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/slab.h>
  44 #include <linux/mutex.h>
  45 #include <linux/scatterlist.h>
  46 #include <linux/bitmap.h>
  47 #include <linux/list.h>
  48
  49 #include <xen/xen.h>
  50 #include <xen/xenbus.h>
  51 #include <xen/grant_table.h>
  52 #include <xen/events.h>
  53 #include <xen/page.h>
  54 #include <xen/platform_pci.h>
  55
  56 #include <xen/interface/grant_table.h>
  57 #include <xen/interface/io/blkif.h>
  58 #include <xen/interface/io/protocols.h>
  59
  60 #include <asm/xen/hypervisor.h>
  61
  62 enum blkif_state {
  63         BLKIF_STATE_DISCONNECTED,
  64         BLKIF_STATE_CONNECTED,
  65         BLKIF_STATE_SUSPENDED,
  66 };
  67
  68 struct grant {
  69         grant_ref_t gref;
  70         unsigned long pfn;
  71         struct list_head node;
  72 };
  73
  74 struct blk_shadow {
  75         struct blkif_request req;
  76         struct request *request;
  77         struct grant **grants_used;
  78         struct grant **indirect_grants;
  79         struct scatterlist *sg;
  80 };
  81
  82 struct split_bio {
  83         struct bio *bio;
  84         atomic_t pending;
  85         int err;
  86 };
  87
  88 static DEFINE_MUTEX(blkfront_mutex);
  89 static const struct block_device_operations xlvbd_block_fops;
  90
  91 /*
  92  * Maximum number of segments in indirect requests, the actual value used by
  93  * the frontend driver is the minimum of this value and the value provided
  94  * by the backend driver.
  95  */
  96
  97 static unsigned int xen_blkif_max_segments = 32;
  98 module_param_named(max, xen_blkif_max_segments, int, S_IRUGO);
  99 MODULE_PARM_DESC(max, "Maximum amount of segments in indirect requests (default is 32)");
 100
 101 #define BLK_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE)
 102
 103 /*
 104  * We have one of these per vbd, whether ide, scsi or 'other'.  They
 105  * hang in private_data off the gendisk structure. We may end up
 106  * putting all kinds of interesting stuff here :-)
 107  */
 108 struct blkfront_info
 109 {
 110         spinlock_t io_lock;
 111         struct mutex mutex;
 112         struct xenbus_device *xbdev;
 113         struct gendisk *gd;
 114         int vdevice;
 115         blkif_vdev_t handle;
 116         enum blkif_state connected;
 117         int ring_ref;
 118         struct blkif_front_ring ring;
 119         unsigned int evtchn, irq;
 120         struct request_queue *rq;
 121         struct work_struct work;
 122         struct gnttab_free_callback callback;
 123         struct blk_shadow shadow[BLK_RING_SIZE];
 124         struct list_head persistent_gnts;
 125         unsigned int persistent_gnts_c;
 126         unsigned long shadow_free;
 127         unsigned int feature_flush;
 128         unsigned int flush_op;
 129         unsigned int feature_discard:1;
 130         unsigned int feature_secdiscard:1;
 131         unsigned int discard_granularity;
 132         unsigned int discard_alignment;
 133         unsigned int feature_persistent:1;
 134         unsigned int max_indirect_segments;
 135         int is_ready;
 136 };
 137
 138 static unsigned int nr_minors;
 139 static unsigned long *minors;
 140 static DEFINE_SPINLOCK(minor_lock);
 141
 142 #define MAXIMUM_OUTSTANDING_BLOCK_REQS \
 143         (BLKIF_MAX_SEGMENTS_PER_REQUEST * BLK_RING_SIZE)
 144 #define GRANT_INVALID_REF       0
 145
 146 #define PARTS_PER_DISK          16
 147 #define PARTS_PER_EXT_DISK      256
 148
 149 #define BLKIF_MAJOR(dev) ((dev)>>8)
 150 #define BLKIF_MINOR(dev) ((dev) & 0xff)
 151
 152 #define EXT_SHIFT 28
 153 #define EXTENDED (1<<EXT_SHIFT)
 154 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED))
 155 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED))
 156 #define EMULATED_HD_DISK_MINOR_OFFSET (0)
 157 #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256)
 158 #define EMULATED_SD_DISK_MINOR_OFFSET (0)
 159 #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256)
 160
 161 #define DEV_NAME        "xvd"   /* name in /dev */
 162
 163 #define SEGS_PER_INDIRECT_FRAME \
 164         (PAGE_SIZE/sizeof(struct blkif_request_segment_aligned))
 165 #define INDIRECT_GREFS(_segs) \
 166         ((_segs + SEGS_PER_INDIRECT_FRAME - 1)/SEGS_PER_INDIRECT_FRAME)
 167
 168 static int blkfront_setup_indirect(struct blkfront_info *info);
 169
 170 static int get_id_from_freelist(struct blkfront_info *info)
 171 {
 172         unsigned long free = info->shadow_free;
 173         BUG_ON(free >= BLK_RING_SIZE);
 174         info->shadow_free = info->shadow[free].req.u.rw.id;
 175         info->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */
 176         return free;
 177 }
 178
 179 static int add_id_to_freelist(struct blkfront_info *info,
 180                                unsigned long id)
 181 {
 182         if (info->shadow[id].req.u.rw.id != id)
 183                 return -EINVAL;
 184         if (info->shadow[id].request == NULL)
 185                 return -EINVAL;
 186         info->shadow[id].req.u.rw.id  = info->shadow_free;
 187         info->shadow[id].request = NULL;
 188         info->shadow_free = id;
 189         return 0;
 190 }
 191
 192 static int fill_grant_buffer(struct blkfront_info *info, int num)
 193 {
 194         struct page *granted_page;
 195         struct grant *gnt_list_entry, *n;
 196         int i = 0;
 197
 198         while(i < num) {
 199                 gnt_list_entry = kzalloc(sizeof(struct grant), GFP_NOIO);
 200                 if (!gnt_list_entry)
 201                         goto out_of_memory;
 202
 203                 granted_page = alloc_page(GFP_NOIO);
 204                 if (!granted_page) {
 205                         kfree(gnt_list_entry);
 206                         goto out_of_memory;
 207                 }
 208
 209                 gnt_list_entry->pfn = page_to_pfn(granted_page);
 210                 gnt_list_entry->gref = GRANT_INVALID_REF;
 211                 list_add(&gnt_list_entry->node, &info->persistent_gnts);
 212                 i++;
 213         }
 214
 215         return 0;
 216
 217 out_of_memory:
 218         list_for_each_entry_safe(gnt_list_entry, n,
 219                                  &info->persistent_gnts, node) {
 220                 list_del(&gnt_list_entry->node);
 221                 __free_page(pfn_to_page(gnt_list_entry->pfn));
 222                 kfree(gnt_list_entry);
 223                 i--;
 224         }
 225         BUG_ON(i != 0);
 226         return -ENOMEM;
 227 }
 228
 229 static struct grant *get_grant(grant_ref_t *gref_head,
 230                                struct blkfront_info *info)
 231 {
 232         struct grant *gnt_list_entry;
 233         unsigned long buffer_mfn;
 234
 235         BUG_ON(list_empty(&info->persistent_gnts));
 236         gnt_list_entry = list_first_entry(&info->persistent_gnts, struct grant,
 237                                           node);
 238         list_del(&gnt_list_entry->node);
 239
 240         if (gnt_list_entry->gref != GRANT_INVALID_REF) {
 241                 info->persistent_gnts_c--;
 242                 return gnt_list_entry;
 243         }
 244
 245         /* Assign a gref to this page */
 246         gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head);
 247         BUG_ON(gnt_list_entry->gref == -ENOSPC);
 248         buffer_mfn = pfn_to_mfn(gnt_list_entry->pfn);
 249         gnttab_grant_foreign_access_ref(gnt_list_entry->gref,
 250                                         info->xbdev->otherend_id,
 251                                         buffer_mfn, 0);
 252         return gnt_list_entry;
 253 }
 254
 255 static const char *op_name(int op)
 256 {
 257         static const char *const names[] = {
 258                 [BLKIF_OP_READ] = "read",
 259                 [BLKIF_OP_WRITE] = "write",
 260                 [BLKIF_OP_WRITE_BARRIER] = "barrier",
 261                 [BLKIF_OP_FLUSH_DISKCACHE] = "flush",
 262                 [BLKIF_OP_DISCARD] = "discard" };
 263
 264         if (op < 0 || op >= ARRAY_SIZE(names))
 265                 return "unknown";
 266
 267         if (!names[op])
 268                 return "reserved";
 269
 270         return names[op];
 271 }
 272 static int xlbd_reserve_minors(unsigned int minor, unsigned int nr)
 273 {
 274         unsigned int end = minor + nr;
 275         int rc;
 276
 277         if (end > nr_minors) {
 278                 unsigned long *bitmap, *old;
 279
 280                 bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap),
 281                                  GFP_KERNEL);
 282                 if (bitmap == NULL)
 283                         return -ENOMEM;
 284
 285                 spin_lock(&minor_lock);
 286                 if (end > nr_minors) {
 287                         old = minors;
 288                         memcpy(bitmap, minors,
 289                                BITS_TO_LONGS(nr_minors) * sizeof(*bitmap));
 290                         minors = bitmap;
 291                         nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG;
 292                 } else
 293                         old = bitmap;
 294                 spin_unlock(&minor_lock);
 295                 kfree(old);
 296         }
 297
 298         spin_lock(&minor_lock);
 299         if (find_next_bit(minors, end, minor) >= end) {
 300                 bitmap_set(minors, minor, nr);
 301                 rc = 0;
 302         } else
 303                 rc = -EBUSY;
 304         spin_unlock(&minor_lock);
