1 /* 2 * Xen SCSI frontend driver 3 * 4 * Copyright (c) 2008, FUJITSU Limited 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License version 2 8 * as published by the Free Software Foundation; or, when distributed 9 * separately from the Linux kernel or incorporated into other 10 * software packages, subject to the following license: 11 * 12 * Permission is hereby granted, free of charge, to any person obtaining a copy 13 * of this source file (the "Software"), to deal in the Software without 14 * restriction, including without limitation the rights to use, copy, modify, 15 * merge, publish, distribute, sublicense, and/or sell copies of the Software, 16 * and to permit persons to whom the Software is furnished to do so, subject to 17 * the following conditions: 18 * 19 * The above copyright notice and this permission notice shall be included in 20 * all copies or substantial portions of the Software. 21 * 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 DEALINGS 28 * IN THE SOFTWARE. 29 */ 30 31 #include <linux/module.h> 32 #include <linux/kernel.h> 33 #include <linux/device.h> 34 #include <linux/wait.h> 35 #include <linux/interrupt.h> 36 #include <linux/mutex.h> 37 #include <linux/spinlock.h> 38 #include <linux/sched.h> 39 #include <linux/blkdev.h> 40 #include <linux/pfn.h> 41 #include <linux/slab.h> 42 #include <linux/bitops.h> 43 44 #include <scsi/scsi_cmnd.h> 45 #include <scsi/scsi_device.h> 46 #include <scsi/scsi.h> 47 #include <scsi/scsi_host.h> 48 49 #include <xen/xen.h> 50 #include <xen/xenbus.h> 51 #include <xen/grant_table.h> 52 #include <xen/events.h> 53 #include <xen/page.h> 54 55 #include <xen/interface/grant_table.h> 56 #include <xen/interface/io/vscsiif.h> 57 #include <xen/interface/io/protocols.h> 58 59 #include <asm/xen/hypervisor.h> 60 61 62 #define GRANT_INVALID_REF 0 63 64 #define VSCSIFRONT_OP_ADD_LUN 1 65 #define VSCSIFRONT_OP_DEL_LUN 2 66 #define VSCSIFRONT_OP_READD_LUN 3 67 68 /* Tuning point. */ 69 #define VSCSIIF_DEFAULT_CMD_PER_LUN 10 70 #define VSCSIIF_MAX_TARGET 64 71 #define VSCSIIF_MAX_LUN 255 72 73 #define VSCSIIF_RING_SIZE __CONST_RING_SIZE(vscsiif, PAGE_SIZE) 74 #define VSCSIIF_MAX_REQS VSCSIIF_RING_SIZE 75 76 #define vscsiif_grants_sg(_sg) (PFN_UP((_sg) * \ 77 sizeof(struct scsiif_request_segment))) 78 79 struct vscsifrnt_shadow { 80 /* command between backend and frontend */ 81 unsigned char act; 82 uint8_t nr_segments; 83 uint16_t rqid; 84 uint16_t ref_rqid; 85 86 unsigned int nr_grants; /* number of grants in gref[] */ 87 struct scsiif_request_segment *sg; /* scatter/gather elements */ 88 struct scsiif_request_segment seg[VSCSIIF_SG_TABLESIZE]; 89 90 /* Do reset or abort function. */ 91 wait_queue_head_t wq_reset; /* reset work queue */ 92 int wait_reset; /* reset work queue condition */ 93 int32_t rslt_reset; /* reset response status: */ 94 /* SUCCESS or FAILED or: */ 95 #define RSLT_RESET_WAITING 0 96 #define RSLT_RESET_ERR -1 97 98 /* Requested struct scsi_cmnd is stored from kernel. */ 99 struct scsi_cmnd *sc; 100 int gref[vscsiif_grants_sg(SG_ALL) + SG_ALL]; 101 }; 102 103 struct vscsifrnt_info { 104 struct xenbus_device *dev; 105 106 struct Scsi_Host *host; 107 int host_active; 108 109 unsigned int evtchn; 110 unsigned int irq; 111 112 grant_ref_t ring_ref; 113 struct vscsiif_front_ring ring; 114 struct vscsiif_response ring_rsp; 115 116 spinlock_t shadow_lock; 117 DECLARE_BITMAP(shadow_free_bitmap, VSCSIIF_MAX_REQS); 118 struct vscsifrnt_shadow *shadow[VSCSIIF_MAX_REQS]; 119 120 /* Following items are protected by the host lock. */ 121 wait_queue_head_t wq_sync; 122 wait_queue_head_t wq_pause; 123 unsigned int wait_ring_available:1; 124 unsigned int waiting_pause:1; 125 unsigned int pause:1; 126 unsigned callers; 127 128 char dev_state_path[64]; 129 struct task_struct *curr; 130 }; 131 132 static DEFINE_MUTEX(scsifront_mutex); 133 134 static void scsifront_wake_up(struct vscsifrnt_info *info) 135 { 136 info->wait_ring_available = 0; 137 