Annotation of sys/ufs/ffs/ffs_vfsops.c, Revision 1.1.1.1
1.1 nbrk 1: /* $OpenBSD: ffs_vfsops.c,v 1.109 2007/08/04 03:33:31 art Exp $ */
2: /* $NetBSD: ffs_vfsops.c,v 1.19 1996/02/09 22:22:26 christos Exp $ */
3:
4: /*
5: * Copyright (c) 1989, 1991, 1993, 1994
6: * The Regents of the University of California. All rights reserved.
7: *
8: * Redistribution and use in source and binary forms, with or without
9: * modification, are permitted provided that the following conditions
10: * are met:
11: * 1. Redistributions of source code must retain the above copyright
12: * notice, this list of conditions and the following disclaimer.
13: * 2. Redistributions in binary form must reproduce the above copyright
14: * notice, this list of conditions and the following disclaimer in the
15: * documentation and/or other materials provided with the distribution.
16: * 3. Neither the name of the University nor the names of its contributors
17: * may be used to endorse or promote products derived from this software
18: * without specific prior written permission.
19: *
20: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30: * SUCH DAMAGE.
31: *
32: * @(#)ffs_vfsops.c 8.14 (Berkeley) 11/28/94
33: */
34:
35: #include <sys/param.h>
36: #include <sys/systm.h>
37: #include <sys/namei.h>
38: #include <sys/proc.h>
39: #include <sys/kernel.h>
40: #include <sys/vnode.h>
41: #include <sys/socket.h>
42: #include <sys/mount.h>
43: #include <sys/buf.h>
44: #include <sys/mbuf.h>
45: #include <sys/file.h>
46: #include <sys/disklabel.h>
47: #include <sys/ioctl.h>
48: #include <sys/errno.h>
49: #include <sys/malloc.h>
50: #include <sys/sysctl.h>
51: #include <sys/pool.h>
52:
53: #include <dev/rndvar.h>
54:
55: #include <miscfs/specfs/specdev.h>
56:
57: #include <ufs/ufs/quota.h>
58: #include <ufs/ufs/ufsmount.h>
59: #include <ufs/ufs/inode.h>
60: #include <ufs/ufs/dir.h>
61: #include <ufs/ufs/ufs_extern.h>
62: #include <ufs/ufs/dirhash.h>
63:
64: #include <ufs/ffs/fs.h>
65: #include <ufs/ffs/ffs_extern.h>
66:
67: int ffs_sbupdate(struct ufsmount *, int);
68: int ffs_reload_vnode(struct vnode *, void *);
69: int ffs_sync_vnode(struct vnode *, void *);
70: int ffs_validate(struct fs *);
71:
72: void ffs1_compat_read(struct fs *, struct ufsmount *, daddr64_t);
73: void ffs1_compat_write(struct fs *, struct ufsmount *);
74:
75: const struct vfsops ffs_vfsops = {
76: ffs_mount,
77: ufs_start,
78: ffs_unmount,
79: ufs_root,
80: ufs_quotactl,
81: ffs_statfs,
82: ffs_sync,
83: ffs_vget,
84: ffs_fhtovp,
85: ffs_vptofh,
86: ffs_init,
87: ffs_sysctl,
88: ufs_check_export
89: };
90:
91: struct inode_vtbl ffs_vtbl = {
92: ffs_truncate,
93: ffs_update,
94: ffs_inode_alloc,
95: ffs_inode_free,
96: ffs_balloc,
97: ffs_bufatoff
98: };
99:
100:
101: /*
102: * Called by main() when ufs is going to be mounted as root.
103: */
104:
105: struct pool ffs_ino_pool;
106: struct pool ffs_dinode1_pool;
107: #ifdef FFS2
108: struct pool ffs_dinode2_pool;
109: #endif
110:
111: int
112: ffs_mountroot(void)
113: {
114: struct fs *fs;
115: struct mount *mp;
116: struct proc *p = curproc; /* XXX */
117: struct ufsmount *ump;
118: int error;
119:
120: /*
121: * Get vnodes for swapdev and rootdev.
122: */
123: swapdev_vp = NULL;
124: if ((error = bdevvp(swapdev, &swapdev_vp)) ||
125: (error = bdevvp(rootdev, &rootvp))) {
126: printf("ffs_mountroot: can't setup bdevvp's\n");
127: if (swapdev_vp)
128: vrele(swapdev_vp);
129: return (error);
130: }
131:
132: if ((error = vfs_rootmountalloc("ffs", "root_device", &mp)) != 0) {
133: vrele(swapdev_vp);
134: vrele(rootvp);
135: return (error);
136: }
137:
138: if ((error = ffs_mountfs(rootvp, mp, p)) != 0) {
139: mp->mnt_vfc->vfc_refcount--;
140: vfs_unbusy(mp);
141: free(mp, M_MOUNT);
142: vrele(swapdev_vp);
143: vrele(rootvp);
144: return (error);
145: }
146:
147: CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
148: ump = VFSTOUFS(mp);
149: fs = ump->um_fs;
150: (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
151: (void)ffs_statfs(mp, &mp->mnt_stat, p);
152: vfs_unbusy(mp);
153: inittodr(fs->fs_time);
154:
155: return (0);
156: }
157:
158: /*
159: * VFS Operations.
160: *
161: * mount system call
162: */
163: int
164: ffs_mount(struct mount *mp, const char *path, void *data,
165: struct nameidata *ndp, struct proc *p)
166: {
167: struct vnode *devvp;
168: struct ufs_args args;
169: struct ufsmount *ump = NULL;
170: struct fs *fs;
171: int error = 0, flags;
172: int ronly;
173: mode_t accessmode;
174: size_t size;
175:
176: error = copyin(data, &args, sizeof (struct ufs_args));
177: if (error)
178: return (error);
179:
180: #ifndef FFS_SOFTUPDATES
181: if (mp->mnt_flag & MNT_SOFTDEP) {
182: printf("WARNING: soft updates isn't compiled in\n");
183: mp->mnt_flag &= ~MNT_SOFTDEP;
184: }
185: #endif
186:
187: /*
188: * Soft updates is incompatible with "async",
189: * so if we are doing softupdates stop the user
190: * from setting the async flag.
191: */
192: if ((mp->mnt_flag & (MNT_SOFTDEP | MNT_ASYNC)) ==
193: (MNT_SOFTDEP | MNT_ASYNC)) {
194: return (EINVAL);
195: }
196: /*
197: * If updating, check whether changing from read-only to
198: * read/write; if there is no device name, that's all we do.
199: */
200: if (mp->mnt_flag & MNT_UPDATE) {
201: ump = VFSTOUFS(mp);
202: fs = ump->um_fs;
203: devvp = ump->um_devvp;
204: error = 0;
205: ronly = fs->fs_ronly;
206:
207: if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
208: /* Flush any dirty data */
209: mp->mnt_flag &= ~MNT_RDONLY;
210: VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
211: mp->mnt_flag |= MNT_RDONLY;
212:
213: /*
214: * Get rid of files open for writing.
