xref: /linux-6.15/include/linux/hyperv.h (revision 85dc2c65)
1 /*
2  *
3  * Copyright (c) 2011, Microsoft Corporation.
4  *
5  * This program is free software; you can redistribute it and/or modify it
6  * under the terms and conditions of the GNU General Public License,
7  * version 2, as published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope it will be useful, but WITHOUT
10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
12  * more details.
13  *
14  * You should have received a copy of the GNU General Public License along with
15  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
16  * Place - Suite 330, Boston, MA 02111-1307 USA.
17  *
18  * Authors:
19  *   Haiyang Zhang <[email protected]>
20  *   Hank Janssen  <[email protected]>
21  *   K. Y. Srinivasan <[email protected]>
22  *
23  */
24 
25 #ifndef _HYPERV_H
26 #define _HYPERV_H
27 
28 #include <uapi/linux/hyperv.h>
29 
30 #include <linux/types.h>
31 #include <linux/scatterlist.h>
32 #include <linux/list.h>
33 #include <linux/timer.h>
34 #include <linux/completion.h>
35 #include <linux/device.h>
36 #include <linux/mod_devicetable.h>
37 #include <linux/interrupt.h>
38 #include <linux/reciprocal_div.h>
39 
40 #define MAX_PAGE_BUFFER_COUNT				32
41 #define MAX_MULTIPAGE_BUFFER_COUNT			32 /* 128K */
42 
43 #pragma pack(push, 1)
44 
45 /* Single-page buffer */
46 struct hv_page_buffer {
47 	u32 len;
48 	u32 offset;
49 	u64 pfn;
50 };
51 
52 /* Multiple-page buffer */
53 struct hv_multipage_buffer {
54 	/* Length and Offset determines the # of pfns in the array */
55 	u32 len;
56 	u32 offset;
57 	u64 pfn_array[MAX_MULTIPAGE_BUFFER_COUNT];
58 };
59 
60 /*
61  * Multiple-page buffer array; the pfn array is variable size:
62  * The number of entries in the PFN array is determined by
63  * "len" and "offset".
64  */
65 struct hv_mpb_array {
66 	/* Length and Offset determines the # of pfns in the array */
67 	u32 len;
68 	u32 offset;
69 	u64 pfn_array[];
70 };
71 
72 /* 0x18 includes the proprietary packet header */
73 #define MAX_PAGE_BUFFER_PACKET		(0x18 +			\
74 					(sizeof(struct hv_page_buffer) * \
75 					 MAX_PAGE_BUFFER_COUNT))
76 #define MAX_MULTIPAGE_BUFFER_PACKET	(0x18 +			\
77 					 sizeof(struct hv_multipage_buffer))
78 
79 
80 #pragma pack(pop)
81 
82 struct hv_ring_buffer {
83 	/* Offset in bytes from the start of ring data below */
84 	u32 write_index;
85 
86 	/* Offset in bytes from the start of ring data below */
87 	u32 read_index;
88 
89 	u32 interrupt_mask;
90 
91 	/*
92 	 * WS2012/Win8 and later versions of Hyper-V implement interrupt
93 	 * driven flow management. The feature bit feat_pending_send_sz
94 	 * is set by the host on the host->guest ring buffer, and by the
95 	 * guest on the guest->host ring buffer.
96 	 *
97 	 * The meaning of the feature bit is a bit complex in that it has
98 	 * semantics that apply to both ring buffers.  If the guest sets
99 	 * the feature bit in the guest->host ring buffer, the guest is
100 	 * telling the host that:
101 	 * 1) It will set the pending_send_sz field in the guest->host ring
102 	 *    buffer when it is waiting for space to become available, and
103 	 * 2) It will read the pending_send_sz field in the host->guest
104 	 *    ring buffer and interrupt the host when it frees enough space
105 	 *
106 	 * Similarly, if the host sets the feature bit in the host->guest
107 	 * ring buffer, the host is telling the guest that:
108 	 * 1) It will set the pending_send_sz field in the host->guest ring
109 	 *    buffer when it is waiting for space to become available, and
110 	 * 2) It will read the pending_send_sz field in the guest->host
111 	 *    ring buffer and interrupt the guest when it frees enough space
112 	 *
113 	 * If either the guest or host does not set the feature bit that it
114 	 * owns, that guest or host must do polling if it encounters a full
115 	 * ring buffer, and not signal the other end with an interrupt.
116 	 */
117 	u32 pending_send_sz;
118 	u32 reserved1[12];
119 	union {
120 		struct {
121 			u32 feat_pending_send_sz:1;
122 		};
123 		u32 value;
124 	} feature_bits;
125 
126 	/* Pad it to PAGE_SIZE so that data starts on page boundary */
127 	u8	reserved2[4028];
128 
129 	/*
130 	 * Ring data starts here + RingDataStartOffset
131 	 * !!! DO NOT place any fields below this !!!
132 	 */
133 	u8 buffer[0];
134 } __packed;
135 
136 struct hv_ring_buffer_info {
137 	struct hv_ring_buffer *ring_buffer;
138 	u32 ring_size;			/* Include the shared header */
139 	struct reciprocal_value ring_size_div10_reciprocal;
140 	spinlock_t ring_lock;
141 
142 	u32 ring_datasize;		/* < ring_size */
143 	u32 priv_read_index;
144 };
145 
146 
147 static inline u32 hv_get_bytes_to_read(const struct hv_ring_buffer_info *rbi)
148 {
149 	u32 read_loc, write_loc, dsize, read;
150 
151 	dsize = rbi->ring_datasize;
152 	read_loc = rbi->ring_buffer->read_index;
153 	write_loc = READ_ONCE(rbi->ring_buffer->write_index);
154 
155 	read = write_loc >= read_loc ? (write_loc - read_loc) :
156 		(dsize - read_loc) + write_loc;
157 
158 	return read;
159 }
160 
161 static inline u32 hv_get_bytes_to_write(const struct hv_ring_buffer_info *rbi)
162 {
163 	u32 read_loc, write_loc, dsize, write;
164 
165 	dsize = rbi->ring_datasize;
166 	read_loc = READ_ONCE(rbi->ring_buffer->read_index);
167 	write_loc = rbi->ring_buffer->write_index;
168 
169 	write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
170 		read_loc - write_loc;
171 	return write;
172 }
173 
174 static inline u32 hv_get_avail_to_write_percent(
175 		const struct hv_ring_buffer_info *rbi)
176 {
177 	u32 avail_write = hv_get_bytes_to_write(rbi);
178 
179 	return reciprocal_divide(
180 			(avail_write  << 3) + (avail_write << 1),
181 			rbi->ring_size_div10_reciprocal);
182 }
183 
184 /*
185  * VMBUS version is 32 bit entity broken up into
186  * two 16 bit quantities: major_number. minor_number.
