SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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 "version_info.h"
  36 #include "vmbus.h"
  37 #include "storvsc_api.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 static int blkvsc_ringbuffer_size = BLKVSC_RING_BUFFER_SIZE;
 153 module_param(blkvsc_ringbuffer_size, int, S_IRUGO);
 154 MODULE_PARM_DESC(ring_size, "Ring buffer size (in bytes)");
 155
 156 /* The one and only one */
 157 static struct blkvsc_driver_context g_blkvsc_drv;
 158
 159 static const struct block_device_operations block_ops = {
 160         .owner = THIS_MODULE,
 161         .open = blkvsc_open,
 162         .release = blkvsc_release,
 163         .media_changed = blkvsc_media_changed,
 164         .revalidate_disk = blkvsc_revalidate_disk,
 165         .getgeo = blkvsc_getgeo,
 166         .ioctl  = blkvsc_ioctl,
 167 };
 168
 169 /*
 170  * blkvsc_drv_init -  BlkVsc driver initialization.
 171  */
 172 static int blkvsc_drv_init(int (*drv_init)(struct hv_driver *drv))
 173 {
 174         struct storvsc_driver_object *storvsc_drv_obj = &g_blkvsc_drv.drv_obj;
 175         struct driver_context *drv_ctx = &g_blkvsc_drv.drv_ctx;
 176         int ret;
 177
 178         DPRINT_ENTER(BLKVSC_DRV);
 179
 180         vmbus_get_interface(&storvsc_drv_obj->Base.VmbusChannelInterface);
 181
 182         storvsc_drv_obj->RingBufferSize = blkvsc_ringbuffer_size;
 183
 184         /* Callback to client driver to complete the initialization */
 185         drv_init(&storvsc_drv_obj->Base);
 186
 187         drv_ctx->driver.name = storvsc_drv_obj->Base.name;
 188         memcpy(&drv_ctx->class_id, &storvsc_drv_obj->Base.deviceType,
 189                sizeof(struct hv_guid));
 190
 191         drv_ctx->probe = blkvsc_probe;
 192         drv_ctx->remove = blkvsc_remove;
 193         drv_ctx->shutdown = blkvsc_shutdown;
 194
 195         /* The driver belongs to vmbus */
 196         ret = vmbus_child_driver_register(drv_ctx);
 197
 198         DPRINT_EXIT(BLKVSC_DRV);
 199
 200         return ret;
 201 }
 202
 203 static int blkvsc_drv_exit_cb(struct device *dev, void *data)
 204 {
 205         struct device **curr = (struct device **)data;
 206         *curr = dev;
 207         return 1; /* stop iterating */
 208 }
 209
 210 static void blkvsc_drv_exit(void)
 211 {
 212         struct storvsc_driver_object *storvsc_drv_obj = &g_blkvsc_drv.drv_obj;
 213         struct driver_context *drv_ctx = &g_blkvsc_drv.drv_ctx;
 214         struct device *current_dev;
 215         int ret;
 216
 217         DPRINT_ENTER(BLKVSC_DRV);
 218
 219         while (1) {
 220                 current_dev = NULL;
 221
 222                 /* Get the device */
 223                 ret = driver_for_each_device(&drv_ctx->driver, NULL,
 224                                              (void *) &current_dev,
 225                                              blkvsc_drv_exit_cb);
 226
 227                 if (ret)
 228                         DPRINT_WARN(BLKVSC_DRV,
 229                                     "driver_for_each_device returned %d", ret);
 230
 231
 232                 if (current_dev == NULL)
 233                         break;
 234
 235                 /* Initiate removal from the top-down */
 236                 device_unregister(current_dev);
 237         }
 238
 239         if (storvsc_drv_obj->Base.OnCleanup)
 240                 storvsc_drv_obj->Base.OnCleanup(&storvsc_drv_obj->Base);
 241
 242         vmbus_child_driver_unregister(drv_ctx);
 243
 244         DPRINT_EXIT(BLKVSC_DRV);
 245
 246         return;
 247 }
 248
 249 /*
 250  * blkvsc_probe - Add a new device for this driver
 251  */
 252 static int blkvsc_probe(struct device *device)
 253 {
 254         struct driver_context *driver_ctx =
 255                                 driver_to_driver_context(device->driver);
 256         struct blkvsc_driver_context *blkvsc_drv_ctx =
 257                                 (struct blkvsc_driver_context *)driver_ctx;
 258         struct storvsc_driver_object *storvsc_drv_obj =
 259                                 &blkvsc_drv_ctx->drv_obj;
 260         struct vm_device *device_ctx = device_to_vm_device(device);
 261         struct hv_device *device_obj = &device_ctx->device_obj;
 262
 263         struct block_device_context *blkdev = NULL;
 264         struct storvsc_device_info device_info;
 265         int major = 0;
 266         int devnum = 0;
 267         int ret = 0;
 268         static int ide0_registered;
 269         static int ide1_registered;
 270
 271         DPRINT_ENTER(BLKVSC_DRV);
 272
 273         DPRINT_DBG(BLKVSC_DRV, "blkvsc_probe - enter");
 274
 275         if (!storvsc_drv_obj->Base.OnDeviceAdd) {
 276                 DPRINT_ERR(BLKVSC_DRV, "OnDeviceAdd() not set");
 277                 ret = -1;
 278                 goto Cleanup;
 279         }
 280
 281         blkdev = kzalloc(sizeof(struct block_device_context), GFP_KERNEL);
 282         if (!blkdev) {
 283                 ret = -ENOMEM;
 284                 goto Cleanup;
 285         }
 286
 287         INIT_LIST_HEAD(&blkdev->pending_list);
 288
 289         /* Initialize what we can here */
 290         spin_lock_init(&blkdev->lock);
 291
 292         /* ASSERT(sizeof(struct blkvsc_request_group) <= */
 293         /*      sizeof(struct blkvsc_request)); */
 294
 295         blkdev->request_pool = kmem_cache_create(dev_name(&device_ctx->device),
 296                                         sizeof(struct blkvsc_request) +
 297                                         storvsc_drv_obj->RequestExtSize, 0,
 298                                         SLAB_HWCACHE_ALIGN, NULL);
 299         if (!blkdev->request_pool) {
 300                 ret = -ENOMEM;
 301                 goto Cleanup;
 302         }
 303
 304
 305         /* Call to the vsc driver to add the device */
 306         ret = storvsc_drv_obj->Base.OnDeviceAdd(device_obj, &device_info);
 307         if (ret != 0) {
 308                 DPRINT_ERR(BLKVSC_DRV, "unable to add blkvsc device");
 309                 goto Cleanup;
 310         }
 311
 312         blkdev->device_ctx = device_ctx;
 313         /* this identified the device 0 or 1 */
 314         blkdev->target = device_info.TargetId;
 315         /* this identified the ide ctrl 0 or 1 */
 316         blkdev->path = device_info.PathId;
 317
 318         dev_set_drvdata(device, blkdev);
 319
 320         /* Calculate the major and device num */
 321         if (blkdev->path == 0) {
 322                 major = IDE0_MAJOR;
 323                 devnum = blkdev->path + blkdev->target;         /* 0 or 1 */
 324
 325                 if (!ide0_registered) {
 326                         ret = register_blkdev(major, "ide");
 327                         if (ret != 0) {
 328                                 DPRINT_ERR(BLKVSC_DRV,
 329                                            "register_blkdev() failed! ret %d",
 330                                            ret);
