2 * Copyright (c) 1989, 1990, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)mfs_vfsops.c 8.11 (Berkeley) 6/19/95
30 * $FreeBSD: src/sys/ufs/mfs/mfs_vfsops.c,v 1.81.2.3 2001/07/04 17:35:21 tegge Exp $
34 #include <sys/param.h>
35 #include <sys/systm.h>
37 #include <sys/device.h>
38 #include <sys/kernel.h>
41 #include <sys/mount.h>
42 #include <sys/signalvar.h>
43 #include <sys/signal2.h>
44 #include <sys/vnode.h>
45 #include <sys/malloc.h>
46 #include <sys/sysproto.h>
48 #include <sys/linker.h>
49 #include <sys/fcntl.h>
50 #include <sys/nlookup.h>
51 #include <sys/devfs.h>
54 #include <vm/vm_object.h>
55 #include <vm/vm_page.h>
56 #include <vm/vm_pager.h>
57 #include <vm/vnode_pager.h>
58 #include <vm/vm_extern.h>
61 #include <sys/thread2.h>
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>
71 #include "mfs_extern.h"
73 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
75 static int mfs_mount (struct mount *mp,
76 char *path, caddr_t data, struct ucred *td);
77 static int mfs_start (struct mount *mp, int flags);
78 static int mfs_statfs (struct mount *mp, struct statfs *sbp,
80 static int mfs_init (struct vfsconf *);
81 static void mfs_doio(struct bio *bio, struct mfsnode *mfsp);
85 d_strategy_t mfsstrategy;
87 static struct dev_ops mfs_ops = {
88 { "MFS", -1, D_DISK },
93 .d_strategy = mfsstrategy,
99 static struct vfsops mfs_vfsops = {
100 .vfs_mount = mfs_mount,
101 .vfs_start = mfs_start,
102 .vfs_unmount = ffs_unmount,
103 .vfs_root = ufs_root,
104 .vfs_quotactl = ufs_quotactl,
105 .vfs_statfs = mfs_statfs,
106 .vfs_sync = ffs_sync,
107 .vfs_vget = ffs_vget,
108 .vfs_fhtovp = ffs_fhtovp,
109 .vfs_checkexp = ufs_check_export,
110 .vfs_vptofh = ffs_vptofh,
114 VFS_SET(mfs_vfsops, mfs, 0);
115 MODULE_VERSION(mfs, 1);
118 * We allow the underlying MFS block device to be opened and read.
121 mfsopen(struct dev_open_args *ap)
123 cdev_t dev = ap->a_head.a_dev;
126 if (ap->a_oflags & FWRITE)
135 mfsclose(struct dev_close_args *ap)
137 cdev_t dev = ap->a_head.a_dev;
138 struct mfsnode *mfsp;
140 if ((mfsp = dev->si_drv1) == NULL)
142 mfsp->mfs_active = 0;
143 wakeup((caddr_t)mfsp);
148 mfsstrategy(struct dev_strategy_args *ap)
150 cdev_t dev = ap->a_head.a_dev;
151 struct bio *bio = ap->a_bio;
152 struct buf *bp = bio->bio_buf;
153 off_t boff = bio->bio_offset;
154 off_t eoff = boff + bp->b_bcount;
155 struct mfsnode *mfsp;
157 if ((mfsp = dev->si_drv1) == NULL) {
163 if (eoff > mfsp->mfs_size) {
164 if (boff > mfsp->mfs_size || (bp->b_flags & B_BNOCLIP))
167 * Return EOF by completing the I/O with 0 bytes transfered.
168 * Set B_INVAL to indicate that any data in the buffer is not
171 if (boff == mfsp->mfs_size) {
172 bp->b_resid = bp->b_bcount;
173 bp->b_flags |= B_INVAL;
176 bp->b_bcount = mfsp->mfs_size - boff;
182 if (mfsp->mfs_td == curthread) {
185 bioq_insert_tail(&mfsp->bio_queue, bio);
186 wakeup((caddr_t)mfsp);
191 * Failure conditions on bio
194 bp->b_error = EINVAL;
196 bp->b_flags |= B_ERROR | B_INVAL;
205 * Called when mounting local physical media
209 * mp mount point structure
210 * path NULL (flag for root mount!!!)
213 * p process (user credentials check [statfs])
216 * mp mount point structure
217 * path path to mount point
218 * data pointer to argument struct in user space
219 * ndp mount point namei() return (used for
220 * credentials on reload), reused to look
222 * p process (user credentials check)
225 * !0 error number (errno.h)
230 * mount point is locked
232 * mount point is locked
235 * A NULL path can be used for a flag since the mount
236 * system call will fail with EFAULT in copyinstr in
237 * namei() if it is a genuine NULL from the user.
241 mfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
244 struct mfs_args args;
245 struct ufsmount *ump;
247 struct mfsnode *mfsp;
248 struct nlookupdata nd;
257 * Use NULL path to flag a root mount
262 * Mounting root file system
267 panic("mfs_mount: mount MFS as root: not configured!");
274 * Mounting non-root file system or updating a file system
278 /* copy in user arguments*/
279 error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args));
284 * If updating, check whether changing from read-only to
285 * read/write; if there is no device name, that's all we do.
287 if (mp->mnt_flag & MNT_UPDATE) {
295 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
297 if (mp->mnt_flag & MNT_FORCE)
299 error = ffs_flushfiles(mp, flags);
303 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
304 /* XXX reopen the device vnode read-write */
307 /* if not updating name...*/
308 if (args.fspec == 0) {
310 * Process export requests. Jumping to "success"
311 * will return the vfs_export() error code.
313 error = vfs_export(mp, &ump->um_export, &args.export);
317 /* XXX MFS does not support name updating*/
322 * Do the MALLOC before the make_dev since doing so afterward
323 * might cause a bogus v_data pointer to get dereferenced
324 * elsewhere if MALLOC should block.
326 mfsp = kmalloc(sizeof *mfsp, M_MFSNODE, M_WAITOK | M_ZERO);
328 minnum = (int)curproc->p_pid;
330 dev = make_dev(&mfs_ops, minnum, UID_ROOT, GID_WHEEL, 0600,
332 /* It is not clear that these will get initialized otherwise */
333 dev->si_bsize_phys = DEV_BSIZE;
334 dev->si_iosize_max = MAXPHYS;
336 mfsp->mfs_baseoff = args.base;
337 mfsp->mfs_size = args.size;
339 mfsp->mfs_td = curthread;
340 mfsp->mfs_active = 1;
341 bioq_init(&mfsp->bio_queue);
343 devfs_config(); /* sync devfs work */
344 ksnprintf(devname, sizeof(devname), "/dev/mfs%d", minnum);
345 nlookup_init(&nd, devname, UIO_SYSSPACE, 0);
347 error = nlookup(&nd);
349 devvp = nd.nl_nch.ncp->nc_vp;
352 error = vget(devvp, LK_SHARED);
361 * Our 'block' device must be backed by a VM object. Theoretically
362 * we could use the anonymous memory VM object supplied by userland,
363 * but it would be somewhat of a complex task to deal with it
364 * that way since it would result in I/O requests which supply
365 * the VM pages from our own object.
367 * vnode_pager_alloc() is typically called when a VM object is
368 * being referenced externally. We have to undo the refs for
369 * the self reference between vnode and object.
371 vnode_pager_setsize(devvp, args.size);
373 /* Save "mounted from" info for mount point (NULL pad)*/
374 copyinstr(args.fspec, /* device name*/
375 mp->mnt_stat.f_mntfromname, /* save area*/
376 MNAMELEN - 1, /* max size*/
377 &size); /* real size*/
378 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
379 /* vref is eaten by mount? */
381 error = ffs_mountfs(devvp, mp, M_MFSNODE);
383 mfsp->mfs_active = 0;
388 * Initialize FS stat information in mount struct; uses
389 * mp->mnt_stat.f_mntfromname.
