net/iucv/af_iucv.c

   1 /*
   2  *  IUCV protocol stack for Linux on zSeries
   3  *
   4  *  Copyright IBM Corp. 2006, 2009
   5  *
   6  *  Author(s):  Jennifer Hunt <jenhunt@us.ibm.com>
   7  *              Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
   8  *  PM functions:
   9  *              Ursula Braun <ursula.braun@de.ibm.com>
  10  */
  11
  12 #define KMSG_COMPONENT "af_iucv"
  13 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  14
  15 #include <linux/module.h>
  16 #include <linux/types.h>
  17 #include <linux/list.h>
  18 #include <linux/errno.h>
  19 #include <linux/kernel.h>
  20 #include <linux/sched.h>
  21 #include <linux/slab.h>
  22 #include <linux/skbuff.h>
  23 #include <linux/init.h>
  24 #include <linux/poll.h>
  25 #include <net/sock.h>
  26 #include <asm/ebcdic.h>
  27 #include <asm/cpcmd.h>
  28 #include <linux/kmod.h>
  29
  30 #include <net/iucv/af_iucv.h>
  31
  32 #define VERSION "1.2"
  33
  34 static char iucv_userid[80];
  35
  36 static const struct proto_ops iucv_sock_ops;
  37
  38 static struct proto iucv_proto = {
  39         .name           = "AF_IUCV",
  40         .owner          = THIS_MODULE,
  41         .obj_size       = sizeof(struct iucv_sock),
  42 };
  43
  44 static struct iucv_interface *pr_iucv;
  45
  46 /* special AF_IUCV IPRM messages */
  47 static const u8 iprm_shutdown[8] =
  48         {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
  49
  50 #define TRGCLS_SIZE     (sizeof(((struct iucv_message *)0)->class))
  51
  52 /* macros to set/get socket control buffer at correct offset */
  53 #define CB_TAG(skb)     ((skb)->cb)             /* iucv message tag */
  54 #define CB_TAG_LEN      (sizeof(((struct iucv_message *) 0)->tag))
  55 #define CB_TRGCLS(skb)  ((skb)->cb + CB_TAG_LEN) /* iucv msg target class */
  56 #define CB_TRGCLS_LEN   (TRGCLS_SIZE)
  57
  58 #define __iucv_sock_wait(sk, condition, timeo, ret)                     \
  59 do {                                                                    \
  60         DEFINE_WAIT(__wait);                                            \
  61         long __timeo = timeo;                                           \
  62         ret = 0;                                                        \
  63         prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);     \
  64         while (!(condition)) {                                          \
  65                 if (!__timeo) {                                         \
  66                         ret = -EAGAIN;                                  \
  67                         break;                                          \
  68                 }                                                       \
  69                 if (signal_pending(current)) {                          \
  70                         ret = sock_intr_errno(__timeo);                 \
  71                         break;                                          \
  72                 }                                                       \
  73                 release_sock(sk);                                       \
  74                 __timeo = schedule_timeout(__timeo);                    \
  75                 lock_sock(sk);                                          \
  76                 ret = sock_error(sk);                                   \
  77                 if (ret)                                                \
  78                         break;                                          \
  79         }                                                               \
  80         finish_wait(sk_sleep(sk), &__wait);                             \
  81 } while (0)
  82
  83 #define iucv_sock_wait(sk, condition, timeo)                            \
  84 ({                                                                      \
  85         int __ret = 0;                                                  \
  86         if (!(condition))                                               \
  87                 __iucv_sock_wait(sk, condition, timeo, __ret);          \
  88         __ret;                                                          \
  89 })
  90
  91 static void iucv_sock_kill(struct sock *sk);
  92 static void iucv_sock_close(struct sock *sk);
  93 static void iucv_sever_path(struct sock *, int);
  94
  95 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
  96         struct packet_type *pt, struct net_device *orig_dev);
  97 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
  98                    struct sk_buff *skb, u8 flags);
  99 static void afiucv_hs_callback_txnotify(struct sk_buff *, enum iucv_tx_notify);
 100
 101 /* Call Back functions */
 102 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
 103 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
 104 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
 105 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
 106                                  u8 ipuser[16]);
 107 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
 108 static void iucv_callback_shutdown(struct iucv_path *, u8 ipuser[16]);
 109
 110 static struct iucv_sock_list iucv_sk_list = {
 111         .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
 112         .autobind_name = ATOMIC_INIT(0)
 113 };
 114
 115 static struct iucv_handler af_iucv_handler = {
 116         .path_pending     = iucv_callback_connreq,
 117         .path_complete    = iucv_callback_connack,
 118         .path_severed     = iucv_callback_connrej,
 119         .message_pending  = iucv_callback_rx,
 120         .message_complete = iucv_callback_txdone,
 121         .path_quiesced    = iucv_callback_shutdown,
 122 };
 123
 124 static inline void high_nmcpy(unsigned char *dst, char *src)
 125 {
 126        memcpy(dst, src, 8);
 127 }
 128
 129 static inline void low_nmcpy(unsigned char *dst, char *src)
 130 {
 131        memcpy(&dst[8], src, 8);
 132 }
 133
 134 static int afiucv_pm_prepare(struct device *dev)
 135 {
 136 #ifdef CONFIG_PM_DEBUG
 137         printk(KERN_WARNING "afiucv_pm_prepare\n");
 138 #endif
 139         return 0;
 140 }
 141
 142 static void afiucv_pm_complete(struct device *dev)
 143 {
 144 #ifdef CONFIG_PM_DEBUG
 145         printk(KERN_WARNING "afiucv_pm_complete\n");
 146 #endif
 147 }
 148
 149 /**
 150  * afiucv_pm_freeze() - Freeze PM callback
 151  * @dev:        AFIUCV dummy device
 152  *
 153  * Sever all established IUCV communication pathes
 154  */
 155 static int afiucv_pm_freeze(struct device *dev)
 156 {
 157         struct iucv_sock *iucv;
 158         struct sock *sk;
 159         int err = 0;
 160
 161 #ifdef CONFIG_PM_DEBUG
 162         printk(KERN_WARNING "afiucv_pm_freeze\n");
 163 #endif
 164         read_lock(&iucv_sk_list.lock);
 165         sk_for_each(sk, &iucv_sk_list.head) {
 166                 iucv = iucv_sk(sk);
 167                 switch (sk->sk_state) {
 168                 case IUCV_DISCONN:
 169                 case IUCV_CLOSING:
 170                 case IUCV_CONNECTED:
 171                         iucv_sever_path(sk, 0);
 172                         break;
 173                 case IUCV_OPEN:
 174                 case IUCV_BOUND:
 175                 case IUCV_LISTEN:
 176                 case IUCV_CLOSED:
 177                 default:
 178                         break;
 179                 }
 180                 skb_queue_purge(&iucv->send_skb_q);
 181                 skb_queue_purge(&iucv->backlog_skb_q);
 182         }
 183         read_unlock(&iucv_sk_list.lock);
 184         return err;
 185 }
 186
 187 /**
 188  * afiucv_pm_restore_thaw() - Thaw and restore PM callback
 189  * @dev:        AFIUCV dummy device
 190  *
 191  * socket clean up after freeze
 192  */
 193 static int afiucv_pm_restore_thaw(struct device *dev)
 194 {
 195         struct sock *sk;
 196
 197 #ifdef CONFIG_PM_DEBUG
 198         printk(KERN_WARNING "afiucv_pm_restore_thaw\n");
 199 #endif
 200         read_lock(&iucv_sk_list.lock);
 201         sk_for_each(sk, &iucv_sk_list.head) {
 202                 switch (sk->sk_state) {
 203                 case IUCV_CONNECTED:
 204                         sk->sk_err = EPIPE;
 205                         sk->sk_state = IUCV_DISCONN;
 206                         sk->sk_state_change(sk);
 207                         break;
 208                 case IUCV_DISCONN:
 209                 case IUCV_CLOSING:
 210                 case IUCV_LISTEN:
 211                 case IUCV_BOUND:
 212                 case IUCV_OPEN:
 213                 default:
 214                         break;
 215                 }
 216         }
 217         read_unlock(&iucv_sk_list.lock);
 218         return 0;
 219 }
 220
 221 static const struct dev_pm_ops afiucv_pm_ops = {
 222         .prepare = afiucv_pm_prepare,
 223         .complete = afiucv_pm_complete,
 224         .freeze = afiucv_pm_freeze,
 225         .thaw = afiucv_pm_restore_thaw,
 226         .restore = afiucv_pm_restore_thaw,
 227 };
 228
 229 static struct device_driver af_iucv_driver = {
 230         .owner = THIS_MODULE,
 231         .name = "afiucv",
 232         .bus  = NULL,
 233         .pm   = &afiucv_pm_ops,
 234 };
 235
 236 /* dummy device used as trigger for PM functions */
 237 static struct device *af_iucv_dev;
 238
 239 /**
 240  * iucv_msg_length() - Returns the length of an iucv message.
 241  * @msg:        Pointer to struct iucv_message, MUST NOT be NULL
 242  *
 243  * The function returns the length of the specified iucv message @msg of data
 244  * stored in a buffer and of data stored in the parameter list (PRMDATA).
 245  *
 246  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
 247  * data:
 248  *      PRMDATA[0..6]   socket data (max 7 bytes);
 249  *      PRMDATA[7]      socket data length value (len is 0xff - PRMDATA[7])
 250  *
 251  * The socket data length is computed by subtracting the socket data length
 252  * value from 0xFF.
 253  * If the socket data len is greater 7, then PRMDATA can be used for special
 254  * notifications (see iucv_sock_shutdown); and further,
 255  * if the socket data len is > 7, the function returns 8.
 256  *
 257  * Use this function to allocate socket buffers to store iucv message data.
 258  */
 259 static inline size_t iucv_msg_length(struct iucv_message *msg)
 260 {
 261         size_t datalen;
 262
 263         if (msg->flags & IUCV_IPRMDATA) {
 264                 datalen = 0xff - msg->rmmsg[7];
 265                 return (datalen < 8) ? datalen : 8;
 266         }
 267         return msg->length;
 268 }
 269
 270 /**
 271  * iucv_sock_in_state() - check for specific states
 272  * @sk:         sock structure
 273  * @state:      first iucv sk state
 274  * @state:      second iucv sk state
 275  *
 276  * Returns true if the socket in either in the first or second state.
