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)
196 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs: specified vnode operation is not implemented (yet)\n");
202 devfs_access(struct vop_access_args *ap)
204 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
207 if (!devfs_node_is_accessible(node))
210 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_access() called!\n");
212 error = vop_helper_access(ap, node->uid, node->gid,
213 node->mode, node->flags);
216 /* XXX: consider possible special cases? terminal, ...? */
221 devfs_inactive(struct vop_inactive_args *ap)
223 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
225 if (node == NULL || (node->flags & DEVFS_NODE_LINKED) == 0)
232 devfs_reclaim(struct vop_reclaim_args *ap)
234 struct devfs_node *node;
238 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_reclaim() called!\n");
241 * Check if it is locked already. if not, we acquire the devfs lock
243 if (!(lockstatus(&devfs_lock, curthread)) == LK_EXCLUSIVE) {
244 lockmgr(&devfs_lock, LK_EXCLUSIVE);
251 * Get rid of the devfs_node if it is no longer linked into the
255 if ((node = DEVFS_NODE(vp)) != NULL) {
256 if ((node->flags & DEVFS_NODE_LINKED) == 0) {
258 /* NOTE: v_data is NULLd out by freep */
261 /* vp->v_data = NULL; handled below */
266 lockmgr(&devfs_lock, LK_RELEASE);
269 * v_rdev was not set with v_associate_rdev (??), so just NULL it
270 * out. Make sure v_data is NULL as well.
280 devfs_readdir(struct vop_readdir_args *ap)
282 struct devfs_node *node;
283 int error2 = 0, r, error = 0;
290 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_readdir() called!\n");
292 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
294 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0)
297 if (!devfs_node_is_accessible(DEVFS_NODE(ap->a_vp)))
300 lockmgr(&devfs_lock, LK_EXCLUSIVE);
302 saveoff = ap->a_uio->uio_offset;
304 if (ap->a_ncookies) {
305 ncookies = ap->a_uio->uio_resid / 16 + 1; /* Why / 16 ?? */
308 cookies = kmalloc(256 * sizeof(off_t), M_TEMP, M_WAITOK);
316 nanotime(&DEVFS_NODE(ap->a_vp)->atime);
319 r = vop_write_dirent(&error, ap->a_uio, DEVFS_NODE(ap->a_vp)->d_dir.d_ino, DT_DIR, 1, ".");
323 cookies[cookie_index] = saveoff;
326 if (cookie_index == ncookies)
331 if (DEVFS_NODE(ap->a_vp)->parent) {
332 r = vop_write_dirent(&error, ap->a_uio,
333 DEVFS_NODE(ap->a_vp)->d_dir.d_ino,
336 r = vop_write_dirent(&error, ap->a_uio,
337 DEVFS_NODE(ap->a_vp)->d_dir.d_ino, DT_DIR, 2, "..");
342 cookies[cookie_index] = saveoff;
345 if (cookie_index == ncookies)
349 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(DEVFS_NODE(ap->a_vp)), link) {
350 if ((node->flags & DEVFS_HIDDEN) || (node->flags & DEVFS_INVISIBLE))
353 if (node->cookie < saveoff)
361 saveoff = node->cookie;
363 error2 = vop_write_dirent(&error, ap->a_uio,
364 node->d_dir.d_ino, node->d_dir.d_type,
365 node->d_dir.d_namlen, node->d_dir.d_name);
373 cookies[cookie_index] = node->cookie;
375 if (cookie_index == ncookies)
380 lockmgr(&devfs_lock, LK_RELEASE);
383 ap->a_uio->uio_offset = saveoff;
384 if (error && cookie_index == 0) {
386 kfree(cookies, M_TEMP);
388 *ap->a_cookies = NULL;
392 *ap->a_ncookies = cookie_index;
393 *ap->a_cookies = cookies;
401 devfs_nresolve(struct vop_nresolve_args *ap)
403 struct devfs_node *node, *found = NULL;
404 struct namecache *ncp;
405 struct vnode *vp = NULL;
410 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve() called!\n");
412 ncp = ap->a_nch->ncp;
415 if (!devfs_node_is_accessible(DEVFS_NODE(ap->a_dvp)))
418 lockmgr(&devfs_lock, LK_EXCLUSIVE);
420 if ((DEVFS_NODE(ap->a_dvp)->node_type != Proot) &&
421 (DEVFS_NODE(ap->a_dvp)->node_type != Pdir)) {
422 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve: ap->a_dvp is not a dir!!!\n");
423 cache_setvp(ap->a_nch, NULL);
428 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -search- \n");
429 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(DEVFS_NODE(ap->a_dvp)), link) {
430 if (len == node->d_dir.d_namlen) {
431 if (!memcmp(ncp->nc_name, node->d_dir.d_name, len)) {
432 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve: found: %s\n", ncp->nc_name);
440 if ((found->node_type == Plink) && (found->link_target))
441 found = found->link_target;
443 if (!(found->flags & DEVFS_HIDDEN))
444 devfs_allocv(/*ap->a_dvp->v_mount, */ &vp, found);
447 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -2- \n");
451 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve vp==NULL \n");
453 /* XXX: len is int, devfs_clone expects size_t*, not int* */
454 if ((!hidden) && (!devfs_clone(ncp->nc_name, &len, NULL, 0, ap->a_cred))) {
458 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -4- \n");
460 cache_setvp(ap->a_nch, NULL);
