VN - Remove use of IO_DIRECT, Improve lock performance for reads
[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 $
4ecf7cc9 42 * $DragonFly: src/sys/dev/disk/vn/vn.c,v 1.38 2008/07/01 02:02:53 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
95db3aac 60#include "use_vn.h"
984263bc
MD
61#include <sys/param.h>
62#include <sys/systm.h>
63#include <sys/kernel.h>
984263bc 64#include <sys/proc.h>
895c1f85 65#include <sys/priv.h>
fad57d0e 66#include <sys/nlookup.h>
984263bc
MD
67#include <sys/buf.h>
68#include <sys/malloc.h>
69#include <sys/mount.h>
70#include <sys/vnode.h>
71#include <sys/fcntl.h>
72#include <sys/conf.h>
984263bc 73#include <sys/diskslice.h>
84f8b009 74#include <sys/disk.h>
984263bc 75#include <sys/stat.h>
984263bc
MD
76#include <sys/module.h>
77#include <sys/vnioctl.h>
78
79#include <vm/vm.h>
80#include <vm/vm_object.h>
81#include <vm/vm_page.h>
82#include <vm/vm_pager.h>
83#include <vm/vm_pageout.h>
84#include <vm/swap_pager.h>
85#include <vm/vm_extern.h>
86#include <vm/vm_zone.h>
2c1e28dd 87#include <sys/devfs.h>
984263bc
MD
88
89static d_ioctl_t vnioctl;
90static d_open_t vnopen;
91static d_close_t vnclose;
92static d_psize_t vnsize;
93static d_strategy_t vnstrategy;
8be7edad 94static d_clone_t vnclone;
cc80c90e 95
8be7edad 96DEVFS_DECLARE_CLONE_BITMAP(vn);
95db3aac
AH
97
98#if NVN <= 1
8be7edad 99#define VN_PREALLOCATED_UNITS 4
95db3aac
AH
100#else
101#define VN_PREALLOCATED_UNITS NVN
102#endif
984263bc
MD
103
104#define CDEV_MAJOR 43
984263bc
MD
105
106#define VN_BSIZE_BEST 8192
107
108/*
fef8985e 109 * dev_ops
984263bc 110 * D_DISK we want to look like a disk
10f3fee5 111 * D_CANFREE We support BUF_CMD_FREEBLKS
984263bc
MD
112 */
113
fef8985e
MD
114static struct dev_ops vn_ops = {
115 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
116 .d_open = vnopen,
117 .d_close = vnclose,
118 .d_read = physread,
119 .d_write = physwrite,
120 .d_ioctl = vnioctl,
121 .d_strategy = vnstrategy,
122 .d_psize = vnsize
984263bc
MD
123};
124
984263bc
MD
125struct vn_softc {
126 int sc_unit;
127 int sc_flags; /* flags */
e0fc5693 128 u_int64_t sc_size; /* size of vn, sc_secsize scale */
984263bc 129 int sc_secsize; /* sector size */
8be7edad 130 struct disk sc_disk;
984263bc
MD
131 struct vnode *sc_vp; /* vnode if not NULL */
132 vm_object_t sc_object; /* backing object if not NULL */
133 struct ucred *sc_cred; /* credentials */
134 int sc_maxactive; /* max # of active requests */
135 struct buf sc_tab; /* transfer queue */
136 u_long sc_options; /* options */
b13267a5 137 cdev_t sc_devlist; /* devices that refer to this unit */
984263bc
MD
138 SLIST_ENTRY(vn_softc) sc_list;
139};
140
141static SLIST_HEAD(, vn_softc) vn_list;
142
143/* sc_flags */
144#define VNF_INITED 0x01
145#define VNF_READONLY 0x02
146
147static u_long vn_options;
148
149#define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
150#define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
151
152static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
153static void vnclear (struct vn_softc *vn);
30c1fde0 154static int vnget (cdev_t dev, struct vn_softc *vn , struct vn_user *vnu);
984263bc 155static int vn_modevent (module_t, int, void *);
b13267a5
MD
