xref: /openbmc/linux/drivers/hv/channel.c (revision e8f6f3b4)
1 /*
2  * Copyright (c) 2009, Microsoft Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * You should have received a copy of the GNU General Public License along with
14  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
15  * Place - Suite 330, Boston, MA 02111-1307 USA.
16  *
17  * Authors:
18  *   Haiyang Zhang <haiyangz@microsoft.com>
19  *   Hank Janssen  <hjanssen@microsoft.com>
20  */
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 
23 #include <linux/kernel.h>
24 #include <linux/sched.h>
25 #include <linux/wait.h>
26 #include <linux/mm.h>
27 #include <linux/slab.h>
28 #include <linux/module.h>
29 #include <linux/hyperv.h>
30 #include <linux/uio.h>
31 
32 #include "hyperv_vmbus.h"
33 
34 #define NUM_PAGES_SPANNED(addr, len) \
35 ((PAGE_ALIGN(addr + len) >> PAGE_SHIFT) - (addr >> PAGE_SHIFT))
36 
37 /*
38  * vmbus_setevent- Trigger an event notification on the specified
39  * channel.
40  */
41 static void vmbus_setevent(struct vmbus_channel *channel)
42 {
43 	struct hv_monitor_page *monitorpage;
44 
45 	if (channel->offermsg.monitor_allocated) {
46 		/* Each u32 represents 32 channels */
47 		sync_set_bit(channel->offermsg.child_relid & 31,
48 			(unsigned long *) vmbus_connection.send_int_page +
49 			(channel->offermsg.child_relid >> 5));
50 
51 		/* Get the child to parent monitor page */
52 		monitorpage = vmbus_connection.monitor_pages[1];
53 
54 		sync_set_bit(channel->monitor_bit,
55 			(unsigned long *)&monitorpage->trigger_group
56 					[channel->monitor_grp].pending);
57 
58 	} else {
59 		vmbus_set_event(channel);
60 	}
61 }
62 
63 /*
64  * vmbus_open - Open the specified channel.
65  */
66 int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
67 		     u32 recv_ringbuffer_size, void *userdata, u32 userdatalen,
68 		     void (*onchannelcallback)(void *context), void *context)
69 {
70 	struct vmbus_channel_open_channel *open_msg;
71 	struct vmbus_channel_msginfo *open_info = NULL;
72 	void *in, *out;
73 	unsigned long flags;
74 	int ret, t, err = 0;
75 
76 	spin_lock_irqsave(&newchannel->sc_lock, flags);
77 	if (newchannel->state == CHANNEL_OPEN_STATE) {
78 		newchannel->state = CHANNEL_OPENING_STATE;
79 	} else {
80 		spin_unlock_irqrestore(&newchannel->sc_lock, flags);
81 		return -EINVAL;
82 	}
83 	spin_unlock_irqrestore(&newchannel->sc_lock, flags);
84 
85 	newchannel->onchannel_callback = onchannelcallback;
86 	newchannel->channel_callback_context = context;
87 
88 	/* Allocate the ring buffer */
89 	out = (void *)__get_free_pages(GFP_KERNEL|__GFP_ZERO,
90 		get_order(send_ringbuffer_size + recv_ringbuffer_size));
91 
92 	if (!out)
93 		return -ENOMEM;
94 
95 
96 	in = (void *)((unsigned long)out + send_ringbuffer_size);
97 
98 	newchannel->ringbuffer_pages = out;
99 	newchannel->ringbuffer_pagecount = (send_ringbuffer_size +
100 					   recv_ringbuffer_size) >> PAGE_SHIFT;
101 
102 	ret = hv_ringbuffer_init(
103 		&newchannel->outbound, out, send_ringbuffer_size);
