Cleanup minor syntax/informational issues.
[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 $
53994a38 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.27 2005/09/03 08:01:32 dillon Exp $
984263bc
MD
36 */
37
38#include <sys/param.h>
39#include <sys/proc.h>
40#include <sys/systm.h>
41#include <sys/kernel.h>
42#include <sys/conf.h>
43#include <sys/buf.h>
335dda38 44#include <sys/device.h>
984263bc
MD
45#include <sys/mount.h>
46#include <sys/vnode.h>
47#include <sys/stat.h>
48#include <sys/fcntl.h>
49#include <sys/vmmeter.h>
50#include <sys/tty.h>
51
52#include <vm/vm.h>
53#include <vm/vm_object.h>
54#include <vm/vm_page.h>
55#include <vm/vm_pager.h>
56
3020e3be
MD
57#include <sys/buf2.h>
58
165dba55
DR
59#include <sys/thread2.h>
60
a6ee311a
RG
61static int spec_advlock (struct vop_advlock_args *);
62static int spec_bmap (struct vop_bmap_args *);
63static int spec_close (struct vop_close_args *);
64static int spec_freeblks (struct vop_freeblks_args *);
65static int spec_fsync (struct vop_fsync_args *);
66static int spec_getpages (struct vop_getpages_args *);
67static int spec_inactive (struct vop_inactive_args *);
68static int spec_ioctl (struct vop_ioctl_args *);
69static int spec_open (struct vop_open_args *);
70static int spec_poll (struct vop_poll_args *);
71static int spec_kqfilter (struct vop_kqfilter_args *);
72static int spec_print (struct vop_print_args *);
73static int spec_read (struct vop_read_args *);
74static int spec_strategy (struct vop_strategy_args *);
75static int spec_write (struct vop_write_args *);
984263bc 76
2d3e977e 77struct vop_ops *spec_vnode_vops;
0961aa92 78struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
2d3e977e
MD
79 { &vop_default_desc, vop_defaultop },
80 { &vop_access_desc, vop_ebadf },
625ddaba
JS
81 { &vop_advlock_desc, (vnodeopv_entry_t) spec_advlock },
82 { &vop_bmap_desc, (vnodeopv_entry_t) spec_bmap },
83 { &vop_close_desc, (vnodeopv_entry_t) spec_close },
2d3e977e 84 { &vop_create_desc, vop_panic },
625ddaba
JS
85 { &vop_freeblks_desc, (vnodeopv_entry_t) spec_freeblks },
86 { &vop_fsync_desc, (vnodeopv_entry_t) spec_fsync },
87 { &vop_getpages_desc, (vnodeopv_entry_t) spec_getpages },
88 { &vop_inactive_desc, (vnodeopv_entry_t) spec_inactive },
89 { &vop_ioctl_desc, (vnodeopv_entry_t) spec_ioctl },
2d3e977e
MD
90 { &vop_lease_desc, vop_null },
91 { &vop_link_desc, vop_panic },
92 { &vop_mkdir_desc, vop_panic },
93 { &vop_mknod_desc, vop_panic },
625ddaba
JS
94 { &vop_open_desc, (vnodeopv_entry_t) spec_open },
95 { &vop_pathconf_desc, (vnodeopv_entry_t) vop_stdpathconf },
96 { &vop_poll_desc, (vnodeopv_entry_t) spec_poll },
97 { &vop_kqfilter_desc, (vnodeopv_entry_t) spec_kqfilter },
98 { &vop_print_desc, (vnodeopv_entry_t) spec_print },
99 { &vop_read_desc, (vnodeopv_entry_t) spec_read },
2d3e977e
MD
100 { &vop_readdir_desc, vop_panic },
101 { &vop_readlink_desc, vop_panic },
102 { &vop_reallocblks_desc, vop_panic },
103 { &vop_reclaim_desc, vop_null },
104 { &vop_remove_desc, vop_panic },
105 { &vop_rename_desc, vop_panic },
106 { &vop_rmdir_desc, vop_panic },
107 { &vop_setattr_desc, vop_ebadf },
625ddaba 108 { &vop_strategy_desc, (vnodeopv_entry_t) spec_strategy },
2d3e977e 109 { &vop_symlink_desc, vop_panic },
625ddaba 110 { &vop_write_desc, (vnodeopv_entry_t) spec_write },
984263bc
MD
111 { NULL, NULL }
112};
113static struct vnodeopv_desc spec_vnodeop_opv_desc =
2d3e977e 114 { &spec_vnode_vops, spec_vnodeop_entries };
984263bc
MD
115VNODEOP_SET(spec_vnodeop_opv_desc);
116
e4c9c0c8
MD
117extern int dev_ref_debug;
118
50775720
CP
119/*
120 * spec_vnoperate(struct vnodeop_desc *a_desc, ...)
