sys/kern/subr_disk.c

   1 /*
   2  * Copyright (c) 2003,2004 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * ----------------------------------------------------------------------------
  35  * "THE BEER-WARE LICENSE" (Revision 42):
  36  * <phk@FreeBSD.ORG> wrote this file.  As long as you retain this notice you
  37  * can do whatever you want with this stuff. If we meet some day, and you think
  38  * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
  39  * ----------------------------------------------------------------------------
  40  *
  41  * Copyright (c) 1982, 1986, 1988, 1993
  42  *      The Regents of the University of California.  All rights reserved.
  43  * (c) UNIX System Laboratories, Inc.
  44  * All or some portions of this file are derived from material licensed
  45  * to the University of California by American Telephone and Telegraph
  46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  47  * the permission of UNIX System Laboratories, Inc.
  48  *
  49  * Redistribution and use in source and binary forms, with or without
  50  * modification, are permitted provided that the following conditions
  51  * are met:
  52  * 1. Redistributions of source code must retain the above copyright
  53  *    notice, this list of conditions and the following disclaimer.
  54  * 2. Redistributions in binary form must reproduce the above copyright
  55  *    notice, this list of conditions and the following disclaimer in the
  56  *    documentation and/or other materials provided with the distribution.
  57  * 3. All advertising materials mentioning features or use of this software
  58  *    must display the following acknowledgement:
  59  *      This product includes software developed by the University of
  60  *      California, Berkeley and its contributors.
  61  * 4. Neither the name of the University nor the names of its contributors
  62  *    may be used to endorse or promote products derived from this software
  63  *    without specific prior written permission.
  64  *
  65  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  66  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  67  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  68  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  69  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  70  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  71  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  72  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  73  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  74  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  75  * SUCH DAMAGE.
  76  *
  77  *      @(#)ufs_disksubr.c      8.5 (Berkeley) 1/21/94
  78  * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
  79  * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
  80  * $DragonFly: src/sys/kern/subr_disk.c,v 1.21 2006/02/17 19:18:06 dillon Exp $
  81  */
  82
  83 #include <sys/param.h>
  84 #include <sys/systm.h>
  85 #include <sys/kernel.h>
  86 #include <sys/proc.h>
  87 #include <sys/sysctl.h>
  88 #include <sys/buf.h>
  89 #include <sys/conf.h>
  90 #include <sys/disklabel.h>
  91 #include <sys/diskslice.h>
  92 #include <sys/disk.h>
  93 #include <sys/malloc.h>
  94 #include <sys/sysctl.h>
  95 #include <machine/md_var.h>
  96 #include <sys/ctype.h>
  97 #include <sys/syslog.h>
  98 #include <sys/device.h>
  99 #include <sys/msgport.h>
 100 #include <sys/msgport2.h>
 101 #include <sys/buf2.h>
 102
 103 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
 104
 105 static d_strategy_t diskstrategy;
 106 static d_open_t diskopen;
 107 static d_close_t diskclose;
 108 static d_ioctl_t diskioctl;
 109 static d_psize_t diskpsize;
 110 static d_clone_t diskclone;
 111 static int disk_putport(lwkt_port_t port, lwkt_msg_t msg);
 112
 113 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
 114
 115 /*
 116  * Create a slice and unit managed disk.
 117  *
 118  * Our port layer will be responsible for assigning blkno and handling
 119  * high level partition operations, then forwarding the requests to the
 120  * raw device.
 121  *
 122  * The disk_create() function clones the provided rawsw for creating a
 123  * managed disk device.  In addition, the cdevsw intercept port is
 124  * changed to disk_putport, which is used to transform requests for the
 125  * managed disk device.
 126  *
 127  * The raw device (based on rawsw) is returned to the caller, NOT the
 128  * slice and unit managed cdev.  The caller typically sets various
 129  * driver parameters and IO limits on the returned rawdev which we must
 130  * inherit when our managed device is opened.