 331                                 goto Remove;
 332                         }
 333
 334                         ide0_registered = 1;
 335                 }
 336         } else if (blkdev->path == 1) {
 337                 major = IDE1_MAJOR;
 338                 devnum = blkdev->path + blkdev->target + 1; /* 2 or 3 */
 339
 340                 if (!ide1_registered) {
 341                         ret = register_blkdev(major, "ide");
 342                         if (ret != 0) {
 343                                 DPRINT_ERR(BLKVSC_DRV,
 344                                            "register_blkdev() failed! ret %d",
 345                                            ret);
 346                                 goto Remove;
 347                         }
 348
 349                         ide1_registered = 1;
 350                 }
 351         } else {
 352                 DPRINT_ERR(BLKVSC_DRV, "invalid pathid");
 353                 ret = -1;
 354                 goto Cleanup;
 355         }
 356
 357         DPRINT_INFO(BLKVSC_DRV, "blkvsc registered for major %d!!", major);
 358
 359         blkdev->gd = alloc_disk(BLKVSC_MINORS);
 360         if (!blkdev->gd) {
 361                 DPRINT_ERR(BLKVSC_DRV, "register_blkdev() failed! ret %d", ret);
 362                 ret = -1;
 363                 goto Cleanup;
 364         }
 365
 366         blkdev->gd->queue = blk_init_queue(blkvsc_request, &blkdev->lock);
 367
 368         blk_queue_max_segment_size(blkdev->gd->queue, PAGE_SIZE);
 369         blk_queue_max_segments(blkdev->gd->queue, MAX_MULTIPAGE_BUFFER_COUNT);
 370         blk_queue_segment_boundary(blkdev->gd->queue, PAGE_SIZE-1);
 371         blk_queue_bounce_limit(blkdev->gd->queue, BLK_BOUNCE_ANY);
 372         blk_queue_dma_alignment(blkdev->gd->queue, 511);
 373
 374         blkdev->gd->major = major;
 375         if (devnum == 1 || devnum == 3)
 376                 blkdev->gd->first_minor = BLKVSC_MINORS;
 377         else
 378                 blkdev->gd->first_minor = 0;
 379         blkdev->gd->fops = &block_ops;
 380         blkdev->gd->private_data = blkdev;
 381         sprintf(blkdev->gd->disk_name, "hd%c", 'a' + devnum);
 382
 383         blkvsc_do_inquiry(blkdev);
 384         if (blkdev->device_type == DVD_TYPE) {
 385                 set_disk_ro(blkdev->gd, 1);
 386                 blkdev->gd->flags |= GENHD_FL_REMOVABLE;
 387                 blkvsc_do_read_capacity(blkdev);
 388         } else {
 389                 blkvsc_do_read_capacity16(blkdev);
 390         }
 391
 392         set_capacity(blkdev->gd, blkdev->capacity * (blkdev->sector_size/512));
 393         blk_queue_logical_block_size(blkdev->gd->queue, blkdev->sector_size);
 394         /* go! */
 395         add_disk(blkdev->gd);
 396
 397         DPRINT_INFO(BLKVSC_DRV, "%s added!! capacity %lu sector_size %d",
 398                     blkdev->gd->disk_name, (unsigned long)blkdev->capacity,
 399                     blkdev->sector_size);
 400
 401         return ret;
 402
 403 Remove:
 404         storvsc_drv_obj->Base.OnDeviceRemove(device_obj);
 405
 406 Cleanup:
 407         if (blkdev) {
 408                 if (blkdev->request_pool) {
 409                         kmem_cache_destroy(blkdev->request_pool);
 410                         blkdev->request_pool = NULL;
 411                 }
 412                 kfree(blkdev);
 413                 blkdev = NULL;
 414         }
 415
 416         DPRINT_EXIT(BLKVSC_DRV);
 417
 418         return ret;
 419 }
 420
 421 static void blkvsc_shutdown(struct device *device)
 422 {
 423         struct block_device_context *blkdev = dev_get_drvdata(device);
 424         unsigned long flags;
 425
 426         if (!blkdev)
 427                 return;
 428
 429         DPRINT_DBG(BLKVSC_DRV, "blkvsc_shutdown - users %d disk %s\n",
 430                    blkdev->users, blkdev->gd->disk_name);
 431
 432         spin_lock_irqsave(&blkdev->lock, flags);
 433
 434         blkdev->shutting_down = 1;
 435
 436         blk_stop_queue(blkdev->gd->queue);
 437
 438         spin_unlock_irqrestore(&blkdev->lock, flags);
 439
 440         while (blkdev->num_outstanding_reqs) {
 441                 DPRINT_INFO(STORVSC, "waiting for %d requests to complete...",
 442                             blkdev->num_outstanding_reqs);
 443                 udelay(100);
 444         }
 445
 446         blkvsc_do_flush(blkdev);
 447
 448         spin_lock_irqsave(&blkdev->lock, flags);
 449
 450         blkvsc_cancel_pending_reqs(blkdev);
 451
 452         spin_unlock_irqrestore(&blkdev->lock, flags);
 453 }
 454
 455 static int blkvsc_do_flush(struct block_device_context *blkdev)
 456 {
 457         struct blkvsc_request *blkvsc_req;
 458
 459         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_flush()\n");
 460
 461         if (blkdev->device_type != HARDDISK_TYPE)
 462                 return 0;
 463
 464         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
 465         if (!blkvsc_req)
 466                 return -ENOMEM;
 467
 468         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
 469         init_waitqueue_head(&blkvsc_req->wevent);
 470         blkvsc_req->dev = blkdev;
 471         blkvsc_req->req = NULL;
 472         blkvsc_req->write = 0;
 473
 474         blkvsc_req->request.DataBuffer.PfnArray[0] = 0;
 475         blkvsc_req->request.DataBuffer.Offset = 0;
 476         blkvsc_req->request.DataBuffer.Length = 0;
 477
 478         blkvsc_req->cmnd[0] = SYNCHRONIZE_CACHE;
 479         blkvsc_req->cmd_len = 10;
 480
 481         /*
 482          * Set this here since the completion routine may be invoked and
 483          * completed before we return
 484          */
 485         blkvsc_req->cond = 0;
 486         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
 487
 488         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
 489
 490         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 491
 492         return 0;
 493 }
 494
 495 /* Do a scsi INQUIRY cmd here to get the device type (ie disk or dvd) */
 496 static int blkvsc_do_inquiry(struct block_device_context *blkdev)
 497 {
 498         struct blkvsc_request *blkvsc_req;
 499         struct page *page_buf;
 500         unsigned char *buf;
 501         unsigned char device_type;
 502
 503         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_inquiry()\n");
 504
 505         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
 506         if (!blkvsc_req)
 507                 return -ENOMEM;
 508
 509         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
 510         page_buf = alloc_page(GFP_KERNEL);
 511         if (!page_buf) {
 512                 kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 513                 return -ENOMEM;
 514         }
 515
 516         init_waitqueue_head(&blkvsc_req->wevent);
 517         blkvsc_req->dev = blkdev;
 518         blkvsc_req->req = NULL;
