| 1 | /* |
| 2 | * Copyright (c) 2011, 2012 The DragonFly Project. All rights reserved. |
| 3 | * |
| 4 | * This code is derived from software contributed to The DragonFly Project |
| 5 | * by Matthew Dillon <dillon@backplane.com> |
| 6 | * |
| 7 | * Redistribution and use in source and binary forms, with or without |
| 8 | * modification, are permitted provided that the following conditions |
| 9 | * are met: |
| 10 | * |
| 11 | * 1. Redistributions of source code must retain the above copyright |
| 12 | * notice, this list of conditions and the following disclaimer. |
| 13 | * 2. Redistributions in binary form must reproduce the above copyright |
| 14 | * notice, this list of conditions and the following disclaimer in |
| 15 | * the documentation and/or other materials provided with the |
| 16 | * distribution. |
| 17 | * 3. Neither the name of The DragonFly Project nor the names of its |
| 18 | * contributors may be used to endorse or promote products derived |
| 19 | * from this software without specific, prior written permission. |
| 20 | * |
| 21 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 22 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 23 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS |
| 24 | * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE |
| 25 | * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 26 | * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, |
| 27 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
| 28 | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED |
| 29 | * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 30 | * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT |
| 31 | * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 32 | * SUCH DAMAGE. |
| 33 | */ |
| 34 | /*- |
| 35 | * Copyright (c) 2005 The NetBSD Foundation, Inc. |
| 36 | * All rights reserved. |
| 37 | * |
| 38 | * This code is derived from software contributed to The NetBSD Foundation |
| 39 | * by Julio M. Merino Vidal, developed as part of Google's Summer of Code |
| 40 | * 2005 program. |
| 41 | * |
| 42 | * Redistribution and use in source and binary forms, with or without |
| 43 | * modification, are permitted provided that the following conditions |
| 44 | * are met: |
| 45 | * 1. Redistributions of source code must retain the above copyright |
| 46 | * notice, this list of conditions and the following disclaimer. |
| 47 | * 2. Redistributions in binary form must reproduce the above copyright |
| 48 | * notice, this list of conditions and the following disclaimer in the |
| 49 | * documentation and/or other materials provided with the distribution. |
| 50 | * |
| 51 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
| 52 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
| 53 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
| 54 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS |
| 55 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
| 56 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
| 57 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
| 58 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
| 59 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
| 60 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| 61 | * POSSIBILITY OF SUCH DAMAGE. |
| 62 | */ |
| 63 | |
| 64 | #include <sys/param.h> |
| 65 | #include <sys/systm.h> |
| 66 | #include <sys/kernel.h> |
| 67 | #include <sys/nlookup.h> |
| 68 | #include <sys/vnode.h> |
| 69 | #include <sys/mount.h> |
| 70 | #include <sys/fcntl.h> |
| 71 | #include <sys/buf.h> |
| 72 | #include <sys/uuid.h> |
| 73 | |
| 74 | #include "hammer2.h" |
| 75 | #include "hammer2_disk.h" |
| 76 | #include "hammer2_mount.h" |
| 77 | |
| 78 | static int hammer2_init(struct vfsconf *conf); |
| 79 | static int hammer2_mount(struct mount *mp, char *path, caddr_t data, |
| 80 | struct ucred *cred); |
| 81 | static int hammer2_remount(struct mount *, char *, struct vnode *, |
| 82 | struct ucred *); |
| 83 | static int hammer2_unmount(struct mount *mp, int mntflags); |
| 84 | static int hammer2_root(struct mount *mp, struct vnode **vpp); |
| 85 | static int hammer2_statfs(struct mount *mp, struct statfs *sbp, |
| 86 | struct ucred *cred); |
