Skip to contentSkip to conversation

Add SELinux policy capability for always checking packet class.

PATCH In review Chris PeBenito 3 participants 9 messages Jun 7, 2012 – Jun 8, 2012 SELinux security module

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 Add SELinux policy capability for always checking packet class.

Currently the packet class in SELinux is not checked if there are no SECMARK rules in the security or mangle netfilter tables. Some systems prefer that packets are always checked, for example, to protect the system should the netfilter rules fail to load or if the nefilter rules were maliciously flushed.

Add the always_check_network policy capability which, when enabled, treats SECMARK as enabled, even if there are no netfilter SECMARK rules.

---
 security/selinux/hooks.c            |    8 ++++++--
 security/selinux/include/security.h |    2 ++
 security/selinux/selinuxfs.c        |    3 ++-
 security/selinux/ss/services.c      |    3 +++
 4 files changed, 13 insertions(+), 3 deletions(-)
security/selinux/hooks.c +6 -2
index 372ec65..ec7151b 100644
@@ -137,12 +137,16 @@ static struct kmem_cache *sel_inode_cache;
  * This function checks the SECMARK reference counter to see if any SECMARK
  * targets are currently configured, if the reference counter is greater than
  * zero SECMARK is considered to be enabled.  Returns true (1) if SECMARK is
- * enabled, false (0) if SECMARK is disabled.
+ * enabled, false (0) if SECMARK is disabled.  If the always_check_network
+ * policy capability is enabled, SECMARK is always considered enabled.
  *
  */
 static int selinux_secmark_enabled(void)
 {
-	return (atomic_read(&selinux_secmark_refcount) > 0);
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (atomic_read(&selinux_secmark_refcount) > 0);
 }
 
 /*
security/selinux/include/security.h +2 -0
index dde2005..981c4ac 100644
@@ -68,12 +68,14 @@ extern int selinux_enabled;
 enum {
 	POLICYDB_CAPABILITY_NETPEER,
 	POLICYDB_CAPABILITY_OPENPERM,
+	POLICYDB_CAPABILITY_ALWAYSNETWORK,
 	__POLICYDB_CAPABILITY_MAX
 };
 #define POLICYDB_CAPABILITY_MAX (__POLICYDB_CAPABILITY_MAX - 1)
 
 extern int selinux_policycap_netpeer;
 extern int selinux_policycap_openperm;
+extern int selinux_policycap_alwaysnetwork;
 
 /*
  * type_datum properties
security/selinux/selinuxfs.c +2 -1
index 3ad2902..cb893f9 100644
@@ -44,7 +44,8 @@
 /* Policy capability filenames */
 static char *policycap_names[] = {
 	"network_peer_controls",
-	"open_perms"
+	"open_perms",
+	"always_check_network"
 };
 
 unsigned int selinux_checkreqprot = CONFIG_SECURITY_SELINUX_CHECKREQPROT_VALUE;
security/selinux/ss/services.c +3 -0
index 4321b8f..e124d8f 100644
@@ -72,6 +72,7 @@
 
 int selinux_policycap_netpeer;
 int selinux_policycap_openperm;
+int selinux_policycap_alwaysnetwork;
 
 static DEFINE_RWLOCK(policy_rwlock);
 
@@ -1812,6 +1813,8 @@ static void security_load_policycaps(void)
 						  POLICYDB_CAPABILITY_NETPEER);
 	selinux_policycap_openperm = ebitmap_get_bit(&policydb.policycaps,
 						  POLICYDB_CAPABILITY_OPENPERM);
+	selinux_policycap_alwaysnetwork = ebitmap_get_bit(&policydb.policycaps,
+						  POLICYDB_CAPABILITY_ALWAYSNETWORK);
 }
 
 static int security_preserve_bools(struct policydb *p);
Signature · 7 lines
1.7.8.6


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#1 lore raw

PATCH 2/2 Update SELinux policy capability to always check peer class.

Update the always_check_network policy capability which, when enabled, treats peer labeling as enabled, even if there is no Netlabel or labeled IPSEC configuration.