104 
105 	if (ret != 0) {
106 		err = ret;
107 		goto error0;
108 	}
109 
110 	ret = hv_ringbuffer_init(
111 		&newchannel->inbound, in, recv_ringbuffer_size);
112 	if (ret != 0) {
113 		err = ret;
114 		goto error0;
115 	}
116 
117 
118 	/* Establish the gpadl for the ring buffer */
119 	newchannel->ringbuffer_gpadlhandle = 0;
120 
121 	ret = vmbus_establish_gpadl(newchannel,
122 					 newchannel->outbound.ring_buffer,
123 					 send_ringbuffer_size +
124 					 recv_ringbuffer_size,
125 					 &newchannel->ringbuffer_gpadlhandle);
126 
127 	if (ret != 0) {
128 		err = ret;
129 		goto error0;
130 	}
131 
132 	/* Create and init the channel open message */
133 	open_info = kmalloc(sizeof(*open_info) +
134 			   sizeof(struct vmbus_channel_open_channel),
135 			   GFP_KERNEL);
136 	if (!open_info) {
137 		err = -ENOMEM;
138 		goto error0;
139 	}
140 
141 	init_completion(&open_info->waitevent);
142 
143 	open_msg = (struct vmbus_channel_open_channel *)open_info->msg;
144 	open_msg->header.msgtype = CHANNELMSG_OPENCHANNEL;
145 	open_msg->openid = newchannel->offermsg.child_relid;
146 	open_msg->child_relid = newchannel->offermsg.child_relid;
147 	open_msg->ringbuffer_gpadlhandle = newchannel->ringbuffer_gpadlhandle;
148 	open_msg->downstream_ringbuffer_pageoffset = send_ringbuffer_size >>
149 						  PAGE_SHIFT;
150 	open_msg->target_vp = newchannel->target_vp;
151 
152 	if (userdatalen > MAX_USER_DEFINED_BYTES) {
153 		err = -EINVAL;
154 		goto error0;
155 	}
156 
157 	if (userdatalen)
158 		memcpy(open_msg->userdata, userdata, userdatalen);
159 
160 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
161 	list_add_tail(&open_info->msglistentry,
162 		      &vmbus_connection.chn_msg_list);
163 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
164 
165 	ret = vmbus_post_msg(open_msg,
166 			       sizeof(struct vmbus_channel_open_channel));
167 
168 	if (ret != 0) {
169 		err = ret;
170 		goto error1;
171 	}
172 
173 	t = wait_for_completion_timeout(&open_info->waitevent, 5*HZ);
174 	if (t == 0) {
175 		err = -ETIMEDOUT;
176 		goto error1;
177 	}
178 
179 
180 	if (open_info->response.open_result.status)
181 		err = open_info->response.open_result.status;
182 
183 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
184 	list_del(&open_info->msglistentry);
185 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
186 
187 	if (err == 0)
188 		newchannel->state = CHANNEL_OPENED_STATE;
189 
190 	kfree(open_info);
191 	return err;
192 
193 error1:
194 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
195 	list_del(&open_info->msglistentry);
196 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
197 
198 error0:
199 	free_pages((unsigned long)out,
200 		get_order(send_ringbuffer_size + recv_ringbuffer_size));
201 	kfree(open_info);
202 	return err;
203 }
204 EXPORT_SYMBOL_GPL(vmbus_open);
205 
206 /*
207  * create_gpadl_header - Creates a gpadl for the specified buffer
208  */
209 static int create_gpadl_header(void *kbuffer, u32 size,
