net/iucv/af_iucv.c

   1 /*
   2  *  IUCV protocol stack for Linux on zSeries
   3  *
   4  *  Copyright IBM Corp. 2006, 2009
   5  *
   6  *  Author(s):  Jennifer Hunt <jenhunt@us.ibm.com>
   7  *              Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
   8  *  PM functions:
   9  *              Ursula Braun <ursula.braun@de.ibm.com>
  10  */
  11
  12 #define KMSG_COMPONENT "af_iucv"
  13 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  14
  15 #include <linux/module.h>
  16 #include <linux/types.h>
  17 #include <linux/list.h>
  18 #include <linux/errno.h>
  19 #include <linux/kernel.h>
  20 #include <linux/sched.h>
  21 #include <linux/slab.h>
  22 #include <linux/skbuff.h>
  23 #include <linux/init.h>
  24 #include <linux/poll.h>
  25 #include <net/sock.h>
  26 #include <asm/ebcdic.h>
  27 #include <asm/cpcmd.h>
  28 #include <linux/kmod.h>
  29
  30 #include <net/iucv/af_iucv.h>
  31
  32 #define VERSION "1.2"
  33
  34 static char iucv_userid[80];
  35
  36 static const struct proto_ops iucv_sock_ops;
  37
  38 static struct proto iucv_proto = {
  39         .name           = "AF_IUCV",
  40         .owner          = THIS_MODULE,
  41         .obj_size       = sizeof(struct iucv_sock),
  42 };
  43
  44 static struct iucv_interface *pr_iucv;
  45
  46 /* special AF_IUCV IPRM messages */
  47 static const u8 iprm_shutdown[8] =
  48         {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
  49
  50 #define TRGCLS_SIZE     (sizeof(((struct iucv_message *)0)->class))
  51
  52 #define __iucv_sock_wait(sk, condition, timeo, ret)                     \
  53 do {                                                                    \
  54         DEFINE_WAIT(__wait);                                            \
  55         long __timeo = timeo;                                           \
  56         ret = 0;                                                        \
  57         prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);     \
  58         while (!(condition)) {                                          \
  59                 if (!__timeo) {                                         \
  60                         ret = -EAGAIN;                                  \
  61                         break;                                          \
  62                 }                                                       \
  63                 if (signal_pending(current)) {                          \
  64                         ret = sock_intr_errno(__timeo);                 \
  65                         break;                                          \
  66                 }                                                       \
  67                 release_sock(sk);                                       \
  68                 __timeo = schedule_timeout(__timeo);                    \
  69                 lock_sock(sk);                                          \
  70                 ret = sock_error(sk);                                   \
  71                 if (ret)                                                \
  72                         break;                                          \
  73         }                                                               \
  74         finish_wait(sk_sleep(sk), &__wait);                             \
  75 } while (0)
  76
  77 #define iucv_sock_wait(sk, condition, timeo)                            \
  78 ({                                                                      \
  79         int __ret = 0;                                                  \
  80         if (!(condition))                                               \
  81                 __iucv_sock_wait(sk, condition, timeo, __ret);          \
  82         __ret;                                                          \
  83 })
  84
  85 static void iucv_sock_kill(struct sock *sk);
  86 static void iucv_sock_close(struct sock *sk);
  87 static void iucv_sever_path(struct sock *, int);
  88
  89 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
  90         struct packet_type *pt, struct net_device *orig_dev);
  91 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
  92                    struct sk_buff *skb, u8 flags);
  93 static void afiucv_hs_callback_txnotify(struct sk_buff *, enum iucv_tx_notify);
  94
  95 /* Call Back functions */
  96 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
  97 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
  98 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
  99 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
 100                                  u8 ipuser[16]);
 101 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
 102 static void iucv_callback_shutdown(struct iucv_path *, u8 ipuser[16]);
 103
 104 static struct iucv_sock_list iucv_sk_list = {
 105         .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
 106         .autobind_name = ATOMIC_INIT(0)
 107 };
 108
 109 static struct iucv_handler af_iucv_handler = {
 110         .path_pending     = iucv_callback_connreq,
 111         .path_complete    = iucv_callback_connack,
 112         .path_severed     = iucv_callback_connrej,
 113         .message_pending  = iucv_callback_rx,
 114         .message_complete = iucv_callback_txdone,
 115         .path_quiesced    = iucv_callback_shutdown,
 116 };
 117
 118 static inline void high_nmcpy(unsigned char *dst, char *src)
 119 {
 120        memcpy(dst, src, 8);
 121 }
 122
 123 static inline void low_nmcpy(unsigned char *dst, char *src)
 124 {
 125        memcpy(&dst[8], src, 8);
 126 }
 127
 128 static int afiucv_pm_prepare(struct device *dev)
 129 {
 130 #ifdef CONFIG_PM_DEBUG
 131         printk(KERN_WARNING "afiucv_pm_prepare\n");
 132 #endif
 133         return 0;
 134 }
 135
 136 static void afiucv_pm_complete(struct device *dev)
 137 {
 138 #ifdef CONFIG_PM_DEBUG
 139         printk(KERN_WARNING "afiucv_pm_complete\n");
 140 #endif
 141 }
 142
 143 /**
 144  * afiucv_pm_freeze() - Freeze PM callback
 145  * @dev:        AFIUCV dummy device
 146  *
 147  * Sever all established IUCV communication pathes
 148  */
 149 static int afiucv_pm_freeze(struct device *dev)
 150 {
 151         struct iucv_sock *iucv;
 152         struct sock *sk;
 153         int err = 0;
 154
 155 #ifdef CONFIG_PM_DEBUG
 156         printk(KERN_WARNING "afiucv_pm_freeze\n");
 157 #endif
 158         read_lock(&iucv_sk_list.lock);
 159         sk_for_each(sk, &iucv_sk_list.head) {
 160                 iucv = iucv_sk(sk);
 161                 switch (sk->sk_state) {
 162                 case IUCV_DISCONN:
 163                 case IUCV_CLOSING:
 164                 case IUCV_CONNECTED:
 165                         iucv_sever_path(sk, 0);
 166                         break;
 167                 case IUCV_OPEN:
 168                 case IUCV_BOUND:
 169                 case IUCV_LISTEN:
 170                 case IUCV_CLOSED:
 171                 default:
 172                         break;
 173                 }
 174                 skb_queue_purge(&iucv->send_skb_q);
 175                 skb_queue_purge(&iucv->backlog_skb_q);
 176         }
 177         read_unlock(&iucv_sk_list.lock);
 178         return err;
 179 }
 180
 181 /**
 182  * afiucv_pm_restore_thaw() - Thaw and restore PM callback
 183  * @dev:        AFIUCV dummy device
 184  *
 185  * socket clean up after freeze
 186  */
 187 static int afiucv_pm_restore_thaw(struct device *dev)
 188 {
 189         struct sock *sk;
 190
 191 #ifdef CONFIG_PM_DEBUG
 192         printk(KERN_WARNING "afiucv_pm_restore_thaw\n");
 193 #endif
 194         read_lock(&iucv_sk_list.lock);
 195         sk_for_each(sk, &iucv_sk_list.head) {
 196                 switch (sk->sk_state) {
 197                 case IUCV_CONNECTED:
 198                         sk->sk_err = EPIPE;
 199                         sk->sk_state = IUCV_DISCONN;
 200                         sk->sk_state_change(sk);
 201                         break;
 202                 case IUCV_DISCONN:
 203                 case IUCV_CLOSING:
 204                 case IUCV_LISTEN:
 205                 case IUCV_BOUND:
 206                 case IUCV_OPEN:
 207                 default:
 208                         break;
 209                 }
 210         }
 211         read_unlock(&iucv_sk_list.lock);
 212         return 0;
 213 }
 214
 215 static const struct dev_pm_ops afiucv_pm_ops = {
 216         .prepare = afiucv_pm_prepare,
 217         .complete = afiucv_pm_complete,
 218         .freeze = afiucv_pm_freeze,
 219         .thaw = afiucv_pm_restore_thaw,
 220         .restore = afiucv_pm_restore_thaw,
 221 };
 222
 223 static struct device_driver af_iucv_driver = {
 224         .owner = THIS_MODULE,
 225         .name = "afiucv",
 226         .bus  = NULL,
 227         .pm   = &afiucv_pm_ops,
 228 };
 229
 230 /* dummy device used as trigger for PM functions */
 231 static struct device *af_iucv_dev;
 232
 233 /**
 234  * iucv_msg_length() - Returns the length of an iucv message.
 235  * @msg:        Pointer to struct iucv_message, MUST NOT be NULL
 236  *
 237  * The function returns the length of the specified iucv message @msg of data
 238  * stored in a buffer and of data stored in the parameter list (PRMDATA).
 239  *
 240  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
 241  * data:
 242  *      PRMDATA[0..6]   socket data (max 7 bytes);
 243  *      PRMDATA[7]      socket data length value (len is 0xff - PRMDATA[7])
 244  *
 245  * The socket data length is computed by subtracting the socket data length
 246  * value from 0xFF.
 247  * If the socket data len is greater 7, then PRMDATA can be used for special
 248  * notifications (see iucv_sock_shutdown); and further,
 249  * if the socket data len is > 7, the function returns 8.
 250  *
 251  * Use this function to allocate socket buffers to store iucv message data.
 252  */
 253 static inline size_t iucv_msg_length(struct iucv_message *msg)
 254 {
 255         size_t datalen;
 256
 257         if (msg->flags & IUCV_IPRMDATA) {
 258                 datalen = 0xff - msg->rmmsg[7];
 259                 return (datalen < 8) ? datalen : 8;
 260         }
 261         return msg->length;
 262 }
 263
 264 /**
 265  * iucv_sock_in_state() - check for specific states
 266  * @sk:         sock structure
 267  * @state:      first iucv sk state
 268  * @state:      second iucv sk state
 269  *
 270  * Returns true if the socket in either in the first or second state.
 271  */
 272 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
 273 {
 274         return (sk->sk_state == state || sk->sk_state == state2);
 275 }
 276
 277 /**
 278  * iucv_below_msglim() - function to check if messages can be sent
 279  * @sk:         sock structure
 280  *
 281  * Returns true if the send queue length is lower than the message limit.
