tcp: Do not use TSO/GSO when there is urgent data
[linux-2.6] / net / iucv / af_iucv.c
1 /*
2  *  linux/net/iucv/af_iucv.c
3  *
4  *  IUCV protocol stack for Linux on zSeries
5  *
6  *  Copyright 2006 IBM Corporation
7  *
8  *  Author(s):  Jennifer Hunt <jenhunt@us.ibm.com>
9  */
10
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/list.h>
14 #include <linux/errno.h>
15 #include <linux/kernel.h>
16 #include <linux/sched.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/init.h>
20 #include <linux/poll.h>
21 #include <net/sock.h>
22 #include <asm/ebcdic.h>
23 #include <asm/cpcmd.h>
24 #include <linux/kmod.h>
25
26 #include <net/iucv/iucv.h>
27 #include <net/iucv/af_iucv.h>
28
29 #define CONFIG_IUCV_SOCK_DEBUG 1
30
31 #define IPRMDATA 0x80
32 #define VERSION "1.0"
33
34 static char iucv_userid[80];
35
36 static 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 void iucv_sock_kill(struct sock *sk);
45 static void iucv_sock_close(struct sock *sk);
46
47 /* Call Back functions */
48 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
49 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
50 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
51 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
52                                  u8 ipuser[16]);
53 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
54
55 static struct iucv_sock_list iucv_sk_list = {
56         .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
57         .autobind_name = ATOMIC_INIT(0)
58 };
59
60 static struct iucv_handler af_iucv_handler = {
61         .path_pending     = iucv_callback_connreq,
62         .path_complete    = iucv_callback_connack,
63         .path_severed     = iucv_callback_connrej,
64         .message_pending  = iucv_callback_rx,
65         .message_complete = iucv_callback_txdone
66 };
67
68 static inline void high_nmcpy(unsigned char *dst, char *src)
69 {
70        memcpy(dst, src, 8);
71 }
72
73 static inline void low_nmcpy(unsigned char *dst, char *src)
74 {
75        memcpy(&dst[8], src, 8);
76 }
77
78 /* Timers */
79 static void iucv_sock_timeout(unsigned long arg)
80 {
81         struct sock *sk = (struct sock *)arg;
82
83         bh_lock_sock(sk);
84         sk->sk_err = ETIMEDOUT;
85         sk->sk_state_change(sk);
86         bh_unlock_sock(sk);
87
88         iucv_sock_kill(sk);
89         sock_put(sk);
90 }
91
92 static void iucv_sock_clear_timer(struct sock *sk)
93 {
94         sk_stop_timer(sk, &sk->sk_timer);
95 }
96
97 static struct sock *__iucv_get_sock_by_name(char *nm)
98 {
99         struct sock *sk;
100         struct hlist_node *node;
101
102         sk_for_each(sk, node, &iucv_sk_list.head)
103                 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
104                         return sk;
105
106         return NULL;
107 }
108
109 static void iucv_sock_destruct(struct sock *sk)
110 {
111         skb_queue_purge(&sk->sk_receive_queue);
112         skb_queue_purge(&sk->sk_write_queue);
113 }
114
115 /* Cleanup Listen */
116 static void iucv_sock_cleanup_listen(struct sock *parent)
117 {
118         struct sock *sk;
119
120         /* Close non-accepted connections */
121         while ((sk = iucv_accept_dequeue(parent, NULL))) {
122                 iucv_sock_close(sk);
123                 iucv_sock_kill(sk);
124         }
125
126         parent->sk_state = IUCV_CLOSED;
127         sock_set_flag(parent, SOCK_ZAPPED);
128 }
129
130 /* Kill socket */
131 static void iucv_sock_kill(struct sock *sk)
132 {
133         if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
134                 return;
135
136         iucv_sock_unlink(&iucv_sk_list, sk);
137         sock_set_flag(sk, SOCK_DEAD);
138         sock_put(sk);
139 }
140
141 /* Close an IUCV socket */
142 static void iucv_sock_close(struct sock *sk)
143 {
144         unsigned char user_data[16];
145         struct iucv_sock *iucv = iucv_sk(sk);
146         int err;
147         unsigned long timeo;
148
149         iucv_sock_clear_timer(sk);
150         lock_sock(sk);
151
152         switch (sk->sk_state) {
153         case IUCV_LISTEN:
154                 iucv_sock_cleanup_listen(sk);
155                 break;
156
157         case IUCV_CONNECTED:
158         case IUCV_DISCONN:
159                 err = 0;
160
161                 sk->sk_state = IUCV_CLOSING;
162                 sk->sk_state_change(sk);
163
164                 if (!skb_queue_empty(&iucv->send_skb_q)) {
165                         if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
166                                 timeo = sk->sk_lingertime;
167                         else
168                                 timeo = IUCV_DISCONN_TIMEOUT;
169                         err = iucv_sock_wait_state(sk, IUCV_CLOSED, 0, timeo);
170                 }
171
172                 sk->sk_state = IUCV_CLOSED;
173                 sk->sk_state_change(sk);
174
175                 if (iucv->path) {
176                         low_nmcpy(user_data, iucv->src_name);
177                         high_nmcpy(user_data, iucv->dst_name);
178                         ASCEBC(user_data, sizeof(user_data));
179                         err = iucv_path_sever(iucv->path, user_data);
180                         iucv_path_free(iucv->path);
181                         iucv->path = NULL;
182                 }
183
184                 sk->sk_err = ECONNRESET;
