1 /* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * %sccs.include.redist.c% 6 * 7 * @(#)ffs_vnops.c 8.1 (Berkeley) 06/11/93 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, ufs_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, ufs_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_pathconf_desc, ufs_pathconf }, /* pathconf */ 74 { &vop_advlock_desc, ufs_advlock }, /* advlock */ 75 { &vop_blkatoff_desc, ffs_blkatoff }, /* blkatoff */ 76 { &vop_valloc_desc, ffs_valloc }, /* valloc */ 77 { &vop_vfree_desc, ffs_vfree }, /* vfree */ 78 { &vop_truncate_desc, ffs_truncate }, /* truncate */ 79 { &vop_update_desc, ffs_update }, /* update */ 80 { &vop_bwrite_desc, vn_bwrite }, 81 { (struct vnodeop_desc*)NULL, (int(*)())NULL } 82 }; 83 struct vnodeopv_desc ffs_vnodeop_opv_desc = 84 { &ffs_vnodeop_p, ffs_vnodeop_entries }; 85 86 int (**ffs_specop_p)(); 87 struct vnodeopv_entry_desc ffs_specop_entries[] = { 88 { &vop_default_desc, vn_default_error }, 89 { &vop_lookup_desc, spec_lookup }, /* lookup */ 90 { &vop_create_desc, spec_create }, /* create */ 91 { &vop_mknod_desc, spec_mknod }, /* mknod */ 92 { &vop_open_desc, spec_open }, /* open */ 93 { &vop_close_desc, ufsspec_close }, /* close */ 94 { &vop_access_desc, ufs_access }, /* access */ 95 { &vop_getattr_desc, ufs_getattr }, /* getattr */ 96 { &vop_setattr_desc, ufs_setattr }, /* setattr */ 97 { &vop_read_desc, ufsspec_read }, /* read */ 98 { &vop_write_desc, ufsspec_write }, /* write */ 99 { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ 100 { &vop_select_desc, spec_select }, /* select */ 101 { &vop_mmap_desc, spec_mmap }, /* mmap */ 102 { &vop_fsync_desc, ffs_fsync }, /* fsync */ 103 { &vop_seek_desc, spec_seek }, /* seek */ 104 { &vop_remove_desc, spec_remove }, /* remove */ 105 { &vop_link_desc, spec_link }, /* link */ 106 { &vop_rename_desc, spec_rename }, /* rename */ 107 { &vop_mkdir_desc, spec_mkdir }, /* mkdir */ 108 { &vop_rmdir_desc, spec_rmdir }, /* rmdir */ 109 { &vop_symlink_desc, spec_symlink }, /* symlink */ 110 { &vop_readdir_desc, spec_readdir }, /* readdir */ 111 { &vop_readlink_desc, spec_readlink }, /* readlink */ 112 { &vop_abortop_desc, spec_abortop }, /* abortop */ 113 { &vop_inactive_desc, ufs_inactive }, /* inactive */ 114 { &vop_reclaim_desc, ufs_reclaim }, /* reclaim */ 115 { &vop_lock_desc, ufs_lock }, /* lock */ 116 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 117 { &vop_bmap_desc, spec_bmap }, /* bmap */ 118 { &vop_strategy_desc, spec_strategy }, /* strategy */ 119 { &vop_print_desc, ufs_print }, /* print */ 120 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 121 { &vop_pathconf_desc, spec_pathconf }, /* pathconf */ 122 { &vop_advlock_desc, spec_advlock }, /* advlock */ 123 { &vop_blkatoff_desc, spec_blkatoff }, /* blkatoff */ 124 { &vop_valloc_desc, spec_valloc }, /* valloc */ 125 { &vop_vfree_desc, ffs_vfree }, /* vfree */ 126 { &vop_truncate_desc, spec_truncate }, /* truncate */ 127 { &vop_update_desc, ffs_update }, /* update */ 128 { &vop_bwrite_desc, vn_bwrite }, 129 { (struct vnodeop_desc*)NULL, (int(*)())NULL } 130 }; 