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 lockmgr(&devfs_lock, LK_RELEASE);
538 devfs_setattr(struct vop_setattr_args *ap)
540 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
544 if (!devfs_node_is_accessible(node))
547 lockmgr(&devfs_lock, LK_EXCLUSIVE);
551 if (vap->va_uid != (uid_t)VNOVAL) {
552 if ((ap->a_cred->cr_uid != node->uid) &&
553 (!groupmember(node->gid, ap->a_cred))) {
554 error = priv_check(curthread, PRIV_VFS_CHOWN);
558 node->uid = vap->va_uid;
561 if (vap->va_gid != (uid_t)VNOVAL) {
562 if ((ap->a_cred->cr_uid != node->uid) &&
563 (!groupmember(node->gid, ap->a_cred))) {
564 error = priv_check(curthread, PRIV_VFS_CHOWN);
568 node->gid = vap->va_gid;
571 if (vap->va_mode != (mode_t)VNOVAL) {
572 if (ap->a_cred->cr_uid != node->uid) {
573 error = priv_check(curthread, PRIV_VFS_ADMIN);
577 node->mode = vap->va_mode;
581 nanotime(&node->ctime);
582 lockmgr(&devfs_lock, LK_RELEASE);
589 devfs_readlink(struct vop_readlink_args *ap)
591 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
594 if (!devfs_node_is_accessible(node))
597 lockmgr(&devfs_lock, LK_EXCLUSIVE);
598 ret = uiomove(node->symlink_name, node->symlink_namelen, ap->a_uio);
599 lockmgr(&devfs_lock, LK_RELEASE);
606 devfs_print(struct vop_print_args *ap)
613 devfs_nsymlink(struct vop_nsymlink_args *ap)
615 struct devfs_node *dnode = DEVFS_NODE(ap->a_dvp);
616 struct devfs_node *node;
619 if (!devfs_node_is_accessible(dnode))
622 ap->a_vap->va_type = VLNK;
624 if ((dnode->node_type != Proot) && (dnode->node_type != Pdir))
627 lockmgr(&devfs_lock, LK_EXCLUSIVE);
628 devfs_allocvp(ap->a_dvp->v_mount, ap->a_vpp, Plink,
629 ap->a_nch->ncp->nc_name, dnode, NULL);
631 targetlen = strlen(ap->a_target);
633 node = DEVFS_NODE(*ap->a_vpp);
634 node->flags |= DEVFS_USER_CREATED;
635 node->symlink_namelen = targetlen;
636 node->symlink_name = kmalloc(targetlen + 1, M_DEVFS, M_WAITOK);
637 memcpy(node->symlink_name, ap->a_target, targetlen);
638 node->symlink_name[targetlen] = '\0';
639 cache_setunresolved(ap->a_nch);
640 cache_setvp(ap->a_nch, *ap->a_vpp);
642 lockmgr(&devfs_lock, LK_RELEASE);
644 return ((*ap->a_vpp == NULL) ? ENOTDIR : 0);
649 devfs_nremove(struct vop_nremove_args *ap)
651 struct devfs_node *dnode = DEVFS_NODE(ap->a_dvp);
652 struct devfs_node *node;
653 struct namecache *ncp;
656 ncp = ap->a_nch->ncp;
658 if (!devfs_node_is_accessible(dnode))
661 lockmgr(&devfs_lock, LK_EXCLUSIVE);
663 if ((dnode->node_type != Proot) && (dnode->node_type != Pdir))
666 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(dnode), link) {
667 if (ncp->nc_nlen != node->d_dir.d_namlen)
669 if (memcmp(ncp->nc_name, node->d_dir.d_name, ncp->nc_nlen))
673 * only allow removal of user created stuff (e.g. symlinks)
675 if ((node->flags & DEVFS_USER_CREATED) == 0) {
680 cache_inval_vp(node->v_node, CINV_DESTROY);
687 cache_setunresolved(ap->a_nch);
688 cache_setvp(ap->a_nch, NULL);
691 lockmgr(&devfs_lock, LK_RELEASE);
697 devfs_spec_open(struct vop_open_args *ap)
699 struct vnode *vp = ap->a_vp;
700 struct vnode *orig_vp = NULL;
701 struct devfs_node *node = DEVFS_NODE(vp);
702 struct devfs_node *newnode;
703 cdev_t dev, ndev = NULL;
707 if (node->d_dev == NULL)
709 if (!devfs_node_is_accessible(node))
713 if ((dev = vp->v_rdev) == NULL)
716 if (node && ap->a_fp) {
717 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -1.1-\n");
718 lockmgr(&devfs_lock, LK_EXCLUSIVE);
720 ndev = devfs_clone(dev, node->d_dir.d_name, node->d_dir.d_namlen,
721 ap->a_mode, ap->a_cred);
723 newnode = devfs_create_device_node(
724 DEVFS_MNTDATA(vp->v_mount)->root_node,
726 /* XXX: possibly destroy device if this happens */
728 if (newnode != NULL) {
732 devfs_debug(DEVFS_DEBUG_DEBUG,
733 "parent here is: %s, node is: |%s|\n",
734 ((node->parent->node_type == Proot) ?