| 87 | static int hammer2_statvfs(struct mount *mp, struct statvfs *sbp, |
| 88 | struct ucred *cred); |
| 89 | static int hammer2_sync(struct mount *mp, int waitfor); |
| 90 | static int hammer2_vget(struct mount *mp, struct vnode *dvp, |
| 91 | ino_t ino, struct vnode **vpp); |
| 92 | static int hammer2_fhtovp(struct mount *mp, struct vnode *rootvp, |
| 93 | struct fid *fhp, struct vnode **vpp); |
| 94 | static int hammer2_vptofh(struct vnode *vp, struct fid *fhp); |
| 95 | static int hammer2_checkexp(struct mount *mp, struct sockaddr *nam, |
| 96 | int *exflagsp, struct ucred **credanonp); |
| 97 | |
| 98 | static int tmpfs_unmount(struct mount *, int); |
| 99 | static int tmpfs_root(struct mount *, struct vnode **); |
| 100 | |
| 101 | /* |
| 102 | * HAMMER2 vfs operations. |
| 103 | */ |
| 104 | static struct vfsops hammer2_vfsops = { |
| 105 | /* From tmpfs */ |
| 106 | .vfs_root = tmpfs_root, |
| 107 | |
| 108 | /* From HAMMER2 */ |
| 109 | .vfs_init = hammer2_init, |
| 110 | .vfs_sync = hammer2_sync, |
| 111 | .vfs_mount = hammer2_mount, |
| 112 | .vfs_unmount = hammer2_unmount, |
| 113 | #ifdef notyet |
| 114 | .vfs_root = hammer2_root, |
| 115 | #endif |
| 116 | .vfs_statfs = hammer2_statfs, |
| 117 | /* If we enable statvfs, we disappear in df, till we implement it. */ |
| 118 | /* That makes debugging difficult :) */ |
| 119 | // .vfs_statvfs = hammer2_statvfs, |
| 120 | .vfs_vget = hammer2_vget, |
| 121 | .vfs_vptofh = hammer2_vptofh, |
| 122 | .vfs_fhtovp = hammer2_fhtovp, |
| 123 | .vfs_checkexp = hammer2_checkexp |
| 124 | }; |
| 125 | |
| 126 | |
| 127 | MALLOC_DEFINE(M_HAMMER2, "HAMMER2-mount", ""); |
| 128 | |
| 129 | VFS_SET(hammer2_vfsops, hammer2, 0); |
| 130 | MODULE_VERSION(hammer2, 1); |
| 131 | |
| 132 | static int |
| 133 | hammer2_init(struct vfsconf *conf) |
| 134 | { |
| 135 | int error; |
| 136 | |
| 137 | error = 0; |
| 138 | |
| 139 | if (HAMMER2_BLOCKREF_SIZE != sizeof(struct hammer2_blockref)) |
| 140 | error = EINVAL; |
| 141 | if (HAMMER2_INODE_SIZE != sizeof(struct hammer2_inode_data)) |
| 142 | error = EINVAL; |
| 143 | if (HAMMER2_ALLOCREF_SIZE != sizeof(struct hammer2_allocref)) |
| 144 | error = EINVAL; |
| 145 | if (HAMMER2_VOLUME_SIZE != sizeof(struct hammer2_volume_data)) |
| 146 | error = EINVAL; |
| 147 | |
| 148 | if (error) |
| 149 | kprintf("HAMMER2 structure size mismatch; cannot continue.\n"); |
| 150 | |
| 151 | return (error); |
| 152 | } |
| 153 | |
| 154 | /* |
| 155 | * Mount or remount HAMMER2 fileystem from physical media |
| 156 | * |
| 157 | * mountroot |
| 158 | * mp mount point structure |
| 159 | * path NULL |
| 160 | * data <unused> |
| 161 | * cred <unused> |
| 162 | * |
| 163 | * mount |
| 164 | * mp mount point structure |
| 165 | * path path to mount point |
| 166 | * data pointer to argument structure in user space |
| 167 | * volume volume path (device@LABEL form) |
| 168 | * hflags user mount flags |
| 169 | * cred user credentials |
| 170 | * |
| 171 | * RETURNS: 0 Success |
| 172 | * !0 error number |
| 173 | */ |
| 174 | static int |
| 175 | hammer2_mount(struct mount *mp, char *path, caddr_t data, |
| 176 | struct ucred *cred) |
| 177 | { |
| 178 | struct hammer2_mount_info info; |
| 179 | struct hammer2_mount *hmp; |
| 180 | struct vnode *devvp; |
| 181 | struct nlookupdata nd; |
| 182 | char devstr[MNAMELEN]; |
| 183 | size_t size; |
| 184 | size_t done; |
| 185 | char *dev, *label; |
| 186 | int ronly; |
| 187 | int error; |
| 188 | int rc; |
| 189 | |
| 190 | hmp = NULL; |
| 191 | dev = label = NULL; |
| 192 | devvp = NULL; |
| 193 | |
| 194 | kprintf("hammer2_mount\n"); |
| 195 | |
| 196 | if (path == NULL) { |
| 197 | /* |
| 198 | * Root mount |
| 199 | */ |
| 200 | |
| 201 | return (EOPNOTSUPP); |
| 202 | } else { |
| 203 | /* |
| 204 | * Non-root mount or updating a mount |
| 205 | */ |
| 206 | |
| 207 | error = copyin(data, &info, sizeof(info)); |
| 208 | if (error) |
| 209 | return (error); |
| 210 | |
| 211 | error = copyinstr(info.volume, devstr, MNAMELEN - 1, &done); |
| 212 | if (error) |
| 213 | return (error); |
| 214 | |
| 215 | /* Extract device and label */ |
| 216 | dev = devstr; |
| 217 | label = strchr(devstr, '@'); |
| 218 | if (label == NULL || |
