drivers/scsi/xen-scsifront.c

   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
  67 /* Tuning point. */
  68 #define VSCSIIF_DEFAULT_CMD_PER_LUN 10
  69 #define VSCSIIF_MAX_TARGET          64
  70 #define VSCSIIF_MAX_LUN             255
  71
  72 #define VSCSIIF_RING_SIZE       __CONST_RING_SIZE(vscsiif, PAGE_SIZE)
  73 #define VSCSIIF_MAX_REQS        VSCSIIF_RING_SIZE
  74
  75 #define vscsiif_grants_sg(_sg)  (PFN_UP((_sg) *         \
  76                                 sizeof(struct scsiif_request_segment)))
  77
  78 struct vscsifrnt_shadow {
  79         /* command between backend and frontend */
  80         unsigned char act;
  81         uint16_t rqid;
  82
  83         unsigned int nr_grants;         /* number of grants in gref[] */
  84         struct scsiif_request_segment *sg;      /* scatter/gather elements */
  85
  86         /* Do reset or abort function. */
  87         wait_queue_head_t wq_reset;     /* reset work queue           */
  88         int wait_reset;                 /* reset work queue condition */
  89         int32_t rslt_reset;             /* reset response status:     */
  90                                         /* SUCCESS or FAILED or:      */
  91 #define RSLT_RESET_WAITING      0
  92 #define RSLT_RESET_ERR          -1
  93
  94         /* Requested struct scsi_cmnd is stored from kernel. */
  95         struct scsi_cmnd *sc;
  96         int gref[vscsiif_grants_sg(SG_ALL) + SG_ALL];
  97 };
  98
  99 struct vscsifrnt_info {
 100         struct xenbus_device *dev;
 101
 102         struct Scsi_Host *host;
 103         int host_active;
 104
 105         unsigned int evtchn;
 106         unsigned int irq;
 107
 108         grant_ref_t ring_ref;
 109         struct vscsiif_front_ring ring;
 110         struct vscsiif_response ring_rsp;
 111
 112         spinlock_t shadow_lock;
 113         DECLARE_BITMAP(shadow_free_bitmap, VSCSIIF_MAX_REQS);
 114         struct vscsifrnt_shadow *shadow[VSCSIIF_MAX_REQS];
 115
 116         wait_queue_head_t wq_sync;
 117         unsigned int wait_ring_available:1;
 118
 119         char dev_state_path[64];
 120         struct task_struct *curr;
 121 };
 122
 123 static DEFINE_MUTEX(scsifront_mutex);
 124
 125 static void scsifront_wake_up(struct vscsifrnt_info *info)
 126 {
 127         info->wait_ring_available = 0;
 128         wake_up(&info->wq_sync);
 129 }
 130
 131 static int scsifront_get_rqid(struct vscsifrnt_info *info)
 132 {
 133         unsigned long flags;
 134         int free;
 135
 136         spin_lock_irqsave(&info->shadow_lock, flags);
 137
 138         free = find_first_bit(info->shadow_free_bitmap, VSCSIIF_MAX_REQS);
 139         __clear_bit(free, info->shadow_free_bitmap);
 140
 141         spin_unlock_irqrestore(&info->shadow_lock, flags);
 142
 143         return free;
 144 }
 145
 146 static int _scsifront_put_rqid(struct vscsifrnt_info *info, uint32_t id)
 147 {
 148         int empty = bitmap_empty(info->shadow_free_bitmap, VSCSIIF_MAX_REQS);
 149
 150         __set_bit(id, info->shadow_free_bitmap);
 151         info->shadow[id] = NULL;
 152
 153         return empty || info->wait_ring_available;
 154 }
 155
 156 static void scsifront_put_rqid(struct vscsifrnt_info *info, uint32_t id)
 157 {
 158         unsigned long flags;
 159         int kick;
 160
 161         spin_lock_irqsave(&info->shadow_lock, flags);
 162         kick = _scsifront_put_rqid(info, id);
 163         spin_unlock_irqrestore(&info->shadow_lock, flags);
 164
 165         if (kick)
 166                 scsifront_wake_up(info);
 167 }
 168
 169 static struct vscsiif_request *scsifront_pre_req(struct vscsifrnt_info *info)
 170 {
 171         struct vscsiif_front_ring *ring = &(info->ring);
 172         struct vscsiif_request *ring_req;
 173         uint32_t id;
 174
 175         id = scsifront_get_rqid(info);  /* use id in response */
 176         if (id >= VSCSIIF_MAX_REQS)
 177                 return NULL;
 178
 179         ring_req = RING_GET_REQUEST(&(info->ring), ring->req_prod_pvt);
 180
 181         ring->req_prod_pvt++;
 182
 183         ring_req->rqid = (uint16_t)id;
 184
 185         return ring_req;
 186 }
 187
 188 static void scsifront_do_request(struct vscsifrnt_info *info)
 189 {
 190         struct vscsiif_front_ring *ring = &(info->ring);
 191         int notify;
 192
 193         RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(ring, notify);