735 "ROOT!" : node->parent->d_dir.d_name),
736 newnode->d_dir.d_name);
737 devfs_debug(DEVFS_DEBUG_DEBUG,
739 ((struct devfs_node *)(TAILQ_LAST(DEVFS_DENODE_HEAD(node->parent), devfs_node_head)))->d_dir.d_name);
742 * orig_vp is set to the original vp if we cloned.
744 /* node->flags |= DEVFS_CLONED; */
745 devfs_allocv(&vp, newnode);
750 lockmgr(&devfs_lock, LK_RELEASE);
753 devfs_debug(DEVFS_DEBUG_DEBUG,
754 "devfs_spec_open() called on %s! \n",
758 * Make this field valid before any I/O in ->d_open
760 if (!dev->si_iosize_max)
761 dev->si_iosize_max = DFLTPHYS;
763 if (dev_dflags(dev) & D_TTY)
764 vp->v_flag |= VISTTY;
767 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_cred);
768 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
771 * Clean up any cloned vp if we error out.
777 /* orig_vp = NULL; */
783 if (dev_dflags(dev) & D_TTY) {
788 devfs_debug(DEVFS_DEBUG_DEBUG,
789 "devfs: no t_stop\n");
790 tp->t_stop = nottystop;
796 if (vn_isdisk(vp, NULL)) {
797 if (!dev->si_bsize_phys)
798 dev->si_bsize_phys = DEV_BSIZE;
799 vinitvmio(vp, IDX_TO_OFF(INT_MAX));
805 nanotime(&node->atime);
811 /* Ugly pty magic, to make pty devices appear once they are opened */
812 if (node && (node->flags & DEVFS_PTY) == DEVFS_PTY)
813 node->flags &= ~DEVFS_INVISIBLE;
816 ap->a_fp->f_type = DTYPE_VNODE;
817 ap->a_fp->f_flag = ap->a_mode & FMASK;
818 ap->a_fp->f_ops = &devfs_dev_fileops;
819 ap->a_fp->f_data = vp;
827 devfs_spec_close(struct vop_close_args *ap)
829 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
830 struct proc *p = curproc;
831 struct vnode *vp = ap->a_vp;
832 cdev_t dev = vp->v_rdev;
836 devfs_debug(DEVFS_DEBUG_DEBUG,
837 "devfs_spec_close() called on %s! \n",
841 * A couple of hacks for devices and tty devices. The
842 * vnode ref count cannot be used to figure out the
843 * last close, but we can use v_opencount now that
844 * revoke works properly.
846 * Detect the last close on a controlling terminal and clear
847 * the session (half-close).
852 if (p && vp->v_opencount <= 1 && vp == p->p_session->s_ttyvp) {
853 p->p_session->s_ttyvp = NULL;
858 * Vnodes can be opened and closed multiple times. Do not really
859 * close the device unless (1) it is being closed forcibly,
860 * (2) the device wants to track closes, or (3) this is the last
861 * vnode doing its last close on the device.
863 * XXX the VXLOCK (force close) case can leave vnodes referencing
864 * a closed device. This might not occur now that our revoke is
867 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -1- \n");
868 if (dev && ((vp->v_flag & VRECLAIMED) ||
869 (dev_dflags(dev) & D_TRACKCLOSE) ||
870 (vp->v_opencount == 1))) {
872 * Unlock around dev_dclose()
875 if (vn_islocked(vp)) {
879 error = dev_dclose(dev, ap->a_fflag, S_IFCHR);
882 * Ugly pty magic, to make pty devices disappear again once
885 if (node && (node->flags & DEVFS_PTY) == DEVFS_PTY)
886 node->flags |= DEVFS_INVISIBLE;
889 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
893 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -2- \n");
896 * Track the actual opens and closes on the vnode. The last close
897 * disassociates the rdev. If the rdev is already disassociated or
898 * the opencount is already 0, the vnode might have been revoked
899 * and no further opencount tracking occurs.