210 					 struct vmbus_channel_msginfo **msginfo,
211 					 u32 *messagecount)
212 {
213 	int i;
214 	int pagecount;
215 	struct vmbus_channel_gpadl_header *gpadl_header;
216 	struct vmbus_channel_gpadl_body *gpadl_body;
217 	struct vmbus_channel_msginfo *msgheader;
218 	struct vmbus_channel_msginfo *msgbody = NULL;
219 	u32 msgsize;
220 
221 	int pfnsum, pfncount, pfnleft, pfncurr, pfnsize;
222 
223 	pagecount = size >> PAGE_SHIFT;
224 
225 	/* do we need a gpadl body msg */
226 	pfnsize = MAX_SIZE_CHANNEL_MESSAGE -
227 		  sizeof(struct vmbus_channel_gpadl_header) -
228 		  sizeof(struct gpa_range);
229 	pfncount = pfnsize / sizeof(u64);
230 
231 	if (pagecount > pfncount) {
232 		/* we need a gpadl body */
233 		/* fill in the header */
234 		msgsize = sizeof(struct vmbus_channel_msginfo) +
235 			  sizeof(struct vmbus_channel_gpadl_header) +
236 			  sizeof(struct gpa_range) + pfncount * sizeof(u64);
237 		msgheader =  kzalloc(msgsize, GFP_KERNEL);
238 		if (!msgheader)
239 			goto nomem;
240 
241 		INIT_LIST_HEAD(&msgheader->submsglist);
242 		msgheader->msgsize = msgsize;
243 
244 		gpadl_header = (struct vmbus_channel_gpadl_header *)
245 			msgheader->msg;
246 		gpadl_header->rangecount = 1;
247 		gpadl_header->range_buflen = sizeof(struct gpa_range) +
248 					 pagecount * sizeof(u64);
249 		gpadl_header->range[0].byte_offset = 0;
250 		gpadl_header->range[0].byte_count = size;
251 		for (i = 0; i < pfncount; i++)
252 			gpadl_header->range[0].pfn_array[i] = slow_virt_to_phys(
253 				kbuffer + PAGE_SIZE * i) >> PAGE_SHIFT;
254 		*msginfo = msgheader;
255 		*messagecount = 1;
256 
257 		pfnsum = pfncount;
258 		pfnleft = pagecount - pfncount;
259 
260 		/* how many pfns can we fit */
261 		pfnsize = MAX_SIZE_CHANNEL_MESSAGE -
262 			  sizeof(struct vmbus_channel_gpadl_body);
263 		pfncount = pfnsize / sizeof(u64);
264 
265 		/* fill in the body */
266 		while (pfnleft) {
267 			if (pfnleft > pfncount)
268 				pfncurr = pfncount;
269 			else
270 				pfncurr = pfnleft;
271 
272 			msgsize = sizeof(struct vmbus_channel_msginfo) +
273 				  sizeof(struct vmbus_channel_gpadl_body) +
274 				  pfncurr * sizeof(u64);
275 			msgbody = kzalloc(msgsize, GFP_KERNEL);
276 
277 			if (!msgbody) {
278 				struct vmbus_channel_msginfo *pos = NULL;
279 				struct vmbus_channel_msginfo *tmp = NULL;
280 				/*
281 				 * Free up all the allocated messages.
282 				 */
283 				list_for_each_entry_safe(pos, tmp,
284 					&msgheader->submsglist,
285 					msglistentry) {
286 
287 					list_del(&pos->msglistentry);
288 					kfree(pos);
289 				}
290 
291 				goto nomem;
292 			}
293 
294 			msgbody->msgsize = msgsize;
295 			(*messagecount)++;
296 			gpadl_body =
297 				(struct vmbus_channel_gpadl_body *)msgbody->msg;
298 
299 			/*
300 			 * Gpadl is u32 and we are using a pointer which could
301 			 * be 64-bit
302 			 * This is governed by the guest/host protocol and
303 			 * so the hypervisor gurantees that this is ok.
