lfs_vnops.c revision 1.169
1 /* $NetBSD: lfs_vnops.c,v 1.169 2006/04/18 21:41:20 perseant Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Konrad E. Schroder <perseant (at) hhhh.org>. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 /* 39 * Copyright (c) 1986, 1989, 1991, 1993, 1995 40 * The Regents of the University of California. All rights reserved. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)lfs_vnops.c 8.13 (Berkeley) 6/10/95 67 */ 68 69 #include <sys/cdefs.h> 70 __KERNEL_RCSID(0, "$NetBSD: lfs_vnops.c,v 1.169 2006/04/18 21:41:20 perseant Exp $"); 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/namei.h> 75 #include <sys/resourcevar.h> 76 #include <sys/kernel.h> 77 #include <sys/file.h> 78 #include <sys/stat.h> 79 #include <sys/buf.h> 80 #include <sys/proc.h> 81 #include <sys/mount.h> 82 #include <sys/vnode.h> 83 #include <sys/pool.h> 84 #include <sys/signalvar.h> 85 86 #include <miscfs/fifofs/fifo.h> 87 #include <miscfs/genfs/genfs.h> 88 #include <miscfs/specfs/specdev.h> 89 90 #include <ufs/ufs/inode.h> 91 #include <ufs/ufs/dir.h> 92 #include <ufs/ufs/ufsmount.h> 93 #include <ufs/ufs/ufs_extern.h> 94 95 #include <uvm/uvm.h> 96 #include <uvm/uvm_pmap.h> 97 #include <uvm/uvm_stat.h> 98 #include <uvm/uvm_pager.h> 99 100 #include <ufs/lfs/lfs.h> 101 #include <ufs/lfs/lfs_extern.h> 102 103 extern pid_t lfs_writer_daemon; 104 105 /* Global vfs data structures for lfs. */ 106 int (**lfs_vnodeop_p)(void *); 107 const struct vnodeopv_entry_desc lfs_vnodeop_entries[] = { 108 { &vop_default_desc, vn_default_error }, 109 { &vop_lookup_desc, ufs_lookup }, /* lookup */ 110 { &vop_create_desc, lfs_create }, /* create */ 111 { &vop_whiteout_desc, ufs_whiteout }, /* whiteout */ 112 { &vop_mknod_desc, lfs_mknod }, /* mknod */ 113 { &vop_open_desc, ufs_open }, /* open */ 114 { &vop_close_desc, lfs_close }, /* close */ 115 { &vop_access_desc, ufs_access }, /* access */ 116 { &vop_getattr_desc, lfs_getattr }, /* getattr */ 117 { &vop_setattr_desc, lfs_setattr }, /* setattr */ 118 { &vop_read_desc, lfs_read }, /* read */ 119 { &vop_write_desc, lfs_write }, /* write */ 120 { &vop_lease_desc, ufs_lease_check }, /* lease */ 121 { &vop_ioctl_desc, ufs_ioctl }, /* ioctl */ 122 { &vop_fcntl_desc, lfs_fcntl }, /* fcntl */ 123 { &vop_poll_desc, ufs_poll }, /* poll */ 124 { &vop_kqfilter_desc, genfs_kqfilter }, /* kqfilter */ 125 { &vop_revoke_desc, ufs_revoke }, /* revoke */ 126 { &vop_mmap_desc, lfs_mmap }, /* mmap */ 127 { &vop_fsync_desc, lfs_fsync }, /* fsync */ 128 { &vop_seek_desc, ufs_seek }, /* seek */ 129 { &vop_remove_desc, lfs_remove }, /* remove */ 130 { &vop_link_desc, lfs_link }, /* link */ 131 { &vop_rename_desc, lfs_rename }, /* rename */ 132 { &vop_mkdir_desc, lfs_mkdir }, /* mkdir */ 133 { &vop_rmdir_desc, lfs_rmdir }, /* rmdir */ 134 { &vop_symlink_desc, lfs_symlink }, /* symlink */ 135 { &vop_readdir_desc, ufs_readdir }, /* readdir */ 136 { &vop_readlink_desc, ufs_readlink }, /* readlink */ 137 { &vop_abortop_desc, ufs_abortop }, /* abortop */ 138 { &vop_inactive_desc, lfs_inactive }, /* inactive */ 139 { &vop_reclaim_desc, lfs_reclaim }, /* reclaim */ 140 { &vop_lock_desc, ufs_lock }, /* lock */ 141 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 142 { &vop_bmap_desc, ufs_bmap }, /* bmap */ 143 { &vop_strategy_desc, lfs_strategy }, /* strategy */ 144 { &vop_print_desc, ufs_print }, /* print */ 145 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 146 { &vop_pathconf_desc, ufs_pathconf }, /* pathconf */ 147 { &vop_advlock_desc, ufs_advlock }, /* advlock */ 148 { &vop_bwrite_desc, lfs_bwrite }, /* bwrite */ 149 { &vop_getpages_desc, lfs_getpages }, /* getpages */ 150 { &vop_putpages_desc, lfs_putpages }, /* putpages */ 151 { NULL, NULL } 152 }; 153 const struct vnodeopv_desc lfs_vnodeop_opv_desc = 154 { &lfs_vnodeop_p, lfs_vnodeop_entries }; 155 156 int (**lfs_specop_p)(void *); 157 const struct vnodeopv_entry_desc lfs_specop_entries[] = { 158 { &vop_default_desc, vn_default_error }, 159 { &vop_lookup_desc, spec_lookup }, /* lookup */ 160 { &vop_create_desc, spec_create }, /* create */ 161 { &vop_mknod_desc, spec_mknod }, /* mknod */ 162 { &vop_open_desc, spec_open }, /* open */ 163 { &vop_close_desc, lfsspec_close }, /* close */ 164 { &vop_access_desc, ufs_access }, /* access */ 165 { &vop_getattr_desc, lfs_getattr }, /* getattr */ 166 { &vop_setattr_desc, lfs_setattr }, /* setattr */ 167 { &vop_read_desc, ufsspec_read }, /* read */ 168 { &vop_write_desc, ufsspec_write }, /* write */ 169 { &vop_lease_desc, spec_lease_check }, /* lease */ 170 { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ 171 { &vop_fcntl_desc, ufs_fcntl }, /* fcntl */ 172 { &vop_poll_desc, spec_poll }, /* poll */ 173 { &vop_kqfilter_desc, spec_kqfilter }, /* kqfilter */ 174 { &vop_revoke_desc, spec_revoke }, /* revoke */ 175 { &vop_mmap_desc, spec_mmap }, /* mmap */ 176 { &vop_fsync_desc, spec_fsync }, /* fsync */ 177 { &vop_seek_desc, spec_seek }, /* seek */ 178 { &vop_remove_desc, spec_remove }, /* remove */ 179 { &vop_link_desc, spec_link }, /* link */ 180 { &vop_rename_desc, spec_rename }, /* rename */ 181 { &vop_mkdir_desc, spec_mkdir }, /* mkdir */ 182 { &vop_rmdir_desc, spec_rmdir }, /* rmdir */ 183 { &vop_symlink_desc, spec_symlink }, /* symlink */ 184 { &vop_readdir_desc, spec_readdir }, /* readdir */ 185 { &vop_readlink_desc, spec_readlink }, /* readlink */ 186 { &vop_abortop_desc, spec_abortop }, /* abortop */ 187 { &vop_inactive_desc, lfs_inactive }, /* inactive */ 188 { &vop_reclaim_desc, lfs_reclaim }, /* reclaim */ 189 { &vop_lock_desc, ufs_lock }, /* lock */ 190 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 191 { &vop_bmap_desc, spec_bmap }, /* bmap */ 192 { &vop_strategy_desc, spec_strategy }, /* strategy */ 193 { &vop_print_desc, ufs_print }, /* print */ 194 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 195 { &vop_pathconf_desc, spec_pathconf }, /* pathconf */ 196 { &vop_advlock_desc, spec_advlock }, /* advlock */ 197 { &vop_bwrite_desc, vn_bwrite }, /* bwrite */ 198 { &vop_getpages_desc, spec_getpages }, /* getpages */ 199 { &vop_putpages_desc, spec_putpages }, /* putpages */ 200 { NULL, NULL } 201 }; 202 const struct vnodeopv_desc lfs_specop_opv_desc = 203 { &lfs_specop_p, lfs_specop_entries }; 204 205 int (**lfs_fifoop_p)(void *); 206 const struct vnodeopv_entry_desc lfs_fifoop_entries[] = { 207 { &vop_default_desc, vn_default_error }, 208 { &vop_lookup_desc, fifo_lookup }, /* lookup */ 209 { &vop_create_desc, fifo_create }, /* create */ 210 { &vop_mknod_desc, fifo_mknod }, /* mknod */ 211 { &vop_open_desc, fifo_open }, /* open */ 212 { &vop_close_desc, lfsfifo_close }, /* close */ 213 { &vop_access_desc, ufs_access }, /* access */ 214 { &vop_getattr_desc, lfs_getattr }, /* getattr */ 215 { &vop_setattr_desc, lfs_setattr }, /* setattr */ 216 { &vop_read_desc, ufsfifo_read }, /* read */ 217 { &vop_write_desc, ufsfifo_write }, /* write */ 218 { &vop_lease_desc, fifo_lease_check }, /* lease */ 219 { &vop_ioctl_desc, fifo_ioctl }, /* ioctl */ 220 { &vop_fcntl_desc, ufs_fcntl }, /* fcntl */ 221 { &vop_poll_desc, fifo_poll }, /* poll */ 222 { &vop_kqfilter_desc, fifo_kqfilter }, /* kqfilter */ 223 { &vop_revoke_desc, fifo_revoke }, /* revoke */ 224 { &vop_mmap_desc, fifo_mmap }, /* mmap */ 225 { &vop_fsync_desc, fifo_fsync }, /* fsync */ 226 { &vop_seek_desc, fifo_seek }, /* seek */ 227 { &vop_remove_desc, fifo_remove }, /* remove */ 228 { &vop_link_desc, fifo_link }, /* link */ 229 { &vop_rename_desc, fifo_rename }, /* rename */ 230 { &vop_mkdir_desc, fifo_mkdir }, /* mkdir */ 231 { &vop_rmdir_desc, fifo_rmdir }, /* rmdir */ 232 { &vop_symlink_desc, fifo_symlink }, /* symlink */ 233 { &vop_readdir_desc, fifo_readdir }, /* readdir */ 234 { &vop_readlink_desc, fifo_readlink }, /* readlink */ 235 { &vop_abortop_desc, fifo_abortop }, /* abortop */ 236 { &vop_inactive_desc, lfs_inactive }, /* inactive */ 237 { &vop_reclaim_desc, lfs_reclaim }, /* reclaim */ 238 { &vop_lock_desc, ufs_lock }, /* lock */ 239 { &vop_unlock_desc, ufs_unlock }, /* unlock */ 240 { &vop_bmap_desc, fifo_bmap }, /* bmap */ 241 { &vop_strategy_desc, fifo_strategy }, /* strategy */ 242 { &vop_print_desc, ufs_print }, /* print */ 243 { &vop_islocked_desc, ufs_islocked }, /* islocked */ 244 { &vop_pathconf_desc, fifo_pathconf }, /* pathconf */ 245 { &vop_advlock_desc, fifo_advlock }, /* advlock */ 246 { &vop_bwrite_desc, lfs_bwrite }, /* bwrite */ 247 { &vop_putpages_desc, fifo_putpages }, /* putpages */ 248 { NULL, NULL } 249 }; 250 const struct vnodeopv_desc lfs_fifoop_opv_desc = 251 { &lfs_fifoop_p, lfs_fifoop_entries }; 252 253 static int check_dirty(struct lfs *, struct vnode *, off_t, off_t, off_t, int, int); 254 255 #define LFS_READWRITE 256 #include <ufs/ufs/ufs_readwrite.c> 257 #undef LFS_READWRITE 258 259 /* 260 * Synch an open file. 261 */ 262 /* ARGSUSED */ 263 int 264 lfs_fsync(void *v) 265 { 266 struct vop_fsync_args /* { 267 struct vnode *a_vp; 268 struct ucred *a_cred; 269 int a_flags; 270 off_t offlo; 271 off_t offhi; 272 struct lwp *a_l; 273 } */ *ap = v; 274 struct vnode *vp = ap->a_vp; 275 int error, wait; 276 277 /* If we're mounted read-only, don't try to sync. */ 278 if (VTOI(vp)->i_lfs->lfs_ronly) 279 return 0; 280 281 /* 282 * Trickle sync checks for need to do a checkpoint after possible 283 * activity from the pagedaemon. 284 */ 285 if (ap->a_flags & FSYNC_LAZY) { 286 simple_lock(&lfs_subsys_lock); 287 wakeup(&lfs_writer_daemon); 288 simple_unlock(&lfs_subsys_lock); 289 return 0; 290 } 291 292 wait = (ap->a_flags & FSYNC_WAIT); 293 simple_lock(&vp->v_interlock); 294 error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo), 295 round_page(ap->a_offhi), 296 PGO_CLEANIT | (wait ? PGO_SYNCIO : 0)); 297 if (error) 298 return error; 299 error = lfs_update(vp, NULL, NULL, wait ? UPDATE_WAIT : 0); 300 if (error == 0 && ap->a_flags & FSYNC_CACHE) { 301 int l = 0; 302 error = VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &l, FWRITE, 303 ap->a_l->l_proc->p_ucred, ap->a_l); 304 } 305 if (wait && !VPISEMPTY(vp)) 306 LFS_SET_UINO(VTOI(vp), IN_MODIFIED); 307 308 return error; 309 } 310 311 /* 312 * Take IN_ADIROP off, then call ufs_inactive. 313 */ 314 int 315 lfs_inactive(void *v) 316 { 317 struct vop_inactive_args /* { 318 struct vnode *a_vp; 319 struct lwp *a_l; 320 } */ *ap = v; 321 322 KASSERT(VTOI(ap->a_vp)->i_nlink == VTOI(ap->a_vp)->i_ffs_effnlink); 323 324 lfs_unmark_vnode(ap->a_vp); 325 326 /* 327 * The Ifile is only ever inactivated on unmount. 328 * Streamline this process by not giving it more dirty blocks. 329 */ 330 if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM) { 331 LFS_CLR_UINO(VTOI(ap->a_vp), IN_ALLMOD); 332 VOP_UNLOCK(ap->a_vp, 0); 333 return 0; 334 } 335 336 return ufs_inactive(v); 337 } 338 339 /* 340 * These macros are used to bracket UFS directory ops, so that we can 341 * identify all the pages touched during directory ops which need to 342 * be ordered and flushed atomically, so that they may be recovered. 343 * 344 * Because we have to mark nodes VDIROP in order to prevent 345 * the cache from reclaiming them while a dirop is in progress, we must 346 * also manage the number of nodes so marked (otherwise we can run out). 347 * We do this by setting lfs_dirvcount to the number of marked vnodes; it 348 * is decremented during segment write, when VDIROP is taken off. 349 */ 350 #define MARK_VNODE(vp) lfs_mark_vnode(vp) 351 #define UNMARK_VNODE(vp) lfs_unmark_vnode(vp) 352 #define SET_DIROP_CREATE(dvp, vpp) lfs_set_dirop_create((dvp), (vpp)) 353 #define SET_DIROP_REMOVE(dvp, vp) lfs_set_dirop((dvp), (vp)) 354 static int lfs_set_dirop_create(struct vnode *, struct vnode **); 355 static int lfs_set_dirop(struct vnode *, struct vnode *); 356 357 static int 358 lfs_set_dirop(struct vnode *dvp, struct vnode *vp) 359 { 360 struct lfs *fs; 361 int error; 362 363 KASSERT(VOP_ISLOCKED(dvp)); 364 KASSERT(vp == NULL || VOP_ISLOCKED(vp)); 365 366 fs = VTOI(dvp)->i_lfs; 367 368 ASSERT_NO_SEGLOCK(fs); 369 /* 370 * LFS_NRESERVE calculates direct and indirect blocks as well 371 * as an inode block; an overestimate in most cases. 372 */ 373 if ((error = lfs_reserve(fs, dvp, vp, LFS_NRESERVE(fs))) != 0) 374 return (error); 375 376 restart: 377 simple_lock(&fs->lfs_interlock); 378 if (fs->lfs_dirops == 0) { 379 simple_unlock(&fs->lfs_interlock); 380 lfs_check(dvp, LFS_UNUSED_LBN, 0); 381 simple_lock(&fs->lfs_interlock); 382 } 383 while (fs->lfs_writer) 384 ltsleep(&fs->lfs_dirops, (PRIBIO + 1), "lfs_sdirop", 0, 385 &fs->lfs_interlock); 386 simple_lock(&lfs_subsys_lock); 387 if (lfs_dirvcount > LFS_MAX_DIROP && fs->lfs_dirops == 0) { 388 wakeup(&lfs_writer_daemon); 389 simple_unlock(&lfs_subsys_lock); 390 simple_unlock(&fs->lfs_interlock); 391 preempt(1); 392 goto restart; 393 } 394 395 if (lfs_dirvcount > LFS_MAX_DIROP) { 396 simple_unlock(&fs->lfs_interlock); 397 DLOG((DLOG_DIROP, "lfs_set_dirop: sleeping with dirops=%d, " 398 "dirvcount=%d\n", fs->lfs_dirops, lfs_dirvcount)); 399 if ((error = ltsleep(&lfs_dirvcount, 400 PCATCH | PUSER | PNORELOCK, "lfs_maxdirop", 0, 401 &lfs_subsys_lock)) != 0) { 402 goto unreserve; 403 } 404 goto restart; 405 } 406 simple_unlock(&lfs_subsys_lock); 407 408 ++fs->lfs_dirops; 409 fs->lfs_doifile = 1; 410 simple_unlock(&fs->lfs_interlock); 411 412 /* Hold a reference so SET_ENDOP will be happy */ 413 vref(dvp); 414 if (vp) { 415 vref(vp); 416 MARK_VNODE(vp); 417 } 418 419 MARK_VNODE(dvp); 420 return 0; 421 422 unreserve: 423 lfs_reserve(fs, dvp, vp, -LFS_NRESERVE(fs)); 424 return error; 425 } 426 427 /* 428 * Get a new vnode *before* adjusting the dirop count, to avoid a deadlock 429 * in getnewvnode(), if we have a stacked filesystem mounted on top 430 * of us. 431 * 432 * NB: this means we have to clear the new vnodes on error. Fortunately 433 * SET_ENDOP is there to do that for us. 434 */ 435 static int 436 lfs_set_dirop_create(struct vnode *dvp, struct vnode **vpp) 437 { 438 int error; 439 struct lfs *fs; 440 441 fs = VFSTOUFS(dvp->v_mount)->um_lfs; 442 ASSERT_NO_SEGLOCK(fs); 443 if (fs->lfs_ronly) 444 return EROFS; 445 if (vpp && (error = getnewvnode(VT_LFS, dvp->v_mount, lfs_vnodeop_p, vpp))) { 446 DLOG((DLOG_ALLOC, "lfs_set_dirop_create: dvp %p error %d\n", 447 dvp, error)); 448 return error; 449 } 450 if ((error = lfs_set_dirop(dvp, NULL)) != 0) { 451 if (vpp) { 452 ungetnewvnode(*vpp); 453 *vpp = NULL; 454 } 455 return error; 456 } 457 return 0; 458 } 459 460 #define SET_ENDOP_BASE(fs, dvp, str) \ 461 do { \ 462 simple_lock(&(fs)->lfs_interlock); \ 463 --(fs)->lfs_dirops; \ 464 if (!(fs)->lfs_dirops) { \ 465 if ((fs)->lfs_nadirop) { \ 466 panic("SET_ENDOP: %s: no dirops but " \ 467 " nadirop=%d", (str), \ 468 (fs)->lfs_nadirop); \ 469 } \ 470 wakeup(&(fs)->lfs_writer); \ 471 simple_unlock(&(fs)->lfs_interlock); \ 472 lfs_check((dvp), LFS_UNUSED_LBN, 0); \ 473 } else \ 474 simple_unlock(&(fs)->lfs_interlock); \ 475 } while(0) 476 #define SET_ENDOP_CREATE(fs, dvp, nvpp, str) \ 477 do { \ 478 UNMARK_VNODE(dvp); \ 479 if (nvpp && *nvpp) \ 480 UNMARK_VNODE(*nvpp); \ 481 /* Check for error return to stem vnode leakage */ \ 482 if (nvpp && *nvpp && !((*nvpp)->v_flag & VDIROP)) \ 483 ungetnewvnode(*(nvpp)); \ 484 SET_ENDOP_BASE((fs), (dvp), (str)); \ 485 lfs_reserve((fs), (dvp), NULL, -LFS_NRESERVE(fs)); \ 486 vrele(dvp); \ 487 } while(0) 488 #define SET_ENDOP_CREATE_AP(ap, str) \ 489 SET_ENDOP_CREATE(VTOI((ap)->a_dvp)->i_lfs, (ap)->a_dvp, \ 490 (ap)->a_vpp, (str)) 491 #define SET_ENDOP_REMOVE(fs, dvp, ovp, str) \ 492 do { \ 493 UNMARK_VNODE(dvp); \ 494 if (ovp) \ 495 UNMARK_VNODE(ovp); \ 496 SET_ENDOP_BASE((fs), (dvp), (str)); \ 497 lfs_reserve((fs), (dvp), (ovp), -LFS_NRESERVE(fs)); \ 498 vrele(dvp); \ 499 if (ovp) \ 500 vrele(ovp); \ 501 } while(0) 502 503 void 504 lfs_mark_vnode(struct vnode *vp) 505 { 506 struct inode *ip = VTOI(vp); 507 struct lfs *fs = ip->i_lfs; 508 509 simple_lock(&fs->lfs_interlock); 510 if (!