freebsd-dev/sys/kern/kern_conf.c

908 lines
20 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1999-2002 Poul-Henning Kamp
* All rights reserved.
*
* 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. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/module.h>
#include <sys/malloc.h>
#include <sys/conf.h>
#include <sys/vnode.h>
#include <sys/queue.h>
#include <sys/poll.h>
#include <sys/ctype.h>
#include <sys/tty.h>
When devfs cloning takes place, provide access to the credential of the process that caused the clone event to take place for the device driver creating the device. This allows cloned device drivers to adapt the device node based on security aspects of the process, such as the uid, gid, and MAC label. - Add a cred reference to struct cdev, so that when a device node is instantiated as a vnode, the cloning credential can be exposed to MAC. - Add make_dev_cred(), a version of make_dev() that additionally accepts the credential to stick in the struct cdev. Implement it and make_dev() in terms of a back-end make_dev_credv(). - Add a new event handler, dev_clone_cred, which can be registered to receive the credential instead of dev_clone, if desired. - Modify the MAC entry point mac_create_devfs_device() to accept an optional credential pointer (may be NULL), so that MAC policies can inspect and act on the label or other elements of the credential when initializing the skeleton device protections. - Modify tty_pty.c to register clone_dev_cred and invoke make_dev_cred(), so that the pty clone credential is exposed to the MAC Framework. While currently primarily focussed on MAC policies, this change is also a prerequisite for changes to allow ptys to be instantiated with the UID of the process looking up the pty. This requires further changes to the pty driver -- in particular, to immediately recycle pty nodes on last close so that the credential-related state can be recreated on next lookup. Submitted by: Andrew Reisse <andrew.reisse@sparta.com> Obtained from: TrustedBSD Project Sponsored by: SPAWAR, SPARTA MFC after: 1 week MFC note: Merge to 6.x, but not 5.x for ABI reasons
2005-07-14 10:22:09 +00:00
#include <sys/ucred.h>
#include <machine/stdarg.h>
#include <fs/devfs/devfs_int.h>
static MALLOC_DEFINE(M_DEVT, "cdev", "cdev storage");
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 mtx devmtx;
static void destroy_devl(struct cdev *dev);
When devfs cloning takes place, provide access to the credential of the process that caused the clone event to take place for the device driver creating the device. This allows cloned device drivers to adapt the device node based on security aspects of the process, such as the uid, gid, and MAC label. - Add a cred reference to struct cdev, so that when a device node is instantiated as a vnode, the cloning credential can be exposed to MAC. - Add make_dev_cred(), a version of make_dev() that additionally accepts the credential to stick in the struct cdev. Implement it and make_dev() in terms of a back-end make_dev_credv(). - Add a new event handler, dev_clone_cred, which can be registered to receive the credential instead of dev_clone, if desired. - Modify the MAC entry point mac_create_devfs_device() to accept an optional credential pointer (may be NULL), so that MAC policies can inspect and act on the label or other elements of the credential when initializing the skeleton device protections. - Modify tty_pty.c to register clone_dev_cred and invoke make_dev_cred(), so that the pty clone credential is exposed to the MAC Framework. While currently primarily focussed on MAC policies, this change is also a prerequisite for changes to allow ptys to be instantiated with the UID of the process looking up the pty. This requires further changes to the pty driver -- in particular, to immediately recycle pty nodes on last close so that the credential-related state can be recreated on next lookup. Submitted by: Andrew Reisse <andrew.reisse@sparta.com> Obtained from: TrustedBSD Project Sponsored by: SPAWAR, SPARTA MFC after: 1 week MFC note: Merge to 6.x, but not 5.x for ABI reasons
