~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/linux/hyperv.h

Version: ~ [ linux-5.2 ] ~ [ linux-5.1.16 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.57 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.132 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.184 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.184 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.69 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.39.4 ] ~ [ linux-2.6.38.8 ] ~ [ linux-2.6.37.6 ] ~ [ linux-2.6.36.4 ] ~ [ linux-2.6.35.14 ] ~ [ linux-2.6.34.15 ] ~ [ linux-2.6.33.20 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 #include <uapi/asm/hyperv.h>
 30 
 31 #include <linux/types.h>
 32 #include <linux/scatterlist.h>
 33 #include <linux/list.h>
 34 #include <linux/timer.h>
 35 #include <linux/workqueue.h>
 36 #include <linux/completion.h>
 37 #include <linux/device.h>
 38 #include <linux/mod_devicetable.h>
 39 
 40 
 41 #define MAX_PAGE_BUFFER_COUNT                           32
 42 #define MAX_MULTIPAGE_BUFFER_COUNT                      32 /* 128K */
 43 
 44 #pragma pack(push, 1)
 45 
 46 /* Single-page buffer */
 47 struct hv_page_buffer {
 48         u32 len;
 49         u32 offset;
 50         u64 pfn;
 51 };
 52 
 53 /* Multiple-page buffer */
 54 struct hv_multipage_buffer {
 55         /* Length and Offset determines the # of pfns in the array */
 56         u32 len;
 57         u32 offset;
 58         u64 pfn_array[MAX_MULTIPAGE_BUFFER_COUNT];
 59 };
 60 
 61 /*
 62  * Multiple-page buffer array; the pfn array is variable size:
 63  * The number of entries in the PFN array is determined by
 64  * "len" and "offset".
 65  */
 66 struct hv_mpb_array {
 67         /* Length and Offset determines the # of pfns in the array */
 68         u32 len;
 69         u32 offset;
 70         u64 pfn_array[];
 71 };
 72 
 73 /* 0x18 includes the proprietary packet header */
 74 #define MAX_PAGE_BUFFER_PACKET          (0x18 +                 \
 75                                         (sizeof(struct hv_page_buffer) * \
 76                                          MAX_PAGE_BUFFER_COUNT))
 77 #define MAX_MULTIPAGE_BUFFER_PACKET     (0x18 +                 \
 78                                          sizeof(struct hv_multipage_buffer))
 79 
 80 
 81 #pragma pack(pop)
 82 
 83 struct hv_ring_buffer {
 84         /* Offset in bytes from the start of ring data below */
 85         u32 write_index;
 86 
 87         /* Offset in bytes from the start of ring data below */
 88         u32 read_index;
 89 
 90         u32 interrupt_mask;
 91 
 92         /*
 93          * Win8 uses some of the reserved bits to implement
 94          * interrupt driven flow management. On the send side
 95          * we can request that the receiver interrupt the sender
 96          * when the ring transitions from being full to being able
 97          * to handle a message of size "pending_send_sz".
 98          *
 99          * Add necessary state for this enhancement.
100          */
101         u32 pending_send_sz;
102 
103         u32 reserved1[12];
104 
105         union {
106                 struct {
107                         u32 feat_pending_send_sz:1;
108                 };
109                 u32 value;
110         } feature_bits;
111 
112         /* Pad it to PAGE_SIZE so that data starts on page boundary */
113         u8      reserved2[4028];
114 
115         /*
116          * Ring data starts here + RingDataStartOffset
117          * !!! DO NOT place any fields below this !!!
118          */
119         u8 buffer[0];
120 } __packed;
121 
122 struct hv_ring_buffer_info {
123         struct hv_ring_buffer *ring_buffer;
124         u32 ring_size;                  /* Include the shared header */
125         spinlock_t ring_lock;
126 
127         u32 ring_datasize;              /* < ring_size */
128         u32 ring_data_startoffset;
129         u32 priv_write_index;
130         u32 priv_read_index;
131         u32 cached_read_index;
132 };
133 
134 /*
135  *
136  * hv_get_ringbuffer_availbytes()
137  *
138  * Get number of bytes available to read and to write to
139  * for the specified ring buffer
140  */
141 static inline void
142 hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
143                           u32 *read, u32 *write)
144 {
145         u32 read_loc, write_loc, dsize;
146 
147         /* Capture the read/write indices before they changed */
148         read_loc = rbi->ring_buffer->read_index;
149         write_loc = rbi->ring_buffer->write_index;
150         dsize = rbi->ring_datasize;
151 
152         *write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
153                 read_loc - write_loc;
154         *read = dsize - *write;
155 }
156 
157 static inline u32 hv_get_bytes_to_read(struct hv_ring_buffer_info *rbi)
158 {
159         u32 read_loc, write_loc, dsize, read;
160 
161         dsize = rbi->ring_datasize;
162         read_loc = rbi->ring_buffer->read_index;
163         write_loc = READ_ONCE(rbi->ring_buffer->write_index);
164 
165         read = write_loc >= read_loc ? (write_loc - read_loc) :
166                 (dsize - read_loc) + write_loc;
167 
168         return read;
169 }
170 
171 static inline u32 hv_get_bytes_to_write(struct hv_ring_buffer_info *rbi)
172 {
173         u32 read_loc, write_loc, dsize, write;
174 
175         dsize = rbi->ring_datasize;
176         read_loc = READ_ONCE(rbi->ring_buffer->read_index);
177         write_loc = rbi->ring_buffer->write_index;
178 
179         write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
180                 read_loc - write_loc;
181         return write;
182 }
183 
184 static inline u32 hv_get_cached_bytes_to_write(
185         const struct hv_ring_buffer_info *rbi)
186 {
187         u32 read_loc, write_loc, dsize, write;
188 
189         dsize = rbi->ring_datasize;
190         read_loc = rbi->cached_read_index;
191         write_loc = rbi->ring_buffer->write_index;
192 
193         write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
194                 read_loc - write_loc;
195         return write;
196 }
197 /*
198  * VMBUS version is 32 bit entity broken up into
199  * two 16 bit quantities: major_number. minor_number.
200  *
201  * 0 . 13 (Windows Server 2008)
202  * 1 . 1  (Windows 7)
203  * 2 . 4  (Windows 8)
204  * 3 . 0  (Windows 8 R2)
205  * 4 . 0  (Windows 10)
206  */
207 
208 #define VERSION_WS2008  ((0 << 16) | (13))
209 #define VERSION_WIN7    ((1 << 16) | (1))
210 #define VERSION_WIN8    ((2 << 16) | (4))
211 #define VERSION_WIN8_1    ((3 << 16) | (0))
212 #define VERSION_WIN10   ((4 << 16) | (0))
213 
214 #define VERSION_INVAL -1
215 
216 #define VERSION_CURRENT VERSION_WIN10
217 
218 /* Make maximum size of pipe payload of 16K */
219 #define MAX_PIPE_DATA_PAYLOAD           (sizeof(u8) * 16384)
220 
221 /* Define PipeMode values. */
222 #define VMBUS_PIPE_TYPE_BYTE            0x00000000
223 #define VMBUS_PIPE_TYPE_MESSAGE         0x00000004
224 
225 /* The size of the user defined data buffer for non-pipe offers. */
226 #define MAX_USER_DEFINED_BYTES          120
227 
228 /* The size of the user defined data buffer for pipe offers. */
229 #define MAX_PIPE_USER_DEFINED_BYTES     116
230 
231 /*
232  * At the center of the Channel Management library is the Channel Offer. This
233  * struct contains the fundamental information about an offer.
