lfs_vfsops.c revision 1.58 1 /* $NetBSD: lfs_vfsops.c,v 1.58 2000/09/09 04:49:55 perseant Exp $ */
2
3 /*-
4 * Copyright (c) 1999, 2000 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 <uvm/uvm_extern.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 abp = NULL;
370 ump = NULL;
371
372 /* Read in the superblock. */
373 error = bread(devvp, LFS_LABELPAD / size, LFS_SBPAD, cred, &bp);
374 if (error)
375 goto out;
376 dfs = (struct dlfs *)bp->b_data;
377
378 /* Check the basics. */
379 if (dfs->dlfs_magic != LFS_MAGIC || dfs->dlfs_bsize > MAXBSIZE ||
380 dfs->dlfs_version > LFS_VERSION ||
381 dfs->dlfs_bsize < sizeof(struct dlfs)) {
382 error = EINVAL; /* XXX needs translation */
383 goto out;
384 }
385
386 /*
387 * Check the second superblock to see which is newer; then mount
388 * using the older of the two. This is necessary to ensure that
389 * the filesystem is valid if it was not unmounted cleanly.
390 */
391 if (dfs->dlfs_sboffs[1] &&
392 dfs->dlfs_sboffs[1]-(LFS_LABELPAD/size) > LFS_SBPAD/size)
393 {
394 error = bread(devvp, dfs->dlfs_sboffs[1], LFS_SBPAD, cred, &abp);
395 if (error)
396 goto out;
397 adfs = (struct dlfs *)abp->b_data;
398
399 if (adfs->dlfs_tstamp < dfs->dlfs_tstamp) /* XXX 1s? */
400 dfs = adfs;
401 } else {
402 printf("lfs_mountfs: invalid alt superblock daddr=0x%x\n",
403 dfs->dlfs_sboffs[1]);
404 error = EINVAL;
405 goto out;
406 }
407
408 /* Allocate the mount structure, copy the superblock into it. */
409 fs = malloc(sizeof(struct lfs), M_UFSMNT, M_WAITOK);
410 memcpy(&fs->lfs_dlfs, dfs, sizeof(struct dlfs));
411 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
412 memset((caddr_t)ump, 0, sizeof *ump);
413 ump->um_lfs = fs;
414 if (sizeof(struct lfs) < LFS_SBPAD) /* XXX why? */
415 bp->b_flags |= B_INVAL;
416 brelse(bp);
417 bp = NULL;
418 brelse(abp);
419 abp = NULL;
420
421 /* Set up the I/O information */
422 fs->lfs_iocount = 0;
423 fs->lfs_diropwait = 0;
424 fs->lfs_activesb = 0;
425 fs->lfs_uinodes = 0;
426 fs->lfs_ravail = 0;
427 #ifdef LFS_CANNOT_ROLLFW
428 fs->lfs_sbactive = 0;
429 #endif
430 #ifdef LFS_TRACK_IOS
431 for (i=0;i<LFS_THROTTLE;i++)
432 fs->lfs_pending[i] = LFS_UNUSED_DADDR;
433 #endif
434
435 /* Set up the ifile and lock aflags */
436 fs->lfs_doifile = 0;
437 fs->lfs_writer = 0;
438 fs->lfs_dirops = 0;
439 fs->lfs_nadirop = 0;
440 fs->lfs_seglock = 0;
441 lockinit(&fs->lfs_freelock, PINOD, "lfs_freelock", 0, 0);
442
443 /* Set the file system readonly/modify bits. */
444 fs->lfs_ronly = ronly;
445 if (ronly == 0)
446 fs->lfs_fmod = 1;
447
448 /* Initialize the mount structure. */
449 dev = devvp->v_rdev;
450 mp->mnt_data = (qaddr_t)ump;
451 mp->mnt_stat.f_fsid.val[0] = (long)dev;
452 mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_LFS);
453 mp->mnt_stat.f_iosize = fs->lfs_bsize;
454 mp->mnt_maxsymlinklen = fs->lfs_maxsymlinklen;
455 mp->mnt_flag |= MNT_LOCAL;
456 ump->um_flags = 0;
457 ump->um_mountp = mp;
458 ump->um_dev = dev;
459 ump->um_devvp = devvp;
460 ump->um_bptrtodb = 0;
461 ump->um_seqinc = 1 << fs->lfs_fsbtodb;
462 ump->um_nindir = fs->lfs_nindir;
463 for (i = 0; i < MAXQUOTAS; i++)
464 ump->um_quotas[i] = NULLVP;
465 devvp->v_specmountpoint = mp;
466
467 /*
468 * We use the ifile vnode for almost every operation. Instead of
469 * retrieving it from the hash table each time we retrieve it here,
470 * artificially increment the reference count and keep a pointer
471 * to it in the incore copy of the superblock.
