ANSI-fication. No operational changes.
[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 $
500df1da 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.16 2004/05/26 00:11:56 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
a6ee311a
RG
59static int spec_advlock (struct vop_advlock_args *);
60static int spec_bmap (struct vop_bmap_args *);
61static int spec_close (struct vop_close_args *);
62static int spec_freeblks (struct vop_freeblks_args *);
63static int spec_fsync (struct vop_fsync_args *);
64static int spec_getpages (struct vop_getpages_args *);
65static int spec_inactive (struct vop_inactive_args *);
66static int spec_ioctl (struct vop_ioctl_args *);
67static int spec_open (struct vop_open_args *);
68static int spec_poll (struct vop_poll_args *);
69static int spec_kqfilter (struct vop_kqfilter_args *);
70static int spec_print (struct vop_print_args *);
71static int spec_read (struct vop_read_args *);
72static int spec_strategy (struct vop_strategy_args *);
73static int spec_write (struct vop_write_args *);
984263bc
MD
74
75vop_t **spec_vnodeop_p;
76static struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
77 { &vop_default_desc, (vop_t *) vop_defaultop },
78 { &vop_access_desc, (vop_t *) vop_ebadf },
79 { &vop_advlock_desc, (vop_t *) spec_advlock },
80 { &vop_bmap_desc, (vop_t *) spec_bmap },
81 { &vop_close_desc, (vop_t *) spec_close },
82 { &vop_create_desc, (vop_t *) vop_panic },
83 { &vop_freeblks_desc, (vop_t *) spec_freeblks },
84 { &vop_fsync_desc, (vop_t *) spec_fsync },
85 { &vop_getpages_desc, (vop_t *) spec_getpages },
86 { &vop_inactive_desc, (vop_t *) spec_inactive },
87 { &vop_ioctl_desc, (vop_t *) spec_ioctl },
88 { &vop_lease_desc, (vop_t *) vop_null },
89 { &vop_link_desc, (vop_t *) vop_panic },
90 { &vop_mkdir_desc, (vop_t *) vop_panic },
91 { &vop_mknod_desc, (vop_t *) vop_panic },
92 { &vop_open_desc, (vop_t *) spec_open },
93 { &vop_pathconf_desc, (vop_t *) vop_stdpathconf },
94 { &vop_poll_desc, (vop_t *) spec_poll },
95 { &vop_kqfilter_desc, (vop_t *) spec_kqfilter },
96 { &vop_print_desc, (vop_t *) spec_print },
97 { &vop_read_desc, (vop_t *) spec_read },
98 { &vop_readdir_desc, (vop_t *) vop_panic },
99 { &vop_readlink_desc, (vop_t *) vop_panic },
100 { &vop_reallocblks_desc, (vop_t *) vop_panic },
101 { &vop_reclaim_desc, (vop_t *) vop_null },
102 { &vop_remove_desc, (vop_t *) vop_panic },
103 { &vop_rename_desc, (vop_t *) vop_panic },
104 { &vop_rmdir_desc, (vop_t *) vop_panic },
105 { &vop_setattr_desc, (vop_t *) vop_ebadf },
106 { &vop_strategy_desc, (vop_t *) spec_strategy },
107 { &vop_symlink_desc, (vop_t *) vop_panic },
108 { &vop_write_desc, (vop_t *) spec_write },
109 { NULL, NULL }
110};
111static struct vnodeopv_desc spec_vnodeop_opv_desc =
112 { &spec_vnodeop_p, spec_vnodeop_entries };
113
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
MD
123{
124 return (VOCALL(spec_vnodeop_p, ap->a_desc->vdesc_offset, ap));
125}
126
a6ee311a 127static void spec_getpages_iodone (struct buf *bp);
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
154 * under us we report it (and, for now, we allow it).