---
 security/selinux/hooks.c |   24 +++++++++++++++++++++---
 1 files changed, 21 insertions(+), 3 deletions(-)
security/selinux/hooks.c +21 -3
index ec7151b..e8f612e 100644
@@ -149,6 +149,24 @@ static int selinux_secmark_enabled(void)
 		return (atomic_read(&selinux_secmark_refcount) > 0);
 }
 
+/**
+ * selinux_peerlbl_enabled - Check to see if peer labeling is currently enabled
+ *
+ * Description:
+ * This function checks if NetLabel or labeled IPSEC is enabled.  Returns true
+ * (1) if any are enabled or false (0) if neither are enabled.  If the
+ * always_check_network policy capability is enabled, peer labeling
+ * is always considered enabled.
+ *
+ */
+static int selinux_peerlbl_enabled(void)
+{
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (netlbl_enabled() || selinux_xfrm_enabled());
+}
+
 /*
  * initialise the security for the init task
  */
@@ -4188,7 +4206,7 @@ static int selinux_socket_sock_rcv_skb(struct sock *sk, struct sk_buff *skb)
 		return selinux_sock_rcv_skb_compat(sk, skb, family);
 
 	secmark_active = selinux_secmark_enabled();
-	peerlbl_active = netlbl_enabled() || selinux_xfrm_enabled();
+	peerlbl_active = selinux_peerlbl_enabled();
 	if (!secmark_active && !peerlbl_active)
 		return 0;
 
@@ -4540,7 +4558,7 @@ static unsigned int selinux_ip_forward(struct sk_buff *skb, int ifindex,
 
 	secmark_active = selinux_secmark_enabled();
 	netlbl_active = netlbl_enabled();
-	peerlbl_active = netlbl_active || selinux_xfrm_enabled();
+	peerlbl_active = selinux_peerlbl_enabled();
 	if (!secmark_active && !peerlbl_active)
 		return NF_ACCEPT;
 
@@ -4692,7 +4710,7 @@ static unsigned int selinux_ip_postroute(struct sk_buff *skb, int ifindex,
 		return NF_ACCEPT;
 #endif
 	secmark_active = selinux_secmark_enabled();
-	peerlbl_active = netlbl_enabled() || selinux_xfrm_enabled();
+	peerlbl_active = selinux_peerlbl_enabled();
 	if (!secmark_active && !peerlbl_active)
 		return NF_ACCEPT;
 
Signature · 7 lines
1.7.8.6


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
maintainer
#2 lore raw

Re: [PATCH 2/2] Update SELinux policy capability to always check peer class.

Quoted 5 lines · Chris PeBenito

Update the always_check_network policy capability which, when enabled, treats peer labeling as enabled, even if there is no Netlabel or labeled IPSEC configuration.

...

security/selinux/hooks.c static int selinux_secmark_enabled(void) · line 155

Quoted 12 lines
security/selinux/hooks.c +7 -0
index ec7151b..e8f612e 100644
@@ -149,6 +149,24 @@ static int selinux_secmark_enabled(void)
+static int selinux_peerlbl_enabled(void)
+{
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (netlbl_enabled() || selinux_xfrm_enabled());
+}

Why not make this more consistent?

return (selinux_policycap_alwaysnetwork || ...

Signature · 8 lines
paul moore
www.paul-moore.com


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#3 lore raw

Re: [PATCH 2/2] Update SELinux policy capability to always check peer class.

Quoted 24 lines · Paul Moore
Quoted 5 lines · Chris PeBenito

Update the always_check_network policy capability which, when enabled, treats peer labeling as enabled, even if there is no Netlabel or labeled IPSEC configuration.

...

security/selinux/hooks.c static int selinux_secmark_enabled(void) · line 155

Quoted 12 lines
security/selinux/hooks.c +7 -0
index ec7151b..e8f612e 100644
@@ -149,6 +149,24 @@ static int selinux_secmark_enabled(void)
+static int selinux_peerlbl_enabled(void)
+{
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (netlbl_enabled() || selinux_xfrm_enabled());
+}

Why not make this more consistent?

Quoted 1 line

return (selinux_policycap_alwaysnetwork || ...

Same response as the other patch.

Signature · 8 lines
Chris PeBenito
Tresys Technology, LLC
www.tresys.com | oss.tresys.com

--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
maintainer
#4 lore raw
In reply to Chris PeBenito
Quoted 10 lines · Chris PeBenito

Currently the packet class in SELinux is not checked if there are no SECMARK rules in the security or mangle netfilter tables. Some systems prefer that packets are always checked, for example, to protect the system should the netfilter rules fail to load or if the nefilter rules were maliciously flushed.

Add the always_check_network policy capability which, when enabled, treats SECMARK as enabled, even if there are no netfilter SECMARK rules.

...

Quoted 4 lines
security/selinux/hooks.c +0 -0
index 372ec65..ec7151b 100644

...

security/selinux/hooks.c static struct kmem_cache *sel_inode_cache; · line 150

Quoted 8 lines

static int selinux_secmark_enabled(void)
{

security/selinux/hooks.c +4 -1
-	return (atomic_read(&selinux_secmark_refcount) > 0);
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (atomic_read(&selinux_secmark_refcount) > 0);
 }

Nit picky, but why not simply:

return (selinux_policycap_alwaysnetwork || atomic_read( ...

