Currently we check kernel taint flags after a failed test and reboot system if any are set.
This is broken for several reasons:
there is no guarantee that taint flags are set only after failing test
certain taint flags are harmless e.g. "System was live patched"
it may take a while between kernel internal data structures are corrupted and the time the bug has been exposed
So all in all we should check for taint flags after each test and attempt reboot the system if any of the serious flags are set I guess that we want to reboot at least on R, M, B, D, W, L flags.
Also we should check the kernel taint flags before the testrun and the runltp-ng should exit if any of the serious flags are set prior the start of the testsuite.
Another part of the problem are tests that may trigger taint flags "module was force unloaded" may be triggered by rmmod tests, so we probably need .sets_taint_flags in tst_test structure in testcases which would get parsed into the test metadata and the testrunner will make use of that information, e.g. schedule such test to be executed as a last one.
Currently we check kernel taint flags after a failed test and reboot system if any are set.
This is broken for several reasons:
So all in all we should check for taint flags after each test and attempt reboot the system if any of the serious flags are set I guess that we want to reboot at least on R, M, B, D, W, L flags.
Also we should check the kernel taint flags before the testrun and the runltp-ng should exit if any of the serious flags are set prior the start of the testsuite.
Another part of the problem are tests that may trigger taint flags "module was force unloaded" may be triggered by rmmod tests, so we probably need .sets_taint_flags in tst_test structure in testcases which would get parsed into the test metadata and the testrunner will make use of that information, e.g. schedule such test to be executed as a last one.