xref: /csrg-svn/sys/ufs/lfs/lfs_vfsops.c (revision 38776)
1 /*
2  * Copyright (c) 1989 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms are permitted
6  * provided that the above copyright notice and this paragraph are
7  * duplicated in all such forms and that any documentation,
8  * advertising materials, and other materials related to such
9  * distribution and use acknowledge that the software was developed
10  * by the University of California, Berkeley.  The name of the
11  * University may not be used to endorse or promote products derived
12  * from this software without specific prior written permission.
13  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
14  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
15  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
16  *
17  *	@(#)lfs_vfsops.c	7.19 (Berkeley) 08/26/89
18  */
19 
20 #include "param.h"
21 #include "systm.h"
22 #include "time.h"
23 #include "kernel.h"
24 #include "namei.h"
25 #include "vnode.h"
26 #include "mount.h"
27 #include "buf.h"
28 #include "ucred.h"
29 #include "file.h"
30 #include "disklabel.h"
31 #include "ioctl.h"
32 #include "errno.h"
33 #include "malloc.h"
34 #include "../ufs/fs.h"
35 #include "../ufs/ufsmount.h"
36 #include "../ufs/inode.h"
37 
38 /*
39  * ufs vfs operations.
40  */
41 int ufs_mount();
42 int ufs_unmount();
43 int ufs_root();
44 int ufs_statfs();
45 int ufs_sync();
46 int ufs_fhtovp();
47 int ufs_vptofh();
48 
49 struct vfsops ufs_vfsops = {
50 	ufs_mount,
51 	ufs_unmount,
52 	ufs_root,
53 	ufs_statfs,
54 	ufs_sync,
55 	ufs_fhtovp,
56 	ufs_vptofh
57 };
58 
59 /*
60  * ufs mount table.
61  */
62 struct ufsmount mounttab[NMOUNT];
63 
64 /*
65  * Called by vfs_mountroot when ufs is going to be mounted as root
66  *
67  * XXX - Need to have a way of figuring the name of the root device
68  */
69 #define ROOTNAME	"root device"
70 
71 ufs_mountroot()
72 {
73 	register struct mount *mp;
74 	extern struct vnode *rootvp;
75 	struct ufsmount *ump;
76 	register struct fs *fs;
77 	u_int size;
78 	int error;
79 
80 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
81 		M_MOUNT, M_WAITOK);
82 	mp->m_op = &ufs_vfsops;
83 	mp->m_flag = 0;
84 	mp->m_exroot = 0;
85 	error = mountfs(rootvp, mp);
86 	if (error) {
87 		free((caddr_t)mp, M_MOUNT);
88 		return (error);
89 	}
90 	error = vfs_add((struct vnode *)0, mp, 0);
91 	if (error) {
92 		(void)ufs_unmount(mp, 0);
93 		free((caddr_t)mp, M_MOUNT);
94 		return (error);
95 	}
96 	ump = VFSTOUFS(mp);
97 	fs = ump->um_fs;
98 	fs->fs_fsmnt[0] = '/';
99 	bzero(fs->fs_fsmnt + 1, sizeof(fs->fs_fsmnt) - 1);
100 	(void) copystr(ROOTNAME, ump->um_mntname, MNAMELEN - 1, &size);
101 	bzero(ump->um_mntname + size, MNAMELEN - size);
102 	vfs_unlock(mp);
103 	inittodr(fs->fs_time);
104 	return (0);
105 }
106 
107 /*
108  * VFS Operations.