2005-07-14 10:22:09 +00:00
static struct cdev *make_dev_credv(struct cdevsw *devsw, int minornr,
struct ucred *cr, uid_t uid, gid_t gid, int mode, const char *fmt,
va_list ap);
void
dev_lock(void)
{
mtx_lock(&devmtx);
}
void
dev_unlock(void)
{
mtx_unlock(&devmtx);
}
void
dev_ref(struct cdev *dev)
{
mtx_assert(&devmtx, MA_NOTOWNED);
mtx_lock(&devmtx);
dev->si_refcount++;
mtx_unlock(&devmtx);
}
void
2005-03-31 06:51:54 +00:00
dev_refl(struct cdev *dev)
{
mtx_assert(&devmtx, MA_OWNED);
dev->si_refcount++;
}
void
dev_rel(struct cdev *dev)
{
int flag = 0;
2004-10-01 06:33:39 +00:00
mtx_assert(&devmtx, MA_NOTOWNED);
dev_lock();
dev->si_refcount--;
KASSERT(dev->si_refcount >= 0,
("dev_rel(%s) gave negative count", devtoname(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
#if 0
if (dev->si_usecount == 0 &&
(dev->si_flags & SI_CHEAPCLONE) && (dev->si_flags & SI_NAMED))
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
;
else
#endif
2006-01-04 17:40:54 +00:00
if (dev->si_devsw == NULL && dev->si_refcount == 0) {
LIST_REMOVE(dev, si_list);
2004-10-01 06:33:39 +00:00
flag = 1;
}
2004-10-01 06:33:39 +00:00
dev_unlock();
if (flag)
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_free(dev);
}
2004-10-01 06:33:39 +00:00
struct cdevsw *
dev_refthread(struct cdev *dev)
{
struct cdevsw *csw;
mtx_assert(&devmtx, MA_NOTOWNED);
dev_lock();
csw = dev->si_devsw;
if (csw != NULL)
dev->si_threadcount++;
dev_unlock();
return (csw);
}
struct cdevsw *
devvn_refthread(struct vnode *vp, struct cdev **devp)
{
struct cdevsw *csw;
mtx_assert(&devmtx, MA_NOTOWNED);
csw = NULL;
dev_lock();
*devp = vp->v_rdev;
if (*devp != NULL) {
csw = (*devp)->si_devsw;
if (csw != NULL)
(*devp)->si_threadcount++;
}
dev_unlock();
return (csw);
}
void
dev_relthread(struct cdev *dev)
{
mtx_assert(&devmtx, MA_NOTOWNED);
dev_lock();
dev->si_threadcount--;
dev_unlock();
}
int
nullop(void)
{
return (0);
}
int
eopnotsupp(void)
{
return (EOPNOTSUPP);
}
static int
enxio(void)
{
return (ENXIO);
}
static int
enodev(void)
{
return (ENODEV);
}
/* Define a dead_cdevsw for use when devices leave unexpectedly. */
#define dead_open (d_open_t *)enxio
#define dead_close (d_close_t *)enxio
#define dead_read (d_read_t *)enxio
#define dead_write (d_write_t *)enxio
#define dead_ioctl (d_ioctl_t *)enxio
#define dead_poll (d_poll_t *)enodev
#define dead_mmap (d_mmap_t *)enodev
static void
dead_strategy(struct bio *bp)
{
biofinish(bp, NULL, ENXIO);
}
#define dead_dump (dumper_t *)enxio
#define dead_kqfilter (d_kqfilter_t *)enxio
static struct cdevsw dead_cdevsw = {
.d_version = D_VERSION,
.d_flags = D_NEEDGIANT, /* XXX: does dead_strategy need this ? */
.d_open = dead_open,
.d_close = dead_close,
.d_read = dead_read,
.d_write = dead_write,
.d_ioctl = dead_ioctl,
.d_poll = dead_poll,
.d_mmap = dead_mmap,
.d_strategy = dead_strategy,
.d_name = "dead",
.d_dump = dead_dump,
.d_kqfilter = dead_kqfilter
};
/* Default methods if driver does not specify method */
#define null_open (d_open_t *)nullop
#define null_close (d_close_t *)nullop
#define no_read (d_read_t *)enodev
#define no_write (d_write_t *)enodev
#define no_ioctl (d_ioctl_t *)enodev
#define no_mmap (d_mmap_t *)enodev
#define no_kqfilter (d_kqfilter_t *)enodev
static void
no_strategy(struct bio *bp)
{
biofinish(bp, NULL, ENODEV);
}
static int
no_poll(struct cdev *dev __unused, int events, struct thread *td __unused)
{
/*
* Return true for read/write. If the user asked for something
* special, return POLLNVAL, so that clients have a way of
* determining reliably whether or not the extended
* functionality is present without hard-coding knowledge
* of specific filesystem implementations.
