2 * Copyright (c) 1988 University of Utah.
3 * Copyright (c) 1990, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * the Systems Programming Group of the University of Utah Computer
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * from: Utah Hdr: vn.c 1.13 94/04/02
40 * from: @(#)vn.c 8.6 (Berkeley) 4/1/94
41 * $FreeBSD: src/sys/dev/vn/vn.c,v 1.105.2.4 2001/11/18 07:11:00 dillon Exp $
42 * $DragonFly: src/sys/dev/disk/vn/vn.c,v 1.33 2007/05/17 03:20:08 dillon Exp $
48 * Block/character interface to a vnode. Allows one to treat a file
49 * as a disk (e.g. build a filesystem in it, mount it, etc.).
51 * NOTE 1: There is a security issue involved with this driver.
52 * Once mounted all access to the contents of the "mapped" file via
53 * the special file is controlled by the permissions on the special
54 * file, the protection of the mapped file is ignored (effectively,
55 * by using root credentials in all transactions).
57 * NOTE 2: Doesn't interact with leases, should it?
60 #include <sys/param.h>
61 #include <sys/systm.h>
62 #include <sys/kernel.h>
64 #include <sys/nlookup.h>
66 #include <sys/malloc.h>
67 #include <sys/mount.h>
68 #include <sys/vnode.h>
69 #include <sys/fcntl.h>
71 #include <sys/disklabel.h>
72 #include <sys/diskslice.h>
75 #include <sys/module.h>
76 #include <sys/vnioctl.h>
79 #include <vm/vm_object.h>
80 #include <vm/vm_page.h>
81 #include <vm/vm_pager.h>
82 #include <vm/vm_pageout.h>
83 #include <vm/swap_pager.h>
84 #include <vm/vm_extern.h>
85 #include <vm/vm_zone.h>
87 static d_ioctl_t vnioctl;
88 static d_open_t vnopen;
89 static d_close_t vnclose;
90 static d_psize_t vnsize;
91 static d_strategy_t vnstrategy;
95 #define VN_BSIZE_BEST 8192
99 * D_DISK we want to look like a disk
100 * D_CANFREE We support BUF_CMD_FREEBLKS
103 static struct dev_ops vn_ops = {
104 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
108 .d_write = physwrite,
110 .d_strategy = vnstrategy,
116 int sc_flags; /* flags */
117 u_int64_t sc_size; /* size of vn, sc_secsize scale */
118 int sc_secsize; /* sector size */
119 struct diskslices *sc_slices; /* XXX fields from struct disk */
120 struct disk_info sc_info; /* XXX fields from struct disk */
121 struct vnode *sc_vp; /* vnode if not NULL */
122 vm_object_t sc_object; /* backing object if not NULL */
123 struct ucred *sc_cred; /* credentials */
124 int sc_maxactive; /* max # of active requests */
125 struct buf sc_tab; /* transfer queue */
126 u_long sc_options; /* options */
127 cdev_t sc_devlist; /* devices that refer to this unit */
128 SLIST_ENTRY(vn_softc) sc_list;
131 static SLIST_HEAD(, vn_softc) vn_list;
134 #define VNF_INITED 0x01
135 #define VNF_READONLY 0x02
137 static u_long vn_options;
139 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
140 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
142 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
143 static void vnclear (struct vn_softc *vn);
144 static int vn_modevent (module_t, int, void *);
145 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
146 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
149 vnclose(struct dev_close_args *ap)
151 cdev_t dev = ap->a_head.a_dev;
152 struct vn_softc *vn = dev->si_drv1;
155 if (vn->sc_slices != NULL)
156 dsclose(dev, ap->a_devtype, vn->sc_slices);
161 * Called only when si_drv1 is NULL. Locate the associated vn node and
162 * attach the device to it.
164 static struct vn_softc *
171 SLIST_FOREACH(vn, &vn_list, sc_list) {
172 if (vn->sc_unit == unit) {
174 dev->si_drv2 = vn->sc_devlist;
175 vn->sc_devlist = dev;
181 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
184 vn->sc_devlist = make_dev(&vn_ops, 0, UID_ROOT,
185 GID_OPERATOR, 0640, "vn%d", unit);
186 if (vn->sc_devlist->si_drv1 == NULL) {
187 reference_dev(vn->sc_devlist);
188 vn->sc_devlist->si_drv1 = vn;
189 vn->sc_devlist->si_drv2 = NULL;
191 if (vn->sc_devlist != dev) {
193 dev->si_drv2 = vn->sc_devlist;
194 vn->sc_devlist = dev;
197 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
203 vnopen(struct dev_open_args *ap)
205 cdev_t dev = ap->a_head.a_dev;
207 struct disk_info *info;
210 * Locate preexisting device
213 if ((vn = dev->si_drv1) == NULL)
217 * Update si_bsize fields for device. This data will be overriden by
218 * the slice/parition code for vn accesses through partitions, and
219 * used directly if you open the 'whole disk' device.
221 * si_bsize_best must be reinitialized in case VN has been
222 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
224 dev->si_bsize_phys = vn->sc_secsize;
225 dev->si_bsize_best = vn->sc_secsize;
226 if (dev->si_bsize_best < VN_BSIZE_BEST)
227 dev->si_bsize_best = VN_BSIZE_BEST;
229 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
233 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
234 devtoname(dev), ap->a_oflags, ap->a_devtype);
240 IFOPT(vn, VN_LABELS) {
241 if (vn->sc_flags & VNF_INITED) {
243 bzero(info, sizeof(*info));
244 info->d_media_blksize = vn->sc_secsize;
245 info->d_media_blocks = vn->sc_size;
247 * reserve mbr sector for backwards compatibility
248 * when no slices exist.
250 info->d_dsflags = DSO_COMPATMBR;
252 info->d_secpertrack = 32;
253 info->d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
254 info->d_secpercyl = info->d_secpertrack *
256 info->d_ncylinders = vn->sc_size / info->d_secpercyl;
258 return (dsopen(dev, ap->a_devtype, 0,
259 &vn->sc_slices, info));
261 if (dkslice(dev) != WHOLE_DISK_SLICE ||
262 dkpart(dev) != RAW_PART ||
263 ap->a_devtype != S_IFCHR) {
273 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
274 * for vnode-backed vn's, and the new vm_pager_strategy() call for
275 * vm_object-backed vn's.
277 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
280 vnstrategy(struct dev_strategy_args *ap)
282 cdev_t dev = ap->a_head.a_dev;
283 struct bio *bio = ap->a_bio;
291 if ((vn = dev->si_drv1) == NULL)
297 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
299 if ((vn->sc_flags & VNF_INITED) == 0) {
301 bp->b_flags |= B_ERROR;
306 bp->b_resid = bp->b_bcount;
308 IFOPT(vn, VN_LABELS) {
310 * The vnode device is using disk/slice label support.
312 * The dscheck() function is called for validating the
313 * slices that exist ON the vnode device itself, and
314 * translate the "slice-relative" block number, again.
315 * dscheck() will call biodone() and return NULL if
316 * we are at EOF or beyond the device size.
318 if (vn->sc_slices == NULL) {
320 } else if ((nbio = dscheck(dev, bio, vn->sc_slices)) == NULL) {
324 int64_t pbn; /* in sc_secsize chunks */
325 long sz; /* in sc_secsize chunks */
328 * Check for required alignment. Transfers must be a valid
329 * multiple of the sector size.
331 if (bp->b_bcount % vn->sc_secsize != 0 ||
332 bio->bio_offset % vn->sc_secsize != 0) {
336 pbn = bio->bio_offset / vn->sc_secsize;
337 sz = howmany(bp->b_bcount, vn->sc_secsize);
340 * Check for an illegal pbn or EOF truncation
344 if (pbn + sz > vn->sc_size) {
345 if (pbn > vn->sc_size || (bp->b_flags & B_BNOCLIP))
347 if (pbn == vn->sc_size) {
348 bp->b_resid = bp->b_bcount;
349 bp->b_flags |= B_INVAL;
352 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
354 nbio = push_bio(bio);
355 nbio->bio_offset = pbn * vn->sc_secsize;
359 * Use the translated nbio from this point on
361 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
363 * Freeblks is not handled for vnode-backed elements yet.
