The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/geom/geom_dev.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 2002 Poul-Henning Kamp
    3  * Copyright (c) 2002 Networks Associates Technology, Inc.
    4  * All rights reserved.
    5  *
    6  * This software was developed for the FreeBSD Project by Poul-Henning Kamp
    7  * and NAI Labs, the Security Research Division of Network Associates, Inc.
    8  * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
    9  * DARPA CHATS research program.
   10  *
   11  * Redistribution and use in source and binary forms, with or without
   12  * modification, are permitted provided that the following conditions
   13  * are met:
   14  * 1. Redistributions of source code must retain the above copyright
   15  *    notice, this list of conditions and the following disclaimer.
   16  * 2. Redistributions in binary form must reproduce the above copyright
   17  *    notice, this list of conditions and the following disclaimer in the
   18  *    documentation and/or other materials provided with the distribution.
   19  * 3. The names of the authors may not be used to endorse or promote
   20  *    products derived from this software without specific prior written
   21  *    permission.
   22  *
   23  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   33  * SUCH DAMAGE.
   34  */
   35 
   36 #include <sys/cdefs.h>
   37 __FBSDID("$FreeBSD$");
   38 
   39 #include <sys/param.h>
   40 #include <sys/systm.h>
   41 #include <sys/malloc.h>
   42 #include <sys/kernel.h>
   43 #include <sys/conf.h>
   44 #include <sys/bio.h>
   45 #include <sys/lock.h>
   46 #include <sys/mutex.h>
   47 #include <sys/proc.h>
   48 #include <sys/errno.h>
   49 #include <sys/time.h>
   50 #include <sys/disk.h>
   51 #include <sys/fcntl.h>
   52 #include <sys/limits.h>
   53 #include <sys/sysctl.h>
   54 #include <geom/geom.h>
   55 #include <geom/geom_int.h>
   56 
   57 static d_open_t         g_dev_open;
   58 static d_close_t        g_dev_close;
   59 static d_strategy_t     g_dev_strategy;
   60 static d_ioctl_t        g_dev_ioctl;
   61 
   62 static struct cdevsw g_dev_cdevsw = {
   63         .d_version =    D_VERSION,
   64         .d_open =       g_dev_open,
   65         .d_close =      g_dev_close,
   66         .d_read =       physread,
   67         .d_write =      physwrite,
   68         .d_ioctl =      g_dev_ioctl,
   69         .d_strategy =   g_dev_strategy,
   70         .d_name =       "g_dev",
   71         .d_flags =      D_DISK | D_TRACKCLOSE,
   72 };
   73 
   74 static g_taste_t g_dev_taste;
   75 static g_orphan_t g_dev_orphan;
   76 
   77 static struct g_class g_dev_class       = {
   78         .name = "DEV",
   79         .version = G_VERSION,
   80         .taste = g_dev_taste,
   81         .orphan = g_dev_orphan,
   82 };
   83 
   84 /*
   85  * We target 262144 (8 x 32768) sectors by default as this significantly
   86  * increases the throughput on commonly used SSD's with a marginal
   87  * increase in non-interruptible request latency.
