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?
61 #include <sys/param.h>
62 #include <sys/systm.h>
63 #include <sys/kernel.h>
66 #include <sys/nlookup.h>
68 #include <sys/malloc.h>
69 #include <sys/mount.h>
70 #include <sys/vnode.h>
71 #include <sys/fcntl.h>
73 #include <sys/diskslice.h>
76 #include <sys/module.h>
77 #include <sys/vnioctl.h>
80 #include <vm/vm_object.h>
81 #include <vm/vm_page.h>
82 #include <vm/vm_pager.h>
83 #include <vm/vm_pageout.h>
84 #include <vm/swap_pager.h>
85 #include <vm/vm_extern.h>
86 #include <vm/vm_zone.h>
87 #include <sys/devfs.h>
89 static d_ioctl_t vnioctl;
90 static d_open_t vnopen;
91 static d_close_t vnclose;
92 static d_psize_t vnsize;
93 static d_strategy_t vnstrategy;
94 static d_clone_t vnclone;
96 DEVFS_DECLARE_CLONE_BITMAP(vn);
99 #define VN_PREALLOCATED_UNITS 4
101 #define VN_PREALLOCATED_UNITS NVN
104 #define CDEV_MAJOR 43
106 #define VN_BSIZE_BEST 8192
110 * D_DISK we want to look like a disk
111 * D_CANFREE We support BUF_CMD_FREEBLKS
114 static struct dev_ops vn_ops = {
115 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
119 .d_write = physwrite,
121 .d_strategy = vnstrategy,
127 int sc_flags; /* flags */
128 u_int64_t sc_size; /* size of vn, sc_secsize scale */
129 int sc_secsize; /* sector size */
131 struct vnode *sc_vp; /* vnode if not NULL */
132 vm_object_t sc_object; /* backing object if not NULL */
133 struct ucred *sc_cred; /* credentials */
134 int sc_maxactive; /* max # of active requests */
135 struct buf sc_tab; /* transfer queue */
136 u_long sc_options; /* options */
137 cdev_t sc_devlist; /* devices that refer to this unit */
138 SLIST_ENTRY(vn_softc) sc_list;
141 static SLIST_HEAD(, vn_softc) vn_list;
144 #define VNF_INITED 0x01
145 #define VNF_READONLY 0x02
147 static u_long vn_options;
149 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
150 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
152 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
153 static void vnclear (struct vn_softc *vn);
154 static int vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu);
155 static int vn_modevent (module_t, int, void *);
156 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
157 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
158 static cdev_t vn_create(int unit, struct devfs_bitmap *bitmap);
161 vnclone(struct dev_clone_args *ap)
165 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(vn), 0);
166 ap->a_dev = vn_create(unit, &DEVFS_CLONE_BITMAP(vn));
172 vnclose(struct dev_close_args *ap)
177 static struct vn_softc *
182 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
187 vninitvn(struct vn_softc *vn, cdev_t dev)
191 KKASSERT(vn != NULL);
192 KKASSERT(dev != NULL);
197 vn->sc_devlist = dev;
198 if (vn->sc_devlist->si_drv1 == NULL) {
199 reference_dev(vn->sc_devlist);
200 vn->sc_devlist->si_drv1 = vn;
201 vn->sc_devlist->si_drv2 = NULL;
203 if (vn->sc_devlist != dev) {
205 dev->si_drv2 = vn->sc_devlist;
206 vn->sc_devlist = dev;
209 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
213 * Called only when si_drv1 is NULL. Locate the associated vn node and
214 * attach the device to it.
216 static struct vn_softc *
223 SLIST_FOREACH(vn, &vn_list, sc_list) {
224 if (vn->sc_unit == unit) {
226 dev->si_drv2 = vn->sc_devlist;
227 vn->sc_devlist = dev;
233 KKASSERT(vn != NULL);
239 vnopen(struct dev_open_args *ap)
241 cdev_t dev = ap->a_head.a_dev;
245 * Locate preexisting device
248 if ((vn = dev->si_drv1) == NULL)
252 * Update si_bsize fields for device. This data will be overriden by
253 * the slice/parition code for vn accesses through partitions, and
254 * used directly if you open the 'whole disk' device.
