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 number of rings/queues blkback supports, allow as many queues as there 88 * are CPUs if user has not specified a value. 89 */ 90 unsigned int xenblk_max_queues; 91 module_param_named(max_queues, xenblk_max_queues, uint, 0644); 92 MODULE_PARM_DESC(max_queues, 93 "Maximum number of hardware queues per virtual disk." \ 94 "By default it is the number of online CPUs."); 95 96 /* 97 * Maximum order of pages to be used for the shared ring between front and 98 * backend, 4KB page granularity is used. 99 */ 100 unsigned int xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; 101 module_param_named(max_ring_page_order, xen_blkif_max_ring_order, int, S_IRUGO); 102 MODULE_PARM_DESC(max_ring_page_order, "Maximum order of pages to be used for the shared ring"); 103 /* 104 * The LRU mechanism to clean the lists of persistent grants needs to 105 * be executed periodically. The time interval between consecutive executions 106 * of the purge mechanism is set in ms. 107 */ 108 #define LRU_INTERVAL 100 109 110 /* 111 * When the persistent grants list is full we will remove unused grants 112 * from the list. The percent number of grants to be removed at each LRU 113 * execution. 114 */ 115 #define LRU_PERCENT_CLEAN 5 116 117 /* Run-time switchable: /sys/module/blkback/parameters/ */ 118 static unsigned int log_stats; 119 module_param(log_stats, int, 0644); 120 121 #define BLKBACK_INVALID_HANDLE (~0) 122 123 /* Number of free pages to remove on each call to gnttab_free_pages */ 124 #define NUM_BATCH_FREE_PAGES 10 125 126 static inline int get_free_page(struct xen_blkif_ring *ring, struct page **page) 127 { 128 unsigned long flags; 129 130 spin_lock_irqsave(&ring->free_pages_lock, flags); 131 if (list_empty(&ring->free_pages)) { 132 BUG_ON(ring->free_pages_num != 0); 133 spin_unlock_irqrestore(&ring->free_pages_lock, flags); 134 return gnttab_alloc_pages(1, page); 135 } 136 BUG_ON(ring->free_pages_num == 0); 137 page[0] = list_first_entry(&ring->free_pages, struct page, lru); 138 list_del(&page[0]->lru); 139 ring->free_pages_num--; 140 spin_unlock_irqrestore(&ring->free_pages_lock, flags); 141 142 return 0; 143 } 144 145 static inline void put_free_pages(struct xen_blkif_ring *ring, struct page **page, 146 int num) 147 { 148 unsigned long flags; 149 int i; 150 151 spin_lock_irqsave(&ring->free_pages_lock, flags); 152 for (i = 0; i < num; i++) 153 list_add(&page[i]->lru, &ring->free_pages); 154 ring->free_pages_num += num; 155 spin_unlock_irqrestore(&ring->free_pages_lock, flags); 156 } 157 158 static inline void shrink_free_pagepool(struct xen_blkif_ring *ring, int num) 159 { 160 /* Remove requested pages in batches of NUM_BATCH_FREE_PAGES */ 161 struct page *page[NUM_BATCH_FREE_PAGES]; 162 unsigned int num_pages = 0; 163 unsigned long flags; 164 165 spin_lock_irqsave(&ring->free_pages_lock, flags); 166 while (ring->free_pages_num > num) { 167 BUG_ON(list_empty(&ring->free_pages)); 168 page[num_pages] = list_first_entry(&ring->free_pages, 169 struct page, lru); 170 list_del(&page[num_pages]->lru); 171 ring->free_pages_num--; 172 if (++num_pages == NUM_BATCH_FREE_PAGES) { 173 spin_unlock_irqrestore(&ring->free_pages_lock, flags); 174 gnttab_free_pages(num_pages, page); 175 spin_lock_irqsave(&ring->free_pages_lock, flags); 176 num_pages = 0; 177 } 178 } 179 spin_unlock_irqrestore(&ring->free_pages_lock, flags); 180 if (num_pages != 0) 181 gnttab_free_pages(num_pages, page); 182 } 183 184 #define vaddr(page) ((unsigned long)pfn_to_kaddr(page_to_pfn(page))) 185 186 static int do_block_io_op(struct xen_blkif_ring *ring); 187 static int dispatch_rw_block_io(struct xen_blkif_ring *ring, 188 struct blkif_request *req, 189 struct pending_req *pending_req); 190 static void make_response(struct xen_blkif_ring *ring, u64 id, 191 unsigned short op, int st); 192 193 #define foreach_grant_safe(pos, n, rbtree, node) \ 194 for ((pos) = container_of(rb_first((rbtree)), typeof(*(pos)), node), \ 195 (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL; \ 196 &(pos)->node != NULL; \ 197 (pos) = container_of(n, typeof(*(pos)), node), \ 198 (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL) 199 200 201 /* 202 * We don't need locking around the persistent grant helpers 203 * because blkback uses a single-thread for each backend, so we 204 * can be sure that this functions will never be called recursively. 205 * 206 * The only exception to that is put_persistent_grant, that can be called 207 * from interrupt context (by xen_blkbk_unmap), so we have to use atomic 208 * bit operations to modify the flags of a persistent grant and to count 209 * the number of used grants. 