109  *
110  * mount system call
111  */
112 ufs_mount(mp, path, data, ndp)
113 	struct mount *mp;
114 	char *path;
115 	caddr_t data;
116 	struct nameidata *ndp;
117 {
118 	struct vnode *devvp;
119 	struct ufs_args args;
120 	struct ufsmount *ump;
121 	register struct fs *fs;
122 	u_int size;
123 	int error;
124 
125 	if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
126 		return (error);
127 	if ((error = getmdev(&devvp, args.fspec, ndp)) != 0)
128 		return (error);
129 	error = mountfs(devvp, mp);
130 	if (error) {
131 		vrele(devvp);
132 		return (error);
133 	}
134 	ump = VFSTOUFS(mp);
135 	fs = ump->um_fs;
136 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
137 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
138 	(void) copyinstr(args.fspec, ump->um_mntname, MNAMELEN - 1, &size);
139 	bzero(ump->um_mntname + size, MNAMELEN - size);
140 	return (0);
141 }
142 
143 /*
144  * Common code for mount and mountroot
145  */
146 mountfs(devvp, mp)
147 	struct vnode *devvp;
148 	struct mount *mp;
149 {
150 	register struct ufsmount *ump;
151 	struct ufsmount *fmp = NULL;
152 	struct buf *bp = NULL;
153 	register struct fs *fs;
154 	dev_t dev = devvp->v_rdev;
155 	struct partinfo dpart;
156 	caddr_t base, space;
157 	int havepart = 0, blks;
158 	int error, i, size;
159 	int needclose = 0;
160 	int ronly = (mp->m_flag & M_RDONLY) != 0;
161 
162 	for (ump = &mounttab[0]; ump < &mounttab[NMOUNT]; ump++) {
163 		if (ump->um_fs == NULL) {
164 			if (fmp == NULL)
165 				fmp = ump;
166 		} else if (dev == ump->um_dev) {
167 			return (EBUSY);		/* needs translation */
168 		}
169 	}
170 	if ((ump = fmp) == NULL)
171 		return (EMFILE);		/* needs translation */
172 	ump->um_fs = (struct fs *)1;		/* just to reserve this slot */
173 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED);
174 	if (error) {
175 		ump->um_fs = NULL;
176 		return (error);
177 	}
178 	needclose = 1;
179 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED) != 0)
180 		size = DEV_BSIZE;
181 	else {
182 		havepart = 1;
183 		size = dpart.disklab->d_secsize;
184 	}
185 	if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) {
186 		ump->um_fs = NULL;
187 		goto out;
188 	}
189 	fs = bp->b_un.b_fs;
190 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
191 	    fs->fs_bsize < sizeof(struct fs)) {
192 		ump->um_fs = NULL;
193 		error = EINVAL;		/* XXX also needs translation */
194 		goto out;
195 	}
196 	ump->um_fs = (struct fs *)malloc((u_long)fs->fs_sbsize, M_SUPERBLK,
197 	    M_WAITOK);
198 	bcopy((caddr_t)bp->b_un.b_addr, (caddr_t)ump->um_fs,
199 	   (u_int)fs->fs_sbsize);
200 	brelse(bp);
201 	bp = NULL;
202 	fs = ump->um_fs;
203 	fs->fs_ronly = ronly;
204 	if (ronly == 0)
205 		fs->fs_fmod = 1;
206 	if (havepart) {
207 		dpart.part->p_fstype = FS_BSDFFS;
208 		dpart.part->p_fsize = fs->fs_fsize;
209 		dpart.part->p_frag = fs->fs_frag;
210 		dpart.part->p_cpg = fs->fs_cpg;
211 	}
212 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
213 	base = space = (caddr_t)malloc((u_long)fs->fs_cssize, M_SUPERBLK,
214 	    M_WAITOK);
215 	for (i = 0; i < blks; i += fs->fs_frag) {
216 		size = fs->fs_bsize;
217 		if (i + fs->fs_frag > blks)
218 			size = (blks - i) * fs->fs_fsize;
219 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
220 			NOCRED, &bp);
221 		if (error) {
222 			free((caddr_t)base, M_SUPERBLK);
223 			goto out;
224 		}
225 		bcopy((caddr_t)bp->b_un.b_addr, space, (u_int)size);
226 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
227 		space += size;
228 		brelse(bp);
229 		bp = NULL;
230 	}
231 	mp->m_data = (qaddr_t)ump;
