ffs_vfsops.c revision 32585
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 1.64 1998/01/06 05:23:41 dyson 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 77struct 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 ffs_fhtovp, 87 ffs_vptofh, 88 ffs_init, 89}; 90 91VFS_SET(ufs_vfsops, ufs, MOUNT_UFS, 0); 92 93/* 94 * ffs_mount 95 * 96 * Called when mounting local physical media 97 * 98 * PARAMETERS: 99 * mountroot 100 * mp mount point structure 101 * path NULL (flag for root mount!!!) 102 * data <unused> 103 * ndp <unused> 104 * p process (user credentials check [statfs]) 105 * 106 * mount 107 * mp mount point structure 108 * path path to mount point 109 * data pointer to argument struct in user space 110 * ndp mount point namei() return (used for 111 * credentials on reload), reused to look 112 * up block device. 113 * p process (user credentials check) 114 * 115 * RETURNS: 0 Success 116 * !0 error number (errno.h) 117 * 118 * LOCK STATE: 119 * 120 * ENTRY 121 * mount point is locked 122 * EXIT 123 * mount point is locked 124 * 125 * NOTES: 126 * A NULL path can be used for a flag since the mount 127 * system call will fail with EFAULT in copyinstr in 128 * namei() if it is a genuine NULL from the user. 129 */ 130static int 131ffs_mount( mp, path, data, ndp, p) 132 struct mount *mp; /* mount struct pointer*/ 133 char *path; /* path to mount point*/ 134 caddr_t data; /* arguments to FS specific mount*/ 135 struct nameidata *ndp; /* mount point credentials*/ 136 struct proc *p; /* process requesting mount*/ 137{ 138 u_int size; 139 int err = 0; 140 struct vnode *devvp; 141 142 struct ufs_args args; 143 struct ufsmount *ump = 0; 144 register struct fs *fs; 145 int flags; 146 147 /* 148 * Use NULL path to flag a root mount 149 */ 150 if( path == NULL) { 151 /* 152 *** 153 * Mounting root file system 154 *** 155 */ 156 157 if ((err = bdevvp(rootdev, &rootvp))) { 158 printf("ffs_mountroot: can't find rootvp"); 159 return (err); 160 } 161 162 if (bdevsw[major(rootdev)]->d_flags & D_NOCLUSTERR) 163 mp->mnt_flag |= MNT_NOCLUSTERR; 164 if (bdevsw[major(rootdev)]->d_flags & D_NOCLUSTERW) 165 mp->mnt_flag |= MNT_NOCLUSTERW; 166 if( ( err = ffs_mountfs(rootvp, mp, p, M_FFSNODE)) != 0) { 167 /* fs specific cleanup (if any)*/ 168 goto error_1; 169 } 170 171 goto dostatfs; /* success*/ 172 173 } 174 175 /* 176 *** 177 * Mounting non-root file system or updating a file system 178 *** 179 */ 180 181 /* copy in user arguments*/ 182 err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)); 183 if (err) 184 goto error_1; /* can't get arguments*/ 185 186 /* 187 * If updating, check whether changing from read-only to 188 * read/write; if there is no device name, that's all we do. 189 * Disallow clearing MNT_NOCLUSTERR and MNT_NOCLUSTERW flags, 190 * if block device requests. 191 */ 192 if (mp->mnt_flag & MNT_UPDATE) { 193 ump = VFSTOUFS(mp); 194 fs = ump->um_fs; 195 err = 0; 196 if (bdevsw[major(ump->um_dev)]->d_flags & D_NOCLUSTERR) 197 mp->mnt_flag |= MNT_NOCLUSTERR; 198 if (bdevsw[major(ump->um_dev)]->d_flags & D_NOCLUSTERW) 199 mp->mnt_flag |= MNT_NOCLUSTERW; 200 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 201 flags = WRITECLOSE; 202 if (mp->mnt_flag & MNT_FORCE) 203 flags |= FORCECLOSE; 204 err = ffs_flushfiles(mp, flags, p); 205 } 206 if (!err && (mp->mnt_flag & MNT_RELOAD)) 207 err = ffs_reload(mp, ndp->ni_cnd.cn_cred, p); 208 if (err) { 209 goto error_1; 210 } 211 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 212 if (!fs->fs_clean) { 213 if (mp->mnt_flag & MNT_FORCE) { 214 printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt); 215 } else { 216 printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n", 217 fs->fs_fsmnt); 218 err = EPERM; 219 goto error_1; 220 } 221 } 222 fs->fs_ronly = 0; 223 } 224 if (fs->fs_ronly == 0) { 225 fs->fs_clean = 0; 226 ffs_sbupdate(ump, MNT_WAIT); 227 } 228 /* if not updating name...