 131  */
 132 dev_t
 133 disk_create(int unit, struct disk *dp, int flags, struct cdevsw *rawsw)
 134 {
 135         dev_t rawdev;
 136         struct cdevsw *devsw;
 137
 138         /*
 139          * Create the raw backing device
 140          */
 141         compile_devsw(rawsw);
 142         rawdev = make_dev(rawsw, dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
 143                             UID_ROOT, GID_OPERATOR, 0640,
 144                             "%s%d", rawsw->d_name, unit);
 145
 146         /*
 147          * Initialize our intercept port
 148          */
 149         bzero(dp, sizeof(*dp));
 150         lwkt_initport(&dp->d_port, NULL);
 151         dp->d_port.mp_putport = disk_putport;
 152         dp->d_rawsw = rawsw;
 153
 154         /*
 155          * We install a custom cdevsw rather then the passed cdevsw,
 156          * and save our disk structure in d_data so we can get at it easily
 157          * without any complex cloning code.
 158          */
 159         devsw = cdevsw_add_override(rawdev, dkunitmask(), dkmakeunit(unit));
 160         devsw->d_port = &dp->d_port;
 161         devsw->d_data = dp;
 162         devsw->d_clone = diskclone;
 163         dp->d_devsw = devsw;
 164         dp->d_rawdev = rawdev;
 165         dp->d_cdev = make_dev(devsw,
 166                             dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
 167                             UID_ROOT, GID_OPERATOR, 0640,
 168                             "%s%d", devsw->d_name, unit);
 169
 170         dp->d_dsflags = flags;
 171         LIST_INSERT_HEAD(&disklist, dp, d_list);
 172         return (dp->d_rawdev);
 173 }
 174
 175 /*
 176  * This routine is called when an adapter detaches.  The higher level
 177  * managed disk device is destroyed while the lower level raw device is
 178  * released.
 179  */
 180 void
 181 disk_destroy(struct disk *disk)
 182 {
 183         if (disk->d_devsw) {
 184             cdevsw_remove(disk->d_devsw, dkunitmask(),
 185                             dkmakeunit(dkunit(disk->d_cdev)));
 186             LIST_REMOVE(disk, d_list);
 187         }
 188         if (disk->d_rawsw) {
 189             destroy_all_dev(disk->d_rawsw, dkunitmask(),
 190                             dkmakeunit(dkunit(disk->d_rawdev)));
 191         }
 192         bzero(disk, sizeof(*disk));
 193 }
 194
 195 int
 196 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize)
 197 {
 198         struct disk *dp;
 199         struct disklabel *dl;
 200         u_int boff;
 201
 202         dp = dev->si_disk;
 203         if (!dp)
 204                 return (ENXIO);
 205         if (!dp->d_slice)
 206                 return (ENXIO);
 207         dl = dsgetlabel(dev, dp->d_slice);
 208         if (!dl)
 209                 return (ENXIO);
 210         *count = Maxmem * (PAGE_SIZE / dl->d_secsize);
 211         if (dumplo <= LABELSECTOR ||
 212             (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size))
 213                 return (EINVAL);
 214         boff = dl->d_partitions[dkpart(dev)].p_offset +
 215             dp->d_slice->dss_slices[dkslice(dev)].ds_offset;
 216         *blkno = boff + dumplo;
 217         *secsize = dl->d_secsize;
 218         return (0);
 219
 220 }
 221
 222 void
 223 disk_invalidate (struct disk *disk)
 224 {
 225         if (disk->d_slice)
 226                 dsgone(&disk->d_slice);
 227 }
 228
 229 struct disk *
 230 disk_enumerate(struct disk *disk)
 231 {
 232         if (!disk)
 233                 return (LIST_FIRST(&disklist));
 234         else
 235                 return (LIST_NEXT(disk, d_list));
 236 }
 237
 238 static
 239 int
 240 sysctl_disks(SYSCTL_HANDLER_ARGS)
 241 {
 242         struct disk *disk;
 243         int error, first;
 244
 245         disk = NULL;
 246         first = 1;
 247
 248         while ((disk = disk_enumerate(disk))) {
 249                 if (!first) {
 250                         error = SYSCTL_OUT(req, " ", 1);
 251                         if (error)
 252                                 return error;
 253                 } else {
 254                         first = 0;
 255                 }
 256                 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
 257                                    strlen(disk->d_rawdev->si_name));
