freebsd-dev/sys/geom/nop/g_nop.c

705 lines
18 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
* All rights reserved.
*
* 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.
2006-02-01 12:06:01 +00:00
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS 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 AUTHORS 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/module.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/bio.h>
#include <sys/sbuf.h>
#include <sys/sysctl.h>
#include <sys/malloc.h>
#include <geom/geom.h>
#include <geom/nop/g_nop.h>
SYSCTL_DECL(_kern_geom);
static SYSCTL_NODE(_kern_geom, OID_AUTO, nop, CTLFLAG_RW, 0, "GEOM_NOP stuff");
static u_int g_nop_debug = 0;
SYSCTL_UINT(_kern_geom_nop, OID_AUTO, debug, CTLFLAG_RW, &g_nop_debug, 0,
"Debug level");
static int g_nop_destroy(struct g_geom *gp, boolean_t force);
static int g_nop_destroy_geom(struct gctl_req *req, struct g_class *mp,
struct g_geom *gp);
static void g_nop_config(struct gctl_req *req, struct g_class *mp,
const char *verb);
static void g_nop_dumpconf(struct sbuf *sb, const char *indent,
struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp);
struct g_class g_nop_class = {
.name = G_NOP_CLASS_NAME,
.version = G_VERSION,
.ctlreq = g_nop_config,
.destroy_geom = g_nop_destroy_geom
};
static void
g_nop_orphan(struct g_consumer *cp)
{
g_topology_assert();
g_nop_destroy(cp->geom, 1);
}
static void
g_nop_resize(struct g_consumer *cp)
{
struct g_nop_softc *sc;
struct g_geom *gp;
struct g_provider *pp;
off_t size;
g_topology_assert();
gp = cp->geom;
sc = gp->softc;
if (sc->sc_explicitsize != 0)
return;
if (cp->provider->mediasize < sc->sc_offset) {
g_nop_destroy(gp, 1);
return;
}
size = cp->provider->mediasize - sc->sc_offset;
LIST_FOREACH(pp, &gp->provider, provider)
g_resize_provider(pp, size);
}
static void
g_nop_start(struct bio *bp)
{
struct g_nop_softc *sc;
struct g_geom *gp;
struct g_provider *pp;
struct bio *cbp;
u_int failprob = 0;
gp = bp->bio_to->geom;
sc = gp->softc;
G_NOP_LOGREQ(bp, "Request received.");
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
mtx_lock(&sc->sc_lock);
switch (bp->bio_cmd) {
case BIO_READ:
sc->sc_reads++;
sc->sc_readbytes += bp->bio_length;
failprob = sc->sc_rfailprob;
break;
case BIO_WRITE:
sc->sc_writes++;
sc->sc_wrotebytes += bp->bio_length;
failprob = sc->sc_wfailprob;
break;
case BIO_DELETE:
sc->sc_deletes++;
break;
case BIO_GETATTR:
sc->sc_getattrs++;
break;
case BIO_FLUSH:
sc->sc_flushes++;
break;
case BIO_CMD0:
sc->sc_cmd0s++;
break;
case BIO_CMD1:
sc->sc_cmd1s++;
break;
case BIO_CMD2:
sc->sc_cmd2s++;
break;
}
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
mtx_unlock(&sc->sc_lock);
if (failprob > 0) {
u_int rval;
rval = arc4random() % 100;
if (rval < failprob) {
G_NOP_LOGREQLVL(1, bp, "Returning error=%d.", sc->sc_error);
g_io_deliver(bp, sc->sc_error);
return;
}
}
cbp = g_clone_bio(bp);
if (cbp == NULL) {
g_io_deliver(bp, ENOMEM);
return;
}
cbp->bio_done = g_std_done;
cbp->bio_offset = bp->bio_offset + sc->sc_offset;
pp = LIST_FIRST(&gp->provider);
KASSERT(pp != NULL, ("NULL pp"));
cbp->bio_to = pp;
G_NOP_LOGREQ(cbp, "Sending request.");
g_io_request(cbp, LIST_FIRST(&gp->consumer));
}
static int
g_nop_access(struct g_provider *pp, int dr, int dw, int de)
{
struct g_geom *gp;
struct g_consumer *cp;
int error;
gp = pp->geom;
cp = LIST_FIRST(&gp->consumer);
error = g_access(cp, dr, dw, de);
return (error);
}
static int
g_nop_create(struct gctl_req *req, struct g_class *mp, struct g_provider *pp,
int ioerror, u_int rfailprob, u_int wfailprob, off_t offset, off_t size,
u_int secsize, u_int stripesize, u_int stripeoffset)
{
struct g_nop_softc *sc;
struct g_geom *gp;
struct g_provider *newpp;
struct g_consumer *cp;