 305
 306         return rc;
 307 }
 308
 309 static void xlbd_release_minors(unsigned int minor, unsigned int nr)
 310 {
 311         unsigned int end = minor + nr;
 312
 313         BUG_ON(end > nr_minors);
 314         spin_lock(&minor_lock);
 315         bitmap_clear(minors,  minor, nr);
 316         spin_unlock(&minor_lock);
 317 }
 318
 319 static void blkif_restart_queue_callback(void *arg)
 320 {
 321         struct blkfront_info *info = (struct blkfront_info *)arg;
 322         schedule_work(&info->work);
 323 }
 324
 325 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg)
 326 {
 327         /* We don't have real geometry info, but let's at least return
 328            values consistent with the size of the device */
 329         sector_t nsect = get_capacity(bd->bd_disk);
 330         sector_t cylinders = nsect;
 331
 332         hg->heads = 0xff;
 333         hg->sectors = 0x3f;
 334         sector_div(cylinders, hg->heads * hg->sectors);
 335         hg->cylinders = cylinders;
 336         if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect)
 337                 hg->cylinders = 0xffff;
 338         return 0;
 339 }
 340
 341 static int blkif_ioctl(struct block_device *bdev, fmode_t mode,
 342                        unsigned command, unsigned long argument)
 343 {
 344         struct blkfront_info *info = bdev->bd_disk->private_data;
 345         int i;
 346
 347         dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n",
 348                 command, (long)argument);
 349
 350         switch (command) {
 351         case CDROMMULTISESSION:
 352                 dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n");
 353                 for (i = 0; i < sizeof(struct cdrom_multisession); i++)
 354                         if (put_user(0, (char __user *)(argument + i)))
 355                                 return -EFAULT;
 356                 return 0;
 357
 358         case CDROM_GET_CAPABILITY: {
 359                 struct gendisk *gd = info->gd;
 360                 if (gd->flags & GENHD_FL_CD)
 361                         return 0;
 362                 return -EINVAL;
 363         }
 364
 365         default:
 366                 /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n",
 367                   command);*/
 368                 return -EINVAL; /* same return as native Linux */
 369         }
 370
 371         return 0;
 372 }
 373
 374 /*
 375  * Generate a Xen blkfront IO request from a blk layer request.  Reads
 376  * and writes are handled as expected.
 377  *
 378  * @req: a request struct
 379  */
 380 static int blkif_queue_request(struct request *req)
 381 {
 382         struct blkfront_info *info = req->rq_disk->private_data;
 383         struct blkif_request *ring_req;
 384         unsigned long id;
 385         unsigned int fsect, lsect;
 386         int i, ref, n;
 387         struct blkif_request_segment_aligned *segments = NULL;
 388
 389         /*
 390          * Used to store if we are able to queue the request by just using
 391          * existing persistent grants, or if we have to get new grants,
 392          * as there are not sufficiently many free.
 393          */
 394         bool new_persistent_gnts;
 395         grant_ref_t gref_head;
 396         struct grant *gnt_list_entry = NULL;
 397         struct scatterlist *sg;
 398         int nseg, max_grefs;
 399
 400         if (unlikely(info->connected != BLKIF_STATE_CONNECTED))
 401                 return 1;
 402
 403         max_grefs = info->max_indirect_segments ?
 404                     info->max_indirect_segments +
 405                     INDIRECT_GREFS(info->max_indirect_segments) :
 406                     BLKIF_MAX_SEGMENTS_PER_REQUEST;
 407
 408         /* Check if we have enough grants to allocate a requests */
 409         if (info->persistent_gnts_c < max_grefs) {
 410                 new_persistent_gnts = 1;
 411                 if (gnttab_alloc_grant_references(
 412                     max_grefs - info->persistent_gnts_c,
 413                     &gref_head) < 0) {
 414                         gnttab_request_free_callback(
 415                                 &info->callback,
 416                                 blkif_restart_queue_callback,
 417                                 info,
 418                                 max_grefs);
 419                         return 1;
 420                 }
 421         } else
 422                 new_persistent_gnts = 0;
 423
 424         /* Fill out a communications ring structure. */
 425         ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
 426         id = get_id_from_freelist(info);
 427         info->shadow[id].request = req;
 428
 429         if (unlikely(req->cmd_flags & (REQ_DISCARD | REQ_SECURE))) {
 430                 ring_req->operation = BLKIF_OP_DISCARD;
 431                 ring_req->u.discard.nr_sectors = blk_rq_sectors(req);
 432                 ring_req->u.discard.id = id;
 433                 ring_req->u.discard.sector_number = (blkif_sector_t)blk_rq_pos(req);
 434                 if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard)
 435                         ring_req->u.discard.flag = BLKIF_DISCARD_SECURE;
 436                 else
 437                         ring_req->u.discard.flag = 0;
 438         } else {
 439                 BUG_ON(info->max_indirect_segments == 0 &&
 440                        req->nr_phys_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST);
 441                 BUG_ON(info->max_indirect_segments &&
 442                        req->nr_phys_segments > info->max_indirect_segments);
 443                 nseg = blk_rq_map_sg(req->q, req, info->shadow[id].sg);
 444                 ring_req->u.rw.id = id;
 445                 if (nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST) {
 446                         /*
 447                          * The indirect operation can only be a BLKIF_OP_READ or
 448                          * BLKIF_OP_WRITE
 449                          */
 450                         BUG_ON(req->cmd_flags & (REQ_FLUSH | REQ_FUA));
 451                         ring_req->operation = BLKIF_OP_INDIRECT;
 452                         ring_req->u.indirect.indirect_op = rq_data_dir(req) ?
 453                                 BLKIF_OP_WRITE : BLKIF_OP_READ;
 454                         ring_req->u.indirect.sector_number = (blkif_sector_t)blk_rq_pos(req);
 455                         ring_req->u.indirect.handle = info->handle;
 456                         ring_req->u.indirect.nr_segments = nseg;
 457                 } else {
 458                         ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req);
 459                         ring_req->u.rw.handle = info->handle;
 460                         ring_req->operation = rq_data_dir(req) ?
 461                                 BLKIF_OP_WRITE : BLKIF_OP_READ;
 462                         if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) {
 463                                 /*
 464                                  * Ideally we can do an unordered flush-to-disk. In case the
 465                                  * backend onlysupports barriers, use that. A barrier request
 466                                  * a superset of FUA, so we can implement it the same
 467                                  * way.  (It's also a FLUSH+FUA, since it is
 468                                  * guaranteed ordered WRT previous writes.)
 469                                  */
 470                                 ring_req->operation = info->flush_op;
 471                         }
 472                         ring_req->u.rw.nr_segments = nseg;
 473                 }
 474                 for_each_sg(info->shadow[id].sg, sg, nseg, i) {
 475                         fsect = sg->offset >> 9;
 476                         lsect = fsect + (sg->length >> 9) - 1;
 477
 478                         if ((ring_req->operation == BLKIF_OP_INDIRECT) &&
 479                             (i % SEGS_PER_INDIRECT_FRAME == 0)) {
 480                                 if (segments)
 481                                         kunmap_atomic(segments);
 482
 483                                 n = i / SEGS_PER_INDIRECT_FRAME;
 484                                 gnt_list_entry = get_grant(&gref_head, info);
 485                                 info->shadow[id].indirect_grants[n] = gnt_list_entry;
 486                                 segments = kmap_atomic(pfn_to_page(gnt_list_entry->pfn));
 487                                 ring_req->u.indirect.indirect_grefs[n] = gnt_list_entry->gref;
 488                         }
 489
 490                         gnt_list_entry = get_grant(&gref_head, info);
 491                         ref = gnt_list_entry->gref;
 492
 493                         info->shadow[id].grants_used[i] = gnt_list_entry;
 494
 495                         if (rq_data_dir(req)) {
 496                                 char *bvec_data;
 497                                 void *shared_data;
 498
 499                                 BUG_ON(sg->offset + sg->length > PAGE_SIZE);
 500
 501                                 shared_data = kmap_atomic(pfn_to_page(gnt_list_entry->pfn));
 502                                 bvec_data = kmap_atomic(sg_page(sg));
 503
 504                                 /*
 505                                  * this does not wipe data stored outside the
 506                                  * range sg->offset..sg->offset+sg->length.
