Back to home page

OSCL-LXR

 
 

    


0001 /* SPDX-License-Identifier: GPL-2.0-or-later */
0002 /*
0003  * INET     An implementation of the TCP/IP protocol suite for the LINUX
0004  *      operating system.  INET is implemented using the  BSD Socket
0005  *      interface as the means of communication with the user level.
0006  *
0007  *      Definitions for inet_sock
0008  *
0009  * Authors: Many, reorganised here by
0010  *      Arnaldo Carvalho de Melo <acme@mandriva.com>
0011  */
0012 #ifndef _INET_SOCK_H
0013 #define _INET_SOCK_H
0014 
0015 #include <linux/bitops.h>
0016 #include <linux/string.h>
0017 #include <linux/types.h>
0018 #include <linux/jhash.h>
0019 #include <linux/netdevice.h>
0020 
0021 #include <net/flow.h>
0022 #include <net/sock.h>
0023 #include <net/request_sock.h>
0024 #include <net/netns/hash.h>
0025 #include <net/tcp_states.h>
0026 #include <net/l3mdev.h>
0027 
0028 /** struct ip_options - IP Options
0029  *
0030  * @faddr - Saved first hop address
0031  * @nexthop - Saved nexthop address in LSRR and SSRR
0032  * @is_strictroute - Strict source route
0033  * @srr_is_hit - Packet destination addr was our one
0034  * @is_changed - IP checksum more not valid
0035  * @rr_needaddr - Need to record addr of outgoing dev
0036  * @ts_needtime - Need to record timestamp
0037  * @ts_needaddr - Need to record addr of outgoing dev
0038  */
0039 struct ip_options {
0040     __be32      faddr;
0041     __be32      nexthop;
0042     unsigned char   optlen;
0043     unsigned char   srr;
0044     unsigned char   rr;
0045     unsigned char   ts;
0046     unsigned char   is_strictroute:1,
0047             srr_is_hit:1,
0048             is_changed:1,
0049             rr_needaddr:1,
0050             ts_needtime:1,
0051             ts_needaddr:1;
0052     unsigned char   router_alert;
0053     unsigned char   cipso;
0054     unsigned char   __pad2;
0055     unsigned char   __data[];
0056 };
0057 
0058 struct ip_options_rcu {
0059     struct rcu_head rcu;
0060     struct ip_options opt;
0061 };
0062 
0063 struct ip_options_data {
0064     struct ip_options_rcu   opt;
0065     char            data[40];
0066 };
0067 
0068 struct inet_request_sock {
0069     struct request_sock req;
0070 #define ir_loc_addr     req.__req_common.skc_rcv_saddr
0071 #define ir_rmt_addr     req.__req_common.skc_daddr
0072 #define ir_num          req.__req_common.skc_num
0073 #define ir_rmt_port     req.__req_common.skc_dport
0074 #define ir_v6_rmt_addr      req.__req_common.skc_v6_daddr
0075 #define ir_v6_loc_addr      req.__req_common.skc_v6_rcv_saddr
0076 #define ir_iif          req.__req_common.skc_bound_dev_if
0077 #define ir_cookie       req.__req_common.skc_cookie
0078 #define ireq_net        req.__req_common.skc_net
0079 #define ireq_state      req.__req_common.skc_state
0080 #define ireq_family     req.__req_common.skc_family
0081 
0082     u16         snd_wscale : 4,
0083                 rcv_wscale : 4,
0084                 tstamp_ok  : 1,
0085                 sack_ok    : 1,
0086                 wscale_ok  : 1,
0087                 ecn_ok     : 1,
0088                 acked      : 1,
0089                 no_srccheck: 1,
0090                 smc_ok     : 1;
0091     u32                     ir_mark;
0092     union {
0093         struct ip_options_rcu __rcu *ireq_opt;
0094 #if IS_ENABLED(CONFIG_IPV6)
0095         struct {
0096             struct ipv6_txoptions   *ipv6_opt;
0097             struct sk_buff      *pktopts;
0098         };
0099 #endif
0100     };
0101 };
0102 
0103 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
0104 {
0105     return (struct inet_request_sock *)sk;
0106 }
0107 
0108 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
0109 {
0110     if (!sk->sk_mark &&
0111         READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept))
0112         return skb->mark;
0113 
0114     return sk->sk_mark;
0115 }
0116 
0117 static inline int inet_request_bound_dev_if(const struct sock *sk,
0118                         struct sk_buff *skb)
0119 {
0120     int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
0121 #ifdef CONFIG_NET_L3_MASTER_DEV
0122     struct net *net = sock_net(sk);
0123 
0124     if (!bound_dev_if && READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
0125         return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
0126 #endif
0127 
0128     return bound_dev_if;
0129 }
0130 
0131 static inline int inet_sk_bound_l3mdev(const struct sock *sk)
0132 {
0133 #ifdef CONFIG_NET_L3_MASTER_DEV
0134     struct net *net = sock_net(sk);
0135 
0136     if (!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
0137         return l3mdev_master_ifindex_by_index(net,
0138                               sk->sk_bound_dev_if);
