diff mbox series

[V4,1/3] selinux-testsuite: Add BPF tests

Message ID 20190917145640.25629-2-richard_c_haines@btinternet.com (mailing list archive)
State Changes Requested
Headers show
Series selinux-testsuite: Add BPF tests | expand

Commit Message

Richard Haines Sept. 17, 2019, 2:56 p.m. UTC
This adds basic BPF tests for map and prog functions.

Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
---
 README.md              |  4 ++-
 defconfig              |  5 +++
 policy/Makefile        |  4 +++
 policy/test_bpf.te     | 78 ++++++++++++++++++++++++++++++++++++++++++
 tests/Makefile         |  7 ++++
 tests/bpf/.gitignore   |  2 ++
 tests/bpf/Makefile     | 10 ++++++
 tests/bpf/bpf_common.c | 53 ++++++++++++++++++++++++++++
 tests/bpf/bpf_common.h | 34 ++++++++++++++++++
 tests/bpf/bpf_test.c   | 77 +++++++++++++++++++++++++++++++++++++++++
 tests/bpf/test         | 64 ++++++++++++++++++++++++++++++++++
 11 files changed, 337 insertions(+), 1 deletion(-)
 create mode 100644 policy/test_bpf.te
 create mode 100644 tests/bpf/.gitignore
 create mode 100644 tests/bpf/Makefile
 create mode 100644 tests/bpf/bpf_common.c
 create mode 100644 tests/bpf/bpf_common.h
 create mode 100644 tests/bpf/bpf_test.c
 create mode 100755 tests/bpf/test

Comments

Stephen Smalley Sept. 18, 2019, 2:58 p.m. UTC | #1
On 9/17/19 10:56 AM, Richard Haines wrote:
> This adds basic BPF tests for map and prog functions.
> 
> Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
> ---
>   README.md              |  4 ++-
>   defconfig              |  5 +++
>   policy/Makefile        |  4 +++
>   policy/test_bpf.te     | 78 ++++++++++++++++++++++++++++++++++++++++++
>   tests/Makefile         |  7 ++++
>   tests/bpf/.gitignore   |  2 ++
>   tests/bpf/Makefile     | 10 ++++++
>   tests/bpf/bpf_common.c | 53 ++++++++++++++++++++++++++++
>   tests/bpf/bpf_common.h | 34 ++++++++++++++++++
>   tests/bpf/bpf_test.c   | 77 +++++++++++++++++++++++++++++++++++++++++
>   tests/bpf/test         | 64 ++++++++++++++++++++++++++++++++++
>   11 files changed, 337 insertions(+), 1 deletion(-)
>   create mode 100644 policy/test_bpf.te
>   create mode 100644 tests/bpf/.gitignore
>   create mode 100644 tests/bpf/Makefile
>   create mode 100644 tests/bpf/bpf_common.c
>   create mode 100644 tests/bpf/bpf_common.h
>   create mode 100644 tests/bpf/bpf_test.c
>   create mode 100755 tests/bpf/test
> 
> diff --git a/README.md b/README.md
> index 26784f8..1396c8e 100644
> --- a/README.md
> +++ b/README.md
> @@ -51,6 +51,7 @@ similar dependencies):
>   * iptables _(to load the `iptables SECMARK` rules during `inet_socket` tests)_
>   * lksctp-tools-devel _(to build the SCTP test programs)_
>   * attr _(tools used by the overlayfs tests)_
> +* libbpf-devel _(tools used by the bpf tests)_
>   
>   On a modern Fedora system you can install these dependencies with the
>   following command:
> @@ -65,7 +66,8 @@ following command:
>   		netlabel_tools \
>   		iptables \
>   		lksctp-tools-devel \
> -		attr
> +		attr \
> +		libbpf-devel
>   
>   The testsuite requires a pre-existing base policy configuration of SELinux,
>   using either the old example policy or the reference policy as the baseline.
> diff --git a/defconfig b/defconfig
> index d7f0ea5..cb57f22 100644
> --- a/defconfig
> +++ b/defconfig
> @@ -62,3 +62,8 @@ CONFIG_ANDROID_BINDER_IPC=y
>   # This will configure the Dynamically Allocated Binder Devices added
>   # to 5.0+ kernels:
>   CONFIG_ANDROID_BINDERFS=y
> +
> +# Test BPF + check in selinux_file_receive and selinux_binder_transfer_files.
> +# They are not required for SELinux operation itself.
> +CONFIG_BPF=y
> +CONFIG_BPF_SYSCALL=y
> diff --git a/policy/Makefile b/policy/Makefile
> index 305b572..16a4469 100644
> --- a/policy/Makefile
> +++ b/policy/Makefile
> @@ -71,6 +71,10 @@ ifeq ($(shell grep -q corenet_sctp_bind_all_nodes $(POLDEV)/include/kernel/coren
>   TARGETS += test_sctp.te
>   endif
>   
> +ifeq ($(shell grep -q bpf $(POLDEV)/include/support/all_perms.spt && echo true),true)
> +TARGETS += test_bpf.te
> +endif
> +
>   ifeq (x$(DISTRO),$(filter x$(DISTRO),xRHEL4 xRHEL5 xRHEL6))
>   TARGETS:=$(filter-out test_overlayfs.te test_mqueue.te, $(TARGETS))
>   endif
> diff --git a/policy/test_bpf.te b/policy/test_bpf.te
> new file mode 100644
> index 0000000..89d240c
> --- /dev/null
> +++ b/policy/test_bpf.te
> @@ -0,0 +1,78 @@
> +#
> +################# BPF selinux-testsuite policy module ######################
> +#
> +
> +attribute bpfdomain;
> +
> +################################### Main ###################################
> +type test_bpf_t;
> +domain_type(test_bpf_t)
> +unconfined_runs_test(test_bpf_t)
> +typeattribute test_bpf_t testdomain;
> +typeattribute test_bpf_t bpfdomain;
> +
> +allow test_bpf_t self:process { setrlimit };
> +#allow test_bpf_t self:capability { sys_resource sys_admin };
> +allow test_bpf_t self:capability { sys_resource };
> +allow test_bpf_t self:bpf { map_create map_read map_write prog_load prog_run };
> +
> +############################## Deny map_create #############################
> +type test_bpf_deny_map_create_t;
> +domain_type(test_bpf_deny_map_create_t)
> +unconfined_runs_test(test_bpf_deny_map_create_t)
> +typeattribute test_bpf_deny_map_create_t testdomain;
> +typeattribute test_bpf_deny_map_create_t bpfdomain;
> +
> +allow test_bpf_deny_map_create_t self:process { setrlimit };
> +allow test_bpf_deny_map_create_t self:capability { sys_resource };
> +allow test_bpf_deny_map_create_t self:bpf { map_read map_write prog_load prog_run };
> +
> +############################## Deny map_read ##############################
> +type test_bpf_deny_map_read_t;
> +domain_type(test_bpf_deny_map_read_t)
> +unconfined_runs_test(test_bpf_deny_map_read_t)
> +typeattribute test_bpf_deny_map_read_t testdomain;
> +typeattribute test_bpf_deny_map_read_t bpfdomain;
> +
> +allow test_bpf_deny_map_read_t self:process { setrlimit };
> +allow test_bpf_deny_map_read_t self:capability { sys_resource };
> +allow test_bpf_deny_map_read_t self:bpf { map_create map_write prog_load prog_run };
> +
> +############################## Deny map_write ##############################
> +type test_bpf_deny_map_write_t;
> +domain_type(test_bpf_deny_map_write_t)
> +unconfined_runs_test(test_bpf_deny_map_write_t)
> +typeattribute test_bpf_deny_map_write_t testdomain;
> +typeattribute test_bpf_deny_map_write_t bpfdomain;
> +
> +allow test_bpf_deny_map_write_t self:process { setrlimit };
> +allow test_bpf_deny_map_write_t self:capability { sys_resource };
> +allow test_bpf_deny_map_write_t self:bpf { map_create map_read prog_load prog_run };
> +
> +############################## Deny prog_load ##############################
> +type test_bpf_deny_prog_load_t;
> +domain_type(test_bpf_deny_prog_load_t)
> +unconfined_runs_test(test_bpf_deny_prog_load_t)
> +typeattribute test_bpf_deny_prog_load_t testdomain;
> +typeattribute test_bpf_deny_prog_load_t bpfdomain;
> +
> +allow test_bpf_deny_prog_load_t self:process { setrlimit };
> +allow test_bpf_deny_prog_load_t self:capability { sys_resource };
> +allow test_bpf_deny_prog_load_t self:bpf { map_create map_read map_write prog_run };
> +
> +############################## Deny prog_run ###############################
> +type test_bpf_deny_prog_run_t;
> +domain_type(test_bpf_deny_prog_run_t)
> +unconfined_runs_test(test_bpf_deny_prog_run_t)
> +typeattribute test_bpf_deny_prog_run_t testdomain;
> +typeattribute test_bpf_deny_prog_run_t bpfdomain;
> +
> +allow test_bpf_deny_prog_run_t self:process { setrlimit };
> +allow test_bpf_deny_prog_run_t self:capability { sys_resource };
> +allow test_bpf_deny_prog_run_t self:bpf { map_create map_read map_write prog_load };
> +
> +#
> +############ Allow these domains to be entered from sysadm domain ############
> +#
> +miscfiles_domain_entry_test_files(bpfdomain)
> +userdom_sysadm_entry_spec_domtrans_to(bpfdomain)
> diff --git a/tests/Makefile b/tests/Makefile
> index 63aa325..80187b7 100644
> --- a/tests/Makefile
> +++ b/tests/Makefile
> @@ -42,6 +42,13 @@ ifeq ($(shell grep -q binder $(POLDEV)/include/support/all_perms.spt && test -e
>   SUBDIRS += binder
>   endif
>   
> +ifeq ($(shell grep -q bpf $(POLDEV)/include/support/all_perms.spt && echo true),true)
> +ifneq ($(shell ./kvercmp $$(uname -r) 4.4),-1)

