ffs_vfsops.c revision 24203
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 1.51 1997/03/23 20:08:19 guido 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/socket.h> 46#include <sys/mount.h> 47#include <sys/buf.h> 48#include <sys/mbuf.h> 49#include <sys/fcntl.h> 50#include <sys/disklabel.h> 51#include <sys/errno.h> 52#include <sys/malloc.h> 53 54#include <miscfs/specfs/specdev.h> 55 56#include <ufs/ufs/quota.h> 57#include <ufs/ufs/ufsmount.h> 58#include <ufs/ufs/inode.h> 59#include <ufs/ufs/ufs_extern.h> 60 61#include <ufs/ffs/fs.h> 62#include <ufs/ffs/ffs_extern.h> 63 64#include <vm/vm.h> 65#include <vm/vm_param.h> 66#include <vm/vm_prot.h> 67#include <vm/vm_page.h> 68#include <vm/vm_object.h> 69#include <vm/vm_extern.h> 70 71static int ffs_sbupdate __P((struct ufsmount *, int)); 72static int ffs_reload __P((struct mount *,struct ucred *,struct proc *)); 73static int ffs_oldfscompat __P((struct fs *)); 74static int ffs_mount __P((struct mount *, char *, caddr_t, 75 struct nameidata *, struct proc *)); 76static int ffs_init __P((struct vfsconf *)); 77 78struct vfsops ufs_vfsops = { 79 ffs_mount, 80 ufs_start, 81 ffs_unmount, 82 ufs_root, 83 ufs_quotactl, 84 ffs_statfs, 85 ffs_sync, 86 ffs_vget, 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 flags; 147 148 /* 149 * Use NULL path to flag a root mount 150 */ 151 if( path == NULL) { 152 /* 153 *** 154 * Mounting root file system 155 *** 156 */ 157 158 /* Get vnode for root device*/ 159 if ((err = bdevvp( rootdev, &rootvp))) { 160 printf("ffs_mountroot: can't setup bdevvp for root"); 161 return (err); 162 } 163 164 /* 165 * Attempt mount 166 */ 167 if( ( err = ffs_mountfs(rootvp, mp, p)) != 0) { 168 /* fs specific cleanup (if any)*/ 169 goto error_1; 170 } 171 172 goto dostatfs; /* success*/ 173 174 } 175 176 /* 177 *** 178 * Mounting non-root file system or updating a file system 179 *** 180 */ 181 182 /* copy in user arguments*/ 183 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)); 184 if (err) 185 goto error_1; /* can't get arguments*/ 186 187 /* 188 * If updating, check whether changing from read-only to 189 * read/write; if there is no device name, that's all we do. 190 */ 191 if (mp->mnt_flag & MNT_UPDATE) { 192 ump = VFSTOUFS(mp); 193 fs = ump->um_fs; 194 err = 0; 195 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 196 flags = WRITECLOSE; 197 if (mp->mnt_flag & MNT_FORCE) 198 flags |= FORCECLOSE; 199 err = ffs_flushfiles(mp, flags, p); 200 } 201 if (!err && (mp->mnt_flag & MNT_RELOAD)) 202 err = ffs_reload(mp, ndp->ni_cnd.cn_cred, p); 203 if (err) { 204 goto error_1; 205 } 206 if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) { 207 if (!fs->fs_clean) { 208 if (mp->mnt_flag & MNT_FORCE) { 209 printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt); 210 } else { 211 printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n", 212 fs->fs_fsmnt); 213 err = EPERM; 214 goto error_1; 215 } 216 } 217 fs->fs_ronly = 0; 218 } 219 if (fs->fs_ronly == 0) { 220 fs->fs_clean = 0; 221 ffs_sbupdate(ump, MNT_WAIT); 222 } 223 /* if not updating name...*/ 224 if (args.fspec == 0) { 225 /* 226 * Process export requests. Jumping to "success" 227 * will return the vfs_export() error code. 228 */ 229 err = vfs_export(mp, &ump->um_export, &args.export); 230 goto success; 231 } 232 } 233 234 /* 235 * Not an update, or updating the name: look up the name 236 * and verify that it refers to a sensible block device. 237 */ 238 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 239 err = namei(ndp); 240 if (err) { 241 /* can't get devvp!