 282  * Always returns true if the socket is not connected (no iucv path for
 283  * checking the message limit).
 284  */
 285 static inline int iucv_below_msglim(struct sock *sk)
 286 {
 287         struct iucv_sock *iucv = iucv_sk(sk);
 288
 289         if (sk->sk_state != IUCV_CONNECTED)
 290                 return 1;
 291         if (iucv->transport == AF_IUCV_TRANS_IUCV)
 292                 return (skb_queue_len(&iucv->send_skb_q) < iucv->path->msglim);
 293         else
 294                 return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
 295                         (atomic_read(&iucv->pendings) <= 0));
 296 }
 297
 298 /**
 299  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
 300  */
 301 static void iucv_sock_wake_msglim(struct sock *sk)
 302 {
 303         struct socket_wq *wq;
 304
 305         rcu_read_lock();
 306         wq = rcu_dereference(sk->sk_wq);
 307         if (wq_has_sleeper(wq))
 308                 wake_up_interruptible_all(&wq->wait);
 309         sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
 310         rcu_read_unlock();
 311 }
 312
 313 /**
 314  * afiucv_hs_send() - send a message through HiperSockets transport
 315  */
 316 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
 317                    struct sk_buff *skb, u8 flags)
 318 {
 319         struct iucv_sock *iucv = iucv_sk(sock);
 320         struct af_iucv_trans_hdr *phs_hdr;
 321         struct sk_buff *nskb;
 322         int err, confirm_recv = 0;
 323
 324         memset(skb->head, 0, ETH_HLEN);
 325         phs_hdr = (struct af_iucv_trans_hdr *)skb_push(skb,
 326                                         sizeof(struct af_iucv_trans_hdr));
 327         skb_reset_mac_header(skb);
 328         skb_reset_network_header(skb);
 329         skb_push(skb, ETH_HLEN);
 330         skb_reset_mac_header(skb);
 331         memset(phs_hdr, 0, sizeof(struct af_iucv_trans_hdr));
 332
 333         phs_hdr->magic = ETH_P_AF_IUCV;
 334         phs_hdr->version = 1;
 335         phs_hdr->flags = flags;
 336         if (flags == AF_IUCV_FLAG_SYN)
 337                 phs_hdr->window = iucv->msglimit;
 338         else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
 339                 confirm_recv = atomic_read(&iucv->msg_recv);
 340                 phs_hdr->window = confirm_recv;
 341                 if (confirm_recv)
 342                         phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
 343         }
 344         memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
 345         memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
 346         memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
 347         memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
 348         ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
 349         ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
 350         ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
 351         ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
 352         if (imsg)
 353                 memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
 354
 355         skb->dev = iucv->hs_dev;
 356         if (!skb->dev)
 357                 return -ENODEV;
 358         if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev))
 359                 return -ENETDOWN;
 360         if (skb->len > skb->dev->mtu) {
 361                 if (sock->sk_type == SOCK_SEQPACKET)
 362                         return -EMSGSIZE;
 363                 else
 364                         skb_trim(skb, skb->dev->mtu);
 365         }
 366         skb->protocol = ETH_P_AF_IUCV;
 367         nskb = skb_clone(skb, GFP_ATOMIC);
 368         if (!nskb)
 369                 return -ENOMEM;
 370         skb_queue_tail(&iucv->send_skb_q, nskb);
 371         err = dev_queue_xmit(skb);
 372         if (net_xmit_eval(err)) {
 373                 skb_unlink(nskb, &iucv->send_skb_q);
 374                 kfree_skb(nskb);
 375         } else {
 376                 atomic_sub(confirm_recv, &iucv->msg_recv);
 377                 WARN_ON(atomic_read(&iucv->msg_recv) < 0);
 378         }
 379         return net_xmit_eval(err);
 380 }
 381
 382 static struct sock *__iucv_get_sock_by_name(char *nm)
 383 {
 384         struct sock *sk;
 385
 386         sk_for_each(sk, &iucv_sk_list.head)
 387                 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
 388                         return sk;
 389
 390         return NULL;
 391 }
 392
 393 static void iucv_sock_destruct(struct sock *sk)
 394 {
 395         skb_queue_purge(&sk->sk_receive_queue);
 396         skb_queue_purge(&sk->sk_error_queue);
 397
 398         sk_mem_reclaim(sk);
 399
 400         if (!sock_flag(sk, SOCK_DEAD)) {
 401                 pr_err("Attempt to release alive iucv socket %p\n", sk);
 402                 return;
 403         }
 404
 405         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
 406         WARN_ON(atomic_read(&sk->sk_wmem_alloc));
 407         WARN_ON(sk->sk_wmem_queued);
 408         WARN_ON(sk->sk_forward_alloc);
 409 }
 410
 411 /* Cleanup Listen */
 412 static void iucv_sock_cleanup_listen(struct sock *parent)
 413 {
 414         struct sock *sk;
 415
 416         /* Close non-accepted connections */
 417         while ((sk = iucv_accept_dequeue(parent, NULL))) {
 418                 iucv_sock_close(sk);
 419                 iucv_sock_kill(sk);
 420         }
 421
 422         parent->sk_state = IUCV_CLOSED;
 423 }
 424
 425 /* Kill socket (only if zapped and orphaned) */
 426 static void iucv_sock_kill(struct sock *sk)
 427 {
 428         if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
 429                 return;
 430
 431         iucv_sock_unlink(&iucv_sk_list, sk);
 432         sock_set_flag(sk, SOCK_DEAD);
 433         sock_put(sk);
 434 }
 435
 436 /* Terminate an IUCV path */
 437 static void iucv_sever_path(struct sock *sk, int with_user_data)
 438 {
 439         unsigned char user_data[16];
 440         struct iucv_sock *iucv = iucv_sk(sk);
 441         struct iucv_path *path = iucv->path;
 442
 443         if (iucv->path) {
 444                 iucv->path = NULL;
 445                 if (with_user_data) {
 446                         low_nmcpy(user_data, iucv->src_name);
 447                         high_nmcpy(user_data, iucv->dst_name);
 448                         ASCEBC(user_data, sizeof(user_data));
 449                         pr_iucv->path_sever(path, user_data);
 450                 } else
 451                         pr_iucv->path_sever(path, NULL);
 452                 iucv_path_free(path);
 453         }
 454 }
 455
 456 /* Send FIN through an IUCV socket for HIPER transport */
 457 static int iucv_send_ctrl(struct sock *sk, u8 flags)
 458 {
 459         int err = 0;
 460         int blen;
 461         struct sk_buff *skb;
 462
 463         blen = sizeof(struct af_iucv_trans_hdr) + ETH_HLEN;
 464         skb = sock_alloc_send_skb(sk, blen, 1, &err);
 465         if (skb) {
 466                 skb_reserve(skb, blen);
 467                 err = afiucv_hs_send(NULL, sk, skb, flags);
 468         }
 469         return err;
 470 }
 471
 472 /* Close an IUCV socket */
 473 static void iucv_sock_close(struct sock *sk)
 474 {
 475         struct iucv_sock *iucv = iucv_sk(sk);
 476         unsigned long timeo;
 477         int err = 0;
 478
 479         lock_sock(sk);
 480
 481         switch (sk->sk_state) {
 482         case IUCV_LISTEN:
 483                 iucv_sock_cleanup_listen(sk);
 484                 break;
 485
 486         case IUCV_CONNECTED:
 487                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
 488                         err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
 489                         sk->sk_state = IUCV_DISCONN;
 490                         sk->sk_state_change(sk);
 491                 }
 492         case IUCV_DISCONN:   /* fall through */
 493                 sk->sk_state = IUCV_CLOSING;
 494                 sk->sk_state_change(sk);
 495
 496                 if (!err && !skb_queue_empty(&iucv->send_skb_q)) {
 497                         if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
 498                                 timeo = sk->sk_lingertime;
 499                         else
 500                                 timeo = IUCV_DISCONN_TIMEOUT;
 501                         iucv_sock_wait(sk,
 502                                         iucv_sock_in_state(sk, IUCV_CLOSED, 0),
 503                                         timeo);
 504                 }
 505
 506         case IUCV_CLOSING:   /* fall through */
 507                 sk->sk_state = IUCV_CLOSED;
 508                 sk->sk_state_change(sk);
 509
 510                 sk->sk_err = ECONNRESET;
 511                 sk->sk_state_change(sk);
 512
 513                 skb_queue_purge(&iucv->send_skb_q);
 514                 skb_queue_purge(&iucv->backlog_skb_q);
 515
 516         default:   /* fall through */
 517                 iucv_sever_path(sk, 1);
 518         }
 519
 520         if (iucv->hs_dev) {
 521                 dev_put(iucv->hs_dev);
 522                 iucv->hs_dev = NULL;
 523                 sk->sk_bound_dev_if = 0;
 524         }
 525
 526         /* mark socket for deletion by iucv_sock_kill() */
 527         sock_set_flag(sk, SOCK_ZAPPED);
 528
 529         release_sock(sk);
 530 }
 531
 532 static void iucv_sock_init(struct sock *sk, struct sock *parent)
 533 {
 534         if (parent)
 535                 sk->sk_type = parent->sk_type;
 536 }
 537
 538 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
 539 {
 540         struct sock *sk;
 541         struct iucv_sock *iucv;
 542
 543         sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
 544         if (!sk)
 545                 return NULL;
 546         iucv = iucv_sk(sk);
 547
 548         sock_init_data(sock, sk);
 549         INIT_LIST_HEAD(&iucv->accept_q);
 550         spin_lock_init(&iucv->accept_q_lock);
 551         skb_queue_head_init(&iucv->send_skb_q);
 552         INIT_LIST_HEAD(&iucv->message_q.list);
 553         spin_lock_init(&iucv->message_q.lock);
 554         skb_queue_head_init(&iucv->backlog_skb_q);
 555         iucv->send_tag = 0;
 556         atomic_set(&iucv->pendings, 0);
 557         iucv->flags = 0;
 558         iucv->msglimit = 0;
 559         atomic_set(&iucv->msg_sent, 0);
 560         atomic_set(&iucv->msg_recv, 0);
 561         iucv->path = NULL;
 562         iucv->sk_txnotify = afiucv_hs_callback_txnotify;
 563         memset(&iucv->src_user_id , 0, 32);
 564         if (pr_iucv)
 565                 iucv->transport = AF_IUCV_TRANS_IUCV;
 566         else
 567                 iucv->transport = AF_IUCV_TRANS_HIPER;
 568
 569         sk->sk_destruct = iucv_sock_destruct;
 570         sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
 571         sk->sk_allocation = GFP_DMA;
 572
 573         sock_reset_flag(sk, SOCK_ZAPPED);
 574
 575         sk->sk_protocol = proto;
 576         sk->sk_state    = IUCV_OPEN;
 577
 578         iucv_sock_link(&iucv_sk_list, sk);
 579         return sk;
 580 }
 581
 582 /* Create an IUCV socket */
 583 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
 584                             int kern)
 585 {
 586         struct sock *sk;
 587
 588         if (protocol && protocol != PF_IUCV)
 589                 return -EPROTONOSUPPORT;
 590
 591         sock->state = SS_UNCONNECTED;
 592
 593         switch (sock->type) {
 594         case SOCK_STREAM:
 595                 sock->ops = &iucv_sock_ops;