The SELinux checks for bpf weren't introduced until Linux v4.15.

> +SUBDIRS += bpf
> +export CFLAGS += -DHAVE_BPF
> +endif
> +endif
> +
>   ifeq ($(shell grep "^SELINUX_INFINIBAND_ENDPORT_TEST=" infiniband_endport/ibendport_test.conf | cut -d'=' -f 2),1)
>   SUBDIRS += infiniband_endport
>   endif
> diff --git a/tests/bpf/.gitignore b/tests/bpf/.gitignore
> new file mode 100644
> index 0000000..1919ff8
> --- /dev/null
> +++ b/tests/bpf/.gitignore
> @@ -0,0 +1,2 @@
> +bpf_test
> +bpf_common
> diff --git a/tests/bpf/Makefile b/tests/bpf/Makefile
> new file mode 100644
> index 0000000..46817a5
> --- /dev/null
> +++ b/tests/bpf/Makefile
> @@ -0,0 +1,10 @@
> +TARGETS = bpf_test
> +DEPS = bpf_common.c bpf_common.h
> +LDLIBS += -lselinux -lbpf
> +
> +all: $(TARGETS)
> +
> +clean:
> +	rm -f $(TARGETS)
> +
> +$(TARGETS): $(DEPS)
> diff --git a/tests/bpf/bpf_common.c b/tests/bpf/bpf_common.c
> new file mode 100644
> index 0000000..738c607
> --- /dev/null
> +++ b/tests/bpf/bpf_common.c
> @@ -0,0 +1,53 @@
> +#include "bpf_common.h"
> +
> +int create_bpf_map(void)
> +{
> +	int map_fd, key;
> +	long long value = 0;
> +
> +	map_fd = bpf_create_map(BPF_MAP_TYPE_ARRAY, sizeof(key),
> +				sizeof(value), 256, 0);
> +	if (map_fd < 0) {
> +		fprintf(stderr, "Failed to create BPF map: %s\n",
> +			strerror(errno));
> +		return -1;
> +	}
> +
> +	return map_fd;
> +}
> +
> +int create_bpf_prog(void)
> +{
> +	int prog_fd;
> +	size_t insns_cnt;
> +
> +	struct bpf_insn prog[] = {
> +		BPF_MOV64_IMM(BPF_REG_0, 1),
> +		BPF_EXIT_INSN(),
> +	};
> +	insns_cnt = sizeof(prog) / sizeof(struct bpf_insn);
> +
> +	prog_fd = bpf_load_program(BPF_PROG_TYPE_SOCKET_FILTER, prog,
> +				   insns_cnt, "GPL", 0, NULL, 0);
> +
> +	if (prog_fd < 0) {
> +		fprintf(stderr, "Failed to load BPF prog: %s\n",
> +			strerror(errno));
> +		return -1;
> +	}
> +
> +	return prog_fd;
> +}
> +
> +void bpf_setrlimit(void)
> +{
> +	int result;
> +	struct rlimit r = { RLIM_INFINITY, RLIM_INFINITY };
> +
> +	result = setrlimit(RLIMIT_MEMLOCK, &r);

Is this required?  Why must it be unlimited especially for such trivial 
map/prog examples?