*/ 229 if (args.fspec == 0) { 230 /* 231 * Process export requests. Jumping to "success" 232 * will return the vfs_export() error code. 233 */ 234 err = vfs_export(mp, &ump->um_export, &args.export); 235 goto success; 236 } 237 } 238 239 /* 240 * Not an update, or updating the name: look up the name 241 * and verify that it refers to a sensible block device. 242 */ 243 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 244 err = namei(ndp); 245 if (err) { 246 /* can't get devvp!*/ 247 goto error_1; 248 } 249 250 devvp = ndp->ni_vp; 251 252 if (devvp->v_type != VBLK) { 253 err = ENOTBLK; 254 goto error_2; 255 } 256 if (major(devvp->v_rdev) >= nblkdev) { 257 err = ENXIO; 258 goto error_2; 259 } 260 if (mp->mnt_flag & MNT_UPDATE) { 261 /* 262 ******************** 263 * UPDATE 264 ******************** 265 */ 266 267 if (devvp != ump->um_devvp) 268 err = EINVAL; /* needs translation */ 269 else 270 vrele(devvp); 271 /* 272 * Update device name only on success 273 */ 274 if( !err) { 275 /* Save "mounted from" info for mount point (NULL pad)*/ 276 copyinstr( args.fspec, 277 mp->mnt_stat.f_mntfromname, 278 MNAMELEN - 1, 279 &size); 280 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 281 } 282 } else { 283 /* 284 ******************** 285 * NEW MOUNT 286 ******************** 287 */ 288 289 if (bdevsw[major(devvp->v_rdev)]->d_flags & D_NOCLUSTERR) 290 mp->mnt_flag |= MNT_NOCLUSTERR; 291 if (bdevsw[major(devvp->v_rdev)]->d_flags & D_NOCLUSTERW) 292 mp->mnt_flag |= MNT_NOCLUSTERW; 293 294 /* 295 * Since this is a new mount, we want the names for 296 * the device and the mount point copied in. If an 297 * error occurs, the mountpoint is discarded by the 298 * upper level code. 299 */ 300 /* Save "last mounted on" info for mount point (NULL pad)*/ 301 copyinstr( path, /* mount point*/ 302 mp->mnt_stat.f_mntonname, /* save area*/ 303 MNAMELEN - 1, /* max size*/ 304 &size); /* real size*/ 305 bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size); 306 307 /* Save "mounted from" info for mount point (NULL pad)*/ 308 copyinstr( args.fspec, /* device name*/ 309 mp->mnt_stat.f_mntfromname, /* save area*/ 310 MNAMELEN - 1, /* max size*/ 311 &size); /* real size*/ 312 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 313 314 err = ffs_mountfs(devvp, mp, p, M_FFSNODE); 315 } 316 if (err) { 317 goto error_2; 318 } 319 320dostatfs: 321 /* 322 * Initialize FS stat information in mount struct; uses both 323 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname 324 * 325 * This code is common to root and non-root mounts 326 */ 327 (void)VFS_STATFS(mp, &mp->mnt_stat, p); 328 329 goto success; 330 331 332error_2: /* error with devvp held*/ 333 334 /* release devvp before failing*/ 335 vrele(devvp); 336 337error_1: /* no state to back out*/ 338 339success: 340 return( err); 341} 342 343/* 344 * Reload all incore data for a filesystem (used after running fsck on 345 * the root filesystem and finding things to fix). The filesystem must 346 * be mounted read-only. 347 * 348 * Things to do to update the mount: 349 * 1) invalidate all cached meta-data. 350 * 2) re-read superblock from disk. 351 * 3) re-read summary information from disk. 352 * 4) invalidate all inactive vnodes. 353 * 5) invalidate all cached file data. 354 * 6) re-read inode data for all active vnodes. 355 */ 356static int 357ffs_reload(mp, cred, p) 358 register struct mount *mp; 359 struct ucred *cred; 360 struct proc *p; 361{ 362 register struct vnode *vp, *nvp, *devvp; 363 struct inode *ip; 364 struct csum *space; 365 struct buf *bp; 366 struct fs *fs, *newfs; 367 struct partinfo dpart; 368 dev_t dev; 369 int i, blks, size, error; 370 int32_t *lp; 371 372 if ((mp->mnt_flag & MNT_RDONLY) == 0) 373 return (EINVAL); 374 /* 375 * Step 1: invalidate all cached meta-data. 376 */ 377 devvp = VFSTOUFS(mp)->um_devvp; 378 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 379 panic("ffs_reload: dirty1"); 380 381 dev = devvp->v_rdev; 382 /* 383 * Only VMIO the backing device if the backing device is a real 384 * block device. This excludes the original MFS implementation. 