234  */
235 struct vmbus_channel_offer {
236         uuid_le if_type;
237         uuid_le if_instance;
238 
239         /*
240          * These two fields are not currently used.
241          */
242         u64 reserved1;
243         u64 reserved2;
244 
245         u16 chn_flags;
246         u16 mmio_megabytes;             /* in bytes * 1024 * 1024 */
247 
248         union {
249                 /* Non-pipes: The user has MAX_USER_DEFINED_BYTES bytes. */
250                 struct {
251                         unsigned char user_def[MAX_USER_DEFINED_BYTES];
252                 } std;
253 
254                 /*
255                  * Pipes:
256                  * The following sructure is an integrated pipe protocol, which
257                  * is implemented on top of standard user-defined data. Pipe
258                  * clients have MAX_PIPE_USER_DEFINED_BYTES left for their own
259                  * use.
260                  */
261                 struct {
262                         u32  pipe_mode;
263                         unsigned char user_def[MAX_PIPE_USER_DEFINED_BYTES];
264                 } pipe;
265         } u;
266         /*
267          * The sub_channel_index is defined in win8.
268          */
269         u16 sub_channel_index;
270         u16 reserved3;
271 } __packed;
272 
273 /* Server Flags */
274 #define VMBUS_CHANNEL_ENUMERATE_DEVICE_INTERFACE        1
275 #define VMBUS_CHANNEL_SERVER_SUPPORTS_TRANSFER_PAGES    2
276 #define VMBUS_CHANNEL_SERVER_SUPPORTS_GPADLS            4
277 #define VMBUS_CHANNEL_NAMED_PIPE_MODE                   0x10
278 #define VMBUS_CHANNEL_LOOPBACK_OFFER                    0x100
279 #define VMBUS_CHANNEL_PARENT_OFFER                      0x200
280 #define VMBUS_CHANNEL_REQUEST_MONITORED_NOTIFICATION    0x400
281 #define VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER              0x2000
282 
283 struct vmpacket_descriptor {
284         u16 type;
285         u16 offset8;
286         u16 len8;
287         u16 flags;
288         u64 trans_id;
289 } __packed;
290 
291 struct vmpacket_header {
292         u32 prev_pkt_start_offset;
293         struct vmpacket_descriptor descriptor;
294 } __packed;
295 
296 struct vmtransfer_page_range {
297         u32 byte_count;
298         u32 byte_offset;
299 } __packed;
300 
301 struct vmtransfer_page_packet_header {
302         struct vmpacket_descriptor d;
303         u16 xfer_pageset_id;
304         u8  sender_owns_set;
305         u8 reserved;
306         u32 range_cnt;
307         struct vmtransfer_page_range ranges[1];
308 } __packed;
309 
310 struct vmgpadl_packet_header {
311         struct vmpacket_descriptor d;
312         u32 gpadl;
313         u32 reserved;
314 } __packed;
315 
316 struct vmadd_remove_transfer_page_set {
317         struct vmpacket_descriptor d;
318         u32 gpadl;
319         u16 xfer_pageset_id;
320         u16 reserved;
321 } __packed;
322 
323 /*
324  * This structure defines a range in guest physical space that can be made to
325  * look virtually contiguous.
326  */
327 struct gpa_range {
328         u32 byte_count;
329         u32 byte_offset;
330         u64 pfn_array[0];
331 };
332 
333 /*
334  * This is the format for an Establish Gpadl packet, which contains a handle by
335  * which this GPADL will be known and a set of GPA ranges associated with it.
336  * This can be converted to a MDL by the guest OS.  If there are multiple GPA
337  * ranges, then the resulting MDL will be "chained," representing multiple VA
338  * ranges.
339  */
340 struct vmestablish_gpadl {
341         struct vmpacket_descriptor d;
342         u32 gpadl;
343         u32 range_cnt;
344         struct gpa_range range[1];
345 } __packed;
346 
347 /*
348  * This is the format for a Teardown Gpadl packet, which indicates that the
349  * GPADL handle in the Establish Gpadl packet will never be referenced again.
350  */
351 struct vmteardown_gpadl {
352         struct vmpacket_descriptor d;
353         u32 gpadl;
354         u32 reserved;   /* for alignment to a 8-byte boundary */
355 } __packed;
356 
357 /*
358  * This is the format for a GPA-Direct packet, which contains a set of GPA
359  * ranges, in addition to commands and/or data.
360  */
361 struct vmdata_gpa_direct {
362         struct vmpacket_descriptor d;
363         u32 reserved;
364         u32 range_cnt;
365         struct gpa_range range[1];
366 } __packed;
367 
368 /* This is the format for a Additional Data Packet. */
369 struct vmadditional_data {
370         struct vmpacket_descriptor d;
371         u64 total_bytes;
372         u32 offset;
373         u32 byte_cnt;
374         unsigned char data[1];
375 } __packed;
376 
377 union vmpacket_largest_possible_header {
378         struct vmpacket_descriptor simple_hdr;
379         struct vmtransfer_page_packet_header xfer_page_hdr;
380         struct vmgpadl_packet_header gpadl_hdr;
381         struct vmadd_remove_transfer_page_set add_rm_xfer_page_hdr;
382         struct vmestablish_gpadl establish_gpadl_hdr;
383         struct vmteardown_gpadl teardown_gpadl_hdr;
384         struct vmdata_gpa_direct data_gpa_direct_hdr;
385 };
386 
387 #define VMPACKET_DATA_START_ADDRESS(__packet)   \
388         (void *)(((unsigned char *)__packet) +  \
389          ((struct vmpacket_descriptor)__packet)->offset8 * 8)
390 
391 #define VMPACKET_DATA_LENGTH(__packet)          \
392         ((((struct vmpacket_descriptor)__packet)->len8 -        \
393           ((struct vmpacket_descriptor)__packet)->offset8) * 8)
394 
395 #define VMPACKET_TRANSFER_MODE(__packet)        \
396         (((struct IMPACT)__packet)->type)
397 
398 enum vmbus_packet_type {
399         VM_PKT_INVALID                          = 0x0,
400         VM_PKT_SYNCH                            = 0x1,
401         VM_PKT_ADD_XFER_PAGESET                 = 0x2,
402         VM_PKT_RM_XFER_PAGESET                  = 0x3,
403         VM_PKT_ESTABLISH_GPADL                  = 0x4,
404         VM_PKT_TEARDOWN_GPADL                   = 0x5,
405         VM_PKT_DATA_INBAND                      = 0x6,
406         VM_PKT_DATA_USING_XFER_PAGES            = 0x7,
407         VM_PKT_DATA_USING_GPADL                 = 0x8,
408         VM_PKT_DATA_USING_GPA_DIRECT            = 0x9,
409         VM_PKT_CANCEL_REQUEST                   = 0xa,
410         VM_PKT_COMP                             = 0xb,
411         VM_PKT_DATA_USING_ADDITIONAL_PKT        = 0xc,
412         VM_PKT_ADDITIONAL_DATA                  = 0xd
413 };
414 
415 #define VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED     1
416 
417 
418 /* Version 1 messages */
419 enum vmbus_channel_message_type {
420         CHANNELMSG_INVALID                      =  0,
421         CHANNELMSG_OFFERCHANNEL         =  1,
422         CHANNELMSG_RESCIND_CHANNELOFFER =  2,
423         CHANNELMSG_REQUESTOFFERS                =  3,
424         CHANNELMSG_ALLOFFERS_DELIVERED  =  4,
425         CHANNELMSG_OPENCHANNEL          =  5,
426         CHANNELMSG_OPENCHANNEL_RESULT           =  6,
427         CHANNELMSG_CLOSECHANNEL         =  7,
428         CHANNELMSG_GPADL_HEADER         =  8,
429         CHANNELMSG_GPADL_BODY                   =  9,
430         CHANNELMSG_GPADL_CREATED                = 10,
431         CHANNELMSG_GPADL_TEARDOWN               = 11,
432         CHANNELMSG_GPADL_TORNDOWN               = 12,
433         CHANNELMSG_RELID_RELEASED               = 13,
434         CHANNELMSG_INITIATE_CONTACT             = 14,
435         CHANNELMSG_VERSION_RESPONSE             = 15,
436         CHANNELMSG_UNLOAD                       = 16,
437         CHANNELMSG_UNLOAD_RESPONSE              = 17,
438         CHANNELMSG_18                           = 18,
439         CHANNELMSG_19                           = 19,
440         CHANNELMSG_20                           = 20,
441         CHANNELMSG_TL_CONNECT_REQUEST           = 21,
442         CHANNELMSG_COUNT
443 };
444 
445 struct vmbus_channel_message_header {
446         enum vmbus_channel_message_type msgtype;
447         u32 padding;
448 } __packed;
449 
450 /* Query VMBus Version parameters */
451 struct vmbus_channel_query_vmbus_version {
452         struct vmbus_channel_message_header header;
453         u32 version;
454 } __packed;
455 
456 /* VMBus Version Supported parameters */
457 struct vmbus_channel_version_supported {
458         struct vmbus_channel_message_header header;
459         u8 version_supported;
460 } __packed;
461 
462 /* Offer Channel parameters */
463 struct vmbus_channel_offer_channel {
464         struct vmbus_channel_message_header header;
465         struct vmbus_channel_offer offer;
466         u32 child_relid;
467         u8 monitorid;
468         /*
469          * win7 and beyond splits this field into a bit field.
