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/major.h> 46 #include <linux/mutex.h> 47 #include <linux/scatterlist.h> 48 #include <linux/bitmap.h> 49 #include <linux/list.h> 50 #include <linux/workqueue.h> 51 #include <linux/sched/mm.h> 52 53 #include <xen/xen.h> 54 #include <xen/xenbus.h> 55 #include <xen/grant_table.h> 56 #include <xen/events.h> 57 #include <xen/page.h> 58 #include <xen/platform_pci.h> 59 60 #include <xen/interface/grant_table.h> 61 #include <xen/interface/io/blkif.h> 62 #include <xen/interface/io/protocols.h> 63 64 #include <asm/xen/hypervisor.h> 65 66 /* 67 * The minimal size of segment supported by the block framework is PAGE_SIZE. 68 * When Linux is using a different page size than Xen, it may not be possible 69 * to put all the data in a single segment. 70 * This can happen when the backend doesn't support indirect descriptor and 71 * therefore the maximum amount of data that a request can carry is 72 * BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE = 44KB 73 * 74 * Note that we only support one extra request. So the Linux page size 75 * should be <= ( 2 * BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE) = 76 * 88KB. 77 */ 78 #define HAS_EXTRA_REQ (BLKIF_MAX_SEGMENTS_PER_REQUEST < XEN_PFN_PER_PAGE) 79 80 enum blkif_state { 81 BLKIF_STATE_DISCONNECTED, 82 BLKIF_STATE_CONNECTED, 83 BLKIF_STATE_SUSPENDED, 84 BLKIF_STATE_ERROR, 85 }; 86 87 struct grant { 88 grant_ref_t gref; 89 struct page *page; 90 struct list_head node; 91 }; 92 93 enum blk_req_status { 94 REQ_PROCESSING, 95 REQ_WAITING, 96 REQ_DONE, 97 REQ_ERROR, 98 REQ_EOPNOTSUPP, 99 }; 100 101 struct blk_shadow { 102 struct blkif_request req; 103 struct request *request; 104 struct grant **grants_used; 105 struct grant **indirect_grants; 106 struct scatterlist *sg; 107 unsigned int num_sg; 108 enum blk_req_status status; 109 110 #define NO_ASSOCIATED_ID ~0UL 111 /* 112 * Id of the sibling if we ever need 2 requests when handling a 113 * block I/O request 114 */ 115 unsigned long associated_id; 116 }; 117 118 struct blkif_req { 119 blk_status_t error; 120 }; 121 122 static inline struct blkif_req *blkif_req(struct request *rq) 123 { 124 return blk_mq_rq_to_pdu(rq); 125 } 126 127 static DEFINE_MUTEX(blkfront_mutex); 128 static const struct block_device_operations xlvbd_block_fops; 129 static struct delayed_work blkfront_work; 130 static LIST_HEAD(info_list); 131 132 /* 133 * Maximum number of segments in indirect requests, the actual value used by 134 * the frontend driver is the minimum of this value and the value provided 135 * by the backend driver. 136 */ 137 138 static unsigned int xen_blkif_max_segments = 32; 139 module_param_named(max_indirect_segments, xen_blkif_max_segments, uint, 0444); 140 MODULE_PARM_DESC(max_indirect_segments, 141 "Maximum amount of segments in indirect requests (default is 32)"); 142 143 static unsigned int xen_blkif_max_queues = 4; 144 module_param_named(max_queues, xen_blkif_max_queues, uint, 0444); 145 MODULE_PARM_DESC(max_queues, "Maximum number of hardware queues/rings used per virtual disk"); 146 147 /* 148 * Maximum order of pages to be used for the shared ring between front and 149 * backend, 4KB page granularity is used. 150 */ 151 static unsigned int xen_blkif_max_ring_order; 152 module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, 0444); 153 MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring"); 154 155 static bool __read_mostly xen_blkif_trusted = true; 156 module_param_named(trusted, xen_blkif_trusted, bool, 0644); 157 MODULE_PARM_DESC(trusted, "Is the backend trusted"); 158 159 #define BLK_RING_SIZE(info) \ 160 __CONST_RING_SIZE(blkif, XEN_PAGE_SIZE * (info)->nr_ring_pages) 161 162 /* 163 * ring-ref%u i=(-1UL) would take 11 characters + 'ring-ref' is 8, so 19 164 * characters are enough. Define to 20 to keep consistent with backend. 165 */ 166 #define RINGREF_NAME_LEN (20) 167 /* 168 * queue-%u would take 7 + 10(UINT_MAX) = 17 characters. 169 */ 170 #define QUEUE_NAME_LEN (17) 171 172 /* 173 * Per-ring info. 174 * Every blkfront device can associate with one or more blkfront_ring_info, 175 * depending on how many hardware queues/rings to be used. 176 */ 177 struct blkfront_ring_info { 178 /* Lock to protect data in every ring buffer. */ 179 spinlock_t ring_lock; 180 struct blkif_front_ring ring; 181 unsigned int ring_ref[XENBUS_MAX_RING_GRANTS]; 182 unsigned int evtchn, irq; 183 struct work_struct work; 184 struct gnttab_free_callback callback; 185 struct list_head indirect_pages; 186 struct list_head grants; 187 unsigned int persistent_gnts_c; 188 unsigned long shadow_free; 189 struct blkfront_info *dev_info; 190 struct blk_shadow shadow[]; 191 }; 192 193 /* 194 * We have one of these per vbd, whether ide, scsi or 'other'. They 195 * hang in private_data off the gendisk structure. We may end up 196 * putting all kinds of interesting stuff here :-) 197 */ 198 struct blkfront_info 199 { 200 struct mutex mutex; 201 struct xenbus_device *xbdev; 202 struct gendisk *gd; 203 u16 sector_size; 204 unsigned int physical_sector_size; 205 unsigned long vdisk_info; 206 int vdevice; 207 blkif_vdev_t handle; 208 enum blkif_state connected; 209 /* Number of pages per ring buffer. */ 210 unsigned int nr_ring_pages; 211 struct request_queue *rq; 212 unsigned int feature_flush:1; 213 unsigned int feature_fua:1; 214 unsigned int feature_discard:1; 215 unsigned int feature_secdiscard:1; 216 unsigned int feature_persistent:1; 217 unsigned int bounce:1; 218 unsigned int discard_granularity; 219 unsigned int discard_alignment; 220 /* Number of 4KB segments handled */ 221 unsigned int max_indirect_segments; 222 int is_ready; 223 struct blk_mq_tag_set tag_set; 224 struct blkfront_ring_info *rinfo; 225 unsigned int nr_rings; 226 unsigned int rinfo_size; 227 /* Save uncomplete reqs and bios for migration. */ 228 struct list_head requests; 229 struct bio_list bio_list; 230 struct list_head info_list; 231 }; 232 233 static unsigned int nr_minors; 234 static unsigned long *minors; 235 static DEFINE_SPINLOCK(minor_lock); 236 237 #define PARTS_PER_DISK 16 238 #define PARTS_PER_EXT_DISK 256 239 240 #define BLKIF_MAJOR(dev) ((dev)>>8) 241 #define BLKIF_MINOR(dev) ((dev) & 0xff) 242 243 #define EXT_SHIFT 28 244 #define EXTENDED (1<<EXT_SHIFT) 245 #define VDEV_IS_EXTENDED(dev) ((dev)&(EXTENDED)) 246 #define BLKIF_MINOR_EXT(dev) ((dev)&(~EXTENDED)) 247 #define EMULATED_HD_DISK_MINOR_OFFSET (0) 248 #define EMULATED_HD_DISK_NAME_OFFSET (EMULATED_HD_DISK_MINOR_OFFSET / 256) 249 #define EMULATED_SD_DISK_MINOR_OFFSET (0) 250 #define EMULATED_SD_DISK_NAME_OFFSET (EMULATED_SD_DISK_MINOR_OFFSET / 256) 251 252 #define DEV_NAME "xvd" /* name in /dev */ 253 254 /* 255 * Grants are always the same size as a Xen page (i.e 4KB). 256 * A physical segment is always the same size as a Linux page. 257 * Number of grants per physical segment 258 */ 259 #define GRANTS_PER_PSEG (PAGE_SIZE / XEN_PAGE_SIZE) 260 261 #define GRANTS_PER_INDIRECT_FRAME \ 262 (XEN_PAGE_SIZE / sizeof(struct blkif_request_segment)) 263 264 #define INDIRECT_GREFS(_grants) \ 265 DIV_ROUND_UP(_grants, GRANTS_PER_INDIRECT_FRAME) 266 267 static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo); 268 static void blkfront_gather_backend_features(struct blkfront_info *info); 269 static int negotiate_mq(struct blkfront_info *info); 270 271 #define for_each_rinfo(info, ptr, idx) \ 272 for ((ptr) = (info)->rinfo, (idx) = 0; \ 273 (idx) < (info)->nr_rings; \ 274 (idx)++, (ptr) = (void *)(ptr) + (info)->rinfo_size) 275 276 static inline struct blkfront_ring_info * 277 get_rinfo(const struct blkfront_info *info, unsigned int i) 278 { 279 BUG_ON(i >= info->nr_rings); 280 return (void *)info->rinfo + i * info->rinfo_size; 281 } 282 283 static int get_id_from_freelist(struct blkfront_ring_info *rinfo) 284 { 285 unsigned long free = rinfo->shadow_free; 286 287 BUG_ON(free >= BLK_RING_SIZE(rinfo->dev_info)); 288 rinfo->shadow_free = rinfo->shadow[free].req.u.rw.id; 289 rinfo->shadow[free].req.u.rw.id = 0x0fffffee; /* debug */ 290 return free; 291 } 292 293 static int add_id_to_freelist(struct blkfront_ring_info *rinfo, 294 unsigned long id) 295 { 296 if (rinfo->shadow[id].req.u.rw.id != id) 297 return -EINVAL; 298 if (rinfo->shadow[id].request == NULL) 299 return -EINVAL; 300 rinfo->shadow[id].req.u.rw.id = rinfo->shadow_free; 301 rinfo->shadow[id].request = NULL; 302 rinfo->shadow_free = id; 303 return 0; 304 } 305 306 static int fill_grant_buffer(struct blkfront_ring_info *rinfo, int num) 307 { 308 struct blkfront_info *info = rinfo->dev_info; 309 struct page *granted_page; 310 struct grant *gnt_list_entry, *n; 311 int i = 0; 312 313 while (i < num) { 314 gnt_list_entry = kzalloc(sizeof(struct grant), GFP_NOIO); 315 if (!gnt_list_entry) 316 goto out_of_memory; 317 318 if (info->bounce) { 319 granted_page = alloc_page(GFP_NOIO | __GFP_ZERO); 320 if (!granted_page) { 321 kfree(gnt_list_entry); 322 goto out_of_memory; 323 } 324 gnt_list_entry->page = granted_page; 325 } 326 327 gnt_list_entry->gref = INVALID_GRANT_REF; 328 list_add(&gnt_list_entry->node, &rinfo->grants); 329 i++; 330 } 331 332 return 0; 333 334 out_of_memory: 335 list_for_each_entry_safe(gnt_list_entry, n, 336 &rinfo->grants, node) { 337 list_del(&gnt_list_entry->node); 338 if (info->bounce) 339 __free_page(gnt_list_entry->page); 340 kfree(gnt_list_entry); 341 i--; 342 } 343 BUG_ON(i != 0); 344 return -ENOMEM; 345 } 346 347 static struct grant *get_free_grant(struct blkfront_ring_info *rinfo) 348 { 349 struct grant *gnt_list_entry; 350 351 BUG_ON(list_empty(&rinfo->grants)); 352 gnt_list_entry = list_first_entry(&rinfo->grants, struct grant, 353 node); 354 list_del(&gnt_list_entry->node); 