freebsd-dev/sys/fs/devfs/devfs_vnops.c

1381 lines
32 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2000-2004
* Poul-Henning Kamp. All rights reserved.
* Copyright (c) 1989, 1992-1993, 1995
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software donated to Berkeley by
* Jan-Simon Pendry.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)kernfs_vnops.c 8.15 (Berkeley) 5/21/95
* From: FreeBSD: src/sys/miscfs/kernfs/kernfs_vnops.c 1.43
*
* $FreeBSD$
*/
/*
* TODO:
* remove empty directories
* mkdir: want it ?
*/
#include "opt_mac.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/conf.h>
#include <sys/dirent.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/filio.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/namei.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/stat.h>
#include <sys/sx.h>
#include <sys/time.h>
#include <sys/ttycom.h>
#include <sys/unistd.h>
#include <sys/vnode.h>
2005-09-12 08:03:15 +00:00
static struct vop_vector devfs_vnodeops;
static struct vop_vector devfs_specops;
static struct fileops devfs_ops_f;
2005-09-12 08:03:15 +00:00
#include <fs/devfs/devfs.h>
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
#include <fs/devfs/devfs_int.h>
#include <security/mac/mac_framework.h>
struct mtx devfs_de_interlock;
MTX_SYSINIT(devfs_de_interlock, &devfs_de_interlock, "devfs interlock", MTX_DEF);
static int
devfs_fp_check(struct file *fp, struct cdev **devp, struct cdevsw **dswp)
{
*dswp = devvn_refthread(fp->f_vnode, devp);
if (*devp != fp->f_data) {
if (*dswp != NULL)
dev_relthread(*devp);
return (ENXIO);
}
KASSERT((*devp)->si_refcount > 0,
("devfs: un-referenced struct cdev *(%s)", devtoname(*devp)));
if (*dswp == NULL)
return (ENXIO);
return (0);
}
/*
* Construct the fully qualified path name relative to the mountpoint
*/
static char *
devfs_fqpn(char *buf, struct vnode *dvp, struct componentname *cnp)
{
int i;
struct devfs_dirent *de, *dd;
struct devfs_mount *dmp;
dmp = VFSTODEVFS(dvp->v_mount);
dd = dvp->v_data;
i = SPECNAMELEN;
buf[i] = '\0';
i -= cnp->cn_namelen;
if (i < 0)
return (NULL);
bcopy(cnp->cn_nameptr, buf + i, cnp->cn_namelen);
de = dd;
while (de != dmp->dm_rootdir) {
i--;
if (i < 0)
return (NULL);
buf[i] = '/';
i -= de->de_dirent->d_namlen;
if (i < 0)
return (NULL);
bcopy(de->de_dirent->d_name, buf + i,
de->de_dirent->d_namlen);
de = TAILQ_FIRST(&de->de_dlist); /* "." */
de = TAILQ_NEXT(de, de_list); /* ".." */
de = de->de_dir;
}
return (buf + i);
}
static int
devfs_allocv_drop_refs(int drop_dm_lock, struct devfs_mount *dmp,
struct devfs_dirent *de)
{
int not_found;
not_found = 0;
if (de->de_flags & DE_DOOMED)
not_found = 1;
if (DEVFS_DE_DROP(de)) {
KASSERT(not_found == 1, ("DEVFS de dropped but not doomed"));
devfs_dirent_free(de);
}
if (DEVFS_DMP_DROP(dmp)) {
KASSERT(not_found == 1,
("DEVFS mount struct freed before dirent"));
not_found = 2;
sx_xunlock(&dmp->dm_lock);
devfs_unmount_final(dmp);
}
if (not_found == 1 || (drop_dm_lock && not_found != 2))
sx_unlock(&dmp->dm_lock);
return (not_found);
}
/*
* devfs_allocv shall be entered with dmp->dm_lock held, and it drops
* it on return.
*/
int
devfs_allocv(struct devfs_dirent *de, struct mount *mp, struct vnode **vpp, struct thread *td)
{
int error;
struct vnode *vp;
struct cdev *dev;
struct devfs_mount *dmp;
KASSERT(td == curthread, ("devfs_allocv: td != curthread"));
dmp = VFSTODEVFS(mp);
if (de->de_flags & DE_DOOMED) {
sx_xunlock(&dmp->dm_lock);
return (ENOENT);
}
loop:
DEVFS_DE_HOLD(de);
DEVFS_DMP_HOLD(dmp);
mtx_lock(&devfs_de_interlock);
vp = de->de_vnode;
if (vp != NULL) {
VI_LOCK(vp);
mtx_unlock(&devfs_de_interlock);
sx_xunlock(&dmp->dm_lock);
error = vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td);
sx_xlock(&dmp->dm_lock);
if (devfs_allocv_drop_refs(0, dmp, de)) {
if (error == 0)
vput(vp);
return (ENOENT);
}
else if (error)
goto loop;
sx_xunlock(&dmp->dm_lock);
*vpp = vp;
return (0);
}
mtx_unlock(&devfs_de_interlock);
if (de->de_dirent->d_type == DT_CHR) {
if (!(de->de_cdp->cdp_flags & CDP_ACTIVE)) {
devfs_allocv_drop_refs(1, dmp, de);
return (ENOENT);
}
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
dev = &de->de_cdp->cdp_c;
} else {
dev = NULL;
}
error = getnewvnode("devfs", mp, &devfs_vnodeops, &vp);
if (error != 0) {
devfs_allocv_drop_refs(1, dmp, de);
printf("devfs_allocv: failed to allocate new vnode\n");
return (error);
}
if (de->de_dirent->d_type == DT_CHR) {
vp->v_type = VCHR;
The remaining part of nmount/omount/rootfs mount changes. I cannot sensibly split the conversion of the remaining three filesystems out from the root mounting changes, so in one go: cd9660: Convert to nmount. Add omount compat shims. Remove dedicated rootfs mounting code. Use vfs_mountedfrom() Rely on vfs_mount.c calling VFS_STATFS() nfs(client): Convert to nmount (the simple way, mount_nfs(8) is still necessary). Add omount compat shims. Drop COMPAT_PRELITE2 mount arg compatibility. ffs: Convert to nmount. Add omount compat shims. Remove dedicated rootfs mounting code. Use vfs_mountedfrom() Rely on vfs_mount.c calling VFS_STATFS() Remove vfs_omount() method, all filesystems are now converted. Remove MNTK_WANTRDWR, handling RO/RW conversions is a filesystem task, and they all do it now. Change rootmounting to use DEVFS trampoline: vfs_mount.c: Mount devfs on /. Devfs needs no 'from' so this is clean. symlink /dev to /. This makes it possible to lookup /dev/foo. Mount "real" root filesystem on /. Surgically move the devfs mountpoint from under the real root filesystem onto /dev in the real root filesystem. Remove now unnecessary getdiskbyname(). kern_init.c: Don't do devfs mounting and rootvnode assignment here, it was already handled by vfs_mount.c. Remove now unused bdevvp(), addaliasu() and addalias(). Put the few necessary lines in devfs where they belong. This eliminates the second-last source of bogo vnodes, leaving only the lemming-syncer. Remove rootdev variable, it doesn't give meaning in a global context and was not trustworth anyway. Correct information is provided by statfs(/).
