Add a fairly bad hack to detect ripouts that might occur during a list
[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 $
71c18fe3 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.44 2006/05/07 19:17:18 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
2d3e977e 79struct vop_ops *spec_vnode_vops;
0961aa92 80struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
2d3e977e
MD
81 { &vop_default_desc, vop_defaultop },
82 { &vop_access_desc, vop_ebadf },
625ddaba
JS
83 { &vop_advlock_desc, (vnodeopv_entry_t) spec_advlock },
84 { &vop_bmap_desc, (vnodeopv_entry_t) spec_bmap },
85 { &vop_close_desc, (vnodeopv_entry_t) spec_close },
e62afb5f 86 { &vop_old_create_desc, vop_panic },
625ddaba
JS
87 { &vop_freeblks_desc, (vnodeopv_entry_t) spec_freeblks },
88 { &vop_fsync_desc, (vnodeopv_entry_t) spec_fsync },
89 { &vop_getpages_desc, (vnodeopv_entry_t) spec_getpages },
90 { &vop_inactive_desc, (vnodeopv_entry_t) spec_inactive },
91 { &vop_ioctl_desc, (vnodeopv_entry_t) spec_ioctl },
e62afb5f
MD
92 { &vop_old_link_desc, vop_panic },
93 { &vop_old_mkdir_desc, vop_panic },
94 { &vop_old_mknod_desc, vop_panic },
625ddaba
JS
95 { &vop_open_desc, (vnodeopv_entry_t) spec_open },
96 { &vop_pathconf_desc, (vnodeopv_entry_t) vop_stdpathconf },
97 { &vop_poll_desc, (vnodeopv_entry_t) spec_poll },
98 { &vop_kqfilter_desc, (vnodeopv_entry_t) spec_kqfilter },
99 { &vop_print_desc, (vnodeopv_entry_t) spec_print },
100 { &vop_read_desc, (vnodeopv_entry_t) spec_read },
2d3e977e
MD
101 { &vop_readdir_desc, vop_panic },
102 { &vop_readlink_desc, vop_panic },
103 { &vop_reallocblks_desc, vop_panic },
104 { &vop_reclaim_desc, vop_null },
e62afb5f
MD
105 { &vop_old_remove_desc, vop_panic },
106 { &vop_old_rename_desc, vop_panic },
107 { &vop_old_rmdir_desc, vop_panic },
2d3e977e 108 { &vop_setattr_desc, vop_ebadf },
625ddaba 109 { &vop_strategy_desc, (vnodeopv_entry_t) spec_strategy },
e62afb5f 110 { &vop_old_symlink_desc, vop_panic },
625ddaba 111 { &vop_write_desc, (vnodeopv_entry_t) spec_write },
984263bc
MD
112 { NULL, NULL }
113};
114static struct vnodeopv_desc spec_vnodeop_opv_desc =
dc1be39c 115 { &spec_vnode_vops, spec_vnodeop_entries, 0 };
984263bc
MD
116VNODEOP_SET(spec_vnodeop_opv_desc);
117
e4c9c0c8
MD
118extern int dev_ref_debug;
119
50775720
CP
120/*
121 * spec_vnoperate(struct vnodeop_desc *a_desc, ...)
122 */
984263bc 123int
50775720 124spec_vnoperate(struct vop_generic_args *ap)
984263bc 125{
0961aa92 126 return (VOCALL(spec_vnode_vops, ap));
984263bc
MD
127}
128
81b5c339 129static void spec_getpages_iodone (struct bio *bio);
984263bc
MD
130
131/*
132 * Open a special file.
50775720
CP
133 *
134 * spec_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
8ddc6004 135 * struct file *a_fp, struct thread *a_td)
984263bc
MD
136 */
137/* ARGSUSED */
138static int
50775720 139spec_open(struct vop_open_args *ap)
984263bc 140{
984263bc 141 struct vnode *vp = ap->a_vp;
e4c9c0c8 142 dev_t dev;
984263bc 143 int error;
e4c9c0c8 144 int isblk = (vp->v_type == VBLK) ? 1 : 0;
984263bc
MD
145 const char *cp;
146
147 /*
148 * Don't allow open if fs is mounted -nodev.
