lfs_vfsops.c revision 1.48 1 /* $NetBSD: lfs_vfsops.c,v 1.48 2000/03/30 12:41:13 augustss 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/device.h>
86 #include <sys/mbuf.h>
87 #include <sys/file.h>
88 #include <sys/disklabel.h>
89 #include <sys/ioctl.h>
90 #include <sys/errno.h>
91 #include <sys/malloc.h>
92 #include <sys/pool.h>
93 #include <sys/socket.h>
94 #include <vm/vm.h>
95 #include <sys/sysctl.h>
96
97 #include <miscfs/specfs/specdev.h>
98
99 #include <ufs/ufs/quota.h>
100 #include <ufs/ufs/inode.h>
101 #include <ufs/ufs/ufsmount.h>
102 #include <ufs/ufs/ufs_extern.h>
103
104 #include <ufs/lfs/lfs.h>
105 #include <ufs/lfs/lfs_extern.h>
106
107 int lfs_mountfs __P((struct vnode *, struct mount *, struct proc *));
108
109 extern struct vnodeopv_desc lfs_vnodeop_opv_desc;
110 extern struct vnodeopv_desc lfs_specop_opv_desc;
111 extern struct vnodeopv_desc lfs_fifoop_opv_desc;
112
113 struct vnodeopv_desc *lfs_vnodeopv_descs[] = {
114 &lfs_vnodeop_opv_desc,
115 &lfs_specop_opv_desc,
116 &lfs_fifoop_opv_desc,
117 NULL,
118 };
119
120 struct vfsops lfs_vfsops = {
121 MOUNT_LFS,
122 lfs_mount,
123 ufs_start,
124 lfs_unmount,
125 ufs_root,
126 ufs_quotactl,
127 lfs_statfs,
128 lfs_sync,
129 lfs_vget,
130 lfs_fhtovp,
131 lfs_vptofh,
132 lfs_init,
133 lfs_done,
134 lfs_sysctl,
135 lfs_mountroot,
136 ufs_check_export,
137 lfs_vnodeopv_descs,
138 };
139
140 struct pool lfs_inode_pool;
141
142 /*
143 * Initialize the filesystem, most work done by ufs_init.
144 */
145 void
146 lfs_init()
147 {
148 ufs_init();
149
150 /*
151 * XXX Same structure as FFS inodes? Should we share a common pool?
152 */
153 pool_init(&lfs_inode_pool, sizeof(struct inode), 0, 0, 0,
154 "lfsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
155 M_LFSNODE);
156 }
157
158 void
159 lfs_done()
160 {
161 ufs_done();
162 pool_destroy(&lfs_inode_pool);
163 }
164
165 /*
166 * Called by main() when ufs is going to be mounted as root.
167 */
168 int
169 lfs_mountroot()
170 {
171 extern struct vnode *rootvp;
172 struct mount *mp;
173 struct proc *p = curproc; /* XXX */
174 int error;
175
176 if (root_device->dv_class != DV_DISK)
177 return (ENODEV);
178
179 if (rootdev == NODEV)
180 return (ENODEV);
181 /*
182 * Get vnodes for swapdev and rootdev.
183 */
184 if ((error = bdevvp(rootdev, &rootvp))) {
185 printf("lfs_mountroot: can't setup bdevvp's");
186 return (error);
187 }
188 if ((error = vfs_rootmountalloc(MOUNT_LFS, "root_device", &mp))) {
189 vrele(rootvp);
190 return (error);
191 }
192 if ((error = lfs_mountfs(rootvp, mp, p))) {
193 mp->mnt_op->vfs_refcount--;
194 vfs_unbusy(mp);
195 free(mp, M_MOUNT);
196 vrele(rootvp);
197 return (error);
198 }
199 simple_lock(&mountlist_slock);
200 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
201 simple_unlock(&mountlist_slock);
202 (void)lfs_statfs(mp, &mp->mnt_stat, p);
203 vfs_unbusy(mp);
204 return (0);
205 }
206
207 /*
208 * VFS Operations.