 507                                  * Therefore, blkback *could* see data from
 508                                  * previous requests. This is OK as long as
 509                                  * persistent grants are shared with just one
 510                                  * domain. It may need refactoring if this
 511                                  * changes
 512                                  */
 513                                 memcpy(shared_data + sg->offset,
 514                                        bvec_data   + sg->offset,
 515                                        sg->length);
 516
 517                                 kunmap_atomic(bvec_data);
 518                                 kunmap_atomic(shared_data);
 519                         }
 520                         if (ring_req->operation != BLKIF_OP_INDIRECT) {
 521                                 ring_req->u.rw.seg[i] =
 522                                                 (struct blkif_request_segment) {
 523                                                         .gref       = ref,
 524                                                         .first_sect = fsect,
 525                                                         .last_sect  = lsect };
 526                         } else {
 527                                 n = i % SEGS_PER_INDIRECT_FRAME;
 528                                 segments[n] =
 529                                         (struct blkif_request_segment_aligned) {
 530                                                         .gref       = ref,
 531                                                         .first_sect = fsect,
 532                                                         .last_sect  = lsect };
 533                         }
 534                 }
 535                 if (segments)
 536                         kunmap_atomic(segments);
 537         }
 538
 539         info->ring.req_prod_pvt++;
 540
 541         /* Keep a private copy so we can reissue requests when recovering. */
 542         info->shadow[id].req = *ring_req;
 543
 544         if (new_persistent_gnts)
 545                 gnttab_free_grant_references(gref_head);
 546
 547         return 0;
 548 }
 549
 550
 551 static inline void flush_requests(struct blkfront_info *info)
 552 {
 553         int notify;
 554
 555         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&info->ring, notify);
 556
 557         if (notify)
 558                 notify_remote_via_irq(info->irq);
 559 }
 560
 561 /*
 562  * do_blkif_request
 563  *  read a block; request is in a request queue
 564  */
 565 static void do_blkif_request(struct request_queue *rq)
 566 {
 567         struct blkfront_info *info = NULL;
 568         struct request *req;
 569         int queued;
 570
 571         pr_debug("Entered do_blkif_request\n");
 572
 573         queued = 0;
 574
 575         while ((req = blk_peek_request(rq)) != NULL) {
 576                 info = req->rq_disk->private_data;
 577
 578                 if (RING_FULL(&info->ring))
 579                         goto wait;
 580
 581                 blk_start_request(req);
 582
 583                 if ((req->cmd_type != REQ_TYPE_FS) ||
 584                     ((req->cmd_flags & (REQ_FLUSH | REQ_FUA)) &&
 585                     !info->flush_op)) {
 586                         __blk_end_request_all(req, -EIO);
 587                         continue;
 588                 }
 589
 590                 pr_debug("do_blk_req %p: cmd %p, sec %lx, "
 591                          "(%u/%u) buffer:%p [%s]\n",
 592                          req, req->cmd, (unsigned long)blk_rq_pos(req),
 593                          blk_rq_cur_sectors(req), blk_rq_sectors(req),
 594                          req->buffer, rq_data_dir(req) ? "write" : "read");
 595
 596                 if (blkif_queue_request(req)) {
 597                         blk_requeue_request(rq, req);
 598 wait:
 599                         /* Avoid pointless unplugs. */
 600                         blk_stop_queue(rq);
 601                         break;
 602                 }
 603
 604                 queued++;
 605         }
 606
 607         if (queued != 0)
 608                 flush_requests(info);
 609 }
 610
 611 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size,
 612                                 unsigned int physical_sector_size,
 613                                 unsigned int segments)
 614 {
 615         struct request_queue *rq;
 616         struct blkfront_info *info = gd->private_data;
 617
 618         rq = blk_init_queue(do_blkif_request, &info->io_lock);
 619         if (rq == NULL)
 620                 return -1;
 621
 622         queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq);
 623
 624         if (info->feature_discard) {
 625                 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq);
 626                 blk_queue_max_discard_sectors(rq, get_capacity(gd));
 627                 rq->limits.discard_granularity = info->discard_granularity;
 628                 rq->limits.discard_alignment = info->discard_alignment;
 629                 if (info->feature_secdiscard)
 630                         queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq);
 631         }
 632
 633         /* Hard sector size and max sectors impersonate the equiv. hardware. */
 634         blk_queue_logical_block_size(rq, sector_size);
 635         blk_queue_physical_block_size(rq, physical_sector_size);
 636         blk_queue_max_hw_sectors(rq, (segments * PAGE_SIZE) / 512);
 637
 638         /* Each segment in a request is up to an aligned page in size. */
 639         blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
 640         blk_queue_max_segment_size(rq, PAGE_SIZE);
 641
 642         /* Ensure a merged request will fit in a single I/O ring slot. */
 643         blk_queue_max_segments(rq, segments);
 644
 645         /* Make sure buffer addresses are sector-aligned. */
 646         blk_queue_dma_alignment(rq, 511);
 647
 648         /* Make sure we don't use bounce buffers. */
 649         blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY);
 650
 651         gd->queue = rq;
 652
 653         return 0;
 654 }
 655
 656
 657 static void xlvbd_flush(struct blkfront_info *info)
 658 {
 659         blk_queue_flush(info->rq, info->feature_flush);
 660         printk(KERN_INFO "blkfront: %s: %s: %s %s %s %s %s\n",
 661                info->gd->disk_name,
 662                info->flush_op == BLKIF_OP_WRITE_BARRIER ?
 663                 "barrier" : (info->flush_op == BLKIF_OP_FLUSH_DISKCACHE ?
 664                 "flush diskcache" : "barrier or flush"),
 665                info->feature_flush ? "enabled;" : "disabled;",
 666                "persistent grants:",
 667                info->feature_persistent ? "enabled;" : "disabled;",
 668                "indirect descriptors:",
 669                info->max_indirect_segments ? "enabled;" : "disabled;");
 670 }
 671
 672 static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset)
 673 {
 674         int major;
 675         major = BLKIF_MAJOR(vdevice);
 676         *minor = BLKIF_MINOR(vdevice);
 677         switch (major) {
 678                 case XEN_IDE0_MAJOR:
 679                         *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET;
 680                         *minor = ((*minor / 64) * PARTS_PER_DISK) +
 681                                 EMULATED_HD_DISK_MINOR_OFFSET;
 682                         break;
 683                 case XEN_IDE1_MAJOR:
 684                         *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET;
 685                         *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) +
 686                                 EMULATED_HD_DISK_MINOR_OFFSET;
 687                         break;
 688                 case XEN_SCSI_DISK0_MAJOR:
 689                         *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET;
 690                         *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET;
 691                         break;
 692                 case XEN_SCSI_DISK1_MAJOR:
 693                 case XEN_SCSI_DISK2_MAJOR:
 694                 case XEN_SCSI_DISK3_MAJOR:
 695                 case XEN_SCSI_DISK4_MAJOR:
 696                 case XEN_SCSI_DISK5_MAJOR:
 697                 case XEN_SCSI_DISK6_MAJOR:
 698                 case XEN_SCSI_DISK7_MAJOR:
 699                         *offset = (*minor / PARTS_PER_DISK) +
 700                                 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) +
 701                                 EMULATED_SD_DISK_NAME_OFFSET;
 702                         *minor = *minor +
 703                                 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) +
 704                                 EMULATED_SD_DISK_MINOR_OFFSET;
 705                         break;
 706                 case XEN_SCSI_DISK8_MAJOR:
 707                 case XEN_SCSI_DISK9_MAJOR:
 708                 case XEN_SCSI_DISK10_MAJOR:
 709                 case XEN_SCSI_DISK11_MAJOR:
 710                 case XEN_SCSI_DISK12_MAJOR:
 711                 case XEN_SCSI_DISK13_MAJOR:
 712                 case XEN_SCSI_DISK14_MAJOR:
 713                 case XEN_SCSI_DISK15_MAJOR:
 714                         *offset = (*minor / PARTS_PER_DISK) +
 715                                 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) +
 716                                 EMULATED_SD_DISK_NAME_OFFSET;
 717                         *minor = *minor +
 718                                 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) +
 719                                 EMULATED_SD_DISK_MINOR_OFFSET;
 720                         break;
 721                 case XENVBD_MAJOR:
 722                         *offset = *minor / PARTS_PER_DISK;
 723                         break;
 724                 default:
 725                         printk(KERN_WARNING "blkfront: your disk configuration is "
 726                                         "incorrect, please use an xvd device instead\n");
 727                         return -ENODEV;
 728         }
 729         return 0;
 730 }
 731
 732 static char *encode_disk_name(char *ptr, unsigned int n)
