xref: /original-bsd/sys/ufs/ffs/ffs_vfsops.c (revision bafc759a)
1 /*
2  * Copyright (c) 1989, 1991 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * %sccs.include.redist.c%
6  *
7  *	@(#)ffs_vfsops.c	7.58 (Berkeley) 11/05/91
8  */
9 
10 #include <sys/param.h>
11 #include <sys/systm.h>
12 #include <sys/namei.h>
13 #include <sys/proc.h>
14 #include <sys/kernel.h>
15 #include <sys/vnode.h>
16 #include <sys/specdev.h>
17 #include <sys/mount.h>
18 #include <sys/buf.h>
19 #include <sys/file.h>
20 #include <sys/disklabel.h>
21 #include <sys/ioctl.h>
22 #include <sys/errno.h>
23 #include <sys/malloc.h>
24 
25 #include <ufs/ufs/quota.h>
26 #include <ufs/ufs/ufsmount.h>
27 #include <ufs/ufs/inode.h>
28 #include <ufs/ufs/ufs_extern.h>
29 
30 #include <ufs/ffs/fs.h>
31 #include <ufs/ffs/ffs_extern.h>
32 
33 static int ffs_sbupdate __P((struct ufsmount *, int));
34 
35 struct vfsops ufs_vfsops = {
36 	ffs_mount,
37 	ufs_start,
38 	ffs_unmount,
39 	ffs_root,
40 	ufs_quotactl,
41 	ffs_statfs,
42 	ffs_sync,
43 	ffs_fhtovp,
44 	ffs_vptofh,
45 	ffs_init,
46 };
47 
48 /*
49  * Called by vfs_mountroot when ufs is going to be mounted as root.
50  *
51  * Name is updated by mount(8) after booting.
52  */
53 #define ROOTNAME	"root_device"
54 
55 ffs_mountroot()
56 {
57 	extern struct vnode *rootvp;
58 	register struct fs *fs;
59 	register struct mount *mp;
60 	struct proc *p = curproc;	/* XXX */
61 	struct ufsmount *ump;
62 	u_int size;
63 	int error;
64 
65 	mp = malloc((u_long)sizeof(struct mount),
66 		M_MOUNT, M_WAITOK);
67 	mp->mnt_op = &ufs_vfsops;
68 	mp->mnt_flag = MNT_RDONLY;
69 	mp->mnt_exroot = 0;
70 	mp->mnt_mounth = NULLVP;
71 	if (error = ffs_mountfs(rootvp, mp, p)) {
72 		free(mp, M_MOUNT);
73 		return (error);
74 	}
75 	if (error = vfs_lock(mp)) {
76 		(void)ffs_unmount(mp, 0, p);
77 		free(mp, M_MOUNT);
78 		return (error);
79 	}
80 	rootfs = mp;
81 	mp->mnt_next = mp;
82 	mp->mnt_prev = mp;
83 	mp->mnt_vnodecovered = NULLVP;
84 	ump = VFSTOUFS(mp);
85 	fs = ump->um_fs;
86 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
87 	fs->fs_fsmnt[0] = '/';
88 	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
89 	    MNAMELEN);
90 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
91 	    &size);
92 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
93 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
94 	vfs_unlock(mp);
95 	inittodr(fs->fs_time);
96 	return (0);
97 }
98 
99 /*
100  * VFS Operations.
101  *
102  * mount system call
103  */
104 static int
105 ffs_mount(mp, path, data, ndp, p)
106 	register struct mount *mp;
107 	char *path;
108 	caddr_t data;
109 	struct nameidata *ndp;
110 	struct proc *p;
111 {
112 	struct vnode *devvp;
113 	struct ufs_args args;
114 	struct ufsmount *ump;
115 	register struct fs *fs;
116 	u_int size;
117 	int error;
118 
119 	if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
120 		return (error);
121 	/*
122 	 * Process export requests.