215: */
216: flags = WRITECLOSE;
217: if (mp->mnt_flag & MNT_FORCE)
218: flags |= FORCECLOSE;
219: if (fs->fs_flags & FS_DOSOFTDEP) {
220: error = softdep_flushfiles(mp, flags, p);
221: mp->mnt_flag &= ~MNT_SOFTDEP;
222: } else
223: error = ffs_flushfiles(mp, flags, p);
224: ronly = 1;
225: }
226:
227: /*
228: * Flush soft dependencies if disabling it via an update
229: * mount. This may leave some items to be processed,
230: * so don't do this yet XXX.
231: */
232: if ((fs->fs_flags & FS_DOSOFTDEP) &&
233: !(mp->mnt_flag & MNT_SOFTDEP) &&
234: !(mp->mnt_flag & MNT_RDONLY) && fs->fs_ronly == 0) {
235: #if 0
236: flags = WRITECLOSE;
237: if (mp->mnt_flag & MNT_FORCE)
238: flags |= FORCECLOSE;
239: error = softdep_flushfiles(mp, flags, p);
240: #elif FFS_SOFTUPDATES
241: mp->mnt_flag |= MNT_SOFTDEP;
242: #endif
243: }
244: /*
245: * When upgrading to a softdep mount, we must first flush
246: * all vnodes. (not done yet -- see above)
247: */
248: if (!(fs->fs_flags & FS_DOSOFTDEP) &&
249: (mp->mnt_flag & MNT_SOFTDEP) && fs->fs_ronly == 0) {
250: #if 0
251: flags = WRITECLOSE;
252: if (mp->mnt_flag & MNT_FORCE)
253: flags |= FORCECLOSE;
254: error = ffs_flushfiles(mp, flags, p);
255: #else
256: mp->mnt_flag &= ~MNT_SOFTDEP;
257: #endif
258: }
259:
260: if (!error && (mp->mnt_flag & MNT_RELOAD))
261: error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
262: if (error)
263: goto error_1;
264:
265: if (ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
266: /*
267: * If upgrade to read-write by non-root, then verify
268: * that user has necessary permissions on the device.
269: */
270: if (suser(p, 0)) {
271: vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
272: error = VOP_ACCESS(devvp, VREAD | VWRITE,
273: p->p_ucred, p);
274: VOP_UNLOCK(devvp, 0, p);
275: if (error)
276: goto error_1;
277: }
278:
279: if (fs->fs_clean == 0) {
280: #if 0
281: /*
282: * It is safe mount unclean file system
283: * if it was previously mounted with softdep
284: * but we may loss space and must
285: * sometimes run fsck manually.
286: */
287: if (fs->fs_flags & FS_DOSOFTDEP)
288: printf(
289: "WARNING: %s was not properly unmounted\n",
290: fs->fs_fsmnt);
291: else
292: #endif
293: if (mp->mnt_flag & MNT_FORCE) {
294: printf(
295: "WARNING: %s was not properly unmounted\n",
296: fs->fs_fsmnt);
297: } else {
298: printf(
299: "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
300: fs->fs_fsmnt);
301: error = EROFS;
302: goto error_1;
303: }
304: }
305:
306: if ((fs->fs_flags & FS_DOSOFTDEP)) {
307: error = softdep_mount(devvp, mp, fs,
308: p->p_ucred);
309: if (error)
310: goto error_1;
311: }
312: fs->fs_contigdirs=(u_int8_t*)malloc((u_long)fs->fs_ncg,
313: M_UFSMNT, M_WAITOK);
314: bzero(fs->fs_contigdirs, fs->fs_ncg);
315:
316: ronly = 0;
317: }
318: if (args.fspec == 0) {
319: /*
320: * Process export requests.
321: */
322: error = vfs_export(mp, &ump->um_export,
323: &args.export_info);
324: if (error)
325: goto error_1;
326: else
327: goto success;
328: }
329: }
330: /*
331: * Not an update, or updating the name: look up the name
332: * and verify that it refers to a sensible block device.
333: */
334: NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
335: if ((error = namei(ndp)) != 0)
336: goto error_1;
337:
338: devvp = ndp->ni_vp;
339:
340: if (devvp->v_type != VBLK) {
341: error = ENOTBLK;
342: goto error_2;
343: }
344:
345: if (major(devvp->v_rdev) >= nblkdev) {
346: error = ENXIO;
347: goto error_2;
348: }
349:
350: /*
351: * If mount by non-root, then verify that user has necessary
352: * permissions on the device.
353: */
354: if (suser(p, 0)) {
355: accessmode = VREAD;
356: if ((mp->mnt_flag & MNT_RDONLY) == 0)
357: accessmode |= VWRITE;
358: vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
359: error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
360: VOP_UNLOCK(devvp, 0, p);
361: if (error)
362: goto error_2;
363: }
364:
365: if (mp->mnt_flag & MNT_UPDATE) {
366: /*
367: * UPDATE
368: * If it's not the same vnode, or at least the same device
369: * then it's not correct.
370: */
371:
372: if (devvp != ump->um_devvp) {
373: if (devvp->v_rdev == ump->um_devvp->v_rdev) {
374: vrele(devvp);
375: } else {
376: error = EINVAL; /* needs translation */
377: }
378: } else
379: vrele(devvp);
380: /*
381: * Update device name only on success
382: */
383: if (!error) {
384: /*
385: * Save "mounted from" info for mount point (NULL pad)
386: */
387: copyinstr(args.fspec,
388: mp->mnt_stat.f_mntfromname,
389: MNAMELEN - 1,
390: &size);
391: bzero(mp->mnt_stat.f_mntfromname + size,
392: MNAMELEN - size);
393: }
394: } else {
395: /*
396: * Since this is a new mount, we want the names for
397: * the device and the mount point copied in. If an
398: * error occurs, the mountpoint is discarded by the
399: * upper level code.
