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: This uses the VOP_BMAP/VOP_STRATEGY interface to the vnode
51 * instead of a simple VOP_RDWR. We do this to avoid distorting the
54 * NOTE 2: There is a security issue involved with this driver.
55 * Once mounted all access to the contents of the "mapped" file via
56 * the special file is controlled by the permissions on the special
57 * file, the protection of the mapped file is ignored (effectively,
58 * by using root credentials in all transactions).
60 * NOTE 3: Doesn't interact with leases, should it?
63 #include <sys/param.h>
64 #include <sys/systm.h>
65 #include <sys/kernel.h>
66 #include <sys/namei.h>
69 #include <sys/malloc.h>
70 #include <sys/mount.h>
71 #include <sys/vnode.h>
72 #include <sys/fcntl.h>
74 #include <sys/disklabel.h>
75 #include <sys/diskslice.h>
78 #include <sys/module.h>
79 #include <sys/vnioctl.h>
82 #include <vm/vm_object.h>
83 #include <vm/vm_page.h>
84 #include <vm/vm_pager.h>
85 #include <vm/vm_pageout.h>
86 #include <vm/swap_pager.h>
87 #include <vm/vm_extern.h>
88 #include <vm/vm_zone.h>
90 static d_ioctl_t vnioctl;
91 static d_open_t vnopen;
92 static d_close_t vnclose;
93 static d_psize_t vnsize;
94 static d_strategy_t vnstrategy;
99 #define VN_BSIZE_BEST 8192
103 * D_DISK we want to look like a disk
104 * D_CANFREE We support B_FREEBUF
107 static struct cdevsw vn_cdevsw = {
111 /* write */ physwrite,
115 /* strategy */ vnstrategy,
117 /* maj */ CDEV_MAJOR,
120 /* flags */ D_DISK|D_CANFREE,
121 /* bmaj */ BDEV_MAJOR
125 ((struct buf *)malloc(sizeof(struct buf), M_DEVBUF, M_WAITOK))
127 #define putvnbuf(bp) \
128 free((caddr_t)(bp), M_DEVBUF)
132 int sc_flags; /* flags */
133 int sc_size; /* size of vn, sc_secsize scale */
134 int sc_secsize; /* sector size */
135 struct diskslices *sc_slices;
136 struct vnode *sc_vp; /* vnode if not NULL */
137 vm_object_t sc_object; /* backing object if not NULL */
138 struct ucred *sc_cred; /* credentials */
139 int sc_maxactive; /* max # of active requests */
140 struct buf sc_tab; /* transfer queue */
141 u_long sc_options; /* options */
142 dev_t sc_devlist; /* devices that refer to this unit */
143 SLIST_ENTRY(vn_softc) sc_list;
146 static SLIST_HEAD(, vn_softc) vn_list;
149 #define VNF_INITED 0x01
150 #define VNF_READONLY 0x02
152 static u_long vn_options;
154 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
155 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
157 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
158 static void vnclear (struct vn_softc *vn);
159 static int vn_modevent (module_t, int, void *);
160 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct proc *p);
161 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct proc *p);
164 vnclose(dev_t dev, int flags, int mode, struct proc *p)
166 struct vn_softc *vn = dev->si_drv1;
169 if (vn->sc_slices != NULL)
170 dsclose(dev, mode, vn->sc_slices);
174 static struct vn_softc *
183 SLIST_FOREACH(vn, &vn_list, sc_list) {
184 if (vn->sc_unit == unit) {
185 dev->si_drv2 = vn->sc_devlist;
186 vn->sc_devlist = dev;
193 vn = malloc(sizeof *vn, M_DEVBUF, M_WAITOK);
196 bzero(vn, sizeof *vn);
199 vn->sc_devlist = make_dev(&vn_cdevsw, 0,
200 UID_ROOT, GID_OPERATOR, 0640, "vn%d", unit);
201 vn->sc_devlist->si_drv1 = vn;
202 vn->sc_devlist->si_drv2 = NULL;
203 if (vn->sc_devlist != dev) {
204 dev->si_drv2 = vn->sc_devlist;
205 vn->sc_devlist = dev;
207 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
213 vnopen(dev_t dev, int flags, int mode, struct proc *p)
218 * Locate preexisting device
221 if ((vn = dev->si_drv1) == NULL)
225 * Update si_bsize fields for device. This data will be overriden by
226 * the slice/parition code for vn accesses through partitions, and
227 * used directly if you open the 'whole disk' device.