 596                 break;
 597         case SOCK_SEQPACKET:
 598                 /* currently, proto ops can handle both sk types */
 599                 sock->ops = &iucv_sock_ops;
 600                 break;
 601         default:
 602                 return -ESOCKTNOSUPPORT;
 603         }
 604
 605         sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
 606         if (!sk)
 607                 return -ENOMEM;
 608
 609         iucv_sock_init(sk, NULL);
 610
 611         return 0;
 612 }
 613
 614 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
 615 {
 616         write_lock_bh(&l->lock);
 617         sk_add_node(sk, &l->head);
 618         write_unlock_bh(&l->lock);
 619 }
 620
 621 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
 622 {
 623         write_lock_bh(&l->lock);
 624         sk_del_node_init(sk);
 625         write_unlock_bh(&l->lock);
 626 }
 627
 628 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
 629 {
 630         unsigned long flags;
 631         struct iucv_sock *par = iucv_sk(parent);
 632
 633         sock_hold(sk);
 634         spin_lock_irqsave(&par->accept_q_lock, flags);
 635         list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
 636         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 637         iucv_sk(sk)->parent = parent;
 638         sk_acceptq_added(parent);
 639 }
 640
 641 void iucv_accept_unlink(struct sock *sk)
 642 {
 643         unsigned long flags;
 644         struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
 645
 646         spin_lock_irqsave(&par->accept_q_lock, flags);
 647         list_del_init(&iucv_sk(sk)->accept_q);
 648         spin_unlock_irqrestore(&par->accept_q_lock, flags);
 649         sk_acceptq_removed(iucv_sk(sk)->parent);
 650         iucv_sk(sk)->parent = NULL;
 651         sock_put(sk);
 652 }
 653
 654 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
 655 {
 656         struct iucv_sock *isk, *n;
 657         struct sock *sk;
 658
 659         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
 660                 sk = (struct sock *) isk;
 661                 lock_sock(sk);
 662
 663                 if (sk->sk_state == IUCV_CLOSED) {
 664                         iucv_accept_unlink(sk);
 665                         release_sock(sk);
 666                         continue;
 667                 }
 668
 669                 if (sk->sk_state == IUCV_CONNECTED ||
 670                     sk->sk_state == IUCV_DISCONN ||
 671                     !newsock) {
 672                         iucv_accept_unlink(sk);
 673                         if (newsock)
 674                                 sock_graft(sk, newsock);
 675
 676                         release_sock(sk);
 677                         return sk;
 678                 }
 679
 680                 release_sock(sk);
 681         }
 682         return NULL;
 683 }
 684
 685 static void __iucv_auto_name(struct iucv_sock *iucv)
 686 {
 687         char name[12];
 688
 689         sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
 690         while (__iucv_get_sock_by_name(name)) {
 691                 sprintf(name, "%08x",
 692                         atomic_inc_return(&iucv_sk_list.autobind_name));
 693         }
 694         memcpy(iucv->src_name, name, 8);
 695 }
 696
 697 /* Bind an unbound socket */
 698 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
 699                           int addr_len)
 700 {
 701         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 702         struct sock *sk = sock->sk;
 703         struct iucv_sock *iucv;
 704         int err = 0;
 705         struct net_device *dev;
 706         char uid[9];
 707
 708         /* Verify the input sockaddr */
 709         if (!addr || addr->sa_family != AF_IUCV)
 710                 return -EINVAL;
 711
 712         lock_sock(sk);
 713         if (sk->sk_state != IUCV_OPEN) {
 714                 err = -EBADFD;
 715                 goto done;
 716         }
 717
 718         write_lock_bh(&iucv_sk_list.lock);
 719
 720         iucv = iucv_sk(sk);
 721         if (__iucv_get_sock_by_name(sa->siucv_name)) {
 722                 err = -EADDRINUSE;
 723                 goto done_unlock;
 724         }
 725         if (iucv->path)
 726                 goto done_unlock;
 727
 728         /* Bind the socket */
 729         if (pr_iucv)
 730                 if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
 731                         goto vm_bind; /* VM IUCV transport */
 732
 733         /* try hiper transport */
 734         memcpy(uid, sa->siucv_user_id, sizeof(uid));
 735         ASCEBC(uid, 8);
 736         rcu_read_lock();
 737         for_each_netdev_rcu(&init_net, dev) {
 738                 if (!memcmp(dev->perm_addr, uid, 8)) {
 739                         memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
 740                         /* Check for unitialized siucv_name */
 741                         if (strncmp(sa->siucv_name, "        ", 8) == 0)
 742                                 __iucv_auto_name(iucv);
 743                         else
 744                                 memcpy(iucv->src_name, sa->siucv_name, 8);
 745                         sk->sk_bound_dev_if = dev->ifindex;
 746                         iucv->hs_dev = dev;
 747                         dev_hold(dev);
 748                         sk->sk_state = IUCV_BOUND;
 749                         iucv->transport = AF_IUCV_TRANS_HIPER;
 750                         if (!iucv->msglimit)
 751                                 iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
 752                         rcu_read_unlock();
 753                         goto done_unlock;
 754                 }
 755         }
 756         rcu_read_unlock();
 757 vm_bind:
 758         if (pr_iucv) {
 759                 /* use local userid for backward compat */
 760                 memcpy(iucv->src_name, sa->siucv_name, 8);
 761                 memcpy(iucv->src_user_id, iucv_userid, 8);
 762                 sk->sk_state = IUCV_BOUND;
 763                 iucv->transport = AF_IUCV_TRANS_IUCV;
 764                 if (!iucv->msglimit)
 765                         iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 766                 goto done_unlock;
 767         }
 768         /* found no dev to bind */
 769         err = -ENODEV;
 770 done_unlock:
 771         /* Release the socket list lock */
 772         write_unlock_bh(&iucv_sk_list.lock);
 773 done:
 774         release_sock(sk);
 775         return err;
 776 }
 777
 778 /* Automatically bind an unbound socket */
 779 static int iucv_sock_autobind(struct sock *sk)
 780 {
 781         struct iucv_sock *iucv = iucv_sk(sk);
 782         int err = 0;
 783
 784         if (unlikely(!pr_iucv))
 785                 return -EPROTO;
 786
 787         memcpy(iucv->src_user_id, iucv_userid, 8);
 788
 789         write_lock_bh(&iucv_sk_list.lock);
 790         __iucv_auto_name(iucv);
 791         write_unlock_bh(&iucv_sk_list.lock);
 792
 793         if (!iucv->msglimit)
 794                 iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
 795
 796         return err;
 797 }
 798
 799 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
 800 {
 801         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 802         struct sock *sk = sock->sk;
 803         struct iucv_sock *iucv = iucv_sk(sk);
 804         unsigned char user_data[16];
 805         int err;
 806
 807         high_nmcpy(user_data, sa->siucv_name);
 808         low_nmcpy(user_data, iucv->src_name);
 809         ASCEBC(user_data, sizeof(user_data));
 810
 811         /* Create path. */
 812         iucv->path = iucv_path_alloc(iucv->msglimit,
 813                                      IUCV_IPRMDATA, GFP_KERNEL);
 814         if (!iucv->path) {
 815                 err = -ENOMEM;
 816                 goto done;
 817         }
 818         err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
 819                                     sa->siucv_user_id, NULL, user_data,
 820                                     sk);
 821         if (err) {
 822                 iucv_path_free(iucv->path);
 823                 iucv->path = NULL;
 824                 switch (err) {
 825                 case 0x0b:      /* Target communicator is not logged on */
 826                         err = -ENETUNREACH;
 827                         break;
 828                 case 0x0d:      /* Max connections for this guest exceeded */
 829                 case 0x0e:      /* Max connections for target guest exceeded */
 830                         err = -EAGAIN;
 831                         break;
 832                 case 0x0f:      /* Missing IUCV authorization */
 833                         err = -EACCES;
 834                         break;
 835                 default:
 836                         err = -ECONNREFUSED;
 837                         break;
 838                 }
 839         }
 840 done:
 841         return err;
 842 }
 843
 844 /* Connect an unconnected socket */
 845 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
 846                              int alen, int flags)
 847 {
 848         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
 849         struct sock *sk = sock->sk;
 850         struct iucv_sock *iucv = iucv_sk(sk);
 851         int err;
 852
 853         if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
 854                 return -EINVAL;
 855
 856         if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
 857                 return -EBADFD;
 858
 859         if (sk->sk_state == IUCV_OPEN &&
 860             iucv->transport == AF_IUCV_TRANS_HIPER)
 861                 return -EBADFD; /* explicit bind required */
 862
 863         if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
 864                 return -EINVAL;
 865
 866         if (sk->sk_state == IUCV_OPEN) {
 867                 err = iucv_sock_autobind(sk);
 868                 if (unlikely(err))
 869                         return err;
 870         }
 871
 872         lock_sock(sk);
 873
 874         /* Set the destination information */
 875         memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
 876         memcpy(iucv->dst_name, sa->siucv_name, 8);
 877
 878         if (iucv->transport == AF_IUCV_TRANS_HIPER)
 879                 err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
 880         else
 881                 err = afiucv_path_connect(sock, addr);
 882         if (err)
 883                 goto done;
 884
 885         if (sk->sk_state != IUCV_CONNECTED)
 886                 err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
 887                                                             IUCV_DISCONN),
 888                                      sock_sndtimeo(sk, flags & O_NONBLOCK));
 889
 890         if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
 891                 err = -ECONNREFUSED;
 892
 893         if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
 894                 iucv_sever_path(sk, 0);
 895
 896 done:
 897         release_sock(sk);
 898         return err;
 899 }
 900
 901 /* Move a socket into listening state. */
 902 static int iucv_sock_listen(struct socket *sock, int backlog)
 903 {
 904         struct sock *sk = sock->sk;
 905         int err;
 906
 907         lock_sock(sk);
 908
 909         err = -EINVAL;
 910         if (sk->sk_state != IUCV_BOUND)
 911                 goto done;
 912
 913         if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
