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
34 * $FreeBSD: src/sys/miscfs/specfs/spec_vnops.c,v 1.131.2.4 2001/02/26 04:23:20 jlemon Exp $
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_kqfilter __P((struct vop_kqfilter_args *));
66 static int spec_print __P((struct vop_print_args *));
67 static int spec_read __P((struct vop_read_args *));
68 static int spec_strategy __P((struct vop_strategy_args *));
69 static int spec_write __P((struct vop_write_args *));
71 vop_t **spec_vnodeop_p;
72 static struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
73 { &vop_default_desc, (vop_t *) vop_defaultop },
74 { &vop_access_desc, (vop_t *) vop_ebadf },
75 { &vop_advlock_desc, (vop_t *) spec_advlock },
76 { &vop_bmap_desc, (vop_t *) spec_bmap },
77 { &vop_close_desc, (vop_t *) spec_close },
78 { &vop_create_desc, (vop_t *) vop_panic },
79 { &vop_freeblks_desc, (vop_t *) spec_freeblks },
80 { &vop_fsync_desc, (vop_t *) spec_fsync },
81 { &vop_getpages_desc, (vop_t *) spec_getpages },
82 { &vop_inactive_desc, (vop_t *) spec_inactive },
83 { &vop_ioctl_desc, (vop_t *) spec_ioctl },
84 { &vop_lease_desc, (vop_t *) vop_null },
85 { &vop_link_desc, (vop_t *) vop_panic },
86 { &vop_mkdir_desc, (vop_t *) vop_panic },
87 { &vop_mknod_desc, (vop_t *) vop_panic },
88 { &vop_open_desc, (vop_t *) spec_open },
89 { &vop_pathconf_desc, (vop_t *) vop_stdpathconf },
90 { &vop_poll_desc, (vop_t *) spec_poll },
91 { &vop_kqfilter_desc, (vop_t *) spec_kqfilter },
92 { &vop_print_desc, (vop_t *) spec_print },
93 { &vop_read_desc, (vop_t *) spec_read },
94 { &vop_readdir_desc, (vop_t *) vop_panic },
95 { &vop_readlink_desc, (vop_t *) vop_panic },
96 { &vop_reallocblks_desc, (vop_t *) vop_panic },
97 { &vop_reclaim_desc, (vop_t *) vop_null },
98 { &vop_remove_desc, (vop_t *) vop_panic },
99 { &vop_rename_desc, (vop_t *) vop_panic },
100 { &vop_rmdir_desc, (vop_t *) vop_panic },
101 { &vop_setattr_desc, (vop_t *) vop_ebadf },
102 { &vop_strategy_desc, (vop_t *) spec_strategy },
103 { &vop_symlink_desc, (vop_t *) vop_panic },
104 { &vop_write_desc, (vop_t *) spec_write },
107 static struct vnodeopv_desc spec_vnodeop_opv_desc =
108 { &spec_vnodeop_p, spec_vnodeop_entries };
110 VNODEOP_SET(spec_vnodeop_opv_desc);
114 struct vop_generic_args /* {
115 struct vnodeop_desc *a_desc;
116 <other random data follows, presumably>
119 return (VOCALL(spec_vnodeop_p, ap->a_desc->vdesc_offset, ap));
122 static void spec_getpages_iodone __P((struct buf *bp));
125 * Open a special file.
130 struct vop_open_args /* {
133 struct ucred *a_cred;
137 struct proc *p = ap->a_p;
138 struct vnode *vp = ap->a_vp;
139 dev_t dev = vp->v_rdev;
145 * Don't allow open if fs is mounted -nodev.
147 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
151 if ( (dsw == NULL) || (dsw->d_open == NULL))
154 /* Make this field valid before any I/O in ->d_open */
155 if (!dev->si_iosize_max)
156 dev->si_iosize_max = DFLTPHYS;
159 * XXX: Disks get special billing here, but it is mostly wrong.
160 * XXX: diskpartitions can overlap and the real checks should
161 * XXX: take this into account, and consequently they need to
162 * XXX: live in the diskslicing code. Some checks do.
