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 <haiyangz@microsoft.com>
20 * Hank Janssen <hjanssen@microsoft.com>
21 * K. Y. Srinivasan <kys@microsoft.com>
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 */
46struct hv_page_buffer {
47 u32 len;
48 u32 offset;
49 u64 pfn;
50};
51
52/* Multiple-page buffer */
53struct 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 */
65struct 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
82struct 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
136struct 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
147static 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
161static 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
174static 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 */
224struct vmbus_channel_offer {
225 guid_t if_type;
226 guid_t 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
272struct vmpacket_descriptor {
273 u16 type;
274 u16 offset8;
275 u16 len8;
276 u16 flags;
277 u64 trans_id;
278} __packed;
279
280struct vmpacket_header {
281 u32 prev_pkt_start_offset;
282 struct vmpacket_descriptor descriptor;
283} __packed;
284
285struct vmtransfer_page_range {
286 u32 byte_count;
287 u32 byte_offset;
288} __packed;
289
290struct 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
299struct vmgpadl_packet_header {
300 struct vmpacket_descriptor d;
301 u32 gpadl;
302 u32 reserved;
303} __packed;
304
305struct 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 */
316struct 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 */
329struct 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 */
340struct 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 */
350struct 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. */
358struct 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
366union 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
387enum 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 */
408enum 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
434struct vmbus_channel_message_header {
435 enum vmbus_channel_message_type msgtype;
436 u32 padding;
437} __packed;
438
439/* Query VMBus Version parameters */
440struct vmbus_channel_query_vmbus_version {
441 struct vmbus_channel_message_header header;
442 u32 version;
443} __packed;
444
445/* VMBus Version Supported parameters */
446struct vmbus_channel_version_supported {
447 struct vmbus_channel_message_header header;
448 u8 version_supported;
449} __packed;
450
451/* Offer Channel parameters */
452struct 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 */
479struct vmbus_channel_rescind_offer {
480 struct vmbus_channel_message_header header;
481 u32 child_relid;
482} __packed;
483
484static inline u32
485hv_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 */
500struct 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 */
534struct 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; */
542struct 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 */
558struct 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. */
568struct vmbus_channel_gpadl_body {
569 struct vmbus_channel_message_header header;
570 u32 msgnumber;
571 u32 gpadl;
572 u64 pfn[0];
573} __packed;
574
575struct 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
582struct vmbus_channel_gpadl_teardown {
583 struct vmbus_channel_message_header header;
584 u32 child_relid;
585 u32 gpadl;
586} __packed;
587
588struct vmbus_channel_gpadl_torndown {
589 struct vmbus_channel_message_header header;
590 u32 gpadl;
591} __packed;
592
593struct vmbus_channel_relid_released {
594 struct vmbus_channel_message_header header;
595 u32 child_relid;
596} __packed;
597
598struct 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 */
615struct vmbus_channel_tl_connect_request {
616 struct vmbus_channel_message_header header;
617 guid_t guest_endpoint_id;
618 guid_t host_service_id;
619} __packed;
620
621struct 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
639enum 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 */
650struct 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
676struct vmbus_close_msg {
677 struct vmbus_channel_msginfo info;
678 struct vmbus_channel_close_channel msg;
679};
680
681/* Define connection identifier type. */
682union hv_connection_id {
683 u32 asu32;
684 struct {
685 u32 id:24;
686 u32 reserved:8;
687 } u;
688};
689
690enum hv_numa_policy {
691 HV_BALANCED = 0,
692 HV_LOCALIZED,
693};
694
695enum 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
715struct vmbus_device {
716 u16 dev_type;
717 guid_t guid;
718 bool perf_device;
719};
720
721struct 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 struct page *ringbuffer_page;
743 u32 ringbuffer_pagecount;
744 u32 ringbuffer_send_offset;
745 struct hv_ring_buffer_info outbound; /* send to parent */
746 struct hv_ring_buffer_info inbound; /* receive from parent */
747
748 struct vmbus_close_msg close_msg;
749
750 /* Statistics */
751 u64 interrupts; /* Host to Guest interrupts */
752 u64 sig_events; /* Guest to Host events */
753
754 /*
755 * Guest to host interrupts caused by the outbound ring buffer changing
756 * from empty to not empty.
757 */
758 u64 intr_out_empty;
759
760 /*
761 * Indicates that a full outbound ring buffer was encountered. The flag
762 * is set to true when a full outbound ring buffer is encountered and
763 * set to false when a write to the outbound ring buffer is completed.
