]> git.karo-electronics.de Git - karo-tx-linux.git/blob - drivers/staging/hv/blkvsc_drv.c
block: push down BKL into .open and .release
[karo-tx-linux.git] / drivers / staging / hv / blkvsc_drv.c
1 /*
2  * Copyright (c) 2009, Microsoft Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * You should have received a copy of the GNU General Public License along with
14  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
15  * Place - Suite 330, Boston, MA 02111-1307 USA.
16  *
17  * Authors:
18  *   Haiyang Zhang <haiyangz@microsoft.com>
19  *   Hank Janssen  <hjanssen@microsoft.com>
20  */
21 #include <linux/init.h>
22 #include <linux/module.h>
23 #include <linux/device.h>
24 #include <linux/blkdev.h>
25 #include <linux/major.h>
26 #include <linux/delay.h>
27 #include <linux/hdreg.h>
28 #include <linux/smp_lock.h>
29 #include <linux/slab.h>
30 #include <scsi/scsi.h>
31 #include <scsi/scsi_cmnd.h>
32 #include <scsi/scsi_eh.h>
33 #include <scsi/scsi_dbg.h>
34 #include "osd.h"
35 #include "logging.h"
36 #include "version_info.h"
37 #include "vmbus.h"
38 #include "storvsc_api.h"
39
40
41 #define BLKVSC_MINORS   64
42
43 enum blkvsc_device_type {
44         UNKNOWN_DEV_TYPE,
45         HARDDISK_TYPE,
46         DVD_TYPE,
47 };
48
49 /*
50  * This request ties the struct request and struct
51  * blkvsc_request/hv_storvsc_request together A struct request may be
52  * represented by 1 or more struct blkvsc_request
53  */
54 struct blkvsc_request_group {
55         int outstanding;
56         int status;
57         struct list_head blkvsc_req_list;       /* list of blkvsc_requests */
58 };
59
60 struct blkvsc_request {
61         /* blkvsc_request_group.blkvsc_req_list */
62         struct list_head req_entry;
63
64         /* block_device_context.pending_list */
65         struct list_head pend_entry;
66
67         /* This may be null if we generate a request internally */
68         struct request *req;
69
70         struct block_device_context *dev;
71
72         /* The group this request is part of. Maybe null */
73         struct blkvsc_request_group *group;
74
75         wait_queue_head_t wevent;
76         int cond;
77
78         int write;
79         sector_t sector_start;
80         unsigned long sector_count;
81
82         unsigned char sense_buffer[SCSI_SENSE_BUFFERSIZE];
83         unsigned char cmd_len;
84         unsigned char cmnd[MAX_COMMAND_SIZE];
85
86         struct hv_storvsc_request request;
87         /*
88          * !!!DO NOT ADD ANYTHING BELOW HERE!!! Otherwise, memory can overlap,
89          * because - The extension buffer falls right here and is pointed to by
90          * request.Extension;
91          * Which sounds like a horrible idea, who designed this?
92          */
93 };
94
95 /* Per device structure */
96 struct block_device_context {
97         /* point back to our device context */
98         struct vm_device *device_ctx;
99         struct kmem_cache *request_pool;
100         spinlock_t lock;
101         struct gendisk *gd;
102         enum blkvsc_device_type device_type;
103         struct list_head pending_list;
104
105         unsigned char device_id[64];
106         unsigned int device_id_len;
107         int num_outstanding_reqs;
108         int shutting_down;
109         int media_not_present;
110         unsigned int sector_size;
111         sector_t capacity;
112         unsigned int port;
113         unsigned char path;
114         unsigned char target;
115         int users;
116 };
117
118 /* Per driver */
119 struct blkvsc_driver_context {
120         /* !! These must be the first 2 fields !! */
121         /* FIXME this is a bug! */
122         struct driver_context drv_ctx;
123         struct storvsc_driver_object drv_obj;
124 };
125
126 /* Static decl */
127 static int blkvsc_probe(struct device *dev);
128 static int blkvsc_remove(struct device *device);
129 static void blkvsc_shutdown(struct device *device);
130
131 static int blkvsc_open(struct block_device *bdev,  fmode_t mode);
132 static int blkvsc_release(struct gendisk *disk, fmode_t mode);
133 static int blkvsc_media_changed(struct gendisk *gd);
134 static int blkvsc_revalidate_disk(struct gendisk *gd);
135 static int blkvsc_getgeo(struct block_device *bd, struct hd_geometry *hg);
136 static int blkvsc_ioctl(struct block_device *bd, fmode_t mode,
137                         unsigned cmd, unsigned long argument);
138 static void blkvsc_request(struct request_queue *queue);
139 static void blkvsc_request_completion(struct hv_storvsc_request *request);
140 static int blkvsc_do_request(struct block_device_context *blkdev,
141                              struct request *req);
142 static int blkvsc_submit_request(struct blkvsc_request *blkvsc_req,
143                 void (*request_completion)(struct hv_storvsc_request *));
144 static void blkvsc_init_rw(struct blkvsc_request *blkvsc_req);
145 static void blkvsc_cmd_completion(struct hv_storvsc_request *request);
146 static int blkvsc_do_inquiry(struct block_device_context *blkdev);
147 static int blkvsc_do_read_capacity(struct block_device_context *blkdev);
148 static int blkvsc_do_read_capacity16(struct block_device_context *blkdev);
149 static int blkvsc_do_flush(struct block_device_context *blkdev);
150 static int blkvsc_cancel_pending_reqs(struct block_device_context *blkdev);
151 static int blkvsc_do_pending_reqs(struct block_device_context *blkdev);
152
153 static int blkvsc_ringbuffer_size = BLKVSC_RING_BUFFER_SIZE;
154 module_param(blkvsc_ringbuffer_size, int, S_IRUGO);
155 MODULE_PARM_DESC(ring_size, "Ring buffer size (in bytes)");
156
157 /* The one and only one */
158 static struct blkvsc_driver_context g_blkvsc_drv;
159
160 static const struct block_device_operations block_ops = {
161         .owner = THIS_MODULE,
162         .open = blkvsc_open,
163         .release = blkvsc_release,
164         .media_changed = blkvsc_media_changed,
165         .revalidate_disk = blkvsc_revalidate_disk,
166         .getgeo = blkvsc_getgeo,
167         .ioctl  = blkvsc_ioctl,
168 };
169
170 /*
171  * blkvsc_drv_init -  BlkVsc driver initialization.