* Stay in sync with vop_nopoll().
*/
if (events & ~POLLSTANDARD)
return (POLLNVAL);
return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
}
#define no_dump (dumper_t *)enodev
static int
giant_open(struct cdev *dev, int oflags, int devtype, struct thread *td)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_open(dev, oflags, devtype, td);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_fdopen(struct cdev *dev, int oflags, struct thread *td, struct file *fp)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_fdopen(dev, oflags, td, fp);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_close(struct cdev *dev, int fflag, int devtype, struct thread *td)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_close(dev, fflag, devtype, td);
mtx_unlock(&Giant);
return (retval);
}
static void
giant_strategy(struct bio *bp)
{
mtx_lock(&Giant);
bp->bio_dev->si_devsw->d_gianttrick->
d_strategy(bp);
mtx_unlock(&Giant);
}
static int
giant_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_ioctl(dev, cmd, data, fflag, td);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_read(struct cdev *dev, struct uio *uio, int ioflag)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_read(dev, uio, ioflag);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_write(struct cdev *dev, struct uio *uio, int ioflag)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_write(dev, uio, ioflag);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_poll(struct cdev *dev, int events, struct thread *td)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_poll(dev, events, td);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_kqfilter(struct cdev *dev, struct knote *kn)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_kqfilter(dev, kn);
mtx_unlock(&Giant);
return (retval);
}
static int
giant_mmap(struct cdev *dev, vm_offset_t offset, vm_paddr_t *paddr, int nprot)
{
int retval;
mtx_lock(&Giant);
retval = dev->si_devsw->d_gianttrick->
d_mmap(dev, offset, paddr, nprot);
mtx_unlock(&Giant);
return (retval);
}
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
/*
* struct cdev * and u_dev_t primitives
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
*/
int
minor(struct cdev *x)
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
{
if (x == NULL)
return NODEV;
return(x->si_drv0 & MAXMINOR);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
}
int
dev2unit(struct cdev *x)
{
if (x == NULL)
return NODEV;
2005-01-29 15:10:30 +00:00
return (minor2unit(minor(x)));
}
2005-03-08 10:40:03 +00:00
u_int
minor2unit(u_int _minor)
{
2005-01-29 16:50:04 +00:00
KASSERT((_minor & ~MAXMINOR) == 0, ("Illegal minor %x", _minor));
2005-03-08 10:40:03 +00:00
return ((_minor & 0xff) | ((_minor >> 8) & 0xffff00));
}
int
unit2minor(int unit)
{
KASSERT(unit <= 0xffffff, ("Invalid unit (%d) in unit2minor", unit));
return ((unit & 0xff) | ((unit << 8) & ~0xffff));
}
static struct cdev *
newdev(struct cdevsw *csw, int y, struct cdev *si)
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
{
struct cdev *si2;
dev_t udev;
mtx_assert(&devmtx, MA_OWNED);
udev = y;
LIST_FOREACH(si2, &csw->d_devs, si_list) {
if (si2->si_drv0 == udev) {
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_free(si);
return (si2);
}
}
si->si_drv0 = udev;
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
si->si_devsw = csw;
