xref: /dragonfly/sys/dev/disk/dm/dm_pdev.c (revision 67640b13)
1 /*        $NetBSD: dm_pdev.c,v 1.6 2010/01/04 00:19:08 haad Exp $      */
2 
3 /*
4  * Copyright (c) 2010-2011 Alex Hornung <alex@alexhornung.com>
5  * Copyright (c) 2008 The NetBSD Foundation, Inc.
6  * All rights reserved.
7  *
8  * This code is derived from software contributed to The NetBSD Foundation
9  * by Adam Hamsik.
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  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 #include <sys/disk.h>
34 #include <sys/fcntl.h>
35 #include <sys/malloc.h>
36 #include <sys/namei.h>
37 #include <sys/nlookup.h>
38 
39 #include <dev/disk/dm/dm.h>
40 
41 static TAILQ_HEAD(, dm_pdev) dm_pdev_list;
42 
43 static struct lock dm_pdev_mutex;
44 
45 static dm_pdev_t *dm_pdev_alloc(const char *);
46 static int dm_pdev_free(dm_pdev_t *);
47 static dm_pdev_t *dm_pdev_lookup_name(const char *);
48 
49 /*
50  * Find used pdev with name == dm_pdev_name.
51  * needs to be called with the dm_pdev_mutex held.
52  */
53 static dm_pdev_t *
54 dm_pdev_lookup_name(const char *dm_pdev_name)
55 {
56 	dm_pdev_t *dmp;
57 
58 	KKASSERT(dm_pdev_name != NULL);
59 
60 	TAILQ_FOREACH(dmp, &dm_pdev_list, next_pdev) {
61 		if (strcmp(dm_pdev_name, dmp->name) == 0)
62 			return dmp;
63 	}
64 
65 	return NULL;
66 }
67 
68 static int
69 dm_dk_lookup(const char *dev_name, struct vnode **vpp)
70 {
71 	struct nlookupdata nd;
72 	int error;
73 
74 	error = nlookup_init(&nd, dev_name, UIO_SYSSPACE, NLC_FOLLOW);
75 	if (error)
76 		return error;
77 
78 	error = vn_open(&nd, NULL, FREAD|FWRITE, 0);
79 	if (error) {
80 		nlookup_done(&nd);
81 		return error;
82 	}
83 
84 	*vpp = nd.nl_open_vp;
85 	nd.nl_open_vp = NULL;
86 	nlookup_done(&nd);
87 
88 	return 0;
89 }
90 
91 /*
92  * Since dm can have arbitrary stacking on any number of disks and any dm
93  * volume is at least stacked onto another disk, we need to adjust the
94  * dumping offset (which is a raw offset from the beginning of the lowest
95  * physical disk) taking into account the offset of the underlying device
96  * which in turn takes into account the offset below it, etc.
97  *
98  * This function adjusts the dumping offset that is passed to the next
99  * dev_ddump() so it is correct for that underlying device.
100  */
101 off_t
102 dm_pdev_correct_dump_offset(dm_pdev_t *pdev, off_t offset)
103 {
104 	off_t noffset;
105 
106 	noffset = pdev->pdev_pinfo.reserved_blocks +
107 	    pdev->pdev_pinfo.media_offset / pdev->pdev_pinfo.media_blksize;
108 	noffset *= DEV_BSIZE;
109 	noffset += offset;
110 
111 	return noffset;
112 }
113 
114 /*
115  * Create entry for device with name dev_name and open vnode for it.
116  * If entry already exists in global TAILQ I will only increment
117  * reference counter.
118  */
119 dm_pdev_t *
120 dm_pdev_insert(const char *dev_name)
121 {
122 	dm_pdev_t *dmp;
123 	struct vattr va;
124 	int error;
125 
126 	KKASSERT(dev_name != NULL);
127 
128 	lockmgr(&dm_pdev_mutex, LK_EXCLUSIVE);
129 	dmp = dm_pdev_lookup_name(dev_name);
130 
131 	if (dmp != NULL) {
132 		dmp->ref_cnt++;
133 		dmdebug("dmp_pdev_insert pdev %s already in tree\n", dev_name);
134 		lockmgr(&dm_pdev_mutex, LK_RELEASE);
135 		return dmp;
136 	}
137 
138 	if ((dmp = dm_pdev_alloc(dev_name)) == NULL) {
139 		lockmgr(&dm_pdev_mutex, LK_RELEASE);
140 		return NULL;
141 	}
142 
143 	error = dm_dk_lookup(dev_name, &dmp->pdev_vnode);
144 	if (error) {
145 		dmdebug("dk_lookup on device: %s failed with error %d!\n",
146 		    dev_name, error);
147 		dm_pdev_free(dmp);
148 		lockmgr(&dm_pdev_mutex, LK_RELEASE);
149 		return NULL;
150 	}
151 	dmp->ref_cnt = 1;
152 
153 	if (dm_pdev_get_vattr(dmp, &va) == -1) {
154 		dmdebug("makeudev %s failed\n", dev_name);
155 		dm_pdev_free(dmp);
156 		lockmgr(&dm_pdev_mutex, LK_RELEASE);
157 		return NULL;
158 	}
159 	ksnprintf(dmp->udev_name, sizeof(dmp->udev_name),
160 		"%d:%d", va.va_rmajor, va.va_rminor);
161 	dmp->udev = dm_pdev_get_udev(dmp);
162 
163 	/*
164 	 * Get us the partinfo from the underlying device, it's needed for
165 	 * dumps.