385 * Note that it is optional that the backing device be VMIOed. This 386 * increases the opportunity for metadata caching. 387 */ 388 if ((devvp->v_type == VBLK) && (major(dev) < nblkdev)) { 389 vfs_object_create(devvp, p, p->p_ucred, 0); 390 } 391 392 /* 393 * Step 2: re-read superblock from disk. 394 */ 395 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 396 size = DEV_BSIZE; 397 else 398 size = dpart.disklab->d_secsize; 399 if (error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, NOCRED,&bp)) 400 return (error); 401 newfs = (struct fs *)bp->b_data; 402 if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE || 403 newfs->fs_bsize < sizeof(struct fs)) { 404 brelse(bp); 405 return (EIO); /* XXX needs translation */ 406 } 407 fs = VFSTOUFS(mp)->um_fs; 408 /* 409 * Copy pointer fields back into superblock before copying in XXX 410 * new superblock. These should really be in the ufsmount. XXX 411 * Note that important parameters (eg fs_ncg) are unchanged. 412 */ 413 bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp)); 414 newfs->fs_maxcluster = fs->fs_maxcluster; 415 bcopy(newfs, fs, (u_int)fs->fs_sbsize); 416 if (fs->fs_sbsize < SBSIZE) 417 bp->b_flags |= B_INVAL; 418 brelse(bp); 419 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 420 ffs_oldfscompat(fs); 421 422 /* 423 * Step 3: re-read summary information from disk. 424 */ 425 blks = howmany(fs->fs_cssize, fs->fs_fsize); 426 space = fs->fs_csp[0]; 427 for (i = 0; i < blks; i += fs->fs_frag) { 428 size = fs->fs_bsize; 429 if (i + fs->fs_frag > blks) 430 size = (blks - i) * fs->fs_fsize; 431 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 432 NOCRED, &bp); 433 if (error) 434 return (error); 435 bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size); 436 brelse(bp); 437 } 438 /* 439 * We no longer know anything about clusters per cylinder group. 440 */ 441 if (fs->fs_contigsumsize > 0) { 442 lp = fs->fs_maxcluster; 443 for (i = 0; i < fs->fs_ncg; i++) 444 *lp++ = fs->fs_contigsumsize; 445 } 446 447loop: 448 simple_lock(&mntvnode_slock); 449 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 450 if (vp->v_mount != mp) { 451 simple_unlock(&mntvnode_slock); 452 goto loop; 453 } 454 nvp = vp->v_mntvnodes.le_next; 455 /* 456 * Step 4: invalidate all inactive vnodes. 457 */ 458 if (vrecycle(vp, &mntvnode_slock, p)) 459 goto loop; 460 /* 461 * Step 5: invalidate all cached file data. 462 */ 463 simple_lock(&vp->v_interlock); 464 simple_unlock(&mntvnode_slock); 465 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 466 goto loop; 467 } 468 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 469 panic("ffs_reload: dirty2"); 470 /* 471 * Step 6: re-read inode data for all active vnodes. 472 */ 473 ip = VTOI(vp); 474 error = 475 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 476 (int)fs->fs_bsize, NOCRED, &bp); 477 if (error) { 478 vput(vp); 479 return (error); 480 } 481 ip->i_din = *((struct dinode *)bp->b_data + 482 ino_to_fsbo(fs, ip->i_number)); 483 brelse(bp); 484 vput(vp); 485 simple_lock(&mntvnode_slock); 486 } 487 simple_unlock(&mntvnode_slock); 488 return (0); 489} 490 491/* 492 * Common code for mount and mountroot 493 */ 494int 495ffs_mountfs(devvp, mp, p, malloctype) 496 register struct vnode *devvp; 497 struct mount *mp; 498 struct proc *p; 499 struct malloc_type *malloctype; 500{ 501 register struct ufsmount *ump; 502 struct buf *bp; 503 register struct fs *fs; 504 dev_t dev; 505 struct partinfo dpart; 506 caddr_t base, space; 507 int error, i, blks, size, ronly; 508 int32_t *lp; 509 struct ucred *cred; 510 u_int64_t maxfilesize; /* XXX */ 511 u_int strsize; 512 int ncount; 513 514 dev = devvp->v_rdev; 515 cred = p ? p->p_ucred : NOCRED; 516 /* 517 * Disallow multiple mounts of the same device. 