185                 sk->sk_state_change(sk);
186
187                 skb_queue_purge(&iucv->send_skb_q);
188                 skb_queue_purge(&iucv->backlog_skb_q);
189
190                 sock_set_flag(sk, SOCK_ZAPPED);
191                 break;
192
193         default:
194                 sock_set_flag(sk, SOCK_ZAPPED);
195                 break;
196         }
197
198         release_sock(sk);
199         iucv_sock_kill(sk);
200 }
201
202 static void iucv_sock_init(struct sock *sk, struct sock *parent)
203 {
204         if (parent)
205                 sk->sk_type = parent->sk_type;
206 }
207
208 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
209 {
210         struct sock *sk;
211
212         sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
213         if (!sk)
214                 return NULL;
215
216         sock_init_data(sock, sk);
217         INIT_LIST_HEAD(&iucv_sk(sk)->accept_q);
218         spin_lock_init(&iucv_sk(sk)->accept_q_lock);
219         skb_queue_head_init(&iucv_sk(sk)->send_skb_q);
220         INIT_LIST_HEAD(&iucv_sk(sk)->message_q.list);
221         spin_lock_init(&iucv_sk(sk)->message_q.lock);
222         skb_queue_head_init(&iucv_sk(sk)->backlog_skb_q);
223         iucv_sk(sk)->send_tag = 0;
224
225         sk->sk_destruct = iucv_sock_destruct;
226         sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
227         sk->sk_allocation = GFP_DMA;
228
229         sock_reset_flag(sk, SOCK_ZAPPED);
230
231         sk->sk_protocol = proto;
232         sk->sk_state    = IUCV_OPEN;
233
234         setup_timer(&sk->sk_timer, iucv_sock_timeout, (unsigned long)sk);
235
236         iucv_sock_link(&iucv_sk_list, sk);
237         return sk;
238 }
239
240 /* Create an IUCV socket */
241 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol)
242 {
243         struct sock *sk;
244
245         if (sock->type != SOCK_STREAM)
246                 return -ESOCKTNOSUPPORT;
247
248         sock->state = SS_UNCONNECTED;
249         sock->ops = &iucv_sock_ops;
250
251         sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
252         if (!sk)
253                 return -ENOMEM;
254
255         iucv_sock_init(sk, NULL);
256
257         return 0;
258 }
259
260 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
261 {
262         write_lock_bh(&l->lock);
263         sk_add_node(sk, &l->head);
264         write_unlock_bh(&l->lock);
265 }
266
267 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
268 {
269         write_lock_bh(&l->lock);
270         sk_del_node_init(sk);
271         write_unlock_bh(&l->lock);
272 }
273
274 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
275 {
276         unsigned long flags;
277         struct iucv_sock *par = iucv_sk(parent);
278
279         sock_hold(sk);
280         spin_lock_irqsave(&par->accept_q_lock, flags);
281         list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
282         spin_unlock_irqrestore(&par->accept_q_lock, flags);
283         iucv_sk(sk)->parent = parent;
284         parent->sk_ack_backlog++;
285 }
286
287 void iucv_accept_unlink(struct sock *sk)
288 {
289         unsigned long flags;
290         struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
291
292         spin_lock_irqsave(&par->accept_q_lock, flags);
293         list_del_init(&iucv_sk(sk)->accept_q);
294         spin_unlock_irqrestore(&par->accept_q_lock, flags);
295         iucv_sk(sk)->parent->sk_ack_backlog--;
296         iucv_sk(sk)->parent = NULL;
297         sock_put(sk);
298 }
299
300 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
301 {
302         struct iucv_sock *isk, *n;
303         struct sock *sk;
304
305         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
306                 sk = (struct sock *) isk;
307                 lock_sock(sk);
308
309                 if (sk->sk_state == IUCV_CLOSED) {
310                         iucv_accept_unlink(sk);
311                         release_sock(sk);
312                         continue;
313                 }
314
315                 if (sk->sk_state == IUCV_CONNECTED ||
316                     sk->sk_state == IUCV_SEVERED ||
317                     !newsock) {
318                         iucv_accept_unlink(sk);
319                         if (newsock)
320                                 sock_graft(sk, newsock);
321
322                         if (sk->sk_state == IUCV_SEVERED)
323                                 sk->sk_state = IUCV_DISCONN;
324
325                         release_sock(sk);
326                         return sk;
327                 }
328
329                 release_sock(sk);
330         }
331         return NULL;
332 }
333
334 int iucv_sock_wait_state(struct sock *sk, int state, int state2,
335                          unsigned long timeo)
336 {
337         DECLARE_WAITQUEUE(wait, current);
338         int err = 0;
339
340         add_wait_queue(sk->sk_sleep, &wait);
341         while (sk->sk_state != state && sk->sk_state != state2) {
342                 set_current_state(TASK_INTERRUPTIBLE);
343
344                 if (!timeo) {
345                         err = -EAGAIN;
346                         break;
347                 }
348
349                 if (signal_pending(current)) {
350                         err = sock_intr_errno(timeo);
351                         break;
352                 }
353
354                 release_sock(sk);
