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 #include <linux/spinlock.h> 38 #include <linux/kthread.h> 39 #include <linux/list.h> 40 #include <linux/delay.h> 41 #include <linux/freezer.h> 42 #include <linux/bitmap.h> 43 44 #include <xen/events.h> 45 #include <xen/page.h> 46 #include <xen/xen.h> 47 #include <asm/xen/hypervisor.h> 48 #include <asm/xen/hypercall.h> 49 #include <xen/balloon.h> 50 #include "common.h" 51 52 /* 53 * These are rather arbitrary. They are fairly large because adjacent requests 54 * pulled from a communication ring are quite likely to end up being part of 55 * the same scatter/gather request at the disc. 56 * 57 * ** TRY INCREASING 'xen_blkif_reqs' IF WRITE SPEEDS SEEM TOO LOW ** 58 * 59 * This will increase the chances of being able to write whole tracks. 60 * 64 should be enough to keep us competitive with Linux. 61 */ 62 static int xen_blkif_reqs = 64; 63 module_param_named(reqs, xen_blkif_reqs, int, 0); 64 MODULE_PARM_DESC(reqs, "Number of blkback requests to allocate"); 65 66 /* Run-time switchable: /sys/module/blkback/parameters/ */ 67 static unsigned int log_stats; 68 module_param(log_stats, int, 0644); 69 70 /* 71 * Each outstanding request that we've passed to the lower device layers has a 72 * 'pending_req' allocated to it. Each buffer_head that completes decrements 73 * the pendcnt towards zero. When it hits zero, the specified domain has a 74 * response queued for it, with the saved 'id' passed back. 75 */ 76 struct pending_req { 77 struct xen_blkif *blkif; 78 u64 id; 79 int nr_pages; 80 atomic_t pendcnt; 81 unsigned short operation; 82 int status; 83 struct list_head free_list; 84 DECLARE_BITMAP(unmap_seg, BLKIF_MAX_SEGMENTS_PER_REQUEST); 85 }; 86 87 #define BLKBACK_INVALID_HANDLE (~0) 88 89 struct xen_blkbk { 90 struct pending_req *pending_reqs; 91 /* List of all 'pending_req' available */ 92 struct list_head pending_free; 93 /* And its spinlock. */ 94 spinlock_t pending_free_lock; 95 wait_queue_head_t pending_free_wq; 96 /* The list of all pages that are available. */ 97 struct page **pending_pages; 98 /* And the grant handles that are available. */ 99 grant_handle_t *pending_grant_handles; 100 }; 101 102 static struct xen_blkbk *blkbk; 103 104 /* 105 * Maximum number of grant pages that can be mapped in blkback. 106 * BLKIF_MAX_SEGMENTS_PER_REQUEST * RING_SIZE is the maximum number of 107 * pages that blkback will persistently map. 108 * Currently, this is: 109 * RING_SIZE = 32 (for all known ring types) 110 * BLKIF_MAX_SEGMENTS_PER_REQUEST = 11 111 * sizeof(struct persistent_gnt) = 48 112 * So the maximum memory used to store the grants is: 113 * 32 * 11 * 48 = 16896 bytes 114 */ 115 static inline unsigned int max_mapped_grant_pages(enum blkif_protocol protocol) 116 { 117 switch (protocol) { 118 case BLKIF_PROTOCOL_NATIVE: 119 return __CONST_RING_SIZE(blkif, PAGE_SIZE) * 120 BLKIF_MAX_SEGMENTS_PER_REQUEST; 121 case BLKIF_PROTOCOL_X86_32: 122 return __CONST_RING_SIZE(blkif_x86_32, PAGE_SIZE) * 123 BLKIF_MAX_SEGMENTS_PER_REQUEST; 124 case BLKIF_PROTOCOL_X86_64: 125 return __CONST_RING_SIZE(blkif_x86_64, PAGE_SIZE) * 126 BLKIF_MAX_SEGMENTS_PER_REQUEST; 127 default: 128 BUG(); 129 } 130 return 0; 131 } 132 133 134 /* 135 * Little helpful macro to figure out the index and virtual address of the 136 * pending_pages[..]. For each 'pending_req' we have have up to 137 * BLKIF_MAX_SEGMENTS_PER_REQUEST (11) pages. The seg would be from 0 through 138 * 10 and would index in the pending_pages[..]. 