210 */ 211 static int add_persistent_gnt(struct xen_blkif_ring *ring, 212 struct persistent_gnt *persistent_gnt) 213 { 214 struct rb_node **new = NULL, *parent = NULL; 215 struct persistent_gnt *this; 216 struct xen_blkif *blkif = ring->blkif; 217 218 if (ring->persistent_gnt_c >= xen_blkif_max_pgrants) { 219 if (!blkif->vbd.overflow_max_grants) 220 blkif->vbd.overflow_max_grants = 1; 221 return -EBUSY; 222 } 223 /* Figure out where to put new node */ 224 new = &ring->persistent_gnts.rb_node; 225 while (*new) { 226 this = container_of(*new, struct persistent_gnt, node); 227 228 parent = *new; 229 if (persistent_gnt->gnt < this->gnt) 230 new = &((*new)->rb_left); 231 else if (persistent_gnt->gnt > this->gnt) 232 new = &((*new)->rb_right); 233 else { 234 pr_alert_ratelimited("trying to add a gref that's already in the tree\n"); 235 return -EINVAL; 236 } 237 } 238 239 bitmap_zero(persistent_gnt->flags, PERSISTENT_GNT_FLAGS_SIZE); 240 set_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags); 241 /* Add new node and rebalance tree. */ 242 rb_link_node(&(persistent_gnt->node), parent, new); 243 rb_insert_color(&(persistent_gnt->node), &ring->persistent_gnts); 244 ring->persistent_gnt_c++; 245 atomic_inc(&ring->persistent_gnt_in_use); 246 return 0; 247 } 248 249 static struct persistent_gnt *get_persistent_gnt(struct xen_blkif_ring *ring, 250 grant_ref_t gref) 251 { 252 struct persistent_gnt *data; 253 struct rb_node *node = NULL; 254 255 node = ring->persistent_gnts.rb_node; 256 while (node) { 257 data = container_of(node, struct persistent_gnt, node); 258 259 if (gref < data->gnt) 260 node = node->rb_left; 261 else if (gref > data->gnt) 262 node = node->rb_right; 263 else { 264 if(test_bit(PERSISTENT_GNT_ACTIVE, data->flags)) { 265 pr_alert_ratelimited("requesting a grant already in use\n"); 266 return NULL; 267 } 268 set_bit(PERSISTENT_GNT_ACTIVE, data->flags); 269 atomic_inc(&ring->persistent_gnt_in_use); 270 return data; 271 } 272 } 273 return NULL; 274 } 275 276 static void put_persistent_gnt(struct xen_blkif_ring *ring, 277 struct persistent_gnt *persistent_gnt) 278 { 279 if(!test_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags)) 280 pr_alert_ratelimited("freeing a grant already unused\n"); 281 set_bit(PERSISTENT_GNT_WAS_ACTIVE, persistent_gnt->flags); 282 clear_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags); 283 atomic_dec(&ring->persistent_gnt_in_use); 284 } 285 286 static void free_persistent_gnts(struct xen_blkif_ring *ring, struct rb_root *root, 287 unsigned int num) 288 { 289 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 290 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 291 struct persistent_gnt *persistent_gnt; 292 struct rb_node *n; 293 int segs_to_unmap = 0; 294 struct gntab_unmap_queue_data unmap_data; 295 296 unmap_data.pages = pages; 297 unmap_data.unmap_ops = unmap; 298 unmap_data.kunmap_ops = NULL; 299 300 foreach_grant_safe(persistent_gnt, n, root, node) { 301 BUG_ON(persistent_gnt->handle == 302 BLKBACK_INVALID_HANDLE); 303 gnttab_set_unmap_op(&unmap[segs_to_unmap], 304 (unsigned long) pfn_to_kaddr(page_to_pfn( 305 persistent_gnt->page)), 306 GNTMAP_host_map, 307 persistent_gnt->handle); 308 309 pages[segs_to_unmap] = persistent_gnt->page; 310 311 if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST || 312 !rb_next(&persistent_gnt->node)) { 313 314 unmap_data.count = segs_to_unmap; 315 BUG_ON(gnttab_unmap_refs_sync(&unmap_data)); 316 317 put_free_pages(ring, pages, segs_to_unmap); 318 segs_to_unmap = 0; 319 } 320 321 rb_erase(&persistent_gnt->node, root); 322 kfree(persistent_gnt); 323 num--; 324 } 325 BUG_ON(num != 0); 326 } 327 328 void xen_blkbk_unmap_purged_grants(struct work_struct *work) 329 { 330 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 331 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 332 struct persistent_gnt *persistent_gnt; 333 int segs_to_unmap = 0; 334 struct xen_blkif_ring *ring = container_of(work, typeof(*ring), persistent_purge_work); 335 struct gntab_unmap_queue_data unmap_data; 336 337 unmap_data.pages = pages; 338 unmap_data.unmap_ops = unmap; 339 unmap_data.kunmap_ops = NULL; 340 341 while(!list_empty(&ring->persistent_purge_list)) { 342 persistent_gnt = list_first_entry(&ring->persistent_purge_list, 343 struct persistent_gnt, 344 remove_node); 345 list_del(&persistent_gnt->remove_node); 346 347 gnttab_set_unmap_op(&unmap[segs_to_unmap], 348 vaddr(persistent_gnt->page), 349 GNTMAP_host_map, 350 persistent_gnt->handle); 351 352 pages[segs_to_unmap] = persistent_gnt->page; 353 354 if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST) { 355 unmap_data.count = segs_to_unmap; 356 BUG_ON(gnttab_unmap_refs_sync(&unmap_data)); 357 put_free_pages(ring, pages, segs_to_unmap); 358 segs_to_unmap = 0; 359 } 360 kfree(persistent_gnt); 361 } 362 if (segs_to_unmap > 0) { 363 unmap_data.count = segs_to_unmap; 364 BUG_ON(gnttab_unmap_refs_sync(&unmap_data)); 365 put_free_pages(ring, pages, segs_to_unmap); 366 } 367 } 368 369 static void purge_persistent_gnt(struct xen_blkif_ring *ring) 370 { 371 struct persistent_gnt *persistent_gnt; 372 struct rb_node *n; 373 unsigned int num_clean, total; 374 bool scan_used = false, clean_used = false; 375 struct rb_root *root; 376 377 if (ring->persistent_gnt_c < xen_blkif_max_pgrants || 378 (ring->persistent_gnt_c == xen_blkif_max_pgrants && 379 !ring->blkif->vbd.overflow_max_grants)) { 380 goto out; 381 } 382 383 if (work_busy(&ring->persistent_purge_work)) { 384 pr_alert_ratelimited("Scheduled work from previous purge is still busy, cannot purge list\n"); 385 goto out; 386 } 387 388 num_clean = (xen_blkif_max_pgrants / 100) * LRU_PERCENT_CLEAN; 389 num_clean = ring->persistent_gnt_c - xen_blkif_max_pgrants + num_clean; 390 num_clean = min(ring->persistent_gnt_c, num_clean); 391 if ((num_clean == 0) || 392 (num_clean > (ring->persistent_gnt_c - atomic_read(&ring->persistent_gnt_in_use)))) 393 goto out; 394 395 /* 396 * At this point, we can assure that there will be no calls 397 * to get_persistent_grant (because we are executing this code from 398 * xen_blkif_schedule), there can only be calls to put_persistent_gnt, 399 * which means that the number of currently used grants will go down, 400 * but never up, so we will always be able to remove the requested 401 * number of grants. 