155 */
156 if (vp->v_rdev != NULL) {
157 dev = udev2dev(vp->v_udev, isblk);
158 if (dev != vp->v_rdev) {
159 printf(
160 "Warning: spec_open: dev %s was lost",
161 vp->v_rdev->si_name);
162 v_release_rdev(vp);
163 error = v_associate_rdev(vp,
164 udev2dev(vp->v_udev, isblk));
165 if (error)
166 printf(", reacquisition failed\n");
167 else
168 printf(", reacquisition successful\n");
169 } else {
170 error = 0;
171 }
172 } else {
173 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
174 }
175 if (error)
176 return(error);
177
500df1da
MD
178 /*
179 * Prevent degenerate open/close sequences from nulling out rdev.
180 */
181 ++vp->v_opencount;
e4c9c0c8 182 dev = vp->v_rdev;
500df1da 183 KKASSERT(dev != NULL);
984263bc 184
e4c9c0c8
MD
185 /*
186 * Make this field valid before any I/O in ->d_open. XXX the
187 * device itself should probably be required to initialize
188 * this field in d_open.
189 */
984263bc
MD
190 if (!dev->si_iosize_max)
191 dev->si_iosize_max = DFLTPHYS;
192
193 /*
194 * XXX: Disks get special billing here, but it is mostly wrong.
195 * XXX: diskpartitions can overlap and the real checks should
196 * XXX: take this into account, and consequently they need to
197 * XXX: live in the diskslicing code. Some checks do.
198 */
199 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
200 (ap->a_mode & FWRITE)) {
201 /*
202 * Never allow opens for write if the device is mounted R/W
203 */
e4c9c0c8
MD
204 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
205 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
206 error = EBUSY;
207 goto done;
208 }
984263bc
MD
209
210 /*
211 * When running in secure mode, do not allow opens
212 * for writing if the device is mounted
213 */
e4c9c0c8
MD
214 if (securelevel >= 1 && vfs_mountedon(vp)) {
215 error = EPERM;
216 goto done;
217 }
984263bc
MD
218
219 /*
220 * When running in very secure mode, do not allow
221 * opens for writing of any devices.
222 */
e4c9c0c8
MD
223 if (securelevel >= 2) {
224 error = EPERM;
225 goto done;
226 }
984263bc
MD
227 }
228
229 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 230 if (dev_dflags(dev) & D_TTY)
984263bc
MD
231 vp->v_flag |= VISTTY;
232
e4c9c0c8
MD
233 /*
234 * dev_dopen() is always called for each open. dev_dclose() is
235 * only called for the last close unless D_TRACKCLOSE is set.
236 */
41a01a4d 237 VOP_UNLOCK(vp, NULL, 0, ap->a_td);
335dda38 238 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
41a01a4d 239 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
984263bc
MD
240
241 if (error)
e4c9c0c8 242 goto done;
984263bc 243
335dda38 244 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
245 if (dev->si_tty) {
246 struct tty *tp;
247 tp = dev->si_tty;
248 if (!tp->t_stop) {
249 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
250 tp->t_stop = nottystop;
251 }
252 }
253 }
254
255 if (vn_isdisk(vp, NULL)) {
256 if (!dev->si_bsize_phys)
257 dev->si_bsize_phys = DEV_BSIZE;
258 }
335dda38 259 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 260 cp = devtoname(dev);
335dda38 261 if (*cp == '#') {
984263bc 262 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 263 dev_dname(dev), cp);
984263bc
MD
264 }
265 }
e4c9c0c8
MD
266 if (dev_ref_debug)
267 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
268done:
269 if (error) {
500df1da
MD
270 KKASSERT(vp->v_opencount > 0);
271 if (--vp->v_opencount == 0)
e4c9c0c8
MD
272 v_release_rdev(vp);
273 }
984263bc
MD
274 return (error);
275}
276
277/*
278 * Vnode op for read
50775720
CP
279 *
280 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
281 * struct ucred *a_cred)
984263bc
MD
282 */
283/* ARGSUSED */
284static int
50775720 285spec_read(struct vop_read_args *ap)
984263bc
MD
286{
287 struct vnode *vp;
dadab5e9 288 struct thread *td;
984263bc
MD
289 struct uio *uio;
290 dev_t dev;
291 int error;
292
293 vp = ap->a_vp;
294 dev = vp->v_rdev;
295 uio = ap->a_uio;
dadab5e9 296 td = uio->uio_td;
984263bc
MD
297
298 if (uio->uio_resid == 0)
299 return (0);
300
41a01a4d 301 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 302 error = dev_dread(dev, uio, ap->a_ioflag);
41a01a4d 303 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
304 return (error);
305}
306
307/*
308 * Vnode op for write
50775720
CP
309 *
310 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
311 * struct ucred *a_cred)
984263bc
MD
312 */
313/* ARGSUSED */
314static int
50775720 315spec_write(struct vop_write_args *ap)
984263bc
MD
316{
317 struct vnode *vp;
dadab5e9 318 struct thread *td;
984263bc
MD
319 struct uio *uio;
320 dev_t dev;
321 int error;
322
323 vp = ap->a_vp;
324 dev = vp->v_rdev;
325 uio = ap->a_uio;
dadab5e9 326 td = uio->uio_td;
984263bc 327
41a01a4d 328 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 329 error = dev_dwrite(dev, uio, ap->a_ioflag);
41a01a4d 330 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
331 return (error);
332}
333
334/*
335 * Device ioctl operation.
