Booted up a new 4.4.21 kernel from my previous 4.4.6 kernel build, exact same kconfig. Reproducible: Always Actual Results: Samba stopped, then sshd stopped... Upon accessing the console line for the server and running dmesg rendered the OOM killer was killing all processes on the box... Expected Results: My atom machine shouldn't have died. Reverting to 4.4.6 didn't reproduce the issue, showing it's a bad kernel and not user-space... Please mask 4.4.21.
I'm lucky because I have console access to the box, this kernel is pretty dangerous to anyone loading the presumably problematic kmodule. Attachment is https://bugs.gentoo.org/attachment.cgi?id=449778
It seems some things are not quite right with this version. My issue is not related to yours, but it is also solved by reverting back to 4.4.6. Like you, the .config file for 4.4.21 is exactly the same as for 4.4.6. Nothing else has changed except the gentoo-sources version: https://bugs.gentoo.org/show_bug.cgi?id=596880
Why did you CC amd64?
(In reply to Tomáš Mózes from comment #3) > Why did you CC amd64? Because this impacts atleast amd64, and is up to amd64 to mask for that architecture? Is that not what Arches are for? This is fixed in sys-kernel/gentoo-sources-4.4.26, clearly an issue with the release... .21 is masked for other reasons, so this is fixed.
(In reply to Kyle Sanderson from comment #4) > (In reply to Tomáš Mózes from comment #3) > > Why did you CC amd64? > > Because this impacts atleast amd64, and is up to amd64 to mask for that > architecture? Is that not what Arches are for? Mostly stabilization ;) The kernel team will take care of that, please do not CC arches.
(In reply to Tomáš Mózes from comment #5) > (In reply to Kyle Sanderson from comment #4) > > (In reply to Tomáš Mózes from comment #3) > > > Why did you CC amd64? > > > > Because this impacts atleast amd64, and is up to amd64 to mask for that > > architecture? Is that not what Arches are for? > > Mostly stabilization ;) The kernel team will take care of that, please do > not CC arches. Right, but that's exactly what this bug is about. This buggy broken kernel was stabilised when it had serious issues. The bug was asking for a hard mask.
(In reply to Kyle Sanderson from comment #6) > (In reply to Tomáš Mózes from comment #5) > > (In reply to Kyle Sanderson from comment #4) > > > (In reply to Tomáš Mózes from comment #3) > > > > Why did you CC amd64? > > > > > > Because this impacts atleast amd64, and is up to amd64 to mask for that > > > architecture? Is that not what Arches are for? > > > > Mostly stabilization ;) The kernel team will take care of that, please do > > not CC arches. > > Right, but that's exactly what this bug is about. This buggy broken kernel > was stabilised when it had serious issues. The bug was asking for a hard > mask. Ok, but that decision is made by the kernel team, that's all. Great you reported this, but it's up to the kernel team to take further steps.
