Add PCI IDs for Intel's ICH7 and ICH7M ATA/SATA hardware, used in
[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 $
e07fef60 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.32 2006/03/27 16:18:59 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 },
e62afb5f 84 { &vop_old_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 },
e62afb5f
MD
90 { &vop_old_link_desc, vop_panic },
91 { &vop_old_mkdir_desc, vop_panic },
92 { &vop_old_mknod_desc, vop_panic },
625ddaba
JS
93 { &vop_open_desc, (vnodeopv_entry_t) spec_open },
94 { &vop_pathconf_desc, (vnodeopv_entry_t) vop_stdpathconf },
95 { &vop_poll_desc, (vnodeopv_entry_t) spec_poll },
96 { &vop_kqfilter_desc, (vnodeopv_entry_t) spec_kqfilter },
97 { &vop_print_desc, (vnodeopv_entry_t) spec_print },
98 { &vop_read_desc, (vnodeopv_entry_t) spec_read },
2d3e977e
MD
99 { &vop_readdir_desc, vop_panic },
100 { &vop_readlink_desc, vop_panic },
101 { &vop_reallocblks_desc, vop_panic },
102 { &vop_reclaim_desc, vop_null },
e62afb5f
MD
103 { &vop_old_remove_desc, vop_panic },
104 { &vop_old_rename_desc, vop_panic },
105 { &vop_old_rmdir_desc, vop_panic },
2d3e977e 106 { &vop_setattr_desc, vop_ebadf },
625ddaba 107 { &vop_strategy_desc, (vnodeopv_entry_t) spec_strategy },
e62afb5f 108 { &vop_old_symlink_desc, vop_panic },
625ddaba 109 { &vop_write_desc, (vnodeopv_entry_t) spec_write },
984263bc
MD
110 { NULL, NULL }
111};
112static struct vnodeopv_desc spec_vnodeop_opv_desc =
dc1be39c 113 { &spec_vnode_vops, spec_vnodeop_entries, 0 };
984263bc
MD
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 123{
0961aa92 124 return (VOCALL(spec_vnode_vops, ap));
984263bc
MD
125}
126
81b5c339 127static void spec_getpages_iodone (struct bio *bio);
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
1fbb5fc0
MD
154 * under us we report it (and, for now, we allow it). Since
155 * v_release_rdev() zero's v_opencount, we have to save and restore
156 * it when replacing the rdev reference.
e4c9c0c8
MD
157 */
158 if (vp->v_rdev != NULL) {
159 dev = udev2dev(vp->v_udev, isblk);
160 if (dev != vp->v_rdev) {
1fbb5fc0 161 int oc = vp->v_opencount;
e4c9c0c8
MD
162 printf(
163 "Warning: spec_open: dev %s was lost",
164 vp->v_rdev->si_name);
165 v_release_rdev(vp);
166 error = v_associate_rdev(vp,
167 udev2dev(vp->v_udev, isblk));
1fbb5fc0 168 if (error) {
e4c9c0c8 169 printf(", reacquisition failed\n");
1fbb5fc0
MD
170 } else {
171 vp->v_opencount = oc;
e4c9c0c8 172 printf(", reacquisition successful\n");
1fbb5fc0 173 }
e4c9c0c8
MD
174 } else {
175 error = 0;
176 }
177 } else {
178 error = v_associate_rdev(vp, udev2dev(vp->v_udev, isblk));
179 }
180 if (error)
181 return(error);
182
500df1da
MD
183 /*
184 * Prevent degenerate open/close sequences from nulling out rdev.
185 */
186 ++vp->v_opencount;
e4c9c0c8 187 dev = vp->v_rdev;
500df1da 188 KKASSERT(dev != NULL);
984263bc 189
e4c9c0c8
MD
190 /*
191 * Make this field valid before any I/O in ->d_open. XXX the
192 * device itself should probably be required to initialize
193 * this field in d_open.
194 */
984263bc
MD
195 if (!dev->si_iosize_max)
196 dev->si_iosize_max = DFLTPHYS;
197
198 /*
199 * XXX: Disks get special billing here, but it is mostly wrong.