232 	mp->m_bsize = fs->fs_bsize;
233 	mp->m_fsize = fs->fs_fsize;
234 	mp->m_fsid.val[0] = (long)dev;
235 	mp->m_fsid.val[1] = MOUNT_UFS;
236 	ump->um_mountp = mp;
237 	ump->um_dev = dev;
238 	ump->um_devvp = devvp;
239 	ump->um_qinod = NULL;
240 	devvp->v_mount = mp;
241 
242 	/* Sanity checks for old file systems.			   XXX */
243 	fs->fs_npsect = MAX(fs->fs_npsect, fs->fs_nsect);	/* XXX */
244 	fs->fs_interleave = MAX(fs->fs_interleave, 1);		/* XXX */
245 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
246 		fs->fs_nrpos = 8;				/* XXX */
247 	return (0);
248 out:
249 	if (needclose)
250 		(void) VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED);
251 	if (ump->um_fs) {
252 		free((caddr_t)ump->um_fs, M_SUPERBLK);
253 		ump->um_fs = NULL;
254 	}
255 	if (bp)
256 		brelse(bp);
257 	return (error);
258 }
259 
260 
261 /*
262  * unmount system call
263  */
264 ufs_unmount(mp, flags)
265 	struct mount *mp;
266 	int flags;
267 {
268 	register struct ufsmount *ump;
269 	register struct fs *fs;
270 	dev_t dev;
271 	int error, ronly;
272 
273 	if (flags & MNT_FORCE)
274 		return (EINVAL);
275 	ump = VFSTOUFS(mp);
276 	dev = ump->um_dev;
277 #ifdef QUOTA
278 	if (error = iflush(dev, ump->um_qinod))
279 #else
280 	if (error = iflush(dev))
281 #endif
282 		return (error);
283 #ifdef QUOTA
284 	(void)closedq(ump);
285 	/*
286 	 * Here we have to iflush again to get rid of the quota inode.
287 	 * A drag, but it would be ugly to cheat, & this doesn't happen often
288 	 */
289 	(void)iflush(dev, (struct inode *)NULL);
290 #endif
291 	fs = ump->um_fs;
292 	ronly = !fs->fs_ronly;
293 	free((caddr_t)fs->fs_csp[0], M_SUPERBLK);
294 	free((caddr_t)fs, M_SUPERBLK);
295 	ump->um_fs = NULL;
296 	ump->um_dev = NODEV;
297 	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED);
298 	ump->um_devvp->v_mount = (struct mount *)0;
299 	vrele(ump->um_devvp);
300 	ump->um_devvp = (struct vnode *)0;
301 	return (error);
302 }
303 
304 /*
305  * Return root of a filesystem
306  */
307 ufs_root(mp, vpp)
308 	struct mount *mp;
309 	struct vnode **vpp;
310 {
311 	struct inode tip, *ip;
312 	int error;
313 
314 	tip.i_dev = VFSTOUFS(mp)->um_dev;
315 	tip.i_vnode.v_mount = mp;
316 	error = iget(&tip, (ino_t)ROOTINO, &ip);
317 	if (error)
318 		return (error);
319 	*vpp = ITOV(ip);
320 	return (0);
321 }
322 
323 /*
324  * Get file system statistics.
325  */
326 ufs_statfs(mp, sbp)
327 	struct mount *mp;
328 	register struct statfs *sbp;
329 {
330 	register struct ufsmount *ump;
331 	register struct fs *fs;
332 
333 	ump = VFSTOUFS(mp);
334 	fs = ump->um_fs;
335 	if (fs->fs_magic != FS_MAGIC)
336 		panic("ufs_statfs");
337 	sbp->f_type = MOUNT_UFS;
338 	sbp->f_flags = mp->m_flag &~ (M_MLOCK|M_MWAIT);
339 	sbp->f_fsize = fs->fs_fsize;
340 	sbp->f_bsize = fs->fs_bsize;
341 	sbp->f_blocks = fs->fs_dsize;
342 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
343 		fs->fs_cstotal.cs_nffree;
344 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
345 		(fs->fs_dsize - sbp->f_bfree);
346 	if (sbp->f_bavail < 0)
347 		sbp->f_bavail = 0;
348 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg;
349 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
350 	sbp->f_fsid = mp->m_fsid;
351 	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
352 	bcopy((caddr_t)ump->um_mntname, (caddr_t)&sbp->f_mntfromname[0],
353 		MNAMELEN);
354 	return (0);
355 }
356 
357 int	syncprt = 0;
358 
359 /*
360  * Go through the disk queues to initiate sandbagged IO;
361  * go through the inodes to write those that have been modified;
362  * initiate the writing of the super block if it has been modified.