402 */ 403 404 total = num_clean; 405 406 pr_debug("Going to purge %u persistent grants\n", num_clean); 407 408 BUG_ON(!list_empty(&ring->persistent_purge_list)); 409 root = &ring->persistent_gnts; 410 purge_list: 411 foreach_grant_safe(persistent_gnt, n, root, node) { 412 BUG_ON(persistent_gnt->handle == 413 BLKBACK_INVALID_HANDLE); 414 415 if (clean_used) { 416 clear_bit(PERSISTENT_GNT_WAS_ACTIVE, persistent_gnt->flags); 417 continue; 418 } 419 420 if (test_bit(PERSISTENT_GNT_ACTIVE, persistent_gnt->flags)) 421 continue; 422 if (!scan_used && 423 (test_bit(PERSISTENT_GNT_WAS_ACTIVE, persistent_gnt->flags))) 424 continue; 425 426 rb_erase(&persistent_gnt->node, root); 427 list_add(&persistent_gnt->remove_node, 428 &ring->persistent_purge_list); 429 if (--num_clean == 0) 430 goto finished; 431 } 432 /* 433 * If we get here it means we also need to start cleaning 434 * grants that were used since last purge in order to cope 435 * with the requested num 436 */ 437 if (!scan_used && !clean_used) { 438 pr_debug("Still missing %u purged frames\n", num_clean); 439 scan_used = true; 440 goto purge_list; 441 } 442 finished: 443 if (!clean_used) { 444 pr_debug("Finished scanning for grants to clean, removing used flag\n"); 445 clean_used = true; 446 goto purge_list; 447 } 448 449 ring->persistent_gnt_c -= (total - num_clean); 450 ring->blkif->vbd.overflow_max_grants = 0; 451 452 /* We can defer this work */ 453 schedule_work(&ring->persistent_purge_work); 454 pr_debug("Purged %u/%u\n", (total - num_clean), total); 455 456 out: 457 return; 458 } 459 460 /* 461 * Retrieve from the 'pending_reqs' a free pending_req structure to be used. 462 */ 463 static struct pending_req *alloc_req(struct xen_blkif_ring *ring) 464 { 465 struct pending_req *req = NULL; 466 unsigned long flags; 467 468 spin_lock_irqsave(&ring->pending_free_lock, flags); 469 if (!list_empty(&ring->pending_free)) { 470 req = list_entry(ring->pending_free.next, struct pending_req, 471 free_list); 472 list_del(&req->free_list); 473 } 474 spin_unlock_irqrestore(&ring->pending_free_lock, flags); 475 return req; 476 } 477 478 /* 479 * Return the 'pending_req' structure back to the freepool. We also 480 * wake up the thread if it was waiting for a free page. 481 */ 482 static void free_req(struct xen_blkif_ring *ring, struct pending_req *req) 483 { 484 unsigned long flags; 485 int was_empty; 486 487 spin_lock_irqsave(&ring->pending_free_lock, flags); 488 was_empty = list_empty(&ring->pending_free); 489 list_add(&req->free_list, &ring->pending_free); 490 spin_unlock_irqrestore(&ring->pending_free_lock, flags); 491 if (was_empty) 492 wake_up(&ring->pending_free_wq); 493 } 494 495 /* 496 * Routines for managing virtual block devices (vbds). 497 */ 498 static int xen_vbd_translate(struct phys_req *req, struct xen_blkif *blkif, 499 int operation) 500 { 501 struct xen_vbd *vbd = &blkif->vbd; 502 int rc = -EACCES; 503 504 if ((operation != READ) && vbd->readonly) 505 goto out; 506 507 if (likely(req->nr_sects)) { 508 blkif_sector_t end = req->sector_number + req->nr_sects; 509 510 if (unlikely(end < req->sector_number)) 511 goto out; 512 if (unlikely(end > vbd_sz(vbd))) 513 goto out; 514 } 515 516 req->dev = vbd->pdevice; 517 req->bdev = vbd->bdev; 518 rc = 0; 519 520 out: 521 return rc; 522 } 523 524 static void xen_vbd_resize(struct xen_blkif *blkif) 525 { 526 struct xen_vbd *vbd = &blkif->vbd; 527 struct xenbus_transaction xbt; 528 int err; 529 struct xenbus_device *dev = xen_blkbk_xenbus(blkif->be); 530 unsigned long long new_size = vbd_sz(vbd); 531 532 pr_info("VBD Resize: Domid: %d, Device: (%d, %d)\n", 533 blkif->domid, MAJOR(vbd->pdevice), MINOR(vbd->pdevice)); 534 pr_info("VBD Resize: new size %llu\n", new_size); 535 vbd->size = new_size; 536 again: 537 err = xenbus_transaction_start(&xbt); 538 if (err) { 539 pr_warn("Error starting transaction\n"); 540 return; 541 } 542 err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu", 543 (unsigned long long)vbd_sz(vbd)); 544 if (err) { 545 pr_warn("Error writing new size\n"); 546 goto abort; 547 } 548 /* 549 * Write the current state; we will use this to synchronize 550 * the front-end. If the current state is "connected" the 551 * front-end will get the new size information online. 552 */ 553 err = xenbus_printf(xbt, dev->nodename, "state", "%d", dev->state); 554 if (err) { 555 pr_warn("Error writing the state\n"); 556 goto abort; 557 } 558 559 err = xenbus_transaction_end(xbt, 0); 560 if (err == -EAGAIN) 561 goto again; 562 if (err) 563 pr_warn("Error ending transaction\n"); 564 return; 565 abort: 566 xenbus_transaction_end(xbt, 1); 567 } 568 569 /* 570 * Notification from the guest OS. 571 */ 572 static void blkif_notify_work(struct xen_blkif_ring *ring) 573 { 574 ring->waiting_reqs = 1; 575 wake_up(&ring->wq); 576 } 577 578 irqreturn_t xen_blkif_be_int(int irq, void *dev_id) 579 { 580 blkif_notify_work(dev_id); 581 return IRQ_HANDLED; 582 } 583 584 /* 585 * SCHEDULER FUNCTIONS 586 */ 587 588 static void print_stats(struct xen_blkif_ring *ring) 589 { 590 pr_info("(%s): oo %3llu | rd %4llu | wr %4llu | f %4llu" 591 " | ds %4llu | pg: %4u/%4d\n", 592 current->comm, ring->st_oo_req, 593 ring->st_rd_req, ring->st_wr_req, 