472 */
473 if ((error = VFS_VGET(mp, LFS_IFILE_INUM, &vp)) != 0)
474 goto out;
475 fs->lfs_ivnode = vp;
476 VREF(vp);
477 vput(vp);
478
479 /*
480 * Mark the current segment as ACTIVE, since we're going to
481 * be writing to it.
482 */
483 LFS_SEGENTRY(sup, fs, datosn(fs, fs->lfs_offset), bp);
484 sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE;
485 (void) VOP_BWRITE(bp);
486
487 return (0);
488 out:
489 if (bp)
490 brelse(bp);
491 if (abp)
492 brelse(abp);
493 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
494 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
495 VOP_UNLOCK(devvp, 0);
496 if (ump) {
497 free(ump->um_lfs, M_UFSMNT);
498 free(ump, M_UFSMNT);
499 mp->mnt_data = (qaddr_t)0;
500 }
501 return (error);
502 }
503
504 /*
505 * unmount system call
506 */
507 int
508 lfs_unmount(mp, mntflags, p)
509 struct mount *mp;
510 int mntflags;
511 struct proc *p;
512 {
513 struct ufsmount *ump;
514 struct lfs *fs;
515 int error, flags, ronly;
516 extern int lfs_allclean_wakeup;
517
518 flags = 0;
519 if (mntflags & MNT_FORCE)
520 flags |= FORCECLOSE;
521
522 ump = VFSTOUFS(mp);
523 fs = ump->um_lfs;
524 #ifdef QUOTA
525 if (mp->mnt_flag & MNT_QUOTA) {
526 int i;
527 error = vflush(mp, fs->lfs_ivnode, SKIPSYSTEM|flags);
528 if (error)
529 return (error);
530 for (i = 0; i < MAXQUOTAS; i++) {
531 if (ump->um_quotas[i] == NULLVP)
532 continue;
533 quotaoff(p, mp, i);
534 }
535 /*
536 * Here we fall through to vflush again to ensure
537 * that we have gotten rid of all the system vnodes.
538 */
539 }
540 #endif
541 if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0)
542 return (error);
543 fs->lfs_clean = 1;
544 if ((error = VFS_SYNC(mp, 1, p->p_ucred, p)) != 0)
545 return (error);
546 if (fs->lfs_ivnode->v_dirtyblkhd.lh_first)
547 panic("lfs_unmount: still dirty blocks on ifile vnode\n");
548 vrele(fs->lfs_ivnode);
549 vgone(fs->lfs_ivnode);
550
551 ronly = !fs->lfs_ronly;
552 if (ump->um_devvp->v_type != VBAD)
553 ump->um_devvp->v_specmountpoint = NULL;
554 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
555 error = VOP_CLOSE(ump->um_devvp,
556 ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
557 vput(ump->um_devvp);
558
559 /* XXX KS - wake up the cleaner so it can die */
560 wakeup(&fs->lfs_nextseg);
561 wakeup(&lfs_allclean_wakeup);
562
563 free(fs, M_UFSMNT);
564 free(ump, M_UFSMNT);
565 mp->mnt_data = (qaddr_t)0;
566 mp->mnt_flag &= ~MNT_LOCAL;
567 return (error);
568 }
569
570 /*
571 * Get file system statistics.