229 * si_bsize_best must be reinitialized in case VN has been
230 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
232 dev->si_bsize_phys = vn->sc_secsize;
233 dev->si_bsize_best = vn->sc_secsize;
234 if (dev->si_bsize_best < VN_BSIZE_BEST)
235 dev->si_bsize_best = VN_BSIZE_BEST;
237 if ((flags & FWRITE) && (vn->sc_flags & VNF_READONLY))
241 printf("vnopen(%s, 0x%x, 0x%x, %p)\n",
242 devtoname(dev), flags, mode, (void *)p);
248 IFOPT(vn, VN_LABELS) {
249 if (vn->sc_flags & VNF_INITED) {
250 struct disklabel label;
252 /* Build label for whole disk. */
253 bzero(&label, sizeof label);
254 label.d_secsize = vn->sc_secsize;
255 label.d_nsectors = 32;
256 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE);
257 label.d_secpercyl = label.d_nsectors * label.d_ntracks;
258 label.d_ncylinders = vn->sc_size / label.d_secpercyl;
259 label.d_secperunit = vn->sc_size;
260 label.d_partitions[RAW_PART].p_size = vn->sc_size;
262 return (dsopen(dev, mode, 0, &vn->sc_slices, &label));
264 if (dkslice(dev) != WHOLE_DISK_SLICE ||
265 dkpart(dev) != RAW_PART ||
276 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
277 * for vnode-backed vn's, and the new vm_pager_strategy() call for
278 * vm_object-backed vn's.
280 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
282 * NOTE: bp->b_blkno is DEV_BSIZE'd. We must generate bp->b_pblkno for
283 * our uio or vn_pager_strategy() call that is vn->sc_secsize'd
287 vnstrategy(struct buf *bp)
293 unit = dkunit(bp->b_dev);
294 vn = bp->b_dev->si_drv1;
296 vn = vnfindvn(bp->b_dev);
299 printf("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;
310 IFOPT(vn, VN_LABELS) {
311 if (vn->sc_slices != NULL && dscheck(bp, vn->sc_slices) <= 0) {
312 bp->b_flags |= B_INVAL;
317 int pbn; /* in sc_secsize chunks */
318 long sz; /* in sc_secsize chunks */
321 * Check for required alignment. Transfers must be a valid
322 * multiple of the sector size.
324 if (bp->b_bcount % vn->sc_secsize != 0 ||
325 bp->b_blkno % (vn->sc_secsize / DEV_BSIZE) != 0) {
326 bp->b_error = EINVAL;
327 bp->b_flags |= B_ERROR | B_INVAL;
332 pbn = bp->b_blkno / (vn->sc_secsize / DEV_BSIZE);
333 sz = howmany(bp->b_bcount, vn->sc_secsize);
336 * If out of bounds return an error. If at the EOF point,
337 * simply read or write less.
339 if (pbn < 0 || pbn >= vn->sc_size) {
340 if (pbn != vn->sc_size) {
341 bp->b_error = EINVAL;
342 bp->b_flags |= B_ERROR | B_INVAL;
349 * If the request crosses EOF, truncate the request.
351 if (pbn + sz > vn->sc_size) {
352 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
353 bp->b_resid = bp->b_bcount;
358 if (vn->sc_vp && (bp->b_flags & B_FREEBUF)) {
360 * Not handled for vnode-backed element yet.