355                 timeo = schedule_timeout(timeo);
356                 lock_sock(sk);
357
358                 err = sock_error(sk);
359                 if (err)
360                         break;
361         }
362         set_current_state(TASK_RUNNING);
363         remove_wait_queue(sk->sk_sleep, &wait);
364         return err;
365 }
366
367 /* Bind an unbound socket */
368 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
369                           int addr_len)
370 {
371         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
372         struct sock *sk = sock->sk;
373         struct iucv_sock *iucv;
374         int err;
375
376         /* Verify the input sockaddr */
377         if (!addr || addr->sa_family != AF_IUCV)
378                 return -EINVAL;
379
380         lock_sock(sk);
381         if (sk->sk_state != IUCV_OPEN) {
382                 err = -EBADFD;
383                 goto done;
384         }
385
386         write_lock_bh(&iucv_sk_list.lock);
387
388         iucv = iucv_sk(sk);
389         if (__iucv_get_sock_by_name(sa->siucv_name)) {
390                 err = -EADDRINUSE;
391                 goto done_unlock;
392         }
393         if (iucv->path) {
394                 err = 0;
395                 goto done_unlock;
396         }
397
398         /* Bind the socket */
399         memcpy(iucv->src_name, sa->siucv_name, 8);
400
401         /* Copy the user id */
402         memcpy(iucv->src_user_id, iucv_userid, 8);
403         sk->sk_state = IUCV_BOUND;
404         err = 0;
405
406 done_unlock:
407         /* Release the socket list lock */
408         write_unlock_bh(&iucv_sk_list.lock);
409 done:
410         release_sock(sk);
411         return err;
412 }
413
414 /* Automatically bind an unbound socket */
415 static int iucv_sock_autobind(struct sock *sk)
416 {
417         struct iucv_sock *iucv = iucv_sk(sk);
418         char query_buffer[80];
419         char name[12];
420         int err = 0;
421
422         /* Set the userid and name */
423         cpcmd("QUERY USERID", query_buffer, sizeof(query_buffer), &err);
424         if (unlikely(err))
425                 return -EPROTO;
426
427         memcpy(iucv->src_user_id, query_buffer, 8);
428
429         write_lock_bh(&iucv_sk_list.lock);
430
431         sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
432         while (__iucv_get_sock_by_name(name)) {
433                 sprintf(name, "%08x",
434                         atomic_inc_return(&iucv_sk_list.autobind_name));
435         }
436
437         write_unlock_bh(&iucv_sk_list.lock);
438
439         memcpy(&iucv->src_name, name, 8);
440
441         return err;
442 }
443
444 /* Connect an unconnected socket */
445 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
446                              int alen, int flags)
447 {
448         struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
449         struct sock *sk = sock->sk;
450         struct iucv_sock *iucv;
451         unsigned char user_data[16];
452         int err;
453
454         if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
455                 return -EINVAL;
456
457         if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
458                 return -EBADFD;
459
460         if (sk->sk_type != SOCK_STREAM)
461                 return -EINVAL;
462
463         iucv = iucv_sk(sk);
464
465         if (sk->sk_state == IUCV_OPEN) {
466                 err = iucv_sock_autobind(sk);
467                 if (unlikely(err))
468                         return err;
469         }
470
471         lock_sock(sk);
472
473         /* Set the destination information */
474         memcpy(iucv_sk(sk)->dst_user_id, sa->siucv_user_id, 8);
475         memcpy(iucv_sk(sk)->dst_name, sa->siucv_name, 8);
476
477         high_nmcpy(user_data, sa->siucv_name);
478         low_nmcpy(user_data, iucv_sk(sk)->src_name);
479         ASCEBC(user_data, sizeof(user_data));
480
481         iucv = iucv_sk(sk);
482         /* Create path. */
483         iucv->path = iucv_path_alloc(IUCV_QUEUELEN_DEFAULT,
484                                      IPRMDATA, GFP_KERNEL);
485         if (!iucv->path) {
486                 err = -ENOMEM;
487                 goto done;
488         }
489         err = iucv_path_connect(iucv->path, &af_iucv_handler,
490                                 sa->siucv_user_id, NULL, user_data, sk);
491         if (err) {
492                 iucv_path_free(iucv->path);
493                 iucv->path = NULL;
494                 err = -ECONNREFUSED;
495                 goto done;
496         }
497
498         if (sk->sk_state != IUCV_CONNECTED) {
499                 err = iucv_sock_wait_state(sk, IUCV_CONNECTED, IUCV_DISCONN,
500                                 sock_sndtimeo(sk, flags & O_NONBLOCK));
501         }
502
503         if (sk->sk_state == IUCV_DISCONN) {
504                 release_sock(sk);
505                 return -ECONNREFUSED;
506         }
507 done:
508         release_sock(sk);
509         return err;
510 }
511
512 /* Move a socket into listening state. */
513 static int iucv_sock_listen(struct socket *sock, int backlog)
514 {
515         struct sock *sk = sock->sk;
516         int err;
517
518         lock_sock(sk);
519
520         err = -EINVAL;
521         if (sk->sk_state != IUCV_BOUND || sock->type != SOCK_STREAM)
522                 goto done;
523
524         sk->sk_max_ack_backlog = backlog;
