Remove owi(4) from build.
[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 $
7540ab49 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.33 2006/03/29 18:45:03 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
3020e3be
MD
57#include <sys/buf2.h>
58
165dba55
DR
59#include <sys/thread2.h>
60
a6ee311a
RG
61static int spec_advlock (struct vop_advlock_args *);
62static int spec_bmap (struct vop_bmap_args *);
63static int spec_close (struct vop_close_args *);
64static int spec_freeblks (struct vop_freeblks_args *);
65static int spec_fsync (struct vop_fsync_args *);
66static int spec_getpages (struct vop_getpages_args *);
67static int spec_inactive (struct vop_inactive_args *);
68static int spec_ioctl (struct vop_ioctl_args *);
69static int spec_open (struct vop_open_args *);
70static int spec_poll (struct vop_poll_args *);
71static int spec_kqfilter (struct vop_kqfilter_args *);
72static int spec_print (struct vop_print_args *);
73static int spec_read (struct vop_read_args *);
74static int spec_strategy (struct vop_strategy_args *);
75static int spec_write (struct vop_write_args *);
984263bc 76
2d3e977e 77struct vop_ops *spec_vnode_vops;
0961aa92 78struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
2d3e977e
MD
79 { &vop_default_desc, vop_defaultop },
80 { &vop_access_desc, vop_ebadf },
625ddaba
JS
81 { &vop_advlock_desc, (vnodeopv_entry_t) spec_advlock },
82 { &vop_bmap_desc, (vnodeopv_entry_t) spec_bmap },
83 { &vop_close_desc, (vnodeopv_entry_t) spec_close },
e62afb5f 84 { &vop_old_create_desc, vop_panic },
625ddaba
JS
85 { &vop_freeblks_desc, (vnodeopv_entry_t) spec_freeblks },
86 { &vop_fsync_desc, (vnodeopv_entry_t) spec_fsync },
87 { &vop_getpages_desc, (vnodeopv_entry_t) spec_getpages },
88 { &vop_inactive_desc, (vnodeopv_entry_t) spec_inactive },
89 { &vop_ioctl_desc, (vnodeopv_entry_t) spec_ioctl },
e62afb5f
MD
90 { &vop_old_link_desc, vop_panic },
91 { &vop_old_mkdir_desc, vop_panic },
92 { &vop_old_mknod_desc, vop_panic },
625ddaba
JS
93 { &vop_open_desc, (vnodeopv_entry_t) spec_open },
94 { &vop_pathconf_desc, (vnodeopv_entry_t) vop_stdpathconf },
95 { &vop_poll_desc, (vnodeopv_entry_t) spec_poll },
96 { &vop_kqfilter_desc, (vnodeopv_entry_t) spec_kqfilter },
97 { &vop_print_desc, (vnodeopv_entry_t) spec_print },
98 { &vop_read_desc, (vnodeopv_entry_t) spec_read },
2d3e977e
MD
99 { &vop_readdir_desc, vop_panic },
100 { &vop_readlink_desc, vop_panic },
101 { &vop_reallocblks_desc, vop_panic },
102 { &vop_reclaim_desc, vop_null },
e62afb5f
MD
103 { &vop_old_remove_desc, vop_panic },
104 { &vop_old_rename_desc, vop_panic },
105 { &vop_old_rmdir_desc, vop_panic },
2d3e977e 106 { &vop_setattr_desc, vop_ebadf },
625ddaba 107 { &vop_strategy_desc, (vnodeopv_entry_t) spec_strategy },
e62afb5f 108 { &vop_old_symlink_desc, vop_panic },
625ddaba 109 { &vop_write_desc, (vnodeopv_entry_t) spec_write },
984263bc
MD
110 { NULL, NULL }
111};
112static struct vnodeopv_desc spec_vnodeop_opv_desc =
dc1be39c 113 { &spec_vnode_vops, spec_vnodeop_entries, 0 };
984263bc
MD
114VNODEOP_SET(spec_vnodeop_opv_desc);
115
e4c9c0c8
MD
116extern int dev_ref_debug;
117
50775720
CP
118/*
119 * spec_vnoperate(struct vnodeop_desc *a_desc, ...)
120 */
984263bc 121int
50775720 122spec_vnoperate(struct vop_generic_args *ap)
984263bc 123{
0961aa92 124 return (VOCALL(spec_vnode_vops, ap));
984263bc
MD
125}
126
81b5c339 127static void spec_getpages_iodone (struct bio *bio);
984263bc
MD
128
129/*
130 * Open a special file.
