2 * Copyright (c) 1989, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
33 * @(#)spec_vnops.c 8.14 (Berkeley) 5/21/95
37 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
43 #include <sys/mount.h>
44 #include <sys/vnode.h>
46 #include <sys/fcntl.h>
47 #include <sys/vmmeter.h>
51 #include <vm/vm_object.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_pager.h>
55 static int spec_advlock __P((struct vop_advlock_args *));
56 static int spec_bmap __P((struct vop_bmap_args *));
57 static int spec_close __P((struct vop_close_args *));
58 static int spec_freeblks __P((struct vop_freeblks_args *));
59 static int spec_fsync __P((struct vop_fsync_args *));
60 static int spec_getpages __P((struct vop_getpages_args *));
61 static int spec_inactive __P((struct vop_inactive_args *));
62 static int spec_ioctl __P((struct vop_ioctl_args *));
63 static int spec_open __P((struct vop_open_args *));
64 static int spec_poll __P((struct vop_poll_args *));
65 static int spec_print __P((struct vop_print_args *));
66 static int spec_read __P((struct vop_read_args *));
67 static int spec_strategy __P((struct vop_strategy_args *));
68 static int spec_write __P((struct vop_write_args *));
70 vop_t **spec_vnodeop_p;
71 static struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
72 { &vop_default_desc, (vop_t *) vop_defaultop },
73 { &vop_access_desc, (vop_t *) vop_ebadf },
74 { &vop_advlock_desc, (vop_t *) spec_advlock },
75 { &vop_bmap_desc, (vop_t *) spec_bmap },
76 { &vop_close_desc, (vop_t *) spec_close },
77 { &vop_create_desc, (vop_t *) vop_panic },
78 { &vop_freeblks_desc, (vop_t *) spec_freeblks },
79 { &vop_fsync_desc, (vop_t *) spec_fsync },
80 { &vop_getpages_desc, (vop_t *) spec_getpages },
81 { &vop_inactive_desc, (vop_t *) spec_inactive },
82 { &vop_ioctl_desc, (vop_t *) spec_ioctl },
83 { &vop_lease_desc, (vop_t *) vop_null },
84 { &vop_link_desc, (vop_t *) vop_panic },
85 { &vop_mkdir_desc, (vop_t *) vop_panic },
86 { &vop_mknod_desc, (vop_t *) vop_panic },
87 { &vop_open_desc, (vop_t *) spec_open },
88 { &vop_pathconf_desc, (vop_t *) vop_stdpathconf },
89 { &vop_poll_desc, (vop_t *) spec_poll },
90 { &vop_print_desc, (vop_t *) spec_print },
91 { &vop_read_desc, (vop_t *) spec_read },
92 { &vop_readdir_desc, (vop_t *) vop_panic },
93 { &vop_readlink_desc, (vop_t *) vop_panic },
94 { &vop_reallocblks_desc, (vop_t *) vop_panic },
95 { &vop_reclaim_desc, (vop_t *) vop_null },
96 { &vop_remove_desc, (vop_t *) vop_panic },
97 { &vop_rename_desc, (vop_t *) vop_panic },
98 { &vop_rmdir_desc, (vop_t *) vop_panic },
99 { &vop_setattr_desc, (vop_t *) vop_ebadf },
100 { &vop_strategy_desc, (vop_t *) spec_strategy },
101 { &vop_symlink_desc, (vop_t *) vop_panic },
102 { &vop_write_desc, (vop_t *) spec_write },
105 static struct vnodeopv_desc spec_vnodeop_opv_desc =
106 { &spec_vnodeop_p, spec_vnodeop_entries };
108 VNODEOP_SET(spec_vnodeop_opv_desc);
112 struct vop_generic_args /* {
113 struct vnodeop_desc *a_desc;
114 <other random data follows, presumably>
117 return (VOCALL(spec_vnodeop_p, ap->a_desc->vdesc_offset, ap));
120 static void spec_getpages_iodone __P((struct buf *bp));
123 * Open a special file.
