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 ---------------------------*/ 66 static void xbd_closing(device_t); 67 static 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 ----------------------------*/ 79 static MALLOC_DEFINE(M_XENBLOCKFRONT, "xbd", "Xen Block Front driver data"); 80 81 static int xbd_enable_indirect = 1; 82 SYSCTL_NODE(_hw, OID_AUTO, xbd, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 83 "xbd driver parameters"); 84 SYSCTL_INT(_hw_xbd, OID_AUTO, xbd_enable_indirect, CTLFLAG_RDTUN, 85 &xbd_enable_indirect, 0, "Enable xbd indirect segments"); 86 87 /*---------------------------- Command Processing ----------------------------*/ 88 static void 89 xbd_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 98 static void 99 xbd_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 } 111 112 static void 113 xbd_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 } 121 122 static void 123 xbd_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 } 131 132 static inline void 133 xbd_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 } 142 143 static void 144 xbd_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 } 157 158 static void 159 xbd_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 unsigned int sector_size) 163 { 164 struct blkif_request_segment *last_block_sg = sg + nsegs; 165 vm_paddr_t buffer_ma; 166 uint64_t fsect, lsect; 167 int ref; 168 169 while (sg < last_block_sg) { 170 KASSERT((segs->ds_addr & (sector_size - 1)) == 0, 171 ("XEN disk driver I/O must be sector aligned")); 172 KASSERT((segs->ds_len & (sector_size - 1)) == 0, 173 ("XEN disk driver I/Os must be a multiple of " 174 "the sector length")); 175 buffer_ma = segs->ds_addr; 176 fsect = (buffer_ma & PAGE_MASK) >> XBD_SECTOR_SHFT; 177 lsect = fsect + (segs->ds_len >> XBD_SECTOR_SHFT) - 1; 178 179 KASSERT(lsect <= 7, ("XEN disk driver data cannot " 180 "cross a page boundary")); 181 182 /* install a grant reference. */ 183 ref = gnttab_claim_grant_reference(gref_head); 184 185 /* 186 * GNTTAB_LIST_END == 0xffffffff, but it is private 187 * to gnttab.c. 188 */ 189 KASSERT(ref != ~0, ("grant_reference failed")); 190 191 gnttab_grant_foreign_access_ref( 192 ref, 193 otherend_id, 194 buffer_ma >> PAGE_SHIFT, 195 readonly); 196 197 *sg_ref = ref; 198 *sg = (struct blkif_request_segment) { 199 .gref = ref, 200 .first_sect = fsect, 201 .last_sect = lsect 202 }; 203 sg++; 204 sg_ref++; 205 segs++; 206 } 207 } 208 209 static void 210 xbd_queue_cb(void *arg, bus_dma_segment_t *segs, int nsegs, int error) 211 { 212 struct xbd_softc *sc; 213 struct xbd_command *cm; 214 int op; 215 216 cm = arg; 217 sc = cm->cm_sc; 218 219 if (error) { 220 cm->cm_bp->bio_error = EIO; 221 biodone(cm->cm_bp); 222 xbd_free_command(cm); 223 return; 224 } 225 226 KASSERT(nsegs <= sc->xbd_max_request_segments, 227 ("Too many segments in a blkfront I/O")); 228 229 if (nsegs <= BLKIF_MAX_SEGMENTS_PER_REQUEST) { 230 blkif_request_t *ring_req; 231 232 /* Fill out a blkif_request_t structure. */ 233 ring_req = (blkif_request_t *) 234 RING_GET_REQUEST(&sc->xbd_ring, sc->xbd_ring.req_prod_pvt); 235 sc->xbd_ring.req_prod_pvt++; 236 ring_req->id = cm->cm_id; 237 ring_req->operation = cm->cm_operation; 238 ring_req->sector_number = cm->cm_sector_number; 239 ring_req->handle = (blkif_vdev_t)(uintptr_t)sc->xbd_disk; 240 ring_req->nr_segments = nsegs; 241 cm->cm_nseg = nsegs; 242 xbd_mksegarray(segs, nsegs, &cm->cm_gref_head, 243 xenbus_get_otherend_id(sc->xbd_dev), 244 cm->cm_operation == BLKIF_OP_WRITE, 245 cm->cm_sg_refs, ring_req->seg, 246 sc->xbd_disk->d_sectorsize); 247 } else { 248 blkif_request_indirect_t *ring_req; 249 250 /* Fill out a blkif_request_indirect_t structure. */ 251 ring_req = (blkif_request_indirect_t *) 252 RING_GET_REQUEST(&sc->xbd_ring, sc->xbd_ring.req_prod_pvt); 253 sc->xbd_ring.req_prod_pvt++; 254 ring_req->id = cm->cm_id; 255 ring_req->operation = BLKIF_OP_INDIRECT; 256 ring_req->indirect_op = cm->cm_operation; 257 ring_req->sector_number = cm->cm_sector_number; 258 ring_req->handle = (blkif_vdev_t)(uintptr_t)sc->xbd_disk; 259 ring_req->nr_segments = nsegs; 260 cm->cm_nseg = nsegs; 261 xbd_mksegarray(segs, nsegs, &cm->cm_gref_head, 262 xenbus_get_otherend_id(sc->xbd_dev), 263 cm->cm_operation == BLKIF_OP_WRITE, 264 cm->cm_sg_refs, cm->cm_indirectionpages, 265 sc->xbd_disk->d_sectorsize); 266 memcpy(ring_req->indirect_grefs, &cm->cm_indirectionrefs, 267 sizeof(grant_ref_t) * sc->xbd_max_request_indirectpages); 268 } 269 270 if (cm->cm_operation == BLKIF_OP_READ) 271 op = BUS_DMASYNC_PREREAD; 272 else if (cm->cm_operation == BLKIF_OP_WRITE) 273 op = BUS_DMASYNC_PREWRITE; 274 else 275 op = 0; 276 bus_dmamap_sync(sc->xbd_io_dmat, cm->cm_map, op); 277 278 gnttab_free_grant_references(cm->cm_gref_head); 279 280 xbd_enqueue_cm(cm, XBD_Q_BUSY); 281 282 /* 283 * If bus dma had to asynchronously call us back to dispatch 284 * this command, we are no longer executing in the context of 285 * xbd_startio(). Thus we cannot rely on xbd_startio()'s call to 286 * xbd_flush_requests() to publish this command to the backend 287 * along with any other commands that it could batch. 