1/*	$NetBSD: mfs_vfsops.c,v 1.116 2022/03/19 13:53:33 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 1.116 2022/03/19 13:53:33 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
69MODULE(MODULE_CLASS_VFS, mfs, "ffs");
70
71kmutex_t mfs_lock;	/* global lock */
72
73/* used for building internal dev_t, minor == 0 reserved for miniroot */
74static devminor_t mfs_minor = 1;
75static int mfs_initcnt;
76
77extern int (**mfs_vnodeop_p)(void *);
78
79/*
80 * mfs vfs operations.
81 */
82
83extern const struct vnodeopv_desc mfs_vnodeop_opv_desc;
84
85const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = {
86	&mfs_vnodeop_opv_desc,
87	NULL,
88};
89
90struct vfsops mfs_vfsops = {
91	.vfs_name = MOUNT_MFS,
92	.vfs_min_mount_data = sizeof (struct mfs_args),
93	.vfs_mount = mfs_mount,
94	.vfs_start = mfs_start,
95	.vfs_unmount = ffs_unmount,
96	.vfs_root = ufs_root,
97	.vfs_quotactl = ufs_quotactl,
98	.vfs_statvfs = mfs_statvfs,
99	.vfs_sync = ffs_sync,
100	.vfs_vget = ufs_vget,
101	.vfs_loadvnode = ffs_loadvnode,
102	.vfs_newvnode = ffs_newvnode,
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 = genfs_suspendctl,
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
117SYSCTL_SETUP(mfs_sysctl_setup, "mfs sysctl")
118{
119
120	sysctl_createv(clog, 0, NULL, NULL,
121		       CTLFLAG_PERMANENT|CTLFLAG_ALIAS,
122		       CTLTYPE_NODE, "mfs",
123		       SYSCTL_DESCR("Memory based file system"),
124		       NULL, 1, NULL, 0,
125		       CTL_VFS, 3, CTL_EOL);
126	/*
127	 * XXX the "1" and the "3" above could be dynamic, thereby
128	 * eliminating one more instance of the "number to vfs"
129	 * mapping problem, but they are in order as taken from
130	 * sys/mount.h
131	 */
132}
133
134static int
135mfs_modcmd(modcmd_t cmd, void *arg)
136{
137	int error;
138
139	switch (cmd) {
140	case MODULE_CMD_INIT:
141		error = vfs_attach(&mfs_vfsops);
142		if (error != 0)
143			break;
144		break;
145	case MODULE_CMD_FINI:
146		error = vfs_detach(&mfs_vfsops);
147		if (error != 0)
148			break;
149		break;
150	default:
151		error = ENOTTY;
152		break;
153	}
154
155	return (error);
156}
157
158/*
159 * Memory based filesystem initialization.
160 */
161void
162mfs_init(void)
163{
164
165	if (mfs_initcnt++ == 0) {
166		mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE);
167		ffs_init();
168	}
169}
170
171void
172mfs_reinit(void)
173{
174
175	ffs_reinit();
176}
177
178void
179mfs_done(void)
180{
181
182	if (--mfs_initcnt == 0) {
183		ffs_done();
184		mutex_destroy(&mfs_lock);
185	}
186}
187
188/*
189 * Called by main() when mfs is going to be mounted as root.
