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 1.24 2006/05/06 02:43:03 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>
64 #include <sys/nlookup.h>
66 #include <sys/malloc.h>
67 #include <sys/mount.h>
68 #include <sys/vnode.h>
69 #include <sys/fcntl.h>
71 #include <sys/disklabel.h>
72 #include <sys/diskslice.h>
74 #include <sys/module.h>
75 #include <sys/vnioctl.h>
78 #include <vm/vm_object.h>
79 #include <vm/vm_page.h>
80 #include <vm/vm_pager.h>
81 #include <vm/vm_pageout.h>
82 #include <vm/swap_pager.h>
83 #include <vm/vm_extern.h>
84 #include <vm/vm_zone.h>
86 static d_ioctl_t vnioctl;
87 static d_open_t vnopen;
88 static d_close_t vnclose;
89 static d_psize_t vnsize;
90 static d_strategy_t vnstrategy;
94 #define VN_BSIZE_BEST 8192
98 * D_DISK we want to look like a disk
99 * D_CANFREE We support BUF_CMD_FREEBLKS
102 static struct cdevsw vn_cdevsw = {
104 /* maj */ CDEV_MAJOR,
105 /* flags */ D_DISK|D_CANFREE,
112 /* write */ physwrite,
116 /* strategy */ vnstrategy,
123 int sc_flags; /* flags */
124 int sc_size; /* size of vn, sc_secsize scale */
125 int sc_secsize; /* sector size */
126 struct diskslices *sc_slices;
127 struct vnode *sc_vp; /* vnode if not NULL */
128 vm_object_t sc_object; /* backing object if not NULL */
129 struct ucred *sc_cred; /* credentials */
130 int sc_maxactive; /* max # of active requests */
131 struct buf sc_tab; /* transfer queue */
132 u_long sc_options; /* options */
133 dev_t sc_devlist; /* devices that refer to this unit */
134 SLIST_ENTRY(vn_softc) sc_list;
137 static SLIST_HEAD(, vn_softc) vn_list;
140 #define VNF_INITED 0x01
141 #define VNF_READONLY 0x02
143 static u_long vn_options;
145 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
146 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
148 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
149 static void vnclear (struct vn_softc *vn);
150 static int vn_modevent (module_t, int, void *);
151 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p);
152 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p);
155 vnclose(dev_t dev, int flags, int mode, struct thread *td)
157 struct vn_softc *vn = dev->si_drv1;
160 if (vn->sc_slices != NULL)
161 dsclose(dev, mode, vn->sc_slices);
166 * Called only when si_drv1 is NULL. Locate the associated vn node and
167 * attach the device to it.
169 static struct vn_softc *
176 SLIST_FOREACH(vn, &vn_list, sc_list) {
177 if (vn->sc_unit == unit) {
179 dev->si_drv2 = vn->sc_devlist;
180 vn->sc_devlist = dev;
186 vn = malloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
189 vn->sc_devlist = make_dev(&vn_cdevsw, 0, UID_ROOT,
190 GID_OPERATOR, 0640, "vn%d", unit);
191 if (vn->sc_devlist->si_drv1 == NULL) {
192 reference_dev(vn->sc_devlist);
193 vn->sc_devlist->si_drv1 = vn;
194 vn->sc_devlist->si_drv2 = NULL;
196 if (vn->sc_devlist != dev) {
198 dev->si_drv2 = vn->sc_devlist;
199 vn->sc_devlist = dev;
202 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
208 vnopen(dev_t dev, int flags, int mode, struct thread *td)
213 * Locate preexisting device
216 if ((vn = dev->si_drv1) == NULL)
220 * Update si_bsize fields for device. This data will be overriden by
221 * the slice/parition code for vn accesses through partitions, and
222 * used directly if you open the 'whole disk' device.
224 * si_bsize_best must be reinitialized in case VN has been
225 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
227 dev->si_bsize_phys = vn->sc_secsize;
228 dev->si_bsize_best = vn->sc_secsize;
229 if (dev->si_bsize_best < VN_BSIZE_BEST)
230 dev->si_bsize_best = VN_BSIZE_BEST;
232 if ((flags & FWRITE) && (vn->sc_flags & VNF_READONLY))
236 printf("vnopen(%s, 0x%x, 0x%x, %p)\n",
237 devtoname(dev), flags, mode, (void *)td);
243 IFOPT(vn, VN_LABELS) {
244 if (vn->sc_flags & VNF_INITED) {
245 struct disklabel label;
247 /* Build label for whole disk. */
248 bzero(&label, sizeof label);
249 label.d_secsize = vn->sc_secsize;
250 label.d_nsectors = 32;
251 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE);
252 label.d_secpercyl = label.d_nsectors * label.d_ntracks;
253 label.d_ncylinders = vn->sc_size / label.d_secpercyl;
254 label.d_secperunit = vn->sc_size;
255 label.d_partitions[RAW_PART].p_size = vn->sc_size;
257 return (dsopen(dev, mode, 0, &vn->sc_slices, &label));
259 if (dkslice(dev) != WHOLE_DISK_SLICE ||
260 dkpart(dev) != RAW_PART ||
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.