525         sk->sk_ack_backlog = 0;
526         sk->sk_state = IUCV_LISTEN;
527         err = 0;
528
529 done:
530         release_sock(sk);
531         return err;
532 }
533
534 /* Accept a pending connection */
535 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
536                             int flags)
537 {
538         DECLARE_WAITQUEUE(wait, current);
539         struct sock *sk = sock->sk, *nsk;
540         long timeo;
541         int err = 0;
542
543         lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
544
545         if (sk->sk_state != IUCV_LISTEN) {
546                 err = -EBADFD;
547                 goto done;
548         }
549
550         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
551
552         /* Wait for an incoming connection */
553         add_wait_queue_exclusive(sk->sk_sleep, &wait);
554         while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
555                 set_current_state(TASK_INTERRUPTIBLE);
556                 if (!timeo) {
557                         err = -EAGAIN;
558                         break;
559                 }
560
561                 release_sock(sk);
562                 timeo = schedule_timeout(timeo);
563                 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
564
565                 if (sk->sk_state != IUCV_LISTEN) {
566                         err = -EBADFD;
567                         break;
568                 }
569
570                 if (signal_pending(current)) {
571                         err = sock_intr_errno(timeo);
572                         break;
573                 }
574         }
575
576         set_current_state(TASK_RUNNING);
577         remove_wait_queue(sk->sk_sleep, &wait);
578
579         if (err)
580                 goto done;
581
582         newsock->state = SS_CONNECTED;
583
584 done:
585         release_sock(sk);
586         return err;
587 }
588
589 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
590                              int *len, int peer)
591 {
592         struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
593         struct sock *sk = sock->sk;
594
595         addr->sa_family = AF_IUCV;
596         *len = sizeof(struct sockaddr_iucv);
597
598         if (peer) {
599                 memcpy(siucv->siucv_user_id, iucv_sk(sk)->dst_user_id, 8);
600                 memcpy(siucv->siucv_name, &iucv_sk(sk)->dst_name, 8);
601         } else {
602                 memcpy(siucv->siucv_user_id, iucv_sk(sk)->src_user_id, 8);
603                 memcpy(siucv->siucv_name, iucv_sk(sk)->src_name, 8);
604         }
605         memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
606         memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
607         memset(siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
608
609         return 0;
610 }
611
612 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
613                              struct msghdr *msg, size_t len)
614 {
615         struct sock *sk = sock->sk;
616         struct iucv_sock *iucv = iucv_sk(sk);
617         struct sk_buff *skb;
618         struct iucv_message txmsg;
619         int err;
620
621         err = sock_error(sk);
622         if (err)
623                 return err;
624
625         if (msg->msg_flags & MSG_OOB)
626                 return -EOPNOTSUPP;
627
628         lock_sock(sk);
629
630         if (sk->sk_shutdown & SEND_SHUTDOWN) {
631                 err = -EPIPE;
632                 goto out;
633         }
634
635         if (sk->sk_state == IUCV_CONNECTED) {
636                 if (!(skb = sock_alloc_send_skb(sk, len,
637                                                 msg->msg_flags & MSG_DONTWAIT,
638                                                 &err)))
639                         goto out;
640
641                 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
642                         err = -EFAULT;
643                         goto fail;
644                 }
645
646                 txmsg.class = 0;
647                 memcpy(&txmsg.class, skb->data, skb->len >= 4 ? 4 : skb->len);
648                 txmsg.tag = iucv->send_tag++;
649                 memcpy(skb->cb, &txmsg.tag, 4);
650                 skb_queue_tail(&iucv->send_skb_q, skb);
651                 err = iucv_message_send(iucv->path, &txmsg, 0, 0,
652                                         (void *) skb->data, skb->len);
653                 if (err) {
654                         if (err == 3)
655                                 printk(KERN_ERR "AF_IUCV msg limit exceeded\n");
656                         skb_unlink(skb, &iucv->send_skb_q);
657                         err = -EPIPE;
658                         goto fail;
659                 }
660
661         } else {
662                 err = -ENOTCONN;
663                 goto out;
664         }
665
666         release_sock(sk);
667         return len;
668
669 fail:
670         kfree_skb(skb);
671 out:
672         release_sock(sk);
673         return err;
674 }
675
676 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
677 {
678         int dataleft, size, copied = 0;
679         struct sk_buff *nskb;
680
681         dataleft = len;
682         while (dataleft) {
683                 if (dataleft >= sk->sk_rcvbuf / 4)
684                         size = sk->sk_rcvbuf / 4;
685                 else
686                         size = dataleft;
