Fix a minor bug that prevented compilation.
[dragonfly.git] / sys / vfs / specfs / spec_vnops.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)spec_vnops.c 8.14 (Berkeley) 5/21/95
34 * $FreeBSD: src/sys/miscfs/specfs/spec_vnops.c,v 1.131.2.4 2001/02/26 04:23:20 jlemon Exp $
31bd717a 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.46 2006/07/19 06:08:14 dillon Exp $
984263bc
MD
36 */
37
38#include <sys/param.h>
39#include <sys/proc.h>
40#include <sys/systm.h>
41#include <sys/kernel.h>
42#include <sys/conf.h>
43#include <sys/buf.h>
335dda38 44#include <sys/device.h>
984263bc
MD
45#include <sys/mount.h>
46#include <sys/vnode.h>
47#include <sys/stat.h>
48#include <sys/fcntl.h>
49#include <sys/vmmeter.h>
50#include <sys/tty.h>
51
52#include <vm/vm.h>
53#include <vm/vm_object.h>
54#include <vm/vm_page.h>
55#include <vm/vm_pager.h>
56
1c843a13
MD
57#include <machine/limits.h>
58
3020e3be
MD
59#include <sys/buf2.h>
60
165dba55
DR
61#include <sys/thread2.h>
62
a6ee311a
RG
63static int spec_advlock (struct vop_advlock_args *);
64static int spec_bmap (struct vop_bmap_args *);
65static int spec_close (struct vop_close_args *);
66static int spec_freeblks (struct vop_freeblks_args *);
67static int spec_fsync (struct vop_fsync_args *);
68static int spec_getpages (struct vop_getpages_args *);
69static int spec_inactive (struct vop_inactive_args *);
70static int spec_ioctl (struct vop_ioctl_args *);
71static int spec_open (struct vop_open_args *);
72static int spec_poll (struct vop_poll_args *);
73static int spec_kqfilter (struct vop_kqfilter_args *);
74static int spec_print (struct vop_print_args *);
75static int spec_read (struct vop_read_args *);
76static int spec_strategy (struct vop_strategy_args *);
77static int spec_write (struct vop_write_args *);
984263bc 78
66a1ddf5
MD
79struct vop_ops spec_vnode_vops = {
80 .vop_default = vop_defaultop,
81 .vop_access = (void *)vop_ebadf,
82 .vop_advlock = spec_advlock,
83 .vop_bmap = spec_bmap,
84 .vop_close = spec_close,
85 .vop_old_create = (void *)vop_panic,
86 .vop_freeblks = spec_freeblks,
87 .vop_fsync = spec_fsync,
88 .vop_getpages = spec_getpages,
89 .vop_inactive = spec_inactive,
90 .vop_ioctl = spec_ioctl,
91 .vop_old_link = (void *)vop_panic,
92 .vop_old_mkdir = (void *)vop_panic,
93 .vop_old_mknod = (void *)vop_panic,
94 .vop_open = spec_open,
95 .vop_pathconf = vop_stdpathconf,
96 .vop_poll = spec_poll,
97 .vop_kqfilter = spec_kqfilter,
98 .vop_print = spec_print,
99 .vop_read = spec_read,
100 .vop_readdir = (void *)vop_panic,
101 .vop_readlink = (void *)vop_panic,
102 .vop_reallocblks = (void *)vop_panic,
103 .vop_reclaim = (void *)vop_null,
104 .vop_old_remove = (void *)vop_panic,
105 .vop_old_rename = (void *)vop_panic,
106 .vop_old_rmdir = (void *)vop_panic,
107 .vop_setattr = (void *)vop_ebadf,
108 .vop_strategy = spec_strategy,
109 .vop_old_symlink = (void *)vop_panic,
110 .vop_write = spec_write
984263bc 111};
66a1ddf5
MD
112
113struct vop_ops *spec_vnode_vops_p = &spec_vnode_vops;
114
115VNODEOP_SET(spec_vnode_vops);
984263bc 116
e4c9c0c8
MD
117extern int dev_ref_debug;
118
50775720 119/*
31bd717a 120 * spec_vnoperate()
50775720 121 */
984263bc 122int
50775720 123spec_vnoperate(struct vop_generic_args *ap)
984263bc 124{
66a1ddf5 125 return (VOCALL(&spec_vnode_vops, ap));
984263bc
MD
126}
127
81b5c339 128static void spec_getpages_iodone (struct bio *bio);
984263bc
MD
129
130/*
131 * Open a special file.
