ALTQ support.
[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 $
fad57d0e 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.22 2004/11/12 00:09:50 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 74
2d3e977e 75struct vop_ops *spec_vnode_vops;
0961aa92 76struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
2d3e977e
MD
77 { &vop_default_desc, vop_defaultop },
78 { &vop_access_desc, vop_ebadf },
79 { &vop_advlock_desc, (void *) spec_advlock },
80 { &vop_bmap_desc, (void *) spec_bmap },
81 { &vop_close_desc, (void *) spec_close },
82 { &vop_create_desc, vop_panic },
83 { &vop_freeblks_desc, (void *) spec_freeblks },
84 { &vop_fsync_desc, (void *) spec_fsync },
85 { &vop_getpages_desc, (void *) spec_getpages },
86 { &vop_inactive_desc, (void *) spec_inactive },
87 { &vop_ioctl_desc, (void *) spec_ioctl },
88 { &vop_lease_desc, vop_null },
89 { &vop_link_desc, vop_panic },
90 { &vop_mkdir_desc, vop_panic },
91 { &vop_mknod_desc, vop_panic },
92 { &vop_open_desc, (void *) spec_open },
93 { &vop_pathconf_desc, (void *) vop_stdpathconf },
94 { &vop_poll_desc, (void *) spec_poll },
95 { &vop_kqfilter_desc, (void *) spec_kqfilter },
96 { &vop_print_desc, (void *) spec_print },
97 { &vop_read_desc, (void *) spec_read },
98 { &vop_readdir_desc, vop_panic },
99 { &vop_readlink_desc, vop_panic },
100 { &vop_reallocblks_desc, vop_panic },
101 { &vop_reclaim_desc, vop_null },
102 { &vop_remove_desc, vop_panic },
103 { &vop_rename_desc, vop_panic },
104 { &vop_rmdir_desc, vop_panic },
105 { &vop_setattr_desc, vop_ebadf },
106 { &vop_strategy_desc, (void *) spec_strategy },
107 { &vop_symlink_desc, vop_panic },
108 { &vop_write_desc, (void *) spec_write },
984263bc
MD
109 { NULL, NULL }
110};
111static struct vnodeopv_desc spec_vnodeop_opv_desc =
2d3e977e 112 { &spec_vnode_vops, spec_vnodeop_entries };
984263bc
MD
113VNODEOP_SET(spec_vnodeop_opv_desc);
114
e4c9c0c8
MD
115extern int dev_ref_debug;
116
50775720
CP
117/*
118 * spec_vnoperate(struct vnodeop_desc *a_desc, ...)
119 */
984263bc 120int
50775720 121spec_vnoperate(struct vop_generic_args *ap)
984263bc 122{
0961aa92 123 return (VOCALL(spec_vnode_vops, ap));
984263bc
MD
124}
125
a6ee311a 126static void spec_getpages_iodone (struct buf *bp);
984263bc
MD
127
128/*
129 * Open a special file.
50775720
CP
130 *
131 * spec_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
132 * struct thread *a_td)
984263bc
MD
133 */
134/* ARGSUSED */
135static int
50775720 136spec_open(struct vop_open_args *ap)
984263bc 137{
984263bc 138 struct vnode *vp = ap->a_vp;
e4c9c0c8 139 dev_t dev;
984263bc 140 int error;
e4c9c0c8 141 int isblk = (vp->v_type == VBLK) ? 1 : 0;
984263bc
MD
142 const char *cp;
143
144 /*
145 * Don't allow open if fs is mounted -nodev.
146 */
147 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
148 return (ENXIO);
149
e4c9c0c8
MD
150 /*
151 * Resolve the device. If the vnode is already open v_rdev may
152 * already be resolved. However, if the device changes out from
1fbb5fc0
MD
153 * under us we report it (and, for now, we allow it). Since
154 * v_release_rdev() zero's v_opencount, we have to save and restore
155 * it when replacing the rdev reference.
e4c9c0c8
MD
156 */
157 if (vp->v_rdev != NULL) {
158 dev = udev2dev(vp->v_udev, isblk);
159 if (dev != vp->v_rdev) {
1fbb5fc0 160 int oc = vp->v_opencount;
e4c9c0c8
MD
161 printf(
162 "Warning: spec_open: dev %s was lost",
163 vp->v_rdev->si_name);
164 v_release_rdev(vp);
165 error = v_associate_rdev(vp,
166 udev2dev(vp->v_udev, isblk));
1fbb5fc0 167 if (error) {
e4c9c0c8 168 printf(", reacquisition failed\n");
1fbb5fc0
MD
169 } else {
170 vp->v_opencount = oc;
e4c9c0c8 171 printf(", reacquisition successful\n");
1fbb5fc0 172 }
e4c9c0c8
MD
173 } else {
174 error = 0;
175 }
176 } else {
177 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
178 }
179 if (error)
180 return(error);
181
500df1da
MD
182 /*
183 * Prevent degenerate open/close sequences from nulling out rdev.
