xref: /linux/net/iucv/af_iucv.c (revision a4eb44a6435d6d8f9e642407a4a06f65eb90ca04)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  IUCV protocol stack for Linux on zSeries
4  *
5  *  Copyright IBM Corp. 2006, 2009
6  *
7  *  Author(s):	Jennifer Hunt <jenhunt@us.ibm.com>
8  *		Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
9  *  PM functions:
10  *		Ursula Braun <ursula.braun@de.ibm.com>
11  */
12 
13 #define KMSG_COMPONENT "af_iucv"
14 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
15 
16 #include <linux/filter.h>
17 #include <linux/module.h>
18 #include <linux/netdevice.h>
19 #include <linux/types.h>
20 #include <linux/limits.h>
21 #include <linux/list.h>
22 #include <linux/errno.h>
23 #include <linux/kernel.h>
24 #include <linux/sched/signal.h>
25 #include <linux/slab.h>
26 #include <linux/skbuff.h>
27 #include <linux/init.h>
28 #include <linux/poll.h>
29 #include <linux/security.h>
30 #include <net/sock.h>
31 #include <asm/ebcdic.h>
32 #include <asm/cpcmd.h>
33 #include <linux/kmod.h>
34 
35 #include <net/iucv/af_iucv.h>
36 
37 #define VERSION "1.2"
38 
39 static char iucv_userid[80];
40 
41 static struct proto iucv_proto = {
42 	.name		= "AF_IUCV",
43 	.owner		= THIS_MODULE,
44 	.obj_size	= sizeof(struct iucv_sock),
45 };
46 
47 static struct iucv_interface *pr_iucv;
48 static struct iucv_handler af_iucv_handler;
49 
50 /* special AF_IUCV IPRM messages */
51 static const u8 iprm_shutdown[8] =
52 	{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
53 
54 #define TRGCLS_SIZE	sizeof_field(struct iucv_message, class)
55 
56 #define __iucv_sock_wait(sk, condition, timeo, ret)			\
57 do {									\
58 	DEFINE_WAIT(__wait);						\
59 	long __timeo = timeo;						\
60 	ret = 0;							\
61 	prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);	\
62 	while (!(condition)) {						\
63 		if (!__timeo) {						\
64 			ret = -EAGAIN;					\
65 			break;						\
66 		}							\
67 		if (signal_pending(current)) {				\
68 			ret = sock_intr_errno(__timeo);			\
69 			break;						\
70 		}							\
71 		release_sock(sk);					\
72 		__timeo = schedule_timeout(__timeo);			\
73 		lock_sock(sk);						\
74 		ret = sock_error(sk);					\
75 		if (ret)						\
76 			break;						\
77 	}								\
78 	finish_wait(sk_sleep(sk), &__wait);				\
79 } while (0)
80 
81 #define iucv_sock_wait(sk, condition, timeo)				\
82 ({									\
83 	int __ret = 0;							\
84 	if (!(condition))						\
85 		__iucv_sock_wait(sk, condition, timeo, __ret);		\
86 	__ret;								\
87 })
88 
89 static struct sock *iucv_accept_dequeue(struct sock *parent,
90 					struct socket *newsock);
91 static void iucv_sock_kill(struct sock *sk);
92 static void iucv_sock_close(struct sock *sk);
93 
94 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify);
95 
96 static struct iucv_sock_list iucv_sk_list = {
97 	.lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
98 	.autobind_name = ATOMIC_INIT(0)
99 };
100 
101 static inline void high_nmcpy(unsigned char *dst, char *src)
102 {
103        memcpy(dst, src, 8);
104 }
105 
106 static inline void low_nmcpy(unsigned char *dst, char *src)
107 {
108        memcpy(&dst[8], src, 8);
109 }
110 
111 /**
112  * iucv_msg_length() - Returns the length of an iucv message.
113  * @msg:	Pointer to struct iucv_message, MUST NOT be NULL
114  *
115  * The function returns the length of the specified iucv message @msg of data
116  * stored in a buffer and of data stored in the parameter list (PRMDATA).
117  *
118  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
119  * data:
120  *	PRMDATA[0..6]	socket data (max 7 bytes);
121  *	PRMDATA[7]	socket data length value (len is 0xff - PRMDATA[7])
122  *
123  * The socket data length is computed by subtracting the socket data length
124  * value from 0xFF.
125  * If the socket data len is greater 7, then PRMDATA can be used for special
126  * notifications (see iucv_sock_shutdown); and further,
127  * if the socket data len is > 7, the function returns 8.
128  *
129  * Use this function to allocate socket buffers to store iucv message data.
130  */
131 static inline size_t iucv_msg_length(struct iucv_message *msg)
132 {
133 	size_t datalen;
134 
135 	if (msg->flags & IUCV_IPRMDATA) {
136 		datalen = 0xff - msg->rmmsg[7];
137 		return (datalen < 8) ? datalen : 8;
138 	}
139 	return msg->length;
140 }
141 
142 /**
143  * iucv_sock_in_state() - check for specific states
144  * @sk:		sock structure
145  * @state:	first iucv sk state
146  * @state2:	second iucv sk state
147  *
148  * Returns true if the socket in either in the first or second state.
149  */
150 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
151 {
152 	return (sk->sk_state == state || sk->sk_state == state2);
153 }
154 
155 /**
156  * iucv_below_msglim() - function to check if messages can be sent
157  * @sk:		sock structure
158  *
159  * Returns true if the send queue length is lower than the message limit.
160  * Always returns true if the socket is not connected (no iucv path for
161  * checking the message limit).
162  */
163 static inline int iucv_below_msglim(struct sock *sk)
164 {
165 	struct iucv_sock *iucv = iucv_sk(sk);
166 
167 	if (sk->sk_state != IUCV_CONNECTED)
168 		return 1;
169 	if (iucv->transport == AF_IUCV_TRANS_IUCV)
170 		return (atomic_read(&iucv->skbs_in_xmit) < iucv->path->msglim);
171 	else
172 		return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
173 			(atomic_read(&iucv->pendings) <= 0));
174 }
175 
176 /*
177  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
178  */
179 static void iucv_sock_wake_msglim(struct sock *sk)
180 {
181 	struct socket_wq *wq;
182 
183 	rcu_read_lock();
184 	wq = rcu_dereference(sk->sk_wq);
185 	if (skwq_has_sleeper(wq))
186 		wake_up_interruptible_all(&wq->wait);
187 	sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
188 	rcu_read_unlock();
189 }
190 
191 /*
192  * afiucv_hs_send() - send a message through HiperSockets transport
193  */
194 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
195 		   struct sk_buff *skb, u8 flags)
196 {
197 	struct iucv_sock *iucv = iucv_sk(sock);
198 	struct af_iucv_trans_hdr *phs_hdr;
199 	int err, confirm_recv = 0;
200 
201 	phs_hdr = skb_push(skb, sizeof(*phs_hdr));
202 	memset(phs_hdr, 0, sizeof(*phs_hdr));
203 	skb_reset_network_header(skb);
204 
205 	phs_hdr->magic = ETH_P_AF_IUCV;
206 	phs_hdr->version = 1;
207 	phs_hdr->flags = flags;
208 	if (flags == AF_IUCV_FLAG_SYN)
209 		phs_hdr->window = iucv->msglimit;
210 	else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
211 		confirm_recv = atomic_read(&iucv->msg_recv);
212 		phs_hdr->window = confirm_recv;
213 		if (confirm_recv)
214 			phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
215 	}
216 	memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
217 	memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
218 	memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
219 	memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
220 	ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
221 	ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
222 	ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
223 	ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
224 	if (imsg)
225 		memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
226 
227 	skb->dev = iucv->hs_dev;
228 	if (!skb->dev) {
229 		err = -ENODEV;
230 		goto err_free;
231 	}
232 
233 	dev_hard_header(skb, skb->dev, ETH_P_AF_IUCV, NULL, NULL, skb->len);
234 
235 	if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev)) {
236 		err = -ENETDOWN;
237 		goto err_free;
238 	}
239 	if (skb->len > skb->dev->mtu) {
240 		if (sock->sk_type == SOCK_SEQPACKET) {
241 			err = -EMSGSIZE;
242 			goto err_free;
243 		}
244 		err = pskb_trim(skb, skb->dev->mtu);
245 		if (err)
246 			goto err_free;
247 	}
248 	skb->protocol = cpu_to_be16(ETH_P_AF_IUCV);
249 
250 	atomic_inc(&iucv->skbs_in_xmit);
251 	err = dev_queue_xmit(skb);
252 	if (net_xmit_eval(err)) {
253 		atomic_dec(&iucv->skbs_in_xmit);
254 	} else {
255 		atomic_sub(confirm_recv, &iucv->msg_recv);
256 		WARN_ON(atomic_read(&iucv->msg_recv) < 0);
257 	}
258 	return net_xmit_eval(err);
259 
260 err_free:
261 	kfree_skb(skb);
262 	return err;
263 }
264 
265 static struct sock *__iucv_get_sock_by_name(char *nm)
266 {
267 	struct sock *sk;
268 
269 	sk_for_each(sk, &iucv_sk_list.head)
270 		if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
271 			return sk;
272 
273 	return NULL;