903 if (vp->v_opencount > 0)
911 devfs_specf_close(struct file *fp)
913 struct vnode *vp = (struct vnode *)fp->f_data;
917 fp->f_ops = &badfileops;
918 error = vn_close(vp, fp->f_flag);
926 * Device-optimized file table vnode read routine.
928 * This bypasses the VOP table and talks directly to the device. Most
929 * filesystems just route to specfs and can make this optimization.
931 * MPALMOSTSAFE - acquires mplock
934 devfs_specf_read(struct file *fp, struct uio *uio,
935 struct ucred *cred, int flags)
937 struct devfs_node *node;
944 KASSERT(uio->uio_td == curthread,
945 ("uio_td %p is not td %p", uio->uio_td, curthread));
947 vp = (struct vnode *)fp->f_data;
948 if (vp == NULL || vp->v_type == VBAD) {
952 node = DEVFS_NODE(vp);
954 if ((dev = vp->v_rdev) == NULL) {
961 if (uio->uio_resid == 0) {
965 if ((flags & O_FOFFSET) == 0)
966 uio->uio_offset = fp->f_offset;
969 if (flags & O_FBLOCKING) {
970 /* ioflag &= ~IO_NDELAY; */
971 } else if (flags & O_FNONBLOCKING) {
973 } else if (fp->f_flag & FNONBLOCK) {
976 if (flags & O_FBUFFERED) {
977 /* ioflag &= ~IO_DIRECT; */
978 } else if (flags & O_FUNBUFFERED) {
980 } else if (fp->f_flag & O_DIRECT) {
983 ioflag |= sequential_heuristic(uio, fp);
985 error = dev_dread(dev, uio, ioflag);
989 nanotime(&node->atime);
990 if ((flags & O_FOFFSET) == 0)
991 fp->f_offset = uio->uio_offset;
992 fp->f_nextoff = uio->uio_offset;
1000 devfs_specf_write(struct file *fp, struct uio *uio,
1001 struct ucred *cred, int flags)
1003 struct devfs_node *node;
1010 KASSERT(uio->uio_td == curthread,
1011 ("uio_td %p is not p %p", uio->uio_td, curthread));
1013 vp = (struct vnode *)fp->f_data;
1014 if (vp == NULL || vp->v_type == VBAD) {
1018 node = DEVFS_NODE(vp);
1019 if (vp->v_type == VREG)
1020 bwillwrite(uio->uio_resid);
1021 vp = (struct vnode *)fp->f_data;
1023 if ((dev = vp->v_rdev) == NULL) {
1029 if ((flags & O_FOFFSET) == 0)
1030 uio->uio_offset = fp->f_offset;
1033 if (vp->v_type == VREG &&
1034 ((fp->f_flag & O_APPEND) || (flags & O_FAPPEND))) {
1035 ioflag |= IO_APPEND;
1038 if (flags & O_FBLOCKING) {
1039 /* ioflag &= ~IO_NDELAY; */
1040 } else if (flags & O_FNONBLOCKING) {
1041 ioflag |= IO_NDELAY;
1042 } else if (fp->f_flag & FNONBLOCK) {
1043 ioflag |= IO_NDELAY;
1045 if (flags & O_FBUFFERED) {
1046 /* ioflag &= ~IO_DIRECT; */
1047 } else if (flags & O_FUNBUFFERED) {
1048 ioflag |= IO_DIRECT;
1049 } else if (fp->f_flag & O_DIRECT) {
1050 ioflag |= IO_DIRECT;
1052 if (flags & O_FASYNCWRITE) {
1053 /* ioflag &= ~IO_SYNC; */
1054 } else if (flags & O_FSYNCWRITE) {
1056 } else if (fp->f_flag & O_FSYNC) {
1060 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))
1062 ioflag |= sequential_heuristic(uio, fp);
1064 error = dev_dwrite(dev, uio, ioflag);
1068 nanotime(&node->atime);
1069 nanotime(&node->mtime);
1072 if ((flags & O_FOFFSET) == 0)
1073 fp->f_offset = uio->uio_offset;
1074 fp->f_nextoff = uio->uio_offset;
1082 devfs_specf_stat(struct file *fp, struct stat *sb, struct ucred *cred)
1088 vp = (struct vnode *)fp->f_data;
1089 error = vn_stat(vp, sb, cred);
1101 error = VOP_GETATTR(vp, vap);
1108 * Zero the spare stat fields
1114 * Copy from vattr table ... or not in case it's a cloned device
1116 if (vap->va_fsid != VNOVAL)
1117 sb->st_dev = vap->va_fsid;
1119 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1121 sb->st_ino = vap->va_fileid;
1123 mode = vap->va_mode;
1127 if (vap->va_nlink > (nlink_t)-1)
1128 sb->st_nlink = (nlink_t)-1;
1130 sb->st_nlink = vap->va_nlink;
1131 sb->st_uid = vap->va_uid;
1132 sb->st_gid = vap->va_gid;
1133 sb->st_rdev = dev2udev(DEVFS_NODE(vp)->d_dev);
1134 sb->st_size = vap->va_size;
1135 sb->st_atimespec = vap->va_atime;
1136 sb->st_mtimespec = vap->va_mtime;
1137 sb->st_ctimespec = vap->va_ctime;
1140 * A VCHR and VBLK device may track the last access and last modified
1141 * time independantly of the filesystem. This is particularly true
1142 * because device read and write calls may bypass the filesystem.