| 219 | ((label + 1) - dev) > done) |
| 220 | return (EINVAL); |
| 221 | *label = '\0'; |
| 222 | label++; |
| 223 | if (*label == '\0') |
| 224 | return (EINVAL); |
| 225 | |
| 226 | if (mp->mnt_flag & MNT_UPDATE) { |
| 227 | /* Update mount */ |
| 228 | /* HAMMER2 implements NFS export via mountctl */ |
| 229 | hmp = MPTOH2(mp); |
| 230 | devvp = hmp->hm_devvp; |
| 231 | return hammer2_remount(mp, path, devvp, cred); |
| 232 | } |
| 233 | } |
| 234 | |
| 235 | /* |
| 236 | * New non-root mount |
| 237 | */ |
| 238 | /* Lookup name and verify it refers to a block device */ |
| 239 | error = nlookup_init(&nd, dev, UIO_SYSSPACE, NLC_FOLLOW); |
| 240 | if (error) |
| 241 | return (error); |
| 242 | error = nlookup(&nd); |
| 243 | if (error) |
| 244 | return (error); |
| 245 | error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp); |
| 246 | if (error) |
| 247 | return (error); |
| 248 | nlookup_done(&nd); |
| 249 | |
| 250 | if (!vn_isdisk(devvp, &error)) { |
| 251 | vrele(devvp); |
| 252 | return (error); |
| 253 | } |
| 254 | |
| 255 | /* |
| 256 | * Common path for new root/non-root mounts; |
| 257 | * devvp is a ref-ed by not locked vnode referring to the fs device |
| 258 | */ |
| 259 | |
| 260 | error = vfs_mountedon(devvp); |
| 261 | if (error) { |
| 262 | vrele(devvp); |
| 263 | return (error); |
| 264 | } |
| 265 | |
| 266 | if (vcount(devvp) > 0) { |
| 267 | vrele(devvp); |
| 268 | return (EBUSY); |
| 269 | } |
| 270 | |
| 271 | /* |
| 272 | * Open the fs device |
| 273 | */ |
| 274 | ronly = (mp->mnt_flag & MNT_RDONLY) != 0; |
| 275 | vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); |
| 276 | error = vinvalbuf(devvp, V_SAVE, 0, 0); |
| 277 | if (error) { |
| 278 | vn_unlock(devvp); |
| 279 | vrele(devvp); |
| 280 | return (error); |
| 281 | } |
| 282 | /* This is correct; however due to an NFS quirk of my setup, FREAD |
| 283 | * is required... */ |
| 284 | /* |
| 285 | error = VOP_OPEN(devvp, ronly ? FREAD : FREAD | FWRITE, FSCRED, NULL); |
| 286 | */ |
| 287 | error = VOP_OPEN(devvp, FREAD, FSCRED, NULL); |
| 288 | vn_unlock(devvp); |
| 289 | if (error) { |
| 290 | vrele(devvp); |
| 291 | return (error); |
| 292 | } |
| 293 | |
| 294 | #ifdef notyet |
| 295 | /* VOP_IOCTL(EXTENDED_DISK_INFO, devvp); */ |
| 296 | /* if vn device, never use bdwrite(); */ |
| 297 | /* check if device supports BUF_CMD_READALL; */ |
| 298 | /* check if device supports BUF_CMD_WRITEALL; */ |
| 299 | #endif |
| 300 | |
| 301 | hmp = kmalloc(sizeof(*hmp), M_HAMMER2, M_WAITOK | M_ZERO); |
| 302 | /*mp->mnt_data = (qaddr_t) hmp;*/ |
| 303 | hmp->hm_mp = mp; |
| 304 | /*hmp->hm_ronly = ronly;*/ |
| 305 | /*hmp->hm_devvp = devvp;*/ |
| 306 | lockinit(&hmp->hm_lk, "h2mp", 0, 0); |
| 307 | kmalloc_create(&hmp->hm_inodes, "HAMMER2-inodes"); |
| 308 | kmalloc_create(&hmp->hm_ipstacks, "HAMMER2-ipstacks"); |
| 309 | |
| 310 | /* Readout volume headers, make sure we have a live filesystem */ |
| 311 | /* Kinda hacky atm */ |
| 312 | { |
| 313 | struct buf *bps[HAMMER2_NUM_VOLHDRS]; |
| 314 | int valid = 0; |
| 315 | int hi_tid = 0; |
| 316 | int hi_num = 0; |
| 317 | int i; |
| 318 | uint32_t crc; |
| 319 | struct hammer2_volume_data *vd; |
| 320 | for (i = 0; i < HAMMER2_NUM_VOLHDRS; i++) { |
| 321 | rc = bread(devvp, i * HAMMER2_RESERVE_ALIGN64, |
| 322 | HAMMER2_BUFSIZE, &bps[i]); |
| 323 | if (rc != 0) { |
| 324 | brelse(bps[i]); |
| 325 | bps[i] = NULL; |
| 326 | continue; |
| 327 | } |
| 328 | |
| 329 | vd = bps[i]->b_data; |
| 330 | if (vd->magic == HAMMER2_VOLUME_ID_HBO) { |
| 331 | uint32_t ccrc; |
| 332 | unsigned char tmp[512]; |
| 333 | bcopy(bps[i]->b_data, &tmp, 512); |
| 334 | bzero(&tmp[512 - 4], 4); |
| 335 | /* Calculate CRC32 w/ crc field zero */ |
| 336 | /* XXX: Can we modify b_data? */ |
| 337 | ccrc = hammer2_icrc32(tmp, 512); |
| 338 | crc = vd->icrc_sect0; |
| 339 | |
| 340 | if (ccrc != crc) { |
| 341 | brelse(bps[i]); |
| 342 | bps[i] = NULL; |
| 343 | continue; |
| 344 | } |
| 345 | |
| 346 | valid++; |
| 347 | if (vd->last_tid > hi_tid) { |
| 348 | hi_tid = vd->last_tid; |
| 349 | hi_num = i; |
| 350 | } |
| 351 | } |
| 352 | } |
| 353 | if (valid) { |
| 354 | /* We have found the hammer volume header w/ |
