Device layer rollup commit.
[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 $
e4c9c0c8 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.15 2004/05/19 22:53:06 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
178 dev = vp->v_rdev;
984263bc 179
e4c9c0c8
MD
180 /*
181 * Make this field valid before any I/O in ->d_open. XXX the
182 * device itself should probably be required to initialize
183 * this field in d_open.
184 */
984263bc
MD
185 if (!dev->si_iosize_max)
186 dev->si_iosize_max = DFLTPHYS;
187
188 /*
189 * XXX: Disks get special billing here, but it is mostly wrong.
190 * XXX: diskpartitions can overlap and the real checks should
191 * XXX: take this into account, and consequently they need to
192 * XXX: live in the diskslicing code. Some checks do.
193 */
194 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
195 (ap->a_mode & FWRITE)) {
196 /*
197 * Never allow opens for write if the device is mounted R/W
198 */
e4c9c0c8
MD
199 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
200 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
201 error = EBUSY;
202 goto done;
203 }
984263bc
MD
204
205 /*
206 * When running in secure mode, do not allow opens
207 * for writing if the device is mounted
208 */
e4c9c0c8
MD
209 if (securelevel >= 1 && vfs_mountedon(vp)) {
210 error = EPERM;
211 goto done;
212 }
984263bc
MD
213
214 /*
215 * When running in very secure mode, do not allow
216 * opens for writing of any devices.
217 */
e4c9c0c8
MD
218 if (securelevel >= 2) {
219 error = EPERM;
220 goto done;
221 }
984263bc
MD
222 }
223
224 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 225 if (dev_dflags(dev) & D_TTY)
984263bc
MD
226 vp->v_flag |= VISTTY;
227
e4c9c0c8
MD
228 /*
229 * dev_dopen() is always called for each open. dev_dclose() is
230 * only called for the last close unless D_TRACKCLOSE is set.
231 */
41a01a4d 232 VOP_UNLOCK(vp, NULL, 0, ap->a_td);
335dda38 233 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
41a01a4d 234 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
984263bc
MD
235
236 if (error)
e4c9c0c8 237 goto done;
984263bc 238
335dda38 239 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
240 if (dev->si_tty) {
241 struct tty *tp;
242 tp = dev->si_tty;
243 if (!tp->t_stop) {
244 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
245 tp->t_stop = nottystop;
246 }
247 }
248 }
249
250 if (vn_isdisk(vp, NULL)) {
251 if (!dev->si_bsize_phys)
252 dev->si_bsize_phys = DEV_BSIZE;
253 }
335dda38 254 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 255 cp = devtoname(dev);
335dda38 256 if (*cp == '#') {
984263bc 257 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 258 dev_dname(dev), cp);
984263bc
MD
259 }
260 }
e4c9c0c8
MD
261 ++vp->v_opencount;
262 if (dev_ref_debug)
263 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
264done:
265 if (error) {
266 if (vp->v_opencount == 0)
267 v_release_rdev(vp);
268 }
984263bc
MD
269 return (error);
270}
271
272/*
273 * Vnode op for read
50775720
CP
274 *
275 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
276 * struct ucred *a_cred)
984263bc
MD
277 */
278/* ARGSUSED */
279static int
50775720 280spec_read(struct vop_read_args *ap)
984263bc
MD
281{
282 struct vnode *vp;
dadab5e9 283 struct thread *td;
984263bc
MD
284 struct uio *uio;
285 dev_t dev;
286 int error;
287
288 vp = ap->a_vp;
289 dev = vp->v_rdev;
290 uio = ap->a_uio;
dadab5e9 291 td = uio->uio_td;
984263bc
MD
292
293 if (uio->uio_resid == 0)
294 return (0);
295
41a01a4d 296 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 297 error = dev_dread(dev, uio, ap->a_ioflag);
41a01a4d 298 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
299 return (error);
300}
301
302/*
303 * Vnode op for write
50775720
CP
304 *
305 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
306 * struct ucred *a_cred)
984263bc
MD
307 */
308/* ARGSUSED */
309static int
50775720 310spec_write(struct vop_write_args *ap)
984263bc
MD
311{
312 struct vnode *vp;
dadab5e9 313 struct thread *td;
984263bc
MD
314 struct uio *uio;
315 dev_t dev;
316 int error;
317
318 vp = ap->a_vp;
319 dev = vp->v_rdev;
320 uio = ap->a_uio;
dadab5e9 321 td = uio->uio_td;
984263bc 322
41a01a4d 323 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 324 error = dev_dwrite(dev, uio, ap->a_ioflag);
41a01a4d 325 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
326 return (error);
327}
328
329/*
330 * Device ioctl operation.