400: */
401: /* Save "last mounted on" info for mount point (NULL pad)*/
402: copyinstr(path, /* mount point*/
403: mp->mnt_stat.f_mntonname, /* save area*/
404: MNAMELEN - 1, /* max size*/
405: &size); /* real size*/
406: bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
407:
408: /* Save "mounted from" info for mount point (NULL pad)*/
409: copyinstr(args.fspec, /* device name*/
410: mp->mnt_stat.f_mntfromname, /* save area*/
411: MNAMELEN - 1, /* max size*/
412: &size); /* real size*/
413: bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
414:
415: error = ffs_mountfs(devvp, mp, p);
416: }
417:
418: if (error)
419: goto error_2;
420:
421: /*
422: * Initialize FS stat information in mount struct; uses both
423: * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
424: *
425: * This code is common to root and non-root mounts
426: */
427: bcopy(&args, &mp->mnt_stat.mount_info.ufs_args, sizeof(args));
428: (void)VFS_STATFS(mp, &mp->mnt_stat, p);
429:
430: success:
431: if (path && (mp->mnt_flag & MNT_UPDATE)) {
432: /* Update clean flag after changing read-onlyness. */
433: fs = ump->um_fs;
434: if (ronly != fs->fs_ronly) {
435: fs->fs_ronly = ronly;
436: fs->fs_clean = ronly &&
437: (fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
438: if (ronly)
439: free(fs->fs_contigdirs, M_UFSMNT);
440: }
441: if (!ronly) {
442: if (mp->mnt_flag & MNT_SOFTDEP)
443: fs->fs_flags |= FS_DOSOFTDEP;
444: else
445: fs->fs_flags &= ~FS_DOSOFTDEP;
446: }
447: ffs_sbupdate(ump, MNT_WAIT);
448: }
449: return (0);
450:
451: error_2: /* error with devvp held */
452: vrele (devvp);
453: error_1: /* no state to back out */
454: return (error);
455: }
456:
457: struct ffs_reload_args {
458: struct fs *fs;
459: struct proc *p;
460: struct ucred *cred;
461: struct vnode *devvp;
462: };
463:
464: int
465: ffs_reload_vnode(struct vnode *vp, void *args)
466: {
467: struct ffs_reload_args *fra = args;
468: struct inode *ip;
469: struct buf *bp;
470: int error;
471:
472: /*
473: * Step 4: invalidate all inactive vnodes.
474: */
475: if (vp->v_usecount == 0) {
476: vgonel(vp, fra->p);
477: return (0);
478: }
479:
480: /*
481: * Step 5: invalidate all cached file data.
482: */
483: if (vget(vp, LK_EXCLUSIVE, fra->p))
484: return (0);
485:
486: if (vinvalbuf(vp, 0, fra->cred, fra->p, 0, 0))
487: panic("ffs_reload: dirty2");
488:
489: /*
490: * Step 6: re-read inode data for all active vnodes.
491: */
492: ip = VTOI(vp);
493:
494: error = bread(fra->devvp,
495: fsbtodb(fra->fs, ino_to_fsba(fra->fs, ip->i_number)),
496: (int)fra->fs->fs_bsize, NOCRED, &bp);
497: if (error) {
498: brelse(bp);
499: vput(vp);
500: return (error);
501: }
502:
503: *ip->i_din1 = *((struct ufs1_dinode *)bp->b_data +
504: ino_to_fsbo(fra->fs, ip->i_number));
505: ip->i_effnlink = DIP(ip, nlink);
506: brelse(bp);
507: vput(vp);
508: return (0);
509: }
510:
511: /*
512: * Reload all incore data for a filesystem (used after running fsck on
513: * the root filesystem and finding things to fix). The filesystem must
514: * be mounted read-only.
515: *
516: * Things to do to update the mount:
517: * 1) invalidate all cached meta-data.
518: * 2) re-read superblock from disk.
519: * 3) re-read summary information from disk.
520: * 4) invalidate all inactive vnodes.
521: * 5) invalidate all cached file data.
522: * 6) re-read inode data for all active vnodes.
523: */
524: int
525: ffs_reload(struct mount *mountp, struct ucred *cred, struct proc *p)
526: {
527: struct vnode *devvp;
528: caddr_t space;
529: struct fs *fs, *newfs;
530: struct partinfo dpart;
531: int i, blks, size, error;
532: int32_t *lp;
533: struct buf *bp = NULL;
534: struct ffs_reload_args fra;
535:
536: if ((mountp->mnt_flag & MNT_RDONLY) == 0)
537: return (EINVAL);
538: /*
539: * Step 1: invalidate all cached meta-data.
540: */
541: devvp = VFSTOUFS(mountp)->um_devvp;
542: vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
543: error = vinvalbuf(devvp, 0, cred, p, 0, 0);
544: VOP_UNLOCK(devvp, 0, p);
545: if (error)
546: panic("ffs_reload: dirty1");
547:
548: /*
549: * Step 2: re-read superblock from disk.
550: */
551: if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
552: size = DEV_BSIZE;
553: else
554: size = dpart.disklab->d_secsize;
555:
556: fs = VFSTOUFS(mountp)->um_fs;
557:
558: error = bread(devvp, (daddr_t)(fs->fs_sblockloc / size), SBSIZE,
559: NOCRED, &bp);
560: if (error) {
561: brelse(bp);
562: return (error);
563: }
564:
565: newfs = (struct fs *)bp->b_data;
566: if (ffs_validate(newfs) == 0) {
567: brelse(bp);
568: return (EINVAL);
569: }
570:
571: /*
572: * Copy pointer fields back into superblock before copying in XXX
573: * new superblock. These should really be in the ufsmount. XXX
574: * Note that important parameters (eg fs_ncg) are unchanged.
575: */
576: newfs->fs_csp = fs->fs_csp;
577: newfs->fs_maxcluster = fs->fs_maxcluster;
578: newfs->fs_ronly = fs->fs_ronly;
579: bcopy(newfs, fs, (u_int)fs->fs_sbsize);
580: if (fs->fs_sbsize < SBSIZE)
581: bp->b_flags |= B_INVAL;
582: brelse(bp);
583: mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
584: ffs1_compat_read(fs, VFSTOUFS(mountp), fs->fs_sblockloc);
585: ffs_oldfscompat(fs);
586: (void)ffs_statfs(mountp, &mountp->mnt_stat, p);
587: /*
588: * Step 3: re-read summary information from disk.
589: */
590: blks = howmany(fs->fs_cssize, fs->fs_fsize);
591: space = (caddr_t)fs->fs_csp;
592: for (i = 0; i < blks; i += fs->fs_frag) {
593: size = fs->fs_bsize;
594: if (i + fs->fs_frag > blks)
595: size = (blks - i) * fs->fs_fsize;
596: error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
597: NOCRED, &bp);
598: if (error) {
599: brelse(bp);
600: return (error);
601: }
602: bcopy(bp->b_data, space, (u_int)size);
603: space += size;
604: brelse(bp);
605: }
606: if ((fs->fs_flags & FS_DOSOFTDEP))
607: (void) softdep_mount(devvp, mountp, fs, cred);
608: /*
609: * We no longer know anything about clusters per cylinder group.
610: */
611: if (fs->fs_contigsumsize > 0) {
612: lp = fs->fs_maxcluster;
613: for (i = 0; i < fs->fs_ncg; i++)
614: *lp++ = fs->fs_contigsumsize;
615: }
616:
617: fra.p = p;
618: fra.cred = cred;
619: fra.fs = fs;
620: fra.devvp = devvp;
621:
622: error = vfs_mount_foreach_vnode(mountp, ffs_reload_vnode, &fra);
623:
624: return (error);
625: }
626:
627: /*
628: * Checks if a super block is sane enough to be mounted.
