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/blk-mq.h> 41 #include <linux/hdreg.h> 42 #include <linux/cdrom.h> 43 #include <linux/module.h> 44 #include <linux/slab.h> 45 #include <linux/mutex.h> 46 #include <linux/scatterlist.h> 47 #include <linux/bitmap.h> 48 #include <linux/list.h> 49 50 #include <xen/xen.h> 51 #include <xen/xenbus.h> 52 #include <xen/grant_table.h> 53 #include <xen/events.h> 54 #include <xen/page.h> 55 #include <xen/platform_pci.h> 56 57 #include <xen/interface/grant_table.h> 58 #include <xen/interface/io/blkif.h> 59 #include <xen/interface/io/protocols.h> 60 61 #include <asm/xen/hypervisor.h> 62 63 /* 64 * The minimal size of segment supported by the block framework is PAGE_SIZE. 65 * When Linux is using a different page size than Xen, it may not be possible 66 * to put all the data in a single segment. 67 * This can happen when the backend doesn't support indirect descriptor and 68 * therefore the maximum amount of data that a request can carry is 69 * BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE = 44KB 70 * 71 * Note that we only support one extra request. So the Linux page size 72 * should be <= ( 2 * BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE) = 73 * 88KB. 74 */ 75 #define HAS_EXTRA_REQ (BLKIF_MAX_SEGMENTS_PER_REQUEST < XEN_PFN_PER_PAGE) 76 77 enum blkif_state { 78 BLKIF_STATE_DISCONNECTED, 79 BLKIF_STATE_CONNECTED, 80 BLKIF_STATE_SUSPENDED, 81 }; 82 83 struct grant { 84 grant_ref_t gref; 85 struct page *page; 86 struct list_head node; 87 }; 88 89 enum blk_req_status { 90 REQ_WAITING, 91 REQ_DONE, 92 REQ_ERROR, 93 REQ_EOPNOTSUPP, 94 }; 95 96 struct blk_shadow { 97 struct blkif_request req; 98 struct request *request; 99 struct grant **grants_used; 100 struct grant **indirect_grants; 101 struct scatterlist *sg; 102 unsigned int num_sg; 103 enum blk_req_status status; 104 105 #define NO_ASSOCIATED_ID ~0UL 106 /* 107 * Id of the sibling if we ever need 2 requests when handling a 108 * block I/O request 109 */ 110 unsigned long associated_id; 111 }; 112 113 struct blkif_req { 114 blk_status_t error; 115 }; 116 117 static inline struct blkif_req *blkif_req(struct request *rq) 118 { 119 return blk_mq_rq_to_pdu(rq); 120 } 121 122 static DEFINE_MUTEX(blkfront_mutex); 123 static const struct block_device_operations xlvbd_block_fops; 124 125 /* 126 * Maximum number of segments in indirect requests, the actual value used by 127 * the frontend driver is the minimum of this value and the value provided 128 * by the backend driver. 129 */ 130 131 static unsigned int xen_blkif_max_segments = 32; 132 module_param_named(max_indirect_segments, xen_blkif_max_segments, uint, 0444); 133 MODULE_PARM_DESC(max_indirect_segments, 134 "Maximum amount of segments in indirect requests (default is 32)"); 135 136 static unsigned int xen_blkif_max_queues = 4; 137 module_param_named(max_queues, xen_blkif_max_queues, uint, 0444); 138 MODULE_PARM_DESC(max_queues, "Maximum number of hardware queues/rings used per virtual disk"); 139 140 /* 141 * Maximum order of pages to be used for the shared ring between front and 142 * backend, 4KB page granularity is used. 143 */ 144 static unsigned int xen_blkif_max_ring_order; 145 module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, 0444); 146 MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring"); 147 148 #define BLK_RING_SIZE(info) \ 149 __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * (info)->nr_ring_pages) 150 151 #define BLK_MAX_RING_SIZE \ 152 __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * XENBUS_MAX_RING_GRANTS) 153 154 /* 155 * ring-ref%u i=(-1UL) would take 11 characters + 'ring-ref' is 8, so 19 156 * characters are enough. Define to 20 to keep consistent with backend. 157 */ 158 #define RINGREF_NAME_LEN (20) 159 /* 160 * queue-%u would take 7 + 10(UINT_MAX) = 17 characters. 161 */ 162 #define QUEUE_NAME_LEN (17) 163 164 /* 165 * Per-ring info. 166 * Every blkfront device can associate with one or more blkfront_ring_info, 167 * depending on how many hardware queues/rings to be used. 168 */ 169 struct blkfront_ring_info { 170 /* Lock to protect data in every ring buffer. */ 171 spinlock_t ring_lock; 172 struct blkif_front_ring ring; 173 unsigned int ring_ref[XENBUS_MAX_RING_GRANTS]; 174 unsigned int evtchn, irq; 175 struct work_struct work; 176 struct gnttab_free_callback callback; 177 struct blk_shadow shadow[BLK_MAX_RING_SIZE]; 178 struct list_head indirect_pages; 179 struct list_head grants; 180 unsigned int persistent_gnts_c; 181 unsigned long shadow_free; 182 struct blkfront_info *dev_info; 183 }; 184 185 /* 186 * We have one of these per vbd, whether ide, scsi or 'other'. They 187 * hang in private_data off the gendisk structure. We may end up 188 * putting all kinds of interesting stuff here :-) 189 */ 190 struct blkfront_info 191 { 192 struct mutex mutex; 193 struct xenbus_device *xbdev; 194 struct gendisk *gd; 195 u16 sector_size; 196 unsigned int physical_sector_size; 197 int vdevice; 198 blkif_vdev_t handle; 199 enum blkif_state connected; 200 /* Number of pages per ring buffer. */ 201 unsigned int nr_ring_pages; 202 struct request_queue *rq; 203 unsigned int feature_flush:1; 204 unsigned int feature_fua:1; 205 unsigned int feature_discard:1; 206 unsigned int feature_secdiscard:1; 207 unsigned int feature_persistent:1; 208 unsigned int discard_granularity; 209 unsigned int discard_alignment; 210 /* Number of 4KB segments handled */ 211 unsigned int max_indirect_segments; 212 int is_ready; 213 struct blk_mq_tag_set tag_set; 214 struct blkfront_ring_info *rinfo; 215 unsigned int nr_rings; 216 /* Save uncomplete reqs and bios for migration. */ 217 struct list_head requests; 218 struct bio_list bio_list; 219 }; 220 221 static unsigned int nr_minors; 222 static unsigned long *minors; 223 static DEFINE_SPINLOCK(minor_lock); 224 225 #define GRANT_INVALID_REF 0 226 227 #define PARTS_PER_DISK 16 228 #define PARTS_PER_EXT_DISK 256 229 230 #define BLKIF_MAJOR(dev) ((dev)>>8) 231 #define BLKIF_MINOR(dev) ((dev) & 0xff) 232 233 #define EXT_SHIFT 28 234 #define EXTENDED (1<<EXT_SHIFT) 235 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED)) 236 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED)) 237 #define EMULATED_HD_DISK_MINOR_OFFSET (0) 238 #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256) 239 #define EMULATED_SD_DISK_MINOR_OFFSET (0) 240 #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256) 241 242 #define DEV_NAME "xvd" /* name in /dev */ 243 244 /* 245 * Grants are always the same size as a Xen page (i.e 4KB). 246 * A physical segment is always the same size as a Linux page. 247 * Number of grants per physical segment 248 */ 249 #define GRANTS_PER_PSEG (PAGE_SIZE / XEN_PAGE_SIZE) 250 251 #define GRANTS_PER_INDIRECT_FRAME \ 252 (XEN_PAGE_SIZE / sizeof(struct blkif_request_segment)) 253 254 #define PSEGS_PER_INDIRECT_FRAME \ 255 (GRANTS_INDIRECT_FRAME / GRANTS_PSEGS) 256 257 #define INDIRECT_GREFS(_grants) \ 258 DIV_ROUND_UP(_grants, GRANTS_PER_INDIRECT_FRAME) 259 260 #define GREFS(_psegs) ((_psegs) * GRANTS_PER_PSEG) 261 262 static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo); 263 static void blkfront_gather_backend_features(struct blkfront_info *info); 264 static int negotiate_mq(struct blkfront_info *info); 265 266 static int get_id_from_freelist(struct blkfront_ring_info *rinfo) 267 { 268 unsigned long free = rinfo->shadow_free; 269 270 BUG_ON(free >= BLK_RING_SIZE(rinfo->dev_info)); 271 rinfo->shadow_free = rinfo->shadow[free].req.u.rw.id; 272 rinfo->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */ 273 return free; 274 } 275 276 static int add_id_to_freelist(struct blkfront_ring_info *rinfo, 277 unsigned long id) 278 { 279 if (rinfo->shadow[id].req.u.rw.id != id) 280 return -EINVAL; 281 if (rinfo->shadow[id].request == NULL) 282 return -EINVAL; 283 rinfo->shadow[id].req.u.rw.id = rinfo->shadow_free; 284 rinfo->shadow[id].request = NULL; 285 rinfo->shadow_free = id; 286 return 0; 287 } 288 289 static int fill_grant_buffer(struct blkfront_ring_info *rinfo, int num) 290 { 291 struct blkfront_info *info = rinfo->dev_info; 292 struct page *granted_page; 293 struct grant *gnt_list_entry, *n; 294 int i = 0; 295 296 while (i < num) { 297 gnt_list_entry = kzalloc(sizeof(struct grant), GFP_NOIO); 298 if (!gnt_list_entry) 299 goto out_of_memory; 300 301 if (info->feature_persistent) { 302 granted_page = alloc_page(GFP_NOIO); 303 if (!granted_page) { 304 kfree(gnt_list_entry); 305 goto out_of_memory; 306 } 307 gnt_list_entry->page = granted_page; 308 } 309 310 gnt_list_entry->gref = GRANT_INVALID_REF; 311 list_add(&gnt_list_entry->node, &rinfo->grants); 312 i++; 313 } 314 315 return 0; 316 317 out_of_memory: 318 list_for_each_entry_safe(gnt_list_entry, n, 319 &rinfo->grants, node) { 320 list_del(&gnt_list_entry->node); 321 if (info->feature_persistent) 322 __free_page(gnt_list_entry->page); 323 kfree(gnt_list_entry); 324 i--; 325 } 326 BUG_ON(i != 0); 327 return -ENOMEM; 328 } 329 330 static struct grant *get_free_grant(struct blkfront_ring_info *rinfo) 331 { 332 struct grant *gnt_list_entry; 333 334 BUG_ON(list_empty(&rinfo->grants)); 335 gnt_list_entry = list_first_entry(&rinfo->grants, struct grant, 336 node); 337 list_del(&gnt_list_entry->node); 338 339 if (gnt_list_entry->gref != GRANT_INVALID_REF) 340 rinfo->persistent_gnts_c--; 341 342 return gnt_list_entry; 343 } 344 345 static inline void grant_foreign_access(const struct grant *gnt_list_entry, 346 const struct blkfront_info *info) 347 { 348 gnttab_page_grant_foreign_access_ref_one(gnt_list_entry->gref, 349 info->xbdev->otherend_id, 350 gnt_list_entry->page, 351 0); 352 } 353 354 static struct grant *get_grant(grant_ref_t *gref_head, 355 unsigned long gfn, 356 struct blkfront_ring_info *rinfo) 357 { 358 struct grant *gnt_list_entry = get_free_grant(rinfo); 359 struct blkfront_info *info = rinfo->dev_info; 360 361 if (gnt_list_entry->gref != GRANT_INVALID_REF) 362 return gnt_list_entry; 363 364 /* Assign a gref to this page */ 365 gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); 366 BUG_ON(gnt_list_entry->gref == -ENOSPC); 367 if (info->feature_persistent) 368 grant_foreign_access(gnt_list_entry, info); 369 else { 370 /* Grant access to the GFN passed by the caller */ 371 gnttab_grant_foreign_access_ref(gnt_list_entry->gref, 372 info->xbdev->otherend_id, 373 gfn, 0); 374 } 375 376 return gnt_list_entry; 377 } 378 379 static struct grant *get_indirect_grant(grant_ref_t *gref_head, 380 struct blkfront_ring_info *rinfo) 381 { 382 struct grant *gnt_list_entry = get_free_grant(rinfo); 383 struct blkfront_info *info = rinfo->dev_info; 384 385 if (gnt_list_entry->gref != GRANT_INVALID_REF) 386 return gnt_list_entry; 387 388 /* Assign a gref to this page */ 389 gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); 390 BUG_ON(gnt_list_entry->gref == -ENOSPC); 391 if (!info->feature_persistent) { 392 struct page *indirect_page; 393 394 /* Fetch a pre-allocated page to use for indirect grefs */ 395 BUG_ON(list_empty(&rinfo->indirect_pages)); 396 indirect_page = list_first_entry(&rinfo->indirect_pages, 397 struct page, lru); 398 list_del(&indirect_page->lru); 399 gnt_list_entry->page = indirect_page; 400 } 401 grant_foreign_access(gnt_list_entry, info); 402 403 return gnt_list_entry; 404 } 405 406 static const char *op_name(int op) 407 { 408 static const char *const names[] = { 409 [BLKIF_OP_READ] = "read", 410 [BLKIF_OP_WRITE] = "write", 411 [BLKIF_OP_WRITE_BARRIER] = "barrier", 412 [BLKIF_OP_FLUSH_DISKCACHE] = "flush", 413 [BLKIF_OP_DISCARD] = "discard" }; 414 415 if (op < 0 || op >= ARRAY_SIZE(names)) 416 return "unknown"; 417 418 if (!names[op]) 419 return "reserved"; 420 421 return names[op]; 422 } 423 static int xlbd_reserve_minors(unsigned int minor, unsigned int nr) 424 { 425 unsigned int end = minor + nr; 426 int rc; 427 428 if (end > nr_minors) { 429 unsigned long *bitmap, *old; 430 431 bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap), 432 GFP_KERNEL); 433 if (bitmap == NULL) 434 return -ENOMEM; 435 436 spin_lock(&minor_lock); 437 if (end > nr_minors) { 438 old = minors; 439 memcpy(bitmap, minors, 440 BITS_TO_LONGS(nr_minors) * sizeof(*bitmap)); 441 minors = bitmap; 442 nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG; 443 } else 444 old = bitmap; 445 spin_unlock(&minor_lock); 446 kfree(old); 447 } 448 449 spin_lock(&minor_lock); 450 if (find_next_bit(minors, end, minor) >= end) { 451 bitmap_set(minors, minor, nr); 452 rc = 0; 453 } else 454 rc = -EBUSY; 455 spin_unlock(&minor_lock); 456 457 return rc; 458 } 459 460 static void xlbd_release_minors(unsigned int minor, unsigned int nr) 461 { 462 unsigned int end = minor + nr; 463 464 BUG_ON(end > nr_minors); 465 spin_lock(&minor_lock); 466 bitmap_clear(minors, minor, nr); 467 spin_unlock(&minor_lock); 468 } 469 470 static void blkif_restart_queue_callback(void *arg) 471 { 472 struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)arg; 473 schedule_work(&rinfo->work); 474 } 475 476 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg) 477 { 478 /* We don't have real geometry info, but let's at least return 479 values consistent with the size of the device */ 480 sector_t nsect = get_capacity(bd->bd_disk); 481 sector_t cylinders = nsect; 482 483 hg->heads = 0xff; 484 hg->sectors = 0x3f; 485 sector_div(cylinders, hg->heads * hg->sectors); 486 hg->cylinders = cylinders; 487 if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect) 488 hg->cylinders = 0xffff; 489 return 0; 490 } 491 492 static int blkif_ioctl(struct block_device *bdev, fmode_t mode, 493 unsigned command, unsigned long argument) 494 { 495 struct blkfront_info *info = bdev->bd_disk->private_data; 496 int i; 497 498 dev_dbg(&info->xbdev->dev, "command: 0x%x, argument: 0x%lx\n", 499 command, (long)argument); 500 501 switch (command) { 502 case CDROMMULTISESSION: 503 dev_dbg(&info->xbdev->dev, "FIXME: support multisession CDs later\n"); 504 for (i = 0; i < sizeof(struct cdrom_multisession); i++) 505 if (put_user(0, (char __user *)(argument + i))) 506 return -EFAULT; 507 return 0; 508 509 case CDROM_GET_CAPABILITY: { 510 struct gendisk *gd = info->gd; 511 if (gd->flags & GENHD_FL_CD) 512 return 0; 513 return -EINVAL; 514 } 515 516 default: 517 /*printk(KERN_ALERT "ioctl %08x not supported by Xen blkdev\n", 518 command);*/ 519 return -EINVAL; /* same return as native Linux */ 520 } 521 522 return 0; 523 } 524 525 static unsigned long blkif_ring_get_request(struct blkfront_ring_info *rinfo, 526 struct request *req, 527 struct blkif_request **ring_req) 528 { 529 unsigned long id; 530 531 *ring_req = RING_GET_REQUEST(&rinfo->ring, rinfo->ring.req_prod_pvt); 532 rinfo->ring.req_prod_pvt++; 533 534 id = get_id_from_freelist(rinfo); 535 rinfo->shadow[id].request = req; 536 rinfo->shadow[id].status = REQ_WAITING; 537 rinfo->shadow[id].associated_id = NO_ASSOCIATED_ID; 538 539 (*ring_req)->u.rw.id = id; 540 541 return id; 542 } 543 544 static int blkif_queue_discard_req(struct request *req, struct blkfront_ring_info *rinfo) 545 { 546 struct blkfront_info *info = rinfo->dev_info; 547 struct blkif_request *ring_req; 548 unsigned long id; 549 550 /* Fill out a communications ring structure. */ 551 id = blkif_ring_get_request(rinfo, req, &ring_req); 552 553 ring_req->operation = BLKIF_OP_DISCARD; 554 ring_req->u.discard.nr_sectors = blk_rq_sectors(req); 555 ring_req->u.discard.id = id; 556 ring_req->u.discard.sector_number = (blkif_sector_t)blk_rq_pos(req); 557 if (req_op(req) == REQ_OP_SECURE_ERASE && info->feature_secdiscard) 558 ring_req->u.discard.flag = BLKIF_DISCARD_SECURE; 559 else 560 ring_req->u.discard.flag = 0; 561 562 /* Keep a private copy so we can reissue requests when recovering. */ 563 rinfo->shadow[id].req = *ring_req; 564 565 return 0; 566 } 567 568 struct setup_rw_req { 569 unsigned int grant_idx; 570 struct blkif_request_segment *segments; 571 struct blkfront_ring_info *rinfo; 572 struct blkif_request *ring_req; 573 grant_ref_t gref_head; 574 unsigned int id; 575 /* Only used when persistent grant is used and it's a read request */ 576 bool need_copy; 577 unsigned int bvec_off; 578 char *bvec_data; 579 580 bool require_extra_req; 581 struct blkif_request *extra_ring_req; 582 }; 583 584 static void blkif_setup_rw_req_grant(unsigned long gfn, unsigned int offset, 585 unsigned int len, void *data) 586 { 587 struct setup_rw_req *setup = data; 588 int n, ref; 589 struct grant *gnt_list_entry; 590 unsigned int fsect, lsect; 591 /* Convenient aliases */ 592 unsigned int grant_idx = setup->grant_idx; 593 struct blkif_request *ring_req = setup->ring_req; 594 struct blkfront_ring_info *rinfo = setup->rinfo; 595 /* 596 * We always use the shadow of the first request to store the list 597 * of grant associated to the block I/O request. This made the 598 * completion more easy to handle even if the block I/O request is 599 * split. 600 */ 601 struct blk_shadow *shadow = &rinfo->shadow[setup->id]; 602 603 if (unlikely(setup->require_extra_req && 604 grant_idx >= BLKIF_MAX_SEGMENTS_PER_REQUEST)) { 605 /* 606 * We are using the second request, setup grant_idx 607 * to be the index of the segment array. 608 */ 609 grant_idx -= BLKIF_MAX_SEGMENTS_PER_REQUEST; 610 ring_req = setup->extra_ring_req; 611 } 612 613 if ((ring_req->operation == BLKIF_OP_INDIRECT) && 614 (grant_idx % GRANTS_PER_INDIRECT_FRAME == 0)) { 615 if (setup->segments) 616 kunmap_atomic(setup->segments); 617 618 n = grant_idx / GRANTS_PER_INDIRECT_FRAME; 619 gnt_list_entry = get_indirect_grant(&setup->gref_head, rinfo); 620 shadow->indirect_grants[n] = gnt_list_entry; 621 setup->segments = kmap_atomic(gnt_list_entry->page); 622 ring_req->u.indirect.indirect_grefs[n] = gnt_list_entry->gref; 623 } 624 625 gnt_list_entry = get_grant(&setup->gref_head, gfn, rinfo); 626 ref = gnt_list_entry->gref; 627 /* 628 * All the grants are stored in the shadow of the first 629 * request. Therefore we have to use the global index. 630 */ 631 shadow->grants_used[setup->grant_idx] = gnt_list_entry; 632 633 if (setup->need_copy) { 634 void *shared_data; 635 636 shared_data = kmap_atomic(gnt_list_entry->page); 637 /* 638 * this does not wipe data stored outside the 639 * range sg->offset..sg->offset+sg->length. 640 * Therefore, blkback *could* see data from 641 * previous requests. This is OK as long as 642 * persistent grants are shared with just one 643 * domain. It may need refactoring if this 644 * changes 645 */ 646 memcpy(shared_data + offset, 647 setup->bvec_data + setup->bvec_off, 648 len); 649 650 kunmap_atomic(shared_data); 651 setup->bvec_off += len; 652 } 653 654 fsect = offset >> 9; 655 lsect = fsect + (len >> 9) - 1; 656 if (ring_req->operation != BLKIF_OP_INDIRECT) { 657 ring_req->u.rw.seg[grant_idx] = 658 (struct blkif_request_segment) { 659 .gref = ref, 660 .first_sect = fsect, 661 .last_sect = lsect }; 662 } else { 663 setup->segments[grant_idx % GRANTS_PER_INDIRECT_FRAME] = 664 (struct blkif_request_segment) { 665 .gref = ref, 666 .first_sect = fsect, 667 .last_sect = lsect }; 668 } 669 670 (setup->grant_idx)++; 671 } 672 673 static void blkif_setup_extra_req(struct blkif_request *first, 674 struct blkif_request *second) 675 { 676 uint16_t nr_segments = first->u.rw.nr_segments; 677 678 /* 679 * The second request is only present when the first request uses 680 * all its segments. It's always the continuity of the first one. 681 */ 682 first->u.rw.nr_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; 683 684 second->u.rw.nr_segments = nr_segments - BLKIF_MAX_SEGMENTS_PER_REQUEST; 685 second->u.rw.sector_number = first->u.rw.sector_number + 686 (BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE) / 512; 687 688 second->u.rw.handle = first->u.rw.handle; 689 second->operation = first->operation; 690 } 691 692 static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *rinfo) 693 { 694 struct blkfront_info *info = rinfo->dev_info; 695 struct blkif_request *ring_req, *extra_ring_req = NULL; 696 unsigned long id, extra_id = NO_ASSOCIATED_ID; 697 bool require_extra_req = false; 698 int i; 699 struct setup_rw_req setup = { 700 .grant_idx = 0, 701 .segments = NULL, 702 .rinfo = rinfo, 703 .need_copy = rq_data_dir(req) && info->feature_persistent, 704 }; 705 706 /* 707 * Used to store if we are able to queue the request by just using 708 * existing persistent grants, or if we have to get new grants, 709 * as there are not sufficiently many free. 710 */ 711 bool new_persistent_gnts = false; 712 struct scatterlist *sg; 713 int num_sg, max_grefs, num_grant; 714 715 max_grefs = req->nr_phys_segments * GRANTS_PER_PSEG; 716 if (max_grefs > BLKIF_MAX_SEGMENTS_PER_REQUEST) 717 /* 718 * If we are using indirect segments we need to account 719 * for the indirect grefs used in the request. 