ipstat needs thread2.h for MP stuff and mycpu.
[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 $
41a01a4d 35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.13 2004/03/01 06:33:23 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
116int
117spec_vnoperate(ap)
118 struct vop_generic_args /* {
119 struct vnodeop_desc *a_desc;
120 <other random data follows, presumably>
121 } */ *ap;
122{
123 return (VOCALL(spec_vnodeop_p, ap->a_desc->vdesc_offset, ap));
124}
125
a6ee311a 126static void spec_getpages_iodone (struct buf *bp);
984263bc
MD
127
128/*
129 * Open a special file.
130 */
131/* ARGSUSED */
132static int
133spec_open(ap)
134 struct vop_open_args /* {
135 struct vnode *a_vp;
136 int a_mode;
137 struct ucred *a_cred;
dadab5e9 138 struct thread *a_td;
984263bc
MD
139 } */ *ap;
140{
984263bc
MD
141 struct vnode *vp = ap->a_vp;
142 dev_t dev = vp->v_rdev;
143 int error;
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
335dda38 152 if (dev_dport(dev) == NULL)
984263bc
MD
153 return ENXIO;
154
155 /* Make this field valid before any I/O in ->d_open */
156 if (!dev->si_iosize_max)
157 dev->si_iosize_max = DFLTPHYS;
158
159 /*
160 * XXX: Disks get special billing here, but it is mostly wrong.
161 * XXX: diskpartitions can overlap and the real checks should
162 * XXX: take this into account, and consequently they need to
163 * XXX: live in the diskslicing code. Some checks do.
164 */
165 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
166 (ap->a_mode & FWRITE)) {
167 /*
168 * Never allow opens for write if the device is mounted R/W
169 */
170 if (vp->v_specmountpoint != NULL &&
171 !(vp->v_specmountpoint->mnt_flag & MNT_RDONLY))
172 return (EBUSY);
173
174 /*
175 * When running in secure mode, do not allow opens
176 * for writing if the device is mounted
177 */
178 if (securelevel >= 1 && vp->v_specmountpoint != NULL)
179 return (EPERM);
180
181 /*
182 * When running in very secure mode, do not allow
183 * opens for writing of any devices.
184 */
185 if (securelevel >= 2)
186 return (EPERM);
187 }
188
189 /* XXX: Special casing of ttys for deadfs. Probably redundant */
335dda38 190 if (dev_dflags(dev) & D_TTY)
984263bc
MD
191 vp->v_flag |= VISTTY;
192
41a01a4d 193 VOP_UNLOCK(vp, NULL, 0, ap->a_td);
335dda38 194 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_td);
41a01a4d 195 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
984263bc
MD
196
197 if (error)
198 return (error);
199
335dda38 200 if (dev_dflags(dev) & D_TTY) {
984263bc
MD
201 if (dev->si_tty) {
202 struct tty *tp;
203 tp = dev->si_tty;
204 if (!tp->t_stop) {
205 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
206 tp->t_stop = nottystop;
207 }
208 }
209 }
210
211 if (vn_isdisk(vp, NULL)) {
212 if (!dev->si_bsize_phys)
213 dev->si_bsize_phys = DEV_BSIZE;
214 }
335dda38 215 if ((dev_dflags(dev) & D_DISK) == 0) {
984263bc 216 cp = devtoname(dev);
335dda38 217 if (*cp == '#') {
984263bc 218 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
335dda38 219 dev_dname(dev), cp);
984263bc
MD
220 }
221 }
222 return (error);
223}
224
225/*
226 * Vnode op for read
227 */
228/* ARGSUSED */
229static int
230spec_read(ap)
231 struct vop_read_args /* {
232 struct vnode *a_vp;
233 struct uio *a_uio;
234 int a_ioflag;
235 struct ucred *a_cred;
236 } */ *ap;
237{
238 struct vnode *vp;
dadab5e9 239 struct thread *td;
984263bc
MD
240 struct uio *uio;
241 dev_t dev;
242 int error;
243
244 vp = ap->a_vp;
245 dev = vp->v_rdev;
246 uio = ap->a_uio;
dadab5e9 247 td = uio->uio_td;
984263bc
MD
248
249 if (uio->uio_resid == 0)
250 return (0);
251
41a01a4d 252 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 253 error = dev_dread(dev, uio, ap->a_ioflag);
41a01a4d 254 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
255 return (error);
256}
257
258/*
259 * Vnode op for write
260 */
261/* ARGSUSED */
262static int
263spec_write(ap)
264 struct vop_write_args /* {
265 struct vnode *a_vp;
266 struct uio *a_uio;
267 int a_ioflag;
268 struct ucred *a_cred;
269 } */ *ap;
270{
271 struct vnode *vp;
dadab5e9 272 struct thread *td;
984263bc
MD
273 struct uio *uio;
274 dev_t dev;
275 int error;
276
277 vp = ap->a_vp;
278 dev = vp->v_rdev;
279 uio = ap->a_uio;
dadab5e9 280 td = uio->uio_td;
984263bc 281
41a01a4d 282 VOP_UNLOCK(vp, NULL, 0, td);
335dda38 283 error = dev_dwrite(dev, uio, ap->a_ioflag);
41a01a4d 284 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
285 return (error);
286}
287
288/*
289 * Device ioctl operation.
