xref: /netbsd-src/sys/ufs/mfs/mfs_vfsops.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: mfs_vfsops.c,v 1.108 2014/05/08 08:21:53 hannken 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 1.108 2014/05/08 08:21:53 hannken 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 = ufs_vget,
103 	.vfs_loadvnode = ffs_loadvnode,
104 	.vfs_fhtovp = ffs_fhtovp,
105 	.vfs_vptofh = ffs_vptofh,
106 	.vfs_init = mfs_init,
107 	.vfs_reinit = mfs_reinit,
108 	.vfs_done = mfs_done,
109 	.vfs_snapshot = (void *)eopnotsupp,
110 	.vfs_extattrctl = vfs_stdextattrctl,
111 	.vfs_suspendctl = (void *)eopnotsupp,
112 	.vfs_renamelock_enter = genfs_renamelock_enter,
113 	.vfs_renamelock_exit = genfs_renamelock_exit,
114 	.vfs_fsync = (void *)eopnotsupp,
115 	.vfs_opv_descs = mfs_vnodeopv_descs
116 };
117 
118 static int
119 mfs_modcmd(modcmd_t cmd, void *arg)
120 {
121 	int error;
122 
123 	switch (cmd) {
124 	case MODULE_CMD_INIT:
125 		error = vfs_attach(&mfs_vfsops);
126 		if (error != 0)
127 			break;
128 		sysctl_createv(&mfs_sysctl_log, 0, NULL, NULL,
129 			       CTLFLAG_PERMANENT|CTLFLAG_ALIAS,
130 			       CTLTYPE_NODE, "mfs",
131 			       SYSCTL_DESCR("Memory based file system"),
132 			       NULL, 1, NULL, 0,
133 			       CTL_VFS, 3, CTL_EOL);
134 		/*
135 		 * XXX the "1" and the "3" above could be dynamic, thereby
136 		 * eliminating one more instance of the "number to vfs"
137 		 * mapping problem, but they are in order as taken from
138 		 * sys/mount.h
139 		 */
140 		break;
141 	case MODULE_CMD_FINI:
142 		error = vfs_detach(&mfs_vfsops);
143 		if (error != 0)
144 			break;
145 		sysctl_teardown(&mfs_sysctl_log);
146 		break;
147 	default:
148 		error = ENOTTY;
149 		break;
150 	}
151 
152 	return (error);
153 }
154 
155 /*
156  * Memory based filesystem initialization.
157  */
158 void
159 mfs_init(void)
160 {
161 
162 	if (mfs_initcnt++ == 0) {
163 		mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE);
164 		ffs_init();
165 	}
166 }
167 
168 void
169 mfs_reinit(void)
170 {
171 
172 	ffs_reinit();
173 }
174 
175 void
176 mfs_done(void)
177 {
178 
179 	if (--mfs_initcnt == 0) {
180 		ffs_done();
181 		mutex_destroy(&mfs_lock);
182 	}
183 }
184 
185 /*
186  * Called by main() when mfs is going to be mounted as root.
187  */
188 
189 int
190 mfs_mountroot(void)
191 {
192 	struct fs *fs;
193 	struct mount *mp;
194 	struct lwp *l = curlwp;		/* XXX */
195 	struct ufsmount *ump;
196 	struct mfsnode *mfsp;
197 	int error = 0;
198 
199 	if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
200 		vrele(rootvp);
201 		return (error);
202 	}
203 
204 	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
205 	rootvp->v_data = mfsp;
206 	rootvp->v_op = mfs_vnodeop_p;
207 	rootvp->v_tag = VT_MFS;
208 	mfsp->mfs_baseoff = mfs_rootbase;
209 	mfsp->mfs_size = mfs_rootsize;
210 	mfsp->mfs_vnode = rootvp;
211 	mfsp->mfs_proc = NULL;		/* indicate kernel space */
212 	mfsp->mfs_shutdown = 0;
213 	cv_init(&mfsp->mfs_cv, "mfs");
214 	mfsp->mfs_refcnt = 1;
215 	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
216 	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
217 		vfs_unbusy(mp, false, NULL);
218 		bufq_free(mfsp->mfs_buflist);
219 		vfs_destroy(mp);
220 		kmem_free(mfsp, sizeof(*mfsp));
221 		return (error);
222 	}
223 	mountlist_append(mp);
224 	mp->mnt_vnodecovered = NULLVP;
225 	ump = VFSTOUFS(mp);
226 	fs = ump->um_fs;
227 	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
228 	(void)ffs_statvfs(mp, &mp->mnt_stat);
229 	vfs_unbusy(mp, false, NULL);
230 	return (0);
231 }
232 
233 /*
234  * VFS Operations.
235  *
236  * mount system call
237  */
238 /* ARGSUSED */
239 int
240 mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
241 {
242 	struct lwp *l = curlwp;
243 	struct vnode *devvp;
244 	struct mfs_args *args = data;
245 	struct ufsmount *ump;
246 	struct fs *fs;
247 	struct mfsnode *mfsp;
248 	struct proc *p;
249 	int flags, error = 0;
250 
251 	if (args == NULL)
252 		return EINVAL;
253 	if (*data_len < sizeof *args)
254 		return EINVAL;
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 		*data_len = sizeof *args;
276 		return 0;
277 	}
278 	/*
279 	 * XXX turn off async to avoid hangs when writing lots of data.
280 	 * the problem is that MFS needs to allocate pages to clean pages,
281 	 * so if we wait until the last minute to clean pages then there
282 	 * may not be any pages available to do the cleaning.
