net/iucv/af_iucv.c

   1 /*
   2  *  IUCV protocol stack for Linux on zSeries
   3  *
   4  *  Copyright IBM Corp. 2006, 2009
   5  *
   6  *  Author(s):  Jennifer Hunt <jenhunt@us.ibm.com>
   7  *              Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
   8  *  PM functions:
   9  *              Ursula Braun <ursula.braun@de.ibm.com>
  10  */
  11
  12 #define KMSG_COMPONENT "af_iucv"
  13 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  14
  15 #include <linux/module.h>
  16 #include <linux/types.h>
  17 #include <linux/list.h>
  18 #include <linux/errno.h>
  19 #include <linux/kernel.h>
  20 #include <linux/sched.h>
  21 #include <linux/slab.h>
  22 #include <linux/skbuff.h>
  23 #include <linux/init.h>
  24 #include <linux/poll.h>
  25 #include <net/sock.h>
  26 #include <asm/ebcdic.h>
  27 #include <asm/cpcmd.h>
  28 #include <linux/kmod.h>
  29
  30 #include <net/iucv/af_iucv.h>
  31
  32 #define VERSION "1.2"
  33
  34 static char iucv_userid[80];
  35
  36 static const struct proto_ops iucv_sock_ops;
  37
  38 static struct proto iucv_proto = {
  39         .name           = "AF_IUCV",
  40         .owner          = THIS_MODULE,
  41         .obj_size       = sizeof(struct iucv_sock),
  42 };
  43
  44 static struct iucv_interface *pr_iucv;
  45
  46 /* special AF_IUCV IPRM messages */
  47 static const u8 iprm_shutdown[8] =
  48         {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
  49
  50 #define TRGCLS_SIZE     (sizeof(((struct iucv_message *)0)->class))
  51
  52 /* macros to set/get socket control buffer at correct offset */
  53 #define CB_TAG(skb)     ((skb)->cb)             /* iucv message tag */
  54 #define CB_TAG_LEN      (sizeof(((struct iucv_message *) 0)->tag))
  55 #define CB_TRGCLS(skb)  ((skb)->cb + CB_TAG_LEN) /* iucv msg target class */
  56 #define CB_TRGCLS_LEN   (TRGCLS_SIZE)
  57
  58 #define __iucv_sock_wait(sk, condition, timeo, ret)                     \
  59 do {                                                                    \
  60         DEFINE_WAIT(__wait);                                            \
  61         long __timeo = timeo;                                           \
  62         ret = 0;                                                        \
  63         prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);     \
  64         while (!(condition)) {                                          \
  65                 if (!__timeo) {                                         \
  66                         ret = -EAGAIN;                                  \
  67                         break;                                          \
  68                 }                                                       \
  69                 if (signal_pending(current)) {                          \
  70                         ret = sock_intr_errno(__timeo);                 \
  71                         break;                                          \
  72                 }                                                       \
  73                 release_sock(sk);                                       \
  74                 __timeo = schedule_timeout(__timeo);                    \
  75                 lock_sock(sk);                                          \
  76                 ret = sock_error(sk);                                   \
  77                 if (ret)                                                \
  78                         break;                                          \
  79         }                                                               \
  80         finish_wait(sk_sleep(sk), &__wait);                             \
  81 } while (0)
  82
  83 #define iucv_sock_wait(sk, condition, timeo)                            \
  84 ({                                                                      \
  85         int __ret = 0;                                                  \
  86         if (!(condition))                                               \
  87                 __iucv_sock_wait(sk, condition, timeo, __ret);          \
  88         __ret;                                                          \
  89 })
  90
  91 static void iucv_sock_kill(struct sock *sk);
  92 static void iucv_sock_close(struct sock *sk);
  93 static void iucv_sever_path(struct sock *, int);
  94
  95 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
  96         struct packet_type *pt, struct net_device *orig_dev);
  97 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
  98                    struct sk_buff *skb, u8 flags);
  99 static void afiucv_hs_callback_txnotify(struct sk_buff *, enum iucv_tx_notify);
 100
 101 /* Call Back functions */
 102 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
 103 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
 104 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
 105 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
 106                                  u8 ipuser[16]);
 107 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
 108 static void iucv_callback_shutdown(struct iucv_path *, u8 ipuser[16]);
 109
 110 static struct iucv_sock_list iucv_sk_list = {
 111         .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
 112         .autobind_name = ATOMIC_INIT(0)
 113 };
 114
 115 static struct iucv_handler af_iucv_handler = {
 116         .path_pending     = iucv_callback_connreq,
 117         .path_complete    = iucv_callback_connack,
 118         .path_severed     = iucv_callback_connrej,
 119         .message_pending  = iucv_callback_rx,
 120         .message_complete = iucv_callback_txdone,
 121         .path_quiesced    = iucv_callback_shutdown,
 122 };
 123
 124 static inline void high_nmcpy(unsigned char *dst, char *src)
 125 {
 126        memcpy(dst, src, 8);
 127 }
 128
 129 static inline void low_nmcpy(unsigned char *dst, char *src)
 130 {
 131        memcpy(&dst[8], src, 8);
 132 }
 133
 134 static int afiucv_pm_prepare(struct device *dev)
 135 {
 136 #ifdef CONFIG_PM_DEBUG
 137         printk(KERN_WARNING "afiucv_pm_prepare\n");
 138 #endif
 139         return 0;
 140 }
 141
 142 static void afiucv_pm_complete(struct device *dev)
 143 {
 144 #ifdef CONFIG_PM_DEBUG
 145         printk(KERN_WARNING "afiucv_pm_complete\n");
 146 #endif
 147 }
 148
 149 /**
 150  * afiucv_pm_freeze() - Freeze PM callback
 151  * @dev:        AFIUCV dummy device
 152  *
 153  * Sever all established IUCV communication pathes
 154  */
 155 static int afiucv_pm_freeze(struct device *dev)
 156 {
 157         struct iucv_sock *iucv;
 158         struct sock *sk;
 159         struct hlist_node *node;
 160         int err = 0;
 161
 162 #ifdef CONFIG_PM_DEBUG
 163         printk(KERN_WARNING "afiucv_pm_freeze\n");
 164 #endif
 165         read_lock(&iucv_sk_list.lock);
 166         sk_for_each(sk, node, &iucv_sk_list.head) {
 167                 iucv = iucv_sk(sk);
 168                 switch (sk->sk_state) {
 169                 case IUCV_DISCONN:
 170                 case IUCV_CLOSING:
 171                 case IUCV_CONNECTED:
 172                         iucv_sever_path(sk, 0);
 173                         break;
 174                 case IUCV_OPEN:
 175                 case IUCV_BOUND:
 176                 case IUCV_LISTEN:
 177                 case IUCV_CLOSED:
 178                 default:
 179                         break;
 180                 }
 181                 skb_queue_purge(&iucv->send_skb_q);
 182                 skb_queue_purge(&iucv->backlog_skb_q);
 183         }
 184         read_unlock(&iucv_sk_list.lock);
 185         return err;
 186 }
 187
 188 /**
 189  * afiucv_pm_restore_thaw() - Thaw and restore PM callback
 190  * @dev:        AFIUCV dummy device
 191  *
 192  * socket clean up after freeze
 193  */
 194 static int afiucv_pm_restore_thaw(struct device *dev)
 195 {
 196         struct sock *sk;
 197         struct hlist_node *node;
 198
 199 #ifdef CONFIG_PM_DEBUG
 200         printk(KERN_WARNING "afiucv_pm_restore_thaw\n");
 201 #endif
 202         read_lock(&iucv_sk_list.lock);
 203         sk_for_each(sk, node, &iucv_sk_list.head) {
 204                 switch (sk->sk_state) {
 205                 case IUCV_CONNECTED:
 206                         sk->sk_err = EPIPE;
 207                         sk->sk_state = IUCV_DISCONN;
 208                         sk->sk_state_change(sk);
 209                         break;
 210                 case IUCV_DISCONN:
 211                 case IUCV_CLOSING:
 212                 case IUCV_LISTEN:
 213                 case IUCV_BOUND:
 214                 case IUCV_OPEN:
 215                 default:
 216                         break;
 217                 }
 218         }
 219         read_unlock(&iucv_sk_list.lock);
 220         return 0;
 221 }
 222
 223 static const struct dev_pm_ops afiucv_pm_ops = {
 224         .prepare = afiucv_pm_prepare,
 225         .complete = afiucv_pm_complete,
 226         .freeze = afiucv_pm_freeze,
 227         .thaw = afiucv_pm_restore_thaw,
 228         .restore = afiucv_pm_restore_thaw,
 229 };
 230
 231 static struct device_driver af_iucv_driver = {
 232         .owner = THIS_MODULE,
 233         .name = "afiucv",
 234         .bus  = NULL,
 235         .pm   = &afiucv_pm_ops,
 236 };
 237
 238 /* dummy device used as trigger for PM functions */
 239 static struct device *af_iucv_dev;
 240
 241 /**
 242  * iucv_msg_length() - Returns the length of an iucv message.
 243  * @msg:        Pointer to struct iucv_message, MUST NOT be NULL
 244  *
 245  * The function returns the length of the specified iucv message @msg of data
 246  * stored in a buffer and of data stored in the parameter list (PRMDATA).
 247  *
 248  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
 249  * data:
 250  *      PRMDATA[0..6]   socket data (max 7 bytes);
 251  *      PRMDATA[7]      socket data length value (len is 0xff - PRMDATA[7])
 252  *
 253  * The socket data length is computed by subtracting the socket data length
 254  * value from 0xFF.
 255  * If the socket data len is greater 7, then PRMDATA can be used for special
 256  * notifications (see iucv_sock_shutdown); and further,
 257  * if the socket data len is > 7, the function returns 8.
 258  *
 259  * Use this function to allocate socket buffers to store iucv message data.
 260  */
 261 static inline size_t iucv_msg_length(struct iucv_message *msg)
 262 {
 263         size_t datalen;
 264
 265         if (msg->flags & IUCV_IPRMDATA) {
 266                 datalen = 0xff - msg->rmmsg[7];
 267                 return (datalen < 8) ? datalen : 8;
 268         }
 269         return msg->length;
 270 }
 271
 272 /**
 273  * iucv_sock_in_state() - check for specific states
 274  * @sk:         sock structure
 275  * @state:      first iucv sk state
 276  * @state:      second iucv sk state
 277  *
 278  * Returns true if the socket in either in the first or second state.
