Re: [patch 1/3] lsm: add bsdjail module

Re: [patch 1/3] lsm: add bsdjail module

[Colin Walters]

[-- Attachment #1: Type: text/plain, Size: 1612 bytes --]

[ Sorry for the lack of References, I'm not on this list and had to grab
the message from the web archives ]

> On Iau, 2004-10-07 at 00:26, Andrew Morton wrote:
> > I don't recall anyone requesting this feature.  Tell me why we should add
> > it to Linux?
> 
> Subject to the code cleanups and stuff you've noted I'd actually like to
> see BSD jail stuff in our security modules because it has the virtue of
> simplicity. If it can be extended to do all of vserver even better. J
> Random Admin has a good chance at configuring BSD jails etups. J Random
> Admin needs some serious tools that don't exist to set up SELinux the
> same way.
> 
> In the security world simplicity is often a virtue, both in code and
> concepts.

BSD jails are a kind of halfway point on the 
"access control <-> virtualization" sliding scale.  Certainly, using
SELinux for virtualization is not trivial, and I think that's fine; it
is not its intended purpose.  But let's be honest - using BSD jails for
strong access control ("security") has its own downsides.  For example,
now J Random Admin has N rpm/dpkg databases to manage on the same system
instead of one.  Or they have hand-rolled chroots which need their own
special treatment, scripts, etc.

In other words, BSD jails and SELinux are certainly not the same thing.
This becomes more obvious when you realize that hey - SELinux is still
useful inside a BSD jail.  Why should the cracked sendmail daemon be
able to destroy your configuration and the rest of the chroot?  You
might even want BSD jails to enforce different SELinux policies.


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [patch 1/3] lsm: add bsdjail module

[Herbert Poetzl]

On Thu, Oct 07, 2004 at 11:52:40AM -0700, Chris Wright wrote:
> * Andrew Morton (akpm@osdl.org) wrote:
> > Chris Wright <chrisw@osdl.org> wrote:
> > > * Andrew Morton (akpm@osdl.org) wrote:
> > > Which feature are you concerned over, the additional hook or the
> > > new module?
> > 
> > I am concerned about the presence of new code - simple as that.
> 
> Understood.
> 
> > We need to be able to demonstrate that the new code is sufficiently useful
> > to a sufficiently large number of people as to warrant the cost of
> > maintaining it in the tree for the rest of eternity.
> 
> That's fine.  Serge, can you enlighten us with an idea of the users of
> this code?
> 
> > >  The module is a no-op for anybody who doesn't want it.
> > 
> > It still needs to be maintained.
> 
> Absolutely.
> 
> > > I can't vouch for the number of users of this module although I've seen
> > > some positive feedback from users.  One nice bit is that it goes a way
> > > towards helping vserver which does have quite a few users.
> > 
> > Tell us more.
> 
> One portion of the vserver project (that which has to do with security
> and isolation) could be largely covered by this work.  And vserver
> is an active project with many users AFAICT.  The vserver maintainer
> has expressed some interest in this as well.  The other portion of the
> project, which does the resource limiting has a decent chance of working
> well with something like CKRM or similar.

well, as 'the vserver project' probably means the
'linux-vserver project', I would like to point out 
why and where the bsdjail LSM, in it's current form
is flawed from the linux-vserver point of view ...

Serge, don't get me wrong, this is neither against 
you nor against the bsdjail LSM, which I consider
an interesting approach, and I'm still confident 
that we find some way of cooperation ...

(copied the jail struct here to comment it)

| struct jail_struct {
| 	struct kref		kref;
|
|	/* these are set on writes to /proc/<pid>/attr/exec */
|	char *root_pathname; /* char * containing path to use as jail / */

linux-vserver uses namespaces to create the vservers, 
only the legacy method uses a simple chroot() to
setup the vserver environment ...

| 	char *ip4_addr_name;  /* char * containing ip4 addr to use for jail */
| 	char *ip6_addr_name;  /* char * containing ip6 addr to use for jail */

linux-vserver is slowly moving from chbind (which 
restricts a process and it's children to a set of 
IPs to an iptable (marking) based approach, which 
is much more flexible

| 	/* these are set when a jail becomes active */
| 	__u32 addr4;		/* internal form of ip4_addr_name */
| 	struct in6_addr addr6;  /* internal form of ip6_addr_name */

up to 16 addresses are currently allowed in this set
in the future the limit will go away (network code is
actually the oldest piece) by using 'markings'
(network is virtualized to allow binding to 0.0.0.0)

| 	struct dentry *dentry;  /* dentry of fs root */
| 	struct vfsmount *mnt;	/* vfsmnt of fs root */
 
| 	/* Resource limits.  0 = no limit */
| 	int max_nrtask; 	/* maximum number of tasks within this jail. */
| 	int cur_nrtask; 	/* current number of tasks within this jail. */

linux-vserver already has a nice and usable resource
management system for most resources, supporting
much more limits than those ...

| 	long maxtimeslice;	/* max timeslice in ms for procs in this jail */
|	long nice;		/* nice level for processes in this jail */
| 	long max_data, max_memlock;  /* equivalent to RLIMIT_{DATA,MEMLOCK} */

the resource limitations should not be part of a 
security module, and the scheduler slice would be
a step in the wrong redirection, as linux-vserver
already uses token buckets to control the scheduler

| 	char jail_flags;
|};

also many distributions (and distribution hosting
is _the_ main application area for linux-vserver)
require the 'jail' to be as similar as possible
to a real host (like a separate init process, or
the ability to renice services) so some of the
'features' of that LSM are contra productive here
not to mention that linux-vserver's security is
mainly based on linux capabilities which are not
handled by this LSM at all ...

aside from that, without the notion of a security
context, which can be controlled and entered from
outside (the host) the 'jail' can not be used for
typical hosting purposes 

> > >  This module
> > > really demonstrates one of the points of LSM...to support multiple
> > > security models.
> > 
> > Sure.  But that doesn't mean that those modules have to live at kernel.org
> > rather than, say, at bsdjail.sourceforge.net.
> 
> I agree, some userbase does wonders to justify mainlining the code.

I'm pretty confident the fact that a big company
seems interested in this LSM will help with the 
integration into mainline, but I can not say that 
'linux-vserver' users or developers will have any 
immediate benefit from it's inclusion ... why?

 - the linux-vserver user would have to apply an
   additional patch anyway (and install special
   tools to control the vservers)

 - the LSM does not provide what linux-vserver
   requires and would need heavy modification
   (missing context, namespaces, network, most
   virtualization, resource isolation)

 - once CKRM will be able to replace the resource
   management currently present in linux-vserver,
   it (CKRM) will collide with the resource stuff
   done in this LSM

so while I'm fine with the idea to move a part of
linux-vserver to the LSM framework (once the LSM
stackering issues are resolved), this part would
not be usable without a decent part of kernel
modifications to do the virtualization and the
resource isolation

of course if there _is_ interest to include 
linux-vserver like features into mainline, then 
there should be some commitment to do it properly 
and this includes not limiting it to a security
module, where it requires much more to be useful
to anyone ...

finally I have no problem to maintain the 'vserver'
patches outside the kernel tree, as they are 
probably only of limited interest for the typical
linux desktop user ...

best,
Herbert 
(linux-vserver maintainer)

PS: for details see the linux-vserver paper at
http://www.linux-vserver.org/index.php?page=Linux-VServer-Paper

> thanks,
> -chris
> -- 
> Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: [patch 1/3] lsm: add bsdjail module

[Serge E. Hallyn]

> * Andrew Morton (akpm@osdl.org) wrote:
> > Chris Wright <chrisw@osdl.org> wrote:
> > > * Andrew Morton (akpm@osdl.org) wrote:
> > > Which feature are you concerned over, the additional hook or the
> > > new module?
> > 
> > I am concerned about the presence of new code - simple as that.
> 
> Understood.

We do have time allotted for maintenance of LSMs, so not only am I
interested in maintaining bsdjail on my own, but I don't even have to do
it in my free time  :)

> > We need to be able to demonstrate that the new code is sufficiently useful
> > to a sufficiently large number of people as to warrant the cost of
> > maintaining it in the tree for the rest of eternity.
> 
> That's fine.  Serge, can you enlighten us with an idea of the users of
> this code?

I am "just a developer", and don't have ready access to any marketers.
There was no customer demand which we were addressing.  We just saw it
as a very useful feature easy to implement.  Some people have privately
expressed interested in the patch over the last few months as I've been
sending out patches.  And as Chris has mentioned, the vserver community
appears to be thriving, and should be partially (though by no means
fully!) served by this module.  If nothing else it should reduce the
size of the patch they need to maintain.