156static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
157static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
cc80c90e
AH
158static cdev_t vn_create(int unit, struct devfs_bitmap *bitmap);
159
160static int
161vnclone(struct dev_clone_args *ap)
162{
163 int unit;
164
165 unit = devfs_clone_bitmap_get(&DEVFS_CLONE_BITMAP(vn), 0);
166 ap->a_dev = vn_create(unit, &DEVFS_CLONE_BITMAP(vn));
167
168 return 0;
169}
984263bc
MD
170
171static int
fef8985e 172vnclose(struct dev_close_args *ap)
984263bc 173{
984263bc
MD
174 return (0);
175}
176
8be7edad
MD
177static struct vn_softc *
178vncreatevn(void)
179{
180 struct vn_softc *vn;
181
182 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
183 return vn;
184}
185
186static void
187vninitvn(struct vn_softc *vn, cdev_t dev)
188{
189 int unit;
190
191 KKASSERT(vn != NULL);
192 KKASSERT(dev != NULL);
193 unit = dkunit(dev);
194
195 vn->sc_unit = unit;
196 dev->si_drv1 = vn;
197 vn->sc_devlist = dev;
198 if (vn->sc_devlist->si_drv1 == NULL) {
199 reference_dev(vn->sc_devlist);
200 vn->sc_devlist->si_drv1 = vn;
201 vn->sc_devlist->si_drv2 = NULL;
202 }
203 if (vn->sc_devlist != dev) {
204 dev->si_drv1 = vn;
205 dev->si_drv2 = vn->sc_devlist;
206 vn->sc_devlist = dev;
207 reference_dev(dev);
208 }
209 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
210}
211
e2f51e29
MD
212/*
213 * Called only when si_drv1 is NULL. Locate the associated vn node and
214 * attach the device to it.
215 */
984263bc 216static struct vn_softc *
b13267a5 217vnfindvn(cdev_t dev)
984263bc
MD
218{
219 int unit;
220 struct vn_softc *vn;
221
222 unit = dkunit(dev);
e4c9c0c8
MD
223 SLIST_FOREACH(vn, &vn_list, sc_list) {
224 if (vn->sc_unit == unit) {
e2f51e29 225 dev->si_drv1 = vn;
e4c9c0c8
MD
226 dev->si_drv2 = vn->sc_devlist;
227 vn->sc_devlist = dev;
228 reference_dev(dev);
e4c9c0c8 229 break;
984263bc
MD
230 }
231 }
8be7edad
MD
232
233 KKASSERT(vn != NULL);
234
235 return vn;
984263bc
MD
236}
237
238static int
fef8985e 239vnopen(struct dev_open_args *ap)
984263bc 240{
b13267a5 241 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
242 struct vn_softc *vn;
243
244 /*
245 * Locate preexisting device
246 */
247
248 if ((vn = dev->si_drv1) == NULL)
249 vn = vnfindvn(dev);
250
251 /*
252 * Update si_bsize fields for device. This data will be overriden by
253 * the slice/parition code for vn accesses through partitions, and
254 * used directly if you open the 'whole disk' device.
255 *
256 * si_bsize_best must be reinitialized in case VN has been
257 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
258 */
259 dev->si_bsize_phys = vn->sc_secsize;
260 dev->si_bsize_best = vn->sc_secsize;
261 if (dev->si_bsize_best < VN_BSIZE_BEST)
262 dev->si_bsize_best = VN_BSIZE_BEST;
263
fef8985e 264 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
984263bc
MD
265 return (EACCES);
266
267 IFOPT(vn, VN_FOLLOW)
e3869ec7 268 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
fef8985e 269 devtoname(dev), ap->a_oflags, ap->a_devtype);
984263bc 270
984263bc
MD
271 return(0);
272}
273
274/*
275 * vnstrategy:
276 *
277 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
107e9bcc 278 * for vnode-backed vn's, and the swap_pager_strategy() call for
984263bc 279 * vm_object-backed vn's.