166 	 */
167 	bzero(&dmp->pdev_pinfo, sizeof(dmp->pdev_pinfo));
168 	error = dev_dioctl(dmp->pdev_vnode->v_rdev, DIOCGPART,
169 	    (void *)&dmp->pdev_pinfo, 0, proc0.p_ucred, NULL, NULL);
170 	if (!error) {
171 		struct partinfo *dpart = &dmp->pdev_pinfo;
172 		dmdebug("dmp_pdev_insert DIOCGPART "
173 			"offset=%ju size=%ju blocks=%ju blksize=%d\n",
174 			dpart->media_offset,
175 			dpart->media_size,
176 			dpart->media_blocks,
177 			dpart->media_blksize);
178 	} else {
179 		kprintf("dmp_pdev_insert DIOCGPART failed %d\n", error);
180 	}
181 
182 	TAILQ_INSERT_TAIL(&dm_pdev_list, dmp, next_pdev);
183 	lockmgr(&dm_pdev_mutex, LK_RELEASE);
184 
185 	dmdebug("dmp_pdev_insert pdev %s %s 0x%016jx\n",
186 		dmp->name, dmp->udev_name, (uintmax_t)dmp->udev);
187 
188 	return dmp;
189 }
190 
191 /*
192  * Allocat new pdev structure if is not already present and
193  * set name.
194  */
195 static dm_pdev_t *
196 dm_pdev_alloc(const char *name)
197 {
198 	dm_pdev_t *dmp;
199 
200 	dmp = kmalloc(sizeof(*dmp), M_DM, M_WAITOK | M_ZERO);
201 	if (dmp == NULL)
202 		return NULL;
203 
204 	if (name)
205 		strlcpy(dmp->name, name, DM_MAX_DEV_NAME);
206 
207 	return dmp;
208 }
209 
210 /*
211  * Destroy allocated dm_pdev.
212  */
213 static int
214 dm_pdev_free(dm_pdev_t *dmp)
215 {
216 	int err;
217 
218 	KKASSERT(dmp != NULL);
219 
220 	if (dmp->pdev_vnode != NULL) {
221 		err = vn_close(dmp->pdev_vnode, FREAD | FWRITE, NULL);
222 		if (err != 0) {
223 			kfree(dmp, M_DM);
224 			return err;
225 		}
226 	}
227 	kfree(dmp, M_DM);
228 
229 	return 0;
230 }
231 
232 /*
233  * This funcion is called from targets' destroy() handler.
234  * When I'm removing device from list, I have to decrement
235  * reference counter. If reference counter is 0 I will remove
236  * dmp from global list and from device list to. And I will CLOSE
237  * dmp vnode too.
238  */
239 /*
240  * Decrement pdev reference counter if 0 remove it.
241  */
242 int
243 dm_pdev_decr(dm_pdev_t *dmp)
244 {
245 	KKASSERT(dmp != NULL);
246 	/*
247 	 * If this was last reference remove dmp from
248 	 * global list also.
249 	 */
250 	lockmgr(&dm_pdev_mutex, LK_EXCLUSIVE);
251 
252 	if (--dmp->ref_cnt == 0) {
253 		TAILQ_REMOVE(&dm_pdev_list, dmp, next_pdev);
254 		lockmgr(&dm_pdev_mutex, LK_RELEASE);
255 		dm_pdev_free(dmp);
256 		return 0;
257 	}
258 	lockmgr(&dm_pdev_mutex, LK_RELEASE);
259 	return 0;
260 }
261 
262 uint64_t
263 dm_pdev_get_udev(dm_pdev_t *dmp)
264 {
265 	struct vattr va;
266 	int ret;
267 
268 	if (dmp->pdev_vnode == NULL)
269 		return (uint64_t)-1;
270 
271 	ret = dm_pdev_get_vattr(dmp, &va);
272 	if (ret)
273 		return (uint64_t)-1;
274 
275 	ret = makeudev(va.va_rmajor, va.va_rminor);
276 
277 	return ret;
278 }
279 
280 int
281 dm_pdev_get_vattr(dm_pdev_t *dmp, struct vattr *vap)
282 {
283 	int ret;
284 
285 	if (dmp->pdev_vnode == NULL)
286 		return -1;
287 
288 	KKASSERT(vap);
289 	ret = VOP_GETATTR(dmp->pdev_vnode, vap);
290 	if (ret)
291 		return -1;
292 
293 	return 0;
294 }
295 
296 /*
297  * Initialize pdev subsystem.
298  */
299 int
300 dm_pdev_init(void)
301 {
302 	TAILQ_INIT(&dm_pdev_list);	/* initialize global pdev list */
303 	lockinit(&dm_pdev_mutex, "dmpdev", 0, LK_CANRECURSE);
304 
305 	return 0;
306 }
307 
308 /*
309  * Destroy all existing pdev's in device-mapper.
310  */
311 int
312 dm_pdev_uninit(void)
313 {
314 	dm_pdev_t *dmp;
315 
316 	lockmgr(&dm_pdev_mutex, LK_EXCLUSIVE);
317 
318 	while ((dmp = TAILQ_FIRST(&dm_pdev_list)) != NULL) {
319 		TAILQ_REMOVE(&dm_pdev_list, dmp, next_pdev);
320 		dm_pdev_free(dmp);
321 	}
322 	KKASSERT(TAILQ_EMPTY(&dm_pdev_list));
323 
324 	lockmgr(&dm_pdev_mutex, LK_RELEASE);
325 
326 	lockuninit(&dm_pdev_mutex);
327 	return 0;
328 }
329