187  *
188  * 0 . 13 (Windows Server 2008)
189  * 1 . 1  (Windows 7)
190  * 2 . 4  (Windows 8)
191  * 3 . 0  (Windows 8 R2)
192  * 4 . 0  (Windows 10)
193  * 5 . 0  (Newer Windows 10)
194  */
195 
196 #define VERSION_WS2008  ((0 << 16) | (13))
197 #define VERSION_WIN7    ((1 << 16) | (1))
198 #define VERSION_WIN8    ((2 << 16) | (4))
199 #define VERSION_WIN8_1    ((3 << 16) | (0))
200 #define VERSION_WIN10	((4 << 16) | (0))
201 #define VERSION_WIN10_V5 ((5 << 16) | (0))
202 
203 #define VERSION_INVAL -1
204 
205 #define VERSION_CURRENT VERSION_WIN10_V5
206 
207 /* Make maximum size of pipe payload of 16K */
208 #define MAX_PIPE_DATA_PAYLOAD		(sizeof(u8) * 16384)
209 
210 /* Define PipeMode values. */
211 #define VMBUS_PIPE_TYPE_BYTE		0x00000000
212 #define VMBUS_PIPE_TYPE_MESSAGE		0x00000004
213 
214 /* The size of the user defined data buffer for non-pipe offers. */
215 #define MAX_USER_DEFINED_BYTES		120
216 
217 /* The size of the user defined data buffer for pipe offers. */
218 #define MAX_PIPE_USER_DEFINED_BYTES	116
219 
220 /*
221  * At the center of the Channel Management library is the Channel Offer. This
222  * struct contains the fundamental information about an offer.
223  */
224 struct vmbus_channel_offer {
225 	uuid_le if_type;
226 	uuid_le if_instance;
227 
228 	/*
229 	 * These two fields are not currently used.
230 	 */
231 	u64 reserved1;
232 	u64 reserved2;
233 
234 	u16 chn_flags;
235 	u16 mmio_megabytes;		/* in bytes * 1024 * 1024 */
236 
237 	union {
238 		/* Non-pipes: The user has MAX_USER_DEFINED_BYTES bytes. */
239 		struct {
240 			unsigned char user_def[MAX_USER_DEFINED_BYTES];
241 		} std;
242 
243 		/*
244 		 * Pipes:
245 		 * The following sructure is an integrated pipe protocol, which
246 		 * is implemented on top of standard user-defined data. Pipe
247 		 * clients have MAX_PIPE_USER_DEFINED_BYTES left for their own
248 		 * use.
249 		 */
250 		struct {
251 			u32  pipe_mode;
252 			unsigned char user_def[MAX_PIPE_USER_DEFINED_BYTES];
253 		} pipe;
254 	} u;
255 	/*
256 	 * The sub_channel_index is defined in win8.
257 	 */
258 	u16 sub_channel_index;
259 	u16 reserved3;
260 } __packed;
261 
262 /* Server Flags */
263 #define VMBUS_CHANNEL_ENUMERATE_DEVICE_INTERFACE	1
264 #define VMBUS_CHANNEL_SERVER_SUPPORTS_TRANSFER_PAGES	2
265 #define VMBUS_CHANNEL_SERVER_SUPPORTS_GPADLS		4
266 #define VMBUS_CHANNEL_NAMED_PIPE_MODE			0x10
267 #define VMBUS_CHANNEL_LOOPBACK_OFFER			0x100
268 #define VMBUS_CHANNEL_PARENT_OFFER			0x200
269 #define VMBUS_CHANNEL_REQUEST_MONITORED_NOTIFICATION	0x400
270 #define VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER		0x2000
271 
272 struct vmpacket_descriptor {
273 	u16 type;
274 	u16 offset8;
275 	u16 len8;
276 	u16 flags;
277 	u64 trans_id;
278 } __packed;
279 
280 struct vmpacket_header {
281 	u32 prev_pkt_start_offset;
282 	struct vmpacket_descriptor descriptor;
283 } __packed;
284 
285 struct vmtransfer_page_range {
286 	u32 byte_count;
287 	u32 byte_offset;
288 } __packed;
289 
290 struct vmtransfer_page_packet_header {
291 	struct vmpacket_descriptor d;
292 	u16 xfer_pageset_id;
293 	u8  sender_owns_set;
294 	u8 reserved;
295 	u32 range_cnt;
296 	struct vmtransfer_page_range ranges[1];
297 } __packed;
298 
299 struct vmgpadl_packet_header {
300 	struct vmpacket_descriptor d;
301 	u32 gpadl;
302 	u32 reserved;
303 } __packed;
304 
305 struct vmadd_remove_transfer_page_set {
306 	struct vmpacket_descriptor d;
307 	u32 gpadl;
308 	u16 xfer_pageset_id;
309 	u16 reserved;
310 } __packed;
311 
312 /*
313  * This structure defines a range in guest physical space that can be made to
314  * look virtually contiguous.
315  */
316 struct gpa_range {
317 	u32 byte_count;
318 	u32 byte_offset;
319 	u64 pfn_array[0];
320 };
321 
322 /*
323  * This is the format for an Establish Gpadl packet, which contains a handle by
324  * which this GPADL will be known and a set of GPA ranges associated with it.
325  * This can be converted to a MDL by the guest OS.  If there are multiple GPA
326  * ranges, then the resulting MDL will be "chained," representing multiple VA
327  * ranges.
328  */
329 struct vmestablish_gpadl {
330 	struct vmpacket_descriptor d;
331 	u32 gpadl;
332 	u32 range_cnt;
333 	struct gpa_range range[1];
334 } __packed;
335 
336 /*
337  * This is the format for a Teardown Gpadl packet, which indicates that the
338  * GPADL handle in the Establish Gpadl packet will never be referenced again.
339  */
340 struct vmteardown_gpadl {
341 	struct vmpacket_descriptor d;
342 	u32 gpadl;
343 	u32 reserved;	/* for alignment to a 8-byte boundary */
344 } __packed;
345 
346 /*
347  * This is the format for a GPA-Direct packet, which contains a set of GPA
348  * ranges, in addition to commands and/or data.