984263bc 280 */
fef8985e
MD
281static int
282vnstrategy(struct dev_strategy_args *ap)
984263bc 283{
b13267a5 284 cdev_t dev = ap->a_head.a_dev;
fef8985e 285 struct bio *bio = ap->a_bio;
81b5c339
MD
286 struct buf *bp;
287 struct bio *nbio;
984263bc
MD
288 int unit;
289 struct vn_softc *vn;
290 int error;
291
81b5c339
MD
292 unit = dkunit(dev);
293 if ((vn = dev->si_drv1) == NULL)
294 vn = vnfindvn(dev);
295
296 bp = bio->bio_buf;
984263bc
MD
297
298 IFOPT(vn, VN_DEBUG)
e3869ec7 299 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
984263bc
MD
300
301 if ((vn->sc_flags & VNF_INITED) == 0) {
302 bp->b_error = ENXIO;
303 bp->b_flags |= B_ERROR;
81b5c339 304 biodone(bio);
fef8985e 305 return(0);
984263bc
MD
306 }
307
308 bp->b_resid = bp->b_bcount;
309
8be7edad
MD
310 /*
311 * The vnode device is using disk/slice label support.
312 *
313 * The dscheck() function is called for validating the
314 * slices that exist ON the vnode device itself, and
315 * translate the "slice-relative" block number, again.
316 * dscheck() will call biodone() and return NULL if
317 * we are at EOF or beyond the device size.
318 */
984263bc 319
8be7edad 320 nbio = bio;
984263bc 321
81b5c339
MD
322 /*
323 * Use the translated nbio from this point on
324 */
10f3fee5 325 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
984263bc 326 /*
10f3fee5 327 * Freeblks is not handled for vnode-backed elements yet.
984263bc 328 */
9a71d53f 329 bp->b_resid = 0;
4414f2c9 330 /* operation complete */
984263bc
MD
331 } else if (vn->sc_vp) {
332 /*
333 * VNODE I/O
334 *
335 * If an error occurs, we set B_ERROR but we do not set
336 * B_INVAL because (for a write anyway), the buffer is
337 * still valid.
338 */
339 struct uio auio;
340 struct iovec aiov;
341
342 bzero(&auio, sizeof(auio));
343
344 aiov.iov_base = bp->b_data;
345 aiov.iov_len = bp->b_bcount;
346 auio.uio_iov = &aiov;
347 auio.uio_iovcnt = 1;
54078292 348 auio.uio_offset = nbio->bio_offset;
984263bc 349 auio.uio_segflg = UIO_SYSSPACE;
10f3fee5 350 if (bp->b_cmd == BUF_CMD_READ)
984263bc
MD
351 auio.uio_rw = UIO_READ;
352 else
353 auio.uio_rw = UIO_WRITE;
354 auio.uio_resid = bp->b_bcount;
dadab5e9 355 auio.uio_td = curthread;
b527c4c5
MD
356
357 /*
358 * Don't use IO_DIRECT here, it really gets in the way
359 * due to typical blocksize differences between the
360 * fs backing the VN device and whatever is running on
361 * the VN device.
362 */
363 if (bp->b_cmd == BUF_CMD_READ) {
364 vn_lock(vn->sc_vp, LK_SHARED | LK_RETRY);
365 error = VOP_READ(vn->sc_vp, &auio, IO_RECURSE,
366 vn->sc_cred);
367 } else {
368 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
369 error = VOP_WRITE(vn->sc_vp, &auio, IO_RECURSE,
370 vn->sc_cred);
371 }
a11aaa81 372 vn_unlock(vn->sc_vp);
984263bc 373 bp->b_resid = auio.uio_resid;
984263bc
MD
374 if (error) {
375 bp->b_error = error;
376 bp->b_flags |= B_ERROR;
377 }
4414f2c9 378 /* operation complete */
984263bc
MD
379 } else if (vn->sc_object) {
380 /*
4414f2c9 381 * OBJT_SWAP I/O (handles read, write, freebuf)
984263bc 382 *
4414f2c9
MD
383 * We have nothing to do if freeing blocks on a reserved
384 * swap area, othrewise execute the op.
