Change the kernel dev_t, representing a pointer to a specinfo structure,
[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 $
b13267a5 42 * $DragonFly: src/sys/dev/disk/vn/vn.c,v 1.28 2006/09/10 01:26:34 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 *
54078292 51 * NOTE 1: There is a security issue involved with this driver.
984263bc
MD
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).
56 *
54078292 57 * NOTE 2: Doesn't interact with leases, should it?
984263bc
MD
58 */
59
60#include <sys/param.h>
61#include <sys/systm.h>
62#include <sys/kernel.h>
984263bc 63#include <sys/proc.h>
fad57d0e 64#include <sys/nlookup.h>
984263bc
MD
65#include <sys/buf.h>
66#include <sys/malloc.h>
67#include <sys/mount.h>
68#include <sys/vnode.h>
69#include <sys/fcntl.h>
70#include <sys/conf.h>
71#include <sys/disklabel.h>
72#include <sys/diskslice.h>
73#include <sys/stat.h>
984263bc
MD
74#include <sys/module.h>
75#include <sys/vnioctl.h>
76
77#include <vm/vm.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>
85
86static d_ioctl_t vnioctl;
87static d_open_t vnopen;
88static d_close_t vnclose;
89static d_psize_t vnsize;
90static d_strategy_t vnstrategy;
91
92#define CDEV_MAJOR 43
984263bc
MD
93
94#define VN_BSIZE_BEST 8192
95
96/*
fef8985e 97 * dev_ops
984263bc 98 * D_DISK we want to look like a disk
10f3fee5 99 * D_CANFREE We support BUF_CMD_FREEBLKS
984263bc
MD
100 */
101
fef8985e
MD
102static struct dev_ops vn_ops = {
103 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
104 .d_open = vnopen,
105 .d_close = vnclose,
106 .d_read = physread,
107 .d_write = physwrite,
108 .d_ioctl = vnioctl,
109 .d_strategy = vnstrategy,
110 .d_psize = vnsize
984263bc
MD
111};
112
984263bc
MD
113struct vn_softc {
114 int sc_unit;
115 int sc_flags; /* flags */
116 int sc_size; /* size of vn, sc_secsize scale */
117 int sc_secsize; /* sector size */
118 struct diskslices *sc_slices;
119 struct vnode *sc_vp; /* vnode if not NULL */
120 vm_object_t sc_object; /* backing object if not NULL */
121 struct ucred *sc_cred; /* credentials */
122 int sc_maxactive; /* max # of active requests */
123 struct buf sc_tab; /* transfer queue */
124 u_long sc_options; /* options */
b13267a5 125 cdev_t sc_devlist; /* devices that refer to this unit */
984263bc
MD
126 SLIST_ENTRY(vn_softc) sc_list;
127};
128
129static SLIST_HEAD(, vn_softc) vn_list;
130
131/* sc_flags */
132#define VNF_INITED 0x01
133#define VNF_READONLY 0x02
134
135static u_long vn_options;
136
137#define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
138#define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
139
140static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
141static void vnclear (struct vn_softc *vn);
142static int vn_modevent (module_t, int, void *);
b13267a5
MD
143static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
144static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
984263bc
MD
145
146static int
fef8985e 147vnclose(struct dev_close_args *ap)
984263bc 148{
b13267a5 149 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
150 struct vn_softc *vn = dev->si_drv1;
151
152 IFOPT(vn, VN_LABELS)
153 if (vn->sc_slices != NULL)
fef8985e 154 dsclose(dev, ap->a_devtype, vn->sc_slices);
984263bc
MD
155 return (0);
156}
157
e2f51e29
MD
158/*
159 * Called only when si_drv1 is NULL. Locate the associated vn node and
160 * attach the device to it.
