xref: /freebsd/sys/geom/geom_disk.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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", MTX_DEF, 0);
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_NOGIANT)
91 		return;
92 	mtx_lock(&Giant);
93 }
94 
95 static void __inline
96 g_disk_unlock_giant(struct disk *dp)
97 {
98 	if (dp->d_flags & DISKFLAG_NOGIANT)
99 		return;
100 	mtx_unlock(&Giant);
101 }
102 
103 static int
104 g_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 	r += pp->acr;
113 	w += pp->acw;
114 	e += pp->ace;
115 	dp = pp->geom->softc;
116 	if (dp == NULL)
117 		return (ENXIO);
118 	error = 0;
119 	if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
120 		if (dp->d_open != NULL) {
121 			g_disk_lock_giant(dp);
122 			error = dp->d_open(dp);
123 			if (error != 0)
124 				printf("Opened disk %s -> %d\n",
125 				    pp->name, error);
126 			g_disk_unlock_giant(dp);
127 		}
128 		pp->mediasize = dp->d_mediasize;
129 		pp->sectorsize = dp->d_sectorsize;
130 		dp->d_flags |= DISKFLAG_OPEN;
131 		if (dp->d_maxsize == 0) {
132 			printf("WARNING: Disk drive %s%d has no d_maxsize\n",
133 			    dp->d_name, dp->d_unit);
134 			dp->d_maxsize = DFLTPHYS;
135 		}
136 	} else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
137 		if (dp->d_close != NULL) {
138 			g_disk_lock_giant(dp);
139 			error = dp->d_close(dp);
140 			if (error != 0)
141 				printf("Closed disk %s -> %d\n",
142 				    pp->name, error);
143 			g_disk_unlock_giant(dp);
144 		}
145 		dp->d_flags &= ~DISKFLAG_OPEN;
146 	}
147 	return (error);
148 }
149 
150 static void
151 g_disk_kerneldump(struct bio *bp, struct disk *dp)
152 {
153 	int error;
154 	struct g_kerneldump *gkd;
155 	struct dumperinfo di;
156 	struct g_geom *gp;
157 
158 	gkd = (struct g_kerneldump*)bp->bio_data;
159 	gp = bp->bio_to->geom;
160 	g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
161 		gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
162 	di.dumper = dp->d_dump;
163 	di.priv = dp;
164 	di.blocksize = dp->d_sectorsize;
165 	di.mediaoffset = gkd->offset;
166 	di.mediasize = gkd->length;
167 	error = set_dumper(&di);
168 	g_io_deliver(bp, error);
169 }
170 
171 static void
172 g_disk_done(struct bio *bp)
173 {
174 	struct bio *bp2;
175 	struct disk *dp;
176 
177 	/* See "notes" for why we need a mutex here */
178 	/* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
179 	mtx_lock(&g_disk_done_mtx);
180 	bp->bio_completed = bp->bio_length - bp->bio_resid;
181 
182 	bp2 = bp->bio_parent;
183 	dp = bp2->bio_to->geom->softc;
184 	if (bp2->bio_error == 0)
185 		bp2->bio_error = bp->bio_error;
186 	bp2->bio_completed += bp->bio_completed;
187 	g_destroy_bio(bp);
188 	bp2->bio_inbed++;
189 	if (bp2->bio_children == bp2->bio_inbed) {
190 		bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
191 		devstat_end_transaction_bio(dp->d_devstat, bp2);
192 		g_io_deliver(bp2, bp2->bio_error);
193 	}
194 	mtx_unlock(&g_disk_done_mtx);
195 }
196 
197 static int
198 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, struct thread *td)
199 {
200 	struct g_geom *gp;
201 	struct disk *dp;
202 	int error;
203 
204 	gp = pp->geom;
205 	dp = gp->softc;
206 
207 	if (dp->d_ioctl == NULL)
208 		return (ENOIOCTL);
209 	g_disk_lock_giant(dp);
210 	error = dp->d_ioctl(dp, cmd, data, 0, td);
211 	g_disk_unlock_giant(dp);
212 	return(error);
213 }
214 
215 static void
216 g_disk_start(struct bio *bp)
217 {
218 	struct bio *bp2, *bp3;
219 	struct disk *dp;
220 	int error;
221 	off_t off;
222 
223 	dp = bp->bio_to->geom->softc;
224 	if (dp == NULL)
225 		g_io_deliver(bp, ENXIO);
226 	error = EJUSTRETURN;
227 	switch(bp->bio_cmd) {
228 	case BIO_DELETE:
229 		if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
230 			error = 0;
231 			break;
232 		}
233 		/* fall-through */
234 	case BIO_READ:
235 	case BIO_WRITE:
236 		off = 0;
237 		bp3 = NULL;
238 		bp2 = g_clone_bio(bp);
239 		if (bp2 == NULL) {
240 			error = ENOMEM;
241 			break;
242 		}
243 		devstat_start_transaction_bio(dp->d_devstat, bp);
244 		do {
245 			bp2->bio_offset += off;
246 			bp2->bio_length -= off;
247 			bp2->bio_data += off;
248 			if (bp2->bio_length > dp->d_maxsize) {
249 				/*
250 				 * XXX: If we have a stripesize we should really
251 				 * use it here.
