2 * Copyright (c) 2009 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Alex Hornung <ahornung@gmail.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include <sys/param.h>
35 #include <sys/systm.h>
37 #include <sys/kernel.h>
39 #include <sys/fcntl.h>
42 #include <sys/signalvar.h>
43 #include <sys/vnode.h>
45 #include <sys/mount.h>
47 #include <sys/fcntl.h>
48 #include <sys/namei.h>
49 #include <sys/dirent.h>
50 #include <sys/malloc.h>
54 #include <vm/vm_pager.h>
55 #include <vm/vm_zone.h>
56 #include <vm/vm_object.h>
57 #include <sys/filio.h>
58 #include <sys/ttycom.h>
59 #include <sys/sysref2.h>
61 #include <vfs/devfs/devfs.h>
62 #include <sys/pioctl.h>
64 #include <machine/limits.h>
66 MALLOC_DECLARE(M_DEVFS);
67 #define DEVFS_BADOP (void *)devfs_badop
69 static int devfs_badop(struct vop_generic_args *);
70 static int devfs_access(struct vop_access_args *);
71 static int devfs_inactive(struct vop_inactive_args *);
72 static int devfs_reclaim(struct vop_reclaim_args *);
73 static int devfs_readdir(struct vop_readdir_args *);
74 static int devfs_getattr(struct vop_getattr_args *);
75 static int devfs_setattr(struct vop_setattr_args *);
76 static int devfs_readlink(struct vop_readlink_args *);
77 static int devfs_print(struct vop_print_args *);
79 static int devfs_nresolve(struct vop_nresolve_args *);
80 static int devfs_nlookupdotdot(struct vop_nlookupdotdot_args *);
81 static int devfs_nsymlink(struct vop_nsymlink_args *);
82 static int devfs_nremove(struct vop_nremove_args *);
84 static int devfs_spec_open(struct vop_open_args *);
85 static int devfs_spec_close(struct vop_close_args *);
86 static int devfs_spec_fsync(struct vop_fsync_args *);
88 static int devfs_spec_read(struct vop_read_args *);
89 static int devfs_spec_write(struct vop_write_args *);
90 static int devfs_spec_ioctl(struct vop_ioctl_args *);
91 static int devfs_spec_poll(struct vop_poll_args *);
92 static int devfs_spec_kqfilter(struct vop_kqfilter_args *);
93 static int devfs_spec_strategy(struct vop_strategy_args *);
94 static void devfs_spec_strategy_done(struct bio *);
95 static int devfs_spec_freeblks(struct vop_freeblks_args *);
96 static int devfs_spec_bmap(struct vop_bmap_args *);
97 static int devfs_spec_advlock(struct vop_advlock_args *);
98 static void devfs_spec_getpages_iodone(struct bio *);
99 static int devfs_spec_getpages(struct vop_getpages_args *);
102 static int devfs_specf_close(struct file *);
103 static int devfs_specf_read(struct file *, struct uio *, struct ucred *, int);
104 static int devfs_specf_write(struct file *, struct uio *, struct ucred *, int);
105 static int devfs_specf_stat(struct file *, struct stat *, struct ucred *);
106 static int devfs_specf_kqfilter(struct file *, struct knote *);
107 static int devfs_specf_poll(struct file *, int, struct ucred *);
108 static int devfs_specf_ioctl(struct file *, u_long, caddr_t, struct ucred *);
111 static __inline int sequential_heuristic(struct uio *, struct file *);
112 extern struct lock devfs_lock;
115 * devfs vnode operations for regular files
117 struct vop_ops devfs_vnode_norm_vops = {
118 .vop_default = vop_defaultop,
119 .vop_access = devfs_access,
120 .vop_advlock = DEVFS_BADOP,
121 .vop_bmap = DEVFS_BADOP,
122 .vop_close = vop_stdclose,
123 .vop_getattr = devfs_getattr,
124 .vop_inactive = devfs_inactive,
125 .vop_ncreate = DEVFS_BADOP,
126 .vop_nresolve = devfs_nresolve,
127 .vop_nlookupdotdot = devfs_nlookupdotdot,
128 .vop_nlink = DEVFS_BADOP,
129 .vop_nmkdir = DEVFS_BADOP,
130 .vop_nmknod = DEVFS_BADOP,
131 .vop_nremove = devfs_nremove,
132 .vop_nrename = DEVFS_BADOP,
133 .vop_nrmdir = DEVFS_BADOP,
134 .vop_nsymlink = devfs_nsymlink,
135 .vop_open = vop_stdopen,
136 .vop_pathconf = vop_stdpathconf,
137 .vop_print = devfs_print,
138 .vop_read = DEVFS_BADOP,
139 .vop_readdir = devfs_readdir,
140 .vop_readlink = devfs_readlink,
141 .vop_reclaim = devfs_reclaim,
142 .vop_setattr = devfs_setattr,
143 .vop_write = DEVFS_BADOP,
144 .vop_ioctl = DEVFS_BADOP
148 * devfs vnode operations for character devices
150 struct vop_ops devfs_vnode_dev_vops = {
151 .vop_default = vop_defaultop,
152 .vop_access = devfs_access,
153 .vop_advlock = devfs_spec_advlock,
154 .vop_bmap = devfs_spec_bmap,
155 .vop_close = devfs_spec_close,
156 .vop_freeblks = devfs_spec_freeblks,
157 .vop_fsync = devfs_spec_fsync,
158 .vop_getattr = devfs_getattr,
159 .vop_getpages = devfs_spec_getpages,
160 .vop_inactive = devfs_inactive,
161 .vop_open = devfs_spec_open,
162 .vop_pathconf = vop_stdpathconf,
163 .vop_print = devfs_print,
164 .vop_poll = devfs_spec_poll,
165 .vop_kqfilter = devfs_spec_kqfilter,
166 .vop_read = devfs_spec_read,
167 .vop_readdir = DEVFS_BADOP,
168 .vop_readlink = DEVFS_BADOP,
169 .vop_reclaim = devfs_reclaim,
170 .vop_setattr = devfs_setattr,
