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 $
47 * Block/character interface to a vnode. Allows one to treat a file
48 * as a disk (e.g. build a filesystem in it, mount it, etc.).
50 * NOTE 1: There is a security issue involved with this driver.
51 * Once mounted all access to the contents of the "mapped" file via
52 * the special file is controlled by the permissions on the special
53 * file, the protection of the mapped file is ignored (effectively,
54 * by using root credentials in all transactions).
56 * 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);
98 #define VN_PREALLOCATED_UNITS 4
100 #define VN_PREALLOCATED_UNITS NVN
103 #define VN_BSIZE_BEST 8192
107 * D_DISK we want to look like a disk
108 * D_CANFREE We support BUF_CMD_FREEBLKS
111 static struct dev_ops vn_ops = {
112 { "vn", 0, D_DISK | D_CANFREE },
116 .d_write = physwrite,
118 .d_strategy = vnstrategy,
124 int sc_flags; /* flags */
125 u_int64_t sc_size; /* size of vn, sc_secsize scale */
126 int sc_secsize; /* sector size */
128 struct vnode *sc_vp; /* vnode if not NULL */
129 vm_object_t sc_object; /* backing object if not NULL */
130 struct ucred *sc_cred; /* credentials */
131 int sc_maxactive; /* max # of active requests */
132 struct buf sc_tab; /* transfer queue */
133 u_long sc_options; /* options */
134 cdev_t sc_dev; /* devices that refer to this unit */
135 SLIST_ENTRY(vn_softc) sc_list;
138 static SLIST_HEAD(, vn_softc) vn_list;
141 #define VNF_INITED 0x01
142 #define VNF_READONLY 0x02
143 #define VNF_OPENED 0x10
144 #define VNF_DESTROY 0x20
146 static u_long vn_options;
148 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
149 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
151 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
152 static void vnclear (struct vn_softc *vn);
153 static int vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu);
154 static int vn_modevent (module_t, int, void *);
155 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
156 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
157 static cdev_t vn_create(int unit, struct devfs_bitmap *bitmap, int clone);
160 vnclone(struct dev_clone_args *ap)
164 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(vn), 0);
165 ap->a_dev = vn_create(unit, &DEVFS_CLONE_BITMAP(vn), 1);
171 vnclose(struct dev_close_args *ap)
173 cdev_t dev = ap->a_head.a_dev;
177 KKASSERT(vn != NULL);
179 vn->sc_flags &= ~VNF_OPENED;
181 /* The disk has been detached and can now be safely destroyed */
182 if (vn->sc_flags & VNF_DESTROY) {
183 KKASSERT(disk_getopencount(&vn->sc_disk) == 0);
184 disk_destroy(&vn->sc_disk);
185 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dkunit(dev));
186 SLIST_REMOVE(&vn_list, vn, vn_softc, sc_list);
192 static struct vn_softc *
197 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
202 vninitvn(struct vn_softc *vn, cdev_t dev)
206 KKASSERT(vn != NULL);
207 KKASSERT(dev != NULL);
214 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
218 vnopen(struct dev_open_args *ap)
220 cdev_t dev = ap->a_head.a_dev;
224 * Locate preexisting device
228 KKASSERT(vn != NULL);
231 * Update si_bsize fields for device. This data will be overriden by
232 * the slice/parition code for vn accesses through partitions, and
233 * used directly if you open the 'whole disk' device.
235 * si_bsize_best must be reinitialized in case VN has been
236 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
238 dev->si_bsize_phys = vn->sc_secsize;
239 dev->si_bsize_best = vn->sc_secsize;
240 if (dev->si_bsize_best < VN_BSIZE_BEST)
241 dev->si_bsize_best = VN_BSIZE_BEST;
243 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
247 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
248 devtoname(dev), ap->a_oflags, ap->a_devtype);
250 vn->sc_flags |= VNF_OPENED;
257 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
258 * for vnode-backed vn's, and the swap_pager_strategy() call for
259 * vm_object-backed vn's.
