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 48 #include <xen/xen.h> 49 #include <xen/xenbus.h> 50 #include <xen/grant_table.h> 51 #include <xen/events.h> 52 #include <xen/page.h> 53 #include <xen/platform_pci.h> 54 55 #include <xen/interface/grant_table.h> 56 #include <xen/interface/io/blkif.h> 57 #include <xen/interface/io/protocols.h> 58 59 #include <asm/xen/hypervisor.h> 60 61 enum blkif_state { 62 BLKIF_STATE_DISCONNECTED, 63 BLKIF_STATE_CONNECTED, 64 BLKIF_STATE_SUSPENDED, 65 }; 66 67 struct blk_shadow { 68 struct blkif_request req; 69 struct request *request; 70 unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 71 }; 72 73 static DEFINE_MUTEX(blkfront_mutex); 74 static const struct block_device_operations xlvbd_block_fops; 75 76 #define BLK_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE) 77 78 /* 79 * We have one of these per vbd, whether ide, scsi or 'other'. They 80 * hang in private_data off the gendisk structure. We may end up 81 * putting all kinds of interesting stuff here :-) 82 */ 83 struct blkfront_info 84 { 85 spinlock_t io_lock; 86 struct mutex mutex; 87 struct xenbus_device *xbdev; 88 struct gendisk *gd; 89 int vdevice; 90 blkif_vdev_t handle; 91 enum blkif_state connected; 92 int ring_ref; 93 struct blkif_front_ring ring; 94 struct scatterlist sg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 95 unsigned int evtchn, irq; 96 struct request_queue *rq; 97 struct work_struct work; 98 struct gnttab_free_callback callback; 99 struct blk_shadow shadow[BLK_RING_SIZE]; 100 unsigned long shadow_free; 101 unsigned int feature_flush; 102 unsigned int flush_op; 103 unsigned int feature_discard:1; 104 unsigned int feature_secdiscard:1; 105 unsigned int discard_granularity; 106 unsigned int discard_alignment; 107 int is_ready; 108 }; 109 110 static unsigned int nr_minors; 111 static unsigned long *minors; 112 static DEFINE_SPINLOCK(minor_lock); 113 114 #define MAXIMUM_OUTSTANDING_BLOCK_REQS \ 115 (BLKIF_MAX_SEGMENTS_PER_REQUEST * BLK_RING_SIZE) 116 #define GRANT_INVALID_REF 0 117 118 #define PARTS_PER_DISK 16 119 #define PARTS_PER_EXT_DISK 256 120 121 #define BLKIF_MAJOR(dev) ((dev)>>8) 122 #define BLKIF_MINOR(dev) ((dev) & 0xff) 123 124 #define EXT_SHIFT 28 125 #define EXTENDED (1<<EXT_SHIFT) 126 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED)) 127 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED)) 128 #define EMULATED_HD_DISK_MINOR_OFFSET (0) 129 #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256) 130 #define EMULATED_SD_DISK_MINOR_OFFSET (0) 131 #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256) 132 133 #define DEV_NAME "xvd" /* name in /dev */ 134 135 static int get_id_from_freelist(struct blkfront_info *info) 136 { 137 unsigned long free = info->shadow_free; 138 BUG_ON(free >= BLK_RING_SIZE); 139 info->shadow_free = info->shadow[free].req.u.rw.id; 140 info->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */ 141 return free; 142 } 143 144 static int add_id_to_freelist(struct blkfront_info *info, 145 unsigned long id) 146 { 147 if (info->shadow[id].req.u.rw.id != id) 148 return -EINVAL; 149 if (info->shadow[id].request == NULL) 150 return -EINVAL; 151 info->shadow[id].req.u.rw.id = info->shadow_free; 152 info->shadow[id].request = NULL; 153 info->shadow_free = id; 154 return 0; 155 } 156 157 static const char *op_name(int op) 158 { 159 static const char *const names[] = { 160 [BLKIF_OP_READ] = "read", 161 [BLKIF_OP_WRITE] = "write", 162 [BLKIF_OP_WRITE_BARRIER] = "barrier", 163 [BLKIF_OP_FLUSH_DISKCACHE] = "flush", 164 [BLKIF_OP_DISCARD] = "discard" }; 165 166 if (op < 0 || op >= ARRAY_SIZE(names)) 167 return "unknown"; 168 169 if (!names[op]) 170 return "reserved"; 171 172 return names[op]; 173 } 174 static int xlbd_reserve_minors(unsigned int minor, unsigned int nr) 175 { 176 unsigned int end = minor + nr; 177 int rc; 178 179 if (end > nr_minors) { 180 unsigned long *bitmap, *old; 181 182 bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap), 183 GFP_KERNEL); 184 if (bitmap == NULL) 185 return -ENOMEM; 186 187 spin_lock(&minor_lock); 188 if (end > nr_minors) { 189 old = minors; 190 memcpy(bitmap, minors, 191 BITS_TO_LONGS(nr_minors) * sizeof(*bitmap)); 192 minors = bitmap; 193 nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG; 194 } else 195 old = bitmap; 196 spin_unlock(&minor_lock); 197 kfree(old); 198 } 199 200 spin_lock(&minor_lock); 201 if (find_next_bit(minors, end, minor) >= end) { 202 bitmap_set(minors, minor, nr); 203 rc = 0; 204 } else 205 rc = -EBUSY; 206 spin_unlock(&minor_lock); 207 208 return rc; 209 } 210 211 static void xlbd_release_minors(unsigned int minor, unsigned int nr) 212 { 213 unsigned int end = minor + nr; 214 215 BUG_ON(end > nr_minors); 216 spin_lock(&minor_lock); 217 bitmap_clear(minors, minor, nr); 218 spin_unlock(&minor_lock); 219 } 220 221 static void blkif_restart_queue_callback(void *arg) 222 { 223 struct blkfront_info *info = (struct blkfront_info *)arg; 224 schedule_work(&info->work); 225 } 226 227 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg) 228 { 229 /* We don't have real geometry info, but let's at least return 230 values consistent with the size of the device */ 