mfs_vfsops.c revision 1.99 
1/*	$NetBSD: mfs_vfsops.c,v 1.99 2008/11/13 11:10:41 ad 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 1.99 2008/11/13 11:10:41 ad 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/kmem.h>
53#include <sys/module.h>
54
55#include <miscfs/genfs/genfs.h>
56#include <miscfs/specfs/specdev.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
69MODULE(MODULE_CLASS_VFS, mfs, "ffs");
70
71void *	mfs_rootbase;	/* address of mini-root in kernel virtual memory */
72u_long	mfs_rootsize;	/* size of mini-root in bytes */
73kmutex_t mfs_lock;	/* global lock */
74
75static int mfs_minor;	/* used for building internal dev_t */
76static int mfs_initcnt;
77
78extern int (**mfs_vnodeop_p)(void *);
79
80static struct sysctllog *mfs_sysctl_log;
81
82/*
83 * mfs vfs operations.
84 */
85
86extern const struct vnodeopv_desc mfs_vnodeop_opv_desc;
87
88const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = {
89	&mfs_vnodeop_opv_desc,
90	NULL,
91};
92
93struct vfsops mfs_vfsops = {
94	MOUNT_MFS,
95	sizeof (struct mfs_args),
96	mfs_mount,
97	mfs_start,
98	ffs_unmount,
99	ufs_root,
100	ufs_quotactl,
101	mfs_statvfs,
102	ffs_sync,
103	ffs_vget,
104	ffs_fhtovp,
105	ffs_vptofh,
106	mfs_init,
107	mfs_reinit,
108	mfs_done,
109	NULL,
110	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
111	vfs_stdextattrctl,
112	(void *)eopnotsupp,	/* vfs_suspendctl */
113	genfs_renamelock_enter,
114	genfs_renamelock_exit,
115	(void *)eopnotsupp,
116	mfs_vnodeopv_descs,
117	0,
118	{ NULL, NULL },
119};
120
121static int
122mfs_modcmd(modcmd_t cmd, void *arg)
123{
124	int error;
125
126	switch (cmd) {
127	case MODULE_CMD_INIT:
128		error = vfs_attach(&mfs_vfsops);
129		if (error != 0)
130			break;
131		sysctl_createv(&mfs_sysctl_log, 0, NULL, NULL,
132			       CTLFLAG_PERMANENT,
133			       CTLTYPE_NODE, "vfs", NULL,
134			       NULL, 0, NULL, 0,
135			       CTL_VFS, CTL_EOL);
136		sysctl_createv(&mfs_sysctl_log, 0, NULL, NULL,
137			       CTLFLAG_PERMANENT|CTLFLAG_ALIAS,
138			       CTLTYPE_NODE, "mfs",
139			       SYSCTL_DESCR("Memory based file system"),
140			       NULL, 1, NULL, 0,
141			       CTL_VFS, 3, CTL_EOL);
142		/*
143		 * XXX the "1" and the "3" above could be dynamic, thereby
144		 * eliminating one more instance of the "number to vfs"
145		 * mapping problem, but they are in order as taken from
146		 * sys/mount.h
147		 */
148		break;
149	case MODULE_CMD_FINI:
150		error = vfs_detach(&mfs_vfsops);
151		if (error != 0)
152			break;
153		sysctl_teardown(&mfs_sysctl_log);
154		break;
155	default:
156		error = ENOTTY;
157		break;
158	}
159
160	return (error);
161}
162
163/*
164 * Memory based filesystem initialization.
165 */
166void
167mfs_init(void)
168{
169
170	if (mfs_initcnt++ == 0) {
171		mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE);
172		ffs_init();
173	}
174}
175
176void
177mfs_reinit(void)
178{
179
180	ffs_reinit();
181}
182
183void
184mfs_done(void)
185{
186
187	if (--mfs_initcnt == 0) {
188		ffs_done();
189		mutex_destroy(&mfs_lock);
190	}
191}
192
193/*
194 * Called by main() when mfs is going to be mounted as root.