*/ 242 goto error_1; 243 } 244 245 devvp = ndp->ni_vp; 246 247 if (devvp->v_type != VBLK) { 248 err = ENOTBLK; 249 goto error_2; 250 } 251 if (major(devvp->v_rdev) >= nblkdev) { 252 err = ENXIO; 253 goto error_2; 254 } 255 if (mp->mnt_flag & MNT_UPDATE) { 256 /* 257 ******************** 258 * UPDATE 259 ******************** 260 */ 261 262 if (devvp != ump->um_devvp) 263 err = EINVAL; /* needs translation */ 264 else 265 vrele(devvp); 266 /* 267 * Update device name only on success 268 */ 269 if( !err) { 270 /* Save "mounted from" info for mount point (NULL pad)*/ 271 copyinstr( args.fspec, 272 mp->mnt_stat.f_mntfromname, 273 MNAMELEN - 1, 274 &size); 275 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 276 } 277 } else { 278 /* 279 ******************** 280 * NEW MOUNT 281 ******************** 282 */ 283 284 /* 285 * Since this is a new mount, we want the names for 286 * the device and the mount point copied in. If an 287 * error occurs, the mountpoint is discarded by the 288 * upper level code. 289 */ 290 /* Save "last mounted on" info for mount point (NULL pad)*/ 291 copyinstr( path, /* mount point*/ 292 mp->mnt_stat.f_mntonname, /* save area*/ 293 MNAMELEN - 1, /* max size*/ 294 &size); /* real size*/ 295 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size); 296 297 /* Save "mounted from" info for mount point (NULL pad)*/ 298 copyinstr( args.fspec, /* device name*/ 299 mp->mnt_stat.f_mntfromname, /* save area*/ 300 MNAMELEN - 1, /* max size*/ 301 &size); /* real size*/ 302 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 303 304 err = ffs_mountfs(devvp, mp, p); 305 } 306 if (err) { 307 goto error_2; 308 } 309 310dostatfs: 311 /* 312 * Initialize FS stat information in mount struct; uses both 313 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname 314 * 315 * This code is common to root and non-root mounts 316 */ 317 (void)VFS_STATFS(mp, &mp->mnt_stat, p); 318 319 goto success; 320 321 322error_2: /* error with devvp held*/ 323 324 /* release devvp before failing*/ 325 vrele(devvp); 326 327error_1: /* no state to back out*/ 328 329success: 330 return( err); 331} 332 333/* 334 * Reload all incore data for a filesystem (used after running fsck on 335 * the root filesystem and finding things to fix). The filesystem must 336 * be mounted read-only. 337 * 338 * Things to do to update the mount: 339 * 1) invalidate all cached meta-data. 340 * 2) re-read superblock from disk. 341 * 3) re-read summary information from disk. 342 * 4) invalidate all inactive vnodes. 343 * 5) invalidate all cached file data. 344 * 6) re-read inode data for all active vnodes. 345 */ 346static int 347ffs_reload(mp, cred, p) 348 register struct mount *mp; 349 struct ucred *cred; 350 struct proc *p; 351{ 352 register struct vnode *vp, *nvp, *devvp; 353 struct inode *ip; 354 struct csum *space; 355 struct buf *bp; 356 struct fs *fs, *newfs; 357 struct partinfo dpart; 358 int i, blks, size, error; 359 int32_t *lp; 360 361 if ((mp->mnt_flag & MNT_RDONLY) == 0) 362 return (EINVAL); 363 /* 364 * Step 1: invalidate all cached meta-data. 365 */ 366 devvp = VFSTOUFS(mp)->um_devvp; 367 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 368 panic("ffs_reload: dirty1"); 369 /* 370 * Step 2: re-read superblock from disk. 371 */ 372 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 373 size = DEV_BSIZE; 374 else 375 size = dpart.disklab->d_secsize; 376 if (error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, NOCRED,&bp)) 377 return (error); 378 newfs = (struct fs *)bp->b_data; 379 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE || 380 newfs->fs_bsize < sizeof(struct fs)) { 381 brelse(bp); 382 return (EIO); /* XXX needs translation */ 383 } 384 fs = VFSTOUFS(mp)->um_fs; 385 /* 386 * Copy pointer fields back into superblock before copying in XXX 387 * new superblock. These should really be in the ufsmount. XXX 388 * Note that important parameters (eg fs_ncg) are unchanged. 