net: add option to enable error queue packets waking select
[linux-3.10.git] / net / iucv / af_iucv.c
1 /*
2  *  IUCV protocol stack for Linux on zSeries
3  *
4  *  Copyright IBM Corp. 2006, 2009
5  *
6  *  Author(s):  Jennifer Hunt <jenhunt@us.ibm.com>
7  *              Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
8  *  PM functions:
9  *              Ursula Braun <ursula.braun@de.ibm.com>
10  */
11
12 #define KMSG_COMPONENT "af_iucv"
13 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
14
15 #include <linux/module.h>
16 #include <linux/types.h>
17 #include <linux/list.h>
18 #include <linux/errno.h>
19 #include <linux/kernel.h>
20 #include <linux/sched.h>
21 #include <linux/slab.h>
22 #include <linux/skbuff.h>
23 #include <linux/init.h>
24 #include <linux/poll.h>
25 #include <net/sock.h>
26 #include <asm/ebcdic.h>
27 #include <asm/cpcmd.h>
28 #include <linux/kmod.h>
29
30 #include <net/iucv/af_iucv.h>
31
32 #define VERSION "1.2"
33
34 static char iucv_userid[80];
35
36 static const struct proto_ops iucv_sock_ops;
37
38 static struct proto iucv_proto = {
39         .name           = "AF_IUCV",
40         .owner          = THIS_MODULE,
41         .obj_size       = sizeof(struct iucv_sock),
42 };
43
44 static struct iucv_interface *pr_iucv;
45
46 /* special AF_IUCV IPRM messages */
47 static const u8 iprm_shutdown[8] =
48         {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
49
50 #define TRGCLS_SIZE     (sizeof(((struct iucv_message *)0)->class))
51
52 /* macros to set/get socket control buffer at correct offset */
53 #define CB_TAG(skb)     ((skb)->cb)             /* iucv message tag */
54 #define CB_TAG_LEN      (sizeof(((struct iucv_message *) 0)->tag))
55 #define CB_TRGCLS(skb)  ((skb)->cb + CB_TAG_LEN) /* iucv msg target class */
56 #define CB_TRGCLS_LEN   (TRGCLS_SIZE)
57
58 #define __iucv_sock_wait(sk, condition, timeo, ret)                     \
59 do {                                                                    \
60         DEFINE_WAIT(__wait);                                            \
61         long __timeo = timeo;                                           \
62         ret = 0;                                                        \
63         prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);     \
64         while (!(condition)) {                                          \
65                 if (!__timeo) {                                         \
66                         ret = -EAGAIN;                                  \
67                         break;                                          \
68                 }                                                       \
69                 if (signal_pending(current)) {                          \
70                         ret = sock_intr_errno(__timeo);                 \
71                         break;                                          \
72                 }                                                       \
73                 release_sock(sk);                                       \
74                 __timeo = schedule_timeout(__timeo);                    \
75                 lock_sock(sk);                                          \
76                 ret = sock_error(sk);                                   \
77                 if (ret)                                                \
78                         break;                                          \
79         }                                                               \
80         finish_wait(sk_sleep(sk), &__wait);                             \
81 } while (0)
82
83 #define iucv_sock_wait(sk, condition, timeo)                            \
84 ({                                                                      \
85         int __ret = 0;                                                  \
86         if (!(condition))                                               \
87                 __iucv_sock_wait(sk, condition, timeo, __ret);          \
88         __ret;                                                          \
89 })
90
91 static void iucv_sock_kill(struct sock *sk);
92 static void iucv_sock_close(struct sock *sk);
93 static void iucv_sever_path(struct sock *, int);
94
95 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
96         struct packet_type *pt, struct net_device *orig_dev);
97 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
98                    struct sk_buff *skb, u8 flags);
99 static void afiucv_hs_callback_txnotify(struct sk_buff *, enum iucv_tx_notify);
100
101 /* Call Back functions */
102 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
103 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
104 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
105 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
106                                  u8 ipuser[16]);
107 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
108 static void iucv_callback_shutdown(struct iucv_path *, u8 ipuser[16]);
109
110 static struct iucv_sock_list iucv_sk_list = {
111         .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
112         .autobind_name = ATOMIC_INIT(0)
113 };
114
115 static struct iucv_handler af_iucv_handler = {
116         .path_pending     = iucv_callback_connreq,
117         .path_complete    = iucv_callback_connack,
118         .path_severed     = iucv_callback_connrej,
119         .message_pending  = iucv_callback_rx,
120         .message_complete = iucv_callback_txdone,
121         .path_quiesced    = iucv_callback_shutdown,
122 };
123
124 static inline void high_nmcpy(unsigned char *dst, char *src)
125 {
126        memcpy(dst, src, 8);
127 }
128
129 static inline void low_nmcpy(unsigned char *dst, char *src)
130 {
131        memcpy(&dst[8], src, 8);
132 }
133
134 static int afiucv_pm_prepare(struct device *dev)
135 {
136 #ifdef CONFIG_PM_DEBUG
137         printk(KERN_WARNING "afiucv_pm_prepare\n");
138 #endif
139         return 0;
140 }
141
142 static void afiucv_pm_complete(struct device *dev)
143 {
144 #ifdef CONFIG_PM_DEBUG
145         printk(KERN_WARNING "afiucv_pm_complete\n");
146 #endif
147 }
148
149 /**
150  * afiucv_pm_freeze() - Freeze PM callback
151  * @dev:        AFIUCV dummy device
152  *
153  * Sever all established IUCV communication pathes
154  */
155 static int afiucv_pm_freeze(struct device *dev)
156 {
157         struct iucv_sock *iucv;
158         struct sock *sk;
159         int err = 0;
160
161 #ifdef CONFIG_PM_DEBUG
162         printk(KERN_WARNING "afiucv_pm_freeze\n");
163 #endif
164         read_lock(&iucv_sk_list.lock);
165         sk_for_each(sk, &iucv_sk_list.head) {
166                 iucv = iucv_sk(sk);
167                 switch (sk->sk_state) {
168                 case IUCV_DISCONN:
169                 case IUCV_CLOSING:
170                 case IUCV_CONNECTED:
171                         iucv_sever_path(sk, 0);
172                         break;
173                 case IUCV_OPEN:
174                 case IUCV_BOUND:
175                 case IUCV_LISTEN:
176                 case IUCV_CLOSED:
177                 default:
178                         break;
179                 }
180                 skb_queue_purge(&iucv->send_skb_q);
181                 skb_queue_purge(&iucv->backlog_skb_q);
182         }
183         read_unlock(&iucv_sk_list.lock);
184         return err;
185 }
186
187 /**
188  * afiucv_pm_restore_thaw() - Thaw and restore PM callback
189  * @dev:        AFIUCV dummy device
190  *
191  * socket clean up after freeze
192  */
193 static int afiucv_pm_restore_thaw(struct device *dev)
194 {
195         struct sock *sk;
196
197 #ifdef CONFIG_PM_DEBUG
198         printk(KERN_WARNING "afiucv_pm_restore_thaw\n");
199 #endif
200         read_lock(&iucv_sk_list.lock);
201         sk_for_each(sk, &iucv_sk_list.head) {
202                 switch (sk->sk_state) {
203                 case IUCV_CONNECTED:
204                         sk->sk_err = EPIPE;
205                         sk->sk_state = IUCV_DISCONN;
206                         sk->sk_state_change(sk);
207                         break;
208                 case IUCV_DISCONN:
209                 case IUCV_CLOSING:
210                 case IUCV_LISTEN:
211                 case IUCV_BOUND:
212                 case IUCV_OPEN:
213                 default:
214                         break;
215                 }
216         }
217         read_unlock(&iucv_sk_list.lock);
218         return 0;
219 }
220
221 static const struct dev_pm_ops afiucv_pm_ops = {
222         .prepare = afiucv_pm_prepare,
223         .complete = afiucv_pm_complete,
224         .freeze = afiucv_pm_freeze,
225         .thaw = afiucv_pm_restore_thaw,
226         .restore = afiucv_pm_restore_thaw,
227 };
228
229 static struct device_driver af_iucv_driver = {
230         .owner = THIS_MODULE,
231         .name = "afiucv",
232         .bus  = NULL,
233         .pm   = &afiucv_pm_ops,
234 };
235
236 /* dummy device used as trigger for PM functions */
237 static struct device *af_iucv_dev;
238
239 /**
240  * iucv_msg_length() - Returns the length of an iucv message.
241  * @msg:        Pointer to struct iucv_message, MUST NOT be NULL
242  *
243  * The function returns the length of the specified iucv message @msg of data
244  * stored in a buffer and of data stored in the parameter list (PRMDATA).
245  *
246  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
247  * data:
248  *      PRMDATA[0..6]   socket data (max 7 bytes);
249  *      PRMDATA[7]      socket data length value (len is 0xff - PRMDATA[7])
250  *
251  * The socket data length is computed by subtracting the socket data length
252  * value from 0xFF.
253  * If the socket data len is greater 7, then PRMDATA can be used for special
254  * notifications (see iucv_sock_shutdown); and further,
255  * if the socket data len is > 7, the function returns 8.
256  *
257  * Use this function to allocate socket buffers to store iucv message data.
258  */
259 static inline size_t iucv_msg_length(struct iucv_message *msg)
260 {
261         size_t datalen;
262
263         if (msg->flags & IUCV_IPRMDATA) {
264                 datalen = 0xff - msg->rmmsg[7];
265                 return (datalen < 8) ? datalen : 8;
266         }
267         return msg->length;
268 }
269
270 /**
271  * iucv_sock_in_state() - check for specific states
272  * @sk:         sock structure
273  * @state:      first iucv sk state
274  * @state:      second iucv sk state
275  *
276  * Returns true if the socket in either in the first or second state.
277  */
278 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
279 {
280         return (sk->sk_state == state || sk->sk_state == state2);
281 }
282
283 /**
284  * iucv_below_msglim() - function to check if messages can be sent
285  * @sk:         sock structure
286  *
287  * Returns true if the send queue length is lower than the message limit.
288  * Always returns true if the socket is not connected (no iucv path for
289  * checking the message limit).
