xref: /freebsd/sys/geom/geom_disk.c (revision 7dfd9569a2f0637fb9a48157b1c1bfe5709faee3)
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("$FreeBSD$");
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 
59 static struct mtx g_disk_done_mtx;
60 
61 static g_access_t g_disk_access;
62 static g_init_t g_disk_init;
63 static g_fini_t g_disk_fini;
64 static g_start_t g_disk_start;
65 static g_ioctl_t g_disk_ioctl;
66 static g_dumpconf_t g_disk_dumpconf;
67 
68 static struct g_class g_disk_class = {
69 	.name = "DISK",
70 	.version = G_VERSION,
71 	.init = g_disk_init,
72 	.fini = g_disk_fini,
73 	.start = g_disk_start,
74 	.access = g_disk_access,
75 	.ioctl = g_disk_ioctl,
76 	.dumpconf = g_disk_dumpconf,
77 };
78 
79 static void
80 g_disk_init(struct g_class *mp __unused)
81 {
82 
83 	mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF);
84 }
85 
86 static void
87 g_disk_fini(struct g_class *mp __unused)
88 {
89 
90 	mtx_destroy(&g_disk_done_mtx);
91 }
92 
93 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
94 
95 static void __inline
96 g_disk_lock_giant(struct disk *dp)
97 {
98 	if (dp->d_flags & DISKFLAG_NEEDSGIANT)
99 		mtx_lock(&Giant);
100 }
101 
102 static void __inline
103 g_disk_unlock_giant(struct disk *dp)
104 {
105 	if (dp->d_flags & DISKFLAG_NEEDSGIANT)
106 		mtx_unlock(&Giant);
107 }
108 
109 static int
110 g_disk_access(struct g_provider *pp, int r, int w, int e)
111 {
112 	struct disk *dp;
113 	int error;
114 
115 	g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
116 	    pp->name, r, w, e);
117 	g_topology_assert();
118 	dp = pp->geom->softc;
119 	if (dp == NULL || dp->d_destroyed) {
120 		/*
121 		 * Allow decreasing access count even if disk is not
122 		 * avaliable anymore.
123 		 */
124 		if (r <= 0 && w <= 0 && e <= 0)
125 			return (0);
126 		return (ENXIO);
127 	}
128 	r += pp->acr;
129 	w += pp->acw;
130 	e += pp->ace;
131 	error = 0;
132 	if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
133 		if (dp->d_open != NULL) {
134 			g_disk_lock_giant(dp);
135 			error = dp->d_open(dp);
136 			if (error != 0)
137 				printf("Opened disk %s -> %d\n",
138 				    pp->name, error);
139 			g_disk_unlock_giant(dp);
140 		}
141 		pp->mediasize = dp->d_mediasize;
142 		pp->sectorsize = dp->d_sectorsize;
143 		dp->d_flags |= DISKFLAG_OPEN;
144 		if (dp->d_maxsize == 0) {
145 			printf("WARNING: Disk drive %s%d has no d_maxsize\n",
146 			    dp->d_name, dp->d_unit);
147 			dp->d_maxsize = DFLTPHYS;
148 		}
149 	} else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
150 		if (dp->d_close != NULL) {
151 			g_disk_lock_giant(dp);
152 			error = dp->d_close(dp);
153 			if (error != 0)
154 				printf("Closed disk %s -> %d\n",
155 				    pp->name, error);
156 			g_disk_unlock_giant(dp);
157 		}
158 		dp->d_flags &= ~DISKFLAG_OPEN;
159 	}
160 	return (error);
161 }
162 
163 static void
164 g_disk_kerneldump(struct bio *bp, struct disk *dp)
165 {
166 	int error;