172  */
173 static int blkvsc_drv_init(int (*drv_init)(struct hv_driver *drv))
174 {
175         struct storvsc_driver_object *storvsc_drv_obj = &g_blkvsc_drv.drv_obj;
176         struct driver_context *drv_ctx = &g_blkvsc_drv.drv_ctx;
177         int ret;
178
179         DPRINT_ENTER(BLKVSC_DRV);
180
181         vmbus_get_interface(&storvsc_drv_obj->Base.VmbusChannelInterface);
182
183         storvsc_drv_obj->RingBufferSize = blkvsc_ringbuffer_size;
184
185         /* Callback to client driver to complete the initialization */
186         drv_init(&storvsc_drv_obj->Base);
187
188         drv_ctx->driver.name = storvsc_drv_obj->Base.name;
189         memcpy(&drv_ctx->class_id, &storvsc_drv_obj->Base.deviceType,
190                sizeof(struct hv_guid));
191
192         drv_ctx->probe = blkvsc_probe;
193         drv_ctx->remove = blkvsc_remove;
194         drv_ctx->shutdown = blkvsc_shutdown;
195
196         /* The driver belongs to vmbus */
197         ret = vmbus_child_driver_register(drv_ctx);
198
199         DPRINT_EXIT(BLKVSC_DRV);
200
201         return ret;
202 }
203
204 static int blkvsc_drv_exit_cb(struct device *dev, void *data)
205 {
206         struct device **curr = (struct device **)data;
207         *curr = dev;
208         return 1; /* stop iterating */
209 }
210
211 static void blkvsc_drv_exit(void)
212 {
213         struct storvsc_driver_object *storvsc_drv_obj = &g_blkvsc_drv.drv_obj;
214         struct driver_context *drv_ctx = &g_blkvsc_drv.drv_ctx;
215         struct device *current_dev;
216         int ret;
217
218         DPRINT_ENTER(BLKVSC_DRV);
219
220         while (1) {
221                 current_dev = NULL;
222
223                 /* Get the device */
224                 ret = driver_for_each_device(&drv_ctx->driver, NULL,
225                                              (void *) &current_dev,
226                                              blkvsc_drv_exit_cb);
227
228                 if (ret)
229                         DPRINT_WARN(BLKVSC_DRV,
230                                     "driver_for_each_device returned %d", ret);
231
232
233                 if (current_dev == NULL)
234                         break;
235
236                 /* Initiate removal from the top-down */
237                 device_unregister(current_dev);
238         }
239
240         if (storvsc_drv_obj->Base.OnCleanup)
241                 storvsc_drv_obj->Base.OnCleanup(&storvsc_drv_obj->Base);
242
243         vmbus_child_driver_unregister(drv_ctx);
244
245         DPRINT_EXIT(BLKVSC_DRV);
246
247         return;
248 }
249
250 /*
251  * blkvsc_probe - Add a new device for this driver
252  */
253 static int blkvsc_probe(struct device *device)
254 {
255         struct driver_context *driver_ctx =
256                                 driver_to_driver_context(device->driver);
257         struct blkvsc_driver_context *blkvsc_drv_ctx =
258                                 (struct blkvsc_driver_context *)driver_ctx;
259         struct storvsc_driver_object *storvsc_drv_obj =
260                                 &blkvsc_drv_ctx->drv_obj;
261         struct vm_device *device_ctx = device_to_vm_device(device);
262         struct hv_device *device_obj = &device_ctx->device_obj;
263
264         struct block_device_context *blkdev = NULL;
265         struct storvsc_device_info device_info;
266         int major = 0;
267         int devnum = 0;
268         int ret = 0;
269         static int ide0_registered;
270         static int ide1_registered;
271
272         DPRINT_ENTER(BLKVSC_DRV);
273
274         DPRINT_DBG(BLKVSC_DRV, "blkvsc_probe - enter");
275
276         if (!storvsc_drv_obj->Base.OnDeviceAdd) {
277                 DPRINT_ERR(BLKVSC_DRV, "OnDeviceAdd() not set");
278                 ret = -1;
279                 goto Cleanup;
280         }
281
282         blkdev = kzalloc(sizeof(struct block_device_context), GFP_KERNEL);
283         if (!blkdev) {
284                 ret = -ENOMEM;
285                 goto Cleanup;
286         }
287
288         INIT_LIST_HEAD(&blkdev->pending_list);
289
290         /* Initialize what we can here */
291         spin_lock_init(&blkdev->lock);
292
293         /* ASSERT(sizeof(struct blkvsc_request_group) <= */
294         /*      sizeof(struct blkvsc_request)); */
295
296         blkdev->request_pool = kmem_cache_create(dev_name(&device_ctx->device),
297                                         sizeof(struct blkvsc_request) +
298                                         storvsc_drv_obj->RequestExtSize, 0,
299                                         SLAB_HWCACHE_ALIGN, NULL);
300         if (!blkdev->request_pool) {
301                 ret = -ENOMEM;
302                 goto Cleanup;
303         }
304
305
306         /* Call to the vsc driver to add the device */
307         ret = storvsc_drv_obj->Base.OnDeviceAdd(device_obj, &device_info);
308         if (ret != 0) {
309                 DPRINT_ERR(BLKVSC_DRV, "unable to add blkvsc device");
310                 goto Cleanup;
311         }
312
313         blkdev->device_ctx = device_ctx;
314         /* this identified the device 0 or 1 */
315         blkdev->target = device_info.TargetId;
316         /* this identified the ide ctrl 0 or 1 */
317         blkdev->path = device_info.PathId;
318
319         dev_set_drvdata(device, blkdev);
320
321         /* Calculate the major and device num */
322         if (blkdev->path == 0) {
323                 major = IDE0_MAJOR;
324                 devnum = blkdev->path + blkdev->target;         /* 0 or 1 */
325
326                 if (!ide0_registered) {
327                         ret = register_blkdev(major, "ide");
328                         if (ret != 0) {
329                                 DPRINT_ERR(BLKVSC_DRV,
330                                            "register_blkdev() failed! ret %d",
331                                            ret);
332                                 goto Remove;
333                         }
334
335                         ide0_registered = 1;
336                 }
337         } else if (blkdev->path == 1) {
338                 major = IDE1_MAJOR;
339                 devnum = blkdev->path + blkdev->target + 1; /* 2 or 3 */
340
341                 if (!ide1_registered) {
342                         ret = register_blkdev(major, "ide");
343                         if (ret != 0) {
344                                 DPRINT_ERR(BLKVSC_DRV,
345                                            "register_blkdev() failed! ret %d",
346                                            ret);
347                                 goto Remove;
348                         }
349
350                         ide1_registered = 1;
351                 }
352         } else {
353                 DPRINT_ERR(BLKVSC_DRV, "invalid pathid");
354                 ret = -1;
355                 goto Cleanup;
356         }
357
358         DPRINT_INFO(BLKVSC_DRV, "blkvsc registered for major %d!!", major);
359
360         blkdev->gd = alloc_disk(BLKVSC_MINORS);
361         if (!blkdev->gd) {
362                 DPRINT_ERR(BLKVSC_DRV, "register_blkdev() failed! ret %d", ret);