139 */ 140 static inline int vaddr_pagenr(struct pending_req *req, int seg) 141 { 142 return (req - blkbk->pending_reqs) * 143 BLKIF_MAX_SEGMENTS_PER_REQUEST + seg; 144 } 145 146 #define pending_page(req, seg) pending_pages[vaddr_pagenr(req, seg)] 147 148 static inline unsigned long vaddr(struct pending_req *req, int seg) 149 { 150 unsigned long pfn = page_to_pfn(blkbk->pending_page(req, seg)); 151 return (unsigned long)pfn_to_kaddr(pfn); 152 } 153 154 #define pending_handle(_req, _seg) \ 155 (blkbk->pending_grant_handles[vaddr_pagenr(_req, _seg)]) 156 157 158 static int do_block_io_op(struct xen_blkif *blkif); 159 static int dispatch_rw_block_io(struct xen_blkif *blkif, 160 struct blkif_request *req, 161 struct pending_req *pending_req); 162 static void make_response(struct xen_blkif *blkif, u64 id, 163 unsigned short op, int st); 164 165 #define foreach_grant_safe(pos, n, rbtree, node) \ 166 for ((pos) = container_of(rb_first((rbtree)), typeof(*(pos)), node), \ 167 (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL; \ 168 &(pos)->node != NULL; \ 169 (pos) = container_of(n, typeof(*(pos)), node), \ 170 (n) = (&(pos)->node != NULL) ? rb_next(&(pos)->node) : NULL) 171 172 173 static void add_persistent_gnt(struct rb_root *root, 174 struct persistent_gnt *persistent_gnt) 175 { 176 struct rb_node **new = &(root->rb_node), *parent = NULL; 177 struct persistent_gnt *this; 178 179 /* Figure out where to put new node */ 180 while (*new) { 181 this = container_of(*new, struct persistent_gnt, node); 182 183 parent = *new; 184 if (persistent_gnt->gnt < this->gnt) 185 new = &((*new)->rb_left); 186 else if (persistent_gnt->gnt > this->gnt) 187 new = &((*new)->rb_right); 188 else { 189 pr_alert(DRV_PFX " trying to add a gref that's already in the tree\n"); 190 BUG(); 191 } 192 } 193 194 /* Add new node and rebalance tree. */ 195 rb_link_node(&(persistent_gnt->node), parent, new); 196 rb_insert_color(&(persistent_gnt->node), root); 197 } 198 199 static struct persistent_gnt *get_persistent_gnt(struct rb_root *root, 200 grant_ref_t gref) 201 { 202 struct persistent_gnt *data; 203 struct rb_node *node = root->rb_node; 204 205 while (node) { 206 data = container_of(node, struct persistent_gnt, node); 207 208 if (gref < data->gnt) 209 node = node->rb_left; 210 else if (gref > data->gnt) 211 node = node->rb_right; 212 else 213 return data; 214 } 215 return NULL; 216 } 217 218 static void free_persistent_gnts(struct rb_root *root, unsigned int num) 219 { 220 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 221 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 222 struct persistent_gnt *persistent_gnt; 223 struct rb_node *n; 224 int ret = 0; 225 int segs_to_unmap = 0; 226 227 foreach_grant_safe(persistent_gnt, n, root, node) { 228 BUG_ON(persistent_gnt->handle == 229 BLKBACK_INVALID_HANDLE); 230 gnttab_set_unmap_op(&unmap[segs_to_unmap], 231 (unsigned long) pfn_to_kaddr(page_to_pfn( 232 persistent_gnt->page)), 233 GNTMAP_host_map, 234 persistent_gnt->handle); 235 236 pages[segs_to_unmap] = persistent_gnt->page; 237 238 if (++segs_to_unmap == BLKIF_MAX_SEGMENTS_PER_REQUEST || 239 !rb_next(&persistent_gnt->node)) { 240 ret = gnttab_unmap_refs(unmap, NULL, pages, 241 segs_to_unmap); 242 BUG_ON(ret); 243 free_xenballooned_pages(segs_to_unmap, pages); 244 segs_to_unmap = 0; 245 } 246 247 rb_erase(&persistent_gnt->node, root); 248 kfree(persistent_gnt); 249 num--; 250 } 251 BUG_ON(num != 0); 252 } 253 254 /* 255 * Retrieve from the 'pending_reqs' a free pending_req structure to be used. 256 */ 257 static struct pending_req *alloc_req(void) 258 { 259 struct pending_req *req = NULL; 260 unsigned long flags; 261 262 spin_lock_irqsave(&blkbk->pending_free_lock, flags); 263 if (!list_empty(&blkbk->pending_free)) { 264 req = list_entry(blkbk->pending_free.next, struct pending_req, 265 free_list); 266 list_del(&req->free_list); 267 } 268 spin_unlock_irqrestore(&blkbk->pending_free_lock, flags); 269 return req; 270 } 271 272 /* 273 * Return the 'pending_req' structure back to the freepool. We also 274 * wake up the thread if it was waiting for a free page. 275 */ 276 static void free_req(struct pending_req *req) 277 { 278 unsigned long flags; 279 int was_empty; 280 281 spin_lock_irqsave(&blkbk->pending_free_lock, flags); 282 was_empty = list_empty(&blkbk->pending_free); 283 list_add(&req->free_list, &blkbk->pending_free); 284 spin_unlock_irqrestore(&blkbk->pending_free_lock, flags); 285 if (was_empty) 286 wake_up(&blkbk->pending_free_wq); 287 } 288 289 /* 290 * Routines for managing virtual block devices (vbds). 