389 */ 390 bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp)); 391 newfs->fs_maxcluster = fs->fs_maxcluster; 392 bcopy(newfs, fs, (u_int)fs->fs_sbsize); 393 if (fs->fs_sbsize < SBSIZE) 394 bp->b_flags |= B_INVAL; 395 brelse(bp); 396 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 397 ffs_oldfscompat(fs); 398 399 /* 400 * Step 3: re-read summary information from disk. 401 */ 402 blks = howmany(fs->fs_cssize, fs->fs_fsize); 403 space = fs->fs_csp[0]; 404 for (i = 0; i < blks; i += fs->fs_frag) { 405 size = fs->fs_bsize; 406 if (i + fs->fs_frag > blks) 407 size = (blks - i) * fs->fs_fsize; 408 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 409 NOCRED, &bp); 410 if (error) 411 return (error); 412 bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size); 413 brelse(bp); 414 } 415 /* 416 * We no longer know anything about clusters per cylinder group. 417 */ 418 if (fs->fs_contigsumsize > 0) { 419 lp = fs->fs_maxcluster; 420 for (i = 0; i < fs->fs_ncg; i++) 421 *lp++ = fs->fs_contigsumsize; 422 } 423 424loop: 425 simple_lock(&mntvnode_slock); 426 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 427 if (vp->v_mount != mp) { 428 simple_unlock(&mntvnode_slock); 429 goto loop; 430 } 431 nvp = vp->v_mntvnodes.le_next; 432 /* 433 * Step 4: invalidate all inactive vnodes. 434 */ 435 if (vrecycle(vp, &mntvnode_slock, p)) 436 goto loop; 437 /* 438 * Step 5: invalidate all cached file data. 439 */ 440 simple_lock(&vp->v_interlock); 441 simple_unlock(&mntvnode_slock); 442 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 443 goto loop; 444 } 445 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 446 panic("ffs_reload: dirty2"); 447 /* 448 * Step 6: re-read inode data for all active vnodes. 449 */ 450 ip = VTOI(vp); 451 error = 452 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 453 (int)fs->fs_bsize, NOCRED, &bp); 454 if (error) { 455 vput(vp); 456 return (error); 457 } 458 ip->i_din = *((struct dinode *)bp->b_data + 459 ino_to_fsbo(fs, ip->i_number)); 460 brelse(bp); 461 vput(vp); 462 simple_lock(&mntvnode_slock); 463 } 464 simple_unlock(&mntvnode_slock); 465 return (0); 466} 467 468/* 469 * Common code for mount and mountroot 470 */ 471int 472ffs_mountfs(devvp, mp, p) 473 register struct vnode *devvp; 474 struct mount *mp; 475 struct proc *p; 476{ 477 register struct ufsmount *ump; 478 struct buf *bp; 479 register struct fs *fs; 480 dev_t dev; 481 struct partinfo dpart; 482 caddr_t base, space; 483 int error, i, blks, size, ronly; 484 int32_t *lp; 485 struct ucred *cred; 486 u_int64_t maxfilesize; /* XXX */ 487 u_int strsize; 488 int ncount; 489 490 dev = devvp->v_rdev; 491 cred = p ? p->p_ucred : NOCRED; 492 /* 493 * Disallow multiple mounts of the same device. 494 * Disallow mounting of a device that is currently in use 495 * (except for root, which might share swap device for miniroot). 