274 }
275 
276 static void iucv_sock_destruct(struct sock *sk)
277 {
278 	skb_queue_purge(&sk->sk_receive_queue);
279 	skb_queue_purge(&sk->sk_error_queue);
280 
281 	sk_mem_reclaim(sk);
282 
283 	if (!sock_flag(sk, SOCK_DEAD)) {
284 		pr_err("Attempt to release alive iucv socket %p\n", sk);
285 		return;
286 	}
287 
288 	WARN_ON(atomic_read(&sk->sk_rmem_alloc));
289 	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
290 	WARN_ON(sk->sk_wmem_queued);
291 	WARN_ON(sk->sk_forward_alloc);
292 }
293 
294 /* Cleanup Listen */
295 static void iucv_sock_cleanup_listen(struct sock *parent)
296 {
297 	struct sock *sk;
298 
299 	/* Close non-accepted connections */
300 	while ((sk = iucv_accept_dequeue(parent, NULL))) {
301 		iucv_sock_close(sk);
302 		iucv_sock_kill(sk);
303 	}
304 
305 	parent->sk_state = IUCV_CLOSED;
306 }
307 
308 static void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
309 {
310 	write_lock_bh(&l->lock);
311 	sk_add_node(sk, &l->head);
312 	write_unlock_bh(&l->lock);
313 }
314 
315 static void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
316 {
317 	write_lock_bh(&l->lock);
318 	sk_del_node_init(sk);
319 	write_unlock_bh(&l->lock);
320 }
321 
322 /* Kill socket (only if zapped and orphaned) */
323 static void iucv_sock_kill(struct sock *sk)
324 {
325 	if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
326 		return;
327 
328 	iucv_sock_unlink(&iucv_sk_list, sk);
329 	sock_set_flag(sk, SOCK_DEAD);
330 	sock_put(sk);
331 }
332 
333 /* Terminate an IUCV path */
334 static void iucv_sever_path(struct sock *sk, int with_user_data)
335 {
336 	unsigned char user_data[16];
337 	struct iucv_sock *iucv = iucv_sk(sk);
338 	struct iucv_path *path = iucv->path;
339 
340 	if (iucv->path) {
341 		iucv->path = NULL;
342 		if (with_user_data) {
343 			low_nmcpy(user_data, iucv->src_name);
344 			high_nmcpy(user_data, iucv->dst_name);
345 			ASCEBC(user_data, sizeof(user_data));
346 			pr_iucv->path_sever(path, user_data);
347 		} else
348 			pr_iucv->path_sever(path, NULL);
349 		iucv_path_free(path);
350 	}
351 }
352 
353 /* Send controlling flags through an IUCV socket for HIPER transport */
354 static int iucv_send_ctrl(struct sock *sk, u8 flags)
355 {
356 	struct iucv_sock *iucv = iucv_sk(sk);
357 	int err = 0;
358 	int blen;
359 	struct sk_buff *skb;
360 	u8 shutdown = 0;
361 
362 	blen = sizeof(struct af_iucv_trans_hdr) +
363 	       LL_RESERVED_SPACE(iucv->hs_dev);
364 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
365 		/* controlling flags should be sent anyway */
366 		shutdown = sk->sk_shutdown;
367 		sk->sk_shutdown &= RCV_SHUTDOWN;
368 	}
369 	skb = sock_alloc_send_skb(sk, blen, 1, &err);
370 	if (skb) {
371 		skb_reserve(skb, blen);
372 		err = afiucv_hs_send(NULL, sk, skb, flags);
373 	}
374 	if (shutdown)
375 		sk->sk_shutdown = shutdown;
376 	return err;
377 }
378 
379 /* Close an IUCV socket */
380 static void iucv_sock_close(struct sock *sk)
381 {
382 	struct iucv_sock *iucv = iucv_sk(sk);
383 	unsigned long timeo;
384 	int err = 0;
385 
386 	lock_sock(sk);
387 
388 	switch (sk->sk_state) {
389 	case IUCV_LISTEN:
390 		iucv_sock_cleanup_listen(sk);
391 		break;
392 
393 	case IUCV_CONNECTED:
394 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
395 			err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
396 			sk->sk_state = IUCV_DISCONN;
397 			sk->sk_state_change(sk);
398 		}
399 		fallthrough;
400 
401 	case IUCV_DISCONN:
402 		sk->sk_state = IUCV_CLOSING;
403 		sk->sk_state_change(sk);
404 
405 		if (!err && atomic_read(&iucv->skbs_in_xmit) > 0) {
406 			if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
407 				timeo = sk->sk_lingertime;
408 			else
409 				timeo = IUCV_DISCONN_TIMEOUT;
410 			iucv_sock_wait(sk,
411 					iucv_sock_in_state(sk, IUCV_CLOSED, 0),
412 					timeo);
413 		}
414 		fallthrough;
415 
416 	case IUCV_CLOSING:
417 		sk->sk_state = IUCV_CLOSED;
418 		sk->sk_state_change(sk);
419 
420 		sk->sk_err = ECONNRESET;
421 		sk->sk_state_change(sk);
422 
423 		skb_queue_purge(&iucv->send_skb_q);
424 		skb_queue_purge(&iucv->backlog_skb_q);
425 		fallthrough;
426 
427 	default:
428 		iucv_sever_path(sk, 1);
429 	}
430 
431 	if (iucv->hs_dev) {
432 		dev_put(iucv->hs_dev);
433 		iucv->hs_dev = NULL;
434 		sk->sk_bound_dev_if = 0;
435 	}
436 
437 	/* mark socket for deletion by iucv_sock_kill() */
438 	sock_set_flag(sk, SOCK_ZAPPED);
439 
440 	release_sock(sk);
441 }
442 
443 static void iucv_sock_init(struct sock *sk, struct sock *parent)
444 {
445 	if (parent) {
446 		sk->sk_type = parent->sk_type;
447 		security_sk_clone(parent, sk);
448 	}
449 }
450 
451 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio, int kern)
452 {
453 	struct sock *sk;
454 	struct iucv_sock *iucv;
455 
456 	sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto, kern);
457 	if (!sk)
458 		return NULL;
459 	iucv = iucv_sk(sk);
460 
461 	sock_init_data(sock, sk);
462 	INIT_LIST_HEAD(&iucv->accept_q);
463 	spin_lock_init(&iucv->accept_q_lock);
464 	skb_queue_head_init(&iucv->send_skb_q);
465 	INIT_LIST_HEAD(&iucv->message_q.list);
466 	spin_lock_init(&iucv->message_q.lock);
467 	skb_queue_head_init(&iucv->backlog_skb_q);
468 	iucv->send_tag = 0;
469 	atomic_set(&iucv->pendings, 0);
470 	iucv->flags = 0;
471 	iucv->msglimit = 0;
472 	atomic_set(&iucv->skbs_in_xmit, 0);
473 	atomic_set(&iucv->msg_sent, 0);
474 	atomic_set(&iucv->msg_recv, 0);
475 	iucv->path = NULL;
476 	iucv->sk_txnotify = afiucv_hs_callback_txnotify;
477 	memset(&iucv->init, 0, sizeof(iucv->init));
478 	if (pr_iucv)
479 		iucv->transport = AF_IUCV_TRANS_IUCV;
480 	else
481 		iucv->transport = AF_IUCV_TRANS_HIPER;
482 
483 	sk->sk_destruct = iucv_sock_destruct;
484 	sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
485 
486 	sock_reset_flag(sk, SOCK_ZAPPED);
487 
488 	sk->sk_protocol = proto;
489 	sk->sk_state	= IUCV_OPEN;
490 
491 	iucv_sock_link(&iucv_sk_list, sk);
492 	return sk;
493 }
494 
495 static void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
496 {
497 	unsigned long flags;
498 	struct iucv_sock *par = iucv_sk(parent);
499 
500 	sock_hold(sk);
501 	spin_lock_irqsave(&par->accept_q_lock, flags);
502 	list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
503 	spin_unlock_irqrestore(&par->accept_q_lock, flags);
504 	iucv_sk(sk)->parent = parent;
505 	sk_acceptq_added(parent);
506 }
507 
508 static void iucv_accept_unlink(struct sock *sk)
509 {
510 	unsigned long flags;
511 	struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
512 
513 	spin_lock_irqsave(&par->accept_q_lock, flags);
514 	list_del_init(&iucv_sk(sk)->accept_q);
515 	spin_unlock_irqrestore(&par->accept_q_lock, flags);
516 	sk_acceptq_removed(iucv_sk(sk)->parent);
517 	iucv_sk(sk)->parent = NULL;
518 	sock_put(sk);
519 }
520 
521 static struct sock *iucv_accept_dequeue(struct sock *parent,
522 					struct socket *newsock)
523 {
524 	struct iucv_sock *isk, *n;
525 	struct sock *sk;
526 
527 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
528 		sk = (struct sock *) isk;
529 		lock_sock(sk);
530 
531 		if (sk->sk_state == IUCV_CLOSED) {
532 			iucv_accept_unlink(sk);
533 			release_sock(sk);
534 			continue;
535 		}
536 
537 		if (sk->sk_state == IUCV_CONNECTED ||
538 		    sk->sk_state == IUCV_DISCONN ||
539 		    !newsock) {
540 			iucv_accept_unlink(sk);
541 			if (newsock)
542 				sock_graft(sk, newsock);
543 
544 			release_sock(sk);
545 			return sk;
546 		}
547 
548 		release_sock(sk);
549 	}
550 	return NULL;
551 }
552 
553 static void __iucv_auto_name(struct iucv_sock *iucv)
554 {
555 	char name[12];
556 
557 	sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
558 	while (__iucv_get_sock_by_name(name)) {
559 		sprintf(name, "%08x",
560 			atomic_inc_return(&iucv_sk_list.autobind_name));
561 	}
562 	memcpy(iucv->src_name, name, 8);
563 }
564 
565 /* Bind an unbound socket */
566 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
567 			  int addr_len)
568 {
569 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
570 	char uid[sizeof(sa->siucv_user_id)];
571 	struct sock *sk = sock->sk;
572 	struct iucv_sock *iucv;
573 	int err = 0;
574 	struct net_device *dev;
575 
576 	/* Verify the input sockaddr */
577 	if (addr_len < sizeof(struct sockaddr_iucv) ||
578 	    addr->sa_family != AF_IUCV)
579 		return -EINVAL;
580 
581 	lock_sock(sk);
582 	if (sk->sk_state != IUCV_OPEN) {
583 		err = -EBADFD;
584 		goto done;
585 	}
586 
587 	write_lock_bh(&iucv_sk_list.lock);