2004-12-07 08:15:41 +00:00
VI_LOCK(vp);
dev_lock();
2005-03-31 06:51:54 +00:00
dev_refl(dev);
The remaining part of nmount/omount/rootfs mount changes. I cannot sensibly split the conversion of the remaining three filesystems out from the root mounting changes, so in one go: cd9660: Convert to nmount. Add omount compat shims. Remove dedicated rootfs mounting code. Use vfs_mountedfrom() Rely on vfs_mount.c calling VFS_STATFS() nfs(client): Convert to nmount (the simple way, mount_nfs(8) is still necessary). Add omount compat shims. Drop COMPAT_PRELITE2 mount arg compatibility. ffs: Convert to nmount. Add omount compat shims. Remove dedicated rootfs mounting code. Use vfs_mountedfrom() Rely on vfs_mount.c calling VFS_STATFS() Remove vfs_omount() method, all filesystems are now converted. Remove MNTK_WANTRDWR, handling RO/RW conversions is a filesystem task, and they all do it now. Change rootmounting to use DEVFS trampoline: vfs_mount.c: Mount devfs on /. Devfs needs no 'from' so this is clean. symlink /dev to /. This makes it possible to lookup /dev/foo. Mount "real" root filesystem on /. Surgically move the devfs mountpoint from under the real root filesystem onto /dev in the real root filesystem. Remove now unnecessary getdiskbyname(). kern_init.c: Don't do devfs mounting and rootvnode assignment here, it was already handled by vfs_mount.c. Remove now unused bdevvp(), addaliasu() and addalias(). Put the few necessary lines in devfs where they belong. This eliminates the second-last source of bogo vnodes, leaving only the lemming-syncer. Remove rootdev variable, it doesn't give meaning in a global context and was not trustworth anyway. Correct information is provided by statfs(/).
2004-12-07 08:15:41 +00:00
vp->v_rdev = dev;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
KASSERT(vp->v_usecount == 1,
("%s %d (%d)\n", __func__, __LINE__, vp->v_usecount));
The remaining part of nmount/omount/rootfs mount changes. I cannot sensibly split the conversion of the remaining three filesystems out from the root mounting changes, so in one go: cd9660: Convert to nmount. Add omount compat shims. Remove dedicated rootfs mounting code. Use vfs_mountedfrom() Rely on vfs_mount.c calling VFS_STATFS() nfs(client): Convert to nmount (the simple way, mount_nfs(8) is still necessary). Add omount compat shims. Drop COMPAT_PRELITE2 mount arg compatibility. ffs: Convert to nmount. Add omount compat shims. Remove dedicated rootfs mounting code. Use vfs_mountedfrom() Rely on vfs_mount.c calling VFS_STATFS() Remove vfs_omount() method, all filesystems are now converted. Remove MNTK_WANTRDWR, handling RO/RW conversions is a filesystem task, and they all do it now. Change rootmounting to use DEVFS trampoline: vfs_mount.c: Mount devfs on /. Devfs needs no 'from' so this is clean. symlink /dev to /. This makes it possible to lookup /dev/foo. Mount "real" root filesystem on /. Surgically move the devfs mountpoint from under the real root filesystem onto /dev in the real root filesystem. Remove now unnecessary getdiskbyname(). kern_init.c: Don't do devfs mounting and rootvnode assignment here, it was already handled by vfs_mount.c. Remove now unused bdevvp(), addaliasu() and addalias(). Put the few necessary lines in devfs where they belong. This eliminates the second-last source of bogo vnodes, leaving only the lemming-syncer. Remove rootdev variable, it doesn't give meaning in a global context and was not trustworth anyway. Correct information is provided by statfs(/).