50775720
CP
132 *
133 * spec_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
8ddc6004 134 * struct file *a_fp, struct thread *a_td)
984263bc
MD
135 */
136/* ARGSUSED */
137static int
50775720 138spec_open(struct vop_open_args *ap)
984263bc 139{
984263bc 140 struct vnode *vp = ap->a_vp;
e4c9c0c8 141 dev_t dev;
984263bc 142 int error;
e4c9c0c8 143 int isblk = (vp->v_type == VBLK) ? 1 : 0;
984263bc
MD
144 const char *cp;
145
146 /*
147 * Don't allow open if fs is mounted -nodev.
148 */
149 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
150 return (ENXIO);
151
e4c9c0c8
MD
152 /*
153 * Resolve the device. If the vnode is already open v_rdev may
154 * already be resolved. However, if the device changes out from
1fbb5fc0
MD
155 * under us we report it (and, for now, we allow it). Since
156 * v_release_rdev() zero's v_opencount, we have to save and restore
157 * it when replacing the rdev reference.
e4c9c0c8
MD
158 */
159 if (vp->v_rdev != NULL) {
160 dev = udev2dev(vp->v_udev, isblk);
161 if (dev != vp->v_rdev) {
1fbb5fc0 162 int oc = vp->v_opencount;
e4c9c0c8
MD
163 printf(
164 "Warning: spec_open: dev %s was lost",
165 vp->v_rdev->si_name);
166 v_release_rdev(vp);
167 error = v_associate_rdev(vp,
168 udev2dev(vp->v_udev, isblk));
1fbb5fc0 169 if (error) {
e4c9c0c8 170 printf(", reacquisition failed\n");
1fbb5fc0
MD
171 } else {
172 vp->v_opencount = oc;
e4c9c0c8 173 printf(", reacquisition successful\n");
1fbb5fc0 174 }
e4c9c0c8
MD
175 } else {
176 error = 0;
177 }
178 } else {
179 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
180 }
181 if (error)
182 return(error);
183
500df1da
MD
184 /*
185 * Prevent degenerate open/close sequences from nulling out rdev.
186 */
e4c9c0c8 187 dev = vp->v_rdev;
500df1da 188 KKASSERT(dev != NULL);
984263bc 189
e4c9c0c8
MD
190 /*
191 * Make this field valid before any I/O in ->d_open. XXX the
192 * device itself should probably be required to initialize
193 * this field in d_open.
194 */
984263bc
MD
195 if (!dev->si_iosize_max)
196 dev->si_iosize_max = DFLTPHYS;
197
198 /*
199 * XXX: Disks get special billing here, but it is mostly wrong.
200 * XXX: diskpartitions can overlap and the real checks should
201 * XXX: take this into account, and consequently they need to
202 * XXX: live in the diskslicing code. Some checks do.
203 */
204 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
205 (ap->a_mode & FWRITE)) {
206 /*
207 * Never allow opens for write if the device is mounted R/W
208 */
e4c9c0c8
MD
209 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
210 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
211 error = EBUSY;
212 goto done;
213 }
984263bc
MD
214
215 /*
216 * When running in secure mode, do not allow opens
217 * for writing if the device is mounted
218 */
e4c9c0c8
MD
219 if (securelevel >= 1 && vfs_mountedon(vp)) {
220 error = EPERM;
221 goto done;
222 }
984263bc
MD
223
224 /*
225 * When running in very secure mode, do not allow
226 * opens for writing of any devices.