588 
589 	iucv = iucv_sk(sk);
590 	if (__iucv_get_sock_by_name(sa->siucv_name)) {
591 		err = -EADDRINUSE;
592 		goto done_unlock;
593 	}
594 	if (iucv->path)
595 		goto done_unlock;
596 
597 	/* Bind the socket */
598 	if (pr_iucv)
599 		if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
600 			goto vm_bind; /* VM IUCV transport */
601 
602 	/* try hiper transport */
603 	memcpy(uid, sa->siucv_user_id, sizeof(uid));
604 	ASCEBC(uid, 8);
605 	rcu_read_lock();
606 	for_each_netdev_rcu(&init_net, dev) {
607 		if (!memcmp(dev->perm_addr, uid, 8)) {
608 			memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
609 			/* Check for uninitialized siucv_name */
610 			if (strncmp(sa->siucv_name, "        ", 8) == 0)
611 				__iucv_auto_name(iucv);
612 			else
613 				memcpy(iucv->src_name, sa->siucv_name, 8);
614 			sk->sk_bound_dev_if = dev->ifindex;
615 			iucv->hs_dev = dev;
616 			dev_hold(dev);
617 			sk->sk_state = IUCV_BOUND;
618 			iucv->transport = AF_IUCV_TRANS_HIPER;
619 			if (!iucv->msglimit)
620 				iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
621 			rcu_read_unlock();
622 			goto done_unlock;
623 		}
624 	}
625 	rcu_read_unlock();
626 vm_bind:
627 	if (pr_iucv) {
628 		/* use local userid for backward compat */
629 		memcpy(iucv->src_name, sa->siucv_name, 8);
630 		memcpy(iucv->src_user_id, iucv_userid, 8);
631 		sk->sk_state = IUCV_BOUND;
632 		iucv->transport = AF_IUCV_TRANS_IUCV;
633 		sk->sk_allocation |= GFP_DMA;
634 		if (!iucv->msglimit)
635 			iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
636 		goto done_unlock;
637 	}
638 	/* found no dev to bind */
639 	err = -ENODEV;
640 done_unlock:
641 	/* Release the socket list lock */
642 	write_unlock_bh(&iucv_sk_list.lock);
643 done:
644 	release_sock(sk);
645 	return err;
646 }
647 
648 /* Automatically bind an unbound socket */
649 static int iucv_sock_autobind(struct sock *sk)
650 {
651 	struct iucv_sock *iucv = iucv_sk(sk);
652 	int err = 0;
653 
654 	if (unlikely(!pr_iucv))
655 		return -EPROTO;
656 
657 	memcpy(iucv->src_user_id, iucv_userid, 8);
658 	iucv->transport = AF_IUCV_TRANS_IUCV;
659 	sk->sk_allocation |= GFP_DMA;
660 
661 	write_lock_bh(&iucv_sk_list.lock);
662 	__iucv_auto_name(iucv);
663 	write_unlock_bh(&iucv_sk_list.lock);
664 
665 	if (!iucv->msglimit)
666 		iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
667 
668 	return err;
669 }
670 
671 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
672 {
673 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
674 	struct sock *sk = sock->sk;
675 	struct iucv_sock *iucv = iucv_sk(sk);
676 	unsigned char user_data[16];
677 	int err;
678 
679 	high_nmcpy(user_data, sa->siucv_name);
680 	low_nmcpy(user_data, iucv->src_name);
681 	ASCEBC(user_data, sizeof(user_data));
682 
683 	/* Create path. */
684 	iucv->path = iucv_path_alloc(iucv->msglimit,
685 				     IUCV_IPRMDATA, GFP_KERNEL);
686 	if (!iucv->path) {
687 		err = -ENOMEM;
688 		goto done;
689 	}
690 	err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
691 				    sa->siucv_user_id, NULL, user_data,
692 				    sk);
693 	if (err) {
694 		iucv_path_free(iucv->path);
695 		iucv->path = NULL;
696 		switch (err) {
697 		case 0x0b:	/* Target communicator is not logged on */
698 			err = -ENETUNREACH;
699 			break;
700 		case 0x0d:	/* Max connections for this guest exceeded */
701 		case 0x0e:	/* Max connections for target guest exceeded */
702 			err = -EAGAIN;
703 			break;
704 		case 0x0f:	/* Missing IUCV authorization */
705 			err = -EACCES;
706 			break;
707 		default:
708 			err = -ECONNREFUSED;
709 			break;
710 		}
711 	}
712 done:
713 	return err;
714 }
715 
716 /* Connect an unconnected socket */
717 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
718 			     int alen, int flags)
719 {
720 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
721 	struct sock *sk = sock->sk;
722 	struct iucv_sock *iucv = iucv_sk(sk);
723 	int err;
724 
725 	if (alen < sizeof(struct sockaddr_iucv) || addr->sa_family != AF_IUCV)
726 		return -EINVAL;
727 
728 	if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
729 		return -EBADFD;
730 
731 	if (sk->sk_state == IUCV_OPEN &&
732 	    iucv->transport == AF_IUCV_TRANS_HIPER)
733 		return -EBADFD; /* explicit bind required */
734 
735 	if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
736 		return -EINVAL;
737 
738 	if (sk->sk_state == IUCV_OPEN) {
739 		err = iucv_sock_autobind(sk);
740 		if (unlikely(err))
741 			return err;
742 	}
743 
744 	lock_sock(sk);
745 
746 	/* Set the destination information */
747 	memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
748 	memcpy(iucv->dst_name, sa->siucv_name, 8);
749 
750 	if (iucv->transport == AF_IUCV_TRANS_HIPER)
751 		err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
752 	else
753 		err = afiucv_path_connect(sock, addr);
754 	if (err)
755 		goto done;
756 
757 	if (sk->sk_state != IUCV_CONNECTED)
758 		err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
759 							    IUCV_DISCONN),
760 				     sock_sndtimeo(sk, flags & O_NONBLOCK));
761 
762 	if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
763 		err = -ECONNREFUSED;
764 
765 	if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
766 		iucv_sever_path(sk, 0);
767 
768 done:
769 	release_sock(sk);
770 	return err;
771 }
772 
773 /* Move a socket into listening state. */
774 static int iucv_sock_listen(struct socket *sock, int backlog)
775 {
776 	struct sock *sk = sock->sk;
777 	int err;
778 
779 	lock_sock(sk);
780 
781 	err = -EINVAL;
782 	if (sk->sk_state != IUCV_BOUND)
783 		goto done;
784 
785 	if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
786 		goto done;
787 
788 	sk->sk_max_ack_backlog = backlog;
789 	sk->sk_ack_backlog = 0;
790 	sk->sk_state = IUCV_LISTEN;
791 	err = 0;
792 
793 done:
794 	release_sock(sk);
795 	return err;
796 }
797 
798 /* Accept a pending connection */
799 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
800 			    int flags, bool kern)
801 {
802 	DECLARE_WAITQUEUE(wait, current);
803 	struct sock *sk = sock->sk, *nsk;
804 	long timeo;
805 	int err = 0;
806 
807 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
808 
809 	if (sk->sk_state != IUCV_LISTEN) {
810 		err = -EBADFD;
811 		goto done;
812 	}
813 
814 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
815 
816 	/* Wait for an incoming connection */
817 	add_wait_queue_exclusive(sk_sleep(sk), &wait);
818 	while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
819 		set_current_state(TASK_INTERRUPTIBLE);
820 		if (!timeo) {
821 			err = -EAGAIN;
822 			break;
823 		}
824 
825 		release_sock(sk);
826 		timeo = schedule_timeout(timeo);
827 		lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
828 
829 		if (sk->sk_state != IUCV_LISTEN) {
830 			err = -EBADFD;
831 			break;
832 		}
833 
834 		if (signal_pending(current)) {
835 			err = sock_intr_errno(timeo);
836 			break;
837 		}
838 	}
839 
840 	set_current_state(TASK_RUNNING);
841 	remove_wait_queue(sk_sleep(sk), &wait);
842 
843 	if (err)
844 		goto done;
845 
846 	newsock->state = SS_CONNECTED;
847 
848 done:
849 	release_sock(sk);
850 	return err;
851 }
852 
853 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
854 			     int peer)
855 {
856 	DECLARE_SOCKADDR(struct sockaddr_iucv *, siucv, addr);
857 	struct sock *sk = sock->sk;
858 	struct iucv_sock *iucv = iucv_sk(sk);
859 
860 	addr->sa_family = AF_IUCV;
861 
862 	if (peer) {
863 		memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
864 		memcpy(siucv->siucv_name, iucv->dst_name, 8);
865 	} else {
866 		memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
867 		memcpy(siucv->siucv_name, iucv->src_name, 8);
868 	}
869 	memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
870 	memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
871 	memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
872 
873 	return sizeof(struct sockaddr_iucv);
874 }
875 
876 /**
877  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
878  * @path:	IUCV path
879  * @msg:	Pointer to a struct iucv_message
880  * @skb:	The socket data to send, skb->len MUST BE <= 7
881  *
882  * Send the socket data in the parameter list in the iucv message
883  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
884  * list and the socket data len at index 7 (last byte).
885  * See also iucv_msg_length().
886  *
887  * Returns the error code from the iucv_message_send() call.
