xref: /dflybsd-src/sys/vfs/mfs/mfs_vfsops.c (revision 6d49aa6ffaff1e5a2ff3abe70c453cc8b47adb73)
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 1.38 2007/02/25 23:17:13 corecode Exp $
36  */
37 
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/conf.h>
42 #include <sys/device.h>
43 #include <sys/kernel.h>
44 #include <sys/proc.h>
45 #include <sys/buf.h>
46 #include <sys/mount.h>
47 #include <sys/signalvar.h>
48 #include <sys/signal2.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 extern struct vop_ops *mfs_vnode_vops_p;
76 
77 static int	mfs_mount (struct mount *mp,
78 			char *path, caddr_t data, struct ucred *td);
79 static int	mfs_start (struct mount *mp, int flags);
80 static int	mfs_statfs (struct mount *mp, struct statfs *sbp,
81 			struct ucred *cred);
82 static int	mfs_init (struct vfsconf *);
83 
84 d_open_t	mfsopen;
85 d_close_t	mfsclose;
86 d_strategy_t	mfsstrategy;
87 
88 #define MFS_CDEV_MAJOR	253
89 
90 static struct dev_ops mfs_ops = {
91 	{ "MFS", MFS_CDEV_MAJOR, D_DISK },
92 	.d_open =	mfsopen,
93 	.d_close =	mfsclose,
94 	.d_read =	physread,
95 	.d_write =	physwrite,
96 	.d_strategy =	mfsstrategy,
97 };
98 
99 /*
100  * mfs vfs operations.
101  */
102 static struct vfsops mfs_vfsops = {
103 	.vfs_mount =     	mfs_mount,
104 	.vfs_start =    	mfs_start,
105 	.vfs_unmount =   	ffs_unmount,
106 	.vfs_root =     	ufs_root,
107 	.vfs_quotactl =  	ufs_quotactl,
108 	.vfs_statfs =   	mfs_statfs,
109 	.vfs_sync =     	ffs_sync,
110 	.vfs_vget =      	ffs_vget,
111 	.vfs_fhtovp =   	ffs_fhtovp,
112 	.vfs_checkexp =  	ufs_check_export,
113 	.vfs_vptofh =   	ffs_vptofh,
114 	.vfs_init =     	mfs_init
115 };
116 
117 VFS_SET(mfs_vfsops, mfs, 0);
118 
119 /*
120  * We allow the underlying MFS block device to be opened and read.
121  */
122 int
123 mfsopen(struct dev_open_args *ap)
124 {
125 	cdev_t dev = ap->a_head.a_dev;
126 
127 	if (ap->a_oflags & FWRITE)
128 		return(EROFS);
129 	if (dev->si_drv1)
130 		return(0);
131 	return(ENXIO);
132 }
133 
134 int
135 mfsclose(struct dev_close_args *ap)
136 {
137 	return(0);
138 }
139 
140 int
141 mfsstrategy(struct dev_strategy_args *ap)
142 {
143 	cdev_t dev = ap->a_head.a_dev;
144 	struct bio *bio = ap->a_bio;
145 	struct buf *bp = bio->bio_buf;
146 	off_t boff = bio->bio_offset;
147 	off_t eoff = boff + bp->b_bcount;
148 	struct mfsnode *mfsp;
149 
150 	if ((mfsp = dev->si_drv1) == NULL) {
151 		bp->b_error = ENXIO;
152 		goto error;
153 	}
154 	if (boff < 0)
155 		goto bad;
156 	if (eoff > mfsp->mfs_size) {
157 		if (boff > mfsp->mfs_size || (bp->b_flags & B_BNOCLIP))
158 			goto bad;
159 		/*
160 		 * Return EOF by completing the I/O with 0 bytes transfered.
161 		 * Set B_INVAL to indicate that any data in the buffer is not
162 		 * valid.
163 		 */
164 		if (boff == mfsp->mfs_size) {
165 			bp->b_resid = bp->b_bcount;
166 			bp->b_flags |= B_INVAL;
167 			goto done;
168 		}
169 		bp->b_bcount = mfsp->mfs_size - boff;
170 	}
171 
172 	/*
173 	 * Initiate I/O
174 	 */
175 	bioq_insert_tail(&mfsp->bio_queue, bio);
176 	wakeup((caddr_t)mfsp);
177 	return(0);
178 
179 	/*
180 	 * Failure conditions on bio
181 	 */
182 bad:
183 	bp->b_error = EINVAL;
184 error:
185 	bp->b_flags |= B_ERROR | B_INVAL;
186 done:
187 	biodone(bio);
188 	return(0);
189 }
190 
191 /*
192  * mfs_mount
193  *
194  * Called when mounting local physical media
195  *
196  * PARAMETERS:
197  *		mountroot
198  *			mp	mount point structure
199  *			path	NULL (flag for root mount!!!)
200  *			data	<unused>
201  *			ndp	<unused>
202  *			p	process (user credentials check [statfs])
203  *
204  *		mount
205  *			mp	mount point structure
206  *			path	path to mount point
207  *			data	pointer to argument struct in user space
208  *			ndp	mount point namei() return (used for
209  *				credentials on reload), reused to look
210  *				up block device.
211  *			p	process (user credentials check)
212  *
213  * RETURNS:	0	Success
214  *		!0	error number (errno.h)
215  *
216  * LOCK STATE:
217  *
218  *		ENTRY
219  *			mount point is locked
220  *		EXIT
221  *			mount point is locked
222  *
223  * NOTES:
224  *		A NULL path can be used for a flag since the mount
225  *		system call will fail with EFAULT in copyinstr in
226  *		namei() if it is a genuine NULL from the user.
227  */
228 /* ARGSUSED */
229 static int
230 mfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
231 {
232 	struct vnode *devvp;
233 	struct mfs_args args;
234 	struct ufsmount *ump;
235 	struct fs *fs;
236 	struct mfsnode *mfsp;
237 	size_t size;
238 	int flags, err;
239 	int minnum;
240 	cdev_t dev;
241 
242 	/*
243 	 * Use NULL path to flag a root mount
244 	 */
245 	if( path == NULL) {
246 		/*
247 		 ***
248 		 * Mounting root file system
249 		 ***
250 		 */
251 
252 		/* you lose */
253 		panic("mfs_mount: mount MFS as root: not configured!");
254 	}
255 
256 	/*
257 	 ***
258 	 * Mounting non-root file system or updating a file system
259 	 ***
260 	 */
261 
262 	/* copy in user arguments*/
263 	if ((err = copyin(data, (caddr_t)&args, sizeof (struct mfs_args))) != 0)
264 		goto error_1;
265 
266 	/*
267 	 * If updating, check whether changing from read-only to
268 	 * read/write; if there is no device name, that's all we do.
269 	 */
270 	if (mp->mnt_flag & MNT_UPDATE) {
271 		/*
272 		 ********************
273 		 * UPDATE
274 		 ********************
275 		 */
276 		ump = VFSTOUFS(mp);
277 		fs = ump->um_fs;
278 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
279 			flags = WRITECLOSE;
280 			if (mp->mnt_flag & MNT_FORCE)
281 				flags |= FORCECLOSE;
282 			err = ffs_flushfiles(mp, flags);
283 			if (err)
284 				goto error_1;
285 		}
286 		if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
287 			/* XXX reopen the device vnode read-write */
288 			fs->fs_ronly = 0;
289 		}
290 		/* if not updating name...*/
291 		if (args.fspec == 0) {
292 			/*
293 			 * Process export requests.  Jumping to "success"
294 			 * will return the vfs_export() error code.
295 			 */
296 			err = vfs_export(mp, &ump->um_export, &args.export);
297 			goto success;
298 		}
299 
300 		/* XXX MFS does not support name updating*/
301 		goto success;
302 	}
303 	/*
304 	 * Do the MALLOC before the getnewvnode since doing so afterward
305 	 * might cause a bogus v_data pointer to get dereferenced
306 	 * elsewhere if MALLOC should block.
307 	 */
308 	MALLOC(mfsp, struct mfsnode *, sizeof *mfsp, M_MFSNODE, M_WAITOK);
309 
310 	err = getspecialvnode(VT_MFS, NULL, &mfs_vnode_vops_p, &devvp, 0, 0);
311 	if (err) {
312 		FREE(mfsp, M_MFSNODE);
313 		goto error_1;
314 	}
315 
316 	minnum = (curproc->p_pid & 0xFF) |
317 		((curproc->p_pid & ~0xFF) << 8);
318 
319 	devvp->v_type = VCHR;
320 	dev = make_dev(&mfs_ops, minnum, UID_ROOT, GID_WHEEL, 0600,
321 			"MFS%d", minnum >> 16);
322 	/* It is not clear that these will get initialized otherwise */
323 	dev->si_bsize_phys = DEV_BSIZE;
324 	dev->si_iosize_max = DFLTPHYS;
325 	dev->si_drv1 = mfsp;
326 	addaliasu(devvp, makeudev(MFS_CDEV_MAJOR, minnum));
327 	devvp->v_data = mfsp;
328 	mfsp->mfs_baseoff = args.base;
329 	mfsp->mfs_size = args.size;
330 	mfsp->mfs_vnode = devvp;
331 	mfsp->mfs_dev = reference_dev(dev);
332 	mfsp->mfs_td = curthread;
333 	mfsp->mfs_active = 1;
334 	bioq_init(&mfsp->bio_queue);
335 
336 	/*
337 	 * Our 'block' device must be backed by a VM object.  Theoretically
338 	 * we could use the anonymous memory VM object supplied by userland,
339 	 * but it would be somewhat of a complex task to deal with it
340 	 * that way since it would result in I/O requests which supply
341 	 * the VM pages from our own object.