(ip->i_flag & IN_ADIROP)) { 511 if (!(vp->v_flag & VDIROP)) { 512 (void)lfs_vref(vp); 513 simple_lock(&lfs_subsys_lock); 514 ++lfs_dirvcount; 515 simple_unlock(&lfs_subsys_lock); 516 TAILQ_INSERT_TAIL(&fs->lfs_dchainhd, ip, i_lfs_dchain); 517 vp->v_flag |= VDIROP; 518 } 519 ++fs->lfs_nadirop; 520 ip->i_flag |= IN_ADIROP; 521 } else 522 KASSERT(vp->v_flag & VDIROP); 523 simple_unlock(&fs->lfs_interlock); 524 } 525 526 void 527 lfs_unmark_vnode(struct vnode *vp) 528 { 529 struct inode *ip = VTOI(vp); 530 531 if (ip && (ip->i_flag & IN_ADIROP)) { 532 KASSERT(vp->v_flag & VDIROP); 533 simple_lock(&ip->i_lfs->lfs_interlock); 534 --ip->i_lfs->lfs_nadirop; 535 simple_unlock(&ip->i_lfs->lfs_interlock); 536 ip->i_flag &= ~IN_ADIROP; 537 } 538 } 539 540 int 541 lfs_symlink(void *v) 542 { 543 struct vop_symlink_args /* { 544 struct vnode *a_dvp; 545 struct vnode **a_vpp; 546 struct componentname *a_cnp; 547 struct vattr *a_vap; 548 char *a_target; 549 } */ *ap = v; 550 int error; 551 552 if ((error = SET_DIROP_CREATE(ap->a_dvp, ap->a_vpp)) != 0) { 553 vput(ap->a_dvp); 554 return error; 555 } 556 error = ufs_symlink(ap); 557 SET_ENDOP_CREATE_AP(ap, "symlink"); 558 return (error); 559 } 560 561 int 562 lfs_mknod(void *v) 563 { 564 struct vop_mknod_args /* { 565 struct vnode *a_dvp; 566 struct vnode **a_vpp; 567 struct componentname *a_cnp; 568 struct vattr *a_vap; 569 } */ *ap = v; 570 struct vattr *vap = ap->a_vap; 571 struct vnode **vpp = ap->a_vpp; 572 struct inode *ip; 573 int error; 574 struct mount *mp; 575 ino_t ino; 576 577 if ((error = SET_DIROP_CREATE(ap->a_dvp, ap->a_vpp)) != 0) { 578 vput(ap->a_dvp); 579 return error; 580 } 581 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode), 582 ap->a_dvp, vpp, ap->a_cnp); 583 584 /* Either way we're done with the dirop at this point */ 585 SET_ENDOP_CREATE_AP(ap, "mknod"); 586 587 if (error) 588 return (error); 589 590 ip = VTOI(*vpp); 591 mp = (*vpp)->v_mount; 592 ino = ip->i_number; 593 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 594 if (vap->va_rdev != VNOVAL) { 595 /* 596 * Want to be able to use this to make badblock 597 * inodes, so don't truncate the dev number. 598 */ 599 #if 0 600 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev, 601 UFS_MPNEEDSWAP((*vpp)->v_mount)); 602 #else 603 ip->i_ffs1_rdev = vap->va_rdev; 604 #endif 605 } 606 607 /* 608 * Call fsync to write the vnode so that we don't have to deal with 609 * flushing it when it's marked VDIROP|VXLOCK. 610 * 611 * XXX KS - If we can't flush we also can't call vgone(), so must 612 * return. But, that leaves this vnode in limbo, also not good. 613 * Can this ever happen (barring hardware failure)? 614 */ 615 if ((error = VOP_FSYNC(*vpp, NOCRED, FSYNC_WAIT, 0, 0, 616 curlwp)) != 0) { 617 panic("lfs_mknod: couldn't fsync (ino %llu)", 618 (unsigned long long)ino); 619 /* return (error); */ 620 } 621 /* 622 * Remove vnode so that it will be reloaded by VFS_VGET and 623 * checked to see if it is an alias of an existing entry in 624 * the inode cache. 625 */ 626 /* Used to be vput, but that causes us to call VOP_INACTIVE twice. */ 627 628 VOP_UNLOCK(*vpp, 0); 629 lfs_vunref(*vpp); 630 (*vpp)->v_type = VNON; 631 vgone(*vpp); 632 error = VFS_VGET(mp, ino, vpp); 633 634 if (error != 0) { 635 *vpp = NULL; 636 return (error); 637 } 638 return (0); 639 } 640 641 int 642 lfs_create(void *v) 643 { 644 struct vop_create_args /* { 645 struct vnode *a_dvp; 646 struct vnode **a_vpp; 647 struct componentname *a_cnp; 648 struct vattr *a_vap; 649 } */ *ap = v; 650 int error; 651 652 if ((error = SET_DIROP_CREATE(ap->a_dvp, ap->a_vpp)) != 0) { 653 vput(ap->a_dvp); 654 return error; 655 } 656 error = ufs_create(ap); 657 SET_ENDOP_CREATE_AP(ap, "create"); 658 return (error); 659 } 660 661 int 662 lfs_mkdir(void *v) 663 { 664 struct vop_mkdir_args /* { 665 struct vnode *a_dvp; 666 struct vnode **a_vpp; 667 struct componentname *a_cnp; 668 struct vattr *a_vap; 669 } */ *ap = v; 670 int error; 671 672 if ((error = SET_DIROP_CREATE(ap->a_dvp, ap->a_vpp)) != 0) { 673 vput(ap->a_dvp); 674 return error; 675 } 676 error = ufs_mkdir(ap); 677 SET_ENDOP_CREATE_AP(ap, "mkdir"); 678 return (error); 679 } 680 681 int 682 lfs_remove(void *v) 683 { 684 struct vop_remove_args /* { 685 struct vnode *a_dvp; 686 struct vnode *a_vp; 687 struct componentname *a_cnp; 688 } */ *ap = v; 689 struct vnode *dvp, *vp; 690 int error; 691 692 dvp = ap->a_dvp; 693 vp = ap->a_vp; 694 if ((error = SET_DIROP_REMOVE(dvp, vp)) != 0) { 695 if (dvp == vp) 696 vrele(vp); 697 else 698 vput(vp); 699 vput(dvp); 700 return error; 701 } 702 error = ufs_remove(ap); 703 SET_ENDOP_REMOVE(VTOI(dvp)->i_lfs, dvp, vp, "remove"); 704 return (error); 705 } 706 707 int 708 lfs_rmdir(void *v) 709 { 710 struct vop_rmdir_args /* { 711 struct vnodeop_desc *a_desc; 712 struct vnode *a_dvp; 713 struct vnode *a_vp; 714 struct componentname *a_cnp; 715 } */ *ap = v; 716 struct vnode *vp; 717 int error; 718 719 vp = ap->a_vp; 720 if ((error = SET_DIROP_REMOVE(ap->a_dvp, ap->a_vp)) != 0) { 721 vrele(ap->a_dvp); 722 if (ap->a_vp != ap->a_dvp) 723 VOP_UNLOCK(ap->a_dvp, 0); 724 vput(vp); 725 return error; 726 } 727 error = ufs_rmdir(ap); 728 SET_ENDOP_REMOVE(VTOI(ap->a_dvp)->i_lfs, ap->a_dvp, vp, "rmdir"); 729 return (error); 730 } 731 732 int 733 lfs_link(void *v) 734 { 735 struct vop_link_args /* { 736 struct vnode *a_dvp; 737 struct vnode *a_vp; 738 struct componentname *a_cnp; 739 } */ *ap = v; 740 int error; 741 struct vnode **vpp = NULL; 742 743 if ((error = SET_DIROP_CREATE(ap->a_dvp, vpp)) != 0) { 744 vput(ap->a_dvp); 745 return error; 746 } 747 error = ufs_link(ap); 748 SET_ENDOP_CREATE(VTOI(ap->a_dvp)->i_lfs, ap->a_dvp, vpp, "link"); 749 return (error); 750 } 751 752 int 753 lfs_rename(void *v) 754 { 755 struct vop_rename_args /* { 756 struct vnode *a_fdvp; 757 struct vnode *a_fvp; 758 struct componentname *a_fcnp; 759 struct vnode *a_tdvp; 760 struct vnode *a_tvp; 761 struct componentname *a_tcnp; 762 } */ *ap = v; 763 struct vnode *tvp, *fvp, *tdvp, *fdvp; 764 struct componentname *tcnp, *fcnp; 765 int error; 766 struct lfs *fs; 767 768 fs = VTOI(ap->a_fdvp)->i_lfs; 769 tvp = ap->a_tvp; 770 tdvp = ap->a_tdvp; 771 tcnp = ap->a_tcnp; 772 fvp = ap->a_fvp; 773 fdvp = ap->a_fdvp; 774 fcnp = ap->a_fcnp; 775 776 /* 777 * Check for cross-device rename. 778 * If it is, we don't want to set dirops, just error out. 779 * (In particular note that MARK_VNODE(tdvp) will DTWT on 780 * a cross-device rename.) 781 * 782 * Copied from ufs_rename. 783 */ 784 if ((fvp->v_mount != tdvp->v_mount) || 785 (tvp && (fvp->v_mount != tvp->v_mount))) { 786 error = EXDEV; 787 goto errout; 788 } 789 790 /* 791 * Check to make sure we're not renaming a vnode onto itself 792 * (deleting a hard link by renaming one name onto another); 793 * if we are we can't recursively call VOP_REMOVE since that 794 * would leave us with an unaccounted-for number of live dirops. 795 * 796 * Inline the relevant section of ufs_rename here, *before* 797 * calling SET_DIROP_REMOVE. 