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 #define __iucv_sock_wait(sk, condition, timeo, ret)                     \
  53 do {                                                                    \
  54         DEFINE_WAIT(__wait);                                            \
  55         long __timeo = timeo;                                           \
  56         ret = 0;                                                        \
  57         prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);     \
  58         while (!(condition)) {                                          \
  59                 if (!__timeo) {                                         \
  60                         ret = -EAGAIN;                                  \
  61                         break;                                          \
  62                 }                                                       \
  63                 if (signal_pending(current)) {                          \
  64                         ret = sock_intr_errno(__timeo);                 \
  65                         break;                                          \
  66                 }                                                       \
  67                 release_sock(sk);                                       \
  68                 __timeo = schedule_timeout(__timeo);                    \
  69                 lock_sock(sk);                                          \
  70                 ret = sock_error(sk);                                   \
  71                 if (ret)                                                \
  72                         break;                                          \
  73         }                                                               \
  74         finish_wait(sk_sleep(sk), &__wait);                             \
  75 } while (0)
  76
  77 #define iucv_sock_wait(sk, condition, timeo)                            \
  78 ({                                                                      \
  79         int __ret = 0;                                                  \
  80         if (!(condition))                                               \
  81                 __iucv_sock_wait(sk, condition, timeo, __ret);          \
  82         __ret;                                                          \
  83 })
  84
  85 static void iucv_sock_kill(struct sock *sk);
  86 static void iucv_sock_close(struct sock *sk);
  87 static void iucv_sever_path(struct sock *, int);
  88
  89 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
  90         struct packet_type *pt, struct net_device *orig_dev);
  91 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
  92                    struct sk_buff *skb, u8 flags);
  93 static void afiucv_hs_callback_txnotify(struct sk_buff *, enum iucv_tx_notify);
  94
  95 /* Call Back functions */
  96 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
  97 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
  98 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
  99 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
 100                                  u8 ipuser[16]);
 101 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
 102 static void iucv_callback_shutdown(struct iucv_path *, u8 ipuser[16]);
 103
 104 static struct iucv_sock_list iucv_sk_list = {
 105         .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
 106         .autobind_name = ATOMIC_INIT(0)
 107 };
 108
 109 static struct iucv_handler af_iucv_handler = {
 110         .path_pending     = iucv_callback_connreq,
 111         .path_complete    = iucv_callback_connack,
 112         .path_severed     = iucv_callback_connrej,
 113         .message_pending  = iucv_callback_rx,
 114         .message_complete = iucv_callback_txdone,
 115         .path_quiesced    = iucv_callback_shutdown,
 116 };
 117
 118 static inline void high_nmcpy(unsigned char *dst, char *src)
 119 {
 120        memcpy(dst, src, 8);
 121 }
 122
 123 static inline void low_nmcpy(unsigned char *dst, char *src)
 124 {
 125        memcpy(&dst[8], src, 8);
 126 }
 127
 128 static int afiucv_pm_prepare(struct device *dev)
 129 {
 130 #ifdef CONFIG_PM_DEBUG
 131         printk(KERN_WARNING "afiucv_pm_prepare\n");
 132 #endif
 133         return 0;
 134 }
 135
 136 static void afiucv_pm_complete(struct device *dev)
 137 {
 138 #ifdef CONFIG_PM_DEBUG
 139         printk(KERN_WARNING "afiucv_pm_complete\n");
 140 #endif
 141 }
 142
 143 /**
 144  * afiucv_pm_freeze() - Freeze PM callback
 145  * @dev:        AFIUCV dummy device
 146  *
 147  * Sever all established IUCV communication pathes
 148  */
 149 static int afiucv_pm_freeze(struct device *dev)
 150 {
 151         struct iucv_sock *iucv;
 152         struct sock *sk;
 153         int err = 0;
 154
 155 #ifdef CONFIG_PM_DEBUG
 156         printk(KERN_WARNING "afiucv_pm_freeze\n");
 157 #endif
 158         read_lock(&iucv_sk_list.lock);
 159         sk_for_each(sk, &iucv_sk_list.head) {
 160                 iucv = iucv_sk(sk);
 161                 switch (sk->sk_state) {
 162                 case IUCV_DISCONN:
 163                 case IUCV_CLOSING:
 164                 case IUCV_CONNECTED:
 165                         iucv_sever_path(sk, 0);
 166                         break;
 167                 case IUCV_OPEN:
 168                 case IUCV_BOUND:
 169                 case IUCV_LISTEN:
 170                 case IUCV_CLOSED:
 171                 default:
 172                         break;
 173                 }
 174                 skb_queue_purge(&iucv->send_skb_q);
 175                 skb_queue_purge(&iucv->backlog_skb_q);
 176         }
 177         read_unlock(&iucv_sk_list.lock);
 178         return err;
 179 }
 180
 181 /**
 182  * afiucv_pm_restore_thaw() - Thaw and restore PM callback
 183  * @dev:        AFIUCV dummy device
 184  *
 185  * socket clean up after freeze
 186  */
 187 static int afiucv_pm_restore_thaw(struct device *dev)
 188 {
 189         struct sock *sk;
 190
 191 #ifdef CONFIG_PM_DEBUG
 192         printk(KERN_WARNING "afiucv_pm_restore_thaw\n");
 193 #endif
 194         read_lock(&iucv_sk_list.lock);
 195         sk_for_each(sk, &iucv_sk_list.head) {
 196                 switch (sk->sk_state) {
 197                 case IUCV_CONNECTED:
 198                         sk->sk_err = EPIPE;
 199                         sk->sk_state = IUCV_DISCONN;
 200                         sk->sk_state_change(sk);
 201                         break;
 202                 case IUCV_DISCONN:
 203                 case IUCV_CLOSING:
 204                 case IUCV_LISTEN:
 205                 case IUCV_BOUND:
 206                 case IUCV_OPEN:
 207                 default:
 208                         break;
 209                 }
 210         }
 211         read_unlock(&iucv_sk_list.lock);
 212         return 0;
 213 }
 214
 215 static const struct dev_pm_ops afiucv_pm_ops = {
 216         .prepare = afiucv_pm_prepare,
 217         .complete = afiucv_pm_complete,
 218         .freeze = afiucv_pm_freeze,
 219         .thaw = afiucv_pm_restore_thaw,
 220         .restore = afiucv_pm_restore_thaw,
 221 };
 222
 223 static struct device_driver af_iucv_driver = {
 224         .owner = THIS_MODULE,
 225         .name = "afiucv",
 226         .bus  = NULL,
 227         .pm   = &afiucv_pm_ops,
 228 };
 229
 230 /* dummy device used as trigger for PM functions */
 231 static struct device *af_iucv_dev;
 232
 233 /**
 234  * iucv_msg_length() - Returns the length of an iucv message.
 235  * @msg:        Pointer to struct iucv_message, MUST NOT be NULL
 236  *
 237  * The function returns the length of the specified iucv message @msg of data
 238  * stored in a buffer and of data stored in the parameter list (PRMDATA).
 239  *
 240  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
 241  * data:
 242  *      PRMDATA[0..6]   socket data (max 7 bytes);
 243  *      PRMDATA[7]      socket data length value (len is 0xff - PRMDATA[7])
 244  *
 245  * The socket data length is computed by subtracting the socket data length
 246  * value from 0xFF.
 247  * If the socket data len is greater 7, then PRMDATA can be used for special
 248  * notifications (see iucv_sock_shutdown); and further,
 249  * if the socket data len is > 7, the function returns 8.
 250  *
 251  * Use this function to allocate socket buffers to store iucv message data.
 252  */
 253 static inline size_t iucv_msg_length(struct iucv_message *msg)
 254 {
 255         size_t datalen;
 256
 257         if (msg->flags & IUCV_IPRMDATA) {
 258                 datalen = 0xff - msg->rmmsg[7];
 259                 return (datalen < 8) ? datalen : 8;
 260         }
 261         return msg->length;
 262 }
 263
 264 /**
 265  * iucv_sock_in_state() - check for specific states
 266  * @sk:         sock structure
 267  * @state:      first iucv sk state
 268  * @state:      second iucv sk state
 269  *
 270  * Returns true if the socket in either in the first or second state.
 271  */
 272 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
 273 {
 274         return (sk->sk_state == state || sk->sk_state == state2);
 275 }
 276
 277 /**
 278  * iucv_below_msglim() - function to check if messages can be sent
 279  * @sk:         sock structure
 280  *
 281  * Returns true if the send queue length is lower than the message limit.
 282  * Always returns true if the socket is not connected (no iucv path for
 283  * checking the message limit).
