_out;       /* Lost packets                 */
+       u32     sacked_out;     /* SACK'd packets                       */
+       u32     fackets_out;    /* FACK'd packets                       */
+       u32     tso_deferred;
+       u32     bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
 
-       struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
+       /* from STCP, retrans queue hinting */
+       struct sk_buff* lost_skb_hint;
+       struct sk_buff *scoreboard_skb_hint;
+       struct sk_buff *retransmit_skb_hint;
 
-       __u32   rcv_wnd;        /* Current receiver window              */
-       __u32   rcv_wup;        /* rcv_nxt on last window update sent   */
-       __u32   write_seq;      /* Tail(+1) of data held in tcp send buffer */
-       __u32   pushed_seq;     /* Last pushed seq, required to talk to windows */
-       __u32   copied_seq;     /* Head of yet unread data              */
+       struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 
-/*     SACKs data      */
+       /* SACKs data, these 2 need to be together (see tcp_build_and_update_options) */
        struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
        struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 
        struct tcp_sack_block recv_sack_cache[4];
 
-       /* from STCP, retrans queue hinting */
-       struct sk_buff* lost_skb_hint;
+       struct sk_buff *highest_sack;   /* highest skb with SACK received
+                                        * (validity guaranteed only if
+                                        * sacked_out > 0)
+                                        */
 
-       struct sk_buff *scoreboard_skb_hint;
-       struct sk_buff *retransmit_skb_hint;
-       struct sk_buff *forward_skb_hint;
-       struct sk_buff *fastpath_skb_hint;
-
-       int     fastpath_cnt_hint;
        int     lost_cnt_hint;
-       int     retransmit_cnt_hint;
-       int     forward_cnt_hint;
+       u32     retransmit_high;        /* L-bits may be on up to this seqno */
+
+       u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 
-       __u16   advmss;         /* Advertised MSS                       */
-       __u16   prior_ssthresh; /* ssthresh saved at recovery start     */
-       __u32   lost_out;       /* Lost packets                 */
-       __u32   sacked_out;     /* SACK'd packets                       */
-       __u32   fackets_out;    /* FACK'd packets                       */
-       __u32   high_seq;       /* snd_nxt at onset of congestion       */
+       u32     prior_ssthresh; /* ssthresh saved at recovery start     */
+       u32     high_seq;       /* snd_nxt at onset of congestion       */
 
-       __u32   retrans_stamp;  /* Timestamp of the last retransmit,
+       u32     retrans_stamp;  /* Timestamp of the last retransmit,
                                 * also used in SYN-SENT to remember stamp of
                                 * the first SYN. */
-       __u32   undo_marker;    /* tracking retrans started here. */
+       u32     undo_marker;    /* tracking retrans started here. */
        int     undo_retrans;   /* number of undoable retransmissions. */
-       __u32   urg_seq;        /* Seq of received urgent pointer */
-       __u16   urg_data;       /* Saved octet of OOB data and control flags */
-       __u8    urg_mode;       /* In urgent mode               */
-       __u8    ecn_flags;      /* ECN status bits.                     */
-       __u32   snd_up;         /* Urgent pointer               */
-
-       __u32   total_retrans;  /* Total retransmits for entire connection */
-       __u32   bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
+       u32     total_retrans;  /* Total retransmits for entire connection */
 
+       u32     urg_seq;        /* Seq of received urgent pointer */
        unsigned int            keepalive_time;   /* time before keep alive takes place */
        unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
-       int                     linger2;
 
-       unsigned long last_synq_overflow; 
+       int                     linger2;
 
 /* Receiver side RTT estimation */
        struct {
-               __u32   rtt;
-               __u32   seq;
-               __u32   time;
+               u32     rtt;
+               u32     seq;
+               u32     time;
        } rcv_rtt_est;
 
 /* Receiver queue space */
        struct {
                int     space;
-               __u32   seq;
-               __u32   time;
+               u32     seq;
+               u32     time;
        } rcvq_space;
+
+/* TCP-specific MTU probe information. */
+       struct {
+               u32               probe_seq_start;
+               u32               probe_seq_end;
+       } mtu_probe;
+
+#ifdef CONFIG_TCP_MD5SIG
+/* TCP AF-Specific parts; only used by MD5 Signature support so far */
+       const struct tcp_sock_af_ops    *af_specific;
+
+/* TCP MD5 Signature Option information */
+       struct tcp_md5sig_info  *md5sig_info;
+#endif
+
+       /* When the cookie options are generated and exchanged, then this
+        * object holds a reference to them (cookie_values->kref).  Also
+        * contains related tcp_cookie_transactions fields.
+        */
+       struct tcp_cookie_values  *cookie_values;
 };
 
 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
@@ -352,11 +463,19 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 
 struct tcp_timewait_sock {
        struct inet_timewait_sock tw_sk;
-       __u32                     tw_rcv_nxt;
-       __u32                     tw_snd_nxt;
-       __u32                     tw_rcv_wnd;
-       __u32                     tw_ts_recent;
+       u32                       tw_rcv_nxt;
+       u32                       tw_snd_nxt;
+       u32                       tw_rcv_wnd;
+       u32                       tw_ts_recent;
        long                      tw_ts_recent_stamp;
+#ifdef CONFIG_TCP_MD5SIG
+       u16                       tw_md5_keylen;
+       u8                        tw_md5_key[TCP_MD5SIG_MAXKEYLEN];
+#endif
+       /* Few sockets in timewait have cookies; in that case, then this
+        * object holds a reference to them (tw_cookie_values->kref).
+        */
+       struct tcp_cookie_values  *tw_cookie_values;
 };
 
 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
@@ -364,6 +483,6 @@ static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
        return (struct tcp_timewait_sock *)sk;
 }
 
-#endif
+#endif /* __KERNEL__ */
 
 #endif /* _LINUX_TCP_H */