363  */
364 ufs_sync(mp, waitfor)
365 	struct mount *mp;
366 	int waitfor;
367 {
368 	register struct inode *ip;
369 	register struct ufsmount *ump = VFSTOUFS(mp);
370 	register struct fs *fs;
371 	int error = 0;
372 	static int updlock = 0;
373 
374 	if (syncprt)
375 		bufstats();
376 	if (updlock)
377 		return (EBUSY);
378 	fs = ump->um_fs;
379 	if (fs == (struct fs *)1)
380 		return (0);
381 	updlock++;
382 	/*
383 	 * Write back modified superblock.
384 	 * Consistency check that the superblock
385 	 * is still in the buffer cache.
386 	 */
387 	if (fs->fs_fmod != 0) {
388 		if (fs->fs_ronly != 0) {		/* XXX */
389 			printf("fs = %s\n", fs->fs_fsmnt);
390 			panic("update: rofs mod");
391 		}
392 		fs->fs_fmod = 0;
393 		fs->fs_time = time.tv_sec;
394 		error = sbupdate(ump, waitfor);
395 	}
396 	/*
397 	 * Write back each (modified) inode.
398 	 */
399 	for (ip = inode; ip < inodeNINODE; ip++) {
400 		if (ip->i_devvp != ump->um_devvp ||
401 		    (ip->i_flag & ILOCKED) != 0 || ITOV(ip)->v_count == 0 ||
402 		    (ip->i_flag & (IMOD|IACC|IUPD|ICHG)) == 0)
403 			continue;
404 		ILOCK(ip);
405 		VREF(ITOV(ip));
406 		error = iupdat(ip, &time, &time, waitfor == MNT_WAIT);
407 		iput(ip);
408 	}
409 	updlock = 0;
410 	/*
411 	 * Force stale buffer cache information to be flushed.
412 	 */
413 	bflush(ump->um_devvp->v_mount);
414 	return (error);
415 }
416 
417 /*
418  * Write a superblock and associated information back to disk.
419  */
420 sbupdate(mp, waitfor)
421 	struct ufsmount *mp;
422 	int waitfor;
423 {
424 	register struct fs *fs = mp->um_fs;
425 	register struct buf *bp;
426 	int blks;
427 	caddr_t space;
428 	int i, size, error = 0;
429 
430 	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize);
431 	bcopy((caddr_t)fs, bp->b_un.b_addr, (u_int)fs->fs_sbsize);
432 	/* Restore compatibility to old file systems.		   XXX */
433 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
434 		bp->b_un.b_fs->fs_nrpos = -1;			/* XXX */
435 	if (waitfor == MNT_WAIT)
436 		error = bwrite(bp);
437 	else
438 		bawrite(bp);
439 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
440 	space = (caddr_t)fs->fs_csp[0];
441 	for (i = 0; i < blks; i += fs->fs_frag) {
442 		size = fs->fs_bsize;
443 		if (i + fs->fs_frag > blks)
444 			size = (blks - i) * fs->fs_fsize;
445 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), size);
446 		bcopy(space, bp->b_un.b_addr, (u_int)size);
447 		space += size;
448 		if (waitfor == MNT_WAIT)
449 			error = bwrite(bp);
450 		else
451 			bawrite(bp);
452 	}
453 	return (error);
454 }
455 
456 /*
457  * Print out statistics on the current allocation of the buffer pool.
458  * Can be enabled to print out on every ``sync'' by setting "syncprt"
459  * above.
