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 1.38 2008/07/01 02:02:53 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>
65 #include <sys/nlookup.h>
67 #include <sys/malloc.h>
68 #include <sys/mount.h>
69 #include <sys/vnode.h>
70 #include <sys/fcntl.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 vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu);
145 static int vn_modevent (module_t, int, void *);
146 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
147 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
150 vnclose(struct dev_close_args *ap)
152 cdev_t dev = ap->a_head.a_dev;
153 struct vn_softc *vn = dev->si_drv1;
156 if (vn->sc_slices != NULL)
157 dsclose(dev, ap->a_devtype, vn->sc_slices);
162 * Called only when si_drv1 is NULL. Locate the associated vn node and
163 * attach the device to it.
165 static struct vn_softc *
172 SLIST_FOREACH(vn, &vn_list, sc_list) {
173 if (vn->sc_unit == unit) {
175 dev->si_drv2 = vn->sc_devlist;
176 vn->sc_devlist = dev;
182 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
185 vn->sc_devlist = make_dev(&vn_ops, 0, UID_ROOT,
186 GID_OPERATOR, 0640, "vn%d", unit);
187 if (vn->sc_devlist->si_drv1 == NULL) {
188 reference_dev(vn->sc_devlist);
189 vn->sc_devlist->si_drv1 = vn;
190 vn->sc_devlist->si_drv2 = NULL;
192 if (vn->sc_devlist != dev) {
194 dev->si_drv2 = vn->sc_devlist;
195 vn->sc_devlist = dev;
198 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
204 vnopen(struct dev_open_args *ap)
206 cdev_t dev = ap->a_head.a_dev;
208 struct disk_info *info;
211 * Locate preexisting device
214 if ((vn = dev->si_drv1) == NULL)
218 * Update si_bsize fields for device. This data will be overriden by
219 * the slice/parition code for vn accesses through partitions, and
220 * used directly if you open the 'whole disk' device.
222 * si_bsize_best must be reinitialized in case VN has been
223 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
225 dev->si_bsize_phys = vn->sc_secsize;
226 dev->si_bsize_best = vn->sc_secsize;
227 if (dev->si_bsize_best < VN_BSIZE_BEST)
228 dev->si_bsize_best = VN_BSIZE_BEST;
230 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
234 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
235 devtoname(dev), ap->a_oflags, ap->a_devtype);
241 IFOPT(vn, VN_LABELS) {
242 if (vn->sc_flags & VNF_INITED) {
244 bzero(info, sizeof(*info));
245 info->d_media_blksize = vn->sc_secsize;
246 info->d_media_blocks = vn->sc_size;
248 * reserve mbr sector for backwards compatibility
249 * when no slices exist.
251 info->d_dsflags = DSO_COMPATMBR;
253 info->d_secpertrack = 32;
254 info->d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
255 info->d_secpercyl = info->d_secpertrack *
257 info->d_ncylinders = vn->sc_size / info->d_secpercyl;
259 return (dsopen(dev, ap->a_devtype, 0,
260 &vn->sc_slices, info));
262 if (dkslice(dev) != WHOLE_DISK_SLICE ||
263 dkpart(dev) != WHOLE_SLICE_PART ||
264 ap->a_devtype != S_IFCHR) {
274 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
275 * for vnode-backed vn's, and the new vm_pager_strategy() call for
276 * vm_object-backed vn's.
279 vnstrategy(struct dev_strategy_args *ap)
281 cdev_t dev = ap->a_head.a_dev;
282 struct bio *bio = ap->a_bio;
290 if ((vn = dev->si_drv1) == NULL)
296 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
298 if ((vn->sc_flags & VNF_INITED) == 0) {
300 bp->b_flags |= B_ERROR;
305 bp->b_resid = bp->b_bcount;
307 IFOPT(vn, VN_LABELS) {
309 * The vnode device is using disk/slice label support.
311 * The dscheck() function is called for validating the
312 * slices that exist ON the vnode device itself, and
313 * translate the "slice-relative" block number, again.
314 * dscheck() will call biodone() and return NULL if
315 * we are at EOF or beyond the device size.
317 if (vn->sc_slices == NULL) {
319 } else if ((nbio = dscheck(dev, bio, vn->sc_slices)) == NULL) {
323 int64_t pbn; /* in sc_secsize chunks */
324 long sz; /* in sc_secsize chunks */
327 * Check for required alignment. Transfers must be a valid
328 * multiple of the sector size.