LIST_INSERT_HEAD(&csw->d_devs, si, si_list);
return (si);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
}
int
uminor(dev_t dev)
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
{
2005-01-29 16:50:04 +00:00
return (dev & MAXMINOR);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
}
int
umajor(dev_t dev)
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
{
2005-01-29 16:50:04 +00:00
return ((dev & ~MAXMINOR) >> 8);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
}
static void
fini_cdevsw(struct cdevsw *devsw)
{
struct cdevsw *gt;
if (devsw->d_gianttrick != NULL) {
gt = devsw->d_gianttrick;
memcpy(devsw, gt, sizeof *devsw);
free(gt, M_DEVT);
devsw->d_gianttrick = NULL;
}
devsw->d_flags &= ~D_INIT;
}
static void
prep_cdevsw(struct cdevsw *devsw)
{
struct cdevsw *dsw2;
if (devsw->d_flags & D_NEEDGIANT)
dsw2 = malloc(sizeof *dsw2, M_DEVT, M_WAITOK);
else
dsw2 = NULL;
dev_lock();
if (devsw->d_version != D_VERSION_01) {
printf(
"WARNING: Device driver \"%s\" has wrong version %s\n",
devsw->d_name == NULL ? "???" : devsw->d_name,
"and is disabled. Recompile KLD module.");
devsw->d_open = dead_open;
devsw->d_close = dead_close;
devsw->d_read = dead_read;
devsw->d_write = dead_write;
devsw->d_ioctl = dead_ioctl;
devsw->d_poll = dead_poll;
devsw->d_mmap = dead_mmap;
devsw->d_strategy = dead_strategy;
devsw->d_dump = dead_dump;
devsw->d_kqfilter = dead_kqfilter;
}
if (devsw->d_flags & D_TTY) {
if (devsw->d_ioctl == NULL) devsw->d_ioctl = ttyioctl;
if (devsw->d_read == NULL) devsw->d_read = ttyread;
if (devsw->d_write == NULL) devsw->d_write = ttywrite;
if (devsw->d_kqfilter == NULL) devsw->d_kqfilter = ttykqfilter;
if (devsw->d_poll == NULL) devsw->d_poll = ttypoll;
}
if (devsw->d_flags & D_NEEDGIANT) {
if (devsw->d_gianttrick == NULL) {
memcpy(dsw2, devsw, sizeof *dsw2);
devsw->d_gianttrick = dsw2;
} else
free(dsw2, M_DEVT);
}
#define FIXUP(member, noop, giant) \
do { \
if (devsw->member == NULL) { \
devsw->member = noop; \
} else if (devsw->d_flags & D_NEEDGIANT) \
devsw->member = giant; \
} \
while (0)
FIXUP(d_open, null_open, giant_open);
FIXUP(d_fdopen, NULL, giant_fdopen);
FIXUP(d_close, null_close, giant_close);
FIXUP(d_read, no_read, giant_read);
FIXUP(d_write, no_write, giant_write);
FIXUP(d_ioctl, no_ioctl, giant_ioctl);
FIXUP(d_poll, no_poll, giant_poll);
FIXUP(d_mmap, no_mmap, giant_mmap);
FIXUP(d_strategy, no_strategy, giant_strategy);
FIXUP(d_kqfilter, no_kqfilter, giant_kqfilter);
if (devsw->d_dump == NULL) devsw->d_dump = no_dump;
LIST_INIT(&devsw->d_devs);
devsw->d_flags |= D_INIT;
dev_unlock();
}
When devfs cloning takes place, provide access to the credential of the process that caused the clone event to take place for the device driver creating the device. This allows cloned device drivers to adapt the device node based on security aspects of the process, such as the uid, gid, and MAC label. - Add a cred reference to struct cdev, so that when a device node is instantiated as a vnode, the cloning credential can be exposed to MAC. - Add make_dev_cred(), a version of make_dev() that additionally accepts the credential to stick in the struct cdev. Implement it and make_dev() in terms of a back-end make_dev_credv(). - Add a new event handler, dev_clone_cred, which can be registered to receive the credential instead of dev_clone, if desired. - Modify the