227 */
e4c9c0c8
MD
228 if (securelevel >= 2) {
229 error = EPERM;
230 goto done;
231 }
984263bc
MD
232 }
233
234 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 235 if (dev_dflags(dev) & D_TTY)
984263bc
MD
236 vp->v_flag |= VISTTY;
237
e4c9c0c8
MD
238 /*
239 * dev_dopen() is always called for each open. dev_dclose() is
240 * only called for the last close unless D_TRACKCLOSE is set.
241 */
ca466bae 242 VOP_UNLOCK(vp, 0);
87de5057 243 error = dev_dopen(dev, ap->a_mode, S_IFCHR, curthread);
ca466bae 244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
245
246 if (error)
e4c9c0c8 247 goto done;
984263bc 248
335dda38 249 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
250 if (dev->si_tty) {
251 struct tty *tp;
252 tp = dev->si_tty;
253 if (!tp->t_stop) {
254 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
255 tp->t_stop = nottystop;
256 }
257 }
258 }
259
7540ab49
MD
260 /*
261 * If this is 'disk' or disk-like device, associate a VM object
262 * with it.
263 */
984263bc
MD
264 if (vn_isdisk(vp, NULL)) {
265 if (!dev->si_bsize_phys)
266 dev->si_bsize_phys = DEV_BSIZE;
1c843a13 267 vinitvmio(vp, IDX_TO_OFF(INT_MAX));
984263bc 268 }
335dda38 269 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 270 cp = devtoname(dev);
335dda38 271 if (*cp == '#') {
984263bc 272 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 273 dev_dname(dev), cp);
984263bc
MD
274 }
275 }
fad57d0e
MD
276
277 /*
278 * If we were handed a file pointer we may be able to install a
279 * shortcut which issues device read and write operations directly
280 * from the fileops rather then having to go through spec_read()
281 * and spec_write().
282 */
283 if (ap->a_fp)
284 vn_setspecops(ap->a_fp);
285
e4c9c0c8
MD
286 if (dev_ref_debug)
287 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
288done:
289 if (error) {
8ddc6004 290 if (vp->v_opencount == 0)
e4c9c0c8 291 v_release_rdev(vp);
8ddc6004
MD
292 } else {
293 vop_stdopen(ap);
e4c9c0c8 294 }
984263bc
MD
295 return (error);
296}
297
298/*
299 * Vnode op for read
50775720
CP
300 *
301 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
302 * struct ucred *a_cred)
984263bc
MD
303 */
304/* ARGSUSED */
305static int
50775720 306spec_read(struct vop_read_args *ap)
984263bc
MD
307{
308 struct vnode *vp;
dadab5e9 309 struct thread *td;
984263bc
MD
310 struct uio *uio;
311 dev_t dev;
312 int error;
313
314 vp = ap->a_vp;
315 dev = vp->v_rdev;
316 uio = ap->a_uio;
dadab5e9 317 td = uio->uio_td;
984263bc 318
72402f22
MD
319 if (dev == NULL) /* device was revoked */
320 return (EBADF);
984263bc
MD
321 if (uio->uio_resid == 0)
322 return (0);
323
ca466bae 324 VOP_UNLOCK(vp, 0);
335dda38 325 error = dev_dread(dev, uio, ap->a_ioflag);
ca466bae 326 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
327 return (error);
328}
329
330/*
331 * Vnode op for write
50775720
CP
332 *
333 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
334 * struct ucred *a_cred)
984263bc
MD
335 */
336/* ARGSUSED */
337static int
50775720 338spec_write(struct vop_write_args *ap)
984263bc
MD
339{
340 struct vnode *vp;
dadab5e9 341 struct thread *td;
984263bc
MD
342 struct uio *uio;
343 dev_t dev;
344 int error;
345
346 vp = ap->a_vp;
347 dev = vp->v_rdev;
348 uio = ap->a_uio;
dadab5e9 349 td = uio->uio_td;
984263bc 350
72402f22
MD
351 if (dev == NULL) /* device was revoked */
352 return (EBADF);
353
ca466bae 354 VOP_UNLOCK(vp, 0);
335dda38 355 error = dev_dwrite(dev, uio, ap->a_ioflag);
ca466bae 356 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
357 return (error);
358}
359
360/*
361 * Device ioctl operation.
