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 $
35 * $DragonFly: src/sys/vfs/specfs/spec_vnops.c,v 1.3 2003/06/19 01:55:06 dillon Exp $
38 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
44 #include <sys/mount.h>
45 #include <sys/vnode.h>
47 #include <sys/fcntl.h>
48 #include <sys/vmmeter.h>
52 #include <vm/vm_object.h>
53 #include <vm/vm_page.h>
54 #include <vm/vm_pager.h>
58 static int spec_advlock __P((struct vop_advlock_args *));
59 static int spec_bmap __P((struct vop_bmap_args *));
60 static int spec_close __P((struct vop_close_args *));
61 static int spec_freeblks __P((struct vop_freeblks_args *));
62 static int spec_fsync __P((struct vop_fsync_args *));
63 static int spec_getpages __P((struct vop_getpages_args *));
64 static int spec_inactive __P((struct vop_inactive_args *));
65 static int spec_ioctl __P((struct vop_ioctl_args *));
66 static int spec_open __P((struct vop_open_args *));
67 static int spec_poll __P((struct vop_poll_args *));
68 static int spec_kqfilter __P((struct vop_kqfilter_args *));
69 static int spec_print __P((struct vop_print_args *));
70 static int spec_read __P((struct vop_read_args *));
71 static int spec_strategy __P((struct vop_strategy_args *));
72 static int spec_write __P((struct vop_write_args *));
74 vop_t **spec_vnodeop_p;
75 static struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
76 { &vop_default_desc, (vop_t *) vop_defaultop },
77 { &vop_access_desc, (vop_t *) vop_ebadf },
78 { &vop_advlock_desc, (vop_t *) spec_advlock },
79 { &vop_bmap_desc, (vop_t *) spec_bmap },
80 { &vop_close_desc, (vop_t *) spec_close },
81 { &vop_create_desc, (vop_t *) vop_panic },
82 { &vop_freeblks_desc, (vop_t *) spec_freeblks },
83 { &vop_fsync_desc, (vop_t *) spec_fsync },
84 { &vop_getpages_desc, (vop_t *) spec_getpages },
85 { &vop_inactive_desc, (vop_t *) spec_inactive },
86 { &vop_ioctl_desc, (vop_t *) spec_ioctl },
87 { &vop_lease_desc, (vop_t *) vop_null },
88 { &vop_link_desc, (vop_t *) vop_panic },
89 { &vop_mkdir_desc, (vop_t *) vop_panic },
90 { &vop_mknod_desc, (vop_t *) vop_panic },
91 { &vop_open_desc, (vop_t *) spec_open },
92 { &vop_pathconf_desc, (vop_t *) vop_stdpathconf },
93 { &vop_poll_desc, (vop_t *) spec_poll },
94 { &vop_kqfilter_desc, (vop_t *) spec_kqfilter },
95 { &vop_print_desc, (vop_t *) spec_print },
96 { &vop_read_desc, (vop_t *) spec_read },
97 { &vop_readdir_desc, (vop_t *) vop_panic },
98 { &vop_readlink_desc, (vop_t *) vop_panic },
99 { &vop_reallocblks_desc, (vop_t *) vop_panic },
100 { &vop_reclaim_desc, (vop_t *) vop_null },
101 { &vop_remove_desc, (vop_t *) vop_panic },
102 { &vop_rename_desc, (vop_t *) vop_panic },
103 { &vop_rmdir_desc, (vop_t *) vop_panic },
104 { &vop_setattr_desc, (vop_t *) vop_ebadf },
105 { &vop_strategy_desc, (vop_t *) spec_strategy },
106 { &vop_symlink_desc, (vop_t *) vop_panic },
107 { &vop_write_desc, (vop_t *) spec_write },
110 static struct vnodeopv_desc spec_vnodeop_opv_desc =
111 { &spec_vnodeop_p, spec_vnodeop_entries };
113 VNODEOP_SET(spec_vnodeop_opv_desc);
117 struct vop_generic_args /* {
118 struct vnodeop_desc *a_desc;
119 <other random data follows, presumably>
122 return (VOCALL(spec_vnodeop_p, ap->a_desc->vdesc_offset, ap));
125 static void spec_getpages_iodone __P((struct buf *bp));
128 * Open a special file.