2004-12-07 08:15:41 +00:00
dev->si_usecount += vp->v_usecount;
dev_unlock();
VI_UNLOCK(vp);
vp->v_op = &devfs_specops;
} else if (de->de_dirent->d_type == DT_DIR) {
vp->v_type = VDIR;
} else if (de->de_dirent->d_type == DT_LNK) {
vp->v_type = VLNK;
} else {
vp->v_type = VBAD;
}
mtx_lock(&devfs_de_interlock);
vp->v_data = de;
de->de_vnode = vp;
mtx_unlock(&devfs_de_interlock);
sx_xunlock(&dmp->dm_lock);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
sx_xlock(&dmp->dm_lock);
if (devfs_allocv_drop_refs(0, dmp, de)) {
vput(vp);
return (ENOENT);
}
#ifdef MAC
mac_associate_vnode_devfs(mp, de, vp);
#endif
sx_xunlock(&dmp->dm_lock);
*vpp = vp;
return (0);
}
static int
2005-09-12 08:03:15 +00:00
devfs_access(struct vop_access_args *ap)
{
struct vnode *vp = ap->a_vp;
struct devfs_dirent *de;
int error;
de = vp->v_data;
if (vp->v_type == VDIR)
de = de->de_dir;
error = vaccess(vp->v_type, de->de_mode, de->de_uid, de->de_gid,
ap->a_mode, ap->a_cred, NULL);
if (!error)
return (error);
if (error != EACCES)
return (error);
/* We do, however, allow access to the controlling terminal */
if (!(ap->a_td->td_proc->p_flag & P_CONTROLT))
return (error);
if (ap->a_td->td_proc->p_session->s_ttyvp == de->de_vnode)
return (0);
return (error);
}
/* ARGSUSED */
static int
2005-09-12 08:03:15 +00:00
devfs_advlock(struct vop_advlock_args *ap)
{
return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL);
}
/* ARGSUSED */
static int
2005-09-12 08:03:15 +00:00
devfs_close(struct vop_close_args *ap)
{
struct vnode *vp = ap->a_vp, *oldvp;
struct thread *td = ap->a_td;
struct cdev *dev = vp->v_rdev;
struct cdevsw *dsw;
int vp_locked, error;
/*
* Hack: a tty device that is a controlling terminal
* has a reference from the session structure.
* We cannot easily tell that a character device is
* a controlling terminal, unless it is the closing
* process' controlling terminal. In that case,
* if the reference count is 2 (this last descriptor
* plus the session), release the reference from the session.
*/
oldvp = NULL;
sx_xlock(&proctree_lock);
if (td && vp == td->td_proc->p_session->s_ttyvp) {
SESS_LOCK(td->td_proc->p_session);
VI_LOCK(vp);
if (count_dev(dev) == 2 && (vp->v_iflag & VI_DOOMED) == 0) {
td->td_proc->p_session->s_ttyvp = NULL;
oldvp = vp;
}
VI_UNLOCK(vp);
SESS_UNLOCK(td->td_proc->p_session);
}
sx_xunlock(&proctree_lock);
if (oldvp != NULL)
vrele(oldvp);
/*
* We do not want to really close the device if it
* is still in use unless we are trying to close it
* forcibly. Since every use (buffer, vnode, swap, cmap)
* holds a reference to the vnode, and because we mark
* any other vnodes that alias this device, when the
* sum of the reference counts on all the aliased
* vnodes descends to one, we are on last close.
*/
dsw = dev_refthread(dev);
if (dsw == NULL)
return (ENXIO);
VI_LOCK(vp);
if (vp->v_iflag & VI_DOOMED) {
/* Forced close. */
} else if (dsw->d_flags & D_TRACKCLOSE) {
/* Keep device updated on status. */
} else if (count_dev(dev) > 1) {
VI_UNLOCK(vp);
dev_relthread(dev);
return (0);
}
vholdl(vp);
VI_UNLOCK(vp);
vp_locked = VOP_ISLOCKED(vp, td);
VOP_UNLOCK(vp, 0, td);
KASSERT(dev->si_refcount > 0,
("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev)));
if (!(dsw->d_flags & D_NEEDGIANT)) {
DROP_GIANT();
error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td);
PICKUP_GIANT();
} else {
error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td);
}
dev_relthread(dev);
vn_lock(vp, vp_locked | LK_RETRY, td);
vdrop(vp);
return (error);
}
static int
devfs_close_f(struct file *fp, struct thread *td)
{
return (vnops.fo_close(fp, td));
}
/* ARGSUSED */
static int
2005-09-12 08:03:15 +00:00
devfs_fsync(struct vop_fsync_args *ap)
{
if (!vn_isdisk(ap->a_vp, NULL))
return (0);
return (vop_stdfsync(ap));
}
static int
2005-09-12 08:03:15 +00:00
devfs_getattr(struct vop_getattr_args *ap)
{
struct vnode *vp = ap->a_vp;
struct vattr *vap = ap->a_vap;
int error = 0;
struct devfs_dirent *de;
struct cdev *dev;
de = vp->v_data;
KASSERT(de != NULL, ("Null dirent in devfs_getattr vp=%p", vp));
if (vp->v_type == VDIR) {
de = de->de_dir;
KASSERT(de != NULL,
("Null dir dirent in devfs_getattr vp=%p", vp));
}
bzero((caddr_t) vap, sizeof(*vap));
vattr_null(vap);
vap->va_uid = de->de_uid;
vap->va_gid = de->de_gid;
vap->va_mode = de->de_mode;
if (vp->v_type == VLNK)
vap->va_size = strlen(de->de_symlink);
else if (vp->v_type == VDIR)
vap->va_size = vap->va_bytes = DEV_BSIZE;
else
vap->va_size = 0;
if (vp->v_type != VDIR)
vap->va_bytes = 0;
vap->va_blocksize = DEV_BSIZE;
vap->va_type = vp->v_type;
#define fix(aa) \
do { \
if ((aa).tv_sec == 0) { \
(aa).tv_sec = boottime.tv_sec; \
(aa).tv_nsec = boottime.tv_usec * 1000; \
} \
} while (0)
if (vp->v_type != VCHR) {
fix(de->de_atime);
vap->va_atime = de->de_atime;
fix(de->de_mtime);
vap->va_mtime = de->de_mtime;
fix(de->de_ctime);
vap->va_ctime = de->de_ctime;
} else {
dev = vp->v_rdev;
fix(dev->si_atime);
vap->va_atime = dev->si_atime;
fix(dev->si_mtime);
vap->va_mtime = dev->si_mtime;
fix(dev->si_ctime);
vap->va_ctime = dev->si_ctime;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
vap->va_rdev = dev->si_priv->cdp_inode;
}
vap->va_gen = 0;
vap->va_flags = 0;
vap->va_nlink = de->de_links;