 284  */
 285 static inline int iucv_below_msglim(struct sock *sk)
 286 {
 287         struct iucv_sock *iucv = iucv_sk(sk);
 288
 289         if (sk->sk_state != IUCV_CONNECTED)
 290                 return 1;
 291         if (iucv->transport == AF_IUCV_TRANS_IUCV)
 292                 return (skb_queue_len(&iucv->send_skb_q) < iucv->path->msglim);
 293         else
 294                 return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
 295                         (atomic_read(&iucv->pendings) <= 0));
 296 }
 297
 298 /**
 299  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
 300  */
 301 static void iucv_sock_wake_msglim(struct sock *sk)
 302 {
 303         struct socket_wq *wq;
 304
 305         rcu_read_lock();
 306         wq = rcu_dereference(sk->sk_wq);
 307         if (wq_has_sleeper(wq))
 308                 wake_up_interruptible_all(&wq->wait);
 309         sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
 310         rcu_read_unlock();
 311 }
 312
 313 /**
 314  * afiucv_hs_send() - send a message through HiperSockets transport
 315  */
 316 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
 317                    struct sk_buff *skb, u8 flags)
 318 {
 319         struct iucv_sock *iucv = iucv_sk(sock);
 320         struct af_iucv_trans_hdr *phs_hdr;
 321         struct sk_buff *nskb;
 322         int err, confirm_recv = 0;
 323
 324         memset(skb->head, 0, ETH_HLEN);
 325         phs_hdr = (struct af_iucv_trans_hdr *)skb_push(skb,
 326                                         sizeof(struct af_iucv_trans_hdr));
 327         skb_reset_mac_header(skb);
 328         skb_reset_network_header(skb);
 329         skb_push(skb, ETH_HLEN);
 330         skb_reset_mac_header(skb);
 331         memset(phs_hdr, 0, sizeof(struct af_iucv_trans_hdr));
 332
 333         phs_hdr->magic = ETH_P_AF_IUCV;
 334         phs_hdr->version = 1;
 335         phs_hdr->flags = flags;
 336         if (flags == AF_IUCV_FLAG_SYN)
 337                 phs_hdr->window = iucv->msglimit;
 338         else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
 339                 confirm_recv = atomic_read(&iucv->msg_recv);
 340                 phs_hdr->window = confirm_recv;
 341                 if (confirm_recv)
 342                         phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
 343         }
 344         memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
 345         memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
 346         memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
 347         memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
 348         ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
 349         ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
 350         ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
 351         ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
 352         if (imsg)
 353                 memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
 354
 355         skb->dev = iucv->hs_dev;
 356         if (!skb->dev)
 357                 return -ENODEV;
 358         if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev))
 359                 return -ENETDOWN;
 360         if (skb->len > skb->dev->mtu) {
 361                 if (sock->sk_type == SOCK_SEQPACKET)
 362                         return -EMSGSIZE;
 363                 else
 364                         skb_trim(skb, skb->dev->mtu);
 365         }
 366         skb->protocol = ETH_P_AF_IUCV;
 367         nskb = skb_clone(skb, GFP_ATOMIC);
 368         if (!nskb)
 369                 return -ENOMEM;
 370         skb_queue_tail(&iucv->send_skb_q, nskb);
 371         err = dev_queue_xmit(skb);
 372         if (net_xmit_eval(err)) {
 373                 skb_unlink(nskb, &iucv->send_skb_q);
 374                 kfree_skb(nskb);
 375         } else {
 376                 atomic_sub(confirm_recv, &iucv->msg_recv);
 377                 WARN_ON(atomic_read(&iucv->msg_recv) < 0);
 378         }
 379         return net_xmit_eval(err);
 380 }
 381
 382 static struct sock *__iucv_get_sock_by_name(char *nm)
 383 {
 384         struct sock *sk;
 385
 386         sk_for_each(sk, &iucv_sk_list.head)
 387                 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
 388                         return sk;
 389
 390         return NULL;
 391 }
 392
 393 static void iucv_sock_destruct(struct sock *sk)
 394 {
 395         skb_queue_purge(&sk->sk_receive_queue);
 396         skb_queue_purge(&sk->sk_error_queue);
 397
 398         sk_mem_reclaim(sk);
 399
 400         if (!sock_flag(sk, SOCK_DEAD)) {
 401                 pr_err("Attempt to release alive iucv socket %p\n", sk);
 402                 return;
 403         }
 404
 405         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
 406         WARN_ON(atomic_read(&sk->sk_wmem_alloc));
 407         WARN_ON(sk->sk_wmem_queued);
 408         WARN_ON(sk->sk_forward_alloc);
 409 }
 410
 411 /* Cleanup Listen */
 412 static void iucv_sock_cleanup_listen(struct sock *parent)
 413 {
 414         struct sock *sk;
 415
 416         /* Close non-accepted connections */
 417         while ((sk = iucv_accept_dequeue(parent, NULL))) {
 418                 iucv_sock_close(sk);
 419                 iucv_sock_kill(sk);
 420         }
 421
 422         parent->sk_state = IUCV_CLOSED;
 423 }
 424
 425 /* Kill socket (only if zapped and orphaned) */
 426 static void iucv_sock_kill(struct sock *sk)
 427 {
 428         if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
 429                 return;
 430
 431         iucv_sock_unlink(&iucv_sk_list, sk);
 432         sock_set_flag(sk, SOCK_DEAD);
 433         sock_put(sk);
 434 }
 435
 436 /* Terminate an IUCV path */
 437 static void iucv_sever_path(struct sock *sk, int with_user_data)
 438 {
 439         unsigned char user_data[16];
 440         struct iucv_sock *iucv = iucv_sk(sk);
 441         struct iucv_path *path = iucv->path;
 442
 443         if (iucv->path) {
 444                 iucv->path = NULL;
 445                 if (with_user_data) {
 446                         low_nmcpy(user_data, iucv->src_name);
 447                         high_nmcpy(user_data, iucv->dst_name);
 448                         ASCEBC(user_data, sizeof(user_data));
 449                         pr_iucv->path_sever(path, user_data);
 450                 } else
 451                         pr_iucv->path_sever(path, NULL);
 452                 iucv_path_free(path);
 453         }
 454 }
 455
 456 /* Send FIN through an IUCV socket for HIPER transport */
 457 static int iucv_send_ctrl(struct sock *sk, u8 flags)
 458 {
 459         int err = 0;
 460         int blen;
 461         struct sk_buff *skb;
 462
 463         blen = sizeof(struct af_iucv_trans_hdr) + ETH_HLEN;
 464         skb = sock_alloc_send_skb(sk, blen, 1, &err);
 465         if (skb) {
 466                 skb_reserve(skb, blen);
 467                 err = afiucv_hs_send(NULL, sk, skb, flags);
 468         }
 469         return err;
 470 }
 471
 472 /* Close an IUCV socket */
 473 static void iucv_sock_close(struct sock *sk)
 474 {
 475         struct iucv_sock *iucv = iucv_sk(sk);
 476         unsigned long timeo;
 477         int err = 0;
 478
 479         lock_sock(sk);
 480
 481         switch (sk->sk_state) {
 482         case IUCV_LISTEN:
 483                 iucv_sock_cleanup_listen(sk);
 484                 break;
 485
 486         case IUCV_CONNECTED:
 487                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
 488                         err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
 489                         sk->sk_state = IUCV_DISCONN;
 490                         sk->sk_state_change(sk);
 491                 }
 492         case IUCV_DISCONN:   /* fall through */
 493                 sk->sk_state = IUCV_CLOSING;
 494                 sk->sk_state_change(sk);
 495
 496                 if (!err && !skb_queue_empty(&iucv->send_skb_q)) {
 497                         if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
 498                                 timeo = sk->sk_lingertime;
 499                         else
 500                                 timeo = IUCV_DISCONN_TIMEOUT;
 501                         iucv_sock_wait(sk,
 502                                         iucv_sock_in_state(sk, IUCV_CLOSED, 0),
 503                                         timeo);
 504                 }
 505
 506         case IUCV_CLOSING:   /* fall through */
 507                 sk->sk_state = IUCV_CLOSED;
 508                 sk->sk_state_change(sk);
 509
 510                 sk->sk_err = ECONNRESET;
 511                 sk->sk_state_change(sk);
 512
 513                 skb_queue_purge(&iucv->send_skb_q);
 514                 skb_queue_purge(&iucv->backlog_skb_q);
 515
 516         default:   /* fall through */
 517                 iucv_sever_path(sk, 1);
 518         }
 519
 520         if (iucv->hs_dev) {
 521                 dev_put(iucv->hs_dev);
 522                 iucv->hs_dev = NULL;
 523                 sk->sk_bound_dev_if = 0;
 524         }
 525
 526         /* mark socket for deletion by iucv_sock_kill() */
 527         sock_set_flag(sk, SOCK_ZAPPED);
 528
 529         release_sock(sk);
 530 }
 531
 532 static void iucv_sock_init(struct sock *sk, struct sock *parent)
 533 {
 534         if (parent)
 535                 sk->sk_type = parent->sk_type;
 536 }
 537
 538 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
 539 {
 540         struct sock *sk;
 541         struct iucv_sock *iucv;
 542
 543         sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
 544         if (!sk)
 545                 return NULL;
 546         iucv = iucv_sk(sk);
 547
 548         sock_init_data(sock, sk);
 549         INIT_LIST_HEAD(&iucv->accept_q);
 550         spin_lock_init(&iucv->accept_q_lock);
 551         skb_queue_head_init(&iucv->send_skb_q);
 552         INIT_LIST_HEAD(&iucv->message_q.list);
 553         spin_lock_init(&iucv->message_q.lock);
 554         skb_queue_head_init(&iucv->backlog_skb_q);
 555         iucv->send_tag = 0;
 556         atomic_set(&iucv->pendings, 0);
 557         iucv->flags = 0;
 558         iucv->msglimit = 0;
 559         atomic_set(&iucv->msg_sent, 0);
 560         atomic_set(&iucv->msg_recv, 0);
 561         iucv->path = NULL;
 562         iucv->sk_txnotify = afiucv_hs_callback_txnotify;
 563         memset(&iucv->src_user_id , 0, 32);
 564         if (pr_iucv)
 565                 iucv->transport = AF_IUCV_TRANS_IUCV;
 566         else
 567                 iucv->transport = AF_IUCV_TRANS_HIPER;
 568
 569         sk->sk_destruct = iucv_sock_destruct;
 570         sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
 571         sk->sk_allocation = GFP_DMA;
 572
 573         sock_reset_flag(sk, SOCK_ZAPPED);
 574
 575         sk->sk_protocol = proto;
 576         sk->sk_state    = IUCV_OPEN;
 577
 578         iucv_sock_link(&iucv_sk_list, sk);
 579         return sk;
 580 }
 581
 582 /* Create an IUCV socket */
 583 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
 584                             int kern)
 585 {
 586         struct sock *sk;
 587
 588         if (protocol && protocol != PF_IUCV)
 589                 return -EPROTONOSUPPORT;
 590
 591         sock->state = SS_UNCONNECTED;
 592
 593         switch (sock->type) {
 594         case SOCK_STREAM:
 595                 sock->ops = &iucv_sock_ops;
 596                 break;
 597         case SOCK_SEQPACKET:
 598                 /* currently, proto ops can handle both sk types */
 599                 sock->ops = &iucv_sock_ops;
 600                 break;
 601         default:
 602                 return -ESOCKTNOSUPPORT;
 603         }
 604
 605         sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
 606         if (!sk)
 607                 return -ENOMEM;
 608
 609         iucv_sock_init(sk, NULL);
 610
 611         return 0;
 612 }
 613
 614 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
 615 {
 616         write_lock_bh(&l->lock);
 617         sk_add_node(sk, &l->head);
 618         write_unlock_bh(&l->lock);
 619 }
 620
 621 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
 622 {
 623         write_lock_bh(&l->lock);
 624         sk_del_node_init(sk);
 625         write_unlock_bh(&l->lock);
 626 }
 627
 628 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
 629 {
 630         unsigned long flags;
 631         struct iucv_sock *par = iucv_sk(parent);
 632
 633         sock_hold(sk);
 634         spin_lock_irqsave(&par->accept_q_lock, flags);
 635         list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
 636         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 637         iucv_sk(sk)->parent = parent;
 638         sk_acceptq_added(parent);
 639 }
 640
 641 void iucv_accept_unlink(struct sock *sk)
 642 {
 643         unsigned long flags;
 644         struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
 645
 646         spin_lock_irqsave(&par->accept_q_lock, flags);
