Re: x86: e820 regression

On Mon, Dec 10, 2018 at 08:54:21AM +0000, Naoya Horiguchi wrote:

Hi Erick,

On Mon, Dec 10, 2018 at 03:28:37AM -0500, Erick Cafferata wrote:

...

The following commit introduced a regression on my system.

124049decbb121ec32742c94fb5d9d6bed8f24d8 x86/e820: put !E820_TYPE_RAM regions into memblock.reserved

and it was backported to stable, stopping the kernel to boot on my system since around 4.17.4. It was reverted on upstream a couple months ago. commit 2a5bda5a624d6471d25e953b9adba5182ab1b51f upstream

This commit seems not a correct pointer. In mainline, commit 124049decbb was reverted by

commit 9fd61bc95130d4971568b89c9548b5e0a4e18e0e
Author: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Date:   Fri Oct 26 15:10:24 2018 -0700

Revert "x86/e820: put !E820_TYPE_RAM regions into memblock.reserved"

and, the original problem was finally fixed by

commit 907ec5fca3dc38d37737de826f06f25b063aa08e
Author: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Date:   Fri Oct 26 15:10:15 2018 -0700

mm: zero remaining unavailable struct pages Patch series "mm: Fix for movable_node boot option", v3.

so I think both patches should be backported onto v4.17.z.

4.17.y and 4.18.y are long end-of-life, there's nothing I can do there.

I can apply the above patches to the 4.19.y tree, is that sufficient?

thanks,

greg k-h