char name[64];
int error;
off_t explicitsize;
g_topology_assert();
gp = NULL;
newpp = NULL;
cp = NULL;
if ((offset % pp->sectorsize) != 0) {
gctl_error(req, "Invalid offset for provider %s.", pp->name);
return (EINVAL);
}
if ((size % pp->sectorsize) != 0) {
gctl_error(req, "Invalid size for provider %s.", pp->name);
return (EINVAL);
}
if (offset >= pp->mediasize) {
gctl_error(req, "Invalid offset for provider %s.", pp->name);
return (EINVAL);
}
explicitsize = size;
if (size == 0)
size = pp->mediasize - offset;
if (offset + size > pp->mediasize) {
gctl_error(req, "Invalid size for provider %s.", pp->name);
return (EINVAL);
}
if (secsize == 0)
secsize = pp->sectorsize;
else if ((secsize % pp->sectorsize) != 0) {
gctl_error(req, "Invalid secsize for provider %s.", pp->name);
return (EINVAL);
}
if (secsize > MAXPHYS) {
gctl_error(req, "secsize is too big.");
return (EINVAL);
}
size -= size % secsize;
if ((stripesize % pp->sectorsize) != 0) {
gctl_error(req, "Invalid stripesize for provider %s.", pp->name);
return (EINVAL);
}
if ((stripeoffset % pp->sectorsize) != 0) {
gctl_error(req, "Invalid stripeoffset for provider %s.", pp->name);
return (EINVAL);
}
if (stripesize != 0 && stripeoffset >= stripesize) {
gctl_error(req, "stripeoffset is too big.");
return (EINVAL);
}
snprintf(name, sizeof(name), "%s%s", pp->name, G_NOP_SUFFIX);
LIST_FOREACH(gp, &mp->geom, geom) {
if (strcmp(gp->name, name) == 0) {
gctl_error(req, "Provider %s already exists.", name);
return (EEXIST);
}
}
gp = g_new_geomf(mp, "%s", name);
sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
sc->sc_offset = offset;
sc->sc_explicitsize = explicitsize;
sc->sc_stripesize = stripesize;
sc->sc_stripeoffset = stripeoffset;
sc->sc_error = ioerror;
sc->sc_rfailprob = rfailprob;
sc->sc_wfailprob = wfailprob;
sc->sc_reads = 0;
sc->sc_writes = 0;
sc->sc_deletes = 0;
sc->sc_getattrs = 0;
sc->sc_flushes = 0;
sc->sc_cmd0s = 0;
sc->sc_cmd1s = 0;
sc->sc_cmd2s = 0;
sc->sc_readbytes = 0;
sc->sc_wrotebytes = 0;
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
mtx_init(&sc->sc_lock, "gnop lock", NULL, MTX_DEF);
gp->softc = sc;
gp->start = g_nop_start;
gp->orphan = g_nop_orphan;
gp->resize = g_nop_resize;
gp->access = g_nop_access;
gp->dumpconf = g_nop_dumpconf;
newpp = g_new_providerf(gp, "%s", gp->name);
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
newpp->flags |= G_PF_DIRECT_SEND | G_PF_DIRECT_RECEIVE;
newpp->mediasize = size;
newpp->sectorsize = secsize;
newpp->stripesize = stripesize;
newpp->stripeoffset = stripeoffset;
cp = g_new_consumer(gp);
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
cp->flags |= G_CF_DIRECT_SEND | G_CF_DIRECT_RECEIVE;
error = g_attach(cp, pp);
if (error != 0) {
gctl_error(req, "Cannot attach to provider %s.", pp->name);
goto fail;
}
newpp->flags |= pp->flags & G_PF_ACCEPT_UNMAPPED;
g_error_provider(newpp, 0);
G_NOP_DEBUG(0, "Device %s created.", gp->name);
return (0);
fail:
if (cp->provider != NULL)
g_detach(cp);
g_destroy_consumer(cp);
g_destroy_provider(newpp);
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
mtx_destroy(&sc->sc_lock);
g_free(gp->softc);
g_destroy_geom(gp);
return (error);
}
static int
g_nop_destroy(struct g_geom *gp, boolean_t force)
{
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
struct g_nop_softc *sc;
struct g_provider *pp;
g_topology_assert();
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
sc = gp->softc;
if (sc == NULL)
return (ENXIO);
pp = LIST_FIRST(&gp->provider);
if (pp != NULL && (pp->acr != 0 || pp->acw != 0 || pp->ace != 0)) {
if (force) {
G_NOP_DEBUG(0, "Device %s is still open, so it "
"can't be definitely removed.", pp->name);
} else {
G_NOP_DEBUG(1, "Device %s is still open (r%dw%de%d).",
pp->name, pp->acr, pp->acw, pp->ace);