1144 if (vp->v_type == VCHR || vp->v_type == VBLK) {
1147 if (dev->si_lastread) {
1148 sb->st_atimespec.tv_sec = dev->si_lastread;
1149 sb->st_atimespec.tv_nsec = 0;
1151 if (dev->si_lastwrite) {
1152 sb->st_atimespec.tv_sec = dev->si_lastwrite;
1153 sb->st_atimespec.tv_nsec = 0;
1159 * According to www.opengroup.org, the meaning of st_blksize is
1160 * "a filesystem-specific preferred I/O block size for this
1161 * object. In some filesystem types, this may vary from file
1163 * Default to PAGE_SIZE after much discussion.
1166 sb->st_blksize = PAGE_SIZE;
1168 sb->st_flags = vap->va_flags;
1170 error = priv_check_cred(cred, PRIV_VFS_GENERATION, 0);
1174 sb->st_gen = (u_int32_t)vap->va_gen;
1176 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1177 sb->st_fsmid = vap->va_fsmid;
1185 devfs_specf_kqfilter(struct file *fp, struct knote *kn)
1187 struct devfs_node *node;
1194 vp = (struct vnode *)fp->f_data;
1195 if (vp == NULL || vp->v_type == VBAD) {
1199 node = DEVFS_NODE(vp);
1201 if ((dev = vp->v_rdev) == NULL) {
1207 error = dev_dkqfilter(dev, kn);
1218 devfs_specf_poll(struct file *fp, int events, struct ucred *cred)
1220 struct devfs_node *node;
1227 vp = (struct vnode *)fp->f_data;
1228 if (vp == NULL || vp->v_type == VBAD) {
1232 node = DEVFS_NODE(vp);
1234 if ((dev = vp->v_rdev) == NULL) {
1239 error = dev_dpoll(dev, events);
1245 nanotime(&node->atime);
1254 * MPALMOSTSAFE - acquires mplock
1257 devfs_specf_ioctl(struct file *fp, u_long com, caddr_t data, struct ucred *ucred)
1259 struct devfs_node *node;
1264 struct fiodname_args *name_args;
1269 vp = ((struct vnode *)fp->f_data);
1270 if ((dev = vp->v_rdev) == NULL) {
1271 error = EBADF; /* device was revoked */
1275 node = DEVFS_NODE(vp);
1277 devfs_debug(DEVFS_DEBUG_DEBUG,
1278 "devfs_specf_ioctl() called! for dev %s\n",
1281 if (com == FIODTYPE) {
1282 *(int *)data = dev_dflags(dev) & D_TYPEMASK;
1285 } else if (com == FIODNAME) {
1286 name_args = (struct fiodname_args *)data;
1287 name = dev->si_name;
1288 namlen = strlen(name) + 1;
1290 devfs_debug(DEVFS_DEBUG_DEBUG,
1291 "ioctl, got: FIODNAME for %s\n", name);
1293 if (namlen <= name_args->len)
1294 error = copyout(dev->si_name, name_args->name, namlen);
1298 devfs_debug(DEVFS_DEBUG_DEBUG,
1299 "ioctl stuff: error: %d\n", error);
1303 error = dev_dioctl(dev, com, data, fp->f_flag, ucred);
1307 nanotime(&node->atime);
1308 nanotime(&node->mtime);
1312 if (com == TIOCSCTTY) {
1313 devfs_debug(DEVFS_DEBUG_DEBUG,
1314 "devfs_specf_ioctl: got TIOCSCTTY on %s\n",
1317 if (error == 0 && com == TIOCSCTTY) {
1318 struct proc *p = curthread->td_proc;
1319 struct session *sess;
1321 devfs_debug(DEVFS_DEBUG_DEBUG,
1322 "devfs_specf_ioctl: dealing with TIOCSCTTY on %s\n",
1328 sess = p->p_session;
1331 * Do nothing if reassigning same control tty
1333 if (sess->s_ttyvp == vp) {
1339 * Get rid of reference to old control tty
1341 ovp = sess->s_ttyvp;
1350 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl() finished! \n");
1356 devfs_spec_fsync(struct vop_fsync_args *ap)
1358 struct vnode *vp = ap->a_vp;
1361 if (!vn_isdisk(vp, NULL))
1365 * Flush all dirty buffers associated with a block device.