161 */
984263bc 162static struct vn_softc *
b13267a5 163vnfindvn(cdev_t dev)
984263bc
MD
164{
165 int unit;
166 struct vn_softc *vn;
167
168 unit = dkunit(dev);
e4c9c0c8
MD
169 SLIST_FOREACH(vn, &vn_list, sc_list) {
170 if (vn->sc_unit == unit) {
e2f51e29 171 dev->si_drv1 = vn;
e4c9c0c8
MD
172 dev->si_drv2 = vn->sc_devlist;
173 vn->sc_devlist = dev;
174 reference_dev(dev);
e4c9c0c8 175 break;
984263bc
MD
176 }
177 }
e4c9c0c8 178 if (vn == NULL) {
efda3bd0 179 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
984263bc
MD
180 vn->sc_unit = unit;
181 dev->si_drv1 = vn;
fef8985e 182 vn->sc_devlist = make_dev(&vn_ops, 0, UID_ROOT,
e4c9c0c8 183 GID_OPERATOR, 0640, "vn%d", unit);
e2f51e29
MD
184 if (vn->sc_devlist->si_drv1 == NULL) {
185 reference_dev(vn->sc_devlist);
186 vn->sc_devlist->si_drv1 = vn;
187 vn->sc_devlist->si_drv2 = NULL;
188 }
984263bc 189 if (vn->sc_devlist != dev) {
e4c9c0c8 190 dev->si_drv1 = vn;
984263bc
MD
191 dev->si_drv2 = vn->sc_devlist;
192 vn->sc_devlist = dev;
e4c9c0c8 193 reference_dev(dev);
984263bc
MD
194 }
195 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
196 }
197 return (vn);
198}
199
200static int
fef8985e 201vnopen(struct dev_open_args *ap)
984263bc 202{
b13267a5 203 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
204 struct vn_softc *vn;
205
206 /*
207 * Locate preexisting device
208 */
209
210 if ((vn = dev->si_drv1) == NULL)
211 vn = vnfindvn(dev);
212
213 /*
214 * Update si_bsize fields for device. This data will be overriden by
215 * the slice/parition code for vn accesses through partitions, and
216 * used directly if you open the 'whole disk' device.
217 *
218 * si_bsize_best must be reinitialized in case VN has been
219 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
220 */
221 dev->si_bsize_phys = vn->sc_secsize;
222 dev->si_bsize_best = vn->sc_secsize;
223 if (dev->si_bsize_best < VN_BSIZE_BEST)
224 dev->si_bsize_best = VN_BSIZE_BEST;
225
fef8985e 226 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
984263bc
MD
227 return (EACCES);
228
229 IFOPT(vn, VN_FOLLOW)
fef8985e
MD
230 printf("vnopen(%s, 0x%x, 0x%x)\n",
231 devtoname(dev), ap->a_oflags, ap->a_devtype);
984263bc
MD
232
233 /*
234 * Initialize label
235 */
236
237 IFOPT(vn, VN_LABELS) {
238 if (vn->sc_flags & VNF_INITED) {
239 struct disklabel label;
240
241 /* Build label for whole disk. */
242 bzero(&label, sizeof label);
243 label.d_secsize = vn->sc_secsize;
244 label.d_nsectors = 32;
245 label.d_ntracks = 64 / (vn->sc_secsize / DEV_BSIZE);
246 label.d_secpercyl = label.d_nsectors * label.d_ntracks;
247 label.d_ncylinders = vn->sc_size / label.d_secpercyl;
248 label.d_secperunit = vn->sc_size;
249 label.d_partitions[RAW_PART].p_size = vn->sc_size;
250
fef8985e 251 return (dsopen(dev, ap->a_devtype, 0, &vn->sc_slices, &label));
984263bc
MD
252 }
253 if (dkslice(dev) != WHOLE_DISK_SLICE ||
254 dkpart(dev) != RAW_PART ||
fef8985e 255 ap->a_devtype != S_IFCHR) {
984263bc
MD
256 return (ENXIO);
257 }
258 }
259 return(0);
260}
261
262/*
263 * vnstrategy:
264 *
265 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
266 * for vnode-backed vn's, and the new vm_pager_strategy() call for
267 * vm_object-backed vn's.
268 *
269 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
984263bc 270 */
fef8985e
MD
271static int
272vnstrategy(struct dev_strategy_args *ap)
984263bc 273{
b13267a5 274 cdev_t dev = ap->a_head.a_dev;
fef8985e 275 struct bio *bio = ap->a_bio;
81b5c339
MD
276 struct buf *bp;
277 struct bio *nbio;
984263bc
MD
278 int unit;
279 struct vn_softc *vn;
280 int error;
281
81b5c339
MD
282 unit = dkunit(dev);
283 if ((vn = dev->si_drv1) == NULL)
284 vn = vnfindvn(dev);
285
286 bp = bio->bio_buf;
984263bc
MD
287
288 IFOPT(vn, VN_DEBUG)
289 printf("vnstrategy(%p): unit %d\n", bp, unit);
290
291 if ((vn->sc_flags & VNF_INITED) == 0) {
292 bp->b_error = ENXIO;
293 bp->b_flags |= B_ERROR;
81b5c339 294 biodone(bio);
fef8985e 295 return(0);
984263bc
MD
296 }
297
298 bp->b_resid = bp->b_bcount;
299
300 IFOPT(vn, VN_LABELS) {
6f76c57e
HP
301 /*
302 * The vnode device is using disk/slice label support.