 258                 if (error)
 259                         return error;
 260         }
 261         error = SYSCTL_OUT(req, "", 1);
 262         return error;
 263 }
 264
 265 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL,
 266     sysctl_disks, "A", "names of available disks");
 267
 268 /*
 269  * The port intercept functions
 270  */
 271 static
 272 int
 273 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg)
 274 {
 275         struct disk *disk = (struct disk *)port;
 276         cdevallmsg_t msg = (cdevallmsg_t)lmsg;
 277         int error;
 278
 279         switch(msg->am_lmsg.ms_cmd.cm_op) {
 280         case CDEV_CMD_OPEN:
 281                 error = diskopen(
 282                             msg->am_open.msg.dev,
 283                             msg->am_open.oflags,
 284                             msg->am_open.devtype,
 285                             msg->am_open.td);
 286                 break;
 287         case CDEV_CMD_CLOSE:
 288                 error = diskclose(
 289                             msg->am_close.msg.dev,
 290                             msg->am_close.fflag,
 291                             msg->am_close.devtype,
 292                             msg->am_close.td);
 293                 break;
 294         case CDEV_CMD_IOCTL:
 295                 error = diskioctl(
 296                             msg->am_ioctl.msg.dev,
 297                             msg->am_ioctl.cmd,
 298                             msg->am_ioctl.data,
 299                             msg->am_ioctl.fflag,
 300                             msg->am_ioctl.td);
 301                 break;
 302         case CDEV_CMD_STRATEGY:
 303                 diskstrategy(msg->am_strategy.msg.dev, msg->am_strategy.bio);
 304                 error = 0;
 305                 break;
 306         case CDEV_CMD_PSIZE:
 307                 msg->am_psize.result = diskpsize(msg->am_psize.msg.dev);
 308                 error = 0;      /* XXX */
 309                 break;
 310         case CDEV_CMD_READ:
 311                 error = physio(msg->am_read.msg.dev,
 312                                 msg->am_read.uio, msg->am_read.ioflag);
 313                 break;
 314         case CDEV_CMD_WRITE:
 315                 error = physio(msg->am_write.msg.dev,
 316                                 msg->am_write.uio, msg->am_write.ioflag);
 317                 break;
 318         case CDEV_CMD_POLL:
 319         case CDEV_CMD_KQFILTER:
 320                 error = ENODEV;
 321         case CDEV_CMD_MMAP:
 322                 error = -1;
 323                 break;
 324         case CDEV_CMD_DUMP:
 325                 error = disk_dumpcheck(msg->am_dump.msg.dev,
 326                                 &msg->am_dump.count,
 327                                 &msg->am_dump.blkno,
 328                                 &msg->am_dump.secsize);
 329                 if (error == 0) {
 330                         msg->am_dump.msg.dev = disk->d_rawdev;
 331                         error = lwkt_forwardmsg(disk->d_rawdev->si_port,
 332                                                 &msg->am_dump.msg.msg);
 333                         printf("error2 %d\n", error);
 334                 }
 335                 break;
 336         default:
 337                 error = ENOTSUP;
 338                 break;
 339         }
 340         return(error);
 341 }
 342
 343 /*
 344  * When new device entries are instantiated, make sure they inherit our
 345  * si_disk structure and block and iosize limits from the raw device.
 346  *
 347  * This routine is always called synchronously in the context of the
 348  * client.
 349  *
 350  * XXX The various io and block size constraints are not always initialized
 351  * properly by devices.
 352  */
 353 static
 354 int
 355 diskclone(dev_t dev)
 356 {
 357         struct disk *dp;
 358
 359         dp = dev->si_devsw->d_data;
 360         KKASSERT(dp != NULL);
 361         dev->si_disk = dp;
 362         dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
 363         dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
 364         dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
 365         return(0);
 366 }
 367
 368 /*
 369  * Open a disk device or partition.
