xref: /netbsd-src/sys/ufs/mfs/mfs_vfsops.c (revision 413d532bcc3f62d122e56d92e13ac64825a40baf)
1 /*	$NetBSD: mfs_vfsops.c,v 1.107 2014/04/16 18:55:19 maxv 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 1.107 2014/04/16 18:55:19 maxv 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 
69 MODULE(MODULE_CLASS_VFS, mfs, "ffs");
70 
71 kmutex_t mfs_lock;	/* global lock */
72 
73 /* used for building internal dev_t, minor == 0 reserved for miniroot */
74 static int mfs_minor = 1;
75 static int mfs_initcnt;
76 
77 extern int (**mfs_vnodeop_p)(void *);
78 
79 static struct sysctllog *mfs_sysctl_log;
80 
81 /*
82  * mfs vfs operations.
83  */
84 
85 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc;
86 
87 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = {
88 	&mfs_vnodeop_opv_desc,
89 	NULL,
90 };
91 
92 struct vfsops mfs_vfsops = {
93 	.vfs_name = MOUNT_MFS,
94 	.vfs_min_mount_data = sizeof (struct mfs_args),
95 	.vfs_mount = mfs_mount,
96 	.vfs_start = mfs_start,
97 	.vfs_unmount = ffs_unmount,
98 	.vfs_root = ufs_root,
99 	.vfs_quotactl = ufs_quotactl,
100 	.vfs_statvfs = mfs_statvfs,
101 	.vfs_sync = ffs_sync,
102 	.vfs_vget = ffs_vget,
103 	.vfs_fhtovp = ffs_fhtovp,
104 	.vfs_vptofh = ffs_vptofh,
105 	.vfs_init = mfs_init,
106 	.vfs_reinit = mfs_reinit,
107 	.vfs_done = mfs_done,
108 	.vfs_snapshot = (void *)eopnotsupp,
109 	.vfs_extattrctl = vfs_stdextattrctl,
110 	.vfs_suspendctl = (void *)eopnotsupp,
111 	.vfs_renamelock_enter = genfs_renamelock_enter,
112 	.vfs_renamelock_exit = genfs_renamelock_exit,
113 	.vfs_fsync = (void *)eopnotsupp,
114 	.vfs_opv_descs = mfs_vnodeopv_descs
115 };
116 
117 static int
118 mfs_modcmd(modcmd_t cmd, void *arg)
119 {
120 	int error;
121 
122 	switch (cmd) {
123 	case MODULE_CMD_INIT:
124 		error = vfs_attach(&mfs_vfsops);
125 		if (error != 0)
126 			break;
127 		sysctl_createv(&mfs_sysctl_log, 0, NULL, NULL,
128 			       CTLFLAG_PERMANENT|CTLFLAG_ALIAS,
129 			       CTLTYPE_NODE, "mfs",
130 			       SYSCTL_DESCR("Memory based file system"),
131 			       NULL, 1, NULL, 0,
132 			       CTL_VFS, 3, CTL_EOL);
133 		/*
134 		 * XXX the "1" and the "3" above could be dynamic, thereby
135 		 * eliminating one more instance of the "number to vfs"
136 		 * mapping problem, but they are in order as taken from
137 		 * sys/mount.h
138 		 */
139 		break;
140 	case MODULE_CMD_FINI:
141 		error = vfs_detach(&mfs_vfsops);
142 		if (error != 0)
143 			break;
144 		sysctl_teardown(&mfs_sysctl_log);
145 		break;
146 	default:
147 		error = ENOTTY;
148 		break;
149 	}
150 
151 	return (error);
152 }
153 
154 /*
155  * Memory based filesystem initialization.
156  */
157 void
158 mfs_init(void)
159 {
160 
161 	if (mfs_initcnt++ == 0) {
162 		mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE);
163 		ffs_init();
164 	}
165 }
166 
167 void
168 mfs_reinit(void)
169 {
170 
171 	ffs_reinit();
172 }
173 
174 void
175 mfs_done(void)
176 {
177 
178 	if (--mfs_initcnt == 0) {
179 		ffs_done();
180 		mutex_destroy(&mfs_lock);
181 	}
182 }
183 
184 /*
185  * Called by main() when mfs is going to be mounted as root.