363                 ret = -1;
364                 goto Cleanup;
365         }
366
367         blkdev->gd->queue = blk_init_queue(blkvsc_request, &blkdev->lock);
368
369         blk_queue_max_segment_size(blkdev->gd->queue, PAGE_SIZE);
370         blk_queue_max_segments(blkdev->gd->queue, MAX_MULTIPAGE_BUFFER_COUNT);
371         blk_queue_segment_boundary(blkdev->gd->queue, PAGE_SIZE-1);
372         blk_queue_bounce_limit(blkdev->gd->queue, BLK_BOUNCE_ANY);
373         blk_queue_dma_alignment(blkdev->gd->queue, 511);
374
375         blkdev->gd->major = major;
376         if (devnum == 1 || devnum == 3)
377                 blkdev->gd->first_minor = BLKVSC_MINORS;
378         else
379                 blkdev->gd->first_minor = 0;
380         blkdev->gd->fops = &block_ops;
381         blkdev->gd->private_data = blkdev;
382         sprintf(blkdev->gd->disk_name, "hd%c", 'a' + devnum);
383
384         blkvsc_do_inquiry(blkdev);
385         if (blkdev->device_type == DVD_TYPE) {
386                 set_disk_ro(blkdev->gd, 1);
387                 blkdev->gd->flags |= GENHD_FL_REMOVABLE;
388                 blkvsc_do_read_capacity(blkdev);
389         } else {
390                 blkvsc_do_read_capacity16(blkdev);
391         }
392
393         set_capacity(blkdev->gd, blkdev->capacity * (blkdev->sector_size/512));
394         blk_queue_logical_block_size(blkdev->gd->queue, blkdev->sector_size);
395         /* go! */
396         add_disk(blkdev->gd);
397
398         DPRINT_INFO(BLKVSC_DRV, "%s added!! capacity %lu sector_size %d",
399                     blkdev->gd->disk_name, (unsigned long)blkdev->capacity,
400                     blkdev->sector_size);
401
402         return ret;
403
404 Remove:
405         storvsc_drv_obj->Base.OnDeviceRemove(device_obj);
406
407 Cleanup:
408         if (blkdev) {
409                 if (blkdev->request_pool) {
410                         kmem_cache_destroy(blkdev->request_pool);
411                         blkdev->request_pool = NULL;
412                 }
413                 kfree(blkdev);
414                 blkdev = NULL;
415         }
416
417         DPRINT_EXIT(BLKVSC_DRV);
418
419         return ret;
420 }
421
422 static void blkvsc_shutdown(struct device *device)
423 {
424         struct block_device_context *blkdev = dev_get_drvdata(device);
425         unsigned long flags;
426
427         if (!blkdev)
428                 return;
429
430         DPRINT_DBG(BLKVSC_DRV, "blkvsc_shutdown - users %d disk %s\n",
431                    blkdev->users, blkdev->gd->disk_name);
432
433         spin_lock_irqsave(&blkdev->lock, flags);
434
435         blkdev->shutting_down = 1;
436
437         blk_stop_queue(blkdev->gd->queue);
438
439         spin_unlock_irqrestore(&blkdev->lock, flags);
440
441         while (blkdev->num_outstanding_reqs) {
442                 DPRINT_INFO(STORVSC, "waiting for %d requests to complete...",
443                             blkdev->num_outstanding_reqs);
444                 udelay(100);
445         }
446
447         blkvsc_do_flush(blkdev);
448
449         spin_lock_irqsave(&blkdev->lock, flags);
450
451         blkvsc_cancel_pending_reqs(blkdev);
452
453         spin_unlock_irqrestore(&blkdev->lock, flags);
454 }
455
456 static int blkvsc_do_flush(struct block_device_context *blkdev)
457 {
458         struct blkvsc_request *blkvsc_req;
459
460         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_flush()\n");
461
462         if (blkdev->device_type != HARDDISK_TYPE)
463                 return 0;
464
465         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
466         if (!blkvsc_req)
467                 return -ENOMEM;
468
469         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
470         init_waitqueue_head(&blkvsc_req->wevent);
471         blkvsc_req->dev = blkdev;
472         blkvsc_req->req = NULL;
473         blkvsc_req->write = 0;
474
475         blkvsc_req->request.DataBuffer.PfnArray[0] = 0;
476         blkvsc_req->request.DataBuffer.Offset = 0;
477         blkvsc_req->request.DataBuffer.Length = 0;
478
479         blkvsc_req->cmnd[0] = SYNCHRONIZE_CACHE;
480         blkvsc_req->cmd_len = 10;
481
482         /*
483          * Set this here since the completion routine may be invoked and
484          * completed before we return
485          */
486         blkvsc_req->cond = 0;
487         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
488
489         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
490
491         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
492
493         return 0;
494 }
495
496 /* Do a scsi INQUIRY cmd here to get the device type (ie disk or dvd) */
497 static int blkvsc_do_inquiry(struct block_device_context *blkdev)
498 {
499         struct blkvsc_request *blkvsc_req;
500         struct page *page_buf;
501         unsigned char *buf;
502         unsigned char device_type;
503
504         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_inquiry()\n");
505
506         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
507         if (!blkvsc_req)
508                 return -ENOMEM;
509
510         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
511         page_buf = alloc_page(GFP_KERNEL);
512         if (!page_buf) {
513                 kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
514                 return -ENOMEM;
515         }
516
517         init_waitqueue_head(&blkvsc_req->wevent);
518         blkvsc_req->dev = blkdev;
519         blkvsc_req->req = NULL;
520         blkvsc_req->write = 0;
521
522         blkvsc_req->request.DataBuffer.PfnArray[0] = page_to_pfn(page_buf);
523         blkvsc_req->request.DataBuffer.Offset = 0;
524         blkvsc_req->request.DataBuffer.Length = 64;
525
526         blkvsc_req->cmnd[0] = INQUIRY;
527         blkvsc_req->cmnd[1] = 0x1;              /* Get product data */
528         blkvsc_req->cmnd[2] = 0x83;             /* mode page 83 */
529         blkvsc_req->cmnd[4] = 64;
530         blkvsc_req->cmd_len = 6;
531
532         /*
533          * Set this here since the completion routine may be invoked and
534          * completed before we return
535          */
536         blkvsc_req->cond = 0;
537
538         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
539
540         DPRINT_DBG(BLKVSC_DRV, "waiting %p to complete - cond %d\n",
541                    blkvsc_req, blkvsc_req->cond);
542
543         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
544
545         buf = kmap(page_buf);
546
547         /* print_hex_dump_bytes("", DUMP_PREFIX_NONE, buf, 64); */
548         /* be to le */
549         device_type = buf[0] & 0x1F;
550
551         if (device_type == 0x0) {
552                 blkdev->device_type = HARDDISK_TYPE;
553         } else if (device_type == 0x5) {
554                 blkdev->device_type = DVD_TYPE;
555         } else {
556                 /* TODO: this is currently unsupported device type */
557                 blkdev->device_type = UNKNOWN_DEV_TYPE;
558         }
559
560         DPRINT_DBG(BLKVSC_DRV, "device type %d\n", device_type);
