drivers/block/xen-blkfront.c

   1 /*
   2  * blkfront.c
   3  *
   4  * XenLinux virtual block device driver.
   5  *
   6  * Copyright (c) 2003-2004, Keir Fraser & Steve Hand
   7  * Modifications by Mark A. Williamson are (c) Intel Research Cambridge
   8  * Copyright (c) 2004, Christian Limpach
   9  * Copyright (c) 2004, Andrew Warfield
  10  * Copyright (c) 2005, Christopher Clark
  11  * Copyright (c) 2005, XenSource Ltd
  12  *
  13  * This program is free software; you can redistribute it and/or
  14  * modify it under the terms of the GNU General Public License version 2
  15  * as published by the Free Software Foundation; or, when distributed
  16  * separately from the Linux kernel or incorporated into other
  17  * software packages, subject to the following license:
  18  *
  19  * Permission is hereby granted, free of charge, to any person obtaining a copy
  20  * of this source file (the "Software"), to deal in the Software without
  21  * restriction, including without limitation the rights to use, copy, modify,
  22  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  23  * and to permit persons to whom the Software is furnished to do so, subject to
  24  * the following conditions:
  25  *
  26  * The above copyright notice and this permission notice shall be included in
  27  * all copies or substantial portions of the Software.
  28  *
  29  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  30  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  31  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  32  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  33  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  34  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  35  * IN THE SOFTWARE.
  36  */
  37
  38 #include <linux/interrupt.h>
  39 #include <linux/blkdev.h>
  40 #include <linux/module.h>
  41
  42 #include <xen/xenbus.h>
  43 #include <xen/grant_table.h>
  44 #include <xen/events.h>
  45 #include <xen/page.h>
  46
  47 #include <xen/interface/grant_table.h>
  48 #include <xen/interface/io/blkif.h>
  49
  50 #include <asm/xen/hypervisor.h>
  51
  52 enum blkif_state {
  53         BLKIF_STATE_DISCONNECTED,
  54         BLKIF_STATE_CONNECTED,
  55         BLKIF_STATE_SUSPENDED,
  56 };
  57
  58 struct blk_shadow {
  59         struct blkif_request req;
  60         unsigned long request;
  61         unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
  62 };
  63
  64 static struct block_device_operations xlvbd_block_fops;
  65
  66 #define BLK_RING_SIZE __RING_SIZE((struct blkif_sring *)0, PAGE_SIZE)
  67
  68 /*
  69  * We have one of these per vbd, whether ide, scsi or 'other'.  They
  70  * hang in private_data off the gendisk structure. We may end up
  71  * putting all kinds of interesting stuff here :-)
  72  */
  73 struct blkfront_info
  74 {
  75         struct xenbus_device *xbdev;
  76         dev_t dev;
  77         struct gendisk *gd;
  78         int vdevice;
  79         blkif_vdev_t handle;
  80         enum blkif_state connected;
  81         int ring_ref;
  82         struct blkif_front_ring ring;
  83         unsigned int evtchn, irq;
  84         struct request_queue *rq;
  85         struct work_struct work;
  86         struct gnttab_free_callback callback;
  87         struct blk_shadow shadow[BLK_RING_SIZE];
  88         unsigned long shadow_free;
  89         int feature_barrier;
  90
  91         /**
  92          * The number of people holding this device open.  We won't allow a
  93          * hot-unplug unless this is 0.
  94          */
  95         int users;
  96 };
  97
  98 static DEFINE_SPINLOCK(blkif_io_lock);
  99
 100 #define MAXIMUM_OUTSTANDING_BLOCK_REQS \
 101         (BLKIF_MAX_SEGMENTS_PER_REQUEST * BLK_RING_SIZE)
 102 #define GRANT_INVALID_REF       0
 103
 104 #define PARTS_PER_DISK          16
 105
 106 #define BLKIF_MAJOR(dev) ((dev)>>8)
 107 #define BLKIF_MINOR(dev) ((dev) & 0xff)
 108
 109 #define DEV_NAME        "xvd"   /* name in /dev */
 110
 111 /* Information about our VBDs. */
 112 #define MAX_VBDS 64
 113 static LIST_HEAD(vbds_list);
 114
 115 static int get_id_from_freelist(struct blkfront_info *info)
 116 {
 117         unsigned long free = info->shadow_free;
 118         BUG_ON(free > BLK_RING_SIZE);
 119         info->shadow_free = info->shadow[free].req.id;
 120         info->shadow[free].req.id = 0x0fffffee; /* debug */
 121         return free;
 122 }
 123
 124 static void add_id_to_freelist(struct blkfront_info *info,
 125                                unsigned long id)
 126 {
 127         info->shadow[id].req.id  = info->shadow_free;
 128         info->shadow[id].request = 0;
 129         info->shadow_free = id;
 130 }
 131
 132 static void blkif_restart_queue_callback(void *arg)
 133 {
 134         struct blkfront_info *info = (struct blkfront_info *)arg;
 135         schedule_work(&info->work);
 136 }
 137
 138 /*
 139  * blkif_queue_request
 140  *
 141  * request block io
 142  *
 143  * id: for guest use only.