 647         list_del_init(&iucv_sk(sk)->accept_q);
 648         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 649         sk_acceptq_removed(iucv_sk(sk)->parent);
 650         iucv_sk(sk)->parent = NULL;
 651         sock_put(sk);
 652 }
 653
 654 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
 655 {
 656         struct iucv_sock *isk, *n;
 657         struct sock *sk;
 658
 659         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
 660                 sk = (struct sock *) isk;
 661                 lock_sock(sk);
 662
 663                 if (sk->sk_state == IUCV_CLOSED) {
 664                         iucv_accept_unlink(sk);
 665                         release_sock(sk);
 666                         continue;
 667                 }
 668
 669                 if (sk->sk_state == IUCV_CONNECTED ||
 670                     sk->sk_state == IUCV_DISCONN ||
 671                     !newsock) {
 672                         iucv_accept_unlink(sk);
 673                         if (newsock)
 674                                 sock_graft(sk, newsock);
 675
 676                         release_sock(sk);
 677                         return sk;
 678                 }
 679
 680                 release_sock(sk);
 681         }
 682         return NULL;
 683 }
 684
 685 /* Bind an unbound socket */
 686 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
 687                           int addr_len)
 688 {
 689         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 690         struct sock *sk = sock->sk;
 691         struct iucv_sock *iucv;
 692         int err = 0;
 693         struct net_device *dev;
 694         char uid[9];
 695
 696         /* Verify the input sockaddr */
 697         if (!addr || addr->sa_family != AF_IUCV)
 698                 return -EINVAL;
 699
 700         lock_sock(sk);
 701         if (sk->sk_state != IUCV_OPEN) {
 702                 err = -EBADFD;
 703                 goto done;
 704         }
 705
 706         write_lock_bh(&iucv_sk_list.lock);
 707
 708         iucv = iucv_sk(sk);
 709         if (__iucv_get_sock_by_name(sa->siucv_name)) {
 710                 err = -EADDRINUSE;
 711                 goto done_unlock;
 712         }
 713         if (iucv->path)
 714                 goto done_unlock;
 715
 716         /* Bind the socket */
 717         if (pr_iucv)
 718                 if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
 719                         goto vm_bind; /* VM IUCV transport */
 720
 721         /* try hiper transport */
 722         memcpy(uid, sa->siucv_user_id, sizeof(uid));
 723         ASCEBC(uid, 8);
 724         rcu_read_lock();
 725         for_each_netdev_rcu(&init_net, dev) {
 726                 if (!memcmp(dev->perm_addr, uid, 8)) {
 727                         memcpy(iucv->src_name, sa->siucv_name, 8);
 728                         memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
 729                         sk->sk_bound_dev_if = dev->ifindex;
 730                         iucv->hs_dev = dev;
 731                         dev_hold(dev);
 732                         sk->sk_state = IUCV_BOUND;
 733                         iucv->transport = AF_IUCV_TRANS_HIPER;
 734                         if (!iucv->msglimit)
 735                                 iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
 736                         rcu_read_unlock();
 737                         goto done_unlock;
 738                 }
 739         }
 740         rcu_read_unlock();
 741 vm_bind:
 742         if (pr_iucv) {
 743                 /* use local userid for backward compat */
 744                 memcpy(iucv->src_name, sa->siucv_name, 8);
 745                 memcpy(iucv->src_user_id, iucv_userid, 8);
 746                 sk->sk_state = IUCV_BOUND;
 747                 iucv->transport = AF_IUCV_TRANS_IUCV;
 748                 if (!iucv->msglimit)
 749                         iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 750                 goto done_unlock;
 751         }
 752         /* found no dev to bind */
 753         err = -ENODEV;
 754 done_unlock:
 755         /* Release the socket list lock */
 756         write_unlock_bh(&iucv_sk_list.lock);
 757 done:
 758         release_sock(sk);
 759         return err;
 760 }
 761
 762 /* Automatically bind an unbound socket */
 763 static int iucv_sock_autobind(struct sock *sk)
 764 {
 765         struct iucv_sock *iucv = iucv_sk(sk);
 766         char name[12];
 767         int err = 0;
 768
 769         if (unlikely(!pr_iucv))
 770                 return -EPROTO;
 771
 772         memcpy(iucv->src_user_id, iucv_userid, 8);
 773
 774         write_lock_bh(&iucv_sk_list.lock);
 775
 776         sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
 777         while (__iucv_get_sock_by_name(name)) {
 778                 sprintf(name, "%08x",
 779                         atomic_inc_return(&iucv_sk_list.autobind_name));
 780         }
 781
 782         write_unlock_bh(&iucv_sk_list.lock);
 783
 784         memcpy(&iucv->src_name, name, 8);
 785
 786         if (!iucv->msglimit)
 787                 iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 788
 789         return err;
 790 }
 791
 792 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
 793 {
 794         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 795         struct sock *sk = sock->sk;
 796         struct iucv_sock *iucv = iucv_sk(sk);
 797         unsigned char user_data[16];
 798         int err;
 799
 800         high_nmcpy(user_data, sa->siucv_name);
 801         low_nmcpy(user_data, iucv->src_name);
 802         ASCEBC(user_data, sizeof(user_data));
 803
 804         /* Create path. */
 805         iucv->path = iucv_path_alloc(iucv->msglimit,
 806                                      IUCV_IPRMDATA, GFP_KERNEL);
 807         if (!iucv->path) {
 808                 err = -ENOMEM;
 809                 goto done;
 810         }
 811         err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
 812                                     sa->siucv_user_id, NULL, user_data,
 813                                     sk);
 814         if (err) {
 815                 iucv_path_free(iucv->path);
 816                 iucv->path = NULL;
 817                 switch (err) {
 818                 case 0x0b:      /* Target communicator is not logged on */
 819                         err = -ENETUNREACH;
 820                         break;
 821                 case 0x0d:      /* Max connections for this guest exceeded */
 822                 case 0x0e:      /* Max connections for target guest exceeded */
 823                         err = -EAGAIN;
 824                         break;
 825                 case 0x0f:      /* Missing IUCV authorization */
 826                         err = -EACCES;
 827                         break;
 828                 default:
 829                         err = -ECONNREFUSED;
 830                         break;
 831                 }
 832         }
 833 done:
 834         return err;
 835 }
 836
 837 /* Connect an unconnected socket */
 838 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
 839                              int alen, int flags)
 840 {
 841         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 842         struct sock *sk = sock->sk;
 843         struct iucv_sock *iucv = iucv_sk(sk);
 844         int err;
 845
 846         if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
 847                 return -EINVAL;
 848
 849         if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
 850                 return -EBADFD;
 851
 852         if (sk->sk_state == IUCV_OPEN &&
 853             iucv->transport == AF_IUCV_TRANS_HIPER)
 854                 return -EBADFD; /* explicit bind required */
 855
 856         if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
 857                 return -EINVAL;
 858
 859         if (sk->sk_state == IUCV_OPEN) {
 860                 err = iucv_sock_autobind(sk);
 861                 if (unlikely(err))
 862                         return err;
 863         }
 864
 865         lock_sock(sk);
 866
 867         /* Set the destination information */
 868         memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
 869         memcpy(iucv->dst_name, sa->siucv_name, 8);
 870
 871         if (iucv->transport == AF_IUCV_TRANS_HIPER)
 872                 err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
 873         else
 874                 err = afiucv_path_connect(sock, addr);
 875         if (err)
 876                 goto done;
 877
 878         if (sk->sk_state != IUCV_CONNECTED)
 879                 err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
 880                                                             IUCV_DISCONN),
 881                                      sock_sndtimeo(sk, flags & O_NONBLOCK));
 882
 883         if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
 884                 err = -ECONNREFUSED;
 885
 886         if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
 887                 iucv_sever_path(sk, 0);
 888
 889 done:
 890         release_sock(sk);
 891         return err;
 892 }
 893
 894 /* Move a socket into listening state. */
 895 static int iucv_sock_listen(struct socket *sock, int backlog)
 896 {
 897         struct sock *sk = sock->sk;
 898         int err;
 899
 900         lock_sock(sk);
 901
 902         err = -EINVAL;
 903         if (sk->sk_state != IUCV_BOUND)
 904                 goto done;
 905
 906         if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
 907                 goto done;
 908
 909         sk->sk_max_ack_backlog = backlog;
 910         sk->sk_ack_backlog = 0;
 911         sk->sk_state = IUCV_LISTEN;
 912         err = 0;
 913
 914 done:
 915         release_sock(sk);
 916         return err;
 917 }
 918
 919 /* Accept a pending connection */
 920 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
 921                             int flags)
 922 {
 923         DECLARE_WAITQUEUE(wait, current);
 924         struct sock *sk = sock->sk, *nsk;
 925         long timeo;
 926         int err = 0;
 927
 928         lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 929
 930         if (sk->sk_state != IUCV_LISTEN) {
 931                 err = -EBADFD;
 932                 goto done;
 933         }
 934
 935         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
 936
 937         /* Wait for an incoming connection */
 938         add_wait_queue_exclusive(sk_sleep(sk), &wait);
 939         while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
 940                 set_current_state(TASK_INTERRUPTIBLE);
 941                 if (!timeo) {
 942                         err = -EAGAIN;
 943                         break;
 944                 }
 945
 946                 release_sock(sk);
 947                 timeo = schedule_timeout(timeo);
 948                 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 949
 950                 if (sk->sk_state != IUCV_LISTEN) {
 951                         err = -EBADFD;
 952                         break;
 953                 }
 954
 955                 if (signal_pending(current)) {
 956                         err = sock_intr_errno(timeo);
 957                         break;
 958                 }
 959         }
 960
 961         set_current_state(TASK_RUNNING);
 962         remove_wait_queue(sk_sleep(sk), &wait);
 963
 964         if (err)
 965                 goto done;
 966
 967         newsock->state = SS_CONNECTED;
 968
 969 done:
 970         release_sock(sk);
 971         return err;
 972 }
 973
 974 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
 975                              int *len, int peer)
 976 {
 977         struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
 978         struct sock *sk = sock->sk;
 979         struct iucv_sock *iucv = iucv_sk(sk);
 980
 981         addr->sa_family = AF_IUCV;
 982         *len = sizeof(struct sockaddr_iucv);
 983
 984         if (peer) {
 985                 memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
 986                 memcpy(siucv->siucv_name, iucv->dst_name, 8);
 987         } else {
 988                 memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
 989                 memcpy(siucv->siucv_name, iucv->src_name, 8);
 990         }
 991         memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
 992         memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
 993         memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
 994
 995         return 0;
 996 }
 997
 998 /**
 999  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
1000  * @path:       IUCV path
1001  * @msg:        Pointer to a struct iucv_message
1002  * @skb:        The socket data to send, skb->len MUST BE <= 7
1003  *
1004  * Send the socket data in the parameter list in the iucv message
1005  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
1006  * list and the socket data len at index 7 (last byte).