231 sector_t nsect = get_capacity(bd->bd_disk); 232 sector_t cylinders = nsect; 233 234 hg->heads = 0xff; 235 hg->sectors = 0x3f; 236 sector_div(cylinders, hg->heads * hg->sectors); 237 hg->cylinders = cylinders; 238 if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect) 239 hg->cylinders = 0xffff; 240 return 0; 241 } 242 243 static int blkif_ioctl(struct block_device *bdev, fmode_t mode, 244 unsigned command, unsigned long argument) 245 { 246 struct blkfront_info *info = bdev->bd_disk->private_data; 247 int i; 248 249 dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n", 250 command, (long)argument); 251 252 switch (command) { 253 case CDROMMULTISESSION: 254 dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n"); 255 for (i = 0; i < sizeof(struct cdrom_multisession); i++) 256 if (put_user(0, (char __user *)(argument + i))) 257 return -EFAULT; 258 return 0; 259 260 case CDROM_GET_CAPABILITY: { 261 struct gendisk *gd = info->gd; 262 if (gd->flags & GENHD_FL_CD) 263 return 0; 264 return -EINVAL; 265 } 266 267 default: 268 /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n", 269 command);*/ 270 return -EINVAL; /* same return as native Linux */ 271 } 272 273 return 0; 274 } 275 276 /* 277 * Generate a Xen blkfront IO request from a blk layer request. Reads 278 * and writes are handled as expected. 279 * 280 * @req: a request struct 281 */ 282 static int blkif_queue_request(struct request *req) 283 { 284 struct blkfront_info *info = req->rq_disk->private_data; 285 unsigned long buffer_mfn; 286 struct blkif_request *ring_req; 287 unsigned long id; 288 unsigned int fsect, lsect; 289 int i, ref; 290 grant_ref_t gref_head; 291 struct scatterlist *sg; 292 293 if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) 294 return 1; 295 296 if (gnttab_alloc_grant_references( 297 BLKIF_MAX_SEGMENTS_PER_REQUEST, &gref_head) < 0) { 298 gnttab_request_free_callback( 299 &info->callback, 300 blkif_restart_queue_callback, 301 info, 302 BLKIF_MAX_SEGMENTS_PER_REQUEST); 303 return 1; 304 } 305 306 /* Fill out a communications ring structure. */ 307 ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt); 308 id = get_id_from_freelist(info); 309 info->shadow[id].request = req; 310 311 ring_req->u.rw.id = id; 312 ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req); 313 ring_req->u.rw.handle = info->handle; 314 315 ring_req->operation = rq_data_dir(req) ? 316 BLKIF_OP_WRITE : BLKIF_OP_READ; 317 318 if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) { 319 /* 320 * Ideally we can do an unordered flush-to-disk. In case the 321 * backend onlysupports barriers, use that. A barrier request 322 * a superset of FUA, so we can implement it the same 323 * way. (It's also a FLUSH+FUA, since it is 324 * guaranteed ordered WRT previous writes.) 325 */ 326 ring_req->operation = info->flush_op; 327 } 328 329 if (unlikely(req->cmd_flags & (REQ_DISCARD | REQ_SECURE))) { 330 /* id, sector_number and handle are set above. */ 331 ring_req->operation = BLKIF_OP_DISCARD; 332 ring_req->u.discard.nr_sectors = blk_rq_sectors(req); 333 if ((req->cmd_flags & REQ_SECURE) && info->feature_secdiscard) 334 ring_req->u.discard.flag = BLKIF_DISCARD_SECURE; 335 else 336 ring_req->u.discard.flag = 0; 337 } else { 338 ring_req->u.rw.nr_segments = blk_rq_map_sg(req->q, req, 339 info->sg); 340 BUG_ON(ring_req->u.rw.nr_segments > 341 BLKIF_MAX_SEGMENTS_PER_REQUEST); 342 343 for_each_sg(info->sg, sg, ring_req->u.rw.nr_segments, i) { 344 buffer_mfn = pfn_to_mfn(page_to_pfn(sg_page(sg))); 345 fsect = sg->offset >> 9; 346 lsect = fsect + (sg->length >> 9) - 1; 347 /* install a grant reference. */ 348 ref = gnttab_claim_grant_reference(&gref_head); 349 BUG_ON(ref == -ENOSPC); 350 351 gnttab_grant_foreign_access_ref( 352 ref, 353 info->xbdev->otherend_id, 354 buffer_mfn, 355 rq_data_dir(req)); 356 357 info->shadow[id].frame[i] = mfn_to_pfn(buffer_mfn); 358 ring_req->u.rw.seg[i] = 359 (struct blkif_request_segment) { 360 .gref = ref, 361 .first_sect = fsect, 362 .last_sect = lsect }; 363 } 364 } 365 366 info->ring.req_prod_pvt++; 367 368 /* Keep a private copy so we can reissue requests when recovering. */ 369 info->shadow[id].req = *ring_req; 370 371 gnttab_free_grant_references(gref_head); 372 373 return 0; 374 } 375 376 377 static inline void flush_requests(struct blkfront_info *info) 378 { 379 int notify; 380 381 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&info->ring, notify); 382 383 if (notify) 384 notify_remote_via_irq(info->irq); 385 } 386 387 /* 388 * do_blkif_request 389 * read a block; request is in a request queue 390 */ 391 static void do_blkif_request(struct request_queue *rq) 392 { 393 struct blkfront_info *info = NULL; 394 struct request *req; 395 int queued; 396 397 pr_debug("Entered do_blkif_request\n"); 398 399 queued = 0; 400 401 while ((req = blk_peek_request(rq)) != NULL) { 402 info = req->rq_disk->private_data; 403 404 if (RING_FULL(&info->ring)) 405 goto wait; 406 407 blk_start_request(req); 408 409 if ((req->cmd_type != REQ_TYPE_FS) || 410 ((req->cmd_flags & (REQ_FLUSH | REQ_FUA)) && 411 !info->flush_op)) { 412 __blk_end_request_all(req, -EIO); 413 continue; 414 } 415 416 pr_debug("do_blk_req %p: cmd %p, sec %lx, " 417 "(%u/%u) buffer:%p [%s]\n", 418 req, req->cmd, (unsigned long)blk_rq_pos(req), 419 blk_rq_cur_sectors(req), blk_rq_sectors(req), 420 req->buffer, rq_data_dir(req) ? "write" : "read"); 421 422 if (blkif_queue_request(req)) { 423 blk_requeue_request(rq, req); 424 wait: 425 /* Avoid pointless unplugs. */ 426 blk_stop_queue(rq); 427 break; 428 } 429 430 queued++; 431 } 432 433 if (queued != 0) 434 flush_requests(info); 435 } 436 437 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size) 438 { 439 struct request_queue *rq; 440 struct blkfront_info *info = gd->private_data; 441 442 rq = blk_init_queue(do_blkif_request, &info->io_lock); 443 if (rq == NULL) 444 return -1; 445 446 queue_flag_set_unlocked(QUEUE_FLAG_VIRT, rq); 447 448 if (info->feature_discard) { 449 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, rq); 450 blk_queue_max_discard_sectors(rq, get_capacity(gd)); 451 rq->limits.discard_granularity = info->discard_granularity; 452 rq->limits.discard_alignment = info->discard_alignment; 453 if (info->feature_secdiscard) 454 queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, rq); 455 } 456 457 /* Hard sector size and max sectors impersonate the equiv. hardware. */ 458 blk_queue_logical_block_size(rq, sector_size); 459 blk_queue_max_hw_sectors(rq, 512); 460 461 /* Each segment in a request is up to an aligned page in size. */ 462 blk_queue_segment_boundary(rq, PAGE_SIZE - 1); 463 blk_queue_max_segment_size(rq, PAGE_SIZE); 464 465 /* Ensure a merged request will fit in a single I/O ring slot. */ 466 blk_queue_max_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST); 467 468 /* Make sure buffer addresses are sector-aligned. */ 469 blk_queue_dma_alignment(rq, 511); 470 471 /* Make sure we don't use bounce buffers. */ 472 blk_queue_bounce_limit(rq, BLK_BOUNCE_ANY); 473 474 gd->queue = rq; 475 476 return 0; 477 } 478 479 480 static void xlvbd_flush(struct blkfront_info *info) 481 { 482 blk_queue_flush(info->rq, info->feature_flush); 483 printk(KERN_INFO "blkfront: %s: %s: %s\n", 484 info->gd->disk_name, 485 info->flush_op == BLKIF_OP_WRITE_BARRIER ? 486 "barrier" : (info->flush_op == BLKIF_OP_FLUSH_DISKCACHE ? 487 "flush diskcache" : "barrier or flush"), 488 info->feature_flush ? "enabled" : "disabled"); 489 } 490 491 static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset) 492 { 493 int major; 494 major = BLKIF_MAJOR(vdevice); 495 *minor = BLKIF_MINOR(vdevice); 496 switch (major) { 497 case XEN_IDE0_MAJOR: 498 *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET; 499 *minor = ((*minor / 64) * PARTS_PER_DISK) + 500 EMULATED_HD_DISK_MINOR_OFFSET; 501 break; 502 case XEN_IDE1_MAJOR: 503 *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET; 504 *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) + 505 EMULATED_HD_DISK_MINOR_OFFSET; 506 break; 507 case XEN_SCSI_DISK0_MAJOR: 508 *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET; 509 *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET; 510 break; 511 case XEN_SCSI_DISK1_MAJOR: 512 case XEN_SCSI_DISK2_MAJOR: 513 case XEN_SCSI_DISK3_MAJOR: 514 case XEN_SCSI_DISK4_MAJOR: 515 case XEN_SCSI_DISK5_MAJOR: 516 case XEN_SCSI_DISK6_MAJOR: 517 case XEN_SCSI_DISK7_MAJOR: 518 *offset = (*minor / PARTS_PER_DISK) + 519 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) + 520 EMULATED_SD_DISK_NAME_OFFSET; 521 *minor = *minor + 522 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) + 523 EMULATED_SD_DISK_MINOR_OFFSET; 524 break; 525 case XEN_SCSI_DISK8_MAJOR: 526 case XEN_SCSI_DISK9_MAJOR: 527 case XEN_SCSI_DISK10_MAJOR: 528 case XEN_SCSI_DISK11_MAJOR: 529 case XEN_SCSI_DISK12_MAJOR: 530 case XEN_SCSI_DISK13_MAJOR: 531 case XEN_SCSI_DISK14_MAJOR: 532 case XEN_SCSI_DISK15_MAJOR: 533 *offset = (*minor / PARTS_PER_DISK) + 534 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) + 535 EMULATED_SD_DISK_NAME_OFFSET; 536 *minor = *minor + 537 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) + 538 EMULATED_SD_DISK_MINOR_OFFSET; 539 break; 540 case XENVBD_MAJOR: 541 *offset = *minor / PARTS_PER_DISK; 542 break; 543 default: 544 printk(KERN_WARNING "blkfront: your disk configuration is " 545 "incorrect, please use an xvd device instead\n"); 546 return -ENODEV; 547 } 548 return 0; 549 } 550 551 static char *encode_disk_name(char *ptr, unsigned int n) 552 { 553 if (n >= 26) 554 ptr = encode_disk_name(ptr, n / 26 - 1); 555 *ptr = 'a' + n % 26; 556 return ptr + 1; 557 } 558 559 static int xlvbd_alloc_gendisk(blkif_sector_t capacity, 560 struct blkfront_info *info, 561 u16 vdisk_info, u16 sector_size) 562 { 563 struct gendisk *gd; 564 int nr_minors = 1; 565 int err; 566 unsigned int offset; 567 int minor; 568 int nr_parts; 569 char *ptr; 570 571 BUG_ON(info->gd != NULL); 572 BUG_ON(info->rq != NULL); 573 574 if ((info->vdevice>>EXT_SHIFT) > 1) { 575 /* this is above the extended range; something is wrong */ 576 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice); 577 return -ENODEV; 578 } 579 580 if (!VDEV_IS_EXTENDED(info->vdevice)) { 581 err = xen_translate_vdev(info->vdevice, &minor, &offset); 582 if (err) 583 return err; 584 nr_parts = PARTS_PER_DISK; 585 } else { 586 minor = BLKIF_MINOR_EXT(info->vdevice); 