Build infrastructure for GCC 3.4
[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 $
72402f22 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.17 2004/06/06 18:47:51 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 297
72402f22
MD
298 if (dev == NULL) /* device was revoked */
299 return (EBADF);
984263bc
MD
300 if (uio->uio_resid == 0)
301 return (0);
302
41a01a4d 303 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 304 error = dev_dread(dev, uio, ap->a_ioflag);
41a01a4d 305 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
306 return (error);
307}
308
309/*
310 * Vnode op for write
50775720
CP
311 *
312 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
313 * struct ucred *a_cred)
984263bc
MD
314 */
315/* ARGSUSED */
316static int
50775720 317spec_write(struct vop_write_args *ap)
984263bc
MD
318{
319 struct vnode *vp;
dadab5e9 320 struct thread *td;
984263bc
MD
321 struct uio *uio;
322 dev_t dev;
323 int error;
324
325 vp = ap->a_vp;
326 dev = vp->v_rdev;
327 uio = ap->a_uio;
dadab5e9 328 td = uio->uio_td;
984263bc 329
72402f22
MD
330 if (dev == NULL) /* device was revoked */
331 return (EBADF);
332
41a01a4d 333 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 334 error = dev_dwrite(dev, uio, ap->a_ioflag);
41a01a4d 335 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
336 return (error);
337}
338
339/*
340 * Device ioctl operation.
50775720
CP
341 *
342 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
343 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
344 */
345/* ARGSUSED */
346static int
50775720 347spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
348{
349 dev_t dev;
350
72402f22
MD
351 if ((dev = ap->a_vp->v_rdev) == NULL)
352 return (EBADF); /* device was revoked */
353
335dda38
MD
354 return (dev_dioctl(dev, ap->a_command, ap->a_data,
355 ap->a_fflag, ap->a_td));
984263bc
MD
356}
357
50775720
CP
358/*
359 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
360 * struct thread *a_td)
361 */
984263bc
MD
362/* ARGSUSED */
363static int
50775720 364spec_poll(struct vop_poll_args *ap)
984263bc
MD
365{
366 dev_t dev;
367
72402f22
MD
368 if ((dev = ap->a_vp->v_rdev) == NULL)
369 return (EBADF); /* device was revoked */
335dda38 370 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
371}
372
50775720
CP
373/*
374 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
375 */
984263bc
MD
376/* ARGSUSED */
377static int
50775720 378spec_kqfilter(struct vop_kqfilter_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_dkqfilter(dev, ap->a_kn));
984263bc
MD
385}
386
387/*
388 * Synch buffers associated with a block device
50775720
CP
389 *
390 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
391 * int a_waitfor, struct thread *a_td)
984263bc
MD
392 */
393/* ARGSUSED */
394static int
50775720 395spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
396{
397 struct vnode *vp = ap->a_vp;
398 struct buf *bp;
399 struct buf *nbp;
400 int s;
401 int maxretry = 10000; /* large, arbitrarily chosen */
402
403 if (!vn_isdisk(vp, NULL))
404 return (0);
405
406loop1:
407 /*
408 * MARK/SCAN initialization to avoid infinite loops
409 */
410 s = splbio();
411 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
412 bp = TAILQ_NEXT(bp, b_vnbufs)) {
413 bp->b_flags &= ~B_SCANNED;
414 }
415 splx(s);
416
417 /*
418 * Flush all dirty buffers associated with a block device.
419 */
420loop2:
421 s = splbio();
422 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
423 nbp = TAILQ_NEXT(bp, b_vnbufs);
424 if ((bp->b_flags & B_SCANNED) != 0)
425 continue;
426 bp->b_flags |= B_SCANNED;
427 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
428 continue;
429 if ((bp->b_flags & B_DELWRI) == 0)
430 panic("spec_fsync: not dirty");
431 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
432 BUF_UNLOCK(bp);
433 vfs_bio_awrite(bp);
434 splx(s);
435 } else {
436 bremfree(bp);
437 splx(s);
438 bawrite(bp);
439 }
440 goto loop2;
441 }
442
443 /*
444 * If synchronous the caller expects us to completely resolve all
445 * dirty buffers in the system. Wait for in-progress I/O to
446 * complete (which could include background bitmap writes), then
447 * retry if dirty blocks still exist.