171 .vop_strategy = devfs_spec_strategy,
172 .vop_write = devfs_spec_write,
173 .vop_ioctl = devfs_spec_ioctl
176 struct vop_ops *devfs_vnode_dev_vops_p = &devfs_vnode_dev_vops;
178 struct fileops devfs_dev_fileops = {
179 .fo_read = devfs_specf_read,
180 .fo_write = devfs_specf_write,
181 .fo_ioctl = devfs_specf_ioctl,
182 .fo_poll = devfs_specf_poll,
183 .fo_kqfilter = devfs_specf_kqfilter,
184 .fo_stat = devfs_specf_stat,
185 .fo_close = devfs_specf_close,
186 .fo_shutdown = nofo_shutdown
191 * generic entry point for unsupported operations
194 devfs_badop(struct vop_generic_args *ap)
201 devfs_access(struct vop_access_args *ap)
203 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
206 if (!devfs_node_is_accessible(node))
208 error = vop_helper_access(ap, node->uid, node->gid,
209 node->mode, node->flags);
216 devfs_inactive(struct vop_inactive_args *ap)
218 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
220 if (node == NULL || (node->flags & DEVFS_NODE_LINKED) == 0)
227 devfs_reclaim(struct vop_reclaim_args *ap)
229 struct devfs_node *node;
234 * Check if it is locked already. if not, we acquire the devfs lock
236 if (!(lockstatus(&devfs_lock, curthread)) == LK_EXCLUSIVE) {
237 lockmgr(&devfs_lock, LK_EXCLUSIVE);
244 * Get rid of the devfs_node if it is no longer linked into the
248 if ((node = DEVFS_NODE(vp)) != NULL) {
249 if ((node->flags & DEVFS_NODE_LINKED) == 0) {
251 /* NOTE: v_data is NULLd out by freep */
254 /* vp->v_data = NULL; handled below */
259 lockmgr(&devfs_lock, LK_RELEASE);
262 * v_rdev needs to be properly released using v_release_rdev
263 * Make sure v_data is NULL as well.
272 devfs_readdir(struct vop_readdir_args *ap)
274 struct devfs_node *dnode = DEVFS_NODE(ap->a_vp);
275 struct devfs_node *node;
284 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_readdir() called!\n");
286 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
288 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0)
291 if (!devfs_node_is_accessible(dnode))
294 lockmgr(&devfs_lock, LK_EXCLUSIVE);
296 saveoff = ap->a_uio->uio_offset;
298 if (ap->a_ncookies) {
299 ncookies = ap->a_uio->uio_resid / 16 + 1; /* Why / 16 ?? */
302 cookies = kmalloc(256 * sizeof(off_t), M_TEMP, M_WAITOK);
310 nanotime(&dnode->atime);
313 r = vop_write_dirent(&error, ap->a_uio, dnode->d_dir.d_ino,
318 cookies[cookie_index] = saveoff;
321 if (cookie_index == ncookies)
327 r = vop_write_dirent(&error, ap->a_uio,
328 dnode->parent->d_dir.d_ino,
331 r = vop_write_dirent(&error, ap->a_uio,
338 cookies[cookie_index] = saveoff;
341 if (cookie_index == ncookies)
345 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(dnode), link) {
346 if ((node->flags & DEVFS_HIDDEN) ||
347 (node->flags & DEVFS_INVISIBLE)) {
352 * If the node type is a valid devfs alias, then we make sure that the
353 * target isn't hidden. If it is, we don't show the link in the
356 if ((node->node_type == Plink) && (node->link_target != NULL) &&
357 (node->link_target->flags & DEVFS_HIDDEN))
360 if (node->cookie < saveoff)
363 saveoff = node->cookie;
365 error2 = vop_write_dirent(&error, ap->a_uio, node->d_dir.d_ino,
367 node->d_dir.d_namlen,
376 cookies[cookie_index] = node->cookie;
378 if (cookie_index == ncookies)
383 lockmgr(&devfs_lock, LK_RELEASE);
386 ap->a_uio->uio_offset = saveoff;
387 if (error && cookie_index == 0) {
389 kfree(cookies, M_TEMP);
391 *ap->a_cookies = NULL;
395 *ap->a_ncookies = cookie_index;
396 *ap->a_cookies = cookies;
404 devfs_nresolve(struct vop_nresolve_args *ap)
406 struct devfs_node *dnode = DEVFS_NODE(ap->a_dvp);
407 struct devfs_node *node, *found = NULL;
408 struct namecache *ncp;
409 struct vnode *vp = NULL;
414 ncp = ap->a_nch->ncp;
417 if (!devfs_node_is_accessible(dnode))
420 lockmgr(&devfs_lock, LK_EXCLUSIVE);
422 if ((dnode->node_type != Proot) && (dnode->node_type != Pdir)) {
424 cache_setvp(ap->a_nch, NULL);
428 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(dnode), link) {
429 if (len == node->d_dir.d_namlen) {
430 if (!memcmp(ncp->nc_name, node->d_dir.d_name, len)) {
438 if ((found->node_type == Plink) && (found->link_target))
439 found = found->link_target;
441 if (!(found->flags & DEVFS_HIDDEN))
442 devfs_allocv(/*ap->a_dvp->v_mount, */ &vp, found);
449 cache_setvp(ap->a_nch, NULL);
455 cache_setvp(ap->a_nch, vp);
458 lockmgr(&devfs_lock, LK_RELEASE);
465 devfs_nlookupdotdot(struct vop_nlookupdotdot_args *ap)
467 struct devfs_node *dnode = DEVFS_NODE(ap->a_dvp);
470 if (!devfs_node_is_accessible(dnode))
473 lockmgr(&devfs_lock, LK_EXCLUSIVE);
474 if (dnode->parent != NULL) {
475 devfs_allocv(ap->a_vpp, dnode->parent);
476 vn_unlock(*ap->a_vpp);
478 lockmgr(&devfs_lock, LK_RELEASE);
480 return ((*ap->a_vpp == NULL) ? ENOENT : 0);
485 devfs_getattr(struct vop_getattr_args *ap)