131 struct vnodeopv_desc ffs_specop_opv_desc = 132 { &ffs_specop_p, ffs_specop_entries }; 133 134 #ifdef FIFO 135 int (**ffs_fifoop_p)(); 136 struct vnodeopv_entry_desc ffs_fifoop_entries[] = { 137 { &vop_default_desc, vn_default_error }, 138 { &vop_lookup_desc, fifo_lookup }, /* lookup */ 139 { &vop_create_desc, fifo_create }, /* create */ 140 { &vop_mknod_desc, fifo_mknod }, /* mknod */ 141 { &vop_open_desc, fifo_open }, /* open */ 142 { &vop_close_desc, ufsfifo_close }, /* close */ 143 { &vop_access_desc, ufs_access }, /* access */ 144 { &vop_getattr_desc, ufs_getattr }, /* getattr */ 145 { &vop_setattr_desc, ufs_setattr }, /* setattr */ 146 { &vop_read_desc, ufsfifo_read }, /* read */ 147 { &vop_write_desc, ufsfifo_write }, /* write */ 148 { &vop_ioctl_desc, fifo_ioctl }, /* ioctl */ 149 { &vop_select_desc, fifo_select }, /* select */ 150 { &vop_mmap_desc, fifo_mmap }, /* mmap */ 151 { &vop_fsync_desc, ffs_fsync }, /* fsync */ 152 { &vop_seek_desc, fifo_seek }, /* seek */ 153 { &vop_remove_desc, fifo_remove }, /* remove */ 154 { &vop_link_desc, fifo_link }, /* link */ 155 { &vop_rename_desc, fifo_rename }, /* rename */ 156 { &vop_mkdir_desc, fifo_mkdir }, /* mkdir */ 157 { &vop_rmdir_desc, fifo_rmdir }, /* rmdir */ 158 { &vop_symlink_desc, fifo_symlink }, /* symlink */ 159 { &vop_readdir_desc, fifo_readdir }, /* readdir */ 160 { &vop_readlink_desc, fifo_readlink }, /* readlink */ 161 { &vop_abortop_desc, fifo_abortop }, /* abortop */ 162 { &vop_inactive_desc, ufs_inactive }, /* inactive */ 163 { &vop_reclaim_desc, ufs_reclaim }, /* reclaim */ 164 { &vop_lock_desc, ufs_lock }, /* lock */ 165 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 166 { &vop_bmap_desc, fifo_bmap }, /* bmap */ 167 { &vop_strategy_desc, fifo_strategy }, /* strategy */ 168 { &vop_print_desc, ufs_print }, /* print */ 169 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 170 { &vop_pathconf_desc, fifo_pathconf }, /* pathconf */ 171 { &vop_advlock_desc, fifo_advlock }, /* advlock */ 172 { &vop_blkatoff_desc, fifo_blkatoff }, /* blkatoff */ 173 { &vop_valloc_desc, fifo_valloc }, /* valloc */ 174 { &vop_vfree_desc, ffs_vfree }, /* vfree */ 175 { &vop_truncate_desc, fifo_truncate }, /* truncate */ 176 { &vop_update_desc, ffs_update }, /* update */ 177 { &vop_bwrite_desc, vn_bwrite }, 178 { (struct vnodeop_desc*)NULL, (int(*)())NULL } 179 }; 180 struct vnodeopv_desc ffs_fifoop_opv_desc = 181 { &ffs_fifoop_p, ffs_fifoop_entries }; 182 #endif /* FIFO */ 183 184 /* 185 * Enabling cluster read/write operations. 186 */ 187 #include <sys/sysctl.h> 188 int doclusterread = 1; 189 struct ctldebug debug11 = { "doclusterread", &doclusterread }; 190 int doclusterwrite = 1; 191 struct ctldebug debug12 = { "doclusterwrite", &doclusterwrite }; 192 193 /* 194 * Vnode op for reading. 