304 			 */
305 			for (i = 0; i < pfncurr; i++)
306 				gpadl_body->pfn[i] = slow_virt_to_phys(
307 					kbuffer + PAGE_SIZE * (pfnsum + i)) >>
308 					PAGE_SHIFT;
309 
310 			/* add to msg header */
311 			list_add_tail(&msgbody->msglistentry,
312 				      &msgheader->submsglist);
313 			pfnsum += pfncurr;
314 			pfnleft -= pfncurr;
315 		}
316 	} else {
317 		/* everything fits in a header */
318 		msgsize = sizeof(struct vmbus_channel_msginfo) +
319 			  sizeof(struct vmbus_channel_gpadl_header) +
320 			  sizeof(struct gpa_range) + pagecount * sizeof(u64);
321 		msgheader = kzalloc(msgsize, GFP_KERNEL);
322 		if (msgheader == NULL)
323 			goto nomem;
324 		msgheader->msgsize = msgsize;
325 
326 		gpadl_header = (struct vmbus_channel_gpadl_header *)
327 			msgheader->msg;
328 		gpadl_header->rangecount = 1;
329 		gpadl_header->range_buflen = sizeof(struct gpa_range) +
330 					 pagecount * sizeof(u64);
331 		gpadl_header->range[0].byte_offset = 0;
332 		gpadl_header->range[0].byte_count = size;
333 		for (i = 0; i < pagecount; i++)
334 			gpadl_header->range[0].pfn_array[i] = slow_virt_to_phys(
335 				kbuffer + PAGE_SIZE * i) >> PAGE_SHIFT;
336 
337 		*msginfo = msgheader;
338 		*messagecount = 1;
339 	}
340 
341 	return 0;
342 nomem:
343 	kfree(msgheader);
344 	kfree(msgbody);
345 	return -ENOMEM;
346 }
347 
348 /*
349  * vmbus_establish_gpadl - Estabish a GPADL for the specified buffer
350  *
351  * @channel: a channel
352  * @kbuffer: from kmalloc or vmalloc
353  * @size: page-size multiple
354  * @gpadl_handle: some funky thing
355  */
356 int vmbus_establish_gpadl(struct vmbus_channel *channel, void *kbuffer,
357 			       u32 size, u32 *gpadl_handle)
358 {
359 	struct vmbus_channel_gpadl_header *gpadlmsg;
360 	struct vmbus_channel_gpadl_body *gpadl_body;
361 	struct vmbus_channel_msginfo *msginfo = NULL;
362 	struct vmbus_channel_msginfo *submsginfo;
363 	u32 msgcount;
364 	struct list_head *curr;
365 	u32 next_gpadl_handle;
366 	unsigned long flags;
367 	int ret = 0;
368 
369 	next_gpadl_handle = atomic_read(&vmbus_connection.next_gpadl_handle);
370 	atomic_inc(&vmbus_connection.next_gpadl_handle);
371 
372 	ret = create_gpadl_header(kbuffer, size, &msginfo, &msgcount);
373 	if (ret)
374 		return ret;
375 
376 	init_completion(&msginfo->waitevent);
377 
378 	gpadlmsg = (struct vmbus_channel_gpadl_header *)msginfo->msg;
379 	gpadlmsg->header.msgtype = CHANNELMSG_GPADL_HEADER;
380 	gpadlmsg->child_relid = channel->offermsg.child_relid;
381 	gpadlmsg->gpadl = next_gpadl_handle;
382 
383 
384 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
385 	list_add_tail(&msginfo->msglistentry,
386 		      &vmbus_connection.chn_msg_list);
387 
388 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
389 
390 	ret = vmbus_post_msg(gpadlmsg, msginfo->msgsize -
391 			       sizeof(*msginfo));
392 	if (ret != 0)
393 		goto cleanup;
394 
395 	if (msgcount > 1) {
396 		list_for_each(curr, &msginfo->submsglist) {
397 
398 			submsginfo = (struct vmbus_channel_msginfo *)curr;
399 			gpadl_body =
400 			     (struct vmbus_channel_gpadl_body *)submsginfo->msg;
401 
402 			gpadl_body->header.msgtype =
403 				CHANNELMSG_GPADL_BODY;
404 			gpadl_body->gpadl = next_gpadl_handle;
405 
406 			ret = vmbus_post_msg(gpadl_body,
407 					       submsginfo->msgsize -
408 					       sizeof(*submsginfo));
409 			if (ret != 0)