720 */ 721 max_grefs += INDIRECT_GREFS(max_grefs); 722 723 /* Check if we have enough persistent grants to allocate a requests */ 724 if (rinfo->persistent_gnts_c < max_grefs) { 725 new_persistent_gnts = true; 726 727 if (gnttab_alloc_grant_references( 728 max_grefs - rinfo->persistent_gnts_c, 729 &setup.gref_head) < 0) { 730 gnttab_request_free_callback( 731 &rinfo->callback, 732 blkif_restart_queue_callback, 733 rinfo, 734 max_grefs - rinfo->persistent_gnts_c); 735 return 1; 736 } 737 } 738 739 /* Fill out a communications ring structure. */ 740 id = blkif_ring_get_request(rinfo, req, &ring_req); 741 742 num_sg = blk_rq_map_sg(req->q, req, rinfo->shadow[id].sg); 743 num_grant = 0; 744 /* Calculate the number of grant used */ 745 for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) 746 num_grant += gnttab_count_grant(sg->offset, sg->length); 747 748 require_extra_req = info->max_indirect_segments == 0 && 749 num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST; 750 BUG_ON(!HAS_EXTRA_REQ && require_extra_req); 751 752 rinfo->shadow[id].num_sg = num_sg; 753 if (num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST && 754 likely(!require_extra_req)) { 755 /* 756 * The indirect operation can only be a BLKIF_OP_READ or 757 * BLKIF_OP_WRITE 758 */ 759 BUG_ON(req_op(req) == REQ_OP_FLUSH || req->cmd_flags & REQ_FUA); 760 ring_req->operation = BLKIF_OP_INDIRECT; 761 ring_req->u.indirect.indirect_op = rq_data_dir(req) ? 762 BLKIF_OP_WRITE : BLKIF_OP_READ; 763 ring_req->u.indirect.sector_number = (blkif_sector_t)blk_rq_pos(req); 764 ring_req->u.indirect.handle = info->handle; 765 ring_req->u.indirect.nr_segments = num_grant; 766 } else { 767 ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req); 768 ring_req->u.rw.handle = info->handle; 769 ring_req->operation = rq_data_dir(req) ? 770 BLKIF_OP_WRITE : BLKIF_OP_READ; 771 if (req_op(req) == REQ_OP_FLUSH || req->cmd_flags & REQ_FUA) { 772 /* 773 * Ideally we can do an unordered flush-to-disk. 774 * In case the backend onlysupports barriers, use that. 775 * A barrier request a superset of FUA, so we can 776 * implement it the same way. (It's also a FLUSH+FUA, 777 * since it is guaranteed ordered WRT previous writes.) 778 */ 779 if (info->feature_flush && info->feature_fua) 780 ring_req->operation = 781 BLKIF_OP_WRITE_BARRIER; 782 else if (info->feature_flush) 783 ring_req->operation = 784 BLKIF_OP_FLUSH_DISKCACHE; 785 else 786 ring_req->operation = 0; 787 } 788 ring_req->u.rw.nr_segments = num_grant; 789 if (unlikely(require_extra_req)) { 790 extra_id = blkif_ring_get_request(rinfo, req, 791 &extra_ring_req); 792 /* 793 * Only the first request contains the scatter-gather 794 * list. 795 */ 796 rinfo->shadow[extra_id].num_sg = 0; 797 798 blkif_setup_extra_req(ring_req, extra_ring_req); 799 800 /* Link the 2 requests together */ 801 rinfo->shadow[extra_id].associated_id = id; 802 rinfo->shadow[id].associated_id = extra_id; 803 } 804 } 805 806 setup.ring_req = ring_req; 807 setup.id = id; 808 809 setup.require_extra_req = require_extra_req; 810 if (unlikely(require_extra_req)) 811 setup.extra_ring_req = extra_ring_req; 812 813 for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) { 814 BUG_ON(sg->offset + sg->length > PAGE_SIZE); 815 816 if (setup.need_copy) { 817 setup.bvec_off = sg->offset; 818 setup.bvec_data = kmap_atomic(sg_page(sg)); 819 } 820 821 gnttab_foreach_grant_in_range(sg_page(sg), 822 sg->offset, 823 sg->length, 824 blkif_setup_rw_req_grant, 825 &setup); 826 827 if (setup.need_copy) 828 kunmap_atomic(setup.bvec_data); 829 } 830 if (setup.segments) 831 kunmap_atomic(setup.segments); 832 833 /* Keep a private copy so we can reissue requests when recovering. */ 834 rinfo->shadow[id].req = *ring_req; 835 if (unlikely(require_extra_req)) 836 rinfo->shadow[extra_id].req = *extra_ring_req; 837 838 if (new_persistent_gnts) 839 gnttab_free_grant_references(setup.gref_head); 840 841 return 0; 842 } 843 844 /* 845 * Generate a Xen blkfront IO request from a blk layer request. Reads 846 * and writes are handled as expected. 847 * 848 * @req: a request struct 849 */ 850 static int blkif_queue_request(struct request *req, struct blkfront_ring_info *rinfo) 851 { 852 if (unlikely(rinfo->dev_info->connected != BLKIF_STATE_CONNECTED)) 853 return 1; 854 855 if (unlikely(req_op(req) == REQ_OP_DISCARD || 856 req_op(req) == REQ_OP_SECURE_ERASE)) 857 return blkif_queue_discard_req(req, rinfo); 858 else 859 return blkif_queue_rw_req(req, rinfo); 860 } 861 862 static inline void flush_requests(struct blkfront_ring_info *rinfo) 863 { 864 int notify; 865 866 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&rinfo->ring, notify); 867 868 if (notify) 869 notify_remote_via_irq(rinfo->irq); 870 } 871 872 static inline bool blkif_request_flush_invalid(struct request *req, 873 struct blkfront_info *info) 874 { 875 return (blk_rq_is_passthrough(req) || 876 ((req_op(req) == REQ_OP_FLUSH) && 877 !info->feature_flush) || 878 ((req->cmd_flags & REQ_FUA) && 879 !info->feature_fua)); 880 } 881 882 static blk_status_t blkif_queue_rq(struct blk_mq_hw_ctx *hctx, 883 const struct blk_mq_queue_data *qd) 884 { 885 unsigned long flags; 886 int qid = hctx->queue_num; 887 struct blkfront_info *info = hctx->queue->queuedata; 888 struct blkfront_ring_info *rinfo = NULL; 889 890 BUG_ON(info->nr_rings <= qid); 891 rinfo = &info->rinfo[qid]; 892 blk_mq_start_request(qd->rq); 893 spin_lock_irqsave(&rinfo->ring_lock, flags); 894 if (RING_FULL(&rinfo->ring)) 895 goto out_busy; 896 897 if (blkif_request_flush_invalid(qd->rq, rinfo->dev_info)) 898 goto out_err; 899 900 if (blkif_queue_request(qd->rq, rinfo)) 901 goto out_busy; 902 903 flush_requests(rinfo); 904 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 905 return BLK_STS_OK; 906 907 out_err: 908 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 909 return BLK_STS_IOERR; 910 911 out_busy: 912 blk_mq_stop_hw_queue(hctx); 913 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 914 return BLK_STS_DEV_RESOURCE; 915 } 916 917 static void blkif_complete_rq(struct request *rq) 918 { 919 blk_mq_end_request(rq, blkif_req(rq)->error); 920 } 921 922 static const struct blk_mq_ops blkfront_mq_ops = { 923 .queue_rq = blkif_queue_rq, 924 .complete = blkif_complete_rq, 925 }; 926 927 static void blkif_set_queue_limits(struct blkfront_info *info) 928 { 929 struct request_queue *rq = info->rq; 930 struct gendisk *gd = info->gd; 931 unsigned int segments = info->max_indirect_segments ? : 932 BLKIF_MAX_SEGMENTS_PER_REQUEST; 933 934 blk_queue_flag_set(QUEUE_FLAG_VIRT, rq); 935 936 if (info->feature_discard) { 937 blk_queue_flag_set(QUEUE_FLAG_DISCARD, rq); 938 blk_queue_max_discard_sectors(rq, get_capacity(gd)); 939 rq->limits.discard_granularity = info->discard_granularity; 940 rq->limits.discard_alignment = info->discard_alignment; 941 if (info->feature_secdiscard) 942 blk_queue_flag_set(QUEUE_FLAG_SECERASE, rq); 943 } 944 945 /* Hard sector size and max sectors impersonate the equiv. hardware. */ 946 blk_queue_logical_block_size(rq, info->sector_size); 947 blk_queue_physical_block_size(rq, info->physical_sector_size); 948 blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512); 949 950 /* Each segment in a request is up to an aligned page in size. */ 951 blk_queue_segment_boundary(rq, PAGE_SIZE - 1); 952 blk_queue_max_segment_size(rq, PAGE_SIZE); 953 954 /* Ensure a merged request will fit in a single I/O ring slot. */ 955 blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG); 956 957 /* Make sure buffer addresses are sector-aligned. */ 958 blk_queue_dma_alignment(rq, 511); 959 } 960 961 static int xlvbd_init_blk_queue(struct gendisk *gd, u16 sector_size, 962 unsigned int physical_sector_size) 963 { 964 struct request_queue *rq; 965 struct blkfront_info *info = gd->private_data; 966 967 memset(&info->tag_set, 0, sizeof(info->tag_set)); 968 info->tag_set.ops = &blkfront_mq_ops; 969 info->tag_set.nr_hw_queues = info->nr_rings; 970 if (HAS_EXTRA_REQ && info->max_indirect_segments == 0) { 971 /* 972 * When indirect descriptior is not supported, the I/O request 973 * will be split between multiple request in the ring. 974 * To avoid problems when sending the request, divide by 975 * 2 the depth of the queue. 