290  */
291 static inline int iucv_below_msglim(struct sock *sk)
292 {
293         struct iucv_sock *iucv = iucv_sk(sk);
294
295         if (sk->sk_state != IUCV_CONNECTED)
296                 return 1;
297         if (iucv->transport == AF_IUCV_TRANS_IUCV)
298                 return (skb_queue_len(&iucv->send_skb_q) < iucv->path->msglim);
299         else
300                 return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
301                         (atomic_read(&iucv->pendings) <= 0));
302 }
303
304 /**
305  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
306  */
307 static void iucv_sock_wake_msglim(struct sock *sk)
308 {
309         struct socket_wq *wq;
310
311         rcu_read_lock();
312         wq = rcu_dereference(sk->sk_wq);
313         if (wq_has_sleeper(wq))
314                 wake_up_interruptible_all(&wq->wait);
315         sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
316         rcu_read_unlock();
317 }
318
319 /**
320  * afiucv_hs_send() - send a message through HiperSockets transport
321  */
322 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
323                    struct sk_buff *skb, u8 flags)
324 {
325         struct iucv_sock *iucv = iucv_sk(sock);
326         struct af_iucv_trans_hdr *phs_hdr;
327         struct sk_buff *nskb;
328         int err, confirm_recv = 0;
329
330         memset(skb->head, 0, ETH_HLEN);
331         phs_hdr = (struct af_iucv_trans_hdr *)skb_push(skb,
332                                         sizeof(struct af_iucv_trans_hdr));
333         skb_reset_mac_header(skb);
334         skb_reset_network_header(skb);
335         skb_push(skb, ETH_HLEN);
336         skb_reset_mac_header(skb);
337         memset(phs_hdr, 0, sizeof(struct af_iucv_trans_hdr));
338
339         phs_hdr->magic = ETH_P_AF_IUCV;
340         phs_hdr->version = 1;
341         phs_hdr->flags = flags;
342         if (flags == AF_IUCV_FLAG_SYN)
343                 phs_hdr->window = iucv->msglimit;
344         else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
345                 confirm_recv = atomic_read(&iucv->msg_recv);
346                 phs_hdr->window = confirm_recv;
347                 if (confirm_recv)
348                         phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
349         }
350         memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
351         memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
352         memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
353         memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
354         ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
355         ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
356         ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
357         ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
358         if (imsg)
359                 memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
360
361         skb->dev = iucv->hs_dev;
362         if (!skb->dev)
363                 return -ENODEV;
364         if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev))
365                 return -ENETDOWN;
366         if (skb->len > skb->dev->mtu) {
367                 if (sock->sk_type == SOCK_SEQPACKET)
368                         return -EMSGSIZE;
369                 else
370                         skb_trim(skb, skb->dev->mtu);
371         }
372         skb->protocol = ETH_P_AF_IUCV;
373         nskb = skb_clone(skb, GFP_ATOMIC);
374         if (!nskb)
375                 return -ENOMEM;
376         skb_queue_tail(&iucv->send_skb_q, nskb);
377         err = dev_queue_xmit(skb);
378         if (net_xmit_eval(err)) {
379                 skb_unlink(nskb, &iucv->send_skb_q);
380                 kfree_skb(nskb);
381         } else {
382                 atomic_sub(confirm_recv, &iucv->msg_recv);
383                 WARN_ON(atomic_read(&iucv->msg_recv) < 0);
384         }
385         return net_xmit_eval(err);
386 }
387
388 static struct sock *__iucv_get_sock_by_name(char *nm)
389 {
390         struct sock *sk;
391
392         sk_for_each(sk, &iucv_sk_list.head)
393                 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
394                         return sk;
395
396         return NULL;
397 }
398
399 static void iucv_sock_destruct(struct sock *sk)
400 {
401         skb_queue_purge(&sk->sk_receive_queue);
402         skb_queue_purge(&sk->sk_error_queue);
403
404         sk_mem_reclaim(sk);
405
406         if (!sock_flag(sk, SOCK_DEAD)) {
407                 pr_err("Attempt to release alive iucv socket %p\n", sk);
408                 return;
409         }
410
411         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
412         WARN_ON(atomic_read(&sk->sk_wmem_alloc));
413         WARN_ON(sk->sk_wmem_queued);
414         WARN_ON(sk->sk_forward_alloc);
415 }
416
417 /* Cleanup Listen */
418 static void iucv_sock_cleanup_listen(struct sock *parent)
419 {
420         struct sock *sk;
421
422         /* Close non-accepted connections */
423         while ((sk = iucv_accept_dequeue(parent, NULL))) {
424                 iucv_sock_close(sk);
425                 iucv_sock_kill(sk);
426         }
427
428         parent->sk_state = IUCV_CLOSED;
429 }
430
431 /* Kill socket (only if zapped and orphaned) */
432 static void iucv_sock_kill(struct sock *sk)
433 {
434         if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
435                 return;
436
437         iucv_sock_unlink(&iucv_sk_list, sk);
438         sock_set_flag(sk, SOCK_DEAD);
439         sock_put(sk);
440 }
441
442 /* Terminate an IUCV path */
443 static void iucv_sever_path(struct sock *sk, int with_user_data)
444 {
445         unsigned char user_data[16];
446         struct iucv_sock *iucv = iucv_sk(sk);
447         struct iucv_path *path = iucv->path;
448
449         if (iucv->path) {
450                 iucv->path = NULL;
451                 if (with_user_data) {
452                         low_nmcpy(user_data, iucv->src_name);
453                         high_nmcpy(user_data, iucv->dst_name);
454                         ASCEBC(user_data, sizeof(user_data));
455                         pr_iucv->path_sever(path, user_data);
456                 } else
457                         pr_iucv->path_sever(path, NULL);
458                 iucv_path_free(path);
459         }
460 }
461
462 /* Send FIN through an IUCV socket for HIPER transport */
463 static int iucv_send_ctrl(struct sock *sk, u8 flags)
464 {
465         int err = 0;
466         int blen;
467         struct sk_buff *skb;
468
469         blen = sizeof(struct af_iucv_trans_hdr) + ETH_HLEN;
470         skb = sock_alloc_send_skb(sk, blen, 1, &err);
471         if (skb) {
472                 skb_reserve(skb, blen);
473                 err = afiucv_hs_send(NULL, sk, skb, flags);
474         }
475         return err;
476 }
477
478 /* Close an IUCV socket */
479 static void iucv_sock_close(struct sock *sk)
480 {
481         struct iucv_sock *iucv = iucv_sk(sk);
482         unsigned long timeo;
483         int err = 0;
484
485         lock_sock(sk);
486
487         switch (sk->sk_state) {
488         case IUCV_LISTEN:
489                 iucv_sock_cleanup_listen(sk);
490                 break;
491
492         case IUCV_CONNECTED:
493                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
494                         err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
495                         sk->sk_state = IUCV_DISCONN;
496                         sk->sk_state_change(sk);
497                 }
498         case IUCV_DISCONN:   /* fall through */
499                 sk->sk_state = IUCV_CLOSING;
500                 sk->sk_state_change(sk);
501
502                 if (!err && !skb_queue_empty(&iucv->send_skb_q)) {
503                         if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
504                                 timeo = sk->sk_lingertime;
505                         else
506                                 timeo = IUCV_DISCONN_TIMEOUT;
507                         iucv_sock_wait(sk,
508                                         iucv_sock_in_state(sk, IUCV_CLOSED, 0),
509                                         timeo);
510                 }
511
512         case IUCV_CLOSING:   /* fall through */
513                 sk->sk_state = IUCV_CLOSED;
514                 sk->sk_state_change(sk);
515
516                 sk->sk_err = ECONNRESET;
517                 sk->sk_state_change(sk);
518
519                 skb_queue_purge(&iucv->send_skb_q);
520                 skb_queue_purge(&iucv->backlog_skb_q);
521
522         default:   /* fall through */
523                 iucv_sever_path(sk, 1);
524         }
525
526         if (iucv->hs_dev) {
527                 dev_put(iucv->hs_dev);
528                 iucv->hs_dev = NULL;
529                 sk->sk_bound_dev_if = 0;
530         }
531
532         /* mark socket for deletion by iucv_sock_kill() */
533         sock_set_flag(sk, SOCK_ZAPPED);
534
535         release_sock(sk);
536 }
537
538 static void iucv_sock_init(struct sock *sk, struct sock *parent)
539 {
540         if (parent)
541                 sk->sk_type = parent->sk_type;
542 }
543
544 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
545 {
546         struct sock *sk;
547         struct iucv_sock *iucv;
548
549         sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
550         if (!sk)
551                 return NULL;
552         iucv = iucv_sk(sk);
553
554         sock_init_data(sock, sk);
555         INIT_LIST_HEAD(&iucv->accept_q);
556         spin_lock_init(&iucv->accept_q_lock);
557         skb_queue_head_init(&iucv->send_skb_q);
558         INIT_LIST_HEAD(&iucv->message_q.list);
559         spin_lock_init(&iucv->message_q.lock);
560         skb_queue_head_init(&iucv->backlog_skb_q);
561         iucv->send_tag = 0;
562         atomic_set(&iucv->pendings, 0);
563         iucv->flags = 0;
564         iucv->msglimit = 0;
565         atomic_set(&iucv->msg_sent, 0);
566         atomic_set(&iucv->msg_recv, 0);
567         iucv->path = NULL;
568         iucv->sk_txnotify = afiucv_hs_callback_txnotify;
569         memset(&iucv->src_user_id , 0, 32);
570         if (pr_iucv)
571                 iucv->transport = AF_IUCV_TRANS_IUCV;
572         else
573                 iucv->transport = AF_IUCV_TRANS_HIPER;
574
575         sk->sk_destruct = iucv_sock_destruct;
576         sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
577         sk->sk_allocation = GFP_DMA;
578
579         sock_reset_flag(sk, SOCK_ZAPPED);
580
581         sk->sk_protocol = proto;
582         sk->sk_state    = IUCV_OPEN;
583
584         iucv_sock_link(&iucv_sk_list, sk);
585         return sk;
586 }
587
588 /* Create an IUCV socket */
589 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
590                             int kern)
591 {
592         struct sock *sk;
593
594         if (protocol && protocol != PF_IUCV)
595                 return -EPROTONOSUPPORT;
596
597         sock->state = SS_UNCONNECTED;
598
599         switch (sock->type) {
600         case SOCK_STREAM:
601                 sock->ops = &iucv_sock_ops;
602                 break;
603         case SOCK_SEQPACKET:
604                 /* currently, proto ops can handle both sk types */
605                 sock->ops = &iucv_sock_ops;
606                 break;
607         default:
608                 return -ESOCKTNOSUPPORT;
609         }
610
611         sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
612         if (!sk)
613                 return -ENOMEM;
614
615         iucv_sock_init(sk, NULL);
616
617         return 0;
618 }
619
620 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
621 {
622         write_lock_bh(&l->lock);
623         sk_add_node(sk, &l->head);
624         write_unlock_bh(&l->lock);
625 }
626
627 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
628 {
629         write_lock_bh(&l->lock);
630         sk_del_node_init(sk);
631         write_unlock_bh(&l->lock);
632 }
633
634 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
635 {
636         unsigned long flags;
637         struct iucv_sock *par = iucv_sk(parent);
638
639         sock_hold(sk);
640         spin_lock_irqsave(&par->accept_q_lock, flags);
641         list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
642         spin_unlock_irqrestore(&par->accept_q_lock, flags);
643         iucv_sk(sk)->parent = parent;
644         sk_acceptq_added(parent);
645 }
646
647 void iucv_accept_unlink(struct sock *sk)
648 {
649         unsigned long flags;
650         struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
651
652         spin_lock_irqsave(&par->accept_q_lock, flags);