410 				goto cleanup;
411 
412 		}
413 	}
414 	wait_for_completion(&msginfo->waitevent);
415 
416 	/* At this point, we received the gpadl created msg */
417 	*gpadl_handle = gpadlmsg->gpadl;
418 
419 cleanup:
420 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
421 	list_del(&msginfo->msglistentry);
422 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
423 
424 	kfree(msginfo);
425 	return ret;
426 }
427 EXPORT_SYMBOL_GPL(vmbus_establish_gpadl);
428 
429 /*
430  * vmbus_teardown_gpadl -Teardown the specified GPADL handle
431  */
432 int vmbus_teardown_gpadl(struct vmbus_channel *channel, u32 gpadl_handle)
433 {
434 	struct vmbus_channel_gpadl_teardown *msg;
435 	struct vmbus_channel_msginfo *info;
436 	unsigned long flags;
437 	int ret;
438 
439 	info = kmalloc(sizeof(*info) +
440 		       sizeof(struct vmbus_channel_gpadl_teardown), GFP_KERNEL);
441 	if (!info)
442 		return -ENOMEM;
443 
444 	init_completion(&info->waitevent);
445 
446 	msg = (struct vmbus_channel_gpadl_teardown *)info->msg;
447 
448 	msg->header.msgtype = CHANNELMSG_GPADL_TEARDOWN;
449 	msg->child_relid = channel->offermsg.child_relid;
450 	msg->gpadl = gpadl_handle;
451 
452 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
453 	list_add_tail(&info->msglistentry,
454 		      &vmbus_connection.chn_msg_list);
455 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
456 	ret = vmbus_post_msg(msg,
457 			       sizeof(struct vmbus_channel_gpadl_teardown));
458 
459 	if (ret)
460 		goto post_msg_err;
461 
462 	wait_for_completion(&info->waitevent);
463 
464 post_msg_err:
465 	spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
466 	list_del(&info->msglistentry);
467 	spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
468 
469 	kfree(info);
470 	return ret;
471 }
472 EXPORT_SYMBOL_GPL(vmbus_teardown_gpadl);
473 
474 static void reset_channel_cb(void *arg)
475 {
476 	struct vmbus_channel *channel = arg;
477 
478 	channel->onchannel_callback = NULL;
479 }
480 
481 static int vmbus_close_internal(struct vmbus_channel *channel)
482 {
483 	struct vmbus_channel_close_channel *msg;
484 	int ret;
485 
486 	channel->state = CHANNEL_OPEN_STATE;
487 	channel->sc_creation_callback = NULL;
488 	/* Stop callback and cancel the timer asap */
489 	if (channel->target_cpu != get_cpu()) {
490 		put_cpu();
491 		smp_call_function_single(channel->target_cpu, reset_channel_cb,
492 					 channel, true);
493 	} else {
494 		reset_channel_cb(channel);
495 		put_cpu();
496 	}
497 
498 	/* Send a closing message */
499 
500 	msg = &channel->close_msg.msg;
501 
502 	msg->header.msgtype = CHANNELMSG_CLOSECHANNEL;
503 	msg->child_relid = channel->offermsg.child_relid;
504 
505 	ret = vmbus_post_msg(msg, sizeof(struct vmbus_channel_close_channel));
506 
507 	if (ret) {
508 		pr_err("Close failed: close post msg return is %d\n", ret);
509 		/*
510 		 * If we failed to post the close msg,
511 		 * it is perhaps better to leak memory.
512 		 */
513 		return ret;
514 	}
515 
516 	/* Tear down the gpadl for the channel's ring buffer */
517 	if (channel->ringbuffer_gpadlhandle) {
518 		ret = vmbus_teardown_gpadl(channel,
519 					   channel->ringbuffer_gpadlhandle);
520 		if (ret) {
521 			pr_err("Close failed: teardown gpadl return %d\n", ret);
522 			/*
523 			 * If we failed to teardown gpadl,
524 			 * it is perhaps better to leak memory.