 144  * operation: BLKIF_OP_{READ,WRITE,PROBE}
 145  * buffer: buffer to read/write into. this should be a
 146  *   virtual address in the guest os.
 147  */
 148 static int blkif_queue_request(struct request *req)
 149 {
 150         struct blkfront_info *info = req->rq_disk->private_data;
 151         unsigned long buffer_mfn;
 152         struct blkif_request *ring_req;
 153         struct req_iterator iter;
 154         struct bio_vec *bvec;
 155         unsigned long id;
 156         unsigned int fsect, lsect;
 157         int ref;
 158         grant_ref_t gref_head;
 159
 160         if (unlikely(info->connected != BLKIF_STATE_CONNECTED))
 161                 return 1;
 162
 163         if (gnttab_alloc_grant_references(
 164                 BLKIF_MAX_SEGMENTS_PER_REQUEST, &gref_head) < 0) {
 165                 gnttab_request_free_callback(
 166                         &info->callback,
 167                         blkif_restart_queue_callback,
 168                         info,
 169                         BLKIF_MAX_SEGMENTS_PER_REQUEST);
 170                 return 1;
 171         }
 172
 173         /* Fill out a communications ring structure. */
 174         ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
 175         id = get_id_from_freelist(info);
 176         info->shadow[id].request = (unsigned long)req;
 177
 178         ring_req->id = id;
 179         ring_req->sector_number = (blkif_sector_t)req->sector;
 180         ring_req->handle = info->handle;
 181
 182         ring_req->operation = rq_data_dir(req) ?
 183                 BLKIF_OP_WRITE : BLKIF_OP_READ;
 184         if (blk_barrier_rq(req))
 185                 ring_req->operation = BLKIF_OP_WRITE_BARRIER;
 186
 187         ring_req->nr_segments = 0;
 188         rq_for_each_segment(bvec, req, iter) {
 189                 BUG_ON(ring_req->nr_segments == BLKIF_MAX_SEGMENTS_PER_REQUEST);
 190                 buffer_mfn = pfn_to_mfn(page_to_pfn(bvec->bv_page));
 191                 fsect = bvec->bv_offset >> 9;
 192                 lsect = fsect + (bvec->bv_len >> 9) - 1;
 193                 /* install a grant reference. */
 194                 ref = gnttab_claim_grant_reference(&gref_head);
 195                 BUG_ON(ref == -ENOSPC);
 196
 197                 gnttab_grant_foreign_access_ref(
 198                                 ref,
 199                                 info->xbdev->otherend_id,
 200                                 buffer_mfn,
 201                                 rq_data_dir(req) );
 202
 203                 info->shadow[id].frame[ring_req->nr_segments] =
 204                                 mfn_to_pfn(buffer_mfn);
 205
 206                 ring_req->seg[ring_req->nr_segments] =
 207                                 (struct blkif_request_segment) {
 208                                         .gref       = ref,
 209                                         .first_sect = fsect,
 210                                         .last_sect  = lsect };
 211
 212                 ring_req->nr_segments++;
 213         }
 214
 215         info->ring.req_prod_pvt++;
 216
 217         /* Keep a private copy so we can reissue requests when recovering. */
 218         info->shadow[id].req = *ring_req;
 219
 220         gnttab_free_grant_references(gref_head);
 221
 222         return 0;
 223 }
 224
 225
 226 static inline void flush_requests(struct blkfront_info *info)
 227 {
 228         int notify;
 229
 230         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&info->ring, notify);
 231
 232         if (notify)
 233                 notify_remote_via_irq(info->irq);
 234 }
 235
 236 /*
 237  * do_blkif_request
 238  *  read a block; request is in a request queue
 239  */
 240 static void do_blkif_request(struct request_queue *rq)
 241 {
 242         struct blkfront_info *info = NULL;
 243         struct request *req;
 244         int queued;
 245
 246         pr_debug("Entered do_blkif_request\n");
 247
 248         queued = 0;
 249
 250         while ((req = elv_next_request(rq)) != NULL) {
 251                 info = req->rq_disk->private_data;
 252                 if (!blk_fs_request(req)) {
 253                         end_request(req, 0);
 254                         continue;
 255                 }
 256
 257                 if (RING_FULL(&info->ring))
 258                         goto wait;
 259
 260                 pr_debug("do_blk_req %p: cmd %p, sec %lx, "