366 /* operation complete */
367 } else if (vn->sc_vp) {
371 * If an error occurs, we set B_ERROR but we do not set
372 * B_INVAL because (for a write anyway), the buffer is
378 bzero(&auio, sizeof(auio));
380 aiov.iov_base = bp->b_data;
381 aiov.iov_len = bp->b_bcount;
382 auio.uio_iov = &aiov;
384 auio.uio_offset = nbio->bio_offset;
385 auio.uio_segflg = UIO_SYSSPACE;
386 if (bp->b_cmd == BUF_CMD_READ)
387 auio.uio_rw = UIO_READ;
389 auio.uio_rw = UIO_WRITE;
390 auio.uio_resid = bp->b_bcount;
391 auio.uio_td = curthread;
392 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
393 if (bp->b_cmd == BUF_CMD_READ)
394 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
396 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
397 vn_unlock(vn->sc_vp);
398 bp->b_resid = auio.uio_resid;
401 bp->b_flags |= B_ERROR;
403 /* operation complete */
404 } else if (vn->sc_object) {
406 * OBJT_SWAP I/O (handles read, write, freebuf)
408 * We have nothing to do if freeing blocks on a reserved
409 * swap area, othrewise execute the op.
411 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
413 /* operation complete */
415 vm_pager_strategy(vn->sc_object, nbio);
420 bp->b_resid = bp->b_bcount;
421 bp->b_flags |= B_ERROR | B_INVAL;
422 bp->b_error = EINVAL;
423 /* operation complete */
429 * Shortcuts / check failures on the original bio (not nbio).
432 bp->b_error = EINVAL;
433 bp->b_flags |= B_ERROR | B_INVAL;
441 vnioctl(struct dev_ioctl_args *ap)
443 cdev_t dev = ap->a_head.a_dev;
445 struct vn_ioctl *vio;
450 IFOPT(vn,VN_FOLLOW) {
451 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
452 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
466 IFOPT(vn,VN_LABELS) {
467 if (vn->sc_slices != NULL) {
468 error = dsioctl(dev, ap->a_cmd, ap->a_data,
470 &vn->sc_slices, &vn->sc_info);
471 if (error != ENOIOCTL)
474 if (dkslice(dev) != WHOLE_DISK_SLICE ||
475 dkpart(dev) != RAW_PART)
481 error = suser_cred(ap->a_cred, 0);
485 vio = (struct vn_ioctl *)ap->a_data;
486 f = (u_long*)ap->a_data;
490 if (vn->sc_flags & VNF_INITED)
493 if (vio->vn_file == NULL)
494 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
496 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
500 if ((vn->sc_flags & VNF_INITED) == 0)
503 * XXX handle i/o in progress. Return EBUSY, or wait, or
505 * XXX handle multiple opens of the device. Return EBUSY,
506 * or revoke the fd's.
507 * How are these problems handled for removable and failing
508 * hardware devices? (Hint: They are not)
512 kprintf("vnioctl: CLRed\n");
526 vn->sc_options |= *f;
531 vn->sc_options &= ~(*f);
545 * Attach a file to a VN partition. Return the size in the vn_size
550 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
551 int flag, struct ucred *cred)
554 struct nlookupdata nd;
558 flags = FREAD|FWRITE;
559 error = nlookup_init(&nd, vio->vn_file,
560 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
563 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
564 if (error != EACCES && error != EPERM && error != EROFS)
568 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
571 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
575 if (vp->v_type != VREG ||
576 (error = VOP_GETATTR(vp, &vattr))) {
582 vn->sc_secsize = DEV_BSIZE;
584 nd.nl_open_vp = NULL;
587 * If the size is specified, override the file attributes. Note that
588 * the vn_size argument is in PAGE_SIZE sized blocks.
591 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
593 vn->sc_size = vattr.va_size / vn->sc_secsize;
594 error = vnsetcred(vn, cred);
600 vn->sc_flags |= VNF_INITED;
602 vn->sc_flags |= VNF_READONLY;
603 IFOPT(vn, VN_LABELS) {
605 * Reopen so that `ds' knows which devices are open.
606 * If this is the first VNIOCSET, then we've
607 * guaranteed that the device is the cdev and that
608 * no other slices or labels are open. Otherwise,
609 * we rely on VNIOCCLR not being abused.
611 error = dev_dopen(dev, flag, S_IFCHR, cred);
616 kprintf("vnioctl: SET vp %p size %llx blks\n",
617 vn->sc_vp, vn->sc_size);
626 * Attach swap backing store to a VN partition of the size specified
631 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
632 int flag, struct ucred *cred)
637 * Range check. Disallow negative sizes or any size less then the
638 * size of a page. Then round to a page.
641 if (vio->vn_size <= 0)
645 * Allocate an OBJT_SWAP object.
647 * sc_secsize is PAGE_SIZE'd
649 * vio->vn_size is in PAGE_SIZE'd chunks.
650 * sc_size must be in PAGE_SIZE'd chunks.
651 * Note the truncation.
654 vn->sc_secsize = PAGE_SIZE;
655 vn->sc_size = vio->vn_size;
656 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
657 vn->sc_secsize * (off_t)vio->vn_size,
659 IFOPT(vn, VN_RESERVE) {
660 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
661 vm_pager_deallocate(vn->sc_object);
662 vn->sc_object = NULL;
666 vn->sc_flags |= VNF_INITED;
668 error = vnsetcred(vn, cred);
670 IFOPT(vn, VN_LABELS) {
672 * Reopen so that `ds' knows which devices are open.
673 * If this is the first VNIOCSET, then we've
674 * guaranteed that the device is the cdev and that
675 * no other slices or labels are open. Otherwise,
676 * we rely on VNIOCCLR not being abused.
678 error = dev_dopen(dev, flag, S_IFCHR, cred);
682 IFOPT(vn, VN_FOLLOW) {
683 kprintf("vnioctl: SET vp %p size %llx\n",
684 vn->sc_vp, vn->sc_size);
693 * Duplicate the current processes' credentials. Since we are called only
694 * as the result of a SET ioctl and only root can do that, any future access
695 * to this "disk" is essentially as root. Note that credentials may change
696 * if some other uid can write directly to the mapped file (NFS).
699 vnsetcred(struct vn_softc *vn, struct ucred *cred)
705 * Set credits in our softc
710 vn->sc_cred = crdup(cred);
713 * Horrible kludge to establish credentials for NFS XXX.
720 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
721 bzero(&auio, sizeof(auio));
723 aiov.iov_base = tmpbuf;
724 aiov.iov_len = vn->sc_secsize;
725 auio.uio_iov = &aiov;
728 auio.uio_rw = UIO_READ;
729 auio.uio_segflg = UIO_SYSSPACE;
730 auio.uio_resid = aiov.iov_len;
731 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
732 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
733 vn_unlock(vn->sc_vp);
734 kfree(tmpbuf, M_TEMP);
740 vnclear(struct vn_softc *vn)
743 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
744 if (vn->sc_slices != NULL)
745 dsgone(&vn->sc_slices);
746 vn->sc_flags &= ~VNF_INITED;
747 if (vn->sc_vp != NULL) {
749 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
752 vn->sc_flags &= ~VNF_READONLY;
757 if (vn->sc_object != NULL) {
758 vm_pager_deallocate(vn->sc_object);
759 vn->sc_object = NULL;
765 vnsize(struct dev_psize_args *ap)
767 cdev_t dev = ap->a_head.a_dev;
773 if ((vn->sc_flags & VNF_INITED) == 0)
775 ap->a_result = (int64_t)vn->sc_size;
780 vn_modevent(module_t mod, int type, void *data)
787 dev_ops_add(&vn_ops, 0, 0);
792 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
793 SLIST_REMOVE_HEAD(&vn_list, sc_list);
794 if (vn->sc_flags & VNF_INITED)
796 /* Cleanup all cdev_t's that refer to this unit */
797 while ((dev = vn->sc_devlist) != NULL) {
798 vn->sc_devlist = dev->si_drv2;
799 dev->si_drv1 = dev->si_drv2 = NULL;
804 dev_ops_remove(&vn_ops, 0, 0);
812 DEV_MODULE(vn, vn_modevent, 0);