764 */
765 bool out_full_flag;
766
767 /* Channel callback's invoked in softirq context */
768 struct tasklet_struct callback_event;
769 void (*onchannel_callback)(void *context);
770 void *channel_callback_context;
771
772 /*
773 * A channel can be marked for one of three modes of reading:
774 * BATCHED - callback called from taslket and should read
775 * channel until empty. Interrupts from the host
776 * are masked while read is in process (default).
777 * DIRECT - callback called from tasklet (softirq).
778 * ISR - callback called in interrupt context and must
779 * invoke its own deferred processing.
780 * Host interrupts are disabled and must be re-enabled
781 * when ring is empty.
782 */
783 enum hv_callback_mode {
784 HV_CALL_BATCHED,
785 HV_CALL_DIRECT,
786 HV_CALL_ISR
787 } callback_mode;
788
789 bool is_dedicated_interrupt;
790 u64 sig_event;
791
792 /*
793 * Starting with win8, this field will be used to specify
794 * the target virtual processor on which to deliver the interrupt for
795 * the host to guest communication.
796 * Prior to win8, incoming channel interrupts would only
797 * be delivered on cpu 0. Setting this value to 0 would
798 * preserve the earlier behavior.
799 */
800 u32 target_vp;
801 /* The corresponding CPUID in the guest */
802 u32 target_cpu;
803 /*
804 * State to manage the CPU affiliation of channels.
805 */
806 struct cpumask alloced_cpus_in_node;
807 int numa_node;
808 /*
809 * Support for sub-channels. For high performance devices,
810 * it will be useful to have multiple sub-channels to support
811 * a scalable communication infrastructure with the host.
812 * The support for sub-channels is implemented as an extention
813 * to the current infrastructure.
814 * The initial offer is considered the primary channel and this
815 * offer message will indicate if the host supports sub-channels.
816 * The guest is free to ask for sub-channels to be offerred and can
817 * open these sub-channels as a normal "primary" channel. However,
818 * all sub-channels will have the same type and instance guids as the
819 * primary channel. Requests sent on a given channel will result in a
820 * response on the same channel.
821 */
822
823 /*
824 * Sub-channel creation callback. This callback will be called in
825 * process context when a sub-channel offer is received from the host.
826 * The guest can open the sub-channel in the context of this callback.
827 */
828 void (*sc_creation_callback)(struct vmbus_channel *new_sc);
829
830 /*
831 * Channel rescind callback. Some channels (the hvsock ones), need to
832 * register a callback which is invoked in vmbus_onoffer_rescind().
833 */
834 void (*chn_rescind_callback)(struct vmbus_channel *channel);
835
836 /*
837 * The spinlock to protect the structure. It is being used to protect
838 * test-and-set access to various attributes of the structure as well
839 * as all sc_list operations.
840 */
841 spinlock_t lock;
842 /*
843 * All Sub-channels of a primary channel are linked here.
844 */
845 struct list_head sc_list;
846 /*
847 * The primary channel this sub-channel belongs to.
848 * This will be NULL for the primary channel.
849 */
850 struct vmbus_channel *primary_channel;
851 /*
852 * Support per-channel state for use by vmbus drivers.
853 */
854 void *per_channel_state;
855 /*
856 * To support per-cpu lookup mapping of relid to channel,
857 * link up channels based on their CPU affinity.
858 */
859 struct list_head percpu_list;
860
861 /*
862 * Defer freeing channel until after all cpu's have
863 * gone through grace period.
864 */
865 struct rcu_head rcu;
866
867 /*
868 * For sysfs per-channel properties.
869 */
870 struct kobject kobj;
871
872 /*
873 * For performance critical channels (storage, networking
874 * etc,), Hyper-V has a mechanism to enhance the throughput
875 * at the expense of latency:
876 * When the host is to be signaled, we just set a bit in a shared page
877 * and this bit will be inspected by the hypervisor within a certain
878 * window and if the bit is set, the host will be signaled. The window
879 * of time is the monitor latency - currently around 100 usecs. This
880 * mechanism improves throughput by:
881 *
882 * A) Making the host more efficient - each time it wakes up,
883 * potentially it will process morev number of packets. The
884 * monitor latency allows a batch to build up.
885 * B) By deferring the hypercall to signal, we will also minimize
886 * the interrupts.