653         list_del_init(&iucv_sk(sk)->accept_q);
654         spin_unlock_irqrestore(&par->accept_q_lock, flags);
655         sk_acceptq_removed(iucv_sk(sk)->parent);
656         iucv_sk(sk)->parent = NULL;
657         sock_put(sk);
658 }
659
660 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
661 {
662         struct iucv_sock *isk, *n;
663         struct sock *sk;
664
665         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
666                 sk = (struct sock *) isk;
667                 lock_sock(sk);
668
669                 if (sk->sk_state == IUCV_CLOSED) {
670                         iucv_accept_unlink(sk);
671                         release_sock(sk);
672                         continue;
673                 }
674
675                 if (sk->sk_state == IUCV_CONNECTED ||
676                     sk->sk_state == IUCV_DISCONN ||
677                     !newsock) {
678                         iucv_accept_unlink(sk);
679                         if (newsock)
680                                 sock_graft(sk, newsock);
681
682                         release_sock(sk);
683                         return sk;
684                 }
685
686                 release_sock(sk);
687         }
688         return NULL;
689 }
690
691 /* Bind an unbound socket */
692 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
693                           int addr_len)
694 {
695         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
696         struct sock *sk = sock->sk;
697         struct iucv_sock *iucv;
698         int err = 0;
699         struct net_device *dev;
700         char uid[9];
701
702         /* Verify the input sockaddr */
703         if (!addr || addr->sa_family != AF_IUCV)
704                 return -EINVAL;
705
706         lock_sock(sk);
707         if (sk->sk_state != IUCV_OPEN) {
708                 err = -EBADFD;
709                 goto done;
710         }
711
712         write_lock_bh(&iucv_sk_list.lock);
713
714         iucv = iucv_sk(sk);
715         if (__iucv_get_sock_by_name(sa->siucv_name)) {
716                 err = -EADDRINUSE;
717                 goto done_unlock;
718         }
719         if (iucv->path)
720                 goto done_unlock;
721
722         /* Bind the socket */
723         if (pr_iucv)
724                 if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
725                         goto vm_bind; /* VM IUCV transport */
726
727         /* try hiper transport */
728         memcpy(uid, sa->siucv_user_id, sizeof(uid));
729         ASCEBC(uid, 8);
730         rcu_read_lock();
731         for_each_netdev_rcu(&init_net, dev) {
732                 if (!memcmp(dev->perm_addr, uid, 8)) {
733                         memcpy(iucv->src_name, sa->siucv_name, 8);
734                         memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
735                         sk->sk_bound_dev_if = dev->ifindex;
736                         iucv->hs_dev = dev;
737                         dev_hold(dev);
738                         sk->sk_state = IUCV_BOUND;
739                         iucv->transport = AF_IUCV_TRANS_HIPER;
740                         if (!iucv->msglimit)
741                                 iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
742                         rcu_read_unlock();
743                         goto done_unlock;
744                 }
745         }
746         rcu_read_unlock();
747 vm_bind:
748         if (pr_iucv) {
749                 /* use local userid for backward compat */
750                 memcpy(iucv->src_name, sa->siucv_name, 8);
751                 memcpy(iucv->src_user_id, iucv_userid, 8);
752                 sk->sk_state = IUCV_BOUND;
753                 iucv->transport = AF_IUCV_TRANS_IUCV;
754                 if (!iucv->msglimit)
755                         iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
756                 goto done_unlock;
757         }
758         /* found no dev to bind */
759         err = -ENODEV;
760 done_unlock:
761         /* Release the socket list lock */
762         write_unlock_bh(&iucv_sk_list.lock);
763 done:
764         release_sock(sk);
765         return err;
766 }
767
768 /* Automatically bind an unbound socket */
769 static int iucv_sock_autobind(struct sock *sk)
770 {
771         struct iucv_sock *iucv = iucv_sk(sk);
772         char name[12];
773         int err = 0;
774
775         if (unlikely(!pr_iucv))
776                 return -EPROTO;
777
778         memcpy(iucv->src_user_id, iucv_userid, 8);
779
780         write_lock_bh(&iucv_sk_list.lock);
781
782         sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
783         while (__iucv_get_sock_by_name(name)) {
784                 sprintf(name, "%08x",
785                         atomic_inc_return(&iucv_sk_list.autobind_name));
786         }
787
788         write_unlock_bh(&iucv_sk_list.lock);
789
790         memcpy(&iucv->src_name, name, 8);
791
792         if (!iucv->msglimit)
793                 iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
794
795         return err;
796 }
797
798 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
799 {
800         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
801         struct sock *sk = sock->sk;
802         struct iucv_sock *iucv = iucv_sk(sk);
803         unsigned char user_data[16];
804         int err;
805
806         high_nmcpy(user_data, sa->siucv_name);
807         low_nmcpy(user_data, iucv->src_name);
808         ASCEBC(user_data, sizeof(user_data));
809
810         /* Create path. */
811         iucv->path = iucv_path_alloc(iucv->msglimit,
812                                      IUCV_IPRMDATA, GFP_KERNEL);
813         if (!iucv->path) {
814                 err = -ENOMEM;
815                 goto done;
816         }
817         err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
818                                     sa->siucv_user_id, NULL, user_data,
819                                     sk);
820         if (err) {
821                 iucv_path_free(iucv->path);
822                 iucv->path = NULL;
823                 switch (err) {
824                 case 0x0b:      /* Target communicator is not logged on */
825                         err = -ENETUNREACH;
826                         break;
827                 case 0x0d:      /* Max connections for this guest exceeded */
828                 case 0x0e:      /* Max connections for target guest exceeded */
829                         err = -EAGAIN;
830                         break;
831                 case 0x0f:      /* Missing IUCV authorization */
832                         err = -EACCES;
833                         break;
834                 default:
835                         err = -ECONNREFUSED;
836                         break;
837                 }
838         }
839 done:
840         return err;
841 }
842
843 /* Connect an unconnected socket */
844 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
845                              int alen, int flags)
846 {
847         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
848         struct sock *sk = sock->sk;
849         struct iucv_sock *iucv = iucv_sk(sk);
850         int err;
851
852         if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
853                 return -EINVAL;
854
855         if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
856                 return -EBADFD;
857
858         if (sk->sk_state == IUCV_OPEN &&
859             iucv->transport == AF_IUCV_TRANS_HIPER)
860                 return -EBADFD; /* explicit bind required */
861
862         if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
863                 return -EINVAL;
864
865         if (sk->sk_state == IUCV_OPEN) {
866                 err = iucv_sock_autobind(sk);
867                 if (unlikely(err))
868                         return err;
869         }
870
871         lock_sock(sk);
872
873         /* Set the destination information */
874         memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
875         memcpy(iucv->dst_name, sa->siucv_name, 8);
876
877         if (iucv->transport == AF_IUCV_TRANS_HIPER)
878                 err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
879         else
880                 err = afiucv_path_connect(sock, addr);
881         if (err)
882                 goto done;
883
884         if (sk->sk_state != IUCV_CONNECTED)
885                 err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
886                                                             IUCV_DISCONN),
887                                      sock_sndtimeo(sk, flags & O_NONBLOCK));
888
889         if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
890                 err = -ECONNREFUSED;
891
892         if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
893                 iucv_sever_path(sk, 0);
894
895 done:
896         release_sock(sk);
897         return err;
898 }
899
900 /* Move a socket into listening state. */
901 static int iucv_sock_listen(struct socket *sock, int backlog)
902 {
903         struct sock *sk = sock->sk;
904         int err;
905
906         lock_sock(sk);
907
908         err = -EINVAL;
909         if (sk->sk_state != IUCV_BOUND)
910                 goto done;
911
912         if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
913                 goto done;
914
915         sk->sk_max_ack_backlog = backlog;
916         sk->sk_ack_backlog = 0;
917         sk->sk_state = IUCV_LISTEN;
918         err = 0;
919
920 done:
921         release_sock(sk);
922         return err;
923 }
924
925 /* Accept a pending connection */
926 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
927                             int flags)
928 {
929         DECLARE_WAITQUEUE(wait, current);
930         struct sock *sk = sock->sk, *nsk;
931         long timeo;
932         int err = 0;
933
934         lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
935
936         if (sk->sk_state != IUCV_LISTEN) {
937                 err = -EBADFD;
938                 goto done;
939         }
940
941         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
942
943         /* Wait for an incoming connection */
944         add_wait_queue_exclusive(sk_sleep(sk), &wait);
945         while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
946                 set_current_state(TASK_INTERRUPTIBLE);
947                 if (!timeo) {
948                         err = -EAGAIN;
949                         break;
950                 }
951
952                 release_sock(sk);
953                 timeo = schedule_timeout(timeo);
954                 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
955
956                 if (sk->sk_state != IUCV_LISTEN) {
957                         err = -EBADFD;
958                         break;
959                 }
960
961                 if (signal_pending(current)) {
962                         err = sock_intr_errno(timeo);
963                         break;
964                 }
965         }
966
967         set_current_state(TASK_RUNNING);
968         remove_wait_queue(sk_sleep(sk), &wait);
969
970         if (err)
971                 goto done;
972
973         newsock->state = SS_CONNECTED;
974
975 done:
976         release_sock(sk);
977         return err;
978 }
979
980 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
981                              int *len, int peer)
982 {
983         struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
984         struct sock *sk = sock->sk;
985         struct iucv_sock *iucv = iucv_sk(sk);
986
987         addr->sa_family = AF_IUCV;
988         *len = sizeof(struct sockaddr_iucv);
989
990         if (peer) {
991                 memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
992                 memcpy(siucv->siucv_name, iucv->dst_name, 8);
993         } else {
994                 memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
995                 memcpy(siucv->siucv_name, iucv->src_name, 8);
996         }
997         memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
998         memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
999         memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
1000
1001         return 0;
1002 }
1003
1004 /**
1005  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
1006  * @path:       IUCV path
1007  * @msg:        Pointer to a struct iucv_message
1008  * @skb:        The socket data to send, skb->len MUST BE <= 7
1009  *
1010  * Send the socket data in the parameter list in the iucv message
1011  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
1012  * list and the socket data len at index 7 (last byte).