984263bc 385 */
10f3fee5 386 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
4414f2c9
MD
387 bp->b_resid = 0;
388 /* operation complete */
984263bc 389 } else {
107e9bcc 390 swap_pager_strategy(vn->sc_object, nbio);
fef8985e 391 return(0);
4414f2c9 392 /* NOT REACHED */
984263bc
MD
393 }
394 } else {
4414f2c9
MD
395 bp->b_resid = bp->b_bcount;
396 bp->b_flags |= B_ERROR | B_INVAL;
984263bc 397 bp->b_error = EINVAL;
4414f2c9 398 /* operation complete */
984263bc 399 }
4414f2c9 400 biodone(nbio);
fef8985e 401 return(0);
984263bc
MD
402}
403
404/* ARGSUSED */
405static int
fef8985e 406vnioctl(struct dev_ioctl_args *ap)
984263bc 407{
b13267a5 408 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
409 struct vn_softc *vn;
410 struct vn_ioctl *vio;
411 int error;
412 u_long *f;
413
414 vn = dev->si_drv1;
fef8985e 415 IFOPT(vn,VN_FOLLOW) {
e3869ec7 416 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
fef8985e 417 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
984263bc 418 dkunit(dev));
fef8985e 419 }
984263bc 420
fef8985e 421 switch (ap->a_cmd) {
984263bc
MD
422 case VNIOCATTACH:
423 case VNIOCDETACH:
424 case VNIOCGSET:
425 case VNIOCGCLEAR:
30c1fde0 426 case VNIOCGET:
984263bc
MD
427 case VNIOCUSET:
428 case VNIOCUCLEAR:
429 goto vn_specific;
430 }
431
8be7edad
MD
432#if 0
433 if (dkslice(dev) != WHOLE_DISK_SLICE ||
434 dkpart(dev) != WHOLE_SLICE_PART)
435 return (ENOTTY);
436#endif
984263bc
MD
437
438 vn_specific:
439
895c1f85 440 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
984263bc
MD
441 if (error)
442 return (error);
443
fef8985e
MD
444 vio = (struct vn_ioctl *)ap->a_data;
445 f = (u_long*)ap->a_data;
984263bc 446
fef8985e 447 switch (ap->a_cmd) {
984263bc
MD
448 case VNIOCATTACH:
449 if (vn->sc_flags & VNF_INITED)
450 return(EBUSY);
451
452 if (vio->vn_file == NULL)
fef8985e 453 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
984263bc 454 else
fef8985e 455 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
984263bc
MD
456 break;
457
458 case VNIOCDETACH:
459 if ((vn->sc_flags & VNF_INITED) == 0)
460 return(ENXIO);
461 /*
462 * XXX handle i/o in progress. Return EBUSY, or wait, or
463 * flush the i/o.
464 * XXX handle multiple opens of the device. Return EBUSY,
465 * or revoke the fd's.
