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