355 356 if (gnt_list_entry->gref != INVALID_GRANT_REF) 357 rinfo->persistent_gnts_c--; 358 359 return gnt_list_entry; 360 } 361 362 static inline void grant_foreign_access(const struct grant *gnt_list_entry, 363 const struct blkfront_info *info) 364 { 365 gnttab_page_grant_foreign_access_ref_one(gnt_list_entry->gref, 366 info->xbdev->otherend_id, 367 gnt_list_entry->page, 368 0); 369 } 370 371 static struct grant *get_grant(grant_ref_t *gref_head, 372 unsigned long gfn, 373 struct blkfront_ring_info *rinfo) 374 { 375 struct grant *gnt_list_entry = get_free_grant(rinfo); 376 struct blkfront_info *info = rinfo->dev_info; 377 378 if (gnt_list_entry->gref != INVALID_GRANT_REF) 379 return gnt_list_entry; 380 381 /* Assign a gref to this page */ 382 gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); 383 BUG_ON(gnt_list_entry->gref == -ENOSPC); 384 if (info->bounce) 385 grant_foreign_access(gnt_list_entry, info); 386 else { 387 /* Grant access to the GFN passed by the caller */ 388 gnttab_grant_foreign_access_ref(gnt_list_entry->gref, 389 info->xbdev->otherend_id, 390 gfn, 0); 391 } 392 393 return gnt_list_entry; 394 } 395 396 static struct grant *get_indirect_grant(grant_ref_t *gref_head, 397 struct blkfront_ring_info *rinfo) 398 { 399 struct grant *gnt_list_entry = get_free_grant(rinfo); 400 struct blkfront_info *info = rinfo->dev_info; 401 402 if (gnt_list_entry->gref != INVALID_GRANT_REF) 403 return gnt_list_entry; 404 405 /* Assign a gref to this page */ 406 gnt_list_entry->gref = gnttab_claim_grant_reference(gref_head); 407 BUG_ON(gnt_list_entry->gref == -ENOSPC); 408 if (!info->bounce) { 409 struct page *indirect_page; 410 411 /* Fetch a pre-allocated page to use for indirect grefs */ 412 BUG_ON(list_empty(&rinfo->indirect_pages)); 413 indirect_page = list_first_entry(&rinfo->indirect_pages, 414 struct page, lru); 415 list_del(&indirect_page->lru); 416 gnt_list_entry->page = indirect_page; 417 } 418 grant_foreign_access(gnt_list_entry, info); 419 420 return gnt_list_entry; 421 } 422 423 static const char *op_name(int op) 424 { 425 static const char *const names[] = { 426 [BLKIF_OP_READ] = "read", 427 [BLKIF_OP_WRITE] = "write", 428 [BLKIF_OP_WRITE_BARRIER] = "barrier", 429 [BLKIF_OP_FLUSH_DISKCACHE] = "flush", 430 [BLKIF_OP_DISCARD] = "discard" }; 431 432 if (op < 0 || op >= ARRAY_SIZE(names)) 433 return "unknown"; 434 435 if (!names[op]) 436 return "reserved"; 437 438 return names[op]; 439 } 440 static int xlbd_reserve_minors(unsigned int minor, unsigned int nr) 441 { 442 unsigned int end = minor + nr; 443 int rc; 444 445 if (end > nr_minors) { 446 unsigned long *bitmap, *old; 447 448 bitmap = kcalloc(BITS_TO_LONGS(end), sizeof(*bitmap), 449 GFP_KERNEL); 450 if (bitmap == NULL) 451 return -ENOMEM; 452 453 spin_lock(&minor_lock); 454 if (end > nr_minors) { 455 old = minors; 456 memcpy(bitmap, minors, 457 BITS_TO_LONGS(nr_minors) * sizeof(*bitmap)); 458 minors = bitmap; 459 nr_minors = BITS_TO_LONGS(end) * BITS_PER_LONG; 460 } else 461 old = bitmap; 462 spin_unlock(&minor_lock); 463 kfree(old); 464 } 465 466 spin_lock(&minor_lock); 467 if (find_next_bit(minors, end, minor) >= end) { 468 bitmap_set(minors, minor, nr); 469 rc = 0; 470 } else 471 rc = -EBUSY; 472 spin_unlock(&minor_lock); 473 474 return rc; 475 } 476 477 static void xlbd_release_minors(unsigned int minor, unsigned int nr) 478 { 479 unsigned int end = minor + nr; 480 481 BUG_ON(end > nr_minors); 482 spin_lock(&minor_lock); 483 bitmap_clear(minors, minor, nr); 484 spin_unlock(&minor_lock); 485 } 486 487 static void blkif_restart_queue_callback(void *arg) 488 { 489 struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)arg; 490 schedule_work(&rinfo->work); 491 } 492 493 static int blkif_getgeo(struct block_device *bd, struct hd_geometry *hg) 494 { 495 /* We don't have real geometry info, but let's at least return 496 values consistent with the size of the device */ 497 sector_t nsect = get_capacity(bd->bd_disk); 498 sector_t cylinders = nsect; 499 500 hg->heads = 0xff; 501 hg->sectors = 0x3f; 502 sector_div(cylinders, hg->heads * hg->sectors); 503 hg->cylinders = cylinders; 504 if ((sector_t)(hg->cylinders + 1) * hg->heads * hg->sectors < nsect) 505 hg->cylinders = 0xffff; 506 return 0; 507 } 508 509 static int blkif_ioctl(struct block_device *bdev, fmode_t mode, 510 unsigned command, unsigned long argument) 511 { 512 struct blkfront_info *info = bdev->bd_disk->private_data; 513 int i; 514 515 switch (command) { 516 case CDROMMULTISESSION: 517 for (i = 0; i < sizeof(struct cdrom_multisession); i++) 518 if (put_user(0, (char __user *)(argument + i))) 519 return -EFAULT; 520 return 0; 521 case CDROM_GET_CAPABILITY: 522 if (!(info->vdisk_info & VDISK_CDROM)) 523 return -EINVAL; 524 return 0; 525 default: 526 return -EINVAL; 527 } 528 } 529 530 static unsigned long blkif_ring_get_request(struct blkfront_ring_info *rinfo, 531 struct request *req, 532 struct blkif_request **ring_req) 533 { 534 unsigned long id; 535 536 *ring_req = RING_GET_REQUEST(&rinfo->ring, rinfo->ring.req_prod_pvt); 537 rinfo->ring.req_prod_pvt++; 538 539 id = get_id_from_freelist(rinfo); 540 rinfo->shadow[id].request = req; 541 rinfo->shadow[id].status = REQ_PROCESSING; 542 rinfo->shadow[id].associated_id = NO_ASSOCIATED_ID; 543 544 rinfo->shadow[id].req.u.rw.id = id; 545 546 return id; 547 } 548 549 static int blkif_queue_discard_req(struct request *req, struct blkfront_ring_info *rinfo) 550 { 551 struct blkfront_info *info = rinfo->dev_info; 552 struct blkif_request *ring_req, *final_ring_req; 553 unsigned long id; 554 555 /* Fill out a communications ring structure. */ 556 id = blkif_ring_get_request(rinfo, req, &final_ring_req); 557 ring_req = &rinfo->shadow[id].req; 558 559 ring_req->operation = BLKIF_OP_DISCARD; 560 ring_req->u.discard.nr_sectors = blk_rq_sectors(req); 561 ring_req->u.discard.id = id; 562 ring_req->u.discard.sector_number = (blkif_sector_t)blk_rq_pos(req); 563 if (req_op(req) == REQ_OP_SECURE_ERASE && info->feature_secdiscard) 564 ring_req->u.discard.flag = BLKIF_DISCARD_SECURE; 565 else 566 ring_req->u.discard.flag = 0; 567 568 /* Copy the request to the ring page. */ 569 *final_ring_req = *ring_req; 570 rinfo->shadow[id].status = REQ_WAITING; 571 572 return 0; 573 } 574 575 struct setup_rw_req { 576 unsigned int grant_idx; 577 struct blkif_request_segment *segments; 578 struct blkfront_ring_info *rinfo; 579 struct blkif_request *ring_req; 580 grant_ref_t gref_head; 581 unsigned int id; 582 /* Only used when persistent grant is used and it's a write request */ 583 bool need_copy; 584 unsigned int bvec_off; 585 char *bvec_data; 586 587 bool require_extra_req; 588 struct blkif_request *extra_ring_req; 589 }; 590 591 static void blkif_setup_rw_req_grant(unsigned long gfn, unsigned int offset, 592 unsigned int len, void *data) 593 { 594 struct setup_rw_req *setup = data; 595 int n, ref; 596 struct grant *gnt_list_entry; 597 unsigned int fsect, lsect; 598 /* Convenient aliases */ 599 unsigned int grant_idx = setup->grant_idx; 600 struct blkif_request *ring_req = setup->ring_req; 601 struct blkfront_ring_info *rinfo = setup->rinfo; 602 /* 603 * We always use the shadow of the first request to store the list 604 * of grant associated to the block I/O request. This made the 605 * completion more easy to handle even if the block I/O request is 606 * split. 607 */ 608 struct blk_shadow *shadow = &rinfo->shadow[setup->id]; 609 610 if (unlikely(setup->require_extra_req && 611 grant_idx >= BLKIF_MAX_SEGMENTS_PER_REQUEST)) { 612 /* 613 * We are using the second request, setup grant_idx 614 * to be the index of the segment array. 615 */ 616 grant_idx -= BLKIF_MAX_SEGMENTS_PER_REQUEST; 617 ring_req = setup->extra_ring_req; 618 } 619 620 if ((ring_req->operation == BLKIF_OP_INDIRECT) && 621 (grant_idx % GRANTS_PER_INDIRECT_FRAME == 0)) { 622 if (setup->segments) 623 kunmap_atomic(setup->segments); 624 625 n = grant_idx / GRANTS_PER_INDIRECT_FRAME; 626 gnt_list_entry = get_indirect_grant(&setup->gref_head, rinfo); 627 shadow->indirect_grants[n] = gnt_list_entry; 628 setup->segments = kmap_atomic(gnt_list_entry->page); 629 ring_req->u.indirect.indirect_grefs[n] = gnt_list_entry->gref; 630 } 631 632 gnt_list_entry = get_grant(&setup->gref_head, gfn, rinfo); 633 ref = gnt_list_entry->gref; 634 /* 635 * All the grants are stored in the shadow of the first 636 * request. Therefore we have to use the global index. 637 */ 638 shadow->grants_used[setup->grant_idx] = gnt_list_entry; 639 640 if (setup->need_copy) { 641 void *shared_data; 642 643 shared_data = kmap_atomic(gnt_list_entry->page); 644 /* 645 * this does not wipe data stored outside the 646 * range sg->offset..sg->offset+sg->length. 