303 *
304 * The dscheck() function is called for validating the
305 * slices that exist ON the vnode device itself, and
306 * translate the "slice-relative" block number, again.
9a71d53f
MD
307 * dscheck() will call biodone() and return NULL if
308 * we are at EOF or beyond the device size.
6f76c57e 309 */
81b5c339
MD
310 if (vn->sc_slices == NULL) {
311 nbio = bio;
312 } else if ((nbio = dscheck(dev, bio, vn->sc_slices)) == NULL) {
4414f2c9 313 goto done;
984263bc
MD
314 }
315 } else {
316 int pbn; /* in sc_secsize chunks */
317 long sz; /* in sc_secsize chunks */
318
319 /*
320 * Check for required alignment. Transfers must be a valid
321 * multiple of the sector size.
322 */
323 if (bp->b_bcount % vn->sc_secsize != 0 ||
54078292 324 bio->bio_offset % vn->sc_secsize != 0) {
4414f2c9 325 goto bad;
984263bc
MD
326 }
327
54078292 328 pbn = bio->bio_offset / vn->sc_secsize;
984263bc
MD
329 sz = howmany(bp->b_bcount, vn->sc_secsize);
330
331 /*
4414f2c9 332 * Check for an illegal pbn or EOF truncation
984263bc 333 */
4414f2c9
MD
334 if (pbn < 0)
335 goto bad;
984263bc 336 if (pbn + sz > vn->sc_size) {
4414f2c9
MD
337 if (pbn > vn->sc_size || (bp->b_flags & B_BNOCLIP))
338 goto bad;
339 if (pbn == vn->sc_size) {
340 bp->b_resid = bp->b_bcount;
341 bp->b_flags |= B_INVAL;
342 goto done;
343 }
984263bc 344 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
984263bc 345 }
81b5c339 346 nbio = push_bio(bio);
54078292 347 nbio->bio_offset = pbn * vn->sc_secsize;
984263bc
MD
348 }
349
81b5c339
MD
350 /*
351 * Use the translated nbio from this point on
352 */
10f3fee5 353 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
984263bc 354 /*
10f3fee5 355 * Freeblks is not handled for vnode-backed elements yet.
984263bc 356 */
9a71d53f 357 bp->b_resid = 0;
4414f2c9 358 /* operation complete */
984263bc
MD
359 } else if (vn->sc_vp) {
360 /*
361 * VNODE I/O
362 *
363 * If an error occurs, we set B_ERROR but we do not set
364 * B_INVAL because (for a write anyway), the buffer is
365 * still valid.
366 */
367 struct uio auio;
368 struct iovec aiov;
369
370 bzero(&auio, sizeof(auio));
371
372 aiov.iov_base = bp->b_data;
373 aiov.iov_len = bp->b_bcount;
374 auio.uio_iov = &aiov;
375 auio.uio_iovcnt = 1;
54078292 376 auio.uio_offset = nbio->bio_offset;
984263bc 377 auio.uio_segflg = UIO_SYSSPACE;
10f3fee5 378 if (bp->b_cmd == BUF_CMD_READ)
984263bc
MD
379 auio.uio_rw = UIO_READ;
380 else
381 auio.uio_rw = UIO_WRITE;
382 auio.uio_resid = bp->b_bcount;
dadab5e9 383 auio.uio_td = curthread;
ca466bae 384 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
10f3fee5 385 if (bp->b_cmd == BUF_CMD_READ)
984263bc
MD
386 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
387 else
388 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
a11aaa81 389 vn_unlock(vn->sc_vp);
984263bc 390 bp->b_resid = auio.uio_resid;
984263bc
MD
391 if (error) {
392 bp->b_error = error;
393 bp->b_flags |= B_ERROR;
394 }
4414f2c9 395 /* operation complete */
984263bc
MD
396 } else if (vn->sc_object) {
397 /*
4414f2c9 398 * OBJT_SWAP I/O (handles read, write, freebuf)
984263bc 399 *
4414f2c9
MD
400 * We have nothing to do if freeing blocks on a reserved
401 * swap area, othrewise execute the op.