470          */
471         u8 monitor_allocated:1;
472         u8 reserved:7;
473         /*
474          * These are new fields added in win7 and later.
475          * Do not access these fields without checking the
476          * negotiated protocol.
477          *
478          * If "is_dedicated_interrupt" is set, we must not set the
479          * associated bit in the channel bitmap while sending the
480          * interrupt to the host.
481          *
482          * connection_id is to be used in signaling the host.
483          */
484         u16 is_dedicated_interrupt:1;
485         u16 reserved1:15;
486         u32 connection_id;
487 } __packed;
488 
489 /* Rescind Offer parameters */
490 struct vmbus_channel_rescind_offer {
491         struct vmbus_channel_message_header header;
492         u32 child_relid;
493 } __packed;
494 
495 /*
496  * Request Offer -- no parameters, SynIC message contains the partition ID
497  * Set Snoop -- no parameters, SynIC message contains the partition ID
498  * Clear Snoop -- no parameters, SynIC message contains the partition ID
499  * All Offers Delivered -- no parameters, SynIC message contains the partition
500  *                         ID
501  * Flush Client -- no parameters, SynIC message contains the partition ID
502  */
503 
504 /* Open Channel parameters */
505 struct vmbus_channel_open_channel {
506         struct vmbus_channel_message_header header;
507 
508         /* Identifies the specific VMBus channel that is being opened. */
509         u32 child_relid;
510 
511         /* ID making a particular open request at a channel offer unique. */
512         u32 openid;
513 
514         /* GPADL for the channel's ring buffer. */
515         u32 ringbuffer_gpadlhandle;
516 
517         /*
518          * Starting with win8, this field will be used to specify
519          * the target virtual processor on which to deliver the interrupt for
520          * the host to guest communication.
521          * Prior to win8, incoming channel interrupts would only
522          * be delivered on cpu 0. Setting this value to 0 would
523          * preserve the earlier behavior.
524          */
525         u32 target_vp;
526 
527         /*
528         * The upstream ring buffer begins at offset zero in the memory
529         * described by RingBufferGpadlHandle. The downstream ring buffer
530         * follows it at this offset (in pages).
531         */
532         u32 downstream_ringbuffer_pageoffset;
533 
534         /* User-specific data to be passed along to the server endpoint. */
535         unsigned char userdata[MAX_USER_DEFINED_BYTES];
536 } __packed;
537 
538 /* Open Channel Result parameters */
539 struct vmbus_channel_open_result {
540         struct vmbus_channel_message_header header;
541         u32 child_relid;
542         u32 openid;
543         u32 status;
544 } __packed;
545 
546 /* Close channel parameters; */
547 struct vmbus_channel_close_channel {
548         struct vmbus_channel_message_header header;
549         u32 child_relid;
550 } __packed;
551 
552 /* Channel Message GPADL */
553 #define GPADL_TYPE_RING_BUFFER          1
554 #define GPADL_TYPE_SERVER_SAVE_AREA     2
555 #define GPADL_TYPE_TRANSACTION          8
556 
557 /*
558  * The number of PFNs in a GPADL message is defined by the number of
559  * pages that would be spanned by ByteCount and ByteOffset.  If the
560  * implied number of PFNs won't fit in this packet, there will be a
561  * follow-up packet that contains more.
562  */
563 struct vmbus_channel_gpadl_header {
564         struct vmbus_channel_message_header header;
565         u32 child_relid;
566         u32 gpadl;
567         u16 range_buflen;
568         u16 rangecount;
569         struct gpa_range range[0];
570 } __packed;
571 
572 /* This is the followup packet that contains more PFNs. */
573 struct vmbus_channel_gpadl_body {
574         struct vmbus_channel_message_header header;
575         u32 msgnumber;
576         u32 gpadl;
577         u64 pfn[0];
578 } __packed;
579 
580 struct vmbus_channel_gpadl_created {
581         struct vmbus_channel_message_header header;
582         u32 child_relid;
583         u32 gpadl;
584         u32 creation_status;
585 } __packed;
586 
587 struct vmbus_channel_gpadl_teardown {
588         struct vmbus_channel_message_header header;
589         u32 child_relid;
590         u32 gpadl;
591 } __packed;
592 
593 struct vmbus_channel_gpadl_torndown {
594         struct vmbus_channel_message_header header;
595         u32 gpadl;
596 } __packed;
597 
598 struct vmbus_channel_relid_released {
599         struct vmbus_channel_message_header header;
600         u32 child_relid;
601 } __packed;
602 
603 struct vmbus_channel_initiate_contact {
604         struct vmbus_channel_message_header header;
605         u32 vmbus_version_requested;
606         u32 target_vcpu; /* The VCPU the host should respond to */
607         u64 interrupt_page;
608         u64 monitor_page1;
609         u64 monitor_page2;
610 } __packed;
611 
612 /* Hyper-V socket: guest's connect()-ing to host */
613 struct vmbus_channel_tl_connect_request {
614         struct vmbus_channel_message_header header;
615         uuid_le guest_endpoint_id;
616         uuid_le host_service_id;
617 } __packed;
618 
619 struct vmbus_channel_version_response {
620         struct vmbus_channel_message_header header;
621         u8 version_supported;
622 } __packed;
623 
624 enum vmbus_channel_state {
625         CHANNEL_OFFER_STATE,
626         CHANNEL_OPENING_STATE,
627         CHANNEL_OPEN_STATE,
628         CHANNEL_OPENED_STATE,
629 };
630 
631 /*
632  * Represents each channel msg on the vmbus connection This is a
633  * variable-size data structure depending on the msg type itself
634  */
635 struct vmbus_channel_msginfo {
636         /* Bookkeeping stuff */
637         struct list_head msglistentry;
638 
639         /* So far, this is only used to handle gpadl body message */
640         struct list_head submsglist;
641 
642         /* Synchronize the request/response if needed */
643         struct completion  waitevent;
644         struct vmbus_channel *waiting_channel;
645         union {
646                 struct vmbus_channel_version_supported version_supported;
647                 struct vmbus_channel_open_result open_result;
648                 struct vmbus_channel_gpadl_torndown gpadl_torndown;
649                 struct vmbus_channel_gpadl_created gpadl_created;
650                 struct vmbus_channel_version_response version_response;
651         } response;
652 
653         u32 msgsize;
654         /*
655          * The channel message that goes out on the "wire".