50775720
CP
362 *
363 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
364 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
365 */
366/* ARGSUSED */
367static int
50775720 368spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
369{
370 dev_t dev;
371
72402f22
MD
372 if ((dev = ap->a_vp->v_rdev) == NULL)
373 return (EBADF); /* device was revoked */
374
335dda38 375 return (dev_dioctl(dev, ap->a_command, ap->a_data,
87de5057 376 ap->a_fflag,curthread));
984263bc
MD
377}
378
50775720
CP
379/*
380 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
381 * struct thread *a_td)
382 */
984263bc
MD
383/* ARGSUSED */
384static int
50775720 385spec_poll(struct vop_poll_args *ap)
984263bc
MD
386{
387 dev_t dev;
388
72402f22
MD
389 if ((dev = ap->a_vp->v_rdev) == NULL)
390 return (EBADF); /* device was revoked */
87de5057 391 return (dev_dpoll(dev, ap->a_events, curthread));
984263bc
MD
392}
393
50775720
CP
394/*
395 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
396 */
984263bc
MD
397/* ARGSUSED */
398static int
50775720 399spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
400{
401 dev_t dev;
402
72402f22
MD
403 if ((dev = ap->a_vp->v_rdev) == NULL)
404 return (EBADF); /* device was revoked */
335dda38 405 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
406}
407
408/*
409 * Synch buffers associated with a block device
50775720
CP
410 *
411 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
412 * int a_waitfor, struct thread *a_td)
984263bc
MD
413 */
414/* ARGSUSED */
415static int
50775720 416spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
417{
418 struct vnode *vp = ap->a_vp;
6bae6177 419 int error;
984263bc
MD
420
421 if (!vn_isdisk(vp, NULL))
422 return (0);
423
984263bc
MD
424 /*
425 * Flush all dirty buffers associated with a block device.
426 */
4e0ecc94 427 error = vfsync(vp, ap->a_waitfor, 10000, NULL, NULL);
6bae6177 428 return (error);
984263bc
MD
429}
430
50775720
CP
431/*
432 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
433 */
984263bc 434static int
50775720 435spec_inactive(struct vop_inactive_args *ap)
984263bc 436{
984263bc
MD
437 return (0);
438}
439
440/*
441 * Just call the device strategy routine
50775720 442 *
81b5c339 443 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
984263bc
MD
444 */
445static int
50775720 446spec_strategy(struct vop_strategy_args *ap)
984263bc 447{
81b5c339
MD
448 struct bio *bio = ap->a_bio;
449 struct buf *bp = bio->bio_buf;
984263bc
MD
450 struct vnode *vp;
451 struct mount *mp;
452
10f3fee5
MD
453 if (bp->b_cmd != BUF_CMD_READ &&
454 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start) {
984263bc 455 (*bioops.io_start)(bp);
10f3fee5 456 }
984263bc
MD
457
458 /*
459 * Collect statistics on synchronous and asynchronous read
460 * and write counts for disks that have associated filesystems.