687
688                 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
689                 if (!nskb)
690                         return -ENOMEM;
691
692                 memcpy(nskb->data, skb->data + copied, size);
693                 copied += size;
694                 dataleft -= size;
695
696                 skb_reset_transport_header(nskb);
697                 skb_reset_network_header(nskb);
698                 nskb->len = size;
699
700                 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
701         }
702
703         return 0;
704 }
705
706 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
707                                  struct iucv_path *path,
708                                  struct iucv_message *msg)
709 {
710         int rc;
711
712         if (msg->flags & IPRMDATA) {
713                 skb->data = NULL;
714                 skb->len = 0;
715         } else {
716                 rc = iucv_message_receive(path, msg, 0, skb->data,
717                                           msg->length, NULL);
718                 if (rc) {
719                         kfree_skb(skb);
720                         return;
721                 }
722                 if (skb->truesize >= sk->sk_rcvbuf / 4) {
723                         rc = iucv_fragment_skb(sk, skb, msg->length);
724                         kfree_skb(skb);
725                         skb = NULL;
726                         if (rc) {
727                                 iucv_path_sever(path, NULL);
728                                 return;
729                         }
730                         skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
731                 } else {
732                         skb_reset_transport_header(skb);
733                         skb_reset_network_header(skb);
734                         skb->len = msg->length;
735                 }
736         }
737
738         if (sock_queue_rcv_skb(sk, skb))
739                 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
740 }
741
742 static void iucv_process_message_q(struct sock *sk)
743 {
744         struct iucv_sock *iucv = iucv_sk(sk);
745         struct sk_buff *skb;
746         struct sock_msg_q *p, *n;
747
748         list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
749                 skb = alloc_skb(p->msg.length, GFP_ATOMIC | GFP_DMA);
750                 if (!skb)
751                         break;
752                 iucv_process_message(sk, skb, p->path, &p->msg);
753                 list_del(&p->list);
754                 kfree(p);
755                 if (!skb_queue_empty(&iucv->backlog_skb_q))
756                         break;
757         }
758 }
759
760 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
761                              struct msghdr *msg, size_t len, int flags)
762 {
763         int noblock = flags & MSG_DONTWAIT;
764         struct sock *sk = sock->sk;
765         struct iucv_sock *iucv = iucv_sk(sk);
766         int target, copied = 0;
767         struct sk_buff *skb, *rskb, *cskb;
768         int err = 0;
769
770         if ((sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_SEVERED) &&
771             skb_queue_empty(&iucv->backlog_skb_q) &&
772             skb_queue_empty(&sk->sk_receive_queue) &&
773             list_empty(&iucv->message_q.list))
774                 return 0;
775
776         if (flags & (MSG_OOB))
777                 return -EOPNOTSUPP;
778
779         target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
780
781         skb = skb_recv_datagram(sk, flags, noblock, &err);
782         if (!skb) {
783                 if (sk->sk_shutdown & RCV_SHUTDOWN)
784                         return 0;
785                 return err;
786         }
787
788         copied = min_t(unsigned int, skb->len, len);
789
790         cskb = skb;
791         if (memcpy_toiovec(msg->msg_iov, cskb->data, copied)) {
792                 skb_queue_head(&sk->sk_receive_queue, skb);
793                 if (copied == 0)
794                         return -EFAULT;
795                 goto done;
796         }
797
798         len -= copied;
799
800         /* Mark read part of skb as used */
801         if (!(flags & MSG_PEEK)) {
802                 skb_pull(skb, copied);
803
804                 if (skb->len) {
805                         skb_queue_head(&sk->sk_receive_queue, skb);
806                         goto done;
807                 }
808
809                 kfree_skb(skb);
810
811                 /* Queue backlog skbs */
812                 rskb = skb_dequeue(&iucv->backlog_skb_q);
813                 while (rskb) {
814                         if (sock_queue_rcv_skb(sk, rskb)) {
815                                 skb_queue_head(&iucv->backlog_skb_q,
816                                                 rskb);
817                                 break;
818                         } else {
819                                 rskb = skb_dequeue(&iucv->backlog_skb_q);
820                         }
821                 }
822                 if (skb_queue_empty(&iucv->backlog_skb_q)) {
823                         spin_lock_bh(&iucv->message_q.lock);
824                         if (!list_empty(&iucv->message_q.list))
825                                 iucv_process_message_q(sk);
826                         spin_unlock_bh(&iucv->message_q.lock);
827                 }
828