252 				 */
253 				bp2->bio_length = dp->d_maxsize;
254 				off += dp->d_maxsize;
255 				/*
256 				 * To avoid a race, we need to grab the next bio
257 				 * before we schedule this one.  See "notes".
258 				 */
259 				bp3 = g_clone_bio(bp);
260 				if (bp3 == NULL)
261 					bp->bio_error = ENOMEM;
262 			}
263 			bp2->bio_done = g_disk_done;
264 			bp2->bio_blkno = bp2->bio_offset >> DEV_BSHIFT;
265 			bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
266 			bp2->bio_bcount = bp2->bio_length;
267 			bp2->bio_disk = dp;
268 			g_disk_lock_giant(dp);
269 			dp->d_strategy(bp2);
270 			g_disk_unlock_giant(dp);
271 			bp2 = bp3;
272 			bp3 = NULL;
273 		} while (bp2 != NULL);
274 		break;
275 	case BIO_GETATTR:
276 		if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors))
277 			break;
278 		else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
279 			break;
280 		else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
281 			break;
282 		else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
283 			g_disk_kerneldump(bp, dp);
284 		else
285 			error = ENOIOCTL;
286 		break;
287 	default:
288 		error = EOPNOTSUPP;
289 		break;
290 	}
291 	if (error != EJUSTRETURN)
292 		g_io_deliver(bp, error);
293 	return;
294 }
295 
296 static void
297 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
298 {
299 	struct disk *dp;
300 
301 	dp = gp->softc;
302 	if (indent == NULL) {
303 		sbuf_printf(sb, " hd %u", dp->d_fwheads);
304 		sbuf_printf(sb, " sc %u", dp->d_fwsectors);
305 		return;
306 	}
307 	if (pp != NULL) {
308 		sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
309 		    indent, dp->d_fwheads);
310 		sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
311 		    indent, dp->d_fwsectors);
312 	}
313 }
314 
315 static void
316 g_disk_create(void *arg, int flag)
317 {
318 	struct g_geom *gp;
319 	struct g_provider *pp;
320 	struct disk *dp;
321 
322 	if (flag == EV_CANCEL)
323 		return;
324 	g_topology_assert();
325 	dp = arg;
326 	gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
327 	gp->start = g_disk_start;
328 	gp->access = g_disk_access;
329 	gp->ioctl = g_disk_ioctl;
330 	gp->softc = dp;
331 	gp->dumpconf = g_disk_dumpconf;
332 	pp = g_new_providerf(gp, "%s", gp->name);
333 	pp->mediasize = dp->d_mediasize;
334 	pp->sectorsize = dp->d_sectorsize;
335 	if (dp->d_flags & DISKFLAG_CANDELETE)
336 		pp->flags |= G_PF_CANDELETE;
337 	pp->stripeoffset = dp->d_stripeoffset;
338 	pp->stripesize = dp->d_stripesize;
339 	if (bootverbose)
340 		printf("GEOM: new disk %s\n", gp->name);
341 	dp->d_geom = gp;
342 	g_error_provider(pp, 0);
343 }
344 
345 static void
346 g_disk_destroy(void *ptr, int flag)
347 {
348 	struct g_geom *gp;
349 
350 	g_topology_assert();
351 	gp = ptr;
352 	gp->softc = NULL;
353 	g_wither_geom(gp, ENXIO);
354 }
355 
356 void
357 disk_create(int unit, struct disk *dp, int flags, void *unused __unused, void * unused2 __unused)
358 {
359 
360 	dp->d_unit = unit;
361 	dp->d_flags = flags;
362 	KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
363 	KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
364 	KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
365 	KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
366 	if (bootverbose || 1)
367 		printf("GEOM: create disk %s%d dp=%p\n",
368 		    dp->d_name, dp->d_unit, dp);
369 	dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
370 	    dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
371 	    DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
372 	dp->d_geom = NULL;
373 	g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
374 }
375 
376 void
377 disk_destroy(struct disk *dp)
378 {
379 	struct g_geom *gp;
380 
381 	if (bootverbose || 1)
382 		printf("GEOM: destroy disk %s%d dp=%p\n",
383 		    dp->d_name, dp->d_unit, dp);
384 	g_cancel_event(dp);
385 	gp = dp->d_geom;
386 	if (gp == NULL)
387 		return;
388 	gp->softc = NULL;
389 	devstat_remove_entry(dp->d_devstat);
390 	g_waitfor_event(g_disk_destroy, gp, M_WAITOK, NULL, NULL);
391 	dp->d_geom = NULL;
392 }
393 
394 static void
395 g_kern_disks(void *p, int flag __unused)
396 {
397 	struct sbuf *sb;
398 	struct g_geom *gp;
399 	char *sp;
400 
401 	sb = p;
402 	sp = "";
403 	g_topology_assert();
404 	LIST_FOREACH(gp, &g_disk_class.geom, geom) {
405 		sbuf_printf(sb, "%s%s", sp, gp->name);
406 		sp = " ";
407 	}
408 	sbuf_finish(sb);
409 }
410 
411 static int
412 sysctl_disks(SYSCTL_HANDLER_ARGS)
413 {
414 	int error;
415 	struct sbuf *sb;
416 
417 	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
418 	sbuf_clear(sb);
419 	g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
420 	error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
421 	sbuf_delete(sb);
422 	return error;
423 }
424 
425 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0,
426     sysctl_disks, "A", "names of available disks");
427 
428