50775720
CP
336 *
337 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
338 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
339 */
340/* ARGSUSED */
341static int
50775720 342spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
343{
344 dev_t dev;
345
346 dev = ap->a_vp->v_rdev;
335dda38
MD
347 return (dev_dioctl(dev, ap->a_command, ap->a_data,
348 ap->a_fflag, ap->a_td));
984263bc
MD
349}
350
50775720
CP
351/*
352 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
353 * struct thread *a_td)
354 */
984263bc
MD
355/* ARGSUSED */
356static int
50775720 357spec_poll(struct vop_poll_args *ap)
984263bc
MD
358{
359 dev_t dev;
360
361 dev = ap->a_vp->v_rdev;
335dda38 362 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
363}
364
50775720
CP
365/*
366 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
367 */
984263bc
MD
368/* ARGSUSED */
369static int
50775720 370spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
371{
372 dev_t dev;
373
374 dev = ap->a_vp->v_rdev;
335dda38 375 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
376}
377
378/*
379 * Synch buffers associated with a block device
50775720
CP
380 *
381 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
382 * int a_waitfor, struct thread *a_td)
984263bc
MD
383 */
384/* ARGSUSED */
385static int
50775720 386spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
387{
388 struct vnode *vp = ap->a_vp;
389 struct buf *bp;
390 struct buf *nbp;
391 int s;
392 int maxretry = 10000; /* large, arbitrarily chosen */
393
394 if (!vn_isdisk(vp, NULL))
395 return (0);
396
397loop1:
398 /*
399 * MARK/SCAN initialization to avoid infinite loops
400 */
401 s = splbio();
402 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
403 bp = TAILQ_NEXT(bp, b_vnbufs)) {
404 bp->b_flags &= ~B_SCANNED;
405 }
406 splx(s);
407
408 /*
409 * Flush all dirty buffers associated with a block device.
410 */
411loop2:
412 s = splbio();
413 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
414 nbp = TAILQ_NEXT(bp, b_vnbufs);
415 if ((bp->b_flags & B_SCANNED) != 0)
416 continue;
417 bp->b_flags |= B_SCANNED;
418 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
419 continue;
420 if ((bp->b_flags & B_DELWRI) == 0)
421 panic("spec_fsync: not dirty");
422 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
423 BUF_UNLOCK(bp);
424 vfs_bio_awrite(bp);
425 splx(s);
426 } else {
427 bremfree(bp);
428 splx(s);
429 bawrite(bp);
430 }
431 goto loop2;
432 }
433
434 /*
435 * If synchronous the caller expects us to completely resolve all
436 * dirty buffers in the system. Wait for in-progress I/O to
437 * complete (which could include background bitmap writes), then
438 * retry if dirty blocks still exist.
