AppArmor: compatibility patch for v5 network controll
authorJohn Johansen <john.johansen@canonical.com>
Mon, 4 Oct 2010 22:03:36 +0000 (15:03 -0700)
committerLeann Ogasawara <leann.ogasawara@canonical.com>
Mon, 28 Mar 2011 13:49:56 +0000 (06:49 -0700)
Add compatibility for v5 network rules.

Signed-off-by: John Johansen <john.johansen@canonical.com>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>

include/linux/lsm_audit.h
security/apparmor/Makefile
security/apparmor/include/net.h [new file with mode: 0644]
security/apparmor/include/policy.h
security/apparmor/lsm.c
security/apparmor/net.c [new file with mode: 0644]
security/apparmor/policy.c
security/apparmor/policy_unpack.c

index 112a550..d5f3dd7 100644 (file)
@@ -123,6 +123,10 @@ struct common_audit_data {
                                        u32 denied;
                                        uid_t ouid;
                                } fs;
+                               struct {
+                                       int type, protocol;
+                                       struct sock *sk;
+                               } net;
                        };
                } apparmor_audit_data;
 #endif
index f204869..a9a1db0 100644 (file)
@@ -4,17 +4,21 @@ obj-$(CONFIG_SECURITY_APPARMOR) += apparmor.o
 
 apparmor-y := apparmorfs.o audit.o capability.o context.o ipc.o lib.o match.o \
               path.o domain.o policy.o policy_unpack.o procattr.o lsm.o \
-              resource.o sid.o file.o
+              resource.o sid.o file.o net.o
 
 clean-files: capability_names.h af_names.h
 
 quiet_cmd_make-caps = GEN     $@
 cmd_make-caps = echo "static const char *capability_names[] = {" > $@ ; sed -n -e "/CAP_FS_MASK/d" -e "s/^\#define[ \\t]\\+CAP_\\([A-Z0-9_]\\+\\)[ \\t]\\+\\([0-9]\\+\\)\$$/[\\2]  = \"\\1\",/p" $< | tr A-Z a-z >> $@ ; echo "};" >> $@
 
+quiet_cmd_make-af = GEN     $@
+cmd_make-af = echo "static const char *address_family_names[] = {" > $@ ; sed -n -e "/AF_MAX/d" -e "/AF_LOCAL/d" -e "s/^\#define[ \\t]\\+AF_\\([A-Z0-9_]\\+\\)[ \\t]\\+\\([0-9]\\+\\)\\(.*\\)\$$/[\\2]  = \"\\1\",/p" $< | tr A-Z a-z >> $@ ; echo "};" >> $@
+
 quiet_cmd_make-rlim = GEN     $@
 cmd_make-rlim = echo "static const char *rlim_names[] = {" > $@ ; sed -n --e "/AF_MAX/d" -e "s/^\# \\?define[ \\t]\\+RLIMIT_\\([A-Z0-9_]\\+\\)[ \\t]\\+\\([0-9]\\+\\)\\(.*\\)\$$/[\\2]  = \"\\1\",/p" $< | tr A-Z a-z >> $@ ; echo "};" >> $@ ; echo "static const int rlim_map[] = {" >> $@ ; sed -n -e "/AF_MAX/d" -e "s/^\# \\?define[ \\t]\\+\\(RLIMIT_[A-Z0-9_]\\+\\)[ \\t]\\+\\([0-9]\\+\\)\\(.*\\)\$$/\\1,/p" $< >> $@ ; echo "};" >> $@
 
 $(obj)/capability.o : $(obj)/capability_names.h
+$(obj)/net.o : $(obj)/af_names.h
 $(obj)/resource.o : $(obj)/rlim_names.h
 $(obj)/capability_names.h : $(srctree)/include/linux/capability.h
        $(call cmd,make-caps)
diff --git a/security/apparmor/include/net.h b/security/apparmor/include/net.h
new file mode 100644 (file)
index 0000000..3c7d599
--- /dev/null
@@ -0,0 +1,40 @@
+/*
+ * AppArmor security module
+ *
+ * This file contains AppArmor network mediation definitions.
+ *
+ * Copyright (C) 1998-2008 Novell/SUSE
+ * Copyright 2009-2010 Canonical Ltd.
+ *
+ * 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, version 2 of the
+ * License.
+ */
+
+#ifndef __AA_NET_H
+#define __AA_NET_H
+
+#include <net/sock.h>
+
+/* struct aa_net - network confinement data
+ * @allowed: basic network families permissions
+ * @audit_network: which network permissions to force audit
+ * @quiet_network: which network permissions to quiet rejects
+ */
+struct aa_net {
+       u16 allow[AF_MAX];
+       u16 audit[AF_MAX];
+       u16 quiet[AF_MAX];
+};
+
+extern int aa_net_perm(int op, struct aa_profile *profile, u16 family,
+                      int type, int protocol, struct sock *sk);
+extern int aa_revalidate_sk(int op, struct sock *sk);
+
+static inline void aa_free_net_rules(struct aa_net *new)
+{
+       /* NOP */
+}
+
+#endif /* __AA_NET_H */
index aeda5cf..6776929 100644 (file)
@@ -27,6 +27,7 @@
 #include "capability.h"
 #include "domain.h"
 #include "file.h"
+#include "net.h"
 #include "resource.h"
 