561
562         blkdev->device_id_len = buf[7];
563         if (blkdev->device_id_len > 64)
564                 blkdev->device_id_len = 64;
565
566         memcpy(blkdev->device_id, &buf[8], blkdev->device_id_len);
567         /* printk_hex_dump_bytes("", DUMP_PREFIX_NONE, blkdev->device_id,
568          * blkdev->device_id_len); */
569
570         kunmap(page_buf);
571
572         __free_page(page_buf);
573
574         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
575
576         return 0;
577 }
578
579 /* Do a scsi READ_CAPACITY cmd here to get the size of the disk */
580 static int blkvsc_do_read_capacity(struct block_device_context *blkdev)
581 {
582         struct blkvsc_request *blkvsc_req;
583         struct page *page_buf;
584         unsigned char *buf;
585         struct scsi_sense_hdr sense_hdr;
586
587         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_read_capacity()\n");
588
589         blkdev->sector_size = 0;
590         blkdev->capacity = 0;
591         blkdev->media_not_present = 0; /* assume a disk is present */
592
593         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
594         if (!blkvsc_req)
595                 return -ENOMEM;
596
597         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
598         page_buf = alloc_page(GFP_KERNEL);
599         if (!page_buf) {
600                 kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
601                 return -ENOMEM;
602         }
603
604         init_waitqueue_head(&blkvsc_req->wevent);
605         blkvsc_req->dev = blkdev;
606         blkvsc_req->req = NULL;
607         blkvsc_req->write = 0;
608
609         blkvsc_req->request.DataBuffer.PfnArray[0] = page_to_pfn(page_buf);
610         blkvsc_req->request.DataBuffer.Offset = 0;
611         blkvsc_req->request.DataBuffer.Length = 8;
612
613         blkvsc_req->cmnd[0] = READ_CAPACITY;
614         blkvsc_req->cmd_len = 16;
615
616         /*
617          * Set this here since the completion routine may be invoked
618          * and completed before we return
619          */
620         blkvsc_req->cond = 0;
621
622         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
623
624         DPRINT_DBG(BLKVSC_DRV, "waiting %p to complete - cond %d\n",
625                    blkvsc_req, blkvsc_req->cond);
626
627         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
628
629         /* check error */
630         if (blkvsc_req->request.Status) {
631                 scsi_normalize_sense(blkvsc_req->sense_buffer,
632                                      SCSI_SENSE_BUFFERSIZE, &sense_hdr);
633
634                 if (sense_hdr.asc == 0x3A) {
635                         /* Medium not present */
636                         blkdev->media_not_present = 1;
637                 }
638                 return 0;
639         }
640         buf = kmap(page_buf);
641
642         /* be to le */
643         blkdev->capacity = ((buf[0] << 24) | (buf[1] << 16) |
644                             (buf[2] << 8) | buf[3]) + 1;
645         blkdev->sector_size = (buf[4] << 24) | (buf[5] << 16) |
646                               (buf[6] << 8) | buf[7];
647
648         kunmap(page_buf);
649
650         __free_page(page_buf);
651
652         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
653
654         return 0;
655 }
656
657 static int blkvsc_do_read_capacity16(struct block_device_context *blkdev)
658 {
659         struct blkvsc_request *blkvsc_req;
660         struct page *page_buf;
661         unsigned char *buf;
662         struct scsi_sense_hdr sense_hdr;
663
664         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_read_capacity16()\n");
665
666         blkdev->sector_size = 0;
667         blkdev->capacity = 0;
668         blkdev->media_not_present = 0; /* assume a disk is present */
669
670         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
671         if (!blkvsc_req)
672                 return -ENOMEM;
673
674         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
675         page_buf = alloc_page(GFP_KERNEL);
676         if (!page_buf) {
677                 kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
678                 return -ENOMEM;
679         }
680
681         init_waitqueue_head(&blkvsc_req->wevent);
682         blkvsc_req->dev = blkdev;
683         blkvsc_req->req = NULL;
684         blkvsc_req->write = 0;
685
686         blkvsc_req->request.DataBuffer.PfnArray[0] = page_to_pfn(page_buf);
687         blkvsc_req->request.DataBuffer.Offset = 0;
688         blkvsc_req->request.DataBuffer.Length = 12;
689
690         blkvsc_req->cmnd[0] = 0x9E; /* READ_CAPACITY16; */
691         blkvsc_req->cmd_len = 16;
692
693         /*
694          * Set this here since the completion routine may be invoked
695          * and completed before we return
696          */
697         blkvsc_req->cond = 0;
698
699         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
700
701         DPRINT_DBG(BLKVSC_DRV, "waiting %p to complete - cond %d\n",
702                    blkvsc_req, blkvsc_req->cond);
703
704         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
705
706         /* check error */
707         if (blkvsc_req->request.Status) {
708                 scsi_normalize_sense(blkvsc_req->sense_buffer,
709                                      SCSI_SENSE_BUFFERSIZE, &sense_hdr);
710                 if (sense_hdr.asc == 0x3A) {
711                         /* Medium not present */
712                         blkdev->media_not_present = 1;
713                 }
714                 return 0;
715         }
716         buf = kmap(page_buf);
717
718         /* be to le */
719         blkdev->capacity = be64_to_cpu(*(unsigned long long *) &buf[0]) + 1;
720         blkdev->sector_size = be32_to_cpu(*(unsigned int *)&buf[8]);
721
722 #if 0
723         blkdev->capacity = ((buf[0] << 24) | (buf[1] << 16) |
724                             (buf[2] << 8) | buf[3]) + 1;
725         blkdev->sector_size = (buf[4] << 24) | (buf[5] << 16) |
726                               (buf[6] << 8) | buf[7];
727 #endif
728
729         kunmap(page_buf);
730
731         __free_page(page_buf);
732
733         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
734
735         return 0;
736 }
737
738 /*
739  * blkvsc_remove() - Callback when our device is removed
740  */
741 static int blkvsc_remove(struct device *device)
742 {
743         struct driver_context *driver_ctx =
744                                 driver_to_driver_context(device->driver);
745         struct blkvsc_driver_context *blkvsc_drv_ctx =
746                                 (struct blkvsc_driver_context *)driver_ctx;
747         struct storvsc_driver_object *storvsc_drv_obj =
748                                 &blkvsc_drv_ctx->drv_obj;
749         struct vm_device *device_ctx = device_to_vm_device(device);
750         struct hv_device *device_obj = &device_ctx->device_obj;
751         struct block_device_context *blkdev = dev_get_drvdata(device);
752         unsigned long flags;
753         int ret;
754