349  */
350 struct vmdata_gpa_direct {
351 	struct vmpacket_descriptor d;
352 	u32 reserved;
353 	u32 range_cnt;
354 	struct gpa_range range[1];
355 } __packed;
356 
357 /* This is the format for a Additional Data Packet. */
358 struct vmadditional_data {
359 	struct vmpacket_descriptor d;
360 	u64 total_bytes;
361 	u32 offset;
362 	u32 byte_cnt;
363 	unsigned char data[1];
364 } __packed;
365 
366 union vmpacket_largest_possible_header {
367 	struct vmpacket_descriptor simple_hdr;
368 	struct vmtransfer_page_packet_header xfer_page_hdr;
369 	struct vmgpadl_packet_header gpadl_hdr;
370 	struct vmadd_remove_transfer_page_set add_rm_xfer_page_hdr;
371 	struct vmestablish_gpadl establish_gpadl_hdr;
372 	struct vmteardown_gpadl teardown_gpadl_hdr;
373 	struct vmdata_gpa_direct data_gpa_direct_hdr;
374 };
375 
376 #define VMPACKET_DATA_START_ADDRESS(__packet)	\
377 	(void *)(((unsigned char *)__packet) +	\
378 	 ((struct vmpacket_descriptor)__packet)->offset8 * 8)
379 
380 #define VMPACKET_DATA_LENGTH(__packet)		\
381 	((((struct vmpacket_descriptor)__packet)->len8 -	\
382 	  ((struct vmpacket_descriptor)__packet)->offset8) * 8)
383 
384 #define VMPACKET_TRANSFER_MODE(__packet)	\
385 	(((struct IMPACT)__packet)->type)
386 
387 enum vmbus_packet_type {
388 	VM_PKT_INVALID				= 0x0,
389 	VM_PKT_SYNCH				= 0x1,
390 	VM_PKT_ADD_XFER_PAGESET			= 0x2,
391 	VM_PKT_RM_XFER_PAGESET			= 0x3,
392 	VM_PKT_ESTABLISH_GPADL			= 0x4,
393 	VM_PKT_TEARDOWN_GPADL			= 0x5,
394 	VM_PKT_DATA_INBAND			= 0x6,
395 	VM_PKT_DATA_USING_XFER_PAGES		= 0x7,
396 	VM_PKT_DATA_USING_GPADL			= 0x8,
397 	VM_PKT_DATA_USING_GPA_DIRECT		= 0x9,
398 	VM_PKT_CANCEL_REQUEST			= 0xa,
399 	VM_PKT_COMP				= 0xb,
400 	VM_PKT_DATA_USING_ADDITIONAL_PKT	= 0xc,
401 	VM_PKT_ADDITIONAL_DATA			= 0xd
402 };
403 
404 #define VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED	1
405 
406 
407 /* Version 1 messages */
408 enum vmbus_channel_message_type {
409 	CHANNELMSG_INVALID			=  0,
410 	CHANNELMSG_OFFERCHANNEL		=  1,
411 	CHANNELMSG_RESCIND_CHANNELOFFER	=  2,
412 	CHANNELMSG_REQUESTOFFERS		=  3,
413 	CHANNELMSG_ALLOFFERS_DELIVERED	=  4,
414 	CHANNELMSG_OPENCHANNEL		=  5,
415 	CHANNELMSG_OPENCHANNEL_RESULT		=  6,
416 	CHANNELMSG_CLOSECHANNEL		=  7,
417 	CHANNELMSG_GPADL_HEADER		=  8,
418 	CHANNELMSG_GPADL_BODY			=  9,
419 	CHANNELMSG_GPADL_CREATED		= 10,
420 	CHANNELMSG_GPADL_TEARDOWN		= 11,
421 	CHANNELMSG_GPADL_TORNDOWN		= 12,
422 	CHANNELMSG_RELID_RELEASED		= 13,
423 	CHANNELMSG_INITIATE_CONTACT		= 14,
424 	CHANNELMSG_VERSION_RESPONSE		= 15,
425 	CHANNELMSG_UNLOAD			= 16,
426 	CHANNELMSG_UNLOAD_RESPONSE		= 17,
427 	CHANNELMSG_18				= 18,
428 	CHANNELMSG_19				= 19,
429 	CHANNELMSG_20				= 20,
430 	CHANNELMSG_TL_CONNECT_REQUEST		= 21,
431 	CHANNELMSG_COUNT
432 };
433 
434 struct vmbus_channel_message_header {
435 	enum vmbus_channel_message_type msgtype;
436 	u32 padding;
437 } __packed;
438 
439 /* Query VMBus Version parameters */
440 struct vmbus_channel_query_vmbus_version {
441 	struct vmbus_channel_message_header header;
442 	u32 version;
443 } __packed;
444 
445 /* VMBus Version Supported parameters */
446 struct vmbus_channel_version_supported {
447 	struct vmbus_channel_message_header header;
448 	u8 version_supported;
449 } __packed;
450 
451 /* Offer Channel parameters */
452 struct vmbus_channel_offer_channel {
453 	struct vmbus_channel_message_header header;
454 	struct vmbus_channel_offer offer;
455 	u32 child_relid;
456 	u8 monitorid;
457 	/*
458 	 * win7 and beyond splits this field into a bit field.
459 	 */
460 	u8 monitor_allocated:1;
461 	u8 reserved:7;
462 	/*
463 	 * These are new fields added in win7 and later.
464 	 * Do not access these fields without checking the
465 	 * negotiated protocol.
466 	 *
467 	 * If "is_dedicated_interrupt" is set, we must not set the
468 	 * associated bit in the channel bitmap while sending the
469 	 * interrupt to the host.
470 	 *
471 	 * connection_id is to be used in signaling the host.
472 	 */
473 	u16 is_dedicated_interrupt:1;
474 	u16 reserved1:15;
475 	u32 connection_id;
476 } __packed;
477 
478 /* Rescind Offer parameters */
479 struct vmbus_channel_rescind_offer {
480 	struct vmbus_channel_message_header header;
481 	u32 child_relid;
482 } __packed;
483 
484 static inline u32
485 hv_ringbuffer_pending_size(const struct hv_ring_buffer_info *rbi)
486 {
487 	return rbi->ring_buffer->pending_send_sz;
488 }
489 
490 /*
491  * Request Offer -- no parameters, SynIC message contains the partition ID
492  * Set Snoop -- no parameters, SynIC message contains the partition ID
493  * Clear Snoop -- no parameters, SynIC message contains the partition ID
494  * All Offers Delivered -- no parameters, SynIC message contains the partition
495  *		           ID
496  * Flush Client -- no parameters, SynIC message contains the partition ID
497  */
498 
499 /* Open Channel parameters */
500 struct vmbus_channel_open_channel {
501 	struct vmbus_channel_message_header header;
502 
503 	/* Identifies the specific VMBus channel that is being opened. */
504 	u32 child_relid;
505 
506 	/* ID making a particular open request at a channel offer unique. */
507 	u32 openid;
508 
509 	/* GPADL for the channel's ring buffer. */
510 	u32 ringbuffer_gpadlhandle;
511 
512 	/*
513 	 * Starting with win8, this field will be used to specify
514 	 * the target virtual processor on which to deliver the interrupt for
515 	 * the host to guest communication.
516 	 * Prior to win8, incoming channel interrupts would only
517 	 * be delivered on cpu 0. Setting this value to 0 would
518 	 * preserve the earlier behavior.
519 	 */
520 	u32 target_vp;
521 
522 	/*
523 	 * The upstream ring buffer begins at offset zero in the memory
524 	 * described by RingBufferGpadlHandle. The downstream ring buffer
525 	 * follows it at this offset (in pages).
526 	 */
527 	u32 downstream_ringbuffer_pageoffset;
528 
529 	/* User-specific data to be passed along to the server endpoint. */
530 	unsigned char userdata[MAX_USER_DEFINED_BYTES];
531 } __packed;
532 
533 /* Open Channel Result parameters */
534 struct vmbus_channel_open_result {
535 	struct vmbus_channel_message_header header;
536 	u32 child_relid;
537 	u32 openid;
538 	u32 status;
539 } __packed;
540 
541 /* Close channel parameters; */
542 struct vmbus_channel_close_channel {
543 	struct vmbus_channel_message_header header;
544 	u32 child_relid;
545 } __packed;
546 
547 /* Channel Message GPADL */
548 #define GPADL_TYPE_RING_BUFFER		1
549 #define GPADL_TYPE_SERVER_SAVE_AREA	2
550 #define GPADL_TYPE_TRANSACTION		8
551 
552 /*
553  * The number of PFNs in a GPADL message is defined by the number of
554  * pages that would be spanned by ByteCount and ByteOffset.  If the
555  * implied number of PFNs won't fit in this packet, there will be a
556  * follow-up packet that contains more.