164 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
165 (ap->a_mode & FWRITE)) {
167 * Never allow opens for write if the device is mounted R/W
169 if (vp->v_specmountpoint != NULL &&
170 !(vp->v_specmountpoint->mnt_flag & MNT_RDONLY))
174 * When running in secure mode, do not allow opens
175 * for writing if the device is mounted
177 if (securelevel >= 1 && vp->v_specmountpoint != NULL)
181 * When running in very secure mode, do not allow
182 * opens for writing of any devices.
184 if (securelevel >= 2)
188 /* XXX: Special casing of ttys for deadfs. Probably redundant */
189 if (dsw->d_flags & D_TTY)
190 vp->v_flag |= VISTTY;
192 VOP_UNLOCK(vp, 0, p);
193 error = (*dsw->d_open)(dev, ap->a_mode, S_IFCHR, p);
194 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
199 if (dsw->d_flags & D_TTY) {
204 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
205 tp->t_stop = nottystop;
210 if (vn_isdisk(vp, NULL)) {
211 if (!dev->si_bsize_phys)
212 dev->si_bsize_phys = DEV_BSIZE;
214 if ((dsw->d_flags & D_DISK) == 0) {
216 if (*cp == '#' && (dsw->d_flags & D_NAGGED) == 0) {
217 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
219 dsw->d_flags |= D_NAGGED;
231 struct vop_read_args /* {
235 struct ucred *a_cred;
249 if (uio->uio_resid == 0)
252 VOP_UNLOCK(vp, 0, p);
253 error = (*devsw(dev)->d_read) (dev, uio, ap->a_ioflag);
254 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
264 struct vop_write_args /* {
268 struct ucred *a_cred;
282 VOP_UNLOCK(vp, 0, p);
283 error = (*devsw(dev)->d_write) (dev, uio, ap->a_ioflag);
284 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
289 * Device ioctl operation.
294 struct vop_ioctl_args /* {
299 struct ucred *a_cred;
305 dev = ap->a_vp->v_rdev;
306 return ((*devsw(dev)->d_ioctl)(dev, ap->a_command,
307 ap->a_data, ap->a_fflag, ap->a_p));
313 struct vop_poll_args /* {
316 struct ucred *a_cred;
322 dev = ap->a_vp->v_rdev;
323 return (*devsw(dev)->d_poll)(dev, ap->a_events, ap->a_p);
329 struct vop_kqfilter_args /* {
336 dev = ap->a_vp->v_rdev;
337 if (devsw(dev)->d_flags & D_KQFILTER)
338 return (*devsw(dev)->d_kqfilter)(dev, ap->a_kn);
343 * Synch buffers associated with a block device
348 struct vop_fsync_args /* {
350 struct ucred *a_cred;
355 struct vnode *vp = ap->a_vp;
359 int maxretry = 10000; /* large, arbitrarily chosen */
361 if (!vn_isdisk(vp, NULL))
366 * MARK/SCAN initialization to avoid infinite loops
369 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
370 bp = TAILQ_NEXT(bp, b_vnbufs)) {
371 bp->b_flags &= ~B_SCANNED;
376 * Flush all dirty buffers associated with a block device.
380 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
381 nbp = TAILQ_NEXT(bp, b_vnbufs);
382 if ((bp->b_flags & B_SCANNED) != 0)
384 bp->b_flags |= B_SCANNED;
385 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
387 if ((bp->b_flags & B_DELWRI) == 0)
388 panic("spec_fsync: not dirty");
389 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
402 * If synchronous the caller expects us to completely resolve all
403 * dirty buffers in the system. Wait for in-progress I/O to
404 * complete (which could include background bitmap writes), then
405 * retry if dirty blocks still exist.