829         } else
830                 skb_queue_head(&sk->sk_receive_queue, skb);
831
832 done:
833         return err ? : copied;
834 }
835
836 static inline unsigned int iucv_accept_poll(struct sock *parent)
837 {
838         struct iucv_sock *isk, *n;
839         struct sock *sk;
840
841         list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
842                 sk = (struct sock *) isk;
843
844                 if (sk->sk_state == IUCV_CONNECTED)
845                         return POLLIN | POLLRDNORM;
846         }
847
848         return 0;
849 }
850
851 unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
852                             poll_table *wait)
853 {
854         struct sock *sk = sock->sk;
855         unsigned int mask = 0;
856
857         poll_wait(file, sk->sk_sleep, wait);
858
859         if (sk->sk_state == IUCV_LISTEN)
860                 return iucv_accept_poll(sk);
861
862         if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
863                 mask |= POLLERR;
864
865         if (sk->sk_shutdown & RCV_SHUTDOWN)
866                 mask |= POLLRDHUP;
867
868         if (sk->sk_shutdown == SHUTDOWN_MASK)
869                 mask |= POLLHUP;
870
871         if (!skb_queue_empty(&sk->sk_receive_queue) ||
872             (sk->sk_shutdown & RCV_SHUTDOWN))
873                 mask |= POLLIN | POLLRDNORM;
874
875         if (sk->sk_state == IUCV_CLOSED)
876                 mask |= POLLHUP;
877
878         if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_SEVERED)
879                 mask |= POLLIN;
880
881         if (sock_writeable(sk))
882                 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
883         else
884                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
885
886         return mask;
887 }
888
889 static int iucv_sock_shutdown(struct socket *sock, int how)
890 {
891         struct sock *sk = sock->sk;
892         struct iucv_sock *iucv = iucv_sk(sk);
893         struct iucv_message txmsg;
894         int err = 0;
895         u8 prmmsg[8] = {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
896
897         how++;
898
899         if ((how & ~SHUTDOWN_MASK) || !how)
900                 return -EINVAL;
901
902         lock_sock(sk);
903         switch (sk->sk_state) {
904         case IUCV_CLOSED:
905                 err = -ENOTCONN;
906                 goto fail;
907
908         default:
909                 sk->sk_shutdown |= how;
910                 break;
911         }
912
913         if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
914                 txmsg.class = 0;
915                 txmsg.tag = 0;
916                 err = iucv_message_send(iucv->path, &txmsg, IUCV_IPRMDATA, 0,
917                                         (void *) prmmsg, 8);
918                 if (err) {
919                         switch (err) {
920                         case 1:
921                                 err = -ENOTCONN;
922                                 break;
923                         case 2:
924                                 err = -ECONNRESET;
925                                 break;
926                         default:
927                                 err = -ENOTCONN;
928                                 break;
929                         }
930                 }
931         }
932
933         if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
934                 err = iucv_path_quiesce(iucv_sk(sk)->path, NULL);
935                 if (err)
936                         err = -ENOTCONN;
937
938                 skb_queue_purge(&sk->sk_receive_queue);
939         }
940
941         /* Wake up anyone sleeping in poll */
942         sk->sk_state_change(sk);
943
944 fail:
945         release_sock(sk);
946         return err;
947 }
948
949 static int iucv_sock_release(struct socket *sock)
950 {
951         struct sock *sk = sock->sk;
952         int err = 0;
953
954         if (!sk)
955                 return 0;
956
957         iucv_sock_close(sk);
958
959         /* Unregister with IUCV base support */
960         if (iucv_sk(sk)->path) {
961                 iucv_path_sever(iucv_sk(sk)->path, NULL);
962                 iucv_path_free(iucv_sk(sk)->path);
963                 iucv_sk(sk)->path = NULL;
964         }
965
966         sock_orphan(sk);
967         iucv_sock_kill(sk);
968         return err;
969 }
970
971 /* Callback wrappers - called from iucv base support */
972 static int iucv_callback_connreq(struct iucv_path *path,
973                                  u8 ipvmid[8], u8 ipuser[16])
974 {
975         unsigned char user_data[16];
976         unsigned char nuser_data[16];
977         unsigned char src_name[8];
978         struct hlist_node *node;
979         struct sock *sk, *nsk;
980         struct iucv_sock *iucv, *niucv;
981         int err;
982
983         memcpy(src_name, ipuser, 8);
984         EBCASC(src_name, 8);
985         /* Find out if this path belongs to af_iucv. */
986         read_lock(&iucv_sk_list.lock);
987         iucv = NULL;
988         sk = NULL;
989         sk_for_each(sk, node, &iucv_sk_list.head)
990                 if (sk->sk_state == IUCV_LISTEN &&
991                     !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
992                         /*
993                          * Found a listening socket with
994                          * src_name == ipuser[0-7].