 261                          "(%u/%li) buffer:%p [%s]\n",
 262                          req, req->cmd, (unsigned long)req->sector,
 263                          req->current_nr_sectors,
 264                          req->nr_sectors, req->buffer,
 265                          rq_data_dir(req) ? "write" : "read");
 266
 267
 268                 blkdev_dequeue_request(req);
 269                 if (blkif_queue_request(req)) {
 270                         blk_requeue_request(rq, req);
 271 wait:
 272                         /* Avoid pointless unplugs. */
 273                         blk_stop_queue(rq);
 274                         break;
 275                 }
 276
 277                 queued++;
 278         }
 279
 280         if (queued != 0)
 281                 flush_requests(info);
 282 }
 283
 284 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size)
 285 {
 286         struct request_queue *rq;
 287
 288         rq = blk_init_queue(do_blkif_request, &blkif_io_lock);
 289         if (rq == NULL)
 290                 return -1;
 291
 292         elevator_init(rq, "noop");
 293
 294         /* Hard sector size and max sectors impersonate the equiv. hardware. */
 295         blk_queue_hardsect_size(rq, sector_size);
 296         blk_queue_max_sectors(rq, 512);
 297
 298         /* Each segment in a request is up to an aligned page in size. */
 299         blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
 300         blk_queue_max_segment_size(rq, PAGE_SIZE);
 301
 302         /* Ensure a merged request will fit in a single I/O ring slot. */
 303         blk_queue_max_phys_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
 304         blk_queue_max_hw_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
 305
 306         /* Make sure buffer addresses are sector-aligned. */
 307         blk_queue_dma_alignment(rq, 511);
 308
 309         gd->queue = rq;
 310
 311         return 0;
 312 }
 313
 314
 315 static int xlvbd_barrier(struct blkfront_info *info)
 316 {
 317         int err;
 318
 319         err = blk_queue_ordered(info->rq,
 320                                 info->feature_barrier ? QUEUE_ORDERED_DRAIN : QUEUE_ORDERED_NONE,
 321                                 NULL);
 322
 323         if (err)
 324                 return err;
 325
 326         printk(KERN_INFO "blkfront: %s: barriers %s\n",
 327                info->gd->disk_name,
 328                info->feature_barrier ? "enabled" : "disabled");
 329         return 0;
 330 }
 331
 332
 333 static int xlvbd_alloc_gendisk(int minor, blkif_sector_t capacity,
 334                                int vdevice, u16 vdisk_info, u16 sector_size,
 335                                struct blkfront_info *info)
 336 {
 337         struct gendisk *gd;
 338         int nr_minors = 1;
 339         int err = -ENODEV;
 340
 341         BUG_ON(info->gd != NULL);
 342         BUG_ON(info->rq != NULL);
 343
 344         if ((minor % PARTS_PER_DISK) == 0)
 345                 nr_minors = PARTS_PER_DISK;
 346
 347         gd = alloc_disk(nr_minors);
 348         if (gd == NULL)
 349                 goto out;
 350
 351         if (nr_minors > 1)
 352                 sprintf(gd->disk_name, "%s%c", DEV_NAME,
 353                         'a' + minor / PARTS_PER_DISK);
 354         else
 355                 sprintf(gd->disk_name, "%s%c%d", DEV_NAME,
 356                         'a' + minor / PARTS_PER_DISK,
 357                         minor % PARTS_PER_DISK);
 358
 359         gd->major = XENVBD_MAJOR;
 360         gd->first_minor = minor;
 361         gd->fops = &xlvbd_block_fops;
 362         gd->private_data = info;
 363         gd->driverfs_dev = &(info->xbdev->dev);
 364         set_capacity(gd, capacity);
 365
 366         if (xlvbd_init_blk_queue(gd, sector_size)) {
 367                 del_gendisk(gd);
 368                 goto out;
 369         }
 370
 371         info->rq = gd->queue;
 372         info->gd = gd;
 373
 374         if (info->feature_barrier)
 375                 xlvbd_barrier(info);
 376
 377         if (vdisk_info & VDISK_READONLY)
 378                 set_disk_ro(gd, 1);
 379
 380         if (vdisk_info & VDISK_REMOVABLE)
 381                 gd->flags |= GENHD_FL_REMOVABLE;
 382
 383         if (vdisk_info & VDISK_CDROM)
 384                 gd->flags |= GENHD_FL_CD;
 385
 386         return 0;
 387
 388  out:
 389         return err;
 390 }
 391