200 * XXX: diskpartitions can overlap and the real checks should
201 * XXX: take this into account, and consequently they need to
202 * XXX: live in the diskslicing code. Some checks do.
203 */
204 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
205 (ap->a_mode & FWRITE)) {
206 /*
207 * Never allow opens for write if the device is mounted R/W
208 */
e4c9c0c8
MD
209 if (vp->v_rdev && vp->v_rdev->si_mountpoint &&
210 !(vp->v_rdev->si_mountpoint->mnt_flag & MNT_RDONLY)) {
211 error = EBUSY;
212 goto done;
213 }
984263bc
MD
214
215 /*
216 * When running in secure mode, do not allow opens
217 * for writing if the device is mounted
218 */
e4c9c0c8
MD
219 if (securelevel >= 1 && vfs_mountedon(vp)) {
220 error = EPERM;
221 goto done;
222 }
984263bc
MD
223
224 /*
225 * When running in very secure mode, do not allow
226 * opens for writing of any devices.
227 */
e4c9c0c8
MD
228 if (securelevel >= 2) {
229 error = EPERM;
230 goto done;
231 }
984263bc
MD
232 }
233
234 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 235 if (dev_dflags(dev) & D_TTY)
984263bc
MD
236 vp->v_flag |= VISTTY;
237
e4c9c0c8
MD
238 /*
239 * dev_dopen() is always called for each open. dev_dclose() is
240 * only called for the last close unless D_TRACKCLOSE is set.
241 */
5fd012e0 242 VOP_UNLOCK(vp, 0, ap->a_td);
335dda38 243 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
5fd012e0 244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
984263bc
MD
245
246 if (error)
e4c9c0c8 247 goto done;
984263bc 248
335dda38 249 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
250 if (dev->si_tty) {
251 struct tty *tp;
252 tp = dev->si_tty;
253 if (!tp->t_stop) {
254 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
255 tp->t_stop = nottystop;
256 }
257 }
258 }
259
260 if (vn_isdisk(vp, NULL)) {
261 if (!dev->si_bsize_phys)
262 dev->si_bsize_phys = DEV_BSIZE;
263 }
335dda38 264 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 265 cp = devtoname(dev);
335dda38 266 if (*cp == '#') {
984263bc 267 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 268 dev_dname(dev), cp);
984263bc
MD
269 }
270 }
fad57d0e
MD
271
272 /*
273 * If we were handed a file pointer we may be able to install a
274 * shortcut which issues device read and write operations directly
275 * from the fileops rather then having to go through spec_read()
276 * and spec_write().
277 */
278 if (ap->a_fp)
279 vn_setspecops(ap->a_fp);
280
e4c9c0c8
MD
281 if (dev_ref_debug)
282 printf("spec_open: %s %d\n", dev->si_name, vp->v_opencount);
283done:
284 if (error) {
500df1da
MD
285 KKASSERT(vp->v_opencount > 0);
286 if (--vp->v_opencount == 0)
e4c9c0c8
MD
287 v_release_rdev(vp);
288 }
984263bc
MD
289 return (error);
290}
291
292/*
293 * Vnode op for read
50775720
CP
294 *
295 * spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
296 * struct ucred *a_cred)
984263bc
MD
297 */
298/* ARGSUSED */
299static int
50775720 300spec_read(struct vop_read_args *ap)
984263bc
MD
301{
302 struct vnode *vp;
dadab5e9 303 struct thread *td;
984263bc
MD
304 struct uio *uio;
305 dev_t dev;
306 int error;
307
308 vp = ap->a_vp;
309 dev = vp->v_rdev;
310 uio = ap->a_uio;
dadab5e9 311 td = uio->uio_td;
984263bc 312
72402f22
MD
313 if (dev == NULL) /* device was revoked */
314 return (EBADF);
984263bc
MD
315 if (uio->uio_resid == 0)
316 return (0);
317
5fd012e0 318 VOP_UNLOCK(vp, 0, td);
335dda38 319 error = dev_dread(dev, uio, ap->a_ioflag);
5fd012e0 320 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
321 return (error);
322}
323
324/*
325 * Vnode op for write
50775720
CP
326 *
327 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
328 * struct ucred *a_cred)
984263bc
MD
329 */
330/* ARGSUSED */
331static int
50775720 332spec_write(struct vop_write_args *ap)
984263bc
MD
333{
334 struct vnode *vp;
dadab5e9 335 struct thread *td;
984263bc
MD
336 struct uio *uio;
337 dev_t dev;
338 int error;
339
340 vp = ap->a_vp;
341 dev = vp->v_rdev;
342 uio = ap->a_uio;
dadab5e9 343 td = uio->uio_td;
984263bc 344
72402f22
MD
345 if (dev == NULL) /* device was revoked */
346 return (EBADF);
347
5fd012e0 348 VOP_UNLOCK(vp, 0, td);
335dda38 349 error = dev_dwrite(dev, uio, ap->a_ioflag);
5fd012e0 350 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
351 return (error);
352}
353
354/*
355 * Device ioctl operation.