209 *
210 * mount system call
211 */
212 int
213 lfs_mount(mp, path, data, ndp, p)
214 struct mount *mp;
215 const char *path;
216 void *data;
217 struct nameidata *ndp;
218 struct proc *p;
219 {
220 struct vnode *devvp;
221 struct ufs_args args;
222 struct ufsmount *ump = NULL;
223 struct lfs *fs = NULL; /* LFS */
224 size_t size;
225 int error;
226 mode_t accessmode;
227
228 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
229 if (error)
230 return (error);
231
232 #if 0
233 /* Until LFS can do NFS right. XXX */
234 if (args.export.ex_flags & MNT_EXPORTED)
235 return (EINVAL);
236 #endif
237
238 /*
239 * If updating, check whether changing from read-only to
240 * read/write; if there is no device name, that's all we do.
241 */
242 if (mp->mnt_flag & MNT_UPDATE) {
243 ump = VFSTOUFS(mp);
244 fs = ump->um_lfs;
245 if (fs->lfs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
246 /*
247 * If upgrade to read-write by non-root, then verify
248 * that user has necessary permissions on the device.
249 */
250 if (p->p_ucred->cr_uid != 0) {
251 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
252 error = VOP_ACCESS(ump->um_devvp, VREAD|VWRITE,
253 p->p_ucred, p);
254 VOP_UNLOCK(ump->um_devvp, 0);
255 if (error)
256 return (error);
257 }
258 fs->lfs_ronly = 0;
259 }
260 if (args.fspec == 0) {
261 /*
262 * Process export requests.
263 */
264 return (vfs_export(mp, &ump->um_export, &args.export));
265 }
266 }
267 /*
268 * Not an update, or updating the name: look up the name
269 * and verify that it refers to a sensible block device.
270 */
271 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
272 if ((error = namei(ndp)) != 0)
273 return (error);
274 devvp = ndp->ni_vp;
275 if (devvp->v_type != VBLK) {
276 vrele(devvp);
277 return (ENOTBLK);
278 }
279 if (major(devvp->v_rdev) >= nblkdev) {
280 vrele(devvp);
281 return (ENXIO);
282 }
283 /*
284 * If mount by non-root, then verify that user has necessary
285 * permissions on the device.
286 */
287 if (p->p_ucred->cr_uid != 0) {
288 accessmode = VREAD;
289 if ((mp->mnt_flag & MNT_RDONLY) == 0)
290 accessmode |= VWRITE;
291 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
292 error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
293 if (error) {
294 vput(devvp);
295 return (error);
296 }
297 VOP_UNLOCK(devvp, 0);
298 }
299 if ((mp->mnt_flag & MNT_UPDATE) == 0)
300 error = lfs_mountfs(devvp, mp, p); /* LFS */
301 else {
302 if (devvp != ump->um_devvp)
303 error = EINVAL; /* needs translation */
304 else
305 vrele(devvp);
306 }
307 if (error) {
308 vrele(devvp);
309 return (error);
310 }
311 ump = VFSTOUFS(mp);
312 fs = ump->um_lfs; /* LFS */
313 (void)copyinstr(path, fs->lfs_fsmnt, sizeof(fs->lfs_fsmnt) - 1, &size);
314 bzero(fs->lfs_fsmnt + size, sizeof(fs->lfs_fsmnt) - size);
315 bcopy(fs->lfs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
316 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
317 &size);
318 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
319 return (0);
320 }
321
322 /*
323 * Common code for mount and mountroot
324 * LFS specific
325 */
326 int
327 lfs_mountfs(devvp, mp, p)
328 struct vnode *devvp;
329 struct mount *mp;
330 struct proc *p;
331 {
332 extern struct vnode *rootvp;
333 struct dlfs *dfs, *adfs;
334 struct lfs *fs;
335 struct ufsmount *ump;
336 struct vnode *vp;
337 struct buf *bp, *abp;
338 struct partinfo dpart;
339 dev_t dev;
340 int error, i, ronly, size;
341 struct ucred *cred;
342 SEGUSE *sup;
343
344 cred = p ? p->p_ucred : NOCRED;
345 /*
346 * Disallow multiple mounts of the same device.
347 * Disallow mounting of a device that is currently in use
348 * (except for root, which might share swap device for miniroot).
349 * Flush out any old buffers remaining from a previous use.