 392 static void kick_pending_request_queues(struct blkfront_info *info)
 393 {
 394         if (!RING_FULL(&info->ring)) {
 395                 /* Re-enable calldowns. */
 396                 blk_start_queue(info->rq);
 397                 /* Kick things off immediately. */
 398                 do_blkif_request(info->rq);
 399         }
 400 }
 401
 402 static void blkif_restart_queue(struct work_struct *work)
 403 {
 404         struct blkfront_info *info = container_of(work, struct blkfront_info, work);
 405
 406         spin_lock_irq(&blkif_io_lock);
 407         if (info->connected == BLKIF_STATE_CONNECTED)
 408                 kick_pending_request_queues(info);
 409         spin_unlock_irq(&blkif_io_lock);
 410 }
 411
 412 static void blkif_free(struct blkfront_info *info, int suspend)
 413 {
 414         /* Prevent new requests being issued until we fix things up. */
 415         spin_lock_irq(&blkif_io_lock);
 416         info->connected = suspend ?
 417                 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED;
 418         /* No more blkif_request(). */
 419         if (info->rq)
 420                 blk_stop_queue(info->rq);
 421         /* No more gnttab callback work. */
 422         gnttab_cancel_free_callback(&info->callback);
 423         spin_unlock_irq(&blkif_io_lock);
 424
 425         /* Flush gnttab callback work. Must be done with no locks held. */
 426         flush_scheduled_work();
 427
 428         /* Free resources associated with old device channel. */
 429         if (info->ring_ref != GRANT_INVALID_REF) {
 430                 gnttab_end_foreign_access(info->ring_ref, 0,
 431                                           (unsigned long)info->ring.sring);
 432                 info->ring_ref = GRANT_INVALID_REF;
 433                 info->ring.sring = NULL;
 434         }
 435         if (info->irq)
 436                 unbind_from_irqhandler(info->irq, info);
 437         info->evtchn = info->irq = 0;
 438
 439 }
 440
 441 static void blkif_completion(struct blk_shadow *s)
 442 {
 443         int i;
 444         for (i = 0; i < s->req.nr_segments; i++)
 445                 gnttab_end_foreign_access(s->req.seg[i].gref, 0, 0UL);
 446 }
 447
 448 static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 449 {
 450         struct request *req;
 451         struct blkif_response *bret;
 452         RING_IDX i, rp;
 453         unsigned long flags;
 454         struct blkfront_info *info = (struct blkfront_info *)dev_id;
 455         int error;
 456
 457         spin_lock_irqsave(&blkif_io_lock, flags);
 458
 459         if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) {
 460                 spin_unlock_irqrestore(&blkif_io_lock, flags);
 461                 return IRQ_HANDLED;
 462         }
 463
 464  again:
 465         rp = info->ring.sring->rsp_prod;
 466         rmb(); /* Ensure we see queued responses up to 'rp'. */
 467
 468         for (i = info->ring.rsp_cons; i != rp; i++) {
 469                 unsigned long id;
 470                 int ret;
 471
 472                 bret = RING_GET_RESPONSE(&info->ring, i);
 473                 id   = bret->id;
 474                 req  = (struct request *)info->shadow[id].request;
 475
 476                 blkif_completion(&info->shadow[id]);
 477
 478                 add_id_to_freelist(info, id);
 479
 480                 error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO;
 481                 switch (bret->operation) {
 482                 case BLKIF_OP_WRITE_BARRIER:
 483                         if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
 484                                 printk(KERN_WARNING "blkfront: %s: write barrier op failed\n",
 485                                        info->gd->disk_name);
 486                                 error = -EOPNOTSUPP;
 487                                 info->feature_barrier = 0;
 488                                 xlvbd_barrier(info);
 489                         }
 490                         /* fall through */
 491                 case BLKIF_OP_READ:
 492                 case BLKIF_OP_WRITE:
 493                         if (unlikely(bret->status != BLKIF_RSP_OKAY))
 494                                 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data "
 495                                         "request: %x\n", bret->status);
 496
 497                         ret = __blk_end_request(req, error, blk_rq_bytes(req));