190 */
191
192int
193mfs_mountroot(void)
194{
195	struct fs *fs;
196	struct mount *mp;
197	struct lwp *l = curlwp;		/* XXX */
198	struct ufsmount *ump;
199	struct mfsnode *mfsp;
200	int error = 0;
201
202	if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
203		vrele(rootvp);
204		return (error);
205	}
206
207	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
208	rootvp->v_data = mfsp;
209	rootvp->v_op = mfs_vnodeop_p;
210	rootvp->v_tag = VT_MFS;
211	mfsp->mfs_baseoff = mfs_rootbase;
212	mfsp->mfs_size = mfs_rootsize;
213	mfsp->mfs_vnode = rootvp;
214	mfsp->mfs_proc = NULL;		/* indicate kernel space */
215	mfsp->mfs_shutdown = 0;
216	cv_init(&mfsp->mfs_cv, "mfs");
217	mfsp->mfs_refcnt = 1;
218	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
219	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
220		vfs_unbusy(mp);
221		bufq_free(mfsp->mfs_buflist);
222		vfs_rele(mp);
223		kmem_free(mfsp, sizeof(*mfsp));
224		return (error);
225	}
226	mountlist_append(mp);
227	mp->mnt_vnodecovered = NULLVP;
228	ump = VFSTOUFS(mp);
229	fs = ump->um_fs;
230	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
231	(void)ffs_statvfs(mp, &mp->mnt_stat);
232	vfs_unbusy(mp);
233	return (0);
234}
235
236/*
237 * VFS Operations.
238 *
239 * mount system call
240 */
241/* ARGSUSED */
242int
243mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
244{
245	struct lwp *l = curlwp;
246	struct vnode *devvp;
247	struct mfs_args *args = data;
248	struct ufsmount *ump;
249	struct fs *fs;
250	struct mfsnode *mfsp;
251	struct proc *p;
252	devminor_t minor;
253	int flags, error = 0;
254
255	if (args == NULL)
256		return EINVAL;
257	if (*data_len < sizeof *args)
258		return EINVAL;
259
260	p = l->l_proc;
261	if (mp->mnt_flag & MNT_GETARGS) {
262		struct vnode *vp;
263
264		ump = VFSTOUFS(mp);
265		if (ump == NULL)
266			return EIO;
267
268		vp = ump->um_devvp;
269		if (vp == NULL)
270			return EIO;
271
272		mfsp = VTOMFS(vp);
273		if (mfsp == NULL)
274			return EIO;
275
276		args->fspec = NULL;
277		args->base = mfsp->mfs_baseoff;
278		args->size = mfsp->mfs_size;
279		*data_len = sizeof *args;
280		return 0;
281	}
282	/*
283	 * XXX turn off async to avoid hangs when writing lots of data.
284	 * the problem is that MFS needs to allocate pages to clean pages,
285	 * so if we wait until the last minute to clean pages then there
286	 * may not be any pages available to do the cleaning.
287	 * ... and since the default partially-synchronous mode turns out
288	 * to not be sufficient under heavy load, make it full synchronous.
289	 */
290	mp->mnt_flag &= ~MNT_ASYNC;
291	mp->mnt_flag |= MNT_SYNCHRONOUS;
292
293	/*
294	 * If updating, check whether changing from read-only to
295	 * read/write; if there is no device name, that's all we do.
296	 */
297	if (mp->mnt_flag & MNT_UPDATE) {
298		ump = VFSTOUFS(mp);
299		fs = ump->um_fs;
300		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
301			flags = WRITECLOSE;
302			if (mp->mnt_flag & MNT_FORCE)
303				flags |= FORCECLOSE;
304			error = ffs_flushfiles(mp, flags, l);
305			if (error)
306				return (error);
307		}
308		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
309			fs->fs_ronly = 0;
310		if (args->fspec == NULL)
311			return EINVAL;
312		return (0);
313	}
314	mutex_enter(&mfs_lock);
315	minor = mfs_minor++;
316	mutex_exit(&mfs_lock);
317	error = bdevvp(makedev(255, minor), &devvp);
318	if (error)
319		return (error);
320	mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
321	/*
322	 * Changing v_op and v_data here is safe as we are
323	 * the exclusive owner of this device node.