350 */
351 if ((error = vfs_mountedon(devvp)) != 0)
352 return (error);
353 if (vcount(devvp) > 1 && devvp != rootvp)
354 return (EBUSY);
355 if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
356 return (error);
357
358 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
359 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
360 if (error)
361 return (error);
362 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
363 size = DEV_BSIZE;
364 else
365 size = dpart.disklab->d_secsize;
366
367 /* Don't free random space on error. */
368 bp = NULL;
369 ump = NULL;
370
371 /* Read in the superblock. */
372 error = bread(devvp, LFS_LABELPAD / size, LFS_SBPAD, cred, &bp);
373 if (error)
374 goto out;
375 dfs = (struct dlfs *)bp->b_data;
376
377 /*
378 * Check the second superblock to see which is newer; then mount
379 * using the older of the two. This is necessary to ensure that
380 * the filesystem is valid if it was not unmounted cleanly.
381 */
382 error = bread(devvp, dfs->dlfs_sboffs[1], LFS_SBPAD, cred, &abp);
383 if (error)
384 goto out;
385 adfs = (struct dlfs *)abp->b_data;
386
387 if (adfs->dlfs_tstamp < dfs->dlfs_tstamp) /* XXX KS - 1s resolution? */
388 dfs = adfs;
389
390 /* Check the basics. */
391 if (dfs->dlfs_magic != LFS_MAGIC || dfs->dlfs_bsize > MAXBSIZE ||
392 dfs->dlfs_version > LFS_VERSION ||
393 dfs->dlfs_bsize < sizeof(struct dlfs)) {
394 error = EINVAL; /* XXX needs translation */
395 goto out;
396 }
397
398 /* Allocate the mount structure, copy the superblock into it. */
399 fs = malloc(sizeof(struct lfs), M_UFSMNT, M_WAITOK);
400 memcpy(&fs->lfs_dlfs, dfs, sizeof(struct dlfs));
401 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
402 memset((caddr_t)ump, 0, sizeof *ump);
403 ump->um_lfs = fs;
404 if (sizeof(struct lfs) < LFS_SBPAD) /* XXX why? */
405 bp->b_flags |= B_INVAL;
406 brelse(bp);
407 bp = NULL;
408 brelse(abp);
409 abp = NULL;
410
411 /* Set up the I/O information */
412 fs->lfs_iocount = 0;
413 fs->lfs_diropwait = 0;
414 fs->lfs_activesb = 0;
415 #ifdef LFS_CANNOT_ROLLFW
416 fs->lfs_sbactive = NULL;
417 #endif
418 #ifdef LFS_TRACK_IOS
419 for (i=0;i<LFS_THROTTLE;i++)
420 fs->lfs_pending[i] = LFS_UNUSED_DADDR;
421 #endif
422
423 /* Set up the ifile and lock aflags */
424 fs->lfs_doifile = 0;
425 fs->lfs_writer = 0;
426 fs->lfs_dirops = 0;
427 fs->lfs_seglock = 0;
428 lockinit(&fs->lfs_freelock, PINOD, "lfs_freelock", 0, 0);
429
430 /* Set the file system readonly/modify bits. */
431 fs->lfs_ronly = ronly;
432 if (ronly == 0)
433 fs->lfs_fmod = 1;
434
435 /* Initialize the mount structure. */
436 dev = devvp->v_rdev;
437 mp->mnt_data = (qaddr_t)ump;
438 mp->mnt_stat.f_fsid.val[0] = (long)dev;
439 mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_LFS);
440 mp->mnt_maxsymlinklen = fs->lfs_maxsymlinklen;
441 mp->mnt_flag |= MNT_LOCAL;
442 ump->um_flags = 0;
443 ump->um_mountp = mp;
444 ump->um_dev = dev;
445 ump->um_devvp = devvp;
446 ump->um_bptrtodb = 0;
447 ump->um_seqinc = 1 << fs->lfs_fsbtodb;
448 ump->um_nindir = fs->lfs_nindir;
449 for (i = 0; i < MAXQUOTAS; i++)
450 ump->um_quotas[i] = NULLVP;
451 devvp->v_specmountpoint = mp;
452
453 /*
454 * We use the ifile vnode for almost every operation. Instead of
455 * retrieving it from the hash table each time we retrieve it here,
456 * artificially increment the reference count and keep a pointer
457 * to it in the incore copy of the superblock.
458 */
459 if ((error = VFS_VGET(mp, LFS_IFILE_INUM, &vp)) != 0)
460 goto out;
461 fs->lfs_ivnode = vp;
462 VREF(vp);
463 vput(vp);
464
465 /*
466 * Mark the current segment as ACTIVE, since we're going to
467 * be writing to it.
