xref: /linux/net/iucv/af_iucv.c (revision f1359c240191e686614847905fc861cbda480b47)
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 pr_fmt(fmt) "af_iucv: " fmt
14 
15 #include <linux/filter.h>
16 #include <linux/module.h>
17 #include <linux/netdevice.h>
18 #include <linux/types.h>
19 #include <linux/limits.h>
20 #include <linux/list.h>
21 #include <linux/errno.h>
22 #include <linux/kernel.h>
23 #include <linux/sched/signal.h>
24 #include <linux/slab.h>
25 #include <linux/skbuff.h>
26 #include <linux/init.h>
27 #include <linux/poll.h>
28 #include <linux/security.h>
29 #include <net/sock.h>
30 #include <asm/machine.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  * Returns: Length of the IUCV message.
132  */
133 static inline size_t iucv_msg_length(struct iucv_message *msg)
134 {
135 	size_t datalen;
136 
137 	if (msg->flags & IUCV_IPRMDATA) {
138 		datalen = 0xff - msg->rmmsg[7];
139 		return (datalen < 8) ? datalen : 8;
140 	}
141 	return msg->length;
142 }
143 
144 /**
145  * iucv_sock_in_state() - check for specific states
146  * @sk:		sock structure
147  * @state:	first iucv sk state
148  * @state2:	second iucv sk state
149  *
150  * Returns: true if the socket is either in the first or second state.
151  */
152 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
153 {
154 	return (sk->sk_state == state || sk->sk_state == state2);
155 }
156 
157 /**
158  * iucv_below_msglim() - function to check if messages can be sent
159  * @sk:		sock structure
160  *
161  * Returns: true, if either the socket is not connected (no iucv path for
162  * checking the message limit) or if the send queue length is lower
163  * than the message limit.
164  */
165 static inline int iucv_below_msglim(struct sock *sk)
166 {
167 	struct iucv_sock *iucv = iucv_sk(sk);
168 
169 	if (sk->sk_state != IUCV_CONNECTED)
170 		return 1;
171 	if (iucv->transport == AF_IUCV_TRANS_IUCV)
172 		return (atomic_read(&iucv->skbs_in_xmit) < iucv->path->msglim);
173 	else
174 		return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
175 			(atomic_read(&iucv->pendings) <= 0));
176 }
177 
178 /*
179  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
180  */
181 static void iucv_sock_wake_msglim(struct sock *sk)
182 {
183 	struct socket_wq *wq;
184 
185 	rcu_read_lock();
186 	wq = rcu_dereference(sk->sk_wq);
187 	if (skwq_has_sleeper(wq))
188 		wake_up_interruptible_all(&wq->wait);
189 	sk_wake_async_rcu(sk, SOCK_WAKE_SPACE, POLL_OUT);
190 	rcu_read_unlock();
191 }
192 
193 /*
194  * afiucv_hs_send() - send a message through HiperSockets transport
195  */
196 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
197 		   struct sk_buff *skb, u8 flags)
198 {
199 	struct iucv_sock *iucv = iucv_sk(sock);
200 	struct af_iucv_trans_hdr *phs_hdr;
201 	int err, confirm_recv = 0;
202 
203 	phs_hdr = skb_push(skb, sizeof(*phs_hdr));
204 	memset(phs_hdr, 0, sizeof(*phs_hdr));
205 	skb_reset_network_header(skb);
206 
207 	phs_hdr->magic = ETH_P_AF_IUCV;
208 	phs_hdr->version = 1;
209 	phs_hdr->flags = flags;
210 	if (flags == AF_IUCV_FLAG_SYN)
211 		phs_hdr->window = iucv->msglimit;
212 	else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
213 		confirm_recv = atomic_read(&iucv->msg_recv);
214 		phs_hdr->window = confirm_recv;
215 		if (confirm_recv)
216 			phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
217 	}
218 	memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
219 	memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
220 	memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
221 	memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
222 	ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
223 	ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
224 	ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
225 	ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
226 	if (imsg)
227 		memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
228 
229 	skb->dev = iucv->hs_dev;
230 	if (!skb->dev) {
231 		err = -ENODEV;
232 		goto err_free;
233 	}
234 
235 	dev_hard_header(skb, skb->dev, ETH_P_AF_IUCV, NULL, NULL, skb->len);
236 
237 	if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev)) {
238 		err = -ENETDOWN;
239 		goto err_free;
240 	}
241 	if (skb->len > skb->dev->mtu) {
242 		if (sock->sk_type == SOCK_SEQPACKET) {
243 			err = -EMSGSIZE;
244 			goto err_free;
245 		}
246 		err = pskb_trim(skb, skb->dev->mtu);
247 		if (err)
248 			goto err_free;
249 	}
250 	skb->protocol = cpu_to_be16(ETH_P_AF_IUCV);
251 
252 	atomic_inc(&iucv->skbs_in_xmit);
253 	err = dev_queue_xmit(skb);
254 	if (net_xmit_eval(err)) {
255 		atomic_dec(&iucv->skbs_in_xmit);
256 	} else {
257 		atomic_sub(confirm_recv, &iucv->msg_recv);
258 		WARN_ON(atomic_read(&iucv->msg_recv) < 0);
259 	}
260 	return net_xmit_eval(err);
261 
262 err_free:
263 	kfree_skb(skb);
264 	return err;
265 }
266 
267 static struct sock *__iucv_get_sock_by_name(char *nm)
268 {
269 	struct sock *sk;
270 
271 	sk_for_each(sk, &iucv_sk_list.head)
272 		if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
273 			return sk;
274 
275 	return NULL;
276 }
277 
278 static void iucv_sock_destruct(struct sock *sk)
279 {
280 	skb_queue_purge(&sk->sk_receive_queue);
281 	skb_queue_purge(&sk->sk_error_queue);
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 	/* Whoever resets the path pointer, must sever and free it. */
341 	if (xchg(&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 	scnprintf(name, sizeof(name),
558 		  "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
559 	while (__iucv_get_sock_by_name(name)) {
560 		scnprintf(name, sizeof(name), "%08x",
561 			  atomic_inc_return(&iucv_sk_list.autobind_name));
562 	}
563 	memcpy(iucv->src_name, name, 8);
564 }
565 
566 /* Bind an unbound socket */
567 static int iucv_sock_bind(struct socket *sock, struct sockaddr_unsized *addr,
568 			  int addr_len)
569 {
570 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
571 	char uid[sizeof(sa->siucv_user_id)];
572 	struct sock *sk = sock->sk;
573 	struct iucv_sock *iucv;
574 	int err = 0;
575 	struct net_device *dev;
576 
577 	/* Verify the input sockaddr */
578 	if (addr_len < sizeof(struct sockaddr_iucv) ||
579 	    addr->sa_family != AF_IUCV)
580 		return -EINVAL;
581 
582 	lock_sock(sk);
583 	if (sk->sk_state != IUCV_OPEN) {
584 		err = -EBADFD;
585 		goto done;
586 	}
587 
588 	write_lock_bh(&iucv_sk_list.lock);
589 
590 	iucv = iucv_sk(sk);
591 	if (__iucv_get_sock_by_name(sa->siucv_name)) {
592 		err = -EADDRINUSE;
593 		goto done_unlock;
594 	}
595 	if (iucv->path)
596 		goto done_unlock;
597 
598 	/* Bind the socket */
599 	if (pr_iucv)
600 		if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
601 			goto vm_bind; /* VM IUCV transport */
602 
603 	/* try hiper transport */
604 	memcpy(uid, sa->siucv_user_id, sizeof(uid));
605 	ASCEBC(uid, 8);
606 	rcu_read_lock();
607 	for_each_netdev_rcu(&init_net, dev) {
608 		if (!memcmp(dev->perm_addr, uid, 8)) {
609 			memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
610 			/* Check for uninitialized siucv_name */
611 			if (strncmp(sa->siucv_name, "        ", 8) == 0)
612 				__iucv_auto_name(iucv);
613 			else
614 				memcpy(iucv->src_name, sa->siucv_name, 8);
615 			sk->sk_bound_dev_if = dev->ifindex;
616 			iucv->hs_dev = dev;
617 			dev_hold(dev);
618 			sk->sk_state = IUCV_BOUND;
619 			iucv->transport = AF_IUCV_TRANS_HIPER;