461 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -5- \n");
465 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -6- \n");
468 cache_setvp(ap->a_nch, vp);
471 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -9- \n");
473 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nresolve -end:10- failed? %s \n", (error)?"FAILED!":"OK!");
474 lockmgr(&devfs_lock, LK_RELEASE);
480 devfs_nlookupdotdot(struct vop_nlookupdotdot_args *ap)
482 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nlookupdotdot() called!\n");
485 if (!devfs_node_is_accessible(DEVFS_NODE(ap->a_dvp)))
488 lockmgr(&devfs_lock, LK_EXCLUSIVE);
489 if (DEVFS_NODE(ap->a_dvp)->parent != NULL) {
490 devfs_allocv(/*ap->a_dvp->v_mount, */ap->a_vpp, DEVFS_NODE(ap->a_dvp)->parent);
491 vn_unlock(*ap->a_vpp);
493 lockmgr(&devfs_lock, LK_RELEASE);
495 return ((*ap->a_vpp == NULL) ? ENOENT : 0);
500 devfs_getattr(struct vop_getattr_args *ap)
502 struct vattr *vap = ap->a_vap;
503 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
506 if (!devfs_node_is_accessible(node))
509 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_getattr() called for %s!\n", DEVFS_NODE(ap->a_vp)->d_dir.d_name);
510 lockmgr(&devfs_lock, LK_EXCLUSIVE);
512 /* start by zeroing out the attributes */
515 /* next do all the common fields */
516 vap->va_type = ap->a_vp->v_type;
517 vap->va_mode = node->mode;
518 vap->va_fileid = DEVFS_NODE(ap->a_vp)->d_dir.d_ino ;
519 vap->va_flags = 0; /* XXX: what should this be? */
520 vap->va_blocksize = DEV_BSIZE;
521 vap->va_bytes = vap->va_size = sizeof(struct devfs_node);
523 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
526 vap->va_atime = node->atime;
527 vap->va_mtime = node->mtime;
528 vap->va_ctime = node->ctime;
530 vap->va_nlink = 1; /* number of references to file */
532 vap->va_uid = node->uid;
533 vap->va_gid = node->gid;
538 if ((DEVFS_NODE(ap->a_vp)->node_type == Pdev) &&
539 (DEVFS_NODE(ap->a_vp)->d_dev)) {
540 devfs_debug(DEVFS_DEBUG_DEBUG, "getattr: dev is: %p\n", DEVFS_NODE(ap->a_vp)->d_dev);
541 reference_dev(DEVFS_NODE(ap->a_vp)->d_dev);
542 vap->va_rminor = DEVFS_NODE(ap->a_vp)->d_dev->si_uminor;
543 release_dev(DEVFS_NODE(ap->a_vp)->d_dev);
546 /* For a softlink the va_size is the length of the softlink */
547 if (DEVFS_NODE(ap->a_vp)->symlink_name != 0) {
548 vap->va_size = DEVFS_NODE(ap->a_vp)->symlink_namelen;
550 nanotime(&node->atime);
551 lockmgr(&devfs_lock, LK_RELEASE);
557 devfs_setattr(struct vop_setattr_args *ap)
559 struct devfs_node *node;
563 node = DEVFS_NODE(ap->a_vp);
565 if (!devfs_node_is_accessible(node))
568 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_setattr() called!\n");
569 lockmgr(&devfs_lock, LK_EXCLUSIVE);
573 if (vap->va_uid != (uid_t)VNOVAL) {
574 if ((ap->a_cred->cr_uid != node->uid) &&
575 (!groupmember(node->gid, ap->a_cred))) {
576 error = priv_check(curthread, PRIV_VFS_CHOWN);
578 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_setattr, erroring out -1-\n");
582 node->uid = vap->va_uid;
585 if (vap->va_gid != (uid_t)VNOVAL) {
586 if ((ap->a_cred->cr_uid != node->uid) &&
587 (!groupmember(node->gid, ap->a_cred))) {
588 error = priv_check(curthread, PRIV_VFS_CHOWN);
590 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_setattr, erroring out -2-\n");
594 node->gid = vap->va_gid;
597 if (vap->va_mode != (mode_t)VNOVAL) {
598 if (ap->a_cred->cr_uid != node->uid) {
599 error = priv_check(curthread, PRIV_VFS_ADMIN);
601 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_setattr, erroring out -3-\n");
605 node->mode = vap->va_mode;
609 nanotime(&node->mtime);
610 lockmgr(&devfs_lock, LK_RELEASE);
616 devfs_readlink(struct vop_readlink_args *ap)
618 struct devfs_node *node = DEVFS_NODE(ap->a_vp);
621 if (!devfs_node_is_accessible(node))
624 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_readlink() called!\n");
626 lockmgr(&devfs_lock, LK_EXCLUSIVE);
627 ret = uiomove(node->symlink_name, node->symlink_namelen, ap->a_uio);
628 lockmgr(&devfs_lock, LK_RELEASE);
635 devfs_print(struct vop_print_args *ap)
637 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_print() called!\n");
639 /* XXX: print some useful debugging about node. */
645 devfs_nsymlink(struct vop_nsymlink_args *ap)
647 size_t targetlen = strlen(ap->a_target);
649 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nsymlink() called!\n");
651 if (!devfs_node_is_accessible(DEVFS_NODE(ap->a_dvp)))
654 ap->a_vap->va_type = VLNK;
656 if ((DEVFS_NODE(ap->a_dvp)->node_type != Proot) &&
657 (DEVFS_NODE(ap->a_dvp)->node_type != Pdir)) {
658 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nsymlink: ap->a_dvp is not a dir!!!\n");
661 lockmgr(&devfs_lock, LK_EXCLUSIVE);