 733 {
 734         if (n >= 26)
 735                 ptr = encode_disk_name(ptr, n / 26 - 1);
 736         *ptr = 'a' + n % 26;
 737         return ptr + 1;
 738 }
 739
 740 static int xlvbd_alloc_gendisk(blkif_sector_t capacity,
 741                                struct blkfront_info *info,
 742                                u16 vdisk_info, u16 sector_size,
 743                                unsigned int physical_sector_size)
 744 {
 745         struct gendisk *gd;
 746         int nr_minors = 1;
 747         int err;
 748         unsigned int offset;
 749         int minor;
 750         int nr_parts;
 751         char *ptr;
 752
 753         BUG_ON(info->gd != NULL);
 754         BUG_ON(info->rq != NULL);
 755
 756         if ((info->vdevice>>EXT_SHIFT) > 1) {
 757                 /* this is above the extended range; something is wrong */
 758                 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice);
 759                 return -ENODEV;
 760         }
 761
 762         if (!VDEV_IS_EXTENDED(info->vdevice)) {
 763                 err = xen_translate_vdev(info->vdevice, &minor, &offset);
 764                 if (err)
 765                         return err;
 766                 nr_parts = PARTS_PER_DISK;
 767         } else {
 768                 minor = BLKIF_MINOR_EXT(info->vdevice);
 769                 nr_parts = PARTS_PER_EXT_DISK;
 770                 offset = minor / nr_parts;
 771                 if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4)
 772                         printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with "
 773                                         "emulated IDE disks,\n\t choose an xvd device name"
 774                                         "from xvde on\n", info->vdevice);
 775         }
 776         if (minor >> MINORBITS) {
 777                 pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n",
 778                         info->vdevice, minor);
 779                 return -ENODEV;
 780         }
 781
 782         if ((minor % nr_parts) == 0)
 783                 nr_minors = nr_parts;
 784
 785         err = xlbd_reserve_minors(minor, nr_minors);
 786         if (err)
 787                 goto out;
 788         err = -ENODEV;
 789
 790         gd = alloc_disk(nr_minors);
 791         if (gd == NULL)
 792                 goto release;
 793
 794         strcpy(gd->disk_name, DEV_NAME);
 795         ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset);
 796         BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN);
 797         if (nr_minors > 1)
 798                 *ptr = 0;
 799         else
 800                 snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr,
 801                          "%d", minor & (nr_parts - 1));
 802
 803         gd->major = XENVBD_MAJOR;
 804         gd->first_minor = minor;
 805         gd->fops = &xlvbd_block_fops;
 806         gd->private_data = info;
 807         gd->driverfs_dev = &(info->xbdev->dev);
 808         set_capacity(gd, capacity);
 809
 810         if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size,
 811                                  info->max_indirect_segments ? :
 812                                  BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
 813                 del_gendisk(gd);
 814                 goto release;
 815         }
 816
 817         info->rq = gd->queue;
 818         info->gd = gd;
 819
 820         xlvbd_flush(info);
 821
 822         if (vdisk_info & VDISK_READONLY)
 823                 set_disk_ro(gd, 1);
 824
 825         if (vdisk_info & VDISK_REMOVABLE)
 826                 gd->flags |= GENHD_FL_REMOVABLE;
 827
 828         if (vdisk_info & VDISK_CDROM)
 829                 gd->flags |= GENHD_FL_CD;
 830
 831         return 0;
 832
 833  release:
 834         xlbd_release_minors(minor, nr_minors);
 835  out:
 836         return err;
 837 }
 838
 839 static void xlvbd_release_gendisk(struct blkfront_info *info)
 840 {
 841         unsigned int minor, nr_minors;
 842         unsigned long flags;
 843
 844         if (info->rq == NULL)
 845                 return;
 846
 847         spin_lock_irqsave(&info->io_lock, flags);
 848
 849         /* No more blkif_request(). */
 850         blk_stop_queue(info->rq);
 851
 852         /* No more gnttab callback work. */
 853         gnttab_cancel_free_callback(&info->callback);
 854         spin_unlock_irqrestore(&info->io_lock, flags);
 855
 856         /* Flush gnttab callback work. Must be done with no locks held. */
 857         flush_work(&info->work);
 858
 859         del_gendisk(info->gd);
 860
 861         minor = info->gd->first_minor;
 862         nr_minors = info->gd->minors;
 863         xlbd_release_minors(minor, nr_minors);
 864
 865         blk_cleanup_queue(info->rq);
 866         info->rq = NULL;
 867
 868         put_disk(info->gd);
 869         info->gd = NULL;
 870 }
 871
 872 static void kick_pending_request_queues(struct blkfront_info *info)
 873 {
 874         if (!RING_FULL(&info->ring)) {
 875                 /* Re-enable calldowns. */
 876                 blk_start_queue(info->rq);
 877                 /* Kick things off immediately. */
 878                 do_blkif_request(info->rq);
 879         }
 880 }
 881
 882 static void blkif_restart_queue(struct work_struct *work)
 883 {
 884         struct blkfront_info *info = container_of(work, struct blkfront_info, work);
 885
 886         spin_lock_irq(&info->io_lock);
 887         if (info->connected == BLKIF_STATE_CONNECTED)
 888                 kick_pending_request_queues(info);
 889         spin_unlock_irq(&info->io_lock);
 890 }
 891
 892 static void blkif_free(struct blkfront_info *info, int suspend)
 893 {
 894         struct grant *persistent_gnt;
 895         struct grant *n;
 896         int i, j, segs;
 897
 898         /* Prevent new requests being issued until we fix things up. */
 899         spin_lock_irq(&info->io_lock);
 900         info->connected = suspend ?
 901                 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED;
 902         /* No more blkif_request(). */
 903         if (info->rq)
 904                 blk_stop_queue(info->rq);
 905
 906         /* Remove all persistent grants */
 907         if (!list_empty(&info->persistent_gnts)) {
 908                 list_for_each_entry_safe(persistent_gnt, n,
 909                                          &info->persistent_gnts, node) {
 910                         list_del(&persistent_gnt->node);
 911                         if (persistent_gnt->gref != GRANT_INVALID_REF) {
 912                                 gnttab_end_foreign_access(persistent_gnt->gref,
 913                                                           0, 0UL);
 914                                 info->persistent_gnts_c--;
 915                         }
 916                         __free_page(pfn_to_page(persistent_gnt->pfn));
 917                         kfree(persistent_gnt);
 918                 }
 919         }
 920         BUG_ON(info->persistent_gnts_c != 0);
 921
 922         for (i = 0; i < BLK_RING_SIZE; i++) {
 923                 /*
 924                  * Clear persistent grants present in requests already
 925                  * on the shared ring
 926                  */
 927                 if (!info->shadow[i].request)
 928                         goto free_shadow;
 929
 930                 segs = info->shadow[i].req.operation == BLKIF_OP_INDIRECT ?
 931                        info->shadow[i].req.u.indirect.nr_segments :
 932                        info->shadow[i].req.u.rw.nr_segments;
 933                 for (j = 0; j < segs; j++) {
 934                         persistent_gnt = info->shadow[i].grants_used[j];
 935                         gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL);
 936                         __free_page(pfn_to_page(persistent_gnt->pfn));
 937                         kfree(persistent_gnt);
 938                 }
 939
 940                 if (info->shadow[i].req.operation != BLKIF_OP_INDIRECT)
 941                         /*
 942                          * If this is not an indirect operation don't try to
 943                          * free indirect segments
 944                          */
 945                         goto free_shadow;
 946
 947                 for (j = 0; j < INDIRECT_GREFS(segs); j++) {
 948                         persistent_gnt = info->shadow[i].indirect_grants[j];
 949                         gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL);
 950                         __free_page(pfn_to_page(persistent_gnt->pfn));
 951                         kfree(persistent_gnt);
 952                 }
 953
 954 free_shadow:
 955                 kfree(info->shadow[i].grants_used);
 956                 info->shadow[i].grants_used = NULL;
 957                 kfree(info->shadow[i].indirect_grants);
 958                 info->shadow[i].indirect_grants = NULL;
 959                 kfree(info->shadow[i].sg);
 960                 info->shadow[i].sg = NULL;
 961         }
 962
 963         /* No more gnttab callback work. */
 964         gnttab_cancel_free_callback(&info->callback);
 965         spin_unlock_irq(&info->io_lock);
 966
 967         /* Flush gnttab callback work. Must be done with no locks held. */
 968         flush_work(&info->work);
 969
 970         /* Free resources associated with old device channel. */
 971         if (info->ring_ref != GRANT_INVALID_REF) {
 972                 gnttab_end_foreign_access(info->ring_ref, 0,
 973                                           (unsigned long)info->ring.sring);
 974                 info->ring_ref = GRANT_INVALID_REF;
 975                 info->ring.sring = NULL;
 976         }
 977         if (info->irq)
 978                 unbind_from_irqhandler(info->irq, info);
 979         info->evtchn = info->irq = 0;
 980
 981 }
 982
 983 static void blkif_completion(struct blk_shadow *s, struct blkfront_info *info,
 984                              struct blkif_response *bret)
 985 {
 986         int i = 0;
 987         struct scatterlist *sg;
 988         char *bvec_data;
 989         void *shared_data;
 990         int nseg;
 991
 992         nseg = s->req.operation == BLKIF_OP_INDIRECT ?