 277  */
 278 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
 279 {
 280         return (sk->sk_state == state || sk->sk_state == state2);
 281 }
 282
 283 /**
 284  * iucv_below_msglim() - function to check if messages can be sent
 285  * @sk:         sock structure
 286  *
 287  * Returns true if the send queue length is lower than the message limit.
 288  * Always returns true if the socket is not connected (no iucv path for
 289  * checking the message limit).
 290  */
 291 static inline int iucv_below_msglim(struct sock *sk)
 292 {
 293         struct iucv_sock *iucv = iucv_sk(sk);
 294
 295         if (sk->sk_state != IUCV_CONNECTED)
 296                 return 1;
 297         if (iucv->transport == AF_IUCV_TRANS_IUCV)
 298                 return (skb_queue_len(&iucv->send_skb_q) < iucv->path->msglim);
 299         else
 300                 return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
 301                         (atomic_read(&iucv->pendings) <= 0));
 302 }
 303
 304 /**
 305  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
 306  */
 307 static void iucv_sock_wake_msglim(struct sock *sk)
 308 {
 309         struct socket_wq *wq;
 310
 311         rcu_read_lock();
 312         wq = rcu_dereference(sk->sk_wq);
 313         if (wq_has_sleeper(wq))
 314                 wake_up_interruptible_all(&wq->wait);
 315         sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
 316         rcu_read_unlock();
 317 }
 318
 319 /**
 320  * afiucv_hs_send() - send a message through HiperSockets transport
 321  */
 322 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
 323                    struct sk_buff *skb, u8 flags)
 324 {
 325         struct iucv_sock *iucv = iucv_sk(sock);
 326         struct af_iucv_trans_hdr *phs_hdr;
 327         struct sk_buff *nskb;
 328         int err, confirm_recv = 0;
 329
 330         memset(skb->head, 0, ETH_HLEN);
 331         phs_hdr = (struct af_iucv_trans_hdr *)skb_push(skb,
 332                                         sizeof(struct af_iucv_trans_hdr));
 333         skb_reset_mac_header(skb);
 334         skb_reset_network_header(skb);
 335         skb_push(skb, ETH_HLEN);
 336         skb_reset_mac_header(skb);
 337         memset(phs_hdr, 0, sizeof(struct af_iucv_trans_hdr));
 338
 339         phs_hdr->magic = ETH_P_AF_IUCV;
 340         phs_hdr->version = 1;
 341         phs_hdr->flags = flags;
 342         if (flags == AF_IUCV_FLAG_SYN)
 343                 phs_hdr->window = iucv->msglimit;
 344         else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
 345                 confirm_recv = atomic_read(&iucv->msg_recv);
 346                 phs_hdr->window = confirm_recv;
 347                 if (confirm_recv)
 348                         phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
 349         }
 350         memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
 351         memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
 352         memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
 353         memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
 354         ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
 355         ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
 356         ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
 357         ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
 358         if (imsg)
 359                 memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
 360
 361         skb->dev = iucv->hs_dev;
 362         if (!skb->dev)
 363                 return -ENODEV;
 364         if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev))
 365                 return -ENETDOWN;
 366         if (skb->len > skb->dev->mtu) {
 367                 if (sock->sk_type == SOCK_SEQPACKET)
 368                         return -EMSGSIZE;
 369                 else
 370                         skb_trim(skb, skb->dev->mtu);
 371         }
 372         skb->protocol = ETH_P_AF_IUCV;
 373         nskb = skb_clone(skb, GFP_ATOMIC);
 374         if (!nskb)
 375                 return -ENOMEM;
 376         skb_queue_tail(&iucv->send_skb_q, nskb);
 377         err = dev_queue_xmit(skb);
 378         if (net_xmit_eval(err)) {
 379                 skb_unlink(nskb, &iucv->send_skb_q);
 380                 kfree_skb(nskb);
 381         } else {
 382                 atomic_sub(confirm_recv, &iucv->msg_recv);
 383                 WARN_ON(atomic_read(&iucv->msg_recv) < 0);
 384         }
 385         return net_xmit_eval(err);
 386 }
 387
 388 static struct sock *__iucv_get_sock_by_name(char *nm)
 389 {
 390         struct sock *sk;
 391
 392         sk_for_each(sk, &iucv_sk_list.head)
 393                 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
 394                         return sk;
 395
 396         return NULL;
 397 }
 398
 399 static void iucv_sock_destruct(struct sock *sk)
 400 {
 401         skb_queue_purge(&sk->sk_receive_queue);
 402         skb_queue_purge(&sk->sk_error_queue);
 403
 404         sk_mem_reclaim(sk);
 405
 406         if (!sock_flag(sk, SOCK_DEAD)) {
 407                 pr_err("Attempt to release alive iucv socket %p\n", sk);
 408                 return;
 409         }
 410
 411         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
 412         WARN_ON(atomic_read(&sk->sk_wmem_alloc));
 413         WARN_ON(sk->sk_wmem_queued);
 414         WARN_ON(sk->sk_forward_alloc);
 415 }
 416
 417 /* Cleanup Listen */
 418 static void iucv_sock_cleanup_listen(struct sock *parent)
 419 {
 420         struct sock *sk;
 421
 422         /* Close non-accepted connections */
 423         while ((sk = iucv_accept_dequeue(parent, NULL))) {
 424                 iucv_sock_close(sk);
 425                 iucv_sock_kill(sk);
 426         }
 427
 428         parent->sk_state = IUCV_CLOSED;
 429 }
 430
 431 /* Kill socket (only if zapped and orphaned) */
 432 static void iucv_sock_kill(struct sock *sk)
 433 {
 434         if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
 435                 return;
 436
 437         iucv_sock_unlink(&iucv_sk_list, sk);
 438         sock_set_flag(sk, SOCK_DEAD);
 439         sock_put(sk);
 440 }
 441
 442 /* Terminate an IUCV path */
 443 static void iucv_sever_path(struct sock *sk, int with_user_data)
 444 {
 445         unsigned char user_data[16];
 446         struct iucv_sock *iucv = iucv_sk(sk);
 447         struct iucv_path *path = iucv->path;
 448
 449         if (iucv->path) {
 450                 iucv->path = NULL;
 451                 if (with_user_data) {
 452                         low_nmcpy(user_data, iucv->src_name);
 453                         high_nmcpy(user_data, iucv->dst_name);
 454                         ASCEBC(user_data, sizeof(user_data));
 455                         pr_iucv->path_sever(path, user_data);
 456                 } else
 457                         pr_iucv->path_sever(path, NULL);
 458                 iucv_path_free(path);
 459         }
 460 }
 461
 462 /* Send FIN through an IUCV socket for HIPER transport */
 463 static int iucv_send_ctrl(struct sock *sk, u8 flags)
 464 {
 465         int err = 0;
 466         int blen;
 467         struct sk_buff *skb;
 468
 469         blen = sizeof(struct af_iucv_trans_hdr) + ETH_HLEN;
 470         skb = sock_alloc_send_skb(sk, blen, 1, &err);
 471         if (skb) {
 472                 skb_reserve(skb, blen);
 473                 err = afiucv_hs_send(NULL, sk, skb, flags);
 474         }
 475         return err;
 476 }
 477
 478 /* Close an IUCV socket */
 479 static void iucv_sock_close(struct sock *sk)
 480 {
 481         struct iucv_sock *iucv = iucv_sk(sk);
 482         unsigned long timeo;
 483         int err = 0;
 484
 485         lock_sock(sk);
 486
 487         switch (sk->sk_state) {
 488         case IUCV_LISTEN:
 489                 iucv_sock_cleanup_listen(sk);
 490                 break;
 491
 492         case IUCV_CONNECTED:
 493                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
 494                         err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
 495                         sk->sk_state = IUCV_DISCONN;
 496                         sk->sk_state_change(sk);
 497                 }
 498         case IUCV_DISCONN:   /* fall through */
 499                 sk->sk_state = IUCV_CLOSING;
 500                 sk->sk_state_change(sk);
 501
 502                 if (!err && !skb_queue_empty(&iucv->send_skb_q)) {
 503                         if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
 504                                 timeo = sk->sk_lingertime;
 505                         else
 506                                 timeo = IUCV_DISCONN_TIMEOUT;
 507                         iucv_sock_wait(sk,
 508                                         iucv_sock_in_state(sk, IUCV_CLOSED, 0),
 509                                         timeo);
 510                 }
 511
 512         case IUCV_CLOSING:   /* fall through */
 513                 sk->sk_state = IUCV_CLOSED;
 514                 sk->sk_state_change(sk);
 515
 516                 sk->sk_err = ECONNRESET;
 517                 sk->sk_state_change(sk);
 518
 519                 skb_queue_purge(&iucv->send_skb_q);
 520                 skb_queue_purge(&iucv->backlog_skb_q);
 521
 522         default:   /* fall through */
 523                 iucv_sever_path(sk, 1);
 524         }
 525
 526         if (iucv->hs_dev) {
 527                 dev_put(iucv->hs_dev);
 528                 iucv->hs_dev = NULL;
 529                 sk->sk_bound_dev_if = 0;
 530         }
 531
 532         /* mark socket for deletion by iucv_sock_kill() */
 533         sock_set_flag(sk, SOCK_ZAPPED);
 534
 535         release_sock(sk);
 536 }
 537
 538 static void iucv_sock_init(struct sock *sk, struct sock *parent)
 539 {
 540         if (parent)
 541                 sk->sk_type = parent->sk_type;
 542 }
 543
 544 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
 545 {
 546         struct sock *sk;
 547         struct iucv_sock *iucv;
 548
 549         sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
 550         if (!sk)
 551                 return NULL;
 552         iucv = iucv_sk(sk);
 553
 554         sock_init_data(sock, sk);
 555         INIT_LIST_HEAD(&iucv->accept_q);
 556         spin_lock_init(&iucv->accept_q_lock);
 557         skb_queue_head_init(&iucv->send_skb_q);
 558         INIT_LIST_HEAD(&iucv->message_q.list);
 559         spin_lock_init(&iucv->message_q.lock);
 560         skb_queue_head_init(&iucv->backlog_skb_q);
 561         iucv->send_tag = 0;
 562         atomic_set(&iucv->pendings, 0);
 563         iucv->flags = 0;
 564         iucv->msglimit = 0;
 565         atomic_set(&iucv->msg_sent, 0);
 566         atomic_set(&iucv->msg_recv, 0);
 567         iucv->path = NULL;
 568         iucv->sk_txnotify = afiucv_hs_callback_txnotify;
 569         memset(&iucv->src_user_id , 0, 32);
 570         if (pr_iucv)
 571                 iucv->transport = AF_IUCV_TRANS_IUCV;
 572         else
 573                 iucv->transport = AF_IUCV_TRANS_HIPER;
 574
 575         sk->sk_destruct = iucv_sock_destruct;
 576         sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
 577         sk->sk_allocation = GFP_DMA;
 578
 579         sock_reset_flag(sk, SOCK_ZAPPED);
 580
 581         sk->sk_protocol = proto;
 582         sk->sk_state    = IUCV_OPEN;
 583
 584         iucv_sock_link(&iucv_sk_list, sk);
 585         return sk;
 586 }
 587
 588 /* Create an IUCV socket */
 589 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
 590                             int kern)
 591 {
 592         struct sock *sk;
 593
 594         if (protocol && protocol != PF_IUCV)
 595                 return -EPROTONOSUPPORT;
 596
 597         sock->state = SS_UNCONNECTED;
 598
 599         switch (sock->type) {
 600         case SOCK_STREAM:
 601                 sock->ops = &iucv_sock_ops;
 602                 break;
 603         case SOCK_SEQPACKET:
 604                 /* currently, proto ops can handle both sk types */
 605                 sock->ops = &iucv_sock_ops;
 606                 break;
 607         default:
 608                 return -ESOCKTNOSUPPORT;
 609         }
 610
 611         sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
 612         if (!sk)
 613                 return -ENOMEM;
 614
 615         iucv_sock_init(sk, NULL);
 616
 617         return 0;
 618 }
 619
 620 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
 621 {
 622         write_lock_bh(&l->lock);
 623         sk_add_node(sk, &l->head);
 624         write_unlock_bh(&l->lock);
 625 }
 626
 627 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
 628 {
 629         write_lock_bh(&l->lock);
 630         sk_del_node_init(sk);
 631         write_unlock_bh(&l->lock);
 632 }
 633
 634 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
 635 {
 636         unsigned long flags;
 637         struct iucv_sock *par = iucv_sk(parent);
 638
 639         sock_hold(sk);
 640         spin_lock_irqsave(&par->accept_q_lock, flags);
 641         list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
 642         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 643         iucv_sk(sk)->parent = parent;
 644         sk_acceptq_added(parent);