572 */
573 int
574 lfs_statfs(mp, sbp, p)
575 struct mount *mp;
576 struct statfs *sbp;
577 struct proc *p;
578 {
579 struct lfs *fs;
580 struct ufsmount *ump;
581
582 ump = VFSTOUFS(mp);
583 fs = ump->um_lfs;
584 if (fs->lfs_magic != LFS_MAGIC)
585 panic("lfs_statfs: magic");
586
587 sbp->f_type = 0;
588 sbp->f_bsize = fs->lfs_fsize;
589 sbp->f_iosize = fs->lfs_bsize;
590 sbp->f_blocks = dbtofrags(fs, LFS_EST_NONMETA(fs));
591 sbp->f_bfree = dbtofrags(fs, LFS_EST_BFREE(fs));
592 sbp->f_bavail = dbtofrags(fs, (long)LFS_EST_BFREE(fs) -
593 (long)LFS_EST_RSVD(fs));
594 sbp->f_files = dbtofsb(fs,fs->lfs_bfree) * INOPB(fs);
595 sbp->f_ffree = sbp->f_files - fs->lfs_nfiles;
596 if (sbp != &mp->mnt_stat) {
597 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
598 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
599 }
600 strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
601 return (0);
602 }
603
604 /*
605 * Go through the disk queues to initiate sandbagged IO;
606 * go through the inodes to write those that have been modified;
607 * initiate the writing of the super block if it has been modified.
608 *
609 * Note: we are always called with the filesystem marked `MPBUSY'.
610 */
611 int
612 lfs_sync(mp, waitfor, cred, p)
613 struct mount *mp;
614 int waitfor;
615 struct ucred *cred;
616 struct proc *p;
617 {
618 int error;
619 struct lfs *fs;
620
621 fs = ((struct ufsmount *)mp->mnt_data)->ufsmount_u.lfs;
622 if (fs->lfs_ronly)
623 return 0;
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 if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
668 return (0);
669
670 if ((error = getnewvnode(VT_LFS, mp, lfs_vnodeop_p, &vp)) != 0) {
671 *vpp = NULL;
672 return (error);
673 }
674
675 do {
676 if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL) {
677 ungetnewvnode(vp);
678 return (0);
679 }
680 } while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
681
682 /* Translate the inode number to a disk address. */
683 fs = ump->um_lfs;
684 if (ino == LFS_IFILE_INUM)
685 daddr = fs->lfs_idaddr;
686 else {
687 LFS_IENTRY(ifp, fs, ino, bp);
688 daddr = ifp->if_daddr;
689 #ifdef LFS_ATIME_IFILE
690 ts = ifp->if_atime; /* structure copy */
691 #endif
692 brelse(bp);
693 if (daddr == LFS_UNUSED_DADDR) {
694 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
695 return (ENOENT);
696 }
697 }
698
699 /* Allocate/init new vnode/inode. */
700 lfs_vcreate(mp, ino, vp);
701
702 /*
703 * Put it onto its hash chain and lock it so that other requests for
704 * this inode will block if they arrive while we are sleeping waiting
705 * for old data structures to be purged or for the contents of the
706 * disk portion of this inode to be read.
707 */
708 ip = VTOI(vp);
709 ufs_ihashins(ip);
710 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
711
712 /*
713 * XXX
714 * This may not need to be here, logically it should go down with
715 * the i_devvp initialization.
716 * Ask Kirk.
717 */
718 ip->i_lfs = ump->um_lfs;
719
720 /* Read in the disk contents for the inode, copy into the inode. */
721 error = bread(ump->um_devvp, daddr, (int)fs->lfs_bsize, NOCRED, &bp);
722 if (error) {
723 /*
724 * The inode does not contain anything useful, so it would
725 * be misleading to leave it on its hash chain. With mode
726 * still zero, it will be unlinked and returned to the free
727 * list by vput().