557  */
558 struct vmbus_channel_gpadl_header {
559 	struct vmbus_channel_message_header header;
560 	u32 child_relid;
561 	u32 gpadl;
562 	u16 range_buflen;
563 	u16 rangecount;
564 	struct gpa_range range[0];
565 } __packed;
566 
567 /* This is the followup packet that contains more PFNs. */
568 struct vmbus_channel_gpadl_body {
569 	struct vmbus_channel_message_header header;
570 	u32 msgnumber;
571 	u32 gpadl;
572 	u64 pfn[0];
573 } __packed;
574 
575 struct vmbus_channel_gpadl_created {
576 	struct vmbus_channel_message_header header;
577 	u32 child_relid;
578 	u32 gpadl;
579 	u32 creation_status;
580 } __packed;
581 
582 struct vmbus_channel_gpadl_teardown {
583 	struct vmbus_channel_message_header header;
584 	u32 child_relid;
585 	u32 gpadl;
586 } __packed;
587 
588 struct vmbus_channel_gpadl_torndown {
589 	struct vmbus_channel_message_header header;
590 	u32 gpadl;
591 } __packed;
592 
593 struct vmbus_channel_relid_released {
594 	struct vmbus_channel_message_header header;
595 	u32 child_relid;
596 } __packed;
597 
598 struct vmbus_channel_initiate_contact {
599 	struct vmbus_channel_message_header header;
600 	u32 vmbus_version_requested;
601 	u32 target_vcpu; /* The VCPU the host should respond to */
602 	union {
603 		u64 interrupt_page;
604 		struct {
605 			u8	msg_sint;
606 			u8	padding1[3];
607 			u32	padding2;
608 		};
609 	};
610 	u64 monitor_page1;
611 	u64 monitor_page2;
612 } __packed;
613 
614 /* Hyper-V socket: guest's connect()-ing to host */
615 struct vmbus_channel_tl_connect_request {
616 	struct vmbus_channel_message_header header;
617 	uuid_le guest_endpoint_id;
618 	uuid_le host_service_id;
619 } __packed;
620 
621 struct vmbus_channel_version_response {
622 	struct vmbus_channel_message_header header;
623 	u8 version_supported;
624 
625 	u8 connection_state;
626 	u16 padding;
627 
628 	/*
629 	 * On new hosts that support VMBus protocol 5.0, we must use
630 	 * VMBUS_MESSAGE_CONNECTION_ID_4 for the Initiate Contact Message,
631 	 * and for subsequent messages, we must use the Message Connection ID
632 	 * field in the host-returned Version Response Message.
633 	 *
634 	 * On old hosts, we should always use VMBUS_MESSAGE_CONNECTION_ID (1).
635 	 */
636 	u32 msg_conn_id;
637 } __packed;
638 
639 enum vmbus_channel_state {
640 	CHANNEL_OFFER_STATE,
641 	CHANNEL_OPENING_STATE,
642 	CHANNEL_OPEN_STATE,
643 	CHANNEL_OPENED_STATE,
644 };
645 
646 /*
647  * Represents each channel msg on the vmbus connection This is a
648  * variable-size data structure depending on the msg type itself
649  */
650 struct vmbus_channel_msginfo {
651 	/* Bookkeeping stuff */
652 	struct list_head msglistentry;
653 
654 	/* So far, this is only used to handle gpadl body message */
655 	struct list_head submsglist;
656 
657 	/* Synchronize the request/response if needed */
658 	struct completion  waitevent;
659 	struct vmbus_channel *waiting_channel;
660 	union {
661 		struct vmbus_channel_version_supported version_supported;
662 		struct vmbus_channel_open_result open_result;
663 		struct vmbus_channel_gpadl_torndown gpadl_torndown;
664 		struct vmbus_channel_gpadl_created gpadl_created;
665 		struct vmbus_channel_version_response version_response;
666 	} response;
667 
668 	u32 msgsize;
669 	/*
670 	 * The channel message that goes out on the "wire".
671 	 * It will contain at minimum the VMBUS_CHANNEL_MESSAGE_HEADER header
672 	 */
673 	unsigned char msg[0];
674 };
675 
676 struct vmbus_close_msg {
677 	struct vmbus_channel_msginfo info;
678 	struct vmbus_channel_close_channel msg;
679 };
680 
681 /* Define connection identifier type. */
682 union hv_connection_id {
683 	u32 asu32;
684 	struct {
685 		u32 id:24;
686 		u32 reserved:8;
687 	} u;
688 };
689 
690 enum hv_numa_policy {
691 	HV_BALANCED = 0,
692 	HV_LOCALIZED,
693 };
694 
695 enum vmbus_device_type {
696 	HV_IDE = 0,
697 	HV_SCSI,
698 	HV_FC,
699 	HV_NIC,
700 	HV_ND,
701 	HV_PCIE,
702 	HV_FB,
703 	HV_KBD,
704 	HV_MOUSE,
705 	HV_KVP,
706 	HV_TS,
707 	HV_HB,
708 	HV_SHUTDOWN,
709 	HV_FCOPY,
710 	HV_BACKUP,
711 	HV_DM,
712 	HV_UNKNOWN,
713 };
714 
715 struct vmbus_device {
716 	u16  dev_type;
717 	uuid_le guid;
718 	bool perf_device;
719 };
720 
721 struct vmbus_channel {
722 	struct list_head listentry;
723 
724 	struct hv_device *device_obj;
725 
726 	enum vmbus_channel_state state;
727 
728 	struct vmbus_channel_offer_channel offermsg;
729 	/*
730 	 * These are based on the OfferMsg.MonitorId.
731 	 * Save it here for easy access.
732 	 */
733 	u8 monitor_grp;
734 	u8 monitor_bit;
735 
736 	bool rescind; /* got rescind msg */
737 	struct completion rescind_event;
738 
739 	u32 ringbuffer_gpadlhandle;
740 
741 	/* Allocated memory for ring buffer */
742 	void *ringbuffer_pages;
743 	u32 ringbuffer_pagecount;
744 	struct hv_ring_buffer_info outbound;	/* send to parent */
745 	struct hv_ring_buffer_info inbound;	/* receive from parent */
746 
747 	struct vmbus_close_msg close_msg;
748 
749 	/* Statistics */
750 	u64	interrupts;	/* Host to Guest interrupts */
751 	u64	sig_events;	/* Guest to Host events */
752 
753 	/* Channel callback's invoked in softirq context */
754 	struct tasklet_struct callback_event;
755 	void (*onchannel_callback)(void *context);
756 	void *channel_callback_context;
757 
758 	/*
759 	 * A channel can be marked for one of three modes of reading:
760 	 *   BATCHED - callback called from taslket and should read
761 	 *            channel until empty. Interrupts from the host
762 	 *            are masked while read is in process (default).
763 	 *   DIRECT - callback called from tasklet (softirq).
764 	 *   ISR - callback called in interrupt context and must
765 	 *         invoke its own deferred processing.
766 	 *         Host interrupts are disabled and must be re-enabled
767 	 *         when ring is empty.
768 	 */
769 	enum hv_callback_mode {
770 		HV_CALL_BATCHED,
771 		HV_CALL_DIRECT,
772 		HV_CALL_ISR
773 	} callback_mode;
774 
775 	bool is_dedicated_interrupt;
776 	u64 sig_event;
777 
778 	/*
779 	 * Starting with win8, this field will be used to specify
780 	 * the target virtual processor on which to deliver the interrupt for
781 	 * the host to guest communication.
782 	 * Prior to win8, incoming channel interrupts would only
783 	 * be delivered on cpu 0. Setting this value to 0 would
784 	 * preserve the earlier behavior.
785 	 */
786 	u32 target_vp;
787 	/* The corresponding CPUID in the guest */
788 	u32 target_cpu;
789 	/*
790 	 * State to manage the CPU affiliation of channels.