888  */
889 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
890 			  struct sk_buff *skb)
891 {
892 	u8 prmdata[8];
893 
894 	memcpy(prmdata, (void *) skb->data, skb->len);
895 	prmdata[7] = 0xff - (u8) skb->len;
896 	return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
897 				 (void *) prmdata, 8);
898 }
899 
900 static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,
901 			     size_t len)
902 {
903 	struct sock *sk = sock->sk;
904 	struct iucv_sock *iucv = iucv_sk(sk);
905 	size_t headroom = 0;
906 	size_t linear;
907 	struct sk_buff *skb;
908 	struct iucv_message txmsg = {0};
909 	struct cmsghdr *cmsg;
910 	int cmsg_done;
911 	long timeo;
912 	char user_id[9];
913 	char appl_id[9];
914 	int err;
915 	int noblock = msg->msg_flags & MSG_DONTWAIT;
916 
917 	err = sock_error(sk);
918 	if (err)
919 		return err;
920 
921 	if (msg->msg_flags & MSG_OOB)
922 		return -EOPNOTSUPP;
923 
924 	/* SOCK_SEQPACKET: we do not support segmented records */
925 	if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
926 		return -EOPNOTSUPP;
927 
928 	lock_sock(sk);
929 
930 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
931 		err = -EPIPE;
932 		goto out;
933 	}
934 
935 	/* Return if the socket is not in connected state */
936 	if (sk->sk_state != IUCV_CONNECTED) {
937 		err = -ENOTCONN;
938 		goto out;
939 	}
940 
941 	/* initialize defaults */
942 	cmsg_done   = 0;	/* check for duplicate headers */
943 
944 	/* iterate over control messages */
945 	for_each_cmsghdr(cmsg, msg) {
946 		if (!CMSG_OK(msg, cmsg)) {
947 			err = -EINVAL;
948 			goto out;
949 		}
950 
951 		if (cmsg->cmsg_level != SOL_IUCV)
952 			continue;
953 
954 		if (cmsg->cmsg_type & cmsg_done) {
955 			err = -EINVAL;
956 			goto out;
957 		}
958 		cmsg_done |= cmsg->cmsg_type;
959 
960 		switch (cmsg->cmsg_type) {
961 		case SCM_IUCV_TRGCLS:
962 			if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
963 				err = -EINVAL;
964 				goto out;
965 			}
966 
967 			/* set iucv message target class */
968 			memcpy(&txmsg.class,
969 				(void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
970 
971 			break;
972 
973 		default:
974 			err = -EINVAL;
975 			goto out;
976 		}
977 	}
978 
979 	/* allocate one skb for each iucv message:
980 	 * this is fine for SOCK_SEQPACKET (unless we want to support
981 	 * segmented records using the MSG_EOR flag), but
982 	 * for SOCK_STREAM we might want to improve it in future */
983 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
984 		headroom = sizeof(struct af_iucv_trans_hdr) +
985 			   LL_RESERVED_SPACE(iucv->hs_dev);
986 		linear = min(len, PAGE_SIZE - headroom);
987 	} else {
988 		if (len < PAGE_SIZE) {
989 			linear = len;
990 		} else {
991 			/* In nonlinear "classic" iucv skb,
992 			 * reserve space for iucv_array
993 			 */
994 			headroom = sizeof(struct iucv_array) *
995 				   (MAX_SKB_FRAGS + 1);
996 			linear = PAGE_SIZE - headroom;
997 		}
998 	}
999 	skb = sock_alloc_send_pskb(sk, headroom + linear, len - linear,
1000 				   noblock, &err, 0);
1001 	if (!skb)
1002 		goto out;
1003 	if (headroom)
1004 		skb_reserve(skb, headroom);
1005 	skb_put(skb, linear);
1006 	skb->len = len;
1007 	skb->data_len = len - linear;
1008 	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
1009 	if (err)
1010 		goto fail;
1011 
1012 	/* wait if outstanding messages for iucv path has reached */
1013 	timeo = sock_sndtimeo(sk, noblock);
1014 	err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1015 	if (err)
1016 		goto fail;
1017 
1018 	/* return -ECONNRESET if the socket is no longer connected */
1019 	if (sk->sk_state != IUCV_CONNECTED) {
1020 		err = -ECONNRESET;
1021 		goto fail;
1022 	}
1023 
1024 	/* increment and save iucv message tag for msg_completion cbk */
1025 	txmsg.tag = iucv->send_tag++;
1026 	IUCV_SKB_CB(skb)->tag = txmsg.tag;
1027 
1028 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1029 		atomic_inc(&iucv->msg_sent);
1030 		err = afiucv_hs_send(&txmsg, sk, skb, 0);
1031 		if (err) {
1032 			atomic_dec(&iucv->msg_sent);
1033 			goto out;
1034 		}
1035 	} else { /* Classic VM IUCV transport */
1036 		skb_queue_tail(&iucv->send_skb_q, skb);
1037 		atomic_inc(&iucv->skbs_in_xmit);
1038 
1039 		if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags) &&
1040 		    skb->len <= 7) {
1041 			err = iucv_send_iprm(iucv->path, &txmsg, skb);
1042 
1043 			/* on success: there is no message_complete callback */
1044 			/* for an IPRMDATA msg; remove skb from send queue   */
1045 			if (err == 0) {
1046 				atomic_dec(&iucv->skbs_in_xmit);
1047 				skb_unlink(skb, &iucv->send_skb_q);
1048 				consume_skb(skb);
1049 			}
1050 
1051 			/* this error should never happen since the	*/
1052 			/* IUCV_IPRMDATA path flag is set... sever path */
1053 			if (err == 0x15) {
1054 				pr_iucv->path_sever(iucv->path, NULL);
1055 				atomic_dec(&iucv->skbs_in_xmit);
1056 				skb_unlink(skb, &iucv->send_skb_q);
1057 				err = -EPIPE;
1058 				goto fail;
1059 			}
1060 		} else if (skb_is_nonlinear(skb)) {
1061 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1062 			int i;
1063 
1064 			/* skip iucv_array lying in the headroom */
1065 			iba[0].address = (u32)(addr_t)skb->data;
1066 			iba[0].length = (u32)skb_headlen(skb);
1067 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1068 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1069 
1070 				iba[i + 1].address =
1071 					(u32)(addr_t)skb_frag_address(frag);
1072 				iba[i + 1].length = (u32)skb_frag_size(frag);
1073 			}
1074 			err = pr_iucv->message_send(iucv->path, &txmsg,
1075 						    IUCV_IPBUFLST, 0,
1076 						    (void *)iba, skb->len);
1077 		} else { /* non-IPRM Linear skb */
1078 			err = pr_iucv->message_send(iucv->path, &txmsg,
1079 					0, 0, (void *)skb->data, skb->len);
1080 		}
1081 		if (err) {
1082 			if (err == 3) {
1083 				user_id[8] = 0;
1084 				memcpy(user_id, iucv->dst_user_id, 8);
1085 				appl_id[8] = 0;
1086 				memcpy(appl_id, iucv->dst_name, 8);
1087 				pr_err(
1088 		"Application %s on z/VM guest %s exceeds message limit\n",
1089 					appl_id, user_id);
1090 				err = -EAGAIN;
1091 			} else {
1092 				err = -EPIPE;
1093 			}
1094 
1095 			atomic_dec(&iucv->skbs_in_xmit);
1096 			skb_unlink(skb, &iucv->send_skb_q);
1097 			goto fail;
1098 		}
1099 	}
1100 
1101 	release_sock(sk);
1102 	return len;
1103 
1104 fail:
1105 	kfree_skb(skb);
1106 out:
1107 	release_sock(sk);
1108 	return err;
1109 }
1110 
1111 static struct sk_buff *alloc_iucv_recv_skb(unsigned long len)
1112 {
1113 	size_t headroom, linear;
1114 	struct sk_buff *skb;
1115 	int err;
1116 
1117 	if (len < PAGE_SIZE) {
1118 		headroom = 0;
1119 		linear = len;
1120 	} else {
1121 		headroom = sizeof(struct iucv_array) * (MAX_SKB_FRAGS + 1);
1122 		linear = PAGE_SIZE - headroom;
1123 	}
1124 	skb = alloc_skb_with_frags(headroom + linear, len - linear,
1125 				   0, &err, GFP_ATOMIC | GFP_DMA);
1126 	WARN_ONCE(!skb,
1127 		  "alloc of recv iucv skb len=%lu failed with errcode=%d\n",
1128 		  len, err);
1129 	if (skb) {
1130 		if (headroom)
1131 			skb_reserve(skb, headroom);
1132 		skb_put(skb, linear);
1133 		skb->len = len;
1134 		skb->data_len = len - linear;
1135 	}
1136 	return skb;
1137 }
1138 
1139 /* iucv_process_message() - Receive a single outstanding IUCV message
1140  *
1141  * Locking: must be called with message_q.lock held
1142  */
1143 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1144 				 struct iucv_path *path,
1145 				 struct iucv_message *msg)
1146 {
1147 	int rc;
1148 	unsigned int len;
1149 
1150 	len = iucv_msg_length(msg);
1151 
1152 	/* store msg target class in the second 4 bytes of skb ctrl buffer */
1153 	/* Note: the first 4 bytes are reserved for msg tag */
1154 	IUCV_SKB_CB(skb)->class = msg->class;
1155 
1156 	/* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1157 	if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1158 		if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1159 			skb->data = NULL;
1160 			skb->len = 0;
1161 		}
1162 	} else {
1163 		if (skb_is_nonlinear(skb)) {
1164 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1165 			int i;
1166 
1167 			iba[0].address = (u32)(addr_t)skb->data;
1168 			iba[0].length = (u32)skb_headlen(skb);
1169 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1170 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1171 
1172 				iba[i + 1].address =
1173 					(u32)(addr_t)skb_frag_address(frag);
1174 				iba[i + 1].length = (u32)skb_frag_size(frag);
1175 			}
1176 			rc = pr_iucv->message_receive(path, msg,
1177 					      IUCV_IPBUFLST,
1178 					      (void *)iba, len, NULL);
1179 		} else {