50775720
CP
131 *
132 * spec_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
133 * struct thread *a_td)
984263bc
MD
134 */
135/* ARGSUSED */
136static int
50775720 137spec_open(struct vop_open_args *ap)
984263bc 138{
984263bc 139 struct vnode *vp = ap->a_vp;
e4c9c0c8 140 dev_t dev;
984263bc 141 int error;
e4c9c0c8 142 int isblk = (vp->v_type == VBLK) ? 1 : 0;
984263bc
MD
143 const char *cp;
144
145 /*
146 * Don't allow open if fs is mounted -nodev.
147 */
148 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
149 return (ENXIO);
150
e4c9c0c8
MD
151 /*
152 * Resolve the device. If the vnode is already open v_rdev may
153 * already be resolved. However, if the device changes out from
1fbb5fc0
MD
154 * under us we report it (and, for now, we allow it). Since
155 * v_release_rdev() zero's v_opencount, we have to save and restore
156 * it when replacing the rdev reference.
e4c9c0c8
MD
157 */
158 if (vp->v_rdev != NULL) {
159 dev = udev2dev(vp->v_udev, isblk);
160 if (dev != vp->v_rdev) {
1fbb5fc0 161 int oc = vp->v_opencount;
e4c9c0c8
MD
162 printf(
163 "Warning: spec_open: dev %s was lost",
164 vp->v_rdev->si_name);
165 v_release_rdev(vp);
166 error = v_associate_rdev(vp,
167 udev2dev(vp->v_udev, isblk));
1fbb5fc0 168 if (error) {
e4c9c0c8 169 printf(", reacquisition failed\n");
1fbb5fc0
MD
170 } else {
171 vp->v_opencount = oc;
e4c9c0c8 172 printf(", reacquisition successful\n");
1fbb5fc0 173 }
e4c9c0c8
MD
174 } else {
175 error = 0;
176 }
177 } else {
178 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
179 }
180 if (error)
181 return(error);
182
500df1da
MD
183 /*
184 * Prevent degenerate open/close sequences from nulling out rdev.
185 */
186 ++vp->v_opencount;
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 */
5fd012e0 242 VOP_UNLOCK(vp, 0, ap->a_td);
335dda38 243 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
5fd012e0 244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
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;
7540ab49 267 vinitvmio(vp);
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) {
500df1da
MD
290 KKASSERT(vp->v_opencount > 0);
291 if (--vp->v_opencount == 0)
e4c9c0c8
MD
292 v_release_rdev(vp);
293 }
984263bc
MD
294 return (error);
295}
296
297/*
298 * Vnode op for read
50775720
CP
299 *
300 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
301 * struct ucred *a_cred)
984263bc
MD
302 */
303/* ARGSUSED */
304static int
50775720 305spec_read(struct vop_read_args *ap)
984263bc
MD
306{
307 struct vnode *vp;
dadab5e9 308 struct thread *td;
984263bc
MD
309 struct uio *uio;
310 dev_t dev;
311 int error;
312
313 vp = ap->a_vp;
314 dev = vp->v_rdev;
315 uio = ap->a_uio;
dadab5e9 316 td = uio->uio_td;
984263bc 317
72402f22
MD
318 if (dev == NULL) /* device was revoked */
319 return (EBADF);
984263bc
MD
320 if (uio->uio_resid == 0)
321 return (0);
322
5fd012e0 323 VOP_UNLOCK(vp, 0, td);
335dda38 324 error = dev_dread(dev, uio, ap->a_ioflag);
5fd012e0 325 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
326 return (error);
327}
328
329/*
330 * Vnode op for write
50775720
CP
331 *
332 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
333 * struct ucred *a_cred)
984263bc
MD
334 */
335/* ARGSUSED */
336static int
50775720 337spec_write(struct vop_write_args *ap)
984263bc
MD
338{
339 struct vnode *vp;
dadab5e9 340 struct thread *td;
984263bc
MD
341 struct uio *uio;
342 dev_t dev;
343 int error;
344
345 vp = ap->a_vp;
346 dev = vp->v_rdev;
347 uio = ap->a_uio;
dadab5e9 348 td = uio->uio_td;
984263bc 349
72402f22
MD
350 if (dev == NULL) /* device was revoked */
351 return (EBADF);
352
5fd012e0 353 VOP_UNLOCK(vp, 0, td);
335dda38 354 error = dev_dwrite(dev, uio, ap->a_ioflag);
5fd012e0 355 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
356 return (error);
357}
358
359/*
360 * Device ioctl operation.