487 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
488 struct vattr *vap = ap->a_vap;
491 if (!devfs_node_is_accessible(node))
494 lockmgr(&devfs_lock, LK_EXCLUSIVE);
496 /* start by zeroing out the attributes */
499 /* next do all the common fields */
500 vap->va_type = ap->a_vp->v_type;
501 vap->va_mode = node->mode;
502 vap->va_fileid = DEVFS_NODE(ap->a_vp)->d_dir.d_ino ;
503 vap->va_flags = 0; /* XXX: what should this be? */
504 vap->va_blocksize = DEV_BSIZE;
505 vap->va_bytes = vap->va_size = sizeof(struct devfs_node);
507 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
509 vap->va_atime = node->atime;
510 vap->va_mtime = node->mtime;
511 vap->va_ctime = node->ctime;
513 vap->va_nlink = 1; /* number of references to file */
515 vap->va_uid = node->uid;
516 vap->va_gid = node->gid;
521 if ((node->node_type == Pdev) && node->d_dev) {
522 reference_dev(node->d_dev);
523 vap->va_rminor = node->d_dev->si_uminor;
524 release_dev(node->d_dev);
527 /* For a softlink the va_size is the length of the softlink */
528 if (node->symlink_name != 0) {
529 vap->va_size = node->symlink_namelen;
531 nanotime(&node->atime);
532 lockmgr(&devfs_lock, LK_RELEASE);
539 devfs_setattr(struct vop_setattr_args *ap)
541 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
545 if (!devfs_node_is_accessible(node))
548 lockmgr(&devfs_lock, LK_EXCLUSIVE);
552 if (vap->va_uid != (uid_t)VNOVAL) {
553 if ((ap->a_cred->cr_uid != node->uid) &&
554 (!groupmember(node->gid, ap->a_cred))) {
555 error = priv_check(curthread, PRIV_VFS_CHOWN);
559 node->uid = vap->va_uid;
562 if (vap->va_gid != (uid_t)VNOVAL) {
563 if ((ap->a_cred->cr_uid != node->uid) &&
564 (!groupmember(node->gid, ap->a_cred))) {
565 error = priv_check(curthread, PRIV_VFS_CHOWN);
569 node->gid = vap->va_gid;
572 if (vap->va_mode != (mode_t)VNOVAL) {
573 if (ap->a_cred->cr_uid != node->uid) {
574 error = priv_check(curthread, PRIV_VFS_ADMIN);
578 node->mode = vap->va_mode;
582 nanotime(&node->mtime);
583 lockmgr(&devfs_lock, LK_RELEASE);
590 devfs_readlink(struct vop_readlink_args *ap)
592 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
595 if (!devfs_node_is_accessible(node))
598 lockmgr(&devfs_lock, LK_EXCLUSIVE);
599 ret = uiomove(node->symlink_name, node->symlink_namelen, ap->a_uio);
600 lockmgr(&devfs_lock, LK_RELEASE);
607 devfs_print(struct vop_print_args *ap)
614 devfs_nsymlink(struct vop_nsymlink_args *ap)
616 struct devfs_node *dnode = DEVFS_NODE(ap->a_dvp);
617 struct devfs_node *node;
620 if (!devfs_node_is_accessible(dnode))
623 ap->a_vap->va_type = VLNK;
625 if ((dnode->node_type != Proot) && (dnode->node_type != Pdir))
628 lockmgr(&devfs_lock, LK_EXCLUSIVE);
629 devfs_allocvp(ap->a_dvp->v_mount, ap->a_vpp, Plink,
630 ap->a_nch->ncp->nc_name, dnode, NULL);
632 targetlen = strlen(ap->a_target);
634 node = DEVFS_NODE(*ap->a_vpp);
635 node->flags |= DEVFS_USER_CREATED;
636 node->symlink_namelen = targetlen;
637 node->symlink_name = kmalloc(targetlen + 1, M_DEVFS, M_WAITOK);
638 memcpy(node->symlink_name, ap->a_target, targetlen);
639 node->symlink_name[targetlen] = '\0';
640 cache_setunresolved(ap->a_nch);
641 cache_setvp(ap->a_nch, *ap->a_vpp);
643 lockmgr(&devfs_lock, LK_RELEASE);
645 return ((*ap->a_vpp == NULL) ? ENOTDIR : 0);
650 devfs_nremove(struct vop_nremove_args *ap)
652 struct devfs_node *dnode = DEVFS_NODE(ap->a_dvp);
653 struct devfs_node *node;
654 struct namecache *ncp;
657 ncp = ap->a_nch->ncp;
659 if (!devfs_node_is_accessible(dnode))
662 lockmgr(&devfs_lock, LK_EXCLUSIVE);
664 if ((dnode->node_type != Proot) && (dnode->node_type != Pdir))
667 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(dnode), link) {
668 if (ncp->nc_nlen != node->d_dir.d_namlen)
670 if (memcmp(ncp->nc_name, node->d_dir.d_name, ncp->nc_nlen))
674 * only allow removal of user created stuff (e.g. symlinks)
676 if ((node->flags & DEVFS_USER_CREATED) == 0) {
681 cache_inval_vp(node->v_node, CINV_DESTROY);
688 cache_setunresolved(ap->a_nch);
689 cache_setvp(ap->a_nch, NULL);
692 lockmgr(&devfs_lock, LK_RELEASE);
698 devfs_spec_open(struct vop_open_args *ap)
700 struct vnode *vp = ap->a_vp;
701 struct vnode *orig_vp = NULL;
702 struct devfs_node *node = DEVFS_NODE(vp);
703 struct devfs_node *newnode;
704 cdev_t dev, ndev = NULL;
709 if (node->d_dev == NULL)
711 if (!devfs_node_is_accessible(node))
715 if ((dev = vp->v_rdev) == NULL)
718 if (node && ap->a_fp) {
719 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -1.1-\n");
720 lockmgr(&devfs_lock, LK_EXCLUSIVE);
721 len = node->d_dir.d_namlen;
722 if (devfs_clone(node->d_dir.d_name, &len,
723 &ndev, 1, ap->a_cred) == 0) {
726 newnode = devfs_create_device_node(
727 DEVFS_MNTDATA(vp->v_mount)->root_node,
730 devfs_debug(DEVFS_DEBUG_DEBUG,
731 "parent here is: %s, node is: |%s|\n",
732 ((node->parent->node_type == Proot) ?
