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>
86 #include <sys/devfs.h>
88 static d_ioctl_t vnioctl;
89 static d_open_t vnopen;
90 static d_close_t vnclose;
91 static d_psize_t vnsize;
92 static d_strategy_t vnstrategy;
93 static d_clone_t vnclone;
95 DEVFS_DECLARE_CLONE_BITMAP(vn);
96 #define VN_PREALLOCATED_UNITS 4
100 #define VN_BSIZE_BEST 8192
104 * D_DISK we want to look like a disk
105 * D_CANFREE We support BUF_CMD_FREEBLKS
108 static struct dev_ops vn_ops = {
109 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
113 .d_write = physwrite,
115 .d_strategy = vnstrategy,
121 int sc_flags; /* flags */
122 u_int64_t sc_size; /* size of vn, sc_secsize scale */
123 int sc_secsize; /* sector size */
125 struct vnode *sc_vp; /* vnode if not NULL */
126 vm_object_t sc_object; /* backing object if not NULL */
127 struct ucred *sc_cred; /* credentials */
128 int sc_maxactive; /* max # of active requests */
129 struct buf sc_tab; /* transfer queue */
130 u_long sc_options; /* options */
131 cdev_t sc_devlist; /* devices that refer to this unit */
132 SLIST_ENTRY(vn_softc) sc_list;
135 static SLIST_HEAD(, vn_softc) vn_list;
138 #define VNF_INITED 0x01
139 #define VNF_READONLY 0x02
141 static u_long vn_options;
143 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
144 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
146 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
147 static void vnclear (struct vn_softc *vn);
148 static int vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu);
149 static int vn_modevent (module_t, int, void *);
150 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
151 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
152 static cdev_t vn_create(int unit, struct devfs_bitmap *bitmap);
155 vnclone(struct dev_clone_args *ap)
159 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(vn), 0);
160 ap->a_dev = vn_create(unit, &DEVFS_CLONE_BITMAP(vn));
166 vnclose(struct dev_close_args *ap)
171 static struct vn_softc *
176 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
181 vninitvn(struct vn_softc *vn, cdev_t dev)
185 KKASSERT(vn != NULL);
186 KKASSERT(dev != NULL);
191 vn->sc_devlist = dev;
192 if (vn->sc_devlist->si_drv1 == NULL) {
193 reference_dev(vn->sc_devlist);
194 vn->sc_devlist->si_drv1 = vn;
195 vn->sc_devlist->si_drv2 = NULL;
197 if (vn->sc_devlist != dev) {
199 dev->si_drv2 = vn->sc_devlist;
200 vn->sc_devlist = dev;
203 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
207 * Called only when si_drv1 is NULL. Locate the associated vn node and
208 * attach the device to it.
210 static struct vn_softc *
217 SLIST_FOREACH(vn, &vn_list, sc_list) {
218 if (vn->sc_unit == unit) {
220 dev->si_drv2 = vn->sc_devlist;
221 vn->sc_devlist = dev;
227 KKASSERT(vn != NULL);
233 vnopen(struct dev_open_args *ap)
235 cdev_t dev = ap->a_head.a_dev;
239 * Locate preexisting device
242 if ((vn = dev->si_drv1) == NULL)
246 * Update si_bsize fields for device. This data will be overriden by
247 * the slice/parition code for vn accesses through partitions, and
248 * used directly if you open the 'whole disk' device.
250 * si_bsize_best must be reinitialized in case VN has been
251 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
253 dev->si_bsize_phys = vn->sc_secsize;
254 dev->si_bsize_best = vn->sc_secsize;
255 if (dev->si_bsize_best < VN_BSIZE_BEST)
256 dev->si_bsize_best = VN_BSIZE_BEST;
258 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
262 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
263 devtoname(dev), ap->a_oflags, ap->a_devtype);
271 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
272 * for vnode-backed vn's, and the new vm_pager_strategy() call for
273 * vm_object-backed vn's.
