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 1.2 2003/06/17 04:28:42 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>
56 static int spec_advlock __P((struct vop_advlock_args *));
57 static int spec_bmap __P((struct vop_bmap_args *));
58 static int spec_close __P((struct vop_close_args *));
59 static int spec_freeblks __P((struct vop_freeblks_args *));
60 static int spec_fsync __P((struct vop_fsync_args *));
61 static int spec_getpages __P((struct vop_getpages_args *));
62 static int spec_inactive __P((struct vop_inactive_args *));
63 static int spec_ioctl __P((struct vop_ioctl_args *));
64 static int spec_open __P((struct vop_open_args *));
65 static int spec_poll __P((struct vop_poll_args *));
66 static int spec_kqfilter __P((struct vop_kqfilter_args *));
67 static int spec_print __P((struct vop_print_args *));
68 static int spec_read __P((struct vop_read_args *));
69 static int spec_strategy __P((struct vop_strategy_args *));
70 static int spec_write __P((struct vop_write_args *));
72 vop_t **spec_vnodeop_p;
73 static struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
74 { &vop_default_desc, (vop_t *) vop_defaultop },
75 { &vop_access_desc, (vop_t *) vop_ebadf },
76 { &vop_advlock_desc, (vop_t *) spec_advlock },
77 { &vop_bmap_desc, (vop_t *) spec_bmap },
78 { &vop_close_desc, (vop_t *) spec_close },
79 { &vop_create_desc, (vop_t *) vop_panic },
80 { &vop_freeblks_desc, (vop_t *) spec_freeblks },
81 { &vop_fsync_desc, (vop_t *) spec_fsync },
82 { &vop_getpages_desc, (vop_t *) spec_getpages },
83 { &vop_inactive_desc, (vop_t *) spec_inactive },
84 { &vop_ioctl_desc, (vop_t *) spec_ioctl },
85 { &vop_lease_desc, (vop_t *) vop_null },
86 { &vop_link_desc, (vop_t *) vop_panic },
87 { &vop_mkdir_desc, (vop_t *) vop_panic },
88 { &vop_mknod_desc, (vop_t *) vop_panic },
89 { &vop_open_desc, (vop_t *) spec_open },
90 { &vop_pathconf_desc, (vop_t *) vop_stdpathconf },
91 { &vop_poll_desc, (vop_t *) spec_poll },
92 { &vop_kqfilter_desc, (vop_t *) spec_kqfilter },
93 { &vop_print_desc, (vop_t *) spec_print },
94 { &vop_read_desc, (vop_t *) spec_read },
95 { &vop_readdir_desc, (vop_t *) vop_panic },
96 { &vop_readlink_desc, (vop_t *) vop_panic },
97 { &vop_reallocblks_desc, (vop_t *) vop_panic },
98 { &vop_reclaim_desc, (vop_t *) vop_null },
99 { &vop_remove_desc, (vop_t *) vop_panic },
100 { &vop_rename_desc, (vop_t *) vop_panic },
101 { &vop_rmdir_desc, (vop_t *) vop_panic },
102 { &vop_setattr_desc, (vop_t *) vop_ebadf },
103 { &vop_strategy_desc, (vop_t *) spec_strategy },
104 { &vop_symlink_desc, (vop_t *) vop_panic },
105 { &vop_write_desc, (vop_t *) spec_write },
108 static struct vnodeopv_desc spec_vnodeop_opv_desc =
109 { &spec_vnodeop_p, spec_vnodeop_entries };
111 VNODEOP_SET(spec_vnodeop_opv_desc);
115 struct vop_generic_args /* {
116 struct vnodeop_desc *a_desc;
117 <other random data follows, presumably>
120 return (VOCALL(spec_vnodeop_p, ap->a_desc->vdesc_offset, ap));
123 static void spec_getpages_iodone __P((struct buf *bp));
126 * Open a special file.
131 struct vop_open_args /* {
134 struct ucred *a_cred;
138 struct proc *p = ap->a_p;
139 struct vnode *vp = ap->a_vp;
140 dev_t dev = vp->v_rdev;
146 * Don't allow open if fs is mounted -nodev.
148 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
152 if ( (dsw == NULL) || (dsw->d_open == NULL))
155 /* Make this field valid before any I/O in ->d_open */
156 if (!dev->si_iosize_max)
157 dev->si_iosize_max = DFLTPHYS;
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.
