xref: /netbsd-src/sys/ufs/mfs/mfs_vfsops.c (revision d710132b4b8ce7f7cccaaf660cb16aa16b4077a0)
1 /*	$NetBSD: mfs_vfsops.c,v 1.48 2003/04/22 17:16:21 christos Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1990, 1993, 1994
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)mfs_vfsops.c	8.11 (Berkeley) 6/19/95
36  */
37 
38 #include <sys/cdefs.h>
39 __KERNEL_RCSID(0, "$NetBSD: mfs_vfsops.c,v 1.48 2003/04/22 17:16:21 christos Exp $");
40 
41 #if defined(_KERNEL_OPT)
42 #include "opt_compat_netbsd.h"
43 #endif
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/time.h>
48 #include <sys/kernel.h>
49 #include <sys/proc.h>
50 #include <sys/buf.h>
51 #include <sys/mount.h>
52 #include <sys/signalvar.h>
53 #include <sys/vnode.h>
54 #include <sys/malloc.h>
55 
56 #include <miscfs/syncfs/syncfs.h>
57 
58 #include <ufs/ufs/quota.h>
59 #include <ufs/ufs/inode.h>
60 #include <ufs/ufs/ufsmount.h>
61 #include <ufs/ufs/ufs_extern.h>
62 
63 #include <ufs/ffs/fs.h>
64 #include <ufs/ffs/ffs_extern.h>
65 
66 #include <ufs/mfs/mfsnode.h>
67 #include <ufs/mfs/mfs_extern.h>
68 
69 caddr_t	mfs_rootbase;	/* address of mini-root in kernel virtual memory */
70 u_long	mfs_rootsize;	/* size of mini-root in bytes */
71 
72 static	int mfs_minor;	/* used for building internal dev_t */
73 
74 extern int (**mfs_vnodeop_p) __P((void *));
75 
76 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
77 
78 /*
79  * mfs vfs operations.
80  */
81 
82 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc;
83 
84 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = {
85 	&mfs_vnodeop_opv_desc,
86 	NULL,
87 };
88 
89 struct vfsops mfs_vfsops = {
90 	MOUNT_MFS,
91 	mfs_mount,
92 	mfs_start,
93 	ffs_unmount,
94 	ufs_root,
95 	ufs_quotactl,
96 	mfs_statfs,
97 	ffs_sync,
98 	ffs_vget,
99 	ffs_fhtovp,
100 	ffs_vptofh,
101 	mfs_init,
102 	mfs_reinit,
103 	mfs_done,
104 	ffs_sysctl,
105 	NULL,
106 	ufs_check_export,
107 	mfs_vnodeopv_descs,
108 };
109 
110 /*
111  * Memory based filesystem initialization.
112  */
113 void
114 mfs_init()
115 {
116 #ifdef _LKM
117 	malloc_type_attach(M_MFSNODE);
118 #endif
119 	/*
120 	 * ffs_init() ensures to initialize necessary resources
121 	 * only once.
122 	 */
123 	ffs_init();
124 }
125 
126 void
127 mfs_reinit()
128 {
129 	ffs_reinit();
130 }
131 
132 void
133 mfs_done()
134 {
135 	/*
136 	 * ffs_done() ensures to free necessary resources
137 	 * only once, when it's no more needed.
138 	 */
139 	ffs_done();
140 #ifdef _LKM
141 	malloc_type_detach(M_MFSNODE);
142 #endif
143 }
144 
145 /*
146  * Called by main() when mfs is going to be mounted as root.
147  */
148 
149 int
150 mfs_mountroot()
151 {
152 	struct fs *fs;
153 	struct mount *mp;
154 	struct proc *p = curproc;	/* XXX */
155 	struct ufsmount *ump;
156 	struct mfsnode *mfsp;
157 	int error = 0;
158 
159 	/*
160 	 * Get vnodes for rootdev.