647 * Therefore, blkback *could* see data from 648 * previous requests. This is OK as long as 649 * persistent grants are shared with just one 650 * domain. It may need refactoring if this 651 * changes 652 */ 653 memcpy(shared_data + offset, 654 setup->bvec_data + setup->bvec_off, 655 len); 656 657 kunmap_atomic(shared_data); 658 setup->bvec_off += len; 659 } 660 661 fsect = offset >> 9; 662 lsect = fsect + (len >> 9) - 1; 663 if (ring_req->operation != BLKIF_OP_INDIRECT) { 664 ring_req->u.rw.seg[grant_idx] = 665 (struct blkif_request_segment) { 666 .gref = ref, 667 .first_sect = fsect, 668 .last_sect = lsect }; 669 } else { 670 setup->segments[grant_idx % GRANTS_PER_INDIRECT_FRAME] = 671 (struct blkif_request_segment) { 672 .gref = ref, 673 .first_sect = fsect, 674 .last_sect = lsect }; 675 } 676 677 (setup->grant_idx)++; 678 } 679 680 static void blkif_setup_extra_req(struct blkif_request *first, 681 struct blkif_request *second) 682 { 683 uint16_t nr_segments = first->u.rw.nr_segments; 684 685 /* 686 * The second request is only present when the first request uses 687 * all its segments. It's always the continuity of the first one. 688 */ 689 first->u.rw.nr_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; 690 691 second->u.rw.nr_segments = nr_segments - BLKIF_MAX_SEGMENTS_PER_REQUEST; 692 second->u.rw.sector_number = first->u.rw.sector_number + 693 (BLKIF_MAX_SEGMENTS_PER_REQUEST * XEN_PAGE_SIZE) / 512; 694 695 second->u.rw.handle = first->u.rw.handle; 696 second->operation = first->operation; 697 } 698 699 static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *rinfo) 700 { 701 struct blkfront_info *info = rinfo->dev_info; 702 struct blkif_request *ring_req, *extra_ring_req = NULL; 703 struct blkif_request *final_ring_req, *final_extra_ring_req = NULL; 704 unsigned long id, extra_id = NO_ASSOCIATED_ID; 705 bool require_extra_req = false; 706 int i; 707 struct setup_rw_req setup = { 708 .grant_idx = 0, 709 .segments = NULL, 710 .rinfo = rinfo, 711 .need_copy = rq_data_dir(req) && info->bounce, 712 }; 713 714 /* 715 * Used to store if we are able to queue the request by just using 716 * existing persistent grants, or if we have to get new grants, 717 * as there are not sufficiently many free. 718 */ 719 bool new_persistent_gnts = false; 720 struct scatterlist *sg; 721 int num_sg, max_grefs, num_grant; 722 723 max_grefs = req->nr_phys_segments * GRANTS_PER_PSEG; 724 if (max_grefs > BLKIF_MAX_SEGMENTS_PER_REQUEST) 725 /* 726 * If we are using indirect segments we need to account 727 * for the indirect grefs used in the request. 728 */ 729 max_grefs += INDIRECT_GREFS(max_grefs); 730 731 /* Check if we have enough persistent grants to allocate a requests */ 732 if (rinfo->persistent_gnts_c < max_grefs) { 733 new_persistent_gnts = true; 734 735 if (gnttab_alloc_grant_references( 736 max_grefs - rinfo->persistent_gnts_c, 737 &setup.gref_head) < 0) { 738 gnttab_request_free_callback( 739 &rinfo->callback, 740 blkif_restart_queue_callback, 741 rinfo, 742 max_grefs - rinfo->persistent_gnts_c); 743 return 1; 744 } 745 } 746 747 /* Fill out a communications ring structure. */ 748 id = blkif_ring_get_request(rinfo, req, &final_ring_req); 749 ring_req = &rinfo->shadow[id].req; 750 751 num_sg = blk_rq_map_sg(req->q, req, rinfo->shadow[id].sg); 752 num_grant = 0; 753 /* Calculate the number of grant used */ 754 for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) 755 num_grant += gnttab_count_grant(sg->offset, sg->length); 756 757 require_extra_req = info->max_indirect_segments == 0 && 758 num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST; 759 BUG_ON(!HAS_EXTRA_REQ && require_extra_req); 760 761 rinfo->shadow[id].num_sg = num_sg; 762 if (num_grant > BLKIF_MAX_SEGMENTS_PER_REQUEST && 763 likely(!require_extra_req)) { 764 /* 765 * The indirect operation can only be a BLKIF_OP_READ or 766 * BLKIF_OP_WRITE 767 */ 768 BUG_ON(req_op(req) == REQ_OP_FLUSH || req->cmd_flags & REQ_FUA); 769 ring_req->operation = BLKIF_OP_INDIRECT; 770 ring_req->u.indirect.indirect_op = rq_data_dir(req) ? 771 BLKIF_OP_WRITE : BLKIF_OP_READ; 772 ring_req->u.indirect.sector_number = (blkif_sector_t)blk_rq_pos(req); 773 ring_req->u.indirect.handle = info->handle; 774 ring_req->u.indirect.nr_segments = num_grant; 775 } else { 776 ring_req->u.rw.sector_number = (blkif_sector_t)blk_rq_pos(req); 777 ring_req->u.rw.handle = info->handle; 778 ring_req->operation = rq_data_dir(req) ? 779 BLKIF_OP_WRITE : BLKIF_OP_READ; 780 if (req_op(req) == REQ_OP_FLUSH || req->cmd_flags & REQ_FUA) { 781 /* 782 * Ideally we can do an unordered flush-to-disk. 783 * In case the backend onlysupports barriers, use that. 784 * A barrier request a superset of FUA, so we can 785 * implement it the same way. (It's also a FLUSH+FUA, 786 * since it is guaranteed ordered WRT previous writes.) 787 */ 788 if (info->feature_flush && info->feature_fua) 789 ring_req->operation = 790 BLKIF_OP_WRITE_BARRIER; 791 else if (info->feature_flush) 792 ring_req->operation = 793 BLKIF_OP_FLUSH_DISKCACHE; 794 else 795 ring_req->operation = 0; 796 } 797 ring_req->u.rw.nr_segments = num_grant; 798 if (unlikely(require_extra_req)) { 799 extra_id = blkif_ring_get_request(rinfo, req, 800 &final_extra_ring_req); 801 extra_ring_req = &rinfo->shadow[extra_id].req; 802 803 /* 804 * Only the first request contains the scatter-gather 805 * list. 806 */ 807 rinfo->shadow[extra_id].num_sg = 0; 808 809 blkif_setup_extra_req(ring_req, extra_ring_req); 810 811 /* Link the 2 requests together */ 812 rinfo->shadow[extra_id].associated_id = id; 813 rinfo->shadow[id].associated_id = extra_id; 814 } 815 } 816 817 setup.ring_req = ring_req; 818 setup.id = id; 819 820 setup.require_extra_req = require_extra_req; 821 if (unlikely(require_extra_req)) 822 setup.extra_ring_req = extra_ring_req; 823 824 for_each_sg(rinfo->shadow[id].sg, sg, num_sg, i) { 825 BUG_ON(sg->offset + sg->length > PAGE_SIZE); 826 827 if (setup.need_copy) { 828 setup.bvec_off = sg->offset; 829 setup.bvec_data = kmap_atomic(sg_page(sg)); 830 } 831 832 gnttab_foreach_grant_in_range(sg_page(sg), 833 sg->offset, 834 sg->length, 835 blkif_setup_rw_req_grant, 836 &setup); 837 838 if (setup.need_copy) 839 kunmap_atomic(setup.bvec_data); 840 } 841 if (setup.segments) 842 kunmap_atomic(setup.segments); 843 844 /* Copy request(s) to the ring page. */ 845 *final_ring_req = *ring_req; 846 rinfo->shadow[id].status = REQ_WAITING; 847 if (unlikely(require_extra_req)) { 848 *final_extra_ring_req = *extra_ring_req; 849 rinfo->shadow[extra_id].status = REQ_WAITING; 850 } 851 852 if (new_persistent_gnts) 853 gnttab_free_grant_references(setup.gref_head); 854 855 return 0; 856 } 857 858 /* 859 * Generate a Xen blkfront IO request from a blk layer request. Reads 860 * and writes are handled as expected. 861 * 862 * @req: a request struct 863 */ 864 static int blkif_queue_request(struct request *req, struct blkfront_ring_info *rinfo) 865 { 866 if (unlikely(rinfo->dev_info->connected != BLKIF_STATE_CONNECTED)) 867 return 1; 868 869 if (unlikely(req_op(req) == REQ_OP_DISCARD || 870 req_op(req) == REQ_OP_SECURE_ERASE)) 871 return blkif_queue_discard_req(req, rinfo); 872 else 873 return blkif_queue_rw_req(req, rinfo); 874 } 875 876 static inline void flush_requests(struct blkfront_ring_info *rinfo) 877 { 878 int notify; 879 880 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&rinfo->ring, notify); 881 882 if (notify) 883 notify_remote_via_irq(rinfo->irq); 884 } 885 886 static inline bool blkif_request_flush_invalid(struct request *req, 887 struct blkfront_info *info) 888 { 889 return (blk_rq_is_passthrough(req) || 890 ((req_op(req) == REQ_OP_FLUSH) && 891 !info->feature_flush) || 892 ((req->cmd_flags & REQ_FUA) && 893 !info->feature_fua)); 894 } 895 896 static blk_status_t blkif_queue_rq(struct blk_mq_hw_ctx *hctx, 897 const struct blk_mq_queue_data *qd) 898 { 899 unsigned long flags; 900 int qid = hctx->queue_num; 901 struct blkfront_info *info = hctx->queue->queuedata; 902 struct blkfront_ring_info *rinfo = NULL; 903 904 rinfo = get_rinfo(info, qid); 905 blk_mq_start_request(qd->rq); 906 spin_lock_irqsave(&rinfo->ring_lock, flags); 907 if (RING_FULL(&rinfo->ring)) 908 goto out_busy; 909 910 if (blkif_request_flush_invalid(qd->rq, rinfo->dev_info)) 911 goto out_err; 912 913 if (blkif_queue_request(qd->rq, rinfo)) 914 goto out_busy; 915 916 flush_requests(rinfo); 917 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 918 return BLK_STS_OK; 919 920 out_err: 921 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 922 return BLK_STS_IOERR; 923 924 out_busy: 925 blk_mq_stop_hw_queue(hctx); 926 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 927 return BLK_STS_DEV_RESOURCE; 928 } 929 930 static void blkif_complete_rq(struct request *rq) 931 { 932 blk_mq_end_request(rq, blkif_req(rq)->error); 933 } 934 935 static const struct blk_mq_ops blkfront_mq_ops = { 936 .queue_rq = blkif_queue_rq, 937 .complete = blkif_complete_rq, 938 }; 939 940 static void blkif_set_queue_limits(struct blkfront_info *info) 941 { 942 struct request_queue *rq = info->rq; 943 struct gendisk *gd = info->gd; 944 unsigned int segments = info->max_indirect_segments ? : 945 BLKIF_MAX_SEGMENTS_PER_REQUEST; 946 947 blk_queue_flag_set(QUEUE_FLAG_VIRT, rq); 948 949 if (info->feature_discard) { 950 blk_queue_max_discard_sectors(rq, get_capacity(gd)); 951 rq->limits.discard_granularity = info->discard_granularity ?: 952 info->physical_sector_size; 953 rq->limits.discard_alignment = info->discard_alignment; 954 if (info->feature_secdiscard) 955 blk_queue_max_secure_erase_sectors(rq, 956 get_capacity(gd)); 957 } 958 959 /* Hard sector size and max sectors impersonate the equiv. hardware. */ 960 blk_queue_logical_block_size(rq, info->sector_size); 961 blk_queue_physical_block_size(rq, info->physical_sector_size); 962 blk_queue_max_hw_sectors(rq, (segments * XEN_PAGE_SIZE) / 512); 963 964 /* Each segment in a request is up to an aligned page in size. */ 965 blk_queue_segment_boundary(rq, PAGE_SIZE - 1); 966 blk_queue_max_segment_size(rq, PAGE_SIZE); 967 968 /* Ensure a merged request will fit in a single I/O ring slot. */ 969 blk_queue_max_segments(rq, segments / GRANTS_PER_PSEG); 970 971 /* Make sure buffer addresses are sector-aligned. */ 972 blk_queue_dma_alignment(rq, 511); 973 } 974 975 static const char *flush_info(struct blkfront_info *info) 976 { 977 if (info->feature_flush && info->feature_fua) 978 return "barrier: enabled;"; 979 else if (info->feature_flush) 980 return "flush diskcache: enabled;"; 981 else 982 return "barrier or flush: disabled;"; 983 } 984 985 static void xlvbd_flush(struct blkfront_info *info) 986 { 987 blk_queue_write_cache(info->rq, info->feature_flush ? true : false, 988 info->feature_fua ? true : false); 989 pr_info("blkfront: %s: %s %s %s %s %s %s %s\n", 990 info->gd->disk_name, flush_info(info), 991 "persistent grants:", info->feature_persistent ? 