freebsd-skq/sys/geom/geom_dev.c
Poul-Henning Kamp c7e1925c7c Properly handle races between open/close and orphan.
KASSERT the race between close and strategy, it is an error in the upper
echelons if this happens,

Add XXX: comment explaining why the ioctl/orphan race is not closed.
2003-04-02 13:09:50 +00:00

463 lines
12 KiB
C

/*-
* Copyright (c) 2002 Poul-Henning Kamp
* Copyright (c) 2002 Networks Associates Technology, Inc.
* All rights reserved.
*
* This software was developed for the FreeBSD Project by Poul-Henning Kamp
* and NAI Labs, the Security Research Division of Network Associates, Inc.
* under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
* DARPA CHATS research program.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The names of the authors may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/conf.h>
#include <sys/bio.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/errno.h>
#include <sys/time.h>
#include <sys/disk.h>
#include <sys/fcntl.h>
#include <geom/geom.h>
#include <geom/geom_int.h>
#include <machine/limits.h>
static d_open_t g_dev_open;
static d_close_t g_dev_close;
static d_strategy_t g_dev_strategy;
static d_ioctl_t g_dev_ioctl;
static struct cdevsw g_dev_cdevsw = {
.d_open = g_dev_open,
.d_close = g_dev_close,
.d_read = physread,
.d_write = physwrite,
.d_ioctl = g_dev_ioctl,
.d_strategy = g_dev_strategy,
.d_name = "g_dev",
.d_maj = GEOM_MAJOR,
.d_flags = D_DISK | D_TRACKCLOSE,
};
static g_taste_t g_dev_taste;
static g_orphan_t g_dev_orphan;
static struct g_class g_dev_class = {
.name = "DEV",
.taste = g_dev_taste,
G_CLASS_INITIALIZER
};
int
g_dev_print(void)
{
struct g_geom *gp;
if (LIST_EMPTY(&g_dev_class.geom))
return (0);
printf("List of GEOM disk devices:\n ");
LIST_FOREACH(gp, &g_dev_class.geom, geom)
printf(" %s", gp->name);
printf("\n");
return (1);
}
/*
* XXX: This is disgusting and wrong in every way imaginable: The only reason
* XXX: we have a clone function is because of the root-mount hack we currently
* XXX: employ. An improvment would be to unregister this cloner once we know
* XXX: we no longer need it. Ideally, root-fs would be mounted through DEVFS
* XXX: eliminating the need for this hack.
*/
static void
g_dev_clone(void *arg __unused, char *name, int namelen __unused, dev_t *dev)
{
struct g_geom *gp;
if (*dev != NODEV)
return;
g_waitidle();
/* g_topology_lock(); */
LIST_FOREACH(gp, &g_dev_class.geom, geom) {
if (strcmp(gp->name, name))
continue;
*dev = gp->softc;
g_trace(G_T_TOPOLOGY, "g_dev_clone(%s) = %p", name, *dev);
return;
}
/* g_topology_unlock(); */
return;
}
static void
g_dev_register_cloner(void *foo __unused)
{
static int once;
/* XXX: why would this happen more than once ?? */
if (!once) {
EVENTHANDLER_REGISTER(dev_clone, g_dev_clone, 0, 1000);
once++;
}
}
SYSINIT(geomdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE,g_dev_register_cloner,NULL);
static struct g_geom *
g_dev_taste(struct g_class *mp, struct g_provider *pp, int insist __unused)
{
struct g_geom *gp;
struct g_consumer *cp;
static int unit = GEOM_MINOR_PROVIDERS;
int error;
dev_t dev;
g_trace(G_T_TOPOLOGY, "dev_taste(%s,%s)", mp->name, pp->name);
g_topology_assert();
LIST_FOREACH(cp, &pp->consumers, consumers)
if (cp->geom->class == mp)
return (NULL);
gp = g_new_geomf(mp, pp->name);
gp->orphan = g_dev_orphan;
cp = g_new_consumer(gp);
error = g_attach(cp, pp);
KASSERT(error == 0,
("g_dev_taste(%s) failed to g_attach, err=%d", pp->name, error));
/*
* XXX: I'm not 100% sure we can call make_dev(9) without Giant
* yet. Once we can, we don't need to drop topology here either.
