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/types.h> 34 35 #include <sys/disk.h> 36 #include <sys/fcntl.h> 37 #include <sys/malloc.h> 38 #include <sys/namei.h> 39 #include <sys/vnode.h> 40 #include <sys/nlookup.h> 41 42 #include <dev/disk/dm/dm.h> 43 44 static TAILQ_HEAD(, dm_pdev) dm_pdev_list; 45 46 static struct lock dm_pdev_mutex; 47 48 static dm_pdev_t *dm_pdev_alloc(const char *); 49 static int dm_pdev_free(dm_pdev_t *); 50 static dm_pdev_t *dm_pdev_lookup_name(const char *); 51 52 /* 53 * Find used pdev with name == dm_pdev_name. 54 * needs to be called with the dm_pdev_mutex held. 55 */ 56 static dm_pdev_t * 57 dm_pdev_lookup_name(const char *dm_pdev_name) 58 { 59 dm_pdev_t *dmp; 60 61 KKASSERT(dm_pdev_name != NULL); 62 63 TAILQ_FOREACH(dmp, &dm_pdev_list, next_pdev) { 64 if (strcmp(dm_pdev_name, dmp->name) == 0) 65 return dmp; 66 } 67 68 return NULL; 69 } 70 71 static int 72 dm_dk_lookup(const char *dev_name, struct vnode **vpp) 73 { 74 struct nlookupdata nd; 75 int error; 76 77 error = nlookup_init(&nd, dev_name, UIO_SYSSPACE, NLC_FOLLOW); 78 if (error) 79 return error; 80 81 error = vn_open(&nd, NULL, FREAD|FWRITE, 0); 82 if (error) { 83 nlookup_done(&nd); 84 return error; 85 } 86 87 *vpp = nd.nl_open_vp; 88 nd.nl_open_vp = NULL; 89 nlookup_done(&nd); 90 91 return 0; 92 } 93 94 /* 95 * Since dm can have arbitrary stacking on any number of disks and any dm 96 * volume is at least stacked onto another disk, we need to adjust the 97 * dumping offset (which is a raw offset from the beginning of the lowest 98 * physical disk) taking into account the offset of the underlying device 99 * which in turn takes into account the offset below it, etc. 100 * 101 * This function adjusts the dumping offset that is passed to the next 102 * dev_ddump() so it is correct for that underlying device. 103 */ 104 off_t 105 dm_pdev_correct_dump_offset(dm_pdev_t *pdev, off_t offset) 106 { 107 off_t noffset; 108 109 noffset = pdev->pdev_pinfo.reserved_blocks + 110 pdev->pdev_pinfo.media_offset / pdev->pdev_pinfo.media_blksize; 111 noffset *= DEV_BSIZE; 112 noffset += offset; 113 114 return noffset; 115 } 116 117 /* 118 * Create entry for device with name dev_name and open vnode for it. 119 * If entry already exists in global TAILQ I will only increment 120 * reference counter. 121 */ 122 dm_pdev_t * 123 dm_pdev_insert(const char *dev_name) 124 { 125 dm_pdev_t *dmp; 126 struct vattr va; 127 int error; 128 129 KKASSERT(dev_name != NULL); 130 131 lockmgr(&dm_pdev_mutex, LK_EXCLUSIVE); 132 dmp = dm_pdev_lookup_name(dev_name); 133 134 if (dmp != NULL) { 135 dmp->ref_cnt++; 136 dmdebug("dmp_pdev_insert pdev %s already in tree\n", dev_name); 137 lockmgr(&dm_pdev_mutex, LK_RELEASE); 138 return dmp; 139 } 140 141 if ((dmp = dm_pdev_alloc(dev_name)) == NULL) { 142 lockmgr(&dm_pdev_mutex, LK_RELEASE); 143 return NULL; 144 } 145 146 error = dm_dk_lookup(dev_name, &dmp->pdev_vnode); 147 if (error) { 148 dmdebug("dk_lookup on device: %s failed with error %d!\n", 149 dev_name, error); 150 dm_pdev_free(dmp); 151 lockmgr(&dm_pdev_mutex, LK_RELEASE); 152 return NULL; 153 } 154 dmp->ref_cnt = 1; 155 156 if (dm_pdev_get_vattr(dmp, &va) == -1) { 157 dmdebug("makeudev %s failed\n", dev_name); 158 dm_pdev_free(dmp); 159 lockmgr(&dm_pdev_mutex, LK_RELEASE); 160 return NULL; 161 } 162 ksnprintf(dmp->udev_name, sizeof(dmp->udev_name), 163 "%d:%d", va.va_rmajor, va.va_rminor); 164 dmp->udev = dm_pdev_get_udev(dmp); 165 166 /* 167 * Get us the partinfo from the underlying device, it's needed for 168 * dumps. 