587 nr_parts = PARTS_PER_EXT_DISK; 588 offset = minor / nr_parts; 589 if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4) 590 printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with " 591 "emulated IDE disks,\n\t choose an xvd device name" 592 "from xvde on\n", info->vdevice); 593 } 594 if (minor >> MINORBITS) { 595 pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n", 596 info->vdevice, minor); 597 return -ENODEV; 598 } 599 600 if ((minor % nr_parts) == 0) 601 nr_minors = nr_parts; 602 603 err = xlbd_reserve_minors(minor, nr_minors); 604 if (err) 605 goto out; 606 err = -ENODEV; 607 608 gd = alloc_disk(nr_minors); 609 if (gd == NULL) 610 goto release; 611 612 strcpy(gd->disk_name, DEV_NAME); 613 ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset); 614 BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN); 615 if (nr_minors > 1) 616 *ptr = 0; 617 else 618 snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr, 619 "%d", minor & (nr_parts - 1)); 620 621 gd->major = XENVBD_MAJOR; 622 gd->first_minor = minor; 623 gd->fops = &xlvbd_block_fops; 624 gd->private_data = info; 625 gd->driverfs_dev = &(info->xbdev->dev); 626 set_capacity(gd, capacity); 627 628 if (xlvbd_init_blk_queue(gd, sector_size)) { 629 del_gendisk(gd); 630 goto release; 631 } 632 633 info->rq = gd->queue; 634 info->gd = gd; 635 636 xlvbd_flush(info); 637 638 if (vdisk_info & VDISK_READONLY) 639 set_disk_ro(gd, 1); 640 641 if (vdisk_info & VDISK_REMOVABLE) 642 gd->flags |= GENHD_FL_REMOVABLE; 643 644 if (vdisk_info & VDISK_CDROM) 645 gd->flags |= GENHD_FL_CD; 646 647 return 0; 648 649 release: 650 xlbd_release_minors(minor, nr_minors); 651 out: 652 return err; 653 } 654 655 static void xlvbd_release_gendisk(struct blkfront_info *info) 656 { 657 unsigned int minor, nr_minors; 658 unsigned long flags; 659 660 if (info->rq == NULL) 661 return; 662 663 spin_lock_irqsave(&info->io_lock, flags); 664 665 /* No more blkif_request(). */ 666 blk_stop_queue(info->rq); 667 668 /* No more gnttab callback work. */ 669 gnttab_cancel_free_callback(&info->callback); 670 spin_unlock_irqrestore(&info->io_lock, flags); 671 672 /* Flush gnttab callback work. Must be done with no locks held. */ 673 flush_work_sync(&info->work); 674 675 del_gendisk(info->gd); 676 677 minor = info->gd->first_minor; 678 nr_minors = info->gd->minors; 679 xlbd_release_minors(minor, nr_minors); 680 681 blk_cleanup_queue(info->rq); 682 info->rq = NULL; 683 684 put_disk(info->gd); 685 info->gd = NULL; 686 } 687 688 static void kick_pending_request_queues(struct blkfront_info *info) 689 { 690 if (!RING_FULL(&info->ring)) { 691 /* Re-enable calldowns. */ 692 blk_start_queue(info->rq); 693 /* Kick things off immediately. */ 694 do_blkif_request(info->rq); 695 } 696 } 697 698 static void blkif_restart_queue(struct work_struct *work) 699 { 700 struct blkfront_info *info = container_of(work, struct blkfront_info, work); 701 702 spin_lock_irq(&info->io_lock); 703 if (info->connected == BLKIF_STATE_CONNECTED) 704 kick_pending_request_queues(info); 705 spin_unlock_irq(&info->io_lock); 706 } 707 708 static void blkif_free(struct blkfront_info *info, int suspend) 709 { 710 /* Prevent new requests being issued until we fix things up. */ 711 spin_lock_irq(&info->io_lock); 712 info->connected = suspend ? 713 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED; 714 /* No more blkif_request(). */ 715 if (info->rq) 716 blk_stop_queue(info->rq); 717 /* No more gnttab callback work. */ 718 gnttab_cancel_free_callback(&info->callback); 719 spin_unlock_irq(&info->io_lock); 720 721 /* Flush gnttab callback work. Must be done with no locks held. */ 722 flush_work_sync(&info->work); 723 724 /* Free resources associated with old device channel. */ 725 if (info->ring_ref != GRANT_INVALID_REF) { 726 gnttab_end_foreign_access(info->ring_ref, 0, 727 (unsigned long)info->ring.sring); 728 info->ring_ref = GRANT_INVALID_REF; 729 info->ring.sring = NULL; 730 } 731 if (info->irq) 732 unbind_from_irqhandler(info->irq, info); 733 info->evtchn = info->irq = 0; 734 735 } 736 737 static void blkif_completion(struct blk_shadow *s) 738 { 739 int i; 740 /* Do not let BLKIF_OP_DISCARD as nr_segment is in the same place 741 * flag. */ 742 for (i = 0; i < s->req.u.rw.nr_segments; i++) 743 gnttab_end_foreign_access(s->req.u.rw.seg[i].gref, 0, 0UL); 744 } 745 746 static irqreturn_t blkif_interrupt(int irq, void *dev_id) 747 { 748 struct request *req; 749 struct blkif_response *bret; 750 RING_IDX i, rp; 751 unsigned long flags; 752 struct blkfront_info *info = (struct blkfront_info *)dev_id; 753 int error; 754 755 spin_lock_irqsave(&info->io_lock, flags); 756 757 if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) { 758 spin_unlock_irqrestore(&info->io_lock, flags); 759 return IRQ_HANDLED; 760 } 761 762 again: 763 rp = info->ring.sring->rsp_prod; 764 rmb(); /* Ensure we see queued responses up to 'rp'. */ 765 766 for (i = info->ring.rsp_cons; i != rp; i++) { 767 unsigned long id; 768 769 bret = RING_GET_RESPONSE(&info->ring, i); 770 id = bret->id; 771 /* 772 * The backend has messed up and given us an id that we would 773 * never have given to it (we stamp it up to BLK_RING_SIZE - 774 * look in get_id_from_freelist. 