133 struct vop_open_args /* {
136 struct ucred *a_cred;
140 struct proc *p = ap->a_p;
141 struct vnode *vp = ap->a_vp;
142 dev_t dev = vp->v_rdev;
148 * Don't allow open if fs is mounted -nodev.
150 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
154 if ( (dsw == NULL) || (dsw->d_open == NULL))
157 /* Make this field valid before any I/O in ->d_open */
158 if (!dev->si_iosize_max)
159 dev->si_iosize_max = DFLTPHYS;
162 * XXX: Disks get special billing here, but it is mostly wrong.
163 * XXX: diskpartitions can overlap and the real checks should
164 * XXX: take this into account, and consequently they need to
165 * XXX: live in the diskslicing code. Some checks do.
167 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
168 (ap->a_mode & FWRITE)) {
170 * Never allow opens for write if the device is mounted R/W
172 if (vp->v_specmountpoint != NULL &&
173 !(vp->v_specmountpoint->mnt_flag & MNT_RDONLY))
177 * When running in secure mode, do not allow opens
178 * for writing if the device is mounted
180 if (securelevel >= 1 && vp->v_specmountpoint != NULL)
184 * When running in very secure mode, do not allow
185 * opens for writing of any devices.
187 if (securelevel >= 2)
191 /* XXX: Special casing of ttys for deadfs. Probably redundant */
192 if (dsw->d_flags & D_TTY)
193 vp->v_flag |= VISTTY;
195 VOP_UNLOCK(vp, 0, p);
196 error = (*dsw->d_open)(dev, ap->a_mode, S_IFCHR, p);
197 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
202 if (dsw->d_flags & D_TTY) {
207 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
208 tp->t_stop = nottystop;
213 if (vn_isdisk(vp, NULL)) {
214 if (!dev->si_bsize_phys)
215 dev->si_bsize_phys = DEV_BSIZE;
217 if ((dsw->d_flags & D_DISK) == 0) {
219 if (*cp == '#' && (dsw->d_flags & D_NAGGED) == 0) {
220 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
222 dsw->d_flags |= D_NAGGED;
234 struct vop_read_args /* {
238 struct ucred *a_cred;
252 if (uio->uio_resid == 0)
255 VOP_UNLOCK(vp, 0, p);
256 error = (*devsw(dev)->d_read) (dev, uio, ap->a_ioflag);
257 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
267 struct vop_write_args /* {
271 struct ucred *a_cred;
285 VOP_UNLOCK(vp, 0, p);
286 error = (*devsw(dev)->d_write) (dev, uio, ap->a_ioflag);
287 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
292 * Device ioctl operation.
297 struct vop_ioctl_args /* {
302 struct ucred *a_cred;
308 dev = ap->a_vp->v_rdev;
309 return ((*devsw(dev)->d_ioctl)(dev, ap->a_command,
310 ap->a_data, ap->a_fflag, ap->a_p));
316 struct vop_poll_args /* {
319 struct ucred *a_cred;
325 dev = ap->a_vp->v_rdev;
326 return (*devsw(dev)->d_poll)(dev, ap->a_events, ap->a_p);
332 struct vop_kqfilter_args /* {
339 dev = ap->a_vp->v_rdev;
340 if (devsw(dev)->d_flags & D_KQFILTER)
341 return (*devsw(dev)->d_kqfilter)(dev, ap->a_kn);
346 * Synch buffers associated with a block device
351 struct vop_fsync_args /* {
353 struct ucred *a_cred;
358 struct vnode *vp = ap->a_vp;
362 int maxretry = 10000; /* large, arbitrarily chosen */
364 if (!vn_isdisk(vp, NULL))
369 * MARK/SCAN initialization to avoid infinite loops
372 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
373 bp = TAILQ_NEXT(bp, b_vnbufs)) {
374 bp->b_flags &= ~B_SCANNED;
379 * Flush all dirty buffers associated with a block device.
383 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
384 nbp = TAILQ_NEXT(bp, b_vnbufs);
385 if ((bp->b_flags & B_SCANNED) != 0)
387 bp->b_flags |= B_SCANNED;
388 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
390 if ((bp->b_flags & B_DELWRI) == 0)
391 panic("spec_fsync: not dirty");
392 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
405 * If synchronous the caller expects us to completely resolve all
406 * dirty buffers in the system. Wait for in-progress I/O to
407 * complete (which could include background bitmap writes), then
408 * retry if dirty blocks still exist.