I wish I had a better answer...

-serge

Re: [patch 1/3] lsm: add bsdjail module

[Chris Wright]

* Andrew Morton (akpm@osdl.org) wrote:
> Chris Wright <chrisw@osdl.org> wrote:
> > * Andrew Morton (akpm@osdl.org) wrote:
> > Which feature are you concerned over, the additional hook or the
> > new module?
> 
> I am concerned about the presence of new code - simple as that.

Understood.

> We need to be able to demonstrate that the new code is sufficiently useful
> to a sufficiently large number of people as to warrant the cost of
> maintaining it in the tree for the rest of eternity.

That's fine.  Serge, can you enlighten us with an idea of the users of
this code?

> >  The module is a no-op for anybody who doesn't want it.
> 
> It still needs to be maintained.

Absolutely.

> > I can't vouch for the number of users of this module although I've seen
> > some positive feedback from users.  One nice bit is that it goes a way
> > towards helping vserver which does have quite a few users.
> 
> Tell us more.

One portion of the vserver project (that which has to do with security
and isolation) could be largely covered by this work.  And vserver
is an active project with many users AFAICT.  The vserver maintainer
has expressed some interest in this as well.  The other portion of the
project, which does the resource limiting has a decent chance of working
well with something like CKRM or similar.

> >  This module
> > really demonstrates one of the points of LSM...to support multiple
> > security models.
> 
> Sure.  But that doesn't mean that those modules have to live at kernel.org
> rather than, say, at bsdjail.sourceforge.net.

I agree, some userbase does wonders to justify mainlining the code.

thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net

Re: [patch 1/3] lsm: add bsdjail module

[Andrew Morton]

Chris Wright <chrisw@osdl.org> wrote:
>
> * Andrew Morton (akpm@osdl.org) wrote:
> > James Morris <jmorris@redhat.com> wrote:
> > > On Thu, 7 Oct 2004, Serge E. Hallyn wrote:
> > > 
> > >  > Because it gives Linux a functionality like FreeBSD's jail and Solaris'
> > >  > zones in an unobtrusive manner, without impacting users who don't wish
> > >  > to use it  (except for the extra security_task_lookup function calls).
> > > 
> > >  Yes, as an LSM module, it can be configured out.  I think it's a good use
> > >  of the LSM framework, and may be useful for people migrating to Linux from
> > >  legacy Solaris and FreeBSD.
> > 
> > Sure, but that's a bit speculative for adding a feature to the mainline
> > kernel.
> 
> Which feature are you concerned over, the additional hook or the
> new module?

I am concerned about the presence of new code - simple as that.

We need to be able to demonstrate that the new code is sufficiently useful
to a sufficiently large number of people as to warrant the cost of
maintaining it in the tree for the rest of eternity.

>  The module is a no-op for anybody who doesn't want it.

It still needs to be maintained.

> I can't vouch for the number of users of this module although I've seen
> some positive feedback from users.  One nice bit is that it goes a way
> towards helping vserver which does have quite a few users.

Tell us more.

>  This module
> really demonstrates one of the points of LSM...to support multiple
> security models.

Sure.  But that doesn't mean that those modules have to live at kernel.org
rather than, say, at bsdjail.sourceforge.net.

Re: [patch 1/3] lsm: add bsdjail module

[Chris Wright]

* Andrew Morton (akpm@osdl.org) wrote:
> James Morris <jmorris@redhat.com> wrote:
> > On Thu, 7 Oct 2004, Serge E. Hallyn wrote:
> > 
> >  > Because it gives Linux a functionality like FreeBSD's jail and Solaris'
> >  > zones in an unobtrusive manner, without impacting users who don't wish
> >  > to use it  (except for the extra security_task_lookup function calls).
> > 
> >  Yes, as an LSM module, it can be configured out.  I think it's a good use
> >  of the LSM framework, and may be useful for people migrating to Linux from
> >  legacy Solaris and FreeBSD.
> 
> Sure, but that's a bit speculative for adding a feature to the mainline
> kernel.

Which feature are you concerned over, the additional hook or the
new module?  The module is a no-op for anybody who doesn't want it.
I can't vouch for the number of users of this module although I've seen
some positive feedback from users.  One nice bit is that it goes a way
towards helping vserver which does have quite a few users.  This module
really demonstrates one of the points of LSM...to support multiple
security models.

thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net

Re: [patch 1/3] lsm: add bsdjail module

[Alan Cox]

On Iau, 2004-10-07 at 00:26, Andrew Morton wrote:
> I don't recall anyone requesting this feature.  Tell me why we should add
> it to Linux?

Subject to the code cleanups and stuff you've noted I'd actually like to
see BSD jail stuff in our security modules because it has the virtue of
simplicity. If it can be extended to do all of vserver even better. J
Random Admin has a good chance at configuring BSD jails etups. J Random
Admin needs some serious tools that don't exist to set up SELinux the
same way.

In the security world simplicity is often a virtue, both in code and
concepts.

Alan

Re: [patch 1/3] lsm: add bsdjail module

[Andrew Morton]

James Morris <jmorris@redhat.com> wrote:
>
> On Thu, 7 Oct 2004, Serge E. Hallyn wrote:
> 
>  > Because it gives Linux a functionality like FreeBSD's jail and Solaris'
>  > zones in an unobtrusive manner, without impacting users who don't wish
>  > to use it  (except for the extra security_task_lookup function calls).
> 
>  Yes, as an LSM module, it can be configured out.  I think it's a good use
>  of the LSM framework, and may be useful for people migrating to Linux from
>  legacy Solaris and FreeBSD.

Sure, but that's a bit speculative for adding a feature to the mainline
kernel.

Is there vendor pull for this feature?  Do IBM have customers requiring it?

"someone might like this" is not a sufficient basis for adding stuff to the
kernel, sorry.

Re: [patch 1/3] lsm: add bsdjail module

[James Morris]

On Thu, 7 Oct 2004, Serge E. Hallyn wrote:

> Because it gives Linux a functionality like FreeBSD's jail and Solaris'
> zones in an unobtrusive manner, without impacting users who don't wish
> to use it  (except for the extra security_task_lookup function calls).

Yes, as an LSM module, it can be configured out.  I think it's a good use
of the LSM framework, and may be useful for people migrating to Linux from
legacy Solaris and FreeBSD.


- James
-- 
James Morris
<jmorris@redhat.com>

Re: [patch 1/3] lsm: add bsdjail module

[Serge E. Hallyn]

Thank you for the feedback.  I have implemented these changes, but
want to run a few tests tomorrow before I send them out to make
sure I didn't break anything...

Quoting Andrew Morton (akpm@osdl.org):
> Serge Hallyn <serue@us.ibm.com> wrote:
> >
> > 
> > Attached is a patch against the security Kconfig and Makefile to support
> > bsdjail, as well as the bsdjail.c file itself. bsdjail offers
> > functionality similar to (but more limited than) the vserver patch.
> 
> I don't recall anyone requesting this feature.  Tell me why we should add
> it to Linux?

Because it gives Linux a functionality like FreeBSD's jail and Solaris'
zones in an unobtrusive manner, without impacting users who don't wish
to use it  (except for the extra security_task_lookup function calls).
It allows me (for instance) to compartmentalize apache and sendmail by
running them in different jails.  Or offer family members, customers,
or whoever, ssh accounts into seemingly distinct boxes, which are simply
sshd's under different jails at different network aliases.  Each would
see their own private filesystems and network, have their own usage
limits, and (mostly) not see processes outside their respective jails.
They can't {un,}load modules, ptrace unjailed processes or send signals
to them, create devices, mount, or umount.  It is functionality which
otherwise would have to be achieved by running vmware or uml, but far
more lightweight, since no new OS needs to be run.  (Once read-only bind
mounts are implemented, it will become even more lightweight, as large
pieces of filesystem trees will be shareable readonly between jails.)

thanks,
-serge

Re: [patch 1/3] lsm: add bsdjail module

[Andrew Morton]

Serge Hallyn <serue@us.ibm.com> wrote:
>
> 
> Attached is a patch against the security Kconfig and Makefile to support
> bsdjail, as well as the bsdjail.c file itself. bsdjail offers
> functionality similar to (but more limited than) the vserver patch.

I don't recall anyone requesting this feature.  Tell me why we should add
it to Linux?