629: */
630: int
631: ffs_validate(struct fs *fsp)
632: {
633: #ifdef FFS2
634: if (fsp->fs_magic != FS_UFS2_MAGIC && fsp->fs_magic != FS_UFS1_MAGIC)
635: return (0); /* Invalid magic */
636: #else
637: if (fsp->fs_magic != FS_UFS1_MAGIC)
638: return (0); /* Invalid magic */
639: #endif /* FFS2 */
640:
641: if ((u_int)fsp->fs_bsize > MAXBSIZE)
642: return (0); /* Invalid block size */
643:
644: if ((u_int)fsp->fs_bsize < sizeof(struct fs))
645: return (0); /* Invalid block size */
646:
647: if ((u_int)fsp->fs_sbsize > SBSIZE)
648: return (0); /* Invalid super block size */
649:
650: if ((u_int)fsp->fs_frag > MAXFRAG || fragtbl[fsp->fs_frag] == NULL)
651: return (0); /* Invalid number of fragments */
652:
653: return (1); /* Super block is okay */
654: }
655:
656: /*
657: * Possible locations for the super-block.
658: */
659: const int sbtry[] = SBLOCKSEARCH;
660:
661: /*
662: * Common code for mount and mountroot
663: */
664: int
665: ffs_mountfs(struct vnode *devvp, struct mount *mp, struct proc *p)
666: {
667: struct ufsmount *ump;
668: struct buf *bp;
669: struct fs *fs;
670: dev_t dev;
671: struct partinfo dpart;
672: caddr_t space;
673: daddr64_t sbloc;
674: int error, i, blks, size, ronly;
675: int32_t *lp;
676: size_t strsize;
677: struct ucred *cred;
678: u_int64_t maxfilesize; /* XXX */
679:
680: dev = devvp->v_rdev;
681: cred = p ? p->p_ucred : NOCRED;
682: /*
683: * Disallow multiple mounts of the same device.
684: * Disallow mounting of a device that is currently in use
685: * (except for root, which might share swap device for miniroot).
686: * Flush out any old buffers remaining from a previous use.
687: */
688: if ((error = vfs_mountedon(devvp)) != 0)
689: return (error);
690: if (vcount(devvp) > 1 && devvp != rootvp)
691: return (EBUSY);
692: vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
693: error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0);
694: VOP_UNLOCK(devvp, 0, p);
695: if (error)
696: return (error);
697:
698: ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
699: error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
700: if (error)
701: return (error);
702: if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
703: size = DEV_BSIZE;
704: else
705: size = dpart.disklab->d_secsize;
706:
707: bp = NULL;
708: ump = NULL;
709:
710: /*
711: * Try reading the super-block in each of its possible locations.
712: */
713: for (i = 0; sbtry[i] != -1; i++) {
714: if (bp != NULL) {
715: bp->b_flags |= B_NOCACHE;
716: brelse(bp);
717: bp = NULL;
718: }
719:
720: error = bread(devvp, sbtry[i] / size, SBSIZE, cred, &bp);
721: if (error)
722: goto out;
723:
724: fs = (struct fs *) bp->b_data;
725: sbloc = sbtry[i];
726:
727: #if 0
728: if (fs->fs_magic == FS_UFS2_MAGIC) {
729: printf("ffs_mountfs(): Sorry, no UFS2 support (yet)\n");
730: error = EFTYPE;
731: goto out;
732: }
733: #endif
734:
735: /*
736: * Do not look for an FFS1 file system at SBLOCK_UFS2. Doing so
737: * will find the wrong super-block for file systems with 64k
738: * block size.
739: */
740: if (fs->fs_magic == FS_UFS1_MAGIC && sbloc == SBLOCK_UFS2)
741: continue;
742:
743: if (ffs_validate(fs))
744: break; /* Super block validated */
745: }
746:
747: if (sbtry[i] == -1) {
748: error = EINVAL;
749: goto out;
750: }
751:
752: fs->fs_fmod = 0;
753: fs->fs_flags &= ~FS_UNCLEAN;
754: if (fs->fs_clean == 0) {
755: #if 0
756: /*
757: * It is safe mount unclean file system
758: * if it was previously mounted with softdep
759: * but we may loss space and must
760: * sometimes run fsck manually.
761: */
762: if (fs->fs_flags & FS_DOSOFTDEP)
763: printf(
764: "WARNING: %s was not properly unmounted\n",
765: fs->fs_fsmnt);
766: else
767: #endif
768: if (ronly || (mp->mnt_flag & MNT_FORCE)) {
769: printf(
770: "WARNING: %s was not properly unmounted\n",
771: fs->fs_fsmnt);
772: } else {
773: printf(
774: "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
775: fs->fs_fsmnt);
776: error = EROFS;
777: goto out;
778: }
779: }
780:
781: if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
782: #ifndef SMALL_KERNEL
783: printf("ffs_mountfs(): obsolete rotational table format, "
784: "please use fsck_ffs(8) -c 1\n");
785: #endif
786: error = EFTYPE;
787: goto out;
788: }
789:
790: ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
791: bzero(ump, sizeof *ump);
792: ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
793: M_WAITOK);
794:
795: if (fs->fs_magic == FS_UFS1_MAGIC)
796: ump->um_fstype = UM_UFS1;
797: #ifdef FFS2
798: else
799: ump->um_fstype = UM_UFS2;
800: #endif
801:
802: bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
803: if (fs->fs_sbsize < SBSIZE)
804: bp->b_flags |= B_INVAL;
805: brelse(bp);
806: bp = NULL;
807: fs = ump->um_fs;
808:
809: ffs1_compat_read(fs, ump, sbloc);
810:
811: if (fs->fs_clean == 0)
812: fs->fs_flags |= FS_UNCLEAN;
813: fs->fs_ronly = ronly;
814: size = fs->fs_cssize;
815: blks = howmany(size, fs->fs_fsize);
816: if (fs->fs_contigsumsize > 0)
817: size += fs->fs_ncg * sizeof(int32_t);
818: space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
819: fs->fs_csp = (struct csum *)space;
820: for (i = 0; i < blks; i += fs->fs_frag) {
821: size = fs->fs_bsize;
822: if (i + fs->fs_frag > blks)
823: size = (blks - i) * fs->fs_fsize;
824: error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
825: cred, &bp);
826: if (error) {
827: free(fs->fs_csp, M_UFSMNT);
828: goto out;
829: }
830: bcopy(bp->b_data, space, (u_int)size);
831: space += size;
832: brelse(bp);
833: bp = NULL;
834: }
835: if (fs->fs_contigsumsize > 0) {
836: fs->fs_maxcluster = lp = (int32_t *)space;
837: for (i = 0; i < fs->fs_ncg; i++)
838: *lp++ = fs->fs_contigsumsize;
839: }
840: mp->mnt_data = (qaddr_t)ump;
841: mp->mnt_stat.f_fsid.val[0] = (long)dev;
842: /* Use on-disk fsid if it exists, else fake it */
843: if (fs->fs_id[0] != 0 && fs->fs_id[1] != 0)
844: mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
845: else
846: mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
847: mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
848: mp->mnt_flag |= MNT_LOCAL;
849: ump->um_mountp = mp;
850: ump->um_dev = dev;
851: ump->um_devvp = devvp;
852: ump->um_nindir = fs->fs_nindir;
853: ump->um_bptrtodb = fs->fs_fsbtodb;
854: ump->um_seqinc = fs->fs_frag;
855: for (i = 0; i < MAXQUOTAS; i++)
856: ump->um_quotas[i] = NULLVP;
857:
858: devvp->v_specmountpoint = mp;
859: ffs_oldfscompat(fs);
860:
861: if (ronly)
862: fs->fs_contigdirs = NULL;
863: else {
864: fs->fs_contigdirs = (u_int8_t*)malloc((u_long)fs->fs_ncg,
865: M_UFSMNT, M_WAITOK);
866: bzero(fs->fs_contigdirs, fs->fs_ncg);
867: }
868:
869: /*
870: * Set FS local "last mounted on" information (NULL pad)
871: */
872: copystr(mp->mnt_stat.f_mntonname, /* mount point*/
873: fs->fs_fsmnt, /* copy area*/
874: sizeof(fs->fs_fsmnt) - 1, /* max size*/
875: &strsize); /* real size*/
876: bzero(fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
877:
878: #if 0
879: if( mp->mnt_flag & MNT_ROOTFS) {
880: /*
881: * Root mount; update timestamp in mount structure.