vap->va_fileid = de->de_inode;
return (error);
}
/* ARGSUSED */
static int
devfs_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td)
{
struct cdev *dev;
struct cdevsw *dsw;
struct vnode *vp;
struct vnode *vpold;
int error, i;
const char *p;
struct fiodgname_arg *fgn;
error = devfs_fp_check(fp, &dev, &dsw);
if (error)
return (error);
if (com == FIODTYPE) {
*(int *)data = dsw->d_flags & D_TYPEMASK;
dev_relthread(dev);
return (0);
} else if (com == FIODGNAME) {
fgn = data;
p = devtoname(dev);
i = strlen(p) + 1;
if (i > fgn->len)
error = EINVAL;
else
error = copyout(p, fgn->buf, i);
dev_relthread(dev);
return (error);
}
error = dsw->d_ioctl(dev, com, data, fp->f_flag, td);
dev_relthread(dev);
if (error == ENOIOCTL)
error = ENOTTY;
if (error == 0 && com == TIOCSCTTY) {
vp = fp->f_vnode;
/* Do nothing if reassigning same control tty */
sx_slock(&proctree_lock);
if (td->td_proc->p_session->s_ttyvp == vp) {
sx_sunlock(&proctree_lock);
return (0);
}
mtx_lock(&Giant);
vpold = td->td_proc->p_session->s_ttyvp;
VREF(vp);
SESS_LOCK(td->td_proc->p_session);
td->td_proc->p_session->s_ttyvp = vp;
SESS_UNLOCK(td->td_proc->p_session);
sx_sunlock(&proctree_lock);
/* Get rid of reference to old control tty */
if (vpold)
vrele(vpold);
mtx_unlock(&Giant);
}
return (error);
}
/* ARGSUSED */
static int
devfs_kqfilter_f(struct file *fp, struct knote *kn)
{
struct cdev *dev;
struct cdevsw *dsw;
int error;
error = devfs_fp_check(fp, &dev, &dsw);
if (error)
return (error);
error = dsw->d_kqfilter(dev, kn);
dev_relthread(dev);
return (error);
}
static int
devfs_lookupx(struct vop_lookup_args *ap, int *dm_unlock)
{
struct componentname *cnp;
struct vnode *dvp, **vpp;
struct thread *td;
struct devfs_dirent *de, *dd;
struct devfs_dirent **dde;
struct devfs_mount *dmp;
struct cdev *cdev;
int error, flags, nameiop;
char specname[SPECNAMELEN + 1], *pname;
cnp = ap->a_cnp;
vpp = ap->a_vpp;
dvp = ap->a_dvp;
pname = cnp->cn_nameptr;
td = cnp->cn_thread;
flags = cnp->cn_flags;
nameiop = cnp->cn_nameiop;
dmp = VFSTODEVFS(dvp->v_mount);
dd = dvp->v_data;
*vpp = NULLVP;
if ((flags & ISLASTCN) && nameiop == RENAME)
return (EOPNOTSUPP);
if (dvp->v_type != VDIR)
return (ENOTDIR);
if ((flags & ISDOTDOT) && (dvp->v_vflag & VV_ROOT))
return (EIO);
error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td);
if (error)
return (error);
if (cnp->cn_namelen == 1 && *pname == '.') {
if ((flags & ISLASTCN) && nameiop != LOOKUP)
return (EINVAL);
*vpp = dvp;
VREF(dvp);
return (0);
}
if (flags & ISDOTDOT) {
if ((flags & ISLASTCN) && nameiop != LOOKUP)
return (EINVAL);
VOP_UNLOCK(dvp, 0, td);
de = TAILQ_FIRST(&dd->de_dlist); /* "." */
de = TAILQ_NEXT(de, de_list); /* ".." */
de = de->de_dir;
error = devfs_allocv(de, dvp->v_mount, vpp, td);
*dm_unlock = 0;
vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
return (error);
}
DEVFS_DMP_HOLD(dmp);
devfs_populate(dmp);
if (DEVFS_DMP_DROP(dmp)) {
*dm_unlock = 0;
sx_xunlock(&dmp->dm_lock);
devfs_unmount_final(dmp);
return (ENOENT);
}
dd = dvp->v_data;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen);
while (de == NULL) { /* While(...) so we can use break */
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
if (nameiop == DELETE)
return (ENOENT);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
/*
* OK, we didn't have an entry for the name we were asked for
* so we try to see if anybody can create it on demand.
*/
pname = devfs_fqpn(specname, dvp, cnp);
if (pname == NULL)
break;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
cdev = NULL;
EVENTHANDLER_INVOKE(dev_clone,
td->td_ucred, pname, strlen(pname), &cdev);
if (cdev == NULL)
break;
DEVFS_DMP_HOLD(dmp);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
devfs_populate(dmp);
if (DEVFS_DMP_DROP(dmp)) {
*dm_unlock = 0;
sx_xunlock(&dmp->dm_lock);
devfs_unmount_final(dmp);
return (ENOENT);
}
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
dev_lock();
dde = &cdev->si_priv->cdp_dirents[dmp->dm_idx];
if (dde != NULL && *dde != NULL)
de = *dde;
dev_unlock();
dev_rel(cdev);
break;
}
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
if (de == NULL || de->de_flags & DE_WHITEOUT) {
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & (LOCKPARENT | WANTPARENT)) && (flags & ISLASTCN)) {
cnp->cn_flags |= SAVENAME;
return (EJUSTRETURN);
}
return (ENOENT);
}
if ((cnp->cn_nameiop == DELETE) && (flags & ISLASTCN)) {
error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td);
if (error)
return (error);
if (*vpp == dvp) {
VREF(dvp);
*vpp = dvp;
return (0);
}
}
error = devfs_allocv(de, dvp->v_mount, vpp, td);
*dm_unlock = 0;
return (error);
}
static int
devfs_lookup(struct vop_lookup_args *ap)
{
int j;
struct devfs_mount *dmp;
int dm_unlock;
dmp = VFSTODEVFS(ap->a_dvp->v_mount);
dm_unlock = 1;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xlock(&dmp->dm_lock);
j = devfs_lookupx(ap, &dm_unlock);
if (dm_unlock == 1)
sx_xunlock(&dmp->dm_lock);
return (j);
}
static int
devfs_mknod(struct vop_mknod_args *ap)
{
struct componentname *cnp;
struct vnode *dvp, **vpp;
struct thread *td;
struct devfs_dirent *dd, *de;
struct devfs_mount *dmp;
int error;
/*
* The only type of node we should be creating here is a
* character device, for anything else return EOPNOTSUPP.