149 */
150 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
151 return (ENXIO);
152
e4c9c0c8
MD
153 /*
154 * Resolve the device. If the vnode is already open v_rdev may
155 * already be resolved. However, if the device changes out from
1fbb5fc0
MD
156 * under us we report it (and, for now, we allow it). Since
157 * v_release_rdev() zero's v_opencount, we have to save and restore
158 * it when replacing the rdev reference.
e4c9c0c8
MD
159 */
160 if (vp->v_rdev != NULL) {
161 dev = udev2dev(vp->v_udev, isblk);
162 if (dev != vp->v_rdev) {
1fbb5fc0 163 int oc = vp->v_opencount;
e4c9c0c8
MD
164 printf(
165 "Warning: spec_open: dev %s was lost",
166 vp->v_rdev->si_name);
167 v_release_rdev(vp);
168 error = v_associate_rdev(vp,
169 udev2dev(vp->v_udev, isblk));
1fbb5fc0 170 if (error) {
e4c9c0c8 171 printf(", reacquisition failed\n");
1fbb5fc0
MD
172 } else {
173 vp->v_opencount = oc;
e4c9c0c8 174 printf(", reacquisition successful\n");
1fbb5fc0 175 }
e4c9c0c8
MD
176 } else {
177 error = 0;
178 }
179 } else {
180 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
181 }
182 if (error)
183 return(error);
184
500df1da
MD
185 /*
186 * Prevent degenerate open/close sequences from nulling out rdev.
187 */
e4c9c0c8 188 dev = vp->v_rdev;
500df1da 189 KKASSERT(dev != NULL);
984263bc 190
e4c9c0c8
MD
191 /*
192 * Make this field valid before any I/O in ->d_open. XXX the
193 * device itself should probably be required to initialize
194 * this field in d_open.
195 */
984263bc
MD
196 if (!dev->si_iosize_max)
197 dev->si_iosize_max = DFLTPHYS;
198
199 /*
200 * XXX: Disks get special billing here, but it is mostly wrong.
201 * XXX: diskpartitions can overlap and the real checks should
202 * XXX: take this into account, and consequently they need to
203 * XXX: live in the diskslicing code. Some checks do.
204 */
205 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
206 (ap->a_mode & FWRITE)) {
207 /*
208 * Never allow opens for write if the device is mounted R/W
209 */
e4c9c0c8
MD
210 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
211 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
212 error = EBUSY;
213 goto done;
214 }
984263bc
MD
215
216 /*
217 * When running in secure mode, do not allow opens
218 * for writing if the device is mounted
219 */
e4c9c0c8
MD
220 if (securelevel >= 1 && vfs_mountedon(vp)) {
221 error = EPERM;
222 goto done;
223 }
984263bc
MD
224
225 /*
226 * When running in very secure mode, do not allow
227 * opens for writing of any devices.
228 */
e4c9c0c8
MD
229 if (securelevel >= 2) {
230 error = EPERM;
231 goto done;
232 }
984263bc
MD
233 }
234
235 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 236 if (dev_dflags(dev) & D_TTY)
984263bc
MD
237 vp->v_flag |= VISTTY;
238
e4c9c0c8
MD
239 /*
240 * dev_dopen() is always called for each open. dev_dclose() is
241 * only called for the last close unless D_TRACKCLOSE is set.
242 */
ca466bae 243 VOP_UNLOCK(vp, 0);
87de5057 244 error = dev_dopen(dev, ap->a_mode, S_IFCHR, curthread);
ca466bae 245 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
246
247 if (error)
e4c9c0c8 248 goto done;
984263bc 249
335dda38 250 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
251 if (dev->si_tty) {
252 struct tty *tp;
253 tp = dev->si_tty;
254 if (!tp->t_stop) {
255 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
256 tp->t_stop = nottystop;
257 }
258 }
259 }
260
7540ab49
MD
261 /*
262 * If this is 'disk' or disk-like device, associate a VM object
263 * with it.
264 */
984263bc
MD
265 if (vn_isdisk(vp, NULL)) {
266 if (!dev->si_bsize_phys)
267 dev->si_bsize_phys = DEV_BSIZE;
1c843a13 268 vinitvmio(vp, IDX_TO_OFF(INT_MAX));
984263bc 269 }
335dda38 270 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 271 cp = devtoname(dev);
335dda38 272 if (*cp == '#') {
984263bc 273 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 274 dev_dname(dev), cp);
984263bc
MD
275 }
276 }
fad57d0e
MD
277
278 /*
279 * If we were handed a file pointer we may be able to install a
280 * shortcut which issues device read and write operations directly
281 * from the fileops rather then having to go through spec_read()
282 * and spec_write().