984263bc 402 */
10f3fee5 403 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
4414f2c9
MD
404 bp->b_resid = 0;
405 /* operation complete */
984263bc 406 } else {
81b5c339 407 vm_pager_strategy(vn->sc_object, nbio);
fef8985e 408 return(0);
4414f2c9 409 /* NOT REACHED */
984263bc
MD
410 }
411 } else {
4414f2c9
MD
412 bp->b_resid = bp->b_bcount;
413 bp->b_flags |= B_ERROR | B_INVAL;
984263bc 414 bp->b_error = EINVAL;
4414f2c9 415 /* operation complete */
984263bc 416 }
4414f2c9 417 biodone(nbio);
fef8985e 418 return(0);
4414f2c9
MD
419
420 /*
421 * Shortcuts / check failures on the original bio (not nbio).
422 */
423bad:
424 bp->b_error = EINVAL;
4414f2c9
MD
425 bp->b_flags |= B_ERROR | B_INVAL;
426done:
427 biodone(bio);
fef8985e 428 return(0);
984263bc
MD
429}
430
431/* ARGSUSED */
432static int
fef8985e 433vnioctl(struct dev_ioctl_args *ap)
984263bc 434{
b13267a5 435 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
436 struct vn_softc *vn;
437 struct vn_ioctl *vio;
438 int error;
439 u_long *f;
440
441 vn = dev->si_drv1;
fef8985e
MD
442 IFOPT(vn,VN_FOLLOW) {
443 printf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
444 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
984263bc 445 dkunit(dev));
fef8985e 446 }
984263bc 447
fef8985e 448 switch (ap->a_cmd) {
984263bc
MD
449 case VNIOCATTACH:
450 case VNIOCDETACH:
451 case VNIOCGSET:
452 case VNIOCGCLEAR:
453 case VNIOCUSET:
454 case VNIOCUCLEAR:
455 goto vn_specific;
456 }
457
458 IFOPT(vn,VN_LABELS) {
459 if (vn->sc_slices != NULL) {
fef8985e
MD
460 error = dsioctl(dev, ap->a_cmd, ap->a_data,
461 ap->a_fflag, &vn->sc_slices);
984263bc
MD
462 if (error != ENOIOCTL)
463 return (error);
464 }
465 if (dkslice(dev) != WHOLE_DISK_SLICE ||
466 dkpart(dev) != RAW_PART)
467 return (ENOTTY);
468 }
469
470 vn_specific:
471
fef8985e 472 error = suser_cred(ap->a_cred, 0);
984263bc
MD
473 if (error)
474 return (error);
475
fef8985e
MD
476 vio = (struct vn_ioctl *)ap->a_data;
477 f = (u_long*)ap->a_data;
984263bc 478
fef8985e 479 switch (ap->a_cmd) {
984263bc
MD
480 case VNIOCATTACH:
481 if (vn->sc_flags & VNF_INITED)
482 return(EBUSY);
483
484 if (vio->vn_file == NULL)
fef8985e 485 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
984263bc 486 else
fef8985e 487 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
984263bc
MD
488 break;
489
490 case VNIOCDETACH:
491 if ((vn->sc_flags & VNF_INITED) == 0)
492 return(ENXIO);
493 /*
494 * XXX handle i/o in progress. Return EBUSY, or wait, or
495 * flush the i/o.
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)
500 */
501 vnclear(vn);
502 IFOPT(vn, VN_FOLLOW)
503 printf("vnioctl: CLRed\n");
504 break;
505
506 case VNIOCGSET:
507 vn_options |= *f;
508 *f = vn_options;
509 break;
510
511 case VNIOCGCLEAR:
512 vn_options &= ~(*f);
513 *f = vn_options;
514 break;
515
516 case VNIOCUSET:
517 vn->sc_options |= *f;
518 *f = vn->sc_options;
519 break;
520
521 case VNIOCUCLEAR:
522 vn->sc_options &= ~(*f);
523 *f = vn->sc_options;
524 break;
525
526 default:
527 error = ENOTTY;
528 break;
529 }
530 return(error);
531}
532
533/*
534 * vniocattach_file:
535 *
536 * Attach a file to a VN partition. Return the size in the vn_size
537 * field.