 645 }
 646
 647 void iucv_accept_unlink(struct sock *sk)
 648 {
 649         unsigned long flags;
 650         struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
 651
 652         spin_lock_irqsave(&par->accept_q_lock, flags);
 653         list_del_init(&iucv_sk(sk)->accept_q);
 654         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 655         sk_acceptq_removed(iucv_sk(sk)->parent);
 656         iucv_sk(sk)->parent = NULL;
 657         sock_put(sk);
 658 }
 659
 660 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
 661 {
 662         struct iucv_sock *isk, *n;
 663         struct sock *sk;
 664
 665         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
 666                 sk = (struct sock *) isk;
 667                 lock_sock(sk);
 668
 669                 if (sk->sk_state == IUCV_CLOSED) {
 670                         iucv_accept_unlink(sk);
 671                         release_sock(sk);
 672                         continue;
 673                 }
 674
 675                 if (sk->sk_state == IUCV_CONNECTED ||
 676                     sk->sk_state == IUCV_DISCONN ||
 677                     !newsock) {
 678                         iucv_accept_unlink(sk);
 679                         if (newsock)
 680                                 sock_graft(sk, newsock);
 681
 682                         release_sock(sk);
 683                         return sk;
 684                 }
 685
 686                 release_sock(sk);
 687         }
 688         return NULL;
 689 }
 690
 691 /* Bind an unbound socket */
 692 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
 693                           int addr_len)
 694 {
 695         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 696         struct sock *sk = sock->sk;
 697         struct iucv_sock *iucv;
 698         int err = 0;
 699         struct net_device *dev;
 700         char uid[9];
 701
 702         /* Verify the input sockaddr */
 703         if (!addr || addr->sa_family != AF_IUCV)
 704                 return -EINVAL;
 705
 706         lock_sock(sk);
 707         if (sk->sk_state != IUCV_OPEN) {
 708                 err = -EBADFD;
 709                 goto done;
 710         }
 711
 712         write_lock_bh(&iucv_sk_list.lock);
 713
 714         iucv = iucv_sk(sk);
 715         if (__iucv_get_sock_by_name(sa->siucv_name)) {
 716                 err = -EADDRINUSE;
 717                 goto done_unlock;
 718         }
 719         if (iucv->path)
 720                 goto done_unlock;
 721
 722         /* Bind the socket */
 723         if (pr_iucv)
 724                 if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
 725                         goto vm_bind; /* VM IUCV transport */
 726
 727         /* try hiper transport */
 728         memcpy(uid, sa->siucv_user_id, sizeof(uid));
 729         ASCEBC(uid, 8);
 730         rcu_read_lock();
 731         for_each_netdev_rcu(&init_net, dev) {
 732                 if (!memcmp(dev->perm_addr, uid, 8)) {
 733                         memcpy(iucv->src_name, sa->siucv_name, 8);
 734                         memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
 735                         sk->sk_bound_dev_if = dev->ifindex;
 736                         iucv->hs_dev = dev;
 737                         dev_hold(dev);
 738                         sk->sk_state = IUCV_BOUND;
 739                         iucv->transport = AF_IUCV_TRANS_HIPER;
 740                         if (!iucv->msglimit)
 741                                 iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
 742                         rcu_read_unlock();
 743                         goto done_unlock;
 744                 }
 745         }
 746         rcu_read_unlock();
 747 vm_bind:
 748         if (pr_iucv) {
 749                 /* use local userid for backward compat */
 750                 memcpy(iucv->src_name, sa->siucv_name, 8);
 751                 memcpy(iucv->src_user_id, iucv_userid, 8);
 752                 sk->sk_state = IUCV_BOUND;
 753                 iucv->transport = AF_IUCV_TRANS_IUCV;
 754                 if (!iucv->msglimit)
 755                         iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 756                 goto done_unlock;
 757         }
 758         /* found no dev to bind */
 759         err = -ENODEV;
 760 done_unlock:
 761         /* Release the socket list lock */
 762         write_unlock_bh(&iucv_sk_list.lock);
 763 done:
 764         release_sock(sk);
 765         return err;
 766 }
 767
 768 /* Automatically bind an unbound socket */
 769 static int iucv_sock_autobind(struct sock *sk)
 770 {
 771         struct iucv_sock *iucv = iucv_sk(sk);
 772         char name[12];
 773         int err = 0;
 774
 775         if (unlikely(!pr_iucv))
 776                 return -EPROTO;
 777
 778         memcpy(iucv->src_user_id, iucv_userid, 8);
 779
 780         write_lock_bh(&iucv_sk_list.lock);
 781
 782         sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
 783         while (__iucv_get_sock_by_name(name)) {
 784                 sprintf(name, "%08x",
 785                         atomic_inc_return(&iucv_sk_list.autobind_name));
 786         }
 787
 788         write_unlock_bh(&iucv_sk_list.lock);
 789
 790         memcpy(&iucv->src_name, name, 8);
 791
 792         if (!iucv->msglimit)
 793                 iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 794
 795         return err;
 796 }
 797
 798 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
 799 {
 800         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 801         struct sock *sk = sock->sk;
 802         struct iucv_sock *iucv = iucv_sk(sk);
 803         unsigned char user_data[16];
 804         int err;
 805
 806         high_nmcpy(user_data, sa->siucv_name);
 807         low_nmcpy(user_data, iucv->src_name);
 808         ASCEBC(user_data, sizeof(user_data));
 809
 810         /* Create path. */
 811         iucv->path = iucv_path_alloc(iucv->msglimit,
 812                                      IUCV_IPRMDATA, GFP_KERNEL);
 813         if (!iucv->path) {
 814                 err = -ENOMEM;
 815                 goto done;
 816         }
 817         err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
 818                                     sa->siucv_user_id, NULL, user_data,
 819                                     sk);
 820         if (err) {
 821                 iucv_path_free(iucv->path);
 822                 iucv->path = NULL;
 823                 switch (err) {
 824                 case 0x0b:      /* Target communicator is not logged on */
 825                         err = -ENETUNREACH;
 826                         break;
 827                 case 0x0d:      /* Max connections for this guest exceeded */
 828                 case 0x0e:      /* Max connections for target guest exceeded */
 829                         err = -EAGAIN;
 830                         break;
 831                 case 0x0f:      /* Missing IUCV authorization */
 832                         err = -EACCES;
 833                         break;
 834                 default:
 835                         err = -ECONNREFUSED;
 836                         break;
 837                 }
 838         }
 839 done:
 840         return err;
 841 }
 842
 843 /* Connect an unconnected socket */
 844 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
 845                              int alen, int flags)
 846 {
 847         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 848         struct sock *sk = sock->sk;
 849         struct iucv_sock *iucv = iucv_sk(sk);
 850         int err;
 851
 852         if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
 853                 return -EINVAL;
 854
 855         if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
 856                 return -EBADFD;
 857
 858         if (sk->sk_state == IUCV_OPEN &&
 859             iucv->transport == AF_IUCV_TRANS_HIPER)
 860                 return -EBADFD; /* explicit bind required */
 861
 862         if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
 863                 return -EINVAL;
 864
 865         if (sk->sk_state == IUCV_OPEN) {
 866                 err = iucv_sock_autobind(sk);
 867                 if (unlikely(err))
 868                         return err;
 869         }
 870
 871         lock_sock(sk);
 872
 873         /* Set the destination information */
 874         memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
 875         memcpy(iucv->dst_name, sa->siucv_name, 8);
 876
 877         if (iucv->transport == AF_IUCV_TRANS_HIPER)
 878                 err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
 879         else
 880                 err = afiucv_path_connect(sock, addr);
 881         if (err)
 882                 goto done;
 883
 884         if (sk->sk_state != IUCV_CONNECTED)
 885                 err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
 886                                                             IUCV_DISCONN),
 887                                      sock_sndtimeo(sk, flags & O_NONBLOCK));
 888
 889         if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
 890                 err = -ECONNREFUSED;
 891
 892         if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
 893                 iucv_sever_path(sk, 0);
 894
 895 done:
 896         release_sock(sk);
 897         return err;
 898 }
 899
 900 /* Move a socket into listening state. */
 901 static int iucv_sock_listen(struct socket *sock, int backlog)
 902 {
 903         struct sock *sk = sock->sk;
 904         int err;
 905
 906         lock_sock(sk);
 907
 908         err = -EINVAL;
 909         if (sk->sk_state != IUCV_BOUND)
 910                 goto done;
 911
 912         if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
 913                 goto done;
 914
 915         sk->sk_max_ack_backlog = backlog;
 916         sk->sk_ack_backlog = 0;
 917         sk->sk_state = IUCV_LISTEN;
 918         err = 0;
 919
 920 done:
 921         release_sock(sk);
 922         return err;
 923 }
 924
 925 /* Accept a pending connection */
 926 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
 927                             int flags)
 928 {
 929         DECLARE_WAITQUEUE(wait, current);
 930         struct sock *sk = sock->sk, *nsk;
 931         long timeo;
 932         int err = 0;
 933
 934         lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 935
 936         if (sk->sk_state != IUCV_LISTEN) {
 937                 err = -EBADFD;
 938                 goto done;
 939         }
 940
 941         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
 942
 943         /* Wait for an incoming connection */
 944         add_wait_queue_exclusive(sk_sleep(sk), &wait);
 945         while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
 946                 set_current_state(TASK_INTERRUPTIBLE);
 947                 if (!timeo) {
 948                         err = -EAGAIN;
 949                         break;
 950                 }
 951
 952                 release_sock(sk);
 953                 timeo = schedule_timeout(timeo);
 954                 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 955
 956                 if (sk->sk_state != IUCV_LISTEN) {
 957                         err = -EBADFD;
 958                         break;
 959                 }
 960
 961                 if (signal_pending(current)) {
 962                         err = sock_intr_errno(timeo);
 963                         break;
 964                 }
 965         }
 966
 967         set_current_state(TASK_RUNNING);
 968         remove_wait_queue(sk_sleep(sk), &wait);
 969
 970         if (err)
 971                 goto done;
 972
 973         newsock->state = SS_CONNECTED;
 974
 975 done:
 976         release_sock(sk);
 977         return err;
 978 }
 979
 980 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
 981                              int *len, int peer)
 982 {
 983         struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
 984         struct sock *sk = sock->sk;
 985         struct iucv_sock *iucv = iucv_sk(sk);
 986
 987         addr->sa_family = AF_IUCV;
 988         *len = sizeof(struct sockaddr_iucv);
 989
 990         if (peer) {
 991                 memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
 992                 memcpy(siucv->siucv_name, iucv->dst_name, 8);
 993         } else {
 994                 memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
 995                 memcpy(siucv->siucv_name, iucv->src_name, 8);
 996         }
 997         memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
 998         memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
 999         memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
1000
1001         return 0;
1002 }
1003
1004 /**
1005  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
1006  * @path:       IUCV path
1007  * @msg:        Pointer to a struct iucv_message
1008  * @skb:        The socket data to send, skb->len MUST BE <= 7
1009  *
1010  * Send the socket data in the parameter list in the iucv message
1011  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
1012  * list and the socket data len at index 7 (last byte).
