Change the kernel dev_t, representing a pointer to a specinfo structure,
[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 $
b13267a5 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.49 2006/09/10 01:26:41 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,
fef8985e 134 * struct file *a_fp)
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;
b13267a5 141 cdev_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 */
a11aaa81 242 vn_unlock(vp);
fef8985e 243 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_cred);
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 == '#') {
b13267a5 272 printf("WARNING: driver %s should register devices with make_dev() (cdev_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 310 struct uio *uio;
b13267a5 311 cdev_t dev;
984263bc
MD
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
a11aaa81 324 vn_unlock(vp);
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 342 struct uio *uio;
b13267a5 343 cdev_t dev;
984263bc
MD
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
a11aaa81 354 vn_unlock(vp);
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,
fef8985e 364 * int a_fflag, struct ucred *a_cred)
984263bc
MD
365 */
366/* ARGSUSED */
367static int
50775720 368spec_ioctl(struct vop_ioctl_args *ap)
984263bc 369{
b13267a5 370 cdev_t dev;
984263bc 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,
fef8985e 376 ap->a_fflag, ap->a_cred));
984263bc
MD
377}
378
50775720 379/*
fef8985e 380 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred)
50775720 381 */
984263bc
MD
382/* ARGSUSED */
383static int
50775720 384spec_poll(struct vop_poll_args *ap)
984263bc 385{
b13267a5 386 cdev_t dev;
984263bc 387
72402f22
MD
388 if ((dev = ap->a_vp->v_rdev) == NULL)
389 return (EBADF); /* device was revoked */
fef8985e 390 return (dev_dpoll(dev, ap->a_events));
984263bc
MD
391}
392
50775720
CP
393/*
394 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
395 */
984263bc
MD
396/* ARGSUSED */
397static int
50775720 398spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc 399{
b13267a5 400 cdev_t dev;
984263bc 401
72402f22
MD
402 if ((dev = ap->a_vp->v_rdev) == NULL)
403 return (EBADF); /* device was revoked */
335dda38 404 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
405}
406
407/*
408 * Synch buffers associated with a block device
50775720
CP
409 *
410 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
411 * int a_waitfor, struct thread *a_td)
984263bc
MD
412 */
413/* ARGSUSED */
414static int
50775720 415spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
416{
417 struct vnode *vp = ap->a_vp;
6bae6177 418 int error;
984263bc
MD
419
420 if (!vn_isdisk(vp, NULL))
421 return (0);
422
984263bc
MD
423 /*
424 * Flush all dirty buffers associated with a block device.
425 */
4e0ecc94 426 error = vfsync(vp, ap->a_waitfor, 10000, NULL, NULL);
6bae6177 427 return (error);
984263bc
MD
428}
429
50775720
CP
430/*
431 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
432 */
984263bc 433static int
50775720 434spec_inactive(struct vop_inactive_args *ap)
984263bc 435{
984263bc
MD
436 return (0);
437}
438
439/*
440 * Just call the device strategy routine
50775720 441 *
81b5c339 442 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
984263bc
MD
443 */
444static int
50775720 445spec_strategy(struct vop_strategy_args *ap)
984263bc 446{
81b5c339
MD
447 struct bio *bio = ap->a_bio;
448 struct buf *bp = bio->bio_buf;
984263bc
MD
449 struct vnode *vp;
450 struct mount *mp;
451
10f3fee5
MD
452 if (bp->b_cmd != BUF_CMD_READ &&
453 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start) {
984263bc 454 (*bioops.io_start)(bp);
10f3fee5 455 }
984263bc
MD
456
457 /*
458 * Collect statistics on synchronous and asynchronous read
459 * and write counts for disks that have associated filesystems.