128 struct vop_open_args /* {
131 struct ucred *a_cred;
135 struct proc *p = ap->a_p;
136 struct vnode *vp = ap->a_vp;
137 dev_t dev = vp->v_rdev;
143 * Don't allow open if fs is mounted -nodev.
145 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
149 if ( (dsw == NULL) || (dsw->d_open == NULL))
152 /* Make this field valid before any I/O in ->d_open */
153 if (!dev->si_iosize_max)
154 dev->si_iosize_max = DFLTPHYS;
157 * XXX: Disks get special billing here, but it is mostly wrong.
158 * XXX: diskpartitions can overlap and the real checks should
159 * XXX: take this into account, and consequently they need to
160 * XXX: live in the diskslicing code. Some checks do.
162 if (vn_isdisk(vp) && ap->a_cred != FSCRED && (ap->a_mode & FWRITE)) {
164 * Never allow opens for write if the device is mounted R/W
166 if (vp->v_specmountpoint != NULL &&
167 !(vp->v_specmountpoint->mnt_flag & MNT_RDONLY))
171 * When running in secure mode, do not allow opens
172 * for writing if the device is mounted
174 if (securelevel >= 1 && vp->v_specmountpoint != NULL)
178 * When running in very secure mode, do not allow
179 * opens for writing of any devices.
181 if (securelevel >= 2)
185 /* XXX: Special casing of ttys for deadfs. Probably redundant */
186 if (dsw->d_flags & D_TTY)
187 vp->v_flag |= VISTTY;
189 VOP_UNLOCK(vp, 0, p);
190 error = (*dsw->d_open)(dev, ap->a_mode, S_IFCHR, p);
191 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
196 if (dsw->d_flags & D_TTY) {
201 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
202 tp->t_stop = nottystop;
208 if (!dev->si_bsize_phys)
209 dev->si_bsize_phys = DEV_BSIZE;
211 if ((dsw->d_flags & D_DISK) == 0) {
213 if (*cp == '#' && (dsw->d_flags & D_NAGGED) == 0) {
214 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
216 dsw->d_flags |= D_NAGGED;
228 struct vop_read_args /* {
232 struct ucred *a_cred;
246 if (uio->uio_resid == 0)
249 VOP_UNLOCK(vp, 0, p);
250 error = (*devsw(dev)->d_read) (dev, uio, ap->a_ioflag);
251 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
261 struct vop_write_args /* {
265 struct ucred *a_cred;
279 VOP_UNLOCK(vp, 0, p);
280 error = (*devsw(dev)->d_write) (dev, uio, ap->a_ioflag);
281 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
286 * Device ioctl operation.
291 struct vop_ioctl_args /* {
296 struct ucred *a_cred;
302 dev = ap->a_vp->v_rdev;
303 return ((*devsw(dev)->d_ioctl)(dev, ap->a_command,
304 ap->a_data, ap->a_fflag, ap->a_p));
310 struct vop_poll_args /* {
313 struct ucred *a_cred;
319 dev = ap->a_vp->v_rdev;
320 return (*devsw(dev)->d_poll)(dev, ap->a_events, ap->a_p);
323 * Synch buffers associated with a block device
328 struct vop_fsync_args /* {
330 struct ucred *a_cred;
335 struct vnode *vp = ap->a_vp;
344 * Flush all dirty buffers associated with a block device.
348 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
349 nbp = TAILQ_NEXT(bp, b_vnbufs);
350 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
352 if ((bp->b_flags & B_DELWRI) == 0)
353 panic("spec_fsync: not dirty");
354 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
365 if (ap->a_waitfor == MNT_WAIT) {
366 while (vp->v_numoutput) {
367 vp->v_flag |= VBWAIT;
368 (void) tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "spfsyn", 0);
371 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
372 vprint("spec_fsync: dirty", vp);
384 struct vop_inactive_args /* {
390 VOP_UNLOCK(ap->a_vp, 0, ap->a_p);
395 * Just call the device strategy routine
399 struct vop_strategy_args /* {
409 if (((bp->b_flags & B_READ) == 0) &&
410 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
411 (*bioops.io_start)(bp);
414 * Collect statistics on synchronous and asynchronous read
415 * and write counts for disks that have associated filesystems.