1013  * See also iucv_msg_length().
1014  *
1015  * Returns the error code from the iucv_message_send() call.
1016  */
1017 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
1018                           struct sk_buff *skb)
1019 {
1020         u8 prmdata[8];
1021
1022         memcpy(prmdata, (void *) skb->data, skb->len);
1023         prmdata[7] = 0xff - (u8) skb->len;
1024         return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
1025                                  (void *) prmdata, 8);
1026 }
1027
1028 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1029                              struct msghdr *msg, size_t len)
1030 {
1031         struct sock *sk = sock->sk;
1032         struct iucv_sock *iucv = iucv_sk(sk);
1033         struct sk_buff *skb;
1034         struct iucv_message txmsg;
1035         struct cmsghdr *cmsg;
1036         int cmsg_done;
1037         long timeo;
1038         char user_id[9];
1039         char appl_id[9];
1040         int err;
1041         int noblock = msg->msg_flags & MSG_DONTWAIT;
1042
1043         err = sock_error(sk);
1044         if (err)
1045                 return err;
1046
1047         if (msg->msg_flags & MSG_OOB)
1048                 return -EOPNOTSUPP;
1049
1050         /* SOCK_SEQPACKET: we do not support segmented records */
1051         if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
1052                 return -EOPNOTSUPP;
1053
1054         lock_sock(sk);
1055
1056         if (sk->sk_shutdown & SEND_SHUTDOWN) {
1057                 err = -EPIPE;
1058                 goto out;
1059         }
1060
1061         /* Return if the socket is not in connected state */
1062         if (sk->sk_state != IUCV_CONNECTED) {
1063                 err = -ENOTCONN;
1064                 goto out;
1065         }
1066
1067         /* initialize defaults */
1068         cmsg_done   = 0;        /* check for duplicate headers */
1069         txmsg.class = 0;
1070
1071         /* iterate over control messages */
1072         for (cmsg = CMSG_FIRSTHDR(msg); cmsg;
1073                 cmsg = CMSG_NXTHDR(msg, cmsg)) {
1074
1075                 if (!CMSG_OK(msg, cmsg)) {
1076                         err = -EINVAL;
1077                         goto out;
1078                 }
1079
1080                 if (cmsg->cmsg_level != SOL_IUCV)
1081                         continue;
1082
1083                 if (cmsg->cmsg_type & cmsg_done) {
1084                         err = -EINVAL;
1085                         goto out;
1086                 }
1087                 cmsg_done |= cmsg->cmsg_type;
1088
1089                 switch (cmsg->cmsg_type) {
1090                 case SCM_IUCV_TRGCLS:
1091                         if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
1092                                 err = -EINVAL;
1093                                 goto out;
1094                         }
1095
1096                         /* set iucv message target class */
1097                         memcpy(&txmsg.class,
1098                                 (void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
1099
1100                         break;
1101
1102                 default:
1103                         err = -EINVAL;
1104                         goto out;
1105                         break;
1106                 }
1107         }
1108
1109         /* allocate one skb for each iucv message:
1110          * this is fine for SOCK_SEQPACKET (unless we want to support
1111          * segmented records using the MSG_EOR flag), but
1112          * for SOCK_STREAM we might want to improve it in future */
1113         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1114                 skb = sock_alloc_send_skb(sk,
1115                         len + sizeof(struct af_iucv_trans_hdr) + ETH_HLEN,
1116                         noblock, &err);
1117         else
1118                 skb = sock_alloc_send_skb(sk, len, noblock, &err);
1119         if (!skb) {
1120                 err = -ENOMEM;
1121                 goto out;
1122         }
1123         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1124                 skb_reserve(skb, sizeof(struct af_iucv_trans_hdr) + ETH_HLEN);
1125         if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1126                 err = -EFAULT;
1127                 goto fail;
1128         }
1129
1130         /* wait if outstanding messages for iucv path has reached */
1131         timeo = sock_sndtimeo(sk, noblock);
1132         err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1133         if (err)
1134                 goto fail;
1135
1136         /* return -ECONNRESET if the socket is no longer connected */
1137         if (sk->sk_state != IUCV_CONNECTED) {
1138                 err = -ECONNRESET;
1139                 goto fail;
1140         }
1141
1142         /* increment and save iucv message tag for msg_completion cbk */
1143         txmsg.tag = iucv->send_tag++;
1144         memcpy(CB_TAG(skb), &txmsg.tag, CB_TAG_LEN);
1145
1146         if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1147                 atomic_inc(&iucv->msg_sent);
1148                 err = afiucv_hs_send(&txmsg, sk, skb, 0);
1149                 if (err) {
1150                         atomic_dec(&iucv->msg_sent);
1151                         goto fail;
1152                 }
1153                 goto release;
1154         }
1155         skb_queue_tail(&iucv->send_skb_q, skb);
1156
1157         if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags)
1158               && skb->len <= 7) {
1159                 err = iucv_send_iprm(iucv->path, &txmsg, skb);
1160
1161                 /* on success: there is no message_complete callback
1162                  * for an IPRMDATA msg; remove skb from send queue */
1163                 if (err == 0) {
1164                         skb_unlink(skb, &iucv->send_skb_q);
1165                         kfree_skb(skb);
1166                 }
1167
1168                 /* this error should never happen since the
1169                  * IUCV_IPRMDATA path flag is set... sever path */
1170                 if (err == 0x15) {
1171                         pr_iucv->path_sever(iucv->path, NULL);
1172                         skb_unlink(skb, &iucv->send_skb_q);
1173                         err = -EPIPE;
1174                         goto fail;
1175                 }
1176         } else
1177                 err = pr_iucv->message_send(iucv->path, &txmsg, 0, 0,
1178                                         (void *) skb->data, skb->len);
1179         if (err) {
1180                 if (err == 3) {
1181                         user_id[8] = 0;
1182                         memcpy(user_id, iucv->dst_user_id, 8);
1183                         appl_id[8] = 0;
1184                         memcpy(appl_id, iucv->dst_name, 8);
1185                         pr_err("Application %s on z/VM guest %s"
1186                                 " exceeds message limit\n",
1187                                 appl_id, user_id);
1188                         err = -EAGAIN;
1189                 } else
1190                         err = -EPIPE;
1191                 skb_unlink(skb, &iucv->send_skb_q);
1192                 goto fail;
1193         }
1194
1195 release:
1196         release_sock(sk);
1197         return len;
1198
1199 fail:
1200         kfree_skb(skb);
1201 out:
1202         release_sock(sk);
1203         return err;
1204 }
1205
1206 /* iucv_fragment_skb() - Fragment a single IUCV message into multiple skb's
1207  *
1208  * Locking: must be called with message_q.lock held
1209  */
1210 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
1211 {
1212         int dataleft, size, copied = 0;
1213         struct sk_buff *nskb;
1214
1215         dataleft = len;
1216         while (dataleft) {
1217                 if (dataleft >= sk->sk_rcvbuf / 4)
1218                         size = sk->sk_rcvbuf / 4;
1219                 else
1220                         size = dataleft;
1221
1222                 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
1223                 if (!nskb)
1224                         return -ENOMEM;
1225
1226                 /* copy target class to control buffer of new skb */
1227                 memcpy(CB_TRGCLS(nskb), CB_TRGCLS(skb), CB_TRGCLS_LEN);
1228
1229                 /* copy data fragment */
1230                 memcpy(nskb->data, skb->data + copied, size);
1231                 copied += size;
1232                 dataleft -= size;
1233
1234                 skb_reset_transport_header(nskb);
1235                 skb_reset_network_header(nskb);
1236                 nskb->len = size;
1237
1238                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
1239         }
1240
1241         return 0;
1242 }
1243
1244 /* iucv_process_message() - Receive a single outstanding IUCV message
1245  *
1246  * Locking: must be called with message_q.lock held
1247  */
1248 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1249                                  struct iucv_path *path,
1250                                  struct iucv_message *msg)
1251 {
1252         int rc;
1253         unsigned int len;
1254
1255         len = iucv_msg_length(msg);
1256
1257         /* store msg target class in the second 4 bytes of skb ctrl buffer */
1258         /* Note: the first 4 bytes are reserved for msg tag */
1259         memcpy(CB_TRGCLS(skb), &msg->class, CB_TRGCLS_LEN);
1260
1261         /* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1262         if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1263                 if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1264                         skb->data = NULL;
1265                         skb->len = 0;
1266                 }
1267         } else {
1268                 rc = pr_iucv->message_receive(path, msg,
1269                                               msg->flags & IUCV_IPRMDATA,
1270                                               skb->data, len, NULL);
1271                 if (rc) {
1272                         kfree_skb(skb);
1273                         return;
1274                 }
1275                 /* we need to fragment iucv messages for SOCK_STREAM only;
1276                  * for SOCK_SEQPACKET, it is only relevant if we support
1277                  * record segmentation using MSG_EOR (see also recvmsg()) */
1278                 if (sk->sk_type == SOCK_STREAM &&
1279                     skb->truesize >= sk->sk_rcvbuf / 4) {
1280                         rc = iucv_fragment_skb(sk, skb, len);
1281                         kfree_skb(skb);
1282                         skb = NULL;
1283                         if (rc) {
1284                                 pr_iucv->path_sever(path, NULL);
1285                                 return;
1286                         }
1287                         skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
