1/* 2 * XenBSD block device driver 3 * 4 * Copyright (c) 2010-2013 Spectra Logic Corporation 5 * Copyright (c) 2009 Scott Long, Yahoo! 6 * Copyright (c) 2009 Frank Suchomel, Citrix 7 * Copyright (c) 2009 Doug F. Rabson, Citrix 8 * Copyright (c) 2005 Kip Macy 9 * Copyright (c) 2003-2004, Keir Fraser & Steve Hand 10 * Modifications by Mark A. Williamson are (c) Intel Research Cambridge 11 * 12 * 13 * Permission is hereby granted, free of charge, to any person obtaining a copy 14 * of this software and associated documentation files (the "Software"), to 15 * deal in the Software without restriction, including without limitation the 16 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 17 * sell copies of the Software, and to permit persons to whom the Software is 18 * furnished to do so, subject to the following conditions: 19 * 20 * The above copyright notice and this permission notice shall be included in 21 * all copies or substantial portions of the Software. 22 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 23 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 24 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 25 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 26 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 27 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 28 * DEALINGS IN THE SOFTWARE. 29 */ 30 31#include <sys/param.h> 32#include <sys/systm.h> 33#include <sys/malloc.h> 34#include <sys/kernel.h> 35#include <vm/vm.h> 36#include <vm/pmap.h> 37 38#include <sys/bio.h> 39#include <sys/bus.h> 40#include <sys/conf.h> 41#include <sys/module.h> 42#include <sys/sysctl.h> 43 44#include <machine/bus.h> 45#include <sys/rman.h> 46#include <machine/resource.h> 47#include <machine/vmparam.h> 48 49#include <xen/xen-os.h> 50#include <xen/hypervisor.h> 51#include <xen/xen_intr.h> 52#include <xen/gnttab.h> 53#include <contrib/xen/grant_table.h> 54#include <contrib/xen/io/protocols.h> 55#include <xen/xenbus/xenbusvar.h> 56 57#include <machine/_inttypes.h> 58 59#include <geom/geom_disk.h> 60 61#include <dev/xen/blkfront/block.h> 62 63#include "xenbus_if.h" 64 65/*--------------------------- Forward Declarations ---------------------------*/ 66static void xbd_closing(device_t); 67static void xbd_startio(struct xbd_softc *sc); 68 69/*---------------------------------- Macros ----------------------------------*/ 70#if 0 71#define DPRINTK(fmt, args...) printf("[XEN] %s:%d: " fmt ".\n", __func__, __LINE__, ##args) 72#else 73#define DPRINTK(fmt, args...) 74#endif 75 76#define XBD_SECTOR_SHFT 9 77 78/*---------------------------- Global Static Data ----------------------------*/ 79static MALLOC_DEFINE(M_XENBLOCKFRONT, "xbd", "Xen Block Front driver data"); 80 81static int xbd_enable_indirect = 1; 82SYSCTL_NODE(_hw, OID_AUTO, xbd, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 83 "xbd driver parameters"); 84SYSCTL_INT(_hw_xbd, OID_AUTO, xbd_enable_indirect, CTLFLAG_RDTUN, 85 &xbd_enable_indirect, 0, "Enable xbd indirect segments"); 86 87/*---------------------------- Command Processing ----------------------------*/ 88static void 89xbd_freeze(struct xbd_softc *sc, xbd_flag_t xbd_flag) 90{ 91 if (xbd_flag != XBDF_NONE && (sc->xbd_flags & xbd_flag) != 0) 92 return; 93 94 sc->xbd_flags |= xbd_flag; 95 sc->xbd_qfrozen_cnt++; 96} 97 98static void 99xbd_thaw(struct xbd_softc *sc, xbd_flag_t xbd_flag) 100{ 101 if (xbd_flag != XBDF_NONE && (sc->xbd_flags & xbd_flag) == 0) 102 return; 103 104 if (sc->xbd_qfrozen_cnt == 0) 105 panic("%s: Thaw with flag 0x%x while not frozen.", 106 __func__, xbd_flag); 107 108 sc->xbd_flags &= ~xbd_flag; 109 sc->xbd_qfrozen_cnt--; 110} 111 112static void 113xbd_cm_freeze(struct xbd_softc *sc, struct xbd_command *cm, xbdc_flag_t cm_flag) 114{ 115 if ((cm->cm_flags & XBDCF_FROZEN) != 0) 116 return; 117 118 cm->cm_flags |= XBDCF_FROZEN|cm_flag; 119 xbd_freeze(sc, XBDF_NONE); 120} 121 122static void 123xbd_cm_thaw(struct xbd_softc *sc, struct xbd_command *cm) 124{ 125 if ((cm->cm_flags & XBDCF_FROZEN) == 0) 126 return; 127 128 cm->cm_flags &= ~XBDCF_FROZEN; 129 xbd_thaw(sc, XBDF_NONE); 130} 131 132static inline void 133xbd_flush_requests(struct xbd_softc *sc) 134{ 135 int notify; 136 137 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->xbd_ring, notify); 138 139 if (notify) 140 xen_intr_signal(sc->xen_intr_handle); 141} 142 143static void 144xbd_free_command(struct xbd_command *cm) 145{ 146 147 KASSERT((cm->cm_flags & XBDCF_Q_MASK) == XBD_Q_NONE, 148 ("Freeing command that is still on queue %d.", 149 cm->cm_flags & XBDCF_Q_MASK)); 150 151 cm->cm_flags = XBDCF_INITIALIZER; 152 cm->cm_bp = NULL; 153 cm->cm_complete = NULL; 154 xbd_enqueue_cm(cm, XBD_Q_FREE); 155 xbd_thaw(cm->cm_sc, XBDF_CM_SHORTAGE); 156} 157 158static void 159xbd_mksegarray(bus_dma_segment_t *segs, int nsegs, 160 grant_ref_t * gref_head, int otherend_id, int readonly, 161 grant_ref_t * sg_ref, struct blkif_request_segment *sg) 162{ 163 struct blkif_request_segment *last_block_sg = sg + nsegs; 164 vm_paddr_t buffer_ma; 165 uint64_t fsect, lsect; 166 int ref; 167 168 while (sg < last_block_sg) { 169 KASSERT(segs->ds_addr % (1 << XBD_SECTOR_SHFT) == 0, 170 ("XEN disk driver I/O must be sector aligned")); 171 KASSERT(segs->ds_len % (1 << XBD_SECTOR_SHFT) == 0, 172 ("XEN disk driver I/Os must be a multiple of " 173 "the sector length")); 174 buffer_ma = segs->ds_addr; 175 fsect = (buffer_ma & PAGE_MASK) >> XBD_SECTOR_SHFT; 176 lsect = fsect + (segs->ds_len >> XBD_SECTOR_SHFT) - 1; 177 178 KASSERT(lsect <= 7, ("XEN disk driver data cannot " 179 "cross a page boundary")); 180 181 /* install a grant reference. */ 182 ref = gnttab_claim_grant_reference(gref_head); 183 184 /* 185 * GNTTAB_LIST_END == 0xffffffff, but it is private 186 * to gnttab.c. 187 */ 188 KASSERT(ref != ~0, ("grant_reference failed")); 189 190 gnttab_grant_foreign_access_ref( 191 ref, 192 otherend_id, 193 buffer_ma >> PAGE_SHIFT, 194 readonly); 195 196 *sg_ref = ref; 197 *sg = (struct blkif_request_segment) { 198 .gref = ref, 199 .first_sect = fsect, 200 .last_sect = lsect 201 }; 202 sg++; 203 sg_ref++; 204 segs++; 205 } 206} 207 208static void 209xbd_queue_cb(void *arg, bus_dma_segment_t *segs, int nsegs, int error) 210{ 211 struct xbd_softc *sc; 212 struct xbd_command *cm; 213 int op; 214 215 cm = arg; 216 sc = cm->cm_sc; 217 218 if (error) { 219 cm->cm_bp->bio_error = EIO; 220 biodone(cm->cm_bp); 221 xbd_free_command(cm); 222 return; 223 } 224 225 KASSERT(nsegs <= sc->xbd_max_request_segments, 226 ("Too many segments in a blkfront I/O")); 227 228 if (nsegs <= BLKIF_MAX_SEGMENTS_PER_REQUEST) { 229 blkif_request_t *ring_req; 230 231 /* Fill out a blkif_request_t structure. */ 232 ring_req = (blkif_request_t *) 233 RING_GET_REQUEST(&sc->xbd_ring, sc->xbd_ring.req_prod_pvt); 234 sc->xbd_ring.req_prod_pvt++; 235 ring_req->id = cm->cm_id; 236 ring_req->operation = cm->cm_operation; 237 ring_req->sector_number = cm->cm_sector_number; 238 ring_req->handle = (blkif_vdev_t)(uintptr_t)sc->xbd_disk; 239 ring_req->nr_segments = nsegs; 240 cm->cm_nseg = nsegs; 241 xbd_mksegarray(segs, nsegs, &cm->cm_gref_head, 242 xenbus_get_otherend_id(sc->xbd_dev), 243 cm->cm_operation == BLKIF_OP_WRITE, 244 cm->cm_sg_refs, ring_req->seg); 245 } else { 246 blkif_request_indirect_t *ring_req; 247 248 /* Fill out a blkif_request_indirect_t structure. */ 249 ring_req = (blkif_request_indirect_t *) 250 RING_GET_REQUEST(&sc->xbd_ring, sc->xbd_ring.req_prod_pvt); 251 sc->xbd_ring.req_prod_pvt++; 252 ring_req->id = cm->cm_id; 253 ring_req->operation = BLKIF_OP_INDIRECT; 254 ring_req->indirect_op = cm->cm_operation; 255 ring_req->sector_number = cm->cm_sector_number; 256 ring_req->handle = (blkif_vdev_t)(uintptr_t)sc->xbd_disk; 257 ring_req->nr_segments = nsegs; 258 cm->cm_nseg = nsegs; 259 xbd_mksegarray(segs, nsegs, &cm->cm_gref_head, 260 xenbus_get_otherend_id(sc->xbd_dev), 261 cm->cm_operation == BLKIF_OP_WRITE, 262 cm->cm_sg_refs, cm->cm_indirectionpages); 263 memcpy(ring_req->indirect_grefs, &cm->cm_indirectionrefs, 264 sizeof(grant_ref_t) * sc->xbd_max_request_indirectpages); 265 } 266 267 if (cm->cm_operation == BLKIF_OP_READ) 268 op = BUS_DMASYNC_PREREAD; 269 else if (cm->cm_operation == BLKIF_OP_WRITE) 270 op = BUS_DMASYNC_PREWRITE; 271 else 272 op = 0; 273 bus_dmamap_sync(sc->xbd_io_dmat, cm->cm_map, op); 274 275 gnttab_free_grant_references(cm->cm_gref_head); 276 277 xbd_enqueue_cm(cm, XBD_Q_BUSY); 278 279 /* 280 * If bus dma had to asynchronously call us back to dispatch 281 * this command, we are no longer executing in the context of 282 * xbd_startio(). Thus we cannot rely on xbd_startio()'s call to 283 * xbd_flush_requests() to publish this command to the backend 284 * along with any other commands that it could batch. 285 */ 286 if ((cm->cm_flags & XBDCF_ASYNC_MAPPING) != 0) 287 xbd_flush_requests(sc); 288 289 return; 290} 291 292static int 293xbd_queue_request(struct xbd_softc *sc, struct xbd_command *cm) 294{ 295 int error; 296 297 if (cm->cm_bp != NULL) 298 error = bus_dmamap_load_bio(sc->xbd_io_dmat, cm->cm_map, 299 cm->cm_bp, xbd_queue_cb, cm, 0); 300 else 301 error = bus_dmamap_load(sc->xbd_io_dmat, cm->cm_map, 302 cm->cm_data, cm->cm_datalen, xbd_queue_cb, cm, 0); 303 if (error == EINPROGRESS) { 304 /* 305 * Maintain queuing order by freezing the queue. The next 306 * command may not require as many resources as the command 307 * we just attempted to map, so we can't rely on bus dma 308 * blocking for it too. 309 */ 310 xbd_cm_freeze(sc, cm, XBDCF_ASYNC_MAPPING); 311 return (0); 312 } 313 314 return (error); 315} 316 317static void 318xbd_restart_queue_callback(void *arg) 319{ 320 struct xbd_softc *sc = arg; 321 322 mtx_lock(&sc->xbd_io_lock); 323 324 xbd_thaw(sc, XBDF_GNT_SHORTAGE); 325 326 xbd_startio(sc); 327 328 mtx_unlock(&sc->xbd_io_lock); 329} 330 331static struct xbd_command * 332xbd_bio_command(struct xbd_softc *sc) 333{ 334 struct xbd_command *cm; 335 struct bio *bp; 336 337 if (__predict_false(sc->xbd_state != XBD_STATE_CONNECTED)) 338 return (NULL); 339 340 bp = xbd_dequeue_bio(sc); 341 if (bp == NULL) 342 return (NULL); 343 344 if ((cm = xbd_dequeue_cm(sc, XBD_Q_FREE)) == NULL) { 345 xbd_freeze(sc, XBDF_CM_SHORTAGE); 346 xbd_requeue_bio(sc, bp); 347 return (NULL); 348 } 349 350 if (gnttab_alloc_grant_references(sc->xbd_max_request_segments, 351 &cm->cm_gref_head) != 0) { 352 gnttab_request_free_callback(&sc->xbd_callback, 353 xbd_restart_queue_callback, sc, 354 sc->xbd_max_request_segments); 355 xbd_freeze(sc, XBDF_GNT_SHORTAGE); 356 xbd_requeue_bio(sc, bp); 357 xbd_enqueue_cm(cm, XBD_Q_FREE); 358 return (NULL); 359 } 360 361 cm->cm_bp = bp; 362 cm->cm_sector_number = (blkif_sector_t)bp->bio_pblkno; 363 364 switch (bp->bio_cmd) { 365 case BIO_READ: 366 cm->cm_operation = BLKIF_OP_READ; 367 break; 368 case BIO_WRITE: 369 cm->cm_operation = BLKIF_OP_WRITE; 370 if ((bp->bio_flags & BIO_ORDERED) != 0) { 371 if ((sc->xbd_flags & XBDF_BARRIER) != 0) { 372 cm->cm_operation = BLKIF_OP_WRITE_BARRIER; 373 } else { 374 /* 375 * Single step this command. 376 */ 377 cm->cm_flags |= XBDCF_Q_FREEZE; 378 if (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 379 /* 380 * Wait for in-flight requests to 381 * finish. 382 */ 383 xbd_freeze(sc, XBDF_WAIT_IDLE); 384 xbd_requeue_cm(cm, XBD_Q_READY); 385 return (NULL); 386 } 387 } 388 } 389 break; 390 case BIO_FLUSH: 391 if ((sc->xbd_flags & XBDF_FLUSH) != 0) 392 cm->cm_operation = BLKIF_OP_FLUSH_DISKCACHE; 393 else if ((sc->xbd_flags & XBDF_BARRIER) != 0) 394 cm->cm_operation = BLKIF_OP_WRITE_BARRIER; 395 else 396 panic("flush request, but no flush support available"); 397 break; 398 default: 399 biofinish(bp, NULL, EOPNOTSUPP); 400 xbd_enqueue_cm(cm, XBD_Q_FREE); 401 return (NULL); 402 } 403 404 return (cm); 405} 406 407/* 408 * Dequeue buffers and place them in the shared communication ring. 409 * Return when no more requests can be accepted or all buffers have 410 * been queued. 