 279  */
 280 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
 281 {
 282         return (sk->sk_state == state || sk->sk_state == state2);
 283 }
 284
 285 /**
 286  * iucv_below_msglim() - function to check if messages can be sent
 287  * @sk:         sock structure
 288  *
 289  * Returns true if the send queue length is lower than the message limit.
 290  * Always returns true if the socket is not connected (no iucv path for
 291  * checking the message limit).
 292  */
 293 static inline int iucv_below_msglim(struct sock *sk)
 294 {
 295         struct iucv_sock *iucv = iucv_sk(sk);
 296
 297         if (sk->sk_state != IUCV_CONNECTED)
 298                 return 1;
 299         if (iucv->transport == AF_IUCV_TRANS_IUCV)
 300                 return (skb_queue_len(&iucv->send_skb_q) < iucv->path->msglim);
 301         else
 302                 return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
 303                         (atomic_read(&iucv->pendings) <= 0));
 304 }
 305
 306 /**
 307  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
 308  */
 309 static void iucv_sock_wake_msglim(struct sock *sk)
 310 {
 311         struct socket_wq *wq;
 312
 313         rcu_read_lock();
 314         wq = rcu_dereference(sk->sk_wq);
 315         if (wq_has_sleeper(wq))
 316                 wake_up_interruptible_all(&wq->wait);
 317         sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
 318         rcu_read_unlock();
 319 }
 320
 321 /**
 322  * afiucv_hs_send() - send a message through HiperSockets transport
 323  */
 324 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
 325                    struct sk_buff *skb, u8 flags)
 326 {
 327         struct iucv_sock *iucv = iucv_sk(sock);
 328         struct af_iucv_trans_hdr *phs_hdr;
 329         struct sk_buff *nskb;
 330         int err, confirm_recv = 0;
 331
 332         memset(skb->head, 0, ETH_HLEN);
 333         phs_hdr = (struct af_iucv_trans_hdr *)skb_push(skb,
 334                                         sizeof(struct af_iucv_trans_hdr));
 335         skb_reset_mac_header(skb);
 336         skb_reset_network_header(skb);
 337         skb_push(skb, ETH_HLEN);
 338         skb_reset_mac_header(skb);
 339         memset(phs_hdr, 0, sizeof(struct af_iucv_trans_hdr));
 340
 341         phs_hdr->magic = ETH_P_AF_IUCV;
 342         phs_hdr->version = 1;
 343         phs_hdr->flags = flags;
 344         if (flags == AF_IUCV_FLAG_SYN)
 345                 phs_hdr->window = iucv->msglimit;
 346         else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
 347                 confirm_recv = atomic_read(&iucv->msg_recv);
 348                 phs_hdr->window = confirm_recv;
 349                 if (confirm_recv)
 350                         phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
 351         }
 352         memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
 353         memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
 354         memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
 355         memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
 356         ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
 357         ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
 358         ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
 359         ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
 360         if (imsg)
 361                 memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
 362
 363         skb->dev = iucv->hs_dev;
 364         if (!skb->dev)
 365                 return -ENODEV;
 366         if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev))
 367                 return -ENETDOWN;
 368         if (skb->len > skb->dev->mtu) {
 369                 if (sock->sk_type == SOCK_SEQPACKET)
 370                         return -EMSGSIZE;
 371                 else
 372                         skb_trim(skb, skb->dev->mtu);
 373         }
 374         skb->protocol = ETH_P_AF_IUCV;
 375         skb_shinfo(skb)->tx_flags |= SKBTX_DRV_NEEDS_SK_REF;
 376         nskb = skb_clone(skb, GFP_ATOMIC);
 377         if (!nskb)
 378                 return -ENOMEM;
 379         skb_queue_tail(&iucv->send_skb_q, nskb);
 380         err = dev_queue_xmit(skb);
 381         if (net_xmit_eval(err)) {
 382                 skb_unlink(nskb, &iucv->send_skb_q);
 383                 kfree_skb(nskb);
 384         } else {
 385                 atomic_sub(confirm_recv, &iucv->msg_recv);
 386                 WARN_ON(atomic_read(&iucv->msg_recv) < 0);
 387         }
 388         return net_xmit_eval(err);
 389 }
 390
 391 static struct sock *__iucv_get_sock_by_name(char *nm)
 392 {
 393         struct sock *sk;
 394         struct hlist_node *node;
 395
 396         sk_for_each(sk, node, &iucv_sk_list.head)
 397                 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
 398                         return sk;
 399
 400         return NULL;
 401 }
 402
 403 static void iucv_sock_destruct(struct sock *sk)
 404 {
 405         skb_queue_purge(&sk->sk_receive_queue);
 406         skb_queue_purge(&sk->sk_error_queue);
 407
 408         sk_mem_reclaim(sk);
 409
 410         if (!sock_flag(sk, SOCK_DEAD)) {
 411                 pr_err("Attempt to release alive iucv socket %p\n", sk);
 412                 return;
 413         }
 414
 415         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
 416         WARN_ON(atomic_read(&sk->sk_wmem_alloc));
 417         WARN_ON(sk->sk_wmem_queued);
 418         WARN_ON(sk->sk_forward_alloc);
 419 }
 420
 421 /* Cleanup Listen */
 422 static void iucv_sock_cleanup_listen(struct sock *parent)
 423 {
 424         struct sock *sk;
 425
 426         /* Close non-accepted connections */
 427         while ((sk = iucv_accept_dequeue(parent, NULL))) {
 428                 iucv_sock_close(sk);
 429                 iucv_sock_kill(sk);
 430         }
 431
 432         parent->sk_state = IUCV_CLOSED;
 433 }
 434
 435 /* Kill socket (only if zapped and orphaned) */
 436 static void iucv_sock_kill(struct sock *sk)
 437 {
 438         if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
 439                 return;
 440
 441         iucv_sock_unlink(&iucv_sk_list, sk);
 442         sock_set_flag(sk, SOCK_DEAD);
 443         sock_put(sk);
 444 }
 445
 446 /* Terminate an IUCV path */
 447 static void iucv_sever_path(struct sock *sk, int with_user_data)
 448 {
 449         unsigned char user_data[16];
 450         struct iucv_sock *iucv = iucv_sk(sk);
 451         struct iucv_path *path = iucv->path;
 452
 453         if (iucv->path) {
 454                 iucv->path = NULL;
 455                 if (with_user_data) {
 456                         low_nmcpy(user_data, iucv->src_name);
 457                         high_nmcpy(user_data, iucv->dst_name);
 458                         ASCEBC(user_data, sizeof(user_data));
 459                         pr_iucv->path_sever(path, user_data);
 460                 } else
 461                         pr_iucv->path_sever(path, NULL);
 462                 iucv_path_free(path);
 463         }
 464 }
 465
 466 /* Send FIN through an IUCV socket for HIPER transport */
 467 static int iucv_send_ctrl(struct sock *sk, u8 flags)
 468 {
 469         int err = 0;
 470         int blen;
 471         struct sk_buff *skb;
 472
 473         blen = sizeof(struct af_iucv_trans_hdr) + ETH_HLEN;
 474         skb = sock_alloc_send_skb(sk, blen, 1, &err);
 475         if (skb) {
 476                 skb_reserve(skb, blen);
 477                 err = afiucv_hs_send(NULL, sk, skb, flags);
 478         }
 479         return err;
 480 }
 481
 482 /* Close an IUCV socket */
 483 static void iucv_sock_close(struct sock *sk)
 484 {
 485         struct iucv_sock *iucv = iucv_sk(sk);
 486         unsigned long timeo;
 487         int err = 0;
 488
 489         lock_sock(sk);
 490
 491         switch (sk->sk_state) {
 492         case IUCV_LISTEN:
 493                 iucv_sock_cleanup_listen(sk);
 494                 break;
 495
 496         case IUCV_CONNECTED:
 497                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
 498                         err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
 499                         sk->sk_state = IUCV_DISCONN;
 500                         sk->sk_state_change(sk);
 501                 }
 502         case IUCV_DISCONN:   /* fall through */
 503                 sk->sk_state = IUCV_CLOSING;
 504                 sk->sk_state_change(sk);
 505
 506                 if (!err && !skb_queue_empty(&iucv->send_skb_q)) {
 507                         if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
 508                                 timeo = sk->sk_lingertime;
 509                         else
 510                                 timeo = IUCV_DISCONN_TIMEOUT;
 511                         iucv_sock_wait(sk,
 512                                         iucv_sock_in_state(sk, IUCV_CLOSED, 0),
 513                                         timeo);
 514                 }
 515
 516         case IUCV_CLOSING:   /* fall through */
 517                 sk->sk_state = IUCV_CLOSED;
 518                 sk->sk_state_change(sk);
 519
 520                 sk->sk_err = ECONNRESET;
 521                 sk->sk_state_change(sk);
 522
 523                 skb_queue_purge(&iucv->send_skb_q);
 524                 skb_queue_purge(&iucv->backlog_skb_q);
 525
 526         default:   /* fall through */
 527                 iucv_sever_path(sk, 1);
 528         }
 529
 530         if (iucv->hs_dev) {
 531                 dev_put(iucv->hs_dev);
 532                 iucv->hs_dev = NULL;
 533                 sk->sk_bound_dev_if = 0;
 534         }
 535
 536         /* mark socket for deletion by iucv_sock_kill() */
 537         sock_set_flag(sk, SOCK_ZAPPED);
 538
 539         release_sock(sk);
 540 }
 541
 542 static void iucv_sock_init(struct sock *sk, struct sock *parent)
 543 {
 544         if (parent)
 545                 sk->sk_type = parent->sk_type;
 546 }
 547
 548 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
 549 {
 550         struct sock *sk;
 551         struct iucv_sock *iucv;
 552
 553         sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
 554         if (!sk)
 555                 return NULL;
 556         iucv = iucv_sk(sk);
 557
 558         sock_init_data(sock, sk);
 559         INIT_LIST_HEAD(&iucv->accept_q);
 560         spin_lock_init(&iucv->accept_q_lock);
 561         skb_queue_head_init(&iucv->send_skb_q);
 562         INIT_LIST_HEAD(&iucv->message_q.list);
 563         spin_lock_init(&iucv->message_q.lock);
 564         skb_queue_head_init(&iucv->backlog_skb_q);
 565         iucv->send_tag = 0;
 566         atomic_set(&iucv->pendings, 0);
 567         iucv->flags = 0;
 568         iucv->msglimit = 0;
 569         atomic_set(&iucv->msg_sent, 0);
 570         atomic_set(&iucv->msg_recv, 0);
 571         iucv->path = NULL;
 572         iucv->sk_txnotify = afiucv_hs_callback_txnotify;
 573         memset(&iucv->src_user_id , 0, 32);
 574         if (pr_iucv)
 575                 iucv->transport = AF_IUCV_TRANS_IUCV;
 576         else
 577                 iucv->transport = AF_IUCV_TRANS_HIPER;
 578
 579         sk->sk_destruct = iucv_sock_destruct;
 580         sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
 581         sk->sk_allocation = GFP_DMA;
 582
 583         sock_reset_flag(sk, SOCK_ZAPPED);
 584
 585         sk->sk_protocol = proto;
 586         sk->sk_state    = IUCV_OPEN;
 587
 588         iucv_sock_link(&iucv_sk_list, sk);
 589         return sk;
 590 }
 591
 592 /* Create an IUCV socket */
 593 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
 594                             int kern)
 595 {
 596         struct sock *sk;
 597
 598         if (protocol && protocol != PF_IUCV)
 599                 return -EPROTONOSUPPORT;
 600
 601         sock->state = SS_UNCONNECTED;
 602
 603         switch (sock->type) {
 604         case SOCK_STREAM:
 605                 sock->ops = &iucv_sock_ops;
 606                 break;
 607         case SOCK_SEQPACKET:
 608                 /* currently, proto ops can handle both sk types */
 609                 sock->ops = &iucv_sock_ops;
 610                 break;
 611         default:
 612                 return -ESOCKTNOSUPPORT;
 613         }
 614
 615         sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
 616         if (!sk)
 617                 return -ENOMEM;
 618
 619         iucv_sock_init(sk, NULL);
 620
 621         return 0;
 622 }
 623
 624 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
 625 {
 626         write_lock_bh(&l->lock);
 627         sk_add_node(sk, &l->head);
 628         write_unlock_bh(&l->lock);
 629 }
 630
 631 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
 632 {
 633         write_lock_bh(&l->lock);
 634         sk_del_node_init(sk);
 635         write_unlock_bh(&l->lock);
 636 }
 637
 638 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
 639 {
 640         unsigned long flags;
 641         struct iucv_sock *par = iucv_sk(parent);
 642
 643         sock_hold(sk);
 644         spin_lock_irqsave(&par->accept_q_lock, flags);
 645         list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
 646         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 647         iucv_sk(sk)->parent = parent;