1288                 } else {
1289                         skb_reset_transport_header(skb);
1290                         skb_reset_network_header(skb);
1291                         skb->len = len;
1292                 }
1293         }
1294
1295         if (sock_queue_rcv_skb(sk, skb))
1296                 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
1297 }
1298
1299 /* iucv_process_message_q() - Process outstanding IUCV messages
1300  *
1301  * Locking: must be called with message_q.lock held
1302  */
1303 static void iucv_process_message_q(struct sock *sk)
1304 {
1305         struct iucv_sock *iucv = iucv_sk(sk);
1306         struct sk_buff *skb;
1307         struct sock_msg_q *p, *n;
1308
1309         list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1310                 skb = alloc_skb(iucv_msg_length(&p->msg), GFP_ATOMIC | GFP_DMA);
1311                 if (!skb)
1312                         break;
1313                 iucv_process_message(sk, skb, p->path, &p->msg);
1314                 list_del(&p->list);
1315                 kfree(p);
1316                 if (!skb_queue_empty(&iucv->backlog_skb_q))
1317                         break;
1318         }
1319 }
1320
1321 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
1322                              struct msghdr *msg, size_t len, int flags)
1323 {
1324         int noblock = flags & MSG_DONTWAIT;
1325         struct sock *sk = sock->sk;
1326         struct iucv_sock *iucv = iucv_sk(sk);
1327         unsigned int copied, rlen;
1328         struct sk_buff *skb, *rskb, *cskb;
1329         int err = 0;
1330
1331         if ((sk->sk_state == IUCV_DISCONN) &&
1332             skb_queue_empty(&iucv->backlog_skb_q) &&
1333             skb_queue_empty(&sk->sk_receive_queue) &&
1334             list_empty(&iucv->message_q.list))
1335                 return 0;
1336
1337         if (flags & (MSG_OOB))
1338                 return -EOPNOTSUPP;
1339
1340         /* receive/dequeue next skb:
1341          * the function understands MSG_PEEK and, thus, does not dequeue skb */
1342         skb = skb_recv_datagram(sk, flags, noblock, &err);
1343         if (!skb) {
1344                 if (sk->sk_shutdown & RCV_SHUTDOWN)
1345                         return 0;
1346                 return err;
1347         }
1348
1349         rlen   = skb->len;              /* real length of skb */
1350         copied = min_t(unsigned int, rlen, len);
1351         if (!rlen)
1352                 sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1353
1354         cskb = skb;
1355         if (skb_copy_datagram_iovec(cskb, 0, msg->msg_iov, copied)) {
1356                 if (!(flags & MSG_PEEK))
1357                         skb_queue_head(&sk->sk_receive_queue, skb);
1358                 return -EFAULT;
1359         }
1360
1361         /* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1362         if (sk->sk_type == SOCK_SEQPACKET) {
1363                 if (copied < rlen)
1364                         msg->msg_flags |= MSG_TRUNC;
1365                 /* each iucv message contains a complete record */
1366                 msg->msg_flags |= MSG_EOR;
1367         }
1368
1369         /* create control message to store iucv msg target class:
1370          * get the trgcls from the control buffer of the skb due to
1371          * fragmentation of original iucv message. */
1372         err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1373                         CB_TRGCLS_LEN, CB_TRGCLS(skb));
1374         if (err) {
1375                 if (!(flags & MSG_PEEK))
1376                         skb_queue_head(&sk->sk_receive_queue, skb);
1377                 return err;
1378         }
1379
1380         /* Mark read part of skb as used */
1381         if (!(flags & MSG_PEEK)) {
1382
1383                 /* SOCK_STREAM: re-queue skb if it contains unreceived data */
1384                 if (sk->sk_type == SOCK_STREAM) {
1385                         skb_pull(skb, copied);
1386                         if (skb->len) {
1387                                 skb_queue_head(&sk->sk_receive_queue, skb);
1388                                 goto done;
1389                         }
1390                 }
1391
1392                 kfree_skb(skb);
1393                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1394                         atomic_inc(&iucv->msg_recv);
1395                         if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1396                                 WARN_ON(1);
1397                                 iucv_sock_close(sk);
1398                                 return -EFAULT;
1399                         }
1400                 }
1401
1402                 /* Queue backlog skbs */
1403                 spin_lock_bh(&iucv->message_q.lock);
1404                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1405                 while (rskb) {
1406                         if (sock_queue_rcv_skb(sk, rskb)) {
1407                                 skb_queue_head(&iucv->backlog_skb_q,
1408                                                 rskb);
1409                                 break;
1410                         } else {
1411                                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1412                         }
1413                 }
1414                 if (skb_queue_empty(&iucv->backlog_skb_q)) {
1415                         if (!list_empty(&iucv->message_q.list))
1416                                 iucv_process_message_q(sk);
1417                         if (atomic_read(&iucv->msg_recv) >=
1418                                                         iucv->msglimit / 2) {
1419                                 err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1420                                 if (err) {
1421                                         sk->sk_state = IUCV_DISCONN;
1422                                         sk->sk_state_change(sk);
1423                                 }
1424                         }
1425                 }
1426                 spin_unlock_bh(&iucv->message_q.lock);
1427         }
1428
1429 done:
1430         /* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1431         if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1432                 copied = rlen;
1433
1434         return copied;
1435 }
1436
1437 static inline unsigned int iucv_accept_poll(struct sock *parent)
1438 {
1439         struct iucv_sock *isk, *n;
1440         struct sock *sk;
1441
1442         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1443                 sk = (struct sock *) isk;
1444
1445                 if (sk->sk_state == IUCV_CONNECTED)
1446                         return POLLIN | POLLRDNORM;
1447         }
1448
1449         return 0;
1450 }
1451
1452 unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
1453                             poll_table *wait)
1454 {
1455         struct sock *sk = sock->sk;
1456         unsigned int mask = 0;
1457
1458         sock_poll_wait(file, sk_sleep(sk), wait);
1459
1460         if (sk->sk_state == IUCV_LISTEN)
1461                 return iucv_accept_poll(sk);
1462
1463         if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1464                 mask |= POLLERR;
1465
1466         if (sk->sk_shutdown & RCV_SHUTDOWN)
1467                 mask |= POLLRDHUP;
1468
1469         if (sk->sk_shutdown == SHUTDOWN_MASK)
1470                 mask |= POLLHUP;
1471
1472         if (!skb_queue_empty(&sk->sk_receive_queue) ||
1473             (sk->sk_shutdown & RCV_SHUTDOWN))
1474                 mask |= POLLIN | POLLRDNORM;
1475
1476         if (sk->sk_state == IUCV_CLOSED)
1477                 mask |= POLLHUP;
1478
1479         if (sk->sk_state == IUCV_DISCONN)
1480                 mask |= POLLIN;
1481
1482         if (sock_writeable(sk) && iucv_below_msglim(sk))
1483                 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1484         else
1485                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1486
1487         return mask;
1488 }
1489
1490 static int iucv_sock_shutdown(struct socket *sock, int how)
1491 {
1492         struct sock *sk = sock->sk;
1493         struct iucv_sock *iucv = iucv_sk(sk);
1494         struct iucv_message txmsg;
1495         int err = 0;
1496
1497         how++;
1498
1499         if ((how & ~SHUTDOWN_MASK) || !how)
1500                 return -EINVAL;
1501
1502         lock_sock(sk);
1503         switch (sk->sk_state) {
1504         case IUCV_LISTEN:
1505         case IUCV_DISCONN:
1506         case IUCV_CLOSING:
1507         case IUCV_CLOSED:
1508                 err = -ENOTCONN;
1509                 goto fail;
1510         default:
1511                 break;
1512         }
1513
1514         if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
1515                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1516                         txmsg.class = 0;
1517                         txmsg.tag = 0;
1518                         err = pr_iucv->message_send(iucv->path, &txmsg,
1519                                 IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1520                         if (err) {
1521                                 switch (err) {
1522                                 case 1:
1523                                         err = -ENOTCONN;
1524                                         break;
1525                                 case 2:
1526                                         err = -ECONNRESET;
1527                                         break;
1528                                 default:
1529                                         err = -ENOTCONN;
1530                                         break;
1531                                 }
1532                         }
1533                 } else
1534                         iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1535         }
1536
1537         sk->sk_shutdown |= how;
1538         if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1539                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1540                         err = pr_iucv->path_quiesce(iucv->path, NULL);
1541                         if (err)
1542                                 err = -ENOTCONN;
1543 /*                      skb_queue_purge(&sk->sk_receive_queue); */
1544                 }
1545                 skb_queue_purge(&sk->sk_receive_queue);
1546         }
1547
1548         /* Wake up anyone sleeping in poll */
1549         sk->sk_state_change(sk);
1550
1551 fail:
1552         release_sock(sk);
1553         return err;
1554 }
1555
1556 static int iucv_sock_release(struct socket *sock)
1557 {
1558         struct sock *sk = sock->sk;
1559         int err = 0;
1560
1561         if (!sk)
1562                 return 0;
1563
1564         iucv_sock_close(sk);
1565
1566         sock_orphan(sk);