   88  */
   89 static uint64_t g_dev_del_max_sectors = 262144;
   90 SYSCTL_DECL(_kern_geom);
   91 SYSCTL_NODE(_kern_geom, OID_AUTO, dev, CTLFLAG_RW, 0, "GEOM_DEV stuff");
   92 SYSCTL_QUAD(_kern_geom_dev, OID_AUTO, delete_max_sectors, CTLFLAG_RW,
   93     &g_dev_del_max_sectors, 0, "Maximum number of sectors in a single "
   94     "delete request sent to the provider. Larger requests are chunked "
   95     "so they can be interrupted. (0 = disable chunking)");
   96 
   97 void
   98 g_dev_print(void)
   99 {
  100         struct g_geom *gp;
  101         char const *p = "";
  102 
  103         LIST_FOREACH(gp, &g_dev_class.geom, geom) {
  104                 printf("%s%s", p, gp->name);
  105                 p = " ";
  106         }
  107         printf("\n");
  108 }
  109 
  110 struct g_provider *
  111 g_dev_getprovider(struct cdev *dev)
  112 {
  113         struct g_consumer *cp;
  114 
  115         g_topology_assert();
  116         if (dev == NULL)
  117                 return (NULL);
  118         if (dev->si_devsw != &g_dev_cdevsw)
  119                 return (NULL);
  120         cp = dev->si_drv2;
  121         return (cp->provider);
  122 }
  123 
  124 
  125 static struct g_geom *
  126 g_dev_taste(struct g_class *mp, struct g_provider *pp, int insist __unused)
  127 {
  128         struct g_geom *gp;
  129         struct g_consumer *cp;
  130         int error, len;
  131         struct cdev *dev, *adev;
  132         char buf[64], *val;
  133 
  134         g_trace(G_T_TOPOLOGY, "dev_taste(%s,%s)", mp->name, pp->name);
  135         g_topology_assert();
  136         LIST_FOREACH(cp, &pp->consumers, consumers)
  137                 if (cp->geom->class == mp)
  138                         return (NULL);
  139         gp = g_new_geomf(mp, pp->name);
  140         cp = g_new_consumer(gp);
  141         error = g_attach(cp, pp);
  142         KASSERT(error == 0,
  143             ("g_dev_taste(%s) failed to g_attach, err=%d", pp->name, error));
  144         dev = make_dev(&g_dev_cdevsw, 0,
  145             UID_ROOT, GID_OPERATOR, 0640, "%s", gp->name);
  146 
  147         /* Search for device alias name and create it if found. */
  148         adev = NULL;
  149         for (len = MIN(strlen(gp->name), sizeof(buf) - 15); len > 0; len--) {
  150                 snprintf(buf, sizeof(buf), "kern.devalias.%s", gp->name);
  151                 buf[14 + len] = 0;
  152                 val = getenv(buf);
  153                 if (val != NULL) {
  154                         snprintf(buf, sizeof(buf), "%s%s",
  155                             val, gp->name + len);
  156                         freeenv(val);
  157                         adev = make_dev_alias(dev, buf);
  158                         break;
  159                 }
  160         }
  161 
  162         if (pp->flags & G_PF_CANDELETE)
  163                 dev->si_flags |= SI_CANDELETE;
  164         dev->si_iosize_max = MAXPHYS;
  165         gp->softc = dev;
  166         dev->si_drv1 = gp;
  167         dev->si_drv2 = cp;
  168         if (adev != NULL) {
  169                 if (pp->flags & G_PF_CANDELETE)
  170                         adev->si_flags |= SI_CANDELETE;
  171                 adev->si_iosize_max = MAXPHYS;
  172                 adev->si_drv1 = gp;
  173                 adev->si_drv2 = cp;
  174         }
  175         return (gp);
  176 }
  177 
  178 static int
  179 g_dev_open(struct cdev *dev, int flags, int fmt, struct thread *td)
  180 {
  181         struct g_geom *gp;
  182         struct g_consumer *cp;
  183         int error, r, w, e;
  184 
  185         gp = dev->si_drv1;
  186         cp = dev->si_drv2;
  187         if (gp == NULL || cp == NULL || gp->softc != dev)
  188                 return(ENXIO);          /* g_dev_taste() not done yet */
  189 
  190         g_trace(G_T_ACCESS, "g_dev_open(%s, %d, %d, %p)",
  191             gp->name, flags, fmt, td);
  192 
  193         r = flags & FREAD ? 1 : 0;
  194         w = flags & FWRITE ? 1 : 0;
  195 #ifdef notyet
  196         e = flags & O_EXCL ? 1 : 0;
  197 #else
  198         e = 0;
  199 #endif
  200         if (w) {
  201                 /*
  202                  * When running in very secure mode, do not allow
  203                  * opens for writing of any disks.