992 "enabled;" : "disabled;", "indirect descriptors:", 993 info->max_indirect_segments ? "enabled;" : "disabled;", 994 "bounce buffer:", info->bounce ? "enabled" : "disabled;"); 995 } 996 997 static int xen_translate_vdev(int vdevice, int *minor, unsigned int *offset) 998 { 999 int major; 1000 major = BLKIF_MAJOR(vdevice); 1001 *minor = BLKIF_MINOR(vdevice); 1002 switch (major) { 1003 case XEN_IDE0_MAJOR: 1004 *offset = (*minor / 64) + EMULATED_HD_DISK_NAME_OFFSET; 1005 *minor = ((*minor / 64) * PARTS_PER_DISK) + 1006 EMULATED_HD_DISK_MINOR_OFFSET; 1007 break; 1008 case XEN_IDE1_MAJOR: 1009 *offset = (*minor / 64) + 2 + EMULATED_HD_DISK_NAME_OFFSET; 1010 *minor = (((*minor / 64) + 2) * PARTS_PER_DISK) + 1011 EMULATED_HD_DISK_MINOR_OFFSET; 1012 break; 1013 case XEN_SCSI_DISK0_MAJOR: 1014 *offset = (*minor / PARTS_PER_DISK) + EMULATED_SD_DISK_NAME_OFFSET; 1015 *minor = *minor + EMULATED_SD_DISK_MINOR_OFFSET; 1016 break; 1017 case XEN_SCSI_DISK1_MAJOR: 1018 case XEN_SCSI_DISK2_MAJOR: 1019 case XEN_SCSI_DISK3_MAJOR: 1020 case XEN_SCSI_DISK4_MAJOR: 1021 case XEN_SCSI_DISK5_MAJOR: 1022 case XEN_SCSI_DISK6_MAJOR: 1023 case XEN_SCSI_DISK7_MAJOR: 1024 *offset = (*minor / PARTS_PER_DISK) + 1025 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16) + 1026 EMULATED_SD_DISK_NAME_OFFSET; 1027 *minor = *minor + 1028 ((major - XEN_SCSI_DISK1_MAJOR + 1) * 16 * PARTS_PER_DISK) + 1029 EMULATED_SD_DISK_MINOR_OFFSET; 1030 break; 1031 case XEN_SCSI_DISK8_MAJOR: 1032 case XEN_SCSI_DISK9_MAJOR: 1033 case XEN_SCSI_DISK10_MAJOR: 1034 case XEN_SCSI_DISK11_MAJOR: 1035 case XEN_SCSI_DISK12_MAJOR: 1036 case XEN_SCSI_DISK13_MAJOR: 1037 case XEN_SCSI_DISK14_MAJOR: 1038 case XEN_SCSI_DISK15_MAJOR: 1039 *offset = (*minor / PARTS_PER_DISK) + 1040 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16) + 1041 EMULATED_SD_DISK_NAME_OFFSET; 1042 *minor = *minor + 1043 ((major - XEN_SCSI_DISK8_MAJOR + 8) * 16 * PARTS_PER_DISK) + 1044 EMULATED_SD_DISK_MINOR_OFFSET; 1045 break; 1046 case XENVBD_MAJOR: 1047 *offset = *minor / PARTS_PER_DISK; 1048 break; 1049 default: 1050 printk(KERN_WARNING "blkfront: your disk configuration is " 1051 "incorrect, please use an xvd device instead\n"); 1052 return -ENODEV; 1053 } 1054 return 0; 1055 } 1056 1057 static char *encode_disk_name(char *ptr, unsigned int n) 1058 { 1059 if (n >= 26) 1060 ptr = encode_disk_name(ptr, n / 26 - 1); 1061 *ptr = 'a' + n % 26; 1062 return ptr + 1; 1063 } 1064 1065 static int xlvbd_alloc_gendisk(blkif_sector_t capacity, 1066 struct blkfront_info *info, u16 sector_size, 1067 unsigned int physical_sector_size) 1068 { 1069 struct gendisk *gd; 1070 int nr_minors = 1; 1071 int err; 1072 unsigned int offset; 1073 int minor; 1074 int nr_parts; 1075 char *ptr; 1076 1077 BUG_ON(info->gd != NULL); 1078 BUG_ON(info->rq != NULL); 1079 1080 if ((info->vdevice>>EXT_SHIFT) > 1) { 1081 /* this is above the extended range; something is wrong */ 1082 printk(KERN_WARNING "blkfront: vdevice 0x%x is above the extended range; ignoring\n", info->vdevice); 1083 return -ENODEV; 1084 } 1085 1086 if (!VDEV_IS_EXTENDED(info->vdevice)) { 1087 err = xen_translate_vdev(info->vdevice, &minor, &offset); 1088 if (err) 1089 return err; 1090 nr_parts = PARTS_PER_DISK; 1091 } else { 1092 minor = BLKIF_MINOR_EXT(info->vdevice); 1093 nr_parts = PARTS_PER_EXT_DISK; 1094 offset = minor / nr_parts; 1095 if (xen_hvm_domain() && offset < EMULATED_HD_DISK_NAME_OFFSET + 4) 1096 printk(KERN_WARNING "blkfront: vdevice 0x%x might conflict with " 1097 "emulated IDE disks,\n\t choose an xvd device name" 1098 "from xvde on\n", info->vdevice); 1099 } 1100 if (minor >> MINORBITS) { 1101 pr_warn("blkfront: %#x's minor (%#x) out of range; ignoring\n", 1102 info->vdevice, minor); 1103 return -ENODEV; 1104 } 1105 1106 if ((minor % nr_parts) == 0) 1107 nr_minors = nr_parts; 1108 1109 err = xlbd_reserve_minors(minor, nr_minors); 1110 if (err) 1111 return err; 1112 1113 memset(&info->tag_set, 0, sizeof(info->tag_set)); 1114 info->tag_set.ops = &blkfront_mq_ops; 1115 info->tag_set.nr_hw_queues = info->nr_rings; 1116 if (HAS_EXTRA_REQ && info->max_indirect_segments == 0) { 1117 /* 1118 * When indirect descriptior is not supported, the I/O request 1119 * will be split between multiple request in the ring. 1120 * To avoid problems when sending the request, divide by 1121 * 2 the depth of the queue. 1122 */ 1123 info->tag_set.queue_depth = BLK_RING_SIZE(info) / 2; 1124 } else 1125 info->tag_set.queue_depth = BLK_RING_SIZE(info); 1126 info->tag_set.numa_node = NUMA_NO_NODE; 1127 info->tag_set.flags = BLK_MQ_F_SHOULD_MERGE; 1128 info->tag_set.cmd_size = sizeof(struct blkif_req); 1129 info->tag_set.driver_data = info; 1130 1131 err = blk_mq_alloc_tag_set(&info->tag_set); 1132 if (err) 1133 goto out_release_minors; 1134 1135 gd = blk_mq_alloc_disk(&info->tag_set, info); 1136 if (IS_ERR(gd)) { 1137 err = PTR_ERR(gd); 1138 goto out_free_tag_set; 1139 } 1140 1141 strcpy(gd->disk_name, DEV_NAME); 1142 ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset); 1143 BUG_ON(ptr >= gd->disk_name + DISK_NAME_LEN); 1144 if (nr_minors > 1) 1145 *ptr = 0; 1146 else 1147 snprintf(ptr, gd->disk_name + DISK_NAME_LEN - ptr, 1148 "%d", minor & (nr_parts - 1)); 1149 1150 gd->major = XENVBD_MAJOR; 1151 gd->first_minor = minor; 1152 gd->minors = nr_minors; 1153 gd->fops = &xlvbd_block_fops; 1154 gd->private_data = info; 1155 set_capacity(gd, capacity); 1156 1157 info->rq = gd->queue; 1158 info->gd = gd; 1159 info->sector_size = sector_size; 1160 info->physical_sector_size = physical_sector_size; 1161 blkif_set_queue_limits(info); 1162 1163 xlvbd_flush(info); 1164 1165 if (info->vdisk_info & VDISK_READONLY) 1166 set_disk_ro(gd, 1); 1167 if (info->vdisk_info & VDISK_REMOVABLE) 1168 gd->flags |= GENHD_FL_REMOVABLE; 1169 1170 return 0; 1171 1172 out_free_tag_set: 1173 blk_mq_free_tag_set(&info->tag_set); 1174 out_release_minors: 1175 xlbd_release_minors(minor, nr_minors); 1176 return err; 1177 } 1178 1179 /* Already hold rinfo->ring_lock. */ 1180 static inline void kick_pending_request_queues_locked(struct blkfront_ring_info *rinfo) 1181 { 1182 if (!RING_FULL(&rinfo->ring)) 1183 blk_mq_start_stopped_hw_queues(rinfo->dev_info->rq, true); 1184 } 1185 1186 static void kick_pending_request_queues(struct blkfront_ring_info *rinfo) 1187 { 1188 unsigned long flags; 1189 1190 spin_lock_irqsave(&rinfo->ring_lock, flags); 1191 kick_pending_request_queues_locked(rinfo); 1192 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 1193 } 1194 1195 static void blkif_restart_queue(struct work_struct *work) 1196 { 1197 struct blkfront_ring_info *rinfo = container_of(work, struct blkfront_ring_info, work); 1198 1199 if (rinfo->dev_info->connected == BLKIF_STATE_CONNECTED) 1200 kick_pending_request_queues(rinfo); 1201 } 1202 1203 static void blkif_free_ring(struct blkfront_ring_info *rinfo) 1204 { 1205 struct grant *persistent_gnt, *n; 1206 struct blkfront_info *info = rinfo->dev_info; 1207 int i, j, segs; 1208 1209 /* 1210 * Remove indirect pages, this only happens when using indirect 1211 * descriptors but not persistent grants 1212 */ 1213 if (!list_empty(&rinfo->indirect_pages)) { 1214 struct page *indirect_page, *n; 1215 1216 BUG_ON(info->bounce); 1217 list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { 1218 list_del(&indirect_page->lru); 1219 __free_page(indirect_page); 1220 } 1221 } 1222 1223 /* Remove all persistent grants. */ 1224 if (!list_empty(&rinfo->grants)) { 1225 list_for_each_entry_safe(persistent_gnt, n, 1226 &rinfo->grants, node) { 1227 list_del(&persistent_gnt->node); 1228 if (persistent_gnt->gref != INVALID_GRANT_REF) { 1229 gnttab_end_foreign_access(persistent_gnt->gref, 1230 NULL); 1231 rinfo->persistent_gnts_c--; 1232 } 1233 if (info->bounce) 1234 __free_page(persistent_gnt->page); 1235 kfree(persistent_gnt); 1236 } 1237 } 1238 BUG_ON(rinfo->persistent_gnts_c != 0); 1239 1240 for (i = 0; i < BLK_RING_SIZE(info); i++) { 1241 /* 1242 * Clear persistent grants present in requests already 1243 * on the shared ring 1244 */ 1245 if (!rinfo->shadow[i].request) 1246 goto free_shadow; 1247 1248 segs = rinfo->shadow[i].req.operation == BLKIF_OP_INDIRECT ? 1249 rinfo->shadow[i].req.u.indirect.nr_segments : 1250 rinfo->shadow[i].req.u.rw.nr_segments; 1251 for (j = 0; j < segs; j++) { 1252 persistent_gnt = rinfo->shadow[i].grants_used[j]; 1253 gnttab_end_foreign_access(persistent_gnt->gref, NULL); 1254 if (info->bounce) 1255 __free_page(persistent_gnt->page); 1256 kfree(persistent_gnt); 1257 } 1258 1259 if (rinfo->shadow[i].req.operation != BLKIF_OP_INDIRECT) 1260 /* 1261 * If this is not an indirect operation don't try to 1262 * free indirect segments 1263 */ 1264 goto free_shadow; 1265 1266 for (j = 0; j < INDIRECT_GREFS(segs); j++) { 1267 persistent_gnt = rinfo->shadow[i].indirect_grants[j]; 1268 gnttab_end_foreign_access(persistent_gnt->gref, NULL); 1269 __free_page(persistent_gnt->page); 1270 kfree(persistent_gnt); 1271 } 1272 1273 free_shadow: 1274 kvfree(rinfo->shadow[i].grants_used); 1275 rinfo->shadow[i].grants_used = NULL; 1276 kvfree(rinfo->shadow[i].indirect_grants); 1277 rinfo->shadow[i].indirect_grants = NULL; 1278 kvfree(rinfo->shadow[i].sg); 1279 rinfo->shadow[i].sg = NULL; 1280 } 1281 1282 /* No more gnttab callback work. */ 1283 gnttab_cancel_free_callback(&rinfo->callback); 1284 1285 /* Flush gnttab callback work. Must be done with no locks held. */ 1286 flush_work(&rinfo->work); 1287 1288 /* Free resources associated with old device channel. */ 1289 xenbus_teardown_ring((void **)&rinfo->ring.sring, info->nr_ring_pages, 1290 rinfo->ring_ref); 1291 1292 if (rinfo->irq) 1293 unbind_from_irqhandler(rinfo->irq, rinfo); 1294 rinfo->evtchn = rinfo->irq = 0; 1295 } 1296 1297 static void blkif_free(struct blkfront_info *info, int suspend) 1298 { 1299 unsigned int i; 1300 struct blkfront_ring_info *rinfo; 1301 1302 /* Prevent new requests being issued until we fix things up. */ 1303 info->connected = suspend ? 