net/ipv4/tcp_fastopen.c

   1 #include <linux/err.h>
   2 #include <linux/init.h>
   3 #include <linux/kernel.h>
   4 #include <linux/list.h>
   5 #include <linux/tcp.h>
   6 #include <linux/rcupdate.h>
   7 #include <linux/rculist.h>
   8 #include <net/inetpeer.h>
   9 #include <net/tcp.h>
  10
  11 int sysctl_tcp_fastopen __read_mostly = TFO_CLIENT_ENABLE;
  12
  13 struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
  14
  15 static DEFINE_SPINLOCK(tcp_fastopen_ctx_lock);
  16
  17 void tcp_fastopen_init_key_once(bool publish)
  18 {
  19         static u8 key[TCP_FASTOPEN_KEY_LENGTH];
  20
  21         /* tcp_fastopen_reset_cipher publishes the new context
  22          * atomically, so we allow this race happening here.
  23          *
  24          * All call sites of tcp_fastopen_cookie_gen also check
  25          * for a valid cookie, so this is an acceptable risk.
  26          */
  27         if (net_get_random_once(key, sizeof(key)) && publish)
  28                 tcp_fastopen_reset_cipher(key, sizeof(key));
  29 }
  30
  31 static void tcp_fastopen_ctx_free(struct rcu_head *head)
  32 {
  33         struct tcp_fastopen_context *ctx =
  34             container_of(head, struct tcp_fastopen_context, rcu);
  35         crypto_free_cipher(ctx->tfm);
  36         kfree(ctx);
  37 }
  38
  39 int tcp_fastopen_reset_cipher(void *key, unsigned int len)
  40 {
  41         int err;
  42         struct tcp_fastopen_context *ctx, *octx;
  43
  44         ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
  45         if (!ctx)
  46                 return -ENOMEM;
  47         ctx->tfm = crypto_alloc_cipher("aes", 0, 0);
  48
  49         if (IS_ERR(ctx->tfm)) {
  50                 err = PTR_ERR(ctx->tfm);
  51 error:          kfree(ctx);
  52                 pr_err("TCP: TFO aes cipher alloc error: %d\n", err);
  53                 return err;
  54         }
  55         err = crypto_cipher_setkey(ctx->tfm, key, len);
  56         if (err) {
  57                 pr_err("TCP: TFO cipher key error: %d\n", err);
  58                 crypto_free_cipher(ctx->tfm);
  59                 goto error;
  60         }
  61         memcpy(ctx->key, key, len);
  62
  63         spin_lock(&tcp_fastopen_ctx_lock);
  64
  65         octx = rcu_dereference_protected(tcp_fastopen_ctx,
  66                                 lockdep_is_held(&tcp_fastopen_ctx_lock));
  67         rcu_assign_pointer(tcp_fastopen_ctx, ctx);
  68         spin_unlock(&tcp_fastopen_ctx_lock);
  69
  70         if (octx)
  71                 call_rcu(&octx->rcu, tcp_fastopen_ctx_free);
  72         return err;
  73 }
  74
  75 static bool __tcp_fastopen_cookie_gen(const void *path,
  76                                       struct tcp_fastopen_cookie *foc)
  77 {
  78         struct tcp_fastopen_context *ctx;
  79         bool ok = false;
  80
  81         tcp_fastopen_init_key_once(true);
  82
  83         rcu_read_lock();
  84         ctx = rcu_dereference(tcp_fastopen_ctx);
  85         if (ctx) {
  86                 crypto_cipher_encrypt_one(ctx->tfm, foc->val, path);
  87                 foc->len = TCP_FASTOPEN_COOKIE_SIZE;
  88                 ok = true;
  89         }
  90         rcu_read_unlock();
  91         return ok;
  92 }
  93
  94 /* Generate the fastopen cookie by doing aes128 encryption on both
  95  * the source and destination addresses. Pad 0s for IPv4 or IPv4-mapped-IPv6
  96  * addresses. For the longer IPv6 addresses use CBC-MAC.
  97  *
  98  * XXX (TFO) - refactor when TCP_FASTOPEN_COOKIE_SIZE != AES_BLOCK_SIZE.