976 */ 977 info->tag_set.queue_depth = BLK_RING_SIZE(info) / 2; 978 } else 979 info->tag_set.queue_depth = BLK_RING_SIZE(info); 980 info->tag_set.numa_node = NUMA_NO_NODE; 981 info->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE; 982 info->tag_set.cmd_size = sizeof(struct blkif_req); 983 info->tag_set.driver_data = info; 984 985 if (blk_mq_alloc_tag_set(&info->tag_set)) 986 return -EINVAL; 987 rq = blk_mq_init_queue(&info->tag_set); 988 if (IS_ERR(rq)) { 989 blk_mq_free_tag_set(&info->tag_set); 990 return PTR_ERR(rq); 991 } 992 993 rq->queuedata = info; 994 info->rq = gd->queue = rq; 995 info->gd = gd; 996 info->sector_size = sector_size; 997 info->physical_sector_size = physical_sector_size; 998 blkif_set_queue_limits(info); 999 1000 return 0; 1001 } 1002 1003 static const char *flush_info(struct blkfront_info *info) 1004 { 1005 if (info->feature_flush && info->feature_fua) 1006 return "barrier: enabled;"; 1007 else if (info->feature_flush) 1008 return "flush diskcache: enabled;"; 1009 else 1010 return "barrier or flush: disabled;"; 1011 } 1012 1013 static void xlvbd_flush(struct blkfront_info *info) 1014 { 1015 blk_queue_write_cache(info->rq, info->feature_flush ? true : false, 1016 info->feature_fua ? true : false); 1017 pr_info("blkfront: %s: %s %s %s %s %s\n", 1018 info->gd->disk_name, flush_info(info), 1019 "persistent grants:", info->feature_persistent ? 1020 "enabled;" : "disabled;", "indirect descriptors:", 1021 info->max_indirect_segments ? "enabled;" : "disabled;"); 1022 } 1023 1024 static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset) 1025 { 1026 int major; 1027 major = BLKIF_MAJOR(vdevice); 1028 *minor = BLKIF_MINOR(vdevice); 1029 switch (major) { 1030 case XEN_IDE0_MAJOR: 1031 *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET; 1032 *minor = ((*minor / 64) * PARTS_PER_DISK) + 1033 EMULATED_HD_DISK_MINOR_OFFSET; 1034 break; 1035 case XEN_IDE1_MAJOR: 1036 *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET; 1037 *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) + 1038 EMULATED_HD_DISK_MINOR_OFFSET; 1039 break; 1040 case XEN_SCSI_DISK0_MAJOR: 1041 *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET; 1042 *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET; 1043 break; 1044 case XEN_SCSI_DISK1_MAJOR: 1045 case XEN_SCSI_DISK2_MAJOR: 1046 case XEN_SCSI_DISK3_MAJOR: 1047 case XEN_SCSI_DISK4_MAJOR: 1048 case XEN_SCSI_DISK5_MAJOR: 1049 case XEN_SCSI_DISK6_MAJOR: 1050 case XEN_SCSI_DISK7_MAJOR: 1051 *offset = (*minor / PARTS_PER_DISK) + 1052 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) + 1053 EMULATED_SD_DISK_NAME_OFFSET; 1054 *minor = *minor + 1055 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) + 1056 EMULATED_SD_DISK_MINOR_OFFSET; 1057 break; 1058 case XEN_SCSI_DISK8_MAJOR: 1059 case XEN_SCSI_DISK9_MAJOR: 1060 case XEN_SCSI_DISK10_MAJOR: 1061 case XEN_SCSI_DISK11_MAJOR: 1062 case XEN_SCSI_DISK12_MAJOR: 1063 case XEN_SCSI_DISK13_MAJOR: 1064 case XEN_SCSI_DISK14_MAJOR: 1065 case XEN_SCSI_DISK15_MAJOR: 1066 *offset = (*minor / PARTS_PER_DISK) + 1067 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) + 1068 EMULATED_SD_DISK_NAME_OFFSET; 1069 *minor = *minor + 1070 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) + 1071 EMULATED_SD_DISK_MINOR_OFFSET; 1072 break; 1073 case XENVBD_MAJOR: 1074 *offset = *minor / PARTS_PER_DISK; 1075 break; 1076 default: 1077 printk(KERN_WARNING "blkfront: your disk configuration is " 1078 "incorrect, please use an xvd device instead\n"); 1079 return -ENODEV; 1080 } 1081 return 0; 1082 } 1083 1084 static char *encode_disk_name(char *ptr, unsigned int n) 1085 { 1086 if (n >= 26) 1087 ptr = encode_disk_name(ptr, n / 26 - 1); 1088 *ptr = 'a' + n % 26; 1089 return ptr + 1; 1090 } 1091 1092 static int xlvbd_alloc_gendisk(blkif_sector_t capacity, 1093 struct blkfront_info *info, 1094 u16 vdisk_info, u16 sector_size, 1095 unsigned int physical_sector_size) 1096 { 1097 struct gendisk *gd; 1098 int nr_minors = 1; 1099 int err; 1100 unsigned int offset; 1101 int minor; 1102 int nr_parts; 1103 char *ptr; 1104 1105 BUG_ON(info->gd != NULL); 1106 BUG_ON(info->rq != NULL); 1107 1108 if ((info->vdevice>>EXT_SHIFT) > 1) { 1109 /* this is above the extended range; something is wrong */ 1110 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice); 1111 return -ENODEV; 1112 } 1113 1114 if (!VDEV_IS_EXTENDED(info->vdevice)) { 1115 err = xen_translate_vdev(info->vdevice, &minor, &offset); 1116 if (err) 1117 return err; 1118 nr_parts = PARTS_PER_DISK; 1119 } else { 1120 minor = BLKIF_MINOR_EXT(info->vdevice); 1121 nr_parts = PARTS_PER_EXT_DISK; 1122 offset = minor / nr_parts; 1123 if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4) 1124 printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with " 1125 "emulated IDE disks,\n\t choose an xvd device name" 1126 "from xvde on\n", info->vdevice); 1127 } 1128 if (minor >> MINORBITS) { 1129 pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n", 1130 info->vdevice, minor); 1131 return -ENODEV; 1132 } 1133 1134 if ((minor % nr_parts) == 0) 1135 nr_minors = nr_parts; 1136 1137 err = xlbd_reserve_minors(minor, nr_minors); 1138 if (err) 1139 goto out; 1140 err = -ENODEV; 1141 1142 gd = alloc_disk(nr_minors); 1143 if (gd == NULL) 1144 goto release; 1145 1146 strcpy(gd->disk_name, DEV_NAME); 1147 ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset); 1148 BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN); 1149 if (nr_minors > 1) 1150 *ptr = 0; 1151 else 1152 snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr, 1153 "%d", minor & (nr_parts - 1)); 1154 1155 gd->major = XENVBD_MAJOR; 1156 gd->first_minor = minor; 1157 gd->fops = &xlvbd_block_fops; 1158 gd->private_data = info; 1159 set_capacity(gd, capacity); 1160 1161 if (xlvbd_init_blk_queue(gd, sector_size, physical_sector_size)) { 1162 del_gendisk(gd); 1163 goto release; 1164 } 1165 1166 xlvbd_flush(info); 1167 1168 if (vdisk_info & VDISK_READONLY) 1169 set_disk_ro(gd, 1); 1170 1171 if (vdisk_info & VDISK_REMOVABLE) 1172 gd->flags |= GENHD_FL_REMOVABLE; 1173 1174 if (vdisk_info & VDISK_CDROM) 1175 gd->flags |= GENHD_FL_CD; 1176 1177 return 0; 1178 1179 release: 1180 xlbd_release_minors(minor, nr_minors); 1181 out: 1182 return err; 1183 } 1184 1185 static void xlvbd_release_gendisk(struct blkfront_info *info) 1186 { 1187 unsigned int minor, nr_minors, i; 1188 1189 if (info->rq == NULL) 1190 return; 1191 1192 /* No more blkif_request(). */ 1193 blk_mq_stop_hw_queues(info->rq); 1194 1195 for (i = 0; i < info->nr_rings; i++) { 1196 struct blkfront_ring_info *rinfo = &info->rinfo[i]; 1197 1198 /* No more gnttab callback work. */ 1199 gnttab_cancel_free_callback(&rinfo->callback); 1200 1201 /* Flush gnttab callback work. Must be done with no locks held. */ 1202 flush_work(&rinfo->work); 1203 } 1204 1205 del_gendisk(info->gd); 1206 1207 minor = info->gd->first_minor; 1208 nr_minors = info->gd->minors; 1209 xlbd_release_minors(minor, nr_minors); 1210 1211 blk_cleanup_queue(info->rq); 1212 blk_mq_free_tag_set(&info->tag_set); 1213 info->rq = NULL; 1214 1215 put_disk(info->gd); 1216 info->gd = NULL; 1217 } 1218 1219 /* Already hold rinfo->ring_lock. */ 1220 static inline void kick_pending_request_queues_locked(struct blkfront_ring_info *rinfo) 1221 { 1222 if (!RING_FULL(&rinfo->ring)) 1223 blk_mq_start_stopped_hw_queues(rinfo->dev_info->rq, true); 1224 } 1225 1226 static void kick_pending_request_queues(struct blkfront_ring_info *rinfo) 1227 { 1228 unsigned long flags; 1229 1230 spin_lock_irqsave(&rinfo->ring_lock, flags); 1231 kick_pending_request_queues_locked(rinfo); 1232 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 1233 } 1234 1235 static void blkif_restart_queue(struct work_struct *work) 1236 { 1237 struct blkfront_ring_info *rinfo = container_of(work, struct blkfront_ring_info, work); 1238 1239 if (rinfo->dev_info->connected == BLKIF_STATE_CONNECTED) 1240 kick_pending_request_queues(rinfo); 1241 } 1242 1243 static void blkif_free_ring(struct blkfront_ring_info *rinfo) 1244 { 1245 struct grant *persistent_gnt, *n; 1246 struct blkfront_info *info = rinfo->dev_info; 1247 int i, j, segs; 1248 1249 /* 1250 * Remove indirect pages, this only happens when using indirect 1251 * descriptors but not persistent grants 1252 */ 1253 if (!list_empty(&rinfo->indirect_pages)) { 1254 struct page *indirect_page, *n; 1255 1256 BUG_ON(info->feature_persistent); 1257 list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { 1258 list_del(&indirect_page->lru); 1259 __free_page(indirect_page); 1260 } 1261 } 1262 1263 /* Remove all persistent grants. */ 1264 if (!list_empty(&rinfo->grants)) { 1265 list_for_each_entry_safe(persistent_gnt, n, 1266 &rinfo->grants, node) { 1267 list_del(&persistent_gnt->node); 1268 if (persistent_gnt->gref != GRANT_INVALID_REF) { 1269 gnttab_end_foreign_access(persistent_gnt->gref, 1270 0, 0UL); 1271 rinfo->persistent_gnts_c--; 1272 } 1273 if (info->feature_persistent) 1274 __free_page(persistent_gnt->page); 1275 kfree(persistent_gnt); 1276 } 1277 } 1278 BUG_ON(rinfo->persistent_gnts_c != 0); 1279 1280 for (i = 0; i < BLK_RING_SIZE(info); i++) { 1281 /* 1282 * Clear persistent grants present in requests already 1283 * on the shared ring 1284 */ 1285 if (!rinfo->shadow[i].request) 1286 goto free_shadow; 1287 1288 segs = rinfo->shadow[i].req.operation == BLKIF_OP_INDIRECT ? 1289 rinfo->shadow[i].req.u.indirect.nr_segments : 1290 rinfo->shadow[i].req.u.rw.nr_segments; 1291 for (j = 0; j < segs; j++) { 1292 persistent_gnt = rinfo->shadow[i].grants_used[j]; 1293 gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL); 1294 if (info->feature_persistent) 1295 __free_page(persistent_gnt->page); 1296 kfree(persistent_gnt); 1297 } 1298 1299 if (rinfo->shadow[i].req.operation != BLKIF_OP_INDIRECT) 1300 /* 1301 * If this is not an indirect operation don't try to 1302 * free indirect segments 1303 */ 1304 goto free_shadow; 1305 1306 for (j = 0; j < INDIRECT_GREFS(segs); j++) { 1307 persistent_gnt = rinfo->shadow[i].indirect_grants[j]; 1308 gnttab_end_foreign_access(persistent_gnt->gref, 0, 0UL); 1309 __free_page(persistent_gnt->page); 1310 kfree(persistent_gnt); 1311 } 1312 1313 free_shadow: 1314 kfree(rinfo->shadow[i].grants_used); 1315 rinfo->shadow[i].grants_used = NULL; 1316 kfree(rinfo->shadow[i].indirect_grants); 1317 rinfo->shadow[i].indirect_grants = NULL; 1318 kfree(rinfo->shadow[i].sg); 1319 rinfo->shadow[i].sg = NULL; 1320 } 1321 1322 /* No more gnttab callback work. */ 1323 gnttab_cancel_free_callback(&rinfo->callback); 1324 1325 /* Flush gnttab callback work. Must be done with no locks held. */ 1326 flush_work(&rinfo->work); 1327 1328 /* Free resources associated with old device channel. */ 1329 for (i = 0; i < info->nr_ring_pages; i++) { 1330 if (rinfo->ring_ref[i] != GRANT_INVALID_REF) { 1331 gnttab_end_foreign_access(rinfo->ring_ref[i], 0, 0); 1332 rinfo->ring_ref[i] = GRANT_INVALID_REF; 1333 } 1334 } 1335 free_pages((unsigned long)rinfo->ring.sring, get_order(info->nr_ring_pages * XEN_PAGE_SIZE)); 1336 rinfo->ring.sring = NULL; 1337 1338 if (rinfo->irq) 1339 unbind_from_irqhandler(rinfo->irq, rinfo); 1340 rinfo->evtchn = rinfo->irq = 0; 1341 } 1342 1343 static void blkif_free(struct blkfront_info *info, int suspend) 1344 { 1345 unsigned int i; 1346 1347 /* Prevent new requests being issued until we fix things up. */ 1348 info->connected = suspend ? 