290 */
291/* ARGSUSED */
292static int
293spec_ioctl(ap)
294 struct vop_ioctl_args /* {
295 struct vnode *a_vp;
296 int a_command;
297 caddr_t a_data;
298 int a_fflag;
299 struct ucred *a_cred;
dadab5e9 300 struct thread *a_td;
984263bc
MD
301 } */ *ap;
302{
303 dev_t dev;
304
305 dev = ap->a_vp->v_rdev;
335dda38
MD
306 return (dev_dioctl(dev, ap->a_command, ap->a_data,
307 ap->a_fflag, ap->a_td));
984263bc
MD
308}
309
310/* ARGSUSED */
311static int
312spec_poll(ap)
313 struct vop_poll_args /* {
314 struct vnode *a_vp;
315 int a_events;
316 struct ucred *a_cred;
dadab5e9 317 struct thread *a_td;
984263bc
MD
318 } */ *ap;
319{
320 dev_t dev;
321
322 dev = ap->a_vp->v_rdev;
335dda38 323 return (dev_dpoll(dev, ap->a_events, ap->a_td));
984263bc
MD
324}
325
326/* ARGSUSED */
327static int
328spec_kqfilter(ap)
329 struct vop_kqfilter_args /* {
330 struct vnode *a_vp;
331 struct knote *a_kn;
332 } */ *ap;
333{
334 dev_t dev;
335
336 dev = ap->a_vp->v_rdev;
335dda38 337 return (dev_dkqfilter(dev, ap->a_kn));
984263bc
MD
338}
339
340/*
341 * Synch buffers associated with a block device
342 */
343/* ARGSUSED */
344static int
345spec_fsync(ap)
346 struct vop_fsync_args /* {
347 struct vnode *a_vp;
348 struct ucred *a_cred;
349 int a_waitfor;
dadab5e9 350 struct thread *a_td;
984263bc
MD
351 } */ *ap;
352{
353 struct vnode *vp = ap->a_vp;
354 struct buf *bp;
355 struct buf *nbp;
356 int s;
357 int maxretry = 10000; /* large, arbitrarily chosen */
358
359 if (!vn_isdisk(vp, NULL))
360 return (0);
361
362loop1:
363 /*
364 * MARK/SCAN initialization to avoid infinite loops
365 */
366 s = splbio();
367 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
368 bp = TAILQ_NEXT(bp, b_vnbufs)) {
369 bp->b_flags &= ~B_SCANNED;
370 }
371 splx(s);
372
373 /*
374 * Flush all dirty buffers associated with a block device.
375 */
376loop2:
377 s = splbio();
378 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
379 nbp = TAILQ_NEXT(bp, b_vnbufs);
380 if ((bp->b_flags & B_SCANNED) != 0)
381 continue;
382 bp->b_flags |= B_SCANNED;
383 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
384 continue;
385 if ((bp->b_flags & B_DELWRI) == 0)
386 panic("spec_fsync: not dirty");
387 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
388 BUF_UNLOCK(bp);
389 vfs_bio_awrite(bp);
390 splx(s);
391 } else {
392 bremfree(bp);
393 splx(s);
394 bawrite(bp);
395 }
396 goto loop2;
397 }
398
399 /*
400 * If synchronous the caller expects us to completely resolve all
401 * dirty buffers in the system. Wait for in-progress I/O to
402 * complete (which could include background bitmap writes), then
403 * retry if dirty blocks still exist.