> +	if (result < 0) {
> +		fprintf(stderr, "Failed to set resource limit: %s\n",
> +			strerror(errno));
> +		exit(-1);
> +	}
> +}
> diff --git a/tests/bpf/bpf_common.h b/tests/bpf/bpf_common.h
> new file mode 100644
> index 0000000..44ac28f
> --- /dev/null
> +++ b/tests/bpf/bpf_common.h
> @@ -0,0 +1,34 @@
> +#include <stdio.h>
> +#include <unistd.h>
> +#include <string.h>
> +#include <stdlib.h>
> +#include <errno.h>
> +#include <stdbool.h>
> +#include <selinux/selinux.h>
> +#include <bpf/bpf.h>
> +#include <linux/bpf.h>
> +#include <sys/resource.h>
> +
> +extern int create_bpf_map(void);
> +extern int create_bpf_prog(void);
> +extern void bpf_setrlimit(void);
> +
> +/* edited eBPF instruction library */
> +/* Short form of mov, dst_reg = imm32 */
> +#define BPF_MOV64_IMM(DST, IMM)				\
> +	((struct bpf_insn) {				\
> +		.code  = BPF_ALU64 | BPF_MOV | BPF_K,	\
> +			 .dst_reg = DST,				\
> +				    .src_reg = 0,				\
> +					       .off   = 0,				\
> +							.imm   = IMM })
> +
> +/* Program exit */
> +#define BPF_EXIT_INSN()				\
> +	((struct bpf_insn) {			\
> +		.code  = BPF_JMP | BPF_EXIT,	\
> +			 .dst_reg = 0,			\
> +				    .src_reg = 0,			\
> +					       .off   = 0,			\
> +							.imm   = 0 })
> +
> diff --git a/tests/bpf/bpf_test.c b/tests/bpf/bpf_test.c
> new file mode 100644
> index 0000000..3c6a29c
> --- /dev/null
> +++ b/tests/bpf/bpf_test.c
> @@ -0,0 +1,77 @@
> +#include "bpf_common.h"
> +
> +static void usage(char *progname)
> +{
> +	fprintf(stderr,
> +		"usage:  %s -m|-p [-v]\n"
> +		"Where:\n\t"
> +		"-m    Create BPF map fd\n\t"
> +		"-p    Create BPF prog fd\n\t"
> +		"-v Print information.\n", progname);
> +	exit(-1);
> +}
> +
> +int main(int argc, char *argv[])
> +{
> +	int opt, result, fd;
> +	bool verbose = false;
> +	char *context;
> +
> +	enum {
> +		MAP_FD = 1,
> +		PROG_FD
> +	} bpf_fd_type;
> +
> +	while ((opt = getopt(argc, argv, "mpv")) != -1) {
> +		switch (opt) {
> +		case 'm':
> +			bpf_fd_type = MAP_FD;
> +			break;
> +		case 'p':
> +			bpf_fd_type = PROG_FD;
> +			break;
> +		case 'v':
> +			verbose = true;
> +			break;
> +		default:
> +			usage(argv[0]);
> +		}
> +	}
> +
> +	result = getcon(&context);
> +	if (result < 0) {
> +		fprintf(stderr, "Failed to obtain SELinux context\n");
> +		exit(-1);
> +	}
> +
> +	if (verbose)
> +		printf("Process context:\n\t%s\n", context);
> +
> +	free(context);
> +
> +	/* If BPF enabled, then need to set limits */
> +	bpf_setrlimit();
> +
> +	switch (bpf_fd_type) {
> +	case MAP_FD:
> +		if (verbose)
> +			printf("Creating BPF map\n");
> +
> +		fd = create_bpf_map();
> +		break;
> +	case PROG_FD:
> +		if (verbose)
> +			printf("Creating BPF prog\n");
> +
> +		fd = create_bpf_prog();
> +		break;
> +	default:
> +		usage(argv[0]);
> +	}
> +
> +	if (fd < 0)
> +		return fd;
> +
> +	close(fd);
> +	return 0;
> +}
> diff --git a/tests/bpf/test b/tests/bpf/test
> new file mode 100755
> index 0000000..ee00a19
> --- /dev/null
> +++ b/tests/bpf/test
> @@ -0,0 +1,64 @@
> +#!/usr/bin/perl
> +use Test::More;
> +
> +BEGIN {
> +    $basedir = $0;
> +    $basedir =~ s|(.*)/[^/]*|$1|;
> +
> +    $test_count = 7;
> +
> +    # allow info to be shown during tests
> +    $v = $ARGV[0];
> +    if ($v) {
> +        if ( $v ne "-v" ) {
> +            plan skip_all => "Invalid option (use -v)";
> +        }
> +    }
> +    else {
> +        $v = " ";
> +    }
> +
> +    plan tests => $test_count;
> +}
> +
> +#
> +# These tests are run with: kernel.unprivileged_bpf_disabled = FALSE
> +#
> +
> +#
> +################ Core BPF Tests #######################
> +#
> +# BPF map - BPF_MAP_TYPE_ARRAY
> +$result = system "runcon -t test_bpf_t $basedir/bpf_test -m $v";
> +ok( $result eq 0 );
> +
> +# BPF prog - BPF_PROG_TYPE_SOCKET_FILTER
> +$result = system "runcon -t test_bpf_t $basedir/bpf_test -p $v";
> +ok( $result eq 0 );
> +
> +# Deny map_create permission
> +$result =
> +  system "runcon -t test_bpf_deny_map_create_t $basedir/bpf_test -m $v 2>&1";
> +ok($result);
> +
> +# Deny map_read permission
> +$result =
> +  system "runcon -t test_bpf_deny_map_read_t $basedir/bpf_test -m $v 2>&1";
> +ok($result);
> +
> +# Deny map_write permission
> +$result =
> +  system "runcon -t test_bpf_deny_map_write_t $basedir/bpf_test -m $v 2>&1";
> +ok($result);
> +
> +# Deny prog_load permission
> +$result =
> +  system "runcon -t test_bpf_deny_prog_load_t $basedir/bpf_test -p $v 2>&1";
> +ok($result);
> +
> +# Deny prog_run permission
> +$result =
> +  system "runcon -t test_bpf_deny_prog_run_t $basedir/bpf_test -p $v 2>&1";
> +ok($result);
> +
> +exit;
>
Richard Haines Sept. 19, 2019, 1:36 p.m. UTC | #2
On Wed, 2019-09-18 at 10:58 -0400, Stephen Smalley wrote:
> On 9/17/19 10:56 AM, Richard Haines wrote:
> > This adds basic BPF tests for map and prog functions.
> > 
> > Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
> > ---
> >   README.md              |  4 ++-
> >   defconfig              |  5 +++
> >   policy/Makefile        |  4 +++
> >   policy/test_bpf.te     | 78
> > ++++++++++++++++++++++++++++++++++++++++++
> >   tests/Makefile         |  7 ++++
> >   tests/bpf/.gitignore   |  2 ++
> >   tests/bpf/Makefile     | 10 ++++++
> >   tests/bpf/bpf_common.c | 53 ++++++++++++++++++++++++++++
> >   tests/bpf/bpf_common.h | 34 ++++++++++++++++++
> >   tests/bpf/bpf_test.c   | 77
> > +++++++++++++++++++++++++++++++++++++++++
> >   tests/bpf/test         | 64 ++++++++++++++++++++++++++++++++++
> >   11 files changed, 337 insertions(+), 1 deletion(-)
> >   create mode 100644 policy/test_bpf.te
> >   create mode 100644 tests/bpf/.gitignore
> >   create mode 100644 tests/bpf/Makefile
> >   create mode 100644 tests/bpf/bpf_common.c
> >   create mode 100644 tests/bpf/bpf_common.h
> >   create mode 100644 tests/bpf/bpf_test.c
> >   create mode 100755 tests/bpf/test
> > 
> > diff --git a/README.md b/README.md
> > index 26784f8..1396c8e 100644
> > --- a/README.md
> > +++ b/README.md
> > @@ -51,6 +51,7 @@ similar dependencies):
> >   * iptables _(to load the `iptables SECMARK` rules during
> > `inet_socket` tests)_
> >   * lksctp-tools-devel _(to build the SCTP test programs)_