Quoted 7 lines

/*

(quoted diff) +0 -0

b/security/selinux/include/security.h index dde2005..981c4ac 100644

security/selinux/include/security.h +0 -0
@@ -68,12 +68,14 @@ extern int selinux_enabled;
 enum {

security/selinux/include/security.h extern int selinux_enabled; · line 78

Quoted 10 lines
security/selinux/include/security.h +2 -0
 	POLICYDB_CAPABILITY_NETPEER,
 	POLICYDB_CAPABILITY_OPENPERM,
+	POLICYDB_CAPABILITY_ALWAYSNETWORK,
 	__POLICYDB_CAPABILITY_MAX
 };
 #define POLICYDB_CAPABILITY_MAX (__POLICYDB_CAPABILITY_MAX - 1)
 
 extern int selinux_policycap_netpeer;
 extern int selinux_policycap_openperm;
+extern int selinux_policycap_alwaysnetwork;

Also nit picky, but it would seem like "selinux_policycap_netalways" is a bit more consistent with the other variables.

security/selinux/selinuxfs.c line 49

Quoted 12 lines
security/selinux/selinuxfs.c +2 -1
index 3ad2902..cb893f9 100644
@@ -44,7 +44,8 @@
 /* Policy capability filenames */
 static char *policycap_names[] = {
 	"network_peer_controls",
-	"open_perms"
+	"open_perms",
+	"always_check_network"
 };

Similarly, I think "network_always" is more consistent.

security/selinux/ss/services.c line 75

Quoted 9 lines

a/security/selinux/ss/services.c b/security/selinux/ss/services.c index 4321b8f..e124d8f 100644

security/selinux/ss/services.c +1 -0
@@ -72,6 +72,7 @@
 
 int selinux_policycap_netpeer;
 int selinux_policycap_openperm;
+int selinux_policycap_alwaysnetwork;

See above.

Signature · 8 lines
paul moore
www.paul-moore.com


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#5 lore raw
Quoted 32 lines · Paul Moore
Quoted 10 lines · Chris PeBenito

Currently the packet class in SELinux is not checked if there are no SECMARK rules in the security or mangle netfilter tables. Some systems prefer that packets are always checked, for example, to protect the system should the netfilter rules fail to load or if the nefilter rules were maliciously flushed.

Add the always_check_network policy capability which, when enabled, treats SECMARK as enabled, even if there are no netfilter SECMARK rules.

...

Quoted 4 lines
security/selinux/hooks.c +0 -0
index 372ec65..ec7151b 100644

...

security/selinux/hooks.c static struct kmem_cache *sel_inode_cache; · line 150

Quoted 8 lines

static int selinux_secmark_enabled(void)
{

security/selinux/hooks.c +4 -1
-	return (atomic_read(&selinux_secmark_refcount) > 0);
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (atomic_read(&selinux_secmark_refcount) > 0);
 }

Nit picky, but why not simply:

Quoted 1 line

return (selinux_policycap_alwaysnetwork || atomic_read( ...

Sure it can be done that way, but I'd argue that the intent is clearer the way its implemented in the patch.

Quoted 14 lines
Quoted 14 lines

/*

(quoted diff) +0 -0

b/security/selinux/include/security.h index dde2005..981c4ac 100644

security/selinux/include/security.h +1 -0
@@ -68,12 +68,14 @@ extern int selinux_enabled;
 enum {
 	POLICYDB_CAPABILITY_NETPEER,
 	POLICYDB_CAPABILITY_OPENPERM,
+	POLICYDB_CAPABILITY_ALWAYSNETWORK,
 	__POLICYDB_CAPABILITY_MAX
 };
 #define POLICYDB_CAPABILITY_MAX (__POLICYDB_CAPABILITY_MAX - 1)
 
Quoted 6 lines

security/selinux/include/security.h extern int selinux_enabled; · line 78

Quoted 3 lines
security/selinux/include/security.h +1 -0
 extern int selinux_policycap_netpeer;
 extern int selinux_policycap_openperm;
+extern int selinux_policycap_alwaysnetwork;

Also nit picky, but it would seem like "selinux_policycap_netalways" is a bit more consistent with the other variables.

This response stems on what you wrote on a later cunch about the capability name, so you should probably read below first.

I don't agree. If you go by current capabilities names/variables, based on the ordering of the words in the capability name, the words in the variable are in the same order. e.g.