887 *
888 * Clearly, these optimizations improve throughput at the expense of
889 * latency. Furthermore, since the channel is shared for both
890 * control and data messages, control messages currently suffer
891 * unnecessary latency adversley impacting performance and boot
892 * time. To fix this issue, permit tagging the channel as being
893 * in "low latency" mode. In this mode, we will bypass the monitor
894 * mechanism.
895 */
896 bool low_latency;
897
898 /*
899 * NUMA distribution policy:
900 * We support two policies:
901 * 1) Balanced: Here all performance critical channels are
902 * distributed evenly amongst all the NUMA nodes.
903 * This policy will be the default policy.
904 * 2) Localized: All channels of a given instance of a
905 * performance critical service will be assigned CPUs
906 * within a selected NUMA node.
907 */
908 enum hv_numa_policy affinity_policy;
909
910 bool probe_done;
911
912 /*
913 * We must offload the handling of the primary/sub channels
914 * from the single-threaded vmbus_connection.work_queue to
915 * two different workqueue, otherwise we can block
916 * vmbus_connection.work_queue and hang: see vmbus_process_offer().
917 */
918 struct work_struct add_channel_work;
919
920 /*
921 * Guest to host interrupts caused by the inbound ring buffer changing
922 * from full to not full while a packet is waiting.
923 */
924 u64 intr_in_full;
925
926 /*
927 * The total number of write operations that encountered a full
928 * outbound ring buffer.
929 */
930 u64 out_full_total;
931
932 /*
933 * The number of write operations that were the first to encounter a
934 * full outbound ring buffer.
935 */
936 u64 out_full_first;
937};
938
939static inline bool is_hvsock_channel(const struct vmbus_channel *c)
940{
941 return !!(c->offermsg.offer.chn_flags &
942 VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER);
943}
944
945static inline void set_channel_affinity_state(struct vmbus_channel *c,
946 enum hv_numa_policy policy)
947{
948 c->affinity_policy = policy;
949}
950
951static inline void set_channel_read_mode(struct vmbus_channel *c,
952 enum hv_callback_mode mode)
953{
954 c->callback_mode = mode;
955}
956
957static inline void set_per_channel_state(struct vmbus_channel *c, void *s)
958{
959 c->per_channel_state = s;
960}
961
962static inline void *get_per_channel_state(struct vmbus_channel *c)
963{
964 return c->per_channel_state;
965}
966
967static inline void set_channel_pending_send_size(struct vmbus_channel *c,
968 u32 size)
969{
970 unsigned long flags;
971
972 if (size) {
973 spin_lock_irqsave(&c->outbound.ring_lock, flags);
974 ++c->out_full_total;
975
976 if (!c->out_full_flag) {
977 ++c->out_full_first;
978 c->out_full_flag = true;
979 }
980 spin_unlock_irqrestore(&c->outbound.ring_lock, flags);
981 } else {
982 c->out_full_flag = false;
983 }
984
985 c->outbound.ring_buffer->pending_send_sz = size;
986}
987
988static inline void set_low_latency_mode(struct vmbus_channel *c)
989{
990 c->low_latency = true;
991}
992
993static inline void clear_low_latency_mode(struct vmbus_channel *c)
994{
995 c->low_latency = false;
996}
997
998void vmbus_onmessage(void *context);
999
1000int vmbus_request_offers(void);
1001
1002/*
1003 * APIs for managing sub-channels.
1004 */
1005
1006void vmbus_set_sc_create_callback(struct vmbus_channel *primary_channel,
1007 void (*sc_cr_cb)(struct vmbus_channel *new_sc));
1008
1009void vmbus_set_chn_rescind_callback(struct vmbus_channel *channel,
1010 void (*chn_rescind_cb)(struct vmbus_channel *));
1011
1012/*
1013 * Check if sub-channels have already been offerred. This API will be useful
1014 * when the driver is unloaded after establishing sub-channels. In this case,
1015 * when the driver is re-loaded, the driver would have to check if the
1016 * subchannels have already been established before attempting to request
1017 * the creation of sub-channels.
1018 * This function returns TRUE to indicate that subchannels have already been
1019 * created.
1020 * This function should be invoked after setting the callback function for
1021 * sub-channel creation.