460 */
461 vp = ap->a_vp;
72402f22 462 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 463 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
10f3fee5 464 if (bp->b_cmd == BUF_CMD_READ) {
df4f70a6 465 if (bp->b_flags & B_ASYNC)
10f3fee5 466 mp->mnt_stat.f_asyncreads++;
984263bc 467 else
10f3fee5 468 mp->mnt_stat.f_syncreads++;
984263bc 469 } else {
df4f70a6 470 if (bp->b_flags & B_ASYNC)
10f3fee5 471 mp->mnt_stat.f_asyncwrites++;
984263bc 472 else
10f3fee5 473 mp->mnt_stat.f_syncwrites++;
984263bc
MD
474 }
475 }
81b5c339 476 dev_dstrategy_chain(vp->v_rdev, bio);
984263bc
MD
477 return (0);
478}
479
50775720
CP
480/*
481 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
482 */
984263bc 483static int
50775720 484spec_freeblks(struct vop_freeblks_args *ap)
984263bc 485{
984263bc
MD
486 struct buf *bp;
487
488 /*
489 * XXX: This assumes that strategy does the deed right away.
490 * XXX: this may not be TRTTD.
491 */
72402f22 492 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 493 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
494 return (0);
495 bp = geteblk(ap->a_length);
10f3fee5 496 bp->b_cmd = BUF_CMD_FREEBLKS;
54078292 497 bp->b_bio1.bio_offset = ap->a_offset;
984263bc 498 bp->b_bcount = ap->a_length;
81b5c339 499 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
984263bc
MD
500 return (0);
501}
502
503/*
504 * Implement degenerate case where the block requested is the block
505 * returned, and assume that the entire device is contiguous in regards
506 * to the contiguous block range (runp and runb).
50775720 507 *
54078292
MD
508 * spec_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
509 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
510 */
511static int
50775720 512spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
513{
514 struct vnode *vp = ap->a_vp;
984263bc
MD
515
516 if (ap->a_vpp != NULL)
517 *ap->a_vpp = vp;
54078292
MD
518 if (ap->a_doffsetp != NULL)
519 *ap->a_doffsetp = ap->a_loffset;
984263bc 520 if (ap->a_runp != NULL)
54078292
MD
521 *ap->a_runp = MAXBSIZE;
522 if (ap->a_runb != NULL) {
523 if (ap->a_loffset < MAXBSIZE)
524 *ap->a_runb = (int)ap->a_loffset;
525 else
526 *ap->a_runb = MAXBSIZE;
527 }
984263bc
MD
528 return (0);
529}
530
531/*
532 * Device close routine
50775720
CP
533 *
534 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
535 * struct thread *a_td)
53994a38
MD
536 *
537 * NOTE: the vnode may or may not be locked on call.
984263bc
MD
538 */
539/* ARGSUSED */
540static int
50775720 541spec_close(struct vop_close_args *ap)
984263bc 542{
87de5057 543 struct proc *p = curproc;
984263bc 544 struct vnode *vp = ap->a_vp;
b13267a5 545 cdev_t dev = vp->v_rdev;
e4c9c0c8 546 int error;
53994a38 547 int needrelock;
984263bc
MD
548
549 /*
550 * Hack: a tty device that is a controlling terminal
551 * has a reference from the session structure.
552 * We cannot easily tell that a character device is
553 * a controlling terminal, unless it is the closing
554 * process' controlling terminal. In that case,
555 * if the reference count is 2 (this last descriptor
556 * plus the session), release the reference from the session.
1fbb5fc0
MD
557 *
558 * It is possible for v_opencount to be 0 or 1 in this case, 0
559 * because the tty might have been revoked.
984263bc 560 */
72402f22
MD
561 if (dev)
562 reference_dev(dev);
1fbb5fc0 563 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 564 p && vp == p->p_session->s_ttyvp) {
984263bc 565 p->p_session->s_ttyvp = NULL;
1fbb5fc0 566 vrele(vp);
984263bc 567 }
e4c9c0c8 568
984263bc 569 /*
e4c9c0c8
MD
570 * Vnodes can be opened and close multiple times. Do not really
571 * close the device unless (1) it is being closed forcibly,
572 * (2) the device wants to track closes, or (3) this is the last
573 * vnode doing its last close on the device.