*/
if (ap->a_vap->va_type != VCHR)
return (EOPNOTSUPP);
dvp = ap->a_dvp;
dmp = VFSTODEVFS(dvp->v_mount);
cnp = ap->a_cnp;
vpp = ap->a_vpp;
td = cnp->cn_thread;
dd = dvp->v_data;
error = ENOENT;
sx_xlock(&dmp->dm_lock);
TAILQ_FOREACH(de, &dd->de_dlist, de_list) {
if (cnp->cn_namelen != de->de_dirent->d_namlen)
continue;
if (bcmp(cnp->cn_nameptr, de->de_dirent->d_name,
de->de_dirent->d_namlen) != 0)
continue;
if (de->de_flags & DE_WHITEOUT)
break;
goto notfound;
}
if (de == NULL)
goto notfound;
de->de_flags &= ~DE_WHITEOUT;
error = devfs_allocv(de, dvp->v_mount, vpp, td);
return (error);
notfound:
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xunlock(&dmp->dm_lock);
return (error);
}
/* ARGSUSED */
static int
2005-09-12 08:03:15 +00:00
devfs_open(struct vop_open_args *ap)
{
struct thread *td = ap->a_td;
struct vnode *vp = ap->a_vp;
struct cdev *dev = vp->v_rdev;
struct file *fp;
int error;
struct cdevsw *dsw;
if (vp->v_type == VBLK)
return (ENXIO);
if (dev == NULL)
return (ENXIO);
/* Make this field valid before any I/O in d_open. */
if (dev->si_iosize_max == 0)
dev->si_iosize_max = DFLTPHYS;
dsw = dev_refthread(dev);
if (dsw == NULL)
return (ENXIO);
/* XXX: Special casing of ttys for deadfs. Probably redundant. */
if (dsw->d_flags & D_TTY)
vp->v_vflag |= VV_ISTTY;
VOP_UNLOCK(vp, 0, td);
if(!(dsw->d_flags & D_NEEDGIANT)) {
DROP_GIANT();
if (dsw->d_fdopen != NULL)
error = dsw->d_fdopen(dev, ap->a_mode, td, ap->a_fdidx);
else
error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td);
PICKUP_GIANT();
} else {
if (dsw->d_fdopen != NULL)
error = dsw->d_fdopen(dev, ap->a_mode, td, ap->a_fdidx);
else
error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td);
}
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
dev_relthread(dev);
if (error)
return (error);
#if 0 /* /dev/console */
KASSERT(ap->a_fdidx >= 0,
("Could not vnode bypass device on fd %d", ap->a_fdidx));
#else
if(ap->a_fdidx < 0)
return (error);
#endif
/*
* This is a pretty disgustingly long chain, but I am not
* sure there is any better way. Passing the fdidx into
* VOP_OPEN() offers us more information than just passing
* the file *.
*/
fp = ap->a_td->td_proc->p_fd->fd_ofiles[ap->a_fdidx];
KASSERT(fp->f_ops == &badfileops,
("Could not vnode bypass device on fdops %p", fp->f_ops));
fp->f_ops = &devfs_ops_f;
fp->f_data = dev;
return (error);
}
static int
2005-09-12 08:03:15 +00:00
devfs_pathconf(struct vop_pathconf_args *ap)
{
switch (ap->a_name) {
case _PC_MAC_PRESENT:
#ifdef MAC
/*
* If MAC is enabled, devfs automatically supports
* trivial non-persistant label storage.
*/
*ap->a_retval = 1;
#else
*ap->a_retval = 0;
#endif
return (0);
default:
return (vop_stdpathconf(ap));
}
/* NOTREACHED */
}
/* ARGSUSED */
static int
devfs_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td)
{
struct cdev *dev;
struct cdevsw *dsw;
int error;
error = devfs_fp_check(fp, &dev, &dsw);
if (error)
return (error);
error = dsw->d_poll(dev, events, td);
dev_relthread(dev);
return(error);
}
/*
* Print out the contents of a special device vnode.
*/
static int
2005-09-12 08:03:15 +00:00
devfs_print(struct vop_print_args *ap)
{
printf("\tdev %s\n", devtoname(ap->a_vp->v_rdev));
return (0);
}
/* ARGSUSED */
static int
devfs_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
{
struct cdev *dev;
int ioflag, error, resid;
struct cdevsw *dsw;
error = devfs_fp_check(fp, &dev, &dsw);
if (error)
return (error);
resid = uio->uio_resid;
ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT);
if (ioflag & O_DIRECT)
ioflag |= IO_DIRECT;
if ((flags & FOF_OFFSET) == 0)
uio->uio_offset = fp->f_offset;
error = dsw->d_read(dev, uio, ioflag);
if (uio->uio_resid != resid || (error == 0 && resid != 0))
vfs_timestamp(&dev->si_atime);
dev_relthread(dev);
if ((flags & FOF_OFFSET) == 0)
fp->f_offset = uio->uio_offset;
fp->f_nextoff = uio->uio_offset;
return (error);
}
static int
2005-09-12 08:03:15 +00:00
devfs_readdir(struct vop_readdir_args *ap)
{
int error;
struct uio *uio;
struct dirent *dp;
struct devfs_dirent *dd;
struct devfs_dirent *de;
struct devfs_mount *dmp;
off_t off, oldoff;
int *tmp_ncookies = NULL;
if (ap->a_vp->v_type != VDIR)
return (ENOTDIR);
uio = ap->a_uio;
if (uio->uio_offset < 0)
return (EINVAL);
/*
* XXX: This is a temporary hack to get around this filesystem not
* supporting cookies. We store the location of the ncookies pointer
* in a temporary variable before calling vfs_subr.c:vfs_read_dirent()
* and set the number of cookies to 0. We then set the pointer to
* NULL so that vfs_read_dirent doesn't try to call realloc() on
* ap->a_cookies. Later in this function, we restore the ap->a_ncookies
* pointer to its original location before returning to the caller.