1567         iucv_sock_kill(sk);
1568         return err;
1569 }
1570
1571 /* getsockopt and setsockopt */
1572 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1573                                 char __user *optval, unsigned int optlen)
1574 {
1575         struct sock *sk = sock->sk;
1576         struct iucv_sock *iucv = iucv_sk(sk);
1577         int val;
1578         int rc;
1579
1580         if (level != SOL_IUCV)
1581                 return -ENOPROTOOPT;
1582
1583         if (optlen < sizeof(int))
1584                 return -EINVAL;
1585
1586         if (get_user(val, (int __user *) optval))
1587                 return -EFAULT;
1588
1589         rc = 0;
1590
1591         lock_sock(sk);
1592         switch (optname) {
1593         case SO_IPRMDATA_MSG:
1594                 if (val)
1595                         iucv->flags |= IUCV_IPRMDATA;
1596                 else
1597                         iucv->flags &= ~IUCV_IPRMDATA;
1598                 break;
1599         case SO_MSGLIMIT:
1600                 switch (sk->sk_state) {
1601                 case IUCV_OPEN:
1602                 case IUCV_BOUND:
1603                         if (val < 1 || val > (u16)(~0))
1604                                 rc = -EINVAL;
1605                         else
1606                                 iucv->msglimit = val;
1607                         break;
1608                 default:
1609                         rc = -EINVAL;
1610                         break;
1611                 }
1612                 break;
1613         default:
1614                 rc = -ENOPROTOOPT;
1615                 break;
1616         }
1617         release_sock(sk);
1618
1619         return rc;
1620 }
1621
1622 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1623                                 char __user *optval, int __user *optlen)
1624 {
1625         struct sock *sk = sock->sk;
1626         struct iucv_sock *iucv = iucv_sk(sk);
1627         unsigned int val;
1628         int len;
1629
1630         if (level != SOL_IUCV)
1631                 return -ENOPROTOOPT;
1632
1633         if (get_user(len, optlen))
1634                 return -EFAULT;
1635
1636         if (len < 0)
1637                 return -EINVAL;
1638
1639         len = min_t(unsigned int, len, sizeof(int));
1640
1641         switch (optname) {
1642         case SO_IPRMDATA_MSG:
1643                 val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1644                 break;
1645         case SO_MSGLIMIT:
1646                 lock_sock(sk);
1647                 val = (iucv->path != NULL) ? iucv->path->msglim /* connected */
1648                                            : iucv->msglimit;    /* default */
1649                 release_sock(sk);
1650                 break;
1651         case SO_MSGSIZE:
1652                 if (sk->sk_state == IUCV_OPEN)
1653                         return -EBADFD;
1654                 val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1655                                 sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1656                                 0x7fffffff;
1657                 break;
1658         default:
1659                 return -ENOPROTOOPT;
1660         }
1661
1662         if (put_user(len, optlen))
1663                 return -EFAULT;
1664         if (copy_to_user(optval, &val, len))
1665                 return -EFAULT;
1666
1667         return 0;
1668 }
1669
1670
1671 /* Callback wrappers - called from iucv base support */
1672 static int iucv_callback_connreq(struct iucv_path *path,
1673                                  u8 ipvmid[8], u8 ipuser[16])
1674 {
1675         unsigned char user_data[16];
1676         unsigned char nuser_data[16];
1677         unsigned char src_name[8];
1678         struct sock *sk, *nsk;
1679         struct iucv_sock *iucv, *niucv;
1680         int err;
1681
1682         memcpy(src_name, ipuser, 8);
1683         EBCASC(src_name, 8);
1684         /* Find out if this path belongs to af_iucv. */
1685         read_lock(&iucv_sk_list.lock);
1686         iucv = NULL;
1687         sk = NULL;
1688         sk_for_each(sk, &iucv_sk_list.head)
1689                 if (sk->sk_state == IUCV_LISTEN &&
1690                     !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1691                         /*
1692                          * Found a listening socket with
1693                          * src_name == ipuser[0-7].
1694                          */
1695                         iucv = iucv_sk(sk);
1696                         break;
1697                 }
1698         read_unlock(&iucv_sk_list.lock);
1699         if (!iucv)
1700                 /* No socket found, not one of our paths. */
1701                 return -EINVAL;
1702
1703         bh_lock_sock(sk);
1704
1705         /* Check if parent socket is listening */
1706         low_nmcpy(user_data, iucv->src_name);
1707         high_nmcpy(user_data, iucv->dst_name);
1708         ASCEBC(user_data, sizeof(user_data));
1709         if (sk->sk_state != IUCV_LISTEN) {
1710                 err = pr_iucv->path_sever(path, user_data);
1711                 iucv_path_free(path);
1712                 goto fail;
1713         }
1714
1715         /* Check for backlog size */
1716         if (sk_acceptq_is_full(sk)) {
1717                 err = pr_iucv->path_sever(path, user_data);
1718                 iucv_path_free(path);
1719                 goto fail;
1720         }
1721
1722         /* Create the new socket */
1723         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1724         if (!nsk) {
1725                 err = pr_iucv->path_sever(path, user_data);
1726                 iucv_path_free(path);
1727                 goto fail;
1728         }
1729
1730         niucv = iucv_sk(nsk);
1731         iucv_sock_init(nsk, sk);
1732
1733         /* Set the new iucv_sock */
1734         memcpy(niucv->dst_name, ipuser + 8, 8);
1735         EBCASC(niucv->dst_name, 8);
1736         memcpy(niucv->dst_user_id, ipvmid, 8);
1737         memcpy(niucv->src_name, iucv->src_name, 8);
1738         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1739         niucv->path = path;
1740
1741         /* Call iucv_accept */
1742         high_nmcpy(nuser_data, ipuser + 8);
1743         memcpy(nuser_data + 8, niucv->src_name, 8);
1744         ASCEBC(nuser_data + 8, 8);
1745
1746         /* set message limit for path based on msglimit of accepting socket */
1747         niucv->msglimit = iucv->msglimit;
1748         path->msglim = iucv->msglimit;
1749         err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1750         if (err) {
1751                 iucv_sever_path(nsk, 1);
1752                 iucv_sock_kill(nsk);
1753                 goto fail;
1754         }
1755
1756         iucv_accept_enqueue(sk, nsk);
1757
1758         /* Wake up accept */
1759         nsk->sk_state = IUCV_CONNECTED;
1760         sk->sk_data_ready(sk, 1);
1761         err = 0;
1762 fail:
1763         bh_unlock_sock(sk);
1764         return 0;
1765 }
1766
1767 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1768 {
1769         struct sock *sk = path->private;
1770
1771         sk->sk_state = IUCV_CONNECTED;
1772         sk->sk_state_change(sk);
1773 }
1774
1775 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1776 {
1777         struct sock *sk = path->private;
1778         struct iucv_sock *iucv = iucv_sk(sk);
1779         struct sk_buff *skb;
1780         struct sock_msg_q *save_msg;
1781         int len;
1782
1783         if (sk->sk_shutdown & RCV_SHUTDOWN) {
1784                 pr_iucv->message_reject(path, msg);
1785                 return;
1786         }
1787
1788         spin_lock(&iucv->message_q.lock);
1789
1790         if (!list_empty(&iucv->message_q.list) ||
1791             !skb_queue_empty(&iucv->backlog_skb_q))
1792                 goto save_message;
1793
1794         len = atomic_read(&sk->sk_rmem_alloc);
1795         len += SKB_TRUESIZE(iucv_msg_length(msg));
1796         if (len > sk->sk_rcvbuf)
1797                 goto save_message;
1798
1799         skb = alloc_skb(iucv_msg_length(msg), GFP_ATOMIC | GFP_DMA);
1800         if (!skb)
1801                 goto save_message;
1802
1803         iucv_process_message(sk, skb, path, msg);
1804         goto out_unlock;
1805
1806 save_message:
1807         save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1808         if (!save_msg)
1809                 goto out_unlock;
1810         save_msg->path = path;
1811         save_msg->msg = *msg;
1812
1813         list_add_tail(&save_msg->list, &iucv->message_q.list);
1814
1815 out_unlock:
1816         spin_unlock(&iucv->message_q.lock);
1817 }
1818
1819 static void iucv_callback_txdone(struct iucv_path *path,
1820                                  struct iucv_message *msg)
1821 {
1822         struct sock *sk = path->private;
1823         struct sk_buff *this = NULL;
1824         struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1825         struct sk_buff *list_skb = list->next;
1826         unsigned long flags;
1827
1828         bh_lock_sock(sk);
1829         if (!skb_queue_empty(list)) {
1830                 spin_lock_irqsave(&list->lock, flags);
1831
1832                 while (list_skb != (struct sk_buff *)list) {
1833                         if (!memcmp(&msg->tag, CB_TAG(list_skb), CB_TAG_LEN)) {
1834                                 this = list_skb;
1835                                 break;
1836                         }
1837                         list_skb = list_skb->next;
1838                 }
1839                 if (this)
1840                         __skb_unlink(this, list);
1841
1842                 spin_unlock_irqrestore(&list->lock, flags);
1843
1844                 if (this) {
1845                         kfree_skb(this);
1846                         /* wake up any process waiting for sending */
1847                         iucv_sock_wake_msglim(sk);
1848                 }
1849         }
1850
1851         if (sk->sk_state == IUCV_CLOSING) {
1852                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1853                         sk->sk_state = IUCV_CLOSED;
1854                         sk->sk_state_change(sk);
1855                 }
1856         }
1857         bh_unlock_sock(sk);
1858
1859 }
1860
1861 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1862 {
1863         struct sock *sk = path->private;
1864
1865         if (sk->sk_state == IUCV_CLOSED)
1866                 return;
1867
1868         bh_lock_sock(sk);
1869         iucv_sever_path(sk, 1);
1870         sk->sk_state = IUCV_DISCONN;
1871
1872         sk->sk_state_change(sk);
1873         bh_unlock_sock(sk);
1874 }
1875
1876 /* called if the other communication side shuts down its RECV direction;
1877  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1878  */
1879 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1880 {