411 * 412 * Signal XEN once the ring has been filled out. 413 */ 414static void 415xbd_startio(struct xbd_softc *sc) 416{ 417 struct xbd_command *cm; 418 int error, queued = 0; 419 420 mtx_assert(&sc->xbd_io_lock, MA_OWNED); 421 422 if (sc->xbd_state != XBD_STATE_CONNECTED) 423 return; 424 425 while (!RING_FULL(&sc->xbd_ring)) { 426 if (sc->xbd_qfrozen_cnt != 0) 427 break; 428 429 cm = xbd_dequeue_cm(sc, XBD_Q_READY); 430 431 if (cm == NULL) 432 cm = xbd_bio_command(sc); 433 434 if (cm == NULL) 435 break; 436 437 if ((cm->cm_flags & XBDCF_Q_FREEZE) != 0) { 438 /* 439 * Single step command. Future work is 440 * held off until this command completes. 441 */ 442 xbd_cm_freeze(sc, cm, XBDCF_Q_FREEZE); 443 } 444 445 if ((error = xbd_queue_request(sc, cm)) != 0) { 446 printf("xbd_queue_request returned %d\n", error); 447 break; 448 } 449 queued++; 450 } 451 452 if (queued != 0) 453 xbd_flush_requests(sc); 454} 455 456static void 457xbd_bio_complete(struct xbd_softc *sc, struct xbd_command *cm) 458{ 459 struct bio *bp; 460 461 bp = cm->cm_bp; 462 463 if (__predict_false(cm->cm_status != BLKIF_RSP_OKAY)) { 464 disk_err(bp, "disk error" , -1, 0); 465 printf(" status: %x\n", cm->cm_status); 466 bp->bio_flags |= BIO_ERROR; 467 } 468 469 if (bp->bio_flags & BIO_ERROR) 470 bp->bio_error = EIO; 471 else 472 bp->bio_resid = 0; 473 474 xbd_free_command(cm); 475 biodone(bp); 476} 477 478static void 479xbd_int(void *xsc) 480{ 481 struct xbd_softc *sc = xsc; 482 struct xbd_command *cm; 483 blkif_response_t *bret; 484 RING_IDX i, rp; 485 int op; 486 487 mtx_lock(&sc->xbd_io_lock); 488 489 if (__predict_false(sc->xbd_state == XBD_STATE_DISCONNECTED)) { 490 mtx_unlock(&sc->xbd_io_lock); 491 return; 492 } 493 494 again: 495 rp = sc->xbd_ring.sring->rsp_prod; 496 rmb(); /* Ensure we see queued responses up to 'rp'. */ 497 498 for (i = sc->xbd_ring.rsp_cons; i != rp;) { 499 bret = RING_GET_RESPONSE(&sc->xbd_ring, i); 500 cm = &sc->xbd_shadow[bret->id]; 501 502 xbd_remove_cm(cm, XBD_Q_BUSY); 503 gnttab_end_foreign_access_references(cm->cm_nseg, 504 cm->cm_sg_refs); 505 i++; 506 507 if (cm->cm_operation == BLKIF_OP_READ) 508 op = BUS_DMASYNC_POSTREAD; 509 else if (cm->cm_operation == BLKIF_OP_WRITE || 510 cm->cm_operation == BLKIF_OP_WRITE_BARRIER) 511 op = BUS_DMASYNC_POSTWRITE; 512 else 513 op = 0; 514 bus_dmamap_sync(sc->xbd_io_dmat, cm->cm_map, op); 515 bus_dmamap_unload(sc->xbd_io_dmat, cm->cm_map); 516 517 /* 518 * Release any hold this command has on future command 519 * dispatch. 520 */ 521 xbd_cm_thaw(sc, cm); 522 523 /* 524 * Directly call the i/o complete routine to save an 525 * an indirection in the common case. 526 */ 527 cm->cm_status = bret->status; 528 if (cm->cm_bp) 529 xbd_bio_complete(sc, cm); 530 else if (cm->cm_complete != NULL) 531 cm->cm_complete(cm); 532 else 533 xbd_free_command(cm); 534 } 535 536 sc->xbd_ring.rsp_cons = i; 537 538 if (i != sc->xbd_ring.req_prod_pvt) { 539 int more_to_do; 540 RING_FINAL_CHECK_FOR_RESPONSES(&sc->xbd_ring, more_to_do); 541 if (more_to_do) 542 goto again; 543 } else { 544 sc->xbd_ring.sring->rsp_event = i + 1; 545 } 546 547 if (xbd_queue_length(sc, XBD_Q_BUSY) == 0) 548 xbd_thaw(sc, XBDF_WAIT_IDLE); 549 550 xbd_startio(sc); 551 552 if (__predict_false(sc->xbd_state == XBD_STATE_SUSPENDED)) 553 wakeup(&sc->xbd_cm_q[XBD_Q_BUSY]); 554 555 mtx_unlock(&sc->xbd_io_lock); 556} 557 558/*------------------------------- Dump Support -------------------------------*/ 559/** 560 * Quiesce the disk writes for a dump file before allowing the next buffer. 561 */ 562static void 563xbd_quiesce(struct xbd_softc *sc) 564{ 565 int mtd; 566 567 // While there are outstanding requests 568 while (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 569 RING_FINAL_CHECK_FOR_RESPONSES(&sc->xbd_ring, mtd); 570 if (mtd) { 571 /* Received request completions, update queue. */ 572 xbd_int(sc); 573 } 574 if (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 575 /* 576 * Still pending requests, wait for the disk i/o 577 * to complete. 578 */ 579 HYPERVISOR_yield(); 580 } 581 } 582} 583 584/* Kernel dump function for a paravirtualized disk device */ 585static void 586xbd_dump_complete(struct xbd_command *cm) 587{ 588 589 xbd_enqueue_cm(cm, XBD_Q_COMPLETE); 590} 591 592static int 593xbd_dump(void *arg, void *virtual, off_t offset, size_t length) 594{ 595 struct disk *dp = arg; 596 struct xbd_softc *sc = dp->d_drv1; 597 struct xbd_command *cm; 598 size_t chunk; 599 int rc = 0; 600 601 if (length == 0) 602 return (0); 603 604 xbd_quiesce(sc); /* All quiet on the western front. */ 605 606 /* 607 * If this lock is held, then this module is failing, and a 608 * successful kernel dump is highly unlikely anyway. 609 */ 610 mtx_lock(&sc->xbd_io_lock); 611 612 /* Split the 64KB block as needed */ 613 while (length > 0) { 614 cm = xbd_dequeue_cm(sc, XBD_Q_FREE); 615 if (cm == NULL) { 616 mtx_unlock(&sc->xbd_io_lock); 617 device_printf(sc->xbd_dev, "dump: no more commands?\n"); 618 return (EBUSY); 619 } 620 621 if (gnttab_alloc_grant_references(sc->xbd_max_request_segments, 622 &cm->cm_gref_head) != 0) { 623 xbd_free_command(cm); 624 mtx_unlock(&sc->xbd_io_lock); 625 device_printf(sc->xbd_dev, "no more grant allocs?\n"); 626 return (EBUSY); 627 } 628 629 chunk = length > sc->xbd_max_request_size ? 