I ran sel4test (12.1.0 kernel, latest sel4test) on an Armv8-A (aarch64) machine with SMP and Hyp mode on, and all tests passed except for CACHEFLUSH0004 [sel4test/cache.c at master · seL4/sel4test · GitHub]. The value of *ptr2
after the frame was deleted and remapped does not return 0, but C0FFEE still (as before the frame was deleted).
Should there be a Page Clean in place before *ptr2
is accessed? (or maybe it’s platform-dependent, though in CI this test passed for all platforms present)
Adding a page clean makes the test pass in my case.
This doesn’t make sense since the memory reserved for the frame is set to “non-cacheable”. Cache maintenance operations shouldn’t be needed.