620 			if (!iucv->msglimit)
621 				iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
622 			rcu_read_unlock();
623 			goto done_unlock;
624 		}
625 	}
626 	rcu_read_unlock();
627 vm_bind:
628 	if (pr_iucv) {
629 		/* use local userid for backward compat */
630 		memcpy(iucv->src_name, sa->siucv_name, 8);
631 		memcpy(iucv->src_user_id, iucv_userid, 8);
632 		sk->sk_state = IUCV_BOUND;
633 		iucv->transport = AF_IUCV_TRANS_IUCV;
634 		sk->sk_allocation |= GFP_DMA;
635 		if (!iucv->msglimit)
636 			iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
637 		goto done_unlock;
638 	}
639 	/* found no dev to bind */
640 	err = -ENODEV;
641 done_unlock:
642 	/* Release the socket list lock */
643 	write_unlock_bh(&iucv_sk_list.lock);
644 done:
645 	release_sock(sk);
646 	return err;
647 }
648 
649 /* Automatically bind an unbound socket */
650 static int iucv_sock_autobind(struct sock *sk)
651 {
652 	struct iucv_sock *iucv = iucv_sk(sk);
653 	int err = 0;
654 
655 	if (unlikely(!pr_iucv))
656 		return -EPROTO;
657 
658 	memcpy(iucv->src_user_id, iucv_userid, 8);
659 	iucv->transport = AF_IUCV_TRANS_IUCV;
660 	sk->sk_allocation |= GFP_DMA;
661 
662 	write_lock_bh(&iucv_sk_list.lock);
663 	__iucv_auto_name(iucv);
664 	write_unlock_bh(&iucv_sk_list.lock);
665 
666 	if (!iucv->msglimit)
667 		iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
668 
669 	return err;
670 }
671 
672 static int afiucv_path_connect(struct socket *sock, struct sockaddr_unsized *addr)
673 {
674 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
675 	struct sock *sk = sock->sk;
676 	struct iucv_sock *iucv = iucv_sk(sk);
677 	unsigned char user_data[16];
678 	int err;
679 
680 	high_nmcpy(user_data, sa->siucv_name);
681 	low_nmcpy(user_data, iucv->src_name);
682 	ASCEBC(user_data, sizeof(user_data));
683 
684 	/* Create path. */
685 	iucv->path = iucv_path_alloc(iucv->msglimit,
686 				     IUCV_IPRMDATA, GFP_KERNEL);
687 	if (!iucv->path) {
688 		err = -ENOMEM;
689 		goto done;
690 	}
691 	err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
692 				    sa->siucv_user_id, NULL, user_data,
693 				    sk);
694 	if (err) {
695 		iucv_path_free(iucv->path);
696 		iucv->path = NULL;
697 		switch (err) {
698 		case 0x0b:	/* Target communicator is not logged on */
699 			err = -ENETUNREACH;
700 			break;
701 		case 0x0d:	/* Max connections for this guest exceeded */
702 		case 0x0e:	/* Max connections for target guest exceeded */
703 			err = -EAGAIN;
704 			break;
705 		case 0x0f:	/* Missing IUCV authorization */
706 			err = -EACCES;
707 			break;
708 		default:
709 			err = -ECONNREFUSED;
710 			break;
711 		}
712 	}
713 done:
714 	return err;
715 }
716 
717 /* Connect an unconnected socket */
718 static int iucv_sock_connect(struct socket *sock, struct sockaddr_unsized *addr,
719 			     int alen, int flags)
720 {
721 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
722 	struct sock *sk = sock->sk;
723 	struct iucv_sock *iucv = iucv_sk(sk);
724 	int err;
725 
726 	if (alen < sizeof(struct sockaddr_iucv) || addr->sa_family != AF_IUCV)
727 		return -EINVAL;
728 
729 	if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
730 		return -EBADFD;
731 
732 	if (sk->sk_state == IUCV_OPEN &&
733 	    iucv->transport == AF_IUCV_TRANS_HIPER)
734 		return -EBADFD; /* explicit bind required */
735 
736 	if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
737 		return -EINVAL;
738 
739 	if (sk->sk_state == IUCV_OPEN) {
740 		err = iucv_sock_autobind(sk);
741 		if (unlikely(err))
742 			return err;
743 	}
744 
745 	lock_sock(sk);
746 
747 	/* Set the destination information */
748 	memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
749 	memcpy(iucv->dst_name, sa->siucv_name, 8);
750 
751 	if (iucv->transport == AF_IUCV_TRANS_HIPER)
752 		err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
753 	else
754 		err = afiucv_path_connect(sock, addr);
755 	if (err)
756 		goto done;
757 
758 	if (sk->sk_state != IUCV_CONNECTED)
759 		err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
760 							    IUCV_DISCONN),
761 				     sock_sndtimeo(sk, flags & O_NONBLOCK));
762 
763 	if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
764 		err = -ECONNREFUSED;
765 
766 	if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
767 		iucv_sever_path(sk, 0);
768 
769 done:
770 	release_sock(sk);
771 	return err;
772 }
773 
774 /* Move a socket into listening state. */
775 static int iucv_sock_listen(struct socket *sock, int backlog)
776 {
777 	struct sock *sk = sock->sk;
778 	int err;
779 
780 	lock_sock(sk);
781 
782 	err = -EINVAL;
783 	if (sk->sk_state != IUCV_BOUND)
784 		goto done;
785 
786 	if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
787 		goto done;
788 
789 	sk->sk_max_ack_backlog = backlog;
790 	sk->sk_ack_backlog = 0;
791 	sk->sk_state = IUCV_LISTEN;
792 	err = 0;
793 
794 done:
795 	release_sock(sk);
796 	return err;
797 }
798 
799 /* Accept a pending connection */
800 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
801 			    struct proto_accept_arg *arg)
802 {
803 	DECLARE_WAITQUEUE(wait, current);
804 	struct sock *sk = sock->sk, *nsk;
805 	long timeo;
806 	int err = 0;
807 
808 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
809 
810 	if (sk->sk_state != IUCV_LISTEN) {
811 		err = -EBADFD;
812 		goto done;
813 	}
814 
815 	timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK);
816 
817 	/* Wait for an incoming connection */
818 	add_wait_queue_exclusive(sk_sleep(sk), &wait);
819 	while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
820 		set_current_state(TASK_INTERRUPTIBLE);
821 		if (!timeo) {
822 			err = -EAGAIN;
823 			break;
824 		}
825 
826 		release_sock(sk);
827 		timeo = schedule_timeout(timeo);
828 		lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
829 
830 		if (sk->sk_state != IUCV_LISTEN) {
831 			err = -EBADFD;
832 			break;
833 		}
834 
835 		if (signal_pending(current)) {
836 			err = sock_intr_errno(timeo);
837 			break;
838 		}
839 	}
840 
841 	set_current_state(TASK_RUNNING);
842 	remove_wait_queue(sk_sleep(sk), &wait);
843 
844 	if (err)
845 		goto done;
846 
847 	newsock->state = SS_CONNECTED;
848 
849 done:
850 	release_sock(sk);
851 	return err;
852 }
853 
854 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
855 			     int peer)
856 {
857 	DECLARE_SOCKADDR(struct sockaddr_iucv *, siucv, addr);
858 	struct sock *sk = sock->sk;
859 	struct iucv_sock *iucv = iucv_sk(sk);
860 
861 	addr->sa_family = AF_IUCV;
862 
863 	if (peer) {
864 		memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
865 		memcpy(siucv->siucv_name, iucv->dst_name, 8);
866 	} else {
867 		memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
868 		memcpy(siucv->siucv_name, iucv->src_name, 8);
869 	}
870 	memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
871 	memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
872 	memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
873 
874 	return sizeof(struct sockaddr_iucv);
875 }
876 
877 /**
878  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
879  * @path:	IUCV path
880  * @msg:	Pointer to a struct iucv_message
881  * @skb:	The socket data to send, skb->len MUST BE <= 7
882  *
883  * Send the socket data in the parameter list in the iucv message
884  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
885  * list and the socket data len at index 7 (last byte).