| 355 | * the highest transaction id. Use it. */ |
| 356 | |
| 357 | bcopy(bps[hi_num]->b_data, &hmp->hm_sb, |
| 358 | HAMMER2_BUFSIZE); |
| 359 | |
| 360 | for (i = 0 ; i < HAMMER2_NUM_VOLHDRS; i++) |
| 361 | brelse(bps[i]); |
| 362 | |
| 363 | kprintf("HAMMER2 volume %d by\n", hmp->hm_sb.volu_size); |
| 364 | } else { |
| 365 | /* XXX More to do! Release structures and stuff */ |
| 366 | return (EINVAL); |
| 367 | } |
| 368 | } |
| 369 | |
| 370 | /* |
| 371 | * Filesystem subroutines are self-synchronized |
| 372 | */ |
| 373 | /*mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;*/ |
| 374 | |
| 375 | |
| 376 | /* Setup root inode */ |
| 377 | hmp->hm_iroot = alloci(hmp); |
| 378 | hmp->hm_iroot->type = HAMMER2_INODE_DIR | HAMMER2_INODE_ROOT; |
| 379 | hmp->hm_iroot->hi_inum = 1; |
| 380 | |
| 381 | /* currently rely on tmpfs routines */ |
| 382 | /*vfs_getnewfsid(mp);*/ |
| 383 | /*vfs_add_vnodeops(mp, &hammer2_vnode_vops, &mp->mnt_vn_norm_ops);*/ |
| 384 | /*vfs_add_vnodeops(mp, &hammer2_spec_vops, &mp->mnt_vn_spec_ops);*/ |
| 385 | /*vfs_add_vnodeops(mp, &hammer2_fifo_vops, &mp->mnt_vn_fifo_ops);*/ |
| 386 | |
| 387 | copystr("hammer2", mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size); |
| 388 | bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); |
| 389 | bzero(mp->mnt_stat.f_mntonname, sizeof(mp->mnt_stat.f_mntonname)); |
| 390 | copyinstr(path, mp->mnt_stat.f_mntonname, |
| 391 | sizeof(mp->mnt_stat.f_mntonname) - 1, |
| 392 | &size); |
| 393 | |
| 394 | hammer2_statfs(mp, &mp->mnt_stat, cred); |
| 395 | |
| 396 | hammer2_inode_unlock_ex(hmp->hm_iroot); |
| 397 | |
| 398 | return (tmpfs_mount(hmp, mp, path, data, cred)); |
| 399 | } |
| 400 | |
| 401 | static int |
| 402 | hammer2_remount(struct mount *mp, char *path, struct vnode *devvp, |
| 403 | struct ucred *cred) |
| 404 | { |
| 405 | return (0); |
| 406 | } |
| 407 | |
| 408 | static int |
| 409 | hammer2_unmount(struct mount *mp, int mntflags) |
| 410 | { |
| 411 | struct hammer2_mount *hmp; |
| 412 | int flags; |
| 413 | int error; |
| 414 | |
| 415 | kprintf("hammer2_unmount\n"); |
| 416 | |
| 417 | hmp = MPTOH2(mp); |
| 418 | flags = 0; |
| 419 | |
| 420 | if (mntflags & MNT_FORCE) |
| 421 | flags |= FORCECLOSE; |
| 422 | |
| 423 | hammer2_mount_exlock(hmp); |
| 424 | |
| 425 | error = vflush(mp, 0, flags); |
| 426 | |
| 427 | /* |
| 428 | * Work to do: |
| 429 | * 1) Wait on the flusher having no work; heat up if needed |
| 430 | * 2) Scan inode RB tree till all the inodes are free |
| 431 | * 3) Destroy the kmalloc inode zone |
| 432 | * 4) Free the mount point |
| 433 | */ |
| 434 | |
| 435 | kmalloc_destroy(&hmp->hm_inodes); |
| 436 | kmalloc_destroy(&hmp->hm_ipstacks); |
| 437 | |
| 438 | hammer2_mount_unlock(hmp); |
| 439 | |
| 440 | // Tmpfs does this |
| 441 | //kfree(hmp, M_HAMMER2); |
| 442 | |
| 443 | return (tmpfs_unmount(mp, mntflags)); |
| 444 | |
| 445 | return (error); |
| 446 | } |
| 447 | |
| 448 | static int |
| 449 | hammer2_vget(struct mount *mp, struct vnode *dvp, |
| 450 | ino_t ino, struct vnode **vpp) |
| 451 | { |
| 452 | kprintf("hammer2_vget\n"); |
| 453 | return (EOPNOTSUPP); |
| 454 | } |
| 455 | |
| 456 | static int |
| 457 | hammer2_root(struct mount *mp, struct vnode **vpp) |
| 458 | { |
| 459 | struct hammer2_mount *hmp; |
| 460 | int error; |
| 461 | struct vnode *vp; |
| 462 | |
| 463 | kprintf("hammer2_root\n"); |
| 464 | |
| 465 | hmp = MPTOH2(mp); |
| 466 | hammer2_mount_lock_ex(hmp); |
| 467 | if (hmp->hm_iroot == NULL) { |
| 468 | *vpp = NULL; |
| 469 | error = EINVAL; |
| 470 | } else { |
| 471 | vp = igetv(hmp->hm_iroot, &error); |
| 472 | *vpp = vp; |
| 473 | if (vp == NULL) |
| 474 | kprintf("vnodefail\n"); |
| 475 | } |
| 476 | hammer2_mount_unlock(hmp); |
| 477 | |
| 478 | return (error); |
| 479 | } |
| 480 | |
| 481 | static int |
| 482 | hammer2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred) |
| 483 | { |
| 484 | struct hammer2_mount *hmp; |
| 485 | |
| 486 | kprintf("hammer2_statfs\n"); |
| 487 | |
| 488 | hmp = MPTOH2(mp); |
| 489 | |
| 490 | sbp->f_iosize = PAGE_SIZE; |
| 491 | sbp->f_bsize = PAGE_SIZE; |
| 492 | |
| 493 | sbp->f_blocks = 10; |
| 494 | sbp->f_bavail = 10; |
| 495 | sbp->f_bfree = 10; |