 370  */
 371 static
 372 int
 373 diskopen(dev_t dev, int oflags, int devtype, struct thread *td)
 374 {
 375         struct disk *dp;
 376         int error;
 377
 378         /*
 379          * dp can't be NULL here XXX.
 380          */
 381         error = 0;
 382         dp = dev->si_disk;
 383         if (dp == NULL)
 384                 return (ENXIO);
 385
 386         /*
 387          * Deal with open races
 388          */
 389         while (dp->d_flags & DISKFLAG_LOCK) {
 390                 dp->d_flags |= DISKFLAG_WANTED;
 391                 error = tsleep(dp, PCATCH, "diskopen", hz);
 392                 if (error)
 393                         return (error);
 394         }
 395         dp->d_flags |= DISKFLAG_LOCK;
 396
 397         /*
 398          * Open the underlying raw device.
 399          */
 400         if (!dsisopen(dp->d_slice)) {
 401 #if 0
 402                 if (!pdev->si_iosize_max)
 403                         pdev->si_iosize_max = dev->si_iosize_max;
 404 #endif
 405                 error = dev_dopen(dp->d_rawdev, oflags, devtype, td);
 406         }
 407
 408         /*
 409          * Inherit properties from the underlying device now that it is
 410          * open.
 411          */
 412         diskclone(dev);
 413
 414         if (error)
 415                 goto out;
 416
 417         error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label);
 418
 419         if (!dsisopen(dp->d_slice))
 420                 dev_dclose(dp->d_rawdev, oflags, devtype, td);
 421 out:
 422         dp->d_flags &= ~DISKFLAG_LOCK;
 423         if (dp->d_flags & DISKFLAG_WANTED) {
 424                 dp->d_flags &= ~DISKFLAG_WANTED;
 425                 wakeup(dp);
 426         }
 427
 428         return(error);
 429 }
 430
 431 /*
 432  * Close a disk device or partition
 433  */
 434 static
 435 int
 436 diskclose(dev_t dev, int fflag, int devtype, struct thread *td)
 437 {
 438         struct disk *dp;
 439         int error;
 440
 441         error = 0;
 442         dp = dev->si_disk;
 443
 444         dsclose(dev, devtype, dp->d_slice);
 445         if (!dsisopen(dp->d_slice))
 446                 error = dev_dclose(dp->d_rawdev, fflag, devtype, td);
 447         return (error);
 448 }
 449
 450 /*
 451  * Execute strategy routine
 452  */
 453 static
 454 void
 455 diskstrategy(dev_t dev, struct bio *bio)
 456 {
 457         struct bio *nbio;
 458         struct disk *dp;
 459
 460         dp = dev->si_disk;
 461
 462         if (dp == NULL) {
 463                 bio->bio_buf->b_error = ENXIO;
 464                 bio->bio_buf->b_flags |= B_ERROR;
 465                 biodone(bio);
 466                 return;
 467         }
 468         KKASSERT(dev->si_disk == dp);
 469
 470         /*
 471          * The dscheck() function will also transform the slice relative
 472          * block number i.e. bio->bio_blkno into a block number that can be
 473          * passed directly to the underlying raw device.
 474          */
 475         nbio = dscheck(dev, bio, dp->d_slice);
 476         if (nbio == NULL) {
 477                 biodone(bio);
 478                 return;
 479         }
 480         dev_dstrategy(dp->d_rawdev, nbio);
 481 }
 482
 483 /*
 484  * First execute the ioctl on the disk device, and if it isn't supported
 485  * try running it on the backing device.