 194         if (notify)
 195                 notify_remote_via_irq(info->irq);
 196 }
 197
 198 static void scsifront_gnttab_done(struct vscsifrnt_info *info, uint32_t id)
 199 {
 200         struct vscsifrnt_shadow *s = info->shadow[id];
 201         int i;
 202
 203         if (s->sc->sc_data_direction == DMA_NONE)
 204                 return;
 205
 206         for (i = 0; i < s->nr_grants; i++) {
 207                 if (unlikely(gnttab_query_foreign_access(s->gref[i]) != 0)) {
 208                         shost_printk(KERN_ALERT, info->host, KBUILD_MODNAME
 209                                      "grant still in use by backend\n");
 210                         BUG();
 211                 }
 212                 gnttab_end_foreign_access(s->gref[i], 0, 0UL);
 213         }
 214
 215         kfree(s->sg);
 216 }
 217
 218 static void scsifront_cdb_cmd_done(struct vscsifrnt_info *info,
 219                                    struct vscsiif_response *ring_rsp)
 220 {
 221         struct scsi_cmnd *sc;
 222         uint32_t id;
 223         uint8_t sense_len;
 224
 225         id = ring_rsp->rqid;
 226         sc = info->shadow[id]->sc;
 227
 228         BUG_ON(sc == NULL);
 229
 230         scsifront_gnttab_done(info, id);
 231         scsifront_put_rqid(info, id);
 232
 233         sc->result = ring_rsp->rslt;
 234         scsi_set_resid(sc, ring_rsp->residual_len);
 235
 236         sense_len = min_t(uint8_t, VSCSIIF_SENSE_BUFFERSIZE,
 237                           ring_rsp->sense_len);
 238
 239         if (sense_len)
 240                 memcpy(sc->sense_buffer, ring_rsp->sense_buffer, sense_len);
 241
 242         sc->scsi_done(sc);
 243 }
 244
 245 static void scsifront_sync_cmd_done(struct vscsifrnt_info *info,
 246                                     struct vscsiif_response *ring_rsp)
 247 {
 248         uint16_t id = ring_rsp->rqid;
 249         unsigned long flags;
 250         struct vscsifrnt_shadow *shadow = info->shadow[id];
 251         int kick;
 252
 253         spin_lock_irqsave(&info->shadow_lock, flags);
 254         shadow->wait_reset = 1;
 255         switch (shadow->rslt_reset) {
 256         case RSLT_RESET_WAITING:
 257                 shadow->rslt_reset = ring_rsp->rslt;
 258                 break;
 259         case RSLT_RESET_ERR:
 260                 kick = _scsifront_put_rqid(info, id);
 261                 spin_unlock_irqrestore(&info->shadow_lock, flags);
 262                 kfree(shadow);
 263                 if (kick)
 264                         scsifront_wake_up(info);
 265                 return;
 266         default:
 267                 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME
 268                              "bad reset state %d, possibly leaking %u\n",
 269                              shadow->rslt_reset, id);
 270                 break;
 271         }
 272         spin_unlock_irqrestore(&info->shadow_lock, flags);
 273
 274         wake_up(&shadow->wq_reset);
 275 }
 276
 277 static int scsifront_cmd_done(struct vscsifrnt_info *info)
 278 {
 279         struct vscsiif_response *ring_rsp;
 280         RING_IDX i, rp;
 281         int more_to_do = 0;
 282         unsigned long flags;
 283
 284         spin_lock_irqsave(info->host->host_lock, flags);
 285
 286         rp = info->ring.sring->rsp_prod;
 287         rmb();  /* ordering required respective to dom0 */
 288         for (i = info->ring.rsp_cons; i != rp; i++) {
 289
 290                 ring_rsp = RING_GET_RESPONSE(&info->ring, i);
 291
 292                 if (WARN(ring_rsp->rqid >= VSCSIIF_MAX_REQS ||
 293                          test_bit(ring_rsp->rqid, info->shadow_free_bitmap),
 294                          "illegal rqid %u returned by backend!\n",
 295                          ring_rsp->rqid))
 296                         continue;
 297
 298                 if (info->shadow[ring_rsp->rqid]->act == VSCSIIF_ACT_SCSI_CDB)
 299                         scsifront_cdb_cmd_done(info, ring_rsp);
 300                 else
 301                         scsifront_sync_cmd_done(info, ring_rsp);
 302         }
 303
 304         info->ring.rsp_cons = i;
 305
 306         if (i != info->ring.req_prod_pvt)
 307                 RING_FINAL_CHECK_FOR_RESPONSES(&info->ring, more_to_do);
 308         else
 309                 info->ring.sring->rsp_event = i + 1;
