lfs_vfsops.c revision 1.36 1 /* $NetBSD: lfs_vfsops.c,v 1.36 1999/09/03 22:48:51 perseant Exp $ */
2
3 /*-
4 * Copyright (c) 1999 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) 1989, 1991, 1993, 1994
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. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Berkeley and its contributors.
54 * 4. Neither the name of the University nor the names of its contributors
55 * may be used to endorse or promote products derived from this software
56 * without specific prior written permission.
57 *
58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * SUCH DAMAGE.
69 *
70 * @(#)lfs_vfsops.c 8.20 (Berkeley) 6/10/95
71 */
72
73 #if defined(_KERNEL) && !defined(_LKM)
74 #include "opt_quota.h"
75 #endif
76
77 #include <sys/param.h>
78 #include <sys/systm.h>
79 #include <sys/namei.h>
80 #include <sys/proc.h>
81 #include <sys/kernel.h>
82 #include <sys/vnode.h>
83 #include <sys/mount.h>
84 #include <sys/buf.h>
85 #include <sys/mbuf.h>
86 #include <sys/file.h>
87 #include <sys/disklabel.h>
88 #include <sys/ioctl.h>
89 #include <sys/errno.h>
90 #include <sys/malloc.h>
91 #include <sys/pool.h>
92 #include <sys/socket.h>
93 #include <vm/vm.h>
94 #include <sys/sysctl.h>
95
96 #include <miscfs/specfs/specdev.h>
97
98 #include <ufs/ufs/quota.h>
99 #include <ufs/ufs/inode.h>
100 #include <ufs/ufs/ufsmount.h>
101 #include <ufs/ufs/ufs_extern.h>
102
103 #include <ufs/lfs/lfs.h>
104 #include <ufs/lfs/lfs_extern.h>
105
106 int lfs_mountfs __P((struct vnode *, struct mount *, struct proc *));
107
108 extern struct vnodeopv_desc lfs_vnodeop_opv_desc;
109 extern struct vnodeopv_desc lfs_specop_opv_desc;
110 extern struct vnodeopv_desc lfs_fifoop_opv_desc;
111
112 struct vnodeopv_desc *lfs_vnodeopv_descs[] = {
113 &lfs_vnodeop_opv_desc,
114 &lfs_specop_opv_desc,
115 &lfs_fifoop_opv_desc,
116 NULL,
117 };
118
119 struct vfsops lfs_vfsops = {
120 MOUNT_LFS,
121 lfs_mount,
122 ufs_start,
123 lfs_unmount,
124 ufs_root,
125 ufs_quotactl,
126 lfs_statfs,
127 lfs_sync,
128 lfs_vget,
129 lfs_fhtovp,
130 lfs_vptofh,
131 lfs_init,
132 lfs_sysctl,
133 lfs_mountroot,
134 ufs_check_export,
135 lfs_vnodeopv_descs,
136 };
137
138 struct pool lfs_inode_pool;
139
140 /*
141 * Initialize the filesystem, most work done by ufs_init.
142 */
143 void
144 lfs_init()
145 {
146 ufs_init();
147
148 /*
149 * XXX Same structure as FFS inodes? Should we share a common pool?
150 */
151 pool_init(&lfs_inode_pool, sizeof(struct inode), 0, 0, 0,
152 "lfsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
153 M_LFSNODE);
154 }
155
156 /*
157 * Called by main() when ufs is going to be mounted as root.
158 */
159 int
160 lfs_mountroot()
161 {
162 extern struct vnode *rootvp;
163 struct mount *mp;
164 struct proc *p = curproc; /* XXX */
165 int error;
166
167 /*
168 * Get vnodes for swapdev and rootdev.
169 */
170 if ((error = bdevvp(rootdev, &rootvp))) {
171 printf("lfs_mountroot: can't setup bdevvp's");
172 return (error);
173 }
174 if ((error = vfs_rootmountalloc(MOUNT_LFS, "root_device", &mp))) {
175 vrele(rootvp);
176 return (error);
177 }
178 if ((error = lfs_mountfs(rootvp, mp, p))) {
179 mp->mnt_op->vfs_refcount--;
180 vfs_unbusy(mp);
181 free(mp, M_MOUNT);
182 vrele(rootvp);
183 return (error);
184 }
185 simple_lock(&mountlist_slock);
186 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
187 simple_unlock(&mountlist_slock);
188 (void)lfs_statfs(mp, &mp->mnt_stat, p);
189 vfs_unbusy(mp);
190 return (0);
191 }
192
193 /*
194 * VFS Operations.