184 */
185 ++vp->v_opencount;
e4c9c0c8 186 dev = vp->v_rdev;
500df1da 187 KKASSERT(dev != NULL);
984263bc 188
e4c9c0c8
MD
189 /*
190 * Make this field valid before any I/O in ->d_open. XXX the
191 * device itself should probably be required to initialize
192 * this field in d_open.
193 */
984263bc
MD
194 if (!dev->si_iosize_max)
195 dev->si_iosize_max = DFLTPHYS;
196
197 /*
198 * XXX: Disks get special billing here, but it is mostly wrong.
199 * XXX: diskpartitions can overlap and the real checks should
200 * XXX: take this into account, and consequently they need to
201 * XXX: live in the diskslicing code. Some checks do.
202 */
203 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
204 (ap->a_mode & FWRITE)) {
205 /*
206 * Never allow opens for write if the device is mounted R/W
207 */
e4c9c0c8
MD
208 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
209 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
210 error = EBUSY;
211 goto done;
212 }
984263bc
MD
213
214 /*
215 * When running in secure mode, do not allow opens
216 * for writing if the device is mounted
217 */
e4c9c0c8
MD
218 if (securelevel >= 1 && vfs_mountedon(vp)) {
219 error = EPERM;
220 goto done;
221 }
984263bc
MD
222
223 /*
224 * When running in very secure mode, do not allow
225 * opens for writing of any devices.
226 */
e4c9c0c8
MD
227 if (securelevel >= 2) {
228 error = EPERM;
229 goto done;
230 }
984263bc
MD
231 }
232
233 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 234 if (dev_dflags(dev) & D_TTY)
984263bc
MD
235 vp->v_flag |= VISTTY;
236
e4c9c0c8
MD
237 /*
238 * dev_dopen() is always called for each open. dev_dclose() is
239 * only called for the last close unless D_TRACKCLOSE is set.
240 */
5fd012e0 241 VOP_UNLOCK(vp, 0, ap->a_td);
335dda38 242 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
5fd012e0 243 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
984263bc
MD
244
245 if (error)
e4c9c0c8 246 goto done;
984263bc 247
335dda38 248 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
249 if (dev->si_tty) {
250 struct tty *tp;
251 tp = dev->si_tty;
252 if (!tp->t_stop) {
253 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
254 tp->t_stop = nottystop;
255 }
256 }
257 }
258
259 if (vn_isdisk(vp, NULL)) {
260 if (!dev->si_bsize_phys)
261 dev->si_bsize_phys = DEV_BSIZE;
262 }
335dda38 263 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 264 cp = devtoname(dev);
335dda38 265 if (*cp == '#') {
984263bc 266 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 267 dev_dname(dev), cp);
984263bc
MD
268 }
269 }
fad57d0e
MD
270
271 /*
272 * If we were handed a file pointer we may be able to install a
273 * shortcut which issues device read and write operations directly
274 * from the fileops rather then having to go through spec_read()
275 * and spec_write().