1013  * See also iucv_msg_length().
1014  *
1015  * Returns the error code from the iucv_message_send() call.
1016  */
1017 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
1018                           struct sk_buff *skb)
1019 {
1020         u8 prmdata[8];
1021
1022         memcpy(prmdata, (void *) skb->data, skb->len);
1023         prmdata[7] = 0xff - (u8) skb->len;
1024         return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
1025                                  (void *) prmdata, 8);
1026 }
1027
1028 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1029                              struct msghdr *msg, size_t len)
1030 {
1031         struct sock *sk = sock->sk;
1032         struct iucv_sock *iucv = iucv_sk(sk);
1033         struct sk_buff *skb;
1034         struct iucv_message txmsg;
1035         struct cmsghdr *cmsg;
1036         int cmsg_done;
1037         long timeo;
1038         char user_id[9];
1039         char appl_id[9];
1040         int err;
1041         int noblock = msg->msg_flags & MSG_DONTWAIT;
1042
1043         err = sock_error(sk);
1044         if (err)
1045                 return err;
1046
1047         if (msg->msg_flags & MSG_OOB)
1048                 return -EOPNOTSUPP;
1049
1050         /* SOCK_SEQPACKET: we do not support segmented records */
1051         if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
1052                 return -EOPNOTSUPP;
1053
1054         lock_sock(sk);
1055
1056         if (sk->sk_shutdown & SEND_SHUTDOWN) {
1057                 err = -EPIPE;
1058                 goto out;
1059         }
1060
1061         /* Return if the socket is not in connected state */
1062         if (sk->sk_state != IUCV_CONNECTED) {
1063                 err = -ENOTCONN;
1064                 goto out;
1065         }
1066
1067         /* initialize defaults */
1068         cmsg_done   = 0;        /* check for duplicate headers */
1069         txmsg.class = 0;
1070
1071         /* iterate over control messages */
1072         for (cmsg = CMSG_FIRSTHDR(msg); cmsg;
1073                 cmsg = CMSG_NXTHDR(msg, cmsg)) {
1074
1075                 if (!CMSG_OK(msg, cmsg)) {
1076                         err = -EINVAL;
1077                         goto out;
1078                 }
1079
1080                 if (cmsg->cmsg_level != SOL_IUCV)
1081                         continue;
1082
1083                 if (cmsg->cmsg_type & cmsg_done) {
1084                         err = -EINVAL;
1085                         goto out;
1086                 }
1087                 cmsg_done |= cmsg->cmsg_type;
1088
1089                 switch (cmsg->cmsg_type) {
1090                 case SCM_IUCV_TRGCLS:
1091                         if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
1092                                 err = -EINVAL;
1093                                 goto out;
1094                         }
1095
1096                         /* set iucv message target class */
1097                         memcpy(&txmsg.class,
1098                                 (void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
1099
1100                         break;
1101
1102                 default:
1103                         err = -EINVAL;
1104                         goto out;
1105                         break;
1106                 }
1107         }
1108
1109         /* allocate one skb for each iucv message:
1110          * this is fine for SOCK_SEQPACKET (unless we want to support
1111          * segmented records using the MSG_EOR flag), but
1112          * for SOCK_STREAM we might want to improve it in future */
1113         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1114                 skb = sock_alloc_send_skb(sk,
1115                         len + sizeof(struct af_iucv_trans_hdr) + ETH_HLEN,
1116                         noblock, &err);
1117         else
1118                 skb = sock_alloc_send_skb(sk, len, noblock, &err);
1119         if (!skb) {
1120                 err = -ENOMEM;
1121                 goto out;
1122         }
1123         if (iucv->transport == AF_IUCV_TRANS_HIPER)
1124                 skb_reserve(skb, sizeof(struct af_iucv_trans_hdr) + ETH_HLEN);
1125         if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1126                 err = -EFAULT;
1127                 goto fail;
1128         }
1129
1130         /* wait if outstanding messages for iucv path has reached */
1131         timeo = sock_sndtimeo(sk, noblock);
1132         err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1133         if (err)
1134                 goto fail;
1135
1136         /* return -ECONNRESET if the socket is no longer connected */
1137         if (sk->sk_state != IUCV_CONNECTED) {
1138                 err = -ECONNRESET;
1139                 goto fail;
1140         }
1141
1142         /* increment and save iucv message tag for msg_completion cbk */
1143         txmsg.tag = iucv->send_tag++;
1144         memcpy(CB_TAG(skb), &txmsg.tag, CB_TAG_LEN);
1145
1146         if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1147                 atomic_inc(&iucv->msg_sent);
1148                 err = afiucv_hs_send(&txmsg, sk, skb, 0);
1149                 if (err) {
1150                         atomic_dec(&iucv->msg_sent);
1151                         goto fail;
1152                 }
1153                 goto release;
1154         }
1155         skb_queue_tail(&iucv->send_skb_q, skb);
1156
1157         if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags)
1158               && skb->len <= 7) {
1159                 err = iucv_send_iprm(iucv->path, &txmsg, skb);
1160
1161                 /* on success: there is no message_complete callback
1162                  * for an IPRMDATA msg; remove skb from send queue */
1163                 if (err == 0) {
1164                         skb_unlink(skb, &iucv->send_skb_q);
1165                         kfree_skb(skb);
1166                 }
1167
1168                 /* this error should never happen since the
1169                  * IUCV_IPRMDATA path flag is set... sever path */
1170                 if (err == 0x15) {
1171                         pr_iucv->path_sever(iucv->path, NULL);
1172                         skb_unlink(skb, &iucv->send_skb_q);
1173                         err = -EPIPE;
1174                         goto fail;
1175                 }
1176         } else
1177                 err = pr_iucv->message_send(iucv->path, &txmsg, 0, 0,
1178                                         (void *) skb->data, skb->len);
1179         if (err) {
1180                 if (err == 3) {
1181                         user_id[8] = 0;
1182                         memcpy(user_id, iucv->dst_user_id, 8);
1183                         appl_id[8] = 0;
1184                         memcpy(appl_id, iucv->dst_name, 8);
1185                         pr_err("Application %s on z/VM guest %s"
1186                                 " exceeds message limit\n",
1187                                 appl_id, user_id);
1188                         err = -EAGAIN;
1189                 } else
1190                         err = -EPIPE;
1191                 skb_unlink(skb, &iucv->send_skb_q);
1192                 goto fail;
1193         }
1194
1195 release:
1196         release_sock(sk);
1197         return len;
1198
1199 fail:
1200         kfree_skb(skb);
1201 out:
1202         release_sock(sk);
1203         return err;
1204 }
1205
1206 /* iucv_fragment_skb() - Fragment a single IUCV message into multiple skb's
1207  *
1208  * Locking: must be called with message_q.lock held
1209  */
1210 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
1211 {
1212         int dataleft, size, copied = 0;
1213         struct sk_buff *nskb;
1214
1215         dataleft = len;
1216         while (dataleft) {
1217                 if (dataleft >= sk->sk_rcvbuf / 4)
1218                         size = sk->sk_rcvbuf / 4;
1219                 else
1220                         size = dataleft;
1221
1222                 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
1223                 if (!nskb)
1224                         return -ENOMEM;
1225
1226                 /* copy target class to control buffer of new skb */
1227                 memcpy(CB_TRGCLS(nskb), CB_TRGCLS(skb), CB_TRGCLS_LEN);
1228
1229                 /* copy data fragment */
1230                 memcpy(nskb->data, skb->data + copied, size);
1231                 copied += size;
1232                 dataleft -= size;
1233
1234                 skb_reset_transport_header(nskb);
1235                 skb_reset_network_header(nskb);
1236                 nskb->len = size;
1237
1238                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
1239         }
1240
1241         return 0;
1242 }
1243
1244 /* iucv_process_message() - Receive a single outstanding IUCV message
1245  *
1246  * Locking: must be called with message_q.lock held
1247  */
1248 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1249                                  struct iucv_path *path,
1250                                  struct iucv_message *msg)
1251 {
1252         int rc;
1253         unsigned int len;
1254
1255         len = iucv_msg_length(msg);
1256
1257         /* store msg target class in the second 4 bytes of skb ctrl buffer */
1258         /* Note: the first 4 bytes are reserved for msg tag */
1259         memcpy(CB_TRGCLS(skb), &msg->class, CB_TRGCLS_LEN);
1260
1261         /* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1262         if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1263                 if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1264                         skb->data = NULL;
1265                         skb->len = 0;
1266                 }
1267         } else {
1268                 rc = pr_iucv->message_receive(path, msg,
1269                                               msg->flags & IUCV_IPRMDATA,
1270                                               skb->data, len, NULL);
1271                 if (rc) {
1272                         kfree_skb(skb);
1273                         return;
1274                 }
1275                 /* we need to fragment iucv messages for SOCK_STREAM only;
1276                  * for SOCK_SEQPACKET, it is only relevant if we support
1277                  * record segmentation using MSG_EOR (see also recvmsg()) */
1278                 if (sk->sk_type == SOCK_STREAM &&
1279                     skb->truesize >= sk->sk_rcvbuf / 4) {
1280                         rc = iucv_fragment_skb(sk, skb, len);
1281                         kfree_skb(skb);
1282                         skb = NULL;
1283                         if (rc) {
1284                                 pr_iucv->path_sever(path, NULL);
1285                                 return;
1286                         }
1287                         skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