538 */
539
540static int
b13267a5 541vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
fef8985e 542 int flag, struct ucred *cred)
984263bc
MD
543{
544 struct vattr vattr;
fad57d0e 545 struct nlookupdata nd;
984263bc 546 int error, flags;
fad57d0e 547 struct vnode *vp;
984263bc
MD
548
549 flags = FREAD|FWRITE;
fad57d0e
MD
550 error = nlookup_init(&nd, vio->vn_file,
551 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
552 if (error)
553 return (error);
554 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
984263bc 555 if (error != EACCES && error != EPERM && error != EROFS)
fad57d0e 556 goto done;
984263bc 557 flags &= ~FWRITE;
fad57d0e
MD
558 nlookup_done(&nd);
559 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
984263bc
MD
560 if (error)
561 return (error);
fad57d0e
MD
562 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
563 goto done;
984263bc 564 }
fad57d0e
MD
565 vp = nd.nl_open_vp;
566 if (vp->v_type != VREG ||
87de5057 567 (error = VOP_GETATTR(vp, &vattr))) {
fad57d0e
MD
568 if (error == 0)
569 error = EINVAL;
570 goto done;
984263bc 571 }
a11aaa81 572 vn_unlock(vp);
984263bc 573 vn->sc_secsize = DEV_BSIZE;
fad57d0e
MD
574 vn->sc_vp = vp;
575 nd.nl_open_vp = NULL;
984263bc
MD
576
577 /*
578 * If the size is specified, override the file attributes. Note that
579 * the vn_size argument is in PAGE_SIZE sized blocks.
580 */
581 if (vio->vn_size)
582 vn->sc_size = (quad_t)vio->vn_size * PAGE_SIZE / vn->sc_secsize;
583 else
584 vn->sc_size = vattr.va_size / vn->sc_secsize;
fef8985e 585 error = vnsetcred(vn, cred);
984263bc 586 if (error) {
fad57d0e 587 vn->sc_vp = NULL;
87de5057 588 vn_close(vp, flags);
fad57d0e 589 goto done;
984263bc
MD
590 }
591 vn->sc_flags |= VNF_INITED;
592 if (flags == FREAD)
593 vn->sc_flags |= VNF_READONLY;
594 IFOPT(vn, VN_LABELS) {
595 /*
596 * Reopen so that `ds' knows which devices are open.
597 * If this is the first VNIOCSET, then we've
598 * guaranteed that the device is the cdev and that
599 * no other slices or labels are open. Otherwise,
600 * we rely on VNIOCCLR not being abused.
601 */
fef8985e 602 error = dev_dopen(dev, flag, S_IFCHR, cred);
984263bc
MD
603 if (error)
604 vnclear(vn);
605 }
606 IFOPT(vn, VN_FOLLOW)
607 printf("vnioctl: SET vp %p size %x blks\n",
608 vn->sc_vp, vn->sc_size);
fad57d0e
MD
609done:
610 nlookup_done(&nd);
611 return(error);
984263bc
MD
612}
613
614/*
615 * vniocattach_swap:
616 *
617 * Attach swap backing store to a VN partition of the size specified
618 * in vn_size.
619 */
620
621static int
b13267a5 622vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
fef8985e 623 int flag, struct ucred *cred)
984263bc
MD
624{
625 int error;
626
627 /*
628 * Range check. Disallow negative sizes or any size less then the
629 * size of a page. Then round to a page.
630 */
631
632 if (vio->vn_size <= 0)
633 return(EDOM);
634
635 /*
636 * Allocate an OBJT_SWAP object.
637 *
638 * sc_secsize is PAGE_SIZE'd
639 *
640 * vio->vn_size is in PAGE_SIZE'd chunks.
641 * sc_size must be in PAGE_SIZE'd chunks.
642 * Note the truncation.
643 */
644
645 vn->sc_secsize = PAGE_SIZE;
646 vn->sc_size = vio->vn_size;
57f7b636
MD
647 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
648 vn->sc_secsize * (off_t)vio->vn_size,
649 VM_PROT_DEFAULT, 0);
984263bc
MD
650 IFOPT(vn, VN_RESERVE) {
651 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
652 vm_pager_deallocate(vn->sc_object);
653 vn->sc_object = NULL;
654 return(EDOM);
655 }
656 }
657 vn->sc_flags |= VNF_INITED;
658
fef8985e 659 error = vnsetcred(vn, cred);
984263bc
MD
660 if (error == 0) {
661 IFOPT(vn, VN_LABELS) {
662 /*
663 * Reopen so that `ds' knows which devices are open.
664 * If this is the first VNIOCSET, then we've
665 * guaranteed that the device is the cdev and that
666 * no other slices or labels are open. Otherwise,
667 * we rely on VNIOCCLR not being abused.