> >   * attr _(tools used by the overlayfs tests)_
> > +* libbpf-devel _(tools used by the bpf tests)_
> >   
> >   On a modern Fedora system you can install these dependencies with
> > the
> >   following command:
> > @@ -65,7 +66,8 @@ following command:
> >   		netlabel_tools \
> >   		iptables \
> >   		lksctp-tools-devel \
> > -		attr
> > +		attr \
> > +		libbpf-devel
> >   
> >   The testsuite requires a pre-existing base policy configuration
> > of SELinux,
> >   using either the old example policy or the reference policy as
> > the baseline.
> > diff --git a/defconfig b/defconfig
> > index d7f0ea5..cb57f22 100644
> > --- a/defconfig
> > +++ b/defconfig
> > @@ -62,3 +62,8 @@ CONFIG_ANDROID_BINDER_IPC=y
> >   # This will configure the Dynamically Allocated Binder Devices
> > added
> >   # to 5.0+ kernels:
> >   CONFIG_ANDROID_BINDERFS=y
> > +
> > +# Test BPF + check in selinux_file_receive and
> > selinux_binder_transfer_files.
> > +# They are not required for SELinux operation itself.
> > +CONFIG_BPF=y
> > +CONFIG_BPF_SYSCALL=y
> > diff --git a/policy/Makefile b/policy/Makefile
> > index 305b572..16a4469 100644
> > --- a/policy/Makefile
> > +++ b/policy/Makefile
> > @@ -71,6 +71,10 @@ ifeq ($(shell grep -q
> > corenet_sctp_bind_all_nodes $(POLDEV)/include/kernel/coren
> >   TARGETS += test_sctp.te
> >   endif
> >   
> > +ifeq ($(shell grep -q bpf $(POLDEV)/include/support/all_perms.spt
> > && echo true),true)
> > +TARGETS += test_bpf.te
> > +endif
> > +
> >   ifeq (x$(DISTRO),$(filter x$(DISTRO),xRHEL4 xRHEL5 xRHEL6))
> >   TARGETS:=$(filter-out test_overlayfs.te test_mqueue.te,
> > $(TARGETS))
> >   endif
> > diff --git a/policy/test_bpf.te b/policy/test_bpf.te
> > new file mode 100644
> > index 0000000..89d240c
> > --- /dev/null
> > +++ b/policy/test_bpf.te
> > @@ -0,0 +1,78 @@
> > +#
> > +################# BPF selinux-testsuite policy module
> > ######################
> > +#
> > +
> > +attribute bpfdomain;
> > +
> > +################################### Main
> > ###################################
> > +type test_bpf_t;
> > +domain_type(test_bpf_t)
> > +unconfined_runs_test(test_bpf_t)
> > +typeattribute test_bpf_t testdomain;
> > +typeattribute test_bpf_t bpfdomain;
> > +
> > +allow test_bpf_t self:process { setrlimit };
> > +#allow test_bpf_t self:capability { sys_resource sys_admin };
> > +allow test_bpf_t self:capability { sys_resource };
> > +allow test_bpf_t self:bpf { map_create map_read map_write
> > prog_load prog_run };
> > +
> > +############################## Deny map_create
> > #############################
> > +type test_bpf_deny_map_create_t;
> > +domain_type(test_bpf_deny_map_create_t)
> > +unconfined_runs_test(test_bpf_deny_map_create_t)
> > +typeattribute test_bpf_deny_map_create_t testdomain;
> > +typeattribute test_bpf_deny_map_create_t bpfdomain;
> > +
> > +allow test_bpf_deny_map_create_t self:process { setrlimit };
> > +allow test_bpf_deny_map_create_t self:capability { sys_resource };
> > +allow test_bpf_deny_map_create_t self:bpf { map_read map_write
> > prog_load prog_run };
> > +
> > +############################## Deny map_read
> > ##############################
> > +type test_bpf_deny_map_read_t;
> > +domain_type(test_bpf_deny_map_read_t)
> > +unconfined_runs_test(test_bpf_deny_map_read_t)
> > +typeattribute test_bpf_deny_map_read_t testdomain;
> > +typeattribute test_bpf_deny_map_read_t bpfdomain;
> > +
> > +allow test_bpf_deny_map_read_t self:process { setrlimit };
> > +allow test_bpf_deny_map_read_t self:capability { sys_resource };
> > +allow test_bpf_deny_map_read_t self:bpf { map_create map_write
> > prog_load prog_run };
> > +
> > +############################## Deny map_write
> > ##############################
> > +type test_bpf_deny_map_write_t;
> > +domain_type(test_bpf_deny_map_write_t)
> > +unconfined_runs_test(test_bpf_deny_map_write_t)
> > +typeattribute test_bpf_deny_map_write_t testdomain;
> > +typeattribute test_bpf_deny_map_write_t bpfdomain;
> > +
> > +allow test_bpf_deny_map_write_t self:process { setrlimit };
> > +allow test_bpf_deny_map_write_t self:capability { sys_resource };
> > +allow test_bpf_deny_map_write_t self:bpf { map_create map_read
> > prog_load prog_run };
> > +
> > +############################## Deny prog_load
> > ##############################
> > +type test_bpf_deny_prog_load_t;
> > +domain_type(test_bpf_deny_prog_load_t)
> > +unconfined_runs_test(test_bpf_deny_prog_load_t)
> > +typeattribute test_bpf_deny_prog_load_t testdomain;
> > +typeattribute test_bpf_deny_prog_load_t bpfdomain;
> > +
> > +allow test_bpf_deny_prog_load_t self:process { setrlimit };
> > +allow test_bpf_deny_prog_load_t self:capability { sys_resource };
> > +allow test_bpf_deny_prog_load_t self:bpf { map_create map_read
> > map_write prog_run };
> > +
> > +############################## Deny prog_run
> > ###############################
> > +type test_bpf_deny_prog_run_t;
> > +domain_type(test_bpf_deny_prog_run_t)
> > +unconfined_runs_test(test_bpf_deny_prog_run_t)
> > +typeattribute test_bpf_deny_prog_run_t testdomain;
> > +typeattribute test_bpf_deny_prog_run_t bpfdomain;
> > +
> > +allow test_bpf_deny_prog_run_t self:process { setrlimit };
> > +allow test_bpf_deny_prog_run_t self:capability { sys_resource };
> > +allow test_bpf_deny_prog_run_t self:bpf { map_create map_read
> > map_write prog_load };
> > +
> > +#
> > +############ Allow these domains to be entered from sysadm domain
> > ############
> > +#
> > +miscfiles_domain_entry_test_files(bpfdomain)
> > +userdom_sysadm_entry_spec_domtrans_to(bpfdomain)
> > diff --git a/tests/Makefile b/tests/Makefile
> > index 63aa325..80187b7 100644
> > --- a/tests/Makefile
> > +++ b/tests/Makefile
> > @@ -42,6 +42,13 @@ ifeq ($(shell grep -q binder
> > $(POLDEV)/include/support/all_perms.spt && test -e
> >   SUBDIRS += binder
> >   endif
> >   
> > +ifeq ($(shell grep -q bpf $(POLDEV)/include/support/all_perms.spt
> > && echo true),true)
> > +ifneq ($(shell ./kvercmp $$(uname -r) 4.4),-1)
> 
> The SELinux checks for bpf weren't introduced until Linux v4.15.