733 "ROOT!" : node->parent->d_dir.d_name),
734 newnode->d_dir.d_name);
735 devfs_debug(DEVFS_DEBUG_DEBUG,
737 ((struct devfs_node *)(TAILQ_LAST(DEVFS_DENODE_HEAD(node->parent), devfs_node_head)))->d_dir.d_name);
740 * orig_vp is set to the original vp if we cloned.
742 /* node->flags |= DEVFS_CLONED; */
743 devfs_allocv(&vp, newnode);
747 lockmgr(&devfs_lock, LK_RELEASE);
750 devfs_debug(DEVFS_DEBUG_DEBUG,
751 "devfs_spec_open() called on %s! \n",
755 * Make this field valid before any I/O in ->d_open
757 if (!dev->si_iosize_max)
758 dev->si_iosize_max = DFLTPHYS;
760 if (dev_dflags(dev) & D_TTY)
761 vp->v_flag |= VISTTY;
764 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_cred);
765 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
768 * Clean up any cloned vp if we error out.
774 /* orig_vp = NULL; */
780 if (dev_dflags(dev) & D_TTY) {
785 devfs_debug(DEVFS_DEBUG_DEBUG,
786 "devfs: no t_stop\n");
787 tp->t_stop = nottystop;
793 if (vn_isdisk(vp, NULL)) {
794 if (!dev->si_bsize_phys)
795 dev->si_bsize_phys = DEV_BSIZE;
796 vinitvmio(vp, IDX_TO_OFF(INT_MAX));
801 nanotime(&node->atime);
806 /* Ugly pty magic, to make pty devices appear once they are opened */
807 if (node && (node->flags & DEVFS_PTY) == DEVFS_PTY)
808 node->flags &= ~DEVFS_INVISIBLE;
811 ap->a_fp->f_type = DTYPE_VNODE;
812 ap->a_fp->f_flag = ap->a_mode & FMASK;
813 ap->a_fp->f_ops = &devfs_dev_fileops;
814 ap->a_fp->f_data = vp;
822 devfs_spec_close(struct vop_close_args *ap)
824 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
825 struct proc *p = curproc;
826 struct vnode *vp = ap->a_vp;
827 cdev_t dev = vp->v_rdev;
831 devfs_debug(DEVFS_DEBUG_DEBUG,
832 "devfs_spec_close() called on %s! \n",
836 * A couple of hacks for devices and tty devices. The
837 * vnode ref count cannot be used to figure out the
838 * last close, but we can use v_opencount now that
839 * revoke works properly.
841 * Detect the last close on a controlling terminal and clear
842 * the session (half-close).
847 if (p && vp->v_opencount <= 1 && vp == p->p_session->s_ttyvp) {
848 p->p_session->s_ttyvp = NULL;
853 * Vnodes can be opened and closed multiple times. Do not really
854 * close the device unless (1) it is being closed forcibly,
855 * (2) the device wants to track closes, or (3) this is the last
856 * vnode doing its last close on the device.
858 * XXX the VXLOCK (force close) case can leave vnodes referencing
859 * a closed device. This might not occur now that our revoke is
862 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -1- \n");
863 if (dev && ((vp->v_flag & VRECLAIMED) ||
864 (dev_dflags(dev) & D_TRACKCLOSE) ||
865 (vp->v_opencount == 1))) {
867 * Unlock around dev_dclose()
870 if (vn_islocked(vp)) {
874 error = dev_dclose(dev, ap->a_fflag, S_IFCHR);
877 * Ugly pty magic, to make pty devices disappear again once
880 if (node && (node->flags & DEVFS_PTY) == DEVFS_PTY)
881 node->flags |= DEVFS_INVISIBLE;
884 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
888 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -2- \n");
891 * Track the actual opens and closes on the vnode. The last close
892 * disassociates the rdev. If the rdev is already disassociated or
893 * the opencount is already 0, the vnode might have been revoked
894 * and no further opencount tracking occurs.
898 if (vp->v_opencount > 0)
906 devfs_specf_close(struct file *fp)
908 struct vnode *vp = (struct vnode *)fp->f_data;
912 fp->f_ops = &badfileops;
913 error = vn_close(vp, fp->f_flag);
921 * Device-optimized file table vnode read routine.
923 * This bypasses the VOP table and talks directly to the device. Most
924 * filesystems just route to specfs and can make this optimization.