195 *
196 * mount system call
197 */
198 int
199 lfs_mount(mp, path, data, ndp, p)
200 register struct mount *mp;
201 const char *path;
202 void *data;
203 struct nameidata *ndp;
204 struct proc *p;
205 {
206 struct vnode *devvp;
207 struct ufs_args args;
208 struct ufsmount *ump = NULL;
209 register struct lfs *fs = NULL; /* LFS */
210 size_t size;
211 int error;
212 mode_t accessmode;
213
214 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
215 if (error)
216 return (error);
217
218 #if 0
219 /* Until LFS can do NFS right. XXX */
220 if (args.export.ex_flags & MNT_EXPORTED)
221 return (EINVAL);
222 #endif
223
224 /*
225 * If updating, check whether changing from read-only to
226 * read/write; if there is no device name, that's all we do.
227 */
228 if (mp->mnt_flag & MNT_UPDATE) {
229 ump = VFSTOUFS(mp);
230 fs = ump->um_lfs;
231 if (fs->lfs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
232 /*
233 * If upgrade to read-write by non-root, then verify
234 * that user has necessary permissions on the device.
235 */
236 if (p->p_ucred->cr_uid != 0) {
237 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
238 error = VOP_ACCESS(ump->um_devvp, VREAD|VWRITE,
239 p->p_ucred, p);
240 VOP_UNLOCK(ump->um_devvp, 0);
241 if (error)
242 return (error);
243 }
244 fs->lfs_ronly = 0;
245 }
246 if (args.fspec == 0) {
247 /*
248 * Process export requests.
249 */
250 return (vfs_export(mp, &ump->um_export, &args.export));
251 }
252 }
253 /*
254 * Not an update, or updating the name: look up the name
255 * and verify that it refers to a sensible block device.
256 */
257 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
258 if ((error = namei(ndp)) != 0)
259 return (error);
260 devvp = ndp->ni_vp;
261 if (devvp->v_type != VBLK) {
262 vrele(devvp);
263 return (ENOTBLK);
264 }
265 if (major(devvp->v_rdev) >= nblkdev) {
266 vrele(devvp);
267 return (ENXIO);
268 }
269 /*
270 * If mount by non-root, then verify that user has necessary
271 * permissions on the device.
272 */
273 if (p->p_ucred->cr_uid != 0) {
274 accessmode = VREAD;
275 if ((mp->mnt_flag & MNT_RDONLY) == 0)
276 accessmode |= VWRITE;
277 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
278 error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
279 if (error) {
280 vput(devvp);
281 return (error);
282 }
283 VOP_UNLOCK(devvp, 0);
284 }
285 if ((mp->mnt_flag & MNT_UPDATE) == 0)
286 error = lfs_mountfs(devvp, mp, p); /* LFS */
287 else {
288 if (devvp != ump->um_devvp)
289 error = EINVAL; /* needs translation */
290 else
291 vrele(devvp);
292 }
293 if (error) {
294 vrele(devvp);
295 return (error);
296 }
297 ump = VFSTOUFS(mp);
298 fs = ump->um_lfs; /* LFS */
299 (void)copyinstr(path, fs->lfs_fsmnt, sizeof(fs->lfs_fsmnt) - 1, &size);
300 bzero(fs->lfs_fsmnt + size, sizeof(fs->lfs_fsmnt) - size);
301 bcopy(fs->lfs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
302 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
303 &size);
304 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
305 return (0);
306 }
307
308 /*
309 * Common code for mount and mountroot
310 * LFS specific
311 */
312 int
313 lfs_mountfs(devvp, mp, p)
314 register struct vnode *devvp;
315 struct mount *mp;
316 struct proc *p;
317 {
318 extern struct vnode *rootvp;
319 struct dlfs *dfs, *adfs;
320 register struct lfs *fs;
321 register struct ufsmount *ump;
322 struct vnode *vp;
323 struct buf *bp, *abp;
324 struct partinfo dpart;
325 dev_t dev;
326 int error, i, ronly, size;
327 struct ucred *cred;
328 SEGUSE *sup;
329
330 cred = p ? p->p_ucred : NOCRED;
331 /*
332 * Disallow multiple mounts of the same device.
333 * Disallow mounting of a device that is currently in use
334 * (except for root, which might share swap device for miniroot).
335 * Flush out any old buffers remaining from a previous use.