195 */ 196 /* ARGSUSED */ 197 ffs_read(ap) 198 struct vop_read_args /* { 199 struct vnode *a_vp; 200 struct uio *a_uio; 201 int a_ioflag; 202 struct ucred *a_cred; 203 } */ *ap; 204 { 205 register struct vnode *vp = ap->a_vp; 206 register struct inode *ip = VTOI(vp); 207 register struct uio *uio = ap->a_uio; 208 register struct fs *fs; 209 struct buf *bp; 210 daddr_t lbn, bn, rablock; 211 off_t diff; 212 int type, rasize, error = 0; 213 long size, n, on; 214 215 type = ip->i_mode & IFMT; 216 #ifdef DIAGNOSTIC 217 if (uio->uio_rw != UIO_READ) 218 panic("ffs_read mode"); 219 if (type != IFDIR && type != IFREG && type != IFLNK) 220 panic("ffs_read type"); 221 if (type == IFLNK && (int)ip->i_size < vp->v_mount->mnt_maxsymlinklen) 222 panic("read short symlink"); 223 #endif 224 if (uio->uio_resid == 0) 225 return (0); 226 fs = ip->i_fs; 227 if ((u_quad_t)uio->uio_offset > fs->fs_maxfilesize) 228 return (EFBIG); 229 ip->i_flag |= IACC; 230 do { 231 lbn = lblkno(fs, uio->uio_offset); 232 on = blkoff(fs, uio->uio_offset); 233 n = min((unsigned)(fs->fs_bsize - on), uio->uio_resid); 234 diff = ip->i_size - uio->uio_offset; 235 if (diff <= 0) 236 return (0); 237 if (diff < n) 238 n = diff; 239 size = blksize(fs, ip, lbn); 240 rablock = lbn + 1; 241 if (doclusterread && lblktosize(fs, rablock) <= ip->i_size) { 242 error = cluster_read(vp, ip->i_size, lbn, size, 243 NOCRED, &bp); 244 } else if (vp->v_lastr + 1 == lbn && 245 lblktosize(fs, rablock) < ip->i_size) { 246 rasize = blksize(fs, ip, rablock); 247 error = breadn(vp, lbn, size, &rablock, 248 &rasize, 1, NOCRED, &bp); 249 } else 250 error = bread(vp, lbn, size, NOCRED, &bp); 251 vp->v_lastr = lbn; 252 n = min(n, size - bp->b_resid); 253 if (error) { 254 brelse(bp); 255 return (error); 256 } 257 error = uiomove(bp->b_un.b_addr + on, (int)n, uio); 258 if (type == IFREG && 259 (n + on == fs->fs_bsize || uio->uio_offset == ip->i_size)) 260 bp->b_flags |= B_AGE; 261 brelse(bp); 262 } while (error == 0 && uio->uio_resid > 0 && n != 0); 263 return (error); 264 } 265 266 /* 267 * Vnode op for writing. 268 */ 269 ffs_write(ap) 270 struct vop_write_args /* { 271 struct vnode *a_vp; 272 struct uio *a_uio; 273 int a_ioflag; 274 struct ucred *a_cred; 275 } */ *ap; 276 { 277 register struct vnode *vp = ap->a_vp; 278 register struct uio *uio = ap->a_uio; 279 register struct inode *ip = VTOI(vp); 280 register struct fs *fs; 281 struct proc *p = uio->uio_procp; 282 int ioflag = ap->a_ioflag; 283 struct timeval tv; 284 struct buf *bp; 285 daddr_t lbn, bn; 286 off_t osize; 287 int n, on, flags; 288 int size, resid, error = 0; 289 290 #ifdef DIAGNOSTIC 291 if (uio->uio_rw != UIO_WRITE) 292 panic("ffs_write mode"); 293 #endif 294 switch (vp->v_type) { 295 case VREG: 296 if (ioflag & IO_APPEND) 297 uio->uio_offset = ip->i_size; 298 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size) 299 return (EPERM); 300 /* fall through */ 301 case VLNK: 302 break; 303 304 case VDIR: 305 if ((ioflag & IO_SYNC) == 0) 306 panic("ffs_write nonsync dir write"); 307 break; 308 309 default: 310 panic("ffs_write type"); 311 } 312 if (uio->uio_resid == 0) 313 return (0); 314 fs = ip->i_fs; 315 if (uio->uio_offset < 0 || 316 (u_quad_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize) 317 return (EFBIG); 318 /* 319 * Maybe this should be above the vnode op call, but so long as 320 * file servers have no limits, i don't