256 * si_bsize_best must be reinitialized in case VN has been
257 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
259 dev->si_bsize_phys = vn->sc_secsize;
260 dev->si_bsize_best = vn->sc_secsize;
261 if (dev->si_bsize_best < VN_BSIZE_BEST)
262 dev->si_bsize_best = VN_BSIZE_BEST;
264 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
268 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
269 devtoname(dev), ap->a_oflags, ap->a_devtype);
277 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
278 * for vnode-backed vn's, and the swap_pager_strategy() call for
279 * vm_object-backed vn's.
282 vnstrategy(struct dev_strategy_args *ap)
284 cdev_t dev = ap->a_head.a_dev;
285 struct bio *bio = ap->a_bio;
293 if ((vn = dev->si_drv1) == NULL)
299 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
301 if ((vn->sc_flags & VNF_INITED) == 0) {
303 bp->b_flags |= B_ERROR;
308 bp->b_resid = bp->b_bcount;
311 * The vnode device is using disk/slice label support.
313 * The dscheck() function is called for validating the
314 * slices that exist ON the vnode device itself, and
315 * translate the "slice-relative" block number, again.
316 * dscheck() will call biodone() and return NULL if
317 * we are at EOF or beyond the device size.
323 * Use the translated nbio from this point on
325 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
327 * Freeblks is not handled for vnode-backed elements yet.
330 /* operation complete */
331 } else if (vn->sc_vp) {
335 * If an error occurs, we set B_ERROR but we do not set
336 * B_INVAL because (for a write anyway), the buffer is
342 bzero(&auio, sizeof(auio));
344 aiov.iov_base = bp->b_data;
345 aiov.iov_len = bp->b_bcount;
346 auio.uio_iov = &aiov;
348 auio.uio_offset = nbio->bio_offset;
349 auio.uio_segflg = UIO_SYSSPACE;
350 if (bp->b_cmd == BUF_CMD_READ)
351 auio.uio_rw = UIO_READ;
353 auio.uio_rw = UIO_WRITE;
354 auio.uio_resid = bp->b_bcount;
355 auio.uio_td = curthread;
358 * Don't use IO_DIRECT here, it really gets in the way
359 * due to typical blocksize differences between the
360 * fs backing the VN device and whatever is running on
363 if (bp->b_cmd == BUF_CMD_READ) {
364 vn_lock(vn->sc_vp, LK_SHARED | LK_RETRY);
365 error = VOP_READ(vn->sc_vp, &auio, IO_RECURSE,
368 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
369 error = VOP_WRITE(vn->sc_vp, &auio, IO_RECURSE,
372 vn_unlock(vn->sc_vp);
373 bp->b_resid = auio.uio_resid;
376 bp->b_flags |= B_ERROR;
378 /* operation complete */
379 } else if (vn->sc_object) {
381 * OBJT_SWAP I/O (handles read, write, freebuf)
383 * We have nothing to do if freeing blocks on a reserved
384 * swap area, othrewise execute the op.
386 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
388 /* operation complete */
390 swap_pager_strategy(vn->sc_object, nbio);
395 bp->b_resid = bp->b_bcount;
396 bp->b_flags |= B_ERROR | B_INVAL;
397 bp->b_error = EINVAL;
398 /* operation complete */
406 vnioctl(struct dev_ioctl_args *ap)
408 cdev_t dev = ap->a_head.a_dev;
410 struct vn_ioctl *vio;
415 IFOPT(vn,VN_FOLLOW) {
416 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
417 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
433 if (dkslice(dev) != WHOLE_DISK_SLICE ||
434 dkpart(dev) != WHOLE_SLICE_PART)
440 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
444 vio = (struct vn_ioctl *)ap->a_data;
445 f = (u_long*)ap->a_data;
449 if (vn->sc_flags & VNF_INITED)
452 if (vio->vn_file == NULL)
453 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
455 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
459 if ((vn->sc_flags & VNF_INITED) == 0)
462 * XXX handle i/o in progress. Return EBUSY, or wait, or
464 * XXX handle multiple opens of the device. Return EBUSY,
465 * or revoke the fd's.