1007  * See also iucv_msg_length().
1008  *
1009  * Returns the error code from the iucv_message_send() call.
1010  */
1011 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
1012                           struct sk_buff *skb)
1013 {
1014         u8 prmdata[8];
1015
1016         memcpy(prmdata, (void *) skb->data, skb->len);
1017         prmdata[7] = 0xff - (u8) skb->len;
1018         return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
1019                                  (void *) prmdata, 8);
1020 }
1021
1022 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1023                              struct msghdr *msg, size_t len)
1024 {
1025         struct sock *sk = sock->sk;
1026         struct iucv_sock *iucv = iucv_sk(sk);
1027         struct sk_buff *skb;
1028         struct iucv_message txmsg;
1029         struct cmsghdr *cmsg;
1030         int cmsg_done;
1031         long timeo;
1032         char user_id[9];
1033         char appl_id[9];
1034         int err;
1035         int noblock = msg->msg_flags & MSG_DONTWAIT;
1036
1037         err = sock_error(sk);
1038         if (err)
1039                 return err;
1040
1041         if (msg->msg_flags & MSG_OOB)
1042                 return -EOPNOTSUPP;
1043
1044         /* SOCK_SEQPACKET: we do not support segmented records */
1045         if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
1046                 return -EOPNOTSUPP;
1047
1048         lock_sock(sk);
1049
1050         if (sk->sk_shutdown & SEND_SHUTDOWN) {
1051                 err = -EPIPE;
1052                 goto out;
1053         }
1054
1055         /* Return if the socket is not in connected state */
1056         if (sk->sk_state != IUCV_CONNECTED) {
1057                 err = -ENOTCONN;
1058                 goto out;
1059         }
1060
1061         /* initialize defaults */
1062         cmsg_done   = 0;        /* check for duplicate headers */
1063         txmsg.class = 0;
1064
1065         /* iterate over control messages */
1066         for (cmsg = CMSG_FIRSTHDR(msg); cmsg;
1067                 cmsg = CMSG_NXTHDR(msg, cmsg)) {
1068
1069                 if (!CMSG_OK(msg, cmsg)) {
1070                         err = -EINVAL;
1071                         goto out;
1072                 }
1073
1074                 if (cmsg->cmsg_level != SOL_IUCV)
1075                         continue;
1076
1077                 if (cmsg->cmsg_type & cmsg_done) {
1078                         err = -EINVAL;
1079                         goto out;
1080                 }
1081                 cmsg_done |= cmsg->cmsg_type;
1082
1083                 switch (cmsg->cmsg_type) {
1084                 case SCM_IUCV_TRGCLS:
1085                         if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
1086                                 err = -EINVAL;
1087                                 goto out;
1088                         }
1089
1090                         /* set iucv message target class */
1091                         memcpy(&txmsg.class,
1092                                 (void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
1093
1094                         break;
1095
1096                 default:
1097                         err = -EINVAL;
1098                         goto out;
1099                         break;
1100                 }
1101         }
1102
1103         /* allocate one skb for each iucv message:
1104          * this is fine for SOCK_SEQPACKET (unless we want to support
1105          * segmented records using the MSG_EOR flag), but
1106          * for SOCK_STREAM we might want to improve it in future */
1107         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1108                 skb = sock_alloc_send_skb(sk,
1109                         len + sizeof(struct af_iucv_trans_hdr) + ETH_HLEN,
1110                         noblock, &err);
1111         else
1112                 skb = sock_alloc_send_skb(sk, len, noblock, &err);
1113         if (!skb) {
1114                 err = -ENOMEM;
1115                 goto out;
1116         }
1117         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1118                 skb_reserve(skb, sizeof(struct af_iucv_trans_hdr) + ETH_HLEN);
1119         if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1120                 err = -EFAULT;
1121                 goto fail;
1122         }
1123
1124         /* wait if outstanding messages for iucv path has reached */
1125         timeo = sock_sndtimeo(sk, noblock);
1126         err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1127         if (err)
1128                 goto fail;
1129
1130         /* return -ECONNRESET if the socket is no longer connected */
1131         if (sk->sk_state != IUCV_CONNECTED) {
1132                 err = -ECONNRESET;
1133                 goto fail;
1134         }
1135
1136         /* increment and save iucv message tag for msg_completion cbk */
1137         txmsg.tag = iucv->send_tag++;
1138         IUCV_SKB_CB(skb)->tag = txmsg.tag;
1139
1140         if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1141                 atomic_inc(&iucv->msg_sent);
1142                 err = afiucv_hs_send(&txmsg, sk, skb, 0);
1143                 if (err) {
1144                         atomic_dec(&iucv->msg_sent);
1145                         goto fail;
1146                 }
1147                 goto release;
1148         }
1149         skb_queue_tail(&iucv->send_skb_q, skb);
1150
1151         if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags)
1152               && skb->len <= 7) {
1153                 err = iucv_send_iprm(iucv->path, &txmsg, skb);
1154
1155                 /* on success: there is no message_complete callback
1156                  * for an IPRMDATA msg; remove skb from send queue */
1157                 if (err == 0) {
1158                         skb_unlink(skb, &iucv->send_skb_q);
1159                         kfree_skb(skb);
1160                 }
1161
1162                 /* this error should never happen since the
1163                  * IUCV_IPRMDATA path flag is set... sever path */
1164                 if (err == 0x15) {
1165                         pr_iucv->path_sever(iucv->path, NULL);
1166                         skb_unlink(skb, &iucv->send_skb_q);
1167                         err = -EPIPE;
1168                         goto fail;
1169                 }
1170         } else
1171                 err = pr_iucv->message_send(iucv->path, &txmsg, 0, 0,
1172                                         (void *) skb->data, skb->len);
1173         if (err) {
1174                 if (err == 3) {
1175                         user_id[8] = 0;
1176                         memcpy(user_id, iucv->dst_user_id, 8);
1177                         appl_id[8] = 0;
1178                         memcpy(appl_id, iucv->dst_name, 8);
1179                         pr_err("Application %s on z/VM guest %s"
1180                                 " exceeds message limit\n",
1181                                 appl_id, user_id);
1182                         err = -EAGAIN;
1183                 } else
1184                         err = -EPIPE;
1185                 skb_unlink(skb, &iucv->send_skb_q);
1186                 goto fail;
1187         }
1188
1189 release:
1190         release_sock(sk);
1191         return len;
1192
1193 fail:
1194         kfree_skb(skb);
1195 out:
1196         release_sock(sk);
1197         return err;
1198 }
1199
1200 /* iucv_fragment_skb() - Fragment a single IUCV message into multiple skb's
1201  *
1202  * Locking: must be called with message_q.lock held
1203  */
1204 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
1205 {
1206         int dataleft, size, copied = 0;
1207         struct sk_buff *nskb;
1208
1209         dataleft = len;
1210         while (dataleft) {
1211                 if (dataleft >= sk->sk_rcvbuf / 4)
1212                         size = sk->sk_rcvbuf / 4;
1213                 else
1214                         size = dataleft;
1215
1216                 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
1217                 if (!nskb)
1218                         return -ENOMEM;
1219
1220                 /* copy target class to control buffer of new skb */
1221                 IUCV_SKB_CB(nskb)->class = IUCV_SKB_CB(skb)->class;
1222
1223                 /* copy data fragment */
1224                 memcpy(nskb->data, skb->data + copied, size);
1225                 copied += size;
1226                 dataleft -= size;
1227
1228                 skb_reset_transport_header(nskb);
1229                 skb_reset_network_header(nskb);
1230                 nskb->len = size;
1231
1232                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
1233         }
1234
1235         return 0;
1236 }
1237
1238 /* iucv_process_message() - Receive a single outstanding IUCV message
1239  *
1240  * Locking: must be called with message_q.lock held
1241  */
1242 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1243                                  struct iucv_path *path,
1244                                  struct iucv_message *msg)
1245 {
1246         int rc;
1247         unsigned int len;
1248
1249         len = iucv_msg_length(msg);
1250
1251         /* store msg target class in the second 4 bytes of skb ctrl buffer */
1252         /* Note: the first 4 bytes are reserved for msg tag */
1253         IUCV_SKB_CB(skb)->class = msg->class;
1254
1255         /* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1256         if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1257                 if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1258                         skb->data = NULL;
1259                         skb->len = 0;
1260                 }
1261         } else {
1262                 rc = pr_iucv->message_receive(path, msg,
1263                                               msg->flags & IUCV_IPRMDATA,
1264                                               skb->data, len, NULL);
1265                 if (rc) {
1266                         kfree_skb(skb);
1267                         return;
1268                 }
1269                 /* we need to fragment iucv messages for SOCK_STREAM only;
1270                  * for SOCK_SEQPACKET, it is only relevant if we support
1271                  * record segmentation using MSG_EOR (see also recvmsg()) */
1272                 if (sk->sk_type == SOCK_STREAM &&
1273                     skb->truesize >= sk->sk_rcvbuf / 4) {
1274                         rc = iucv_fragment_skb(sk, skb, len);
1275                         kfree_skb(skb);
1276                         skb = NULL;
1277                         if (rc) {
1278                                 pr_iucv->path_sever(path, NULL);
1279                                 return;
1280                         }
1281                         skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
1282                 } else {
1283                         skb_reset_transport_header(skb);
1284                         skb_reset_network_header(skb);
1285                         skb->len = len;
1286                 }
1287         }
1288
1289         IUCV_SKB_CB(skb)->offset = 0;
1290         if (sock_queue_rcv_skb(sk, skb))
1291                 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
1292 }
1293
1294 /* iucv_process_message_q() - Process outstanding IUCV messages
1295  *
1296  * Locking: must be called with message_q.lock held
1297  */
1298 static void iucv_process_message_q(struct sock *sk)
1299 {
1300         struct iucv_sock *iucv = iucv_sk(sk);
1301         struct sk_buff *skb;
1302         struct sock_msg_q *p, *n;
1303
1304         list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1305                 skb = alloc_skb(iucv_msg_length(&p->msg), GFP_ATOMIC | GFP_DMA);
1306                 if (!skb)
1307                         break;
1308                 iucv_process_message(sk, skb, p->path, &p->msg);