662 devfs_allocvp(ap->a_dvp->v_mount, ap->a_vpp, Plink,
663 ap->a_nch->ncp->nc_name, DEVFS_NODE(ap->a_dvp), NULL);
666 DEVFS_NODE(*ap->a_vpp)->flags |= DEVFS_USER_CREATED;
668 DEVFS_NODE(*ap->a_vpp)->symlink_namelen = targetlen;
669 DEVFS_NODE(*ap->a_vpp)->symlink_name = kmalloc(targetlen + 1, M_DEVFS, M_WAITOK);
670 memcpy(DEVFS_NODE(*ap->a_vpp)->symlink_name, ap->a_target, targetlen);
671 DEVFS_NODE(*ap->a_vpp)->symlink_name[targetlen] = '\0';
672 cache_setunresolved(ap->a_nch);
673 cache_setvp(ap->a_nch, *ap->a_vpp);
675 lockmgr(&devfs_lock, LK_RELEASE);
677 return ((*ap->a_vpp == NULL) ? ENOTDIR : 0);
683 devfs_nremove(struct vop_nremove_args *ap)
685 struct devfs_node *node;
686 struct namecache *ncp;
689 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nremove() called!\n");
691 ncp = ap->a_nch->ncp;
693 if (!devfs_node_is_accessible(DEVFS_NODE(ap->a_dvp)))
696 lockmgr(&devfs_lock, LK_EXCLUSIVE);
698 if ((DEVFS_NODE(ap->a_dvp)->node_type != Proot) &&
699 (DEVFS_NODE(ap->a_dvp)->node_type != Pdir)) {
700 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_nremove: ap->a_dvp is not a dir!!!\n");
704 TAILQ_FOREACH(node, DEVFS_DENODE_HEAD(DEVFS_NODE(ap->a_dvp)), link) {
705 if (ncp->nc_nlen == node->d_dir.d_namlen) {
706 if (!memcmp(ncp->nc_name, node->d_dir.d_name, ncp->nc_nlen)) {
707 /* only allow removal of user created stuff (e.g. symlinks) */
708 if ((node->flags & DEVFS_USER_CREATED) == 0) {
713 cache_inval_vp(node->v_node, CINV_DESTROY);
723 cache_setunresolved(ap->a_nch);
724 cache_setvp(ap->a_nch, NULL);
727 lockmgr(&devfs_lock, LK_RELEASE);
733 devfs_spec_open(struct vop_open_args *ap)
735 struct vnode *vp = ap->a_vp;
736 struct vnode *orig_vp = NULL;
737 cdev_t dev, ndev = NULL;
738 struct devfs_node *node = NULL;
742 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open() called\n");
744 if (DEVFS_NODE(vp)) {
745 if (DEVFS_NODE(vp)->d_dev == NULL)
747 if (!devfs_node_is_accessible(DEVFS_NODE(vp)))
751 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -1-\n");
753 if ((dev = vp->v_rdev) == NULL)
756 if (DEVFS_NODE(vp) && ap->a_fp) {
757 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -1.1-\n");
758 lockmgr(&devfs_lock, LK_EXCLUSIVE);
759 len = DEVFS_NODE(vp)->d_dir.d_namlen;
760 if (!(devfs_clone(DEVFS_NODE(vp)->d_dir.d_name, &len, &ndev, 1, ap->a_cred))) {
761 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -1.2- |%s|\n", ndev->si_name);
765 node = devfs_create_device_node(DEVFS_MNTDATA(vp->v_mount)->root_node, dev, NULL, NULL);
767 devfs_debug(DEVFS_DEBUG_DEBUG, "parent here is: %s, node is: |%s|\n", (DEVFS_NODE(vp)->parent->node_type == Proot)?"ROOT!":DEVFS_NODE(vp)->parent->d_dir.d_name, node->d_dir.d_name);
768 devfs_debug(DEVFS_DEBUG_DEBUG, "test: %s\n", ((struct devfs_node *)(TAILQ_LAST(DEVFS_DENODE_HEAD(DEVFS_NODE(vp)->parent), devfs_node_head)))->d_dir.d_name);
771 * orig_vp is set to the original vp if we cloned.
773 /* node->flags |= DEVFS_CLONED; */
774 devfs_allocv(&vp, node);
778 lockmgr(&devfs_lock, LK_RELEASE);
781 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open() called on %s! \n", dev->si_name);
783 * Make this field valid before any I/O in ->d_open
785 if (!dev->si_iosize_max)
786 dev->si_iosize_max = DFLTPHYS;
788 if (dev_dflags(dev) & D_TTY)
789 vp->v_flag |= VISTTY;
792 error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_cred);
793 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
796 * Clean up any cloned vp if we error out.
799 devfs_debug(DEVFS_DEBUG_DEBUG,
800 "devfs_spec_open() error out: %x\n", error);
804 /* orig_vp = NULL; */
810 if (dev_dflags(dev) & D_TTY) {
815 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs: no t_stop\n");
816 tp->t_stop = nottystop;
822 if (vn_isdisk(vp, NULL)) {
823 if (!dev->si_bsize_phys)
824 dev->si_bsize_phys = DEV_BSIZE;
825 vinitvmio(vp, IDX_TO_OFF(INT_MAX));
830 nanotime(&DEVFS_NODE(vp)->atime);
835 /* Ugly pty magic, to make pty devices appear once they are opened */
836 if (DEVFS_NODE(vp) && ((DEVFS_NODE(vp)->flags & DEVFS_PTY) == DEVFS_PTY))
837 DEVFS_NODE(vp)->flags &= ~DEVFS_INVISIBLE;
840 ap->a_fp->f_type = DTYPE_VNODE;
841 ap->a_fp->f_flag = ap->a_mode & FMASK;
842 ap->a_fp->f_ops = &devfs_dev_fileops;
843 ap->a_fp->f_data = vp;
846 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -end:3-\n");
853 devfs_spec_close(struct vop_close_args *ap)
855 struct proc *p = curproc;
856 struct vnode *vp = ap->a_vp;
857 cdev_t dev = vp->v_rdev;
861 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() called on %s! \n", dev->si_name);
864 * A couple of hacks for devices and tty devices. The
865 * vnode ref count cannot be used to figure out the
866 * last close, but we can use v_opencount now that
867 * revoke works properly.