 extern const char *profile_mode_names[];
@@ -145,6 +146,7 @@ struct aa_namespace {
  * @size: the memory consumed by this profiles rules
  * @file: The set of rules governing basic file access and domain transitions
  * @caps: capabilities for the profile
+ * @net: network controls for the profile
  * @rlimits: rlimits for the profile
  *
  * The AppArmor profile contains the basic confinement data.  Each profile
@@ -181,6 +183,7 @@ struct aa_profile {
 
        struct aa_file_rules file;
        struct aa_caps caps;
+       struct aa_net net;
        struct aa_rlimit rlimits;
 };
 
index b7106f1..fa778a7 100644 (file)
@@ -31,6 +31,7 @@
 #include "include/context.h"
 #include "include/file.h"
 #include "include/ipc.h"
+#include "include/net.h"
 #include "include/path.h"
 #include "include/policy.h"
 #include "include/procattr.h"
@@ -619,6 +620,104 @@ static int apparmor_task_setrlimit(struct task_struct *task,
        return error;
 }
 
+static int apparmor_socket_create(int family, int type, int protocol, int kern)
+{
+       struct aa_profile *profile;
+       int error = 0;
+
+       if (kern)
+               return 0;
+
+       profile = __aa_current_profile();
+       if (!unconfined(profile))
+               error = aa_net_perm(OP_CREATE, profile, family, type, protocol,
+                                   NULL);
+       return error;
+}
+
+static int apparmor_socket_bind(struct socket *sock,
+                               struct sockaddr *address, int addrlen)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_BIND, sk);
+}
+
+static int apparmor_socket_connect(struct socket *sock,
+                                  struct sockaddr *address, int addrlen)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_CONNECT, sk);
+}
+
+static int apparmor_socket_listen(struct socket *sock, int backlog)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_LISTEN, sk);
+}
+
+static int apparmor_socket_accept(struct socket *sock, struct socket *newsock)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_ACCEPT, sk);
+}
+
+static int apparmor_socket_sendmsg(struct socket *sock,
+                                  struct msghdr *msg, int size)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_SENDMSG, sk);
+}
+
+static int apparmor_socket_recvmsg(struct socket *sock,
+                                  struct msghdr *msg, int size, int flags)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_RECVMSG, sk);
+}
+
+static int apparmor_socket_getsockname(struct socket *sock)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_GETSOCKNAME, sk);
+}
+
+static int apparmor_socket_getpeername(struct socket *sock)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_GETPEERNAME, sk);
+}
+
+static int apparmor_socket_getsockopt(struct socket *sock, int level,
+                                     int optname)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_GETSOCKOPT, sk);
+}
+
+static int apparmor_socket_setsockopt(struct socket *sock, int level,
+                                     int optname)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_SETSOCKOPT, sk);
+}
+
+static int apparmor_socket_shutdown(struct socket *sock, int how)
+{
+       struct sock *sk = sock->sk;
+
+       return aa_revalidate_sk(OP_SOCK_SHUTDOWN, sk);
+}
+
 static struct security_operations apparmor_ops = {
        .name =                         "apparmor",
 
@@ -650,6 +749,19 @@ static struct security_operations apparmor_ops = {
        .getprocattr =                  apparmor_getprocattr,
        .setprocattr =                  apparmor_setprocattr,
 