1349 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED; 1350 /* No more blkif_request(). */ 1351 if (info->rq) 1352 blk_mq_stop_hw_queues(info->rq); 1353 1354 for (i = 0; i < info->nr_rings; i++) 1355 blkif_free_ring(&info->rinfo[i]); 1356 1357 kfree(info->rinfo); 1358 info->rinfo = NULL; 1359 info->nr_rings = 0; 1360 } 1361 1362 struct copy_from_grant { 1363 const struct blk_shadow *s; 1364 unsigned int grant_idx; 1365 unsigned int bvec_offset; 1366 char *bvec_data; 1367 }; 1368 1369 static void blkif_copy_from_grant(unsigned long gfn, unsigned int offset, 1370 unsigned int len, void *data) 1371 { 1372 struct copy_from_grant *info = data; 1373 char *shared_data; 1374 /* Convenient aliases */ 1375 const struct blk_shadow *s = info->s; 1376 1377 shared_data = kmap_atomic(s->grants_used[info->grant_idx]->page); 1378 1379 memcpy(info->bvec_data + info->bvec_offset, 1380 shared_data + offset, len); 1381 1382 info->bvec_offset += len; 1383 info->grant_idx++; 1384 1385 kunmap_atomic(shared_data); 1386 } 1387 1388 static enum blk_req_status blkif_rsp_to_req_status(int rsp) 1389 { 1390 switch (rsp) 1391 { 1392 case BLKIF_RSP_OKAY: 1393 return REQ_DONE; 1394 case BLKIF_RSP_EOPNOTSUPP: 1395 return REQ_EOPNOTSUPP; 1396 case BLKIF_RSP_ERROR: 1397 /* Fallthrough. */ 1398 default: 1399 return REQ_ERROR; 1400 } 1401 } 1402 1403 /* 1404 * Get the final status of the block request based on two ring response 1405 */ 1406 static int blkif_get_final_status(enum blk_req_status s1, 1407 enum blk_req_status s2) 1408 { 1409 BUG_ON(s1 == REQ_WAITING); 1410 BUG_ON(s2 == REQ_WAITING); 1411 1412 if (s1 == REQ_ERROR || s2 == REQ_ERROR) 1413 return BLKIF_RSP_ERROR; 1414 else if (s1 == REQ_EOPNOTSUPP || s2 == REQ_EOPNOTSUPP) 1415 return BLKIF_RSP_EOPNOTSUPP; 1416 return BLKIF_RSP_OKAY; 1417 } 1418 1419 static bool blkif_completion(unsigned long *id, 1420 struct blkfront_ring_info *rinfo, 1421 struct blkif_response *bret) 1422 { 1423 int i = 0; 1424 struct scatterlist *sg; 1425 int num_sg, num_grant; 1426 struct blkfront_info *info = rinfo->dev_info; 1427 struct blk_shadow *s = &rinfo->shadow[*id]; 1428 struct copy_from_grant data = { 1429 .grant_idx = 0, 1430 }; 1431 1432 num_grant = s->req.operation == BLKIF_OP_INDIRECT ? 1433 s->req.u.indirect.nr_segments : s->req.u.rw.nr_segments; 1434 1435 /* The I/O request may be split in two. */ 1436 if (unlikely(s->associated_id != NO_ASSOCIATED_ID)) { 1437 struct blk_shadow *s2 = &rinfo->shadow[s->associated_id]; 1438 1439 /* Keep the status of the current response in shadow. */ 1440 s->status = blkif_rsp_to_req_status(bret->status); 1441 1442 /* Wait the second response if not yet here. */ 1443 if (s2->status == REQ_WAITING) 1444 return 0; 1445 1446 bret->status = blkif_get_final_status(s->status, 1447 s2->status); 1448 1449 /* 1450 * All the grants is stored in the first shadow in order 1451 * to make the completion code simpler. 1452 */ 1453 num_grant += s2->req.u.rw.nr_segments; 1454 1455 /* 1456 * The two responses may not come in order. Only the 1457 * first request will store the scatter-gather list. 1458 */ 1459 if (s2->num_sg != 0) { 1460 /* Update "id" with the ID of the first response. */ 1461 *id = s->associated_id; 1462 s = s2; 1463 } 1464 1465 /* 1466 * We don't need anymore the second request, so recycling 1467 * it now. 1468 */ 1469 if (add_id_to_freelist(rinfo, s->associated_id)) 1470 WARN(1, "%s: can't recycle the second part (id = %ld) of the request\n", 1471 info->gd->disk_name, s->associated_id); 1472 } 1473 1474 data.s = s; 1475 num_sg = s->num_sg; 1476 1477 if (bret->operation == BLKIF_OP_READ && info->feature_persistent) { 1478 for_each_sg(s->sg, sg, num_sg, i) { 1479 BUG_ON(sg->offset + sg->length > PAGE_SIZE); 1480 1481 data.bvec_offset = sg->offset; 1482 data.bvec_data = kmap_atomic(sg_page(sg)); 1483 1484 gnttab_foreach_grant_in_range(sg_page(sg), 1485 sg->offset, 1486 sg->length, 1487 blkif_copy_from_grant, 1488 &data); 1489 1490 kunmap_atomic(data.bvec_data); 1491 } 1492 } 1493 /* Add the persistent grant into the list of free grants */ 1494 for (i = 0; i < num_grant; i++) { 1495 if (gnttab_query_foreign_access(s->grants_used[i]->gref)) { 1496 /* 1497 * If the grant is still mapped by the backend (the 1498 * backend has chosen to make this grant persistent) 1499 * we add it at the head of the list, so it will be 1500 * reused first. 1501 */ 1502 if (!info->feature_persistent) 1503 pr_alert_ratelimited("backed has not unmapped grant: %u\n", 1504 s->grants_used[i]->gref); 1505 list_add(&s->grants_used[i]->node, &rinfo->grants); 1506 rinfo->persistent_gnts_c++; 1507 } else { 1508 /* 1509 * If the grant is not mapped by the backend we end the 1510 * foreign access and add it to the tail of the list, 1511 * so it will not be picked again unless we run out of 1512 * persistent grants. 1513 */ 1514 gnttab_end_foreign_access(s->grants_used[i]->gref, 0, 0UL); 1515 s->grants_used[i]->gref = GRANT_INVALID_REF; 1516 list_add_tail(&s->grants_used[i]->node, &rinfo->grants); 1517 } 1518 } 1519 if (s->req.operation == BLKIF_OP_INDIRECT) { 1520 for (i = 0; i < INDIRECT_GREFS(num_grant); i++) { 1521 if (gnttab_query_foreign_access(s->indirect_grants[i]->gref)) { 1522 if (!info->feature_persistent) 1523 pr_alert_ratelimited("backed has not unmapped grant: %u\n", 1524 s->indirect_grants[i]->gref); 1525 list_add(&s->indirect_grants[i]->node, &rinfo->grants); 1526 rinfo->persistent_gnts_c++; 1527 } else { 1528 struct page *indirect_page; 1529 1530 gnttab_end_foreign_access(s->indirect_grants[i]->gref, 0, 0UL); 1531 /* 1532 * Add the used indirect page back to the list of 1533 * available pages for indirect grefs. 1534 */ 1535 if (!info->feature_persistent) { 1536 indirect_page = s->indirect_grants[i]->page; 1537 list_add(&indirect_page->lru, &rinfo->indirect_pages); 1538 } 1539 s->indirect_grants[i]->gref = GRANT_INVALID_REF; 1540 list_add_tail(&s->indirect_grants[i]->node, &rinfo->grants); 1541 } 1542 } 1543 } 1544 1545 return 1; 1546 } 1547 1548 static irqreturn_t blkif_interrupt(int irq, void *dev_id) 1549 { 1550 struct request *req; 1551 struct blkif_response *bret; 1552 RING_IDX i, rp; 1553 unsigned long flags; 1554 struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)dev_id; 1555 struct blkfront_info *info = rinfo->dev_info; 1556 1557 if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) 1558 return IRQ_HANDLED; 1559 1560 spin_lock_irqsave(&rinfo->ring_lock, flags); 1561 again: 1562 rp = rinfo->ring.sring->rsp_prod; 1563 rmb(); /* Ensure we see queued responses up to 'rp'. */ 1564 1565 for (i = rinfo->ring.rsp_cons; i != rp; i++) { 1566 unsigned long id; 1567 1568 bret = RING_GET_RESPONSE(&rinfo->ring, i); 1569 id = bret->id; 1570 /* 1571 * The backend has messed up and given us an id that we would 1572 * never have given to it (we stamp it up to BLK_RING_SIZE - 1573 * look in get_id_from_freelist. 1574 */ 1575 if (id >= BLK_RING_SIZE(info)) { 1576 WARN(1, "%s: response to %s has incorrect id (%ld)\n", 1577 info->gd->disk_name, op_name(bret->operation), id); 1578 /* We can't safely get the 'struct request' as 1579 * the id is busted. */ 1580 continue; 1581 } 1582 req = rinfo->shadow[id].request; 1583 1584 if (bret->operation != BLKIF_OP_DISCARD) { 1585 /* 1586 * We may need to wait for an extra response if the 1587 * I/O request is split in 2 1588 */ 1589 if (!blkif_completion(&id, rinfo, bret)) 1590 continue; 1591 } 1592 1593 if (add_id_to_freelist(rinfo, id)) { 1594 WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n", 1595 info->gd->disk_name, op_name(bret->operation), id); 1596 continue; 1597 } 1598 1599 if (bret->status == BLKIF_RSP_OKAY) 1600 blkif_req(req)->error = BLK_STS_OK; 1601 else 1602 blkif_req(req)->error = BLK_STS_IOERR; 1603 1604 switch (bret->operation) { 1605 case BLKIF_OP_DISCARD: 1606 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { 1607 struct request_queue *rq = info->rq; 1608 printk(KERN_WARNING "blkfront: %s: %s op failed\n", 1609 info->gd->disk_name, op_name(bret->operation)); 1610 blkif_req(req)->error = BLK_STS_NOTSUPP; 1611 info->feature_discard = 0; 1612 info->feature_secdiscard = 0; 1613 blk_queue_flag_clear(QUEUE_FLAG_DISCARD, rq); 1614 blk_queue_flag_clear(QUEUE_FLAG_SECERASE, rq); 1615 } 1616 break; 1617 case BLKIF_OP_FLUSH_DISKCACHE: 1618 case BLKIF_OP_WRITE_BARRIER: 1619 if (unlikely(bret->status == BLKIF_RSP_EOPNOTSUPP)) { 1620 printk(KERN_WARNING "blkfront: %s: %s op failed\n", 1621 info->gd->disk_name, op_name(bret->operation)); 1622 blkif_req(req)->error = BLK_STS_NOTSUPP; 1623 } 1624 if (unlikely(bret->status == BLKIF_RSP_ERROR && 1625 rinfo->shadow[id].req.u.rw.nr_segments == 0)) { 1626 printk(KERN_WARNING "blkfront: %s: empty %s op failed\n", 1627 info->gd->disk_name, op_name(bret->operation)); 1628 blkif_req(req)->error = BLK_STS_NOTSUPP; 1629 } 1630 if (unlikely(blkif_req(req)->error)) { 1631 if (blkif_req(req)->error == BLK_STS_NOTSUPP) 1632 blkif_req(req)->error = BLK_STS_OK; 1633 info->feature_fua = 0; 1634 info->feature_flush = 0; 1635 