ffs_vfsops.c revision 34961
1/* 2 * Copyright (c) 1989, 1991, 1993, 1994 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95 34 * $Id: ffs_vfsops.c,v 1.77 1998/03/27 14:20:57 peter Exp $ 35 */ 36 37#include "opt_quota.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/namei.h> 42#include <sys/proc.h> 43#include <sys/kernel.h> 44#include <sys/vnode.h> 45#include <sys/mount.h> 46#include <sys/buf.h> 47#include <sys/conf.h> 48#include <sys/fcntl.h> 49#include <sys/disklabel.h> 50#include <sys/malloc.h> 51 52#include <miscfs/specfs/specdev.h> 53 54#include <ufs/ufs/quota.h> 55#include <ufs/ufs/ufsmount.h> 56#include <ufs/ufs/inode.h> 57#include <ufs/ufs/ufs_extern.h> 58 59#include <ufs/ffs/fs.h> 60#include <ufs/ffs/ffs_extern.h> 61 62#include <vm/vm.h> 63#include <vm/vm_prot.h> 64#include <vm/vm_page.h> 65#include <vm/vm_extern.h> 66#include <vm/vm_object.h> 67 68static MALLOC_DEFINE(M_FFSNODE, "FFS node", "FFS vnode private part"); 69 70static int ffs_sbupdate __P((struct ufsmount *, int)); 71static int ffs_reload __P((struct mount *,struct ucred *,struct proc *)); 72static int ffs_oldfscompat __P((struct fs *)); 73static int ffs_mount __P((struct mount *, char *, caddr_t, 74 struct nameidata *, struct proc *)); 75static int ffs_init __P((struct vfsconf *)); 76 77static struct vfsops ufs_vfsops = { 78 ffs_mount, 79 ufs_start, 80 ffs_unmount, 81 ufs_root, 82 ufs_quotactl, 83 ffs_statfs, 84 ffs_sync, 85 ffs_vget, 86 vfs_vrele, 87 ffs_fhtovp, 88 ffs_vptofh, 89 ffs_init, 90}; 91 92VFS_SET(ufs_vfsops, ufs, MOUNT_UFS, 0); 93 94/* 95 * ffs_mount 96 * 97 * Called when mounting local physical media 98 * 99 * PARAMETERS: 100 * mountroot 101 * mp mount point structure 102 * path NULL (flag for root mount!!!) 103 * data <unused> 104 * ndp <unused> 105 * p process (user credentials check [statfs]) 106 * 107 * mount 108 * mp mount point structure 109 * path path to mount point 110 * data pointer to argument struct in user space 111 * ndp mount point namei() return (used for 112 * credentials on reload), reused to look 113 * up block device. 114 * p process (user credentials check) 115 * 116 * RETURNS: 0 Success 117 * !0 error number (errno.h) 118 * 119 * LOCK STATE: 120 * 121 * ENTRY 122 * mount point is locked 123 * EXIT 124 * mount point is locked 125 * 126 * NOTES: 127 * A NULL path can be used for a flag since the mount 128 * system call will fail with EFAULT in copyinstr in 129 * namei() if it is a genuine NULL from the user. 130 */ 131static int 132ffs_mount( mp, path, data, ndp, p) 133 struct mount *mp; /* mount struct pointer*/ 134 char *path; /* path to mount point*/ 135 caddr_t data; /* arguments to FS specific mount*/ 136 struct nameidata *ndp; /* mount point credentials*/ 137 struct proc *p; /* process requesting mount*/ 138{ 139 u_int size; 140 int err = 0; 141 struct vnode *devvp; 142 143 struct ufs_args args; 144 struct ufsmount *ump = 0; 145 register struct fs *fs; 146 int error, flags; 147 mode_t accessmode; 148 149 /* 150 * Use NULL path to flag a root mount 151 */ 152 if( path == NULL) { 153 /* 154 *** 155 * Mounting root file system 156 *** 157 */ 158 159 if ((err = bdevvp(rootdev, &rootvp))) { 160 printf("ffs_mountroot: can't find rootvp"); 161 return (err); 162 } 163 164 if (bdevsw[major(rootdev)]->d_flags & D_NOCLUSTERR) 165 mp->mnt_flag |= MNT_NOCLUSTERR; 166 if (bdevsw[major(rootdev)]->d_flags & D_NOCLUSTERW) 167 mp->mnt_flag |= MNT_NOCLUSTERW; 168 if( ( err = ffs_mountfs(rootvp, mp, p, M_FFSNODE)) != 0) { 169 /* fs specific cleanup (if any)*/ 170 goto error_1; 171 } 172 173 goto dostatfs; /* success*/ 174 175 } 176 177 /* 178 *** 179 * Mounting non-root file system or updating a file system 180 *** 181 */ 182 183 /* copy in user arguments*/ 184 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)); 185 if (err) 186 goto error_1; /* can't get arguments*/ 187 188 /* 189 * If updating, check whether changing from read-only to 190 * read/write; if there is no device name, that's all we do. 191 * Disallow clearing MNT_NOCLUSTERR and MNT_NOCLUSTERW flags, 192 * if block device requests. 