1180 			rc = pr_iucv->message_receive(path, msg,
1181 					      msg->flags & IUCV_IPRMDATA,
1182 					      skb->data, len, NULL);
1183 		}
1184 		if (rc) {
1185 			kfree_skb(skb);
1186 			return;
1187 		}
1188 		WARN_ON_ONCE(skb->len != len);
1189 	}
1190 
1191 	IUCV_SKB_CB(skb)->offset = 0;
1192 	if (sk_filter(sk, skb)) {
1193 		atomic_inc(&sk->sk_drops);	/* skb rejected by filter */
1194 		kfree_skb(skb);
1195 		return;
1196 	}
1197 	if (__sock_queue_rcv_skb(sk, skb))	/* handle rcv queue full */
1198 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
1199 }
1200 
1201 /* iucv_process_message_q() - Process outstanding IUCV messages
1202  *
1203  * Locking: must be called with message_q.lock held
1204  */
1205 static void iucv_process_message_q(struct sock *sk)
1206 {
1207 	struct iucv_sock *iucv = iucv_sk(sk);
1208 	struct sk_buff *skb;
1209 	struct sock_msg_q *p, *n;
1210 
1211 	list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1212 		skb = alloc_iucv_recv_skb(iucv_msg_length(&p->msg));
1213 		if (!skb)
1214 			break;
1215 		iucv_process_message(sk, skb, p->path, &p->msg);
1216 		list_del(&p->list);
1217 		kfree(p);
1218 		if (!skb_queue_empty(&iucv->backlog_skb_q))
1219 			break;
1220 	}
1221 }
1222 
1223 static int iucv_sock_recvmsg(struct socket *sock, struct msghdr *msg,
1224 			     size_t len, int flags)
1225 {
1226 	int noblock = flags & MSG_DONTWAIT;
1227 	struct sock *sk = sock->sk;
1228 	struct iucv_sock *iucv = iucv_sk(sk);
1229 	unsigned int copied, rlen;
1230 	struct sk_buff *skb, *rskb, *cskb;
1231 	int err = 0;
1232 	u32 offset;
1233 
1234 	if ((sk->sk_state == IUCV_DISCONN) &&
1235 	    skb_queue_empty(&iucv->backlog_skb_q) &&
1236 	    skb_queue_empty(&sk->sk_receive_queue) &&
1237 	    list_empty(&iucv->message_q.list))
1238 		return 0;
1239 
1240 	if (flags & (MSG_OOB))
1241 		return -EOPNOTSUPP;
1242 
1243 	/* receive/dequeue next skb:
1244 	 * the function understands MSG_PEEK and, thus, does not dequeue skb */
1245 	skb = skb_recv_datagram(sk, flags, noblock, &err);
1246 	if (!skb) {
1247 		if (sk->sk_shutdown & RCV_SHUTDOWN)
1248 			return 0;
1249 		return err;
1250 	}
1251 
1252 	offset = IUCV_SKB_CB(skb)->offset;
1253 	rlen   = skb->len - offset;		/* real length of skb */
1254 	copied = min_t(unsigned int, rlen, len);
1255 	if (!rlen)
1256 		sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1257 
1258 	cskb = skb;
1259 	if (skb_copy_datagram_msg(cskb, offset, msg, copied)) {
1260 		if (!(flags & MSG_PEEK))
1261 			skb_queue_head(&sk->sk_receive_queue, skb);
1262 		return -EFAULT;
1263 	}
1264 
1265 	/* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1266 	if (sk->sk_type == SOCK_SEQPACKET) {
1267 		if (copied < rlen)
1268 			msg->msg_flags |= MSG_TRUNC;
1269 		/* each iucv message contains a complete record */
1270 		msg->msg_flags |= MSG_EOR;
1271 	}
1272 
1273 	/* create control message to store iucv msg target class:
1274 	 * get the trgcls from the control buffer of the skb due to
1275 	 * fragmentation of original iucv message. */
1276 	err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1277 		       sizeof(IUCV_SKB_CB(skb)->class),
1278 		       (void *)&IUCV_SKB_CB(skb)->class);
1279 	if (err) {
1280 		if (!(flags & MSG_PEEK))
1281 			skb_queue_head(&sk->sk_receive_queue, skb);
1282 		return err;
1283 	}
1284 
1285 	/* Mark read part of skb as used */
1286 	if (!(flags & MSG_PEEK)) {
1287 
1288 		/* SOCK_STREAM: re-queue skb if it contains unreceived data */
1289 		if (sk->sk_type == SOCK_STREAM) {
1290 			if (copied < rlen) {
1291 				IUCV_SKB_CB(skb)->offset = offset + copied;
1292 				skb_queue_head(&sk->sk_receive_queue, skb);
1293 				goto done;
1294 			}
1295 		}
1296 
1297 		consume_skb(skb);
1298 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1299 			atomic_inc(&iucv->msg_recv);
1300 			if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1301 				WARN_ON(1);
1302 				iucv_sock_close(sk);
1303 				return -EFAULT;
1304 			}
1305 		}
1306 
1307 		/* Queue backlog skbs */
1308 		spin_lock_bh(&iucv->message_q.lock);
1309 		rskb = skb_dequeue(&iucv->backlog_skb_q);
1310 		while (rskb) {
1311 			IUCV_SKB_CB(rskb)->offset = 0;
1312 			if (__sock_queue_rcv_skb(sk, rskb)) {
1313 				/* handle rcv queue full */
1314 				skb_queue_head(&iucv->backlog_skb_q,
1315 						rskb);
1316 				break;
1317 			}
1318 			rskb = skb_dequeue(&iucv->backlog_skb_q);
1319 		}
1320 		if (skb_queue_empty(&iucv->backlog_skb_q)) {
1321 			if (!list_empty(&iucv->message_q.list))
1322 				iucv_process_message_q(sk);
1323 			if (atomic_read(&iucv->msg_recv) >=
1324 							iucv->msglimit / 2) {
1325 				err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1326 				if (err) {
1327 					sk->sk_state = IUCV_DISCONN;
1328 					sk->sk_state_change(sk);
1329 				}
1330 			}
1331 		}
1332 		spin_unlock_bh(&iucv->message_q.lock);
1333 	}
1334 
1335 done:
1336 	/* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1337 	if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1338 		copied = rlen;
1339 
1340 	return copied;
1341 }
1342 
1343 static inline __poll_t iucv_accept_poll(struct sock *parent)
1344 {
1345 	struct iucv_sock *isk, *n;
1346 	struct sock *sk;
1347 
1348 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1349 		sk = (struct sock *) isk;
1350 
1351 		if (sk->sk_state == IUCV_CONNECTED)
1352 			return EPOLLIN | EPOLLRDNORM;
1353 	}
1354 
1355 	return 0;
1356 }
1357 
1358 static __poll_t iucv_sock_poll(struct file *file, struct socket *sock,
1359 			       poll_table *wait)
1360 {
1361 	struct sock *sk = sock->sk;
1362 	__poll_t mask = 0;
1363 
1364 	sock_poll_wait(file, sock, wait);
1365 
1366 	if (sk->sk_state == IUCV_LISTEN)
1367 		return iucv_accept_poll(sk);
1368 
1369 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1370 		mask |= EPOLLERR |
1371 			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
1372 
1373 	if (sk->sk_shutdown & RCV_SHUTDOWN)
1374 		mask |= EPOLLRDHUP;
1375 
1376 	if (sk->sk_shutdown == SHUTDOWN_MASK)
1377 		mask |= EPOLLHUP;
1378 
1379 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
1380 	    (sk->sk_shutdown & RCV_SHUTDOWN))
1381 		mask |= EPOLLIN | EPOLLRDNORM;
1382 
1383 	if (sk->sk_state == IUCV_CLOSED)
1384 		mask |= EPOLLHUP;
1385 
1386 	if (sk->sk_state == IUCV_DISCONN)
1387 		mask |= EPOLLIN;
1388 
1389 	if (sock_writeable(sk) && iucv_below_msglim(sk))
1390 		mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1391 	else
1392 		sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1393 
1394 	return mask;
1395 }
1396 
1397 static int iucv_sock_shutdown(struct socket *sock, int how)
1398 {
1399 	struct sock *sk = sock->sk;
1400 	struct iucv_sock *iucv = iucv_sk(sk);
1401 	struct iucv_message txmsg;
1402 	int err = 0;
1403 
1404 	how++;
1405 
1406 	if ((how & ~SHUTDOWN_MASK) || !how)
1407 		return -EINVAL;
1408 
1409 	lock_sock(sk);
1410 	switch (sk->sk_state) {
1411 	case IUCV_LISTEN:
1412 	case IUCV_DISCONN:
1413 	case IUCV_CLOSING:
1414 	case IUCV_CLOSED:
1415 		err = -ENOTCONN;
1416 		goto fail;
1417 	default:
1418 		break;
1419 	}
1420 
1421 	if ((how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) &&
1422 	    sk->sk_state == IUCV_CONNECTED) {
1423 		if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1424 			txmsg.class = 0;
1425 			txmsg.tag = 0;
1426 			err = pr_iucv->message_send(iucv->path, &txmsg,
1427 				IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1428 			if (err) {
1429 				switch (err) {
1430 				case 1:
1431 					err = -ENOTCONN;
1432 					break;
1433 				case 2:
1434 					err = -ECONNRESET;
1435 					break;
1436 				default:
1437 					err = -ENOTCONN;
1438 					break;
1439 				}
1440 			}
1441 		} else
1442 			iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1443 	}
1444 
1445 	sk->sk_shutdown |= how;
1446 	if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1447 		if ((iucv->transport == AF_IUCV_TRANS_IUCV) &&
1448 		    iucv->path) {
1449 			err = pr_iucv->path_quiesce(iucv->path, NULL);
1450 			if (err)
1451 				err = -ENOTCONN;
1452 /*			skb_queue_purge(&sk->sk_receive_queue); */
1453 		}
1454 		skb_queue_purge(&sk->sk_receive_queue);
1455 	}
1456 
1457 	/* Wake up anyone sleeping in poll */
1458 	sk->sk_state_change(sk);
1459 
1460 fail:
1461 	release_sock(sk);
1462 	return err;
1463 }
1464 
1465 static int iucv_sock_release(struct socket *sock)
1466 {
1467 	struct sock *sk = sock->sk;
1468 	int err = 0;
1469 
1470 	if (!sk)
1471 		return 0;
1472 
1473 	iucv_sock_close(sk);
1474 
1475 	sock_orphan(sk);
1476 	iucv_sock_kill(sk);
1477 	return err;
1478 }
1479 
1480 /* getsockopt and setsockopt */
1481 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1482 				sockptr_t optval, unsigned int optlen)