262 vnstrategy(struct dev_strategy_args *ap)
264 cdev_t dev = ap->a_head.a_dev;
265 struct bio *bio = ap->a_bio;
274 KKASSERT(vn != NULL);
279 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
281 if ((vn->sc_flags & VNF_INITED) == 0) {
283 bp->b_flags |= B_ERROR;
288 bp->b_resid = bp->b_bcount;
291 * The vnode device is using disk/slice label support.
293 * The dscheck() function is called for validating the
294 * slices that exist ON the vnode device itself, and
295 * translate the "slice-relative" block number, again.
296 * dscheck() will call biodone() and return NULL if
297 * we are at EOF or beyond the device size.
303 * Use the translated nbio from this point on
305 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
307 * Freeblks is not handled for vnode-backed elements yet.
310 /* operation complete */
311 } else if (vn->sc_vp) {
315 * If an error occurs, we set B_ERROR but we do not set
316 * B_INVAL because (for a write anyway), the buffer is
322 bzero(&auio, sizeof(auio));
324 aiov.iov_base = bp->b_data;
325 aiov.iov_len = bp->b_bcount;
326 auio.uio_iov = &aiov;
328 auio.uio_offset = nbio->bio_offset;
329 auio.uio_segflg = UIO_SYSSPACE;
330 if (bp->b_cmd == BUF_CMD_READ)
331 auio.uio_rw = UIO_READ;
333 auio.uio_rw = UIO_WRITE;
334 auio.uio_resid = bp->b_bcount;
335 auio.uio_td = curthread;
338 * Don't use IO_DIRECT here, it really gets in the way
339 * due to typical blocksize differences between the
340 * fs backing the VN device and whatever is running on
343 if (bp->b_cmd == BUF_CMD_READ) {
344 vn_lock(vn->sc_vp, LK_SHARED | LK_RETRY);
345 error = VOP_READ(vn->sc_vp, &auio, IO_RECURSE,
348 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
349 error = VOP_WRITE(vn->sc_vp, &auio, IO_RECURSE,
352 vn_unlock(vn->sc_vp);
353 bp->b_resid = auio.uio_resid;
356 bp->b_flags |= B_ERROR;
358 /* operation complete */
359 } else if (vn->sc_object) {
361 * OBJT_SWAP I/O (handles read, write, freebuf)
363 * We have nothing to do if freeing blocks on a reserved
364 * swap area, othrewise execute the op.
366 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
368 /* operation complete */
370 swap_pager_strategy(vn->sc_object, nbio);
375 bp->b_resid = bp->b_bcount;
376 bp->b_flags |= B_ERROR | B_INVAL;
377 bp->b_error = EINVAL;
378 /* operation complete */
386 vnioctl(struct dev_ioctl_args *ap)
388 cdev_t dev = ap->a_head.a_dev;
390 struct vn_ioctl *vio;
395 IFOPT(vn,VN_FOLLOW) {
396 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
397 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
413 if (dkslice(dev) != WHOLE_DISK_SLICE ||
414 dkpart(dev) != WHOLE_SLICE_PART)
420 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
424 vio = (struct vn_ioctl *)ap->a_data;
425 f = (u_long*)ap->a_data;
429 if (vn->sc_flags & VNF_INITED)
432 if (vn->sc_flags & VNF_DESTROY)
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 ((disk_getopencount(&vn->sc_disk)) > 1)
457 kprintf("vnioctl: CLRed\n");
459 if (dkunit(dev) >= VN_PREALLOCATED_UNITS) {
460 vn->sc_flags |= VNF_DESTROY;
466 error = vnget(dev, vn, (struct vn_user *) ap->a_data);
480 vn->sc_options |= *f;
485 vn->sc_options &= ~(*f);
499 * Attach a file to a VN partition. Return the size in the vn_size
504 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
505 int flag, struct ucred *cred)
508 struct nlookupdata nd;
511 struct disk_info info;
513 flags = FREAD|FWRITE;
514 error = nlookup_init(&nd, vio->vn_file,
515 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
518 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
519 if (error != EACCES && error != EPERM && error != EROFS)
523 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
526 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
530 if (vp->v_type != VREG ||
531 (error = VOP_GETATTR(vp, &vattr))) {
537 vn->sc_secsize = DEV_BSIZE;
539 nd.nl_open_vp = NULL;
542 * If the size is specified, override the file attributes. Note that
543 * the vn_size argument is in PAGE_SIZE sized blocks.