1022 */
1023bool vmbus_are_subchannels_present(struct vmbus_channel *primary);
1024
1025/* The format must be the same as struct vmdata_gpa_direct */
1026struct vmbus_channel_packet_page_buffer {
1027 u16 type;
1028 u16 dataoffset8;
1029 u16 length8;
1030 u16 flags;
1031 u64 transactionid;
1032 u32 reserved;
1033 u32 rangecount;
1034 struct hv_page_buffer range[MAX_PAGE_BUFFER_COUNT];
1035} __packed;
1036
1037/* The format must be the same as struct vmdata_gpa_direct */
1038struct vmbus_channel_packet_multipage_buffer {
1039 u16 type;
1040 u16 dataoffset8;
1041 u16 length8;
1042 u16 flags;
1043 u64 transactionid;
1044 u32 reserved;
1045 u32 rangecount; /* Always 1 in this case */
1046 struct hv_multipage_buffer range;
1047} __packed;
1048
1049/* The format must be the same as struct vmdata_gpa_direct */
1050struct vmbus_packet_mpb_array {
1051 u16 type;
1052 u16 dataoffset8;
1053 u16 length8;
1054 u16 flags;
1055 u64 transactionid;
1056 u32 reserved;
1057 u32 rangecount; /* Always 1 in this case */
1058 struct hv_mpb_array range;
1059} __packed;
1060
1061int vmbus_alloc_ring(struct vmbus_channel *channel,
1062 u32 send_size, u32 recv_size);
1063void vmbus_free_ring(struct vmbus_channel *channel);
1064
1065int vmbus_connect_ring(struct vmbus_channel *channel,
1066 void (*onchannel_callback)(void *context),
1067 void *context);
1068int vmbus_disconnect_ring(struct vmbus_channel *channel);
1069
1070extern int vmbus_open(struct vmbus_channel *channel,
1071 u32 send_ringbuffersize,
1072 u32 recv_ringbuffersize,
1073 void *userdata,
1074 u32 userdatalen,
1075 void (*onchannel_callback)(void *context),
1076 void *context);
1077
1078extern void vmbus_close(struct vmbus_channel *channel);
1079
1080extern int vmbus_sendpacket(struct vmbus_channel *channel,
1081 void *buffer,
1082 u32 bufferLen,
1083 u64 requestid,
1084 enum vmbus_packet_type type,
1085 u32 flags);
1086
1087extern int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
1088 struct hv_page_buffer pagebuffers[],
1089 u32 pagecount,
1090 void *buffer,
1091 u32 bufferlen,
1092 u64 requestid);
1093
1094extern int vmbus_sendpacket_mpb_desc(struct vmbus_channel *channel,
1095 struct vmbus_packet_mpb_array *mpb,
1096 u32 desc_size,
1097 void *buffer,
1098 u32 bufferlen,
1099 u64 requestid);
1100
1101extern int vmbus_establish_gpadl(struct vmbus_channel *channel,
1102 void *kbuffer,
1103 u32 size,
1104 u32 *gpadl_handle);
1105
1106extern int vmbus_teardown_gpadl(struct vmbus_channel *channel,
1107 u32 gpadl_handle);
1108
1109void vmbus_reset_channel_cb(struct vmbus_channel *channel);
1110
1111extern int vmbus_recvpacket(struct vmbus_channel *channel,
1112 void *buffer,
1113 u32 bufferlen,
1114 u32 *buffer_actual_len,
1115 u64 *requestid);
1116
1117extern int vmbus_recvpacket_raw(struct vmbus_channel *channel,
1118 void *buffer,
1119 u32 bufferlen,
1120 u32 *buffer_actual_len,
1121 u64 *requestid);
1122
1123
1124extern void vmbus_ontimer(unsigned long data);
1125
1126/* Base driver object */
1127struct hv_driver {
1128 const char *name;
1129
1130 /*
1131 * A hvsock offer, which has a VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER
1132 * channel flag, actually doesn't mean a synthetic device because the
1133 * offer's if_type/if_instance can change for every new hvsock
1134 * connection.
1135 *
1136 * However, to facilitate the notification of new-offer/rescind-offer
1137 * from vmbus driver to hvsock driver, we can handle hvsock offer as
1138 * a special vmbus device, and hence we need the below flag to
1139 * indicate if the driver is the hvsock driver or not: we need to
1140 * specially treat the hvosck offer & driver in vmbus_match().