291 */ 292 static int xen_vbd_translate(struct phys_req *req, struct xen_blkif *blkif, 293 int operation) 294 { 295 struct xen_vbd *vbd = &blkif->vbd; 296 int rc = -EACCES; 297 298 if ((operation != READ) && vbd->readonly) 299 goto out; 300 301 if (likely(req->nr_sects)) { 302 blkif_sector_t end = req->sector_number + req->nr_sects; 303 304 if (unlikely(end < req->sector_number)) 305 goto out; 306 if (unlikely(end > vbd_sz(vbd))) 307 goto out; 308 } 309 310 req->dev = vbd->pdevice; 311 req->bdev = vbd->bdev; 312 rc = 0; 313 314 out: 315 return rc; 316 } 317 318 static void xen_vbd_resize(struct xen_blkif *blkif) 319 { 320 struct xen_vbd *vbd = &blkif->vbd; 321 struct xenbus_transaction xbt; 322 int err; 323 struct xenbus_device *dev = xen_blkbk_xenbus(blkif->be); 324 unsigned long long new_size = vbd_sz(vbd); 325 326 pr_info(DRV_PFX "VBD Resize: Domid: %d, Device: (%d, %d)\n", 327 blkif->domid, MAJOR(vbd->pdevice), MINOR(vbd->pdevice)); 328 pr_info(DRV_PFX "VBD Resize: new size %llu\n", new_size); 329 vbd->size = new_size; 330 again: 331 err = xenbus_transaction_start(&xbt); 332 if (err) { 333 pr_warn(DRV_PFX "Error starting transaction"); 334 return; 335 } 336 err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu", 337 (unsigned long long)vbd_sz(vbd)); 338 if (err) { 339 pr_warn(DRV_PFX "Error writing new size"); 340 goto abort; 341 } 342 /* 343 * Write the current state; we will use this to synchronize 344 * the front-end. If the current state is "connected" the 345 * front-end will get the new size information online. 346 */ 347 err = xenbus_printf(xbt, dev->nodename, "state", "%d", dev->state); 348 if (err) { 349 pr_warn(DRV_PFX "Error writing the state"); 350 goto abort; 351 } 352 353 err = xenbus_transaction_end(xbt, 0); 354 if (err == -EAGAIN) 355 goto again; 356 if (err) 357 pr_warn(DRV_PFX "Error ending transaction"); 358 return; 359 abort: 360 xenbus_transaction_end(xbt, 1); 361 } 362 363 /* 364 * Notification from the guest OS. 365 */ 366 static void blkif_notify_work(struct xen_blkif *blkif) 367 { 368 blkif->waiting_reqs = 1; 369 wake_up(&blkif->wq); 370 } 371 372 irqreturn_t xen_blkif_be_int(int irq, void *dev_id) 373 { 374 blkif_notify_work(dev_id); 375 return IRQ_HANDLED; 376 } 377 378 /* 379 * SCHEDULER FUNCTIONS 380 */ 381 382 static void print_stats(struct xen_blkif *blkif) 383 { 384 pr_info("xen-blkback (%s): oo %3llu | rd %4llu | wr %4llu | f %4llu" 385 " | ds %4llu\n", 386 current->comm, blkif->st_oo_req, 387 blkif->st_rd_req, blkif->st_wr_req, 388 blkif->st_f_req, blkif->st_ds_req); 389 blkif->st_print = jiffies + msecs_to_jiffies(10 * 1000); 390 blkif->st_rd_req = 0; 391 blkif->st_wr_req = 0; 392 blkif->st_oo_req = 0; 393 blkif->st_ds_req = 0; 394 } 395 396 int xen_blkif_schedule(void *arg) 397 { 398 struct xen_blkif *blkif = arg; 399 struct xen_vbd *vbd = &blkif->vbd; 400 401 xen_blkif_get(blkif); 402 403 while (!kthread_should_stop()) { 404 if (try_to_freeze()) 405 continue; 406 if (unlikely(vbd->size != vbd_sz(vbd))) 407 xen_vbd_resize(blkif); 408 409 wait_event_interruptible( 410 blkif->wq, 411 blkif->waiting_reqs || kthread_should_stop()); 412 wait_event_interruptible( 413 blkbk->pending_free_wq, 414 !list_empty(&blkbk->pending_free) || 415 kthread_should_stop()); 416 417 blkif->waiting_reqs = 0; 418 smp_mb(); /* clear flag *before* checking for work */ 419 420 if (do_block_io_op(blkif)) 421 blkif->waiting_reqs = 1; 422 423 if (log_stats && time_after(jiffies, blkif->st_print)) 424 print_stats(blkif); 425 } 426 427 /* Free all persistent grant pages */ 428 if (!RB_EMPTY_ROOT(&blkif->persistent_gnts)) 429 free_persistent_gnts(&blkif->persistent_gnts, 430 blkif->persistent_gnt_c); 431 432 BUG_ON(!RB_EMPTY_ROOT(&blkif->persistent_gnts)); 433 blkif->persistent_gnt_c = 0; 434 435 if (log_stats) 436 print_stats(blkif); 437 438 blkif->xenblkd = NULL; 439 xen_blkif_put(blkif); 440 441 return 0; 442 } 443 444 struct seg_buf { 445 unsigned int offset; 446 unsigned int nsec; 447 }; 448 /* 449 * Unmap the grant references, and also remove the M2P over-rides 450 * used in the 'pending_req'. 