On Fri, Dec 05, 2014 at 05:27:02PM +0000, Russell King - ARM Linux wrote:

On Fri, Dec 05, 2014 at 05:07:45PM +0000, Catalin Marinas wrote:

quoted

From 8e317c6be00abe280de4dcdd598d2e92009174b6 Mon Sep 17 00:00:00 2001
From: Catalin Marinas <catalin.marinas@arm.com>
Date: Fri, 5 Dec 2014 16:41:52 +0000
Subject: [PATCH] Revert "ARM: 8167/1: extend the reserved memory for initrd to
 be page aligned"

This reverts commit 421520ba98290a73b35b7644e877a48f18e06004. There is
no guarantee that the boot-loader places other images like dtb in a
different page than initrd start/end. When this happens, such pages must
not be freed. The free_reserved_area() already takes care of rounding up
"start" and rounding down "end" to avoid freeing partially used pages.

In addition to the revert, this patch also removes the arm32
PAGE_ALIGN(end) when calculating the size of the memory to be poisoned.

which makes the summary line rather misleading, and I really don't think
we need to do this on ARM for the simple reason that we've been doing it
for soo long that it can't be an issue.

I started this as a revert and then realised that it doesn't solve
anything for arm32 without changing the poisoning.

Anyway, if you are happy with how it is, I'll drop the arm32 part. As I
said yesterday, the issue is worse for arm64 with 64K pages.

-- 
Catalin