1367 error = vfsync(vp, ap->a_waitfor, 10000, NULL, NULL);
1372 devfs_spec_read(struct vop_read_args *ap)
1374 struct devfs_node *node;
1383 node = DEVFS_NODE(vp);
1385 if (dev == NULL) /* device was revoked */
1387 if (uio->uio_resid == 0)
1391 error = dev_dread(dev, uio, ap->a_ioflag);
1392 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1395 nanotime(&node->atime);
1401 * Vnode op for write
1403 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1404 * struct ucred *a_cred)
1407 devfs_spec_write(struct vop_write_args *ap)
1409 struct devfs_node *node;
1418 node = DEVFS_NODE(vp);
1420 KKASSERT(uio->uio_segflg != UIO_NOCOPY);
1422 if (dev == NULL) /* device was revoked */
1426 error = dev_dwrite(dev, uio, ap->a_ioflag);
1427 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1430 nanotime(&node->atime);
1431 nanotime(&node->mtime);
1438 * Device ioctl operation.
1440 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
1441 * int a_fflag, struct ucred *a_cred)
1444 devfs_spec_ioctl(struct vop_ioctl_args *ap)
1446 struct vnode *vp = ap->a_vp;
1447 struct devfs_node *node;
1450 if ((dev = vp->v_rdev) == NULL)
1451 return (EBADF); /* device was revoked */
1452 node = DEVFS_NODE(vp);
1456 nanotime(&node->atime);
1457 nanotime(&node->mtime);
1461 return (dev_dioctl(dev, ap->a_command, ap->a_data,
1462 ap->a_fflag, ap->a_cred));
1466 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred)
1470 devfs_spec_poll(struct vop_poll_args *ap)
1472 struct vnode *vp = ap->a_vp;
1473 struct devfs_node *node;
1476 if ((dev = vp->v_rdev) == NULL)
1477 return (EBADF); /* device was revoked */
1478 node = DEVFS_NODE(vp);
1482 nanotime(&node->atime);
1485 return (dev_dpoll(dev, ap->a_events));
1489 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
1493 devfs_spec_kqfilter(struct vop_kqfilter_args *ap)
1495 struct vnode *vp = ap->a_vp;
1496 struct devfs_node *node;
1499 if ((dev = vp->v_rdev) == NULL)
1500 return (EBADF); /* device was revoked */
1501 node = DEVFS_NODE(vp);
1505 nanotime(&node->atime);
1508 return (dev_dkqfilter(dev, ap->a_kn));
1512 * Convert a vnode strategy call into a device strategy call. Vnode strategy
1513 * calls are not limited to device DMA limits so we have to deal with the
1516 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
1519 devfs_spec_strategy(struct vop_strategy_args *ap)
1521 struct bio *bio = ap->a_bio;
1522 struct buf *bp = bio->bio_buf;
1529 if (bp->b_cmd != BUF_CMD_READ && LIST_FIRST(&bp->b_dep) != NULL)
1533 * Collect statistics on synchronous and asynchronous read
1534 * and write counts for disks that have associated filesystems.
1537 KKASSERT(vp->v_rdev != NULL); /* XXX */
1538 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
1539 if (bp->b_cmd == BUF_CMD_READ) {
1540 if (bp->b_flags & BIO_SYNC)
1541 mp->mnt_stat.f_syncreads++;
1543 mp->mnt_stat.f_asyncreads++;
1545 if (bp->b_flags & BIO_SYNC)
1546 mp->mnt_stat.f_syncwrites++;
1548 mp->mnt_stat.f_asyncwrites++;
1553 * Device iosize limitations only apply to read and write. Shortcut
1554 * the I/O if it fits.
