xref: /dflybsd-src/sys/vfs/mfs/mfs_vfsops.c (revision 41871674d0079dec70d55eb824f39d07dc7b3310)
1 /*
2  * Copyright (c) 1989, 1990, 1993, 1994
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)mfs_vfsops.c	8.11 (Berkeley) 6/19/95
34  * $FreeBSD: src/sys/ufs/mfs/mfs_vfsops.c,v 1.81.2.3 2001/07/04 17:35:21 tegge Exp $
35  * $DragonFly: src/sys/vfs/mfs/mfs_vfsops.c,v 1.28 2006/04/02 01:35:34 dillon Exp $
36  */
37 
38 
39 #include "opt_mfs.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/conf.h>
44 #include <sys/kernel.h>
45 #include <sys/proc.h>
46 #include <sys/buf.h>
47 #include <sys/mount.h>
48 #include <sys/signalvar.h>
49 #include <sys/vnode.h>
50 #include <sys/malloc.h>
51 #include <sys/linker.h>
52 #include <sys/fcntl.h>
53 
54 #include <vm/vm.h>
55 #include <vm/vm_object.h>
56 #include <vm/vm_page.h>
57 #include <vm/vm_pager.h>
58 #include <vm/vnode_pager.h>
59 
60 #include <sys/buf2.h>
61 #include <sys/thread2.h>
62 
63 #include <vfs/ufs/quota.h>
64 #include <vfs/ufs/inode.h>
65 #include <vfs/ufs/ufsmount.h>
66 #include <vfs/ufs/ufs_extern.h>
67 #include <vfs/ufs/fs.h>
68 #include <vfs/ufs/ffs_extern.h>
69 
70 #include "mfsnode.h"
71 #include "mfs_extern.h"
72 
73 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
74 
75 
76 extern struct vop_ops *mfs_vnode_vops;
77 
78 static int	mfs_mount (struct mount *mp,
79 			char *path, caddr_t data, struct thread *td);
80 static int	mfs_start (struct mount *mp, int flags, struct thread *td);
81 static int	mfs_statfs (struct mount *mp, struct statfs *sbp,
82 			struct thread *td);
83 static int	mfs_init (struct vfsconf *);
84 
85 d_open_t	mfsopen;
86 d_close_t	mfsclose;
87 d_strategy_t	mfsstrategy;
88 
89 #define MFS_CDEV_MAJOR	253
90 
91 static struct cdevsw mfs_cdevsw = {
92 	/* name */      "MFS",
93 	/* maj */       MFS_CDEV_MAJOR,
94 	/* flags */     D_DISK,
95 	/* port */	NULL,
96 	/* clone */	NULL,
97 
98 	/* open */      mfsopen,
99 	/* close */     mfsclose,
100 	/* read */      physread,
101 	/* write */     physwrite,
102 	/* ioctl */     noioctl,
103 	/* poll */      nopoll,
104 	/* mmap */      nommap,
105 	/* strategy */  mfsstrategy,
106 	/* dump */      nodump,
107 	/* psize */     nopsize
108 };
109 
110 /*
111  * mfs vfs operations.
112  */
113 static struct vfsops mfs_vfsops = {
114 	.vfs_mount =     	mfs_mount,
115 	.vfs_start =    	mfs_start,
116 	.vfs_unmount =   	ffs_unmount,
117 	.vfs_root =     	ufs_root,
118 	.vfs_quotactl =  	ufs_quotactl,
119 	.vfs_statfs =   	mfs_statfs,
120 	.vfs_sync =     	ffs_sync,
121 	.vfs_vget =      	ffs_vget,
122 	.vfs_fhtovp =   	ffs_fhtovp,
123 	.vfs_checkexp =  	ufs_check_export,
124 	.vfs_vptofh =   	ffs_vptofh,
125 	.vfs_init =     	mfs_init
126 };
127 
128 VFS_SET(mfs_vfsops, mfs, 0);
129 
130 /*
131  * We allow the underlying MFS block device to be opened and read.
132  */
133 int
134 mfsopen(dev_t dev, int flags, int mode, struct thread *td)
135 {
136 	if (flags & FWRITE)
137 		return(EROFS);
138 	if (dev->si_drv1)
139 		return(0);
140 	return(ENXIO);
141 }
142 
143 int
144 mfsclose(dev_t dev, int flags, int mode, struct thread *td)
145 {
146 	return(0);
147 }
148 
149 void
150 mfsstrategy(dev_t dev, struct bio *bio)
151 {
152 	struct buf *bp = bio->bio_buf;
153 	struct mfsnode *mfsp;
154 
155 	if ((mfsp = dev->si_drv1) != NULL) {
156 		off_t boff = bio->bio_offset;
157 		off_t eoff = boff + bp->b_bcount;
158 
159 		if (boff < 0) {
160 			bp->b_error = EINVAL;
161 			biodone(bio);
162 		} else if (eoff <= mfsp->mfs_size) {
163 			bioq_insert_tail(&mfsp->bio_queue, bio);
164 			wakeup((caddr_t)mfsp);
165 		} else if (boff < mfsp->mfs_size) {
166 			bp->b_bcount = mfsp->mfs_size - boff;
167 			bioq_insert_tail(&mfsp->bio_queue, bio);
168 			wakeup((caddr_t)mfsp);
169 		} else if (boff == mfsp->mfs_size) {
170 			bp->b_resid = bp->b_bcount;
171 			biodone(bio);
172 		} else {
173 			bp->b_error = EINVAL;
174 			biodone(bio);
175 		}
176 	} else {
177 		bp->b_error = ENXIO;
178 		bp->b_flags |= B_ERROR;
179 		biodone(bio);
180 	}
181 }
182 
183 /*
184  * mfs_mount
185  *
186  * Called when mounting local physical media
187  *
188  * PARAMETERS:
189  *		mountroot
190  *			mp	mount point structure
191  *			path	NULL (flag for root mount!!!)
192  *			data	<unused>
193  *			ndp	<unused>
194  *			p	process (user credentials check [statfs])
195  *
196  *		mount
197  *			mp	mount point structure
198  *			path	path to mount point
199  *			data	pointer to argument struct in user space
200  *			ndp	mount point namei() return (used for
201  *				credentials on reload), reused to look
202  *				up block device.
203  *			p	process (user credentials check)
204  *
205  * RETURNS:	0	Success
206  *		!0	error number (errno.h)
207  *
208  * LOCK STATE:
209  *
210  *		ENTRY
211  *			mount point is locked
212  *		EXIT
213  *			mount point is locked
214  *
215  * NOTES:
216  *		A NULL path can be used for a flag since the mount
217  *		system call will fail with EFAULT in copyinstr in
218  *		namei() if it is a genuine NULL from the user.
219  */
220 /* ARGSUSED */
221 static int
222 mfs_mount(struct mount *mp, char *path, caddr_t data, struct thread *td)
223 {
224 	struct vnode *devvp;
225 	struct mfs_args args;
226 	struct ufsmount *ump;
227 	struct fs *fs;
228 	struct mfsnode *mfsp;
229 	size_t size;
230 	int flags, err;
231 	int minnum;
232 	dev_t dev;
233 
234 	/*
235 	 * Use NULL path to flag a root mount
236 	 */
237 	if( path == NULL) {
238 		/*
239 		 ***
240 		 * Mounting root file system
241 		 ***
242 		 */
243 
244 		/* you lose */
245 		panic("mfs_mount: mount MFS as root: not configured!");
246 	}
247 
248 	/*
249 	 ***
250 	 * Mounting non-root file system or updating a file system
251 	 ***
252 	 */
253 
254 	/* copy in user arguments*/
255 	if ((err = copyin(data, (caddr_t)&args, sizeof (struct mfs_args))) != 0)
256 		goto error_1;
257 
258 	/*
259 	 * If updating, check whether changing from read-only to
260 	 * read/write; if there is no device name, that's all we do.
