xref: /freebsd/sys/geom/geom_disk.c (revision 2546665afcaf0d53dc2c7058fee96354b3680f5a)
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 
65 struct g_class g_disk_class = {
66 	.name = "DISK",
67 	.init = g_disk_init,
68 	.fini = g_disk_fini,
69 };
70 
71 static void
72 g_disk_init(struct g_class *mp __unused)
73 {
74 
75 	mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF);
76 }
77 
78 static void
79 g_disk_fini(struct g_class *mp __unused)
80 {
81 
82 	mtx_destroy(&g_disk_done_mtx);
83 }
84 
85 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
86 
87 static void __inline
88 g_disk_lock_giant(struct disk *dp)
89 {
90 	if (dp->d_flags & DISKFLAG_NEEDSGIANT)
91 		mtx_lock(&Giant);
92 }
93 
94 static void __inline
95 g_disk_unlock_giant(struct disk *dp)
96 {
97 	if (dp->d_flags & DISKFLAG_NEEDSGIANT)
98 		mtx_unlock(&Giant);
99 }
100 
101 static int
102 g_disk_access(struct g_provider *pp, int r, int w, int e)
103 {
104 	struct disk *dp;
105 	int error;
106 
107 	g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
108 	    pp->name, r, w, e);
109 	g_topology_assert();
110 	dp = pp->geom->softc;
111 	if (dp == NULL || dp->d_destroyed) {
112 		/*
113 		 * Allow decreasing access count even if disk is not
114 		 * avaliable anymore.
115 		 */
116 		if (r <= 0 && w <= 0 && e <= 0)
117 			return (0);
118 		return (ENXIO);
119 	}
120 	r += pp->acr;
121 	w += pp->acw;
122 	e += pp->ace;
123 	error = 0;
124 	if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
125 		if (dp->d_open != NULL) {
126 			g_disk_lock_giant(dp);
127 			error = dp->d_open(dp);
128 			if (error != 0)
129 				printf("Opened disk %s -> %d\n",
130 				    pp->name, error);
131 			g_disk_unlock_giant(dp);
132 		}
133 		pp->mediasize = dp->d_mediasize;
134 		pp->sectorsize = dp->d_sectorsize;
135 		dp->d_flags |= DISKFLAG_OPEN;
136 		if (dp->d_maxsize == 0) {
137 			printf("WARNING: Disk drive %s%d has no d_maxsize\n",
138 			    dp->d_name, dp->d_unit);
139 			dp->d_maxsize = DFLTPHYS;
140 		}
141 	} else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
142 		if (dp->d_close != NULL) {
143 			g_disk_lock_giant(dp);
144 			error = dp->d_close(dp);
145 			if (error != 0)
146 				printf("Closed disk %s -> %d\n",
147 				    pp->name, error);
148 			g_disk_unlock_giant(dp);
149 		}
150 		dp->d_flags &= ~DISKFLAG_OPEN;
151 	}
152 	return (error);
153 }
154 
155 static void
156 g_disk_kerneldump(struct bio *bp, struct disk *dp)
157 {
158 	int error;
159 	struct g_kerneldump *gkd;
160 	struct dumperinfo di;
161 	struct g_geom *gp;
162 
163 	gkd = (struct g_kerneldump*)bp->bio_data;
164 	gp = bp->bio_to->geom;
165 	g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