418 if (vn_isdisk(vp) && (mp = vp->v_specmountpoint) != NULL) {
419 if ((bp->b_flags & B_READ) == 0) {
420 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
421 mp->mnt_stat.f_asyncwrites++;
423 mp->mnt_stat.f_syncwrites++;
425 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
426 mp->mnt_stat.f_asyncreads++;
428 mp->mnt_stat.f_syncreads++;
431 KASSERT(devsw(bp->b_dev) != NULL,
432 ("No devsw on dev %s responsible for buffer %p\n",
433 devtoname(bp->b_dev), bp));
434 KASSERT(devsw(bp->b_dev)->d_strategy != NULL,
435 ("No strategy on dev %s responsible for buffer %p\n",
436 devtoname(bp->b_dev), bp));
443 struct vop_freeblks_args /* {
453 * XXX: This assumes that strategy does the deed right away.
454 * XXX: this may not be TRTTD.
456 bsw = devsw(ap->a_vp->v_rdev);
457 if ((bsw->d_flags & D_CANFREE) == 0)
459 bp = geteblk(ap->a_length);
460 bp->b_flags |= B_FREEBUF;
461 bp->b_dev = ap->a_vp->v_rdev;
462 bp->b_blkno = ap->a_addr;
463 bp->b_offset = dbtob(ap->a_addr);
464 bp->b_bcount = ap->a_length;
470 * Implement degenerate case where the block requested is the block
471 * returned, and assume that the entire device is contiguous in regards
472 * to the contiguous block range (runp and runb).
476 struct vop_bmap_args /* {
479 struct vnode **a_vpp;
485 struct vnode *vp = ap->a_vp;
489 if (ap->a_vpp != NULL)
491 if (ap->a_bnp != NULL)
492 *ap->a_bnp = ap->a_bn;
493 if (vp->v_mount != NULL)
494 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
495 if (ap->a_runp != NULL)
497 if (ap->a_runb != NULL)
503 * Device close routine
508 struct vop_close_args /* {
511 struct ucred *a_cred;
515 struct vnode *vp = ap->a_vp;
516 struct proc *p = ap->a_p;
517 dev_t dev = vp->v_rdev;
520 * Hack: a tty device that is a controlling terminal
521 * has a reference from the session structure.
522 * We cannot easily tell that a character device is
523 * a controlling terminal, unless it is the closing
524 * process' controlling terminal. In that case,
525 * if the reference count is 2 (this last descriptor
526 * plus the session), release the reference from the session.
528 if (vcount(vp) == 2 && p && (vp->v_flag & VXLOCK) == 0 &&
529 vp == p->p_session->s_ttyvp) {
531 p->p_session->s_ttyvp = NULL;
534 * We do not want to really close the device if it
535 * is still in use unless we are trying to close it
536 * forcibly. Since every use (buffer, vnode, swap, cmap)
537 * holds a reference to the vnode, and because we mark
538 * any other vnodes that alias this device, when the
539 * sum of the reference counts on all the aliased
540 * vnodes descends to one, we are on last close.
542 if (vp->v_flag & VXLOCK) {
544 } else if (devsw(dev)->d_flags & D_TRACKCLOSE) {
545 /* Keep device updated on status */
546 } else if (vcount(vp) > 1) {
549 return (devsw(dev)->d_close(dev, ap->a_fflag, S_IFCHR, p));
553 * Print out the contents of a special device vnode.
557 struct vop_print_args /* {
562 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
567 * Special device advisory byte-level locks.