193 */ 194 if (mp->mnt_flag & MNT_UPDATE) { 195 ump = VFSTOUFS(mp); 196 fs = ump->um_fs; 197 devvp = ump->um_devvp; 198 err = 0; 199 if (bdevsw[major(ump->um_dev)]->d_flags & D_NOCLUSTERR) 200 mp->mnt_flag |= MNT_NOCLUSTERR; 201 if (bdevsw[major(ump->um_dev)]->d_flags & D_NOCLUSTERW) 202 mp->mnt_flag |= MNT_NOCLUSTERW; 203 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 204 flags = WRITECLOSE; 205 if (mp->mnt_flag & MNT_FORCE) 206 flags |= FORCECLOSE; 207 if (mp->mnt_flag & MNT_SOFTDEP) { 208 err = softdep_flushfiles(mp, flags, p); 209 } else { 210 err = ffs_flushfiles(mp, flags, p); 211 } 212 } 213 if (!err && (mp->mnt_flag & MNT_RELOAD)) 214 err = ffs_reload(mp, ndp->ni_cnd.cn_cred, p); 215 if (err) { 216 goto error_1; 217 } 218 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 219 if (!fs->fs_clean) { 220 if (mp->mnt_flag & MNT_FORCE) { 221 printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt); 222 } else { 223 printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n", 224 fs->fs_fsmnt); 225 err = EPERM; 226 goto error_1; 227 } 228 } 229 230 /* 231 * If upgrade to read-write by non-root, then verify 232 * that user has necessary permissions on the device. 233 */ 234 if (p->p_ucred->cr_uid != 0) { 235 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 236 if (error = VOP_ACCESS(devvp, VREAD | VWRITE, 237 p->p_ucred, p)) { 238 VOP_UNLOCK(devvp, 0, p); 239 return (error); 240 } 241 VOP_UNLOCK(devvp, 0, p); 242 } 243 244 /* check to see if we need to start softdep */ 245 if (fs->fs_flags & FS_DOSOFTDEP) { 246 err = softdep_mount(devvp, mp, fs, p->p_ucred); 247 if (err) 248 goto error_1; 249 } 250 251 fs->fs_ronly = 0; 252 } 253 if (fs->fs_ronly == 0) { 254 fs->fs_clean = 0; 255 ffs_sbupdate(ump, MNT_WAIT); 256 } 257 /* if not updating name...*/ 258 if (args.fspec == 0) { 259 /* 260 * Process export requests. Jumping to "success" 261 * will return the vfs_export() error code. 262 */ 263 err = vfs_export(mp, &ump->um_export, &args.export); 264 goto success; 265 } 266 } 267 268 /* 269 * Not an update, or updating the name: look up the name 270 * and verify that it refers to a sensible block device. 271 */ 272 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 273 err = namei(ndp); 274 if (err) { 275 /* can't get devvp!*/ 276 goto error_1; 277 } 278 279 devvp = ndp->ni_vp; 280 281 if (devvp->v_type != VBLK) { 282 err = ENOTBLK; 283 goto error_2; 284 } 285 if (major(devvp->v_rdev) >= nblkdev) { 286 err = ENXIO; 287 goto error_2; 288 } 289 290 /* 291 * If mount by non-root, then verify that user has necessary 292 * permissions on the device. 293 */ 294 if (p->p_ucred->cr_uid != 0) { 295 accessmode = VREAD; 296 if ((mp->mnt_flag & MNT_RDONLY) == 0) 297 accessmode |= VWRITE; 298 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 299 if (error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) { 300 vput(devvp); 301 return (error); 302 } 303 VOP_UNLOCK(devvp, 0, p); 304 } 305 306 if (mp->mnt_flag & MNT_UPDATE) { 307 /* 308 ******************** 309 * UPDATE 310 ******************** 311 */ 312 313 if (devvp != ump->um_devvp) 314 err = EINVAL; /* needs translation */ 315 else 316 vrele(devvp); 317 /* 318 * Update device name only on success 319 */ 320 if( !err) { 321 /* Save "mounted from" info for mount point (NULL pad)*/ 322 copyinstr( args.fspec, 323 mp->mnt_stat.f_mntfromname, 324 MNAMELEN - 1, 325 &size); 326 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 327 } 328 } else { 329 /* 330 ******************** 331 * NEW MOUNT 332 ******************** 333 */ 334 335 if (bdevsw[major(devvp->v_rdev)]->d_flags & D_NOCLUSTERR) 336 mp->mnt_flag |= MNT_NOCLUSTERR; 337 if (bdevsw[major(devvp->v_rdev)]->d_flags & D_NOCLUSTERW) 338 mp->mnt_flag |= MNT_NOCLUSTERW; 339 340 /* 341 * Since this is a new mount, we want the names for 342 * the device and the mount point copied in. If an 343 * error occurs, the mountpoint is discarded by the 344 * upper level code. 345 */ 346 /* Save "last mounted on" info for mount point (NULL pad)*/ 347 copyinstr( path, /* mount point*/ 348 mp->mnt_stat.f_mntonname, /* save area*/ 349 MNAMELEN - 1, /* max size*/ 350 &size); /* real size*/ 351 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size); 352 353 /* Save "mounted from" info for mount point (NULL pad)*/ 354 copyinstr( args.fspec, /* device name*/ 355 mp->mnt_stat.f_mntfromname, /* save area*/ 356 MNAMELEN - 1, /* max size*/ 357 &size); /* real size*/ 358 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 359 360 err = ffs_mountfs(devvp, mp, p, M_FFSNODE); 361 } 362 if (err) { 363 goto error_2; 364 } 365 366dostatfs: 367 /* 368 * Initialize FS stat information in mount struct; uses both 369 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname 370 * 371 * This code is common to root and non-root mounts 372 */ 373 (void)VFS_STATFS(mp, &mp->mnt_stat, p); 374 375 goto success; 376 377 378error_2: /* error with devvp held*/ 379 380 /* release devvp before failing*/ 381 vrele(devvp); 382 383error_1: /* no state to back out*/ 384 385success: 386 return( err); 387} 388 389/* 390 * Reload all incore data for a filesystem (used after running fsck on 391 * the root filesystem and finding things to fix). The filesystem must 392 * be mounted read-only. 393 * 394 * Things to do to update the mount: 395 * 1) invalidate all cached meta-data. 396 * 2) re-read superblock from disk. 397 * 3) re-read summary information from disk. 398 * 4) invalidate all inactive vnodes. 399 * 5) invalidate all cached file data. 400 * 6) re-read inode data for all active vnodes. 401 */ 402static int 403ffs_reload(mp, cred, p) 404 register struct mount *mp; 405 struct ucred *cred; 406 struct proc *p; 407{ 408 register struct vnode *vp, *nvp, *devvp; 409 struct inode *ip; 410 struct csum *space; 411 struct buf *bp; 412 struct fs *fs, *newfs; 413 struct partinfo dpart; 414 dev_t dev; 415 int i, blks, size, error; 416 int32_t *lp; 417 418 if ((mp->mnt_flag & MNT_RDONLY) == 0) 419 return (EINVAL); 420 /* 421 * Step 1: invalidate all cached meta-data. 422 */ 423 devvp = VFSTOUFS(mp)->um_devvp; 424 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 425 error = vinvalbuf(devvp, 0, cred, p, 0, 0); 426 VOP_UNLOCK(devvp, 0, p); 427 if (error) 428 panic("ffs_reload: dirty1"); 429 430 dev = devvp->v_rdev; 431 /* 432 * Only VMIO the backing device if the backing device is a real 433 * block device. This excludes the original MFS implementation. 434 * Note that it is optional that the backing device be VMIOed. This 435 * increases the opportunity for metadata caching. 436 */ 437 if ((devvp->v_type == VBLK) && (major(dev) < nblkdev)) { 438 simple_lock(&devvp->v_interlock); 439 vfs_object_create(devvp, p, p->p_ucred, 0); 440 } 441 442 /* 443 * Step 2: re-read superblock from disk. 444 */ 445 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 446 size = DEV_BSIZE; 447 else 448 size = dpart.disklab->d_secsize; 449 if (error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, NOCRED,&bp)) 450 return (error); 451 newfs = (struct fs *)bp->b_data; 452 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE || 453 newfs->fs_bsize < sizeof(struct fs)) { 454 brelse(bp); 455 return (EIO); /* XXX needs translation */ 456 } 457 fs = VFSTOUFS(mp)->um_fs; 458 /* 459 * Copy pointer fields back into superblock before copying in XXX 460 * new superblock. These should really be in the ufsmount. XXX 461 * Note that important parameters (eg fs_ncg) are unchanged. 462 */ 463 bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp)); 464 newfs->fs_maxcluster = fs->fs_maxcluster; 465 bcopy(newfs, fs, (u_int)fs->fs_sbsize); 466 if (fs->fs_sbsize < SBSIZE) 467 bp->b_flags |= B_INVAL; 468 brelse(bp); 469 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 470 ffs_oldfscompat(fs); 471 472 /* 473 * Step 3: re-read summary information from disk. 474 */ 475 blks = howmany(fs->fs_cssize, fs->fs_fsize); 476 space = fs->fs_csp[0]; 477 for (i = 0; i < blks; i += fs->fs_frag) { 478 size = fs->fs_bsize; 479 if (i + fs->fs_frag > blks) 480 size = (blks - i) * fs->fs_fsize; 481 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 482 NOCRED, &bp); 483 if (error) 484 return (error); 485 bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size); 486 brelse(bp); 487 } 488 /* 489 * We no longer know anything about clusters per cylinder group. 490 */ 491 if (fs->fs_contigsumsize > 0) { 492 lp = fs->fs_maxcluster; 493 for (i = 0; i < fs->fs_ncg; i++) 494 *lp++ = fs->fs_contigsumsize; 495 } 496 497loop: 498 simple_lock(&mntvnode_slock); 499 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 500 if (vp->v_mount != mp) { 501 simple_unlock(&mntvnode_slock); 502 goto loop; 503 } 504 nvp = vp->v_mntvnodes.le_next; 505 /* 506 * Step 4: invalidate all inactive vnodes. 