1 /****************************************************************************** 2 * 3 * Back-end of the driver for virtual block devices. This portion of the 4 * driver exports a 'unified' block-device interface that can be accessed 5 * by any operating system that implements a compatible front end. A 6 * reference front-end implementation can be found in: 7 * drivers/block/xen-blkfront.c 8 * 9 * Copyright (c) 2003-2004, Keir Fraser & Steve Hand 10 * Copyright (c) 2005, Christopher Clark 11 * 12 * This program is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU General Public License version 2 14 * as published by the Free Software Foundation; or, when distributed 15 * separately from the Linux kernel or incorporated into other 16 * software packages, subject to the following license: 17 * 18 * Permission is hereby granted, free of charge, to any person obtaining a copy 19 * of this source file (the "Software"), to deal in the Software without 20 * restriction, including without limitation the rights to use, copy, modify, 21 * merge, publish, distribute, sublicense, and/or sell copies of the Software, 22 * and to permit persons to whom the Software is furnished to do so, subject to 23 * the following conditions: 24 * 25 * The above copyright notice and this permission notice shall be included in 26 * all copies or substantial portions of the Software. 27 * 28 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 29 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 30 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 31 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 32 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 33 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 34 * IN THE SOFTWARE. 35 */ 36 37 #define pr_fmt(fmt) "xen-blkback: " fmt 38 39 #include <linux/spinlock.h> 40 #include <linux/kthread.h> 41 #include <linux/list.h> 42 #include <linux/delay.h> 43 #include <linux/freezer.h> 44 #include <linux/bitmap.h> 45 46 #include <xen/events.h> 47 #include <xen/page.h> 48 #include <xen/xen.h> 49 #include <asm/xen/hypervisor.h> 50 #include <asm/xen/hypercall.h> 51 #include <xen/balloon.h> 52 #include <xen/grant_table.h> 53 #include "common.h" 54 55 /* 56 * Maximum number of unused free pages to keep in the internal buffer. 57 * Setting this to a value too low will reduce memory used in each backend, 58 * but can have a performance penalty. 59 * 60 * A sane value is xen_blkif_reqs * BLKIF_MAX_SEGMENTS_PER_REQUEST, but can 61 * be set to a lower value that might degrade performance on some intensive 62 * IO workloads. 63 */ 64 65 static int xen_blkif_max_buffer_pages = 1024; 66 module_param_named(max_buffer_pages, xen_blkif_max_buffer_pages, int, 0644); 67 MODULE_PARM_DESC(max_buffer_pages, 68 "Maximum number of free pages to keep in each block backend buffer"); 69 70 /* 71 * Maximum number of grants to map persistently in blkback. For maximum 72 * performance this should be the total numbers of grants that can be used 73 * to fill the ring, but since this might become too high, specially with 74 * the use of indirect descriptors, we set it to a value that provides good 75 * performance without using too much memory. 76 * 77 * When the list of persistent grants is full we clean it up using a LRU 78 * algorithm. 79 */ 80 81 static int xen_blkif_max_pgrants = 1056; 82 module_param_named(max_persistent_grants, xen_blkif_max_pgrants, int, 0644); 83 MODULE_PARM_DESC(max_persistent_grants, 84 "Maximum number of grants to map persistently"); 85 86 /* 87 * Maximum order of pages to be used for the shared ring between front and 88 * backend, 4KB page granularity is used. 89 */ 90 unsigned int xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; 91 module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, S_IRUGO); 92 MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring"); 93 /* 94 * The LRU mechanism to clean the lists of persistent grants needs to 95 * be executed periodically. The time interval between consecutive executions 96 * of the purge mechanism is set in ms. 97 */ 98 #define LRU_INTERVAL 100 99 100 /* 101 * When the persistent grants list is full we will remove unused grants 102 * from the list. The percent number of grants to be removed at each LRU 103 * execution. 104 */ 105 #define LRU_PERCENT_CLEAN 5 106 107 /* Run-time switchable: /sys/module/blkback/parameters/ */ 108 static unsigned int log_stats; 109 module_param(log_stats, int, 0644); 110 111 #define BLKBACK_INVALID_HANDLE (~0) 112 113 /* Number of free pages to remove on each call to gnttab_free_pages */ 114 #define NUM_BATCH_FREE_PAGES 10 115 116 static inline int get_free_page(struct xen_blkif *blkif, struct page **page) 117 { 118 unsigned long flags; 119 120 spin_lock_irqsave(&blkif->free_pages_lock, flags); 121 if (list_empty(&blkif->free_pages)) { 122 BUG_ON(blkif->free_pages_num != 0); 123 spin_unlock_irqrestore(&blkif->free_pages_lock, flags); 124 return gnttab_alloc_pages(1, page); 125 } 126 BUG_ON(blkif->free_pages_num == 0); 127 page[0] = list_first_entry(&blkif->free_pages, struct page, lru); 128 list_del(&page[0]->lru); 129 blkif->free_pages_num--; 130 spin_unlock_irqrestore(&blkif->free_pages_lock, flags); 131 132 return 0; 133 } 134 135 static inline void put_free_pages(struct xen_blkif *blkif, struct page **page, 136 int num) 137 { 138 unsigned long flags; 139 int i; 140 141 spin_lock_irqsave(&blkif->free_pages_lock, flags); 142 for (i = 0; i < num; i++) 143 list_add(&page[i]->lru, &blkif->free_pages); 144 blkif->free_pages_num += num; 145 spin_unlock_irqrestore(&blkif->free_pages_lock, flags); 146 } 147 148 static inline void shrink_free_pagepool(struct xen_blkif *blkif, int num) 149 { 150 /* Remove requested pages in batches of NUM_BATCH_FREE_PAGES */ 151 struct page *page[NUM_BATCH_FREE_PAGES]; 152 unsigned int num_pages = 0; 153 unsigned long flags; 154 155 spin_lock_irqsave(&blkif->free_pages_lock, flags); 156 while (blkif->free_pages_num > num) { 157 BUG_ON(list_empty(&blkif->free_pages)); 158 page[num_pages] = list_first_entry(&blkif->free_pages, 159 struct page, lru); 160 list_del(&page[num_pages]->lru); 161 blkif->free_pages_num--; 162 if (++num_pages == NUM_BATCH_FREE_PAGES) { 163 spin_unlock_irqrestore(&blkif->free_pages_lock, flags); 164 gnttab_free_pages(num_pages, page); 165 spin_lock_irqsave(&blkif->free_pages_lock, flags); 166 num_pages = 0; 167 } 168 } 169 spin_unlock_irqrestore(&blkif->free_pages_lock, flags); 170 if (num_pages != 0) 171 gnttab_free_pages(num_pages, page); 172 } 173 174 #define vaddr(page) ((unsigned long)pfn_to_kaddr(page_to_pfn(page))) 175 176 static int do_block_io_op(struct xen_blkif *blkif); 177 static int dispatch_rw_block_io(struct xen_blkif *blkif, 178 struct blkif_request *req, 179 struct pending_req *pending_req); 180 static void make_response(struct xen_blkif *blkif, u64 id, 181 unsigned short op, int st); 182 183 #define foreach_grant_safe(pos, n, rbtree, node) \ 184 for ((pos) = container_of(rb_first((rbtree)), typeof(*(pos)), node), \ 185 (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL; \ 186 &(pos)->node != NULL; \ 187 (pos) = container_of(n, typeof(*(pos)), node), \ 188 (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL) 189 190 191 /* 192 * We don't need locking around the persistent grant helpers 193 * because blkback uses a single-thread for each backed, so we 194 * can be sure that this functions will never be called recursively. 195 * 196 * The only exception to that is put_persistent_grant, that can be called 197 * from interrupt context (by xen_blkbk_unmap), so we have to use atomic 198 * bit operations to modify the flags of a persistent grant and to count 199 * the number of used grants. 200 */ 201 static int add_persistent_gnt(struct xen_blkif *blkif, 202 struct persistent_gnt *persistent_gnt) 203 { 204 struct rb_node **new = NULL, *parent = NULL; 205 struct persistent_gnt *this; 206 207 if (blkif->persistent_gnt_c >= xen_blkif_max_pgrants) { 208 if (!blkif->vbd.overflow_max_grants) 209 blkif->vbd.overflow_max_grants = 1; 210 return -EBUSY; 211 } 212 /* Figure out where to put new node */ 213 new = &blkif->persistent_gnts.rb_node; 214 while (*new) { 215 this = container_of(*new, struct persistent_gnt, node); 216 217 parent = *new; 218 if (persistent_gnt->gnt < this->gnt) 219 new = &((*new)->rb_left); 220 else if (persistent_gnt->gnt > this->gnt) 221 new = &((*new)->rb_right); 222 else { 223 pr_alert_ratelimited("trying to add a gref that's already in the tree\n"); 224 return -EINVAL; 225 } 226 } 227 228 bitmap_zero(persistent_gnt->flags, PERSISTENT_GNT_FLAGS_SIZE); 229 set_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags); 230 /* Add new node and rebalance tree. */ 231 rb_link_node(&(persistent_gnt->node), parent, new); 232 rb_insert_color(&(persistent_gnt->node), &blkif->persistent_gnts); 233 blkif->persistent_gnt_c++; 234 atomic_inc(&blkif->persistent_gnt_in_use); 235 return 0; 236 } 237 238 static struct persistent_gnt *get_persistent_gnt(struct xen_blkif *blkif, 239 grant_ref_t gref) 240 { 241 struct persistent_gnt *data; 242 struct rb_node *node = NULL; 243 244 node = blkif->persistent_gnts.rb_node; 245 while (node) { 246 data = container_of(node, struct persistent_gnt, node); 247 248 if (gref < data->gnt) 249 node = node->rb_left; 250 else if (gref > data->gnt) 251 node = node->rb_right; 252 else { 253 if(test_bit(PERSISTENT_GNT_ACTIVE, data->flags)) { 254 pr_alert_ratelimited("requesting a grant already in use\n"); 255 return NULL; 256 } 257 set_bit(PERSISTENT_GNT_ACTIVE, data->flags); 258 atomic_inc(&blkif->persistent_gnt_in_use); 259 return data; 260 } 261 } 262 return NULL; 263 } 264 265 static void put_persistent_gnt(struct xen_blkif *blkif, 266 struct persistent_gnt *persistent_gnt) 267 { 268 if(!test_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags)) 269 pr_alert_ratelimited("freeing a grant already unused\n"); 270 set_bit(PERSISTENT_GNT_WAS_ACTIVE, persistent_gnt->flags); 271 clear_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags); 272 atomic_dec(&blkif->persistent_gnt_in_use); 273 } 274 275 static void free_persistent_gnts(struct xen_blkif *blkif, struct rb_root *root, 276 unsigned int num) 277 { 278 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 279 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 280 struct persistent_gnt *persistent_gnt; 281 struct rb_node *n; 282 int segs_to_unmap = 0; 283 struct gntab_unmap_queue_data unmap_data; 284 285 unmap_data.pages = pages; 286 unmap_data.unmap_ops = unmap; 287 unmap_data.kunmap_ops = NULL; 288 289 foreach_grant_safe(persistent_gnt, n, root, node) { 290 BUG_ON(persistent_gnt->handle == 291 BLKBACK_INVALID_HANDLE); 292 gnttab_set_unmap_op(&unmap[segs_to_unmap], 293 (unsigned long) pfn_to_kaddr(page_to_pfn( 294 persistent_gnt->page)), 295 GNTMAP_host_map, 296 persistent_gnt->handle); 297 298 pages[segs_to_unmap] = persistent_gnt->page; 299 300 if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST || 301 !rb_next(&persistent_gnt->node)) { 302 303 unmap_data.count = segs_to_unmap; 304 BUG_ON(gnttab_unmap_refs_sync(&unmap_data)); 305 306 put_free_pages(blkif, pages, segs_to_unmap); 307 segs_to_unmap = 0; 308 } 309 310 rb_erase(&persistent_gnt->node, root); 311 kfree(persistent_gnt); 312 num--; 313 } 314 BUG_ON(num != 0); 315 } 316 317 void xen_blkbk_unmap_purged_grants(struct work_struct *work) 318 { 319 