886  * See also iucv_msg_length().
887  *
888  * Returns: the return code from the iucv_message_send() call.
889  */
890 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
891 			  struct sk_buff *skb)
892 {
893 	u8 prmdata[8];
894 
895 	memcpy(prmdata, (void *) skb->data, skb->len);
896 	prmdata[7] = 0xff - (u8) skb->len;
897 	return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
898 				 (void *) prmdata, 8);
899 }
900 
901 static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,
902 			     size_t len)
903 {
904 	struct sock *sk = sock->sk;
905 	struct iucv_sock *iucv = iucv_sk(sk);
906 	size_t headroom = 0;
907 	size_t linear;
908 	struct sk_buff *skb;
909 	struct iucv_message txmsg = {0};
910 	struct cmsghdr *cmsg;
911 	int cmsg_done;
912 	long timeo;
913 	char user_id[9];
914 	char appl_id[9];
915 	int err;
916 	int noblock = msg->msg_flags & MSG_DONTWAIT;
917 
918 	err = sock_error(sk);
919 	if (err)
920 		return err;
921 
922 	if (msg->msg_flags & MSG_OOB)
923 		return -EOPNOTSUPP;
924 
925 	/* SOCK_SEQPACKET: we do not support segmented records */
926 	if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
927 		return -EOPNOTSUPP;
928 
929 	lock_sock(sk);
930 
931 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
932 		err = -EPIPE;
933 		goto out;
934 	}
935 
936 	/* Return if the socket is not in connected state */
937 	if (sk->sk_state != IUCV_CONNECTED) {
938 		err = -ENOTCONN;
939 		goto out;
940 	}
941 
942 	/* initialize defaults */
943 	cmsg_done   = 0;	/* check for duplicate headers */
944 
945 	/* iterate over control messages */
946 	for_each_cmsghdr(cmsg, msg) {
947 		if (!CMSG_OK(msg, cmsg)) {
948 			err = -EINVAL;
949 			goto out;
950 		}
951 
952 		if (cmsg->cmsg_level != SOL_IUCV)
953 			continue;
954 
955 		if (cmsg->cmsg_type & cmsg_done) {
956 			err = -EINVAL;
957 			goto out;
958 		}
959 		cmsg_done |= cmsg->cmsg_type;
960 
961 		switch (cmsg->cmsg_type) {
962 		case SCM_IUCV_TRGCLS:
963 			if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
964 				err = -EINVAL;
965 				goto out;
966 			}
967 
968 			/* set iucv message target class */
969 			memcpy(&txmsg.class,
970 				(void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
971 
972 			break;
973 
974 		default:
975 			err = -EINVAL;
976 			goto out;
977 		}
978 	}
979 
980 	/* allocate one skb for each iucv message:
981 	 * this is fine for SOCK_SEQPACKET (unless we want to support
982 	 * segmented records using the MSG_EOR flag), but
983 	 * for SOCK_STREAM we might want to improve it in future */
984 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
985 		headroom = sizeof(struct af_iucv_trans_hdr) +
986 			   LL_RESERVED_SPACE(iucv->hs_dev);
987 		linear = min(len, PAGE_SIZE - headroom);
988 	} else {
989 		if (len < PAGE_SIZE) {
990 			linear = len;
991 		} else {
992 			/* In nonlinear "classic" iucv skb,
993 			 * reserve space for iucv_array
994 			 */
995 			headroom = sizeof(struct iucv_array) *
996 				   (MAX_SKB_FRAGS + 1);
997 			linear = PAGE_SIZE - headroom;
998 		}
999 	}
1000 	skb = sock_alloc_send_pskb(sk, headroom + linear, len - linear,
1001 				   noblock, &err, 0);
1002 	if (!skb)
1003 		goto out;
1004 	if (headroom)
1005 		skb_reserve(skb, headroom);
1006 	skb_put(skb, linear);
1007 	skb->len = len;
1008 	skb->data_len = len - linear;
1009 	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
1010 	if (err)
1011 		goto fail;
1012 
1013 	/* wait if outstanding messages for iucv path has reached */
1014 	timeo = sock_sndtimeo(sk, noblock);
1015 	err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1016 	if (err)
1017 		goto fail;
1018 
1019 	/* return -ECONNRESET if the socket is no longer connected */
1020 	if (sk->sk_state != IUCV_CONNECTED) {
1021 		err = -ECONNRESET;
1022 		goto fail;
1023 	}
1024 
1025 	/* increment and save iucv message tag for msg_completion cbk */
1026 	txmsg.tag = iucv->send_tag++;
1027 	IUCV_SKB_CB(skb)->tag = txmsg.tag;
1028 
1029 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1030 		atomic_inc(&iucv->msg_sent);
1031 		err = afiucv_hs_send(&txmsg, sk, skb, 0);
1032 		if (err) {
1033 			atomic_dec(&iucv->msg_sent);
1034 			goto out;
1035 		}
1036 	} else { /* Classic VM IUCV transport */
1037 		skb_queue_tail(&iucv->send_skb_q, skb);
1038 		atomic_inc(&iucv->skbs_in_xmit);
1039 
1040 		if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags) &&
1041 		    skb->len <= 7) {
1042 			err = iucv_send_iprm(iucv->path, &txmsg, skb);
1043 
1044 			/* on success: there is no message_complete callback */
1045 			/* for an IPRMDATA msg; remove skb from send queue   */
1046 			if (err == 0) {
1047 				atomic_dec(&iucv->skbs_in_xmit);
1048 				skb_unlink(skb, &iucv->send_skb_q);
1049 				consume_skb(skb);
1050 			}
1051 
1052 			/* this error should never happen since the	*/
1053 			/* IUCV_IPRMDATA path flag is set... sever path */
1054 			if (err == 0x15) {
1055 				pr_iucv->path_sever(iucv->path, NULL);
1056 				atomic_dec(&iucv->skbs_in_xmit);
1057 				skb_unlink(skb, &iucv->send_skb_q);
1058 				err = -EPIPE;
1059 				goto fail;
1060 			}
1061 		} else if (skb_is_nonlinear(skb)) {
1062 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1063 			int i;
1064 
1065 			/* skip iucv_array lying in the headroom */
1066 			iba[0].address = virt_to_dma32(skb->data);
1067 			iba[0].length = (u32)skb_headlen(skb);
1068 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1069 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1070 
1071 				iba[i + 1].address = virt_to_dma32(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 = virt_to_dma32(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 = virt_to_dma32(skb_frag_address(frag));
1173 				iba[i + 1].length = (u32)skb_frag_size(frag);
1174 			}
1175 			rc = pr_iucv->message_receive(path, msg,
1176 					      IUCV_IPBUFLST,
1177 					      (void *)iba, len, NULL);
1178 		} else {
1179 			rc = pr_iucv->message_receive(path, msg,
1180 					      msg->flags & IUCV_IPRMDATA,
1181 					      skb->data, len, NULL);
1182 		}
1183 		if (rc) {
1184 			kfree_skb(skb);
1185 			return;
1186 		}
1187 		WARN_ON_ONCE(skb->len != len);
1188 	}
1189 
1190 	IUCV_SKB_CB(skb)->offset = 0;
1191 	if (sk_filter(sk, skb)) {
1192 		sk_drops_inc(sk);	/* skb rejected by filter */
1193 		kfree_skb(skb);
1194 		return;
1195 	}
1196 	if (__sock_queue_rcv_skb(sk, skb))	/* handle rcv queue full */
1197 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
1198 }
1199 
1200 /* iucv_process_message_q() - Process outstanding IUCV messages
1201  *
1202  * Locking: must be called with message_q.lock held
1203  */
1204 static void iucv_process_message_q(struct sock *sk)
1205 {
1206 	struct iucv_sock *iucv = iucv_sk(sk);
1207 	struct sk_buff *skb;
1208 	struct sock_msg_q *p, *n;
1209 
1210 	list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1211 		skb = alloc_iucv_recv_skb(iucv_msg_length(&p->msg));
1212 		if (!skb)
1213 			break;
1214 		iucv_process_message(sk, skb, p->path, &p->msg);
1215 		list_del(&p->list);
1216 		kfree(p);
1217 		if (!skb_queue_empty(&iucv->backlog_skb_q))
1218 			break;
1219 	}
1220 }
1221 
1222 static int iucv_sock_recvmsg(struct socket *sock, struct msghdr *msg,
1223 			     size_t len, int flags)
1224 {
1225 	struct sock *sk = sock->sk;
1226 	struct iucv_sock *iucv = iucv_sk(sk);
1227 	unsigned int copied, rlen;
1228 	struct sk_buff *skb, *rskb, *cskb;
1229 	int err = 0;
1230 	u32 offset;
1231 
1232 	if ((sk->sk_state == IUCV_DISCONN) &&
1233 	    skb_queue_empty(&iucv->backlog_skb_q) &&
1234 	    skb_queue_empty(&sk->sk_receive_queue) &&
1235 	    list_empty(&iucv->message_q.list))
1236 		return 0;
1237 
1238 	if (flags & (MSG_OOB))
1239 		return -EOPNOTSUPP;
1240 
1241 	/* receive/dequeue next skb:
1242 	 * the function understands MSG_PEEK and, thus, does not dequeue skb
1243 	 * only refcount is increased.