> +
> +#define in_use(x) (x->jail_flags & IN_USE)
> +#define set_in_use(x) (x->jail_flags |= IN_USE)
> +
> +#define got_network(x) (x->jail_flags & (GOT_IPV4 | GOT_IPV6))
> +#define got_ipv4(x) (x->jail_flags & (GOT_IPV4))
> +#define got_ipv6(x) (x->jail_flags & (GOT_IPV6))
> +#define set_ipv4(x) (x->jail_flags |= GOT_IPV4)
> +#define set_ipv6(x) (x->jail_flags |= GOT_IPV6)
> +#define unset_got_ipv4(x) (x->jail_flags &= ~GOT_IPV4)
> +#define unset_got_ipv6(x) (x->jail_flags &= ~GOT_IPV6)
> +#define get_task_security(task) (task->security)
> +#define get_inode_security(inode) (inode->i_security)
> +#define get_sock_security(sock) (sock->sk_security)
> +#define get_file_security(file) (file->f_security)
> +#define get_ipc_security(ipc)	(ipc->security)
> +#define jail_of(proc) (get_task_security(proc))
> +

The above tricks may make the code easier to type, but I find they make the
code harder for others to read, and that's more important.  We prefer to
open-code such things.

> +	if (tsec->root_pathname)
> +		kfree(tsec->root_pathname);
> +	if (tsec->ip4_addr_name)
> +		kfree(tsec->ip4_addr_name);
> +	if (tsec->ip6_addr_name)
> +		kfree(tsec->ip6_addr_name);

kfree(0) is permitted.  Some people like to do the double test anyway but I
don't think it adds much here.

> +		set_task_security(task,NULL);

whitespace nit: In some places you have spaces after the commas and in
others you do not.

> +	kref_put(&tsec->kref, release_jail);

This is the preferred style.

Re: [patch 1/3] lsm: add bsdjail module

[Serge Hallyn]

Attached is a patch against the security Kconfig and Makefile to support
bsdjail, as well as the bsdjail.c file itself. bsdjail offers
functionality similar to (but more limited than) the vserver patch.

A process in a jail lives under a chroot which is not vulnerable to the
well-known chdir(...)(etc)chroot(.) attack against normal chroots, and
may be locked to one ip address. For additional features, please see
Documentation/bsdjail.txt, which is included in the next patch.

Changelog:
	Sep 10, 2004: original version
	Sep 12, 2004: add ipv6 support
	Sep 13, 2004: support simultaneous ipv4+ipv6
	Oct  6, 2004: move kref release function to kref_put from kref_init

Please apply.

Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>

diff -Nrup linux-2.6.9-rc3-bk6/security/bsdjail.c linux-2.6.9-rc3-bk6-jail/security/bsdjail.c
--- linux-2.6.9-rc3-bk6/security/bsdjail.c	1969-12-31 18:00:00.000000000 -0600
+++ linux-2.6.9-rc3-bk6-jail/security/bsdjail.c	2004-10-06 12:20:54.000000000 -0500
@@ -0,0 +1,1525 @@
+/*
+ * File: linux/security/bsdjail.c
+ * Author: Serge Hallyn (serue@us.ibm.com)
+ * Date: Sep 12, 2004
+ *
+ * (See Documentation/bsdjail.txt for more information)
+ *
+ * Copyright (C) 2004 International Business Machines <serue@us.ibm.com>
+ *
+ *   This program is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation; either version 2 of the License, or
+ *   (at your option) any later version.
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/security.h>
+#include <linux/namei.h>
+#include <linux/namespace.h>
+#include <linux/proc_fs.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/pagemap.h>
+#include <linux/ip.h>
+#include <net/ipv6.h>
+#include <linux/mount.h>
+#include <asm/uaccess.h>
+#include <linux/netdevice.h>
+#include <linux/inetdevice.h>
+#include <linux/seq_file.h>
+#include <linux/un.h>
+#include <linux/smp_lock.h>
+#include <linux/kref.h>
+
+static int jail_debug = 0;
+MODULE_PARM(jail_debug, "i");
+MODULE_PARM_DESC(jail_debug, "Print bsd jail debugging messages.\n");
+
+#define DBG 0
+#define WARN 1
+#define bsdj_debug(how, fmt, arg... ) \
+	do { \
+		if ( how || jail_debug ) \
+			printk(KERN_NOTICE "%s: %s: " fmt, \
+				MY_NAME, __FUNCTION__, \
+				## arg ); \
+	} while ( 0 )
+
+#define MY_NAME "bsdjail"
+
+/* flag to keep track of how we were registered */
+static int secondary = 0;
+
+/*
+ * The task structure holding jail information.
+ * Taskp->security points to one of these (or is null).
+ * There is exactly one jail_struct for each jail.  If >1 process
+ * are in the same jail, they share the same jail_struct.
+ */
+struct jail_struct {
+	struct kref		kref;
+
+	/* these are set on writes to /proc/<pid>/attr/exec */
+	char *root_pathname; /* char * containing path to use as jail / */
+	char *ip4_addr_name;  /* char * containing ip4 addr to use for jail */
+	char *ip6_addr_name;  /* char * containing ip6 addr to use for jail */
+
+	/* these are set when a jail becomes active */
+	__u32 addr4;      	/* internal form of ip4_addr_name */
+	struct in6_addr addr6;	/* internal form of ip6_addr_name */
+
+	struct dentry *dentry;  /* dentry of fs root */
+	struct vfsmount *mnt;   /* vfsmnt of fs root */
+
+	/* Resource limits.  0 = no limit */
+	int max_nrtask;		/* maximum number of tasks within this jail. */
+	int cur_nrtask;	/* current number of tasks within this jail. */
+	long maxtimeslice;      /* max timeslice in ms for procs in this jail */
+	long nice;      	/* nice level for processes in this jail */
+	long max_data, max_memlock;  /* equivalent to RLIMIT_{DATA,MEMLOCK} */
+/* values for the jail_flags field */
+#define IN_USE 1	 /* if 0, task is setting up jail, not yet in it */
+#define GOT_IPV4 2
+#define GOT_IPV6 4	 /* if 0, ipv4, else ipv6 */
+	char jail_flags;
+};
+
+#define in_use(x) (x->jail_flags & IN_USE)
+#define set_in_use(x) (x->jail_flags |= IN_USE)
+
+#define got_network(x) (x->jail_flags & (GOT_IPV4 | GOT_IPV6))
+#define got_ipv4(x) (x->jail_flags & (GOT_IPV4))
+#define got_ipv6(x) (x->jail_flags & (GOT_IPV6))
+#define set_ipv4(x) (x->jail_flags |= GOT_IPV4)
+#define set_ipv6(x) (x->jail_flags |= GOT_IPV6)
+#define unset_got_ipv4(x) (x->jail_flags &= ~GOT_IPV4)
+#define unset_got_ipv6(x) (x->jail_flags &= ~GOT_IPV6)
+
+/*
+ * structs, defines, and functions to cope with stacking
+ */
+
+#define get_task_security(task) (task->security)
+#define get_inode_security(inode) (inode->i_security)
+#define get_sock_security(sock) (sock->sk_security)
+#define get_file_security(file) (file->f_security)
+#define get_ipc_security(ipc)	(ipc->security)
+
+#define jail_of(proc) (get_task_security(proc))
+
+/*
+ * disable_jail:  A jail which was in use, but has no references
+ * left, is disabled - we free up the mountpoint and dentry, and
+ * give up our reference on the module.
+ *
+ *   don't need to put namespace, it will be done automatically
+ *     when the last process in jail is put.
+ *   DO need to put the dentry and vfsmount
+ */
+static void
+disable_jail(struct jail_struct *tsec)
+{
+	dput(tsec->dentry);
+	mntput(tsec->mnt);
+	module_put(THIS_MODULE);
+}
+
+
+static void free_jail(struct jail_struct *tsec)
+{
+	if (!tsec)
+		return;
+
+	if (tsec->root_pathname)
+		kfree(tsec->root_pathname);
+	if (tsec->ip4_addr_name)
+		kfree(tsec->ip4_addr_name);
+	if (tsec->ip6_addr_name)
+		kfree(tsec->ip6_addr_name);
+	kfree(tsec);
+}
+
+/* release_jail:
+ * Callback for kref_put to use for releasing a jail when its
+ * last user exits.
+ */
+static void release_jail(struct kref *kref)
+{
+	struct jail_struct *tsec;
+
+	tsec = container_of(kref,struct jail_struct,kref);
+	disable_jail(tsec);
+	free_jail(tsec);
+}
+
+#define set_task_security(task,data) task->security = data
+#define set_inode_security(inode,data) inode->i_security = data
+#define set_sock_security(sock,data) sock->sk_security = data
+#define set_file_security(file,data) file->f_security = data
+#define set_ipc_security(ipc,data)   ipc.security = data
+
+/*
+ * jail_task_free_security: this is the callback hooked into LSM.
+ * If there was no task->security field for bsdjail, do nothing.
+ * If there was, but it was never put into use, free the jail.
+ * If there was, and the jail is in use, then decrement the usage
+ *  count, and disable and free the jail if the usage count hits 0.
+ */
+static void jail_task_free_security(struct task_struct *task)
+{
+	struct jail_struct *tsec;
+
+	tsec = get_task_security(task);
+
+	if (!tsec)
+		return;
+
+	if (!in_use(tsec)) {
+		/*
+		 * someone did 'echo -n x > /proc/<pid>/attr/exec' but
+		 * then forked before execing.  Nuke the old info.
+		 */
+		free_jail(tsec);
+		set_task_security(task,NULL);
+		return;
+	}
+	tsec->cur_nrtask--;
+	/* If this was the last process in the jail, delete the jail */
+	kref_put(&tsec->kref, release_jail);
+}
+
+static struct jail_struct *
+alloc_task_security(struct task_struct *tsk)
+{
+	struct jail_struct *tsec;
+	tsec = kmalloc(sizeof(struct jail_struct), GFP_KERNEL);
+	if (!tsec)
+		return ERR_PTR(-ENOMEM);
+	memset(tsec, 0, sizeof(struct jail_struct));
+	set_task_security(tsk, tsec);
+	return tsec;
+}
+
+static inline int
+in_jail(struct task_struct *t)
+{
+	struct jail_struct *tsec = jail_of(t);
+
+	if (tsec && in_use(tsec))
+		return 1;
+
+	return 0;
+}
+
+/*
+ * If a network address was passed into /proc/<pid>/attr/exec,
+ * then process in its jail will only be allowed to bind/listen
+ * to that address.
+ */
+static void
+setup_netaddress(struct jail_struct *tsec)
+{
+	unsigned int a,b,c,d, i;
+	unsigned int x[8];
+
+	unset_got_ipv4(tsec);
+	tsec->addr4 = 0;
+	unset_got_ipv6(tsec);
+	ipv6_addr_set(&tsec->addr6, 0, 0, 0, 0);
+
+	if (tsec->ip4_addr_name) {
+		if (sscanf(tsec->ip4_addr_name,"%u.%u.%u.%u",&a,&b,&c,&d)!=4)
+			return;
+		if (a>255 || b>255 || c>255 || d>255)
+			return;
+		tsec->addr4 = htonl((a<<24)|(b<<16)|(c<<8)|d);
+		set_ipv4(tsec);
+		bsdj_debug(DBG, "Network (ipv4) set up (%s)\n",
+			tsec->ip4_addr_name);
+	}
+
+	if (tsec->ip6_addr_name) {
+		if (sscanf(tsec->ip6_addr_name,"%x:%x:%x:%x:%x:%x:%x:%x",
+			&x[0], &x[1], &x[2], &x[3], &x[4], &x[5], &x[6],
+			&x[7]) != 8) {
+			printk(KERN_INFO "%s: bad ipv6 addr %s\n", __FUNCTION__,
+				tsec->ip6_addr_name);
+			return;
+		}
+		for (i=0; i<8; i++) {
+			if (x[i] > 65535) {
+				printk("%s: %x > 65535 at %d\n", __FUNCTION__, x[i], i);
+				return;
+			}
+			tsec->addr6.in6_u.u6_addr16[i] = htons(x[i]);
+		}
+		set_ipv6(tsec);
+		bsdj_debug(DBG, "Network (ipv6) set up (%s)\n",
+			tsec->ip6_addr_name);
+	}
+}
+
+/*
+ * enable_jail:
+ * Called when a process is placed into a new jail to handle the
+ * actual creation of the jail.
+ *   Creates namespace
+ *   Sets process root+pwd
+ *   Stores the requested ip address
+ *   Registers a unique pseudo-proc filesystem for this jail
+ */
+static int enable_jail(struct task_struct *tsk)
+{
+	struct nameidata nd;
+	struct jail_struct *tsec;
+	int retval = -EFAULT;
+
+	tsec = jail_of(tsk);
+	if (!tsec || !tsec->root_pathname)
+		goto out;
+
+	/*
+	 * USE_JAIL_NAMESPACE: could be useful, so that future mounts outside
+	 * the jail don't affect the jail.  But it's not necessary, and
+	 * requires exporting copy_namespace from fs/namespace.c
+	 *
+	 * Actually, it would also be useful for truly hiding
+	 * information about mounts which do not exist in this jail.
+#define USE_JAIL_NAMESPACE
+	 */
+#ifdef USE_JAIL_NAMESPACE
+	bsdj_debug(DBG, "bsdjail: copying namespace.\n");
+	retval = -EPERM;
+	if (copy_namespace(CLONE_NEWNS, tsk))
+		goto out;
+	bsdj_debug(DBG, "bsdjail: copied namespace.\n");
+#endif
+
+	/* find our new root directory */
+	bsdj_debug(DBG, "bsdjail: looking up %s\n", tsec->root_pathname);
+	retval = path_lookup(tsec->root_pathname, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &nd);
+	if (retval)
+		goto out;
+
+	bsdj_debug(DBG, "bsdjail: got %s, setting root to it\n", tsec->root_pathname);
+
+	/* and set the fsroot to it */
+	set_fs_root(tsk->fs, nd.mnt, nd.dentry);
+	set_fs_pwd(tsk->fs, nd.mnt, nd.dentry);
+
+	bsdj_debug(DBG, "bsdjail: root has been set.  Have fun.\n");
+
+	/* set up networking */
+	if (tsec->ip4_addr_name || tsec->ip6_addr_name)