 310
 311         info->wait_ring_available = 0;
 312
 313         spin_unlock_irqrestore(info->host->host_lock, flags);
 314
 315         wake_up(&info->wq_sync);
 316
 317         return more_to_do;
 318 }
 319
 320 static irqreturn_t scsifront_irq_fn(int irq, void *dev_id)
 321 {
 322         struct vscsifrnt_info *info = dev_id;
 323
 324         while (scsifront_cmd_done(info))
 325                 /* Yield point for this unbounded loop. */
 326                 cond_resched();
 327
 328         return IRQ_HANDLED;
 329 }
 330
 331 static int map_data_for_request(struct vscsifrnt_info *info,
 332                                 struct scsi_cmnd *sc,
 333                                 struct vscsiif_request *ring_req,
 334                                 struct vscsifrnt_shadow *shadow)
 335 {
 336         grant_ref_t gref_head;
 337         struct page *page;
 338         int err, ref, ref_cnt = 0;
 339         int grant_ro = (sc->sc_data_direction == DMA_TO_DEVICE);
 340         unsigned int i, off, len, bytes;
 341         unsigned int data_len = scsi_bufflen(sc);
 342         unsigned int data_grants = 0, seg_grants = 0;
 343         struct scatterlist *sg;
 344         unsigned long mfn;
 345         struct scsiif_request_segment *seg;
 346
 347         ring_req->nr_segments = 0;
 348         if (sc->sc_data_direction == DMA_NONE || !data_len)
 349                 return 0;
 350
 351         scsi_for_each_sg(sc, sg, scsi_sg_count(sc), i)
 352                 data_grants += PFN_UP(sg->offset + sg->length);
 353
 354         if (data_grants > VSCSIIF_SG_TABLESIZE) {
 355                 if (data_grants > info->host->sg_tablesize) {
 356                         shost_printk(KERN_ERR, info->host, KBUILD_MODNAME
 357                              "Unable to map request_buffer for command!\n");
 358                         return -E2BIG;
 359                 }
 360                 seg_grants = vscsiif_grants_sg(data_grants);
 361                 shadow->sg = kcalloc(data_grants,
 362                         sizeof(struct scsiif_request_segment), GFP_ATOMIC);
 363                 if (!shadow->sg)
 364                         return -ENOMEM;
 365         }
 366         seg = shadow->sg ? : ring_req->seg;
 367
 368         err = gnttab_alloc_grant_references(seg_grants + data_grants,
 369                                             &gref_head);
 370         if (err) {
 371                 kfree(shadow->sg);
 372                 shost_printk(KERN_ERR, info->host, KBUILD_MODNAME
 373                              "gnttab_alloc_grant_references() error\n");
 374                 return -ENOMEM;
 375         }
 376
 377         if (seg_grants) {
 378                 page = virt_to_page(seg);
 379                 off = (unsigned long)seg & ~PAGE_MASK;
 380                 len = sizeof(struct scsiif_request_segment) * data_grants;
 381                 while (len > 0) {
 382                         bytes = min_t(unsigned int, len, PAGE_SIZE - off);
 383
 384                         ref = gnttab_claim_grant_reference(&gref_head);
 385                         BUG_ON(ref == -ENOSPC);
 386
 387                         mfn = pfn_to_mfn(page_to_pfn(page));
 388                         gnttab_grant_foreign_access_ref(ref,
 389                                 info->dev->otherend_id, mfn, 1);
 390                         shadow->gref[ref_cnt] = ref;
 391                         ring_req->seg[ref_cnt].gref   = ref;
 392                         ring_req->seg[ref_cnt].offset = (uint16_t)off;
 393                         ring_req->seg[ref_cnt].length = (uint16_t)bytes;
 394
 395                         page++;
 396                         len -= bytes;
 397                         off = 0;
 398                         ref_cnt++;
 399                 }
 400                 BUG_ON(seg_grants < ref_cnt);
 401                 seg_grants = ref_cnt;
 402         }
 403
 404         scsi_for_each_sg(sc, sg, scsi_sg_count(sc), i) {
 405                 page = sg_page(sg);
 406                 off = sg->offset;
 407                 len = sg->length;
 408
 409                 while (len > 0 && data_len > 0) {
 410                         /*
 411                          * sg sends a scatterlist that is larger than
 412                          * the data_len it wants transferred for certain
 413                          * IO sizes.