121 */
984263bc 122int
50775720 123spec_vnoperate(struct vop_generic_args *ap)
984263bc 124{
0961aa92 125 return (VOCALL(spec_vnode_vops, ap));
984263bc
MD
126}
127
a6ee311a 128static void spec_getpages_iodone (struct buf *bp);
984263bc
MD
129
130/*
131 * Open a special file.
50775720
CP
132 *
133 * spec_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
134 * struct thread *a_td)
984263bc
MD
135 */
136/* ARGSUSED */
137static int
50775720 138spec_open(struct vop_open_args *ap)
984263bc 139{
984263bc 140 struct vnode *vp = ap->a_vp;
e4c9c0c8 141 dev_t dev;
984263bc 142 int error;
e4c9c0c8 143 int isblk = (vp->v_type == VBLK) ? 1 : 0;
984263bc
MD
144 const char *cp;
145
146 /*
147 * Don't allow open if fs is mounted -nodev.
148 */
149 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
150 return (ENXIO);
151
e4c9c0c8
MD
152 /*
153 * Resolve the device. If the vnode is already open v_rdev may
154 * already be resolved. However, if the device changes out from
1fbb5fc0
MD
155 * under us we report it (and, for now, we allow it). Since
156 * v_release_rdev() zero's v_opencount, we have to save and restore
157 * it when replacing the rdev reference.
e4c9c0c8
MD
158 */
159 if (vp->v_rdev != NULL) {
160 dev = udev2dev(vp->v_udev, isblk);
161 if (dev != vp->v_rdev) {
1fbb5fc0 162 int oc = vp->v_opencount;
e4c9c0c8
MD
163 printf(
164 "Warning: spec_open: dev %s was lost",
165 vp->v_rdev->si_name);
166 v_release_rdev(vp);
167 error = v_associate_rdev(vp,
168 udev2dev(vp->v_udev, isblk));
1fbb5fc0 169 if (error) {
e4c9c0c8 170 printf(", reacquisition failed\n");
1fbb5fc0
MD
171 } else {
172 vp->v_opencount = oc;
e4c9c0c8 173 printf(", reacquisition successful\n");
1fbb5fc0 174 }
e4c9c0c8
MD
175 } else {
176 error = 0;
177 }
178 } else {
179 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
180 }
181 if (error)
182 return(error);
183
500df1da
MD
184 /*
185 * Prevent degenerate open/close sequences from nulling out rdev.
186 */
187 ++vp->v_opencount;
e4c9c0c8 188 dev = vp->v_rdev;
500df1da 189 KKASSERT(dev != NULL);
984263bc 190
e4c9c0c8
MD
191 /*
192 * Make this field valid before any I/O in ->d_open. XXX the
193 * device itself should probably be required to initialize
194 * this field in d_open.
195 */
984263bc
MD
196 if (!dev->si_iosize_max)
197 dev->si_iosize_max = DFLTPHYS;
198
199 /*
200 * XXX: Disks get special billing here, but it is mostly wrong.
