mfs_vfsops.c revision 1.95 1 /* $NetBSD: mfs_vfsops.c,v 1.95 2008/04/30 12:49:17 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 1.95 2008/04/30 12:49:17 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
54 #include <miscfs/genfs/genfs.h>
55 #include <miscfs/specfs/specdev.h>
56 #include <miscfs/syncfs/syncfs.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 void * mfs_rootbase; /* address of mini-root in kernel virtual memory */
70 u_long mfs_rootsize; /* size of mini-root in bytes */
71 kmutex_t mfs_lock; /* global lock */
72
73 static int mfs_minor; /* used for building internal dev_t */
74 static int mfs_initcnt;
75
76 extern int (**mfs_vnodeop_p)(void *);
77
78 /*
79 * mfs vfs operations.
80 */
81
82 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc;
83
84 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = {
85 &mfs_vnodeop_opv_desc,
86 NULL,
87 };
88
89 struct vfsops mfs_vfsops = {
90 MOUNT_MFS,
91 sizeof (struct mfs_args),
92 mfs_mount,
93 mfs_start,
94 ffs_unmount,
95 ufs_root,
96 ufs_quotactl,
97 mfs_statvfs,
98 ffs_sync,
99 ffs_vget,
100 ffs_fhtovp,
101 ffs_vptofh,
102 mfs_init,
103 mfs_reinit,
104 mfs_done,
105 NULL,
106 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
107 vfs_stdextattrctl,
108 (void *)eopnotsupp, /* vfs_suspendctl */
109 genfs_renamelock_enter,
110 genfs_renamelock_exit,
111 (void *)eopnotsupp,
112 mfs_vnodeopv_descs,
113 0,
114 { NULL, NULL },
115 };
116 VFS_ATTACH(mfs_vfsops);
117
118 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup")
119 {
120
121 sysctl_createv(clog, 0, NULL, NULL,
122 CTLFLAG_PERMANENT,
123 CTLTYPE_NODE, "vfs", NULL,
124 NULL, 0, NULL, 0,
125 CTL_VFS, CTL_EOL);
126 sysctl_createv(clog, 0, NULL, NULL,
127 CTLFLAG_PERMANENT|CTLFLAG_ALIAS,
128 CTLTYPE_NODE, "mfs",
129 SYSCTL_DESCR("Memory based file system"),
130 NULL, 1, NULL, 0,
131 CTL_VFS, 3, CTL_EOL);
132 /*
133 * XXX the "1" and the "3" above could be dynamic, thereby
134 * eliminating one more instance of the "number to vfs"
135 * mapping problem, but they are in order as taken from
136 * sys/mount.h
137 */
138 }
139
140 /*
141 * Memory based filesystem initialization.
142 */
143 void
144 mfs_init(void)
145 {
146
147 if (mfs_initcnt++ == 0) {
148 mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE);
149 ffs_init();
150 }
151 }
152
153 void
154 mfs_reinit(void)
155 {
156
157 ffs_reinit();
158 }
159
160 void
161 mfs_done(void)
162 {
163
164 if (--mfs_initcnt == 0) {
165 ffs_done();
166 mutex_destroy(&mfs_lock);
167 }
168 }
169
170 /*
171 * Called by main() when mfs is going to be mounted as root.
172 */
173
174 int
175 mfs_mountroot(void)
176 {
177 struct fs *fs;
178 struct mount *mp;
179 struct lwp *l = curlwp; /* XXX */
180 struct ufsmount *ump;
181 struct mfsnode *mfsp;
182 int error = 0;
183
184 if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) {
185 vrele(rootvp);
186 return (error);
187 }
188
189 mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
190 rootvp->v_data = mfsp;
191 rootvp->v_op = mfs_vnodeop_p;
192 rootvp->v_tag = VT_MFS;
193 mfsp->mfs_baseoff = mfs_rootbase;
194 mfsp->mfs_size = mfs_rootsize;
195 mfsp->mfs_vnode = rootvp;
196 mfsp->mfs_proc = NULL; /* indicate kernel space */
197 mfsp->mfs_shutdown = 0;
198 cv_init(&mfsp->mfs_cv, "mfs");
199 mfsp->mfs_refcnt = 1;
200 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
201 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
202 vfs_unbusy(mp, false, NULL);
203 bufq_free(mfsp->mfs_buflist);
204 vfs_destroy(mp, false);
205 kmem_free(mfsp, sizeof(*mfsp));
206 return (error);
207 }
208 mutex_enter(&mountlist_lock);
209 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
210 mp->mnt_iflag |= IMNT_ONLIST;
211 mutex_exit(&mountlist_lock);
212 mp->mnt_vnodecovered = NULLVP;
213 ump = VFSTOUFS(mp);
214 fs = ump->um_fs;
215 (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
216 (void)ffs_statvfs(mp, &mp->mnt_stat);
217 vfs_unbusy(mp, false, NULL);
218 return (0);
219 }
220
221 /*
222 * This is called early in boot to set the base address and size
223 * of the mini-root.
