proc->thread stage 5: BUF/VFS clearance! Remove the ucred argument from
[dragonfly.git] / sys / dev / disk / vn / vn.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1988 University of Utah.
3 * Copyright (c) 1990, 1993
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * the Systems Programming Group of the University of Utah Computer
8 * Science Department.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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.
25 *
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
36 * SUCH DAMAGE.
37 *
38 * from: Utah Hdr: vn.c 1.13 94/04/02
39 *
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 $
3b568787 42 * $DragonFly: src/sys/dev/disk/vn/vn.c,v 1.5 2003/06/26 05:55:11 dillon Exp $
984263bc
MD
43 */
44
45/*
46 * Vnode disk driver.
47 *
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.).
50 *
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
53 * local buffer cache.
54 *
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).
60 *
61 * NOTE 3: Doesn't interact with leases, should it?
62 */
63
64#include <sys/param.h>
65#include <sys/systm.h>
66#include <sys/kernel.h>
984263bc 67#include <sys/proc.h>
dadab5e9 68#include <sys/namei.h>
984263bc
MD
69#include <sys/buf.h>
70#include <sys/malloc.h>
71#include <sys/mount.h>
72#include <sys/vnode.h>
73#include <sys/fcntl.h>
74#include <sys/conf.h>
75#include <sys/disklabel.h>
76#include <sys/diskslice.h>
77#include <sys/stat.h>
78#include <sys/conf.h>
79#include <sys/module.h>
80#include <sys/vnioctl.h>
81
82#include <vm/vm.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>
90
91static d_ioctl_t vnioctl;
92static d_open_t vnopen;
93static d_close_t vnclose;
94static d_psize_t vnsize;
95static d_strategy_t vnstrategy;
96
97#define CDEV_MAJOR 43
98#define BDEV_MAJOR 15
99
100#define VN_BSIZE_BEST 8192
101
102/*
103 * cdevsw
104 * D_DISK we want to look like a disk
105 * D_CANFREE We support B_FREEBUF
106 */
107
108static struct cdevsw vn_cdevsw = {
109 /* open */ vnopen,
110 /* close */ vnclose,
111 /* read */ physread,
112 /* write */ physwrite,
113 /* ioctl */ vnioctl,
114 /* poll */ nopoll,
115 /* mmap */ nommap,
116 /* strategy */ vnstrategy,
117 /* name */ "vn",
118 /* maj */ CDEV_MAJOR,
119 /* dump */ nodump,
120 /* psize */ vnsize,
121 /* flags */ D_DISK|D_CANFREE,
122 /* bmaj */ BDEV_MAJOR
123};
124
125#define getvnbuf() \
126 ((struct buf *)malloc(sizeof(struct buf), M_DEVBUF, M_WAITOK))
127
128#define putvnbuf(bp) \
129 free((caddr_t)(bp), M_DEVBUF)
130
131struct vn_softc {
132 int sc_unit;
133 int sc_flags; /* flags */
134 int sc_size; /* size of vn, sc_secsize scale */
135 int sc_secsize; /* sector size */
136 struct diskslices *sc_slices;
137 struct vnode *sc_vp; /* vnode if not NULL */
138 vm_object_t sc_object; /* backing object if not NULL */
139 struct ucred *sc_cred; /* credentials */
140 int sc_maxactive; /* max # of active requests */
141 struct buf sc_tab; /* transfer queue */
142 u_long sc_options; /* options */
143 dev_t sc_devlist; /* devices that refer to this unit */
144 SLIST_ENTRY(vn_softc) sc_list;
145};
146
147static SLIST_HEAD(, vn_softc) vn_list;
148
149/* sc_flags */
150#define VNF_INITED 0x01
151#define VNF_READONLY 0x02
152
153static u_long vn_options;
154
155#define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
156#define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
157