 648         sk_acceptq_added(parent);
 649 }
 650
 651 void iucv_accept_unlink(struct sock *sk)
 652 {
 653         unsigned long flags;
 654         struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
 655
 656         spin_lock_irqsave(&par->accept_q_lock, flags);
 657         list_del_init(&iucv_sk(sk)->accept_q);
 658         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 659         sk_acceptq_removed(iucv_sk(sk)->parent);
 660         iucv_sk(sk)->parent = NULL;
 661         sock_put(sk);
 662 }
 663
 664 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
 665 {
 666         struct iucv_sock *isk, *n;
 667         struct sock *sk;
 668
 669         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
 670                 sk = (struct sock *) isk;
 671                 lock_sock(sk);
 672
 673                 if (sk->sk_state == IUCV_CLOSED) {
 674                         iucv_accept_unlink(sk);
 675                         release_sock(sk);
 676                         continue;
 677                 }
 678
 679                 if (sk->sk_state == IUCV_CONNECTED ||
 680                     sk->sk_state == IUCV_DISCONN ||
 681                     !newsock) {
 682                         iucv_accept_unlink(sk);
 683                         if (newsock)
 684                                 sock_graft(sk, newsock);
 685
 686                         release_sock(sk);
 687                         return sk;
 688                 }
 689
 690                 release_sock(sk);
 691         }
 692         return NULL;
 693 }
 694
 695 /* Bind an unbound socket */
 696 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
 697                           int addr_len)
 698 {
 699         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 700         struct sock *sk = sock->sk;
 701         struct iucv_sock *iucv;
 702         int err = 0;
 703         struct net_device *dev;
 704         char uid[9];
 705
 706         /* Verify the input sockaddr */
 707         if (!addr || addr->sa_family != AF_IUCV)
 708                 return -EINVAL;
 709
 710         lock_sock(sk);
 711         if (sk->sk_state != IUCV_OPEN) {
 712                 err = -EBADFD;
 713                 goto done;
 714         }
 715
 716         write_lock_bh(&iucv_sk_list.lock);
 717
 718         iucv = iucv_sk(sk);
 719         if (__iucv_get_sock_by_name(sa->siucv_name)) {
 720                 err = -EADDRINUSE;
 721                 goto done_unlock;
 722         }
 723         if (iucv->path)
 724                 goto done_unlock;
 725
 726         /* Bind the socket */
 727         if (pr_iucv)
 728                 if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
 729                         goto vm_bind; /* VM IUCV transport */
 730
 731         /* try hiper transport */
 732         memcpy(uid, sa->siucv_user_id, sizeof(uid));
 733         ASCEBC(uid, 8);
 734         rcu_read_lock();
 735         for_each_netdev_rcu(&init_net, dev) {
 736                 if (!memcmp(dev->perm_addr, uid, 8)) {
 737                         memcpy(iucv->src_name, sa->siucv_name, 8);
 738                         memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
 739                         sk->sk_bound_dev_if = dev->ifindex;
 740                         iucv->hs_dev = dev;
 741                         dev_hold(dev);
 742                         sk->sk_state = IUCV_BOUND;
 743                         iucv->transport = AF_IUCV_TRANS_HIPER;
 744                         if (!iucv->msglimit)
 745                                 iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
 746                         rcu_read_unlock();
 747                         goto done_unlock;
 748                 }
 749         }
 750         rcu_read_unlock();
 751 vm_bind:
 752         if (pr_iucv) {
 753                 /* use local userid for backward compat */
 754                 memcpy(iucv->src_name, sa->siucv_name, 8);
 755                 memcpy(iucv->src_user_id, iucv_userid, 8);
 756                 sk->sk_state = IUCV_BOUND;
 757                 iucv->transport = AF_IUCV_TRANS_IUCV;
 758                 if (!iucv->msglimit)
 759                         iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 760                 goto done_unlock;
 761         }
 762         /* found no dev to bind */
 763         err = -ENODEV;
 764 done_unlock:
 765         /* Release the socket list lock */
 766         write_unlock_bh(&iucv_sk_list.lock);
 767 done:
 768         release_sock(sk);
 769         return err;
 770 }
 771
 772 /* Automatically bind an unbound socket */
 773 static int iucv_sock_autobind(struct sock *sk)
 774 {
 775         struct iucv_sock *iucv = iucv_sk(sk);
 776         char name[12];
 777         int err = 0;
 778
 779         if (unlikely(!pr_iucv))
 780                 return -EPROTO;
 781
 782         memcpy(iucv->src_user_id, iucv_userid, 8);
 783
 784         write_lock_bh(&iucv_sk_list.lock);
 785
 786         sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
 787         while (__iucv_get_sock_by_name(name)) {
 788                 sprintf(name, "%08x",
 789                         atomic_inc_return(&iucv_sk_list.autobind_name));
 790         }
 791
 792         write_unlock_bh(&iucv_sk_list.lock);
 793
 794         memcpy(&iucv->src_name, name, 8);
 795
 796         if (!iucv->msglimit)
 797                 iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 798
 799         return err;
 800 }
 801
 802 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
 803 {
 804         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 805         struct sock *sk = sock->sk;
 806         struct iucv_sock *iucv = iucv_sk(sk);
 807         unsigned char user_data[16];
 808         int err;
 809
 810         high_nmcpy(user_data, sa->siucv_name);
 811         low_nmcpy(user_data, iucv->src_name);
 812         ASCEBC(user_data, sizeof(user_data));
 813
 814         /* Create path. */
 815         iucv->path = iucv_path_alloc(iucv->msglimit,
 816                                      IUCV_IPRMDATA, GFP_KERNEL);
 817         if (!iucv->path) {
 818                 err = -ENOMEM;
 819                 goto done;
 820         }
 821         err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
 822                                     sa->siucv_user_id, NULL, user_data,
 823                                     sk);
 824         if (err) {
 825                 iucv_path_free(iucv->path);
 826                 iucv->path = NULL;
 827                 switch (err) {
 828                 case 0x0b:      /* Target communicator is not logged on */
 829                         err = -ENETUNREACH;
 830                         break;
 831                 case 0x0d:      /* Max connections for this guest exceeded */
 832                 case 0x0e:      /* Max connections for target guest exceeded */
 833                         err = -EAGAIN;
 834                         break;
 835                 case 0x0f:      /* Missing IUCV authorization */
 836                         err = -EACCES;
 837                         break;
 838                 default:
 839                         err = -ECONNREFUSED;
 840                         break;
 841                 }
 842         }
 843 done:
 844         return err;
 845 }
 846
 847 /* Connect an unconnected socket */
 848 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
 849                              int alen, int flags)
 850 {
 851         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 852         struct sock *sk = sock->sk;
 853         struct iucv_sock *iucv = iucv_sk(sk);
 854         int err;
 855
 856         if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
 857                 return -EINVAL;
 858
 859         if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
 860                 return -EBADFD;
 861
 862         if (sk->sk_state == IUCV_OPEN &&
 863             iucv->transport == AF_IUCV_TRANS_HIPER)
 864                 return -EBADFD; /* explicit bind required */
 865
 866         if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
 867                 return -EINVAL;
 868
 869         if (sk->sk_state == IUCV_OPEN) {
 870                 err = iucv_sock_autobind(sk);
 871                 if (unlikely(err))
 872                         return err;
 873         }
 874
 875         lock_sock(sk);
 876
 877         /* Set the destination information */
 878         memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
 879         memcpy(iucv->dst_name, sa->siucv_name, 8);
 880
 881         if (iucv->transport == AF_IUCV_TRANS_HIPER)
 882                 err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
 883         else
 884                 err = afiucv_path_connect(sock, addr);
 885         if (err)
 886                 goto done;
 887
 888         if (sk->sk_state != IUCV_CONNECTED)
 889                 err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
 890                                                             IUCV_DISCONN),
 891                                      sock_sndtimeo(sk, flags & O_NONBLOCK));
 892
 893         if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
 894                 err = -ECONNREFUSED;
 895
 896         if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
 897                 iucv_sever_path(sk, 0);
 898
 899 done:
 900         release_sock(sk);
 901         return err;
 902 }
 903
 904 /* Move a socket into listening state. */
 905 static int iucv_sock_listen(struct socket *sock, int backlog)
 906 {
 907         struct sock *sk = sock->sk;
 908         int err;
 909
 910         lock_sock(sk);
 911
 912         err = -EINVAL;
 913         if (sk->sk_state != IUCV_BOUND)
 914                 goto done;
 915
 916         if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
 917                 goto done;
 918
 919         sk->sk_max_ack_backlog = backlog;
 920         sk->sk_ack_backlog = 0;
 921         sk->sk_state = IUCV_LISTEN;
 922         err = 0;
 923
 924 done:
 925         release_sock(sk);
 926         return err;
 927 }
 928
 929 /* Accept a pending connection */
 930 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
 931                             int flags)
 932 {
 933         DECLARE_WAITQUEUE(wait, current);
 934         struct sock *sk = sock->sk, *nsk;
 935         long timeo;
 936         int err = 0;
 937
 938         lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 939
 940         if (sk->sk_state != IUCV_LISTEN) {
 941                 err = -EBADFD;
 942                 goto done;
 943         }
 944
 945         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
 946
 947         /* Wait for an incoming connection */
 948         add_wait_queue_exclusive(sk_sleep(sk), &wait);
 949         while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
 950                 set_current_state(TASK_INTERRUPTIBLE);
 951                 if (!timeo) {
 952                         err = -EAGAIN;
 953                         break;
 954                 }
 955
 956                 release_sock(sk);
 957                 timeo = schedule_timeout(timeo);
 958                 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 959
 960                 if (sk->sk_state != IUCV_LISTEN) {
 961                         err = -EBADFD;
 962                         break;
 963                 }
 964
 965                 if (signal_pending(current)) {
 966                         err = sock_intr_errno(timeo);
 967                         break;
 968                 }
 969         }
 970
 971         set_current_state(TASK_RUNNING);
 972         remove_wait_queue(sk_sleep(sk), &wait);
 973
 974         if (err)
 975                 goto done;
 976
 977         newsock->state = SS_CONNECTED;
 978
 979 done:
 980         release_sock(sk);
 981         return err;
 982 }
 983
 984 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
 985                              int *len, int peer)
 986 {
 987         struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
 988         struct sock *sk = sock->sk;
 989         struct iucv_sock *iucv = iucv_sk(sk);
 990
 991         addr->sa_family = AF_IUCV;
 992         *len = sizeof(struct sockaddr_iucv);
 993
 994         if (peer) {
 995                 memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
 996                 memcpy(siucv->siucv_name, iucv->dst_name, 8);
 997         } else {
 998                 memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
 999                 memcpy(siucv->siucv_name, iucv->src_name, 8);
1000         }
1001         memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
1002         memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
1003         memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
1004
1005         return 0;
1006 }
1007
1008 /**
1009  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
1010  * @path:       IUCV path
1011  * @msg:        Pointer to a struct iucv_message
1012  * @skb:        The socket data to send, skb->len MUST BE <= 7
1013  *
1014  * Send the socket data in the parameter list in the iucv message
1015  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
1016  * list and the socket data len at index 7 (last byte).