283 */
284 if (ap->a_fp)
285 vn_setspecops(ap->a_fp);
286
e4c9c0c8
MD
287 if (dev_ref_debug)
288 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
289done:
290 if (error) {
8ddc6004 291 if (vp->v_opencount == 0)
e4c9c0c8 292 v_release_rdev(vp);
8ddc6004
MD
293 } else {
294 vop_stdopen(ap);
e4c9c0c8 295 }
984263bc
MD
296 return (error);
297}
298
299/*
300 * Vnode op for read
50775720
CP
301 *
302 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
303 * struct ucred *a_cred)
984263bc
MD
304 */
305/* ARGSUSED */
306static int
50775720 307spec_read(struct vop_read_args *ap)
984263bc
MD
308{
309 struct vnode *vp;
dadab5e9 310 struct thread *td;
984263bc
MD
311 struct uio *uio;
312 dev_t dev;
313 int error;
314
315 vp = ap->a_vp;
316 dev = vp->v_rdev;
317 uio = ap->a_uio;
dadab5e9 318 td = uio->uio_td;
984263bc 319
72402f22
MD
320 if (dev == NULL) /* device was revoked */
321 return (EBADF);
984263bc
MD
322 if (uio->uio_resid == 0)
323 return (0);
324
ca466bae 325 VOP_UNLOCK(vp, 0);
335dda38 326 error = dev_dread(dev, uio, ap->a_ioflag);
ca466bae 327 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
328 return (error);
329}
330
331/*
332 * Vnode op for write
50775720
CP
333 *
334 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
335 * struct ucred *a_cred)
984263bc
MD
336 */
337/* ARGSUSED */
338static int
50775720 339spec_write(struct vop_write_args *ap)
984263bc
MD
340{
341 struct vnode *vp;
dadab5e9 342 struct thread *td;
984263bc
MD
343 struct uio *uio;
344 dev_t dev;
345 int error;
346
347 vp = ap->a_vp;
348 dev = vp->v_rdev;
349 uio = ap->a_uio;
dadab5e9 350 td = uio->uio_td;
984263bc 351
72402f22
MD
352 if (dev == NULL) /* device was revoked */
353 return (EBADF);
354
ca466bae 355 VOP_UNLOCK(vp, 0);
335dda38 356 error = dev_dwrite(dev, uio, ap->a_ioflag);
ca466bae 357 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
358 return (error);
359}
360
361/*
362 * Device ioctl operation.
50775720
CP
363 *
364 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
365 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
366 */
367/* ARGSUSED */
368static int
50775720 369spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
370{
371 dev_t dev;
372
72402f22
MD
373 if ((dev = ap->a_vp->v_rdev) == NULL)
374 return (EBADF); /* device was revoked */
375
335dda38 376 return (dev_dioctl(dev, ap->a_command, ap->a_data,
87de5057 377 ap->a_fflag,curthread));
984263bc
MD
378}
379
50775720
CP
380/*
381 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
382 * struct thread *a_td)
383 */
984263bc
MD
384/* ARGSUSED */
385static int
50775720 386spec_poll(struct vop_poll_args *ap)
984263bc
MD
387{
388 dev_t dev;
389
72402f22
MD
390 if ((dev = ap->a_vp->v_rdev) == NULL)
391 return (EBADF); /* device was revoked */
87de5057 392 return (dev_dpoll(dev, ap->a_events, curthread));
984263bc
MD
393}
394
50775720
CP
395/*
396 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
397 */
984263bc
MD
398/* ARGSUSED */
399static int
50775720 400spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
401{
402 dev_t dev;
403
72402f22
MD
404 if ((dev = ap->a_vp->v_rdev) == NULL)
405 return (EBADF); /* device was revoked */
335dda38 406 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
407}
408
409/*
410 * Synch buffers associated with a block device
50775720
CP
411 *
412 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
413 * int a_waitfor, struct thread *a_td)
984263bc
MD
414 */
415/* ARGSUSED */
416static int
50775720 417spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
418{
419 struct vnode *vp = ap->a_vp;
6bae6177 420 int error;
984263bc
MD
421
422 if (!vn_isdisk(vp, NULL))
423 return (0);
424
984263bc
MD
425 /*
426 * Flush all dirty buffers associated with a block device.