775 */ 776 if (id >= BLK_RING_SIZE) { 777 WARN(1, "%s: response to %s has incorrect id (%ld)\n", 778 info->gd->disk_name, op_name(bret->operation), id); 779 /* We can't safely get the 'struct request' as 780 * the id is busted. */ 781 continue; 782 } 783 req = info->shadow[id].request; 784 785 if (bret->operation != BLKIF_OP_DISCARD) 786 blkif_completion(&info->shadow[id]); 787 788 if (add_id_to_freelist(info, id)) { 789 WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n", 790 info->gd->disk_name, op_name(bret->operation), id); 791 continue; 792 } 793 794 error = (bret->status == BLKIF_RSP_OKAY) ? 0 : -EIO; 795 switch (bret->operation) { 796 case BLKIF_OP_DISCARD: 797 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { 798 struct request_queue *rq = info->rq; 799 printk(KERN_WARNING "blkfront: %s: %s op failed\n", 800 info->gd->disk_name, op_name(bret->operation)); 801 error = -EOPNOTSUPP; 802 info->feature_discard = 0; 803 info->feature_secdiscard = 0; 804 queue_flag_clear(QUEUE_FLAG_DISCARD, rq); 805 queue_flag_clear(QUEUE_FLAG_SECDISCARD, rq); 806 } 807 __blk_end_request_all(req, error); 808 break; 809 case BLKIF_OP_FLUSH_DISKCACHE: 810 case BLKIF_OP_WRITE_BARRIER: 811 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { 812 printk(KERN_WARNING "blkfront: %s: %s op failed\n", 813 info->gd->disk_name, op_name(bret->operation)); 814 error = -EOPNOTSUPP; 815 } 816 if (unlikely(bret->status == BLKIF_RSP_ERROR && 817 info->shadow[id].req.u.rw.nr_segments == 0)) { 818 printk(KERN_WARNING "blkfront: %s: empty %s op failed\n", 819 info->gd->disk_name, op_name(bret->operation)); 820 error = -EOPNOTSUPP; 821 } 822 if (unlikely(error)) { 823 if (error == -EOPNOTSUPP) 824 error = 0; 825 info->feature_flush = 0; 826 info->flush_op = 0; 827 xlvbd_flush(info); 828 } 829 /* fall through */ 830 case BLKIF_OP_READ: 831 case BLKIF_OP_WRITE: 832 if (unlikely(bret->status != BLKIF_RSP_OKAY)) 833 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data " 834 "request: %x\n", bret->status); 835 836 __blk_end_request_all(req, error); 837 break; 838 default: 839 BUG(); 840 } 841 } 842 843 info->ring.rsp_cons = i; 844 845 if (i != info->ring.req_prod_pvt) { 846 int more_to_do; 847 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do); 848 if (more_to_do) 849 goto again; 850 } else 851 info->ring.sring->rsp_event = i + 1; 852 853 kick_pending_request_queues(info); 854 855 spin_unlock_irqrestore(&info->io_lock, flags); 856 857 return IRQ_HANDLED; 858 } 859 860 861 static int setup_blkring(struct xenbus_device *dev, 862 struct blkfront_info *info) 863 { 864 struct blkif_sring *sring; 865 int err; 866 867 info->ring_ref = GRANT_INVALID_REF; 868 869 sring = (struct blkif_sring *)__get_free_page(GFP_NOIO | __GFP_HIGH); 870 if (!sring) { 871 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring"); 872 return -ENOMEM; 873 } 874 SHARED_RING_INIT(sring); 875 FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE); 876 877 sg_init_table(info->sg, BLKIF_MAX_SEGMENTS_PER_REQUEST); 878 879 err = xenbus_grant_ring(dev, virt_to_mfn(info->ring.sring)); 880 if (err < 0) { 881 free_page((unsigned long)sring); 882 info->ring.sring = NULL; 883 goto fail; 884 } 885 info->ring_ref = err; 886 887 err = xenbus_alloc_evtchn(dev, &info->evtchn); 888 if (err) 889 goto fail; 890 891 err = bind_evtchn_to_irqhandler(info->evtchn, 892 blkif_interrupt, 893 IRQF_SAMPLE_RANDOM, "blkif", info); 894 if (err <= 0) { 895 xenbus_dev_fatal(dev, err, 896 "bind_evtchn_to_irqhandler failed"); 897 goto fail; 898 } 899 info->irq = err; 900 901 return 0; 902 fail: 903 blkif_free(info, 0); 904 return err; 905 } 906 907 908 /* Common code used when first setting up, and when resuming. */ 909 static int talk_to_blkback(struct xenbus_device *dev, 910 struct blkfront_info *info) 911 { 912 const char *message = NULL; 913 struct xenbus_transaction xbt; 914 int err; 915 916 /* Create shared ring, alloc event channel. */ 917 err = setup_blkring(dev, info); 918 if (err) 919 goto out; 920 921 again: 922 err = xenbus_transaction_start(&xbt); 923 if (err) { 924 xenbus_dev_fatal(dev, err, "starting transaction"); 925 goto destroy_blkring; 926 } 927 928 err = xenbus_printf(xbt, dev->nodename, 929 "ring-ref", "%u", info->ring_ref); 930 if (err) { 931 message = "writing ring-ref"; 932 goto abort_transaction; 933 } 934 err = xenbus_printf(xbt, dev->nodename, 935 "event-channel", "%u", info->evtchn); 936 if (err) { 937 message = "writing event-channel"; 938 goto abort_transaction; 939 } 940 err = xenbus_printf(xbt, dev->nodename, "protocol", "%s", 941 XEN_IO_PROTO_ABI_NATIVE); 942 if (err) { 943 message = "writing protocol"; 944 goto abort_transaction; 945 } 946 947 err = xenbus_transaction_end(xbt, 0); 948 if (err) { 949 if (err == -EAGAIN) 950 goto again; 951 xenbus_dev_fatal(dev, err, "completing transaction"); 952 goto destroy_blkring; 953 } 954 955 xenbus_switch_state(dev, XenbusStateInitialised); 956 957 return 0; 958 959 abort_transaction: 960 xenbus_transaction_end(xbt, 1); 961 if (message) 962 xenbus_dev_fatal(dev, err, "%s", message); 963 destroy_blkring: 964 blkif_free(info, 0); 965 out: 966 return err; 967 } 968 969 /** 970 * Entry point to this code when a new device is created. Allocate the basic 971 * structures and the ring buffer for communication with the backend, and 972 * inform the backend of the appropriate details for those. Switch to 973 * Initialised state. 