Re: Linux 5.13-rc6 regression to 5.12.x: kernel OOM and panic during kernel boot in low memory Xen VM's (256MB assigned memory).

From: Sander Eikelenboom
Date: Mon Jun 21 2021 - 17:36:57 EST


On 21/06/2021 18:54, Rasmus Villemoes wrote:
On 18/06/2021 03.06, Sander Eikelenboom wrote:
On 17/06/2021 21:39, Sander Eikelenboom wrote:


OK, done some experimentation and it seems with 256M assigned to the VM
it was almost at the edge of OOM with the 5.12 kernel as well in the
config I am using it.
With v5.12 when I assign 240M it boots, with 230M it doesn't. With 5.13
the tipping point seems to be around 265M and 270M, so my config was
already quite close to the edge.

The "direct kernel boot" feature I'm using just seems somewhat memory
hungry, but using another compression algorithm for the kernel and
initramfs already helped in my case.

So sorry for the noise, clearly user-error.

Hm, perhaps, but I'm still a bit nervous about that report from Oliver
Sang/kernel test robot, which was for a VM equipped with 16G of memory.
But despite quite a few attempts, I haven't been able to reproduce that
locally, so unfortunately I have no idea what's going on.

Rasmus


Hmm I just tried to switch all VM's to a 5.13-rc7 kernel.
Some worked since i reduced the size, but some still fail.

The difference seems the be the number of vcpu's I assign to the VM's

The ones with 1 vcpu now boot with 256MB assigned (that was what I tested before),
but the ones with 2 vcpu's assigned don't and still OOM
on the same kernel and initramfs that I pass in from the host.

Could that box from the test-robot have a massive amount of cpu-cores
and that it is some how related to that ?

--
Sander