Fixed
> 
> > +SUBDIRS += bpf
> > +export CFLAGS += -DHAVE_BPF
> > +endif
> > +endif
> > +
> >   ifeq ($(shell grep "^SELINUX_INFINIBAND_ENDPORT_TEST="
> > infiniband_endport/ibendport_test.conf | cut -d'=' -f 2),1)
> >   SUBDIRS += infiniband_endport
> >   endif
> > diff --git a/tests/bpf/.gitignore b/tests/bpf/.gitignore
> > new file mode 100644
> > index 0000000..1919ff8
> > --- /dev/null
> > +++ b/tests/bpf/.gitignore
> > @@ -0,0 +1,2 @@
> > +bpf_test
> > +bpf_common
> > diff --git a/tests/bpf/Makefile b/tests/bpf/Makefile
> > new file mode 100644
> > index 0000000..46817a5
> > --- /dev/null
> > +++ b/tests/bpf/Makefile
> > @@ -0,0 +1,10 @@
> > +TARGETS = bpf_test
> > +DEPS = bpf_common.c bpf_common.h
> > +LDLIBS += -lselinux -lbpf
> > +
> > +all: $(TARGETS)
> > +
> > +clean:
> > +	rm -f $(TARGETS)
> > +
> > +$(TARGETS): $(DEPS)
> > diff --git a/tests/bpf/bpf_common.c b/tests/bpf/bpf_common.c
> > new file mode 100644
> > index 0000000..738c607
> > --- /dev/null
> > +++ b/tests/bpf/bpf_common.c
> > @@ -0,0 +1,53 @@
> > +#include "bpf_common.h"
> > +
> > +int create_bpf_map(void)
> > +{
> > +	int map_fd, key;
> > +	long long value = 0;
> > +
> > +	map_fd = bpf_create_map(BPF_MAP_TYPE_ARRAY, sizeof(key),
> > +				sizeof(value), 256, 0);
> > +	if (map_fd < 0) {
> > +		fprintf(stderr, "Failed to create BPF map: %s\n",
> > +			strerror(errno));
> > +		return -1;
> > +	}
> > +
> > +	return map_fd;
> > +}
> > +
> > +int create_bpf_prog(void)
> > +{
> > +	int prog_fd;
> > +	size_t insns_cnt;
> > +
> > +	struct bpf_insn prog[] = {
> > +		BPF_MOV64_IMM(BPF_REG_0, 1),
> > +		BPF_EXIT_INSN(),
> > +	};
> > +	insns_cnt = sizeof(prog) / sizeof(struct bpf_insn);
> > +
> > +	prog_fd = bpf_load_program(BPF_PROG_TYPE_SOCKET_FILTER, prog,
> > +				   insns_cnt, "GPL", 0, NULL, 0);
> > +
> > +	if (prog_fd < 0) {
> > +		fprintf(stderr, "Failed to load BPF prog: %s\n",
> > +			strerror(errno));
> > +		return -1;
> > +	}
> > +
> > +	return prog_fd;
> > +}
> > +
> > +void bpf_setrlimit(void)
> > +{
> > +	int result;
> > +	struct rlimit r = { RLIM_INFINITY, RLIM_INFINITY };
> > +
> > +	result = setrlimit(RLIMIT_MEMLOCK, &r);
> 
> Is this required?  Why must it be unlimited especially for such
> trivial 
> map/prog examples?