 486  */
 487 static
 488 int
 489 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
 490 {
 491         struct disk *dp;
 492         int error;
 493
 494         dp = dev->si_disk;
 495         if (dp == NULL)
 496                 return (ENXIO);
 497
 498         error = dsioctl(dev, cmd, data, fflag, &dp->d_slice);
 499         if (error == ENOIOCTL)
 500                 error = dev_dioctl(dp->d_rawdev, cmd, data, fflag, td);
 501         return (error);
 502 }
 503
 504 /*
 505  *
 506  */
 507 static
 508 int
 509 diskpsize(dev_t dev)
 510 {
 511         struct disk *dp;
 512
 513         dp = dev->si_disk;
 514         if (dp == NULL)
 515                 return (-1);
 516         return(dssize(dev, &dp->d_slice));
 517 #if 0
 518         if (dp != dev->si_disk) {
 519                 dev->si_drv1 = pdev->si_drv1;
 520                 dev->si_drv2 = pdev->si_drv2;
 521                 /* XXX: don't set bp->b_dev->si_disk (?) */
 522         }
 523 #endif
 524 }
 525
 526 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD,
 527     0, sizeof(struct disklabel), "sizeof(struct disklabel)");
 528
 529 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
 530     0, sizeof(struct diskslices), "sizeof(struct diskslices)");
 531
 532 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
 533     0, sizeof(struct disk), "sizeof(struct disk)");
 534
 535
 536 /*
 537  * Seek sort for disks.
 538  *
 539  * The bio_queue keep two queues, sorted in ascending block order.  The first
 540  * queue holds those requests which are positioned after the current block
 541  * (in the first request); the second, which starts at queue->switch_point,
 542  * holds requests which came in after their block number was passed.  Thus
 543  * we implement a one way scan, retracting after reaching the end of the drive
 544  * to the first request on the second queue, at which time it becomes the
 545  * first queue.
 546  *
 547  * A one-way scan is natural because of the way UNIX read-ahead blocks are
 548  * allocated.
 549  */
 550 void
 551 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
 552 {
 553         struct bio *bq;
 554         struct bio *bn;
 555         struct bio *be;
 556
 557         be = TAILQ_LAST(&bioq->queue, bio_queue);
 558         /*
 559          * If the queue is empty or we are an
 560          * ordered transaction, then it's easy.
 561          */
 562         if ((bq = bioq_first(bioq)) == NULL ||
 563             (bio->bio_buf->b_flags & B_ORDERED) != 0) {
 564                 bioq_insert_tail(bioq, bio);
 565                 return;
 566         } else if (bioq->insert_point != NULL) {
 567
 568                 /*
 569                  * A certain portion of the list is
 570                  * "locked" to preserve ordering, so
 571                  * we can only insert after the insert
 572                  * point.
 573                  */
 574                 bq = bioq->insert_point;
 575         } else {
 576
 577                 /*
 578                  * If we lie before the last removed (currently active)
 579                  * request, and are not inserting ourselves into the
 580                  * "locked" portion of the list, then we must add ourselves
 581                  * to the second request list.
 582                  */
 583                 if (bio->bio_blkno < bioq->last_blkno) {
 584
 585                         bq = bioq->switch_point;
 586                         /*
 587                          * If we are starting a new secondary list,
 588                          * then it's easy.
 589                          */
 590                         if (bq == NULL) {
 591                                 bioq->switch_point = bio;
 592                                 bioq_insert_tail(bioq, bio);
 593                                 return;
 594                         }
 595                         /*
 596                          * If we lie ahead of the current switch point,
 597                          * insert us before the switch point and move
 598                          * the switch point.
 599                          */
 600                         if (bio->bio_blkno < bq->bio_blkno) {
 601                                 bioq->switch_point = bio;
 602                                 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
 603                                 return;
 604                         }
 605                 } else {
 606                         if (bioq->switch_point != NULL)
 607                                 be = TAILQ_PREV(bioq->switch_point,
 608                                                 bio_queue, bio_act);
 609                         /*
 610                          * If we lie between last_blkno and bq,
 611                          * insert before bq.
 612                          */
 613                         if (bio->bio_blkno < bq->bio_blkno) {
 614                                 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
 615                                 return;
 616                         }
 617                 }
 618         }
 619
 620         /*
 621          * Request is at/after our current position in the list.
 622          * Optimize for sequential I/O by seeing if we go at the tail.
 623          */
 624         if (bio->bio_blkno > be->bio_blkno) {
 625                 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
 626                 return;
 627         }
 628
 629         /* Otherwise, insertion sort */
 630         while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
 631
 632                 /*
 633                  * We want to go after the current request if it is the end
 634                  * of the first request list, or if the next request is a
 635                  * larger cylinder than our request.