123 	 */
124 	if ((args.exflags & MNT_EXPORTED) || (mp->mnt_flag & MNT_EXPORTED)) {
125 		if (args.exflags & MNT_EXPORTED)
126 			mp->mnt_flag |= MNT_EXPORTED;
127 		else
128 			mp->mnt_flag &= ~MNT_EXPORTED;
129 		if (args.exflags & MNT_EXRDONLY)
130 			mp->mnt_flag |= MNT_EXRDONLY;
131 		else
132 			mp->mnt_flag &= ~MNT_EXRDONLY;
133 		mp->mnt_exroot = args.exroot;
134 	}
135 	/*
136 	 * If updating, check whether changing from read-only to
137 	 * read/write; if there is no device name, that's all we do.
138 	 */
139 	if (mp->mnt_flag & MNT_UPDATE) {
140 		ump = VFSTOUFS(mp);
141 		fs = ump->um_fs;
142 		if (fs->fs_ronly && (mp->mnt_flag & MNT_RDONLY) == 0)
143 			fs->fs_ronly = 0;
144 		if (args.fspec == 0)
145 			return (0);
146 	}
147 	/*
148 	 * Not an update, or updating the name: look up the name
149 	 * and verify that it refers to a sensible block device.
150 	 */
151 	ndp->ni_nameiop = LOOKUP | FOLLOW;
152 	ndp->ni_segflg = UIO_USERSPACE;
153 	ndp->ni_dirp = args.fspec;
154 	if (error = namei(ndp, p))
155 		return (error);
156 	devvp = ndp->ni_vp;
157 
158 	if (devvp->v_type != VBLK) {
159 		vrele(devvp);
160 		return (ENOTBLK);
161 	}
162 	if (major(devvp->v_rdev) >= nblkdev) {
163 		vrele(devvp);
164 		return (ENXIO);
165 	}
166 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
167 		error = ffs_mountfs(devvp, mp, p);
168 	else {
169 		if (devvp != ump->um_devvp)
170 			error = EINVAL;	/* needs translation */
171 		else
172 			vrele(devvp);
173 	}
174 	if (error) {
175 		vrele(devvp);
176 		return (error);
177 	}
178 	ump = VFSTOUFS(mp);
179 	fs = ump->um_fs;
180 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
181 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
182 	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
183 	    MNAMELEN);
184 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
185 	    &size);
186 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
187 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
188 	return (0);
189 }
190 
191 /*
192  * Common code for mount and mountroot
193  */
194 int
195 ffs_mountfs(devvp, mp, p)
196 	register struct vnode *devvp;
197 	struct mount *mp;
198 	struct proc *p;
199 {
200 	register struct ufsmount *ump = (struct ufsmount *)0;
201 	struct buf *bp = NULL;
202 	register struct fs *fs;
203 	dev_t dev = devvp->v_rdev;
204 	struct partinfo dpart;
205 	caddr_t base, space;
206 	int havepart = 0, blks;
207 	int error, i, size;
208 	int needclose = 0;
209 	int ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
210 	extern struct vnode *rootvp;
211 
212 	/*
213 	 * Disallow multiple mounts of the same device.
214 	 * Disallow mounting of a device that is currently in use
215 	 * (except for root, which might share swap device for miniroot).
216 	 * Flush out any old buffers remaining from a previous use.