1309                 list_del(&p->list);
1310                 kfree(p);
1311                 if (!skb_queue_empty(&iucv->backlog_skb_q))
1312                         break;
1313         }
1314 }
1315
1316 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
1317                              struct msghdr *msg, size_t len, int flags)
1318 {
1319         int noblock = flags & MSG_DONTWAIT;
1320         struct sock *sk = sock->sk;
1321         struct iucv_sock *iucv = iucv_sk(sk);
1322         unsigned int copied, rlen;
1323         struct sk_buff *skb, *rskb, *cskb;
1324         int err = 0;
1325         u32 offset;
1326
1327         msg->msg_namelen = 0;
1328
1329         if ((sk->sk_state == IUCV_DISCONN) &&
1330             skb_queue_empty(&iucv->backlog_skb_q) &&
1331             skb_queue_empty(&sk->sk_receive_queue) &&
1332             list_empty(&iucv->message_q.list))
1333                 return 0;
1334
1335         if (flags & (MSG_OOB))
1336                 return -EOPNOTSUPP;
1337
1338         /* receive/dequeue next skb:
1339          * the function understands MSG_PEEK and, thus, does not dequeue skb */
1340         skb = skb_recv_datagram(sk, flags, noblock, &err);
1341         if (!skb) {
1342                 if (sk->sk_shutdown & RCV_SHUTDOWN)
1343                         return 0;
1344                 return err;
1345         }
1346
1347         offset = IUCV_SKB_CB(skb)->offset;
1348         rlen   = skb->len - offset;             /* real length of skb */
1349         copied = min_t(unsigned int, rlen, len);
1350         if (!rlen)
1351                 sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1352
1353         cskb = skb;
1354         if (skb_copy_datagram_iovec(cskb, offset, msg->msg_iov, copied)) {
1355                 if (!(flags & MSG_PEEK))
1356                         skb_queue_head(&sk->sk_receive_queue, skb);
1357                 return -EFAULT;
1358         }
1359
1360         /* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1361         if (sk->sk_type == SOCK_SEQPACKET) {
1362                 if (copied < rlen)
1363                         msg->msg_flags |= MSG_TRUNC;
1364                 /* each iucv message contains a complete record */
1365                 msg->msg_flags |= MSG_EOR;
1366         }
1367
1368         /* create control message to store iucv msg target class:
1369          * get the trgcls from the control buffer of the skb due to
1370          * fragmentation of original iucv message. */
1371         err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1372                        sizeof(IUCV_SKB_CB(skb)->class),
1373                        (void *)&IUCV_SKB_CB(skb)->class);
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                         if (copied < rlen) {
1386                                 IUCV_SKB_CB(skb)->offset = offset + copied;
1387                                 goto done;
1388                         }
1389                 }
1390
1391                 kfree_skb(skb);
1392                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1393                         atomic_inc(&iucv->msg_recv);
1394                         if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1395                                 WARN_ON(1);
1396                                 iucv_sock_close(sk);
1397                                 return -EFAULT;
1398                         }
1399                 }
1400
1401                 /* Queue backlog skbs */
1402                 spin_lock_bh(&iucv->message_q.lock);
1403                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1404                 while (rskb) {
1405                         IUCV_SKB_CB(rskb)->offset = 0;
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                         (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
1466
1467         if (sk->sk_shutdown & RCV_SHUTDOWN)
1468                 mask |= POLLRDHUP;
1469
1470         if (sk->sk_shutdown == SHUTDOWN_MASK)
1471                 mask |= POLLHUP;
1472
1473         if (!skb_queue_empty(&sk->sk_receive_queue) ||
1474             (sk->sk_shutdown & RCV_SHUTDOWN))
1475                 mask |= POLLIN | POLLRDNORM;
1476
1477         if (sk->sk_state == IUCV_CLOSED)
1478                 mask |= POLLHUP;
1479
1480         if (sk->sk_state == IUCV_DISCONN)
1481                 mask |= POLLIN;
1482
1483         if (sock_writeable(sk) && iucv_below_msglim(sk))
1484                 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1485         else
1486                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1487
1488         return mask;
1489 }
1490
1491 static int iucv_sock_shutdown(struct socket *sock, int how)
1492 {
1493         struct sock *sk = sock->sk;
1494         struct iucv_sock *iucv = iucv_sk(sk);
1495         struct iucv_message txmsg;
1496         int err = 0;
1497
1498         how++;
1499
1500         if ((how & ~SHUTDOWN_MASK) || !how)
1501                 return -EINVAL;
1502
1503         lock_sock(sk);
1504         switch (sk->sk_state) {
1505         case IUCV_LISTEN:
1506         case IUCV_DISCONN:
1507         case IUCV_CLOSING:
1508         case IUCV_CLOSED:
1509                 err = -ENOTCONN;
1510                 goto fail;
1511         default:
1512                 break;
1513         }
1514
1515         if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
1516                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1517                         txmsg.class = 0;
1518                         txmsg.tag = 0;
1519                         err = pr_iucv->message_send(iucv->path, &txmsg,
1520                                 IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1521                         if (err) {
1522                                 switch (err) {
1523                                 case 1:
1524                                         err = -ENOTCONN;
1525                                         break;
1526                                 case 2:
1527                                         err = -ECONNRESET;
1528                                         break;
1529                                 default:
1530                                         err = -ENOTCONN;
1531                                         break;
1532                                 }
1533                         }
1534                 } else
1535                         iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1536         }
1537
1538         sk->sk_shutdown |= how;
1539         if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1540                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1541                         err = pr_iucv->path_quiesce(iucv->path, NULL);
1542                         if (err)
1543                                 err = -ENOTCONN;
1544 /*                      skb_queue_purge(&sk->sk_receive_queue); */
1545                 }
1546                 skb_queue_purge(&sk->sk_receive_queue);
1547         }
1548
1549         /* Wake up anyone sleeping in poll */
1550         sk->sk_state_change(sk);
1551
1552 fail:
1553         release_sock(sk);
1554         return err;
1555 }
1556
1557 static int iucv_sock_release(struct socket *sock)
1558 {
1559         struct sock *sk = sock->sk;
1560         int err = 0;
1561
1562         if (!sk)
1563                 return 0;
1564
1565         iucv_sock_close(sk);
1566
1567         sock_orphan(sk);
1568         iucv_sock_kill(sk);
1569         return err;
1570 }
1571
1572 /* getsockopt and setsockopt */
1573 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1574                                 char __user *optval, unsigned int optlen)
1575 {
1576         struct sock *sk = sock->sk;
1577         struct iucv_sock *iucv = iucv_sk(sk);
1578         int val;
1579         int rc;
1580
1581         if (level != SOL_IUCV)
1582                 return -ENOPROTOOPT;
1583
1584         if (optlen < sizeof(int))
1585                 return -EINVAL;
1586
1587         if (get_user(val, (int __user *) optval))
1588                 return -EFAULT;
1589
1590         rc = 0;
1591
1592         lock_sock(sk);
1593         switch (optname) {
1594         case SO_IPRMDATA_MSG:
1595                 if (val)
1596                         iucv->flags |= IUCV_IPRMDATA;
1597                 else
1598                         iucv->flags &= ~IUCV_IPRMDATA;
1599                 break;
1600         case SO_MSGLIMIT:
1601                 switch (sk->sk_state) {
1602                 case IUCV_OPEN:
1603                 case IUCV_BOUND:
1604                         if (val < 1 || val > (u16)(~0))
1605                                 rc = -EINVAL;
1606                         else
1607                                 iucv->msglimit = val;
1608                         break;
1609                 default:
1610                         rc = -EINVAL;
1611                         break;
1612                 }
1613                 break;
1614         default:
1615                 rc = -ENOPROTOOPT;
1616                 break;
1617         }
1618         release_sock(sk);
1619
1620         return rc;
1621 }
1622
1623 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1624                                 char __user *optval, int __user *optlen)
1625 {
1626         struct sock *sk = sock->sk;
1627         struct iucv_sock *iucv = iucv_sk(sk);
1628         unsigned int val;
1629         int len;
1630
1631         if (level != SOL_IUCV)
1632                 return -ENOPROTOOPT;
1633
1634         if (get_user(len, optlen))
1635                 return -EFAULT;
1636
1637         if (len < 0)
1638                 return -EINVAL;
1639
1640         len = min_t(unsigned int, len, sizeof(int));
1641
1642         switch (optname) {
1643         case SO_IPRMDATA_MSG:
1644                 val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1645                 break;
1646         case SO_MSGLIMIT:
1647                 lock_sock(sk);
1648                 val = (iucv->path != NULL) ? iucv->path->msglim /* connected */
1649                                            : iucv->msglimit;    /* default */
1650                 release_sock(sk);
1651                 break;
1652         case SO_MSGSIZE:
1653                 if (sk->sk_state == IUCV_OPEN)
1654                         return -EBADFD;
1655                 val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1656                                 sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1657                                 0x7fffffff;
1658                 break;
1659         default:
1660                 return -ENOPROTOOPT;
1661         }
1662
1663         if (put_user(len, optlen))
1664                 return -EFAULT;
1665         if (copy_to_user(optval, &val, len))
1666                 return -EFAULT;
1667
1668         return 0;
1669 }
1670
1671
1672 /* Callback wrappers - called from iucv base support */
1673 static int iucv_callback_connreq(struct iucv_path *path,
1674                                  u8 ipvmid[8], u8 ipuser[16])
1675 {
1676         unsigned char user_data[16];
1677         unsigned char nuser_data[16];
1678         unsigned char src_name[8];
1679         struct sock *sk, *nsk;
1680         struct iucv_sock *iucv, *niucv;
1681         int err;
1682
1683         memcpy(src_name, ipuser, 8);
1684         EBCASC(src_name, 8);
1685         /* Find out if this path belongs to af_iucv. */
1686         read_lock(&iucv_sk_list.lock);
1687         iucv = NULL;
1688         sk = NULL;
1689         sk_for_each(sk, &iucv_sk_list.head)
1690                 if (sk->sk_state == IUCV_LISTEN &&
1691                     !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1692                         /*
1693                          * Found a listening socket with
1694                          * src_name == ipuser[0-7].