166 		gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
167 	if (dp->d_dump == NULL) {
168 		g_io_deliver(bp, ENODEV);
169 		return;
170 	}
171 	di.dumper = dp->d_dump;
172 	di.priv = dp;
173 	di.blocksize = dp->d_sectorsize;
174 	di.mediaoffset = gkd->offset;
175 	di.mediasize = gkd->length;
176 	error = set_dumper(&di);
177 	g_io_deliver(bp, error);
178 }
179 
180 static void
181 g_disk_done(struct bio *bp)
182 {
183 	struct bio *bp2;
184 	struct disk *dp;
185 
186 	/* See "notes" for why we need a mutex here */
187 	/* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
188 	mtx_lock(&g_disk_done_mtx);
189 	bp->bio_completed = bp->bio_length - bp->bio_resid;
190 
191 	bp2 = bp->bio_parent;
192 	if (bp2->bio_error == 0)
193 		bp2->bio_error = bp->bio_error;
194 	bp2->bio_completed += bp->bio_completed;
195 	g_destroy_bio(bp);
196 	bp2->bio_inbed++;
197 	if (bp2->bio_children == bp2->bio_inbed) {
198 		bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
199 		if ((dp = bp2->bio_to->geom->softc))
200 			devstat_end_transaction_bio(dp->d_devstat, bp2);
201 		g_io_deliver(bp2, bp2->bio_error);
202 	}
203 	mtx_unlock(&g_disk_done_mtx);
204 }
205 
206 static int
207 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, struct thread *td)
208 {
209 	struct g_geom *gp;
210 	struct disk *dp;
211 	int error;
212 
213 	gp = pp->geom;
214 	dp = gp->softc;
215 
216 	if (dp->d_ioctl == NULL)
217 		return (ENOIOCTL);
218 	g_disk_lock_giant(dp);
219 	error = dp->d_ioctl(dp, cmd, data, 0, td);
220 	g_disk_unlock_giant(dp);
221 	return(error);
222 }
223 
224 static void
225 g_disk_start(struct bio *bp)
226 {
227 	struct bio *bp2, *bp3;
228 	struct disk *dp;
229 	int error;
230 	off_t off;
231 
232 	dp = bp->bio_to->geom->softc;
233 	if (dp == NULL || dp->d_destroyed)
234 		g_io_deliver(bp, ENXIO);
235 	error = EJUSTRETURN;
236 	switch(bp->bio_cmd) {
237 	case BIO_DELETE:
238 		if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
239 			error = 0;
240 			break;
241 		}
242 		/* fall-through */
243 	case BIO_READ:
244 	case BIO_WRITE:
245 		off = 0;
246 		bp3 = NULL;
247 		bp2 = g_clone_bio(bp);
248 		if (bp2 == NULL) {
249 			error = ENOMEM;
250 			break;
251 		}
252 		devstat_start_transaction_bio(dp->d_devstat, bp);
253 		do {
254 			bp2->bio_offset += off;
255 			bp2->bio_length -= off;
256 			bp2->bio_data += off;
257 			if (bp2->bio_length > dp->d_maxsize) {
258 				/*
259 				 * XXX: If we have a stripesize we should really
260 				 * use it here.
261 				 */
262 				bp2->bio_length = dp->d_maxsize;
263 				off += dp->d_maxsize;
264 				/*
265 				 * To avoid a race, we need to grab the next bio
266 				 * before we schedule this one.  See "notes".