1244 	 */
1245 	skb = skb_recv_datagram(sk, flags, &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 		err = -EFAULT;
1261 		goto err_out;
1262 	}
1263 
1264 	/* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1265 	if (sk->sk_type == SOCK_SEQPACKET) {
1266 		if (copied < rlen)
1267 			msg->msg_flags |= MSG_TRUNC;
1268 		/* each iucv message contains a complete record */
1269 		msg->msg_flags |= MSG_EOR;
1270 	}
1271 
1272 	/* create control message to store iucv msg target class:
1273 	 * get the trgcls from the control buffer of the skb due to
1274 	 * fragmentation of original iucv message. */
1275 	err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1276 		       sizeof(IUCV_SKB_CB(skb)->class),
1277 		       (void *)&IUCV_SKB_CB(skb)->class);
1278 	if (err)
1279 		goto err_out;
1280 
1281 	/* Mark read part of skb as used */
1282 	if (!(flags & MSG_PEEK)) {
1283 
1284 		/* SOCK_STREAM: re-queue skb if it contains unreceived data */
1285 		if (sk->sk_type == SOCK_STREAM) {
1286 			if (copied < rlen) {
1287 				IUCV_SKB_CB(skb)->offset = offset + copied;
1288 				skb_queue_head(&sk->sk_receive_queue, skb);
1289 				goto done;
1290 			}
1291 		}
1292 
1293 		consume_skb(skb);
1294 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1295 			atomic_inc(&iucv->msg_recv);
1296 			if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1297 				WARN_ON(1);
1298 				iucv_sock_close(sk);
1299 				return -EFAULT;
1300 			}
1301 		}
1302 
1303 		/* Queue backlog skbs */
1304 		spin_lock_bh(&iucv->message_q.lock);
1305 		rskb = skb_dequeue(&iucv->backlog_skb_q);
1306 		while (rskb) {
1307 			IUCV_SKB_CB(rskb)->offset = 0;
1308 			if (__sock_queue_rcv_skb(sk, rskb)) {
1309 				/* handle rcv queue full */
1310 				skb_queue_head(&iucv->backlog_skb_q,
1311 						rskb);
1312 				break;
1313 			}
1314 			rskb = skb_dequeue(&iucv->backlog_skb_q);
1315 		}
1316 		if (skb_queue_empty(&iucv->backlog_skb_q)) {
1317 			if (!list_empty(&iucv->message_q.list))
1318 				iucv_process_message_q(sk);
1319 			if (atomic_read(&iucv->msg_recv) >=
1320 							iucv->msglimit / 2) {
1321 				err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1322 				if (err) {
1323 					sk->sk_state = IUCV_DISCONN;
1324 					sk->sk_state_change(sk);
1325 				}
1326 			}
1327 		}
1328 		spin_unlock_bh(&iucv->message_q.lock);
1329 	}
1330 
1331 done:
1332 	/* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1333 	if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1334 		copied = rlen;
1335 	if (flags & MSG_PEEK)
1336 		skb_unref(skb);
1337 
1338 	return copied;
1339 
1340 err_out:
1341 	if (!(flags & MSG_PEEK))
1342 		skb_queue_head(&sk->sk_receive_queue, skb);
1343 	else
1344 		skb_unref(skb);
1345 
1346 	return err;
1347 }
1348 
1349 static inline __poll_t iucv_accept_poll(struct sock *parent)
1350 {
1351 	struct iucv_sock *isk, *n;
1352 	struct sock *sk;
1353 
1354 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1355 		sk = (struct sock *) isk;
1356 
1357 		if (sk->sk_state == IUCV_CONNECTED)
1358 			return EPOLLIN | EPOLLRDNORM;
1359 	}
1360 
1361 	return 0;
1362 }
1363 
1364 static __poll_t iucv_sock_poll(struct file *file, struct socket *sock,
1365 			       poll_table *wait)
1366 {
1367 	struct sock *sk = sock->sk;
1368 	__poll_t mask = 0;
1369 
1370 	sock_poll_wait(file, sock, wait);
1371 
1372 	if (sk->sk_state == IUCV_LISTEN)
1373 		return iucv_accept_poll(sk);
1374 
1375 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1376 		mask |= EPOLLERR |
1377 			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
1378 
1379 	if (sk->sk_shutdown & RCV_SHUTDOWN)
1380 		mask |= EPOLLRDHUP;
1381 
1382 	if (sk->sk_shutdown == SHUTDOWN_MASK)
1383 		mask |= EPOLLHUP;
1384 
1385 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
1386 	    (sk->sk_shutdown & RCV_SHUTDOWN))
1387 		mask |= EPOLLIN | EPOLLRDNORM;
1388 
1389 	if (sk->sk_state == IUCV_CLOSED)
1390 		mask |= EPOLLHUP;
1391 
1392 	if (sk->sk_state == IUCV_DISCONN)
1393 		mask |= EPOLLIN;
1394 
1395 	if (sock_writeable(sk) && iucv_below_msglim(sk))
1396 		mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1397 	else
1398 		sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1399 
1400 	return mask;
1401 }
1402 
1403 static int iucv_sock_shutdown(struct socket *sock, int how)
1404 {
1405 	struct sock *sk = sock->sk;
1406 	struct iucv_sock *iucv = iucv_sk(sk);
1407 	struct iucv_message txmsg;
1408 	int err = 0;
1409 
1410 	how++;
1411 
1412 	if ((how & ~SHUTDOWN_MASK) || !how)
1413 		return -EINVAL;
1414 
1415 	lock_sock(sk);
1416 	switch (sk->sk_state) {
1417 	case IUCV_LISTEN:
1418 	case IUCV_DISCONN:
1419 	case IUCV_CLOSING:
1420 	case IUCV_CLOSED:
1421 		err = -ENOTCONN;