261 	 */
262 	if (mp->mnt_flag & MNT_UPDATE) {
263 		/*
264 		 ********************
265 		 * UPDATE
266 		 ********************
267 		 */
268 		ump = VFSTOUFS(mp);
269 		fs = ump->um_fs;
270 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
271 			flags = WRITECLOSE;
272 			if (mp->mnt_flag & MNT_FORCE)
273 				flags |= FORCECLOSE;
274 			err = ffs_flushfiles(mp, flags, td);
275 			if (err)
276 				goto error_1;
277 		}
278 		if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
279 			/* XXX reopen the device vnode read-write */
280 			fs->fs_ronly = 0;
281 		}
282 		/* if not updating name...*/
283 		if (args.fspec == 0) {
284 			/*
285 			 * Process export requests.  Jumping to "success"
286 			 * will return the vfs_export() error code.
287 			 */
288 			err = vfs_export(mp, &ump->um_export, &args.export);
289 			goto success;
290 		}
291 
292 		/* XXX MFS does not support name updating*/
293 		goto success;
294 	}
295 	/*
296 	 * Do the MALLOC before the getnewvnode since doing so afterward
297 	 * might cause a bogus v_data pointer to get dereferenced
298 	 * elsewhere if MALLOC should block.
299 	 */
300 	MALLOC(mfsp, struct mfsnode *, sizeof *mfsp, M_MFSNODE, M_WAITOK);
301 
302 	err = getspecialvnode(VT_MFS, NULL, &mfs_vnode_vops, &devvp, 0, 0);
303 	if (err) {
304 		FREE(mfsp, M_MFSNODE);
305 		goto error_1;
306 	}
307 
308 	minnum = (curproc->p_pid & 0xFF) |
309 		((curproc->p_pid & ~0xFF) << 8);
310 
311 	devvp->v_type = VCHR;
312 	dev = make_dev(&mfs_cdevsw, minnum, UID_ROOT, GID_WHEEL, 0600,
313 			"MFS%d", minnum >> 16);
314 	/* It is not clear that these will get initialized otherwise */
315 	dev->si_bsize_phys = DEV_BSIZE;
316 	dev->si_iosize_max = DFLTPHYS;
317 	dev->si_drv1 = mfsp;
318 	addaliasu(devvp, makeudev(MFS_CDEV_MAJOR, minnum));
319 	devvp->v_data = mfsp;
320 	mfsp->mfs_baseoff = args.base;
321 	mfsp->mfs_size = args.size;
322 	mfsp->mfs_vnode = devvp;
323 	mfsp->mfs_dev = reference_dev(dev);
324 	mfsp->mfs_td = td;
325 	mfsp->mfs_active = 1;
326 	bioq_init(&mfsp->bio_queue);
327 
328 	/*
329 	 * Our 'block' device must be backed by a VM object.  Theoretically
330 	 * we could use the anonymous memory VM object supplied by userland,
331 	 * but it would be somewhat of a complex task to deal with it
332 	 * that way since it would result in I/O requests which supply
333 	 * the VM pages from our own object.
334 	 *
335 	 * vnode_pager_alloc() is typically called when a VM object is
336 	 * being referenced externally.  We have to undo the refs for
337 	 * the self reference between vnode and object.
338 	 */
339 	vnode_pager_alloc(devvp, args.size, 0, 0);
340 	--devvp->v_usecount;
341 	--devvp->v_object->ref_count;
342 
343 	/* Save "mounted from" info for mount point (NULL pad)*/
344 	copyinstr(	args.fspec,			/* device name*/
345 			mp->mnt_stat.f_mntfromname,	/* save area*/
346 			MNAMELEN - 1,			/* max size*/
347 			&size);				/* real size*/
348 	bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
349 
350 	vx_unlock(devvp);
351 	if ((err = ffs_mountfs(devvp, mp, td, M_MFSNODE)) != 0) {
352 		mfsp->mfs_active = 0;
353 		goto error_2;
354 	}
355 
356 	/*
357 	 * Initialize FS stat information in mount struct; uses
358 	 * mp->mnt_stat.f_mntfromname.
359 	 *
360 	 * This code is common to root and non-root mounts
361 	 */
362 	VFS_STATFS(mp, &mp->mnt_stat, td);
363 
364 	goto success;
365 
366 error_2:	/* error with devvp held*/
367 
368 	/* release devvp before failing*/
369 	vrele(devvp);
370 
371 error_1:	/* no state to back out*/
372 
373 success:
374 	return( err);
375 }
376 
377 /*
378  * Used to grab the process and keep it in the kernel to service
379  * memory filesystem I/O requests.
380  *
381  * Loop servicing I/O requests.
382  * Copy the requested data into or out of the memory filesystem
383  * address space.
384  */
385 /* ARGSUSED */
386 static int
387 mfs_start(struct mount *mp, int flags, struct thread *td)
388 {
389 	struct vnode *vp = VFSTOUFS(mp)->um_devvp;
390 	struct mfsnode *mfsp = VTOMFS(vp);
391 	struct bio *bio;
392 	struct buf *bp;
393 	int gotsig = 0, sig;
394 
395 	/*
396 	 * We must prevent the system from trying to swap
397 	 * out or kill ( when swap space is low, see vm/pageout.c ) the
398 	 * process.  A deadlock can occur if the process is swapped out,
399 	 * and the system can loop trying to kill the unkillable ( while
400 	 * references exist ) MFS process when swap space is low.
401 	 */
402 	KKASSERT(curproc);
403 	PHOLD(curproc);
404 
405 	while (mfsp->mfs_active) {
406 		crit_enter();
407 
408 		while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) {
409 			bioq_remove(&mfsp->bio_queue, bio);
410 			crit_exit();
411 			bp = bio->bio_buf;
412 			mfs_doio(bio, mfsp);
413 			wakeup(bp);
414 			crit_enter();
415 		}
416 
417 		crit_exit();
418 
419 		/*
420 		 * If a non-ignored signal is received, try to unmount.
421 		 * If that fails, clear the signal (it has been "processed"),
422 		 * otherwise we will loop here, as tsleep will always return
423 		 * EINTR/ERESTART.
424 		 */
425 		/*
426 		 * Note that dounmount() may fail if work was queued after
427 		 * we slept. We have to jump hoops here to make sure that we
428 		 * process any buffers after the sleep, before we dounmount()
429 		 */
430 		if (gotsig) {
431 			gotsig = 0;
432 			if (dounmount(mp, 0, td) != 0) {
433 				KKASSERT(td->td_proc);
434 				sig = CURSIG(td->td_proc);
435 				if (sig)
436 					SIGDELSET(td->td_proc->p_siglist, sig);
437 			}
438 		}
439 		else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0))
440 			gotsig++;	/* try to unmount in next pass */
441 	}
442 	PRELE(curproc);
443 	v_release_rdev(vp);	/* hack because we do not implement CLOSE */
444 	/* XXX destroy/release devvp */
445 	return (0);
446 }
447 
448 /*
449  * Get file system statistics.
450  */
451 static int
452 mfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
453 {
454 	int error;
455 
456 	error = ffs_statfs(mp, sbp, td);
457 	sbp->f_type = mp->mnt_vfc->vfc_typenum;
458 	return (error);
459 }
460 
461 /*
462  * Memory based filesystem initialization.
463  */
464 static int
465 mfs_init(struct vfsconf *vfsp)
466 {
467 	cdevsw_add(&mfs_cdevsw, 0, 0);
468 	return (0);
469 }
470