 914                 goto done;
 915
 916         sk->sk_max_ack_backlog = backlog;
 917         sk->sk_ack_backlog = 0;
 918         sk->sk_state = IUCV_LISTEN;
 919         err = 0;
 920
 921 done:
 922         release_sock(sk);
 923         return err;
 924 }
 925
 926 /* Accept a pending connection */
 927 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
 928                             int flags)
 929 {
 930         DECLARE_WAITQUEUE(wait, current);
 931         struct sock *sk = sock->sk, *nsk;
 932         long timeo;
 933         int err = 0;
 934
 935         lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 936
 937         if (sk->sk_state != IUCV_LISTEN) {
 938                 err = -EBADFD;
 939                 goto done;
 940         }
 941
 942         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
 943
 944         /* Wait for an incoming connection */
 945         add_wait_queue_exclusive(sk_sleep(sk), &wait);
 946         while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
 947                 set_current_state(TASK_INTERRUPTIBLE);
 948                 if (!timeo) {
 949                         err = -EAGAIN;
 950                         break;
 951                 }
 952
 953                 release_sock(sk);
 954                 timeo = schedule_timeout(timeo);
 955                 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 956
 957                 if (sk->sk_state != IUCV_LISTEN) {
 958                         err = -EBADFD;
 959                         break;
 960                 }
 961
 962                 if (signal_pending(current)) {
 963                         err = sock_intr_errno(timeo);
 964                         break;
 965                 }
 966         }
 967
 968         set_current_state(TASK_RUNNING);
 969         remove_wait_queue(sk_sleep(sk), &wait);
 970
 971         if (err)
 972                 goto done;
 973
 974         newsock->state = SS_CONNECTED;
 975
 976 done:
 977         release_sock(sk);
 978         return err;
 979 }
 980
 981 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
 982                              int *len, int peer)
 983 {
 984         struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
 985         struct sock *sk = sock->sk;
 986         struct iucv_sock *iucv = iucv_sk(sk);
 987
 988         addr->sa_family = AF_IUCV;
 989         *len = sizeof(struct sockaddr_iucv);
 990
 991         if (peer) {
 992                 memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
 993                 memcpy(siucv->siucv_name, iucv->dst_name, 8);
 994         } else {
 995                 memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
 996                 memcpy(siucv->siucv_name, iucv->src_name, 8);
 997         }
 998         memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
 999         memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
1000         memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
1001
1002         return 0;
1003 }
1004
1005 /**
1006  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
1007  * @path:       IUCV path
1008  * @msg:        Pointer to a struct iucv_message
1009  * @skb:        The socket data to send, skb->len MUST BE <= 7
1010  *
1011  * Send the socket data in the parameter list in the iucv message
1012  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
1013  * list and the socket data len at index 7 (last byte).
1014  * See also iucv_msg_length().
1015  *
1016  * Returns the error code from the iucv_message_send() call.
1017  */
1018 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
1019                           struct sk_buff *skb)
1020 {
1021         u8 prmdata[8];
1022
1023         memcpy(prmdata, (void *) skb->data, skb->len);
1024         prmdata[7] = 0xff - (u8) skb->len;
1025         return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
1026                                  (void *) prmdata, 8);
1027 }
1028
1029 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1030                              struct msghdr *msg, size_t len)
1031 {
1032         struct sock *sk = sock->sk;
1033         struct iucv_sock *iucv = iucv_sk(sk);
1034         struct sk_buff *skb;
1035         struct iucv_message txmsg;
1036         struct cmsghdr *cmsg;
1037         int cmsg_done;
1038         long timeo;
1039         char user_id[9];
1040         char appl_id[9];
1041         int err;
1042         int noblock = msg->msg_flags & MSG_DONTWAIT;
1043
1044         err = sock_error(sk);
1045         if (err)
1046                 return err;
1047
1048         if (msg->msg_flags & MSG_OOB)
1049                 return -EOPNOTSUPP;
1050
1051         /* SOCK_SEQPACKET: we do not support segmented records */
1052         if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
1053                 return -EOPNOTSUPP;
1054
1055         lock_sock(sk);
1056
1057         if (sk->sk_shutdown & SEND_SHUTDOWN) {
1058                 err = -EPIPE;
1059                 goto out;
1060         }
1061
1062         /* Return if the socket is not in connected state */
1063         if (sk->sk_state != IUCV_CONNECTED) {
1064                 err = -ENOTCONN;
1065                 goto out;
1066         }
1067
1068         /* initialize defaults */
1069         cmsg_done   = 0;        /* check for duplicate headers */
1070         txmsg.class = 0;
1071
1072         /* iterate over control messages */
1073         for (cmsg = CMSG_FIRSTHDR(msg); cmsg;
1074                 cmsg = CMSG_NXTHDR(msg, cmsg)) {
1075
1076                 if (!CMSG_OK(msg, cmsg)) {
1077                         err = -EINVAL;
1078                         goto out;
1079                 }
1080
1081                 if (cmsg->cmsg_level != SOL_IUCV)
1082                         continue;
1083
1084                 if (cmsg->cmsg_type & cmsg_done) {
1085                         err = -EINVAL;
1086                         goto out;
1087                 }
1088                 cmsg_done |= cmsg->cmsg_type;
1089
1090                 switch (cmsg->cmsg_type) {
1091                 case SCM_IUCV_TRGCLS:
1092                         if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
1093                                 err = -EINVAL;
1094                                 goto out;
1095                         }
1096
1097                         /* set iucv message target class */
1098                         memcpy(&txmsg.class,
1099                                 (void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
1100
1101                         break;
1102
1103                 default:
1104                         err = -EINVAL;
1105                         goto out;
1106                         break;
1107                 }
1108         }
1109
1110         /* allocate one skb for each iucv message:
1111          * this is fine for SOCK_SEQPACKET (unless we want to support
1112          * segmented records using the MSG_EOR flag), but
1113          * for SOCK_STREAM we might want to improve it in future */
1114         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1115                 skb = sock_alloc_send_skb(sk,
1116                         len + sizeof(struct af_iucv_trans_hdr) + ETH_HLEN,
1117                         noblock, &err);
1118         else
1119                 skb = sock_alloc_send_skb(sk, len, noblock, &err);
1120         if (!skb) {
1121                 err = -ENOMEM;
1122                 goto out;
1123         }
1124         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1125                 skb_reserve(skb, sizeof(struct af_iucv_trans_hdr) + ETH_HLEN);
1126         if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1127                 err = -EFAULT;
1128                 goto fail;
1129         }
1130
1131         /* wait if outstanding messages for iucv path has reached */
1132         timeo = sock_sndtimeo(sk, noblock);
1133         err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1134         if (err)
1135                 goto fail;
1136
1137         /* return -ECONNRESET if the socket is no longer connected */
1138         if (sk->sk_state != IUCV_CONNECTED) {
1139                 err = -ECONNRESET;
1140                 goto fail;
1141         }
1142
1143         /* increment and save iucv message tag for msg_completion cbk */
1144         txmsg.tag = iucv->send_tag++;
1145         IUCV_SKB_CB(skb)->tag = txmsg.tag;
1146
1147         if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1148                 atomic_inc(&iucv->msg_sent);
1149                 err = afiucv_hs_send(&txmsg, sk, skb, 0);
1150                 if (err) {
1151                         atomic_dec(&iucv->msg_sent);
1152                         goto fail;
1153                 }
1154                 goto release;
1155         }
1156         skb_queue_tail(&iucv->send_skb_q, skb);
1157
1158         if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags)
1159               && skb->len <= 7) {
1160                 err = iucv_send_iprm(iucv->path, &txmsg, skb);
1161
1162                 /* on success: there is no message_complete callback
1163                  * for an IPRMDATA msg; remove skb from send queue */
1164                 if (err == 0) {
1165                         skb_unlink(skb, &iucv->send_skb_q);
1166                         kfree_skb(skb);
1167                 }
1168
1169                 /* this error should never happen since the
1170                  * IUCV_IPRMDATA path flag is set... sever path */
1171                 if (err == 0x15) {
1172                         pr_iucv->path_sever(iucv->path, NULL);
1173                         skb_unlink(skb, &iucv->send_skb_q);
1174                         err = -EPIPE;
1175                         goto fail;
1176                 }
1177         } else
1178                 err = pr_iucv->message_send(iucv->path, &txmsg, 0, 0,
1179                                         (void *) skb->data, skb->len);
1180         if (err) {
1181                 if (err == 3) {
1182                         user_id[8] = 0;
1183                         memcpy(user_id, iucv->dst_user_id, 8);
1184                         appl_id[8] = 0;
1185                         memcpy(appl_id, iucv->dst_name, 8);
1186                         pr_err("Application %s on z/VM guest %s"
1187                                 " exceeds message limit\n",
1188                                 appl_id, user_id);
1189                         err = -EAGAIN;
1190                 } else
1191                         err = -EPIPE;
1192                 skb_unlink(skb, &iucv->send_skb_q);
1193                 goto fail;
1194         }
1195
1196 release:
1197         release_sock(sk);
1198         return len;
1199
1200 fail:
1201         kfree_skb(skb);
1202 out:
1203         release_sock(sk);
1204         return err;
1205 }
1206
1207 /* iucv_fragment_skb() - Fragment a single IUCV message into multiple skb's
1208  *
1209  * Locking: must be called with message_q.lock held
1210  */
1211 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
1212 {
1213         int dataleft, size, copied = 0;
1214         struct sk_buff *nskb;
1215
1216         dataleft = len;
1217         while (dataleft) {
1218                 if (dataleft >= sk->sk_rcvbuf / 4)
1219                         size = sk->sk_rcvbuf / 4;
1220                 else
1221                         size = dataleft;
1222
1223                 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
1224                 if (!nskb)
1225                         return -ENOMEM;