50775720
CP
331 *
332 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
333 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
334 */
335/* ARGSUSED */
336static int
50775720 337spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
338{
339 dev_t dev;
340
341 dev = ap->a_vp->v_rdev;
335dda38
MD
342 return (dev_dioctl(dev, ap->a_command, ap->a_data,
343 ap->a_fflag, ap->a_td));
984263bc
MD
344}
345
50775720
CP
346/*
347 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
348 * struct thread *a_td)
349 */
984263bc
MD
350/* ARGSUSED */
351static int
50775720 352spec_poll(struct vop_poll_args *ap)
984263bc
MD
353{
354 dev_t dev;
355
356 dev = ap->a_vp->v_rdev;
335dda38 357 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
358}
359
50775720
CP
360/*
361 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
362 */
984263bc
MD
363/* ARGSUSED */
364static int
50775720 365spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
366{
367 dev_t dev;
368
369 dev = ap->a_vp->v_rdev;
335dda38 370 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
371}
372
373/*
374 * Synch buffers associated with a block device
50775720
CP
375 *
376 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
377 * int a_waitfor, struct thread *a_td)
984263bc
MD
378 */
379/* ARGSUSED */
380static int
50775720 381spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
382{
383 struct vnode *vp = ap->a_vp;
384 struct buf *bp;
385 struct buf *nbp;
386 int s;
387 int maxretry = 10000; /* large, arbitrarily chosen */
388
389 if (!vn_isdisk(vp, NULL))
390 return (0);
391
392loop1:
393 /*
394 * MARK/SCAN initialization to avoid infinite loops
395 */
396 s = splbio();
397 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
398 bp = TAILQ_NEXT(bp, b_vnbufs)) {
399 bp->b_flags &= ~B_SCANNED;
400 }
401 splx(s);
402
403 /*
404 * Flush all dirty buffers associated with a block device.
405 */
406loop2:
407 s = splbio();
408 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
409 nbp = TAILQ_NEXT(bp, b_vnbufs);
410 if ((bp->b_flags & B_SCANNED) != 0)
411 continue;
412 bp->b_flags |= B_SCANNED;
413 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
414 continue;
415 if ((bp->b_flags & B_DELWRI) == 0)
416 panic("spec_fsync: not dirty");
417 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
418 BUF_UNLOCK(bp);
419 vfs_bio_awrite(bp);
420 splx(s);
421 } else {
422 bremfree(bp);
423 splx(s);
424 bawrite(bp);
425 }
426 goto loop2;
427 }
428
429 /*
430 * If synchronous the caller expects us to completely resolve all
431 * dirty buffers in the system. Wait for in-progress I/O to
432 * complete (which could include background bitmap writes), then
433 * retry if dirty blocks still exist.