 519         blkvsc_req->write = 0;
 520
 521         blkvsc_req->request.DataBuffer.PfnArray[0] = page_to_pfn(page_buf);
 522         blkvsc_req->request.DataBuffer.Offset = 0;
 523         blkvsc_req->request.DataBuffer.Length = 64;
 524
 525         blkvsc_req->cmnd[0] = INQUIRY;
 526         blkvsc_req->cmnd[1] = 0x1;              /* Get product data */
 527         blkvsc_req->cmnd[2] = 0x83;             /* mode page 83 */
 528         blkvsc_req->cmnd[4] = 64;
 529         blkvsc_req->cmd_len = 6;
 530
 531         /*
 532          * Set this here since the completion routine may be invoked and
 533          * completed before we return
 534          */
 535         blkvsc_req->cond = 0;
 536
 537         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
 538
 539         DPRINT_DBG(BLKVSC_DRV, "waiting %p to complete - cond %d\n",
 540                    blkvsc_req, blkvsc_req->cond);
 541
 542         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
 543
 544         buf = kmap(page_buf);
 545
 546         /* print_hex_dump_bytes("", DUMP_PREFIX_NONE, buf, 64); */
 547         /* be to le */
 548         device_type = buf[0] & 0x1F;
 549
 550         if (device_type == 0x0) {
 551                 blkdev->device_type = HARDDISK_TYPE;
 552         } else if (device_type == 0x5) {
 553                 blkdev->device_type = DVD_TYPE;
 554         } else {
 555                 /* TODO: this is currently unsupported device type */
 556                 blkdev->device_type = UNKNOWN_DEV_TYPE;
 557         }
 558
 559         DPRINT_DBG(BLKVSC_DRV, "device type %d\n", device_type);
 560
 561         blkdev->device_id_len = buf[7];
 562         if (blkdev->device_id_len > 64)
 563                 blkdev->device_id_len = 64;
 564
 565         memcpy(blkdev->device_id, &buf[8], blkdev->device_id_len);
 566         /* printk_hex_dump_bytes("", DUMP_PREFIX_NONE, blkdev->device_id,
 567          * blkdev->device_id_len); */
 568
 569         kunmap(page_buf);
 570
 571         __free_page(page_buf);
 572
 573         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 574
 575         return 0;
 576 }
 577
 578 /* Do a scsi READ_CAPACITY cmd here to get the size of the disk */
 579 static int blkvsc_do_read_capacity(struct block_device_context *blkdev)
 580 {
 581         struct blkvsc_request *blkvsc_req;
 582         struct page *page_buf;
 583         unsigned char *buf;
 584         struct scsi_sense_hdr sense_hdr;
 585
 586         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_read_capacity()\n");
 587
 588         blkdev->sector_size = 0;
 589         blkdev->capacity = 0;
 590         blkdev->media_not_present = 0; /* assume a disk is present */
 591
 592         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
 593         if (!blkvsc_req)
 594                 return -ENOMEM;
 595
 596         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
 597         page_buf = alloc_page(GFP_KERNEL);
 598         if (!page_buf) {
 599                 kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 600                 return -ENOMEM;
 601         }
 602
 603         init_waitqueue_head(&blkvsc_req->wevent);
 604         blkvsc_req->dev = blkdev;
 605         blkvsc_req->req = NULL;
 606         blkvsc_req->write = 0;
 607
 608         blkvsc_req->request.DataBuffer.PfnArray[0] = page_to_pfn(page_buf);
 609         blkvsc_req->request.DataBuffer.Offset = 0;
 610         blkvsc_req->request.DataBuffer.Length = 8;
 611
 612         blkvsc_req->cmnd[0] = READ_CAPACITY;
 613         blkvsc_req->cmd_len = 16;
 614
 615         /*
 616          * Set this here since the completion routine may be invoked
 617          * and completed before we return
 618          */
 619         blkvsc_req->cond = 0;
 620
 621         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
 622
 623         DPRINT_DBG(BLKVSC_DRV, "waiting %p to complete - cond %d\n",
 624                    blkvsc_req, blkvsc_req->cond);
 625
 626         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
 627
 628         /* check error */
 629         if (blkvsc_req->request.Status) {
 630                 scsi_normalize_sense(blkvsc_req->sense_buffer,
 631                                      SCSI_SENSE_BUFFERSIZE, &sense_hdr);
 632
 633                 if (sense_hdr.asc == 0x3A) {
 634                         /* Medium not present */
 635                         blkdev->media_not_present = 1;
 636                 }
 637                 return 0;
 638         }
 639         buf = kmap(page_buf);
 640
 641         /* be to le */
 642         blkdev->capacity = ((buf[0] << 24) | (buf[1] << 16) |
 643                             (buf[2] << 8) | buf[3]) + 1;
 644         blkdev->sector_size = (buf[4] << 24) | (buf[5] << 16) |
 645                               (buf[6] << 8) | buf[7];
 646
 647         kunmap(page_buf);
 648
 649         __free_page(page_buf);
 650
 651         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 652
 653         return 0;
 654 }
 655
 656 static int blkvsc_do_read_capacity16(struct block_device_context *blkdev)
 657 {
 658         struct blkvsc_request *blkvsc_req;
 659         struct page *page_buf;
 660         unsigned char *buf;
 661         struct scsi_sense_hdr sense_hdr;
 662
 663         DPRINT_DBG(BLKVSC_DRV, "blkvsc_do_read_capacity16()\n");
 664
 665         blkdev->sector_size = 0;
 666         blkdev->capacity = 0;
 667         blkdev->media_not_present = 0; /* assume a disk is present */
 668
 669         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_KERNEL);
 670         if (!blkvsc_req)
 671                 return -ENOMEM;
 672
 673         memset(blkvsc_req, 0, sizeof(struct blkvsc_request));
 674         page_buf = alloc_page(GFP_KERNEL);
 675         if (!page_buf) {
 676                 kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 677                 return -ENOMEM;
 678         }
 679
 680         init_waitqueue_head(&blkvsc_req->wevent);
 681         blkvsc_req->dev = blkdev;
 682         blkvsc_req->req = NULL;
 683         blkvsc_req->write = 0;
 684
 685         blkvsc_req->request.DataBuffer.PfnArray[0] = page_to_pfn(page_buf);
 686         blkvsc_req->request.DataBuffer.Offset = 0;
 687         blkvsc_req->request.DataBuffer.Length = 12;
 688
 689         blkvsc_req->cmnd[0] = 0x9E; /* READ_CAPACITY16; */
 690         blkvsc_req->cmd_len = 16;
 691
 692         /*
 693          * Set this here since the completion routine may be invoked
 694          * and completed before we return
 695          */
 696         blkvsc_req->cond = 0;
 697
 698         blkvsc_submit_request(blkvsc_req, blkvsc_cmd_completion);
 699
 700         DPRINT_DBG(BLKVSC_DRV, "waiting %p to complete - cond %d\n",
 701                    blkvsc_req, blkvsc_req->cond);
 702
 703         wait_event_interruptible(blkvsc_req->wevent, blkvsc_req->cond);
 704
 705         /* check error */
