[apparmor] [PATCH] tests: Conditionalize when to build/run stacking tests

[Steve Beattie]

On Sat, Mar 19, 2016 at 02:35:07AM -0500, Tyler Hicks wrote:
> On 2016-03-19 00:30:01, Steve Beattie wrote:
> > On Fri, Mar 18, 2016 at 06:05:47PM -0500, Tyler Hicks wrote:
> > > The stacking test binary links against libapparmor for
> > > aa_stack_profile() and aa_stack_onexec(), which will be present in 2.11.
> > > This means that regression test builds using the system libapparmor
> > > should not build the stacking test binary unless the libapparmor 2.11 or
> > > newer is present.
> > > 
> > > Signed-off-by: Tyler Hicks <tyhicks at canonical.com>
> > 
> > > ---
> > >  tests/regression/apparmor/Makefile | 21 ++++++++++++++-------
> > >  1 file changed, 14 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/tests/regression/apparmor/Makefile b/tests/regression/apparmor/Makefile
> > > index 21c7fb3..c61f141 100644
> > > --- a/tests/regression/apparmor/Makefile
> > > +++ b/tests/regression/apparmor/Makefile
> > > @@ -119,7 +119,6 @@ SRC=access.c \
> > >      readdir.c \
> > >      rw.c \
> > >      socketpair.c \
> > > -    stacking.c \
> > >      symlink.c \
> > >      syscall_mknod.c \
> > >      swap.c \
> > > @@ -160,16 +159,26 @@ endif
> > >  ifdef USE_SYSTEM
> > >    ifneq (,$(shell pkg-config --atleast-version 2.10 libapparmor && echo TRUE))
> > >      SRC+=aa_policy_cache.c
> > > -    AA_POLICY_CACHE_TEST=aa_policy_cache
> > > +    CONDITIONAL_TESTS+=aa_policy_cache
> > >    else
> > >      $(warning ${nl}\
> > >      ************************************************************************${nl}\
> > >      Skipping aa_policy_cache tests: requires libapparmor 2.10 or newer ...${nl}\
> > >      ************************************************************************${nl})
> > >    endif
> > > +
> > > +  ifneq (,$(shell pkg-config --atleast-version 2.11 libapparmor && echo TRUE))
> > > +    SRC+=stacking.c
> > > +    CONDITIONAL_TESTS+=stackonexec stackprofile
> > > +  else
> > > +    $(warning ${nl}\
> > > +    ************************************************************************${nl}\
> > > +    Skipping stacking tests: requires libapparmor 2.11 or newer ...${nl}\
> > > +    ************************************************************************${nl})
> > > +  endif
> > >  else
> > > -  SRC+=aa_policy_cache.c
> > > -  AA_POLICY_CACHE_TEST=aa_policy_cache
> > > +  SRC+=aa_policy_cache.c stacking.c
> > > +  CONDITIONAL_TESTS+=aa_policy_cache stackonexec stackprofile
> > >  endif
> > >  
> > >  EXEC=$(SRC:%.c=%)
> > > @@ -219,8 +228,6 @@ TESTS=aa_exec \
> > >        swap \
> > >        sd_flags \
> > >        setattr \
> > > -      stackonexec \
> > > -      stackprofile \
> > >        symlink \
> > >        syscall \
> > >        tcp \
> > > @@ -237,7 +244,7 @@ ifneq (,$(shell pkg-config --exists dbus-1 && echo TRUE))
> > >  TESTS+=dbus_eavesdrop dbus_message dbus_service dbus_unrequested_reply
> > >  endif
> > >  
> > > -TESTS+=$(AA_POLICY_CACHE_TEST)
> > > +TESTS+=$(CONDITIONAL_TESTS)
> > >  
> > >  # Tests that can crash the kernel should be placed here
> > >  RISKY_TESTS=
> > 
> > The added exec_stack tests should also not be used if stacking.c cannot
> > be built. It should be removed from the default set and added to the
> > $(CONDITIONAL_TESTS). With that change made,
> > Acked-by: Steve Beattie <steve at nxnw.org>
> 
> The reason I didn't do that was because exec_stack.sh doesn't depend on
> any programs that link against libapparmor for aa_stack_*().

Hrm? It uses the stacking binary. That it doesn't use the aa_stack_*()
function references within it doesn't really matter, if stacking.c
can't compile successfully. (Unless you're planning some conditional
compilation juju that #ifdefs those references away if the library
is too old, that is.)

> The top of the exec_stack.sh file contains:
> 
>   requires_kernel_features domain/stack
> 
> So the entire test is skipped if stacking support isn't present in the
> kernel.

Consider a kernel that supports stacking but the system libapparmor
does not, The stacking binary fails to build, but the exec_stack.sh
script still tries to run it.

> Are you fine with this or should I still move it into CONDITIONAL_TESTS?

I think it should still go in CONDITIONAL_TESTS.

That said, one of the tests should verify the interactions between
aa_stack, aa_stack_onexec, and x -> &stack.

-- 
Steve Beattie
<sbeattie at ubuntu.com>
http://NxNW.org/~steve/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <https://lists.ubuntu.com/archives/apparmor/attachments/20160319/00749bc2/attachment.pgp>