496 * Flush out any old buffers remaining from a previous use. 497 */ 498 error = vfs_mountedon(devvp); 499 if (error) 500 return (error); 501 ncount = vcount(devvp); 502 if (devvp->v_object) 503 ncount -= 1; 504 if (ncount > 1 && devvp != rootvp) 505 return (EBUSY); 506 if (error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) 507 return (error); 508 509 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 510 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p); 511 if (error) 512 return (error); 513 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0) 514 size = DEV_BSIZE; 515 else 516 size = dpart.disklab->d_secsize; 517 518 bp = NULL; 519 ump = NULL; 520 if (error = bread(devvp, SBLOCK, SBSIZE, cred, &bp)) 521 goto out; 522 fs = (struct fs *)bp->b_data; 523 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE || 524 fs->fs_bsize < sizeof(struct fs)) { 525 error = EINVAL; /* XXX needs translation */ 526 goto out; 527 } 528 fs->fs_fmod = 0; 529 if (!fs->fs_clean) { 530 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 531 printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt); 532 } else { 533 printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt); 534 error = EPERM; 535 goto out; 536 } 537 } 538 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */ 539 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) { 540 error = EROFS; /* needs translation */ 541 goto out; 542 } 543 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 544 bzero((caddr_t)ump, sizeof *ump); 545 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT, 546 M_WAITOK); 547 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize); 548 if (fs->fs_sbsize < SBSIZE) 549 bp->b_flags |= B_INVAL; 550 brelse(bp); 551 bp = NULL; 552 fs = ump->um_fs; 553 fs->fs_ronly = ronly; 554 if (ronly == 0) { 555 fs->fs_fmod = 1; 556 fs->fs_clean = 0; 557 } 558 size = fs->fs_cssize; 559 blks = howmany(size, fs->fs_fsize); 560 if (fs->fs_contigsumsize > 0) 561 size += fs->fs_ncg * sizeof(int32_t); 562 base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK); 563 for (i = 0; i < blks; i += fs->fs_frag) { 564 size = fs->fs_bsize; 565 if (i + fs->fs_frag > blks) 566 size = (blks - i) * fs->fs_fsize; 567 if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 568 cred, &bp)) { 569 free(base, M_UFSMNT); 570 goto out; 571 } 572 bcopy(bp->b_data, space, (u_int)size); 573 fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space; 574 space += size; 575 brelse(bp); 576 bp = NULL; 577 } 578 if (fs->fs_contigsumsize > 0) { 579 fs->fs_maxcluster = lp = (int32_t *)space; 580 for (i = 0; i < fs->fs_ncg; i++) 581 *lp++ = fs->fs_contigsumsize; 582 } 583 mp->mnt_data = (qaddr_t)ump; 584 mp->mnt_stat.f_fsid.val[0] = (long)dev; 585 if (fs->fs_id[0] != 0 && fs->fs_id[1] != 0) 586 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1]; 587 else 588 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 589 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 590 mp->mnt_flag |= MNT_LOCAL; 591 ump->um_mountp = mp; 592 ump->um_dev = dev; 593 ump->um_devvp = devvp; 594 ump->um_nindir = fs->fs_nindir; 595 ump->um_bptrtodb = fs->fs_fsbtodb; 596 ump->um_seqinc = fs->fs_frag; 597 for (i = 0; i < MAXQUOTAS; i++) 598 ump->um_quotas[i] = NULLVP; 599 devvp->v_specflags |= SI_MOUNTEDON; 600 ffs_oldfscompat(fs); 601 602 /* 603 * Set FS local "last mounted on" information (NULL pad) 604 */ 605 copystr( mp->mnt_stat.f_mntonname, /* mount point*/ 606 fs->fs_fsmnt, /* copy area*/ 607 sizeof(fs->fs_fsmnt) - 1, /* max size*/ 608 &strsize); /* real size*/ 609 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize); 610 611 if( mp->mnt_flag & MNT_ROOTFS) { 612 /* 613 * Root mount; update timestamp in mount structure. 614 * this will be used by the common root mount code 615 * to update the system clock. 616 */ 617 mp->mnt_time = fs->fs_time; 618 } 619 620 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */ 621 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */ 622 if (fs->fs_maxfilesize > maxfilesize) /* XXX */ 623 fs->fs_maxfilesize = maxfilesize; /* XXX */ 624 if (ronly == 0) { 625 fs->fs_clean = 0; 626 (void) ffs_sbupdate(ump, MNT_WAIT); 627 } 628 /* 629 * Only VMIO the backing device if the backing device is a real 630 * block device. This excludes the original MFS implementation. 