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/hdreg.h>
  41 #include <linux/cdrom.h>
  42 #include <linux/module.h>
  43
  44 #include <xen/xenbus.h>
  45 #include <xen/grant_table.h>
  46 #include <xen/events.h>
  47 #include <xen/page.h>
  48
  49 #include <xen/interface/grant_table.h>
  50 #include <xen/interface/io/blkif.h>
  51 #include <xen/interface/io/protocols.h>
  52
  53 #include <asm/xen/hypervisor.h>
  54
  55 enum blkif_state {
  56         BLKIF_STATE_DISCONNECTED,
  57         BLKIF_STATE_CONNECTED,
  58         BLKIF_STATE_SUSPENDED,
  59 };
  60
  61 struct blk_shadow {
  62         struct blkif_request req;
  63         unsigned long request;
  64         unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
  65 };
  66
  67 static struct block_device_operations xlvbd_block_fops;
  68
  69 #define BLK_RING_SIZE __RING_SIZE((struct blkif_sring *)0, PAGE_SIZE)
  70
  71 /*
  72  * We have one of these per vbd, whether ide, scsi or 'other'.  They
  73  * hang in private_data off the gendisk structure. We may end up
  74  * putting all kinds of interesting stuff here :-)
  75  */
  76 struct blkfront_info
  77 {
  78         struct xenbus_device *xbdev;
  79         struct gendisk *gd;
  80         int vdevice;
  81         blkif_vdev_t handle;
  82         enum blkif_state connected;
  83         int ring_ref;
  84         struct blkif_front_ring ring;
  85         unsigned int evtchn, irq;
  86         struct request_queue *rq;
  87         struct work_struct work;
  88         struct gnttab_free_callback callback;
  89         struct blk_shadow shadow[BLK_RING_SIZE];
  90         unsigned long shadow_free;
  91         int feature_barrier;
  92         int is_ready;
  93
  94         /**
  95          * The number of people holding this device open.  We won't allow a
  96          * hot-unplug unless this is 0.
  97          */
  98         int users;
  99 };
 100
 101 static DEFINE_SPINLOCK(blkif_io_lock);
 102
 103 #define MAXIMUM_OUTSTANDING_BLOCK_REQS \
 104         (BLKIF_MAX_SEGMENTS_PER_REQUEST * BLK_RING_SIZE)
 105 #define GRANT_INVALID_REF       0
 106
 107 #define PARTS_PER_DISK          16
 108 #define PARTS_PER_EXT_DISK      256
 109
 110 #define BLKIF_MAJOR(dev) ((dev)>>8)
 111 #define BLKIF_MINOR(dev) ((dev) & 0xff)
 112
 113 #define EXT_SHIFT 28
 114 #define EXTENDED (1<<EXT_SHIFT)
 115 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED))
 116 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED))
 117
 118 #define DEV_NAME        "xvd"   /* name in /dev */
 119
 120 static int get_id_from_freelist(struct blkfront_info *info)
 121 {
 122         unsigned long free = info->shadow_free;
 123         BUG_ON(free > BLK_RING_SIZE);
 124         info->shadow_free = info->shadow[free].req.id;
 125         info->shadow[free].req.id = 0x0fffffee; /* debug */
 126         return free;
 127 }
 128
 129 static void add_id_to_freelist(struct blkfront_info *info,
 130                                unsigned long id)
 131 {
 132         info->shadow[id].req.id  = info->shadow_free;
 133         info->shadow[id].request = 0;
 134         info->shadow_free = id;
 135 }
 136
 137 static void blkif_restart_queue_callback(void *arg)
 138 {
 139         struct blkfront_info *info = (struct blkfront_info *)arg;
 140         schedule_work(&info->work);
 141 }
 142
 143 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg)
 144 {
 145         /* We don't have real geometry info, but let's at least return
 146            values consistent with the size of the device */
 147         sector_t nsect = get_capacity(bd->bd_disk);
 148         sector_t cylinders = nsect;
 149
 150         hg->heads = 0xff;
 151         hg->sectors = 0x3f;
 152         sector_div(cylinders, hg->heads * hg->sectors);
 153         hg->cylinders = cylinders;
 154         if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect)
 155                 hg->cylinders = 0xffff;
 156         return 0;
 157 }
 158
 159 static int blkif_ioctl(struct block_device *bdev, fmode_t mode,
 160                        unsigned command, unsigned long argument)
 161 {
 162         struct blkfront_info *info = bdev->bd_disk->private_data;
 163         int i;
 164
 165         dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n",
 166                 command, (long)argument);
 167
 168         switch (command) {
 169         case CDROMMULTISESSION:
 170                 dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n");
 171                 for (i = 0; i < sizeof(struct cdrom_multisession); i++)
 172                         if (put_user(0, (char __user *)(argument + i)))
 173                                 return -EFAULT;
 174                 return 0;
 175
 176         case CDROM_GET_CAPABILITY: {
 177                 struct gendisk *gd = info->gd;
 178                 if (gd->flags & GENHD_FL_CD)
 179                         return 0;
 180                 return -EINVAL;
 181         }
 182
 183         default:
 184                 /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n",
 185                   command);*/
 186                 return -EINVAL; /* same return as native Linux */
 187         }
 188
 189         return 0;
 190 }
 191
 192 /*
 193  * blkif_queue_request
 194  *
 195  * request block io
 196  *
 197  * id: for guest use only.