276 */
277 if (ap->a_fp)
278 vn_setspecops(ap->a_fp);
279
e4c9c0c8
MD
280 if (dev_ref_debug)
281 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
282done:
283 if (error) {
500df1da
MD
284 KKASSERT(vp->v_opencount > 0);
285 if (--vp->v_opencount == 0)
e4c9c0c8
MD
286 v_release_rdev(vp);
287 }
984263bc
MD
288 return (error);
289}
290
291/*
292 * Vnode op for read
50775720
CP
293 *
294 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
295 * struct ucred *a_cred)
984263bc
MD
296 */
297/* ARGSUSED */
298static int
50775720 299spec_read(struct vop_read_args *ap)
984263bc
MD
300{
301 struct vnode *vp;
dadab5e9 302 struct thread *td;
984263bc
MD
303 struct uio *uio;
304 dev_t dev;
305 int error;
306
307 vp = ap->a_vp;
308 dev = vp->v_rdev;
309 uio = ap->a_uio;
dadab5e9 310 td = uio->uio_td;
984263bc 311
72402f22
MD
312 if (dev == NULL) /* device was revoked */
313 return (EBADF);
984263bc
MD
314 if (uio->uio_resid == 0)
315 return (0);
316
5fd012e0 317 VOP_UNLOCK(vp, 0, td);
335dda38 318 error = dev_dread(dev, uio, ap->a_ioflag);
5fd012e0 319 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
320 return (error);
321}
322
323/*
324 * Vnode op for write
50775720
CP
325 *
326 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
327 * struct ucred *a_cred)
984263bc
MD
328 */
329/* ARGSUSED */
330static int
50775720 331spec_write(struct vop_write_args *ap)
984263bc
MD
332{
333 struct vnode *vp;
dadab5e9 334 struct thread *td;
984263bc
MD
335 struct uio *uio;
336 dev_t dev;
337 int error;
338
339 vp = ap->a_vp;
340 dev = vp->v_rdev;
341 uio = ap->a_uio;
dadab5e9 342 td = uio->uio_td;
984263bc 343
72402f22
MD
344 if (dev == NULL) /* device was revoked */
345 return (EBADF);
346
5fd012e0 347 VOP_UNLOCK(vp, 0, td);
335dda38 348 error = dev_dwrite(dev, uio, ap->a_ioflag);
5fd012e0 349 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
350 return (error);
351}
352
353/*
354 * Device ioctl operation.
50775720
CP
355 *
356 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
357 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
358 */
359/* ARGSUSED */
360static int
50775720 361spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
362{
363 dev_t dev;
364
72402f22
MD
365 if ((dev = ap->a_vp->v_rdev) == NULL)
366 return (EBADF); /* device was revoked */
367
335dda38
MD
368 return (dev_dioctl(dev, ap->a_command, ap->a_data,
369 ap->a_fflag, ap->a_td));
984263bc
MD
370}
371
50775720
CP
372/*
373 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
374 * struct thread *a_td)
375 */
984263bc
MD
376/* ARGSUSED */
377static int
50775720 378spec_poll(struct vop_poll_args *ap)
984263bc
MD
379{
380 dev_t dev;
381
72402f22
MD
382 if ((dev = ap->a_vp->v_rdev) == NULL)
383 return (EBADF); /* device was revoked */
335dda38 384 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
385}
386
50775720
CP
387/*
388 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
389 */
984263bc
MD
390/* ARGSUSED */
391static int
50775720 392spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
393{
394 dev_t dev;
395
72402f22
MD
396 if ((dev = ap->a_vp->v_rdev) == NULL)
397 return (EBADF); /* device was revoked */
335dda38 398 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
399}
400
401/*
402 * Synch buffers associated with a block device
50775720
CP
403 *
404 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
405 * int a_waitfor, struct thread *a_td)
984263bc
MD
406 */
407/* ARGSUSED */
408static int
50775720 409spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
410{
411 struct vnode *vp = ap->a_vp;
412 struct buf *bp;
413 struct buf *nbp;
414 int s;
415 int maxretry = 10000; /* large, arbitrarily chosen */
416
417 if (!vn_isdisk(vp, NULL))
418 return (0);
419
420loop1:
421 /*
422 * MARK/SCAN initialization to avoid infinite loops
423 */
424 s = splbio();
425 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
426 bp = TAILQ_NEXT(bp, b_vnbufs)) {
427 bp->b_flags &= ~B_SCANNED;
428 }
429 splx(s);
430
431 /*
432 * Flush all dirty buffers associated with a block device.
433 */
434loop2:
435 s = splbio();
436 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
437 nbp = TAILQ_NEXT(bp, b_vnbufs);
438 if ((bp->b_flags & B_SCANNED) != 0)
439 continue;
440 bp->b_flags |= B_SCANNED;
441 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
442 continue;
443 if ((bp->b_flags & B_DELWRI) == 0)
444 panic("spec_fsync: not dirty");
445 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
446 BUF_UNLOCK(bp);
447 vfs_bio_awrite(bp);
448 splx(s);
449 } else {
450 bremfree(bp);
451 splx(s);
452 bawrite(bp);
453 }
454 goto loop2;
455 }
456
457 /*
458 * If synchronous the caller expects us to completely resolve all
459 * dirty buffers in the system. Wait for in-progress I/O to
460 * complete (which could include background bitmap writes), then
461 * retry if dirty blocks still exist.
