xref: /dragonfly/sys/vfs/mfs/mfs_vfsops.c (revision 6bd457ed)
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 1.23 2005/07/26 15:43:35 hmp 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 <sys/buf2.h>
55 
56 #include <sys/thread2.h>
57 
58 #include <vfs/ufs/quota.h>
59 #include <vfs/ufs/inode.h>
60 #include <vfs/ufs/ufsmount.h>
61 #include <vfs/ufs/ufs_extern.h>
62 #include <vfs/ufs/fs.h>
63 #include <vfs/ufs/ffs_extern.h>
64 
65 #include "mfsnode.h"
66 #include "mfs_extern.h"
67 
68 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
69 
70 
71 extern struct vop_ops *mfs_vnode_vops;
72 
73 static int	mfs_mount (struct mount *mp,
74 			char *path, caddr_t data, struct thread *td);
75 static int	mfs_start (struct mount *mp, int flags, struct thread *td);
76 static int	mfs_statfs (struct mount *mp, struct statfs *sbp,
77 			struct thread *td);
78 static int	mfs_init (struct vfsconf *);
79 
80 d_open_t	mfsopen;
81 d_close_t	mfsclose;
82 d_strategy_t	mfsstrategy;
83 
84 #define MFS_CDEV_MAJOR	253
85 
86 static struct cdevsw mfs_cdevsw = {
87 	/* name */      "MFS",
88 	/* maj */       MFS_CDEV_MAJOR,
89 	/* flags */     D_DISK,
90 	/* port */	NULL,
91 	/* clone */	NULL,
92 
93 	/* open */      mfsopen,
94 	/* close */     mfsclose,
95 	/* read */      physread,
96 	/* write */     physwrite,
97 	/* ioctl */     noioctl,
98 	/* poll */      nopoll,
99 	/* mmap */      nommap,
100 	/* strategy */  mfsstrategy,
101 	/* dump */      nodump,
102 	/* psize */     nopsize
103 };
104 
105 /*
106  * mfs vfs operations.
107  */
108 static struct vfsops mfs_vfsops = {
109 	.vfs_mount =     	mfs_mount,
110 	.vfs_start =    	mfs_start,
111 	.vfs_unmount =   	ffs_unmount,
112 	.vfs_root =     	ufs_root,
113 	.vfs_quotactl =  	ufs_quotactl,
114 	.vfs_statfs =   	mfs_statfs,
115 	.vfs_sync =     	ffs_sync,
116 	.vfs_vget =      	ffs_vget,
117 	.vfs_fhtovp =   	ffs_fhtovp,
118 	.vfs_checkexp =  	ufs_check_export,
119 	.vfs_vptofh =   	ffs_vptofh,
120 	.vfs_init =     	mfs_init
121 };
122 
123 VFS_SET(mfs_vfsops, mfs, 0);
124 
125 /*
126  * We allow the underlying MFS block device to be opened and read.
127  */
128 int
129 mfsopen(dev_t dev, int flags, int mode, struct thread *td)
130 {
131 	if (flags & FWRITE)
132 		return(EROFS);
133 	if (dev->si_drv1)
134 		return(0);
135 	return(ENXIO);
136 }
137 
138 int
139 mfsclose(dev_t dev, int flags, int mode, struct thread *td)
140 {
141 	return(0);
142 }
143 
144 void
145 mfsstrategy(struct buf *bp)
146 {
147 	struct mfsnode *mfsp;
148 
149 	if ((mfsp = bp->b_dev->si_drv1) != NULL) {
150 		off_t boff = (off_t)bp->b_blkno << DEV_BSHIFT;
151 		off_t eoff = boff + bp->b_bcount;
152 
153 		if (eoff <= mfsp->mfs_size) {
154 			bufq_insert_tail(&mfsp->buf_queue, bp);
155 			wakeup((caddr_t)mfsp);
156 		} else if (boff < mfsp->mfs_size) {
157 			bp->b_bcount = mfsp->mfs_size - boff;
158 			bufq_insert_tail(&mfsp->buf_queue, bp);
159 			wakeup((caddr_t)mfsp);
160 		} else if (boff == mfsp->mfs_size) {
161 			bp->b_resid = bp->b_bcount;
162 			biodone(bp);
163 		} else {
164 			bp->b_error = EINVAL;
165 			biodone(bp);
166 		}
167 	} else {
168 		bp->b_error = ENXIO;
169 		bp->b_flags |= B_ERROR;
170 		biodone(bp);
171 	}
172 }
173 
174 /*
175  * mfs_mount
176  *
177  * Called when mounting local physical media
178  *
179  * PARAMETERS:
180  *		mountroot
181  *			mp	mount point structure
182  *			path	NULL (flag for root mount!!!)
183  *			data	<unused>
184  *			ndp	<unused>
185  *			p	process (user credentials check [statfs])
186  *
187  *		mount
188  *			mp	mount point structure
189  *			path	path to mount point
190  *			data	pointer to argument struct in user space
191  *			ndp	mount point namei() return (used for
192  *				credentials on reload), reused to look
193  *				up block device.
194  *			p	process (user credentials check)
195  *
196  * RETURNS:	0	Success
197  *		!0	error number (errno.h)
198  *
199  * LOCK STATE:
200  *
201  *		ENTRY
202  *			mount point is locked
203  *		EXIT
204  *			mount point is locked
205  *
206  * NOTES:
207  *		A NULL path can be used for a flag since the mount
208  *		system call will fail with EFAULT in copyinstr in
209  *		namei() if it is a genuine NULL from the user.
210  */
211 /* ARGSUSED */
212 static int
213 mfs_mount(struct mount *mp, char *path, caddr_t data, struct thread *td)
214 {
215 	struct vnode *devvp;
216 	struct mfs_args args;
217 	struct ufsmount *ump;
218 	struct fs *fs;
219 	struct mfsnode *mfsp;
220 	size_t size;
221 	int flags, err;
222 	int minnum;
223 	dev_t dev;
224 
225 	/*
226 	 * Use NULL path to flag a root mount
227 	 */
228 	if( path == NULL) {
229 		/*
230 		 ***
231 		 * Mounting root file system
232 		 ***
233 		 */
234 
235 		/* you lose */
236 		panic("mfs_mount: mount MFS as root: not configured!");
237 	}
238 
239 	/*
240 	 ***
241 	 * Mounting non-root file system or updating a file system
242 	 ***
243 	 */
244 
245 	/* copy in user arguments*/
246 	if ((err = copyin(data, (caddr_t)&args, sizeof (struct mfs_args))) != 0)
247 		goto error_1;
248 
249 	/*
250 	 * If updating, check whether changing from read-only to
251 	 * read/write; if there is no device name, that's all we do.
252 	 */
253 	if (mp->mnt_flag & MNT_UPDATE) {
254 		/*
255 		 ********************
256 		 * UPDATE
257 		 ********************
258 		 */
259 		ump = VFSTOUFS(mp);
260 		fs = ump->um_fs;
261 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
262 			flags = WRITECLOSE;
263 			if (mp->mnt_flag & MNT_FORCE)
264 				flags |= FORCECLOSE;
265 			err = ffs_flushfiles(mp, flags, td);
266 			if (err)
267 				goto error_1;
268 		}
269 		if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR))
270 			fs->fs_ronly = 0;
271 		/* if not updating name...*/
272 		if (args.fspec == 0) {
273 			/*
274 			 * Process export requests.  Jumping to "success"
275 			 * will return the vfs_export() error code.
276 			 */
277 			err = vfs_export(mp, &ump->um_export, &args.export);
278 			goto success;
279 		}
280 
281 		/* XXX MFS does not support name updating*/
282 		goto success;
283 	}
284 	/*
285 	 * Do the MALLOC before the getnewvnode since doing so afterward
286 	 * might cause a bogus v_data pointer to get dereferenced
287 	 * elsewhere if MALLOC should block.
288 	 */
289 	MALLOC(mfsp, struct mfsnode *, sizeof *mfsp, M_MFSNODE, M_WAITOK);
290 
291 	err = getspecialvnode(VT_MFS, NULL, &mfs_vnode_vops, &devvp, 0, 0);
292 	if (err) {
293 		FREE(mfsp, M_MFSNODE);
294 		goto error_1;
295 	}
296 
297 	minnum = (curproc->p_pid & 0xFF) |
298 		((curproc->p_pid & ~0xFF) << 8);
299 
300 	devvp->v_type = VCHR;
301 	dev = make_dev(&mfs_cdevsw, minnum, UID_ROOT, GID_WHEEL, 0600,
302 			"MFS%d", minnum >> 16);
303 	/* It is not clear that these will get initialized otherwise */
304 	dev->si_bsize_phys = DEV_BSIZE;
305 	dev->si_iosize_max = DFLTPHYS;
306 	dev->si_drv1 = mfsp;
307 	addaliasu(devvp, makeudev(MFS_CDEV_MAJOR, minnum));
308 	devvp->v_data = mfsp;
309 	mfsp->mfs_baseoff = args.base;
310 	mfsp->mfs_size = args.size;
311 	mfsp->mfs_vnode = devvp;
312 	mfsp->mfs_dev = reference_dev(dev);
313 	mfsp->mfs_td = td;
314 	mfsp->mfs_active = 1;
315 	bufq_init(&mfsp->buf_queue);
316 
317 	/* Save "mounted from" info for mount point (NULL pad)*/
318 	copyinstr(	args.fspec,			/* device name*/
319 			mp->mnt_stat.f_mntfromname,	/* save area*/
320 			MNAMELEN - 1,			/* max size*/
321 			&size);				/* real size*/
322 	bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
323 
324 	vx_unlock(devvp);
325 	if ((err = ffs_mountfs(devvp, mp, td, M_MFSNODE)) != 0) {
326 		mfsp->mfs_active = 0;
327 		goto error_2;
328 	}
329 
330 	/*
331 	 * Initialize FS stat information in mount struct; uses
332 	 * mp->mnt_stat.f_mntfromname.