518 * Disallow mounting of a device that is currently in use 519 * (except for root, which might share swap device for miniroot). 520 * Flush out any old buffers remaining from a previous use. 521 */ 522 error = vfs_mountedon(devvp); 523 if (error) 524 return (error); 525 ncount = vcount(devvp); 526/* 527 if (devvp->v_object) 528 ncount -= 1; 529*/ 530 if (ncount > 1 && devvp != rootvp) 531 return (EBUSY); 532 if (error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) 533 return (error); 534 535 /* 536 * Only VMIO the backing device if the backing device is a real 537 * block device. This excludes the original MFS implementation. 538 * Note that it is optional that the backing device be VMIOed. This 539 * increases the opportunity for metadata caching. 540 */ 541 if ((devvp->v_type == VBLK) && (major(dev) < nblkdev)) { 542 vfs_object_create(devvp, p, p->p_ucred, 0); 543 } 544 545 546 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 547 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p); 548 if (error) 549 return (error); 550 551 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0) 552 size = DEV_BSIZE; 553 else 554 size = dpart.disklab->d_secsize; 555 556 bp = NULL; 557 ump = NULL; 558 if (error = bread(devvp, SBLOCK, SBSIZE, cred, &bp)) 559 goto out; 560 fs = (struct fs *)bp->b_data; 561 if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE || 562 fs->fs_bsize < sizeof(struct fs)) { 563 error = EINVAL; /* XXX needs translation */ 564 goto out; 565 } 566 fs->fs_fmod = 0; 567 if (!fs->fs_clean) { 568 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 569 printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt); 570 } else { 571 printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt); 572 error = EPERM; 573 goto out; 574 } 575 } 576 /* XXX updating 4.2 FFS superblocks trashes rotational layout tables */ 577 if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) { 578 error = EROFS; /* needs translation */ 579 goto out; 580 } 581 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 582 bzero((caddr_t)ump, sizeof *ump); 583 ump->um_malloctype = malloctype; 584 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT, 585 M_WAITOK); 586 ump->um_blkatoff = ffs_blkatoff; 587 ump->um_truncate = ffs_truncate; 588 ump->um_update = ffs_update; 589 ump->um_valloc = ffs_valloc; 590 ump->um_vfree = ffs_vfree; 591 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize); 592 if (fs->fs_sbsize < SBSIZE) 593 bp->b_flags |= B_INVAL; 594 brelse(bp); 595 bp = NULL; 596 fs = ump->um_fs; 597 fs->fs_ronly = ronly; 598 if (ronly == 0) { 599 fs->fs_fmod = 1; 600 fs->fs_clean = 0; 601 } 602 size = fs->fs_cssize; 603 blks = howmany(size, fs->fs_fsize); 604 if (fs->fs_contigsumsize > 0) 605 size += fs->fs_ncg * sizeof(int32_t); 606 base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK); 607 for (i = 0; i < blks; i += fs->fs_frag) { 608 size = fs->fs_bsize; 609 if (i + fs->fs_frag > blks) 610 size = (blks - i) * fs->fs_fsize; 611 if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 612 cred, &bp)) { 613 free(base, M_UFSMNT); 614 goto out; 615 } 616 bcopy(bp->b_data, space, (u_int)size); 617 fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space; 618 space += size; 619 brelse(bp); 620 bp = NULL; 621 } 622 if (fs->fs_contigsumsize > 0) { 623 fs->fs_maxcluster = lp = (int32_t *)space; 624 for (i = 0; i < fs->fs_ncg; i++) 625 *lp++ = fs->fs_contigsumsize; 626 } 627 mp->mnt_data = (qaddr_t)ump; 628 mp->mnt_stat.f_fsid.val[0] = (long)dev; 629 if (fs->fs_id[0] != 0 && fs->fs_id[1] != 0) 630 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1]; 