631 * Note that it is optional that the backing device be VMIOed. This 632 * increases the opportunity for metadata caching. 633 */ 634 if ((devvp->v_type == VBLK) && (major(devvp->v_rdev) < nblkdev)) { 635 vfs_object_create(devvp, p, p->p_ucred, 0); 636 } 637 return (0); 638out: 639 if (bp) 640 brelse(bp); 641 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p); 642 if (ump) { 643 free(ump->um_fs, M_UFSMNT); 644 free(ump, M_UFSMNT); 645 mp->mnt_data = (qaddr_t)0; 646 } 647 return (error); 648} 649 650/* 651 * Sanity checks for old file systems. 652 * 653 * XXX - goes away some day. 654 */ 655static int 656ffs_oldfscompat(fs) 657 struct fs *fs; 658{ 659 660 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */ 661 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */ 662 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */ 663 fs->fs_nrpos = 8; /* XXX */ 664 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 665#if 0 666 int i; /* XXX */ 667 u_int64_t sizepb = fs->fs_bsize; /* XXX */ 668 /* XXX */ 669 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */ 670 for (i = 0; i < NIADDR; i++) { /* XXX */ 671 sizepb *= NINDIR(fs); /* XXX */ 672 fs->fs_maxfilesize += sizepb; /* XXX */ 673 } /* XXX */ 674#endif 675 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 676 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */ 677 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */ 678 } /* XXX */ 679 return (0); 680} 681 682/* 683 * unmount system call 684 */ 685int 686ffs_unmount(mp, mntflags, p) 687 struct mount *mp; 688 int mntflags; 689 struct proc *p; 690{ 691 register struct ufsmount *ump; 692 register struct fs *fs; 693 int error, flags; 694 695 flags = 0; 696 if (mntflags & MNT_FORCE) { 697 flags |= FORCECLOSE; 698 } 699 error = ffs_flushfiles(mp, flags, p); 700 if (error) 701 return (error); 702 ump = VFSTOUFS(mp); 703 fs = ump->um_fs; 704 if (fs->fs_ronly == 0) { 705 fs->fs_clean = 1; 706 error = ffs_sbupdate(ump, MNT_WAIT); 707 if (error) { 708 fs->fs_clean = 0; 709 return (error); 710 } 711 } 712 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON; 713 714 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p); 715 vnode_pager_uncache(ump->um_devvp, p); 716 VOP_UNLOCK(ump->um_devvp, 0, p); 717 718 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, 719 NOCRED, p); 720 721 vrele(ump->um_devvp); 722 723 free(fs->fs_csp[0], M_UFSMNT); 724 free(fs, M_UFSMNT); 725 free(ump, M_UFSMNT); 726 mp->mnt_data = (qaddr_t)0; 727 mp->mnt_flag &= ~MNT_LOCAL; 728 return (error); 729} 730 731/* 732 * Flush out all the files in a filesystem. 733 */ 734int 735ffs_flushfiles(mp, flags, p) 736 register struct mount *mp; 737 int flags; 738 struct proc *p; 739{ 740 register struct ufsmount *ump; 741 int error; 742 743 ump = VFSTOUFS(mp); 744#ifdef QUOTA 745 if (mp->mnt_flag & MNT_QUOTA) { 746 int i; 747 error = vflush(mp, NULLVP, SKIPSYSTEM|flags); 748 if (error) 749 return (error); 750 for (i = 0; i < MAXQUOTAS; i++) { 751 if (ump->um_quotas[i] == NULLVP) 752 continue; 753 quotaoff(p, mp, i); 754 } 755 /* 756 * Here we fall through to vflush again to ensure 757 * that we have gotten rid of all the system vnodes. 758 */ 759 } 760#endif 761 error = vflush(mp, NULLVP, flags); 762 return (error); 763} 764 765/* 766 * Get file system statistics. 