xref: /netbsd-src/sys/ufs/mfs/mfs_vfsops.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: mfs_vfsops.c,v 1.74 2006/10/12 01:32:52 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. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)mfs_vfsops.c	8.11 (Berkeley) 6/19/95
32  */
33 
34 #include <sys/cdefs.h>
35 __KERNEL_RCSID(0, "$NetBSD: mfs_vfsops.c,v 1.74 2006/10/12 01:32:52 christos Exp $");
36 
37 #if defined(_KERNEL_OPT)
38 #include "opt_compat_netbsd.h"
39 #endif
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/sysctl.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/proc.h>
47 #include <sys/buf.h>
48 #include <sys/bufq.h>
49 #include <sys/mount.h>
50 #include <sys/signalvar.h>
51 #include <sys/vnode.h>
52 #include <sys/malloc.h>
53 
54 #include <miscfs/syncfs/syncfs.h>
55 
56 #include <ufs/ufs/quota.h>
57 #include <ufs/ufs/inode.h>
58 #include <ufs/ufs/ufsmount.h>
59 #include <ufs/ufs/ufs_extern.h>
60 
61 #include <ufs/ffs/fs.h>
62 #include <ufs/ffs/ffs_extern.h>
63 
64 #include <ufs/mfs/mfsnode.h>
65 #include <ufs/mfs/mfs_extern.h>
66 
67 caddr_t	mfs_rootbase;	/* address of mini-root in kernel virtual memory */
68 u_long	mfs_rootsize;	/* size of mini-root in bytes */
69 
70 static	int mfs_minor;	/* used for building internal dev_t */
71 
72 extern int (**mfs_vnodeop_p)(void *);
73 
74 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
75 
76 /*
77  * mfs vfs operations.
78  */
79 
80 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc;
81 
82 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = {
83 	&mfs_vnodeop_opv_desc,
84 	NULL,
85 };
86 
87 struct vfsops mfs_vfsops = {
88 	MOUNT_MFS,
89 	mfs_mount,
90 	mfs_start,
91 	ffs_unmount,
92 	ufs_root,
93 	ufs_quotactl,
94 	mfs_statvfs,
95 	ffs_sync,
96 	ffs_vget,
97 	ffs_fhtovp,
98 	ffs_vptofh,
99 	mfs_init,
100 	mfs_reinit,
101 	mfs_done,
102 	NULL,
103 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
104 	vfs_stdextattrctl,
105 	mfs_vnodeopv_descs,
106 	0,
107 	{ NULL, NULL },
108 };
109 VFS_ATTACH(mfs_vfsops);
110 
111 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup")
112 {
113 
114 	sysctl_createv(clog, 0, NULL, NULL,
115 		       CTLFLAG_PERMANENT,
116 		       CTLTYPE_NODE, "vfs", NULL,
117 		       NULL, 0, NULL, 0,
118 		       CTL_VFS, CTL_EOL);
119 	sysctl_createv(clog, 0, NULL, NULL,
120 		       CTLFLAG_PERMANENT|CTLFLAG_ALIAS,
121 		       CTLTYPE_NODE, "mfs",
122 		       SYSCTL_DESCR("Memory based file system"),
123 		       NULL, 1, NULL, 0,
124 		       CTL_VFS, 3, CTL_EOL);
125 	/*
126 	 * XXX the "1" and the "3" above could be dynamic, thereby
127 	 * eliminating one more instance of the "number to vfs"
128 	 * mapping problem, but they are in order as taken from
129 	 * sys/mount.h
130 	 */
131 }
132 
133 /*
134  * Memory based filesystem initialization.
135  */
136 void
137 mfs_init(void)
138 {
139 #ifdef _LKM
140 	malloc_type_attach(M_MFSNODE);
141 #endif
142 	/*
143 	 * ffs_init() ensures to initialize necessary resources
144 	 * only once.
145 	 */
146 	ffs_init();
147 }
148 
149 void
150 mfs_reinit(void)
151 {
152 	ffs_reinit();
153 }
154 
155 void
156 mfs_done(void)
157 {
158 	/*
159 	 * ffs_done() ensures to free necessary resources
160 	 * only once, when it's no more needed.
161 	 */
162 	ffs_done();
163 #ifdef _LKM
164 	malloc_type_detach(M_MFSNODE);
165 #endif
166 }
167 
168 /*
169  * Called by main() when mfs is going to be mounted as root.