630 sc->xbd_max_request_size : length; 631 cm->cm_data = virtual; 632 cm->cm_datalen = chunk; 633 cm->cm_operation = BLKIF_OP_WRITE; 634 cm->cm_sector_number = offset / dp->d_sectorsize; 635 cm->cm_complete = xbd_dump_complete; 636 637 xbd_enqueue_cm(cm, XBD_Q_READY); 638 639 length -= chunk; 640 offset += chunk; 641 virtual = (char *) virtual + chunk; 642 } 643 644 /* Tell DOM0 to do the I/O */ 645 xbd_startio(sc); 646 mtx_unlock(&sc->xbd_io_lock); 647 648 /* Poll for the completion. */ 649 xbd_quiesce(sc); /* All quite on the eastern front */ 650 651 /* If there were any errors, bail out... */ 652 while ((cm = xbd_dequeue_cm(sc, XBD_Q_COMPLETE)) != NULL) { 653 if (cm->cm_status != BLKIF_RSP_OKAY) { 654 device_printf(sc->xbd_dev, 655 "Dump I/O failed at sector %jd\n", 656 cm->cm_sector_number); 657 rc = EIO; 658 } 659 xbd_free_command(cm); 660 } 661 662 return (rc); 663} 664 665/*----------------------------- Disk Entrypoints -----------------------------*/ 666static int 667xbd_open(struct disk *dp) 668{ 669 struct xbd_softc *sc = dp->d_drv1; 670 671 if (sc == NULL) { 672 printf("xbd%d: not found", dp->d_unit); 673 return (ENXIO); 674 } 675 676 sc->xbd_flags |= XBDF_OPEN; 677 sc->xbd_users++; 678 return (0); 679} 680 681static int 682xbd_close(struct disk *dp) 683{ 684 struct xbd_softc *sc = dp->d_drv1; 685 686 if (sc == NULL) 687 return (ENXIO); 688 sc->xbd_flags &= ~XBDF_OPEN; 689 if (--(sc->xbd_users) == 0) { 690 /* 691 * Check whether we have been instructed to close. We will 692 * have ignored this request initially, as the device was 693 * still mounted. 694 */ 695 if (xenbus_get_otherend_state(sc->xbd_dev) == 696 XenbusStateClosing) 697 xbd_closing(sc->xbd_dev); 698 } 699 return (0); 700} 701 702static int 703xbd_ioctl(struct disk *dp, u_long cmd, void *addr, int flag, struct thread *td) 704{ 705 struct xbd_softc *sc = dp->d_drv1; 706 707 if (sc == NULL) 708 return (ENXIO); 709 710 return (ENOTTY); 711} 712 713/* 714 * Read/write routine for a buffer. Finds the proper unit, place it on 715 * the sortq and kick the controller. 716 */ 717static void 718xbd_strategy(struct bio *bp) 719{ 720 struct xbd_softc *sc = bp->bio_disk->d_drv1; 721 722 /* bogus disk? */ 723 if (sc == NULL) { 724 bp->bio_error = EINVAL; 725 bp->bio_flags |= BIO_ERROR; 726 bp->bio_resid = bp->bio_bcount; 727 biodone(bp); 728 return; 729 } 730 731 /* 732 * Place it in the queue of disk activities for this disk 733 */ 734 mtx_lock(&sc->xbd_io_lock); 735 736 xbd_enqueue_bio(sc, bp); 737 xbd_startio(sc); 738 739 mtx_unlock(&sc->xbd_io_lock); 740 return; 741} 742 743/*------------------------------ Ring Management -----------------------------*/ 744static int 745xbd_alloc_ring(struct xbd_softc *sc) 746{ 747 blkif_sring_t *sring; 748 uintptr_t sring_page_addr; 749 int error; 750 int i; 751 752 sring = malloc(sc->xbd_ring_pages * PAGE_SIZE, M_XENBLOCKFRONT, 753 M_NOWAIT|M_ZERO); 754 if (sring == NULL) { 755 xenbus_dev_fatal(sc->xbd_dev, ENOMEM, "allocating shared ring"); 756 return (ENOMEM); 757 } 758 SHARED_RING_INIT(sring); 759 FRONT_RING_INIT(&sc->xbd_ring, sring, sc->xbd_ring_pages * PAGE_SIZE); 760 761 for (i = 0, sring_page_addr = (uintptr_t)sring; 762 i < sc->xbd_ring_pages; 763 i++, sring_page_addr += PAGE_SIZE) { 764 error = xenbus_grant_ring(sc->xbd_dev, 765 (vtophys(sring_page_addr) >> PAGE_SHIFT), 766 &sc->xbd_ring_ref[i]); 767 if (error) { 768 xenbus_dev_fatal(sc->xbd_dev, error, 769 "granting ring_ref(%d)", i); 770 return (error); 771 } 772 } 773 if (sc->xbd_ring_pages == 1) { 774 error = xs_printf(XST_NIL, xenbus_get_node(sc->xbd_dev), 775 "ring-ref", "%u", sc->xbd_ring_ref[0]); 776 if (error) { 777 xenbus_dev_fatal(sc->xbd_dev, error, 778 "writing %s/ring-ref", 779 xenbus_get_node(sc->xbd_dev)); 780 return (error); 781 } 782 } else { 783 for (i = 0; i < sc->xbd_ring_pages; i++) { 784 char ring_ref_name[]= "ring_refXX"; 785 786 snprintf(ring_ref_name, sizeof(ring_ref_name), 787 "ring-ref%u", i); 788 error = xs_printf(XST_NIL, xenbus_get_node(sc->xbd_dev), 789 ring_ref_name, "%u", sc->xbd_ring_ref[i]); 790 if (error) { 791 xenbus_dev_fatal(sc->xbd_dev, error, 792 "writing %s/%s", 793 xenbus_get_node(sc->xbd_dev), 794 ring_ref_name); 795 return (error); 796 } 797 } 798 } 799 800 error = xen_intr_alloc_and_bind_local_port(sc->xbd_dev, 801 xenbus_get_otherend_id(sc->xbd_dev), NULL, xbd_int, sc, 802 INTR_TYPE_BIO | INTR_MPSAFE, &sc->xen_intr_handle); 803 if (error) { 804 xenbus_dev_fatal(sc->xbd_dev, error, 805 "xen_intr_alloc_and_bind_local_port failed"); 806 return (error); 807 } 808 809 return (0); 810} 811 812static void 813xbd_free_ring(struct xbd_softc *sc) 814{ 815 int i; 816 817 if (sc->xbd_ring.sring == NULL) 818 return; 819 820 for (i = 0; i < sc->xbd_ring_pages; i++) { 821 if (sc->xbd_ring_ref[i] != GRANT_REF_INVALID) { 822 gnttab_end_foreign_access_ref(sc->xbd_ring_ref[i]); 823 sc->xbd_ring_ref[i] = GRANT_REF_INVALID; 824 } 825 } 826 free(sc->xbd_ring.sring, M_XENBLOCKFRONT); 827 sc->xbd_ring.sring = NULL; 828} 829 830/*-------------------------- Initialization/Teardown -------------------------*/ 831static int 832xbd_feature_string(struct xbd_softc *sc, char *features, size_t len) 833{ 834 struct sbuf sb; 835 int feature_cnt; 836 837 sbuf_new(&sb, features, len, SBUF_FIXEDLEN); 838 839 feature_cnt = 0; 840 if ((sc->xbd_flags & XBDF_FLUSH) != 0) { 841 sbuf_printf(&sb, "flush"); 842 feature_cnt++; 843 } 844 845 if ((sc->xbd_flags & XBDF_BARRIER) != 0) { 846 if (feature_cnt != 0) 847 sbuf_printf(&sb, ", "); 848 sbuf_printf(&sb, "write_barrier"); 849 feature_cnt++; 850 } 851 852 if ((sc->xbd_flags & XBDF_DISCARD) != 0) { 853 if (feature_cnt != 0) 854 sbuf_printf(&sb, ", "); 855 sbuf_printf(&sb, "discard"); 856 feature_cnt++; 857 } 858 859 if ((sc->xbd_flags & XBDF_PERSISTENT) != 0) { 860 if (feature_cnt != 0) 861 sbuf_printf(&sb, ", "); 862 sbuf_printf(&sb, "persistent_grants"); 863 feature_cnt++; 864 } 865 