| 496 | |
| 497 | sbp->f_files = 10; |
| 498 | sbp->f_ffree = 10; |
| 499 | sbp->f_owner = 0; |
| 500 | |
| 501 | return (0); |
| 502 | } |
| 503 | |
| 504 | static int |
| 505 | hammer2_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred) |
| 506 | { |
| 507 | kprintf("hammer2_statvfs\n"); |
| 508 | return (EOPNOTSUPP); |
| 509 | } |
| 510 | |
| 511 | /* |
| 512 | * Sync the entire filesystem; this is called from the filesystem syncer |
| 513 | * process periodically and whenever a user calls sync(1) on the hammer |
| 514 | * mountpoint. |
| 515 | * |
| 516 | * Currently is actually called from the syncer! \o/ |
| 517 | * |
| 518 | * This task will have to snapshot the state of the dirty inode chain. |
| 519 | * From that, it will have to make sure all of the inodes on the dirty |
| 520 | * chain have IO initiated. We make sure that io is initiated for the root |
| 521 | * block. |
| 522 | * |
| 523 | * If waitfor is set, we wait for media to acknowledge the new rootblock. |
| 524 | * |
| 525 | * THINKS: side A vs side B, to have sync not stall all I/O? |
| 526 | */ |
| 527 | static int |
| 528 | hammer2_sync(struct mount *mp, int waitfor) |
| 529 | { |
| 530 | struct hammer2_mount *hmp; |
| 531 | struct hammer2_inode *ip; |
| 532 | |
| 533 | kprintf("hammer2_sync \n"); |
| 534 | |
| 535 | // hmp = MPTOH2(mp); |
| 536 | |
| 537 | return (0); |
| 538 | } |
| 539 | |
| 540 | static int |
| 541 | hammer2_vptofh(struct vnode *vp, struct fid *fhp) |
| 542 | { |
| 543 | return (0); |
| 544 | } |
| 545 | |
| 546 | static int |
| 547 | hammer2_fhtovp(struct mount *mp, struct vnode *rootvp, |
| 548 | struct fid *fhp, struct vnode **vpp) |
| 549 | { |
| 550 | return (0); |
| 551 | } |
| 552 | |
| 553 | static int |
| 554 | hammer2_checkexp(struct mount *mp, struct sockaddr *nam, |
| 555 | int *exflagsp, struct ucred **credanonp) |
| 556 | { |
| 557 | return (0); |
| 558 | } |
| 559 | |
| 560 | /* |
| 561 | * Efficient memory file system. |
| 562 | * |
| 563 | * tmpfs is a file system that uses NetBSD's virtual memory sub-system |
| 564 | * (the well-known UVM) to store file data and metadata in an efficient |
| 565 | * way. This means that it does not follow the structure of an on-disk |
| 566 | * file system because it simply does not need to. Instead, it uses |
| 567 | * memory-specific data structures and algorithms to automatically |
| 568 | * allocate and release resources. |
| 569 | */ |
| 570 | |
| 571 | #include <sys/conf.h> |
| 572 | #include <sys/param.h> |
| 573 | #include <sys/limits.h> |
| 574 | #include <sys/lock.h> |
| 575 | #include <sys/mutex.h> |
| 576 | #include <sys/kernel.h> |
| 577 | #include <sys/stat.h> |
| 578 | #include <sys/systm.h> |
| 579 | #include <sys/sysctl.h> |
| 580 | #include <sys/objcache.h> |
| 581 | |
| 582 | #include <vm/vm.h> |
| 583 | #include <vm/vm_object.h> |
| 584 | #include <vm/vm_param.h> |
| 585 | |
| 586 | #include "hammer2.h" |
| 587 | |
| 588 | /* |
| 589 | * Default permission for root node |
| 590 | */ |
| 591 | #define TMPFS_DEFAULT_ROOT_MODE (S_IRWXU|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH) |
| 592 | |
| 593 | /* --------------------------------------------------------------------- */ |
| 594 | int |
| 595 | tmpfs_node_ctor(void *obj, void *privdata, int flags) |
| 596 | { |
| 597 | struct tmpfs_node *node = (struct tmpfs_node *)obj; |
| 598 | |
| 599 | node->tn_gen++; |
| 600 | node->tn_size = 0; |
| 601 | node->tn_status = 0; |
| 602 | node->tn_flags = 0; |
| 603 | node->tn_links = 0; |
| 604 | node->tn_vnode = NULL; |
| 605 | node->tn_vpstate = TMPFS_VNODE_WANT; |
| 606 | bzero(&node->tn_spec, sizeof(node->tn_spec)); |
| 607 | |
| 608 | return (1); |
| 609 | } |
| 610 | |
| 611 | static void |
| 612 | tmpfs_node_dtor(void *obj, void *privdata) |
| 613 | { |
| 614 | struct tmpfs_node *node = (struct tmpfs_node *)obj; |
| 615 | node->tn_type = VNON; |
| 616 | node->tn_vpstate = TMPFS_VNODE_DOOMED; |
| 617 | } |
| 618 | |
| 619 | static void* |
| 620 | tmpfs_node_init(void *args, int flags) |
| 621 | { |
| 622 | struct tmpfs_node *node = (struct tmpfs_node *)objcache_malloc_alloc(args, flags); |
| 623 | if (node == NULL) |
| 624 | return (NULL); |
| 625 | node->tn_id = 0; |