50775720
CP
356 *
357 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
358 * int a_fflag, struct ucred *a_cred, struct thread *a_td)
984263bc
MD
359 */
360/* ARGSUSED */
361static int
50775720 362spec_ioctl(struct vop_ioctl_args *ap)
984263bc
MD
363{
364 dev_t dev;
365
72402f22
MD
366 if ((dev = ap->a_vp->v_rdev) == NULL)
367 return (EBADF); /* device was revoked */
368
335dda38
MD
369 return (dev_dioctl(dev, ap->a_command, ap->a_data,
370 ap->a_fflag, ap->a_td));
984263bc
MD
371}
372
50775720
CP
373/*
374 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
375 * struct thread *a_td)
376 */
984263bc
MD
377/* ARGSUSED */
378static int
50775720 379spec_poll(struct vop_poll_args *ap)
984263bc
MD
380{
381 dev_t dev;
382
72402f22
MD
383 if ((dev = ap->a_vp->v_rdev) == NULL)
384 return (EBADF); /* device was revoked */
335dda38 385 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
386}
387
50775720
CP
388/*
389 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
390 */
984263bc
MD
391/* ARGSUSED */
392static int
50775720 393spec_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
394{
395 dev_t dev;
396
72402f22
MD
397 if ((dev = ap->a_vp->v_rdev) == NULL)
398 return (EBADF); /* device was revoked */
335dda38 399 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
400}
401
402/*
403 * Synch buffers associated with a block device
50775720
CP
404 *
405 * spec_fsync(struct vnode *a_vp, struct ucred *a_cred,
406 * int a_waitfor, struct thread *a_td)
984263bc
MD
407 */
408/* ARGSUSED */
409static int
50775720 410spec_fsync(struct vop_fsync_args *ap)
984263bc
MD
411{
412 struct vnode *vp = ap->a_vp;
6bae6177 413 int error;
984263bc
MD
414
415 if (!vn_isdisk(vp, NULL))
416 return (0);
417
984263bc
MD
418 /*
419 * Flush all dirty buffers associated with a block device.
420 */
54078292 421 error = vfsync(vp, ap->a_waitfor, 10000, NOOFFSET, NULL, NULL);
6bae6177 422 return (error);
984263bc
MD
423}
424
50775720
CP
425/*
426 * spec_inactive(struct vnode *a_vp, struct thread *a_td)
427 */
984263bc 428static int
50775720 429spec_inactive(struct vop_inactive_args *ap)
984263bc 430{
984263bc
MD
431 return (0);
432}
433
434/*
435 * Just call the device strategy routine
50775720 436 *
81b5c339 437 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
984263bc
MD
438 */
439static int
50775720 440spec_strategy(struct vop_strategy_args *ap)
984263bc 441{
81b5c339
MD
442 struct bio *bio = ap->a_bio;
443 struct buf *bp = bio->bio_buf;
984263bc
MD
444 struct vnode *vp;
445 struct mount *mp;
446
984263bc
MD
447 if (((bp->b_flags & B_READ) == 0) &&
448 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
449 (*bioops.io_start)(bp);
450
451 /*
452 * Collect statistics on synchronous and asynchronous read
453 * and write counts for disks that have associated filesystems.