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 320 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 321 struct persistent_gnt *persistent_gnt; 322 int segs_to_unmap = 0; 323 struct xen_blkif *blkif = container_of(work, typeof(*blkif), persistent_purge_work); 324 struct gntab_unmap_queue_data unmap_data; 325 326 unmap_data.pages = pages; 327 unmap_data.unmap_ops = unmap; 328 unmap_data.kunmap_ops = NULL; 329 330 while(!list_empty(&blkif->persistent_purge_list)) { 331 persistent_gnt = list_first_entry(&blkif->persistent_purge_list, 332 struct persistent_gnt, 333 remove_node); 334 list_del(&persistent_gnt->remove_node); 335 336 gnttab_set_unmap_op(&unmap[segs_to_unmap], 337 vaddr(persistent_gnt->page), 338 GNTMAP_host_map, 339 persistent_gnt->handle); 340 341 pages[segs_to_unmap] = persistent_gnt->page; 342 343 if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) { 344 unmap_data.count = segs_to_unmap; 345 BUG_ON(gnttab_unmap_refs_sync(&unmap_data)); 346 put_free_pages(blkif, pages, segs_to_unmap); 347 segs_to_unmap = 0; 348 } 349 kfree(persistent_gnt); 350 } 351 if (segs_to_unmap > 0) { 352 unmap_data.count = segs_to_unmap; 353 BUG_ON(gnttab_unmap_refs_sync(&unmap_data)); 354 put_free_pages(blkif, pages, segs_to_unmap); 355 } 356 } 357 358 static void purge_persistent_gnt(struct xen_blkif *blkif) 359 { 360 struct persistent_gnt *persistent_gnt; 361 struct rb_node *n; 362 unsigned int num_clean, total; 363 bool scan_used = false, clean_used = false; 364 struct rb_root *root; 365 366 if (blkif->persistent_gnt_c < xen_blkif_max_pgrants || 367 (blkif->persistent_gnt_c == xen_blkif_max_pgrants && 368 !blkif->vbd.overflow_max_grants)) { 369 return; 370 } 371 372 if (work_busy(&blkif->persistent_purge_work)) { 373 pr_alert_ratelimited("Scheduled work from previous purge is still busy, cannot purge list\n"); 374 return; 375 } 376 377 num_clean = (xen_blkif_max_pgrants / 100) * LRU_PERCENT_CLEAN; 378 num_clean = blkif->persistent_gnt_c - xen_blkif_max_pgrants + num_clean; 379 num_clean = min(blkif->persistent_gnt_c, num_clean); 380 if ((num_clean == 0) || 381 (num_clean > (blkif->persistent_gnt_c - atomic_read(&blkif->persistent_gnt_in_use)))) 382 return; 383 384 /* 385 * At this point, we can assure that there will be no calls 386 * to get_persistent_grant (because we are executing this code from 387 * xen_blkif_schedule), there can only be calls to put_persistent_gnt, 388 * which means that the number of currently used grants will go down, 389 * but never up, so we will always be able to remove the requested 390 * number of grants. 391 */ 392 393 total = num_clean; 394 395 pr_debug("Going to purge %u persistent grants\n", num_clean); 396 397 BUG_ON(!list_empty(&blkif->persistent_purge_list)); 398 root = &blkif->persistent_gnts; 399 purge_list: 400 foreach_grant_safe(persistent_gnt, n, root, node) { 401 BUG_ON(persistent_gnt->handle == 402 BLKBACK_INVALID_HANDLE); 403 404 if (clean_used) { 405 clear_bit(PERSISTENT_GNT_WAS_ACTIVE, persistent_gnt->flags); 406 continue; 407 } 408 409 if (test_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags)) 410 continue; 411 if (!scan_used && 412 (test_bit(PERSISTENT_GNT_WAS_ACTIVE, persistent_gnt->flags))) 413 continue; 414 415 rb_erase(&persistent_gnt->node, root); 416 list_add(&persistent_gnt->remove_node, 417 &blkif->persistent_purge_list); 418 if (--num_clean == 0) 419 goto finished; 420 } 421 /* 422 * If we get here it means we also need to start cleaning 423 * grants that were used since last purge in order to cope 424 * with the requested num 425 */ 426 if (!scan_used && !clean_used) { 427 pr_debug("Still missing %u purged frames\n", num_clean); 428 scan_used = true; 429 goto purge_list; 430 } 431 finished: 432 if (!clean_used) { 433 pr_debug("Finished scanning for grants to clean, removing used flag\n"); 434 clean_used = true; 435 goto purge_list; 436 } 437 438 blkif->persistent_gnt_c -= (total - num_clean); 439 blkif->vbd.overflow_max_grants = 0; 440 441 /* We can defer this work */ 442 schedule_work(&blkif->persistent_purge_work); 443 pr_debug("Purged %u/%u\n", (total - num_clean), total); 444 return; 445 } 446 447 /* 448 * Retrieve from the 'pending_reqs' a free pending_req structure to be used. 449 */ 450 static struct pending_req *alloc_req(struct xen_blkif *blkif) 451 { 452 struct pending_req *req = NULL; 453 unsigned long flags; 454 455 spin_lock_irqsave(&blkif->pending_free_lock, flags); 456 if (!list_empty(&blkif->pending_free)) { 457 req = list_entry(blkif->pending_free.next, struct pending_req, 458 free_list); 459 list_del(&req->free_list); 460 } 461 spin_unlock_irqrestore(&blkif->pending_free_lock, flags); 462 return req; 463 } 464 465 /* 466 * Return the 'pending_req' structure back to the freepool. We also 467 * wake up the thread if it was waiting for a free page. 468 */ 469 static void free_req(struct xen_blkif *blkif, struct pending_req *req) 470 { 471 unsigned long flags; 472 int was_empty; 473 474 spin_lock_irqsave(&blkif->pending_free_lock, flags); 475 was_empty = list_empty(&blkif->pending_free); 476 list_add(&req->free_list, &blkif->pending_free); 477 spin_unlock_irqrestore(&blkif->pending_free_lock, flags); 478 if (was_empty) 479 wake_up(&blkif->pending_free_wq); 480 } 481 482 /* 483 * Routines for managing virtual block devices (vbds). 