Re: [PATCH v2] selftests: livepatch: Test atomic replace against multiple modules

[Marcos Paulo de Souza]

On Mon, 2024-06-03 at 14:52 +0200, Petr Mladek wrote:
> On Fri 2024-05-31 18:06:48, Marcos Paulo de Souza wrote:
> > On Fri, 2024-05-31 at 15:44 -0400, Joe Lawrence wrote:
> > > On Sat, May 25, 2024 at 11:34:08AM -0300, Marcos Paulo de Souza
> > > wrote:
> > > > Adapt the current test-livepatch.sh script to account the
> > > > number of
> > > > applied livepatches and ensure that an atomic replace livepatch
> > > > disables
> > > > all previously applied livepatches.
> > > > 
> > > > Signed-off-by: Marcos Paulo de Souza <mpdesouza@xxxxxxxx>
> > > > ---
> > > > Changes since v1:
> > > > * Added checks in the existing test-livepatch.sh instead of
> > > > creating a
> > > >   new test file. (Joe)
> > > > * Fixed issues reported by ShellCheck (Joe)
> > > > ---
> > > >  .../testing/selftests/livepatch/test-livepatch.sh  | 46
> > > > ++++++++++++++++++++--
> > > >  1 file changed, 42 insertions(+), 4 deletions(-)
> > > > 
> > > > diff --git a/tools/testing/selftests/livepatch/test-
> > > > livepatch.sh
> > > > b/tools/testing/selftests/livepatch/test-livepatch.sh
> > > > index e3455a6b1158..d85405d18e54 100755
> > > > --- a/tools/testing/selftests/livepatch/test-livepatch.sh
> > > > +++ b/tools/testing/selftests/livepatch/test-livepatch.sh
> > > > @@ -107,9 +107,12 @@ livepatch: '$MOD_LIVEPATCH': unpatching
> > > > complete
> > > >  
> > > >  # - load a livepatch that modifies the output from
> > > > /proc/cmdline
> > > > and
> > > >  #   verify correct behavior
> > > > -# - load an atomic replace livepatch and verify that only the
> > > > second is active
> > > > -# - remove the first livepatch and verify that the atomic
> > > > replace
> > > > livepatch
> > > > -#   is still active
> > > > +# - load two addtional livepatches and check the number of
> > > > livepatch modules
> > > > +#   applied
> > > > +# - load an atomic replace livepatch and check that the other
> > > > three modules were
> > > > +#   disabled
> > > > +# - remove all livepatches besides the atomic replace one and
> > > > verify that the
> > > > +#   atomic replace livepatch is still active
> > > >  # - remove the atomic replace livepatch and verify that none
> > > > are
> > > > active
> > > >  
> > > >  start_test "atomic replace livepatch"
> > > > @@ -119,12 +122,31 @@ load_lp $MOD_LIVEPATCH
> > > >  grep 'live patched' /proc/cmdline > /dev/kmsg
> > > >  grep 'live patched' /proc/meminfo > /dev/kmsg
> > > >  
> > > > +for mod in test_klp_syscall test_klp_callbacks_demo; do
> > > 
> > > Slightly nitpicky here, but the tests were originally written
> > > with
> > > the
> > > livepatch module names via variables like $MOD_LIVEPATCH.  Would
> > > using
> > > $MOD_LIVEPATCH{1,2,3} help indicate that their specifics aren't
> > > really
> > > interesting, that we just need 3 of them?
> > 
> > Makes sense. I thought about it when I was changing the code, but I
> > didn't want to change it too much, so it was the result. But that
> > makes
> > sense to have the modules better named.
> 
> I like this.
> 
> > > > +	load_lp $mod
> > > > +done
> > > > +
> > > > +mods=(/sys/kernel/livepatch/*)
> > > > +nmods=${#mods[@]}
> > > > +if [ "$nmods" -ne 3 ]; then
> > > > +	die "Expecting three modules listed, found $nmods"
> > > > +fi
> > > > +
> > > 
> > > I was going to suggest that we might protect against a situation
> > > where
> > > other livepatch modules were active, that a simple count wouldn't
> > > be
> > > sufficient.  But then I thought about this test, atomic replace!
> > > Anything previously loaded is going to be pushed aside anyway.
> > > 
> > > So maybe (in another patch or set) it would be worth enhancing
> > > functions.sh :: start_test() do a quick sanity check to see that
> > > the
> > > initial conditions are safe?  That might also prevent some
> > > collateral
> > > damage when test A fails and leaves the world a strange place for
> > > tests
> > > B, C, etc.
> > 
> > We have been discussing about start/end functions that would check
> > for
> > leftover modules... maybe should be a good think to implement soon
> > as
> > we land more tests.
> 
> Makes sense :-)
> 
> > > >  load_lp $MOD_REPLACE replace=1
> > > >  
> > > >  grep 'live patched' /proc/cmdline > /dev/kmsg
> > > >  grep 'live patched' /proc/meminfo > /dev/kmsg
> > > >  
> > > > -unload_lp $MOD_LIVEPATCH
> > > > +mods=(/sys/kernel/livepatch/*)
> > > > +nmods=${#mods[@]}
> > > > +if [ "$nmods" -ne 1 ]; then
> > > > +	die "Expecting only one moduled listed, found $nmods"
> > > > +fi
> > > > +
> > > > +# These modules were disabled by the atomic replace
> > > > +for mod in test_klp_callbacks_demo test_klp_syscall
> > > > $MOD_LIVEPATCH; do
> > > > +	unload_lp "$mod"
> > > > +done
> > > >  
> > > >  grep 'live patched' /proc/cmdline > /dev/kmsg
> > > >  grep 'live patched' /proc/meminfo > /dev/kmsg
> > > > @@ -142,6 +164,20 @@ livepatch: '$MOD_LIVEPATCH': starting
> > > > patching
> > > > transition
> > > >  livepatch: '$MOD_LIVEPATCH': completing patching transition
> > > >  livepatch: '$MOD_LIVEPATCH': patching complete
> > > >  $MOD_LIVEPATCH: this has been live patched
> > > > +% insmod test_modules/test_klp_syscall.ko
> > > 
> > > Similar minor nit here, too.  If we think copy/pasting all the
> > > $MOD_FOO
> > > is annoying, I am fine with leaving this as is.  I don't have a
> > > strong
> > > opinion other than following some convention.
> > > 
> > > With that, I'm happy to ack as-is or with variable names.
> > 
> > Thanks Joe! I think that is Petr's call, either way I can rework
> > this
> > patch, or send additional ones to adjust the tests.
> 
> I would prefer if you did respin this patch. The use of
> $MOD_LIVEPATCH{1,2,3} would make even the patch easier to follow.

Done in v3. About the pre-check, I discussed with Miroslav about having
an easier way to skip tests. The idea was to split each "test" into a
different file, like fstests already does. Using this approach, each
start_test function will be placed in a different file to test
specifically one functionality. This way we can skip a test if we don't
have some requirements (like a sysfs attribute for example, or the
there were leftover modules).

I plan to send a patch starting this move when the v3 of this patchset
is accepted.

> 
> Best Regards,
> Petr