 198  * operation: BLKIF_OP_{READ,WRITE,PROBE}
 199  * buffer: buffer to read/write into. this should be a
 200  *   virtual address in the guest os.
 201  */
 202 static int blkif_queue_request(struct request *req)
 203 {
 204         struct blkfront_info *info = req->rq_disk->private_data;
 205         unsigned long buffer_mfn;
 206         struct blkif_request *ring_req;
 207         struct req_iterator iter;
 208         struct bio_vec *bvec;
 209         unsigned long id;
 210         unsigned int fsect, lsect;
 211         int ref;
 212         grant_ref_t gref_head;
 213
 214         if (unlikely(info->connected != BLKIF_STATE_CONNECTED))
 215                 return 1;
 216
 217         if (gnttab_alloc_grant_references(
 218                 BLKIF_MAX_SEGMENTS_PER_REQUEST, &gref_head) < 0) {
 219                 gnttab_request_free_callback(
 220                         &info->callback,
 221                         blkif_restart_queue_callback,
 222                         info,
 223                         BLKIF_MAX_SEGMENTS_PER_REQUEST);
 224                 return 1;
 225         }
 226
 227         /* Fill out a communications ring structure. */
 228         ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
 229         id = get_id_from_freelist(info);
 230         info->shadow[id].request = (unsigned long)req;
 231
 232         ring_req->id = id;
 233         ring_req->sector_number = (blkif_sector_t)req->sector;
 234         ring_req->handle = info->handle;
 235
 236         ring_req->operation = rq_data_dir(req) ?
 237                 BLKIF_OP_WRITE : BLKIF_OP_READ;
 238         if (blk_barrier_rq(req))
 239                 ring_req->operation = BLKIF_OP_WRITE_BARRIER;
 240
 241         ring_req->nr_segments = 0;
 242         rq_for_each_segment(bvec, req, iter) {
 243                 BUG_ON(ring_req->nr_segments == BLKIF_MAX_SEGMENTS_PER_REQUEST);
 244                 buffer_mfn = pfn_to_mfn(page_to_pfn(bvec->bv_page));
 245                 fsect = bvec->bv_offset >> 9;
 246                 lsect = fsect + (bvec->bv_len >> 9) - 1;
 247                 /* install a grant reference. */
 248                 ref = gnttab_claim_grant_reference(&gref_head);
 249                 BUG_ON(ref == -ENOSPC);
 250
 251                 gnttab_grant_foreign_access_ref(
 252                                 ref,
 253                                 info->xbdev->otherend_id,
 254                                 buffer_mfn,
 255                                 rq_data_dir(req) );
 256
 257                 info->shadow[id].frame[ring_req->nr_segments] =
 258                                 mfn_to_pfn(buffer_mfn);
 259
 260                 ring_req->seg[ring_req->nr_segments] =
 261                                 (struct blkif_request_segment) {
 262                                         .gref       = ref,
 263                                         .first_sect = fsect,
 264                                         .last_sect  = lsect };
 265
 266                 ring_req->nr_segments++;
 267         }
 268
 269         info->ring.req_prod_pvt++;
 270
 271         /* Keep a private copy so we can reissue requests when recovering. */
 272         info->shadow[id].req = *ring_req;
 273
 274         gnttab_free_grant_references(gref_head);
 275
 276         return 0;
 277 }
 278
 279
 280 static inline void flush_requests(struct blkfront_info *info)
 281 {
 282         int notify;
 283
 284         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&info->ring, notify);
 285
 286         if (notify)
 287                 notify_remote_via_irq(info->irq);
 288 }
 289
 290 /*
 291  * do_blkif_request
 292  *  read a block; request is in a request queue
 293  */
 294 static void do_blkif_request(struct request_queue *rq)
 295 {
 296         struct blkfront_info *info = NULL;
 297         struct request *req;
 298         int queued;
 299
 300         pr_debug("Entered do_blkif_request\n");
 301
 302         queued = 0;
 303
 304         while ((req = elv_next_request(rq)) != NULL) {
 305                 info = req->rq_disk->private_data;
 306                 if (!blk_fs_request(req)) {
 307                         end_request(req, 0);
 308                         continue;
 309                 }
 310
 311                 if (RING_FULL(&info->ring))
 312                         goto wait;
 313
 314                 pr_debug("do_blk_req %p: cmd %p, sec %lx, "
 315                          "(%u/%li) buffer:%p [%s]\n",
 316                          req, req->cmd, (unsigned long)req->sector,
 317                          req->current_nr_sectors,
 318                          req->nr_sectors, req->buffer,
 319                          rq_data_dir(req) ? "write" : "read");
 320
 321
 322                 blkdev_dequeue_request(req);
 323                 if (blkif_queue_request(req)) {
 324                         blk_requeue_request(rq, req);