484 */ 485 static int xen_vbd_translate(struct phys_req *req, struct xen_blkif *blkif, 486 int operation) 487 { 488 struct xen_vbd *vbd = &blkif->vbd; 489 int rc = -EACCES; 490 491 if ((operation != READ) && vbd->readonly) 492 goto out; 493 494 if (likely(req->nr_sects)) { 495 blkif_sector_t end = req->sector_number + req->nr_sects; 496 497 if (unlikely(end < req->sector_number)) 498 goto out; 499 if (unlikely(end > vbd_sz(vbd))) 500 goto out; 501 } 502 503 req->dev = vbd->pdevice; 504 req->bdev = vbd->bdev; 505 rc = 0; 506 507 out: 508 return rc; 509 } 510 511 static void xen_vbd_resize(struct xen_blkif *blkif) 512 { 513 struct xen_vbd *vbd = &blkif->vbd; 514 struct xenbus_transaction xbt; 515 int err; 516 struct xenbus_device *dev = xen_blkbk_xenbus(blkif->be); 517 unsigned long long new_size = vbd_sz(vbd); 518 519 pr_info("VBD Resize: Domid: %d, Device: (%d, %d)\n", 520 blkif->domid, MAJOR(vbd->pdevice), MINOR(vbd->pdevice)); 521 pr_info("VBD Resize: new size %llu\n", new_size); 522 vbd->size = new_size; 523 again: 524 err = xenbus_transaction_start(&xbt); 525 if (err) { 526 pr_warn("Error starting transaction\n"); 527 return; 528 } 529 err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu", 530 (unsigned long long)vbd_sz(vbd)); 531 if (err) { 532 pr_warn("Error writing new size\n"); 533 goto abort; 534 } 535 /* 536 * Write the current state; we will use this to synchronize 537 * the front-end. If the current state is "connected" the 538 * front-end will get the new size information online. 539 */ 540 err = xenbus_printf(xbt, dev->nodename, "state", "%d", dev->state); 541 if (err) { 542 pr_warn("Error writing the state\n"); 543 goto abort; 544 } 545 546 err = xenbus_transaction_end(xbt, 0); 547 if (err == -EAGAIN) 548 goto again; 549 if (err) 550 pr_warn("Error ending transaction\n"); 551 return; 552 abort: 553 xenbus_transaction_end(xbt, 1); 554 } 555 556 /* 557 * Notification from the guest OS. 558 */ 559 static void blkif_notify_work(struct xen_blkif *blkif) 560 { 561 blkif->waiting_reqs = 1; 562 wake_up(&blkif->wq); 563 } 564 565 irqreturn_t xen_blkif_be_int(int irq, void *dev_id) 566 { 567 blkif_notify_work(dev_id); 568 return IRQ_HANDLED; 569 } 570 571 /* 572 * SCHEDULER FUNCTIONS 573 */ 574 575 static void print_stats(struct xen_blkif *blkif) 576 { 577 pr_info("(%s): oo %3llu | rd %4llu | wr %4llu | f %4llu" 578 " | ds %4llu | pg: %4u/%4d\n", 579 current->comm, blkif->st_oo_req, 580 blkif->st_rd_req, blkif->st_wr_req, 581 blkif->st_f_req, blkif->st_ds_req, 582 blkif->persistent_gnt_c, 583 xen_blkif_max_pgrants); 584 blkif->st_print = jiffies + msecs_to_jiffies(10 * 1000); 585 blkif->st_rd_req = 0; 586 blkif->st_wr_req = 0; 587 blkif->st_oo_req = 0; 588 blkif->st_ds_req = 0; 589 } 590 591 int xen_blkif_schedule(void *arg) 592 { 593 struct xen_blkif *blkif = arg; 594 struct xen_vbd *vbd = &blkif->vbd; 595 unsigned long timeout; 596 int ret; 597 598 xen_blkif_get(blkif); 599 600 while (!kthread_should_stop()) { 601 if (try_to_freeze()) 602 continue; 603 if (unlikely(vbd->size != vbd_sz(vbd))) 604 xen_vbd_resize(blkif); 605 606 timeout = msecs_to_jiffies(LRU_INTERVAL); 607 608 timeout = wait_event_interruptible_timeout( 609 blkif->wq, 610 blkif->waiting_reqs || kthread_should_stop(), 611 timeout); 612 if (timeout == 0) 613 goto purge_gnt_list; 614 timeout = wait_event_interruptible_timeout( 615 blkif->pending_free_wq, 616 !list_empty(&blkif->pending_free) || 617 kthread_should_stop(), 618 timeout); 619 if (timeout == 0) 620 goto purge_gnt_list; 621 622 blkif->waiting_reqs = 0; 623 smp_mb(); /* clear flag *before* checking for work */ 624 625 ret = do_block_io_op(blkif); 626 if (ret > 0) 627 blkif->waiting_reqs = 1; 628 if (ret == -EACCES) 629 wait_event_interruptible(blkif->shutdown_wq, 630 kthread_should_stop()); 631 632 purge_gnt_list: 633 if (blkif->vbd.feature_gnt_persistent && 634 time_after(jiffies, blkif->next_lru)) { 635 purge_persistent_gnt(blkif); 636 blkif->next_lru = jiffies + msecs_to_jiffies(LRU_INTERVAL); 637 } 638 639 /* Shrink if we have more than xen_blkif_max_buffer_pages */ 640 shrink_free_pagepool(blkif, xen_blkif_max_buffer_pages); 641 642 if (log_stats && time_after(jiffies, blkif->st_print)) 643 print_stats(blkif); 644 } 645 646 /* Drain pending purge work */ 647 flush_work(&blkif->persistent_purge_work); 648 649 if (log_stats) 650 print_stats(blkif); 651 652 blkif->xenblkd = NULL; 653 xen_blkif_put(blkif); 654 655 return 0; 656 } 657 658 /* 659 * Remove persistent grants and empty the pool of free pages 660 */ 661 void xen_blkbk_free_caches(struct xen_blkif *blkif) 662 { 663 /* Free all persistent grant pages */ 664 if (!RB_EMPTY_ROOT(&blkif->persistent_gnts)) 665 free_persistent_gnts(blkif, &blkif->persistent_gnts, 666 blkif->persistent_gnt_c); 667 668 BUG_ON(!RB_EMPTY_ROOT(&blkif->persistent_gnts)); 669 blkif->persistent_gnt_c = 0; 670 671 /* Since we are shutting down remove all pages from the buffer */ 672 shrink_free_pagepool(blkif, 0 /* All */); 673 } 674 675 static unsigned int xen_blkbk_unmap_prepare( 676 struct xen_blkif *blkif, 677 struct grant_page **pages, 678 unsigned int num, 679 struct gnttab_unmap_grant_ref *unmap_ops, 680 struct page **unmap_pages) 681 { 682 unsigned int i, invcount = 0; 683 684 for (i = 0; i < num; i++) { 685 if (pages[i]->persistent_gnt != NULL) { 686 put_persistent_gnt(blkif, pages[i]->persistent_gnt); 687 continue; 688 } 689 if (pages[i]->handle == BLKBACK_INVALID_HANDLE) 690 continue; 691 unmap_pages[invcount] = pages[i]->page; 692 gnttab_set_unmap_op(&unmap_ops[invcount], vaddr(pages[i]->page), 693 GNTMAP_host_map, pages[i]->handle); 694 pages[i]->handle = BLKBACK_INVALID_HANDLE; 695 invcount++; 696 } 697 698 return invcount; 699 } 700 701 static void xen_blkbk_unmap_and_respond_callback(int result, struct gntab_unmap_queue_data *data) 702 { 703 struct pending_req* pending_req = (struct pending_req*) (data->data); 704 struct xen_blkif *blkif = pending_req->blkif; 705 706 /* BUG_ON used to reproduce existing behaviour, 707 but is this the best way to deal with this? */ 708 BUG_ON(result); 709 710 put_free_pages(blkif, data->pages, data->count); 711 make_response(blkif, pending_req->id, 712 pending_req->operation, pending_req->status); 713 free_req(blkif, pending_req); 714 /* 715 * Make sure the request is freed before releasing blkif, 716 * or there could be a race between free_req and the 717 * cleanup done in xen_blkif_free during shutdown. 