1 /* 2 * Copyright (c) 1982, 1986, 1989 Regents of the University of California. 3 * All rights reserved. 4 * 5 * %sccs.include.redist.c% 6 * 7 * @(#)ffs_vnops.c 7.92 (Berkeley) 07/17/92 8 */ 9 10 #include <sys/param.h> 11 #include <sys/systm.h> 12 #include <sys/resourcevar.h> 13 #include <sys/kernel.h> 14 #include <sys/file.h> 15 #include <sys/stat.h> 16 #include <sys/buf.h> 17 #include <sys/proc.h> 18 #include <sys/conf.h> 19 #include <sys/mount.h> 20 #include <sys/vnode.h> 21 #include <sys/malloc.h> 22 23 #include <vm/vm.h> 24 25 #include <miscfs/specfs/specdev.h> 26 #include <miscfs/fifofs/fifo.h> 27 28 #include <ufs/ufs/lockf.h> 29 #include <ufs/ufs/quota.h> 30 #include <ufs/ufs/inode.h> 31 #include <ufs/ufs/dir.h> 32 #include <ufs/ufs/ufs_extern.h> 33 34 #include <ufs/ffs/fs.h> 35 #include <ufs/ffs/ffs_extern.h> 36 37 /* Global vfs data structures for ufs. */ 38 int (**ffs_vnodeop_p)(); 39 struct vnodeopv_entry_desc ffs_vnodeop_entries[] = { 40 { &vop_default_desc, vn_default_error }, 41 { &vop_lookup_desc, ufs_lookup }, /* lookup */ 42 { &vop_create_desc, ufs_create }, /* create */ 43 { &vop_mknod_desc, ufs_mknod }, /* mknod */ 44 { &vop_open_desc, ufs_open }, /* open */ 45 { &vop_close_desc, ufs_close }, /* close */ 46 { &vop_access_desc, ufs_access }, /* access */ 47 { &vop_getattr_desc, ufs_getattr }, /* getattr */ 48 { &vop_setattr_desc, ufs_setattr }, /* setattr */ 49 { &vop_read_desc, ffs_read }, /* read */ 50 { &vop_write_desc, ffs_write }, /* write */ 51 { &vop_ioctl_desc, ufs_ioctl }, /* ioctl */ 52 { &vop_select_desc, ufs_select }, /* select */ 53 { &vop_mmap_desc, ufs_mmap }, /* mmap */ 54 { &vop_fsync_desc, ffs_fsync }, /* fsync */ 55 { &vop_seek_desc, ufs_seek }, /* seek */ 56 { &vop_remove_desc, ufs_remove }, /* remove */ 57 { &vop_link_desc, ufs_link }, /* link */ 58 { &vop_rename_desc, ufs_rename }, /* rename */ 59 { &vop_mkdir_desc, ufs_mkdir }, /* mkdir */ 60 { &vop_rmdir_desc, ufs_rmdir }, /* rmdir */ 61 { &vop_symlink_desc, ufs_symlink }, /* symlink */ 62 { &vop_readdir_desc, ufs_readdir }, /* readdir */ 63 { &vop_readlink_desc, ufs_readlink }, /* readlink */ 64 { &vop_abortop_desc, ufs_abortop }, /* abortop */ 65 { &vop_inactive_desc, ffs_inactive }, /* inactive */ 66 { &vop_reclaim_desc, ufs_reclaim }, /* reclaim */ 67 { &vop_lock_desc, ufs_lock }, /* lock */ 68 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 69 { &vop_bmap_desc, ffs_bmap }, /* bmap */ 70 { &vop_strategy_desc, ufs_strategy }, /* strategy */ 71 { &vop_print_desc, ufs_print }, /* print */ 72 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 73 { &vop_advlock_desc, ufs_advlock }, /* advlock */ 74 { &vop_blkatoff_desc, ffs_blkatoff }, /* blkatoff */ 75 { &vop_valloc_desc, ffs_valloc }, /* valloc */ 76 { &vop_vfree_desc, ffs_vfree }, /* vfree */ 77 { &vop_truncate_desc, ffs_truncate }, /* truncate */ 78 { &vop_update_desc, ffs_update }, /* update */ 79 { &vop_bwrite_desc, vn_bwrite }, 80 { (struct vnodeop_desc*)NULL, (int(*)())NULL } 81 }; 82 struct vnodeopv_desc ffs_vnodeop_opv_desc = 83 { &ffs_vnodeop_p, ffs_vnodeop_entries }; 84 85 int (**ffs_specop_p)(); 86 struct vnodeopv_entry_desc ffs_specop_entries[] = { 87 { &vop_default_desc, vn_default_error }, 88 { &vop_lookup_desc, spec_lookup }, /* lookup */ 89 { &vop_create_desc, spec_create }, /* create */ 90 { &vop_mknod_desc, spec_mknod }, /* mknod */ 91 { &vop_open_desc, spec_open }, /* open */ 92 { &vop_close_desc, ufsspec_close }, /* close */ 93 { &vop_access_desc, ufs_access }, /* access */ 94 { &vop_getattr_desc, ufs_getattr }, /* getattr */ 95 { &vop_setattr_desc, ufs_setattr }, /* setattr */ 96 { &vop_read_desc, ufsspec_read }, /* read */ 97 { &vop_write_desc, ufsspec_write }, /* write */ 98 { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ 99 { &vop_select_desc, spec_select }, /* select */ 100 { &vop_mmap_desc, spec_mmap }, /* mmap */ 101 { &vop_fsync_desc, ffs_fsync }, /* fsync */ 102 { &vop_seek_desc, spec_seek }, /* seek */ 103 { &vop_remove_desc, spec_remove }, /* remove */ 104 { &vop_link_desc, spec_link }, /* link */ 105 { &vop_rename_desc, spec_rename }, /* rename */ 106 { &vop_mkdir_desc, spec_mkdir }, /* mkdir */ 107 { &vop_rmdir_desc, spec_rmdir }, /* rmdir */ 108 { &vop_symlink_desc, spec_symlink }, /* symlink */ 109 { &vop_readdir_desc, spec_readdir }, /* readdir */ 110 { &vop_readlink_desc, spec_readlink }, /* readlink */ 111 { &vop_abortop_desc, spec_abortop }, /* abortop */ 112 { &vop_inactive_desc, ffs_inactive }, /* inactive */ 113 { &vop_reclaim_desc, ufs_reclaim }, /* reclaim */ 114 { &vop_lock_desc, ufs_lock }, /* lock */ 115 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 116 { &vop_bmap_desc, spec_bmap }, /* bmap */ 117 { &vop_strategy_desc, spec_strategy }, /* strategy */ 118 { &vop_print_desc, ufs_print }, /* print */ 119 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 120 { &vop_advlock_desc, spec_advlock }, /* advlock */ 121 { &vop_blkatoff_desc, spec_blkatoff }, /* blkatoff */ 122 { &vop_valloc_desc, spec_valloc }, /* valloc */ 123 { &vop_vfree_desc, ffs_vfree }, /* vfree */ 124 { &vop_truncate_desc, spec_truncate }, /* truncate */ 125 { &vop_update_desc, ffs_update }, /* update */ 126 { &vop_bwrite_desc, vn_bwrite }, 127 { (struct vnodeop_desc*)NULL, (int(*)())NULL } 128 }; 129 struct vnodeopv_desc ffs_specop_opv_desc = 130 { &ffs_specop_p, ffs_specop_entries }; 131 132 #ifdef FIFO 133 int (**ffs_fifoop_p)(); 134 struct vnodeopv_entry_desc ffs_fifoop_entries[] = { 135 { &vop_default_desc, vn_default_error }, 136 { &vop_lookup_desc, fifo_lookup }, /* lookup */ 137 { &vop_create_desc, fifo_create }, /* create */ 138 { &vop_mknod_desc, fifo_mknod }, /* mknod */ 139 { &vop_open_desc, fifo_open }, /* open */ 140 { &vop_close_desc, ufsfifo_close }, /* close */ 141 { &vop_access_desc, ufs_access }, /* access */ 142 { &vop_getattr_desc, ufs_getattr }, /* getattr */ 143 { &vop_setattr_desc, ufs_setattr }, /* setattr */ 144 { &vop_read_desc, ufsfifo_read }, /* read */ 145 { &vop_write_desc, ufsfifo_write }, /* write */ 146 { &vop_ioctl_desc, fifo_ioctl }, /* ioctl */ 147 { &vop_select_desc, fifo_select }, /* select */ 148 { &vop_mmap_desc, fifo_mmap }, /* mmap */ 149 { &vop_fsync_desc, ffs_fsync }, /* fsync */ 150 { &vop_seek_desc, fifo_seek }, /* seek */ 151 { &vop_remove_desc, fifo_remove }, /* remove */ 152 { &vop_link_desc, fifo_link }, /* link */ 153 { &vop_rename_desc, fifo_rename }, /* rename */ 154 { &vop_mkdir_desc, fifo_mkdir }, /* mkdir */ 155 { &vop_rmdir_desc, fifo_rmdir }, /* rmdir */ 156 { &vop_symlink_desc, fifo_symlink }, /* symlink */ 157 { &vop_readdir_desc, fifo_readdir }, /* readdir */ 158 { &vop_readlink_desc, fifo_readlink }, /* readlink */ 159 { &vop_abortop_desc, fifo_abortop }, /* abortop */ 160 { &vop_inactive_desc, ffs_inactive }, /* inactive */ 161 { &vop_reclaim_desc, ufs_reclaim }, /* reclaim */ 162 { &vop_lock_desc, ufs_lock }, /* lock */ 163 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 164 { &vop_bmap_desc, fifo_bmap }, /* bmap */ 165 { &vop_strategy_desc, fifo_strategy }, /* strategy */ 166 { &vop_print_desc, ufs_print }, /* print */ 167 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 168 { &vop_advlock_desc, fifo_advlock }, /* advlock */ 169 { &vop_blkatoff_desc, fifo_blkatoff }, /* blkatoff */ 170 { &vop_valloc_desc, fifo_valloc }, /* valloc */ 171 { &vop_vfree_desc, ffs_vfree }, /* vfree */ 172 { &vop_truncate_desc, fifo_truncate }, /* truncate */ 173 { &vop_update_desc, ffs_update }, /* update */ 174 { &vop_bwrite_desc, vn_bwrite }, 175 { (struct vnodeop_desc*)NULL, (int(*)())NULL } 176 }; 177 struct vnodeopv_desc ffs_fifoop_opv_desc = 178 { &ffs_fifoop_p, ffs_fifoop_entries }; 179 #endif /* FIFO */ 180 181 182 /* 183 * Vnode op for reading. 184 */ 185 /* ARGSUSED */ 186 ffs_read(ap) 187 struct vop_read_args /* { 188 struct vnode *a_vp; 189 struct uio *a_uio; 190 int a_ioflag; 191 struct ucred *a_cred; 192 } */ *ap; 193 { 194 register struct vnode *vp = ap->a_vp; 195 register struct inode *ip = VTOI(vp); 196 register struct uio *uio = ap->a_uio; 197 register struct fs *fs; 198 struct buf *bp; 199 daddr_t lbn, bn, rablock; 200 off_t diff; 201 int rasize, error = 0; 202 long size, n, on; 203 204 #ifdef DIAGNOSTIC 205 int type; 206 if (uio->uio_rw != UIO_READ) 207 panic("ffs_read mode"); 208 type = ip->i_mode & IFMT; 209 if (type != IFDIR && type != IFREG && type != IFLNK) 210 panic("ffs_read type"); 211 if (type == IFLNK && (int)ip->i_size < vp->v_mount->mnt_maxsymlinklen) 212 panic("read short symlink"); 213 #endif 214 if (uio->uio_resid == 0) 215 return (0); 216 fs = ip->i_fs; 217 if ((u_quad_t)uio->uio_offset > fs->fs_maxfilesize) 218 return (EFBIG); 219 ip->i_flag |= IACC; 220 do { 221 lbn = lblkno(fs, uio->uio_offset); 222 on = blkoff(fs, uio->uio_offset); 223 n = min((unsigned)(fs->fs_bsize - on), uio->uio_resid); 224 diff = ip->i_size - uio->uio_offset; 225 if (diff <= 0) 226 return (0); 227 if (diff < n) 228 n = diff; 229 size = blksize(fs, ip, lbn); 230 rablock = lbn + 1; 231 if (vp->v_lastr + 1 == lbn && 232 lblktosize(fs, rablock) < ip->i_size) { 233 rasize = blksize(fs, ip, rablock); 234 error = breadn(vp, lbn, size, &rablock, 235 &rasize, 1, NOCRED, &bp); 236 } else 237 error = bread(vp, lbn, size, NOCRED, &bp); 238 vp->v_lastr = lbn; 239 n = min(n, size - bp->b_resid); 240 if (error) { 241 brelse(bp); 242 return (error); 243 } 244 error = uiomove(bp->b_un.b_addr + on, (int)n, uio); 245 if (n + on == fs->fs_bsize || uio->uio_offset == ip->i_size) 246 bp->b_flags |= B_AGE; 247 brelse(bp); 248 } while (error == 0 && uio->uio_resid > 0 && n != 0); 249 return (error); 250 } 251 252 /* 253 * Vnode op for writing. 254 */ 255 ffs_write(ap) 256 struct vop_write_args /* { 257 struct vnode *a_vp; 258 struct uio *a_uio; 259 int a_ioflag; 260 struct ucred *a_cred; 261 } */ *ap; 262 { 263 register struct vnode *vp = ap->a_vp; 264 register struct uio *uio = ap->a_uio; 265 register struct inode *ip = VTOI(vp); 266 register struct fs *fs; 267 struct proc *p = uio->uio_procp; 268 int ioflag = ap->a_ioflag; 269 struct timeval tv; 270 struct buf *bp; 271 daddr_t lbn, bn; 272 off_t osize; 273 int n, on, flags; 274 int size, resid, error = 0; 275 276 #ifdef DIAGNOSTIC 277 if (uio->uio_rw != UIO_WRITE) 278 panic("ffs_write mode"); 279 #endif 280 switch (vp->v_type) { 281 case VREG: 282 if (ioflag & IO_APPEND) 283 uio->uio_offset = ip->i_size; 284 /* fall through */ 285 case VLNK: 286 break; 287 288 case VDIR: 289 if ((ioflag & IO_SYNC) == 0) 290 panic("ffs_write nonsync dir write"); 291 break; 292 293 default: 294 panic("ffs_write type"); 295 } 296 if (uio->uio_resid == 0) 297 return (0); 298 fs = ip->i_fs; 299 if (uio->uio_offset < 0 || 300 (u_quad_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize) 301 return (EFBIG); 302 /* 303 * Maybe this should be above the vnode op call, but so long as 304 * file servers have no limits, i don't think it matters 305 */ 306 if (vp->v_type == VREG && p && 307 uio->uio_offset + uio->uio_resid > 308 p->p_rlimit[RLIMIT_FSIZE].rlim_cur) { 309 psignal(p, SIGXFSZ); 310 return (EFBIG); 311 } 312 resid = uio->uio_resid; 313 osize = ip->i_size; 314 flags = 0; 315 if (ioflag & IO_SYNC) 316 flags = B_SYNC; 317 do { 318 lbn = lblkno(fs, uio->uio_offset); 319 on = blkoff(fs, uio->uio_offset); 320 n = min((unsigned)(fs->fs_bsize - on), uio->uio_resid); 321 if (n < fs->fs_bsize) 322 flags |= B_CLRBUF; 323 else 324 flags &= ~B_CLRBUF; 325 if (error = ffs_balloc(ip, lbn, on + n, ap->a_cred, &bp, flags)) 326 break; 327 bn = bp->b_blkno; 328 if (uio->uio_offset + n > ip->i_size) { 329 ip->i_size = uio->uio_offset + n; 330 vnode_pager_setsize(vp, (u_long)ip->i_size); 331 } 332 size = blksize(fs, ip, lbn); 333 (void) vnode_pager_uncache(vp); 334 n = min(n, size - bp->b_resid); 335 error = uiomove(bp->b_un.b_addr + on, n, uio); 336 if (ioflag & IO_SYNC) 337 (void) bwrite(bp); 338 else if (n + on == fs->fs_bsize) { 339 bp->b_flags |= B_AGE; 340 bawrite(bp); 341 } else 342 bdwrite(bp); 343 ip->i_flag |= IUPD|ICHG; 344 if (ap->a_cred->cr_uid != 0) 345 ip->i_mode &= ~(ISUID|ISGID); 346 } while (error == 0 && uio->uio_resid > 0 && n != 0); 347 if (error && (ioflag & IO_UNIT)) { 348 (void)VOP_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred, 349 uio->uio_procp); 350 uio->uio_offset -= resid - uio->uio_resid; 351 uio->uio_resid = resid; 352 } 353 if (!error && (ioflag & IO_SYNC)) { 354 tv = time; 355 error = VOP_UPDATE(vp, &tv, &tv, 1); 356 } 357 return (error); 358 } 359 360 /* 361 * Synch an open file. 362 */ 363 /* ARGSUSED */ 364 int 365 ffs_fsync(ap) 366 struct vop_fsync_args /* { 367 struct vnode *a_vp; 368 struct ucred *a_cred; 369 int a_waitfor; 370 struct proc *a_p; 371 } */ *ap; 372 { 373 register struct vnode *vp = ap->a_vp; 374 struct inode *ip = VTOI(vp); 375 register struct buf *bp; 376 struct timeval tv; 377 struct buf *nbp; 378 int s; 379 380 /* 381 * Flush all dirty buffers associated with a vnode. 382 */ 383 loop: 384 s = splbio(); 385 for (bp = vp->v_dirtyblkhd; bp; bp = nbp) { 386 nbp = bp->b_blockf; 387 if ((bp->b_flags & B_BUSY)) 388 continue; 389 if ((bp->b_flags & B_DELWRI) == 0) 390 panic("ffs_fsync: not dirty"); 391 bremfree(bp); 392 bp->b_flags |= B_BUSY; 393 splx(s); 394 /* 395 * Wait for I/O associated with indirect blocks to complete, 396 * since there is no way to quickly wait for them below. 397 */ 398 if (bp->b_vp == vp || ap->a_waitfor == MNT_NOWAIT) 399 (void) bawrite(bp); 400 else 401 (void) bwrite(bp); 402 goto loop; 403 } 404 if (ap->a_waitfor == MNT_WAIT) { 405 while (vp->v_numoutput) { 406 vp->v_flag |= VBWAIT; 407 sleep((caddr_t)&vp->v_numoutput, PRIBIO + 1); 408 } 409 #ifdef DIAGNOSTIC 410 if (vp->v_dirtyblkhd) { 411 vprint("ffs_fsync: dirty", vp); 412 goto loop; 413 } 414 #endif 415 } 416 splx(s); 417 tv = time; 418 return (VOP_UPDATE(ap->a_vp, &tv, &tv, ap->a_waitfor == MNT_WAIT)); 419 } 420 421 /* 422 * Last reference to an inode, write the inode out and if necessary, 423 * truncate and deallocate the file. 424 */ 425 int 426 ffs_inactive(ap) 427 struct vop_inactive_args /* { 428 struct vnode *a_vp; 429 } */ *ap; 430 { 431 register struct vnode *vp = ap->a_vp; 432 register struct inode *ip = VTOI(vp); 433 struct timeval tv; 434 int mode, error; 435 extern int prtactive; 436 437 if (prtactive && vp->v_usecount != 0) 438 vprint("ffs_inactive: pushing active", vp); 439 440 /* Get rid of inodes related to stale file handles. */ 441 if (ip->i_mode == 0) { 442 if ((vp->v_flag & VXLOCK) == 0) 443 vgone(vp); 444 return (0); 445 } 446 447 error = 0; 448 ILOCK(ip); 449 if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 450 #ifdef QUOTA 451 if (!getinoquota(ip)) 452 (void)chkiq(ip, -1, NOCRED, 0); 453 #endif 454 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL); 455 mode = ip->i_mode; 456 ip->i_mode = 0; 457 ip->i_rdev = 0; 458 ip->i_flag |= IUPD|ICHG; 459 VOP_VFREE(vp, ip->i_number, mode); 460 } 461 if (ip->i_flag&(IUPD|IACC|ICHG|IMOD)) { 462 tv = time; 463 VOP_UPDATE(vp, &tv, &tv, 0); 464 } 465 IUNLOCK(ip); 466 ip->i_flag = 0; 467 /* 468 * If we are done with the inode, reclaim it 469 * so that it can be reused immediately. 470 */ 471 if (vp->v_usecount == 0 && ip->i_mode == 0) 472 vgone(vp); 473 return (error); 474 } 475