wake_up(&info->wq_sync); 138 } 139 140 static int scsifront_get_rqid(struct vscsifrnt_info *info) 141 { 142 unsigned long flags; 143 int free; 144 145 spin_lock_irqsave(&info->shadow_lock, flags); 146 147 free = find_first_bit(info->shadow_free_bitmap, VSCSIIF_MAX_REQS); 148 __clear_bit(free, info->shadow_free_bitmap); 149 150 spin_unlock_irqrestore(&info->shadow_lock, flags); 151 152 return free; 153 } 154 155 static int _scsifront_put_rqid(struct vscsifrnt_info *info, uint32_t id) 156 { 157 int empty = bitmap_empty(info->shadow_free_bitmap, VSCSIIF_MAX_REQS); 158 159 __set_bit(id, info->shadow_free_bitmap); 160 info->shadow[id] = NULL; 161 162 return empty || info->wait_ring_available; 163 } 164 165 static void scsifront_put_rqid(struct vscsifrnt_info *info, uint32_t id) 166 { 167 unsigned long flags; 168 int kick; 169 170 spin_lock_irqsave(&info->shadow_lock, flags); 171 kick = _scsifront_put_rqid(info, id); 172 spin_unlock_irqrestore(&info->shadow_lock, flags); 173 174 if (kick) 175 scsifront_wake_up(info); 176 } 177 178 static int scsifront_do_request(struct vscsifrnt_info *info, 179 struct vscsifrnt_shadow *shadow) 180 { 181 struct vscsiif_front_ring *ring = &(info->ring); 182 struct vscsiif_request *ring_req; 183 struct scsi_cmnd *sc = shadow->sc; 184 uint32_t id; 185 int i, notify; 186 187 if (RING_FULL(&info->ring)) 188 return -EBUSY; 189 190 id = scsifront_get_rqid(info); /* use id in response */ 191 if (id >= VSCSIIF_MAX_REQS) 192 return -EBUSY; 193 194 info->shadow[id] = shadow; 195 shadow->rqid = id; 196 197 ring_req = RING_GET_REQUEST(&(info->ring), ring->req_prod_pvt); 198 ring->req_prod_pvt++; 199 200 ring_req->rqid = id; 201 ring_req->act = shadow->act; 202 ring_req->ref_rqid = shadow->ref_rqid; 203 ring_req->nr_segments = shadow->nr_segments; 204 205 ring_req->id = sc->device->id; 206 ring_req->lun = sc->device->lun; 207 ring_req->channel = sc->device->channel; 208 ring_req->cmd_len = sc->cmd_len; 209 210 BUG_ON(sc->cmd_len > VSCSIIF_MAX_COMMAND_SIZE); 211 212 memcpy(ring_req->cmnd, sc->cmnd, sc->cmd_len); 213 214 ring_req->sc_data_direction = (uint8_t)sc->sc_data_direction; 215 ring_req->timeout_per_command = scsi_cmd_to_rq(sc)->timeout / HZ; 216 217 for (i = 0; i < (shadow->nr_segments & ~VSCSIIF_SG_GRANT); i++) 218 ring_req->seg[i] = shadow->seg[i]; 219 220 RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(ring, notify); 221 if (notify) 222 notify_remote_via_irq(info->irq); 223 224 return 0; 225 } 226 227 static void scsifront_gnttab_done(struct vscsifrnt_info *info, 228 struct vscsifrnt_shadow *shadow) 229 { 230 int i; 231 232 if (shadow->sc->sc_data_direction == DMA_NONE) 233 return; 234 235 for (i = 0; i < shadow->nr_grants; i++) { 236 if (unlikely(!gnttab_try_end_foreign_access(shadow->gref[i]))) { 237 shost_printk(KERN_ALERT, info->host, KBUILD_MODNAME 238 "grant still in use by backend\n"); 239 BUG(); 240 } 241 } 242 243 kfree(shadow->sg); 244 } 245 246 static void scsifront_cdb_cmd_done(struct vscsifrnt_info *info, 247 struct vscsiif_response *ring_rsp) 248 { 249 struct vscsifrnt_shadow *shadow; 250 struct scsi_cmnd *sc; 251 uint32_t id; 252 uint8_t sense_len; 253 int result; 254 255 id = ring_rsp->rqid; 256 shadow = info->shadow[id]; 257 sc = shadow->sc; 258 259 BUG_ON(sc == NULL); 260 261 scsifront_gnttab_done(info, shadow); 262 scsifront_put_rqid(info, id); 263 264 result = ring_rsp->rslt; 265 if (result >> 24) 266 set_host_byte(sc, DID_ERROR); 267 else 268 set_host_byte(sc, host_byte(result)); 269 set_status_byte(sc, result & 0xff); 270 scsi_set_resid(sc, ring_rsp->residual_len); 271 272 sense_len = min_t(uint8_t, VSCSIIF_SENSE_BUFFERSIZE, 273 ring_rsp->sense_len); 274 275 if (sense_len) 276 memcpy(sc->sense_buffer, ring_rsp->sense_buffer, sense_len); 277 278 scsi_done(sc); 279 } 280 281 static void scsifront_sync_cmd_done(struct vscsifrnt_info *info, 282 struct vscsiif_response *ring_rsp) 283 { 284 uint16_t id = ring_rsp->rqid; 285 unsigned long flags; 286 struct