167 	struct g_kerneldump *gkd;
168 	struct dumperinfo di;
169 	struct g_geom *gp;
170 
171 	gkd = (struct g_kerneldump*)bp->bio_data;
172 	gp = bp->bio_to->geom;
173 	g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
174 		gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
175 	if (dp->d_dump == NULL) {
176 		g_io_deliver(bp, ENODEV);
177 		return;
178 	}
179 	di.dumper = dp->d_dump;
180 	di.priv = dp;
181 	di.blocksize = dp->d_sectorsize;
182 	di.mediaoffset = gkd->offset;
183 	if ((gkd->offset + gkd->length) > dp->d_mediasize)
184 		gkd->length = dp->d_mediasize - gkd->offset;
185 	di.mediasize = gkd->length;
186 	error = set_dumper(&di);
187 	g_io_deliver(bp, error);
188 }
189 
190 static void
191 g_disk_done(struct bio *bp)
192 {
193 	struct bio *bp2;
194 	struct disk *dp;
195 
196 	/* See "notes" for why we need a mutex here */
197 	/* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
198 	mtx_lock(&g_disk_done_mtx);
199 	bp->bio_completed = bp->bio_length - bp->bio_resid;
200 
201 	bp2 = bp->bio_parent;
202 	if (bp2->bio_error == 0)
203 		bp2->bio_error = bp->bio_error;
204 	bp2->bio_completed += bp->bio_completed;
205 	if ((dp = bp2->bio_to->geom->softc))
206 		devstat_end_transaction_bio(dp->d_devstat, bp);
207 	g_destroy_bio(bp);
208 	bp2->bio_inbed++;
209 	if (bp2->bio_children == bp2->bio_inbed) {
210 		bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
211 		g_io_deliver(bp2, bp2->bio_error);
212 	}
213 	mtx_unlock(&g_disk_done_mtx);
214 }
215 
216 static int
217 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
218 {
219 	struct g_geom *gp;
220 	struct disk *dp;
221 	int error;
222 
223 	gp = pp->geom;
224 	dp = gp->softc;
225 
226 	if (dp->d_ioctl == NULL)
227 		return (ENOIOCTL);
228 	g_disk_lock_giant(dp);
229 	error = dp->d_ioctl(dp, cmd, data, fflag, td);
230 	g_disk_unlock_giant(dp);
231 	return(error);
232 }
233 
234 static void
235 g_disk_start(struct bio *bp)
236 {
237 	struct bio *bp2, *bp3;
238 	struct disk *dp;
239 	int error;
240 	off_t off;
241 
242 	dp = bp->bio_to->geom->softc;
243 	if (dp == NULL || dp->d_destroyed) {
244 		g_io_deliver(bp, ENXIO);
245 		return;
246 	}
247 	error = EJUSTRETURN;
248 	switch(bp->bio_cmd) {
249 	case BIO_DELETE:
250 		if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
251 			error = 0;
252 			break;
253 		}
254 		/* fall-through */
255 	case BIO_READ:
256 	case BIO_WRITE:
257 		off = 0;
258 		bp3 = NULL;
259 		bp2 = g_clone_bio(bp);
260 		if (bp2 == NULL) {
261 			error = ENOMEM;
262 			break;
263 		}
264 		do {
265 			bp2->bio_offset += off;
266 			bp2->bio_length -= off;
267 			bp2->bio_data += off;
268 			if (bp2->bio_length > dp->d_maxsize) {
269 				/*
270 				 * XXX: If we have a stripesize we should really
271 				 * use it here.
272 				 */
273 				bp2->bio_length = dp->d_maxsize;
274 				off += dp->d_maxsize;
275 				/*
276 				 * To avoid a race, we need to grab the next bio
277 				 * before we schedule this one.  See "notes".