195 */
196
197int
198mfs_mountroot(void)
199{
200	struct fs *fs;
201	struct mount *mp;
202	struct lwp *l = curlwp;		/* XXX */
203	struct ufsmount *ump;
204	struct mfsnode *mfsp;
205	int error = 0;
206
207	if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
208		vrele(rootvp);
209		return (error);
210	}
211
212	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
213	rootvp->v_data = mfsp;
214	rootvp->v_op = mfs_vnodeop_p;
215	rootvp->v_tag = VT_MFS;
216	mfsp->mfs_baseoff = mfs_rootbase;
217	mfsp->mfs_size = mfs_rootsize;
218	mfsp->mfs_vnode = rootvp;
219	mfsp->mfs_proc = NULL;		/* indicate kernel space */
220	mfsp->mfs_shutdown = 0;
221	cv_init(&mfsp->mfs_cv, "mfs");
222	mfsp->mfs_refcnt = 1;
223	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
224	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
225		vfs_unbusy(mp, false, NULL);
226		bufq_free(mfsp->mfs_buflist);
227		vfs_destroy(mp);
228		kmem_free(mfsp, sizeof(*mfsp));
229		return (error);
230	}
231	mutex_enter(&mountlist_lock);
232	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
233	mutex_exit(&mountlist_lock);
234	mp->mnt_vnodecovered = NULLVP;
235	ump = VFSTOUFS(mp);
236	fs = ump->um_fs;
237	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
238	(void)ffs_statvfs(mp, &mp->mnt_stat);
239	vfs_unbusy(mp, false, NULL);
240	return (0);
241}
242
243/*
244 * This is called early in boot to set the base address and size
245 * of the mini-root.
246 */
247int
248mfs_initminiroot(void *base)
249{
250	struct fs *fs = (struct fs *)((char *)base + SBLOCK_UFS1);
251
252	/* check for valid super block */
253	if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE ||
254	    fs->fs_bsize < sizeof(struct fs))
255		return (0);
256	mountroot = mfs_mountroot;
257	mfs_rootbase = base;
258	mfs_rootsize = fs->fs_fsize * fs->fs_size;
259	rootdev = makedev(255, mfs_minor);
260	mfs_minor++;
261	return (mfs_rootsize);
262}
263
264/*
265 * VFS Operations.
266 *
267 * mount system call
268 */
269/* ARGSUSED */
270int
271mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
272{
273	struct lwp *l = curlwp;
274	struct vnode *devvp;
275	struct mfs_args *args = data;
276	struct ufsmount *ump;
277	struct fs *fs;
278	struct mfsnode *mfsp;
279	struct proc *p;
280	int flags, error = 0;
281
282	if (*data_len < sizeof *args)
283		return EINVAL;
284
285	p = l->l_proc;
286	if (mp->mnt_flag & MNT_GETARGS) {
287		struct vnode *vp;
288
289		ump = VFSTOUFS(mp);
290		if (ump == NULL)
291			return EIO;
292
293		vp = ump->um_devvp;
294		if (vp == NULL)
295			return EIO;
296
297		mfsp = VTOMFS(vp);
298		if (mfsp == NULL)
299			return EIO;
300
301		args->fspec = NULL;
302		args->base = mfsp->mfs_baseoff;
303		args->size = mfsp->mfs_size;
304		*data_len = sizeof *args;
305		return 0;
306	}
307	/*
308	 * XXX turn off async to avoid hangs when writing lots of data.
309	 * the problem is that MFS needs to allocate pages to clean pages,
310	 * so if we wait until the last minute to clean pages then there
311	 * may not be any pages available to do the cleaning.
312	 * ... and since the default partially-synchronous mode turns out
313	 * to not be sufficient under heavy load, make it full synchronous.
314	 */
315	mp->mnt_flag &= ~MNT_ASYNC;
316	mp->mnt_flag |= MNT_SYNCHRONOUS;
317
318	/*
319	 * If updating, check whether changing from read-only to
320	 * read/write; if there is no device name, that's all we do.