276 vnstrategy(struct dev_strategy_args *ap)
278 cdev_t dev = ap->a_head.a_dev;
279 struct bio *bio = ap->a_bio;
287 if ((vn = dev->si_drv1) == NULL)
293 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
295 if ((vn->sc_flags & VNF_INITED) == 0) {
297 bp->b_flags |= B_ERROR;
302 bp->b_resid = bp->b_bcount;
305 * The vnode device is using disk/slice label support.
307 * The dscheck() function is called for validating the
308 * slices that exist ON the vnode device itself, and
309 * translate the "slice-relative" block number, again.
310 * dscheck() will call biodone() and return NULL if
311 * we are at EOF or beyond the device size.
317 * Use the translated nbio from this point on
319 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
321 * Freeblks is not handled for vnode-backed elements yet.
324 /* operation complete */
325 } else if (vn->sc_vp) {
329 * If an error occurs, we set B_ERROR but we do not set
330 * B_INVAL because (for a write anyway), the buffer is
336 bzero(&auio, sizeof(auio));
338 aiov.iov_base = bp->b_data;
339 aiov.iov_len = bp->b_bcount;
340 auio.uio_iov = &aiov;
342 auio.uio_offset = nbio->bio_offset;
343 auio.uio_segflg = UIO_SYSSPACE;
344 if (bp->b_cmd == BUF_CMD_READ)
345 auio.uio_rw = UIO_READ;
347 auio.uio_rw = UIO_WRITE;
348 auio.uio_resid = bp->b_bcount;
349 auio.uio_td = curthread;
350 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
351 if (bp->b_cmd == BUF_CMD_READ)
352 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred);
354 error = VOP_WRITE(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred);
355 vn_unlock(vn->sc_vp);
356 bp->b_resid = auio.uio_resid;
359 bp->b_flags |= B_ERROR;
361 /* operation complete */
362 } else if (vn->sc_object) {
364 * OBJT_SWAP I/O (handles read, write, freebuf)
366 * We have nothing to do if freeing blocks on a reserved
367 * swap area, othrewise execute the op.
369 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
371 /* operation complete */
373 vm_pager_strategy(vn->sc_object, nbio);
378 bp->b_resid = bp->b_bcount;
379 bp->b_flags |= B_ERROR | B_INVAL;
380 bp->b_error = EINVAL;
381 /* operation complete */
389 vnioctl(struct dev_ioctl_args *ap)
391 cdev_t dev = ap->a_head.a_dev;
393 struct vn_ioctl *vio;
398 IFOPT(vn,VN_FOLLOW) {
399 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
400 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
416 if (dkslice(dev) != WHOLE_DISK_SLICE ||
417 dkpart(dev) != WHOLE_SLICE_PART)
423 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
427 vio = (struct vn_ioctl *)ap->a_data;
428 f = (u_long*)ap->a_data;
432 if (vn->sc_flags & VNF_INITED)
435 if (vio->vn_file == NULL)
436 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
438 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
442 if ((vn->sc_flags & VNF_INITED) == 0)
445 * XXX handle i/o in progress. Return EBUSY, or wait, or
447 * XXX handle multiple opens of the device. Return EBUSY,
448 * or revoke the fd's.
449 * How are these problems handled for removable and failing
450 * hardware devices? (Hint: They are not)
452 if (count_dev(vn->sc_devlist) > 1)
457 kprintf("vnioctl: CLRed\n");
459 if (dkunit(dev) >= VN_PREALLOCATED_UNITS) {
460 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dkunit(dev));
461 disk_destroy(&vn->sc_disk);
467 error = vnget(dev, vn, (struct vn_user *) ap->a_data);
481 vn->sc_options |= *f;
486 vn->sc_options &= ~(*f);
500 * Attach a file to a VN partition. Return the size in the vn_size
505 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
506 int flag, struct ucred *cred)
509 struct nlookupdata nd;
512 struct disk_info info;
514 flags = FREAD|FWRITE;
515 error = nlookup_init(&nd, vio->vn_file,
516 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
519 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
520 if (error != EACCES && error != EPERM && error != EROFS)
524 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
527 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
531 if (vp->v_type != VREG ||
532 (error = VOP_GETATTR(vp, &vattr))) {
538 vn->sc_secsize = DEV_BSIZE;
540 nd.nl_open_vp = NULL;
543 * If the size is specified, override the file attributes. Note that
544 * the vn_size argument is in PAGE_SIZE sized blocks.