462 */
463 if (ap->a_waitfor == MNT_WAIT) {
464 while (vp->v_numoutput) {
465 vp->v_flag |= VBWAIT;
377d4740 466 (void) tsleep((caddr_t)&vp->v_numoutput, 0, "spfsyn", 0);
984263bc
MD
467 }
468 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
469 if (--maxretry != 0) {
470 splx(s);
471 goto loop1;
472 }
473 vprint("spec_fsync: giving up on dirty", vp);
474 }
475 }
476 splx(s);
477 return (0);
478}
479
50775720
CP
480/*
481 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
482 */
984263bc 483static int
50775720 484spec_inactive(struct vop_inactive_args *ap)
984263bc 485{
984263bc
MD
486 return (0);
487}
488
489/*
490 * Just call the device strategy routine
50775720
CP
491 *
492 * spec_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc
MD
493 */
494static int
50775720 495spec_strategy(struct vop_strategy_args *ap)
984263bc
MD
496{
497 struct buf *bp;
498 struct vnode *vp;
499 struct mount *mp;
500
501 bp = ap->a_bp;
502 if (((bp->b_flags & B_READ) == 0) &&
503 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
504 (*bioops.io_start)(bp);
505
506 /*
507 * Collect statistics on synchronous and asynchronous read
508 * and write counts for disks that have associated filesystems.
509 */
510 vp = ap->a_vp;
72402f22 511 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 512 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 513 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 514 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
515 mp->mnt_stat.f_asyncwrites++;
516 else
517 mp->mnt_stat.f_syncwrites++;
518 } else {
dadab5e9 519 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
520 mp->mnt_stat.f_asyncreads++;
521 else
522 mp->mnt_stat.f_syncreads++;
523 }
524 }
e4c9c0c8 525 bp->b_dev = vp->v_rdev;
984263bc
MD
526 BUF_STRATEGY(bp, 0);
527 return (0);
528}
529
50775720
CP
530/*
531 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
532 */
984263bc 533static int
50775720 534spec_freeblks(struct vop_freeblks_args *ap)
984263bc 535{
984263bc
MD
536 struct buf *bp;
537
538 /*
539 * XXX: This assumes that strategy does the deed right away.
540 * XXX: this may not be TRTTD.
541 */
72402f22 542 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 543 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
544 return (0);
545 bp = geteblk(ap->a_length);
546 bp->b_flags |= B_FREEBUF;
547 bp->b_dev = ap->a_vp->v_rdev;
548 bp->b_blkno = ap->a_addr;
549 bp->b_offset = dbtob(ap->a_addr);
550 bp->b_bcount = ap->a_length;
551 BUF_STRATEGY(bp, 0);
552 return (0);
553}
554
555/*
556 * Implement degenerate case where the block requested is the block
557 * returned, and assume that the entire device is contiguous in regards
558 * to the contiguous block range (runp and runb).
50775720
CP
559 *
560 * spec_bmap(struct vnode *a_vp, daddr_t a_bn, struct vnode **a_vpp,
561 * daddr_t *a_bnp, int *a_runp, int *a_runb)
984263bc
MD
562 */
563static int
50775720 564spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
565{
566 struct vnode *vp = ap->a_vp;
567 int runp = 0;
568 int runb = 0;
569
570 if (ap->a_vpp != NULL)
571 *ap->a_vpp = vp;
572 if (ap->a_bnp != NULL)
573 *ap->a_bnp = ap->a_bn;
574 if (vp->v_mount != NULL)
575 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
576 if (ap->a_runp != NULL)
577 *ap->a_runp = runp;
578 if (ap->a_runb != NULL)
579 *ap->a_runb = runb;
580 return (0);
581}
582
583/*
584 * Device close routine
50775720
CP
585 *
586 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
587 * struct thread *a_td)
984263bc
MD
588 */
589/* ARGSUSED */
590static int
50775720 591spec_close(struct vop_close_args *ap)
984263bc 592{
dadab5e9 593 struct proc *p = ap->a_td->td_proc;
984263bc 594 struct vnode *vp = ap->a_vp;
984263bc 595 dev_t dev = vp->v_rdev;
e4c9c0c8 596 int error;
984263bc
MD
597
598 /*
599 * Hack: a tty device that is a controlling terminal
600 * has a reference from the session structure.