460  */
461 bufstats()
462 {
463 	int s, i, j, count;
464 	register struct buf *bp, *dp;
465 	int counts[MAXBSIZE/CLBYTES+1];
466 	static char *bname[BQUEUES] = { "LOCKED", "LRU", "AGE", "EMPTY" };
467 
468 	for (bp = bfreelist, i = 0; bp < &bfreelist[BQUEUES]; bp++, i++) {
469 		count = 0;
470 		for (j = 0; j <= MAXBSIZE/CLBYTES; j++)
471 			counts[j] = 0;
472 		s = splbio();
473 		for (dp = bp->av_forw; dp != bp; dp = dp->av_forw) {
474 			counts[dp->b_bufsize/CLBYTES]++;
475 			count++;
476 		}
477 		splx(s);
478 		printf("%s: total-%d", bname[i], count);
479 		for (j = 0; j <= MAXBSIZE/CLBYTES; j++)
480 			if (counts[j] != 0)
481 				printf(", %d-%d", j * CLBYTES, counts[j]);
482 		printf("\n");
483 	}
484 }
485 
486 /*
487  * File handle to vnode
488  *
489  * Have to be really careful about stale file handles:
490  * - check that the inode number is in range
491  * - call iget() to get the locked inode
492  * - check for an unallocated inode (i_mode == 0)
493  * - check that the generation number matches
494  */
495 ufs_fhtovp(mp, fhp, vpp)
496 	register struct mount *mp;
497 	struct fid *fhp;
498 	struct vnode **vpp;
499 {
500 	register struct ufid *ufhp;
501 	register struct fs *fs;
502 	struct inode tip, *ip;
503 	int error;
504 
505 	ufhp = (struct ufid *)fhp;
506 	fs = VFSTOUFS(mp)->um_fs;
507 	if (ufhp->ufid_ino < ROOTINO ||
508 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) {
509 		*vpp = (struct vnode *)0;
510 		return (EINVAL);
511 	}
512 	tip.i_dev = VFSTOUFS(mp)->um_dev;
513 	tip.i_vnode.v_mount = mp;
514 	if (error = iget(&tip, ufhp->ufid_ino, &ip)) {
515 		*vpp = (struct vnode *)0;
516 		return (error);
517 	}
518 	if (ip->i_mode == 0) {
519 		iput(ip);
520 		*vpp = (struct vnode *)0;
521 		return (EINVAL);
522 	}
523 	if (ip->i_gen != ufhp->ufid_gen) {
524 		iput(ip);
525 		*vpp = (struct vnode *)0;
526 		return (EINVAL);
527 	}
528 	*vpp = ITOV(ip);
529 	return (0);
530 }
531 
532 /*
533  * Vnode pointer to File handle
534  */
535 /* ARGSUSED */
536 ufs_vptofh(vp, fhp)
537 	struct vnode *vp;
538 	struct fid *fhp;
539 {
540 	register struct inode *ip = VTOI(vp);
541 	register struct ufid *ufhp;
542 
543 	ufhp = (struct ufid *)fhp;
544 	ufhp->ufid_len = sizeof(struct ufid);
545 	ufhp->ufid_ino = ip->i_number;
546 	ufhp->ufid_gen = ip->i_gen;
547 	return (0);
548 }
549 
550 /*
551  * Common code for mount and quota.
552  * Check that the user's argument is a reasonable
553  * thing on which to mount, and return the device number if so.
554  */
555 getmdev(devvpp, fname, ndp)
556 	struct vnode **devvpp;
557 	caddr_t fname;
558 	register struct nameidata *ndp;
559 {
560 	register struct vnode *vp;
561 	int error;
562 
563 	ndp->ni_nameiop = LOOKUP | LOCKLEAF | FOLLOW;
564 	ndp->ni_segflg = UIO_USERSPACE;
565 	ndp->ni_dirp = fname;
566 	if (error = namei(ndp)) {
567 		if (error == ENOENT)
568 			return (ENODEV);	/* needs translation */
569 		return (error);
570 	}
571 	vp = ndp->ni_vp;
572 	if (vp->v_type != VBLK) {
573 		vput(vp);
574 		return (ENOTBLK);
575 	}
576 	if (major(vp->v_rdev) >= nblkdev)
577 		return (ENXIO);
578 	iunlock(VTOI(vp));
579 	*devvpp = vp;
580 	return (0);
581 }
582