201 * XXX: diskpartitions can overlap and the real checks should
202 * XXX: take this into account, and consequently they need to
203 * XXX: live in the diskslicing code. Some checks do.
204 */
205 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
206 (ap->a_mode & FWRITE)) {
207 /*
208 * Never allow opens for write if the device is mounted R/W
209 */
e4c9c0c8
MD
210 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
211 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
212 error = EBUSY;
213 goto done;
214 }
984263bc
MD
215
216 /*
217 * When running in secure mode, do not allow opens
218 * for writing if the device is mounted
219 */
e4c9c0c8
MD
220 if (securelevel >= 1 && vfs_mountedon(vp)) {
221 error = EPERM;
222 goto done;
223 }
984263bc
MD
224
225 /*
226 * When running in very secure mode, do not allow
227 * opens for writing of any devices.
228 */
e4c9c0c8
MD
229 if (securelevel >= 2) {
230 error = EPERM;
231 goto done;
232 }
984263bc
MD
233 }
234
235 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 236 if (dev_dflags(dev) & D_TTY)
984263bc
MD
237 vp->v_flag |= VISTTY;
238
e4c9c0c8
MD
239 /*
240 * dev_dopen() is always called for each open. dev_dclose() is
241 * only called for the last close unless D_TRACKCLOSE is set.
242 */
5fd012e0 243 VOP_UNLOCK(vp, 0, ap->a_td);
335dda38 244 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
5fd012e0 245 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
984263bc
MD
246
247 if (error)
e4c9c0c8 248 goto done;
984263bc 249
335dda38 250 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
251 if (dev->si_tty) {
252 struct tty *tp;
253 tp = dev->si_tty;
254 if (!tp->t_stop) {
255 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
256 tp->t_stop = nottystop;
257 }
258 }
259 }
260
261 if (vn_isdisk(vp, NULL)) {
262 if (!dev->si_bsize_phys)
263 dev->si_bsize_phys = DEV_BSIZE;
264 }
335dda38 265 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 266 cp = devtoname(dev);
335dda38 267 if (*cp == '#') {
984263bc 268 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 269 dev_dname(dev), cp);
984263bc
MD
270 }
271 }
fad57d0e
MD
272
273 /*
274 * If we were handed a file pointer we may be able to install a
275 * shortcut which issues device read and write operations directly
276 * from the fileops rather then having to go through spec_read()
277 * and spec_write().
278 */
279 if (ap->a_fp)
280 vn_setspecops(ap->a_fp);
281
e4c9c0c8
MD
282 if (dev_ref_debug)
283 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
284done:
285 if (error) {
500df1da
MD
286 KKASSERT(vp->v_opencount > 0);
287 if (--vp->v_opencount == 0)
e4c9c0c8
MD
288 v_release_rdev(vp);
289 }
984263bc
MD
290 return (error);
291}
292
293/*
294 * Vnode op for read
50775720
CP
295 *
296 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
297 * struct ucred *a_cred)
984263bc
MD
298 */
299/* ARGSUSED */
300static int
50775720 301spec_read(struct vop_read_args *ap)
984263bc
MD
302{
303 struct vnode *vp;
dadab5e9 304 struct thread *td;
984263bc
MD
305 struct uio *uio;
306 dev_t dev;
307 int error;
308
309 vp = ap->a_vp;
310 dev = vp->v_rdev;
311 uio = ap->a_uio;
dadab5e9 312 td = uio->uio_td;
984263bc 313
72402f22
MD
314 if (dev == NULL) /* device was revoked */
315 return (EBADF);
984263bc
MD
316 if (uio->uio_resid == 0)
317 return (0);
318
5fd012e0 319 VOP_UNLOCK(vp, 0, td);
335dda38 320 error = dev_dread(dev, uio, ap->a_ioflag);
5fd012e0 321 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
322 return (error);
323}
324
325/*
326 * Vnode op for write
50775720
CP
327 *
328 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
329 * struct ucred *a_cred)
984263bc
MD
330 */
331/* ARGSUSED */
332static int
50775720 333spec_write(struct vop_write_args *ap)
984263bc
MD
334{
335 struct vnode *vp;
dadab5e9 336 struct thread *td;
984263bc
MD
337 struct uio *uio;
338 dev_t dev;
339 int error;
340
341 vp = ap->a_vp;
342 dev = vp->v_rdev;
343 uio = ap->a_uio;
dadab5e9 344 td = uio->uio_td;
984263bc 345
72402f22
MD
346 if (dev == NULL) /* device was revoked */
347 return (EBADF);
348
5fd012e0 349 VOP_UNLOCK(vp, 0, td);
335dda38 350 error = dev_dwrite(dev, uio, ap->a_ioflag);
5fd012e0 351 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
352 return (error);
353}
354
355/*
356 * Device ioctl operation.