461 */
462 vp = ap->a_vp;
72402f22 463 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 464 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
10f3fee5 465 if (bp->b_cmd == BUF_CMD_READ) {
df4f70a6 466 if (bp->b_flags & B_ASYNC)
10f3fee5 467 mp->mnt_stat.f_asyncreads++;
984263bc 468 else
10f3fee5 469 mp->mnt_stat.f_syncreads++;
984263bc 470 } else {
df4f70a6 471 if (bp->b_flags & B_ASYNC)
10f3fee5 472 mp->mnt_stat.f_asyncwrites++;
984263bc 473 else
10f3fee5 474 mp->mnt_stat.f_syncwrites++;
984263bc
MD
475 }
476 }
81b5c339 477 dev_dstrategy_chain(vp->v_rdev, bio);
984263bc
MD
478 return (0);
479}
480
50775720
CP
481/*
482 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
483 */
984263bc 484static int
50775720 485spec_freeblks(struct vop_freeblks_args *ap)
984263bc 486{
984263bc
MD
487 struct buf *bp;
488
489 /*
490 * XXX: This assumes that strategy does the deed right away.
491 * XXX: this may not be TRTTD.
492 */
72402f22 493 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 494 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
495 return (0);
496 bp = geteblk(ap->a_length);
10f3fee5 497 bp->b_cmd = BUF_CMD_FREEBLKS;
54078292 498 bp->b_bio1.bio_offset = ap->a_offset;
984263bc 499 bp->b_bcount = ap->a_length;
81b5c339 500 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
984263bc
MD
501 return (0);
502}
503
504/*
505 * Implement degenerate case where the block requested is the block
506 * returned, and assume that the entire device is contiguous in regards
507 * to the contiguous block range (runp and runb).
50775720 508 *
54078292
MD
509 * spec_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
510 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
511 */
512static int
50775720 513spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
514{
515 struct vnode *vp = ap->a_vp;
984263bc
MD
516
517 if (ap->a_vpp != NULL)
518 *ap->a_vpp = vp;
54078292
MD
519 if (ap->a_doffsetp != NULL)
520 *ap->a_doffsetp = ap->a_loffset;
984263bc 521 if (ap->a_runp != NULL)
54078292
MD
522 *ap->a_runp = MAXBSIZE;
523 if (ap->a_runb != NULL) {
524 if (ap->a_loffset < MAXBSIZE)
525 *ap->a_runb = (int)ap->a_loffset;
526 else
527 *ap->a_runb = MAXBSIZE;
528 }
984263bc
MD
529 return (0);
530}
531
532/*
533 * Device close routine
50775720
CP
534 *
535 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
536 * struct thread *a_td)
53994a38
MD
537 *
538 * NOTE: the vnode may or may not be locked on call.
984263bc
MD
539 */
540/* ARGSUSED */
541static int
50775720 542spec_close(struct vop_close_args *ap)
984263bc 543{
87de5057 544 struct proc *p = curproc;
984263bc 545 struct vnode *vp = ap->a_vp;
984263bc 546 dev_t dev = vp->v_rdev;
e4c9c0c8 547 int error;
53994a38 548 int needrelock;
984263bc
MD
549
550 /*
551 * Hack: a tty device that is a controlling terminal
552 * has a reference from the session structure.
553 * We cannot easily tell that a character device is
554 * a controlling terminal, unless it is the closing
555 * process' controlling terminal. In that case,
556 * if the reference count is 2 (this last descriptor
557 * plus the session), release the reference from the session.
1fbb5fc0
MD
558 *
559 * It is possible for v_opencount to be 0 or 1 in this case, 0
560 * because the tty might have been revoked.
984263bc 561 */
72402f22
MD
562 if (dev)
563 reference_dev(dev);
1fbb5fc0 564 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 565 p && vp == p->p_session->s_ttyvp) {
984263bc 566 p->p_session->s_ttyvp = NULL;
1fbb5fc0 567 vrele(vp);
984263bc 568 }
e4c9c0c8 569
984263bc 570 /*
e4c9c0c8
MD
571 * Vnodes can be opened and close multiple times. Do not really
572 * close the device unless (1) it is being closed forcibly,
573 * (2) the device wants to track closes, or (3) this is the last
574 * vnode doing its last close on the device.
575 *
576 * XXX the VXLOCK (force close) case can leave vnodes referencing
577 * a closed device.