1483 {
1484 	struct sock *sk = sock->sk;
1485 	struct iucv_sock *iucv = iucv_sk(sk);
1486 	int val;
1487 	int rc;
1488 
1489 	if (level != SOL_IUCV)
1490 		return -ENOPROTOOPT;
1491 
1492 	if (optlen < sizeof(int))
1493 		return -EINVAL;
1494 
1495 	if (copy_from_sockptr(&val, optval, sizeof(int)))
1496 		return -EFAULT;
1497 
1498 	rc = 0;
1499 
1500 	lock_sock(sk);
1501 	switch (optname) {
1502 	case SO_IPRMDATA_MSG:
1503 		if (val)
1504 			iucv->flags |= IUCV_IPRMDATA;
1505 		else
1506 			iucv->flags &= ~IUCV_IPRMDATA;
1507 		break;
1508 	case SO_MSGLIMIT:
1509 		switch (sk->sk_state) {
1510 		case IUCV_OPEN:
1511 		case IUCV_BOUND:
1512 			if (val < 1 || val > U16_MAX)
1513 				rc = -EINVAL;
1514 			else
1515 				iucv->msglimit = val;
1516 			break;
1517 		default:
1518 			rc = -EINVAL;
1519 			break;
1520 		}
1521 		break;
1522 	default:
1523 		rc = -ENOPROTOOPT;
1524 		break;
1525 	}
1526 	release_sock(sk);
1527 
1528 	return rc;
1529 }
1530 
1531 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1532 				char __user *optval, int __user *optlen)
1533 {
1534 	struct sock *sk = sock->sk;
1535 	struct iucv_sock *iucv = iucv_sk(sk);
1536 	unsigned int val;
1537 	int len;
1538 
1539 	if (level != SOL_IUCV)
1540 		return -ENOPROTOOPT;
1541 
1542 	if (get_user(len, optlen))
1543 		return -EFAULT;
1544 
1545 	if (len < 0)
1546 		return -EINVAL;
1547 
1548 	len = min_t(unsigned int, len, sizeof(int));
1549 
1550 	switch (optname) {
1551 	case SO_IPRMDATA_MSG:
1552 		val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1553 		break;
1554 	case SO_MSGLIMIT:
1555 		lock_sock(sk);
1556 		val = (iucv->path != NULL) ? iucv->path->msglim	/* connected */
1557 					   : iucv->msglimit;	/* default */
1558 		release_sock(sk);
1559 		break;
1560 	case SO_MSGSIZE:
1561 		if (sk->sk_state == IUCV_OPEN)
1562 			return -EBADFD;
1563 		val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1564 				sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1565 				0x7fffffff;
1566 		break;
1567 	default:
1568 		return -ENOPROTOOPT;
1569 	}
1570 
1571 	if (put_user(len, optlen))
1572 		return -EFAULT;
1573 	if (copy_to_user(optval, &val, len))
1574 		return -EFAULT;
1575 
1576 	return 0;
1577 }
1578 
1579 
1580 /* Callback wrappers - called from iucv base support */
1581 static int iucv_callback_connreq(struct iucv_path *path,
1582 				 u8 ipvmid[8], u8 ipuser[16])
1583 {
1584 	unsigned char user_data[16];
1585 	unsigned char nuser_data[16];
1586 	unsigned char src_name[8];
1587 	struct sock *sk, *nsk;
1588 	struct iucv_sock *iucv, *niucv;
1589 	int err;
1590 
1591 	memcpy(src_name, ipuser, 8);
1592 	EBCASC(src_name, 8);
1593 	/* Find out if this path belongs to af_iucv. */
1594 	read_lock(&iucv_sk_list.lock);
1595 	iucv = NULL;
1596 	sk = NULL;
1597 	sk_for_each(sk, &iucv_sk_list.head)
1598 		if (sk->sk_state == IUCV_LISTEN &&
1599 		    !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1600 			/*
1601 			 * Found a listening socket with
1602 			 * src_name == ipuser[0-7].
1603 			 */
1604 			iucv = iucv_sk(sk);
1605 			break;
1606 		}
1607 	read_unlock(&iucv_sk_list.lock);
1608 	if (!iucv)
1609 		/* No socket found, not one of our paths. */
1610 		return -EINVAL;
1611 
1612 	bh_lock_sock(sk);
1613 
1614 	/* Check if parent socket is listening */
1615 	low_nmcpy(user_data, iucv->src_name);
1616 	high_nmcpy(user_data, iucv->dst_name);
1617 	ASCEBC(user_data, sizeof(user_data));
1618 	if (sk->sk_state != IUCV_LISTEN) {
1619 		err = pr_iucv->path_sever(path, user_data);
1620 		iucv_path_free(path);
1621 		goto fail;
1622 	}
1623 
1624 	/* Check for backlog size */
1625 	if (sk_acceptq_is_full(sk)) {
1626 		err = pr_iucv->path_sever(path, user_data);
1627 		iucv_path_free(path);
1628 		goto fail;
1629 	}
1630 
1631 	/* Create the new socket */
1632 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1633 	if (!nsk) {
1634 		err = pr_iucv->path_sever(path, user_data);
1635 		iucv_path_free(path);
1636 		goto fail;
1637 	}
1638 
1639 	niucv = iucv_sk(nsk);
1640 	iucv_sock_init(nsk, sk);
1641 	niucv->transport = AF_IUCV_TRANS_IUCV;
1642 	nsk->sk_allocation |= GFP_DMA;
1643 
1644 	/* Set the new iucv_sock */
1645 	memcpy(niucv->dst_name, ipuser + 8, 8);
1646 	EBCASC(niucv->dst_name, 8);
1647 	memcpy(niucv->dst_user_id, ipvmid, 8);
1648 	memcpy(niucv->src_name, iucv->src_name, 8);
1649 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1650 	niucv->path = path;
1651 
1652 	/* Call iucv_accept */
1653 	high_nmcpy(nuser_data, ipuser + 8);
1654 	memcpy(nuser_data + 8, niucv->src_name, 8);
1655 	ASCEBC(nuser_data + 8, 8);
1656 
1657 	/* set message limit for path based on msglimit of accepting socket */
1658 	niucv->msglimit = iucv->msglimit;
1659 	path->msglim = iucv->msglimit;
1660 	err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1661 	if (err) {
1662 		iucv_sever_path(nsk, 1);
1663 		iucv_sock_kill(nsk);
1664 		goto fail;
1665 	}
1666 
1667 	iucv_accept_enqueue(sk, nsk);
1668 
1669 	/* Wake up accept */
1670 	nsk->sk_state = IUCV_CONNECTED;
1671 	sk->sk_data_ready(sk);
1672 	err = 0;
1673 fail:
1674 	bh_unlock_sock(sk);
1675 	return 0;
1676 }
1677 
1678 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1679 {
1680 	struct sock *sk = path->private;
1681 
1682 	sk->sk_state = IUCV_CONNECTED;
1683 	sk->sk_state_change(sk);
1684 }
1685 
1686 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1687 {
1688 	struct sock *sk = path->private;
1689 	struct iucv_sock *iucv = iucv_sk(sk);
1690 	struct sk_buff *skb;
1691 	struct sock_msg_q *save_msg;
1692 	int len;
1693 
1694 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1695 		pr_iucv->message_reject(path, msg);
1696 		return;
1697 	}
1698 
1699 	spin_lock(&iucv->message_q.lock);
1700 
1701 	if (!list_empty(&iucv->message_q.list) ||
1702 	    !skb_queue_empty(&iucv->backlog_skb_q))
1703 		goto save_message;
1704 
1705 	len = atomic_read(&sk->sk_rmem_alloc);
1706 	len += SKB_TRUESIZE(iucv_msg_length(msg));
1707 	if (len > sk->sk_rcvbuf)
1708 		goto save_message;
1709 
1710 	skb = alloc_iucv_recv_skb(iucv_msg_length(msg));
1711 	if (!skb)
1712 		goto save_message;
1713 
1714 	iucv_process_message(sk, skb, path, msg);
1715 	goto out_unlock;
1716 
1717 save_message:
1718 	save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1719 	if (!save_msg)
1720 		goto out_unlock;
1721 	save_msg->path = path;
1722 	save_msg->msg = *msg;
1723 
1724 	list_add_tail(&save_msg->list, &iucv->message_q.list);
1725 
1726 out_unlock:
1727 	spin_unlock(&iucv->message_q.lock);
1728 }
1729 
1730 static void iucv_callback_txdone(struct iucv_path *path,
1731 				 struct iucv_message *msg)
1732 {
1733 	struct sock *sk = path->private;
1734 	struct sk_buff *this = NULL;
1735 	struct sk_buff_head *list;
1736 	struct sk_buff *list_skb;
1737 	struct iucv_sock *iucv;
1738 	unsigned long flags;
1739 
1740 	iucv = iucv_sk(sk);
1741 	list = &iucv->send_skb_q;
1742 
1743 	bh_lock_sock(sk);
1744 
1745 	spin_lock_irqsave(&list->lock, flags);
1746 	skb_queue_walk(list, list_skb) {
1747 		if (msg->tag == IUCV_SKB_CB(list_skb)->tag) {
1748 			this = list_skb;
1749 			break;
1750 		}
1751 	}
1752 	if (this) {
1753 		atomic_dec(&iucv->skbs_in_xmit);
1754 		__skb_unlink(this, list);
1755 	}
1756 
1757 	spin_unlock_irqrestore(&list->lock, flags);
1758 
1759 	if (this) {
1760 		consume_skb(this);
1761 		/* wake up any process waiting for sending */
1762 		iucv_sock_wake_msglim(sk);
1763 	}
1764 
1765 	if (sk->sk_state == IUCV_CLOSING) {
1766 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
1767 			sk->sk_state = IUCV_CLOSED;
1768 			sk->sk_state_change(sk);
1769 		}
1770 	}
1771 	bh_unlock_sock(sk);
1772 
1773 }
1774 
1775 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1776 {
1777 	struct sock *sk = path->private;
1778 
1779 	if (sk->sk_state == IUCV_CLOSED)
1780 		return;
1781 
1782 	bh_lock_sock(sk);
1783 	iucv_sever_path(sk, 1);
1784 	sk->sk_state = IUCV_DISCONN;
1785 
1786 	sk->sk_state_change(sk);
1787 	bh_unlock_sock(sk);
1788 }
1789 
1790 /* called if the other communication side shuts down its RECV direction;
1791  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1792  */
1793 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1794 {
1795 	struct sock *sk = path->private;
1796 
1797 	bh_lock_sock(sk);
1798 	if (sk->sk_state != IUCV_CLOSED) {
1799 		sk->sk_shutdown |= SEND_SHUTDOWN;
1800 		sk->sk_state_change(sk);
1801 	}
1802 	bh_unlock_sock(sk);
1803 }
1804 
1805 static struct iucv_handler af_iucv_handler = {
1806 	.path_pending		= iucv_callback_connreq,
1807 	.path_complete		= iucv_callback_connack,
1808 	.path_severed		= iucv_callback_connrej,