601 * We cannot easily tell that a character device is
602 * a controlling terminal, unless it is the closing
603 * process' controlling terminal. In that case,
604 * if the reference count is 2 (this last descriptor
605 * plus the session), release the reference from the session.
1fbb5fc0
MD
606 *
607 * It is possible for v_opencount to be 0 or 1 in this case, 0
608 * because the tty might have been revoked.
984263bc 609 */
72402f22
MD
610 if (dev)
611 reference_dev(dev);
1fbb5fc0 612 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 613 p && vp == p->p_session->s_ttyvp) {
984263bc 614 p->p_session->s_ttyvp = NULL;
1fbb5fc0 615 vrele(vp);
984263bc 616 }
e4c9c0c8 617
984263bc 618 /*
e4c9c0c8
MD
619 * Vnodes can be opened and close multiple times. Do not really
620 * close the device unless (1) it is being closed forcibly,
621 * (2) the device wants to track closes, or (3) this is the last
622 * vnode doing its last close on the device.
623 *
624 * XXX the VXLOCK (force close) case can leave vnodes referencing
625 * a closed device.
984263bc 626 */
5fd012e0 627 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 628 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 629 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
e4c9c0c8
MD
630 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
631 } else {
632 error = 0;
984263bc 633 }
e4c9c0c8
MD
634
635 /*
636 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
637 * disassociates the rdev. If the rdev is already disassociated
638 * the vnode might have been revoked and no further opencount
639 * tracking occurs.
e4c9c0c8 640 */
1fbb5fc0
MD
641 if (dev) {
642 KKASSERT(vp->v_opencount > 0);
643 if (dev_ref_debug) {
72402f22
MD
644 printf("spec_close: %s %d\n",
645 dev->si_name, vp->v_opencount - 1);
72402f22 646 }
1fbb5fc0
MD
647 if (--vp->v_opencount == 0)
648 v_release_rdev(vp);
72402f22 649 release_dev(dev);
1fbb5fc0 650 }
e4c9c0c8 651 return(error);
984263bc
MD
652}
653
654/*
655 * Print out the contents of a special device vnode.
50775720
CP
656 *
657 * spec_print(struct vnode *a_vp)
984263bc
MD
658 */
659static int
50775720 660spec_print(struct vop_print_args *ap)
984263bc 661{
984263bc
MD
662 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
663 return (0);
664}
665
666/*
667 * Special device advisory byte-level locks.
50775720
CP
668 *
669 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
670 * struct flock *a_fl, int a_flags)
984263bc
MD
671 */
672/* ARGSUSED */
673static int
50775720 674spec_advlock(struct vop_advlock_args *ap)
984263bc 675{
984263bc
MD
676 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
677}
678
679static void
50775720 680spec_getpages_iodone(struct buf *bp)
984263bc 681{
984263bc
MD
682 bp->b_flags |= B_DONE;
683 wakeup(bp);
684}
685
686static int
50775720 687spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
688{
689 vm_offset_t kva;
690 int error;
691 int i, pcount, size, s;
692 daddr_t blkno;
693 struct buf *bp;
694 vm_page_t m;
695 vm_ooffset_t offset;
696 int toff, nextoff, nread;
697 struct vnode *vp = ap->a_vp;
698 int blksiz;
699 int gotreqpage;
700
701 error = 0;
702 pcount = round_page(ap->a_count) / PAGE_SIZE;
703
704 /*
705 * Calculate the offset of the transfer and do sanity check.
706 * FreeBSD currently only supports an 8 TB range due to b_blkno
707 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
708 * VOP_STRATEGY. XXX
709 */
710 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
711
712#define DADDR_T_BIT (sizeof(daddr_t)*8)
713#define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
714
715 if (offset < 0 || offset > OFFSET_MAX) {
716 /* XXX still no %q in kernel. */
717 printf("spec_getpages: preposterous offset 0x%x%08x\n",
718 (u_int)((u_quad_t)offset >> 32),
719 (u_int)(offset & 0xffffffff));
720 return (VM_PAGER_ERROR);
721 }
722
723 blkno = btodb(offset);
724
725 /*
726 * Round up physical size for real devices. We cannot round using
727 * v_mount's block size data because v_mount has nothing to do with
728 * the device. i.e. it's usually '/dev'. We need the physical block
729 * size for the device itself.