 706         if (blkvsc_req->request.Status) {
 707                 scsi_normalize_sense(blkvsc_req->sense_buffer,
 708                                      SCSI_SENSE_BUFFERSIZE, &sense_hdr);
 709                 if (sense_hdr.asc == 0x3A) {
 710                         /* Medium not present */
 711                         blkdev->media_not_present = 1;
 712                 }
 713                 return 0;
 714         }
 715         buf = kmap(page_buf);
 716
 717         /* be to le */
 718         blkdev->capacity = be64_to_cpu(*(unsigned long long *) &buf[0]) + 1;
 719         blkdev->sector_size = be32_to_cpu(*(unsigned int *)&buf[8]);
 720
 721 #if 0
 722         blkdev->capacity = ((buf[0] << 24) | (buf[1] << 16) |
 723                             (buf[2] << 8) | buf[3]) + 1;
 724         blkdev->sector_size = (buf[4] << 24) | (buf[5] << 16) |
 725                               (buf[6] << 8) | buf[7];
 726 #endif
 727
 728         kunmap(page_buf);
 729
 730         __free_page(page_buf);
 731
 732         kmem_cache_free(blkvsc_req->dev->request_pool, blkvsc_req);
 733
 734         return 0;
 735 }
 736
 737 /*
 738  * blkvsc_remove() - Callback when our device is removed
 739  */
 740 static int blkvsc_remove(struct device *device)
 741 {
 742         struct driver_context *driver_ctx =
 743                                 driver_to_driver_context(device->driver);
 744         struct blkvsc_driver_context *blkvsc_drv_ctx =
 745                                 (struct blkvsc_driver_context *)driver_ctx;
 746         struct storvsc_driver_object *storvsc_drv_obj =
 747                                 &blkvsc_drv_ctx->drv_obj;
 748         struct vm_device *device_ctx = device_to_vm_device(device);
 749         struct hv_device *device_obj = &device_ctx->device_obj;
 750         struct block_device_context *blkdev = dev_get_drvdata(device);
 751         unsigned long flags;
 752         int ret;
 753
 754         DPRINT_ENTER(BLKVSC_DRV);
 755
 756         DPRINT_DBG(BLKVSC_DRV, "blkvsc_remove()\n");
 757
 758         if (!storvsc_drv_obj->Base.OnDeviceRemove) {
 759                 DPRINT_EXIT(BLKVSC_DRV);
 760                 return -1;
 761         }
 762
 763         /*
 764          * Call to the vsc driver to let it know that the device is being
 765          * removed
 766          */
 767         ret = storvsc_drv_obj->Base.OnDeviceRemove(device_obj);
 768         if (ret != 0) {
 769                 /* TODO: */
 770                 DPRINT_ERR(BLKVSC_DRV,
 771                            "unable to remove blkvsc device (ret %d)", ret);
 772         }
 773
 774         /* Get to a known state */
 775         spin_lock_irqsave(&blkdev->lock, flags);
 776
 777         blkdev->shutting_down = 1;
 778
 779         blk_stop_queue(blkdev->gd->queue);
 780
 781         spin_unlock_irqrestore(&blkdev->lock, flags);
 782
 783         while (blkdev->num_outstanding_reqs) {
 784                 DPRINT_INFO(STORVSC, "waiting for %d requests to complete...",
 785                             blkdev->num_outstanding_reqs);
 786                 udelay(100);
 787         }
 788
 789         blkvsc_do_flush(blkdev);
 790
 791         spin_lock_irqsave(&blkdev->lock, flags);
 792
 793         blkvsc_cancel_pending_reqs(blkdev);
 794
 795         spin_unlock_irqrestore(&blkdev->lock, flags);
 796
 797         blk_cleanup_queue(blkdev->gd->queue);
 798
 799         del_gendisk(blkdev->gd);
 800
 801         kmem_cache_destroy(blkdev->request_pool);
 802
 803         kfree(blkdev);
 804
 805         DPRINT_EXIT(BLKVSC_DRV);
 806
 807         return ret;
 808 }
 809
 810 static void blkvsc_init_rw(struct blkvsc_request *blkvsc_req)
 811 {
 812         /* ASSERT(blkvsc_req->req); */
 813         /* ASSERT(blkvsc_req->sector_count <= (MAX_MULTIPAGE_BUFFER_COUNT*8)); */
 814
 815         blkvsc_req->cmd_len = 16;
 816
 817         if (blkvsc_req->sector_start > 0xffffffff) {
 818                 if (rq_data_dir(blkvsc_req->req)) {
 819                         blkvsc_req->write = 1;
 820                         blkvsc_req->cmnd[0] = WRITE_16;
 821                 } else {
 822                         blkvsc_req->write = 0;
 823                         blkvsc_req->cmnd[0] = READ_16;
 824                 }
 825
 826                 blkvsc_req->cmnd[1] |= blk_fua_rq(blkvsc_req->req) ? 0x8 : 0;
 827
 828                 *(unsigned long long *)&blkvsc_req->cmnd[2] =
 829                                 cpu_to_be64(blkvsc_req->sector_start);
 830                 *(unsigned int *)&blkvsc_req->cmnd[10] =
 831                                 cpu_to_be32(blkvsc_req->sector_count);
 832         } else if ((blkvsc_req->sector_count > 0xff) ||
 833                    (blkvsc_req->sector_start > 0x1fffff)) {
 834                 if (rq_data_dir(blkvsc_req->req)) {
 835                         blkvsc_req->write = 1;
 836                         blkvsc_req->cmnd[0] = WRITE_10;
 837                 } else {
 838                         blkvsc_req->write = 0;
 839                         blkvsc_req->cmnd[0] = READ_10;
 840                 }
 841
 842                 blkvsc_req->cmnd[1] |= blk_fua_rq(blkvsc_req->req) ? 0x8 : 0;
 843
 844                 *(unsigned int *)&blkvsc_req->cmnd[2] =
 845                                 cpu_to_be32(blkvsc_req->sector_start);
 846                 *(unsigned short *)&blkvsc_req->cmnd[7] =
 847                                 cpu_to_be16(blkvsc_req->sector_count);
 848         } else {
 849                 if (rq_data_dir(blkvsc_req->req)) {
 850                         blkvsc_req->write = 1;
 851                         blkvsc_req->cmnd[0] = WRITE_6;
 852                 } else {
 853                         blkvsc_req->write = 0;
 854                         blkvsc_req->cmnd[0] = READ_6;
 855                 }
 856
 857                 *(unsigned int *)&blkvsc_req->cmnd[1] =
 858                                 cpu_to_be32(blkvsc_req->sector_start) >> 8;
 859                 blkvsc_req->cmnd[1] &= 0x1f;
 860                 blkvsc_req->cmnd[4] = (unsigned char)blkvsc_req->sector_count;
 861         }
 862 }
 863
 864 static int blkvsc_submit_request(struct blkvsc_request *blkvsc_req,
 865                         void (*request_completion)(struct hv_storvsc_request *))
 866 {
 867         struct block_device_context *blkdev = blkvsc_req->dev;
 868         struct vm_device *device_ctx = blkdev->device_ctx;
 869         struct driver_context *driver_ctx =
 870                         driver_to_driver_context(device_ctx->device.driver);
 871         struct blkvsc_driver_context *blkvsc_drv_ctx =
 872                         (struct blkvsc_driver_context *)driver_ctx;
 873         struct storvsc_driver_object *storvsc_drv_obj =
 874                         &blkvsc_drv_ctx->drv_obj;
 875         struct hv_storvsc_request *storvsc_req;
 876         int ret;
 877
 878         DPRINT_DBG(BLKVSC_DRV, "blkvsc_submit_request() - "