1288                 } else {
1289                         skb_reset_transport_header(skb);
1290                         skb_reset_network_header(skb);
1291                         skb->len = len;
1292                 }
1293         }
1294
1295         if (sock_queue_rcv_skb(sk, skb))
1296                 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
1297 }
1298
1299 /* iucv_process_message_q() - Process outstanding IUCV messages
1300  *
1301  * Locking: must be called with message_q.lock held
1302  */
1303 static void iucv_process_message_q(struct sock *sk)
1304 {
1305         struct iucv_sock *iucv = iucv_sk(sk);
1306         struct sk_buff *skb;
1307         struct sock_msg_q *p, *n;
1308
1309         list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1310                 skb = alloc_skb(iucv_msg_length(&p->msg), GFP_ATOMIC | GFP_DMA);
1311                 if (!skb)
1312                         break;
1313                 iucv_process_message(sk, skb, p->path, &p->msg);
1314                 list_del(&p->list);
1315                 kfree(p);
1316                 if (!skb_queue_empty(&iucv->backlog_skb_q))
1317                         break;
1318         }
1319 }
1320
1321 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
1322                              struct msghdr *msg, size_t len, int flags)
1323 {
1324         int noblock = flags & MSG_DONTWAIT;
1325         struct sock *sk = sock->sk;
1326         struct iucv_sock *iucv = iucv_sk(sk);
1327         unsigned int copied, rlen;
1328         struct sk_buff *skb, *rskb, *cskb;
1329         int err = 0;
1330
1331         if ((sk->sk_state == IUCV_DISCONN) &&
1332             skb_queue_empty(&iucv->backlog_skb_q) &&
1333             skb_queue_empty(&sk->sk_receive_queue) &&
1334             list_empty(&iucv->message_q.list))
1335                 return 0;
1336
1337         if (flags & (MSG_OOB))
1338                 return -EOPNOTSUPP;
1339
1340         /* receive/dequeue next skb:
1341          * the function understands MSG_PEEK and, thus, does not dequeue skb */
1342         skb = skb_recv_datagram(sk, flags, noblock, &err);
1343         if (!skb) {
1344                 if (sk->sk_shutdown & RCV_SHUTDOWN)
1345                         return 0;
1346                 return err;
1347         }
1348
1349         rlen   = skb->len;              /* real length of skb */
1350         copied = min_t(unsigned int, rlen, len);
1351         if (!rlen)
1352                 sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1353
1354         cskb = skb;
1355         if (skb_copy_datagram_iovec(cskb, 0, msg->msg_iov, copied)) {
1356                 if (!(flags & MSG_PEEK))
1357                         skb_queue_head(&sk->sk_receive_queue, skb);
1358                 return -EFAULT;
1359         }
1360
1361         /* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1362         if (sk->sk_type == SOCK_SEQPACKET) {
1363                 if (copied < rlen)
1364                         msg->msg_flags |= MSG_TRUNC;
1365                 /* each iucv message contains a complete record */
1366                 msg->msg_flags |= MSG_EOR;
1367         }
1368
1369         /* create control message to store iucv msg target class:
1370          * get the trgcls from the control buffer of the skb due to
1371          * fragmentation of original iucv message. */
1372         err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1373                         CB_TRGCLS_LEN, CB_TRGCLS(skb));
1374         if (err) {
1375                 if (!(flags & MSG_PEEK))
1376                         skb_queue_head(&sk->sk_receive_queue, skb);
1377                 return err;
1378         }
1379
1380         /* Mark read part of skb as used */
1381         if (!(flags & MSG_PEEK)) {
1382
1383                 /* SOCK_STREAM: re-queue skb if it contains unreceived data */
1384                 if (sk->sk_type == SOCK_STREAM) {
1385                         skb_pull(skb, copied);
1386                         if (skb->len) {
1387                                 skb_queue_head(&sk->sk_receive_queue, skb);
1388                                 goto done;
1389                         }
1390                 }
1391
1392                 kfree_skb(skb);
1393                 if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1394                         atomic_inc(&iucv->msg_recv);
1395                         if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1396                                 WARN_ON(1);
1397                                 iucv_sock_close(sk);
1398                                 return -EFAULT;
1399                         }
1400                 }
1401
1402                 /* Queue backlog skbs */
1403                 spin_lock_bh(&iucv->message_q.lock);
1404                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1405                 while (rskb) {
1406                         if (sock_queue_rcv_skb(sk, rskb)) {
1407                                 skb_queue_head(&iucv->backlog_skb_q,
1408                                                 rskb);
1409                                 break;
1410                         } else {
1411                                 rskb = skb_dequeue(&iucv->backlog_skb_q);
1412                         }
1413                 }
1414                 if (skb_queue_empty(&iucv->backlog_skb_q)) {
1415                         if (!list_empty(&iucv->message_q.list))
1416                                 iucv_process_message_q(sk);
1417                         if (atomic_read(&iucv->msg_recv) >=
1418                                                         iucv->msglimit / 2) {
1419                                 err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1420                                 if (err) {
1421                                         sk->sk_state = IUCV_DISCONN;
1422                                         sk->sk_state_change(sk);
1423                                 }
1424                         }
1425                 }
1426                 spin_unlock_bh(&iucv->message_q.lock);
1427         }
1428
1429 done:
1430         /* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1431         if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1432                 copied = rlen;
1433
1434         return copied;
1435 }
1436
1437 static inline unsigned int iucv_accept_poll(struct sock *parent)
1438 {
1439         struct iucv_sock *isk, *n;
1440         struct sock *sk;
1441
1442         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1443                 sk = (struct sock *) isk;
1444
1445                 if (sk->sk_state == IUCV_CONNECTED)
1446                         return POLLIN | POLLRDNORM;
1447         }
1448
1449         return 0;
1450 }
1451
1452 unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
1453                             poll_table *wait)
1454 {
1455         struct sock *sk = sock->sk;
1456         unsigned int mask = 0;
1457
1458         sock_poll_wait(file, sk_sleep(sk), wait);
1459
1460         if (sk->sk_state == IUCV_LISTEN)
1461                 return iucv_accept_poll(sk);
1462
1463         if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1464                 mask |= POLLERR |
1465                         sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0;
1466
1467         if (sk->sk_shutdown & RCV_SHUTDOWN)
1468                 mask |= POLLRDHUP;
1469
1470         if (sk->sk_shutdown == SHUTDOWN_MASK)
1471                 mask |= POLLHUP;
1472
1473         if (!skb_queue_empty(&sk->sk_receive_queue) ||
1474             (sk->sk_shutdown & RCV_SHUTDOWN))
1475                 mask |= POLLIN | POLLRDNORM;
1476
1477         if (sk->sk_state == IUCV_CLOSED)
1478                 mask |= POLLHUP;
1479
1480         if (sk->sk_state == IUCV_DISCONN)
1481                 mask |= POLLIN;
1482
1483         if (sock_writeable(sk) && iucv_below_msglim(sk))
1484                 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1485         else
1486                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1487
1488         return mask;
1489 }
1490
1491 static int iucv_sock_shutdown(struct socket *sock, int how)
1492 {
1493         struct sock *sk = sock->sk;
1494         struct iucv_sock *iucv = iucv_sk(sk);
1495         struct iucv_message txmsg;
1496         int err = 0;
1497
1498         how++;
1499
1500         if ((how & ~SHUTDOWN_MASK) || !how)
1501                 return -EINVAL;
1502
1503         lock_sock(sk);
1504         switch (sk->sk_state) {
1505         case IUCV_LISTEN:
1506         case IUCV_DISCONN:
1507         case IUCV_CLOSING:
1508         case IUCV_CLOSED:
1509                 err = -ENOTCONN;
1510                 goto fail;
1511         default:
1512                 break;
1513         }
1514
1515         if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
1516                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1517                         txmsg.class = 0;
1518                         txmsg.tag = 0;
1519                         err = pr_iucv->message_send(iucv->path, &txmsg,
1520                                 IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1521                         if (err) {
1522                                 switch (err) {
1523                                 case 1:
1524                                         err = -ENOTCONN;
1525                                         break;
1526                                 case 2:
1527                                         err = -ECONNRESET;
1528                                         break;
1529                                 default:
1530                                         err = -ENOTCONN;
1531                                         break;
1532                                 }
1533                         }
1534                 } else
1535                         iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1536         }
1537
1538         sk->sk_shutdown |= how;
1539         if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1540                 if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1541                         err = pr_iucv->path_quiesce(iucv->path, NULL);
1542                         if (err)
1543                                 err = -ENOTCONN;
1544 /*                      skb_queue_purge(&sk->sk_receive_queue); */
1545                 }
1546                 skb_queue_purge(&sk->sk_receive_queue);
1547         }
1548
1549         /* Wake up anyone sleeping in poll */
1550         sk->sk_state_change(sk);
1551
1552 fail:
1553         release_sock(sk);
1554         return err;
1555 }
1556
1557 static int iucv_sock_release(struct socket *sock)
1558 {
1559         struct sock *sk = sock->sk;
1560         int err = 0;
1561
1562         if (!sk)
1563                 return 0;
1564
1565         iucv_sock_close(sk);
