We've been exercising the rtm images that include the latest workaround now for a considerable amount of time repeating the test that originally triggered the issue.
On mako, 2 devices running:
device #1, 329 iterations - no corruption observed
device #2, 251 iterations - no corruption observed
And also on another spare device using rtm image,
device #3, 176 iterations, - no corruption observed
I can continue running these for another few days if need be, but I think it is fairly conclusive that the data=journal change prevents the apparmor metadata from ending up on the /var/lib/apparmor/profiles/* files.
We've been exercising the rtm images that include the latest workaround now for a considerable amount of time repeating the test that originally triggered the issue.
On mako, 2 devices running:
device #1, 329 iterations - no corruption observed
device #2, 251 iterations - no corruption observed
And also on another spare device using rtm image,
device #3, 176 iterations, - no corruption observed
I can continue running these for another few days if need be, but I think it is fairly conclusive that the data=journal change prevents the apparmor metadata from ending up on the /var/lib/ apparmor/ profiles/ * files.