geom_disk.c revision 125651 
1/*-
2 * Copyright (c) 2002 Poul-Henning Kamp
3 * Copyright (c) 2002 Networks Associates Technology, Inc.
4 * All rights reserved.
5 *
6 * This software was developed for the FreeBSD Project by Poul-Henning Kamp
7 * and NAI Labs, the Security Research Division of Network Associates, Inc.
8 * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
9 * DARPA CHATS research program.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 *    notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 *    notice, this list of conditions and the following disclaimer in the
18 *    documentation and/or other materials provided with the distribution.
19 * 3. The names of the authors may not be used to endorse or promote
20 *    products derived from this software without specific prior written
21 *    permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 */
35
36#include <sys/cdefs.h>
37__FBSDID("$FreeBSD: head/sys/geom/geom_disk.c 125651 2004-02-10 10:54:19Z phk $");
38
39#include "opt_geom.h"
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/kernel.h>
44#include <sys/sysctl.h>
45#include <sys/bio.h>
46#include <sys/conf.h>
47#include <sys/fcntl.h>
48#include <sys/malloc.h>
49#include <sys/sysctl.h>
50#include <sys/devicestat.h>
51#include <machine/md_var.h>
52
53#include <sys/lock.h>
54#include <sys/mutex.h>
55#include <geom/geom.h>
56#include <geom/geom_disk.h>
57#include <geom/geom_int.h>
58
59static struct mtx g_disk_done_mtx;
60
61static g_access_t g_disk_access;
62static g_init_t g_disk_init;
63static g_fini_t g_disk_fini;
64
65struct g_class g_disk_class = {
66	.name = "DISK",
67	.init = g_disk_init,
68	.fini = g_disk_fini,
69};
70
71static void
72g_disk_init(struct g_class *mp __unused)
73{
74
75	mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF);
76}
77
78static void
79g_disk_fini(struct g_class *mp __unused)
80{
81
82	mtx_destroy(&g_disk_done_mtx);
83}
84
85DECLARE_GEOM_CLASS(g_disk_class, g_disk);
86
87static void __inline
88g_disk_lock_giant(struct disk *dp)
89{
90	if (dp->d_flags & DISKFLAG_NOGIANT)
91		return;
92	mtx_lock(&Giant);
93}
94
95static void __inline
96g_disk_unlock_giant(struct disk *dp)
97{
98	if (dp->d_flags & DISKFLAG_NOGIANT)
99		return;
100	mtx_unlock(&Giant);
101}
102
103static int
104g_disk_access(struct g_provider *pp, int r, int w, int e)
105{
106	struct disk *dp;
107	int error;
108
109	g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
110	    pp->name, r, w, e);
111	g_topology_assert();
112	dp = pp->geom->softc;
113	if (dp == NULL) {
114		/*
115		 * Allow decreasing access count even if disk is not
116		 * avaliable anymore.