572 struct vop_advlock_args /* {
581 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
585 spec_getpages_iodone(bp)
589 bp->b_flags |= B_DONE;
595 struct vop_getpages_args *ap;
599 int i, pcount, size, s;
604 int toff, nextoff, nread;
605 struct vnode *vp = ap->a_vp;
610 pcount = round_page(ap->a_count) / PAGE_SIZE;
613 * Calculate the offset of the transfer and do sanity check.
614 * FreeBSD currently only supports an 8 TB range due to b_blkno
615 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
618 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
620 #define DADDR_T_BIT (sizeof(daddr_t)*8)
621 #define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
623 if (offset < 0 || offset > OFFSET_MAX) {
624 /* XXX still no %q in kernel. */
625 printf("spec_getpages: preposterous offset 0x%x%08x\n",
626 (u_int)((u_quad_t)offset >> 32),
627 (u_int)(offset & 0xffffffff));
628 return (VM_PAGER_ERROR);
631 blkno = btodb(offset);
634 * Round up physical size for real devices. We cannot round using
635 * v_mount's block size data because v_mount has nothing to do with
636 * the device. i.e. it's usually '/dev'. We need the physical block
637 * size for the device itself.
639 * We can't use v_specmountpoint because it only exists when the
640 * block device is mounted. However, we can use v_rdev.
644 blksiz = vp->v_rdev->si_bsize_phys;
648 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
651 kva = (vm_offset_t)bp->b_data;
654 * Map the pages to be read into the kva.
656 pmap_qenter(kva, ap->a_m, pcount);
658 /* Build a minimal buffer header. */
659 bp->b_flags = B_READ | B_CALL;
660 bp->b_iodone = spec_getpages_iodone;
662 /* B_PHYS is not set, but it is nice to fill this in. */
663 bp->b_rcred = bp->b_wcred = curproc->p_ucred;
664 if (bp->b_rcred != NOCRED)
666 if (bp->b_wcred != NOCRED)
669 bp->b_lblkno = blkno;
670 pbgetvp(ap->a_vp, bp);
672 bp->b_bufsize = size;
676 cnt.v_vnodepgsin += pcount;
679 VOP_STRATEGY(bp->b_vp, bp);
683 /* We definitely need to be at splbio here. */
684 while ((bp->b_flags & B_DONE) == 0)
685 tsleep(bp, PVM, "spread", 0);
689 if ((bp->b_flags & B_ERROR) != 0) {
696 nread = size - bp->b_resid;
698 if (nread < ap->a_count) {
699 bzero((caddr_t)kva + nread,
700 ap->a_count - nread);
702 pmap_qremove(kva, pcount);
706 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
707 nextoff = toff + PAGE_SIZE;
710 m->flags &= ~PG_ZERO;
712 if (nextoff <= nread) {
713 m->valid = VM_PAGE_BITS_ALL;
715 } else if (toff < nread) {
717 * Since this is a VM request, we have to supply the
718 * unaligned offset to allow vm_page_set_validclean()
719 * to zero sub-DEV_BSIZE'd portions of the page.
721 vm_page_set_validclean(m, 0, nread - toff);
727 if (i != ap->a_reqpage) {
729 * Just in case someone was asking for this page we
730 * now tell them that it is ok to use.
732 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
734 if (m->flags & PG_WANTED) {
737 vm_page_deactivate(m);
746 } else if (m->valid) {
749 * Since this is a VM request, we need to make the
750 * entire page presentable by zeroing invalid sections.
752 if (m->valid != VM_PAGE_BITS_ALL)
753 vm_page_zero_invalid(m, FALSE);
757 m = ap->a_m[ap->a_reqpage];
760 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
761 devtoname(bp->b_dev), error, bp, bp->b_vp);
763 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
764 size, bp->b_resid, ap->a_count, m->valid);
766 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
767 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
770 * Free the buffer header back to the swap buffer pool.
773 return VM_PAGER_ERROR;
776 * Free the buffer header back to the swap buffer pool.