+		setup_netaddress(tsec);
+
+	tsec->cur_nrtask = 1;
+	if (tsec->nice)
+		set_user_nice(current, tsec->nice);
+	if (tsec->max_data) {
+		current->rlim[RLIMIT_DATA].rlim_cur = tsec->max_data;
+		current->rlim[RLIMIT_DATA].rlim_max = tsec->max_data;
+	}
+	if (tsec->max_memlock) {
+		current->rlim[RLIMIT_MEMLOCK].rlim_cur = tsec->max_memlock;
+		current->rlim[RLIMIT_MEMLOCK].rlim_max = tsec->max_memlock;
+	}
+	if (tsec->maxtimeslice) {
+		current->rlim[RLIMIT_CPU].rlim_cur = tsec->maxtimeslice;
+		current->rlim[RLIMIT_CPU].rlim_max = tsec->maxtimeslice;
+	}
+	/* success and end */
+	tsec->mnt = mntget(nd.mnt);
+	tsec->dentry = dget(nd.dentry);
+	path_release(&nd);
+	kref_init(&tsec->kref);
+	set_in_use(tsec);
+
+	/* won't let ourselves be removed until this jail goes away */
+	try_module_get(THIS_MODULE);
+
+	return 0;
+
+out:
+	return retval;
+}
+
+/*
+ * LSM /proc/<pid>/attr hooks.
+ * You may write into /proc/<pid>/attr/exec:
+ *    root /some/path
+ *    ip 2.2.2.2
+ * These values will be used on the next exec() to set up your jail
+ *  (assuming you're not already in a jail)
+ */
+static int
+jail_setprocattr(struct task_struct *p, char *name, void *value, size_t size)
+{
+	struct jail_struct *tsec = jail_of(current);
+	long val;
+	int start, len;
+
+	if (tsec && in_use(tsec))
+		return -EINVAL;  /* let them guess why */
+
+	if (p != current || strcmp(name, "exec"))
+		return -EPERM;
+
+	if (strncmp(value, "root ", 5)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		if (tsec->root_pathname)
+			kfree(tsec->root_pathname);
+		start = 5;
+		len = size-start;
+		tsec->root_pathname = kmalloc(len+1, GFP_KERNEL);
+		if (!tsec->root_pathname)
+			return -ENOMEM;
+		strlcpy(tsec->root_pathname, value+start, len+1);
+	} else if (strncmp(value, "ip ", 3)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		if (tsec->ip4_addr_name)
+			kfree(tsec->ip4_addr_name);
+		start = 3;
+		len = size-start;
+		tsec->ip4_addr_name = kmalloc(len+1, GFP_KERNEL);
+		if (!tsec->ip4_addr_name)
+			return -ENOMEM;
+		strlcpy(tsec->ip4_addr_name, value+start, len+1);
+	} else if (strncmp(value, "ip6 ", 4) == 0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		if (tsec->ip6_addr_name)
+			kfree(tsec->ip6_addr_name);
+		start = 4;
+		len = size-start;
+		tsec->ip6_addr_name = kmalloc(len+1, GFP_KERNEL);
+		if (!tsec->ip6_addr_name)
+			return -ENOMEM;
+		strlcpy(tsec->ip6_addr_name, value+start, len+1);
+
+	/* the next two are equivalent */
+	} else if (strncmp(value, "slice ", 6)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		val = simple_strtoul(value+6, NULL, 0);
+		tsec->maxtimeslice = val;
+	} else if (strncmp(value, "timeslice ", 10)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		val = simple_strtoul(value+10, NULL, 0);
+		tsec->maxtimeslice = val;
+	} else if (strncmp(value, "nrtask ", 7)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		val = (int) simple_strtol(value+7, NULL, 0);
+		if (val < 1)
+			return -EINVAL;
+		tsec->max_nrtask = val;
+	} else if (strncmp(value, "memlock ", 8)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		val = simple_strtoul(value+8, NULL, 0);
+		tsec->max_memlock = val;
+	} else if (strncmp(value, "data ", 5)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		val = simple_strtoul(value+5, NULL, 0);
+		tsec->max_data = val;
+	} else if (strncmp(value, "nice ", 5)==0) {
+		if (!tsec)
+			tsec = alloc_task_security(current);
+		if (IS_ERR(tsec))
+			return -ENOMEM;
+
+		val = simple_strtoul(value+5, NULL, 0);
+		tsec->nice = val;
+	} else
+		return -EINVAL;
+
+	return size;
+}
+
+static int print_jail_net_info(struct jail_struct *j, char *buf, int maxcnt)
+{
+	int len = 0;
+
+	if (j->ip4_addr_name)
+		len += snprintf(buf, maxcnt, "%s\n", j->ip4_addr_name);
+	if (j->ip6_addr_name)
+		len += snprintf(buf, maxcnt-len, "%s\n", j->ip6_addr_name);
+
+	return snprintf(buf, maxcnt, "No network information\n");
+}
+
+/*
+ * LSM /proc/<pid>/attr read hook.
+ *
+ * /proc/$$/attr/current output:
+ * If the reading process, say process 1001, is in a jail, then
+ *   cat /proc/999/attr/current
+ * will print networking information.
+ * If the reading process, say process 1001, is not in a jail, then
+ *   cat /proc/999/attr/current
+ * will return
+ *   root: (root of jail)
+ *   ip:   (ip address of jail)
+ * if 999 is in a jail, or
+ *   -EINVAL
+ * if 999 is not in a jail.
+ *
+ * /proc/$$/attr/exec output:
+ * A process in a jail gets -EINVAL for /proc/$$/attr/exec.
+ * A process not in a jail gets hints on starting a jail.
+ */
+static int
+jail_getprocattr(struct task_struct *p, char *name, void *value, size_t size)
+{
+	struct jail_struct *tsec;
+	int err = 0;
+
+	if (in_jail(current)) {
+		if (strcmp(name, "current")==0) {
+			/* provide network info */
+			err = print_jail_net_info(jail_of(current), value,
+				size);
+			return err;
+		}
+		return -EINVAL;  /* let them guess why */
+	}
+
+	if (strcmp(name, "exec") == 0) {
+		/* Print usage some help */
+		err = snprintf(value, size,
+			"Valid keywords:\n"
+			"root    <pathname>\n"
+			"ip      <ip4-addr>\n"
+			"ip6     <ip6-addr>\n"
+			"nrtask  <max number of tasks in this jail>\n"
+			"nice    <nice level for processes in this jail>\n"
+			"slice   <max timeslice per process in msecs>\n"
+			"data    <max data size per process in bytes>\n"
+			"memlock <max lockable memory per process in bytes>\n");
+		return err;
+	}
+
+	if (strcmp(name, "current"))
+		return -EPERM;
+
+	tsec = jail_of(p);
+	if (!tsec || !in_use(tsec)) {
+		err = snprintf(value, size, "Not Jailed\n");
+	} else {
+		err = snprintf(value, size,
+			"Root: %s\nIPv4: %s\nIPv6: %s\n"
+			"max_nrtask %d current nrtask %d max_timeslice %lu "
+			"nice %lu\n"
+			"max_memlock %lu max_data %lu\n",
+			tsec->root_pathname,
+			tsec->ip4_addr_name ? tsec->ip4_addr_name : "(none)",
+			tsec->ip6_addr_name ? tsec->ip6_addr_name : "(none)",
+			tsec->max_nrtask, tsec->cur_nrtask, tsec->maxtimeslice,
+			tsec->nice, tsec->max_data, tsec->max_memlock);
+	}
+
+	return err;
+}
+
+/*
+ * Forbid a process in a jail from sending a signal to a process in another
+ * (or no) jail through file sigio.
+ *
+ * We consider the process which set the fowner to be the one sending the
+ * signal, rather than the one writing to the file.  Therefore we store the
+ * jail of a process during jail_file_set_fowner, then check that against
+ * the jail of the process receiving the signal.
+ */
+static int
+jail_file_send_sigiotask(struct task_struct *tsk, struct fown_struct *fown,
+       int fd, int reason)
+{
+	struct file *file;
+	struct jail_struct *tsec, *fsec;
+
+	if (!in_jail(current))
+		return 0;
+
+        file = (struct file *)((long)fown - offsetof(struct file,f_owner));
+	tsec = jail_of(tsk);
+	fsec = get_file_security(file);
+
+	if (fsec != tsec)
+		return -EPERM;
+
+	return 0;
+}
+
+static int
+jail_file_set_fowner(struct file *file)
+{
+	struct jail_struct *tsec;
+
+	tsec = jail_of(current);
+	set_file_security(file, tsec);
+	if (tsec)
+		kref_get(&tsec->kref);
+
+	return 0;
+}
+
+static void free_ipc_security(struct kern_ipc_perm *ipc)
+{
+	struct jail_struct *tsec;
+
+	tsec = get_ipc_security(ipc);
+	if (!tsec)
+		return;
+	kref_put(&tsec->kref, release_jail);
+	set_ipc_security((*ipc), NULL);
+}
+
+static void free_file_security(struct file *file)
+{
+	struct jail_struct *tsec;
+
+	tsec = get_file_security(file);
+	if (!tsec)
+		return;
+	kref_put(&tsec->kref, release_jail);
+	set_file_security(file, NULL);
+}
+
+static void free_inode_security(struct inode *inode)
+{
+	struct jail_struct *tsec;
+
+	tsec = get_inode_security(inode);
+	if (!tsec)
+		return;
+	kref_put(&tsec->kref, release_jail);
+	set_inode_security(inode, NULL);
+}
+
+/*
+ * LSM ptrace hook:
+ * process in jail may not ptrace process not in the same jail
+ */
+static int
+jail_ptrace (struct task_struct *tracer, struct task_struct *tracee)
+{
+	struct jail_struct *tsec = jail_of(tracer);
+
+	if (tsec && in_use(tsec)) {
+		if (tsec == jail_of(tracee))
+			return 0;
+		return -EPERM;
+	}
+	return 0;
+}
+
+/*
+ * process in jail may only use one (aliased) ip address.  If they try to
+ * attach to 127.0.0.1, that is remapped to their own address.  If some
+ * other address (and not their own), deny permission
+ */
+static int jail_socket_unix_bind(struct socket *sock, struct sockaddr *address,
+		int addrlen);
+
+#define loopbackaddr htonl((127 << 24) | 1)
+
+static inline int jail_inet4_bind(struct socket *sock, struct sockaddr *address,
+	int addrlen, struct jail_struct *tsec)
+{
+	struct sockaddr_in *inaddr;
+	__u32 sin_addr, jailaddr;
+
+	if (!got_ipv4(tsec))
+		return -EPERM;
+
+	inaddr = (struct sockaddr_in *)address;
+	sin_addr = inaddr->sin_addr.s_addr;
+	jailaddr = tsec->addr4;
+
+	if (sin_addr == jailaddr)
+		return 0;
+
+	if (sin_addr == loopbackaddr || !sin_addr) {
+		bsdj_debug(DBG, "Got a loopback or 0 address\n");
+		sin_addr = jailaddr;
+		bsdj_debug(DBG, "Converted to: %u.%u.%u.%u\n",
+			NIPQUAD(sin_addr));
+		return 0;
+	}
+
+	return -EPERM;
+}
+
+static inline int
+jail_inet6_bind(struct socket *sock, struct sockaddr *address, int addrlen,
+	struct jail_struct *tsec)
+{
+	struct sockaddr_in6 *inaddr6;
+	struct in6_addr *sin6_addr, *jailaddr;
+
+	if (!got_ipv6(tsec))
+		return -EPERM;
+
+	inaddr6 = (struct sockaddr_in6 *)address;
+	sin6_addr = &inaddr6->sin6_addr;
+	jailaddr = &tsec->addr6;
+
+	if (ipv6_addr_cmp(jailaddr, sin6_addr)==0)
+		return 0;
+
+	if (ipv6_addr_cmp(sin6_addr, &in6addr_loopback)==0) {
