xref: /freebsd/sys/geom/geom_vfs.c (revision 77013d11e6483b970af25e13c9b892075742f7e5)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2004 Poul-Henning Kamp
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/bio.h>
35 #include <sys/kernel.h>
36 #include <sys/lock.h>
37 #include <sys/malloc.h>
38 #include <sys/mutex.h>
39 #include <sys/sbuf.h>
40 #include <sys/vnode.h>
41 #include <sys/mount.h>
42 
43 #include <geom/geom.h>
44 #include <geom/geom_vfs.h>
45 
46 /*
47  * subroutines for use by filesystems.
48  *
49  * XXX: should maybe live somewhere else ?
50  */
51 #include <sys/buf.h>
52 
53 struct g_vfs_softc {
54 	struct mtx	 sc_mtx;
55 	struct bufobj	*sc_bo;
56 	struct g_event	*sc_event;
57 	int		 sc_active;
58 	int		 sc_orphaned;
59 	int		 sc_enxio_active;
60 };
61 
62 static struct buf_ops __g_vfs_bufops = {
63 	.bop_name =	"GEOM_VFS",
64 	.bop_write =	bufwrite,
65 	.bop_strategy =	g_vfs_strategy,
66 	.bop_sync =	bufsync,
67 	.bop_bdflush =	bufbdflush
68 };
69 
70 struct buf_ops *g_vfs_bufops = &__g_vfs_bufops;
71 
72 static g_orphan_t g_vfs_orphan;
73 
74 static struct g_class g_vfs_class = {
75 	.name =		"VFS",
76 	.version =	G_VERSION,
77 	.orphan =	g_vfs_orphan,
78 };
79 
80 DECLARE_GEOM_CLASS(g_vfs_class, g_vfs);
81 
82 static void
83 g_vfs_destroy(void *arg, int flags __unused)
84 {
85 	struct g_consumer *cp;
86 
87 	g_topology_assert();
88 	cp = arg;
89 	if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0)
90 		g_access(cp, -cp->acr, -cp->acw, -cp->ace);
91 	g_detach(cp);
92 	if (cp->geom->softc == NULL)
93 		g_wither_geom(cp->geom, ENXIO);
94 }
95 
96 static void
97 g_vfs_done(struct bio *bip)
98 {
99 	struct g_consumer *cp;
100 	struct g_event *event;
101 	struct g_vfs_softc *sc;
102 	struct buf *bp;
103 	int destroy;
104 	struct mount *mp;
105 	struct vnode *vp;
106 	struct cdev *cdevp;
107 
108 	/*
109 	 * Collect statistics on synchronous and asynchronous read
110 	 * and write counts for disks that have associated filesystems.
111 	 */
112 	bp = bip->bio_caller2;
113 	vp = bp->b_vp;
114 	if (vp != NULL) {
115 		/*
116 		 * If not a disk vnode, use its associated mount point
117 		 * otherwise use the mountpoint associated with the disk.
118 		 */
119 		VI_LOCK(vp);
120 		if (vp->v_type != VCHR ||
121 		    (cdevp = vp->v_rdev) == NULL ||
122 		    cdevp->si_devsw == NULL ||
123 		    (cdevp->si_devsw->d_flags & D_DISK) == 0)
124 			mp = vp->v_mount;
125 		else
126 			mp = cdevp->si_mountpt;
127 		if (mp != NULL) {
128 			if (bp->b_iocmd == BIO_READ) {
129 				if (LK_HOLDER(bp->b_lock.lk_lock) == LK_KERNPROC)
130 					mp->mnt_stat.f_asyncreads++;
131 				else
132 					mp->mnt_stat.f_syncreads++;
133 			} else if (bp->b_iocmd == BIO_WRITE) {
134 				if (LK_HOLDER(bp->b_lock.lk_lock) == LK_KERNPROC)
135 					mp->mnt_stat.f_asyncwrites++;
136 				else
137 					mp->mnt_stat.f_syncwrites++;
138 			}
139 		}
140 		VI_UNLOCK(vp);
141 	}
142 
143 	cp = bip->bio_from;
144 	sc = cp->geom->softc;
145 	if (bip->bio_error != 0 && bip->bio_error != EOPNOTSUPP) {
146 		if ((bp->b_xflags & BX_CVTENXIO) != 0)
147 			sc->sc_enxio_active = 1;
148 		if (sc->sc_enxio_active)
149 			bip->bio_error = ENXIO;
150 		g_print_bio("g_vfs_done():", bip, "error = %d",
151 		    bip->bio_error);
152 	}
153 	bp->b_error = bip->bio_error;
154 	bp->b_ioflags = bip->bio_flags;
155 	if (bip->bio_error)
156 		bp->b_ioflags |= BIO_ERROR;
157 	bp->b_resid = bp->b_bcount - bip->bio_completed;
158 	g_destroy_bio(bip);
159 
160 	mtx_lock(&sc->sc_mtx);
161 	destroy = ((--sc->sc_active) == 0 && sc->sc_orphaned);
162 	if (destroy) {
163 		event = sc->sc_event;
164 		sc->sc_event = NULL;
165 	} else
166 		event = NULL;
167 	mtx_unlock(&sc->sc_mtx);
168 	if (destroy)
169 		g_post_event_ep(g_vfs_destroy, cp, event, NULL);
170 
171 	bufdone(bp);
172 }
173 
174 void
175 g_vfs_strategy(struct bufobj *bo, struct buf *bp)
176 {
177 	struct g_vfs_softc *sc;
178 	struct g_consumer *cp;
179 	struct bio *bip;
180 
181 	cp = bo->bo_private;
182 	sc = cp->geom->softc;
183 
184 	/*
185 	 * If the provider has orphaned us, just return ENXIO.