336 */
337 if ((error = vfs_mountedon(devvp)) != 0)
338 return (error);
339 if (vcount(devvp) > 1 && devvp != rootvp)
340 return (EBUSY);
341 if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
342 return (error);
343
344 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
345 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
346 if (error)
347 return (error);
348 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
349 size = DEV_BSIZE;
350 else
351 size = dpart.disklab->d_secsize;
352
353 /* Don't free random space on error. */
354 bp = NULL;
355 ump = NULL;
356
357 /* Read in the superblock. */
358 error = bread(devvp, LFS_LABELPAD / size, LFS_SBPAD, cred, &bp);
359 if (error)
360 goto out;
361 dfs = (struct dlfs *)bp->b_data;
362
363 /*
364 * Check the second superblock to see which is newer; then mount
365 * using the older of the two. This is necessary to ensure that
366 * the filesystem is valid if it was not unmounted cleanly.
367 */
368 error = bread(devvp, dfs->dlfs_sboffs[1], LFS_SBPAD, cred, &abp);
369 if (error)
370 goto out;
371 adfs = (struct dlfs *)abp->b_data;
372
373 if (adfs->dlfs_tstamp < dfs->dlfs_tstamp) /* XXX KS - 1s resolution? */
374 dfs = adfs;
375
376 /* Check the basics. */
377 if (dfs->dlfs_magic != LFS_MAGIC || dfs->dlfs_bsize > MAXBSIZE ||
378 dfs->dlfs_version > LFS_VERSION ||
379 dfs->dlfs_bsize < sizeof(struct dlfs)) {
380 error = EINVAL; /* XXX needs translation */
381 goto out;
382 }
383
384 /* Allocate the mount structure, copy the superblock into it. */
385 fs = malloc(sizeof(struct lfs), M_UFSMNT, M_WAITOK);
386 memcpy(&fs->lfs_dlfs, dfs, sizeof(struct dlfs));
387 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
388 memset((caddr_t)ump, 0, sizeof *ump);
389 ump->um_lfs = fs;
390 if (sizeof(struct lfs) < LFS_SBPAD) /* XXX why? */
391 bp->b_flags |= B_INVAL;
392 brelse(bp);
393 bp = NULL;
394 brelse(abp);
395 abp = NULL;
396
397 /* Set up the I/O information */
398 fs->lfs_iocount = 0;
399 fs->lfs_dirvcount = 0;
400 fs->lfs_diropwait = 0;
401 fs->lfs_activesb = 0;
402 #ifdef LFS_CANNOT_ROLLFW
403 fs->lfs_sbactive = NULL;
404 #endif
405 #ifdef LFS_TRACK_IOS
406 for (i=0;i<LFS_THROTTLE;i++)
407 fs->lfs_pending[i] = LFS_UNUSED_DADDR;
408 #endif
409
410 /* Set up the ifile and lock aflags */
411 fs->lfs_doifile = 0;
412 fs->lfs_writer = 0;
413 fs->lfs_dirops = 0;
414 fs->lfs_seglock = 0;
415
416 /* Set the file system readonly/modify bits. */
417 fs->lfs_ronly = ronly;
418 if (ronly == 0)
419 fs->lfs_fmod = 1;
420
421 /* Initialize the mount structure. */
422 dev = devvp->v_rdev;
423 mp->mnt_data = (qaddr_t)ump;
424 mp->mnt_stat.f_fsid.val[0] = (long)dev;
425 mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_LFS);
426 mp->mnt_maxsymlinklen = fs->lfs_maxsymlinklen;
427 mp->mnt_flag |= MNT_LOCAL;
428 ump->um_flags = 0;
429 ump->um_mountp = mp;
430 ump->um_dev = dev;
431 ump->um_devvp = devvp;
432 ump->um_bptrtodb = 0;
433 ump->um_seqinc = 1 << fs->lfs_fsbtodb;
434 ump->um_nindir = fs->lfs_nindir;
435 for (i = 0; i < MAXQUOTAS; i++)
436 ump->um_quotas[i] = NULLVP;
437 devvp->v_specflags |= SI_MOUNTEDON;
438
439 /*
440 * We use the ifile vnode for almost every operation. Instead of
441 * retrieving it from the hash table each time we retrieve it here,
442 * artificially increment the reference count and keep a pointer
443 * to it in the incore copy of the superblock.