410 if (ap->a_waitfor == MNT_WAIT) {
411 while (vp->v_numoutput) {
412 vp->v_flag |= VBWAIT;
413 (void) tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "spfsyn", 0);
415 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
416 if (--maxretry != 0) {
420 vprint("spec_fsync: giving up on dirty", vp);
429 struct vop_inactive_args /* {
435 VOP_UNLOCK(ap->a_vp, 0, ap->a_p);
440 * Just call the device strategy routine
444 struct vop_strategy_args /* {
454 if (((bp->b_flags & B_READ) == 0) &&
455 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
456 (*bioops.io_start)(bp);
459 * Collect statistics on synchronous and asynchronous read
460 * and write counts for disks that have associated filesystems.
463 if (vn_isdisk(vp, NULL) && (mp = vp->v_specmountpoint) != NULL) {
464 if ((bp->b_flags & B_READ) == 0) {
465 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
466 mp->mnt_stat.f_asyncwrites++;
468 mp->mnt_stat.f_syncwrites++;
470 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
471 mp->mnt_stat.f_asyncreads++;
473 mp->mnt_stat.f_syncreads++;
476 KASSERT(devsw(bp->b_dev) != NULL,
477 ("No devsw on dev %s responsible for buffer %p\n",
478 devtoname(bp->b_dev), bp));
479 KASSERT(devsw(bp->b_dev)->d_strategy != NULL,
480 ("No strategy on dev %s responsible for buffer %p\n",
481 devtoname(bp->b_dev), bp));
488 struct vop_freeblks_args /* {
498 * XXX: This assumes that strategy does the deed right away.
499 * XXX: this may not be TRTTD.
501 bsw = devsw(ap->a_vp->v_rdev);
502 if ((bsw->d_flags & D_CANFREE) == 0)
504 bp = geteblk(ap->a_length);
505 bp->b_flags |= B_FREEBUF;
506 bp->b_dev = ap->a_vp->v_rdev;
507 bp->b_blkno = ap->a_addr;
508 bp->b_offset = dbtob(ap->a_addr);
509 bp->b_bcount = ap->a_length;
515 * Implement degenerate case where the block requested is the block
516 * returned, and assume that the entire device is contiguous in regards
517 * to the contiguous block range (runp and runb).
521 struct vop_bmap_args /* {
524 struct vnode **a_vpp;
530 struct vnode *vp = ap->a_vp;
534 if (ap->a_vpp != NULL)
536 if (ap->a_bnp != NULL)
537 *ap->a_bnp = ap->a_bn;
538 if (vp->v_mount != NULL)
539 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
540 if (ap->a_runp != NULL)
542 if (ap->a_runb != NULL)
548 * Device close routine
553 struct vop_close_args /* {
556 struct ucred *a_cred;
560 struct vnode *vp = ap->a_vp;
561 struct proc *p = ap->a_p;
562 dev_t dev = vp->v_rdev;
565 * Hack: a tty device that is a controlling terminal
566 * has a reference from the session structure.
567 * We cannot easily tell that a character device is
568 * a controlling terminal, unless it is the closing
569 * process' controlling terminal. In that case,
570 * if the reference count is 2 (this last descriptor
571 * plus the session), release the reference from the session.
573 if (vcount(vp) == 2 && p && (vp->v_flag & VXLOCK) == 0 &&
574 vp == p->p_session->s_ttyvp) {
576 p->p_session->s_ttyvp = NULL;
579 * We do not want to really close the device if it
580 * is still in use unless we are trying to close it
581 * forcibly. Since every use (buffer, vnode, swap, cmap)
582 * holds a reference to the vnode, and because we mark
583 * any other vnodes that alias this device, when the
584 * sum of the reference counts on all the aliased
585 * vnodes descends to one, we are on last close.
587 if (vp->v_flag & VXLOCK) {
589 } else if (devsw(dev)->d_flags & D_TRACKCLOSE) {
590 /* Keep device updated on status */
591 } else if (vcount(vp) > 1) {
594 return (devsw(dev)->d_close(dev, ap->a_fflag, S_IFCHR, p));
598 * Print out the contents of a special device vnode.
