1 /*- 2 * Copyright (c) 2004 Poul-Henning Kamp 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/param.h> 31 #include <sys/systm.h> 32 #include <sys/bio.h> 33 #include <sys/kernel.h> 34 #include <sys/lock.h> 35 #include <sys/malloc.h> 36 #include <sys/mutex.h> 37 #include <sys/vnode.h> 38 #include <sys/mount.h> /* XXX Temporary for VFS_LOCK_GIANT */ 39 40 #include <geom/geom.h> 41 #include <geom/geom_vfs.h> 42 43 /* 44 * subroutines for use by filesystems. 45 * 46 * XXX: should maybe live somewhere else ? 47 */ 48 #include <sys/buf.h> 49 50 struct g_vfs_softc { 51 struct mtx sc_mtx; 52 struct bufobj *sc_bo; 53 int sc_active; 54 int sc_orphaned; 55 }; 56 57 static struct buf_ops __g_vfs_bufops = { 58 .bop_name = "GEOM_VFS", 59 .bop_write = bufwrite, 60 .bop_strategy = g_vfs_strategy, 61 .bop_sync = bufsync, 62 .bop_bdflush = bufbdflush 63 }; 64 65 struct buf_ops *g_vfs_bufops = &__g_vfs_bufops; 66 67 static g_orphan_t g_vfs_orphan; 68 69 static struct g_class g_vfs_class = { 70 .name = "VFS", 71 .version = G_VERSION, 72 .orphan = g_vfs_orphan, 73 }; 74 75 DECLARE_GEOM_CLASS(g_vfs_class, g_vfs); 76 77 static void 78 g_vfs_destroy(void *arg, int flags __unused) 79 { 80 struct g_consumer *cp; 81 82 g_topology_assert(); 83 cp = arg; 84 if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0) 85 g_access(cp, -cp->acr, -cp->acw, -cp->ace); 86 g_detach(cp); 87 if (cp->geom->softc == NULL) 88 g_wither_geom(cp->geom, ENXIO); 89 } 90 91 static void 92 g_vfs_done(struct bio *bip) 93 { 94 struct g_consumer *cp; 95 struct g_vfs_softc *sc; 96 struct buf *bp; 97 int vfslocked, destroy; 98 struct mount *mp; 99 struct vnode *vp; 100 101 /* 102 * Collect statistics on synchronous and asynchronous read 103 * and write counts for disks that have associated filesystems. 104 * Since this run by the g_up thread it is single threaded and 105 * we do not need to use atomic increments on the counters. 106 */ 107 bp = bip->bio_caller2; 108 vp = bp->b_vp; 109 if (vp == NULL) 110 mp = NULL; 111 else if (vn_isdisk(vp, NULL)) 112 mp = vp->v_rdev->si_mountpt; 113 else 114 mp = vp->v_mount; 115 if (mp != NULL) { 116 if (bp->b_iocmd == BIO_WRITE) { 117 if (LK_HOLDER(bp->b_lock.lk_lock) == LK_KERNPROC) 118 mp->mnt_stat.f_asyncwrites++; 119 else 120 mp->mnt_stat.f_syncwrites++; 121 } else { 122 if (LK_HOLDER(bp->b_lock.lk_lock) == LK_KERNPROC) 123 mp->mnt_stat.f_asyncreads++; 124 else 125 mp->mnt_stat.f_syncreads++; 126 } 127 } 128 129 cp = bip->bio_from; 130 sc = cp->geom->softc; 131 if (bip->bio_error) { 132 printf("g_vfs_done():"); 133 g_print_bio(bip); 134 printf("error = %d\n", bip->bio_error); 135 } 136 bp->b_error = bip->bio_error; 137 bp->b_ioflags = bip->bio_flags; 138 if (bip->bio_error) 139 bp->b_ioflags |= BIO_ERROR; 140 bp->b_resid = bp->b_bcount - bip->bio_completed; 141 g_destroy_bio(bip); 142 143 mtx_lock(&sc->sc_mtx); 144 destroy = ((--sc->sc_active) == 0 && sc->sc_orphaned); 145 mtx_unlock(&sc->sc_mtx); 146 if (destroy) 147 g_post_event(g_vfs_destroy, cp, M_WAITOK, NULL); 148 149 vfslocked = VFS_LOCK_GIANT(((struct mount *)NULL)); 