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 1.13 2004/11/12 00:09:04 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: This uses the VOP_BMAP/VOP_STRATEGY interface to the vnode
52 * instead of a simple VOP_RDWR. We do this to avoid distorting the
55 * NOTE 2: There is a security issue involved with this driver.
56 * Once mounted all access to the contents of the "mapped" file via
57 * the special file is controlled by the permissions on the special
58 * file, the protection of the mapped file is ignored (effectively,
59 * by using root credentials in all transactions).
61 * NOTE 3: Doesn't interact with leases, should it?
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/kernel.h>
68 #include <sys/nlookup.h>
70 #include <sys/malloc.h>
71 #include <sys/mount.h>
72 #include <sys/vnode.h>
73 #include <sys/fcntl.h>
75 #include <sys/disklabel.h>
76 #include <sys/diskslice.h>
79 #include <sys/module.h>
80 #include <sys/vnioctl.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/vm_pager.h>
86 #include <vm/vm_pageout.h>
87 #include <vm/swap_pager.h>
88 #include <vm/vm_extern.h>
89 #include <vm/vm_zone.h>
91 static d_ioctl_t vnioctl;
92 static d_open_t vnopen;
93 static d_close_t vnclose;
94 static d_psize_t vnsize;
95 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 = {
109 /* maj */ CDEV_MAJOR,
110 /* flags */ D_DISK|D_CANFREE,
117 /* write */ physwrite,
121 /* strategy */ vnstrategy,
127 ((struct buf *)malloc(sizeof(struct buf), M_DEVBUF, M_WAITOK))
129 #define putvnbuf(bp) \
130 free((caddr_t)(bp), M_DEVBUF)
134 int sc_flags; /* flags */
135 int sc_size; /* size of vn, sc_secsize scale */
136 int sc_secsize; /* sector size */
137 struct diskslices *sc_slices;
138 struct vnode *sc_vp; /* vnode if not NULL */
139 vm_object_t sc_object; /* backing object if not NULL */
140 struct ucred *sc_cred; /* credentials */
141 int sc_maxactive; /* max # of active requests */
142 struct buf sc_tab; /* transfer queue */
143 u_long sc_options; /* options */
144 dev_t sc_devlist; /* devices that refer to this unit */
145 SLIST_ENTRY(vn_softc) sc_list;
148 static SLIST_HEAD(, vn_softc) vn_list;
151 #define VNF_INITED 0x01
152 #define VNF_READONLY 0x02
154 static u_long vn_options;
156 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
157 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
159 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
160 static void vnclear (struct vn_softc *vn);
161 static int vn_modevent (module_t, int, void *);
162 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p);
163 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p);
166 vnclose(dev_t dev, int flags, int mode, struct thread *td)
168 struct vn_softc *vn = dev->si_drv1;
171 if (vn->sc_slices != NULL)
172 dsclose(dev, mode, vn->sc_slices);
177 * Called only when si_drv1 is NULL. Locate the associated vn node and
178 * attach the device to it.