*/
if (ap->a_ncookies != NULL) {
tmp_ncookies = ap->a_ncookies;
*ap->a_ncookies = 0;
ap->a_ncookies = NULL;
}
dmp = VFSTODEVFS(ap->a_vp->v_mount);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xlock(&dmp->dm_lock);
DEVFS_DMP_HOLD(dmp);
devfs_populate(dmp);
if (DEVFS_DMP_DROP(dmp)) {
sx_xunlock(&dmp->dm_lock);
devfs_unmount_final(dmp);
if (tmp_ncookies != NULL)
ap->a_ncookies = tmp_ncookies;
return (EIO);
}
error = 0;
de = ap->a_vp->v_data;
off = 0;
oldoff = uio->uio_offset;
TAILQ_FOREACH(dd, &de->de_dlist, de_list) {
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__));
if (dd->de_flags & DE_WHITEOUT)
continue;
if (dd->de_dirent->d_type == DT_DIR)
de = dd->de_dir;
else
de = dd;
dp = dd->de_dirent;
if (dp->d_reclen > uio->uio_resid)
break;
dp->d_fileno = de->de_inode;
if (off >= uio->uio_offset) {
2005-09-12 08:03:15 +00:00
error = vfs_read_dirent(ap, dp, off);
if (error)
break;
}
off += dp->d_reclen;
}
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xunlock(&dmp->dm_lock);
uio->uio_offset = off;
/*
* Restore ap->a_ncookies if it wasn't originally NULL in the first
* place.
*/
if (tmp_ncookies != NULL)
ap->a_ncookies = tmp_ncookies;
return (error);
}
static int
2005-09-12 08:03:15 +00:00
devfs_readlink(struct vop_readlink_args *ap)
{
struct devfs_dirent *de;
de = ap->a_vp->v_data;
2005-09-15 10:28:19 +00:00
return (uiomove(de->de_symlink, strlen(de->de_symlink), ap->a_uio));
}
static int
2005-09-12 08:03:15 +00:00
devfs_reclaim(struct vop_reclaim_args *ap)
{
struct vnode *vp = ap->a_vp;
struct devfs_dirent *de;
struct cdev *dev;
mtx_lock(&devfs_de_interlock);
de = vp->v_data;
if (de != NULL) {
de->de_vnode = NULL;
vp->v_data = NULL;
}
mtx_unlock(&devfs_de_interlock);
vnode_destroy_vobject(vp);
dev_lock();
dev = vp->v_rdev;
vp->v_rdev = NULL;
if (dev == NULL) {
dev_unlock();
return (0);
}
dev->si_usecount -= vp->v_usecount;
dev_unlock();
dev_rel(dev);
return (0);
}
static int
2005-09-12 08:03:15 +00:00
devfs_remove(struct vop_remove_args *ap)
{
struct vnode *vp = ap->a_vp;
struct devfs_dirent *dd;
struct devfs_dirent *de;
struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xlock(&dmp->dm_lock);
dd = ap->a_dvp->v_data;
de = vp->v_data;
if (de->de_cdp == NULL) {
TAILQ_REMOVE(&dd->de_dlist, de, de_list);
devfs_delete(dmp, de, 1);
} else {
de->de_flags |= DE_WHITEOUT;
}
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xunlock(&dmp->dm_lock);
return (0);
}
/*
* Revoke is called on a tty when a terminal session ends. The vnode
* is orphaned by setting v_op to deadfs so we need to let go of it
* as well so that we create a new one next time around.
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
*
*/
static int
2005-09-12 08:03:15 +00:00
devfs_revoke(struct vop_revoke_args *ap)
{
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
struct vnode *vp = ap->a_vp, *vp2;
struct cdev *dev;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
struct cdev_priv *cdp;
struct devfs_dirent *de;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
int i;
KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL"));
dev = vp->v_rdev;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
cdp = dev->si_priv;
dev_lock();
cdp->cdp_inuse++;
dev_unlock();
vhold(vp);
vgone(vp);
vdrop(vp);
VOP_UNLOCK(vp,0,curthread);
loop:
for (;;) {
mtx_lock(&devfs_de_interlock);
dev_lock();
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
vp2 = NULL;
for (i = 0; i <= cdp->cdp_maxdirent; i++) {
de = cdp->cdp_dirents[i];
if (de == NULL)
continue;
vp2 = de->de_vnode;
if (vp2 != NULL) {
dev_unlock();
VI_LOCK(vp2);
mtx_unlock(&devfs_de_interlock);
if (vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK,
curthread))
goto loop;
vhold(vp2);
vgone(vp2);
vdrop(vp2);
vput(vp2);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
break;
}
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
}
if (vp2 != NULL) {
continue;
}
dev_unlock();
mtx_unlock(&devfs_de_interlock);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
break;
}
dev_lock();
cdp->cdp_inuse--;
if (!(cdp->cdp_flags & CDP_ACTIVE) && cdp->cdp_inuse == 0) {
TAILQ_REMOVE(&cdevp_list, cdp, cdp_list);
dev_unlock();
dev_rel(&cdp->cdp_c);
} else
dev_unlock();
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, curthread);
return (0);
}
static int
2005-09-12 08:03:15 +00:00
devfs_rioctl(struct vop_ioctl_args *ap)
{
int error;
struct devfs_mount *dmp;
dmp = VFSTODEVFS(ap->a_vp->v_mount);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xlock(&dmp->dm_lock);
DEVFS_DMP_HOLD(dmp);
devfs_populate(dmp);
if (DEVFS_DMP_DROP(dmp)) {
sx_xunlock(&dmp->dm_lock);
devfs_unmount_final(dmp);
return (ENOENT);
}
2005-09-15 10:28:19 +00:00
error = devfs_rules_ioctl(dmp, ap->a_command, ap->a_data, ap->a_td);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xunlock(&dmp->dm_lock);
return (error);
}
static int
2005-09-12 08:03:15 +00:00
devfs_rread(struct vop_read_args *ap)
{