882: * this will be used by the common root mount code
883: * to update the system clock.
884: */
885: mp->mnt_time = fs->fs_time;
886: }
887: #endif
888:
889: /*
890: * XXX
891: * Limit max file size. Even though ffs can handle files up to 16TB,
892: * we do limit the max file to 2^31 pages to prevent overflow of
893: * a 32-bit unsigned int. The buffer cache has its own checks but
894: * a little added paranoia never hurts.
895: */
896: ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
897: maxfilesize = (u_int64_t)0x80000000 * MIN(PAGE_SIZE, fs->fs_bsize) - 1;
898: if (fs->fs_maxfilesize > maxfilesize) /* XXX */
899: fs->fs_maxfilesize = maxfilesize; /* XXX */
900: if (ronly == 0) {
901: if ((fs->fs_flags & FS_DOSOFTDEP) &&
902: (error = softdep_mount(devvp, mp, fs, cred)) != 0) {
903: free(fs->fs_csp, M_UFSMNT);
904: free(fs->fs_contigdirs, M_UFSMNT);
905: goto out;
906: }
907: fs->fs_fmod = 1;
908: fs->fs_clean = 0;
909: if (mp->mnt_flag & MNT_SOFTDEP)
910: fs->fs_flags |= FS_DOSOFTDEP;
911: else
912: fs->fs_flags &= ~FS_DOSOFTDEP;
913: (void) ffs_sbupdate(ump, MNT_WAIT);
914: }
915: return (0);
916: out:
917: devvp->v_specmountpoint = NULL;
918: if (bp)
919: brelse(bp);
920: (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
921: if (ump) {
922: free(ump->um_fs, M_UFSMNT);
923: free(ump, M_UFSMNT);
924: mp->mnt_data = (qaddr_t)0;
925: }
926: return (error);
927: }
928:
929: /*
930: * Sanity checks for old file systems.
931: */
932: int
933: ffs_oldfscompat(struct fs *fs)
934: {
935: int i;
936:
937: fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
938: fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
939: if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
940: fs->fs_nrpos = 8; /* XXX */
941: if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
942: u_int64_t sizepb = fs->fs_bsize; /* XXX */
943: /* XXX */
944: fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
945: for (i = 0; i < NIADDR; i++) { /* XXX */
946: sizepb *= NINDIR(fs); /* XXX */
947: fs->fs_maxfilesize += sizepb; /* XXX */
948: } /* XXX */
949: fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
950: fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
951: } /* XXX */
952: if (fs->fs_avgfilesize <= 0) /* XXX */
953: fs->fs_avgfilesize = AVFILESIZ; /* XXX */
954: if (fs->fs_avgfpdir <= 0) /* XXX */
955: fs->fs_avgfpdir = AFPDIR; /* XXX */
956: return (0);
957: }
958:
959: /*
960: * Auxiliary function for reading FFS1 super blocks.
961: */
962: void
963: ffs1_compat_read(struct fs *fs, struct ufsmount *ump, daddr64_t sbloc)
964: {
965: if (fs->fs_magic == FS_UFS2_MAGIC)
966: return; /* UFS2 */
967: #if 0
968: if (fs->fs_ffs1_flags & FS_FLAGS_UPDATED)
969: return; /* Already updated */
970: #endif
971: fs->fs_flags = fs->fs_ffs1_flags;
972: fs->fs_sblockloc = sbloc;
973: fs->fs_maxbsize = fs->fs_bsize;
974: fs->fs_time = fs->fs_ffs1_time;
975: fs->fs_size = fs->fs_ffs1_size;
976: fs->fs_dsize = fs->fs_ffs1_dsize;
977: fs->fs_csaddr = fs->fs_ffs1_csaddr;
978: fs->fs_cstotal.cs_ndir = fs->fs_ffs1_cstotal.cs_ndir;
979: fs->fs_cstotal.cs_nbfree = fs->fs_ffs1_cstotal.cs_nbfree;
980: fs->fs_cstotal.cs_nifree = fs->fs_ffs1_cstotal.cs_nifree;
981: fs->fs_cstotal.cs_nffree = fs->fs_ffs1_cstotal.cs_nffree;
982: fs->fs_ffs1_flags |= FS_FLAGS_UPDATED;
983: }
984:
985: /*
986: * Auxiliary function for writing FFS1 super blocks.