 879                    "req %p type %s start_sector %lu count %ld offset %d "
 880                    "len %d\n", blkvsc_req,
 881                    (blkvsc_req->write) ? "WRITE" : "READ",
 882                    (unsigned long) blkvsc_req->sector_start,
 883                    blkvsc_req->sector_count,
 884                    blkvsc_req->request.DataBuffer.Offset,
 885                    blkvsc_req->request.DataBuffer.Length);
 886 #if 0
 887         for (i = 0; i < (blkvsc_req->request.DataBuffer.Length >> 12); i++) {
 888                 DPRINT_DBG(BLKVSC_DRV, "blkvsc_submit_request() - "
 889                            "req %p pfn[%d] %llx\n",
 890                            blkvsc_req, i,
 891                            blkvsc_req->request.DataBuffer.PfnArray[i]);
 892         }
 893 #endif
 894
 895         storvsc_req = &blkvsc_req->request;
 896         storvsc_req->Extension = (void *)((unsigned long)blkvsc_req +
 897                                           sizeof(struct blkvsc_request));
 898
 899         storvsc_req->Type = blkvsc_req->write ? WRITE_TYPE : READ_TYPE;
 900
 901         storvsc_req->OnIOCompletion = request_completion;
 902         storvsc_req->Context = blkvsc_req;
 903
 904         storvsc_req->Host = blkdev->port;
 905         storvsc_req->Bus = blkdev->path;
 906         storvsc_req->TargetId = blkdev->target;
 907         storvsc_req->LunId = 0;  /* this is not really used at all */
 908
 909         storvsc_req->CdbLen = blkvsc_req->cmd_len;
 910         storvsc_req->Cdb = blkvsc_req->cmnd;
 911
 912         storvsc_req->SenseBuffer = blkvsc_req->sense_buffer;
 913         storvsc_req->SenseBufferSize = SCSI_SENSE_BUFFERSIZE;
 914
 915         ret = storvsc_drv_obj->OnIORequest(&blkdev->device_ctx->device_obj,
 916                                            &blkvsc_req->request);
 917         if (ret == 0)
 918                 blkdev->num_outstanding_reqs++;
 919
 920         return ret;
 921 }
 922
 923 /*
 924  * We break the request into 1 or more blkvsc_requests and submit
 925  * them.  If we cant submit them all, we put them on the
 926  * pending_list. The blkvsc_request() will work on the pending_list.
 927  */
 928 static int blkvsc_do_request(struct block_device_context *blkdev,
 929                              struct request *req)
 930 {
 931         struct bio *bio = NULL;
 932         struct bio_vec *bvec = NULL;
 933         struct bio_vec *prev_bvec = NULL;
 934         struct blkvsc_request *blkvsc_req = NULL;
 935         struct blkvsc_request *tmp;
 936         int databuf_idx = 0;
 937         int seg_idx = 0;
 938         sector_t start_sector;
 939         unsigned long num_sectors = 0;
 940         int ret = 0;
 941         int pending = 0;
 942         struct blkvsc_request_group *group = NULL;
 943
 944         DPRINT_DBG(BLKVSC_DRV, "blkdev %p req %p sect %lu\n", blkdev, req,
 945                   (unsigned long)blk_rq_pos(req));
 946
 947         /* Create a group to tie req to list of blkvsc_reqs */
 948         group = kmem_cache_alloc(blkdev->request_pool, GFP_ATOMIC);
 949         if (!group)
 950                 return -ENOMEM;
 951
 952         INIT_LIST_HEAD(&group->blkvsc_req_list);
 953         group->outstanding = group->status = 0;
 954
 955         start_sector = blk_rq_pos(req);
 956
 957         /* foreach bio in the request */
 958         if (req->bio) {
 959                 for (bio = req->bio; bio; bio = bio->bi_next) {
 960                         /*
 961                          * Map this bio into an existing or new storvsc request
 962                          */
 963                         bio_for_each_segment(bvec, bio, seg_idx) {
 964                                 DPRINT_DBG(BLKVSC_DRV, "bio_for_each_segment() "
 965                                            "- req %p bio %p bvec %p seg_idx %d "
 966                                            "databuf_idx %d\n", req, bio, bvec,
 967                                            seg_idx, databuf_idx);
 968
 969                                 /* Get a new storvsc request */
 970                                 /* 1st-time */
 971                                 if ((!blkvsc_req) ||
 972                                     (databuf_idx >= MAX_MULTIPAGE_BUFFER_COUNT)
 973                                     /* hole at the begin of page */
 974                                     || (bvec->bv_offset != 0) ||
 975                                     /* hold at the end of page */
 976                                     (prev_bvec &&
 977                                      (prev_bvec->bv_len != PAGE_SIZE))) {
 978                                         /* submit the prev one */
 979                                         if (blkvsc_req) {
 980                                                 blkvsc_req->sector_start = start_sector;
 981                                                 sector_div(blkvsc_req->sector_start, (blkdev->sector_size >> 9));
 982
 983                                                 blkvsc_req->sector_count = num_sectors / (blkdev->sector_size >> 9);
 984                                                 blkvsc_init_rw(blkvsc_req);
 985                                         }
 986
 987                                         /*
 988                                          * Create new blkvsc_req to represent
 989                                          * the current bvec
 990                                          */
 991                                         blkvsc_req = kmem_cache_alloc(blkdev->request_pool, GFP_ATOMIC);
 992                                         if (!blkvsc_req) {
 993                                                 /* free up everything */
 994                                                 list_for_each_entry_safe(
 995                                                         blkvsc_req, tmp,
 996                                                         &group->blkvsc_req_list,
 997                                                         req_entry) {
 998                                                         list_del(&blkvsc_req->req_entry);
 999                                                         kmem_cache_free(blkdev->request_pool, blkvsc_req);
1000                                                 }
1001
1002                                                 kmem_cache_free(blkdev->request_pool, group);
1003                                                 return -ENOMEM;
1004                                         }
1005
1006                                         memset(blkvsc_req, 0,
1007                                                sizeof(struct blkvsc_request));
1008
1009                                         blkvsc_req->dev = blkdev;
1010                                         blkvsc_req->req = req;
1011                                         blkvsc_req->request.DataBuffer.Offset = bvec->bv_offset;
1012                                         blkvsc_req->request.DataBuffer.Length = 0;