631 else 632 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 633 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 634 mp->mnt_flag |= MNT_LOCAL; 635 ump->um_mountp = mp; 636 ump->um_dev = dev; 637 ump->um_devvp = devvp; 638 ump->um_nindir = fs->fs_nindir; 639 ump->um_bptrtodb = fs->fs_fsbtodb; 640 ump->um_seqinc = fs->fs_frag; 641 for (i = 0; i < MAXQUOTAS; i++) 642 ump->um_quotas[i] = NULLVP; 643 devvp->v_specflags |= SI_MOUNTEDON; 644 ffs_oldfscompat(fs); 645 646 /* 647 * Set FS local "last mounted on" information (NULL pad) 648 */ 649 copystr( mp->mnt_stat.f_mntonname, /* mount point*/ 650 fs->fs_fsmnt, /* copy area*/ 651 sizeof(fs->fs_fsmnt) - 1, /* max size*/ 652 &strsize); /* real size*/ 653 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize); 654 655 if( mp->mnt_flag & MNT_ROOTFS) { 656 /* 657 * Root mount; update timestamp in mount structure. 658 * this will be used by the common root mount code 659 * to update the system clock. 660 */ 661 mp->mnt_time = fs->fs_time; 662 } 663 664 ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */ 665 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */ 666 if (fs->fs_maxfilesize > maxfilesize) /* XXX */ 667 fs->fs_maxfilesize = maxfilesize; /* XXX */ 668 if (ronly == 0) { 669 fs->fs_clean = 0; 670 (void) ffs_sbupdate(ump, MNT_WAIT); 671 } 672 return (0); 673out: 674 if (bp) 675 brelse(bp); 676 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p); 677 if (ump) { 678 free(ump->um_fs, M_UFSMNT); 679 free(ump, M_UFSMNT); 680 mp->mnt_data = (qaddr_t)0; 681 } 682 return (error); 683} 684 685/* 686 * Sanity checks for old file systems. 687 * 688 * XXX - goes away some day. 689 */ 690static int 691ffs_oldfscompat(fs) 692 struct fs *fs; 693{ 694 695 fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */ 696 fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */ 697 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */ 698 fs->fs_nrpos = 8; /* XXX */ 699 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 700#if 0 701 int i; /* XXX */ 702 u_int64_t sizepb = fs->fs_bsize; /* XXX */ 703 /* XXX */ 704 fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */ 705 for (i = 0; i < NIADDR; i++) { /* XXX */ 706 sizepb *= NINDIR(fs); /* XXX */ 707 fs->fs_maxfilesize += sizepb; /* XXX */ 708 } /* XXX */ 709#endif 710 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 711 fs->fs_qbmask = ~fs->fs_bmask; /* XXX */ 712 fs->fs_qfmask = ~fs->fs_fmask; /* XXX */ 713 } /* XXX */ 714 return (0); 715} 716 717/* 718 * unmount system call 719 */ 720int 721ffs_unmount(mp, mntflags, p) 722 struct mount *mp; 723 int mntflags; 724 struct proc *p; 725{ 726 register struct ufsmount *ump; 727 register struct fs *fs; 728 int error, flags; 729 730 flags = 0; 731 if (mntflags & MNT_FORCE) { 732 flags |= FORCECLOSE; 733 } 734 error = ffs_flushfiles(mp, flags, p); 735 if (error) 736 return (error); 737 ump = VFSTOUFS(mp); 738 fs = ump->um_fs; 739 if (fs->fs_ronly == 0) { 740 fs->fs_clean = 1; 741 error = ffs_sbupdate(ump, MNT_WAIT); 742 if (error) { 743 fs->fs_clean = 0; 744 return (error); 745 } 746 } 747 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON; 748 749 vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, p, 0, 0); 750 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE, 751 NOCRED, p); 752 753 vrele(ump->um_devvp); 754 755 free(fs->fs_csp[0], M_UFSMNT); 756 free(fs, M_UFSMNT); 757 free(ump, M_UFSMNT); 758 mp->mnt_data = (qaddr_t)0; 759 mp->mnt_flag &= ~MNT_LOCAL; 760 return (error); 761} 762 763/* 764 * Flush out all the files in a filesystem. 