0139 #endif
0140 
0141     return 0;
0142 }
0143 
0144 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
0145                      int dif, int sdif)
0146 {
0147     if (!bound_dev_if)
0148         return !sdif || l3mdev_accept;
0149     return bound_dev_if == dif || bound_dev_if == sdif;
0150 }
0151 
0152 static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if,
0153                     int dif, int sdif)
0154 {
0155 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
0156     return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
0157                  bound_dev_if, dif, sdif);
0158 #else
0159     return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
0160 #endif
0161 }
0162 
0163 struct inet_cork {
0164     unsigned int        flags;
0165     __be32          addr;
0166     struct ip_options   *opt;
0167     unsigned int        fragsize;
0168     int         length; /* Total length of all frames */
0169     struct dst_entry    *dst;
0170     u8          tx_flags;
0171     __u8            ttl;
0172     __s16           tos;
0173     char            priority;
0174     __u16           gso_size;
0175     u64         transmit_time;
0176     u32         mark;
0177 };
0178 
0179 struct inet_cork_full {
0180     struct inet_cork    base;
0181     struct flowi        fl;
0182 };
0183 
0184 struct ip_mc_socklist;
0185 struct ipv6_pinfo;
0186 struct rtable;
0187 
0188 /** struct inet_sock - representation of INET sockets
0189  *
0190  * @sk - ancestor class
0191  * @pinet6 - pointer to IPv6 control block
0192  * @inet_daddr - Foreign IPv4 addr
0193  * @inet_rcv_saddr - Bound local IPv4 addr
0194  * @inet_dport - Destination port
0195  * @inet_num - Local port
0196  * @inet_saddr - Sending source
0197  * @uc_ttl - Unicast TTL
0198  * @inet_sport - Source port
0199  * @inet_id - ID counter for DF pkts
0200  * @tos - TOS
0201  * @mc_ttl - Multicasting TTL
0202  * @is_icsk - is this an inet_connection_sock?
0203  * @uc_index - Unicast outgoing device index
0204  * @mc_index - Multicast device index
0205  * @mc_list - Group array
0206  * @cork - info to build ip hdr on each ip frag while socket is corked
0207  */
0208 struct inet_sock {
0209     /* sk and pinet6 has to be the first two members of inet_sock */
0210     struct sock     sk;
0211 #if IS_ENABLED(CONFIG_IPV6)
0212     struct ipv6_pinfo   *pinet6;
0213 #endif
0214     /* Socket demultiplex comparisons on incoming packets. */
0215 #define inet_daddr      sk.__sk_common.skc_daddr
0216 #define inet_rcv_saddr      sk.__sk_common.skc_rcv_saddr
0217 #define inet_dport      sk.__sk_common.skc_dport
0218 #define inet_num        sk.__sk_common.skc_num
0219 
0220     __be32          inet_saddr;
0221     __s16           uc_ttl;
0222     __u16           cmsg_flags;
0223     struct ip_options_rcu __rcu *inet_opt;
0224     __be16          inet_sport;
0225     __u16           inet_id;
0226 
0227     __u8            tos;
0228     __u8            min_ttl;
0229     __u8            mc_ttl;
0230     __u8            pmtudisc;
0231     __u8            recverr:1,
0232                 is_icsk:1,
0233                 freebind:1,
0234                 hdrincl:1,
0235                 mc_loop:1,
0236                 transparent:1,
0237                 mc_all:1,
0238                 nodefrag:1;
0239     __u8            bind_address_no_port:1,
0240                 recverr_rfc4884:1,
0241                 defer_connect:1; /* Indicates that fastopen_connect is set
0242                           * and cookie exists so we defer connect
0243                           * until first data frame is written
0244                           */
0245     __u8            rcv_tos;
0246     __u8            convert_csum;
0247     int         uc_index;
0248     int         mc_index;
0249     __be32          mc_addr;
0250     struct ip_mc_socklist __rcu *mc_list;
0251     struct inet_cork_full   cork;
0252 };
0253 
0254 #define IPCORK_OPT  1   /* ip-options has been held in ipcork.opt */
0255 #define IPCORK_ALLFRAG  2   /* always fragment (for ipv6 for now) */
0256 
0257 /* cmsg flags for inet */
0258 #define IP_CMSG_PKTINFO     BIT(0)
0259 #define IP_CMSG_TTL     BIT(1)
0260 #define IP_CMSG_TOS     BIT(2)
0261 #define IP_CMSG_RECVOPTS    BIT(3)
0262 #define IP_CMSG_RETOPTS     BIT(4)
0263 #define IP_CMSG_PASSSEC     BIT(5)
0264 #define IP_CMSG_ORIGDSTADDR BIT(6)
0265 #define IP_CMSG_CHECKSUM    BIT(7)
0266 #define IP_CMSG_RECVFRAGSIZE    BIT(8)
0267 
0268 static inline bool sk_is_inet(struct sock *sk)
0269 {
0270     return sk->sk_family == AF_INET || sk->sk_family == AF_INET6;
0271 }
0272 
0273 /**
0274  * sk_to_full_sk - Access to a full socket
0275  * @sk: pointer to a socket
0276  *
0277  * SYNACK messages might be attached to request sockets.