+       .socket_create =                apparmor_socket_create,
+       .socket_bind =                  apparmor_socket_bind,
+       .socket_connect =               apparmor_socket_connect,
+       .socket_listen =                apparmor_socket_listen,
+       .socket_accept =                apparmor_socket_accept,
+       .socket_sendmsg =               apparmor_socket_sendmsg,
+       .socket_recvmsg =               apparmor_socket_recvmsg,
+       .socket_getsockname =           apparmor_socket_getsockname,
+       .socket_getpeername =           apparmor_socket_getpeername,
+       .socket_getsockopt =            apparmor_socket_getsockopt,
+       .socket_setsockopt =            apparmor_socket_setsockopt,
+       .socket_shutdown =              apparmor_socket_shutdown,
+
        .cred_alloc_blank =             apparmor_cred_alloc_blank,
        .cred_free =                    apparmor_cred_free,
        .cred_prepare =                 apparmor_cred_prepare,
diff --git a/security/apparmor/net.c b/security/apparmor/net.c
new file mode 100644 (file)
index 0000000..1765901
--- /dev/null
@@ -0,0 +1,170 @@
+/*
+ * AppArmor security module
+ *
+ * This file contains AppArmor network mediation
+ *
+ * Copyright (C) 1998-2008 Novell/SUSE
+ * Copyright 2009-2010 Canonical Ltd.
+ *
+ * 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, version 2 of the
+ * License.
+ */
+
+#include "include/apparmor.h"
+#include "include/audit.h"
+#include "include/context.h"
+#include "include/net.h"
+#include "include/policy.h"
+
+#include "af_names.h"
+
+static const char *sock_type_names[] = {
+       "unknown(0)",
+       "stream",
+       "dgram",
+       "raw",
+       "rdm",
+       "seqpacket",
+       "dccp",
+       "unknown(7)",
+       "unknown(8)",
+       "unknown(9)",
+       "packet",
+};
+
+/* audit callback for net specific fields */
+static void audit_cb(struct audit_buffer *ab, void *va)
+{
+       struct common_audit_data *sa = va;
+
+       audit_log_format(ab, " family=");
+       if (address_family_names[sa->u.net.family]) {
+               audit_log_string(ab, address_family_names[sa->u.net.family]);
+       } else {
+               audit_log_format(ab, " \"unknown(%d)\"", sa->u.net.family);
+       }
+
+       audit_log_format(ab, " sock_type=");
+       if (sock_type_names[sa->aad.net.type]) {
+               audit_log_string(ab, sock_type_names[sa->aad.net.type]);
+       } else {
+               audit_log_format(ab, "\"unknown(%d)\"", sa->aad.net.type);
+       }
+
+       audit_log_format(ab, " protocol=%d", sa->aad.net.protocol);
+}
+
+/**
+ * audit_net - audit network access
+ * @profile: profile being enforced  (NOT NULL)
+ * @op: operation being checked
+ * @family: network family
+ * @type:   network type
+ * @protocol: network protocol
+ * @sk: socket auditing is being applied to
+ * @error: error code for failure else 0
+ *
+ * Returns: %0 or sa->error else other errorcode on failure
+ */
+static int audit_net(struct aa_profile *profile, int op, u16 family, int type,
+                    int protocol, struct sock *sk, int error)
+{
+       int audit_type = AUDIT_APPARMOR_AUTO;
+       struct common_audit_data sa;
+       if (sk) {
+               COMMON_AUDIT_DATA_INIT(&sa, NET);
+       } else {
+               COMMON_AUDIT_DATA_INIT(&sa, NONE);
+       }
+       /* todo fill in socket addr info */
+
+       sa.aad.op = op,
+       sa.u.net.family = family;
+       sa.u.net.sk = sk;
+       sa.aad.net.type = type;
+       sa.aad.net.protocol = protocol;
+       sa.aad.error = error;
+
+       if (likely(!sa.aad.error)) {
+               u16 audit_mask = profile->net.audit[sa.u.net.family];
+               if (likely((AUDIT_MODE(profile) != AUDIT_ALL) &&
+                          !(1 << sa.aad.net.type & audit_mask)))
+                       return 0;
+               audit_type = AUDIT_APPARMOR_AUDIT;
+       } else {
+               u16 quiet_mask = profile->net.quiet[sa.u.net.family];
+               u16 kill_mask = 0;
+               u16 denied = (1 << sa.aad.net.type) & ~quiet_mask;
+
+               if (denied & kill_mask)
+                       audit_type = AUDIT_APPARMOR_KILL;
+
+               if ((denied & quiet_mask) &&
+                   AUDIT_MODE(profile) != AUDIT_NOQUIET &&
+                   AUDIT_MODE(profile) != AUDIT_ALL)
+                       return COMPLAIN_MODE(profile) ? 0 : sa.aad.error;
+       }
+
+       return aa_audit(audit_type, profile, GFP_KERNEL, &sa, audit_cb);
+}
+
+/**
+ * aa_net_perm - very course network access check
+ * @op: operation being checked
+ * @profile: profile being enforced  (NOT NULL)
+ * @family: network family
+ * @type:   network type
+ * @protocol: network protocol
+ *
+ * Returns: %0 else error if permission denied
+ */
+int aa_net_perm(int op, struct aa_profile *profile, u16 family, int type,
+               int protocol, struct sock *sk)
+{
+       u16 family_mask;
+       int error;
+
+       if ((family < 0) || (family >= AF_MAX))
+               return -EINVAL;
+
+       if ((type < 0) || (type >= SOCK_MAX))
+               return -EINVAL;
+
+       /* unix domain and netlink sockets are handled by ipc */
+       if (family == AF_UNIX || family == AF_NETLINK)
+               return 0;
+
+       family_mask = profile->net.allow[family];
+
+       error = (family_mask & (1 << type)) ? 0 : -EACCES;
+
+       return audit_net(profile, op, family, type, protocol, sk, error);
+}
+
+/**
+ * aa_revalidate_sk - Revalidate access to a sock
+ * @op: operation being checked
+ * @sk: sock being revalidated  (NOT NULL)
+ *
+ * Returns: %0 else error if permission denied
+ */
+int aa_revalidate_sk(int op, struct sock *sk)
+{
+       struct aa_profile *profile;
+       int error = 0;
+
+       /* aa_revalidate_sk should not be called from interrupt context
+        * don't mediate these calls as they are not task related
+        */
+       if (in_interrupt())
+               return 0;
+
+       profile = __aa_current_profile();
+       if (!unconfined(profile))
+               error = aa_net_perm(op, profile, sk->sk_family, sk->sk_type,
+                                   sk->sk_protocol, sk);
+
+       return error;
+}
index 4f0eade..4d5ce13 100644 (file)
@@ -745,6 +745,7 @@ static void free_profile(struct aa_profile *profile)
 