547 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
549 vn->sc_size = vattr.va_size / vn->sc_secsize;
550 error = vnsetcred(vn, cred);
556 vn->sc_flags |= VNF_INITED;
558 vn->sc_flags |= VNF_READONLY;
561 * Set the disk info so that probing is triggered
563 bzero(&info, sizeof(struct disk_info));
564 info.d_media_blksize = vn->sc_secsize;
565 info.d_media_blocks = vn->sc_size;
567 * reserve mbr sector for backwards compatibility
568 * when no slices exist.
570 info.d_dsflags = DSO_COMPATMBR;
571 info.d_secpertrack = 32;
572 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
573 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
574 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
575 disk_setdiskinfo_sync(&vn->sc_disk, &info);
577 error = dev_dopen(dev, flag, S_IFCHR, cred);
582 kprintf("vnioctl: SET vp %p size %llx blks\n",
583 vn->sc_vp, (long long)vn->sc_size);
592 * Attach swap backing store to a VN partition of the size specified
597 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
598 int flag, struct ucred *cred)
601 struct disk_info info;
604 * Range check. Disallow negative sizes or any size less then the
605 * size of a page. Then round to a page.
608 if (vio->vn_size <= 0)
612 * Allocate an OBJT_SWAP object.
614 * sc_secsize is PAGE_SIZE'd
616 * vio->vn_size is in PAGE_SIZE'd chunks.
617 * sc_size must be in PAGE_SIZE'd chunks.
618 * Note the truncation.
621 vn->sc_secsize = PAGE_SIZE;
622 vn->sc_size = vio->vn_size;
623 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
624 vn->sc_secsize * (off_t)vio->vn_size,
626 IFOPT(vn, VN_RESERVE) {
627 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
628 vm_pager_deallocate(vn->sc_object);
629 vn->sc_object = NULL;
633 vn->sc_flags |= VNF_INITED;
635 error = vnsetcred(vn, cred);
638 * Set the disk info so that probing is triggered
640 bzero(&info, sizeof(struct disk_info));
641 info.d_media_blksize = vn->sc_secsize;
642 info.d_media_blocks = vn->sc_size;
644 * reserve mbr sector for backwards compatibility
645 * when no slices exist.
647 info.d_dsflags = DSO_COMPATMBR;
648 info.d_secpertrack = 32;
649 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
650 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
651 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
652 disk_setdiskinfo_sync(&vn->sc_disk, &info);
654 error = dev_dopen(dev, flag, S_IFCHR, cred);
657 IFOPT(vn, VN_FOLLOW) {
658 kprintf("vnioctl: SET vp %p size %llx\n",
659 vn->sc_vp, (long long)vn->sc_size);
668 * Duplicate the current processes' credentials. Since we are called only
669 * as the result of a SET ioctl and only root can do that, any future access
670 * to this "disk" is essentially as root. Note that credentials may change
671 * if some other uid can write directly to the mapped file (NFS).
674 vnsetcred(struct vn_softc *vn, struct ucred *cred)
680 * Set credits in our softc
685 vn->sc_cred = crdup(cred);
688 * Horrible kludge to establish credentials for NFS XXX.