1809 	.message_pending	= iucv_callback_rx,
1810 	.message_complete	= iucv_callback_txdone,
1811 	.path_quiesced		= iucv_callback_shutdown,
1812 };
1813 
1814 /***************** HiperSockets transport callbacks ********************/
1815 static void afiucv_swap_src_dest(struct sk_buff *skb)
1816 {
1817 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1818 	char tmpID[8];
1819 	char tmpName[8];
1820 
1821 	ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1822 	ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1823 	ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1824 	ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1825 	memcpy(tmpID, trans_hdr->srcUserID, 8);
1826 	memcpy(tmpName, trans_hdr->srcAppName, 8);
1827 	memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1828 	memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1829 	memcpy(trans_hdr->destUserID, tmpID, 8);
1830 	memcpy(trans_hdr->destAppName, tmpName, 8);
1831 	skb_push(skb, ETH_HLEN);
1832 	memset(skb->data, 0, ETH_HLEN);
1833 }
1834 
1835 /*
1836  * afiucv_hs_callback_syn - react on received SYN
1837  */
1838 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1839 {
1840 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1841 	struct sock *nsk;
1842 	struct iucv_sock *iucv, *niucv;
1843 	int err;
1844 
1845 	iucv = iucv_sk(sk);
1846 	if (!iucv) {
1847 		/* no sock - connection refused */
1848 		afiucv_swap_src_dest(skb);
1849 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1850 		err = dev_queue_xmit(skb);
1851 		goto out;
1852 	}
1853 
1854 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1855 	bh_lock_sock(sk);
1856 	if ((sk->sk_state != IUCV_LISTEN) ||
1857 	    sk_acceptq_is_full(sk) ||
1858 	    !nsk) {
1859 		/* error on server socket - connection refused */
1860 		afiucv_swap_src_dest(skb);
1861 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1862 		err = dev_queue_xmit(skb);
1863 		iucv_sock_kill(nsk);
1864 		bh_unlock_sock(sk);
1865 		goto out;
1866 	}
1867 
1868 	niucv = iucv_sk(nsk);
1869 	iucv_sock_init(nsk, sk);
1870 	niucv->transport = AF_IUCV_TRANS_HIPER;
1871 	niucv->msglimit = iucv->msglimit;
1872 	if (!trans_hdr->window)
1873 		niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1874 	else
1875 		niucv->msglimit_peer = trans_hdr->window;
1876 	memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1877 	memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1878 	memcpy(niucv->src_name, iucv->src_name, 8);
1879 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1880 	nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1881 	niucv->hs_dev = iucv->hs_dev;
1882 	dev_hold(niucv->hs_dev);
1883 	afiucv_swap_src_dest(skb);
1884 	trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1885 	trans_hdr->window = niucv->msglimit;
1886 	/* if receiver acks the xmit connection is established */
1887 	err = dev_queue_xmit(skb);
1888 	if (!err) {
1889 		iucv_accept_enqueue(sk, nsk);
1890 		nsk->sk_state = IUCV_CONNECTED;
1891 		sk->sk_data_ready(sk);
1892 	} else
1893 		iucv_sock_kill(nsk);
1894 	bh_unlock_sock(sk);
1895 
1896 out:
1897 	return NET_RX_SUCCESS;
1898 }
1899 
1900 /*
1901  * afiucv_hs_callback_synack() - react on received SYN-ACK
1902  */
1903 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1904 {
1905 	struct iucv_sock *iucv = iucv_sk(sk);
1906 
1907 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1908 		kfree_skb(skb);
1909 		return NET_RX_SUCCESS;
1910 	}
1911 
1912 	bh_lock_sock(sk);
1913 	iucv->msglimit_peer = iucv_trans_hdr(skb)->window;
1914 	sk->sk_state = IUCV_CONNECTED;
1915 	sk->sk_state_change(sk);
1916 	bh_unlock_sock(sk);
1917 	consume_skb(skb);
1918 	return NET_RX_SUCCESS;
1919 }
1920 
1921 /*
1922  * afiucv_hs_callback_synfin() - react on received SYN_FIN
1923  */
1924 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
1925 {
1926 	struct iucv_sock *iucv = iucv_sk(sk);
1927 
1928 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1929 		kfree_skb(skb);
1930 		return NET_RX_SUCCESS;
1931 	}
1932 
1933 	bh_lock_sock(sk);
1934 	sk->sk_state = IUCV_DISCONN;
1935 	sk->sk_state_change(sk);
1936 	bh_unlock_sock(sk);
1937 	consume_skb(skb);
1938 	return NET_RX_SUCCESS;
1939 }
1940 
1941 /*
1942  * afiucv_hs_callback_fin() - react on received FIN
1943  */
1944 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
1945 {
1946 	struct iucv_sock *iucv = iucv_sk(sk);
1947 
1948 	/* other end of connection closed */
1949 	if (!iucv) {
1950 		kfree_skb(skb);
1951 		return NET_RX_SUCCESS;
1952 	}
1953 
1954 	bh_lock_sock(sk);
1955 	if (sk->sk_state == IUCV_CONNECTED) {
1956 		sk->sk_state = IUCV_DISCONN;
1957 		sk->sk_state_change(sk);
1958 	}
1959 	bh_unlock_sock(sk);
1960 	consume_skb(skb);
1961 	return NET_RX_SUCCESS;
1962 }
1963 
1964 /*
1965  * afiucv_hs_callback_win() - react on received WIN
1966  */
1967 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
1968 {
1969 	struct iucv_sock *iucv = iucv_sk(sk);
1970 
1971 	if (!iucv)
1972 		return NET_RX_SUCCESS;
1973 
1974 	if (sk->sk_state != IUCV_CONNECTED)
1975 		return NET_RX_SUCCESS;
1976 
1977 	atomic_sub(iucv_trans_hdr(skb)->window, &iucv->msg_sent);
1978 	iucv_sock_wake_msglim(sk);
1979 	return NET_RX_SUCCESS;
1980 }
1981 
1982 /*
1983  * afiucv_hs_callback_rx() - react on received data
1984  */
1985 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
1986 {
1987 	struct iucv_sock *iucv = iucv_sk(sk);
1988 
1989 	if (!iucv) {
1990 		kfree_skb(skb);
1991 		return NET_RX_SUCCESS;
1992 	}
1993 
1994 	if (sk->sk_state != IUCV_CONNECTED) {
1995 		kfree_skb(skb);
1996 		return NET_RX_SUCCESS;
1997 	}
1998 
1999 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2000 		kfree_skb(skb);
2001 		return NET_RX_SUCCESS;
2002 	}
2003 
2004 	/* write stuff from iucv_msg to skb cb */
2005 	skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2006 	skb_reset_transport_header(skb);
2007 	skb_reset_network_header(skb);
2008 	IUCV_SKB_CB(skb)->offset = 0;
2009 	if (sk_filter(sk, skb)) {
2010 		atomic_inc(&sk->sk_drops);	/* skb rejected by filter */
2011 		kfree_skb(skb);
2012 		return NET_RX_SUCCESS;
2013 	}
2014 
2015 	spin_lock(&iucv->message_q.lock);
2016 	if (skb_queue_empty(&iucv->backlog_skb_q)) {
2017 		if (__sock_queue_rcv_skb(sk, skb))
2018 			/* handle rcv queue full */
2019 			skb_queue_tail(&iucv->backlog_skb_q, skb);
2020 	} else
2021 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2022 	spin_unlock(&iucv->message_q.lock);
2023 	return NET_RX_SUCCESS;
2024 }
2025 
2026 /*
2027  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2028  *                   transport
2029  *                   called from netif RX softirq
2030  */
2031 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2032 	struct packet_type *pt, struct net_device *orig_dev)
2033 {
2034 	struct sock *sk;
2035 	struct iucv_sock *iucv;
2036 	struct af_iucv_trans_hdr *trans_hdr;
2037 	int err = NET_RX_SUCCESS;
2038 	char nullstring[8];
2039 
2040 	if (!pskb_may_pull(skb, sizeof(*trans_hdr))) {
2041 		kfree_skb(skb);
2042 		return NET_RX_SUCCESS;
2043 	}
2044 
2045 	trans_hdr = iucv_trans_hdr(skb);
2046 	EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2047 	EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2048 	EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2049 	EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2050 	memset(nullstring, 0, sizeof(nullstring));
2051 	iucv = NULL;
2052 	sk = NULL;
2053 	read_lock(&iucv_sk_list.lock);
2054 	sk_for_each(sk, &iucv_sk_list.head) {
2055 		if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2056 			if ((!memcmp(&iucv_sk(sk)->src_name,
2057 				     trans_hdr->destAppName, 8)) &&
2058 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2059 				     trans_hdr->destUserID, 8)) &&
2060 			    (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2061 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2062 				     nullstring, 8))) {
2063 				iucv = iucv_sk(sk);
2064 				break;
2065 			}
2066 		} else {
2067 			if ((!memcmp(&iucv_sk(sk)->src_name,
2068 				     trans_hdr->destAppName, 8)) &&
2069 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2070 				     trans_hdr->destUserID, 8)) &&
2071 			    (!memcmp(&iucv_sk(sk)->dst_name,
2072 				     trans_hdr->srcAppName, 8)) &&
2073 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2074 				     trans_hdr->srcUserID, 8))) {
2075 				iucv = iucv_sk(sk);
2076 				break;
2077 			}
2078 		}
2079 	}
2080 	read_unlock(&iucv_sk_list.lock);
2081 	if (!iucv)
2082 		sk = NULL;
2083 
2084 	/* no sock
2085 	how should we send with no sock
2086 	1) send without sock no send rc checking?