MAC entry point mac_create_devfs_device() to accept an optional credential pointer (may be NULL), so that MAC policies can inspect and act on the label or other elements of the credential when initializing the skeleton device protections. - Modify tty_pty.c to register clone_dev_cred and invoke make_dev_cred(), so that the pty clone credential is exposed to the MAC Framework. While currently primarily focussed on MAC policies, this change is also a prerequisite for changes to allow ptys to be instantiated with the UID of the process looking up the pty. This requires further changes to the pty driver -- in particular, to immediately recycle pty nodes on last close so that the credential-related state can be recreated on next lookup. Submitted by: Andrew Reisse <andrew.reisse@sparta.com> Obtained from: TrustedBSD Project Sponsored by: SPAWAR, SPARTA MFC after: 1 week MFC note: Merge to 6.x, but not 5.x for ABI reasons
2005-07-14 10:22:09 +00:00
static struct cdev *
make_dev_credv(struct cdevsw *devsw, int minornr, struct ucred *cr, uid_t uid,
gid_t gid, int mode, const char *fmt, va_list ap)
{
struct cdev *dev;
int i;
2005-01-29 16:50:04 +00:00
KASSERT((minornr & ~MAXMINOR) == 0,
("Invalid minor (0x%x) in make_dev", minornr));
if (!(devsw->d_flags & D_INIT))
prep_cdevsw(devsw);
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 = devfs_alloc();
dev_lock();
dev = newdev(devsw, minornr, dev);
if (dev->si_flags & SI_CHEAPCLONE &&
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->si_flags & SI_NAMED) {
/*
* This is allowed as it removes races and generally
* simplifies cloning devices.
* XXX: still ??
*/
dev_unlock();
return (dev);
}
KASSERT(!(dev->si_flags & SI_NAMED),
("make_dev() by driver %s on pre-existing device (min=%x, name=%s)",
devsw->d_name, minor(dev), devtoname(dev)));
i = vsnrprintf(dev->__si_namebuf, sizeof dev->__si_namebuf, 32, fmt, ap);
if (i > (sizeof dev->__si_namebuf - 1)) {
2004-08-30 01:10:20 +00:00
printf("WARNING: Device name truncated! (%s)\n",
dev->__si_namebuf);
}
dev->si_flags |= SI_NAMED;
When devfs cloning takes place, provide access to the credential of the process that caused the clone event to take place for the device driver creating the device. This allows cloned device drivers to adapt the device node based on security aspects of the process, such as the uid, gid, and MAC label. - Add a cred reference to struct cdev, so that when a device node is instantiated as a vnode, the cloning credential can be exposed to MAC. - Add make_dev_cred(), a version of make_dev() that additionally accepts the credential to stick in the struct cdev. Implement it and make_dev() in terms of a back-end make_dev_credv(). - Add a new event handler, dev_clone_cred, which can be registered to receive the credential instead of dev_clone, if desired. - Modify the MAC entry point mac_create_devfs_device() to accept an optional credential pointer (may be NULL), so that MAC policies can inspect and act on the label or other elements of the credential when initializing the skeleton device protections. - Modify tty_pty.c to register clone_dev_cred and invoke make_dev_cred(), so that the pty clone credential is exposed to the MAC Framework. While currently primarily focussed on MAC policies, this change is also a prerequisite for changes to allow ptys to be instantiated with the UID of the process looking up the pty. This requires further changes to the pty driver -- in particular, to immediately recycle pty nodes on last close so that the credential-related state can be recreated on next lookup. Submitted by: Andrew Reisse <andrew.reisse@sparta.com> Obtained from: TrustedBSD Project Sponsored by: SPAWAR, SPARTA MFC after: 1 week MFC note: Merge to 6.x, but not 5.x for ABI reasons