466 * How are these problems handled for removable and failing
467 * hardware devices? (Hint: They are not)
469 if (count_dev(vn->sc_devlist) > 1)
474 kprintf("vnioctl: CLRed\n");
476 if (dkunit(dev) >= VN_PREALLOCATED_UNITS) {
477 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dkunit(dev));
478 disk_destroy(&vn->sc_disk);
479 SLIST_REMOVE(&vn_list, vn, vn_softc, sc_list);
485 error = vnget(dev, vn, (struct vn_user *) ap->a_data);
499 vn->sc_options |= *f;
504 vn->sc_options &= ~(*f);
518 * Attach a file to a VN partition. Return the size in the vn_size
523 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
524 int flag, struct ucred *cred)
527 struct nlookupdata nd;
530 struct disk_info info;
532 flags = FREAD|FWRITE;
533 error = nlookup_init(&nd, vio->vn_file,
534 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
537 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
538 if (error != EACCES && error != EPERM && error != EROFS)
542 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
545 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
549 if (vp->v_type != VREG ||
550 (error = VOP_GETATTR(vp, &vattr))) {
556 vn->sc_secsize = DEV_BSIZE;
558 nd.nl_open_vp = NULL;
561 * If the size is specified, override the file attributes. Note that
562 * the vn_size argument is in PAGE_SIZE sized blocks.
565 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
567 vn->sc_size = vattr.va_size / vn->sc_secsize;
568 error = vnsetcred(vn, cred);
574 vn->sc_flags |= VNF_INITED;
576 vn->sc_flags |= VNF_READONLY;
579 * Set the disk info so that probing is triggered
581 bzero(&info, sizeof(struct disk_info));
582 info.d_media_blksize = vn->sc_secsize;
583 info.d_media_blocks = vn->sc_size;
585 * reserve mbr sector for backwards compatibility
586 * when no slices exist.
588 info.d_dsflags = DSO_COMPATMBR;
589 info.d_secpertrack = 32;
590 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
591 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
592 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
593 disk_setdiskinfo_sync(&vn->sc_disk, &info);
595 error = dev_dopen(dev, flag, S_IFCHR, cred);
600 kprintf("vnioctl: SET vp %p size %llx blks\n",
601 vn->sc_vp, (long long)vn->sc_size);
610 * Attach swap backing store to a VN partition of the size specified
615 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
616 int flag, struct ucred *cred)
619 struct disk_info info;
622 * Range check. Disallow negative sizes or any size less then the
623 * size of a page. Then round to a page.
626 if (vio->vn_size <= 0)
630 * Allocate an OBJT_SWAP object.
632 * sc_secsize is PAGE_SIZE'd
634 * vio->vn_size is in PAGE_SIZE'd chunks.
635 * sc_size must be in PAGE_SIZE'd chunks.
636 * Note the truncation.
639 vn->sc_secsize = PAGE_SIZE;
640 vn->sc_size = vio->vn_size;
641 vn->sc_object = swap_pager_alloc(NULL,
642 vn->sc_secsize * (off_t)vio->vn_size,
644 IFOPT(vn, VN_RESERVE) {
645 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
646 vm_pager_deallocate(vn->sc_object);
647 vn->sc_object = NULL;
651 vn->sc_flags |= VNF_INITED;
653 error = vnsetcred(vn, cred);
656 * Set the disk info so that probing is triggered
658 bzero(&info, sizeof(struct disk_info));
659 info.d_media_blksize = vn->sc_secsize;
660 info.d_media_blocks = vn->sc_size;
662 * reserve mbr sector for backwards compatibility
663 * when no slices exist.
665 info.d_dsflags = DSO_COMPATMBR;
666 info.d_secpertrack = 32;
667 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
668 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
669 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
670 disk_setdiskinfo_sync(&vn->sc_disk, &info);
672 error = dev_dopen(dev, flag, S_IFCHR, cred);
675 IFOPT(vn, VN_FOLLOW) {
676 kprintf("vnioctl: SET vp %p size %llx\n",
677 vn->sc_vp, (long long)vn->sc_size);
686 * Duplicate the current processes' credentials. Since we are called only
687 * as the result of a SET ioctl and only root can do that, any future access
688 * to this "disk" is essentially as root. Note that credentials may change
689 * if some other uid can write directly to the mapped file (NFS).