1017  * See also iucv_msg_length().
1018  *
1019  * Returns the error code from the iucv_message_send() call.
1020  */
1021 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
1022                           struct sk_buff *skb)
1023 {
1024         u8 prmdata[8];
1025
1026         memcpy(prmdata, (void *) skb->data, skb->len);
1027         prmdata[7] = 0xff - (u8) skb->len;
1028         return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
1029                                  (void *) prmdata, 8);
1030 }
1031
1032 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1033                              struct msghdr *msg, size_t len)
1034 {
1035         struct sock *sk = sock->sk;
1036         struct iucv_sock *iucv = iucv_sk(sk);
1037         struct sk_buff *skb;
1038         struct iucv_message txmsg;
1039         struct cmsghdr *cmsg;
1040         int cmsg_done;
1041         long timeo;
1042         char user_id[9];
1043         char appl_id[9];
1044         int err;
1045         int noblock = msg->msg_flags & MSG_DONTWAIT;
1046
1047         err = sock_error(sk);
1048         if (err)
1049                 return err;
1050
1051         if (msg->msg_flags & MSG_OOB)
1052                 return -EOPNOTSUPP;
1053
1054         /* SOCK_SEQPACKET: we do not support segmented records */
1055         if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
1056                 return -EOPNOTSUPP;
1057
1058         lock_sock(sk);
1059
1060         if (sk->sk_shutdown & SEND_SHUTDOWN) {
1061                 err = -EPIPE;
1062                 goto out;
1063         }
1064
1065         /* Return if the socket is not in connected state */
1066         if (sk->sk_state != IUCV_CONNECTED) {
1067                 err = -ENOTCONN;
1068                 goto out;
1069         }
1070
1071         /* initialize defaults */
1072         cmsg_done   = 0;        /* check for duplicate headers */
1073         txmsg.class = 0;
1074
1075         /* iterate over control messages */
1076         for (cmsg = CMSG_FIRSTHDR(msg); cmsg;
1077                 cmsg = CMSG_NXTHDR(msg, cmsg)) {
1078
1079                 if (!CMSG_OK(msg, cmsg)) {
1080                         err = -EINVAL;
1081                         goto out;
1082                 }
1083
1084                 if (cmsg->cmsg_level != SOL_IUCV)
1085                         continue;
1086
1087                 if (cmsg->cmsg_type & cmsg_done) {
1088                         err = -EINVAL;
1089                         goto out;
1090                 }
1091                 cmsg_done |= cmsg->cmsg_type;
1092
1093                 switch (cmsg->cmsg_type) {
1094                 case SCM_IUCV_TRGCLS:
1095                         if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
1096                                 err = -EINVAL;
1097                                 goto out;
1098                         }
1099
1100                         /* set iucv message target class */
1101                         memcpy(&txmsg.class,
1102                                 (void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
1103
1104                         break;
1105
1106                 default:
1107                         err = -EINVAL;
1108                         goto out;
1109                         break;
1110                 }
1111         }
1112
1113         /* allocate one skb for each iucv message:
1114          * this is fine for SOCK_SEQPACKET (unless we want to support
1115          * segmented records using the MSG_EOR flag), but
1116          * for SOCK_STREAM we might want to improve it in future */
1117         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1118                 skb = sock_alloc_send_skb(sk,
1119                         len + sizeof(struct af_iucv_trans_hdr) + ETH_HLEN,
1120                         noblock, &err);
1121         else
1122                 skb = sock_alloc_send_skb(sk, len, noblock, &err);
1123         if (!skb) {
1124                 err = -ENOMEM;
1125                 goto out;
1126         }
1127         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1128                 skb_reserve(skb, sizeof(struct af_iucv_trans_hdr) + ETH_HLEN);
1129         if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1130                 err = -EFAULT;
1131                 goto fail;
1132         }
1133
1134         /* wait if outstanding messages for iucv path has reached */
1135         timeo = sock_sndtimeo(sk, noblock);
1136         err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1137         if (err)
1138                 goto fail;
1139
1140         /* return -ECONNRESET if the socket is no longer connected */
1141         if (sk->sk_state != IUCV_CONNECTED) {
1142                 err = -ECONNRESET;
1143                 goto fail;
1144         }
1145
1146         /* increment and save iucv message tag for msg_completion cbk */
1147         txmsg.tag = iucv->send_tag++;
1148         memcpy(CB_TAG(skb), &txmsg.tag, CB_TAG_LEN);
1149
1150         if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1151                 atomic_inc(&iucv->msg_sent);
1152                 err = afiucv_hs_send(&txmsg, sk, skb, 0);
1153                 if (err) {
1154                         atomic_dec(&iucv->msg_sent);
1155                         goto fail;
1156                 }
1157                 goto release;
1158         }
1159         skb_queue_tail(&iucv->send_skb_q, skb);
1160
1161         if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags)
1162               && skb->len <= 7) {
1163                 err = iucv_send_iprm(iucv->path, &txmsg, skb);
1164
1165                 /* on success: there is no message_complete callback
1166                  * for an IPRMDATA msg; remove skb from send queue */
1167                 if (err == 0) {
1168                         skb_unlink(skb, &iucv->send_skb_q);
1169                         kfree_skb(skb);
1170                 }
1171
1172                 /* this error should never happen since the
1173                  * IUCV_IPRMDATA path flag is set... sever path */
1174                 if (err == 0x15) {
1175                         pr_iucv->path_sever(iucv->path, NULL);
1176                         skb_unlink(skb, &iucv->send_skb_q);
1177                         err = -EPIPE;
1178                         goto fail;
1179                 }
1180         } else
1181                 err = pr_iucv->message_send(iucv->path, &txmsg, 0, 0,
1182                                         (void *) skb->data, skb->len);
1183         if (err) {
1184                 if (err == 3) {
1185                         user_id[8] = 0;
1186                         memcpy(user_id, iucv->dst_user_id, 8);
1187                         appl_id[8] = 0;
1188                         memcpy(appl_id, iucv->dst_name, 8);
1189                         pr_err("Application %s on z/VM guest %s"
1190                                 " exceeds message limit\n",
1191                                 appl_id, user_id);
1192                         err = -EAGAIN;
1193                 } else
1194                         err = -EPIPE;
1195                 skb_unlink(skb, &iucv->send_skb_q);
1196                 goto fail;
1197         }
1198
1199 release:
1200         release_sock(sk);
1201         return len;
1202
1203 fail:
1204         kfree_skb(skb);
1205 out:
1206         release_sock(sk);
1207         return err;
1208 }
1209
1210 /* iucv_fragment_skb() - Fragment a single IUCV message into multiple skb's
1211  *
1212  * Locking: must be called with message_q.lock held
1213  */
1214 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
1215 {
1216         int dataleft, size, copied = 0;
1217         struct sk_buff *nskb;
1218
1219         dataleft = len;
1220         while (dataleft) {
1221                 if (dataleft >= sk->sk_rcvbuf / 4)
1222                         size = sk->sk_rcvbuf / 4;
1223                 else
1224                         size = dataleft;
1225
1226                 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
1227                 if (!nskb)
1228                         return -ENOMEM;
1229
1230                 /* copy target class to control buffer of new skb */
1231                 memcpy(CB_TRGCLS(nskb), CB_TRGCLS(skb), CB_TRGCLS_LEN);
1232
1233                 /* copy data fragment */
1234                 memcpy(nskb->data, skb->data + copied, size);
1235                 copied += size;
1236                 dataleft -= size;
1237
1238                 skb_reset_transport_header(nskb);
1239                 skb_reset_network_header(nskb);
1240                 nskb->len = size;
1241
1242                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
1243         }
1244
1245         return 0;
1246 }
1247
1248 /* iucv_process_message() - Receive a single outstanding IUCV message
1249  *
1250  * Locking: must be called with message_q.lock held
1251  */
1252 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1253                                  struct iucv_path *path,
1254                                  struct iucv_message *msg)
1255 {
1256         int rc;
1257         unsigned int len;
1258
1259         len = iucv_msg_length(msg);
1260
1261         /* store msg target class in the second 4 bytes of skb ctrl buffer */
1262         /* Note: the first 4 bytes are reserved for msg tag */
1263         memcpy(CB_TRGCLS(skb), &msg->class, CB_TRGCLS_LEN);
1264
1265         /* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1266         if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1267                 if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1268                         skb->data = NULL;
1269                         skb->len = 0;
1270                 }
1271         } else {
1272                 rc = pr_iucv->message_receive(path, msg,
1273                                               msg->flags & IUCV_IPRMDATA,
1274                                               skb->data, len, NULL);
1275                 if (rc) {
1276                         kfree_skb(skb);
1277                         return;
1278                 }
1279                 /* we need to fragment iucv messages for SOCK_STREAM only;
1280                  * for SOCK_SEQPACKET, it is only relevant if we support
1281                  * record segmentation using MSG_EOR (see also recvmsg()) */
1282                 if (sk->sk_type == SOCK_STREAM &&
1283                     skb->truesize >= sk->sk_rcvbuf / 4) {
1284                         rc = iucv_fragment_skb(sk, skb, len);
1285                         kfree_skb(skb);
1286                         skb = NULL;
1287                         if (rc) {
1288                                 pr_iucv->path_sever(path, NULL);
1289                                 return;
1290                         }
1291                         skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