| 626 | |
| 627 | lockinit(&node->tn_interlock, "tmpfs node interlock", 0, LK_CANRECURSE); |
| 628 | node->tn_gen = karc4random(); |
| 629 | |
| 630 | return node; |
| 631 | } |
| 632 | |
| 633 | static void |
| 634 | tmpfs_node_fini(void *obj, void *args) |
| 635 | { |
| 636 | struct tmpfs_node *node = (struct tmpfs_node *)obj; |
| 637 | lockuninit(&node->tn_interlock); |
| 638 | objcache_malloc_free(obj, args); |
| 639 | } |
| 640 | |
| 641 | int |
| 642 | tmpfs_mount(struct hammer2_mount *hmp, |
| 643 | struct mount *mp, char *path, caddr_t data, struct ucred *cred) |
| 644 | { |
| 645 | // struct tmpfs_mount *tmp; |
| 646 | struct tmpfs_node *root; |
| 647 | // struct tmpfs_args args; |
| 648 | vm_pindex_t pages; |
| 649 | vm_pindex_t pages_limit; |
| 650 | ino_t nodes; |
| 651 | u_int64_t maxfsize; |
| 652 | int error; |
| 653 | /* Size counters. */ |
| 654 | ino_t nodes_max; |
| 655 | off_t size_max; |
| 656 | size_t maxfsize_max; |
| 657 | size_t size; |
| 658 | |
| 659 | /* Root node attributes. */ |
| 660 | uid_t root_uid = cred->cr_uid; |
| 661 | gid_t root_gid = cred->cr_gid; |
| 662 | mode_t root_mode = (VREAD | VWRITE); |
| 663 | |
| 664 | if (mp->mnt_flag & MNT_UPDATE) { |
| 665 | /* XXX: There is no support yet to update file system |
| 666 | * settings. Should be added. */ |
| 667 | |
| 668 | return EOPNOTSUPP; |
| 669 | } |
| 670 | |
| 671 | kprintf("tmpfs_mount\n"); |
| 672 | |
| 673 | /* |
| 674 | * mount info |
| 675 | */ |
| 676 | // bzero(&args, sizeof(args)); |
| 677 | size_max = 0; |
| 678 | nodes_max = 0; |
| 679 | maxfsize_max = 0; |
| 680 | |
| 681 | if (path) { |
| 682 | if (data) { |
| 683 | // error = copyin(data, &args, sizeof(args)); |
| 684 | // if (error) |
| 685 | // return (error); |
| 686 | } |
| 687 | /* |
| 688 | size_max = args.ta_size_max; |
| 689 | nodes_max = args.ta_nodes_max; |
| 690 | maxfsize_max = args.ta_maxfsize_max; |
| 691 | root_uid = args.ta_root_uid; |
| 692 | root_gid = args.ta_root_gid; |
| 693 | root_mode = args.ta_root_mode; |
| 694 | */ |
| 695 | } |
| 696 | |
| 697 | /* |
| 698 | * If mount by non-root, then verify that user has necessary |
| 699 | * permissions on the device. |
| 700 | */ |
| 701 | if (cred->cr_uid != 0) { |
| 702 | root_mode = VREAD; |
| 703 | if ((mp->mnt_flag & MNT_RDONLY) == 0) |
| 704 | root_mode |= VWRITE; |
| 705 | } |
| 706 | |
| 707 | pages_limit = vm_swap_max + vmstats.v_page_count / 2; |
| 708 | |
| 709 | if (size_max == 0) |
| 710 | pages = pages_limit / 2; |
| 711 | else if (size_max < PAGE_SIZE) |
| 712 | pages = 1; |
| 713 | else if (OFF_TO_IDX(size_max) > pages_limit) |
| 714 | pages = pages_limit; |
| 715 | else |
| 716 | pages = OFF_TO_IDX(size_max); |
| 717 | |
| 718 | if (nodes_max == 0) |
| 719 | nodes = 3 + pages * PAGE_SIZE / 1024; |
| 720 | else if (nodes_max < 3) |
| 721 | nodes = 3; |
| 722 | else if (nodes_max > pages) |
| 723 | nodes = pages; |
| 724 | else |
| 725 | nodes = nodes_max; |
| 726 | |
| 727 | maxfsize = IDX_TO_OFF(pages_limit); |
| 728 | if (maxfsize_max != 0 && maxfsize > maxfsize_max) |
| 729 | maxfsize = maxfsize_max; |
| 730 | |
| 731 | /* Allocate the tmpfs mount structure and fill it. */ |
| 732 | // tmp = kmalloc(sizeof(*tmp), M_HAMMER2, M_WAITOK | M_ZERO); |
| 733 | |
| 734 | struct hammer2_mount *tmp = hmp; |
| 735 | lockinit(&(tmp->allnode_lock), "tmpfs allnode lock", 0, LK_CANRECURSE); |
| 736 | tmp->tm_nodes_max = nodes; |
| 737 | tmp->tm_nodes_inuse = 0; |
| 738 | tmp->tm_maxfilesize = maxfsize; |
| 739 | LIST_INIT(&tmp->tm_nodes_used); |
| 740 | |
| 741 | tmp->tm_pages_max = pages; |
| 742 | tmp->tm_pages_used = 0; |
| 743 | |
| 744 | kmalloc_create(&tmp->tm_node_zone, "tmpfs node"); |
| 745 | kmalloc_create(&tmp->tm_dirent_zone, "tmpfs dirent"); |
| 746 | kmalloc_create(&tmp->tm_name_zone, "tmpfs name zone"); |
| 747 | |
| 748 | kmalloc_raise_limit(tmp->tm_node_zone, sizeof(struct tmpfs_node) * |
| 749 | tmp->tm_nodes_max); |
| 750 | |
| 751 | tmp->tm_node_zone_malloc_args.objsize = sizeof(struct tmpfs_node); |
| 752 | tmp->tm_node_zone_malloc_args.mtype = tmp->tm_node_zone; |
| 753 | |
| 754 | tmp->tm_dirent_zone_malloc_args.objsize = sizeof(struct tmpfs_dirent); |
| 755 | tmp->tm_dirent_zone_malloc_args.mtype = tmp->tm_dirent_zone; |