267 				 */
268 				bp3 = g_clone_bio(bp);
269 				if (bp3 == NULL)
270 					bp->bio_error = ENOMEM;
271 			}
272 			bp2->bio_done = g_disk_done;
273 			bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
274 			bp2->bio_bcount = bp2->bio_length;
275 			bp2->bio_disk = dp;
276 			g_disk_lock_giant(dp);
277 			dp->d_strategy(bp2);
278 			g_disk_unlock_giant(dp);
279 			bp2 = bp3;
280 			bp3 = NULL;
281 		} while (bp2 != NULL);
282 		break;
283 	case BIO_GETATTR:
284 		if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors))
285 			break;
286 		else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
287 			break;
288 		else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
289 			break;
290 		else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
291 			g_disk_kerneldump(bp, dp);
292 		else
293 			error = ENOIOCTL;
294 		break;
295 	default:
296 		error = EOPNOTSUPP;
297 		break;
298 	}
299 	if (error != EJUSTRETURN)
300 		g_io_deliver(bp, error);
301 	return;
302 }
303 
304 static void
305 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
306 {
307 	struct disk *dp;
308 
309 	dp = gp->softc;
310 	if (dp == NULL)
311 		return;
312 	if (indent == NULL) {
313 		sbuf_printf(sb, " hd %u", dp->d_fwheads);
314 		sbuf_printf(sb, " sc %u", dp->d_fwsectors);
315 		return;
316 	}
317 	if (pp != NULL) {
318 		sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
319 		    indent, dp->d_fwheads);
320 		sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
321 		    indent, dp->d_fwsectors);
322 	}
323 }
324 
325 static void
326 g_disk_create(void *arg, int flag)
327 {
328 	struct g_geom *gp;
329 	struct g_provider *pp;
330 	struct disk *dp;
331 
332 	if (flag == EV_CANCEL)
333 		return;
334 	g_topology_assert();
335 	dp = arg;
336 	gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
337 	gp->start = g_disk_start;
338 	gp->access = g_disk_access;
339 	gp->ioctl = g_disk_ioctl;
340 	gp->softc = dp;
341 	gp->dumpconf = g_disk_dumpconf;
342 	pp = g_new_providerf(gp, "%s", gp->name);
343 	pp->mediasize = dp->d_mediasize;
344 	pp->sectorsize = dp->d_sectorsize;
345 	if (dp->d_flags & DISKFLAG_CANDELETE)
346 		pp->flags |= G_PF_CANDELETE;
347 	pp->stripeoffset = dp->d_stripeoffset;
348 	pp->stripesize = dp->d_stripesize;
349 	if (bootverbose)
350 		printf("GEOM: new disk %s\n", gp->name);
351 	dp->d_geom = gp;
352 	g_error_provider(pp, 0);
353 }
354 
355 static void
356 g_disk_destroy(void *ptr, int flag)
357 {
358 	struct disk *dp;
359 	struct g_geom *gp;
360 
361 	g_topology_assert();
362 	dp = ptr;
363 	gp = dp->d_geom;
364 	gp->softc = NULL;
365 	g_wither_geom(gp, ENXIO);
366 	g_free(dp);
367 }
368 
369 struct disk *
370 disk_alloc()
371 {
372 	struct disk *dp;
373 
374 	dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO);
375 	return (dp);
376 }
377 
378 void
379 disk_create(struct disk *dp, int version)
380 {
381 	if (version != DISK_VERSION_00) {
382 		printf("WARNING: Attempt to add disk %s%d %s",
383 		    dp->d_name, dp->d_unit,
384 		    " using incompatible ABI version of disk(9)\n");
385 		printf("WARNING: Ignoring disk %s%d\n",
386 		    dp->d_name, dp->d_unit);
387 		return;
388 	}
389 	KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
390 	KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
391 	KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
392 	KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
393 	if (dp->d_devstat == NULL)
394 		dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
395 		    dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
396 		    DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
397 	dp->d_geom = NULL;
398 	g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
399 }
400 
401 void
402 disk_destroy(struct disk *dp)
403 {
404 
405 	g_cancel_event(dp);
406 	dp->d_destroyed = 1;
407 	if (dp->d_devstat != NULL)
408 		devstat_remove_entry(dp->d_devstat);
409 	g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
410 }
411 
412 static void
413 g_kern_disks(void *p, int flag __unused)
414 {
415 	struct sbuf *sb;
416 	struct g_geom *gp;
417 	char *sp;
418 
419 	sb = p;
420 	sp = "";
421 	g_topology_assert();
422 	LIST_FOREACH(gp, &g_disk_class.geom, geom) {
423 		sbuf_printf(sb, "%s%s", sp, gp->name);
424 		sp = " ";
425 	}
426 	sbuf_finish(sb);
427 }
428 
429 static int
430 sysctl_disks(SYSCTL_HANDLER_ARGS)
431 {
432 	int error;
433 	struct sbuf *sb;
434 
435 	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
436 	g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
437 	error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
438 	sbuf_delete(sb);
439 	return error;
440 }
441 
442 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0,
443     sysctl_disks, "A", "names of available disks");
444 
445