926 * MPALMOSTSAFE - acquires mplock
929 devfs_specf_read(struct file *fp, struct uio *uio,
930 struct ucred *cred, int flags)
932 struct devfs_node *node;
939 KASSERT(uio->uio_td == curthread,
940 ("uio_td %p is not td %p", uio->uio_td, curthread));
942 vp = (struct vnode *)fp->f_data;
943 if (vp == NULL || vp->v_type == VBAD) {
947 node = DEVFS_NODE(vp);
949 if ((dev = vp->v_rdev) == NULL) {
956 if (uio->uio_resid == 0) {
960 if ((flags & O_FOFFSET) == 0)
961 uio->uio_offset = fp->f_offset;
964 if (flags & O_FBLOCKING) {
965 /* ioflag &= ~IO_NDELAY; */
966 } else if (flags & O_FNONBLOCKING) {
968 } else if (fp->f_flag & FNONBLOCK) {
971 if (flags & O_FBUFFERED) {
972 /* ioflag &= ~IO_DIRECT; */
973 } else if (flags & O_FUNBUFFERED) {
975 } else if (fp->f_flag & O_DIRECT) {
978 ioflag |= sequential_heuristic(uio, fp);
980 error = dev_dread(dev, uio, ioflag);
984 nanotime(&node->atime);
985 if ((flags & O_FOFFSET) == 0)
986 fp->f_offset = uio->uio_offset;
987 fp->f_nextoff = uio->uio_offset;
995 devfs_specf_write(struct file *fp, struct uio *uio,
996 struct ucred *cred, int flags)
998 struct devfs_node *node;
1005 KASSERT(uio->uio_td == curthread,
1006 ("uio_td %p is not p %p", uio->uio_td, curthread));
1008 vp = (struct vnode *)fp->f_data;
1009 if (vp == NULL || vp->v_type == VBAD) {
1013 node = DEVFS_NODE(vp);
1014 if (vp->v_type == VREG)
1015 bwillwrite(uio->uio_resid);
1016 vp = (struct vnode *)fp->f_data;
1018 if ((dev = vp->v_rdev) == NULL) {
1024 if ((flags & O_FOFFSET) == 0)
1025 uio->uio_offset = fp->f_offset;
1028 if (vp->v_type == VREG &&
1029 ((fp->f_flag & O_APPEND) || (flags & O_FAPPEND))) {
1030 ioflag |= IO_APPEND;
1033 if (flags & O_FBLOCKING) {
1034 /* ioflag &= ~IO_NDELAY; */
1035 } else if (flags & O_FNONBLOCKING) {
1036 ioflag |= IO_NDELAY;
1037 } else if (fp->f_flag & FNONBLOCK) {
1038 ioflag |= IO_NDELAY;
1040 if (flags & O_FBUFFERED) {
1041 /* ioflag &= ~IO_DIRECT; */
1042 } else if (flags & O_FUNBUFFERED) {
1043 ioflag |= IO_DIRECT;
1044 } else if (fp->f_flag & O_DIRECT) {
1045 ioflag |= IO_DIRECT;
1047 if (flags & O_FASYNCWRITE) {
1048 /* ioflag &= ~IO_SYNC; */
1049 } else if (flags & O_FSYNCWRITE) {
1051 } else if (fp->f_flag & O_FSYNC) {
1055 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))
1057 ioflag |= sequential_heuristic(uio, fp);
1059 error = dev_dwrite(dev, uio, ioflag);
1063 nanotime(&node->mtime);
1065 if ((flags & O_FOFFSET) == 0)
1066 fp->f_offset = uio->uio_offset;
1067 fp->f_nextoff = uio->uio_offset;
1075 devfs_specf_stat(struct file *fp, struct stat *sb, struct ucred *cred)
1081 vp = (struct vnode *)fp->f_data;
1082 error = vn_stat(vp, sb, cred);
1094 error = VOP_GETATTR(vp, vap);
1101 * Zero the spare stat fields
1107 * Copy from vattr table ... or not in case it's a cloned device
1109 if (vap->va_fsid != VNOVAL)
1110 sb->st_dev = vap->va_fsid;
1112 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1114 sb->st_ino = vap->va_fileid;
1116 mode = vap->va_mode;
1120 if (vap->va_nlink > (nlink_t)-1)
1121 sb->st_nlink = (nlink_t)-1;
1123 sb->st_nlink = vap->va_nlink;
1124 sb->st_uid = vap->va_uid;
1125 sb->st_gid = vap->va_gid;
1126 sb->st_rdev = dev2udev(DEVFS_NODE(vp)->d_dev);
1127 sb->st_size = vap->va_size;
1128 sb->st_atimespec = vap->va_atime;
1129 sb->st_mtimespec = vap->va_mtime;
1130 sb->st_ctimespec = vap->va_ctime;
1133 * A VCHR and VBLK device may track the last access and last modified
1134 * time independantly of the filesystem. This is particularly true
1135 * because device read and write calls may bypass the filesystem.
1137 if (vp->v_type == VCHR || vp->v_type == VBLK) {
1140 if (dev->si_lastread) {
1141 sb->st_atimespec.tv_sec = dev->si_lastread;
1142 sb->st_atimespec.tv_nsec = 0;
1144 if (dev->si_lastwrite) {
1145 sb->st_atimespec.tv_sec = dev->si_lastwrite;
1146 sb->st_atimespec.tv_nsec = 0;
1152 * According to www.opengroup.org, the meaning of st_blksize is
1153 * "a filesystem-specific preferred I/O block size for this
1154 * object. In some filesystem types, this may vary from file
1156 * Default to PAGE_SIZE after much discussion.
1159 sb->st_blksize = PAGE_SIZE;
1161 sb->st_flags = vap->va_flags;
1163 error = priv_check_cred(cred, PRIV_VFS_GENERATION, 0);
1167 sb->st_gen = (u_int32_t)vap->va_gen;
1169 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1170 sb->st_fsmid = vap->va_fsmid;
1178 devfs_specf_kqfilter(struct file *fp, struct knote *kn)
1180 struct devfs_node *node;
1187 vp = (struct vnode *)fp->f_data;
1188 if (vp == NULL || vp->v_type == VBAD) {
1192 node = DEVFS_NODE(vp);
1194 if ((dev = vp->v_rdev) == NULL) {
1200 error = dev_dkqfilter(dev, kn);
1205 nanotime(&node->atime);
1213 devfs_specf_poll(struct file *fp, int events, struct ucred *cred)
1215 struct devfs_node *node;
1222 vp = (struct vnode *)fp->f_data;
1223 if (vp == NULL || vp->v_type == VBAD) {
1227 node = DEVFS_NODE(vp);
1229 if ((dev = vp->v_rdev) == NULL) {
1234 error = dev_dpoll(dev, events);
1239 nanotime(&node->atime);
1247 * MPALMOSTSAFE - acquires mplock
1250 devfs_specf_ioctl(struct file *fp, u_long com, caddr_t data, struct ucred *ucred)
1252 struct devfs_node *node;
1257 struct fiodname_args *name_args;
1262 vp = ((struct vnode *)fp->f_data);
1263 if ((dev = vp->v_rdev) == NULL) {
1264 error = EBADF; /* device was revoked */
1268 node = DEVFS_NODE(vp);
1270 devfs_debug(DEVFS_DEBUG_DEBUG,
1271 "devfs_specf_ioctl() called! for dev %s\n",
1274 if (com == FIODTYPE) {
1275 *(int *)data = dev_dflags(dev) & D_TYPEMASK;
1278 } else if (com == FIODNAME) {
1279 name_args = (struct fiodname_args *)data;
1280 name = dev->si_name;
1281 namlen = strlen(name) + 1;
1283 devfs_debug(DEVFS_DEBUG_DEBUG,
1284 "ioctl, got: FIODNAME for %s\n", name);
1286 if (namlen <= name_args->len)
1287 error = copyout(dev->si_name, name_args->name, namlen);
1291 devfs_debug(DEVFS_DEBUG_DEBUG,
1292 "ioctl stuff: error: %d\n", error);
1296 error = dev_dioctl(dev, com, data, fp->f_flag, ucred);
1299 nanotime(&node->atime);
1300 nanotime(&node->mtime);
1303 if (com == TIOCSCTTY) {
1304 devfs_debug(DEVFS_DEBUG_DEBUG,
1305 "devfs_specf_ioctl: got TIOCSCTTY on %s\n",
1308 if (error == 0 && com == TIOCSCTTY) {
1309 struct proc *p = curthread->td_proc;
1310 struct session *sess;
1312 devfs_debug(DEVFS_DEBUG_DEBUG,
1313 "devfs_specf_ioctl: dealing with TIOCSCTTY on %s\n",
1319 sess = p->p_session;
1322 * Do nothing if reassigning same control tty
1324 if (sess->s_ttyvp == vp) {
1330 * Get rid of reference to old control tty
1332 ovp = sess->s_ttyvp;
1341 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl() finished! \n");
1347 devfs_spec_fsync(struct vop_fsync_args *ap)
1349 struct vnode *vp = ap->a_vp;
1352 if (!vn_isdisk(vp, NULL))
1356 * Flush all dirty buffers associated with a block device.