1226
1227                 /* copy target class to control buffer of new skb */
1228                 IUCV_SKB_CB(nskb)->class = IUCV_SKB_CB(skb)->class;
1229
1230                 /* copy data fragment */
1231                 memcpy(nskb->data, skb->data + copied, size);
1232                 copied += size;
1233                 dataleft -= size;
1234
1235                 skb_reset_transport_header(nskb);
1236                 skb_reset_network_header(nskb);
1237                 nskb->len = size;
1238
1239                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
1240         }
1241
1242         return 0;
1243 }
1244
1245 /* iucv_process_message() - Receive a single outstanding IUCV message
1246  *
1247  * Locking: must be called with message_q.lock held
1248  */
1249 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1250                                  struct iucv_path *path,
1251                                  struct iucv_message *msg)
1252 {
1253         int rc;
1254         unsigned int len;
1255
1256         len = iucv_msg_length(msg);
1257
1258         /* store msg target class in the second 4 bytes of skb ctrl buffer */
1259         /* Note: the first 4 bytes are reserved for msg tag */
1260         IUCV_SKB_CB(skb)->class = msg->class;
1261
1262         /* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1263         if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1264                 if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1265                         skb->data = NULL;
1266                         skb->len = 0;
1267                 }
1268         } else {
1269                 rc = pr_iucv->message_receive(path, msg,
1270                                               msg->flags & IUCV_IPRMDATA,
1271                                               skb->data, len, NULL);
1272                 if (rc) {
1273                         kfree_skb(skb);
1274                         return;
1275                 }
1276                 /* we need to fragment iucv messages for SOCK_STREAM only;
1277                  * for SOCK_SEQPACKET, it is only relevant if we support
1278                  * record segmentation using MSG_EOR (see also recvmsg()) */
1279                 if (sk->sk_type == SOCK_STREAM &&
1280                     skb->truesize >= sk->sk_rcvbuf / 4) {
1281                         rc = iucv_fragment_skb(sk, skb, len);
1282                         kfree_skb(skb);
1283                         skb = NULL;
1284                         if (rc) {
1285                                 pr_iucv->path_sever(path, NULL);
1286                                 return;
1287                         }
1288                         skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
1289                 } else {
1290                         skb_reset_transport_header(skb);
1291                         skb_reset_network_header(skb);
1292                         skb->len = len;
1293                 }
1294         }
1295
1296         IUCV_SKB_CB(skb)->offset = 0;
1297         if (sock_queue_rcv_skb(sk, skb))
1298                 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
1299 }
1300
1301 /* iucv_process_message_q() - Process outstanding IUCV messages
1302  *
1303  * Locking: must be called with message_q.lock held
1304  */
1305 static void iucv_process_message_q(struct sock *sk)
1306 {
1307         struct iucv_sock *iucv = iucv_sk(sk);
1308         struct sk_buff *skb;
1309         struct sock_msg_q *p, *n;
1310
1311         list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1312                 skb = alloc_skb(iucv_msg_length(&p->msg), GFP_ATOMIC | GFP_DMA);
1313                 if (!skb)
1314                         break;
1315                 iucv_process_message(sk, skb, p->path, &p->msg);
1316                 list_del(&p->list);
1317                 kfree(p);
1318                 if (!skb_queue_empty(&iucv->backlog_skb_q))
1319                         break;
1320         }
1321 }
1322
1323 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
1324                              struct msghdr *msg, size_t len, int flags)
1325 {
1326         int noblock = flags & MSG_DONTWAIT;
1327         struct sock *sk = sock->sk;
1328         struct iucv_sock *iucv = iucv_sk(sk);
1329         unsigned int copied, rlen;
1330         struct sk_buff *skb, *rskb, *cskb;
1331         int err = 0;
1332         u32 offset;
1333
1334         if ((sk->sk_state == IUCV_DISCONN) &&
1335             skb_queue_empty(&iucv->backlog_skb_q) &&
1336             skb_queue_empty(&sk->sk_receive_queue) &&
1337             list_empty(&iucv->message_q.list))
1338                 return 0;
1339
1340         if (flags & (MSG_OOB))
1341                 return -EOPNOTSUPP;
1342
1343         /* receive/dequeue next skb:
1344          * the function understands MSG_PEEK and, thus, does not dequeue skb */
1345         skb = skb_recv_datagram(sk, flags, noblock, &err);
1346         if (!skb) {
1347                 if (sk->sk_shutdown & RCV_SHUTDOWN)
1348                         return 0;
1349                 return err;
1350         }
1351
1352         offset = IUCV_SKB_CB(skb)->offset;
1353         rlen   = skb->len - offset;             /* 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, offset, 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                        sizeof(IUCV_SKB_CB(skb)->class),
1378                        (void *)&IUCV_SKB_CB(skb)->class);
1379         if (err) {
1380                 if (!(flags & MSG_PEEK))
1381                         skb_queue_head(&sk->sk_receive_queue, skb);
1382                 return err;
1383         }
1384
1385         /* Mark read part of skb as used */
1386         if (!(flags & MSG_PEEK)) {
1387
1388                 /* SOCK_STREAM: re-queue skb if it contains unreceived data */
1389                 if (sk->sk_type == SOCK_STREAM) {
1390                         if (copied < rlen) {
1391                                 IUCV_SKB_CB(skb)->offset = offset + copied;
1392                                 skb_queue_head(&sk->sk_receive_queue, skb);
1393                                 goto done;
1394                         }
1395                 }
1396
1397                 kfree_skb(skb);
1398                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1399                         atomic_inc(&iucv->msg_recv);
1400                         if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1401                                 WARN_ON(1);
1402                                 iucv_sock_close(sk);
1403                                 return -EFAULT;
1404                         }
1405                 }
1406
1407                 /* Queue backlog skbs */
1408                 spin_lock_bh(&iucv->message_q.lock);
1409                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1410                 while (rskb) {
1411                         IUCV_SKB_CB(rskb)->offset = 0;
1412                         if (sock_queue_rcv_skb(sk, rskb)) {
1413                                 skb_queue_head(&iucv->backlog_skb_q,
1414                                                 rskb);
1415                                 break;
1416                         } else {
1417                                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1418                         }
1419                 }
1420                 if (skb_queue_empty(&iucv->backlog_skb_q)) {
1421                         if (!list_empty(&iucv->message_q.list))
1422                                 iucv_process_message_q(sk);
1423                         if (atomic_read(&iucv->msg_recv) >=
1424                                                         iucv->msglimit / 2) {
1425                                 err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1426                                 if (err) {
1427                                         sk->sk_state = IUCV_DISCONN;
1428                                         sk->sk_state_change(sk);
1429                                 }
1430                         }
1431                 }
1432                 spin_unlock_bh(&iucv->message_q.lock);
1433         }
1434
1435 done:
1436         /* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1437         if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1438                 copied = rlen;
1439
1440         return copied;
1441 }
1442
1443 static inline unsigned int iucv_accept_poll(struct sock *parent)
1444 {
1445         struct iucv_sock *isk, *n;
1446         struct sock *sk;
1447
1448         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1449                 sk = (struct sock *) isk;
1450
1451                 if (sk->sk_state == IUCV_CONNECTED)
1452                         return POLLIN | POLLRDNORM;
1453         }
1454
1455         return 0;
1456 }
1457
1458 unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
1459                             poll_table *wait)
1460 {
1461         struct sock *sk = sock->sk;
1462         unsigned int mask = 0;
1463
1464         sock_poll_wait(file, sk_sleep(sk), wait);
1465
1466         if (sk->sk_state == IUCV_LISTEN)
1467                 return iucv_accept_poll(sk);
1468
1469         if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1470                 mask |= POLLERR |
1471                         (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
1472
1473         if (sk->sk_shutdown & RCV_SHUTDOWN)
1474                 mask |= POLLRDHUP;
1475
1476         if (sk->sk_shutdown == SHUTDOWN_MASK)
1477                 mask |= POLLHUP;
1478
1479         if (!skb_queue_empty(&sk->sk_receive_queue) ||
1480             (sk->sk_shutdown & RCV_SHUTDOWN))
1481                 mask |= POLLIN | POLLRDNORM;
1482
1483         if (sk->sk_state == IUCV_CLOSED)
1484                 mask |= POLLHUP;
1485
1486         if (sk->sk_state == IUCV_DISCONN)
1487                 mask |= POLLIN;
1488
1489         if (sock_writeable(sk) && iucv_below_msglim(sk))
1490                 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1491         else
1492                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1493
1494         return mask;
1495 }
1496
1497 static int iucv_sock_shutdown(struct socket *sock, int how)
1498 {
1499         struct sock *sk = sock->sk;
1500         struct iucv_sock *iucv = iucv_sk(sk);
1501         struct iucv_message txmsg;
1502         int err = 0;
1503
1504         how++;
1505
1506         if ((how & ~SHUTDOWN_MASK) || !how)
1507                 return -EINVAL;
1508
1509         lock_sock(sk);
1510         switch (sk->sk_state) {
1511         case IUCV_LISTEN:
1512         case IUCV_DISCONN:
1513         case IUCV_CLOSING:
1514         case IUCV_CLOSED:
1515                 err = -ENOTCONN;
1516                 goto fail;
1517         default:
1518                 break;
1519         }
1520
1521         if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
1522                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1523                         txmsg.class = 0;
1524                         txmsg.tag = 0;
1525                         err = pr_iucv->message_send(iucv->path, &txmsg,
1526                                 IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1527                         if (err) {
1528                                 switch (err) {
1529                                 case 1:
1530                                         err = -ENOTCONN;
1531                                         break;