  99  */
 100 static bool tcp_fastopen_cookie_gen(struct request_sock *req,
 101                                     struct sk_buff *syn,
 102                                     struct tcp_fastopen_cookie *foc)
 103 {
 104         if (req->rsk_ops->family == AF_INET) {
 105                 const struct iphdr *iph = ip_hdr(syn);
 106
 107                 __be32 path[4] = { iph->saddr, iph->daddr, 0, 0 };
 108                 return __tcp_fastopen_cookie_gen(path, foc);
 109         }
 110
 111 #if IS_ENABLED(CONFIG_IPV6)
 112         if (req->rsk_ops->family == AF_INET6) {
 113                 const struct ipv6hdr *ip6h = ipv6_hdr(syn);
 114                 struct tcp_fastopen_cookie tmp;
 115
 116                 if (__tcp_fastopen_cookie_gen(&ip6h->saddr, &tmp)) {
 117                         struct in6_addr *buf = (struct in6_addr *) tmp.val;
 118                         int i;
 119
 120                         for (i = 0; i < 4; i++)
 121                                 buf->s6_addr32[i] ^= ip6h->daddr.s6_addr32[i];
 122                         return __tcp_fastopen_cookie_gen(buf, foc);
 123                 }
 124         }
 125 #endif
 126         return false;
 127 }
 128
 129 static bool tcp_fastopen_create_child(struct sock *sk,
 130                                       struct sk_buff *skb,
 131                                       struct dst_entry *dst,
 132                                       struct request_sock *req)
 133 {
 134         struct tcp_sock *tp;
 135         struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
 136         struct sock *child;
 137
 138         req->num_retrans = 0;
 139         req->num_timeout = 0;
 140         req->sk = NULL;
 141
 142         child = inet_csk(sk)->icsk_af_ops->syn_recv_sock(sk, skb, req, NULL);
 143         if (child == NULL)
 144                 return false;
 145
 146         spin_lock(&queue->fastopenq->lock);
 147         queue->fastopenq->qlen++;
 148         spin_unlock(&queue->fastopenq->lock);
 149
 150         /* Initialize the child socket. Have to fix some values to take
 151          * into account the child is a Fast Open socket and is created
 152          * only out of the bits carried in the SYN packet.
 153          */
 154         tp = tcp_sk(child);
 155
 156         tp->fastopen_rsk = req;
 157         /* Do a hold on the listner sk so that if the listener is being
 158          * closed, the child that has been accepted can live on and still
 159          * access listen_lock.
 160          */
 161         sock_hold(sk);
 162         tcp_rsk(req)->listener = sk;
 163
 164         /* RFC1323: The window in SYN & SYN/ACK segments is never
 165          * scaled. So correct it appropriately.
 166          */
 167         tp->snd_wnd = ntohs(tcp_hdr(skb)->window);
 168
 169         /* Activate the retrans timer so that SYNACK can be retransmitted.
 170          * The request socket is not added to the SYN table of the parent
 171          * because it's been added to the accept queue directly.
 172          */
 173         inet_csk_reset_xmit_timer(child, ICSK_TIME_RETRANS,
 174                                   TCP_TIMEOUT_INIT, TCP_RTO_MAX);
 175
 176         /* Add the child socket directly into the accept queue */
 177         inet_csk_reqsk_queue_add(sk, req, child);
 178
 179         /* Now finish processing the fastopen child socket. */
 180         inet_csk(child)->icsk_af_ops->rebuild_header(child);
 181         tcp_init_congestion_control(child);
 182         tcp_mtup_init(child);
 183         tcp_init_metrics(child);
 184         tcp_init_buffer_space(child);
 185
 186         /* Queue the data carried in the SYN packet. We need to first
 187          * bump skb's refcnt because the caller will attempt to free it.
 188          *
 189          * XXX (TFO) - we honor a zero-payload TFO request for now,
 190          * (any reason not to?) but no need to queue the skb since
 191          * there is no data. How about SYN+FIN?