1556 if ((maxiosize = vp->v_rdev->si_iosize_max) == 0) {
1557 devfs_debug(DEVFS_DEBUG_DEBUG,
1558 "%s: si_iosize_max not set!\n",
1559 dev_dname(vp->v_rdev));
1560 maxiosize = MAXPHYS;
1562 #if SPEC_CHAIN_DEBUG & 2
1565 if (bp->b_bcount <= maxiosize ||
1566 (bp->b_cmd != BUF_CMD_READ && bp->b_cmd != BUF_CMD_WRITE)) {
1567 dev_dstrategy_chain(vp->v_rdev, bio);
1572 * Clone the buffer and set up an I/O chain to chunk up the I/O.
1574 nbp = kmalloc(sizeof(*bp), M_DEVBUF, M_INTWAIT|M_ZERO);
1578 BUF_LOCK(nbp, LK_EXCLUSIVE);
1581 nbp->b_flags = B_PAGING | (bp->b_flags & B_BNOCLIP);
1582 nbp->b_data = bp->b_data;
1583 nbp->b_bio1.bio_done = devfs_spec_strategy_done;
1584 nbp->b_bio1.bio_offset = bio->bio_offset;
1585 nbp->b_bio1.bio_caller_info1.ptr = bio;
1588 * Start the first transfer
1590 if (vn_isdisk(vp, NULL))
1591 chunksize = vp->v_rdev->si_bsize_phys;
1593 chunksize = DEV_BSIZE;
1594 chunksize = maxiosize / chunksize * chunksize;
1595 #if SPEC_CHAIN_DEBUG & 1
1596 devfs_debug(DEVFS_DEBUG_DEBUG,
1597 "spec_strategy chained I/O chunksize=%d\n",
1600 nbp->b_cmd = bp->b_cmd;
1601 nbp->b_bcount = chunksize;
1602 nbp->b_bufsize = chunksize; /* used to detect a short I/O */
1603 nbp->b_bio1.bio_caller_info2.index = chunksize;
1605 #if SPEC_CHAIN_DEBUG & 1
1606 devfs_debug(DEVFS_DEBUG_DEBUG,
1607 "spec_strategy: chain %p offset %d/%d bcount %d\n",
1608 bp, 0, bp->b_bcount, nbp->b_bcount);
1611 dev_dstrategy(vp->v_rdev, &nbp->b_bio1);
1613 if (DEVFS_NODE(vp)) {
1614 nanotime(&DEVFS_NODE(vp)->atime);
1615 nanotime(&DEVFS_NODE(vp)->mtime);
1622 * Chunked up transfer completion routine - chain transfers until done
1626 devfs_spec_strategy_done(struct bio *nbio)
1628 struct buf *nbp = nbio->bio_buf;
1629 struct bio *bio = nbio->bio_caller_info1.ptr; /* original bio */
1630 struct buf *bp = bio->bio_buf; /* original bp */
1631 int chunksize = nbio->bio_caller_info2.index; /* chunking */
1632 int boffset = nbp->b_data - bp->b_data;
1634 if (nbp->b_flags & B_ERROR) {
1636 * An error terminates the chain, propogate the error back
1637 * to the original bp
1639 bp->b_flags |= B_ERROR;
1640 bp->b_error = nbp->b_error;
1641 bp->b_resid = bp->b_bcount - boffset +
1642 (nbp->b_bcount - nbp->b_resid);
1643 #if SPEC_CHAIN_DEBUG & 1
1644 devfs_debug(DEVFS_DEBUG_DEBUG,
1645 "spec_strategy: chain %p error %d bcount %d/%d\n",
1646 bp, bp->b_error, bp->b_bcount,
1647 bp->b_bcount - bp->b_resid);
1649 kfree(nbp, M_DEVBUF);
1651 } else if (nbp->b_resid) {
1653 * A short read or write terminates the chain
1655 bp->b_error = nbp->b_error;
1656 bp->b_resid = bp->b_bcount - boffset +
1657 (nbp->b_bcount - nbp->b_resid);
1658 #if SPEC_CHAIN_DEBUG & 1
1659 devfs_debug(DEVFS_DEBUG_DEBUG,
1660 "spec_strategy: chain %p short read(1) "
1662 bp, bp->b_bcount - bp->b_resid, bp->b_bcount);
1664 kfree(nbp, M_DEVBUF);
1666 } else if (nbp->b_bcount != nbp->b_bufsize) {
1668 * A short read or write can also occur by truncating b_bcount
1670 #if SPEC_CHAIN_DEBUG & 1
1671 devfs_debug(DEVFS_DEBUG_DEBUG,
1672 "spec_strategy: chain %p short read(2) "
1674 bp, nbp->b_bcount + boffset, bp->b_bcount);
1677 bp->b_bcount = nbp->b_bcount + boffset;