1532                                 case 2:
1533                                         err = -ECONNRESET;
1534                                         break;
1535                                 default:
1536                                         err = -ENOTCONN;
1537                                         break;
1538                                 }
1539                         }
1540                 } else
1541                         iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1542         }
1543
1544         sk->sk_shutdown |= how;
1545         if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1546                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1547                         err = pr_iucv->path_quiesce(iucv->path, NULL);
1548                         if (err)
1549                                 err = -ENOTCONN;
1550 /*                      skb_queue_purge(&sk->sk_receive_queue); */
1551                 }
1552                 skb_queue_purge(&sk->sk_receive_queue);
1553         }
1554
1555         /* Wake up anyone sleeping in poll */
1556         sk->sk_state_change(sk);
1557
1558 fail:
1559         release_sock(sk);
1560         return err;
1561 }
1562
1563 static int iucv_sock_release(struct socket *sock)
1564 {
1565         struct sock *sk = sock->sk;
1566         int err = 0;
1567
1568         if (!sk)
1569                 return 0;
1570
1571         iucv_sock_close(sk);
1572
1573         sock_orphan(sk);
1574         iucv_sock_kill(sk);
1575         return err;
1576 }
1577
1578 /* getsockopt and setsockopt */
1579 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1580                                 char __user *optval, unsigned int optlen)
1581 {
1582         struct sock *sk = sock->sk;
1583         struct iucv_sock *iucv = iucv_sk(sk);
1584         int val;
1585         int rc;
1586
1587         if (level != SOL_IUCV)
1588                 return -ENOPROTOOPT;
1589
1590         if (optlen < sizeof(int))
1591                 return -EINVAL;
1592
1593         if (get_user(val, (int __user *) optval))
1594                 return -EFAULT;
1595
1596         rc = 0;
1597
1598         lock_sock(sk);
1599         switch (optname) {
1600         case SO_IPRMDATA_MSG:
1601                 if (val)
1602                         iucv->flags |= IUCV_IPRMDATA;
1603                 else
1604                         iucv->flags &= ~IUCV_IPRMDATA;
1605                 break;
1606         case SO_MSGLIMIT:
1607                 switch (sk->sk_state) {
1608                 case IUCV_OPEN:
1609                 case IUCV_BOUND:
1610                         if (val < 1 || val > (u16)(~0))
1611                                 rc = -EINVAL;
1612                         else
1613                                 iucv->msglimit = val;
1614                         break;
1615                 default:
1616                         rc = -EINVAL;
1617                         break;
1618                 }
1619                 break;
1620         default:
1621                 rc = -ENOPROTOOPT;
1622                 break;
1623         }
1624         release_sock(sk);
1625
1626         return rc;
1627 }
1628
1629 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1630                                 char __user *optval, int __user *optlen)
1631 {
1632         struct sock *sk = sock->sk;
1633         struct iucv_sock *iucv = iucv_sk(sk);
1634         unsigned int val;
1635         int len;
1636
1637         if (level != SOL_IUCV)
1638                 return -ENOPROTOOPT;
1639
1640         if (get_user(len, optlen))
1641                 return -EFAULT;
1642
1643         if (len < 0)
1644                 return -EINVAL;
1645
1646         len = min_t(unsigned int, len, sizeof(int));
1647
1648         switch (optname) {
1649         case SO_IPRMDATA_MSG:
1650                 val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1651                 break;
1652         case SO_MSGLIMIT:
1653                 lock_sock(sk);
1654                 val = (iucv->path != NULL) ? iucv->path->msglim /* connected */
1655                                            : iucv->msglimit;    /* default */
1656                 release_sock(sk);
1657                 break;
1658         case SO_MSGSIZE:
1659                 if (sk->sk_state == IUCV_OPEN)
1660                         return -EBADFD;
1661                 val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1662                                 sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1663                                 0x7fffffff;
1664                 break;
1665         default:
1666                 return -ENOPROTOOPT;
1667         }
1668
1669         if (put_user(len, optlen))
1670                 return -EFAULT;
1671         if (copy_to_user(optval, &val, len))
1672                 return -EFAULT;
1673
1674         return 0;
1675 }
1676
1677
1678 /* Callback wrappers - called from iucv base support */
1679 static int iucv_callback_connreq(struct iucv_path *path,
1680                                  u8 ipvmid[8], u8 ipuser[16])
1681 {
1682         unsigned char user_data[16];
1683         unsigned char nuser_data[16];
1684         unsigned char src_name[8];
1685         struct sock *sk, *nsk;
1686         struct iucv_sock *iucv, *niucv;
1687         int err;
1688
1689         memcpy(src_name, ipuser, 8);
1690         EBCASC(src_name, 8);
1691         /* Find out if this path belongs to af_iucv. */
1692         read_lock(&iucv_sk_list.lock);
1693         iucv = NULL;
1694         sk = NULL;
1695         sk_for_each(sk, &iucv_sk_list.head)
1696                 if (sk->sk_state == IUCV_LISTEN &&
1697                     !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1698                         /*
1699                          * Found a listening socket with
1700                          * src_name == ipuser[0-7].
1701                          */
1702                         iucv = iucv_sk(sk);
1703                         break;
1704                 }
1705         read_unlock(&iucv_sk_list.lock);
1706         if (!iucv)
1707                 /* No socket found, not one of our paths. */
1708                 return -EINVAL;
1709
1710         bh_lock_sock(sk);
1711
1712         /* Check if parent socket is listening */
1713         low_nmcpy(user_data, iucv->src_name);
1714         high_nmcpy(user_data, iucv->dst_name);
1715         ASCEBC(user_data, sizeof(user_data));
1716         if (sk->sk_state != IUCV_LISTEN) {
1717                 err = pr_iucv->path_sever(path, user_data);
1718                 iucv_path_free(path);
1719                 goto fail;
1720         }
1721
1722         /* Check for backlog size */
1723         if (sk_acceptq_is_full(sk)) {
1724                 err = pr_iucv->path_sever(path, user_data);
1725                 iucv_path_free(path);
1726                 goto fail;
1727         }
1728
1729         /* Create the new socket */
1730         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1731         if (!nsk) {
1732                 err = pr_iucv->path_sever(path, user_data);
1733                 iucv_path_free(path);
1734                 goto fail;
1735         }
1736
1737         niucv = iucv_sk(nsk);
1738         iucv_sock_init(nsk, sk);
1739
1740         /* Set the new iucv_sock */
1741         memcpy(niucv->dst_name, ipuser + 8, 8);
1742         EBCASC(niucv->dst_name, 8);
1743         memcpy(niucv->dst_user_id, ipvmid, 8);
1744         memcpy(niucv->src_name, iucv->src_name, 8);
1745         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1746         niucv->path = path;
1747
1748         /* Call iucv_accept */
1749         high_nmcpy(nuser_data, ipuser + 8);
1750         memcpy(nuser_data + 8, niucv->src_name, 8);
1751         ASCEBC(nuser_data + 8, 8);
1752
1753         /* set message limit for path based on msglimit of accepting socket */
1754         niucv->msglimit = iucv->msglimit;
1755         path->msglim = iucv->msglimit;
1756         err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1757         if (err) {
1758                 iucv_sever_path(nsk, 1);
1759                 iucv_sock_kill(nsk);
1760                 goto fail;
1761         }
1762
1763         iucv_accept_enqueue(sk, nsk);
1764
1765         /* Wake up accept */
1766         nsk->sk_state = IUCV_CONNECTED;
1767         sk->sk_data_ready(sk);
1768         err = 0;
1769 fail:
1770         bh_unlock_sock(sk);
1771         return 0;
1772 }
1773
1774 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1775 {
1776         struct sock *sk = path->private;
1777
1778         sk->sk_state = IUCV_CONNECTED;
1779         sk->sk_state_change(sk);
1780 }
1781
1782 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1783 {
1784         struct sock *sk = path->private;
1785         struct iucv_sock *iucv = iucv_sk(sk);
1786         struct sk_buff *skb;
1787         struct sock_msg_q *save_msg;
1788         int len;
1789
1790         if (sk->sk_shutdown & RCV_SHUTDOWN) {
1791                 pr_iucv->message_reject(path, msg);
1792                 return;
1793         }
1794
1795         spin_lock(&iucv->message_q.lock);
1796
1797         if (!list_empty(&iucv->message_q.list) ||
1798             !skb_queue_empty(&iucv->backlog_skb_q))
1799                 goto save_message;
1800
1801         len = atomic_read(&sk->sk_rmem_alloc);
1802         len += SKB_TRUESIZE(iucv_msg_length(msg));
1803         if (len > sk->sk_rcvbuf)
1804                 goto save_message;
1805
1806         skb = alloc_skb(iucv_msg_length(msg), GFP_ATOMIC | GFP_DMA);
1807         if (!skb)
1808                 goto save_message;
1809
1810         iucv_process_message(sk, skb, path, msg);
1811         goto out_unlock;
1812
1813 save_message:
1814         save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1815         if (!save_msg)
1816                 goto out_unlock;
1817         save_msg->path = path;
1818         save_msg->msg = *msg;
1819
1820         list_add_tail(&save_msg->list, &iucv->message_q.list);
1821
1822 out_unlock:
1823         spin_unlock(&iucv->message_q.lock);
1824 }
1825
1826 static void iucv_callback_txdone(struct iucv_path *path,
1827                                  struct iucv_message *msg)
1828 {
1829         struct sock *sk = path->private;
1830         struct sk_buff *this = NULL;
1831         struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1832         struct sk_buff *list_skb = list->next;
1833         unsigned long flags;
1834
1835         bh_lock_sock(sk);
1836         if (!skb_queue_empty(list)) {
1837                 spin_lock_irqsave(&list->lock, flags);
1838
1839                 while (list_skb != (struct sk_buff *)list) {
1840                         if (msg->tag == IUCV_SKB_CB(list_skb)->tag) {
1841                                 this = list_skb;
1842                                 break;
1843                         }
1844                         list_skb = list_skb->next;
1845                 }
1846                 if (this)
1847                         __skb_unlink(this, list);
1848
1849                 spin_unlock_irqrestore(&list->lock, flags);
1850
1851                 if (this) {
1852                         kfree_skb(this);
1853                         /* wake up any process waiting for sending */
1854                         iucv_sock_wake_msglim(sk);
1855                 }
1856         }
1857
1858         if (sk->sk_state == IUCV_CLOSING) {
1859                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1860                         sk->sk_state = IUCV_CLOSED;