1013
1014                                         /* Add to the group */
1015                                         blkvsc_req->group = group;
1016                                         blkvsc_req->group->outstanding++;
1017                                         list_add_tail(&blkvsc_req->req_entry,
1018                                                 &blkvsc_req->group->blkvsc_req_list);
1019
1020                                         start_sector += num_sectors;
1021                                         num_sectors = 0;
1022                                         databuf_idx = 0;
1023                                 }
1024
1025                                 /* Add the curr bvec/segment to the curr blkvsc_req */
1026                                 blkvsc_req->request.DataBuffer.PfnArray[databuf_idx] = page_to_pfn(bvec->bv_page);
1027                                 blkvsc_req->request.DataBuffer.Length += bvec->bv_len;
1028
1029                                 prev_bvec = bvec;
1030
1031                                 databuf_idx++;
1032                                 num_sectors += bvec->bv_len >> 9;
1033
1034                         } /* bio_for_each_segment */
1035
1036                 } /* rq_for_each_bio */
1037         }
1038
1039         /* Handle the last one */
1040         if (blkvsc_req) {
1041                 DPRINT_DBG(BLKVSC_DRV, "blkdev %p req %p group %p count %d\n",
1042                            blkdev, req, blkvsc_req->group,
1043                            blkvsc_req->group->outstanding);
1044
1045                 blkvsc_req->sector_start = start_sector;
1046                 sector_div(blkvsc_req->sector_start,
1047                            (blkdev->sector_size >> 9));
1048
1049                 blkvsc_req->sector_count = num_sectors /
1050                                            (blkdev->sector_size >> 9);
1051
1052                 blkvsc_init_rw(blkvsc_req);
1053         }
1054
1055         list_for_each_entry(blkvsc_req, &group->blkvsc_req_list, req_entry) {
1056                 if (pending) {
1057                         DPRINT_DBG(BLKVSC_DRV, "adding blkvsc_req to "
1058                                    "pending_list - blkvsc_req %p start_sect %lu"
1059                                    " sect_count %ld (%lu %ld)\n", blkvsc_req,
1060                                    (unsigned long)blkvsc_req->sector_start,
1061                                    blkvsc_req->sector_count,
1062                                    (unsigned long)start_sector,
1063                                    (unsigned long)num_sectors);
1064
1065                         list_add_tail(&blkvsc_req->pend_entry,
1066                                       &blkdev->pending_list);
1067                 } else {
1068                         ret = blkvsc_submit_request(blkvsc_req,
1069                                                     blkvsc_request_completion);
1070                         if (ret == -1) {
1071                                 pending = 1;
1072                                 list_add_tail(&blkvsc_req->pend_entry,
1073                                               &blkdev->pending_list);
1074                         }
1075
1076                         DPRINT_DBG(BLKVSC_DRV, "submitted blkvsc_req %p "
1077                                    "start_sect %lu sect_count %ld (%lu %ld) "
1078                                    "ret %d\n", blkvsc_req,
1079                                    (unsigned long)blkvsc_req->sector_start,
1080                                    blkvsc_req->sector_count,
1081                                    (unsigned long)start_sector,
1082                                    num_sectors, ret);
1083                 }
1084         }
1085
1086         return pending;
1087 }
1088
1089 static void blkvsc_cmd_completion(struct hv_storvsc_request *request)
1090 {
1091         struct blkvsc_request *blkvsc_req =
1092                         (struct blkvsc_request *)request->Context;
1093         struct block_device_context *blkdev =
1094                         (struct block_device_context *)blkvsc_req->dev;
1095         struct scsi_sense_hdr sense_hdr;
1096
1097         DPRINT_DBG(BLKVSC_DRV, "blkvsc_cmd_completion() - req %p\n",
1098                    blkvsc_req);
1099
1100         blkdev->num_outstanding_reqs--;
1101
1102         if (blkvsc_req->request.Status)
1103                 if (scsi_normalize_sense(blkvsc_req->sense_buffer,
1104                                          SCSI_SENSE_BUFFERSIZE, &sense_hdr))
1105                         scsi_print_sense_hdr("blkvsc", &sense_hdr);
1106
1107         blkvsc_req->cond = 1;
1108         wake_up_interruptible(&blkvsc_req->wevent);
1109 }
1110
1111 static void blkvsc_request_completion(struct hv_storvsc_request *request)
1112 {
1113         struct blkvsc_request *blkvsc_req =
1114                         (struct blkvsc_request *)request->Context;
1115         struct block_device_context *blkdev =
1116                         (struct block_device_context *)blkvsc_req->dev;
1117         unsigned long flags;
1118         struct blkvsc_request *comp_req, *tmp;
1119
1120         /* ASSERT(blkvsc_req->group); */
1121
1122         DPRINT_DBG(BLKVSC_DRV, "blkdev %p blkvsc_req %p group %p type %s "
1123                    "sect_start %lu sect_count %ld len %d group outstd %d "
1124                    "total outstd %d\n",
1125                    blkdev, blkvsc_req, blkvsc_req->group,
1126                    (blkvsc_req->write) ? "WRITE" : "READ",
1127                    (unsigned long)blkvsc_req->sector_start,
1128                    blkvsc_req->sector_count,
1129                    blkvsc_req->request.DataBuffer.Length,
1130                    blkvsc_req->group->outstanding,
1131                    blkdev->num_outstanding_reqs);
1132
1133         spin_lock_irqsave(&blkdev->lock, flags);
1134
1135         blkdev->num_outstanding_reqs--;
1136         blkvsc_req->group->outstanding--;
1137
1138         /*
1139          * Only start processing when all the blkvsc_reqs are
1140          * completed. This guarantees no out-of-order blkvsc_req
1141          * completion when calling end_that_request_first()
1142          */
1143         if (blkvsc_req->group->outstanding == 0) {
1144                 list_for_each_entry_safe(comp_req, tmp,
1145                                          &blkvsc_req->group->blkvsc_req_list,
1146                                          req_entry) {
1147                         DPRINT_DBG(BLKVSC_DRV, "completing blkvsc_req %p "
1148                                    "sect_start %lu sect_count %ld\n",
1149                                    comp_req,
1150                                    (unsigned long)comp_req->sector_start,
1151                                    comp_req->sector_count);
1152
1153                         list_del(&comp_req->req_entry);
1154