+		ipv6_addr_copy(sin6_addr, jailaddr);
+		return 0;
+	}
+
+	printk(KERN_NOTICE "%s: DENYING\n", __FUNCTION__);
+	printk(KERN_NOTICE "%s: a %04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x "
+		"j %04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n",
+		__FUNCTION__,
+		NIP6(*sin6_addr),
+		NIP6(*jailaddr));
+
+	return -EPERM;
+}
+
+static int
+jail_socket_bind(struct socket *sock, struct sockaddr *address, int addrlen)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+
+	if (sock->sk->sk_family == AF_UNIX)
+		return jail_socket_unix_bind(sock, address, addrlen);
+
+	if (!got_network(tsec))
+		/* If we want to be strict, we could just
+		 * deny net access when lacking a pseudo ip.
+		 * For now we just allow it. */
+		return 0;
+
+	switch(address->sa_family) {
+		case AF_INET:
+			return jail_inet4_bind(sock, address, addrlen, tsec);
+
+		case AF_INET6:
+			return jail_inet6_bind(sock, address, addrlen, tsec);
+
+		default:
+			return 0;
+	}
+}
+
+/*
+ * If locked in an ipv6 jail, don't let them use ipv4, and vice versa
+ */
+static int
+jail_socket_create(int family, int type, int protocol, int kern)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec) || kern || !got_network(tsec))
+		return 0;
+
+	switch(family) {
+		case AF_INET:
+			if (got_ipv4(tsec))
+				return 0;
+			return -EPERM;
+		case AF_INET6:
+			if (got_ipv6(tsec))
+				return 0;
+			return -EPERM;
+		default:
+			return 0;
+	};
+
+	return 0;
+}
+
+static void
+jail_socket_post_create(struct socket *sock, int family, int type,
+	int protocol, int kern)
+{
+	struct inet_opt *inet;
+	struct ipv6_pinfo *inet6;
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec) || kern || !got_network(tsec))
+		return;
+
+	switch(family) {
+		case AF_INET:
+			inet = inet_sk(sock->sk);
+			inet->saddr = tsec->addr4;
+			break;
+		case AF_INET6:
+			inet6 = inet6_sk(sock->sk);
+			ipv6_addr_copy(&inet6->saddr, &tsec->addr6);
+			break;
+		default:
+			break;
+	};
+
+	return;
+}
+
+static int
+jail_socket_listen(struct socket *sock, int backlog)
+{
+	struct inet_opt *inet;
+	struct ipv6_pinfo *inet6;
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec) || !got_network(tsec))
+		return 0;
+
+	switch (sock->sk->sk_family) {
+		case AF_INET:
+			inet = inet_sk(sock->sk);
+			if (inet->saddr == tsec->addr4)
+				return 0;
+			return -EPERM;
+
+		case AF_INET6:
+			inet6 = inet6_sk(sock->sk);
+			if (ipv6_addr_cmp(&inet6->saddr, &tsec->addr6)==0)
+				return 0;
+			return -EPERM;
+
+		default:
+			return 0;
+
+	}
+}
+
+static void free_sock_security(struct sock *sk)
+{
+	struct jail_struct *tsec;
+
+	tsec = get_sock_security(sk);
+	if (!tsec)
+		return;
+	kref_put(&tsec->kref, release_jail);
+	set_sock_security(sk, NULL);
+}
+
+/*
+ * The next three (socket) hooks prevent a process in a jail from sending
+ * data to a abstract unix domain socket which was bound outside the jail.
+ */
+static int
+jail_socket_unix_bind(struct socket *sock, struct sockaddr *address,
+	int addrlen)
+{
+	struct sockaddr_un *sunaddr;
+	struct jail_struct *tsec;
+
+	if (sock->sk->sk_family != AF_UNIX)
+		return 0;
+
+	sunaddr = (struct sockaddr_un *)address;
+	if (sunaddr->sun_path[0] != 0)
+		return 0;
+
+	tsec = jail_of(current);
+	set_sock_security(sock->sk, tsec);
+	if (tsec)
+		kref_get(&tsec->kref);
+	return 0;
+}
+
+/*
+ * Note - we deny sends  both from unjailed to jailed, and from jailed
+ * to unjailed.  As well as, of course between different jails.
+ */
+static int
+jail_socket_unix_may_send(struct socket *sock, struct socket *other)
+{
+	struct jail_struct *tsec, *ssec;
+
+	tsec = jail_of(current);  /* jail of sending process */
+	ssec = get_sock_security(other->sk);  /* jail of receiver */
+
+	if (tsec != ssec)
+		return -EPERM;
+
+	return 0;
+}
+
+static int
+jail_socket_unix_stream_connect(struct socket *sock,
+	      struct socket *other, struct sock *newsk)
+{
+	struct jail_struct *tsec, *ssec;
+
+	tsec = jail_of(current);  /* jail of sending process */
+	ssec = get_sock_security(other->sk);  /* jail of receiver */
+
+	if (tsec != ssec)
+		return -EPERM;
+
+	return 0;
+}
+
+static int
+jail_mount(char * dev_name, struct nameidata *nd, char * type,
+                         unsigned long flags, void * data)
+{
+	if (in_jail(current))
+		return -EPERM;
+
+	return 0;
+}
+
+static int
+jail_umount(struct vfsmount *mnt, int flags)
+{
+	if (in_jail(current))
+		return -EPERM;
+
+	return 0;
+}
+
+/*
+ * process in jail may not:
+ *   use nice
+ *   change network config
+ *   load/unload modules
+ */
+static int
+jail_capable (struct task_struct *tsk, int cap)
+{
+	if (in_jail(tsk)) {
+		if (cap == CAP_SYS_NICE)
+			return -EPERM;
+		if (cap == CAP_NET_ADMIN)
+			return -EPERM;
+		if (cap == CAP_SYS_MODULE)
+			return -EPERM;
+		if (cap == CAP_SYS_RAWIO)
+			return -EPERM;
+	}
+
+	if (cap_is_fs_cap (cap) ? tsk->fsuid == 0 : tsk->euid == 0)
+		return 0;
+	return -EPERM;
+}
+
+/*
+ * jail_security_task_create:
+ *
+ * If the current process is ina a jail, and that jail is about to exceed a
+ * maximum number of processes, then refuse to fork.  If the maximum number
+ * of jails is listed as 0, then there is no limit for this jail, and we allow
+ * all forks.
+ */
+static inline int
+jail_security_task_create (unsigned long clone_flags)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+
+	if (tsec->max_nrtask && tsec->cur_nrtask >= tsec->max_nrtask)
+		return -EPERM;
+	return 0;
+}
+
+/*
+ * The child of a process in a jail belongs in the same jail
+ */
+static int
+jail_task_alloc_security(struct task_struct *tsk)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+
+	set_task_security(tsk, tsec);
+	kref_get(&tsec->kref);
+	tsec->cur_nrtask++;
+	if (tsec->maxtimeslice) {
+		tsk->rlim[RLIMIT_CPU].rlim_max = tsec->maxtimeslice;
+		tsk->rlim[RLIMIT_CPU].rlim_cur = tsec->maxtimeslice;
+	}
+	if (tsec->max_data) {
+		tsk->rlim[RLIMIT_CPU].rlim_max = tsec->max_data;
+		tsk->rlim[RLIMIT_CPU].rlim_cur = tsec->max_data;
+	}
+	if (tsec->max_memlock) {
+		tsk->rlim[RLIMIT_CPU].rlim_max = tsec->max_memlock;
+		tsk->rlim[RLIMIT_CPU].rlim_cur = tsec->max_memlock;
+	}
+	if (tsec->nice)
+		set_user_nice(current, tsec->nice);
+
+	return 0;
+}
+
+static int
+jail_bprm_alloc_security(struct linux_binprm *bprm)
+{
+	struct jail_struct *tsec;
+	int ret;
+
+	tsec = jail_of(current);
+	if (!tsec)
+		return 0;
+
+	if (in_use(tsec))
+		return 0;
+
+	if (tsec->root_pathname) {
+		ret = enable_jail(current);
+		if (ret) {
+			/* if we failed, nix out the root/ip requests */
+			jail_task_free_security(current);
+			return ret;
+		}
+	}
+	return 0;
+}
+
+/*
+ * Process in jail may not create devices
+ * Thanks to Brad Spender for pointing out fifos should be allowed.
+ */
+/* TODO: We may want to allow /dev/log, at least... */
+static int
+jail_inode_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
+{
+	if (!in_jail(current))
+		return 0;
+
+	if (S_ISFIFO(mode))
+		return 0;
+
+	return -EPERM;
+}
+
+/* yanked from fs/proc/base.c */
+static unsigned name_to_int(struct dentry *dentry)
+{
+	const char *name = dentry->d_name.name;
+	int len = dentry->d_name.len;
+	unsigned n = 0;
+
+	if (len > 1 && *name == '0')
+		goto out;
+	while (len-- > 0) {
+		unsigned c = *name++ - '0';
+		if (c > 9)
+			goto out;
+		if (n >= (~0U-9)/10)
+			goto out;
+		n *= 10;
+		n += c;
+	}
+	return n;
+out:
+	return ~0U;
+}
+
+/*
+ * jail_proc_inode_permission:
+ *   called only when current is in a jail, and is trying to reach
+ *   /proc/<pid>.  We check whether <pid> is in the same jail as
+ *   current.  If not, permission is denied.
+ *
+ * NOTE:  On the one hand, the task_to_inode(inode)->i_security
+ * approach seems cleaner, but on the other, this prevents us
+ * from unloading bsdjail for awhile...
+ */
+static int
+jail_proc_inode_permission(struct inode *inode, int mask,
+				    struct nameidata *nd)
+{
+	struct jail_struct *tsec = jail_of(current);
+	struct dentry *dentry = nd->dentry;
+	unsigned pid;
+
+	pid = name_to_int(dentry);
+	if (pid == ~0U) {
+		struct qstr *dname = &dentry->d_name;
+		if (strcmp(dname->name, "scsi")==0 ||
+			strcmp(dname->name, "sys")==0 ||
+			strcmp(dname->name, "ide")==0)
+			return -EPERM;
+		return 0;
+	}
+
+	if (dentry->d_parent != dentry->d_sb->s_root)
+		return 0;
+	if (get_inode_security(inode) != tsec)
+		return -ENOENT;
+
+	return 0;
+}
+
+/*
+ * Here is our attempt to prevent chroot escapes.
+ */
+static int
+is_jailroot_parent(struct dentry *candidate, struct dentry *root,
+	struct vfsmount *rootmnt)
+{
+	if (candidate == root)
+		return 0;
+
+	/* simple case:  fs->root/.. == candidate */
+	if (root->d_parent == candidate)
+		return 1;
+
+	/*
+	 * now more complicated:  if fs->root is a mounted directory,
+	 * then chdir(..) out of fs->root, at follow_dotdot, will follow
+	 * the fs->root mount point. So we must check the parent dir of
+	 * the fs->root mount point.
+	 */
+	if (rootmnt->mnt_root == root && rootmnt->mnt_mountpoint!=root) {
+		root = rootmnt->mnt_mountpoint;
+		rootmnt = rootmnt->mnt_parent;
+		return is_jailroot_parent(candidate, root, rootmnt);
+	}
+
+	return 0;
+}
+
+/*
+ * A process in a jail may not see that /proc/<pid> exists for
+ * process not in its jail
+ * Unfortunately we can't pretend that pid for the starting process
+ * is 1, as vserver does.
+ */
+static int jail_task_lookup(struct task_struct *p)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec)
+		return 0;
+	if (tsec == jail_of(p))
+		return 0;