 325 wait:
 326                         /* Avoid pointless unplugs. */
 327                         blk_stop_queue(rq);
 328                         break;
 329                 }
 330
 331                 queued++;
 332         }
 333
 334         if (queued != 0)
 335                 flush_requests(info);
 336 }
 337
 338 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size)
 339 {
 340         struct request_queue *rq;
 341         elevator_t *old_e;
 342
 343         rq = blk_init_queue(do_blkif_request, &blkif_io_lock);
 344         if (rq == NULL)
 345                 return -1;
 346
 347         old_e = rq->elevator;
 348         if (IS_ERR_VALUE(elevator_init(rq, "noop")))
 349                 printk(KERN_WARNING
 350                         "blkfront: Switch elevator failed, use default\n");
 351         else
 352                 elevator_exit(old_e);
 353
 354         /* Hard sector size and max sectors impersonate the equiv. hardware. */
 355         blk_queue_hardsect_size(rq, sector_size);
 356         blk_queue_max_sectors(rq, 512);
 357
 358         /* Each segment in a request is up to an aligned page in size. */
 359         blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
 360         blk_queue_max_segment_size(rq, PAGE_SIZE);
 361
 362         /* Ensure a merged request will fit in a single I/O ring slot. */
 363         blk_queue_max_phys_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
 364         blk_queue_max_hw_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
 365
 366         /* Make sure buffer addresses are sector-aligned. */
 367         blk_queue_dma_alignment(rq, 511);
 368
 369         /* Make sure we don't use bounce buffers. */
 370         blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY);
 371
 372         gd->queue = rq;
 373
 374         return 0;
 375 }
 376
 377
 378 static int xlvbd_barrier(struct blkfront_info *info)
 379 {
 380         int err;
 381
 382         err = blk_queue_ordered(info->rq,
 383                                 info->feature_barrier ? QUEUE_ORDERED_DRAIN : QUEUE_ORDERED_NONE,
 384                                 NULL);
 385
 386         if (err)
 387                 return err;
 388
 389         printk(KERN_INFO "blkfront: %s: barriers %s\n",
 390                info->gd->disk_name,
 391                info->feature_barrier ? "enabled" : "disabled");
 392         return 0;
 393 }
 394
 395
 396 static int xlvbd_alloc_gendisk(blkif_sector_t capacity,
 397                                struct blkfront_info *info,
 398                                u16 vdisk_info, u16 sector_size)
 399 {
 400         struct gendisk *gd;
 401         int nr_minors = 1;
 402         int err = -ENODEV;
 403         unsigned int offset;
 404         int minor;
 405         int nr_parts;
 406
 407         BUG_ON(info->gd != NULL);
 408         BUG_ON(info->rq != NULL);
 409
 410         if ((info->vdevice>>EXT_SHIFT) > 1) {
 411                 /* this is above the extended range; something is wrong */
 412                 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice);
 413                 return -ENODEV;
 414         }
 415
 416         if (!VDEV_IS_EXTENDED(info->vdevice)) {
 417                 minor = BLKIF_MINOR(info->vdevice);
 418                 nr_parts = PARTS_PER_DISK;
 419         } else {
 420                 minor = BLKIF_MINOR_EXT(info->vdevice);
 421                 nr_parts = PARTS_PER_EXT_DISK;
 422         }
 423
 424         if ((minor % nr_parts) == 0)
 425                 nr_minors = nr_parts;
 426
 427         gd = alloc_disk(nr_minors);
 428         if (gd == NULL)
 429                 goto out;
 430
 431         offset = minor / nr_parts;
 432
 433         if (nr_minors > 1) {
 434                 if (offset < 26)
 435                         sprintf(gd->disk_name, "%s%c", DEV_NAME, 'a' + offset);
 436                 else
 437                         sprintf(gd->disk_name, "%s%c%c", DEV_NAME,
 438                                 'a' + ((offset / 26)-1), 'a' + (offset % 26));
 439         } else {
 440                 if (offset < 26)
 441                         sprintf(gd->disk_name, "%s%c%d", DEV_NAME,
 442                                 'a' + offset,
 443                                 minor & (nr_parts - 1));
 444                 else
 445                         sprintf(gd->disk_name, "%s%c%c%d", DEV_NAME,
 446                                 'a' + ((offset / 26) - 1),
 447                                 'a' + (offset % 26),
 448                                 minor & (nr_parts - 1));
 449         }
 450
 451         gd->major = XENVBD_MAJOR;
 452         gd->first_minor = minor;
 453         gd->fops = &xlvbd_block_fops;
 454         gd->private_data = info;
 455         gd->driverfs_dev = &(info->xbdev->dev);
 456         set_capacity(gd, capacity);
 457