656          * It will contain at minimum the VMBUS_CHANNEL_MESSAGE_HEADER header
657          */
658         unsigned char msg[0];
659 };
660 
661 struct vmbus_close_msg {
662         struct vmbus_channel_msginfo info;
663         struct vmbus_channel_close_channel msg;
664 };
665 
666 /* Define connection identifier type. */
667 union hv_connection_id {
668         u32 asu32;
669         struct {
670                 u32 id:24;
671                 u32 reserved:8;
672         } u;
673 };
674 
675 /* Definition of the hv_signal_event hypercall input structure. */
676 struct hv_input_signal_event {
677         union hv_connection_id connectionid;
678         u16 flag_number;
679         u16 rsvdz;
680 };
681 
682 struct hv_input_signal_event_buffer {
683         u64 align8;
684         struct hv_input_signal_event event;
685 };
686 
687 enum hv_signal_policy {
688         HV_SIGNAL_POLICY_DEFAULT = 0,
689         HV_SIGNAL_POLICY_EXPLICIT,
690 };
691 
692 enum hv_numa_policy {
693         HV_BALANCED = 0,
694         HV_LOCALIZED,
695 };
696 
697 enum vmbus_device_type {
698         HV_IDE = 0,
699         HV_SCSI,
700         HV_FC,
701         HV_NIC,
702         HV_ND,
703         HV_PCIE,
704         HV_FB,
705         HV_KBD,
706         HV_MOUSE,
707         HV_KVP,
708         HV_TS,
709         HV_HB,
710         HV_SHUTDOWN,
711         HV_FCOPY,
712         HV_BACKUP,
713         HV_DM,
714         HV_UNKNOWN,
715 };
716 
717 struct vmbus_device {
718         u16  dev_type;
719         uuid_le guid;
720         bool perf_device;
721 };
722 
723 struct vmbus_channel {
724         struct list_head listentry;
725 
726         struct hv_device *device_obj;
727 
728         enum vmbus_channel_state state;
729 
730         struct vmbus_channel_offer_channel offermsg;
731         /*
732          * These are based on the OfferMsg.MonitorId.
733          * Save it here for easy access.
734          */
735         u8 monitor_grp;
736         u8 monitor_bit;
737 
738         bool rescind; /* got rescind msg */
739 
740         u32 ringbuffer_gpadlhandle;
741 
742         /* Allocated memory for ring buffer */
743         void *ringbuffer_pages;
744         u32 ringbuffer_pagecount;
745         struct hv_ring_buffer_info outbound;    /* send to parent */
746         struct hv_ring_buffer_info inbound;     /* receive from parent */
747         spinlock_t inbound_lock;
748 
749         struct vmbus_close_msg close_msg;
750 
751         /* Channel callback are invoked in this workqueue context */
752         /* HANDLE dataWorkQueue; */
753 
754         void (*onchannel_callback)(void *context);
755         void *channel_callback_context;
756 
757         /*
758          * A channel can be marked for efficient (batched)
759          * reading:
760          * If batched_reading is set to "true", we read until the
761          * channel is empty and hold off interrupts from the host
762          * during the entire read process.
763          * If batched_reading is set to "false", the client is not
764          * going to perform batched reading.
765          *
766          * By default we will enable batched reading; specific
767          * drivers that don't want this behavior can turn it off.
768          */
769 
770         bool batched_reading;
771 
772         bool is_dedicated_interrupt;
773         struct hv_input_signal_event_buffer sig_buf;
774         struct hv_input_signal_event *sig_event;
775 
776         /*
777          * Starting with win8, this field will be used to specify
778          * the target virtual processor on which to deliver the interrupt for
779          * the host to guest communication.
780          * Prior to win8, incoming channel interrupts would only
781          * be delivered on cpu 0. Setting this value to 0 would
782          * preserve the earlier behavior.
783          */
784         u32 target_vp;
785         /* The corresponding CPUID in the guest */
786         u32 target_cpu;
787         /*
788          * State to manage the CPU affiliation of channels.
789          */
790         struct cpumask alloced_cpus_in_node;
791         int numa_node;
792         /*
793          * Support for sub-channels. For high performance devices,
794          * it will be useful to have multiple sub-channels to support
795          * a scalable communication infrastructure with the host.
796          * The support for sub-channels is implemented as an extention
797          * to the current infrastructure.
798          * The initial offer is considered the primary channel and this
799          * offer message will indicate if the host supports sub-channels.
800          * The guest is free to ask for sub-channels to be offerred and can
801          * open these sub-channels as a normal "primary" channel. However,
802          * all sub-channels will have the same type and instance guids as the
803          * primary channel. Requests sent on a given channel will result in a
804          * response on the same channel.
805          */
806 
807         /*
808          * Sub-channel creation callback. This callback will be called in
809          * process context when a sub-channel offer is received from the host.
810          * The guest can open the sub-channel in the context of this callback.
811          */
812         void (*sc_creation_callback)(struct vmbus_channel *new_sc);
813 
814         /*
815          * Channel rescind callback. Some channels (the hvsock ones), need to
816          * register a callback which is invoked in vmbus_onoffer_rescind().
817          */
818         void (*chn_rescind_callback)(struct vmbus_channel *channel);
819 
820         /*
821          * The spinlock to protect the structure. It is being used to protect
822          * test-and-set access to various attributes of the structure as well
823          * as all sc_list operations.
824          */
825         spinlock_t lock;
826         /*
827          * All Sub-channels of a primary channel are linked here.
828          */
829         struct list_head sc_list;
830         /*
831          * Current number of sub-channels.
832          */
833         int num_sc;
834         /*
835          * Number of a sub-channel (position within sc_list) which is supposed
836          * to be used as the next outgoing channel.
837          */
838         int next_oc;
839         /*
840          * The primary channel this sub-channel belongs to.
841          * This will be NULL for the primary channel.
842          */
843         struct vmbus_channel *primary_channel;
844         /*
845          * Support per-channel state for use by vmbus drivers.
846          */
847         void *per_channel_state;
848         /*
849          * To support per-cpu lookup mapping of relid to channel,
850          * link up channels based on their CPU affinity.
851          */
852         struct list_head percpu_list;
853         /*
854          * Host signaling policy: The default policy will be
855          * based on the ring buffer state. We will also support
856          * a policy where the client driver can have explicit
857          * signaling control.
858          */
859         enum hv_signal_policy  signal_policy;
860         /*
861          * On the channel send side, many of the VMBUS
862          * device drivers explicity serialize access to the
863          * outgoing ring buffer. Give more control to the
864          * VMBUS device drivers in terms how to serialize
865          * accesss to the outgoing ring buffer.