407 if (ap->a_waitfor == MNT_WAIT) {
408 while (vp->v_numoutput) {
409 vp->v_flag |= VBWAIT;
410 (void) tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "spfsyn", 0);
412 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
413 if (--maxretry != 0) {
417 vprint("spec_fsync: giving up on dirty", vp);
426 struct vop_inactive_args /* {
432 VOP_UNLOCK(ap->a_vp, 0, ap->a_p);
437 * Just call the device strategy routine
441 struct vop_strategy_args /* {
451 if (((bp->b_flags & B_READ) == 0) &&
452 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
453 (*bioops.io_start)(bp);
456 * Collect statistics on synchronous and asynchronous read
457 * and write counts for disks that have associated filesystems.
460 if (vn_isdisk(vp, NULL) && (mp = vp->v_specmountpoint) != NULL) {
461 if ((bp->b_flags & B_READ) == 0) {
462 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
463 mp->mnt_stat.f_asyncwrites++;
465 mp->mnt_stat.f_syncwrites++;
467 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
468 mp->mnt_stat.f_asyncreads++;
470 mp->mnt_stat.f_syncreads++;
473 KASSERT(devsw(bp->b_dev) != NULL,
474 ("No devsw on dev %s responsible for buffer %p\n",
475 devtoname(bp->b_dev), bp));
476 KASSERT(devsw(bp->b_dev)->d_strategy != NULL,
477 ("No strategy on dev %s responsible for buffer %p\n",
478 devtoname(bp->b_dev), bp));
485 struct vop_freeblks_args /* {
495 * XXX: This assumes that strategy does the deed right away.
496 * XXX: this may not be TRTTD.
498 bsw = devsw(ap->a_vp->v_rdev);
499 if ((bsw->d_flags & D_CANFREE) == 0)
501 bp = geteblk(ap->a_length);
502 bp->b_flags |= B_FREEBUF;
503 bp->b_dev = ap->a_vp->v_rdev;
504 bp->b_blkno = ap->a_addr;
505 bp->b_offset = dbtob(ap->a_addr);
506 bp->b_bcount = ap->a_length;
512 * Implement degenerate case where the block requested is the block
513 * returned, and assume that the entire device is contiguous in regards
514 * to the contiguous block range (runp and runb).
518 struct vop_bmap_args /* {
521 struct vnode **a_vpp;
527 struct vnode *vp = ap->a_vp;
531 if (ap->a_vpp != NULL)
533 if (ap->a_bnp != NULL)
534 *ap->a_bnp = ap->a_bn;
535 if (vp->v_mount != NULL)
536 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
537 if (ap->a_runp != NULL)
539 if (ap->a_runb != NULL)
545 * Device close routine
550 struct vop_close_args /* {
553 struct ucred *a_cred;
557 struct vnode *vp = ap->a_vp;
558 struct proc *p = ap->a_p;
559 dev_t dev = vp->v_rdev;
562 * Hack: a tty device that is a controlling terminal
563 * has a reference from the session structure.
564 * We cannot easily tell that a character device is
565 * a controlling terminal, unless it is the closing
566 * process' controlling terminal. In that case,
567 * if the reference count is 2 (this last descriptor
568 * plus the session), release the reference from the session.
570 if (vcount(vp) == 2 && p && (vp->v_flag & VXLOCK) == 0 &&
571 vp == p->p_session->s_ttyvp) {
573 p->p_session->s_ttyvp = NULL;
576 * We do not want to really close the device if it
577 * is still in use unless we are trying to close it
578 * forcibly. Since every use (buffer, vnode, swap, cmap)
579 * holds a reference to the vnode, and because we mark
580 * any other vnodes that alias this device, when the
581 * sum of the reference counts on all the aliased
582 * vnodes descends to one, we are on last close.
584 if (vp->v_flag & VXLOCK) {
586 } else if (devsw(dev)->d_flags & D_TRACKCLOSE) {
587 /* Keep device updated on status */
588 } else if (vcount(vp) > 1) {
591 return (devsw(dev)->d_close(dev, ap->a_fflag, S_IFCHR, p));
595 * Print out the contents of a special device vnode.
599 struct vop_print_args /* {
604 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
609 * Special device advisory byte-level locks.