466 * How are these problems handled for removable and failing
467 * hardware devices? (Hint: They are not)
468 */
c9f4065d
SK
469 if (count_dev(vn->sc_devlist) > 1)
470 return (EBUSY);
471
984263bc
MD
472 vnclear(vn);
473 IFOPT(vn, VN_FOLLOW)
e3869ec7 474 kprintf("vnioctl: CLRed\n");
cc80c90e
AH
475
476 if (dkunit(dev) >= VN_PREALLOCATED_UNITS) {
477 devfs_clone_bitmap_put(&DEVFS_CLONE_BITMAP(vn), dkunit(dev));
478 disk_destroy(&vn->sc_disk);
479 }
480
984263bc
MD
481 break;
482
30c1fde0
CT
483 case VNIOCGET:
484 error = vnget(dev, vn, (struct vn_user *) ap->a_data);
485 break;
486
984263bc
MD
487 case VNIOCGSET:
488 vn_options |= *f;
489 *f = vn_options;
490 break;
491
492 case VNIOCGCLEAR:
493 vn_options &= ~(*f);
494 *f = vn_options;
495 break;
496
497 case VNIOCUSET:
498 vn->sc_options |= *f;
499 *f = vn->sc_options;
500 break;
501
502 case VNIOCUCLEAR:
503 vn->sc_options &= ~(*f);
504 *f = vn->sc_options;
505 break;
506
507 default:
508 error = ENOTTY;
509 break;
510 }
511 return(error);
512}
513
514/*
515 * vniocattach_file:
516 *
517 * Attach a file to a VN partition. Return the size in the vn_size
518 * field.
519 */
520
521static int
b13267a5 522vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
fef8985e 523 int flag, struct ucred *cred)
984263bc
MD
524{
525 struct vattr vattr;
fad57d0e 526 struct nlookupdata nd;
984263bc 527 int error, flags;
fad57d0e 528 struct vnode *vp;
8be7edad 529 struct disk_info info;
984263bc
MD
530
531 flags = FREAD|FWRITE;
fad57d0e
MD
532 error = nlookup_init(&nd, vio->vn_file,
533 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
534 if (error)
535 return (error);
536 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
984263bc 537 if (error != EACCES && error != EPERM && error != EROFS)
fad57d0e 538 goto done;
984263bc 539 flags &= ~FWRITE;
fad57d0e
MD
540 nlookup_done(&nd);
541 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
984263bc
MD
542 if (error)
543 return (error);
fad57d0e
MD
544 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
545 goto done;
984263bc 546 }
fad57d0e
MD
547 vp = nd.nl_open_vp;
548 if (vp->v_type != VREG ||
87de5057 549 (error = VOP_GETATTR(vp, &vattr))) {
fad57d0e
MD
550 if (error == 0)
551 error = EINVAL;
552 goto done;
984263bc 553 }
a11aaa81 554 vn_unlock(vp);
984263bc 555 vn->sc_secsize = DEV_BSIZE;
fad57d0e
MD
556 vn->sc_vp = vp;
557 nd.nl_open_vp = NULL;
984263bc
MD
558
559 /*
560 * If the size is specified, override the file attributes. Note that
561 * the vn_size argument is in PAGE_SIZE sized blocks.
562 */
563 if (vio->vn_size)
e0fc5693 564 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
984263bc
MD
565 else
566 vn->sc_size = vattr.va_size / vn->sc_secsize;
fef8985e 567 error = vnsetcred(vn, cred);
984263bc 568 if (error) {
fad57d0e 569 vn->sc_vp = NULL;
87de5057 570 vn_close(vp, flags);
fad57d0e 571 goto done;
984263bc
MD
572 }
573 vn->sc_flags |= VNF_INITED;
574 if (flags == FREAD)
575 vn->sc_flags |= VNF_READONLY;
8be7edad
MD
576
577 /*
578 * Set the disk info so that probing is triggered
579 */
580 bzero(&info, sizeof(struct disk_info));
581 info.d_media_blksize = vn->sc_secsize;
582 info.d_media_blocks = vn->sc_size;
583 /*
584 * reserve mbr sector for backwards compatibility
585 * when no slices exist.