507 */ 508 if (vrecycle(vp, &mntvnode_slock, p)) 509 goto loop; 510 /* 511 * Step 5: invalidate all cached file data. 512 */ 513 simple_lock(&vp->v_interlock); 514 simple_unlock(&mntvnode_slock); 515 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 516 goto loop; 517 } 518 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 519 panic("ffs_reload: dirty2"); 520 /* 521 * Step 6: re-read inode data for all active vnodes. 522 */ 523 ip = VTOI(vp); 524 error = 525 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 526 (int)fs->fs_bsize, NOCRED, &bp); 527 if (error) { 528 vput(vp); 529 return (error); 530 } 531 ip->i_din = *((struct dinode *)bp->b_data + 532 ino_to_fsbo(fs, ip->i_number)); 533 ip->i_effnlink = ip->i_nlink; 534 brelse(bp); 535 vput(vp); 536 simple_lock(&mntvnode_slock); 537 } 538 simple_unlock(&mntvnode_slock); 539 return (0); 540} 541 542/* 543 * Common code for mount and mountroot 544 */ 545int 546ffs_mountfs(devvp, mp, p, malloctype) 547 register struct vnode *devvp; 548 struct mount *mp; 549 struct proc *p; 550 struct malloc_type *malloctype; 551{ 552 register struct ufsmount *ump; 553 struct buf *bp; 554 register struct fs *fs; 555 struct cg *cgp; 556 dev_t dev; 557 struct partinfo dpart; 558 struct csum cstotal; 559 caddr_t base, space; 560 int error, i, cyl, blks, size, ronly; 561 int32_t *lp; 562 struct ucred *cred; 563 u_int64_t maxfilesize; /* XXX */ 564 u_int strsize; 565 int ncount; 566 567 dev = devvp->v_rdev; 568 cred = p ? p->p_ucred : NOCRED; 569 /* 570 * Disallow multiple mounts of the same device. 571 * Disallow mounting of a device that is currently in use 572 * (except for root, which might share swap device for miniroot). 573 * Flush out any old buffers remaining from a previous use. 574 */ 575 error = vfs_mountedon(devvp); 576 if (error) 577 return (error); 578 ncount = vcount(devvp); 579 580 if (ncount > 1 && devvp != rootvp) 581 return (EBUSY); 582 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 583 error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0); 584 VOP_UNLOCK(devvp, 0, p); 585 if (error) 586 return (error); 587 588 /* 589 * Only VMIO the backing device if the backing device is a real 590 * block device. This excludes the original MFS implementation. 591 * Note that it is optional that the backing device be VMIOed. This 592 * increases the opportunity for metadata caching. 593 */ 594 if ((devvp->v_type == VBLK) && (major(dev) < nblkdev)) { 595 simple_lock(&devvp->v_interlock); 596 vfs_object_create(devvp, p, p->p_ucred, 0); 597 } 598 599 600 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 601 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p); 602 if (error) 603 return (error); 604 605 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0) 606 size = DEV_BSIZE; 607 else 608 size = dpart.disklab->d_secsize; 609 610 bp = NULL; 611 ump = NULL; 612 if (error = bread(devvp, SBLOCK, SBSIZE, cred, &bp)) 613 goto out; 614 fs = (struct fs *)bp->b_data; 615 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE || 616 fs->fs_bsize < sizeof(struct fs)) { 617 error = EINVAL; /* XXX needs translation */ 618 goto out; 619 } 620 fs->fs_fmod = 0; 621 if (!fs->fs_clean) { 622 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 623 printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt); 624 } else { 625 printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt); 626 error = EPERM; 627 goto out; 628 } 629 } 630 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */ 631 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) { 632 error = EROFS; /* needs translation */ 633 goto out; 634 } 635 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 636 bzero((caddr_t)ump, sizeof *ump); 637 ump->um_malloctype = malloctype; 638 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT, 639 M_WAITOK); 640 ump->um_blkatoff = ffs_blkatoff; 641 ump->um_truncate = ffs_truncate; 642 ump->um_update = ffs_update; 643 ump->um_valloc = ffs_valloc; 644 ump->um_vfree = ffs_vfree; 645 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize); 646 if (fs->fs_sbsize < SBSIZE) 647 bp->b_flags |= B_INVAL; 648 brelse(bp); 649 bp = NULL; 650 fs = ump->um_fs; 651 fs->fs_ronly = ronly; 652 if (ronly == 0) { 653 fs->fs_fmod = 1; 654 fs->fs_clean = 0; 655 } 656 size = fs->fs_cssize; 657 blks = howmany(size, fs->fs_fsize); 658 if (fs->fs_contigsumsize > 0) 659 size += fs->fs_ncg * sizeof(int32_t); 660 