1141 */
1142 bool hvsock;
1143
1144 /* the device type supported by this driver */
1145 guid_t dev_type;
1146 const struct hv_vmbus_device_id *id_table;
1147
1148 struct device_driver driver;
1149
1150 /* dynamic device GUID's */
1151 struct {
1152 spinlock_t lock;
1153 struct list_head list;
1154 } dynids;
1155
1156 int (*probe)(struct hv_device *, const struct hv_vmbus_device_id *);
1157 int (*remove)(struct hv_device *);
1158 void (*shutdown)(struct hv_device *);
1159
1160};
1161
1162/* Base device object */
1163struct hv_device {
1164 /* the device type id of this device */
1165 guid_t dev_type;
1166
1167 /* the device instance id of this device */
1168 guid_t dev_instance;
1169 u16 vendor_id;
1170 u16 device_id;
1171
1172 struct device device;
1173 char *driver_override; /* Driver name to force a match */
1174
1175 struct vmbus_channel *channel;
1176 struct kset *channels_kset;
1177};
1178
1179
1180static inline struct hv_device *device_to_hv_device(struct device *d)
1181{
1182 return container_of(d, struct hv_device, device);
1183}
1184
1185static inline struct hv_driver *drv_to_hv_drv(struct device_driver *d)
1186{
1187 return container_of(d, struct hv_driver, driver);
1188}
1189
1190static inline void hv_set_drvdata(struct hv_device *dev, void *data)
1191{
1192 dev_set_drvdata(&dev->device, data);
1193}
1194
1195static inline void *hv_get_drvdata(struct hv_device *dev)
1196{
1197 return dev_get_drvdata(&dev->device);
1198}
1199
1200struct hv_ring_buffer_debug_info {
1201 u32 current_interrupt_mask;
1202 u32 current_read_index;
1203 u32 current_write_index;
1204 u32 bytes_avail_toread;
1205 u32 bytes_avail_towrite;
1206};
1207
1208
1209int hv_ringbuffer_get_debuginfo(const struct hv_ring_buffer_info *ring_info,
1210 struct hv_ring_buffer_debug_info *debug_info);
1211
1212/* Vmbus interface */
1213#define vmbus_driver_register(driver) \
1214 __vmbus_driver_register(driver, THIS_MODULE, KBUILD_MODNAME)
1215int __must_check __vmbus_driver_register(struct hv_driver *hv_driver,
1216 struct module *owner,
1217 const char *mod_name);
1218void vmbus_driver_unregister(struct hv_driver *hv_driver);
1219
1220void vmbus_hvsock_device_unregister(struct vmbus_channel *channel);
1221
1222int vmbus_allocate_mmio(struct resource **new, struct hv_device *device_obj,
1223 resource_size_t min, resource_size_t max,
1224 resource_size_t size, resource_size_t align,
1225 bool fb_overlap_ok);
1226void vmbus_free_mmio(resource_size_t start, resource_size_t size);
1227
1228/*
1229 * GUID definitions of various offer types - services offered to the guest.
1230 */
1231
1232/*
1233 * Network GUID
1234 * {f8615163-df3e-46c5-913f-f2d2f965ed0e}
1235 */
1236#define HV_NIC_GUID \
1237 .guid = GUID_INIT(0xf8615163, 0xdf3e, 0x46c5, 0x91, 0x3f, \
1238 0xf2, 0xd2, 0xf9, 0x65, 0xed, 0x0e)
1239
1240/*
1241 * IDE GUID
1242 * {32412632-86cb-44a2-9b5c-50d1417354f5}