186  */
187 
188 int
189 mfs_mountroot(void)
190 {
191 	struct fs *fs;
192 	struct mount *mp;
193 	struct lwp *l = curlwp;		/* XXX */
194 	struct ufsmount *ump;
195 	struct mfsnode *mfsp;
196 	int error = 0;
197 
198 	if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
199 		vrele(rootvp);
200 		return (error);
201 	}
202 
203 	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
204 	rootvp->v_data = mfsp;
205 	rootvp->v_op = mfs_vnodeop_p;
206 	rootvp->v_tag = VT_MFS;
207 	mfsp->mfs_baseoff = mfs_rootbase;
208 	mfsp->mfs_size = mfs_rootsize;
209 	mfsp->mfs_vnode = rootvp;
210 	mfsp->mfs_proc = NULL;		/* indicate kernel space */
211 	mfsp->mfs_shutdown = 0;
212 	cv_init(&mfsp->mfs_cv, "mfs");
213 	mfsp->mfs_refcnt = 1;
214 	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
215 	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
216 		vfs_unbusy(mp, false, NULL);
217 		bufq_free(mfsp->mfs_buflist);
218 		vfs_destroy(mp);
219 		kmem_free(mfsp, sizeof(*mfsp));
220 		return (error);
221 	}
222 	mountlist_append(mp);
223 	mp->mnt_vnodecovered = NULLVP;
224 	ump = VFSTOUFS(mp);
225 	fs = ump->um_fs;
226 	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
227 	(void)ffs_statvfs(mp, &mp->mnt_stat);
228 	vfs_unbusy(mp, false, NULL);
229 	return (0);
230 }
231 
232 /*
233  * VFS Operations.
234  *
235  * mount system call
236  */
237 /* ARGSUSED */
238 int
239 mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
240 {
241 	struct lwp *l = curlwp;
242 	struct vnode *devvp;
243 	struct mfs_args *args = data;
244 	struct ufsmount *ump;
245 	struct fs *fs;
246 	struct mfsnode *mfsp;
247 	struct proc *p;
248 	int flags, error = 0;
249 
250 	if (args == NULL)
251 		return EINVAL;
252 	if (*data_len < sizeof *args)
253 		return EINVAL;
254 
255 	p = l->l_proc;
256 	if (mp->mnt_flag & MNT_GETARGS) {
257 		struct vnode *vp;
258 
259 		ump = VFSTOUFS(mp);
260 		if (ump == NULL)
261 			return EIO;
262 
263 		vp = ump->um_devvp;
264 		if (vp == NULL)
265 			return EIO;
266 
267 		mfsp = VTOMFS(vp);
268 		if (mfsp == NULL)
269 			return EIO;
270 
271 		args->fspec = NULL;
272 		args->base = mfsp->mfs_baseoff;
273 		args->size = mfsp->mfs_size;
274 		*data_len = sizeof *args;
275 		return 0;
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 	/*
289 	 * If updating, check whether changing from read-only to
290 	 * read/write; if there is no device name, that's all we do.
291 	 */
292 	if (mp->mnt_flag & MNT_UPDATE) {
293 		ump = VFSTOUFS(mp);
294 		fs = ump->um_fs;
295 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
296 			flags = WRITECLOSE;
297 			if (mp->mnt_flag & MNT_FORCE)
298 				flags |= FORCECLOSE;
299 			error = ffs_flushfiles(mp, flags, l);
300 			if (error)
301 				return (error);
302 		}
303 		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
304 			fs->fs_ronly = 0;
305 		if (args->fspec == NULL)
306 			return EINVAL;
307 		return (0);
308 	}
309 	error = getnewvnode(VT_MFS, NULL, mfs_vnodeop_p, NULL, &devvp);
310 	if (error)
311 		return (error);
312 	devvp->v_vflag |= VV_MPSAFE;
313 	devvp->v_type = VBLK;
314 	spec_node_init(devvp, makedev(255, mfs_minor));
315 	mfs_minor++;
316 	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
317 	devvp->v_data = mfsp;
318 	mfsp->mfs_baseoff = args->base;
319 	mfsp->mfs_size = args->size;
320 	mfsp->mfs_vnode = devvp;
321 	mfsp->mfs_proc = p;
322 	mfsp->mfs_shutdown = 0;
323 	cv_init(&mfsp->mfs_cv, "mfsidl");
324 	mfsp->mfs_refcnt = 1;
325 	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
326 	if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
327 		mfsp->mfs_shutdown = 1;
328 		vrele(devvp);
329 		return (error);
330 	}
331 	ump = VFSTOUFS(mp);
332 	fs = ump->um_fs;
333 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
334 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
335 	if (error)
336 		return error;
337 	(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
338 		sizeof(fs->fs_fsmnt));
339 	fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
340 	/* XXX: cleanup on error */
341 	return 0;
342 }
343 
344 /*
345  * Used to grab the process and keep it in the kernel to service
346  * memory filesystem I/O requests.