117		 */
118		if (r <= 0 && w <= 0 && e <= 0)
119			return (0);
120		else
121			return (ENXIO);
122	}
123	r += pp->acr;
124	w += pp->acw;
125	e += pp->ace;
126	error = 0;
127	if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
128		if (dp->d_open != NULL) {
129			g_disk_lock_giant(dp);
130			error = dp->d_open(dp);
131			if (error != 0)
132				printf("Opened disk %s -> %d\n",
133				    pp->name, error);
134			g_disk_unlock_giant(dp);
135		}
136		pp->mediasize = dp->d_mediasize;
137		pp->sectorsize = dp->d_sectorsize;
138		dp->d_flags |= DISKFLAG_OPEN;
139		if (dp->d_maxsize == 0) {
140			printf("WARNING: Disk drive %s%d has no d_maxsize\n",
141			    dp->d_name, dp->d_unit);
142			dp->d_maxsize = DFLTPHYS;
143		}
144	} else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
145		if (dp->d_close != NULL) {
146			g_disk_lock_giant(dp);
147			error = dp->d_close(dp);
148			if (error != 0)
149				printf("Closed disk %s -> %d\n",
150				    pp->name, error);
151			g_disk_unlock_giant(dp);
152		}
153		dp->d_flags &= ~DISKFLAG_OPEN;
154	}
155	return (error);
156}
157
158static void
159g_disk_kerneldump(struct bio *bp, struct disk *dp)
160{
161	int error;
162	struct g_kerneldump *gkd;
163	struct dumperinfo di;
164	struct g_geom *gp;
165
166	gkd = (struct g_kerneldump*)bp->bio_data;
167	gp = bp->bio_to->geom;
168	g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
169		gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
170	if (dp->d_dump == NULL) {
171		g_io_deliver(bp, ENODEV);
172		return;
173	}
174	di.dumper = dp->d_dump;
175	di.priv = dp;
176	di.blocksize = dp->d_sectorsize;
177	di.mediaoffset = gkd->offset;
178	di.mediasize = gkd->length;
179	error = set_dumper(&di);
180	g_io_deliver(bp, error);
181}
182
183static void
184g_disk_done(struct bio *bp)
185{
186	struct bio *bp2;
187	struct disk *dp;
188
189	/* See "notes" for why we need a mutex here */
190	/* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
191	mtx_lock(&g_disk_done_mtx);
192	bp->bio_completed = bp->bio_length - bp->bio_resid;
193
194	bp2 = bp->bio_parent;
195	dp = bp2->bio_to->geom->softc;
196	if (bp2->bio_error == 0)
197		bp2->bio_error = bp->bio_error;
198	bp2->bio_completed += bp->bio_completed;
199	g_destroy_bio(bp);
200	bp2->bio_inbed++;
201	if (bp2->bio_children == bp2->bio_inbed) {
202		bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
203		devstat_end_transaction_bio(dp->d_devstat, bp2);
204		g_io_deliver(bp2, bp2->bio_error);
205	}
206	mtx_unlock(&g_disk_done_mtx);
207}
208
209static int
210g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, struct thread *td)
211{
212	struct g_geom *gp;
213	struct disk *dp;
214	int error;
215
216	gp = pp->geom;
217	dp = gp->softc;
218
219	if (dp->d_ioctl == NULL)
220		return (ENOIOCTL);
221	g_disk_lock_giant(dp);
222	error = dp->d_ioctl(dp, cmd, data, 0, td);
223	g_disk_unlock_giant(dp);
224	return(error);
225}
226
227static void
228g_disk_start(struct bio *bp)
229{
230	struct bio *bp2, *bp3;
231	struct disk *dp;
232	int error;
233	off_t off;
234
235	dp = bp->bio_to->geom->softc;
236	if (dp == NULL)
237		g_io_deliver(bp, ENXIO);
238	error = EJUSTRETURN;
239	switch(bp->bio_cmd) {
240	case BIO_DELETE:
241		if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
242			error = 0;
243			break;
244		}
245		/* fall-through */
246	case BIO_READ:
247	case BIO_WRITE:
248		off = 0;
249		bp3 = NULL;
250		bp2 = g_clone_bio(bp);
251		if (bp2 == NULL) {
252			error = ENOMEM;
253			break;
254		}
255		devstat_start_transaction_bio(dp->d_devstat, bp);
256		do {
257			bp2->bio_offset += off;
258			bp2->bio_length -= off;
259			bp2->bio_data += off;
260			if (bp2->bio_length > dp->d_maxsize) {
261				/*
262				 * XXX: If we have a stripesize we should really
263				 * use it here.
264				 */
265				bp2->bio_length = dp->d_maxsize;
266				off += dp->d_maxsize;
267				/*
268				 * To avoid a race, we need to grab the next bio
269				 * before we schedule this one.  See "notes".