  204                  */
  205                 error = securelevel_ge(td->td_ucred, 2);
  206                 if (error)
  207                         return (error);
  208         }
  209         g_topology_lock();
  210         if (dev->si_devsw == NULL)
  211                 error = ENXIO;          /* We were orphaned */
  212         else
  213                 error = g_access(cp, r, w, e);
  214         g_topology_unlock();
  215         return(error);
  216 }
  217 
  218 static int
  219 g_dev_close(struct cdev *dev, int flags, int fmt, struct thread *td)
  220 {
  221         struct g_geom *gp;
  222         struct g_consumer *cp;
  223         int error, r, w, e, i;
  224 
  225         gp = dev->si_drv1;
  226         cp = dev->si_drv2;
  227         if (gp == NULL || cp == NULL)
  228                 return(ENXIO);
  229         g_trace(G_T_ACCESS, "g_dev_close(%s, %d, %d, %p)",
  230             gp->name, flags, fmt, td);
  231         r = flags & FREAD ? -1 : 0;
  232         w = flags & FWRITE ? -1 : 0;
  233 #ifdef notyet
  234         e = flags & O_EXCL ? -1 : 0;
  235 #else
  236         e = 0;
  237 #endif
  238         g_topology_lock();
  239         if (dev->si_devsw == NULL)
  240                 error = ENXIO;          /* We were orphaned */
  241         else
  242                 error = g_access(cp, r, w, e);
  243         for (i = 0; i < 10 * hz;) {
  244                 if (cp->acr != 0 || cp->acw != 0)
  245                         break;
  246                 if (cp->nstart == cp->nend)
  247                         break;
  248                 pause("gdevwclose", hz / 10);
  249                 i += hz / 10;
  250         }
  251         if (cp->acr == 0 && cp->acw == 0 && cp->nstart != cp->nend) {
  252                 printf("WARNING: Final close of geom_dev(%s) %s %s\n",
  253                     gp->name,
  254                     "still has outstanding I/O after 10 seconds.",
  255                     "Completing close anyway, panic may happen later.");
  256         }
  257         g_topology_unlock();
  258         return (error);
  259 }
  260 
  261 /*
  262  * XXX: Until we have unmessed the ioctl situation, there is a race against
  263  * XXX: a concurrent orphanization.  We cannot close it by holding topology
  264  * XXX: since that would prevent us from doing our job, and stalling events
  265  * XXX: will break (actually: stall) the BSD disklabel hacks.
  266  */
  267 static int
  268 g_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
  269 {
  270         struct g_geom *gp;
  271         struct g_consumer *cp;
  272         struct g_provider *pp;
  273         struct g_kerneldump kd;
  274         off_t offset, length, chunk;
  275         int i, error;
  276         u_int u;
  277 
  278         gp = dev->si_drv1;
  279         cp = dev->si_drv2;
  280         pp = cp->provider;
  281 
  282         error = 0;
  283         KASSERT(cp->acr || cp->acw,
  284             ("Consumer with zero access count in g_dev_ioctl"));
  285 
  286         i = IOCPARM_LEN(cmd);
  287         switch (cmd) {
  288         case DIOCGSECTORSIZE:
  289                 *(u_int *)data = cp->provider->sectorsize;
  290                 if (*(u_int *)data == 0)
  291                         error = ENOENT;
  292                 break;
  293         case DIOCGMEDIASIZE:
  294                 *(off_t *)data = cp->provider->mediasize;
  295                 if (*(off_t *)data == 0)
  296                         error = ENOENT;
  297                 break;
  298         case DIOCGFWSECTORS:
  299                 error = g_io_getattr("GEOM::fwsectors", cp, &i, data);
  300                 if (error == 0 && *(u_int *)data == 0)
  301                         error = ENOENT;
  302                 break;
  303         case DIOCGFWHEADS:
  304                 error = g_io_getattr("GEOM::fwheads", cp, &i, data);
  305                 if (error == 0 && *(u_int *)data == 0)