1304 BLKIF_STATE_SUSPENDED : BLKIF_STATE_DISCONNECTED; 1305 /* No more blkif_request(). */ 1306 if (info->rq) 1307 blk_mq_stop_hw_queues(info->rq); 1308 1309 for_each_rinfo(info, rinfo, i) 1310 blkif_free_ring(rinfo); 1311 1312 kvfree(info->rinfo); 1313 info->rinfo = NULL; 1314 info->nr_rings = 0; 1315 } 1316 1317 struct copy_from_grant { 1318 const struct blk_shadow *s; 1319 unsigned int grant_idx; 1320 unsigned int bvec_offset; 1321 char *bvec_data; 1322 }; 1323 1324 static void blkif_copy_from_grant(unsigned long gfn, unsigned int offset, 1325 unsigned int len, void *data) 1326 { 1327 struct copy_from_grant *info = data; 1328 char *shared_data; 1329 /* Convenient aliases */ 1330 const struct blk_shadow *s = info->s; 1331 1332 shared_data = kmap_atomic(s->grants_used[info->grant_idx]->page); 1333 1334 memcpy(info->bvec_data + info->bvec_offset, 1335 shared_data + offset, len); 1336 1337 info->bvec_offset += len; 1338 info->grant_idx++; 1339 1340 kunmap_atomic(shared_data); 1341 } 1342 1343 static enum blk_req_status blkif_rsp_to_req_status(int rsp) 1344 { 1345 switch (rsp) 1346 { 1347 case BLKIF_RSP_OKAY: 1348 return REQ_DONE; 1349 case BLKIF_RSP_EOPNOTSUPP: 1350 return REQ_EOPNOTSUPP; 1351 case BLKIF_RSP_ERROR: 1352 default: 1353 return REQ_ERROR; 1354 } 1355 } 1356 1357 /* 1358 * Get the final status of the block request based on two ring response 1359 */ 1360 static int blkif_get_final_status(enum blk_req_status s1, 1361 enum blk_req_status s2) 1362 { 1363 BUG_ON(s1 < REQ_DONE); 1364 BUG_ON(s2 < REQ_DONE); 1365 1366 if (s1 == REQ_ERROR || s2 == REQ_ERROR) 1367 return BLKIF_RSP_ERROR; 1368 else if (s1 == REQ_EOPNOTSUPP || s2 == REQ_EOPNOTSUPP) 1369 return BLKIF_RSP_EOPNOTSUPP; 1370 return BLKIF_RSP_OKAY; 1371 } 1372 1373 /* 1374 * Return values: 1375 * 1 response processed. 1376 * 0 missing further responses. 1377 * -1 error while processing. 1378 */ 1379 static int blkif_completion(unsigned long *id, 1380 struct blkfront_ring_info *rinfo, 1381 struct blkif_response *bret) 1382 { 1383 int i = 0; 1384 struct scatterlist *sg; 1385 int num_sg, num_grant; 1386 struct blkfront_info *info = rinfo->dev_info; 1387 struct blk_shadow *s = &rinfo->shadow[*id]; 1388 struct copy_from_grant data = { 1389 .grant_idx = 0, 1390 }; 1391 1392 num_grant = s->req.operation == BLKIF_OP_INDIRECT ? 1393 s->req.u.indirect.nr_segments : s->req.u.rw.nr_segments; 1394 1395 /* The I/O request may be split in two. */ 1396 if (unlikely(s->associated_id != NO_ASSOCIATED_ID)) { 1397 struct blk_shadow *s2 = &rinfo->shadow[s->associated_id]; 1398 1399 /* Keep the status of the current response in shadow. */ 1400 s->status = blkif_rsp_to_req_status(bret->status); 1401 1402 /* Wait the second response if not yet here. */ 1403 if (s2->status < REQ_DONE) 1404 return 0; 1405 1406 bret->status = blkif_get_final_status(s->status, 1407 s2->status); 1408 1409 /* 1410 * All the grants is stored in the first shadow in order 1411 * to make the completion code simpler. 1412 */ 1413 num_grant += s2->req.u.rw.nr_segments; 1414 1415 /* 1416 * The two responses may not come in order. Only the 1417 * first request will store the scatter-gather list. 1418 */ 1419 if (s2->num_sg != 0) { 1420 /* Update "id" with the ID of the first response. */ 1421 *id = s->associated_id; 1422 s = s2; 1423 } 1424 1425 /* 1426 * We don't need anymore the second request, so recycling 1427 * it now. 1428 */ 1429 if (add_id_to_freelist(rinfo, s->associated_id)) 1430 WARN(1, "%s: can't recycle the second part (id = %ld) of the request\n", 1431 info->gd->disk_name, s->associated_id); 1432 } 1433 1434 data.s = s; 1435 num_sg = s->num_sg; 1436 1437 if (bret->operation == BLKIF_OP_READ && info->bounce) { 1438 for_each_sg(s->sg, sg, num_sg, i) { 1439 BUG_ON(sg->offset + sg->length > PAGE_SIZE); 1440 1441 data.bvec_offset = sg->offset; 1442 data.bvec_data = kmap_atomic(sg_page(sg)); 1443 1444 gnttab_foreach_grant_in_range(sg_page(sg), 1445 sg->offset, 1446 sg->length, 1447 blkif_copy_from_grant, 1448 &data); 1449 1450 kunmap_atomic(data.bvec_data); 1451 } 1452 } 1453 /* Add the persistent grant into the list of free grants */ 1454 for (i = 0; i < num_grant; i++) { 1455 if (!gnttab_try_end_foreign_access(s->grants_used[i]->gref)) { 1456 /* 1457 * If the grant is still mapped by the backend (the 1458 * backend has chosen to make this grant persistent) 1459 * we add it at the head of the list, so it will be 1460 * reused first. 1461 */ 1462 if (!info->feature_persistent) { 1463 pr_alert("backed has not unmapped grant: %u\n", 1464 s->grants_used[i]->gref); 1465 return -1; 1466 } 1467 list_add(&s->grants_used[i]->node, &rinfo->grants); 1468 rinfo->persistent_gnts_c++; 1469 } else { 1470 /* 1471 * If the grant is not mapped by the backend we add it 1472 * to the tail of the list, so it will not be picked 1473 * again unless we run out of persistent grants. 1474 */ 1475 s->grants_used[i]->gref = INVALID_GRANT_REF; 1476 list_add_tail(&s->grants_used[i]->node, &rinfo->grants); 1477 } 1478 } 1479 if (s->req.operation == BLKIF_OP_INDIRECT) { 1480 for (i = 0; i < INDIRECT_GREFS(num_grant); i++) { 1481 if (!gnttab_try_end_foreign_access(s->indirect_grants[i]->gref)) { 1482 if (!info->feature_persistent) { 1483 pr_alert("backed has not unmapped grant: %u\n", 1484 s->indirect_grants[i]->gref); 1485 return -1; 1486 } 1487 list_add(&s->indirect_grants[i]->node, &rinfo->grants); 1488 rinfo->persistent_gnts_c++; 1489 } else { 1490 struct page *indirect_page; 1491 1492 /* 1493 * Add the used indirect page back to the list of 1494 * available pages for indirect grefs. 1495 */ 1496 if (!info->bounce) { 1497 indirect_page = s->indirect_grants[i]->page; 1498 list_add(&indirect_page->lru, &rinfo->indirect_pages); 1499 } 1500 s->indirect_grants[i]->gref = INVALID_GRANT_REF; 1501 list_add_tail(&s->indirect_grants[i]->node, &rinfo->grants); 1502 } 1503 } 1504 } 1505 1506 return 1; 1507 } 1508 1509 static irqreturn_t blkif_interrupt(int irq, void *dev_id) 1510 { 1511 struct request *req; 1512 struct blkif_response bret; 1513 RING_IDX i, rp; 1514 unsigned long flags; 1515 struct blkfront_ring_info *rinfo = (struct blkfront_ring_info *)dev_id; 1516 struct blkfront_info *info = rinfo->dev_info; 1517 unsigned int eoiflag = XEN_EOI_FLAG_SPURIOUS; 1518 1519 if (unlikely(info->connected != BLKIF_STATE_CONNECTED)) { 1520 xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); 1521 return IRQ_HANDLED; 1522 } 1523 1524 spin_lock_irqsave(&rinfo->ring_lock, flags); 1525 again: 1526 rp = READ_ONCE(rinfo->ring.sring->rsp_prod); 1527 virt_rmb(); /* Ensure we see queued responses up to 'rp'. */ 1528 if (RING_RESPONSE_PROD_OVERFLOW(&rinfo->ring, rp)) { 1529 pr_alert("%s: illegal number of responses %u\n", 1530 info->gd->disk_name, rp - rinfo->ring.rsp_cons); 1531 goto err; 1532 } 1533 1534 for (i = rinfo->ring.rsp_cons; i != rp; i++) { 1535 unsigned long id; 1536 unsigned int op; 1537 1538 eoiflag = 0; 1539 1540 RING_COPY_RESPONSE(&rinfo->ring, i, &bret); 1541 id = bret.id; 1542 1543 /* 1544 * The backend has messed up and given us an id that we would 1545 * never have given to it (we stamp it up to BLK_RING_SIZE - 1546 * look in get_id_from_freelist. 1547 */ 1548 if (id >= BLK_RING_SIZE(info)) { 1549 pr_alert("%s: response has incorrect id (%ld)\n", 1550 info->gd->disk_name, id); 1551 goto err; 1552 } 1553 if (rinfo->shadow[id].status != REQ_WAITING) { 1554 pr_alert("%s: response references no pending request\n", 1555 info->gd->disk_name); 1556 goto err; 1557 } 1558 1559 rinfo->shadow[id].status = REQ_PROCESSING; 1560 req = rinfo->shadow[id].request; 1561 1562 op = rinfo->shadow[id].req.operation; 1563 if (op == BLKIF_OP_INDIRECT) 1564 op = rinfo->shadow[id].req.u.indirect.indirect_op; 1565 if (bret.operation != op) { 1566 pr_alert("%s: response has wrong operation (%u instead of %u)\n", 1567 info->gd->disk_name, bret.operation, op); 1568 goto err; 1569 } 1570 1571 if (bret.operation != BLKIF_OP_DISCARD) { 1572 int ret; 1573 1574 /* 1575 * We may need to wait for an extra response if the 1576 * I/O request is split in 2 1577 */ 1578 ret = blkif_completion(&id, rinfo, &bret); 1579 if (!ret) 1580 continue; 1581 if (unlikely(ret < 0)) 1582 goto err; 1583 } 1584 1585 if (add_id_to_freelist(rinfo, id)) { 1586 WARN(1, "%s: response to %s (id %ld) couldn't be recycled!