1695                          */
1696                         iucv = iucv_sk(sk);
1697                         break;
1698                 }
1699         read_unlock(&iucv_sk_list.lock);
1700         if (!iucv)
1701                 /* No socket found, not one of our paths. */
1702                 return -EINVAL;
1703
1704         bh_lock_sock(sk);
1705
1706         /* Check if parent socket is listening */
1707         low_nmcpy(user_data, iucv->src_name);
1708         high_nmcpy(user_data, iucv->dst_name);
1709         ASCEBC(user_data, sizeof(user_data));
1710         if (sk->sk_state != IUCV_LISTEN) {
1711                 err = pr_iucv->path_sever(path, user_data);
1712                 iucv_path_free(path);
1713                 goto fail;
1714         }
1715
1716         /* Check for backlog size */
1717         if (sk_acceptq_is_full(sk)) {
1718                 err = pr_iucv->path_sever(path, user_data);
1719                 iucv_path_free(path);
1720                 goto fail;
1721         }
1722
1723         /* Create the new socket */
1724         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1725         if (!nsk) {
1726                 err = pr_iucv->path_sever(path, user_data);
1727                 iucv_path_free(path);
1728                 goto fail;
1729         }
1730
1731         niucv = iucv_sk(nsk);
1732         iucv_sock_init(nsk, sk);
1733
1734         /* Set the new iucv_sock */
1735         memcpy(niucv->dst_name, ipuser + 8, 8);
1736         EBCASC(niucv->dst_name, 8);
1737         memcpy(niucv->dst_user_id, ipvmid, 8);
1738         memcpy(niucv->src_name, iucv->src_name, 8);
1739         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1740         niucv->path = path;
1741
1742         /* Call iucv_accept */
1743         high_nmcpy(nuser_data, ipuser + 8);
1744         memcpy(nuser_data + 8, niucv->src_name, 8);
1745         ASCEBC(nuser_data + 8, 8);
1746
1747         /* set message limit for path based on msglimit of accepting socket */
1748         niucv->msglimit = iucv->msglimit;
1749         path->msglim = iucv->msglimit;
1750         err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1751         if (err) {
1752                 iucv_sever_path(nsk, 1);
1753                 iucv_sock_kill(nsk);
1754                 goto fail;
1755         }
1756
1757         iucv_accept_enqueue(sk, nsk);
1758
1759         /* Wake up accept */
1760         nsk->sk_state = IUCV_CONNECTED;
1761         sk->sk_data_ready(sk, 1);
1762         err = 0;
1763 fail:
1764         bh_unlock_sock(sk);
1765         return 0;
1766 }
1767
1768 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1769 {
1770         struct sock *sk = path->private;
1771
1772         sk->sk_state = IUCV_CONNECTED;
1773         sk->sk_state_change(sk);
1774 }
1775
1776 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1777 {
1778         struct sock *sk = path->private;
1779         struct iucv_sock *iucv = iucv_sk(sk);
1780         struct sk_buff *skb;
1781         struct sock_msg_q *save_msg;
1782         int len;
1783
1784         if (sk->sk_shutdown & RCV_SHUTDOWN) {
1785                 pr_iucv->message_reject(path, msg);
1786                 return;
1787         }
1788
1789         spin_lock(&iucv->message_q.lock);
1790
1791         if (!list_empty(&iucv->message_q.list) ||
1792             !skb_queue_empty(&iucv->backlog_skb_q))
1793                 goto save_message;
1794
1795         len = atomic_read(&sk->sk_rmem_alloc);
1796         len += SKB_TRUESIZE(iucv_msg_length(msg));
1797         if (len > sk->sk_rcvbuf)
1798                 goto save_message;
1799
1800         skb = alloc_skb(iucv_msg_length(msg), GFP_ATOMIC | GFP_DMA);
1801         if (!skb)
1802                 goto save_message;
1803
1804         iucv_process_message(sk, skb, path, msg);
1805         goto out_unlock;
1806
1807 save_message:
1808         save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1809         if (!save_msg)
1810                 goto out_unlock;
1811         save_msg->path = path;
1812         save_msg->msg = *msg;
1813
1814         list_add_tail(&save_msg->list, &iucv->message_q.list);
1815
1816 out_unlock:
1817         spin_unlock(&iucv->message_q.lock);
1818 }
1819
1820 static void iucv_callback_txdone(struct iucv_path *path,
1821                                  struct iucv_message *msg)
1822 {
1823         struct sock *sk = path->private;
1824         struct sk_buff *this = NULL;
1825         struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1826         struct sk_buff *list_skb = list->next;
1827         unsigned long flags;
1828
1829         bh_lock_sock(sk);
1830         if (!skb_queue_empty(list)) {
1831                 spin_lock_irqsave(&list->lock, flags);
1832
1833                 while (list_skb != (struct sk_buff *)list) {
1834                         if (msg->tag != IUCV_SKB_CB(list_skb)->tag) {
1835                                 this = list_skb;
1836                                 break;
1837                         }
1838                         list_skb = list_skb->next;
1839                 }
1840                 if (this)
1841                         __skb_unlink(this, list);
1842
1843                 spin_unlock_irqrestore(&list->lock, flags);
1844
1845                 if (this) {
1846                         kfree_skb(this);
1847                         /* wake up any process waiting for sending */
1848                         iucv_sock_wake_msglim(sk);
1849                 }
1850         }
1851
1852         if (sk->sk_state == IUCV_CLOSING) {
1853                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1854                         sk->sk_state = IUCV_CLOSED;
1855                         sk->sk_state_change(sk);
1856                 }
1857         }
1858         bh_unlock_sock(sk);
1859
1860 }
1861
1862 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1863 {
1864         struct sock *sk = path->private;
1865
1866         if (sk->sk_state == IUCV_CLOSED)
1867                 return;
1868
1869         bh_lock_sock(sk);
1870         iucv_sever_path(sk, 1);
1871         sk->sk_state = IUCV_DISCONN;
1872
1873         sk->sk_state_change(sk);
1874         bh_unlock_sock(sk);
1875 }
1876
1877 /* called if the other communication side shuts down its RECV direction;
1878  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1879  */
1880 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1881 {
1882         struct sock *sk = path->private;
1883
1884         bh_lock_sock(sk);
1885         if (sk->sk_state != IUCV_CLOSED) {
1886                 sk->sk_shutdown |= SEND_SHUTDOWN;
1887                 sk->sk_state_change(sk);
1888         }
1889         bh_unlock_sock(sk);
1890 }
1891
1892 /***************** HiperSockets transport callbacks ********************/
1893 static void afiucv_swap_src_dest(struct sk_buff *skb)
1894 {
1895         struct af_iucv_trans_hdr *trans_hdr =
1896                                 (struct af_iucv_trans_hdr *)skb->data;
1897         char tmpID[8];
1898         char tmpName[8];
1899
1900         ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1901         ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1902         ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1903         ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1904         memcpy(tmpID, trans_hdr->srcUserID, 8);
1905         memcpy(tmpName, trans_hdr->srcAppName, 8);
1906         memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1907         memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1908         memcpy(trans_hdr->destUserID, tmpID, 8);
1909         memcpy(trans_hdr->destAppName, tmpName, 8);
1910         skb_push(skb, ETH_HLEN);
1911         memset(skb->data, 0, ETH_HLEN);
1912 }
1913
1914 /**
1915  * afiucv_hs_callback_syn - react on received SYN
1916  **/
1917 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1918 {
1919         struct sock *nsk;
1920         struct iucv_sock *iucv, *niucv;
1921         struct af_iucv_trans_hdr *trans_hdr;
1922         int err;
1923
1924         iucv = iucv_sk(sk);
1925         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
1926         if (!iucv) {
1927                 /* no sock - connection refused */
1928                 afiucv_swap_src_dest(skb);
1929                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1930                 err = dev_queue_xmit(skb);
1931                 goto out;
1932         }
1933
1934         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1935         bh_lock_sock(sk);
1936         if ((sk->sk_state != IUCV_LISTEN) ||
1937             sk_acceptq_is_full(sk) ||
1938             !nsk) {