525 			 */
526 			return ret;
527 		}
528 	}
529 
530 	/* Cleanup the ring buffers for this channel */
531 	hv_ringbuffer_cleanup(&channel->outbound);
532 	hv_ringbuffer_cleanup(&channel->inbound);
533 
534 	free_pages((unsigned long)channel->ringbuffer_pages,
535 		get_order(channel->ringbuffer_pagecount * PAGE_SIZE));
536 
537 	return ret;
538 }
539 
540 /*
541  * vmbus_close - Close the specified channel
542  */
543 void vmbus_close(struct vmbus_channel *channel)
544 {
545 	struct list_head *cur, *tmp;
546 	struct vmbus_channel *cur_channel;
547 
548 	if (channel->primary_channel != NULL) {
549 		/*
550 		 * We will only close sub-channels when
551 		 * the primary is closed.
552 		 */
553 		return;
554 	}
555 	/*
556 	 * Close all the sub-channels first and then close the
557 	 * primary channel.
558 	 */
559 	list_for_each_safe(cur, tmp, &channel->sc_list) {
560 		cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
561 		if (cur_channel->state != CHANNEL_OPENED_STATE)
562 			continue;
563 		vmbus_close_internal(cur_channel);
564 	}
565 	/*
566 	 * Now close the primary.
567 	 */
568 	vmbus_close_internal(channel);
569 }
570 EXPORT_SYMBOL_GPL(vmbus_close);
571 
572 /**
573  * vmbus_sendpacket() - Send the specified buffer on the given channel
574  * @channel: Pointer to vmbus_channel structure.
575  * @buffer: Pointer to the buffer you want to receive the data into.
576  * @bufferlen: Maximum size of what the the buffer will hold
577  * @requestid: Identifier of the request
578  * @type: Type of packet that is being send e.g. negotiate, time
579  * packet etc.
580  *
581  * Sends data in @buffer directly to hyper-v via the vmbus
582  * This will send the data unparsed to hyper-v.
583  *
584  * Mainly used by Hyper-V drivers.
585  */
586 int vmbus_sendpacket(struct vmbus_channel *channel, void *buffer,
587 			   u32 bufferlen, u64 requestid,
588 			   enum vmbus_packet_type type, u32 flags)
589 {
590 	struct vmpacket_descriptor desc;
591 	u32 packetlen = sizeof(struct vmpacket_descriptor) + bufferlen;
592 	u32 packetlen_aligned = ALIGN(packetlen, sizeof(u64));
593 	struct kvec bufferlist[3];
594 	u64 aligned_data = 0;
595 	int ret;
596 	bool signal = false;
597 
598 
599 	/* Setup the descriptor */
600 	desc.type = type; /* VmbusPacketTypeDataInBand; */
601 	desc.flags = flags; /* VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED; */
602 	/* in 8-bytes granularity */
603 	desc.offset8 = sizeof(struct vmpacket_descriptor) >> 3;
604 	desc.len8 = (u16)(packetlen_aligned >> 3);
605 	desc.trans_id = requestid;
606 
607 	bufferlist[0].iov_base = &desc;
608 	bufferlist[0].iov_len = sizeof(struct vmpacket_descriptor);
609 	bufferlist[1].iov_base = buffer;
610 	bufferlist[1].iov_len = bufferlen;
611 	bufferlist[2].iov_base = &aligned_data;
612 	bufferlist[2].iov_len = (packetlen_aligned - packetlen);
613 
614 	ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3, &signal);
615 
616 	if (ret == 0 && signal)
617 		vmbus_setevent(channel);
618 
619 	return ret;
620 }
621 EXPORT_SYMBOL(vmbus_sendpacket);
622 
623 /*
624  * vmbus_sendpacket_pagebuffer - Send a range of single-page buffer
625  * packets using a GPADL Direct packet type.