275 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
278 vnstrategy(dev_t dev, struct bio *bio)
287 if ((vn = dev->si_drv1) == NULL)
293 printf("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;
304 IFOPT(vn, VN_LABELS) {
306 * The vnode device is using disk/slice label support.
308 * The dscheck() function is called for validating the
309 * slices that exist ON the vnode device itself, and
310 * translate the "slice-relative" block number, again.
311 * dscheck() will call biodone() and return NULL if
312 * we are at EOF or beyond the device size.
314 if (vn->sc_slices == NULL) {
316 } else if ((nbio = dscheck(dev, bio, vn->sc_slices)) == NULL) {
320 int pbn; /* in sc_secsize chunks */
321 long sz; /* in sc_secsize chunks */
324 * Check for required alignment. Transfers must be a valid
325 * multiple of the sector size.
327 if (bp->b_bcount % vn->sc_secsize != 0 ||
328 bio->bio_offset % vn->sc_secsize != 0) {
332 pbn = bio->bio_offset / vn->sc_secsize;
333 sz = howmany(bp->b_bcount, vn->sc_secsize);
336 * Check for an illegal pbn or EOF truncation
340 if (pbn + sz > vn->sc_size) {
341 if (pbn > vn->sc_size || (bp->b_flags & B_BNOCLIP))
343 if (pbn == vn->sc_size) {
344 bp->b_resid = bp->b_bcount;
345 bp->b_flags |= B_INVAL;
348 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
350 nbio = push_bio(bio);
351 nbio->bio_offset = pbn * vn->sc_secsize;
355 * Use the translated nbio from this point on
357 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
359 * Freeblks is not handled for vnode-backed elements yet.
362 /* operation complete */
363 } else if (vn->sc_vp) {
367 * If an error occurs, we set B_ERROR but we do not set
368 * B_INVAL because (for a write anyway), the buffer is
374 bzero(&auio, sizeof(auio));
376 aiov.iov_base = bp->b_data;
377 aiov.iov_len = bp->b_bcount;
378 auio.uio_iov = &aiov;
380 auio.uio_offset = nbio->bio_offset;
381 auio.uio_segflg = UIO_SYSSPACE;
382 if (bp->b_cmd == BUF_CMD_READ)
383 auio.uio_rw = UIO_READ;
385 auio.uio_rw = UIO_WRITE;
386 auio.uio_resid = bp->b_bcount;
387 auio.uio_td = curthread;
388 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
389 if (bp->b_cmd == BUF_CMD_READ)
390 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
392 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
393 VOP_UNLOCK(vn->sc_vp, 0);
394 bp->b_resid = auio.uio_resid;
397 bp->b_flags |= B_ERROR;
399 /* operation complete */
400 } else if (vn->sc_object) {
402 * OBJT_SWAP I/O (handles read, write, freebuf)
404 * We have nothing to do if freeing blocks on a reserved
405 * swap area, othrewise execute the op.
407 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
409 /* operation complete */
411 vm_pager_strategy(vn->sc_object, nbio);
416 bp->b_resid = bp->b_bcount;
417 bp->b_flags |= B_ERROR | B_INVAL;
418 bp->b_error = EINVAL;
419 /* operation complete */
425 * Shortcuts / check failures on the original bio (not nbio).
428 bp->b_error = EINVAL;
429 bp->b_flags |= B_ERROR | B_INVAL;
436 vnioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct thread *td)
439 struct vn_ioctl *vio;
445 printf("vnioctl(%s, 0x%lx, %p, 0x%x, %p): unit %d\n",
446 devtoname(dev), cmd, (void *)data, flag, (void *)td,
459 IFOPT(vn,VN_LABELS) {
460 if (vn->sc_slices != NULL) {
461 error = dsioctl(dev, cmd, data, flag, &vn->sc_slices);
462 if (error != ENOIOCTL)
465 if (dkslice(dev) != WHOLE_DISK_SLICE ||
466 dkpart(dev) != RAW_PART)
476 vio = (struct vn_ioctl *)data;
481 if (vn->sc_flags & VNF_INITED)
484 if (vio->vn_file == NULL)
485 error = vniocattach_swap(vn, vio, dev, flag, td);
487 error = vniocattach_file(vn, vio, dev, flag, td);
491 if ((vn->sc_flags & VNF_INITED) == 0)
494 * XXX handle i/o in progress. Return EBUSY, or wait, or
496 * XXX handle multiple opens of the device. Return EBUSY,
497 * or revoke the fd's.