434 */
435 if (ap->a_waitfor == MNT_WAIT) {
436 while (vp->v_numoutput) {
437 vp->v_flag |= VBWAIT;
377d4740 438 (void) tsleep((caddr_t)&vp->v_numoutput, 0, "spfsyn", 0);
984263bc
MD
439 }
440 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
441 if (--maxretry != 0) {
442 splx(s);
443 goto loop1;
444 }
445 vprint("spec_fsync: giving up on dirty", vp);
446 }
447 }
448 splx(s);
449 return (0);
450}
451
50775720
CP
452/*
453 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
454 */
984263bc 455static int
50775720 456spec_inactive(struct vop_inactive_args *ap)
984263bc 457{
41a01a4d 458 VOP_UNLOCK(ap->a_vp, NULL, 0, ap->a_td);
984263bc
MD
459 return (0);
460}
461
462/*
463 * Just call the device strategy routine
50775720
CP
464 *
465 * spec_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc
MD
466 */
467static int
50775720 468spec_strategy(struct vop_strategy_args *ap)
984263bc
MD
469{
470 struct buf *bp;
471 struct vnode *vp;
472 struct mount *mp;
473
474 bp = ap->a_bp;
475 if (((bp->b_flags & B_READ) == 0) &&
476 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
477 (*bioops.io_start)(bp);
478
479 /*
480 * Collect statistics on synchronous and asynchronous read
481 * and write counts for disks that have associated filesystems.
482 */
483 vp = ap->a_vp;
e4c9c0c8 484 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 485 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 486 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
487 mp->mnt_stat.f_asyncwrites++;
488 else
489 mp->mnt_stat.f_syncwrites++;
490 } else {
dadab5e9 491 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
492 mp->mnt_stat.f_asyncreads++;
493 else
494 mp->mnt_stat.f_syncreads++;
495 }
496 }
e4c9c0c8 497 bp->b_dev = vp->v_rdev;
984263bc
MD
498 BUF_STRATEGY(bp, 0);
499 return (0);
500}
501
50775720
CP
502/*
503 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
504 */
984263bc 505static int
50775720 506spec_freeblks(struct vop_freeblks_args *ap)
984263bc 507{
984263bc
MD
508 struct buf *bp;
509
510 /*
511 * XXX: This assumes that strategy does the deed right away.
512 * XXX: this may not be TRTTD.
513 */
335dda38 514 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
515 return (0);
516 bp = geteblk(ap->a_length);
517 bp->b_flags |= B_FREEBUF;
518 bp->b_dev = ap->a_vp->v_rdev;
519 bp->b_blkno = ap->a_addr;
520 bp->b_offset = dbtob(ap->a_addr);
521 bp->b_bcount = ap->a_length;
522 BUF_STRATEGY(bp, 0);
523 return (0);
524}
525
526/*
527 * Implement degenerate case where the block requested is the block
528 * returned, and assume that the entire device is contiguous in regards
529 * to the contiguous block range (runp and runb).
50775720
CP
530 *
531 * spec_bmap(struct vnode *a_vp, daddr_t a_bn, struct vnode **a_vpp,
532 * daddr_t *a_bnp, int *a_runp, int *a_runb)
984263bc
MD
533 */
534static int
50775720 535spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
536{
537 struct vnode *vp = ap->a_vp;
538 int runp = 0;
539 int runb = 0;
540
541 if (ap->a_vpp != NULL)
542 *ap->a_vpp = vp;
543 if (ap->a_bnp != NULL)
544 *ap->a_bnp = ap->a_bn;
545 if (vp->v_mount != NULL)
546 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
547 if (ap->a_runp != NULL)
548 *ap->a_runp = runp;
549 if (ap->a_runb != NULL)
550 *ap->a_runb = runb;
551 return (0);
552}
553
554/*
555 * Device close routine
50775720
CP
556 *
557 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
558 * struct thread *a_td)
984263bc
MD
559 */
560/* ARGSUSED */
561static int
50775720 562spec_close(struct vop_close_args *ap)
984263bc 563{
dadab5e9 564 struct proc *p = ap->a_td->td_proc;
984263bc 565 struct vnode *vp = ap->a_vp;
984263bc 566 dev_t dev = vp->v_rdev;
e4c9c0c8 567 int error;
984263bc
MD
568
569 /*
570 * Hack: a tty device that is a controlling terminal
571 * has a reference from the session structure.