454 */
455 vp = ap->a_vp;
72402f22 456 KKASSERT(vp->v_rdev != NULL); /* XXX */
e4c9c0c8 457 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
984263bc 458 if ((bp->b_flags & B_READ) == 0) {
dadab5e9 459 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
460 mp->mnt_stat.f_asyncwrites++;
461 else
462 mp->mnt_stat.f_syncwrites++;
463 } else {
dadab5e9 464 if (bp->b_lock.lk_lockholder == LK_KERNTHREAD)
984263bc
MD
465 mp->mnt_stat.f_asyncreads++;
466 else
467 mp->mnt_stat.f_syncreads++;
468 }
469 }
81b5c339 470 dev_dstrategy_chain(vp->v_rdev, bio);
984263bc
MD
471 return (0);
472}
473
50775720
CP
474/*
475 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
476 */
984263bc 477static int
50775720 478spec_freeblks(struct vop_freeblks_args *ap)
984263bc 479{
984263bc
MD
480 struct buf *bp;
481
482 /*
483 * XXX: This assumes that strategy does the deed right away.
484 * XXX: this may not be TRTTD.
485 */
72402f22 486 KKASSERT(ap->a_vp->v_rdev != NULL);
335dda38 487 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
488 return (0);
489 bp = geteblk(ap->a_length);
490 bp->b_flags |= B_FREEBUF;
54078292 491 bp->b_bio1.bio_offset = ap->a_offset;
984263bc 492 bp->b_bcount = ap->a_length;
81b5c339 493 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
984263bc
MD
494 return (0);
495}
496
497/*
498 * Implement degenerate case where the block requested is the block
499 * returned, and assume that the entire device is contiguous in regards
500 * to the contiguous block range (runp and runb).
50775720 501 *
54078292
MD
502 * spec_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp,
503 * off_t *a_doffsetp, int *a_runp, int *a_runb)
984263bc
MD
504 */
505static int
50775720 506spec_bmap(struct vop_bmap_args *ap)
984263bc
MD
507{
508 struct vnode *vp = ap->a_vp;
984263bc
MD
509
510 if (ap->a_vpp != NULL)
511 *ap->a_vpp = vp;
54078292
MD
512 if (ap->a_doffsetp != NULL)
513 *ap->a_doffsetp = ap->a_loffset;
984263bc 514 if (ap->a_runp != NULL)
54078292
MD
515 *ap->a_runp = MAXBSIZE;
516 if (ap->a_runb != NULL) {
517 if (ap->a_loffset < MAXBSIZE)
518 *ap->a_runb = (int)ap->a_loffset;
519 else
520 *ap->a_runb = MAXBSIZE;
521 }
984263bc
MD
522 return (0);
523}
524
525/*
526 * Device close routine
50775720
CP
527 *
528 * spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
529 * struct thread *a_td)
53994a38
MD
530 *
531 * NOTE: the vnode may or may not be locked on call.
984263bc
MD
532 */
533/* ARGSUSED */
534static int
50775720 535spec_close(struct vop_close_args *ap)
984263bc 536{
dadab5e9 537 struct proc *p = ap->a_td->td_proc;
984263bc 538 struct vnode *vp = ap->a_vp;
984263bc 539 dev_t dev = vp->v_rdev;
e4c9c0c8 540 int error;
53994a38 541 int needrelock;
984263bc
MD
542
543 /*
544 * Hack: a tty device that is a controlling terminal
545 * has a reference from the session structure.
546 * We cannot easily tell that a character device is
547 * a controlling terminal, unless it is the closing
548 * process' controlling terminal. In that case,
549 * if the reference count is 2 (this last descriptor
550 * plus the session), release the reference from the session.
1fbb5fc0
MD
551 *
552 * It is possible for v_opencount to be 0 or 1 in this case, 0
553 * because the tty might have been revoked.