1678 bp->b_resid = nbp->b_resid;
1679 kfree(nbp, M_DEVBUF);
1681 } else if (nbp->b_bcount + boffset == bp->b_bcount) {
1683 * No more data terminates the chain
1685 #if SPEC_CHAIN_DEBUG & 1
1686 devfs_debug(DEVFS_DEBUG_DEBUG,
1687 "spec_strategy: chain %p finished bcount %d\n",
1692 kfree(nbp, M_DEVBUF);
1696 * Continue the chain
1698 boffset += nbp->b_bcount;
1699 nbp->b_data = bp->b_data + boffset;
1700 nbp->b_bcount = bp->b_bcount - boffset;
1701 if (nbp->b_bcount > chunksize)
1702 nbp->b_bcount = chunksize;
1703 nbp->b_bio1.bio_done = devfs_spec_strategy_done;
1704 nbp->b_bio1.bio_offset = bio->bio_offset + boffset;
1706 #if SPEC_CHAIN_DEBUG & 1
1707 devfs_debug(DEVFS_DEBUG_DEBUG,
1708 "spec_strategy: chain %p offset %d/%d bcount %d\n",
1709 bp, boffset, bp->b_bcount, nbp->b_bcount);
1712 dev_dstrategy(nbp->b_vp->v_rdev, &nbp->b_bio1);
1717 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
1720 devfs_spec_freeblks(struct vop_freeblks_args *ap)
1725 * XXX: This assumes that strategy does the deed right away.
1726 * XXX: this may not be TRTTD.
1728 KKASSERT(ap->a_vp->v_rdev != NULL);
1729 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
1731 bp = geteblk(ap->a_length);
1732 bp->b_cmd = BUF_CMD_FREEBLKS;
1733 bp->b_bio1.bio_offset = ap->a_offset;
1734 bp->b_bcount = ap->a_length;
1735 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
1740 * Implement degenerate case where the block requested is the block
1741 * returned, and assume that the entire device is contiguous in regards
1742 * to the contiguous block range (runp and runb).
1744 * spec_bmap(struct vnode *a_vp, off_t a_loffset,
1745 * off_t *a_doffsetp, int *a_runp, int *a_runb)
1748 devfs_spec_bmap(struct vop_bmap_args *ap)
1750 if (ap->a_doffsetp != NULL)
1751 *ap->a_doffsetp = ap->a_loffset;
1752 if (ap->a_runp != NULL)
1753 *ap->a_runp = MAXBSIZE;
1754 if (ap->a_runb != NULL) {
1755 if (ap->a_loffset < MAXBSIZE)
1756 *ap->a_runb = (int)ap->a_loffset;
1758 *ap->a_runb = MAXBSIZE;
1765 * Special device advisory byte-level locks.
1767 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
1768 * struct flock *a_fl, int a_flags)
1772 devfs_spec_advlock(struct vop_advlock_args *ap)
1774 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
1778 devfs_spec_getpages_iodone(struct bio *bio)
1780 bio->bio_buf->b_cmd = BUF_CMD_DONE;
1781 wakeup(bio->bio_buf);
1785 * spec_getpages() - get pages associated with device vnode.
1787 * Note that spec_read and spec_write do not use the buffer cache, so we
1788 * must fully implement getpages here.
1791 devfs_spec_getpages(struct vop_getpages_args *ap)
1795 int i, pcount, size;
1798 vm_ooffset_t offset;
1799 int toff, nextoff, nread;
1800 struct vnode *vp = ap->a_vp;
1805 pcount = round_page(ap->a_count) / PAGE_SIZE;
1808 * Calculate the offset of the transfer and do sanity check.
1810 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
1813 * Round up physical size for real devices. We cannot round using
1814 * v_mount's block size data because v_mount has nothing to do with
1815 * the device. i.e. it's usually '/dev'. We need the physical block
1816 * size for the device itself.