 993                 s->req.u.indirect.nr_segments : s->req.u.rw.nr_segments;
 994
 995         if (bret->operation == BLKIF_OP_READ) {
 996                 /*
 997                  * Copy the data received from the backend into the bvec.
 998                  * Since bv_offset can be different than 0, and bv_len different
 999                  * than PAGE_SIZE, we have to keep track of the current offset,
1000                  * to be sure we are copying the data from the right shared page.
1001                  */
1002                 for_each_sg(s->sg, sg, nseg, i) {
1003                         BUG_ON(sg->offset + sg->length > PAGE_SIZE);
1004                         shared_data = kmap_atomic(
1005                                 pfn_to_page(s->grants_used[i]->pfn));
1006                         bvec_data = kmap_atomic(sg_page(sg));
1007                         memcpy(bvec_data   + sg->offset,
1008                                shared_data + sg->offset,
1009                                sg->length);
1010                         kunmap_atomic(bvec_data);
1011                         kunmap_atomic(shared_data);
1012                 }
1013         }
1014         /* Add the persistent grant into the list of free grants */
1015         for (i = 0; i < nseg; i++) {
1016                 list_add(&s->grants_used[i]->node, &info->persistent_gnts);
1017                 info->persistent_gnts_c++;
1018         }
1019         if (s->req.operation == BLKIF_OP_INDIRECT) {
1020                 for (i = 0; i < INDIRECT_GREFS(nseg); i++) {
1021                         list_add(&s->indirect_grants[i]->node, &info->persistent_gnts);
1022                         info->persistent_gnts_c++;
1023                 }
1024         }
1025 }
1026
1027 static irqreturn_t blkif_interrupt(int irq, void *dev_id)
1028 {
1029         struct request *req;
1030         struct blkif_response *bret;
1031         RING_IDX i, rp;
1032         unsigned long flags;
1033         struct blkfront_info *info = (struct blkfront_info *)dev_id;
1034         int error;
1035
1036         spin_lock_irqsave(&info->io_lock, flags);
1037
1038         if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) {
1039                 spin_unlock_irqrestore(&info->io_lock, flags);
1040                 return IRQ_HANDLED;
1041         }
1042
1043  again:
1044         rp = info->ring.sring->rsp_prod;
1045         rmb(); /* Ensure we see queued responses up to 'rp'. */
1046
1047         for (i = info->ring.rsp_cons; i != rp; i++) {
1048                 unsigned long id;
1049
1050                 bret = RING_GET_RESPONSE(&info->ring, i);
1051                 id   = bret->id;
1052                 /*
1053                  * The backend has messed up and given us an id that we would
1054                  * never have given to it (we stamp it up to BLK_RING_SIZE -
1055                  * look in get_id_from_freelist.
1056                  */
1057                 if (id >= BLK_RING_SIZE) {
1058                         WARN(1, "%s: response to %s has incorrect id (%ld)\n",
1059                              info->gd->disk_name, op_name(bret->operation), id);
1060                         /* We can't safely get the 'struct request' as
1061                          * the id is busted. */
1062                         continue;
1063                 }
1064                 req  = info->shadow[id].request;
1065
1066                 if (bret->operation != BLKIF_OP_DISCARD)
1067                         blkif_completion(&info->shadow[id], info, bret);
1068
1069                 if (add_id_to_freelist(info, id)) {
1070                         WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n",
1071                              info->gd->disk_name, op_name(bret->operation), id);
1072                         continue;
1073                 }
1074
1075                 error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO;
1076                 switch (bret->operation) {
1077                 case BLKIF_OP_DISCARD:
1078                         if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
1079                                 struct request_queue *rq = info->rq;
1080                                 printk(KERN_WARNING "blkfront: %s: %s op failed\n",
1081                                            info->gd->disk_name, op_name(bret->operation));
1082                                 error = -EOPNOTSUPP;
1083                                 info->feature_discard = 0;
1084                                 info->feature_secdiscard = 0;
1085                                 queue_flag_clear(QUEUE_FLAG_DISCARD, rq);
1086                                 queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq);
1087                         }
1088                         __blk_end_request_all(req, error);
1089                         break;
1090                 case BLKIF_OP_FLUSH_DISKCACHE:
1091                 case BLKIF_OP_WRITE_BARRIER:
1092                         if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) {
1093                                 printk(KERN_WARNING "blkfront: %s: %s op failed\n",
1094                                        info->gd->disk_name, op_name(bret->operation));
1095                                 error = -EOPNOTSUPP;
1096                         }
1097                         if (unlikely(bret->status == BLKIF_RSP_ERROR &&
1098                                      info->shadow[id].req.u.rw.nr_segments == 0)) {
1099                                 printk(KERN_WARNING "blkfront: %s: empty %s op failed\n",
1100                                        info->gd->disk_name, op_name(bret->operation));
1101                                 error = -EOPNOTSUPP;
1102                         }
1103                         if (unlikely(error)) {
1104                                 if (error == -EOPNOTSUPP)
1105                                         error = 0;
1106                                 info->feature_flush = 0;
1107                                 info->flush_op = 0;
1108                                 xlvbd_flush(info);
1109                         }
1110                         /* fall through */
1111                 case BLKIF_OP_READ:
1112                 case BLKIF_OP_WRITE:
1113                         if (unlikely(bret->status != BLKIF_RSP_OKAY))
1114                                 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data "
1115                                         "request: %x\n", bret->status);
1116
1117                         __blk_end_request_all(req, error);
1118                         break;
1119                 default:
1120                         BUG();
1121                 }
1122         }
1123
1124         info->ring.rsp_cons = i;
1125
1126         if (i != info->ring.req_prod_pvt) {
1127                 int more_to_do;
1128                 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
1129                 if (more_to_do)
1130                         goto again;
1131         } else
1132                 info->ring.sring->rsp_event = i + 1;
1133
1134         kick_pending_request_queues(info);
1135
1136         spin_unlock_irqrestore(&info->io_lock, flags);
1137
1138         return IRQ_HANDLED;
1139 }
1140
1141
1142 static int setup_blkring(struct xenbus_device *dev,
1143                          struct blkfront_info *info)
1144 {
1145         struct blkif_sring *sring;
1146         int err;
1147
1148         info->ring_ref = GRANT_INVALID_REF;
1149
1150         sring = (struct blkif_sring *)__get_free_page(GFP_NOIO | __GFP_HIGH);
1151         if (!sring) {
1152                 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring");
1153                 return -ENOMEM;
1154         }
1155         SHARED_RING_INIT(sring);
1156         FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE);
1157
1158         err = xenbus_grant_ring(dev, virt_to_mfn(info->ring.sring));
1159         if (err < 0) {
1160                 free_page((unsigned long)sring);
1161                 info->ring.sring = NULL;
1162                 goto fail;
1163         }
1164         info->ring_ref = err;
1165
1166         err = xenbus_alloc_evtchn(dev, &info->evtchn);
1167         if (err)
1168                 goto fail;
1169
1170         err = bind_evtchn_to_irqhandler(info->evtchn, blkif_interrupt, 0,
1171                                         "blkif", info);
1172         if (err <= 0) {
1173                 xenbus_dev_fatal(dev, err,
1174                                  "bind_evtchn_to_irqhandler failed");
1175                 goto fail;
1176         }
1177         info->irq = err;
1178
1179         return 0;
1180 fail:
1181         blkif_free(info, 0);
1182         return err;
1183 }
1184
1185
1186 /* Common code used when first setting up, and when resuming. */
1187 static int talk_to_blkback(struct xenbus_device *dev,
1188                            struct blkfront_info *info)
1189 {
1190         const char *message = NULL;
1191         struct xenbus_transaction xbt;
1192         int err;
1193
1194         /* Create shared ring, alloc event channel. */
1195         err = setup_blkring(dev, info);
1196         if (err)
1197                 goto out;
1198
1199 again:
1200         err = xenbus_transaction_start(&xbt);
1201         if (err) {
1202                 xenbus_dev_fatal(dev, err, "starting transaction");
1203                 goto destroy_blkring;
1204         }
1205
1206         err = xenbus_printf(xbt, dev->nodename,
1207                             "ring-ref", "%u", info->ring_ref);
1208         if (err) {
1209                 message = "writing ring-ref";
1210                 goto abort_transaction;
1211         }
1212         err = xenbus_printf(xbt, dev->nodename,
1213                             "event-channel", "%u", info->evtchn);
1214         if (err) {
1215                 message = "writing event-channel";
1216                 goto abort_transaction;
1217         }
1218         err = xenbus_printf(xbt, dev->nodename, "protocol", "%s",
1219                             XEN_IO_PROTO_ABI_NATIVE);
1220         if (err) {
1221                 message = "writing protocol";
1222                 goto abort_transaction;
1223         }
1224         err = xenbus_printf(xbt, dev->nodename,
1225                             "feature-persistent", "%u", 1);
1226         if (err)
1227                 dev_warn(&dev->dev,
1228                          "writing persistent grants feature to xenbus");
1229
1230         err = xenbus_transaction_end(xbt, 0);
1231         if (err) {
1232                 if (err == -EAGAIN)
1233                         goto again;
1234                 xenbus_dev_fatal(dev, err, "completing transaction");
1235                 goto destroy_blkring;
1236         }
1237
1238         xenbus_switch_state(dev, XenbusStateInitialised);
1239
1240         return 0;
1241
1242  abort_transaction:
1243         xenbus_transaction_end(xbt, 1);
1244         if (message)
1245                 xenbus_dev_fatal(dev, err, "%s", message);
1246  destroy_blkring:
1247         blkif_free(info, 0);
1248  out:
1249         return err;
1250 }
1251
1252 /**
1253  * Entry point to this code when a new device is created.  Allocate the basic
1254  * structures and the ring buffer for communication with the backend, and
1255  * inform the backend of the appropriate details for those.  Switch to
1256  * Initialised state.