984263bc 554 */
72402f22
MD
555 if (dev)
556 reference_dev(dev);
1fbb5fc0 557 if (vcount(vp) == 2 && vp->v_opencount <= 1 &&
5fd012e0 558 p && vp == p->p_session->s_ttyvp) {
984263bc 559 p->p_session->s_ttyvp = NULL;
1fbb5fc0 560 vrele(vp);
984263bc 561 }
e4c9c0c8 562
984263bc 563 /*
e4c9c0c8
MD
564 * Vnodes can be opened and close multiple times. Do not really
565 * close the device unless (1) it is being closed forcibly,
566 * (2) the device wants to track closes, or (3) this is the last
567 * vnode doing its last close on the device.
568 *
569 * XXX the VXLOCK (force close) case can leave vnodes referencing
570 * a closed device.
984263bc 571 */
5fd012e0 572 if (dev && ((vp->v_flag & VRECLAIMED) ||
e4c9c0c8 573 (dev_dflags(dev) & D_TRACKCLOSE) ||
72402f22 574 (vcount(vp) <= 1 && vp->v_opencount == 1))) {
53994a38
MD
575 needrelock = 0;
576 if (VOP_ISLOCKED(vp, ap->a_td)) {
577 needrelock = 1;
578 VOP_UNLOCK(vp, 0, ap->a_td);
579 }
e4c9c0c8 580 error = dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td);
53994a38
MD
581 if (needrelock)
582 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
e4c9c0c8
MD
583 } else {
584 error = 0;
984263bc 585 }
e4c9c0c8
MD
586
587 /*
588 * Track the actual opens and closes on the vnode. The last close
1fbb5fc0
MD
589 * disassociates the rdev. If the rdev is already disassociated
590 * the vnode might have been revoked and no further opencount
591 * tracking occurs.
e4c9c0c8 592 */
1fbb5fc0
MD
593 if (dev) {
594 KKASSERT(vp->v_opencount > 0);
595 if (dev_ref_debug) {
72402f22
MD
596 printf("spec_close: %s %d\n",
597 dev->si_name, vp->v_opencount - 1);
72402f22 598 }
1fbb5fc0
MD
599 if (--vp->v_opencount == 0)
600 v_release_rdev(vp);
72402f22 601 release_dev(dev);
1fbb5fc0 602 }
e4c9c0c8 603 return(error);
984263bc
MD
604}
605
606/*
607 * Print out the contents of a special device vnode.
50775720
CP
608 *
609 * spec_print(struct vnode *a_vp)
984263bc
MD
610 */
611static int
50775720 612spec_print(struct vop_print_args *ap)
984263bc 613{
984263bc
MD
614 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
615 return (0);
616}
617
618/*
619 * Special device advisory byte-level locks.
50775720
CP
620 *
621 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
622 * struct flock *a_fl, int a_flags)
984263bc
MD
623 */
624/* ARGSUSED */
625static int
50775720 626spec_advlock(struct vop_advlock_args *ap)
984263bc 627{
984263bc
MD
628 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
629}
630
631static void
81b5c339 632spec_getpages_iodone(struct bio *bio)
984263bc 633{
81b5c339
MD
634 bio->bio_buf->b_flags |= B_DONE;
635 wakeup(bio->bio_buf);
984263bc
MD
636}
637
638static int
50775720 639spec_getpages(struct vop_getpages_args *ap)
984263bc
MD
640{
641 vm_offset_t kva;
642 int error;
165dba55 643 int i, pcount, size;
984263bc
MD
644 struct buf *bp;
645 vm_page_t m;
646 vm_ooffset_t offset;
647 int toff, nextoff, nread;
648 struct vnode *vp = ap->a_vp;
649 int blksiz;
650 int gotreqpage;
651
652 error = 0;
653 pcount = round_page(ap->a_count) / PAGE_SIZE;
654
655 /*
656 * Calculate the offset of the transfer and do sanity check.
984263bc
MD
657 */
658 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
659
984263bc
MD
660 /*
661 * Round up physical size for real devices. We cannot round using
662 * v_mount's block size data because v_mount has nothing to do with
663 * the device. i.e. it's usually '/dev'. We need the physical block
664 * size for the device itself.
665 *
e4c9c0c8 666 * We can't use v_rdev->si_mountpoint because it only exists when the
984263bc
MD
667 * block device is mounted. However, we can use v_rdev.