288 */ 289 if ((cm->cm_flags & XBDCF_ASYNC_MAPPING) != 0) 290 xbd_flush_requests(sc); 291 292 return; 293 } 294 295 static int 296 xbd_queue_request(struct xbd_softc *sc, struct xbd_command *cm) 297 { 298 int error; 299 300 if (cm->cm_bp != NULL) 301 error = bus_dmamap_load_bio(sc->xbd_io_dmat, cm->cm_map, 302 cm->cm_bp, xbd_queue_cb, cm, 0); 303 else 304 error = bus_dmamap_load(sc->xbd_io_dmat, cm->cm_map, 305 cm->cm_data, cm->cm_datalen, xbd_queue_cb, cm, 0); 306 if (error == EINPROGRESS) { 307 /* 308 * Maintain queuing order by freezing the queue. The next 309 * command may not require as many resources as the command 310 * we just attempted to map, so we can't rely on bus dma 311 * blocking for it too. 312 */ 313 xbd_cm_freeze(sc, cm, XBDCF_ASYNC_MAPPING); 314 return (0); 315 } 316 317 return (error); 318 } 319 320 static void 321 xbd_restart_queue_callback(void *arg) 322 { 323 struct xbd_softc *sc = arg; 324 325 mtx_lock(&sc->xbd_io_lock); 326 327 xbd_thaw(sc, XBDF_GNT_SHORTAGE); 328 329 xbd_startio(sc); 330 331 mtx_unlock(&sc->xbd_io_lock); 332 } 333 334 static struct xbd_command * 335 xbd_bio_command(struct xbd_softc *sc) 336 { 337 struct xbd_command *cm; 338 struct bio *bp; 339 340 if (__predict_false(sc->xbd_state != XBD_STATE_CONNECTED)) 341 return (NULL); 342 343 bp = xbd_dequeue_bio(sc); 344 if (bp == NULL) 345 return (NULL); 346 347 if ((cm = xbd_dequeue_cm(sc, XBD_Q_FREE)) == NULL) { 348 xbd_freeze(sc, XBDF_CM_SHORTAGE); 349 xbd_requeue_bio(sc, bp); 350 return (NULL); 351 } 352 353 if (gnttab_alloc_grant_references(sc->xbd_max_request_segments, 354 &cm->cm_gref_head) != 0) { 355 gnttab_request_free_callback(&sc->xbd_callback, 356 xbd_restart_queue_callback, sc, 357 sc->xbd_max_request_segments); 358 xbd_freeze(sc, XBDF_GNT_SHORTAGE); 359 xbd_requeue_bio(sc, bp); 360 xbd_enqueue_cm(cm, XBD_Q_FREE); 361 return (NULL); 362 } 363 364 cm->cm_bp = bp; 365 cm->cm_sector_number = 366 ((blkif_sector_t)bp->bio_pblkno * sc->xbd_disk->d_sectorsize) >> 367 XBD_SECTOR_SHFT; 368 369 switch (bp->bio_cmd) { 370 case BIO_READ: 371 cm->cm_operation = BLKIF_OP_READ; 372 break; 373 case BIO_WRITE: 374 cm->cm_operation = BLKIF_OP_WRITE; 375 if ((bp->bio_flags & BIO_ORDERED) != 0) { 376 if ((sc->xbd_flags & XBDF_BARRIER) != 0) { 377 cm->cm_operation = BLKIF_OP_WRITE_BARRIER; 378 } else { 379 /* 380 * Single step this command. 381 */ 382 cm->cm_flags |= XBDCF_Q_FREEZE; 383 if (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 384 /* 385 * Wait for in-flight requests to 386 * finish. 387 */ 388 xbd_freeze(sc, XBDF_WAIT_IDLE); 389 xbd_requeue_cm(cm, XBD_Q_READY); 390 return (NULL); 391 } 392 } 393 } 394 break; 395 case BIO_FLUSH: 396 if ((sc->xbd_flags & XBDF_FLUSH) != 0) 397 cm->cm_operation = BLKIF_OP_FLUSH_DISKCACHE; 398 else if ((sc->xbd_flags & XBDF_BARRIER) != 0) 399 cm->cm_operation = BLKIF_OP_WRITE_BARRIER; 400 else 401 panic("flush request, but no flush support available"); 402 break; 403 default: 404 biofinish(bp, NULL, EOPNOTSUPP); 405 xbd_enqueue_cm(cm, XBD_Q_FREE); 406 return (NULL); 407 } 408 409 return (cm); 410 } 411 412 /* 413 * Dequeue buffers and place them in the shared communication ring. 414 * Return when no more requests can be accepted or all buffers have 415 * been queued. 416 * 417 * Signal XEN once the ring has been filled out. 418 */ 419 static void 420 xbd_startio(struct xbd_softc *sc) 421 { 422 struct xbd_command *cm; 423 int error, queued = 0; 424 425 mtx_assert(&sc->xbd_io_lock, MA_OWNED); 426 427 if (sc->xbd_state != XBD_STATE_CONNECTED) 428 return; 429 430 while (!RING_FULL(&sc->xbd_ring)) { 431 if (sc->xbd_qfrozen_cnt != 0) 432 break; 433 434 cm = xbd_dequeue_cm(sc, XBD_Q_READY); 435 436 if (cm == NULL) 437 cm = xbd_bio_command(sc); 438 439 if (cm == NULL) 440 break; 441 442 if ((cm->cm_flags & XBDCF_Q_FREEZE) != 0) { 443 /* 444 * Single step command. Future work is 445 * held off until this command completes. 446 */ 447 xbd_cm_freeze(sc, cm, XBDCF_Q_FREEZE); 448 } 449 450 if ((error = xbd_queue_request(sc, cm)) != 0) { 451 printf("xbd_queue_request returned %d\n", error); 452 break; 453 } 454 queued++; 455 } 456 457 if (queued != 0) 458 xbd_flush_requests(sc); 459 } 460 461 static void 462 xbd_bio_complete(struct xbd_softc *sc, struct xbd_command *cm) 463 { 464 struct bio *bp; 465 466 bp = cm->cm_bp; 467 468 if (__predict_false(cm->cm_status != BLKIF_RSP_OKAY)) { 469 disk_err(bp, "disk error" , -1, 0); 470 printf(" status: %x\n", cm->cm_status); 471 bp->bio_flags |= BIO_ERROR; 472 } 473 474 if (bp->bio_flags & BIO_ERROR) 475 bp->bio_error = EIO; 476 else 477 bp->bio_resid = 0; 478 479 xbd_free_command(cm); 480 biodone(bp); 481 } 482 483 static void 484 xbd_int(void *xsc) 485 { 486 struct xbd_softc *sc = xsc; 487 struct xbd_command *cm; 488 blkif_response_t *bret; 489 RING_IDX i, rp; 490 int op; 491 492 mtx_lock(&sc->xbd_io_lock); 493 494 if (__predict_false(sc->xbd_state == XBD_STATE_DISCONNECTED)) { 495 mtx_unlock(&sc->xbd_io_lock); 496 return; 497 } 498 499 again: 500 rp = sc->xbd_ring.sring->rsp_prod; 501 rmb(); /* Ensure we see queued responses up to 'rp'. */ 502 503 for (i = sc->xbd_ring.rsp_cons; i != rp;) { 504 bret = RING_GET_RESPONSE(&sc->xbd_ring, i); 505 cm = &sc->xbd_shadow[bret->id]; 506 507 xbd_remove_cm(cm, XBD_Q_BUSY); 508 gnttab_end_foreign_access_references(cm->cm_nseg, 509 cm->cm_sg_refs); 510 i++; 511 512 if (cm->cm_operation == BLKIF_OP_READ) 513 op = BUS_DMASYNC_POSTREAD; 514 else if (cm->cm_operation == BLKIF_OP_WRITE || 515 cm->cm_operation == BLKIF_OP_WRITE_BARRIER) 516 op = BUS_DMASYNC_POSTWRITE; 517 else 518 op = 0; 519 bus_dmamap_sync(sc->xbd_io_dmat, cm->cm_map, op); 520 bus_dmamap_unload(sc->xbd_io_dmat, cm->cm_map); 521 522 /* 523 * Release any hold this command has on future command 524 * dispatch. 