984263bc 578 */
5fd012e0 579 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 580 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 581 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
53994a38 582 needrelock = 0;
87de5057 583 if (VOP_ISLOCKED(vp, curthread)) {
53994a38 584 needrelock = 1;
ca466bae 585 VOP_UNLOCK(vp, 0);
53994a38 586 }
87de5057 587 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, curthread);
53994a38 588 if (needrelock)
ca466bae 589 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
e4c9c0c8
MD
590 } else {
591 error = 0;
984263bc 592 }
e4c9c0c8
MD
593
594 /*
595 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
596 * disassociates the rdev. If the rdev is already disassociated
597 * the vnode might have been revoked and no further opencount
598 * tracking occurs.
e4c9c0c8 599 */
1fbb5fc0 600 if (dev) {
8ddc6004 601 /*KKASSERT(vp->v_opencount > 0);*/
1fbb5fc0 602 if (dev_ref_debug) {
72402f22
MD
603 printf("spec_close: %s %d\n",
604 dev->si_name, vp->v_opencount - 1);
72402f22 605 }
8ddc6004 606 if (vp->v_opencount == 1)
1fbb5fc0 607 v_release_rdev(vp);
72402f22 608 release_dev(dev);
1fbb5fc0 609 }
8ddc6004 610 vop_stdclose(ap);
e4c9c0c8 611 return(error);
984263bc
MD
612}
613
614/*
615 * Print out the contents of a special device vnode.
50775720
CP
616 *
617 * spec_print(struct vnode *a_vp)
984263bc
MD
618 */
619static int
50775720 620spec_print(struct vop_print_args *ap)
984263bc 621{
984263bc
MD
622 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
623 return (0);
624}
625
626/*
627 * Special device advisory byte-level locks.
50775720
CP
628 *
629 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
630 * struct flock *a_fl, int a_flags)
984263bc
MD
631 */
632/* ARGSUSED */
633static int
50775720 634spec_advlock(struct vop_advlock_args *ap)
984263bc 635{
71c18fe3 636 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
984263bc
MD
637}
638
639static void
81b5c339 640spec_getpages_iodone(struct bio *bio)
984263bc 641{
10f3fee5 642 bio->bio_buf->b_cmd = BUF_CMD_DONE;
81b5c339 643 wakeup(bio->bio_buf);
984263bc
MD
644}
645
646static int
50775720 647spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
648{
649 vm_offset_t kva;
650 int error;
165dba55 651 int i, pcount, size;
984263bc
MD
652 struct buf *bp;
653 vm_page_t m;
654 vm_ooffset_t offset;
655 int toff, nextoff, nread;
656 struct vnode *vp = ap->a_vp;
657 int blksiz;
658 int gotreqpage;
659
660 error = 0;
661 pcount = round_page(ap->a_count) / PAGE_SIZE;
662
663 /*
664 * Calculate the offset of the transfer and do sanity check.
984263bc
MD
665 */
666 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
667
984263bc
MD
668 /*
669 * Round up physical size for real devices. We cannot round using
670 * v_mount's block size data because v_mount has nothing to do with
671 * the device. i.e. it's usually '/dev'. We need the physical block
672 * size for the device itself.
673 *
e4c9c0c8 674 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
675 * block device is mounted. However, we can use v_rdev.
676 */
677
678 if (vn_isdisk(vp, NULL))
679 blksiz = vp->v_rdev->si_bsize_phys;
680 else
681 blksiz = DEV_BSIZE;
682
683 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
684
685 bp = getpbuf(NULL);
686 kva = (vm_offset_t)bp->b_data;
687
688 /*
689 * Map the pages to be read into the kva.
