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 7.101 (Berkeley) 04/18/93 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_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, ufs_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, ufs_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 * Enabling cluster read/write operations. 183 */ 184 #include <sys/sysctl.h> 185 int doclusterread = 1; 186 struct ctldebug debug11 = { "doclusterread", &doclusterread }; 187 int doclusterwrite = 1; 188 struct ctldebug debug12 = { "doclusterwrite", &doclusterwrite }; 189 190 /* 191 * Vnode op for reading. 192 */ 193 /* ARGSUSED */ 194 ffs_read(ap) 195 struct vop_read_args /* { 196 struct vnode *a_vp; 197 struct uio *a_uio; 198 int a_ioflag; 199 struct ucred *a_cred; 200 } */ *ap; 201 { 202 register struct vnode *vp = ap->a_vp; 203 register struct inode *ip = VTOI(vp); 204 register struct uio *uio = ap->a_uio; 205 register struct fs *fs; 206 struct buf *bp; 207 daddr_t lbn, bn, rablock; 208 off_t diff; 209 int type, rasize, error = 0; 210 long size, n, on; 211 212 type = ip->i_mode & IFMT; 213 #ifdef DIAGNOSTIC 214 if (uio->uio_rw != UIO_READ) 215 panic("ffs_read mode"); 216 if (type != IFDIR && type != IFREG && type != IFLNK) 217 panic("ffs_read type"); 218 if (type == IFLNK && (int)ip->i_size < vp->v_mount->mnt_maxsymlinklen) 219 panic("read short symlink"); 220 #endif 221 if (uio->uio_resid == 0) 222 return (0); 223 fs = ip->i_fs; 224 if ((u_quad_t)uio->uio_offset > fs->fs_maxfilesize) 225 return (EFBIG); 226 ip->i_flag |= IACC; 227 do { 228 lbn = lblkno(fs, uio->uio_offset); 229 on = blkoff(fs, uio->uio_offset); 230 n = min((unsigned)(fs->fs_bsize - on), uio->uio_resid); 231 diff = ip->i_size - uio->uio_offset; 232 if (diff <= 0) 233 return (0); 234 if (diff < n) 235 n = diff; 236 size = blksize(fs, ip, lbn); 237 rablock = lbn + 1; 238 if (doclusterread && lblktosize(fs, rablock) <= ip->i_size) { 239 error = cluster_read(vp, ip->i_size, lbn, size, 240 NOCRED, &bp); 241 } else if (vp->v_lastr + 1 == lbn && 242 lblktosize(fs, rablock) < ip->i_size) { 243 rasize = blksize(fs, ip, rablock); 244 error = breadn(vp, lbn, size, &rablock, 245 &rasize, 1, NOCRED, &bp); 246 } else 247 error = bread(vp, lbn, size, NOCRED, &bp); 248 vp->v_lastr = lbn; 249 n = min(n, size - bp->b_resid); 250 if (error) { 251 brelse(bp); 252 return (error); 253 } 254 error = uiomove(bp->b_un.b_addr + on, (int)n, uio); 255 if (type == IFREG && 256 (n + on == fs->fs_bsize || uio->uio_offset == ip->i_size)) 257 bp->b_flags |= B_AGE; 258 brelse(bp); 259 } while (error == 0 && uio->uio_resid > 0 && n != 0); 260 return (error); 261 } 262 263 /* 264 * Vnode op for writing. 265 */ 266 ffs_write(ap) 267 struct vop_write_args /* { 268 struct vnode *a_vp; 269 struct uio *a_uio; 270 int a_ioflag; 271 struct ucred *a_cred; 272 } */ *ap; 273 { 274 register struct vnode *vp = ap->a_vp; 275 register struct uio *uio = ap->a_uio; 276 register struct inode *ip = VTOI(vp); 277 register struct fs *fs; 278 struct proc *p = uio->uio_procp; 279 int ioflag = ap->a_ioflag; 280 struct timeval tv; 281 struct buf *bp; 282 daddr_t lbn, bn; 283 off_t osize; 284 int n, on, flags; 285 int size, resid, error = 0; 286 287 #ifdef DIAGNOSTIC 288 if (uio->uio_rw != UIO_WRITE) 289 panic("ffs_write mode"); 290 #endif 291 switch (vp->v_type) { 292 case VREG: 293 if (ioflag & IO_APPEND) 294 uio->uio_offset = ip->i_size; 295 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size) 296 return (EPERM); 297 /* fall through */ 298 case VLNK: 299 break; 300 