[PATCH] UBUNTU: SAUCE: dove: avoid page table overwrite when resuming from hibernation

[Tim Gardner]

On 05/27/2010 09:03 AM, Eric Miao wrote:
>  From 1ae0436f8c2159936ca42d9ef938055b68af64f7 Mon Sep 17 00:00:00 2001
> From: Eric Miao<eric.miao at canonical.com>
> Date: Fri, 23 Apr 2010 14:16:17 +0800
> Subject: [PATCH] UBUNTU: SAUCE: dove: avoid page table overwrite when
> resuming from hibernation
>
> BugLink: http://bugs.launchpad.net/bugs/509006
>
> Resuming from hibernation is OK if 'resume=/dev/sdaX' is explicitly
> specified on the kernel command line, but it fails if scripts in
> initramfs are used to trigger the resume. It turned out to be page
> table being overwritten when restoring the memory content because
> it's using a normal user process's page table in the latter case,
> which is not safe and could be overwritten. Fix this by using the
> safe swapper_pg_dir during restoring.
>
> Signed-off-by: Eric Miao<eric.miao at canonical.com>
> ---
>
> It was posted weeks ago, thought it was merged but not. This is an updated
> version addressing the previous comments to make it cleaner.
>
>   arch/arm/mach-dove/Makefile |    2 ++
>   arch/arm/mach-dove/swsusp.S |    8 ++++++++
>   2 files changed, 10 insertions(+), 0 deletions(-)
>
> diff --git a/arch/arm/mach-dove/Makefile b/arch/arm/mach-dove/Makefile
> index 0be1e1c..c5c028f 100755
> --- a/arch/arm/mach-dove/Makefile
> +++ b/arch/arm/mach-dove/Makefile
> @@ -1,3 +1,5 @@
> +AFLAGS_swsusp.o			:= -DTEXT_OFFSET=$(TEXT_OFFSET)
> +
>   obj-y				+= clock.o common.o addr-map.o irq.o pcie.o mpp.o \
>   				sdhci_cam_mbus.o
>   obj-$(CONFIG_MACH_DOVE_RD_AVNG)	+= dove-rd-avng-setup.o
> diff --git a/arch/arm/mach-dove/swsusp.S b/arch/arm/mach-dove/swsusp.S
> index 4f4a884..9b752a0 100644
> --- a/arch/arm/mach-dove/swsusp.S
> +++ b/arch/arm/mach-dove/swsusp.S
> @@ -28,6 +28,7 @@
>    */
>
>   #include<linux/linkage.h>
> +#include<asm/memory.h>
>   #include<asm/segment.h>
>   #include<asm/page.h>
>   #include<asm/asm-offsets.h>
> @@ -209,8 +210,15 @@ FUNC(swsusp_arch_suspend)
>
>   FUNC_END(swsusp_arch_suspend)
>
> +#define KERNEL_RAM_PADDR	(PHYS_OFFSET + TEXT_OFFSET)
> +#define SWAPPER_PG_DIR		(KERNEL_RAM_PADDR - 0x4000)
> +
>   FUNC(swsusp_arch_resume)
>   	/* set page table if needed */
> +	ldr	r0, =SWAPPER_PG_DIR
> +	mcr	p15, 0, r0, c2, c0, 0		@ load page table pointer
> +	mcr	p15, 0, r0, c8, c7, 0		@ invalidate I,D TLBs
> +	mcr	p15, 0, r0, c7, c5, 4		@ ISB
>
>   	/*
>   	 * retore "nr_copy_pages" pages which are saved and specified

Test results in the LP report look good. Has this patch been upstreamed? 
I cannot remember from the original discussions a few weeks ago.

Acked-by: Tim Gardner <tim.gardner at canonical.com>

-- 
Tim Gardner tim.gardner at canonical.com