186 	 */
187 	mtx_lock(&sc->sc_mtx);
188 	if (sc->sc_orphaned || sc->sc_enxio_active) {
189 		mtx_unlock(&sc->sc_mtx);
190 		bp->b_error = ENXIO;
191 		bp->b_ioflags |= BIO_ERROR;
192 		bufdone(bp);
193 		return;
194 	}
195 	sc->sc_active++;
196 	mtx_unlock(&sc->sc_mtx);
197 
198 	bip = g_alloc_bio();
199 	bip->bio_cmd = bp->b_iocmd;
200 	bip->bio_offset = bp->b_iooffset;
201 	bip->bio_length = bp->b_bcount;
202 	bdata2bio(bp, bip);
203 	if ((bp->b_flags & B_BARRIER) != 0) {
204 		bip->bio_flags |= BIO_ORDERED;
205 		bp->b_flags &= ~B_BARRIER;
206 	}
207 	if (bp->b_iocmd == BIO_SPEEDUP)
208 		bip->bio_flags |= bp->b_ioflags;
209 	bip->bio_done = g_vfs_done;
210 	bip->bio_caller2 = bp;
211 #if defined(BUF_TRACKING) || defined(FULL_BUF_TRACKING)
212 	buf_track(bp, __func__);
213 	bip->bio_track_bp = bp;
214 #endif
215 	g_io_request(bip, cp);
216 }
217 
218 static void
219 g_vfs_orphan(struct g_consumer *cp)
220 {
221 	struct g_geom *gp;
222 	struct g_event *event;
223 	struct g_vfs_softc *sc;
224 	int destroy;
225 
226 	g_topology_assert();
227 
228 	gp = cp->geom;
229 	g_trace(G_T_TOPOLOGY, "g_vfs_orphan(%p(%s))", cp, gp->name);
230 	sc = gp->softc;
231 	if (sc == NULL)
232 		return;
233 	event = g_alloc_event(M_WAITOK);
234 	mtx_lock(&sc->sc_mtx);
235 	KASSERT(sc->sc_event == NULL, ("g_vfs %p already has an event", sc));
236 	sc->sc_orphaned = 1;
237 	destroy = (sc->sc_active == 0);
238 	if (!destroy) {
239 		sc->sc_event = event;
240 		event = NULL;
241 	}
242 	mtx_unlock(&sc->sc_mtx);
243 	if (destroy) {
244 		g_free(event);
245 		g_vfs_destroy(cp, 0);
246 	}
247 
248 	/*
249 	 * Do not destroy the geom.  Filesystem will do that during unmount.
250 	 */
251 }
252 
253 int
254 g_vfs_open(struct vnode *vp, struct g_consumer **cpp, const char *fsname, int wr)
255 {
256 	struct g_geom *gp;
257 	struct g_provider *pp;
258 	struct g_consumer *cp;
259 	struct g_vfs_softc *sc;
260 	struct bufobj *bo;
261 	int error;
262 
263 	g_topology_assert();
264 
265 	*cpp = NULL;
266 	bo = &vp->v_bufobj;
267 	if (bo->bo_private != vp)
268 		return (EBUSY);
269 
270 	pp = g_dev_getprovider(vp->v_rdev);
271 	if (pp == NULL)
272 		return (ENOENT);
273 	gp = g_new_geomf(&g_vfs_class, "%s.%s", fsname, pp->name);
274 	sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
275 	mtx_init(&sc->sc_mtx, "g_vfs", NULL, MTX_DEF);
276 	sc->sc_bo = bo;
277 	gp->softc = sc;
278 	cp = g_new_consumer(gp);
279 	error = g_attach(cp, pp);
280 	if (error) {
281 		g_wither_geom(gp, ENXIO);
282 		return (error);
283 	}
284 	error = g_access(cp, 1, wr, wr);
285 	if (error) {
286 		g_wither_geom(gp, ENXIO);
287 		return (error);
288 	}
289 	vnode_create_vobject(vp, pp->mediasize, curthread);
290 	*cpp = cp;
291 	cp->private = vp;
292 	cp->flags |= G_CF_DIRECT_SEND | G_CF_DIRECT_RECEIVE;
293 	bo->bo_ops = g_vfs_bufops;
294 	bo->bo_private = cp;
295 	bo->bo_bsize = pp->sectorsize;
296 
297 	return (error);
298 }
299 
300 void
301 g_vfs_close(struct g_consumer *cp)
302 {
303 	struct g_geom *gp;
304 	struct g_vfs_softc *sc;
305 
306 	g_topology_assert();
307 
308 	gp = cp->geom;
309 	sc = gp->softc;
310 	bufobj_invalbuf(sc->sc_bo, V_SAVE, 0, 0);
311 	sc->sc_bo->bo_private = cp->private;
312 	gp->softc = NULL;
313 	mtx_destroy(&sc->sc_mtx);
314 	if (!sc->sc_orphaned || cp->provider == NULL)
315 		g_wither_geom_close(gp, ENXIO);
316 	KASSERT(sc->sc_event == NULL, ("g_vfs %p event is non-NULL", sc));
317 	g_free(sc);
318 }
319