\n", 1587 info->gd->disk_name, op_name(bret.operation), id); 1588 continue; 1589 } 1590 1591 if (bret.status == BLKIF_RSP_OKAY) 1592 blkif_req(req)->error = BLK_STS_OK; 1593 else 1594 blkif_req(req)->error = BLK_STS_IOERR; 1595 1596 switch (bret.operation) { 1597 case BLKIF_OP_DISCARD: 1598 if (unlikely(bret.status == BLKIF_RSP_EOPNOTSUPP)) { 1599 struct request_queue *rq = info->rq; 1600 1601 pr_warn_ratelimited("blkfront: %s: %s op failed\n", 1602 info->gd->disk_name, op_name(bret.operation)); 1603 blkif_req(req)->error = BLK_STS_NOTSUPP; 1604 info->feature_discard = 0; 1605 info->feature_secdiscard = 0; 1606 blk_queue_max_discard_sectors(rq, 0); 1607 blk_queue_max_secure_erase_sectors(rq, 0); 1608 } 1609 break; 1610 case BLKIF_OP_FLUSH_DISKCACHE: 1611 case BLKIF_OP_WRITE_BARRIER: 1612 if (unlikely(bret.status == BLKIF_RSP_EOPNOTSUPP)) { 1613 pr_warn_ratelimited("blkfront: %s: %s op failed\n", 1614 info->gd->disk_name, op_name(bret.operation)); 1615 blkif_req(req)->error = BLK_STS_NOTSUPP; 1616 } 1617 if (unlikely(bret.status == BLKIF_RSP_ERROR && 1618 rinfo->shadow[id].req.u.rw.nr_segments == 0)) { 1619 pr_warn_ratelimited("blkfront: %s: empty %s op failed\n", 1620 info->gd->disk_name, op_name(bret.operation)); 1621 blkif_req(req)->error = BLK_STS_NOTSUPP; 1622 } 1623 if (unlikely(blkif_req(req)->error)) { 1624 if (blkif_req(req)->error == BLK_STS_NOTSUPP) 1625 blkif_req(req)->error = BLK_STS_OK; 1626 info->feature_fua = 0; 1627 info->feature_flush = 0; 1628 xlvbd_flush(info); 1629 } 1630 fallthrough; 1631 case BLKIF_OP_READ: 1632 case BLKIF_OP_WRITE: 1633 if (unlikely(bret.status != BLKIF_RSP_OKAY)) 1634 dev_dbg_ratelimited(&info->xbdev->dev, 1635 "Bad return from blkdev data request: %#x\n", 1636 bret.status); 1637 1638 break; 1639 default: 1640 BUG(); 1641 } 1642 1643 if (likely(!blk_should_fake_timeout(req->q))) 1644 blk_mq_complete_request(req); 1645 } 1646 1647 rinfo->ring.rsp_cons = i; 1648 1649 if (i != rinfo->ring.req_prod_pvt) { 1650 int more_to_do; 1651 RING_FINAL_CHECK_FOR_RESPONSES(&rinfo->ring, more_to_do); 1652 if (more_to_do) 1653 goto again; 1654 } else 1655 rinfo->ring.sring->rsp_event = i + 1; 1656 1657 kick_pending_request_queues_locked(rinfo); 1658 1659 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 1660 1661 xen_irq_lateeoi(irq, eoiflag); 1662 1663 return IRQ_HANDLED; 1664 1665 err: 1666 info->connected = BLKIF_STATE_ERROR; 1667 1668 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 1669 1670 /* No EOI in order to avoid further interrupts. */ 1671 1672 pr_alert("%s disabled for further use\n", info->gd->disk_name); 1673 return IRQ_HANDLED; 1674 } 1675 1676 1677 static int setup_blkring(struct xenbus_device *dev, 1678 struct blkfront_ring_info *rinfo) 1679 { 1680 struct blkif_sring *sring; 1681 int err; 1682 struct blkfront_info *info = rinfo->dev_info; 1683 unsigned long ring_size = info->nr_ring_pages * XEN_PAGE_SIZE; 1684 1685 err = xenbus_setup_ring(dev, GFP_NOIO, (void **)&sring, 1686 info->nr_ring_pages, rinfo->ring_ref); 1687 if (err) 1688 goto fail; 1689 1690 XEN_FRONT_RING_INIT(&rinfo->ring, sring, ring_size); 1691 1692 err = xenbus_alloc_evtchn(dev, &rinfo->evtchn); 1693 if (err) 1694 goto fail; 1695 1696 err = bind_evtchn_to_irqhandler_lateeoi(rinfo->evtchn, blkif_interrupt, 1697 0, "blkif", rinfo); 1698 if (err <= 0) { 1699 xenbus_dev_fatal(dev, err, 1700 "bind_evtchn_to_irqhandler failed"); 1701 goto fail; 1702 } 1703 rinfo->irq = err; 1704 1705 return 0; 1706 fail: 1707 blkif_free(info, 0); 1708 return err; 1709 } 1710 1711 /* 1712 * Write out per-ring/queue nodes including ring-ref and event-channel, and each 1713 * ring buffer may have multi pages depending on ->nr_ring_pages. 1714 */ 1715 static int write_per_ring_nodes(struct xenbus_transaction xbt, 1716 struct blkfront_ring_info *rinfo, const char *dir) 1717 { 1718 int err; 1719 unsigned int i; 1720 const char *message = NULL; 1721 struct blkfront_info *info = rinfo->dev_info; 1722 1723 if (info->nr_ring_pages == 1) { 1724 err = xenbus_printf(xbt, dir, "ring-ref", "%u", rinfo->ring_ref[0]); 1725 if (err) { 1726 message = "writing ring-ref"; 1727 goto abort_transaction; 1728 } 1729 } else { 1730 for (i = 0; i < info->nr_ring_pages; i++) { 1731 char ring_ref_name[RINGREF_NAME_LEN]; 1732 1733 snprintf(ring_ref_name, RINGREF_NAME_LEN, "ring-ref%u", i); 1734 err = xenbus_printf(xbt, dir, ring_ref_name, 1735 "%u", rinfo->ring_ref[i]); 1736 if (err) { 1737 message = "writing ring-ref"; 1738 goto abort_transaction; 1739 } 1740 } 1741 } 1742 1743 err = xenbus_printf(xbt, dir, "event-channel", "%u", rinfo->evtchn); 1744 if (err) { 1745 message = "writing event-channel"; 1746 goto abort_transaction; 1747 } 1748 1749 return 0; 1750 1751 abort_transaction: 1752 xenbus_transaction_end(xbt, 1); 1753 if (message) 1754 xenbus_dev_fatal(info->xbdev, err, "%s", message); 1755 1756 return err; 1757 } 1758 1759 /* Common code used when first setting up, and when resuming. */ 1760 static int talk_to_blkback(struct xenbus_device *dev, 1761 struct blkfront_info *info) 1762 { 1763 const char *message = NULL; 1764 struct xenbus_transaction xbt; 1765 int err; 1766 unsigned int i, max_page_order; 1767 unsigned int ring_page_order; 1768 struct blkfront_ring_info *rinfo; 1769 1770 if (!info) 1771 return -ENODEV; 1772 1773 /* Check if backend is trusted. */ 1774 info->bounce = !xen_blkif_trusted || 1775 !xenbus_read_unsigned(dev->nodename, "trusted", 1); 1776 1777 max_page_order = xenbus_read_unsigned(info->xbdev->otherend, 1778 "max-ring-page-order", 0); 1779 ring_page_order = min(xen_blkif_max_ring_order, max_page_order); 1780 info->nr_ring_pages = 1 << ring_page_order; 1781 1782 err = negotiate_mq(info); 1783 if (err) 1784 goto destroy_blkring; 1785 1786 for_each_rinfo(info, rinfo, i) { 1787 /* Create shared ring, alloc event channel. */ 1788 err = setup_blkring(dev, rinfo); 1789 if (err) 1790 goto destroy_blkring; 1791 } 1792 1793 again: 1794 err = xenbus_transaction_start(&xbt); 1795 if (err) { 1796 xenbus_dev_fatal(dev, err, "starting transaction"); 1797 goto destroy_blkring; 1798 } 1799 1800 if (info->nr_ring_pages > 1) { 1801 err = xenbus_printf(xbt, dev->nodename, "ring-page-order", "%u", 1802 ring_page_order); 1803 if (err) { 1804 message = "writing ring-page-order"; 1805 goto abort_transaction; 1806 } 1807 } 1808 1809 /* We already got the number of queues/rings in _probe */ 1810 if (info->nr_rings == 1) { 1811 err = write_per_ring_nodes(xbt, info->rinfo, dev->nodename); 1812 if (err) 1813 goto destroy_blkring; 1814 } else { 1815 char *path; 1816 size_t pathsize; 1817 1818 err = xenbus_printf(xbt, dev->nodename, "multi-queue-num-queues", "%u", 1819 info->nr_rings); 1820 if (err) { 1821 message = "writing multi-queue-num-queues"; 1822 goto abort_transaction; 1823 } 1824 1825 pathsize = strlen(dev->nodename) + QUEUE_NAME_LEN; 1826 path = kmalloc(pathsize, GFP_KERNEL); 1827 if (!path) { 1828 err = -ENOMEM; 1829 message = "ENOMEM while writing ring references"; 1830 goto abort_transaction; 1831 } 1832 1833 for_each_rinfo(info, rinfo, i) { 1834 memset(path, 0, pathsize); 1835 snprintf(path, pathsize, "%s/queue-%u", dev->nodename, i); 1836 err = write_per_ring_nodes(xbt, rinfo, path); 1837 if (err) { 1838 kfree(path); 1839 goto destroy_blkring; 1840 } 1841 } 1842 kfree(path); 1843 } 1844 err = xenbus_printf(xbt, dev->nodename, "protocol", "%s", 1845 XEN_IO_PROTO_ABI_NATIVE); 1846 if (err) { 1847 message = "writing protocol"; 1848 goto abort_transaction; 1849 } 1850 err = xenbus_printf(xbt, dev->nodename, "feature-persistent", "%u", 1851 info->feature_persistent); 1852 if (err) 1853 dev_warn(&dev->dev, 1854 "writing persistent grants feature to xenbus"); 1855 1856 err = xenbus_transaction_end(xbt, 0); 1857 if (err) { 1858 if (err == -EAGAIN) 1859 goto again; 1860 xenbus_dev_fatal(dev, err, "completing transaction"); 1861 goto destroy_blkring; 1862 } 1863 1864 for_each_rinfo(info, rinfo, i) { 1865 unsigned int j; 1866 1867 for (j = 0; j < BLK_RING_SIZE(info); j++) 1868 rinfo->shadow[j].req.u.rw.id = j + 1; 1869 rinfo->shadow[BLK_RING_SIZE(info)-1].req.u.rw.id = 0x0fffffff; 1870 } 1871 xenbus_switch_state(dev, XenbusStateInitialised); 1872 1873 return 0; 1874 1875 abort_transaction: 1876 xenbus_transaction_end(xbt, 1); 1877 if (message) 1878 xenbus_dev_fatal(dev, err, "%s", message); 1879 destroy_blkring: 1880 blkif_free(info, 0); 1881 return err; 1882 } 1883 1884 static int negotiate_mq(struct blkfront_info *info) 1885 { 1886 unsigned int backend_max_queues; 1887 unsigned int i; 1888 struct blkfront_ring_info *rinfo; 1889 1890 BUG_ON(info->nr_rings); 1891 1892 /* Check if backend supports multiple queues. */ 1893 backend_max_queues = xenbus_read_unsigned(info->xbdev->otherend, 1894 "multi-queue-max-queues", 1); 1895 info->nr_rings = min(backend_max_queues, xen_blkif_max_queues); 1896 /* We need at least one ring. */ 1897 if (!info->nr_rings) 1898 info->nr_rings = 1; 1899 1900 info->rinfo_size = struct_size(info->rinfo, shadow, 1901 BLK_RING_SIZE(info)); 1902 info->rinfo = kvcalloc(info->nr_rings, info->rinfo_size, GFP_KERNEL); 1903 if (!info->rinfo) { 1904 xenbus_dev_fatal(info->xbdev, -ENOMEM, "allocating ring_info structure"); 1905 info->nr_rings = 0; 1906 return -ENOMEM; 1907 } 1908 1909 for_each_rinfo(info, rinfo, i) { 1910 INIT_LIST_HEAD(&rinfo->indirect_pages); 1911 INIT_LIST_HEAD(&rinfo->grants); 1912 rinfo->dev_info = info; 1913 INIT_WORK(&rinfo->work, blkif_restart_queue); 1914 spin_lock_init(&rinfo->ring_lock); 1915 } 1916 return 0; 1917 } 1918 1919 /* Enable the persistent grants feature. */ 1920 static bool feature_persistent = true; 1921 module_param(feature_persistent, bool, 0644); 1922 MODULE_PARM_DESC(feature_persistent, 1923 "Enables the persistent grants feature"); 1924 1925 /* 1926 * Entry point to this code when a new device is created. Allocate the basic 1927 * structures and the ring buffer for communication with the backend, and 1928 * inform the backend of the appropriate details for those. Switch to 1929 * Initialised state. 