718 * 719 * NB: The fact that we might try to wake up pending_free_wq 720 * before drain_complete (in case there's a drain going on) 721 * it's not a problem with our current implementation 722 * because we can assure there's no thread waiting on 723 * pending_free_wq if there's a drain going on, but it has 724 * to be taken into account if the current model is changed. 725 */ 726 if (atomic_dec_and_test(&blkif->inflight) && atomic_read(&blkif->drain)) { 727 complete(&blkif->drain_complete); 728 } 729 xen_blkif_put(blkif); 730 } 731 732 static void xen_blkbk_unmap_and_respond(struct pending_req *req) 733 { 734 struct gntab_unmap_queue_data* work = &req->gnttab_unmap_data; 735 struct xen_blkif *blkif = req->blkif; 736 struct grant_page **pages = req->segments; 737 unsigned int invcount; 738 739 invcount = xen_blkbk_unmap_prepare(blkif, pages, req->nr_segs, 740 req->unmap, req->unmap_pages); 741 742 work->data = req; 743 work->done = xen_blkbk_unmap_and_respond_callback; 744 work->unmap_ops = req->unmap; 745 work->kunmap_ops = NULL; 746 work->pages = req->unmap_pages; 747 work->count = invcount; 748 749 gnttab_unmap_refs_async(&req->gnttab_unmap_data); 750 } 751 752 753 /* 754 * Unmap the grant references. 755 * 756 * This could accumulate ops up to the batch size to reduce the number 757 * of hypercalls, but since this is only used in error paths there's 758 * no real need. 759 */ 760 static void xen_blkbk_unmap(struct xen_blkif *blkif, 761 struct grant_page *pages[], 762 int num) 763 { 764 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 765 struct page *unmap_pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 766 unsigned int invcount = 0; 767 int ret; 768 769 while (num) { 770 unsigned int batch = min(num, BLKIF_MAX_SEGMENTS_PER_REQUEST); 771 772 invcount = xen_blkbk_unmap_prepare(blkif, pages, batch, 773 unmap, unmap_pages); 774 if (invcount) { 775 ret = gnttab_unmap_refs(unmap, NULL, unmap_pages, invcount); 776 BUG_ON(ret); 777 put_free_pages(blkif, unmap_pages, invcount); 778 } 779 pages += batch; 780 num -= batch; 781 } 782 } 783 784 static int xen_blkbk_map(struct xen_blkif *blkif, 785 struct grant_page *pages[], 786 int num, bool ro) 787 { 788 struct gnttab_map_grant_ref map[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 789 struct page *pages_to_gnt[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 790 struct persistent_gnt *persistent_gnt = NULL; 791 phys_addr_t addr = 0; 792 int i, seg_idx, new_map_idx; 793 int segs_to_map = 0; 794 int ret = 0; 795 int last_map = 0, map_until = 0; 796 int use_persistent_gnts; 797 798 use_persistent_gnts = (blkif->vbd.feature_gnt_persistent); 799 800 /* 801 * Fill out preq.nr_sects with proper amount of sectors, and setup 802 * assign map[..] with the PFN of the page in our domain with the 803 * corresponding grant reference for each page. 804 */ 805 again: 806 for (i = map_until; i < num; i++) { 807 uint32_t flags; 808 809 if (use_persistent_gnts) 810 persistent_gnt = get_persistent_gnt( 811 blkif, 812 pages[i]->gref); 813 814 if (persistent_gnt) { 815 /* 816 * We are using persistent grants and 817 * the grant is already mapped 818 */ 819 pages[i]->page = persistent_gnt->page; 820 pages[i]->persistent_gnt = persistent_gnt; 821 } else { 822 if (get_free_page(blkif, &pages[i]->page)) 823 goto out_of_memory; 824 addr = vaddr(pages[i]->page); 825 pages_to_gnt[segs_to_map] = pages[i]->page; 826 pages[i]->persistent_gnt = NULL; 827 flags = GNTMAP_host_map; 828 if (!use_persistent_gnts && ro) 829 flags |= GNTMAP_readonly; 830 gnttab_set_map_op(&map[segs_to_map++], addr, 831 flags, pages[i]->gref, 832 blkif->domid); 833 } 834 map_until = i + 1; 835 if (segs_to_map == BLKIF_MAX_SEGMENTS_PER_REQUEST) 836 break; 837 } 838 839 if (segs_to_map) { 840 ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map); 841 BUG_ON(ret); 842 } 843 844 /* 845 * Now swizzle the MFN in our domain with the MFN from the other domain 846 * so that when we access vaddr(pending_req,i) it has the contents of 847 * the page from the other domain. 848 */ 849 for (seg_idx = last_map, new_map_idx = 0; seg_idx < map_until; seg_idx++) { 850 if (!pages[seg_idx]->persistent_gnt) { 851 /* This is a newly mapped grant */ 852 BUG_ON(new_map_idx >= segs_to_map); 853 if (unlikely(map[new_map_idx].status != 0)) { 854 pr_debug("invalid buffer -- could not remap it\n"); 855 put_free_pages(blkif, &pages[seg_idx]->page, 1); 856 pages[seg_idx]->handle = BLKBACK_INVALID_HANDLE; 857 ret |= 1; 858 goto next; 859 } 860 pages[seg_idx]->handle = map[new_map_idx].handle; 861 } else { 862 continue; 863 } 864 if (use_persistent_gnts && 865 blkif->persistent_gnt_c < xen_blkif_max_pgrants) { 866 /* 867 * We are using persistent grants, the grant is 868 * not mapped but we might have room for it. 869 */ 870 persistent_gnt = kmalloc(sizeof(struct persistent_gnt), 871 GFP_KERNEL); 872 if (!persistent_gnt) { 873 /* 874 * If we don't have enough memory to 875 * allocate the persistent_gnt struct 876 * map this grant non-persistenly 877 */ 878 goto next; 879 } 880 persistent_gnt->gnt = map[new_map_idx].ref; 881 persistent_gnt->handle = map[new_map_idx].handle; 882 persistent_gnt->page = pages[seg_idx]->page; 883 if (add_persistent_gnt(blkif, 884 persistent_gnt)) { 885 kfree(persistent_gnt); 886 persistent_gnt = NULL; 887 goto next; 888 } 889 pages[seg_idx]->persistent_gnt = persistent_gnt; 890 pr_debug("grant %u added to the tree of persistent grants, using %u/%u\n", 891 persistent_gnt->gnt, blkif->persistent_gnt_c, 892 xen_blkif_max_pgrants); 893 goto next; 894 } 895 if (use_persistent_gnts && !blkif->vbd.overflow_max_grants) { 896 blkif->vbd.overflow_max_grants = 1; 897 pr_debug("domain %u, device %#x is using maximum number of persistent grants\n", 898 blkif->domid, blkif->vbd.handle); 899 } 900 /* 901 * We could not map this grant persistently, so use it as 902 * a non-persistent grant. 