728 */
729 vput(vp);
730 brelse(bp);
731 *vpp = NULL;
732 return (error);
733 }
734 ip->i_din.ffs_din = *lfs_ifind(fs, ino, bp);
735 ip->i_ffs_effnlink = ip->i_ffs_nlink;
736 ip->i_lfs_effnblks = ip->i_ffs_blocks;
737 #ifdef LFS_ATIME_IFILE
738 ip->i_ffs_atime = ts.tv_sec;
739 ip->i_ffs_atimensec = ts.tv_nsec;
740 #endif
741 brelse(bp);
742
743 /*
744 * Initialize the vnode from the inode, check for aliases. In all
745 * cases re-init ip, the underlying vnode/inode may have changed.
746 */
747 error = ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
748 if (error) {
749 vput(vp);
750 *vpp = NULL;
751 return (error);
752 }
753 #ifdef DIAGNOSTIC
754 if(vp->v_type == VNON) {
755 panic("lfs_vget: ino %d is type VNON! (ifmt %o)\n",
756 ip->i_number, (ip->i_ffs_mode & IFMT) >> 12);
757 }
758 #endif
759 /*
760 * Finish inode initialization now that aliasing has been resolved.
761 */
762 ip->i_devvp = ump->um_devvp;
763 VREF(ip->i_devvp);
764 *vpp = vp;
765
766 return (0);
767 }
768
769 /*
770 * File handle to vnode
771 *
772 * Have to be really careful about stale file handles:
773 * - check that the inode number is valid
774 * - call lfs_vget() to get the locked inode
775 * - check for an unallocated inode (i_mode == 0)
776 *
777 * XXX
778 * use ifile to see if inode is allocated instead of reading off disk
779 * what is the relationship between my generational number and the NFS
780 * generational number.
781 */
782 int
783 lfs_fhtovp(mp, fhp, vpp)
784 struct mount *mp;
785 struct fid *fhp;
786 struct vnode **vpp;
787 {
788 struct ufid *ufhp;
789
790 ufhp = (struct ufid *)fhp;
791 if (ufhp->ufid_ino < ROOTINO)
792 return (ESTALE);
793 return (ufs_fhtovp(mp, ufhp, vpp));
794 }
795
796 /*
797 * Vnode pointer to File handle
798 */
799 /* ARGSUSED */
800 int
801 lfs_vptofh(vp, fhp)
802 struct vnode *vp;
803 struct fid *fhp;
804 {
805 struct inode *ip;
806 struct ufid *ufhp;
807
808 ip = VTOI(vp);
809 ufhp = (struct ufid *)fhp;
810 ufhp->ufid_len = sizeof(struct ufid);
811 ufhp->ufid_ino = ip->i_number;
812 ufhp->ufid_gen = ip->i_ffs_gen;
813 return (0);
814 }
815
816 int
817 lfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
818 int *name;
819 u_int namelen;
820 void *oldp;
821 size_t *oldlenp;
822 void *newp;
823 size_t newlen;
824 struct proc *p;
825 {
826 extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead;
827 extern struct lfs_stats lfs_stats;
828 int error;
829
830 /* all sysctl names at this level are terminal */
831 if (namelen != 1)
832 return (ENOTDIR);
833
834 switch (name[0]) {
835 case LFS_WRITEINDIR:
836 return (sysctl_int(oldp, oldlenp, newp, newlen,
837 &lfs_writeindir));
838 case LFS_CLEAN_VNHEAD:
839 return (sysctl_int(oldp, oldlenp, newp, newlen,
840 &lfs_clean_vnhead));
841 case LFS_DOSTATS:
842 if((error = sysctl_int(oldp, oldlenp, newp, newlen,
843 &lfs_dostats)))
844 return error;
845 if(lfs_dostats == 0)
846 memset(&lfs_stats,0,sizeof(lfs_stats));
847 return 0;
848 case LFS_STATS:
849 return (sysctl_rdstruct(oldp, oldlenp, newp,
850 &lfs_stats, sizeof(lfs_stats)));
851 default:
852 return (EOPNOTSUPP);
853 }
854 /* NOTREACHED */
855 }
856