995                          */
996                         iucv = iucv_sk(sk);
997                         break;
998                 }
999         read_unlock(&iucv_sk_list.lock);
1000         if (!iucv)
1001                 /* No socket found, not one of our paths. */
1002                 return -EINVAL;
1003
1004         bh_lock_sock(sk);
1005
1006         /* Check if parent socket is listening */
1007         low_nmcpy(user_data, iucv->src_name);
1008         high_nmcpy(user_data, iucv->dst_name);
1009         ASCEBC(user_data, sizeof(user_data));
1010         if (sk->sk_state != IUCV_LISTEN) {
1011                 err = iucv_path_sever(path, user_data);
1012                 goto fail;
1013         }
1014
1015         /* Check for backlog size */
1016         if (sk_acceptq_is_full(sk)) {
1017                 err = iucv_path_sever(path, user_data);
1018                 goto fail;
1019         }
1020
1021         /* Create the new socket */
1022         nsk = iucv_sock_alloc(NULL, SOCK_STREAM, GFP_ATOMIC);
1023         if (!nsk) {
1024                 err = iucv_path_sever(path, user_data);
1025                 goto fail;
1026         }
1027
1028         niucv = iucv_sk(nsk);
1029         iucv_sock_init(nsk, sk);
1030
1031         /* Set the new iucv_sock */
1032         memcpy(niucv->dst_name, ipuser + 8, 8);
1033         EBCASC(niucv->dst_name, 8);
1034         memcpy(niucv->dst_user_id, ipvmid, 8);
1035         memcpy(niucv->src_name, iucv->src_name, 8);
1036         memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1037         niucv->path = path;
1038
1039         /* Call iucv_accept */
1040         high_nmcpy(nuser_data, ipuser + 8);
1041         memcpy(nuser_data + 8, niucv->src_name, 8);
1042         ASCEBC(nuser_data + 8, 8);
1043
1044         path->msglim = IUCV_QUEUELEN_DEFAULT;
1045         err = iucv_path_accept(path, &af_iucv_handler, nuser_data, nsk);
1046         if (err) {
1047                 err = iucv_path_sever(path, user_data);
1048                 goto fail;
1049         }
1050
1051         iucv_accept_enqueue(sk, nsk);
1052
1053         /* Wake up accept */
1054         nsk->sk_state = IUCV_CONNECTED;
1055         sk->sk_data_ready(sk, 1);
1056         err = 0;
1057 fail:
1058         bh_unlock_sock(sk);
1059         return 0;
1060 }
1061
1062 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1063 {
1064         struct sock *sk = path->private;
1065
1066         sk->sk_state = IUCV_CONNECTED;
1067         sk->sk_state_change(sk);
1068 }
1069
1070 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1071 {
1072         struct sock *sk = path->private;
1073         struct iucv_sock *iucv = iucv_sk(sk);
1074         struct sk_buff *skb;
1075         struct sock_msg_q *save_msg;
1076         int len;
1077
1078         if (sk->sk_shutdown & RCV_SHUTDOWN)
1079                 return;
1080
1081         if (!list_empty(&iucv->message_q.list) ||
1082             !skb_queue_empty(&iucv->backlog_skb_q))
1083                 goto save_message;
1084
1085         len = atomic_read(&sk->sk_rmem_alloc);
1086         len += msg->length + sizeof(struct sk_buff);
1087         if (len > sk->sk_rcvbuf)
1088                 goto save_message;
1089
1090         skb = alloc_skb(msg->length, GFP_ATOMIC | GFP_DMA);
1091         if (!skb)
1092                 goto save_message;
1093
1094         spin_lock(&iucv->message_q.lock);
1095         iucv_process_message(sk, skb, path, msg);
1096         spin_unlock(&iucv->message_q.lock);
1097
1098         return;
1099
1100 save_message:
1101         save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1102         if (!save_msg)
1103                 return;
1104         save_msg->path = path;
1105         save_msg->msg = *msg;
1106
1107         spin_lock(&iucv->message_q.lock);
1108         list_add_tail(&save_msg->list, &iucv->message_q.list);
1109         spin_unlock(&iucv->message_q.lock);
1110 }
1111
1112 static void iucv_callback_txdone(struct iucv_path *path,
1113                                  struct iucv_message *msg)
1114 {
1115         struct sock *sk = path->private;
1116         struct sk_buff *this = NULL;