180 static struct vn_softc *
187 SLIST_FOREACH(vn, &vn_list, sc_list) {
188 if (vn->sc_unit == unit) {
190 dev->si_drv2 = vn->sc_devlist;
191 vn->sc_devlist = dev;
197 vn = malloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
200 vn->sc_devlist = make_dev(&vn_cdevsw, 0, UID_ROOT,
201 GID_OPERATOR, 0640, "vn%d", unit);
202 if (vn->sc_devlist->si_drv1 == NULL) {
203 reference_dev(vn->sc_devlist);
204 vn->sc_devlist->si_drv1 = vn;
205 vn->sc_devlist->si_drv2 = NULL;
207 if (vn->sc_devlist != dev) {
209 dev->si_drv2 = vn->sc_devlist;
210 vn->sc_devlist = dev;
213 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
219 vnopen(dev_t dev, int flags, int mode, struct thread *td)
224 * Locate preexisting device
227 if ((vn = dev->si_drv1) == 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 ((flags & FWRITE) && (vn->sc_flags & VNF_READONLY))
247 printf("vnopen(%s, 0x%x, 0x%x, %p)\n",
248 devtoname(dev), flags, mode, (void *)td);
254 IFOPT(vn, VN_LABELS) {
255 if (vn->sc_flags & VNF_INITED) {
256 struct disklabel label;
258 /* Build label for whole disk. */
259 bzero(&label, sizeof label);
260 label.d_secsize = vn->sc_secsize;
261 label.d_nsectors = 32;
262 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE);
263 label.d_secpercyl = label.d_nsectors * label.d_ntracks;
264 label.d_ncylinders = vn->sc_size / label.d_secpercyl;
265 label.d_secperunit = vn->sc_size;
266 label.d_partitions[RAW_PART].p_size = vn->sc_size;
268 return (dsopen(dev, mode, 0, &vn->sc_slices, &label));
270 if (dkslice(dev) != WHOLE_DISK_SLICE ||
271 dkpart(dev) != RAW_PART ||
282 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
283 * for vnode-backed vn's, and the new vm_pager_strategy() call for
284 * vm_object-backed vn's.
286 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
288 * NOTE: bp->b_blkno is DEV_BSIZE'd. We must generate bp->b_pblkno for
289 * our uio or vn_pager_strategy() call that is vn->sc_secsize'd
293 vnstrategy(struct buf *bp)
299 unit = dkunit(bp->b_dev);
300 if ((vn = bp->b_dev->si_drv1) == NULL)
301 vn = vnfindvn(bp->b_dev);
304 printf("vnstrategy(%p): unit %d\n", bp, unit);
306 if ((vn->sc_flags & VNF_INITED) == 0) {
308 bp->b_flags |= B_ERROR;
313 bp->b_resid = bp->b_bcount;
315 IFOPT(vn, VN_LABELS) {
316 if (vn->sc_slices != NULL && dscheck(bp, vn->sc_slices) <= 0) {
317 bp->b_flags |= B_INVAL;
322 int pbn; /* in sc_secsize chunks */
323 long sz; /* in sc_secsize chunks */
326 * Check for required alignment. Transfers must be a valid
327 * multiple of the sector size.
329 if (bp->b_bcount % vn->sc_secsize != 0 ||
330 bp->b_blkno % (vn->sc_secsize / DEV_BSIZE) != 0) {
331 bp->b_error = EINVAL;
332 bp->b_flags |= B_ERROR | B_INVAL;
337 pbn = bp->b_blkno / (vn->sc_secsize / DEV_BSIZE);
338 sz = howmany(bp->b_bcount, vn->sc_secsize);
341 * If out of bounds return an error. If at the EOF point,
342 * simply read or write less.
344 if (pbn < 0 || pbn >= vn->sc_size) {
345 if (pbn != vn->sc_size) {
346 bp->b_error = EINVAL;
347 bp->b_flags |= B_ERROR | B_INVAL;
354 * If the request crosses EOF, truncate the request.
356 if (pbn + sz > vn->sc_size) {
357 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
358 bp->b_resid = bp->b_bcount;
363 if (vn->sc_vp && (bp->b_flags & B_FREEBUF)) {
365 * Not handled for vnode-backed element yet.