755         DPRINT_ENTER(BLKVSC_DRV);
756
757         DPRINT_DBG(BLKVSC_DRV, "blkvsc_remove()\n");
758
759         if (!storvsc_drv_obj->Base.OnDeviceRemove) {
760                 DPRINT_EXIT(BLKVSC_DRV);
761                 return -1;
762         }
763
764         /*
765          * Call to the vsc driver to let it know that the device is being
766          * removed
767          */
768         ret = storvsc_drv_obj->Base.OnDeviceRemove(device_obj);
769         if (ret != 0) {
770                 /* TODO: */
771                 DPRINT_ERR(BLKVSC_DRV,
772                            "unable to remove blkvsc device (ret %d)", ret);
773         }
774
775         /* Get to a known state */
776         spin_lock_irqsave(&blkdev->lock, flags);
777
778         blkdev->shutting_down = 1;
779
780         blk_stop_queue(blkdev->gd->queue);
781
782         spin_unlock_irqrestore(&blkdev->lock, flags);
783
784         while (blkdev->num_outstanding_reqs) {
785                 DPRINT_INFO(STORVSC, "waiting for %d requests to complete...",
786                             blkdev->num_outstanding_reqs);
787                 udelay(100);
788         }
789
790         blkvsc_do_flush(blkdev);
791
792         spin_lock_irqsave(&blkdev->lock, flags);
793
794         blkvsc_cancel_pending_reqs(blkdev);
795
796         spin_unlock_irqrestore(&blkdev->lock, flags);
797
798         blk_cleanup_queue(blkdev->gd->queue);
799
800         del_gendisk(blkdev->gd);
801
802         kmem_cache_destroy(blkdev->request_pool);
803
804         kfree(blkdev);
805
806         DPRINT_EXIT(BLKVSC_DRV);
807
808         return ret;
809 }
810
811 static void blkvsc_init_rw(struct blkvsc_request *blkvsc_req)
812 {
813         /* ASSERT(blkvsc_req->req); */
814         /* ASSERT(blkvsc_req->sector_count <= (MAX_MULTIPAGE_BUFFER_COUNT*8)); */
815
816         blkvsc_req->cmd_len = 16;
817
818         if (blkvsc_req->sector_start > 0xffffffff) {
819                 if (rq_data_dir(blkvsc_req->req)) {
820                         blkvsc_req->write = 1;
821                         blkvsc_req->cmnd[0] = WRITE_16;
822                 } else {
823                         blkvsc_req->write = 0;
824                         blkvsc_req->cmnd[0] = READ_16;
825                 }
826
827                 blkvsc_req->cmnd[1] |=
828                         (blkvsc_req->req->cmd_flags & REQ_FUA) ? 0x8 : 0;
829
830                 *(unsigned long long *)&blkvsc_req->cmnd[2] =
831                                 cpu_to_be64(blkvsc_req->sector_start);
832                 *(unsigned int *)&blkvsc_req->cmnd[10] =
833                                 cpu_to_be32(blkvsc_req->sector_count);
834         } else if ((blkvsc_req->sector_count > 0xff) ||
835                    (blkvsc_req->sector_start > 0x1fffff)) {
836                 if (rq_data_dir(blkvsc_req->req)) {
837                         blkvsc_req->write = 1;
838                         blkvsc_req->cmnd[0] = WRITE_10;
839                 } else {
840                         blkvsc_req->write = 0;
841                         blkvsc_req->cmnd[0] = READ_10;
842                 }
843
844                 blkvsc_req->cmnd[1] |=
845                         (blkvsc_req->req->cmd_flags & REQ_FUA) ? 0x8 : 0;
846
847                 *(unsigned int *)&blkvsc_req->cmnd[2] =
848                                 cpu_to_be32(blkvsc_req->sector_start);
849                 *(unsigned short *)&blkvsc_req->cmnd[7] =
850                                 cpu_to_be16(blkvsc_req->sector_count);
851         } else {
852                 if (rq_data_dir(blkvsc_req->req)) {
853                         blkvsc_req->write = 1;
854                         blkvsc_req->cmnd[0] = WRITE_6;
855                 } else {
856                         blkvsc_req->write = 0;
857                         blkvsc_req->cmnd[0] = READ_6;
858                 }
859
860                 *(unsigned int *)&blkvsc_req->cmnd[1] =
861                                 cpu_to_be32(blkvsc_req->sector_start) >> 8;
862                 blkvsc_req->cmnd[1] &= 0x1f;
863                 blkvsc_req->cmnd[4] = (unsigned char)blkvsc_req->sector_count;
864         }
865 }
866
867 static int blkvsc_submit_request(struct blkvsc_request *blkvsc_req,
868                         void (*request_completion)(struct hv_storvsc_request *))
869 {
870         struct block_device_context *blkdev = blkvsc_req->dev;
871         struct vm_device *device_ctx = blkdev->device_ctx;
872         struct driver_context *driver_ctx =
873                         driver_to_driver_context(device_ctx->device.driver);
874         struct blkvsc_driver_context *blkvsc_drv_ctx =
875                         (struct blkvsc_driver_context *)driver_ctx;
876         struct storvsc_driver_object *storvsc_drv_obj =
877                         &blkvsc_drv_ctx->drv_obj;
878         struct hv_storvsc_request *storvsc_req;
879         int ret;
880
881         DPRINT_DBG(BLKVSC_DRV, "blkvsc_submit_request() - "
882                    "req %p type %s start_sector %lu count %ld offset %d "
883                    "len %d\n", blkvsc_req,
884                    (blkvsc_req->write) ? "WRITE" : "READ",
885                    (unsigned long) blkvsc_req->sector_start,
886                    blkvsc_req->sector_count,
887                    blkvsc_req->request.DataBuffer.Offset,
888                    blkvsc_req->request.DataBuffer.Length);
889 #if 0
890         for (i = 0; i < (blkvsc_req->request.DataBuffer.Length >> 12); i++) {
891                 DPRINT_DBG(BLKVSC_DRV, "blkvsc_submit_request() - "
892                            "req %p pfn[%d] %llx\n",
893                            blkvsc_req, i,
894                            blkvsc_req->request.DataBuffer.PfnArray[i]);
895         }
896 #endif
897
898         storvsc_req = &blkvsc_req->request;
899         storvsc_req->Extension = (void *)((unsigned long)blkvsc_req +
900                                           sizeof(struct blkvsc_request));
901
902         storvsc_req->Type = blkvsc_req->write ? WRITE_TYPE : READ_TYPE;
903
904         storvsc_req->OnIOCompletion = request_completion;
905         storvsc_req->Context = blkvsc_req;
906
907         storvsc_req->Host = blkdev->port;
908         storvsc_req->Bus = blkdev->path;
909         storvsc_req->TargetId = blkdev->target;
910         storvsc_req->LunId = 0;  /* this is not really used at all */
911
912         storvsc_req->CdbLen = blkvsc_req->cmd_len;
913         storvsc_req->Cdb = blkvsc_req->cmnd;
914
915         storvsc_req->SenseBuffer = blkvsc_req->sense_buffer;
916         storvsc_req->SenseBufferSize = SCSI_SENSE_BUFFERSIZE;
917
918         ret = storvsc_drv_obj->OnIORequest(&blkdev->device_ctx->device_obj,
919                                            &blkvsc_req->request);
920         if (ret == 0)
921                 blkdev->num_outstanding_reqs++;
922
923         return ret;
924 }
925
926 /*
927  * We break the request into 1 or more blkvsc_requests and submit
928  * them.  If we cant submit them all, we put them on the
929  * pending_list. The blkvsc_request() will work on the pending_list.