546 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
548 vn->sc_size = vattr.va_size / vn->sc_secsize;
549 error = vnsetcred(vn, cred);
555 vn->sc_flags |= VNF_INITED;
557 vn->sc_flags |= VNF_READONLY;
560 * Set the disk info so that probing is triggered
562 bzero(&info, sizeof(struct disk_info));
563 info.d_media_blksize = vn->sc_secsize;
564 info.d_media_blocks = vn->sc_size;
566 * reserve mbr sector for backwards compatibility
567 * when no slices exist.
569 info.d_dsflags = DSO_COMPATMBR;
570 info.d_secpertrack = 32;
571 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
572 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
573 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
574 disk_setdiskinfo_sync(&vn->sc_disk, &info);
576 error = dev_dopen(dev, flag, S_IFCHR, cred);
581 kprintf("vnioctl: SET vp %p size %llx blks\n",
582 vn->sc_vp, (long long)vn->sc_size);
591 * Attach swap backing store to a VN partition of the size specified
596 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
597 int flag, struct ucred *cred)
600 struct disk_info info;
603 * Range check. Disallow negative sizes or any size less then the
604 * size of a page. Then round to a page.
607 if (vio->vn_size <= 0)
611 * Allocate an OBJT_SWAP object.
613 * sc_secsize is PAGE_SIZE'd
615 * vio->vn_size is in PAGE_SIZE'd chunks.
616 * sc_size must be in PAGE_SIZE'd chunks.
617 * Note the truncation.
620 vn->sc_secsize = PAGE_SIZE;
621 vn->sc_size = vio->vn_size;
622 vn->sc_object = swap_pager_alloc(NULL,
623 vn->sc_secsize * (off_t)vio->vn_size,
625 IFOPT(vn, VN_RESERVE) {
626 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
627 vm_pager_deallocate(vn->sc_object);
628 vn->sc_object = NULL;
632 vn->sc_flags |= VNF_INITED;
634 error = vnsetcred(vn, cred);
637 * Set the disk info so that probing is triggered
639 bzero(&info, sizeof(struct disk_info));
640 info.d_media_blksize = vn->sc_secsize;
641 info.d_media_blocks = vn->sc_size;
643 * reserve mbr sector for backwards compatibility
644 * when no slices exist.
646 info.d_dsflags = DSO_COMPATMBR;
647 info.d_secpertrack = 32;
648 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
649 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
650 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
651 disk_setdiskinfo_sync(&vn->sc_disk, &info);
653 error = dev_dopen(dev, flag, S_IFCHR, cred);
656 IFOPT(vn, VN_FOLLOW) {
657 kprintf("vnioctl: SET vp %p size %llx\n",
658 vn->sc_vp, (long long)vn->sc_size);
667 * Duplicate the current processes' credentials. Since we are called only
668 * as the result of a SET ioctl and only root can do that, any future access
669 * to this "disk" is essentially as root. Note that credentials may change
670 * if some other uid can write directly to the mapped file (NFS).
673 vnsetcred(struct vn_softc *vn, struct ucred *cred)
679 * Set credits in our softc
684 vn->sc_cred = crdup(cred);
687 * Horrible kludge to establish credentials for NFS XXX.
694 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
695 bzero(&auio, sizeof(auio));
697 aiov.iov_base = tmpbuf;
698 aiov.iov_len = vn->sc_secsize;
699 auio.uio_iov = &aiov;
702 auio.uio_rw = UIO_READ;
703 auio.uio_segflg = UIO_SYSSPACE;
704 auio.uio_resid = aiov.iov_len;
705 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
706 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
707 vn_unlock(vn->sc_vp);
708 kfree(tmpbuf, M_TEMP);
714 vnclear(struct vn_softc *vn)
717 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
718 vn->sc_flags &= ~VNF_INITED;
719 if (vn->sc_vp != NULL) {
721 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
724 vn->sc_flags &= ~VNF_READONLY;
729 if (vn->sc_object != NULL) {
730 vm_pager_deallocate(vn->sc_object);
731 vn->sc_object = NULL;
734 disk_unprobe(&vn->sc_disk);
742 * populate a struct vn_user for the VNIOCGET ioctl.