50775720
CP
357 *
358 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
359 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
360 */
361/* ARGSUSED */
362static int
50775720 363spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
364{
365 dev_t dev;
366
72402f22
MD
367 if ((dev = ap->a_vp->v_rdev) == NULL)
368 return (EBADF); /* device was revoked */
369
335dda38
MD
370 return (dev_dioctl(dev, ap->a_command, ap->a_data,
371 ap->a_fflag, ap->a_td));
984263bc
MD
372}
373
50775720
CP
374/*
375 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
376 * struct thread *a_td)
377 */
984263bc
MD
378/* ARGSUSED */
379static int
50775720 380spec_poll(struct vop_poll_args *ap)
984263bc
MD
381{
382 dev_t dev;
383
72402f22
MD
384 if ((dev = ap->a_vp->v_rdev) == NULL)
385 return (EBADF); /* device was revoked */
335dda38 386 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
387}
388
50775720
CP
389/*
390 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
391 */
984263bc
MD
392/* ARGSUSED */
393static int
50775720 394spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
395{
396 dev_t dev;
397
72402f22
MD
398 if ((dev = ap->a_vp->v_rdev) == NULL)
399 return (EBADF); /* device was revoked */
335dda38 400 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
401}
402
403/*
404 * Synch buffers associated with a block device
50775720
CP
405 *
406 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
407 * int a_waitfor, struct thread *a_td)
984263bc
MD
408 */
409/* ARGSUSED */
410static int
50775720 411spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
412{
413 struct vnode *vp = ap->a_vp;
6bae6177 414 int error;
984263bc
MD
415
416 if (!vn_isdisk(vp, NULL))
417 return (0);
418
984263bc
MD
419 /*
420 * Flush all dirty buffers associated with a block device.
421 */
6bae6177
MD
422 error = vfsync(vp, ap->a_waitfor, 10000, (daddr_t)-1, NULL, NULL);
423 return (error);
984263bc
MD
424}
425
50775720
CP
426/*
427 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
428 */
984263bc 429static int
50775720 430spec_inactive(struct vop_inactive_args *ap)
984263bc 431{
984263bc
MD
432 return (0);
433}
434
435/*
436 * Just call the device strategy routine
50775720
CP
437 *
438 * spec_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc
MD
439 */
440static int
50775720 441spec_strategy(struct vop_strategy_args *ap)
984263bc
MD
442{
443 struct buf *bp;
444 struct vnode *vp;
445 struct mount *mp;
446
447 bp = ap->a_bp;
448 if (((bp->b_flags & B_READ) == 0) &&
449 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
450 (*bioops.io_start)(bp);
451
452 /*
453 * Collect statistics on synchronous and asynchronous read
454 * and write counts for disks that have associated filesystems.