1818 * We can't use v_rdev->si_mountpoint because it only exists when the
1819 * block device is mounted. However, we can use v_rdev.
1821 if (vn_isdisk(vp, NULL))
1822 blksiz = vp->v_rdev->si_bsize_phys;
1826 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
1829 kva = (vm_offset_t)bp->b_data;
1832 * Map the pages to be read into the kva.
1834 pmap_qenter(kva, ap->a_m, pcount);
1836 /* Build a minimal buffer header. */
1837 bp->b_cmd = BUF_CMD_READ;
1838 bp->b_bcount = size;
1840 bp->b_runningbufspace = size;
1842 runningbufspace += bp->b_runningbufspace;
1846 bp->b_bio1.bio_offset = offset;
1847 bp->b_bio1.bio_done = devfs_spec_getpages_iodone;
1849 mycpu->gd_cnt.v_vnodein++;
1850 mycpu->gd_cnt.v_vnodepgsin += pcount;
1853 vn_strategy(ap->a_vp, &bp->b_bio1);
1857 /* We definitely need to be at splbio here. */
1858 while (bp->b_cmd != BUF_CMD_DONE)
1859 tsleep(bp, 0, "spread", 0);
1863 if (bp->b_flags & B_ERROR) {
1865 error = bp->b_error;
1871 * If EOF is encountered we must zero-extend the result in order
1872 * to ensure that the page does not contain garabge. When no
1873 * error occurs, an early EOF is indicated if b_bcount got truncated.
1874 * b_resid is relative to b_bcount and should be 0, but some devices
1875 * might indicate an EOF with b_resid instead of truncating b_bcount.
1877 nread = bp->b_bcount - bp->b_resid;
1878 if (nread < ap->a_count)
1879 bzero((caddr_t)kva + nread, ap->a_count - nread);
1880 pmap_qremove(kva, pcount);
1883 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
1884 nextoff = toff + PAGE_SIZE;
1887 m->flags &= ~PG_ZERO;
1889 if (nextoff <= nread) {
1890 m->valid = VM_PAGE_BITS_ALL;
1892 } else if (toff < nread) {
1894 * Since this is a VM request, we have to supply the
1895 * unaligned offset to allow vm_page_set_validclean()
1896 * to zero sub-DEV_BSIZE'd portions of the page.
1898 vm_page_set_validclean(m, 0, nread - toff);
1904 if (i != ap->a_reqpage) {
1906 * Just in case someone was asking for this page we
1907 * now tell them that it is ok to use.
1909 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
1911 if (m->flags & PG_WANTED) {
1912 vm_page_activate(m);
1914 vm_page_deactivate(m);
1923 } else if (m->valid) {
1926 * Since this is a VM request, we need to make the
1927 * entire page presentable by zeroing invalid sections.
1929 if (m->valid != VM_PAGE_BITS_ALL)
1930 vm_page_zero_invalid(m, FALSE);
1934 m = ap->a_m[ap->a_reqpage];
1935 devfs_debug(DEVFS_DEBUG_WARNING,
1936 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
1937 devtoname(vp->v_rdev), error, bp, bp->b_vp);
1938 devfs_debug(DEVFS_DEBUG_WARNING,
1939 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
1940 size, bp->b_resid, ap->a_count, m->valid);
1941 devfs_debug(DEVFS_DEBUG_WARNING,
1942 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
1943 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
1945 * Free the buffer header back to the swap buffer pool.
1948 return VM_PAGER_ERROR;
1951 * Free the buffer header back to the swap buffer pool.
1954 if (DEVFS_NODE(ap->a_vp))
1955 nanotime(&DEVFS_NODE(ap->a_vp)->mtime);
1961 sequential_heuristic(struct uio *uio, struct file *fp)
1964 * Sequential heuristic - detect sequential operation
1966 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
1967 uio->uio_offset == fp->f_nextoff) {
1969 * XXX we assume that the filesystem block size is
1970 * the default. Not true, but still gives us a pretty
1971 * good indicator of how sequential the read operations
1974 int tmpseq = fp->f_seqcount;
1976 tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
1977 if (tmpseq > IO_SEQMAX)
1979 fp->f_seqcount = tmpseq;
1980 return(fp->f_seqcount << IO_SEQSHIFT);
1984 * Not sequential, quick draw-down of seqcount
1986 if (fp->f_seqcount > 1)