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;
94 static d_clone_t vnclone;
96 DEVFS_DECLARE_CLONE_BITMAP(vn);
97 #define VN_PREALLOCATED_UNITS 4
101 #define VN_BSIZE_BEST 8192
105 * D_DISK we want to look like a disk
106 * D_CANFREE We support BUF_CMD_FREEBLKS
109 static struct dev_ops vn_ops = {
110 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
114 .d_write = physwrite,
116 .d_strategy = vnstrategy,
122 int sc_flags; /* flags */
123 u_int64_t sc_size; /* size of vn, sc_secsize scale */
124 int sc_secsize; /* sector size */
126 struct vnode *sc_vp; /* vnode if not NULL */
127 vm_object_t sc_object; /* backing object if not NULL */
128 struct ucred *sc_cred; /* credentials */
129 int sc_maxactive; /* max # of active requests */
130 struct buf sc_tab; /* transfer queue */
131 u_long sc_options; /* options */
132 cdev_t sc_devlist; /* devices that refer to this unit */
133 SLIST_ENTRY(vn_softc) sc_list;
136 static SLIST_HEAD(, vn_softc) vn_list;
139 #define VNF_INITED 0x01
140 #define VNF_READONLY 0x02
142 static u_long vn_options;
144 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
145 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
147 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
148 static void vnclear (struct vn_softc *vn);
149 static int vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu);
150 static int vn_modevent (module_t, int, void *);
151 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
152 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
155 vnclose(struct dev_close_args *ap)
158 cdev_t dev = ap->a_head.a_dev;
159 struct vn_softc *vn = dev->si_drv1;
160 if (dev->si_uminor >= VN_PREALLOCATED_UNITS) {
161 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dev->si_uminor);
168 static struct vn_softc *
173 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
178 vninitvn(struct vn_softc *vn, cdev_t dev)
182 KKASSERT(vn != NULL);
183 KKASSERT(dev != NULL);
188 vn->sc_devlist = dev;
189 if (vn->sc_devlist->si_drv1 == NULL) {
190 reference_dev(vn->sc_devlist);
191 vn->sc_devlist->si_drv1 = vn;
192 vn->sc_devlist->si_drv2 = NULL;
194 if (vn->sc_devlist != dev) {
196 dev->si_drv2 = vn->sc_devlist;
197 vn->sc_devlist = dev;
200 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
204 * Called only when si_drv1 is NULL. Locate the associated vn node and
205 * attach the device to it.
207 static struct vn_softc *
214 SLIST_FOREACH(vn, &vn_list, sc_list) {
215 if (vn->sc_unit == unit) {
217 dev->si_drv2 = vn->sc_devlist;
218 vn->sc_devlist = dev;
224 KKASSERT(vn != NULL);
230 vnopen(struct dev_open_args *ap)
232 cdev_t dev = ap->a_head.a_dev;
236 * Locate preexisting device
239 if ((vn = dev->si_drv1) == NULL)
243 * Update si_bsize fields for device. This data will be overriden by
244 * the slice/parition code for vn accesses through partitions, and
245 * used directly if you open the 'whole disk' device.
247 * si_bsize_best must be reinitialized in case VN has been
248 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
250 dev->si_bsize_phys = vn->sc_secsize;
251 dev->si_bsize_best = vn->sc_secsize;
252 if (dev->si_bsize_best < VN_BSIZE_BEST)
253 dev->si_bsize_best = VN_BSIZE_BEST;
255 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
259 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
260 devtoname(dev), ap->a_oflags, ap->a_devtype);
268 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
269 * for vnode-backed vn's, and the new vm_pager_strategy() call for
270 * vm_object-backed vn's.
273 vnstrategy(struct dev_strategy_args *ap)
275 cdev_t dev = ap->a_head.a_dev;
276 struct bio *bio = ap->a_bio;
284 if ((vn = dev->si_drv1) == NULL)
290 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
292 if ((vn->sc_flags & VNF_INITED) == 0) {
294 bp->b_flags |= B_ERROR;
299 bp->b_resid = bp->b_bcount;
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.
314 * Use the translated nbio from this point on
316 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
318 * Freeblks is not handled for vnode-backed elements yet.