161 	 */
162 	if (bdevvp(rootdev, &rootvp)) {
163 		printf("mfs_mountroot: can't setup bdevvp's");
164 		return (error);
165 	}
166 
167 	if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
168 		vrele(rootvp);
169 		return (error);
170 	}
171 
172 	mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK);
173 	rootvp->v_data = mfsp;
174 	rootvp->v_op = mfs_vnodeop_p;
175 	rootvp->v_tag = VT_MFS;
176 	mfsp->mfs_baseoff = mfs_rootbase;
177 	mfsp->mfs_size = mfs_rootsize;
178 	mfsp->mfs_vnode = rootvp;
179 	mfsp->mfs_proc = NULL;		/* indicate kernel space */
180 	mfsp->mfs_shutdown = 0;
181 	bufq_alloc(&mfsp->mfs_buflist, BUFQ_FCFS);
182 	if ((error = ffs_mountfs(rootvp, mp, p)) != 0) {
183 		mp->mnt_op->vfs_refcount--;
184 		vfs_unbusy(mp);
185 		bufq_free(&mfsp->mfs_buflist);
186 		free(mp, M_MOUNT);
187 		free(mfsp, M_MFSNODE);
188 		vrele(rootvp);
189 		return (error);
190 	}
191 	simple_lock(&mountlist_slock);
192 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
193 	simple_unlock(&mountlist_slock);
194 	mp->mnt_vnodecovered = NULLVP;
195 	ump = VFSTOUFS(mp);
196 	fs = ump->um_fs;
197 	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
198 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
199 	vfs_unbusy(mp);
200 	inittodr((time_t)0);
201 	return (0);
202 }
203 
204 /*
205  * This is called early in boot to set the base address and size
206  * of the mini-root.
207  */
208 int
209 mfs_initminiroot(base)
210 	caddr_t base;
211 {
212 	struct fs *fs = (struct fs *)(base + SBLOCK_UFS1);
213 
214 	/* check for valid super block */
215 	if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE ||
216 	    fs->fs_bsize < sizeof(struct fs))
217 		return (0);
218 	mountroot = mfs_mountroot;
219 	mfs_rootbase = base;
220 	mfs_rootsize = fs->fs_fsize * fs->fs_size;
221 	rootdev = makedev(255, mfs_minor);
222 	mfs_minor++;
223 	return (mfs_rootsize);
224 }
225 
226 /*
227  * VFS Operations.
228  *
229  * mount system call
230  */
231 /* ARGSUSED */
232 int
233 mfs_mount(mp, path, data, ndp, p)
234 	struct mount *mp;
235 	const char *path;
236 	void *data;
237 	struct nameidata *ndp;
238 	struct proc *p;
239 {
240 	struct vnode *devvp;
241 	struct mfs_args args;
242 	struct ufsmount *ump;
243 	struct fs *fs;
244 	struct mfsnode *mfsp;
245 	int flags, error;
246 
247 	if (mp->mnt_flag & MNT_GETARGS) {
248 		struct vnode *vp;
249 		struct mfsnode *mfsp;
250 
251 		ump = VFSTOUFS(mp);
252 		if (ump == NULL)
253 			return EIO;
254 
255 		vp = ump->um_devvp;
256 		if (vp == NULL)
257 			return EIO;
258 
259 		mfsp = VTOMFS(vp);
260 		if (mfsp == NULL)
261 			return EIO;
262 
263 		args.fspec = NULL;
264 		vfs_showexport(mp, &args.export, &ump->um_export);
265 		args.base = mfsp->mfs_baseoff;
266 		args.size = mfsp->mfs_size;
267 		return copyout(&args, data, sizeof(args));
268 	}
269 	/*
270 	 * XXX turn off async to avoid hangs when writing lots of data.
271 	 * the problem is that MFS needs to allocate pages to clean pages,
272 	 * so if we wait until the last minute to clean pages then there
273 	 * may not be any pages available to do the cleaning.
274 	 * ... and since the default partially-synchronous mode turns out
275 	 * to not be sufficient under heavy load, make it full synchronous.
276 	 */
277 	mp->mnt_flag &= ~MNT_ASYNC;
278 	mp->mnt_flag |= MNT_SYNCHRONOUS;
279 
280 	error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args));
281 	if (error)
282 		return (error);
283 
284 	/*
285 	 * If updating, check whether changing from read-only to
286 	 * read/write; if there is no device name, that's all we do.