586 */
587 info.d_dsflags = DSO_COMPATMBR;
588 info.d_secpertrack = 32;
589 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
590 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
591 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
592 disk_setdiskinfo_sync(&vn->sc_disk, &info);
593
594 error = dev_dopen(dev, flag, S_IFCHR, cred);
595 if (error)
596 vnclear(vn);
597
984263bc 598 IFOPT(vn, VN_FOLLOW)
e0fc5693 599 kprintf("vnioctl: SET vp %p size %llx blks\n",
665d7d25 600 vn->sc_vp, (long long)vn->sc_size);
fad57d0e
MD
601done:
602 nlookup_done(&nd);
603 return(error);
984263bc
MD
604}
605
606/*
607 * vniocattach_swap:
608 *
609 * Attach swap backing store to a VN partition of the size specified
610 * in vn_size.
611 */
612
613static int
b13267a5 614vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
fef8985e 615 int flag, struct ucred *cred)
984263bc
MD
616{
617 int error;
8be7edad 618 struct disk_info info;
984263bc
MD
619
620 /*
621 * Range check. Disallow negative sizes or any size less then the
622 * size of a page. Then round to a page.
623 */
624
625 if (vio->vn_size <= 0)
626 return(EDOM);
627
628 /*
629 * Allocate an OBJT_SWAP object.
630 *
631 * sc_secsize is PAGE_SIZE'd
632 *
633 * vio->vn_size is in PAGE_SIZE'd chunks.
634 * sc_size must be in PAGE_SIZE'd chunks.
635 * Note the truncation.
636 */
637
638 vn->sc_secsize = PAGE_SIZE;
639 vn->sc_size = vio->vn_size;
5a648714
MD
640 vn->sc_object = swap_pager_alloc(NULL,
641 vn->sc_secsize * (off_t)vio->vn_size,
642 VM_PROT_DEFAULT, 0);
984263bc
MD
643 IFOPT(vn, VN_RESERVE) {
644 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
645 vm_pager_deallocate(vn->sc_object);
646 vn->sc_object = NULL;
647 return(EDOM);
648 }
649 }
650 vn->sc_flags |= VNF_INITED;
651
fef8985e 652 error = vnsetcred(vn, cred);
984263bc 653 if (error == 0) {
8be7edad
MD
654 /*
655 * Set the disk info so that probing is triggered
656 */
657 bzero(&info, sizeof(struct disk_info));
658 info.d_media_blksize = vn->sc_secsize;
659 info.d_media_blocks = vn->sc_size;
660 /*
661 * reserve mbr sector for backwards compatibility
662 * when no slices exist.
663 */
664 info.d_dsflags = DSO_COMPATMBR;
665 info.d_secpertrack = 32;
666 info.d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
667 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
668 info.d_ncylinders = vn->sc_size / info.d_secpercyl;
669 disk_setdiskinfo_sync(&vn->sc_disk, &info);
670
671 error = dev_dopen(dev, flag, S_IFCHR, cred);
984263bc
MD
672 }
673 if (error == 0) {
674 IFOPT(vn, VN_FOLLOW) {
e0fc5693 675 kprintf("vnioctl: SET vp %p size %llx\n",
665d7d25 676 vn->sc_vp, (long long)vn->sc_size);
984263bc
MD
677 }
678 }
679 if (error)
680 vnclear(vn);
681 return(error);
682}
683
684/*
685 * Duplicate the current processes' credentials. Since we are called only
686 * as the result of a SET ioctl and only root can do that, any future access
687 * to this "disk" is essentially as root. Note that credentials may change
688 * if some other uid can write directly to the mapped file (NFS).
689 */
690int
691vnsetcred(struct vn_softc *vn, struct ucred *cred)
692{
693 char *tmpbuf;
694 int error = 0;
695
696 /*
697 * Set credits in our softc
698 */
699
700 if (vn->sc_cred)
701 crfree(vn->sc_cred);
702 vn->sc_cred = crdup(cred);
703
704 /*
705 * Horrible kludge to establish credentials for NFS XXX.