451 */ 452 static void xen_blkbk_unmap(struct pending_req *req) 453 { 454 struct gnttab_unmap_grant_ref unmap[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 455 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 456 unsigned int i, invcount = 0; 457 grant_handle_t handle; 458 int ret; 459 460 for (i = 0; i < req->nr_pages; i++) { 461 if (!test_bit(i, req->unmap_seg)) 462 continue; 463 handle = pending_handle(req, i); 464 if (handle == BLKBACK_INVALID_HANDLE) 465 continue; 466 gnttab_set_unmap_op(&unmap[invcount], vaddr(req, i), 467 GNTMAP_host_map, handle); 468 pending_handle(req, i) = BLKBACK_INVALID_HANDLE; 469 pages[invcount] = virt_to_page(vaddr(req, i)); 470 invcount++; 471 } 472 473 ret = gnttab_unmap_refs(unmap, NULL, pages, invcount); 474 BUG_ON(ret); 475 } 476 477 static int xen_blkbk_map(struct blkif_request *req, 478 struct pending_req *pending_req, 479 struct seg_buf seg[], 480 struct page *pages[]) 481 { 482 struct gnttab_map_grant_ref map[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 483 struct persistent_gnt *persistent_gnts[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 484 struct page *pages_to_gnt[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 485 struct persistent_gnt *persistent_gnt = NULL; 486 struct xen_blkif *blkif = pending_req->blkif; 487 phys_addr_t addr = 0; 488 int i, j; 489 bool new_map; 490 int nseg = req->u.rw.nr_segments; 491 int segs_to_map = 0; 492 int ret = 0; 493 int use_persistent_gnts; 494 495 use_persistent_gnts = (blkif->vbd.feature_gnt_persistent); 496 497 BUG_ON(blkif->persistent_gnt_c > 498 max_mapped_grant_pages(pending_req->blkif->blk_protocol)); 499 500 /* 501 * Fill out preq.nr_sects with proper amount of sectors, and setup 502 * assign map[..] with the PFN of the page in our domain with the 503 * corresponding grant reference for each page. 504 */ 505 for (i = 0; i < nseg; i++) { 506 uint32_t flags; 507 508 if (use_persistent_gnts) 509 persistent_gnt = get_persistent_gnt( 510 &blkif->persistent_gnts, 511 req->u.rw.seg[i].gref); 512 513 if (persistent_gnt) { 514 /* 515 * We are using persistent grants and 516 * the grant is already mapped 517 */ 518 new_map = false; 519 } else if (use_persistent_gnts && 520 blkif->persistent_gnt_c < 521 max_mapped_grant_pages(blkif->blk_protocol)) { 522 /* 523 * We are using persistent grants, the grant is 524 * not mapped but we have room for it 525 */ 526 new_map = true; 527 persistent_gnt = kmalloc( 528 sizeof(struct persistent_gnt), 529 GFP_KERNEL); 530 if (!persistent_gnt) 531 return -ENOMEM; 532 if (alloc_xenballooned_pages(1, &persistent_gnt->page, 533 false)) { 534 kfree(persistent_gnt); 535 return -ENOMEM; 536 } 537 persistent_gnt->gnt = req->u.rw.seg[i].gref; 538 persistent_gnt->handle = BLKBACK_INVALID_HANDLE; 539 540 pages_to_gnt[segs_to_map] = 541 persistent_gnt->page; 542 addr = (unsigned long) pfn_to_kaddr( 543 page_to_pfn(persistent_gnt->page)); 544 545 add_persistent_gnt(&blkif->persistent_gnts, 546 persistent_gnt); 547 blkif->persistent_gnt_c++; 548 pr_debug(DRV_PFX " grant %u added to the tree of persistent grants, using %u/%u\n", 549 persistent_gnt->gnt, blkif->persistent_gnt_c, 550 max_mapped_grant_pages(blkif->blk_protocol)); 551 } else { 552 /* 553 * We are either using persistent grants and 554 * hit the maximum limit of grants mapped, 555 * or we are not using persistent grants. 556 */ 557 if (use_persistent_gnts && 558 !blkif->vbd.overflow_max_grants) { 559 blkif->vbd.overflow_max_grants = 1; 560 pr_alert(DRV_PFX " domain %u, device %#x is using maximum number of persistent grants\n", 561 blkif->domid, blkif->vbd.handle); 562 } 563 new_map = true; 564 pages[i] = blkbk->pending_page(pending_req, i); 565 addr = vaddr(pending_req, i); 566 pages_to_gnt[segs_to_map] = 567 blkbk->pending_page(pending_req, i); 568 } 569 570 if (persistent_gnt) { 571 pages[i] = persistent_gnt->page; 572 persistent_gnts[i] = persistent_gnt; 573 } else { 574 persistent_gnts[i] = NULL; 575 } 576 577 if (new_map) { 578 flags = GNTMAP_host_map; 579 if (!persistent_gnt && 580 (pending_req->operation != BLKIF_OP_READ)) 581 flags |= GNTMAP_readonly; 582 gnttab_set_map_op(&map[segs_to_map++], addr, 583 flags, req->u.rw.seg[i].gref, 584 blkif->domid); 585 } 586 } 587 588 if (segs_to_map) { 589 ret = gnttab_map_refs(map, NULL, pages_to_gnt, segs_to_map); 590 BUG_ON(ret); 591 } 592 593 /* 594 * Now swizzle the MFN in our domain with the MFN from the other domain 595 * so that when we access vaddr(pending_req,i) it has the contents of 596 * the page from the other domain. 