2005-07-14 10:22:09 +00:00
if (cr != NULL)
dev->si_cred = crhold(cr);
else
dev->si_cred = NULL;
dev->si_uid = uid;
dev->si_gid = gid;
dev->si_mode = mode;
devfs_create(dev);
dev_unlock();
return (dev);
}
When devfs cloning takes place, provide access to the credential of the process that caused the clone event to take place for the device driver creating the device. This allows cloned device drivers to adapt the device node based on security aspects of the process, such as the uid, gid, and MAC label. - Add a cred reference to struct cdev, so that when a device node is instantiated as a vnode, the cloning credential can be exposed to MAC. - Add make_dev_cred(), a version of make_dev() that additionally accepts the credential to stick in the struct cdev. Implement it and make_dev() in terms of a back-end make_dev_credv(). - Add a new event handler, dev_clone_cred, which can be registered to receive the credential instead of dev_clone, if desired. - Modify the MAC entry point mac_create_devfs_device() to accept an optional credential pointer (may be NULL), so that MAC policies can inspect and act on the label or other elements of the credential when initializing the skeleton device protections. - Modify tty_pty.c to register clone_dev_cred and invoke make_dev_cred(), so that the pty clone credential is exposed to the MAC Framework. While currently primarily focussed on MAC policies, this change is also a prerequisite for changes to allow ptys to be instantiated with the UID of the process looking up the pty. This requires further changes to the pty driver -- in particular, to immediately recycle pty nodes on last close so that the credential-related state can be recreated on next lookup. Submitted by: Andrew Reisse <andrew.reisse@sparta.com> Obtained from: TrustedBSD Project Sponsored by: SPAWAR, SPARTA MFC after: 1 week MFC note: Merge to 6.x, but not 5.x for ABI reasons
2005-07-14 10:22:09 +00:00
struct cdev *
make_dev(struct cdevsw *devsw, int minornr, uid_t uid, gid_t gid, int mode,
const char *fmt, ...)
{
struct cdev *dev;
va_list ap;
va_start(ap, fmt);
dev = make_dev_credv(devsw, minornr, NULL, uid, gid, mode, fmt, ap);
va_end(ap);
return (dev);
}
struct cdev *
make_dev_cred(struct cdevsw *devsw, int minornr, struct ucred *cr, uid_t uid,
gid_t gid, int mode, const char *fmt, ...)
{
struct cdev *dev;
va_list ap;
va_start(ap, fmt);
dev = make_dev_credv(devsw, minornr, cr, uid, gid, mode, fmt, ap);
va_end(ap);
return (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
static void
dev_dependsl(struct cdev *pdev, struct cdev *cdev)
{
cdev->si_parent = pdev;
cdev->si_flags |= SI_CHILD;
LIST_INSERT_HEAD(&pdev->si_children, cdev, si_siblings);
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
}
void
dev_depends(struct cdev *pdev, struct cdev *cdev)
{
dev_lock();
dev_dependsl(pdev, cdev);
dev_unlock();
}
struct cdev *
make_dev_alias(struct cdev *pdev, const char *fmt, ...)