base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK); 661 for (i = 0; i < blks; i += fs->fs_frag) { 662 size = fs->fs_bsize; 663 if (i + fs->fs_frag > blks) 664 size = (blks - i) * fs->fs_fsize; 665 if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 666 cred, &bp)) { 667 free(base, M_UFSMNT); 668 goto out; 669 } 670 bcopy(bp->b_data, space, (u_int)size); 671 fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space; 672 space += size; 673 brelse(bp); 674 bp = NULL; 675 } 676 if (fs->fs_contigsumsize > 0) { 677 fs->fs_maxcluster = lp = (int32_t *)space; 678 for (i = 0; i < fs->fs_ncg; i++) 679 *lp++ = fs->fs_contigsumsize; 680 } 681 mp->mnt_data = (qaddr_t)ump; 682 mp->mnt_stat.f_fsid.val[0] = (long)dev; 683 if (fs->fs_id[0] != 0 && fs->fs_id[1] != 0) 684 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1]; 685 else 686 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 687 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 688 mp->mnt_flag |= MNT_LOCAL; 689 ump->um_mountp = mp; 690 ump->um_dev = dev; 691 ump->um_devvp = devvp; 692 ump->um_nindir = fs->fs_nindir; 693 ump->um_bptrtodb = fs->fs_fsbtodb; 694 ump->um_seqinc = fs->fs_frag; 695 for (i = 0; i < MAXQUOTAS; i++) 696 ump->um_quotas[i] = NULLVP; 697 devvp->v_specmountpoint = mp; 698 ffs_oldfscompat(fs); 699 700 /* 701 * Set FS local "last mounted on" information (NULL pad) 702 */ 703 copystr( mp->mnt_stat.f_mntonname, /* mount point*/ 704 fs->fs_fsmnt, /* copy area*/ 705 sizeof(fs->fs_fsmnt) - 1, /* max size*/ 706 &strsize); /* real size*/ 707 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize); 708 709 if( mp->mnt_flag & MNT_ROOTFS) { 710 /* 711 * Root mount; update timestamp in mount structure. 712 * this will be used by the common root mount code 713 * to update the system clock. 714 */ 715 mp->mnt_time = fs->fs_time; 716 } 717 718 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */ 719 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */ 720 if (fs->fs_maxfilesize > maxfilesize) /* XXX */ 721 fs->fs_maxfilesize = maxfilesize; /* XXX */ 722 if (ronly == 0) { 723 if ((fs->fs_flags & FS_DOSOFTDEP) && 724 (error = softdep_mount(devvp, mp, fs, cred)) != 0) { 725 free(base, M_UFSMNT); 726 goto out; 727 } 728 fs->fs_clean = 0; 729 (void) ffs_sbupdate(ump, MNT_WAIT); 730 } 731 return (0); 732out: 733 devvp->v_specmountpoint = NULL; 734 if (bp) 735 brelse(bp); 736 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p); 737 if (ump) { 738 free(ump->um_fs, M_UFSMNT); 739 free(ump, M_UFSMNT); 740 mp->mnt_data = (qaddr_t)0; 741 } 742 return (error); 743} 744 745/* 746 * Sanity checks for old file systems. 747 * 748 * XXX - goes away some day. 749 */ 750static int 751ffs_oldfscompat(fs) 752 struct fs *fs; 753{ 754 755 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */ 756 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */ 757 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */ 758 fs->fs_nrpos = 8; /* XXX */ 759 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 760#if 0 761 int i; /* XXX */ 762 u_int64_t sizepb = fs->fs_bsize; /* XXX */ 763 /* XXX */ 764 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */ 765 for (i = 0; i < NIADDR; i++) { /* XXX */ 766 sizepb *= NINDIR(fs); /* XXX */ 767 fs->fs_maxfilesize += sizepb; /* XXX */ 768 } /* XXX */ 769#endif 770 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 771 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */ 772 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */ 773 } /* XXX */ 774 return (0); 775} 776 777/* 778 * unmount system call 779 */ 780int 781ffs_unmount(mp, mntflags, p) 782 struct mount *mp; 783 int mntflags; 784 struct proc *p; 785{ 786 register struct ufsmount *ump; 787 register struct fs *fs; 788 int error, flags; 789 790 flags = 0; 791 if (mntflags & MNT_FORCE) { 792 flags |= FORCECLOSE; 793 } 794 if (mp->mnt_flag & MNT_SOFTDEP) { 795 if ((error = softdep_flushfiles(mp, flags, p)) != 0) 796 return (error); 797 } else { 798 if ((error = ffs_flushfiles(mp, flags, p)) != 0) 799 return (error); 800 } 801 ump = VFSTOUFS(mp); 802 fs = ump->um_fs; 803 if (fs->fs_ronly == 0) { 804 fs->fs_clean = 1; 805 error = ffs_sbupdate(ump, MNT_WAIT); 806 if (error) { 807 fs->fs_clean = 0; 808 return (error); 809 } 810 } 811 ump->um_devvp->v_specmountpoint = NULL; 812 813 vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, p, 0, 0); 814 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, 815 NOCRED, p); 816 817 vrele(ump->um_devvp); 818 819 free(fs->fs_csp[0], M_UFSMNT); 820 free(fs, M_UFSMNT); 821 free(ump, M_UFSMNT); 822 mp->mnt_data = (qaddr_t)0; 823 mp->mnt_flag &= ~MNT_LOCAL; 824 return (error); 825} 826 827/* 828 * Flush out all the files in a filesystem. 