706 */
707
708 if (vn->sc_vp) {
709 struct uio auio;
710 struct iovec aiov;
711
efda3bd0 712 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
984263bc
MD
713 bzero(&auio, sizeof(auio));
714
715 aiov.iov_base = tmpbuf;
716 aiov.iov_len = vn->sc_secsize;
717 auio.uio_iov = &aiov;
718 auio.uio_iovcnt = 1;
719 auio.uio_offset = 0;
720 auio.uio_rw = UIO_READ;
721 auio.uio_segflg = UIO_SYSSPACE;
722 auio.uio_resid = aiov.iov_len;
ca466bae 723 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
984263bc 724 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
a11aaa81 725 vn_unlock(vn->sc_vp);
efda3bd0 726 kfree(tmpbuf, M_TEMP);
984263bc
MD
727 }
728 return (error);
729}
730
731void
732vnclear(struct vn_softc *vn)
733{
984263bc 734 IFOPT(vn, VN_FOLLOW)
e3869ec7 735 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
984263bc
MD
736 vn->sc_flags &= ~VNF_INITED;
737 if (vn->sc_vp != NULL) {
87de5057
MD
738 vn_close(vn->sc_vp,
739 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
984263bc
MD
740 vn->sc_vp = NULL;
741 }
742 vn->sc_flags &= ~VNF_READONLY;
743 if (vn->sc_cred) {
744 crfree(vn->sc_cred);
745 vn->sc_cred = NULL;
746 }
747 if (vn->sc_object != NULL) {
748 vm_pager_deallocate(vn->sc_object);
749 vn->sc_object = NULL;
750 }
8be7edad
MD
751
752 disk_unprobe(&vn->sc_disk);
753
984263bc
MD
754 vn->sc_size = 0;
755}
756
30c1fde0
CT
757/*
758 * vnget:
759 *
760 * populate a struct vn_user for the VNIOCGET ioctl.
761 * interface conventions defined in sys/sys/vnioctl.h.
762 */
763
764static int
765vnget(cdev_t dev, struct vn_softc *vn, struct vn_user *vnu)
766{
767 int error, found = 0;
768 char *freepath, *fullpath;
769 struct vattr vattr;
770
771 if (vnu->vnu_unit == -1) {
772 vnu->vnu_unit = dkunit(dev);
773 }
774 else if (vnu->vnu_unit < 0)
775 return (EINVAL);
776
777 SLIST_FOREACH(vn, &vn_list, sc_list) {
778
779 if(vn->sc_unit != vnu->vnu_unit)
780 continue;
781
782 found = 1;
783
784 if (vn->sc_flags & VNF_INITED && vn->sc_vp != NULL) {
785
786 /* note: u_cred checked in vnioctl above */
787 error = VOP_GETATTR(vn->sc_vp, &vattr);
788 if (error) {
789 kprintf("vnget: VOP_GETATTR for %p failed\n",
790 vn->sc_vp);
791 return (error);
792 }
793
794 error = vn_fullpath(curproc, vn->sc_vp,
795 &fullpath, &freepath);
796
797 if (error) {
798 kprintf("vnget: unable to resolve vp %p\n",
799 vn->sc_vp);
800 return(error);
801 }
802
803 strlcpy(vnu->vnu_file, fullpath,
804 sizeof(vnu->vnu_file));
805 kfree(freepath, M_TEMP);
806 vnu->vnu_dev = vattr.va_fsid;
807 vnu->vnu_ino = vattr.va_fileid;
808
809 }
810 else if (vn->sc_flags & VNF_INITED && vn->sc_object != NULL){
811
812 strlcpy(vnu->vnu_file, _VN_USER_SWAP,
813 sizeof(vnu->vnu_file));
814 vnu->vnu_size = vn->sc_size;