448 */
449 if (ap->a_waitfor == MNT_WAIT) {
450 while (vp->v_numoutput) {
451 vp->v_flag |= VBWAIT;
377d4740 452 (void) tsleep((caddr_t)&vp->v_numoutput, 0, "spfsyn", 0);
984263bc
MD
453 }
454 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
455 if (--maxretry != 0) {
456 splx(s);
457 goto loop1;
458 }
459 vprint("spec_fsync: giving up on dirty", vp);
460 }
461 }
462 splx(s);
463 return (0);
464}
465
50775720
CP
466/*
467 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
468 */
984263bc 469static int
50775720 470spec_inactive(struct vop_inactive_args *ap)
984263bc 471{
41a01a4d 472 VOP_UNLOCK(ap->a_vp, NULL, 0, ap->a_td);
984263bc
MD
473 return (0);
474}
475
476/*
477 * Just call the device strategy routine
50775720
CP
478 *
479 * spec_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc
MD
480 */
481static int
50775720 482spec_strategy(struct vop_strategy_args *ap)
984263bc
MD
483{
484 struct buf *bp;
485 struct vnode *vp;
486 struct mount *mp;
487
488 bp = ap->a_bp;
489 if (((bp->b_flags & B_READ) == 0) &&
490 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
491 (*bioops.io_start)(bp);
492
493 /*
494 * Collect statistics on synchronous and asynchronous read
495 * and write counts for disks that have associated filesystems.
496 */
497 vp = ap->a_vp;
72402f22 498 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 499 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 500 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 501 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
502 mp->mnt_stat.f_asyncwrites++;
503 else
504 mp->mnt_stat.f_syncwrites++;
505 } else {
dadab5e9 506 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
507 mp->mnt_stat.f_asyncreads++;
508 else
509 mp->mnt_stat.f_syncreads++;
510 }
511 }
e4c9c0c8 512 bp->b_dev = vp->v_rdev;
984263bc
MD
513 BUF_STRATEGY(bp, 0);
514 return (0);
515}
516
50775720
CP
517/*
518 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
519 */
984263bc 520static int
50775720 521spec_freeblks(struct vop_freeblks_args *ap)
984263bc 522{
984263bc
MD
523 struct buf *bp;
524
525 /*
526 * XXX: This assumes that strategy does the deed right away.
527 * XXX: this may not be TRTTD.
528 */
72402f22 529 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 530 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
531 return (0);
532 bp = geteblk(ap->a_length);
533 bp->b_flags |= B_FREEBUF;
534 bp->b_dev = ap->a_vp->v_rdev;
535 bp->b_blkno = ap->a_addr;
536 bp->b_offset = dbtob(ap->a_addr);
537 bp->b_bcount = ap->a_length;
538 BUF_STRATEGY(bp, 0);
539 return (0);
540}
541
542/*
543 * Implement degenerate case where the block requested is the block
544 * returned, and assume that the entire device is contiguous in regards
545 * to the contiguous block range (runp and runb).
50775720
CP
546 *
547 * spec_bmap(struct vnode *a_vp, daddr_t a_bn, struct vnode **a_vpp,
548 * daddr_t *a_bnp, int *a_runp, int *a_runb)
984263bc
MD
549 */
550static int
50775720 551spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
552{
553 struct vnode *vp = ap->a_vp;
554 int runp = 0;
555 int runb = 0;
556
557 if (ap->a_vpp != NULL)
558 *ap->a_vpp = vp;
559 if (ap->a_bnp != NULL)
560 *ap->a_bnp = ap->a_bn;
561 if (vp->v_mount != NULL)
562 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
563 if (ap->a_runp != NULL)
564 *ap->a_runp = runp;
565 if (ap->a_runb != NULL)
566 *ap->a_runb = runb;
567 return (0);
568}
569
570/*
571 * Device close routine
50775720
CP
572 *
573 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
574 * struct thread *a_td)
984263bc
MD
575 */
576/* ARGSUSED */
577static int
50775720 578spec_close(struct vop_close_args *ap)
984263bc 579{
dadab5e9 580 struct proc *p = ap->a_td->td_proc;
984263bc 581 struct vnode *vp = ap->a_vp;
984263bc 582 dev_t dev = vp->v_rdev;
e4c9c0c8 583 int error;
984263bc
MD
584
585 /*
586 * Hack: a tty device that is a controlling terminal
587 * has a reference from the session structure.