170  */
171 
172 int
173 mfs_mountroot(void)
174 {
175 	struct fs *fs;
176 	struct mount *mp;
177 	struct lwp *l = curlwp;		/* XXX */
178 	struct ufsmount *ump;
179 	struct mfsnode *mfsp;
180 	int error = 0;
181 
182 	if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
183 		vrele(rootvp);
184 		return (error);
185 	}
186 
187 	mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK);
188 	rootvp->v_data = mfsp;
189 	rootvp->v_op = mfs_vnodeop_p;
190 	rootvp->v_tag = VT_MFS;
191 	mfsp->mfs_baseoff = mfs_rootbase;
192 	mfsp->mfs_size = mfs_rootsize;
193 	mfsp->mfs_vnode = rootvp;
194 	mfsp->mfs_proc = NULL;		/* indicate kernel space */
195 	mfsp->mfs_shutdown = 0;
196 	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
197 	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
198 		mp->mnt_op->vfs_refcount--;
199 		vfs_unbusy(mp);
200 		bufq_free(mfsp->mfs_buflist);
201 		free(mp, M_MOUNT);
202 		free(mfsp, M_MFSNODE);
203 		return (error);
204 	}
205 	simple_lock(&mountlist_slock);
206 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
207 	simple_unlock(&mountlist_slock);
208 	mp->mnt_vnodecovered = NULLVP;
209 	ump = VFSTOUFS(mp);
210 	fs = ump->um_fs;
211 	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
212 	(void)ffs_statvfs(mp, &mp->mnt_stat, l);
213 	vfs_unbusy(mp);
214 	return (0);
215 }
216 
217 /*
218  * This is called early in boot to set the base address and size
219  * of the mini-root.
220  */
221 int
222 mfs_initminiroot(caddr_t base)
223 {
224 	struct fs *fs = (struct fs *)(base + SBLOCK_UFS1);
225 
226 	/* check for valid super block */
227 	if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE ||
228 	    fs->fs_bsize < sizeof(struct fs))
229 		return (0);
230 	mountroot = mfs_mountroot;
231 	mfs_rootbase = base;
232 	mfs_rootsize = fs->fs_fsize * fs->fs_size;
233 	rootdev = makedev(255, mfs_minor);
234 	mfs_minor++;
235 	return (mfs_rootsize);
236 }
237 
238 /*
239  * VFS Operations.
240  *
241  * mount system call
242  */
243 /* ARGSUSED */
244 int
245 mfs_mount(struct mount *mp, const char *path, void *data,
246     struct nameidata *ndp __unused, struct lwp *l)
247 {
248 	struct vnode *devvp;
249 	struct mfs_args args;
250 	struct ufsmount *ump;
251 	struct fs *fs;
252 	struct mfsnode *mfsp;
253 	struct proc *p;
254 	int flags, error;
255 
256 	p = l->l_proc;
257 	if (mp->mnt_flag & MNT_GETARGS) {
258 		struct vnode *vp;
259 
260 		ump = VFSTOUFS(mp);
261 		if (ump == NULL)
262 			return EIO;
263 
264 		vp = ump->um_devvp;
265 		if (vp == NULL)
266 			return EIO;
267 
268 		mfsp = VTOMFS(vp);
269 		if (mfsp == NULL)
270 			return EIO;
271 
272 		args.fspec = NULL;
273 		args.base = mfsp->mfs_baseoff;
274 		args.size = mfsp->mfs_size;
275 		return copyout(&args, data, sizeof(args));
276 	}
277 	/*
278 	 * XXX turn off async to avoid hangs when writing lots of data.
279 	 * the problem is that MFS needs to allocate pages to clean pages,
280 	 * so if we wait until the last minute to clean pages then there
281 	 * may not be any pages available to do the cleaning.
282 	 * ... and since the default partially-synchronous mode turns out
283 	 * to not be sufficient under heavy load, make it full synchronous.
284 	 */
285 	mp->mnt_flag &= ~MNT_ASYNC;
286 	mp->mnt_flag |= MNT_SYNCHRONOUS;
287 
288 	error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args));
289 	if (error)
290 		return (error);
291 
292 	/*
293 	 * If updating, check whether changing from read-only to
294 	 * read/write; if there is no device name, that's all we do.