594 ring->st_f_req, ring->st_ds_req, 595 ring->persistent_gnt_c, 596 xen_blkif_max_pgrants); 597 ring->st_print = jiffies + msecs_to_jiffies(10 * 1000); 598 ring->st_rd_req = 0; 599 ring->st_wr_req = 0; 600 ring->st_oo_req = 0; 601 ring->st_ds_req = 0; 602 } 603 604 int xen_blkif_schedule(void *arg) 605 { 606 struct xen_blkif_ring *ring = arg; 607 struct xen_blkif *blkif = ring->blkif; 608 struct xen_vbd *vbd = &blkif->vbd; 609 unsigned long timeout; 610 int ret; 611 612 xen_blkif_get(blkif); 613 614 set_freezable(); 615 while (!kthread_should_stop()) { 616 if (try_to_freeze()) 617 continue; 618 if (unlikely(vbd->size != vbd_sz(vbd))) 619 xen_vbd_resize(blkif); 620 621 timeout = msecs_to_jiffies(LRU_INTERVAL); 622 623 timeout = wait_event_interruptible_timeout( 624 ring->wq, 625 ring->waiting_reqs || kthread_should_stop(), 626 timeout); 627 if (timeout == 0) 628 goto purge_gnt_list; 629 timeout = wait_event_interruptible_timeout( 630 ring->pending_free_wq, 631 !list_empty(&ring->pending_free) || 632 kthread_should_stop(), 633 timeout); 634 if (timeout == 0) 635 goto purge_gnt_list; 636 637 ring->waiting_reqs = 0; 638 smp_mb(); /* clear flag *before* checking for work */ 639 640 ret = do_block_io_op(ring); 641 if (ret > 0) 642 ring->waiting_reqs = 1; 643 if (ret == -EACCES) 644 wait_event_interruptible(ring->shutdown_wq, 645 kthread_should_stop()); 646 647 purge_gnt_list: 648 if (blkif->vbd.feature_gnt_persistent && 649 time_after(jiffies, ring->next_lru)) { 650 purge_persistent_gnt(ring); 651 ring->next_lru = jiffies + msecs_to_jiffies(LRU_INTERVAL); 652 } 653 654 /* Shrink if we have more than xen_blkif_max_buffer_pages */ 655 shrink_free_pagepool(ring, xen_blkif_max_buffer_pages); 656 657 if (log_stats && time_after(jiffies, ring->st_print)) 658 print_stats(ring); 659 } 660 661 /* Drain pending purge work */ 662 flush_work(&ring->persistent_purge_work); 663 664 if (log_stats) 665 print_stats(ring); 666 667 ring->xenblkd = NULL; 668 xen_blkif_put(blkif); 669 670 return 0; 671 } 672 673 /* 674 * Remove persistent grants and empty the pool of free pages 675 */ 676 void xen_blkbk_free_caches(struct xen_blkif_ring *ring) 677 { 678 /* Free all persistent grant pages */ 679 if (!RB_EMPTY_ROOT(&ring->persistent_gnts)) 680 free_persistent_gnts(ring, &ring->persistent_gnts, 681 ring->persistent_gnt_c); 682 683 BUG_ON(!RB_EMPTY_ROOT(&ring->persistent_gnts)); 684 ring->persistent_gnt_c = 0; 685 686 /* Since we are shutting down remove all pages from the buffer */ 687 shrink_free_pagepool(ring, 0 /* All */); 688 } 689 690 static unsigned int xen_blkbk_unmap_prepare( 691 struct xen_blkif_ring *ring, 692 struct grant_page **pages, 693 unsigned int num, 694 struct gnttab_unmap_grant_ref *unmap_ops, 695 struct page **unmap_pages) 696 { 697 unsigned int i, invcount = 0; 698 699 for (i = 0; i < num; i++) { 700 if (pages[i]->persistent_gnt != NULL) { 701 put_persistent_gnt(ring, pages[i]->persistent_gnt); 702 continue; 703 } 704 if (pages[i]->handle == BLKBACK_INVALID_HANDLE) 705 continue; 706 unmap_pages[invcount] = pages[i]->page; 707 gnttab_set_unmap_op(&unmap_ops[invcount], vaddr(pages[i]->page), 708 GNTMAP_host_map, pages[i]->handle); 709 pages[i]->handle = BLKBACK_INVALID_HANDLE; 710 invcount++; 711 } 712 713 return invcount; 714 } 715 716 static void xen_blkbk_unmap_and_respond_callback(int result, struct gntab_unmap_queue_data *data) 717 { 718 struct pending_req *pending_req = (struct pending_req *)(data->data); 719 struct xen_blkif_ring *ring = pending_req->ring; 720 struct xen_blkif *blkif = ring->blkif; 721 722 /* BUG_ON used to reproduce existing behaviour, 723 but is this the best way to deal with this? */ 724 BUG_ON(result); 725 726 put_free_pages(ring, data->pages, data->count); 727 make_response(ring, pending_req->id, 728 pending_req->operation, pending_req->status); 729 free_req(ring, pending_req); 730 /* 731 * Make sure the request is freed before releasing blkif, 732 * or there could be a race between free_req and the 733 * cleanup done in xen_blkif_free during shutdown. 734 * 735 * NB: The fact that we might try to wake up pending_free_wq 736 * before drain_complete (in case there's a drain going on) 737 * it's not a problem with our current implementation 738 * because we can assure there's no thread waiting on 739 * pending_free_wq if there's a drain going on, but it has 740 * to be taken into account if the current model is changed. 741 */ 742 if (atomic_dec_and_test(&ring->inflight) && atomic_read(&blkif->drain)) { 743 complete(&blkif->drain_complete); 744 } 745 xen_blkif_put(blkif); 746 } 747 748 static void xen_blkbk_unmap_and_respond(struct pending_req *req) 749 { 750 struct gntab_unmap_queue_data* work = &req->gnttab_unmap_data; 751 struct xen_blkif_ring *ring = req->ring; 752 struct grant_page **pages = req->segments; 753 unsigned int invcount; 754 755 invcount = xen_blkbk_unmap_prepare(ring, pages, req->nr_segs, 756 req->unmap, req->unmap_pages); 757 758 work->data = req; 759 work->done = xen_blkbk_unmap_and_respond_callback; 760 work->unmap_ops = req->unmap; 761 work->kunmap_ops = NULL; 762 work->pages = req->unmap_pages; 763 work->count = invcount; 764 765 gnttab_unmap_refs_async(&req->gnttab_unmap_data); 766 } 767 768 769 /* 770 * Unmap the grant references. 