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 1.15 2005/12/11 01:54:08 swildner 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) {
317 * The vnode device is using disk/slice label support.
319 * The dscheck() function is called for validating the
320 * slices that exist ON the vnode device itself, and
321 * translate the "slice-relative" block number, again.
323 if (vn->sc_slices != NULL && dscheck(bp, vn->sc_slices) <= 0) {
324 bp->b_flags |= B_INVAL;
329 int pbn; /* in sc_secsize chunks */
330 long sz; /* in sc_secsize chunks */
333 * Check for required alignment. Transfers must be a valid
334 * multiple of the sector size.
336 if (bp->b_bcount % vn->sc_secsize != 0 ||
337 bp->b_blkno % (vn->sc_secsize / DEV_BSIZE) != 0) {
338 bp->b_error = EINVAL;
339 bp->b_flags |= B_ERROR | B_INVAL;
344 pbn = bp->b_blkno / (vn->sc_secsize / DEV_BSIZE);
345 sz = howmany(bp->b_bcount, vn->sc_secsize);
348 * If out of bounds return an error. If at the EOF point,
349 * simply read or write less.
351 if (pbn < 0 || pbn >= vn->sc_size) {
352 if (pbn != vn->sc_size) {
353 bp->b_error = EINVAL;
354 bp->b_flags |= B_ERROR | B_INVAL;
361 * If the request crosses EOF, truncate the request.
363 if (pbn + sz > vn->sc_size) {
364 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
365 bp->b_resid = bp->b_bcount;
370 if (vn->sc_vp && (bp->b_flags & B_FREEBUF)) {
372 * Not handled for vnode-backed element yet.
375 } else if (vn->sc_vp) {
379 * If an error occurs, we set B_ERROR but we do not set
380 * B_INVAL because (for a write anyway), the buffer is
386 bzero(&auio, sizeof(auio));
388 aiov.iov_base = bp->b_data;
389 aiov.iov_len = bp->b_bcount;
390 auio.uio_iov = &aiov;
392 auio.uio_offset = (vm_ooffset_t)bp->b_pblkno * vn->sc_secsize;
393 auio.uio_segflg = UIO_SYSSPACE;
394 if( bp->b_flags & B_READ)
395 auio.uio_rw = UIO_READ;
397 auio.uio_rw = UIO_WRITE;
398 auio.uio_resid = bp->b_bcount;
399 auio.uio_td = curthread;
400 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curthread);
401 if (bp->b_flags & B_READ)
402 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
404 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
405 VOP_UNLOCK(vn->sc_vp, 0, curthread);
406 bp->b_resid = auio.uio_resid;
410 bp->b_flags |= B_ERROR;
413 } else if (vn->sc_object) {
417 * ( handles read, write, freebuf )
419 * Note: if we pre-reserved swap, B_FREEBUF is disabled
421 KASSERT((bp->b_bufsize & (vn->sc_secsize - 1)) == 0,
422 ("vnstrategy: buffer %p too small for physio", bp));
424 if ((bp->b_flags & B_FREEBUF) && TESTOPT(vn, VN_RESERVE)) {
427 vm_pager_strategy(vn->sc_object, bp);
430 bp->b_flags |= B_ERROR;
431 bp->b_error = EINVAL;
438 vnioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct thread *td)
441 struct vn_ioctl *vio;
447 printf("vnioctl(%s, 0x%lx, %p, 0x%x, %p): unit %d\n",
448 devtoname(dev), cmd, (void *)data, flag, (void *)td,
461 IFOPT(vn,VN_LABELS) {
462 if (vn->sc_slices != NULL) {
463 error = dsioctl(dev, cmd, data, flag, &vn->sc_slices);
464 if (error != ENOIOCTL)
467 if (dkslice(dev) != WHOLE_DISK_SLICE ||
468 dkpart(dev) != RAW_PART)
478 vio = (struct vn_ioctl *)data;
483 if (vn->sc_flags & VNF_INITED)
486 if (vio->vn_file == NULL)
487 error = vniocattach_swap(vn, vio, dev, flag, td);
489 error = vniocattach_file(vn, vio, dev, flag, td);
493 if ((vn->sc_flags & VNF_INITED) == 0)
496 * XXX handle i/o in progress. Return EBUSY, or wait, or
498 * XXX handle multiple opens of the device. Return EBUSY,
499 * or revoke the fd's.
500 * How are these problems handled for removable and failing
501 * hardware devices? (Hint: They are not)
505 printf("vnioctl: CLRed\n");
519 vn->sc_options |= *f;
524 vn->sc_options &= ~(*f);
538 * Attach a file to a VN partition. Return the size in the vn_size
543 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, dev_t dev,
544 int flag, struct thread *td)
547 struct nlookupdata nd;
550 struct proc *p = td->td_proc;
554 flags = FREAD|FWRITE;
555 error = nlookup_init(&nd, vio->vn_file,
556 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
559 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
560 if (error != EACCES && error != EPERM && error != EROFS)
564 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
567 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
571 if (vp->v_type != VREG ||
572 (error = VOP_GETATTR(vp, &vattr, td))) {
577 VOP_UNLOCK(vp, 0, td);
578 vn->sc_secsize = DEV_BSIZE;
580 nd.nl_open_vp = NULL;
583 * If the size is specified, override the file attributes. Note that
584 * the vn_size argument is in PAGE_SIZE sized blocks.