50775720
CP
361 *
362 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
363 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
364 */
365/* ARGSUSED */
366static int
50775720 367spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
368{
369 dev_t dev;
370
72402f22
MD
371 if ((dev = ap->a_vp->v_rdev) == NULL)
372 return (EBADF); /* device was revoked */
373
335dda38
MD
374 return (dev_dioctl(dev, ap->a_command, ap->a_data,
375 ap->a_fflag, ap->a_td));
984263bc
MD
376}
377
50775720
CP
378/*
379 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
380 * struct thread *a_td)
381 */
984263bc
MD
382/* ARGSUSED */
383static int
50775720 384spec_poll(struct vop_poll_args *ap)
984263bc
MD
385{
386 dev_t dev;
387
72402f22
MD
388 if ((dev = ap->a_vp->v_rdev) == NULL)
389 return (EBADF); /* device was revoked */
335dda38 390 return (dev_dpoll(dev, ap->a_events, ap->a_td));
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
MD
399{
400 dev_t dev;
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 */
54078292 426 error = vfsync(vp, ap->a_waitfor, 10000, NOOFFSET, 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
984263bc
MD
452 if (((bp->b_flags & B_READ) == 0) &&
453 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
454 (*bioops.io_start)(bp);
455
456 /*
457 * Collect statistics on synchronous and asynchronous read
458 * and write counts for disks that have associated filesystems.
459 */
460 vp = ap->a_vp;
72402f22 461 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 462 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 463 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 464 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
465 mp->mnt_stat.f_asyncwrites++;
466 else
467 mp->mnt_stat.f_syncwrites++;
468 } else {
dadab5e9 469 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
470 mp->mnt_stat.f_asyncreads++;
471 else
472 mp->mnt_stat.f_syncreads++;
473 }
474 }
81b5c339 475 dev_dstrategy_chain(vp->v_rdev, bio);
984263bc
MD
476 return (0);
477}
478
50775720
CP
479/*
480 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
481 */
984263bc 482static int
50775720 483spec_freeblks(struct vop_freeblks_args *ap)
984263bc 484{
984263bc
MD
485 struct buf *bp;
486
487 /*
488 * XXX: This assumes that strategy does the deed right away.
489 * XXX: this may not be TRTTD.
490 */
72402f22 491 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 492 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
493 return (0);
494 bp = geteblk(ap->a_length);
495 bp->b_flags |= B_FREEBUF;
54078292 496 bp->b_bio1.bio_offset = ap->a_offset;
984263bc 497 bp->b_bcount = ap->a_length;
81b5c339 498 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
984263bc
MD
499 return (0);
500}
501
502/*
503 * Implement degenerate case where the block requested is the block
504 * returned, and assume that the entire device is contiguous in regards
505 * to the contiguous block range (runp and runb).
50775720 506 *
54078292
MD
507 * spec_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
508 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
509 */
510static int
50775720 511spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
512{
513 struct vnode *vp = ap->a_vp;
984263bc
MD
514
515 if (ap->a_vpp != NULL)
516 *ap->a_vpp = vp;
54078292
MD
517 if (ap->a_doffsetp != NULL)
518 *ap->a_doffsetp = ap->a_loffset;
984263bc 519 if (ap->a_runp != NULL)
54078292
MD
520 *ap->a_runp = MAXBSIZE;
521 if (ap->a_runb != NULL) {
522 if (ap->a_loffset < MAXBSIZE)
523 *ap->a_runb = (int)ap->a_loffset;
524 else
525 *ap->a_runb = MAXBSIZE;
526 }
984263bc
MD
527 return (0);
528}
529
530/*
531 * Device close routine
50775720
CP
532 *
533 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
534 * struct thread *a_td)
53994a38
MD
535 *
536 * NOTE: the vnode may or may not be locked on call.
984263bc
MD
537 */
538/* ARGSUSED */
539static int
50775720 540spec_close(struct vop_close_args *ap)
984263bc 541{
dadab5e9 542 struct proc *p = ap->a_td->td_proc;
984263bc 543 struct vnode *vp = ap->a_vp;
984263bc 544 dev_t dev = vp->v_rdev;
e4c9c0c8 545 int error;
53994a38 546 int needrelock;
984263bc
MD
547
548 /*
549 * Hack: a tty device that is a controlling terminal
550 * has a reference from the session structure.
551 * We cannot easily tell that a character device is
552 * a controlling terminal, unless it is the closing
553 * process' controlling terminal. In that case,
554 * if the reference count is 2 (this last descriptor
555 * plus the session), release the reference from the session.
1fbb5fc0
MD
556 *
557 * It is possible for v_opencount to be 0 or 1 in this case, 0
558 * because the tty might have been revoked.
984263bc 559 */
72402f22
MD
560 if (dev)
561 reference_dev(dev);
1fbb5fc0 562 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 563 p && vp == p->p_session->s_ttyvp) {
984263bc 564 p->p_session->s_ttyvp = NULL;
1fbb5fc0 565 vrele(vp);
984263bc 566 }
e4c9c0c8 567
984263bc 568 /*
e4c9c0c8
MD
569 * Vnodes can be opened and close multiple times. Do not really
570 * close the device unless (1) it is being closed forcibly,
571 * (2) the device wants to track closes, or (3) this is the last
572 * vnode doing its last close on the device.