368 } else if (vn->sc_vp) {
372 * If an error occurs, we set B_ERROR but we do not set
373 * B_INVAL because (for a write anyway), the buffer is
379 bzero(&auio, sizeof(auio));
381 aiov.iov_base = bp->b_data;
382 aiov.iov_len = bp->b_bcount;
383 auio.uio_iov = &aiov;
385 auio.uio_offset = (vm_ooffset_t)bp->b_pblkno * vn->sc_secsize;
386 auio.uio_segflg = UIO_SYSSPACE;
387 if( bp->b_flags & B_READ)
388 auio.uio_rw = UIO_READ;
390 auio.uio_rw = UIO_WRITE;
391 auio.uio_resid = bp->b_bcount;
392 auio.uio_td = curthread;
393 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curthread);
394 if (bp->b_flags & B_READ)
395 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
397 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
398 VOP_UNLOCK(vn->sc_vp, 0, curthread);
399 bp->b_resid = auio.uio_resid;
403 bp->b_flags |= B_ERROR;
406 } else if (vn->sc_object) {
410 * ( handles read, write, freebuf )
412 * Note: if we pre-reserved swap, B_FREEBUF is disabled
414 KASSERT((bp->b_bufsize & (vn->sc_secsize - 1)) == 0,
415 ("vnstrategy: buffer %p too small for physio", bp));
417 if ((bp->b_flags & B_FREEBUF) && TESTOPT(vn, VN_RESERVE)) {
420 vm_pager_strategy(vn->sc_object, bp);
423 bp->b_flags |= B_ERROR;
424 bp->b_error = EINVAL;
431 vnioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct thread *td)
434 struct vn_ioctl *vio;
440 printf("vnioctl(%s, 0x%lx, %p, 0x%x, %p): unit %d\n",
441 devtoname(dev), cmd, (void *)data, flag, (void *)td,
454 IFOPT(vn,VN_LABELS) {
455 if (vn->sc_slices != NULL) {
456 error = dsioctl(dev, cmd, data, flag, &vn->sc_slices);
457 if (error != ENOIOCTL)
460 if (dkslice(dev) != WHOLE_DISK_SLICE ||
461 dkpart(dev) != RAW_PART)
471 vio = (struct vn_ioctl *)data;
476 if (vn->sc_flags & VNF_INITED)
479 if (vio->vn_file == NULL)
480 error = vniocattach_swap(vn, vio, dev, flag, td);
482 error = vniocattach_file(vn, vio, dev, flag, td);
486 if ((vn->sc_flags & VNF_INITED) == 0)
489 * XXX handle i/o in progress. Return EBUSY, or wait, or
491 * XXX handle multiple opens of the device. Return EBUSY,
492 * or revoke the fd's.
493 * How are these problems handled for removable and failing
494 * hardware devices? (Hint: They are not)
498 printf("vnioctl: CLRed\n");
512 vn->sc_options |= *f;
517 vn->sc_options &= ~(*f);
531 * Attach a file to a VN partition. Return the size in the vn_size
536 vniocattach_file(vn, vio, dev, flag, td)
538 struct vn_ioctl *vio;
544 struct nlookupdata nd;
547 struct proc *p = td->td_proc;
551 flags = FREAD|FWRITE;
552 error = nlookup_init(&nd, vio->vn_file,
553 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
556 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
557 if (error != EACCES && error != EPERM && error != EROFS)
561 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
564 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
568 if (vp->v_type != VREG ||
569 (error = VOP_GETATTR(vp, &vattr, td))) {
574 VOP_UNLOCK(vp, 0, td);
575 vn->sc_secsize = DEV_BSIZE;
577 nd.nl_open_vp = NULL;
580 * If the size is specified, override the file attributes. Note that
581 * the vn_size argument is in PAGE_SIZE sized blocks.
584 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
586 vn->sc_size = vattr.va_size / vn->sc_secsize;
587 error = vnsetcred(vn, p->p_ucred);
590 vn_close(vp, flags, td);
593 vn->sc_flags |= VNF_INITED;
595 vn->sc_flags |= VNF_READONLY;
596 IFOPT(vn, VN_LABELS) {
598 * Reopen so that `ds' knows which devices are open.
599 * If this is the first VNIOCSET, then we've
600 * guaranteed that the device is the cdev and that
601 * no other slices or labels are open. Otherwise,
602 * we rely on VNIOCCLR not being abused.
604 error = vnopen(dev, flag, S_IFCHR, td);
609 printf("vnioctl: SET vp %p size %x blks\n",
610 vn->sc_vp, vn->sc_size);
619 * Attach swap backing store to a VN partition of the size specified
624 vniocattach_swap(vn, vio, dev, flag, td)
626 struct vn_ioctl *vio;
632 struct proc *p = td->td_proc;
636 * Range check. Disallow negative sizes or any size less then the
637 * size of a page. Then round to a page.