1930 */ 1931 static int blkfront_probe(struct xenbus_device *dev, 1932 const struct xenbus_device_id *id) 1933 { 1934 int err, vdevice; 1935 struct blkfront_info *info; 1936 1937 /* FIXME: Use dynamic device id if this is not set. */ 1938 err = xenbus_scanf(XBT_NIL, dev->nodename, 1939 "virtual-device", "%i", &vdevice); 1940 if (err != 1) { 1941 /* go looking in the extended area instead */ 1942 err = xenbus_scanf(XBT_NIL, dev->nodename, "virtual-device-ext", 1943 "%i", &vdevice); 1944 if (err != 1) { 1945 xenbus_dev_fatal(dev, err, "reading virtual-device"); 1946 return err; 1947 } 1948 } 1949 1950 if (xen_hvm_domain()) { 1951 char *type; 1952 int len; 1953 /* no unplug has been done: do not hook devices != xen vbds */ 1954 if (xen_has_pv_and_legacy_disk_devices()) { 1955 int major; 1956 1957 if (!VDEV_IS_EXTENDED(vdevice)) 1958 major = BLKIF_MAJOR(vdevice); 1959 else 1960 major = XENVBD_MAJOR; 1961 1962 if (major != XENVBD_MAJOR) { 1963 printk(KERN_INFO 1964 "%s: HVM does not support vbd %d as xen block device\n", 1965 __func__, vdevice); 1966 return -ENODEV; 1967 } 1968 } 1969 /* do not create a PV cdrom device if we are an HVM guest */ 1970 type = xenbus_read(XBT_NIL, dev->nodename, "device-type", &len); 1971 if (IS_ERR(type)) 1972 return -ENODEV; 1973 if (strncmp(type, "cdrom", 5) == 0) { 1974 kfree(type); 1975 return -ENODEV; 1976 } 1977 kfree(type); 1978 } 1979 info = kzalloc(sizeof(*info), GFP_KERNEL); 1980 if (!info) { 1981 xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure"); 1982 return -ENOMEM; 1983 } 1984 1985 info->xbdev = dev; 1986 1987 mutex_init(&info->mutex); 1988 info->vdevice = vdevice; 1989 info->connected = BLKIF_STATE_DISCONNECTED; 1990 1991 info->feature_persistent = feature_persistent; 1992 1993 /* Front end dir is a number, which is used as the id. */ 1994 info->handle = simple_strtoul(strrchr(dev->nodename, '/')+1, NULL, 0); 1995 dev_set_drvdata(&dev->dev, info); 1996 1997 mutex_lock(&blkfront_mutex); 1998 list_add(&info->info_list, &info_list); 1999 mutex_unlock(&blkfront_mutex); 2000 2001 return 0; 2002 } 2003 2004 static int blkif_recover(struct blkfront_info *info) 2005 { 2006 unsigned int r_index; 2007 struct request *req, *n; 2008 int rc; 2009 struct bio *bio; 2010 unsigned int segs; 2011 struct blkfront_ring_info *rinfo; 2012 2013 blkfront_gather_backend_features(info); 2014 /* Reset limits changed by blk_mq_update_nr_hw_queues(). */ 2015 blkif_set_queue_limits(info); 2016 segs = info->max_indirect_segments ? : BLKIF_MAX_SEGMENTS_PER_REQUEST; 2017 blk_queue_max_segments(info->rq, segs / GRANTS_PER_PSEG); 2018 2019 for_each_rinfo(info, rinfo, r_index) { 2020 rc = blkfront_setup_indirect(rinfo); 2021 if (rc) 2022 return rc; 2023 } 2024 xenbus_switch_state(info->xbdev, XenbusStateConnected); 2025 2026 /* Now safe for us to use the shared ring */ 2027 info->connected = BLKIF_STATE_CONNECTED; 2028 2029 for_each_rinfo(info, rinfo, r_index) { 2030 /* Kick any other new requests queued since we resumed */ 2031 kick_pending_request_queues(rinfo); 2032 } 2033 2034 list_for_each_entry_safe(req, n, &info->requests, queuelist) { 2035 /* Requeue pending requests (flush or discard) */ 2036 list_del_init(&req->queuelist); 2037 BUG_ON(req->nr_phys_segments > segs); 2038 blk_mq_requeue_request(req, false); 2039 } 2040 blk_mq_start_stopped_hw_queues(info->rq, true); 2041 blk_mq_kick_requeue_list(info->rq); 2042 2043 while ((bio = bio_list_pop(&info->bio_list)) != NULL) { 2044 /* Traverse the list of pending bios and re-queue them */ 2045 submit_bio(bio); 2046 } 2047 2048 return 0; 2049 } 2050 2051 /* 2052 * We are reconnecting to the backend, due to a suspend/resume, or a backend 2053 * driver restart. We tear down our blkif structure and recreate it, but 2054 * leave the device-layer structures intact so that this is transparent to the 2055 * rest of the kernel. 2056 */ 2057 static int blkfront_resume(struct xenbus_device *dev) 2058 { 2059 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 2060 int err = 0; 2061 unsigned int i, j; 2062 struct blkfront_ring_info *rinfo; 2063 2064 dev_dbg(&dev->dev, "blkfront_resume: %s\n", dev->nodename); 2065 2066 bio_list_init(&info->bio_list); 2067 INIT_LIST_HEAD(&info->requests); 2068 for_each_rinfo(info, rinfo, i) { 2069 struct bio_list merge_bio; 2070 struct blk_shadow *shadow = rinfo->shadow; 2071 2072 for (j = 0; j < BLK_RING_SIZE(info); j++) { 2073 /* Not in use? */ 2074 if (!shadow[j].request) 2075 continue; 2076 2077 /* 2078 * Get the bios in the request so we can re-queue them. 2079 */ 2080 if (req_op(shadow[j].request) == REQ_OP_FLUSH || 2081 req_op(shadow[j].request) == REQ_OP_DISCARD || 2082 req_op(shadow[j].request) == REQ_OP_SECURE_ERASE || 2083 shadow[j].request->cmd_flags & REQ_FUA) { 2084 /* 2085 * Flush operations don't contain bios, so 2086 * we need to requeue the whole request 2087 * 2088 * XXX: but this doesn't make any sense for a 2089 * write with the FUA flag set.. 2090 */ 2091 list_add(&shadow[j].request->queuelist, &info->requests); 2092 continue; 2093 } 2094 merge_bio.head = shadow[j].request->bio; 2095 merge_bio.tail = shadow[j].request->biotail; 2096 bio_list_merge(&info->bio_list, &merge_bio); 2097 shadow[j].request->bio = NULL; 2098 blk_mq_end_request(shadow[j].request, BLK_STS_OK); 2099 } 2100 } 2101 2102 blkif_free(info, info->connected == BLKIF_STATE_CONNECTED); 2103 2104 err = talk_to_blkback(dev, info); 2105 if (!err) 2106 blk_mq_update_nr_hw_queues(&info->tag_set, info->nr_rings); 2107 2108 /* 2109 * We have to wait for the backend to switch to 2110 * connected state, since we want to read which 2111 * features it supports. 2112 */ 2113 2114 return err; 2115 } 2116 2117 static void blkfront_closing(struct blkfront_info *info) 2118 { 2119 struct xenbus_device *xbdev = info->xbdev; 2120 struct blkfront_ring_info *rinfo; 2121 unsigned int i; 2122 2123 if (xbdev->state == XenbusStateClosing) 2124 return; 2125 2126 /* No more blkif_request(). */ 2127 if (info->rq && info->gd) { 2128 blk_mq_stop_hw_queues(info->rq); 2129 blk_mark_disk_dead(info->gd); 2130 set_capacity(info->gd, 0); 2131 } 2132 2133 for_each_rinfo(info, rinfo, i) { 2134 /* No more gnttab callback work. */ 2135 gnttab_cancel_free_callback(&rinfo->callback); 2136 2137 /* Flush gnttab callback work. Must be done with no locks held. */ 2138 flush_work(&rinfo->work); 2139 } 2140 2141 xenbus_frontend_closed(xbdev); 2142 } 2143 2144 static void blkfront_setup_discard(struct blkfront_info *info) 2145 { 2146 info->feature_discard = 1; 2147 info->discard_granularity = xenbus_read_unsigned(info->xbdev->otherend, 2148 "discard-granularity", 2149 0); 2150 info->discard_alignment = xenbus_read_unsigned(info->xbdev->otherend, 2151 "discard-alignment", 0); 2152 info->feature_secdiscard = 2153 !!xenbus_read_unsigned(info->xbdev->otherend, "discard-secure", 2154 0); 2155 } 2156 2157 static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo) 2158 { 2159 unsigned int psegs, grants, memflags; 2160 int err, i; 2161 struct blkfront_info *info = rinfo->dev_info; 2162 2163 memflags = memalloc_noio_save(); 2164 2165 if (info->max_indirect_segments == 0) { 2166 if (!HAS_EXTRA_REQ) 2167 grants = BLKIF_MAX_SEGMENTS_PER_REQUEST; 2168 else { 2169 /* 2170 * When an extra req is required, the maximum 2171 * grants supported is related to the size of the 2172 * Linux block segment. 2173 */ 2174 grants = GRANTS_PER_PSEG; 2175 } 2176 } 2177 else 2178 grants = info->max_indirect_segments; 2179 psegs = DIV_ROUND_UP(grants, GRANTS_PER_PSEG); 2180 2181 err = fill_grant_buffer(rinfo, 2182 (grants + INDIRECT_GREFS(grants)) * BLK_RING_SIZE(info)); 2183 if (err) 2184 goto out_of_memory; 2185 2186 if (!info->bounce && info->max_indirect_segments) { 2187 /* 2188 * We are using indirect descriptors but don't have a bounce 2189 * buffer, we need to allocate a set of pages that can be 2190 * used for mapping indirect grefs 2191 */ 2192 int num = INDIRECT_GREFS(grants) * BLK_RING_SIZE(info); 2193 2194 BUG_ON(!list_empty(&rinfo->indirect_pages)); 2195 for (i = 0; i < num; i++) { 2196 struct page *indirect_page = alloc_page(GFP_KERNEL | 2197 __GFP_ZERO); 2198 if (!indirect_page) 2199 goto out_of_memory; 2200 list_add(&indirect_page->lru, &rinfo->indirect_pages); 2201 } 2202 } 2203 2204 for (i = 0; i < BLK_RING_SIZE(info); i++) { 2205 rinfo->shadow[i].grants_used = 2206 kvcalloc(grants, 2207 sizeof(rinfo->shadow[i].grants_used[0]), 2208 GFP_KERNEL); 2209 rinfo->shadow[i].sg = kvcalloc(psegs, 2210 sizeof(rinfo->shadow[i].sg[0]), 2211 GFP_KERNEL); 2212 if (info->max_indirect_segments) 2213 rinfo->shadow[i].indirect_grants = 2214 kvcalloc(INDIRECT_GREFS(grants), 2215 sizeof(rinfo->shadow[i].indirect_grants[0]), 2216 GFP_KERNEL); 2217 if ((rinfo->shadow[i].grants_used == NULL) || 2218 (rinfo->shadow[i].sg == NULL) || 2219 (info->max_indirect_segments && 2220 (rinfo->shadow[i].indirect_grants == NULL))) 2221 goto out_of_memory; 2222 sg_init_table(rinfo->shadow[i].sg, psegs); 2223 } 2224 2225 memalloc_noio_restore(memflags); 2226 2227 return 0; 2228 2229 out_of_memory: 2230 for (i = 0; i < BLK_RING_SIZE(info); i++) { 2231 kvfree(rinfo->shadow[i].grants_used); 2232 rinfo->shadow[i].grants_used = NULL; 2233 kvfree(rinfo->shadow[i].sg); 2234 rinfo->shadow[i].sg = NULL; 2235 kvfree(rinfo->shadow[i].indirect_grants); 2236 rinfo->shadow[i].indirect_grants = NULL; 2237 } 2238 if (!list_empty(&rinfo->indirect_pages)) { 2239 struct page *indirect_page, *n; 2240 list_for_each_entry_safe(indirect_page, n, &rinfo->indirect_pages, lru) { 2241 list_del(&indirect_page->lru); 2242 __free_page(indirect_page); 2243 } 2244 } 2245 2246 memalloc_noio_restore(memflags); 2247 2248 return -ENOMEM; 2249 } 2250 2251 /* 2252 * Gather all backend feature-* 2253 */ 2254 static void blkfront_gather_backend_features(struct blkfront_info *info) 2255 { 2256 unsigned int indirect_segments; 2257 2258 info->feature_flush = 0; 2259 info->feature_fua = 0; 2260 2261 /* 2262 * If there's no "feature-barrier" defined, then it means 2263 * we're dealing with a very old backend which writes 2264 * synchronously; nothing to do. 2265 * 2266 * If there are barriers, then we use flush. 2267 */ 2268 if (xenbus_read_unsigned(info->xbdev->otherend, "feature-barrier", 0)) { 2269 info->feature_flush = 1; 2270 info->feature_fua = 1; 2271 } 2272 2273 /* 2274 * And if there is "feature-flush-cache" use that above 2275 * barriers. 