588 * We cannot easily tell that a character device is
589 * a controlling terminal, unless it is the closing
590 * process' controlling terminal. In that case,
591 * if the reference count is 2 (this last descriptor
592 * plus the session), release the reference from the session.
593 */
72402f22
MD
594 if (dev)
595 reference_dev(dev);
e4c9c0c8
MD
596 if (vcount(vp) == 2 && vp->v_opencount == 1 &&
597 p && (vp->v_flag & VXLOCK) == 0 && vp == p->p_session->s_ttyvp) {
984263bc
MD
598 vrele(vp);
599 p->p_session->s_ttyvp = NULL;
600 }
e4c9c0c8 601
984263bc 602 /*
e4c9c0c8
MD
603 * Vnodes can be opened and close multiple times. Do not really
604 * close the device unless (1) it is being closed forcibly,
605 * (2) the device wants to track closes, or (3) this is the last
606 * vnode doing its last close on the device.
607 *
608 * XXX the VXLOCK (force close) case can leave vnodes referencing
609 * a closed device.
984263bc 610 */
72402f22 611 if (dev && ((vp->v_flag & VXLOCK) ||
e4c9c0c8 612 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 613 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
e4c9c0c8
MD
614 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
615 } else {
616 error = 0;
984263bc 617 }
e4c9c0c8
MD
618
619 /*
620 * Track the actual opens and closes on the vnode. The last close
621 * disassociates the rdev.
622 */
623 KKASSERT(vp->v_opencount > 0);
72402f22
MD
624 if (dev_ref_debug) {
625 if (dev) {
626 printf("spec_close: %s %d\n",
627 dev->si_name, vp->v_opencount - 1);
628 } else {
629 printf("spec_close: closed revoked device %08x\n",
630 vp->v_udev);
631 }
632 }
e4c9c0c8
MD
633 if (--vp->v_opencount == 0)
634 v_release_rdev(vp);
72402f22
MD
635 if (dev)
636 release_dev(dev);
e4c9c0c8 637 return(error);
984263bc
MD
638}
639
640/*
641 * Print out the contents of a special device vnode.
50775720
CP
642 *
643 * spec_print(struct vnode *a_vp)
984263bc
MD
644 */
645static int
50775720 646spec_print(struct vop_print_args *ap)
984263bc 647{
984263bc
MD
648 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
649 return (0);
650}
651
652/*
653 * Special device advisory byte-level locks.
50775720
CP
654 *
655 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
656 * struct flock *a_fl, int a_flags)
984263bc
MD
657 */
658/* ARGSUSED */
659static int
50775720 660spec_advlock(struct vop_advlock_args *ap)
984263bc 661{
984263bc
MD
662 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
663}
664
665static void
50775720 666spec_getpages_iodone(struct buf *bp)
984263bc 667{
984263bc
MD
668 bp->b_flags |= B_DONE;
669 wakeup(bp);
670}
671
672static int
50775720 673spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
674{
675 vm_offset_t kva;
676 int error;
677 int i, pcount, size, s;
678 daddr_t blkno;
679 struct buf *bp;
680 vm_page_t m;
681 vm_ooffset_t offset;
682 int toff, nextoff, nread;
683 struct vnode *vp = ap->a_vp;
684 int blksiz;
685 int gotreqpage;
686
687 error = 0;
688 pcount = round_page(ap->a_count) / PAGE_SIZE;
689
690 /*
691 * Calculate the offset of the transfer and do sanity check.
692 * FreeBSD currently only supports an 8 TB range due to b_blkno
693 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
694 * VOP_STRATEGY. XXX
695 */
696 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
697
698#define DADDR_T_BIT (sizeof(daddr_t)*8)
699#define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
700
701 if (offset < 0 || offset > OFFSET_MAX) {
702 /* XXX still no %q in kernel. */
703 printf("spec_getpages: preposterous offset 0x%x%08x\n",
704 (u_int)((u_quad_t)offset >> 32),
705 (u_int)(offset & 0xffffffff));
706 return (VM_PAGER_ERROR);
707 }
708
709 blkno = btodb(offset);
710
711 /*
712 * Round up physical size for real devices. We cannot round using
713 * v_mount's block size data because v_mount has nothing to do with
714 * the device. i.e. it's usually '/dev'. We need the physical block
715 * size for the device itself.