602 struct vop_print_args /* {
607 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
612 * Special device advisory byte-level locks.
617 struct vop_advlock_args /* {
626 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
630 spec_getpages_iodone(bp)
634 bp->b_flags |= B_DONE;
640 struct vop_getpages_args *ap;
644 int i, pcount, size, s;
649 int toff, nextoff, nread;
650 struct vnode *vp = ap->a_vp;
655 pcount = round_page(ap->a_count) / PAGE_SIZE;
658 * Calculate the offset of the transfer and do sanity check.
659 * FreeBSD currently only supports an 8 TB range due to b_blkno
660 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
663 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
665 #define DADDR_T_BIT (sizeof(daddr_t)*8)
666 #define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
668 if (offset < 0 || offset > OFFSET_MAX) {
669 /* XXX still no %q in kernel. */
670 printf("spec_getpages: preposterous offset 0x%x%08x\n",
671 (u_int)((u_quad_t)offset >> 32),
672 (u_int)(offset & 0xffffffff));
673 return (VM_PAGER_ERROR);
676 blkno = btodb(offset);
679 * Round up physical size for real devices. We cannot round using
680 * v_mount's block size data because v_mount has nothing to do with
681 * the device. i.e. it's usually '/dev'. We need the physical block
682 * size for the device itself.
684 * We can't use v_specmountpoint because it only exists when the
685 * block device is mounted. However, we can use v_rdev.
688 if (vn_isdisk(vp, NULL))
689 blksiz = vp->v_rdev->si_bsize_phys;
693 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
696 kva = (vm_offset_t)bp->b_data;
699 * Map the pages to be read into the kva.
701 pmap_qenter(kva, ap->a_m, pcount);
703 /* Build a minimal buffer header. */
704 bp->b_flags = B_READ | B_CALL;
705 bp->b_iodone = spec_getpages_iodone;
707 /* B_PHYS is not set, but it is nice to fill this in. */
708 bp->b_rcred = bp->b_wcred = curproc->p_ucred;
709 if (bp->b_rcred != NOCRED)
711 if (bp->b_wcred != NOCRED)
714 bp->b_lblkno = blkno;
715 pbgetvp(ap->a_vp, bp);
717 bp->b_bufsize = size;
719 bp->b_runningbufspace = bp->b_bufsize;
720 runningbufspace += bp->b_runningbufspace;
723 cnt.v_vnodepgsin += pcount;
726 VOP_STRATEGY(bp->b_vp, bp);
730 /* We definitely need to be at splbio here. */
731 while ((bp->b_flags & B_DONE) == 0)
732 tsleep(bp, PVM, "spread", 0);
736 if ((bp->b_flags & B_ERROR) != 0) {
743 nread = size - bp->b_resid;
745 if (nread < ap->a_count) {
746 bzero((caddr_t)kva + nread,
747 ap->a_count - nread);
749 pmap_qremove(kva, pcount);
753 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
754 nextoff = toff + PAGE_SIZE;
757 m->flags &= ~PG_ZERO;
759 if (nextoff <= nread) {
760 m->valid = VM_PAGE_BITS_ALL;
762 } else if (toff < nread) {
764 * Since this is a VM request, we have to supply the
765 * unaligned offset to allow vm_page_set_validclean()
766 * to zero sub-DEV_BSIZE'd portions of the page.
768 vm_page_set_validclean(m, 0, nread - toff);
774 if (i != ap->a_reqpage) {
776 * Just in case someone was asking for this page we
777 * now tell them that it is ok to use.
779 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
781 if (m->flags & PG_WANTED) {
784 vm_page_deactivate(m);
793 } else if (m->valid) {
796 * Since this is a VM request, we need to make the
797 * entire page presentable by zeroing invalid sections.
799 if (m->valid != VM_PAGE_BITS_ALL)
800 vm_page_zero_invalid(m, FALSE);
804 m = ap->a_m[ap->a_reqpage];
806 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
807 devtoname(bp->b_dev), error, bp, bp->b_vp);
809 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
810 size, bp->b_resid, ap->a_count, m->valid);
812 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
813 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
815 * Free the buffer header back to the swap buffer pool.
818 return VM_PAGER_ERROR;
821 * Free the buffer header back to the swap buffer pool.