1257  */
1258 static int blkfront_probe(struct xenbus_device *dev,
1259                           const struct xenbus_device_id *id)
1260 {
1261         int err, vdevice, i;
1262         struct blkfront_info *info;
1263
1264         /* FIXME: Use dynamic device id if this is not set. */
1265         err = xenbus_scanf(XBT_NIL, dev->nodename,
1266                            "virtual-device", "%i", &vdevice);
1267         if (err != 1) {
1268                 /* go looking in the extended area instead */
1269                 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext",
1270                                    "%i", &vdevice);
1271                 if (err != 1) {
1272                         xenbus_dev_fatal(dev, err, "reading virtual-device");
1273                         return err;
1274                 }
1275         }
1276
1277         if (xen_hvm_domain()) {
1278                 char *type;
1279                 int len;
1280                 /* no unplug has been done: do not hook devices != xen vbds */
1281                 if (xen_platform_pci_unplug & XEN_UNPLUG_UNNECESSARY) {
1282                         int major;
1283
1284                         if (!VDEV_IS_EXTENDED(vdevice))
1285                                 major = BLKIF_MAJOR(vdevice);
1286                         else
1287                                 major = XENVBD_MAJOR;
1288
1289                         if (major != XENVBD_MAJOR) {
1290                                 printk(KERN_INFO
1291                                                 "%s: HVM does not support vbd %d as xen block device\n",
1292                                                 __FUNCTION__, vdevice);
1293                                 return -ENODEV;
1294                         }
1295                 }
1296                 /* do not create a PV cdrom device if we are an HVM guest */
1297                 type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len);
1298                 if (IS_ERR(type))
1299                         return -ENODEV;
1300                 if (strncmp(type, "cdrom", 5) == 0) {
1301                         kfree(type);
1302                         return -ENODEV;
1303                 }
1304                 kfree(type);
1305         }
1306         info = kzalloc(sizeof(*info), GFP_KERNEL);
1307         if (!info) {
1308                 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
1309                 return -ENOMEM;
1310         }
1311
1312         mutex_init(&info->mutex);
1313         spin_lock_init(&info->io_lock);
1314         info->xbdev = dev;
1315         info->vdevice = vdevice;
1316         INIT_LIST_HEAD(&info->persistent_gnts);
1317         info->persistent_gnts_c = 0;
1318         info->connected = BLKIF_STATE_DISCONNECTED;
1319         INIT_WORK(&info->work, blkif_restart_queue);
1320
1321         for (i = 0; i < BLK_RING_SIZE; i++)
1322                 info->shadow[i].req.u.rw.id = i+1;
1323         info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff;
1324
1325         /* Front end dir is a number, which is used as the id. */
1326         info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0);
1327         dev_set_drvdata(&dev->dev, info);
1328
1329         err = talk_to_blkback(dev, info);
1330         if (err) {
1331                 kfree(info);
1332                 dev_set_drvdata(&dev->dev, NULL);
1333                 return err;
1334         }
1335
1336         return 0;
1337 }
1338
1339 static void split_bio_end(struct bio *bio, int error)
1340 {
1341         struct split_bio *split_bio = bio->bi_private;
1342
1343         if (error)
1344                 split_bio->err = error;
1345
1346         if (atomic_dec_and_test(&split_bio->pending)) {
1347                 split_bio->bio->bi_phys_segments = 0;
1348                 bio_endio(split_bio->bio, split_bio->err);
1349                 kfree(split_bio);
1350         }
1351         bio_put(bio);
1352 }
1353
1354 static int blkif_recover(struct blkfront_info *info)
1355 {
1356         int i;
1357         struct request *req, *n;
1358         struct blk_shadow *copy;
1359         int rc;
1360         struct bio *bio, *cloned_bio;
1361         struct bio_list bio_list, merge_bio;
1362         unsigned int segs, offset;
1363         int pending, size;
1364         struct split_bio *split_bio;
1365         struct list_head requests;
1366
1367         /* Stage 1: Make a safe copy of the shadow state. */
1368         copy = kmemdup(info->shadow, sizeof(info->shadow),
1369                        GFP_NOIO | __GFP_REPEAT | __GFP_HIGH);
1370         if (!copy)
1371                 return -ENOMEM;
1372
1373         /* Stage 2: Set up free list. */
1374         memset(&info->shadow, 0, sizeof(info->shadow));
1375         for (i = 0; i < BLK_RING_SIZE; i++)
1376                 info->shadow[i].req.u.rw.id = i+1;
1377         info->shadow_free = info->ring.req_prod_pvt;
1378         info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff;
1379
1380         rc = blkfront_setup_indirect(info);
1381         if (rc) {
1382                 kfree(copy);
1383                 return rc;
1384         }
1385
1386         segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST;
1387         blk_queue_max_segments(info->rq, segs);
1388         bio_list_init(&bio_list);
1389         INIT_LIST_HEAD(&requests);
1390         for (i = 0; i < BLK_RING_SIZE; i++) {
1391                 /* Not in use? */
1392                 if (!copy[i].request)
1393                         continue;
1394
1395                 /*
1396                  * Get the bios in the request so we can re-queue them.
1397                  */
1398                 if (copy[i].request->cmd_flags &
1399                     (REQ_FLUSH | REQ_FUA | REQ_DISCARD | REQ_SECURE)) {
1400                         /*
1401                          * Flush operations don't contain bios, so
1402                          * we need to requeue the whole request
1403                          */
1404                         list_add(&copy[i].request->queuelist, &requests);
1405                         continue;
1406                 }
1407                 merge_bio.head = copy[i].request->bio;
1408                 merge_bio.tail = copy[i].request->biotail;
1409                 bio_list_merge(&bio_list, &merge_bio);
1410                 copy[i].request->bio = NULL;
1411                 blk_put_request(copy[i].request);
1412         }
1413
1414         kfree(copy);
1415
1416         /*
1417          * Empty the queue, this is important because we might have
1418          * requests in the queue with more segments than what we
1419          * can handle now.
1420          */
1421         spin_lock_irq(&info->io_lock);
1422         while ((req = blk_fetch_request(info->rq)) != NULL) {
1423                 if (req->cmd_flags &
1424                     (REQ_FLUSH | REQ_FUA | REQ_DISCARD | REQ_SECURE)) {
1425                         list_add(&req->queuelist, &requests);
1426                         continue;
1427                 }
1428                 merge_bio.head = req->bio;
1429                 merge_bio.tail = req->biotail;
1430                 bio_list_merge(&bio_list, &merge_bio);
1431                 req->bio = NULL;
1432                 if (req->cmd_flags & (REQ_FLUSH | REQ_FUA))
1433                         pr_alert("diskcache flush request found!\n");
1434                 __blk_put_request(info->rq, req);
1435         }
1436         spin_unlock_irq(&info->io_lock);
1437
1438         xenbus_switch_state(info->xbdev, XenbusStateConnected);
1439
1440         spin_lock_irq(&info->io_lock);
1441
1442         /* Now safe for us to use the shared ring */
1443         info->connected = BLKIF_STATE_CONNECTED;
1444
1445         /* Kick any other new requests queued since we resumed */
1446         kick_pending_request_queues(info);
1447
1448         list_for_each_entry_safe(req, n, &requests, queuelist) {
1449                 /* Requeue pending requests (flush or discard) */
1450                 list_del_init(&req->queuelist);
1451                 BUG_ON(req->nr_phys_segments > segs);
1452                 blk_requeue_request(info->rq, req);
1453         }
1454         spin_unlock_irq(&info->io_lock);
1455
1456         while ((bio = bio_list_pop(&bio_list)) != NULL) {
1457                 /* Traverse the list of pending bios and re-queue them */
1458                 if (bio_segments(bio) > segs) {
1459                         /*
1460                          * This bio has more segments than what we can
1461                          * handle, we have to split it.