903 */ 904 next: 905 new_map_idx++; 906 } 907 segs_to_map = 0; 908 last_map = map_until; 909 if (map_until != num) 910 goto again; 911 912 return ret; 913 914 out_of_memory: 915 pr_alert("%s: out of memory\n", __func__); 916 put_free_pages(blkif, pages_to_gnt, segs_to_map); 917 return -ENOMEM; 918 } 919 920 static int xen_blkbk_map_seg(struct pending_req *pending_req) 921 { 922 int rc; 923 924 rc = xen_blkbk_map(pending_req->blkif, pending_req->segments, 925 pending_req->nr_segs, 926 (pending_req->operation != BLKIF_OP_READ)); 927 928 return rc; 929 } 930 931 static int xen_blkbk_parse_indirect(struct blkif_request *req, 932 struct pending_req *pending_req, 933 struct seg_buf seg[], 934 struct phys_req *preq) 935 { 936 struct grant_page **pages = pending_req->indirect_pages; 937 struct xen_blkif *blkif = pending_req->blkif; 938 int indirect_grefs, rc, n, nseg, i; 939 struct blkif_request_segment *segments = NULL; 940 941 nseg = pending_req->nr_segs; 942 indirect_grefs = INDIRECT_PAGES(nseg); 943 BUG_ON(indirect_grefs > BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST); 944 945 for (i = 0; i < indirect_grefs; i++) 946 pages[i]->gref = req->u.indirect.indirect_grefs[i]; 947 948 rc = xen_blkbk_map(blkif, pages, indirect_grefs, true); 949 if (rc) 950 goto unmap; 951 952 for (n = 0, i = 0; n < nseg; n++) { 953 if ((n % SEGS_PER_INDIRECT_FRAME) == 0) { 954 /* Map indirect segments */ 955 if (segments) 956 kunmap_atomic(segments); 957 segments = kmap_atomic(pages[n/SEGS_PER_INDIRECT_FRAME]->page); 958 } 959 i = n % SEGS_PER_INDIRECT_FRAME; 960 pending_req->segments[n]->gref = segments[i].gref; 961 seg[n].nsec = segments[i].last_sect - 962 segments[i].first_sect + 1; 963 seg[n].offset = (segments[i].first_sect << 9); 964 if ((segments[i].last_sect >= (XEN_PAGE_SIZE >> 9)) || 965 (segments[i].last_sect < segments[i].first_sect)) { 966 rc = -EINVAL; 967 goto unmap; 968 } 969 preq->nr_sects += seg[n].nsec; 970 } 971 972 unmap: 973 if (segments) 974 kunmap_atomic(segments); 975 xen_blkbk_unmap(blkif, pages, indirect_grefs); 976 return rc; 977 } 978 979 static int dispatch_discard_io(struct xen_blkif *blkif, 980 struct blkif_request *req) 981 { 982 int err = 0; 983 int status = BLKIF_RSP_OKAY; 984 struct block_device *bdev = blkif->vbd.bdev; 985 unsigned long secure; 986 struct phys_req preq; 987 988 xen_blkif_get(blkif); 989 990 preq.sector_number = req->u.discard.sector_number; 991 preq.nr_sects = req->u.discard.nr_sectors; 992 993 err = xen_vbd_translate(&preq, blkif, WRITE); 994 if (err) { 995 pr_warn("access denied: DISCARD [%llu->%llu] on dev=%04x\n", 996 preq.sector_number, 997 preq.sector_number + preq.nr_sects, blkif->vbd.pdevice); 998 goto fail_response; 999 } 1000 blkif->st_ds_req++; 1001 1002 secure = (blkif->vbd.discard_secure && 1003 (req->u.discard.flag & BLKIF_DISCARD_SECURE)) ? 1004 BLKDEV_DISCARD_SECURE : 0; 1005 1006 err = blkdev_issue_discard(bdev, req->u.discard.sector_number, 1007 req->u.discard.nr_sectors, 1008 GFP_KERNEL, secure); 1009 fail_response: 1010 if (err == -EOPNOTSUPP) { 1011 pr_debug("discard op failed, not supported\n"); 1012 status = BLKIF_RSP_EOPNOTSUPP; 1013 } else if (err) 1014 status = BLKIF_RSP_ERROR; 1015 1016 make_response(blkif, req->u.discard.id, req->operation, status); 1017 xen_blkif_put(blkif); 1018 return err; 1019 } 1020 1021 static int dispatch_other_io(struct xen_blkif *blkif, 1022 struct blkif_request *req, 1023 struct pending_req *pending_req) 1024 { 1025 free_req(blkif, pending_req); 1026 make_response(blkif, req->u.other.id, req->operation, 1027 BLKIF_RSP_EOPNOTSUPP); 1028 return -EIO; 1029 } 1030 1031 static void xen_blk_drain_io(struct xen_blkif *blkif) 1032 { 1033 atomic_set(&blkif->drain, 1); 1034 do { 1035 if (atomic_read(&blkif->inflight) == 0) 1036 break; 1037 wait_for_completion_interruptible_timeout( 1038 &blkif->drain_complete, HZ); 1039 1040 if (!atomic_read(&blkif->drain)) 1041 break; 1042 } while (!kthread_should_stop()); 1043 atomic_set(&blkif->drain, 0); 1044 } 1045 1046 /* 1047 * Completion callback on the bio's. Called as bh->b_end_io() 1048 */ 1049 1050 static void __end_block_io_op(struct pending_req *pending_req, int error) 1051 { 1052 /* An error fails the entire request. */ 1053 if ((pending_req->operation == BLKIF_OP_FLUSH_DISKCACHE) && 1054 (error == -EOPNOTSUPP)) { 1055 pr_debug("flush diskcache op failed, not supported\n"); 1056 xen_blkbk_flush_diskcache(XBT_NIL, pending_req->blkif->be, 0); 1057 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 1058 } else if ((pending_req->operation == BLKIF_OP_WRITE_BARRIER) && 1059 (error == -EOPNOTSUPP)) { 1060 pr_debug("write barrier op failed, not supported\n"); 1061 xen_blkbk_barrier(XBT_NIL, pending_req->blkif->be, 0); 1062 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 1063 } else if (error) { 1064 pr_debug("Buffer not up-to-date at end of operation," 1065 " error=%d\n", error); 1066 pending_req->status = BLKIF_RSP_ERROR; 1067 } 1068 1069 /* 1070 * If all of the bio's have completed it is time to unmap 1071 * the grant references associated with 'request' and provide 1072 * the proper response on the ring. 1073 */ 1074 if (atomic_dec_and_test(&pending_req->pendcnt)) 1075 xen_blkbk_unmap_and_respond(pending_req); 1076 } 1077 1078 /* 1079 * bio callback. 