 498                         BUG_ON(ret);
 499                         break;
 500                 default:
 501                         BUG();
 502                 }
 503         }
 504
 505         info->ring.rsp_cons = i;
 506
 507         if (i != info->ring.req_prod_pvt) {
 508                 int more_to_do;
 509                 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
 510                 if (more_to_do)
 511                         goto again;
 512         } else
 513                 info->ring.sring->rsp_event = i + 1;
 514
 515         kick_pending_request_queues(info);
 516
 517         spin_unlock_irqrestore(&blkif_io_lock, flags);
 518
 519         return IRQ_HANDLED;
 520 }
 521
 522
 523 static int setup_blkring(struct xenbus_device *dev,
 524                          struct blkfront_info *info)
 525 {
 526         struct blkif_sring *sring;
 527         int err;
 528
 529         info->ring_ref = GRANT_INVALID_REF;
 530
 531         sring = (struct blkif_sring *)__get_free_page(GFP_KERNEL);
 532         if (!sring) {
 533                 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring");
 534                 return -ENOMEM;
 535         }
 536         SHARED_RING_INIT(sring);
 537         FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE);
 538
 539         err = xenbus_grant_ring(dev, virt_to_mfn(info->ring.sring));
 540         if (err < 0) {
 541                 free_page((unsigned long)sring);
 542                 info->ring.sring = NULL;
 543                 goto fail;
 544         }
 545         info->ring_ref = err;
 546
 547         err = xenbus_alloc_evtchn(dev, &info->evtchn);
 548         if (err)
 549                 goto fail;
 550
 551         err = bind_evtchn_to_irqhandler(info->evtchn,
 552                                         blkif_interrupt,
 553                                         IRQF_SAMPLE_RANDOM, "blkif", info);
 554         if (err <= 0) {
 555                 xenbus_dev_fatal(dev, err,
 556                                  "bind_evtchn_to_irqhandler failed");
 557                 goto fail;
 558         }
 559         info->irq = err;
 560
 561         return 0;
 562 fail:
 563         blkif_free(info, 0);
 564         return err;
 565 }
 566
 567
 568 /* Common code used when first setting up, and when resuming. */
 569 static int talk_to_backend(struct xenbus_device *dev,
 570                            struct blkfront_info *info)
 571 {
 572         const char *message = NULL;
 573         struct xenbus_transaction xbt;
 574         int err;
 575
 576         /* Create shared ring, alloc event channel. */
 577         err = setup_blkring(dev, info);
 578         if (err)
 579                 goto out;
 580
 581 again:
 582         err = xenbus_transaction_start(&xbt);
 583         if (err) {
 584                 xenbus_dev_fatal(dev, err, "starting transaction");
 585                 goto destroy_blkring;
 586         }
 587
 588         err = xenbus_printf(xbt, dev->nodename,
 589                             "ring-ref", "%u", info->ring_ref);
 590         if (err) {
 591                 message = "writing ring-ref";
 592                 goto abort_transaction;
 593         }
 594         err = xenbus_printf(xbt, dev->nodename,
 595                             "event-channel", "%u", info->evtchn);
 596         if (err) {
 597                 message = "writing event-channel";
 598                 goto abort_transaction;
 599         }
 600
 601         err = xenbus_transaction_end(xbt, 0);
 602         if (err) {
 603                 if (err == -EAGAIN)
 604                         goto again;
 605                 xenbus_dev_fatal(dev, err, "completing transaction");
 606                 goto destroy_blkring;
 607         }
 608
 609         xenbus_switch_state(dev, XenbusStateInitialised);
 610
 611         return 0;
 612
 613  abort_transaction:
 614         xenbus_transaction_end(xbt, 1);
 615         if (message)
 616                 xenbus_dev_fatal(dev, err, "%s", message);
 617  destroy_blkring:
 618         blkif_free(info, 0);
 619  out:
 620         return err;
 621 }
 622
 623
 624 /**
 625  * Entry point to this code when a new device is created.  Allocate the basic
 626  * structures and the ring buffer for communication with the backend, and
 627  * inform the backend of the appropriate details for those.  Switch to
 628  * Initialised state.