1358 error = vfsync(vp, ap->a_waitfor, 10000, NULL, NULL);
1363 devfs_spec_read(struct vop_read_args *ap)
1365 struct devfs_node *node;
1374 node = DEVFS_NODE(vp);
1376 if (dev == NULL) /* device was revoked */
1378 if (uio->uio_resid == 0)
1382 error = dev_dread(dev, uio, ap->a_ioflag);
1383 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1386 nanotime(&node->atime);
1392 * Vnode op for write
1394 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1395 * struct ucred *a_cred)
1398 devfs_spec_write(struct vop_write_args *ap)
1400 struct devfs_node *node;
1409 node = DEVFS_NODE(vp);
1411 KKASSERT(uio->uio_segflg != UIO_NOCOPY);
1413 if (dev == NULL) /* device was revoked */
1417 error = dev_dwrite(dev, uio, ap->a_ioflag);
1418 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1421 nanotime(&node->mtime);
1427 * Device ioctl operation.
1429 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
1430 * int a_fflag, struct ucred *a_cred)
1433 devfs_spec_ioctl(struct vop_ioctl_args *ap)
1435 struct vnode *vp = ap->a_vp;
1436 struct devfs_node *node;
1439 if ((dev = vp->v_rdev) == NULL)
1440 return (EBADF); /* device was revoked */
1441 node = DEVFS_NODE(vp);
1444 nanotime(&node->atime);
1445 nanotime(&node->mtime);
1448 return (dev_dioctl(dev, ap->a_command, ap->a_data,
1449 ap->a_fflag, ap->a_cred));
1453 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred)
1457 devfs_spec_poll(struct vop_poll_args *ap)
1459 struct vnode *vp = ap->a_vp;
1460 struct devfs_node *node;
1463 if ((dev = vp->v_rdev) == NULL)
1464 return (EBADF); /* device was revoked */
1465 node = DEVFS_NODE(vp);
1468 nanotime(&node->atime);
1470 return (dev_dpoll(dev, ap->a_events));
1474 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
1478 devfs_spec_kqfilter(struct vop_kqfilter_args *ap)
1480 struct vnode *vp = ap->a_vp;
1481 struct devfs_node *node;
1484 if ((dev = vp->v_rdev) == NULL)
1485 return (EBADF); /* device was revoked */
1486 node = DEVFS_NODE(vp);
1489 nanotime(&node->atime);
1491 return (dev_dkqfilter(dev, ap->a_kn));
1495 * Convert a vnode strategy call into a device strategy call. Vnode strategy
1496 * calls are not limited to device DMA limits so we have to deal with the
1499 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
1502 devfs_spec_strategy(struct vop_strategy_args *ap)
1504 struct bio *bio = ap->a_bio;
1505 struct buf *bp = bio->bio_buf;
1512 if (bp->b_cmd != BUF_CMD_READ && LIST_FIRST(&bp->b_dep) != NULL)
1516 * Collect statistics on synchronous and asynchronous read
1517 * and write counts for disks that have associated filesystems.
1520 KKASSERT(vp->v_rdev != NULL); /* XXX */
1521 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
1522 if (bp->b_cmd == BUF_CMD_READ) {
1523 if (bp->b_flags & BIO_SYNC)
1524 mp->mnt_stat.f_syncreads++;
1526 mp->mnt_stat.f_asyncreads++;
1528 if (bp->b_flags & BIO_SYNC)
1529 mp->mnt_stat.f_syncwrites++;
1531 mp->mnt_stat.f_asyncwrites++;
1536 * Device iosize limitations only apply to read and write. Shortcut
1537 * the I/O if it fits.
1539 if ((maxiosize = vp->v_rdev->si_iosize_max) == 0) {
1540 devfs_debug(DEVFS_DEBUG_DEBUG,
1541 "%s: si_iosize_max not set!\n",
1542 dev_dname(vp->v_rdev));
1543 maxiosize = MAXPHYS;
1545 #if SPEC_CHAIN_DEBUG & 2
1548 if (bp->b_bcount <= maxiosize ||
1549 (bp->b_cmd != BUF_CMD_READ && bp->b_cmd != BUF_CMD_WRITE)) {
1550 dev_dstrategy_chain(vp->v_rdev, bio);
1555 * Clone the buffer and set up an I/O chain to chunk up the I/O.