930  */
931 static int blkvsc_do_request(struct block_device_context *blkdev,
932                              struct request *req)
933 {
934         struct bio *bio = NULL;
935         struct bio_vec *bvec = NULL;
936         struct bio_vec *prev_bvec = NULL;
937         struct blkvsc_request *blkvsc_req = NULL;
938         struct blkvsc_request *tmp;
939         int databuf_idx = 0;
940         int seg_idx = 0;
941         sector_t start_sector;
942         unsigned long num_sectors = 0;
943         int ret = 0;
944         int pending = 0;
945         struct blkvsc_request_group *group = NULL;
946
947         DPRINT_DBG(BLKVSC_DRV, "blkdev %p req %p sect %lu\n", blkdev, req,
948                   (unsigned long)blk_rq_pos(req));
949
950         /* Create a group to tie req to list of blkvsc_reqs */
951         group = kmem_cache_alloc(blkdev->request_pool, GFP_ATOMIC);
952         if (!group)
953                 return -ENOMEM;
954
955         INIT_LIST_HEAD(&group->blkvsc_req_list);
956         group->outstanding = group->status = 0;
957
958         start_sector = blk_rq_pos(req);
959
960         /* foreach bio in the request */
961         if (req->bio) {
962                 for (bio = req->bio; bio; bio = bio->bi_next) {
963                         /*
964                          * Map this bio into an existing or new storvsc request
965                          */
966                         bio_for_each_segment(bvec, bio, seg_idx) {
967                                 DPRINT_DBG(BLKVSC_DRV, "bio_for_each_segment() "
968                                            "- req %p bio %p bvec %p seg_idx %d "
969                                            "databuf_idx %d\n", req, bio, bvec,
970                                            seg_idx, databuf_idx);
971
972                                 /* Get a new storvsc request */
973                                 /* 1st-time */
974                                 if ((!blkvsc_req) ||
975                                     (databuf_idx >= MAX_MULTIPAGE_BUFFER_COUNT)
976                                     /* hole at the begin of page */
977                                     || (bvec->bv_offset != 0) ||
978                                     /* hold at the end of page */
979                                     (prev_bvec &&
980                                      (prev_bvec->bv_len != PAGE_SIZE))) {
981                                         /* submit the prev one */
982                                         if (blkvsc_req) {
983                                                 blkvsc_req->sector_start = start_sector;
984                                                 sector_div(blkvsc_req->sector_start, (blkdev->sector_size >> 9));
985
986                                                 blkvsc_req->sector_count = num_sectors / (blkdev->sector_size >> 9);
987                                                 blkvsc_init_rw(blkvsc_req);
988                                         }
989
990                                         /*
991                                          * Create new blkvsc_req to represent
992                                          * the current bvec
993                                          */
994                                         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_ATOMIC);
995                                         if (!blkvsc_req) {
996                                                 /* free up everything */
997                                                 list_for_each_entry_safe(
998                                                         blkvsc_req, tmp,
999                                                         &group->blkvsc_req_list,
1000                                                         req_entry) {
1001                                                         list_del(&blkvsc_req->req_entry);
1002                                                         kmem_cache_free(blkdev->request_pool, blkvsc_req);
1003                                                 }
1004
1005                                                 kmem_cache_free(blkdev->request_pool, group);
1006                                                 return -ENOMEM;
1007                                         }
1008
1009                                         memset(blkvsc_req, 0,
1010                                                sizeof(struct blkvsc_request));
1011
1012                                         blkvsc_req->dev = blkdev;
1013                                         blkvsc_req->req = req;
1014                                         blkvsc_req->request.DataBuffer.Offset = bvec->bv_offset;
1015                                         blkvsc_req->request.DataBuffer.Length = 0;
1016
1017                                         /* Add to the group */
1018                                         blkvsc_req->group = group;
1019                                         blkvsc_req->group->outstanding++;
1020                                         list_add_tail(&blkvsc_req->req_entry,
1021                                                 &blkvsc_req->group->blkvsc_req_list);
1022
1023                                         start_sector += num_sectors;
1024                                         num_sectors = 0;
1025                                         databuf_idx = 0;
1026                                 }
1027
1028                                 /* Add the curr bvec/segment to the curr blkvsc_req */
1029                                 blkvsc_req->request.DataBuffer.PfnArray[databuf_idx] = page_to_pfn(bvec->bv_page);
1030                                 blkvsc_req->request.DataBuffer.Length += bvec->bv_len;
1031
1032                                 prev_bvec = bvec;
1033
1034                                 databuf_idx++;
1035                                 num_sectors += bvec->bv_len >> 9;
1036
1037                         } /* bio_for_each_segment */
1038
1039                 } /* rq_for_each_bio */
1040         }
1041
1042         /* Handle the last one */
1043         if (blkvsc_req) {
1044                 DPRINT_DBG(BLKVSC_DRV, "blkdev %p req %p group %p count %d\n",
1045                            blkdev, req, blkvsc_req->group,
1046                            blkvsc_req->group->outstanding);
1047
1048                 blkvsc_req->sector_start = start_sector;
1049                 sector_div(blkvsc_req->sector_start,
1050                            (blkdev->sector_size >> 9));
1051
1052                 blkvsc_req->sector_count = num_sectors /
1053                                            (blkdev->sector_size >> 9);
1054
1055                 blkvsc_init_rw(blkvsc_req);
1056         }
1057
1058         list_for_each_entry(blkvsc_req, &group->blkvsc_req_list, req_entry) {
1059                 if (pending) {
1060                         DPRINT_DBG(BLKVSC_DRV, "adding blkvsc_req to "
1061                                    "pending_list - blkvsc_req %p start_sect %lu"
1062                                    " sect_count %ld (%lu %ld)\n", blkvsc_req,
1063                                    (unsigned long)blkvsc_req->sector_start,
1064                                    blkvsc_req->sector_count,
1065                                    (unsigned long)start_sector,
1066                                    (unsigned long)num_sectors);
1067
1068                         list_add_tail(&blkvsc_req->pend_entry,
1069                                       &blkdev->pending_list);
1070                 } else {
1071                         ret = blkvsc_submit_request(blkvsc_req,
1072                                                     blkvsc_request_completion);
1073                         if (ret == -1) {
1074                                 pending = 1;
1075                                 list_add_tail(&blkvsc_req->pend_entry,
1076                                               &blkdev->pending_list);
1077                         }
1078
1079                         DPRINT_DBG(BLKVSC_DRV, "submitted blkvsc_req %p "
1080                                    "start_sect %lu sect_count %ld (%lu %ld) "
1081                                    "ret %d\n", blkvsc_req,
1082                                    (unsigned long)blkvsc_req->sector_start,
1083                                    blkvsc_req->sector_count,
1084                                    (unsigned long)start_sector,
1085                                    num_sectors, ret);
1086                 }
1087         }
1088
1089         return pending;
1090 }
1091