1422 		goto fail;
1423 	default:
1424 		break;
1425 	}
1426 
1427 	if ((how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) &&
1428 	    sk->sk_state == IUCV_CONNECTED) {
1429 		if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1430 			txmsg.class = 0;
1431 			txmsg.tag = 0;
1432 			err = pr_iucv->message_send(iucv->path, &txmsg,
1433 				IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1434 			if (err) {
1435 				switch (err) {
1436 				case 1:
1437 					err = -ENOTCONN;
1438 					break;
1439 				case 2:
1440 					err = -ECONNRESET;
1441 					break;
1442 				default:
1443 					err = -ENOTCONN;
1444 					break;
1445 				}
1446 			}
1447 		} else
1448 			iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1449 	}
1450 
1451 	sk->sk_shutdown |= how;
1452 	if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1453 		if ((iucv->transport == AF_IUCV_TRANS_IUCV) &&
1454 		    iucv->path) {
1455 			err = pr_iucv->path_quiesce(iucv->path, NULL);
1456 			if (err)
1457 				err = -ENOTCONN;
1458 /*			skb_queue_purge(&sk->sk_receive_queue); */
1459 		}
1460 		skb_queue_purge(&sk->sk_receive_queue);
1461 	}
1462 
1463 	/* Wake up anyone sleeping in poll */
1464 	sk->sk_state_change(sk);
1465 
1466 fail:
1467 	release_sock(sk);
1468 	return err;
1469 }
1470 
1471 static int iucv_sock_release(struct socket *sock)
1472 {
1473 	struct sock *sk = sock->sk;
1474 	int err = 0;
1475 
1476 	if (!sk)
1477 		return 0;
1478 
1479 	iucv_sock_close(sk);
1480 
1481 	sock_orphan(sk);
1482 	iucv_sock_kill(sk);
1483 	return err;
1484 }
1485 
1486 /* getsockopt and setsockopt */
1487 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1488 				sockptr_t optval, unsigned int optlen)
1489 {
1490 	struct sock *sk = sock->sk;
1491 	struct iucv_sock *iucv = iucv_sk(sk);
1492 	int val;
1493 	int rc;
1494 
1495 	if (level != SOL_IUCV)
1496 		return -ENOPROTOOPT;
1497 
1498 	if (optlen < sizeof(int))
1499 		return -EINVAL;
1500 
1501 	if (copy_from_sockptr(&val, optval, sizeof(int)))
1502 		return -EFAULT;
1503 
1504 	rc = 0;
1505 
1506 	lock_sock(sk);
1507 	switch (optname) {
1508 	case SO_IPRMDATA_MSG:
1509 		if (val)
1510 			iucv->flags |= IUCV_IPRMDATA;
1511 		else
1512 			iucv->flags &= ~IUCV_IPRMDATA;
1513 		break;
1514 	case SO_MSGLIMIT:
1515 		switch (sk->sk_state) {
1516 		case IUCV_OPEN:
1517 		case IUCV_BOUND:
1518 			if (val < 1 || val > U16_MAX)
1519 				rc = -EINVAL;
1520 			else
1521 				iucv->msglimit = val;
1522 			break;
1523 		default:
1524 			rc = -EINVAL;
1525 			break;
1526 		}
1527 		break;
1528 	default:
1529 		rc = -ENOPROTOOPT;
1530 		break;
1531 	}
1532 	release_sock(sk);
1533 
1534 	return rc;
1535 }
1536 
1537 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1538 				char __user *optval, int __user *optlen)
1539 {
1540 	struct sock *sk = sock->sk;
1541 	struct iucv_sock *iucv = iucv_sk(sk);
1542 	unsigned int val;
1543 	int len;
1544 
1545 	if (level != SOL_IUCV)
1546 		return -ENOPROTOOPT;
1547 
1548 	if (get_user(len, optlen))
1549 		return -EFAULT;
1550 
1551 	if (len < 0)
1552 		return -EINVAL;
1553 
1554 	len = min_t(unsigned int, len, sizeof(int));
1555 
1556 	switch (optname) {
1557 	case SO_IPRMDATA_MSG:
1558 		val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1559 		break;
1560 	case SO_MSGLIMIT:
1561 		lock_sock(sk);
1562 		val = (iucv->path != NULL) ? iucv->path->msglim	/* connected */
1563 					   : iucv->msglimit;	/* default */
1564 		release_sock(sk);
1565 		break;
1566 	case SO_MSGSIZE:
1567 		if (sk->sk_state == IUCV_OPEN)
1568 			return -EBADFD;
1569 		val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1570 				sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1571 				0x7fffffff;
1572 		break;
1573 	default:
1574 		return -ENOPROTOOPT;
1575 	}
1576 
1577 	if (put_user(len, optlen))
1578 		return -EFAULT;
1579 	if (copy_to_user(optval, &val, len))
1580 		return -EFAULT;
1581 
1582 	return 0;
1583 }
1584 
1585 
1586 /* Callback wrappers - called from iucv base support */
1587 static int iucv_callback_connreq(struct iucv_path *path,
1588 				 u8 ipvmid[8], u8 ipuser[16])
1589 {
1590 	unsigned char user_data[16];
1591 	unsigned char nuser_data[16];
1592 	unsigned char src_name[8];
1593 	struct sock *sk, *nsk;
1594 	struct iucv_sock *iucv, *niucv;
1595 	int err;
1596 
1597 	memcpy(src_name, ipuser, 8);
1598 	EBCASC(src_name, 8);
1599 	/* Find out if this path belongs to af_iucv. */
1600 	read_lock(&iucv_sk_list.lock);
1601 	iucv = NULL;
1602 	sk = NULL;
1603 	sk_for_each(sk, &iucv_sk_list.head)
1604 		if (sk->sk_state == IUCV_LISTEN &&
1605 		    !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1606 			/*
1607 			 * Found a listening socket with
1608 			 * src_name == ipuser[0-7].