158static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
159static void vnclear (struct vn_softc *vn);
160static int vn_modevent (module_t, int, void *);
41c20dac
MD
161static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p);
162static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, dev_t dev, int flag, struct thread *p);
984263bc
MD
163
164static int
41c20dac 165vnclose(dev_t dev, int flags, int mode, struct thread *td)
984263bc
MD
166{
167 struct vn_softc *vn = dev->si_drv1;
168
169 IFOPT(vn, VN_LABELS)
170 if (vn->sc_slices != NULL)
171 dsclose(dev, mode, vn->sc_slices);
172 return (0);
173}
174
175static struct vn_softc *
176vnfindvn(dev_t dev)
177{
178 int unit;
179 struct vn_softc *vn;
180
181 unit = dkunit(dev);
182 vn = dev->si_drv1;
183 if (!vn) {
184 SLIST_FOREACH(vn, &vn_list, sc_list) {
185 if (vn->sc_unit == unit) {
186 dev->si_drv2 = vn->sc_devlist;
187 vn->sc_devlist = dev;
188 dev->si_drv1 = vn;
189 break;
190 }
191 }
192 }
193 if (!vn) {
194 vn = malloc(sizeof *vn, M_DEVBUF, M_WAITOK);
195 if (!vn)
196 return (NULL);
197 bzero(vn, sizeof *vn);
198 vn->sc_unit = unit;
199 dev->si_drv1 = vn;
200 vn->sc_devlist = make_dev(&vn_cdevsw, 0,
201 UID_ROOT, GID_OPERATOR, 0640, "vn%d", unit);
202 vn->sc_devlist->si_drv1 = vn;
203 vn->sc_devlist->si_drv2 = NULL;
204 if (vn->sc_devlist != dev) {
205 dev->si_drv2 = vn->sc_devlist;
206 vn->sc_devlist = dev;
207 }
208 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
209 }
210 return (vn);
211}
212
213static int
41c20dac 214vnopen(dev_t dev, int flags, int mode, struct thread *td)
984263bc
MD
215{
216 struct vn_softc *vn;
217
218 /*
219 * Locate preexisting device
220 */
221
222 if ((vn = dev->si_drv1) == NULL)
223 vn = vnfindvn(dev);
224
225 /*
226 * Update si_bsize fields for device. This data will be overriden by
227 * the slice/parition code for vn accesses through partitions, and
228 * used directly if you open the 'whole disk' device.
229 *
230 * si_bsize_best must be reinitialized in case VN has been
231 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
232 */
233 dev->si_bsize_phys = vn->sc_secsize;
234 dev->si_bsize_best = vn->sc_secsize;
235 if (dev->si_bsize_best < VN_BSIZE_BEST)
236 dev->si_bsize_best = VN_BSIZE_BEST;
237
238 if ((flags & FWRITE) && (vn->sc_flags & VNF_READONLY))
239 return (EACCES);
240
241 IFOPT(vn, VN_FOLLOW)
242 printf("vnopen(%s, 0x%x, 0x%x, %p)\n",
41c20dac 243 devtoname(dev), flags, mode, (void *)td);
984263bc
MD
244
245 /*
246 * Initialize label
247 */
248
249 IFOPT(vn, VN_LABELS) {
250 if (vn->sc_flags & VNF_INITED) {
251 struct disklabel label;
252
253 /* Build label for whole disk. */
254 bzero(&label, sizeof label);
255 label.d_secsize = vn->sc_secsize;
256 label.d_nsectors = 32;
257 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE);
258 label.d_secpercyl = label.d_nsectors * label.d_ntracks;
259 label.d_ncylinders = vn->sc_size / label.d_secpercyl;
260 label.d_secperunit = vn->sc_size;
261 label.d_partitions[RAW_PART].p_size = vn->sc_size;
262
263 return (dsopen(dev, mode, 0, &vn->sc_slices, &label));
264 }
265 if (dkslice(dev) != WHOLE_DISK_SLICE ||
266 dkpart(dev) != RAW_PART ||
267 mode != S_IFCHR) {
268 return (ENXIO);
269 }
270 }
271 return(0);
272}
273
274/*
275 * vnstrategy:
276 *
277 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
278 * for vnode-backed vn's, and the new vm_pager_strategy() call for
279 * vm_object-backed vn's.