 192          */
 193         if (TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq + 1) {
 194                 skb = skb_get(skb);
 195                 skb_dst_drop(skb);
 196                 __skb_pull(skb, tcp_hdr(skb)->doff * 4);
 197                 skb_set_owner_r(skb, child);
 198                 __skb_queue_tail(&child->sk_receive_queue, skb);
 199                 tp->syn_data_acked = 1;
 200         }
 201         tcp_rsk(req)->rcv_nxt = tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
 202         sk->sk_data_ready(sk);
 203         bh_unlock_sock(child);
 204         sock_put(child);
 205         WARN_ON(req->sk == NULL);
 206         return true;
 207 }
 208 EXPORT_SYMBOL(tcp_fastopen_create_child);
 209
 210 static bool tcp_fastopen_queue_check(struct sock *sk)
 211 {
 212         struct fastopen_queue *fastopenq;
 213
 214         /* Make sure the listener has enabled fastopen, and we don't
 215          * exceed the max # of pending TFO requests allowed before trying
 216          * to validating the cookie in order to avoid burning CPU cycles
 217          * unnecessarily.
 218          *
 219          * XXX (TFO) - The implication of checking the max_qlen before
 220          * processing a cookie request is that clients can't differentiate
 221          * between qlen overflow causing Fast Open to be disabled
 222          * temporarily vs a server not supporting Fast Open at all.
 223          */
 224         fastopenq = inet_csk(sk)->icsk_accept_queue.fastopenq;
 225         if (fastopenq == NULL || fastopenq->max_qlen == 0)
 226                 return false;
 227
 228         if (fastopenq->qlen >= fastopenq->max_qlen) {
 229                 struct request_sock *req1;
 230                 spin_lock(&fastopenq->lock);
 231                 req1 = fastopenq->rskq_rst_head;
 232                 if ((req1 == NULL) || time_after(req1->expires, jiffies)) {
 233                         spin_unlock(&fastopenq->lock);
 234                         NET_INC_STATS_BH(sock_net(sk),
 235                                          LINUX_MIB_TCPFASTOPENLISTENOVERFLOW);
 236                         return false;
 237                 }
 238                 fastopenq->rskq_rst_head = req1->dl_next;
 239                 fastopenq->qlen--;
 240                 spin_unlock(&fastopenq->lock);
 241                 reqsk_free(req1);
 242         }
 243         return true;
 244 }
 245
 246 /* Returns true if we should perform Fast Open on the SYN. The cookie (foc)
 247  * may be updated and return the client in the SYN-ACK later. E.g., Fast Open
 248  * cookie request (foc->len == 0).
 249  */
 250 bool tcp_try_fastopen(struct sock *sk, struct sk_buff *skb,
 251                       struct request_sock *req,
 252                       struct tcp_fastopen_cookie *foc,
 253                       struct dst_entry *dst)
 254 {
 255         struct tcp_fastopen_cookie valid_foc = { .len = -1 };
 256         bool syn_data = TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq + 1;
 257
 258         if (!((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) &&
 259               (syn_data || foc->len >= 0) &&
 260               tcp_fastopen_queue_check(sk))) {
 261                 foc->len = -1;
 262                 return false;
 263         }
 264
 265         if (syn_data && (sysctl_tcp_fastopen & TFO_SERVER_COOKIE_NOT_REQD))
 266                 goto fastopen;
 267
 268         if (tcp_fastopen_cookie_gen(req, skb, &valid_foc) &&
 269             foc->len == TCP_FASTOPEN_COOKIE_SIZE &&
 270             foc->len == valid_foc.len &&
 271             !memcmp(foc->val, valid_foc.val, foc->len)) {
 272                 /* Cookie is valid. Create a (full) child socket to accept
 273                  * the data in SYN before returning a SYN-ACK to ack the
 274                  * data. If we fail to create the socket, fall back and
 275                  * ack the ISN only but includes the same cookie.
 276                  *
 277                  * Note: Data-less SYN with valid cookie is allowed to send
 278                  * data in SYN_RECV state.
 279                  */
 280 fastopen:
 281                 if (tcp_fastopen_create_child(sk, skb, dst, req)) {
 282                         foc->len = -1;
 283                         NET_INC_STATS_BH(sock_net(sk),
 284                                          LINUX_MIB_TCPFASTOPENPASSIVE);
 285                         return true;
 286                 }
 287         }
 288
 289         NET_INC_STATS_BH(sock_net(sk), foc->len ?
 290                          LINUX_MIB_TCPFASTOPENPASSIVEFAIL :
 291                          LINUX_MIB_TCPFASTOPENCOOKIEREQD);
 292         *foc = valid_foc;
 293         return false;
 294 }
 295 EXPORT_SYMBOL(tcp_try_fastopen);