730 *
e4c9c0c8 731 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
732 * block device is mounted. However, we can use v_rdev.
733 */
734
735 if (vn_isdisk(vp, NULL))
736 blksiz = vp->v_rdev->si_bsize_phys;
737 else
738 blksiz = DEV_BSIZE;
739
740 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
741
742 bp = getpbuf(NULL);
743 kva = (vm_offset_t)bp->b_data;
744
745 /*
746 * Map the pages to be read into the kva.
747 */
748 pmap_qenter(kva, ap->a_m, pcount);
749
750 /* Build a minimal buffer header. */
751 bp->b_flags = B_READ | B_CALL;
752 bp->b_iodone = spec_getpages_iodone;
753
754 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
755 bp->b_blkno = blkno;
756 bp->b_lblkno = blkno;
757 pbgetvp(ap->a_vp, bp);
758 bp->b_bcount = size;
759 bp->b_bufsize = size;
760 bp->b_resid = 0;
761 bp->b_runningbufspace = bp->b_bufsize;
762 runningbufspace += bp->b_runningbufspace;
763
12e4aaff
MD
764 mycpu->gd_cnt.v_vnodein++;
765 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
766
767 /* Do the input. */
768 VOP_STRATEGY(bp->b_vp, bp);
769
770 s = splbio();
771
772 /* We definitely need to be at splbio here. */
26a0694b 773 while ((bp->b_flags & B_DONE) == 0) {
377d4740 774 tsleep(bp, 0, "spread", 0);
26a0694b 775 }
984263bc
MD
776
777 splx(s);
778
779 if ((bp->b_flags & B_ERROR) != 0) {
780 if (bp->b_error)
781 error = bp->b_error;
782 else
783 error = EIO;
784 }
785
786 nread = size - bp->b_resid;
787
788 if (nread < ap->a_count) {
789 bzero((caddr_t)kva + nread,
790 ap->a_count - nread);
791 }
792 pmap_qremove(kva, pcount);
793
794
795 gotreqpage = 0;
796 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
797 nextoff = toff + PAGE_SIZE;
798 m = ap->a_m[i];
799
800 m->flags &= ~PG_ZERO;
801
802 if (nextoff <= nread) {
803 m->valid = VM_PAGE_BITS_ALL;
804 vm_page_undirty(m);
805 } else if (toff < nread) {
806 /*
807 * Since this is a VM request, we have to supply the
808 * unaligned offset to allow vm_page_set_validclean()
809 * to zero sub-DEV_BSIZE'd portions of the page.
810 */
811 vm_page_set_validclean(m, 0, nread - toff);
812 } else {
813 m->valid = 0;
814 vm_page_undirty(m);
815 }
816
817 if (i != ap->a_reqpage) {
818 /*
819 * Just in case someone was asking for this page we
820 * now tell them that it is ok to use.
821 */
822 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
823 if (m->valid) {
824 if (m->flags & PG_WANTED) {
825 vm_page_activate(m);
826 } else {
827 vm_page_deactivate(m);
828 }
829 vm_page_wakeup(m);
830 } else {
831 vm_page_free(m);
832 }
833 } else {
834 vm_page_free(m);
835 }
836 } else if (m->valid) {
837 gotreqpage = 1;
838 /*
839 * Since this is a VM request, we need to make the
840 * entire page presentable by zeroing invalid sections.
841 */
842 if (m->valid != VM_PAGE_BITS_ALL)
843 vm_page_zero_invalid(m, FALSE);
844 }
845 }
846 if (!gotreqpage) {
847 m = ap->a_m[ap->a_reqpage];
848 printf(
849 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
850 devtoname(bp->b_dev), error, bp, bp->b_vp);
851 printf(
852 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
853 size, bp->b_resid, ap->a_count, m->valid);
854 printf(
855 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
856 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
857 /*
858 * Free the buffer header back to the swap buffer pool.
859 */
860 relpbuf(bp, NULL);
861 return VM_PAGER_ERROR;
862 }
863 /*
864 * Free the buffer header back to the swap buffer pool.
865 */
866 relpbuf(bp, NULL);
867 return VM_PAGER_OK;
868}