 629  */
 630 static int blkfront_probe(struct xenbus_device *dev,
 631                           const struct xenbus_device_id *id)
 632 {
 633         int err, vdevice, i;
 634         struct blkfront_info *info;
 635
 636         /* FIXME: Use dynamic device id if this is not set. */
 637         err = xenbus_scanf(XBT_NIL, dev->nodename,
 638                            "virtual-device", "%i", &vdevice);
 639         if (err != 1) {
 640                 xenbus_dev_fatal(dev, err, "reading virtual-device");
 641                 return err;
 642         }
 643
 644         info = kzalloc(sizeof(*info), GFP_KERNEL);
 645         if (!info) {
 646                 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
 647                 return -ENOMEM;
 648         }
 649
 650         info->xbdev = dev;
 651         info->vdevice = vdevice;
 652         info->connected = BLKIF_STATE_DISCONNECTED;
 653         INIT_WORK(&info->work, blkif_restart_queue);
 654
 655         for (i = 0; i < BLK_RING_SIZE; i++)
 656                 info->shadow[i].req.id = i+1;
 657         info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
 658
 659         /* Front end dir is a number, which is used as the id. */
 660         info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
 661         dev->dev.driver_data = info;
 662
 663         err = talk_to_backend(dev, info);
 664         if (err) {
 665                 kfree(info);
 666                 dev->dev.driver_data = NULL;
 667                 return err;
 668         }
 669
 670         return 0;
 671 }
 672
 673
 674 static int blkif_recover(struct blkfront_info *info)
 675 {
 676         int i;
 677         struct blkif_request *req;
 678         struct blk_shadow *copy;
 679         int j;
 680
 681         /* Stage 1: Make a safe copy of the shadow state. */
 682         copy = kmalloc(sizeof(info->shadow), GFP_KERNEL);
 683         if (!copy)
 684                 return -ENOMEM;
 685         memcpy(copy, info->shadow, sizeof(info->shadow));
 686
 687         /* Stage 2: Set up free list. */
 688         memset(&info->shadow, 0, sizeof(info->shadow));
 689         for (i = 0; i < BLK_RING_SIZE; i++)
 690                 info->shadow[i].req.id = i+1;
 691         info->shadow_free = info->ring.req_prod_pvt;
 692         info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
 693
 694         /* Stage 3: Find pending requests and requeue them. */
 695         for (i = 0; i < BLK_RING_SIZE; i++) {
 696                 /* Not in use? */
 697                 if (copy[i].request == 0)
 698                         continue;
 699
 700                 /* Grab a request slot and copy shadow state into it. */
 701                 req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
 702                 *req = copy[i].req;
 703
 704                 /* We get a new request id, and must reset the shadow state. */
 705                 req->id = get_id_from_freelist(info);
 706                 memcpy(&info->shadow[req->id], &copy[i], sizeof(copy[i]));
 707
 708                 /* Rewrite any grant references invalidated by susp/resume. */
 709                 for (j = 0; j < req->nr_segments; j++)
 710                         gnttab_grant_foreign_access_ref(
 711                                 req->seg[j].gref,
 712                                 info->xbdev->otherend_id,
 713                                 pfn_to_mfn(info->shadow[req->id].frame[j]),
 714                                 rq_data_dir(
 715                                         (struct request *)
 716                                         info->shadow[req->id].request));
 717                 info->shadow[req->id].req = *req;
 718
 719                 info->ring.req_prod_pvt++;
 720         }
 721
 722         kfree(copy);
 723
 724         xenbus_switch_state(info->xbdev, XenbusStateConnected);
 725
 726         spin_lock_irq(&blkif_io_lock);
 727
 728         /* Now safe for us to use the shared ring */
 729         info->connected = BLKIF_STATE_CONNECTED;
 730
 731         /* Send off requeued requests */
 732         flush_requests(info);
 733
 734         /* Kick any other new requests queued since we resumed */
 735         kick_pending_request_queues(info);
 736
 737         spin_unlock_irq(&blkif_io_lock);
 738
 739         return 0;
 740 }
 741
 742 /**
 743  * We are reconnecting to the backend, due to a suspend/resume, or a backend
 744  * driver restart.  We tear down our blkif structure and recreate it, but
 745  * leave the device-layer structures intact so that this is transparent to the
 746  * rest of the kernel.