695 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
696 bzero(&auio, sizeof(auio));
698 aiov.iov_base = tmpbuf;
699 aiov.iov_len = vn->sc_secsize;
700 auio.uio_iov = &aiov;
703 auio.uio_rw = UIO_READ;
704 auio.uio_segflg = UIO_SYSSPACE;
705 auio.uio_resid = aiov.iov_len;
706 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
707 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
708 vn_unlock(vn->sc_vp);
709 kfree(tmpbuf, M_TEMP);
715 vnclear(struct vn_softc *vn)
718 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
719 vn->sc_flags &= ~VNF_INITED;
720 if (vn->sc_vp != NULL) {
722 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
725 vn->sc_flags &= ~VNF_READONLY;
730 if (vn->sc_object != NULL) {
731 vm_pager_deallocate(vn->sc_object);
732 vn->sc_object = NULL;
735 disk_unprobe(&vn->sc_disk);
743 * populate a struct vn_user for the VNIOCGET ioctl.
744 * interface conventions defined in sys/sys/vnioctl.h.
748 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu)
750 int error, found = 0;
751 char *freepath, *fullpath;
754 if (vnu->vnu_unit == -1) {
755 vnu->vnu_unit = dkunit(dev);
757 else if (vnu->vnu_unit < 0)
760 SLIST_FOREACH(vn, &vn_list, sc_list) {
762 if(vn->sc_unit != vnu->vnu_unit)
767 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) {
769 /* note: u_cred checked in vnioctl above */
770 error = VOP_GETATTR(vn->sc_vp, &vattr);
772 kprintf("vnget: VOP_GETATTR for %p failed\n",
777 error = vn_fullpath(curproc, vn->sc_vp,
778 &fullpath, &freepath);
781 kprintf("vnget: unable to resolve vp %p\n",
786 strlcpy(vnu->vnu_file, fullpath,
787 sizeof(vnu->vnu_file));
788 kfree(freepath, M_TEMP);
789 vnu->vnu_dev = vattr.va_fsid;
790 vnu->vnu_ino = vattr.va_fileid;
793 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){
795 strlcpy(vnu->vnu_file, _VN_USER_SWAP,
796 sizeof(vnu->vnu_file));
797 vnu->vnu_size = vn->sc_size;
798 vnu->vnu_secsize = vn->sc_secsize;
802 bzero(vnu->vnu_file, sizeof(vnu->vnu_file));
817 vnsize(struct dev_psize_args *ap)
819 cdev_t dev = ap->a_head.a_dev;
825 if ((vn->sc_flags & VNF_INITED) == 0)
827 ap->a_result = (int64_t)vn->sc_size;
832 vn_create(int unit, struct devfs_bitmap *bitmap)
835 struct disk_info info;
839 dev = disk_create(unit, &vn->sc_disk, &vn_ops);
842 bzero(&info, sizeof(struct disk_info));
843 info.d_media_blksize = 512;
844 info.d_media_blocks = 0;
845 info.d_dsflags = DSO_MBRQUIET;
846 info.d_secpertrack = 32;
848 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
849 info.d_ncylinders = 0;
850 disk_setdiskinfo_sync(&vn->sc_disk, &info);
853 devfs_clone_bitmap_set(bitmap, unit);
859 vn_modevent(module_t mod, int type, void *data)
862 struct disk_info info;
863 static cdev_t dev = NULL;
868 dev = make_autoclone_dev(&vn_ops, &DEVFS_CLONE_BITMAP(vn), vnclone, UID_ROOT,
869 GID_OPERATOR, 0640, "vn");
871 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) {
872 vn_create(i, &DEVFS_CLONE_BITMAP(vn));
876 destroy_autoclone_dev(dev, &DEVFS_CLONE_BITMAP(vn));
879 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
880 SLIST_REMOVE_HEAD(&vn_list, sc_list);
881 if (vn->sc_flags & VNF_INITED)
883 /* Cleanup all cdev_t's that refer to this unit */
884 while ((dev = vn->sc_devlist) != NULL) {
885 vn->sc_devlist = dev->si_drv2;
886 dev->si_drv1 = dev->si_drv2 = NULL;
891 dev_ops_remove_all(&vn_ops);
899 DEV_MODULE(vn, vn_modevent, 0);