791 	 */
792 	struct cpumask alloced_cpus_in_node;
793 	int numa_node;
794 	/*
795 	 * Support for sub-channels. For high performance devices,
796 	 * it will be useful to have multiple sub-channels to support
797 	 * a scalable communication infrastructure with the host.
798 	 * The support for sub-channels is implemented as an extention
799 	 * to the current infrastructure.
800 	 * The initial offer is considered the primary channel and this
801 	 * offer message will indicate if the host supports sub-channels.
802 	 * The guest is free to ask for sub-channels to be offerred and can
803 	 * open these sub-channels as a normal "primary" channel. However,
804 	 * all sub-channels will have the same type and instance guids as the
805 	 * primary channel. Requests sent on a given channel will result in a
806 	 * response on the same channel.
807 	 */
808 
809 	/*
810 	 * Sub-channel creation callback. This callback will be called in
811 	 * process context when a sub-channel offer is received from the host.
812 	 * The guest can open the sub-channel in the context of this callback.
813 	 */
814 	void (*sc_creation_callback)(struct vmbus_channel *new_sc);
815 
816 	/*
817 	 * Channel rescind callback. Some channels (the hvsock ones), need to
818 	 * register a callback which is invoked in vmbus_onoffer_rescind().
819 	 */
820 	void (*chn_rescind_callback)(struct vmbus_channel *channel);
821 
822 	/*
823 	 * The spinlock to protect the structure. It is being used to protect
824 	 * test-and-set access to various attributes of the structure as well
825 	 * as all sc_list operations.
826 	 */
827 	spinlock_t lock;
828 	/*
829 	 * All Sub-channels of a primary channel are linked here.
830 	 */
831 	struct list_head sc_list;
832 	/*
833 	 * Current number of sub-channels.
834 	 */
835 	int num_sc;
836 	/*
837 	 * Number of a sub-channel (position within sc_list) which is supposed
838 	 * to be used as the next outgoing channel.
839 	 */
840 	int next_oc;
841 	/*
842 	 * The primary channel this sub-channel belongs to.
843 	 * This will be NULL for the primary channel.
844 	 */
845 	struct vmbus_channel *primary_channel;
846 	/*
847 	 * Support per-channel state for use by vmbus drivers.
848 	 */
849 	void *per_channel_state;
850 	/*
851 	 * To support per-cpu lookup mapping of relid to channel,
852 	 * link up channels based on their CPU affinity.
853 	 */
854 	struct list_head percpu_list;
855 
856 	/*
857 	 * Defer freeing channel until after all cpu's have
858 	 * gone through grace period.
859 	 */
860 	struct rcu_head rcu;
861 
862 	/*
863 	 * For sysfs per-channel properties.
864 	 */
865 	struct kobject			kobj;
866 
867 	/*
868 	 * For performance critical channels (storage, networking
869 	 * etc,), Hyper-V has a mechanism to enhance the throughput
870 	 * at the expense of latency:
871 	 * When the host is to be signaled, we just set a bit in a shared page
872 	 * and this bit will be inspected by the hypervisor within a certain
873 	 * window and if the bit is set, the host will be signaled. The window
874 	 * of time is the monitor latency - currently around 100 usecs. This
875 	 * mechanism improves throughput by:
876 	 *
877 	 * A) Making the host more efficient - each time it wakes up,
878 	 *    potentially it will process morev number of packets. The
879 	 *    monitor latency allows a batch to build up.
880 	 * B) By deferring the hypercall to signal, we will also minimize
881 	 *    the interrupts.
882 	 *
883 	 * Clearly, these optimizations improve throughput at the expense of
884 	 * latency. Furthermore, since the channel is shared for both
885 	 * control and data messages, control messages currently suffer
886 	 * unnecessary latency adversley impacting performance and boot
887 	 * time. To fix this issue, permit tagging the channel as being
888 	 * in "low latency" mode. In this mode, we will bypass the monitor
889 	 * mechanism.
890 	 */
891 	bool low_latency;
892 
893 	/*
894 	 * NUMA distribution policy:
895 	 * We support two policies:
896 	 * 1) Balanced: Here all performance critical channels are
897 	 *    distributed evenly amongst all the NUMA nodes.
898 	 *    This policy will be the default policy.
899 	 * 2) Localized: All channels of a given instance of a
900 	 *    performance critical service will be assigned CPUs
901 	 *    within a selected NUMA node.
902 	 */
903 	enum hv_numa_policy affinity_policy;
904 
905 	bool probe_done;
906 
907 };
908 
909 static inline bool is_hvsock_channel(const struct vmbus_channel *c)
910 {
911 	return !!(c->offermsg.offer.chn_flags &
912 		  VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER);
913 }
914 
915 static inline void set_channel_affinity_state(struct vmbus_channel *c,
916 					      enum hv_numa_policy policy)
917 {
918 	c->affinity_policy = policy;
919 }
920 
921 static inline void set_channel_read_mode(struct vmbus_channel *c,
922 					enum hv_callback_mode mode)
923 {
924 	c->callback_mode = mode;
925 }
926 
927 static inline void set_per_channel_state(struct vmbus_channel *c, void *s)
928 {
929 	c->per_channel_state = s;
930 }
931 
932 static inline void *get_per_channel_state(struct vmbus_channel *c)
933 {
934 	return c->per_channel_state;
935 }
936 
937 static inline void set_channel_pending_send_size(struct vmbus_channel *c,
938 						 u32 size)
939 {
940 	c->outbound.ring_buffer->pending_send_sz = size;
941 }
942 
943 static inline void set_low_latency_mode(struct vmbus_channel *c)
944 {
945 	c->low_latency = true;
946 }
947 
948 static inline void clear_low_latency_mode(struct vmbus_channel *c)
949 {
950 	c->low_latency = false;
951 }
952 
953 void vmbus_onmessage(void *context);
954 
955 int vmbus_request_offers(void);
956 
957 /*
958  * APIs for managing sub-channels.
959  */
960 
961 void vmbus_set_sc_create_callback(struct vmbus_channel *primary_channel,
962 			void (*sc_cr_cb)(struct vmbus_channel *new_sc));
963 
964 void vmbus_set_chn_rescind_callback(struct vmbus_channel *channel,
965 		void (*chn_rescind_cb)(struct vmbus_channel *));
966 
967 /*
968  * Retrieve the (sub) channel on which to send an outgoing request.
969  * When a primary channel has multiple sub-channels, we choose a
970  * channel whose VCPU binding is closest to the VCPU on which
971  * this call is being made.
972  */
973 struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary);
974 
975 /*
976  * Check if sub-channels have already been offerred. This API will be useful
977  * when the driver is unloaded after establishing sub-channels. In this case,
978  * when the driver is re-loaded, the driver would have to check if the
979  * subchannels have already been established before attempting to request
980  * the creation of sub-channels.
981  * This function returns TRUE to indicate that subchannels have already been
982  * created.
983  * This function should be invoked after setting the callback function for
984  * sub-channel creation.