1292                 } else {
1293                         skb_reset_transport_header(skb);
1294                         skb_reset_network_header(skb);
1295                         skb->len = len;
1296                 }
1297         }
1298
1299         if (sock_queue_rcv_skb(sk, skb))
1300                 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
1301 }
1302
1303 /* iucv_process_message_q() - Process outstanding IUCV messages
1304  *
1305  * Locking: must be called with message_q.lock held
1306  */
1307 static void iucv_process_message_q(struct sock *sk)
1308 {
1309         struct iucv_sock *iucv = iucv_sk(sk);
1310         struct sk_buff *skb;
1311         struct sock_msg_q *p, *n;
1312
1313         list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1314                 skb = alloc_skb(iucv_msg_length(&p->msg), GFP_ATOMIC | GFP_DMA);
1315                 if (!skb)
1316                         break;
1317                 iucv_process_message(sk, skb, p->path, &p->msg);
1318                 list_del(&p->list);
1319                 kfree(p);
1320                 if (!skb_queue_empty(&iucv->backlog_skb_q))
1321                         break;
1322         }
1323 }
1324
1325 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
1326                              struct msghdr *msg, size_t len, int flags)
1327 {
1328         int noblock = flags & MSG_DONTWAIT;
1329         struct sock *sk = sock->sk;
1330         struct iucv_sock *iucv = iucv_sk(sk);
1331         unsigned int copied, rlen;
1332         struct sk_buff *skb, *rskb, *cskb;
1333         int err = 0;
1334
1335         if ((sk->sk_state == IUCV_DISCONN) &&
1336             skb_queue_empty(&iucv->backlog_skb_q) &&
1337             skb_queue_empty(&sk->sk_receive_queue) &&
1338             list_empty(&iucv->message_q.list))
1339                 return 0;
1340
1341         if (flags & (MSG_OOB))
1342                 return -EOPNOTSUPP;
1343
1344         /* receive/dequeue next skb:
1345          * the function understands MSG_PEEK and, thus, does not dequeue skb */
1346         skb = skb_recv_datagram(sk, flags, noblock, &err);
1347         if (!skb) {
1348                 if (sk->sk_shutdown & RCV_SHUTDOWN)
1349                         return 0;
1350                 return err;
1351         }
1352
1353         rlen   = skb->len;              /* real length of skb */
1354         copied = min_t(unsigned int, rlen, len);
1355         if (!rlen)
1356                 sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1357
1358         cskb = skb;
1359         if (skb_copy_datagram_iovec(cskb, 0, msg->msg_iov, copied)) {
1360                 if (!(flags & MSG_PEEK))
1361                         skb_queue_head(&sk->sk_receive_queue, skb);
1362                 return -EFAULT;
1363         }
1364
1365         /* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1366         if (sk->sk_type == SOCK_SEQPACKET) {
1367                 if (copied < rlen)
1368                         msg->msg_flags |= MSG_TRUNC;
1369                 /* each iucv message contains a complete record */
1370                 msg->msg_flags |= MSG_EOR;
1371         }
1372
1373         /* create control message to store iucv msg target class:
1374          * get the trgcls from the control buffer of the skb due to
1375          * fragmentation of original iucv message. */
1376         err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1377                         CB_TRGCLS_LEN, CB_TRGCLS(skb));
1378         if (err) {
1379                 if (!(flags & MSG_PEEK))
1380                         skb_queue_head(&sk->sk_receive_queue, skb);
1381                 return err;
1382         }
1383
1384         /* Mark read part of skb as used */
1385         if (!(flags & MSG_PEEK)) {
1386
1387                 /* SOCK_STREAM: re-queue skb if it contains unreceived data */
1388                 if (sk->sk_type == SOCK_STREAM) {
1389                         skb_pull(skb, copied);
1390                         if (skb->len) {
1391                                 skb_queue_head(&sk->sk_receive_queue, skb);
1392                                 goto done;
1393                         }
1394                 }
1395
1396                 kfree_skb(skb);
1397                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1398                         atomic_inc(&iucv->msg_recv);
1399                         if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1400                                 WARN_ON(1);
1401                                 iucv_sock_close(sk);
1402                                 return -EFAULT;
1403                         }
1404                 }
1405
1406                 /* Queue backlog skbs */
1407                 spin_lock_bh(&iucv->message_q.lock);
1408                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1409                 while (rskb) {
1410                         if (sock_queue_rcv_skb(sk, rskb)) {
1411                                 skb_queue_head(&iucv->backlog_skb_q,
1412                                                 rskb);
1413                                 break;
1414                         } else {
1415                                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1416                         }
1417                 }
1418                 if (skb_queue_empty(&iucv->backlog_skb_q)) {
1419                         if (!list_empty(&iucv->message_q.list))
1420                                 iucv_process_message_q(sk);
1421                         if (atomic_read(&iucv->msg_recv) >=
1422                                                         iucv->msglimit / 2) {
1423                                 err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1424                                 if (err) {
1425                                         sk->sk_state = IUCV_DISCONN;
1426                                         sk->sk_state_change(sk);
1427                                 }
1428                         }
1429                 }
1430                 spin_unlock_bh(&iucv->message_q.lock);
1431         }
1432
1433 done:
1434         /* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1435         if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1436                 copied = rlen;
1437
1438         return copied;
1439 }
1440
1441 static inline unsigned int iucv_accept_poll(struct sock *parent)
1442 {
1443         struct iucv_sock *isk, *n;
1444         struct sock *sk;
1445
1446         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1447                 sk = (struct sock *) isk;
1448
1449                 if (sk->sk_state == IUCV_CONNECTED)
1450                         return POLLIN | POLLRDNORM;
1451         }
1452
1453         return 0;
1454 }
1455
1456 unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
1457                             poll_table *wait)
1458 {
1459         struct sock *sk = sock->sk;
1460         unsigned int mask = 0;
1461
1462         sock_poll_wait(file, sk_sleep(sk), wait);
1463
1464         if (sk->sk_state == IUCV_LISTEN)
1465                 return iucv_accept_poll(sk);
1466
1467         if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1468                 mask |= POLLERR;
1469
1470         if (sk->sk_shutdown & RCV_SHUTDOWN)
1471                 mask |= POLLRDHUP;
1472
1473         if (sk->sk_shutdown == SHUTDOWN_MASK)
1474                 mask |= POLLHUP;
1475
1476         if (!skb_queue_empty(&sk->sk_receive_queue) ||
1477             (sk->sk_shutdown & RCV_SHUTDOWN))
1478                 mask |= POLLIN | POLLRDNORM;
1479
1480         if (sk->sk_state == IUCV_CLOSED)
1481                 mask |= POLLHUP;
1482
1483         if (sk->sk_state == IUCV_DISCONN)
1484                 mask |= POLLIN;
1485
1486         if (sock_writeable(sk) && iucv_below_msglim(sk))
1487                 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1488         else
1489                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1490
1491         return mask;
1492 }
1493
1494 static int iucv_sock_shutdown(struct socket *sock, int how)
1495 {
1496         struct sock *sk = sock->sk;
1497         struct iucv_sock *iucv = iucv_sk(sk);
1498         struct iucv_message txmsg;
1499         int err = 0;
1500
1501         how++;
1502
1503         if ((how & ~SHUTDOWN_MASK) || !how)
1504                 return -EINVAL;
1505
1506         lock_sock(sk);
1507         switch (sk->sk_state) {
1508         case IUCV_LISTEN:
1509         case IUCV_DISCONN:
1510         case IUCV_CLOSING:
1511         case IUCV_CLOSED:
1512                 err = -ENOTCONN;
1513                 goto fail;
1514         default:
1515                 break;
1516         }
1517
1518         if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
1519                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1520                         txmsg.class = 0;
1521                         txmsg.tag = 0;
1522                         err = pr_iucv->message_send(iucv->path, &txmsg,
1523                                 IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1524                         if (err) {
1525                                 switch (err) {
1526                                 case 1:
1527                                         err = -ENOTCONN;
1528                                         break;
1529                                 case 2:
1530                                         err = -ECONNRESET;
1531                                         break;
1532                                 default:
1533                                         err = -ENOTCONN;
1534                                         break;
1535                                 }
1536                         }
1537                 } else
1538                         iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1539         }
1540
1541         sk->sk_shutdown |= how;
1542         if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1543                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1544                         err = pr_iucv->path_quiesce(iucv->path, NULL);
1545                         if (err)
1546                                 err = -ENOTCONN;
1547 /*                      skb_queue_purge(&sk->sk_receive_queue); */
1548                 }
1549                 skb_queue_purge(&sk->sk_receive_queue);
1550         }
1551
1552         /* Wake up anyone sleeping in poll */
1553         sk->sk_state_change(sk);
1554
1555 fail:
1556         release_sock(sk);
1557         return err;
1558 }
1559
1560 static int iucv_sock_release(struct socket *sock)
1561 {
1562         struct sock *sk = sock->sk;
1563         int err = 0;
1564
1565         if (!sk)
1566                 return 0;
1567
1568         iucv_sock_close(sk);
1569
1570         sock_orphan(sk);
1571         iucv_sock_kill(sk);