404 */
405 if (ap->a_waitfor == MNT_WAIT) {
406 while (vp->v_numoutput) {
407 vp->v_flag |= VBWAIT;
377d4740 408 (void) tsleep((caddr_t)&vp->v_numoutput, 0, "spfsyn", 0);
984263bc
MD
409 }
410 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
411 if (--maxretry != 0) {
412 splx(s);
413 goto loop1;
414 }
415 vprint("spec_fsync: giving up on dirty", vp);
416 }
417 }
418 splx(s);
419 return (0);
420}
421
422static int
423spec_inactive(ap)
424 struct vop_inactive_args /* {
425 struct vnode *a_vp;
dadab5e9 426 struct thread *a_td;
984263bc
MD
427 } */ *ap;
428{
429
41a01a4d 430 VOP_UNLOCK(ap->a_vp, NULL, 0, ap->a_td);
984263bc
MD
431 return (0);
432}
433
434/*
435 * Just call the device strategy routine
436 */
437static int
438spec_strategy(ap)
439 struct vop_strategy_args /* {
440 struct vnode *a_vp;
441 struct buf *a_bp;
442 } */ *ap;
443{
444 struct buf *bp;
445 struct vnode *vp;
446 struct mount *mp;
447
448 bp = ap->a_bp;
449 if (((bp->b_flags & B_READ) == 0) &&
450 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
451 (*bioops.io_start)(bp);
452
453 /*
454 * Collect statistics on synchronous and asynchronous read
455 * and write counts for disks that have associated filesystems.
456 */
457 vp = ap->a_vp;
458 if (vn_isdisk(vp, NULL) && (mp = vp->v_specmountpoint) != NULL) {
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 }
335dda38 471#if 0
984263bc
MD
472 KASSERT(devsw(bp->b_dev) != NULL,
473 ("No devsw on dev %s responsible for buffer %p\n",
474 devtoname(bp->b_dev), bp));
475 KASSERT(devsw(bp->b_dev)->d_strategy != NULL,
476 ("No strategy on dev %s responsible for buffer %p\n",
477 devtoname(bp->b_dev), bp));
335dda38 478#endif
984263bc
MD
479 BUF_STRATEGY(bp, 0);
480 return (0);
481}
482
483static int
484spec_freeblks(ap)
485 struct vop_freeblks_args /* {
486 struct vnode *a_vp;
487 daddr_t a_addr;
488 daddr_t a_length;
489 } */ *ap;
490{
984263bc
MD
491 struct buf *bp;
492
493 /*
494 * XXX: This assumes that strategy does the deed right away.
495 * XXX: this may not be TRTTD.
496 */
335dda38 497 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
984263bc
MD
498 return (0);
499 bp = geteblk(ap->a_length);
500 bp->b_flags |= B_FREEBUF;
501 bp->b_dev = ap->a_vp->v_rdev;
502 bp->b_blkno = ap->a_addr;
503 bp->b_offset = dbtob(ap->a_addr);
504 bp->b_bcount = ap->a_length;
505 BUF_STRATEGY(bp, 0);
506 return (0);
507}
508
509/*
510 * Implement degenerate case where the block requested is the block
511 * returned, and assume that the entire device is contiguous in regards
512 * to the contiguous block range (runp and runb).
513 */
514static int
515spec_bmap(ap)
516 struct vop_bmap_args /* {
517 struct vnode *a_vp;
518 daddr_t a_bn;
519 struct vnode **a_vpp;
520 daddr_t *a_bnp;
521 int *a_runp;
522 int *a_runb;
523 } */ *ap;
524{
525 struct vnode *vp = ap->a_vp;
526 int runp = 0;
527 int runb = 0;
528
529 if (ap->a_vpp != NULL)
530 *ap->a_vpp = vp;
531 if (ap->a_bnp != NULL)
532 *ap->a_bnp = ap->a_bn;
533 if (vp->v_mount != NULL)
534 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
535 if (ap->a_runp != NULL)
536 *ap->a_runp = runp;
537 if (ap->a_runb != NULL)
538 *ap->a_runb = runb;
539 return (0);
540}
541
542/*
543 * Device close routine
544 */
545/* ARGSUSED */
546static int
547spec_close(ap)
548 struct vop_close_args /* {
549 struct vnode *a_vp;
550 int a_fflag;
551 struct ucred *a_cred;
dadab5e9 552 struct thread *a_td;
984263bc
MD
553 } */ *ap;
554{
dadab5e9 555 struct proc *p = ap->a_td->td_proc;
984263bc 556 struct vnode *vp = ap->a_vp;
984263bc
MD
557 dev_t dev = vp->v_rdev;
558
559 /*
560 * Hack: a tty device that is a controlling terminal
561 * has a reference from the session structure.