866 (void) sbuf_finish(&sb); 867 return (sbuf_len(&sb)); 868} 869 870static int 871xbd_sysctl_features(SYSCTL_HANDLER_ARGS) 872{ 873 char features[80]; 874 struct xbd_softc *sc = arg1; 875 int error; 876 int len; 877 878 error = sysctl_wire_old_buffer(req, 0); 879 if (error != 0) 880 return (error); 881 882 len = xbd_feature_string(sc, features, sizeof(features)); 883 884 /* len is -1 on error, which will make the SYSCTL_OUT a no-op. */ 885 return (SYSCTL_OUT(req, features, len + 1/*NUL*/)); 886} 887 888static void 889xbd_setup_sysctl(struct xbd_softc *xbd) 890{ 891 struct sysctl_ctx_list *sysctl_ctx = NULL; 892 struct sysctl_oid *sysctl_tree = NULL; 893 struct sysctl_oid_list *children; 894 895 sysctl_ctx = device_get_sysctl_ctx(xbd->xbd_dev); 896 if (sysctl_ctx == NULL) 897 return; 898 899 sysctl_tree = device_get_sysctl_tree(xbd->xbd_dev); 900 if (sysctl_tree == NULL) 901 return; 902 903 children = SYSCTL_CHILDREN(sysctl_tree); 904 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 905 "max_requests", CTLFLAG_RD, &xbd->xbd_max_requests, -1, 906 "maximum outstanding requests (negotiated)"); 907 908 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 909 "max_request_segments", CTLFLAG_RD, 910 &xbd->xbd_max_request_segments, 0, 911 "maximum number of pages per requests (negotiated)"); 912 913 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 914 "max_request_size", CTLFLAG_RD, &xbd->xbd_max_request_size, 0, 915 "maximum size in bytes of a request (negotiated)"); 916 917 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 918 "ring_pages", CTLFLAG_RD, &xbd->xbd_ring_pages, 0, 919 "communication channel pages (negotiated)"); 920 921 SYSCTL_ADD_PROC(sysctl_ctx, children, OID_AUTO, 922 "features", CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, xbd, 923 0, xbd_sysctl_features, "A", "protocol features (negotiated)"); 924} 925 926/* 927 * Translate Linux major/minor to an appropriate name and unit 928 * number. For HVM guests, this allows us to use the same drive names 929 * with blkfront as the emulated drives, easing transition slightly. 930 */ 931static void 932xbd_vdevice_to_unit(uint32_t vdevice, int *unit, const char **name) 933{ 934 static struct vdev_info { 935 int major; 936 int shift; 937 int base; 938 const char *name; 939 } info[] = { 940 {3, 6, 0, "ada"}, /* ide0 */ 941 {22, 6, 2, "ada"}, /* ide1 */ 942 {33, 6, 4, "ada"}, /* ide2 */ 943 {34, 6, 6, "ada"}, /* ide3 */ 944 {56, 6, 8, "ada"}, /* ide4 */ 945 {57, 6, 10, "ada"}, /* ide5 */ 946 {88, 6, 12, "ada"}, /* ide6 */ 947 {89, 6, 14, "ada"}, /* ide7 */ 948 {90, 6, 16, "ada"}, /* ide8 */ 949 {91, 6, 18, "ada"}, /* ide9 */ 950 951 {8, 4, 0, "da"}, /* scsi disk0 */ 952 {65, 4, 16, "da"}, /* scsi disk1 */ 953 {66, 4, 32, "da"}, /* scsi disk2 */ 954 {67, 4, 48, "da"}, /* scsi disk3 */ 955 {68, 4, 64, "da"}, /* scsi disk4 */ 956 {69, 4, 80, "da"}, /* scsi disk5 */ 957 {70, 4, 96, "da"}, /* scsi disk6 */ 958 {71, 4, 112, "da"}, /* scsi disk7 */ 959 {128, 4, 128, "da"}, /* scsi disk8 */ 960 {129, 4, 144, "da"}, /* scsi disk9 */ 961 {130, 4, 160, "da"}, /* scsi disk10 */ 962 {131, 4, 176, "da"}, /* scsi disk11 */ 963 {132, 4, 192, "da"}, /* scsi disk12 */ 964 {133, 4, 208, "da"}, /* scsi disk13 */ 965 {134, 4, 224, "da"}, /* scsi disk14 */ 966 {135, 4, 240, "da"}, /* scsi disk15 */ 967 968 {202, 4, 0, "xbd"}, /* xbd */ 969 970 {0, 0, 0, NULL}, 971 }; 972 int major = vdevice >> 8; 973 int minor = vdevice & 0xff; 974 int i; 975 976 if (vdevice & (1 << 28)) { 977 *unit = (vdevice & ((1 << 28) - 1)) >> 8; 978 *name = "xbd"; 979 return; 980 } 981 982 for (i = 0; info[i].major; i++) { 983 if (info[i].major == major) { 984 *unit = info[i].base + (minor >> info[i].shift); 985 *name = info[i].name; 986 return; 987 } 988 } 989 990 *unit = minor >> 4; 991 *name = "xbd"; 992} 993 994int 995xbd_instance_create(struct xbd_softc *sc, blkif_sector_t sectors, 996 int vdevice, uint16_t vdisk_info, unsigned long sector_size, 997 unsigned long phys_sector_size) 998{ 999 char features[80]; 1000 int unit, error = 0; 1001 const char *name; 1002 1003 xbd_vdevice_to_unit(vdevice, &unit, &name); 1004 1005 sc->xbd_unit = unit; 1006 1007 if (strcmp(name, "xbd") != 0) 1008 device_printf(sc->xbd_dev, "attaching as %s%d\n", name, unit); 1009 1010 if (xbd_feature_string(sc, features, sizeof(features)) > 0) { 1011 device_printf(sc->xbd_dev, "features: %s\n", 1012 features); 1013 } 1014 1015 sc->xbd_disk = disk_alloc(); 1016 sc->xbd_disk->d_unit = sc->xbd_unit; 1017 sc->xbd_disk->d_open = xbd_open; 1018 sc->xbd_disk->d_close = xbd_close; 1019 sc->xbd_disk->d_ioctl = xbd_ioctl; 1020 sc->xbd_disk->d_strategy = xbd_strategy; 1021 sc->xbd_disk->d_dump = xbd_dump; 1022 sc->xbd_disk->d_name = name; 1023 sc->xbd_disk->d_drv1 = sc; 1024 sc->xbd_disk->d_sectorsize = sector_size; 1025 sc->xbd_disk->d_stripesize = phys_sector_size; 1026 sc->xbd_disk->d_stripeoffset = 0; 1027 1028 sc->xbd_disk->d_mediasize = sectors * sector_size; 1029 sc->xbd_disk->d_maxsize = sc->xbd_max_request_size; 1030 sc->xbd_disk->d_flags = DISKFLAG_UNMAPPED_BIO; 1031 if ((sc->xbd_flags & (XBDF_FLUSH|XBDF_BARRIER)) != 0) { 1032 sc->xbd_disk->d_flags |= DISKFLAG_CANFLUSHCACHE; 1033 device_printf(sc->xbd_dev, 1034 "synchronize cache commands enabled.