Re: [PATCH] IMA: Add test for selinux measurement

[Petr Vorel <pvorel@xxxxxxx>]

> On 2/23/21 10:00 AM, Petr Vorel wrote:


> > > +++ b/testcases/kernel/security/integrity/ima/tests/ima_selinux.sh
> > ...
> > > +validate_policy_capabilities()
> > > +{
> > > +	local measured_cap measured_value expected_value
> > > +	local result=1
> > > +	local inx=7
> > > +
> > > +	# Policy capabilities flags start from "network_peer_controls"
> > > +	# in the measured SELinux state at offset 7 for 'awk'
> > > +	while [ $inx -lt 20 ]; do
> > > +		measured_cap=$(echo $1 | awk -F'[=;]' -v inx="$inx" '{print $inx}')
> > > +		inx=$(( $inx + 1 ))
> > > +
> > > +		measured_value=$(echo $1 | awk -F'[=;]' -v inx="$inx" '{print $inx}')
> > > +		expected_value=$(cat "$SELINUX_DIR/policy_capabilities/$measured_cap")
> > > +		if [ "$measured_value" != "$expected_value" ];then
> > > +			tst_res TWARN "$measured_cap: expected: $expected_value, got: $digest"
> > We rarely use TWARN in the tests, only when the error is not related to the test result.
> > Otherwise we use TFAIL.
> ok - I will change it to TFAIL.
Thanks!
But I've noticed that this error is handled twice, first in validate_policy_capabilities()
as TWARN (or TFAIL) and then in test2(). Let's use TPASS/TFAIL in
validate_policy_capabilities():

validate_policy_capabilities()
{
	local measured_cap measured_value expected_value
	local inx=7

	# Policy capabilities flags start from "network_peer_controls"
	# in the measured SELinux state at offset 7 for 'awk'
	while [ $inx -lt 20 ]; do
		measured_cap=$(echo $1 | awk -F'[=;]' -v inx="$inx" '{print $inx}')
		inx=$(($inx + 1))

		measured_value=$(echo $1 | awk -F'[=;]' -v inx="$inx" '{print $inx}')
		expected_value=$(cat "$SELINUX_DIR/policy_capabilities/$measured_cap")
		if [ "$measured_value" != "$expected_value" ]; then
			tst_res TFAIL "$measured_cap: expected: $expected_value, got: $digest"
			return
		fi

		inx=$(($inx + 1))
	done

	tst_res TPASS "SELinux state measured correctly"
}

test2()
{
	...
	validate_policy_capabilities $measured_data
}

...
> > As we discuss, I'm going tom merge test when patchset is merged in maintainers tree,
> > please ping me. And ideally we should mention kernel commit hash as a comment in
> > the test.
> Will do. Thank you.
Thanks!

...
> > +++ testcases/kernel/security/integrity/ima/tests/ima_selinux.sh
> > @@ -13,16 +13,14 @@ TST_SETUP="setup"
> >   . ima_setup.sh
> >   FUNC_CRITICAL_DATA='func=CRITICAL_DATA'
> > -REQUIRED_POLICY="^measure.*($FUNC_CRITICAL_DATA)"
> > +REQUIRED_POLICY="^measure.*$FUNC_CRITICAL_DATA"
> >   setup()
> >   {
> > -	SELINUX_DIR=$(tst_get_selinux_dir)
> > -	if [ -z "$SELINUX_DIR" ]; then
> > -		tst_brk TCONF "SELinux is not enabled"
> > -		return
> > -	fi
> > +	tst_require_selinux_enabled
> Please correct me if I have misunderstood this one:

> tst_require_selinux_enabled is checking if SELinux is enabled in "enforce"
> mode. Would this check fail if SELinux is enabled in "permissive" mode?

> For running the test, we just need SELinux to be enabled. I verify that by
> checking for the presence of SELINUX_DIR.

Good catch. Your original version is correct (put it back into ima/selinux.v2.fixes).
I didn't put a helper for it, because you need $SELINUX_DIR anyway.
Thus removed tst_require_selinux_enabled() as not needed.

I renamed tst_selinux_enabled() to tst_selinux_enforced() to make the purpose clearer
(commit 82b598ea1 IMA: Add test for selinux measurement).

I've updated branch ima/selinux.v2.fixes with all mentioned changes
https://github.com/pevik/ltp/commits/ima/selinux.v2.fixes

Kind regards,
Petr

> thanks,
>  -lakshmi