1572         return err;
1573 }
1574
1575 /* getsockopt and setsockopt */
1576 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1577                                 char __user *optval, unsigned int optlen)
1578 {
1579         struct sock *sk = sock->sk;
1580         struct iucv_sock *iucv = iucv_sk(sk);
1581         int val;
1582         int rc;
1583
1584         if (level != SOL_IUCV)
1585                 return -ENOPROTOOPT;
1586
1587         if (optlen < sizeof(int))
1588                 return -EINVAL;
1589
1590         if (get_user(val, (int __user *) optval))
1591                 return -EFAULT;
1592
1593         rc = 0;
1594
1595         lock_sock(sk);
1596         switch (optname) {
1597         case SO_IPRMDATA_MSG:
1598                 if (val)
1599                         iucv->flags |= IUCV_IPRMDATA;
1600                 else
1601                         iucv->flags &= ~IUCV_IPRMDATA;
1602                 break;
1603         case SO_MSGLIMIT:
1604                 switch (sk->sk_state) {
1605                 case IUCV_OPEN:
1606                 case IUCV_BOUND:
1607                         if (val < 1 || val > (u16)(~0))
1608                                 rc = -EINVAL;
1609                         else
1610                                 iucv->msglimit = val;
1611                         break;
1612                 default:
1613                         rc = -EINVAL;
1614                         break;
1615                 }
1616                 break;
1617         default:
1618                 rc = -ENOPROTOOPT;
1619                 break;
1620         }
1621         release_sock(sk);
1622
1623         return rc;
1624 }
1625
1626 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1627                                 char __user *optval, int __user *optlen)
1628 {
1629         struct sock *sk = sock->sk;
1630         struct iucv_sock *iucv = iucv_sk(sk);
1631         unsigned int val;
1632         int len;
1633
1634         if (level != SOL_IUCV)
1635                 return -ENOPROTOOPT;
1636
1637         if (get_user(len, optlen))
1638                 return -EFAULT;
1639
1640         if (len < 0)
1641                 return -EINVAL;
1642
1643         len = min_t(unsigned int, len, sizeof(int));
1644
1645         switch (optname) {
1646         case SO_IPRMDATA_MSG:
1647                 val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1648                 break;
1649         case SO_MSGLIMIT:
1650                 lock_sock(sk);
1651                 val = (iucv->path != NULL) ? iucv->path->msglim /* connected */
1652                                            : iucv->msglimit;    /* default */
1653                 release_sock(sk);
1654                 break;
1655         case SO_MSGSIZE:
1656                 if (sk->sk_state == IUCV_OPEN)
1657                         return -EBADFD;
1658                 val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1659                                 sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1660                                 0x7fffffff;
1661                 break;
1662         default:
1663                 return -ENOPROTOOPT;
1664         }
1665
1666         if (put_user(len, optlen))
1667                 return -EFAULT;
1668         if (copy_to_user(optval, &val, len))
1669                 return -EFAULT;
1670
1671         return 0;
1672 }
1673
1674
1675 /* Callback wrappers - called from iucv base support */
1676 static int iucv_callback_connreq(struct iucv_path *path,
1677                                  u8 ipvmid[8], u8 ipuser[16])
1678 {
1679         unsigned char user_data[16];
1680         unsigned char nuser_data[16];
1681         unsigned char src_name[8];
1682         struct hlist_node *node;
1683         struct sock *sk, *nsk;
1684         struct iucv_sock *iucv, *niucv;
1685         int err;
1686
1687         memcpy(src_name, ipuser, 8);
1688         EBCASC(src_name, 8);
1689         /* Find out if this path belongs to af_iucv. */
1690         read_lock(&iucv_sk_list.lock);
1691         iucv = NULL;
1692         sk = NULL;
1693         sk_for_each(sk, node, &iucv_sk_list.head)
1694                 if (sk->sk_state == IUCV_LISTEN &&
1695                     !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1696                         /*
1697                          * Found a listening socket with
1698                          * src_name == ipuser[0-7].
1699                          */
1700                         iucv = iucv_sk(sk);
1701                         break;
1702                 }
1703         read_unlock(&iucv_sk_list.lock);
1704         if (!iucv)
1705                 /* No socket found, not one of our paths. */
1706                 return -EINVAL;
1707
1708         bh_lock_sock(sk);
1709
1710         /* Check if parent socket is listening */
1711         low_nmcpy(user_data, iucv->src_name);
1712         high_nmcpy(user_data, iucv->dst_name);
1713         ASCEBC(user_data, sizeof(user_data));
1714         if (sk->sk_state != IUCV_LISTEN) {
1715                 err = pr_iucv->path_sever(path, user_data);
1716                 iucv_path_free(path);
1717                 goto fail;
1718         }
1719
1720         /* Check for backlog size */
1721         if (sk_acceptq_is_full(sk)) {
1722                 err = pr_iucv->path_sever(path, user_data);
1723                 iucv_path_free(path);
1724                 goto fail;
1725         }
1726
1727         /* Create the new socket */
1728         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1729         if (!nsk) {
1730                 err = pr_iucv->path_sever(path, user_data);
1731                 iucv_path_free(path);
1732                 goto fail;
1733         }
1734
1735         niucv = iucv_sk(nsk);
1736         iucv_sock_init(nsk, sk);
1737
1738         /* Set the new iucv_sock */
1739         memcpy(niucv->dst_name, ipuser + 8, 8);
1740         EBCASC(niucv->dst_name, 8);
1741         memcpy(niucv->dst_user_id, ipvmid, 8);
1742         memcpy(niucv->src_name, iucv->src_name, 8);
1743         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1744         niucv->path = path;
1745
1746         /* Call iucv_accept */
1747         high_nmcpy(nuser_data, ipuser + 8);
1748         memcpy(nuser_data + 8, niucv->src_name, 8);
1749         ASCEBC(nuser_data + 8, 8);
1750
1751         /* set message limit for path based on msglimit of accepting socket */
1752         niucv->msglimit = iucv->msglimit;
1753         path->msglim = iucv->msglimit;
1754         err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1755         if (err) {
1756                 iucv_sever_path(nsk, 1);
1757                 iucv_sock_kill(nsk);
1758                 goto fail;
1759         }
1760
1761         iucv_accept_enqueue(sk, nsk);
1762
1763         /* Wake up accept */
1764         nsk->sk_state = IUCV_CONNECTED;
1765         sk->sk_data_ready(sk, 1);
1766         err = 0;
1767 fail:
1768         bh_unlock_sock(sk);
1769         return 0;
1770 }
1771
1772 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1773 {
1774         struct sock *sk = path->private;
1775
1776         sk->sk_state = IUCV_CONNECTED;
1777         sk->sk_state_change(sk);
1778 }
1779
1780 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1781 {
1782         struct sock *sk = path->private;
1783         struct iucv_sock *iucv = iucv_sk(sk);
1784         struct sk_buff *skb;
1785         struct sock_msg_q *save_msg;
1786         int len;
1787
1788         if (sk->sk_shutdown & RCV_SHUTDOWN) {
1789                 pr_iucv->message_reject(path, msg);
1790                 return;
1791         }
1792
1793         spin_lock(&iucv->message_q.lock);
1794
1795         if (!list_empty(&iucv->message_q.list) ||
1796             !skb_queue_empty(&iucv->backlog_skb_q))
1797                 goto save_message;
1798
1799         len = atomic_read(&sk->sk_rmem_alloc);
1800         len += SKB_TRUESIZE(iucv_msg_length(msg));
1801         if (len > sk->sk_rcvbuf)
1802                 goto save_message;
1803
1804         skb = alloc_skb(iucv_msg_length(msg), GFP_ATOMIC | GFP_DMA);
1805         if (!skb)
1806                 goto save_message;
1807
1808         iucv_process_message(sk, skb, path, msg);
1809         goto out_unlock;
1810
1811 save_message:
1812         save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1813         if (!save_msg)
1814                 goto out_unlock;
1815         save_msg->path = path;
1816         save_msg->msg = *msg;
1817
1818         list_add_tail(&save_msg->list, &iucv->message_q.list);
1819
1820 out_unlock:
1821         spin_unlock(&iucv->message_q.lock);
1822 }
1823
1824 static void iucv_callback_txdone(struct iucv_path *path,
1825                                  struct iucv_message *msg)
1826 {
1827         struct sock *sk = path->private;
1828         struct sk_buff *this = NULL;
1829         struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1830         struct sk_buff *list_skb = list->next;
1831         unsigned long flags;
1832
1833         bh_lock_sock(sk);
1834         if (!skb_queue_empty(list)) {
1835                 spin_lock_irqsave(&list->lock, flags);
1836
1837                 while (list_skb != (struct sk_buff *)list) {
1838                         if (!memcmp(&msg->tag, CB_TAG(list_skb), CB_TAG_LEN)) {
1839                                 this = list_skb;
1840                                 break;
1841                         }
1842                         list_skb = list_skb->next;
1843                 }
1844                 if (this)
1845                         __skb_unlink(this, list);
1846
1847                 spin_unlock_irqrestore(&list->lock, flags);
1848
1849                 if (this) {
1850                         kfree_skb(this);
1851                         /* wake up any process waiting for sending */
1852                         iucv_sock_wake_msglim(sk);
1853                 }
1854         }
1855
1856         if (sk->sk_state == IUCV_CLOSING) {
1857                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1858                         sk->sk_state = IUCV_CLOSED;
1859                         sk->sk_state_change(sk);
1860                 }
1861         }
1862         bh_unlock_sock(sk);
1863
1864 }
1865
1866 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1867 {
1868         struct sock *sk = path->private;
1869
1870         if (sk->sk_state == IUCV_CLOSED)
1871                 return;
1872
1873         bh_lock_sock(sk);
1874         iucv_sever_path(sk, 1);
1875         sk->sk_state = IUCV_DISCONN;
1876
1877         sk->sk_state_change(sk);
1878         bh_unlock_sock(sk);
1879 }
1880
1881 /* called if the other communication side shuts down its RECV direction;
1882  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1883  */
1884 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1885 {