 747  */
 748 static int blkfront_resume(struct xenbus_device *dev)
 749 {
 750         struct blkfront_info *info = dev->dev.driver_data;
 751         int err;
 752
 753         dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename);
 754
 755         blkif_free(info, info->connected == BLKIF_STATE_CONNECTED);
 756
 757         err = talk_to_backend(dev, info);
 758         if (info->connected == BLKIF_STATE_SUSPENDED && !err)
 759                 err = blkif_recover(info);
 760
 761         return err;
 762 }
 763
 764
 765 /*
 766  * Invoked when the backend is finally 'ready' (and has told produced
 767  * the details about the physical device - #sectors, size, etc).
 768  */
 769 static void blkfront_connect(struct blkfront_info *info)
 770 {
 771         unsigned long long sectors;
 772         unsigned long sector_size;
 773         unsigned int binfo;
 774         int err;
 775
 776         if ((info->connected == BLKIF_STATE_CONNECTED) ||
 777             (info->connected == BLKIF_STATE_SUSPENDED) )
 778                 return;
 779
 780         dev_dbg(&info->xbdev->dev, "%s:%s.\n",
 781                 __func__, info->xbdev->otherend);
 782
 783         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
 784                             "sectors", "%llu", &sectors,
 785                             "info", "%u", &binfo,
 786                             "sector-size", "%lu", &sector_size,
 787                             NULL);
 788         if (err) {
 789                 xenbus_dev_fatal(info->xbdev, err,
 790                                  "reading backend fields at %s",
 791                                  info->xbdev->otherend);
 792                 return;
 793         }
 794
 795         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
 796                             "feature-barrier", "%lu", &info->feature_barrier,
 797                             NULL);
 798         if (err)
 799                 info->feature_barrier = 0;
 800
 801         err = xlvbd_alloc_gendisk(BLKIF_MINOR(info->vdevice),
 802                                   sectors, info->vdevice,
 803                                   binfo, sector_size, info);
 804         if (err) {
 805                 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s",
 806                                  info->xbdev->otherend);
 807                 return;
 808         }
 809
 810         xenbus_switch_state(info->xbdev, XenbusStateConnected);
 811
 812         /* Kick pending requests. */
 813         spin_lock_irq(&blkif_io_lock);
 814         info->connected = BLKIF_STATE_CONNECTED;
 815         kick_pending_request_queues(info);
 816         spin_unlock_irq(&blkif_io_lock);
 817
 818         add_disk(info->gd);
 819 }
 820
 821 /**
 822  * Handle the change of state of the backend to Closing.  We must delete our
 823  * device-layer structures now, to ensure that writes are flushed through to
 824  * the backend.  Once is this done, we can switch to Closed in
 825  * acknowledgement.
 826  */
 827 static void blkfront_closing(struct xenbus_device *dev)
 828 {
 829         struct blkfront_info *info = dev->dev.driver_data;
 830         unsigned long flags;
 831
 832         dev_dbg(&dev->dev, "blkfront_closing: %s removed\n", dev->nodename);
 833
 834         if (info->rq == NULL)
 835                 goto out;
 836
 837         spin_lock_irqsave(&blkif_io_lock, flags);
 838
 839         del_gendisk(info->gd);
 840
 841         /* No more blkif_request(). */
 842         blk_stop_queue(info->rq);
 843
 844         /* No more gnttab callback work. */
 845         gnttab_cancel_free_callback(&info->callback);
 846         spin_unlock_irqrestore(&blkif_io_lock, flags);
 847
 848         /* Flush gnttab callback work. Must be done with no locks held. */
 849         flush_scheduled_work();
 850
 851         blk_cleanup_queue(info->rq);
 852         info->rq = NULL;
 853
 854  out:
 855         xenbus_frontend_closed(dev);
 856 }
 857
 858 /**
 859  * Callback received when the backend's state changes.