525 */ 526 xbd_cm_thaw(sc, cm); 527 528 /* 529 * Directly call the i/o complete routine to save an 530 * an indirection in the common case. 531 */ 532 cm->cm_status = bret->status; 533 if (cm->cm_bp) 534 xbd_bio_complete(sc, cm); 535 else if (cm->cm_complete != NULL) 536 cm->cm_complete(cm); 537 else 538 xbd_free_command(cm); 539 } 540 541 sc->xbd_ring.rsp_cons = i; 542 543 if (i != sc->xbd_ring.req_prod_pvt) { 544 int more_to_do; 545 RING_FINAL_CHECK_FOR_RESPONSES(&sc->xbd_ring, more_to_do); 546 if (more_to_do) 547 goto again; 548 } else { 549 sc->xbd_ring.sring->rsp_event = i + 1; 550 } 551 552 if (xbd_queue_length(sc, XBD_Q_BUSY) == 0) 553 xbd_thaw(sc, XBDF_WAIT_IDLE); 554 555 xbd_startio(sc); 556 557 if (__predict_false(sc->xbd_state == XBD_STATE_SUSPENDED)) 558 wakeup(&sc->xbd_cm_q[XBD_Q_BUSY]); 559 560 mtx_unlock(&sc->xbd_io_lock); 561 } 562 563 /*------------------------------- Dump Support -------------------------------*/ 564 /** 565 * Quiesce the disk writes for a dump file before allowing the next buffer. 566 */ 567 static void 568 xbd_quiesce(struct xbd_softc *sc) 569 { 570 int mtd; 571 572 // While there are outstanding requests 573 while (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 574 RING_FINAL_CHECK_FOR_RESPONSES(&sc->xbd_ring, mtd); 575 if (mtd) { 576 /* Received request completions, update queue. */ 577 xbd_int(sc); 578 } 579 if (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 580 /* 581 * Still pending requests, wait for the disk i/o 582 * to complete. 583 */ 584 HYPERVISOR_yield(); 585 } 586 } 587 } 588 589 /* Kernel dump function for a paravirtualized disk device */ 590 static void 591 xbd_dump_complete(struct xbd_command *cm) 592 { 593 594 xbd_enqueue_cm(cm, XBD_Q_COMPLETE); 595 } 596 597 static int 598 xbd_dump(void *arg, void *virtual, off_t offset, size_t length) 599 { 600 struct disk *dp = arg; 601 struct xbd_softc *sc = dp->d_drv1; 602 struct xbd_command *cm; 603 size_t chunk; 604 int rc = 0; 605 606 if (length == 0) 607 return (0); 608 609 xbd_quiesce(sc); /* All quiet on the western front. */ 610 611 /* 612 * If this lock is held, then this module is failing, and a 613 * successful kernel dump is highly unlikely anyway. 614 */ 615 mtx_lock(&sc->xbd_io_lock); 616 617 /* Split the 64KB block as needed */ 618 while (length > 0) { 619 cm = xbd_dequeue_cm(sc, XBD_Q_FREE); 620 if (cm == NULL) { 621 mtx_unlock(&sc->xbd_io_lock); 622 device_printf(sc->xbd_dev, "dump: no more commands?\n"); 623 return (EBUSY); 624 } 625 626 if (gnttab_alloc_grant_references(sc->xbd_max_request_segments, 627 &cm->cm_gref_head) != 0) { 628 xbd_free_command(cm); 629 mtx_unlock(&sc->xbd_io_lock); 630 device_printf(sc->xbd_dev, "no more grant allocs?\n"); 631 return (EBUSY); 632 } 633 634 chunk = length > sc->xbd_max_request_size ? 635 sc->xbd_max_request_size : length; 636 cm->cm_data = virtual; 637 cm->cm_datalen = chunk; 638 cm->cm_operation = BLKIF_OP_WRITE; 639 cm->cm_sector_number = offset >> XBD_SECTOR_SHFT; 640 cm->cm_complete = xbd_dump_complete; 641 642 xbd_enqueue_cm(cm, XBD_Q_READY); 643 644 length -= chunk; 645 offset += chunk; 646 virtual = (char *) virtual + chunk; 647 } 648 649 /* Tell DOM0 to do the I/O */ 650 xbd_startio(sc); 651 mtx_unlock(&sc->xbd_io_lock); 652 653 /* Poll for the completion. */ 654 xbd_quiesce(sc); /* All quite on the eastern front */ 655 656 /* If there were any errors, bail out... */ 657 while ((cm = xbd_dequeue_cm(sc, XBD_Q_COMPLETE)) != NULL) { 658 if (cm->cm_status != BLKIF_RSP_OKAY) { 659 device_printf(sc->xbd_dev, 660 "Dump I/O failed at sector %jd\n", 661 cm->cm_sector_number); 662 rc = EIO; 663 } 664 xbd_free_command(cm); 665 } 666 667 return (rc); 668 } 669 670 /*----------------------------- Disk Entrypoints -----------------------------*/ 671 static int 672 xbd_open(struct disk *dp) 673 { 674 struct xbd_softc *sc = dp->d_drv1; 675 676 if (sc == NULL) { 677 printf("xbd%d: not found", dp->d_unit); 678 return (ENXIO); 679 } 680 681 sc->xbd_flags |= XBDF_OPEN; 682 sc->xbd_users++; 683 return (0); 684 } 685 686 static int 687 xbd_close(struct disk *dp) 688 { 689 struct xbd_softc *sc = dp->d_drv1; 690 691 if (sc == NULL) 692 return (ENXIO); 693 sc->xbd_flags &= ~XBDF_OPEN; 694 if (--(sc->xbd_users) == 0) { 695 /* 696 * Check whether we have been instructed to close. We will 697 * have ignored this request initially, as the device was 698 * still mounted. 699 */ 700 if (xenbus_get_otherend_state(sc->xbd_dev) == 701 XenbusStateClosing) 702 xbd_closing(sc->xbd_dev); 703 } 704 return (0); 705 } 706 707 static int 708 xbd_ioctl(struct disk *dp, u_long cmd, void *addr, int flag, struct thread *td) 709 { 710 struct xbd_softc *sc = dp->d_drv1; 711 712 if (sc == NULL) 713 return (ENXIO); 714 715 return (ENOTTY); 716 } 717 718 /* 719 * Read/write routine for a buffer. Finds the proper unit, place it on 720 * the sortq and kick the controller. 