xlvbd_flush(info); 1636 } 1637 /* fall through */ 1638 case BLKIF_OP_READ: 1639 case BLKIF_OP_WRITE: 1640 if (unlikely(bret->status != BLKIF_RSP_OKAY)) 1641 dev_dbg(&info->xbdev->dev, "Bad return from blkdev data " 1642 "request: %x\n", bret->status); 1643 1644 break; 1645 default: 1646 BUG(); 1647 } 1648 1649 blk_mq_complete_request(req); 1650 } 1651 1652 rinfo->ring.rsp_cons = i; 1653 1654 if (i != rinfo->ring.req_prod_pvt) { 1655 int more_to_do; 1656 RING_FINAL_CHECK_FOR_RESPONSES(&rinfo->ring, more_to_do); 1657 if (more_to_do) 1658 goto again; 1659 } else 1660 rinfo->ring.sring->rsp_event = i + 1; 1661 1662 kick_pending_request_queues_locked(rinfo); 1663 1664 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 1665 1666 return IRQ_HANDLED; 1667 } 1668 1669 1670 static int setup_blkring(struct xenbus_device *dev, 1671 struct blkfront_ring_info *rinfo) 1672 { 1673 struct blkif_sring *sring; 1674 int err, i; 1675 struct blkfront_info *info = rinfo->dev_info; 1676 unsigned long ring_size = info->nr_ring_pages * XEN_PAGE_SIZE; 1677 grant_ref_t gref[XENBUS_MAX_RING_GRANTS]; 1678 1679 for (i = 0; i < info->nr_ring_pages; i++) 1680 rinfo->ring_ref[i] = GRANT_INVALID_REF; 1681 1682 sring = (struct blkif_sring *)__get_free_pages(GFP_NOIO | __GFP_HIGH, 1683 get_order(ring_size)); 1684 if (!sring) { 1685 xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring"); 1686 return -ENOMEM; 1687 } 1688 SHARED_RING_INIT(sring); 1689 FRONT_RING_INIT(&rinfo->ring, sring, ring_size); 1690 1691 err = xenbus_grant_ring(dev, rinfo->ring.sring, info->nr_ring_pages, gref); 1692 if (err < 0) { 1693 free_pages((unsigned long)sring, get_order(ring_size)); 1694 rinfo->ring.sring = NULL; 1695 goto fail; 1696 } 1697 for (i = 0; i < info->nr_ring_pages; i++) 1698 rinfo->ring_ref[i] = gref[i]; 1699 1700 err = xenbus_alloc_evtchn(dev, &rinfo->evtchn); 1701 if (err) 1702 goto fail; 1703 1704 err = bind_evtchn_to_irqhandler(rinfo->evtchn, blkif_interrupt, 0, 1705 "blkif", rinfo); 1706 if (err <= 0) { 1707 xenbus_dev_fatal(dev, err, 1708 "bind_evtchn_to_irqhandler failed"); 1709 goto fail; 1710 } 1711 rinfo->irq = err; 1712 1713 return 0; 1714 fail: 1715 blkif_free(info, 0); 1716 return err; 1717 } 1718 1719 /* 1720 * Write out per-ring/queue nodes including ring-ref and event-channel, and each 1721 * ring buffer may have multi pages depending on ->nr_ring_pages. 1722 */ 1723 static int write_per_ring_nodes(struct xenbus_transaction xbt, 1724 struct blkfront_ring_info *rinfo, const char *dir) 1725 { 1726 int err; 1727 unsigned int i; 1728 const char *message = NULL; 1729 struct blkfront_info *info = rinfo->dev_info; 1730 1731 if (info->nr_ring_pages == 1) { 1732 err = xenbus_printf(xbt, dir, "ring-ref", "%u", rinfo->ring_ref[0]); 1733 if (err) { 1734 message = "writing ring-ref"; 1735 goto abort_transaction; 1736 } 1737 } else { 1738 for (i = 0; i < info->nr_ring_pages; i++) { 1739 char ring_ref_name[RINGREF_NAME_LEN]; 1740 1741 snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i); 1742 err = xenbus_printf(xbt, dir, ring_ref_name, 1743 "%u", rinfo->ring_ref[i]); 1744 if (err) { 1745 message = "writing ring-ref"; 1746 goto abort_transaction; 1747 } 1748 } 1749 } 1750 1751 err = xenbus_printf(xbt, dir, "event-channel", "%u", rinfo->evtchn); 1752 if (err) { 1753 message = "writing event-channel"; 1754 goto abort_transaction; 1755 } 1756 1757 return 0; 1758 1759 abort_transaction: 1760 xenbus_transaction_end(xbt, 1); 1761 if (message) 1762 xenbus_dev_fatal(info->xbdev, err, "%s", message); 1763 1764 return err; 1765 } 1766 1767 /* Common code used when first setting up, and when resuming. */ 1768 static int talk_to_blkback(struct xenbus_device *dev, 1769 struct blkfront_info *info) 1770 { 1771 const char *message = NULL; 1772 struct xenbus_transaction xbt; 1773 int err; 1774 unsigned int i, max_page_order; 1775 unsigned int ring_page_order; 1776 1777 if (!info) 1778 return -ENODEV; 1779 1780 max_page_order = xenbus_read_unsigned(info->xbdev->otherend, 1781 "max-ring-page-order", 0); 1782 ring_page_order = min(xen_blkif_max_ring_order, max_page_order); 1783 info->nr_ring_pages = 1 << ring_page_order; 1784 1785 err = negotiate_mq(info); 1786 if (err) 1787 goto destroy_blkring; 1788 1789 for (i = 0; i < info->nr_rings; i++) { 1790 struct blkfront_ring_info *rinfo = &info->rinfo[i]; 1791 1792 /* Create shared ring, alloc event channel. */ 1793 err = setup_blkring(dev, rinfo); 1794 if (err) 1795 goto destroy_blkring; 1796 } 1797 1798 again: 1799 err = xenbus_transaction_start(&xbt); 1800 if (err) { 1801 xenbus_dev_fatal(dev, err, "starting transaction"); 1802 goto destroy_blkring; 1803 } 1804 1805 if (info->nr_ring_pages > 1) { 1806 err = xenbus_printf(xbt, dev->nodename, "ring-page-order", "%u", 1807 ring_page_order); 1808 if (err) { 1809 message = "writing ring-page-order"; 1810 goto abort_transaction; 1811 } 1812 } 1813 1814 /* We already got the number of queues/rings in _probe */ 1815 if (info->nr_rings == 1) { 1816 err = write_per_ring_nodes(xbt, &info->rinfo[0], dev->nodename); 1817 if (err) 1818 goto destroy_blkring; 1819 } else { 1820 char *path; 1821 size_t pathsize; 1822 1823 err = xenbus_printf(xbt, dev->nodename, "multi-queue-num-queues", "%u", 1824 info->nr_rings); 1825 if (err) { 1826 message = "writing multi-queue-num-queues"; 1827 goto abort_transaction; 1828 } 1829 1830 pathsize = strlen(dev->nodename) + QUEUE_NAME_LEN; 1831 path = kmalloc(pathsize, GFP_KERNEL); 1832 if (!path) { 1833 err = -ENOMEM; 1834 message = "ENOMEM while writing ring references"; 1835 goto abort_transaction; 1836 } 1837 1838 for (i = 0; i < info->nr_rings; i++) { 1839 memset(path, 0, pathsize); 1840 snprintf(path, pathsize, "%s/queue-%u", dev->nodename, i); 1841 err = write_per_ring_nodes(xbt, &info->rinfo[i], path); 1842 if (err) { 1843 kfree(path); 1844 goto destroy_blkring; 1845 } 1846 } 1847 kfree(path); 1848 } 1849 err = xenbus_printf(xbt, dev->nodename, "protocol", "%s", 1850 XEN_IO_PROTO_ABI_NATIVE); 1851 if (err) { 1852 message = "writing protocol"; 1853 goto abort_transaction; 1854 } 1855 err = xenbus_printf(xbt, dev->nodename, 1856 "feature-persistent", "%u", 1); 1857 if (err) 1858 dev_warn(&dev->dev, 1859 "writing persistent grants feature to xenbus"); 1860 1861 err = xenbus_transaction_end(xbt, 0); 1862 if (err) { 1863 if (err == -EAGAIN) 1864 goto again; 1865 xenbus_dev_fatal(dev, err, "completing transaction"); 1866 goto destroy_blkring; 1867 } 1868 1869 for (i = 0; i < info->nr_rings; i++) { 1870 unsigned int j; 1871 struct blkfront_ring_info *rinfo = &info->rinfo[i]; 1872 1873 for (j = 0; j < BLK_RING_SIZE(info); j++) 1874 rinfo->shadow[j].req.u.rw.id = j + 1; 1875 rinfo->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff; 1876 } 1877 xenbus_switch_state(dev, XenbusStateInitialised); 1878 1879 return 0; 1880 1881 abort_transaction: 1882 xenbus_transaction_end(xbt, 1); 1883 if (message) 1884 xenbus_dev_fatal(dev, err, "%s", message); 1885 destroy_blkring: 1886 blkif_free(info, 0); 1887 1888 kfree(info); 1889 dev_set_drvdata(&dev->dev, NULL); 1890 1891 return err; 1892 } 1893 1894 static int negotiate_mq(struct blkfront_info *info) 1895 { 1896 unsigned int backend_max_queues; 1897 unsigned int i; 1898 1899 BUG_ON(info->nr_rings); 1900 1901 /* Check if backend supports multiple queues. */ 1902 backend_max_queues = xenbus_read_unsigned(info->xbdev->otherend, 1903 "multi-queue-max-queues", 1); 1904 info->nr_rings = min(backend_max_queues, xen_blkif_max_queues); 1905 /* We need at least one ring. */ 1906 if (!info->nr_rings) 1907 info->nr_rings = 1; 1908 1909 info->rinfo = kcalloc(info->nr_rings, 1910 sizeof(struct blkfront_ring_info), 1911 GFP_KERNEL); 1912 if (!info->rinfo) { 1913 xenbus_dev_fatal(info->xbdev, -ENOMEM, "allocating ring_info structure"); 1914 return -ENOMEM; 1915 } 1916 1917 for (i = 0; i < info->nr_rings; i++) { 1918 struct blkfront_ring_info *rinfo; 1919 1920 rinfo = &info->rinfo[i]; 1921 INIT_LIST_HEAD(&rinfo->indirect_pages); 1922 INIT_LIST_HEAD(&rinfo->grants); 1923 rinfo->dev_info = info; 1924 INIT_WORK(&rinfo->work, blkif_restart_queue); 1925 spin_lock_init(&rinfo->ring_lock); 1926 } 1927 return 0; 1928 } 1929 /** 1930 * Entry point to this code when a new device is created. Allocate the basic 1931 * structures and the ring buffer for communication with the backend, and 1932 * inform the backend of the appropriate details for those. Switch to 1933 * Initialised state. 