-_-

Test that failed:

Assertion 'personality(PER_LINUX | ADDR_NO_RANDOMIZE) == -1 && errno == EPERM' failed at ../src/test/test-seccomp.c:590, function test_lock_personality(). Aborting.
lockpersonalityseccomp terminated by signal ABRT.
Assertion 'wait_for_terminate_and_warn("lockpersonalityseccomp", pid, true) == EXIT_SUCCESS' failed at ../src/test/test-seccomp.c:629, function test_lock_personality(). Aborting.
FAIL: test-seccomp (code: 134)
Aborted (core dumped)

Someone mind helping me understand how it core dumped? I really don't think adding to hwdb would cause this.. but still pretty interesting.

My tests are attached:
testlog.txt

Yes, it's unrelated and due to #6585.

The problem is with i386 build. Earlier in the logs we have
current personality=0
That is, opinionated_personality() returned PER_LINUX. The logic in opinionated_personality() assumes that a 32 bit personality uses PER_LINUX32 (or PER_LINUX) but the build machine seems to be instead using PER_LINUX_32BIT (aka PER_LINUX | ADDR_LIMIT_32BIT) and opinionated_personality() removes ADDR_LIMIT_32BIT.

But the build problem could be instead that the build machine is not a real i386 but x86_64 with i386 chroot or something and so the x86_64 kernel could be running in different personality mode than real i386 kernel would (maybe PER_LINUX or PER_LINUX32), so opinionated_personality() would be correct. @poettering @martinpitt ?

@topimiettinen yeah, something's fishy there, and it affects various PRs, not just this one. Let's open a new issue about this, and discuss things there...

I figure we need to submit some PR that just shows the actual personality set, and then look at what the CI prints for that, and then adjust opinionated_personality() accordingly...

cc @whot

(and please ignore the seccomp/lockpersnality CI issue)

@tommydrum let's not use the full dmi string please, we usually stick to the minimum required to identify the product reliably. in your case that shoud be
evdev:name:Synaptics TM2438-005:dmi:*svnRazer:pnBladeStealth*

Please give that a try and amend the PR as required, thanks.

Updated with that change :)

thanks!