391 * This code is common to root and non-root mounts
393 VFS_STATFS(mp, &mp->mnt_stat, cred);
396 * Mark VFS_START MPSAFE; this is to avoid accessing
397 * per-mount token after VFS_START exits
399 mp->mnt_kern_flag |= MNTK_ST_MPSAFE;
403 error_2: /* error with devvp held*/
406 error_1: /* no state to back out*/
409 destroy_dev(mfsp->mfs_dev);
410 mfsp->mfs_dev = NULL;
412 kfree(mfsp, M_MFSNODE);
420 * Used to grab the process and keep it in the kernel to service
421 * memory filesystem I/O requests.
423 * Loop servicing I/O requests.
424 * Copy the requested data into or out of the memory filesystem
429 mfs_start(struct mount *mp, int flags)
431 struct vnode *vp = VFSTOUFS(mp)->um_devvp;
432 struct mfsnode *mfsp = vp->v_rdev->si_drv1;
436 thread_t td = curthread;
439 * We must prevent the system from trying to swap
440 * out or kill ( when swap space is low, see vm/pageout.c ) the
441 * process. A deadlock can occur if the process is swapped out,
442 * and the system can loop trying to kill the unkillable ( while
443 * references exist ) MFS process when swap space is low.
450 while (mfsp->mfs_active) {
453 while ((bio = bioq_takefirst(&mfsp->bio_queue)) != NULL) {
464 * If a non-ignored signal is received, try to unmount.
465 * If that fails, clear the signal (it has been "processed"),
466 * otherwise we will loop here, as tsleep will always return
470 * Note that dounmount() may fail if work was queued after
471 * we slept. We have to jump hoops here to make sure that we
472 * process any buffers after the sleep, before we dounmount()
476 if (dounmount(mp, 0) != 0) {
477 KKASSERT(td->td_proc);
478 sig = CURSIG(td->td_lwp);
480 spin_lock(&td->td_lwp->lwp_spin);
481 lwp_delsig(td->td_lwp, sig);
482 spin_unlock(&td->td_lwp->lwp_spin);
486 else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0))
487 gotsig++; /* try to unmount in next pass */
491 destroy_dev(mfsp->mfs_dev);
492 mfsp->mfs_dev = NULL;
494 kfree(mfsp, M_MFSNODE);
499 * Get file system statistics.
502 mfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
506 error = ffs_statfs(mp, sbp, cred);
507 sbp->f_type = mp->mnt_vfc->vfc_typenum;
512 * Memory based filesystem initialization.
515 mfs_init(struct vfsconf *vfsp)
521 * Memory file system I/O.
523 * Trivial on the HP since buffer has already been mapping into KVA space.
525 * Read and Write are handled with a simple copyin and copyout.
527 * We also partially support VOP_FREEBLKS(). We can't implement
528 * completely -- for example, on fragments or inode metadata, but we can
529 * implement it for page-aligned requests.
532 mfs_doio(struct bio *bio, struct mfsnode *mfsp)
534 struct buf *bp = bio->bio_buf;
535 caddr_t base = mfsp->mfs_baseoff + bio->bio_offset;
539 case BUF_CMD_FREEBLKS:
541 * Implement FREEBLKS, which allows the filesystem to tell
542 * a block device when blocks are no longer needed (like when
543 * a file is deleted). We use the hook to MADV_FREE the VM.
544 * This makes an MFS filesystem work as well or better then
545 * a sun-style swap-mounted filesystem.
547 bytes = bp->b_bcount;
549 if ((vm_offset_t)base & PAGE_MASK) {
550 int n = PAGE_SIZE - ((vm_offset_t)base & PAGE_MASK);
555 struct madvise_args uap;
559 bzero(&uap, sizeof(uap));
562 uap.behav = MADV_FREE;
570 * Read data from our 'memory' disk
572 bp->b_error = copyin(base, bp->b_data, bp->b_bcount);
576 * Write data to our 'memory' disk
578 bp->b_error = copyout(bp->b_data, base, bp->b_bcount);
581 panic("mfs: bad b_cmd %d", bp->b_cmd);
584 bp->b_flags |= B_ERROR;