1861                         sk->sk_state_change(sk);
1862                 }
1863         }
1864         bh_unlock_sock(sk);
1865
1866 }
1867
1868 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1869 {
1870         struct sock *sk = path->private;
1871
1872         if (sk->sk_state == IUCV_CLOSED)
1873                 return;
1874
1875         bh_lock_sock(sk);
1876         iucv_sever_path(sk, 1);
1877         sk->sk_state = IUCV_DISCONN;
1878
1879         sk->sk_state_change(sk);
1880         bh_unlock_sock(sk);
1881 }
1882
1883 /* called if the other communication side shuts down its RECV direction;
1884  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1885  */
1886 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1887 {
1888         struct sock *sk = path->private;
1889
1890         bh_lock_sock(sk);
1891         if (sk->sk_state != IUCV_CLOSED) {
1892                 sk->sk_shutdown |= SEND_SHUTDOWN;
1893                 sk->sk_state_change(sk);
1894         }
1895         bh_unlock_sock(sk);
1896 }
1897
1898 /***************** HiperSockets transport callbacks ********************/
1899 static void afiucv_swap_src_dest(struct sk_buff *skb)
1900 {
1901         struct af_iucv_trans_hdr *trans_hdr =
1902                                 (struct af_iucv_trans_hdr *)skb->data;
1903         char tmpID[8];
1904         char tmpName[8];
1905
1906         ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1907         ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1908         ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1909         ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1910         memcpy(tmpID, trans_hdr->srcUserID, 8);
1911         memcpy(tmpName, trans_hdr->srcAppName, 8);
1912         memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1913         memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1914         memcpy(trans_hdr->destUserID, tmpID, 8);
1915         memcpy(trans_hdr->destAppName, tmpName, 8);
1916         skb_push(skb, ETH_HLEN);
1917         memset(skb->data, 0, ETH_HLEN);
1918 }
1919
1920 /**
1921  * afiucv_hs_callback_syn - react on received SYN
1922  **/
1923 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1924 {
1925         struct sock *nsk;
1926         struct iucv_sock *iucv, *niucv;
1927         struct af_iucv_trans_hdr *trans_hdr;
1928         int err;
1929
1930         iucv = iucv_sk(sk);
1931         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
1932         if (!iucv) {
1933                 /* no sock - connection refused */
1934                 afiucv_swap_src_dest(skb);
1935                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1936                 err = dev_queue_xmit(skb);
1937                 goto out;
1938         }
1939
1940         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1941         bh_lock_sock(sk);
1942         if ((sk->sk_state != IUCV_LISTEN) ||
1943             sk_acceptq_is_full(sk) ||
1944             !nsk) {
1945                 /* error on server socket - connection refused */
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                 iucv_sock_kill(nsk);
1950                 bh_unlock_sock(sk);
1951                 goto out;
1952         }
1953
1954         niucv = iucv_sk(nsk);
1955         iucv_sock_init(nsk, sk);
1956         niucv->transport = AF_IUCV_TRANS_HIPER;
1957         niucv->msglimit = iucv->msglimit;
1958         if (!trans_hdr->window)
1959                 niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1960         else
1961                 niucv->msglimit_peer = trans_hdr->window;
1962         memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1963         memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1964         memcpy(niucv->src_name, iucv->src_name, 8);
1965         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1966         nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1967         niucv->hs_dev = iucv->hs_dev;
1968         dev_hold(niucv->hs_dev);
1969         afiucv_swap_src_dest(skb);
1970         trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1971         trans_hdr->window = niucv->msglimit;
1972         /* if receiver acks the xmit connection is established */
1973         err = dev_queue_xmit(skb);
1974         if (!err) {
1975                 iucv_accept_enqueue(sk, nsk);
1976                 nsk->sk_state = IUCV_CONNECTED;
1977                 sk->sk_data_ready(sk);
1978         } else
1979                 iucv_sock_kill(nsk);
1980         bh_unlock_sock(sk);
1981
1982 out:
1983         return NET_RX_SUCCESS;
1984 }
1985
1986 /**
1987  * afiucv_hs_callback_synack() - react on received SYN-ACK
1988  **/
1989 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1990 {
1991         struct iucv_sock *iucv = iucv_sk(sk);
1992         struct af_iucv_trans_hdr *trans_hdr =
1993                                         (struct af_iucv_trans_hdr *)skb->data;
1994
1995         if (!iucv)
1996                 goto out;
1997         if (sk->sk_state != IUCV_BOUND)
1998                 goto out;
1999         bh_lock_sock(sk);
2000         iucv->msglimit_peer = trans_hdr->window;
2001         sk->sk_state = IUCV_CONNECTED;
2002         sk->sk_state_change(sk);
2003         bh_unlock_sock(sk);
2004 out:
2005         kfree_skb(skb);
2006         return NET_RX_SUCCESS;
2007 }
2008
2009 /**
2010  * afiucv_hs_callback_synfin() - react on received SYN_FIN
2011  **/
2012 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
2013 {
2014         struct iucv_sock *iucv = iucv_sk(sk);
2015
2016         if (!iucv)
2017                 goto out;
2018         if (sk->sk_state != IUCV_BOUND)
2019                 goto out;
2020         bh_lock_sock(sk);
2021         sk->sk_state = IUCV_DISCONN;
2022         sk->sk_state_change(sk);
2023         bh_unlock_sock(sk);
2024 out:
2025         kfree_skb(skb);
2026         return NET_RX_SUCCESS;
2027 }
2028
2029 /**
2030  * afiucv_hs_callback_fin() - react on received FIN
2031  **/
2032 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
2033 {
2034         struct iucv_sock *iucv = iucv_sk(sk);
2035
2036         /* other end of connection closed */
2037         if (!iucv)
2038                 goto out;
2039         bh_lock_sock(sk);
2040         if (sk->sk_state == IUCV_CONNECTED) {
2041                 sk->sk_state = IUCV_DISCONN;
2042                 sk->sk_state_change(sk);
2043         }
2044         bh_unlock_sock(sk);
2045 out:
2046         kfree_skb(skb);
2047         return NET_RX_SUCCESS;
2048 }
2049
2050 /**
2051  * afiucv_hs_callback_win() - react on received WIN
2052  **/
2053 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
2054 {
2055         struct iucv_sock *iucv = iucv_sk(sk);
2056         struct af_iucv_trans_hdr *trans_hdr =
2057                                         (struct af_iucv_trans_hdr *)skb->data;
2058
2059         if (!iucv)
2060                 return NET_RX_SUCCESS;
2061
2062         if (sk->sk_state != IUCV_CONNECTED)
2063                 return NET_RX_SUCCESS;
2064
2065         atomic_sub(trans_hdr->window, &iucv->msg_sent);
2066         iucv_sock_wake_msglim(sk);
2067         return NET_RX_SUCCESS;
2068 }
2069
2070 /**
2071  * afiucv_hs_callback_rx() - react on received data
2072  **/
2073 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
2074 {
2075         struct iucv_sock *iucv = iucv_sk(sk);
2076
2077         if (!iucv) {
2078                 kfree_skb(skb);
2079                 return NET_RX_SUCCESS;
2080         }
2081
2082         if (sk->sk_state != IUCV_CONNECTED) {
2083                 kfree_skb(skb);
2084                 return NET_RX_SUCCESS;
2085         }
2086
2087         if (sk->sk_shutdown & RCV_SHUTDOWN) {
2088                 kfree_skb(skb);
2089                 return NET_RX_SUCCESS;
2090         }
2091
2092                 /* write stuff from iucv_msg to skb cb */
2093         if (skb->len < sizeof(struct af_iucv_trans_hdr)) {
2094                 kfree_skb(skb);
2095                 return NET_RX_SUCCESS;
2096         }
2097         skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2098         skb_reset_transport_header(skb);
2099         skb_reset_network_header(skb);
2100         IUCV_SKB_CB(skb)->offset = 0;
2101         spin_lock(&iucv->message_q.lock);
2102         if (skb_queue_empty(&iucv->backlog_skb_q)) {
2103                 if (sock_queue_rcv_skb(sk, skb)) {
2104                         /* handle rcv queue full */
2105                         skb_queue_tail(&iucv->backlog_skb_q, skb);
2106                 }
2107         } else
2108                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2109         spin_unlock(&iucv->message_q.lock);
2110         return NET_RX_SUCCESS;
2111 }
2112
2113 /**
2114  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2115  *                   transport
2116  *                   called from netif RX softirq
2117  **/
2118 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2119         struct packet_type *pt, struct net_device *orig_dev)
2120 {
2121         struct sock *sk;
2122         struct iucv_sock *iucv;
2123         struct af_iucv_trans_hdr *trans_hdr;
2124         char nullstring[8];
2125         int err = 0;
2126
2127         skb_pull(skb, ETH_HLEN);
2128         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
2129         EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2130         EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2131         EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2132         EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2133         memset(nullstring, 0, sizeof(nullstring));
2134         iucv = NULL;
2135         sk = NULL;
2136         read_lock(&iucv_sk_list.lock);
2137         sk_for_each(sk, &iucv_sk_list.head) {
2138                 if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2139                         if ((!memcmp(&iucv_sk(sk)->src_name,
2140                                      trans_hdr->destAppName, 8)) &&
2141                             (!memcmp(&iucv_sk(sk)->src_user_id,
2142                                      trans_hdr->destUserID, 8)) &&
2143                             (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2144                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2145                                      nullstring, 8))) {
2146                                 iucv = iucv_sk(sk);
2147                                 break;
2148                         }
2149                 } else {
2150                         if ((!memcmp(&iucv_sk(sk)->src_name,
2151                                      trans_hdr->destAppName, 8)) &&
2152                             (!memcmp(&iucv_sk(sk)->src_user_id,
2153                                      trans_hdr->destUserID, 8)) &&
2154                             (!memcmp(&iucv_sk(sk)->dst_name,
2155                                      trans_hdr->srcAppName, 8)) &&
2156                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2157                                      trans_hdr->srcUserID, 8))) {
2158                                 iucv = iucv_sk(sk);
2159                                 break;
2160                         }
2161                 }
2162         }
2163         read_unlock(&iucv_sk_list.lock);
2164         if (!iucv)
2165                 sk = NULL;
2166
2167         /* no sock
2168         how should we send with no sock
2169         1) send without sock no send rc checking?