 860  */
 861 static void backend_changed(struct xenbus_device *dev,
 862                             enum xenbus_state backend_state)
 863 {
 864         struct blkfront_info *info = dev->dev.driver_data;
 865         struct block_device *bd;
 866
 867         dev_dbg(&dev->dev, "blkfront:backend_changed.\n");
 868
 869         switch (backend_state) {
 870         case XenbusStateInitialising:
 871         case XenbusStateInitWait:
 872         case XenbusStateInitialised:
 873         case XenbusStateUnknown:
 874         case XenbusStateClosed:
 875                 break;
 876
 877         case XenbusStateConnected:
 878                 blkfront_connect(info);
 879                 break;
 880
 881         case XenbusStateClosing:
 882                 bd = bdget(info->dev);
 883                 if (bd == NULL)
 884                         xenbus_dev_fatal(dev, -ENODEV, "bdget failed");
 885
 886                 mutex_lock(&bd->bd_mutex);
 887                 if (info->users > 0)
 888                         xenbus_dev_error(dev, -EBUSY,
 889                                          "Device in use; refusing to close");
 890                 else
 891                         blkfront_closing(dev);
 892                 mutex_unlock(&bd->bd_mutex);
 893                 bdput(bd);
 894                 break;
 895         }
 896 }
 897
 898 static int blkfront_remove(struct xenbus_device *dev)
 899 {
 900         struct blkfront_info *info = dev->dev.driver_data;
 901
 902         dev_dbg(&dev->dev, "blkfront_remove: %s removed\n", dev->nodename);
 903
 904         blkif_free(info, 0);
 905
 906         kfree(info);
 907
 908         return 0;
 909 }
 910
 911 static int blkif_open(struct inode *inode, struct file *filep)
 912 {
 913         struct blkfront_info *info = inode->i_bdev->bd_disk->private_data;
 914         info->users++;
 915         return 0;
 916 }
 917
 918 static int blkif_release(struct inode *inode, struct file *filep)
 919 {
 920         struct blkfront_info *info = inode->i_bdev->bd_disk->private_data;
 921         info->users--;
 922         if (info->users == 0) {
 923                 /* Check whether we have been instructed to close.  We will
 924                    have ignored this request initially, as the device was
 925                    still mounted. */
 926                 struct xenbus_device *dev = info->xbdev;
 927                 enum xenbus_state state = xenbus_read_driver_state(dev->otherend);
 928
 929                 if (state == XenbusStateClosing)
 930                         blkfront_closing(dev);
 931         }
 932         return 0;
 933 }
 934
 935 static struct block_device_operations xlvbd_block_fops =
 936 {
 937         .owner = THIS_MODULE,
 938         .open = blkif_open,
 939         .release = blkif_release,
 940 };
 941
 942
 943 static struct xenbus_device_id blkfront_ids[] = {
 944         { "vbd" },
 945         { "" }
 946 };
 947
 948 static struct xenbus_driver blkfront = {
 949         .name = "vbd",
 950         .owner = THIS_MODULE,
 951         .ids = blkfront_ids,
 952         .probe = blkfront_probe,
 953         .remove = blkfront_remove,
 954         .resume = blkfront_resume,
 955         .otherend_changed = backend_changed,
 956 };
 957
 958 static int __init xlblk_init(void)
 959 {
 960         if (!is_running_on_xen())
 961                 return -ENODEV;
 962
 963         if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) {
 964                 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n",
 965                        XENVBD_MAJOR, DEV_NAME);
 966                 return -ENODEV;
 967         }
 968
 969         return xenbus_register_frontend(&blkfront);
 970 }
 971 module_init(xlblk_init);
 972
 973
 974 static void xlblk_exit(void)
 975 {
 976         return xenbus_unregister_driver(&blkfront);
 977 }
 978 module_exit(xlblk_exit);
 979
 980 MODULE_DESCRIPTION("Xen virtual block device frontend");
 981 MODULE_LICENSE("GPL");
 982 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR);