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 1.28 2006/09/10 01:26:34 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>
74 #include <sys/module.h>
75 #include <sys/vnioctl.h>
78 #include <vm/vm_object.h>
79 #include <vm/vm_page.h>
80 #include <vm/vm_pager.h>
81 #include <vm/vm_pageout.h>
82 #include <vm/swap_pager.h>
83 #include <vm/vm_extern.h>
84 #include <vm/vm_zone.h>
86 static d_ioctl_t vnioctl;
87 static d_open_t vnopen;
88 static d_close_t vnclose;
89 static d_psize_t vnsize;
90 static d_strategy_t vnstrategy;
94 #define VN_BSIZE_BEST 8192
98 * D_DISK we want to look like a disk
99 * D_CANFREE We support BUF_CMD_FREEBLKS
102 static struct dev_ops vn_ops = {
103 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
107 .d_write = physwrite,
109 .d_strategy = vnstrategy,
115 int sc_flags; /* flags */
116 int sc_size; /* size of vn, sc_secsize scale */
117 int sc_secsize; /* sector size */
118 struct diskslices *sc_slices;
119 struct vnode *sc_vp; /* vnode if not NULL */
120 vm_object_t sc_object; /* backing object if not NULL */
121 struct ucred *sc_cred; /* credentials */
122 int sc_maxactive; /* max # of active requests */
123 struct buf sc_tab; /* transfer queue */
124 u_long sc_options; /* options */
125 cdev_t sc_devlist; /* devices that refer to this unit */
126 SLIST_ENTRY(vn_softc) sc_list;
129 static SLIST_HEAD(, vn_softc) vn_list;
132 #define VNF_INITED 0x01
133 #define VNF_READONLY 0x02
135 static u_long vn_options;
137 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
138 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
140 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
141 static void vnclear (struct vn_softc *vn);
142 static int vn_modevent (module_t, int, void *);
143 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
144 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
147 vnclose(struct dev_close_args *ap)
149 cdev_t dev = ap->a_head.a_dev;
150 struct vn_softc *vn = dev->si_drv1;
153 if (vn->sc_slices != NULL)
154 dsclose(dev, ap->a_devtype, vn->sc_slices);
159 * Called only when si_drv1 is NULL. Locate the associated vn node and
160 * attach the device to it.
162 static struct vn_softc *
169 SLIST_FOREACH(vn, &vn_list, sc_list) {
170 if (vn->sc_unit == unit) {
172 dev->si_drv2 = vn->sc_devlist;
173 vn->sc_devlist = dev;
179 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
182 vn->sc_devlist = make_dev(&vn_ops, 0, UID_ROOT,
183 GID_OPERATOR, 0640, "vn%d", unit);
184 if (vn->sc_devlist->si_drv1 == NULL) {
185 reference_dev(vn->sc_devlist);
186 vn->sc_devlist->si_drv1 = vn;
187 vn->sc_devlist->si_drv2 = NULL;
189 if (vn->sc_devlist != dev) {
191 dev->si_drv2 = vn->sc_devlist;
192 vn->sc_devlist = dev;
195 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
201 vnopen(struct dev_open_args *ap)
203 cdev_t dev = ap->a_head.a_dev;
207 * Locate preexisting device
210 if ((vn = dev->si_drv1) == NULL)
214 * Update si_bsize fields for device. This data will be overriden by
215 * the slice/parition code for vn accesses through partitions, and
216 * used directly if you open the 'whole disk' device.
218 * si_bsize_best must be reinitialized in case VN has been
219 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
221 dev->si_bsize_phys = vn->sc_secsize;
222 dev->si_bsize_best = vn->sc_secsize;
223 if (dev->si_bsize_best < VN_BSIZE_BEST)
224 dev->si_bsize_best = VN_BSIZE_BEST;
226 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
230 printf("vnopen(%s, 0x%x, 0x%x)\n",
231 devtoname(dev), ap->a_oflags, ap->a_devtype);
237 IFOPT(vn, VN_LABELS) {
238 if (vn->sc_flags & VNF_INITED) {
239 struct disklabel label;
241 /* Build label for whole disk. */
242 bzero(&label, sizeof label);
243 label.d_secsize = vn->sc_secsize;
244 label.d_nsectors = 32;
245 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE);
246 label.d_secpercyl = label.d_nsectors * label.d_ntracks;
247 label.d_ncylinders = vn->sc_size / label.d_secpercyl;
248 label.d_secperunit = vn->sc_size;
249 label.d_partitions[RAW_PART].p_size = vn->sc_size;
251 return (dsopen(dev, ap->a_devtype, 0, &vn->sc_slices, &label));
253 if (dkslice(dev) != WHOLE_DISK_SLICE ||
254 dkpart(dev) != RAW_PART ||
255 ap->a_devtype != S_IFCHR) {
265 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
266 * for vnode-backed vn's, and the new vm_pager_strategy() call for
267 * vm_object-backed vn's.