270				 */
271				bp3 = g_clone_bio(bp);
272				if (bp3 == NULL)
273					bp->bio_error = ENOMEM;
274			}
275			bp2->bio_done = g_disk_done;
276			bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
277			bp2->bio_bcount = bp2->bio_length;
278			bp2->bio_disk = dp;
279			g_disk_lock_giant(dp);
280			dp->d_strategy(bp2);
281			g_disk_unlock_giant(dp);
282			bp2 = bp3;
283			bp3 = NULL;
284		} while (bp2 != NULL);
285		break;
286	case BIO_GETATTR:
287		if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors))
288			break;
289		else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
290			break;
291		else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
292			break;
293		else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
294			g_disk_kerneldump(bp, dp);
295		else
296			error = ENOIOCTL;
297		break;
298	default:
299		error = EOPNOTSUPP;
300		break;
301	}
302	if (error != EJUSTRETURN)
303		g_io_deliver(bp, error);
304	return;
305}
306
307static void
308g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
309{
310	struct disk *dp;
311
312	dp = gp->softc;
313	if (dp == NULL)
314		return;
315	if (indent == NULL) {
316		sbuf_printf(sb, " hd %u", dp->d_fwheads);
317		sbuf_printf(sb, " sc %u", dp->d_fwsectors);
318		return;
319	}
320	if (pp != NULL) {
321		sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
322		    indent, dp->d_fwheads);
323		sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
324		    indent, dp->d_fwsectors);
325	}
326}
327
328static void
329g_disk_create(void *arg, int flag)
330{
331	struct g_geom *gp;
332	struct g_provider *pp;
333	struct disk *dp;
334
335	if (flag == EV_CANCEL)
336		return;
337	g_topology_assert();
338	dp = arg;
339	gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
340	gp->start = g_disk_start;
341	gp->access = g_disk_access;
342	gp->ioctl = g_disk_ioctl;
343	gp->softc = dp;
344	gp->dumpconf = g_disk_dumpconf;
345	pp = g_new_providerf(gp, "%s", gp->name);
346	pp->mediasize = dp->d_mediasize;
347	pp->sectorsize = dp->d_sectorsize;
348	if (dp->d_flags & DISKFLAG_CANDELETE)
349		pp->flags |= G_PF_CANDELETE;
350	pp->stripeoffset = dp->d_stripeoffset;
351	pp->stripesize = dp->d_stripesize;
352	if (bootverbose)
353		printf("GEOM: new disk %s\n", gp->name);
354	dp->d_geom = gp;
355	g_error_provider(pp, 0);
356}
357
358static void
359g_disk_destroy(void *ptr, int flag)
360{
361	struct g_geom *gp;
362
363	g_topology_assert();
364	gp = ptr;
365	gp->softc = NULL;
366	g_wither_geom(gp, ENXIO);
367}
368
369void
370disk_create(int unit, struct disk *dp, int flags, void *unused __unused, void * unused2 __unused)
371{
372
373	dp->d_unit = unit;
374	dp->d_flags = flags;
375	KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
376	KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
377	KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
378	KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
379	if (dp->d_devstat == NULL)
380		dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
381		    dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
382		    DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
383	dp->d_geom = NULL;
384	g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
385}
386
387void
388disk_destroy(struct disk *dp)
389{
390	struct g_geom *gp;
391
392	g_cancel_event(dp);
393	gp = dp->d_geom;
394	if (gp == NULL)
395		return;
396	gp->softc = NULL;
397	devstat_remove_entry(dp->d_devstat);
398	g_waitfor_event(g_disk_destroy, gp, M_WAITOK, NULL, NULL);
399	dp->d_geom = NULL;
400}
401
402static void
403g_kern_disks(void *p, int flag __unused)
404{
405	struct sbuf *sb;
406	struct g_geom *gp;
407	char *sp;
408
409	sb = p;
410	sp = "";
411	g_topology_assert();
412	LIST_FOREACH(gp, &g_disk_class.geom, geom) {
413		sbuf_printf(sb, "%s%s", sp, gp->name);
414		sp = " ";
415	}
416	sbuf_finish(sb);
417}
418
419static int
420sysctl_disks(SYSCTL_HANDLER_ARGS)
421{
422	int error;
423	struct sbuf *sb;
424
425	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
426	g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
427	error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
428	sbuf_delete(sb);
429	return error;
430}
431
432SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0,
433    sysctl_disks, "A", "names of available disks");
434
435