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 = (vm_ooffset_t)bp->b_pblkno * vn->sc_secsize;
381 auio.uio_segflg = UIO_SYSSPACE;
382 if( bp->b_flags & B_READ)
383 auio.uio_rw = UIO_READ;
385 auio.uio_rw = UIO_WRITE;
386 auio.uio_resid = bp->b_bcount;
387 auio.uio_procp = curproc;
388 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curproc);
389 if (bp->b_flags & B_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, curproc);
394 bp->b_resid = auio.uio_resid;
398 bp->b_flags |= B_ERROR;
401 } else if (vn->sc_object) {
405 * ( handles read, write, freebuf )
407 * Note: if we pre-reserved swap, B_FREEBUF is disabled
409 KASSERT((bp->b_bufsize & (vn->sc_secsize - 1)) == 0,
410 ("vnstrategy: buffer %p too small for physio", bp));
412 if ((bp->b_flags & B_FREEBUF) && TESTOPT(vn, VN_RESERVE)) {
415 vm_pager_strategy(vn->sc_object, bp);
418 bp->b_flags |= B_ERROR;
419 bp->b_error = EINVAL;
426 vnioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct proc *p)
429 struct vn_ioctl *vio;
435 printf("vnioctl(%s, 0x%lx, %p, 0x%x, %p): unit %d\n",
436 devtoname(dev), cmd, (void *)data, flag, (void *)p,
449 IFOPT(vn,VN_LABELS) {
450 if (vn->sc_slices != NULL) {
451 error = dsioctl(dev, cmd, data, flag, &vn->sc_slices);
452 if (error != ENOIOCTL)
455 if (dkslice(dev) != WHOLE_DISK_SLICE ||
456 dkpart(dev) != RAW_PART)
466 vio = (struct vn_ioctl *)data;
471 if (vn->sc_flags & VNF_INITED)
474 if (vio->vn_file == NULL)
475 error = vniocattach_swap(vn, vio, dev, flag, p);
477 error = vniocattach_file(vn, vio, dev, flag, p);
481 if ((vn->sc_flags & VNF_INITED) == 0)
484 * XXX handle i/o in progress. Return EBUSY, or wait, or
486 * XXX handle multiple opens of the device. Return EBUSY,
487 * or revoke the fd's.
488 * How are these problems handled for removable and failing
489 * hardware devices? (Hint: They are not)
493 printf("vnioctl: CLRed\n");
507 vn->sc_options |= *f;
512 vn->sc_options &= ~(*f);
526 * Attach a file to a VN partition. Return the size in the vn_size
531 vniocattach_file(vn, vio, dev, flag, p)
533 struct vn_ioctl *vio;
542 flags = FREAD|FWRITE;
543 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, vio->vn_file, p);
544 error = vn_open(&nd, flags, 0);
546 if (error != EACCES && error != EPERM && error != EROFS)
549 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, vio->vn_file, p);
550 error = vn_open(&nd, flags, 0);
554 NDFREE(&nd, NDF_ONLY_PNBUF);
555 if (nd.ni_vp->v_type != VREG ||
556 (error = VOP_GETATTR(nd.ni_vp, &vattr, p->p_ucred, p))) {
557 VOP_UNLOCK(nd.ni_vp, 0, p);
558 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p);
559 return (error ? error : EINVAL);
561 VOP_UNLOCK(nd.ni_vp, 0, p);
562 vn->sc_secsize = DEV_BSIZE;
563 vn->sc_vp = nd.ni_vp;
566 * If the size is specified, override the file attributes. Note that
567 * the vn_size argument is in PAGE_SIZE sized blocks.
570 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
572 vn->sc_size = vattr.va_size / vn->sc_secsize;
573 error = vnsetcred(vn, p->p_ucred);
575 (void) vn_close(nd.ni_vp, flags, p->p_ucred, p);
578 vn->sc_flags |= VNF_INITED;
580 vn->sc_flags |= VNF_READONLY;
581 IFOPT(vn, VN_LABELS) {
583 * Reopen so that `ds' knows which devices are open.
584 * If this is the first VNIOCSET, then we've
585 * guaranteed that the device is the cdev and that
586 * no other slices or labels are open. Otherwise,
587 * we rely on VNIOCCLR not being abused.
589 error = vnopen(dev, flag, S_IFCHR, p);
594 printf("vnioctl: SET vp %p size %x blks\n",
595 vn->sc_vp, vn->sc_size);
602 * Attach swap backing store to a VN partition of the size specified
607 vniocattach_swap(vn, vio, dev, flag, p)
609 struct vn_ioctl *vio;
617 * Range check. Disallow negative sizes or any size less then the
618 * size of a page. Then round to a page.