vscsifrnt_shadow *shadow = info->shadow[id]; 287 int kick; 288 289 spin_lock_irqsave(&info->shadow_lock, flags); 290 shadow->wait_reset = 1; 291 switch (shadow->rslt_reset) { 292 case RSLT_RESET_WAITING: 293 shadow->rslt_reset = ring_rsp->rslt; 294 break; 295 case RSLT_RESET_ERR: 296 kick = _scsifront_put_rqid(info, id); 297 spin_unlock_irqrestore(&info->shadow_lock, flags); 298 kfree(shadow); 299 if (kick) 300 scsifront_wake_up(info); 301 return; 302 default: 303 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME 304 "bad reset state %d, possibly leaking %u\n", 305 shadow->rslt_reset, id); 306 break; 307 } 308 spin_unlock_irqrestore(&info->shadow_lock, flags); 309 310 wake_up(&shadow->wq_reset); 311 } 312 313 static void scsifront_do_response(struct vscsifrnt_info *info, 314 struct vscsiif_response *ring_rsp) 315 { 316 if (WARN(ring_rsp->rqid >= VSCSIIF_MAX_REQS || 317 test_bit(ring_rsp->rqid, info->shadow_free_bitmap), 318 "illegal rqid %u returned by backend!\n", ring_rsp->rqid)) 319 return; 320 321 if (info->shadow[ring_rsp->rqid]->act == VSCSIIF_ACT_SCSI_CDB) 322 scsifront_cdb_cmd_done(info, ring_rsp); 323 else 324 scsifront_sync_cmd_done(info, ring_rsp); 325 } 326 327 static int scsifront_ring_drain(struct vscsifrnt_info *info) 328 { 329 struct vscsiif_response *ring_rsp; 330 RING_IDX i, rp; 331 int more_to_do = 0; 332 333 rp = info->ring.sring->rsp_prod; 334 rmb(); /* ordering required respective to dom0 */ 335 for (i = info->ring.rsp_cons; i != rp; i++) { 336 ring_rsp = RING_GET_RESPONSE(&info->ring, i); 337 scsifront_do_response(info, ring_rsp); 338 } 339 340 info->ring.rsp_cons = i; 341 342 if (i != info->ring.req_prod_pvt) 343 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do); 344 else 345 info->ring.sring->rsp_event = i + 1; 346 347 return more_to_do; 348 } 349 350 static int scsifront_cmd_done(struct vscsifrnt_info *info) 351 { 352 int more_to_do; 353 unsigned long flags; 354 355 spin_lock_irqsave(info->host->host_lock, flags); 356 357 more_to_do = scsifront_ring_drain(info); 358 359 info->wait_ring_available = 0; 360 361 spin_unlock_irqrestore(info->host->host_lock, flags); 362 363 wake_up(&info->wq_sync); 364 365 return more_to_do; 366 } 367 368 static irqreturn_t scsifront_irq_fn(int irq, void *dev_id) 369 { 370 struct vscsifrnt_info *info = dev_id; 371 372 while (scsifront_cmd_done(info)) 373 /* Yield point for this unbounded loop. */ 374 cond_resched(); 375 376 return IRQ_HANDLED; 377 } 378 379 static void scsifront_finish_all(struct vscsifrnt_info *info) 380 { 381 unsigned i; 382 struct vscsiif_response resp; 383 384 scsifront_ring_drain(info); 385 386 for (i = 0; i < VSCSIIF_MAX_REQS; i++) { 387 if (test_bit(i, info->shadow_free_bitmap)) 388 continue; 389 resp.rqid = i; 390 resp.sense_len = 0; 391 resp.rslt = DID_RESET << 16; 392 resp.residual_len = 0; 393 scsifront_do_response(info, &resp); 394 } 395 } 396 397 static int map_data_for_request(struct vscsifrnt_info *info, 398 struct scsi_cmnd *sc, 399 struct vscsifrnt_shadow *shadow) 400 { 401 grant_ref_t gref_head; 402 struct page *page; 403 int err, ref, ref_cnt = 0; 404 int grant_ro = (sc->sc_data_direction == DMA_TO_DEVICE); 405 unsigned int i, off, len, bytes; 406 unsigned int data_len = scsi_bufflen(sc); 407 unsigned int data_grants = 0, seg_grants = 0; 408 struct scatterlist *sg; 409 struct scsiif_request_segment *seg; 410 411 if (sc->sc_data_direction == DMA_NONE || !data_len) 412 return 0; 413 414 scsi_for_each_sg(sc, sg, scsi_sg_count(sc), i) 415 data_grants += PFN_UP(sg->offset + sg->length); 416 417 if (data_grants > VSCSIIF_SG_TABLESIZE) { 418 if (data_grants > info->host->sg_tablesize) { 419 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME 420 "Unable to map request_buffer for command!