597 */ 598 bitmap_zero(pending_req->unmap_seg, BLKIF_MAX_SEGMENTS_PER_REQUEST); 599 for (i = 0, j = 0; i < nseg; i++) { 600 if (!persistent_gnts[i] || 601 persistent_gnts[i]->handle == BLKBACK_INVALID_HANDLE) { 602 /* This is a newly mapped grant */ 603 BUG_ON(j >= segs_to_map); 604 if (unlikely(map[j].status != 0)) { 605 pr_debug(DRV_PFX "invalid buffer -- could not remap it\n"); 606 map[j].handle = BLKBACK_INVALID_HANDLE; 607 ret |= 1; 608 if (persistent_gnts[i]) { 609 rb_erase(&persistent_gnts[i]->node, 610 &blkif->persistent_gnts); 611 blkif->persistent_gnt_c--; 612 kfree(persistent_gnts[i]); 613 persistent_gnts[i] = NULL; 614 } 615 } 616 } 617 if (persistent_gnts[i]) { 618 if (persistent_gnts[i]->handle == 619 BLKBACK_INVALID_HANDLE) { 620 /* 621 * If this is a new persistent grant 622 * save the handler 623 */ 624 persistent_gnts[i]->handle = map[j++].handle; 625 } 626 pending_handle(pending_req, i) = 627 persistent_gnts[i]->handle; 628 629 if (ret) 630 continue; 631 } else { 632 pending_handle(pending_req, i) = map[j++].handle; 633 bitmap_set(pending_req->unmap_seg, i, 1); 634 635 if (ret) 636 continue; 637 } 638 seg[i].offset = (req->u.rw.seg[i].first_sect << 9); 639 } 640 return ret; 641 } 642 643 static int dispatch_discard_io(struct xen_blkif *blkif, 644 struct blkif_request *req) 645 { 646 int err = 0; 647 int status = BLKIF_RSP_OKAY; 648 struct block_device *bdev = blkif->vbd.bdev; 649 unsigned long secure; 650 651 blkif->st_ds_req++; 652 653 xen_blkif_get(blkif); 654 secure = (blkif->vbd.discard_secure && 655 (req->u.discard.flag & BLKIF_DISCARD_SECURE)) ? 656 BLKDEV_DISCARD_SECURE : 0; 657 658 err = blkdev_issue_discard(bdev, req->u.discard.sector_number, 659 req->u.discard.nr_sectors, 660 GFP_KERNEL, secure); 661 662 if (err == -EOPNOTSUPP) { 663 pr_debug(DRV_PFX "discard op failed, not supported\n"); 664 status = BLKIF_RSP_EOPNOTSUPP; 665 } else if (err) 666 status = BLKIF_RSP_ERROR; 667 668 make_response(blkif, req->u.discard.id, req->operation, status); 669 xen_blkif_put(blkif); 670 return err; 671 } 672 673 static int dispatch_other_io(struct xen_blkif *blkif, 674 struct blkif_request *req, 675 struct pending_req *pending_req) 676 { 677 free_req(pending_req); 678 make_response(blkif, req->u.other.id, req->operation, 679 BLKIF_RSP_EOPNOTSUPP); 680 return -EIO; 681 } 682 683 static void xen_blk_drain_io(struct xen_blkif *blkif) 684 { 685 atomic_set(&blkif->drain, 1); 686 do { 687 /* The initial value is one, and one refcnt taken at the 688 * start of the xen_blkif_schedule thread. */ 689 if (atomic_read(&blkif->refcnt) <= 2) 690 break; 691 wait_for_completion_interruptible_timeout( 692 &blkif->drain_complete, HZ); 693 694 if (!atomic_read(&blkif->drain)) 695 break; 696 } while (!kthread_should_stop()); 697 atomic_set(&blkif->drain, 0); 698 } 699 700 /* 701 * Completion callback on the bio's. Called as bh->b_end_io() 702 */ 703 704 static void __end_block_io_op(struct pending_req *pending_req, int error) 705 { 706 /* An error fails the entire request. */ 707 if ((pending_req->operation == BLKIF_OP_FLUSH_DISKCACHE) && 708 (error == -EOPNOTSUPP)) { 709 pr_debug(DRV_PFX "flush diskcache op failed, not supported\n"); 710 xen_blkbk_flush_diskcache(XBT_NIL, pending_req->blkif->be, 0); 711 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 712 } else if ((pending_req->operation == BLKIF_OP_WRITE_BARRIER) && 713 (error == -EOPNOTSUPP)) { 714 pr_debug(DRV_PFX "write barrier op failed, not supported\n"); 