985  */
986 bool vmbus_are_subchannels_present(struct vmbus_channel *primary);
987 
988 /* The format must be the same as struct vmdata_gpa_direct */
989 struct vmbus_channel_packet_page_buffer {
990 	u16 type;
991 	u16 dataoffset8;
992 	u16 length8;
993 	u16 flags;
994 	u64 transactionid;
995 	u32 reserved;
996 	u32 rangecount;
997 	struct hv_page_buffer range[MAX_PAGE_BUFFER_COUNT];
998 } __packed;
999 
1000 /* The format must be the same as struct vmdata_gpa_direct */
1001 struct vmbus_channel_packet_multipage_buffer {
1002 	u16 type;
1003 	u16 dataoffset8;
1004 	u16 length8;
1005 	u16 flags;
1006 	u64 transactionid;
1007 	u32 reserved;
1008 	u32 rangecount;		/* Always 1 in this case */
1009 	struct hv_multipage_buffer range;
1010 } __packed;
1011 
1012 /* The format must be the same as struct vmdata_gpa_direct */
1013 struct vmbus_packet_mpb_array {
1014 	u16 type;
1015 	u16 dataoffset8;
1016 	u16 length8;
1017 	u16 flags;
1018 	u64 transactionid;
1019 	u32 reserved;
1020 	u32 rangecount;         /* Always 1 in this case */
1021 	struct hv_mpb_array range;
1022 } __packed;
1023 
1024 
1025 extern int vmbus_open(struct vmbus_channel *channel,
1026 			    u32 send_ringbuffersize,
1027 			    u32 recv_ringbuffersize,
1028 			    void *userdata,
1029 			    u32 userdatalen,
1030 			    void (*onchannel_callback)(void *context),
1031 			    void *context);
1032 
1033 extern void vmbus_close(struct vmbus_channel *channel);
1034 
1035 extern int vmbus_sendpacket(struct vmbus_channel *channel,
1036 				  void *buffer,
1037 				  u32 bufferLen,
1038 				  u64 requestid,
1039 				  enum vmbus_packet_type type,
1040 				  u32 flags);
1041 
1042 extern int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
1043 					    struct hv_page_buffer pagebuffers[],
1044 					    u32 pagecount,
1045 					    void *buffer,
1046 					    u32 bufferlen,
1047 					    u64 requestid);
1048 
1049 extern int vmbus_sendpacket_mpb_desc(struct vmbus_channel *channel,
1050 				     struct vmbus_packet_mpb_array *mpb,
1051 				     u32 desc_size,
1052 				     void *buffer,
1053 				     u32 bufferlen,
1054 				     u64 requestid);
1055 
1056 extern int vmbus_establish_gpadl(struct vmbus_channel *channel,
1057 				      void *kbuffer,
1058 				      u32 size,
1059 				      u32 *gpadl_handle);
1060 
1061 extern int vmbus_teardown_gpadl(struct vmbus_channel *channel,
1062 				     u32 gpadl_handle);
1063 
1064 void vmbus_reset_channel_cb(struct vmbus_channel *channel);
1065 
1066 extern int vmbus_recvpacket(struct vmbus_channel *channel,
1067 				  void *buffer,
1068 				  u32 bufferlen,
1069 				  u32 *buffer_actual_len,
1070 				  u64 *requestid);
1071 
1072 extern int vmbus_recvpacket_raw(struct vmbus_channel *channel,
1073 				     void *buffer,
1074 				     u32 bufferlen,
1075 				     u32 *buffer_actual_len,
1076 				     u64 *requestid);
1077 
1078 
1079 extern void vmbus_ontimer(unsigned long data);
1080 
1081 /* Base driver object */
1082 struct hv_driver {
1083 	const char *name;
1084 
1085 	/*
1086 	 * A hvsock offer, which has a VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER
1087 	 * channel flag, actually doesn't mean a synthetic device because the
1088 	 * offer's if_type/if_instance can change for every new hvsock
1089 	 * connection.
1090 	 *
1091 	 * However, to facilitate the notification of new-offer/rescind-offer
1092 	 * from vmbus driver to hvsock driver, we can handle hvsock offer as
1093 	 * a special vmbus device, and hence we need the below flag to
1094 	 * indicate if the driver is the hvsock driver or not: we need to
1095 	 * specially treat the hvosck offer & driver in vmbus_match().
1096 	 */
1097 	bool hvsock;
1098 
1099 	/* the device type supported by this driver */
1100 	uuid_le dev_type;
1101 	const struct hv_vmbus_device_id *id_table;
1102 
1103 	struct device_driver driver;
1104 
1105 	/* dynamic device GUID's */
1106 	struct  {
1107 		spinlock_t lock;
1108 		struct list_head list;
1109 	} dynids;
1110 
1111 	int (*probe)(struct hv_device *, const struct hv_vmbus_device_id *);
1112 	int (*remove)(struct hv_device *);
1113 	void (*shutdown)(struct hv_device *);
1114 
1115 };
1116 
1117 /* Base device object */
1118 struct hv_device {
1119 	/* the device type id of this device */
1120 	uuid_le dev_type;
1121 
1122 	/* the device instance id of this device */
1123 	uuid_le dev_instance;
1124 	u16 vendor_id;
1125 	u16 device_id;
1126 
1127 	struct device device;
1128 	char *driver_override; /* Driver name to force a match */
1129 
1130 	struct vmbus_channel *channel;
1131 	struct kset	     *channels_kset;
1132 };
1133 
1134 
1135 static inline struct hv_device *device_to_hv_device(struct device *d)
1136 {
1137 	return container_of(d, struct hv_device, device);
1138 }
1139 
1140 static inline struct hv_driver *drv_to_hv_drv(struct device_driver *d)
1141 {
1142 	return container_of(d, struct hv_driver, driver);
1143 }
1144 
1145 static inline void hv_set_drvdata(struct hv_device *dev, void *data)
1146 {
1147 	dev_set_drvdata(&dev->device, data);
1148 }
1149 
1150 static inline void *hv_get_drvdata(struct hv_device *dev)
1151 {
1152 	return dev_get_drvdata(&dev->device);
1153 }
1154 
1155 struct hv_ring_buffer_debug_info {
1156 	u32 current_interrupt_mask;
1157 	u32 current_read_index;
1158 	u32 current_write_index;
1159 	u32 bytes_avail_toread;
1160 	u32 bytes_avail_towrite;
1161 };
1162 
1163 void hv_ringbuffer_get_debuginfo(const struct hv_ring_buffer_info *ring_info,
1164 			    struct hv_ring_buffer_debug_info *debug_info);
1165 
1166 /* Vmbus interface */
1167 #define vmbus_driver_register(driver)	\
1168 	__vmbus_driver_register(driver, THIS_MODULE, KBUILD_MODNAME)
1169 int __must_check __vmbus_driver_register(struct hv_driver *hv_driver,
1170 					 struct module *owner,
1171 					 const char *mod_name);
1172 void vmbus_driver_unregister(struct hv_driver *hv_driver);
1173 
1174 void vmbus_hvsock_device_unregister(struct vmbus_channel *channel);
1175 
1176 int vmbus_allocate_mmio(struct resource **new, struct hv_device *device_obj,
1177 			resource_size_t min, resource_size_t max,
1178 			resource_size_t size, resource_size_t align,
1179 			bool fb_overlap_ok);
1180 void vmbus_free_mmio(resource_size_t start, resource_size_t size);
1181 
1182 /*
1183  * GUID definitions of various offer types - services offered to the guest.