869 * Detect the last close on a controlling terminal and clear
870 * the session (half-close).
875 if (p && vp->v_opencount <= 1 && vp == p->p_session->s_ttyvp) {
876 p->p_session->s_ttyvp = NULL;
881 * Vnodes can be opened and closed multiple times. Do not really
882 * close the device unless (1) it is being closed forcibly,
883 * (2) the device wants to track closes, or (3) this is the last
884 * vnode doing its last close on the device.
886 * XXX the VXLOCK (force close) case can leave vnodes referencing
887 * a closed device. This might not occur now that our revoke is
890 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -1- \n");
891 if (dev && ((vp->v_flag & VRECLAIMED) ||
892 (dev_dflags(dev) & D_TRACKCLOSE) ||
893 (vp->v_opencount == 1))) {
895 if (vn_islocked(vp)) {
899 error = dev_dclose(dev, ap->a_fflag, S_IFCHR);
901 if (DEVFS_NODE(vp) && (DEVFS_NODE(vp)->flags & DEVFS_CLONED) == DEVFS_CLONED) {
902 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close: last of the cloned ones, so delete node %s\n", dev->si_name);
903 devfs_unlinkp(DEVFS_NODE(vp));
904 devfs_freep(DEVFS_NODE(vp));
907 /* Ugly pty magic, to make pty devices disappear again once they are closed */
908 if (DEVFS_NODE(vp) && ((DEVFS_NODE(vp)->flags & DEVFS_PTY) == DEVFS_PTY))
909 DEVFS_NODE(vp)->flags |= DEVFS_INVISIBLE;
912 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
916 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -2- \n");
918 * Track the actual opens and closes on the vnode. The last close
919 * disassociates the rdev. If the rdev is already disassociated or the
920 * opencount is already 0, the vnode might have been revoked and no
921 * further opencount tracking occurs.
924 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -3- \n");
925 if (vp->v_opencount == 1) {
926 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -3.5- \n");
930 if (vp->v_opencount > 0) {
931 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -4- \n");
933 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -5- \n");
936 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_close() -end:6- \n");
943 devfs_specf_close(struct file *fp)
946 struct vnode *vp = (struct vnode *)fp->f_data;
948 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_close() called! \n");
950 fp->f_ops = &badfileops;
952 error = vn_close(vp, fp->f_flag);
960 * Device-optimized file table vnode read routine.
962 * This bypasses the VOP table and talks directly to the device. Most
963 * filesystems just route to specfs and can make this optimization.
965 * MPALMOSTSAFE - acquires mplock
968 devfs_specf_read(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
976 KASSERT(uio->uio_td == curthread,
977 ("uio_td %p is not td %p", uio->uio_td, curthread));
979 vp = (struct vnode *)fp->f_data;
980 if (vp == NULL || vp->v_type == VBAD) {
985 if ((dev = vp->v_rdev) == NULL) {
992 if (uio->uio_resid == 0) {
996 if ((flags & O_FOFFSET) == 0)
997 uio->uio_offset = fp->f_offset;
1000 if (flags & O_FBLOCKING) {
1001 /* ioflag &= ~IO_NDELAY; */
1002 } else if (flags & O_FNONBLOCKING) {
1003 ioflag |= IO_NDELAY;
1004 } else if (fp->f_flag & FNONBLOCK) {
1005 ioflag |= IO_NDELAY;
1007 if (flags & O_FBUFFERED) {
1008 /* ioflag &= ~IO_DIRECT; */
1009 } else if (flags & O_FUNBUFFERED) {
1010 ioflag |= IO_DIRECT;
1011 } else if (fp->f_flag & O_DIRECT) {
1012 ioflag |= IO_DIRECT;
1014 ioflag |= sequential_heuristic(uio, fp);
1016 error = dev_dread(dev, uio, ioflag);
1020 nanotime(&DEVFS_NODE(vp)->atime);
1021 if ((flags & O_FOFFSET) == 0)
1022 fp->f_offset = uio->uio_offset;
1023 fp->f_nextoff = uio->uio_offset;
1031 devfs_specf_write(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
1039 KASSERT(uio->uio_td == curthread,
1040 ("uio_td %p is not p %p", uio->uio_td, curthread));
1042 vp = (struct vnode *)fp->f_data;
1043 if (vp == NULL || vp->v_type == VBAD) {
1047 if (vp->v_type == VREG)
1048 bwillwrite(uio->uio_resid);
1049 vp = (struct vnode *)fp->f_data;
1051 if ((dev = vp->v_rdev) == NULL) {
1057 if ((flags & O_FOFFSET) == 0)
1058 uio->uio_offset = fp->f_offset;
1061 if (vp->v_type == VREG &&
1062 ((fp->f_flag & O_APPEND) || (flags & O_FAPPEND))) {
1063 ioflag |= IO_APPEND;
1066 if (flags & O_FBLOCKING) {
1067 /* ioflag &= ~IO_NDELAY; */
1068 } else if (flags & O_FNONBLOCKING) {
1069 ioflag |= IO_NDELAY;
1070 } else if (fp->f_flag & FNONBLOCK) {
1071 ioflag |= IO_NDELAY;
1073 if (flags & O_FBUFFERED) {
1074 /* ioflag &= ~IO_DIRECT; */
1075 } else if (flags & O_FUNBUFFERED) {
1076 ioflag |= IO_DIRECT;
1077 } else if (fp->f_flag & O_DIRECT) {
1078 ioflag |= IO_DIRECT;
1080 if (flags & O_FASYNCWRITE) {
1081 /* ioflag &= ~IO_SYNC; */
1082 } else if (flags & O_FSYNCWRITE) {
1084 } else if (fp->f_flag & O_FSYNC) {
1088 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))