716 *
e4c9c0c8 717 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
718 * block device is mounted. However, we can use v_rdev.
719 */
720
721 if (vn_isdisk(vp, NULL))
722 blksiz = vp->v_rdev->si_bsize_phys;
723 else
724 blksiz = DEV_BSIZE;
725
726 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
727
728 bp = getpbuf(NULL);
729 kva = (vm_offset_t)bp->b_data;
730
731 /*
732 * Map the pages to be read into the kva.
733 */
734 pmap_qenter(kva, ap->a_m, pcount);
735
736 /* Build a minimal buffer header. */
737 bp->b_flags = B_READ | B_CALL;
738 bp->b_iodone = spec_getpages_iodone;
739
740 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
741 bp->b_blkno = blkno;
742 bp->b_lblkno = blkno;
743 pbgetvp(ap->a_vp, bp);
744 bp->b_bcount = size;
745 bp->b_bufsize = size;
746 bp->b_resid = 0;
747 bp->b_runningbufspace = bp->b_bufsize;
748 runningbufspace += bp->b_runningbufspace;
749
12e4aaff
MD
750 mycpu->gd_cnt.v_vnodein++;
751 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
752
753 /* Do the input. */
754 VOP_STRATEGY(bp->b_vp, bp);
755
756 s = splbio();
757
758 /* We definitely need to be at splbio here. */
26a0694b 759 while ((bp->b_flags & B_DONE) == 0) {
377d4740 760 tsleep(bp, 0, "spread", 0);
26a0694b 761 }
984263bc
MD
762
763 splx(s);
764
765 if ((bp->b_flags & B_ERROR) != 0) {
766 if (bp->b_error)
767 error = bp->b_error;
768 else
769 error = EIO;
770 }
771
772 nread = size - bp->b_resid;
773
774 if (nread < ap->a_count) {
775 bzero((caddr_t)kva + nread,
776 ap->a_count - nread);
777 }
778 pmap_qremove(kva, pcount);
779
780
781 gotreqpage = 0;
782 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
783 nextoff = toff + PAGE_SIZE;
784 m = ap->a_m[i];
785
786 m->flags &= ~PG_ZERO;
787
788 if (nextoff <= nread) {
789 m->valid = VM_PAGE_BITS_ALL;
790 vm_page_undirty(m);
791 } else if (toff < nread) {
792 /*
793 * Since this is a VM request, we have to supply the
794 * unaligned offset to allow vm_page_set_validclean()
795 * to zero sub-DEV_BSIZE'd portions of the page.
796 */
797 vm_page_set_validclean(m, 0, nread - toff);
798 } else {
799 m->valid = 0;
800 vm_page_undirty(m);
801 }
802
803 if (i != ap->a_reqpage) {
804 /*
805 * Just in case someone was asking for this page we
806 * now tell them that it is ok to use.
807 */
808 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
809 if (m->valid) {
810 if (m->flags & PG_WANTED) {
811 vm_page_activate(m);
812 } else {
813 vm_page_deactivate(m);
814 }
815 vm_page_wakeup(m);
816 } else {
817 vm_page_free(m);
818 }
819 } else {
820 vm_page_free(m);
821 }
822 } else if (m->valid) {
823 gotreqpage = 1;
824 /*
825 * Since this is a VM request, we need to make the
826 * entire page presentable by zeroing invalid sections.
827 */
828 if (m->valid != VM_PAGE_BITS_ALL)
829 vm_page_zero_invalid(m, FALSE);
830 }
831 }
832 if (!gotreqpage) {
833 m = ap->a_m[ap->a_reqpage];
834 printf(
835 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
836 devtoname(bp->b_dev), error, bp, bp->b_vp);
837 printf(
838 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
839 size, bp->b_resid, ap->a_count, m->valid);
840 printf(
841 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
842 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
843 /*
844 * Free the buffer header back to the swap buffer pool.
845 */
846 relpbuf(bp, NULL);
847 return VM_PAGER_ERROR;
848 }
849 /*
850 * Free the buffer header back to the swap buffer pool.
851 */
852 relpbuf(bp, NULL);
853 return VM_PAGER_OK;
854}