 636                  */
 637                 if (bn == bioq->switch_point
 638                  || bio->bio_blkno < bn->bio_blkno)
 639                         break;
 640                 bq = bn;
 641         }
 642         TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
 643 }
 644
 645
 646 /*
 647  * Attempt to read a disk label from a device using the indicated strategy
 648  * routine.  The label must be partly set up before this: secpercyl, secsize
 649  * and anything required in the strategy routine (e.g., dummy bounds for the
 650  * partition containing the label) must be filled in before calling us.
 651  * Returns NULL on success and an error string on failure.
 652  */
 653 char *
 654 readdisklabel(dev_t dev, struct disklabel *lp)
 655 {
 656         struct buf *bp;
 657         struct disklabel *dlp;
 658         char *msg = NULL;
 659
 660         bp = geteblk((int)lp->d_secsize);
 661         bp->b_bio1.bio_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
 662         bp->b_bcount = lp->d_secsize;
 663         bp->b_flags &= ~B_INVAL;
 664         bp->b_flags |= B_READ;
 665         dev_dstrategy(dev, &bp->b_bio1);
 666         if (biowait(bp))
 667                 msg = "I/O error";
 668         else for (dlp = (struct disklabel *)bp->b_data;
 669             dlp <= (struct disklabel *)((char *)bp->b_data +
 670             lp->d_secsize - sizeof(*dlp));
 671             dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
 672                 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
 673                         if (msg == NULL)
 674                                 msg = "no disk label";
 675                 } else if (dlp->d_npartitions > MAXPARTITIONS ||
 676                            dkcksum(dlp) != 0)
 677                         msg = "disk label corrupted";
 678                 else {
 679                         *lp = *dlp;
 680                         msg = NULL;
 681                         break;
 682                 }
 683         }
 684         bp->b_flags |= B_INVAL | B_AGE;
 685         brelse(bp);
 686         return (msg);
 687 }
 688
 689 /*
 690  * Check new disk label for sensibility before setting it.
 691  */
 692 int
 693 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask)
 694 {
 695         int i;
 696         struct partition *opp, *npp;
 697
 698         /*
 699          * Check it is actually a disklabel we are looking at.
 700          */
 701         if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
 702             dkcksum(nlp) != 0)
 703                 return (EINVAL);
 704         /*
 705          * For each partition that we think is open,
 706          */
 707         while ((i = ffs((long)openmask)) != 0) {
 708                 i--;
 709                 /*
 710                  * Check it is not changing....
 711                  */
 712                 openmask &= ~(1 << i);
 713                 if (nlp->d_npartitions <= i)
 714                         return (EBUSY);
 715                 opp = &olp->d_partitions[i];
 716                 npp = &nlp->d_partitions[i];
 717                 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
 718                         return (EBUSY);
 719                 /*
 720                  * Copy internally-set partition information
 721                  * if new label doesn't include it.             XXX
 722                  * (If we are using it then we had better stay the same type)
 723                  * This is possibly dubious, as someone else noted (XXX)
 724                  */
 725                 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
 726                         npp->p_fstype = opp->p_fstype;
 727                         npp->p_fsize = opp->p_fsize;
 728                         npp->p_frag = opp->p_frag;
 729                         npp->p_cpg = opp->p_cpg;
 730                 }
 731         }
 732         nlp->d_checksum = 0;
 733         nlp->d_checksum = dkcksum(nlp);
 734         *olp = *nlp;
 735         return (0);
 736 }
 737
 738 /*
 739  * Write disk label back to device after modification.
 740  */
 741 int
 742 writedisklabel(dev_t dev, struct disklabel *lp)
 743 {
 744         struct buf *bp;
 745         struct disklabel *dlp;
 746         int error = 0;
 747
 748         if (lp->d_partitions[RAW_PART].p_offset != 0)
 749                 return (EXDEV);                 /* not quite right */
 750         bp = geteblk((int)lp->d_secsize);
 751         bp->b_bio1.bio_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
 752         bp->b_bcount = lp->d_secsize;
 753 #if 1
 754         /*
 755          * We read the label first to see if it's there,
 756          * in which case we will put ours at the same offset into the block..