return (EBUSY);
}
} else {
G_NOP_DEBUG(0, "Device %s removed.", gp->name);
}
gp->softc = NULL;
Merge GEOM direct dispatch changes from the projects/camlock branch. When safety requirements are met, it allows to avoid passing I/O requests to GEOM g_up/g_down thread, executing them directly in the caller context. That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid several context switches per I/O. The defined now safety requirements are: - caller should not hold any locks and should be reenterable; - callee should not depend on GEOM dual-threaded concurency semantics; - on the way down, if request is unmapped while callee doesn't support it, the context should be sleepable; - kernel thread stack usage should be below 50%. To keep compatibility with GEOM classes not meeting above requirements new provider and consumer flags added: - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request); - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done); - G_PF_DIRECT_SEND -- provider code meets caller requirements (done); - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request). Capable GEOM class can set them, allowing direct dispatch in cases where it is safe. If any of requirements are not met, request is queued to g_up or g_down thread same as before. Such GEOM classes were reviewed and updated to support direct dispatch: CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE, VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL, MAP, FLASHMAP, etc). To declare direct completion capability disk(9) KPI got new flag equivalent to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION. da(4) and ada(4) disk drivers got it set now thanks to earlier CAM locking work. This change more then twice increases peak block storage performance on systems with manu CPUs, together with earlier CAM locking changes reaching more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to 256 user-level threads). Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-22 08:22:19 +00:00
mtx_destroy(&sc->sc_lock);
g_free(sc);
g_wither_geom(gp, ENXIO);
return (0);
}
static int
g_nop_destroy_geom(struct gctl_req *req, struct g_class *mp, struct g_geom *gp)
{
return (g_nop_destroy(gp, 0));
}
static void
g_nop_ctl_create(struct gctl_req *req, struct g_class *mp)
{
struct g_provider *pp;
intmax_t *error, *rfailprob, *wfailprob, *offset, *secsize, *size,
*stripesize, *stripeoffset;
const char *name;
char param[16];
int i, *nargs;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument", "nargs");
return;
}
if (*nargs <= 0) {
gctl_error(req, "Missing device(s).");
return;
}
error = gctl_get_paraml(req, "error", sizeof(*error));
if (error == NULL) {
gctl_error(req, "No '%s' argument", "error");
return;
}
rfailprob = gctl_get_paraml(req, "rfailprob", sizeof(*rfailprob));
if (rfailprob == NULL) {
gctl_error(req, "No '%s' argument", "rfailprob");
return;
}
if (*rfailprob < -1 || *rfailprob > 100) {
gctl_error(req, "Invalid '%s' argument", "rfailprob");
return;
}
wfailprob = gctl_get_paraml(req, "wfailprob", sizeof(*wfailprob));
if (wfailprob == NULL) {
gctl_error(req, "No '%s' argument", "wfailprob");
return;
}
if (*wfailprob < -1 || *wfailprob > 100) {
gctl_error(req, "Invalid '%s' argument", "wfailprob");
return;
}
offset = gctl_get_paraml(req, "offset", sizeof(*offset));
if (offset == NULL) {
gctl_error(req, "No '%s' argument", "offset");
return;
}
if (*offset < 0) {
gctl_error(req, "Invalid '%s' argument", "offset");
return;
}
size = gctl_get_paraml(req, "size", sizeof(*size));
if (size == NULL) {
gctl_error(req, "No '%s' argument", "size");
return;
}
if (*size < 0) {
gctl_error(req, "Invalid '%s' argument", "size");
return;
}
secsize = gctl_get_paraml(req, "secsize", sizeof(*secsize));