217 	 */
218 	if (error = ufs_mountedon(devvp))
219 		return (error);
220 	if (vcount(devvp) > 1 && devvp != rootvp)
221 		return (EBUSY);
222 	vinvalbuf(devvp, 1);
223 	if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p))
224 		return (error);
225 	needclose = 1;
226 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
227 		size = DEV_BSIZE;
228 	else {
229 		havepart = 1;
230 		size = dpart.disklab->d_secsize;
231 	}
232 	if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
233 		goto out;
234 	fs = bp->b_un.b_fs;
235 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
236 	    fs->fs_bsize < sizeof(struct fs)) {
237 		error = EINVAL;		/* XXX needs translation */
238 		goto out;
239 	}
240 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
241 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_SUPERBLK,
242 	    M_WAITOK);
243 	bcopy((caddr_t)bp->b_un.b_addr, (caddr_t)ump->um_fs,
244 	   (u_int)fs->fs_sbsize);
245 	if (fs->fs_sbsize < SBSIZE)
246 		bp->b_flags |= B_INVAL;
247 	brelse(bp);
248 	bp = NULL;
249 	fs = ump->um_fs;
250 	fs->fs_ronly = ronly;
251 	if (ronly == 0)
252 		fs->fs_fmod = 1;
253 	if (havepart) {
254 		dpart.part->p_fstype = FS_BSDFFS;
255 		dpart.part->p_fsize = fs->fs_fsize;
256 		dpart.part->p_frag = fs->fs_frag;
257 		dpart.part->p_cpg = fs->fs_cpg;
258 	}
259 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
260 	base = space = malloc((u_long)fs->fs_cssize, M_SUPERBLK,
261 	    M_WAITOK);
262 	for (i = 0; i < blks; i += fs->fs_frag) {
263 		size = fs->fs_bsize;
264 		if (i + fs->fs_frag > blks)
265 			size = (blks - i) * fs->fs_fsize;
266 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
267 			NOCRED, &bp);
268 		if (error) {
269 			free(base, M_SUPERBLK);
270 			goto out;
271 		}
272 		bcopy((caddr_t)bp->b_un.b_addr, space, (u_int)size);
273 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
274 		space += size;
275 		brelse(bp);
276 		bp = NULL;
277 	}
278 	mp->mnt_data = (qaddr_t)ump;
279 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
280 	mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
281 	mp->mnt_flag |= MNT_LOCAL;
282 	ump->um_mountp = mp;
283 	ump->um_dev = dev;
284 	ump->um_devvp = devvp;
285 	for (i = 0; i < MAXQUOTAS; i++)
286 		ump->um_quotas[i] = NULLVP;
287 	devvp->v_specflags |= SI_MOUNTEDON;
288 
289 	/* Sanity checks for old file systems.			   XXX */
290 	fs->fs_npsect = MAX(fs->fs_npsect, fs->fs_nsect);	/* XXX */
291 	fs->fs_interleave = MAX(fs->fs_interleave, 1);		/* XXX */
292 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
293 		fs->fs_nrpos = 8;				/* XXX */
294 	return (0);
295 out:
296 	if (bp)
297 		brelse(bp);
298 	if (needclose)
299 		(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
300 	if (ump) {
301 		free(ump->um_fs, M_SUPERBLK);
302 		free(ump, M_UFSMNT);
303 		mp->mnt_data = (qaddr_t)0;
304 	}
305 	return (error);
306 }
307 
308 /*
309  * unmount system call
310  */
311 int
312 ffs_unmount(mp, mntflags, p)
313 	struct mount *mp;
314 	int mntflags;
315 	struct proc *p;
316 {
317 	extern int doforce;
318 	register struct ufsmount *ump;
319 	register struct fs *fs;
320 	int i, error, ronly, flags = 0;
321 
322 	if (mntflags & MNT_FORCE) {
323 		if (!doforce || mp == rootfs)
324 			return (EINVAL);
325 		flags |= FORCECLOSE;
326 	}
327 	mntflushbuf(mp, 0);
328 	if (mntinvalbuf(mp))
329 		return (EBUSY);
330 	ump = VFSTOUFS(mp);
331 #ifdef QUOTA
332 	if (mp->mnt_flag & MNT_QUOTA) {
333 		if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
334 			return (error);
335 		for (i = 0; i < MAXQUOTAS; i++) {
336 			if (ump->um_quotas[i] == NULLVP)
337 				continue;
338 			quotaoff(p, mp, i);
339 		}
340 		/*
341 		 * Here we fall through to vflush again to ensure
342 		 * that we have gotten rid of all the system vnodes.