866          * The default behavior will be to aquire the
867          * ring lock to preserve the current behavior.
868          */
869         bool acquire_ring_lock;
870         /*
871          * For performance critical channels (storage, networking
872          * etc,), Hyper-V has a mechanism to enhance the throughput
873          * at the expense of latency:
874          * When the host is to be signaled, we just set a bit in a shared page
875          * and this bit will be inspected by the hypervisor within a certain
876          * window and if the bit is set, the host will be signaled. The window
877          * of time is the monitor latency - currently around 100 usecs. This
878          * mechanism improves throughput by:
879          *
880          * A) Making the host more efficient - each time it wakes up,
881          *    potentially it will process morev number of packets. The
882          *    monitor latency allows a batch to build up.
883          * B) By deferring the hypercall to signal, we will also minimize
884          *    the interrupts.
885          *
886          * Clearly, these optimizations improve throughput at the expense of
887          * latency. Furthermore, since the channel is shared for both
888          * control and data messages, control messages currently suffer
889          * unnecessary latency adversley impacting performance and boot
890          * time. To fix this issue, permit tagging the channel as being
891          * in "low latency" mode. In this mode, we will bypass the monitor
892          * mechanism.
893          */
894         bool low_latency;
895 
896         /*
897          * NUMA distribution policy:
898          * We support teo policies:
899          * 1) Balanced: Here all performance critical channels are
900          *    distributed evenly amongst all the NUMA nodes.
901          *    This policy will be the default policy.
902          * 2) Localized: All channels of a given instance of a
903          *    performance critical service will be assigned CPUs
904          *    within a selected NUMA node.
905          */
906         enum hv_numa_policy affinity_policy;
907 
908 };
909 
910 static inline void set_channel_lock_state(struct vmbus_channel *c, bool state)
911 {
912         c->acquire_ring_lock = state;
913 }
914 
915 static inline bool is_hvsock_channel(const struct vmbus_channel *c)
916 {
917         return !!(c->offermsg.offer.chn_flags &
918                   VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER);
919 }
920 
921 static inline void set_channel_signal_state(struct vmbus_channel *c,
922                                             enum hv_signal_policy policy)
923 {
924         c->signal_policy = policy;
925 }
926 
927 static inline void set_channel_affinity_state(struct vmbus_channel *c,
928                                               enum hv_numa_policy policy)
929 {
930         c->affinity_policy = policy;
931 }
932 
933 static inline void set_channel_read_state(struct vmbus_channel *c, bool state)
934 {
935         c->batched_reading = state;
936 }
937 
938 static inline void set_per_channel_state(struct vmbus_channel *c, void *s)
939 {
940         c->per_channel_state = s;
941 }
942 
943 static inline void *get_per_channel_state(struct vmbus_channel *c)
944 {
945         return c->per_channel_state;
946 }
947 
948 static inline void set_channel_pending_send_size(struct vmbus_channel *c,
949                                                  u32 size)
950 {
951         c->outbound.ring_buffer->pending_send_sz = size;
952 }
953 
954 static inline void set_low_latency_mode(struct vmbus_channel *c)
955 {
956         c->low_latency = true;
957 }
958 
959 static inline void clear_low_latency_mode(struct vmbus_channel *c)
960 {
961         c->low_latency = false;
962 }
963 
964 void vmbus_onmessage(void *context);
965 
966 int vmbus_request_offers(void);
967 
968 /*
969  * APIs for managing sub-channels.
970  */
971 
972 void vmbus_set_sc_create_callback(struct vmbus_channel *primary_channel,
973                         void (*sc_cr_cb)(struct vmbus_channel *new_sc));
974 
975 void vmbus_set_chn_rescind_callback(struct vmbus_channel *channel,
976                 void (*chn_rescind_cb)(struct vmbus_channel *));
977 
978 /*
979  * Retrieve the (sub) channel on which to send an outgoing request.
980  * When a primary channel has multiple sub-channels, we choose a
981  * channel whose VCPU binding is closest to the VCPU on which
982  * this call is being made.
983  */
984 struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary);
985 
986 /*
987  * Check if sub-channels have already been offerred. This API will be useful
988  * when the driver is unloaded after establishing sub-channels. In this case,
989  * when the driver is re-loaded, the driver would have to check if the
990  * subchannels have already been established before attempting to request
991  * the creation of sub-channels.
992  * This function returns TRUE to indicate that subchannels have already been
993  * created.
994  * This function should be invoked after setting the callback function for
995  * sub-channel creation.
996  */
997 bool vmbus_are_subchannels_present(struct vmbus_channel *primary);
998 
999 /* The format must be the same as struct vmdata_gpa_direct */
1000 struct vmbus_channel_packet_page_buffer {
1001         u16 type;
1002         u16 dataoffset8;
1003         u16 length8;
1004         u16 flags;
1005         u64 transactionid;
1006         u32 reserved;
1007         u32 rangecount;
1008         struct hv_page_buffer range[MAX_PAGE_BUFFER_COUNT];
1009 } __packed;
1010 
1011 /* The format must be the same as struct vmdata_gpa_direct */
1012 struct vmbus_channel_packet_multipage_buffer {
1013         u16 type;
1014         u16 dataoffset8;
1015         u16 length8;
1016         u16 flags;
1017         u64 transactionid;
1018         u32 reserved;
1019         u32 rangecount;         /* Always 1 in this case */
1020         struct hv_multipage_buffer range;
1021 } __packed;
1022 
1023 /* The format must be the same as struct vmdata_gpa_direct */
1024 struct vmbus_packet_mpb_array {
1025         u16 type;
1026         u16 dataoffset8;
1027         u16 length8;
1028         u16 flags;
1029         u64 transactionid;
1030         u32 reserved;
1031         u32 rangecount;         /* Always 1 in this case */
1032         struct hv_mpb_array range;
1033 } __packed;
1034 
1035 
1036 extern int vmbus_open(struct vmbus_channel *channel,
1037                             u32 send_ringbuffersize,
1038                             u32 recv_ringbuffersize,
1039                             void *userdata,
1040                             u32 userdatalen,
1041                             void(*onchannel_callback)(void *context),
1042                             void *context);
1043 
1044 extern void vmbus_close(struct vmbus_channel *channel);
1045 
1046 extern int vmbus_sendpacket(struct vmbus_channel *channel,
1047                                   void *buffer,
1048                                   u32 bufferLen,
1049                                   u64 requestid,
1050                                   enum vmbus_packet_type type,
1051                                   u32 flags);
1052 
1053 extern int vmbus_sendpacket_ctl(struct vmbus_channel *channel,
1054                                   void *buffer,
1055                                   u32 bufferLen,
1056                                   u64 requestid,
1057                                   enum vmbus_packet_type type,
1058                                   u32 flags,
1059                                   bool kick_q);
1060 
1061 extern int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
1062                                             struct hv_page_buffer pagebuffers[],
1063                                             u32 pagecount,
1064                                             void *buffer,
1065                                             u32 bufferlen,
1066                                             u64 requestid);
1067 
1068 extern int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
1069                                            struct hv_page_buffer pagebuffers[],
1070                                            u32 pagecount,
1071                                            void *buffer,
1072                                            u32 bufferlen,
1073                                            u64 requestid,
1074                                            u32 flags,
1075                                            bool kick_q);
1076 
1077 extern int vmbus_sendpacket_multipagebuffer(struct vmbus_channel *channel,
1078                                         struct hv_multipage_buffer *mpb,
1079                                         void *buffer,
1080                                         u32 bufferlen,
1081                                         u64 requestid);
1082 
1083 extern int vmbus_sendpacket_mpb_desc(struct vmbus_channel *channel,
1084                                      struct vmbus_packet_mpb_array *mpb,
1085                                      u32 desc_size,
1086                                      void *buffer,
1087                                      u32 bufferlen,
1088                                      u64 requestid);
1089 
1090 extern int vmbus_establish_gpadl(struct vmbus_channel *channel,
1091                                       void *kbuffer,
1092                                       u32 size,
1093                                       u32 *gpadl_handle);
1094 
1095 extern int vmbus_teardown_gpadl(struct vmbus_channel *channel,
1096                                      u32 gpadl_handle);
1097 
1098 extern int vmbus_recvpacket(struct vmbus_channel *channel,
1099                                   void *buffer,
1100                                   u32 bufferlen,
1101                                   u32 *buffer_actual_len,
1102                                   u64 *requestid);
1103 
1104 extern int vmbus_recvpacket_raw(struct vmbus_channel *channel,
1105                                      void *buffer,
1106                                      u32 bufferlen,
1107                                      u32 *buffer_actual_len,
1108                                      u64 *requestid);
1109 
1110 
1111 extern void vmbus_ontimer(unsigned long data);
1112 
1113 /* Base driver object */
1114 struct hv_driver {
1115         const char *name;
1116 
1117         /*
1118          * A hvsock offer, which has a VMBUS_CHANNEL_TLNPI_PROVIDER_OFFER
1119          * channel flag, actually doesn't mean a synthetic device because the
1120          * offer's if_type/if_instance can change for every new hvsock
1121          * connection.