347  *
348  * Loop servicing I/O requests.
349  * Copy the requested data into or out of the memory filesystem
350  * address space.
351  */
352 /* ARGSUSED */
353 int
354 mfs_start(struct mount *mp, int flags)
355 {
356 	struct vnode *vp;
357 	struct mfsnode *mfsp;
358 	struct proc *p;
359 	struct buf *bp;
360 	void *base;
361 	int sleepreturn = 0, refcnt, error;
362 	ksiginfoq_t kq;
363 
364 	/*
365 	 * Ensure that file system is still mounted when getting mfsnode.
366 	 * Add a reference to the mfsnode to prevent it disappearing in
367 	 * this routine.
368 	 */
369 	if ((error = vfs_busy(mp, NULL)) != 0)
370 		return error;
371 	vp = VFSTOUFS(mp)->um_devvp;
372 	mfsp = VTOMFS(vp);
373 	mutex_enter(&mfs_lock);
374 	mfsp->mfs_refcnt++;
375 	mutex_exit(&mfs_lock);
376 	vfs_unbusy(mp, false, NULL);
377 
378 	base = mfsp->mfs_baseoff;
379 	mutex_enter(&mfs_lock);
380 	while (mfsp->mfs_shutdown != 1) {
381 		while ((bp = bufq_get(mfsp->mfs_buflist)) != NULL) {
382 			mutex_exit(&mfs_lock);
383 			mfs_doio(bp, base);
384 			mutex_enter(&mfs_lock);
385 		}
386 		/*
387 		 * If a non-ignored signal is received, try to unmount.
388 		 * If that fails, or the filesystem is already in the
389 		 * process of being unmounted, clear the signal (it has been
390 		 * "processed"), otherwise we will loop here, as tsleep
391 		 * will always return EINTR/ERESTART.
392 		 */
393 		if (sleepreturn != 0) {
394 			mutex_exit(&mfs_lock);
395 			if (dounmount(mp, 0, curlwp) != 0) {
396 				p = curproc;
397 				ksiginfo_queue_init(&kq);
398 				mutex_enter(p->p_lock);
399 				sigclearall(p, NULL, &kq);
400 				mutex_exit(p->p_lock);
401 				ksiginfo_queue_drain(&kq);
402 			}
403 			sleepreturn = 0;
404 			mutex_enter(&mfs_lock);
405 			continue;
406 		}
407 
408 		sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock);
409 	}
410 	KASSERT(bufq_peek(mfsp->mfs_buflist) == NULL);
411 	refcnt = --mfsp->mfs_refcnt;
412 	mutex_exit(&mfs_lock);
413 	if (refcnt == 0) {
414 		bufq_free(mfsp->mfs_buflist);
415 		cv_destroy(&mfsp->mfs_cv);
416 		kmem_free(mfsp, sizeof(*mfsp));
417 	}
418 	return (sleepreturn);
419 }
420 
421 /*
422  * Get file system statistics.
423  */
424 int
425 mfs_statvfs(struct mount *mp, struct statvfs *sbp)
426 {
427 	int error;
428 
429 	error = ffs_statvfs(mp, sbp);
430 	if (error)
431 		return error;
432 	(void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name,
433 	    sizeof(sbp->f_fstypename));
434 	sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0';
435 	return 0;
436 }
437