1092 static void blkvsc_cmd_completion(struct hv_storvsc_request *request)
1093 {
1094         struct blkvsc_request *blkvsc_req =
1095                         (struct blkvsc_request *)request->Context;
1096         struct block_device_context *blkdev =
1097                         (struct block_device_context *)blkvsc_req->dev;
1098         struct scsi_sense_hdr sense_hdr;
1099
1100         DPRINT_DBG(BLKVSC_DRV, "blkvsc_cmd_completion() - req %p\n",
1101                    blkvsc_req);
1102
1103         blkdev->num_outstanding_reqs--;
1104
1105         if (blkvsc_req->request.Status)
1106                 if (scsi_normalize_sense(blkvsc_req->sense_buffer,
1107                                          SCSI_SENSE_BUFFERSIZE, &sense_hdr))
1108                         scsi_print_sense_hdr("blkvsc", &sense_hdr);
1109
1110         blkvsc_req->cond = 1;
1111         wake_up_interruptible(&blkvsc_req->wevent);
1112 }
1113
1114 static void blkvsc_request_completion(struct hv_storvsc_request *request)
1115 {
1116         struct blkvsc_request *blkvsc_req =
1117                         (struct blkvsc_request *)request->Context;
1118         struct block_device_context *blkdev =
1119                         (struct block_device_context *)blkvsc_req->dev;
1120         unsigned long flags;
1121         struct blkvsc_request *comp_req, *tmp;
1122
1123         /* ASSERT(blkvsc_req->group); */
1124
1125         DPRINT_DBG(BLKVSC_DRV, "blkdev %p blkvsc_req %p group %p type %s "
1126                    "sect_start %lu sect_count %ld len %d group outstd %d "
1127                    "total outstd %d\n",
1128                    blkdev, blkvsc_req, blkvsc_req->group,
1129                    (blkvsc_req->write) ? "WRITE" : "READ",
1130                    (unsigned long)blkvsc_req->sector_start,
1131                    blkvsc_req->sector_count,
1132                    blkvsc_req->request.DataBuffer.Length,
1133                    blkvsc_req->group->outstanding,
1134                    blkdev->num_outstanding_reqs);
1135
1136         spin_lock_irqsave(&blkdev->lock, flags);
1137
1138         blkdev->num_outstanding_reqs--;
1139         blkvsc_req->group->outstanding--;
1140
1141         /*
1142          * Only start processing when all the blkvsc_reqs are
1143          * completed. This guarantees no out-of-order blkvsc_req
1144          * completion when calling end_that_request_first()
1145          */
1146         if (blkvsc_req->group->outstanding == 0) {
1147                 list_for_each_entry_safe(comp_req, tmp,
1148                                          &blkvsc_req->group->blkvsc_req_list,
1149                                          req_entry) {
1150                         DPRINT_DBG(BLKVSC_DRV, "completing blkvsc_req %p "
1151                                    "sect_start %lu sect_count %ld\n",
1152                                    comp_req,
1153                                    (unsigned long)comp_req->sector_start,
1154                                    comp_req->sector_count);
1155
1156                         list_del(&comp_req->req_entry);
1157
1158                         if (!__blk_end_request(comp_req->req,
1159                                 (!comp_req->request.Status ? 0 : -EIO),
1160                                 comp_req->sector_count * blkdev->sector_size)) {
1161                                 /*
1162                                  * All the sectors have been xferred ie the
1163                                  * request is done
1164                                  */
1165                                 DPRINT_DBG(BLKVSC_DRV, "req %p COMPLETED\n",
1166                                            comp_req->req);
1167                                 kmem_cache_free(blkdev->request_pool,
1168                                                 comp_req->group);
1169                         }
1170
1171                         kmem_cache_free(blkdev->request_pool, comp_req);
1172                 }
1173
1174                 if (!blkdev->shutting_down) {
1175                         blkvsc_do_pending_reqs(blkdev);
1176                         blk_start_queue(blkdev->gd->queue);
1177                         blkvsc_request(blkdev->gd->queue);
1178                 }
1179         }
1180
1181         spin_unlock_irqrestore(&blkdev->lock, flags);
1182 }
1183
1184 static int blkvsc_cancel_pending_reqs(struct block_device_context *blkdev)
1185 {
1186         struct blkvsc_request *pend_req, *tmp;
1187         struct blkvsc_request *comp_req, *tmp2;
1188
1189         int ret = 0;
1190
1191         DPRINT_DBG(BLKVSC_DRV, "blkvsc_cancel_pending_reqs()");
1192
1193         /* Flush the pending list first */
1194         list_for_each_entry_safe(pend_req, tmp, &blkdev->pending_list,
1195                                  pend_entry) {
1196                 /*
1197                  * The pend_req could be part of a partially completed
1198                  * request. If so, complete those req first until we
1199                  * hit the pend_req
1200                  */
1201                 list_for_each_entry_safe(comp_req, tmp2,
1202                                          &pend_req->group->blkvsc_req_list,
1203                                          req_entry) {
1204                         DPRINT_DBG(BLKVSC_DRV, "completing blkvsc_req %p "
1205                                    "sect_start %lu sect_count %ld\n",
1206                                    comp_req,
1207                                    (unsigned long) comp_req->sector_start,
1208                                    comp_req->sector_count);
1209
1210                         if (comp_req == pend_req)
1211                                 break;
1212
1213                         list_del(&comp_req->req_entry);
1214
1215                         if (comp_req->req) {
1216                                 ret = __blk_end_request(comp_req->req,
1217                                         (!comp_req->request.Status ? 0 : -EIO),
1218                                         comp_req->sector_count *
1219                                         blkdev->sector_size);
1220
1221                                 /* FIXME: shouldn't this do more than return? */
1222                                 if (ret)
1223                                         goto out;
1224                         }
1225
1226                         kmem_cache_free(blkdev->request_pool, comp_req);
1227                 }
1228
1229                 DPRINT_DBG(BLKVSC_DRV, "cancelling pending request - %p\n",
1230                            pend_req);
1231
1232                 list_del(&pend_req->pend_entry);
1233
1234                 list_del(&pend_req->req_entry);
1235
1236                 if (comp_req->req) {
1237                         if (!__blk_end_request(pend_req->req, -EIO,
1238                                                pend_req->sector_count *
1239                                                blkdev->sector_size)) {
1240                                 /*
1241                                  * All the sectors have been xferred ie the
1242                                  * request is done
1243                                  */
1244                                 DPRINT_DBG(BLKVSC_DRV,
1245                                            "blkvsc_cancel_pending_reqs() - "
1246                                            "req %p COMPLETED\n", pend_req->req);
1247                                 kmem_cache_free(blkdev->request_pool,
1248                                                 pend_req->group);
1249                         }
1250                 }
1251
1252                 kmem_cache_free(blkdev->request_pool, pend_req);
1253         }
1254
1255 out:
1256         return ret;
1257 }
1258
1259 static int blkvsc_do_pending_reqs(struct block_device_context *blkdev)
1260 {
1261         struct blkvsc_request *pend_req, *tmp;
1262         int ret = 0;
1263
1264         /* Flush the pending list first */
1265         list_for_each_entry_safe(pend_req, tmp, &blkdev->pending_list,
1266                                  pend_entry) {
1267                 DPRINT_DBG(BLKVSC_DRV, "working off pending_list - %p\n",
1268                            pend_req);
1269
1270                 ret = blkvsc_submit_request(pend_req,
1271                                             blkvsc_request_completion);
1272                 if (ret != 0)
1273                         break;
1274                 else
1275                         list_del(&pend_req->pend_entry);
1276         }
1277
1278         return ret;
1279 }
1280