1090 ioflag |= sequential_heuristic(uio, fp);
1092 error = dev_dwrite(dev, uio, ioflag);
1096 nanotime(&DEVFS_NODE(vp)->mtime);
1098 if ((flags & O_FOFFSET) == 0)
1099 fp->f_offset = uio->uio_offset;
1100 fp->f_nextoff = uio->uio_offset;
1108 devfs_specf_stat(struct file *fp, struct stat *sb, struct ucred *cred)
1113 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_stat() called\n");
1116 vp = (struct vnode *)fp->f_data;
1117 error = vn_stat(vp, sb, cred);
1129 error = VOP_GETATTR(vp, vap);
1136 * Zero the spare stat fields
1142 * Copy from vattr table ... or not in case it's a cloned device
1144 if (vap->va_fsid != VNOVAL)
1145 sb->st_dev = vap->va_fsid;
1147 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1149 sb->st_ino = vap->va_fileid;
1151 mode = vap->va_mode;
1155 if (vap->va_nlink > (nlink_t)-1)
1156 sb->st_nlink = (nlink_t)-1;
1158 sb->st_nlink = vap->va_nlink;
1159 sb->st_uid = vap->va_uid;
1160 sb->st_gid = vap->va_gid;
1161 sb->st_rdev = dev2udev(DEVFS_NODE(vp)->d_dev);
1162 sb->st_size = vap->va_size;
1163 sb->st_atimespec = vap->va_atime;
1164 sb->st_mtimespec = vap->va_mtime;
1165 sb->st_ctimespec = vap->va_ctime;
1168 * A VCHR and VBLK device may track the last access and last modified
1169 * time independantly of the filesystem. This is particularly true
1170 * because device read and write calls may bypass the filesystem.
1172 if (vp->v_type == VCHR || vp->v_type == VBLK) {
1175 if (dev->si_lastread) {
1176 sb->st_atimespec.tv_sec = dev->si_lastread;
1177 sb->st_atimespec.tv_nsec = 0;
1179 if (dev->si_lastwrite) {
1180 sb->st_atimespec.tv_sec = dev->si_lastwrite;
1181 sb->st_atimespec.tv_nsec = 0;
1187 * According to www.opengroup.org, the meaning of st_blksize is
1188 * "a filesystem-specific preferred I/O block size for this
1189 * object. In some filesystem types, this may vary from file
1191 * Default to PAGE_SIZE after much discussion.
1194 sb->st_blksize = PAGE_SIZE;
1196 sb->st_flags = vap->va_flags;
1198 error = priv_check_cred(cred, PRIV_VFS_GENERATION, 0);
1202 sb->st_gen = (u_int32_t)vap->va_gen;
1204 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1205 sb->st_fsmid = vap->va_fsmid;
1213 devfs_specf_kqfilter(struct file *fp, struct knote *kn)
1221 vp = (struct vnode *)fp->f_data;
1222 if (vp == NULL || vp->v_type == VBAD) {
1227 if ((dev = vp->v_rdev) == NULL) {
1233 error = dev_dkqfilter(dev, kn);
1238 nanotime(&DEVFS_NODE(vp)->atime);
1246 devfs_specf_poll(struct file *fp, int events, struct ucred *cred)
1254 vp = (struct vnode *)fp->f_data;
1255 if (vp == NULL || vp->v_type == VBAD) {
1260 if ((dev = vp->v_rdev) == NULL) {
1265 error = dev_dpoll(dev, events);
1270 nanotime(&DEVFS_NODE(vp)->atime);
1278 * MPALMOSTSAFE - acquires mplock
1281 devfs_specf_ioctl(struct file *fp, u_long com, caddr_t data, struct ucred *ucred)
1283 struct vnode *vp = ((struct vnode *)fp->f_data);
1287 struct fiodname_args *name_args;
1291 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl() called! \n");
1295 if ((dev = vp->v_rdev) == NULL) {
1296 error = EBADF; /* device was revoked */
1299 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl() called! for dev %s\n", dev->si_name);
1301 if (!(dev_dflags(dev) & D_TTY))
1302 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl() called on %s! com is: %x\n", dev->si_name, com);
1304 if (com == FIODTYPE) {
1305 *(int *)data = dev_dflags(dev) & D_TYPEMASK;
1308 } else if (com == FIODNAME) {
1309 name_args = (struct fiodname_args *)data;
1310 name = dev->si_name;
1311 namlen = strlen(name) + 1;
1313 devfs_debug(DEVFS_DEBUG_DEBUG, "ioctl, got: FIODNAME for %s\n", name);
1315 if (namlen <= name_args->len)
1316 error = copyout(dev->si_name, name_args->name, namlen);
1320 devfs_debug(DEVFS_DEBUG_DEBUG, "ioctl stuff: error: %d\n", error);
1324 error = dev_dioctl(dev, com, data, fp->f_flag, ucred);
1326 if (DEVFS_NODE(vp)) {
1327 nanotime(&DEVFS_NODE(vp)->atime);
1328 nanotime(&DEVFS_NODE(vp)->mtime);
1331 if (com == TIOCSCTTY)
1332 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl: got TIOCSCTTY on %s\n", dev->si_name);
1333 if (error == 0 && com == TIOCSCTTY) {
1334 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl: dealing with TIOCSCTTY on %s\n", dev->si_name);
1335 struct proc *p = curthread->td_proc;
1336 struct session *sess;
1341 sess = p->p_session;
1342 /* Do nothing if reassigning same control tty */
1343 if (sess->s_ttyvp == vp) {
1347 /* Get rid of reference to old control tty */
1348 ovp = sess->s_ttyvp;
1357 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_specf_ioctl() finished! \n");
1363 devfs_spec_fsync(struct vop_fsync_args *ap)
1365 struct vnode *vp = ap->a_vp;
1368 if (!vn_isdisk(vp, NULL))
1372 * Flush all dirty buffers associated with a block device.