1557 nbp = kmalloc(sizeof(*bp), M_DEVBUF, M_INTWAIT|M_ZERO);
1561 BUF_LOCK(nbp, LK_EXCLUSIVE);
1564 nbp->b_flags = B_PAGING | (bp->b_flags & B_BNOCLIP);
1565 nbp->b_data = bp->b_data;
1566 nbp->b_bio1.bio_done = devfs_spec_strategy_done;
1567 nbp->b_bio1.bio_offset = bio->bio_offset;
1568 nbp->b_bio1.bio_caller_info1.ptr = bio;
1571 * Start the first transfer
1573 if (vn_isdisk(vp, NULL))
1574 chunksize = vp->v_rdev->si_bsize_phys;
1576 chunksize = DEV_BSIZE;
1577 chunksize = maxiosize / chunksize * chunksize;
1578 #if SPEC_CHAIN_DEBUG & 1
1579 devfs_debug(DEVFS_DEBUG_DEBUG,
1580 "spec_strategy chained I/O chunksize=%d\n",
1583 nbp->b_cmd = bp->b_cmd;
1584 nbp->b_bcount = chunksize;
1585 nbp->b_bufsize = chunksize; /* used to detect a short I/O */
1586 nbp->b_bio1.bio_caller_info2.index = chunksize;
1588 #if SPEC_CHAIN_DEBUG & 1
1589 devfs_debug(DEVFS_DEBUG_DEBUG,
1590 "spec_strategy: chain %p offset %d/%d bcount %d\n",
1591 bp, 0, bp->b_bcount, nbp->b_bcount);
1594 dev_dstrategy(vp->v_rdev, &nbp->b_bio1);
1596 if (DEVFS_NODE(vp)) {
1597 nanotime(&DEVFS_NODE(vp)->atime);
1598 nanotime(&DEVFS_NODE(vp)->mtime);
1605 * Chunked up transfer completion routine - chain transfers until done
1609 devfs_spec_strategy_done(struct bio *nbio)
1611 struct buf *nbp = nbio->bio_buf;
1612 struct bio *bio = nbio->bio_caller_info1.ptr; /* original bio */
1613 struct buf *bp = bio->bio_buf; /* original bp */
1614 int chunksize = nbio->bio_caller_info2.index; /* chunking */
1615 int boffset = nbp->b_data - bp->b_data;
1617 if (nbp->b_flags & B_ERROR) {
1619 * An error terminates the chain, propogate the error back
1620 * to the original bp
1622 bp->b_flags |= B_ERROR;
1623 bp->b_error = nbp->b_error;
1624 bp->b_resid = bp->b_bcount - boffset +
1625 (nbp->b_bcount - nbp->b_resid);
1626 #if SPEC_CHAIN_DEBUG & 1
1627 devfs_debug(DEVFS_DEBUG_DEBUG,
1628 "spec_strategy: chain %p error %d bcount %d/%d\n",
1629 bp, bp->b_error, bp->b_bcount,
1630 bp->b_bcount - bp->b_resid);
1632 kfree(nbp, M_DEVBUF);
1634 } else if (nbp->b_resid) {
1636 * A short read or write terminates the chain
1638 bp->b_error = nbp->b_error;
1639 bp->b_resid = bp->b_bcount - boffset +
1640 (nbp->b_bcount - nbp->b_resid);
1641 #if SPEC_CHAIN_DEBUG & 1
1642 devfs_debug(DEVFS_DEBUG_DEBUG,
1643 "spec_strategy: chain %p short read(1) "
1645 bp, bp->b_bcount - bp->b_resid, bp->b_bcount);
1647 kfree(nbp, M_DEVBUF);
1649 } else if (nbp->b_bcount != nbp->b_bufsize) {
1651 * A short read or write can also occur by truncating b_bcount
1653 #if SPEC_CHAIN_DEBUG & 1
1654 devfs_debug(DEVFS_DEBUG_DEBUG,
1655 "spec_strategy: chain %p short read(2) "
1657 bp, nbp->b_bcount + boffset, bp->b_bcount);
1660 bp->b_bcount = nbp->b_bcount + boffset;
1661 bp->b_resid = nbp->b_resid;
1662 kfree(nbp, M_DEVBUF);
1664 } else if (nbp->b_bcount + boffset == bp->b_bcount) {
1666 * No more data terminates the chain
1668 #if SPEC_CHAIN_DEBUG & 1
1669 devfs_debug(DEVFS_DEBUG_DEBUG,
1670 "spec_strategy: chain %p finished bcount %d\n",
1675 kfree(nbp, M_DEVBUF);
1679 * Continue the chain
1681 boffset += nbp->b_bcount;
1682 nbp->b_data = bp->b_data + boffset;
1683 nbp->b_bcount = bp->b_bcount - boffset;
1684 if (nbp->b_bcount > chunksize)
1685 nbp->b_bcount = chunksize;
1686 nbp->b_bio1.bio_done = devfs_spec_strategy_done;
1687 nbp->b_bio1.bio_offset = bio->bio_offset + boffset;
1689 #if SPEC_CHAIN_DEBUG & 1
1690 devfs_debug(DEVFS_DEBUG_DEBUG,
1691 "spec_strategy: chain %p offset %d/%d bcount %d\n",
1692 bp, boffset, bp->b_bcount, nbp->b_bcount);
1695 dev_dstrategy(nbp->b_vp->v_rdev, &nbp->b_bio1);
1700 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
1703 devfs_spec_freeblks(struct vop_freeblks_args *ap)
1708 * XXX: This assumes that strategy does the deed right away.
1709 * XXX: this may not be TRTTD.
1711 KKASSERT(ap->a_vp->v_rdev != NULL);
1712 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
1714 bp = geteblk(ap->a_length);
1715 bp->b_cmd = BUF_CMD_FREEBLKS;
1716 bp->b_bio1.bio_offset = ap->a_offset;
1717 bp->b_bcount = ap->a_length;
1718 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
1723 * Implement degenerate case where the block requested is the block
1724 * returned, and assume that the entire device is contiguous in regards
1725 * to the contiguous block range (runp and runb).