333 	 *
334 	 * This code is common to root and non-root mounts
335 	 */
336 	(void) VFS_STATFS(mp, &mp->mnt_stat, td);
337 
338 	goto success;
339 
340 error_2:	/* error with devvp held*/
341 
342 	/* release devvp before failing*/
343 	vrele(devvp);
344 
345 error_1:	/* no state to back out*/
346 
347 success:
348 	return( err);
349 }
350 
351 /*
352  * Used to grab the process and keep it in the kernel to service
353  * memory filesystem I/O requests.
354  *
355  * Loop servicing I/O requests.
356  * Copy the requested data into or out of the memory filesystem
357  * address space.
358  */
359 /* ARGSUSED */
360 static int
361 mfs_start(struct mount *mp, int flags, struct thread *td)
362 {
363 	struct vnode *vp = VFSTOUFS(mp)->um_devvp;
364 	struct mfsnode *mfsp = VTOMFS(vp);
365 	struct buf *bp;
366 	int gotsig = 0, sig;
367 
368 	/*
369 	 * We must prevent the system from trying to swap
370 	 * out or kill ( when swap space is low, see vm/pageout.c ) the
371 	 * process.  A deadlock can occur if the process is swapped out,
372 	 * and the system can loop trying to kill the unkillable ( while
373 	 * references exist ) MFS process when swap space is low.
374 	 */
375 	KKASSERT(curproc);
376 	PHOLD(curproc);
377 
378 	while (mfsp->mfs_active) {
379 
380 		crit_enter();
381 
382 		while ((bp = bufq_first(&mfsp->buf_queue)) != NULL) {
383 			bufq_remove(&mfsp->buf_queue, bp);
384 			crit_exit();
385 			mfs_doio(bp, mfsp);
386 			wakeup((caddr_t)bp);
387 			crit_enter();
388 		}
389 
390 		crit_exit();
391 
392 		/*
393 		 * If a non-ignored signal is received, try to unmount.
394 		 * If that fails, clear the signal (it has been "processed"),
395 		 * otherwise we will loop here, as tsleep will always return
396 		 * EINTR/ERESTART.
397 		 */
398 		/*
399 		 * Note that dounmount() may fail if work was queued after
400 		 * we slept. We have to jump hoops here to make sure that we
401 		 * process any buffers after the sleep, before we dounmount()
402 		 */
403 		if (gotsig) {
404 			gotsig = 0;
405 			if (dounmount(mp, 0, td) != 0) {
406 				KKASSERT(td->td_proc);
407 				sig = CURSIG(td->td_proc);
408 				if (sig)
409 					SIGDELSET(td->td_proc->p_siglist, sig);
410 			}
411 		}
412 		else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0))
413 			gotsig++;	/* try to unmount in next pass */
414 	}
415 	PRELE(curproc);
416 	v_release_rdev(vp);	/* hack because we do not implement CLOSE */
417 	/* XXX destroy/release devvp */
418 	return (0);
419 }
420 
421 /*
422  * Get file system statistics.
423  */
424 static int
425 mfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
426 {
427 	int error;
428 
429 	error = ffs_statfs(mp, sbp, td);
430 	sbp->f_type = mp->mnt_vfc->vfc_typenum;
431 	return (error);
432 }
433 
434 /*
435  * Memory based filesystem initialization.
436  */
437 static int
438 mfs_init(struct vfsconf *vfsp)
439 {
440 	cdevsw_add(&mfs_cdevsw, 0, 0);
441 	return (0);
442 }
443