1374 error = vfsync(vp, ap->a_waitfor, 10000, NULL, NULL);
1398 devfs_spec_read(struct vop_read_args *ap)
1409 if (dev == NULL) /* device was revoked */
1411 if (uio->uio_resid == 0)
1415 error = dev_dread(dev, uio, ap->a_ioflag);
1416 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1419 nanotime(&DEVFS_NODE(vp)->atime);
1425 * Vnode op for write
1427 * spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1428 * struct ucred *a_cred)
1432 devfs_spec_write(struct vop_write_args *ap)
1443 KKASSERT(uio->uio_segflg != UIO_NOCOPY);
1445 if (dev == NULL) /* device was revoked */
1449 error = dev_dwrite(dev, uio, ap->a_ioflag);
1450 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1453 nanotime(&DEVFS_NODE(vp)->mtime);
1459 * Device ioctl operation.
1461 * spec_ioctl(struct vnode *a_vp, int a_command, caddr_t a_data,
1462 * int a_fflag, struct ucred *a_cred)
1466 devfs_spec_ioctl(struct vop_ioctl_args *ap)
1469 struct vnode *vp = ap->a_vp;
1471 if ((dev = vp->v_rdev) == NULL)
1472 return (EBADF); /* device was revoked */
1473 if ( ap->a_command == TIOCSCTTY )
1474 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_*SPEC*_ioctl: got TIOCSCTTY\n");
1476 if (DEVFS_NODE(vp)) {
1477 nanotime(&DEVFS_NODE(vp)->atime);
1478 nanotime(&DEVFS_NODE(vp)->mtime);
1481 return (dev_dioctl(dev, ap->a_command, ap->a_data,
1482 ap->a_fflag, ap->a_cred));
1486 * spec_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred)
1490 devfs_spec_poll(struct vop_poll_args *ap)
1493 struct vnode *vp = ap->a_vp;
1495 if ((dev = vp->v_rdev) == NULL)
1496 return (EBADF); /* device was revoked */
1499 nanotime(&DEVFS_NODE(vp)->atime);
1501 return (dev_dpoll(dev, ap->a_events));
1505 * spec_kqfilter(struct vnode *a_vp, struct knote *a_kn)
1509 devfs_spec_kqfilter(struct vop_kqfilter_args *ap)
1512 struct vnode *vp = ap->a_vp;
1514 if ((dev = vp->v_rdev) == NULL)
1515 return (EBADF); /* device was revoked */
1518 nanotime(&DEVFS_NODE(vp)->atime);
1520 return (dev_dkqfilter(dev, ap->a_kn));
1565 * Convert a vnode strategy call into a device strategy call. Vnode strategy
1566 * calls are not limited to device DMA limits so we have to deal with the
1569 * spec_strategy(struct vnode *a_vp, struct bio *a_bio)
1572 devfs_spec_strategy(struct vop_strategy_args *ap)
1574 struct bio *bio = ap->a_bio;
1575 struct buf *bp = bio->bio_buf;
1582 if (bp->b_cmd != BUF_CMD_READ && LIST_FIRST(&bp->b_dep) != NULL)
1586 * Collect statistics on synchronous and asynchronous read
1587 * and write counts for disks that have associated filesystems.
1590 KKASSERT(vp->v_rdev != NULL); /* XXX */
1591 if (vn_isdisk(vp, NULL) && (mp = vp->v_rdev->si_mountpoint) != NULL) {
1592 if (bp->b_cmd == BUF_CMD_READ) {
1593 if (bp->b_flags & BIO_SYNC)
1594 mp->mnt_stat.f_syncreads++;
1596 mp->mnt_stat.f_asyncreads++;
1598 if (bp->b_flags & BIO_SYNC)
1599 mp->mnt_stat.f_syncwrites++;
1601 mp->mnt_stat.f_asyncwrites++;
1606 * Device iosize limitations only apply to read and write. Shortcut
1607 * the I/O if it fits.
1609 if ((maxiosize = vp->v_rdev->si_iosize_max) == 0) {
1610 devfs_debug(DEVFS_DEBUG_DEBUG, "%s: si_iosize_max not set!\n", dev_dname(vp->v_rdev));
1611 maxiosize = MAXPHYS;
1613 #if SPEC_CHAIN_DEBUG & 2
1616 if (bp->b_bcount <= maxiosize ||
1617 (bp->b_cmd != BUF_CMD_READ && bp->b_cmd != BUF_CMD_WRITE)) {
1618 dev_dstrategy_chain(vp->v_rdev, bio);
1623 * Clone the buffer and set up an I/O chain to chunk up the I/O.