Most examples I've seen do this, however reading [1] I've now updated
and added comment:
/*
 * The default RLIMIT_MEMLOCK is normally 64K, however BPF map requires
 * more than this so double it unless RLIM_INFINITY is set.
 */

[1] https://lkml.org/lkml/2019/7/17/713

> 
> > +	if (result < 0) {
> > +		fprintf(stderr, "Failed to set resource limit: %s\n",
> > +			strerror(errno));
> > +		exit(-1);
> > +	}
> > +}
> > diff --git a/tests/bpf/bpf_common.h b/tests/bpf/bpf_common.h
> > new file mode 100644
> > index 0000000..44ac28f
> > --- /dev/null
> > +++ b/tests/bpf/bpf_common.h
> > @@ -0,0 +1,34 @@
> > +#include <stdio.h>
> > +#include <unistd.h>
> > +#include <string.h>
> > +#include <stdlib.h>
> > +#include <errno.h>
> > +#include <stdbool.h>
> > +#include <selinux/selinux.h>
> > +#include <bpf/bpf.h>
> > +#include <linux/bpf.h>
> > +#include <sys/resource.h>
> > +
> > +extern int create_bpf_map(void);
> > +extern int create_bpf_prog(void);
> > +extern void bpf_setrlimit(void);
> > +
> > +/* edited eBPF instruction library */
> > +/* Short form of mov, dst_reg = imm32 */
> > +#define BPF_MOV64_IMM(DST, IMM)				\
> > +	((struct bpf_insn) {				\
> > +		.code  = BPF_ALU64 | BPF_MOV | BPF_K,	\
> > +			 .dst_reg = DST,				\
> > +				    .src_reg = 0,				
> > \
> > +					       .off   = 0,			
> > 	\
> > +							.imm   = IMM })
> > +
> > +/* Program exit */
> > +#define BPF_EXIT_INSN()				\
> > +	((struct bpf_insn) {			\
> > +		.code  = BPF_JMP | BPF_EXIT,	\
> > +			 .dst_reg = 0,			\
> > +				    .src_reg = 0,			\
> > +					       .off   = 0,			
> > \
> > +							.imm   = 0 })
> > +
> > diff --git a/tests/bpf/bpf_test.c b/tests/bpf/bpf_test.c
> > new file mode 100644
> > index 0000000..3c6a29c
> > --- /dev/null
> > +++ b/tests/bpf/bpf_test.c
> > @@ -0,0 +1,77 @@
> > +#include "bpf_common.h"
> > +
> > +static void usage(char *progname)
> > +{
> > +	fprintf(stderr,
> > +		"usage:  %s -m|-p [-v]\n"
> > +		"Where:\n\t"
> > +		"-m    Create BPF map fd\n\t"
> > +		"-p    Create BPF prog fd\n\t"
> > +		"-v Print information.\n", progname);
> > +	exit(-1);
> > +}
> > +
> > +int main(int argc, char *argv[])
> > +{
> > +	int opt, result, fd;
> > +	bool verbose = false;
> > +	char *context;
> > +
> > +	enum {
> > +		MAP_FD = 1,
> > +		PROG_FD
> > +	} bpf_fd_type;
> > +
> > +	while ((opt = getopt(argc, argv, "mpv")) != -1) {
> > +		switch (opt) {
> > +		case 'm':
> > +			bpf_fd_type = MAP_FD;
> > +			break;
> > +		case 'p':
> > +			bpf_fd_type = PROG_FD;
> > +			break;
> > +		case 'v':
> > +			verbose = true;
> > +			break;
> > +		default:
> > +			usage(argv[0]);
> > +		}
> > +	}
> > +
> > +	result = getcon(&context);
> > +	if (result < 0) {
> > +		fprintf(stderr, "Failed to obtain SELinux context\n");
> > +		exit(-1);
> > +	}
> > +
> > +	if (verbose)
> > +		printf("Process context:\n\t%s\n", context);
> > +
> > +	free(context);
> > +
> > +	/* If BPF enabled, then need to set limits */
> > +	bpf_setrlimit();
> > +
> > +	switch (bpf_fd_type) {
> > +	case MAP_FD:
> > +		if (verbose)
> > +			printf("Creating BPF map\n");
> > +
> > +		fd = create_bpf_map();
> > +		break;
> > +	case PROG_FD:
> > +		if (verbose)
> > +			printf("Creating BPF prog\n");
> > +
> > +		fd = create_bpf_prog();
> > +		break;
> > +	default:
> > +		usage(argv[0]);
> > +	}
> > +
> > +	if (fd < 0)
> > +		return fd;
> > +
> > +	close(fd);
> > +	return 0;
> > +}
> > diff --git a/tests/bpf/test b/tests/bpf/test
> > new file mode 100755
> > index 0000000..ee00a19
> > --- /dev/null
> > +++ b/tests/bpf/test
> > @@ -0,0 +1,64 @@
> > +#!/usr/bin/perl
> > +use Test::More;
> > +
> > +BEGIN {
> > +    $basedir = $0;
> > +    $basedir =~ s|(.*)/[^/]*|$1|;
> > +
> > +    $test_count = 7;
> > +
> > +    # allow info to be shown during tests
> > +    $v = $ARGV[0];
> > +    if ($v) {
> > +        if ( $v ne "-v" ) {
> > +            plan skip_all => "Invalid option (use -v)";
> > +        }
> > +    }
> > +    else {
> > +        $v = " ";
> > +    }
> > +
> > +    plan tests => $test_count;
> > +}
> > +
> > +#
> > +# These tests are run with: kernel.unprivileged_bpf_disabled =
> > FALSE
> > +#
> > +
> > +#
> > +################ Core BPF Tests #######################
> > +#
> > +# BPF map - BPF_MAP_TYPE_ARRAY
> > +$result = system "runcon -t test_bpf_t $basedir/bpf_test -m $v";
> > +ok( $result eq 0 );
> > +
> > +# BPF prog - BPF_PROG_TYPE_SOCKET_FILTER
> > +$result = system "runcon -t test_bpf_t $basedir/bpf_test -p $v";
> > +ok( $result eq 0 );
> > +
> > +# Deny map_create permission
> > +$result =
> > +  system "runcon -t test_bpf_deny_map_create_t $basedir/bpf_test
> > -m $v 2>&1";
> > +ok($result);
> > +
> > +# Deny map_read permission
> > +$result =
> > +  system "runcon -t test_bpf_deny_map_read_t $basedir/bpf_test -m
> > $v 2>&1";
> > +ok($result);
> > +
> > +# Deny map_write permission
> > +$result =
> > +  system "runcon -t test_bpf_deny_map_write_t $basedir/bpf_test -m
> > $v 2>&1";
> > +ok($result);
> > +
> > +# Deny prog_load permission
> > +$result =
> > +  system "runcon -t test_bpf_deny_prog_load_t $basedir/bpf_test -p
> > $v 2>&1";
> > +ok($result);
> > +
> > +# Deny prog_run permission
> > +$result =
> > +  system "runcon -t test_bpf_deny_prog_run_t $basedir/bpf_test -p
> > $v 2>&1";
> > +ok($result);
> > +
> > +exit;
> >
diff mbox series

Patch

diff --git a/README.md b/README.md
index 26784f8..1396c8e 100644
--- a/README.md
+++ b/README.md
@@ -51,6 +51,7 @@  similar dependencies):
 * iptables _(to load the `iptables SECMARK` rules during `inet_socket` tests)_
 * lksctp-tools-devel _(to build the SCTP test programs)_
 * attr _(tools used by the overlayfs tests)_
+* libbpf-devel _(tools used by the bpf tests)_
 
 On a modern Fedora system you can install these dependencies with the
 following command:
@@ -65,7 +66,8 @@  following command:
 		netlabel_tools \
 		iptables \
 		lksctp-tools-devel \
-		attr
+		attr \
+		libbpf-devel
 
 The testsuite requires a pre-existing base policy configuration of SELinux,
 using either the old example policy or the reference policy as the baseline.
diff --git a/defconfig b/defconfig
index d7f0ea5..cb57f22 100644
--- a/defconfig
+++ b/defconfig
@@ -62,3 +62,8 @@  CONFIG_ANDROID_BINDER_IPC=y
 # This will configure the Dynamically Allocated Binder Devices added
 # to 5.0+ kernels:
 CONFIG_ANDROID_BINDERFS=y
+
+# Test BPF + check in selinux_file_receive and selinux_binder_transfer_files.
+# They are not required for SELinux operation itself.
+CONFIG_BPF=y
+CONFIG_BPF_SYSCALL=y
diff --git a/policy/Makefile b/policy/Makefile
index 305b572..16a4469 100644
--- a/policy/Makefile
+++ b/policy/Makefile
@@ -71,6 +71,10 @@  ifeq ($(shell grep -q corenet_sctp_bind_all_nodes $(POLDEV)/include/kernel/coren
 TARGETS += test_sctp.te
 endif
 
