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