767 */ 768int 769ffs_statfs(mp, sbp, p) 770 struct mount *mp; 771 register struct statfs *sbp; 772 struct proc *p; 773{ 774 register struct ufsmount *ump; 775 register struct fs *fs; 776 777 ump = VFSTOUFS(mp); 778 fs = ump->um_fs; 779 if (fs->fs_magic != FS_MAGIC) 780 panic("ffs_statfs"); 781 sbp->f_bsize = fs->fs_fsize; 782 sbp->f_iosize = fs->fs_bsize; 783 sbp->f_blocks = fs->fs_dsize; 784 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag + 785 fs->fs_cstotal.cs_nffree; 786 sbp->f_bavail = freespace(fs, fs->fs_minfree); 787 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO; 788 sbp->f_ffree = fs->fs_cstotal.cs_nifree; 789 if (sbp != &mp->mnt_stat) { 790 sbp->f_type = mp->mnt_vfc->vfc_typenum; 791 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 792 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 793 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 794 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 795 } 796 return (0); 797} 798 799/* 800 * Go through the disk queues to initiate sandbagged IO; 801 * go through the inodes to write those that have been modified; 802 * initiate the writing of the super block if it has been modified. 803 * 804 * Note: we are always called with the filesystem marked `MPBUSY'. 805 */ 806int 807ffs_sync(mp, waitfor, cred, p) 808 struct mount *mp; 809 int waitfor; 810 struct ucred *cred; 811 struct proc *p; 812{ 813 struct vnode *nvp, *vp; 814 struct inode *ip; 815 struct ufsmount *ump = VFSTOUFS(mp); 816 struct fs *fs; 817 struct timeval tv; 818 int error, allerror = 0; 819 820 fs = ump->um_fs; 821 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 822 printf("fs = %s\n", fs->fs_fsmnt); 823 panic("ffs_sync: rofs mod"); 824 } 825 /* 826 * Write back each (modified) inode. 827 */ 828 simple_lock(&mntvnode_slock); 829loop: 830 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 831 /* 832 * If the vnode that we are about to sync is no longer 833 * associated with this mount point, start over. 834 */ 835 if (vp->v_mount != mp) 836 goto loop; 837 simple_lock(&vp->v_interlock); 838 nvp = vp->v_mntvnodes.le_next; 839 ip = VTOI(vp); 840 if (((ip->i_flag & 841 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0) && 842 vp->v_dirtyblkhd.lh_first == NULL) { 843 simple_unlock(&vp->v_interlock); 844 continue; 845 } 846 if (vp->v_type != VCHR) { 847 simple_unlock(&mntvnode_slock); 848 error = 849 vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p); 850 if (error) { 851 simple_lock(&mntvnode_slock); 852 if (error == ENOENT) 853 goto loop; 854 continue; 855 } 856 if (error = VOP_FSYNC(vp, cred, waitfor, p)) 857 allerror = error; 858 VOP_UNLOCK(vp, 0, p); 859 vrele(vp); 860 simple_lock(&mntvnode_slock); 861 } else { 862 simple_unlock(&mntvnode_slock); 863 simple_unlock(&vp->v_interlock); 864 gettime(&tv); 865 /* VOP_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT); */ 866 VOP_UPDATE(vp, &tv, &tv, 0); 867 simple_lock(&mntvnode_slock); 868 } 869 } 870 simple_unlock(&mntvnode_slock); 871 /* 872 * Force stale file system control information to be flushed. 873 */ 874 error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p); 875 if (error) 876 allerror = error; 877#ifdef QUOTA 878 qsync(mp); 879#endif 880 /* 881 * Write back modified superblock. 882 */ 883 if (fs->fs_fmod != 0) { 884 fs->fs_fmod = 0; 885 fs->fs_time = time.tv_sec; 886 if (error = ffs_sbupdate(ump, waitfor)) 887 allerror = error; 888 } 889 return (allerror); 890} 891 892/* 893 * Look up a FFS dinode number to find its incore vnode, otherwise read it 894 * in from disk. If it is in core, wait for the lock bit to clear, then 895 * return the inode locked. Detection and handling of mount points must be 896 * done by the calling routine. 