280 *
281 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
282 *
283 * NOTE: bp->b_blkno is DEV_BSIZE'd. We must generate bp->b_pblkno for
284 * our uio or vn_pager_strategy() call that is vn->sc_secsize'd
285 */
286
287static void
288vnstrategy(struct buf *bp)
289{
290 int unit;
291 struct vn_softc *vn;
292 int error;
293
294 unit = dkunit(bp->b_dev);
295 vn = bp->b_dev->si_drv1;
296 if (!vn)
297 vn = vnfindvn(bp->b_dev);
298
299 IFOPT(vn, VN_DEBUG)
300 printf("vnstrategy(%p): unit %d\n", bp, unit);
301
302 if ((vn->sc_flags & VNF_INITED) == 0) {
303 bp->b_error = ENXIO;
304 bp->b_flags |= B_ERROR;
305 biodone(bp);
306 return;
307 }
308
309 bp->b_resid = bp->b_bcount;
310
311 IFOPT(vn, VN_LABELS) {
312 if (vn->sc_slices != NULL && dscheck(bp, vn->sc_slices) <= 0) {
313 bp->b_flags |= B_INVAL;
314 biodone(bp);
315 return;
316 }
317 } else {
318 int pbn; /* in sc_secsize chunks */
319 long sz; /* in sc_secsize chunks */
320
321 /*
322 * Check for required alignment. Transfers must be a valid
323 * multiple of the sector size.
324 */
325 if (bp->b_bcount % vn->sc_secsize != 0 ||
326 bp->b_blkno % (vn->sc_secsize / DEV_BSIZE) != 0) {
327 bp->b_error = EINVAL;
328 bp->b_flags |= B_ERROR | B_INVAL;
329 biodone(bp);
330 return;
331 }
332
333 pbn = bp->b_blkno / (vn->sc_secsize / DEV_BSIZE);
334 sz = howmany(bp->b_bcount, vn->sc_secsize);
335
336 /*
337 * If out of bounds return an error. If at the EOF point,
338 * simply read or write less.
339 */
340 if (pbn < 0 || pbn >= vn->sc_size) {
341 if (pbn != vn->sc_size) {
342 bp->b_error = EINVAL;
343 bp->b_flags |= B_ERROR | B_INVAL;
344 }
345 biodone(bp);
346 return;
347 }
348
349 /*
350 * If the request crosses EOF, truncate the request.
351 */
352 if (pbn + sz > vn->sc_size) {
353 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
354 bp->b_resid = bp->b_bcount;
355 }
356 bp->b_pblkno = pbn;
357 }
358
359 if (vn->sc_vp && (bp->b_flags & B_FREEBUF)) {
360 /*
361 * Not handled for vnode-backed element yet.
362 */
363 biodone(bp);
364 } else if (vn->sc_vp) {
365 /*
366 * VNODE I/O
367 *
368 * If an error occurs, we set B_ERROR but we do not set
369 * B_INVAL because (for a write anyway), the buffer is
370 * still valid.
371 */
372 struct uio auio;
373 struct iovec aiov;
374
375 bzero(&auio, sizeof(auio));
376
377 aiov.iov_base = bp->b_data;
378 aiov.iov_len = bp->b_bcount;
379 auio.uio_iov = &aiov;
380 auio.uio_iovcnt = 1;
381 auio.uio_offset = (vm_ooffset_t)bp->b_pblkno * vn->sc_secsize;
382 auio.uio_segflg = UIO_SYSSPACE;
383 if( bp->b_flags & B_READ)
384 auio.uio_rw = UIO_READ;
385 else
386 auio.uio_rw = UIO_WRITE;
387 auio.uio_resid = bp->b_bcount;
dadab5e9
MD
388 auio.uio_td = curthread;
389 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curthread);
984263bc
MD
390 if (bp->b_flags & B_READ)
391 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
392 else
393 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
dadab5e9 394 VOP_UNLOCK(vn->sc_vp, 0, curthread);
984263bc
MD
395 bp->b_resid = auio.uio_resid;
396
397 if (error) {
398 bp->b_error = error;
399 bp->b_flags |= B_ERROR;
400 }
401 biodone(bp);
402 } else if (vn->sc_object) {
403 /*
404 * OBJT_SWAP I/O
405 *
406 * ( handles read, write, freebuf )
407 *
408 * Note: if we pre-reserved swap, B_FREEBUF is disabled
409 */
410 KASSERT((bp->b_bufsize & (vn->sc_secsize - 1)) == 0,
411 ("vnstrategy: buffer %p too small for physio", bp));
412
413 if ((bp->b_flags & B_FREEBUF) && TESTOPT(vn, VN_RESERVE)) {
414 biodone(bp);
415 } else {
416 vm_pager_strategy(vn->sc_object, bp);
417 }
418 } else {
419 bp->b_flags |= B_ERROR;
420 bp->b_error = EINVAL;
421 biodone(bp);
422 }
423}
424
425/* ARGSUSED */
426static int