468 */
469 LFS_SEGENTRY(sup, fs, datosn(fs, fs->lfs_offset), bp);
470 sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE;
471 (void) VOP_BWRITE(bp);
472
473 return (0);
474 out:
475 if (bp)
476 brelse(bp);
477 if (abp)
478 brelse(abp);
479 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
480 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
481 VOP_UNLOCK(devvp, 0);
482 if (ump) {
483 free(ump->um_lfs, M_UFSMNT);
484 free(ump, M_UFSMNT);
485 mp->mnt_data = (qaddr_t)0;
486 }
487 return (error);
488 }
489
490 /*
491 * unmount system call
492 */
493 int
494 lfs_unmount(mp, mntflags, p)
495 struct mount *mp;
496 int mntflags;
497 struct proc *p;
498 {
499 struct ufsmount *ump;
500 struct lfs *fs;
501 int error, flags, ronly;
502 extern int lfs_allclean_wakeup;
503
504 flags = 0;
505 if (mntflags & MNT_FORCE)
506 flags |= FORCECLOSE;
507
508 ump = VFSTOUFS(mp);
509 fs = ump->um_lfs;
510 #ifdef QUOTA
511 if (mp->mnt_flag & MNT_QUOTA) {
512 int i;
513 error = vflush(mp, fs->lfs_ivnode, SKIPSYSTEM|flags);
514 if (error)
515 return (error);
516 for (i = 0; i < MAXQUOTAS; i++) {
517 if (ump->um_quotas[i] == NULLVP)
518 continue;
519 quotaoff(p, mp, i);
520 }
521 /*
522 * Here we fall through to vflush again to ensure
523 * that we have gotten rid of all the system vnodes.
524 */
525 }
526 #endif
527 if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0)
528 return (error);
529 fs->lfs_clean = 1;
530 if ((error = VFS_SYNC(mp, 1, p->p_ucred, p)) != 0)
531 return (error);
532 if (fs->lfs_ivnode->v_dirtyblkhd.lh_first)
533 panic("lfs_unmount: still dirty blocks on ifile vnode\n");
534 vrele(fs->lfs_ivnode);
535 vgone(fs->lfs_ivnode);
536
537 ronly = !fs->lfs_ronly;
538 if (ump->um_devvp->v_type != VBAD)
539 ump->um_devvp->v_specmountpoint = NULL;
540 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
541 error = VOP_CLOSE(ump->um_devvp,
542 ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
543 vput(ump->um_devvp);
544
545 /* XXX KS - wake up the cleaner so it can die */
546 wakeup(&fs->lfs_nextseg);
547 wakeup(&lfs_allclean_wakeup);
548
549 free(fs, M_UFSMNT);
550 free(ump, M_UFSMNT);
551 mp->mnt_data = (qaddr_t)0;
552 mp->mnt_flag &= ~MNT_LOCAL;
553 return (error);
554 }
555
556 /*
557 * Get file system statistics.
558 */
559 int
560 lfs_statfs(mp, sbp, p)
561 struct mount *mp;
562 struct statfs *sbp;
563 struct proc *p;
564 {
565 struct lfs *fs;
566 struct ufsmount *ump;
567
568 ump = VFSTOUFS(mp);
569 fs = ump->um_lfs;
570 if (fs->lfs_magic != LFS_MAGIC)
571 panic("lfs_statfs: magic");
572 sbp->f_type = 0;
573 sbp->f_bsize = fs->lfs_fsize;
574 sbp->f_iosize = fs->lfs_bsize;
575 sbp->f_blocks = dbtofrags(fs, fs->lfs_dsize);
576 sbp->f_bfree = dbtofrags(fs, fs->lfs_bfree);
577 /*
578 * To compute the available space. Subtract the minimum free
579 * from the total number of blocks in the file system. Set avail
580 * to the smaller of this number and fs->lfs_bfree.
581 *
582 * XXX KS - is my modification below what is desired? (This
583 * will, e.g., change the report when the cleaner runs.)