897 */ 898static int ffs_inode_hash_lock; 899 900int 901ffs_vget(mp, ino, vpp) 902 struct mount *mp; 903 ino_t ino; 904 struct vnode **vpp; 905{ 906 struct fs *fs; 907 struct inode *ip; 908 struct ufsmount *ump; 909 struct buf *bp; 910 struct vnode *vp; 911 dev_t dev; 912 int type, error; 913 914 ump = VFSTOUFS(mp); 915 dev = ump->um_dev; 916restart: 917 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 918 return (0); 919 920 /* 921 * Lock out the creation of new entries in the FFS hash table in 922 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 923 * may occur! 924 */ 925 if (ffs_inode_hash_lock) { 926 while (ffs_inode_hash_lock) { 927 ffs_inode_hash_lock = -1; 928 tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0); 929 } 930 goto restart; 931 } 932 ffs_inode_hash_lock = 1; 933 934 /* 935 * If this MALLOC() is performed after the getnewvnode() 936 * it might block, leaving a vnode with a NULL v_data to be 937 * found by ffs_sync() if a sync happens to fire right then, 938 * which will cause a panic because ffs_sync() blindly 939 * dereferences vp->v_data (as well it should). 940 */ 941 type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */ 942 MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK); 943 944 /* Allocate a new vnode/inode. */ 945 error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp); 946 if (error) { 947 if (ffs_inode_hash_lock < 0) 948 wakeup(&ffs_inode_hash_lock); 949 ffs_inode_hash_lock = 0; 950 *vpp = NULL; 951 FREE(ip, type); 952 return (error); 953 } 954 bzero((caddr_t)ip, sizeof(struct inode)); 955 lockinit(&ip->i_lock, PINOD, "inode", 0, 0); 956 vp->v_data = ip; 957 ip->i_vnode = vp; 958 ip->i_fs = fs = ump->um_fs; 959 ip->i_dev = dev; 960 ip->i_number = ino; 961#ifdef QUOTA 962 { 963 int i; 964 for (i = 0; i < MAXQUOTAS; i++) 965 ip->i_dquot[i] = NODQUOT; 966 } 967#endif 968 /* 969 * Put it onto its hash chain and lock it so that other requests for 970 * this inode will block if they arrive while we are sleeping waiting 971 * for old data structures to be purged or for the contents of the 972 * disk portion of this inode to be read. 973 */ 974 ufs_ihashins(ip); 975 976 if (ffs_inode_hash_lock < 0) 977 wakeup(&ffs_inode_hash_lock); 978 ffs_inode_hash_lock = 0; 979 980 /* Read in the disk contents for the inode, copy into the inode. */ 981 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 982 (int)fs->fs_bsize, NOCRED, &bp); 983 if (error) { 984 /* 985 * The inode does not contain anything useful, so it would 986 * be misleading to leave it on its hash chain. With mode 987 * still zero, it will be unlinked and returned to the free 988 * list by vput(). 989 */ 990 brelse(bp); 991 vput(vp); 992 *vpp = NULL; 993 return (error); 994 } 995 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino)); 996 bqrelse(bp); 997 998 /* 999 * Initialize the vnode from the inode, check for aliases. 1000 * Note that the underlying vnode may have changed. 1001 */ 1002 error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1003 if (error) { 1004 vput(vp); 1005 *vpp = NULL; 1006 return (error); 1007 } 1008 /* 1009 * Finish inode initialization now that aliasing has been resolved. 1010 */ 1011 ip->i_devvp = ump->um_devvp; 1012 VREF(ip->i_devvp); 1013 /* 1014 * Set up a generation number for this inode if it does not 1015 * already have one. This should only happen on old filesystems. 1016 */ 1017 if (ip->i_gen == 0) { 1018 ip->i_gen = random() / 2 + 1; 1019 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1020 ip->i_flag |= IN_MODIFIED; 1021 } 1022 /* 1023 * Ensure that uid and gid are correct. This is a temporary 1024 * fix until fsck has been changed to do the update. 