562 * We cannot easily tell that a character device is
563 * a controlling terminal, unless it is the closing
564 * process' controlling terminal. In that case,
565 * if the reference count is 2 (this last descriptor
566 * plus the session), release the reference from the session.
567 */
568 if (vcount(vp) == 2 && p && (vp->v_flag & VXLOCK) == 0 &&
569 vp == p->p_session->s_ttyvp) {
570 vrele(vp);
571 p->p_session->s_ttyvp = NULL;
572 }
573 /*
574 * We do not want to really close the device if it
575 * is still in use unless we are trying to close it
576 * forcibly. Since every use (buffer, vnode, swap, cmap)
577 * holds a reference to the vnode, and because we mark
578 * any other vnodes that alias this device, when the
579 * sum of the reference counts on all the aliased
580 * vnodes descends to one, we are on last close.
581 */
582 if (vp->v_flag & VXLOCK) {
583 /* Forced close */
335dda38 584 } else if (dev_dflags(dev) & D_TRACKCLOSE) {
984263bc
MD
585 /* Keep device updated on status */
586 } else if (vcount(vp) > 1) {
587 return (0);
588 }
335dda38 589 return (dev_dclose(dev, ap->a_fflag, S_IFCHR, ap->a_td));
984263bc
MD
590}
591
592/*
593 * Print out the contents of a special device vnode.
594 */
595static int
596spec_print(ap)
597 struct vop_print_args /* {
598 struct vnode *a_vp;
599 } */ *ap;
600{
601
602 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
603 return (0);
604}
605
606/*
607 * Special device advisory byte-level locks.
608 */
609/* ARGSUSED */
610static int
611spec_advlock(ap)
612 struct vop_advlock_args /* {
613 struct vnode *a_vp;
614 caddr_t a_id;
615 int a_op;
616 struct flock *a_fl;
617 int a_flags;
618 } */ *ap;
619{
620
621 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
622}
623
624static void
625spec_getpages_iodone(bp)
626 struct buf *bp;
627{
628
629 bp->b_flags |= B_DONE;
630 wakeup(bp);
631}
632
633static int
634spec_getpages(ap)
635 struct vop_getpages_args *ap;
636{
637 vm_offset_t kva;
638 int error;
639 int i, pcount, size, s;
640 daddr_t blkno;
641 struct buf *bp;
642 vm_page_t m;
643 vm_ooffset_t offset;
644 int toff, nextoff, nread;
645 struct vnode *vp = ap->a_vp;
646 int blksiz;
647 int gotreqpage;
648
649 error = 0;
650 pcount = round_page(ap->a_count) / PAGE_SIZE;
651
652 /*
653 * Calculate the offset of the transfer and do sanity check.
654 * FreeBSD currently only supports an 8 TB range due to b_blkno
655 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
656 * VOP_STRATEGY. XXX
657 */
658 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
659
660#define DADDR_T_BIT (sizeof(daddr_t)*8)
661#define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
662
663 if (offset < 0 || offset > OFFSET_MAX) {
664 /* XXX still no %q in kernel. */
665 printf("spec_getpages: preposterous offset 0x%x%08x\n",
666 (u_int)((u_quad_t)offset >> 32),
667 (u_int)(offset & 0xffffffff));
668 return (VM_PAGER_ERROR);
669 }
670
671 blkno = btodb(offset);
672
673 /*
674 * Round up physical size for real devices. We cannot round using
675 * v_mount's block size data because v_mount has nothing to do with
676 * the device. i.e. it's usually '/dev'. We need the physical block
677 * size for the device itself.