301 case VDIR: 302 if ((ioflag & IO_SYNC) == 0) 303 panic("ffs_write nonsync dir write"); 304 break; 305 306 default: 307 panic("ffs_write type"); 308 } 309 if (uio->uio_resid == 0) 310 return (0); 311 fs = ip->i_fs; 312 if (uio->uio_offset < 0 || 313 (u_quad_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize) 314 return (EFBIG); 315 /* 316 * Maybe this should be above the vnode op call, but so long as 317 * file servers have no limits, i don't think it matters 318 */ 319 if (vp->v_type == VREG && p && 320 uio->uio_offset + uio->uio_resid > 321 p->p_rlimit[RLIMIT_FSIZE].rlim_cur) { 322 psignal(p, SIGXFSZ); 323 return (EFBIG); 324 } 325 resid = uio->uio_resid; 326 osize = ip->i_size; 327 flags = 0; 328 if (ioflag & IO_SYNC) 329 flags = B_SYNC; 330 do { 331 lbn = lblkno(fs, uio->uio_offset); 332 on = blkoff(fs, uio->uio_offset); 333 n = min((unsigned)(fs->fs_bsize - on), uio->uio_resid); 334 if (n < fs->fs_bsize) 335 flags |= B_CLRBUF; 336 else 337 flags &= ~B_CLRBUF; 338 if (error = ffs_balloc(ip, lbn, on + n, ap->a_cred, &bp, flags)) 339 break; 340 bn = bp->b_blkno; 341 if (uio->uio_offset + n > ip->i_size) { 342 ip->i_size = uio->uio_offset + n; 343 vnode_pager_setsize(vp, (u_long)ip->i_size); 344 } 345 size = blksize(fs, ip, lbn); 346 (void) vnode_pager_uncache(vp); 347 n = min(n, size - bp->b_resid); 348 error = uiomove(bp->b_un.b_addr + on, n, uio); 349 if (ioflag & IO_SYNC) 350 (void) bwrite(bp); 351 else if (n + on == fs->fs_bsize) { 352 if (doclusterwrite) { 353 cluster_write(bp, ip->i_size); 354 } else { 355 bp->b_flags |= B_AGE; 356 bawrite(bp); 357 } 358 } else 359 bdwrite(bp); 360 ip->i_flag |= IUPD|ICHG; 361 if (ap->a_cred && ap->a_cred->cr_uid != 0) 362 ip->i_mode &= ~(ISUID|ISGID); 363 } while (error == 0 && uio->uio_resid > 0 && n != 0); 364 if (error && (ioflag & IO_UNIT)) { 365 (void)VOP_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred, 366 uio->uio_procp); 367 uio->uio_offset -= resid - uio->uio_resid; 368 uio->uio_resid = resid; 369 } 370 if (!error && (ioflag & IO_SYNC)) { 371 tv = time; 372 error = VOP_UPDATE(vp, &tv, &tv, 1); 373 } 374 return (error); 375 } 376 377 /* 378 * Synch an open file. 379 */ 380 /* ARGSUSED */ 381 int 382 ffs_fsync(ap) 383 struct vop_fsync_args /* { 384 struct vnode *a_vp; 385 struct ucred *a_cred; 386 int a_waitfor; 387 struct proc *a_p; 388 } */ *ap; 389 { 390 register struct vnode *vp = ap->a_vp; 391 struct inode *ip = VTOI(vp); 392 register struct buf *bp; 393 struct timeval tv; 394 struct buf *nbp; 395 int s; 396 397 /* 398 * Flush all dirty buffers associated with a vnode. 399 */ 400 loop: 401 s = splbio(); 402 for (bp = vp->v_dirtyblkhd.le_next; bp; bp = nbp) { 403 nbp = bp->b_vnbufs.qe_next; 404 if ((bp->b_flags & B_BUSY)) 405 continue; 406 if ((bp->b_flags & B_DELWRI) == 0) 407 panic("ffs_fsync: not dirty"); 408 bremfree(bp); 409 bp->b_flags |= B_BUSY; 410 splx(s); 411 /* 412 * Wait for I/O associated with indirect blocks to complete, 413 * since there is no way to quickly wait for them below. 414 */ 415 if (bp->b_vp == vp || ap->a_waitfor == MNT_NOWAIT) 416 (void) bawrite(bp); 417 else 418 (void) bwrite(bp); 419 goto loop; 420 } 421 if (ap->a_waitfor == MNT_WAIT) { 422 while (vp->v_numoutput) { 423 vp->v_flag |= VBWAIT; 424 sleep((caddr_t)&vp->v_numoutput, PRIBIO + 1); 425 } 426 #ifdef DIAGNOSTIC 427 if (vp->v_dirtyblkhd.le_next) { 428 vprint("ffs_fsync: dirty", vp); 429 goto loop; 430 } 431 #endif 432 } 433 splx(s); 434 tv = time; 435 return (VOP_UPDATE(ap->a_vp, &tv, &tv, ap->a_waitfor == MNT_WAIT)); 436 } 437 438