1184  */
1185 
1186 /*
1187  * Network GUID
1188  * {f8615163-df3e-46c5-913f-f2d2f965ed0e}
1189  */
1190 #define HV_NIC_GUID \
1191 	.guid = UUID_LE(0xf8615163, 0xdf3e, 0x46c5, 0x91, 0x3f, \
1192 			0xf2, 0xd2, 0xf9, 0x65, 0xed, 0x0e)
1193 
1194 /*
1195  * IDE GUID
1196  * {32412632-86cb-44a2-9b5c-50d1417354f5}
1197  */
1198 #define HV_IDE_GUID \
1199 	.guid = UUID_LE(0x32412632, 0x86cb, 0x44a2, 0x9b, 0x5c, \
1200 			0x50, 0xd1, 0x41, 0x73, 0x54, 0xf5)
1201 
1202 /*
1203  * SCSI GUID
1204  * {ba6163d9-04a1-4d29-b605-72e2ffb1dc7f}
1205  */
1206 #define HV_SCSI_GUID \
1207 	.guid = UUID_LE(0xba6163d9, 0x04a1, 0x4d29, 0xb6, 0x05, \
1208 			0x72, 0xe2, 0xff, 0xb1, 0xdc, 0x7f)
1209 
1210 /*
1211  * Shutdown GUID
1212  * {0e0b6031-5213-4934-818b-38d90ced39db}
1213  */
1214 #define HV_SHUTDOWN_GUID \
1215 	.guid = UUID_LE(0x0e0b6031, 0x5213, 0x4934, 0x81, 0x8b, \
1216 			0x38, 0xd9, 0x0c, 0xed, 0x39, 0xdb)
1217 
1218 /*
1219  * Time Synch GUID
1220  * {9527E630-D0AE-497b-ADCE-E80AB0175CAF}
1221  */
1222 #define HV_TS_GUID \
1223 	.guid = UUID_LE(0x9527e630, 0xd0ae, 0x497b, 0xad, 0xce, \
1224 			0xe8, 0x0a, 0xb0, 0x17, 0x5c, 0xaf)
1225 
1226 /*
1227  * Heartbeat GUID
1228  * {57164f39-9115-4e78-ab55-382f3bd5422d}
1229  */
1230 #define HV_HEART_BEAT_GUID \
1231 	.guid = UUID_LE(0x57164f39, 0x9115, 0x4e78, 0xab, 0x55, \
1232 			0x38, 0x2f, 0x3b, 0xd5, 0x42, 0x2d)
1233 
1234 /*
1235  * KVP GUID
1236  * {a9a0f4e7-5a45-4d96-b827-8a841e8c03e6}
1237  */
1238 #define HV_KVP_GUID \
1239 	.guid = UUID_LE(0xa9a0f4e7, 0x5a45, 0x4d96, 0xb8, 0x27, \
1240 			0x8a, 0x84, 0x1e, 0x8c, 0x03, 0xe6)
1241 
1242 /*
1243  * Dynamic memory GUID
1244  * {525074dc-8985-46e2-8057-a307dc18a502}
1245  */
1246 #define HV_DM_GUID \
1247 	.guid = UUID_LE(0x525074dc, 0x8985, 0x46e2, 0x80, 0x57, \
1248 			0xa3, 0x07, 0xdc, 0x18, 0xa5, 0x02)
1249 
1250 /*
1251  * Mouse GUID
1252  * {cfa8b69e-5b4a-4cc0-b98b-8ba1a1f3f95a}
1253  */
1254 #define HV_MOUSE_GUID \
1255 	.guid = UUID_LE(0xcfa8b69e, 0x5b4a, 0x4cc0, 0xb9, 0x8b, \
1256 			0x8b, 0xa1, 0xa1, 0xf3, 0xf9, 0x5a)
1257 
1258 /*
1259  * Keyboard GUID
1260  * {f912ad6d-2b17-48ea-bd65-f927a61c7684}
1261  */
1262 #define HV_KBD_GUID \
1263 	.guid = UUID_LE(0xf912ad6d, 0x2b17, 0x48ea, 0xbd, 0x65, \
1264 			0xf9, 0x27, 0xa6, 0x1c, 0x76, 0x84)
1265 
1266 /*
1267  * VSS (Backup/Restore) GUID
1268  */
1269 #define HV_VSS_GUID \
1270 	.guid = UUID_LE(0x35fa2e29, 0xea23, 0x4236, 0x96, 0xae, \
1271 			0x3a, 0x6e, 0xba, 0xcb, 0xa4, 0x40)
1272 /*
1273  * Synthetic Video GUID
1274  * {DA0A7802-E377-4aac-8E77-0558EB1073F8}
1275  */
1276 #define HV_SYNTHVID_GUID \
1277 	.guid = UUID_LE(0xda0a7802, 0xe377, 0x4aac, 0x8e, 0x77, \
1278 			0x05, 0x58, 0xeb, 0x10, 0x73, 0xf8)
1279 
1280 /*
1281  * Synthetic FC GUID
1282  * {2f9bcc4a-0069-4af3-b76b-6fd0be528cda}
1283  */
1284 #define HV_SYNTHFC_GUID \
1285 	.guid = UUID_LE(0x2f9bcc4a, 0x0069, 0x4af3, 0xb7, 0x6b, \
1286 			0x6f, 0xd0, 0xbe, 0x52, 0x8c, 0xda)
1287 
1288 /*
1289  * Guest File Copy Service
1290  * {34D14BE3-DEE4-41c8-9AE7-6B174977C192}
1291  */
1292 
1293 #define HV_FCOPY_GUID \
1294 	.guid = UUID_LE(0x34d14be3, 0xdee4, 0x41c8, 0x9a, 0xe7, \
1295 			0x6b, 0x17, 0x49, 0x77, 0xc1, 0x92)
1296 
1297 /*
1298  * NetworkDirect. This is the guest RDMA service.
1299  * {8c2eaf3d-32a7-4b09-ab99-bd1f1c86b501}
1300  */
1301 #define HV_ND_GUID \
1302 	.guid = UUID_LE(0x8c2eaf3d, 0x32a7, 0x4b09, 0xab, 0x99, \
1303 			0xbd, 0x1f, 0x1c, 0x86, 0xb5, 0x01)
1304 
1305 /*
1306  * PCI Express Pass Through
1307  * {44C4F61D-4444-4400-9D52-802E27EDE19F}
1308  */
1309 
1310 #define HV_PCIE_GUID \
1311 	.guid = UUID_LE(0x44c4f61d, 0x4444, 0x4400, 0x9d, 0x52, \
1312 			0x80, 0x2e, 0x27, 0xed, 0xe1, 0x9f)
1313 
1314 /*
1315  * Linux doesn't support the 3 devices: the first two are for
1316  * Automatic Virtual Machine Activation, and the third is for
1317  * Remote Desktop Virtualization.
1318  * {f8e65716-3cb3-4a06-9a60-1889c5cccab5}
1319  * {3375baf4-9e15-4b30-b765-67acb10d607b}
1320  * {276aacf4-ac15-426c-98dd-7521ad3f01fe}
1321  */
1322 
1323 #define HV_AVMA1_GUID \
1324 	.guid = UUID_LE(0xf8e65716, 0x3cb3, 0x4a06, 0x9a, 0x60, \
1325 			0x18, 0x89, 0xc5, 0xcc, 0xca, 0xb5)
1326 
1327 #define HV_AVMA2_GUID \
1328 	.guid = UUID_LE(0x3375baf4, 0x9e15, 0x4b30, 0xb7, 0x65, \
1329 			0x67, 0xac, 0xb1, 0x0d, 0x60, 0x7b)
1330 
1331 #define HV_RDV_GUID \
1332 	.guid = UUID_LE(0x276aacf4, 0xac15, 0x426c, 0x98, 0xdd, \
1333 			0x75, 0x21, 0xad, 0x3f, 0x01, 0xfe)
1334 
1335 /*
1336  * Common header for Hyper-V ICs
1337  */
1338 
1339 #define ICMSGTYPE_NEGOTIATE		0
1340 #define ICMSGTYPE_HEARTBEAT		1
1341 #define ICMSGTYPE_KVPEXCHANGE		2
1342 #define ICMSGTYPE_SHUTDOWN		3
1343 #define ICMSGTYPE_TIMESYNC		4
1344 #define ICMSGTYPE_VSS			5
1345 
1346 #define ICMSGHDRFLAG_TRANSACTION	1
1347 #define ICMSGHDRFLAG_REQUEST		2
1348 #define ICMSGHDRFLAG_RESPONSE		4
1349 
1350 
1351 /*
1352  * While we want to handle util services as regular devices,
1353  * there is only one instance of each of these services; so
1354  * we statically allocate the service specific state.