343 		 */
344 	}
345 #endif
346 	if (error = vflush(mp, NULLVP, flags))
347 		return (error);
348 	fs = ump->um_fs;
349 	ronly = !fs->fs_ronly;
350 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
351 	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
352 		NOCRED, p);
353 	vrele(ump->um_devvp);
354 	free(fs->fs_csp[0], M_SUPERBLK);
355 	free(fs, M_SUPERBLK);
356 	free(ump, M_UFSMNT);
357 	mp->mnt_data = (qaddr_t)0;
358 	mp->mnt_flag &= ~MNT_LOCAL;
359 	return (error);
360 }
361 
362 /*
363  * Return root of a filesystem
364  */
365 int
366 ffs_root(mp, vpp)
367 	struct mount *mp;
368 	struct vnode **vpp;
369 {
370 	struct vnode *nvp;
371 	int error;
372 
373 	if (error = ffs_vget(mp, (ino_t)ROOTINO, &nvp))
374 		return (error);
375 	*vpp = nvp;
376 	return (0);
377 }
378 
379 /*
380  * Get file system statistics.
381  */
382 int
383 ffs_statfs(mp, sbp, p)
384 	struct mount *mp;
385 	register struct statfs *sbp;
386 	struct proc *p;
387 {
388 	register struct ufsmount *ump;
389 	register struct fs *fs;
390 
391 	ump = VFSTOUFS(mp);
392 	fs = ump->um_fs;
393 	if (fs->fs_magic != FS_MAGIC)
394 		panic("ffs_statfs");
395 	sbp->f_type = MOUNT_UFS;
396 	sbp->f_fsize = fs->fs_fsize;
397 	sbp->f_bsize = fs->fs_bsize;
398 	sbp->f_blocks = fs->fs_dsize;
399 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
400 		fs->fs_cstotal.cs_nffree;
401 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
402 		(fs->fs_dsize - sbp->f_bfree);
403 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
404 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
405 	if (sbp != &mp->mnt_stat) {
406 		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
407 			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
408 		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
409 			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
410 	}
411 	return (0);
412 }
413 
414 /*
415  * Go through the disk queues to initiate sandbagged IO;
416  * go through the inodes to write those that have been modified;
417  * initiate the writing of the super block if it has been modified.
418  *
419  * Note: we are always called with the filesystem marked `MPBUSY'.
420  */
421 int
422 ffs_sync(mp, waitfor)
423 	struct mount *mp;
424 	int waitfor;
425 {
426 	extern int syncprt;
427 	register struct vnode *vp;
428 	register struct inode *ip;
429 	register struct ufsmount *ump = VFSTOUFS(mp);
430 	register struct fs *fs;
431 	int error, allerror = 0;
432 
433 	if (syncprt)
434 		ufs_bufstats();
435 	fs = ump->um_fs;
436 	/*
437 	 * Write back modified superblock.
438 	 * Consistency check that the superblock
439 	 * is still in the buffer cache.
440 	 */
441 	if (fs->fs_fmod != 0) {
442 		if (fs->fs_ronly != 0) {		/* XXX */
443 			printf("fs = %s\n", fs->fs_fsmnt);
444 			panic("update: rofs mod");
445 		}
446 		fs->fs_fmod = 0;
447 		fs->fs_time = time.tv_sec;
448 		allerror = ffs_sbupdate(ump, waitfor);
449 	}
450 	/*
451 	 * Write back each (modified) inode.
452 	 */
453 loop:
454 	for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) {
455 		/*
456 		 * If the vnode that we are about to sync is no longer
457 		 * associated with this mount point, start over.
