freebsd-dev/sys/geom/geom_vfs.c
Alexander Motin 40ea77a036 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

285 lines
6.6 KiB
C

/*-
* Copyright (c) 2004 Poul-Henning Kamp
* 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.
*
* 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <geom/geom.h>
#include <geom/geom_vfs.h>
/*
* subroutines for use by filesystems.
*
* XXX: should maybe live somewhere else ?
*/
#include <sys/buf.h>
struct g_vfs_softc {
struct mtx sc_mtx;
struct bufobj *sc_bo;
int sc_active;
int sc_orphaned;
};
static struct buf_ops __g_vfs_bufops = {
.bop_name = "GEOM_VFS",
.bop_write = bufwrite,
.bop_strategy = g_vfs_strategy,
.bop_sync = bufsync,
.bop_bdflush = bufbdflush
};
struct buf_ops *g_vfs_bufops = &__g_vfs_bufops;
static g_orphan_t g_vfs_orphan;
static struct g_class g_vfs_class = {
.name = "VFS",
.version = G_VERSION,
.orphan = g_vfs_orphan,
};
DECLARE_GEOM_CLASS(g_vfs_class, g_vfs);
static void
g_vfs_destroy(void *arg, int flags __unused)
{
struct g_consumer *cp;
g_topology_assert();
cp = arg;
if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0)
g_access(cp, -cp->acr, -cp->acw, -cp->ace);
g_detach(cp);
if (cp->geom->softc == NULL)
g_wither_geom(cp->geom, ENXIO);
}
static void
g_vfs_done(struct bio *bip)
{
struct g_consumer *cp;
struct g_vfs_softc *sc;
struct buf *bp;
int destroy;
struct mount *mp;
struct vnode *vp;
struct cdev *cdevp;
/*
* Collect statistics on synchronous and asynchronous read
* and write counts for disks that have associated filesystems.
*/
bp = bip->bio_caller2;
vp = bp->b_vp;
if (vp != NULL) {
/*
* If not a disk vnode, use its associated mount point
* otherwise use the mountpoint associated with the disk.
*/
VI_LOCK(vp);
if (vp->v_type != VCHR ||
(cdevp = vp->v_rdev) == NULL ||
cdevp->si_devsw == NULL ||
(cdevp->si_devsw->d_flags & D_DISK) == 0)
mp = vp->v_mount;
else
mp = cdevp->si_mountpt;
if (mp != NULL) {
if (bp->b_iocmd == BIO_READ) {
if (LK_HOLDER(bp->b_lock.lk_lock) == LK_KERNPROC)
mp->mnt_stat.f_asyncreads++;
else
mp->mnt_stat.f_syncreads++;
} else if (bp->b_iocmd == BIO_WRITE) {
if (LK_HOLDER(bp->b_lock.lk_lock) == LK_KERNPROC)
mp->mnt_stat.f_asyncwrites++;
else
mp->mnt_stat.f_syncwrites++;
}
}
VI_UNLOCK(vp);
}
cp = bip->bio_from;
sc = cp->geom->softc;
if (bip->bio_error) {
printf("g_vfs_done():");
g_print_bio(bip);
printf("error = %d\n", bip->bio_error);
}
bp->b_error = bip->bio_error;
bp->b_ioflags = bip->bio_flags;
if (bip->bio_error)
bp->b_ioflags |= BIO_ERROR;
bp->b_resid = bp->b_bcount - bip->bio_completed;
g_destroy_bio(bip);
mtx_lock(&sc->sc_mtx);
destroy = ((--sc->sc_active) == 0 && sc->sc_orphaned);
mtx_unlock(&sc->sc_mtx);
if (destroy)
g_post_event(g_vfs_destroy, cp, M_WAITOK, NULL);
bufdone(bp);
}
void
g_vfs_strategy(struct bufobj *bo, struct buf *bp)
{
struct g_vfs_softc *sc;
struct g_consumer *cp;
struct bio *bip;
cp = bo->bo_private;
sc = cp->geom->softc;
/*
* If the provider has orphaned us, just return EXIO.
*/
mtx_lock(&sc->sc_mtx);
if (sc->sc_orphaned) {
mtx_unlock(&sc->sc_mtx);
bp->b_error = ENXIO;
bp->b_ioflags |= BIO_ERROR;
bufdone(bp);
return;
}
sc->sc_active++;
mtx_unlock(&sc->sc_mtx);
bip = g_alloc_bio();
bip->bio_cmd = bp->b_iocmd;
bip->bio_offset = bp->b_iooffset;
bip->bio_length = bp->b_bcount;
bdata2bio(bp, bip);
if ((bp->b_flags & B_BARRIER) != 0) {
bip->bio_flags |= BIO_ORDERED;
bp->b_flags &= ~B_BARRIER;
}
bip->bio_done = g_vfs_done;
bip->bio_caller2 = bp;
g_io_request(bip, cp);
}
static void
g_vfs_orphan(struct g_consumer *cp)
{
struct g_geom *gp;
struct g_vfs_softc *sc;
int destroy;
g_topology_assert();
gp = cp->geom;
g_trace(G_T_TOPOLOGY, "g_vfs_orphan(%p(%s))", cp, gp->name);
sc = gp->softc;
if (sc == NULL)
return;
mtx_lock(&sc->sc_mtx);
sc->sc_orphaned = 1;
destroy = (sc->sc_active == 0);
mtx_unlock(&sc->sc_mtx);
if (destroy)
g_vfs_destroy(cp, 0);
/*
* Do not destroy the geom. Filesystem will do that during unmount.
*/
}
int
g_vfs_open(struct vnode *vp, struct g_consumer **cpp, const char *fsname, int wr)
{
struct g_geom *gp;
struct g_provider *pp;
struct g_consumer *cp;
struct g_vfs_softc *sc;
struct bufobj *bo;
int error;
g_topology_assert();
*cpp = NULL;
bo = &vp->v_bufobj;
if (bo->bo_private != vp)
return (EBUSY);
pp = g_dev_getprovider(vp->v_rdev);
if (pp == NULL)
return (ENOENT);
gp = g_new_geomf(&g_vfs_class, "%s.%s", fsname, pp->name);
sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
mtx_init(&sc->sc_mtx, "g_vfs", NULL, MTX_DEF);
sc->sc_bo = bo;
gp->softc = sc;
cp = g_new_consumer(gp);
g_attach(cp, pp);
error = g_access(cp, 1, wr, wr);
if (error) {
g_wither_geom(gp, ENXIO);
return (error);
}
vnode_create_vobject(vp, pp->mediasize, curthread);
*cpp = cp;
cp->private = vp;
cp->flags |= G_CF_DIRECT_SEND | G_CF_DIRECT_RECEIVE;
bo->bo_ops = g_vfs_bufops;
bo->bo_private = cp;
bo->bo_bsize = pp->sectorsize;
return (error);
}
void
g_vfs_close(struct g_consumer *cp)
{
struct g_geom *gp;
struct g_vfs_softc *sc;
g_topology_assert();
gp = cp->geom;
sc = gp->softc;
bufobj_invalbuf(sc->sc_bo, V_SAVE, 0, 0);
sc->sc_bo->bo_private = cp->private;
gp->softc = NULL;
mtx_destroy(&sc->sc_mtx);
if (!sc->sc_orphaned || cp->provider == NULL)
g_wither_geom_close(gp, ENXIO);
g_free(sc);
}