{
struct cdev *dev;
va_list ap;
int 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
dev = devfs_alloc();
dev_lock();
dev->si_flags |= SI_ALIAS;
dev->si_flags |= SI_NAMED;
va_start(ap, fmt);
i = vsnrprintf(dev->__si_namebuf, sizeof dev->__si_namebuf, 32, fmt, ap);
if (i > (sizeof dev->__si_namebuf - 1)) {
2004-08-30 01:10:20 +00:00
printf("WARNING: Device name truncated! (%s)\n",
dev->__si_namebuf);
}
va_end(ap);
devfs_create(dev);
dev_unlock();
dev_depends(pdev, dev);
return (dev);
}
static void
destroy_devl(struct cdev *dev)
{
struct cdevsw *csw;
mtx_assert(&devmtx, MA_OWNED);
KASSERT(dev->si_flags & SI_NAMED,
("WARNING: Driver mistake: destroy_dev on %d\n", minor(dev)));
2006-01-04 17:40:54 +00:00
devfs_destroy(dev);
/* Remove name marking */
dev->si_flags &= ~SI_NAMED;
/* If we are a child, remove us from the parents list */
if (dev->si_flags & SI_CHILD) {
LIST_REMOVE(dev, si_siblings);
dev->si_flags &= ~SI_CHILD;
}
/* Kill our children */
while (!LIST_EMPTY(&dev->si_children))
destroy_devl(LIST_FIRST(&dev->si_children));
/* Remove from clone list */
if (dev->si_flags & SI_CLONELIST) {
LIST_REMOVE(dev, si_clone);
dev->si_flags &= ~SI_CLONELIST;
}
dev->si_refcount++; /* Avoid race with dev_rel() */
csw = dev->si_devsw;
dev->si_devsw = NULL; /* already NULL for SI_ALIAS */
while (csw != NULL && csw->d_purge != NULL && dev->si_threadcount) {
csw->d_purge(dev);
msleep(csw, &devmtx, PRIBIO, "devprg", hz/10);
if (dev->si_threadcount)
printf("Still %lu threads in %s\n",
dev->si_threadcount, devtoname(dev));
}
while (dev->si_threadcount != 0) {
/* Use unique dummy wait ident */
msleep(&csw, &devmtx, PRIBIO, "devdrn", hz / 10);
}
dev->si_drv1 = 0;
dev->si_drv2 = 0;
bzero(&dev->__si_u, sizeof(dev->__si_u));
if (!(dev->si_flags & SI_ALIAS)) {
/* Remove from cdevsw list */
LIST_REMOVE(dev, si_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
/* If cdevsw has no more struct cdev *'s, clean it */
2004-09-27 06:34:30 +00:00
if (LIST_EMPTY(&csw->d_devs))
fini_cdevsw(csw);
}
dev->si_flags &= ~SI_ALIAS;
dev->si_refcount--; /* Avoid race with dev_rel() */
if (dev->si_refcount > 0) {
LIST_INSERT_HEAD(&dead_cdevsw.d_devs, dev, si_list);
} else {
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_free(dev);
}
}
void
destroy_dev(struct cdev *dev)
{
dev_lock();
destroy_devl(dev);
dev_unlock();
}
const char *
devtoname(struct cdev *dev)
{
char *p;
struct cdevsw *csw;
int mynor;
if (dev->si_name[0] == '#' || dev->si_name[0] == '\0') {
p = dev->si_name;
csw = dev_refthread(dev);
if (csw != NULL) {
sprintf(p, "(%s)", csw->d_name);
dev_relthread(dev);
}
p += strlen(p);
mynor = minor(dev);
if (mynor < 0 || mynor > 255)
sprintf(p, "/%#x", (u_int)mynor);
else
sprintf(p, "/%d", mynor);
}
return (dev->si_name);
}
int
dev_stdclone(char *name, char **namep, const char *stem, int *unit)
{
int u, i;
i = strlen(stem);
2001-04-14 21:33:58 +00:00
if (bcmp(stem, name, i) != 0)
return (0);
if (!isdigit(name[i]))
return (0);
u = 0;
if (name[i] == '0' && isdigit(name[i+1]))
return (0);
while (isdigit(name[i])) {
u *= 10;
u += name[i++] - '0';
}
if (u > 0xffffff)
return (0);
*unit = u;
if (namep)
*namep = &name[i];
if (name[i])
return (2);
return (1);
}
/*
* Helper functions for cloning device drivers.
*
* The objective here is to make it unnecessary for the device drivers to
* use rman or similar to manage their unit number space. Due to the way
* we do "on-demand" devices, using rman or other "private" methods
* will be very tricky to lock down properly once we lock down this file.
*
* Instead we give the drivers these routines which puts the struct cdev *'s
* that are to be managed on their own list, and gives the driver the ability
* to ask for the first free unit number or a given specified unit number.