771 * 772 * This could accumulate ops up to the batch size to reduce the number 773 * of hypercalls, but since this is only used in error paths there's 774 * no real need. 775 */ 776 static void xen_blkbk_unmap(struct xen_blkif_ring *ring, 777 struct grant_page *pages[], 778 int num) 779 { 780 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 781 struct page *unmap_pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 782 unsigned int invcount = 0; 783 int ret; 784 785 while (num) { 786 unsigned int batch = min(num, BLKIF_MAX_SEGMENTS_PER_REQUEST); 787 788 invcount = xen_blkbk_unmap_prepare(ring, pages, batch, 789 unmap, unmap_pages); 790 if (invcount) { 791 ret = gnttab_unmap_refs(unmap, NULL, unmap_pages, invcount); 792 BUG_ON(ret); 793 put_free_pages(ring, unmap_pages, invcount); 794 } 795 pages += batch; 796 num -= batch; 797 } 798 } 799 800 static int xen_blkbk_map(struct xen_blkif_ring *ring, 801 struct grant_page *pages[], 802 int num, bool ro) 803 { 804 struct gnttab_map_grant_ref map[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 805 struct page *pages_to_gnt[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 806 struct persistent_gnt *persistent_gnt = NULL; 807 phys_addr_t addr = 0; 808 int i, seg_idx, new_map_idx; 809 int segs_to_map = 0; 810 int ret = 0; 811 int last_map = 0, map_until = 0; 812 int use_persistent_gnts; 813 struct xen_blkif *blkif = ring->blkif; 814 815 use_persistent_gnts = (blkif->vbd.feature_gnt_persistent); 816 817 /* 818 * Fill out preq.nr_sects with proper amount of sectors, and setup 819 * assign map[..] with the PFN of the page in our domain with the 820 * corresponding grant reference for each page. 821 */ 822 again: 823 for (i = map_until; i < num; i++) { 824 uint32_t flags; 825 826 if (use_persistent_gnts) { 827 persistent_gnt = get_persistent_gnt( 828 ring, 829 pages[i]->gref); 830 } 831 832 if (persistent_gnt) { 833 /* 834 * We are using persistent grants and 835 * the grant is already mapped 836 */ 837 pages[i]->page = persistent_gnt->page; 838 pages[i]->persistent_gnt = persistent_gnt; 839 } else { 840 if (get_free_page(ring, &pages[i]->page)) 841 goto out_of_memory; 842 addr = vaddr(pages[i]->page); 843 pages_to_gnt[segs_to_map] = pages[i]->page; 844 pages[i]->persistent_gnt = NULL; 845 flags = GNTMAP_host_map; 846 if (!use_persistent_gnts && ro) 847 flags |= GNTMAP_readonly; 848 gnttab_set_map_op(&map[segs_to_map++], addr, 849 flags, pages[i]->gref, 850 blkif->domid); 851 } 852 map_until = i + 1; 853 if (segs_to_map == BLKIF_MAX_SEGMENTS_PER_REQUEST) 854 break; 855 } 856 857 if (segs_to_map) { 858 ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map); 859 BUG_ON(ret); 860 } 861 862 /* 863 * Now swizzle the MFN in our domain with the MFN from the other domain 864 * so that when we access vaddr(pending_req,i) it has the contents of 865 * the page from the other domain. 866 */ 867 for (seg_idx = last_map, new_map_idx = 0; seg_idx < map_until; seg_idx++) { 868 if (!pages[seg_idx]->persistent_gnt) { 869 /* This is a newly mapped grant */ 870 BUG_ON(new_map_idx >= segs_to_map); 871 if (unlikely(map[new_map_idx].status != 0)) { 872 pr_debug("invalid buffer -- could not remap it\n"); 873 put_free_pages(ring, &pages[seg_idx]->page, 1); 874 pages[seg_idx]->handle = BLKBACK_INVALID_HANDLE; 875 ret |= 1; 876 goto next; 877 } 878 pages[seg_idx]->handle = map[new_map_idx].handle; 879 } else { 880 continue; 881 } 882 if (use_persistent_gnts && 883 ring->persistent_gnt_c < xen_blkif_max_pgrants) { 884 /* 885 * We are using persistent grants, the grant is 886 * not mapped but we might have room for it. 887 */ 888 persistent_gnt = kmalloc(sizeof(struct persistent_gnt), 889 GFP_KERNEL); 890 if (!persistent_gnt) { 891 /* 892 * If we don't have enough memory to 893 * allocate the persistent_gnt struct 894 * map this grant non-persistenly 895 */ 896 goto next; 897 } 898 persistent_gnt->gnt = map[new_map_idx].ref; 899 persistent_gnt->handle = map[new_map_idx].handle; 900 persistent_gnt->page = pages[seg_idx]->page; 901 if (add_persistent_gnt(ring, 902 persistent_gnt)) { 903 kfree(persistent_gnt); 904 persistent_gnt = NULL; 905 goto next; 906 } 907 pages[seg_idx]->persistent_gnt = persistent_gnt; 908 pr_debug("grant %u added to the tree of persistent grants, using %u/%u\n", 909 persistent_gnt->gnt, ring->persistent_gnt_c, 910 xen_blkif_max_pgrants); 911 goto next; 912 } 913 if (use_persistent_gnts && !blkif->vbd.overflow_max_grants) { 914 blkif->vbd.overflow_max_grants = 1; 915 pr_debug("domain %u, device %#x is using maximum number of persistent grants\n", 916 blkif->domid, blkif->vbd.handle); 917 } 918 /* 919 * We could not map this grant persistently, so use it as 920 * a non-persistent grant. 