1155                         if (!__blk_end_request(comp_req->req,
1156                                 (!comp_req->request.Status ? 0 : -EIO),
1157                                 comp_req->sector_count * blkdev->sector_size)) {
1158                                 /*
1159                                  * All the sectors have been xferred ie the
1160                                  * request is done
1161                                  */
1162                                 DPRINT_DBG(BLKVSC_DRV, "req %p COMPLETED\n",
1163                                            comp_req->req);
1164                                 kmem_cache_free(blkdev->request_pool,
1165                                                 comp_req->group);
1166                         }
1167
1168                         kmem_cache_free(blkdev->request_pool, comp_req);
1169                 }
1170
1171                 if (!blkdev->shutting_down) {
1172                         blkvsc_do_pending_reqs(blkdev);
1173                         blk_start_queue(blkdev->gd->queue);
1174                         blkvsc_request(blkdev->gd->queue);
1175                 }
1176         }
1177
1178         spin_unlock_irqrestore(&blkdev->lock, flags);
1179 }
1180
1181 static int blkvsc_cancel_pending_reqs(struct block_device_context *blkdev)
1182 {
1183         struct blkvsc_request *pend_req, *tmp;
1184         struct blkvsc_request *comp_req, *tmp2;
1185
1186         int ret = 0;
1187
1188         DPRINT_DBG(BLKVSC_DRV, "blkvsc_cancel_pending_reqs()");
1189
1190         /* Flush the pending list first */
1191         list_for_each_entry_safe(pend_req, tmp, &blkdev->pending_list,
1192                                  pend_entry) {
1193                 /*
1194                  * The pend_req could be part of a partially completed
1195                  * request. If so, complete those req first until we
1196                  * hit the pend_req
1197                  */
1198                 list_for_each_entry_safe(comp_req, tmp2,
1199                                          &pend_req->group->blkvsc_req_list,
1200                                          req_entry) {
1201                         DPRINT_DBG(BLKVSC_DRV, "completing blkvsc_req %p "
1202                                    "sect_start %lu sect_count %ld\n",
1203                                    comp_req,
1204                                    (unsigned long) comp_req->sector_start,
1205                                    comp_req->sector_count);
1206
1207                         if (comp_req == pend_req)
1208                                 break;
1209
1210                         list_del(&comp_req->req_entry);
1211
1212                         if (comp_req->req) {
1213                                 ret = __blk_end_request(comp_req->req,
1214                                         (!comp_req->request.Status ? 0 : -EIO),
1215                                         comp_req->sector_count *
1216                                         blkdev->sector_size);
1217
1218                                 /* FIXME: shouldn't this do more than return? */
1219                                 if (ret)
1220                                         goto out;
1221                         }
1222
1223                         kmem_cache_free(blkdev->request_pool, comp_req);
1224                 }
1225
1226                 DPRINT_DBG(BLKVSC_DRV, "cancelling pending request - %p\n",
1227                            pend_req);
1228
1229                 list_del(&pend_req->pend_entry);
1230
1231                 list_del(&pend_req->req_entry);
1232
1233                 if (comp_req->req) {
1234                         if (!__blk_end_request(pend_req->req, -EIO,
1235                                                pend_req->sector_count *
1236                                                blkdev->sector_size)) {
1237                                 /*
1238                                  * All the sectors have been xferred ie the
1239                                  * request is done
1240                                  */
1241                                 DPRINT_DBG(BLKVSC_DRV,
1242                                            "blkvsc_cancel_pending_reqs() - "
1243                                            "req %p COMPLETED\n", pend_req->req);
1244                                 kmem_cache_free(blkdev->request_pool,
1245                                                 pend_req->group);
1246                         }
1247                 }
1248
1249                 kmem_cache_free(blkdev->request_pool, pend_req);
1250         }
1251
1252 out:
1253         return ret;
1254 }
1255
1256 static int blkvsc_do_pending_reqs(struct block_device_context *blkdev)
1257 {
1258         struct blkvsc_request *pend_req, *tmp;
1259         int ret = 0;
1260
1261         /* Flush the pending list first */
1262         list_for_each_entry_safe(pend_req, tmp, &blkdev->pending_list,
1263                                  pend_entry) {
1264                 DPRINT_DBG(BLKVSC_DRV, "working off pending_list - %p\n",
1265                            pend_req);
1266
1267                 ret = blkvsc_submit_request(pend_req,
1268                                             blkvsc_request_completion);
1269                 if (ret != 0)
1270                         break;
1271                 else
1272                         list_del(&pend_req->pend_entry);
1273         }
1274
1275         return ret;
1276 }
1277
1278 static void blkvsc_request(struct request_queue *queue)
1279 {
1280         struct block_device_context *blkdev = NULL;
1281         struct request *req;
1282         int ret = 0;
1283
1284         DPRINT_DBG(BLKVSC_DRV, "- enter\n");
1285         while ((req = blk_peek_request(queue)) != NULL) {
1286                 DPRINT_DBG(BLKVSC_DRV, "- req %p\n", req);
1287
1288                 blkdev = req->rq_disk->private_data;
1289                 if (blkdev->shutting_down || !blk_fs_request(req) ||
1290                     blkdev->media_not_present) {
1291                         __blk_end_request_cur(req, 0);
1292                         continue;
1293                 }
1294
1295                 ret = blkvsc_do_pending_reqs(blkdev);
1296
1297                 if (ret != 0) {
1298                         DPRINT_DBG(BLKVSC_DRV,
1299                                    "- stop queue - pending_list not empty\n");
1300                         blk_stop_queue(queue);
1301                         break;
1302                 }
1303
1304                 blk_start_request(req);
1305
1306                 ret = blkvsc_do_request(blkdev, req);
1307                 if (ret > 0) {
1308                         DPRINT_DBG(BLKVSC_DRV, "- stop queue - no room\n");
1309                         blk_stop_queue(queue);
1310                         break;
1311                 } else if (ret < 0) {
1312                         DPRINT_DBG(BLKVSC_DRV, "- stop queue - no mem\n");