295 	 */
296 	if (mp->mnt_flag & MNT_UPDATE) {
297 		ump = VFSTOUFS(mp);
298 		fs = ump->um_fs;
299 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
300 			flags = WRITECLOSE;
301 			if (mp->mnt_flag & MNT_FORCE)
302 				flags |= FORCECLOSE;
303 			error = ffs_flushfiles(mp, flags, l);
304 			if (error)
305 				return (error);
306 		}
307 		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
308 			fs->fs_ronly = 0;
309 		if (args.fspec == NULL)
310 			return EINVAL;
311 		return (0);
312 	}
313 	error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp);
314 	if (error)
315 		return (error);
316 	devvp->v_type = VBLK;
317 	if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0))
318 		panic("mfs_mount: dup dev");
319 	mfs_minor++;
320 	mfsp = (struct mfsnode *)malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK);
321 	devvp->v_data = mfsp;
322 	mfsp->mfs_baseoff = args.base;
323 	mfsp->mfs_size = args.size;
324 	mfsp->mfs_vnode = devvp;
325 	mfsp->mfs_proc = p;
326 	mfsp->mfs_shutdown = 0;
327 	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
328 	if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
329 		mfsp->mfs_shutdown = 1;
330 		vrele(devvp);
331 		return (error);
332 	}
333 	ump = VFSTOUFS(mp);
334 	fs = ump->um_fs;
335 	error = set_statvfs_info(path, UIO_USERSPACE, args.fspec,
336 	    UIO_USERSPACE, mp, l);
337 	if (error)
338 		return error;
339 	(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
340 		sizeof(fs->fs_fsmnt));
341 	fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
342 	/* XXX: cleanup on error */
343 	return 0;
344 }
345 
346 int	mfs_pri = PWAIT | PCATCH;		/* XXX prob. temp */
347 
348 /*
349  * Used to grab the process and keep it in the kernel to service
350  * memory filesystem I/O requests.
351  *
352  * Loop servicing I/O requests.
353  * Copy the requested data into or out of the memory filesystem
354  * address space.
355  */
356 /* ARGSUSED */
357 int
358 mfs_start(struct mount *mp, int flags __unused, struct lwp *l)
359 {
360 	struct vnode *vp = VFSTOUFS(mp)->um_devvp;
361 	struct mfsnode *mfsp = VTOMFS(vp);
362 	struct buf *bp;
363 	caddr_t base;
364 	int sleepreturn = 0;
365 
366 	base = mfsp->mfs_baseoff;
367 	while (mfsp->mfs_shutdown != 1) {
368 		while ((bp = BUFQ_GET(mfsp->mfs_buflist)) != NULL) {
369 			mfs_doio(bp, base);
370 			wakeup((caddr_t)bp);
371 		}
372 		/*
373 		 * If a non-ignored signal is received, try to unmount.
374 		 * If that fails, or the filesystem is already in the
375 		 * process of being unmounted, clear the signal (it has been
376 		 * "processed"), otherwise we will loop here, as tsleep
377 		 * will always return EINTR/ERESTART.
378 		 */
379 		if (sleepreturn != 0) {
380 			/*
381 			 * XXX Freeze syncer.  Must do this before locking
382 			 * the mount point.  See dounmount() for details.
383 			 */
384 			lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
385 			if (vfs_busy(mp, LK_NOWAIT, 0) != 0)
386 				lockmgr(&syncer_lock, LK_RELEASE, NULL);
387 			else if (dounmount(mp, 0, l) != 0)
388 				CLRSIG(l);
389 			sleepreturn = 0;
390 			continue;
391 		}
392 
393 		sleepreturn = tsleep(vp, mfs_pri, "mfsidl", 0);
394 	}
395 	KASSERT(BUFQ_PEEK(mfsp->mfs_buflist) == NULL);
396 	bufq_free(mfsp->mfs_buflist);
397 	return (sleepreturn);
398 }
399 
400 /*
401  * Get file system statistics.
402  */
403 int
404 mfs_statvfs(struct mount *mp, struct statvfs *sbp, struct lwp *l)
405 {
406 	int error;
407 
408 	error = ffs_statvfs(mp, sbp, l);
409 	if (error)
410 		return error;
411 	(void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name,
412 	    sizeof(sbp->f_fstypename));
413 	sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0';
414 	return 0;
415 }
416