829 */ 830int 831ffs_flushfiles(mp, flags, p) 832 register struct mount *mp; 833 int flags; 834 struct proc *p; 835{ 836 register struct ufsmount *ump; 837 int error; 838 839 ump = VFSTOUFS(mp); 840#ifdef QUOTA 841 if (mp->mnt_flag & MNT_QUOTA) { 842 int i; 843 error = vflush(mp, NULLVP, SKIPSYSTEM|flags); 844 if (error) 845 return (error); 846 for (i = 0; i < MAXQUOTAS; i++) { 847 if (ump->um_quotas[i] == NULLVP) 848 continue; 849 quotaoff(p, mp, i); 850 } 851 /* 852 * Here we fall through to vflush again to ensure 853 * that we have gotten rid of all the system vnodes. 854 */ 855 } 856#endif 857 /* 858 * Flush all the files. 859 */ 860 if ((error = vflush(mp, NULL, flags)) != 0) 861 return (error); 862 /* 863 * Flush filesystem metadata. 864 */ 865 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p); 866 error = VOP_FSYNC(ump->um_devvp, p->p_ucred, MNT_WAIT, p); 867 VOP_UNLOCK(ump->um_devvp, 0, p); 868 return (error); 869} 870 871/* 872 * Get file system statistics. 873 */ 874int 875ffs_statfs(mp, sbp, p) 876 struct mount *mp; 877 register struct statfs *sbp; 878 struct proc *p; 879{ 880 register struct ufsmount *ump; 881 register struct fs *fs; 882 883 ump = VFSTOUFS(mp); 884 fs = ump->um_fs; 885 if (fs->fs_magic != FS_MAGIC) 886 panic("ffs_statfs"); 887 sbp->f_bsize = fs->fs_fsize; 888 sbp->f_iosize = fs->fs_bsize; 889 sbp->f_blocks = fs->fs_dsize; 890 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag + 891 fs->fs_cstotal.cs_nffree; 892 sbp->f_bavail = freespace(fs, fs->fs_minfree); 893 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO; 894 sbp->f_ffree = fs->fs_cstotal.cs_nifree; 895 if (sbp != &mp->mnt_stat) { 896 sbp->f_type = mp->mnt_vfc->vfc_typenum; 897 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 898 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 899 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 900 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 901 } 902 return (0); 903} 904 905/* 906 * Go through the disk queues to initiate sandbagged IO; 907 * go through the inodes to write those that have been modified; 908 * initiate the writing of the super block if it has been modified. 909 * 910 * Note: we are always called with the filesystem marked `MPBUSY'. 911 */ 912int 913ffs_sync(mp, waitfor, cred, p) 914 struct mount *mp; 915 int waitfor; 916 struct ucred *cred; 917 struct proc *p; 918{ 919 struct vnode *nvp, *vp; 920 struct inode *ip; 921 struct ufsmount *ump = VFSTOUFS(mp); 922 struct fs *fs; 923 struct timeval tv; 924 int error, allerror = 0; 925 926 fs = ump->um_fs; 927 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 928 printf("fs = %s\n", fs->fs_fsmnt); 929 panic("ffs_sync: rofs mod"); 930 } 931 /* 932 * Write back each (modified) inode. 933 */ 934 simple_lock(&mntvnode_slock); 935loop: 936 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 937 /* 938 * If the vnode that we are about to sync is no longer 939 * associated with this mount point, start over. 940 */ 941 if (vp->v_mount != mp) 942 goto loop; 943 simple_lock(&vp->v_interlock); 944 nvp = vp->v_mntvnodes.le_next; 945 ip = VTOI(vp); 946 if ((vp->v_type == VNON) || ((ip->i_flag & 947 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0) && 948 ((vp->v_dirtyblkhd.lh_first == NULL) || (waitfor == MNT_LAZY))) { 949 simple_unlock(&vp->v_interlock); 950 continue; 951 } 952 if (vp->v_type != VCHR) { 953 simple_unlock(&mntvnode_slock); 954 error = 955 vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p); 956 if (error) { 957 simple_lock(&mntvnode_slock); 958 if (error == ENOENT) 959 goto loop; 960 continue; 961 } 962 if (error = VOP_FSYNC(vp, cred, waitfor, p)) 963 allerror = error; 964 VOP_UNLOCK(vp, 0, p); 965 vrele(vp); 966 simple_lock(&mntvnode_slock); 967 } else { 968 simple_unlock(&mntvnode_slock); 969 simple_unlock(&vp->v_interlock); 970 getmicrotime(&tv); 971 /* UFS_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT); */ 972 UFS_UPDATE(vp, &tv, &tv, 0); 973 simple_lock(&mntvnode_slock); 974 } 975 } 976 simple_unlock(&mntvnode_slock); 977 /* 978 * Force stale file system control information to be flushed. 