+	return -EPERM;
+}
+/*
+ * security_task_to_inode:
+ * Set inode->security = task's jail.
+ */
+static void jail_task_to_inode(struct task_struct *p, struct inode *inode)
+{
+	struct jail_struct *tsec = jail_of(p);
+
+	if (!tsec || !in_use(tsec))
+		return;
+	if (get_inode_security(inode))
+		return;
+	kref_get(&tsec->kref);
+	set_inode_security(inode, tsec);
+}
+
+/*
+ * inode_permission:
+ * If we are trying to look into certain /proc files from in a jail, we
+ * 	may deny permission.
+ * If we are trying to cd(..), but the cwd is the root of our jail, then
+ * permission is denied.
+ */
+static int
+jail_inode_permission(struct inode *inode, int mask,
+				    struct nameidata *nd)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+
+	if (!nd)
+		return 0;
+
+	if (nd->dentry &&
+		strcmp(nd->dentry->d_sb->s_type->name, "proc")==0) {
+		return jail_proc_inode_permission(inode, mask, nd);
+
+	}
+
+	if (!(mask&MAY_EXEC))
+		return 0;
+	if (!inode || !S_ISDIR(inode->i_mode))
+		return 0;
+
+	if (is_jailroot_parent(nd->dentry, tsec->dentry, tsec->mnt)) {
+		bsdj_debug(WARN,"Attempt to chdir(..) out of jail!\n"
+				"(%s is a subdir of %s)\n",
+				tsec->dentry->d_name.name,
+				nd->dentry->d_name.name);
+		return -EPERM;
+	}
+
+	return 0;
+}
+
+/*
+ * A function which returns -ENOENT if dentry is the dentry for
+ * a /proc/<pid> directory.  It returns 0 otherwise.
+ */
+static inline int
+generic_procpid_check(struct dentry *dentry)
+{
+	struct jail_struct *jail = jail_of(current);
+	unsigned pid = name_to_int(dentry);
+
+	if (!jail || !in_use(jail))
+		return 0;
+	if (pid == ~0U)
+		return 0;
+	if (strcmp(dentry->d_sb->s_type->name, "proc")!=0)
+		return 0;
+	if (dentry->d_parent != dentry->d_sb->s_root)
+		return 0;
+	if (get_inode_security(dentry->d_inode) != jail)
+		return -ENOENT;
+	return 0;
+}
+
+/*
+ * We want getattr to fail on /proc/<pid> to prevent leakage through, for
+ * instance, ls -d.
+ */
+static int
+jail_inode_getattr(struct vfsmount *mnt, struct dentry *dentry)
+{
+	return generic_procpid_check(dentry);
+}
+
+/* This probably is not necessary - /proc does not support xattrs? */
+static int
+jail_inode_getxattr(struct dentry *dentry, char *name)
+{
+	return generic_procpid_check(dentry);
+}
+
+/* process in jail may not send signal to process not in the same jail */
+static int
+jail_task_kill(struct task_struct *p, struct siginfo *info, int sig)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+
+	if (tsec == jail_of(p))
+		return 0;
+
+	if (sig==SIGCHLD)
+		return 0;
+
+	return -EPERM;
+}
+
+/*
+ * LSM hooks to limit jailed process' abilities to muck with resource
+ * limits
+ */
+static int jail_task_setrlimit (unsigned int resource, struct rlimit *new_rlim)
+{
+	if (!in_jail(current))
+		return 0;
+
+	return -EPERM;
+}
+
+static int jail_task_setscheduler (struct task_struct *p, int policy,
+				    struct sched_param *lp)
+{
+	if (!in_jail(current))
+		return 0;
+
+	return -EPERM;
+}
+
+/*
+ * LSM hooks to limit IPC access.
+ */
+
+static inline int
+basic_ipc_security_check(struct kern_ipc_perm *p, struct task_struct *target)
+{
+	struct jail_struct *tsec = jail_of(target);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+
+	if (get_ipc_security(p) != tsec)
+		return -EPERM;
+
+	return 0;
+}
+
+static int
+jail_ipc_permission(struct kern_ipc_perm *ipcp, short flag)
+{
+	return basic_ipc_security_check(ipcp, current);
+}
+
+static int
+jail_shm_alloc_security (struct shmid_kernel *shp)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+	set_ipc_security(shp->shm_perm, tsec);
+	kref_get(&tsec->kref);
+	return 0;
+}
+
+static void
+jail_shm_free_security (struct shmid_kernel *shp)
+{
+	free_ipc_security(&shp->shm_perm);
+}
+
+static int
+jail_shm_associate (struct shmid_kernel *shp, int shmflg)
+{
+	return basic_ipc_security_check(&shp->shm_perm, current);
+}
+
+static int
+jail_shm_shmctl(struct shmid_kernel *shp, int cmd)
+{
+	if (cmd == IPC_INFO || cmd == SHM_INFO)
+		return 0;
+
+	return basic_ipc_security_check(&shp->shm_perm, current);
+}
+
+static int
+jail_shm_shmat(struct shmid_kernel *shp, char *shmaddr, int shmflg)
+{
+	return basic_ipc_security_check(&shp->shm_perm, current);
+}
+
+static int
+jail_msg_queue_alloc(struct msg_queue *msq)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+	set_ipc_security(msq->q_perm, tsec);
+	kref_get(&tsec->kref);
+	return 0;
+}
+
+static void
+jail_msg_queue_free(struct msg_queue *msq)
+{
+	free_ipc_security(&msq->q_perm);
+}
+
+static int jail_msg_queue_associate(struct msg_queue *msq, int flag)
+{
+	return basic_ipc_security_check(&msq->q_perm, current);
+}
+
+static int
+jail_msg_queue_msgctl(struct msg_queue *msq, int cmd)
+{
+	if (cmd == IPC_INFO || cmd == MSG_INFO)
+		return 0;
+
+	return basic_ipc_security_check(&msq->q_perm, current);
+}
+
+static int
+jail_msg_queue_msgsnd(struct msg_queue *msq, struct msg_msg *msg, int msqflg)
+{
+	return basic_ipc_security_check(&msq->q_perm, current);
+}
+
+static int
+jail_msg_queue_msgrcv(struct msg_queue *msq, struct msg_msg *msg,
+		struct task_struct *target, long type, int mode)
+
+{
+	return basic_ipc_security_check(&msq->q_perm, target);
+}
+
+static int
+jail_sem_alloc_security(struct sem_array *sma)
+{
+	struct jail_struct *tsec = jail_of(current);
+
+	if (!tsec || !in_use(tsec))
+		return 0;
+	set_ipc_security(sma->sem_perm, tsec);
+	kref_get(&tsec->kref);
+	return 0;
+}
+
+static void
+jail_sem_free_security(struct sem_array *sma)
+{
+	free_ipc_security(&sma->sem_perm);
+}
+
+static int
+jail_sem_associate(struct sem_array *sma, int semflg)
+{
+	return basic_ipc_security_check(&sma->sem_perm, current);
+}
+
+static int
+jail_sem_semctl(struct sem_array *sma, int cmd)
+{
+	if (cmd == IPC_INFO || cmd == SEM_INFO)
+		return 0;
+	return basic_ipc_security_check(&sma->sem_perm, current);
+}
+
+static int
+jail_sem_semop(struct sem_array *sma, struct sembuf *sops, unsigned nsops,
+	int alter)
+{
+	return basic_ipc_security_check(&sma->sem_perm, current);
+}
+
+static struct security_operations bsdjail_security_ops = {
+	.ptrace  =			jail_ptrace,
+	.capable =			jail_capable,
+
+	.task_kill =			jail_task_kill,
+	.task_alloc_security =		jail_task_alloc_security,
+	.task_free_security =		jail_task_free_security,
+	.bprm_alloc_security =		jail_bprm_alloc_security,
+	.task_create =			jail_security_task_create,
+	.task_to_inode =		jail_task_to_inode,
+	.task_lookup =			jail_task_lookup,
+
+	.task_setrlimit =		jail_task_setrlimit,
+	.task_setscheduler =		jail_task_setscheduler,
+
+	.setprocattr =                  jail_setprocattr,
+	.getprocattr =                  jail_getprocattr,
+
+	.file_set_fowner =		jail_file_set_fowner,
+	.file_send_sigiotask =		jail_file_send_sigiotask,
+	.file_free_security =		free_file_security,
+
+	.socket_bind =			jail_socket_bind,
+	.socket_listen =		jail_socket_listen,
+	.socket_create =		jail_socket_create,
+	.socket_post_create =		jail_socket_post_create,
+        .unix_stream_connect =		jail_socket_unix_stream_connect,
+	.unix_may_send =		jail_socket_unix_may_send,
+	.sk_free_security =		free_sock_security,
+
+	.inode_mknod =			jail_inode_mknod,
+	.inode_permission =		jail_inode_permission,
+	.inode_free_security =		free_inode_security,
+	.inode_getattr =		jail_inode_getattr,
+	.inode_getxattr =		jail_inode_getxattr,
+	.sb_mount =			jail_mount,
+	.sb_umount =			jail_umount,
+
+	.ipc_permission =		jail_ipc_permission,
+	.shm_alloc_security = 		jail_shm_alloc_security,
+	.shm_free_security = 		jail_shm_free_security,
+	.shm_associate =		jail_shm_associate,
+	.shm_shmctl =			jail_shm_shmctl,
+	.shm_shmat =			jail_shm_shmat,
+
+	.msg_queue_alloc_security =	jail_msg_queue_alloc,
+	.msg_queue_free_security =	jail_msg_queue_free,
+	.msg_queue_associate =		jail_msg_queue_associate,
+	.msg_queue_msgctl =		jail_msg_queue_msgctl,
+	.msg_queue_msgsnd =		jail_msg_queue_msgsnd,
+	.msg_queue_msgrcv =		jail_msg_queue_msgrcv,
+
+	.sem_alloc_security = 		jail_sem_alloc_security,
+	.sem_free_security =  		jail_sem_free_security,
+	.sem_associate =		jail_sem_associate,
+	.sem_semctl =			jail_sem_semctl,
+	.sem_semop =			jail_sem_semop,
+};
+
+static int __init bsdjail_init (void)
+{
+	int rc = 0;
+
+	if (register_security (&bsdjail_security_ops)) {
+		printk (KERN_INFO
+			"Failure registering BSD Jail module with the kernel\n");
+
+		rc = mod_reg_security(MY_NAME, &bsdjail_security_ops);
+		if (rc < 0) {
+			printk (KERN_INFO "Failure registering BSD Jail "
+				" module with primary security module.\n");
+			return -EINVAL;
+		}
+		secondary = 1;
+	}
+	printk (KERN_INFO "BSD Jail module initialized.\n");
+
+	return 0;
+}
+
+static void __exit bsdjail_exit (void)
+{
+	if (secondary) {
+		if (mod_unreg_security (MY_NAME, &bsdjail_security_ops))
+			printk (KERN_INFO "Failure unregistering BSD Jail "
+				" module with primary module.\n");
+	} else {
+		if (unregister_security (&bsdjail_security_ops)) {
+			printk (KERN_INFO "Failure unregistering BSD Jail "
+				"module with the kernel\n");
+		}
+	}
+
+	printk (KERN_INFO "BSD Jail module removed\n");
+}
+
+security_initcall (bsdjail_init);
+module_exit (bsdjail_exit);
+
+MODULE_DESCRIPTION("BSD Jail LSM.");
+MODULE_LICENSE("GPL");
diff -Nrup linux-2.6.9-rc3-bk6/security/Kconfig linux-2.6.9-rc3-bk6-jail/security/Kconfig
--- linux-2.6.9-rc3-bk6/security/Kconfig	2004-10-06 10:08:02.000000000 -0500
+++ linux-2.6.9-rc3-bk6-jail/security/Kconfig	2004-10-06 10:52:13.000000000 -0500
@@ -46,5 +46,16 @@ config SECURITY_ROOTPLUG
 