+ifeq ($(shell grep -q bpf $(POLDEV)/include/support/all_perms.spt && echo true),true)
+TARGETS += test_bpf.te
+endif
+
 ifeq (x$(DISTRO),$(filter x$(DISTRO),xRHEL4 xRHEL5 xRHEL6))
 TARGETS:=$(filter-out test_overlayfs.te test_mqueue.te, $(TARGETS))
 endif
diff --git a/policy/test_bpf.te b/policy/test_bpf.te
new file mode 100644
index 0000000..89d240c
--- /dev/null
+++ b/policy/test_bpf.te
@@ -0,0 +1,78 @@ 
+#
+################# BPF selinux-testsuite policy module ######################
+#
+
+attribute bpfdomain;
+
+################################### Main ###################################
+type test_bpf_t;
+domain_type(test_bpf_t)
+unconfined_runs_test(test_bpf_t)
+typeattribute test_bpf_t testdomain;
+typeattribute test_bpf_t bpfdomain;
+
+allow test_bpf_t self:process { setrlimit };
+#allow test_bpf_t self:capability { sys_resource sys_admin };
+allow test_bpf_t self:capability { sys_resource };
+allow test_bpf_t self:bpf { map_create map_read map_write prog_load prog_run };
+
+############################## Deny map_create #############################
+type test_bpf_deny_map_create_t;
+domain_type(test_bpf_deny_map_create_t)
+unconfined_runs_test(test_bpf_deny_map_create_t)
+typeattribute test_bpf_deny_map_create_t testdomain;
+typeattribute test_bpf_deny_map_create_t bpfdomain;
+
+allow test_bpf_deny_map_create_t self:process { setrlimit };
+allow test_bpf_deny_map_create_t self:capability { sys_resource };
+allow test_bpf_deny_map_create_t self:bpf { map_read map_write prog_load prog_run };
+
+############################## Deny map_read ##############################
+type test_bpf_deny_map_read_t;
+domain_type(test_bpf_deny_map_read_t)
+unconfined_runs_test(test_bpf_deny_map_read_t)
+typeattribute test_bpf_deny_map_read_t testdomain;
+typeattribute test_bpf_deny_map_read_t bpfdomain;
+
+allow test_bpf_deny_map_read_t self:process { setrlimit };
+allow test_bpf_deny_map_read_t self:capability { sys_resource };
+allow test_bpf_deny_map_read_t self:bpf { map_create map_write prog_load prog_run };
+
+############################## Deny map_write ##############################
+type test_bpf_deny_map_write_t;
+domain_type(test_bpf_deny_map_write_t)
+unconfined_runs_test(test_bpf_deny_map_write_t)
+typeattribute test_bpf_deny_map_write_t testdomain;
+typeattribute test_bpf_deny_map_write_t bpfdomain;
+
+allow test_bpf_deny_map_write_t self:process { setrlimit };
+allow test_bpf_deny_map_write_t self:capability { sys_resource };
+allow test_bpf_deny_map_write_t self:bpf { map_create map_read prog_load prog_run };
+
+############################## Deny prog_load ##############################
+type test_bpf_deny_prog_load_t;
+domain_type(test_bpf_deny_prog_load_t)
+unconfined_runs_test(test_bpf_deny_prog_load_t)
+typeattribute test_bpf_deny_prog_load_t testdomain;
+typeattribute test_bpf_deny_prog_load_t bpfdomain;
+
+allow test_bpf_deny_prog_load_t self:process { setrlimit };
+allow test_bpf_deny_prog_load_t self:capability { sys_resource };
+allow test_bpf_deny_prog_load_t self:bpf { map_create map_read map_write prog_run };
+
+############################## Deny prog_run ###############################
+type test_bpf_deny_prog_run_t;
+domain_type(test_bpf_deny_prog_run_t)
+unconfined_runs_test(test_bpf_deny_prog_run_t)
+typeattribute test_bpf_deny_prog_run_t testdomain;
+typeattribute test_bpf_deny_prog_run_t bpfdomain;
+
+allow test_bpf_deny_prog_run_t self:process { setrlimit };
+allow test_bpf_deny_prog_run_t self:capability { sys_resource };
+allow test_bpf_deny_prog_run_t self:bpf { map_create map_read map_write prog_load };
+
+#
+############ Allow these domains to be entered from sysadm domain ############
+#
+miscfiles_domain_entry_test_files(bpfdomain)
+userdom_sysadm_entry_spec_domtrans_to(bpfdomain)
diff --git a/tests/Makefile b/tests/Makefile
index 63aa325..80187b7 100644
--- a/tests/Makefile
+++ b/tests/Makefile
@@ -42,6 +42,13 @@  ifeq ($(shell grep -q binder $(POLDEV)/include/support/all_perms.spt && test -e
 SUBDIRS += binder
 endif
 