1625 nbp = kmalloc(sizeof(*bp), M_DEVBUF, M_INTWAIT|M_ZERO);
1629 BUF_LOCK(nbp, LK_EXCLUSIVE);
1632 nbp->b_flags = B_PAGING | (bp->b_flags & B_BNOCLIP);
1633 nbp->b_data = bp->b_data;
1634 nbp->b_bio1.bio_done = devfs_spec_strategy_done;
1635 nbp->b_bio1.bio_offset = bio->bio_offset;
1636 nbp->b_bio1.bio_caller_info1.ptr = bio;
1639 * Start the first transfer
1641 if (vn_isdisk(vp, NULL))
1642 chunksize = vp->v_rdev->si_bsize_phys;
1644 chunksize = DEV_BSIZE;
1645 chunksize = maxiosize / chunksize * chunksize;
1646 #if SPEC_CHAIN_DEBUG & 1
1647 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy chained I/O chunksize=%d\n", chunksize);
1649 nbp->b_cmd = bp->b_cmd;
1650 nbp->b_bcount = chunksize;
1651 nbp->b_bufsize = chunksize; /* used to detect a short I/O */
1652 nbp->b_bio1.bio_caller_info2.index = chunksize;
1654 #if SPEC_CHAIN_DEBUG & 1
1655 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy: chain %p offset %d/%d bcount %d\n",
1656 bp, 0, bp->b_bcount, nbp->b_bcount);
1659 dev_dstrategy(vp->v_rdev, &nbp->b_bio1);
1661 if (DEVFS_NODE(vp)) {
1662 nanotime(&DEVFS_NODE(vp)->atime);
1663 nanotime(&DEVFS_NODE(vp)->mtime);
1670 * Chunked up transfer completion routine - chain transfers until done
1674 devfs_spec_strategy_done(struct bio *nbio)
1676 struct buf *nbp = nbio->bio_buf;
1677 struct bio *bio = nbio->bio_caller_info1.ptr; /* original bio */
1678 struct buf *bp = bio->bio_buf; /* original bp */
1679 int chunksize = nbio->bio_caller_info2.index; /* chunking */
1680 int boffset = nbp->b_data - bp->b_data;
1682 if (nbp->b_flags & B_ERROR) {
1684 * An error terminates the chain, propogate the error back
1685 * to the original bp
1687 bp->b_flags |= B_ERROR;
1688 bp->b_error = nbp->b_error;
1689 bp->b_resid = bp->b_bcount - boffset +
1690 (nbp->b_bcount - nbp->b_resid);
1691 #if SPEC_CHAIN_DEBUG & 1
1692 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy: chain %p error %d bcount %d/%d\n",
1693 bp, bp->b_error, bp->b_bcount,
1694 bp->b_bcount - bp->b_resid);
1696 kfree(nbp, M_DEVBUF);
1698 } else if (nbp->b_resid) {
1700 * A short read or write terminates the chain
1702 bp->b_error = nbp->b_error;
1703 bp->b_resid = bp->b_bcount - boffset +
1704 (nbp->b_bcount - nbp->b_resid);
1705 #if SPEC_CHAIN_DEBUG & 1
1706 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy: chain %p short read(1) bcount %d/%d\n",
1707 bp, bp->b_bcount - bp->b_resid, bp->b_bcount);
1709 kfree(nbp, M_DEVBUF);
1711 } else if (nbp->b_bcount != nbp->b_bufsize) {
1713 * A short read or write can also occur by truncating b_bcount
1715 #if SPEC_CHAIN_DEBUG & 1
1716 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy: chain %p short read(2) bcount %d/%d\n",
1717 bp, nbp->b_bcount + boffset, bp->b_bcount);
1720 bp->b_bcount = nbp->b_bcount + boffset;
1721 bp->b_resid = nbp->b_resid;
1722 kfree(nbp, M_DEVBUF);
1724 } else if (nbp->b_bcount + boffset == bp->b_bcount) {
1726 * No more data terminates the chain
1728 #if SPEC_CHAIN_DEBUG & 1
1729 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy: chain %p finished bcount %d\n",
1734 kfree(nbp, M_DEVBUF);
1738 * Continue the chain
1740 boffset += nbp->b_bcount;
1741 nbp->b_data = bp->b_data + boffset;
1742 nbp->b_bcount = bp->b_bcount - boffset;
1743 if (nbp->b_bcount > chunksize)
1744 nbp->b_bcount = chunksize;
1745 nbp->b_bio1.bio_done = devfs_spec_strategy_done;
1746 nbp->b_bio1.bio_offset = bio->bio_offset + boffset;
1748 #if SPEC_CHAIN_DEBUG & 1
1749 devfs_debug(DEVFS_DEBUG_DEBUG, "spec_strategy: chain %p offset %d/%d bcount %d\n",
1750 bp, boffset, bp->b_bcount, nbp->b_bcount);
1753 dev_dstrategy(nbp->b_vp->v_rdev, &nbp->b_bio1);
1758 * spec_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
1761 devfs_spec_freeblks(struct vop_freeblks_args *ap)
1766 * XXX: This assumes that strategy does the deed right away.
1767 * XXX: this may not be TRTTD.
1769 KKASSERT(ap->a_vp->v_rdev != NULL);
1770 if ((dev_dflags(ap->a_vp->v_rdev) & D_CANFREE) == 0)
1772 bp = geteblk(ap->a_length);
1773 bp->b_cmd = BUF_CMD_FREEBLKS;
1774 bp->b_bio1.bio_offset = ap->a_offset;
1775 bp->b_bcount = ap->a_length;
1776 dev_dstrategy(ap->a_vp->v_rdev, &bp->b_bio1);
1781 * Implement degenerate case where the block requested is the block
1782 * returned, and assume that the entire device is contiguous in regards
1783 * to the contiguous block range (runp and runb).
1785 * spec_bmap(struct vnode *a_vp, off_t a_loffset,
1786 * off_t *a_doffsetp, int *a_runp, int *a_runb)
1789 devfs_spec_bmap(struct vop_bmap_args *ap)
1791 if (ap->a_doffsetp != NULL)
1792 *ap->a_doffsetp = ap->a_loffset;
1793 if (ap->a_runp != NULL)
1794 *ap->a_runp = MAXBSIZE;
1795 if (ap->a_runb != NULL) {
1796 if (ap->a_loffset < MAXBSIZE)
1797 *ap->a_runb = (int)ap->a_loffset;
1799 *ap->a_runb = MAXBSIZE;
1806 * Special device advisory byte-level locks.