\n"); 421 return -E2BIG; 422 } 423 seg_grants = vscsiif_grants_sg(data_grants); 424 shadow->sg = kcalloc(data_grants, 425 sizeof(struct scsiif_request_segment), GFP_ATOMIC); 426 if (!shadow->sg) 427 return -ENOMEM; 428 } 429 seg = shadow->sg ? : shadow->seg; 430 431 err = gnttab_alloc_grant_references(seg_grants + data_grants, 432 &gref_head); 433 if (err) { 434 kfree(shadow->sg); 435 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME 436 "gnttab_alloc_grant_references() error\n"); 437 return -ENOMEM; 438 } 439 440 if (seg_grants) { 441 page = virt_to_page(seg); 442 off = offset_in_page(seg); 443 len = sizeof(struct scsiif_request_segment) * data_grants; 444 while (len > 0) { 445 bytes = min_t(unsigned int, len, PAGE_SIZE - off); 446 447 ref = gnttab_claim_grant_reference(&gref_head); 448 BUG_ON(ref == -ENOSPC); 449 450 gnttab_grant_foreign_access_ref(ref, 451 info->dev->otherend_id, 452 xen_page_to_gfn(page), 1); 453 shadow->gref[ref_cnt] = ref; 454 shadow->seg[ref_cnt].gref = ref; 455 shadow->seg[ref_cnt].offset = (uint16_t)off; 456 shadow->seg[ref_cnt].length = (uint16_t)bytes; 457 458 page++; 459 len -= bytes; 460 off = 0; 461 ref_cnt++; 462 } 463 BUG_ON(seg_grants < ref_cnt); 464 seg_grants = ref_cnt; 465 } 466 467 scsi_for_each_sg(sc, sg, scsi_sg_count(sc), i) { 468 page = sg_page(sg); 469 off = sg->offset; 470 len = sg->length; 471 472 while (len > 0 && data_len > 0) { 473 /* 474 * sg sends a scatterlist that is larger than 475 * the data_len it wants transferred for certain 476 * IO sizes. 477 */ 478 bytes = min_t(unsigned int, len, PAGE_SIZE - off); 479 bytes = min(bytes, data_len); 480 481 ref = gnttab_claim_grant_reference(&gref_head); 482 BUG_ON(ref == -ENOSPC); 483 484 gnttab_grant_foreign_access_ref(ref, 485 info->dev->otherend_id, 486 xen_page_to_gfn(page), 487 grant_ro); 488 489 shadow->gref[ref_cnt] = ref; 490 seg->gref = ref; 491 seg->offset = (uint16_t)off; 492 seg->length = (uint16_t)bytes; 493 494 page++; 495 seg++; 496 len -= bytes; 497 data_len -= bytes; 498 off = 0; 499 ref_cnt++; 500 } 501 } 502 503 if (seg_grants) 504 shadow->nr_segments = VSCSIIF_SG_GRANT | seg_grants; 505 else 506 shadow->nr_segments = (uint8_t)ref_cnt; 507 shadow->nr_grants = ref_cnt; 508 509 return 0; 510 } 511 512 static int scsifront_enter(struct vscsifrnt_info *info) 513 { 514 if (info->pause) 515 return 1; 516 info->callers++; 517 return 0; 518 } 519 520 static void scsifront_return(struct vscsifrnt_info *info) 521 { 522 info->callers--; 523 if (info->callers) 524 return; 525 526 if (!info->waiting_pause) 527 return; 528 529 info->waiting_pause = 0; 530 wake_up(&info->wq_pause); 531 } 532 533 static int scsifront_queuecommand(struct Scsi_Host *shost, 534 struct scsi_cmnd *sc) 535 { 536 struct vscsifrnt_info *info = shost_priv(shost); 537 struct vscsifrnt_shadow *shadow = scsi_cmd_priv(sc); 538 unsigned long flags; 539 int err; 540 541 sc->result = 0; 542 543 shadow->sc = sc; 544 shadow->act = VSCSIIF_ACT_SCSI_CDB; 545 546 spin_lock_irqsave(shost->host_lock, flags); 547 if (scsifront_enter(info)) { 548 spin_unlock_irqrestore(shost->host_lock, flags); 549 return SCSI_MLQUEUE_HOST_BUSY; 550 } 551 552 err = map_data_for_request(info, sc, shadow); 553 if (err < 0) { 554 pr_debug("%s: err %d\n", __func__, err); 555 scsifront_return(info); 556 spin_unlock_irqrestore(shost->host_lock, flags); 557 if (err == -ENOMEM) 558 return SCSI_MLQUEUE_HOST_BUSY; 559 sc->result = DID_ERROR << 16; 560 scsi_done(sc); 561 return 0; 562 } 563 564 if (scsifront_do_request(info, shadow)) { 565 scsifront_gnttab_done(info, shadow); 566 goto busy; 567 } 568 569 scsifront_return(info); 570 spin_unlock_irqrestore(shost->host_lock, flags); 571 572 return 0; 573 574 busy: 575 scsifront_return(info); 576 spin_unlock_irqrestore(shost->host_lock, flags); 577 pr_debug("%s: busy\n", __func__); 578 return SCSI_MLQUEUE_HOST_BUSY; 579 } 580 581 /* 582 * Any exception handling (reset or abort) must be forwarded to the backend. 583 * We have to wait until an answer is returned. This answer contains the 584 * result to be returned to the requestor. 