458 		 */
459 		if (vp->v_mount != mp)
460 			goto loop;
461 		if (VOP_ISLOCKED(vp))
462 			continue;
463 		ip = VTOI(vp);
464 		if ((ip->i_flag & (IMOD|IACC|IUPD|ICHG)) == 0 &&
465 		    vp->v_dirtyblkhd == NULL)
466 			continue;
467 		if (vget(vp))
468 			goto loop;
469 		if (vp->v_dirtyblkhd)
470 			vflushbuf(vp, 0);
471 		if ((ip->i_flag & (IMOD|IACC|IUPD|ICHG)) &&
472 		    (error = ffs_update(vp, &time, &time, 0)))
473 			allerror = error;
474 		vput(vp);
475 	}
476 	/*
477 	 * Force stale file system control information to be flushed.
478 	 */
479 	vflushbuf(ump->um_devvp, waitfor == MNT_WAIT ? B_SYNC : 0);
480 #ifdef QUOTA
481 	qsync(mp);
482 #endif
483 	return (allerror);
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 valid
491  * - call ffs_vget() to get the locked inode
492  * - check for an unallocated inode (i_mode == 0)
493  * - check that the generation number matches
494  */
495 int
496 ffs_fhtovp(mp, fhp, vpp)
497 	register struct mount *mp;
498 	struct fid *fhp;
499 	struct vnode **vpp;
500 {
501 	register struct inode *ip;
502 	register struct ufid *ufhp;
503 	struct fs *fs;
504 	struct vnode *nvp;
505 	int error;
506 
507 	ufhp = (struct ufid *)fhp;
508 	fs = VFSTOUFS(mp)->um_fs;
509 	if (ufhp->ufid_ino < ROOTINO ||
510 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
511 		return (EINVAL);
512 	if (error = ffs_vget(mp, ufhp->ufid_ino, &nvp)) {
513 		*vpp = NULLVP;
514 		return (error);
515 	}
516 	ip = VTOI(nvp);
517 	if (ip->i_mode == 0) {
518 		ufs_iput(ip);
519 		*vpp = NULLVP;
520 		return (EINVAL);
521 	}
522 	if (ip->i_gen != ufhp->ufid_gen) {
523 		ufs_iput(ip);
524 		*vpp = NULLVP;
525 		return (EINVAL);
526 	}
527 	*vpp = nvp;
528 	return (0);
529 }
530 
531 /*
532  * Vnode pointer to File handle
533  */
534 /* ARGSUSED */
535 ffs_vptofh(vp, fhp)
536 	struct vnode *vp;
537 	struct fid *fhp;
538 {
539 	register struct inode *ip;
540 	register struct ufid *ufhp;
541 
542 	ip = VTOI(vp);
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  * Write a superblock and associated information back to disk.
552  */
553 static int
554 ffs_sbupdate(mp, waitfor)
555 	struct ufsmount *mp;
556 	int waitfor;
557 {
558 	register struct fs *fs = mp->um_fs;
559 	register struct buf *bp;
560 	int blks;
561 	caddr_t space;
562 	int i, size, error = 0;
563 
564 	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize);
565 	bcopy((caddr_t)fs, bp->b_un.b_addr, (u_int)fs->fs_sbsize);
566 	/* Restore compatibility to old file systems.		   XXX */
567 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
568 		bp->b_un.b_fs->fs_nrpos = -1;			/* XXX */
569 	if (waitfor == MNT_WAIT)
570 		error = bwrite(bp);
571 	else
572 		bawrite(bp);
573 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
574 	space = (caddr_t)fs->fs_csp[0];
575 	for (i = 0; i < blks; i += fs->fs_frag) {
576 		size = fs->fs_bsize;
577 		if (i + fs->fs_frag > blks)
578 			size = (blks - i) * fs->fs_fsize;
579 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), size);
580 		bcopy(space, bp->b_un.b_addr, (u_int)size);
581 		space += size;
582 		if (waitfor == MNT_WAIT)
583 			error = bwrite(bp);
584 		else
585 			bawrite(bp);
586 	}
587 	return (error);
588 }
589