427 */
4e0ecc94 428 error = vfsync(vp, ap->a_waitfor, 10000, NULL, NULL);
6bae6177 429 return (error);
984263bc
MD
430}
431
50775720
CP
432/*
433 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
434 */
984263bc 435static int
50775720 436spec_inactive(struct vop_inactive_args *ap)
984263bc 437{
984263bc
MD
438 return (0);
439}
440
441/*
442 * Just call the device strategy routine
50775720 443 *
81b5c339 444 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
984263bc
MD
445 */
446static int
50775720 447spec_strategy(struct vop_strategy_args *ap)
984263bc 448{
81b5c339
MD
449 struct bio *bio = ap->a_bio;
450 struct buf *bp = bio->bio_buf;
984263bc
MD
451 struct vnode *vp;
452 struct mount *mp;
453
10f3fee5
MD
454 if (bp->b_cmd != BUF_CMD_READ &&
455 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start) {
984263bc 456 (*bioops.io_start)(bp);
10f3fee5 457 }
984263bc
MD
458
459 /*
460 * Collect statistics on synchronous and asynchronous read
461 * and write counts for disks that have associated filesystems.
462 */
463 vp = ap->a_vp;
72402f22 464 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 465 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
10f3fee5 466 if (bp->b_cmd == BUF_CMD_READ) {
df4f70a6 467 if (bp->b_flags & B_ASYNC)
10f3fee5 468 mp->mnt_stat.f_asyncreads++;
984263bc 469 else
10f3fee5 470 mp->mnt_stat.f_syncreads++;
984263bc 471 } else {
df4f70a6 472 if (bp->b_flags & B_ASYNC)
10f3fee5 473 mp->mnt_stat.f_asyncwrites++;
984263bc 474 else
10f3fee5 475 mp->mnt_stat.f_syncwrites++;
984263bc
MD
476 }
477 }
81b5c339 478 dev_dstrategy_chain(vp->v_rdev, bio);
984263bc
MD
479 return (0);
480}
481
50775720
CP
482/*
483 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
484 */
984263bc 485static int
50775720 486spec_freeblks(struct vop_freeblks_args *ap)
984263bc 487{
984263bc
MD
488 struct buf *bp;
489
490 /*
491 * XXX: This assumes that strategy does the deed right away.
492 * XXX: this may not be TRTTD.
493 */
72402f22 494 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 495 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
496 return (0);
497 bp = geteblk(ap->a_length);
10f3fee5 498 bp->b_cmd = BUF_CMD_FREEBLKS;
54078292 499 bp->b_bio1.bio_offset = ap->a_offset;
984263bc 500 bp->b_bcount = ap->a_length;
81b5c339 501 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
984263bc
MD
502 return (0);
503}
504
505/*
506 * Implement degenerate case where the block requested is the block
507 * returned, and assume that the entire device is contiguous in regards
508 * to the contiguous block range (runp and runb).
50775720 509 *
54078292
MD
510 * spec_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
511 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
512 */
513static int
50775720 514spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
515{
516 struct vnode *vp = ap->a_vp;
984263bc
MD
517
518 if (ap->a_vpp != NULL)
519 *ap->a_vpp = vp;
54078292
MD
520 if (ap->a_doffsetp != NULL)
521 *ap->a_doffsetp = ap->a_loffset;
984263bc 522 if (ap->a_runp != NULL)
54078292
MD
523 *ap->a_runp = MAXBSIZE;
524 if (ap->a_runb != NULL) {
525 if (ap->a_loffset < MAXBSIZE)
526 *ap->a_runb = (int)ap->a_loffset;
527 else
528 *ap->a_runb = MAXBSIZE;
529 }
984263bc
MD
530 return (0);
531}
532
533/*
534 * Device close routine
50775720
CP
535 *
536 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
537 * struct thread *a_td)
53994a38
MD
538 *
539 * NOTE: the vnode may or may not be locked on call.
984263bc
MD
540 */
541/* ARGSUSED */
542static int
50775720 543spec_close(struct vop_close_args *ap)
984263bc 544{
87de5057 545 struct proc *p = curproc;
984263bc 546 struct vnode *vp = ap->a_vp;
984263bc 547 dev_t dev = vp->v_rdev;
e4c9c0c8 548 int error;
53994a38 549 int needrelock;
984263bc
MD
550
551 /*
552 * Hack: a tty device that is a controlling terminal
553 * has a reference from the session structure.