1808 * spec_advlock(struct vnode *a_vp, caddr_t a_id, int a_op,
1809 * struct flock *a_fl, int a_flags)
1813 devfs_spec_advlock(struct vop_advlock_args *ap)
1815 return ((ap->a_flags & F_POSIX) ? EINVAL : EOPNOTSUPP);
1819 devfs_spec_getpages_iodone(struct bio *bio)
1821 bio->bio_buf->b_cmd = BUF_CMD_DONE;
1822 wakeup(bio->bio_buf);
1826 * spec_getpages() - get pages associated with device vnode.
1828 * Note that spec_read and spec_write do not use the buffer cache, so we
1829 * must fully implement getpages here.
1832 devfs_spec_getpages(struct vop_getpages_args *ap)
1836 int i, pcount, size;
1839 vm_ooffset_t offset;
1840 int toff, nextoff, nread;
1841 struct vnode *vp = ap->a_vp;
1846 pcount = round_page(ap->a_count) / PAGE_SIZE;
1849 * Calculate the offset of the transfer and do sanity check.
1851 offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
1854 * Round up physical size for real devices. We cannot round using
1855 * v_mount's block size data because v_mount has nothing to do with
1856 * the device. i.e. it's usually '/dev'. We need the physical block
1857 * size for the device itself.
1859 * We can't use v_rdev->si_mountpoint because it only exists when the
1860 * block device is mounted. However, we can use v_rdev.
1863 if (vn_isdisk(vp, NULL))
1864 blksiz = vp->v_rdev->si_bsize_phys;
1868 size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
1871 kva = (vm_offset_t)bp->b_data;
1874 * Map the pages to be read into the kva.
1876 pmap_qenter(kva, ap->a_m, pcount);
1878 /* Build a minimal buffer header. */
1879 bp->b_cmd = BUF_CMD_READ;
1880 bp->b_bcount = size;
1882 bp->b_runningbufspace = size;
1884 runningbufspace += bp->b_runningbufspace;
1888 bp->b_bio1.bio_offset = offset;
1889 bp->b_bio1.bio_done = devfs_spec_getpages_iodone;
1891 mycpu->gd_cnt.v_vnodein++;
1892 mycpu->gd_cnt.v_vnodepgsin += pcount;
1895 vn_strategy(ap->a_vp, &bp->b_bio1);
1899 /* We definitely need to be at splbio here. */
1900 while (bp->b_cmd != BUF_CMD_DONE)
1901 tsleep(bp, 0, "spread", 0);
1905 if (bp->b_flags & B_ERROR) {
1907 error = bp->b_error;
1913 * If EOF is encountered we must zero-extend the result in order
1914 * to ensure that the page does not contain garabge. When no
1915 * error occurs, an early EOF is indicated if b_bcount got truncated.
1916 * b_resid is relative to b_bcount and should be 0, but some devices
1917 * might indicate an EOF with b_resid instead of truncating b_bcount.
1919 nread = bp->b_bcount - bp->b_resid;
1920 if (nread < ap->a_count)
1921 bzero((caddr_t)kva + nread, ap->a_count - nread);
1922 pmap_qremove(kva, pcount);
1925 for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
1926 nextoff = toff + PAGE_SIZE;
1929 m->flags &= ~PG_ZERO;
1931 if (nextoff <= nread) {
1932 m->valid = VM_PAGE_BITS_ALL;
1934 } else if (toff < nread) {
1936 * Since this is a VM request, we have to supply the
1937 * unaligned offset to allow vm_page_set_validclean()
1938 * to zero sub-DEV_BSIZE'd portions of the page.
1940 vm_page_set_validclean(m, 0, nread - toff);
1946 if (i != ap->a_reqpage) {
1948 * Just in case someone was asking for this page we
1949 * now tell them that it is ok to use.
1951 if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
1953 if (m->flags & PG_WANTED) {
1954 vm_page_activate(m);
1956 vm_page_deactivate(m);
1965 } else if (m->valid) {
1968 * Since this is a VM request, we need to make the
1969 * entire page presentable by zeroing invalid sections.
1971 if (m->valid != VM_PAGE_BITS_ALL)
1972 vm_page_zero_invalid(m, FALSE);
1976 m = ap->a_m[ap->a_reqpage];
1977 devfs_debug(DEVFS_DEBUG_WARNING,
1978 "spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
1979 devtoname(vp->v_rdev), error, bp, bp->b_vp);
1980 devfs_debug(DEVFS_DEBUG_WARNING,
1981 " size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
1982 size, bp->b_resid, ap->a_count, m->valid);
1983 devfs_debug(DEVFS_DEBUG_WARNING,
1984 " nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
1985 nread, ap->a_reqpage, (u_long)m->pindex, pcount);
1987 * Free the buffer header back to the swap buffer pool.
1990 return VM_PAGER_ERROR;
1993 * Free the buffer header back to the swap buffer pool.
2039 sequential_heuristic(struct uio *uio, struct file *fp)
2042 * Sequential heuristic - detect sequential operation
2044 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
2045 uio->uio_offset == fp->f_nextoff) {
2046 int tmpseq = fp->f_seqcount;
2048 * XXX we assume that the filesystem block size is
2049 * the default. Not true, but still gives us a pretty
2050 * good indicator of how sequential the read operations
2053 tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
2054 if (tmpseq > IO_SEQMAX)
2056 fp->f_seqcount = tmpseq;
2057 return(fp->f_seqcount << IO_SEQSHIFT);
2061 * Not sequential, quick draw-down of seqcount
2063 if (fp->f_seqcount > 1)