 source security/selinux/Kconfig
 
+config SECURITY_BSDJAIL
+	tristate "BSD Jail LSM"
+	depends on SECURITY
+	select SECURITY_NETWORK
+	help
+	  Provides BSD Jail compartmentalization functionality.
+	  See Documentation/bsdjail.txt for more information and
+	  usage instructions.
+	  
+	  If you are unsure how to answer this question, answer N.
+
 endmenu
 
diff -Nrup linux-2.6.9-rc3-bk6/security/Makefile linux-2.6.9-rc3-bk6-jail/security/Makefile
--- linux-2.6.9-rc3-bk6/security/Makefile	2004-08-14 00:37:26.000000000 -0500
+++ linux-2.6.9-rc3-bk6-jail/security/Makefile	2004-10-06 10:52:13.000000000 -0500
@@ -15,3 +15,4 @@ obj-$(CONFIG_SECURITY)			+= security.o d
 obj-$(CONFIG_SECURITY_SELINUX)		+= selinux/built-in.o
 obj-$(CONFIG_SECURITY_CAPABILITIES)	+= commoncap.o capability.o
 obj-$(CONFIG_SECURITY_ROOTPLUG)		+= commoncap.o root_plug.o
+obj-$(CONFIG_SECURITY_BSDJAIL)		+= bsdjail.o