585 */ 586 static int scsifront_action_handler(struct scsi_cmnd *sc, uint8_t act) 587 { 588 struct Scsi_Host *host = sc->device->host; 589 struct vscsifrnt_info *info = shost_priv(host); 590 struct vscsifrnt_shadow *shadow, *s = scsi_cmd_priv(sc); 591 int err = 0; 592 593 shadow = kzalloc(sizeof(*shadow), GFP_NOIO); 594 if (!shadow) 595 return FAILED; 596 597 shadow->act = act; 598 shadow->rslt_reset = RSLT_RESET_WAITING; 599 shadow->sc = sc; 600 shadow->ref_rqid = s->rqid; 601 init_waitqueue_head(&shadow->wq_reset); 602 603 spin_lock_irq(host->host_lock); 604 605 for (;;) { 606 if (scsifront_enter(info)) 607 goto fail; 608 609 if (!scsifront_do_request(info, shadow)) 610 break; 611 612 scsifront_return(info); 613 if (err) 614 goto fail; 615 info->wait_ring_available = 1; 616 spin_unlock_irq(host->host_lock); 617 err = wait_event_interruptible(info->wq_sync, 618 !info->wait_ring_available); 619 spin_lock_irq(host->host_lock); 620 } 621 622 spin_unlock_irq(host->host_lock); 623 err = wait_event_interruptible(shadow->wq_reset, shadow->wait_reset); 624 spin_lock_irq(host->host_lock); 625 626 if (!err) { 627 err = shadow->rslt_reset; 628 scsifront_put_rqid(info, shadow->rqid); 629 kfree(shadow); 630 } else { 631 spin_lock(&info->shadow_lock); 632 shadow->rslt_reset = RSLT_RESET_ERR; 633 spin_unlock(&info->shadow_lock); 634 err = FAILED; 635 } 636 637 scsifront_return(info); 638 spin_unlock_irq(host->host_lock); 639 return err; 640 641 fail: 642 spin_unlock_irq(host->host_lock); 643 kfree(shadow); 644 return FAILED; 645 } 646 647 static int scsifront_eh_abort_handler(struct scsi_cmnd *sc) 648 { 649 pr_debug("%s\n", __func__); 650 return scsifront_action_handler(sc, VSCSIIF_ACT_SCSI_ABORT); 651 } 652 653 static int scsifront_dev_reset_handler(struct scsi_cmnd *sc) 654 { 655 pr_debug("%s\n", __func__); 656 return scsifront_action_handler(sc, VSCSIIF_ACT_SCSI_RESET); 657 } 658 659 static int scsifront_sdev_configure(struct scsi_device *sdev) 660 { 661 struct vscsifrnt_info *info = shost_priv(sdev->host); 662 int err; 663 664 if (info && current == info->curr) { 665 err = xenbus_printf(XBT_NIL, info->dev->nodename, 666 info->dev_state_path, "%d", XenbusStateConnected); 667 if (err) { 668 xenbus_dev_error(info->dev, err, 669 "%s: writing dev_state_path", __func__); 670 return err; 671 } 672 } 673 674 return 0; 675 } 676 677 static void scsifront_sdev_destroy(struct scsi_device *sdev) 678 { 679 struct vscsifrnt_info *info = shost_priv(sdev->host); 680 int err; 681 682 if (info && current == info->curr) { 683 err = xenbus_printf(XBT_NIL, info->dev->nodename, 684 info->dev_state_path, "%d", XenbusStateClosed); 685 if (err) 686 xenbus_dev_error(info->dev, err, 687 "%s: writing dev_state_path", __func__); 688 } 689 } 690 691 static struct scsi_host_template scsifront_sht = { 692 .module = THIS_MODULE, 693 .name = "Xen SCSI frontend driver", 694 .queuecommand = scsifront_queuecommand, 695 .eh_abort_handler = scsifront_eh_abort_handler, 696 .eh_device_reset_handler = scsifront_dev_reset_handler, 697 .slave_configure = scsifront_sdev_configure, 698 .slave_destroy = scsifront_sdev_destroy, 699 .cmd_per_lun = VSCSIIF_DEFAULT_CMD_PER_LUN, 700 .can_queue = VSCSIIF_MAX_REQS, 701 .this_id = -1, 702 .cmd_size = sizeof(struct vscsifrnt_shadow), 703 .sg_tablesize = VSCSIIF_SG_TABLESIZE, 704 .proc_name = "scsifront", 705 }; 706 707 static int scsifront_alloc_ring(struct vscsifrnt_info *info) 708 { 709 struct xenbus_device *dev = info->dev; 710 struct vscsiif_sring *sring; 711 grant_ref_t gref; 712 int err = -ENOMEM; 713 714 /***** Frontend to Backend ring start *****/ 715 sring = (struct vscsiif_sring *)__get_free_page(GFP_KERNEL); 716 if (!sring) { 717 xenbus_dev_fatal(dev, err, 718 "fail to allocate shared ring (Front to Back)"); 719 return err; 720 } 721 SHARED_RING_INIT(sring); 722 FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE); 723 724 err = xenbus_grant_ring(dev, sring, 1, &gref); 725 if (err < 0) { 726 free_page((unsigned long)sring); 727 xenbus_dev_fatal(dev, err, 728 "fail to grant shared ring (Front