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