621 if (vio->vn_size <= 0)
625 * Allocate an OBJT_SWAP object.
627 * sc_secsize is PAGE_SIZE'd
629 * vio->vn_size is in PAGE_SIZE'd chunks.
630 * sc_size must be in PAGE_SIZE'd chunks.
631 * Note the truncation.
634 vn->sc_secsize = PAGE_SIZE;
635 vn->sc_size = vio->vn_size;
637 vm_pager_allocate(OBJT_SWAP, NULL, vn->sc_secsize * (vm_ooffset_t)vio->vn_size, VM_PROT_DEFAULT, 0);
638 IFOPT(vn, VN_RESERVE) {
639 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
640 vm_pager_deallocate(vn->sc_object);
641 vn->sc_object = NULL;
645 vn->sc_flags |= VNF_INITED;
647 error = vnsetcred(vn, p->p_ucred);
649 IFOPT(vn, VN_LABELS) {
651 * Reopen so that `ds' knows which devices are open.
652 * If this is the first VNIOCSET, then we've
653 * guaranteed that the device is the cdev and that
654 * no other slices or labels are open. Otherwise,
655 * we rely on VNIOCCLR not being abused.
657 error = vnopen(dev, flag, S_IFCHR, p);
661 IFOPT(vn, VN_FOLLOW) {
662 printf("vnioctl: SET vp %p size %x\n",
663 vn->sc_vp, vn->sc_size);
672 * Duplicate the current processes' credentials. Since we are called only
673 * as the result of a SET ioctl and only root can do that, any future access
674 * to this "disk" is essentially as root. Note that credentials may change
675 * if some other uid can write directly to the mapped file (NFS).
678 vnsetcred(struct vn_softc *vn, struct ucred *cred)
684 * Set credits in our softc
689 vn->sc_cred = crdup(cred);
692 * Horrible kludge to establish credentials for NFS XXX.
699 tmpbuf = malloc(vn->sc_secsize, M_TEMP, M_WAITOK);
700 bzero(&auio, sizeof(auio));
702 aiov.iov_base = tmpbuf;
703 aiov.iov_len = vn->sc_secsize;
704 auio.uio_iov = &aiov;
707 auio.uio_rw = UIO_READ;
708 auio.uio_segflg = UIO_SYSSPACE;
709 auio.uio_resid = aiov.iov_len;
710 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curproc);
711 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
712 VOP_UNLOCK(vn->sc_vp, 0, curproc);
713 free(tmpbuf, M_TEMP);
719 vnclear(struct vn_softc *vn)
721 struct proc *p = curproc; /* XXX */
724 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
725 if (vn->sc_slices != NULL)
726 dsgone(&vn->sc_slices);
727 vn->sc_flags &= ~VNF_INITED;
728 if (vn->sc_vp != NULL) {
729 (void)vn_close(vn->sc_vp, vn->sc_flags & VNF_READONLY ?
730 FREAD : (FREAD|FWRITE), vn->sc_cred, p);
733 vn->sc_flags &= ~VNF_READONLY;
738 if (vn->sc_object != NULL) {
739 vm_pager_deallocate(vn->sc_object);
740 vn->sc_object = NULL;
753 if ((vn->sc_flags & VNF_INITED) == 0)
760 vn_modevent(module_t mod, int type, void *data)
767 cdevsw_add(&vn_cdevsw);
774 vn = SLIST_FIRST(&vn_list);
777 SLIST_REMOVE_HEAD(&vn_list, sc_list);
778 if (vn->sc_flags & VNF_INITED)
780 /* Cleanup all dev_t's that refer to this unit */
781 while ((dev = vn->sc_devlist) != NULL) {
782 vn->sc_devlist = dev->si_drv2;
783 dev->si_drv1 = dev->si_drv2 = NULL;
784 /* If the last one, destroy it. */
785 if (vn->sc_devlist == NULL)
790 cdevsw_remove(&vn_cdevsw);
798 DEV_MODULE(vn, vn_modevent, 0);