2276 */ 2277 if (xenbus_read_unsigned(info->xbdev->otherend, "feature-flush-cache", 2278 0)) { 2279 info->feature_flush = 1; 2280 info->feature_fua = 0; 2281 } 2282 2283 if (xenbus_read_unsigned(info->xbdev->otherend, "feature-discard", 0)) 2284 blkfront_setup_discard(info); 2285 2286 if (info->feature_persistent) 2287 info->feature_persistent = 2288 !!xenbus_read_unsigned(info->xbdev->otherend, 2289 "feature-persistent", 0); 2290 if (info->feature_persistent) 2291 info->bounce = true; 2292 2293 indirect_segments = xenbus_read_unsigned(info->xbdev->otherend, 2294 "feature-max-indirect-segments", 0); 2295 if (indirect_segments > xen_blkif_max_segments) 2296 indirect_segments = xen_blkif_max_segments; 2297 if (indirect_segments <= BLKIF_MAX_SEGMENTS_PER_REQUEST) 2298 indirect_segments = 0; 2299 info->max_indirect_segments = indirect_segments; 2300 2301 if (info->feature_persistent) { 2302 mutex_lock(&blkfront_mutex); 2303 schedule_delayed_work(&blkfront_work, HZ * 10); 2304 mutex_unlock(&blkfront_mutex); 2305 } 2306 } 2307 2308 /* 2309 * Invoked when the backend is finally 'ready' (and has told produced 2310 * the details about the physical device - #sectors, size, etc). 2311 */ 2312 static void blkfront_connect(struct blkfront_info *info) 2313 { 2314 unsigned long long sectors; 2315 unsigned long sector_size; 2316 unsigned int physical_sector_size; 2317 int err, i; 2318 struct blkfront_ring_info *rinfo; 2319 2320 switch (info->connected) { 2321 case BLKIF_STATE_CONNECTED: 2322 /* 2323 * Potentially, the back-end may be signalling 2324 * a capacity change; update the capacity. 2325 */ 2326 err = xenbus_scanf(XBT_NIL, info->xbdev->otherend, 2327 "sectors", "%Lu", §ors); 2328 if (XENBUS_EXIST_ERR(err)) 2329 return; 2330 printk(KERN_INFO "Setting capacity to %Lu\n", 2331 sectors); 2332 set_capacity_and_notify(info->gd, sectors); 2333 2334 return; 2335 case BLKIF_STATE_SUSPENDED: 2336 /* 2337 * If we are recovering from suspension, we need to wait 2338 * for the backend to announce it's features before 2339 * reconnecting, at least we need to know if the backend 2340 * supports indirect descriptors, and how many. 2341 */ 2342 blkif_recover(info); 2343 return; 2344 2345 default: 2346 break; 2347 } 2348 2349 dev_dbg(&info->xbdev->dev, "%s:%s.\n", 2350 __func__, info->xbdev->otherend); 2351 2352 err = xenbus_gather(XBT_NIL, info->xbdev->otherend, 2353 "sectors", "%llu", §ors, 2354 "info", "%u", &info->vdisk_info, 2355 "sector-size", "%lu", §or_size, 2356 NULL); 2357 if (err) { 2358 xenbus_dev_fatal(info->xbdev, err, 2359 "reading backend fields at %s", 2360 info->xbdev->otherend); 2361 return; 2362 } 2363 2364 /* 2365 * physical-sector-size is a newer field, so old backends may not 2366 * provide this. Assume physical sector size to be the same as 2367 * sector_size in that case. 2368 */ 2369 physical_sector_size = xenbus_read_unsigned(info->xbdev->otherend, 2370 "physical-sector-size", 2371 sector_size); 2372 blkfront_gather_backend_features(info); 2373 for_each_rinfo(info, rinfo, i) { 2374 err = blkfront_setup_indirect(rinfo); 2375 if (err) { 2376 xenbus_dev_fatal(info->xbdev, err, "setup_indirect at %s", 2377 info->xbdev->otherend); 2378 blkif_free(info, 0); 2379 break; 2380 } 2381 } 2382 2383 err = xlvbd_alloc_gendisk(sectors, info, sector_size, 2384 physical_sector_size); 2385 if (err) { 2386 xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s", 2387 info->xbdev->otherend); 2388 goto fail; 2389 } 2390 2391 xenbus_switch_state(info->xbdev, XenbusStateConnected); 2392 2393 /* Kick pending requests. */ 2394 info->connected = BLKIF_STATE_CONNECTED; 2395 for_each_rinfo(info, rinfo, i) 2396 kick_pending_request_queues(rinfo); 2397 2398 err = device_add_disk(&info->xbdev->dev, info->gd, NULL); 2399 if (err) { 2400 blk_cleanup_disk(info->gd); 2401 blk_mq_free_tag_set(&info->tag_set); 2402 info->rq = NULL; 2403 goto fail; 2404 } 2405 2406 info->is_ready = 1; 2407 return; 2408 2409 fail: 2410 blkif_free(info, 0); 2411 return; 2412 } 2413 2414 /* 2415 * Callback received when the backend's state changes. 2416 */ 2417 static void blkback_changed(struct xenbus_device *dev, 2418 enum xenbus_state backend_state) 2419 { 2420 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 2421 2422 dev_dbg(&dev->dev, "blkfront:blkback_changed to state %d.\n", backend_state); 2423 2424 switch (backend_state) { 2425 case XenbusStateInitWait: 2426 if (dev->state != XenbusStateInitialising) 2427 break; 2428 if (talk_to_blkback(dev, info)) 2429 break; 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 fallthrough; 2463 case XenbusStateClosing: 2464 blkfront_closing(info); 2465 break; 2466 } 2467 } 2468 2469 static int blkfront_remove(struct xenbus_device *xbdev) 2470 { 2471 struct blkfront_info *info = dev_get_drvdata(&xbdev->dev); 2472 2473 dev_dbg(&xbdev->dev, "%s removed", xbdev->nodename); 2474 2475 if (info->gd) 2476 del_gendisk(info->gd); 2477 2478 mutex_lock(&blkfront_mutex); 2479 list_del(&info->info_list); 2480 mutex_unlock(&blkfront_mutex); 2481 2482 blkif_free(info, 0); 2483 if (info->gd) { 2484 xlbd_release_minors(info->gd->first_minor, info->gd->minors); 2485 blk_cleanup_disk(info->gd); 2486 blk_mq_free_tag_set(&info->tag_set); 2487 } 2488 2489 kfree(info); 2490 return 0; 2491 } 2492 2493 static int blkfront_is_ready(struct xenbus_device *dev) 2494 { 2495 struct blkfront_info *info = dev_get_drvdata(&dev->dev); 2496 2497 return info->is_ready && info->xbdev; 2498 } 2499 2500 static const struct block_device_operations xlvbd_block_fops = 2501 { 2502 .owner = THIS_MODULE, 2503 .getgeo = blkif_getgeo, 2504 .ioctl = blkif_ioctl, 2505 .compat_ioctl = blkdev_compat_ptr_ioctl, 2506 }; 2507 2508 2509 static const struct xenbus_device_id blkfront_ids[] = { 2510 { "vbd" }, 2511 { "" } 2512 }; 2513 2514 static struct xenbus_driver blkfront_driver = { 2515 .ids = blkfront_ids, 2516 .probe = blkfront_probe, 2517 .remove = blkfront_remove, 2518 .resume = blkfront_resume, 2519 .otherend_changed = blkback_changed, 2520 .is_ready = blkfront_is_ready, 2521 }; 2522 2523 static void purge_persistent_grants(struct blkfront_info *info) 2524 { 2525 unsigned int i; 2526 unsigned long flags; 2527 struct blkfront_ring_info *rinfo; 2528 2529 for_each_rinfo(info, rinfo, i) { 2530 struct grant *gnt_list_entry, *tmp; 2531 LIST_HEAD(grants); 2532 2533 spin_lock_irqsave(&rinfo->ring_lock, flags); 2534 2535 if (rinfo->persistent_gnts_c == 0) { 2536 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 2537 continue; 2538 } 2539 2540 list_for_each_entry_safe(gnt_list_entry, tmp, &rinfo->grants, 2541 node) { 2542 if (gnt_list_entry->gref == INVALID_GRANT_REF || 2543 !gnttab_try_end_foreign_access(gnt_list_entry->gref)) 2544 continue; 2545 2546 list_del(&gnt_list_entry->node); 2547 rinfo->persistent_gnts_c--; 2548 gnt_list_entry->gref = INVALID_GRANT_REF; 2549 list_add_tail(&gnt_list_entry->node, &grants); 2550 } 2551 2552 list_splice_tail(&grants, &rinfo->grants); 2553 2554 spin_unlock_irqrestore(&rinfo->ring_lock, flags); 2555 } 2556 } 2557 2558 static void blkfront_delay_work(struct work_struct *work) 2559 { 2560 struct blkfront_info *info; 2561 bool need_schedule_work = false; 2562 2563 /* 2564 * Note that when using bounce buffers but not persistent grants 2565 * there's no need to run blkfront_delay_work because grants are 2566 * revoked in blkif_completion or else an error is reported and the 2567 * connection is closed. 2568 */ 2569 2570 mutex_lock(&blkfront_mutex); 2571 2572 list_for_each_entry(info, &info_list, info_list) { 2573 if (info->feature_persistent) { 2574 need_schedule_work = true; 2575 mutex_lock(&info->mutex); 2576 purge_persistent_grants(info); 2577 mutex_unlock(&info->mutex); 2578 } 2579 } 2580 2581 if (need_schedule_work) 2582 schedule_delayed_work(&blkfront_work, HZ * 10); 2583 2584 mutex_unlock(&blkfront_mutex); 2585 } 2586 2587 static int __init xlblk_init(void) 2588 { 2589 int ret; 2590 int nr_cpus = num_online_cpus(); 2591 2592 if (!xen_domain()) 2593 return -ENODEV; 2594 2595 if (!xen_has_pv_disk_devices()) 2596 return -ENODEV; 2597 2598 if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) { 2599 pr_warn("xen_blk: can't get major %d with name %s\n", 2600 XENVBD_MAJOR, DEV_NAME); 2601 return -ENODEV; 2602 } 2603 2604 if (xen_blkif_max_segments < BLKIF_MAX_SEGMENTS_PER_REQUEST) 2605 xen_blkif_max_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; 2606 2607 if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { 2608 pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", 2609 xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); 2610 xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; 2611 } 2612 2613 if (xen_blkif_max_queues > nr_cpus) { 2614 pr_info("Invalid max_queues (%d), will use default max: %d.\n", 2615 xen_blkif_max_queues, nr_cpus); 2616 xen_blkif_max_queues = nr_cpus; 2617 } 2618 2619 INIT_DELAYED_WORK(&blkfront_work, blkfront_delay_work); 2620 2621 ret = xenbus_register_frontend(&blkfront_driver); 2622 if (ret) { 2623 unregister_blkdev(XENVBD_MAJOR, DEV_NAME); 2624 return ret; 2625 } 2626 2627 return 0; 2628 } 2629 module_init(xlblk_init); 2630 2631 2632 static void __exit xlblk_exit(void) 2633 { 2634 cancel_delayed_work_sync(&blkfront_work); 2635 2636 xenbus_unregister_driver(&blkfront_driver); 2637 unregister_blkdev(XENVBD_MAJOR, DEV_NAME); 2638 kfree(minors); 2639 } 2640 module_exit(xlblk_exit); 2641 2642 MODULE_DESCRIPTION("Xen virtual block device frontend"); 2643 MODULE_LICENSE("GPL"); 2644 MODULE_ALIAS_BLOCKDEV_MAJOR(XENVBD_MAJOR); 2645 MODULE_ALIAS("xen:vbd"); 2646 MODULE_ALIAS("xenblk"); 2647