2170         2) introduce default sock to handle this cases
2171
2172          SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2173          data -> send FIN
2174          SYN|ACK, SYN|FIN, FIN -> no action? */
2175
2176         switch (trans_hdr->flags) {
2177         case AF_IUCV_FLAG_SYN:
2178                 /* connect request */
2179                 err = afiucv_hs_callback_syn(sk, skb);
2180                 break;
2181         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2182                 /* connect request confirmed */
2183                 err = afiucv_hs_callback_synack(sk, skb);
2184                 break;
2185         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2186                 /* connect request refused */
2187                 err = afiucv_hs_callback_synfin(sk, skb);
2188                 break;
2189         case (AF_IUCV_FLAG_FIN):
2190                 /* close request */
2191                 err = afiucv_hs_callback_fin(sk, skb);
2192                 break;
2193         case (AF_IUCV_FLAG_WIN):
2194                 err = afiucv_hs_callback_win(sk, skb);
2195                 if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2196                         kfree_skb(skb);
2197                         break;
2198                 }
2199                 /* fall through and receive non-zero length data */
2200         case (AF_IUCV_FLAG_SHT):
2201                 /* shutdown request */
2202                 /* fall through and receive zero length data */
2203         case 0:
2204                 /* plain data frame */
2205                 IUCV_SKB_CB(skb)->class = trans_hdr->iucv_hdr.class;
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
2230         read_lock_irqsave(&iucv_sk_list.lock, flags);
2231         sk_for_each(sk, &iucv_sk_list.head)
2232                 if (sk == isk) {
2233                         iucv = iucv_sk(sk);
2234                         break;
2235                 }
2236         read_unlock_irqrestore(&iucv_sk_list.lock, flags);
2237
2238         if (!iucv || sock_flag(sk, SOCK_ZAPPED))
2239                 return;
2240
2241         list = &iucv->send_skb_q;
2242         spin_lock_irqsave(&list->lock, flags);
2243         if (skb_queue_empty(list))
2244                 goto out_unlock;
2245         list_skb = list->next;
2246         nskb = list_skb->next;
2247         while (list_skb != (struct sk_buff *)list) {
2248                 if (skb_shinfo(list_skb) == skb_shinfo(skb)) {
2249                         switch (n) {
2250                         case TX_NOTIFY_OK:
2251                                 __skb_unlink(list_skb, list);
2252                                 kfree_skb(list_skb);
2253                                 iucv_sock_wake_msglim(sk);
2254                                 break;
2255                         case TX_NOTIFY_PENDING:
2256                                 atomic_inc(&iucv->pendings);
2257                                 break;
2258                         case TX_NOTIFY_DELAYED_OK:
2259                                 __skb_unlink(list_skb, list);
2260                                 atomic_dec(&iucv->pendings);
2261                                 if (atomic_read(&iucv->pendings) <= 0)
2262                                         iucv_sock_wake_msglim(sk);
2263                                 kfree_skb(list_skb);
2264                                 break;
2265                         case TX_NOTIFY_UNREACHABLE:
2266                         case TX_NOTIFY_DELAYED_UNREACHABLE:
2267                         case TX_NOTIFY_TPQFULL: /* not yet used */
2268                         case TX_NOTIFY_GENERALERROR:
2269                         case TX_NOTIFY_DELAYED_GENERALERROR:
2270                                 __skb_unlink(list_skb, list);
2271                                 kfree_skb(list_skb);
2272                                 if (sk->sk_state == IUCV_CONNECTED) {
2273                                         sk->sk_state = IUCV_DISCONN;
2274                                         sk->sk_state_change(sk);
2275                                 }
2276                                 break;
2277                         }
2278                         break;
2279                 }
2280                 list_skb = nskb;
2281                 nskb = nskb->next;
2282         }
2283 out_unlock:
2284         spin_unlock_irqrestore(&list->lock, flags);
2285
2286         if (sk->sk_state == IUCV_CLOSING) {
2287                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
2288                         sk->sk_state = IUCV_CLOSED;
2289                         sk->sk_state_change(sk);
2290                 }
2291         }
2292
2293 }
2294
2295 /*
2296  * afiucv_netdev_event: handle netdev notifier chain events
2297  */
2298 static int afiucv_netdev_event(struct notifier_block *this,
2299                                unsigned long event, void *ptr)
2300 {
2301         struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
2302         struct sock *sk;
2303         struct iucv_sock *iucv;
2304
2305         switch (event) {
2306         case NETDEV_REBOOT:
2307         case NETDEV_GOING_DOWN:
2308                 sk_for_each(sk, &iucv_sk_list.head) {
2309                         iucv = iucv_sk(sk);
2310                         if ((iucv->hs_dev == event_dev) &&
2311                             (sk->sk_state == IUCV_CONNECTED)) {
2312                                 if (event == NETDEV_GOING_DOWN)
2313                                         iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2314                                 sk->sk_state = IUCV_DISCONN;
2315                                 sk->sk_state_change(sk);
2316                         }
2317                 }
2318                 break;
2319         case NETDEV_DOWN:
2320         case NETDEV_UNREGISTER:
2321         default:
2322                 break;
2323         }
2324         return NOTIFY_DONE;
2325 }
2326
2327 static struct notifier_block afiucv_netdev_notifier = {
2328         .notifier_call = afiucv_netdev_event,
2329 };
2330
2331 static const struct proto_ops iucv_sock_ops = {
2332         .family         = PF_IUCV,
2333         .owner          = THIS_MODULE,
2334         .release        = iucv_sock_release,
2335         .bind           = iucv_sock_bind,
2336         .connect        = iucv_sock_connect,
2337         .listen         = iucv_sock_listen,
2338         .accept         = iucv_sock_accept,
2339         .getname        = iucv_sock_getname,
2340         .sendmsg        = iucv_sock_sendmsg,
2341         .recvmsg        = iucv_sock_recvmsg,
2342         .poll           = iucv_sock_poll,
2343         .ioctl          = sock_no_ioctl,
2344         .mmap           = sock_no_mmap,
2345         .socketpair     = sock_no_socketpair,
2346         .shutdown       = iucv_sock_shutdown,
2347         .setsockopt     = iucv_sock_setsockopt,
2348         .getsockopt     = iucv_sock_getsockopt,
2349 };
2350
2351 static const struct net_proto_family iucv_sock_family_ops = {
2352         .family = AF_IUCV,
2353         .owner  = THIS_MODULE,
2354         .create = iucv_sock_create,
2355 };
2356
2357 static struct packet_type iucv_packet_type = {
2358         .type = cpu_to_be16(ETH_P_AF_IUCV),
2359         .func = afiucv_hs_rcv,
2360 };
2361
2362 static int afiucv_iucv_init(void)
2363 {
2364         int err;
2365
2366         err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2367         if (err)
2368                 goto out;
2369         /* establish dummy device */
2370         af_iucv_driver.bus = pr_iucv->bus;
2371         err = driver_register(&af_iucv_driver);
2372         if (err)
2373                 goto out_iucv;
2374         af_iucv_dev = kzalloc(sizeof(struct device), GFP_KERNEL);
2375         if (!af_iucv_dev) {
2376                 err = -ENOMEM;
2377                 goto out_driver;
2378         }
2379         dev_set_name(af_iucv_dev, "af_iucv");
2380         af_iucv_dev->bus = pr_iucv->bus;
2381         af_iucv_dev->parent = pr_iucv->root;
2382         af_iucv_dev->release = (void (*)(struct device *))kfree;
2383         af_iucv_dev->driver = &af_iucv_driver;
2384         err = device_register(af_iucv_dev);
2385         if (err)
2386                 goto out_driver;
2387         return 0;
2388
2389 out_driver:
2390         driver_unregister(&af_iucv_driver);
2391 out_iucv:
2392         pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2393 out:
2394         return err;
2395 }
2396
2397 static int __init afiucv_init(void)
2398 {
2399         int err;
2400
2401         if (MACHINE_IS_VM) {
2402                 cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2403                 if (unlikely(err)) {
2404                         WARN_ON(err);
2405                         err = -EPROTONOSUPPORT;
2406                         goto out;
2407                 }
2408
2409                 pr_iucv = try_then_request_module(symbol_get(iucv_if), "iucv");
2410                 if (!pr_iucv) {
2411                         printk(KERN_WARNING "iucv_if lookup failed\n");
2412                         memset(&iucv_userid, 0, sizeof(iucv_userid));
2413                 }
2414         } else {
2415                 memset(&iucv_userid, 0, sizeof(iucv_userid));
2416                 pr_iucv = NULL;
2417         }
2418
2419         err = proto_register(&iucv_proto, 0);
2420         if (err)
2421                 goto out;
2422         err = sock_register(&iucv_sock_family_ops);
2423         if (err)
2424                 goto out_proto;
2425
2426         if (pr_iucv) {
2427                 err = afiucv_iucv_init();
2428                 if (err)
2429                         goto out_sock;
2430         } else
2431                 register_netdevice_notifier(&afiucv_netdev_notifier);
2432         dev_add_pack(&iucv_packet_type);
2433         return 0;
2434
2435 out_sock:
2436         sock_unregister(PF_IUCV);
2437 out_proto:
2438         proto_unregister(&iucv_proto);
2439 out:
2440         if (pr_iucv)
2441                 symbol_put(iucv_if);
2442         return err;
2443 }
2444
2445 static void __exit afiucv_exit(void)
2446 {
2447         if (pr_iucv) {
2448                 device_unregister(af_iucv_dev);
2449                 driver_unregister(&af_iucv_driver);
2450                 pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2451                 symbol_put(iucv_if);
2452         } else
2453                 unregister_netdevice_notifier(&afiucv_netdev_notifier);
2454         dev_remove_pack(&iucv_packet_type);
2455         sock_unregister(PF_IUCV);
2456         proto_unregister(&iucv_proto);
2457 }
2458
2459 module_init(afiucv_init);
2460 module_exit(afiucv_exit);
2461
2462 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2463 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2464 MODULE_VERSION(VERSION);
2465 MODULE_LICENSE("GPL");
2466 MODULE_ALIAS_NETPROTO(PF_IUCV);
2467