439 */
440 if (ap->a_waitfor == MNT_WAIT) {
441 while (vp->v_numoutput) {
442 vp->v_flag |= VBWAIT;
377d4740 443 (void) tsleep((caddr_t)&vp->v_numoutput, 0, "spfsyn", 0);
984263bc
MD
444 }
445 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
446 if (--maxretry != 0) {
447 splx(s);
448 goto loop1;
449 }
450 vprint("spec_fsync: giving up on dirty", vp);
451 }
452 }
453 splx(s);
454 return (0);
455}
456
50775720
CP
457/*
458 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
459 */
984263bc 460static int
50775720 461spec_inactive(struct vop_inactive_args *ap)
984263bc 462{
41a01a4d 463 VOP_UNLOCK(ap->a_vp, NULL, 0, ap->a_td);
984263bc
MD
464 return (0);
465}
466
467/*
468 * Just call the device strategy routine
50775720
CP
469 *
470 * spec_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc
MD
471 */
472static int
50775720 473spec_strategy(struct vop_strategy_args *ap)
984263bc
MD
474{
475 struct buf *bp;
476 struct vnode *vp;
477 struct mount *mp;
478
479 bp = ap->a_bp;
480 if (((bp->b_flags & B_READ) == 0) &&
481 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
482 (*bioops.io_start)(bp);
483
484 /*
485 * Collect statistics on synchronous and asynchronous read
486 * and write counts for disks that have associated filesystems.
487 */
488 vp = ap->a_vp;
e4c9c0c8 489 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 490 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 491 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
492 mp->mnt_stat.f_asyncwrites++;
493 else
494 mp->mnt_stat.f_syncwrites++;
495 } else {
dadab5e9 496 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
497 mp->mnt_stat.f_asyncreads++;
498 else
499 mp->mnt_stat.f_syncreads++;
500 }
501 }
e4c9c0c8 502 bp->b_dev = vp->v_rdev;
984263bc
MD
503 BUF_STRATEGY(bp, 0);
504 return (0);
505}
506
50775720
CP
507/*
508 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
509 */
984263bc 510static int
50775720 511spec_freeblks(struct vop_freeblks_args *ap)
984263bc 512{
984263bc
MD
513 struct buf *bp;
514
515 /*
516 * XXX: This assumes that strategy does the deed right away.
517 * XXX: this may not be TRTTD.
518 */
335dda38 519 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
520 return (0);
521 bp = geteblk(ap->a_length);
522 bp->b_flags |= B_FREEBUF;
523 bp->b_dev = ap->a_vp->v_rdev;
524 bp->b_blkno = ap->a_addr;
525 bp->b_offset = dbtob(ap->a_addr);
526 bp->b_bcount = ap->a_length;
527 BUF_STRATEGY(bp, 0);
528 return (0);
529}
530
531/*
532 * Implement degenerate case where the block requested is the block
533 * returned, and assume that the entire device is contiguous in regards
534 * to the contiguous block range (runp and runb).
50775720
CP
535 *
536 * spec_bmap(struct vnode *a_vp, daddr_t a_bn, struct vnode **a_vpp,
537 * daddr_t *a_bnp, int *a_runp, int *a_runb)
984263bc
MD
538 */
539static int
50775720 540spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
541{
542 struct vnode *vp = ap->a_vp;
543 int runp = 0;
544 int runb = 0;
545
546 if (ap->a_vpp != NULL)
547 *ap->a_vpp = vp;
548 if (ap->a_bnp != NULL)
549 *ap->a_bnp = ap->a_bn;
550 if (vp->v_mount != NULL)
551 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
552 if (ap->a_runp != NULL)
553 *ap->a_runp = runp;
554 if (ap->a_runb != NULL)
555 *ap->a_runb = runb;
556 return (0);
557}
558
559/*
560 * Device close routine
50775720
CP
561 *
562 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
563 * struct thread *a_td)
984263bc
MD
564 */
565/* ARGSUSED */
566static int
50775720 567spec_close(struct vop_close_args *ap)
984263bc 568{
dadab5e9 569 struct proc *p = ap->a_td->td_proc;
984263bc 570 struct vnode *vp = ap->a_vp;
984263bc 571 dev_t dev = vp->v_rdev;
e4c9c0c8 572 int error;
984263bc
MD
573
574 /*
575 * Hack: a tty device that is a controlling terminal
576 * has a reference from the session structure.