692 vnsetcred(struct vn_softc *vn, struct ucred *cred)
698 * Set credits in our softc
703 vn->sc_cred = crdup(cred);
706 * Horrible kludge to establish credentials for NFS XXX.
713 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
714 bzero(&auio, sizeof(auio));
716 aiov.iov_base = tmpbuf;
717 aiov.iov_len = vn->sc_secsize;
718 auio.uio_iov = &aiov;
721 auio.uio_rw = UIO_READ;
722 auio.uio_segflg = UIO_SYSSPACE;
723 auio.uio_resid = aiov.iov_len;
724 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
725 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
726 vn_unlock(vn->sc_vp);
727 kfree(tmpbuf, M_TEMP);
733 vnclear(struct vn_softc *vn)
736 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
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;
753 disk_unprobe(&vn->sc_disk);
761 * populate a struct vn_user for the VNIOCGET ioctl.
762 * interface conventions defined in sys/sys/vnioctl.h.
766 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu)
768 int error, found = 0;
769 char *freepath, *fullpath;
772 if (vnu->vnu_unit == -1) {
773 vnu->vnu_unit = dkunit(dev);
775 else if (vnu->vnu_unit < 0)
778 SLIST_FOREACH(vn, &vn_list, sc_list) {
780 if(vn->sc_unit != vnu->vnu_unit)
785 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) {
787 /* note: u_cred checked in vnioctl above */
788 error = VOP_GETATTR(vn->sc_vp, &vattr);
790 kprintf("vnget: VOP_GETATTR for %p failed\n",
795 error = vn_fullpath(curproc, vn->sc_vp,
796 &fullpath, &freepath, 0);
799 kprintf("vnget: unable to resolve vp %p\n",
804 strlcpy(vnu->vnu_file, fullpath,
805 sizeof(vnu->vnu_file));
806 kfree(freepath, M_TEMP);
807 vnu->vnu_dev = vattr.va_fsid;
808 vnu->vnu_ino = vattr.va_fileid;
811 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){
813 strlcpy(vnu->vnu_file, _VN_USER_SWAP,
814 sizeof(vnu->vnu_file));
815 vnu->vnu_size = vn->sc_size;
816 vnu->vnu_secsize = vn->sc_secsize;
820 bzero(vnu->vnu_file, sizeof(vnu->vnu_file));
835 vnsize(struct dev_psize_args *ap)
837 cdev_t dev = ap->a_head.a_dev;
843 if ((vn->sc_flags & VNF_INITED) == 0)
845 ap->a_result = (int64_t)vn->sc_size;
850 vn_create(int unit, struct devfs_bitmap *bitmap)
853 struct disk_info info;
857 dev = disk_create(unit, &vn->sc_disk, &vn_ops);
860 bzero(&info, sizeof(struct disk_info));
861 info.d_media_blksize = 512;
862 info.d_media_blocks = 0;
863 info.d_dsflags = DSO_MBRQUIET;
864 info.d_secpertrack = 32;
866 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
867 info.d_ncylinders = 0;
868 disk_setdiskinfo_sync(&vn->sc_disk, &info);
871 devfs_clone_bitmap_set(bitmap, unit);
877 vn_modevent(module_t mod, int type, void *data)
880 static cdev_t dev = NULL;
885 dev = make_autoclone_dev(&vn_ops, &DEVFS_CLONE_BITMAP(vn), vnclone, UID_ROOT,
886 GID_OPERATOR, 0640, "vn");
888 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) {
889 vn_create(i, &DEVFS_CLONE_BITMAP(vn));
893 destroy_autoclone_dev(dev, &DEVFS_CLONE_BITMAP(vn));
896 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
897 SLIST_REMOVE_HEAD(&vn_list, sc_list);
898 if (vn->sc_flags & VNF_INITED)
900 /* Cleanup all cdev_t's that refer to this unit */
901 disk_destroy(&vn->sc_disk);
902 while ((dev = vn->sc_devlist) != NULL) {
903 vn->sc_devlist = dev->si_drv2;
904 dev->si_drv1 = dev->si_drv2 = NULL;
909 dev_ops_remove_all(&vn_ops);
917 DEV_MODULE(vn, vn_modevent, 0);