224 */
225 int
226 mfs_initminiroot(void *base)
227 {
228 struct fs *fs = (struct fs *)((char *)base + SBLOCK_UFS1);
229
230 /* check for valid super block */
231 if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE ||
232 fs->fs_bsize < sizeof(struct fs))
233 return (0);
234 mountroot = mfs_mountroot;
235 mfs_rootbase = base;
236 mfs_rootsize = fs->fs_fsize * fs->fs_size;
237 rootdev = makedev(255, mfs_minor);
238 mfs_minor++;
239 return (mfs_rootsize);
240 }
241
242 /*
243 * VFS Operations.
244 *
245 * mount system call
246 */
247 /* ARGSUSED */
248 int
249 mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
250 {
251 struct lwp *l = curlwp;
252 struct vnode *devvp;
253 struct mfs_args *args = data;
254 struct ufsmount *ump;
255 struct fs *fs;
256 struct mfsnode *mfsp;
257 struct proc *p;
258 int flags, error = 0;
259
260 if (*data_len < sizeof *args)
261 return EINVAL;
262
263 p = l->l_proc;
264 if (mp->mnt_flag & MNT_GETARGS) {
265 struct vnode *vp;
266
267 ump = VFSTOUFS(mp);
268 if (ump == NULL)
269 return EIO;
270
271 vp = ump->um_devvp;
272 if (vp == NULL)
273 return EIO;
274
275 mfsp = VTOMFS(vp);
276 if (mfsp == NULL)
277 return EIO;
278
279 args->fspec = NULL;
280 args->base = mfsp->mfs_baseoff;
281 args->size = mfsp->mfs_size;
282 *data_len = sizeof *args;
283 return 0;
284 }
285 /*
286 * XXX turn off async to avoid hangs when writing lots of data.
287 * the problem is that MFS needs to allocate pages to clean pages,
288 * so if we wait until the last minute to clean pages then there
289 * may not be any pages available to do the cleaning.
290 * ... and since the default partially-synchronous mode turns out
291 * to not be sufficient under heavy load, make it full synchronous.
292 */
293 mp->mnt_flag &= ~MNT_ASYNC;
294 mp->mnt_flag |= MNT_SYNCHRONOUS;
295
296 /*
297 * If updating, check whether changing from read-only to
298 * read/write; if there is no device name, that's all we do.
299 */
300 if (mp->mnt_flag & MNT_UPDATE) {
301 ump = VFSTOUFS(mp);
302 fs = ump->um_fs;
303 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
304 flags = WRITECLOSE;
305 if (mp->mnt_flag & MNT_FORCE)
306 flags |= FORCECLOSE;
307 error = ffs_flushfiles(mp, flags, l);
308 if (error)
309 return (error);
310 }
311 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
312 fs->fs_ronly = 0;
313 if (args->fspec == NULL)
314 return EINVAL;
315 return (0);
316 }
317 error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp);
318 if (error)
319 return (error);
320 devvp->v_vflag |= VV_MPSAFE;
321 devvp->v_type = VBLK;
322 spec_node_init(devvp, makedev(255, mfs_minor));
323 mfs_minor++;
324 mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
325 devvp->v_data = mfsp;
326 mfsp->mfs_baseoff = args->base;
327 mfsp->mfs_size = args->size;
328 mfsp->mfs_vnode = devvp;
329 mfsp->mfs_proc = p;
330 mfsp->mfs_shutdown = 0;
331 cv_init(&mfsp->mfs_cv, "mfsidl");
332 mfsp->mfs_refcnt = 1;
333 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
334 if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
335 mfsp->mfs_shutdown = 1;
336 vrele(devvp);
337 return (error);
338 }
339 ump = VFSTOUFS(mp);
340 fs = ump->um_fs;
341 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
342 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
343 if (error)
344 return error;
345 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
346 sizeof(fs->fs_fsmnt));
347 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
348 /* XXX: cleanup on error */
349 return 0;
350 }
351
352 /*
353 * Used to grab the process and keep it in the kernel to service
354 * memory filesystem I/O requests.