690 */
691 pmap_qenter(kva, ap->a_m, pcount);
692
693 /* Build a minimal buffer header. */
10f3fee5 694 bp->b_cmd = BUF_CMD_READ;
984263bc 695 bp->b_bcount = size;
984263bc 696 bp->b_resid = 0;
9a71d53f 697 bp->b_runningbufspace = size;
984263bc
MD
698 runningbufspace += bp->b_runningbufspace;
699
54078292 700 bp->b_bio1.bio_offset = offset;
81b5c339
MD
701 bp->b_bio1.bio_done = spec_getpages_iodone;
702
12e4aaff
MD
703 mycpu->gd_cnt.v_vnodein++;
704 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
705
706 /* Do the input. */
81b5c339 707 vn_strategy(ap->a_vp, &bp->b_bio1);
984263bc 708
165dba55 709 crit_enter();
984263bc
MD
710
711 /* We definitely need to be at splbio here. */
10f3fee5 712 while (bp->b_cmd != BUF_CMD_DONE)
377d4740 713 tsleep(bp, 0, "spread", 0);
984263bc 714
165dba55 715 crit_exit();
984263bc 716
9a71d53f 717 if (bp->b_flags & B_ERROR) {
984263bc
MD
718 if (bp->b_error)
719 error = bp->b_error;
720 else
721 error = EIO;
722 }
723
9a71d53f
MD
724 /*
725 * If EOF is encountered we must zero-extend the result in order
726 * to ensure that the page does not contain garabge. When no
727 * error occurs, an early EOF is indicated if b_bcount got truncated.
728 * b_resid is relative to b_bcount and should be 0, but some devices
729 * might indicate an EOF with b_resid instead of truncating b_bcount.
730 */
731 nread = bp->b_bcount - bp->b_resid;
732 if (nread < ap->a_count)
733 bzero((caddr_t)kva + nread, ap->a_count - nread);
984263bc
MD
734 pmap_qremove(kva, pcount);
735
984263bc
MD
736 gotreqpage = 0;
737 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
738 nextoff = toff + PAGE_SIZE;
739 m = ap->a_m[i];
740
741 m->flags &= ~PG_ZERO;
742
743 if (nextoff <= nread) {
744 m->valid = VM_PAGE_BITS_ALL;
745 vm_page_undirty(m);
746 } else if (toff < nread) {
747 /*
748 * Since this is a VM request, we have to supply the
749 * unaligned offset to allow vm_page_set_validclean()
750 * to zero sub-DEV_BSIZE'd portions of the page.
751 */
752 vm_page_set_validclean(m, 0, nread - toff);
753 } else {
754 m->valid = 0;
755 vm_page_undirty(m);
756 }
757
758 if (i != ap->a_reqpage) {
759 /*
760 * Just in case someone was asking for this page we
761 * now tell them that it is ok to use.
762 */
763 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
764 if (m->valid) {
765 if (m->flags & PG_WANTED) {
766 vm_page_activate(m);
767 } else {
768 vm_page_deactivate(m);
769 }
770 vm_page_wakeup(m);
771 } else {
772 vm_page_free(m);
773 }
774 } else {
775 vm_page_free(m);
776 }
777 } else if (m->valid) {
778 gotreqpage = 1;
779 /*
780 * Since this is a VM request, we need to make the
781 * entire page presentable by zeroing invalid sections.
782 */
783 if (m->valid != VM_PAGE_BITS_ALL)
784 vm_page_zero_invalid(m, FALSE);
785 }
786 }
787 if (!gotreqpage) {
788 m = ap->a_m[ap->a_reqpage];
789 printf(
790 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
81b5c339 791 devtoname(vp->v_rdev), error, bp, bp->b_vp);
984263bc 792 printf(
54078292 793 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
984263bc
MD
794 size, bp->b_resid, ap->a_count, m->valid);
795 printf(
796 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
797 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
798 /*
799 * Free the buffer header back to the swap buffer pool.
800 */
801 relpbuf(bp, NULL);
802 return VM_PAGER_ERROR;
803 }
804 /*
805 * Free the buffer header back to the swap buffer pool.
806 */
807 relpbuf(bp, NULL);
808 return VM_PAGER_OK;
809}