1243 */
1244#define HV_IDE_GUID \
1245 .guid = GUID_INIT(0x32412632, 0x86cb, 0x44a2, 0x9b, 0x5c, \
1246 0x50, 0xd1, 0x41, 0x73, 0x54, 0xf5)
1247
1248/*
1249 * SCSI GUID
1250 * {ba6163d9-04a1-4d29-b605-72e2ffb1dc7f}
1251 */
1252#define HV_SCSI_GUID \
1253 .guid = GUID_INIT(0xba6163d9, 0x04a1, 0x4d29, 0xb6, 0x05, \
1254 0x72, 0xe2, 0xff, 0xb1, 0xdc, 0x7f)
1255
1256/*
1257 * Shutdown GUID
1258 * {0e0b6031-5213-4934-818b-38d90ced39db}
1259 */
1260#define HV_SHUTDOWN_GUID \
1261 .guid = GUID_INIT(0x0e0b6031, 0x5213, 0x4934, 0x81, 0x8b, \
1262 0x38, 0xd9, 0x0c, 0xed, 0x39, 0xdb)
1263
1264/*
1265 * Time Synch GUID
1266 * {9527E630-D0AE-497b-ADCE-E80AB0175CAF}
1267 */
1268#define HV_TS_GUID \
1269 .guid = GUID_INIT(0x9527e630, 0xd0ae, 0x497b, 0xad, 0xce, \
1270 0xe8, 0x0a, 0xb0, 0x17, 0x5c, 0xaf)
1271
1272/*
1273 * Heartbeat GUID
1274 * {57164f39-9115-4e78-ab55-382f3bd5422d}
1275 */
1276#define HV_HEART_BEAT_GUID \
1277 .guid = GUID_INIT(0x57164f39, 0x9115, 0x4e78, 0xab, 0x55, \
1278 0x38, 0x2f, 0x3b, 0xd5, 0x42, 0x2d)
1279
1280/*
1281 * KVP GUID
1282 * {a9a0f4e7-5a45-4d96-b827-8a841e8c03e6}
1283 */
1284#define HV_KVP_GUID \
1285 .guid = GUID_INIT(0xa9a0f4e7, 0x5a45, 0x4d96, 0xb8, 0x27, \
1286 0x8a, 0x84, 0x1e, 0x8c, 0x03, 0xe6)
1287
1288/*
1289 * Dynamic memory GUID
1290 * {525074dc-8985-46e2-8057-a307dc18a502}
1291 */
1292#define HV_DM_GUID \
1293 .guid = GUID_INIT(0x525074dc, 0x8985, 0x46e2, 0x80, 0x57, \
1294 0xa3, 0x07, 0xdc, 0x18, 0xa5, 0x02)
1295
1296/*
1297 * Mouse GUID
1298 * {cfa8b69e-5b4a-4cc0-b98b-8ba1a1f3f95a}
1299 */
1300#define HV_MOUSE_GUID \
1301 .guid = GUID_INIT(0xcfa8b69e, 0x5b4a, 0x4cc0, 0xb9, 0x8b, \
1302 0x8b, 0xa1, 0xa1, 0xf3, 0xf9, 0x5a)
1303
1304/*
1305 * Keyboard GUID
1306 * {f912ad6d-2b17-48ea-bd65-f927a61c7684}
1307 */
1308#define HV_KBD_GUID \
1309 .guid = GUID_INIT(0xf912ad6d, 0x2b17, 0x48ea, 0xbd, 0x65, \
1310 0xf9, 0x27, 0xa6, 0x1c, 0x76, 0x84)
1311
1312/*
1313 * VSS (Backup/Restore) GUID
1314 */
1315#define HV_VSS_GUID \
1316 .guid = GUID_INIT(0x35fa2e29, 0xea23, 0x4236, 0x96, 0xae, \
1317 0x3a, 0x6e, 0xba, 0xcb, 0xa4, 0x40)
1318/*
1319 * Synthetic Video GUID
1320 * {DA0A7802-E377-4aac-8E77-0558EB1073F8}
1321 */
1322#define HV_SYNTHVID_GUID \
1323 .guid = GUID_INIT(0xda0a7802, 0xe377, 0x4aac, 0x8e, 0x77, \
1324 0x05, 0x58, 0xeb, 0x10, 0x73, 0xf8)
1325
1326/*
1327 * Synthetic FC GUID
1328 * {2f9bcc4a-0069-4af3-b76b-6fd0be528cda}
1329 */
1330#define HV_SYNTHFC_GUID \
1331 .guid = GUID_INIT(0x2f9bcc4a, 0x0069, 0x4af3, 0xb7, 0x6b, \
1332 0x6f, 0xd0, 0xbe, 0x52, 0x8c, 0xda)
1333
1334/*
1335 * Guest File Copy Service
1336 * {34D14BE3-DEE4-41c8-9AE7-6B174977C192}
1337 */
1338
1339#define HV_FCOPY_GUID \
1340 .guid = GUID_INIT(0x34d14be3, 0xdee4, 0x41c8, 0x9a, 0xe7, \
1341 0x6b, 0x17, 0x49, 0x77, 0xc1, 0x92)
1342
1343/*
1344 * NetworkDirect. This is the guest RDMA service.