715 xen_blkbk_barrier(XBT_NIL, pending_req->blkif->be, 0); 716 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 717 } else if (error) { 718 pr_debug(DRV_PFX "Buffer not up-to-date at end of operation," 719 " error=%d\n", error); 720 pending_req->status = BLKIF_RSP_ERROR; 721 } 722 723 /* 724 * If all of the bio's have completed it is time to unmap 725 * the grant references associated with 'request' and provide 726 * the proper response on the ring. 727 */ 728 if (atomic_dec_and_test(&pending_req->pendcnt)) { 729 xen_blkbk_unmap(pending_req); 730 make_response(pending_req->blkif, pending_req->id, 731 pending_req->operation, pending_req->status); 732 xen_blkif_put(pending_req->blkif); 733 if (atomic_read(&pending_req->blkif->refcnt) <= 2) { 734 if (atomic_read(&pending_req->blkif->drain)) 735 complete(&pending_req->blkif->drain_complete); 736 } 737 free_req(pending_req); 738 } 739 } 740 741 /* 742 * bio callback. 743 */ 744 static void end_block_io_op(struct bio *bio, int error) 745 { 746 __end_block_io_op(bio->bi_private, error); 747 bio_put(bio); 748 } 749 750 751 752 /* 753 * Function to copy the from the ring buffer the 'struct blkif_request' 754 * (which has the sectors we want, number of them, grant references, etc), 755 * and transmute it to the block API to hand it over to the proper block disk. 756 */ 757 static int 758 __do_block_io_op(struct xen_blkif *blkif) 759 { 760 union blkif_back_rings *blk_rings = &blkif->blk_rings; 761 struct blkif_request req; 762 struct pending_req *pending_req; 763 RING_IDX rc, rp; 764 int more_to_do = 0; 765 766 rc = blk_rings->common.req_cons; 767 rp = blk_rings->common.sring->req_prod; 768 rmb(); /* Ensure we see queued requests up to 'rp'. */ 769 770 while (rc != rp) { 771 772 if (RING_REQUEST_CONS_OVERFLOW(&blk_rings->common, rc)) 773 break; 774 775 if (kthread_should_stop()) { 776 more_to_do = 1; 777 break; 778 } 779 780 pending_req = alloc_req(); 781 if (NULL == pending_req) { 782 blkif->st_oo_req++; 783 more_to_do = 1; 784 break; 785 } 786 787 switch (blkif->blk_protocol) { 788 case BLKIF_PROTOCOL_NATIVE: 789 memcpy(&req, RING_GET_REQUEST(&blk_rings->native, rc), sizeof(req)); 790 break; 791 case BLKIF_PROTOCOL_X86_32: 792 blkif_get_x86_32_req(&req, RING_GET_REQUEST(&blk_rings->x86_32, rc)); 793 break; 794 case BLKIF_PROTOCOL_X86_64: 795 blkif_get_x86_64_req(&req, RING_GET_REQUEST(&blk_rings->x86_64, rc)); 796 break; 797 default: 798 BUG(); 799 } 800 blk_rings->common.req_cons = ++rc; /* before make_response() */ 801 802 /* Apply all sanity checks to /private copy/ of request. */ 803 barrier(); 804 805 switch (req.operation) { 806 case BLKIF_OP_READ: 807 case BLKIF_OP_WRITE: 808 case BLKIF_OP_WRITE_BARRIER: 809 case BLKIF_OP_FLUSH_DISKCACHE: 810 if (dispatch_rw_block_io(blkif, &req, pending_req)) 811 goto done; 812 break; 813 case BLKIF_OP_DISCARD: 814 free_req(pending_req); 815 if (dispatch_discard_io(blkif, &req)) 816 goto done; 817 break; 818 default: 819 if (dispatch_other_io(blkif, &req, pending_req)) 820 goto done; 821 break; 822 } 823 824 /* Yield point for this unbounded loop. */ 825 cond_resched(); 826 } 827 done: 828 return more_to_do; 829 } 830 831 static int 832 do_block_io_op(struct xen_blkif *blkif) 833 { 834 union blkif_back_rings *blk_rings = &blkif->blk_rings; 835 int more_to_do; 836 837 do { 838 more_to_do = __do_block_io_op(blkif); 839 if (more_to_do) 840 break; 841 842 RING_FINAL_CHECK_FOR_REQUESTS(&blk_rings->common, more_to_do); 843 } while (more_to_do); 844 845 return more_to_do; 846 } 847 /* 848 * Transmutation of the 'struct blkif_request' to a proper 'struct bio' 849 * and call the 'submit_bio' to pass it to the underlying storage. 