283 	 * ... and since the default partially-synchronous mode turns out
284 	 * to not be sufficient under heavy load, make it full synchronous.
285 	 */
286 	mp->mnt_flag &= ~MNT_ASYNC;
287 	mp->mnt_flag |= MNT_SYNCHRONOUS;
288 
289 	/*
290 	 * If updating, check whether changing from read-only to
291 	 * read/write; if there is no device name, that's all we do.
292 	 */
293 	if (mp->mnt_flag & MNT_UPDATE) {
294 		ump = VFSTOUFS(mp);
295 		fs = ump->um_fs;
296 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
297 			flags = WRITECLOSE;
298 			if (mp->mnt_flag & MNT_FORCE)
299 				flags |= FORCECLOSE;
300 			error = ffs_flushfiles(mp, flags, l);
301 			if (error)
302 				return (error);
303 		}
304 		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
305 			fs->fs_ronly = 0;
306 		if (args->fspec == NULL)
307 			return EINVAL;
308 		return (0);
309 	}
310 	error = getnewvnode(VT_MFS, NULL, mfs_vnodeop_p, NULL, &devvp);
311 	if (error)
312 		return (error);
313 	devvp->v_vflag |= VV_MPSAFE;
314 	devvp->v_type = VBLK;
315 	spec_node_init(devvp, makedev(255, mfs_minor));
316 	mfs_minor++;
317 	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
318 	devvp->v_data = mfsp;
319 	mfsp->mfs_baseoff = args->base;
320 	mfsp->mfs_size = args->size;
321 	mfsp->mfs_vnode = devvp;
322 	mfsp->mfs_proc = p;
323 	mfsp->mfs_shutdown = 0;
324 	cv_init(&mfsp->mfs_cv, "mfsidl");
325 	mfsp->mfs_refcnt = 1;
326 	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
327 	if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
328 		mfsp->mfs_shutdown = 1;
329 		vrele(devvp);
330 		return (error);
331 	}
332 	ump = VFSTOUFS(mp);
333 	fs = ump->um_fs;
334 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
335 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
336 	if (error)
337 		return error;
338 	(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
339 		sizeof(fs->fs_fsmnt));
340 	fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
341 	/* XXX: cleanup on error */
342 	return 0;
343 }
344 
345 /*
346  * Used to grab the process and keep it in the kernel to service
347  * memory filesystem I/O requests.
348  *
349  * Loop servicing I/O requests.
350  * Copy the requested data into or out of the memory filesystem
351  * address space.
352  */
353 /* ARGSUSED */
354 int
355 mfs_start(struct mount *mp, int flags)
356 {
357 	struct vnode *vp;
358 	struct mfsnode *mfsp;
359 	struct proc *p;
360 	struct buf *bp;
361 	void *base;
362 	int sleepreturn = 0, refcnt, error;
363 	ksiginfoq_t kq;
364 
365 	/*
366 	 * Ensure that file system is still mounted when getting mfsnode.
367 	 * Add a reference to the mfsnode to prevent it disappearing in
368 	 * this routine.
369 	 */
370 	if ((error = vfs_busy(mp, NULL)) != 0)
371 		return error;
372 	vp = VFSTOUFS(mp)->um_devvp;
373 	mfsp = VTOMFS(vp);
374 	mutex_enter(&mfs_lock);
375 	mfsp->mfs_refcnt++;
376 	mutex_exit(&mfs_lock);
377 	vfs_unbusy(mp, false, NULL);
378 
379 	base = mfsp->mfs_baseoff;
380 	mutex_enter(&mfs_lock);
381 	while (mfsp->mfs_shutdown != 1) {
382 		while ((bp = bufq_get(mfsp->mfs_buflist)) != NULL) {
383 			mutex_exit(&mfs_lock);
384 			mfs_doio(bp, base);
385 			mutex_enter(&mfs_lock);
386 		}
387 		/*
388 		 * If a non-ignored signal is received, try to unmount.
389 		 * If that fails, or the filesystem is already in the
390 		 * process of being unmounted, clear the signal (it has been
391 		 * "processed"), otherwise we will loop here, as tsleep
392 		 * will always return EINTR/ERESTART.
393 		 */
394 		if (sleepreturn != 0) {
395 			mutex_exit(&mfs_lock);
396 			if (dounmount(mp, 0, curlwp) != 0) {
397 				p = curproc;
398 				ksiginfo_queue_init(&kq);
399 				mutex_enter(p->p_lock);
400 				sigclearall(p, NULL, &kq);
401 				mutex_exit(p->p_lock);
402 				ksiginfo_queue_drain(&kq);
403 			}
404 			sleepreturn = 0;
405 			mutex_enter(&mfs_lock);
406 			continue;
407 		}
408 
409 		sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock);
410 	}
411 	KASSERT(bufq_peek(mfsp->mfs_buflist) == NULL);
412 	refcnt = --mfsp->mfs_refcnt;
413 	mutex_exit(&mfs_lock);
414 	if (refcnt == 0) {
415 		bufq_free(mfsp->mfs_buflist);
416 		cv_destroy(&mfsp->mfs_cv);
417 		kmem_free(mfsp, sizeof(*mfsp));
418 	}
419 	return (sleepreturn);
420 }
421 
422 /*
423  * Get file system statistics.
424  */
425 int
426 mfs_statvfs(struct mount *mp, struct statvfs *sbp)
427 {
428 	int error;
429 
430 	error = ffs_statvfs(mp, sbp);
431 	if (error)
432 		return error;
433 	(void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name,
434 	    sizeof(sbp->f_fstypename));
435 	sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0';
436 	return 0;
437 }
438