1566
1567         sock_orphan(sk);
1568         iucv_sock_kill(sk);
1569         return err;
1570 }
1571
1572 /* getsockopt and setsockopt */
1573 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1574                                 char __user *optval, unsigned int optlen)
1575 {
1576         struct sock *sk = sock->sk;
1577         struct iucv_sock *iucv = iucv_sk(sk);
1578         int val;
1579         int rc;
1580
1581         if (level != SOL_IUCV)
1582                 return -ENOPROTOOPT;
1583
1584         if (optlen < sizeof(int))
1585                 return -EINVAL;
1586
1587         if (get_user(val, (int __user *) optval))
1588                 return -EFAULT;
1589
1590         rc = 0;
1591
1592         lock_sock(sk);
1593         switch (optname) {
1594         case SO_IPRMDATA_MSG:
1595                 if (val)
1596                         iucv->flags |= IUCV_IPRMDATA;
1597                 else
1598                         iucv->flags &= ~IUCV_IPRMDATA;
1599                 break;
1600         case SO_MSGLIMIT:
1601                 switch (sk->sk_state) {
1602                 case IUCV_OPEN:
1603                 case IUCV_BOUND:
1604                         if (val < 1 || val > (u16)(~0))
1605                                 rc = -EINVAL;
1606                         else
1607                                 iucv->msglimit = val;
1608                         break;
1609                 default:
1610                         rc = -EINVAL;
1611                         break;
1612                 }
1613                 break;
1614         default:
1615                 rc = -ENOPROTOOPT;
1616                 break;
1617         }
1618         release_sock(sk);
1619
1620         return rc;
1621 }
1622
1623 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1624                                 char __user *optval, int __user *optlen)
1625 {
1626         struct sock *sk = sock->sk;
1627         struct iucv_sock *iucv = iucv_sk(sk);
1628         unsigned int val;
1629         int len;
1630
1631         if (level != SOL_IUCV)
1632                 return -ENOPROTOOPT;
1633
1634         if (get_user(len, optlen))
1635                 return -EFAULT;
1636
1637         if (len < 0)
1638                 return -EINVAL;
1639
1640         len = min_t(unsigned int, len, sizeof(int));
1641
1642         switch (optname) {
1643         case SO_IPRMDATA_MSG:
1644                 val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1645                 break;
1646         case SO_MSGLIMIT:
1647                 lock_sock(sk);
1648                 val = (iucv->path != NULL) ? iucv->path->msglim /* connected */
1649                                            : iucv->msglimit;    /* default */
1650                 release_sock(sk);
1651                 break;
1652         case SO_MSGSIZE:
1653                 if (sk->sk_state == IUCV_OPEN)
1654                         return -EBADFD;
1655                 val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1656                                 sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1657                                 0x7fffffff;
1658                 break;
1659         default:
1660                 return -ENOPROTOOPT;
1661         }
1662
1663         if (put_user(len, optlen))
1664                 return -EFAULT;
1665         if (copy_to_user(optval, &val, len))
1666                 return -EFAULT;
1667
1668         return 0;
1669 }
1670
1671
1672 /* Callback wrappers - called from iucv base support */
1673 static int iucv_callback_connreq(struct iucv_path *path,
1674                                  u8 ipvmid[8], u8 ipuser[16])
1675 {
1676         unsigned char user_data[16];
1677         unsigned char nuser_data[16];
1678         unsigned char src_name[8];
1679         struct sock *sk, *nsk;
1680         struct iucv_sock *iucv, *niucv;
1681         int err;
1682
1683         memcpy(src_name, ipuser, 8);
1684         EBCASC(src_name, 8);
1685         /* Find out if this path belongs to af_iucv. */
1686         read_lock(&iucv_sk_list.lock);
1687         iucv = NULL;
1688         sk = NULL;
1689         sk_for_each(sk, &iucv_sk_list.head)
1690                 if (sk->sk_state == IUCV_LISTEN &&
1691                     !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1692                         /*
1693                          * Found a listening socket with
1694                          * src_name == ipuser[0-7].
1695                          */
1696                         iucv = iucv_sk(sk);
1697                         break;
1698                 }
1699         read_unlock(&iucv_sk_list.lock);
1700         if (!iucv)
1701                 /* No socket found, not one of our paths. */
1702                 return -EINVAL;
1703
1704         bh_lock_sock(sk);
1705
1706         /* Check if parent socket is listening */
1707         low_nmcpy(user_data, iucv->src_name);
1708         high_nmcpy(user_data, iucv->dst_name);
1709         ASCEBC(user_data, sizeof(user_data));
1710         if (sk->sk_state != IUCV_LISTEN) {
1711                 err = pr_iucv->path_sever(path, user_data);
1712                 iucv_path_free(path);
1713                 goto fail;
1714         }
1715
1716         /* Check for backlog size */
1717         if (sk_acceptq_is_full(sk)) {
1718                 err = pr_iucv->path_sever(path, user_data);
1719                 iucv_path_free(path);
1720                 goto fail;
1721         }
1722
1723         /* Create the new socket */
1724         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1725         if (!nsk) {
1726                 err = pr_iucv->path_sever(path, user_data);
1727                 iucv_path_free(path);
1728                 goto fail;
1729         }
1730
1731         niucv = iucv_sk(nsk);
1732         iucv_sock_init(nsk, sk);
1733
1734         /* Set the new iucv_sock */
1735         memcpy(niucv->dst_name, ipuser + 8, 8);
1736         EBCASC(niucv->dst_name, 8);
1737         memcpy(niucv->dst_user_id, ipvmid, 8);
1738         memcpy(niucv->src_name, iucv->src_name, 8);
1739         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1740         niucv->path = path;
1741
1742         /* Call iucv_accept */
1743         high_nmcpy(nuser_data, ipuser + 8);
1744         memcpy(nuser_data + 8, niucv->src_name, 8);
1745         ASCEBC(nuser_data + 8, 8);
1746
1747         /* set message limit for path based on msglimit of accepting socket */
1748         niucv->msglimit = iucv->msglimit;
1749         path->msglim = iucv->msglimit;
1750         err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1751         if (err) {
1752                 iucv_sever_path(nsk, 1);
1753                 iucv_sock_kill(nsk);
1754                 goto fail;
1755         }
1756
1757         iucv_accept_enqueue(sk, nsk);
1758
1759         /* Wake up accept */
1760         nsk->sk_state = IUCV_CONNECTED;
1761         sk->sk_data_ready(sk, 1);
1762         err = 0;
1763 fail:
1764         bh_unlock_sock(sk);
1765         return 0;
1766 }
1767
1768 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1769 {
1770         struct sock *sk = path->private;
1771
1772         sk->sk_state = IUCV_CONNECTED;
1773         sk->sk_state_change(sk);
1774 }
1775
1776 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1777 {
1778         struct sock *sk = path->private;
1779         struct iucv_sock *iucv = iucv_sk(sk);
1780         struct sk_buff *skb;
1781         struct sock_msg_q *save_msg;
1782         int len;
1783
1784         if (sk->sk_shutdown & RCV_SHUTDOWN) {
1785                 pr_iucv->message_reject(path, msg);
1786                 return;
1787         }
1788
1789         spin_lock(&iucv->message_q.lock);
1790
1791         if (!list_empty(&iucv->message_q.list) ||
1792             !skb_queue_empty(&iucv->backlog_skb_q))
1793                 goto save_message;
1794
1795         len = atomic_read(&sk->sk_rmem_alloc);
1796         len += SKB_TRUESIZE(iucv_msg_length(msg));
1797         if (len > sk->sk_rcvbuf)
1798                 goto save_message;
1799
1800         skb = alloc_skb(iucv_msg_length(msg), GFP_ATOMIC | GFP_DMA);
1801         if (!skb)
1802                 goto save_message;
1803
1804         iucv_process_message(sk, skb, path, msg);
1805         goto out_unlock;
1806
1807 save_message:
1808         save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1809         if (!save_msg)
1810                 goto out_unlock;
1811         save_msg->path = path;
1812         save_msg->msg = *msg;
1813
1814         list_add_tail(&save_msg->list, &iucv->message_q.list);
1815
1816 out_unlock:
1817         spin_unlock(&iucv->message_q.lock);
1818 }
1819
1820 static void iucv_callback_txdone(struct iucv_path *path,
1821                                  struct iucv_message *msg)
1822 {
1823         struct sock *sk = path->private;
1824         struct sk_buff *this = NULL;
1825         struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1826         struct sk_buff *list_skb = list->next;
1827         unsigned long flags;
1828
1829         bh_lock_sock(sk);
1830         if (!skb_queue_empty(list)) {
1831                 spin_lock_irqsave(&list->lock, flags);
1832
1833                 while (list_skb != (struct sk_buff *)list) {
1834                         if (!memcmp(&msg->tag, CB_TAG(list_skb), CB_TAG_LEN)) {
1835                                 this = list_skb;
1836                                 break;
1837                         }
1838                         list_skb = list_skb->next;
1839                 }
1840                 if (this)
1841                         __skb_unlink(this, list);
1842
1843                 spin_unlock_irqrestore(&list->lock, flags);
1844
1845                 if (this) {
1846                         kfree_skb(this);
1847                         /* wake up any process waiting for sending */
1848                         iucv_sock_wake_msglim(sk);
1849                 }
1850         }
1851
1852         if (sk->sk_state == IUCV_CLOSING) {
1853                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1854                         sk->sk_state = IUCV_CLOSED;
1855                         sk->sk_state_change(sk);
1856                 }
1857         }
1858         bh_unlock_sock(sk);
1859
1860 }
1861
1862 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1863 {
1864         struct sock *sk = path->private;
1865
1866         if (sk->sk_state == IUCV_CLOSED)
1867                 return;
1868
1869         bh_lock_sock(sk);
1870         iucv_sever_path(sk, 1);
1871         sk->sk_state = IUCV_DISCONN;
1872
1873         sk->sk_state_change(sk);
1874         bh_unlock_sock(sk);
1875 }
1876
1877 /* called if the other communication side shuts down its RECV direction;
1878  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1879  */
1880 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1881 {