  306                         error = ENOENT;
  307                 break;
  308         case DIOCGFRONTSTUFF:
  309                 error = g_io_getattr("GEOM::frontstuff", cp, &i, data);
  310                 break;
  311         case DIOCSKERNELDUMP:
  312                 u = *((u_int *)data);
  313                 if (!u) {
  314                         set_dumper(NULL);
  315                         error = 0;
  316                         break;
  317                 }
  318                 kd.offset = 0;
  319                 kd.length = OFF_MAX;
  320                 i = sizeof kd;
  321                 error = g_io_getattr("GEOM::kerneldump", cp, &i, &kd);
  322                 if (!error) {
  323                         error = set_dumper(&kd.di);
  324                         if (!error)
  325                                 dev->si_flags |= SI_DUMPDEV;
  326                 }
  327                 break;
  328         case DIOCGFLUSH:
  329                 error = g_io_flush(cp);
  330                 break;
  331         case DIOCGDELETE:
  332                 offset = ((off_t *)data)[0];
  333                 length = ((off_t *)data)[1];
  334                 if ((offset % cp->provider->sectorsize) != 0 ||
  335                     (length % cp->provider->sectorsize) != 0 || length <= 0) {
  336                         printf("%s: offset=%jd length=%jd\n", __func__, offset,
  337                             length);
  338                         error = EINVAL;
  339                         break;
  340                 }
  341                 while (length > 0) { 
  342                         chunk = length;
  343                         if (g_dev_del_max_sectors != 0 && chunk >
  344                             g_dev_del_max_sectors * cp->provider->sectorsize) {
  345                                 chunk = g_dev_del_max_sectors *
  346                                     cp->provider->sectorsize;
  347                         }
  348                         error = g_delete_data(cp, offset, chunk);
  349                         length -= chunk;
  350                         offset += chunk;
  351                         if (error)
  352                                 break;
  353                         /*
  354                          * Since the request size can be large, the service
  355                          * time can be is likewise.  We make this ioctl
  356                          * interruptible by checking for signals for each bio.
  357                          */
  358                         if (SIGPENDING(td))
  359                                 break;
  360                 }
  361                 break;
  362         case DIOCGIDENT:
  363                 error = g_io_getattr("GEOM::ident", cp, &i, data);
  364                 break;
  365         case DIOCGPROVIDERNAME:
  366                 if (pp == NULL)
  367                         return (ENOENT);
  368                 strlcpy(data, pp->name, i);
  369                 break;
  370         case DIOCGSTRIPESIZE:
  371                 *(off_t *)data = cp->provider->stripesize;
  372                 break;
  373         case DIOCGSTRIPEOFFSET:
  374                 *(off_t *)data = cp->provider->stripeoffset;
  375                 break;
  376         default:
  377                 if (cp->provider->geom->ioctl != NULL) {
  378                         error = cp->provider->geom->ioctl(cp->provider, cmd, data, fflag, td);
  379                 } else {
  380                         error = ENOIOCTL;
  381                 }
  382         }
  383 
  384         return (error);
  385 }
  386 
  387 static void
  388 g_dev_done(struct bio *bp2)
  389 {
  390         struct bio *bp;
  391 
  392         bp = bp2->bio_parent;
  393         bp->bio_error = bp2->bio_error;
  394         if (bp->bio_error != 0) {
  395                 g_trace(G_T_BIO, "g_dev_done(%p) had error %d",
  396                     bp2, bp->bio_error);