1934 */ 1935 static int blkfront_probe(struct xenbus_device *dev, 1936 const struct xenbus_device_id *id) 1937 { 1938 int err, vdevice; 1939 struct blkfront_info *info; 1940 1941 /* FIXME: Use dynamic device id if this is not set. */ 1942 err = xenbus_scanf(XBT_NIL, dev->nodename, 1943 "virtual-device", "%i", &vdevice); 1944 if (err != 1) { 1945 /* go looking in the extended area instead */ 1946 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext", 1947 "%i", &vdevice); 1948 if (err != 1) { 1949 xenbus_dev_fatal(dev, err, "reading virtual-device"); 1950 return err; 1951 } 1952 } 1953 1954 if (xen_hvm_domain()) { 1955 char *type; 1956 int len; 1957 /* no unplug has been done: do not hook devices != xen vbds */ 1958 if (xen_has_pv_and_legacy_disk_devices()) { 1959 int major; 1960 1961 if (!VDEV_IS_EXTENDED(vdevice)) 1962 major = BLKIF_MAJOR(vdevice); 1963 else 1964 major = XENVBD_MAJOR; 1965 1966 if (major != XENVBD_MAJOR) { 1967 printk(KERN_INFO 1968 "%s: HVM does not support vbd %d as xen block device\n", 1969 __func__, vdevice); 1970 return -ENODEV; 1971 } 1972 } 1973 /* do not create a PV cdrom device if we are an HVM guest */ 1974 type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len); 1975 if (IS_ERR(type)) 1976 return -ENODEV; 1977 if (strncmp(type, "cdrom", 5) == 0) { 1978 kfree(type); 1979 return -ENODEV; 1980 } 1981 kfree(type); 1982 } 1983 info = kzalloc(sizeof(*info), GFP_KERNEL); 1984 if (!info) { 1985 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure"); 1986 return -ENOMEM; 1987 } 1988 1989 info->xbdev = dev; 1990 1991 mutex_init(&info->mutex); 1992 info->vdevice = vdevice; 1993 info->connected = BLKIF_STATE_DISCONNECTED; 1994 1995 /* Front end dir is a number, which is used as the id. */ 1996 info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0); 1997 dev_set_drvdata(&dev->dev, info); 1998 1999 return 0; 2000 } 2001 2002 static int blkif_recover(struct blkfront_info *info) 2003 { 2004 unsigned int r_index; 2005 struct request *req, *n; 2006 int rc; 2007 struct bio *bio; 2008 unsigned int segs; 2009 2010 blkfront_gather_backend_features(info); 2011 /* Reset limits changed by blk_mq_update_nr_hw_queues(). */ 2012 blkif_set_queue_limits(info); 2013 segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST; 2014 blk_queue_max_segments(info->rq, segs / GRANTS_PER_PSEG); 2015 2016 for (r_index = 0; r_index < info->nr_rings; r_index++) { 2017 struct blkfront_ring_info *rinfo = &info->rinfo[r_index]; 2018 2019 rc = blkfront_setup_indirect(rinfo); 2020 if (rc) 2021 return rc; 2022 } 2023 xenbus_switch_state(info->xbdev, XenbusStateConnected); 2024 2025 /* Now safe for us to use the shared ring */ 2026 info->connected = BLKIF_STATE_CONNECTED; 2027 2028 for (r_index = 0; r_index < info->nr_rings; r_index++) { 2029 struct blkfront_ring_info *rinfo; 2030 2031 rinfo = &info->rinfo[r_index]; 2032 /* Kick any other new requests queued since we resumed */ 2033 kick_pending_request_queues(rinfo); 2034 } 2035 2036 list_for_each_entry_safe(req, n, &info->requests, queuelist) { 2037 /* Requeue pending requests (flush or discard) */ 2038 list_del_init(&req->queuelist); 2039 BUG_ON(req->nr_phys_segments > segs); 2040 blk_mq_requeue_request(req, false); 2041 } 2042 blk_mq_start_stopped_hw_queues(info->rq, true); 2043 blk_mq_kick_requeue_list(info->rq); 2044 2045 while ((bio = bio_list_pop(&info->bio_list)) != NULL) { 2046 /* Traverse the list of pending bios and re-queue them */ 2047 submit_bio(bio); 2048 } 2049 2050 return 0; 2051 } 2052 2053 /** 2054 * We are reconnecting to the backend, due to a suspend/resume, or a backend 2055 * driver restart. We tear down our blkif structure and recreate it, but 2056 * leave the device-layer structures intact so that this is transparent to the 2057 * rest of the kernel. 2058 */ 2059 static int blkfront_resume(struct xenbus_device *dev) 2060 { 2061 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 2062 int err = 0; 2063 unsigned int i, j; 2064 2065 dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename); 2066 2067 bio_list_init(&info->bio_list); 2068 INIT_LIST_HEAD(&info->requests); 2069 for (i = 0; i < info->nr_rings; i++) { 2070 struct blkfront_ring_info *rinfo = &info->rinfo[i]; 2071 struct bio_list merge_bio; 2072 struct blk_shadow *shadow = rinfo->shadow; 2073 2074 for (j = 0; j < BLK_RING_SIZE(info); j++) { 2075 /* Not in use? */ 2076 if (!shadow[j].request) 2077 continue; 2078 2079 /* 2080 * Get the bios in the request so we can re-queue them. 2081 */ 2082 if (req_op(shadow[j].request) == REQ_OP_FLUSH || 2083 req_op(shadow[j].request) == REQ_OP_DISCARD || 2084 req_op(shadow[j].request) == REQ_OP_SECURE_ERASE || 2085 shadow[j].request->cmd_flags & REQ_FUA) { 2086 /* 2087 * Flush operations don't contain bios, so 2088 * we need to requeue the whole request 2089 * 2090 * XXX: but this doesn't make any sense for a 2091 * write with the FUA flag set.. 2092 */ 2093 list_add(&shadow[j].request->queuelist, &info->requests); 2094 continue; 2095 } 2096 merge_bio.head = shadow[j].request->bio; 2097 merge_bio.tail = shadow[j].request->biotail; 2098 bio_list_merge(&info->bio_list, &merge_bio); 2099 shadow[j].request->bio = NULL; 2100 blk_mq_end_request(shadow[j].request, BLK_STS_OK); 2101 } 2102 } 2103 2104 blkif_free(info, info->connected == BLKIF_STATE_CONNECTED); 2105 2106 err = talk_to_blkback(dev, info); 2107 if (!err) 2108 blk_mq_update_nr_hw_queues(&info->tag_set, info->nr_rings); 2109 2110 /* 2111 * We have to wait for the backend to switch to 2112 * connected state, since we want to read which 2113 * features it supports. 2114 */ 2115 2116 return err; 2117 } 2118 2119 static void blkfront_closing(struct blkfront_info *info) 2120 { 2121 struct xenbus_device *xbdev = info->xbdev; 2122 struct block_device *bdev = NULL; 2123 2124 mutex_lock(&info->mutex); 2125 2126 if (xbdev->state == XenbusStateClosing) { 2127 mutex_unlock(&info->mutex); 2128 return; 2129 } 2130 2131 if (info->gd) 2132 bdev = bdget_disk(info->gd, 0); 2133 2134 mutex_unlock(&info->mutex); 2135 2136 if (!bdev) { 2137 xenbus_frontend_closed(xbdev); 2138 return; 2139 } 2140 2141 mutex_lock(&bdev->bd_mutex); 2142 2143 if (bdev->bd_openers) { 2144 xenbus_dev_error(xbdev, -EBUSY, 2145 "Device in use; refusing to close"); 2146 xenbus_switch_state(xbdev, XenbusStateClosing); 2147 } else { 2148 xlvbd_release_gendisk(info); 2149 xenbus_frontend_closed(xbdev); 2150 } 2151 2152 mutex_unlock(&bdev->bd_mutex); 2153 bdput(bdev); 2154 } 2155 2156 static void blkfront_setup_discard(struct blkfront_info *info) 2157 { 2158 int err; 2159 unsigned int discard_granularity; 2160 unsigned int discard_alignment; 2161 2162 info->feature_discard = 1; 2163 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 2164 "discard-granularity", "%u", &discard_granularity, 2165 "discard-alignment", "%u", &discard_alignment, 2166 NULL); 2167 if (!err) { 2168 info->discard_granularity = discard_granularity; 2169 info->discard_alignment = discard_alignment; 2170 } 2171 info->feature_secdiscard = 2172 !!xenbus_read_unsigned(info->xbdev->otherend, "discard-secure", 2173 0); 2174 } 2175 2176 static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo) 2177 { 2178 unsigned int psegs, grants; 2179 int err, i; 2180 struct blkfront_info *info = rinfo->dev_info; 2181 2182 if (info->max_indirect_segments == 0) { 2183 if (!HAS_EXTRA_REQ) 2184 grants = BLKIF_MAX_SEGMENTS_PER_REQUEST; 2185 else { 2186 /* 2187 * When an extra req is required, the maximum 2188 * grants supported is related to the size of the 2189 * Linux block segment. 2190 */ 2191 grants = GRANTS_PER_PSEG; 2192 } 2193 } 2194 else 2195 grants = info->max_indirect_segments; 2196 psegs = DIV_ROUND_UP(grants, GRANTS_PER_PSEG); 2197 2198 err = fill_grant_buffer(rinfo, 2199 (grants + INDIRECT_GREFS(grants)) * BLK_RING_SIZE(info)); 2200 if (err) 2201 goto out_of_memory; 2202 2203 if (!info->feature_persistent && info->max_indirect_segments) { 2204 /* 2205 * We are using indirect descriptors but not persistent 2206 * grants, we need to allocate a set of pages that can be 2207 * used for mapping indirect grefs 2208 */ 2209 int num = INDIRECT_GREFS(grants) * BLK_RING_SIZE(info); 2210 2211 BUG_ON(!list_empty(&rinfo->indirect_pages)); 2212 for (i = 0; i < num; i++) { 2213 struct page *indirect_page = alloc_page(GFP_NOIO); 2214 if (!indirect_page) 2215 goto out_of_memory; 2216 list_add(&indirect_page->lru, &rinfo->indirect_pages); 2217 } 2218 } 2219 2220 for (i = 0; i < BLK_RING_SIZE(info); i++) { 2221 rinfo->shadow[i].grants_used = 2222 kcalloc(grants, 2223 sizeof(rinfo->shadow[i].grants_used[0]), 2224 GFP_NOIO); 2225 rinfo->shadow[i].sg = kcalloc(psegs, 2226 sizeof(rinfo->shadow[i].sg[0]), 2227 GFP_NOIO); 2228 if (info->max_indirect_segments) 2229 rinfo->shadow[i].indirect_grants = 2230 kcalloc(INDIRECT_GREFS(grants), 2231 sizeof(rinfo->shadow[i].indirect_grants[0]), 2232 GFP_NOIO); 2233 if ((rinfo->shadow[i].grants_used == NULL) || 2234 (rinfo->shadow[i].sg == NULL) || 2235 (info->max_indirect_segments && 2236 (rinfo->shadow[i].indirect_grants == NULL))) 2237 goto out_of_memory; 2238 sg_init_table(rinfo->shadow[i].sg, psegs); 2239 } 2240 2241 2242 return 0; 2243 2244 out_of_memory: 2245 for (i = 0; i < BLK_RING_SIZE(info); i++) { 2246 kfree(rinfo->shadow[i].grants_used); 2247 rinfo->shadow[i].grants_used = NULL; 2248 kfree(rinfo->shadow[i].sg); 2249 rinfo->shadow[i].sg = NULL; 2250 kfree(rinfo->shadow[i].indirect_grants); 2251 rinfo->shadow[i].indirect_grants = NULL; 2252 } 2253 if (!list_empty(&rinfo->indirect_pages)) { 2254 struct page *indirect_page, *n; 2255 list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { 2256 list_del(&indirect_page->lru); 2257 __free_page(indirect_page); 2258 } 2259 } 2260 return -ENOMEM; 2261 } 2262 2263 /* 2264 * Gather all backend feature-* 2265 */ 2266 static void blkfront_gather_backend_features(struct blkfront_info *info) 2267 { 2268 unsigned int indirect_segments; 2269 2270 info->feature_flush = 0; 2271 info->feature_fua = 0; 2272 2273 /* 2274 * If there's no "feature-barrier" defined, then it means 2275 * we're dealing with a very old backend which writes 2276 * synchronously; nothing to do. 2277 * 2278 * If there are barriers, then we use flush. 2279 */ 2280 if (xenbus_read_unsigned(info->xbdev->otherend, "feature-barrier", 0)) { 2281 info->feature_flush = 1; 2282 info->feature_fua = 1; 2283 } 2284 2285 /* 2286 * And if there is "feature-flush-cache" use that above 2287 * barriers. 2288 */ 2289 if (xenbus_read_unsigned(info->xbdev->otherend, "feature-flush-cache", 2290 0)) { 2291 info->feature_flush = 1; 2292 info->feature_fua = 0; 2293 } 2294 2295 if (xenbus_read_unsigned(info->xbdev->otherend, "feature-discard", 0)) 2296 blkfront_setup_discard(info); 2297 2298 info->feature_persistent = 2299 !!xenbus_read_unsigned(info->xbdev->otherend, 2300 "feature-persistent", 0); 2301 2302 indirect_segments = xenbus_read_unsigned(info->xbdev->otherend, 2303 "feature-max-indirect-segments", 0); 2304 if (indirect_segments > xen_blkif_max_segments) 2305 indirect_segments = xen_blkif_max_segments; 2306 if (indirect_segments <= BLKIF_MAX_SEGMENTS_PER_REQUEST) 2307 indirect_segments = 0; 2308 info->max_indirect_segments = indirect_segments; 2309 } 2310 2311 /* 2312 * Invoked when the backend is finally 'ready' (and has told produced 2313 * the details about the physical device - #sectors, size, etc). 