1881         struct sock *sk = path->private;
1882
1883         bh_lock_sock(sk);
1884         if (sk->sk_state != IUCV_CLOSED) {
1885                 sk->sk_shutdown |= SEND_SHUTDOWN;
1886                 sk->sk_state_change(sk);
1887         }
1888         bh_unlock_sock(sk);
1889 }
1890
1891 /***************** HiperSockets transport callbacks ********************/
1892 static void afiucv_swap_src_dest(struct sk_buff *skb)
1893 {
1894         struct af_iucv_trans_hdr *trans_hdr =
1895                                 (struct af_iucv_trans_hdr *)skb->data;
1896         char tmpID[8];
1897         char tmpName[8];
1898
1899         ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1900         ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1901         ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1902         ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1903         memcpy(tmpID, trans_hdr->srcUserID, 8);
1904         memcpy(tmpName, trans_hdr->srcAppName, 8);
1905         memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1906         memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1907         memcpy(trans_hdr->destUserID, tmpID, 8);
1908         memcpy(trans_hdr->destAppName, tmpName, 8);
1909         skb_push(skb, ETH_HLEN);
1910         memset(skb->data, 0, ETH_HLEN);
1911 }
1912
1913 /**
1914  * afiucv_hs_callback_syn - react on received SYN
1915  **/
1916 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1917 {
1918         struct sock *nsk;
1919         struct iucv_sock *iucv, *niucv;
1920         struct af_iucv_trans_hdr *trans_hdr;
1921         int err;
1922
1923         iucv = iucv_sk(sk);
1924         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
1925         if (!iucv) {
1926                 /* no sock - connection refused */
1927                 afiucv_swap_src_dest(skb);
1928                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1929                 err = dev_queue_xmit(skb);
1930                 goto out;
1931         }
1932
1933         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1934         bh_lock_sock(sk);
1935         if ((sk->sk_state != IUCV_LISTEN) ||
1936             sk_acceptq_is_full(sk) ||
1937             !nsk) {
1938                 /* error on server socket - connection refused */
1939                 if (nsk)
1940                         sk_free(nsk);
1941                 afiucv_swap_src_dest(skb);
1942                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1943                 err = dev_queue_xmit(skb);
1944                 bh_unlock_sock(sk);
1945                 goto out;
1946         }
1947
1948         niucv = iucv_sk(nsk);
1949         iucv_sock_init(nsk, sk);
1950         niucv->transport = AF_IUCV_TRANS_HIPER;
1951         niucv->msglimit = iucv->msglimit;
1952         if (!trans_hdr->window)
1953                 niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1954         else
1955                 niucv->msglimit_peer = trans_hdr->window;
1956         memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1957         memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1958         memcpy(niucv->src_name, iucv->src_name, 8);
1959         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1960         nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1961         niucv->hs_dev = iucv->hs_dev;
1962         dev_hold(niucv->hs_dev);
1963         afiucv_swap_src_dest(skb);
1964         trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1965         trans_hdr->window = niucv->msglimit;
1966         /* if receiver acks the xmit connection is established */
1967         err = dev_queue_xmit(skb);
1968         if (!err) {
1969                 iucv_accept_enqueue(sk, nsk);
1970                 nsk->sk_state = IUCV_CONNECTED;
1971                 sk->sk_data_ready(sk, 1);
1972         } else
1973                 iucv_sock_kill(nsk);
1974         bh_unlock_sock(sk);
1975
1976 out:
1977         return NET_RX_SUCCESS;
1978 }
1979
1980 /**
1981  * afiucv_hs_callback_synack() - react on received SYN-ACK
1982  **/
1983 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1984 {
1985         struct iucv_sock *iucv = iucv_sk(sk);
1986         struct af_iucv_trans_hdr *trans_hdr =
1987                                         (struct af_iucv_trans_hdr *)skb->data;
1988
1989         if (!iucv)
1990                 goto out;
1991         if (sk->sk_state != IUCV_BOUND)
1992                 goto out;
1993         bh_lock_sock(sk);
1994         iucv->msglimit_peer = trans_hdr->window;
1995         sk->sk_state = IUCV_CONNECTED;
1996         sk->sk_state_change(sk);
1997         bh_unlock_sock(sk);
1998 out:
1999         kfree_skb(skb);
2000         return NET_RX_SUCCESS;
2001 }
2002
2003 /**
2004  * afiucv_hs_callback_synfin() - react on received SYN_FIN
2005  **/
2006 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
2007 {
2008         struct iucv_sock *iucv = iucv_sk(sk);
2009
2010         if (!iucv)
2011                 goto out;
2012         if (sk->sk_state != IUCV_BOUND)
2013                 goto out;
2014         bh_lock_sock(sk);
2015         sk->sk_state = IUCV_DISCONN;
2016         sk->sk_state_change(sk);
2017         bh_unlock_sock(sk);
2018 out:
2019         kfree_skb(skb);
2020         return NET_RX_SUCCESS;
2021 }
2022
2023 /**
2024  * afiucv_hs_callback_fin() - react on received FIN
2025  **/
2026 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
2027 {
2028         struct iucv_sock *iucv = iucv_sk(sk);
2029
2030         /* other end of connection closed */
2031         if (!iucv)
2032                 goto out;
2033         bh_lock_sock(sk);
2034         if (sk->sk_state == IUCV_CONNECTED) {
2035                 sk->sk_state = IUCV_DISCONN;
2036                 sk->sk_state_change(sk);
2037         }
2038         bh_unlock_sock(sk);
2039 out:
2040         kfree_skb(skb);
2041         return NET_RX_SUCCESS;
2042 }
2043
2044 /**
2045  * afiucv_hs_callback_win() - react on received WIN
2046  **/
2047 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
2048 {
2049         struct iucv_sock *iucv = iucv_sk(sk);
2050         struct af_iucv_trans_hdr *trans_hdr =
2051                                         (struct af_iucv_trans_hdr *)skb->data;
2052
2053         if (!iucv)
2054                 return NET_RX_SUCCESS;
2055
2056         if (sk->sk_state != IUCV_CONNECTED)
2057                 return NET_RX_SUCCESS;
2058
2059         atomic_sub(trans_hdr->window, &iucv->msg_sent);
2060         iucv_sock_wake_msglim(sk);
2061         return NET_RX_SUCCESS;
2062 }
2063
2064 /**
2065  * afiucv_hs_callback_rx() - react on received data
2066  **/
2067 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
2068 {
2069         struct iucv_sock *iucv = iucv_sk(sk);
2070
2071         if (!iucv) {
2072                 kfree_skb(skb);
2073                 return NET_RX_SUCCESS;
2074         }
2075
2076         if (sk->sk_state != IUCV_CONNECTED) {
2077                 kfree_skb(skb);
2078                 return NET_RX_SUCCESS;
2079         }
2080
2081         if (sk->sk_shutdown & RCV_SHUTDOWN) {
2082                 kfree_skb(skb);
2083                 return NET_RX_SUCCESS;
2084         }
2085
2086                 /* write stuff from iucv_msg to skb cb */
2087         if (skb->len < sizeof(struct af_iucv_trans_hdr)) {
2088                 kfree_skb(skb);
2089                 return NET_RX_SUCCESS;
2090         }
2091         skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2092         skb_reset_transport_header(skb);
2093         skb_reset_network_header(skb);
2094         spin_lock(&iucv->message_q.lock);
2095         if (skb_queue_empty(&iucv->backlog_skb_q)) {
2096                 if (sock_queue_rcv_skb(sk, skb)) {
2097                         /* handle rcv queue full */
2098                         skb_queue_tail(&iucv->backlog_skb_q, skb);
2099                 }
2100         } else
2101                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2102         spin_unlock(&iucv->message_q.lock);
2103         return NET_RX_SUCCESS;
2104 }
2105
2106 /**
2107  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2108  *                   transport
2109  *                   called from netif RX softirq
2110  **/
2111 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2112         struct packet_type *pt, struct net_device *orig_dev)
2113 {
2114         struct sock *sk;
2115         struct iucv_sock *iucv;
2116         struct af_iucv_trans_hdr *trans_hdr;
2117         char nullstring[8];
2118         int err = 0;
2119
2120         skb_pull(skb, ETH_HLEN);
2121         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
2122         EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2123         EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2124         EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2125         EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2126         memset(nullstring, 0, sizeof(nullstring));
2127         iucv = NULL;
2128         sk = NULL;
2129         read_lock(&iucv_sk_list.lock);
2130         sk_for_each(sk, &iucv_sk_list.head) {
2131                 if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2132                         if ((!memcmp(&iucv_sk(sk)->src_name,
2133                                      trans_hdr->destAppName, 8)) &&
2134                             (!memcmp(&iucv_sk(sk)->src_user_id,
2135                                      trans_hdr->destUserID, 8)) &&
2136                             (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2137                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2138                                      nullstring, 8))) {
2139                                 iucv = iucv_sk(sk);
2140                                 break;
2141                         }
2142                 } else {
2143                         if ((!memcmp(&iucv_sk(sk)->src_name,
2144                                      trans_hdr->destAppName, 8)) &&
2145                             (!memcmp(&iucv_sk(sk)->src_user_id,
2146                                      trans_hdr->destUserID, 8)) &&
2147                             (!memcmp(&iucv_sk(sk)->dst_name,
2148                                      trans_hdr->srcAppName, 8)) &&
2149                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2150                                      trans_hdr->srcUserID, 8))) {
2151                                 iucv = iucv_sk(sk);
2152                                 break;
2153                         }
2154                 }
2155         }
2156         read_unlock(&iucv_sk_list.lock);
2157         if (!iucv)
2158                 sk = NULL;
2159
2160         /* no sock
2161         how should we send with no sock
2162         1) send without sock no send rc checking?