668 */
fef8985e 669 error = dev_dopen(dev, flag, S_IFCHR, cred);
984263bc
MD
670 }
671 }
672 if (error == 0) {
673 IFOPT(vn, VN_FOLLOW) {
674 printf("vnioctl: SET vp %p size %x\n",
675 vn->sc_vp, vn->sc_size);
676 }
677 }
678 if (error)
679 vnclear(vn);
680 return(error);
681}
682
683/*
684 * Duplicate the current processes' credentials. Since we are called only
685 * as the result of a SET ioctl and only root can do that, any future access
686 * to this "disk" is essentially as root. Note that credentials may change
687 * if some other uid can write directly to the mapped file (NFS).
688 */
689int
690vnsetcred(struct vn_softc *vn, struct ucred *cred)
691{
692 char *tmpbuf;
693 int error = 0;
694
695 /*
696 * Set credits in our softc
697 */
698
699 if (vn->sc_cred)
700 crfree(vn->sc_cred);
701 vn->sc_cred = crdup(cred);
702
703 /*
704 * Horrible kludge to establish credentials for NFS XXX.
705 */
706
707 if (vn->sc_vp) {
708 struct uio auio;
709 struct iovec aiov;
710
efda3bd0 711 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
984263bc
MD
712 bzero(&auio, sizeof(auio));
713
714 aiov.iov_base = tmpbuf;
715 aiov.iov_len = vn->sc_secsize;
716 auio.uio_iov = &aiov;
717 auio.uio_iovcnt = 1;
718 auio.uio_offset = 0;
719 auio.uio_rw = UIO_READ;
720 auio.uio_segflg = UIO_SYSSPACE;
721 auio.uio_resid = aiov.iov_len;
ca466bae 722 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
984263bc 723 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
a11aaa81 724 vn_unlock(vn->sc_vp);
efda3bd0 725 kfree(tmpbuf, M_TEMP);
984263bc
MD
726 }
727 return (error);
728}
729
730void
731vnclear(struct vn_softc *vn)
732{
984263bc
MD
733 IFOPT(vn, VN_FOLLOW)
734 printf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
735 if (vn->sc_slices != NULL)
736 dsgone(&vn->sc_slices);
737 vn->sc_flags &= ~VNF_INITED;
738 if (vn->sc_vp != NULL) {
87de5057
MD
739 vn_close(vn->sc_vp,
740 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
984263bc
MD
741 vn->sc_vp = NULL;
742 }
743 vn->sc_flags &= ~VNF_READONLY;
744 if (vn->sc_cred) {
745 crfree(vn->sc_cred);
746 vn->sc_cred = NULL;
747 }
748 if (vn->sc_object != NULL) {
749 vm_pager_deallocate(vn->sc_object);
750 vn->sc_object = NULL;
751 }
752 vn->sc_size = 0;
753}
754
fef8985e
MD
755static int
756vnsize(struct dev_psize_args *ap)
984263bc 757{
b13267a5 758 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
759 struct vn_softc *vn;
760
761 vn = dev->si_drv1;
762 if (!vn)
fef8985e 763 return(ENXIO);
984263bc 764 if ((vn->sc_flags & VNF_INITED) == 0)
fef8985e
MD
765 return(ENXIO);
766 ap->a_result = vn->sc_size;
767 return(0);
984263bc
MD
768}
769
770static int
771vn_modevent(module_t mod, int type, void *data)
772{
773 struct vn_softc *vn;
b13267a5 774 cdev_t dev;
984263bc
MD
775
776 switch (type) {
777 case MOD_LOAD:
fef8985e 778 dev_ops_add(&vn_ops, 0, 0);
984263bc 779 break;
984263bc
MD
780 case MOD_UNLOAD:
781 /* fall through */
782 case MOD_SHUTDOWN:
e2f51e29 783 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
984263bc
MD
784 SLIST_REMOVE_HEAD(&vn_list, sc_list);
785 if (vn->sc_flags & VNF_INITED)
786 vnclear(vn);
b13267a5 787 /* Cleanup all cdev_t's that refer to this unit */
984263bc
MD
788 while ((dev = vn->sc_devlist) != NULL) {
789 vn->sc_devlist = dev->si_drv2;
790 dev->si_drv1 = dev->si_drv2 = NULL;
e4c9c0c8 791 destroy_dev(dev);
984263bc 792 }
efda3bd0 793 kfree(vn, M_DEVBUF);
984263bc 794 }
fef8985e 795 dev_ops_remove(&vn_ops, 0, 0);
984263bc
MD
796 break;
797 default:
798 break;
799 }
800 return 0;
801}
802
803DEV_MODULE(vn, vn_modevent, 0);