| 756 | |
| 757 | tmp->tm_dirent_pool = objcache_create( "tmpfs dirent cache", |
| 758 | 0, 0, |
| 759 | NULL, NULL, NULL, |
| 760 | objcache_malloc_alloc, objcache_malloc_free, |
| 761 | &tmp->tm_dirent_zone_malloc_args); |
| 762 | tmp->tm_node_pool = objcache_create( "tmpfs node cache", |
| 763 | 0, 0, |
| 764 | tmpfs_node_ctor, tmpfs_node_dtor, NULL, |
| 765 | tmpfs_node_init, tmpfs_node_fini, |
| 766 | &tmp->tm_node_zone_malloc_args); |
| 767 | |
| 768 | /* Allocate the root node. */ |
| 769 | error = tmpfs_alloc_node(tmp, VDIR, root_uid, root_gid, |
| 770 | root_mode & ALLPERMS, NULL, NULL, |
| 771 | VNOVAL, VNOVAL, &root); |
| 772 | |
| 773 | /* |
| 774 | * We are backed by swap, set snocache chflags flag so we |
| 775 | * don't trip over swapcache. |
| 776 | */ |
| 777 | root->tn_flags = SF_NOCACHE; |
| 778 | |
| 779 | if (error != 0 || root == NULL) { |
| 780 | objcache_destroy(tmp->tm_node_pool); |
| 781 | objcache_destroy(tmp->tm_dirent_pool); |
| 782 | kfree(tmp, M_HAMMER2); |
| 783 | return error; |
| 784 | } |
| 785 | KASSERT(root->tn_id >= 0, ("tmpfs root with invalid ino: %d", (int)root->tn_id)); |
| 786 | tmp->tm_root = root; |
| 787 | |
| 788 | mp->mnt_flag |= MNT_LOCAL; |
| 789 | #if 0 |
| 790 | mp->mnt_kern_flag |= MNTK_RD_MPSAFE | MNTK_WR_MPSAFE | MNTK_GA_MPSAFE | |
| 791 | MNTK_IN_MPSAFE | MNTK_SG_MPSAFE; |
| 792 | #endif |
| 793 | mp->mnt_kern_flag |= MNTK_RD_MPSAFE | MNTK_GA_MPSAFE | MNTK_SG_MPSAFE; |
| 794 | mp->mnt_kern_flag |= MNTK_WR_MPSAFE; |
| 795 | mp->mnt_kern_flag |= MNTK_NOMSYNC; |
| 796 | mp->mnt_kern_flag |= MNTK_THR_SYNC; |
| 797 | mp->mnt_data = (qaddr_t)tmp; |
| 798 | vfs_getnewfsid(mp); |
| 799 | |
| 800 | vfs_add_vnodeops(mp, &tmpfs_vnode_vops, &mp->mnt_vn_norm_ops); |
| 801 | vfs_add_vnodeops(mp, &tmpfs_fifo_vops, &mp->mnt_vn_fifo_ops); |
| 802 | |
| 803 | hammer2_statfs(mp, &mp->mnt_stat, cred); |
| 804 | |
| 805 | return 0; |
| 806 | } |
| 807 | |
| 808 | /* --------------------------------------------------------------------- */ |
| 809 | |
| 810 | /* ARGSUSED2 */ |
| 811 | static int |
| 812 | tmpfs_unmount(struct mount *mp, int mntflags) |
| 813 | { |
| 814 | int error; |
| 815 | int flags = 0; |
| 816 | int found; |
| 817 | struct hammer2_mount *tmp; |
| 818 | struct tmpfs_node *node; |
| 819 | |
| 820 | kprintf("tmpfs_umount\n"); |
| 821 | |
| 822 | /* Handle forced unmounts. */ |
| 823 | if (mntflags & MNT_FORCE) |
| 824 | flags |= FORCECLOSE; |
| 825 | |
| 826 | tmp = VFS_TO_TMPFS(mp); |
| 827 | |
| 828 | /* |
| 829 | * Finalize all pending I/O. In the case of tmpfs we want |
| 830 | * to throw all the data away so clean out the buffer cache |
| 831 | * and vm objects before calling vflush(). |
| 832 | */ |
| 833 | LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) { |
| 834 | if (node->tn_type == VREG && node->tn_vnode) { |
| 835 | ++node->tn_links; |
| 836 | TMPFS_NODE_LOCK(node); |
| 837 | vx_get(node->tn_vnode); |
| 838 | tmpfs_truncate(node->tn_vnode, 0); |
| 839 | vx_put(node->tn_vnode); |
| 840 | TMPFS_NODE_UNLOCK(node); |
| 841 | --node->tn_links; |
| 842 | } |
| 843 | } |
| 844 | error = vflush(mp, 0, flags); |
| 845 | if (error != 0) |
| 846 | return error; |
| 847 | |
| 848 | /* |
| 849 | * First pass get rid of all the directory entries and |
| 850 | * vnode associations. The directory structure will |
| 851 | * remain via the extra link count representing tn_dir.tn_parent. |
| 852 | * |
| 853 | * No vnodes should remain after the vflush above. |
| 854 | */ |
| 855 | LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) { |
| 856 | ++node->tn_links; |
| 857 | TMPFS_NODE_LOCK(node); |
| 858 | if (node->tn_type == VDIR) { |
| 859 | struct tmpfs_dirent *de; |
| 860 | |
| 861 | while (!TAILQ_EMPTY(&node->tn_dir.tn_dirhead)) { |
| 862 | de = TAILQ_FIRST(&node->tn_dir.tn_dirhead); |
| 863 | tmpfs_dir_detach(node, de); |
| 864 | tmpfs_free_dirent(tmp, de); |
| 865 | node->tn_size -= sizeof(struct tmpfs_dirent); |
| 866 | } |
| 867 | } |
| 868 | KKASSERT(node->tn_vnode == NULL); |
| 869 | #if 0 |
| 870 | vp = node->tn_vnode; |
| 871 | if (vp != NULL) { |
| 872 | tmpfs_free_vp(vp); |
| 873 | vrecycle(vp); |
| 874 | node->tn_vnode = NULL; |
| 875 | } |