1462                          */
1463                         pending = (bio_segments(bio) + segs - 1) / segs;
1464                         split_bio = kzalloc(sizeof(*split_bio), GFP_NOIO);
1465                         BUG_ON(split_bio == NULL);
1466                         atomic_set(&split_bio->pending, pending);
1467                         split_bio->bio = bio;
1468                         for (i = 0; i < pending; i++) {
1469                                 offset = (i * segs * PAGE_SIZE) >> 9;
1470                                 size = min((unsigned int)(segs * PAGE_SIZE) >> 9,
1471                                            (unsigned int)(bio->bi_size >> 9) - offset);
1472                                 cloned_bio = bio_clone(bio, GFP_NOIO);
1473                                 BUG_ON(cloned_bio == NULL);
1474                                 bio_trim(cloned_bio, offset, size);
1475                                 cloned_bio->bi_private = split_bio;
1476                                 cloned_bio->bi_end_io = split_bio_end;
1477                                 submit_bio(cloned_bio->bi_rw, cloned_bio);
1478                         }
1479                         /*
1480                          * Now we have to wait for all those smaller bios to
1481                          * end, so we can also end the "parent" bio.
1482                          */
1483                         continue;
1484                 }
1485                 /* We don't need to split this bio */
1486                 submit_bio(bio->bi_rw, bio);
1487         }
1488
1489         return 0;
1490 }
1491
1492 /**
1493  * We are reconnecting to the backend, due to a suspend/resume, or a backend
1494  * driver restart.  We tear down our blkif structure and recreate it, but
1495  * leave the device-layer structures intact so that this is transparent to the
1496  * rest of the kernel.
1497  */
1498 static int blkfront_resume(struct xenbus_device *dev)
1499 {
1500         struct blkfront_info *info = dev_get_drvdata(&dev->dev);
1501         int err;
1502
1503         dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename);
1504
1505         blkif_free(info, info->connected == BLKIF_STATE_CONNECTED);
1506
1507         err = talk_to_blkback(dev, info);
1508
1509         /*
1510          * We have to wait for the backend to switch to
1511          * connected state, since we want to read which
1512          * features it supports.
1513          */
1514
1515         return err;
1516 }
1517
1518 static void
1519 blkfront_closing(struct blkfront_info *info)
1520 {
1521         struct xenbus_device *xbdev = info->xbdev;
1522         struct block_device *bdev = NULL;
1523
1524         mutex_lock(&info->mutex);
1525
1526         if (xbdev->state == XenbusStateClosing) {
1527                 mutex_unlock(&info->mutex);
1528                 return;
1529         }
1530
1531         if (info->gd)
1532                 bdev = bdget_disk(info->gd, 0);
1533
1534         mutex_unlock(&info->mutex);
1535
1536         if (!bdev) {
1537                 xenbus_frontend_closed(xbdev);
1538                 return;
1539         }
1540
1541         mutex_lock(&bdev->bd_mutex);
1542
1543         if (bdev->bd_openers) {
1544                 xenbus_dev_error(xbdev, -EBUSY,
1545                                  "Device in use; refusing to close");
1546                 xenbus_switch_state(xbdev, XenbusStateClosing);
1547         } else {
1548                 xlvbd_release_gendisk(info);
1549                 xenbus_frontend_closed(xbdev);
1550         }
1551
1552         mutex_unlock(&bdev->bd_mutex);
1553         bdput(bdev);
1554 }
1555
1556 static void blkfront_setup_discard(struct blkfront_info *info)
1557 {
1558         int err;
1559         char *type;
1560         unsigned int discard_granularity;
1561         unsigned int discard_alignment;
1562         unsigned int discard_secure;
1563
1564         type = xenbus_read(XBT_NIL, info->xbdev->otherend, "type", NULL);
1565         if (IS_ERR(type))
1566                 return;
1567
1568         info->feature_secdiscard = 0;
1569         if (strncmp(type, "phy", 3) == 0) {
1570                 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1571                         "discard-granularity", "%u", &discard_granularity,
1572                         "discard-alignment", "%u", &discard_alignment,
1573                         NULL);
1574                 if (!err) {
1575                         info->feature_discard = 1;
1576                         info->discard_granularity = discard_granularity;
1577                         info->discard_alignment = discard_alignment;
1578                 }
1579                 err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1580                             "discard-secure", "%d", &discard_secure,
1581                             NULL);
1582                 if (!err)
1583                         info->feature_secdiscard = discard_secure;
1584
1585         } else if (strncmp(type, "file", 4) == 0)
1586                 info->feature_discard = 1;
1587
1588         kfree(type);
1589 }
1590
1591 static int blkfront_setup_indirect(struct blkfront_info *info)
1592 {
1593         unsigned int indirect_segments, segs;
1594         int err, i;
1595
1596         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1597                             "feature-max-indirect-segments", "%u", &indirect_segments,
1598                             NULL);
1599         if (err) {
1600                 info->max_indirect_segments = 0;
1601                 segs = BLKIF_MAX_SEGMENTS_PER_REQUEST;
1602         } else {
1603                 info->max_indirect_segments = min(indirect_segments,
1604                                                   xen_blkif_max_segments);
1605                 segs = info->max_indirect_segments;
1606         }
1607
1608         err = fill_grant_buffer(info, (segs + INDIRECT_GREFS(segs)) * BLK_RING_SIZE);
1609         if (err)
1610                 goto out_of_memory;
1611
1612         for (i = 0; i < BLK_RING_SIZE; i++) {
1613                 info->shadow[i].grants_used = kzalloc(
1614                         sizeof(info->shadow[i].grants_used[0]) * segs,
1615                         GFP_NOIO);
1616                 info->shadow[i].sg = kzalloc(sizeof(info->shadow[i].sg[0]) * segs, GFP_NOIO);
1617                 if (info->max_indirect_segments)
1618                         info->shadow[i].indirect_grants = kzalloc(
1619                                 sizeof(info->shadow[i].indirect_grants[0]) *
1620                                 INDIRECT_GREFS(segs),
1621                                 GFP_NOIO);
1622                 if ((info->shadow[i].grants_used == NULL) ||
1623                         (info->shadow[i].sg == NULL) ||
1624                      (info->max_indirect_segments &&
1625                      (info->shadow[i].indirect_grants == NULL)))
1626                         goto out_of_memory;
1627                 sg_init_table(info->shadow[i].sg, segs);
1628         }
1629
1630
1631         return 0;
1632
1633 out_of_memory:
1634         for (i = 0; i < BLK_RING_SIZE; i++) {
1635                 kfree(info->shadow[i].grants_used);
1636                 info->shadow[i].grants_used = NULL;
1637                 kfree(info->shadow[i].sg);
1638                 info->shadow[i].sg = NULL;
1639                 kfree(info->shadow[i].indirect_grants);
1640                 info->shadow[i].indirect_grants = NULL;
1641         }
1642         return -ENOMEM;
1643 }
1644
1645 /*
1646  * Invoked when the backend is finally 'ready' (and has told produced
1647  * the details about the physical device - #sectors, size, etc).
1648  */
1649 static void blkfront_connect(struct blkfront_info *info)
1650 {
1651         unsigned long long sectors;
1652         unsigned long sector_size;
1653         unsigned int physical_sector_size;
1654         unsigned int binfo;
1655         int err;
1656         int barrier, flush, discard, persistent;
1657
1658         switch (info->connected) {
1659         case BLKIF_STATE_CONNECTED:
1660                 /*
1661                  * Potentially, the back-end may be signalling
1662                  * a capacity change; update the capacity.
1663                  */
1664                 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
1665                                    "sectors", "%Lu", &sectors);
1666                 if (XENBUS_EXIST_ERR(err))
1667                         return;
1668                 printk(KERN_INFO "Setting capacity to %Lu\n",
1669                        sectors);
1670                 set_capacity(info->gd, sectors);
1671                 revalidate_disk(info->gd);
1672
1673                 return;
1674         case BLKIF_STATE_SUSPENDED:
1675                 /*
1676                  * If we are recovering from suspension, we need to wait
1677                  * for the backend to announce it's features before
1678                  * reconnecting, at least we need to know if the backend
1679                  * supports indirect descriptors, and how many.
1680                  */
1681                 blkif_recover(info);
1682                 return;
1683
1684         default:
1685                 break;
1686         }
1687
1688         dev_dbg(&info->xbdev->dev, "%s:%s.\n",
1689                 __func__, info->xbdev->otherend);
1690
1691         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1692                             "sectors", "%llu", &sectors,
1693                             "info", "%u", &binfo,
1694                             "sector-size", "%lu", &sector_size,
1695                             NULL);
1696         if (err) {
1697                 xenbus_dev_fatal(info->xbdev, err,
1698                                  "reading backend fields at %s",
1699                                  info->xbdev->otherend);
1700                 return;
1701         }
1702
1703         /*
1704          * physcial-sector-size is a newer field, so old backends may not
1705          * provide this. Assume physical sector size to be the same as
1706          * sector_size in that case.