1939                 /* error on server socket - connection refused */
1940                 if (nsk)
1941                         sk_free(nsk);
1942                 afiucv_swap_src_dest(skb);
1943                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1944                 err = dev_queue_xmit(skb);
1945                 bh_unlock_sock(sk);
1946                 goto out;
1947         }
1948
1949         niucv = iucv_sk(nsk);
1950         iucv_sock_init(nsk, sk);
1951         niucv->transport = AF_IUCV_TRANS_HIPER;
1952         niucv->msglimit = iucv->msglimit;
1953         if (!trans_hdr->window)
1954                 niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1955         else
1956                 niucv->msglimit_peer = trans_hdr->window;
1957         memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1958         memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1959         memcpy(niucv->src_name, iucv->src_name, 8);
1960         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1961         nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1962         niucv->hs_dev = iucv->hs_dev;
1963         dev_hold(niucv->hs_dev);
1964         afiucv_swap_src_dest(skb);
1965         trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1966         trans_hdr->window = niucv->msglimit;
1967         /* if receiver acks the xmit connection is established */
1968         err = dev_queue_xmit(skb);
1969         if (!err) {
1970                 iucv_accept_enqueue(sk, nsk);
1971                 nsk->sk_state = IUCV_CONNECTED;
1972                 sk->sk_data_ready(sk, 1);
1973         } else
1974                 iucv_sock_kill(nsk);
1975         bh_unlock_sock(sk);
1976
1977 out:
1978         return NET_RX_SUCCESS;
1979 }
1980
1981 /**
1982  * afiucv_hs_callback_synack() - react on received SYN-ACK
1983  **/
1984 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1985 {
1986         struct iucv_sock *iucv = iucv_sk(sk);
1987         struct af_iucv_trans_hdr *trans_hdr =
1988                                         (struct af_iucv_trans_hdr *)skb->data;
1989
1990         if (!iucv)
1991                 goto out;
1992         if (sk->sk_state != IUCV_BOUND)
1993                 goto out;
1994         bh_lock_sock(sk);
1995         iucv->msglimit_peer = trans_hdr->window;
1996         sk->sk_state = IUCV_CONNECTED;
1997         sk->sk_state_change(sk);
1998         bh_unlock_sock(sk);
1999 out:
2000         kfree_skb(skb);
2001         return NET_RX_SUCCESS;
2002 }
2003
2004 /**
2005  * afiucv_hs_callback_synfin() - react on received SYN_FIN
2006  **/
2007 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
2008 {
2009         struct iucv_sock *iucv = iucv_sk(sk);
2010
2011         if (!iucv)
2012                 goto out;
2013         if (sk->sk_state != IUCV_BOUND)
2014                 goto out;
2015         bh_lock_sock(sk);
2016         sk->sk_state = IUCV_DISCONN;
2017         sk->sk_state_change(sk);
2018         bh_unlock_sock(sk);
2019 out:
2020         kfree_skb(skb);
2021         return NET_RX_SUCCESS;
2022 }
2023
2024 /**
2025  * afiucv_hs_callback_fin() - react on received FIN
2026  **/
2027 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
2028 {
2029         struct iucv_sock *iucv = iucv_sk(sk);
2030
2031         /* other end of connection closed */
2032         if (!iucv)
2033                 goto out;
2034         bh_lock_sock(sk);
2035         if (sk->sk_state == IUCV_CONNECTED) {
2036                 sk->sk_state = IUCV_DISCONN;
2037                 sk->sk_state_change(sk);
2038         }
2039         bh_unlock_sock(sk);
2040 out:
2041         kfree_skb(skb);
2042         return NET_RX_SUCCESS;
2043 }
2044
2045 /**
2046  * afiucv_hs_callback_win() - react on received WIN
2047  **/
2048 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
2049 {
2050         struct iucv_sock *iucv = iucv_sk(sk);
2051         struct af_iucv_trans_hdr *trans_hdr =
2052                                         (struct af_iucv_trans_hdr *)skb->data;
2053
2054         if (!iucv)
2055                 return NET_RX_SUCCESS;
2056
2057         if (sk->sk_state != IUCV_CONNECTED)
2058                 return NET_RX_SUCCESS;
2059
2060         atomic_sub(trans_hdr->window, &iucv->msg_sent);
2061         iucv_sock_wake_msglim(sk);
2062         return NET_RX_SUCCESS;
2063 }
2064
2065 /**
2066  * afiucv_hs_callback_rx() - react on received data
2067  **/
2068 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
2069 {
2070         struct iucv_sock *iucv = iucv_sk(sk);
2071
2072         if (!iucv) {
2073                 kfree_skb(skb);
2074                 return NET_RX_SUCCESS;
2075         }
2076
2077         if (sk->sk_state != IUCV_CONNECTED) {
2078                 kfree_skb(skb);
2079                 return NET_RX_SUCCESS;
2080         }
2081
2082         if (sk->sk_shutdown & RCV_SHUTDOWN) {
2083                 kfree_skb(skb);
2084                 return NET_RX_SUCCESS;
2085         }
2086
2087                 /* write stuff from iucv_msg to skb cb */
2088         if (skb->len < sizeof(struct af_iucv_trans_hdr)) {
2089                 kfree_skb(skb);
2090                 return NET_RX_SUCCESS;
2091         }
2092         skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2093         skb_reset_transport_header(skb);
2094         skb_reset_network_header(skb);
2095         IUCV_SKB_CB(skb)->offset = 0;
2096         spin_lock(&iucv->message_q.lock);
2097         if (skb_queue_empty(&iucv->backlog_skb_q)) {
2098                 if (sock_queue_rcv_skb(sk, skb)) {
2099                         /* handle rcv queue full */
2100                         skb_queue_tail(&iucv->backlog_skb_q, skb);
2101                 }
2102         } else
2103                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2104         spin_unlock(&iucv->message_q.lock);
2105         return NET_RX_SUCCESS;
2106 }
2107
2108 /**
2109  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2110  *                   transport
2111  *                   called from netif RX softirq
2112  **/
2113 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2114         struct packet_type *pt, struct net_device *orig_dev)
2115 {
2116         struct sock *sk;
2117         struct iucv_sock *iucv;
2118         struct af_iucv_trans_hdr *trans_hdr;
2119         char nullstring[8];
2120         int err = 0;
2121
2122         skb_pull(skb, ETH_HLEN);
2123         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
2124         EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2125         EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2126         EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2127         EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2128         memset(nullstring, 0, sizeof(nullstring));
2129         iucv = NULL;
2130         sk = NULL;
2131         read_lock(&iucv_sk_list.lock);
2132         sk_for_each(sk, &iucv_sk_list.head) {
2133                 if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2134                         if ((!memcmp(&iucv_sk(sk)->src_name,
2135                                      trans_hdr->destAppName, 8)) &&
2136                             (!memcmp(&iucv_sk(sk)->src_user_id,
2137                                      trans_hdr->destUserID, 8)) &&
2138                             (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2139                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2140                                      nullstring, 8))) {
2141                                 iucv = iucv_sk(sk);
2142                                 break;
2143                         }
2144                 } else {
2145                         if ((!memcmp(&iucv_sk(sk)->src_name,
2146                                      trans_hdr->destAppName, 8)) &&
2147                             (!memcmp(&iucv_sk(sk)->src_user_id,
2148                                      trans_hdr->destUserID, 8)) &&
2149                             (!memcmp(&iucv_sk(sk)->dst_name,
2150                                      trans_hdr->srcAppName, 8)) &&
2151                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2152                                      trans_hdr->srcUserID, 8))) {
2153                                 iucv = iucv_sk(sk);
2154                                 break;
2155                         }
2156                 }
2157         }
2158         read_unlock(&iucv_sk_list.lock);
2159         if (!iucv)
2160                 sk = NULL;
2161
2162         /* no sock
2163         how should we send with no sock
2164         1) send without sock no send rc checking?