765 */ 766int 767ffs_flushfiles(mp, flags, p) 768 register struct mount *mp; 769 int flags; 770 struct proc *p; 771{ 772 register struct ufsmount *ump; 773 int error; 774 775 ump = VFSTOUFS(mp); 776#ifdef QUOTA 777 if (mp->mnt_flag & MNT_QUOTA) { 778 int i; 779 error = vflush(mp, NULLVP, SKIPSYSTEM|flags); 780 if (error) 781 return (error); 782 for (i = 0; i < MAXQUOTAS; i++) { 783 if (ump->um_quotas[i] == NULLVP) 784 continue; 785 quotaoff(p, mp, i); 786 } 787 /* 788 * Here we fall through to vflush again to ensure 789 * that we have gotten rid of all the system vnodes. 790 */ 791 } 792#endif 793 error = vflush(mp, NULLVP, flags); 794 return (error); 795} 796 797/* 798 * Get file system statistics. 799 */ 800int 801ffs_statfs(mp, sbp, p) 802 struct mount *mp; 803 register struct statfs *sbp; 804 struct proc *p; 805{ 806 register struct ufsmount *ump; 807 register struct fs *fs; 808 809 ump = VFSTOUFS(mp); 810 fs = ump->um_fs; 811 if (fs->fs_magic != FS_MAGIC) 812 panic("ffs_statfs"); 813 sbp->f_bsize = fs->fs_fsize; 814 sbp->f_iosize = fs->fs_bsize; 815 sbp->f_blocks = fs->fs_dsize; 816 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag + 817 fs->fs_cstotal.cs_nffree; 818 sbp->f_bavail = freespace(fs, fs->fs_minfree); 819 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO; 820 sbp->f_ffree = fs->fs_cstotal.cs_nifree; 821 if (sbp != &mp->mnt_stat) { 822 sbp->f_type = mp->mnt_vfc->vfc_typenum; 823 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 824 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 825 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 826 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 827 } 828 return (0); 829} 830 831/* 832 * Go through the disk queues to initiate sandbagged IO; 833 * go through the inodes to write those that have been modified; 834 * initiate the writing of the super block if it has been modified. 835 * 836 * Note: we are always called with the filesystem marked `MPBUSY'. 837 */ 838int 839ffs_sync(mp, waitfor, cred, p) 840 struct mount *mp; 841 int waitfor; 842 struct ucred *cred; 843 struct proc *p; 844{ 845 struct vnode *nvp, *vp; 846 struct inode *ip; 847 struct ufsmount *ump = VFSTOUFS(mp); 848 struct fs *fs; 849 struct timeval tv; 850 int error, allerror = 0; 851 852 fs = ump->um_fs; 853 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 854 printf("fs = %s\n", fs->fs_fsmnt); 855 panic("ffs_sync: rofs mod"); 856 } 857 /* 858 * Write back each (modified) inode. 859 */ 860 simple_lock(&mntvnode_slock); 861loop: 862 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 863 /* 864 * If the vnode that we are about to sync is no longer 865 * associated with this mount point, start over. 866 */ 867 if (vp->v_mount != mp) 868 goto loop; 869 simple_lock(&vp->v_interlock); 870 nvp = vp->v_mntvnodes.le_next; 871 ip = VTOI(vp); 872 if (((ip->i_flag & 873 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0) && 874 vp->v_dirtyblkhd.lh_first == NULL) { 875 simple_unlock(&vp->v_interlock); 876 continue; 877 } 878 if (vp->v_type != VCHR) { 879 simple_unlock(&mntvnode_slock); 880 error = 881 vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p); 882 if (error) { 883 simple_lock(&mntvnode_slock); 884 if (error == ENOENT) 885 goto loop; 886 continue; 887 } 888 if (error = VOP_FSYNC(vp, cred, waitfor, p)) 889 allerror = error; 890 VOP_UNLOCK(vp, 0, p); 891 vrele(vp); 892 simple_lock(&mntvnode_slock); 893 } else { 894 simple_unlock(&mntvnode_slock); 895 simple_unlock(&vp->v_interlock); 896 gettime(&tv); 897 /* UFS_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT); */ 898 UFS_UPDATE(vp, &tv, &tv, 0); 899 simple_lock(&mntvnode_slock); 900 } 901 } 902 simple_unlock(&mntvnode_slock); 903 /* 904 * Force stale file system control information to be flushed. 905 */ 906 error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p); 907 if (error) 908 allerror = error; 909#ifdef QUOTA 910 qsync(mp); 911#endif 912 /* 913 * Write back modified superblock. 