1122          *
1123          * However, to facilitate the notification of new-offer/rescind-offer
1124          * from vmbus driver to hvsock driver, we can handle hvsock offer as
1125          * a special vmbus device, and hence we need the below flag to
1126          * indicate if the driver is the hvsock driver or not: we need to
1127          * specially treat the hvosck offer & driver in vmbus_match().
1128          */
1129         bool hvsock;
1130 
1131         /* the device type supported by this driver */
1132         uuid_le dev_type;
1133         const struct hv_vmbus_device_id *id_table;
1134 
1135         struct device_driver driver;
1136 
1137         /* dynamic device GUID's */
1138         struct  {
1139                 spinlock_t lock;
1140                 struct list_head list;
1141         } dynids;
1142 
1143         int (*probe)(struct hv_device *, const struct hv_vmbus_device_id *);
1144         int (*remove)(struct hv_device *);
1145         void (*shutdown)(struct hv_device *);
1146 
1147 };
1148 
1149 /* Base device object */
1150 struct hv_device {
1151         /* the device type id of this device */
1152         uuid_le dev_type;
1153 
1154         /* the device instance id of this device */
1155         uuid_le dev_instance;
1156         u16 vendor_id;
1157         u16 device_id;
1158 
1159         struct device device;
1160 
1161         struct vmbus_channel *channel;
1162 };
1163 
1164 
1165 static inline struct hv_device *device_to_hv_device(struct device *d)
1166 {
1167         return container_of(d, struct hv_device, device);
1168 }
1169 
1170 static inline struct hv_driver *drv_to_hv_drv(struct device_driver *d)
1171 {
1172         return container_of(d, struct hv_driver, driver);
1173 }
1174 
1175 static inline void hv_set_drvdata(struct hv_device *dev, void *data)
1176 {
1177         dev_set_drvdata(&dev->device, data);
1178 }
1179 
1180 static inline void *hv_get_drvdata(struct hv_device *dev)
1181 {
1182         return dev_get_drvdata(&dev->device);
1183 }
1184 
1185 /* Vmbus interface */
1186 #define vmbus_driver_register(driver)   \
1187         __vmbus_driver_register(driver, THIS_MODULE, KBUILD_MODNAME)
1188 int __must_check __vmbus_driver_register(struct hv_driver *hv_driver,
1189                                          struct module *owner,
1190                                          const char *mod_name);
1191 void vmbus_driver_unregister(struct hv_driver *hv_driver);
1192 
1193 void vmbus_hvsock_device_unregister(struct vmbus_channel *channel);
1194 
1195 int vmbus_allocate_mmio(struct resource **new, struct hv_device *device_obj,
1196                         resource_size_t min, resource_size_t max,
1197                         resource_size_t size, resource_size_t align,
1198                         bool fb_overlap_ok);
1199 void vmbus_free_mmio(resource_size_t start, resource_size_t size);
1200 int vmbus_cpu_number_to_vp_number(int cpu_number);
1201 u64 hv_do_hypercall(u64 control, void *input, void *output);
1202 
1203 /*
1204  * GUID definitions of various offer types - services offered to the guest.
1205  */
1206 
1207 /*
1208  * Network GUID
1209  * {f8615163-df3e-46c5-913f-f2d2f965ed0e}
1210  */
1211 #define HV_NIC_GUID \
1212         .guid = UUID_LE(0xf8615163, 0xdf3e, 0x46c5, 0x91, 0x3f, \
1213                         0xf2, 0xd2, 0xf9, 0x65, 0xed, 0x0e)
1214 
1215 /*
1216  * IDE GUID
1217  * {32412632-86cb-44a2-9b5c-50d1417354f5}
1218  */
1219 #define HV_IDE_GUID \
1220         .guid = UUID_LE(0x32412632, 0x86cb, 0x44a2, 0x9b, 0x5c, \
1221                         0x50, 0xd1, 0x41, 0x73, 0x54, 0xf5)
1222 
1223 /*
1224  * SCSI GUID
1225  * {ba6163d9-04a1-4d29-b605-72e2ffb1dc7f}
1226  */
1227 #define HV_SCSI_GUID \
1228         .guid = UUID_LE(0xba6163d9, 0x04a1, 0x4d29, 0xb6, 0x05, \
1229                         0x72, 0xe2, 0xff, 0xb1, 0xdc, 0x7f)
1230 
1231 /*
1232  * Shutdown GUID
1233  * {0e0b6031-5213-4934-818b-38d90ced39db}
1234  */
1235 #define HV_SHUTDOWN_GUID \
1236         .guid = UUID_LE(0x0e0b6031, 0x5213, 0x4934, 0x81, 0x8b, \
1237                         0x38, 0xd9, 0x0c, 0xed, 0x39, 0xdb)
1238 
1239 /*
1240  * Time Synch GUID
1241  * {9527E630-D0AE-497b-ADCE-E80AB0175CAF}
1242  */
1243 #define HV_TS_GUID \
1244         .guid = UUID_LE(0x9527e630, 0xd0ae, 0x497b, 0xad, 0xce, \
1245                         0xe8, 0x0a, 0xb0, 0x17, 0x5c, 0xaf)
1246 
1247 /*