think it matters 321 */ 322 if (vp->v_type == VREG && p && 323 uio->uio_offset + uio->uio_resid > 324 p->p_rlimit[RLIMIT_FSIZE].rlim_cur) { 325 psignal(p, SIGXFSZ); 326 return (EFBIG); 327 } 328 resid = uio->uio_resid; 329 osize = ip->i_size; 330 flags = 0; 331 if (ioflag & IO_SYNC) 332 flags = B_SYNC; 333 do { 334 lbn = lblkno(fs, uio->uio_offset); 335 on = blkoff(fs, uio->uio_offset); 336 n = min((unsigned)(fs->fs_bsize - on), uio->uio_resid); 337 if (n < fs->fs_bsize) 338 flags |= B_CLRBUF; 339 else 340 flags &= ~B_CLRBUF; 341 if (error = ffs_balloc(ip, lbn, on + n, ap->a_cred, &bp, flags)) 342 break; 343 bn = bp->b_blkno; 344 if (uio->uio_offset + n > ip->i_size) { 345 ip->i_size = uio->uio_offset + n; 346 vnode_pager_setsize(vp, (u_long)ip->i_size); 347 } 348 size = blksize(fs, ip, lbn); 349 (void) vnode_pager_uncache(vp); 350 n = min(n, size - bp->b_resid); 351 error = uiomove(bp->b_un.b_addr + on, n, uio); 352 if (ioflag & IO_SYNC) 353 (void) bwrite(bp); 354 else if (n + on == fs->fs_bsize) { 355 if (doclusterwrite) { 356 cluster_write(bp, ip->i_size); 357 } else { 358 bp->b_flags |= B_AGE; 359 bawrite(bp); 360 } 361 } else 362 bdwrite(bp); 363 ip->i_flag |= IUPD|ICHG; 364 if (ap->a_cred && ap->a_cred->cr_uid != 0) 365 ip->i_mode &= ~(ISUID|ISGID); 366 } while (error == 0 && uio->uio_resid > 0 && n != 0); 367 if (error && (ioflag & IO_UNIT)) { 368 (void)VOP_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred, 369 uio->uio_procp); 370 uio->uio_offset -= resid - uio->uio_resid; 371 uio->uio_resid = resid; 372 } 373 if (!error && (ioflag & IO_SYNC)) { 374 tv = time; 375 error = VOP_UPDATE(vp, &tv, &tv, 1); 376 } 377 return (error); 378 } 379 380 /* 381 * Synch an open file. 382 */ 383 /* ARGSUSED */ 384 int 385 ffs_fsync(ap) 386 struct vop_fsync_args /* { 387 struct vnode *a_vp; 388 struct ucred *a_cred; 389 int a_waitfor; 390 struct proc *a_p; 391 } */ *ap; 392 { 393 register struct vnode *vp = ap->a_vp; 394 struct inode *ip = VTOI(vp); 395 register struct buf *bp; 396 struct timeval tv; 397 struct buf *nbp; 398 int s; 399 400 /* 401 * Flush all dirty buffers associated with a vnode. 402 */ 403 loop: 404 s = splbio(); 405 for (bp = vp->v_dirtyblkhd.le_next; bp; bp = nbp) { 406 nbp = bp->b_vnbufs.qe_next; 407 if ((bp->b_flags & B_BUSY)) 408 continue; 409 if ((bp->b_flags & B_DELWRI) == 0) 410 panic("ffs_fsync: not dirty"); 411 bremfree(bp); 412 bp->b_flags |= B_BUSY; 413 splx(s); 414 /* 415 * Wait for I/O associated with indirect blocks to complete, 416 * since there is no way to quickly wait for them below. 417 */ 418 if (bp->b_vp == vp || ap->a_waitfor == MNT_NOWAIT) 419 (void) bawrite(bp); 420 else 421 (void) bwrite(bp); 422 goto loop; 423 } 424 if (ap->a_waitfor == MNT_WAIT) { 425 while (vp->v_numoutput) { 426 vp->v_flag |= VBWAIT; 427 sleep((caddr_t)&vp->v_numoutput, PRIBIO + 1); 428 } 429 #ifdef DIAGNOSTIC 430 if (vp->v_dirtyblkhd.le_next) { 431 vprint("ffs_fsync: dirty", vp); 432 goto loop; 433 } 434 #endif 435 } 436 splx(s); 437 tv = time; 438 return (VOP_UPDATE(ap->a_vp, &tv, &tv, ap->a_waitfor == MNT_WAIT)); 439 } 440 441