1609 			 */
1610 			iucv = iucv_sk(sk);
1611 			break;
1612 		}
1613 	read_unlock(&iucv_sk_list.lock);
1614 	if (!iucv)
1615 		/* No socket found, not one of our paths. */
1616 		return -EINVAL;
1617 
1618 	bh_lock_sock(sk);
1619 
1620 	/* Check if parent socket is listening */
1621 	low_nmcpy(user_data, iucv->src_name);
1622 	high_nmcpy(user_data, iucv->dst_name);
1623 	ASCEBC(user_data, sizeof(user_data));
1624 	if (sk->sk_state != IUCV_LISTEN) {
1625 		err = pr_iucv->path_sever(path, user_data);
1626 		iucv_path_free(path);
1627 		goto fail;
1628 	}
1629 
1630 	/* Check for backlog size */
1631 	if (sk_acceptq_is_full(sk)) {
1632 		err = pr_iucv->path_sever(path, user_data);
1633 		iucv_path_free(path);
1634 		goto fail;
1635 	}
1636 
1637 	/* Create the new socket */
1638 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1639 	if (!nsk) {
1640 		err = pr_iucv->path_sever(path, user_data);
1641 		iucv_path_free(path);
1642 		goto fail;
1643 	}
1644 
1645 	niucv = iucv_sk(nsk);
1646 	iucv_sock_init(nsk, sk);
1647 	niucv->transport = AF_IUCV_TRANS_IUCV;
1648 	nsk->sk_allocation |= GFP_DMA;
1649 
1650 	/* Set the new iucv_sock */
1651 	memcpy(niucv->dst_name, ipuser + 8, 8);
1652 	EBCASC(niucv->dst_name, 8);
1653 	memcpy(niucv->dst_user_id, ipvmid, 8);
1654 	memcpy(niucv->src_name, iucv->src_name, 8);
1655 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1656 	niucv->path = path;
1657 
1658 	/* Call iucv_accept */
1659 	high_nmcpy(nuser_data, ipuser + 8);
1660 	memcpy(nuser_data + 8, niucv->src_name, 8);
1661 	ASCEBC(nuser_data + 8, 8);
1662 
1663 	/* set message limit for path based on msglimit of accepting socket */
1664 	niucv->msglimit = iucv->msglimit;
1665 	path->msglim = iucv->msglimit;
1666 	err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1667 	if (err) {
1668 		iucv_sever_path(nsk, 1);
1669 		iucv_sock_kill(nsk);
1670 		goto fail;
1671 	}
1672 
1673 	iucv_accept_enqueue(sk, nsk);
1674 
1675 	/* Wake up accept */
1676 	nsk->sk_state = IUCV_CONNECTED;
1677 	sk->sk_data_ready(sk);
1678 	err = 0;
1679 fail:
1680 	bh_unlock_sock(sk);
1681 	return 0;
1682 }
1683 
1684 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1685 {
1686 	struct sock *sk = path->private;
1687 
1688 	sk->sk_state = IUCV_CONNECTED;
1689 	sk->sk_state_change(sk);
1690 }
1691 
1692 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1693 {
1694 	struct sock *sk = path->private;
1695 	struct iucv_sock *iucv = iucv_sk(sk);
1696 	struct sk_buff *skb;
1697 	struct sock_msg_q *save_msg;
1698 	int len;
1699 
1700 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1701 		pr_iucv->message_reject(path, msg);
1702 		return;
1703 	}
1704 
1705 	spin_lock(&iucv->message_q.lock);
1706 
1707 	if (!list_empty(&iucv->message_q.list) ||
1708 	    !skb_queue_empty(&iucv->backlog_skb_q))
1709 		goto save_message;
1710 
1711 	len = atomic_read(&sk->sk_rmem_alloc);
1712 	len += SKB_TRUESIZE(iucv_msg_length(msg));
1713 	if (len > sk->sk_rcvbuf)
1714 		goto save_message;
1715 
1716 	skb = alloc_iucv_recv_skb(iucv_msg_length(msg));
1717 	if (!skb)
1718 		goto save_message;
1719 
1720 	iucv_process_message(sk, skb, path, msg);
1721 	goto out_unlock;
1722 
1723 save_message:
1724 	save_msg = kzalloc_obj(struct sock_msg_q, GFP_ATOMIC | GFP_DMA);
1725 	if (!save_msg)
1726 		goto out_unlock;
1727 	save_msg->path = path;
1728 	save_msg->msg = *msg;
1729 
1730 	list_add_tail(&save_msg->list, &iucv->message_q.list);
1731 
1732 out_unlock:
1733 	spin_unlock(&iucv->message_q.lock);
1734 }
1735 
1736 static void iucv_callback_txdone(struct iucv_path *path,
1737 				 struct iucv_message *msg)
1738 {
1739 	struct sock *sk = path->private;
1740 	struct sk_buff *this = NULL;
1741 	struct sk_buff_head *list;
1742 	struct sk_buff *list_skb;
1743 	struct iucv_sock *iucv;
1744 	unsigned long flags;
1745 
1746 	iucv = iucv_sk(sk);
1747 	list = &iucv->send_skb_q;
1748 
1749 	bh_lock_sock(sk);
1750 
1751 	spin_lock_irqsave(&list->lock, flags);
1752 	skb_queue_walk(list, list_skb) {
1753 		if (msg->tag == IUCV_SKB_CB(list_skb)->tag) {
1754 			this = list_skb;
1755 			break;
1756 		}
1757 	}
1758 	if (this) {
1759 		atomic_dec(&iucv->skbs_in_xmit);
1760 		__skb_unlink(this, list);
1761 	}
1762 
1763 	spin_unlock_irqrestore(&list->lock, flags);
1764 
1765 	if (this) {
1766 		consume_skb(this);
1767 		/* wake up any process waiting for sending */
1768 		iucv_sock_wake_msglim(sk);
1769 	}
1770 
1771 	if (sk->sk_state == IUCV_CLOSING) {
1772 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
1773 			sk->sk_state = IUCV_CLOSED;
1774 			sk->sk_state_change(sk);
1775 		}
1776 	}
1777 	bh_unlock_sock(sk);
1778 
1779 }
1780 
1781 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1782 {
1783 	struct sock *sk = path->private;
1784 
1785 	if (sk->sk_state == IUCV_CLOSED)
1786 		return;
1787 
1788 	bh_lock_sock(sk);
1789 	iucv_sever_path(sk, 1);
1790 	sk->sk_state = IUCV_DISCONN;
1791 
1792 	sk->sk_state_change(sk);
1793 	bh_unlock_sock(sk);
1794 }
1795 
1796 /* called if the other communication side shuts down its RECV direction;
1797  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1798  */
1799 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1800 {
1801 	struct sock *sk = path->private;
1802 
1803 	bh_lock_sock(sk);
1804 	if (sk->sk_state != IUCV_CLOSED) {
1805 		sk->sk_shutdown |= SEND_SHUTDOWN;
1806 		sk->sk_state_change(sk);
1807 	}
1808 	bh_unlock_sock(sk);
1809 }
1810 
1811 static struct iucv_handler af_iucv_handler = {
1812 	.path_pending		= iucv_callback_connreq,
1813 	.path_complete		= iucv_callback_connack,
1814 	.path_severed		= iucv_callback_connrej,
1815 	.message_pending	= iucv_callback_rx,
1816 	.message_complete	= iucv_callback_txdone,
1817 	.path_quiesced		= iucv_callback_shutdown,
1818 };
1819 
1820 /***************** HiperSockets transport callbacks ********************/
1821 static void afiucv_swap_src_dest(struct sk_buff *skb)
1822 {
1823 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1824 	char tmpID[8];
1825 	char tmpName[8];
1826 
1827 	ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1828 	ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1829 	ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1830 	ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1831 	memcpy(tmpID, trans_hdr->srcUserID, 8);
1832 	memcpy(tmpName, trans_hdr->srcAppName, 8);
1833 	memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1834 	memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1835 	memcpy(trans_hdr->destUserID, tmpID, 8);
1836 	memcpy(trans_hdr->destAppName, tmpName, 8);
1837 	skb_push(skb, ETH_HLEN);
1838 	memset(skb->data, 0, ETH_HLEN);
1839 }
1840 
1841 /*
1842  * afiucv_hs_callback_syn - react on received SYN
1843  */
1844 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1845 {
1846 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1847 	struct sock *nsk;
1848 	struct iucv_sock *iucv, *niucv;
1849 	int err;
1850 
1851 	iucv = iucv_sk(sk);
1852 	if (!iucv) {
1853 		/* no sock - connection refused */
1854 		afiucv_swap_src_dest(skb);
1855 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1856 		err = dev_queue_xmit(skb);