165 if (vn_isdisk(vp, NULL) && ap->a_cred != FSCRED &&
166 (ap->a_mode & FWRITE)) {
168 * Never allow opens for write if the device is mounted R/W
170 if (vp->v_specmountpoint != NULL &&
171 !(vp->v_specmountpoint->mnt_flag & MNT_RDONLY))
175 * When running in secure mode, do not allow opens
176 * for writing if the device is mounted
178 if (securelevel >= 1 && vp->v_specmountpoint != NULL)
182 * When running in very secure mode, do not allow
183 * opens for writing of any devices.
185 if (securelevel >= 2)
189 /* XXX: Special casing of ttys for deadfs. Probably redundant */
190 if (dsw->d_flags & D_TTY)
191 vp->v_flag |= VISTTY;
193 VOP_UNLOCK(vp, 0, p);
194 error = (*dsw->d_open)(dev, ap->a_mode, S_IFCHR, p);
195 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
200 if (dsw->d_flags & D_TTY) {
205 printf("Warning:%s: no t_stop, using nottystop\n", devtoname(dev));
206 tp->t_stop = nottystop;
211 if (vn_isdisk(vp, NULL)) {
212 if (!dev->si_bsize_phys)
213 dev->si_bsize_phys = DEV_BSIZE;
215 if ((dsw->d_flags & D_DISK) == 0) {
217 if (*cp == '#' && (dsw->d_flags & D_NAGGED) == 0) {
218 printf("WARNING: driver %s should register devices with make_dev() (dev_t = \"%s\")\n",
220 dsw->d_flags |= D_NAGGED;
232 struct vop_read_args /* {
236 struct ucred *a_cred;
250 if (uio->uio_resid == 0)
253 VOP_UNLOCK(vp, 0, p);
254 error = (*devsw(dev)->d_read) (dev, uio, ap->a_ioflag);
255 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
265 struct vop_write_args /* {
269 struct ucred *a_cred;
283 VOP_UNLOCK(vp, 0, p);
284 error = (*devsw(dev)->d_write) (dev, uio, ap->a_ioflag);
285 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
290 * Device ioctl operation.
295 struct vop_ioctl_args /* {
300 struct ucred *a_cred;
306 dev = ap->a_vp->v_rdev;
307 return ((*devsw(dev)->d_ioctl)(dev, ap->a_command,
308 ap->a_data, ap->a_fflag, ap->a_p));
314 struct vop_poll_args /* {
317 struct ucred *a_cred;
323 dev = ap->a_vp->v_rdev;
324 return (*devsw(dev)->d_poll)(dev, ap->a_events, ap->a_p);
330 struct vop_kqfilter_args /* {
337 dev = ap->a_vp->v_rdev;
338 if (devsw(dev)->d_flags & D_KQFILTER)
339 return (*devsw(dev)->d_kqfilter)(dev, ap->a_kn);
344 * Synch buffers associated with a block device
349 struct vop_fsync_args /* {
351 struct ucred *a_cred;
356 struct vnode *vp = ap->a_vp;
360 int maxretry = 10000; /* large, arbitrarily chosen */
362 if (!vn_isdisk(vp, NULL))
367 * MARK/SCAN initialization to avoid infinite loops
370 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp;
371 bp = TAILQ_NEXT(bp, b_vnbufs)) {
372 bp->b_flags &= ~B_SCANNED;
377 * Flush all dirty buffers associated with a block device.
381 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
382 nbp = TAILQ_NEXT(bp, b_vnbufs);
383 if ((bp->b_flags & B_SCANNED) != 0)
385 bp->b_flags |= B_SCANNED;
386 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
388 if ((bp->b_flags & B_DELWRI) == 0)
389 panic("spec_fsync: not dirty");
390 if ((vp->v_flag & VOBJBUF) && (bp->b_flags & B_CLUSTEROK)) {
403 * If synchronous the caller expects us to completely resolve all
404 * dirty buffers in the system. Wait for in-progress I/O to
405 * complete (which could include background bitmap writes), then
406 * retry if dirty blocks still exist.