850 */ 851 static int dispatch_rw_block_io(struct xen_blkif *blkif, 852 struct blkif_request *req, 853 struct pending_req *pending_req) 854 { 855 struct phys_req preq; 856 struct seg_buf seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 857 unsigned int nseg; 858 struct bio *bio = NULL; 859 struct bio *biolist[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 860 int i, nbio = 0; 861 int operation; 862 struct blk_plug plug; 863 bool drain = false; 864 struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 865 866 switch (req->operation) { 867 case BLKIF_OP_READ: 868 blkif->st_rd_req++; 869 operation = READ; 870 break; 871 case BLKIF_OP_WRITE: 872 blkif->st_wr_req++; 873 operation = WRITE_ODIRECT; 874 break; 875 case BLKIF_OP_WRITE_BARRIER: 876 drain = true; 877 case BLKIF_OP_FLUSH_DISKCACHE: 878 blkif->st_f_req++; 879 operation = WRITE_FLUSH; 880 break; 881 default: 882 operation = 0; /* make gcc happy */ 883 goto fail_response; 884 break; 885 } 886 887 /* Check that the number of segments is sane. */ 888 nseg = req->u.rw.nr_segments; 889 890 if (unlikely(nseg == 0 && operation != WRITE_FLUSH) || 891 unlikely(nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) { 892 pr_debug(DRV_PFX "Bad number of segments in request (%d)\n", 893 nseg); 894 /* Haven't submitted any bio's yet. */ 895 goto fail_response; 896 } 897 898 preq.sector_number = req->u.rw.sector_number; 899 preq.nr_sects = 0; 900 901 pending_req->blkif = blkif; 902 pending_req->id = req->u.rw.id; 903 pending_req->operation = req->operation; 904 pending_req->status = BLKIF_RSP_OKAY; 905 pending_req->nr_pages = nseg; 906 907 for (i = 0; i < nseg; i++) { 908 seg[i].nsec = req->u.rw.seg[i].last_sect - 909 req->u.rw.seg[i].first_sect + 1; 910 if ((req->u.rw.seg[i].last_sect >= (PAGE_SIZE >> 9)) || 911 (req->u.rw.seg[i].last_sect < req->u.rw.seg[i].first_sect)) 912 goto fail_response; 913 preq.nr_sects += seg[i].nsec; 914 915 } 916 917 if (xen_vbd_translate(&preq, blkif, operation) != 0) { 918 pr_debug(DRV_PFX "access denied: %s of [%llu,%llu] on dev=%04x\n", 919 operation == READ ? "read" : "write", 920 preq.sector_number, 921 preq.sector_number + preq.nr_sects, 922 blkif->vbd.pdevice); 923 goto fail_response; 924 } 925 926 /* 927 * This check _MUST_ be done after xen_vbd_translate as the preq.bdev 928 * is set there. 929 */ 930 for (i = 0; i < nseg; i++) { 931 if (((int)preq.sector_number|(int)seg[i].nsec) & 932 ((bdev_logical_block_size(preq.bdev) >> 9) - 1)) { 933 pr_debug(DRV_PFX "Misaligned I/O request from domain %d", 934 blkif->domid); 935 goto fail_response; 936 } 937 } 938 939 /* Wait on all outstanding I/O's and once that has been completed 940 * issue the WRITE_FLUSH. 941 */ 942 if (drain) 943 xen_blk_drain_io(pending_req->blkif); 944 945 /* 946 * If we have failed at this point, we need to undo the M2P override, 947 * set gnttab_set_unmap_op on all of the grant references and perform 948 * the hypercall to unmap the grants - that is all done in 949 * xen_blkbk_unmap. 950 */ 951 if (xen_blkbk_map(req, pending_req, seg, pages)) 952 goto fail_flush; 953 954 /* 955 * This corresponding xen_blkif_put is done in __end_block_io_op, or 956 * below (in "!bio") if we are handling a BLKIF_OP_DISCARD. 957 */ 958 xen_blkif_get(blkif); 959 960 for (i = 0; i < nseg; i++) { 961 while ((bio == NULL) || 962 (bio_add_page(bio, 963 pages[i], 964 seg[i].nsec << 9, 965 seg[i].offset) == 0)) { 966 967 bio = bio_alloc(GFP_KERNEL, nseg-i); 968 if (unlikely(bio == NULL)) 969 goto fail_put_bio; 970 971 biolist[nbio++] = bio; 972 bio->bi_bdev = preq.bdev; 973 bio->bi_private = pending_req; 974 bio->bi_end_io = end_block_io_op; 975 bio->bi_sector = preq.sector_number; 976 } 977 978 preq.sector_number += seg[i].nsec; 979 } 980 981 /* This will be hit if the operation was a flush or discard. */ 982 if (!bio) { 983 BUG_ON(operation != WRITE_FLUSH); 984 985 bio = bio_alloc(GFP_KERNEL, 0); 986 if (unlikely(bio == NULL)) 987 goto fail_put_bio; 988 989 biolist[nbio++] = bio; 990 bio->bi_bdev = preq.bdev; 991 bio->bi_private = pending_req; 992 bio->bi_end_io = end_block_io_op; 993 } 994 995 atomic_set(&pending_req->pendcnt, nbio); 996 blk_start_plug(&plug); 997 998 for (i = 0; i < nbio; i++) 999 submit_bio(operation, biolist[i]); 1000 1001 /* Let the I/Os go.. */ 1002 blk_finish_plug(&plug); 1003 1004 if (operation == READ) 1005 blkif->st_rd_sect += preq.nr_sects; 1006 else