321 /* operation complete */
322 } else if (vn->sc_vp) {
326 * If an error occurs, we set B_ERROR but we do not set
327 * B_INVAL because (for a write anyway), the buffer is
333 bzero(&auio, sizeof(auio));
335 aiov.iov_base = bp->b_data;
336 aiov.iov_len = bp->b_bcount;
337 auio.uio_iov = &aiov;
339 auio.uio_offset = nbio->bio_offset;
340 auio.uio_segflg = UIO_SYSSPACE;
341 if (bp->b_cmd == BUF_CMD_READ)
342 auio.uio_rw = UIO_READ;
344 auio.uio_rw = UIO_WRITE;
345 auio.uio_resid = bp->b_bcount;
346 auio.uio_td = curthread;
347 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
348 if (bp->b_cmd == BUF_CMD_READ)
349 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred);
351 error = VOP_WRITE(vn->sc_vp, &auio, IO_DIRECT | IO_RECURSE, vn->sc_cred);
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 vm_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 (vio->vn_file == NULL)
433 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
435 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
439 if ((vn->sc_flags & VNF_INITED) == 0)
442 * XXX handle i/o in progress. Return EBUSY, or wait, or
444 * XXX handle multiple opens of the device. Return EBUSY,
445 * or revoke the fd's.
446 * How are these problems handled for removable and failing
447 * hardware devices? (Hint: They are not)
449 if (count_dev(vn->sc_devlist) > 1)
454 kprintf("vnioctl: CLRed\n");
458 error = vnget(dev, vn, (struct vn_user *) ap->a_data);
472 vn->sc_options |= *f;
477 vn->sc_options &= ~(*f);
491 * Attach a file to a VN partition. Return the size in the vn_size
496 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
497 int flag, struct ucred *cred)
500 struct nlookupdata nd;
503 struct disk_info info;
505 flags = FREAD|FWRITE;
506 error = nlookup_init(&nd, vio->vn_file,
507 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
510 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
511 if (error != EACCES && error != EPERM && error != EROFS)
515 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
518 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
522 if (vp->v_type != VREG ||
523 (error = VOP_GETATTR(vp, &vattr))) {
529 vn->sc_secsize = DEV_BSIZE;
531 nd.nl_open_vp = NULL;
534 * If the size is specified, override the file attributes. Note that
535 * the vn_size argument is in PAGE_SIZE sized blocks.
538 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
540 vn->sc_size = vattr.va_size / vn->sc_secsize;
541 error = vnsetcred(vn, cred);
547 vn->sc_flags |= VNF_INITED;
549 vn->sc_flags |= VNF_READONLY;
552 * Set the disk info so that probing is triggered
554 bzero(&info, sizeof(struct disk_info));
555 info.d_media_blksize = vn->sc_secsize;
556 info.d_media_blocks = vn->sc_size;
558 * reserve mbr sector for backwards compatibility
559 * when no slices exist.
561 info.d_dsflags = DSO_COMPATMBR;
562 info.d_secpertrack = 32;
563 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
564 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
565 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
566 disk_setdiskinfo_sync(&vn->sc_disk, &info);
568 error = dev_dopen(dev, flag, S_IFCHR, cred);
573 kprintf("vnioctl: SET vp %p size %llx blks\n",
574 vn->sc_vp, (long long)vn->sc_size);
583 * Attach swap backing store to a VN partition of the size specified
588 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
589 int flag, struct ucred *cred)
592 struct disk_info info;
595 * Range check. Disallow negative sizes or any size less then the
596 * size of a page. Then round to a page.
599 if (vio->vn_size <= 0)
603 * Allocate an OBJT_SWAP object.
605 * sc_secsize is PAGE_SIZE'd
607 * vio->vn_size is in PAGE_SIZE'd chunks.
608 * sc_size must be in PAGE_SIZE'd chunks.
609 * Note the truncation.
612 vn->sc_secsize = PAGE_SIZE;
613 vn->sc_size = vio->vn_size;
614 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
615 vn->sc_secsize * (off_t)vio->vn_size,
617 IFOPT(vn, VN_RESERVE) {
618 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
619 vm_pager_deallocate(vn->sc_object);
620 vn->sc_object = NULL;
624 vn->sc_flags |= VNF_INITED;
626 error = vnsetcred(vn, cred);
629 * Set the disk info so that probing is triggered
631 bzero(&info, sizeof(struct disk_info));
632 info.d_media_blksize = vn->sc_secsize;
633 info.d_media_blocks = vn->sc_size;
635 * reserve mbr sector for backwards compatibility
636 * when no slices exist.
638 info.d_dsflags = DSO_COMPATMBR;
639 info.d_secpertrack = 32;
640 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
641 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
642 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
643 disk_setdiskinfo_sync(&vn->sc_disk, &info);
645 error = dev_dopen(dev, flag, S_IFCHR, cred);
648 IFOPT(vn, VN_FOLLOW) {
649 kprintf("vnioctl: SET vp %p size %llx\n",
650 vn->sc_vp, (long long)vn->sc_size);
659 * Duplicate the current processes' credentials. Since we are called only
660 * as the result of a SET ioctl and only root can do that, any future access
661 * to this "disk" is essentially as root. Note that credentials may change
662 * if some other uid can write directly to the mapped file (NFS).