41c20dac 427vnioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct thread *td)
984263bc
MD
428{
429 struct vn_softc *vn;
430 struct vn_ioctl *vio;
431 int error;
432 u_long *f;
433
434 vn = dev->si_drv1;
435 IFOPT(vn,VN_FOLLOW)
436 printf("vnioctl(%s, 0x%lx, %p, 0x%x, %p): unit %d\n",
41c20dac 437 devtoname(dev), cmd, (void *)data, flag, (void *)td,
984263bc
MD
438 dkunit(dev));
439
440 switch (cmd) {
441 case VNIOCATTACH:
442 case VNIOCDETACH:
443 case VNIOCGSET:
444 case VNIOCGCLEAR:
445 case VNIOCUSET:
446 case VNIOCUCLEAR:
447 goto vn_specific;
448 }
449
450 IFOPT(vn,VN_LABELS) {
451 if (vn->sc_slices != NULL) {
452 error = dsioctl(dev, cmd, data, flag, &vn->sc_slices);
453 if (error != ENOIOCTL)
454 return (error);
455 }
456 if (dkslice(dev) != WHOLE_DISK_SLICE ||
457 dkpart(dev) != RAW_PART)
458 return (ENOTTY);
459 }
460
461 vn_specific:
462
dadab5e9 463 error = suser(td);
984263bc
MD
464 if (error)
465 return (error);
466
467 vio = (struct vn_ioctl *)data;
468 f = (u_long*)data;
469 switch (cmd) {
470
471 case VNIOCATTACH:
472 if (vn->sc_flags & VNF_INITED)
473 return(EBUSY);
474
475 if (vio->vn_file == NULL)
41c20dac 476 error = vniocattach_swap(vn, vio, dev, flag, td);
984263bc 477 else
41c20dac 478 error = vniocattach_file(vn, vio, dev, flag, td);
984263bc
MD
479 break;
480
481 case VNIOCDETACH:
482 if ((vn->sc_flags & VNF_INITED) == 0)
483 return(ENXIO);
484 /*
485 * XXX handle i/o in progress. Return EBUSY, or wait, or
486 * flush the i/o.
487 * XXX handle multiple opens of the device. Return EBUSY,
488 * or revoke the fd's.
489 * How are these problems handled for removable and failing
490 * hardware devices? (Hint: They are not)
491 */
492 vnclear(vn);
493 IFOPT(vn, VN_FOLLOW)
494 printf("vnioctl: CLRed\n");
495 break;
496
497 case VNIOCGSET:
498 vn_options |= *f;
499 *f = vn_options;
500 break;
501
502 case VNIOCGCLEAR:
503 vn_options &= ~(*f);
504 *f = vn_options;
505 break;
506
507 case VNIOCUSET:
508 vn->sc_options |= *f;
509 *f = vn->sc_options;
510 break;
511
512 case VNIOCUCLEAR:
513 vn->sc_options &= ~(*f);
514 *f = vn->sc_options;
515 break;
516
517 default:
518 error = ENOTTY;
519 break;
520 }
521 return(error);
522}
523
524/*
525 * vniocattach_file:
526 *
527 * Attach a file to a VN partition. Return the size in the vn_size
528 * field.
529 */
530
531static int
41c20dac 532vniocattach_file(vn, vio, dev, flag, td)
984263bc
MD
533 struct vn_softc *vn;
534 struct vn_ioctl *vio;
535 dev_t dev;
536 int flag;
41c20dac 537 struct thread *td;
984263bc
MD
538{
539 struct vattr vattr;
540 struct nameidata nd;
541 int error, flags;
41c20dac
MD
542 struct proc *p = td->td_proc;
543
544 KKASSERT(p != NULL);
984263bc
MD
545
546 flags = FREAD|FWRITE;
dadab5e9 547 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, vio->vn_file, td);
984263bc
MD
548 error = vn_open(&nd, flags, 0);
549 if (error) {
550 if (error != EACCES && error != EPERM && error != EROFS)
551 return (error);
552 flags &= ~FWRITE;
dadab5e9 553 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, vio->vn_file, td);
984263bc
MD
554 error = vn_open(&nd, flags, 0);
555 if (error)
556 return (error);
557 }
558 NDFREE(&nd, NDF_ONLY_PNBUF);
559 if (nd.ni_vp->v_type != VREG ||
3b568787 560 (error = VOP_GETATTR(nd.ni_vp, &vattr, td))) {
dadab5e9 561 VOP_UNLOCK(nd.ni_vp, 0, td);
3b568787 562 (void) vn_close(nd.ni_vp, flags, td);
984263bc
MD
563 return (error ? error : EINVAL);
564 }
dadab5e9 565 VOP_UNLOCK(nd.ni_vp, 0, td);
984263bc
MD
566 vn->sc_secsize = DEV_BSIZE;
567 vn->sc_vp = nd.ni_vp;
568
569 /*
570 * If the size is specified, override the file attributes. Note that
571 * the vn_size argument is in PAGE_SIZE sized blocks.
