[j/n][linux-aws][PATCH 0/2 v2] aws: Backport linear memory map change

[Philip Cox]

BugLink: https://bugs.launchpad.net/bugs/2069352

SRU Justification:

[Impact]

AWS discovered an issue of long EC2 instance launch times on the upcoming R8g instance types. 

[Fix]

They have identified the following fixes that address this issue:

* https://patchew.org/linux/20240412131908.433043-1-ryan.roberts@arm.com/20240412131908.433043-2-ryan.roberts@arm.com/
* https://patchew.org/linux/20240412131908.433043-1-ryan.roberts@arm.com/20240412131908.433043-3-ryan.roberts@arm.com/

They would like this fixed in the jammy, and noble kernels. 


For reference, here are the links to the above patches in 6.10-rc3:

* arm64: mm: Don't remap pgtables per-cont(pte|pmd) block
https://github.com/torvalds/linux/commit/5c63db59c5f89925add57642be4f789d0d671ccd
* arm64: mm: Batch dsb and isb when populating pgtables
https://github.com/torvalds/linux/commit/1fcb7cea8a5f7747e02230f816c2c80b060d9517


AWS has requested only the first two patches of the series of three patches be back ported.

[Test Plan]
I have tested this, as has AWS.

[Where problems could occur]
With the reordering of tlb invalidation and memory barriers, there is some risk, but this chance for regressions on this change is fairly small as the scope is fairly narrow.

[Other info]
SF# 00387444


Changes from v1:

Noticed patches tagged to mantic rather than noble.  Fixed subject titles.