1281 static void blkvsc_request(struct request_queue *queue)
1282 {
1283         struct block_device_context *blkdev = NULL;
1284         struct request *req;
1285         int ret = 0;
1286
1287         DPRINT_DBG(BLKVSC_DRV, "- enter\n");
1288         while ((req = blk_peek_request(queue)) != NULL) {
1289                 DPRINT_DBG(BLKVSC_DRV, "- req %p\n", req);
1290
1291                 blkdev = req->rq_disk->private_data;
1292                 if (blkdev->shutting_down || req->cmd_type != REQ_TYPE_FS ||
1293                     blkdev->media_not_present) {
1294                         __blk_end_request_cur(req, 0);
1295                         continue;
1296                 }
1297
1298                 ret = blkvsc_do_pending_reqs(blkdev);
1299
1300                 if (ret != 0) {
1301                         DPRINT_DBG(BLKVSC_DRV,
1302                                    "- stop queue - pending_list not empty\n");
1303                         blk_stop_queue(queue);
1304                         break;
1305                 }
1306
1307                 blk_start_request(req);
1308
1309                 ret = blkvsc_do_request(blkdev, req);
1310                 if (ret > 0) {
1311                         DPRINT_DBG(BLKVSC_DRV, "- stop queue - no room\n");
1312                         blk_stop_queue(queue);
1313                         break;
1314                 } else if (ret < 0) {
1315                         DPRINT_DBG(BLKVSC_DRV, "- stop queue - no mem\n");
1316                         blk_requeue_request(queue, req);
1317                         blk_stop_queue(queue);
1318                         break;
1319                 }
1320         }
1321 }
1322
1323 static int blkvsc_open(struct block_device *bdev, fmode_t mode)
1324 {
1325         struct block_device_context *blkdev = bdev->bd_disk->private_data;
1326
1327         DPRINT_DBG(BLKVSC_DRV, "- users %d disk %s\n", blkdev->users,
1328                    blkdev->gd->disk_name);
1329
1330         lock_kernel();
1331         spin_lock(&blkdev->lock);
1332
1333         if (!blkdev->users && blkdev->device_type == DVD_TYPE) {
1334                 spin_unlock(&blkdev->lock);
1335                 check_disk_change(bdev);
1336                 spin_lock(&blkdev->lock);
1337         }
1338
1339         blkdev->users++;
1340
1341         spin_unlock(&blkdev->lock);
1342         unlock_kernel();
1343         return 0;
1344 }
1345
1346 static int blkvsc_release(struct gendisk *disk, fmode_t mode)
1347 {
1348         struct block_device_context *blkdev = disk->private_data;
1349
1350         DPRINT_DBG(BLKVSC_DRV, "- users %d disk %s\n", blkdev->users,
1351                    blkdev->gd->disk_name);
1352
1353         lock_kernel();
1354         spin_lock(&blkdev->lock);
1355         if (blkdev->users == 1) {
1356                 spin_unlock(&blkdev->lock);
1357                 blkvsc_do_flush(blkdev);
1358                 spin_lock(&blkdev->lock);
1359         }
1360
1361         blkdev->users--;
1362
1363         spin_unlock(&blkdev->lock);
1364         unlock_kernel();
1365         return 0;
1366 }
1367
1368 static int blkvsc_media_changed(struct gendisk *gd)
1369 {
1370         DPRINT_DBG(BLKVSC_DRV, "- enter\n");
1371         return 1;
1372 }
1373
1374 static int blkvsc_revalidate_disk(struct gendisk *gd)
1375 {
1376         struct block_device_context *blkdev = gd->private_data;
1377
1378         DPRINT_DBG(BLKVSC_DRV, "- enter\n");
1379
1380         if (blkdev->device_type == DVD_TYPE) {
1381                 blkvsc_do_read_capacity(blkdev);
1382                 set_capacity(blkdev->gd, blkdev->capacity *
1383                             (blkdev->sector_size/512));
1384                 blk_queue_logical_block_size(gd->queue, blkdev->sector_size);
1385         }
1386         return 0;
1387 }
1388
1389 static int blkvsc_getgeo(struct block_device *bd, struct hd_geometry *hg)
1390 {
1391         sector_t total_sectors = get_capacity(bd->bd_disk);
1392         sector_t cylinder_times_heads = 0;
1393         sector_t temp = 0;
1394
1395         int sectors_per_track = 0;
1396         int heads = 0;
1397         int cylinders = 0;
1398         int rem = 0;
1399
1400         if (total_sectors > (65535 * 16 * 255))
1401                 total_sectors = (65535 * 16 * 255);
1402
1403         if (total_sectors >= (65535 * 16 * 63)) {
1404                 sectors_per_track = 255;
1405                 heads = 16;
1406
1407                 cylinder_times_heads = total_sectors;
1408                 /* sector_div stores the quotient in cylinder_times_heads */
1409                 rem = sector_div(cylinder_times_heads, sectors_per_track);
1410         } else {
1411                 sectors_per_track = 17;
1412
1413                 cylinder_times_heads = total_sectors;
1414                 /* sector_div stores the quotient in cylinder_times_heads */
1415                 rem = sector_div(cylinder_times_heads, sectors_per_track);
1416
1417                 temp = cylinder_times_heads + 1023;
1418                 /* sector_div stores the quotient in temp */
1419                 rem = sector_div(temp, 1024);
1420
1421                 heads = temp;
1422
1423                 if (heads < 4)
1424                         heads = 4;
1425
1426
1427                 if (cylinder_times_heads >= (heads * 1024) || (heads > 16)) {
1428                         sectors_per_track = 31;
1429                         heads = 16;
1430
1431                         cylinder_times_heads = total_sectors;
1432                         /*
1433                          * sector_div stores the quotient in
1434                          * cylinder_times_heads
1435                          */
1436                         rem = sector_div(cylinder_times_heads,
1437                                          sectors_per_track);
1438                 }
1439
1440                 if (cylinder_times_heads >= (heads * 1024)) {
1441                         sectors_per_track = 63;
1442                         heads = 16;
1443
1444                         cylinder_times_heads = total_sectors;
1445                         /*
1446                          * sector_div stores the quotient in
1447                          * cylinder_times_heads
1448                          */
1449                         rem = sector_div(cylinder_times_heads,
1450                                          sectors_per_track);
1451                 }
1452         }
1453
1454         temp = cylinder_times_heads;
1455         /* sector_div stores the quotient in temp */
1456         rem = sector_div(temp, heads);
1457         cylinders = temp;
1458
1459         hg->heads = heads;
1460         hg->sectors = sectors_per_track;
1461         hg->cylinders = cylinders;
1462
1463         DPRINT_INFO(BLKVSC_DRV, "CHS (%d, %d, %d)", cylinders, heads,
1464                     sectors_per_track);
1465
1466     return 0;
1467 }
1468
1469 static int blkvsc_ioctl(struct block_device *bd, fmode_t mode,
1470                         unsigned cmd, unsigned long argument)
1471 {
1472 /*      struct block_device_context *blkdev = bd->bd_disk->private_data; */
1473         int ret;
1474
1475         switch (cmd) {
1476         /*
1477          * TODO: I think there is certain format for HDIO_GET_IDENTITY rather
1478          * than just a GUID. Commented it out for now.
1479          */
1480 #if 0
1481         case HDIO_GET_IDENTITY:
1482                 DPRINT_INFO(BLKVSC_DRV, "HDIO_GET_IDENTITY\n");
1483                 if (copy_to_user((void __user *)arg, blkdev->device_id,
1484                                  blkdev->device_id_len))
1485                         ret = -EFAULT;
1486                 break;
1487 #endif
1488         default:
1489                 ret = -EINVAL;
1490                 break;
1491         }
1492
1493         return ret;
1494 }
1495
1496 static int __init blkvsc_init(void)
1497 {
1498         int ret;
1499
1500         BUILD_BUG_ON(sizeof(sector_t) != 8);
1501
1502         DPRINT_ENTER(BLKVSC_DRV);
1503
1504         DPRINT_INFO(BLKVSC_DRV, "Blkvsc initializing....");
1505
1506         ret = blkvsc_drv_init(BlkVscInitialize);
1507
1508         DPRINT_EXIT(BLKVSC_DRV);
1509
1510         return ret;
1511 }
1512
1513 static void __exit blkvsc_exit(void)
1514 {
1515         DPRINT_ENTER(BLKVSC_DRV);
1516         blkvsc_drv_exit();
1517         DPRINT_ENTER(BLKVSC_DRV);
1518 }
1519
1520 MODULE_LICENSE("GPL");
1521 MODULE_VERSION(HV_DRV_VERSION);
1522 MODULE_DESCRIPTION("Microsoft Hyper-V virtual block driver");
1523 module_init(blkvsc_init);
1524 module_exit(blkvsc_exit);