 458         if (xlvbd_init_blk_queue(gd, sector_size)) {
 459                 del_gendisk(gd);
 460                 goto out;
 461         }
 462
 463         info->rq = gd->queue;
 464         info->gd = gd;
 465
 466         if (info->feature_barrier)
 467                 xlvbd_barrier(info);
 468
 469         if (vdisk_info & VDISK_READONLY)
 470                 set_disk_ro(gd, 1);
 471
 472         if (vdisk_info & VDISK_REMOVABLE)
 473                 gd->flags |= GENHD_FL_REMOVABLE;
 474
 475         if (vdisk_info & VDISK_CDROM)
 476                 gd->flags |= GENHD_FL_CD;
 477
 478         return 0;
 479
 480  out:
 481         return err;
 482 }
 483
 484 static void kick_pending_request_queues(struct blkfront_info *info)
 485 {
 486         if (!RING_FULL(&info->ring)) {
 487                 /* Re-enable calldowns. */
 488                 blk_start_queue(info->rq);
 489                 /* Kick things off immediately. */
 490                 do_blkif_request(info->rq);
 491         }
 492 }
 493
 494 static void blkif_restart_queue(struct work_struct *work)
 495 {
 496         struct blkfront_info *info = container_of(work, struct blkfront_info, work);
 497
 498         spin_lock_irq(&blkif_io_lock);
 499         if (info->connected == BLKIF_STATE_CONNECTED)
 500                 kick_pending_request_queues(info);
 501         spin_unlock_irq(&blkif_io_lock);
 502 }
 503
 504 static void blkif_free(struct blkfront_info *info, int suspend)
 505 {
 506         /* Prevent new requests being issued until we fix things up. */
 507         spin_lock_irq(&blkif_io_lock);
 508         info->connected = suspend ?
 509                 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED;
 510         /* No more blkif_request(). */
 511         if (info->rq)
 512                 blk_stop_queue(info->rq);
 513         /* No more gnttab callback work. */
 514         gnttab_cancel_free_callback(&info->callback);
 515         spin_unlock_irq(&blkif_io_lock);
 516
 517         /* Flush gnttab callback work. Must be done with no locks held. */
 518         flush_scheduled_work();
 519
 520         /* Free resources associated with old device channel. */
 521         if (info->ring_ref != GRANT_INVALID_REF) {
 522                 gnttab_end_foreign_access(info->ring_ref, 0,
 523                                           (unsigned long)info->ring.sring);
 524                 info->ring_ref = GRANT_INVALID_REF;
 525                 info->ring.sring = NULL;
 526         }
 527         if (info->irq)
 528                 unbind_from_irqhandler(info->irq, info);
 529         info->evtchn = info->irq = 0;
 530
 531 }
 532
 533 static void blkif_completion(struct blk_shadow *s)
 534 {
 535         int i;
 536         for (i = 0; i < s->req.nr_segments; i++)
 537                 gnttab_end_foreign_access(s->req.seg[i].gref, 0, 0UL);
 538 }
 539
 540 static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 541 {
 542         struct request *req;
 543         struct blkif_response *bret;
 544         RING_IDX i, rp;
 545         unsigned long flags;
 546         struct blkfront_info *info = (struct blkfront_info *)dev_id;
 547         int error;
 548
 549         spin_lock_irqsave(&blkif_io_lock, flags);
 550
 551         if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) {
 552                 spin_unlock_irqrestore(&blkif_io_lock, flags);
 553                 return IRQ_HANDLED;
 554         }
 555
 556  again:
 557         rp = info->ring.sring->rsp_prod;
 558         rmb(); /* Ensure we see queued responses up to 'rp'. */
 559
 560         for (i = info->ring.rsp_cons; i != rp; i++) {
 561                 unsigned long id;
 562                 int ret;
 563
 564                 bret = RING_GET_RESPONSE(&info->ring, i);
 565                 id   = bret->id;
 566                 req  = (struct request *)info->shadow[id].request;
 567
 568                 blkif_completion(&info->shadow[id]);
 569
 570                 add_id_to_freelist(info, id);
 571
 572                 error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO;
 573                 switch (bret->operation) {
 574                 case BLKIF_OP_WRITE_BARRIER:
 575                         if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
 576                                 printk(KERN_WARNING "blkfront: %s: write barrier op failed\n",
 577                                        info->gd->disk_name);
 578                                 error = -EOPNOTSUPP;
 579                                 info->feature_barrier = 0;
 580                                 xlvbd_barrier(info);
 581                         }
 582                         /* fall through */
 583                 case BLKIF_OP_READ:
 584                 case BLKIF_OP_WRITE:
 585                         if (unlikely(bret->status != BLKIF_RSP_OKAY))
 586                                 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data "