if (secsize == NULL) {
gctl_error(req, "No '%s' argument", "secsize");
return;
}
if (*secsize < 0) {
gctl_error(req, "Invalid '%s' argument", "secsize");
return;
}
stripesize = gctl_get_paraml(req, "stripesize", sizeof(*stripesize));
if (stripesize == NULL) {
gctl_error(req, "No '%s' argument", "stripesize");
return;
}
if (*stripesize < 0) {
gctl_error(req, "Invalid '%s' argument", "stripesize");
return;
}
stripeoffset = gctl_get_paraml(req, "stripeoffset", sizeof(*stripeoffset));
if (stripeoffset == NULL) {
gctl_error(req, "No '%s' argument", "stripeoffset");
return;
}
if (*stripeoffset < 0) {
gctl_error(req, "Invalid '%s' argument", "stripeoffset");
return;
}
for (i = 0; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%d", i);
2006-02-01 12:06:01 +00:00
name = gctl_get_asciiparam(req, param);
if (name == NULL) {
gctl_error(req, "No 'arg%d' argument", i);
return;
}
if (strncmp(name, "/dev/", strlen("/dev/")) == 0)
name += strlen("/dev/");
pp = g_provider_by_name(name);
if (pp == NULL) {
G_NOP_DEBUG(1, "Provider %s is invalid.", name);
gctl_error(req, "Provider %s is invalid.", name);
2006-02-01 12:06:01 +00:00
return;
}
if (g_nop_create(req, mp, pp,
*error == -1 ? EIO : (int)*error,
*rfailprob == -1 ? 0 : (u_int)*rfailprob,
*wfailprob == -1 ? 0 : (u_int)*wfailprob,
(off_t)*offset, (off_t)*size, (u_int)*secsize,
(u_int)*stripesize, (u_int)*stripeoffset) != 0) {
return;
}
}
}
static void
g_nop_ctl_configure(struct gctl_req *req, struct g_class *mp)
{
struct g_nop_softc *sc;
struct g_provider *pp;
intmax_t *error, *rfailprob, *wfailprob;
const char *name;
char param[16];
int i, *nargs;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument", "nargs");
return;
}
if (*nargs <= 0) {
gctl_error(req, "Missing device(s).");
return;
}
error = gctl_get_paraml(req, "error", sizeof(*error));
if (error == NULL) {
gctl_error(req, "No '%s' argument", "error");
return;
}
rfailprob = gctl_get_paraml(req, "rfailprob", sizeof(*rfailprob));
if (rfailprob == NULL) {
gctl_error(req, "No '%s' argument", "rfailprob");
return;
}
if (*rfailprob < -1 || *rfailprob > 100) {
gctl_error(req, "Invalid '%s' argument", "rfailprob");
return;
}
wfailprob = gctl_get_paraml(req, "wfailprob", sizeof(*wfailprob));
if (wfailprob == NULL) {
gctl_error(req, "No '%s' argument", "wfailprob");
return;
}
if (*wfailprob < -1 || *wfailprob > 100) {
gctl_error(req, "Invalid '%s' argument", "wfailprob");
return;
}
for (i = 0; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%d", i);
2006-02-01 12:06:01 +00:00
name = gctl_get_asciiparam(req, param);
if (name == NULL) {
gctl_error(req, "No 'arg%d' argument", i);
return;
}
if (strncmp(name, "/dev/", strlen("/dev/")) == 0)
name += strlen("/dev/");
pp = g_provider_by_name(name);
if (pp == NULL || pp->geom->class != mp) {
G_NOP_DEBUG(1, "Provider %s is invalid.", name);
gctl_error(req, "Provider %s is invalid.", name);
2006-02-01 12:06:01 +00:00
return;
}
sc = pp->geom->softc;
if (*error != -1)
sc->sc_error = (int)*error;
if (*rfailprob != -1)
sc->sc_rfailprob = (u_int)*rfailprob;
if (*wfailprob != -1)
sc->sc_wfailprob = (u_int)*wfailprob;
}
}
static struct g_geom *
g_nop_find_geom(struct g_class *mp, const char *name)
{
struct g_geom *gp;
LIST_FOREACH(gp, &mp->geom, geom) {
if (strcmp(gp->name, name) == 0)
return (gp);
}
return (NULL);
}
static void
g_nop_ctl_destroy(struct gctl_req *req, struct g_class *mp)
{
int *nargs, *force, error, i;
struct g_geom *gp;
const char *name;
char param[16];
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument", "nargs");
return;
}
if (*nargs <= 0) {
gctl_error(req, "Missing device(s).");
return;
}
force = gctl_get_paraml(req, "force", sizeof(*force));
if (force == NULL) {