if (ap->a_vp->v_type != VDIR)
return (EINVAL);
return (VOP_READDIR(ap->a_vp, ap->a_uio, ap->a_cred, NULL, NULL, NULL));
}
static int
2005-09-12 08:03:15 +00:00
devfs_setattr(struct vop_setattr_args *ap)
{
struct devfs_dirent *de;
struct vattr *vap;
struct vnode *vp;
int c, error;
uid_t uid;
gid_t gid;
vap = ap->a_vap;
vp = ap->a_vp;
if ((vap->va_type != VNON) ||
(vap->va_nlink != VNOVAL) ||
(vap->va_fsid != VNOVAL) ||
(vap->va_fileid != VNOVAL) ||
(vap->va_blocksize != VNOVAL) ||
(vap->va_flags != VNOVAL && vap->va_flags != 0) ||
(vap->va_rdev != VNOVAL) ||
((int)vap->va_bytes != VNOVAL) ||
(vap->va_gen != VNOVAL)) {
return (EINVAL);
}
de = vp->v_data;
if (vp->v_type == VDIR)
de = de->de_dir;
error = c = 0;
if (vap->va_uid == (uid_t)VNOVAL)
uid = de->de_uid;
else
uid = vap->va_uid;
if (vap->va_gid == (gid_t)VNOVAL)
gid = de->de_gid;
else
gid = vap->va_gid;
if (uid != de->de_uid || gid != de->de_gid) {
if ((ap->a_cred->cr_uid != de->de_uid) || uid != de->de_uid ||
(gid != de->de_gid && !groupmember(gid, ap->a_cred))) {
error = priv_check_cred(ap->a_td->td_ucred,
PRIV_VFS_CHOWN, SUSER_ALLOWJAIL);
if (error)
return (error);
}
de->de_uid = uid;
de->de_gid = gid;
c = 1;
}
if (vap->va_mode != (mode_t)VNOVAL) {
if (ap->a_cred->cr_uid != de->de_uid) {
error = priv_check_cred(ap->a_td->td_ucred,
PRIV_VFS_ADMIN, SUSER_ALLOWJAIL);
if (error)
return (error);
}
de->de_mode = vap->va_mode;
c = 1;
}
if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
/* See the comment in ufs_vnops::ufs_setattr(). */
if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, ap->a_td)) &&
((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
(error = VOP_ACCESS(vp, VWRITE, ap->a_cred, ap->a_td))))
return (error);
if (vap->va_atime.tv_sec != VNOVAL) {
if (vp->v_type == VCHR)
vp->v_rdev->si_atime = vap->va_atime;
else
de->de_atime = vap->va_atime;
}
if (vap->va_mtime.tv_sec != VNOVAL) {
if (vp->v_type == VCHR)
vp->v_rdev->si_mtime = vap->va_mtime;
else
de->de_mtime = vap->va_mtime;
}
c = 1;
}
if (c) {
if (vp->v_type == VCHR)
vfs_timestamp(&vp->v_rdev->si_ctime);
else
vfs_timestamp(&de->de_mtime);
}
return (0);
}
#ifdef MAC
static int
2005-09-12 08:03:15 +00:00
devfs_setlabel(struct vop_setlabel_args *ap)
{
struct vnode *vp;
struct devfs_dirent *de;
vp = ap->a_vp;
de = vp->v_data;
mac_relabel_vnode(ap->a_cred, vp, ap->a_label);
mac_update_devfsdirent(vp->v_mount, de, vp);
return (0);
}
#endif
static int
devfs_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td)
{
return (vnops.fo_stat(fp, sb, cred, td));
}
static int
2005-09-12 08:03:15 +00:00
devfs_symlink(struct vop_symlink_args *ap)
{
int i, error;
struct devfs_dirent *dd;
struct devfs_dirent *de;
struct devfs_mount *dmp;
struct thread *td;
td = ap->a_cnp->cn_thread;
KASSERT(td == curthread, ("devfs_symlink: td != curthread"));
error = priv_check(td, PRIV_DEVFS_SYMLINK);
if (error)
return(error);
dmp = VFSTODEVFS(ap->a_dvp->v_mount);
dd = ap->a_dvp->v_data;
de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen);
de->de_uid = 0;
de->de_gid = 0;
de->de_mode = 0755;
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
de->de_inode = alloc_unr(devfs_inos);
de->de_dirent->d_type = DT_LNK;
i = strlen(ap->a_target) + 1;
2005-09-15 10:28:19 +00:00
de->de_symlink = malloc(i, M_DEVFS, M_WAITOK);
bcopy(ap->a_target, de->de_symlink, i);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
sx_xlock(&dmp->dm_lock);
#ifdef MAC
mac_create_devfs_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de);
#endif
TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list);
return (devfs_allocv(de, ap->a_dvp->v_mount, ap->a_vpp, td));
}
/* ARGSUSED */
static int
devfs_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td)
{
struct cdev *dev;
int error, ioflag, resid;
struct cdevsw *dsw;
error = devfs_fp_check(fp, &dev, &dsw);
if (error)
return (error);
KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT | O_FSYNC);
if (ioflag & O_DIRECT)
ioflag |= IO_DIRECT;
if ((flags & FOF_OFFSET) == 0)
uio->uio_offset = fp->f_offset;
resid = uio->uio_resid;
error = dsw->d_write(dev, uio, ioflag);
if (uio->uio_resid != resid || (error == 0 && resid != 0)) {
vfs_timestamp(&dev->si_ctime);
dev->si_mtime = dev->si_ctime;
}
dev_relthread(dev);
if ((flags & FOF_OFFSET) == 0)
fp->f_offset = uio->uio_offset;
fp->f_nextoff = uio->uio_offset;
return (error);
}
dev_t
dev2udev(struct cdev *x)
{
if (x == NULL)
return (NODEV);