572 */
573 if (vio->vn_size)
574 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
575 else
576 vn->sc_size = vattr.va_size / vn->sc_secsize;
577 error = vnsetcred(vn, p->p_ucred);
578 if (error) {
3b568787 579 (void) vn_close(nd.ni_vp, flags, td);
984263bc
MD
580 return(error);
581 }
582 vn->sc_flags |= VNF_INITED;
583 if (flags == FREAD)
584 vn->sc_flags |= VNF_READONLY;
585 IFOPT(vn, VN_LABELS) {
586 /*
587 * Reopen so that `ds' knows which devices are open.
588 * If this is the first VNIOCSET, then we've
589 * guaranteed that the device is the cdev and that
590 * no other slices or labels are open. Otherwise,
591 * we rely on VNIOCCLR not being abused.
592 */
41c20dac 593 error = vnopen(dev, flag, S_IFCHR, td);
984263bc
MD
594 if (error)
595 vnclear(vn);
596 }
597 IFOPT(vn, VN_FOLLOW)
598 printf("vnioctl: SET vp %p size %x blks\n",
599 vn->sc_vp, vn->sc_size);
600 return(0);
601}
602
603/*
604 * vniocattach_swap:
605 *
606 * Attach swap backing store to a VN partition of the size specified
607 * in vn_size.
608 */
609
610static int
41c20dac 611vniocattach_swap(vn, vio, dev, flag, td)
984263bc
MD
612 struct vn_softc *vn;
613 struct vn_ioctl *vio;
614 dev_t dev;
615 int flag;
41c20dac 616 struct thread *td;
984263bc
MD
617{
618 int error;
41c20dac 619 struct proc *p = td->td_proc;
984263bc 620
41c20dac 621 KKASSERT(p != NULL);
984263bc
MD
622 /*
623 * Range check. Disallow negative sizes or any size less then the
624 * size of a page. Then round to a page.
625 */
626
627 if (vio->vn_size <= 0)
628 return(EDOM);
629
630 /*
631 * Allocate an OBJT_SWAP object.
632 *
633 * sc_secsize is PAGE_SIZE'd
634 *
635 * vio->vn_size is in PAGE_SIZE'd chunks.
636 * sc_size must be in PAGE_SIZE'd chunks.
637 * Note the truncation.
638 */
639
640 vn->sc_secsize = PAGE_SIZE;
641 vn->sc_size = vio->vn_size;
642 vn->sc_object =
643 vm_pager_allocate(OBJT_SWAP, NULL, vn->sc_secsize * (vm_ooffset_t)vio->vn_size, VM_PROT_DEFAULT, 0);
644 IFOPT(vn, VN_RESERVE) {
645 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
646 vm_pager_deallocate(vn->sc_object);
647 vn->sc_object = NULL;
648 return(EDOM);
649 }
650 }
651 vn->sc_flags |= VNF_INITED;
652
653 error = vnsetcred(vn, p->p_ucred);
654 if (error == 0) {
655 IFOPT(vn, VN_LABELS) {
656 /*
657 * Reopen so that `ds' knows which devices are open.
658 * If this is the first VNIOCSET, then we've
659 * guaranteed that the device is the cdev and that
660 * no other slices or labels are open. Otherwise,
661 * we rely on VNIOCCLR not being abused.