572 * We cannot easily tell that a character device is
573 * a controlling terminal, unless it is the closing
574 * process' controlling terminal. In that case,
575 * if the reference count is 2 (this last descriptor
576 * plus the session), release the reference from the session.
577 */
e4c9c0c8
MD
578 reference_dev(dev);
579 if (vcount(vp) == 2 && vp->v_opencount == 1 &&
580 p && (vp->v_flag & VXLOCK) == 0 && vp == p->p_session->s_ttyvp) {
984263bc
MD
581 vrele(vp);
582 p->p_session->s_ttyvp = NULL;
583 }
e4c9c0c8 584
984263bc 585 /*
e4c9c0c8
MD
586 * Vnodes can be opened and close multiple times. Do not really
587 * close the device unless (1) it is being closed forcibly,
588 * (2) the device wants to track closes, or (3) this is the last
589 * vnode doing its last close on the device.
590 *
591 * XXX the VXLOCK (force close) case can leave vnodes referencing
592 * a closed device.
984263bc 593 */
e4c9c0c8
MD
594 if ((vp->v_flag & VXLOCK) ||
595 (dev_dflags(dev) & D_TRACKCLOSE) ||
596 (vcount(vp) <= 1 && vp->v_opencount == 1)) {
597 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
598 } else {
599 error = 0;
984263bc 600 }
e4c9c0c8
MD
601
602 /*
603 * Track the actual opens and closes on the vnode. The last close
604 * disassociates the rdev.
605 */
606 KKASSERT(vp->v_opencount > 0);
607 if (dev_ref_debug)
608 printf("spec_close: %s %d\n", dev->si_name, vp->v_opencount - 1);
609 if (--vp->v_opencount == 0)
610 v_release_rdev(vp);
611 release_dev(dev);
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{
984263bc
MD
637 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
638}
639
640static void
50775720 641spec_getpages_iodone(struct buf *bp)
984263bc 642{
984263bc
MD
643 bp->b_flags |= B_DONE;
644 wakeup(bp);
645}
646
647static int
50775720 648spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
649{
650 vm_offset_t kva;
651 int error;
652 int i, pcount, size, s;
653 daddr_t blkno;
654 struct buf *bp;
655 vm_page_t m;
656 vm_ooffset_t offset;
657 int toff, nextoff, nread;
658 struct vnode *vp = ap->a_vp;
659 int blksiz;
660 int gotreqpage;
661
662 error = 0;
663 pcount = round_page(ap->a_count) / PAGE_SIZE;
664
665 /*
666 * Calculate the offset of the transfer and do sanity check.
667 * FreeBSD currently only supports an 8 TB range due to b_blkno
668 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
669 * VOP_STRATEGY. XXX
670 */
671 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
672
673#define DADDR_T_BIT (sizeof(daddr_t)*8)
674#define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
675
676 if (offset < 0 || offset > OFFSET_MAX) {
677 /* XXX still no %q in kernel. */
678 printf("spec_getpages: preposterous offset 0x%x%08x\n",
679 (u_int)((u_quad_t)offset >> 32),
680 (u_int)(offset & 0xffffffff));
681 return (VM_PAGER_ERROR);
682 }
683
684 blkno = btodb(offset);
685
686 /*
687 * Round up physical size for real devices. We cannot round using
688 * v_mount's block size data because v_mount has nothing to do with
689 * the device. i.e. it's usually '/dev'. We need the physical block
690 * size for the device itself.
691 *
e4c9c0c8 692 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
693 * block device is mounted. However, we can use v_rdev.