1882         struct sock *sk = path->private;
1883
1884         bh_lock_sock(sk);
1885         if (sk->sk_state != IUCV_CLOSED) {
1886                 sk->sk_shutdown |= SEND_SHUTDOWN;
1887                 sk->sk_state_change(sk);
1888         }
1889         bh_unlock_sock(sk);
1890 }
1891
1892 /***************** HiperSockets transport callbacks ********************/
1893 static void afiucv_swap_src_dest(struct sk_buff *skb)
1894 {
1895         struct af_iucv_trans_hdr *trans_hdr =
1896                                 (struct af_iucv_trans_hdr *)skb->data;
1897         char tmpID[8];
1898         char tmpName[8];
1899
1900         ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1901         ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1902         ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1903         ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1904         memcpy(tmpID, trans_hdr->srcUserID, 8);
1905         memcpy(tmpName, trans_hdr->srcAppName, 8);
1906         memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1907         memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1908         memcpy(trans_hdr->destUserID, tmpID, 8);
1909         memcpy(trans_hdr->destAppName, tmpName, 8);
1910         skb_push(skb, ETH_HLEN);
1911         memset(skb->data, 0, ETH_HLEN);
1912 }
1913
1914 /**
1915  * afiucv_hs_callback_syn - react on received SYN
1916  **/
1917 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1918 {
1919         struct sock *nsk;
1920         struct iucv_sock *iucv, *niucv;
1921         struct af_iucv_trans_hdr *trans_hdr;
1922         int err;
1923
1924         iucv = iucv_sk(sk);
1925         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
1926         if (!iucv) {
1927                 /* no sock - connection refused */
1928                 afiucv_swap_src_dest(skb);
1929                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1930                 err = dev_queue_xmit(skb);
1931                 goto out;
1932         }
1933
1934         nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1935         bh_lock_sock(sk);
1936         if ((sk->sk_state != IUCV_LISTEN) ||
1937             sk_acceptq_is_full(sk) ||
1938             !nsk) {
1939                 /* error on server socket - connection refused */
1940                 if (nsk)
1941                         sk_free(nsk);
1942                 afiucv_swap_src_dest(skb);
1943                 trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1944                 err = dev_queue_xmit(skb);
1945                 bh_unlock_sock(sk);
1946                 goto out;
1947         }
1948
1949         niucv = iucv_sk(nsk);
1950         iucv_sock_init(nsk, sk);
1951         niucv->transport = AF_IUCV_TRANS_HIPER;
1952         niucv->msglimit = iucv->msglimit;
1953         if (!trans_hdr->window)
1954                 niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1955         else
1956                 niucv->msglimit_peer = trans_hdr->window;
1957         memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1958         memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1959         memcpy(niucv->src_name, iucv->src_name, 8);
1960         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1961         nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1962         niucv->hs_dev = iucv->hs_dev;
1963         dev_hold(niucv->hs_dev);
1964         afiucv_swap_src_dest(skb);
1965         trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1966         trans_hdr->window = niucv->msglimit;
1967         /* if receiver acks the xmit connection is established */
1968         err = dev_queue_xmit(skb);
1969         if (!err) {
1970                 iucv_accept_enqueue(sk, nsk);
1971                 nsk->sk_state = IUCV_CONNECTED;
1972                 sk->sk_data_ready(sk, 1);
1973         } else
1974                 iucv_sock_kill(nsk);
1975         bh_unlock_sock(sk);
1976
1977 out:
1978         return NET_RX_SUCCESS;
1979 }
1980
1981 /**
1982  * afiucv_hs_callback_synack() - react on received SYN-ACK
1983  **/
1984 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1985 {
1986         struct iucv_sock *iucv = iucv_sk(sk);
1987         struct af_iucv_trans_hdr *trans_hdr =
1988                                         (struct af_iucv_trans_hdr *)skb->data;
1989
1990         if (!iucv)
1991                 goto out;
1992         if (sk->sk_state != IUCV_BOUND)
1993                 goto out;
1994         bh_lock_sock(sk);
1995         iucv->msglimit_peer = trans_hdr->window;
1996         sk->sk_state = IUCV_CONNECTED;
1997         sk->sk_state_change(sk);
1998         bh_unlock_sock(sk);
1999 out:
2000         kfree_skb(skb);
2001         return NET_RX_SUCCESS;
2002 }
2003
2004 /**
2005  * afiucv_hs_callback_synfin() - react on received SYN_FIN
2006  **/
2007 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
2008 {
2009         struct iucv_sock *iucv = iucv_sk(sk);
2010
2011         if (!iucv)
2012                 goto out;
2013         if (sk->sk_state != IUCV_BOUND)
2014                 goto out;
2015         bh_lock_sock(sk);
2016         sk->sk_state = IUCV_DISCONN;
2017         sk->sk_state_change(sk);
2018         bh_unlock_sock(sk);
2019 out:
2020         kfree_skb(skb);
2021         return NET_RX_SUCCESS;
2022 }
2023
2024 /**
2025  * afiucv_hs_callback_fin() - react on received FIN
2026  **/
2027 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
2028 {
2029         struct iucv_sock *iucv = iucv_sk(sk);
2030
2031         /* other end of connection closed */
2032         if (!iucv)
2033                 goto out;
2034         bh_lock_sock(sk);
2035         if (sk->sk_state == IUCV_CONNECTED) {
2036                 sk->sk_state = IUCV_DISCONN;
2037                 sk->sk_state_change(sk);
2038         }
2039         bh_unlock_sock(sk);
2040 out:
2041         kfree_skb(skb);
2042         return NET_RX_SUCCESS;
2043 }
2044
2045 /**
2046  * afiucv_hs_callback_win() - react on received WIN
2047  **/
2048 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
2049 {
2050         struct iucv_sock *iucv = iucv_sk(sk);
2051         struct af_iucv_trans_hdr *trans_hdr =
2052                                         (struct af_iucv_trans_hdr *)skb->data;
2053
2054         if (!iucv)
2055                 return NET_RX_SUCCESS;
2056
2057         if (sk->sk_state != IUCV_CONNECTED)
2058                 return NET_RX_SUCCESS;
2059
2060         atomic_sub(trans_hdr->window, &iucv->msg_sent);
2061         iucv_sock_wake_msglim(sk);
2062         return NET_RX_SUCCESS;
2063 }
2064
2065 /**
2066  * afiucv_hs_callback_rx() - react on received data
2067  **/
2068 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
2069 {
2070         struct iucv_sock *iucv = iucv_sk(sk);
2071
2072         if (!iucv) {
2073                 kfree_skb(skb);
2074                 return NET_RX_SUCCESS;
2075         }
2076
2077         if (sk->sk_state != IUCV_CONNECTED) {
2078                 kfree_skb(skb);
2079                 return NET_RX_SUCCESS;
2080         }
2081
2082         if (sk->sk_shutdown & RCV_SHUTDOWN) {
2083                 kfree_skb(skb);
2084                 return NET_RX_SUCCESS;
2085         }
2086
2087                 /* write stuff from iucv_msg to skb cb */
2088         if (skb->len < sizeof(struct af_iucv_trans_hdr)) {
2089                 kfree_skb(skb);
2090                 return NET_RX_SUCCESS;
2091         }
2092         skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2093         skb_reset_transport_header(skb);
2094         skb_reset_network_header(skb);
2095         spin_lock(&iucv->message_q.lock);
2096         if (skb_queue_empty(&iucv->backlog_skb_q)) {
2097                 if (sock_queue_rcv_skb(sk, skb)) {
2098                         /* handle rcv queue full */
2099                         skb_queue_tail(&iucv->backlog_skb_q, skb);
2100                 }
2101         } else
2102                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2103         spin_unlock(&iucv->message_q.lock);
2104         return NET_RX_SUCCESS;
2105 }
2106
2107 /**
2108  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2109  *                   transport
2110  *                   called from netif RX softirq
2111  **/
2112 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2113         struct packet_type *pt, struct net_device *orig_dev)
2114 {
2115         struct sock *sk;
2116         struct iucv_sock *iucv;
2117         struct af_iucv_trans_hdr *trans_hdr;
2118         char nullstring[8];
2119         int err = 0;
2120
2121         skb_pull(skb, ETH_HLEN);
2122         trans_hdr = (struct af_iucv_trans_hdr *)skb->data;
2123         EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2124         EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2125         EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2126         EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2127         memset(nullstring, 0, sizeof(nullstring));
2128         iucv = NULL;
2129         sk = NULL;
2130         read_lock(&iucv_sk_list.lock);
2131         sk_for_each(sk, &iucv_sk_list.head) {
2132                 if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2133                         if ((!memcmp(&iucv_sk(sk)->src_name,
2134                                      trans_hdr->destAppName, 8)) &&
2135                             (!memcmp(&iucv_sk(sk)->src_user_id,
2136                                      trans_hdr->destUserID, 8)) &&
2137                             (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2138                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2139                                      nullstring, 8))) {
2140                                 iucv = iucv_sk(sk);
2141                                 break;
2142                         }
2143                 } else {
2144                         if ((!memcmp(&iucv_sk(sk)->src_name,
2145                                      trans_hdr->destAppName, 8)) &&
2146                             (!memcmp(&iucv_sk(sk)->src_user_id,
2147                                      trans_hdr->destUserID, 8)) &&
2148                             (!memcmp(&iucv_sk(sk)->dst_name,
2149                                      trans_hdr->srcAppName, 8)) &&
2150                             (!memcmp(&iucv_sk(sk)->dst_user_id,
2151                                      trans_hdr->srcUserID, 8))) {
2152                                 iucv = iucv_sk(sk);
2153                                 break;
2154                         }
2155                 }
2156         }
2157         read_unlock(&iucv_sk_list.lock);
2158         if (!iucv)
2159                 sk = NULL;
2160
2161         /* no sock
2162         how should we send with no sock
2163         1) send without sock no send rc checking?