1857 		goto out;
1858 	}
1859 
1860 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1861 	bh_lock_sock(sk);
1862 	if ((sk->sk_state != IUCV_LISTEN) ||
1863 	    sk_acceptq_is_full(sk) ||
1864 	    !nsk) {
1865 		/* error on server socket - connection refused */
1866 		afiucv_swap_src_dest(skb);
1867 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1868 		err = dev_queue_xmit(skb);
1869 		iucv_sock_kill(nsk);
1870 		bh_unlock_sock(sk);
1871 		goto out;
1872 	}
1873 
1874 	niucv = iucv_sk(nsk);
1875 	iucv_sock_init(nsk, sk);
1876 	niucv->transport = AF_IUCV_TRANS_HIPER;
1877 	niucv->msglimit = iucv->msglimit;
1878 	if (!trans_hdr->window)
1879 		niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1880 	else
1881 		niucv->msglimit_peer = trans_hdr->window;
1882 	memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1883 	memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1884 	memcpy(niucv->src_name, iucv->src_name, 8);
1885 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1886 	nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1887 	niucv->hs_dev = iucv->hs_dev;
1888 	dev_hold(niucv->hs_dev);
1889 	afiucv_swap_src_dest(skb);
1890 	trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1891 	trans_hdr->window = niucv->msglimit;
1892 	/* if receiver acks the xmit connection is established */
1893 	err = dev_queue_xmit(skb);
1894 	if (!err) {
1895 		iucv_accept_enqueue(sk, nsk);
1896 		nsk->sk_state = IUCV_CONNECTED;
1897 		sk->sk_data_ready(sk);
1898 	} else
1899 		iucv_sock_kill(nsk);
1900 	bh_unlock_sock(sk);
1901 
1902 out:
1903 	return NET_RX_SUCCESS;
1904 }
1905 
1906 /*
1907  * afiucv_hs_callback_synack() - react on received SYN-ACK
1908  */
1909 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1910 {
1911 	struct iucv_sock *iucv = iucv_sk(sk);
1912 
1913 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1914 		kfree_skb(skb);
1915 		return NET_RX_SUCCESS;
1916 	}
1917 
1918 	bh_lock_sock(sk);
1919 	iucv->msglimit_peer = iucv_trans_hdr(skb)->window;
1920 	sk->sk_state = IUCV_CONNECTED;
1921 	sk->sk_state_change(sk);
1922 	bh_unlock_sock(sk);
1923 	consume_skb(skb);
1924 	return NET_RX_SUCCESS;
1925 }
1926 
1927 /*
1928  * afiucv_hs_callback_synfin() - react on received SYN_FIN
1929  */
1930 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
1931 {
1932 	struct iucv_sock *iucv = iucv_sk(sk);
1933 
1934 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1935 		kfree_skb(skb);
1936 		return NET_RX_SUCCESS;
1937 	}
1938 
1939 	bh_lock_sock(sk);
1940 	sk->sk_state = IUCV_DISCONN;
1941 	sk->sk_state_change(sk);
1942 	bh_unlock_sock(sk);
1943 	consume_skb(skb);
1944 	return NET_RX_SUCCESS;
1945 }
1946 
1947 /*
1948  * afiucv_hs_callback_fin() - react on received FIN
1949  */
1950 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
1951 {
1952 	struct iucv_sock *iucv = iucv_sk(sk);
1953 
1954 	/* other end of connection closed */
1955 	if (!iucv) {
1956 		kfree_skb(skb);
1957 		return NET_RX_SUCCESS;
1958 	}
1959 
1960 	bh_lock_sock(sk);
1961 	if (sk->sk_state == IUCV_CONNECTED) {
1962 		sk->sk_state = IUCV_DISCONN;
1963 		sk->sk_state_change(sk);
1964 	}
1965 	bh_unlock_sock(sk);
1966 	consume_skb(skb);
1967 	return NET_RX_SUCCESS;
1968 }
1969 
1970 /*
1971  * afiucv_hs_callback_win() - react on received WIN
1972  */
1973 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
1974 {
1975 	struct iucv_sock *iucv = iucv_sk(sk);
1976 
1977 	if (!iucv)
1978 		return NET_RX_SUCCESS;
1979 
1980 	if (sk->sk_state != IUCV_CONNECTED)
1981 		return NET_RX_SUCCESS;
1982 
1983 	atomic_sub(iucv_trans_hdr(skb)->window, &iucv->msg_sent);
1984 	iucv_sock_wake_msglim(sk);
1985 	return NET_RX_SUCCESS;
1986 }
1987 
1988 /*
1989  * afiucv_hs_callback_rx() - react on received data
1990  */
1991 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
1992 {
1993 	struct iucv_sock *iucv = iucv_sk(sk);
1994 
1995 	if (!iucv) {
1996 		kfree_skb(skb);
1997 		return NET_RX_SUCCESS;
1998 	}
1999 
2000 	if (sk->sk_state != IUCV_CONNECTED) {
2001 		kfree_skb(skb);
2002 		return NET_RX_SUCCESS;
2003 	}
2004 
2005 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2006 		kfree_skb(skb);
2007 		return NET_RX_SUCCESS;
2008 	}
2009 
2010 	/* write stuff from iucv_msg to skb cb */
2011 	skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2012 	skb_reset_transport_header(skb);
2013 	skb_reset_network_header(skb);
2014 	IUCV_SKB_CB(skb)->offset = 0;
2015 	if (sk_filter(sk, skb)) {
2016 		sk_drops_inc(sk);	/* skb rejected by filter */
2017 		kfree_skb(skb);
2018 		return NET_RX_SUCCESS;
2019 	}
2020 
2021 	spin_lock(&iucv->message_q.lock);
2022 	if (skb_queue_empty(&iucv->backlog_skb_q)) {
2023 		if (__sock_queue_rcv_skb(sk, skb))
2024 			/* handle rcv queue full */
2025 			skb_queue_tail(&iucv->backlog_skb_q, skb);
2026 	} else
2027 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2028 	spin_unlock(&iucv->message_q.lock);
2029 	return NET_RX_SUCCESS;
2030 }
2031 
2032 /*
2033  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2034  *                   transport
2035  *                   called from netif RX softirq
2036  */
2037 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2038 	struct packet_type *pt, struct net_device *orig_dev)
2039 {
2040 	struct sock *sk;
2041 	struct iucv_sock *iucv;
2042 	struct af_iucv_trans_hdr *trans_hdr;
2043 	int err = NET_RX_SUCCESS;
2044 	char nullstring[8];
2045 
2046 	if (!pskb_may_pull(skb, sizeof(*trans_hdr))) {
2047 		kfree_skb(skb);
2048 		return NET_RX_SUCCESS;
2049 	}
2050 
2051 	trans_hdr = iucv_trans_hdr(skb);
2052 	EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2053 	EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2054 	EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2055 	EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2056 	memset(nullstring, 0, sizeof(nullstring));
2057 	iucv = NULL;
2058 	sk = NULL;
2059 	read_lock(&iucv_sk_list.lock);
2060 	sk_for_each(sk, &iucv_sk_list.head) {
2061 		if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2062 			if ((!memcmp(&iucv_sk(sk)->src_name,
2063 				     trans_hdr->destAppName, 8)) &&
2064 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2065 				     trans_hdr->destUserID, 8)) &&
2066 			    (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2067 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2068 				     nullstring, 8))) {
2069 				iucv = iucv_sk(sk);
2070 				break;
2071 			}
2072 		} else {
2073 			if ((!memcmp(&iucv_sk(sk)->src_name,
2074 				     trans_hdr->destAppName, 8)) &&
2075 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2076 				     trans_hdr->destUserID, 8)) &&
2077 			    (!memcmp(&iucv_sk(sk)->dst_name,
2078 				     trans_hdr->srcAppName, 8)) &&
2079 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2080 				     trans_hdr->srcUserID, 8))) {
2081 				iucv = iucv_sk(sk);
2082 				break;
2083 			}
2084 		}
2085 	}
2086 	read_unlock(&iucv_sk_list.lock);
2087 	if (!iucv)
2088 		sk = NULL;
2089 
2090 	/* no sock
2091 	how should we send with no sock
2092 	1) send without sock no send rc checking?