584 */
585 #if 0
586 sbp->f_bavail = (long) ((u_int64_t) fs->lfs_dsize * (u_int64_t)
587 (100 - fs->lfs_minfree) / (u_int64_t) 100);
588 sbp->f_bavail =
589 sbp->f_bavail > fs->lfs_bfree ? fs->lfs_bfree : sbp->f_bavail;
590 #else
591 sbp->f_bavail = (long) ((u_int64_t) fs->lfs_dsize * (u_int64_t)
592 (100 - fs->lfs_minfree) / (u_int64_t) 100)
593 - (u_int64_t)(fs->lfs_dsize - fs->lfs_bfree);
594 #endif
595 sbp->f_bavail = dbtofrags(fs, sbp->f_bavail);
596 sbp->f_files = dbtofsb(fs,fs->lfs_bfree) * INOPB(fs);
597 sbp->f_ffree = sbp->f_files - fs->lfs_nfiles;
598 if (sbp != &mp->mnt_stat) {
599 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
600 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
601 }
602 strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
603 return (0);
604 }
605
606 /*
607 * Go through the disk queues to initiate sandbagged IO;
608 * go through the inodes to write those that have been modified;
609 * initiate the writing of the super block if it has been modified.
610 *
611 * Note: we are always called with the filesystem marked `MPBUSY'.
612 */
613 int
614 lfs_sync(mp, waitfor, cred, p)
615 struct mount *mp;
616 int waitfor;
617 struct ucred *cred;
618 struct proc *p;
619 {
620 int error;
621 struct lfs *fs;
622
623 fs = ((struct ufsmount *)mp->mnt_data)->ufsmount_u.lfs;
624 while(fs->lfs_dirops)
625 error = tsleep(&fs->lfs_dirops, PRIBIO + 1, "lfs_dirops", 0);
626 fs->lfs_writer++;
627
628 /* All syncs must be checkpoints until roll-forward is implemented. */
629 error = lfs_segwrite(mp, SEGM_CKP | (waitfor ? SEGM_SYNC : 0));
630 if(--fs->lfs_writer==0)
631 wakeup(&fs->lfs_dirops);
632 #ifdef QUOTA
633 qsync(mp);
634 #endif
635 return (error);
636 }
637
638 extern struct lock ufs_hashlock;
639
640 /*
641 * Look up an LFS dinode number to find its incore vnode. If not already
642 * in core, read it in from the specified device. Return the inode locked.
643 * Detection and handling of mount points must be done by the calling routine.
644 */
645 int
646 lfs_vget(mp, ino, vpp)
647 struct mount *mp;
648 ino_t ino;
649 struct vnode **vpp;
650 {
651 struct lfs *fs;
652 struct inode *ip;
653 struct buf *bp;
654 struct ifile *ifp;
655 struct vnode *vp;
656 struct ufsmount *ump;
657 ufs_daddr_t daddr;
658 dev_t dev;
659 int error;
660 #ifdef LFS_ATIME_IFILE
661 struct timespec ts;
662 #endif
663
664 ump = VFSTOUFS(mp);
665 dev = ump->um_dev;
666
667 do {
668 if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
669 return (0);
670 } while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
671
672 /* Translate the inode number to a disk address. */
673 fs = ump->um_lfs;
674 if (ino == LFS_IFILE_INUM)
675 daddr = fs->lfs_idaddr;
676 else {
677 LFS_IENTRY(ifp, fs, ino, bp);
678 daddr = ifp->if_daddr;
679 #ifdef LFS_ATIME_IFILE
680 ts = ifp->if_atime; /* structure copy */
681 #endif
682 brelse(bp);
683 if (daddr == LFS_UNUSED_DADDR) {
684 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
685 return (ENOENT);
686 }
687 }
688
689 /* Allocate new vnode/inode. */
690 if ((error = lfs_vcreate(mp, ino, &vp)) != 0) {
691 *vpp = NULL;
692 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
693 return (error);
694 }
695
696 /*
697 * Put it onto its hash chain and lock it so that other requests for
698 * this inode will block if they arrive while we are sleeping waiting
699 * for old data structures to be purged or for the contents of the
700 * disk portion of this inode to be read.
701 */
702 ip = VTOI(vp);
703 ufs_ihashins(ip);
704 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
705
706 /*
707 * XXX
708 * This may not need to be here, logically it should go down with
709 * the i_devvp initialization.
710 * Ask Kirk.