1248  * Heartbeat GUID
1249  * {57164f39-9115-4e78-ab55-382f3bd5422d}
1250  */
1251 #define HV_HEART_BEAT_GUID \
1252         .guid = UUID_LE(0x57164f39, 0x9115, 0x4e78, 0xab, 0x55, \
1253                         0x38, 0x2f, 0x3b, 0xd5, 0x42, 0x2d)
1254 
1255 /*
1256  * KVP GUID
1257  * {a9a0f4e7-5a45-4d96-b827-8a841e8c03e6}
1258  */
1259 #define HV_KVP_GUID \
1260         .guid = UUID_LE(0xa9a0f4e7, 0x5a45, 0x4d96, 0xb8, 0x27, \
1261                         0x8a, 0x84, 0x1e, 0x8c, 0x03, 0xe6)
1262 
1263 /*
1264  * Dynamic memory GUID
1265  * {525074dc-8985-46e2-8057-a307dc18a502}
1266  */
1267 #define HV_DM_GUID \
1268         .guid = UUID_LE(0x525074dc, 0x8985, 0x46e2, 0x80, 0x57, \
1269                         0xa3, 0x07, 0xdc, 0x18, 0xa5, 0x02)
1270 
1271 /*
1272  * Mouse GUID
1273  * {cfa8b69e-5b4a-4cc0-b98b-8ba1a1f3f95a}
1274  */
1275 #define HV_MOUSE_GUID \
1276         .guid = UUID_LE(0xcfa8b69e, 0x5b4a, 0x4cc0, 0xb9, 0x8b, \
1277                         0x8b, 0xa1, 0xa1, 0xf3, 0xf9, 0x5a)
1278 
1279 /*
1280  * Keyboard GUID
1281  * {f912ad6d-2b17-48ea-bd65-f927a61c7684}
1282  */
1283 #define HV_KBD_GUID \
1284         .guid = UUID_LE(0xf912ad6d, 0x2b17, 0x48ea, 0xbd, 0x65, \
1285                         0xf9, 0x27, 0xa6, 0x1c, 0x76, 0x84)
1286 
1287 /*
1288  * VSS (Backup/Restore) GUID
1289  */
1290 #define HV_VSS_GUID \
1291         .guid = UUID_LE(0x35fa2e29, 0xea23, 0x4236, 0x96, 0xae, \
1292                         0x3a, 0x6e, 0xba, 0xcb, 0xa4, 0x40)
1293 /*
1294  * Synthetic Video GUID
1295  * {DA0A7802-E377-4aac-8E77-0558EB1073F8}
1296  */
1297 #define HV_SYNTHVID_GUID \
1298         .guid = UUID_LE(0xda0a7802, 0xe377, 0x4aac, 0x8e, 0x77, \
1299                         0x05, 0x58, 0xeb, 0x10, 0x73, 0xf8)
1300 
1301 /*
1302  * Synthetic FC GUID
1303  * {2f9bcc4a-0069-4af3-b76b-6fd0be528cda}
1304  */
1305 #define HV_SYNTHFC_GUID \
1306         .guid = UUID_LE(0x2f9bcc4a, 0x0069, 0x4af3, 0xb7, 0x6b, \
1307                         0x6f, 0xd0, 0xbe, 0x52, 0x8c, 0xda)
1308 
1309 /*
1310  * Guest File Copy Service
1311  * {34D14BE3-DEE4-41c8-9AE7-6B174977C192}
1312  */
1313 
1314 #define HV_FCOPY_GUID \
1315         .guid = UUID_LE(0x34d14be3, 0xdee4, 0x41c8, 0x9a, 0xe7, \
1316                         0x6b, 0x17, 0x49, 0x77, 0xc1, 0x92)
1317 
1318 /*
1319  * NetworkDirect. This is the guest RDMA service.
1320  * {8c2eaf3d-32a7-4b09-ab99-bd1f1c86b501}
1321  */
1322 #define HV_ND_GUID \
1323         .guid = UUID_LE(0x8c2eaf3d, 0x32a7, 0x4b09, 0xab, 0x99, \
1324                         0xbd, 0x1f, 0x1c, 0x86, 0xb5, 0x01)
1325 
1326 /*
1327  * PCI Express Pass Through
1328  * {44C4F61D-4444-4400-9D52-802E27EDE19F}
1329  */
1330 
1331 #define HV_PCIE_GUID \
1332         .guid = UUID_LE(0x44c4f61d, 0x4444, 0x4400, 0x9d, 0x52, \
1333                         0x80, 0x2e, 0x27, 0xed, 0xe1, 0x9f)
1334 
1335 /*
1336  * Linux doesn't support the 3 devices: the first two are for
1337  * Automatic Virtual Machine Activation, and the third is for
1338  * Remote Desktop Virtualization.
1339  * {f8e65716-3cb3-4a06-9a60-1889c5cccab5}
1340  * {3375baf4-9e15-4b30-b765-67acb10d607b}
1341  * {276aacf4-ac15-426c-98dd-7521ad3f01fe}
1342  */
1343 
1344 #define HV_AVMA1_GUID \
1345         .guid = UUID_LE(0xf8e65716, 0x3cb3, 0x4a06, 0x9a, 0x60, \
1346                         0x18, 0x89, 0xc5, 0xcc, 0xca, 0xb5)
1347 
1348 #define HV_AVMA2_GUID \
1349         .guid = UUID_LE(0x3375baf4, 0x9e15, 0x4b30, 0xb7, 0x65, \
1350                         0x67, 0xac, 0xb1, 0x0d, 0x60, 0x7b)
1351 
1352 #define HV_RDV_GUID \
1353         .guid = UUID_LE(0x276aacf4, 0xac15, 0x426c, 0x98, 0xdd, \
1354                         0x75, 0x21, 0xad, 0x3f, 0x01, 0xfe)
1355 
1356 /*
1357  * Common header for Hyper-V ICs
1358  */
1359 
1360 #define ICMSGTYPE_NEGOTIATE             0
1361 #define ICMSGTYPE_HEARTBEAT             1
1362 #define ICMSGTYPE_KVPEXCHANGE           2
1363 #define ICMSGTYPE_SHUTDOWN              3
1364 #define ICMSGTYPE_TIMESYNC              4
1365 #define ICMSGTYPE_VSS                   5
1366 
1367 #define ICMSGHDRFLAG_TRANSACTION        1
1368 #define ICMSGHDRFLAG_REQUEST            2
1369 #define ICMSGHDRFLAG_RESPONSE           4
1370 
1371 
1372 /*
1373  * While we want to handle util services as regular devices,
1374  * there is only one instance of each of these services; so
1375  * we statically allocate the service specific state.