\n"); 1035 } 1036 disk_create(sc->xbd_disk, DISK_VERSION); 1037 1038 return error; 1039} 1040 1041static void 1042xbd_free(struct xbd_softc *sc) 1043{ 1044 int i; 1045 1046 /* Prevent new requests being issued until we fix things up. */ 1047 mtx_lock(&sc->xbd_io_lock); 1048 sc->xbd_state = XBD_STATE_DISCONNECTED; 1049 mtx_unlock(&sc->xbd_io_lock); 1050 1051 /* Free resources associated with old device channel. */ 1052 xbd_free_ring(sc); 1053 if (sc->xbd_shadow) { 1054 for (i = 0; i < sc->xbd_max_requests; i++) { 1055 struct xbd_command *cm; 1056 1057 cm = &sc->xbd_shadow[i]; 1058 if (cm->cm_sg_refs != NULL) { 1059 free(cm->cm_sg_refs, M_XENBLOCKFRONT); 1060 cm->cm_sg_refs = NULL; 1061 } 1062 1063 if (cm->cm_indirectionpages != NULL) { 1064 gnttab_end_foreign_access_references( 1065 sc->xbd_max_request_indirectpages, 1066 &cm->cm_indirectionrefs[0]); 1067 contigfree(cm->cm_indirectionpages, PAGE_SIZE * 1068 sc->xbd_max_request_indirectpages, 1069 M_XENBLOCKFRONT); 1070 cm->cm_indirectionpages = NULL; 1071 } 1072 1073 bus_dmamap_destroy(sc->xbd_io_dmat, cm->cm_map); 1074 } 1075 free(sc->xbd_shadow, M_XENBLOCKFRONT); 1076 sc->xbd_shadow = NULL; 1077 1078 bus_dma_tag_destroy(sc->xbd_io_dmat); 1079 1080 xbd_initq_cm(sc, XBD_Q_FREE); 1081 xbd_initq_cm(sc, XBD_Q_READY); 1082 xbd_initq_cm(sc, XBD_Q_COMPLETE); 1083 } 1084 1085 xen_intr_unbind(&sc->xen_intr_handle); 1086 1087} 1088 1089/*--------------------------- State Change Handlers --------------------------*/ 1090static void 1091xbd_initialize(struct xbd_softc *sc) 1092{ 1093 const char *otherend_path; 1094 const char *node_path; 1095 uint32_t max_ring_page_order; 1096 int error; 1097 1098 if (xenbus_get_state(sc->xbd_dev) != XenbusStateInitialising) { 1099 /* Initialization has already been performed. */ 1100 return; 1101 } 1102 1103 /* 1104 * Protocol defaults valid even if negotiation for a 1105 * setting fails. 1106 */ 1107 max_ring_page_order = 0; 1108 sc->xbd_ring_pages = 1; 1109 1110 /* 1111 * Protocol negotiation. 1112 * 1113 * \note xs_gather() returns on the first encountered error, so 1114 * we must use independent calls in order to guarantee 1115 * we don't miss information in a sparsly populated back-end 1116 * tree. 1117 * 1118 * \note xs_scanf() does not update variables for unmatched 1119 * fields. 1120 */ 1121 otherend_path = xenbus_get_otherend_path(sc->xbd_dev); 1122 node_path = xenbus_get_node(sc->xbd_dev); 1123 1124 /* Support both backend schemes for relaying ring page limits. */ 1125 (void)xs_scanf(XST_NIL, otherend_path, 1126 "max-ring-page-order", NULL, "%" PRIu32, 1127 &max_ring_page_order); 1128 sc->xbd_ring_pages = 1 << max_ring_page_order; 1129 (void)xs_scanf(XST_NIL, otherend_path, 1130 "max-ring-pages", NULL, "%" PRIu32, 1131 &sc->xbd_ring_pages); 1132 if (sc->xbd_ring_pages < 1) 1133 sc->xbd_ring_pages = 1; 1134 1135 if (sc->xbd_ring_pages > XBD_MAX_RING_PAGES) { 1136 device_printf(sc->xbd_dev, 1137 "Back-end specified ring-pages of %u " 1138 "limited to front-end limit of %u.\n", 1139 sc->xbd_ring_pages, XBD_MAX_RING_PAGES); 1140 sc->xbd_ring_pages = XBD_MAX_RING_PAGES; 1141 } 1142 1143 if (powerof2(sc->xbd_ring_pages) == 0) { 1144 uint32_t new_page_limit; 1145 1146 new_page_limit = 0x01 << (fls(sc->xbd_ring_pages) - 1); 1147 device_printf(sc->xbd_dev, 1148 "Back-end specified ring-pages of %u " 1149 "is not a power of 2. Limited to %u.\n", 1150 sc->xbd_ring_pages, new_page_limit); 1151 sc->xbd_ring_pages = new_page_limit; 1152 } 1153 1154 sc->xbd_max_requests = 1155 BLKIF_MAX_RING_REQUESTS(sc->xbd_ring_pages * PAGE_SIZE); 1156 if (sc->xbd_max_requests > XBD_MAX_REQUESTS) { 1157 device_printf(sc->xbd_dev, 1158 "Back-end specified max_requests of %u " 1159 "limited to front-end limit of %zu.\n", 1160 sc->xbd_max_requests, XBD_MAX_REQUESTS); 1161 sc->xbd_max_requests = XBD_MAX_REQUESTS; 1162 } 1163 1164 if (xbd_alloc_ring(sc) != 0) 1165 return; 1166 1167 /* Support both backend schemes for relaying ring page limits. */ 1168 if (sc->xbd_ring_pages > 1) { 1169 error = xs_printf(XST_NIL, node_path, 1170 "num-ring-pages","%u", 1171 sc->xbd_ring_pages); 1172 if (error) { 1173 xenbus_dev_fatal(sc->xbd_dev, error, 1174 "writing %s/num-ring-pages", 1175 node_path); 1176 return; 1177 } 1178 1179 error = xs_printf(XST_NIL, node_path, 1180 "ring-page-order", "%u", 1181 fls(sc->xbd_ring_pages) - 1); 1182 if (error) { 1183 xenbus_dev_fatal(sc->xbd_dev, error, 1184 "writing %s/ring-page-order", 1185 node_path); 1186 return; 1187 } 1188 } 1189 1190 error = xs_printf(XST_NIL, node_path, "event-channel", 1191 "%u", xen_intr_port(sc->xen_intr_handle)); 1192 if (error) { 1193 xenbus_dev_fatal(sc->xbd_dev, error, 1194 "writing %s/event-channel", 1195 node_path); 1196 return; 1197 } 1198 1199 error = xs_printf(XST_NIL, node_path, "protocol", 1200 "%s", XEN_IO_PROTO_ABI_NATIVE); 1201 if (error) { 1202 xenbus_dev_fatal(sc->xbd_dev, error, 1203 "writing %s/protocol", 1204 node_path); 1205 return; 1206 } 1207 1208 xenbus_set_state(sc->xbd_dev, XenbusStateInitialised); 1209} 1210 1211/* 1212 * Invoked when the backend is finally 'ready' (and has published 1213 * the details about the physical device - #sectors, size, etc). 1214 */ 1215static void 1216xbd_connect(struct xbd_softc *sc) 1217{ 1218 device_t dev = sc->xbd_dev; 1219 blkif_sector_t sectors; 1220 unsigned long sector_size, phys_sector_size; 1221 unsigned int binfo; 1222 int err, feature_barrier, feature_flush; 1223 int i, j; 1224 1225 DPRINTK("blkfront.c:connect:%s.\n", xenbus_get_otherend_path(dev)); 1226 1227 if (sc->xbd_state == XBD_STATE_SUSPENDED) { 1228 return; 1229 } 1230 1231 if (sc->xbd_state == XBD_STATE_CONNECTED) { 1232 struct disk *disk; 1233 1234 disk = sc->xbd_disk; 1235 if (disk == NULL) { 1236 return; 1237 } 1238 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1239 "sectors", "%"PRIu64, §ors, NULL); 1240 if (err != 0) { 1241 xenbus_dev_error(dev, err, 1242 "reading sectors at %s", 1243 xenbus_get_otherend_path(dev)); 1244 return; 1245 } 1246 disk->d_mediasize = disk->d_sectorsize * sectors; 1247 err = disk_resize(disk, M_NOWAIT); 1248 if (err) { 1249 xenbus_dev_error(dev, err, 1250 "unable to resize disk %s%u", 1251 disk->d_name, disk->d_unit); 1252 return; 1253 } 1254 device_printf(sc->xbd_dev, 1255 "changed capacity to %jd\n", 1256 (intmax_t)disk->d_mediasize); 1257 return; 1258 } 1259 1260 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1261 "sectors", "%"PRIu64, §ors, 1262 "info", "%u", &binfo, 1263 "sector-size", "%lu", §or_size, 1264 NULL); 1265 if (err) { 1266 xenbus_dev_fatal(dev, err, 1267 "reading backend fields at %s", 1268 xenbus_get_otherend_path(dev)); 1269 return; 1270 } 1271 if ((sectors == 0) || (sector_size == 0)) { 1272 xenbus_dev_fatal(dev, 0, 1273 "invalid parameters from %s:" 1274 " sectors = %"PRIu64", sector_size = %lu", 1275 xenbus_get_otherend_path(dev), 1276 sectors, sector_size); 1277 return; 1278 } 1279 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1280 "physical-sector-size", "%lu", &phys_sector_size, 1281 NULL); 1282 if (err || phys_sector_size <= sector_size) 1283 phys_sector_size = 0; 1284 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1285 "feature-barrier", "%d", &feature_barrier, 1286 NULL); 1287 if (err == 0 && feature_barrier != 0) 1288 sc->xbd_flags |= XBDF_BARRIER; 1289 1290 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1291 "feature-flush-cache", "%d", &feature_flush, 1292 NULL); 1293 if (err == 0 && feature_flush != 0) 1294 sc->xbd_flags |= XBDF_FLUSH; 1295 1296 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1297 "feature-max-indirect-segments", "%" PRIu32, 1298 &sc->xbd_max_request_segments, NULL); 1299 if ((err != 0) || (xbd_enable_indirect == 0)) 1300 sc->xbd_max_request_segments = 0; 1301 if (sc->xbd_max_request_segments > XBD_MAX_INDIRECT_SEGMENTS) 1302 sc->xbd_max_request_segments = XBD_MAX_INDIRECT_SEGMENTS; 1303 if (sc->xbd_max_request_segments > XBD_SIZE_TO_SEGS(maxphys)) 1304 sc->xbd_max_request_segments = XBD_SIZE_TO_SEGS(maxphys); 1305 sc->xbd_max_request_indirectpages = 1306 XBD_INDIRECT_SEGS_TO_PAGES(sc->xbd_max_request_segments); 1307 if (sc->xbd_max_request_segments < BLKIF_MAX_SEGMENTS_PER_REQUEST) 1308 sc->xbd_max_request_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; 1309 sc->xbd_max_request_size = 1310 XBD_SEGS_TO_SIZE(sc->xbd_max_request_segments); 1311 1312 /* Allocate datastructures based on negotiated values. */ 1313 err = bus_dma_tag_create( 1314 bus_get_dma_tag(sc->xbd_dev), /* parent */ 1315 512, PAGE_SIZE, /* algnmnt, boundary */ 1316 BUS_SPACE_MAXADDR, /* lowaddr */ 1317 BUS_SPACE_MAXADDR, /* highaddr */ 1318 NULL, NULL, /* filter, filterarg */ 1319 sc->xbd_max_request_size, 1320 sc->xbd_max_request_segments, 1321 PAGE_SIZE, /* maxsegsize */ 1322 BUS_DMA_ALLOCNOW, /* flags */ 1323 busdma_lock_mutex, /* lockfunc */ 1324 &sc->xbd_io_lock, /* lockarg */ 1325 &sc->xbd_io_dmat); 1326 if (err != 0) { 1327 xenbus_dev_fatal(sc->xbd_dev, err, 1328 "Cannot allocate parent DMA tag\n"); 1329 return; 1330 } 1331 1332 /* Per-transaction data allocation. */ 1333 sc->xbd_shadow = malloc(sizeof(*sc->xbd_shadow) * sc->xbd_max_requests, 1334 M_XENBLOCKFRONT, M_NOWAIT|M_ZERO); 1335 if (sc->xbd_shadow == NULL) { 1336 bus_dma_tag_destroy(sc->xbd_io_dmat); 1337 xenbus_dev_fatal(sc->xbd_dev, ENOMEM, 1338 "Cannot allocate request structures\n"); 1339 return; 1340 } 1341 1342 for (i = 0; i < sc->xbd_max_requests; i++) { 1343 struct xbd_command *cm; 1344 void * indirectpages; 1345 1346 cm = &sc->xbd_shadow[i]; 1347 cm->cm_sg_refs = malloc( 1348 sizeof(grant_ref_t) * sc->xbd_max_request_segments, 1349 M_XENBLOCKFRONT, M_NOWAIT); 1350 if (cm->cm_sg_refs == NULL) 1351 break; 1352 cm->cm_id = i; 1353 cm->cm_flags = XBDCF_INITIALIZER; 1354 cm->cm_sc = sc; 1355 if (bus_dmamap_create(sc->xbd_io_dmat, 0, &cm->cm_map) != 0) 1356 break; 1357 if (sc->xbd_max_request_indirectpages > 0) { 1358 indirectpages = contigmalloc( 1359 PAGE_SIZE * sc->xbd_max_request_indirectpages, 1360 M_XENBLOCKFRONT, M_ZERO | M_NOWAIT, 0, ~0, 1361 PAGE_SIZE, 0); 1362 if (indirectpages == NULL) 1363 sc->xbd_max_request_indirectpages = 0; 1364 } else { 1365 indirectpages = NULL; 1366 } 1367 for (j = 0; j < sc->xbd_max_request_indirectpages; j++) { 1368 if (gnttab_grant_foreign_access( 1369 xenbus_get_otherend_id(sc->xbd_dev), 1370 (vtophys(indirectpages) >> PAGE_SHIFT) + j, 1371 1 /* grant read-only access */, 1372 &cm->cm_indirectionrefs[j])) 1373 break; 1374 } 1375 if (j < sc->xbd_max_request_indirectpages) { 1376 contigfree(indirectpages, 1377 PAGE_SIZE * sc->xbd_max_request_indirectpages, 1378 M_XENBLOCKFRONT); 1379 break; 1380 } 1381 cm->cm_indirectionpages = indirectpages; 1382 xbd_free_command(cm); 1383 } 1384 1385 if (sc->xbd_disk == NULL) { 1386 device_printf(dev, "%juMB <%s> at %s", 1387 (uintmax_t) sectors / (1048576 / sector_size), 1388 device_get_desc(dev), 1389 xenbus_get_node(dev)); 1390 bus_print_child_footer(device_get_parent(dev), dev); 1391 1392 xbd_instance_create(sc, sectors, sc->xbd_vdevice, binfo, 1393 sector_size, phys_sector_size); 1394 } 1395 1396 (void)xenbus_set_state(dev, XenbusStateConnected); 1397 1398 /* Kick pending requests. */ 1399 mtx_lock(&sc->xbd_io_lock); 1400 sc->xbd_state = XBD_STATE_CONNECTED; 1401 xbd_startio(sc); 1402 sc->xbd_flags |= XBDF_READY; 1403 mtx_unlock(&sc->xbd_io_lock); 1404} 1405 1406/** 1407 * Handle the change of state of the backend to Closing. We must delete our 1408 * device-layer structures now, to ensure that writes are flushed through to 1409 * the backend. Once this is done, we can switch to Closed in 1410 * acknowledgement. 1411 */ 1412static void 1413xbd_closing(device_t dev) 1414{ 1415 struct xbd_softc *sc = device_get_softc(dev); 1416 1417 xenbus_set_state(dev, XenbusStateClosing); 1418 1419 DPRINTK("xbd_closing: %s removed\n", xenbus_get_node(dev)); 1420 1421 if (sc->xbd_disk != NULL) { 1422 disk_destroy(sc->xbd_disk); 1423 sc->xbd_disk = NULL; 1424 } 1425 1426 xenbus_set_state(dev, XenbusStateClosed); 1427} 1428 1429/*---------------------------- NewBus Entrypoints ----------------------------*/ 1430static int 1431xbd_probe(device_t dev) 1432{ 1433 if (strcmp(xenbus_get_type(dev), "vbd") != 0) 1434 return (ENXIO); 1435 1436 if (xen_pv_disks_disabled()) 1437 return (ENXIO); 1438 1439 if (xen_hvm_domain()) { 1440 int error; 1441 char *type; 1442 1443 /* 1444 * When running in an HVM domain, IDE disk emulation is 1445 * disabled early in boot so that native drivers will 1446 * not see emulated hardware. However, CDROM device 1447 * emulation cannot be disabled. 