 587                                         "request: %x\n", bret->status);
 588
 589                         ret = __blk_end_request(req, error, blk_rq_bytes(req));
 590                         BUG_ON(ret);
 591                         break;
 592                 default:
 593                         BUG();
 594                 }
 595         }
 596
 597         info->ring.rsp_cons = i;
 598
 599         if (i != info->ring.req_prod_pvt) {
 600                 int more_to_do;
 601                 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
 602                 if (more_to_do)
 603                         goto again;
 604         } else
 605                 info->ring.sring->rsp_event = i + 1;
 606
 607         kick_pending_request_queues(info);
 608
 609         spin_unlock_irqrestore(&blkif_io_lock, flags);
 610
 611         return IRQ_HANDLED;
 612 }
 613
 614
 615 static int setup_blkring(struct xenbus_device *dev,
 616                          struct blkfront_info *info)
 617 {
 618         struct blkif_sring *sring;
 619         int err;
 620
 621         info->ring_ref = GRANT_INVALID_REF;
 622
 623         sring = (struct blkif_sring *)__get_free_page(GFP_NOIO | __GFP_HIGH);
 624         if (!sring) {
 625                 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring");
 626                 return -ENOMEM;
 627         }
 628         SHARED_RING_INIT(sring);
 629         FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE);
 630
 631         err = xenbus_grant_ring(dev, virt_to_mfn(info->ring.sring));
 632         if (err < 0) {
 633                 free_page((unsigned long)sring);
 634                 info->ring.sring = NULL;
 635                 goto fail;
 636         }
 637         info->ring_ref = err;
 638
 639         err = xenbus_alloc_evtchn(dev, &info->evtchn);
 640         if (err)
 641                 goto fail;
 642
 643         err = bind_evtchn_to_irqhandler(info->evtchn,
 644                                         blkif_interrupt,
 645                                         IRQF_SAMPLE_RANDOM, "blkif", info);
 646         if (err <= 0) {
 647                 xenbus_dev_fatal(dev, err,
 648                                  "bind_evtchn_to_irqhandler failed");
 649                 goto fail;
 650         }
 651         info->irq = err;
 652
 653         return 0;
 654 fail:
 655         blkif_free(info, 0);
 656         return err;
 657 }
 658
 659
 660 /* Common code used when first setting up, and when resuming. */
 661 static int talk_to_backend(struct xenbus_device *dev,
 662                            struct blkfront_info *info)
 663 {
 664         const char *message = NULL;
 665         struct xenbus_transaction xbt;
 666         int err;
 667
 668         /* Create shared ring, alloc event channel. */
 669         err = setup_blkring(dev, info);
 670         if (err)
 671                 goto out;
 672
 673 again:
 674         err = xenbus_transaction_start(&xbt);
 675         if (err) {
 676                 xenbus_dev_fatal(dev, err, "starting transaction");
 677                 goto destroy_blkring;
 678         }
 679
 680         err = xenbus_printf(xbt, dev->nodename,
 681                             "ring-ref", "%u", info->ring_ref);
 682         if (err) {
 683                 message = "writing ring-ref";
 684                 goto abort_transaction;
 685         }
 686         err = xenbus_printf(xbt, dev->nodename,
 687                             "event-channel", "%u", info->evtchn);
 688         if (err) {
 689                 message = "writing event-channel";
 690                 goto abort_transaction;
 691         }
 692         err = xenbus_printf(xbt, dev->nodename, "protocol", "%s",
 693                             XEN_IO_PROTO_ABI_NATIVE);
 694         if (err) {
 695                 message = "writing protocol";
 696                 goto abort_transaction;
 697         }
 698
 699         err = xenbus_transaction_end(xbt, 0);
 700         if (err) {
 701                 if (err == -EAGAIN)
 702                         goto again;
 703                 xenbus_dev_fatal(dev, err, "completing transaction");
 704                 goto destroy_blkring;
 705         }
 706
 707         xenbus_switch_state(dev, XenbusStateInitialised);
 708
 709         return 0;
 710
 711  abort_transaction:
 712         xenbus_transaction_end(xbt, 1);
 713         if (message)
 714                 xenbus_dev_fatal(dev, err, "%s", message);
 715  destroy_blkring:
 716         blkif_free(info, 0);
 717  out:
 718         return err;
 719 }
 720
 721
 722 /**
 723  * Entry point to this code when a new device is created.  Allocate the basic
 724  * structures and the ring buffer for communication with the backend, and
 725  * inform the backend of the appropriate details for those.  Switch to
 726  * Initialised state.