408 if (ap->a_waitfor == MNT_WAIT) {
409 while (vp->v_numoutput) {
410 vp->v_flag |= VBWAIT;
411 (void) tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "spfsyn", 0);
413 if (!TAILQ_EMPTY(&vp->v_dirtyblkhd)) {
414 if (--maxretry != 0) {
418 vprint("spec_fsync: giving up on dirty", vp);
427 struct vop_inactive_args /* {
433 VOP_UNLOCK(ap->a_vp, 0, ap->a_p);
438 * Just call the device strategy routine
442 struct vop_strategy_args /* {
452 if (((bp->b_flags & B_READ) == 0) &&
453 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start)
454 (*bioops.io_start)(bp);
457 * Collect statistics on synchronous and asynchronous read
458 * and write counts for disks that have associated filesystems.
461 if (vn_isdisk(vp, NULL) && (mp = vp->v_specmountpoint) != NULL) {
462 if ((bp->b_flags & B_READ) == 0) {
463 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
464 mp->mnt_stat.f_asyncwrites++;
466 mp->mnt_stat.f_syncwrites++;
468 if (bp->b_lock.lk_lockholder == LK_KERNPROC)
469 mp->mnt_stat.f_asyncreads++;
471 mp->mnt_stat.f_syncreads++;
474 KASSERT(devsw(bp->b_dev) != NULL,
475 ("No devsw on dev %s responsible for buffer %p\n",
476 devtoname(bp->b_dev), bp));
477 KASSERT(devsw(bp->b_dev)->d_strategy != NULL,
478 ("No strategy on dev %s responsible for buffer %p\n",
479 devtoname(bp->b_dev), bp));
486 struct vop_freeblks_args /* {
496 * XXX: This assumes that strategy does the deed right away.
497 * XXX: this may not be TRTTD.
499 bsw = devsw(ap->a_vp->v_rdev);
500 if ((bsw->d_flags & D_CANFREE) == 0)
502 bp = geteblk(ap->a_length);
503 bp->b_flags |= B_FREEBUF;
504 bp->b_dev = ap->a_vp->v_rdev;
505 bp->b_blkno = ap->a_addr;
506 bp->b_offset = dbtob(ap->a_addr);
507 bp->b_bcount = ap->a_length;
513 * Implement degenerate case where the block requested is the block
514 * returned, and assume that the entire device is contiguous in regards
515 * to the contiguous block range (runp and runb).
519 struct vop_bmap_args /* {
522 struct vnode **a_vpp;
528 struct vnode *vp = ap->a_vp;
532 if (ap->a_vpp != NULL)
534 if (ap->a_bnp != NULL)
535 *ap->a_bnp = ap->a_bn;
536 if (vp->v_mount != NULL)
537 runp = runb = MAXBSIZE / vp->v_mount->mnt_stat.f_iosize;
538 if (ap->a_runp != NULL)
540 if (ap->a_runb != NULL)
546 * Device close routine
551 struct vop_close_args /* {
554 struct ucred *a_cred;
558 struct vnode *vp = ap->a_vp;
559 struct proc *p = ap->a_p;
560 dev_t dev = vp->v_rdev;
563 * Hack: a tty device that is a controlling terminal
564 * has a reference from the session structure.
565 * We cannot easily tell that a character device is
566 * a controlling terminal, unless it is the closing
567 * process' controlling terminal. In that case,
568 * if the reference count is 2 (this last descriptor
569 * plus the session), release the reference from the session.
571 if (vcount(vp) == 2 && p && (vp->v_flag & VXLOCK) == 0 &&
572 vp == p->p_session->s_ttyvp) {
574 p->p_session->s_ttyvp = NULL;
577 * We do not want to really close the device if it
578 * is still in use unless we are trying to close it
579 * forcibly. Since every use (buffer, vnode, swap, cmap)
580 * holds a reference to the vnode, and because we mark
581 * any other vnodes that alias this device, when the
582 * sum of the reference counts on all the aliased
583 * vnodes descends to one, we are on last close.
585 if (vp->v_flag & VXLOCK) {
587 } else if (devsw(dev)->d_flags & D_TRACKCLOSE) {
588 /* Keep device updated on status */
589 } else if (vcount(vp) > 1) {
592 return (devsw(dev)->d_close(dev, ap->a_fflag, S_IFCHR, p));
596 * Print out the contents of a special device vnode.
600 struct vop_print_args /* {
605 printf("tag VT_NON, dev %s\n", devtoname(ap->a_vp->v_rdev));
610 * Special device advisory byte-level locks.