815 vnu->vnu_secsize = vn->sc_secsize;
816
817 } else {
818
819 bzero(vnu->vnu_file, sizeof(vnu->vnu_file));
820 vnu->vnu_dev = 0;
821 vnu->vnu_ino = 0;
822
823 }
824 break;
825 }
826
827 if (!found)
828 return(ENXIO);
829
830 return(0);
831}
832
fef8985e
MD
833static int
834vnsize(struct dev_psize_args *ap)
984263bc 835{
b13267a5 836 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
837 struct vn_softc *vn;
838
839 vn = dev->si_drv1;
840 if (!vn)
fef8985e 841 return(ENXIO);
984263bc 842 if ((vn->sc_flags & VNF_INITED) == 0)
fef8985e 843 return(ENXIO);
e0fc5693 844 ap->a_result = (int64_t)vn->sc_size;
fef8985e 845 return(0);
984263bc
MD
846}
847
cc80c90e
AH
848static cdev_t
849vn_create(int unit, struct devfs_bitmap *bitmap)
8be7edad 850{
cc80c90e
AH
851 struct vn_softc *vn;
852 struct disk_info info;
853 cdev_t dev;
8be7edad 854
cc80c90e
AH
855 vn = vncreatevn();
856 dev = disk_create(unit, &vn->sc_disk, &vn_ops);
857 vninitvn(vn, dev);
8be7edad 858
cc80c90e
AH
859 bzero(&info, sizeof(struct disk_info));
860 info.d_media_blksize = 512;
861 info.d_media_blocks = 0;
862 info.d_dsflags = DSO_MBRQUIET;
863 info.d_secpertrack = 32;
864 info.d_nheads = 64;
865 info.d_secpercyl = info.d_secpertrack * info.d_nheads;
866 info.d_ncylinders = 0;
867 disk_setdiskinfo_sync(&vn->sc_disk, &info);
868
869 if (bitmap != NULL)
870 devfs_clone_bitmap_set(bitmap, unit);
871
872 return dev;
8be7edad 873}
8be7edad 874
984263bc
MD
875static int
876vn_modevent(module_t mod, int type, void *data)
877{
878 struct vn_softc *vn;
cc80c90e 879 static cdev_t dev = NULL;
8be7edad 880 int i;
984263bc
MD
881
882 switch (type) {
883 case MOD_LOAD:
cc80c90e
AH
884 dev = make_autoclone_dev(&vn_ops, &DEVFS_CLONE_BITMAP(vn), vnclone, UID_ROOT,
885 GID_OPERATOR, 0640, "vn");
886
8be7edad 887 for (i = 0; i < VN_PREALLOCATED_UNITS; i++) {
cc80c90e 888 vn_create(i, &DEVFS_CLONE_BITMAP(vn));
8be7edad 889 }
984263bc 890 break;
984263bc 891 case MOD_UNLOAD:
cc80c90e 892 destroy_autoclone_dev(dev, &DEVFS_CLONE_BITMAP(vn));
984263bc
MD
893 /* fall through */
894 case MOD_SHUTDOWN:
e2f51e29 895 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
984263bc
MD
896 SLIST_REMOVE_HEAD(&vn_list, sc_list);
897 if (vn->sc_flags & VNF_INITED)
898 vnclear(vn);
b13267a5 899 /* Cleanup all cdev_t's that refer to this unit */
f31d07d7 900 disk_destroy(&vn->sc_disk);
984263bc
MD
901 while ((dev = vn->sc_devlist) != NULL) {
902 vn->sc_devlist = dev->si_drv2;
903 dev->si_drv1 = dev->si_drv2 = NULL;
e4c9c0c8 904 destroy_dev(dev);
984263bc 905 }
efda3bd0 906 kfree(vn, M_DEVBUF);
984263bc 907 }
cd29885a 908 dev_ops_remove_all(&vn_ops);
984263bc
MD
909 break;
910 default:
911 break;
912 }
913 return 0;
914}
915
916DEV_MODULE(vn, vn_modevent, 0);