169 */ 170 bzero(&dmp->pdev_pinfo, sizeof(dmp->pdev_pinfo)); 171 error = dev_dioctl(dmp->pdev_vnode->v_rdev, DIOCGPART, 172 (void *)&dmp->pdev_pinfo, 0, proc0.p_ucred, NULL, NULL); 173 if (!error) { 174 struct partinfo *dpart = &dmp->pdev_pinfo; 175 dmdebug("dmp_pdev_insert DIOCGPART " 176 "offset=%ju size=%ju blocks=%ju blksize=%d\n", 177 dpart->media_offset, 178 dpart->media_size, 179 dpart->media_blocks, 180 dpart->media_blksize); 181 } else { 182 kprintf("dmp_pdev_insert DIOCGPART failed %d\n", error); 183 } 184 185 TAILQ_INSERT_TAIL(&dm_pdev_list, dmp, next_pdev); 186 lockmgr(&dm_pdev_mutex, LK_RELEASE); 187 188 dmdebug("dmp_pdev_insert pdev %s %s 0x%016jx\n", 189 dmp->name, dmp->udev_name, (uintmax_t)dmp->udev); 190 191 return dmp; 192 } 193 194 /* 195 * Allocat new pdev structure if is not already present and 196 * set name. 197 */ 198 static dm_pdev_t * 199 dm_pdev_alloc(const char *name) 200 { 201 dm_pdev_t *dmp; 202 203 dmp = kmalloc(sizeof(*dmp), M_DM, M_WAITOK | M_ZERO); 204 if (dmp == NULL) 205 return NULL; 206 207 if (name) 208 strlcpy(dmp->name, name, MAX_DEV_NAME); 209 210 return dmp; 211 } 212 /* 213 * Destroy allocated dm_pdev. 214 */ 215 static int 216 dm_pdev_free(dm_pdev_t *dmp) 217 { 218 int err; 219 220 KKASSERT(dmp != NULL); 221 222 if (dmp->pdev_vnode != NULL) { 223 err = vn_close(dmp->pdev_vnode, FREAD | FWRITE, NULL); 224 if (err != 0) { 225 kfree(dmp, M_DM); 226 return err; 227 } 228 } 229 kfree(dmp, M_DM); 230 231 return 0; 232 } 233 234 /* 235 * This funcion is called from dm_dev_remove_ioctl. 236 * When I'm removing device from list, I have to decrement 237 * reference counter. If reference counter is 0 I will remove 238 * dmp from global list and from device list to. And I will CLOSE 239 * dmp vnode too. 240 */ 241 242 /* 243 * Decrement pdev reference counter if 0 remove it. 244 */ 245 int 246 dm_pdev_decr(dm_pdev_t *dmp) 247 { 248 KKASSERT(dmp != NULL); 249 /* 250 * If this was last reference remove dmp from 251 * global list also. 252 */ 253 lockmgr(&dm_pdev_mutex, LK_EXCLUSIVE); 254 255 if (--dmp->ref_cnt == 0) { 256 TAILQ_REMOVE(&dm_pdev_list, dmp, next_pdev); 257 lockmgr(&dm_pdev_mutex, LK_RELEASE); 258 dm_pdev_free(dmp); 259 return 0; 260 } 261 lockmgr(&dm_pdev_mutex, LK_RELEASE); 262 return 0; 263 } 264 265 uint64_t 266 dm_pdev_get_udev(dm_pdev_t *dmp) 267 { 268 struct vattr va; 269 int ret; 270 271 if (dmp->pdev_vnode == NULL) 272 return (uint64_t)-1; 273 274 ret = dm_pdev_get_vattr(dmp, &va); 275 if (ret) 276 return (uint64_t)-1; 277 278 ret = makeudev(va.va_rmajor, va.va_rminor); 279 280 return ret; 281 } 282 283 int 284 dm_pdev_get_vattr(dm_pdev_t *dmp, struct vattr *vap) 285 { 286 int ret; 287 288 if (dmp->pdev_vnode == NULL) 289 return -1; 290 291 KKASSERT(vap); 292 ret = VOP_GETATTR(dmp->pdev_vnode, vap); 293 if (ret) 294 return -1; 295 296 return 0; 297 } 298 299 /* 300 * Initialize pdev subsystem. 301 */ 302 int 303 dm_pdev_init(void) 304 { 305 TAILQ_INIT(&dm_pdev_list); /* initialize global pdev list */ 306 lockinit(&dm_pdev_mutex, "dmpdev", 0, LK_CANRECURSE); 307 308 return 0; 309 } 310 311 /* 312 * Destroy all existing pdev's in device-mapper. 313 */ 314 int 315 dm_pdev_uninit(void) 316 { 317 dm_pdev_t *dmp; 318 319 lockmgr(&dm_pdev_mutex, LK_EXCLUSIVE); 320 321 while ((dmp = TAILQ_FIRST(&dm_pdev_list)) != NULL) { 322 TAILQ_REMOVE(&dm_pdev_list, dmp, next_pdev); 323 dm_pdev_free(dmp); 324 } 325 KKASSERT(TAILQ_EMPTY(&dm_pdev_list)); 326 327 lockmgr(&dm_pdev_mutex, LK_RELEASE); 328 329 lockuninit(&dm_pdev_mutex); 330 return 0; 331 } 332