498 * How are these problems handled for removable and failing
499 * hardware devices? (Hint: They are not)
503 printf("vnioctl: CLRed\n");
517 vn->sc_options |= *f;
522 vn->sc_options &= ~(*f);
536 * Attach a file to a VN partition. Return the size in the vn_size
541 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, dev_t dev,
542 int flag, struct thread *td)
545 struct nlookupdata nd;
548 struct proc *p = td->td_proc;
552 flags = FREAD|FWRITE;
553 error = nlookup_init(&nd, vio->vn_file,
554 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
557 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
558 if (error != EACCES && error != EPERM && error != EROFS)
562 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
565 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
569 if (vp->v_type != VREG ||
570 (error = VOP_GETATTR(vp, &vattr))) {
576 vn->sc_secsize = DEV_BSIZE;
578 nd.nl_open_vp = NULL;
581 * If the size is specified, override the file attributes. Note that
582 * the vn_size argument is in PAGE_SIZE sized blocks.
585 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
587 vn->sc_size = vattr.va_size / vn->sc_secsize;
588 error = vnsetcred(vn, p->p_ucred);
594 vn->sc_flags |= VNF_INITED;
596 vn->sc_flags |= VNF_READONLY;
597 IFOPT(vn, VN_LABELS) {
599 * Reopen so that `ds' knows which devices are open.
600 * If this is the first VNIOCSET, then we've
601 * guaranteed that the device is the cdev and that
602 * no other slices or labels are open. Otherwise,
603 * we rely on VNIOCCLR not being abused.
605 error = vnopen(dev, flag, S_IFCHR, td);
610 printf("vnioctl: SET vp %p size %x blks\n",
611 vn->sc_vp, vn->sc_size);
620 * Attach swap backing store to a VN partition of the size specified
625 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, dev_t dev,
626 int flag, struct thread *td)
629 struct proc *p = td->td_proc;
633 * Range check. Disallow negative sizes or any size less then the
634 * size of a page. Then round to a page.
637 if (vio->vn_size <= 0)
641 * Allocate an OBJT_SWAP object.
643 * sc_secsize is PAGE_SIZE'd
645 * vio->vn_size is in PAGE_SIZE'd chunks.
646 * sc_size must be in PAGE_SIZE'd chunks.
647 * Note the truncation.
650 vn->sc_secsize = PAGE_SIZE;
651 vn->sc_size = vio->vn_size;
652 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
653 vn->sc_secsize * (off_t)vio->vn_size,
655 IFOPT(vn, VN_RESERVE) {
656 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
657 vm_pager_deallocate(vn->sc_object);
658 vn->sc_object = NULL;
662 vn->sc_flags |= VNF_INITED;
664 error = vnsetcred(vn, p->p_ucred);
666 IFOPT(vn, VN_LABELS) {
668 * Reopen so that `ds' knows which devices are open.
669 * If this is the first VNIOCSET, then we've
670 * guaranteed that the device is the cdev and that
671 * no other slices or labels are open. Otherwise,
672 * we rely on VNIOCCLR not being abused.
674 error = vnopen(dev, flag, S_IFCHR, td);
678 IFOPT(vn, VN_FOLLOW) {
679 printf("vnioctl: SET vp %p size %x\n",
680 vn->sc_vp, vn->sc_size);
689 * Duplicate the current processes' credentials. Since we are called only
690 * as the result of a SET ioctl and only root can do that, any future access
691 * to this "disk" is essentially as root. Note that credentials may change
692 * if some other uid can write directly to the mapped file (NFS).
695 vnsetcred(struct vn_softc *vn, struct ucred *cred)
701 * Set credits in our softc
706 vn->sc_cred = crdup(cred);
709 * Horrible kludge to establish credentials for NFS XXX.
716 tmpbuf = malloc(vn->sc_secsize, M_TEMP, M_WAITOK);
717 bzero(&auio, sizeof(auio));
719 aiov.iov_base = tmpbuf;
720 aiov.iov_len = vn->sc_secsize;
721 auio.uio_iov = &aiov;
724 auio.uio_rw = UIO_READ;
725 auio.uio_segflg = UIO_SYSSPACE;
726 auio.uio_resid = aiov.iov_len;
727 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
728 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
729 VOP_UNLOCK(vn->sc_vp, 0);
730 free(tmpbuf, M_TEMP);
736 vnclear(struct vn_softc *vn)
739 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
740 if (vn->sc_slices != NULL)
741 dsgone(&vn->sc_slices);
742 vn->sc_flags &= ~VNF_INITED;
743 if (vn->sc_vp != NULL) {
745 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
748 vn->sc_flags &= ~VNF_READONLY;
753 if (vn->sc_object != NULL) {
754 vm_pager_deallocate(vn->sc_object);
755 vn->sc_object = NULL;
768 if ((vn->sc_flags & VNF_INITED) == 0)
775 vn_modevent(module_t mod, int type, void *data)
782 cdevsw_add(&vn_cdevsw, 0, 0);
787 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
788 SLIST_REMOVE_HEAD(&vn_list, sc_list);
789 if (vn->sc_flags & VNF_INITED)
791 /* Cleanup all dev_t's that refer to this unit */
792 while ((dev = vn->sc_devlist) != NULL) {
793 vn->sc_devlist = dev->si_drv2;
794 dev->si_drv1 = dev->si_drv2 = NULL;
799 cdevsw_remove(&vn_cdevsw, 0, 0);
807 DEV_MODULE(vn, vn_modevent, 0);