921 */ 922 next: 923 new_map_idx++; 924 } 925 segs_to_map = 0; 926 last_map = map_until; 927 if (map_until != num) 928 goto again; 929 930 return ret; 931 932 out_of_memory: 933 pr_alert("%s: out of memory\n", __func__); 934 put_free_pages(ring, pages_to_gnt, segs_to_map); 935 return -ENOMEM; 936 } 937 938 static int xen_blkbk_map_seg(struct pending_req *pending_req) 939 { 940 int rc; 941 942 rc = xen_blkbk_map(pending_req->ring, pending_req->segments, 943 pending_req->nr_segs, 944 (pending_req->operation != BLKIF_OP_READ)); 945 946 return rc; 947 } 948 949 static int xen_blkbk_parse_indirect(struct blkif_request *req, 950 struct pending_req *pending_req, 951 struct seg_buf seg[], 952 struct phys_req *preq) 953 { 954 struct grant_page **pages = pending_req->indirect_pages; 955 struct xen_blkif_ring *ring = pending_req->ring; 956 int indirect_grefs, rc, n, nseg, i; 957 struct blkif_request_segment *segments = NULL; 958 959 nseg = pending_req->nr_segs; 960 indirect_grefs = INDIRECT_PAGES(nseg); 961 BUG_ON(indirect_grefs > BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST); 962 963 for (i = 0; i < indirect_grefs; i++) 964 pages[i]->gref = req->u.indirect.indirect_grefs[i]; 965 966 rc = xen_blkbk_map(ring, pages, indirect_grefs, true); 967 if (rc) 968 goto unmap; 969 970 for (n = 0, i = 0; n < nseg; n++) { 971 uint8_t first_sect, last_sect; 972 973 if ((n % SEGS_PER_INDIRECT_FRAME) == 0) { 974 /* Map indirect segments */ 975 if (segments) 976 kunmap_atomic(segments); 977 segments = kmap_atomic(pages[n/SEGS_PER_INDIRECT_FRAME]->page); 978 } 979 i = n % SEGS_PER_INDIRECT_FRAME; 980 981 pending_req->segments[n]->gref = segments[i].gref; 982 983 first_sect = READ_ONCE(segments[i].first_sect); 984 last_sect = READ_ONCE(segments[i].last_sect); 985 if (last_sect >= (XEN_PAGE_SIZE >> 9) || last_sect < first_sect) { 986 rc = -EINVAL; 987 goto unmap; 988 } 989 990 seg[n].nsec = last_sect - first_sect + 1; 991 seg[n].offset = first_sect << 9; 992 preq->nr_sects += seg[n].nsec; 993 } 994 995 unmap: 996 if (segments) 997 kunmap_atomic(segments); 998 xen_blkbk_unmap(ring, pages, indirect_grefs); 999 return rc; 1000 } 1001 1002 static int dispatch_discard_io(struct xen_blkif_ring *ring, 1003 struct blkif_request *req) 1004 { 1005 int err = 0; 1006 int status = BLKIF_RSP_OKAY; 1007 struct xen_blkif *blkif = ring->blkif; 1008 struct block_device *bdev = blkif->vbd.bdev; 1009 unsigned long secure; 1010 struct phys_req preq; 1011 1012 xen_blkif_get(blkif); 1013 1014 preq.sector_number = req->u.discard.sector_number; 1015 preq.nr_sects = req->u.discard.nr_sectors; 1016 1017 err = xen_vbd_translate(&preq, blkif, WRITE); 1018 if (err) { 1019 pr_warn("access denied: DISCARD [%llu->%llu] on dev=%04x\n", 1020 preq.sector_number, 1021 preq.sector_number + preq.nr_sects, blkif->vbd.pdevice); 1022 goto fail_response; 1023 } 1024 ring->st_ds_req++; 1025 1026 secure = (blkif->vbd.discard_secure && 1027 (req->u.discard.flag & BLKIF_DISCARD_SECURE)) ? 1028 BLKDEV_DISCARD_SECURE : 0; 1029 1030 err = blkdev_issue_discard(bdev, req->u.discard.sector_number, 1031 req->u.discard.nr_sectors, 1032 GFP_KERNEL, secure); 1033 fail_response: 1034 if (err == -EOPNOTSUPP) { 1035 pr_debug("discard op failed, not supported\n"); 1036 status = BLKIF_RSP_EOPNOTSUPP; 1037 } else if (err) 1038 status = BLKIF_RSP_ERROR; 1039 1040 make_response(ring, req->u.discard.id, req->operation, status); 1041 xen_blkif_put(blkif); 1042 return err; 1043 } 1044 1045 static int dispatch_other_io(struct xen_blkif_ring *ring, 1046 struct blkif_request *req, 1047 struct pending_req *pending_req) 1048 { 1049 free_req(ring, pending_req); 1050 make_response(ring, req->u.other.id, req->operation, 1051 BLKIF_RSP_EOPNOTSUPP); 1052 return -EIO; 1053 } 1054 1055 static void xen_blk_drain_io(struct xen_blkif_ring *ring) 1056 { 1057 struct xen_blkif *blkif = ring->blkif; 1058 1059 atomic_set(&blkif->drain, 1); 1060 do { 1061 if (atomic_read(&ring->inflight) == 0) 1062 break; 1063 wait_for_completion_interruptible_timeout( 1064 &blkif->drain_complete, HZ); 1065 1066 if (!atomic_read(&blkif->drain)) 1067 break; 1068 } while (!kthread_should_stop()); 1069 atomic_set(&blkif->drain, 0); 1070 } 1071 1072 /* 1073 * Completion callback on the bio's. Called as bh->b_end_io() 1074 */ 1075 1076 static void __end_block_io_op(struct pending_req *pending_req, int error) 1077 { 1078 /* An error fails the entire request. */ 1079 if ((pending_req->operation == BLKIF_OP_FLUSH_DISKCACHE) && 1080 (error == -EOPNOTSUPP)) { 1081 pr_debug("flush diskcache op failed, not supported\n"); 1082 xen_blkbk_flush_diskcache(XBT_NIL, pending_req->ring->blkif->be, 0); 1083 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 1084 } else if ((pending_req->operation == BLKIF_OP_WRITE_BARRIER) && 1085 (error == -EOPNOTSUPP)) { 1086 pr_debug("write barrier op failed, not supported\n"); 1087 xen_blkbk_barrier(XBT_NIL, pending_req->ring->blkif->be, 0); 1088 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 1089 } else if (error) { 1090 pr_debug("Buffer not up-to-date at end of operation," 1091 " error=%d\n", error); 1092 pending_req->status = BLKIF_RSP_ERROR; 1093 } 1094 1095 /* 1096 * If all of the bio's have completed it is time to unmap 1097 * the grant references associated with 'request' and provide 1098 * the proper response on the ring. 1099 */ 1100 if (atomic_dec_and_test(&pending_req->pendcnt)) 1101 xen_blkbk_unmap_and_respond(pending_req); 1102 } 1103 1104 /* 1105 * bio callback. 