979 */ 980 if (waitfor != MNT_LAZY) { 981 if (ump->um_mountp->mnt_flag & MNT_SOFTDEP) 982 waitfor = MNT_NOWAIT; 983 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p); 984 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0) 985 allerror = error; 986 VOP_UNLOCK(ump->um_devvp, 0, p); 987 } 988#ifdef QUOTA 989 qsync(mp); 990#endif 991 /* 992 * Write back modified superblock. 993 */ 994 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0) 995 allerror = error; 996 return (allerror); 997} 998 999/* 1000 * Look up a FFS dinode number to find its incore vnode, otherwise read it 1001 * in from disk. If it is in core, wait for the lock bit to clear, then 1002 * return the inode locked. Detection and handling of mount points must be 1003 * done by the calling routine. 1004 */ 1005static int ffs_inode_hash_lock; 1006 1007int 1008ffs_vget(mp, ino, vpp) 1009 struct mount *mp; 1010 ino_t ino; 1011 struct vnode **vpp; 1012{ 1013 struct fs *fs; 1014 struct inode *ip; 1015 struct ufsmount *ump; 1016 struct buf *bp; 1017 struct vnode *vp; 1018 dev_t dev; 1019 int error; 1020 1021 ump = VFSTOUFS(mp); 1022 dev = ump->um_dev; 1023restart: 1024 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) { 1025 return (0); 1026 } 1027 1028 /* 1029 * Lock out the creation of new entries in the FFS hash table in 1030 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 1031 * may occur! 1032 */ 1033 if (ffs_inode_hash_lock) { 1034 while (ffs_inode_hash_lock) { 1035 ffs_inode_hash_lock = -1; 1036 tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0); 1037 } 1038 goto restart; 1039 } 1040 ffs_inode_hash_lock = 1; 1041 1042 /* 1043 * If this MALLOC() is performed after the getnewvnode() 1044 * it might block, leaving a vnode with a NULL v_data to be 1045 * found by ffs_sync() if a sync happens to fire right then, 1046 * which will cause a panic because ffs_sync() blindly 1047 * dereferences vp->v_data (as well it should). 1048 */ 1049 MALLOC(ip, struct inode *, sizeof(struct inode), 1050 ump->um_malloctype, M_WAITOK); 1051 1052 /* Allocate a new vnode/inode. */ 1053 error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp); 1054 if (error) { 1055 if (ffs_inode_hash_lock < 0) 1056 wakeup(&ffs_inode_hash_lock); 1057 ffs_inode_hash_lock = 0; 1058 *vpp = NULL; 1059 FREE(ip, ump->um_malloctype); 1060 return (error); 1061 } 1062 bzero((caddr_t)ip, sizeof(struct inode)); 1063 lockinit(&ip->i_lock, PINOD, "inode", 0, 0); 1064 vp->v_data = ip; 1065 ip->i_vnode = vp; 1066 ip->i_fs = fs = ump->um_fs; 1067 ip->i_dev = dev; 1068 ip->i_number = ino; 1069#ifdef QUOTA 1070 { 1071 int i; 1072 for (i = 0; i < MAXQUOTAS; i++) 1073 ip->i_dquot[i] = NODQUOT; 1074 } 1075#endif 1076 /* 1077 * Put it onto its hash chain and lock it so that other requests for 1078 * this inode will block if they arrive while we are sleeping waiting 1079 * for old data structures to be purged or for the contents of the 1080 * disk portion of this inode to be read. 1081 */ 1082 ufs_ihashins(ip); 1083 1084 if (ffs_inode_hash_lock < 0) 1085 wakeup(&ffs_inode_hash_lock); 1086 ffs_inode_hash_lock = 0; 1087 1088 /* Read in the disk contents for the inode, copy into the inode. */ 1089 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1090 (int)fs->fs_bsize, NOCRED, &bp); 1091 if (error) { 1092 /* 1093 * The inode does not contain anything useful, so it would 1094 * be misleading to leave it on its hash chain. With mode 1095 * still zero, it will be unlinked and returned to the free 1096 * list by vput(). 1097 */ 1098 brelse(bp); 1099 vput(vp); 1100 *vpp = NULL; 1101 return (error); 1102 } 1103 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino)); 1104 if (DOINGSOFTDEP(vp)) 1105 softdep_load_inodeblock(ip); 1106 else 1107 ip->i_effnlink = ip->i_nlink; 1108 bqrelse(bp); 1109 1110 /* 1111 * Initialize the vnode from the inode, check for aliases. 1112 * Note that the underlying vnode may have changed. 1113 */ 1114 error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1115 if (error) { 1116 vput(vp); 1117 *vpp = NULL; 1118 return (error); 1119 } 1120 /* 1121 * Finish inode initialization now that aliasing has been resolved. 