721 */ 722 static void 723 xbd_strategy(struct bio *bp) 724 { 725 struct xbd_softc *sc = bp->bio_disk->d_drv1; 726 727 /* bogus disk? */ 728 if (sc == NULL) { 729 bp->bio_error = EINVAL; 730 bp->bio_flags |= BIO_ERROR; 731 bp->bio_resid = bp->bio_bcount; 732 biodone(bp); 733 return; 734 } 735 736 /* 737 * Place it in the queue of disk activities for this disk 738 */ 739 mtx_lock(&sc->xbd_io_lock); 740 741 xbd_enqueue_bio(sc, bp); 742 xbd_startio(sc); 743 744 mtx_unlock(&sc->xbd_io_lock); 745 return; 746 } 747 748 /*------------------------------ Ring Management -----------------------------*/ 749 static int 750 xbd_alloc_ring(struct xbd_softc *sc) 751 { 752 blkif_sring_t *sring; 753 uintptr_t sring_page_addr; 754 int error; 755 int i; 756 757 sring = malloc(sc->xbd_ring_pages * PAGE_SIZE, M_XENBLOCKFRONT, 758 M_NOWAIT|M_ZERO); 759 if (sring == NULL) { 760 xenbus_dev_fatal(sc->xbd_dev, ENOMEM, "allocating shared ring"); 761 return (ENOMEM); 762 } 763 SHARED_RING_INIT(sring); 764 FRONT_RING_INIT(&sc->xbd_ring, sring, sc->xbd_ring_pages * PAGE_SIZE); 765 766 for (i = 0, sring_page_addr = (uintptr_t)sring; 767 i < sc->xbd_ring_pages; 768 i++, sring_page_addr += PAGE_SIZE) { 769 error = xenbus_grant_ring(sc->xbd_dev, 770 (vtophys(sring_page_addr) >> PAGE_SHIFT), 771 &sc->xbd_ring_ref[i]); 772 if (error) { 773 xenbus_dev_fatal(sc->xbd_dev, error, 774 "granting ring_ref(%d)", i); 775 return (error); 776 } 777 } 778 if (sc->xbd_ring_pages == 1) { 779 error = xs_printf(XST_NIL, xenbus_get_node(sc->xbd_dev), 780 "ring-ref", "%u", sc->xbd_ring_ref[0]); 781 if (error) { 782 xenbus_dev_fatal(sc->xbd_dev, error, 783 "writing %s/ring-ref", 784 xenbus_get_node(sc->xbd_dev)); 785 return (error); 786 } 787 } else { 788 for (i = 0; i < sc->xbd_ring_pages; i++) { 789 char ring_ref_name[]= "ring_refXX"; 790 791 snprintf(ring_ref_name, sizeof(ring_ref_name), 792 "ring-ref%u", i); 793 error = xs_printf(XST_NIL, xenbus_get_node(sc->xbd_dev), 794 ring_ref_name, "%u", sc->xbd_ring_ref[i]); 795 if (error) { 796 xenbus_dev_fatal(sc->xbd_dev, error, 797 "writing %s/%s", 798 xenbus_get_node(sc->xbd_dev), 799 ring_ref_name); 800 return (error); 801 } 802 } 803 } 804 805 error = xen_intr_alloc_and_bind_local_port(sc->xbd_dev, 806 xenbus_get_otherend_id(sc->xbd_dev), NULL, xbd_int, sc, 807 INTR_TYPE_BIO | INTR_MPSAFE, &sc->xen_intr_handle); 808 if (error) { 809 xenbus_dev_fatal(sc->xbd_dev, error, 810 "xen_intr_alloc_and_bind_local_port failed"); 811 return (error); 812 } 813 814 return (0); 815 } 816 817 static void 818 xbd_free_ring(struct xbd_softc *sc) 819 { 820 int i; 821 822 if (sc->xbd_ring.sring == NULL) 823 return; 824 825 for (i = 0; i < sc->xbd_ring_pages; i++) { 826 if (sc->xbd_ring_ref[i] != GRANT_REF_INVALID) { 827 gnttab_end_foreign_access_ref(sc->xbd_ring_ref[i]); 828 sc->xbd_ring_ref[i] = GRANT_REF_INVALID; 829 } 830 } 831 free(sc->xbd_ring.sring, M_XENBLOCKFRONT); 832 sc->xbd_ring.sring = NULL; 833 } 834 835 /*-------------------------- Initialization/Teardown -------------------------*/ 836 static int 837 xbd_feature_string(struct xbd_softc *sc, char *features, size_t len) 838 { 839 struct sbuf sb; 840 int feature_cnt; 841 842 sbuf_new(&sb, features, len, SBUF_FIXEDLEN); 843 844 feature_cnt = 0; 845 if ((sc->xbd_flags & XBDF_FLUSH) != 0) { 846 sbuf_printf(&sb, "flush"); 847 feature_cnt++; 848 } 849 850 if ((sc->xbd_flags & XBDF_BARRIER) != 0) { 851 if (feature_cnt != 0) 852 sbuf_printf(&sb, ", "); 853 sbuf_printf(&sb, "write_barrier"); 854 feature_cnt++; 855 } 856 857 if ((sc->xbd_flags & XBDF_DISCARD) != 0) { 858 if (feature_cnt != 0) 859 sbuf_printf(&sb, ", "); 860 sbuf_printf(&sb, "discard"); 861 feature_cnt++; 862 } 863 864 if ((sc->xbd_flags & XBDF_PERSISTENT) != 0) { 865 if (feature_cnt != 0) 866 sbuf_printf(&sb, ", "); 867 sbuf_printf(&sb, "persistent_grants"); 868 feature_cnt++; 869 } 870 871 (void) sbuf_finish(&sb); 872 return (sbuf_len(&sb)); 873 } 874 875 static int 876 xbd_sysctl_features(SYSCTL_HANDLER_ARGS) 877 { 878 char features[80]; 879 struct xbd_softc *sc = arg1; 880 int error; 881 int len; 882 883 error = sysctl_wire_old_buffer(req, 0); 884 if (error != 0) 885 return (error); 886 887 len = xbd_feature_string(sc, features, sizeof(features)); 888 889 /* len is -1 on error, which will make the SYSCTL_OUT a no-op. */ 890 return (SYSCTL_OUT(req, features, len + 1/*NUL*/)); 891 } 892 893 static void 894 xbd_setup_sysctl(struct xbd_softc *xbd) 895 { 896 struct sysctl_ctx_list *sysctl_ctx = NULL; 897 struct sysctl_oid *sysctl_tree = NULL; 898 struct sysctl_oid_list *children; 899 900 sysctl_ctx = device_get_sysctl_ctx(xbd->xbd_dev); 901 if (sysctl_ctx == NULL) 902 return; 903 904 sysctl_tree = device_get_sysctl_tree(xbd->xbd_dev); 905 if (sysctl_tree == NULL) 906 return; 907 908 children = SYSCTL_CHILDREN(sysctl_tree); 909 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 910 "max_requests", CTLFLAG_RD, &xbd->xbd_max_requests, -1, 911 "maximum outstanding requests (negotiated)"); 912 913 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 914 "max_request_segments", CTLFLAG_RD, 915 &xbd->xbd_max_request_segments, 0, 916 "maximum number of pages per requests (negotiated)"); 917 918 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 919 "max_request_size", CTLFLAG_RD, &xbd->xbd_max_request_size, 0, 920 "maximum size in bytes of a request (negotiated)"); 921 922 SYSCTL_ADD_UINT(sysctl_ctx, children, OID_AUTO, 923 "ring_pages", CTLFLAG_RD, &xbd->xbd_ring_pages, 0, 924 "communication channel pages (negotiated)"); 925 926 SYSCTL_ADD_PROC(sysctl_ctx, children, OID_AUTO, 927 "features", CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, xbd, 928 0, xbd_sysctl_features, "A", "protocol features (negotiated)"); 929 } 930 931 /* 932 * Translate Linux major/minor to an appropriate name and unit 933 * number. For HVM guests, this allows us to use the same drive names 934 * with blkfront as the emulated drives, easing transition slightly. 