278 				 */
279 				bp3 = g_clone_bio(bp);
280 				if (bp3 == NULL)
281 					bp->bio_error = ENOMEM;
282 			}
283 			bp2->bio_done = g_disk_done;
284 			bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
285 			bp2->bio_bcount = bp2->bio_length;
286 			bp2->bio_disk = dp;
287 			devstat_start_transaction_bio(dp->d_devstat, bp2);
288 			g_disk_lock_giant(dp);
289 			dp->d_strategy(bp2);
290 			g_disk_unlock_giant(dp);
291 			bp2 = bp3;
292 			bp3 = NULL;
293 		} while (bp2 != NULL);
294 		break;
295 	case BIO_GETATTR:
296 		if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors))
297 			break;
298 		else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
299 			break;
300 		else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
301 			break;
302 		else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
303 			g_disk_kerneldump(bp, dp);
304 		else
305 			error = ENOIOCTL;
306 		break;
307 	default:
308 		error = EOPNOTSUPP;
309 		break;
310 	}
311 	if (error != EJUSTRETURN)
312 		g_io_deliver(bp, error);
313 	return;
314 }
315 
316 static void
317 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
318 {
319 	struct disk *dp;
320 
321 	dp = gp->softc;
322 	if (dp == NULL)
323 		return;
324 	if (indent == NULL) {
325 		sbuf_printf(sb, " hd %u", dp->d_fwheads);
326 		sbuf_printf(sb, " sc %u", dp->d_fwsectors);
327 		return;
328 	}
329 	if (pp != NULL) {
330 		sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
331 		    indent, dp->d_fwheads);
332 		sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
333 		    indent, dp->d_fwsectors);
334 	}
335 }
336 
337 static void
338 g_disk_create(void *arg, int flag)
339 {
340 	struct g_geom *gp;
341 	struct g_provider *pp;
342 	struct disk *dp;
343 
344 	if (flag == EV_CANCEL)
345 		return;
346 	g_topology_assert();
347 	dp = arg;
348 	gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
349 	gp->softc = dp;
350 	pp = g_new_providerf(gp, "%s", gp->name);
351 	pp->mediasize = dp->d_mediasize;
352 	pp->sectorsize = dp->d_sectorsize;
353 	if (dp->d_flags & DISKFLAG_CANDELETE)
354 		pp->flags |= G_PF_CANDELETE;
355 	pp->stripeoffset = dp->d_stripeoffset;
356 	pp->stripesize = dp->d_stripesize;
357 	if (bootverbose)
358 		printf("GEOM: new disk %s\n", gp->name);
359 	dp->d_geom = gp;
360 	g_error_provider(pp, 0);
361 }
362 
363 static void
364 g_disk_destroy(void *ptr, int flag)
365 {
366 	struct disk *dp;
367 	struct g_geom *gp;
368 
369 	g_topology_assert();
370 	dp = ptr;
371 	gp = dp->d_geom;
372 	if (gp != NULL) {
373 		gp->softc = NULL;
374 		g_wither_geom(gp, ENXIO);
375 	}
376 	g_free(dp);
377 }
378 
379 struct disk *
380 disk_alloc()
381 {
382 	struct disk *dp;
383 
384 	dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO);
385 	return (dp);
386 }
387 
388 void
389 disk_create(struct disk *dp, int version)
390 {
391 	if (version != DISK_VERSION_00) {
392 		printf("WARNING: Attempt to add disk %s%d %s",
393 		    dp->d_name, dp->d_unit,
394 		    " using incompatible ABI version of disk(9)\n");
395 		printf("WARNING: Ignoring disk %s%d\n",
396 		    dp->d_name, dp->d_unit);
397 		return;
398 	}
399 	KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
400 	KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
401 	KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
402 	KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
403 	if (dp->d_devstat == NULL)
404 		dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
405 		    dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
406 		    DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
407 	dp->d_geom = NULL;
408 	g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
409 }
410 
411 void
412 disk_destroy(struct disk *dp)
413 {
414 
415 	g_cancel_event(dp);
416 	dp->d_destroyed = 1;
417 	if (dp->d_devstat != NULL)
418 		devstat_remove_entry(dp->d_devstat);
419 	g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
420 }
421 
422 void
423 disk_gone(struct disk *dp)
424 {
425 	struct g_geom *gp;
426 	struct g_provider *pp;
427 
428 	gp = dp->d_geom;
429 	if (gp != NULL)
430 		LIST_FOREACH(pp, &gp->provider, provider)
431 			g_wither_provider(pp, ENXIO);
432 }
433 
434 static void
435 g_kern_disks(void *p, int flag __unused)
436 {
437 	struct sbuf *sb;
438 	struct g_geom *gp;
439 	char *sp;
440 
441 	sb = p;
442 	sp = "";
443 	g_topology_assert();
444 	LIST_FOREACH(gp, &g_disk_class.geom, geom) {
445 		sbuf_printf(sb, "%s%s", sp, gp->name);
446 		sp = " ";
447 	}
448 	sbuf_finish(sb);
449 }
450 
451 static int
452 sysctl_disks(SYSCTL_HANDLER_ARGS)
453 {
454 	int error;
455 	struct sbuf *sb;
456 
457 	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
458 	g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
459 	error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
460 	sbuf_delete(sb);
461 	return error;
462 }
463 
464 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0,
465     sysctl_disks, "A", "names of available disks");
466 
467