1727 * spec_bmap(struct vnode *a_vp, off_t a_loffset,
1728 * off_t *a_doffsetp, int *a_runp, int *a_runb)
1731 devfs_spec_bmap(struct vop_bmap_args *ap)
1733 if (ap->a_doffsetp != NULL)
1734 *ap->a_doffsetp = ap->a_loffset;
1735 if (ap->a_runp != NULL)
1736 *ap->a_runp = MAXBSIZE;
1737 if (ap->a_runb != NULL) {
1738 if (ap->a_loffset < MAXBSIZE)
1739 *ap->a_runb = (int)ap->a_loffset;
1741 *ap->a_runb = MAXBSIZE;
1748 * Special device advisory byte-level locks.
1750 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
1751 * struct flock *a_fl, int a_flags)
1755 devfs_spec_advlock(struct vop_advlock_args *ap)
1757 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
1761 devfs_spec_getpages_iodone(struct bio *bio)
1763 bio->bio_buf->b_cmd = BUF_CMD_DONE;
1764 wakeup(bio->bio_buf);
1768 * spec_getpages() - get pages associated with device vnode.
1770 * Note that spec_read and spec_write do not use the buffer cache, so we
1771 * must fully implement getpages here.
1774 devfs_spec_getpages(struct vop_getpages_args *ap)
1778 int i, pcount, size;
1781 vm_ooffset_t offset;
1782 int toff, nextoff, nread;
1783 struct vnode *vp = ap->a_vp;
1788 pcount = round_page(ap->a_count) / PAGE_SIZE;
1791 * Calculate the offset of the transfer and do sanity check.
1793 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
1796 * Round up physical size for real devices. We cannot round using
1797 * v_mount's block size data because v_mount has nothing to do with
1798 * the device. i.e. it's usually '/dev'. We need the physical block
1799 * size for the device itself.
1801 * We can't use v_rdev->si_mountpoint because it only exists when the
1802 * block device is mounted. However, we can use v_rdev.
1804 if (vn_isdisk(vp, NULL))
1805 blksiz = vp->v_rdev->si_bsize_phys;
1809 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
1812 kva = (vm_offset_t)bp->b_data;
1815 * Map the pages to be read into the kva.
1817 pmap_qenter(kva, ap->a_m, pcount);
1819 /* Build a minimal buffer header. */
1820 bp->b_cmd = BUF_CMD_READ;
1821 bp->b_bcount = size;
1823 bp->b_runningbufspace = size;
1825 runningbufspace += bp->b_runningbufspace;
1829 bp->b_bio1.bio_offset = offset;
1830 bp->b_bio1.bio_done = devfs_spec_getpages_iodone;
1832 mycpu->gd_cnt.v_vnodein++;
1833 mycpu->gd_cnt.v_vnodepgsin += pcount;
1836 vn_strategy(ap->a_vp, &bp->b_bio1);
1840 /* We definitely need to be at splbio here. */
1841 while (bp->b_cmd != BUF_CMD_DONE)
1842 tsleep(bp, 0, "spread", 0);
1846 if (bp->b_flags & B_ERROR) {
1848 error = bp->b_error;
1854 * If EOF is encountered we must zero-extend the result in order
1855 * to ensure that the page does not contain garabge. When no
1856 * error occurs, an early EOF is indicated if b_bcount got truncated.
1857 * b_resid is relative to b_bcount and should be 0, but some devices
1858 * might indicate an EOF with b_resid instead of truncating b_bcount.
1860 nread = bp->b_bcount - bp->b_resid;
1861 if (nread < ap->a_count)
1862 bzero((caddr_t)kva + nread, ap->a_count - nread);
1863 pmap_qremove(kva, pcount);
1866 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
1867 nextoff = toff + PAGE_SIZE;
1870 m->flags &= ~PG_ZERO;
1872 if (nextoff <= nread) {
1873 m->valid = VM_PAGE_BITS_ALL;
1875 } else if (toff < nread) {
1877 * Since this is a VM request, we have to supply the
1878 * unaligned offset to allow vm_page_set_validclean()
1879 * to zero sub-DEV_BSIZE'd portions of the page.
1881 vm_page_set_validclean(m, 0, nread - toff);
1887 if (i != ap->a_reqpage) {
1889 * Just in case someone was asking for this page we
1890 * now tell them that it is ok to use.
1892 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
1894 if (m->flags & PG_WANTED) {
1895 vm_page_activate(m);
1897 vm_page_deactivate(m);
1906 } else if (m->valid) {
1909 * Since this is a VM request, we need to make the
1910 * entire page presentable by zeroing invalid sections.
1912 if (m->valid != VM_PAGE_BITS_ALL)
1913 vm_page_zero_invalid(m, FALSE);
1917 m = ap->a_m[ap->a_reqpage];
1918 devfs_debug(DEVFS_DEBUG_WARNING,
1919 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
1920 devtoname(vp->v_rdev), error, bp, bp->b_vp);
1921 devfs_debug(DEVFS_DEBUG_WARNING,
1922 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
1923 size, bp->b_resid, ap->a_count, m->valid);
1924 devfs_debug(DEVFS_DEBUG_WARNING,
1925 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
1926 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
1928 * Free the buffer header back to the swap buffer pool.
1931 return VM_PAGER_ERROR;
1934 * Free the buffer header back to the swap buffer pool.
1942 sequential_heuristic(struct uio *uio, struct file *fp)
1945 * Sequential heuristic - detect sequential operation
1947 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
1948 uio->uio_offset == fp->f_nextoff) {
1950 * XXX we assume that the filesystem block size is
1951 * the default. Not true, but still gives us a pretty
1952 * good indicator of how sequential the read operations
1955 int tmpseq = fp->f_seqcount;
1957 tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
1958 if (tmpseq > IO_SEQMAX)
1960 fp->f_seqcount = tmpseq;
1961 return(fp->f_seqcount << IO_SEQSHIFT);
1965 * Not sequential, quick draw-down of seqcount
1967 if (fp->f_seqcount > 1)