 414                          */
 415                         bytes = min_t(unsigned int, len, PAGE_SIZE - off);
 416                         bytes = min(bytes, data_len);
 417
 418                         ref = gnttab_claim_grant_reference(&gref_head);
 419                         BUG_ON(ref == -ENOSPC);
 420
 421                         mfn = pfn_to_mfn(page_to_pfn(page));
 422                         gnttab_grant_foreign_access_ref(ref,
 423                                 info->dev->otherend_id, mfn, grant_ro);
 424
 425                         shadow->gref[ref_cnt] = ref;
 426                         seg->gref   = ref;
 427                         seg->offset = (uint16_t)off;
 428                         seg->length = (uint16_t)bytes;
 429
 430                         page++;
 431                         seg++;
 432                         len -= bytes;
 433                         data_len -= bytes;
 434                         off = 0;
 435                         ref_cnt++;
 436                 }
 437         }
 438
 439         if (seg_grants)
 440                 ring_req->nr_segments = VSCSIIF_SG_GRANT | seg_grants;
 441         else
 442                 ring_req->nr_segments = (uint8_t)ref_cnt;
 443         shadow->nr_grants = ref_cnt;
 444
 445         return 0;
 446 }
 447
 448 static struct vscsiif_request *scsifront_command2ring(
 449                 struct vscsifrnt_info *info, struct scsi_cmnd *sc,
 450                 struct vscsifrnt_shadow *shadow)
 451 {
 452         struct vscsiif_request *ring_req;
 453
 454         memset(shadow, 0, sizeof(*shadow));
 455
 456         ring_req = scsifront_pre_req(info);
 457         if (!ring_req)
 458                 return NULL;
 459
 460         info->shadow[ring_req->rqid] = shadow;
 461         shadow->rqid = ring_req->rqid;
 462
 463         ring_req->id      = sc->device->id;
 464         ring_req->lun     = sc->device->lun;
 465         ring_req->channel = sc->device->channel;
 466         ring_req->cmd_len = sc->cmd_len;
 467
 468         BUG_ON(sc->cmd_len > VSCSIIF_MAX_COMMAND_SIZE);
 469
 470         memcpy(ring_req->cmnd, sc->cmnd, sc->cmd_len);
 471
 472         ring_req->sc_data_direction   = (uint8_t)sc->sc_data_direction;
 473         ring_req->timeout_per_command = sc->request->timeout / HZ;
 474
 475         return ring_req;
 476 }
 477
 478 static int scsifront_queuecommand(struct Scsi_Host *shost,
 479                                   struct scsi_cmnd *sc)
 480 {
 481         struct vscsifrnt_info *info = shost_priv(shost);
 482         struct vscsiif_request *ring_req;
 483         struct vscsifrnt_shadow *shadow = scsi_cmd_priv(sc);
 484         unsigned long flags;
 485         int err;
 486         uint16_t rqid;
 487
 488         spin_lock_irqsave(shost->host_lock, flags);
 489         if (RING_FULL(&info->ring))
 490                 goto busy;
 491
 492         ring_req = scsifront_command2ring(info, sc, shadow);
 493         if (!ring_req)
 494                 goto busy;
 495
 496         sc->result = 0;
 497
 498         rqid = ring_req->rqid;
 499         ring_req->act = VSCSIIF_ACT_SCSI_CDB;
 500
 501         shadow->sc  = sc;
 502         shadow->act = VSCSIIF_ACT_SCSI_CDB;
 503
 504         err = map_data_for_request(info, sc, ring_req, shadow);
 505         if (err < 0) {
 506                 pr_debug("%s: err %d\n", __func__, err);
 507                 scsifront_put_rqid(info, rqid);
 508                 spin_unlock_irqrestore(shost->host_lock, flags);
 509                 if (err == -ENOMEM)
 510                         return SCSI_MLQUEUE_HOST_BUSY;
 511                 sc->result = DID_ERROR << 16;
 512                 sc->scsi_done(sc);
 513                 return 0;
 514         }
 515
 516         scsifront_do_request(info);
 517         spin_unlock_irqrestore(shost->host_lock, flags);
 518
 519         return 0;
 520
 521 busy:
 522         spin_unlock_irqrestore(shost->host_lock, flags);
 523         pr_debug("%s: busy\n", __func__);
 524         return SCSI_MLQUEUE_HOST_BUSY;
 525 }
 526
 527 /*
 528  * Any exception handling (reset or abort) must be forwarded to the backend.
 529  * We have to wait until an answer is returned. This answer contains the
 530  * result to be returned to the requestor.