798 */ 799 if (tvp && ((VTOI(tvp)->i_flags & (IMMUTABLE | APPEND)) || 800 (VTOI(tdvp)->i_flags & APPEND))) { 801 error = EPERM; 802 goto errout; 803 } 804 if (fvp == tvp) { 805 if (fvp->v_type == VDIR) { 806 error = EINVAL; 807 goto errout; 808 } 809 810 /* Release destination completely. */ 811 VOP_ABORTOP(tdvp, tcnp); 812 vput(tdvp); 813 vput(tvp); 814 815 /* Delete source. */ 816 vrele(fvp); 817 fcnp->cn_flags &= ~(MODMASK | SAVESTART); 818 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 819 fcnp->cn_nameiop = DELETE; 820 if ((error = relookup(fdvp, &fvp, fcnp))){ 821 /* relookup blew away fdvp */ 822 return (error); 823 } 824 return (VOP_REMOVE(fdvp, fvp, fcnp)); 825 } 826 827 if ((error = SET_DIROP_REMOVE(tdvp, tvp)) != 0) 828 goto errout; 829 MARK_VNODE(fdvp); 830 MARK_VNODE(fvp); 831 832 error = ufs_rename(ap); 833 UNMARK_VNODE(fdvp); 834 UNMARK_VNODE(fvp); 835 SET_ENDOP_REMOVE(fs, tdvp, tvp, "rename"); 836 return (error); 837 838 errout: 839 VOP_ABORTOP(tdvp, ap->a_tcnp); /* XXX, why not in NFS? */ 840 if (tdvp == tvp) 841 vrele(tdvp); 842 else 843 vput(tdvp); 844 if (tvp) 845 vput(tvp); 846 VOP_ABORTOP(fdvp, ap->a_fcnp); /* XXX, why not in NFS? */ 847 vrele(fdvp); 848 vrele(fvp); 849 return (error); 850 } 851 852 /* XXX hack to avoid calling ITIMES in getattr */ 853 int 854 lfs_getattr(void *v) 855 { 856 struct vop_getattr_args /* { 857 struct vnode *a_vp; 858 struct vattr *a_vap; 859 struct ucred *a_cred; 860 struct lwp *a_l; 861 } */ *ap = v; 862 struct vnode *vp = ap->a_vp; 863 struct inode *ip = VTOI(vp); 864 struct vattr *vap = ap->a_vap; 865 struct lfs *fs = ip->i_lfs; 866 /* 867 * Copy from inode table 868 */ 869 vap->va_fsid = ip->i_dev; 870 vap->va_fileid = ip->i_number; 871 vap->va_mode = ip->i_mode & ~IFMT; 872 vap->va_nlink = ip->i_nlink; 873 vap->va_uid = ip->i_uid; 874 vap->va_gid = ip->i_gid; 875 vap->va_rdev = (dev_t)ip->i_ffs1_rdev; 876 vap->va_size = vp->v_size; 877 vap->va_atime.tv_sec = ip->i_ffs1_atime; 878 vap->va_atime.tv_nsec = ip->i_ffs1_atimensec; 879 vap->va_mtime.tv_sec = ip->i_ffs1_mtime; 880 vap->va_mtime.tv_nsec = ip->i_ffs1_mtimensec; 881 vap->va_ctime.tv_sec = ip->i_ffs1_ctime; 882 vap->va_ctime.tv_nsec = ip->i_ffs1_ctimensec; 883 vap->va_flags = ip->i_flags; 884 vap->va_gen = ip->i_gen; 885 /* this doesn't belong here */ 886 if (vp->v_type == VBLK) 887 vap->va_blocksize = BLKDEV_IOSIZE; 888 else if (vp->v_type == VCHR) 889 vap->va_blocksize = MAXBSIZE; 890 else 891 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize; 892 vap->va_bytes = fsbtob(fs, (u_quad_t)ip->i_lfs_effnblks); 893 vap->va_type = vp->v_type; 894 vap->va_filerev = ip->i_modrev; 895 return (0); 896 } 897 898 /* 899 * Check to make sure the inode blocks won't choke the buffer 900 * cache, then call ufs_setattr as usual. 901 */ 902 int 903 lfs_setattr(void *v) 904 { 905 struct vop_setattr_args /* { 906 struct vnode *a_vp; 907 struct vattr *a_vap; 908 struct ucred *a_cred; 909 struct lwp *a_l; 910 } */ *ap = v; 911 struct vnode *vp = ap->a_vp; 912 913 lfs_check(vp, LFS_UNUSED_LBN, 0); 914 return ufs_setattr(v); 915 } 916 917 /* 918 * Close called 919 * 920 * XXX -- we were using ufs_close, but since it updates the 921 * times on the inode, we might need to bump the uinodes 922 * count. 923 */ 924 /* ARGSUSED */ 925 int 926 lfs_close(void *v) 927 { 928 struct vop_close_args /* { 929 struct vnode *a_vp; 930 int a_fflag; 931 struct ucred *a_cred; 932 struct lwp *a_l; 933 } */ *ap = v; 934 struct vnode *vp = ap->a_vp; 935 struct inode *ip = VTOI(vp); 936 937 if (vp == ip->i_lfs->lfs_ivnode && 938 vp->v_mount->mnt_iflag & IMNT_UNMOUNT) 939 return 0; 940 941 if (vp->v_usecount > 1 && vp != ip->i_lfs->lfs_ivnode) { 942 LFS_ITIMES(ip, NULL, NULL, NULL); 943 } 944 return (0); 945 } 946 947 /* 948 * Close wrapper for special devices. 949 * 950 * Update the times on the inode then do device close. 951 */ 952 int 953 lfsspec_close(void *v) 954 { 955 struct vop_close_args /* { 956 struct vnode *a_vp; 957 int a_fflag; 958 struct ucred *a_cred; 959 struct lwp *a_l; 960 } */ *ap = v; 961 struct vnode *vp; 962 struct inode *ip; 963 964 vp = ap->a_vp; 965 ip = VTOI(vp); 966 if (vp->v_usecount > 1) { 967 LFS_ITIMES(ip, NULL, NULL, NULL); 968 } 969 return (VOCALL (spec_vnodeop_p, VOFFSET(vop_close), ap)); 970 } 971 972 /* 973 * Close wrapper for fifo's. 974 * 975 * Update the times on the inode then do device close. 976 */ 977 int 978 lfsfifo_close(void *v) 979 { 980 struct vop_close_args /* { 981 struct vnode *a_vp; 982 int a_fflag; 983 struct ucred *a_cred; 984 struct lwp *a_l; 985 } */ *ap = v; 986 struct vnode *vp; 987 struct inode *ip; 988 989 vp = ap->a_vp; 990 ip = VTOI(vp); 991 if (ap->a_vp->v_usecount > 1) { 992 LFS_ITIMES(ip, NULL, NULL, NULL); 993 } 994 return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_close), ap)); 995 } 996 997 /* 998 * Reclaim an inode so that it can be used for other purposes. 999 */ 1000 1001 int 1002 lfs_reclaim(void *v) 1003 { 1004 struct vop_reclaim_args /* { 1005 struct vnode *a_vp; 1006 struct lwp *a_l; 1007 } */ *ap = v; 1008 struct vnode *vp = ap->a_vp; 1009 struct inode *ip = VTOI(vp); 1010 int error; 1011 1012 KASSERT(ip->i_nlink == ip->i_ffs_effnlink); 1013 1014 LFS_CLR_UINO(ip, IN_ALLMOD); 1015 if ((error = ufs_reclaim(vp, ap->a_l))) 1016 return (error); 1017 pool_put(&lfs_dinode_pool, ip->i_din.ffs1_din); 1018 lfs_deregister_all(vp); 1019 pool_put(&lfs_inoext_pool, ip->inode_ext.lfs); 1020 ip->inode_ext.lfs = NULL; 1021 pool_put(&lfs_inode_pool, vp->v_data); 1022 vp->v_data = NULL; 1023 return (0); 1024 } 1025 1026 /* 1027 * Read a block from a storage device. 1028 * In order to avoid reading blocks that are in the process of being 1029 * written by the cleaner---and hence are not mutexed by the normal 1030 * buffer cache / page cache mechanisms---check for collisions before 1031 * reading. 1032 * 1033 * We inline ufs_strategy to make sure that the VOP_BMAP occurs *before* 1034 * the active cleaner test. 1035 * 1036 * XXX This code assumes that lfs_markv makes synchronous checkpoints. 1037 */ 1038 int 1039 lfs_strategy(void *v) 1040 { 1041 struct vop_strategy_args /* { 1042 struct vnode *a_vp; 1043 struct buf *a_bp; 1044 } */ *ap = v; 1045 struct buf *bp; 1046 struct lfs *fs; 1047 struct vnode *vp; 1048 struct inode *ip; 1049 daddr_t tbn; 1050 int i, sn, error, slept; 1051 1052 bp = ap->a_bp; 1053 vp = ap->a_vp; 1054 ip = VTOI(vp); 1055 fs = ip->i_lfs; 1056 1057 /* lfs uses its strategy routine only for read */ 1058 KASSERT(bp->b_flags & B_READ); 1059 1060 if (vp->v_type == VBLK || vp->v_type == VCHR) 1061 panic("lfs_strategy: spec"); 1062 KASSERT(bp->b_bcount != 0); 1063 if (bp->b_blkno == bp->b_lblkno) { 1064 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, 1065 NULL); 1066 if (error) { 1067 bp->b_error = error; 1068 bp->b_flags |= B_ERROR; 1069 biodone(bp); 1070 return (error); 1071 } 1072 if ((long)bp->b_blkno == -1) /* no valid data */ 1073 clrbuf(bp); 1074 } 1075 if ((long)bp->b_blkno < 0) { /* block is not on disk */ 1076 biodone(bp); 1077 return (0); 1078 } 1079 1080 slept = 1; 1081 simple_lock(&fs->lfs_interlock); 1082 while (slept && fs->lfs_seglock) { 1083 simple_unlock(&fs->lfs_interlock); 1084 /* 1085 * Look through list of intervals. 1086 * There will only be intervals to look through 1087 * if the cleaner holds the seglock. 1088 * Since the cleaner is synchronous, we can trust 1089 * the list of intervals to be current. 1090 */ 1091 tbn = dbtofsb(fs, bp->b_blkno); 1092 sn = dtosn(fs, tbn); 1093 slept = 0; 1094 for (i = 0; i < fs->lfs_cleanind; i++) { 1095 if (sn == dtosn(fs, fs->lfs_cleanint[i]) && 1096 tbn >= fs->lfs_cleanint[i]) { 1097 DLOG((DLOG_CLEAN, 1098 "lfs_strategy: ino %d lbn %" PRId64 1099 " ind %d sn %d fsb %" PRIx32 1100 " given sn %d fsb %" PRIx64 "\n", 1101 ip->i_number, bp->b_lblkno, i, 1102 dtosn(fs, fs->lfs_cleanint[i]), 1103 fs->lfs_cleanint[i], sn, tbn)); 1104 DLOG((DLOG_CLEAN, 1105 "lfs_strategy: sleeping on ino %d lbn %" 1106 PRId64 "\n", ip->i_number, bp->b_lblkno)); 1107 simple_lock(&fs->lfs_interlock); 1108 if (fs->lfs_seglock) { 1109 ltsleep(&fs->lfs_seglock, 1110 (PRIBIO + 1) | PNORELOCK, 1111 "lfs_strategy", 0, 1112 &fs->lfs_interlock); 1113 slept = 1; 1114 break; 1115 } 1116 simple_unlock(&fs->lfs_interlock); 1117 } 1118 } 1119 simple_lock(&fs->lfs_interlock); 1120 } 1121 simple_unlock(&fs->lfs_interlock); 1122 1123 vp = ip->i_devvp; 1124 VOP_STRATEGY(vp, bp); 1125 return (0); 1126 } 1127 1128 static void 1129 lfs_flush_dirops(struct lfs *fs) 1130 { 1131 struct inode *ip, *nip; 1132 struct vnode *vp; 1133 extern int lfs_dostats; 1134 struct segment *sp; 1135 int needunlock; 1136 1137 ASSERT_MAYBE_SEGLOCK(fs); 1138 1139 if (fs->lfs_ronly) 1140 return; 1141 1142 simple_lock(&fs->lfs_interlock); 1143 if (TAILQ_FIRST(&fs->lfs_dchainhd) == NULL) { 1144 simple_unlock(&fs->lfs_interlock); 1145 return; 1146 } else 1147 simple_unlock(&fs->lfs_interlock); 1148 1149 if (lfs_dostats) 1150 ++lfs_stats.flush_invoked; 1151 1152 /* 1153 * Inline lfs_segwrite/lfs_writevnodes, but just for dirops. 1154 * Technically this is a checkpoint (the on-disk state is valid) 1155 * even though we are leaving out all the file data. 1156 */ 1157 lfs_imtime(fs); 1158 lfs_seglock(fs, SEGM_CKP); 1159 sp = fs->lfs_sp; 1160 1161 /* 1162 * lfs_writevnodes, optimized to get dirops out of the way. 1163 * Only write dirops, and don't flush files' pages, only 1164 * blocks from the directories. 1165 * 1166 * We don't need to vref these files because they are 1167 * dirops and so hold an extra reference until the 1168 * segunlock clears them of that status. 1169 * 1170 * We don't need to check for IN_ADIROP because we know that 1171 * no dirops are active. 1172 * 1173 */ 1174 simple_lock(&fs->lfs_interlock); 1175 for (ip = TAILQ_FIRST(&fs->lfs_dchainhd); ip != NULL; ip = nip) { 1176 nip = TAILQ_NEXT(ip, i_lfs_dchain); 1177 simple_unlock(&fs->lfs_interlock); 1178 vp = ITOV(ip); 1179 1180 /* 1181 * All writes to directories come from dirops; all 1182 * writes to files' direct blocks go through the page 1183 * cache, which we're not touching. Reads to files 1184 * and/or directories will not be affected by writing 1185 * directory blocks inodes and file inodes. So we don't 1186 * really need to lock. If we don't lock, though, 1187 * make sure that we don't clear IN_MODIFIED 1188 * unnecessarily. 1189 */ 1190 if (vp->v_flag & (VXLOCK | VFREEING)) { 1191 simple_lock(&fs->lfs_interlock); 1192 continue; 1193 } 1194 if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT) == 0) { 1195 needunlock = 1; 1196 } else { 1197 DLOG((DLOG_VNODE, "lfs_flush_dirops: flushing locked ino %d\n", 1198 VTOI(vp)->i_number)); 1199 needunlock = 0; 1200 } 1201 if (vp->v_type != VREG && 1202 ((ip->i_flag & IN_ALLMOD) || !VPISEMPTY(vp))) { 1203 lfs_writefile(fs, sp, vp); 1204 if (!VPISEMPTY(vp) && !WRITEINPROG(vp) && 1205 !(ip->i_flag & IN_ALLMOD)) { 1206 LFS_SET_UINO(ip, IN_MODIFIED); 1207 } 1208 } 1209 (void) lfs_writeinode(fs, sp, ip); 1210 if (needunlock) 1211 VOP_UNLOCK(vp, 0); 1212 else 1213 LFS_SET_UINO(ip, IN_MODIFIED); 1214 simple_lock(&fs->lfs_interlock); 1215 } 1216 simple_unlock(&fs->lfs_interlock); 1217 /* We've written all the dirops there are */ 1218 ((SEGSUM *)(sp->segsum))->ss_flags &= ~(SS_CONT); 1219 (void) lfs_writeseg(fs, sp); 1220 lfs_segunlock(fs); 1221 } 1222 1223 /* 1224 * Flush all vnodes for which the pagedaemon has requested pageouts. 1225 * Skip over any files that are marked VDIROP (since lfs_flush_dirop() 1226 * has just run, this would be an error). If we have to skip a vnode 1227 * for any reason, just skip it; if we have to wait for the cleaner, 1228 * abort. The writer daemon will call us again later. 1229 */ 1230 void 1231 lfs_flush_pchain(struct lfs *fs) 1232 { 1233 struct inode *ip, *nip; 1234 struct vnode *vp; 1235 extern int lfs_dostats; 1236 struct segment *sp; 1237 int error; 1238 1239 ASSERT_NO_SEGLOCK(fs); 1240 1241 if (fs->lfs_ronly) 1242 return; 1243 1244 simple_lock(&fs->lfs_interlock); 1245 if (TAILQ_FIRST(&fs->lfs_pchainhd) == NULL) { 1246 simple_unlock(&fs->lfs_interlock); 1247 return; 1248 } else 1249 simple_unlock(&fs->lfs_interlock); 1250 1251 /* Get dirops out of the way */ 1252 lfs_flush_dirops(fs); 1253 1254 if (lfs_dostats) 1255 ++lfs_stats.flush_invoked; 1256 1257 /* 1258 * Inline lfs_segwrite/lfs_writevnodes, but just for pageouts. 1259 */ 1260 lfs_imtime(fs); 1261 lfs_seglock(fs, 0); 1262 sp = fs->lfs_sp; 1263 1264 /* 1265 * lfs_writevnodes, optimized to clear pageout requests. 1266 * Only write non-dirop files that are in the pageout queue. 1267 * We're very conservative about what we write; we want to be 1268 * fast and async. 1269 */ 1270 simple_lock(&fs->lfs_interlock); 1271 top: 1272 for (ip = TAILQ_FIRST(&fs->lfs_pchainhd); ip != NULL; ip = nip) { 1273 nip = TAILQ_NEXT(ip, i_lfs_pchain); 1274 vp = ITOV(ip); 1275 1276 if (!(ip->i_flags & IN_PAGING)) 1277 goto top; 1278 1279 if (vp->v_flag & (VXLOCK|VDIROP)) 1280 continue; 1281 if (vp->v_type != VREG) 1282 continue; 1283 if (lfs_vref(vp)) 1284 continue; 1285 simple_unlock(&fs->lfs_interlock); 1286 1287 if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT) != 0) { 1288 lfs_vunref(vp); 1289 simple_lock(&fs->lfs_interlock); 1290 continue; 1291 } 1292 1293 error = lfs_writefile(fs, sp, vp); 1294 if (!VPISEMPTY(vp) && !WRITEINPROG(vp) && 1295 !(ip->i_flag & IN_ALLMOD)) { 1296 LFS_SET_UINO(ip, IN_MODIFIED); 1297 } 1298 (void) lfs_writeinode(fs, sp, ip); 1299 1300 VOP_UNLOCK(vp, 0); 1301 lfs_vunref(vp); 1302 1303 simple_lock(&fs->lfs_interlock); 1304 1305 if (error == EAGAIN) { 1306 lfs_writeseg(fs, sp); 1307 break; 1308 } 1309 } 1310 simple_unlock(&fs->lfs_interlock); 1311 (void) lfs_writeseg(fs, sp); 1312 lfs_segunlock(fs); 1313 } 1314 1315 /* 1316 * Provide a fcntl interface to sys_lfs_{segwait,bmapv,markv}. 1317 */ 1318 int 1319 lfs_fcntl(void *v) 1320 { 1321 struct vop_fcntl_args /* { 1322 struct vnode *a_vp; 1323 u_long a_command; 1324 caddr_t a_data; 1325 int a_fflag; 1326 struct ucred *a_cred; 1327 struct lwp *a_l; 1328 } */ *ap = v; 1329 struct timeval *tvp; 1330 BLOCK_INFO *blkiov; 1331 CLEANERINFO *cip; 1332 SEGUSE *sup; 1333 int blkcnt, error, oclean; 1334 struct lfs_fcntl_markv blkvp; 1335 struct proc *p; 1336 fsid_t *fsidp; 1337 struct lfs *fs; 1338 struct buf *bp; 1339 fhandle_t *fhp; 1340 daddr_t off; 1341 1342 /* Only respect LFS fcntls on fs root or Ifile */ 1343 if (VTOI(ap->a_vp)->i_number != ROOTINO && 1344 VTOI(ap->a_vp)->i_number != LFS_IFILE_INUM) { 1345 return ufs_fcntl(v); 1346 } 1347 1348 /* Avoid locking a draining lock */ 1349 if (ap->a_vp->v_mount->mnt_iflag & IMNT_UNMOUNT) { 1350 return ESHUTDOWN; 1351 } 1352 1353 p = ap->a_l->l_proc; 1354 fs = VTOI(ap->a_vp)->i_lfs; 1355 fsidp = &ap->a_vp->v_mount->mnt_stat.f_fsidx; 1356 1357 switch (ap->a_command) { 1358 case LFCNSEGWAITALL: 1359 case LFCNSEGWAITALL_COMPAT: 1360 fsidp = NULL; 1361 /* FALLSTHROUGH */ 1362 case LFCNSEGWAIT: 1363 case LFCNSEGWAIT_COMPAT: 1364 tvp = (struct timeval *)ap->a_data; 1365 simple_lock(&fs->lfs_interlock); 1366 ++fs->lfs_sleepers; 1367 simple_unlock(&fs->lfs_interlock); 1368 VOP_UNLOCK(ap->a_vp, 0); 1369 1370 error = lfs_segwait(fsidp, tvp); 1371 1372 VOP_LOCK(ap->a_vp, LK_EXCLUSIVE); 1373 simple_lock(&fs->lfs_interlock); 1374 if (--fs->lfs_sleepers == 0) 1375 wakeup(&fs->lfs_sleepers); 1376 simple_unlock(&fs->lfs_interlock); 1377 return error; 1378 1379 case LFCNBMAPV: 1380 case LFCNMARKV: 1381 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1382 return (error); 1383 blkvp = *(struct lfs_fcntl_markv *)ap->a_data; 1384 1385 blkcnt = blkvp.blkcnt; 1386 if ((u_int) blkcnt > LFS_MARKV_MAXBLKCNT) 1387 return (EINVAL); 1388 blkiov = lfs_malloc(fs, blkcnt * sizeof(BLOCK_INFO), LFS_NB_BLKIOV); 1389 if ((error = copyin(blkvp.blkiov, blkiov, 1390 blkcnt * sizeof(BLOCK_INFO))) != 0) { 1391 lfs_free(fs, blkiov, LFS_NB_BLKIOV); 1392 return error; 1393 } 1394 1395 simple_lock(&fs->lfs_interlock); 1396 ++fs->lfs_sleepers; 1397 simple_unlock(&fs->lfs_interlock); 1398 VOP_UNLOCK(ap->a_vp, 0); 1399 if (ap->a_command == LFCNBMAPV) 1400 error = lfs_bmapv(p, fsidp, blkiov, blkcnt); 1401 else /* LFCNMARKV */ 1402 error = lfs_markv(p, fsidp, blkiov, blkcnt); 1403 if (error == 0) 1404 error = copyout(blkiov, blkvp.blkiov, 1405 blkcnt * sizeof(BLOCK_INFO)); 1406 VOP_LOCK(ap->a_vp, LK_EXCLUSIVE); 1407 simple_lock(&fs->lfs_interlock); 1408 if (--fs->lfs_sleepers == 0) 1409 wakeup(&fs->lfs_sleepers); 1410 simple_unlock(&fs->lfs_interlock); 1411 lfs_free(fs, blkiov, LFS_NB_BLKIOV); 1412 return error; 1413 1414 case LFCNRECLAIM: 1415 /* 1416 * Flush dirops and write Ifile, allowing empty segments 1417 * to be immediately reclaimed. 1418 */ 1419 VOP_UNLOCK(ap->a_vp, 0); 1420 lfs_writer_enter(fs, "pndirop"); 1421 off = fs->lfs_offset; 1422 lfs_seglock(fs, SEGM_FORCE_CKP | SEGM_CKP); 1423 lfs_flush_dirops(fs); 1424 LFS_CLEANERINFO(cip, fs, bp); 1425 oclean = cip->clean; 1426 LFS_SYNC_CLEANERINFO(cip, fs, bp, 1); 1427 lfs_segwrite(ap->a_vp->v_mount, SEGM_FORCE_CKP); 1428 fs->lfs_sp->seg_flags |= SEGM_PROT; 1429 lfs_segunlock(fs); 1430 lfs_writer_leave(fs); 1431 1432 #ifdef DEBUG 1433 LFS_CLEANERINFO(cip, fs, bp); 1434 DLOG((DLOG_CLEAN, "lfs_fcntl: reclaim wrote %" PRId64 1435 " blocks, cleaned %" PRId32 " segments (activesb %d)\n", 1436 fs->lfs_offset - off, cip->clean - oclean, 1437 fs->lfs_activesb)); 1438 LFS_SYNC_CLEANERINFO(cip, fs, bp, 0); 1439 #endif 1440 1441 VOP_LOCK(ap->a_vp, LK_EXCLUSIVE); 1442 return 0; 1443 1444 case LFCNIFILEFH: 1445 /* Return the filehandle of the Ifile */ 1446 if ((error = suser(ap->a_l->l_proc->p_ucred, &ap->a_l->l_proc->p_acflag)) != 0) 1447 return (error); 1448 fhp = (struct fhandle *)ap->a_data; 1449 fhp->fh_fsid = *fsidp; 1450 return lfs_vptofh(fs->lfs_ivnode, &(fhp->fh_fid)); 1451 1452 case LFCNREWIND: 1453 /* Move lfs_offset to the lowest-numbered segment */ 1454 return lfs_rewind(fs, *(int *)ap->a_data); 1455 1456 case LFCNINVAL: 1457 /* Mark a segment SEGUSE_INVAL */ 1458 LFS_SEGENTRY(sup, fs, *(int *)ap->a_data, bp); 1459 if (sup->su_nbytes > 0) { 1460 brelse(bp); 1461 lfs_unset_inval_all(fs); 1462 return EBUSY; 1463 } 1464 sup->su_flags |= SEGUSE_INVAL; 1465 VOP_BWRITE(bp); 1466 return 0; 1467 1468 case LFCNRESIZE: 1469 /* Resize the filesystem */ 1470 return lfs_resize_fs(fs, *(int *)ap->a_data); 1471 1472 case LFCNWRAPSTOP: 1473 /* 1474 * Hold lfs_newseg at segment 0; sleep until the filesystem 1475 * wraps around. For debugging purposes, so an external 1476 * agent can log every segment in the filesystem as it 1477 * was written, and we can regression-test checkpoint 1478 * validity in the general case. 1479 */ 1480 VOP_UNLOCK(ap->a_vp, 0); 1481 simple_lock(&fs->lfs_interlock); 1482 fs->lfs_nowrap = 1; 1483 error = ltsleep(&fs->lfs_nowrap, PCATCH | PUSER | PNORELOCK, 1484 "segwrap", 0, &fs->lfs_interlock); 1485 if (error) { 1486 fs->lfs_nowrap = 0; 1487 wakeup(&fs->lfs_nowrap); 1488 } 1489 VOP_LOCK(ap->a_vp, LK_EXCLUSIVE); 1490 return 0; 1491 1492 case LFCNWRAPGO: 1493 /* 1494 * Having done its work, the agent wakes up the writer. 1495 * It sleeps until a new segment is selected. 1496 */ 1497 VOP_UNLOCK(ap->a_vp, 0); 1498 simple_lock(&fs->lfs_interlock); 1499 fs->lfs_nowrap = 0; 1500 wakeup(&fs->lfs_nowrap); 1501 ltsleep(&fs->lfs_nextseg, PCATCH | PUSER | PNORELOCK, 1502 "segment", 0, &fs->lfs_interlock); 1503 VOP_LOCK(ap->a_vp, LK_EXCLUSIVE); 1504 return 0; 1505 1506 default: 1507 return ufs_fcntl(v); 1508 } 1509 return 0; 1510 } 1511 1512 int 1513 lfs_getpages(void *v) 1514 { 1515 struct vop_getpages_args /* { 1516 struct vnode *a_vp; 1517 voff_t a_offset; 1518 struct vm_page **a_m; 1519 int *a_count; 1520 int a_centeridx; 1521 vm_prot_t a_access_type; 1522 int a_advice; 1523 int a_flags; 1524 } */ *ap = v; 1525 1526 if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM && 1527 (ap->a_access_type & VM_PROT_WRITE) != 0) { 1528 return EPERM; 1529 } 1530 if ((ap->a_access_type & VM_PROT_WRITE) != 0) { 1531 LFS_SET_UINO(VTOI(ap->a_vp), IN_MODIFIED); 1532 } 1533 1534 /* 1535 * we're relying on the fact that genfs_getpages() always read in 1536 * entire filesystem blocks. 1537 */ 1538 return genfs_getpages(v); 1539 } 1540 1541 /* 1542 * Make sure that for all pages in every block in the given range, 1543 * either all are dirty or all are clean. If any of the pages 1544 * we've seen so far are dirty, put the vnode on the paging chain, 1545 * and mark it IN_PAGING. 1546 * 1547 * If checkfirst != 0, don't check all the pages but return at the 1548 * first dirty page. 1549 */ 1550 static int 1551 check_dirty(struct lfs *fs, struct vnode *vp, 1552 off_t startoffset, off_t endoffset, off_t blkeof, 1553 int flags, int checkfirst) 1554 { 1555 int by_list; 1556 struct vm_page *curpg = NULL; /* XXX: gcc */ 1557 struct vm_page *pgs[MAXBSIZE / PAGE_SIZE], *pg; 1558 off_t soff = 0; /* XXX: gcc */ 1559 voff_t off; 1560 int i; 1561 int nonexistent; 1562 int any_dirty; /* number of dirty pages */ 1563 int dirty; /* number of dirty pages in a block */ 1564 int tdirty; 1565 int pages_per_block = fs->lfs_bsize >> PAGE_SHIFT; 1566 int pagedaemon = (curproc == uvm.pagedaemon_proc); 1567 1568 ASSERT_MAYBE_SEGLOCK(fs); 1569 top: 1570 by_list = (vp->v_uobj.uo_npages <= 1571 ((endoffset - startoffset) >> PAGE_SHIFT) * 1572 UVM_PAGE_HASH_PENALTY); 1573 any_dirty = 0; 1574 1575 if (by_list) { 1576 curpg = TAILQ_FIRST(&vp->v_uobj.memq); 1577 } else { 1578 soff = startoffset; 1579 } 1580 while (by_list || soff < MIN(blkeof, endoffset)) { 1581 if (by_list) { 1582 /* 1583 * Find the first page in a block. Skip 1584 * blocks outside our area of interest or beyond 1585 * the end of file. 1586 */ 1587 if (pages_per_block > 1) { 1588 while (curpg && 1589 ((curpg->offset & fs->lfs_bmask) || 1590 curpg->offset >= vp->v_size || 1591 curpg->offset >= endoffset)) 1592 curpg = TAILQ_NEXT(curpg, listq); 1593 } 1594 if (curpg == NULL) 1595 break; 1596 soff = curpg->offset; 1597 } 1598 1599 /* 1600 * Mark all pages in extended range busy; find out if any 1601 * of them are dirty. 1602 */ 1603 nonexistent = dirty = 0; 1604 for (i = 0; i == 0 || i < pages_per_block; i++) { 1605 if (by_list && pages_per_block <= 1) { 1606 pgs[i] = pg = curpg; 1607 } else { 1608 off = soff + (i << PAGE_SHIFT); 1609 pgs[i] = pg = uvm_pagelookup(&vp->v_uobj, off); 1610 if (pg == NULL) { 1611 ++nonexistent; 1612 continue; 1613 } 1614 } 1615 KASSERT(pg != NULL); 1616 1617 /* 1618 * If we're holding the segment lock, we can deadlocked 1619 * against a process that has our page and is waiting 1620 * for the cleaner, while the cleaner waits for the 1621 * segment lock. Just bail in that case. 1622 */ 1623 if ((pg->flags & PG_BUSY) && 1624 (pagedaemon || LFS_SEGLOCK_HELD(fs))) { 1625 if (by_list && i > 0) 1626 uvm_page_unbusy(pgs, i); 1627 DLOG((DLOG_PAGE, "lfs_putpages: avoiding 3-way or pagedaemon deadlock\n")); 1628 return -1; 1629 } 1630 1631 while (pg->flags & PG_BUSY) { 1632 pg->flags |= PG_WANTED; 1633 UVM_UNLOCK_AND_WAIT(pg, &vp->v_interlock, 0, 1634 "lfsput", 0); 1635 simple_lock(&vp->v_interlock); 1636 if (by_list) { 1637 if (i > 0) 1638 uvm_page_unbusy(pgs, i); 1639 goto top; 1640 } 1641 } 1642 pg->flags |= PG_BUSY; 1643 UVM_PAGE_OWN(pg, "lfs_putpages"); 1644 1645 pmap_page_protect(pg, VM_PROT_NONE); 1646 tdirty = (pmap_clear_modify(pg) || 1647 (pg->flags & PG_CLEAN) == 0); 1648 dirty += tdirty; 1649 } 1650 if (pages_per_block > 0 && nonexistent >= pages_per_block) { 1651 if (by_list) { 1652 curpg = TAILQ_NEXT(curpg, listq); 1653 } else { 1654 soff += fs->lfs_bsize; 1655 } 1656 continue; 1657 } 1658 1659 any_dirty += dirty; 1660 KASSERT(nonexistent == 0); 1661 1662 /* 1663 * If any are dirty make all dirty; unbusy them, 1664 * but if we were asked to clean, wire them so that 1665 * the pagedaemon doesn't bother us about them while 1666 * they're on their way to disk. 1667 */ 1668 for (i = 0; i == 0 || i < pages_per_block; i++) { 1669 pg = pgs[i]; 1670 KASSERT(!((pg->flags & PG_CLEAN) && (pg->flags & PG_DELWRI))); 1671 if (dirty) { 1672 pg->flags &= ~PG_CLEAN; 1673 if (flags & PGO_FREE) { 1674 /* 1675 * Wire the page so that 1676 * pdaemon doesn't see it again. 1677 */ 1678 uvm_lock_pageq(); 1679 uvm_pagewire(pg); 1680 uvm_unlock_pageq(); 1681 1682 /* Suspended write flag */ 1683 pg->flags |= PG_DELWRI; 1684 } 1685 } 1686 if (pg->flags & PG_WANTED) 1687 wakeup(pg); 1688 pg->flags &= ~(PG_WANTED|PG_BUSY); 1689 UVM_PAGE_OWN(pg, NULL); 1690 } 1691 1692 if (checkfirst && any_dirty) 1693 break; 1694 1695 if (by_list) { 1696 curpg = TAILQ_NEXT(curpg, listq); 1697 } else { 1698 soff += MAX(PAGE_SIZE, fs->lfs_bsize); 1699 } 1700 } 1701 1702 return any_dirty; 1703 } 1704 1705 /* 1706 * lfs_putpages functions like genfs_putpages except that 1707 * 1708 * (1) It needs to bounds-check the incoming requests to ensure that 1709 * they are block-aligned; if they are not, expand the range and 1710 * do the right thing in case, e.g., the requested range is clean 1711 * but the expanded range is dirty. 1712 * (2) It needs to explicitly send blocks to be written when it is done. 1713 * VOP_PUTPAGES is not ever called with the seglock held, so 1714 * we simply take the seglock and let lfs_segunlock wait for us. 1715 * XXX Actually we can be called with the seglock held, if we have 1716 * XXX to flush a vnode while lfs_markv is in operation. As of this 1717 * XXX writing we panic in this case. 1718 * 1719 * Assumptions: 1720 * 1721 * (1) The caller does not hold any pages in this vnode busy. If it does, 1722 * there is a danger that when we expand the page range and busy the 1723 * pages we will deadlock. 1724 * (2) We are called with vp->v_interlock held; we must return with it 1725 * released. 1726 * (3) We don't absolutely have to free pages right away, provided that 1727 * the request does not have PGO_SYNCIO. When the pagedaemon gives 1728 * us a request with PGO_FREE, we take the pages out of the paging 1729 * queue and wake up the writer, which will handle freeing them for us. 1730 * 1731 * We ensure that for any filesystem block, all pages for that 1732 * block are either resident or not, even if those pages are higher 1733 * than EOF; that means that we will be getting requests to free 1734 * "unused" pages above EOF all the time, and should ignore them. 1735 * 1736 * XXX note that we're (ab)using PGO_LOCKED as "seglock held". 1737 */ 1738 1739 int 1740 lfs_putpages(void *v) 1741 { 1742 int error; 1743 struct vop_putpages_args /* { 1744 struct vnode *a_vp; 1745 voff_t a_offlo; 1746 voff_t a_offhi; 1747 int a_flags; 1748 } */ *ap = v; 1749 struct vnode *vp; 1750 struct inode *ip; 1751 struct lfs *fs; 1752 struct segment *sp; 1753 off_t origoffset, startoffset, endoffset, origendoffset, blkeof; 1754 off_t off, max_endoffset; 1755 int s; 1756 boolean_t seglocked, sync, pagedaemon; 1757 struct vm_page *pg; 1758 UVMHIST_FUNC("lfs_putpages"); UVMHIST_CALLED(ubchist); 1759 1760 vp = ap->a_vp; 1761 ip = VTOI(vp); 1762 fs = ip->i_lfs; 1763 sync = (ap->a_flags & PGO_SYNCIO) != 0; 1764 pagedaemon = (curproc == uvm.pagedaemon_proc); 1765 1766 /* Putpages does nothing for metadata. */ 1767 if (vp == fs->lfs_ivnode || vp->v_type != VREG) { 1768 simple_unlock(&vp->v_interlock); 1769 return 0; 1770 } 1771 1772 /* 1773 * If there are no pages, don't do anything. 1774 */ 1775 if (vp->v_uobj.uo_npages == 0) { 1776 s = splbio(); 1777 if (LIST_FIRST(&vp->v_dirtyblkhd) == NULL && 1778 (vp->v_flag & VONWORKLST)) { 1779 vp->v_flag &= ~VONWORKLST; 1780 LIST_REMOVE(vp, v_synclist); 1781 } 1782 splx(s); 1783 simple_unlock(&vp->v_interlock); 1784 1785 /* Remove us from paging queue, if we were on it */ 1786 simple_lock(&fs->lfs_interlock); 1787 if (ip->i_flags & IN_PAGING) { 1788 ip->i_flags &= ~IN_PAGING; 1789 TAILQ_REMOVE(&fs->lfs_pchainhd, ip, i_lfs_pchain); 1790 } 1791 simple_unlock(&fs->lfs_interlock); 1792 return 0; 1793 } 1794 1795 blkeof = blkroundup(fs, ip->i_size); 1796 1797 /* 1798 * Ignore requests to free pages past EOF but in the same block 1799 * as EOF, unless the request is synchronous. (If the request is 1800 * sync, it comes from lfs_truncate.) 1801 * XXXUBC Make these pages look "active" so the pagedaemon won't 1802 * XXXUBC bother us with them again. 1803 */ 1804 if (!sync && ap->a_offlo >= ip->i_size && ap->a_offlo < blkeof) { 1805 origoffset = ap->a_offlo; 1806 for (off = origoffset; off < blkeof; off += fs->lfs_bsize) { 1807 pg = uvm_pagelookup(&vp->v_uobj, off); 1808 KASSERT(pg != NULL); 1809 while (pg->flags & PG_BUSY) { 1810 pg->flags |= PG_WANTED; 1811 UVM_UNLOCK_AND_WAIT(pg, &vp->v_interlock, 0, 1812 "lfsput2", 0); 1813 simple_lock(&vp->v_interlock); 1814 } 1815 uvm_lock_pageq(); 1816 uvm_pageactivate(pg); 1817 uvm_unlock_pageq(); 1818 } 1819 ap->a_offlo = blkeof; 1820 if (ap->a_offhi > 0 && ap->a_offhi <= ap->a_offlo) { 1821 simple_unlock(&vp->v_interlock); 1822 return 0; 1823 } 1824 } 1825 1826 /* 1827 * Extend page range to start and end at block boundaries. 1828 * (For the purposes of VOP_PUTPAGES, fragments don't exist.) 1829 */ 1830 origoffset = ap->a_offlo; 1831 origendoffset = ap->a_offhi; 1832 startoffset = origoffset & ~(fs->lfs_bmask); 1833 max_endoffset = (trunc_page(LLONG_MAX) >> fs->lfs_bshift) 1834 << fs->lfs_bshift; 1835 1836 if (origendoffset == 0 || ap->a_flags & PGO_ALLPAGES) { 1837 endoffset = max_endoffset; 1838 origendoffset = endoffset; 1839 } else { 1840 origendoffset = round_page(ap->a_offhi); 1841 endoffset = round_page(blkroundup(fs, origendoffset)); 1842 } 1843 1844 KASSERT(startoffset > 0 || endoffset >= startoffset); 1845 if (startoffset == endoffset) { 1846 /* Nothing to do, why were we called? */ 1847 simple_unlock(&vp->v_interlock); 1848 DLOG((DLOG_PAGE, "lfs_putpages: startoffset = endoffset = %" 1849 PRId64 "\n", startoffset)); 1850 return 0; 1851 } 1852 1853 ap->a_offlo = startoffset; 1854 ap->a_offhi = endoffset; 1855 1856 if (!(ap->a_flags & PGO_CLEANIT)) 1857 return genfs_putpages(v); 1858 1859 /* 1860 * If there are more than one page per block, we don't want 1861 * to get caught locking them backwards; so set PGO_BUSYFAIL 1862 * to avoid deadlocks. 1863 */ 1864 ap->a_flags |= PGO_BUSYFAIL; 1865 1866 do { 1867 int r; 1868 1869 /* If no pages are dirty, we can just use genfs_putpages. */ 1870 r = check_dirty(fs, vp, startoffset, endoffset, blkeof, 1871 ap->a_flags, 1); 1872 if (r < 0) { 1873 simple_unlock(&vp->v_interlock); 1874 return EDEADLK; 1875 } 1876 if (r > 0) 1877 break; 1878 1879 /* 1880 * Sometimes pages are dirtied between the time that 1881 * we check and the time we try to clean them. 1882 * Instruct lfs_gop_write to return EDEADLK in this case 1883 * so we can write them properly. 1884 */ 1885 ip->i_lfs_iflags |= LFSI_NO_GOP_WRITE; 1886 r = genfs_putpages(v); 1887 ip->i_lfs_iflags &= ~LFSI_NO_GOP_WRITE; 1888 if (r != EDEADLK) 1889 return r; 1890 1891 /* Start over. */ 1892 preempt(1); 1893 simple_lock(&vp->v_interlock); 1894 } while(1); 1895 1896 /* 1897 * Dirty and asked to clean. 1898 * 1899 * Pagedaemon can't actually write LFS pages; wake up 1900 * the writer to take care of that. The writer will 1901 * notice the pager inode queue and act on that. 1902 */ 1903 if (pagedaemon) { 1904 simple_lock(&fs->lfs_interlock); 1905 if (!(ip->i_flags & IN_PAGING)) { 1906 ip->i_flags |= IN_PAGING; 1907 TAILQ_INSERT_TAIL(&fs->lfs_pchainhd, ip, i_lfs_pchain); 1908 } 1909 simple_lock(&lfs_subsys_lock); 1910 wakeup(&lfs_writer_daemon); 1911 simple_unlock(&lfs_subsys_lock); 1912 simple_unlock(&fs->lfs_interlock); 1913 simple_unlock(&vp->v_interlock); 1914 preempt(1); 1915 return EWOULDBLOCK; 1916 } 1917 1918 /* 1919 * If this is a file created in a recent dirop, we can't flush its 1920 * inode until the dirop is complete. Drain dirops, then flush the 1921 * filesystem (taking care of any other pending dirops while we're 1922 * at it). 1923 */ 1924 if ((ap->a_flags & (PGO_CLEANIT|PGO_LOCKED)) == PGO_CLEANIT && 1925 (vp->v_flag & VDIROP)) { 1926 int locked; 1927 1928 DLOG((DLOG_PAGE, "lfs_putpages: flushing VDIROP\n")); 1929 locked = VOP_ISLOCKED(vp) && /* XXX */ 1930 vp->v_lock.lk_lockholder == curproc->p_pid; 1931 simple_unlock(&vp->v_interlock); 1932 lfs_writer_enter(fs, "ppdirop"); 1933 if (locked) 1934 VOP_UNLOCK(vp, 0); 1935 1936 simple_lock(&fs->lfs_interlock); 1937 lfs_flush_fs(fs, sync ? SEGM_SYNC : 0); 1938 simple_unlock(&fs->lfs_interlock); 1939 1940 simple_lock(&vp->v_interlock); 1941 if (locked) { 1942 VOP_LOCK(vp, LK_EXCLUSIVE | LK_INTERLOCK); 1943 simple_lock(&vp->v_interlock); 1944 } 1945 lfs_writer_leave(fs); 1946 1947 /* XXX the flush should have taken care of this one too! */ 1948 } 1949 1950 /* 1951 * This is it. We are going to write some pages. From here on 1952 * down it's all just mechanics. 1953 * 1954 * Don't let genfs_putpages wait; lfs_segunlock will wait for us. 1955 */ 1956 ap->a_flags &= ~PGO_SYNCIO; 1957 1958 /* 1959 * If we've already got the seglock, flush the node and return. 1960 * The FIP has already been set up for us by lfs_writefile, 1961 * and FIP cleanup and lfs_updatemeta will also be done there, 1962 * unless genfs_putpages returns EDEADLK; then we must flush 1963 * what we have, and correct FIP and segment header accounting. 1964 */ 1965 1966 seglocked = (ap->a_flags & PGO_LOCKED) != 0; 1967 if (!seglocked) { 1968 simple_unlock(&vp->v_interlock); 1969 /* 1970 * Take the seglock, because we are going to be writing pages. 1971 */ 1972 error = lfs_seglock(fs, SEGM_PROT | (sync ? SEGM_SYNC : 0)); 1973 if (error != 0) 1974 return error; 1975 simple_lock(&vp->v_interlock); 1976 } 1977 1978 /* 1979 * VOP_PUTPAGES should not be called while holding the seglock. 1980 * XXXUBC fix lfs_markv, or do this properly. 1981 */ 1982 #ifdef notyet 1983 KASSERT(fs->lfs_seglock == 1); 1984 #endif /* notyet */ 1985 1986 /* 1987 * We assume we're being called with sp->fip pointing at blank space. 1988 * Account for a new FIP in the segment header, and set sp->vp. 1989 * (This should duplicate the setup at the top of lfs_writefile().) 1990 */ 1991 sp = fs->lfs_sp; 1992 if (!seglocked) { 1993 if (sp->seg_bytes_left < fs->lfs_bsize || 1994 sp->sum_bytes_left < sizeof(struct finfo)) 1995 (void) lfs_writeseg(fs, fs->lfs_sp); 1996 1997 sp->sum_bytes_left -= FINFOSIZE; 1998 ++((SEGSUM *)(sp->segsum))->ss_nfinfo; 1999 } 2000 KASSERT(sp->vp == NULL); 2001 sp->vp = vp; 2002 2003 if (!seglocked) { 2004 if (vp->v_flag & VDIROP) 2005 ((SEGSUM *)(sp->segsum))->ss_flags |= (SS_DIROP|SS_CONT); 2006 } 2007 2008 sp->fip->fi_nblocks = 0; 2009 sp->fip->fi_ino = ip->i_number; 2010 sp->fip->fi_version = ip->i_gen; 2011 2012 /* 2013 * Loop through genfs_putpages until all pages are gathered. 2014 * genfs_putpages() drops the interlock, so reacquire it if necessary. 2015 * Whenever we lose the interlock we have to rerun check_dirty, as 2016 * well. 2017 */ 2018 again: 2019 if (check_dirty(fs, vp, startoffset, endoffset, blkeof, 2020 ap->a_flags, 0) < 0) { 2021 simple_unlock(&vp->v_interlock); 2022 sp->vp = NULL; 2023 if (!seglocked) 2024 lfs_segunlock(fs); 2025 return EDEADLK; 2026 } 2027 2028 error = genfs_putpages(v); 2029 if (error == EDEADLK || error == EAGAIN) { 2030 DLOG((DLOG_PAGE, "lfs_putpages: genfs_putpages returned" 2031 " EDEADLK [2] ino %d off %x (seg %d)\n", 2032 ip->i_number, fs->lfs_offset, 2033 dtosn(fs, fs->lfs_offset))); 2034 /* If nothing to write, short-circuit */ 2035 if (sp->cbpp - sp->bpp > 1) { 2036 /* Write gathered pages */ 2037 lfs_updatemeta(sp); 2038 (void) lfs_writeseg(fs, sp); 2039 2040 /* 2041 * Reinitialize brand new FIP and add us to it. 2042 * (This should duplicate the fixup in 2043 * lfs_gatherpages().) 2044 */ 2045 KASSERT(sp->vp == vp); 2046 sp->fip->fi_version = ip->i_gen; 2047 sp->fip->fi_ino = ip->i_number; 2048 /* Add us to the new segment summary. */ 2049 ++((SEGSUM *)(sp->segsum))->ss_nfinfo; 2050 sp->sum_bytes_left -= FINFOSIZE; 2051 } 2052 2053 /* Give the write a chance to complete */ 2054 preempt(1); 2055 2056 /* We've lost the interlock. Start over. */ 2057 if (error == EDEADLK) { 2058 simple_lock(&vp->v_interlock); 2059 goto again; 2060 } 2061 } 2062 2063 KASSERT(sp->vp == vp); 2064 if (!seglocked) { 2065 sp->vp = NULL; /* XXX lfs_gather below will set this */ 2066 2067 /* Write indirect blocks as well */ 2068 lfs_gather(fs, fs->lfs_sp, vp, lfs_match_indir); 2069 lfs_gather(fs, fs->lfs_sp, vp, lfs_match_dindir); 2070 lfs_gather(fs, fs->lfs_sp, vp, lfs_match_tindir); 2071 2072 KASSERT(sp->vp == NULL); 2073 sp->vp = vp; 2074 } 2075 2076 /* 2077 * Blocks are now gathered into a segment waiting to be written. 2078 * All that's left to do is update metadata, and write them. 2079 */ 2080 lfs_updatemeta(sp); 2081 KASSERT(sp->vp == vp); 2082 sp->vp = NULL; 2083 2084 if (seglocked) { 2085 /* we're called by lfs_writefile. */ 2086 return error; 2087 } 2088 2089 /* 2090 * Clean up FIP, since we're done writing this file. 2091 * This should duplicate cleanup at the end of lfs_writefile(). 2092 */ 2093 if (sp->fip->fi_nblocks != 0) { 2094 sp->fip = (FINFO*)((caddr_t)sp->fip + FINFOSIZE + 2095 sizeof(int32_t) * sp->fip->fi_nblocks); 2096 sp->start_lbp = &sp->fip->fi_blocks[0]; 2097 } else { 2098 sp->sum_bytes_left += FINFOSIZE; 2099 --((SEGSUM *)(sp->segsum))->ss_nfinfo; 2100 } 2101 lfs_writeseg(fs, fs->lfs_sp); 2102 2103 /* 2104 * Remove us from paging queue, since we've now written all our 2105 * pages. 2106 */ 2107 simple_lock(&fs->lfs_interlock); 2108 if (ip->i_flags & IN_PAGING) { 2109 ip->i_flags &= ~IN_PAGING; 2110 TAILQ_REMOVE(&fs->lfs_pchainhd, ip, i_lfs_pchain); 2111 } 2112 simple_unlock(&fs->lfs_interlock); 2113 2114 /* 2115 * XXX - with the malloc/copy writeseg, the pages are freed by now 2116 * even if we don't wait (e.g. if we hold a nested lock). This 2117 * will not be true if we stop using malloc/copy. 2118 */ 2119 KASSERT(fs->lfs_sp->seg_flags & SEGM_PROT); 2120 lfs_segunlock(fs); 2121 2122 /* 2123 * Wait for v_numoutput to drop to zero. The seglock should 2124 * take care of this, but there is a slight possibility that 2125 * aiodoned might not have got around to our buffers yet. 2126 */ 2127 if (sync) { 2128 s = splbio(); 2129 simple_lock(&global_v_numoutput_slock); 2130 while (vp->v_numoutput > 0) { 2131 DLOG((DLOG_PAGE, "lfs_putpages: ino %d sleeping on" 2132 " num %d\n", ip->i_number, vp->v_numoutput)); 2133 vp->v_flag |= VBWAIT; 2134 ltsleep(&vp->v_numoutput, PRIBIO + 1, "lfs_vn", 0, 2135 &global_v_numoutput_slock); 2136 } 2137 simple_unlock(&global_v_numoutput_slock); 2138 splx(s); 2139 } 2140 return error; 2141 } 2142 2143 /* 2144 * Return the last logical file offset that should be written for this file 2145 * if we're doing a write that ends at "size". If writing, we need to know 2146 * about sizes on disk, i.e. fragments if there are any; if reading, we need 2147 * to know about entire blocks. 2148 */ 2149 void 2150 lfs_gop_size(struct vnode *vp, off_t size, off_t *eobp, int flags) 2151 { 2152 struct inode *ip = VTOI(vp); 2153 struct lfs *fs = ip->i_lfs; 2154 daddr_t olbn, nlbn; 2155 2156 olbn = lblkno(fs, ip->i_size); 2157 nlbn = lblkno(fs, size); 2158 if (!(flags & GOP_SIZE_MEM) && nlbn < NDADDR && olbn <= nlbn) { 2159 *eobp = fragroundup(fs, size); 2160 } else { 2161 *eobp = blkroundup(fs, size); 2162 } 2163 } 2164 2165 #ifdef DEBUG 2166 void lfs_dump_vop(void *); 2167 2168 void 2169 lfs_dump_vop(void *v) 2170 { 2171 struct vop_putpages_args /* { 2172 struct vnode *a_vp; 2173 voff_t a_offlo; 2174 voff_t a_offhi; 2175 int a_flags; 2176 } */ *ap = v; 2177 2178 #ifdef DDB 2179 vfs_vnode_print(ap->a_vp, 0, printf); 2180 #endif 2181 lfs_dump_dinode(VTOI(ap->a_vp)->i_din.ffs1_din); 2182 } 2183 #endif 2184 2185 int 2186 lfs_mmap(void *v) 2187 { 2188 struct vop_mmap_args /* { 2189 const struct vnodeop_desc *a_desc; 2190 struct vnode *a_vp; 2191 int a_fflags; 2192 struct ucred *a_cred; 2193 struct lwp *a_l; 2194 } */ *ap = v; 2195 2196 if (VTOI(ap->a_vp)->i_number == LFS_IFILE_INUM) 2197 return EOPNOTSUPP; 2198 return ufs_mmap(v); 2199 } 2200
Indexes created Fri Sep 26 20:09:58 GMT 2025