615 struct vop_advlock_args /* {
624 return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
628 spec_getpages_iodone(bp)
632 bp->b_flags |= B_DONE;
638 struct vop_getpages_args *ap;
642 int i, pcount, size, s;
647 int toff, nextoff, nread;
648 struct vnode *vp = ap->a_vp;
653 pcount = round_page(ap->a_count) / PAGE_SIZE;
656 * Calculate the offset of the transfer and do sanity check.
657 * FreeBSD currently only supports an 8 TB range due to b_blkno
658 * being in DEV_BSIZE ( usually 512 ) byte chunks on call to
661 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
663 #define DADDR_T_BIT (sizeof(daddr_t)*8)
664 #define OFFSET_MAX ((1LL << (DADDR_T_BIT + DEV_BSHIFT)) - 1)
666 if (offset < 0 || offset > OFFSET_MAX) {
667 /* XXX still no %q in kernel. */
668 printf("spec_getpages: preposterous offset 0x%x%08x\n",
669 (u_int)((u_quad_t)offset >> 32),
670 (u_int)(offset & 0xffffffff));
671 return (VM_PAGER_ERROR);
674 blkno = btodb(offset);
677 * Round up physical size for real devices. We cannot round using
678 * v_mount's block size data because v_mount has nothing to do with
679 * the device. i.e. it's usually '/dev'. We need the physical block
680 * size for the device itself.
682 * We can't use v_specmountpoint because it only exists when the
683 * block device is mounted. However, we can use v_rdev.
686 if (vn_isdisk(vp, NULL))
687 blksiz = vp->v_rdev->si_bsize_phys;
691 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
694 kva = (vm_offset_t)bp->b_data;
697 * Map the pages to be read into the kva.
699 pmap_qenter(kva, ap->a_m, pcount);
701 /* Build a minimal buffer header. */
702 bp->b_flags = B_READ | B_CALL;
703 bp->b_iodone = spec_getpages_iodone;
705 /* B_PHYS is not set, but it is nice to fill this in. */
706 bp->b_rcred = bp->b_wcred = curproc->p_ucred;
707 if (bp->b_rcred != NOCRED)
709 if (bp->b_wcred != NOCRED)
712 bp->b_lblkno = blkno;
713 pbgetvp(ap->a_vp, bp);
715 bp->b_bufsize = size;
717 bp->b_runningbufspace = bp->b_bufsize;
718 runningbufspace += bp->b_runningbufspace;
721 cnt.v_vnodepgsin += pcount;
724 VOP_STRATEGY(bp->b_vp, bp);
728 /* We definitely need to be at splbio here. */
729 while ((bp->b_flags & B_DONE) == 0)
730 tsleep(bp, PVM, "spread", 0);
734 if ((bp->b_flags & B_ERROR) != 0) {
741 nread = size - bp->b_resid;
743 if (nread < ap->a_count) {
744 bzero((caddr_t)kva + nread,
745 ap->a_count - nread);
747 pmap_qremove(kva, pcount);
751 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
752 nextoff = toff + PAGE_SIZE;
755 m->flags &= ~PG_ZERO;
757 if (nextoff <= nread) {
758 m->valid = VM_PAGE_BITS_ALL;
760 } else if (toff < nread) {
762 * Since this is a VM request, we have to supply the
763 * unaligned offset to allow vm_page_set_validclean()
764 * to zero sub-DEV_BSIZE'd portions of the page.
766 vm_page_set_validclean(m, 0, nread - toff);
772 if (i != ap->a_reqpage) {
774 * Just in case someone was asking for this page we
775 * now tell them that it is ok to use.
777 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
779 if (m->flags & PG_WANTED) {
782 vm_page_deactivate(m);
791 } else if (m->valid) {
794 * Since this is a VM request, we need to make the
795 * entire page presentable by zeroing invalid sections.
797 if (m->valid != VM_PAGE_BITS_ALL)
798 vm_page_zero_invalid(m, FALSE);
802 m = ap->a_m[ap->a_reqpage];
804 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
805 devtoname(bp->b_dev), error, bp, bp->b_vp);
807 " size: %d, resid: %ld, a_count: %d, valid: 0x%x\n",
808 size, bp->b_resid, ap->a_count, m->valid);
810 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
811 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
813 * Free the buffer header back to the swap buffer pool.
816 return VM_PAGER_ERROR;
819 * Free the buffer header back to the swap buffer pool.