NETwork_PEER_controls -> selinux_policycap_netpeer OPEN_PERMs -> selinux_policycap_openperm

hence

ALWAYS_check_NETWORK -> selinux_policycap_alwaysnetwork

Quoted 6 lines
Quoted 6 lines
security/selinux/selinuxfs.c +0 -0
index 3ad2902..cb893f9 100644
@@ -44,7 +44,8 @@
 /* Policy capability filenames */
Quoted 8 lines

security/selinux/selinuxfs.c line 49

Quoted 6 lines
security/selinux/selinuxfs.c +2 -1
 static char *policycap_names[] = {
 	"network_peer_controls",
-	"open_perms"
+	"open_perms",
+	"always_check_network"
 };

Similarly, I think "network_always" is more consistent.

Earlier discussions concluded with this capability name. (to be specific the discussion resolved with always_check_packets, but that was before the peer class was included)

Signature · 8 lines
Chris PeBenito
Tresys Technology, LLC
www.tresys.com | oss.tresys.com

--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
maintainer
#6 lore raw
Quoted 11 lines · Christopher J. PeBenito
Quoted 10 lines · Paul Moore
Quoted 9 lines · Chris PeBenito

static int selinux_secmark_enabled(void)
{

security/selinux/hooks.c +4 -1
-	return (atomic_read(&selinux_secmark_refcount) > 0);
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (atomic_read(&selinux_secmark_refcount) > 0);
Quoted 7 lines
Quoted 4 lines
Quoted 1 line

}

Nit picky, but why not simply:
return (selinux_policycap_alwaysnetwork || atomic_read( ...

Sure it can be done that way, but I'd argue that the intent is clearer the way its implemented in the patch.

It looks very hacky and inconsistent to me.

Oh well, you can sort it out with Eric.

Quoted 21 lines
Quoted 21 lines
Quoted 21 lines
(quoted diff) +0 -0

b/security/selinux/include/security.h index dde2005..981c4ac 100644

security/selinux/include/security.h +1 -0
@@ -68,12 +68,14 @@ extern int selinux_enabled;
 
 enum {
 
 	POLICYDB_CAPABILITY_NETPEER,
 	POLICYDB_CAPABILITY_OPENPERM,
 
+	POLICYDB_CAPABILITY_ALWAYSNETWORK,
 
 	__POLICYDB_CAPABILITY_MAX
 
 };
 #define POLICYDB_CAPABILITY_MAX (__POLICYDB_CAPABILITY_MAX - 1)
 
 extern int selinux_policycap_netpeer;
 extern int selinux_policycap_openperm;
Quoted 7 lines
Quoted 4 lines
Quoted 1 line

+extern int selinux_policycap_alwaysnetwork;

Also nit picky, but it would seem like "selinux_policycap_netalways" is a bit more consistent with the other variables.

This response stems on what you wrote on a later cunch about the capability name, so you should probably read below first.

What?

I don't believe I ever commented on the capability names until now, please correct me with a link to the archives if I'm mistaken.

Quoted 10 lines

I don't agree. If you go by current capabilities names/variables, based on the ordering of the words in the capability name, the words in the variable are in the same order. e.g.

NETwork_PEER_controls -> selinux_policycap_netpeer OPEN_PERMs -> selinux_policycap_openperm

hence

ALWAYS_check_NETWORK -> selinux_policycap_alwaysnetwork

I look at it as a bit of "namespacing" ... "netpeer" enables the network peer controls, "openperm" affects the open permission. Since this capability affects the network controls I believe "netalways" makes more sense.

Quoted 15 lines
Quoted 15 lines
Quoted 15 lines
security/selinux/selinuxfs.c +2 -1
index 3ad2902..cb893f9 100644
@@ -44,7 +44,8 @@
 
 /* Policy capability filenames */
 static char *policycap_names[] = {
 
 	"network_peer_controls",
 
-	"open_perms"
+	"open_perms",
+	"always_check_network"
Quoted 7 lines
Quoted 3 lines
Quoted 1 line

};

Similarly, I think "network_always" is more consistent.

Earlier discussions concluded with this capability name. (to be specific the discussion resolved with always_check_packets, but that was before the peer class was included)

Once again, I don't believe I ever commented on the name of the capability, but I am now. Disregard my comments or incorporate them, that is up to you, and ultimately Eric, to decide.