711 */
712 ip->i_lfs = ump->um_lfs;
713
714 /* Read in the disk contents for the inode, copy into the inode. */
715 error = bread(ump->um_devvp, daddr, (int)fs->lfs_bsize, NOCRED, &bp);
716 if (error) {
717 /*
718 * The inode does not contain anything useful, so it would
719 * be misleading to leave it on its hash chain. With mode
720 * still zero, it will be unlinked and returned to the free
721 * list by vput().
722 */
723 vput(vp);
724 brelse(bp);
725 *vpp = NULL;
726 return (error);
727 }
728 ip->i_din.ffs_din = *lfs_ifind(fs, ino, bp);
729 ip->i_ffs_effnlink = ip->i_ffs_nlink;
730 #ifdef LFS_ATIME_IFILE
731 ip->i_ffs_atime = ts.tv_sec;
732 ip->i_ffs_atimensec = ts.tv_nsec;
733 #endif
734 brelse(bp);
735
736 /*
737 * Initialize the vnode from the inode, check for aliases. In all
738 * cases re-init ip, the underlying vnode/inode may have changed.
739 */
740 error = ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
741 if (error) {
742 vput(vp);
743 *vpp = NULL;
744 return (error);
745 }
746 if(vp->v_type == VNON) {
747 printf("lfs_vget: ino %d is type VNON! (ifmt %o)\n", ip->i_number, (ip->i_ffs_mode&IFMT)>>12);
748 #ifdef DDB
749 Debugger();
750 #endif
751 }
752 /*
753 * Finish inode initialization now that aliasing has been resolved.
754 */
755 ip->i_devvp = ump->um_devvp;
756 VREF(ip->i_devvp);
757 *vpp = vp;
758
759 return (0);
760 }
761
762 /*
763 * File handle to vnode
764 *
765 * Have to be really careful about stale file handles:
766 * - check that the inode number is valid
767 * - call lfs_vget() to get the locked inode
768 * - check for an unallocated inode (i_mode == 0)
769 *
770 * XXX
771 * use ifile to see if inode is allocated instead of reading off disk
772 * what is the relationship between my generational number and the NFS
773 * generational number.
774 */
775 int
776 lfs_fhtovp(mp, fhp, vpp)
777 struct mount *mp;
778 struct fid *fhp;
779 struct vnode **vpp;
780 {
781 struct ufid *ufhp;
782
783 ufhp = (struct ufid *)fhp;
784 if (ufhp->ufid_ino < ROOTINO)
785 return (ESTALE);
786 return (ufs_fhtovp(mp, ufhp, vpp));
787 }
788
789 /*
790 * Vnode pointer to File handle
791 */
792 /* ARGSUSED */
793 int
794 lfs_vptofh(vp, fhp)
795 struct vnode *vp;
796 struct fid *fhp;
797 {
798 struct inode *ip;
799 struct ufid *ufhp;
800
801 ip = VTOI(vp);
802 ufhp = (struct ufid *)fhp;
803 ufhp->ufid_len = sizeof(struct ufid);
804 ufhp->ufid_ino = ip->i_number;
805 ufhp->ufid_gen = ip->i_ffs_gen;
806 return (0);
807 }
808
809 int
810 lfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
811 int *name;
812 u_int namelen;
813 void *oldp;
814 size_t *oldlenp;
815 void *newp;
816 size_t newlen;
817 struct proc *p;
818 {
819 extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead;
820 extern struct lfs_stats lfs_stats;
821 int error;
822
823 /* all sysctl names at this level are terminal */
824 if (namelen != 1)
825 return (ENOTDIR);
826
827 switch (name[0]) {
828 case LFS_WRITEINDIR:
829 return (sysctl_int(oldp, oldlenp, newp, newlen,
830 &lfs_writeindir));
831 case LFS_CLEAN_VNHEAD:
832 return (sysctl_int(oldp, oldlenp, newp, newlen,
833 &lfs_clean_vnhead));
834 case LFS_DOSTATS:
835 if((error = sysctl_int(oldp, oldlenp, newp, newlen,
836 &lfs_dostats)))
837 return error;
838 if(lfs_dostats == 0)
839 memset(&lfs_stats,0,sizeof(lfs_stats));
840 return 0;
841 case LFS_STATS:
842 return (sysctl_rdstruct(oldp, oldlenp, newp,
843 &lfs_stats, sizeof(lfs_stats)));
844 default:
845 return (EOPNOTSUPP);
846 }
847 /* NOTREACHED */
848 }
849