 531  */
 532 static int scsifront_action_handler(struct scsi_cmnd *sc, uint8_t act)
 533 {
 534         struct Scsi_Host *host = sc->device->host;
 535         struct vscsifrnt_info *info = shost_priv(host);
 536         struct vscsifrnt_shadow *shadow, *s = scsi_cmd_priv(sc);
 537         struct vscsiif_request *ring_req;
 538         int err = 0;
 539
 540         shadow = kmalloc(sizeof(*shadow), GFP_NOIO);
 541         if (!shadow)
 542                 return FAILED;
 543
 544         spin_lock_irq(host->host_lock);
 545
 546         for (;;) {
 547                 if (!RING_FULL(&info->ring)) {
 548                         ring_req = scsifront_command2ring(info, sc, shadow);
 549                         if (ring_req)
 550                                 break;
 551                 }
 552                 if (err) {
 553                         spin_unlock_irq(host->host_lock);
 554                         kfree(shadow);
 555                         return FAILED;
 556                 }
 557                 info->wait_ring_available = 1;
 558                 spin_unlock_irq(host->host_lock);
 559                 err = wait_event_interruptible(info->wq_sync,
 560                                                !info->wait_ring_available);
 561                 spin_lock_irq(host->host_lock);
 562         }
 563
 564         ring_req->act = act;
 565         ring_req->ref_rqid = s->rqid;
 566
 567         shadow->act = act;
 568         shadow->rslt_reset = RSLT_RESET_WAITING;
 569         init_waitqueue_head(&shadow->wq_reset);
 570
 571         ring_req->nr_segments = 0;
 572
 573         scsifront_do_request(info);
 574
 575         spin_unlock_irq(host->host_lock);
 576         err = wait_event_interruptible(shadow->wq_reset, shadow->wait_reset);
 577         spin_lock_irq(host->host_lock);
 578
 579         if (!err) {
 580                 err = shadow->rslt_reset;
 581                 scsifront_put_rqid(info, shadow->rqid);
 582                 kfree(shadow);
 583         } else {
 584                 spin_lock(&info->shadow_lock);
 585                 shadow->rslt_reset = RSLT_RESET_ERR;
 586                 spin_unlock(&info->shadow_lock);
 587                 err = FAILED;
 588         }
 589
 590         spin_unlock_irq(host->host_lock);
 591         return err;
 592 }
 593
 594 static int scsifront_eh_abort_handler(struct scsi_cmnd *sc)
 595 {
 596         pr_debug("%s\n", __func__);
 597         return scsifront_action_handler(sc, VSCSIIF_ACT_SCSI_ABORT);
 598 }
 599
 600 static int scsifront_dev_reset_handler(struct scsi_cmnd *sc)
 601 {
 602         pr_debug("%s\n", __func__);
 603         return scsifront_action_handler(sc, VSCSIIF_ACT_SCSI_RESET);
 604 }
 605
 606 static int scsifront_sdev_configure(struct scsi_device *sdev)
 607 {
 608         struct vscsifrnt_info *info = shost_priv(sdev->host);
 609
 610         if (info && current == info->curr)
 611                 xenbus_printf(XBT_NIL, info->dev->nodename,
 612                               info->dev_state_path, "%d", XenbusStateConnected);
 613
 614         return 0;
 615 }
 616
 617 static void scsifront_sdev_destroy(struct scsi_device *sdev)
 618 {
 619         struct vscsifrnt_info *info = shost_priv(sdev->host);
 620
 621         if (info && current == info->curr)
 622                 xenbus_printf(XBT_NIL, info->dev->nodename,
 623                               info->dev_state_path, "%d", XenbusStateClosed);
 624 }
 625
 626 static struct scsi_host_template scsifront_sht = {
 627         .module                 = THIS_MODULE,
 628         .name                   = "Xen SCSI frontend driver",
 629         .queuecommand           = scsifront_queuecommand,
 630         .eh_abort_handler       = scsifront_eh_abort_handler,
 631         .eh_device_reset_handler = scsifront_dev_reset_handler,
 632         .slave_configure        = scsifront_sdev_configure,
 633         .slave_destroy          = scsifront_sdev_destroy,
 634         .cmd_per_lun            = VSCSIIF_DEFAULT_CMD_PER_LUN,
 635         .can_queue              = VSCSIIF_MAX_REQS,
 636         .this_id                = -1,
 637         .cmd_size               = sizeof(struct vscsifrnt_shadow),
 638         .sg_tablesize           = VSCSIIF_SG_TABLESIZE,
 639         .use_clustering         = DISABLE_CLUSTERING,
 640         .proc_name              = "scsifront",
 641 };
 642
 643 static int scsifront_alloc_ring(struct vscsifrnt_info *info)
 644 {