 727  */
 728 static int blkfront_probe(struct xenbus_device *dev,
 729                           const struct xenbus_device_id *id)
 730 {
 731         int err, vdevice, i;
 732         struct blkfront_info *info;
 733
 734         /* FIXME: Use dynamic device id if this is not set. */
 735         err = xenbus_scanf(XBT_NIL, dev->nodename,
 736                            "virtual-device", "%i", &vdevice);
 737         if (err != 1) {
 738                 /* go looking in the extended area instead */
 739                 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext",
 740                                    "%i", &vdevice);
 741                 if (err != 1) {
 742                         xenbus_dev_fatal(dev, err, "reading virtual-device");
 743                         return err;
 744                 }
 745         }
 746
 747         info = kzalloc(sizeof(*info), GFP_KERNEL);
 748         if (!info) {
 749                 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
 750                 return -ENOMEM;
 751         }
 752
 753         info->xbdev = dev;
 754         info->vdevice = vdevice;
 755         info->connected = BLKIF_STATE_DISCONNECTED;
 756         INIT_WORK(&info->work, blkif_restart_queue);
 757
 758         for (i = 0; i < BLK_RING_SIZE; i++)
 759                 info->shadow[i].req.id = i+1;
 760         info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
 761
 762         /* Front end dir is a number, which is used as the id. */
 763         info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
 764         dev->dev.driver_data = info;
 765
 766         err = talk_to_backend(dev, info);
 767         if (err) {
 768                 kfree(info);
 769                 dev->dev.driver_data = NULL;
 770                 return err;
 771         }
 772
 773         return 0;
 774 }
 775
 776
 777 static int blkif_recover(struct blkfront_info *info)
 778 {
 779         int i;
 780         struct blkif_request *req;
 781         struct blk_shadow *copy;
 782         int j;
 783
 784         /* Stage 1: Make a safe copy of the shadow state. */
 785         copy = kmalloc(sizeof(info->shadow),
 786                        GFP_NOIO | __GFP_REPEAT | __GFP_HIGH);
 787         if (!copy)
 788                 return -ENOMEM;
 789         memcpy(copy, info->shadow, sizeof(info->shadow));
 790
 791         /* Stage 2: Set up free list. */
 792         memset(&info->shadow, 0, sizeof(info->shadow));
 793         for (i = 0; i < BLK_RING_SIZE; i++)
 794                 info->shadow[i].req.id = i+1;
 795         info->shadow_free = info->ring.req_prod_pvt;
 796         info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
 797
 798         /* Stage 3: Find pending requests and requeue them. */
 799         for (i = 0; i < BLK_RING_SIZE; i++) {
 800                 /* Not in use? */
 801                 if (copy[i].request == 0)
 802                         continue;
 803
 804                 /* Grab a request slot and copy shadow state into it. */
 805                 req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
 806                 *req = copy[i].req;
 807
 808                 /* We get a new request id, and must reset the shadow state. */
 809                 req->id = get_id_from_freelist(info);
 810                 memcpy(&info->shadow[req->id], &copy[i], sizeof(copy[i]));
 811
 812                 /* Rewrite any grant references invalidated by susp/resume. */
 813                 for (j = 0; j < req->nr_segments; j++)
 814                         gnttab_grant_foreign_access_ref(
 815                                 req->seg[j].gref,
 816                                 info->xbdev->otherend_id,
 817                                 pfn_to_mfn(info->shadow[req->id].frame[j]),
 818                                 rq_data_dir(
 819                                         (struct request *)
 820                                         info->shadow[req->id].request));
 821                 info->shadow[req->id].req = *req;
 822
 823                 info->ring.req_prod_pvt++;
 824         }
 825
 826         kfree(copy);
 827
 828         xenbus_switch_state(info->xbdev, XenbusStateConnected);
 829
 830         spin_lock_irq(&blkif_io_lock);
 831
 832         /* Now safe for us to use the shared ring */
 833         info->connected = BLKIF_STATE_CONNECTED;
 834
 835         /* Send off requeued requests */
 836         flush_requests(info);
 837
 838         /* Kick any other new requests queued since we resumed */
 839         kick_pending_request_queues(info);
 840
 841         spin_unlock_irq(&blkif_io_lock);
 842
 843         return 0;
 844 }
 845
 846 /**
 847  * We are reconnecting to the backend, due to a suspend/resume, or a backend
 848  * driver restart.  We tear down our blkif structure and recreate it, but
 849  * leave the device-layer structures intact so that this is transparent to the
 850  * rest of the kernel.