150 bufdone(bp); 151 VFS_UNLOCK_GIANT(vfslocked); 152 } 153 154 void 155 g_vfs_strategy(struct bufobj *bo, struct buf *bp) 156 { 157 struct g_vfs_softc *sc; 158 struct g_consumer *cp; 159 struct bio *bip; 160 int vfslocked; 161 162 cp = bo->bo_private; 163 sc = cp->geom->softc; 164 165 /* 166 * If the provider has orphaned us, just return EXIO. 167 */ 168 mtx_lock(&sc->sc_mtx); 169 if (sc->sc_orphaned) { 170 mtx_unlock(&sc->sc_mtx); 171 bp->b_error = ENXIO; 172 bp->b_ioflags |= BIO_ERROR; 173 vfslocked = VFS_LOCK_GIANT(((struct mount *)NULL)); 174 bufdone(bp); 175 VFS_UNLOCK_GIANT(vfslocked); 176 return; 177 } 178 sc->sc_active++; 179 mtx_unlock(&sc->sc_mtx); 180 181 bip = g_alloc_bio(); 182 bip->bio_cmd = bp->b_iocmd; 183 bip->bio_offset = bp->b_iooffset; 184 bip->bio_data = bp->b_data; 185 bip->bio_done = g_vfs_done; 186 bip->bio_caller2 = bp; 187 bip->bio_length = bp->b_bcount; 188 g_io_request(bip, cp); 189 } 190 191 static void 192 g_vfs_orphan(struct g_consumer *cp) 193 { 194 struct g_geom *gp; 195 struct g_vfs_softc *sc; 196 int destroy; 197 198 g_topology_assert(); 199 200 gp = cp->geom; 201 g_trace(G_T_TOPOLOGY, "g_vfs_orphan(%p(%s))", cp, gp->name); 202 sc = gp->softc; 203 if (sc == NULL) 204 return; 205 mtx_lock(&sc->sc_mtx); 206 sc->sc_orphaned = 1; 207 destroy = (sc->sc_active == 0); 208 mtx_unlock(&sc->sc_mtx); 209 if (destroy) 210 g_vfs_destroy(cp, 0); 211 212 /* 213 * Do not destroy the geom. Filesystem will do that during unmount. 214 */ 215 } 216 217 int 218 g_vfs_open(struct vnode *vp, struct g_consumer **cpp, const char *fsname, int wr) 219 { 220 struct g_geom *gp; 221 struct g_provider *pp; 222 struct g_consumer *cp; 223 struct g_vfs_softc *sc; 224 struct bufobj *bo; 225 int vfslocked; 226 int error; 227 228 g_topology_assert(); 229 230 *cpp = NULL; 231 bo = &vp->v_bufobj; 232 if (bo->bo_private != vp) 233 return (EBUSY); 234 235 pp = g_dev_getprovider(vp->v_rdev); 236 if (pp == NULL) 237 return (ENOENT); 238 gp = g_new_geomf(&g_vfs_class, "%s.%s", fsname, pp->name); 239 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); 240 mtx_init(&sc->sc_mtx, "g_vfs", NULL, MTX_DEF); 241 sc->sc_bo = bo; 242 gp->softc = sc; 243 cp = g_new_consumer(gp); 244 g_attach(cp, pp); 245 error = g_access(cp, 1, wr, wr); 246 if (error) { 247 g_wither_geom(gp, ENXIO); 248 return (error); 249 } 250 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 251 vnode_create_vobject(vp, pp->mediasize, curthread); 252 VFS_UNLOCK_GIANT(vfslocked); 253 *cpp = cp; 254 cp->private = vp; 255 bo->bo_ops = g_vfs_bufops; 256 bo->bo_private = cp; 257 bo->bo_bsize = pp->sectorsize; 258 259 return (error); 260 } 261 262 void 263 g_vfs_close(struct g_consumer *cp) 264 { 265 struct g_geom *gp; 266 struct g_vfs_softc *sc; 267 268 g_topology_assert(); 269 270 gp = cp->geom; 271 sc = gp->softc; 272 bufobj_invalbuf(sc->sc_bo, V_SAVE, 0, 0); 273 sc->sc_bo->bo_private = cp->private; 274 gp->softc = NULL; 275 mtx_destroy(&sc->sc_mtx); 276 if (!sc->sc_orphaned || cp->provider == NULL) 277 g_wither_geom_close(gp, ENXIO); 278 g_free(sc); 279 } 280