1345 * {8c2eaf3d-32a7-4b09-ab99-bd1f1c86b501}
1346 */
1347#define HV_ND_GUID \
1348 .guid = GUID_INIT(0x8c2eaf3d, 0x32a7, 0x4b09, 0xab, 0x99, \
1349 0xbd, 0x1f, 0x1c, 0x86, 0xb5, 0x01)
1350
1351/*
1352 * PCI Express Pass Through
1353 * {44C4F61D-4444-4400-9D52-802E27EDE19F}
1354 */
1355
1356#define HV_PCIE_GUID \
1357 .guid = GUID_INIT(0x44c4f61d, 0x4444, 0x4400, 0x9d, 0x52, \
1358 0x80, 0x2e, 0x27, 0xed, 0xe1, 0x9f)
1359
1360/*
1361 * Linux doesn't support the 3 devices: the first two are for
1362 * Automatic Virtual Machine Activation, and the third is for
1363 * Remote Desktop Virtualization.
1364 * {f8e65716-3cb3-4a06-9a60-1889c5cccab5}
1365 * {3375baf4-9e15-4b30-b765-67acb10d607b}
1366 * {276aacf4-ac15-426c-98dd-7521ad3f01fe}
1367 */
1368
1369#define HV_AVMA1_GUID \
1370 .guid = GUID_INIT(0xf8e65716, 0x3cb3, 0x4a06, 0x9a, 0x60, \
1371 0x18, 0x89, 0xc5, 0xcc, 0xca, 0xb5)
1372
1373#define HV_AVMA2_GUID \
1374 .guid = GUID_INIT(0x3375baf4, 0x9e15, 0x4b30, 0xb7, 0x65, \
1375 0x67, 0xac, 0xb1, 0x0d, 0x60, 0x7b)
1376
1377#define HV_RDV_GUID \
1378 .guid = GUID_INIT(0x276aacf4, 0xac15, 0x426c, 0x98, 0xdd, \
1379 0x75, 0x21, 0xad, 0x3f, 0x01, 0xfe)
1380
1381/*
1382 * Common header for Hyper-V ICs
1383 */
1384
1385#define ICMSGTYPE_NEGOTIATE 0
1386#define ICMSGTYPE_HEARTBEAT 1
1387#define ICMSGTYPE_KVPEXCHANGE 2
1388#define ICMSGTYPE_SHUTDOWN 3
1389#define ICMSGTYPE_TIMESYNC 4
1390#define ICMSGTYPE_VSS 5
1391
1392#define ICMSGHDRFLAG_TRANSACTION 1
1393#define ICMSGHDRFLAG_REQUEST 2
1394#define ICMSGHDRFLAG_RESPONSE 4
1395
1396
1397/*
1398 * While we want to handle util services as regular devices,
1399 * there is only one instance of each of these services; so
1400 * we statically allocate the service specific state.
1401 */
1402
1403struct hv_util_service {
1404 u8 *recv_buffer;
1405 void *channel;
1406 void (*util_cb)(void *);
1407 int (*util_init)(struct hv_util_service *);
1408 void (*util_deinit)(void);
1409};
1410
1411struct vmbuspipe_hdr {
1412 u32 flags;
1413 u32 msgsize;
1414} __packed;
1415
1416struct ic_version {
1417 u16 major;
1418 u16 minor;
1419} __packed;
1420
1421struct icmsg_hdr {
1422 struct ic_version icverframe;
1423 u16 icmsgtype;
1424 struct ic_version icvermsg;
1425 u16 icmsgsize;
1426 u32 status;
1427 u8 ictransaction_id;
1428 u8 icflags;
1429 u8 reserved[2];
1430} __packed;
1431
1432struct icmsg_negotiate {
1433 u16 icframe_vercnt;
1434 u16 icmsg_vercnt;
1435 u32 reserved;
1436 struct ic_version icversion_data[1]; /* any size array */
1437} __packed;
1438
1439struct shutdown_msg_data {
1440 u32 reason_code;
1441 u32 timeout_seconds;
1442 u32 flags;
1443 u8 display_message[2048];
1444} __packed;
1445
1446struct heartbeat_msg_data {
1447 u64 seq_num;
1448 u32 reserved[8];
1449} __packed;
1450
1451/* Time Sync IC defs */
1452#define ICTIMESYNCFLAG_PROBE 0
1453#define ICTIMESYNCFLAG_SYNC 1
1454#define ICTIMESYNCFLAG_SAMPLE 2
1455
1456#ifdef __x86_64__
1457#define WLTIMEDELTA 116444736000000000L /* in 100ns unit */
1458#else
1459#define WLTIMEDELTA 116444736000000000LL
1460#endif
1461
1462struct ictimesync_data {
1463 u64 parenttime;
1464 u64 childtime;
1465 u64 roundtriptime;
1466 u8 flags;
1467} __packed;
1468
1469struct ictimesync_ref_data {
1470 u64 parenttime;
1471 u64 vmreferencetime;
1472 u8 flags;
1473 char leapflags;
1474 char stratum;
1475 u8 reserved[3];
1476} __packed;
1477
1478struct hyperv_service_callback {
1479 u8 msg_type;
1480 char *log_msg;
1481 guid_t data;
1482 struct vmbus_channel *channel;
1483 void (*callback)(void *context);
1484};
1485
1486#define MAX_SRV_VER 0x7ffffff
1487extern bool vmbus_prep_negotiate_resp(struct icmsg_hdr *icmsghdrp, u8 *buf,
1488 const int *fw_version, int fw_vercnt,
1489 const int *srv_version, int srv_vercnt,
1490 int *nego_fw_version, int *nego_srv_version);
1491
1492void hv_process_channel_removal(struct vmbus_channel *channel);
1493
1494void vmbus_setevent(struct vmbus_channel *channel);
1495/*
1496 * Negotiated version with the Host.