1122 */ 1123 ip->i_devvp = ump->um_devvp; 1124 VREF(ip->i_devvp); 1125 /* 1126 * Set up a generation number for this inode if it does not 1127 * already have one. This should only happen on old filesystems. 1128 */ 1129 if (ip->i_gen == 0) { 1130 ip->i_gen = random() / 2 + 1; 1131 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1132 ip->i_flag |= IN_MODIFIED; 1133 } 1134 /* 1135 * Ensure that uid and gid are correct. This is a temporary 1136 * fix until fsck has been changed to do the update. 1137 */ 1138 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 1139 ip->i_uid = ip->i_din.di_ouid; /* XXX */ 1140 ip->i_gid = ip->i_din.di_ogid; /* XXX */ 1141 } /* XXX */ 1142 1143 *vpp = vp; 1144 return (0); 1145} 1146 1147/* 1148 * File handle to vnode 1149 * 1150 * Have to be really careful about stale file handles: 1151 * - check that the inode number is valid 1152 * - call ffs_vget() to get the locked inode 1153 * - check for an unallocated inode (i_mode == 0) 1154 * - check that the given client host has export rights and return 1155 * those rights via. exflagsp and credanonp 1156 */ 1157int 1158ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) 1159 register struct mount *mp; 1160 struct fid *fhp; 1161 struct sockaddr *nam; 1162 struct vnode **vpp; 1163 int *exflagsp; 1164 struct ucred **credanonp; 1165{ 1166 register struct ufid *ufhp; 1167 struct fs *fs; 1168 1169 ufhp = (struct ufid *)fhp; 1170 fs = VFSTOUFS(mp)->um_fs; 1171 if (ufhp->ufid_ino < ROOTINO || 1172 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) 1173 return (ESTALE); 1174 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)); 1175} 1176 1177/* 1178 * Vnode pointer to File handle 1179 */ 1180/* ARGSUSED */ 1181int 1182ffs_vptofh(vp, fhp) 1183 struct vnode *vp; 1184 struct fid *fhp; 1185{ 1186 register struct inode *ip; 1187 register struct ufid *ufhp; 1188 1189 ip = VTOI(vp); 1190 ufhp = (struct ufid *)fhp; 1191 ufhp->ufid_len = sizeof(struct ufid); 1192 ufhp->ufid_ino = ip->i_number; 1193 ufhp->ufid_gen = ip->i_gen; 1194 return (0); 1195} 1196 1197/* 1198 * Initialize the filesystem; just use ufs_init. 1199 */ 1200static int 1201ffs_init(vfsp) 1202 struct vfsconf *vfsp; 1203{ 1204 1205 softdep_initialize(); 1206 return (ufs_init(vfsp)); 1207} 1208 1209/* 1210 * Write a superblock and associated information back to disk. 1211 */ 1212static int 1213ffs_sbupdate(mp, waitfor) 1214 struct ufsmount *mp; 1215 int waitfor; 1216{ 1217 register struct fs *dfs, *fs = mp->um_fs; 1218 register struct buf *bp; 1219 int blks; 1220 caddr_t space; 1221 int i, size, error, allerror = 0; 1222 1223 /* 1224 * First write back the summary information. 1225 */ 1226 blks = howmany(fs->fs_cssize, fs->fs_fsize); 1227 space = (caddr_t)fs->fs_csp[0]; 1228 for (i = 0; i < blks; i += fs->fs_frag) { 1229 size = fs->fs_bsize; 1230 if (i + fs->fs_frag > blks) 1231 size = (blks - i) * fs->fs_fsize; 1232 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), 1233 size, 0, 0); 1234 bcopy(space, bp->b_data, (u_int)size); 1235 space += size; 1236 if (waitfor != MNT_WAIT) 1237 bawrite(bp); 1238 else if (error = bwrite(bp)) 1239 allerror = error; 1240 } 1241 /* 1242 * Now write back the superblock itself. If any errors occurred 1243 * up to this point, then fail so that the superblock avoids 1244 * being written out as clean. 1245 */ 1246 if (allerror) 1247 return (allerror); 1248 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0); 1249 fs->fs_fmod = 0; 1250 fs->fs_time = time_second; 1251 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); 1252 /* Restore compatibility to old file systems. XXX */ 1253 dfs = (struct fs *)bp->b_data; /* XXX */ 1254 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */ 1255 dfs->fs_nrpos = -1; /* XXX */ 1256 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 1257 int32_t *lp, tmp; /* XXX */ 1258 /* XXX */ 1259 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */ 1260 tmp = lp[4]; /* XXX */ 1261 for (i = 4; i > 0; i--) /* XXX */ 1262 lp[i] = lp[i-1]; /* XXX */ 1263 lp[0] = tmp; /* XXX */ 1264 } /* XXX */ 1265 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */ 1266 if (waitfor != MNT_WAIT) 1267 bawrite(bp); 1268 else if (error = bwrite(bp)) 1269 allerror = error; 1270 return (allerror); 1271} 1272