 851  */
 852 static int blkfront_resume(struct xenbus_device *dev)
 853 {
 854         struct blkfront_info *info = dev->dev.driver_data;
 855         int err;
 856
 857         dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename);
 858
 859         blkif_free(info, info->connected == BLKIF_STATE_CONNECTED);
 860
 861         err = talk_to_backend(dev, info);
 862         if (info->connected == BLKIF_STATE_SUSPENDED && !err)
 863                 err = blkif_recover(info);
 864
 865         return err;
 866 }
 867
 868
 869 /*
 870  * Invoked when the backend is finally 'ready' (and has told produced
 871  * the details about the physical device - #sectors, size, etc).
 872  */
 873 static void blkfront_connect(struct blkfront_info *info)
 874 {
 875         unsigned long long sectors;
 876         unsigned long sector_size;
 877         unsigned int binfo;
 878         int err;
 879
 880         if ((info->connected == BLKIF_STATE_CONNECTED) ||
 881             (info->connected == BLKIF_STATE_SUSPENDED) )
 882                 return;
 883
 884         dev_dbg(&info->xbdev->dev, "%s:%s.\n",
 885                 __func__, info->xbdev->otherend);
 886
 887         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
 888                             "sectors", "%llu", &sectors,
 889                             "info", "%u", &binfo,
 890                             "sector-size", "%lu", &sector_size,
 891                             NULL);
 892         if (err) {
 893                 xenbus_dev_fatal(info->xbdev, err,
 894                                  "reading backend fields at %s",
 895                                  info->xbdev->otherend);
 896                 return;
 897         }
 898
 899         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
 900                             "feature-barrier", "%lu", &info->feature_barrier,
 901                             NULL);
 902         if (err)
 903                 info->feature_barrier = 0;
 904
 905         err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size);
 906         if (err) {
 907                 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s",
 908                                  info->xbdev->otherend);
 909                 return;
 910         }
 911
 912         xenbus_switch_state(info->xbdev, XenbusStateConnected);
 913
 914         /* Kick pending requests. */
 915         spin_lock_irq(&blkif_io_lock);
 916         info->connected = BLKIF_STATE_CONNECTED;
 917         kick_pending_request_queues(info);
 918         spin_unlock_irq(&blkif_io_lock);
 919
 920         add_disk(info->gd);
 921
 922         info->is_ready = 1;
 923 }
 924
 925 /**
 926  * Handle the change of state of the backend to Closing.  We must delete our
 927  * device-layer structures now, to ensure that writes are flushed through to
 928  * the backend.  Once is this done, we can switch to Closed in
 929  * acknowledgement.
 930  */
 931 static void blkfront_closing(struct xenbus_device *dev)
 932 {
 933         struct blkfront_info *info = dev->dev.driver_data;
 934         unsigned long flags;
 935
 936         dev_dbg(&dev->dev, "blkfront_closing: %s removed\n", dev->nodename);
 937
 938         if (info->rq == NULL)
 939                 goto out;
 940
 941         spin_lock_irqsave(&blkif_io_lock, flags);
 942
 943         del_gendisk(info->gd);
 944
 945         /* No more blkif_request(). */
 946         blk_stop_queue(info->rq);
 947
 948         /* No more gnttab callback work. */
 949         gnttab_cancel_free_callback(&info->callback);
 950         spin_unlock_irqrestore(&blkif_io_lock, flags);
 951
 952         /* Flush gnttab callback work. Must be done with no locks held. */
 953         flush_scheduled_work();
 954
 955         blk_cleanup_queue(info->rq);
 956         info->rq = NULL;
 957
 958  out:
 959         xenbus_frontend_closed(dev);
 960 }
 961
 962 /**
 963  * Callback received when the backend's state changes.