626  */
627 int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
628 				     struct hv_page_buffer pagebuffers[],
629 				     u32 pagecount, void *buffer, u32 bufferlen,
630 				     u64 requestid)
631 {
632 	int ret;
633 	int i;
634 	struct vmbus_channel_packet_page_buffer desc;
635 	u32 descsize;
636 	u32 packetlen;
637 	u32 packetlen_aligned;
638 	struct kvec bufferlist[3];
639 	u64 aligned_data = 0;
640 	bool signal = false;
641 
642 	if (pagecount > MAX_PAGE_BUFFER_COUNT)
643 		return -EINVAL;
644 
645 
646 	/*
647 	 * Adjust the size down since vmbus_channel_packet_page_buffer is the
648 	 * largest size we support
649 	 */
650 	descsize = sizeof(struct vmbus_channel_packet_page_buffer) -
651 			  ((MAX_PAGE_BUFFER_COUNT - pagecount) *
652 			  sizeof(struct hv_page_buffer));
653 	packetlen = descsize + bufferlen;
654 	packetlen_aligned = ALIGN(packetlen, sizeof(u64));
655 
656 	/* Setup the descriptor */
657 	desc.type = VM_PKT_DATA_USING_GPA_DIRECT;
658 	desc.flags = VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED;
659 	desc.dataoffset8 = descsize >> 3; /* in 8-bytes grandularity */
660 	desc.length8 = (u16)(packetlen_aligned >> 3);
661 	desc.transactionid = requestid;
662 	desc.rangecount = pagecount;
663 
664 	for (i = 0; i < pagecount; i++) {
665 		desc.range[i].len = pagebuffers[i].len;
666 		desc.range[i].offset = pagebuffers[i].offset;
667 		desc.range[i].pfn	 = pagebuffers[i].pfn;
668 	}
669 
670 	bufferlist[0].iov_base = &desc;
671 	bufferlist[0].iov_len = descsize;
672 	bufferlist[1].iov_base = buffer;
673 	bufferlist[1].iov_len = bufferlen;
674 	bufferlist[2].iov_base = &aligned_data;
675 	bufferlist[2].iov_len = (packetlen_aligned - packetlen);
676 
677 	ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3, &signal);
678 
679 	if (ret == 0 && signal)
680 		vmbus_setevent(channel);
681 
682 	return ret;
683 }
684 EXPORT_SYMBOL_GPL(vmbus_sendpacket_pagebuffer);
685 
686 /*
687  * vmbus_sendpacket_multipagebuffer - Send a multi-page buffer packet
688  * using a GPADL Direct packet type.
689  */
690 int vmbus_sendpacket_multipagebuffer(struct vmbus_channel *channel,
691 				struct hv_multipage_buffer *multi_pagebuffer,
692 				void *buffer, u32 bufferlen, u64 requestid)
693 {
694 	int ret;
695 	struct vmbus_channel_packet_multipage_buffer desc;
696 	u32 descsize;
697 	u32 packetlen;
698 	u32 packetlen_aligned;
699 	struct kvec bufferlist[3];
700 	u64 aligned_data = 0;
701 	bool signal = false;
702 	u32 pfncount = NUM_PAGES_SPANNED(multi_pagebuffer->offset,
703 					 multi_pagebuffer->len);
704 
705 	if (pfncount > MAX_MULTIPAGE_BUFFER_COUNT)
706 		return -EINVAL;
707 
708 	/*
709 	 * Adjust the size down since vmbus_channel_packet_multipage_buffer is
710 	 * the largest size we support
711 	 */
712 	descsize = sizeof(struct vmbus_channel_packet_multipage_buffer) -
713 			  ((MAX_MULTIPAGE_BUFFER_COUNT - pfncount) *
714 			  sizeof(u64));
715 	packetlen = descsize + bufferlen;
716 	packetlen_aligned = ALIGN(packetlen, sizeof(u64));
717 
718 
719 	/* Setup the descriptor */
720 	desc.type = VM_PKT_DATA_USING_GPA_DIRECT;
721 	desc.flags = VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED;
722 	desc.dataoffset8 = descsize >> 3; /* in 8-bytes grandularity */
723 	desc.length8 = (u16)(packetlen_aligned >> 3);
724 	desc.transactionid = requestid;
725 	desc.rangecount = 1;
726 
727 	desc.range.len = multi_pagebuffer->len;
728 	desc.range.offset = multi_pagebuffer->offset;
729 
730 	memcpy(desc.range.pfn_array, multi_pagebuffer->pfn_array,
731 	       pfncount * sizeof(u64));
732 
733 	bufferlist[0].iov_base = &desc;
734 	bufferlist[0].iov_len = descsize;
735 	bufferlist[1].iov_base = buffer;
736 	bufferlist[1].iov_len = bufferlen;
737 	bufferlist[2].iov_base = &aligned_data;
738 	bufferlist[2].iov_len = (packetlen_aligned - packetlen);
739 
740 	ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3, &signal);
741 
742 	if (ret == 0 && signal)
743 		vmbus_setevent(channel);
744 
745 	return ret;
746 }
747 EXPORT_SYMBOL_GPL(vmbus_sendpacket_multipagebuffer);
748 
749 /**
750  * vmbus_recvpacket() - Retrieve the user packet on the specified channel
751  * @channel: Pointer to vmbus_channel structure.