935 */ 936 static void 937 xbd_vdevice_to_unit(uint32_t vdevice, int *unit, const char **name) 938 { 939 static struct vdev_info { 940 int major; 941 int shift; 942 int base; 943 const char *name; 944 } info[] = { 945 {3, 6, 0, "ada"}, /* ide0 */ 946 {22, 6, 2, "ada"}, /* ide1 */ 947 {33, 6, 4, "ada"}, /* ide2 */ 948 {34, 6, 6, "ada"}, /* ide3 */ 949 {56, 6, 8, "ada"}, /* ide4 */ 950 {57, 6, 10, "ada"}, /* ide5 */ 951 {88, 6, 12, "ada"}, /* ide6 */ 952 {89, 6, 14, "ada"}, /* ide7 */ 953 {90, 6, 16, "ada"}, /* ide8 */ 954 {91, 6, 18, "ada"}, /* ide9 */ 955 956 {8, 4, 0, "da"}, /* scsi disk0 */ 957 {65, 4, 16, "da"}, /* scsi disk1 */ 958 {66, 4, 32, "da"}, /* scsi disk2 */ 959 {67, 4, 48, "da"}, /* scsi disk3 */ 960 {68, 4, 64, "da"}, /* scsi disk4 */ 961 {69, 4, 80, "da"}, /* scsi disk5 */ 962 {70, 4, 96, "da"}, /* scsi disk6 */ 963 {71, 4, 112, "da"}, /* scsi disk7 */ 964 {128, 4, 128, "da"}, /* scsi disk8 */ 965 {129, 4, 144, "da"}, /* scsi disk9 */ 966 {130, 4, 160, "da"}, /* scsi disk10 */ 967 {131, 4, 176, "da"}, /* scsi disk11 */ 968 {132, 4, 192, "da"}, /* scsi disk12 */ 969 {133, 4, 208, "da"}, /* scsi disk13 */ 970 {134, 4, 224, "da"}, /* scsi disk14 */ 971 {135, 4, 240, "da"}, /* scsi disk15 */ 972 973 {202, 4, 0, "xbd"}, /* xbd */ 974 975 {0, 0, 0, NULL}, 976 }; 977 int major = vdevice >> 8; 978 int minor = vdevice & 0xff; 979 int i; 980 981 if (vdevice & (1 << 28)) { 982 *unit = (vdevice & ((1 << 28) - 1)) >> 8; 983 *name = "xbd"; 984 return; 985 } 986 987 for (i = 0; info[i].major; i++) { 988 if (info[i].major == major) { 989 *unit = info[i].base + (minor >> info[i].shift); 990 *name = info[i].name; 991 return; 992 } 993 } 994 995 *unit = minor >> 4; 996 *name = "xbd"; 997 } 998 999 int 1000 xbd_instance_create(struct xbd_softc *sc, blkif_sector_t sectors, 1001 int vdevice, uint16_t vdisk_info, unsigned long sector_size, 1002 unsigned long phys_sector_size) 1003 { 1004 char features[80]; 1005 int unit, error = 0; 1006 const char *name; 1007 1008 xbd_vdevice_to_unit(vdevice, &unit, &name); 1009 1010 sc->xbd_unit = unit; 1011 1012 if (strcmp(name, "xbd") != 0) 1013 device_printf(sc->xbd_dev, "attaching as %s%d\n", name, unit); 1014 1015 if (xbd_feature_string(sc, features, sizeof(features)) > 0) { 1016 device_printf(sc->xbd_dev, "features: %s\n", 1017 features); 1018 } 1019 1020 sc->xbd_disk = disk_alloc(); 1021 sc->xbd_disk->d_unit = sc->xbd_unit; 1022 sc->xbd_disk->d_open = xbd_open; 1023 sc->xbd_disk->d_close = xbd_close; 1024 sc->xbd_disk->d_ioctl = xbd_ioctl; 1025 sc->xbd_disk->d_strategy = xbd_strategy; 1026 sc->xbd_disk->d_dump = xbd_dump; 1027 sc->xbd_disk->d_name = name; 1028 sc->xbd_disk->d_drv1 = sc; 1029 sc->xbd_disk->d_sectorsize = sector_size; 1030 sc->xbd_disk->d_stripesize = phys_sector_size; 1031 sc->xbd_disk->d_stripeoffset = 0; 1032 1033 /* 1034 * The 'sectors' xenbus node is always in units of 512b, regardless of 1035 * the 'sector-size' xenbus node value. 1036 */ 1037 sc->xbd_disk->d_mediasize = sectors << XBD_SECTOR_SHFT; 1038 if ((sc->xbd_disk->d_mediasize % sc->xbd_disk->d_sectorsize) != 0) { 1039 error = EINVAL; 1040 xenbus_dev_fatal(sc->xbd_dev, error, 1041 "Disk size (%ju) not a multiple of sector size (%ju)", 1042 (uintmax_t)sc->xbd_disk->d_mediasize, 1043 (uintmax_t)sc->xbd_disk->d_sectorsize); 1044 return (error); 1045 } 1046 sc->xbd_disk->d_maxsize = sc->xbd_max_request_size; 1047 sc->xbd_disk->d_flags = DISKFLAG_UNMAPPED_BIO; 1048 if ((sc->xbd_flags & (XBDF_FLUSH|XBDF_BARRIER)) != 0) { 1049 sc->xbd_disk->d_flags |= DISKFLAG_CANFLUSHCACHE; 1050 device_printf(sc->xbd_dev, 1051 "synchronize cache commands enabled.\n"); 1052 } 1053 disk_create(sc->xbd_disk, DISK_VERSION); 1054 1055 return error; 1056 } 1057 1058 static void 1059 xbd_free(struct xbd_softc *sc) 1060 { 1061 int i; 1062 1063 /* Prevent new requests being issued until we fix things up. */ 1064 mtx_lock(&sc->xbd_io_lock); 1065 sc->xbd_state = XBD_STATE_DISCONNECTED; 1066 mtx_unlock(&sc->xbd_io_lock); 1067 1068 /* Free resources associated with old device channel. */ 1069 xbd_free_ring(sc); 1070 if (sc->xbd_shadow) { 1071 for (i = 0; i < sc->xbd_max_requests; i++) { 1072 struct xbd_command *cm; 1073 1074 cm = &sc->xbd_shadow[i]; 1075 if (cm->cm_sg_refs != NULL) { 1076 free(cm->cm_sg_refs, M_XENBLOCKFRONT); 1077 cm->cm_sg_refs = NULL; 1078 } 1079 1080 if (cm->cm_indirectionpages != NULL) { 1081 gnttab_end_foreign_access_references( 1082 sc->xbd_max_request_indirectpages, 1083 &cm->cm_indirectionrefs[0]); 1084 free(cm->cm_indirectionpages, M_XENBLOCKFRONT); 1085 cm->cm_indirectionpages = NULL; 1086 } 1087 1088 bus_dmamap_destroy(sc->xbd_io_dmat, cm->cm_map); 1089 } 1090 free(sc->xbd_shadow, M_XENBLOCKFRONT); 1091 sc->xbd_shadow = NULL; 1092 1093 bus_dma_tag_destroy(sc->xbd_io_dmat); 1094 1095 xbd_initq_cm(sc, XBD_Q_FREE); 1096 xbd_initq_cm(sc, XBD_Q_READY); 1097 xbd_initq_cm(sc, XBD_Q_COMPLETE); 1098 } 1099 1100 xen_intr_unbind(&sc->xen_intr_handle); 1101 1102 } 1103 1104 /*--------------------------- State Change Handlers --------------------------*/ 1105 static void 1106 xbd_initialize(struct xbd_softc *sc) 1107 { 1108 const char *otherend_path; 1109 const char *node_path; 1110 uint32_t max_ring_page_order; 1111 int error; 1112 1113 if (xenbus_get_state(sc->xbd_dev) != XenbusStateInitialising) { 1114 /* Initialization has already been performed. */ 1115 return; 1116 } 1117 1118 /* 1119 * Protocol defaults valid even if negotiation for a 1120 * setting fails. 