987: */
988: void
989: ffs1_compat_write(struct fs *fs, struct ufsmount *ump)
990: {
991: if (fs->fs_magic != FS_UFS1_MAGIC)
992: return; /* UFS2 */
993:
994: fs->fs_ffs1_time = fs->fs_time;
995: fs->fs_ffs1_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
996: fs->fs_ffs1_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
997: fs->fs_ffs1_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
998: fs->fs_ffs1_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
999: }
1000:
1001: /*
1002: * unmount system call
1003: */
1004: int
1005: ffs_unmount(struct mount *mp, int mntflags, struct proc *p)
1006: {
1007: struct ufsmount *ump;
1008: struct fs *fs;
1009: int error, flags;
1010:
1011: flags = 0;
1012: if (mntflags & MNT_FORCE)
1013: flags |= FORCECLOSE;
1014:
1015: ump = VFSTOUFS(mp);
1016: fs = ump->um_fs;
1017: if (mp->mnt_flag & MNT_SOFTDEP)
1018: error = softdep_flushfiles(mp, flags, p);
1019: else
1020: error = ffs_flushfiles(mp, flags, p);
1021: if (error != 0)
1022: return (error);
1023:
1024: if (fs->fs_ronly == 0) {
1025: fs->fs_clean = (fs->fs_flags & FS_UNCLEAN) ? 0 : 1;
1026: error = ffs_sbupdate(ump, MNT_WAIT);
1027: /* ignore write errors if mounted RW on read-only device */
1028: if (error && error != EROFS) {
1029: fs->fs_clean = 0;
1030: return (error);
1031: }
1032: free(fs->fs_contigdirs, M_UFSMNT);
1033: }
1034: ump->um_devvp->v_specmountpoint = NULL;
1035:
1036: vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, p, 0, 0);
1037: error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
1038: NOCRED, p);
1039: vrele(ump->um_devvp);
1040: free(fs->fs_csp, M_UFSMNT);
1041: free(fs, M_UFSMNT);
1042: free(ump, M_UFSMNT);
1043: mp->mnt_data = (qaddr_t)0;
1044: mp->mnt_flag &= ~MNT_LOCAL;
1045: return (error);
1046: }
1047:
1048: /*
1049: * Flush out all the files in a filesystem.
1050: */
1051: int
1052: ffs_flushfiles(struct mount *mp, int flags, struct proc *p)
1053: {
1054: struct ufsmount *ump;
1055: int error;
1056:
1057: ump = VFSTOUFS(mp);
1058: if (mp->mnt_flag & MNT_QUOTA) {
1059: int i;
1060: if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0)
1061: return (error);
1062: for (i = 0; i < MAXQUOTAS; i++) {
1063: if (ump->um_quotas[i] == NULLVP)
1064: continue;
1065: quotaoff(p, mp, i);
1066: }
1067: /*
1068: * Here we fall through to vflush again to ensure
1069: * that we have gotten rid of all the system vnodes.
1070: */
1071: }
1072:
1073: /*
1074: * Flush all the files.
1075: */
1076: if ((error = vflush(mp, NULL, flags)) != 0)
1077: return (error);
1078: /*
1079: * Flush filesystem metadata.
1080: */
1081: vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p);
1082: error = VOP_FSYNC(ump->um_devvp, p->p_ucred, MNT_WAIT, p);
1083: VOP_UNLOCK(ump->um_devvp, 0, p);
1084: return (error);
1085: }
1086:
1087: /*
1088: * Get file system statistics.
1089: */
1090: int
1091: ffs_statfs(struct mount *mp, struct statfs *sbp, struct proc *p)
1092: {
1093: struct ufsmount *ump;
1094: struct fs *fs;
1095:
1096: ump = VFSTOUFS(mp);
1097: fs = ump->um_fs;
1098:
1099: #ifdef FFS2
1100: if (fs->fs_magic != FS_MAGIC && fs->fs_magic != FS_UFS2_MAGIC)
1101: panic("ffs_statfs");
1102: #else
1103: if (fs->fs_magic != FS_MAGIC)
1104: panic("ffs_statfs");
1105: #endif /* FFS2 */
1106:
1107: sbp->f_bsize = fs->fs_fsize;
1108: sbp->f_iosize = fs->fs_bsize;
1109: sbp->f_blocks = fs->fs_dsize;
1110: sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
1111: fs->fs_cstotal.cs_nffree;
1112: sbp->f_bavail = sbp->f_bfree - ((int64_t)fs->fs_dsize * fs->fs_minfree / 100);
1113: sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
1114: sbp->f_ffree = fs->fs_cstotal.cs_nifree;
1115: if (sbp != &mp->mnt_stat) {
1116: bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
1117: bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
1118: bcopy(&mp->mnt_stat.mount_info.ufs_args,
1119: &sbp->mount_info.ufs_args, sizeof(struct ufs_args));
1120: }
1121: strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN);
1122:
1123: return (0);
1124: }
1125:
1126: struct ffs_sync_args {
1127: int allerror;
1128: struct proc *p;
1129: int waitfor;
1130: struct ucred *cred;
1131: };
1132:
1133: int
1134: ffs_sync_vnode(struct vnode *vp, void *arg) {
1135: struct ffs_sync_args *fsa = arg;
1136: struct inode *ip;
1137: int error;
1138:
1139: ip = VTOI(vp);
1140: if (vp->v_type == VNON ||
1141: ((ip->i_flag &
1142: (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1143: LIST_EMPTY(&vp->v_dirtyblkhd)) ) {
1144: return (0);
1145: }
1146:
1147: if (vget(vp, LK_EXCLUSIVE | LK_NOWAIT, fsa->p))
1148: return (0);
1149:
1150: if ((error = VOP_FSYNC(vp, fsa->cred, fsa->waitfor, fsa->p)))
1151: fsa->allerror = error;
1152: VOP_UNLOCK(vp, 0, fsa->p);
1153: vrele(vp);
1154:
1155: return (0);
1156: }
1157:
1158: /*
1159: * Go through the disk queues to initiate sandbagged IO;
1160: * go through the inodes to write those that have been modified;
1161: * initiate the writing of the super block if it has been modified.
1162: *
1163: * Should always be called with the mount point locked.
1164: */
1165: int
1166: ffs_sync(struct mount *mp, int waitfor, struct ucred *cred, struct proc *p)
1167: {
1168: struct ufsmount *ump = VFSTOUFS(mp);
1169: struct fs *fs;
1170: int error, allerror = 0, count;
1171: struct ffs_sync_args fsa;
1172:
1173: fs = ump->um_fs;
1174: /*
1175: * Write back modified superblock.
1176: * Consistency check that the superblock
1177: * is still in the buffer cache.
1178: */
1179: if (fs->fs_fmod != 0 && fs->fs_ronly != 0) {
1180: printf("fs = %s\n", fs->fs_fsmnt);
1181: panic("update: rofs mod");
1182: }
1183: loop:
1184: /*
1185: * Write back each (modified) inode.
1186: */
1187: fsa.allerror = 0;
1188: fsa.p = p;
1189: fsa.cred = cred;
1190: fsa.waitfor = waitfor;
1191:
1192: /*
1193: * Don't traverse the vnode list if we want to skip all of them.
1194: */
1195: if (waitfor != MNT_LAZY) {
1196: vfs_mount_foreach_vnode(mp, ffs_sync_vnode, &fsa);
1197: allerror = fsa.allerror;
1198: }
1199:
1200: /*
1201: * Force stale file system control information to be flushed.
1202: */
1203: if ((ump->um_mountp->mnt_flag & MNT_SOFTDEP) && waitfor == MNT_WAIT) {
1204: if ((error = softdep_flushworklist(ump->um_mountp, &count, p)))
1205: allerror = error;
1206: /* Flushed work items may create new vnodes to clean */
1207: if (count)
1208: goto loop;
1209: }
1210: if (waitfor != MNT_LAZY) {
1211: vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p);
1212: if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
1213: allerror = error;
1214: VOP_UNLOCK(ump->um_devvp, 0, p);
1215: }
1216: qsync(mp);
1217: /*
1218: * Write back modified superblock.