974 */ 975 static int blkfront_probe(struct xenbus_device *dev, 976 const struct xenbus_device_id *id) 977 { 978 int err, vdevice, i; 979 struct blkfront_info *info; 980 981 /* FIXME: Use dynamic device id if this is not set. */ 982 err = xenbus_scanf(XBT_NIL, dev->nodename, 983 "virtual-device", "%i", &vdevice); 984 if (err != 1) { 985 /* go looking in the extended area instead */ 986 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext", 987 "%i", &vdevice); 988 if (err != 1) { 989 xenbus_dev_fatal(dev, err, "reading virtual-device"); 990 return err; 991 } 992 } 993 994 if (xen_hvm_domain()) { 995 char *type; 996 int len; 997 /* no unplug has been done: do not hook devices != xen vbds */ 998 if (xen_platform_pci_unplug & XEN_UNPLUG_UNNECESSARY) { 999 int major; 1000 1001 if (!VDEV_IS_EXTENDED(vdevice)) 1002 major = BLKIF_MAJOR(vdevice); 1003 else 1004 major = XENVBD_MAJOR; 1005 1006 if (major != XENVBD_MAJOR) { 1007 printk(KERN_INFO 1008 "%s: HVM does not support vbd %d as xen block device\n", 1009 __FUNCTION__, vdevice); 1010 return -ENODEV; 1011 } 1012 } 1013 /* do not create a PV cdrom device if we are an HVM guest */ 1014 type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len); 1015 if (IS_ERR(type)) 1016 return -ENODEV; 1017 if (strncmp(type, "cdrom", 5) == 0) { 1018 kfree(type); 1019 return -ENODEV; 1020 } 1021 kfree(type); 1022 } 1023 info = kzalloc(sizeof(*info), GFP_KERNEL); 1024 if (!info) { 1025 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure"); 1026 return -ENOMEM; 1027 } 1028 1029 mutex_init(&info->mutex); 1030 spin_lock_init(&info->io_lock); 1031 info->xbdev = dev; 1032 info->vdevice = vdevice; 1033 info->connected = BLKIF_STATE_DISCONNECTED; 1034 INIT_WORK(&info->work, blkif_restart_queue); 1035 1036 for (i = 0; i < BLK_RING_SIZE; i++) 1037 info->shadow[i].req.u.rw.id = i+1; 1038 info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff; 1039 1040 /* Front end dir is a number, which is used as the id. */ 1041 info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0); 1042 dev_set_drvdata(&dev->dev, info); 1043 1044 err = talk_to_blkback(dev, info); 1045 if (err) { 1046 kfree(info); 1047 dev_set_drvdata(&dev->dev, NULL); 1048 return err; 1049 } 1050 1051 return 0; 1052 } 1053 1054 1055 static int blkif_recover(struct blkfront_info *info) 1056 { 1057 int i; 1058 struct blkif_request *req; 1059 struct blk_shadow *copy; 1060 int j; 1061 1062 /* Stage 1: Make a safe copy of the shadow state. */ 1063 copy = kmalloc(sizeof(info->shadow), 1064 GFP_NOIO | __GFP_REPEAT | __GFP_HIGH); 1065 if (!copy) 1066 return -ENOMEM; 1067 memcpy(copy, info->shadow, sizeof(info->shadow)); 1068 1069 /* Stage 2: Set up free list. */ 1070 memset(&info->shadow, 0, sizeof(info->shadow)); 1071 for (i = 0; i < BLK_RING_SIZE; i++) 1072 info->shadow[i].req.u.rw.id = i+1; 1073 info->shadow_free = info->ring.req_prod_pvt; 1074 info->shadow[BLK_RING_SIZE-1].req.u.rw.id = 0x0fffffff; 1075 1076 /* Stage 3: Find pending requests and requeue them. */ 1077 for (i = 0; i < BLK_RING_SIZE; i++) { 1078 /* Not in use? */ 1079 if (!copy[i].request) 1080 continue; 1081 1082 /* Grab a request slot and copy shadow state into it. */ 1083 req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt); 1084 *req = copy[i].req; 1085 1086 /* We get a new request id, and must reset the shadow state. */ 1087 req->u.rw.id = get_id_from_freelist(info); 1088 memcpy(&info->shadow[req->u.rw.id], ©[i], sizeof(copy[i])); 1089 1090 if (req->operation != BLKIF_OP_DISCARD) { 1091 /* Rewrite any grant references invalidated by susp/resume. */ 1092 for (j = 0; j < req->u.rw.nr_segments; j++) 1093 gnttab_grant_foreign_access_ref( 1094 req->u.rw.seg[j].gref, 1095 info->xbdev->otherend_id, 1096 pfn_to_mfn(info->shadow[req->u.rw.id].frame[j]), 1097 rq_data_dir(info->shadow[req->u.rw.id].request)); 1098 } 1099 info->shadow[req->u.rw.id].req = *req; 1100 1101 info->ring.req_prod_pvt++; 1102 } 1103 1104 kfree(copy); 1105 1106 xenbus_switch_state(info->xbdev, XenbusStateConnected); 1107 1108 spin_lock_irq(&info->io_lock); 1109 1110 /* Now safe for us to use the shared ring */ 1111 info->connected = BLKIF_STATE_CONNECTED; 1112 1113 /* Send off requeued requests */ 1114 flush_requests(info); 1115 1116 /* Kick any other new requests queued since we resumed */ 1117 kick_pending_request_queues(info); 1118 1119 spin_unlock_irq(&info->io_lock); 1120 1121 return 0; 1122 } 1123 1124 /** 1125 * We are reconnecting to the backend, due to a suspend/resume, or a backend 1126 * driver restart. We tear down our blkif structure and recreate it, but 1127 * leave the device-layer structures intact so that this is transparent to the 1128 * rest of the kernel. 