1121 */ 1122 max_ring_page_order = 0; 1123 sc->xbd_ring_pages = 1; 1124 1125 /* 1126 * Protocol negotiation. 1127 * 1128 * \note xs_gather() returns on the first encountered error, so 1129 * we must use independent calls in order to guarantee 1130 * we don't miss information in a sparsly populated back-end 1131 * tree. 1132 * 1133 * \note xs_scanf() does not update variables for unmatched 1134 * fields. 1135 */ 1136 otherend_path = xenbus_get_otherend_path(sc->xbd_dev); 1137 node_path = xenbus_get_node(sc->xbd_dev); 1138 1139 /* Support both backend schemes for relaying ring page limits. */ 1140 (void)xs_scanf(XST_NIL, otherend_path, 1141 "max-ring-page-order", NULL, "%" PRIu32, 1142 &max_ring_page_order); 1143 sc->xbd_ring_pages = 1 << max_ring_page_order; 1144 (void)xs_scanf(XST_NIL, otherend_path, 1145 "max-ring-pages", NULL, "%" PRIu32, 1146 &sc->xbd_ring_pages); 1147 if (sc->xbd_ring_pages < 1) 1148 sc->xbd_ring_pages = 1; 1149 1150 if (sc->xbd_ring_pages > XBD_MAX_RING_PAGES) { 1151 device_printf(sc->xbd_dev, 1152 "Back-end specified ring-pages of %u " 1153 "limited to front-end limit of %u.\n", 1154 sc->xbd_ring_pages, XBD_MAX_RING_PAGES); 1155 sc->xbd_ring_pages = XBD_MAX_RING_PAGES; 1156 } 1157 1158 if (powerof2(sc->xbd_ring_pages) == 0) { 1159 uint32_t new_page_limit; 1160 1161 new_page_limit = 0x01 << (fls(sc->xbd_ring_pages) - 1); 1162 device_printf(sc->xbd_dev, 1163 "Back-end specified ring-pages of %u " 1164 "is not a power of 2. Limited to %u.\n", 1165 sc->xbd_ring_pages, new_page_limit); 1166 sc->xbd_ring_pages = new_page_limit; 1167 } 1168 1169 sc->xbd_max_requests = 1170 BLKIF_MAX_RING_REQUESTS(sc->xbd_ring_pages * PAGE_SIZE); 1171 if (sc->xbd_max_requests > XBD_MAX_REQUESTS) { 1172 device_printf(sc->xbd_dev, 1173 "Back-end specified max_requests of %u " 1174 "limited to front-end limit of %zu.\n", 1175 sc->xbd_max_requests, XBD_MAX_REQUESTS); 1176 sc->xbd_max_requests = XBD_MAX_REQUESTS; 1177 } 1178 1179 if (xbd_alloc_ring(sc) != 0) 1180 return; 1181 1182 /* Support both backend schemes for relaying ring page limits. */ 1183 if (sc->xbd_ring_pages > 1) { 1184 error = xs_printf(XST_NIL, node_path, 1185 "num-ring-pages","%u", 1186 sc->xbd_ring_pages); 1187 if (error) { 1188 xenbus_dev_fatal(sc->xbd_dev, error, 1189 "writing %s/num-ring-pages", 1190 node_path); 1191 return; 1192 } 1193 1194 error = xs_printf(XST_NIL, node_path, 1195 "ring-page-order", "%u", 1196 fls(sc->xbd_ring_pages) - 1); 1197 if (error) { 1198 xenbus_dev_fatal(sc->xbd_dev, error, 1199 "writing %s/ring-page-order", 1200 node_path); 1201 return; 1202 } 1203 } 1204 1205 error = xs_printf(XST_NIL, node_path, "event-channel", 1206 "%u", xen_intr_port(sc->xen_intr_handle)); 1207 if (error) { 1208 xenbus_dev_fatal(sc->xbd_dev, error, 1209 "writing %s/event-channel", 1210 node_path); 1211 return; 1212 } 1213 1214 error = xs_printf(XST_NIL, node_path, "protocol", 1215 "%s", XEN_IO_PROTO_ABI_NATIVE); 1216 if (error) { 1217 xenbus_dev_fatal(sc->xbd_dev, error, 1218 "writing %s/protocol", 1219 node_path); 1220 return; 1221 } 1222 1223 xenbus_set_state(sc->xbd_dev, XenbusStateInitialised); 1224 } 1225 1226 /* 1227 * Invoked when the backend is finally 'ready' (and has published 1228 * the details about the physical device - #sectors, size, etc). 1229 */ 1230 static void 1231 xbd_connect(struct xbd_softc *sc) 1232 { 1233 device_t dev = sc->xbd_dev; 1234 blkif_sector_t sectors; 1235 unsigned long sector_size, phys_sector_size; 1236 unsigned int binfo; 1237 int err, feature_barrier, feature_flush; 1238 int i, j; 1239 1240 DPRINTK("blkfront.c:connect:%s.\n", xenbus_get_otherend_path(dev)); 1241 1242 if (sc->xbd_state == XBD_STATE_SUSPENDED) { 1243 return; 1244 } 1245 1246 if (sc->xbd_state == XBD_STATE_CONNECTED) { 1247 struct disk *disk; 1248 1249 disk = sc->xbd_disk; 1250 if (disk == NULL) { 1251 return; 1252 } 1253 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1254 "sectors", "%"PRIu64, §ors, NULL); 1255 if (err != 0) { 1256 xenbus_dev_error(dev, err, 1257 "reading sectors at %s", 1258 xenbus_get_otherend_path(dev)); 1259 return; 1260 } 1261 disk->d_mediasize = disk->d_sectorsize * sectors; 1262 err = disk_resize(disk, M_NOWAIT); 1263 if (err) { 1264 xenbus_dev_error(dev, err, 1265 "unable to resize disk %s%u", 1266 disk->d_name, disk->d_unit); 1267 return; 1268 } 1269 device_printf(sc->xbd_dev, 1270 "changed capacity to %jd\n", 1271 (intmax_t)disk->d_mediasize); 1272 return; 1273 } 1274 1275 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1276 "sectors", "%"PRIu64, §ors, 1277 "info", "%u", &binfo, 1278 "sector-size", "%lu", §or_size, 1279 NULL); 1280 if (err) { 1281 xenbus_dev_fatal(dev, err, 1282 "reading backend fields at %s", 1283 xenbus_get_otherend_path(dev)); 1284 return; 1285 } 1286 if ((sectors == 0) || (sector_size == 0)) { 1287 xenbus_dev_fatal(dev, 0, 1288 "invalid parameters from %s:" 1289 " sectors = %"PRIu64", sector_size = %lu", 1290 xenbus_get_otherend_path(dev), 1291 sectors, sector_size); 1292 return; 1293 } 1294 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1295 "physical-sector-size", "%lu", &phys_sector_size, 1296 NULL); 1297 if (err || phys_sector_size <= sector_size) 1298 phys_sector_size = 0; 1299 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1300 "feature-barrier", "%d", &feature_barrier, 1301 NULL); 1302 if (err == 0 && feature_barrier != 0) 1303 sc->xbd_flags |= XBDF_BARRIER; 1304 1305 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1306 "feature-flush-cache", "%d", &feature_flush, 1307 NULL); 1308 if (err == 0 && feature_flush != 0) 1309 sc->xbd_flags |= XBDF_FLUSH; 1310 1311 err = xs_gather(XST_NIL, xenbus_get_otherend_path(dev), 1312 "feature-max-indirect-segments", "%" PRIu32, 1313 &sc->xbd_max_request_segments, NULL); 1314 if ((err != 0) || (xbd_enable_indirect == 0)) 1315 sc->xbd_max_request_segments = 0; 1316 if (sc->xbd_max_request_segments > XBD_MAX_INDIRECT_SEGMENTS) 1317 sc->xbd_max_request_segments = XBD_MAX_INDIRECT_SEGMENTS; 1318 if (sc->xbd_max_request_segments > XBD_SIZE_TO_SEGS(maxphys)) 1319 sc->xbd_max_request_segments = XBD_SIZE_TO_SEGS(maxphys); 1320 sc->xbd_max_request_indirectpages = 1321 XBD_INDIRECT_SEGS_TO_PAGES(sc->xbd_max_request_segments); 1322 if (sc->xbd_max_request_segments < BLKIF_MAX_SEGMENTS_PER_REQUEST) 1323 sc->xbd_max_request_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; 1324 sc->xbd_max_request_size = 1325 XBD_SEGS_TO_SIZE(sc->xbd_max_request_segments); 1326 1327 /* Allocate datastructures based on negotiated values. */ 1328 err = bus_dma_tag_create( 1329 bus_get_dma_tag(sc->xbd_dev), /* parent */ 1330 sector_size, PAGE_SIZE, /* algnmnt, boundary */ 1331 BUS_SPACE_MAXADDR, /* lowaddr */ 1332 BUS_SPACE_MAXADDR, /* highaddr */ 1333 NULL, NULL, /* filter, filterarg */ 1334 sc->xbd_max_request_size, 1335 sc->xbd_max_request_segments, 1336 PAGE_SIZE, /* maxsegsize */ 1337 BUS_DMA_ALLOCNOW, /* flags */ 1338 busdma_lock_mutex, /* lockfunc */ 1339 &sc->xbd_io_lock, /* lockarg */ 1340 &sc->xbd_io_dmat); 1341 if (err != 0) { 1342 xenbus_dev_fatal(sc->xbd_dev, err, 1343 "Cannot allocate parent DMA tag\n"); 1344 return; 1345 } 1346 1347 /* Per-transaction data allocation. */ 1348 sc->xbd_shadow = malloc(sizeof(*sc->xbd_shadow) * sc->xbd_max_requests, 1349 M_XENBLOCKFRONT, M_NOWAIT|M_ZERO); 1350 if (sc->xbd_shadow == NULL) { 1351 bus_dma_tag_destroy(sc->xbd_io_dmat); 1352 xenbus_dev_fatal(sc->xbd_dev, ENOMEM, 1353 "Cannot allocate request structures\n"); 1354 return; 1355 } 1356 1357 for (i = 0; i < sc->xbd_max_requests; i++) { 1358 struct xbd_command *cm; 1359 void * indirectpages; 1360 1361 cm = &sc->xbd_shadow[i]; 1362 cm->cm_sg_refs = malloc( 1363 sizeof(grant_ref_t) * sc->xbd_max_request_segments, 1364 M_XENBLOCKFRONT, M_NOWAIT); 1365 if (cm->cm_sg_refs == NULL) 1366 break; 1367 cm->cm_id = i; 1368 cm->cm_flags = XBDCF_INITIALIZER; 1369 cm->cm_sc = sc; 1370 if (bus_dmamap_create(sc->xbd_io_dmat, 0, &cm->cm_map) != 0) 1371 break; 1372 if (sc->xbd_max_request_indirectpages > 0) { 1373 indirectpages = contigmalloc( 1374 PAGE_SIZE * sc->xbd_max_request_indirectpages, 1375 M_XENBLOCKFRONT, M_ZERO | M_NOWAIT, 0, ~0, 1376 PAGE_SIZE, 0); 1377 if (indirectpages == NULL) 1378 sc->xbd_max_request_indirectpages = 0; 1379 } else { 1380 indirectpages = NULL; 1381 } 1382 for (j = 0; j < sc->xbd_max_request_indirectpages; j++) { 1383 if (gnttab_grant_foreign_access( 1384 xenbus_get_otherend_id(sc->xbd_dev), 1385 (vtophys(indirectpages) >> PAGE_SHIFT) + j, 1386 1 /* grant read-only access */, 1387 &cm->cm_indirectionrefs[j])) 1388 break; 1389 } 1390 if (j < sc->xbd_max_request_indirectpages) { 1391 free(indirectpages, M_XENBLOCKFRONT); 1392 break; 1393 } 1394 cm->cm_indirectionpages = indirectpages; 1395 xbd_free_command(cm); 1396 } 1397 1398 if (sc->xbd_disk == NULL) { 1399 device_printf(dev, "%juMB <%s> at %s", 1400 (uintmax_t)((sectors << XBD_SECTOR_SHFT) / 1048576), 1401 device_get_desc(dev), 1402 xenbus_get_node(dev)); 1403 bus_print_child_footer(device_get_parent(dev), dev); 1404 1405 err = xbd_instance_create(sc, sectors, sc->xbd_vdevice, binfo, 1406 sector_size, phys_sector_size); 1407 if (err != 0) { 1408 xenbus_dev_fatal(dev, err, "Unable to create instance"); 1409 return; 1410 } 1411 } 1412 1413 (void)xenbus_set_state(dev, XenbusStateConnected); 1414 1415 /* Kick pending requests. */ 1416 mtx_lock(&sc->xbd_io_lock); 1417 sc->xbd_state = XBD_STATE_CONNECTED; 1418 xbd_startio(sc); 1419 sc->xbd_flags |= XBDF_READY; 1420 mtx_unlock(&sc->xbd_io_lock); 1421 } 1422 1423 /** 1424 * Handle the change of state of the backend to Closing. We must delete our 1425 * device-layer structures now, to ensure that writes are flushed through to 1426 * the backend. Once this is done, we can switch to Closed in 1427 * acknowledgement. 1428 */ 1429 static void 1430 xbd_closing(device_t dev) 1431 { 1432 struct xbd_softc *sc = device_get_softc(dev); 1433 1434 xenbus_set_state(dev, XenbusStateClosing); 1435 1436 DPRINTK("xbd_closing: %s removed\n", xenbus_get_node(dev)); 1437 1438 if (sc->xbd_disk != NULL) { 1439 disk_destroy(sc->xbd_disk); 1440 sc->xbd_disk = NULL; 1441 } 1442 1443 xenbus_set_state(dev, XenbusStateClosed); 1444 } 1445 1446 /*---------------------------- NewBus Entrypoints ----------------------------*/ 1447 static int 1448 xbd_probe(device_t dev) 1449 { 1450 if (strcmp(xenbus_get_type(dev), "vbd") != 0) 1451 return (ENXIO); 1452 1453 if (xen_pv_disks_disabled()) 1454 return (ENXIO); 1455 1456 if (xen_hvm_domain()) { 1457 int error; 1458 char *type; 1459 1460 /* 1461 * When running in an HVM domain, IDE disk emulation is 1462 * disabled early in boot so that native drivers will 1463 * not see emulated hardware. However, CDROM device 1464 * emulation cannot be disabled. 