678 *
679 * We can't use v_specmountpoint because it only exists when the
680 * block device is mounted. However, we can use v_rdev.
681 */
682
683 if (vn_isdisk(vp, NULL))
684 blksiz = vp->v_rdev->si_bsize_phys;
685 else
686 blksiz = DEV_BSIZE;
687
688 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
689
690 bp = getpbuf(NULL);
691 kva = (vm_offset_t)bp->b_data;
692
693 /*
694 * Map the pages to be read into the kva.
695 */
696 pmap_qenter(kva, ap->a_m, pcount);
697
698 /* Build a minimal buffer header. */
699 bp->b_flags = B_READ | B_CALL;
700 bp->b_iodone = spec_getpages_iodone;
701
702 /* B_PHYS is not set, but it is nice to fill this in. */
984263bc
MD
703 bp->b_blkno = blkno;
704 bp->b_lblkno = blkno;
705 pbgetvp(ap->a_vp, bp);
706 bp->b_bcount = size;
707 bp->b_bufsize = size;
708 bp->b_resid = 0;
709 bp->b_runningbufspace = bp->b_bufsize;
710 runningbufspace += bp->b_runningbufspace;
711
12e4aaff
MD
712 mycpu->gd_cnt.v_vnodein++;
713 mycpu->gd_cnt.v_vnodepgsin += pcount;
984263bc
MD
714
715 /* Do the input. */
716 VOP_STRATEGY(bp->b_vp, bp);
717
718 s = splbio();
719
720 /* We definitely need to be at splbio here. */
26a0694b 721 while ((bp->b_flags & B_DONE) == 0) {
377d4740 722 tsleep(bp, 0, "spread", 0);
26a0694b 723 }
984263bc
MD
724
725 splx(s);
726
727 if ((bp->b_flags & B_ERROR) != 0) {
728 if (bp->b_error)
729 error = bp->b_error;
730 else
731 error = EIO;
732 }
733
734 nread = size - bp->b_resid;
735
736 if (nread < ap->a_count) {
737 bzero((caddr_t)kva + nread,
738 ap->a_count - nread);
739 }
740 pmap_qremove(kva, pcount);
741
742
743 gotreqpage = 0;
744 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
745 nextoff = toff + PAGE_SIZE;
746 m = ap->a_m[i];
747
748 m->flags &= ~PG_ZERO;
749
750 if (nextoff <= nread) {
751 m->valid = VM_PAGE_BITS_ALL;
752 vm_page_undirty(m);
753 } else if (toff < nread) {
754 /*
755 * Since this is a VM request, we have to supply the
756 * unaligned offset to allow vm_page_set_validclean()
757 * to zero sub-DEV_BSIZE'd portions of the page.
758 */
759 vm_page_set_validclean(m, 0, nread - toff);
760 } else {
761 m->valid = 0;
762 vm_page_undirty(m);
763 }
764
765 if (i != ap->a_reqpage) {
766 /*
767 * Just in case someone was asking for this page we
768 * now tell them that it is ok to use.
769 */
770 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
771 if (m->valid) {
772 if (m->flags & PG_WANTED) {
773 vm_page_activate(m);
774 } else {
775 vm_page_deactivate(m);
776 }
777 vm_page_wakeup(m);
778 } else {
779 vm_page_free(m);
780 }
781 } else {
782 vm_page_free(m);
783 }
784 } else if (m->valid) {
785 gotreqpage = 1;
786 /*
787 * Since this is a VM request, we need to make the
788 * entire page presentable by zeroing invalid sections.
789 */
790 if (m->valid != VM_PAGE_BITS_ALL)
791 vm_page_zero_invalid(m, FALSE);
792 }
793 }
794 if (!gotreqpage) {
795 m = ap->a_m[ap->a_reqpage];
796 printf(
797 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
798 devtoname(bp->b_dev), error, bp, bp->b_vp);
799 printf(
800 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
801 size, bp->b_resid, ap->a_count, m->valid);
802 printf(
803 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
804 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
805 /*
806 * Free the buffer header back to the swap buffer pool.
807 */
808 relpbuf(bp, NULL);
809 return VM_PAGER_ERROR;
810 }
811 /*
812 * Free the buffer header back to the swap buffer pool.
813 */
814 relpbuf(bp, NULL);
815 return VM_PAGER_OK;
816}