1448 * 1449 * Through use of FreeBSD's vm_guest and xen_hvm_domain() 1450 * APIs, we could modify the native CDROM driver to fail its 1451 * probe when running under Xen. Unfortunatlely, the PV 1452 * CDROM support in XenServer (up through at least version 1453 * 6.2) isn't functional, so we instead rely on the emulated 1454 * CDROM instance, and fail to attach the PV one here in 1455 * the blkfront driver. 1456 */ 1457 error = xs_read(XST_NIL, xenbus_get_node(dev), 1458 "device-type", NULL, (void **) &type); 1459 if (error) 1460 return (ENXIO); 1461 1462 if (strncmp(type, "cdrom", 5) == 0) { 1463 free(type, M_XENSTORE); 1464 return (ENXIO); 1465 } 1466 free(type, M_XENSTORE); 1467 } 1468 1469 device_set_desc(dev, "Virtual Block Device"); 1470 device_quiet(dev); 1471 return (0); 1472} 1473 1474/* 1475 * Setup supplies the backend dir, virtual device. We place an event 1476 * channel and shared frame entries. We watch backend to wait if it's 1477 * ok. 1478 */ 1479static int 1480xbd_attach(device_t dev) 1481{ 1482 struct xbd_softc *sc; 1483 const char *name; 1484 uint32_t vdevice; 1485 int error; 1486 int i; 1487 int unit; 1488 1489 /* FIXME: Use dynamic device id if this is not set. */ 1490 error = xs_scanf(XST_NIL, xenbus_get_node(dev), 1491 "virtual-device", NULL, "%" PRIu32, &vdevice); 1492 if (error) 1493 error = xs_scanf(XST_NIL, xenbus_get_node(dev), 1494 "virtual-device-ext", NULL, "%" PRIu32, &vdevice); 1495 if (error) { 1496 xenbus_dev_fatal(dev, error, "reading virtual-device"); 1497 device_printf(dev, "Couldn't determine virtual device.\n"); 1498 return (error); 1499 } 1500 1501 xbd_vdevice_to_unit(vdevice, &unit, &name); 1502 if (!strcmp(name, "xbd")) 1503 device_set_unit(dev, unit); 1504 1505 sc = device_get_softc(dev); 1506 mtx_init(&sc->xbd_io_lock, "blkfront i/o lock", NULL, MTX_DEF); 1507 xbd_initqs(sc); 1508 for (i = 0; i < XBD_MAX_RING_PAGES; i++) 1509 sc->xbd_ring_ref[i] = GRANT_REF_INVALID; 1510 1511 sc->xbd_dev = dev; 1512 sc->xbd_vdevice = vdevice; 1513 sc->xbd_state = XBD_STATE_DISCONNECTED; 1514 1515 xbd_setup_sysctl(sc); 1516 1517 /* Wait for backend device to publish its protocol capabilities. */ 1518 xenbus_set_state(dev, XenbusStateInitialising); 1519 1520 return (0); 1521} 1522 1523static int 1524xbd_detach(device_t dev) 1525{ 1526 struct xbd_softc *sc = device_get_softc(dev); 1527 1528 DPRINTK("%s: %s removed\n", __func__, xenbus_get_node(dev)); 1529 1530 xbd_free(sc); 1531 mtx_destroy(&sc->xbd_io_lock); 1532 1533 return 0; 1534} 1535 1536static int 1537xbd_suspend(device_t dev) 1538{ 1539 struct xbd_softc *sc = device_get_softc(dev); 1540 int retval; 1541 int saved_state; 1542 1543 /* Prevent new requests being issued until we fix things up. */ 1544 mtx_lock(&sc->xbd_io_lock); 1545 saved_state = sc->xbd_state; 1546 sc->xbd_state = XBD_STATE_SUSPENDED; 1547 1548 /* Wait for outstanding I/O to drain. */ 1549 retval = 0; 1550 while (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 1551 if (msleep(&sc->xbd_cm_q[XBD_Q_BUSY], &sc->xbd_io_lock, 1552 PRIBIO, "blkf_susp", 30 * hz) == EWOULDBLOCK) { 1553 retval = EBUSY; 1554 break; 1555 } 1556 } 1557 mtx_unlock(&sc->xbd_io_lock); 1558 1559 if (retval != 0) 1560 sc->xbd_state = saved_state; 1561 1562 return (retval); 1563} 1564 1565static int 1566xbd_resume(device_t dev) 1567{ 1568 struct xbd_softc *sc = device_get_softc(dev); 1569 1570 if (xen_suspend_cancelled) { 1571 sc->xbd_state = XBD_STATE_CONNECTED; 1572 return (0); 1573 } 1574 1575 DPRINTK("xbd_resume: %s\n", xenbus_get_node(dev)); 1576 1577 xbd_free(sc); 1578 xbd_initialize(sc); 1579 return (0); 1580} 1581 1582/** 1583 * Callback received when the backend's state changes. 1584 */ 1585static void 1586xbd_backend_changed(device_t dev, XenbusState backend_state) 1587{ 1588 struct xbd_softc *sc = device_get_softc(dev); 1589 1590 DPRINTK("backend_state=%d\n", backend_state); 1591 1592 switch (backend_state) { 1593 case XenbusStateUnknown: 1594 case XenbusStateInitialising: 1595 case XenbusStateReconfigured: 1596 case XenbusStateReconfiguring: 1597 case XenbusStateClosed: 1598 break; 1599 1600 case XenbusStateInitWait: 1601 case XenbusStateInitialised: 1602 xbd_initialize(sc); 1603 break; 1604 1605 case XenbusStateConnected: 1606 xbd_initialize(sc); 1607 xbd_connect(sc); 1608 break; 1609 1610 case XenbusStateClosing: 1611 if (sc->xbd_users > 0) { 1612 device_printf(dev, "detaching with pending users\n"); 1613 KASSERT(sc->xbd_disk != NULL, 1614 ("NULL disk with pending users\n")); 1615 disk_gone(sc->xbd_disk); 1616 } else { 1617 xbd_closing(dev); 1618 } 1619 break; 1620 } 1621} 1622 1623/*---------------------------- NewBus Registration ---------------------------*/ 1624static device_method_t xbd_methods[] = { 1625 /* Device interface */ 1626 DEVMETHOD(device_probe, xbd_probe), 1627 DEVMETHOD(device_attach, xbd_attach), 1628 DEVMETHOD(device_detach, xbd_detach), 1629 DEVMETHOD(device_shutdown, bus_generic_shutdown), 1630 DEVMETHOD(device_suspend, xbd_suspend), 1631 DEVMETHOD(device_resume, xbd_resume), 1632 1633 /* Xenbus interface */ 1634 DEVMETHOD(xenbus_otherend_changed, xbd_backend_changed), 1635 1636 DEVMETHOD_END 1637}; 1638 1639static driver_t xbd_driver = { 1640 "xbd", 1641 xbd_methods, 1642 sizeof(struct xbd_softc), 1643}; 1644 1645DRIVER_MODULE(xbd, xenbusb_front, xbd_driver, 0, 0); 1646