444 */
445 if ((error = VFS_VGET(mp, LFS_IFILE_INUM, &vp)) != 0)
446 goto out;
447 fs->lfs_ivnode = vp;
448 VREF(vp);
449 vput(vp);
450
451 /*
452 * Mark the current segment as ACTIVE, since we're going to
453 * be writing to it.
454 */
455 LFS_SEGENTRY(sup, fs, datosn(fs, fs->lfs_offset), bp);
456 sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE;
457 (void) VOP_BWRITE(bp);
458
459 return (0);
460 out:
461 if (bp)
462 brelse(bp);
463 if (abp)
464 brelse(abp);
465 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
466 if (ump) {
467 free(ump->um_lfs, M_UFSMNT);
468 free(ump, M_UFSMNT);
469 mp->mnt_data = (qaddr_t)0;
470 }
471 return (error);
472 }
473
474 /*
475 * unmount system call
476 */
477 int
478 lfs_unmount(mp, mntflags, p)
479 struct mount *mp;
480 int mntflags;
481 struct proc *p;
482 {
483 register struct ufsmount *ump;
484 register struct lfs *fs;
485 int error, flags, ronly;
486 extern int lfs_allclean_wakeup;
487
488 flags = 0;
489 if (mntflags & MNT_FORCE)
490 flags |= FORCECLOSE;
491
492 ump = VFSTOUFS(mp);
493 fs = ump->um_lfs;
494 #ifdef QUOTA
495 if (mp->mnt_flag & MNT_QUOTA) {
496 int i;
497 error = vflush(mp, fs->lfs_ivnode, SKIPSYSTEM|flags);
498 if (error)
499 return (error);
500 for (i = 0; i < MAXQUOTAS; i++) {
501 if (ump->um_quotas[i] == NULLVP)
502 continue;
503 quotaoff(p, mp, i);
504 }
505 /*
506 * Here we fall through to vflush again to ensure
507 * that we have gotten rid of all the system vnodes.
508 */
509 }
510 #endif
511 if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0)
512 return (error);
513 fs->lfs_clean = 1;
514 if ((error = VFS_SYNC(mp, 1, p->p_ucred, p)) != 0)
515 return (error);
516 if (fs->lfs_ivnode->v_dirtyblkhd.lh_first)
517 panic("lfs_unmount: still dirty blocks on ifile vnode\n");
518 vrele(fs->lfs_ivnode);
519 vgone(fs->lfs_ivnode);
520
521 ronly = !fs->lfs_ronly;
522 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
523 error = VOP_CLOSE(ump->um_devvp,
524 ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
525 vrele(ump->um_devvp);
526
527 /* XXX KS - wake up the cleaner so it can die */
528 wakeup(&fs->lfs_nextseg);
529 wakeup(&lfs_allclean_wakeup);
530
531 free(fs, M_UFSMNT);
532 free(ump, M_UFSMNT);
533 mp->mnt_data = (qaddr_t)0;
534 mp->mnt_flag &= ~MNT_LOCAL;
535 return (error);
536 }
537
538 /*
539 * Get file system statistics.
540 */
541 int
542 lfs_statfs(mp, sbp, p)
543 struct mount *mp;
544 register struct statfs *sbp;
545 struct proc *p;
546 {
547 register struct lfs *fs;
548 register struct ufsmount *ump;
549
550 ump = VFSTOUFS(mp);
551 fs = ump->um_lfs;
552 if (fs->lfs_magic != LFS_MAGIC)
553 panic("lfs_statfs: magic");
554 sbp->f_type = 0;
555 sbp->f_bsize = fs->lfs_fsize;
556 sbp->f_iosize = fs->lfs_bsize;
557 sbp->f_blocks = dbtofrags(fs, fs->lfs_dsize);
558 sbp->f_bfree = dbtofrags(fs, fs->lfs_bfree);
559 /*
560 * To compute the available space. Subtract the minimum free
561 * from the total number of blocks in the file system. Set avail
562 * to the smaller of this number and fs->lfs_bfree.
563 *
564 * XXX KS - is my modification below what is desired? (This
565 * will, e.g., change the report when the cleaner runs.)