1886         struct sock *sk = path->private;
1887
1888         bh_lock_sock(sk);
1889         if (sk->sk_state != IUCV_CLOSED) {
1890                 sk->sk_shutdown |= SEND_SHUTDOWN;
1891                 sk->sk_state_change(sk);
1892         }
1893         bh_unlock_sock(sk);
1894 }
1895
1896 /***************** HiperSockets transport callbacks ********************/
1897 static void afiucv_swap_src_dest(struct sk_buff *skb)
1898 {
1899         struct af_iucv_trans_hdr *trans_hdr =
1900                                 (struct af_iucv_trans_hdr *)skb->data;
1901         char tmpID[8];
1902         char tmpName[8];
1903
1904         ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1905         ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1906         ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1907         ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1908         memcpy(tmpID, trans_hdr->srcUserID, 8);
1909         memcpy(tmpName, trans_hdr->srcAppName, 8);
1910         memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1911         memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1912         memcpy(trans_hdr->destUserID, tmpID, 8);
1913         memcpy(trans_hdr->destAppName, tmpName, 8);
1914         skb_push(skb, ETH_HLEN);
1915         memset(skb->data, 0, ETH_HLEN);
1916 }
1917
1918 /**
1919  * afiucv_hs_callback_syn - react on received SYN
1920  **/
1921 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1922 {
1923         struct sock *nsk;
1924         struct iucv_sock *iucv, *niucv;
1925         struct af_iucv_trans_hdr *trans_hdr;
1926         int err;
1927
1928         iucv = iucv_sk(sk);
1929         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
1930         if (!iucv) {
1931                 /* no sock - connection refused */
1932                 afiucv_swap_src_dest(skb);
1933                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1934                 err = dev_queue_xmit(skb);
1935                 goto out;
1936         }
1937
1938         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1939         bh_lock_sock(sk);
1940         if ((sk->sk_state != IUCV_LISTEN) ||
1941             sk_acceptq_is_full(sk) ||
1942             !nsk) {
1943                 /* error on server socket - connection refused */
1944                 if (nsk)
1945                         sk_free(nsk);
1946                 afiucv_swap_src_dest(skb);
1947                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1948                 err = dev_queue_xmit(skb);
1949                 bh_unlock_sock(sk);
1950                 goto out;
1951         }
1952
1953         niucv = iucv_sk(nsk);
1954         iucv_sock_init(nsk, sk);
1955         niucv->transport = AF_IUCV_TRANS_HIPER;
1956         niucv->msglimit = iucv->msglimit;
1957         if (!trans_hdr->window)
1958                 niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1959         else
1960                 niucv->msglimit_peer = trans_hdr->window;
1961         memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1962         memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1963         memcpy(niucv->src_name, iucv->src_name, 8);
1964         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1965         nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1966         niucv->hs_dev = iucv->hs_dev;
1967         dev_hold(niucv->hs_dev);
1968         afiucv_swap_src_dest(skb);
1969         trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1970         trans_hdr->window = niucv->msglimit;
1971         /* if receiver acks the xmit connection is established */
1972         err = dev_queue_xmit(skb);
1973         if (!err) {
1974                 iucv_accept_enqueue(sk, nsk);
1975                 nsk->sk_state = IUCV_CONNECTED;
1976                 sk->sk_data_ready(sk, 1);
1977         } else
1978                 iucv_sock_kill(nsk);
1979         bh_unlock_sock(sk);
1980
1981 out:
1982         return NET_RX_SUCCESS;
1983 }
1984
1985 /**
1986  * afiucv_hs_callback_synack() - react on received SYN-ACK
1987  **/
1988 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1989 {
1990         struct iucv_sock *iucv = iucv_sk(sk);
1991         struct af_iucv_trans_hdr *trans_hdr =
1992                                         (struct af_iucv_trans_hdr *)skb->data;
1993
1994         if (!iucv)
1995                 goto out;
1996         if (sk->sk_state != IUCV_BOUND)
1997                 goto out;
1998         bh_lock_sock(sk);
1999         iucv->msglimit_peer = trans_hdr->window;
2000         sk->sk_state = IUCV_CONNECTED;
2001         sk->sk_state_change(sk);
2002         bh_unlock_sock(sk);
2003 out:
2004         kfree_skb(skb);
2005         return NET_RX_SUCCESS;
2006 }
2007
2008 /**
2009  * afiucv_hs_callback_synfin() - react on received SYN_FIN
2010  **/
2011 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
2012 {
2013         struct iucv_sock *iucv = iucv_sk(sk);
2014
2015         if (!iucv)
2016                 goto out;
2017         if (sk->sk_state != IUCV_BOUND)
2018                 goto out;
2019         bh_lock_sock(sk);
2020         sk->sk_state = IUCV_DISCONN;
2021         sk->sk_state_change(sk);
2022         bh_unlock_sock(sk);
2023 out:
2024         kfree_skb(skb);
2025         return NET_RX_SUCCESS;
2026 }
2027
2028 /**
2029  * afiucv_hs_callback_fin() - react on received FIN
2030  **/
2031 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
2032 {
2033         struct iucv_sock *iucv = iucv_sk(sk);
2034
2035         /* other end of connection closed */
2036         if (!iucv)
2037                 goto out;
2038         bh_lock_sock(sk);
2039         if (sk->sk_state == IUCV_CONNECTED) {
2040                 sk->sk_state = IUCV_DISCONN;
2041                 sk->sk_state_change(sk);
2042         }
2043         bh_unlock_sock(sk);
2044 out:
2045         kfree_skb(skb);
2046         return NET_RX_SUCCESS;
2047 }
2048
2049 /**
2050  * afiucv_hs_callback_win() - react on received WIN
2051  **/
2052 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
2053 {
2054         struct iucv_sock *iucv = iucv_sk(sk);
2055         struct af_iucv_trans_hdr *trans_hdr =
2056                                         (struct af_iucv_trans_hdr *)skb->data;
2057
2058         if (!iucv)
2059                 return NET_RX_SUCCESS;
2060
2061         if (sk->sk_state != IUCV_CONNECTED)
2062                 return NET_RX_SUCCESS;
2063
2064         atomic_sub(trans_hdr->window, &iucv->msg_sent);
2065         iucv_sock_wake_msglim(sk);
2066         return NET_RX_SUCCESS;
2067 }
2068
2069 /**
2070  * afiucv_hs_callback_rx() - react on received data
2071  **/
2072 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
2073 {
2074         struct iucv_sock *iucv = iucv_sk(sk);
2075
2076         if (!iucv) {
2077                 kfree_skb(skb);
2078                 return NET_RX_SUCCESS;
2079         }
2080
2081         if (sk->sk_state != IUCV_CONNECTED) {
2082                 kfree_skb(skb);
2083                 return NET_RX_SUCCESS;
2084         }
2085
2086         if (sk->sk_shutdown & RCV_SHUTDOWN) {
2087                 kfree_skb(skb);
2088                 return NET_RX_SUCCESS;
2089         }
2090
2091                 /* write stuff from iucv_msg to skb cb */
2092         if (skb->len < sizeof(struct af_iucv_trans_hdr)) {
2093                 kfree_skb(skb);
2094                 return NET_RX_SUCCESS;
2095         }
2096         skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2097         skb_reset_transport_header(skb);
2098         skb_reset_network_header(skb);
2099         spin_lock(&iucv->message_q.lock);
2100         if (skb_queue_empty(&iucv->backlog_skb_q)) {
2101                 if (sock_queue_rcv_skb(sk, skb)) {
2102                         /* handle rcv queue full */
2103                         skb_queue_tail(&iucv->backlog_skb_q, skb);
2104                 }
2105         } else
2106                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2107         spin_unlock(&iucv->message_q.lock);
2108         return NET_RX_SUCCESS;
2109 }
2110
2111 /**
2112  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2113  *                   transport
2114  *                   called from netif RX softirq
2115  **/
2116 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2117         struct packet_type *pt, struct net_device *orig_dev)
2118 {
2119         struct hlist_node *node;
2120         struct sock *sk;
2121         struct iucv_sock *iucv;
2122         struct af_iucv_trans_hdr *trans_hdr;
2123         char nullstring[8];
2124         int err = 0;
2125
2126         skb_pull(skb, ETH_HLEN);
2127         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
2128         EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2129         EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2130         EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2131         EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2132         memset(nullstring, 0, sizeof(nullstring));
2133         iucv = NULL;
2134         sk = NULL;
2135         read_lock(&iucv_sk_list.lock);
2136         sk_for_each(sk, node, &iucv_sk_list.head) {
2137                 if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2138                         if ((!memcmp(&iucv_sk(sk)->src_name,
2139                                      trans_hdr->destAppName, 8)) &&
2140                             (!memcmp(&iucv_sk(sk)->src_user_id,
2141                                      trans_hdr->destUserID, 8)) &&
2142                             (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2143                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2144                                      nullstring, 8))) {
2145                                 iucv = iucv_sk(sk);
2146                                 break;
2147                         }
2148                 } else {
2149                         if ((!memcmp(&iucv_sk(sk)->src_name,
2150                                      trans_hdr->destAppName, 8)) &&
2151                             (!memcmp(&iucv_sk(sk)->src_user_id,
2152                                      trans_hdr->destUserID, 8)) &&
2153                             (!memcmp(&iucv_sk(sk)->dst_name,
2154                                      trans_hdr->srcAppName, 8)) &&
2155                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2156                                      trans_hdr->srcUserID, 8))) {
2157                                 iucv = iucv_sk(sk);
2158                                 break;
2159                         }
2160                 }
2161         }
2162         read_unlock(&iucv_sk_list.lock);
2163         if (!iucv)
2164                 sk = NULL;
2165
2166         /* no sock
2167         how should we send with no sock
2168         1) send without sock no send rc checking?