321	 */
322	if (mp->mnt_flag & MNT_UPDATE) {
323		ump = VFSTOUFS(mp);
324		fs = ump->um_fs;
325		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
326			flags = WRITECLOSE;
327			if (mp->mnt_flag & MNT_FORCE)
328				flags |= FORCECLOSE;
329			error = ffs_flushfiles(mp, flags, l);
330			if (error)
331				return (error);
332		}
333		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
334			fs->fs_ronly = 0;
335		if (args->fspec == NULL)
336			return EINVAL;
337		return (0);
338	}
339	error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp);
340	if (error)
341		return (error);
342	devvp->v_vflag |= VV_MPSAFE;
343	devvp->v_type = VBLK;
344	spec_node_init(devvp, makedev(255, mfs_minor));
345	mfs_minor++;
346	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
347	devvp->v_data = mfsp;
348	mfsp->mfs_baseoff = args->base;
349	mfsp->mfs_size = args->size;
350	mfsp->mfs_vnode = devvp;
351	mfsp->mfs_proc = p;
352	mfsp->mfs_shutdown = 0;
353	cv_init(&mfsp->mfs_cv, "mfsidl");
354	mfsp->mfs_refcnt = 1;
355	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
356	if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
357		mfsp->mfs_shutdown = 1;
358		vrele(devvp);
359		return (error);
360	}
361	ump = VFSTOUFS(mp);
362	fs = ump->um_fs;
363	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
364	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
365	if (error)
366		return error;
367	(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
368		sizeof(fs->fs_fsmnt));
369	fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
370	/* XXX: cleanup on error */
371	return 0;
372}
373
374/*
375 * Used to grab the process and keep it in the kernel to service
376 * memory filesystem I/O requests.
377 *
378 * Loop servicing I/O requests.
379 * Copy the requested data into or out of the memory filesystem
380 * address space.
381 */
382/* ARGSUSED */
383int
384mfs_start(struct mount *mp, int flags)
385{
386	struct vnode *vp;
387	struct mfsnode *mfsp;
388	struct proc *p;
389	struct buf *bp;
390	void *base;
391	int sleepreturn = 0, refcnt, error;
392	ksiginfoq_t kq;
393
394	/*
395	 * Ensure that file system is still mounted when getting mfsnode.
396	 * Add a reference to the mfsnode to prevent it disappearing in
397	 * this routine.
398	 */
399	if ((error = vfs_busy(mp, NULL)) != 0)
400		return error;
401	vp = VFSTOUFS(mp)->um_devvp;
402	mfsp = VTOMFS(vp);
403	mutex_enter(&mfs_lock);
404	mfsp->mfs_refcnt++;
405	mutex_exit(&mfs_lock);
406	vfs_unbusy(mp, false, NULL);
407
408	base = mfsp->mfs_baseoff;
409	mutex_enter(&mfs_lock);
410	while (mfsp->mfs_shutdown != 1) {
411		while ((bp = BUFQ_GET(mfsp->mfs_buflist)) != NULL) {
412			mutex_exit(&mfs_lock);
413			mfs_doio(bp, base);
414			mutex_enter(&mfs_lock);
415		}
416		/*
417		 * If a non-ignored signal is received, try to unmount.
418		 * If that fails, or the filesystem is already in the
419		 * process of being unmounted, clear the signal (it has been
420		 * "processed"), otherwise we will loop here, as tsleep
421		 * will always return EINTR/ERESTART.
422		 */
423		if (sleepreturn != 0) {
424			mutex_exit(&mfs_lock);
425			if (dounmount(mp, 0, curlwp) != 0) {
426				p = curproc;
427				ksiginfo_queue_init(&kq);
428				mutex_enter(p->p_lock);
429				sigclearall(p, NULL, &kq);
430				mutex_exit(p->p_lock);
431				ksiginfo_queue_drain(&kq);
432			}
433			sleepreturn = 0;
434			mutex_enter(&mfs_lock);
435			continue;
436		}
437
438		sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock);
439	}
440	KASSERT(BUFQ_PEEK(mfsp->mfs_buflist) == NULL);
441	refcnt = --mfsp->mfs_refcnt;
442	mutex_exit(&mfs_lock);
443	if (refcnt == 0) {
444		bufq_free(mfsp->mfs_buflist);
445		cv_destroy(&mfsp->mfs_cv);
446		kmem_free(mfsp, sizeof(*mfsp));
447	}
448	return (sleepreturn);
449}
450
451/*
452 * Get file system statistics.
453 */
454int
455mfs_statvfs(struct mount *mp, struct statvfs *sbp)
456{
457	int error;
458
459	error = ffs_statvfs(mp, sbp);
460	if (error)
461		return error;
462	(void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name,
463	    sizeof(sbp->f_fstypename));
464	sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0';
465	return 0;
466}
467