577 * We cannot easily tell that a character device is
578 * a controlling terminal, unless it is the closing
579 * process' controlling terminal. In that case,
580 * if the reference count is 2 (this last descriptor
581 * plus the session), release the reference from the session.
582 */
e4c9c0c8
MD
583 reference_dev(dev);
584 if (vcount(vp) == 2 && vp->v_opencount == 1 &&
585 p && (vp->v_flag & VXLOCK) == 0 && vp == p->p_session->s_ttyvp) {
984263bc
MD
586 vrele(vp);
587 p->p_session->s_ttyvp = NULL;
588 }
e4c9c0c8 589
984263bc 590 /*
e4c9c0c8
MD
591 * Vnodes can be opened and close multiple times. Do not really
592 * close the device unless (1) it is being closed forcibly,
593 * (2) the device wants to track closes, or (3) this is the last
594 * vnode doing its last close on the device.
595 *
596 * XXX the VXLOCK (force close) case can leave vnodes referencing
597 * a closed device.
984263bc 598 */
e4c9c0c8
MD
599 if ((vp->v_flag & VXLOCK) ||
600 (dev_dflags(dev) & D_TRACKCLOSE) ||
601 (vcount(vp) <= 1 && vp->v_opencount == 1)) {
602 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
603 } else {
604 error = 0;
984263bc 605 }
e4c9c0c8
MD
606
607 /*
608 * Track the actual opens and closes on the vnode. The last close
609 * disassociates the rdev.
610 */
611 KKASSERT(vp->v_opencount > 0);
612 if (dev_ref_debug)
613 printf("spec_close: %s %d\n", dev->si_name, vp->v_opencount - 1);
614 if (--vp->v_opencount == 0)
615 v_release_rdev(vp);
616 release_dev(dev);
617 return(error);
984263bc
MD
618}
619
620/*
621 * Print out the contents of a special device vnode.
50775720
CP
622 *
623 * spec_print(struct vnode *a_vp)
984263bc
MD
624 */
625static int
50775720 626spec_print(struct vop_print_args *ap)
984263bc 627{
984263bc
MD
628 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
629 return (0);
630}
631
632/*
633 * Special device advisory byte-level locks.
50775720
CP
634 *
635 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
636 * struct flock *a_fl, int a_flags)
984263bc
MD
637 */
638/* ARGSUSED */
639static int
50775720 640spec_advlock(struct vop_advlock_args *ap)
984263bc 641{
984263bc
MD
642 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
643}
644
645static void
50775720 646spec_getpages_iodone(struct buf *bp)
984263bc 647{
984263bc
MD
648 bp->b_flags |= B_DONE;
649 wakeup(bp);
650}
651
652static int
50775720 653spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
654{
655 vm_offset_t kva;
656 int error;
657 int i, pcount, size, s;
658 daddr_t blkno;
659 struct buf *bp;
660 vm_page_t m;
661 vm_ooffset_t offset;
662 int toff, nextoff, nread;
663 struct vnode *vp = ap->a_vp;
664 int blksiz;
665 int gotreqpage;
666
667 error = 0;
668 pcount = round_page(ap->a_count) / PAGE_SIZE;
669
670 /*
671 * Calculate the offset of the transfer and do sanity check.
672 * FreeBSD currently only supports an 8 TB range due to b_blkno
673 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
674 * VOP_STRATEGY. XXX
675 */
676 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
677
678#define DADDR_T_BIT (sizeof(daddr_t)*8)
679#define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
680
681 if (offset < 0 || offset > OFFSET_MAX) {
682 /* XXX still no %q in kernel. */
683 printf("spec_getpages: preposterous offset 0x%x%08x\n",
684 (u_int)((u_quad_t)offset >> 32),
685 (u_int)(offset & 0xffffffff));
686 return (VM_PAGER_ERROR);
687 }
688
689 blkno = btodb(offset);
690
691 /*
692 * Round up physical size for real devices. We cannot round using
693 * v_mount's block size data because v_mount has nothing to do with
694 * the device. i.e. it's usually '/dev'. We need the physical block
695 * size for the device itself.