914 */ 915 if (fs->fs_fmod != 0) { 916 fs->fs_fmod = 0; 917 fs->fs_time = time.tv_sec; 918 if (error = ffs_sbupdate(ump, waitfor)) 919 allerror = error; 920 } 921 return (allerror); 922} 923 924/* 925 * Look up a FFS dinode number to find its incore vnode, otherwise read it 926 * in from disk. If it is in core, wait for the lock bit to clear, then 927 * return the inode locked. Detection and handling of mount points must be 928 * done by the calling routine. 929 */ 930static int ffs_inode_hash_lock; 931 932int 933ffs_vget(mp, ino, vpp) 934 struct mount *mp; 935 ino_t ino; 936 struct vnode **vpp; 937{ 938 struct fs *fs; 939 struct inode *ip; 940 struct ufsmount *ump; 941 struct buf *bp; 942 struct vnode *vp; 943 dev_t dev; 944 int error; 945 946 ump = VFSTOUFS(mp); 947 dev = ump->um_dev; 948restart: 949 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 950 return (0); 951 952 /* 953 * Lock out the creation of new entries in the FFS hash table in 954 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 955 * may occur! 956 */ 957 if (ffs_inode_hash_lock) { 958 while (ffs_inode_hash_lock) { 959 ffs_inode_hash_lock = -1; 960 tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0); 961 } 962 goto restart; 963 } 964 ffs_inode_hash_lock = 1; 965 966 /* 967 * If this MALLOC() is performed after the getnewvnode() 968 * it might block, leaving a vnode with a NULL v_data to be 969 * found by ffs_sync() if a sync happens to fire right then, 970 * which will cause a panic because ffs_sync() blindly 971 * dereferences vp->v_data (as well it should). 972 */ 973 MALLOC(ip, struct inode *, sizeof(struct inode), 974 ump->um_malloctype, M_WAITOK); 975 976 /* Allocate a new vnode/inode. */ 977 error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp); 978 if (error) { 979 if (ffs_inode_hash_lock < 0) 980 wakeup(&ffs_inode_hash_lock); 981 ffs_inode_hash_lock = 0; 982 *vpp = NULL; 983 FREE(ip, ump->um_malloctype); 984 return (error); 985 } 986 bzero((caddr_t)ip, sizeof(struct inode)); 987 lockinit(&ip->i_lock, PINOD, "inode", 0, 0); 988 vp->v_data = ip; 989 ip->i_vnode = vp; 990 ip->i_fs = fs = ump->um_fs; 991 ip->i_dev = dev; 992 ip->i_number = ino; 993#ifdef QUOTA 994 { 995 int i; 996 for (i = 0; i < MAXQUOTAS; i++) 997 ip->i_dquot[i] = NODQUOT; 998 } 999#endif 1000 /* 1001 * Put it onto its hash chain and lock it so that other requests for 1002 * this inode will block if they arrive while we are sleeping waiting 1003 * for old data structures to be purged or for the contents of the 1004 * disk portion of this inode to be read. 1005 */ 1006 ufs_ihashins(ip); 1007 1008 if (ffs_inode_hash_lock < 0) 1009 wakeup(&ffs_inode_hash_lock); 1010 ffs_inode_hash_lock = 0; 1011 1012 /* Read in the disk contents for the inode, copy into the inode. */ 1013 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1014 (int)fs->fs_bsize, NOCRED, &bp); 1015 if (error) { 1016 /* 1017 * The inode does not contain anything useful, so it would 1018 * be misleading to leave it on its hash chain. With mode 1019 * still zero, it will be unlinked and returned to the free 1020 * list by vput(). 1021 */ 1022 brelse(bp); 1023 vput(vp); 1024 *vpp = NULL; 1025 return (error); 1026 } 1027 ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino)); 1028 bqrelse(bp); 1029 1030 /* 1031 * Initialize the vnode from the inode, check for aliases. 1032 * Note that the underlying vnode may have changed. 1033 */ 1034 error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1035 if (error) { 1036 vput(vp); 1037 *vpp = NULL; 1038 return (error); 1039 } 1040 /* 1041 * Finish inode initialization now that aliasing has been resolved. 1042 */ 1043 ip->i_devvp = ump->um_devvp; 1044 VREF(ip->i_devvp); 1045 /* 1046 * Set up a generation number for this inode if it does not 1047 * already have one. This should only happen on old filesystems. 