574 *
575 * XXX the VXLOCK (force close) case can leave vnodes referencing
576 * a closed device.
984263bc 577 */
5fd012e0 578 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 579 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 580 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
53994a38 581 needrelock = 0;
a11aaa81 582 if (vn_islocked(vp)) {
53994a38 583 needrelock = 1;
a11aaa81 584 vn_unlock(vp);
53994a38 585 }
fef8985e 586 error = dev_dclose(dev, ap->a_fflag, S_IFCHR);
53994a38 587 if (needrelock)
ca466bae 588 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
e4c9c0c8
MD
589 } else {
590 error = 0;
984263bc 591 }
e4c9c0c8
MD
592
593 /*
594 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
595 * disassociates the rdev. If the rdev is already disassociated
596 * the vnode might have been revoked and no further opencount
597 * tracking occurs.
e4c9c0c8 598 */
1fbb5fc0 599 if (dev) {
8ddc6004 600 /*KKASSERT(vp->v_opencount > 0);*/
1fbb5fc0 601 if (dev_ref_debug) {
72402f22
MD
602 printf("spec_close: %s %d\n",
603 dev->si_name, vp->v_opencount - 1);
72402f22 604 }
8ddc6004 605 if (vp->v_opencount == 1)
1fbb5fc0 606 v_release_rdev(vp);
72402f22 607 release_dev(dev);
1fbb5fc0 608 }
8ddc6004 609 vop_stdclose(ap);
e4c9c0c8 610 return(error);
984263bc
MD
611}
612
613/*
614 * Print out the contents of a special device vnode.
50775720
CP
615 *
616 * spec_print(struct vnode *a_vp)
984263bc
MD
617 */
618static int
50775720 619spec_print(struct vop_print_args *ap)
984263bc 620{
984263bc
MD
621 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
622 return (0);
623}
624
625/*
626 * Special device advisory byte-level locks.
50775720
CP
627 *
628 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
629 * struct flock *a_fl, int a_flags)
984263bc
MD
630 */
631/* ARGSUSED */
632static int
50775720 633spec_advlock(struct vop_advlock_args *ap)
984263bc 634{
71c18fe3 635 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
984263bc
MD
636}
637
638static void
81b5c339 639spec_getpages_iodone(struct bio *bio)
984263bc 640{
10f3fee5 641 bio->bio_buf->b_cmd = BUF_CMD_DONE;
81b5c339 642 wakeup(bio->bio_buf);
984263bc
MD
643}
644
645static int
50775720 646spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
647{
648 vm_offset_t kva;
649 int error;
165dba55 650 int i, pcount, size;
984263bc
MD
651 struct buf *bp;
652 vm_page_t m;
653 vm_ooffset_t offset;
654 int toff, nextoff, nread;
655 struct vnode *vp = ap->a_vp;
656 int blksiz;
657 int gotreqpage;
658
659 error = 0;
660 pcount = round_page(ap->a_count) / PAGE_SIZE;
661
662 /*
663 * Calculate the offset of the transfer and do sanity check.
984263bc
MD
664 */
665 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
666
984263bc
MD
667 /*
668 * Round up physical size for real devices. We cannot round using
669 * v_mount's block size data because v_mount has nothing to do with
670 * the device. i.e. it's usually '/dev'. We need the physical block
671 * size for the device itself.
672 *
e4c9c0c8 673 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
674 * block device is mounted. However, we can use v_rdev.
675 */
676
677 if (vn_isdisk(vp, NULL))
678 blksiz = vp->v_rdev->si_bsize_phys;
679 else
680 blksiz = DEV_BSIZE;
681
682 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
683
684 bp = getpbuf(NULL);
685 kva = (vm_offset_t)bp->b_data;
686
687 /*
688 * Map the pages to be read into the kva.