640 if (vio->vn_size <= 0)
644 * Allocate an OBJT_SWAP object.
646 * sc_secsize is PAGE_SIZE'd
648 * vio->vn_size is in PAGE_SIZE'd chunks.
649 * sc_size must be in PAGE_SIZE'd chunks.
650 * Note the truncation.
653 vn->sc_secsize = PAGE_SIZE;
654 vn->sc_size = vio->vn_size;
656 vm_pager_allocate(OBJT_SWAP, NULL, vn->sc_secsize * (vm_ooffset_t)vio->vn_size, VM_PROT_DEFAULT, 0);
657 IFOPT(vn, VN_RESERVE) {
658 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
659 vm_pager_deallocate(vn->sc_object);
660 vn->sc_object = NULL;
664 vn->sc_flags |= VNF_INITED;
666 error = vnsetcred(vn, p->p_ucred);
668 IFOPT(vn, VN_LABELS) {
670 * Reopen so that `ds' knows which devices are open.
671 * If this is the first VNIOCSET, then we've
672 * guaranteed that the device is the cdev and that
673 * no other slices or labels are open. Otherwise,
674 * we rely on VNIOCCLR not being abused.
676 error = vnopen(dev, flag, S_IFCHR, td);
680 IFOPT(vn, VN_FOLLOW) {
681 printf("vnioctl: SET vp %p size %x\n",
682 vn->sc_vp, vn->sc_size);
691 * Duplicate the current processes' credentials. Since we are called only
692 * as the result of a SET ioctl and only root can do that, any future access
693 * to this "disk" is essentially as root. Note that credentials may change
694 * if some other uid can write directly to the mapped file (NFS).
697 vnsetcred(struct vn_softc *vn, struct ucred *cred)
703 * Set credits in our softc
708 vn->sc_cred = crdup(cred);
711 * Horrible kludge to establish credentials for NFS XXX.
718 tmpbuf = malloc(vn->sc_secsize, M_TEMP, M_WAITOK);
719 bzero(&auio, sizeof(auio));
721 aiov.iov_base = tmpbuf;
722 aiov.iov_len = vn->sc_secsize;
723 auio.uio_iov = &aiov;
726 auio.uio_rw = UIO_READ;
727 auio.uio_segflg = UIO_SYSSPACE;
728 auio.uio_resid = aiov.iov_len;
729 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curthread);
730 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
731 VOP_UNLOCK(vn->sc_vp, 0, curthread);
732 free(tmpbuf, M_TEMP);
738 vnclear(struct vn_softc *vn)
740 struct thread *td = curthread; /* XXX */
743 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
744 if (vn->sc_slices != NULL)
745 dsgone(&vn->sc_slices);
746 vn->sc_flags &= ~VNF_INITED;
747 if (vn->sc_vp != NULL) {
748 vn_close(vn->sc_vp, vn->sc_flags & VNF_READONLY ?
749 FREAD : (FREAD|FWRITE), td);
752 vn->sc_flags &= ~VNF_READONLY;
757 if (vn->sc_object != NULL) {
758 vm_pager_deallocate(vn->sc_object);
759 vn->sc_object = NULL;
772 if ((vn->sc_flags & VNF_INITED) == 0)
779 vn_modevent(module_t mod, int type, void *data)
786 cdevsw_add(&vn_cdevsw, 0, 0);
791 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
792 SLIST_REMOVE_HEAD(&vn_list, sc_list);
793 if (vn->sc_flags & VNF_INITED)
795 /* Cleanup all dev_t's that refer to this unit */
796 while ((dev = vn->sc_devlist) != NULL) {
797 vn->sc_devlist = dev->si_drv2;
798 dev->si_drv1 = dev->si_drv2 = NULL;
803 cdevsw_remove(&vn_cdevsw, 0, 0);
811 DEV_MODULE(vn, vn_modevent, 0);