668 */
669
670 if (vn_isdisk(vp, NULL))
671 blksiz = vp->v_rdev->si_bsize_phys;
672 else
673 blksiz = DEV_BSIZE;
674
675 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
676
677 bp = getpbuf(NULL);
678 kva = (vm_offset_t)bp->b_data;
679
680 /*
681 * Map the pages to be read into the kva.
682 */
683 pmap_qenter(kva, ap->a_m, pcount);
684
685 /* Build a minimal buffer header. */
5ff4cfb5 686 bp->b_flags = B_READ;
984263bc
MD
687
688 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
689 pbgetvp(ap->a_vp, bp);
690 bp->b_bcount = size;
691 bp->b_bufsize = size;
692 bp->b_resid = 0;
693 bp->b_runningbufspace = bp->b_bufsize;
694 runningbufspace += bp->b_runningbufspace;
695
54078292 696 bp->b_bio1.bio_offset = offset;
81b5c339
MD
697 bp->b_bio1.bio_done = spec_getpages_iodone;
698
12e4aaff
MD
699 mycpu->gd_cnt.v_vnodein++;
700 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
701
702 /* Do the input. */
81b5c339 703 vn_strategy(ap->a_vp, &bp->b_bio1);
984263bc 704
165dba55 705 crit_enter();
984263bc
MD
706
707 /* We definitely need to be at splbio here. */
26a0694b 708 while ((bp->b_flags & B_DONE) == 0) {
377d4740 709 tsleep(bp, 0, "spread", 0);
26a0694b 710 }
984263bc 711
165dba55 712 crit_exit();
984263bc
MD
713
714 if ((bp->b_flags & B_ERROR) != 0) {
715 if (bp->b_error)
716 error = bp->b_error;
717 else
718 error = EIO;
719 }
720
721 nread = size - bp->b_resid;
722
723 if (nread < ap->a_count) {
724 bzero((caddr_t)kva + nread,
725 ap->a_count - nread);
726 }
727 pmap_qremove(kva, pcount);
728
729
730 gotreqpage = 0;
731 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
732 nextoff = toff + PAGE_SIZE;
733 m = ap->a_m[i];
734
735 m->flags &= ~PG_ZERO;
736
737 if (nextoff <= nread) {
738 m->valid = VM_PAGE_BITS_ALL;
739 vm_page_undirty(m);
740 } else if (toff < nread) {
741 /*
742 * Since this is a VM request, we have to supply the
743 * unaligned offset to allow vm_page_set_validclean()
744 * to zero sub-DEV_BSIZE'd portions of the page.
745 */
746 vm_page_set_validclean(m, 0, nread - toff);
747 } else {
748 m->valid = 0;
749 vm_page_undirty(m);
750 }
751
752 if (i != ap->a_reqpage) {
753 /*
754 * Just in case someone was asking for this page we
755 * now tell them that it is ok to use.
756 */
757 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
758 if (m->valid) {
759 if (m->flags & PG_WANTED) {
760 vm_page_activate(m);
761 } else {
762 vm_page_deactivate(m);
763 }
764 vm_page_wakeup(m);
765 } else {
766 vm_page_free(m);
767 }
768 } else {
769 vm_page_free(m);
770 }
771 } else if (m->valid) {
772 gotreqpage = 1;
773 /*
774 * Since this is a VM request, we need to make the
775 * entire page presentable by zeroing invalid sections.
776 */
777 if (m->valid != VM_PAGE_BITS_ALL)
778 vm_page_zero_invalid(m, FALSE);
779 }
780 }
781 if (!gotreqpage) {
782 m = ap->a_m[ap->a_reqpage];
783 printf(
784 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
81b5c339 785 devtoname(vp->v_rdev), error, bp, bp->b_vp);
984263bc 786 printf(
54078292 787 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
984263bc
MD
788 size, bp->b_resid, ap->a_count, m->valid);
789 printf(
790 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
791 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
792 /*
793 * Free the buffer header back to the swap buffer pool.
794 */
795 relpbuf(bp, NULL);
796 return VM_PAGER_ERROR;
797 }
798 /*
799 * Free the buffer header back to the swap buffer pool.
800 */
801 relpbuf(bp, NULL);
802 return VM_PAGER_OK;
803}