        aa_free_file_rules(&profile->file);
        aa_free_cap_rules(&profile->caps);
+       aa_free_net_rules(&profile->net);
        aa_free_rlimit_rules(&profile->rlimits);
 
        aa_free_sid(profile->sid);
index eb3700e..c2b6225 100644 (file)
@@ -190,6 +190,19 @@ fail:
        return 0;
 }
 
+static bool unpack_u16(struct aa_ext *e, u16 *data, const char *name)
+{
+       if (unpack_nameX(e, AA_U16, name)) {
+               if (!inbounds(e, sizeof(u16)))
+                       return 0;
+               if (data)
+                       *data = le16_to_cpu(get_unaligned((u16 *) e->pos));
+               e->pos += sizeof(u16);
+               return 1;
+       }
+       return 0;
+}
+
 static bool unpack_u32(struct aa_ext *e, u32 *data, const char *name)
 {
        if (unpack_nameX(e, AA_U32, name)) {
@@ -468,7 +481,8 @@ static struct aa_profile *unpack_profile(struct aa_ext *e)
 {
        struct aa_profile *profile = NULL;
        const char *name = NULL;
-       int error = -EPROTO;
+       size_t size = 0;
+       int i, error = -EPROTO;
        kernel_cap_t tmpcap;
        u32 tmp;
 
@@ -559,6 +573,38 @@ static struct aa_profile *unpack_profile(struct aa_ext *e)
        if (!unpack_rlimits(e, profile))
                goto fail;
 
+       size = unpack_array(e, "net_allowed_af");
+       if (size) {
+
+               for (i = 0; i < size; i++) {
+                       /* discard extraneous rules that this kernel will
+                        * never request
+                        */
+                       if (i > AF_MAX) {
+                               u16 tmp;
+                               if (!unpack_u16(e, &tmp, NULL) ||
+                                   !unpack_u16(e, &tmp, NULL) ||
+                                   !unpack_u16(e, &tmp, NULL))
+                                       goto fail;
+                               continue;
+                       }
+                       if (!unpack_u16(e, &profile->net.allow[i], NULL))
+                               goto fail;
+                       if (!unpack_u16(e, &profile->net.audit[i], NULL))
+                               goto fail;
+                       if (!unpack_u16(e, &profile->net.quiet[i], NULL))
+                               goto fail;
+               }
+               if (!unpack_nameX(e, AA_ARRAYEND, NULL))
+                       goto fail;
+               /*
+                * allow unix domain and netlink sockets they are handled
+                * by IPC
+                */
+       }
+       profile->net.allow[AF_UNIX] = 0xffff;
+       profile->net.allow[AF_NETLINK] = 0xffff;
+
        /* get file rules */
        profile->file.dfa = unpack_dfa(e);
        if (IS_ERR(profile->file.dfa)) {