Rewamp DEVFS internals pretty severely [1]. Give DEVFS a proper inode called struct cdev_priv. It is important to keep in mind that this "inode" is shared between all DEVFS mountpoints, therefore it is protected by the global device mutex. Link the cdev_priv's into a list, protected by the global device mutex. Keep track of each cdev_priv's state with a flag bit and of references from mountpoints with a dedicated usecount. Reap the benefits of much improved kernel memory allocator and the generally better defined device driver APIs to get rid of the tables of pointers + serial numbers, their overflow tables, the atomics to muck about in them and all the trouble that resulted in. This makes RAM the only limit on how many devices we can have. The cdev_priv is actually a super struct containing the normal cdev as the "public" part, and therefore allocation and freeing has moved to devfs_devs.c from kern_conf.c. The overall responsibility is (to be) split such that kern/kern_conf.c is the stuff that deals with drivers and struct cdev and fs/devfs handles filesystems and struct cdev_priv and their private liason exposed only in devfs_int.h. Move the inode number from cdev to cdev_priv and allocate inode numbers properly with unr. Local dirents in the mountpoints (directories, symlinks) allocate inodes from the same pool to guarantee against overlaps. Various other fields are going to migrate from cdev to cdev_priv in the future in order to hide them. A few fields may migrate from devfs_dirent to cdev_priv as well. Protect the DEVFS mountpoint with an sx lock instead of lockmgr, this lock also protects the directory tree of the mountpoint. Give each mountpoint a unique integer index, allocated with unr. Use it into an array of devfs_dirent pointers in each cdev_priv. Initially the array points to a single element also inside cdev_priv, but as more devfs instances are mounted, the array is extended with malloc(9) as necessary when the filesystem populates its directory tree. Retire the cdev alias lists, the cdev_priv now know about all the relevant devfs_dirents (and their vnodes) and devfs_revoke() will pick them up from there. We still spelunk into other mountpoints and fondle their data without 100% good locking. It may make better sense to vector the revoke event into the tty code and there do a destroy_dev/make_dev on the tty's devices, but that's for further study. Lots of shuffling of stuff and churn of bits for no good reason[2]. XXX: There is still nothing preventing the dev_clone EVENTHANDLER from being invoked at the same time in two devfs mountpoints. It is not obvious what the best course of action is here. XXX: comment out an if statement that lost its body, until I can find out what should go there so it doesn't do damage in the meantime. XXX: Leave in a few extra malloc types and KASSERTS to help track down any remaining issues. Much testing provided by: Kris Much confusion caused by (races in): md(4) [1] You are not supposed to understand anything past this point. [2] This line should simplify life for the peanut gallery.
2005-09-19 19:56:48 +00:00
return (x->si_priv->cdp_inode);
}
2005-09-12 08:03:15 +00:00
static struct fileops devfs_ops_f = {
.fo_read = devfs_read_f,
.fo_write = devfs_write_f,
.fo_ioctl = devfs_ioctl_f,
.fo_poll = devfs_poll_f,
.fo_kqfilter = devfs_kqfilter_f,
.fo_stat = devfs_stat_f,
.fo_close = devfs_close_f,
.fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
};
static struct vop_vector devfs_vnodeops = {
.vop_default = &default_vnodeops,
.vop_access = devfs_access,
.vop_getattr = devfs_getattr,
.vop_ioctl = devfs_rioctl,
.vop_lookup = devfs_lookup,
.vop_mknod = devfs_mknod,
.vop_pathconf = devfs_pathconf,
.vop_read = devfs_rread,
.vop_readdir = devfs_readdir,
.vop_readlink = devfs_readlink,
.vop_reclaim = devfs_reclaim,
.vop_remove = devfs_remove,
.vop_revoke = devfs_revoke,
.vop_setattr = devfs_setattr,
#ifdef MAC
.vop_setlabel = devfs_setlabel,
#endif
.vop_symlink = devfs_symlink,
};
static struct vop_vector devfs_specops = {
.vop_default = &default_vnodeops,
.vop_access = devfs_access,
.vop_advlock = devfs_advlock,
.vop_bmap = VOP_PANIC,
.vop_close = devfs_close,
.vop_create = VOP_PANIC,
.vop_fsync = devfs_fsync,
.vop_getattr = devfs_getattr,
.vop_lease = VOP_NULL,
.vop_link = VOP_PANIC,
.vop_mkdir = VOP_PANIC,
.vop_mknod = VOP_PANIC,
.vop_open = devfs_open,
.vop_pathconf = devfs_pathconf,
.vop_print = devfs_print,
.vop_read = VOP_PANIC,
.vop_readdir = VOP_PANIC,
.vop_readlink = VOP_PANIC,
.vop_reallocblks = VOP_PANIC,
.vop_reclaim = devfs_reclaim,
.vop_remove = devfs_remove,
.vop_rename = VOP_PANIC,
.vop_revoke = devfs_revoke,
.vop_rmdir = VOP_PANIC,
.vop_setattr = devfs_setattr,
#ifdef MAC
.vop_setlabel = devfs_setlabel,
#endif
.vop_strategy = VOP_PANIC,
.vop_symlink = VOP_PANIC,
.vop_write = VOP_PANIC,
};
/*
* Our calling convention to the device drivers used to be that we passed
* vnode.h IO_* flags to read()/write(), but we're moving to fcntl.h O_
* flags instead since that's what open(), close() and ioctl() takes and
* we don't really want vnode.h in device drivers.
* We solved the source compatibility by redefining some vnode flags to
* be the same as the fcntl ones and by sending down the bitwise OR of
* the respective fcntl/vnode flags. These CTASSERTS make sure nobody
* pulls the rug out under this.
*/
CTASSERT(O_NONBLOCK == IO_NDELAY);
CTASSERT(O_FSYNC == IO_SYNC);