1219: */
1220:
1221: if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
1222: allerror = error;
1223:
1224: return (allerror);
1225: }
1226:
1227: /*
1228: * Look up a FFS dinode number to find its incore vnode, otherwise read it
1229: * in from disk. If it is in core, wait for the lock bit to clear, then
1230: * return the inode locked. Detection and handling of mount points must be
1231: * done by the calling routine.
1232: */
1233: int
1234: ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1235: {
1236: struct fs *fs;
1237: struct inode *ip;
1238: struct ufs1_dinode *dp1;
1239: #ifdef FFS2
1240: struct ufs2_dinode *dp2;
1241: #endif
1242: struct ufsmount *ump;
1243: struct buf *bp;
1244: struct vnode *vp;
1245: dev_t dev;
1246: int error;
1247:
1248: ump = VFSTOUFS(mp);
1249: dev = ump->um_dev;
1250: retry:
1251: if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
1252: return (0);
1253:
1254: /* Allocate a new vnode/inode. */
1255: if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
1256: *vpp = NULL;
1257: return (error);
1258: }
1259: #ifdef VFSDEBUG
1260: vp->v_flag |= VLOCKSWORK;
1261: #endif
1262: /* XXX - we use the same pool for ffs and mfs */
1263: ip = pool_get(&ffs_ino_pool, PR_WAITOK);
1264: bzero((caddr_t)ip, sizeof(struct inode));
1265: lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
1266: ip->i_ump = ump;
1267: VREF(ip->i_devvp);
1268: vp->v_data = ip;
1269: ip->i_vnode = vp;
1270: ip->i_fs = fs = ump->um_fs;
1271: ip->i_dev = dev;
1272: ip->i_number = ino;
1273: ip->i_vtbl = &ffs_vtbl;
1274:
1275: /*
1276: * Put it onto its hash chain and lock it so that other requests for
1277: * this inode will block if they arrive while we are sleeping waiting
1278: * for old data structures to be purged or for the contents of the
1279: * disk portion of this inode to be read.
1280: */
1281: error = ufs_ihashins(ip);
1282:
1283: if (error) {
1284: /*
1285: * VOP_INACTIVE will treat this as a stale file
1286: * and recycle it quickly
1287: */
1288: vrele(vp);
1289:
1290: if (error == EEXIST)
1291: goto retry;
1292:
1293: return (error);
1294: }
1295:
1296:
1297: /* Read in the disk contents for the inode, copy into the inode. */
1298: error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1299: (int)fs->fs_bsize, NOCRED, &bp);
1300: if (error) {
1301: /*
1302: * The inode does not contain anything useful, so it would
1303: * be misleading to leave it on its hash chain. With mode
1304: * still zero, it will be unlinked and returned to the free
1305: * list by vput().
1306: */
1307: vput(vp);
1308: brelse(bp);
1309: *vpp = NULL;
1310: return (error);
1311: }
1312:
1313: #ifdef FFS2
1314: if (ip->i_ump->um_fstype == UM_UFS2) {
1315: ip->i_din2 = pool_get(&ffs_dinode2_pool, PR_WAITOK);
1316: dp2 = (struct ufs2_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
1317: *ip->i_din2 = *dp2;
1318: } else
1319: #endif
1320: {
1321: ip->i_din1 = pool_get(&ffs_dinode1_pool, PR_WAITOK);
1322: dp1 = (struct ufs1_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
1323: *ip->i_din1 = *dp1;
1324: }
1325:
1326: brelse(bp);
1327:
1328: if (DOINGSOFTDEP(vp))
1329: softdep_load_inodeblock(ip);
1330: else
1331: ip->i_effnlink = DIP(ip, nlink);
1332:
1333: /*
1334: * Initialize the vnode from the inode, check for aliases.
1335: * Note that the underlying vnode may have changed.
1336: */
1337: error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
1338: if (error) {
1339: vput(vp);
1340: *vpp = NULL;
1341: return (error);
1342: }
1343:
1344: /*
1345: * Set up a generation number for this inode if it does not
1346: * already have one. This should only happen on old filesystems.
1347: */
1348: if (DIP(ip, gen) == 0) {
1349: DIP_ASSIGN(ip, gen, arc4random() & INT_MAX);
1350: if (DIP(ip, gen) == 0 || DIP(ip, gen) == -1)
1351: DIP_ASSIGN(ip, gen, 1); /* Shouldn't happen */
1352: if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1353: ip->i_flag |= IN_MODIFIED;
1354: }
1355:
1356: /*
1357: * Ensure that uid and gid are correct. This is a temporary
1358: * fix until fsck has been changed to do the update.
1359: */
1360: if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
1361: ip->i_ffs1_uid = ip->i_din1->di_ouid;
1362: ip->i_ffs1_gid = ip->i_din1->di_ogid;
1363: }
1364:
1365: *vpp = vp;
1366:
1367: return (0);
1368: }
1369:
1370: /*
1371: * File handle to vnode
1372: *
1373: * Have to be really careful about stale file handles:
1374: * - check that the inode number is valid
1375: * - call ffs_vget() to get the locked inode
1376: * - check for an unallocated inode (i_mode == 0)
1377: */
1378: int
1379: ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1380: {
1381: struct ufid *ufhp;
1382: struct fs *fs;
1383:
1384: ufhp = (struct ufid *)fhp;
1385: fs = VFSTOUFS(mp)->um_fs;
1386: if (ufhp->ufid_ino < ROOTINO ||
1387: ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1388: return (ESTALE);
1389: return (ufs_fhtovp(mp, ufhp, vpp));
1390: }
1391:
1392: /*
1393: * Vnode pointer to File handle
1394: */
1395: /* ARGSUSED */
1396: int
1397: ffs_vptofh(struct vnode *vp, struct fid *fhp)
1398: {
1399: struct inode *ip;
1400: struct ufid *ufhp;
1401:
1402: ip = VTOI(vp);
1403: ufhp = (struct ufid *)fhp;
1404: ufhp->ufid_len = sizeof(struct ufid);
1405: ufhp->ufid_ino = ip->i_number;
1406: ufhp->ufid_gen = DIP(ip, gen);
1407:
1408: return (0);
1409: }
1410:
1411: /*
1412: * Write a superblock and associated information back to disk.
1413: */
1414: int
1415: ffs_sbupdate(struct ufsmount *mp, int waitfor)
1416: {
1417: struct fs *dfs, *fs = mp->um_fs;
1418: struct buf *bp;
1419: int blks;
1420: caddr_t space;
1421: int i, size, error, allerror = 0;
1422:
1423: /*
1424: * First write back the summary information.