1080 */ 1081 static void end_block_io_op(struct bio *bio) 1082 { 1083 __end_block_io_op(bio->bi_private, bio->bi_error); 1084 bio_put(bio); 1085 } 1086 1087 1088 1089 /* 1090 * Function to copy the from the ring buffer the 'struct blkif_request' 1091 * (which has the sectors we want, number of them, grant references, etc), 1092 * and transmute it to the block API to hand it over to the proper block disk. 1093 */ 1094 static int 1095 __do_block_io_op(struct xen_blkif *blkif) 1096 { 1097 union blkif_back_rings *blk_rings = &blkif->blk_rings; 1098 struct blkif_request req; 1099 struct pending_req *pending_req; 1100 RING_IDX rc, rp; 1101 int more_to_do = 0; 1102 1103 rc = blk_rings->common.req_cons; 1104 rp = blk_rings->common.sring->req_prod; 1105 rmb(); /* Ensure we see queued requests up to 'rp'. */ 1106 1107 if (RING_REQUEST_PROD_OVERFLOW(&blk_rings->common, rp)) { 1108 rc = blk_rings->common.rsp_prod_pvt; 1109 pr_warn("Frontend provided bogus ring requests (%d - %d = %d). Halting ring processing on dev=%04x\n", 1110 rp, rc, rp - rc, blkif->vbd.pdevice); 1111 return -EACCES; 1112 } 1113 while (rc != rp) { 1114 1115 if (RING_REQUEST_CONS_OVERFLOW(&blk_rings->common, rc)) 1116 break; 1117 1118 if (kthread_should_stop()) { 1119 more_to_do = 1; 1120 break; 1121 } 1122 1123 pending_req = alloc_req(blkif); 1124 if (NULL == pending_req) { 1125 blkif->st_oo_req++; 1126 more_to_do = 1; 1127 break; 1128 } 1129 1130 switch (blkif->blk_protocol) { 1131 case BLKIF_PROTOCOL_NATIVE: 1132 memcpy(&req, RING_GET_REQUEST(&blk_rings->native, rc), sizeof(req)); 1133 break; 1134 case BLKIF_PROTOCOL_X86_32: 1135 blkif_get_x86_32_req(&req, RING_GET_REQUEST(&blk_rings->x86_32, rc)); 1136 break; 1137 case BLKIF_PROTOCOL_X86_64: 1138 blkif_get_x86_64_req(&req, RING_GET_REQUEST(&blk_rings->x86_64, rc)); 1139 break; 1140 default: 1141 BUG(); 1142 } 1143 blk_rings->common.req_cons = ++rc; /* before make_response() */ 1144 1145 /* Apply all sanity checks to /private copy/ of request. */ 1146 barrier(); 1147 1148 switch (req.operation) { 1149 case BLKIF_OP_READ: 1150 case BLKIF_OP_WRITE: 1151 case BLKIF_OP_WRITE_BARRIER: 1152 case BLKIF_OP_FLUSH_DISKCACHE: 1153 case BLKIF_OP_INDIRECT: 1154 if (dispatch_rw_block_io(blkif, &req, pending_req)) 1155 goto done; 1156 break; 1157 case BLKIF_OP_DISCARD: 1158 free_req(blkif, pending_req); 1159 if (dispatch_discard_io(blkif, &req)) 1160 goto done; 1161 break; 1162 default: 1163 if (dispatch_other_io(blkif, &req, pending_req)) 1164 goto done; 1165 break; 1166 } 1167 1168 /* Yield point for this unbounded loop. */ 1169 cond_resched(); 1170 } 1171 done: 1172 return more_to_do; 1173 } 1174 1175 static int 1176 do_block_io_op(struct xen_blkif *blkif) 1177 { 1178 union blkif_back_rings *blk_rings = &blkif->blk_rings; 1179 int more_to_do; 1180 1181 do { 1182 more_to_do = __do_block_io_op(blkif); 1183 if (more_to_do) 1184 break; 1185 1186 RING_FINAL_CHECK_FOR_REQUESTS(&blk_rings->common, more_to_do); 1187 } while (more_to_do); 1188 1189 return more_to_do; 1190 } 1191 /* 1192 * Transmutation of the 'struct blkif_request' to a proper 'struct bio' 1193 * and call the 'submit_bio' to pass it to the underlying storage. 1194 */ 1195 static int dispatch_rw_block_io(struct xen_blkif *blkif, 1196 struct blkif_request *req, 1197 struct pending_req *pending_req) 1198 { 1199 struct phys_req preq; 1200 struct seg_buf *seg = pending_req->seg; 1201 unsigned int nseg; 1202 struct bio *bio = NULL; 1203 struct bio **biolist = pending_req->biolist; 1204 int i, nbio = 0; 1205 int operation; 1206 struct blk_plug plug; 1207 bool drain = false; 1208 struct grant_page **pages = pending_req->segments; 1209 unsigned short req_operation; 1210 1211 req_operation = req->operation == BLKIF_OP_INDIRECT ? 1212 req->u.indirect.indirect_op : req->operation; 1213 1214 if ((req->operation == BLKIF_OP_INDIRECT) && 1215 (req_operation != BLKIF_OP_READ) && 1216 (req_operation != BLKIF_OP_WRITE)) { 1217 pr_debug("Invalid indirect operation (%u)\n", req_operation); 1218 goto fail_response; 1219 } 1220 1221 switch (req_operation) { 1222 case BLKIF_OP_READ: 1223 blkif->st_rd_req++; 1224 operation = READ; 1225 break; 1226 case BLKIF_OP_WRITE: 1227 blkif->st_wr_req++; 1228 operation = WRITE_ODIRECT; 1229 break; 1230 case BLKIF_OP_WRITE_BARRIER: 1231 drain = true; 1232 case BLKIF_OP_FLUSH_DISKCACHE: 1233 blkif->st_f_req++; 1234 operation = WRITE_FLUSH; 1235 break; 1236 default: 1237 operation = 0; /* make gcc happy */ 1238 goto fail_response; 1239 break; 1240 } 1241 1242 /* Check that the number of segments is sane. */ 1243 nseg = req->operation == BLKIF_OP_INDIRECT ? 1244 req->u.indirect.nr_segments : req->u.rw.nr_segments; 1245 1246 if (unlikely(nseg == 0 && operation != WRITE_FLUSH) || 1247 unlikely((req->operation != BLKIF_OP_INDIRECT) && 1248 (nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) || 1249 unlikely((req->operation == BLKIF_OP_INDIRECT) && 1250 (nseg > MAX_INDIRECT_SEGMENTS))) { 1251 pr_debug("Bad number of segments in request (%d)\n", nseg); 1252 /* Haven't submitted any bio's yet. */ 1253 goto fail_response; 1254 } 1255 1256 preq.nr_sects = 0; 1257 1258 pending_req->blkif = blkif; 1259 pending_req->id = req->u.rw.id; 1260 pending_req->operation = req_operation; 1261 pending_req->status = BLKIF_RSP_OKAY; 1262 pending_req->nr_segs = nseg; 1263 1264 if (req->operation != BLKIF_OP_INDIRECT) { 1265 preq.dev = req->u.rw.handle; 1266 preq.sector_number = req->u.rw.sector_number; 1267 for (i = 0; i < nseg; i++) { 1268 pages[i]->gref = req->u.rw.seg[i].gref; 1269 seg[i].nsec = req->u.rw.seg[i].last_sect - 1270 req->u.rw.seg[i].first_sect + 1; 1271 seg[i].offset = (req->u.rw.seg[i].first_sect << 9); 1272 if ((req->u.rw.seg[i].last_sect >= (XEN_PAGE_SIZE >> 9)) || 1273 (req->u.rw.seg[i].last_sect < 1274 req->u.rw.seg[i].first_sect)) 1275 goto fail_response; 1276 preq.nr_sects += seg[i].nsec; 1277 } 1278 } else { 1279 preq.dev = req->u.indirect.handle; 1280 preq.sector_number = req->u.indirect.sector_number; 1281 if (xen_blkbk_parse_indirect(req, pending_req, seg, &preq)) 1282 goto fail_response; 1283 } 1284 1285 if (xen_vbd_translate(&preq, blkif, operation) != 0) { 1286 pr_debug("access denied: %s of [%llu,%llu] on dev=%04x\n", 1287 operation == READ ? "read" : "write", 1288 preq.sector_number, 1289 preq.sector_number + preq.nr_sects, 1290 blkif->vbd.pdevice); 1291 goto fail_response; 1292 } 1293 1294 /* 1295 * This check _MUST_ be done after xen_vbd_translate as the preq.bdev 1296 * is set there. 