On Mon, Mar 23, 2020 at 04:51:11PM +0100, Michal Hocko wrote:
On Mon 23-03-20 11:41:59, Rafael Aquini wrote:
On Mon, Mar 23, 2020 at 04:12:56PM +0100, Michal Hocko wrote:
On Mon 23-03-20 11:02:59, Rafael Aquini wrote:
[...]
The selftest also checks the kernel visible effect, via /proc/kpageflags, and that's where it fails after 9c4e6b1a7027f.
I really fail to see your point. Even if you are right that the self test is somehow evaluating the kernel implementation which I am not sure is the scope of the selft thest but anyway. The mere fact that the kernel test fails on a perfectly valid change should just suggest that the test is leading to false positives and therefore should be fixed. Your proposed fix is simply suboptimal because it relies on yet another side effect which might change anytime in the future and still lead to a correctly behaving kernel. See my point?
OK, I concede your point on the bogusness of checking the page flags in this particular test and expect certain valuse there, given that no other selftest seems to be doing that level of inner kenrel detail scrutiny.
I'll repost this fix suggestion getting rif of those related checkpoints.
Here is what I have after I had to context switch to something else before finishing it. Feel free to reuse if you feel like. It is likely to not even compile.
I'm OK with it, if you want to go ahead and do the kill.
Thanks -- Rafael