to Back)"); 729 return err; 730 } 731 info->ring_ref = gref; 732 733 err = xenbus_alloc_evtchn(dev, &info->evtchn); 734 if (err) { 735 xenbus_dev_fatal(dev, err, "xenbus_alloc_evtchn"); 736 goto free_gnttab; 737 } 738 739 err = bind_evtchn_to_irq(info->evtchn); 740 if (err <= 0) { 741 xenbus_dev_fatal(dev, err, "bind_evtchn_to_irq"); 742 goto free_gnttab; 743 } 744 745 info->irq = err; 746 747 err = request_threaded_irq(info->irq, NULL, scsifront_irq_fn, 748 IRQF_ONESHOT, "scsifront", info); 749 if (err) { 750 xenbus_dev_fatal(dev, err, "request_threaded_irq"); 751 goto free_irq; 752 } 753 754 return 0; 755 756 /* free resource */ 757 free_irq: 758 unbind_from_irqhandler(info->irq, info); 759 free_gnttab: 760 gnttab_end_foreign_access(info->ring_ref, 761 (unsigned long)info->ring.sring); 762 763 return err; 764 } 765 766 static void scsifront_free_ring(struct vscsifrnt_info *info) 767 { 768 unbind_from_irqhandler(info->irq, info); 769 gnttab_end_foreign_access(info->ring_ref, 770 (unsigned long)info->ring.sring); 771 } 772 773 static int scsifront_init_ring(struct vscsifrnt_info *info) 774 { 775 struct xenbus_device *dev = info->dev; 776 struct xenbus_transaction xbt; 777 int err; 778 779 pr_debug("%s\n", __func__); 780 781 err = scsifront_alloc_ring(info); 782 if (err) 783 return err; 784 pr_debug("%s: %u %u\n", __func__, info->ring_ref, info->evtchn); 785 786 again: 787 err = xenbus_transaction_start(&xbt); 788 if (err) 789 xenbus_dev_fatal(dev, err, "starting transaction"); 790 791 err = xenbus_printf(xbt, dev->nodename, "ring-ref", "%u", 792 info->ring_ref); 793 if (err) { 794 xenbus_dev_fatal(dev, err, "%s", "writing ring-ref"); 795 goto fail; 796 } 797 798 err = xenbus_printf(xbt, dev->nodename, "event-channel", "%u", 799 info->evtchn); 800 801 if (err) { 802 xenbus_dev_fatal(dev, err, "%s", "writing event-channel"); 803 goto fail; 804 } 805 806 err = xenbus_transaction_end(xbt, 0); 807 if (err) { 808 if (err == -EAGAIN) 809 goto again; 810 xenbus_dev_fatal(dev, err, "completing transaction"); 811 goto free_sring; 812 } 813 814 return 0; 815 816 fail: 817 xenbus_transaction_end(xbt, 1); 818 free_sring: 819 scsifront_free_ring(info); 820 821 return err; 822 } 823 824 825 static int scsifront_probe(struct xenbus_device *dev, 826 const struct xenbus_device_id *id) 827 { 828 struct vscsifrnt_info *info; 829 struct Scsi_Host *host; 830 int err = -ENOMEM; 831 char name[TASK_COMM_LEN]; 832 833 host = scsi_host_alloc(&scsifront_sht, sizeof(*info)); 834 if (!host) { 835 xenbus_dev_fatal(dev, err, "fail to allocate scsi host"); 836 return err; 837 } 838 info = (struct vscsifrnt_info *)host->hostdata; 839 840 dev_set_drvdata(&dev->dev, info); 841 info->dev = dev; 842 843 bitmap_fill(info->shadow_free_bitmap, VSCSIIF_MAX_REQS); 844 845 err = scsifront_init_ring(info); 846 if (err) { 847 scsi_host_put(host); 848 return err; 849 } 850 851 init_waitqueue_head(&info->wq_sync); 852 init_waitqueue_head(&info->wq_pause); 853 spin_lock_init(&info->shadow_lock); 854 855 snprintf(name, TASK_COMM_LEN, "vscsiif.%d", host->host_no); 856 857 host->max_id = VSCSIIF_MAX_TARGET; 858 host->max_channel = 0; 859 host->max_lun = VSCSIIF_MAX_LUN; 860 host->max_sectors = (host->sg_tablesize - 1) * PAGE_SIZE / 512; 861 host->max_cmd_len = VSCSIIF_MAX_COMMAND_SIZE; 862 863 err = scsi_add_host(host, &dev->dev); 864 if (err) { 865 dev_err(&dev->dev, "fail to add scsi host %d\n", err); 866 goto free_sring; 867 } 868 info->host = host; 869 info->host_active = 1; 870 871 xenbus_switch_state(dev, XenbusStateInitialised); 872 873 return 0; 874 875 free_sring: 876 scsifront_free_ring(info); 877 scsi_host_put(host); 878 return err; 879 } 880 881 static int scsifront_resume(struct xenbus_device *dev) 882 { 883 struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev); 884 struct Scsi_Host *host = info->host; 885 int err; 886 887 spin_lock_irq(host->host_lock); 888 889 /* Finish all still pending commands. */ 890 scsifront_finish_all(info); 