355 *
356 * Loop servicing I/O requests.
357 * Copy the requested data into or out of the memory filesystem
358 * address space.
359 */
360 /* ARGSUSED */
361 int
362 mfs_start(struct mount *mp, int flags)
363 {
364 struct vnode *vp;
365 struct mfsnode *mfsp;
366 struct proc *p;
367 struct buf *bp;
368 void *base;
369 int sleepreturn = 0, refcnt, error;
370 ksiginfoq_t kq;
371
372 /*
373 * Ensure that file system is still mounted when getting mfsnode.
374 * Add a reference to the mfsnode to prevent it disappearing in
375 * this routine.
376 */
377 if ((error = vfs_busy(mp, RW_READER)) != 0)
378 return error;
379 vp = VFSTOUFS(mp)->um_devvp;
380 mfsp = VTOMFS(vp);
381 mutex_enter(&mfs_lock);
382 mfsp->mfs_refcnt++;
383 mutex_exit(&mfs_lock);
384 vfs_unbusy(mp, false, NULL);
385
386 base = mfsp->mfs_baseoff;
387 mutex_enter(&mfs_lock);
388 while (mfsp->mfs_shutdown != 1) {
389 while ((bp = BUFQ_GET(mfsp->mfs_buflist)) != NULL) {
390 mutex_exit(&mfs_lock);
391 mfs_doio(bp, base);
392 mutex_enter(&mfs_lock);
393 }
394 /*
395 * If a non-ignored signal is received, try to unmount.
396 * If that fails, or the filesystem is already in the
397 * process of being unmounted, clear the signal (it has been
398 * "processed"), otherwise we will loop here, as tsleep
399 * will always return EINTR/ERESTART.
400 */
401 if (sleepreturn != 0) {
402 mutex_exit(&mfs_lock);
403 /*
404 * XXX Freeze syncer. Must do this before locking
405 * the mount point. See dounmount() for details.
406 */
407 mutex_enter(&syncer_mutex);
408 if (vfs_trybusy(mp, RW_WRITER, NULL) != 0)
409 mutex_exit(&syncer_mutex);
410 else if (dounmount(mp, 0, curlwp) != 0) {
411 p = curproc;
412 ksiginfo_queue_init(&kq);
413 mutex_enter(p->p_lock);
414 sigclearall(p, NULL, &kq);
415 mutex_exit(p->p_lock);
416 ksiginfo_queue_drain(&kq);
417 }
418 sleepreturn = 0;
419 mutex_enter(&mfs_lock);
420 continue;
421 }
422
423 sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock);
424 }
425 KASSERT(BUFQ_PEEK(mfsp->mfs_buflist) == NULL);
426 refcnt = --mfsp->mfs_refcnt;
427 mutex_exit(&mfs_lock);
428 if (refcnt == 0) {
429 bufq_free(mfsp->mfs_buflist);
430 cv_destroy(&mfsp->mfs_cv);
431 kmem_free(mfsp, sizeof(*mfsp));
432 }
433 return (sleepreturn);
434 }
435
436 /*
437 * Get file system statistics.
438 */
439 int
440 mfs_statvfs(struct mount *mp, struct statvfs *sbp)
441 {
442 int error;
443
444 error = ffs_statvfs(mp, sbp);
445 if (error)
446 return error;
447 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name,
448 sizeof(sbp->f_fstypename));
449 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0';
450 return 0;
451 }
452