*/
g_topology_unlock();
mtx_lock(&Giant);
dev = make_dev(&g_dev_cdevsw, unit2minor(unit++),
UID_ROOT, GID_OPERATOR, 0640, gp->name);
if (pp->flags & G_PF_CANDELETE)
dev->si_flags |= SI_CANDELETE;
mtx_unlock(&Giant);
g_topology_lock();
dev->si_iosize_max = MAXPHYS;
dev->si_stripesize = pp->stripesize;
dev->si_stripeoffset = pp->stripeoffset;
gp->softc = dev;
dev->si_drv1 = gp;
dev->si_drv2 = cp;
return (gp);
}
static int
g_dev_open(dev_t dev, int flags, int fmt, struct thread *td)
{
struct g_geom *gp;
struct g_consumer *cp;
int error, r, w, e;
gp = dev->si_drv1;
cp = dev->si_drv2;
if (gp == NULL || cp == NULL || gp->softc != dev)
return(ENXIO); /* g_dev_taste() not done yet */
g_trace(G_T_ACCESS, "g_dev_open(%s, %d, %d, %p)",
gp->name, flags, fmt, td);
r = flags & FREAD ? 1 : 0;
w = flags & FWRITE ? 1 : 0;
#ifdef notyet
e = flags & O_EXCL ? 1 : 0;
#else
e = 0;
#endif
DROP_GIANT();
g_topology_lock();
if (dev->si_devsw == NULL)
error = ENXIO; /* We were orphaned */
else
error = g_access_rel(cp, r, w, e);
g_topology_unlock();
PICKUP_GIANT();
g_waitidle();
if (!error)
dev->si_bsize_phys = cp->provider->sectorsize;
return(error);
}
static int
g_dev_close(dev_t dev, int flags, int fmt, struct thread *td)
{
struct g_geom *gp;
struct g_consumer *cp;
int error, r, w, e;
gp = dev->si_drv1;
cp = dev->si_drv2;
if (gp == NULL || cp == NULL)
return(ENXIO);
g_trace(G_T_ACCESS, "g_dev_close(%s, %d, %d, %p)",
gp->name, flags, fmt, td);
r = flags & FREAD ? -1 : 0;
w = flags & FWRITE ? -1 : 0;
#ifdef notyet
e = flags & O_EXCL ? -1 : 0;
#else
e = 0;
#endif
DROP_GIANT();
g_topology_lock();
if (dev->si_devsw == NULL)
error = ENXIO; /* We were orphaned */
else
error = g_access_rel(cp, r, w, e);
KASSERT((cp->acr || cp->acw) || (cp->nstart == cp->nend),
("final g_dev_close() with outstanding bios"));
g_topology_unlock();
PICKUP_GIANT();
g_waitidle();
return (error);
}
/*
* XXX: Until we have unmessed the ioctl situation, there is a race against
* XXX: a concurrent orphanization. We cannot close it by holding topology
* XXX: since that would prevent us from doing our job, and stalling events
* XXX: will break (actually: stall) the BSD disklabel hacks.
*/
static int
g_dev_ioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
{
struct g_geom *gp, *gp2;
struct g_consumer *cp;
struct g_provider *pp2;
struct g_kerneldump kd;
int i, error;
u_int u;
struct g_ioctl *gio;
gp = dev->si_drv1;
cp = dev->si_drv2;
pp2 = cp->provider;
gp2 = pp2->geom;
gio = NULL;
error = 0;
KASSERT(cp->acr || cp->acw,
("Consumer with zero access count in g_dev_ioctl"));
DROP_GIANT();
gio = NULL;
i = IOCPARM_LEN(cmd);
switch (cmd) {
case DIOCGSECTORSIZE:
*(u_int *)data = cp->provider->sectorsize;
if (*(u_int *)data == 0)
error = ENOENT;
break;
case DIOCGMEDIASIZE:
*(off_t *)data = cp->provider->mediasize;
if (*(off_t *)data == 0)
error = ENOENT;
break;
case DIOCGFWSECTORS:
error = g_io_getattr("GEOM::fwsectors", cp, &i, data);
if (error == 0 && *(u_int *)data == 0)
error = ENOENT;
break;
case DIOCGFWHEADS:
error = g_io_getattr("GEOM::fwheads", cp, &i, data);
if (error == 0 && *(u_int *)data == 0)
error = ENOENT;
break;
case DIOCGFRONTSTUFF:
error = g_io_getattr("GEOM::frontstuff", cp, &i, data);
break;
case DIOCSKERNELDUMP:
u = *((u_int *)data);
if (!u) {
set_dumper(NULL);
error = 0;
break;
}
kd.offset = 0;
kd.length = OFF_MAX;
i = sizeof kd;
error = g_io_getattr("GEOM::kerneldump", cp, &i, &kd);
if (!error)
dev->si_flags |= SI_DUMPDEV;
break;
default:
gio = g_malloc(sizeof *gio, M_WAITOK | M_ZERO);
gio->cmd = cmd;
gio->data = data;
gio->fflag = fflag;