1025 */ 1026 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 1027 ip->i_uid = ip->i_din.di_ouid; /* XXX */ 1028 ip->i_gid = ip->i_din.di_ogid; /* XXX */ 1029 } /* XXX */ 1030 1031 *vpp = vp; 1032 return (0); 1033} 1034 1035/* 1036 * File handle to vnode 1037 * 1038 * Have to be really careful about stale file handles: 1039 * - check that the inode number is valid 1040 * - call ffs_vget() to get the locked inode 1041 * - check for an unallocated inode (i_mode == 0) 1042 * - check that the given client host has export rights and return 1043 * those rights via. exflagsp and credanonp 1044 */ 1045int 1046ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) 1047 register struct mount *mp; 1048 struct fid *fhp; 1049 struct mbuf *nam; 1050 struct vnode **vpp; 1051 int *exflagsp; 1052 struct ucred **credanonp; 1053{ 1054 register struct ufid *ufhp; 1055 struct fs *fs; 1056 1057 ufhp = (struct ufid *)fhp; 1058 fs = VFSTOUFS(mp)->um_fs; 1059 if (ufhp->ufid_ino < ROOTINO || 1060 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) 1061 return (ESTALE); 1062 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)); 1063} 1064 1065/* 1066 * Vnode pointer to File handle 1067 */ 1068/* ARGSUSED */ 1069int 1070ffs_vptofh(vp, fhp) 1071 struct vnode *vp; 1072 struct fid *fhp; 1073{ 1074 register struct inode *ip; 1075 register struct ufid *ufhp; 1076 1077 ip = VTOI(vp); 1078 ufhp = (struct ufid *)fhp; 1079 ufhp->ufid_len = sizeof(struct ufid); 1080 ufhp->ufid_ino = ip->i_number; 1081 ufhp->ufid_gen = ip->i_gen; 1082 return (0); 1083} 1084 1085/* 1086 * Initialize the filesystem; just use ufs_init. 1087 */ 1088static int 1089ffs_init(vfsp) 1090 struct vfsconf *vfsp; 1091{ 1092 1093 return (ufs_init(vfsp)); 1094} 1095 1096/* 1097 * Write a superblock and associated information back to disk. 1098 */ 1099static int 1100ffs_sbupdate(mp, waitfor) 1101 struct ufsmount *mp; 1102 int waitfor; 1103{ 1104 register struct fs *dfs, *fs = mp->um_fs; 1105 register struct buf *bp; 1106 int blks; 1107 caddr_t space; 1108 int i, size, error, allerror = 0; 1109 1110 /* 1111 * First write back the summary information. 1112 */ 1113 blks = howmany(fs->fs_cssize, fs->fs_fsize); 1114 space = (caddr_t)fs->fs_csp[0]; 1115 for (i = 0; i < blks; i += fs->fs_frag) { 1116 size = fs->fs_bsize; 1117 if (i + fs->fs_frag > blks) 1118 size = (blks - i) * fs->fs_fsize; 1119 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), 1120 size, 0, 0); 1121 bcopy(space, bp->b_data, (u_int)size); 1122 space += size; 1123 if (waitfor != MNT_WAIT) 1124 bawrite(bp); 1125 else if (error = bwrite(bp)) 1126 allerror = error; 1127 } 1128 /* 1129 * Now write back the superblock itself. If any errors occurred 1130 * up to this point, then fail so that the superblock avoids 1131 * being written out as clean. 1132 */ 1133 if (allerror) 1134 return (allerror); 1135 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0); 1136 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); 1137 /* Restore compatibility to old file systems. XXX */ 1138 dfs = (struct fs *)bp->b_data; /* XXX */ 1139 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */ 1140 dfs->fs_nrpos = -1; /* XXX */ 1141 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 1142 int32_t *lp, tmp; /* XXX */ 1143 /* XXX */ 1144 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */ 1145 tmp = lp[4]; /* XXX */ 1146 for (i = 4; i > 0; i--) /* XXX */ 1147 lp[i] = lp[i-1]; /* XXX */ 1148 lp[0] = tmp; /* XXX */ 1149 } /* XXX */ 1150 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */ 1151 if (waitfor != MNT_WAIT) 1152 bawrite(bp); 1153 else if (error = bwrite(bp)) 1154 allerror = error; 1155 return (allerror); 1156} 1157