1707          */
1708         err = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
1709                            "physical-sector-size", "%u", &physical_sector_size);
1710         if (err != 1)
1711                 physical_sector_size = sector_size;
1712
1713         info->feature_flush = 0;
1714         info->flush_op = 0;
1715
1716         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1717                             "feature-barrier", "%d", &barrier,
1718                             NULL);
1719
1720         /*
1721          * If there's no "feature-barrier" defined, then it means
1722          * we're dealing with a very old backend which writes
1723          * synchronously; nothing to do.
1724          *
1725          * If there are barriers, then we use flush.
1726          */
1727         if (!err && barrier) {
1728                 info->feature_flush = REQ_FLUSH | REQ_FUA;
1729                 info->flush_op = BLKIF_OP_WRITE_BARRIER;
1730         }
1731         /*
1732          * And if there is "feature-flush-cache" use that above
1733          * barriers.
1734          */
1735         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1736                             "feature-flush-cache", "%d", &flush,
1737                             NULL);
1738
1739         if (!err && flush) {
1740                 info->feature_flush = REQ_FLUSH;
1741                 info->flush_op = BLKIF_OP_FLUSH_DISKCACHE;
1742         }
1743
1744         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1745                             "feature-discard", "%d", &discard,
1746                             NULL);
1747
1748         if (!err && discard)
1749                 blkfront_setup_discard(info);
1750
1751         err = xenbus_gather(XBT_NIL, info->xbdev->otherend,
1752                             "feature-persistent", "%u", &persistent,
1753                             NULL);
1754         if (err)
1755                 info->feature_persistent = 0;
1756         else
1757                 info->feature_persistent = persistent;
1758
1759         err = blkfront_setup_indirect(info);
1760         if (err) {
1761                 xenbus_dev_fatal(info->xbdev, err, "setup_indirect at %s",
1762                                  info->xbdev->otherend);
1763                 return;
1764         }
1765
1766         err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size,
1767                                   physical_sector_size);
1768         if (err) {
1769                 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s",
1770                                  info->xbdev->otherend);
1771                 return;
1772         }
1773
1774         xenbus_switch_state(info->xbdev, XenbusStateConnected);
1775
1776         /* Kick pending requests. */
1777         spin_lock_irq(&info->io_lock);
1778         info->connected = BLKIF_STATE_CONNECTED;
1779         kick_pending_request_queues(info);
1780         spin_unlock_irq(&info->io_lock);
1781
1782         add_disk(info->gd);
1783
1784         info->is_ready = 1;
1785 }
1786
1787 /**
1788  * Callback received when the backend's state changes.
1789  */
1790 static void blkback_changed(struct xenbus_device *dev,
1791                             enum xenbus_state backend_state)
1792 {
1793         struct blkfront_info *info = dev_get_drvdata(&dev->dev);
1794
1795         dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state);
1796
1797         switch (backend_state) {
1798         case XenbusStateInitialising:
1799         case XenbusStateInitWait:
1800         case XenbusStateInitialised:
1801         case XenbusStateReconfiguring:
1802         case XenbusStateReconfigured:
1803         case XenbusStateUnknown:
1804         case XenbusStateClosed:
1805                 break;
1806
1807         case XenbusStateConnected:
1808                 blkfront_connect(info);
1809                 break;
1810
1811         case XenbusStateClosing:
1812                 blkfront_closing(info);
1813                 break;
1814         }
1815 }
1816
1817 static int blkfront_remove(struct xenbus_device *xbdev)
1818 {
1819         struct blkfront_info *info = dev_get_drvdata(&xbdev->dev);
1820         struct block_device *bdev = NULL;
1821         struct gendisk *disk;
1822
1823         dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename);
1824
1825         blkif_free(info, 0);
1826
1827         mutex_lock(&info->mutex);
1828
1829         disk = info->gd;
1830         if (disk)
1831                 bdev = bdget_disk(disk, 0);
1832
1833         info->xbdev = NULL;
1834         mutex_unlock(&info->mutex);
1835
1836         if (!bdev) {
1837                 kfree(info);
1838                 return 0;
1839         }
1840
1841         /*
1842          * The xbdev was removed before we reached the Closed
1843          * state. See if it's safe to remove the disk. If the bdev
1844          * isn't closed yet, we let release take care of it.
1845          */
1846
1847         mutex_lock(&bdev->bd_mutex);
1848         info = disk->private_data;
1849
1850         dev_warn(disk_to_dev(disk),
1851                  "%s was hot-unplugged, %d stale handles\n",
1852                  xbdev->nodename, bdev->bd_openers);
1853
1854         if (info && !bdev->bd_openers) {
1855                 xlvbd_release_gendisk(info);
1856                 disk->private_data = NULL;
1857                 kfree(info);
1858         }
1859
1860         mutex_unlock(&bdev->bd_mutex);
1861         bdput(bdev);
1862
1863         return 0;
1864 }
1865
1866 static int blkfront_is_ready(struct xenbus_device *dev)
1867 {
1868         struct blkfront_info *info = dev_get_drvdata(&dev->dev);
1869
1870         return info->is_ready && info->xbdev;
1871 }
1872
1873 static int blkif_open(struct block_device *bdev, fmode_t mode)
1874 {
1875         struct gendisk *disk = bdev->bd_disk;
1876         struct blkfront_info *info;
1877         int err = 0;
1878
1879         mutex_lock(&blkfront_mutex);
1880
1881         info = disk->private_data;
1882         if (!info) {
1883                 /* xbdev gone */
1884                 err = -ERESTARTSYS;
1885                 goto out;
1886         }
1887
1888         mutex_lock(&info->mutex);
1889
1890         if (!info->gd)
1891                 /* xbdev is closed */
1892                 err = -ERESTARTSYS;
1893
1894         mutex_unlock(&info->mutex);
1895
1896 out:
1897         mutex_unlock(&blkfront_mutex);
1898         return err;
1899 }
1900
1901 static void blkif_release(struct gendisk *disk, fmode_t mode)
1902 {
1903         struct blkfront_info *info = disk->private_data;
1904         struct block_device *bdev;
1905         struct xenbus_device *xbdev;
1906
1907         mutex_lock(&blkfront_mutex);
1908
1909         bdev = bdget_disk(disk, 0);
1910
1911         if (bdev->bd_openers)
1912                 goto out;
1913
1914         /*
1915          * Check if we have been instructed to close. We will have
1916          * deferred this request, because the bdev was still open.
1917          */
1918
1919         mutex_lock(&info->mutex);
1920         xbdev = info->xbdev;
1921
1922         if (xbdev && xbdev->state == XenbusStateClosing) {
1923                 /* pending switch to state closed */
1924                 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n");
1925                 xlvbd_release_gendisk(info);
1926                 xenbus_frontend_closed(info->xbdev);
1927         }
1928
1929         mutex_unlock(&info->mutex);
1930
1931         if (!xbdev) {
1932                 /* sudden device removal */
1933                 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n");
1934                 xlvbd_release_gendisk(info);
1935                 disk->private_data = NULL;
1936                 kfree(info);
1937         }
1938
1939 out:
1940         bdput(bdev);
1941         mutex_unlock(&blkfront_mutex);
1942 }
1943
1944 static const struct block_device_operations xlvbd_block_fops =
1945 {
1946         .owner = THIS_MODULE,
1947         .open = blkif_open,
1948         .release = blkif_release,
1949         .getgeo = blkif_getgeo,
1950         .ioctl = blkif_ioctl,
1951 };
1952
1953
1954 static const struct xenbus_device_id blkfront_ids[] = {
1955         { "vbd" },
1956         { "" }
1957 };
1958
1959 static DEFINE_XENBUS_DRIVER(blkfront, ,
1960         .probe = blkfront_probe,
1961         .remove = blkfront_remove,
1962         .resume = blkfront_resume,
1963         .otherend_changed = blkback_changed,
1964         .is_ready = blkfront_is_ready,
1965 );
1966
1967 static int __init xlblk_init(void)
1968 {
1969         int ret;
1970
1971         if (!xen_domain())
1972                 return -ENODEV;
1973
1974         if (xen_hvm_domain() && !xen_platform_pci_unplug)
1975                 return -ENODEV;
1976
1977         if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) {
1978                 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n",
1979                        XENVBD_MAJOR, DEV_NAME);
1980                 return -ENODEV;
1981         }
1982
1983         ret = xenbus_register_frontend(&blkfront_driver);
1984         if (ret) {
1985                 unregister_blkdev(XENVBD_MAJOR, DEV_NAME);
1986                 return ret;
1987         }
1988
1989         return 0;
1990 }
1991 module_init(xlblk_init);
1992
1993
1994 static void __exit xlblk_exit(void)
1995 {
1996         xenbus_unregister_driver(&blkfront_driver);
1997         unregister_blkdev(XENVBD_MAJOR, DEV_NAME);
1998         kfree(minors);
1999 }
2000 module_exit(xlblk_exit);
2001
2002 MODULE_DESCRIPTION("Xen virtual block device frontend");
2003 MODULE_LICENSE("GPL");
2004 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR);
2005 MODULE_ALIAS("xen:vbd");
2006 MODULE_ALIAS("xenblk");