1425: */
1426: blks = howmany(fs->fs_cssize, fs->fs_fsize);
1427: space = (caddr_t)fs->fs_csp;
1428: for (i = 0; i < blks; i += fs->fs_frag) {
1429: size = fs->fs_bsize;
1430: if (i + fs->fs_frag > blks)
1431: size = (blks - i) * fs->fs_fsize;
1432: bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1433: size, 0, 0);
1434: bcopy(space, bp->b_data, (u_int)size);
1435: space += size;
1436: if (waitfor != MNT_WAIT)
1437: bawrite(bp);
1438: else if ((error = bwrite(bp)))
1439: allerror = error;
1440: }
1441:
1442: /*
1443: * Now write back the superblock itself. If any errors occurred
1444: * up to this point, then fail so that the superblock avoids
1445: * being written out as clean.
1446: */
1447: if (allerror) {
1448: return (allerror);
1449: }
1450:
1451: bp = getblk(mp->um_devvp,
1452: fs->fs_sblockloc >> (fs->fs_fshift - fs->fs_fsbtodb),
1453: (int)fs->fs_sbsize, 0, 0);
1454: fs->fs_fmod = 0;
1455: fs->fs_time = time_second;
1456: bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
1457: /* Restore compatibility to old file systems. XXX */
1458: dfs = (struct fs *)bp->b_data; /* XXX */
1459: if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
1460: dfs->fs_nrpos = -1; /* XXX */
1461: if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1462: int32_t *lp, tmp; /* XXX */
1463: /* XXX */
1464: lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
1465: tmp = lp[4]; /* XXX */
1466: for (i = 4; i > 0; i--) /* XXX */
1467: lp[i] = lp[i-1]; /* XXX */
1468: lp[0] = tmp; /* XXX */
1469: } /* XXX */
1470: dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
1471:
1472: ffs1_compat_write(dfs, mp);
1473:
1474: if (waitfor != MNT_WAIT)
1475: bawrite(bp);
1476: else if ((error = bwrite(bp)))
1477: allerror = error;
1478:
1479: return (allerror);
1480: }
1481:
1482: int
1483: ffs_init(struct vfsconf *vfsp)
1484: {
1485: static int done;
1486:
1487: if (done)
1488: return (0);
1489:
1490: done = 1;
1491:
1492: pool_init(&ffs_ino_pool, sizeof(struct inode), 0, 0, 0, "ffsino",
1493: &pool_allocator_nointr);
1494: pool_init(&ffs_dinode1_pool, sizeof(struct ufs1_dinode), 0, 0, 0,
1495: "dino1pl", &pool_allocator_nointr);
1496: #ifdef FFS2
1497: pool_init(&ffs_dinode2_pool, sizeof(struct ufs2_dinode), 0, 0, 0,
1498: "dino2pl", &pool_allocator_nointr);
1499: #endif
1500:
1501: softdep_initialize();
1502:
1503: return (ufs_init(vfsp));
1504: }
1505:
1506: /*
1507: * fast filesystem related variables.
1508: */
1509: int
1510: ffs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1511: size_t newlen, struct proc *p)
1512: {
1513: extern int doclusterread, doclusterwrite, doreallocblks, doasyncfree;
1514: #ifdef FFS_SOFTUPDATES
1515: extern int max_softdeps, tickdelay, stat_worklist_push;
1516: extern int stat_blk_limit_push, stat_ino_limit_push, stat_blk_limit_hit;
1517: extern int stat_ino_limit_hit, stat_sync_limit_hit, stat_indir_blk_ptrs;
1518: extern int stat_inode_bitmap, stat_direct_blk_ptrs, stat_dir_entry;
1519: #endif
1520:
1521: /* all sysctl names at this level are terminal */
1522: if (namelen != 1)
1523: return (ENOTDIR); /* overloaded */
1524:
1525: switch (name[0]) {
1526: case FFS_CLUSTERREAD:
1527: return (sysctl_int(oldp, oldlenp, newp, newlen,
1528: &doclusterread));
1529: case FFS_CLUSTERWRITE:
1530: return (sysctl_int(oldp, oldlenp, newp, newlen,
1531: &doclusterwrite));
1532: case FFS_REALLOCBLKS:
1533: return (sysctl_int(oldp, oldlenp, newp, newlen,
1534: &doreallocblks));
1535: case FFS_ASYNCFREE:
1536: return (sysctl_int(oldp, oldlenp, newp, newlen, &doasyncfree));
1537: #ifdef FFS_SOFTUPDATES
1538: case FFS_MAX_SOFTDEPS:
1539: return (sysctl_int(oldp, oldlenp, newp, newlen, &max_softdeps));
1540: case FFS_SD_TICKDELAY:
1541: return (sysctl_int(oldp, oldlenp, newp, newlen, &tickdelay));
1542: case FFS_SD_WORKLIST_PUSH:
1543: return (sysctl_rdint(oldp, oldlenp, newp, stat_worklist_push));
1544: case FFS_SD_BLK_LIMIT_PUSH:
1545: return (sysctl_rdint(oldp, oldlenp, newp, stat_blk_limit_push));
1546: case FFS_SD_INO_LIMIT_PUSH:
1547: return (sysctl_rdint(oldp, oldlenp, newp, stat_ino_limit_push));
1548: case FFS_SD_BLK_LIMIT_HIT:
1549: return (sysctl_rdint(oldp, oldlenp, newp, stat_blk_limit_hit));
1550: case FFS_SD_INO_LIMIT_HIT:
1551: return (sysctl_rdint(oldp, oldlenp, newp, stat_ino_limit_hit));
1552: case FFS_SD_SYNC_LIMIT_HIT:
1553: return (sysctl_rdint(oldp, oldlenp, newp, stat_sync_limit_hit));
1554: case FFS_SD_INDIR_BLK_PTRS:
1555: return (sysctl_rdint(oldp, oldlenp, newp, stat_indir_blk_ptrs));
1556: case FFS_SD_INODE_BITMAP:
1557: return (sysctl_rdint(oldp, oldlenp, newp, stat_inode_bitmap));
1558: case FFS_SD_DIRECT_BLK_PTRS:
1559: return (sysctl_rdint(oldp, oldlenp, newp, stat_direct_blk_ptrs));
1560: case FFS_SD_DIR_ENTRY:
1561: return (sysctl_rdint(oldp, oldlenp, newp, stat_dir_entry));
1562: #endif
1563: #ifdef UFS_DIRHASH
1564: case FFS_DIRHASH_DIRSIZE:
1565: return (sysctl_int(oldp, oldlenp, newp, newlen,
1566: &ufs_mindirhashsize));
1567: case FFS_DIRHASH_MAXMEM:
1568: return (sysctl_int(oldp, oldlenp, newp, newlen,
1569: &ufs_dirhashmaxmem));
1570: case FFS_DIRHASH_MEM:
1571: return (sysctl_rdint(oldp, oldlenp, newp, ufs_dirhashmem));
1572: #endif
1573:
1574: default:
1575: return (EOPNOTSUPP);
1576: }
1577: /* NOTREACHED */
1578: }
CVSweb