689 */
690 pmap_qenter(kva, ap->a_m, pcount);
691
692 /* Build a minimal buffer header. */
10f3fee5 693 bp->b_cmd = BUF_CMD_READ;
984263bc 694 bp->b_bcount = size;
984263bc 695 bp->b_resid = 0;
9a71d53f 696 bp->b_runningbufspace = size;
984263bc
MD
697 runningbufspace += bp->b_runningbufspace;
698
54078292 699 bp->b_bio1.bio_offset = offset;
81b5c339
MD
700 bp->b_bio1.bio_done = spec_getpages_iodone;
701
12e4aaff
MD
702 mycpu->gd_cnt.v_vnodein++;
703 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
704
705 /* Do the input. */
81b5c339 706 vn_strategy(ap->a_vp, &bp->b_bio1);
984263bc 707
165dba55 708 crit_enter();
984263bc
MD
709
710 /* We definitely need to be at splbio here. */
10f3fee5 711 while (bp->b_cmd != BUF_CMD_DONE)
377d4740 712 tsleep(bp, 0, "spread", 0);
984263bc 713
165dba55 714 crit_exit();
984263bc 715
9a71d53f 716 if (bp->b_flags & B_ERROR) {
984263bc
MD
717 if (bp->b_error)
718 error = bp->b_error;
719 else
720 error = EIO;
721 }
722
9a71d53f
MD
723 /*
724 * If EOF is encountered we must zero-extend the result in order
725 * to ensure that the page does not contain garabge. When no
726 * error occurs, an early EOF is indicated if b_bcount got truncated.
727 * b_resid is relative to b_bcount and should be 0, but some devices
728 * might indicate an EOF with b_resid instead of truncating b_bcount.
729 */
730 nread = bp->b_bcount - bp->b_resid;
731 if (nread < ap->a_count)
732 bzero((caddr_t)kva + nread, ap->a_count - nread);
984263bc
MD
733 pmap_qremove(kva, pcount);
734
984263bc
MD
735 gotreqpage = 0;
736 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
737 nextoff = toff + PAGE_SIZE;
738 m = ap->a_m[i];
739
740 m->flags &= ~PG_ZERO;
741
742 if (nextoff <= nread) {
743 m->valid = VM_PAGE_BITS_ALL;
744 vm_page_undirty(m);
745 } else if (toff < nread) {
746 /*
747 * Since this is a VM request, we have to supply the
748 * unaligned offset to allow vm_page_set_validclean()
749 * to zero sub-DEV_BSIZE'd portions of the page.
750 */
751 vm_page_set_validclean(m, 0, nread - toff);
752 } else {
753 m->valid = 0;
754 vm_page_undirty(m);
755 }
756
757 if (i != ap->a_reqpage) {
758 /*
759 * Just in case someone was asking for this page we
760 * now tell them that it is ok to use.
761 */
762 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
763 if (m->valid) {
764 if (m->flags & PG_WANTED) {
765 vm_page_activate(m);
766 } else {
767 vm_page_deactivate(m);
768 }
769 vm_page_wakeup(m);
770 } else {
771 vm_page_free(m);
772 }
773 } else {
774 vm_page_free(m);
775 }
776 } else if (m->valid) {
777 gotreqpage = 1;
778 /*
779 * Since this is a VM request, we need to make the
780 * entire page presentable by zeroing invalid sections.
781 */
782 if (m->valid != VM_PAGE_BITS_ALL)
783 vm_page_zero_invalid(m, FALSE);
784 }
785 }
786 if (!gotreqpage) {
787 m = ap->a_m[ap->a_reqpage];
788 printf(
789 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
81b5c339 790 devtoname(vp->v_rdev), error, bp, bp->b_vp);
984263bc 791 printf(
54078292 792 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
984263bc
MD
793 size, bp->b_resid, ap->a_count, m->valid);
794 printf(
795 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
796 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
797 /*
798 * Free the buffer header back to the swap buffer pool.
799 */
800 relpbuf(bp, NULL);
801 return VM_PAGER_ERROR;
802 }
803 /*
804 * Free the buffer header back to the swap buffer pool.
805 */
806 relpbuf(bp, NULL);
807 return VM_PAGER_OK;
808}