1106 */ 1107 static void end_block_io_op(struct bio *bio) 1108 { 1109 __end_block_io_op(bio->bi_private, bio->bi_error); 1110 bio_put(bio); 1111 } 1112 1113 1114 1115 /* 1116 * Function to copy the from the ring buffer the 'struct blkif_request' 1117 * (which has the sectors we want, number of them, grant references, etc), 1118 * and transmute it to the block API to hand it over to the proper block disk. 1119 */ 1120 static int 1121 __do_block_io_op(struct xen_blkif_ring *ring) 1122 { 1123 union blkif_back_rings *blk_rings = &ring->blk_rings; 1124 struct blkif_request req; 1125 struct pending_req *pending_req; 1126 RING_IDX rc, rp; 1127 int more_to_do = 0; 1128 1129 rc = blk_rings->common.req_cons; 1130 rp = blk_rings->common.sring->req_prod; 1131 rmb(); /* Ensure we see queued requests up to 'rp'. */ 1132 1133 if (RING_REQUEST_PROD_OVERFLOW(&blk_rings->common, rp)) { 1134 rc = blk_rings->common.rsp_prod_pvt; 1135 pr_warn("Frontend provided bogus ring requests (%d - %d = %d). Halting ring processing on dev=%04x\n", 1136 rp, rc, rp - rc, ring->blkif->vbd.pdevice); 1137 return -EACCES; 1138 } 1139 while (rc != rp) { 1140 1141 if (RING_REQUEST_CONS_OVERFLOW(&blk_rings->common, rc)) 1142 break; 1143 1144 if (kthread_should_stop()) { 1145 more_to_do = 1; 1146 break; 1147 } 1148 1149 pending_req = alloc_req(ring); 1150 if (NULL == pending_req) { 1151 ring->st_oo_req++; 1152 more_to_do = 1; 1153 break; 1154 } 1155 1156 switch (ring->blkif->blk_protocol) { 1157 case BLKIF_PROTOCOL_NATIVE: 1158 memcpy(&req, RING_GET_REQUEST(&blk_rings->native, rc), sizeof(req)); 1159 break; 1160 case BLKIF_PROTOCOL_X86_32: 1161 blkif_get_x86_32_req(&req, RING_GET_REQUEST(&blk_rings->x86_32, rc)); 1162 break; 1163 case BLKIF_PROTOCOL_X86_64: 1164 blkif_get_x86_64_req(&req, RING_GET_REQUEST(&blk_rings->x86_64, rc)); 1165 break; 1166 default: 1167 BUG(); 1168 } 1169 blk_rings->common.req_cons = ++rc; /* before make_response() */ 1170 1171 /* Apply all sanity checks to /private copy/ of request. */ 1172 barrier(); 1173 1174 switch (req.operation) { 1175 case BLKIF_OP_READ: 1176 case BLKIF_OP_WRITE: 1177 case BLKIF_OP_WRITE_BARRIER: 1178 case BLKIF_OP_FLUSH_DISKCACHE: 1179 case BLKIF_OP_INDIRECT: 1180 if (dispatch_rw_block_io(ring, &req, pending_req)) 1181 goto done; 1182 break; 1183 case BLKIF_OP_DISCARD: 1184 free_req(ring, pending_req); 1185 if (dispatch_discard_io(ring, &req)) 1186 goto done; 1187 break; 1188 default: 1189 if (dispatch_other_io(ring, &req, pending_req)) 1190 goto done; 1191 break; 1192 } 1193 1194 /* Yield point for this unbounded loop. */ 1195 cond_resched(); 1196 } 1197 done: 1198 return more_to_do; 1199 } 1200 1201 static int 1202 do_block_io_op(struct xen_blkif_ring *ring) 1203 { 1204 union blkif_back_rings *blk_rings = &ring->blk_rings; 1205 int more_to_do; 1206 1207 do { 1208 more_to_do = __do_block_io_op(ring); 1209 if (more_to_do) 1210 break; 1211 1212 RING_FINAL_CHECK_FOR_REQUESTS(&blk_rings->common, more_to_do); 1213 } while (more_to_do); 1214 1215 return more_to_do; 1216 } 1217 /* 1218 * Transmutation of the 'struct blkif_request' to a proper 'struct bio' 1219 * and call the 'submit_bio' to pass it to the underlying storage. 1220 */ 1221 static int dispatch_rw_block_io(struct xen_blkif_ring *ring, 1222 struct blkif_request *req, 1223 struct pending_req *pending_req) 1224 { 1225 struct phys_req preq; 1226 struct seg_buf *seg = pending_req->seg; 1227 unsigned int nseg; 1228 struct bio *bio = NULL; 1229 struct bio **biolist = pending_req->biolist; 1230 int i, nbio = 0; 1231 int operation; 1232 struct blk_plug plug; 1233 bool drain = false; 1234 struct grant_page **pages = pending_req->segments; 1235 unsigned short req_operation; 1236 1237 req_operation = req->operation == BLKIF_OP_INDIRECT ? 1238 req->u.indirect.indirect_op : req->operation; 1239 1240 if ((req->operation == BLKIF_OP_INDIRECT) && 1241 (req_operation != BLKIF_OP_READ) && 1242 (req_operation != BLKIF_OP_WRITE)) { 1243 pr_debug("Invalid indirect operation (%u)\n", req_operation); 1244 goto fail_response; 1245 } 1246 1247 switch (req_operation) { 1248 case BLKIF_OP_READ: 1249 ring->st_rd_req++; 1250 operation = READ; 1251 break; 1252 case BLKIF_OP_WRITE: 1253 ring->st_wr_req++; 1254 operation = WRITE_ODIRECT; 1255 break; 1256 case BLKIF_OP_WRITE_BARRIER: 1257 drain = true; 1258 case BLKIF_OP_FLUSH_DISKCACHE: 1259 ring->st_f_req++; 1260 operation = WRITE_FLUSH; 1261 break; 1262 default: 1263 operation = 0; /* make gcc happy */ 1264 goto fail_response; 1265 break; 1266 } 1267 1268 /* Check that the number of segments is sane. */ 1269 nseg = req->operation == BLKIF_OP_INDIRECT ? 1270 req->u.indirect.nr_segments : req->u.rw.nr_segments; 1271 1272 if (unlikely(nseg == 0 && operation != WRITE_FLUSH) || 1273 unlikely((req->operation != BLKIF_OP_INDIRECT) && 1274 (nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) || 1275 unlikely((req->operation == BLKIF_OP_INDIRECT) && 1276 (nseg > MAX_INDIRECT_SEGMENTS))) { 1277 pr_debug("Bad number of segments in request (%d)\n", nseg); 1278 /* Haven't submitted any bio's yet. */ 1279 goto fail_response; 1280 } 1281 1282 preq.nr_sects = 0; 1283 1284 pending_req->ring = ring; 1285 pending_req->id = req->u.rw.id; 1286 pending_req->operation = req_operation; 1287 pending_req->status = BLKIF_RSP_OKAY; 1288 pending_req->nr_segs = nseg; 1289 1290 if (req->operation != BLKIF_OP_INDIRECT) { 1291 preq.dev = req->u.rw.handle; 1292 preq.sector_number = req->u.rw.sector_number; 1293 for (i = 0; i < nseg; i++) { 1294 pages[i]->gref = req->u.rw.seg[i].gref; 1295 seg[i].nsec = req->u.rw.seg[i].last_sect - 1296 req->u.rw.seg[i].first_sect + 1; 1297 seg[i].offset = (req->u.rw.seg[i].first_sect << 9); 1298 if ((req->u.rw.seg[i].last_sect >= (XEN_PAGE_SIZE >> 9)) || 1299 (req->u.rw.seg[i].last_sect < 1300 req->u.rw.seg[i].first_sect)) 1301 goto fail_response; 1302 preq.nr_sects += seg[i].nsec; 1303 } 1304 } else { 1305 preq.dev = req->u.indirect.handle; 1306 preq.sector_number = req->u.indirect.sector_number; 1307 if (xen_blkbk_parse_indirect(req, pending_req, seg, &preq)) 1308 goto fail_response; 1309 } 1310 1311 if (xen_vbd_translate(&preq, ring->blkif, operation) != 0) { 1312 pr_debug("access denied: %s of [%llu,%llu] on dev=%04x\n", 1313 operation == READ ? "read" : "write", 1314 preq.sector_number, 1315 preq.sector_number + preq.nr_sects, 1316 ring->blkif->vbd.pdevice); 1317 goto fail_response; 1318 } 1319 1320 /* 1321 * This check _MUST_ be done after xen_vbd_translate as the preq.bdev 1322 * is set there. 