696 *
e4c9c0c8 697 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
698 * block device is mounted. However, we can use v_rdev.
699 */
700
701 if (vn_isdisk(vp, NULL))
702 blksiz = vp->v_rdev->si_bsize_phys;
703 else
704 blksiz = DEV_BSIZE;
705
706 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
707
708 bp = getpbuf(NULL);
709 kva = (vm_offset_t)bp->b_data;
710
711 /*
712 * Map the pages to be read into the kva.
713 */
714 pmap_qenter(kva, ap->a_m, pcount);
715
716 /* Build a minimal buffer header. */
717 bp->b_flags = B_READ | B_CALL;
718 bp->b_iodone = spec_getpages_iodone;
719
720 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
721 bp->b_blkno = blkno;
722 bp->b_lblkno = blkno;
723 pbgetvp(ap->a_vp, bp);
724 bp->b_bcount = size;
725 bp->b_bufsize = size;
726 bp->b_resid = 0;
727 bp->b_runningbufspace = bp->b_bufsize;
728 runningbufspace += bp->b_runningbufspace;
729
12e4aaff
MD
730 mycpu->gd_cnt.v_vnodein++;
731 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
732
733 /* Do the input. */
734 VOP_STRATEGY(bp->b_vp, bp);
735
736 s = splbio();
737
738 /* We definitely need to be at splbio here. */
26a0694b 739 while ((bp->b_flags & B_DONE) == 0) {
377d4740 740 tsleep(bp, 0, "spread", 0);
26a0694b 741 }
984263bc
MD
742
743 splx(s);
744
745 if ((bp->b_flags & B_ERROR) != 0) {
746 if (bp->b_error)
747 error = bp->b_error;
748 else
749 error = EIO;
750 }
751
752 nread = size - bp->b_resid;
753
754 if (nread < ap->a_count) {
755 bzero((caddr_t)kva + nread,
756 ap->a_count - nread);
757 }
758 pmap_qremove(kva, pcount);
759
760
761 gotreqpage = 0;
762 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
763 nextoff = toff + PAGE_SIZE;
764 m = ap->a_m[i];
765
766 m->flags &= ~PG_ZERO;
767
768 if (nextoff <= nread) {
769 m->valid = VM_PAGE_BITS_ALL;
770 vm_page_undirty(m);
771 } else if (toff < nread) {
772 /*
773 * Since this is a VM request, we have to supply the
774 * unaligned offset to allow vm_page_set_validclean()
775 * to zero sub-DEV_BSIZE'd portions of the page.
776 */
777 vm_page_set_validclean(m, 0, nread - toff);
778 } else {
779 m->valid = 0;
780 vm_page_undirty(m);
781 }
782
783 if (i != ap->a_reqpage) {
784 /*
785 * Just in case someone was asking for this page we
786 * now tell them that it is ok to use.
787 */
788 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
789 if (m->valid) {
790 if (m->flags & PG_WANTED) {
791 vm_page_activate(m);
792 } else {
793 vm_page_deactivate(m);
794 }
795 vm_page_wakeup(m);
796 } else {
797 vm_page_free(m);
798 }
799 } else {
800 vm_page_free(m);
801 }
802 } else if (m->valid) {
803 gotreqpage = 1;
804 /*
805 * Since this is a VM request, we need to make the
806 * entire page presentable by zeroing invalid sections.
807 */
808 if (m->valid != VM_PAGE_BITS_ALL)
809 vm_page_zero_invalid(m, FALSE);
810 }
811 }
812 if (!gotreqpage) {
813 m = ap->a_m[ap->a_reqpage];
814 printf(
815 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
816 devtoname(bp->b_dev), error, bp, bp->b_vp);
817 printf(
818 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
819 size, bp->b_resid, ap->a_count, m->valid);
820 printf(
821 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
822 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
823 /*
824 * Free the buffer header back to the swap buffer pool.
825 */
826 relpbuf(bp, NULL);
827 return VM_PAGER_ERROR;
828 }
829 /*
830 * Free the buffer header back to the swap buffer pool.
831 */
832 relpbuf(bp, NULL);
833 return VM_PAGER_OK;
834}