743 * interface conventions defined in sys/sys/vnioctl.h.
747 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu)
749 int error, found = 0;
750 char *freepath, *fullpath;
753 if (vnu->vnu_unit == -1) {
754 vnu->vnu_unit = dkunit(dev);
756 else if (vnu->vnu_unit < 0)
759 SLIST_FOREACH(vn, &vn_list, sc_list) {
761 if(vn->sc_unit != vnu->vnu_unit)
766 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) {
768 /* note: u_cred checked in vnioctl above */
769 error = VOP_GETATTR(vn->sc_vp, &vattr);
771 kprintf("vnget: VOP_GETATTR for %p failed\n",
776 error = vn_fullpath(curproc, vn->sc_vp,
777 &fullpath, &freepath, 0);
780 kprintf("vnget: unable to resolve vp %p\n",
785 strlcpy(vnu->vnu_file, fullpath,
786 sizeof(vnu->vnu_file));
787 kfree(freepath, M_TEMP);
788 vnu->vnu_dev = vattr.va_fsid;
789 vnu->vnu_ino = vattr.va_fileid;
792 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){
794 strlcpy(vnu->vnu_file, _VN_USER_SWAP,
795 sizeof(vnu->vnu_file));
796 vnu->vnu_size = vn->sc_size;
797 vnu->vnu_secsize = vn->sc_secsize;
801 bzero(vnu->vnu_file, sizeof(vnu->vnu_file));
816 vnsize(struct dev_psize_args *ap)
818 cdev_t dev = ap->a_head.a_dev;
824 if ((vn->sc_flags & VNF_INITED) == 0)
826 ap->a_result = (int64_t)vn->sc_size;
831 vn_create(int unit, struct devfs_bitmap *bitmap, int clone)
834 struct disk_info info;
840 * For clone devices we need to return the top-level cdev,
841 * not the raw dev we'd normally work with.
843 dev = disk_create_clone(unit, &vn->sc_disk, &vn_ops);
844 ret_dev = vn->sc_disk.d_cdev;
846 ret_dev = dev = disk_create(unit, &vn->sc_disk, &vn_ops);
850 bzero(&info, sizeof(struct disk_info));
851 info.d_media_blksize = 512;
852 info.d_media_blocks = 0;
853 info.d_dsflags = DSO_MBRQUIET;
854 info.d_secpertrack = 32;
856 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
857 info.d_ncylinders = 0;
858 disk_setdiskinfo_sync(&vn->sc_disk, &info);
861 devfs_clone_bitmap_set(bitmap, unit);
867 vn_modevent(module_t mod, int type, void *data)
870 static cdev_t dev = NULL;
875 dev = make_autoclone_dev(&vn_ops, &DEVFS_CLONE_BITMAP(vn), vnclone, UID_ROOT,
876 GID_OPERATOR, 0640, "vn");
878 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) {
879 vn_create(i, &DEVFS_CLONE_BITMAP(vn), 0);
885 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
887 * XXX: no idea if we can return EBUSY even in the
888 * shutdown case, so err on the side of caution
889 * and just rip stuff out on shutdown.
891 if (type != MOD_SHUTDOWN) {
892 if (vn->sc_flags & VNF_OPENED)
896 disk_destroy(&vn->sc_disk);
898 SLIST_REMOVE_HEAD(&vn_list, sc_list);
900 if (vn->sc_flags & VNF_INITED)
905 destroy_autoclone_dev(dev, &DEVFS_CLONE_BITMAP(vn));
906 dev_ops_remove_all(&vn_ops);
914 DEV_MODULE(vn, vn_modevent, 0);