*
* In addition these routines support paired devices (pty, nmdm and similar)
* by respecting a number of "flag" bits in the minor number.
*
*/
struct clonedevs {
LIST_HEAD(,cdev) head;
};
void
clone_setup(struct clonedevs **cdp)
{
*cdp = malloc(sizeof **cdp, M_DEVBUF, M_WAITOK | M_ZERO);
LIST_INIT(&(*cdp)->head);
}
int
clone_create(struct clonedevs **cdp, struct cdevsw *csw, int *up, struct cdev **dp, int extra)
{
struct clonedevs *cd;
struct cdev *dev, *ndev, *dl, *de;
int unit, low, u;
KASSERT(*cdp != NULL,
("clone_setup() not called in driver \"%s\"", csw->d_name));
KASSERT(!(extra & CLONE_UNITMASK),
("Illegal extra bits (0x%x) in clone_create", extra));
KASSERT(*up <= CLONE_UNITMASK,
("Too high unit (0x%x) in clone_create", *up));
if (!(csw->d_flags & D_INIT))
prep_cdevsw(csw);
/*
* Search the list for a lot of things in one go:
* A preexisting match is returned immediately.
* The lowest free unit number if we are passed -1, and the place
* in the list where we should insert that new element.
* The place to insert a specified unit number, if applicable
* the end of the list.
*/
unit = *up;
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
ndev = devfs_alloc();
dev_lock();
low = extra;
de = dl = NULL;
cd = *cdp;
LIST_FOREACH(dev, &cd->head, si_clone) {
KASSERT(dev->si_flags & SI_CLONELIST,
("Dev %p(%s) should be on clonelist", dev, dev->si_name));
u = dev2unit(dev);
if (u == (unit | extra)) {
*dp = 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
devfs_free(ndev);
dev_unlock();
return (0);
}
if (unit == -1 && u == low) {
low++;
de = dev;
continue;
} else if (u < (unit | extra)) {
de = dev;
continue;
} else if (u > (unit | extra)) {
dl = dev;
break;
}
}
if (unit == -1)
unit = low & CLONE_UNITMASK;
dev = newdev(csw, unit2minor(unit | extra), ndev);
if (dev->si_flags & SI_CLONELIST) {
printf("dev %p (%s) is on clonelist\n", dev, dev->si_name);
printf("unit=%d, low=%d, extra=0x%x\n", unit, low, extra);
LIST_FOREACH(dev, &cd->head, si_clone) {
printf("\t%p %s\n", dev, dev->si_name);
}
panic("foo");
}
KASSERT(!(dev->si_flags & SI_CLONELIST),
("Dev %p(%s) should not be on clonelist", dev, dev->si_name));
if (dl != NULL)
LIST_INSERT_BEFORE(dl, dev, si_clone);
else if (de != NULL)
LIST_INSERT_AFTER(de, dev, si_clone);
else
LIST_INSERT_HEAD(&cd->head, dev, si_clone);
dev->si_flags |= SI_CLONELIST;
*up = unit;
dev_unlock();
return (1);
}
/*
* Kill everything still on the list. The driver should already have
* disposed of any softc hung of the struct cdev *'s at this time.
*/
void
clone_cleanup(struct clonedevs **cdp)
{
struct cdev *dev, *tdev;
struct clonedevs *cd;
cd = *cdp;
if (cd == NULL)
return;
dev_lock();
LIST_FOREACH_SAFE(dev, &cd->head, si_clone, tdev) {
KASSERT(dev->si_flags & SI_CLONELIST,
("Dev %p(%s) should be on clonelist", dev, dev->si_name));
KASSERT(dev->si_flags & SI_NAMED,
("Driver has goofed in cloning underways udev %x", dev->si_drv0));
destroy_devl(dev);
}
dev_unlock();
free(cd, M_DEVBUF);
*cdp = NULL;
}