455 */
456 vp = ap->a_vp;
72402f22 457 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 458 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 459 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 460 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
461 mp->mnt_stat.f_asyncwrites++;
462 else
463 mp->mnt_stat.f_syncwrites++;
464 } else {
dadab5e9 465 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
466 mp->mnt_stat.f_asyncreads++;
467 else
468 mp->mnt_stat.f_syncreads++;
469 }
470 }
e4c9c0c8 471 bp->b_dev = vp->v_rdev;
984263bc
MD
472 BUF_STRATEGY(bp, 0);
473 return (0);
474}
475
50775720
CP
476/*
477 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
478 */
984263bc 479static int
50775720 480spec_freeblks(struct vop_freeblks_args *ap)
984263bc 481{
984263bc
MD
482 struct buf *bp;
483
484 /*
485 * XXX: This assumes that strategy does the deed right away.
486 * XXX: this may not be TRTTD.
487 */
72402f22 488 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 489 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
490 return (0);
491 bp = geteblk(ap->a_length);
492 bp->b_flags |= B_FREEBUF;
493 bp->b_dev = ap->a_vp->v_rdev;
494 bp->b_blkno = ap->a_addr;
495 bp->b_offset = dbtob(ap->a_addr);
496 bp->b_bcount = ap->a_length;
497 BUF_STRATEGY(bp, 0);
498 return (0);
499}
500
501/*
502 * Implement degenerate case where the block requested is the block
503 * returned, and assume that the entire device is contiguous in regards
504 * to the contiguous block range (runp and runb).
50775720
CP
505 *
506 * spec_bmap(struct vnode *a_vp, daddr_t a_bn, struct vnode **a_vpp,
507 * daddr_t *a_bnp, int *a_runp, int *a_runb)
984263bc
MD
508 */
509static int
50775720 510spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
511{
512 struct vnode *vp = ap->a_vp;
513 int runp = 0;
514 int runb = 0;
515
516 if (ap->a_vpp != NULL)
517 *ap->a_vpp = vp;
518 if (ap->a_bnp != NULL)
519 *ap->a_bnp = ap->a_bn;
520 if (vp->v_mount != NULL)
521 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
522 if (ap->a_runp != NULL)
523 *ap->a_runp = runp;
524 if (ap->a_runb != NULL)
525 *ap->a_runb = runb;
526 return (0);
527}
528
529/*
530 * Device close routine
50775720
CP
531 *
532 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
533 * struct thread *a_td)
53994a38
MD
534 *
535 * NOTE: the vnode may or may not be locked on call.
984263bc
MD
536 */
537/* ARGSUSED */
538static int
50775720 539spec_close(struct vop_close_args *ap)
984263bc 540{
dadab5e9 541 struct proc *p = ap->a_td->td_proc;
984263bc 542 struct vnode *vp = ap->a_vp;
984263bc 543 dev_t dev = vp->v_rdev;
e4c9c0c8 544 int error;
53994a38 545 int needrelock;
984263bc
MD
546
547 /*
548 * Hack: a tty device that is a controlling terminal
549 * has a reference from the session structure.
550 * We cannot easily tell that a character device is
551 * a controlling terminal, unless it is the closing
552 * process' controlling terminal. In that case,
553 * if the reference count is 2 (this last descriptor
554 * plus the session), release the reference from the session.
1fbb5fc0
MD
555 *
556 * It is possible for v_opencount to be 0 or 1 in this case, 0
557 * because the tty might have been revoked.
984263bc 558 */
72402f22
MD
559 if (dev)
560 reference_dev(dev);
1fbb5fc0 561 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 562 p && vp == p->p_session->s_ttyvp) {
984263bc 563 p->p_session->s_ttyvp = NULL;
1fbb5fc0 564 vrele(vp);
984263bc 565 }
e4c9c0c8 566
984263bc 567 /*
e4c9c0c8
MD
568 * Vnodes can be opened and close multiple times. Do not really
569 * close the device unless (1) it is being closed forcibly,
570 * (2) the device wants to track closes, or (3) this is the last
571 * vnode doing its last close on the device.
572 *
573 * XXX the VXLOCK (force close) case can leave vnodes referencing
574 * a closed device.