| 876 | #endif |
| 877 | TMPFS_NODE_UNLOCK(node); |
| 878 | --node->tn_links; |
| 879 | } |
| 880 | |
| 881 | /* |
| 882 | * Now get rid of all nodes. We can remove any node with a |
| 883 | * link count of 0 or any directory node with a link count of |
| 884 | * 1. The parents will not be destroyed until all their children |
| 885 | * have been destroyed. |
| 886 | * |
| 887 | * Recursion in tmpfs_free_node() can further modify the list so |
| 888 | * we cannot use a next pointer here. |
| 889 | * |
| 890 | * The root node will be destroyed by this loop (it will be last). |
| 891 | */ |
| 892 | while (!LIST_EMPTY(&tmp->tm_nodes_used)) { |
| 893 | found = 0; |
| 894 | LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) { |
| 895 | if (node->tn_links == 0 || |
| 896 | (node->tn_links == 1 && node->tn_type == VDIR)) { |
| 897 | TMPFS_NODE_LOCK(node); |
| 898 | tmpfs_free_node(tmp, node); |
| 899 | /* eats lock */ |
| 900 | found = 1; |
| 901 | break; |
| 902 | } |
| 903 | } |
| 904 | if (found == 0) { |
| 905 | kprintf("tmpfs: Cannot free entire node tree!"); |
| 906 | break; |
| 907 | } |
| 908 | } |
| 909 | |
| 910 | KKASSERT(tmp->tm_root == NULL); |
| 911 | |
| 912 | objcache_destroy(tmp->tm_dirent_pool); |
| 913 | objcache_destroy(tmp->tm_node_pool); |
| 914 | |
| 915 | kmalloc_destroy(&tmp->tm_name_zone); |
| 916 | kmalloc_destroy(&tmp->tm_dirent_zone); |
| 917 | kmalloc_destroy(&tmp->tm_node_zone); |
| 918 | |
| 919 | tmp->tm_node_zone = tmp->tm_dirent_zone = NULL; |
| 920 | |
| 921 | lockuninit(&tmp->allnode_lock); |
| 922 | KKASSERT(tmp->tm_pages_used == 0); |
| 923 | KKASSERT(tmp->tm_nodes_inuse == 0); |
| 924 | |
| 925 | /* Throw away the hammer2_mount structure. */ |
| 926 | kfree(tmp, M_HAMMER2); |
| 927 | mp->mnt_data = NULL; |
| 928 | |
| 929 | mp->mnt_flag &= ~MNT_LOCAL; |
| 930 | return 0; |
| 931 | } |
| 932 | |
| 933 | /* --------------------------------------------------------------------- */ |
| 934 | |
| 935 | static int |
| 936 | tmpfs_root(struct mount *mp, struct vnode **vpp) |
| 937 | { |
| 938 | struct hammer2_mount *tmp; |
| 939 | int error; |
| 940 | |
| 941 | kprintf("tmpfs_root\n"); |
| 942 | |
| 943 | tmp = VFS_TO_TMPFS(mp); |
| 944 | if (tmp->tm_root == NULL) { |
| 945 | kprintf("tmpfs_root: called without root node %p\n", mp); |
| 946 | print_backtrace(-1); |
| 947 | *vpp = NULL; |
| 948 | error = EINVAL; |
| 949 | } else { |
| 950 | error = tmpfs_alloc_vp(mp, tmp->tm_root, LK_EXCLUSIVE, vpp); |
| 951 | (*vpp)->v_flag |= VROOT; |
| 952 | (*vpp)->v_type = VDIR; |
| 953 | } |
| 954 | return error; |
| 955 | } |
| 956 | |
| 957 | /* --------------------------------------------------------------------- */ |
| 958 | |
| 959 | static int |
| 960 | tmpfs_fhtovp(struct mount *mp, struct vnode *rootvp, struct fid *fhp, struct vnode **vpp) |
| 961 | { |
| 962 | boolean_t found; |
| 963 | struct tmpfs_fid *tfhp; |
| 964 | struct hammer2_mount *tmp; |
| 965 | struct tmpfs_node *node; |
| 966 | |
| 967 | tmp = VFS_TO_TMPFS(mp); |
| 968 | |
| 969 | tfhp = (struct tmpfs_fid *)fhp; |
| 970 | if (tfhp->tf_len != sizeof(struct tmpfs_fid)) |
| 971 | return EINVAL; |
| 972 | |
| 973 | if (tfhp->tf_id >= tmp->tm_nodes_max) |
| 974 | return EINVAL; |
| 975 | |
| 976 | found = FALSE; |
| 977 | |
| 978 | TMPFS_LOCK(tmp); |
| 979 | LIST_FOREACH(node, &tmp->tm_nodes_used, tn_entries) { |
| 980 | if (node->tn_id == tfhp->tf_id && |
| 981 | node->tn_gen == tfhp->tf_gen) { |
| 982 | found = TRUE; |
| 983 | break; |
| 984 | } |
| 985 | } |
| 986 | TMPFS_UNLOCK(tmp); |
| 987 | |
| 988 | if (found) |
| 989 | return (tmpfs_alloc_vp(mp, node, LK_EXCLUSIVE, vpp)); |
| 990 | |
| 991 | return (EINVAL); |
| 992 | } |
| 993 | |
| 994 | /* --------------------------------------------------------------------- */ |
| 995 | |
| 996 | static int |
| 997 | tmpfs_vptofh(struct vnode *vp, struct fid *fhp) |
| 998 | { |
| 999 | struct tmpfs_node *node; |
| 1000 | struct tmpfs_fid tfh; |
| 1001 | node = VP_TO_TMPFS_NODE(vp); |
| 1002 | memset(&tfh, 0, sizeof(tfh)); |
| 1003 | tfh.tf_len = sizeof(struct tmpfs_fid); |
| 1004 | tfh.tf_gen = node->tn_gen; |
| 1005 | tfh.tf_id = node->tn_id; |
| 1006 | memcpy(fhp, &tfh, sizeof(tfh)); |
| 1007 | return (0); |
| 1008 | } |