 645         struct xenbus_device *dev = info->dev;
 646         struct vscsiif_sring *sring;
 647         int err = -ENOMEM;
 648
 649         /***** Frontend to Backend ring start *****/
 650         sring = (struct vscsiif_sring *)__get_free_page(GFP_KERNEL);
 651         if (!sring) {
 652                 xenbus_dev_fatal(dev, err,
 653                         "fail to allocate shared ring (Front to Back)");
 654                 return err;
 655         }
 656         SHARED_RING_INIT(sring);
 657         FRONT_RING_INIT(&info->ring, sring, PAGE_SIZE);
 658
 659         err = xenbus_grant_ring(dev, virt_to_mfn(sring));
 660         if (err < 0) {
 661                 free_page((unsigned long)sring);
 662                 xenbus_dev_fatal(dev, err,
 663                         "fail to grant shared ring (Front to Back)");
 664                 return err;
 665         }
 666         info->ring_ref = err;
 667
 668         err = xenbus_alloc_evtchn(dev, &info->evtchn);
 669         if (err) {
 670                 xenbus_dev_fatal(dev, err, "xenbus_alloc_evtchn");
 671                 goto free_gnttab;
 672         }
 673
 674         err = bind_evtchn_to_irq(info->evtchn);
 675         if (err <= 0) {
 676                 xenbus_dev_fatal(dev, err, "bind_evtchn_to_irq");
 677                 goto free_gnttab;
 678         }
 679
 680         info->irq = err;
 681
 682         err = request_threaded_irq(info->irq, NULL, scsifront_irq_fn,
 683                                    IRQF_ONESHOT, "scsifront", info);
 684         if (err) {
 685                 xenbus_dev_fatal(dev, err, "request_threaded_irq");
 686                 goto free_irq;
 687         }
 688
 689         return 0;
 690
 691 /* free resource */
 692 free_irq:
 693         unbind_from_irqhandler(info->irq, info);
 694 free_gnttab:
 695         gnttab_end_foreign_access(info->ring_ref, 0,
 696                                   (unsigned long)info->ring.sring);
 697
 698         return err;
 699 }
 700
 701 static int scsifront_init_ring(struct vscsifrnt_info *info)
 702 {
 703         struct xenbus_device *dev = info->dev;
 704         struct xenbus_transaction xbt;
 705         int err;
 706
 707         pr_debug("%s\n", __func__);
 708
 709         err = scsifront_alloc_ring(info);
 710         if (err)
 711                 return err;
 712         pr_debug("%s: %u %u\n", __func__, info->ring_ref, info->evtchn);
 713
 714 again:
 715         err = xenbus_transaction_start(&xbt);
 716         if (err)
 717                 xenbus_dev_fatal(dev, err, "starting transaction");
 718
 719         err = xenbus_printf(xbt, dev->nodename, "ring-ref", "%u",
 720                             info->ring_ref);
 721         if (err) {
 722                 xenbus_dev_fatal(dev, err, "%s", "writing ring-ref");
 723                 goto fail;
 724         }
 725
 726         err = xenbus_printf(xbt, dev->nodename, "event-channel", "%u",
 727                             info->evtchn);
 728
 729         if (err) {
 730                 xenbus_dev_fatal(dev, err, "%s", "writing event-channel");
 731                 goto fail;
 732         }
 733
 734         err = xenbus_transaction_end(xbt, 0);
 735         if (err) {
 736                 if (err == -EAGAIN)
 737                         goto again;
 738                 xenbus_dev_fatal(dev, err, "completing transaction");
 739                 goto free_sring;
 740         }
 741
 742         return 0;
 743
 744 fail:
 745         xenbus_transaction_end(xbt, 1);
 746 free_sring:
 747         unbind_from_irqhandler(info->irq, info);
 748         gnttab_end_foreign_access(info->ring_ref, 0,
 749                                   (unsigned long)info->ring.sring);
 750
 751         return err;
 752 }
 753
 754
 755 static int scsifront_probe(struct xenbus_device *dev,
 756                            const struct xenbus_device_id *id)
 757 {
 758         struct vscsifrnt_info *info;
 759         struct Scsi_Host *host;
 760         int err = -ENOMEM;
 761         char name[TASK_COMM_LEN];
 762
 763         host = scsi_host_alloc(&scsifront_sht, sizeof(*info));
 764         if (!host) {
 765                 xenbus_dev_fatal(dev, err, "fail to allocate scsi host");
 766                 return err;
 767         }
 768         info = (struct vscsifrnt_info *)host->hostdata;
 769
 770         dev_set_drvdata(&dev->dev, info);
 771         info->dev = dev;
 772
 773         bitmap_fill(info->shadow_free_bitmap, VSCSIIF_MAX_REQS);
 774