1297 */ 1298 for (i = 0; i < nseg; i++) { 1299 if (((int)preq.sector_number|(int)seg[i].nsec) & 1300 ((bdev_logical_block_size(preq.bdev) >> 9) - 1)) { 1301 pr_debug("Misaligned I/O request from domain %d\n", 1302 blkif->domid); 1303 goto fail_response; 1304 } 1305 } 1306 1307 /* Wait on all outstanding I/O's and once that has been completed 1308 * issue the WRITE_FLUSH. 1309 */ 1310 if (drain) 1311 xen_blk_drain_io(pending_req->blkif); 1312 1313 /* 1314 * If we have failed at this point, we need to undo the M2P override, 1315 * set gnttab_set_unmap_op on all of the grant references and perform 1316 * the hypercall to unmap the grants - that is all done in 1317 * xen_blkbk_unmap. 1318 */ 1319 if (xen_blkbk_map_seg(pending_req)) 1320 goto fail_flush; 1321 1322 /* 1323 * This corresponding xen_blkif_put is done in __end_block_io_op, or 1324 * below (in "!bio") if we are handling a BLKIF_OP_DISCARD. 1325 */ 1326 xen_blkif_get(blkif); 1327 atomic_inc(&blkif->inflight); 1328 1329 for (i = 0; i < nseg; i++) { 1330 while ((bio == NULL) || 1331 (bio_add_page(bio, 1332 pages[i]->page, 1333 seg[i].nsec << 9, 1334 seg[i].offset) == 0)) { 1335 1336 int nr_iovecs = min_t(int, (nseg-i), BIO_MAX_PAGES); 1337 bio = bio_alloc(GFP_KERNEL, nr_iovecs); 1338 if (unlikely(bio == NULL)) 1339 goto fail_put_bio; 1340 1341 biolist[nbio++] = bio; 1342 bio->bi_bdev = preq.bdev; 1343 bio->bi_private = pending_req; 1344 bio->bi_end_io = end_block_io_op; 1345 bio->bi_iter.bi_sector = preq.sector_number; 1346 } 1347 1348 preq.sector_number += seg[i].nsec; 1349 } 1350 1351 /* This will be hit if the operation was a flush or discard. */ 1352 if (!bio) { 1353 BUG_ON(operation != WRITE_FLUSH); 1354 1355 bio = bio_alloc(GFP_KERNEL, 0); 1356 if (unlikely(bio == NULL)) 1357 goto fail_put_bio; 1358 1359 biolist[nbio++] = bio; 1360 bio->bi_bdev = preq.bdev; 1361 bio->bi_private = pending_req; 1362 bio->bi_end_io = end_block_io_op; 1363 } 1364 1365 atomic_set(&pending_req->pendcnt, nbio); 1366 blk_start_plug(&plug); 1367 1368 for (i = 0; i < nbio; i++) 1369 submit_bio(operation, biolist[i]); 1370 1371 /* Let the I/Os go.. */ 1372 blk_finish_plug(&plug); 1373 1374 if (operation == READ) 1375 blkif->st_rd_sect += preq.nr_sects; 1376 else if (operation & WRITE) 1377 blkif->st_wr_sect += preq.nr_sects; 1378 1379 return 0; 1380 1381 fail_flush: 1382 xen_blkbk_unmap(blkif, pending_req->segments, 1383 pending_req->nr_segs); 1384 fail_response: 1385 /* Haven't submitted any bio's yet. */ 1386 make_response(blkif, req->u.rw.id, req_operation, BLKIF_RSP_ERROR); 1387 free_req(blkif, pending_req); 1388 msleep(1); /* back off a bit */ 1389 return -EIO; 1390 1391 fail_put_bio: 1392 for (i = 0; i < nbio; i++) 1393 bio_put(biolist[i]); 1394 atomic_set(&pending_req->pendcnt, 1); 1395 __end_block_io_op(pending_req, -EINVAL); 1396 msleep(1); /* back off a bit */ 1397 return -EIO; 1398 } 1399 1400 1401 1402 /* 1403 * Put a response on the ring on how the operation fared. 1404 */ 1405 static void make_response(struct xen_blkif *blkif, u64 id, 1406 unsigned short op, int st) 1407 { 1408 struct blkif_response resp; 1409 unsigned long flags; 1410 union blkif_back_rings *blk_rings = &blkif->blk_rings; 1411 int notify; 1412 1413 resp.id = id; 1414 resp.operation = op; 1415 resp.status = st; 1416 1417 spin_lock_irqsave(&blkif->blk_ring_lock, flags); 1418 /* Place on the response ring for the relevant domain. */ 1419 switch (blkif->blk_protocol) { 1420 case BLKIF_PROTOCOL_NATIVE: 1421 memcpy(RING_GET_RESPONSE(&blk_rings->native, blk_rings->native.rsp_prod_pvt), 1422 &resp, sizeof(resp)); 1423 break; 1424 case BLKIF_PROTOCOL_X86_32: 1425 memcpy(RING_GET_RESPONSE(&blk_rings->x86_32, blk_rings->x86_32.rsp_prod_pvt), 1426 &resp, sizeof(resp)); 1427 break; 1428 case BLKIF_PROTOCOL_X86_64: 1429 memcpy(RING_GET_RESPONSE(&blk_rings->x86_64, blk_rings->x86_64.rsp_prod_pvt), 1430 &resp, sizeof(resp)); 1431 break; 1432 default: 1433 BUG(); 1434 } 1435 blk_rings->common.rsp_prod_pvt++; 1436 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&blk_rings->common, notify); 1437 spin_unlock_irqrestore(&blkif->blk_ring_lock, flags); 1438 if (notify) 1439 notify_remote_via_irq(blkif->irq); 1440 } 1441 1442 static int __init xen_blkif_init(void) 1443 { 1444 int rc = 0; 1445 1446 if (!xen_domain()) 1447 return -ENODEV; 1448 1449 if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { 1450 pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", 1451 xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); 1452 xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; 1453 } 1454 1455 rc = xen_blkif_interface_init(); 1456 if (rc) 1457 goto failed_init; 1458 1459 rc = xen_blkif_xenbus_init(); 1460 if (rc) 1461 goto failed_init; 1462 1463 failed_init: 1464 return rc; 1465 } 1466 1467 module_init(xen_blkif_init); 1468 1469 MODULE_LICENSE("Dual BSD/GPL"); 1470 MODULE_ALIAS("xen-backend:vbd"); 1471