1048 */ 1049 if (ip->i_gen == 0) { 1050 ip->i_gen = random() / 2 + 1; 1051 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1052 ip->i_flag |= IN_MODIFIED; 1053 } 1054 /* 1055 * Ensure that uid and gid are correct. This is a temporary 1056 * fix until fsck has been changed to do the update. 1057 */ 1058 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 1059 ip->i_uid = ip->i_din.di_ouid; /* XXX */ 1060 ip->i_gid = ip->i_din.di_ogid; /* XXX */ 1061 } /* XXX */ 1062 1063 *vpp = vp; 1064 return (0); 1065} 1066 1067/* 1068 * File handle to vnode 1069 * 1070 * Have to be really careful about stale file handles: 1071 * - check that the inode number is valid 1072 * - call ffs_vget() to get the locked inode 1073 * - check for an unallocated inode (i_mode == 0) 1074 * - check that the given client host has export rights and return 1075 * those rights via. exflagsp and credanonp 1076 */ 1077int 1078ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) 1079 register struct mount *mp; 1080 struct fid *fhp; 1081 struct sockaddr *nam; 1082 struct vnode **vpp; 1083 int *exflagsp; 1084 struct ucred **credanonp; 1085{ 1086 register struct ufid *ufhp; 1087 struct fs *fs; 1088 1089 ufhp = (struct ufid *)fhp; 1090 fs = VFSTOUFS(mp)->um_fs; 1091 if (ufhp->ufid_ino < ROOTINO || 1092 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) 1093 return (ESTALE); 1094 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)); 1095} 1096 1097/* 1098 * Vnode pointer to File handle 1099 */ 1100/* ARGSUSED */ 1101int 1102ffs_vptofh(vp, fhp) 1103 struct vnode *vp; 1104 struct fid *fhp; 1105{ 1106 register struct inode *ip; 1107 register struct ufid *ufhp; 1108 1109 ip = VTOI(vp); 1110 ufhp = (struct ufid *)fhp; 1111 ufhp->ufid_len = sizeof(struct ufid); 1112 ufhp->ufid_ino = ip->i_number; 1113 ufhp->ufid_gen = ip->i_gen; 1114 return (0); 1115} 1116 1117/* 1118 * Initialize the filesystem; just use ufs_init. 1119 */ 1120static int 1121ffs_init(vfsp) 1122 struct vfsconf *vfsp; 1123{ 1124 1125 return (ufs_init(vfsp)); 1126} 1127 1128/* 1129 * Write a superblock and associated information back to disk. 1130 */ 1131static int 1132ffs_sbupdate(mp, waitfor) 1133 struct ufsmount *mp; 1134 int waitfor; 1135{ 1136 register struct fs *dfs, *fs = mp->um_fs; 1137 register struct buf *bp; 1138 int blks; 1139 caddr_t space; 1140 int i, size, error, allerror = 0; 1141 1142 /* 1143 * First write back the summary information. 1144 */ 1145 blks = howmany(fs->fs_cssize, fs->fs_fsize); 1146 space = (caddr_t)fs->fs_csp[0]; 1147 for (i = 0; i < blks; i += fs->fs_frag) { 1148 size = fs->fs_bsize; 1149 if (i + fs->fs_frag > blks) 1150 size = (blks - i) * fs->fs_fsize; 1151 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), 1152 size, 0, 0); 1153 bcopy(space, bp->b_data, (u_int)size); 1154 space += size; 1155 if (waitfor != MNT_WAIT) 1156 bawrite(bp); 1157 else if (error = bwrite(bp)) 1158 allerror = error; 1159 } 1160 /* 1161 * Now write back the superblock itself. If any errors occurred 1162 * up to this point, then fail so that the superblock avoids 1163 * being written out as clean. 1164 */ 1165 if (allerror) 1166 return (allerror); 1167 bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0); 1168 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); 1169 /* Restore compatibility to old file systems. XXX */ 1170 dfs = (struct fs *)bp->b_data; /* XXX */ 1171 if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */ 1172 dfs->fs_nrpos = -1; /* XXX */ 1173 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */ 1174 int32_t *lp, tmp; /* XXX */ 1175 /* XXX */ 1176 lp = (int32_t *)&dfs->fs_qbmask; /* XXX */ 1177 tmp = lp[4]; /* XXX */ 1178 for (i = 4; i > 0; i--) /* XXX */ 1179 lp[i] = lp[i-1]; /* XXX */ 1180 lp[0] = tmp; /* XXX */ 1181 } /* XXX */ 1182 dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */ 1183 if (waitfor != MNT_WAIT) 1184 bawrite(bp); 1185 else if (error = bwrite(bp)) 1186 allerror = error; 1187 return (allerror); 1188} 1189