 775         err = scsifront_init_ring(info);
 776         if (err) {
 777                 scsi_host_put(host);
 778                 return err;
 779         }
 780
 781         init_waitqueue_head(&info->wq_sync);
 782         spin_lock_init(&info->shadow_lock);
 783
 784         snprintf(name, TASK_COMM_LEN, "vscsiif.%d", host->host_no);
 785
 786         host->max_id      = VSCSIIF_MAX_TARGET;
 787         host->max_channel = 0;
 788         host->max_lun     = VSCSIIF_MAX_LUN;
 789         host->max_sectors = (host->sg_tablesize - 1) * PAGE_SIZE / 512;
 790         host->max_cmd_len = VSCSIIF_MAX_COMMAND_SIZE;
 791
 792         err = scsi_add_host(host, &dev->dev);
 793         if (err) {
 794                 dev_err(&dev->dev, "fail to add scsi host %d\n", err);
 795                 goto free_sring;
 796         }
 797         info->host = host;
 798         info->host_active = 1;
 799
 800         xenbus_switch_state(dev, XenbusStateInitialised);
 801
 802         return 0;
 803
 804 free_sring:
 805         unbind_from_irqhandler(info->irq, info);
 806         gnttab_end_foreign_access(info->ring_ref, 0,
 807                                   (unsigned long)info->ring.sring);
 808         scsi_host_put(host);
 809         return err;
 810 }
 811
 812 static int scsifront_remove(struct xenbus_device *dev)
 813 {
 814         struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev);
 815
 816         pr_debug("%s: %s removed\n", __func__, dev->nodename);
 817
 818         mutex_lock(&scsifront_mutex);
 819         if (info->host_active) {
 820                 /* Scsi_host not yet removed */
 821                 scsi_remove_host(info->host);
 822                 info->host_active = 0;
 823         }
 824         mutex_unlock(&scsifront_mutex);
 825
 826         gnttab_end_foreign_access(info->ring_ref, 0,
 827                                   (unsigned long)info->ring.sring);
 828         unbind_from_irqhandler(info->irq, info);
 829
 830         scsi_host_put(info->host);
 831
 832         return 0;
 833 }
 834
 835 static void scsifront_disconnect(struct vscsifrnt_info *info)
 836 {
 837         struct xenbus_device *dev = info->dev;
 838         struct Scsi_Host *host = info->host;
 839
 840         pr_debug("%s: %s disconnect\n", __func__, dev->nodename);
 841
 842         /*
 843          * When this function is executed, all devices of
 844          * Frontend have been deleted.
 845          * Therefore, it need not block I/O before remove_host.
 846          */
 847
 848         mutex_lock(&scsifront_mutex);
 849         if (info->host_active) {
 850                 scsi_remove_host(host);
 851                 info->host_active = 0;
 852         }
 853         mutex_unlock(&scsifront_mutex);
 854
 855         xenbus_frontend_closed(dev);
 856 }
 857
 858 static void scsifront_do_lun_hotplug(struct vscsifrnt_info *info, int op)
 859 {
 860         struct xenbus_device *dev = info->dev;
 861         int i, err = 0;
 862         char str[64];
 863         char **dir;
 864         unsigned int dir_n = 0;
 865         unsigned int device_state;
 866         unsigned int hst, chn, tgt, lun;
 867         struct scsi_device *sdev;
 868
 869         dir = xenbus_directory(XBT_NIL, dev->otherend, "vscsi-devs", &dir_n);
 870         if (IS_ERR(dir))
 871                 return;
 872
 873         /* mark current task as the one allowed to modify device states */
 874         BUG_ON(info->curr);
 875         info->curr = current;
 876
 877         for (i = 0; i < dir_n; i++) {
 878                 /* read status */
 879                 snprintf(str, sizeof(str), "vscsi-devs/%s/state", dir[i]);
 880                 err = xenbus_scanf(XBT_NIL, dev->otherend, str, "%u",
 881                                    &device_state);
 882                 if (XENBUS_EXIST_ERR(err))
 883                         continue;
 884
 885                 /* virtual SCSI device */
 886                 snprintf(str, sizeof(str), "vscsi-devs/%s/v-dev", dir[i]);
 887                 err = xenbus_scanf(XBT_NIL, dev->otherend, str,
 888                                    "%u:%u:%u:%u", &hst, &chn, &tgt, &lun);
 889                 if (XENBUS_EXIST_ERR(err))
 890                         continue;
 891
 892                 /*
 893                  * Front device state path, used in slave_configure called
 894                  * on successfull scsi_add_device, and in slave_destroy called
 895                  * on remove of a device.