 964  */
 965 static void backend_changed(struct xenbus_device *dev,
 966                             enum xenbus_state backend_state)
 967 {
 968         struct blkfront_info *info = dev->dev.driver_data;
 969         struct block_device *bd;
 970
 971         dev_dbg(&dev->dev, "blkfront:backend_changed.\n");
 972
 973         switch (backend_state) {
 974         case XenbusStateInitialising:
 975         case XenbusStateInitWait:
 976         case XenbusStateInitialised:
 977         case XenbusStateUnknown:
 978         case XenbusStateClosed:
 979                 break;
 980
 981         case XenbusStateConnected:
 982                 blkfront_connect(info);
 983                 break;
 984
 985         case XenbusStateClosing:
 986                 bd = bdget_disk(info->gd, 0);
 987                 if (bd == NULL)
 988                         xenbus_dev_fatal(dev, -ENODEV, "bdget failed");
 989
 990                 mutex_lock(&bd->bd_mutex);
 991                 if (info->users > 0)
 992                         xenbus_dev_error(dev, -EBUSY,
 993                                          "Device in use; refusing to close");
 994                 else
 995                         blkfront_closing(dev);
 996                 mutex_unlock(&bd->bd_mutex);
 997                 bdput(bd);
 998                 break;
 999         }
1000 }
1001
1002 static int blkfront_remove(struct xenbus_device *dev)
1003 {
1004         struct blkfront_info *info = dev->dev.driver_data;
1005
1006         dev_dbg(&dev->dev, "blkfront_remove: %s removed\n", dev->nodename);
1007
1008         blkif_free(info, 0);
1009
1010         kfree(info);
1011
1012         return 0;
1013 }
1014
1015 static int blkfront_is_ready(struct xenbus_device *dev)
1016 {
1017         struct blkfront_info *info = dev->dev.driver_data;
1018
1019         return info->is_ready;
1020 }
1021
1022 static int blkif_open(struct block_device *bdev, fmode_t mode)
1023 {
1024         struct blkfront_info *info = bdev->bd_disk->private_data;
1025         info->users++;
1026         return 0;
1027 }
1028
1029 static int blkif_release(struct gendisk *disk, fmode_t mode)
1030 {
1031         struct blkfront_info *info = disk->private_data;
1032         info->users--;
1033         if (info->users == 0) {
1034                 /* Check whether we have been instructed to close.  We will
1035                    have ignored this request initially, as the device was
1036                    still mounted. */
1037                 struct xenbus_device *dev = info->xbdev;
1038                 enum xenbus_state state = xenbus_read_driver_state(dev->otherend);
1039
1040                 if (state == XenbusStateClosing && info->is_ready)
1041                         blkfront_closing(dev);
1042         }
1043         return 0;
1044 }
1045
1046 static struct block_device_operations xlvbd_block_fops =
1047 {
1048         .owner = THIS_MODULE,
1049         .open = blkif_open,
1050         .release = blkif_release,
1051         .getgeo = blkif_getgeo,
1052         .locked_ioctl = blkif_ioctl,
1053 };
1054
1055
1056 static struct xenbus_device_id blkfront_ids[] = {
1057         { "vbd" },
1058         { "" }
1059 };
1060
1061 static struct xenbus_driver blkfront = {
1062         .name = "vbd",
1063         .owner = THIS_MODULE,
1064         .ids = blkfront_ids,
1065         .probe = blkfront_probe,
1066         .remove = blkfront_remove,
1067         .resume = blkfront_resume,
1068         .otherend_changed = backend_changed,
1069         .is_ready = blkfront_is_ready,
1070 };
1071
1072 static int __init xlblk_init(void)
1073 {
1074         if (!xen_domain())
1075                 return -ENODEV;
1076
1077         if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) {
1078                 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n",
1079                        XENVBD_MAJOR, DEV_NAME);
1080                 return -ENODEV;
1081         }
1082
1083         return xenbus_register_frontend(&blkfront);
1084 }
1085 module_init(xlblk_init);
1086
1087
1088 static void __exit xlblk_exit(void)
1089 {
1090         return xenbus_unregister_driver(&blkfront);
1091 }
1092 module_exit(xlblk_exit);
1093
1094 MODULE_DESCRIPTION("Xen virtual block device frontend");
1095 MODULE_LICENSE("GPL");
1096 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR);
1097 MODULE_ALIAS("xen:vbd");
1098 MODULE_ALIAS("xenblk");