1497 */
1498
1499extern __u32 vmbus_proto_version;
1500
1501int vmbus_send_tl_connect_request(const guid_t *shv_guest_servie_id,
1502 const guid_t *shv_host_servie_id);
1503void vmbus_set_event(struct vmbus_channel *channel);
1504
1505/* Get the start of the ring buffer. */
1506static inline void *
1507hv_get_ring_buffer(const struct hv_ring_buffer_info *ring_info)
1508{
1509 return ring_info->ring_buffer->buffer;
1510}
1511
1512/*
1513 * Mask off host interrupt callback notifications
1514 */
1515static inline void hv_begin_read(struct hv_ring_buffer_info *rbi)
1516{
1517 rbi->ring_buffer->interrupt_mask = 1;
1518
1519 /* make sure mask update is not reordered */
1520 virt_mb();
1521}
1522
1523/*
1524 * Re-enable host callback and return number of outstanding bytes
1525 */
1526static inline u32 hv_end_read(struct hv_ring_buffer_info *rbi)
1527{
1528
1529 rbi->ring_buffer->interrupt_mask = 0;
1530
1531 /* make sure mask update is not reordered */
1532 virt_mb();
1533
1534 /*
1535 * Now check to see if the ring buffer is still empty.
1536 * If it is not, we raced and we need to process new
1537 * incoming messages.
1538 */
1539 return hv_get_bytes_to_read(rbi);
1540}
1541
1542/*
1543 * An API to support in-place processing of incoming VMBUS packets.
1544 */
1545
1546/* Get data payload associated with descriptor */
1547static inline void *hv_pkt_data(const struct vmpacket_descriptor *desc)
1548{
1549 return (void *)((unsigned long)desc + (desc->offset8 << 3));
1550}
1551
1552/* Get data size associated with descriptor */
1553static inline u32 hv_pkt_datalen(const struct vmpacket_descriptor *desc)
1554{
1555 return (desc->len8 << 3) - (desc->offset8 << 3);
1556}
1557
1558
1559struct vmpacket_descriptor *
1560hv_pkt_iter_first(struct vmbus_channel *channel);
1561
1562struct vmpacket_descriptor *
1563__hv_pkt_iter_next(struct vmbus_channel *channel,
1564 const struct vmpacket_descriptor *pkt);
1565
1566void hv_pkt_iter_close(struct vmbus_channel *channel);
1567
1568/*
1569 * Get next packet descriptor from iterator
1570 * If at end of list, return NULL and update host.
1571 */
1572static inline struct vmpacket_descriptor *
1573hv_pkt_iter_next(struct vmbus_channel *channel,
1574 const struct vmpacket_descriptor *pkt)
1575{
1576 struct vmpacket_descriptor *nxt;
1577
1578 nxt = __hv_pkt_iter_next(channel, pkt);
1579 if (!nxt)
1580 hv_pkt_iter_close(channel);
1581
1582 return nxt;
1583}
1584
1585#define foreach_vmbus_pkt(pkt, channel) \
1586 for (pkt = hv_pkt_iter_first(channel); pkt; \
1587 pkt = hv_pkt_iter_next(channel, pkt))
1588
1589#endif /* _HYPERV_H */
1590