1129 */ 1130 static int blkfront_resume(struct xenbus_device *dev) 1131 { 1132 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 1133 int err; 1134 1135 dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename); 1136 1137 blkif_free(info, info->connected == BLKIF_STATE_CONNECTED); 1138 1139 err = talk_to_blkback(dev, info); 1140 if (info->connected == BLKIF_STATE_SUSPENDED && !err) 1141 err = blkif_recover(info); 1142 1143 return err; 1144 } 1145 1146 static void 1147 blkfront_closing(struct blkfront_info *info) 1148 { 1149 struct xenbus_device *xbdev = info->xbdev; 1150 struct block_device *bdev = NULL; 1151 1152 mutex_lock(&info->mutex); 1153 1154 if (xbdev->state == XenbusStateClosing) { 1155 mutex_unlock(&info->mutex); 1156 return; 1157 } 1158 1159 if (info->gd) 1160 bdev = bdget_disk(info->gd, 0); 1161 1162 mutex_unlock(&info->mutex); 1163 1164 if (!bdev) { 1165 xenbus_frontend_closed(xbdev); 1166 return; 1167 } 1168 1169 mutex_lock(&bdev->bd_mutex); 1170 1171 if (bdev->bd_openers) { 1172 xenbus_dev_error(xbdev, -EBUSY, 1173 "Device in use; refusing to close"); 1174 xenbus_switch_state(xbdev, XenbusStateClosing); 1175 } else { 1176 xlvbd_release_gendisk(info); 1177 xenbus_frontend_closed(xbdev); 1178 } 1179 1180 mutex_unlock(&bdev->bd_mutex); 1181 bdput(bdev); 1182 } 1183 1184 static void blkfront_setup_discard(struct blkfront_info *info) 1185 { 1186 int err; 1187 char *type; 1188 unsigned int discard_granularity; 1189 unsigned int discard_alignment; 1190 unsigned int discard_secure; 1191 1192 type = xenbus_read(XBT_NIL, info->xbdev->otherend, "type", NULL); 1193 if (IS_ERR(type)) 1194 return; 1195 1196 info->feature_secdiscard = 0; 1197 if (strncmp(type, "phy", 3) == 0) { 1198 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 1199 "discard-granularity", "%u", &discard_granularity, 1200 "discard-alignment", "%u", &discard_alignment, 1201 NULL); 1202 if (!err) { 1203 info->feature_discard = 1; 1204 info->discard_granularity = discard_granularity; 1205 info->discard_alignment = discard_alignment; 1206 } 1207 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 1208 "discard-secure", "%d", &discard_secure, 1209 NULL); 1210 if (!err) 1211 info->feature_secdiscard = discard_secure; 1212 1213 } else if (strncmp(type, "file", 4) == 0) 1214 info->feature_discard = 1; 1215 1216 kfree(type); 1217 } 1218 1219 /* 1220 * Invoked when the backend is finally 'ready' (and has told produced 1221 * the details about the physical device - #sectors, size, etc). 1222 */ 1223 static void blkfront_connect(struct blkfront_info *info) 1224 { 1225 unsigned long long sectors; 1226 unsigned long sector_size; 1227 unsigned int binfo; 1228 int err; 1229 int barrier, flush, discard; 1230 1231 switch (info->connected) { 1232 case BLKIF_STATE_CONNECTED: 1233 /* 1234 * Potentially, the back-end may be signalling 1235 * a capacity change; update the capacity. 1236 */ 1237 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, 1238 "sectors", "%Lu", §ors); 1239 if (XENBUS_EXIST_ERR(err)) 1240 return; 1241 printk(KERN_INFO "Setting capacity to %Lu\n", 1242 sectors); 1243 set_capacity(info->gd, sectors); 1244 revalidate_disk(info->gd); 1245 1246 /* fall through */ 1247 case BLKIF_STATE_SUSPENDED: 1248 return; 1249 1250 default: 1251 break; 1252 } 1253 1254 dev_dbg(&info->xbdev->dev, "%s:%s.\n", 1255 __func__, info->xbdev->otherend); 1256 1257 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 1258 "sectors", "%llu", §ors, 1259 "info", "%u", &binfo, 1260 "sector-size", "%lu", §or_size, 1261 NULL); 1262 if (err) { 1263 xenbus_dev_fatal(info->xbdev, err, 1264 "reading backend fields at %s", 1265 info->xbdev->otherend); 1266 return; 1267 } 1268 1269 info->feature_flush = 0; 1270 info->flush_op = 0; 1271 1272 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 1273 "feature-barrier", "%d", &barrier, 1274 NULL); 1275 1276 /* 1277 * If there's no "feature-barrier" defined, then it means 1278 * we're dealing with a very old backend which writes 1279 * synchronously; nothing to do. 1280 * 1281 * If there are barriers, then we use flush. 1282 */ 1283 if (!err && barrier) { 1284 info->feature_flush = REQ_FLUSH | REQ_FUA; 1285 info->flush_op = BLKIF_OP_WRITE_BARRIER; 1286 } 1287 /* 1288 * And if there is "feature-flush-cache" use that above 1289 * barriers. 1290 */ 1291 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 1292 "feature-flush-cache", "%d", &flush, 1293 NULL); 1294 1295 if (!err && flush) { 1296 info->feature_flush = REQ_FLUSH; 1297 info->flush_op = BLKIF_OP_FLUSH_DISKCACHE; 1298 } 1299 1300 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 1301 "feature-discard", "%d", &discard, 1302 NULL); 1303 1304 if (!err && discard) 1305 blkfront_setup_discard(info); 1306 1307 err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size); 1308 if (err) { 1309 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s", 1310 info->xbdev->otherend); 1311 return; 1312 } 1313 1314 xenbus_switch_state(info->xbdev, XenbusStateConnected); 1315 1316 /* Kick pending requests. */ 1317 spin_lock_irq(&info->io_lock); 1318 info->connected = BLKIF_STATE_CONNECTED; 1319 kick_pending_request_queues(info); 1320 spin_unlock_irq(&info->io_lock); 1321 1322 add_disk(info->gd); 1323 1324 info->is_ready = 1; 1325 } 1326 1327 /** 1328 * Callback received when the backend's state changes. 