752  * @buffer: Pointer to the buffer you want to receive the data into.
753  * @bufferlen: Maximum size of what the the buffer will hold
754  * @buffer_actual_len: The actual size of the data after it was received
755  * @requestid: Identifier of the request
756  *
757  * Receives directly from the hyper-v vmbus and puts the data it received
758  * into Buffer. This will receive the data unparsed from hyper-v.
759  *
760  * Mainly used by Hyper-V drivers.
761  */
762 int vmbus_recvpacket(struct vmbus_channel *channel, void *buffer,
763 			u32 bufferlen, u32 *buffer_actual_len, u64 *requestid)
764 {
765 	struct vmpacket_descriptor desc;
766 	u32 packetlen;
767 	u32 userlen;
768 	int ret;
769 	bool signal = false;
770 
771 	*buffer_actual_len = 0;
772 	*requestid = 0;
773 
774 
775 	ret = hv_ringbuffer_peek(&channel->inbound, &desc,
776 			     sizeof(struct vmpacket_descriptor));
777 	if (ret != 0)
778 		return 0;
779 
780 	packetlen = desc.len8 << 3;
781 	userlen = packetlen - (desc.offset8 << 3);
782 
783 	*buffer_actual_len = userlen;
784 
785 	if (userlen > bufferlen) {
786 
787 		pr_err("Buffer too small - got %d needs %d\n",
788 			   bufferlen, userlen);
789 		return -ETOOSMALL;
790 	}
791 
792 	*requestid = desc.trans_id;
793 
794 	/* Copy over the packet to the user buffer */
795 	ret = hv_ringbuffer_read(&channel->inbound, buffer, userlen,
796 			     (desc.offset8 << 3), &signal);
797 
798 	if (signal)
799 		vmbus_setevent(channel);
800 
801 	return 0;
802 }
803 EXPORT_SYMBOL(vmbus_recvpacket);
804 
805 /*
806  * vmbus_recvpacket_raw - Retrieve the raw packet on the specified channel
807  */
808 int vmbus_recvpacket_raw(struct vmbus_channel *channel, void *buffer,
809 			      u32 bufferlen, u32 *buffer_actual_len,
810 			      u64 *requestid)
811 {
812 	struct vmpacket_descriptor desc;
813 	u32 packetlen;
814 	int ret;
815 	bool signal = false;
816 
817 	*buffer_actual_len = 0;
818 	*requestid = 0;
819 
820 
821 	ret = hv_ringbuffer_peek(&channel->inbound, &desc,
822 			     sizeof(struct vmpacket_descriptor));
823 	if (ret != 0)
824 		return 0;
825 
826 
827 	packetlen = desc.len8 << 3;
828 
829 	*buffer_actual_len = packetlen;
830 
831 	if (packetlen > bufferlen)
832 		return -ENOBUFS;
833 
834 	*requestid = desc.trans_id;
835 
836 	/* Copy over the entire packet to the user buffer */
837 	ret = hv_ringbuffer_read(&channel->inbound, buffer, packetlen, 0,
838 				 &signal);
839 
840 	if (signal)
841 		vmbus_setevent(channel);
842 
843 	return ret;
844 }
845 EXPORT_SYMBOL_GPL(vmbus_recvpacket_raw);
846