573 *
574 * XXX the VXLOCK (force close) case can leave vnodes referencing
575 * a closed device.
984263bc 576 */
5fd012e0 577 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 578 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 579 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
53994a38
MD
580 needrelock = 0;
581 if (VOP_ISLOCKED(vp, ap->a_td)) {
582 needrelock = 1;
583 VOP_UNLOCK(vp, 0, ap->a_td);
584 }
e4c9c0c8 585 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
53994a38
MD
586 if (needrelock)
587 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
e4c9c0c8
MD
588 } else {
589 error = 0;
984263bc 590 }
e4c9c0c8
MD
591
592 /*
593 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
594 * disassociates the rdev. If the rdev is already disassociated
595 * the vnode might have been revoked and no further opencount
596 * tracking occurs.
e4c9c0c8 597 */
1fbb5fc0
MD
598 if (dev) {
599 KKASSERT(vp->v_opencount > 0);
600 if (dev_ref_debug) {
72402f22
MD
601 printf("spec_close: %s %d\n",
602 dev->si_name, vp->v_opencount - 1);
72402f22 603 }
1fbb5fc0
MD
604 if (--vp->v_opencount == 0)
605 v_release_rdev(vp);
72402f22 606 release_dev(dev);
1fbb5fc0 607 }
e4c9c0c8 608 return(error);
984263bc
MD
609}
610
611/*
612 * Print out the contents of a special device vnode.
50775720
CP
613 *
614 * spec_print(struct vnode *a_vp)
984263bc
MD
615 */
616static int
50775720 617spec_print(struct vop_print_args *ap)
984263bc 618{
984263bc
MD
619 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
620 return (0);
621}
622
623/*
624 * Special device advisory byte-level locks.
50775720
CP
625 *
626 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
627 * struct flock *a_fl, int a_flags)
984263bc
MD
628 */
629/* ARGSUSED */
630static int
50775720 631spec_advlock(struct vop_advlock_args *ap)
984263bc 632{
984263bc
MD
633 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
634}
635
636static void
81b5c339 637spec_getpages_iodone(struct bio *bio)
984263bc 638{
81b5c339
MD
639 bio->bio_buf->b_flags |= B_DONE;
640 wakeup(bio->bio_buf);
984263bc
MD
641}
642
643static int
50775720 644spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
645{
646 vm_offset_t kva;
647 int error;
165dba55 648 int i, pcount, size;
984263bc
MD
649 struct buf *bp;
650 vm_page_t m;
651 vm_ooffset_t offset;
652 int toff, nextoff, nread;
653 struct vnode *vp = ap->a_vp;
654 int blksiz;
655 int gotreqpage;
656
657 error = 0;
658 pcount = round_page(ap->a_count) / PAGE_SIZE;
659
660 /*
661 * Calculate the offset of the transfer and do sanity check.
984263bc
MD
662 */
663 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
664
984263bc
MD
665 /*
666 * Round up physical size for real devices. We cannot round using
667 * v_mount's block size data because v_mount has nothing to do with
668 * the device. i.e. it's usually '/dev'. We need the physical block
669 * size for the device itself.
670 *
e4c9c0c8 671 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
672 * block device is mounted. However, we can use v_rdev.
673 */
674
675 if (vn_isdisk(vp, NULL))
676 blksiz = vp->v_rdev->si_bsize_phys;
677 else
678 blksiz = DEV_BSIZE;
679
680 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
681
682 bp = getpbuf(NULL);
683 kva = (vm_offset_t)bp->b_data;
684
685 /*
686 * Map the pages to be read into the kva.
687 */
688 pmap_qenter(kva, ap->a_m, pcount);
689
690 /* Build a minimal buffer header. */
5ff4cfb5 691 bp->b_flags = B_READ;
984263bc
MD
692
693 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
694 pbgetvp(ap->a_vp, bp);
695 bp->b_bcount = size;
696 bp->b_bufsize = size;
697 bp->b_resid = 0;
698 bp->b_runningbufspace = bp->b_bufsize;
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. */
26a0694b 713 while ((bp->b_flags & B_DONE) == 0) {
377d4740 714 tsleep(bp, 0, "spread", 0);
26a0694b 715 }
984263bc 716
165dba55 717 crit_exit();
984263bc
MD
718
719 if ((bp->b_flags & B_ERROR) != 0) {
720 if (bp->b_error)
721 error = bp->b_error;
722 else
723 error = EIO;
724 }
725
726 nread = size - bp->b_resid;
727
728 if (nread < ap->a_count) {
729 bzero((caddr_t)kva + nread,
730 ap->a_count - nread);
731 }
732 pmap_qremove(kva, pcount);
733
734
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}