2314 */ 2315 static void blkfront_connect(struct blkfront_info *info) 2316 { 2317 unsigned long long sectors; 2318 unsigned long sector_size; 2319 unsigned int physical_sector_size; 2320 unsigned int binfo; 2321 char *envp[] = { "RESIZE=1", NULL }; 2322 int err, i; 2323 2324 switch (info->connected) { 2325 case BLKIF_STATE_CONNECTED: 2326 /* 2327 * Potentially, the back-end may be signalling 2328 * a capacity change; update the capacity. 2329 */ 2330 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, 2331 "sectors", "%Lu", §ors); 2332 if (XENBUS_EXIST_ERR(err)) 2333 return; 2334 printk(KERN_INFO "Setting capacity to %Lu\n", 2335 sectors); 2336 set_capacity(info->gd, sectors); 2337 revalidate_disk(info->gd); 2338 kobject_uevent_env(&disk_to_dev(info->gd)->kobj, 2339 KOBJ_CHANGE, envp); 2340 2341 return; 2342 case BLKIF_STATE_SUSPENDED: 2343 /* 2344 * If we are recovering from suspension, we need to wait 2345 * for the backend to announce it's features before 2346 * reconnecting, at least we need to know if the backend 2347 * supports indirect descriptors, and how many. 2348 */ 2349 blkif_recover(info); 2350 return; 2351 2352 default: 2353 break; 2354 } 2355 2356 dev_dbg(&info->xbdev->dev, "%s:%s.\n", 2357 __func__, info->xbdev->otherend); 2358 2359 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 2360 "sectors", "%llu", §ors, 2361 "info", "%u", &binfo, 2362 "sector-size", "%lu", §or_size, 2363 NULL); 2364 if (err) { 2365 xenbus_dev_fatal(info->xbdev, err, 2366 "reading backend fields at %s", 2367 info->xbdev->otherend); 2368 return; 2369 } 2370 2371 /* 2372 * physcial-sector-size is a newer field, so old backends may not 2373 * provide this. Assume physical sector size to be the same as 2374 * sector_size in that case. 2375 */ 2376 physical_sector_size = xenbus_read_unsigned(info->xbdev->otherend, 2377 "physical-sector-size", 2378 sector_size); 2379 blkfront_gather_backend_features(info); 2380 for (i = 0; i < info->nr_rings; i++) { 2381 err = blkfront_setup_indirect(&info->rinfo[i]); 2382 if (err) { 2383 xenbus_dev_fatal(info->xbdev, err, "setup_indirect at %s", 2384 info->xbdev->otherend); 2385 blkif_free(info, 0); 2386 break; 2387 } 2388 } 2389 2390 err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size, 2391 physical_sector_size); 2392 if (err) { 2393 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s", 2394 info->xbdev->otherend); 2395 goto fail; 2396 } 2397 2398 xenbus_switch_state(info->xbdev, XenbusStateConnected); 2399 2400 /* Kick pending requests. */ 2401 info->connected = BLKIF_STATE_CONNECTED; 2402 for (i = 0; i < info->nr_rings; i++) 2403 kick_pending_request_queues(&info->rinfo[i]); 2404 2405 device_add_disk(&info->xbdev->dev, info->gd); 2406 2407 info->is_ready = 1; 2408 return; 2409 2410 fail: 2411 blkif_free(info, 0); 2412 return; 2413 } 2414 2415 /** 2416 * Callback received when the backend's state changes. 2417 */ 2418 static void blkback_changed(struct xenbus_device *dev, 2419 enum xenbus_state backend_state) 2420 { 2421 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 2422 2423 dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state); 2424 2425 switch (backend_state) { 2426 case XenbusStateInitWait: 2427 if (dev->state != XenbusStateInitialising) 2428 break; 2429 if (talk_to_blkback(dev, info)) 2430 break; 2431 case XenbusStateInitialising: 2432 case XenbusStateInitialised: 2433 case XenbusStateReconfiguring: 2434 case XenbusStateReconfigured: 2435 case XenbusStateUnknown: 2436 break; 2437 2438 case XenbusStateConnected: 2439 /* 2440 * talk_to_blkback sets state to XenbusStateInitialised 2441 * and blkfront_connect sets it to XenbusStateConnected 2442 * (if connection went OK). 2443 * 2444 * If the backend (or toolstack) decides to poke at backend 2445 * state (and re-trigger the watch by setting the state repeatedly 2446 * to XenbusStateConnected (4)) we need to deal with this. 2447 * This is allowed as this is used to communicate to the guest 2448 * that the size of disk has changed! 2449 */ 2450 if ((dev->state != XenbusStateInitialised) && 2451 (dev->state != XenbusStateConnected)) { 2452 if (talk_to_blkback(dev, info)) 2453 break; 2454 } 2455 2456 blkfront_connect(info); 2457 break; 2458 2459 case XenbusStateClosed: 2460 if (dev->state == XenbusStateClosed) 2461 break; 2462 /* fall through */ 2463 case XenbusStateClosing: 2464 if (info) 2465 blkfront_closing(info); 2466 break; 2467 } 2468 } 2469 2470 static int blkfront_remove(struct xenbus_device *xbdev) 2471 { 2472 struct blkfront_info *info = dev_get_drvdata(&xbdev->dev); 2473 struct block_device *bdev = NULL; 2474 struct gendisk *disk; 2475 2476 dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename); 2477 2478 blkif_free(info, 0); 2479 2480 mutex_lock(&info->mutex); 2481 2482 disk = info->gd; 2483 if (disk) 2484 bdev = bdget_disk(disk, 0); 2485 2486 info->xbdev = NULL; 2487 mutex_unlock(&info->mutex); 2488 2489 if (!bdev) { 2490 kfree(info); 2491 return 0; 2492 } 2493 2494 /* 2495 * The xbdev was removed before we reached the Closed 2496 * state. See if it's safe to remove the disk. If the bdev 2497 * isn't closed yet, we let release take care of it. 2498 */ 2499 2500 mutex_lock(&bdev->bd_mutex); 2501 info = disk->private_data; 2502 2503 dev_warn(disk_to_dev(disk), 2504 "%s was hot-unplugged, %d stale handles\n", 2505 xbdev->nodename, bdev->bd_openers); 2506 2507 if (info && !bdev->bd_openers) { 2508 xlvbd_release_gendisk(info); 2509 disk->private_data = NULL; 2510 kfree(info); 2511 } 2512 2513 mutex_unlock(&bdev->bd_mutex); 2514 bdput(bdev); 2515 2516 return 0; 2517 } 2518 2519 static int blkfront_is_ready(struct xenbus_device *dev) 2520 { 2521 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 2522 2523 return info->is_ready && info->xbdev; 2524 } 2525 2526 static int blkif_open(struct block_device *bdev, fmode_t mode) 2527 { 2528 struct gendisk *disk = bdev->bd_disk; 2529 struct blkfront_info *info; 2530 int err = 0; 2531 2532 mutex_lock(&blkfront_mutex); 2533 2534 info = disk->private_data; 2535 if (!info) { 2536 /* xbdev gone */ 2537 err = -ERESTARTSYS; 2538 goto out; 2539 } 2540 2541 mutex_lock(&info->mutex); 2542 2543 if (!info->gd) 2544 /* xbdev is closed */ 2545 err = -ERESTARTSYS; 2546 2547 mutex_unlock(&info->mutex); 2548 2549 out: 2550 mutex_unlock(&blkfront_mutex); 2551 return err; 2552 } 2553 2554 static void blkif_release(struct gendisk *disk, fmode_t mode) 2555 { 2556 struct blkfront_info *info = disk->private_data; 2557 struct block_device *bdev; 2558 struct xenbus_device *xbdev; 2559 2560 mutex_lock(&blkfront_mutex); 2561 2562 bdev = bdget_disk(disk, 0); 2563 2564 if (!bdev) { 2565 WARN(1, "Block device %s yanked out from us!\n", disk->disk_name); 2566 goto out_mutex; 2567 } 2568 if (bdev->bd_openers) 2569 goto out; 2570 2571 /* 2572 * Check if we have been instructed to close. We will have 2573 * deferred this request, because the bdev was still open. 2574 */ 2575 2576 mutex_lock(&info->mutex); 2577 xbdev = info->xbdev; 2578 2579 if (xbdev && xbdev->state == XenbusStateClosing) { 2580 /* pending switch to state closed */ 2581 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); 2582 xlvbd_release_gendisk(info); 2583 xenbus_frontend_closed(info->xbdev); 2584 } 2585 2586 mutex_unlock(&info->mutex); 2587 2588 if (!xbdev) { 2589 /* sudden device removal */ 2590 dev_info(disk_to_dev(bdev->bd_disk), "releasing disk\n"); 2591 xlvbd_release_gendisk(info); 2592 disk->private_data = NULL; 2593 kfree(info); 2594 } 2595 2596 out: 2597 bdput(bdev); 2598 out_mutex: 2599 mutex_unlock(&blkfront_mutex); 2600 } 2601 2602 static const struct block_device_operations xlvbd_block_fops = 2603 { 2604 .owner = THIS_MODULE, 2605 .open = blkif_open, 2606 .release = blkif_release, 2607 .getgeo = blkif_getgeo, 2608 .ioctl = blkif_ioctl, 2609 }; 2610 2611 2612 static const struct xenbus_device_id blkfront_ids[] = { 2613 { "vbd" }, 2614 { "" } 2615 }; 2616 2617 static struct xenbus_driver blkfront_driver = { 2618 .ids = blkfront_ids, 2619 .probe = blkfront_probe, 2620 .remove = blkfront_remove, 2621 .resume = blkfront_resume, 2622 .otherend_changed = blkback_changed, 2623 .is_ready = blkfront_is_ready, 2624 }; 2625 2626 static int __init xlblk_init(void) 2627 { 2628 int ret; 2629 int nr_cpus = num_online_cpus(); 2630 2631 if (!xen_domain()) 2632 return -ENODEV; 2633 2634 if (xen_blkif_max_segments < BLKIF_MAX_SEGMENTS_PER_REQUEST) 2635 xen_blkif_max_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; 2636 2637 if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { 2638 pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", 2639 xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); 2640 xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; 2641 } 2642 2643 if (xen_blkif_max_queues > nr_cpus) { 2644 pr_info("Invalid max_queues (%d), will use default max: %d.\n", 2645 xen_blkif_max_queues, nr_cpus); 2646 xen_blkif_max_queues = nr_cpus; 2647 } 2648 2649 if (!xen_has_pv_disk_devices()) 2650 return -ENODEV; 2651 2652 if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) { 2653 printk(KERN_WARNING "xen_blk: can't get major %d with name %s\n", 2654 XENVBD_MAJOR, DEV_NAME); 2655 return -ENODEV; 2656 } 2657 2658 ret = xenbus_register_frontend(&blkfront_driver); 2659 if (ret) { 2660 unregister_blkdev(XENVBD_MAJOR, DEV_NAME); 2661 return ret; 2662 } 2663 2664 return 0; 2665 } 2666 module_init(xlblk_init); 2667 2668 2669 static void __exit xlblk_exit(void) 2670 { 2671 xenbus_unregister_driver(&blkfront_driver); 2672 unregister_blkdev(XENVBD_MAJOR, DEV_NAME); 2673 kfree(minors); 2674 } 2675 module_exit(xlblk_exit); 2676 2677 MODULE_DESCRIPTION("Xen virtual block device frontend"); 2678 MODULE_LICENSE("GPL"); 2679 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR); 2680 MODULE_ALIAS("xen:vbd"); 2681 MODULE_ALIAS("xenblk"); 2682