330 if (bp->b_bcount % vn->sc_secsize != 0 ||
331 bio->bio_offset % vn->sc_secsize != 0) {
335 pbn = bio->bio_offset / vn->sc_secsize;
336 sz = howmany(bp->b_bcount, vn->sc_secsize);
339 * Check for an illegal pbn or EOF truncation
343 if (pbn + sz > vn->sc_size) {
344 if (pbn > vn->sc_size || (bp->b_flags & B_BNOCLIP))
346 if (pbn == vn->sc_size) {
347 bp->b_resid = bp->b_bcount;
348 bp->b_flags |= B_INVAL;
351 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
353 nbio = push_bio(bio);
354 nbio->bio_offset = pbn * vn->sc_secsize;
358 * Use the translated nbio from this point on
360 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
362 * Freeblks is not handled for vnode-backed elements yet.
365 /* operation complete */
366 } else if (vn->sc_vp) {
370 * If an error occurs, we set B_ERROR but we do not set
371 * B_INVAL because (for a write anyway), the buffer is
377 bzero(&auio, sizeof(auio));
379 aiov.iov_base = bp->b_data;
380 aiov.iov_len = bp->b_bcount;
381 auio.uio_iov = &aiov;
383 auio.uio_offset = nbio->bio_offset;
384 auio.uio_segflg = UIO_SYSSPACE;
385 if (bp->b_cmd == BUF_CMD_READ)
386 auio.uio_rw = UIO_READ;
388 auio.uio_rw = UIO_WRITE;
389 auio.uio_resid = bp->b_bcount;
390 auio.uio_td = curthread;
391 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
392 if (bp->b_cmd == BUF_CMD_READ)
393 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred);
395 error = VOP_WRITE(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred);
396 vn_unlock(vn->sc_vp);
397 bp->b_resid = auio.uio_resid;
400 bp->b_flags |= B_ERROR;
402 /* operation complete */
403 } else if (vn->sc_object) {
405 * OBJT_SWAP I/O (handles read, write, freebuf)
407 * We have nothing to do if freeing blocks on a reserved
408 * swap area, othrewise execute the op.
410 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
412 /* operation complete */
414 vm_pager_strategy(vn->sc_object, nbio);
419 bp->b_resid = bp->b_bcount;
420 bp->b_flags |= B_ERROR | B_INVAL;
421 bp->b_error = EINVAL;
422 /* operation complete */
428 * Shortcuts / check failures on the original bio (not nbio).
431 bp->b_error = EINVAL;
432 bp->b_flags |= B_ERROR | B_INVAL;
440 vnioctl(struct dev_ioctl_args *ap)
442 cdev_t dev = ap->a_head.a_dev;
444 struct vn_ioctl *vio;
449 IFOPT(vn,VN_FOLLOW) {
450 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
451 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) != WHOLE_SLICE_PART)
481 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 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)
510 if (count_dev(vn->sc_devlist) > 1)
515 kprintf("vnioctl: CLRed\n");
519 error = vnget(dev, vn, (struct vn_user *) ap->a_data);
533 vn->sc_options |= *f;
538 vn->sc_options &= ~(*f);
552 * Attach a file to a VN partition. Return the size in the vn_size
557 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
558 int flag, struct ucred *cred)
561 struct nlookupdata nd;
565 flags = FREAD|FWRITE;
566 error = nlookup_init(&nd, vio->vn_file,
567 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
570 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
571 if (error != EACCES && error != EPERM && error != EROFS)
575 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
578 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
582 if (vp->v_type != VREG ||
583 (error = VOP_GETATTR(vp, &vattr))) {
589 vn->sc_secsize = DEV_BSIZE;
591 nd.nl_open_vp = NULL;
594 * If the size is specified, override the file attributes. Note that
595 * the vn_size argument is in PAGE_SIZE sized blocks.
598 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
600 vn->sc_size = vattr.va_size / vn->sc_secsize;
601 error = vnsetcred(vn, cred);
607 vn->sc_flags |= VNF_INITED;
609 vn->sc_flags |= VNF_READONLY;
610 IFOPT(vn, VN_LABELS) {
612 * Reopen so that `ds' knows which devices are open.
613 * If this is the first VNIOCSET, then we've
614 * guaranteed that the device is the cdev and that
615 * no other slices or labels are open. Otherwise,
616 * we rely on VNIOCCLR not being abused.
618 error = dev_dopen(dev, flag, S_IFCHR, cred);
623 kprintf("vnioctl: SET vp %p size %llx blks\n",
624 vn->sc_vp, vn->sc_size);
633 * Attach swap backing store to a VN partition of the size specified
638 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
639 int flag, struct ucred *cred)
644 * Range check. Disallow negative sizes or any size less then the
645 * size of a page. Then round to a page.
648 if (vio->vn_size <= 0)
652 * Allocate an OBJT_SWAP object.
654 * sc_secsize is PAGE_SIZE'd
656 * vio->vn_size is in PAGE_SIZE'd chunks.
657 * sc_size must be in PAGE_SIZE'd chunks.
658 * Note the truncation.