2164         2) introduce default sock to handle this cases
2165
2166          SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2167          data -> send FIN
2168          SYN|ACK, SYN|FIN, FIN -> no action? */
2169
2170         switch (trans_hdr->flags) {
2171         case AF_IUCV_FLAG_SYN:
2172                 /* connect request */
2173                 err = afiucv_hs_callback_syn(sk, skb);
2174                 break;
2175         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2176                 /* connect request confirmed */
2177                 err = afiucv_hs_callback_synack(sk, skb);
2178                 break;
2179         case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2180                 /* connect request refused */
2181                 err = afiucv_hs_callback_synfin(sk, skb);
2182                 break;
2183         case (AF_IUCV_FLAG_FIN):
2184                 /* close request */
2185                 err = afiucv_hs_callback_fin(sk, skb);
2186                 break;
2187         case (AF_IUCV_FLAG_WIN):
2188                 err = afiucv_hs_callback_win(sk, skb);
2189                 if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2190                         kfree_skb(skb);
2191                         break;
2192                 }
2193                 /* fall through and receive non-zero length data */
2194         case (AF_IUCV_FLAG_SHT):
2195                 /* shutdown request */
2196                 /* fall through and receive zero length data */
2197         case 0:
2198                 /* plain data frame */
2199                 memcpy(CB_TRGCLS(skb), &trans_hdr->iucv_hdr.class,
2200                        CB_TRGCLS_LEN);
2201                 err = afiucv_hs_callback_rx(sk, skb);
2202                 break;
2203         default:
2204                 ;
2205         }
2206
2207         return err;
2208 }
2209
2210 /**
2211  * afiucv_hs_callback_txnotify() - handle send notifcations from HiperSockets
2212  *                                 transport
2213  **/
2214 static void afiucv_hs_callback_txnotify(struct sk_buff *skb,
2215                                         enum iucv_tx_notify n)
2216 {
2217         struct sock *isk = skb->sk;
2218         struct sock *sk = NULL;
2219         struct iucv_sock *iucv = NULL;
2220         struct sk_buff_head *list;
2221         struct sk_buff *list_skb;
2222         struct sk_buff *nskb;
2223         unsigned long flags;
2224
2225         read_lock_irqsave(&iucv_sk_list.lock, flags);
2226         sk_for_each(sk, &iucv_sk_list.head)
2227                 if (sk == isk) {
2228                         iucv = iucv_sk(sk);
2229                         break;
2230                 }
2231         read_unlock_irqrestore(&iucv_sk_list.lock, flags);
2232
2233         if (!iucv || sock_flag(sk, SOCK_ZAPPED))
2234                 return;
2235
2236         list = &iucv->send_skb_q;
2237         spin_lock_irqsave(&list->lock, flags);
2238         if (skb_queue_empty(list))
2239                 goto out_unlock;
2240         list_skb = list->next;
2241         nskb = list_skb->next;
2242         while (list_skb != (struct sk_buff *)list) {
2243                 if (skb_shinfo(list_skb) == skb_shinfo(skb)) {
2244                         switch (n) {
2245                         case TX_NOTIFY_OK:
2246                                 __skb_unlink(list_skb, list);
2247                                 kfree_skb(list_skb);
2248                                 iucv_sock_wake_msglim(sk);
2249                                 break;
2250                         case TX_NOTIFY_PENDING:
2251                                 atomic_inc(&iucv->pendings);
2252                                 break;
2253                         case TX_NOTIFY_DELAYED_OK:
2254                                 __skb_unlink(list_skb, list);
2255                                 atomic_dec(&iucv->pendings);
2256                                 if (atomic_read(&iucv->pendings) <= 0)
2257                                         iucv_sock_wake_msglim(sk);
2258                                 kfree_skb(list_skb);
2259                                 break;
2260                         case TX_NOTIFY_UNREACHABLE:
2261                         case TX_NOTIFY_DELAYED_UNREACHABLE:
2262                         case TX_NOTIFY_TPQFULL: /* not yet used */
2263                         case TX_NOTIFY_GENERALERROR:
2264                         case TX_NOTIFY_DELAYED_GENERALERROR:
2265                                 __skb_unlink(list_skb, list);
2266                                 kfree_skb(list_skb);
2267                                 if (sk->sk_state == IUCV_CONNECTED) {
2268                                         sk->sk_state = IUCV_DISCONN;
2269                                         sk->sk_state_change(sk);
2270                                 }
2271                                 break;
2272                         }
2273                         break;
2274                 }
2275                 list_skb = nskb;
2276                 nskb = nskb->next;
2277         }
2278 out_unlock:
2279         spin_unlock_irqrestore(&list->lock, flags);
2280
2281         if (sk->sk_state == IUCV_CLOSING) {
2282                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
2283                         sk->sk_state = IUCV_CLOSED;
2284                         sk->sk_state_change(sk);
2285                 }
2286         }
2287
2288 }
2289
2290 /*
2291  * afiucv_netdev_event: handle netdev notifier chain events
2292  */
2293 static int afiucv_netdev_event(struct notifier_block *this,
2294                                unsigned long event, void *ptr)
2295 {
2296         struct net_device *event_dev = (struct net_device *)ptr;
2297         struct sock *sk;
2298         struct iucv_sock *iucv;
2299
2300         switch (event) {
2301         case NETDEV_REBOOT:
2302         case NETDEV_GOING_DOWN:
2303                 sk_for_each(sk, &iucv_sk_list.head) {
2304                         iucv = iucv_sk(sk);
2305                         if ((iucv->hs_dev == event_dev) &&
2306                             (sk->sk_state == IUCV_CONNECTED)) {
2307                                 if (event == NETDEV_GOING_DOWN)
2308                                         iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2309                                 sk->sk_state = IUCV_DISCONN;
2310                                 sk->sk_state_change(sk);
2311                         }
2312                 }
2313                 break;
2314         case NETDEV_DOWN:
2315         case NETDEV_UNREGISTER:
2316         default:
2317                 break;
2318         }
2319         return NOTIFY_DONE;
2320 }
2321
2322 static struct notifier_block afiucv_netdev_notifier = {
2323         .notifier_call = afiucv_netdev_event,
2324 };
2325
2326 static const struct proto_ops iucv_sock_ops = {
2327         .family         = PF_IUCV,
2328         .owner          = THIS_MODULE,
2329         .release        = iucv_sock_release,
2330         .bind           = iucv_sock_bind,
2331         .connect        = iucv_sock_connect,
2332         .listen         = iucv_sock_listen,
2333         .accept         = iucv_sock_accept,
2334         .getname        = iucv_sock_getname,
2335         .sendmsg        = iucv_sock_sendmsg,
2336         .recvmsg        = iucv_sock_recvmsg,
2337         .poll           = iucv_sock_poll,
2338         .ioctl          = sock_no_ioctl,
2339         .mmap           = sock_no_mmap,
2340         .socketpair     = sock_no_socketpair,
2341         .shutdown       = iucv_sock_shutdown,
2342         .setsockopt     = iucv_sock_setsockopt,
2343         .getsockopt     = iucv_sock_getsockopt,
2344 };
2345
2346 static const struct net_proto_family iucv_sock_family_ops = {
2347         .family = AF_IUCV,
2348         .owner  = THIS_MODULE,
2349         .create = iucv_sock_create,
2350 };
2351
2352 static struct packet_type iucv_packet_type = {
2353         .type = cpu_to_be16(ETH_P_AF_IUCV),
2354         .func = afiucv_hs_rcv,
2355 };
2356
2357 static int afiucv_iucv_init(void)
2358 {
2359         int err;
2360
2361         err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2362         if (err)
2363                 goto out;
2364         /* establish dummy device */
2365         af_iucv_driver.bus = pr_iucv->bus;
2366         err = driver_register(&af_iucv_driver);
2367         if (err)
2368                 goto out_iucv;
2369         af_iucv_dev = kzalloc(sizeof(struct device), GFP_KERNEL);
2370         if (!af_iucv_dev) {
2371                 err = -ENOMEM;
2372                 goto out_driver;
2373         }
2374         dev_set_name(af_iucv_dev, "af_iucv");
2375         af_iucv_dev->bus = pr_iucv->bus;
2376         af_iucv_dev->parent = pr_iucv->root;
2377         af_iucv_dev->release = (void (*)(struct device *))kfree;
2378         af_iucv_dev->driver = &af_iucv_driver;
2379         err = device_register(af_iucv_dev);
2380         if (err)
2381                 goto out_driver;
2382         return 0;
2383
2384 out_driver:
2385         driver_unregister(&af_iucv_driver);
2386 out_iucv:
2387         pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2388 out:
2389         return err;
2390 }
2391
2392 static int __init afiucv_init(void)
2393 {
2394         int err;
2395
2396         if (MACHINE_IS_VM) {
2397                 cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2398                 if (unlikely(err)) {
2399                         WARN_ON(err);
2400                         err = -EPROTONOSUPPORT;
2401                         goto out;
2402                 }
2403
2404                 pr_iucv = try_then_request_module(symbol_get(iucv_if), "iucv");
2405                 if (!pr_iucv) {
2406                         printk(KERN_WARNING "iucv_if lookup failed\n");
2407                         memset(&iucv_userid, 0, sizeof(iucv_userid));
2408                 }
2409         } else {
2410                 memset(&iucv_userid, 0, sizeof(iucv_userid));
2411                 pr_iucv = NULL;
2412         }
2413
2414         err = proto_register(&iucv_proto, 0);
2415         if (err)
2416                 goto out;
2417         err = sock_register(&iucv_sock_family_ops);
2418         if (err)
2419                 goto out_proto;
2420
2421         if (pr_iucv) {
2422                 err = afiucv_iucv_init();
2423                 if (err)
2424                         goto out_sock;
2425         } else
2426                 register_netdevice_notifier(&afiucv_netdev_notifier);
2427         dev_add_pack(&iucv_packet_type);
2428         return 0;
2429
2430 out_sock:
2431         sock_unregister(PF_IUCV);
2432 out_proto:
2433         proto_unregister(&iucv_proto);
2434 out:
2435         if (pr_iucv)
2436                 symbol_put(iucv_if);
2437         return err;
2438 }
2439
2440 static void __exit afiucv_exit(void)
2441 {
2442         if (pr_iucv) {
2443                 device_unregister(af_iucv_dev);
2444                 driver_unregister(&af_iucv_driver);
2445                 pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2446                 symbol_put(iucv_if);
2447         } else
2448                 unregister_netdevice_notifier(&afiucv_netdev_notifier);
2449         dev_remove_pack(&iucv_packet_type);
2450         sock_unregister(PF_IUCV);
2451         proto_unregister(&iucv_proto);
2452 }
2453
2454 module_init(afiucv_init);
2455 module_exit(afiucv_exit);
2456
2457 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2458 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2459 MODULE_VERSION(VERSION);
2460 MODULE_LICENSE("GPL");
2461 MODULE_ALIAS_NETPROTO(PF_IUCV);
2462