 896                  */
 897                 snprintf(info->dev_state_path, sizeof(info->dev_state_path),
 898                          "vscsi-devs/%s/state", dir[i]);
 899
 900                 switch (op) {
 901                 case VSCSIFRONT_OP_ADD_LUN:
 902                         if (device_state != XenbusStateInitialised)
 903                                 break;
 904
 905                         if (scsi_add_device(info->host, chn, tgt, lun)) {
 906                                 dev_err(&dev->dev, "scsi_add_device\n");
 907                                 xenbus_printf(XBT_NIL, dev->nodename,
 908                                               info->dev_state_path,
 909                                               "%d", XenbusStateClosed);
 910                         }
 911                         break;
 912                 case VSCSIFRONT_OP_DEL_LUN:
 913                         if (device_state != XenbusStateClosing)
 914                                 break;
 915
 916                         sdev = scsi_device_lookup(info->host, chn, tgt, lun);
 917                         if (sdev) {
 918                                 scsi_remove_device(sdev);
 919                                 scsi_device_put(sdev);
 920                         }
 921                         break;
 922                 default:
 923                         break;
 924                 }
 925         }
 926
 927         info->curr = NULL;
 928
 929         kfree(dir);
 930 }
 931
 932 static void scsifront_read_backend_params(struct xenbus_device *dev,
 933                                           struct vscsifrnt_info *info)
 934 {
 935         unsigned int sg_grant;
 936         int ret;
 937         struct Scsi_Host *host = info->host;
 938
 939         ret = xenbus_scanf(XBT_NIL, dev->otherend, "feature-sg-grant", "%u",
 940                            &sg_grant);
 941         if (ret == 1 && sg_grant) {
 942                 sg_grant = min_t(unsigned int, sg_grant, SG_ALL);
 943                 sg_grant = max_t(unsigned int, sg_grant, VSCSIIF_SG_TABLESIZE);
 944                 host->sg_tablesize = min_t(unsigned int, sg_grant,
 945                         VSCSIIF_SG_TABLESIZE * PAGE_SIZE /
 946                         sizeof(struct scsiif_request_segment));
 947                 host->max_sectors = (host->sg_tablesize - 1) * PAGE_SIZE / 512;
 948         }
 949         dev_info(&dev->dev, "using up to %d SG entries\n", host->sg_tablesize);
 950 }
 951
 952 static void scsifront_backend_changed(struct xenbus_device *dev,
 953                                       enum xenbus_state backend_state)
 954 {
 955         struct vscsifrnt_info *info = dev_get_drvdata(&dev->dev);
 956
 957         pr_debug("%s: %p %u %u\n", __func__, dev, dev->state, backend_state);
 958
 959         switch (backend_state) {
 960         case XenbusStateUnknown:
 961         case XenbusStateInitialising:
 962         case XenbusStateInitWait:
 963         case XenbusStateInitialised:
 964                 break;
 965
 966         case XenbusStateConnected:
 967                 scsifront_read_backend_params(dev, info);
 968                 if (xenbus_read_driver_state(dev->nodename) ==
 969                     XenbusStateInitialised)
 970                         scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_ADD_LUN);
 971
 972                 if (dev->state != XenbusStateConnected)
 973                         xenbus_switch_state(dev, XenbusStateConnected);
 974                 break;
 975
 976         case XenbusStateClosed:
 977                 if (dev->state == XenbusStateClosed)
 978                         break;
 979                 /* Missed the backend's Closing state -- fallthrough */
 980         case XenbusStateClosing:
 981                 scsifront_disconnect(info);
 982                 break;
 983
 984         case XenbusStateReconfiguring:
 985                 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_DEL_LUN);
 986                 xenbus_switch_state(dev, XenbusStateReconfiguring);
 987                 break;
 988
 989         case XenbusStateReconfigured:
 990                 scsifront_do_lun_hotplug(info, VSCSIFRONT_OP_ADD_LUN);
 991                 xenbus_switch_state(dev, XenbusStateConnected);
 992                 break;
 993         }
 994 }
 995
 996 static const struct xenbus_device_id scsifront_ids[] = {
 997         { "vscsi" },
 998         { "" }
 999 };
1000
1001 static struct xenbus_driver scsifront_driver = {
1002         .ids                    = scsifront_ids,
1003         .probe                  = scsifront_probe,
1004         .remove                 = scsifront_remove,
1005         .otherend_changed       = scsifront_backend_changed,
1006 };
1007
1008 static int __init scsifront_init(void)
1009 {
1010         if (!xen_domain())
1011                 return -ENODEV;
1012
1013         return xenbus_register_frontend(&scsifront_driver);
1014 }
1015 module_init(scsifront_init);
1016
1017 static void __exit scsifront_exit(void)
1018 {
1019         xenbus_unregister_driver(&scsifront_driver);
1020 }
1021 module_exit(scsifront_exit);
1022
1023 MODULE_DESCRIPTION("Xen SCSI frontend driver");
1024 MODULE_LICENSE("GPL");
1025 MODULE_ALIAS("xen:vscsi");
1026 MODULE_AUTHOR("Juergen Gross <jgross@suse.com>");