661 vn->sc_secsize = PAGE_SIZE;
662 vn->sc_size = vio->vn_size;
663 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
664 vn->sc_secsize * (off_t)vio->vn_size,
666 IFOPT(vn, VN_RESERVE) {
667 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
668 vm_pager_deallocate(vn->sc_object);
669 vn->sc_object = NULL;
673 vn->sc_flags |= VNF_INITED;
675 error = vnsetcred(vn, cred);
677 IFOPT(vn, VN_LABELS) {
679 * Reopen so that `ds' knows which devices are open.
680 * If this is the first VNIOCSET, then we've
681 * guaranteed that the device is the cdev and that
682 * no other slices or labels are open. Otherwise,
683 * we rely on VNIOCCLR not being abused.
685 error = dev_dopen(dev, flag, S_IFCHR, cred);
689 IFOPT(vn, VN_FOLLOW) {
690 kprintf("vnioctl: SET vp %p size %llx\n",
691 vn->sc_vp, vn->sc_size);
700 * Duplicate the current processes' credentials. Since we are called only
701 * as the result of a SET ioctl and only root can do that, any future access
702 * to this "disk" is essentially as root. Note that credentials may change
703 * if some other uid can write directly to the mapped file (NFS).
706 vnsetcred(struct vn_softc *vn, struct ucred *cred)
712 * Set credits in our softc
717 vn->sc_cred = crdup(cred);
720 * Horrible kludge to establish credentials for NFS XXX.
727 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
728 bzero(&auio, sizeof(auio));
730 aiov.iov_base = tmpbuf;
731 aiov.iov_len = vn->sc_secsize;
732 auio.uio_iov = &aiov;
735 auio.uio_rw = UIO_READ;
736 auio.uio_segflg = UIO_SYSSPACE;
737 auio.uio_resid = aiov.iov_len;
738 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
739 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
740 vn_unlock(vn->sc_vp);
741 kfree(tmpbuf, M_TEMP);
747 vnclear(struct vn_softc *vn)
750 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
751 if (vn->sc_slices != NULL)
752 dsgone(&vn->sc_slices);
753 vn->sc_flags &= ~VNF_INITED;
754 if (vn->sc_vp != NULL) {
756 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
759 vn->sc_flags &= ~VNF_READONLY;
764 if (vn->sc_object != NULL) {
765 vm_pager_deallocate(vn->sc_object);
766 vn->sc_object = NULL;
774 * populate a struct vn_user for the VNIOCGET ioctl.
775 * interface conventions defined in sys/sys/vnioctl.h.
779 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu)
781 int error, found = 0;
782 char *freepath, *fullpath;
785 if (vnu->vnu_unit == -1) {
786 vnu->vnu_unit = dkunit(dev);
788 else if (vnu->vnu_unit < 0)
791 SLIST_FOREACH(vn, &vn_list, sc_list) {
793 if(vn->sc_unit != vnu->vnu_unit)
798 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) {
800 /* note: u_cred checked in vnioctl above */
801 error = VOP_GETATTR(vn->sc_vp, &vattr);
803 kprintf("vnget: VOP_GETATTR for %p failed\n",
808 error = vn_fullpath(curproc, vn->sc_vp,
809 &fullpath, &freepath);
812 kprintf("vnget: unable to resolve vp %p\n",
817 strlcpy(vnu->vnu_file, fullpath,
818 sizeof(vnu->vnu_file));
819 kfree(freepath, M_TEMP);
820 vnu->vnu_dev = vattr.va_fsid;
821 vnu->vnu_ino = vattr.va_fileid;
824 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){
826 strlcpy(vnu->vnu_file, _VN_USER_SWAP,
827 sizeof(vnu->vnu_file));
828 vnu->vnu_size = vn->sc_size;
829 vnu->vnu_secsize = vn->sc_secsize;
833 bzero(vnu->vnu_file, sizeof(vnu->vnu_file));
848 vnsize(struct dev_psize_args *ap)
850 cdev_t dev = ap->a_head.a_dev;
856 if ((vn->sc_flags & VNF_INITED) == 0)
858 ap->a_result = (int64_t)vn->sc_size;
863 vn_modevent(module_t mod, int type, void *data)
870 dev_ops_add(&vn_ops, 0, 0);
875 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
876 SLIST_REMOVE_HEAD(&vn_list, sc_list);
877 if (vn->sc_flags & VNF_INITED)
879 /* Cleanup all cdev_t's that refer to this unit */
880 while ((dev = vn->sc_devlist) != NULL) {
881 vn->sc_devlist = dev->si_drv2;
882 dev->si_drv1 = dev->si_drv2 = NULL;
887 dev_ops_remove_all(&vn_ops);
895 DEV_MODULE(vn, vn_modevent, 0);