566 */
567 #if 0
568 sbp->f_bavail = (long) ((u_int64_t) fs->lfs_dsize * (u_int64_t)
569 (100 - fs->lfs_minfree) / (u_int64_t) 100);
570 sbp->f_bavail =
571 sbp->f_bavail > fs->lfs_bfree ? fs->lfs_bfree : sbp->f_bavail;
572 #else
573 sbp->f_bavail = (long) ((u_int64_t) fs->lfs_dsize * (u_int64_t)
574 (100 - fs->lfs_minfree) / (u_int64_t) 100)
575 - (u_int64_t)(fs->lfs_dsize - fs->lfs_bfree);
576 #endif
577 sbp->f_bavail = dbtofrags(fs, sbp->f_bavail);
578 sbp->f_files = dbtofsb(fs,fs->lfs_bfree) * INOPB(fs);
579 sbp->f_ffree = sbp->f_files - fs->lfs_nfiles;
580 if (sbp != &mp->mnt_stat) {
581 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
582 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
583 }
584 strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
585 return (0);
586 }
587
588 /*
589 * Go through the disk queues to initiate sandbagged IO;
590 * go through the inodes to write those that have been modified;
591 * initiate the writing of the super block if it has been modified.
592 *
593 * Note: we are always called with the filesystem marked `MPBUSY'.
594 */
595 int
596 lfs_sync(mp, waitfor, cred, p)
597 struct mount *mp;
598 int waitfor;
599 struct ucred *cred;
600 struct proc *p;
601 {
602 int error;
603 struct lfs *fs;
604
605 fs = ((struct ufsmount *)mp->mnt_data)->ufsmount_u.lfs;
606 while(fs->lfs_dirops)
607 error = tsleep(&fs->lfs_dirops, PRIBIO + 1, "lfs_dirops", 0);
608 fs->lfs_writer++;
609
610 /* All syncs must be checkpoints until roll-forward is implemented. */
611 error = lfs_segwrite(mp, SEGM_CKP | (waitfor ? SEGM_SYNC : 0));
612 if(--fs->lfs_writer==0)
613 wakeup(&fs->lfs_dirops);
614 #ifdef QUOTA
615 qsync(mp);
616 #endif
617 return (error);
618 }
619
620 #ifdef USE_UFS_HASHLOCK
621 extern struct lock ufs_hashlock;
622 #endif
623
624 /*
625 * Look up an LFS dinode number to find its incore vnode. If not already
626 * in core, read it in from the specified device. Return the inode locked.
627 * Detection and handling of mount points must be done by the calling routine.
628 */
629 int
630 lfs_vget(mp, ino, vpp)
631 struct mount *mp;
632 ino_t ino;
633 struct vnode **vpp;
634 {
635 register struct lfs *fs;
636 register struct inode *ip;
637 struct buf *bp;
638 struct ifile *ifp;
639 struct vnode *vp;
640 struct ufsmount *ump;
641 ufs_daddr_t daddr;
642 dev_t dev;
643 int error;
644 #ifdef LFS_ATIME_IFILE
645 struct timespec ts;
646 #endif
647
648 ump = VFSTOUFS(mp);
649 dev = ump->um_dev;
650
651 #ifdef USE_UFS_HASHLOCK
652 do {
653 #endif
654 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
655 return (0);
656 #ifdef USE_UFS_HASHLOCK
657 } while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
658 #endif
659
660 /* Translate the inode number to a disk address. */
661 fs = ump->um_lfs;
662 if (ino == LFS_IFILE_INUM)
663 daddr = fs->lfs_idaddr;
664 else {
665 LFS_IENTRY(ifp, fs, ino, bp);
666 daddr = ifp->if_daddr;
667 #ifdef LFS_ATIME_IFILE
668 ts = ifp->if_atime; /* structure copy */
669 #endif
670 brelse(bp);
671 if (daddr == LFS_UNUSED_DADDR) {
672 #ifdef USE_UFS_HASHLOCK
673 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
674 #endif
675 return (ENOENT);
676 }
677 }
678
679 /* Allocate new vnode/inode. */
680 if ((error = lfs_vcreate(mp, ino, &vp)) != 0) {
681 *vpp = NULL;
682 #ifdef USE_UFS_HASHLOCK
683 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
684 #endif
685 return (error);
686 }
687
688 /*
689 * Put it onto its hash chain and lock it so that other requests for
690 * this inode will block if they arrive while we are sleeping waiting
691 * for old data structures to be purged or for the contents of the
692 * disk portion of this inode to be read.
693 */
694 ip = VTOI(vp);
695 ufs_ihashins(ip);
696 #ifdef USE_UFS_HASHLOCK
697 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
698 #endif
699
700 /*
701 * XXX
702 * This may not need to be here, logically it should go down with
703 * the i_devvp initialization.
704 * Ask Kirk.
