Linux 5.12's very bad, double ungood day

Posted Mar 9, 2021 8:21 UTC (Tue) by matthias (subscriber, #94967)
In reply to: Linux 5.12's very bad, double ungood day by tome
Parent article: Linux 5.12's very bad, double ungood day

If the test does not verify that the data is written to the correct location in the swapfile, the test will probably pass. The data can be read back fine. Even if you hit some used location this will manifest as random filesystem corruption. You only notice this if you verify the filesystem afterwards, but why should a test of the swapping code verify the correctness of a totally unrelated filesystem?

Linux 5.12's very bad, double ungood day

Posted Mar 9, 2021 8:44 UTC (Tue) by marcH (subscriber, #57642) [Link]

Tests may pass, but some time later the machine will behave erratically which will raise suspicion. Pretty much what happened but too far away from the origin of the patch and too late.

So back to the original question: which "swaptorture" test suite was used to validate this patch and did that test suite include configuration(s) with swap file(s)? The main article wonders "how things could have been done better" and asks some interesting code review questions but does not seem to mention anything like "test gap".