2165         2) introduce default sock to handle this cases
2166
2167          SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2168          data -> send FIN
2169          SYN|ACK, SYN|FIN, FIN -> no action? */
2170
2171         switch (trans_hdr->flags) {
2172         case AF_IUCV_FLAG_SYN:
2173                 /* connect request */
2174                 err = afiucv_hs_callback_syn(sk, skb);
2175                 break;
2176         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2177                 /* connect request confirmed */
2178                 err = afiucv_hs_callback_synack(sk, skb);
2179                 break;
2180         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2181                 /* connect request refused */
2182                 err = afiucv_hs_callback_synfin(sk, skb);
2183                 break;
2184         case (AF_IUCV_FLAG_FIN):
2185                 /* close request */
2186                 err = afiucv_hs_callback_fin(sk, skb);
2187                 break;
2188         case (AF_IUCV_FLAG_WIN):
2189                 err = afiucv_hs_callback_win(sk, skb);
2190                 if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2191                         kfree_skb(skb);
2192                         break;
2193                 }
2194                 /* fall through and receive non-zero length data */
2195         case (AF_IUCV_FLAG_SHT):
2196                 /* shutdown request */
2197                 /* fall through and receive zero length data */
2198         case 0:
2199                 /* plain data frame */
2200                 IUCV_SKB_CB(skb)->class = trans_hdr->iucv_hdr.class;
2201                 err = afiucv_hs_callback_rx(sk, skb);
2202                 break;
2203         default:
2204                 ;
2205         }
2206
2207         return err;
2208 }
2209
2210 /**
2211  * afiucv_hs_callback_txnotify() - handle send notifcations from HiperSockets
2212  *                                 transport
2213  **/
2214 static void afiucv_hs_callback_txnotify(struct sk_buff *skb,
2215                                         enum iucv_tx_notify n)
2216 {
2217         struct sock *isk = skb->sk;
2218         struct sock *sk = NULL;
2219         struct iucv_sock *iucv = NULL;
2220         struct sk_buff_head *list;
2221         struct sk_buff *list_skb;
2222         struct sk_buff *nskb;
2223         unsigned long flags;
2224
2225         read_lock_irqsave(&iucv_sk_list.lock, flags);
2226         sk_for_each(sk, &iucv_sk_list.head)
2227                 if (sk == isk) {
2228                         iucv = iucv_sk(sk);
2229                         break;
2230                 }
2231         read_unlock_irqrestore(&iucv_sk_list.lock, flags);
2232
2233         if (!iucv || sock_flag(sk, SOCK_ZAPPED))
2234                 return;
2235
2236         list = &iucv->send_skb_q;
2237         spin_lock_irqsave(&list->lock, flags);
2238         if (skb_queue_empty(list))
2239                 goto out_unlock;
2240         list_skb = list->next;
2241         nskb = list_skb->next;
2242         while (list_skb != (struct sk_buff *)list) {
2243                 if (skb_shinfo(list_skb) == skb_shinfo(skb)) {
2244                         switch (n) {
2245                         case TX_NOTIFY_OK:
2246                                 __skb_unlink(list_skb, list);
2247                                 kfree_skb(list_skb);
2248                                 iucv_sock_wake_msglim(sk);
2249                                 break;
2250                         case TX_NOTIFY_PENDING:
2251                                 atomic_inc(&iucv->pendings);
2252                                 break;
2253                         case TX_NOTIFY_DELAYED_OK:
2254                                 __skb_unlink(list_skb, list);
2255                                 atomic_dec(&iucv->pendings);
2256                                 if (atomic_read(&iucv->pendings) <= 0)
2257                                         iucv_sock_wake_msglim(sk);
2258                                 kfree_skb(list_skb);
2259                                 break;
2260                         case TX_NOTIFY_UNREACHABLE:
2261                         case TX_NOTIFY_DELAYED_UNREACHABLE:
2262                         case TX_NOTIFY_TPQFULL: /* not yet used */
2263                         case TX_NOTIFY_GENERALERROR:
2264                         case TX_NOTIFY_DELAYED_GENERALERROR:
2265                                 __skb_unlink(list_skb, list);
2266                                 kfree_skb(list_skb);
2267                                 if (sk->sk_state == IUCV_CONNECTED) {
2268                                         sk->sk_state = IUCV_DISCONN;
2269                                         sk->sk_state_change(sk);
2270                                 }
2271                                 break;
2272                         }
2273                         break;
2274                 }
2275                 list_skb = nskb;
2276                 nskb = nskb->next;
2277         }
2278 out_unlock:
2279         spin_unlock_irqrestore(&list->lock, flags);
2280
2281         if (sk->sk_state == IUCV_CLOSING) {
2282                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
2283                         sk->sk_state = IUCV_CLOSED;
2284                         sk->sk_state_change(sk);
2285                 }
2286         }
2287
2288 }
2289
2290 /*
2291  * afiucv_netdev_event: handle netdev notifier chain events
2292  */
2293 static int afiucv_netdev_event(struct notifier_block *this,
2294                                unsigned long event, void *ptr)
2295 {
2296         struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
2297         struct sock *sk;
2298         struct iucv_sock *iucv;
2299
2300         switch (event) {
2301         case NETDEV_REBOOT:
2302         case NETDEV_GOING_DOWN:
2303                 sk_for_each(sk, &iucv_sk_list.head) {
2304                         iucv = iucv_sk(sk);
2305                         if ((iucv->hs_dev == event_dev) &&
2306                             (sk->sk_state == IUCV_CONNECTED)) {
2307                                 if (event == NETDEV_GOING_DOWN)
2308                                         iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2309                                 sk->sk_state = IUCV_DISCONN;
2310                                 sk->sk_state_change(sk);
2311                         }
2312                 }
2313                 break;
2314         case NETDEV_DOWN:
2315         case NETDEV_UNREGISTER:
2316         default:
2317                 break;
2318         }
2319         return NOTIFY_DONE;
2320 }
2321
2322 static struct notifier_block afiucv_netdev_notifier = {
2323         .notifier_call = afiucv_netdev_event,
2324 };
2325
2326 static const struct proto_ops iucv_sock_ops = {
2327         .family         = PF_IUCV,
2328         .owner          = THIS_MODULE,
2329         .release        = iucv_sock_release,
2330         .bind           = iucv_sock_bind,
2331         .connect        = iucv_sock_connect,
2332         .listen         = iucv_sock_listen,
2333         .accept         = iucv_sock_accept,
2334         .getname        = iucv_sock_getname,
2335         .sendmsg        = iucv_sock_sendmsg,
2336         .recvmsg        = iucv_sock_recvmsg,
2337         .poll           = iucv_sock_poll,
2338         .ioctl          = sock_no_ioctl,
2339         .mmap           = sock_no_mmap,
2340         .socketpair     = sock_no_socketpair,
2341         .shutdown       = iucv_sock_shutdown,
2342         .setsockopt     = iucv_sock_setsockopt,
2343         .getsockopt     = iucv_sock_getsockopt,
2344 };
2345
2346 static const struct net_proto_family iucv_sock_family_ops = {
2347         .family = AF_IUCV,
2348         .owner  = THIS_MODULE,
2349         .create = iucv_sock_create,
2350 };
2351
2352 static struct packet_type iucv_packet_type = {
2353         .type = cpu_to_be16(ETH_P_AF_IUCV),
2354         .func = afiucv_hs_rcv,
2355 };
2356
2357 static int afiucv_iucv_init(void)
2358 {
2359         int err;
2360
2361         err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2362         if (err)
2363                 goto out;
2364         /* establish dummy device */
2365         af_iucv_driver.bus = pr_iucv->bus;
2366         err = driver_register(&af_iucv_driver);
2367         if (err)
2368                 goto out_iucv;
2369         af_iucv_dev = kzalloc(sizeof(struct device), GFP_KERNEL);
2370         if (!af_iucv_dev) {
2371                 err = -ENOMEM;
2372                 goto out_driver;
2373         }
2374         dev_set_name(af_iucv_dev, "af_iucv");
2375         af_iucv_dev->bus = pr_iucv->bus;
2376         af_iucv_dev->parent = pr_iucv->root;
2377         af_iucv_dev->release = (void (*)(struct device *))kfree;
2378         af_iucv_dev->driver = &af_iucv_driver;
2379         err = device_register(af_iucv_dev);
2380         if (err)
2381                 goto out_driver;
2382         return 0;
2383
2384 out_driver:
2385         driver_unregister(&af_iucv_driver);
2386 out_iucv:
2387         pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2388 out:
2389         return err;
2390 }
2391
2392 static int __init afiucv_init(void)
2393 {
2394         int err;
2395
2396         if (MACHINE_IS_VM) {
2397                 cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2398                 if (unlikely(err)) {
2399                         WARN_ON(err);
2400                         err = -EPROTONOSUPPORT;
2401                         goto out;
2402                 }
2403
2404                 pr_iucv = try_then_request_module(symbol_get(iucv_if), "iucv");
2405                 if (!pr_iucv) {
2406                         printk(KERN_WARNING "iucv_if lookup failed\n");
2407                         memset(&iucv_userid, 0, sizeof(iucv_userid));
2408                 }
2409         } else {
2410                 memset(&iucv_userid, 0, sizeof(iucv_userid));
2411                 pr_iucv = NULL;
2412         }
2413
2414         err = proto_register(&iucv_proto, 0);
2415         if (err)
2416                 goto out;
2417         err = sock_register(&iucv_sock_family_ops);
2418         if (err)
2419                 goto out_proto;
2420
2421         if (pr_iucv) {
2422                 err = afiucv_iucv_init();
2423                 if (err)
2424                         goto out_sock;
2425         } else
2426                 register_netdevice_notifier(&afiucv_netdev_notifier);
2427         dev_add_pack(&iucv_packet_type);
2428         return 0;
2429
2430 out_sock:
2431         sock_unregister(PF_IUCV);
2432 out_proto:
2433         proto_unregister(&iucv_proto);
2434 out:
2435         if (pr_iucv)
2436                 symbol_put(iucv_if);
2437         return err;
2438 }
2439
2440 static void __exit afiucv_exit(void)
2441 {
2442         if (pr_iucv) {
2443                 device_unregister(af_iucv_dev);
2444                 driver_unregister(&af_iucv_driver);
2445                 pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2446                 symbol_put(iucv_if);
2447         } else
2448                 unregister_netdevice_notifier(&afiucv_netdev_notifier);
2449         dev_remove_pack(&iucv_packet_type);
2450         sock_unregister(PF_IUCV);
2451         proto_unregister(&iucv_proto);
2452 }
2453
2454 module_init(afiucv_init);
2455 module_exit(afiucv_exit);
2456
2457 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2458 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2459 MODULE_VERSION(VERSION);
2460 MODULE_LICENSE("GPL");
2461 MODULE_ALIAS_NETPROTO(PF_IUCV);
2462