554 * We cannot easily tell that a character device is
555 * a controlling terminal, unless it is the closing
556 * process' controlling terminal. In that case,
557 * if the reference count is 2 (this last descriptor
558 * plus the session), release the reference from the session.
1fbb5fc0
MD
559 *
560 * It is possible for v_opencount to be 0 or 1 in this case, 0
561 * because the tty might have been revoked.
984263bc 562 */
72402f22
MD
563 if (dev)
564 reference_dev(dev);
1fbb5fc0 565 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 566 p && vp == p->p_session->s_ttyvp) {
984263bc 567 p->p_session->s_ttyvp = NULL;
1fbb5fc0 568 vrele(vp);
984263bc 569 }
e4c9c0c8 570
984263bc 571 /*
e4c9c0c8
MD
572 * Vnodes can be opened and close multiple times. Do not really
573 * close the device unless (1) it is being closed forcibly,
574 * (2) the device wants to track closes, or (3) this is the last
575 * vnode doing its last close on the device.
576 *
577 * XXX the VXLOCK (force close) case can leave vnodes referencing
578 * a closed device.
984263bc 579 */
5fd012e0 580 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 581 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 582 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
53994a38 583 needrelock = 0;
87de5057 584 if (VOP_ISLOCKED(vp, curthread)) {
53994a38 585 needrelock = 1;
ca466bae 586 VOP_UNLOCK(vp, 0);
53994a38 587 }
87de5057 588 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, curthread);
53994a38 589 if (needrelock)
ca466bae 590 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
e4c9c0c8
MD
591 } else {
592 error = 0;
984263bc 593 }
e4c9c0c8
MD
594
595 /*
596 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
597 * disassociates the rdev. If the rdev is already disassociated
598 * the vnode might have been revoked and no further opencount
599 * tracking occurs.
e4c9c0c8 600 */
1fbb5fc0 601 if (dev) {
8ddc6004 602 /*KKASSERT(vp->v_opencount > 0);*/
1fbb5fc0 603 if (dev_ref_debug) {
72402f22
MD
604 printf("spec_close: %s %d\n",
605 dev->si_name, vp->v_opencount - 1);
72402f22 606 }
8ddc6004 607 if (vp->v_opencount == 1)
1fbb5fc0 608 v_release_rdev(vp);
72402f22 609 release_dev(dev);
1fbb5fc0 610 }
8ddc6004 611 vop_stdclose(ap);
e4c9c0c8 612 return(error);
984263bc
MD
613}
614
615/*
616 * Print out the contents of a special device vnode.
50775720
CP
617 *
618 * spec_print(struct vnode *a_vp)
984263bc
MD
619 */
620static int
50775720 621spec_print(struct vop_print_args *ap)
984263bc 622{
984263bc
MD
623 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
624 return (0);
625}
626
627/*
628 * Special device advisory byte-level locks.
50775720
CP
629 *
630 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
631 * struct flock *a_fl, int a_flags)
984263bc
MD
632 */
633/* ARGSUSED */
634static int
50775720 635spec_advlock(struct vop_advlock_args *ap)
984263bc 636{
71c18fe3 637 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
984263bc
MD
638}
639
640static void
81b5c339 641spec_getpages_iodone(struct bio *bio)
984263bc 642{
10f3fee5 643 bio->bio_buf->b_cmd = BUF_CMD_DONE;
81b5c339 644 wakeup(bio->bio_buf);
984263bc
MD
645}
646
647static int
50775720 648spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
649{
650 vm_offset_t kva;
651 int error;
165dba55 652 int i, pcount, size;
984263bc
MD
653 struct buf *bp;
654 vm_page_t m;
655 vm_ooffset_t offset;
656 int toff, nextoff, nread;
657 struct vnode *vp = ap->a_vp;
658 int blksiz;
659 int gotreqpage;
660
661 error = 0;
662 pcount = round_page(ap->a_count) / PAGE_SIZE;
663
664 /*
665 * Calculate the offset of the transfer and do sanity check.
984263bc
MD
666 */
667 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
668
984263bc
MD
669 /*
670 * Round up physical size for real devices. We cannot round using
671 * v_mount's block size data because v_mount has nothing to do with
672 * the device. i.e. it's usually '/dev'. We need the physical block
673 * size for the device itself.
674 *
e4c9c0c8 675 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
676 * block device is mounted. However, we can use v_rdev.