2087 	2) introduce default sock to handle this cases
2088 
2089 	 SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2090 	 data -> send FIN
2091 	 SYN|ACK, SYN|FIN, FIN -> no action? */
2092 
2093 	switch (trans_hdr->flags) {
2094 	case AF_IUCV_FLAG_SYN:
2095 		/* connect request */
2096 		err = afiucv_hs_callback_syn(sk, skb);
2097 		break;
2098 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2099 		/* connect request confirmed */
2100 		err = afiucv_hs_callback_synack(sk, skb);
2101 		break;
2102 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2103 		/* connect request refused */
2104 		err = afiucv_hs_callback_synfin(sk, skb);
2105 		break;
2106 	case (AF_IUCV_FLAG_FIN):
2107 		/* close request */
2108 		err = afiucv_hs_callback_fin(sk, skb);
2109 		break;
2110 	case (AF_IUCV_FLAG_WIN):
2111 		err = afiucv_hs_callback_win(sk, skb);
2112 		if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2113 			consume_skb(skb);
2114 			break;
2115 		}
2116 		fallthrough;	/* and receive non-zero length data */
2117 	case (AF_IUCV_FLAG_SHT):
2118 		/* shutdown request */
2119 		fallthrough;	/* and receive zero length data */
2120 	case 0:
2121 		/* plain data frame */
2122 		IUCV_SKB_CB(skb)->class = trans_hdr->iucv_hdr.class;
2123 		err = afiucv_hs_callback_rx(sk, skb);
2124 		break;
2125 	default:
2126 		kfree_skb(skb);
2127 	}
2128 
2129 	return err;
2130 }
2131 
2132 /*
2133  * afiucv_hs_callback_txnotify() - handle send notifications from HiperSockets
2134  *                                 transport
2135  */
2136 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify n)
2137 {
2138 	struct iucv_sock *iucv = iucv_sk(sk);
2139 
2140 	if (sock_flag(sk, SOCK_ZAPPED))
2141 		return;
2142 
2143 	switch (n) {
2144 	case TX_NOTIFY_OK:
2145 		atomic_dec(&iucv->skbs_in_xmit);
2146 		iucv_sock_wake_msglim(sk);
2147 		break;
2148 	case TX_NOTIFY_PENDING:
2149 		atomic_inc(&iucv->pendings);
2150 		break;
2151 	case TX_NOTIFY_DELAYED_OK:
2152 		atomic_dec(&iucv->skbs_in_xmit);
2153 		if (atomic_dec_return(&iucv->pendings) <= 0)
2154 			iucv_sock_wake_msglim(sk);
2155 		break;
2156 	default:
2157 		atomic_dec(&iucv->skbs_in_xmit);
2158 		if (sk->sk_state == IUCV_CONNECTED) {
2159 			sk->sk_state = IUCV_DISCONN;
2160 			sk->sk_state_change(sk);
2161 		}
2162 	}
2163 
2164 	if (sk->sk_state == IUCV_CLOSING) {
2165 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
2166 			sk->sk_state = IUCV_CLOSED;
2167 			sk->sk_state_change(sk);
2168 		}
2169 	}
2170 }
2171 
2172 /*
2173  * afiucv_netdev_event: handle netdev notifier chain events
2174  */
2175 static int afiucv_netdev_event(struct notifier_block *this,
2176 			       unsigned long event, void *ptr)
2177 {
2178 	struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
2179 	struct sock *sk;
2180 	struct iucv_sock *iucv;
2181 
2182 	switch (event) {
2183 	case NETDEV_REBOOT:
2184 	case NETDEV_GOING_DOWN:
2185 		sk_for_each(sk, &iucv_sk_list.head) {
2186 			iucv = iucv_sk(sk);
2187 			if ((iucv->hs_dev == event_dev) &&
2188 			    (sk->sk_state == IUCV_CONNECTED)) {
2189 				if (event == NETDEV_GOING_DOWN)
2190 					iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2191 				sk->sk_state = IUCV_DISCONN;
2192 				sk->sk_state_change(sk);
2193 			}
2194 		}
2195 		break;
2196 	case NETDEV_DOWN:
2197 	case NETDEV_UNREGISTER:
2198 	default:
2199 		break;
2200 	}
2201 	return NOTIFY_DONE;
2202 }
2203 
2204 static struct notifier_block afiucv_netdev_notifier = {
2205 	.notifier_call = afiucv_netdev_event,
2206 };
2207 
2208 static const struct proto_ops iucv_sock_ops = {
2209 	.family		= PF_IUCV,
2210 	.owner		= THIS_MODULE,
2211 	.release	= iucv_sock_release,
2212 	.bind		= iucv_sock_bind,
2213 	.connect	= iucv_sock_connect,
2214 	.listen		= iucv_sock_listen,
2215 	.accept		= iucv_sock_accept,
2216 	.getname	= iucv_sock_getname,
2217 	.sendmsg	= iucv_sock_sendmsg,
2218 	.recvmsg	= iucv_sock_recvmsg,
2219 	.poll		= iucv_sock_poll,
2220 	.ioctl		= sock_no_ioctl,
2221 	.mmap		= sock_no_mmap,
2222 	.socketpair	= sock_no_socketpair,
2223 	.shutdown	= iucv_sock_shutdown,
2224 	.setsockopt	= iucv_sock_setsockopt,
2225 	.getsockopt	= iucv_sock_getsockopt,
2226 };
2227 
2228 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
2229 			    int kern)
2230 {
2231 	struct sock *sk;
2232 
2233 	if (protocol && protocol != PF_IUCV)
2234 		return -EPROTONOSUPPORT;
2235 
2236 	sock->state = SS_UNCONNECTED;
2237 
2238 	switch (sock->type) {
2239 	case SOCK_STREAM:
2240 	case SOCK_SEQPACKET:
2241 		/* currently, proto ops can handle both sk types */
2242 		sock->ops = &iucv_sock_ops;
2243 		break;
2244 	default:
2245 		return -ESOCKTNOSUPPORT;
2246 	}
2247 
2248 	sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL, kern);
2249 	if (!sk)
2250 		return -ENOMEM;
2251 
2252 	iucv_sock_init(sk, NULL);
2253 
2254 	return 0;
2255 }
2256 
2257 static const struct net_proto_family iucv_sock_family_ops = {
2258 	.family	= AF_IUCV,
2259 	.owner	= THIS_MODULE,
2260 	.create	= iucv_sock_create,
2261 };
2262 
2263 static struct packet_type iucv_packet_type = {
2264 	.type = cpu_to_be16(ETH_P_AF_IUCV),
2265 	.func = afiucv_hs_rcv,
2266 };
2267 
2268 static int __init afiucv_init(void)
2269 {
2270 	int err;
2271 
2272 	if (MACHINE_IS_VM && IS_ENABLED(CONFIG_IUCV)) {
2273 		cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2274 		if (unlikely(err)) {
2275 			WARN_ON(err);
2276 			err = -EPROTONOSUPPORT;
2277 			goto out;
2278 		}
2279 
2280 		pr_iucv = &iucv_if;
2281 	} else {
2282 		memset(&iucv_userid, 0, sizeof(iucv_userid));
2283 		pr_iucv = NULL;
2284 	}
2285 
2286 	err = proto_register(&iucv_proto, 0);
2287 	if (err)
2288 		goto out;
2289 	err = sock_register(&iucv_sock_family_ops);
2290 	if (err)
2291 		goto out_proto;
2292 
2293 	if (pr_iucv) {
2294 		err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2295 		if (err)
2296 			goto out_sock;
2297 	}
2298 
2299 	err = register_netdevice_notifier(&afiucv_netdev_notifier);
2300 	if (err)
2301 		goto out_notifier;
2302 
2303 	dev_add_pack(&iucv_packet_type);
2304 	return 0;
2305 
2306 out_notifier:
2307 	if (pr_iucv)
2308 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2309 out_sock:
2310 	sock_unregister(PF_IUCV);
2311 out_proto:
2312 	proto_unregister(&iucv_proto);
2313 out:
2314 	return err;
2315 }
2316 
2317 static void __exit afiucv_exit(void)
2318 {
2319 	if (pr_iucv)
2320 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2321 
2322 	unregister_netdevice_notifier(&afiucv_netdev_notifier);
2323 	dev_remove_pack(&iucv_packet_type);
2324 	sock_unregister(PF_IUCV);
2325 	proto_unregister(&iucv_proto);
2326 }
2327 
2328 module_init(afiucv_init);
2329 module_exit(afiucv_exit);
2330 
2331 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2332 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2333 MODULE_VERSION(VERSION);
2334 MODULE_LICENSE("GPL");
2335 MODULE_ALIAS_NETPROTO(PF_IUCV);
2336