  397                 bp->bio_flags |= BIO_ERROR;
  398         } else {
  399                 g_trace(G_T_BIO, "g_dev_done(%p/%p) resid %ld completed %jd",
  400                     bp2, bp, bp->bio_resid, (intmax_t)bp2->bio_completed);
  401         }
  402         bp->bio_resid = bp->bio_length - bp2->bio_completed;
  403         bp->bio_completed = bp2->bio_completed;
  404         g_destroy_bio(bp2);
  405         biodone(bp);
  406 }
  407 
  408 static void
  409 g_dev_strategy(struct bio *bp)
  410 {
  411         struct g_consumer *cp;
  412         struct bio *bp2;
  413         struct cdev *dev;
  414 
  415         KASSERT(bp->bio_cmd == BIO_READ ||
  416                 bp->bio_cmd == BIO_WRITE ||
  417                 bp->bio_cmd == BIO_DELETE,
  418                 ("Wrong bio_cmd bio=%p cmd=%d", bp, bp->bio_cmd));
  419         dev = bp->bio_dev;
  420         cp = dev->si_drv2;
  421         KASSERT(cp->acr || cp->acw,
  422             ("Consumer with zero access count in g_dev_strategy"));
  423 #ifdef INVARIANTS
  424         if ((bp->bio_offset % cp->provider->sectorsize) != 0 ||
  425             (bp->bio_bcount % cp->provider->sectorsize) != 0) {
  426                 bp->bio_resid = bp->bio_bcount;
  427                 biofinish(bp, NULL, EINVAL);
  428                 return;
  429         }
  430 #endif
  431         for (;;) {
  432                 /*
  433                  * XXX: This is not an ideal solution, but I belive it to
  434                  * XXX: deadlock safe, all things considered.
  435                  */
  436                 bp2 = g_clone_bio(bp);
  437                 if (bp2 != NULL)
  438                         break;
  439                 pause("gdstrat", hz / 10);
  440         }
  441         KASSERT(bp2 != NULL, ("XXX: ENOMEM in a bad place"));
  442         bp2->bio_done = g_dev_done;
  443         g_trace(G_T_BIO,
  444             "g_dev_strategy(%p/%p) offset %jd length %jd data %p cmd %d",
  445             bp, bp2, (intmax_t)bp->bio_offset, (intmax_t)bp2->bio_length,
  446             bp2->bio_data, bp2->bio_cmd);
  447         g_io_request(bp2, cp);
  448         KASSERT(cp->acr || cp->acw,
  449             ("g_dev_strategy raced with g_dev_close and lost"));
  450 
  451 }
  452 
  453 /*
  454  * g_dev_orphan()
  455  *
  456  * Called from below when the provider orphaned us.
  457  * - Clear any dump settings.
  458  * - Destroy the struct cdev *to prevent any more request from coming in.  The
  459  *   provider is already marked with an error, so anything which comes in
  460  *   in the interrim will be returned immediately.
  461  * - Wait for any outstanding I/O to finish.
  462  * - Set our access counts to zero, whatever they were.
  463  * - Detach and self-destruct.
  464  */
  465 
  466 static void
  467 g_dev_orphan(struct g_consumer *cp)
  468 {
  469         struct g_geom *gp;
  470         struct cdev *dev;
  471 
  472         g_topology_assert();
  473         gp = cp->geom;
  474         dev = gp->softc;
  475         g_trace(G_T_TOPOLOGY, "g_dev_orphan(%p(%s))", cp, gp->name);
  476 
  477         /* Reset any dump-area set on this device */
  478         if (dev->si_flags & SI_DUMPDEV)
  479                 set_dumper(NULL);
  480 
  481         /* Destroy the struct cdev *so we get no more requests */
  482         destroy_dev(dev);
  483 
  484         /* Wait for the cows to come home */
  485         while (cp->nstart != cp->nend)
  486                 pause("gdevorphan", hz / 10);
  487 
  488         if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0)
  489                 g_access(cp, -cp->acr, -cp->acw, -cp->ace);
  490 
  491         g_detach(cp);
  492         g_destroy_consumer(cp);
  493         g_destroy_geom(gp);
  494 }
  495 
  496 DECLARE_GEOM_CLASS(g_dev_class, g_dev);

Cache object: 381f97d4b4adbdb48cf16f702845885b


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.