891 892 spin_unlock_irq(host->host_lock); 893 894 /* Reconnect to dom0. */ 895 scsifront_free_ring(info); 896 err = scsifront_init_ring(info); 897 if (err) { 898 dev_err(&dev->dev, "fail to resume %d\n", err); 899 scsi_host_put(host); 900 return err; 901 } 902 903 xenbus_switch_state(dev, XenbusStateInitialised); 904 905 return 0; 906 } 907 908 static int scsifront_suspend(struct xenbus_device *dev) 909 { 910 struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev); 911 struct Scsi_Host *host = info->host; 912 int err = 0; 913 914 /* No new commands for the backend. */ 915 spin_lock_irq(host->host_lock); 916 info->pause = 1; 917 while (info->callers && !err) { 918 info->waiting_pause = 1; 919 info->wait_ring_available = 0; 920 spin_unlock_irq(host->host_lock); 921 wake_up(&info->wq_sync); 922 err = wait_event_interruptible(info->wq_pause, 923 !info->waiting_pause); 924 spin_lock_irq(host->host_lock); 925 } 926 spin_unlock_irq(host->host_lock); 927 return err; 928 } 929 930 static int scsifront_remove(struct xenbus_device *dev) 931 { 932 struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev); 933 934 pr_debug("%s: %s removed\n", __func__, dev->nodename); 935 936 mutex_lock(&scsifront_mutex); 937 if (info->host_active) { 938 /* Scsi_host not yet removed */ 939 scsi_remove_host(info->host); 940 info->host_active = 0; 941 } 942 mutex_unlock(&scsifront_mutex); 943 944 scsifront_free_ring(info); 945 scsi_host_put(info->host); 946 947 return 0; 948 } 949 950 static void scsifront_disconnect(struct vscsifrnt_info *info) 951 { 952 struct xenbus_device *dev = info->dev; 953 struct Scsi_Host *host = info->host; 954 955 pr_debug("%s: %s disconnect\n", __func__, dev->nodename); 956 957 /* 958 * When this function is executed, all devices of 959 * Frontend have been deleted. 960 * Therefore, it need not block I/O before remove_host. 961 */ 962 963 mutex_lock(&scsifront_mutex); 964 if (info->host_active) { 965 scsi_remove_host(host); 966 info->host_active = 0; 967 } 968 mutex_unlock(&scsifront_mutex); 969 970 xenbus_frontend_closed(dev); 971 } 972 973 static void scsifront_do_lun_hotplug(struct vscsifrnt_info *info, int op) 974 { 975 struct xenbus_device *dev = info->dev; 976 int i, err = 0; 977 char str[64]; 978 char **dir; 979 unsigned int dir_n = 0; 980 unsigned int device_state; 981 unsigned int hst, chn, tgt, lun; 982 struct scsi_device *sdev; 983 984 dir = xenbus_directory(XBT_NIL, dev->otherend, "vscsi-devs", &dir_n); 985 if (IS_ERR(dir)) 986 return; 987 988 /* mark current task as the one allowed to modify device states */ 989 BUG_ON(info->curr); 990 info->curr = current; 991 992 for (i = 0; i < dir_n; i++) { 993 /* read status */ 994 snprintf(str, sizeof(str), "vscsi-devs/%s/state", dir[i]); 995 err = xenbus_scanf(XBT_NIL, dev->otherend, str, "%u", 996 &device_state); 997 if (XENBUS_EXIST_ERR(err)) 998 continue; 999 1000 /* virtual SCSI device */ 1001 snprintf(str, sizeof(str), "vscsi-devs/%s/v-dev", dir[i]); 1002 err = xenbus_scanf(XBT_NIL, dev->otherend, str, 1003 "%u:%u:%u:%u", &hst, &chn, &tgt, &lun); 1004 if (XENBUS_EXIST_ERR(err)) 1005 continue; 1006 1007 /* 1008 * Front device state path, used in slave_configure called 1009 * on successfull scsi_add_device, and in slave_destroy called 1010 * on remove of a device. 1011 */ 1012 snprintf(info->dev_state_path, sizeof(info->dev_state_path), 1013 "vscsi-devs/%s/state", dir[i]); 1014 1015 switch (op) { 1016 case VSCSIFRONT_OP_ADD_LUN: 1017 if (device_state != XenbusStateInitialised) 1018 break; 1019 1020 if (scsi_add_device(info->host, chn, tgt, lun)) { 1021 dev_err(&dev->dev, "scsi_add_device\n"); 1022 err = xenbus_printf(XBT_NIL, dev->nodename, 1023 info->dev_state_path, 1024 "%d", XenbusStateClosed); 1025 if (err) 1026 xenbus_dev_error(dev, err, 1027 "%s: writing dev_state_path", __func__); 1028 } 1029 break; 1030 case VSCSIFRONT_OP_DEL_LUN: 1031 if (device_state != XenbusStateClosing) 1032 break; 1033 1034 sdev = scsi_device_lookup(info->host, chn, tgt, lun); 1035 if (sdev) { 1036 scsi_remove_device(sdev); 1037 scsi_device_put(sdev); 1038 } 1039 break; 1040 case VSCSIFRONT_OP_READD_LUN: 1041 if (device_state == XenbusStateConnected) { 1042 err = xenbus_printf(XBT_NIL, dev->nodename, 1043 info->dev_state_path, 1044 "%d", XenbusStateConnected); 1045 if (err) 1046 xenbus_dev_error(dev, err, 1047 "%s: writing dev_state_path", __func__); 1048 } 1049 break; 1050 default: 1051 break; 1052 } 1053 } 1054 1055 info->curr = NULL; 1056 1057 kfree(dir); 1058 } 1059 1060 static void scsifront_read_backend_params(struct xenbus_device *dev, 1061 struct vscsifrnt_info *info) 1062 { 1063 unsigned int sg_grant, nr_segs; 1064 struct Scsi_Host *host = info->host; 1065 1066 sg_grant = xenbus_read_unsigned(dev->otherend, "feature-sg-grant", 0); 1067 nr_segs = min_t(unsigned int, sg_grant, SG_ALL); 1068 nr_segs = max_t(unsigned int, nr_segs, VSCSIIF_SG_TABLESIZE); 1069 nr_segs = min_t(unsigned int, nr_segs, 1070 VSCSIIF_SG_TABLESIZE * PAGE_SIZE / 1071 sizeof(struct scsiif_request_segment)); 1072 1073 if (!info->pause && sg_grant) 1074 dev_info(&dev->dev, "using up to %d SG entries\n", nr_segs); 1075 else if (info->pause && nr_segs < host->sg_tablesize) 1076 dev_warn(&dev->dev, 1077 "SG entries decreased from %d to %u - device may not work properly anymore\n", 1078 host->sg_tablesize, nr_segs); 1079 1080 host->sg_tablesize = nr_segs; 1081 host->max_sectors = (nr_segs - 1) * PAGE_SIZE / 512; 1082 } 1083 1084 static void scsifront_backend_changed(struct xenbus_device *dev, 1085 enum xenbus_state backend_state) 1086 { 1087 struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev); 1088 1089 pr_debug("%s: %p %u %u\n", __func__, dev, dev->state, backend_state); 1090 1091 switch (backend_state) { 1092 case XenbusStateUnknown: 1093 case XenbusStateInitialising: 1094 case XenbusStateInitWait: 1095 case XenbusStateInitialised: 1096 break; 1097 1098 case XenbusStateConnected: 1099 scsifront_read_backend_params(dev, info); 1100 1101 if (info->pause) { 1102 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_READD_LUN); 1103 xenbus_switch_state(dev, XenbusStateConnected); 1104 info->pause = 0; 1105 return; 1106 } 1107 1108 if (xenbus_read_driver_state(dev->nodename) == 1109 XenbusStateInitialised) 1110 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_ADD_LUN); 1111 1112 if (dev->state != XenbusStateConnected) 1113 xenbus_switch_state(dev, XenbusStateConnected); 1114 break; 1115 1116 case XenbusStateClosed: 1117 if (dev->state == XenbusStateClosed) 1118 break; 1119 fallthrough; /* Missed the backend's Closing state */ 1120 case XenbusStateClosing: 1121 scsifront_disconnect(info); 1122 break; 1123 1124 case XenbusStateReconfiguring: 1125 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_DEL_LUN); 1126 xenbus_switch_state(dev, XenbusStateReconfiguring); 1127 break; 1128 1129 case XenbusStateReconfigured: 1130 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_ADD_LUN); 1131 xenbus_switch_state(dev, XenbusStateConnected); 1132 break; 1133 } 1134 } 1135 1136 static const struct xenbus_device_id scsifront_ids[] = { 1137 { "vscsi" }, 1138 { "" } 1139 }; 1140 1141 static struct xenbus_driver scsifront_driver = { 1142 .ids = scsifront_ids, 1143 .probe = scsifront_probe, 1144 .remove = scsifront_remove, 1145 .resume = scsifront_resume, 1146 .suspend = scsifront_suspend, 1147 .otherend_changed = scsifront_backend_changed, 1148 }; 1149 1150 static int __init scsifront_init(void) 1151 { 1152 if (!xen_domain()) 1153 return -ENODEV; 1154 1155 return xenbus_register_frontend(&scsifront_driver); 1156 } 1157 module_init(scsifront_init); 1158 1159 static void __exit scsifront_exit(void) 1160 { 1161 xenbus_unregister_driver(&scsifront_driver); 1162 } 1163 module_exit(scsifront_exit); 1164 1165 MODULE_DESCRIPTION("Xen SCSI frontend driver"); 1166 MODULE_LICENSE("GPL"); 1167 MODULE_ALIAS("xen:vscsi"); 1168 MODULE_AUTHOR("Juergen Gross <jgross@suse.com>"); 1169