1376  */
1377 
1378 struct hv_util_service {
1379         u8 *recv_buffer;
1380         void *channel;
1381         void (*util_cb)(void *);
1382         int (*util_init)(struct hv_util_service *);
1383         void (*util_deinit)(void);
1384 };
1385 
1386 struct vmbuspipe_hdr {
1387         u32 flags;
1388         u32 msgsize;
1389 } __packed;
1390 
1391 struct ic_version {
1392         u16 major;
1393         u16 minor;
1394 } __packed;
1395 
1396 struct icmsg_hdr {
1397         struct ic_version icverframe;
1398         u16 icmsgtype;
1399         struct ic_version icvermsg;
1400         u16 icmsgsize;
1401         u32 status;
1402         u8 ictransaction_id;
1403         u8 icflags;
1404         u8 reserved[2];
1405 } __packed;
1406 
1407 struct icmsg_negotiate {
1408         u16 icframe_vercnt;
1409         u16 icmsg_vercnt;
1410         u32 reserved;
1411         struct ic_version icversion_data[1]; /* any size array */
1412 } __packed;
1413 
1414 struct shutdown_msg_data {
1415         u32 reason_code;
1416         u32 timeout_seconds;
1417         u32 flags;
1418         u8  display_message[2048];
1419 } __packed;
1420 
1421 struct heartbeat_msg_data {
1422         u64 seq_num;
1423         u32 reserved[8];
1424 } __packed;
1425 
1426 /* Time Sync IC defs */
1427 #define ICTIMESYNCFLAG_PROBE    0
1428 #define ICTIMESYNCFLAG_SYNC     1
1429 #define ICTIMESYNCFLAG_SAMPLE   2
1430 
1431 #ifdef __x86_64__
1432 #define WLTIMEDELTA     116444736000000000L     /* in 100ns unit */
1433 #else
1434 #define WLTIMEDELTA     116444736000000000LL
1435 #endif
1436 
1437 struct ictimesync_data {
1438         u64 parenttime;
1439         u64 childtime;
1440         u64 roundtriptime;
1441         u8 flags;
1442 } __packed;
1443 
1444 struct ictimesync_ref_data {
1445         u64 parenttime;
1446         u64 vmreferencetime;
1447         u8 flags;
1448         char leapflags;
1449         char stratum;
1450         u8 reserved[3];
1451 } __packed;
1452 
1453 struct hyperv_service_callback {
1454         u8 msg_type;
1455         char *log_msg;
1456         uuid_le data;
1457         struct vmbus_channel *channel;
1458         void (*callback) (void *context);
1459 };
1460 
1461 #define MAX_SRV_VER     0x7ffffff
1462 extern bool vmbus_prep_negotiate_resp(struct icmsg_hdr *,
1463                                         struct icmsg_negotiate *, u8 *, int,
1464                                         int);
1465 
1466 void hv_event_tasklet_disable(struct vmbus_channel *channel);
1467 void hv_event_tasklet_enable(struct vmbus_channel *channel);
1468 
1469 void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid);
1470 
1471 void vmbus_setevent(struct vmbus_channel *channel);
1472 /*
1473  * Negotiated version with the Host.
1474  */
1475 
1476 extern __u32 vmbus_proto_version;
1477 
1478 int vmbus_send_tl_connect_request(const uuid_le *shv_guest_servie_id,
1479                                   const uuid_le *shv_host_servie_id);
1480 void vmbus_set_event(struct vmbus_channel *channel);
1481 
1482 /* Get the start of the ring buffer. */
1483 static inline void *
1484 hv_get_ring_buffer(struct hv_ring_buffer_info *ring_info)
1485 {
1486         return (void *)ring_info->ring_buffer->buffer;
1487 }
1488 
1489 /*
1490  * To optimize the flow management on the send-side,
1491  * when the sender is blocked because of lack of
1492  * sufficient space in the ring buffer, potential the
1493  * consumer of the ring buffer can signal the producer.
1494  * This is controlled by the following parameters:
1495  *
1496  * 1. pending_send_sz: This is the size in bytes that the
1497  *    producer is trying to send.
1498  * 2. The feature bit feat_pending_send_sz set to indicate if
1499  *    the consumer of the ring will signal when the ring
1500  *    state transitions from being full to a state where
1501  *    there is room for the producer to send the pending packet.
1502  */
1503 
1504 static inline  void hv_signal_on_read(struct vmbus_channel *channel)
1505 {
1506         u32 cur_write_sz, cached_write_sz;
1507         u32 pending_sz;
1508         struct hv_ring_buffer_info *rbi = &channel->inbound;
1509 
1510         /*
1511          * Issue a full memory barrier before making the signaling decision.
1512          * Here is the reason for having this barrier:
1513          * If the reading of the pend_sz (in this function)
1514          * were to be reordered and read before we commit the new read
1515          * index (in the calling function)  we could
1516          * have a problem. If the host were to set the pending_sz after we
1517          * have sampled pending_sz and go to sleep before we commit the
1518          * read index, we could miss sending the interrupt. Issue a full
1519          * memory barrier to address this.
1520          */
1521         virt_mb();
1522 
1523         pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
1524         /* If the other end is not blocked on write don't bother. */
1525         if (pending_sz == 0)
1526                 return;
1527 
1528         cur_write_sz = hv_get_bytes_to_write(rbi);
1529 
1530         if (cur_write_sz < pending_sz)
1531                 return;
1532 
1533         cached_write_sz = hv_get_cached_bytes_to_write(rbi);
1534         if (cached_write_sz < pending_sz)
1535                 vmbus_setevent(channel);
1536 
1537         return;
1538 }
1539 
1540 static inline void
1541 init_cached_read_index(struct vmbus_channel *channel)
1542 {
1543         struct hv_ring_buffer_info *rbi = &channel->inbound;
1544 
1545         rbi->cached_read_index = rbi->ring_buffer->read_index;
1546 }
1547 
1548 /*
1549  * An API to support in-place processing of incoming VMBUS packets.
1550  */
1551 #define VMBUS_PKT_TRAILER       8
1552 
1553 static inline struct vmpacket_descriptor *
1554 get_next_pkt_raw(struct vmbus_channel *channel)
1555 {
1556         struct hv_ring_buffer_info *ring_info = &channel->inbound;
1557         u32 priv_read_loc = ring_info->priv_read_index;
1558         void *ring_buffer = hv_get_ring_buffer(ring_info);
1559         u32 dsize = ring_info->ring_datasize;
1560         /*
1561          * delta is the difference between what is available to read and
1562          * what was already consumed in place. We commit read index after
1563          * the whole batch is processed.
1564          */
1565         u32 delta = priv_read_loc >= ring_info->ring_buffer->read_index ?
1566                 priv_read_loc - ring_info->ring_buffer->read_index :
1567                 (dsize - ring_info->ring_buffer->read_index) + priv_read_loc;
1568         u32 bytes_avail_toread = (hv_get_bytes_to_read(ring_info) - delta);
1569 
1570         if (bytes_avail_toread < sizeof(struct vmpacket_descriptor))
1571                 return NULL;
1572 
1573         return ring_buffer + priv_read_loc;
1574 }
1575 
1576 /*
1577  * A helper function to step through packets "in-place"
1578  * This API is to be called after each successful call
1579  * get_next_pkt_raw().
1580  */
1581 static inline void put_pkt_raw(struct vmbus_channel *channel,
1582                                 struct vmpacket_descriptor *desc)
1583 {
1584         struct hv_ring_buffer_info *ring_info = &channel->inbound;
1585         u32 packetlen = desc->len8 << 3;
1586         u32 dsize = ring_info->ring_datasize;
1587 
1588         /*
1589          * Include the packet trailer.
1590          */
1591         ring_info->priv_read_index += packetlen + VMBUS_PKT_TRAILER;
1592         ring_info->priv_read_index %= dsize;
1593 }
1594 
1595 /*
1596  * This call commits the read index and potentially signals the host.
1597  * Here is the pattern for using the "in-place" consumption APIs:
1598  *
1599  * init_cached_read_index();
1600  *
1601  * while (get_next_pkt_raw() {
1602  *      process the packet "in-place";
1603  *      put_pkt_raw();
1604  * }
1605  * if (packets processed in place)
1606  *      commit_rd_index();
1607  */
1608 static inline void commit_rd_index(struct vmbus_channel *channel)
1609 {
1610         struct hv_ring_buffer_info *ring_info = &channel->inbound;
1611         /*
1612          * Make sure all reads are done before we update the read index since
1613          * the writer may start writing to the read area once the read index
1614          * is updated.
1615          */
1616         virt_rmb();
1617         ring_info->ring_buffer->read_index = ring_info->priv_read_index;
1618 
1619         hv_signal_on_read(channel);
1620 }
1621 
1622 
1623 #endif /* _HYPERV_H */
1624 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp