1 /* $NetBSD: lfs_vfsops.c,v 1.317 2013/11/27 17:24:44 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2002, 2003, 2007, 2007 5 * The NetBSD Foundation, Inc. 6 * All rights reserved. 7 * 8 * This code is derived from software contributed to The NetBSD Foundation 9 * by Konrad E. Schroder <perseant@hhhh.org>. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 /*- 33 * Copyright (c) 1989, 1991, 1993, 1994 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)lfs_vfsops.c 8.20 (Berkeley) 6/10/95 61 */ 62 63 #include <sys/cdefs.h> 64 __KERNEL_RCSID(0, "$NetBSD: lfs_vfsops.c,v 1.317 2013/11/27 17:24:44 christos Exp $"); 65 66 #if defined(_KERNEL_OPT) 67 #include "opt_lfs.h" 68 #include "opt_quota.h" 69 #endif 70 71 #include <sys/param.h> 72 #include <sys/systm.h> 73 #include <sys/namei.h> 74 #include <sys/proc.h> 75 #include <sys/kernel.h> 76 #include <sys/vnode.h> 77 #include <sys/mount.h> 78 #include <sys/kthread.h> 79 #include <sys/buf.h> 80 #include <sys/device.h> 81 #include <sys/mbuf.h> 82 #include <sys/file.h> 83 #include <sys/disklabel.h> 84 #include <sys/ioctl.h> 85 #include <sys/errno.h> 86 #include <sys/malloc.h> 87 #include <sys/pool.h> 88 #include <sys/socket.h> 89 #include <sys/syslog.h> 90 #include <uvm/uvm_extern.h> 91 #include <sys/sysctl.h> 92 #include <sys/conf.h> 93 #include <sys/kauth.h> 94 #include <sys/module.h> 95 #include <sys/syscallvar.h> 96 #include <sys/syscall.h> 97 #include <sys/syscallargs.h> 98 99 #include <miscfs/specfs/specdev.h> 100 101 #include <ufs/lfs/ulfs_quotacommon.h> 102 #include <ufs/lfs/ulfs_inode.h> 103 #include <ufs/lfs/ulfsmount.h> 104 #include <ufs/lfs/ulfs_extern.h> 105 106 #include <uvm/uvm.h> 107 #include <uvm/uvm_stat.h> 108 #include <uvm/uvm_pager.h> 109 #include <uvm/uvm_pdaemon.h> 110 111 #include <ufs/lfs/lfs.h> 112 #include <ufs/lfs/lfs_kernel.h> 113 #include <ufs/lfs/lfs_extern.h> 114 115 #include <miscfs/genfs/genfs.h> 116 #include <miscfs/genfs/genfs_node.h> 117 118 MODULE(MODULE_CLASS_VFS, lfs, NULL); 119 120 static int lfs_gop_write(struct vnode *, struct vm_page **, int, int); 121 static int lfs_mountfs(struct vnode *, struct mount *, struct lwp *); 122 static int lfs_extattrctl(struct mount *, int, struct vnode *, int, 123 const char *); 124 125 static struct sysctllog *lfs_sysctl_log; 126 127 extern const struct vnodeopv_desc lfs_vnodeop_opv_desc; 128 extern const struct vnodeopv_desc lfs_specop_opv_desc; 129 extern const struct vnodeopv_desc lfs_fifoop_opv_desc; 130 131 pid_t lfs_writer_daemon = 0; 132 lwpid_t lfs_writer_lid = 0; 133 int lfs_do_flush = 0; 134 #ifdef LFS_KERNEL_RFW 135 int lfs_do_rfw = 0; 136 #endif 137 138 const struct vnodeopv_desc * const lfs_vnodeopv_descs[] = { 139 &lfs_vnodeop_opv_desc, 140 &lfs_specop_opv_desc, 141 &lfs_fifoop_opv_desc, 142 NULL, 143 }; 144 145 struct vfsops lfs_vfsops = { 146 MOUNT_LFS, 147 sizeof (struct ulfs_args), 148 lfs_mount, 149 ulfs_start, 150 lfs_unmount, 151 ulfs_root, 152 ulfs_quotactl, 153 lfs_statvfs, 154 lfs_sync, 155 lfs_vget, 156 lfs_fhtovp, 157 lfs_vptofh, 158 lfs_init, 159 lfs_reinit, 160 lfs_done, 161 lfs_mountroot, 162 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 163 lfs_extattrctl, 164 (void *)eopnotsupp, /* vfs_suspendctl */ 165 genfs_renamelock_enter, 166 genfs_renamelock_exit, 167 (void *)eopnotsupp, 168 lfs_vnodeopv_descs, 169 0, 170 { NULL, NULL }, 171 }; 172 173 const struct genfs_ops lfs_genfsops = { 174 .gop_size = lfs_gop_size, 175 .gop_alloc = ulfs_gop_alloc, 176 .gop_write = lfs_gop_write, 177 .gop_markupdate = ulfs_gop_markupdate, 178 }; 179 180 struct shortlong { 181 const char *sname; 182 const char *lname; 183 }; 184 185 static int 186 sysctl_lfs_dostats(SYSCTLFN_ARGS) 187 { 188 extern struct lfs_stats lfs_stats; 189 extern int lfs_dostats; 190 int error; 191 192 error = sysctl_lookup(SYSCTLFN_CALL(rnode)); 193 if (error || newp == NULL) 194 return (error); 195 196 if (lfs_dostats == 0) 197 memset(&lfs_stats, 0, sizeof(lfs_stats)); 198 199 return (0); 200 } 201 202 static void 203 lfs_sysctl_setup(struct sysctllog **clog) 204 { 205 int i; 206 extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead, 207 lfs_fs_pagetrip, lfs_ignore_lazy_sync; 208 #ifdef DEBUG 209 extern int lfs_debug_log_subsys[DLOG_MAX]; 210 struct shortlong dlog_names[DLOG_MAX] = { /* Must match lfs.h ! */ 211 { "rollforward", "Debug roll-forward code" }, 212 { "alloc", "Debug inode allocation and free list" }, 213 { "avail", "Debug space-available-now accounting" }, 214 { "flush", "Debug flush triggers" }, 215 { "lockedlist", "Debug locked list accounting" }, 216 { "vnode_verbose", "Verbose per-vnode-written debugging" }, 217 { "vnode", "Debug vnode use during segment write" }, 218 { "segment", "Debug segment writing" }, 219 { "seguse", "Debug segment used-bytes accounting" }, 220 { "cleaner", "Debug cleaning routines" }, 221 { "mount", "Debug mount/unmount routines" }, 222 { "pagecache", "Debug UBC interactions" }, 223 { "dirop", "Debug directory-operation accounting" }, 224 { "malloc", "Debug private malloc accounting" }, 225 }; 226 #endif /* DEBUG */ 227 struct shortlong stat_names[] = { /* Must match lfs.h! */ 228 { "segsused", "Number of new segments allocated" }, 229 { "psegwrites", "Number of partial-segment writes" }, 230 { "psyncwrites", "Number of synchronous partial-segment" 231 " writes" }, 232 { "pcleanwrites", "Number of partial-segment writes by the" 233 " cleaner" }, 234 { "blocktot", "Number of blocks written" }, 235 { "cleanblocks", "Number of blocks written by the cleaner" }, 236 { "ncheckpoints", "Number of checkpoints made" }, 237 { "nwrites", "Number of whole writes" }, 238 { "nsync_writes", "Number of synchronous writes" }, 239 { "wait_exceeded", "Number of times writer waited for" 240 " cleaner" }, 241 { "write_exceeded", "Number of times writer invoked flush" }, 242 { "flush_invoked", "Number of times flush was invoked" }, 243 { "vflush_invoked", "Number of time vflush was called" }, 244 { "clean_inlocked", "Number of vnodes skipped for VI_XLOCK" }, 245 { "clean_vnlocked", "Number of vnodes skipped for vget failure" }, 246 { "segs_reclaimed", "Number of segments reclaimed" }, 247 }; 248 249 sysctl_createv(clog, 0, NULL, NULL, 250 CTLFLAG_PERMANENT, 251 CTLTYPE_NODE, "vfs", NULL, 252 NULL, 0, NULL, 0, 253 CTL_VFS, CTL_EOL); 254 sysctl_createv(clog, 0, NULL, NULL, 255 CTLFLAG_PERMANENT, 256 CTLTYPE_NODE, "lfs", 257 SYSCTL_DESCR("Log-structured file system"), 258 NULL, 0, NULL, 0, 259 CTL_VFS, 5, CTL_EOL); 260 /* 261 * XXX the "5" above could be dynamic, thereby eliminating one 262 * more instance of the "number to vfs" mapping problem, but 263 * "5" is the order as taken from sys/mount.h 264 */ 265 266 sysctl_createv(clog, 0, NULL, NULL, 267 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 268 CTLTYPE_INT, "flushindir", NULL, 269 NULL, 0, &lfs_writeindir, 0, 270 CTL_VFS, 5, LFS_WRITEINDIR, CTL_EOL); 271 sysctl_createv(clog, 0, NULL, NULL, 272 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 273 CTLTYPE_INT, "clean_vnhead", NULL, 274 NULL, 0, &lfs_clean_vnhead, 0, 275 CTL_VFS, 5, LFS_CLEAN_VNHEAD, CTL_EOL); 276 sysctl_createv(clog, 0, NULL, NULL, 277 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 278 CTLTYPE_INT, "dostats", 279 SYSCTL_DESCR("Maintain statistics on LFS operations"), 280 sysctl_lfs_dostats, 0, &lfs_dostats, 0, 281 CTL_VFS, 5, LFS_DOSTATS, CTL_EOL); 282 sysctl_createv(clog, 0, NULL, NULL, 283 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 284 CTLTYPE_INT, "pagetrip", 285 SYSCTL_DESCR("How many dirty pages in fs triggers" 286 " a flush"), 287 NULL, 0, &lfs_fs_pagetrip, 0, 288 CTL_VFS, 5, LFS_FS_PAGETRIP, CTL_EOL); 289 sysctl_createv(clog, 0, NULL, NULL, 290 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 291 CTLTYPE_INT, "ignore_lazy_sync", 292 SYSCTL_DESCR("Lazy Sync is ignored entirely"), 293 NULL, 0, &lfs_ignore_lazy_sync, 0, 294 CTL_VFS, 5, LFS_IGNORE_LAZY_SYNC, CTL_EOL); 295 #ifdef LFS_KERNEL_RFW 296 sysctl_createv(clog, 0, NULL, NULL, 297 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 298 CTLTYPE_INT, "rfw", 299 SYSCTL_DESCR("Use in-kernel roll-forward on mount"), 300 NULL, 0, &lfs_do_rfw, 0, 301 CTL_VFS, 5, LFS_DO_RFW, CTL_EOL); 302 #endif 303 304 sysctl_createv(clog, 0, NULL, NULL, 305 CTLFLAG_PERMANENT, 306 CTLTYPE_NODE, "stats", 307 SYSCTL_DESCR("Debugging options"), 308 NULL, 0, NULL, 0, 309 CTL_VFS, 5, LFS_STATS, CTL_EOL); 310 for (i = 0; i < sizeof(struct lfs_stats) / sizeof(u_int); i++) { 311 sysctl_createv(clog, 0, NULL, NULL, 312 CTLFLAG_PERMANENT|CTLFLAG_READONLY, 313 CTLTYPE_INT, stat_names[i].sname, 314 SYSCTL_DESCR(stat_names[i].lname), 315 NULL, 0, &(((u_int *)&lfs_stats.segsused)[i]), 316 0, CTL_VFS, 5, LFS_STATS, i, CTL_EOL); 317 } 318 319 #ifdef DEBUG 320 sysctl_createv(clog, 0, NULL, NULL, 321 CTLFLAG_PERMANENT, 322 CTLTYPE_NODE, "debug", 323 SYSCTL_DESCR("Debugging options"), 324 NULL, 0, NULL, 0, 325 CTL_VFS, 5, LFS_DEBUGLOG, CTL_EOL); 326 for (i = 0; i < DLOG_MAX; i++) { 327 sysctl_createv(clog, 0, NULL, NULL, 328 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 329 CTLTYPE_INT, dlog_names[i].sname, 330 SYSCTL_DESCR(dlog_names[i].lname), 331 NULL, 0, &(lfs_debug_log_subsys[i]), 0, 332 CTL_VFS, 5, LFS_DEBUGLOG, i, CTL_EOL); 333 } 334 #endif 335 } 336 337 /* old cleaner syscall interface. see VOP_FCNTL() */ 338 static const struct syscall_package lfs_syscalls[] = { 339 { SYS_lfs_bmapv, 0, (sy_call_t *)sys_lfs_bmapv }, 340 { SYS_lfs_markv, 0, (sy_call_t *)sys_lfs_markv }, 341 { SYS_lfs_segclean, 0, (sy_call_t *)sys___lfs_segwait50 }, 342 { 0, 0, NULL }, 343 }; 344 345 static int 346 lfs_modcmd(modcmd_t cmd, void *arg) 347 { 348 int error; 349 350 switch (cmd) { 351 case MODULE_CMD_INIT: 352 error = syscall_establish(NULL, lfs_syscalls); 353 if (error) 354 return error; 355 error = vfs_attach(&lfs_vfsops); 356 if (error != 0) { 357 syscall_disestablish(NULL, lfs_syscalls); 358 break; 359 } 360 lfs_sysctl_setup(&lfs_sysctl_log); 361 break; 362 case MODULE_CMD_FINI: 363 error = vfs_detach(&lfs_vfsops); 364 if (error != 0) 365 break; 366 syscall_disestablish(NULL, lfs_syscalls); 367 sysctl_teardown(&lfs_sysctl_log); 368 break; 369 default: 370 error = ENOTTY; 371 break; 372 } 373 374 return (error); 375 } 376 377 /* 378 * XXX Same structure as FFS inodes? Should we share a common pool? 379 */ 380 struct pool lfs_inode_pool; 381 struct pool lfs_dinode_pool; 382 struct pool lfs_inoext_pool; 383 struct pool lfs_lbnentry_pool; 384 385 /* 386 * The writer daemon. UVM keeps track of how many dirty pages we are holding 387 * in lfs_subsys_pages; the daemon flushes the filesystem when this value 388 * crosses the (user-defined) threshhold LFS_MAX_PAGES. 389 */ 390 static void 391 lfs_writerd(void *arg) 392 { 393 struct mount *mp, *nmp; 394 struct lfs *fs; 395 struct vfsops *vfs = NULL; 396 int fsflags; 397 int skipc; 398 int lfsc; 399 int wrote_something = 0; 400 401 mutex_enter(&lfs_lock); 402 lfs_writer_daemon = curproc->p_pid; 403 lfs_writer_lid = curlwp->l_lid; 404 mutex_exit(&lfs_lock); 405 406 /* Take an extra reference to the LFS vfsops. */ 407 vfs = vfs_getopsbyname(MOUNT_LFS); 408 409 mutex_enter(&lfs_lock); 410 for (;;) { 411 KASSERT(mutex_owned(&lfs_lock)); 412 if (wrote_something == 0) 413 mtsleep(&lfs_writer_daemon, PVM, "lfswriter", hz/10 + 1, 414 &lfs_lock); 415 416 KASSERT(mutex_owned(&lfs_lock)); 417 wrote_something = 0; 418 419 /* 420 * If global state wants a flush, flush everything. 421 */ 422 if (lfs_do_flush || locked_queue_count > LFS_MAX_BUFS || 423 locked_queue_bytes > LFS_MAX_BYTES || 424 lfs_subsys_pages > LFS_MAX_PAGES) { 425 426 if (lfs_do_flush) { 427 DLOG((DLOG_FLUSH, "lfs_writerd: lfs_do_flush\n")); 428 } 429 if (locked_queue_count > LFS_MAX_BUFS) { 430 DLOG((DLOG_FLUSH, "lfs_writerd: lqc = %d, max %d\n", 431 locked_queue_count, LFS_MAX_BUFS)); 432 } 433 if (locked_queue_bytes > LFS_MAX_BYTES) { 434 DLOG((DLOG_FLUSH, "lfs_writerd: lqb = %ld, max %ld\n", 435 locked_queue_bytes, LFS_MAX_BYTES)); 436 } 437 if (lfs_subsys_pages > LFS_MAX_PAGES) { 438 DLOG((DLOG_FLUSH, "lfs_writerd: lssp = %d, max %d\n", 439 lfs_subsys_pages, LFS_MAX_PAGES)); 440 } 441 442 lfs_flush(NULL, SEGM_WRITERD, 0); 443 lfs_do_flush = 0; 444 KASSERT(mutex_owned(&lfs_lock)); 445 continue; 446 } 447 KASSERT(mutex_owned(&lfs_lock)); 448 mutex_exit(&lfs_lock); 449 450 /* 451 * Look through the list of LFSs to see if any of them 452 * have requested pageouts. 453 */ 454 mutex_enter(&mountlist_lock); 455 lfsc = 0; 456 skipc = 0; 457 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 458 if (vfs_busy(mp, &nmp)) { 459 ++skipc; 460 continue; 461 } 462 KASSERT(!mutex_owned(&lfs_lock)); 463 if (strncmp(mp->mnt_stat.f_fstypename, MOUNT_LFS, 464 sizeof(mp->mnt_stat.f_fstypename)) == 0) { 465 ++lfsc; 466 fs = VFSTOULFS(mp)->um_lfs; 467 int32_t ooffset = 0; 468 fsflags = SEGM_SINGLE; 469 470 mutex_enter(&lfs_lock); 471 ooffset = fs->lfs_offset; 472 473 if (fs->lfs_nextseg < fs->lfs_curseg && fs->lfs_nowrap) { 474 /* Don't try to write if we're suspended */ 475 mutex_exit(&lfs_lock); 476 vfs_unbusy(mp, false, &nmp); 477 continue; 478 } 479 if (LFS_STARVED_FOR_SEGS(fs)) { 480 mutex_exit(&lfs_lock); 481 482 DLOG((DLOG_FLUSH, "lfs_writerd: need cleaning before writing possible\n")); 483 lfs_wakeup_cleaner(fs); 484 vfs_unbusy(mp, false, &nmp); 485 continue; 486 } 487 488 if ((fs->lfs_dirvcount > LFS_MAX_FSDIROP(fs) || 489 lfs_dirvcount > LFS_MAX_DIROP) && 490 fs->lfs_dirops == 0) { 491 fsflags &= ~SEGM_SINGLE; 492 fsflags |= SEGM_CKP; 493 DLOG((DLOG_FLUSH, "lfs_writerd: checkpoint\n")); 494 lfs_flush_fs(fs, fsflags); 495 } else if (fs->lfs_pdflush) { 496 DLOG((DLOG_FLUSH, "lfs_writerd: pdflush set\n")); 497 lfs_flush_fs(fs, fsflags); 498 } else if (!TAILQ_EMPTY(&fs->lfs_pchainhd)) { 499 DLOG((DLOG_FLUSH, "lfs_writerd: pchain non-empty\n")); 500 mutex_exit(&lfs_lock); 501 lfs_writer_enter(fs, "wrdirop"); 502 lfs_flush_pchain(fs); 503 lfs_writer_leave(fs); 504 mutex_enter(&lfs_lock); 505 } 506 if (fs->lfs_offset != ooffset) 507 ++wrote_something; 508 mutex_exit(&lfs_lock); 509 } 510 KASSERT(!mutex_owned(&lfs_lock)); 511 vfs_unbusy(mp, false, &nmp); 512 } 513 if (lfsc + skipc == 0) { 514 mutex_enter(&lfs_lock); 515 lfs_writer_daemon = 0; 516 lfs_writer_lid = 0; 517 mutex_exit(&lfs_lock); 518 mutex_exit(&mountlist_lock); 519 break; 520 } 521 mutex_exit(&mountlist_lock); 522 523 mutex_enter(&lfs_lock); 524 } 525 KASSERT(!mutex_owned(&lfs_lock)); 526 KASSERT(!mutex_owned(&mountlist_lock)); 527 528 /* Give up our extra reference so the module can be unloaded. */ 529 mutex_enter(&vfs_list_lock); 530 if (vfs != NULL) 531 vfs->vfs_refcount--; 532 mutex_exit(&vfs_list_lock); 533 534 /* Done! */ 535 kthread_exit(0); 536 } 537 538 /* 539 * Initialize the filesystem, most work done by ulfs_init. 540 */ 541 void 542 lfs_init(void) 543 { 544 545 malloc_type_attach(M_SEGMENT); 546 pool_init(&lfs_inode_pool, sizeof(struct inode), 0, 0, 0, 547 "lfsinopl", &pool_allocator_nointr, IPL_NONE); 548 pool_init(&lfs_dinode_pool, sizeof(struct ulfs1_dinode), 0, 0, 0, 549 "lfsdinopl", &pool_allocator_nointr, IPL_NONE); 550 pool_init(&lfs_inoext_pool, sizeof(struct lfs_inode_ext), 8, 0, 0, 551 "lfsinoextpl", &pool_allocator_nointr, IPL_NONE); 552 pool_init(&lfs_lbnentry_pool, sizeof(struct lbnentry), 0, 0, 0, 553 "lfslbnpool", &pool_allocator_nointr, IPL_NONE); 554 ulfs_init(); 555 556 #ifdef DEBUG 557 memset(lfs_log, 0, sizeof(lfs_log)); 558 #endif 559 mutex_init(&lfs_lock, MUTEX_DEFAULT, IPL_NONE); 560 cv_init(&locked_queue_cv, "lfsbuf"); 561 cv_init(&lfs_writing_cv, "lfsflush"); 562 } 563 564 void 565 lfs_reinit(void) 566 { 567 ulfs_reinit(); 568 } 569 570 void 571 lfs_done(void) 572 { 573 ulfs_done(); 574 mutex_destroy(&lfs_lock); 575 cv_destroy(&locked_queue_cv); 576 cv_destroy(&lfs_writing_cv); 577 pool_destroy(&lfs_inode_pool); 578 pool_destroy(&lfs_dinode_pool); 579 pool_destroy(&lfs_inoext_pool); 580 pool_destroy(&lfs_lbnentry_pool); 581 malloc_type_detach(M_SEGMENT); 582 } 583 584 /* 585 * Called by main() when ulfs is going to be mounted as root. 586 */ 587 int 588 lfs_mountroot(void) 589 { 590 extern struct vnode *rootvp; 591 struct lfs *fs = NULL; /* LFS */ 592 struct mount *mp; 593 struct lwp *l = curlwp; 594 struct ulfsmount *ump; 595 int error; 596 597 if (device_class(root_device) != DV_DISK) 598 return (ENODEV); 599 600 if (rootdev == NODEV) 601 return (ENODEV); 602 if ((error = vfs_rootmountalloc(MOUNT_LFS, "root_device", &mp))) { 603 vrele(rootvp); 604 return (error); 605 } 606 if ((error = lfs_mountfs(rootvp, mp, l))) { 607 vfs_unbusy(mp, false, NULL); 608 vfs_destroy(mp); 609 return (error); 610 } 611 mountlist_append(mp); 612 ump = VFSTOULFS(mp); 613 fs = ump->um_lfs; 614 memset(fs->lfs_fsmnt, 0, sizeof(fs->lfs_fsmnt)); 615 (void)copystr(mp->mnt_stat.f_mntonname, fs->lfs_fsmnt, MNAMELEN - 1, 0); 616 (void)lfs_statvfs(mp, &mp->mnt_stat); 617 vfs_unbusy(mp, false, NULL); 618 setrootfstime((time_t)(VFSTOULFS(mp)->um_lfs->lfs_tstamp)); 619 return (0); 620 } 621 622 /* 623 * VFS Operations. 624 * 625 * mount system call 626 */ 627 int 628 lfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 629 { 630 struct lwp *l = curlwp; 631 struct vnode *devvp; 632 struct ulfs_args *args = data; 633 struct ulfsmount *ump = NULL; 634 struct lfs *fs = NULL; /* LFS */ 635 int error = 0, update; 636 mode_t accessmode; 637 638 if (*data_len < sizeof *args) 639 return EINVAL; 640 641 if (mp->mnt_flag & MNT_GETARGS) { 642 ump = VFSTOULFS(mp); 643 if (ump == NULL) 644 return EIO; 645 args->fspec = NULL; 646 *data_len = sizeof *args; 647 return 0; 648 } 649 650 update = mp->mnt_flag & MNT_UPDATE; 651 652 /* Check arguments */ 653 if (args->fspec != NULL) { 654 /* 655 * Look up the name and verify that it's sane. 656 */ 657 error = namei_simple_user(args->fspec, 658 NSM_FOLLOW_NOEMULROOT, &devvp); 659 if (error != 0) 660 return (error); 661 662 if (!update) { 663 /* 664 * Be sure this is a valid block device 665 */ 666 if (devvp->v_type != VBLK) 667 error = ENOTBLK; 668 else if (bdevsw_lookup(devvp->v_rdev) == NULL) 669 error = ENXIO; 670 } else { 671 /* 672 * Be sure we're still naming the same device 673 * used for our initial mount 674 */ 675 ump = VFSTOULFS(mp); 676 if (devvp != ump->um_devvp) { 677 if (devvp->v_rdev != ump->um_devvp->v_rdev) 678 error = EINVAL; 679 else { 680 vrele(devvp); 681 devvp = ump->um_devvp; 682 vref(devvp); 683 } 684 } 685 } 686 } else { 687 if (!update) { 688 /* New mounts must have a filename for the device */ 689 return (EINVAL); 690 } else { 691 /* Use the extant mount */ 692 ump = VFSTOULFS(mp); 693 devvp = ump->um_devvp; 694 vref(devvp); 695 } 696 } 697 698 699 /* 700 * If mount by non-root, then verify that user has necessary 701 * permissions on the device. 702 */ 703 if (error == 0) { 704 accessmode = VREAD; 705 if (update ? 706 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 : 707 (mp->mnt_flag & MNT_RDONLY) == 0) 708 accessmode |= VWRITE; 709 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 710 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 711 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, 712 KAUTH_ARG(accessmode)); 713 VOP_UNLOCK(devvp); 714 } 715 716 if (error) { 717 vrele(devvp); 718 return (error); 719 } 720 721 if (!update) { 722 int flags; 723 724 if (mp->mnt_flag & MNT_RDONLY) 725 flags = FREAD; 726 else 727 flags = FREAD|FWRITE; 728 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 729 error = VOP_OPEN(devvp, flags, FSCRED); 730 VOP_UNLOCK(devvp); 731 if (error) 732 goto fail; 733 error = lfs_mountfs(devvp, mp, l); /* LFS */ 734 if (error) { 735 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 736 (void)VOP_CLOSE(devvp, flags, NOCRED); 737 VOP_UNLOCK(devvp); 738 goto fail; 739 } 740 741 ump = VFSTOULFS(mp); 742 fs = ump->um_lfs; 743 } else { 744 /* 745 * Update the mount. 746 */ 747 748 /* 749 * The initial mount got a reference on this 750 * device, so drop the one obtained via 751 * namei(), above. 752 */ 753 vrele(devvp); 754 755 ump = VFSTOULFS(mp); 756 fs = ump->um_lfs; 757 758 if (fs->lfs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 759 /* 760 * Changing from read/write to read-only. 761 * XXX: shouldn't we sync here? or does vfs do that? 762 */ 763 #ifdef LFS_QUOTA2 764 /* XXX: quotas should remain on when readonly */ 765 if (fs->lfs_use_quota2) { 766 error = lfsquota2_umount(mp, 0); 767 if (error) { 768 return error; 769 } 770 } 771 #endif 772 } 773 774 if (fs->lfs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) { 775 /* 776 * Changing from read-only to read/write. 777 * Note in the superblocks that we're writing. 778 */ 779 780 /* XXX: quotas should have been on even if readonly */ 781 if (fs->lfs_use_quota2) { 782 #ifdef LFS_QUOTA2 783 error = lfs_quota2_mount(mp); 784 #else 785 uprintf("%s: no kernel support for this " 786 "filesystem's quotas\n", 787 mp->mnt_stat.f_mntonname); 788 if (mp->mnt_flag & MNT_FORCE) { 789 uprintf("%s: mounting anyway; " 790 "fsck afterwards\n", 791 mp->mnt_stat.f_mntonname); 792 } else { 793 error = EINVAL; 794 } 795 #endif 796 if (error) { 797 return error; 798 } 799 } 800 801 fs->lfs_ronly = 0; 802 if (fs->lfs_pflags & LFS_PF_CLEAN) { 803 fs->lfs_pflags &= ~LFS_PF_CLEAN; 804 lfs_writesuper(fs, fs->lfs_sboffs[0]); 805 lfs_writesuper(fs, fs->lfs_sboffs[1]); 806 } 807 } 808 if (args->fspec == NULL) 809 return EINVAL; 810 } 811 812 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 813 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 814 if (error == 0) 815 (void)strncpy(fs->lfs_fsmnt, mp->mnt_stat.f_mntonname, 816 sizeof(fs->lfs_fsmnt)); 817 return error; 818 819 fail: 820 vrele(devvp); 821 return (error); 822 } 823 824 825 /* 826 * Common code for mount and mountroot 827 * LFS specific 828 */ 829 int 830 lfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l) 831 { 832 struct dlfs *tdfs, *dfs, *adfs; 833 struct lfs *fs; 834 struct ulfsmount *ump; 835 struct vnode *vp; 836 struct buf *bp, *abp; 837 dev_t dev; 838 int error, i, ronly, fsbsize; 839 kauth_cred_t cred; 840 CLEANERINFO *cip; 841 SEGUSE *sup; 842 daddr_t sb_addr; 843 844 cred = l ? l->l_cred : NOCRED; 845 846 /* 847 * Flush out any old buffers remaining from a previous use. 848 */ 849 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 850 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0); 851 VOP_UNLOCK(devvp); 852 if (error) 853 return (error); 854 855 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 856 857 /* Don't free random space on error. */ 858 bp = NULL; 859 abp = NULL; 860 ump = NULL; 861 862 sb_addr = LFS_LABELPAD / DEV_BSIZE; 863 while (1) { 864 /* Read in the superblock. */ 865 error = bread(devvp, sb_addr, LFS_SBPAD, cred, 0, &bp); 866 if (error) 867 goto out; 868 dfs = (struct dlfs *)bp->b_data; 869 870 /* Check the basics. */ 871 if (dfs->dlfs_magic != LFS_MAGIC || dfs->dlfs_bsize > MAXBSIZE || 872 dfs->dlfs_version > LFS_VERSION || 873 dfs->dlfs_bsize < sizeof(struct dlfs)) { 874 DLOG((DLOG_MOUNT, "lfs_mountfs: primary superblock sanity failed\n")); 875 error = EINVAL; /* XXX needs translation */ 876 goto out; 877 } 878 if (dfs->dlfs_inodefmt > LFS_MAXINODEFMT) { 879 DLOG((DLOG_MOUNT, "lfs_mountfs: unknown inode format %d\n", 880 dfs->dlfs_inodefmt)); 881 error = EINVAL; 882 goto out; 883 } 884 885 if (dfs->dlfs_version == 1) 886 fsbsize = DEV_BSIZE; 887 else { 888 fsbsize = 1 << dfs->dlfs_ffshift; 889 /* 890 * Could be, if the frag size is large enough, that we 891 * don't have the "real" primary superblock. If that's 892 * the case, get the real one, and try again. 893 */ 894 if (sb_addr != (dfs->dlfs_sboffs[0] << (dfs->dlfs_ffshift - DEV_BSHIFT))) { 895 DLOG((DLOG_MOUNT, "lfs_mountfs: sb daddr" 896 " 0x%llx is not right, trying 0x%llx\n", 897 (long long)sb_addr, 898 (long long)(dfs->dlfs_sboffs[0] << (dfs->dlfs_ffshift - DEV_BSHIFT)))); 899 sb_addr = dfs->dlfs_sboffs[0] << (dfs->dlfs_ffshift - DEV_BSHIFT); 900 brelse(bp, 0); 901 continue; 902 } 903 } 904 break; 905 } 906 907 /* 908 * Check the second superblock to see which is newer; then mount 909 * using the older of the two. This is necessary to ensure that 910 * the filesystem is valid if it was not unmounted cleanly. 911 */ 912 913 if (dfs->dlfs_sboffs[1] && 914 dfs->dlfs_sboffs[1] - LFS_LABELPAD / fsbsize > LFS_SBPAD / fsbsize) 915 { 916 error = bread(devvp, dfs->dlfs_sboffs[1] * (fsbsize / DEV_BSIZE), 917 LFS_SBPAD, cred, 0, &abp); 918 if (error) 919 goto out; 920 adfs = (struct dlfs *)abp->b_data; 921 922 if (dfs->dlfs_version == 1) { 923 /* 1s resolution comparison */ 924 if (adfs->dlfs_tstamp < dfs->dlfs_tstamp) 925 tdfs = adfs; 926 else 927 tdfs = dfs; 928 } else { 929 /* monotonic infinite-resolution comparison */ 930 if (adfs->dlfs_serial < dfs->dlfs_serial) 931 tdfs = adfs; 932 else 933 tdfs = dfs; 934 } 935 936 /* Check the basics. */ 937 if (tdfs->dlfs_magic != LFS_MAGIC || 938 tdfs->dlfs_bsize > MAXBSIZE || 939 tdfs->dlfs_version > LFS_VERSION || 940 tdfs->dlfs_bsize < sizeof(struct dlfs)) { 941 DLOG((DLOG_MOUNT, "lfs_mountfs: alt superblock" 942 " sanity failed\n")); 943 error = EINVAL; /* XXX needs translation */ 944 goto out; 945 } 946 } else { 947 DLOG((DLOG_MOUNT, "lfs_mountfs: invalid alt superblock" 948 " daddr=0x%x\n", dfs->dlfs_sboffs[1])); 949 error = EINVAL; 950 goto out; 951 } 952 953 /* Allocate the mount structure, copy the superblock into it. */ 954 fs = kmem_zalloc(sizeof(struct lfs), KM_SLEEP); 955 memcpy(&fs->lfs_dlfs, tdfs, sizeof(struct dlfs)); 956 957 /* Compatibility */ 958 if (fs->lfs_version < 2) { 959 fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE; 960 fs->lfs_ibsize = fs->lfs_bsize; 961 fs->lfs_start = fs->lfs_sboffs[0]; 962 fs->lfs_tstamp = fs->lfs_otstamp; 963 fs->lfs_fsbtodb = 0; 964 } 965 if (fs->lfs_resvseg == 0) 966 fs->lfs_resvseg = MIN(fs->lfs_minfreeseg - 1, \ 967 MAX(MIN_RESV_SEGS, fs->lfs_minfreeseg / 2 + 1)); 968 969 /* 970 * If we aren't going to be able to write meaningfully to this 971 * filesystem, and were not mounted readonly, bomb out now. 972 */ 973 if (lfs_fsbtob(fs, LFS_NRESERVE(fs)) > LFS_MAX_BYTES && !ronly) { 974 DLOG((DLOG_MOUNT, "lfs_mount: to mount this filesystem read/write," 975 " we need BUFPAGES >= %lld\n", 976 (long long)((bufmem_hiwater / bufmem_lowater) * 977 LFS_INVERSE_MAX_BYTES( 978 lfs_fsbtob(fs, LFS_NRESERVE(fs))) >> PAGE_SHIFT))); 979 kmem_free(fs, sizeof(struct lfs)); 980 error = EFBIG; /* XXX needs translation */ 981 goto out; 982 } 983 984 /* Before rolling forward, lock so vget will sleep for other procs */ 985 if (l != NULL) { 986 fs->lfs_flags = LFS_NOTYET; 987 fs->lfs_rfpid = l->l_proc->p_pid; 988 } 989 990 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP); 991 ump->um_lfs = fs; 992 ump->um_fstype = ULFS1; 993 if (sizeof(struct lfs) < LFS_SBPAD) { /* XXX why? */ 994 brelse(bp, BC_INVAL); 995 brelse(abp, BC_INVAL); 996 } else { 997 brelse(bp, 0); 998 brelse(abp, 0); 999 } 1000 bp = NULL; 1001 abp = NULL; 1002 1003 1004 /* Set up the I/O information */ 1005 fs->lfs_devbsize = DEV_BSIZE; 1006 fs->lfs_iocount = 0; 1007 fs->lfs_diropwait = 0; 1008 fs->lfs_activesb = 0; 1009 fs->lfs_uinodes = 0; 1010 fs->lfs_ravail = 0; 1011 fs->lfs_favail = 0; 1012 fs->lfs_sbactive = 0; 1013 1014 /* Set up the ifile and lock aflags */ 1015 fs->lfs_doifile = 0; 1016 fs->lfs_writer = 0; 1017 fs->lfs_dirops = 0; 1018 fs->lfs_nadirop = 0; 1019 fs->lfs_seglock = 0; 1020 fs->lfs_pdflush = 0; 1021 fs->lfs_sleepers = 0; 1022 fs->lfs_pages = 0; 1023 rw_init(&fs->lfs_fraglock); 1024 rw_init(&fs->lfs_iflock); 1025 cv_init(&fs->lfs_stopcv, "lfsstop"); 1026 1027 /* Set the file system readonly/modify bits. */ 1028 fs->lfs_ronly = ronly; 1029 if (ronly == 0) 1030 fs->lfs_fmod = 1; 1031 1032 /* ulfs-level information */ 1033 fs->um_flags = 0; 1034 fs->um_bptrtodb = fs->lfs_ffshift - DEV_BSHIFT; 1035 fs->um_seqinc = fs->lfs_frag; 1036 fs->um_nindir = fs->lfs_nindir; 1037 fs->um_lognindir = ffs(fs->lfs_nindir) - 1; 1038 fs->um_maxsymlinklen = fs->lfs_maxsymlinklen; 1039 fs->um_dirblksiz = LFS_DIRBLKSIZ; 1040 fs->um_maxfilesize = fs->lfs_maxfilesize; 1041 1042 /* quota stuff */ 1043 /* XXX: these need to come from the on-disk superblock to be used */ 1044 fs->lfs_use_quota2 = 0; 1045 fs->lfs_quota_magic = 0; 1046 fs->lfs_quota_flags = 0; 1047 fs->lfs_quotaino[0] = 0; 1048 fs->lfs_quotaino[1] = 0; 1049 1050 /* Initialize the mount structure. */ 1051 dev = devvp->v_rdev; 1052 mp->mnt_data = ump; 1053 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev; 1054 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_LFS); 1055 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 1056 mp->mnt_stat.f_namemax = LFS_MAXNAMLEN; 1057 mp->mnt_stat.f_iosize = fs->lfs_bsize; 1058 mp->mnt_flag |= MNT_LOCAL; 1059 mp->mnt_fs_bshift = fs->lfs_bshift; 1060 if (fs->um_maxsymlinklen > 0) 1061 mp->mnt_iflag |= IMNT_DTYPE; 1062 1063 ump->um_mountp = mp; 1064 ump->um_dev = dev; 1065 ump->um_devvp = devvp; 1066 for (i = 0; i < ULFS_MAXQUOTAS; i++) 1067 ump->um_quotas[i] = NULLVP; 1068 spec_node_setmountedfs(devvp, mp); 1069 1070 /* Set up reserved memory for pageout */ 1071 lfs_setup_resblks(fs); 1072 /* Set up vdirop tailq */ 1073 TAILQ_INIT(&fs->lfs_dchainhd); 1074 /* and paging tailq */ 1075 TAILQ_INIT(&fs->lfs_pchainhd); 1076 /* and delayed segment accounting for truncation list */ 1077 LIST_INIT(&fs->lfs_segdhd); 1078 1079 /* 1080 * We use the ifile vnode for almost every operation. Instead of 1081 * retrieving it from the hash table each time we retrieve it here, 1082 * artificially increment the reference count and keep a pointer 1083 * to it in the incore copy of the superblock. 1084 */ 1085 if ((error = VFS_VGET(mp, LFS_IFILE_INUM, &vp)) != 0) { 1086 DLOG((DLOG_MOUNT, "lfs_mountfs: ifile vget failed, error=%d\n", error)); 1087 goto out; 1088 } 1089 fs->lfs_ivnode = vp; 1090 vref(vp); 1091 1092 /* Set up inode bitmap and order free list */ 1093 lfs_order_freelist(fs); 1094 1095 /* Set up segment usage flags for the autocleaner. */ 1096 fs->lfs_nactive = 0; 1097 fs->lfs_suflags = (u_int32_t **)malloc(2 * sizeof(u_int32_t *), 1098 M_SEGMENT, M_WAITOK); 1099 fs->lfs_suflags[0] = (u_int32_t *)malloc(fs->lfs_nseg * sizeof(u_int32_t), 1100 M_SEGMENT, M_WAITOK); 1101 fs->lfs_suflags[1] = (u_int32_t *)malloc(fs->lfs_nseg * sizeof(u_int32_t), 1102 M_SEGMENT, M_WAITOK); 1103 memset(fs->lfs_suflags[1], 0, fs->lfs_nseg * sizeof(u_int32_t)); 1104 for (i = 0; i < fs->lfs_nseg; i++) { 1105 int changed; 1106 1107 LFS_SEGENTRY(sup, fs, i, bp); 1108 changed = 0; 1109 if (!ronly) { 1110 if (sup->su_nbytes == 0 && 1111 !(sup->su_flags & SEGUSE_EMPTY)) { 1112 sup->su_flags |= SEGUSE_EMPTY; 1113 ++changed; 1114 } else if (!(sup->su_nbytes == 0) && 1115 (sup->su_flags & SEGUSE_EMPTY)) { 1116 sup->su_flags &= ~SEGUSE_EMPTY; 1117 ++changed; 1118 } 1119 if (sup->su_flags & (SEGUSE_ACTIVE|SEGUSE_INVAL)) { 1120 sup->su_flags &= ~(SEGUSE_ACTIVE|SEGUSE_INVAL); 1121 ++changed; 1122 } 1123 } 1124 fs->lfs_suflags[0][i] = sup->su_flags; 1125 if (changed) 1126 LFS_WRITESEGENTRY(sup, fs, i, bp); 1127 else 1128 brelse(bp, 0); 1129 } 1130 1131 /* 1132 * XXX: if the fs has quotas, quotas should be on even if 1133 * readonly. Otherwise you can't query the quota info! 1134 * However, that's not how the quota2 code got written and I 1135 * don't know if it'll behave itself if enabled while 1136 * readonly, so for now use the same enable logic as ffs. 1137 * 1138 * XXX: also, if you use the -f behavior allowed here (and 1139 * equivalently above for remount) it will corrupt the fs. It 1140 * ought not to allow that. It should allow mounting readonly 1141 * if there are quotas and the kernel doesn't have the quota 1142 * code, but only readonly. 1143 * 1144 * XXX: and if you use the -f behavior allowed here it will 1145 * likely crash at unmount time (or remount time) because we 1146 * think quotas are active. 1147 * 1148 * Although none of this applies until there's a way to set 1149 * lfs_use_quota2 and have quotas in the fs at all. 1150 */ 1151 if (!ronly && fs->lfs_use_quota2) { 1152 #ifdef LFS_QUOTA2 1153 error = lfs_quota2_mount(mp); 1154 #else 1155 uprintf("%s: no kernel support for this filesystem's quotas\n", 1156 mp->mnt_stat.f_mntonname); 1157 if (mp->mnt_flag & MNT_FORCE) { 1158 uprintf("%s: mounting anyway; fsck afterwards\n", 1159 mp->mnt_stat.f_mntonname); 1160 } else { 1161 error = EINVAL; 1162 } 1163 #endif 1164 if (error) { 1165 /* XXX XXX must clean up the stuff immediately above */ 1166 printf("lfs_mountfs: sorry, leaking some memory\n"); 1167 goto out; 1168 } 1169 } 1170 1171 #ifdef LFS_EXTATTR 1172 /* 1173 * Initialize file-backed extended attributes for ULFS1 file 1174 * systems. 1175 * 1176 * XXX: why is this limited to ULFS1? 1177 */ 1178 if (ump->um_fstype == ULFS1) { 1179 ulfs_extattr_uepm_init(&ump->um_extattr); 1180 } 1181 #endif 1182 1183 #ifdef LFS_KERNEL_RFW 1184 lfs_roll_forward(fs, mp, l); 1185 #endif 1186 1187 /* If writing, sb is not clean; record in case of immediate crash */ 1188 if (!fs->lfs_ronly) { 1189 fs->lfs_pflags &= ~LFS_PF_CLEAN; 1190 lfs_writesuper(fs, fs->lfs_sboffs[0]); 1191 lfs_writesuper(fs, fs->lfs_sboffs[1]); 1192 } 1193 1194 /* Allow vget now that roll-forward is complete */ 1195 fs->lfs_flags &= ~(LFS_NOTYET); 1196 wakeup(&fs->lfs_flags); 1197 1198 /* 1199 * Initialize the ifile cleaner info with information from 1200 * the superblock. 1201 */ 1202 LFS_CLEANERINFO(cip, fs, bp); 1203 cip->clean = fs->lfs_nclean; 1204 cip->dirty = fs->lfs_nseg - fs->lfs_nclean; 1205 cip->avail = fs->lfs_avail; 1206 cip->bfree = fs->lfs_bfree; 1207 (void) LFS_BWRITE_LOG(bp); /* Ifile */ 1208 1209 /* 1210 * Mark the current segment as ACTIVE, since we're going to 1211 * be writing to it. 1212 */ 1213 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, fs->lfs_offset), bp); 1214 sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE; 1215 fs->lfs_nactive++; 1216 LFS_WRITESEGENTRY(sup, fs, lfs_dtosn(fs, fs->lfs_offset), bp); /* Ifile */ 1217 1218 /* Now that roll-forward is done, unlock the Ifile */ 1219 vput(vp); 1220 1221 /* Start the pagedaemon-anticipating daemon */ 1222 mutex_enter(&lfs_lock); 1223 if (lfs_writer_daemon == 0 && lfs_writer_lid == 0 && 1224 kthread_create(PRI_BIO, 0, NULL, 1225 lfs_writerd, NULL, NULL, "lfs_writer") != 0) 1226 panic("fork lfs_writer"); 1227 mutex_exit(&lfs_lock); 1228 1229 printf("WARNING: the log-structured file system is experimental\n" 1230 "WARNING: it may cause system crashes and/or corrupt data\n"); 1231 1232 return (0); 1233 1234 out: 1235 if (bp) 1236 brelse(bp, 0); 1237 if (abp) 1238 brelse(abp, 0); 1239 if (ump) { 1240 kmem_free(ump->um_lfs, sizeof(struct lfs)); 1241 kmem_free(ump, sizeof(*ump)); 1242 mp->mnt_data = NULL; 1243 } 1244 1245 return (error); 1246 } 1247 1248 /* 1249 * unmount system call 1250 */ 1251 int 1252 lfs_unmount(struct mount *mp, int mntflags) 1253 { 1254 struct lwp *l = curlwp; 1255 struct ulfsmount *ump; 1256 struct lfs *fs; 1257 int error, flags, ronly; 1258 vnode_t *vp; 1259 1260 flags = 0; 1261 if (mntflags & MNT_FORCE) 1262 flags |= FORCECLOSE; 1263 1264 ump = VFSTOULFS(mp); 1265 fs = ump->um_lfs; 1266 1267 /* Two checkpoints */ 1268 lfs_segwrite(mp, SEGM_CKP | SEGM_SYNC); 1269 lfs_segwrite(mp, SEGM_CKP | SEGM_SYNC); 1270 1271 /* wake up the cleaner so it can die */ 1272 /* XXX: shouldn't this be *after* the error cases below? */ 1273 lfs_wakeup_cleaner(fs); 1274 mutex_enter(&lfs_lock); 1275 while (fs->lfs_sleepers) 1276 mtsleep(&fs->lfs_sleepers, PRIBIO + 1, "lfs_sleepers", 0, 1277 &lfs_lock); 1278 mutex_exit(&lfs_lock); 1279 1280 #ifdef LFS_EXTATTR 1281 if (ump->um_fstype == ULFS1) { 1282 if (ump->um_extattr.uepm_flags & ULFS_EXTATTR_UEPM_STARTED) { 1283 ulfs_extattr_stop(mp, curlwp); 1284 } 1285 if (ump->um_extattr.uepm_flags & ULFS_EXTATTR_UEPM_INITIALIZED) { 1286 ulfs_extattr_uepm_destroy(&ump->um_extattr); 1287 } 1288 } 1289 #endif 1290 #ifdef LFS_QUOTA 1291 if ((error = lfsquota1_umount(mp, flags)) != 0) 1292 return (error); 1293 #endif 1294 #ifdef LFS_QUOTA2 1295 if ((error = lfsquota2_umount(mp, flags)) != 0) 1296 return (error); 1297 #endif 1298 if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0) 1299 return (error); 1300 if ((error = VFS_SYNC(mp, 1, l->l_cred)) != 0) 1301 return (error); 1302 vp = fs->lfs_ivnode; 1303 mutex_enter(vp->v_interlock); 1304 if (LIST_FIRST(&vp->v_dirtyblkhd)) 1305 panic("lfs_unmount: still dirty blocks on ifile vnode"); 1306 mutex_exit(vp->v_interlock); 1307 1308 /* Explicitly write the superblock, to update serial and pflags */ 1309 fs->lfs_pflags |= LFS_PF_CLEAN; 1310 lfs_writesuper(fs, fs->lfs_sboffs[0]); 1311 lfs_writesuper(fs, fs->lfs_sboffs[1]); 1312 mutex_enter(&lfs_lock); 1313 while (fs->lfs_iocount) 1314 mtsleep(&fs->lfs_iocount, PRIBIO + 1, "lfs_umount", 0, 1315 &lfs_lock); 1316 mutex_exit(&lfs_lock); 1317 1318 /* Finish with the Ifile, now that we're done with it */ 1319 vgone(fs->lfs_ivnode); 1320 1321 ronly = !fs->lfs_ronly; 1322 if (ump->um_devvp->v_type != VBAD) 1323 spec_node_setmountedfs(ump->um_devvp, NULL); 1324 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1325 error = VOP_CLOSE(ump->um_devvp, 1326 ronly ? FREAD : FREAD|FWRITE, NOCRED); 1327 vput(ump->um_devvp); 1328 1329 /* Complain about page leakage */ 1330 if (fs->lfs_pages > 0) 1331 printf("lfs_unmount: still claim %d pages (%d in subsystem)\n", 1332 fs->lfs_pages, lfs_subsys_pages); 1333 1334 /* Free per-mount data structures */ 1335 free(fs->lfs_ino_bitmap, M_SEGMENT); 1336 free(fs->lfs_suflags[0], M_SEGMENT); 1337 free(fs->lfs_suflags[1], M_SEGMENT); 1338 free(fs->lfs_suflags, M_SEGMENT); 1339 lfs_free_resblks(fs); 1340 cv_destroy(&fs->lfs_stopcv); 1341 rw_destroy(&fs->lfs_fraglock); 1342 rw_destroy(&fs->lfs_iflock); 1343 1344 kmem_free(fs, sizeof(struct lfs)); 1345 kmem_free(ump, sizeof(*ump)); 1346 1347 mp->mnt_data = NULL; 1348 mp->mnt_flag &= ~MNT_LOCAL; 1349 return (error); 1350 } 1351 1352 /* 1353 * Get file system statistics. 1354 * 1355 * NB: We don't lock to access the superblock here, because it's not 1356 * really that important if we get it wrong. 1357 */ 1358 int 1359 lfs_statvfs(struct mount *mp, struct statvfs *sbp) 1360 { 1361 struct lfs *fs; 1362 struct ulfsmount *ump; 1363 1364 ump = VFSTOULFS(mp); 1365 fs = ump->um_lfs; 1366 if (fs->lfs_magic != LFS_MAGIC) 1367 panic("lfs_statvfs: magic"); 1368 1369 sbp->f_bsize = fs->lfs_bsize; 1370 sbp->f_frsize = fs->lfs_fsize; 1371 sbp->f_iosize = fs->lfs_bsize; 1372 sbp->f_blocks = LFS_EST_NONMETA(fs) - VTOI(fs->lfs_ivnode)->i_lfs_effnblks; 1373 1374 sbp->f_bfree = LFS_EST_BFREE(fs); 1375 KASSERT(sbp->f_bfree <= fs->lfs_dsize); 1376 #if 0 1377 if (sbp->f_bfree < 0) 1378 sbp->f_bfree = 0; 1379 #endif 1380 1381 sbp->f_bresvd = LFS_EST_RSVD(fs); 1382 if (sbp->f_bfree > sbp->f_bresvd) 1383 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd; 1384 else 1385 sbp->f_bavail = 0; 1386 1387 sbp->f_files = fs->lfs_bfree / lfs_btofsb(fs, fs->lfs_ibsize) 1388 * LFS_INOPB(fs); 1389 sbp->f_ffree = sbp->f_files - fs->lfs_nfiles; 1390 sbp->f_favail = sbp->f_ffree; 1391 sbp->f_fresvd = 0; 1392 copy_statvfs_info(sbp, mp); 1393 return (0); 1394 } 1395 1396 /* 1397 * Go through the disk queues to initiate sandbagged IO; 1398 * go through the inodes to write those that have been modified; 1399 * initiate the writing of the super block if it has been modified. 1400 * 1401 * Note: we are always called with the filesystem marked `MPBUSY'. 1402 */ 1403 int 1404 lfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred) 1405 { 1406 int error; 1407 struct lfs *fs; 1408 1409 fs = VFSTOULFS(mp)->um_lfs; 1410 if (fs->lfs_ronly) 1411 return 0; 1412 1413 /* Snapshots should not hose the syncer */ 1414 /* 1415 * XXX Sync can block here anyway, since we don't have a very 1416 * XXX good idea of how much data is pending. If it's more 1417 * XXX than a segment and lfs_nextseg is close to the end of 1418 * XXX the log, we'll likely block. 1419 */ 1420 mutex_enter(&lfs_lock); 1421 if (fs->lfs_nowrap && fs->lfs_nextseg < fs->lfs_curseg) { 1422 mutex_exit(&lfs_lock); 1423 return 0; 1424 } 1425 mutex_exit(&lfs_lock); 1426 1427 lfs_writer_enter(fs, "lfs_dirops"); 1428 1429 /* All syncs must be checkpoints until roll-forward is implemented. */ 1430 DLOG((DLOG_FLUSH, "lfs_sync at 0x%x\n", fs->lfs_offset)); 1431 error = lfs_segwrite(mp, SEGM_CKP | (waitfor ? SEGM_SYNC : 0)); 1432 lfs_writer_leave(fs); 1433 #ifdef LFS_QUOTA 1434 lfs_qsync(mp); 1435 #endif 1436 return (error); 1437 } 1438 1439 /* 1440 * Look up an LFS dinode number to find its incore vnode. If not already 1441 * in core, read it in from the specified device. Return the inode locked. 1442 * Detection and handling of mount points must be done by the calling routine. 1443 */ 1444 int 1445 lfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp) 1446 { 1447 struct lfs *fs; 1448 struct ulfs1_dinode *dip; 1449 struct inode *ip; 1450 struct buf *bp; 1451 struct ifile *ifp; 1452 struct vnode *vp; 1453 struct ulfsmount *ump; 1454 daddr_t daddr; 1455 dev_t dev; 1456 int error, retries; 1457 struct timespec ts; 1458 1459 memset(&ts, 0, sizeof ts); /* XXX gcc */ 1460 1461 ump = VFSTOULFS(mp); 1462 dev = ump->um_dev; 1463 fs = ump->um_lfs; 1464 1465 /* 1466 * If the filesystem is not completely mounted yet, suspend 1467 * any access requests (wait for roll-forward to complete). 1468 */ 1469 mutex_enter(&lfs_lock); 1470 while ((fs->lfs_flags & LFS_NOTYET) && curproc->p_pid != fs->lfs_rfpid) 1471 mtsleep(&fs->lfs_flags, PRIBIO+1, "lfs_notyet", 0, 1472 &lfs_lock); 1473 mutex_exit(&lfs_lock); 1474 1475 retry: 1476 if ((*vpp = ulfs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL) 1477 return (0); 1478 1479 error = getnewvnode(VT_LFS, mp, lfs_vnodeop_p, NULL, &vp); 1480 if (error) { 1481 *vpp = NULL; 1482 return (error); 1483 } 1484 1485 mutex_enter(&ulfs_hashlock); 1486 if (ulfs_ihashget(dev, ino, 0) != NULL) { 1487 mutex_exit(&ulfs_hashlock); 1488 ungetnewvnode(vp); 1489 goto retry; 1490 } 1491 1492 /* Translate the inode number to a disk address. */ 1493 if (ino == LFS_IFILE_INUM) 1494 daddr = fs->lfs_idaddr; 1495 else { 1496 /* XXX bounds-check this too */ 1497 LFS_IENTRY(ifp, fs, ino, bp); 1498 daddr = ifp->if_daddr; 1499 if (fs->lfs_version > 1) { 1500 ts.tv_sec = ifp->if_atime_sec; 1501 ts.tv_nsec = ifp->if_atime_nsec; 1502 } 1503 1504 brelse(bp, 0); 1505 if (daddr == LFS_UNUSED_DADDR) { 1506 *vpp = NULLVP; 1507 mutex_exit(&ulfs_hashlock); 1508 ungetnewvnode(vp); 1509 return (ENOENT); 1510 } 1511 } 1512 1513 /* Allocate/init new vnode/inode. */ 1514 lfs_vcreate(mp, ino, vp); 1515 1516 /* 1517 * Put it onto its hash chain and lock it so that other requests for 1518 * this inode will block if they arrive while we are sleeping waiting 1519 * for old data structures to be purged or for the contents of the 1520 * disk portion of this inode to be read. 1521 */ 1522 ip = VTOI(vp); 1523 ulfs_ihashins(ip); 1524 mutex_exit(&ulfs_hashlock); 1525 1526 /* 1527 * XXX 1528 * This may not need to be here, logically it should go down with 1529 * the i_devvp initialization. 1530 * Ask Kirk. 1531 */ 1532 ip->i_lfs = ump->um_lfs; 1533 1534 /* Read in the disk contents for the inode, copy into the inode. */ 1535 retries = 0; 1536 again: 1537 error = bread(ump->um_devvp, LFS_FSBTODB(fs, daddr), 1538 (fs->lfs_version == 1 ? fs->lfs_bsize : fs->lfs_ibsize), 1539 NOCRED, 0, &bp); 1540 if (error) { 1541 /* 1542 * The inode does not contain anything useful, so it would 1543 * be misleading to leave it on its hash chain. With mode 1544 * still zero, it will be unlinked and returned to the free 1545 * list by vput(). 1546 */ 1547 vput(vp); 1548 *vpp = NULL; 1549 return (error); 1550 } 1551 1552 dip = lfs_ifind(fs, ino, bp); 1553 if (dip == NULL) { 1554 /* Assume write has not completed yet; try again */ 1555 brelse(bp, BC_INVAL); 1556 ++retries; 1557 if (retries > LFS_IFIND_RETRIES) { 1558 #ifdef DEBUG 1559 /* If the seglock is held look at the bpp to see 1560 what is there anyway */ 1561 mutex_enter(&lfs_lock); 1562 if (fs->lfs_seglock > 0) { 1563 struct buf **bpp; 1564 struct ulfs1_dinode *dp; 1565 int i; 1566 1567 for (bpp = fs->lfs_sp->bpp; 1568 bpp != fs->lfs_sp->cbpp; ++bpp) { 1569 if ((*bpp)->b_vp == fs->lfs_ivnode && 1570 bpp != fs->lfs_sp->bpp) { 1571 /* Inode block */ 1572 printf("lfs_vget: block 0x%" PRIx64 ": ", 1573 (*bpp)->b_blkno); 1574 dp = (struct ulfs1_dinode *)(*bpp)->b_data; 1575 for (i = 0; i < LFS_INOPB(fs); i++) 1576 if (dp[i].di_inumber) 1577 printf("%d ", dp[i].di_inumber); 1578 printf("\n"); 1579 } 1580 } 1581 } 1582 mutex_exit(&lfs_lock); 1583 #endif /* DEBUG */ 1584 panic("lfs_vget: dinode not found"); 1585 } 1586 mutex_enter(&lfs_lock); 1587 if (fs->lfs_iocount) { 1588 DLOG((DLOG_VNODE, "lfs_vget: dinode %d not found, retrying...\n", ino)); 1589 (void)mtsleep(&fs->lfs_iocount, PRIBIO + 1, 1590 "lfs ifind", 1, &lfs_lock); 1591 } else 1592 retries = LFS_IFIND_RETRIES; 1593 mutex_exit(&lfs_lock); 1594 goto again; 1595 } 1596 *ip->i_din.ffs1_din = *dip; 1597 brelse(bp, 0); 1598 1599 if (fs->lfs_version > 1) { 1600 ip->i_ffs1_atime = ts.tv_sec; 1601 ip->i_ffs1_atimensec = ts.tv_nsec; 1602 } 1603 1604 lfs_vinit(mp, &vp); 1605 1606 *vpp = vp; 1607 1608 KASSERT(VOP_ISLOCKED(vp)); 1609 1610 return (0); 1611 } 1612 1613 /* 1614 * File handle to vnode 1615 */ 1616 int 1617 lfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 1618 { 1619 struct lfid lfh; 1620 struct buf *bp; 1621 IFILE *ifp; 1622 int32_t daddr; 1623 struct lfs *fs; 1624 vnode_t *vp; 1625 1626 if (fhp->fid_len != sizeof(struct lfid)) 1627 return EINVAL; 1628 1629 memcpy(&lfh, fhp, sizeof(lfh)); 1630 if (lfh.lfid_ino < LFS_IFILE_INUM) 1631 return ESTALE; 1632 1633 fs = VFSTOULFS(mp)->um_lfs; 1634 if (lfh.lfid_ident != fs->lfs_ident) 1635 return ESTALE; 1636 1637 if (lfh.lfid_ino > 1638 ((VTOI(fs->lfs_ivnode)->i_ffs1_size >> fs->lfs_bshift) - 1639 fs->lfs_cleansz - fs->lfs_segtabsz) * fs->lfs_ifpb) 1640 return ESTALE; 1641 1642 mutex_enter(&ulfs_ihash_lock); 1643 vp = ulfs_ihashlookup(VFSTOULFS(mp)->um_dev, lfh.lfid_ino); 1644 mutex_exit(&ulfs_ihash_lock); 1645 if (vp == NULL) { 1646 LFS_IENTRY(ifp, fs, lfh.lfid_ino, bp); 1647 daddr = ifp->if_daddr; 1648 brelse(bp, 0); 1649 if (daddr == LFS_UNUSED_DADDR) 1650 return ESTALE; 1651 } 1652 1653 return (ulfs_fhtovp(mp, &lfh.lfid_ufid, vpp)); 1654 } 1655 1656 /* 1657 * Vnode pointer to File handle 1658 */ 1659 /* ARGSUSED */ 1660 int 1661 lfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size) 1662 { 1663 struct inode *ip; 1664 struct lfid lfh; 1665 1666 if (*fh_size < sizeof(struct lfid)) { 1667 *fh_size = sizeof(struct lfid); 1668 return E2BIG; 1669 } 1670 *fh_size = sizeof(struct lfid); 1671 ip = VTOI(vp); 1672 memset(&lfh, 0, sizeof(lfh)); 1673 lfh.lfid_len = sizeof(struct lfid); 1674 lfh.lfid_ino = ip->i_number; 1675 lfh.lfid_gen = ip->i_gen; 1676 lfh.lfid_ident = ip->i_lfs->lfs_ident; 1677 memcpy(fhp, &lfh, sizeof(lfh)); 1678 return (0); 1679 } 1680 1681 /* 1682 * ulfs_bmaparray callback function for writing. 1683 * 1684 * Since blocks will be written to the new segment anyway, 1685 * we don't care about current daddr of them. 1686 */ 1687 static bool 1688 lfs_issequential_hole(const struct lfs *fs, 1689 daddr_t daddr0, daddr_t daddr1) 1690 { 1691 (void)fs; /* not used */ 1692 1693 daddr0 = (daddr_t)((int32_t)daddr0); /* XXX ondisk32 */ 1694 daddr1 = (daddr_t)((int32_t)daddr1); /* XXX ondisk32 */ 1695 1696 KASSERT(daddr0 == UNWRITTEN || 1697 (0 <= daddr0 && daddr0 <= LFS_MAX_DADDR)); 1698 KASSERT(daddr1 == UNWRITTEN || 1699 (0 <= daddr1 && daddr1 <= LFS_MAX_DADDR)); 1700 1701 /* NOTE: all we want to know here is 'hole or not'. */ 1702 /* NOTE: UNASSIGNED is converted to 0 by ulfs_bmaparray. */ 1703 1704 /* 1705 * treat UNWRITTENs and all resident blocks as 'contiguous' 1706 */ 1707 if (daddr0 != 0 && daddr1 != 0) 1708 return true; 1709 1710 /* 1711 * both are in hole? 1712 */ 1713 if (daddr0 == 0 && daddr1 == 0) 1714 return true; /* all holes are 'contiguous' for us. */ 1715 1716 return false; 1717 } 1718 1719 /* 1720 * lfs_gop_write functions exactly like genfs_gop_write, except that 1721 * (1) it requires the seglock to be held by its caller, and sp->fip 1722 * to be properly initialized (it will return without re-initializing 1723 * sp->fip, and without calling lfs_writeseg). 1724 * (2) it uses the remaining space in the segment, rather than VOP_BMAP, 1725 * to determine how large a block it can write at once (though it does 1726 * still use VOP_BMAP to find holes in the file); 1727 * (3) it calls lfs_gatherblock instead of VOP_STRATEGY on its blocks 1728 * (leaving lfs_writeseg to deal with the cluster blocks, so we might 1729 * now have clusters of clusters, ick.) 1730 */ 1731 static int 1732 lfs_gop_write(struct vnode *vp, struct vm_page **pgs, int npages, 1733 int flags) 1734 { 1735 int i, error, run, haveeof = 0; 1736 int fs_bshift; 1737 vaddr_t kva; 1738 off_t eof, offset, startoffset = 0; 1739 size_t bytes, iobytes, skipbytes; 1740 bool async = (flags & PGO_SYNCIO) == 0; 1741 daddr_t lbn, blkno; 1742 struct vm_page *pg; 1743 struct buf *mbp, *bp; 1744 struct vnode *devvp = VTOI(vp)->i_devvp; 1745 struct inode *ip = VTOI(vp); 1746 struct lfs *fs = ip->i_lfs; 1747 struct segment *sp = fs->lfs_sp; 1748 UVMHIST_FUNC("lfs_gop_write"); UVMHIST_CALLED(ubchist); 1749 const char * failreason = NULL; 1750 1751 ASSERT_SEGLOCK(fs); 1752 1753 /* The Ifile lives in the buffer cache */ 1754 KASSERT(vp != fs->lfs_ivnode); 1755 1756 /* 1757 * We don't want to fill the disk before the cleaner has a chance 1758 * to make room for us. If we're in danger of doing that, fail 1759 * with EAGAIN. The caller will have to notice this, unlock 1760 * so the cleaner can run, relock and try again. 1761 * 1762 * We must write everything, however, if our vnode is being 1763 * reclaimed. 1764 */ 1765 if (LFS_STARVED_FOR_SEGS(fs) && !(vp->v_iflag & VI_XLOCK)) { 1766 failreason = "Starved for segs and not flushing vp"; 1767 goto tryagain; 1768 } 1769 1770 /* 1771 * Sometimes things slip past the filters in lfs_putpages, 1772 * and the pagedaemon tries to write pages---problem is 1773 * that the pagedaemon never acquires the segment lock. 1774 * 1775 * Alternatively, pages that were clean when we called 1776 * genfs_putpages may have become dirty in the meantime. In this 1777 * case the segment header is not properly set up for blocks 1778 * to be added to it. 1779 * 1780 * Unbusy and unclean the pages, and put them on the ACTIVE 1781 * queue under the hypothesis that they couldn't have got here 1782 * unless they were modified *quite* recently. 1783 * 1784 * XXXUBC that last statement is an oversimplification of course. 1785 */ 1786 if (!LFS_SEGLOCK_HELD(fs)) { 1787 failreason = "Seglock not held"; 1788 goto tryagain; 1789 } 1790 if (ip->i_lfs_iflags & LFSI_NO_GOP_WRITE) { 1791 failreason = "Inode with no_gop_write"; 1792 goto tryagain; 1793 } 1794 if ((pgs[0]->offset & fs->lfs_bmask) != 0) { 1795 failreason = "Bad page offset"; 1796 goto tryagain; 1797 } 1798 1799 UVMHIST_LOG(ubchist, "vp %p pgs %p npages %d flags 0x%x", 1800 vp, pgs, npages, flags); 1801 1802 GOP_SIZE(vp, vp->v_size, &eof, 0); 1803 haveeof = 1; 1804 1805 if (vp->v_type == VREG) 1806 fs_bshift = vp->v_mount->mnt_fs_bshift; 1807 else 1808 fs_bshift = DEV_BSHIFT; 1809 error = 0; 1810 pg = pgs[0]; 1811 startoffset = pg->offset; 1812 KASSERT(eof >= 0); 1813 1814 if (startoffset >= eof) { 1815 failreason = "Offset beyond EOF"; 1816 goto tryagain; 1817 } else 1818 bytes = MIN(npages << PAGE_SHIFT, eof - startoffset); 1819 skipbytes = 0; 1820 1821 KASSERT(bytes != 0); 1822 1823 /* Swap PG_DELWRI for PG_PAGEOUT */ 1824 for (i = 0; i < npages; i++) { 1825 if (pgs[i]->flags & PG_DELWRI) { 1826 KASSERT(!(pgs[i]->flags & PG_PAGEOUT)); 1827 pgs[i]->flags &= ~PG_DELWRI; 1828 pgs[i]->flags |= PG_PAGEOUT; 1829 uvm_pageout_start(1); 1830 mutex_enter(vp->v_interlock); 1831 mutex_enter(&uvm_pageqlock); 1832 uvm_pageunwire(pgs[i]); 1833 mutex_exit(&uvm_pageqlock); 1834 mutex_exit(vp->v_interlock); 1835 } 1836 } 1837 1838 /* 1839 * Check to make sure we're starting on a block boundary. 1840 * We'll check later to make sure we always write entire 1841 * blocks (or fragments). 1842 */ 1843 if (startoffset & fs->lfs_bmask) 1844 printf("%" PRId64 " & %" PRId64 " = %" PRId64 "\n", 1845 startoffset, fs->lfs_bmask, 1846 startoffset & fs->lfs_bmask); 1847 KASSERT((startoffset & fs->lfs_bmask) == 0); 1848 if (bytes & fs->lfs_ffmask) { 1849 printf("lfs_gop_write: asked to write %ld bytes\n", (long)bytes); 1850 panic("lfs_gop_write: non-integer blocks"); 1851 } 1852 1853 /* 1854 * We could deadlock here on pager_map with UVMPAGER_MAPIN_WAITOK. 1855 * If we would, write what we have and try again. If we don't 1856 * have anything to write, we'll have to sleep. 1857 */ 1858 if ((kva = uvm_pagermapin(pgs, npages, UVMPAGER_MAPIN_WRITE | 1859 (((SEGSUM *)(sp->segsum))->ss_nfinfo < 1 ? 1860 UVMPAGER_MAPIN_WAITOK : 0))) == 0x0) { 1861 DLOG((DLOG_PAGE, "lfs_gop_write: forcing write\n")); 1862 #if 0 1863 " with nfinfo=%d at offset 0x%x\n", 1864 (int)((SEGSUM *)(sp->segsum))->ss_nfinfo, 1865 (unsigned)fs->lfs_offset)); 1866 #endif 1867 lfs_updatemeta(sp); 1868 lfs_release_finfo(fs); 1869 (void) lfs_writeseg(fs, sp); 1870 1871 lfs_acquire_finfo(fs, ip->i_number, ip->i_gen); 1872 1873 /* 1874 * Having given up all of the pager_map we were holding, 1875 * we can now wait for aiodoned to reclaim it for us 1876 * without fear of deadlock. 1877 */ 1878 kva = uvm_pagermapin(pgs, npages, UVMPAGER_MAPIN_WRITE | 1879 UVMPAGER_MAPIN_WAITOK); 1880 } 1881 1882 mbp = getiobuf(NULL, true); 1883 UVMHIST_LOG(ubchist, "vp %p mbp %p num now %d bytes 0x%x", 1884 vp, mbp, vp->v_numoutput, bytes); 1885 mbp->b_bufsize = npages << PAGE_SHIFT; 1886 mbp->b_data = (void *)kva; 1887 mbp->b_resid = mbp->b_bcount = bytes; 1888 mbp->b_cflags = BC_BUSY|BC_AGE; 1889 mbp->b_iodone = uvm_aio_biodone; 1890 1891 bp = NULL; 1892 for (offset = startoffset; 1893 bytes > 0; 1894 offset += iobytes, bytes -= iobytes) { 1895 lbn = offset >> fs_bshift; 1896 error = ulfs_bmaparray(vp, lbn, &blkno, NULL, NULL, &run, 1897 lfs_issequential_hole); 1898 if (error) { 1899 UVMHIST_LOG(ubchist, "ulfs_bmaparray() -> %d", 1900 error,0,0,0); 1901 skipbytes += bytes; 1902 bytes = 0; 1903 break; 1904 } 1905 1906 iobytes = MIN((((off_t)lbn + 1 + run) << fs_bshift) - offset, 1907 bytes); 1908 if (blkno == (daddr_t)-1) { 1909 skipbytes += iobytes; 1910 continue; 1911 } 1912 1913 /* 1914 * Discover how much we can really pack into this buffer. 1915 */ 1916 /* If no room in the current segment, finish it up */ 1917 if (sp->sum_bytes_left < sizeof(int32_t) || 1918 sp->seg_bytes_left < (1 << fs->lfs_bshift)) { 1919 int vers; 1920 1921 lfs_updatemeta(sp); 1922 vers = sp->fip->fi_version; 1923 lfs_release_finfo(fs); 1924 (void) lfs_writeseg(fs, sp); 1925 1926 lfs_acquire_finfo(fs, ip->i_number, vers); 1927 } 1928 /* Check both for space in segment and space in segsum */ 1929 iobytes = MIN(iobytes, (sp->seg_bytes_left >> fs_bshift) 1930 << fs_bshift); 1931 iobytes = MIN(iobytes, (sp->sum_bytes_left / sizeof(int32_t)) 1932 << fs_bshift); 1933 KASSERT(iobytes > 0); 1934 1935 /* if it's really one i/o, don't make a second buf */ 1936 if (offset == startoffset && iobytes == bytes) { 1937 bp = mbp; 1938 /* 1939 * All the LFS output is done by the segwriter. It 1940 * will increment numoutput by one for all the bufs it 1941 * recieves. However this buffer needs one extra to 1942 * account for aiodone. 1943 */ 1944 mutex_enter(vp->v_interlock); 1945 vp->v_numoutput++; 1946 mutex_exit(vp->v_interlock); 1947 } else { 1948 bp = getiobuf(NULL, true); 1949 UVMHIST_LOG(ubchist, "vp %p bp %p num now %d", 1950 vp, bp, vp->v_numoutput, 0); 1951 nestiobuf_setup(mbp, bp, offset - pg->offset, iobytes); 1952 /* 1953 * LFS doesn't like async I/O here, dies with 1954 * an assert in lfs_bwrite(). Is that assert 1955 * valid? I retained non-async behaviour when 1956 * converted this to use nestiobuf --pooka 1957 */ 1958 bp->b_flags &= ~B_ASYNC; 1959 } 1960 1961 /* XXX This is silly ... is this necessary? */ 1962 mutex_enter(&bufcache_lock); 1963 mutex_enter(vp->v_interlock); 1964 bgetvp(vp, bp); 1965 mutex_exit(vp->v_interlock); 1966 mutex_exit(&bufcache_lock); 1967 1968 bp->b_lblkno = lfs_lblkno(fs, offset); 1969 bp->b_private = mbp; 1970 if (devvp->v_type == VBLK) { 1971 bp->b_dev = devvp->v_rdev; 1972 } 1973 VOP_BWRITE(bp->b_vp, bp); 1974 while (lfs_gatherblock(sp, bp, NULL)) 1975 continue; 1976 } 1977 1978 nestiobuf_done(mbp, skipbytes, error); 1979 if (skipbytes) { 1980 UVMHIST_LOG(ubchist, "skipbytes %d", skipbytes, 0,0,0); 1981 } 1982 UVMHIST_LOG(ubchist, "returning 0", 0,0,0,0); 1983 1984 if (!async) { 1985 /* Start a segment write. */ 1986 UVMHIST_LOG(ubchist, "flushing", 0,0,0,0); 1987 mutex_enter(&lfs_lock); 1988 lfs_flush(fs, 0, 1); 1989 mutex_exit(&lfs_lock); 1990 } 1991 1992 if ((sp->seg_flags & SEGM_SINGLE) && fs->lfs_curseg != fs->lfs_startseg) 1993 return EAGAIN; 1994 1995 return (0); 1996 1997 tryagain: 1998 /* 1999 * We can't write the pages, for whatever reason. 2000 * Clean up after ourselves, and make the caller try again. 2001 */ 2002 mutex_enter(vp->v_interlock); 2003 2004 /* Tell why we're here, if we know */ 2005 if (failreason != NULL) { 2006 DLOG((DLOG_PAGE, "lfs_gop_write: %s\n", failreason)); 2007 } 2008 if (haveeof && startoffset >= eof) { 2009 DLOG((DLOG_PAGE, "lfs_gop_write: ino %d start 0x%" PRIx64 2010 " eof 0x%" PRIx64 " npages=%d\n", VTOI(vp)->i_number, 2011 pgs[0]->offset, eof, npages)); 2012 } 2013 2014 mutex_enter(&uvm_pageqlock); 2015 for (i = 0; i < npages; i++) { 2016 pg = pgs[i]; 2017 2018 if (pg->flags & PG_PAGEOUT) 2019 uvm_pageout_done(1); 2020 if (pg->flags & PG_DELWRI) { 2021 uvm_pageunwire(pg); 2022 } 2023 uvm_pageactivate(pg); 2024 pg->flags &= ~(PG_CLEAN|PG_DELWRI|PG_PAGEOUT|PG_RELEASED); 2025 DLOG((DLOG_PAGE, "pg[%d] = %p (vp %p off %" PRIx64 ")\n", i, pg, 2026 vp, pg->offset)); 2027 DLOG((DLOG_PAGE, "pg[%d]->flags = %x\n", i, pg->flags)); 2028 DLOG((DLOG_PAGE, "pg[%d]->pqflags = %x\n", i, pg->pqflags)); 2029 DLOG((DLOG_PAGE, "pg[%d]->uanon = %p\n", i, pg->uanon)); 2030 DLOG((DLOG_PAGE, "pg[%d]->uobject = %p\n", i, pg->uobject)); 2031 DLOG((DLOG_PAGE, "pg[%d]->wire_count = %d\n", i, 2032 pg->wire_count)); 2033 DLOG((DLOG_PAGE, "pg[%d]->loan_count = %d\n", i, 2034 pg->loan_count)); 2035 } 2036 /* uvm_pageunbusy takes care of PG_BUSY, PG_WANTED */ 2037 uvm_page_unbusy(pgs, npages); 2038 mutex_exit(&uvm_pageqlock); 2039 mutex_exit(vp->v_interlock); 2040 return EAGAIN; 2041 } 2042 2043 /* 2044 * finish vnode/inode initialization. 2045 * used by lfs_vget and lfs_fastvget. 2046 */ 2047 void 2048 lfs_vinit(struct mount *mp, struct vnode **vpp) 2049 { 2050 struct vnode *vp = *vpp; 2051 struct inode *ip = VTOI(vp); 2052 struct ulfsmount *ump = VFSTOULFS(mp); 2053 struct lfs *fs = ump->um_lfs; 2054 int i; 2055 2056 ip->i_mode = ip->i_ffs1_mode; 2057 ip->i_nlink = ip->i_ffs1_nlink; 2058 ip->i_lfs_osize = ip->i_size = ip->i_ffs1_size; 2059 ip->i_flags = ip->i_ffs1_flags; 2060 ip->i_gen = ip->i_ffs1_gen; 2061 ip->i_uid = ip->i_ffs1_uid; 2062 ip->i_gid = ip->i_ffs1_gid; 2063 2064 ip->i_lfs_effnblks = ip->i_ffs1_blocks; 2065 ip->i_lfs_odnlink = ip->i_ffs1_nlink; 2066 2067 /* 2068 * Initialize the vnode from the inode, check for aliases. In all 2069 * cases re-init ip, the underlying vnode/inode may have changed. 2070 */ 2071 ulfs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp); 2072 ip = VTOI(vp); 2073 2074 memset(ip->i_lfs_fragsize, 0, ULFS_NDADDR * sizeof(*ip->i_lfs_fragsize)); 2075 if (vp->v_type != VLNK || ip->i_size >= ip->i_lfs->um_maxsymlinklen) { 2076 #ifdef DEBUG 2077 for (i = (ip->i_size + fs->lfs_bsize - 1) >> fs->lfs_bshift; 2078 i < ULFS_NDADDR; i++) { 2079 if ((vp->v_type == VBLK || vp->v_type == VCHR) && 2080 i == 0) 2081 continue; 2082 if (ip->i_ffs1_db[i] != 0) { 2083 lfs_dump_dinode(ip->i_din.ffs1_din); 2084 panic("inconsistent inode (direct)"); 2085 } 2086 } 2087 for ( ; i < ULFS_NDADDR + ULFS_NIADDR; i++) { 2088 if (ip->i_ffs1_ib[i - ULFS_NDADDR] != 0) { 2089 lfs_dump_dinode(ip->i_din.ffs1_din); 2090 panic("inconsistent inode (indirect)"); 2091 } 2092 } 2093 #endif /* DEBUG */ 2094 for (i = 0; i < ULFS_NDADDR; i++) 2095 if (ip->i_ffs1_db[i] != 0) 2096 ip->i_lfs_fragsize[i] = lfs_blksize(fs, ip, i); 2097 } 2098 2099 #ifdef DIAGNOSTIC 2100 if (vp->v_type == VNON) { 2101 # ifdef DEBUG 2102 lfs_dump_dinode(ip->i_din.ffs1_din); 2103 # endif 2104 panic("lfs_vinit: ino %llu is type VNON! (ifmt=%o)\n", 2105 (unsigned long long)ip->i_number, 2106 (ip->i_mode & LFS_IFMT) >> 12); 2107 } 2108 #endif /* DIAGNOSTIC */ 2109 2110 /* 2111 * Finish inode initialization now that aliasing has been resolved. 2112 */ 2113 2114 ip->i_devvp = ump->um_devvp; 2115 vref(ip->i_devvp); 2116 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2) 2117 ulfsquota_init(ip); 2118 #endif 2119 genfs_node_init(vp, &lfs_genfsops); 2120 uvm_vnp_setsize(vp, ip->i_size); 2121 2122 /* Initialize hiblk from file size */ 2123 ip->i_lfs_hiblk = lfs_lblkno(ip->i_lfs, ip->i_size + ip->i_lfs->lfs_bsize - 1) - 1; 2124 2125 *vpp = vp; 2126 } 2127 2128 /* 2129 * Resize the filesystem to contain the specified number of segments. 2130 */ 2131 int 2132 lfs_resize_fs(struct lfs *fs, int newnsegs) 2133 { 2134 SEGUSE *sup; 2135 struct buf *bp, *obp; 2136 daddr_t olast, nlast, ilast, noff, start, end; 2137 struct vnode *ivp; 2138 struct inode *ip; 2139 int error, badnews, inc, oldnsegs; 2140 int sbbytes, csbbytes, gain, cgain; 2141 int i; 2142 2143 /* Only support v2 and up */ 2144 if (fs->lfs_version < 2) 2145 return EOPNOTSUPP; 2146 2147 /* If we're doing nothing, do it fast */ 2148 oldnsegs = fs->lfs_nseg; 2149 if (newnsegs == oldnsegs) 2150 return 0; 2151 2152 /* We always have to have two superblocks */ 2153 if (newnsegs <= lfs_dtosn(fs, fs->lfs_sboffs[1])) 2154 return EFBIG; 2155 2156 ivp = fs->lfs_ivnode; 2157 ip = VTOI(ivp); 2158 error = 0; 2159 2160 /* Take the segment lock so no one else calls lfs_newseg() */ 2161 lfs_seglock(fs, SEGM_PROT); 2162 2163 /* 2164 * Make sure the segments we're going to be losing, if any, 2165 * are in fact empty. We hold the seglock, so their status 2166 * cannot change underneath us. Count the superblocks we lose, 2167 * while we're at it. 2168 */ 2169 sbbytes = csbbytes = 0; 2170 cgain = 0; 2171 for (i = newnsegs; i < oldnsegs; i++) { 2172 LFS_SEGENTRY(sup, fs, i, bp); 2173 badnews = sup->su_nbytes || !(sup->su_flags & SEGUSE_INVAL); 2174 if (sup->su_flags & SEGUSE_SUPERBLOCK) 2175 sbbytes += LFS_SBPAD; 2176 if (!(sup->su_flags & SEGUSE_DIRTY)) { 2177 ++cgain; 2178 if (sup->su_flags & SEGUSE_SUPERBLOCK) 2179 csbbytes += LFS_SBPAD; 2180 } 2181 brelse(bp, 0); 2182 if (badnews) { 2183 error = EBUSY; 2184 goto out; 2185 } 2186 } 2187 2188 /* Note old and new segment table endpoints, and old ifile size */ 2189 olast = fs->lfs_cleansz + fs->lfs_segtabsz; 2190 nlast = howmany(newnsegs, fs->lfs_sepb) + fs->lfs_cleansz; 2191 ilast = ivp->v_size >> fs->lfs_bshift; 2192 noff = nlast - olast; 2193 2194 /* 2195 * Make sure no one can use the Ifile while we change it around. 2196 * Even after taking the iflock we need to make sure no one still 2197 * is holding Ifile buffers, so we get each one, to drain them. 2198 * (XXX this could be done better.) 2199 */ 2200 rw_enter(&fs->lfs_iflock, RW_WRITER); 2201 for (i = 0; i < ilast; i++) { 2202 /* XXX what to do if bread fails? */ 2203 bread(ivp, i, fs->lfs_bsize, NOCRED, 0, &bp); 2204 brelse(bp, 0); 2205 } 2206 2207 /* Allocate new Ifile blocks */ 2208 for (i = ilast; i < ilast + noff; i++) { 2209 if (lfs_balloc(ivp, i * fs->lfs_bsize, fs->lfs_bsize, NOCRED, 0, 2210 &bp) != 0) 2211 panic("balloc extending ifile"); 2212 memset(bp->b_data, 0, fs->lfs_bsize); 2213 VOP_BWRITE(bp->b_vp, bp); 2214 } 2215 2216 /* Register new ifile size */ 2217 ip->i_size += noff * fs->lfs_bsize; 2218 ip->i_ffs1_size = ip->i_size; 2219 uvm_vnp_setsize(ivp, ip->i_size); 2220 2221 /* Copy the inode table to its new position */ 2222 if (noff != 0) { 2223 if (noff < 0) { 2224 start = nlast; 2225 end = ilast + noff; 2226 inc = 1; 2227 } else { 2228 start = ilast + noff - 1; 2229 end = nlast - 1; 2230 inc = -1; 2231 } 2232 for (i = start; i != end; i += inc) { 2233 if (bread(ivp, i, fs->lfs_bsize, NOCRED, 2234 B_MODIFY, &bp) != 0) 2235 panic("resize: bread dst blk failed"); 2236 if (bread(ivp, i - noff, fs->lfs_bsize, 2237 NOCRED, 0, &obp)) 2238 panic("resize: bread src blk failed"); 2239 memcpy(bp->b_data, obp->b_data, fs->lfs_bsize); 2240 VOP_BWRITE(bp->b_vp, bp); 2241 brelse(obp, 0); 2242 } 2243 } 2244 2245 /* If we are expanding, write the new empty SEGUSE entries */ 2246 if (newnsegs > oldnsegs) { 2247 for (i = oldnsegs; i < newnsegs; i++) { 2248 if ((error = bread(ivp, i / fs->lfs_sepb + 2249 fs->lfs_cleansz, fs->lfs_bsize, 2250 NOCRED, B_MODIFY, &bp)) != 0) 2251 panic("lfs: ifile read: %d", error); 2252 while ((i + 1) % fs->lfs_sepb && i < newnsegs) { 2253 sup = &((SEGUSE *)bp->b_data)[i % fs->lfs_sepb]; 2254 memset(sup, 0, sizeof(*sup)); 2255 i++; 2256 } 2257 VOP_BWRITE(bp->b_vp, bp); 2258 } 2259 } 2260 2261 /* Zero out unused superblock offsets */ 2262 for (i = 2; i < LFS_MAXNUMSB; i++) 2263 if (lfs_dtosn(fs, fs->lfs_sboffs[i]) >= newnsegs) 2264 fs->lfs_sboffs[i] = 0x0; 2265 2266 /* 2267 * Correct superblock entries that depend on fs size. 2268 * The computations of these are as follows: 2269 * 2270 * size = lfs_segtod(fs, nseg) 2271 * dsize = lfs_segtod(fs, nseg - minfreeseg) - lfs_btofsb(#super * LFS_SBPAD) 2272 * bfree = dsize - lfs_btofsb(fs, bsize * nseg / 2) - blocks_actually_used 2273 * avail = lfs_segtod(fs, nclean) - lfs_btofsb(#clean_super * LFS_SBPAD) 2274 * + (lfs_segtod(fs, 1) - (offset - curseg)) 2275 * - lfs_segtod(fs, minfreeseg - (minfreeseg / 2)) 2276 * 2277 * XXX - we should probably adjust minfreeseg as well. 2278 */ 2279 gain = (newnsegs - oldnsegs); 2280 fs->lfs_nseg = newnsegs; 2281 fs->lfs_segtabsz = nlast - fs->lfs_cleansz; 2282 fs->lfs_size += gain * lfs_btofsb(fs, fs->lfs_ssize); 2283 fs->lfs_dsize += gain * lfs_btofsb(fs, fs->lfs_ssize) - lfs_btofsb(fs, sbbytes); 2284 fs->lfs_bfree += gain * lfs_btofsb(fs, fs->lfs_ssize) - lfs_btofsb(fs, sbbytes) 2285 - gain * lfs_btofsb(fs, fs->lfs_bsize / 2); 2286 if (gain > 0) { 2287 fs->lfs_nclean += gain; 2288 fs->lfs_avail += gain * lfs_btofsb(fs, fs->lfs_ssize); 2289 } else { 2290 fs->lfs_nclean -= cgain; 2291 fs->lfs_avail -= cgain * lfs_btofsb(fs, fs->lfs_ssize) - 2292 lfs_btofsb(fs, csbbytes); 2293 } 2294 2295 /* Resize segment flag cache */ 2296 fs->lfs_suflags[0] = realloc(fs->lfs_suflags[0], 2297 fs->lfs_nseg * sizeof(u_int32_t), M_SEGMENT, M_WAITOK); 2298 fs->lfs_suflags[1] = realloc(fs->lfs_suflags[1], 2299 fs->lfs_nseg * sizeof(u_int32_t), M_SEGMENT, M_WAITOK); 2300 for (i = oldnsegs; i < newnsegs; i++) 2301 fs->lfs_suflags[0][i] = fs->lfs_suflags[1][i] = 0x0; 2302 2303 /* Truncate Ifile if necessary */ 2304 if (noff < 0) 2305 lfs_truncate(ivp, ivp->v_size + (noff << fs->lfs_bshift), 0, 2306 NOCRED); 2307 2308 /* Update cleaner info so the cleaner can die */ 2309 /* XXX what to do if bread fails? */ 2310 bread(ivp, 0, fs->lfs_bsize, NOCRED, B_MODIFY, &bp); 2311 ((CLEANERINFO *)bp->b_data)->clean = fs->lfs_nclean; 2312 ((CLEANERINFO *)bp->b_data)->dirty = fs->lfs_nseg - fs->lfs_nclean; 2313 VOP_BWRITE(bp->b_vp, bp); 2314 2315 /* Let Ifile accesses proceed */ 2316 rw_exit(&fs->lfs_iflock); 2317 2318 out: 2319 lfs_segunlock(fs); 2320 return error; 2321 } 2322 2323 /* 2324 * Extended attribute dispatch 2325 */ 2326 static int 2327 lfs_extattrctl(struct mount *mp, int cmd, struct vnode *vp, 2328 int attrnamespace, const char *attrname) 2329 { 2330 #ifdef LFS_EXTATTR 2331 struct ulfsmount *ump; 2332 2333 ump = VFSTOULFS(mp); 2334 if (ump->um_fstype == ULFS1) { 2335 return ulfs_extattrctl(mp, cmd, vp, attrnamespace, attrname); 2336 } 2337 #endif 2338 return vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname); 2339 } 2340