2169         2) introduce default sock to handle this cases
2170
2171          SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2172          data -> send FIN
2173          SYN|ACK, SYN|FIN, FIN -> no action? */
2174
2175         switch (trans_hdr->flags) {
2176         case AF_IUCV_FLAG_SYN:
2177                 /* connect request */
2178                 err = afiucv_hs_callback_syn(sk, skb);
2179                 break;
2180         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2181                 /* connect request confirmed */
2182                 err = afiucv_hs_callback_synack(sk, skb);
2183                 break;
2184         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2185                 /* connect request refused */
2186                 err = afiucv_hs_callback_synfin(sk, skb);
2187                 break;
2188         case (AF_IUCV_FLAG_FIN):
2189                 /* close request */
2190                 err = afiucv_hs_callback_fin(sk, skb);
2191                 break;
2192         case (AF_IUCV_FLAG_WIN):
2193                 err = afiucv_hs_callback_win(sk, skb);
2194                 if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2195                         kfree_skb(skb);
2196                         break;
2197                 }
2198                 /* fall through and receive non-zero length data */
2199         case (AF_IUCV_FLAG_SHT):
2200                 /* shutdown request */
2201                 /* fall through and receive zero length data */
2202         case 0:
2203                 /* plain data frame */
2204                 memcpy(CB_TRGCLS(skb), &trans_hdr->iucv_hdr.class,
2205                        CB_TRGCLS_LEN);
2206                 err = afiucv_hs_callback_rx(sk, skb);
2207                 break;
2208         default:
2209                 ;
2210         }
2211
2212         return err;
2213 }
2214
2215 /**
2216  * afiucv_hs_callback_txnotify() - handle send notifcations from HiperSockets
2217  *                                 transport
2218  **/
2219 static void afiucv_hs_callback_txnotify(struct sk_buff *skb,
2220                                         enum iucv_tx_notify n)
2221 {
2222         struct sock *isk = skb->sk;
2223         struct sock *sk = NULL;
2224         struct iucv_sock *iucv = NULL;
2225         struct sk_buff_head *list;
2226         struct sk_buff *list_skb;
2227         struct sk_buff *nskb;
2228         unsigned long flags;
2229         struct hlist_node *node;
2230
2231         read_lock_irqsave(&iucv_sk_list.lock, flags);
2232         sk_for_each(sk, node, &iucv_sk_list.head)
2233                 if (sk == isk) {
2234                         iucv = iucv_sk(sk);
2235                         break;
2236                 }
2237         read_unlock_irqrestore(&iucv_sk_list.lock, flags);
2238
2239         if (!iucv || sock_flag(sk, SOCK_ZAPPED))
2240                 return;
2241
2242         list = &iucv->send_skb_q;
2243         spin_lock_irqsave(&list->lock, flags);
2244         if (skb_queue_empty(list))
2245                 goto out_unlock;
2246         list_skb = list->next;
2247         nskb = list_skb->next;
2248         while (list_skb != (struct sk_buff *)list) {
2249                 if (skb_shinfo(list_skb) == skb_shinfo(skb)) {
2250                         switch (n) {
2251                         case TX_NOTIFY_OK:
2252                                 __skb_unlink(list_skb, list);
2253                                 kfree_skb(list_skb);
2254                                 iucv_sock_wake_msglim(sk);
2255                                 break;
2256                         case TX_NOTIFY_PENDING:
2257                                 atomic_inc(&iucv->pendings);
2258                                 break;
2259                         case TX_NOTIFY_DELAYED_OK:
2260                                 __skb_unlink(list_skb, list);
2261                                 atomic_dec(&iucv->pendings);
2262                                 if (atomic_read(&iucv->pendings) <= 0)
2263                                         iucv_sock_wake_msglim(sk);
2264                                 kfree_skb(list_skb);
2265                                 break;
2266                         case TX_NOTIFY_UNREACHABLE:
2267                         case TX_NOTIFY_DELAYED_UNREACHABLE:
2268                         case TX_NOTIFY_TPQFULL: /* not yet used */
2269                         case TX_NOTIFY_GENERALERROR:
2270                         case TX_NOTIFY_DELAYED_GENERALERROR:
2271                                 __skb_unlink(list_skb, list);
2272                                 kfree_skb(list_skb);
2273                                 if (sk->sk_state == IUCV_CONNECTED) {
2274                                         sk->sk_state = IUCV_DISCONN;
2275                                         sk->sk_state_change(sk);
2276                                 }
2277                                 break;
2278                         }
2279                         break;
2280                 }
2281                 list_skb = nskb;
2282                 nskb = nskb->next;
2283         }
2284 out_unlock:
2285         spin_unlock_irqrestore(&list->lock, flags);
2286
2287         if (sk->sk_state == IUCV_CLOSING) {
2288                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
2289                         sk->sk_state = IUCV_CLOSED;
2290                         sk->sk_state_change(sk);
2291                 }
2292         }
2293
2294 }
2295
2296 /*
2297  * afiucv_netdev_event: handle netdev notifier chain events
2298  */
2299 static int afiucv_netdev_event(struct notifier_block *this,
2300                                unsigned long event, void *ptr)
2301 {
2302         struct net_device *event_dev = (struct net_device *)ptr;
2303         struct hlist_node *node;
2304         struct sock *sk;
2305         struct iucv_sock *iucv;
2306
2307         switch (event) {
2308         case NETDEV_REBOOT:
2309         case NETDEV_GOING_DOWN:
2310                 sk_for_each(sk, node, &iucv_sk_list.head) {
2311                         iucv = iucv_sk(sk);
2312                         if ((iucv->hs_dev == event_dev) &&
2313                             (sk->sk_state == IUCV_CONNECTED)) {
2314                                 if (event == NETDEV_GOING_DOWN)
2315                                         iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2316                                 sk->sk_state = IUCV_DISCONN;
2317                                 sk->sk_state_change(sk);
2318                         }
2319                 }
2320                 break;
2321         case NETDEV_DOWN:
2322         case NETDEV_UNREGISTER:
2323         default:
2324                 break;
2325         }
2326         return NOTIFY_DONE;
2327 }
2328
2329 static struct notifier_block afiucv_netdev_notifier = {
2330         .notifier_call = afiucv_netdev_event,
2331 };
2332
2333 static const struct proto_ops iucv_sock_ops = {
2334         .family         = PF_IUCV,
2335         .owner          = THIS_MODULE,
2336         .release        = iucv_sock_release,
2337         .bind           = iucv_sock_bind,
2338         .connect        = iucv_sock_connect,
2339         .listen         = iucv_sock_listen,
2340         .accept         = iucv_sock_accept,
2341         .getname        = iucv_sock_getname,
2342         .sendmsg        = iucv_sock_sendmsg,
2343         .recvmsg        = iucv_sock_recvmsg,
2344         .poll           = iucv_sock_poll,
2345         .ioctl          = sock_no_ioctl,
2346         .mmap           = sock_no_mmap,
2347         .socketpair     = sock_no_socketpair,
2348         .shutdown       = iucv_sock_shutdown,
2349         .setsockopt     = iucv_sock_setsockopt,
2350         .getsockopt     = iucv_sock_getsockopt,
2351 };
2352
2353 static const struct net_proto_family iucv_sock_family_ops = {
2354         .family = AF_IUCV,
2355         .owner  = THIS_MODULE,
2356         .create = iucv_sock_create,
2357 };
2358
2359 static struct packet_type iucv_packet_type = {
2360         .type = cpu_to_be16(ETH_P_AF_IUCV),
2361         .func = afiucv_hs_rcv,
2362 };
2363
2364 static int afiucv_iucv_init(void)
2365 {
2366         int err;
2367
2368         err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2369         if (err)
2370                 goto out;
2371         /* establish dummy device */
2372         af_iucv_driver.bus = pr_iucv->bus;
2373         err = driver_register(&af_iucv_driver);
2374         if (err)
2375                 goto out_iucv;
2376         af_iucv_dev = kzalloc(sizeof(struct device), GFP_KERNEL);
2377         if (!af_iucv_dev) {
2378                 err = -ENOMEM;
2379                 goto out_driver;
2380         }
2381         dev_set_name(af_iucv_dev, "af_iucv");
2382         af_iucv_dev->bus = pr_iucv->bus;
2383         af_iucv_dev->parent = pr_iucv->root;
2384         af_iucv_dev->release = (void (*)(struct device *))kfree;
2385         af_iucv_dev->driver = &af_iucv_driver;
2386         err = device_register(af_iucv_dev);
2387         if (err)
2388                 goto out_driver;
2389         return 0;
2390
2391 out_driver:
2392         driver_unregister(&af_iucv_driver);
2393 out_iucv:
2394         pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2395 out:
2396         return err;
2397 }
2398
2399 static int __init afiucv_init(void)
2400 {
2401         int err;
2402
2403         if (MACHINE_IS_VM) {
2404                 cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2405                 if (unlikely(err)) {
2406                         WARN_ON(err);
2407                         err = -EPROTONOSUPPORT;
2408                         goto out;
2409                 }
2410
2411                 pr_iucv = try_then_request_module(symbol_get(iucv_if), "iucv");
2412                 if (!pr_iucv) {
2413                         printk(KERN_WARNING "iucv_if lookup failed\n");
2414                         memset(&iucv_userid, 0, sizeof(iucv_userid));
2415                 }
2416         } else {
2417                 memset(&iucv_userid, 0, sizeof(iucv_userid));
2418                 pr_iucv = NULL;
2419         }
2420
2421         err = proto_register(&iucv_proto, 0);
2422         if (err)
2423                 goto out;
2424         err = sock_register(&iucv_sock_family_ops);
2425         if (err)
2426                 goto out_proto;
2427
2428         if (pr_iucv) {
2429                 err = afiucv_iucv_init();
2430                 if (err)
2431                         goto out_sock;
2432         } else
2433                 register_netdevice_notifier(&afiucv_netdev_notifier);
2434         dev_add_pack(&iucv_packet_type);
2435         return 0;
2436
2437 out_sock:
2438         sock_unregister(PF_IUCV);
2439 out_proto:
2440         proto_unregister(&iucv_proto);
2441 out:
2442         if (pr_iucv)
2443                 symbol_put(iucv_if);
2444         return err;
2445 }
2446
2447 static void __exit afiucv_exit(void)
2448 {
2449         if (pr_iucv) {
2450                 device_unregister(af_iucv_dev);
2451                 driver_unregister(&af_iucv_driver);
2452                 pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2453                 symbol_put(iucv_if);
2454         } else
2455                 unregister_netdevice_notifier(&afiucv_netdev_notifier);
2456         dev_remove_pack(&iucv_packet_type);
2457         sock_unregister(PF_IUCV);
2458         proto_unregister(&iucv_proto);
2459 }
2460
2461 module_init(afiucv_init);
2462 module_exit(afiucv_exit);
2463
2464 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2465 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2466 MODULE_VERSION(VERSION);
2467 MODULE_LICENSE("GPL");
2468 MODULE_ALIAS_NETPROTO(PF_IUCV);
2469