694 */
695
696 if (vn_isdisk(vp, NULL))
697 blksiz = vp->v_rdev->si_bsize_phys;
698 else
699 blksiz = DEV_BSIZE;
700
701 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
702
703 bp = getpbuf(NULL);
704 kva = (vm_offset_t)bp->b_data;
705
706 /*
707 * Map the pages to be read into the kva.
708 */
709 pmap_qenter(kva, ap->a_m, pcount);
710
711 /* Build a minimal buffer header. */
712 bp->b_flags = B_READ | B_CALL;
713 bp->b_iodone = spec_getpages_iodone;
714
715 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
716 bp->b_blkno = blkno;
717 bp->b_lblkno = blkno;
718 pbgetvp(ap->a_vp, bp);
719 bp->b_bcount = size;
720 bp->b_bufsize = size;
721 bp->b_resid = 0;
722 bp->b_runningbufspace = bp->b_bufsize;
723 runningbufspace += bp->b_runningbufspace;
724
12e4aaff
MD
725 mycpu->gd_cnt.v_vnodein++;
726 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
727
728 /* Do the input. */
729 VOP_STRATEGY(bp->b_vp, bp);
730
731 s = splbio();
732
733 /* We definitely need to be at splbio here. */
26a0694b 734 while ((bp->b_flags & B_DONE) == 0) {
377d4740 735 tsleep(bp, 0, "spread", 0);
26a0694b 736 }
984263bc
MD
737
738 splx(s);
739
740 if ((bp->b_flags & B_ERROR) != 0) {
741 if (bp->b_error)
742 error = bp->b_error;
743 else
744 error = EIO;
745 }
746
747 nread = size - bp->b_resid;
748
749 if (nread < ap->a_count) {
750 bzero((caddr_t)kva + nread,
751 ap->a_count - nread);
752 }
753 pmap_qremove(kva, pcount);
754
755
756 gotreqpage = 0;
757 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
758 nextoff = toff + PAGE_SIZE;
759 m = ap->a_m[i];
760
761 m->flags &= ~PG_ZERO;
762
763 if (nextoff <= nread) {
764 m->valid = VM_PAGE_BITS_ALL;
765 vm_page_undirty(m);
766 } else if (toff < nread) {
767 /*
768 * Since this is a VM request, we have to supply the
769 * unaligned offset to allow vm_page_set_validclean()
770 * to zero sub-DEV_BSIZE'd portions of the page.
771 */
772 vm_page_set_validclean(m, 0, nread - toff);
773 } else {
774 m->valid = 0;
775 vm_page_undirty(m);
776 }
777
778 if (i != ap->a_reqpage) {
779 /*
780 * Just in case someone was asking for this page we
781 * now tell them that it is ok to use.
782 */
783 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
784 if (m->valid) {
785 if (m->flags & PG_WANTED) {
786 vm_page_activate(m);
787 } else {
788 vm_page_deactivate(m);
789 }
790 vm_page_wakeup(m);
791 } else {
792 vm_page_free(m);
793 }
794 } else {
795 vm_page_free(m);
796 }
797 } else if (m->valid) {
798 gotreqpage = 1;
799 /*
800 * Since this is a VM request, we need to make the
801 * entire page presentable by zeroing invalid sections.
802 */
803 if (m->valid != VM_PAGE_BITS_ALL)
804 vm_page_zero_invalid(m, FALSE);
805 }
806 }
807 if (!gotreqpage) {
808 m = ap->a_m[ap->a_reqpage];
809 printf(
810 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
811 devtoname(bp->b_dev), error, bp, bp->b_vp);
812 printf(
813 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
814 size, bp->b_resid, ap->a_count, m->valid);
815 printf(
816 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
817 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
818 /*
819 * Free the buffer header back to the swap buffer pool.
820 */
821 relpbuf(bp, NULL);
822 return VM_PAGER_ERROR;
823 }
824 /*
825 * Free the buffer header back to the swap buffer pool.
826 */
827 relpbuf(bp, NULL);
828 return VM_PAGER_OK;
829}