984263bc 575 */
5fd012e0 576 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 577 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 578 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
53994a38
MD
579 needrelock = 0;
580 if (VOP_ISLOCKED(vp, ap->a_td)) {
581 needrelock = 1;
582 VOP_UNLOCK(vp, 0, ap->a_td);
583 }
e4c9c0c8 584 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
53994a38
MD
585 if (needrelock)
586 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
e4c9c0c8
MD
587 } else {
588 error = 0;
984263bc 589 }
e4c9c0c8
MD
590
591 /*
592 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
593 * disassociates the rdev. If the rdev is already disassociated
594 * the vnode might have been revoked and no further opencount
595 * tracking occurs.
e4c9c0c8 596 */
1fbb5fc0
MD
597 if (dev) {
598 KKASSERT(vp->v_opencount > 0);
599 if (dev_ref_debug) {
72402f22
MD
600 printf("spec_close: %s %d\n",
601 dev->si_name, vp->v_opencount - 1);
72402f22 602 }
1fbb5fc0
MD
603 if (--vp->v_opencount == 0)
604 v_release_rdev(vp);
72402f22 605 release_dev(dev);
1fbb5fc0 606 }
e4c9c0c8 607 return(error);
984263bc
MD
608}
609
610/*
611 * Print out the contents of a special device vnode.
50775720
CP
612 *
613 * spec_print(struct vnode *a_vp)
984263bc
MD
614 */
615static int
50775720 616spec_print(struct vop_print_args *ap)
984263bc 617{
984263bc
MD
618 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
619 return (0);
620}
621
622/*
623 * Special device advisory byte-level locks.
50775720
CP
624 *
625 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
626 * struct flock *a_fl, int a_flags)
984263bc
MD
627 */
628/* ARGSUSED */
629static int
50775720 630spec_advlock(struct vop_advlock_args *ap)
984263bc 631{
984263bc
MD
632 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
633}
634
635static void
50775720 636spec_getpages_iodone(struct buf *bp)
984263bc 637{
984263bc
MD
638 bp->b_flags |= B_DONE;
639 wakeup(bp);
640}
641
642static int
50775720 643spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
644{
645 vm_offset_t kva;
646 int error;
165dba55 647 int i, pcount, size;
984263bc
MD
648 daddr_t blkno;
649 struct buf *bp;
650 vm_page_t m;
651 vm_ooffset_t offset;
652 int toff, nextoff, nread;
653 struct vnode *vp = ap->a_vp;
654 int blksiz;
655 int gotreqpage;
656
657 error = 0;
658 pcount = round_page(ap->a_count) / PAGE_SIZE;
659
660 /*
661 * Calculate the offset of the transfer and do sanity check.
662 * FreeBSD currently only supports an 8 TB range due to b_blkno
663 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
664 * VOP_STRATEGY. XXX
665 */
666 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
667
668#define DADDR_T_BIT (sizeof(daddr_t)*8)
669#define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
670
671 if (offset < 0 || offset > OFFSET_MAX) {
672 /* XXX still no %q in kernel. */
673 printf("spec_getpages: preposterous offset 0x%x%08x\n",
674 (u_int)((u_quad_t)offset >> 32),
675 (u_int)(offset & 0xffffffff));
676 return (VM_PAGER_ERROR);
677 }
678
679 blkno = btodb(offset);
680
681 /*
682 * Round up physical size for real devices. We cannot round using
683 * v_mount's block size data because v_mount has nothing to do with
684 * the device. i.e. it's usually '/dev'. We need the physical block
685 * size for the device itself.
686 *
e4c9c0c8 687 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
688 * block device is mounted. However, we can use v_rdev.
689 */
690
691 if (vn_isdisk(vp, NULL))
692 blksiz = vp->v_rdev->si_bsize_phys;
693 else
694 blksiz = DEV_BSIZE;
695
696 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
697
698 bp = getpbuf(NULL);
699 kva = (vm_offset_t)bp->b_data;
700
701 /*
702 * Map the pages to be read into the kva.