gctl_error(req, "No 'force' argument");
return;
}
for (i = 0; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%d", i);
2006-02-01 12:06:01 +00:00
name = gctl_get_asciiparam(req, param);
if (name == NULL) {
gctl_error(req, "No 'arg%d' argument", i);
return;
}
if (strncmp(name, "/dev/", strlen("/dev/")) == 0)
name += strlen("/dev/");
gp = g_nop_find_geom(mp, name);
if (gp == NULL) {
G_NOP_DEBUG(1, "Device %s is invalid.", name);
gctl_error(req, "Device %s is invalid.", name);
2006-02-01 12:06:01 +00:00
return;
}
error = g_nop_destroy(gp, *force);
if (error != 0) {
gctl_error(req, "Cannot destroy device %s (error=%d).",
gp->name, error);
return;
}
}
}
static void
g_nop_ctl_reset(struct gctl_req *req, struct g_class *mp)
{
struct g_nop_softc *sc;
struct g_provider *pp;
const char *name;
char param[16];
int i, *nargs;
g_topology_assert();
nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
if (nargs == NULL) {
gctl_error(req, "No '%s' argument", "nargs");
return;
}
if (*nargs <= 0) {
gctl_error(req, "Missing device(s).");
return;
}
for (i = 0; i < *nargs; i++) {
snprintf(param, sizeof(param), "arg%d", i);
2006-02-01 12:06:01 +00:00
name = gctl_get_asciiparam(req, param);
if (name == NULL) {
gctl_error(req, "No 'arg%d' argument", i);
return;
}
if (strncmp(name, "/dev/", strlen("/dev/")) == 0)
name += strlen("/dev/");
pp = g_provider_by_name(name);
if (pp == NULL || pp->geom->class != mp) {
G_NOP_DEBUG(1, "Provider %s is invalid.", name);
gctl_error(req, "Provider %s is invalid.", name);
2006-02-01 12:06:01 +00:00
return;
}
sc = pp->geom->softc;
sc->sc_reads = 0;
sc->sc_writes = 0;
sc->sc_deletes = 0;
sc->sc_getattrs = 0;
sc->sc_flushes = 0;
sc->sc_cmd0s = 0;
sc->sc_cmd1s = 0;
sc->sc_cmd2s = 0;
sc->sc_readbytes = 0;
sc->sc_wrotebytes = 0;
}
}
static void
g_nop_config(struct gctl_req *req, struct g_class *mp, const char *verb)
{
uint32_t *version;
g_topology_assert();
version = gctl_get_paraml(req, "version", sizeof(*version));
if (version == NULL) {
gctl_error(req, "No '%s' argument.", "version");
return;
}
if (*version != G_NOP_VERSION) {
gctl_error(req, "Userland and kernel parts are out of sync.");
return;
}
if (strcmp(verb, "create") == 0) {
g_nop_ctl_create(req, mp);
return;
} else if (strcmp(verb, "configure") == 0) {
g_nop_ctl_configure(req, mp);
return;
} else if (strcmp(verb, "destroy") == 0) {
g_nop_ctl_destroy(req, mp);
return;
} else if (strcmp(verb, "reset") == 0) {
g_nop_ctl_reset(req, mp);
return;
}
gctl_error(req, "Unknown verb.");
}
static void
g_nop_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp,
struct g_consumer *cp, struct g_provider *pp)
{
struct g_nop_softc *sc;
if (pp != NULL || cp != NULL)
return;
sc = gp->softc;
sbuf_printf(sb, "%s<Offset>%jd</Offset>\n", indent,
(intmax_t)sc->sc_offset);
sbuf_printf(sb, "%s<ReadFailProb>%u</ReadFailProb>\n", indent,
sc->sc_rfailprob);
sbuf_printf(sb, "%s<WriteFailProb>%u</WriteFailProb>\n", indent,
sc->sc_wfailprob);
sbuf_printf(sb, "%s<Error>%d</Error>\n", indent, sc->sc_error);
sbuf_printf(sb, "%s<Reads>%ju</Reads>\n", indent, sc->sc_reads);
sbuf_printf(sb, "%s<Writes>%ju</Writes>\n", indent, sc->sc_writes);
sbuf_printf(sb, "%s<Deletes>%ju</Deletes>\n", indent, sc->sc_deletes);
sbuf_printf(sb, "%s<Getattrs>%ju</Getattrs>\n", indent, sc->sc_getattrs);
sbuf_printf(sb, "%s<Flushes>%ju</Flushes>\n", indent, sc->sc_flushes);
sbuf_printf(sb, "%s<Cmd0s>%ju</Cmd0s>\n", indent, sc->sc_cmd0s);
sbuf_printf(sb, "%s<Cmd1s>%ju</Cmd1s>\n", indent, sc->sc_cmd1s);
sbuf_printf(sb, "%s<Cmd2s>%ju</Cmd2s>\n", indent, sc->sc_cmd2s);
sbuf_printf(sb, "%s<ReadBytes>%ju</ReadBytes>\n", indent,
sc->sc_readbytes);
sbuf_printf(sb, "%s<WroteBytes>%ju</WroteBytes>\n", indent,
sc->sc_wrotebytes);
}
DECLARE_GEOM_CLASS(g_nop_class, g_nop);