1117         struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1118         struct sk_buff *list_skb = list->next;
1119         unsigned long flags;
1120
1121         if (!skb_queue_empty(list)) {
1122                 spin_lock_irqsave(&list->lock, flags);
1123
1124                 while (list_skb != (struct sk_buff *)list) {
1125                         if (!memcmp(&msg->tag, list_skb->cb, 4)) {
1126                                 this = list_skb;
1127                                 break;
1128                         }
1129                         list_skb = list_skb->next;
1130                 }
1131                 if (this)
1132                         __skb_unlink(this, list);
1133
1134                 spin_unlock_irqrestore(&list->lock, flags);
1135
1136                 if (this)
1137                         kfree_skb(this);
1138         }
1139         BUG_ON(!this);
1140
1141         if (sk->sk_state == IUCV_CLOSING) {
1142                 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1143                         sk->sk_state = IUCV_CLOSED;
1144                         sk->sk_state_change(sk);
1145                 }
1146         }
1147
1148 }
1149
1150 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1151 {
1152         struct sock *sk = path->private;
1153
1154         if (!list_empty(&iucv_sk(sk)->accept_q))
1155                 sk->sk_state = IUCV_SEVERED;
1156         else
1157                 sk->sk_state = IUCV_DISCONN;
1158
1159         sk->sk_state_change(sk);
1160 }
1161
1162 static struct proto_ops iucv_sock_ops = {
1163         .family         = PF_IUCV,
1164         .owner          = THIS_MODULE,
1165         .release        = iucv_sock_release,
1166         .bind           = iucv_sock_bind,
1167         .connect        = iucv_sock_connect,
1168         .listen         = iucv_sock_listen,
1169         .accept         = iucv_sock_accept,
1170         .getname        = iucv_sock_getname,
1171         .sendmsg        = iucv_sock_sendmsg,
1172         .recvmsg        = iucv_sock_recvmsg,
1173         .poll           = iucv_sock_poll,
1174         .ioctl          = sock_no_ioctl,
1175         .mmap           = sock_no_mmap,
1176         .socketpair     = sock_no_socketpair,
1177         .shutdown       = iucv_sock_shutdown,
1178         .setsockopt     = sock_no_setsockopt,
1179         .getsockopt     = sock_no_getsockopt
1180 };
1181
1182 static struct net_proto_family iucv_sock_family_ops = {
1183         .family = AF_IUCV,
1184         .owner  = THIS_MODULE,
1185         .create = iucv_sock_create,
1186 };
1187
1188 static int __init afiucv_init(void)
1189 {
1190         int err;
1191
1192         if (!MACHINE_IS_VM) {
1193                 printk(KERN_ERR "AF_IUCV connection needs VM as base\n");
1194                 err = -EPROTONOSUPPORT;
1195                 goto out;
1196         }
1197         cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
1198         if (unlikely(err)) {
1199                 WARN_ON(err);
1200                 err = -EPROTONOSUPPORT;
1201                 goto out;
1202         }
1203
1204         err = iucv_register(&af_iucv_handler, 0);
1205         if (err)
1206                 goto out;
1207         err = proto_register(&iucv_proto, 0);
1208         if (err)
1209                 goto out_iucv;
1210         err = sock_register(&iucv_sock_family_ops);
1211         if (err)
1212                 goto out_proto;
1213         return 0;
1214
1215 out_proto:
1216         proto_unregister(&iucv_proto);
1217 out_iucv:
1218         iucv_unregister(&af_iucv_handler, 0);
1219 out:
1220         return err;
1221 }
1222
1223 static void __exit afiucv_exit(void)
1224 {
1225         sock_unregister(PF_IUCV);
1226         proto_unregister(&iucv_proto);
1227         iucv_unregister(&af_iucv_handler, 0);
1228 }
1229
1230 module_init(afiucv_init);
1231 module_exit(afiucv_exit);
1232
1233 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
1234 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
1235 MODULE_VERSION(VERSION);
1236 MODULE_LICENSE("GPL");
1237 MODULE_ALIAS_NETPROTO(PF_IUCV);