 757          * (I think this is stupid [Julian])
 758          * Note that you can't write a label out over a corrupted label!
 759          * (also stupid.. how do you write the first one? by raw writes?)
 760          */
 761         bp->b_flags &= ~B_INVAL;
 762         bp->b_flags |= B_READ;
 763         dev_dstrategy(dkmodpart(dev, RAW_PART), &bp->b_bio1);
 764         error = biowait(bp);
 765         if (error)
 766                 goto done;
 767         for (dlp = (struct disklabel *)bp->b_data;
 768             dlp <= (struct disklabel *)
 769               ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
 770             dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
 771                 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
 772                     dkcksum(dlp) == 0) {
 773                         *dlp = *lp;
 774                         bp->b_flags &= ~(B_DONE | B_READ);
 775                         bp->b_flags |= B_WRITE;
 776                         dev_dstrategy(dkmodpart(dev, RAW_PART), &bp->b_bio1);
 777                         error = biowait(bp);
 778                         goto done;
 779                 }
 780         }
 781         error = ESRCH;
 782 done:
 783 #else
 784         bzero(bp->b_data, lp->d_secsize);
 785         dlp = (struct disklabel *)bp->b_data;
 786         *dlp = *lp;
 787         bp->b_flags &= ~B_INVAL;
 788         bp->b_flags |= B_WRITE;
 789         BUF_STRATEGY(bp, 1);
 790         error = biowait(bp);
 791 #endif
 792         bp->b_flags |= B_INVAL | B_AGE;
 793         brelse(bp);
 794         return (error);
 795 }
 796
 797 /*
 798  * Disk error is the preface to plaintive error messages
 799  * about failing disk transfers.  It prints messages of the form
 800
 801 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
 802
 803  * if the offset of the error in the transfer and a disk label
 804  * are both available.  blkdone should be -1 if the position of the error
 805  * is unknown; the disklabel pointer may be null from drivers that have not
 806  * been converted to use them.  The message is printed with printf
 807  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
 808  * The message should be completed (with at least a newline) with printf
 809  * or addlog, respectively.  There is no trailing space.
 810  */
 811 void
 812 diskerr(struct bio *bio, dev_t dev, const char *what, int pri,
 813         int blkdone, struct disklabel *lp)
 814 {
 815         struct buf *bp = bio->bio_buf;
 816         int unit = dkunit(dev);
 817         int slice = dkslice(dev);
 818         int part = dkpart(dev);
 819         char partname[2];
 820         char *sname;
 821         daddr_t sn;
 822
 823         sname = dsname(dev, unit, slice, part, partname);
 824         printf("%s%s: %s %sing fsbn ", sname, partname, what,
 825               bp->b_flags & B_READ ? "read" : "writ");
 826         sn = bio->bio_blkno;
 827         if (bp->b_bcount <= DEV_BSIZE) {
 828                 printf("%ld", (long)sn);
 829         } else {
 830                 if (blkdone >= 0) {
 831                         sn += blkdone;
 832                         printf("%ld of ", (long)sn);
 833                 }
 834                 printf("%ld-%ld", (long)bio->bio_blkno,
 835                     (long)(bio->bio_blkno + (bp->b_bcount - 1) / DEV_BSIZE));
 836         }
 837         if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
 838                 sn += lp->d_partitions[part].p_offset;
 839                 /*
 840                  * XXX should add slice offset and not print the slice,
 841                  * but we don't know the slice pointer.
 842                  * XXX should print bio->bio_blkno so that this will work
 843                  * independent of slices, labels and bad sector remapping,
 844                  * but some drivers don't set bio->bio_blkno.
 845                  */
 846                 printf(" (%s bn %ld; cn %ld", sname, (long)sn,
 847                     (long)(sn / lp->d_secpercyl));
 848                 sn %= (long)lp->d_secpercyl;
 849                 printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors),
 850                     (long)(sn % lp->d_nsectors));
 851         }
 852 }
 853