0278  * Some places want to reach the listener in this case.
0279  */
0280 static inline struct sock *sk_to_full_sk(struct sock *sk)
0281 {
0282 #ifdef CONFIG_INET
0283     if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
0284         sk = inet_reqsk(sk)->rsk_listener;
0285 #endif
0286     return sk;
0287 }
0288 
0289 /* sk_to_full_sk() variant with a const argument */
0290 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
0291 {
0292 #ifdef CONFIG_INET
0293     if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
0294         sk = ((const struct request_sock *)sk)->rsk_listener;
0295 #endif
0296     return sk;
0297 }
0298 
0299 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
0300 {
0301     return sk_to_full_sk(skb->sk);
0302 }
0303 
0304 static inline struct inet_sock *inet_sk(const struct sock *sk)
0305 {
0306     return (struct inet_sock *)sk;
0307 }
0308 
0309 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
0310                          const struct sock *sk_from,
0311                          const int ancestor_size)
0312 {
0313     memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
0314            sk_from->sk_prot->obj_size - ancestor_size);
0315 }
0316 
0317 int inet_sk_rebuild_header(struct sock *sk);
0318 
0319 /**
0320  * inet_sk_state_load - read sk->sk_state for lockless contexts
0321  * @sk: socket pointer
0322  *
0323  * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
0324  * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
0325  */
0326 static inline int inet_sk_state_load(const struct sock *sk)
0327 {
0328     /* state change might impact lockless readers. */
0329     return smp_load_acquire(&sk->sk_state);
0330 }
0331 
0332 /**
0333  * inet_sk_state_store - update sk->sk_state
0334  * @sk: socket pointer
0335  * @newstate: new state
0336  *
0337  * Paired with inet_sk_state_load(). Should be used in contexts where
0338  * state change might impact lockless readers.
0339  */
0340 void inet_sk_state_store(struct sock *sk, int newstate);
0341 
0342 void inet_sk_set_state(struct sock *sk, int state);
0343 
0344 static inline unsigned int __inet_ehashfn(const __be32 laddr,
0345                       const __u16 lport,
0346                       const __be32 faddr,
0347                       const __be16 fport,
0348                       u32 initval)
0349 {
0350     return jhash_3words((__force __u32) laddr,
0351                 (__force __u32) faddr,
0352                 ((__u32) lport) << 16 | (__force __u32)fport,
0353                 initval);
0354 }
0355 
0356 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
0357                       struct sock *sk_listener,
0358                       bool attach_listener);
0359 
0360 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
0361 {
0362     __u8 flags = 0;
0363 
0364     if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl)
0365         flags |= FLOWI_FLAG_ANYSRC;
0366     return flags;
0367 }
0368 
0369 static inline void inet_inc_convert_csum(struct sock *sk)
0370 {
0371     inet_sk(sk)->convert_csum++;
0372 }
0373 
0374 static inline void inet_dec_convert_csum(struct sock *sk)
0375 {
0376     if (inet_sk(sk)->convert_csum > 0)
0377         inet_sk(sk)->convert_csum--;
0378 }
0379 
0380 static inline bool inet_get_convert_csum(struct sock *sk)
0381 {
0382     return !!inet_sk(sk)->convert_csum;
0383 }
0384 
0385 
0386 static inline bool inet_can_nonlocal_bind(struct net *net,
0387                       struct inet_sock *inet)
0388 {
0389     return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) ||
0390         inet->freebind || inet->transparent;
0391 }
0392 
0393 static inline bool inet_addr_valid_or_nonlocal(struct net *net,
0394                            struct inet_sock *inet,
0395                            __be32 addr,
0396                            int addr_type)
0397 {
0398     return inet_can_nonlocal_bind(net, inet) ||
0399         addr == htonl(INADDR_ANY) ||
0400         addr_type == RTN_LOCAL ||
0401         addr_type == RTN_MULTICAST ||
0402         addr_type == RTN_BROADCAST;
0403 }
0404 
0405 #endif  /* _INET_SOCK_H */