665 vnsetcred(struct vn_softc *vn, struct ucred *cred)
671 * Set credits in our softc
676 vn->sc_cred = crdup(cred);
679 * Horrible kludge to establish credentials for NFS XXX.
686 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
687 bzero(&auio, sizeof(auio));
689 aiov.iov_base = tmpbuf;
690 aiov.iov_len = vn->sc_secsize;
691 auio.uio_iov = &aiov;
694 auio.uio_rw = UIO_READ;
695 auio.uio_segflg = UIO_SYSSPACE;
696 auio.uio_resid = aiov.iov_len;
697 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
698 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
699 vn_unlock(vn->sc_vp);
700 kfree(tmpbuf, M_TEMP);
706 vnclear(struct vn_softc *vn)
709 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
710 vn->sc_flags &= ~VNF_INITED;
711 if (vn->sc_vp != NULL) {
713 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
716 vn->sc_flags &= ~VNF_READONLY;
721 if (vn->sc_object != NULL) {
722 vm_pager_deallocate(vn->sc_object);
723 vn->sc_object = NULL;
726 disk_unprobe(&vn->sc_disk);
734 * populate a struct vn_user for the VNIOCGET ioctl.
735 * interface conventions defined in sys/sys/vnioctl.h.
739 vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu)
741 int error, found = 0;
742 char *freepath, *fullpath;
745 if (vnu->vnu_unit == -1) {
746 vnu->vnu_unit = dkunit(dev);
748 else if (vnu->vnu_unit < 0)
751 SLIST_FOREACH(vn, &vn_list, sc_list) {
753 if(vn->sc_unit != vnu->vnu_unit)
758 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) {
760 /* note: u_cred checked in vnioctl above */
761 error = VOP_GETATTR(vn->sc_vp, &vattr);
763 kprintf("vnget: VOP_GETATTR for %p failed\n",
768 error = vn_fullpath(curproc, vn->sc_vp,
769 &fullpath, &freepath);
772 kprintf("vnget: unable to resolve vp %p\n",
777 strlcpy(vnu->vnu_file, fullpath,
778 sizeof(vnu->vnu_file));
779 kfree(freepath, M_TEMP);
780 vnu->vnu_dev = vattr.va_fsid;
781 vnu->vnu_ino = vattr.va_fileid;
784 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){
786 strlcpy(vnu->vnu_file, _VN_USER_SWAP,
787 sizeof(vnu->vnu_file));
788 vnu->vnu_size = vn->sc_size;
789 vnu->vnu_secsize = vn->sc_secsize;
793 bzero(vnu->vnu_file, sizeof(vnu->vnu_file));
808 vnsize(struct dev_psize_args *ap)
810 cdev_t dev = ap->a_head.a_dev;
816 if ((vn->sc_flags & VNF_INITED) == 0)
818 ap->a_result = (int64_t)vn->sc_size;
824 vnclone(struct dev_clone_args *ap)
828 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(vn), 0);
829 ap->a_dev = make_only_dev(&vn_ops, unit, UID_ROOT, GID_OPERATOR, 0600,
837 vn_modevent(module_t mod, int type, void *data)
840 struct disk_info info;
847 make_dev(&vn_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "vn");
849 devfs_clone_bitmap_init(&DEVFS_CLONE_BITMAP(vn));
850 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) {
852 dev = disk_create(i, &vn->sc_disk, &vn_ops);
855 bzero(&info, sizeof(struct disk_info));
856 info.d_media_blksize = 512;
857 info.d_media_blocks = 0;
858 info.d_dsflags = DSO_MBRQUIET;
859 info.d_secpertrack = 32;
861 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
862 info.d_ncylinders = 0;
863 disk_setdiskinfo_sync(&vn->sc_disk, &info);
865 devfs_clone_bitmap_set(&DEVFS_CLONE_BITMAP(vn), i);
868 devfs_clone_handler_add("vn", vnclone);
872 devfs_clone_bitmap_uninit(&DEVFS_CLONE_BITMAP(vn));
874 devfs_clone_handler_del("vn");
878 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
879 SLIST_REMOVE_HEAD(&vn_list, sc_list);
880 if (vn->sc_flags & VNF_INITED)
882 /* Cleanup all cdev_t's that refer to this unit */
883 while ((dev = vn->sc_devlist) != NULL) {
884 vn->sc_devlist = dev->si_drv2;
885 dev->si_drv1 = dev->si_drv2 = NULL;
890 dev_ops_remove_all(&vn_ops);
898 DEV_MODULE(vn, vn_modevent, 0);