1355  */
1356 
1357 struct hv_util_service {
1358 	u8 *recv_buffer;
1359 	void *channel;
1360 	void (*util_cb)(void *);
1361 	int (*util_init)(struct hv_util_service *);
1362 	void (*util_deinit)(void);
1363 };
1364 
1365 struct vmbuspipe_hdr {
1366 	u32 flags;
1367 	u32 msgsize;
1368 } __packed;
1369 
1370 struct ic_version {
1371 	u16 major;
1372 	u16 minor;
1373 } __packed;
1374 
1375 struct icmsg_hdr {
1376 	struct ic_version icverframe;
1377 	u16 icmsgtype;
1378 	struct ic_version icvermsg;
1379 	u16 icmsgsize;
1380 	u32 status;
1381 	u8 ictransaction_id;
1382 	u8 icflags;
1383 	u8 reserved[2];
1384 } __packed;
1385 
1386 struct icmsg_negotiate {
1387 	u16 icframe_vercnt;
1388 	u16 icmsg_vercnt;
1389 	u32 reserved;
1390 	struct ic_version icversion_data[1]; /* any size array */
1391 } __packed;
1392 
1393 struct shutdown_msg_data {
1394 	u32 reason_code;
1395 	u32 timeout_seconds;
1396 	u32 flags;
1397 	u8  display_message[2048];
1398 } __packed;
1399 
1400 struct heartbeat_msg_data {
1401 	u64 seq_num;
1402 	u32 reserved[8];
1403 } __packed;
1404 
1405 /* Time Sync IC defs */
1406 #define ICTIMESYNCFLAG_PROBE	0
1407 #define ICTIMESYNCFLAG_SYNC	1
1408 #define ICTIMESYNCFLAG_SAMPLE	2
1409 
1410 #ifdef __x86_64__
1411 #define WLTIMEDELTA	116444736000000000L	/* in 100ns unit */
1412 #else
1413 #define WLTIMEDELTA	116444736000000000LL
1414 #endif
1415 
1416 struct ictimesync_data {
1417 	u64 parenttime;
1418 	u64 childtime;
1419 	u64 roundtriptime;
1420 	u8 flags;
1421 } __packed;
1422 
1423 struct ictimesync_ref_data {
1424 	u64 parenttime;
1425 	u64 vmreferencetime;
1426 	u8 flags;
1427 	char leapflags;
1428 	char stratum;
1429 	u8 reserved[3];
1430 } __packed;
1431 
1432 struct hyperv_service_callback {
1433 	u8 msg_type;
1434 	char *log_msg;
1435 	uuid_le data;
1436 	struct vmbus_channel *channel;
1437 	void (*callback)(void *context);
1438 };
1439 
1440 #define MAX_SRV_VER	0x7ffffff
1441 extern bool vmbus_prep_negotiate_resp(struct icmsg_hdr *icmsghdrp, u8 *buf,
1442 				const int *fw_version, int fw_vercnt,
1443 				const int *srv_version, int srv_vercnt,
1444 				int *nego_fw_version, int *nego_srv_version);
1445 
1446 void hv_process_channel_removal(u32 relid);
1447 
1448 void vmbus_setevent(struct vmbus_channel *channel);
1449 /*
1450  * Negotiated version with the Host.
1451  */
1452 
1453 extern __u32 vmbus_proto_version;
1454 
1455 int vmbus_send_tl_connect_request(const uuid_le *shv_guest_servie_id,
1456 				  const uuid_le *shv_host_servie_id);
1457 void vmbus_set_event(struct vmbus_channel *channel);
1458 
1459 /* Get the start of the ring buffer. */
1460 static inline void *
1461 hv_get_ring_buffer(const struct hv_ring_buffer_info *ring_info)
1462 {
1463 	return ring_info->ring_buffer->buffer;
1464 }
1465 
1466 /*
1467  * Mask off host interrupt callback notifications
1468  */
1469 static inline void hv_begin_read(struct hv_ring_buffer_info *rbi)
1470 {
1471 	rbi->ring_buffer->interrupt_mask = 1;
1472 
1473 	/* make sure mask update is not reordered */
1474 	virt_mb();
1475 }
1476 
1477 /*
1478  * Re-enable host callback and return number of outstanding bytes
1479  */
1480 static inline u32 hv_end_read(struct hv_ring_buffer_info *rbi)
1481 {
1482 
1483 	rbi->ring_buffer->interrupt_mask = 0;
1484 
1485 	/* make sure mask update is not reordered */
1486 	virt_mb();
1487 
1488 	/*
1489 	 * Now check to see if the ring buffer is still empty.
1490 	 * If it is not, we raced and we need to process new
1491 	 * incoming messages.
1492 	 */
1493 	return hv_get_bytes_to_read(rbi);
1494 }
1495 
1496 /*
1497  * An API to support in-place processing of incoming VMBUS packets.
1498  */
1499 
1500 /* Get data payload associated with descriptor */
1501 static inline void *hv_pkt_data(const struct vmpacket_descriptor *desc)
1502 {
1503 	return (void *)((unsigned long)desc + (desc->offset8 << 3));
1504 }
1505 
1506 /* Get data size associated with descriptor */
1507 static inline u32 hv_pkt_datalen(const struct vmpacket_descriptor *desc)
1508 {
1509 	return (desc->len8 << 3) - (desc->offset8 << 3);
1510 }
1511 
1512 
1513 struct vmpacket_descriptor *
1514 hv_pkt_iter_first(struct vmbus_channel *channel);
1515 
1516 struct vmpacket_descriptor *
1517 __hv_pkt_iter_next(struct vmbus_channel *channel,
1518 		   const struct vmpacket_descriptor *pkt);
1519 
1520 void hv_pkt_iter_close(struct vmbus_channel *channel);
1521 
1522 /*
1523  * Get next packet descriptor from iterator
1524  * If at end of list, return NULL and update host.
1525  */
1526 static inline struct vmpacket_descriptor *
1527 hv_pkt_iter_next(struct vmbus_channel *channel,
1528 		 const struct vmpacket_descriptor *pkt)
1529 {
1530 	struct vmpacket_descriptor *nxt;
1531 
1532 	nxt = __hv_pkt_iter_next(channel, pkt);
1533 	if (!nxt)
1534 		hv_pkt_iter_close(channel);
1535 
1536 	return nxt;
1537 }
1538 
1539 #define foreach_vmbus_pkt(pkt, channel) \
1540 	for (pkt = hv_pkt_iter_first(channel); pkt; \
1541 	    pkt = hv_pkt_iter_next(channel, pkt))
1542 
1543 #endif /* _HYPERV_H */
1544