703 */
704 pmap_qenter(kva, ap->a_m, pcount);
705
706 /* Build a minimal buffer header. */
5ff4cfb5 707 bp->b_flags = B_READ;
984263bc
MD
708 bp->b_iodone = spec_getpages_iodone;
709
710 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
711 bp->b_blkno = blkno;
712 bp->b_lblkno = blkno;
713 pbgetvp(ap->a_vp, bp);
714 bp->b_bcount = size;
715 bp->b_bufsize = size;
716 bp->b_resid = 0;
717 bp->b_runningbufspace = bp->b_bufsize;
718 runningbufspace += bp->b_runningbufspace;
719
12e4aaff
MD
720 mycpu->gd_cnt.v_vnodein++;
721 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
722
723 /* Do the input. */
724 VOP_STRATEGY(bp->b_vp, bp);
725
165dba55 726 crit_enter();
984263bc
MD
727
728 /* We definitely need to be at splbio here. */
26a0694b 729 while ((bp->b_flags & B_DONE) == 0) {
377d4740 730 tsleep(bp, 0, "spread", 0);
26a0694b 731 }
984263bc 732
165dba55 733 crit_exit();
984263bc
MD
734
735 if ((bp->b_flags & B_ERROR) != 0) {
736 if (bp->b_error)
737 error = bp->b_error;
738 else
739 error = EIO;
740 }
741
742 nread = size - bp->b_resid;
743
744 if (nread < ap->a_count) {
745 bzero((caddr_t)kva + nread,
746 ap->a_count - nread);
747 }
748 pmap_qremove(kva, pcount);
749
750
751 gotreqpage = 0;
752 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
753 nextoff = toff + PAGE_SIZE;
754 m = ap->a_m[i];
755
756 m->flags &= ~PG_ZERO;
757
758 if (nextoff <= nread) {
759 m->valid = VM_PAGE_BITS_ALL;
760 vm_page_undirty(m);
761 } else if (toff < nread) {
762 /*
763 * Since this is a VM request, we have to supply the
764 * unaligned offset to allow vm_page_set_validclean()
765 * to zero sub-DEV_BSIZE'd portions of the page.
766 */
767 vm_page_set_validclean(m, 0, nread - toff);
768 } else {
769 m->valid = 0;
770 vm_page_undirty(m);
771 }
772
773 if (i != ap->a_reqpage) {
774 /*
775 * Just in case someone was asking for this page we
776 * now tell them that it is ok to use.
777 */
778 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
779 if (m->valid) {
780 if (m->flags & PG_WANTED) {
781 vm_page_activate(m);
782 } else {
783 vm_page_deactivate(m);
784 }
785 vm_page_wakeup(m);
786 } else {
787 vm_page_free(m);
788 }
789 } else {
790 vm_page_free(m);
791 }
792 } else if (m->valid) {
793 gotreqpage = 1;
794 /*
795 * Since this is a VM request, we need to make the
796 * entire page presentable by zeroing invalid sections.
797 */
798 if (m->valid != VM_PAGE_BITS_ALL)
799 vm_page_zero_invalid(m, FALSE);
800 }
801 }
802 if (!gotreqpage) {
803 m = ap->a_m[ap->a_reqpage];
804 printf(
805 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
806 devtoname(bp->b_dev), error, bp, bp->b_vp);
807 printf(
808 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
809 size, bp->b_resid, ap->a_count, m->valid);
810 printf(
811 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
812 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
813 /*
814 * Free the buffer header back to the swap buffer pool.
815 */
816 relpbuf(bp, NULL);
817 return VM_PAGER_ERROR;
818 }
819 /*
820 * Free the buffer header back to the swap buffer pool.
821 */
822 relpbuf(bp, NULL);
823 return VM_PAGER_OK;
824}