614 struct vop_advlock_args /* {
623 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
627 spec_getpages_iodone(bp)
631 bp->b_flags |= B_DONE;
637 struct vop_getpages_args *ap;
641 int i, pcount, size, s;
646 int toff, nextoff, nread;
647 struct vnode *vp = ap->a_vp;
652 pcount = round_page(ap->a_count) / PAGE_SIZE;
655 * Calculate the offset of the transfer and do sanity check.
656 * FreeBSD currently only supports an 8 TB range due to b_blkno
657 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
660 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
662 #define DADDR_T_BIT (sizeof(daddr_t)*8)
663 #define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
665 if (offset < 0 || offset > OFFSET_MAX) {
666 /* XXX still no %q in kernel. */
667 printf("spec_getpages: preposterous offset 0x%x%08x\n",
668 (u_int)((u_quad_t)offset >> 32),
669 (u_int)(offset & 0xffffffff));
670 return (VM_PAGER_ERROR);
673 blkno = btodb(offset);
676 * Round up physical size for real devices. We cannot round using
677 * v_mount's block size data because v_mount has nothing to do with
678 * the device. i.e. it's usually '/dev'. We need the physical block
679 * size for the device itself.
681 * We can't use v_specmountpoint because it only exists when the
682 * block device is mounted. However, we can use v_rdev.
685 if (vn_isdisk(vp, NULL))
686 blksiz = vp->v_rdev->si_bsize_phys;
690 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
693 kva = (vm_offset_t)bp->b_data;
696 * Map the pages to be read into the kva.
698 pmap_qenter(kva, ap->a_m, pcount);
700 /* Build a minimal buffer header. */
701 bp->b_flags = B_READ | B_CALL;
702 bp->b_iodone = spec_getpages_iodone;
704 /* B_PHYS is not set, but it is nice to fill this in. */
705 bp->b_rcred = bp->b_wcred = curproc->p_ucred;
706 if (bp->b_rcred != NOCRED)
708 if (bp->b_wcred != NOCRED)
711 bp->b_lblkno = blkno;
712 pbgetvp(ap->a_vp, bp);
714 bp->b_bufsize = size;
716 bp->b_runningbufspace = bp->b_bufsize;
717 runningbufspace += bp->b_runningbufspace;
720 cnt.v_vnodepgsin += pcount;
723 VOP_STRATEGY(bp->b_vp, bp);
727 /* We definitely need to be at splbio here. */
728 while ((bp->b_flags & B_DONE) == 0)
729 tsleep(bp, PVM, "spread", 0);
733 if ((bp->b_flags & B_ERROR) != 0) {
740 nread = size - bp->b_resid;
742 if (nread < ap->a_count) {
743 bzero((caddr_t)kva + nread,
744 ap->a_count - nread);
746 pmap_qremove(kva, pcount);
750 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
751 nextoff = toff + PAGE_SIZE;
754 m->flags &= ~PG_ZERO;
756 if (nextoff <= nread) {
757 m->valid = VM_PAGE_BITS_ALL;
759 } else if (toff < nread) {
761 * Since this is a VM request, we have to supply the
762 * unaligned offset to allow vm_page_set_validclean()
763 * to zero sub-DEV_BSIZE'd portions of the page.
765 vm_page_set_validclean(m, 0, nread - toff);
771 if (i != ap->a_reqpage) {
773 * Just in case someone was asking for this page we
774 * now tell them that it is ok to use.
776 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
778 if (m->flags & PG_WANTED) {
781 vm_page_deactivate(m);
790 } else if (m->valid) {
793 * Since this is a VM request, we need to make the
794 * entire page presentable by zeroing invalid sections.
796 if (m->valid != VM_PAGE_BITS_ALL)
797 vm_page_zero_invalid(m, FALSE);
801 m = ap->a_m[ap->a_reqpage];
803 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
804 devtoname(bp->b_dev), error, bp, bp->b_vp);
806 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
807 size, bp->b_resid, ap->a_count, m->valid);
809 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
810 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
812 * Free the buffer header back to the swap buffer pool.
815 return VM_PAGER_ERROR;
818 * Free the buffer header back to the swap buffer pool.