587 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
589 vn->sc_size = vattr.va_size / vn->sc_secsize;
590 error = vnsetcred(vn, p->p_ucred);
593 vn_close(vp, flags, td);
596 vn->sc_flags |= VNF_INITED;
598 vn->sc_flags |= VNF_READONLY;
599 IFOPT(vn, VN_LABELS) {
601 * Reopen so that `ds' knows which devices are open.
602 * If this is the first VNIOCSET, then we've
603 * guaranteed that the device is the cdev and that
604 * no other slices or labels are open. Otherwise,
605 * we rely on VNIOCCLR not being abused.
607 error = vnopen(dev, flag, S_IFCHR, td);
612 printf("vnioctl: SET vp %p size %x blks\n",
613 vn->sc_vp, vn->sc_size);
622 * Attach swap backing store to a VN partition of the size specified
627 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, dev_t dev,
628 int flag, struct thread *td)
631 struct proc *p = td->td_proc;
635 * Range check. Disallow negative sizes or any size less then the
636 * size of a page. Then round to a page.
639 if (vio->vn_size <= 0)
643 * Allocate an OBJT_SWAP object.
645 * sc_secsize is PAGE_SIZE'd
647 * vio->vn_size is in PAGE_SIZE'd chunks.
648 * sc_size must be in PAGE_SIZE'd chunks.
649 * Note the truncation.
652 vn->sc_secsize = PAGE_SIZE;
653 vn->sc_size = vio->vn_size;
655 vm_pager_allocate(OBJT_SWAP, NULL, vn->sc_secsize * (vm_ooffset_t)vio->vn_size, VM_PROT_DEFAULT, 0);
656 IFOPT(vn, VN_RESERVE) {
657 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
658 vm_pager_deallocate(vn->sc_object);
659 vn->sc_object = NULL;
663 vn->sc_flags |= VNF_INITED;
665 error = vnsetcred(vn, p->p_ucred);
667 IFOPT(vn, VN_LABELS) {
669 * Reopen so that `ds' knows which devices are open.
670 * If this is the first VNIOCSET, then we've
671 * guaranteed that the device is the cdev and that
672 * no other slices or labels are open. Otherwise,
673 * we rely on VNIOCCLR not being abused.
675 error = vnopen(dev, flag, S_IFCHR, td);
679 IFOPT(vn, VN_FOLLOW) {
680 printf("vnioctl: SET vp %p size %x\n",
681 vn->sc_vp, vn->sc_size);
690 * Duplicate the current processes' credentials. Since we are called only
691 * as the result of a SET ioctl and only root can do that, any future access
692 * to this "disk" is essentially as root. Note that credentials may change
693 * if some other uid can write directly to the mapped file (NFS).
696 vnsetcred(struct vn_softc *vn, struct ucred *cred)
702 * Set credits in our softc
707 vn->sc_cred = crdup(cred);
710 * Horrible kludge to establish credentials for NFS XXX.
717 tmpbuf = malloc(vn->sc_secsize, M_TEMP, M_WAITOK);
718 bzero(&auio, sizeof(auio));
720 aiov.iov_base = tmpbuf;
721 aiov.iov_len = vn->sc_secsize;
722 auio.uio_iov = &aiov;
725 auio.uio_rw = UIO_READ;
726 auio.uio_segflg = UIO_SYSSPACE;
727 auio.uio_resid = aiov.iov_len;
728 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curthread);
729 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
730 VOP_UNLOCK(vn->sc_vp, 0, curthread);
731 free(tmpbuf, M_TEMP);
737 vnclear(struct vn_softc *vn)
739 struct thread *td = curthread; /* XXX */
742 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
743 if (vn->sc_slices != NULL)
744 dsgone(&vn->sc_slices);
745 vn->sc_flags &= ~VNF_INITED;
746 if (vn->sc_vp != NULL) {
747 vn_close(vn->sc_vp, vn->sc_flags & VNF_READONLY ?
748 FREAD : (FREAD|FWRITE), td);
751 vn->sc_flags &= ~VNF_READONLY;
756 if (vn->sc_object != NULL) {
757 vm_pager_deallocate(vn->sc_object);
758 vn->sc_object = NULL;
771 if ((vn->sc_flags & VNF_INITED) == 0)
778 vn_modevent(module_t mod, int type, void *data)
785 cdevsw_add(&vn_cdevsw, 0, 0);
790 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
791 SLIST_REMOVE_HEAD(&vn_list, sc_list);
792 if (vn->sc_flags & VNF_INITED)
794 /* Cleanup all dev_t's that refer to this unit */
795 while ((dev = vn->sc_devlist) != NULL) {
796 vn->sc_devlist = dev->si_drv2;
797 dev->si_drv1 = dev->si_drv2 = NULL;
802 cdevsw_remove(&vn_cdevsw, 0, 0);
810 DEV_MODULE(vn, vn_modevent, 0);