1329 */ 1330 static void blkback_changed(struct xenbus_device *dev, 1331 enum xenbus_state backend_state) 1332 { 1333 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 1334 1335 dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state); 1336 1337 switch (backend_state) { 1338 case XenbusStateInitialising: 1339 case XenbusStateInitWait: 1340 case XenbusStateInitialised: 1341 case XenbusStateReconfiguring: 1342 case XenbusStateReconfigured: 1343 case XenbusStateUnknown: 1344 case XenbusStateClosed: 1345 break; 1346 1347 case XenbusStateConnected: 1348 blkfront_connect(info); 1349 break; 1350 1351 case XenbusStateClosing: 1352 blkfront_closing(info); 1353 break; 1354 } 1355 } 1356 1357 static int blkfront_remove(struct xenbus_device *xbdev) 1358 { 1359 struct blkfront_info *info = dev_get_drvdata(&xbdev->dev); 1360 struct block_device *bdev = NULL; 1361 struct gendisk *disk; 1362 1363 dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename); 1364 1365 blkif_free(info, 0); 1366 1367 mutex_lock(&info->mutex); 1368 1369 disk = info->gd; 1370 if (disk) 1371 bdev = bdget_disk(disk, 0); 1372 1373 info->xbdev = NULL; 1374 mutex_unlock(&info->mutex); 1375 1376 if (!bdev) { 1377 kfree(info); 1378 return 0; 1379 } 1380 1381 /* 1382 * The xbdev was removed before we reached the Closed 1383 * state. See if it's safe to remove the disk. If the bdev 1384 * isn't closed yet, we let release take care of it. 1385 */ 1386 1387 mutex_lock(&bdev->bd_mutex); 1388 info = disk->private_data; 1389 1390 dev_warn(disk_to_dev(disk), 1391 "%s was hot-unplugged, %d stale handles\n", 1392 xbdev->nodename, bdev->bd_openers); 1393 1394 if (info && !bdev->bd_openers) { 1395 xlvbd_release_gendisk(info); 1396 disk->private_data = NULL; 1397 kfree(info); 1398 } 1399 1400 mutex_unlock(&bdev->bd_mutex); 1401 bdput(bdev); 1402 1403 return 0; 1404 } 1405 1406 static int blkfront_is_ready(struct xenbus_device *dev) 1407 { 1408 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 1409 1410 return info->is_ready && info->xbdev; 1411 } 1412 1413 static int blkif_open(struct block_device *bdev, fmode_t mode) 1414 { 1415 struct gendisk *disk = bdev->bd_disk; 1416 struct blkfront_info *info; 1417 int err = 0; 1418 1419 mutex_lock(&blkfront_mutex); 1420 1421 info = disk->private_data; 1422 if (!info) { 1423 /* xbdev gone */ 1424 err = -ERESTARTSYS; 1425 goto out; 1426 } 1427 1428 mutex_lock(&info->mutex); 1429 1430 if (!info->gd) 1431 /* xbdev is closed */ 1432 err = -ERESTARTSYS; 1433 1434 mutex_unlock(&info->mutex); 1435 1436 out: 1437 mutex_unlock(&blkfront_mutex); 1438 return err; 1439 } 1440 1441 static int blkif_release(struct gendisk *disk, fmode_t mode) 1442 { 1443 struct blkfront_info *info = disk->private_data; 1444 struct block_device *bdev; 1445 struct xenbus_device *xbdev; 1446 1447 mutex_lock(&blkfront_mutex); 1448 1449 bdev = bdget_disk(disk, 0); 1450 1451 if (bdev->bd_openers) 1452 goto out; 1453 1454 /* 1455 * Check if we have been instructed to close. We will have 1456 * deferred this request, because the bdev was still open. 1457 */ 1458 1459 mutex_lock(&info->mutex); 1460 xbdev = info->xbdev; 1461 1462 if (xbdev && xbdev->state == XenbusStateClosing) { 1463 /* pending switch to state closed */ 1464 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); 1465 xlvbd_release_gendisk(info); 1466 xenbus_frontend_closed(info->xbdev); 1467 } 1468 1469 mutex_unlock(&info->mutex); 1470 1471 if (!xbdev) { 1472 /* sudden device removal */ 1473 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); 1474 xlvbd_release_gendisk(info); 1475 disk->private_data = NULL; 1476 kfree(info); 1477 } 1478 1479 out: 1480 bdput(bdev); 1481 mutex_unlock(&blkfront_mutex); 1482 return 0; 1483 } 1484 1485 static const struct block_device_operations xlvbd_block_fops = 1486 { 1487 .owner = THIS_MODULE, 1488 .open = blkif_open, 1489 .release = blkif_release, 1490 .getgeo = blkif_getgeo, 1491 .ioctl = blkif_ioctl, 1492 }; 1493 1494 1495 static const struct xenbus_device_id blkfront_ids[] = { 1496 { "vbd" }, 1497 { "" } 1498 }; 1499 1500 static DEFINE_XENBUS_DRIVER(blkfront, , 1501 .probe = blkfront_probe, 1502 .remove = blkfront_remove, 1503 .resume = blkfront_resume, 1504 .otherend_changed = blkback_changed, 1505 .is_ready = blkfront_is_ready, 1506 ); 1507 1508 static int __init xlblk_init(void) 1509 { 1510 int ret; 1511 1512 if (!xen_domain()) 1513 return -ENODEV; 1514 1515 if (xen_hvm_domain() && !xen_platform_pci_unplug) 1516 return -ENODEV; 1517 1518 if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) { 1519 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n", 1520 XENVBD_MAJOR, DEV_NAME); 1521 return -ENODEV; 1522 } 1523 1524 ret = xenbus_register_frontend(&blkfront_driver); 1525 if (ret) { 1526 unregister_blkdev(XENVBD_MAJOR, DEV_NAME); 1527 return ret; 1528 } 1529 1530 return 0; 1531 } 1532 module_init(xlblk_init); 1533 1534 1535 static void __exit xlblk_exit(void) 1536 { 1537 xenbus_unregister_driver(&blkfront_driver); 1538 unregister_blkdev(XENVBD_MAJOR, DEV_NAME); 1539 kfree(minors); 1540 } 1541 module_exit(xlblk_exit); 1542 1543 MODULE_DESCRIPTION("Xen virtual block device frontend"); 1544 MODULE_LICENSE("GPL"); 1545 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR); 1546 MODULE_ALIAS("xen:vbd"); 1547 MODULE_ALIAS("xenblk"); 1548