2093 	2) introduce default sock to handle this cases
2094 
2095 	 SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2096 	 data -> send FIN
2097 	 SYN|ACK, SYN|FIN, FIN -> no action? */
2098 
2099 	switch (trans_hdr->flags) {
2100 	case AF_IUCV_FLAG_SYN:
2101 		/* connect request */
2102 		err = afiucv_hs_callback_syn(sk, skb);
2103 		break;
2104 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2105 		/* connect request confirmed */
2106 		err = afiucv_hs_callback_synack(sk, skb);
2107 		break;
2108 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2109 		/* connect request refused */
2110 		err = afiucv_hs_callback_synfin(sk, skb);
2111 		break;
2112 	case (AF_IUCV_FLAG_FIN):
2113 		/* close request */
2114 		err = afiucv_hs_callback_fin(sk, skb);
2115 		break;
2116 	case (AF_IUCV_FLAG_WIN):
2117 		err = afiucv_hs_callback_win(sk, skb);
2118 		if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2119 			consume_skb(skb);
2120 			break;
2121 		}
2122 		fallthrough;	/* and receive non-zero length data */
2123 	case (AF_IUCV_FLAG_SHT):
2124 		/* shutdown request */
2125 		fallthrough;	/* and receive zero length data */
2126 	case 0:
2127 		/* plain data frame */
2128 		IUCV_SKB_CB(skb)->class = trans_hdr->iucv_hdr.class;
2129 		err = afiucv_hs_callback_rx(sk, skb);
2130 		break;
2131 	default:
2132 		kfree_skb(skb);
2133 	}
2134 
2135 	return err;
2136 }
2137 
2138 /*
2139  * afiucv_hs_callback_txnotify() - handle send notifications from HiperSockets
2140  *                                 transport
2141  */
2142 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify n)
2143 {
2144 	struct iucv_sock *iucv = iucv_sk(sk);
2145 
2146 	if (sock_flag(sk, SOCK_ZAPPED))
2147 		return;
2148 
2149 	switch (n) {
2150 	case TX_NOTIFY_OK:
2151 		atomic_dec(&iucv->skbs_in_xmit);
2152 		iucv_sock_wake_msglim(sk);
2153 		break;
2154 	case TX_NOTIFY_PENDING:
2155 		atomic_inc(&iucv->pendings);
2156 		break;
2157 	case TX_NOTIFY_DELAYED_OK:
2158 		atomic_dec(&iucv->skbs_in_xmit);
2159 		if (atomic_dec_return(&iucv->pendings) <= 0)
2160 			iucv_sock_wake_msglim(sk);
2161 		break;
2162 	default:
2163 		atomic_dec(&iucv->skbs_in_xmit);
2164 		if (sk->sk_state == IUCV_CONNECTED) {
2165 			sk->sk_state = IUCV_DISCONN;
2166 			sk->sk_state_change(sk);
2167 		}
2168 	}
2169 
2170 	if (sk->sk_state == IUCV_CLOSING) {
2171 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
2172 			sk->sk_state = IUCV_CLOSED;
2173 			sk->sk_state_change(sk);
2174 		}
2175 	}
2176 }
2177 
2178 /*
2179  * afiucv_netdev_event: handle netdev notifier chain events
2180  */
2181 static int afiucv_netdev_event(struct notifier_block *this,
2182 			       unsigned long event, void *ptr)
2183 {
2184 	struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
2185 	struct sock *sk;
2186 	struct iucv_sock *iucv;
2187 
2188 	switch (event) {
2189 	case NETDEV_REBOOT:
2190 	case NETDEV_GOING_DOWN:
2191 		sk_for_each(sk, &iucv_sk_list.head) {
2192 			iucv = iucv_sk(sk);
2193 			if ((iucv->hs_dev == event_dev) &&
2194 			    (sk->sk_state == IUCV_CONNECTED)) {
2195 				if (event == NETDEV_GOING_DOWN)
2196 					iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2197 				sk->sk_state = IUCV_DISCONN;
2198 				sk->sk_state_change(sk);
2199 			}
2200 		}
2201 		break;
2202 	case NETDEV_DOWN:
2203 	case NETDEV_UNREGISTER:
2204 	default:
2205 		break;
2206 	}
2207 	return NOTIFY_DONE;
2208 }
2209 
2210 static struct notifier_block afiucv_netdev_notifier = {
2211 	.notifier_call = afiucv_netdev_event,
2212 };
2213 
2214 static const struct proto_ops iucv_sock_ops = {
2215 	.family		= PF_IUCV,
2216 	.owner		= THIS_MODULE,
2217 	.release	= iucv_sock_release,
2218 	.bind		= iucv_sock_bind,
2219 	.connect	= iucv_sock_connect,
2220 	.listen		= iucv_sock_listen,
2221 	.accept		= iucv_sock_accept,
2222 	.getname	= iucv_sock_getname,
2223 	.sendmsg	= iucv_sock_sendmsg,
2224 	.recvmsg	= iucv_sock_recvmsg,
2225 	.poll		= iucv_sock_poll,
2226 	.ioctl		= sock_no_ioctl,
2227 	.mmap		= sock_no_mmap,
2228 	.socketpair	= sock_no_socketpair,
2229 	.shutdown	= iucv_sock_shutdown,
2230 	.setsockopt	= iucv_sock_setsockopt,
2231 	.getsockopt	= iucv_sock_getsockopt,
2232 };
2233 
2234 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
2235 			    int kern)
2236 {
2237 	struct sock *sk;
2238 
2239 	if (protocol && protocol != PF_IUCV)
2240 		return -EPROTONOSUPPORT;
2241 
2242 	sock->state = SS_UNCONNECTED;
2243 
2244 	switch (sock->type) {
2245 	case SOCK_STREAM:
2246 	case SOCK_SEQPACKET:
2247 		/* currently, proto ops can handle both sk types */
2248 		sock->ops = &iucv_sock_ops;
2249 		break;
2250 	default:
2251 		return -ESOCKTNOSUPPORT;
2252 	}
2253 
2254 	sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL, kern);
2255 	if (!sk)
2256 		return -ENOMEM;
2257 
2258 	iucv_sock_init(sk, NULL);
2259 
2260 	return 0;
2261 }
2262 
2263 static const struct net_proto_family iucv_sock_family_ops = {
2264 	.family	= AF_IUCV,
2265 	.owner	= THIS_MODULE,
2266 	.create	= iucv_sock_create,
2267 };
2268 
2269 static struct packet_type iucv_packet_type = {
2270 	.type = cpu_to_be16(ETH_P_AF_IUCV),
2271 	.func = afiucv_hs_rcv,
2272 };
2273 
2274 static int __init afiucv_init(void)
2275 {
2276 	int err;
2277 
2278 	if (machine_is_vm() && IS_ENABLED(CONFIG_IUCV)) {
2279 		cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2280 		if (unlikely(err)) {
2281 			WARN_ON(err);
2282 			err = -EPROTONOSUPPORT;
2283 			goto out;
2284 		}
2285 
2286 		pr_iucv = &iucv_if;
2287 	} else {
2288 		memset(&iucv_userid, 0, sizeof(iucv_userid));
2289 		pr_iucv = NULL;
2290 	}
2291 
2292 	err = proto_register(&iucv_proto, 0);
2293 	if (err)
2294 		goto out;
2295 	err = sock_register(&iucv_sock_family_ops);
2296 	if (err)
2297 		goto out_proto;
2298 
2299 	if (pr_iucv) {
2300 		err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2301 		if (err)
2302 			goto out_sock;
2303 	}
2304 
2305 	err = register_netdevice_notifier(&afiucv_netdev_notifier);
2306 	if (err)
2307 		goto out_notifier;
2308 
2309 	dev_add_pack(&iucv_packet_type);
2310 	return 0;
2311 
2312 out_notifier:
2313 	if (pr_iucv)
2314 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2315 out_sock:
2316 	sock_unregister(PF_IUCV);
2317 out_proto:
2318 	proto_unregister(&iucv_proto);
2319 out:
2320 	return err;
2321 }
2322 
2323 static void __exit afiucv_exit(void)
2324 {
2325 	if (pr_iucv)
2326 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2327 
2328 	unregister_netdevice_notifier(&afiucv_netdev_notifier);
2329 	dev_remove_pack(&iucv_packet_type);
2330 	sock_unregister(PF_IUCV);
2331 	proto_unregister(&iucv_proto);
2332 }
2333 
2334 module_init(afiucv_init);
2335 module_exit(afiucv_exit);
2336 
2337 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2338 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2339 MODULE_VERSION(VERSION);
2340 MODULE_LICENSE("GPL");
2341 MODULE_ALIAS_NETPROTO(PF_IUCV);
2342