1323 */ 1324 for (i = 0; i < nseg; i++) { 1325 if (((int)preq.sector_number|(int)seg[i].nsec) & 1326 ((bdev_logical_block_size(preq.bdev) >> 9) - 1)) { 1327 pr_debug("Misaligned I/O request from domain %d\n", 1328 ring->blkif->domid); 1329 goto fail_response; 1330 } 1331 } 1332 1333 /* Wait on all outstanding I/O's and once that has been completed 1334 * issue the WRITE_FLUSH. 1335 */ 1336 if (drain) 1337 xen_blk_drain_io(pending_req->ring); 1338 1339 /* 1340 * If we have failed at this point, we need to undo the M2P override, 1341 * set gnttab_set_unmap_op on all of the grant references and perform 1342 * the hypercall to unmap the grants - that is all done in 1343 * xen_blkbk_unmap. 1344 */ 1345 if (xen_blkbk_map_seg(pending_req)) 1346 goto fail_flush; 1347 1348 /* 1349 * This corresponding xen_blkif_put is done in __end_block_io_op, or 1350 * below (in "!bio") if we are handling a BLKIF_OP_DISCARD. 1351 */ 1352 xen_blkif_get(ring->blkif); 1353 atomic_inc(&ring->inflight); 1354 1355 for (i = 0; i < nseg; i++) { 1356 while ((bio == NULL) || 1357 (bio_add_page(bio, 1358 pages[i]->page, 1359 seg[i].nsec << 9, 1360 seg[i].offset) == 0)) { 1361 1362 int nr_iovecs = min_t(int, (nseg-i), BIO_MAX_PAGES); 1363 bio = bio_alloc(GFP_KERNEL, nr_iovecs); 1364 if (unlikely(bio == NULL)) 1365 goto fail_put_bio; 1366 1367 biolist[nbio++] = bio; 1368 bio->bi_bdev = preq.bdev; 1369 bio->bi_private = pending_req; 1370 bio->bi_end_io = end_block_io_op; 1371 bio->bi_iter.bi_sector = preq.sector_number; 1372 } 1373 1374 preq.sector_number += seg[i].nsec; 1375 } 1376 1377 /* This will be hit if the operation was a flush or discard. */ 1378 if (!bio) { 1379 BUG_ON(operation != WRITE_FLUSH); 1380 1381 bio = bio_alloc(GFP_KERNEL, 0); 1382 if (unlikely(bio == NULL)) 1383 goto fail_put_bio; 1384 1385 biolist[nbio++] = bio; 1386 bio->bi_bdev = preq.bdev; 1387 bio->bi_private = pending_req; 1388 bio->bi_end_io = end_block_io_op; 1389 } 1390 1391 atomic_set(&pending_req->pendcnt, nbio); 1392 blk_start_plug(&plug); 1393 1394 for (i = 0; i < nbio; i++) 1395 submit_bio(operation, biolist[i]); 1396 1397 /* Let the I/Os go.. */ 1398 blk_finish_plug(&plug); 1399 1400 if (operation == READ) 1401 ring->st_rd_sect += preq.nr_sects; 1402 else if (operation & WRITE) 1403 ring->st_wr_sect += preq.nr_sects; 1404 1405 return 0; 1406 1407 fail_flush: 1408 xen_blkbk_unmap(ring, pending_req->segments, 1409 pending_req->nr_segs); 1410 fail_response: 1411 /* Haven't submitted any bio's yet. */ 1412 make_response(ring, req->u.rw.id, req_operation, BLKIF_RSP_ERROR); 1413 free_req(ring, pending_req); 1414 msleep(1); /* back off a bit */ 1415 return -EIO; 1416 1417 fail_put_bio: 1418 for (i = 0; i < nbio; i++) 1419 bio_put(biolist[i]); 1420 atomic_set(&pending_req->pendcnt, 1); 1421 __end_block_io_op(pending_req, -EINVAL); 1422 msleep(1); /* back off a bit */ 1423 return -EIO; 1424 } 1425 1426 1427 1428 /* 1429 * Put a response on the ring on how the operation fared. 1430 */ 1431 static void make_response(struct xen_blkif_ring *ring, u64 id, 1432 unsigned short op, int st) 1433 { 1434 struct blkif_response resp; 1435 unsigned long flags; 1436 union blkif_back_rings *blk_rings; 1437 int notify; 1438 1439 resp.id = id; 1440 resp.operation = op; 1441 resp.status = st; 1442 1443 spin_lock_irqsave(&ring->blk_ring_lock, flags); 1444 blk_rings = &ring->blk_rings; 1445 /* Place on the response ring for the relevant domain. */ 1446 switch (ring->blkif->blk_protocol) { 1447 case BLKIF_PROTOCOL_NATIVE: 1448 memcpy(RING_GET_RESPONSE(&blk_rings->native, blk_rings->native.rsp_prod_pvt), 1449 &resp, sizeof(resp)); 1450 break; 1451 case BLKIF_PROTOCOL_X86_32: 1452 memcpy(RING_GET_RESPONSE(&blk_rings->x86_32, blk_rings->x86_32.rsp_prod_pvt), 1453 &resp, sizeof(resp)); 1454 break; 1455 case BLKIF_PROTOCOL_X86_64: 1456 memcpy(RING_GET_RESPONSE(&blk_rings->x86_64, blk_rings->x86_64.rsp_prod_pvt), 1457 &resp, sizeof(resp)); 1458 break; 1459 default: 1460 BUG(); 1461 } 1462 blk_rings->common.rsp_prod_pvt++; 1463 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&blk_rings->common, notify); 1464 spin_unlock_irqrestore(&ring->blk_ring_lock, flags); 1465 if (notify) 1466 notify_remote_via_irq(ring->irq); 1467 } 1468 1469 static int __init xen_blkif_init(void) 1470 { 1471 int rc = 0; 1472 1473 if (!xen_domain()) 1474 return -ENODEV; 1475 1476 if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { 1477 pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", 1478 xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); 1479 xen_blkif_max_ring_order = XENBUS_MAX_RING_GRANT_ORDER; 1480 } 1481 1482 if (xenblk_max_queues == 0) 1483 xenblk_max_queues = num_online_cpus(); 1484 1485 rc = xen_blkif_interface_init(); 1486 if (rc) 1487 goto failed_init; 1488 1489 rc = xen_blkif_xenbus_init(); 1490 if (rc) 1491 goto failed_init; 1492 1493 failed_init: 1494 return rc; 1495 } 1496 1497 module_init(xen_blkif_init); 1498 1499 MODULE_LICENSE("Dual BSD/GPL"); 1500 MODULE_ALIAS("xen-backend:vbd"); 1501