705 */
706 ip->i_lfs = ump->um_lfs;
707
708 /* Read in the disk contents for the inode, copy into the inode. */
709 error = bread(ump->um_devvp, daddr, (int)fs->lfs_bsize, NOCRED, &bp);
710 if (error) {
711 /*
712 * The inode does not contain anything useful, so it would
713 * be misleading to leave it on its hash chain. With mode
714 * still zero, it will be unlinked and returned to the free
715 * list by vput().
716 */
717 vput(vp);
718 brelse(bp);
719 *vpp = NULL;
720 return (error);
721 }
722 ip->i_din.ffs_din = *lfs_ifind(fs, ino, (struct dinode *)bp->b_data);
723 #ifdef LFS_ATIME_IFILE
724 ip->i_ffs_atime = ts.tv_sec;
725 ip->i_ffs_atimensec = ts.tv_nsec;
726 #endif
727 brelse(bp);
728
729 /*
730 * Initialize the vnode from the inode, check for aliases. In all
731 * cases re-init ip, the underlying vnode/inode may have changed.
732 */
733 error = ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
734 if (error) {
735 vput(vp);
736 *vpp = NULL;
737 return (error);
738 }
739 if(vp->v_type == VNON) {
740 printf("lfs_vget: ino %d is type VNON! (ifmt %o)\n", ip->i_number, (ip->i_ffs_mode&IFMT)>>12);
741 #ifdef DDB
742 Debugger();
743 #endif
744 }
745 /*
746 * Finish inode initialization now that aliasing has been resolved.
747 */
748 ip->i_devvp = ump->um_devvp;
749 VREF(ip->i_devvp);
750 *vpp = vp;
751
752 return (0);
753 }
754
755 /*
756 * File handle to vnode
757 *
758 * Have to be really careful about stale file handles:
759 * - check that the inode number is valid
760 * - call lfs_vget() to get the locked inode
761 * - check for an unallocated inode (i_mode == 0)
762 *
763 * XXX
764 * use ifile to see if inode is allocated instead of reading off disk
765 * what is the relationship between my generational number and the NFS
766 * generational number.
767 */
768 int
769 lfs_fhtovp(mp, fhp, vpp)
770 register struct mount *mp;
771 struct fid *fhp;
772 struct vnode **vpp;
773 {
774 register struct ufid *ufhp;
775
776 ufhp = (struct ufid *)fhp;
777 if (ufhp->ufid_ino < ROOTINO)
778 return (ESTALE);
779 return (ufs_fhtovp(mp, ufhp, vpp));
780 }
781
782 /*
783 * Vnode pointer to File handle
784 */
785 /* ARGSUSED */
786 int
787 lfs_vptofh(vp, fhp)
788 struct vnode *vp;
789 struct fid *fhp;
790 {
791 register struct inode *ip;
792 register struct ufid *ufhp;
793
794 ip = VTOI(vp);
795 ufhp = (struct ufid *)fhp;
796 ufhp->ufid_len = sizeof(struct ufid);
797 ufhp->ufid_ino = ip->i_number;
798 ufhp->ufid_gen = ip->i_ffs_gen;
799 return (0);
800 }
801
802 int
803 lfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
804 int *name;
805 u_int namelen;
806 void *oldp;
807 size_t *oldlenp;
808 void *newp;
809 size_t newlen;
810 struct proc *p;
811 {
812 extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead;
813 extern struct lfs_stats lfs_stats;
814 int error;
815
816 /* all sysctl names at this level are terminal */
817 if (namelen != 1)
818 return (ENOTDIR);
819
820 switch (name[0]) {
821 case LFS_WRITEINDIR:
822 return (sysctl_int(oldp, oldlenp, newp, newlen,
823 &lfs_writeindir));
824 case LFS_CLEAN_VNHEAD:
825 return (sysctl_int(oldp, oldlenp, newp, newlen,
826 &lfs_clean_vnhead));
827 case LFS_DOSTATS:
828 if((error = sysctl_int(oldp, oldlenp, newp, newlen,
829 &lfs_dostats)))
830 return error;
831 if(lfs_dostats == 0)
832 memset(&lfs_stats,0,sizeof(lfs_stats));
833 return 0;
834 case LFS_STATS:
835 return (sysctl_rdstruct(oldp, oldlenp, newp,
836 &lfs_stats, sizeof(lfs_stats)));
837 default:
838 return (EOPNOTSUPP);
839 }
840 /* NOTREACHED */
841 }
842