gio->td = td;
i = sizeof *gio;
/*
* We always issue ioctls as getattr since the direction of data
* movement in ioctl is no indication of the ioctl being a "set"
* or "get" type ioctl or if such simplistic terms even apply
*/
error = g_io_getattr("GEOM::ioctl", cp, &i, gio);
break;
}
PICKUP_GIANT();
if (error == EDIRIOCTL) {
KASSERT(gio != NULL, ("NULL gio but EDIRIOCTL"));
KASSERT(gio->func != NULL, ("NULL function but EDIRIOCTL"));
error = (gio->func)(gio->dev, cmd, data, fflag, td);
}
g_waitidle();
if (gio != NULL && (error == EOPNOTSUPP || error == ENOIOCTL)) {
if (g_debugflags & G_T_TOPOLOGY) {
i = IOCGROUP(cmd);
printf("IOCTL(0x%lx) \"%s\"", cmd, gp->name);
if (i > ' ' && i <= '~')
printf(" '%c'", (int)IOCGROUP(cmd));
else
printf(" 0x%lx", IOCGROUP(cmd));
printf("/%ld ", cmd & 0xff);
if (cmd & IOC_IN)
printf("I");
if (cmd & IOC_OUT)
printf("O");
printf("(%ld) = ENOIOCTL\n", IOCPARM_LEN(cmd));
}
error = ENOTTY;
}
if (gio != NULL)
g_free(gio);
return (error);
}
static void
g_dev_done(struct bio *bp2)
{
struct bio *bp;
bp = bp2->bio_parent;
bp->bio_error = bp2->bio_error;
if (bp->bio_error != 0) {
g_trace(G_T_BIO, "g_dev_done(%p) had error %d",
bp2, bp->bio_error);
bp->bio_flags |= BIO_ERROR;
} else {
g_trace(G_T_BIO, "g_dev_done(%p/%p) resid %ld completed %jd",
bp2, bp, bp->bio_resid, (intmax_t)bp2->bio_completed);
}
bp->bio_resid = bp->bio_bcount - bp2->bio_completed;
g_destroy_bio(bp2);
mtx_lock(&Giant);
biodone(bp);
mtx_unlock(&Giant);
}
static void
g_dev_strategy(struct bio *bp)
{
struct g_geom *gp;
struct g_consumer *cp;
struct bio *bp2;
dev_t dev;
KASSERT(bp->bio_cmd == BIO_READ ||
bp->bio_cmd == BIO_WRITE ||
bp->bio_cmd == BIO_DELETE,
("Wrong bio_cmd bio=%p cmd=%d", bp, bp->bio_cmd));
dev = bp->bio_dev;
gp = dev->si_drv1;
cp = dev->si_drv2;
KASSERT(cp->acr || cp->acw,
("Consumer with zero access count in g_dev_strategy"));
bp2 = g_clone_bio(bp);
KASSERT(bp2 != NULL, ("XXX: ENOMEM in a bad place"));
bp2->bio_offset = (off_t)bp->bio_blkno << DEV_BSHIFT;
KASSERT(bp2->bio_offset >= 0,
("Negative bio_offset (%jd) on bio %p",
(intmax_t)bp2->bio_offset, bp));
bp2->bio_length = (off_t)bp->bio_bcount;
bp2->bio_done = g_dev_done;
g_trace(G_T_BIO,
"g_dev_strategy(%p/%p) offset %jd length %jd data %p cmd %d",
bp, bp2, (intmax_t)bp->bio_offset, (intmax_t)bp2->bio_length,
bp2->bio_data, bp2->bio_cmd);
g_io_request(bp2, cp);
KASSERT(cp->acr || cp->acw,
("g_dev_strategy raced with g_dev_close and lost"));
}
/*
* g_dev_orphan()
*
* Called from below when the provider orphaned us.
* - Clear any dump settings.
* - Destroy the dev_t to prevent any more request from coming in. The
* provider is already marked with an error, so anything which comes in
* in the interrim will be returned immediately.
* - Wait for any outstanding I/O to finish.
* - Set our access counts to zero, whatever they were.
* - Detach and self-destruct.
*/
static void
g_dev_orphan(struct g_consumer *cp)
{
struct g_geom *gp;
dev_t dev;
g_topology_assert();
gp = cp->geom;
dev = gp->softc;
g_trace(G_T_TOPOLOGY, "g_dev_orphan(%p(%s))", cp, gp->name);
/* Reset any dump-area set on this device */
if (dev->si_flags & SI_DUMPDEV)
set_dumper(NULL);
/* Destroy the dev_t so we get no more requests */
destroy_dev(dev);
/* Wait for the cows to come home */
while (cp->nstart != cp->nend)
msleep(&dev, NULL, PRIBIO, "gdevorphan", hz / 10);
if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0)
g_access_rel(cp, -cp->acr, -cp->acw, -cp->ace);
g_detach(cp);
g_destroy_consumer(cp);
g_destroy_geom(gp);
}
DECLARE_GEOM_CLASS(g_dev_class, g_dev);