662 */
41c20dac 663 error = vnopen(dev, flag, S_IFCHR, td);
984263bc
MD
664 }
665 }
666 if (error == 0) {
667 IFOPT(vn, VN_FOLLOW) {
668 printf("vnioctl: SET vp %p size %x\n",
669 vn->sc_vp, vn->sc_size);
670 }
671 }
672 if (error)
673 vnclear(vn);
674 return(error);
675}
676
677/*
678 * Duplicate the current processes' credentials. Since we are called only
679 * as the result of a SET ioctl and only root can do that, any future access
680 * to this "disk" is essentially as root. Note that credentials may change
681 * if some other uid can write directly to the mapped file (NFS).
682 */
683int
684vnsetcred(struct vn_softc *vn, struct ucred *cred)
685{
686 char *tmpbuf;
687 int error = 0;
688
689 /*
690 * Set credits in our softc
691 */
692
693 if (vn->sc_cred)
694 crfree(vn->sc_cred);
695 vn->sc_cred = crdup(cred);
696
697 /*
698 * Horrible kludge to establish credentials for NFS XXX.
699 */
700
701 if (vn->sc_vp) {
702 struct uio auio;
703 struct iovec aiov;
704
705 tmpbuf = malloc(vn->sc_secsize, M_TEMP, M_WAITOK);
706 bzero(&auio, sizeof(auio));
707
708 aiov.iov_base = tmpbuf;
709 aiov.iov_len = vn->sc_secsize;
710 auio.uio_iov = &aiov;
711 auio.uio_iovcnt = 1;
712 auio.uio_offset = 0;
713 auio.uio_rw = UIO_READ;
714 auio.uio_segflg = UIO_SYSSPACE;
715 auio.uio_resid = aiov.iov_len;
dadab5e9 716 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY, curthread);
984263bc 717 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
dadab5e9 718 VOP_UNLOCK(vn->sc_vp, 0, curthread);
984263bc
MD
719 free(tmpbuf, M_TEMP);
720 }
721 return (error);
722}
723
724void
725vnclear(struct vn_softc *vn)
726{
dadab5e9 727 struct thread *td = curthread; /* XXX */
984263bc
MD
728
729 IFOPT(vn, VN_FOLLOW)
730 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
731 if (vn->sc_slices != NULL)
732 dsgone(&vn->sc_slices);
733 vn->sc_flags &= ~VNF_INITED;
734 if (vn->sc_vp != NULL) {
735 (void)vn_close(vn->sc_vp, vn->sc_flags & VNF_READONLY ?
3b568787 736 FREAD : (FREAD|FWRITE), td);
984263bc
MD
737 vn->sc_vp = NULL;
738 }
739 vn->sc_flags &= ~VNF_READONLY;
740 if (vn->sc_cred) {
741 crfree(vn->sc_cred);
742 vn->sc_cred = NULL;
743 }
744 if (vn->sc_object != NULL) {
745 vm_pager_deallocate(vn->sc_object);
746 vn->sc_object = NULL;
747 }
748 vn->sc_size = 0;
749}
750
751static int
752vnsize(dev_t dev)
753{
754 struct vn_softc *vn;
755
756 vn = dev->si_drv1;
757 if (!vn)
758 return(-1);
759 if ((vn->sc_flags & VNF_INITED) == 0)
760 return(-1);
761
762 return(vn->sc_size);
763}
764
765static int
766vn_modevent(module_t mod, int type, void *data)
767{
768 struct vn_softc *vn;
769 dev_t dev;
770
771 switch (type) {
772 case MOD_LOAD:
773 cdevsw_add(&vn_cdevsw);
774 break;
775
776 case MOD_UNLOAD:
777 /* fall through */
778 case MOD_SHUTDOWN:
779 for (;;) {
780 vn = SLIST_FIRST(&vn_list);
781 if (!vn)
782 break;
783 SLIST_REMOVE_HEAD(&vn_list, sc_list);
784 if (vn->sc_flags & VNF_INITED)
785 vnclear(vn);
786 /* Cleanup all dev_t's that refer to this unit */
787 while ((dev = vn->sc_devlist) != NULL) {
788 vn->sc_devlist = dev->si_drv2;
789 dev->si_drv1 = dev->si_drv2 = NULL;
790 /* If the last one, destroy it. */
791 if (vn->sc_devlist == NULL)
792 destroy_dev(dev);
793 }
794 free(vn, M_DEVBUF);
795 }
796 cdevsw_remove(&vn_cdevsw);
797 break;
798 default:
799 break;
800 }
801 return 0;
802}
803
804DEV_MODULE(vn, vn_modevent, 0);