1465 * 1466 * Through use of FreeBSD's vm_guest and xen_hvm_domain() 1467 * APIs, we could modify the native CDROM driver to fail its 1468 * probe when running under Xen. Unfortunatlely, the PV 1469 * CDROM support in XenServer (up through at least version 1470 * 6.2) isn't functional, so we instead rely on the emulated 1471 * CDROM instance, and fail to attach the PV one here in 1472 * the blkfront driver. 1473 */ 1474 error = xs_read(XST_NIL, xenbus_get_node(dev), 1475 "device-type", NULL, (void **) &type); 1476 if (error) 1477 return (ENXIO); 1478 1479 if (strncmp(type, "cdrom", 5) == 0) { 1480 free(type, M_XENSTORE); 1481 return (ENXIO); 1482 } 1483 free(type, M_XENSTORE); 1484 } 1485 1486 device_set_desc(dev, "Virtual Block Device"); 1487 device_quiet(dev); 1488 return (0); 1489 } 1490 1491 /* 1492 * Setup supplies the backend dir, virtual device. We place an event 1493 * channel and shared frame entries. We watch backend to wait if it's 1494 * ok. 1495 */ 1496 static int 1497 xbd_attach(device_t dev) 1498 { 1499 struct xbd_softc *sc; 1500 const char *name; 1501 uint32_t vdevice; 1502 int error; 1503 int i; 1504 int unit; 1505 1506 /* FIXME: Use dynamic device id if this is not set. */ 1507 error = xs_scanf(XST_NIL, xenbus_get_node(dev), 1508 "virtual-device", NULL, "%" PRIu32, &vdevice); 1509 if (error) 1510 error = xs_scanf(XST_NIL, xenbus_get_node(dev), 1511 "virtual-device-ext", NULL, "%" PRIu32, &vdevice); 1512 if (error) { 1513 xenbus_dev_fatal(dev, error, "reading virtual-device"); 1514 device_printf(dev, "Couldn't determine virtual device.\n"); 1515 return (error); 1516 } 1517 1518 xbd_vdevice_to_unit(vdevice, &unit, &name); 1519 if (!strcmp(name, "xbd")) 1520 device_set_unit(dev, unit); 1521 1522 sc = device_get_softc(dev); 1523 mtx_init(&sc->xbd_io_lock, "blkfront i/o lock", NULL, MTX_DEF); 1524 xbd_initqs(sc); 1525 for (i = 0; i < XBD_MAX_RING_PAGES; i++) 1526 sc->xbd_ring_ref[i] = GRANT_REF_INVALID; 1527 1528 sc->xbd_dev = dev; 1529 sc->xbd_vdevice = vdevice; 1530 sc->xbd_state = XBD_STATE_DISCONNECTED; 1531 1532 xbd_setup_sysctl(sc); 1533 1534 /* Wait for backend device to publish its protocol capabilities. */ 1535 xenbus_set_state(dev, XenbusStateInitialising); 1536 1537 return (0); 1538 } 1539 1540 static int 1541 xbd_detach(device_t dev) 1542 { 1543 struct xbd_softc *sc = device_get_softc(dev); 1544 1545 DPRINTK("%s: %s removed\n", __func__, xenbus_get_node(dev)); 1546 1547 xbd_free(sc); 1548 mtx_destroy(&sc->xbd_io_lock); 1549 1550 return 0; 1551 } 1552 1553 static int 1554 xbd_suspend(device_t dev) 1555 { 1556 struct xbd_softc *sc = device_get_softc(dev); 1557 int retval; 1558 int saved_state; 1559 1560 /* Prevent new requests being issued until we fix things up. */ 1561 mtx_lock(&sc->xbd_io_lock); 1562 saved_state = sc->xbd_state; 1563 sc->xbd_state = XBD_STATE_SUSPENDED; 1564 1565 /* Wait for outstanding I/O to drain. */ 1566 retval = 0; 1567 while (xbd_queue_length(sc, XBD_Q_BUSY) != 0) { 1568 if (msleep(&sc->xbd_cm_q[XBD_Q_BUSY], &sc->xbd_io_lock, 1569 PRIBIO, "blkf_susp", 30 * hz) == EWOULDBLOCK) { 1570 retval = EBUSY; 1571 break; 1572 } 1573 } 1574 mtx_unlock(&sc->xbd_io_lock); 1575 1576 if (retval != 0) 1577 sc->xbd_state = saved_state; 1578 1579 return (retval); 1580 } 1581 1582 static int 1583 xbd_resume(device_t dev) 1584 { 1585 struct xbd_softc *sc = device_get_softc(dev); 1586 1587 if (xen_suspend_cancelled) { 1588 sc->xbd_state = XBD_STATE_CONNECTED; 1589 return (0); 1590 } 1591 1592 DPRINTK("xbd_resume: %s\n", xenbus_get_node(dev)); 1593 1594 xbd_free(sc); 1595 xbd_initialize(sc); 1596 return (0); 1597 } 1598 1599 /** 1600 * Callback received when the backend's state changes. 1601 */ 1602 static void 1603 xbd_backend_changed(device_t dev, XenbusState backend_state) 1604 { 1605 struct xbd_softc *sc = device_get_softc(dev); 1606 1607 DPRINTK("backend_state=%d\n", backend_state); 1608 1609 switch (backend_state) { 1610 case XenbusStateUnknown: 1611 case XenbusStateInitialising: 1612 case XenbusStateReconfigured: 1613 case XenbusStateReconfiguring: 1614 case XenbusStateClosed: 1615 break; 1616 1617 case XenbusStateInitWait: 1618 case XenbusStateInitialised: 1619 xbd_initialize(sc); 1620 break; 1621 1622 case XenbusStateConnected: 1623 xbd_initialize(sc); 1624 xbd_connect(sc); 1625 break; 1626 1627 case XenbusStateClosing: 1628 if (sc->xbd_users > 0) { 1629 device_printf(dev, "detaching with pending users\n"); 1630 KASSERT(sc->xbd_disk != NULL, 1631 ("NULL disk with pending users\n")); 1632 disk_gone(sc->xbd_disk); 1633 } else { 1634 xbd_closing(dev); 1635 } 1636 break; 1637 } 1638 } 1639 1640 /*---------------------------- NewBus Registration ---------------------------*/ 1641 static device_method_t xbd_methods[] = { 1642 /* Device interface */ 1643 DEVMETHOD(device_probe, xbd_probe), 1644 DEVMETHOD(device_attach, xbd_attach), 1645 DEVMETHOD(device_detach, xbd_detach), 1646 DEVMETHOD(device_shutdown, bus_generic_shutdown), 1647 DEVMETHOD(device_suspend, xbd_suspend), 1648 DEVMETHOD(device_resume, xbd_resume), 1649 1650 /* Xenbus interface */ 1651 DEVMETHOD(xenbus_otherend_changed, xbd_backend_changed), 1652 1653 DEVMETHOD_END 1654 }; 1655 1656 static driver_t xbd_driver = { 1657 "xbd", 1658 xbd_methods, 1659 sizeof(struct xbd_softc), 1660 }; 1661 1662 DRIVER_MODULE(xbd, xenbusb_front, xbd_driver, 0, 0); 1663