Signature · 8 lines
paul moore
www.paul-moore.com


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#7 lore raw
Quoted 21 lines · Paul Moore
Quoted 19 lines · Christopher J. PeBenito
Quoted 18 lines
Quoted 15 lines
security/selinux/selinuxfs.c +2 -1
index 3ad2902..cb893f9 100644
@@ -44,7 +44,8 @@
 
 /* Policy capability filenames */
 static char *policycap_names[] = {
 
 	"network_peer_controls",
 
-	"open_perms"
+	"open_perms",
+	"always_check_network"
Quoted 1 line

};

Similarly, I think "network_always" is more consistent.

Quoted 7 lines
Quoted 3 lines

Earlier discussions concluded with this capability name. (to be specific the discussion resolved with always_check_packets, but that was before the peer class was included)

Once again, I don't believe I ever commented on the name of the capability, but I am now. Disregard my comments or incorporate them, that is up to you, and ultimately Eric, to decide.

My apologies, you did not comment on the capability name in the previous thread.

Signature · 8 lines
Chris PeBenito
Tresys Technology, LLC
www.tresys.com | oss.tresys.com

--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#8 lore raw
In reply to Christopher J. PeBenito
Quoted 30 lines · Christopher J. PeBenito
Quoted 29 lines · Paul Moore
Quoted 10 lines · Chris PeBenito

Currently the packet class in SELinux is not checked if there are no SECMARK rules in the security or mangle netfilter tables. Some systems prefer that packets are always checked, for example, to protect the system should the netfilter rules fail to load or if the nefilter rules were maliciously flushed.

Add the always_check_network policy capability which, when enabled, treats SECMARK as enabled, even if there are no netfilter SECMARK rules.

...

Quoted 4 lines
security/selinux/hooks.c +0 -0
index 372ec65..ec7151b 100644

...

security/selinux/hooks.c static struct kmem_cache *sel_inode_cache; · line 150

Quoted 8 lines

static int selinux_secmark_enabled(void)
{

security/selinux/hooks.c +4 -1
-	return (atomic_read(&selinux_secmark_refcount) > 0);
+	if (selinux_policycap_alwaysnetwork)
+		return 1;
+	else
+		return (atomic_read(&selinux_secmark_refcount) > 0);
 }

Nit picky, but why not simply:

Quoted 2 lines
Quoted 1 line

return (selinux_policycap_alwaysnetwork || atomic_read( ...

Sure it can be done that way, but I'd argue that the intent is clearer the way its implemented in the patch.

Oh come on, no one does

    if (...)
            return x;
    else
            return y;

Much better to do

    if (...)
            return x;
    return y;

The else is unnecessary.

Quoted 45 lines
Quoted 16 lines
Quoted 16 lines

/*

(quoted diff) +0 -0

b/security/selinux/include/security.h index dde2005..981c4ac 100644

security/selinux/include/security.h +1 -0
@@ -68,12 +68,14 @@ extern int selinux_enabled;
 enum {
 	POLICYDB_CAPABILITY_NETPEER,
 	POLICYDB_CAPABILITY_OPENPERM,
+	POLICYDB_CAPABILITY_ALWAYSNETWORK,
 	__POLICYDB_CAPABILITY_MAX
 };
 #define POLICYDB_CAPABILITY_MAX (__POLICYDB_CAPABILITY_MAX - 1)
 
Quoted 5 lines

security/selinux/include/security.h extern int selinux_enabled; · line 78

Quoted 3 lines
security/selinux/include/security.h +1 -0
 extern int selinux_policycap_netpeer;
 extern int selinux_policycap_openperm;
+extern int selinux_policycap_alwaysnetwork;

Also nit picky, but it would seem like "selinux_policycap_netalways" is a bit more consistent with the other variables.

This response stems on what you wrote on a later cunch about the capability name, so you should probably read below first.

I don't agree. If you go by current capabilities names/variables, based on the ordering of the words in the capability name, the words in the variable are in the same order. e.g.

NETwork_PEER_controls -> selinux_policycap_netpeer OPEN_PERMs -> selinux_policycap_openperm

hence

ALWAYS_check_NETWORK -> selinux_policycap_alwaysnetwork

Quoted 5 lines
Quoted 5 lines
security/selinux/selinuxfs.c +0 -0
index 3ad2902..cb893f9 100644
@@ -44,7 +44,8 @@
Quoted 8 lines

security/selinux/selinuxfs.c line 49

Quoted 7 lines
security/selinux/selinuxfs.c +2 -1
 /* Policy capability filenames */
 static char *policycap_names[] = {
 	"network_peer_controls",
-	"open_perms"
+	"open_perms",
+	"always_check_network"
 };

Similarly, I think "network_always" is more consistent.

Quoted 3 lines

Earlier discussions concluded with this capability name. (to be specific the discussion resolved with always_check_packets, but that was before the peer class was included)

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.