269 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
272 vnstrategy(struct dev_strategy_args *ap)
274 cdev_t dev = ap->a_head.a_dev;
275 struct bio *bio = ap->a_bio;
283 if ((vn = dev->si_drv1) == NULL)
289 printf("vnstrategy(%p): unit %d\n", bp, unit);
291 if ((vn->sc_flags & VNF_INITED) == 0) {
293 bp->b_flags |= B_ERROR;
298 bp->b_resid = bp->b_bcount;
300 IFOPT(vn, VN_LABELS) {
302 * The vnode device is using disk/slice label support.
304 * The dscheck() function is called for validating the
305 * slices that exist ON the vnode device itself, and
306 * translate the "slice-relative" block number, again.
307 * dscheck() will call biodone() and return NULL if
308 * we are at EOF or beyond the device size.
310 if (vn->sc_slices == NULL) {
312 } else if ((nbio = dscheck(dev, bio, vn->sc_slices)) == NULL) {
316 int pbn; /* in sc_secsize chunks */
317 long sz; /* in sc_secsize chunks */
320 * Check for required alignment. Transfers must be a valid
321 * multiple of the sector size.
323 if (bp->b_bcount % vn->sc_secsize != 0 ||
324 bio->bio_offset % vn->sc_secsize != 0) {
328 pbn = bio->bio_offset / vn->sc_secsize;
329 sz = howmany(bp->b_bcount, vn->sc_secsize);
332 * Check for an illegal pbn or EOF truncation
336 if (pbn + sz > vn->sc_size) {
337 if (pbn > vn->sc_size || (bp->b_flags & B_BNOCLIP))
339 if (pbn == vn->sc_size) {
340 bp->b_resid = bp->b_bcount;
341 bp->b_flags |= B_INVAL;
344 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
346 nbio = push_bio(bio);
347 nbio->bio_offset = pbn * vn->sc_secsize;
351 * Use the translated nbio from this point on
353 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
355 * Freeblks is not handled for vnode-backed elements yet.
358 /* operation complete */
359 } else if (vn->sc_vp) {
363 * If an error occurs, we set B_ERROR but we do not set
364 * B_INVAL because (for a write anyway), the buffer is
370 bzero(&auio, sizeof(auio));
372 aiov.iov_base = bp->b_data;
373 aiov.iov_len = bp->b_bcount;
374 auio.uio_iov = &aiov;
376 auio.uio_offset = nbio->bio_offset;
377 auio.uio_segflg = UIO_SYSSPACE;
378 if (bp->b_cmd == BUF_CMD_READ)
379 auio.uio_rw = UIO_READ;
381 auio.uio_rw = UIO_WRITE;
382 auio.uio_resid = bp->b_bcount;
383 auio.uio_td = curthread;
384 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
385 if (bp->b_cmd == BUF_CMD_READ)
386 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
388 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
389 vn_unlock(vn->sc_vp);
390 bp->b_resid = auio.uio_resid;
393 bp->b_flags |= B_ERROR;
395 /* operation complete */
396 } else if (vn->sc_object) {
398 * OBJT_SWAP I/O (handles read, write, freebuf)
400 * We have nothing to do if freeing blocks on a reserved
401 * swap area, othrewise execute the op.
403 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
405 /* operation complete */
407 vm_pager_strategy(vn->sc_object, nbio);
412 bp->b_resid = bp->b_bcount;
413 bp->b_flags |= B_ERROR | B_INVAL;
414 bp->b_error = EINVAL;
415 /* operation complete */
421 * Shortcuts / check failures on the original bio (not nbio).
424 bp->b_error = EINVAL;
425 bp->b_flags |= B_ERROR | B_INVAL;
433 vnioctl(struct dev_ioctl_args *ap)
435 cdev_t dev = ap->a_head.a_dev;
437 struct vn_ioctl *vio;
442 IFOPT(vn,VN_FOLLOW) {
443 printf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
444 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
458 IFOPT(vn,VN_LABELS) {
459 if (vn->sc_slices != NULL) {
460 error = dsioctl(dev, ap->a_cmd, ap->a_data,
461 ap->a_fflag, &vn->sc_slices);
462 if (error != ENOIOCTL)
465 if (dkslice(dev) != WHOLE_DISK_SLICE ||
466 dkpart(dev) != RAW_PART)
472 error = suser_cred(ap->a_cred, 0);
476 vio = (struct vn_ioctl *)ap->a_data;
477 f = (u_long*)ap->a_data;
481 if (vn->sc_flags & VNF_INITED)
484 if (vio->vn_file == NULL)
485 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
487 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
491 if ((vn->sc_flags & VNF_INITED) == 0)
494 * XXX handle i/o in progress. Return EBUSY, or wait, or
496 * XXX handle multiple opens of the device. Return EBUSY,
497 * or revoke the fd's.