if (operation & WRITE) 1007 blkif->st_wr_sect += preq.nr_sects; 1008 1009 return 0; 1010 1011 fail_flush: 1012 xen_blkbk_unmap(pending_req); 1013 fail_response: 1014 /* Haven't submitted any bio's yet. */ 1015 make_response(blkif, req->u.rw.id, req->operation, BLKIF_RSP_ERROR); 1016 free_req(pending_req); 1017 msleep(1); /* back off a bit */ 1018 return -EIO; 1019 1020 fail_put_bio: 1021 for (i = 0; i < nbio; i++) 1022 bio_put(biolist[i]); 1023 atomic_set(&pending_req->pendcnt, 1); 1024 __end_block_io_op(pending_req, -EINVAL); 1025 msleep(1); /* back off a bit */ 1026 return -EIO; 1027 } 1028 1029 1030 1031 /* 1032 * Put a response on the ring on how the operation fared. 1033 */ 1034 static void make_response(struct xen_blkif *blkif, u64 id, 1035 unsigned short op, int st) 1036 { 1037 struct blkif_response resp; 1038 unsigned long flags; 1039 union blkif_back_rings *blk_rings = &blkif->blk_rings; 1040 int notify; 1041 1042 resp.id = id; 1043 resp.operation = op; 1044 resp.status = st; 1045 1046 spin_lock_irqsave(&blkif->blk_ring_lock, flags); 1047 /* Place on the response ring for the relevant domain. */ 1048 switch (blkif->blk_protocol) { 1049 case BLKIF_PROTOCOL_NATIVE: 1050 memcpy(RING_GET_RESPONSE(&blk_rings->native, blk_rings->native.rsp_prod_pvt), 1051 &resp, sizeof(resp)); 1052 break; 1053 case BLKIF_PROTOCOL_X86_32: 1054 memcpy(RING_GET_RESPONSE(&blk_rings->x86_32, blk_rings->x86_32.rsp_prod_pvt), 1055 &resp, sizeof(resp)); 1056 break; 1057 case BLKIF_PROTOCOL_X86_64: 1058 memcpy(RING_GET_RESPONSE(&blk_rings->x86_64, blk_rings->x86_64.rsp_prod_pvt), 1059 &resp, sizeof(resp)); 1060 break; 1061 default: 1062 BUG(); 1063 } 1064 blk_rings->common.rsp_prod_pvt++; 1065 RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&blk_rings->common, notify); 1066 spin_unlock_irqrestore(&blkif->blk_ring_lock, flags); 1067 if (notify) 1068 notify_remote_via_irq(blkif->irq); 1069 } 1070 1071 static int __init xen_blkif_init(void) 1072 { 1073 int i, mmap_pages; 1074 int rc = 0; 1075 1076 if (!xen_domain()) 1077 return -ENODEV; 1078 1079 blkbk = kzalloc(sizeof(struct xen_blkbk), GFP_KERNEL); 1080 if (!blkbk) { 1081 pr_alert(DRV_PFX "%s: out of memory!\n", __func__); 1082 return -ENOMEM; 1083 } 1084 1085 mmap_pages = xen_blkif_reqs * BLKIF_MAX_SEGMENTS_PER_REQUEST; 1086 1087 blkbk->pending_reqs = kzalloc(sizeof(blkbk->pending_reqs[0]) * 1088 xen_blkif_reqs, GFP_KERNEL); 1089 blkbk->pending_grant_handles = kmalloc(sizeof(blkbk->pending_grant_handles[0]) * 1090 mmap_pages, GFP_KERNEL); 1091 blkbk->pending_pages = kzalloc(sizeof(blkbk->pending_pages[0]) * 1092 mmap_pages, GFP_KERNEL); 1093 1094 if (!blkbk->pending_reqs || !blkbk->pending_grant_handles || 1095 !blkbk->pending_pages) { 1096 rc = -ENOMEM; 1097 goto out_of_memory; 1098 } 1099 1100 for (i = 0; i < mmap_pages; i++) { 1101 blkbk->pending_grant_handles[i] = BLKBACK_INVALID_HANDLE; 1102 blkbk->pending_pages[i] = alloc_page(GFP_KERNEL); 1103 if (blkbk->pending_pages[i] == NULL) { 1104 rc = -ENOMEM; 1105 goto out_of_memory; 1106 } 1107 } 1108 rc = xen_blkif_interface_init(); 1109 if (rc) 1110 goto failed_init; 1111 1112 INIT_LIST_HEAD(&blkbk->pending_free); 1113 spin_lock_init(&blkbk->pending_free_lock); 1114 init_waitqueue_head(&blkbk->pending_free_wq); 1115 1116 for (i = 0; i < xen_blkif_reqs; i++) 1117 list_add_tail(&blkbk->pending_reqs[i].free_list, 1118 &blkbk->pending_free); 1119 1120 rc = xen_blkif_xenbus_init(); 1121 if (rc) 1122 goto failed_init; 1123 1124 return 0; 1125 1126 out_of_memory: 1127 pr_alert(DRV_PFX "%s: out of memory\n", __func__); 1128 failed_init: 1129 kfree(blkbk->pending_reqs); 1130 kfree(blkbk->pending_grant_handles); 1131 if (blkbk->pending_pages) { 1132 for (i = 0; i < mmap_pages; i++) { 1133 if (blkbk->pending_pages[i]) 1134 __free_page(blkbk->pending_pages[i]); 1135 } 1136 kfree(blkbk->pending_pages); 1137 } 1138 kfree(blkbk); 1139 blkbk = NULL; 1140 return rc; 1141 } 1142 1143 module_init(xen_blkif_init); 1144 1145 MODULE_LICENSE("Dual BSD/GPL"); 1146 MODULE_ALIAS("xen-backend:vbd"); 1147