342 	 *
343 	 * vnode_pager_alloc() is typically called when a VM object is
344 	 * being referenced externally.  We have to undo the refs for
345 	 * the self reference between vnode and object.
346 	 */
347 	vnode_pager_alloc(devvp, args.size, 0, 0);
348 	--devvp->v_usecount;
349 	--devvp->v_object->ref_count;
350 
351 	/* Save "mounted from" info for mount point (NULL pad)*/
352 	copyinstr(	args.fspec,			/* device name*/
353 			mp->mnt_stat.f_mntfromname,	/* save area*/
354 			MNAMELEN - 1,			/* max size*/
355 			&size);				/* real size*/
356 	bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
357 
358 	vx_unlock(devvp);
359 	if ((err = ffs_mountfs(devvp, mp, M_MFSNODE)) != 0) {
360 		mfsp->mfs_active = 0;
361 		goto error_2;
362 	}
363 
364 	/*
365 	 * Initialize FS stat information in mount struct; uses
366 	 * mp->mnt_stat.f_mntfromname.
367 	 *
368 	 * This code is common to root and non-root mounts
369 	 */
370 	VFS_STATFS(mp, &mp->mnt_stat, cred);
371 
372 	goto success;
373 
374 error_2:	/* error with devvp held*/
375 
376 	/* release devvp before failing*/
377 	vrele(devvp);
378 
379 error_1:	/* no state to back out*/
380 
381 success:
382 	return( err);
383 }
384 
385 /*
386  * Used to grab the process and keep it in the kernel to service
387  * memory filesystem I/O requests.
388  *
389  * Loop servicing I/O requests.
390  * Copy the requested data into or out of the memory filesystem
391  * address space.
392  */
393 /* ARGSUSED */
394 static int
395 mfs_start(struct mount *mp, int flags)
396 {
397 	struct vnode *vp = VFSTOUFS(mp)->um_devvp;
398 	struct mfsnode *mfsp = VTOMFS(vp);
399 	struct bio *bio;
400 	struct buf *bp;
401 	int gotsig = 0, sig;
402 	thread_t td = curthread;
403 
404 	/*
405 	 * We must prevent the system from trying to swap
406 	 * out or kill ( when swap space is low, see vm/pageout.c ) the
407 	 * process.  A deadlock can occur if the process is swapped out,
408 	 * and the system can loop trying to kill the unkillable ( while
409 	 * references exist ) MFS process when swap space is low.
410 	 */
411 	KKASSERT(curproc);
412 	PHOLD(curproc);
413 
414 	mfsp->mfs_td = td;
415 
416 	while (mfsp->mfs_active) {
417 		crit_enter();
418 
419 		while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) {
420 			bioq_remove(&mfsp->bio_queue, bio);
421 			crit_exit();
422 			bp = bio->bio_buf;
423 			mfs_doio(bio, mfsp);
424 			wakeup(bp);
425 			crit_enter();
426 		}
427 
428 		crit_exit();
429 
430 		/*
431 		 * If a non-ignored signal is received, try to unmount.
432 		 * If that fails, clear the signal (it has been "processed"),
433 		 * otherwise we will loop here, as tsleep will always return
434 		 * EINTR/ERESTART.
435 		 */
436 		/*
437 		 * Note that dounmount() may fail if work was queued after
438 		 * we slept. We have to jump hoops here to make sure that we
439 		 * process any buffers after the sleep, before we dounmount()
440 		 */
441 		if (gotsig) {
442 			gotsig = 0;
443 			if (dounmount(mp, 0) != 0) {
444 				KKASSERT(td->td_proc);
445 				sig = CURSIG(td->td_lwp);
446 				if (sig)
447 					lwp_delsig(td->td_lwp, sig);
448 			}
449 		}
450 		else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0))
451 			gotsig++;	/* try to unmount in next pass */
452 	}
453 	PRELE(curproc);
454 	v_release_rdev(vp);	/* hack because we do not implement CLOSE */
455 	/* XXX destroy/release devvp */
456 	return (0);
457 }
458 
459 /*
460  * Get file system statistics.
461  */
462 static int
463 mfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
464 {
465 	int error;
466 
467 	error = ffs_statfs(mp, sbp, cred);
468 	sbp->f_type = mp->mnt_vfc->vfc_typenum;
469 	return (error);
470 }
471 
472 /*
473  * Memory based filesystem initialization.
474  */
475 static int
476 mfs_init(struct vfsconf *vfsp)
477 {
478 	dev_ops_add(&mfs_ops, 0, 0);
479 	return (0);
480 }
481