324	 */
325	KASSERT(devvp->v_op == spec_vnodeop_p);
326	KASSERT(devvp->v_data == NULL);
327	devvp->v_op = mfs_vnodeop_p;
328	devvp->v_data = mfsp;
329	mfsp->mfs_baseoff = args->base;
330	mfsp->mfs_size = args->size;
331	mfsp->mfs_vnode = devvp;
332	mfsp->mfs_proc = p;
333	mfsp->mfs_shutdown = 0;
334	cv_init(&mfsp->mfs_cv, "mfsidl");
335	mfsp->mfs_refcnt = 1;
336	bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
337	if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
338		mfsp->mfs_shutdown = 1;
339		vrele(devvp);
340		return (error);
341	}
342	ump = VFSTOUFS(mp);
343	fs = ump->um_fs;
344	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
345	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
346	if (error)
347		return error;
348	(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
349		sizeof(fs->fs_fsmnt));
350	fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
351	/* XXX: cleanup on error */
352	return 0;
353}
354
355/*
356 * Used to grab the process and keep it in the kernel to service
357 * memory filesystem I/O requests.
358 *
359 * Loop servicing I/O requests.
360 * Copy the requested data into or out of the memory filesystem
361 * address space.
362 */
363/* ARGSUSED */
364int
365mfs_start(struct mount *mp, int flags)
366{
367	struct vnode *vp;
368	struct mfsnode *mfsp;
369	struct proc *p;
370	struct buf *bp;
371	void *base;
372	int sleepreturn = 0, refcnt, error;
373	ksiginfoq_t kq;
374
375	/*
376	 * Ensure that file system is still mounted when getting mfsnode.
377	 * Add a reference to the mfsnode to prevent it disappearing in
378	 * this routine.
379	 */
380	if ((error = vfs_busy(mp)) != 0)
381		return error;
382	vp = VFSTOUFS(mp)->um_devvp;
383	mfsp = VTOMFS(vp);
384	mutex_enter(&mfs_lock);
385	mfsp->mfs_refcnt++;
386	mutex_exit(&mfs_lock);
387	vfs_unbusy(mp);
388
389	base = mfsp->mfs_baseoff;
390	mutex_enter(&mfs_lock);
391	while (mfsp->mfs_shutdown != 1) {
392		while ((bp = bufq_get(mfsp->mfs_buflist)) != NULL) {
393			mutex_exit(&mfs_lock);
394			mfs_doio(bp, base);
395			mutex_enter(&mfs_lock);
396		}
397		/*
398		 * If a non-ignored signal is received, try to unmount.
399		 * If that fails, or the filesystem is already in the
400		 * process of being unmounted, clear the signal (it has been
401		 * "processed"), otherwise we will loop here, as tsleep
402		 * will always return EINTR/ERESTART.
403		 */
404		if (sleepreturn != 0) {
405			mutex_exit(&mfs_lock);
406			if (dounmount(mp, 0, curlwp) != 0) {
407				p = curproc;
408				ksiginfo_queue_init(&kq);
409				mutex_enter(p->p_lock);
410				sigclearall(p, NULL, &kq);
411				mutex_exit(p->p_lock);
412				ksiginfo_queue_drain(&kq);
413			}
414			sleepreturn = 0;
415			mutex_enter(&mfs_lock);
416			continue;
417		}
418
419		sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock);
420	}
421	KASSERT(bufq_peek(mfsp->mfs_buflist) == NULL);
422	refcnt = --mfsp->mfs_refcnt;
423	mutex_exit(&mfs_lock);
424	if (refcnt == 0) {
425		bufq_free(mfsp->mfs_buflist);
426		cv_destroy(&mfsp->mfs_cv);
427		kmem_free(mfsp, sizeof(*mfsp));
428	}
429	return (sleepreturn);
430}
431
432/*
433 * Get file system statistics.
434 */
435int
436mfs_statvfs(struct mount *mp, struct statvfs *sbp)
437{
438	int error;
439
440	error = ffs_statvfs(mp, sbp);
441	if (error)
442		return error;
443	(void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name,
444	    sizeof(sbp->f_fstypename));
445	sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0';
446	return 0;
447}
448