287 	 */
288 	if (mp->mnt_flag & MNT_UPDATE) {
289 		ump = VFSTOUFS(mp);
290 		fs = ump->um_fs;
291 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
292 			flags = WRITECLOSE;
293 			if (mp->mnt_flag & MNT_FORCE)
294 				flags |= FORCECLOSE;
295 			error = ffs_flushfiles(mp, flags, p);
296 			if (error)
297 				return (error);
298 		}
299 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR))
300 			fs->fs_ronly = 0;
301 		if (args.fspec == 0)
302 			return (vfs_export(mp, &ump->um_export, &args.export));
303 		return (0);
304 	}
305 	error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp);
306 	if (error)
307 		return (error);
308 	devvp->v_type = VBLK;
309 	if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0))
310 		panic("mfs_mount: dup dev");
311 	mfs_minor++;
312 	mfsp = (struct mfsnode *)malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK);
313 	devvp->v_data = mfsp;
314 	mfsp->mfs_baseoff = args.base;
315 	mfsp->mfs_size = args.size;
316 	mfsp->mfs_vnode = devvp;
317 	mfsp->mfs_proc = p;
318 	mfsp->mfs_shutdown = 0;
319 	bufq_alloc(&mfsp->mfs_buflist, BUFQ_FCFS);
320 	if ((error = ffs_mountfs(devvp, mp, p)) != 0) {
321 		mfsp->mfs_shutdown = 1;
322 		vrele(devvp);
323 		return (error);
324 	}
325 	ump = VFSTOUFS(mp);
326 	fs = ump->um_fs;
327 	error = set_statfs_info(path, UIO_USERSPACE, args.fspec,
328 	    UIO_USERSPACE, mp, p);
329 	(void)memcpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
330 	    sizeof(mp->mnt_stat.f_mntonname));
331 	return error;
332 }
333 
334 int	mfs_pri = PWAIT | PCATCH;		/* XXX prob. temp */
335 
336 /*
337  * Used to grab the process and keep it in the kernel to service
338  * memory filesystem I/O requests.
339  *
340  * Loop servicing I/O requests.
341  * Copy the requested data into or out of the memory filesystem
342  * address space.
343  */
344 /* ARGSUSED */
345 int
346 mfs_start(mp, flags, p)
347 	struct mount *mp;
348 	int flags;
349 	struct proc *p;
350 {
351 	struct vnode *vp = VFSTOUFS(mp)->um_devvp;
352 	struct mfsnode *mfsp = VTOMFS(vp);
353 	struct buf *bp;
354 	caddr_t base;
355 	int sleepreturn = 0;
356 	struct lwp *l; /* XXX NJWLWP */
357 
358 	/* XXX NJWLWP the vnode interface again gives us a proc in a
359 	 * place where we want a execution context. Cheat.
360 	 */
361 	KASSERT(curproc == p);
362 	l = curlwp;
363 	base = mfsp->mfs_baseoff;
364 	while (mfsp->mfs_shutdown != 1) {
365 		while ((bp = BUFQ_GET(&mfsp->mfs_buflist)) != NULL) {
366 			mfs_doio(bp, base);
367 			wakeup((caddr_t)bp);
368 		}
369 		/*
370 		 * If a non-ignored signal is received, try to unmount.
371 		 * If that fails, or the filesystem is already in the
372 		 * process of being unmounted, clear the signal (it has been
373 		 * "processed"), otherwise we will loop here, as tsleep
374 		 * will always return EINTR/ERESTART.
375 		 */
376 		if (sleepreturn != 0) {
377 			/*
378 			 * XXX Freeze syncer.  Must do this before locking
379 			 * the mount point.  See dounmount() for details.
380 			 */
381 			lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
382 			if (vfs_busy(mp, LK_NOWAIT, 0) != 0)
383 				lockmgr(&syncer_lock, LK_RELEASE, NULL);
384 			else if (dounmount(mp, 0, p) != 0)
385 				CLRSIG(p, CURSIG(l));
386 			sleepreturn = 0;
387 			continue;
388 		}
389 
390 		sleepreturn = tsleep(vp, mfs_pri, "mfsidl", 0);
391 	}
392 	KASSERT(BUFQ_PEEK(&mfsp->mfs_buflist) == NULL);
393 	bufq_free(&mfsp->mfs_buflist);
394 	return (sleepreturn);
395 }
396 
397 /*
398  * Get file system statistics.
399  */
400 int
401 mfs_statfs(mp, sbp, p)
402 	struct mount *mp;
403 	struct statfs *sbp;
404 	struct proc *p;
405 {
406 	int error;
407 
408 	error = ffs_statfs(mp, sbp, p);
409 #ifdef COMPAT_09
410 	sbp->f_type = 3;
411 #else
412 	sbp->f_type = 0;
413 #endif
414 	strncpy(&sbp->f_fstypename[0], mp->mnt_op->vfs_name, MFSNAMELEN);
415 	return (error);
416 }
417