498 * How are these problems handled for removable and failing
499 * hardware devices? (Hint: They are not)
503 printf("vnioctl: CLRed\n");
517 vn->sc_options |= *f;
522 vn->sc_options &= ~(*f);
536 * Attach a file to a VN partition. Return the size in the vn_size
541 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
542 int flag, struct ucred *cred)
545 struct nlookupdata nd;
549 flags = FREAD|FWRITE;
550 error = nlookup_init(&nd, vio->vn_file,
551 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
554 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
555 if (error != EACCES && error != EPERM && error != EROFS)
559 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
562 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
566 if (vp->v_type != VREG ||
567 (error = VOP_GETATTR(vp, &vattr))) {
573 vn->sc_secsize = DEV_BSIZE;
575 nd.nl_open_vp = NULL;
578 * If the size is specified, override the file attributes. Note that
579 * the vn_size argument is in PAGE_SIZE sized blocks.
582 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
584 vn->sc_size = vattr.va_size / vn->sc_secsize;
585 error = vnsetcred(vn, cred);
591 vn->sc_flags |= VNF_INITED;
593 vn->sc_flags |= VNF_READONLY;
594 IFOPT(vn, VN_LABELS) {
596 * Reopen so that `ds' knows which devices are open.
597 * If this is the first VNIOCSET, then we've
598 * guaranteed that the device is the cdev and that
599 * no other slices or labels are open. Otherwise,
600 * we rely on VNIOCCLR not being abused.
602 error = dev_dopen(dev, flag, S_IFCHR, cred);
607 printf("vnioctl: SET vp %p size %x blks\n",
608 vn->sc_vp, vn->sc_size);
617 * Attach swap backing store to a VN partition of the size specified
622 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
623 int flag, struct ucred *cred)
628 * Range check. Disallow negative sizes or any size less then the
629 * size of a page. Then round to a page.
632 if (vio->vn_size <= 0)
636 * Allocate an OBJT_SWAP object.
638 * sc_secsize is PAGE_SIZE'd
640 * vio->vn_size is in PAGE_SIZE'd chunks.
641 * sc_size must be in PAGE_SIZE'd chunks.
642 * Note the truncation.
645 vn->sc_secsize = PAGE_SIZE;
646 vn->sc_size = vio->vn_size;
647 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
648 vn->sc_secsize * (off_t)vio->vn_size,
650 IFOPT(vn, VN_RESERVE) {
651 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
652 vm_pager_deallocate(vn->sc_object);
653 vn->sc_object = NULL;
657 vn->sc_flags |= VNF_INITED;
659 error = vnsetcred(vn, cred);
661 IFOPT(vn, VN_LABELS) {
663 * Reopen so that `ds' knows which devices are open.
664 * If this is the first VNIOCSET, then we've
665 * guaranteed that the device is the cdev and that
666 * no other slices or labels are open. Otherwise,
667 * we rely on VNIOCCLR not being abused.
669 error = dev_dopen(dev, flag, S_IFCHR, cred);
673 IFOPT(vn, VN_FOLLOW) {
674 printf("vnioctl: SET vp %p size %x\n",
675 vn->sc_vp, vn->sc_size);
684 * Duplicate the current processes' credentials. Since we are called only
685 * as the result of a SET ioctl and only root can do that, any future access
686 * to this "disk" is essentially as root. Note that credentials may change
687 * if some other uid can write directly to the mapped file (NFS).
690 vnsetcred(struct vn_softc *vn, struct ucred *cred)
696 * Set credits in our softc
701 vn->sc_cred = crdup(cred);
704 * Horrible kludge to establish credentials for NFS XXX.
711 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
712 bzero(&auio, sizeof(auio));
714 aiov.iov_base = tmpbuf;
715 aiov.iov_len = vn->sc_secsize;
716 auio.uio_iov = &aiov;
719 auio.uio_rw = UIO_READ;
720 auio.uio_segflg = UIO_SYSSPACE;
721 auio.uio_resid = aiov.iov_len;
722 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
723 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
724 vn_unlock(vn->sc_vp);
725 kfree(tmpbuf, M_TEMP);
731 vnclear(struct vn_softc *vn)
734 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
735 if (vn->sc_slices != NULL)
736 dsgone(&vn->sc_slices);
737 vn->sc_flags &= ~VNF_INITED;
738 if (vn->sc_vp != NULL) {
740 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
743 vn->sc_flags &= ~VNF_READONLY;
748 if (vn->sc_object != NULL) {
749 vm_pager_deallocate(vn->sc_object);
750 vn->sc_object = NULL;
756 vnsize(struct dev_psize_args *ap)
758 cdev_t dev = ap->a_head.a_dev;
764 if ((vn->sc_flags & VNF_INITED) == 0)
766 ap->a_result = vn->sc_size;
771 vn_modevent(module_t mod, int type, void *data)
778 dev_ops_add(&vn_ops, 0, 0);
783 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
784 SLIST_REMOVE_HEAD(&vn_list, sc_list);
785 if (vn->sc_flags & VNF_INITED)
787 /* Cleanup all cdev_t's that refer to this unit */
788 while ((dev = vn->sc_devlist) != NULL) {
789 vn->sc_devlist = dev->si_drv2;
790 dev->si_drv1 = dev->si_drv2 = NULL;
795 dev_ops_remove(&vn_ops, 0, 0);
803 DEV_MODULE(vn, vn_modevent, 0);