677 */
678
679 if (vn_isdisk(vp, NULL))
680 blksiz = vp->v_rdev->si_bsize_phys;
681 else
682 blksiz = DEV_BSIZE;
683
684 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
685
686 bp = getpbuf(NULL);
687 kva = (vm_offset_t)bp->b_data;
688
689 /*
690 * Map the pages to be read into the kva.
691 */
692 pmap_qenter(kva, ap->a_m, pcount);
693
694 /* Build a minimal buffer header. */
10f3fee5 695 bp->b_cmd = BUF_CMD_READ;
984263bc 696 bp->b_bcount = size;
984263bc 697 bp->b_resid = 0;
9a71d53f 698 bp->b_runningbufspace = size;
984263bc
MD
699 runningbufspace += bp->b_runningbufspace;
700
54078292 701 bp->b_bio1.bio_offset = offset;
81b5c339
MD
702 bp->b_bio1.bio_done = spec_getpages_iodone;
703
12e4aaff
MD
704 mycpu->gd_cnt.v_vnodein++;
705 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
706
707 /* Do the input. */
81b5c339 708 vn_strategy(ap->a_vp, &bp->b_bio1);
984263bc 709
165dba55 710 crit_enter();
984263bc
MD
711
712 /* We definitely need to be at splbio here. */
10f3fee5 713 while (bp->b_cmd != BUF_CMD_DONE)
377d4740 714 tsleep(bp, 0, "spread", 0);
984263bc 715
165dba55 716 crit_exit();
984263bc 717
9a71d53f 718 if (bp->b_flags & B_ERROR) {
984263bc
MD
719 if (bp->b_error)
720 error = bp->b_error;
721 else
722 error = EIO;
723 }
724
9a71d53f
MD
725 /*
726 * If EOF is encountered we must zero-extend the result in order
727 * to ensure that the page does not contain garabge. When no
728 * error occurs, an early EOF is indicated if b_bcount got truncated.
729 * b_resid is relative to b_bcount and should be 0, but some devices
730 * might indicate an EOF with b_resid instead of truncating b_bcount.
731 */
732 nread = bp->b_bcount - bp->b_resid;
733 if (nread < ap->a_count)
734 bzero((caddr_t)kva + nread, ap->a_count - nread);
984263bc
MD
735 pmap_qremove(kva, pcount);
736
984263bc
MD
737 gotreqpage = 0;
738 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
739 nextoff = toff + PAGE_SIZE;
740 m = ap->a_m[i];
741
742 m->flags &= ~PG_ZERO;
743
744 if (nextoff <= nread) {
745 m->valid = VM_PAGE_BITS_ALL;
746 vm_page_undirty(m);
747 } else if (toff < nread) {
748 /*
749 * Since this is a VM request, we have to supply the
750 * unaligned offset to allow vm_page_set_validclean()
751 * to zero sub-DEV_BSIZE'd portions of the page.
752 */
753 vm_page_set_validclean(m, 0, nread - toff);
754 } else {
755 m->valid = 0;
756 vm_page_undirty(m);
757 }
758
759 if (i != ap->a_reqpage) {
760 /*
761 * Just in case someone was asking for this page we
762 * now tell them that it is ok to use.
763 */
764 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
765 if (m->valid) {
766 if (m->flags & PG_WANTED) {
767 vm_page_activate(m);
768 } else {
769 vm_page_deactivate(m);
770 }
771 vm_page_wakeup(m);
772 } else {
773 vm_page_free(m);
774 }
775 } else {
776 vm_page_free(m);
777 }
778 } else if (m->valid) {
779 gotreqpage = 1;
780 /*
781 * Since this is a VM request, we need to make the
782 * entire page presentable by zeroing invalid sections.
783 */
784 if (m->valid != VM_PAGE_BITS_ALL)
785 vm_page_zero_invalid(m, FALSE);
786 }
787 }
788 if (!gotreqpage) {
789 m = ap->a_m[ap->a_reqpage];
790 printf(
791 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
81b5c339 792 devtoname(vp->v_rdev), error, bp, bp->b_vp);
984263bc 793 printf(
54078292 794 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
984263bc
MD
795 size, bp->b_resid, ap->a_count, m->valid);
796 printf(
797 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
798 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
799 /*
800 * Free the buffer header back to the swap buffer pool.
801 */
802 relpbuf(bp, NULL);
803 return VM_PAGER_ERROR;
804 }
805 /*
806 * Free the buffer header back to the swap buffer pool.
807 */
808 relpbuf(bp, NULL);
809 return VM_PAGER_OK;
810}