1313                         blk_requeue_request(queue, req);
1314                         blk_stop_queue(queue);
1315                         break;
1316                 }
1317         }
1318 }
1319
1320 static int blkvsc_open(struct block_device *bdev, fmode_t mode)
1321 {
1322         struct block_device_context *blkdev = bdev->bd_disk->private_data;
1323
1324         DPRINT_DBG(BLKVSC_DRV, "- users %d disk %s\n", blkdev->users,
1325                    blkdev->gd->disk_name);
1326
1327         spin_lock(&blkdev->lock);
1328
1329         if (!blkdev->users && blkdev->device_type == DVD_TYPE) {
1330                 spin_unlock(&blkdev->lock);
1331                 check_disk_change(bdev);
1332                 spin_lock(&blkdev->lock);
1333         }
1334
1335         blkdev->users++;
1336
1337         spin_unlock(&blkdev->lock);
1338         return 0;
1339 }
1340
1341 static int blkvsc_release(struct gendisk *disk, fmode_t mode)
1342 {
1343         struct block_device_context *blkdev = disk->private_data;
1344
1345         DPRINT_DBG(BLKVSC_DRV, "- users %d disk %s\n", blkdev->users,
1346                    blkdev->gd->disk_name);
1347
1348         spin_lock(&blkdev->lock);
1349         if (blkdev->users == 1) {
1350                 spin_unlock(&blkdev->lock);
1351                 blkvsc_do_flush(blkdev);
1352                 spin_lock(&blkdev->lock);
1353         }
1354
1355         blkdev->users--;
1356
1357         spin_unlock(&blkdev->lock);
1358         return 0;
1359 }
1360
1361 static int blkvsc_media_changed(struct gendisk *gd)
1362 {
1363         DPRINT_DBG(BLKVSC_DRV, "- enter\n");
1364         return 1;
1365 }
1366
1367 static int blkvsc_revalidate_disk(struct gendisk *gd)
1368 {
1369         struct block_device_context *blkdev = gd->private_data;
1370
1371         DPRINT_DBG(BLKVSC_DRV, "- enter\n");
1372
1373         if (blkdev->device_type == DVD_TYPE) {
1374                 blkvsc_do_read_capacity(blkdev);
1375                 set_capacity(blkdev->gd, blkdev->capacity *
1376                             (blkdev->sector_size/512));
1377                 blk_queue_logical_block_size(gd->queue, blkdev->sector_size);
1378         }
1379         return 0;
1380 }
1381
1382 static int blkvsc_getgeo(struct block_device *bd, struct hd_geometry *hg)
1383 {
1384         sector_t total_sectors = get_capacity(bd->bd_disk);
1385         sector_t cylinder_times_heads = 0;
1386         sector_t temp = 0;
1387
1388         int sectors_per_track = 0;
1389         int heads = 0;
1390         int cylinders = 0;
1391         int rem = 0;
1392
1393         if (total_sectors > (65535 * 16 * 255))
1394                 total_sectors = (65535 * 16 * 255);
1395
1396         if (total_sectors >= (65535 * 16 * 63)) {
1397                 sectors_per_track = 255;
1398                 heads = 16;
1399
1400                 cylinder_times_heads = total_sectors;
1401                 /* sector_div stores the quotient in cylinder_times_heads */
1402                 rem = sector_div(cylinder_times_heads, sectors_per_track);
1403         } else {
1404                 sectors_per_track = 17;
1405
1406                 cylinder_times_heads = total_sectors;
1407                 /* sector_div stores the quotient in cylinder_times_heads */
1408                 rem = sector_div(cylinder_times_heads, sectors_per_track);
1409
1410                 temp = cylinder_times_heads + 1023;
1411                 /* sector_div stores the quotient in temp */
1412                 rem = sector_div(temp, 1024);
1413
1414                 heads = temp;
1415
1416                 if (heads < 4)
1417                         heads = 4;
1418
1419
1420                 if (cylinder_times_heads >= (heads * 1024) || (heads > 16)) {
1421                         sectors_per_track = 31;
1422                         heads = 16;
1423
1424                         cylinder_times_heads = total_sectors;
1425                         /*
1426                          * sector_div stores the quotient in
1427                          * cylinder_times_heads
1428                          */
1429                         rem = sector_div(cylinder_times_heads,
1430                                          sectors_per_track);
1431                 }
1432
1433                 if (cylinder_times_heads >= (heads * 1024)) {
1434                         sectors_per_track = 63;
1435                         heads = 16;
1436
1437                         cylinder_times_heads = total_sectors;
1438                         /*
1439                          * sector_div stores the quotient in
1440                          * cylinder_times_heads
1441                          */
1442                         rem = sector_div(cylinder_times_heads,
1443                                          sectors_per_track);
1444                 }
1445         }
1446
1447         temp = cylinder_times_heads;
1448         /* sector_div stores the quotient in temp */
1449         rem = sector_div(temp, heads);
1450         cylinders = temp;
1451
1452         hg->heads = heads;
1453         hg->sectors = sectors_per_track;
1454         hg->cylinders = cylinders;
1455
1456         DPRINT_INFO(BLKVSC_DRV, "CHS (%d, %d, %d)", cylinders, heads,
1457                     sectors_per_track);
1458
1459     return 0;
1460 }
1461
1462 static int blkvsc_ioctl(struct block_device *bd, fmode_t mode,
1463                         unsigned cmd, unsigned long argument)
1464 {
1465 /*      struct block_device_context *blkdev = bd->bd_disk->private_data; */
1466         int ret;
1467
1468         switch (cmd) {
1469         /*
1470          * TODO: I think there is certain format for HDIO_GET_IDENTITY rather
1471          * than just a GUID. Commented it out for now.
1472          */
1473 #if 0
1474         case HDIO_GET_IDENTITY:
1475                 DPRINT_INFO(BLKVSC_DRV, "HDIO_GET_IDENTITY\n");
1476                 if (copy_to_user((void __user *)arg, blkdev->device_id,
1477                                  blkdev->device_id_len))
1478                         ret = -EFAULT;
1479                 break;
1480 #endif
1481         default:
1482                 ret = -EINVAL;
1483                 break;
1484         }
1485
1486         return ret;
1487 }
1488
1489 static int __init blkvsc_init(void)
1490 {
1491         int ret;
1492
1493         BUILD_BUG_ON(sizeof(sector_t) != 8);
1494
1495         DPRINT_ENTER(BLKVSC_DRV);
1496
1497         DPRINT_INFO(BLKVSC_DRV, "Blkvsc initializing....");
1498
1499         ret = blkvsc_drv_init(BlkVscInitialize);
1500
1501         DPRINT_EXIT(BLKVSC_DRV);
1502
1503         return ret;
1504 }
1505
1506 static void __exit blkvsc_exit(void)
1507 {
1508         DPRINT_ENTER(BLKVSC_DRV);
1509         blkvsc_drv_exit();
1510         DPRINT_ENTER(BLKVSC_DRV);
1511 }
1512
1513 MODULE_LICENSE("GPL");
1514 MODULE_VERSION(HV_DRV_VERSION);
1515 MODULE_DESCRIPTION("Microsoft Hyper-V virtual block driver");
1516 module_init(blkvsc_init);
1517 module_exit(blkvsc_exit);