2163         2) introduce default sock to handle this cases
2164
2165          SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2166          data -> send FIN
2167          SYN|ACK, SYN|FIN, FIN -> no action? */
2168
2169         switch (trans_hdr->flags) {
2170         case AF_IUCV_FLAG_SYN:
2171                 /* connect request */
2172                 err = afiucv_hs_callback_syn(sk, skb);
2173                 break;
2174         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2175                 /* connect request confirmed */
2176                 err = afiucv_hs_callback_synack(sk, skb);
2177                 break;
2178         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2179                 /* connect request refused */
2180                 err = afiucv_hs_callback_synfin(sk, skb);
2181                 break;
2182         case (AF_IUCV_FLAG_FIN):
2183                 /* close request */
2184                 err = afiucv_hs_callback_fin(sk, skb);
2185                 break;
2186         case (AF_IUCV_FLAG_WIN):
2187                 err = afiucv_hs_callback_win(sk, skb);
2188                 if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2189                         kfree_skb(skb);
2190                         break;
2191                 }
2192                 /* fall through and receive non-zero length data */
2193         case (AF_IUCV_FLAG_SHT):
2194                 /* shutdown request */
2195                 /* fall through and receive zero length data */
2196         case 0:
2197                 /* plain data frame */
2198                 memcpy(CB_TRGCLS(skb), &trans_hdr->iucv_hdr.class,
2199                        CB_TRGCLS_LEN);
2200                 err = afiucv_hs_callback_rx(sk, skb);
2201                 break;
2202         default:
2203                 ;
2204         }
2205
2206         return err;
2207 }
2208
2209 /**
2210  * afiucv_hs_callback_txnotify() - handle send notifcations from HiperSockets
2211  *                                 transport
2212  **/
2213 static void afiucv_hs_callback_txnotify(struct sk_buff *skb,
2214                                         enum iucv_tx_notify n)
2215 {
2216         struct sock *isk = skb->sk;
2217         struct sock *sk = NULL;
2218         struct iucv_sock *iucv = NULL;
2219         struct sk_buff_head *list;
2220         struct sk_buff *list_skb;
2221         struct sk_buff *nskb;
2222         unsigned long flags;
2223
2224         read_lock_irqsave(&iucv_sk_list.lock, flags);
2225         sk_for_each(sk, &iucv_sk_list.head)
2226                 if (sk == isk) {
2227                         iucv = iucv_sk(sk);
2228                         break;
2229                 }
2230         read_unlock_irqrestore(&iucv_sk_list.lock, flags);
2231
2232         if (!iucv || sock_flag(sk, SOCK_ZAPPED))
2233                 return;
2234
2235         list = &iucv->send_skb_q;
2236         spin_lock_irqsave(&list->lock, flags);
2237         if (skb_queue_empty(list))
2238                 goto out_unlock;
2239         list_skb = list->next;
2240         nskb = list_skb->next;
2241         while (list_skb != (struct sk_buff *)list) {
2242                 if (skb_shinfo(list_skb) == skb_shinfo(skb)) {
2243                         switch (n) {
2244                         case TX_NOTIFY_OK:
2245                                 __skb_unlink(list_skb, list);
2246                                 kfree_skb(list_skb);
2247                                 iucv_sock_wake_msglim(sk);
2248                                 break;
2249                         case TX_NOTIFY_PENDING:
2250                                 atomic_inc(&iucv->pendings);
2251                                 break;
2252                         case TX_NOTIFY_DELAYED_OK:
2253                                 __skb_unlink(list_skb, list);
2254                                 atomic_dec(&iucv->pendings);
2255                                 if (atomic_read(&iucv->pendings) <= 0)
2256                                         iucv_sock_wake_msglim(sk);
2257                                 kfree_skb(list_skb);
2258                                 break;
2259                         case TX_NOTIFY_UNREACHABLE:
2260                         case TX_NOTIFY_DELAYED_UNREACHABLE:
2261                         case TX_NOTIFY_TPQFULL: /* not yet used */
2262                         case TX_NOTIFY_GENERALERROR:
2263                         case TX_NOTIFY_DELAYED_GENERALERROR:
2264                                 __skb_unlink(list_skb, list);
2265                                 kfree_skb(list_skb);
2266                                 if (sk->sk_state == IUCV_CONNECTED) {
2267                                         sk->sk_state = IUCV_DISCONN;
2268                                         sk->sk_state_change(sk);
2269                                 }
2270                                 break;
2271                         }
2272                         break;
2273                 }
2274                 list_skb = nskb;
2275                 nskb = nskb->next;
2276         }
2277 out_unlock:
2278         spin_unlock_irqrestore(&list->lock, flags);
2279
2280         if (sk->sk_state == IUCV_CLOSING) {
2281                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
2282                         sk->sk_state = IUCV_CLOSED;
2283                         sk->sk_state_change(sk);
2284                 }
2285         }
2286
2287 }
2288
2289 /*
2290  * afiucv_netdev_event: handle netdev notifier chain events
2291  */
2292 static int afiucv_netdev_event(struct notifier_block *this,
2293                                unsigned long event, void *ptr)
2294 {
2295         struct net_device *event_dev = (struct net_device *)ptr;
2296         struct sock *sk;
2297         struct iucv_sock *iucv;
2298
2299         switch (event) {
2300         case NETDEV_REBOOT:
2301         case NETDEV_GOING_DOWN:
2302                 sk_for_each(sk, &iucv_sk_list.head) {
2303                         iucv = iucv_sk(sk);
2304                         if ((iucv->hs_dev == event_dev) &&
2305                             (sk->sk_state == IUCV_CONNECTED)) {
2306                                 if (event == NETDEV_GOING_DOWN)
2307                                         iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2308                                 sk->sk_state = IUCV_DISCONN;
2309                                 sk->sk_state_change(sk);
2310                         }
2311                 }
2312                 break;
2313         case NETDEV_DOWN:
2314         case NETDEV_UNREGISTER:
2315         default:
2316                 break;
2317         }
2318         return NOTIFY_DONE;
2319 }
2320
2321 static struct notifier_block afiucv_netdev_notifier = {
2322         .notifier_call = afiucv_netdev_event,
2323 };
2324
2325 static const struct proto_ops iucv_sock_ops = {
2326         .family         = PF_IUCV,
2327         .owner          = THIS_MODULE,
2328         .release        = iucv_sock_release,
2329         .bind           = iucv_sock_bind,
2330         .connect        = iucv_sock_connect,
2331         .listen         = iucv_sock_listen,
2332         .accept         = iucv_sock_accept,
2333         .getname        = iucv_sock_getname,
2334         .sendmsg        = iucv_sock_sendmsg,
2335         .recvmsg        = iucv_sock_recvmsg,
2336         .poll           = iucv_sock_poll,
2337         .ioctl          = sock_no_ioctl,
2338         .mmap           = sock_no_mmap,
2339         .socketpair     = sock_no_socketpair,
2340         .shutdown       = iucv_sock_shutdown,
2341         .setsockopt     = iucv_sock_setsockopt,
2342         .getsockopt     = iucv_sock_getsockopt,
2343 };
2344
2345 static const struct net_proto_family iucv_sock_family_ops = {
2346         .family = AF_IUCV,
2347         .owner  = THIS_MODULE,
2348         .create = iucv_sock_create,
2349 };
2350
2351 static struct packet_type iucv_packet_type = {
2352         .type = cpu_to_be16(ETH_P_AF_IUCV),
2353         .func = afiucv_hs_rcv,
2354 };
2355
2356 static int afiucv_iucv_init(void)
2357 {
2358         int err;
2359
2360         err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2361         if (err)
2362                 goto out;
2363         /* establish dummy device */
2364         af_iucv_driver.bus = pr_iucv->bus;
2365         err = driver_register(&af_iucv_driver);
2366         if (err)
2367                 goto out_iucv;
2368         af_iucv_dev = kzalloc(sizeof(struct device), GFP_KERNEL);
2369         if (!af_iucv_dev) {
2370                 err = -ENOMEM;
2371                 goto out_driver;
2372         }
2373         dev_set_name(af_iucv_dev, "af_iucv");
2374         af_iucv_dev->bus = pr_iucv->bus;
2375         af_iucv_dev->parent = pr_iucv->root;
2376         af_iucv_dev->release = (void (*)(struct device *))kfree;
2377         af_iucv_dev->driver = &af_iucv_driver;
2378         err = device_register(af_iucv_dev);
2379         if (err)
2380                 goto out_driver;
2381         return 0;
2382
2383 out_driver:
2384         driver_unregister(&af_iucv_driver);
2385 out_iucv:
2386         pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2387 out:
2388         return err;
2389 }
2390
2391 static int __init afiucv_init(void)
2392 {
2393         int err;
2394
2395         if (MACHINE_IS_VM) {
2396                 cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2397                 if (unlikely(err)) {
2398                         WARN_ON(err);
2399                         err = -EPROTONOSUPPORT;
2400                         goto out;
2401                 }
2402
2403                 pr_iucv = try_then_request_module(symbol_get(iucv_if), "iucv");
2404                 if (!pr_iucv) {
2405                         printk(KERN_WARNING "iucv_if lookup failed\n");
2406                         memset(&iucv_userid, 0, sizeof(iucv_userid));
2407                 }
2408         } else {
2409                 memset(&iucv_userid, 0, sizeof(iucv_userid));
2410                 pr_iucv = NULL;
2411         }
2412
2413         err = proto_register(&iucv_proto, 0);
2414         if (err)
2415                 goto out;
2416         err = sock_register(&iucv_sock_family_ops);
2417         if (err)
2418                 goto out_proto;
2419
2420         if (pr_iucv) {
2421                 err = afiucv_iucv_init();
2422                 if (err)
2423                         goto out_sock;
2424         } else
2425                 register_netdevice_notifier(&afiucv_netdev_notifier);
2426         dev_add_pack(&iucv_packet_type);
2427         return 0;
2428
2429 out_sock:
2430         sock_unregister(PF_IUCV);
2431 out_proto:
2432         proto_unregister(&iucv_proto);
2433 out:
2434         if (pr_iucv)
2435                 symbol_put(iucv_if);
2436         return err;
2437 }
2438
2439 static void __exit afiucv_exit(void)
2440 {
2441         if (pr_iucv) {
2442                 device_unregister(af_iucv_dev);
2443                 driver_unregister(&af_iucv_driver);
2444                 pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2445                 symbol_put(iucv_if);
2446         } else
2447                 unregister_netdevice_notifier(&afiucv_netdev_notifier);
2448         dev_remove_pack(&iucv_packet_type);
2449         sock_unregister(PF_IUCV);
2450         proto_unregister(&iucv_proto);
2451 }
2452
2453 module_init(afiucv_init);
2454 module_exit(afiucv_exit);
2455
2456 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2457 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2458 MODULE_VERSION(VERSION);
2459 MODULE_LICENSE("GPL");
2460 MODULE_ALIAS_NETPROTO(PF_IUCV);
2461