+ifeq ($(shell grep -q bpf $(POLDEV)/include/support/all_perms.spt && echo true),true)
+ifneq ($(shell ./kvercmp $$(uname -r) 4.4),-1)
+SUBDIRS += bpf
+export CFLAGS += -DHAVE_BPF
+endif
+endif
+
 ifeq ($(shell grep "^SELINUX_INFINIBAND_ENDPORT_TEST=" infiniband_endport/ibendport_test.conf | cut -d'=' -f 2),1)
 SUBDIRS += infiniband_endport
 endif
diff --git a/tests/bpf/.gitignore b/tests/bpf/.gitignore
new file mode 100644
index 0000000..1919ff8
--- /dev/null
+++ b/tests/bpf/.gitignore
@@ -0,0 +1,2 @@ 
+bpf_test
+bpf_common
diff --git a/tests/bpf/Makefile b/tests/bpf/Makefile
new file mode 100644
index 0000000..46817a5
--- /dev/null
+++ b/tests/bpf/Makefile
@@ -0,0 +1,10 @@ 
+TARGETS = bpf_test
+DEPS = bpf_common.c bpf_common.h
+LDLIBS += -lselinux -lbpf
+
+all: $(TARGETS)
+
+clean:
+	rm -f $(TARGETS)
+
+$(TARGETS): $(DEPS)
diff --git a/tests/bpf/bpf_common.c b/tests/bpf/bpf_common.c
new file mode 100644
index 0000000..738c607
--- /dev/null
+++ b/tests/bpf/bpf_common.c
@@ -0,0 +1,53 @@ 
+#include "bpf_common.h"
+
+int create_bpf_map(void)
+{
+	int map_fd, key;
+	long long value = 0;
+
+	map_fd = bpf_create_map(BPF_MAP_TYPE_ARRAY, sizeof(key),
+				sizeof(value), 256, 0);
+	if (map_fd < 0) {
+		fprintf(stderr, "Failed to create BPF map: %s\n",
+			strerror(errno));
+		return -1;
+	}
+
+	return map_fd;
+}
+
+int create_bpf_prog(void)
+{
+	int prog_fd;
+	size_t insns_cnt;
+
+	struct bpf_insn prog[] = {
+		BPF_MOV64_IMM(BPF_REG_0, 1),
+		BPF_EXIT_INSN(),
+	};
+	insns_cnt = sizeof(prog) / sizeof(struct bpf_insn);
+
+	prog_fd = bpf_load_program(BPF_PROG_TYPE_SOCKET_FILTER, prog,
+				   insns_cnt, "GPL", 0, NULL, 0);
+
+	if (prog_fd < 0) {
+		fprintf(stderr, "Failed to load BPF prog: %s\n",
+			strerror(errno));
+		return -1;
+	}
+
+	return prog_fd;
+}
+
+void bpf_setrlimit(void)
+{
+	int result;
+	struct rlimit r = { RLIM_INFINITY, RLIM_INFINITY };
+
+	result = setrlimit(RLIMIT_MEMLOCK, &r);
+	if (result < 0) {
+		fprintf(stderr, "Failed to set resource limit: %s\n",
+			strerror(errno));
+		exit(-1);
+	}
+}
diff --git a/tests/bpf/bpf_common.h b/tests/bpf/bpf_common.h
new file mode 100644
index 0000000..44ac28f
--- /dev/null
+++ b/tests/bpf/bpf_common.h
@@ -0,0 +1,34 @@ 
+#include <stdio.h>
+#include <unistd.h>
+#include <string.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <stdbool.h>
+#include <selinux/selinux.h>
+#include <bpf/bpf.h>
+#include <linux/bpf.h>
+#include <sys/resource.h>
+
+extern int create_bpf_map(void);
+extern int create_bpf_prog(void);
+extern void bpf_setrlimit(void);
+
+/* edited eBPF instruction library */
+/* Short form of mov, dst_reg = imm32 */
+#define BPF_MOV64_IMM(DST, IMM)				\
+	((struct bpf_insn) {				\
+		.code  = BPF_ALU64 | BPF_MOV | BPF_K,	\
+			 .dst_reg = DST,				\
+				    .src_reg = 0,				\
+					       .off   = 0,				\
+							.imm   = IMM })
+
+/* Program exit */
+#define BPF_EXIT_INSN()				\
+	((struct bpf_insn) {			\
+		.code  = BPF_JMP | BPF_EXIT,	\
+			 .dst_reg = 0,			\
+				    .src_reg = 0,			\
+					       .off   = 0,			\
+							.imm   = 0 })
+
diff --git a/tests/bpf/bpf_test.c b/tests/bpf/bpf_test.c
new file mode 100644
index 0000000..3c6a29c
--- /dev/null
+++ b/tests/bpf/bpf_test.c
@@ -0,0 +1,77 @@ 
+#include "bpf_common.h"
+
+static void usage(char *progname)
+{
+	fprintf(stderr,
+		"usage:  %s -m|-p [-v]\n"
+		"Where:\n\t"
+		"-m    Create BPF map fd\n\t"
+		"-p    Create BPF prog fd\n\t"
+		"-v Print information.\n", progname);
+	exit(-1);
+}
+
+int main(int argc, char *argv[])
+{
+	int opt, result, fd;
+	bool verbose = false;
+	char *context;
+
+	enum {
+		MAP_FD = 1,
+		PROG_FD
+	} bpf_fd_type;
+
+	while ((opt = getopt(argc, argv, "mpv")) != -1) {
+		switch (opt) {
+		case 'm':
+			bpf_fd_type = MAP_FD;
+			break;
+		case 'p':
+			bpf_fd_type = PROG_FD;
+			break;
+		case 'v':
+			verbose = true;
+			break;
+		default:
+			usage(argv[0]);
+		}
+	}
+
+	result = getcon(&context);
+	if (result < 0) {
+		fprintf(stderr, "Failed to obtain SELinux context\n");
+		exit(-1);
+	}
+
+	if (verbose)
+		printf("Process context:\n\t%s\n", context);
+
+	free(context);
+
+	/* If BPF enabled, then need to set limits */
+	bpf_setrlimit();
+
+	switch (bpf_fd_type) {
+	case MAP_FD:
+		if (verbose)
+			printf("Creating BPF map\n");
+
+		fd = create_bpf_map();
+		break;
+	case PROG_FD:
+		if (verbose)
+			printf("Creating BPF prog\n");
+
+		fd = create_bpf_prog();
+		break;
+	default:
+		usage(argv[0]);
+	}
+
+	if (fd < 0)
+		return fd;
+
+	close(fd);
+	return 0;
+}
diff --git a/tests/bpf/test b/tests/bpf/test
new file mode 100755
index 0000000..ee00a19
--- /dev/null
+++ b/tests/bpf/test
@@ -0,0 +1,64 @@ 
+#!/usr/bin/perl
+use Test::More;
+
+BEGIN {
+    $basedir = $0;
+    $basedir =~ s|(.*)/[^/]*|$1|;
+
+    $test_count = 7;
+
+    # allow info to be shown during tests
+    $v = $ARGV[0];
+    if ($v) {
+        if ( $v ne "-v" ) {
+            plan skip_all => "Invalid option (use -v)";
+        }
+    }
+    else {
+        $v = " ";
+    }
+
+    plan tests => $test_count;
+}
+
+#
+# These tests are run with: kernel.unprivileged_bpf_disabled = FALSE
+#
+
+#
+################ Core BPF Tests #######################
+#
+# BPF map - BPF_MAP_TYPE_ARRAY
+$result = system "runcon -t test_bpf_t $basedir/bpf_test -m $v";
+ok( $result eq 0 );
+
+# BPF prog - BPF_PROG_TYPE_SOCKET_FILTER
+$result = system "runcon -t test_bpf_t $basedir/bpf_test -p $v";
+ok( $result eq 0 );
+
+# Deny map_create permission
+$result =
+  system "runcon -t test_bpf_deny_map_create_t $basedir/bpf_test -m $v 2>&1";
+ok($result);
+
+# Deny map_read permission
+$result =
+  system "runcon -t test_bpf_deny_map_read_t $basedir/bpf_test -m $v 2>&1";
+ok($result);
+
+# Deny map_write permission
+$result =
+  system "runcon -t test_bpf_deny_map_write_t $basedir/bpf_test -m $v 2>&1";
+ok($result);
+
+# Deny prog_load permission
+$result =
+  system "runcon -t test_bpf_deny_prog_load_t $basedir/bpf_test -p $v 2>&1";
+ok($result);
+
+# Deny prog_run permission
+$result =
+  system "runcon -t test_bpf_deny_prog_run_t $basedir/bpf_test -p $v 2>&1";
+ok($result);
+
+exit;