[Acked] [Utopic workaround] Temporarily reduce KASLR range

Andy Whitcroft apw at canonical.com
Tue Aug 26 14:55:57 UTC 2014


On Tue, Aug 26, 2014 at 01:00:22PM +0200, Stefan Bader wrote:
> From 2cd5f1b7735c5efafcb77eaf49d0d6dbf0fea9bf Mon Sep 17 00:00:00 2001
> From: Stefan Bader <stefan.bader at canonical.com>
> Date: Tue, 26 Aug 2014 12:49:15 +0200
> Subject: [PATCH] UBUNTU: SAUCE: (config) Temporarily reduce KASLR range
> 
> BugLink: http:/bugs.launchpad.net/bugs/1350522
> 
> This regression was indirectly introduced by enabling Kernel
> Address Space Location Randomization. On 64bit this changes
> the split between kernel and modules which in turn seems to
> mess up paravirt Xen guests (and dom0 is PV, too).
> 
> The boot visible effect is vmalloc failures which can even cause
> crashes (dom0 rather crashes while domU may survive in some undefined
> state).
> 
> I am still trying to figure out what exactly goes wrong. But for
> the mean-time it would allow Utopic dom0 and domU guests to come
> up again if we changed the kernel/module split back to 512M/~1.5G
> (as it was before).
> 
> This will make the randomization less effective (since the kernel
> then can only move within the 512M range) but at least those kernels
> can be used in EC2 (and set up Xen hosts based on Utopic).
> 
> Signed-off-by: Stefan Bader <stefan.bader at canonical.com>
> ---
>  debian.master/config/amd64/config.common.amd64 | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/debian.master/config/amd64/config.common.amd64 b/debian.master/config/amd64/config.common.amd64
> index 9e82d9b..753124c 100644
> --- a/debian.master/config/amd64/config.common.amd64
> +++ b/debian.master/config/amd64/config.common.amd64
> @@ -92,7 +92,7 @@ CONFIG_PROBE_EVENTS=y
>  # CONFIG_PSTORE_CONSOLE is not set
>  CONFIG_PSTORE_RAM=m
>  CONFIG_PTP_1588_CLOCK=m
> -CONFIG_RANDOMIZE_BASE_MAX_OFFSET=0x40000000
> +CONFIG_RANDOMIZE_BASE_MAX_OFFSET=0x20000000
>  CONFIG_RCU_CPU_STALL_TIMEOUT=60
>  CONFIG_RCU_FANOUT=64
>  CONFIG_RCU_FAST_NO_HZ=y

On the assumption we are looking to understand the underlying issue
longer term: 

Acked-by: Andy Whitcroft <apw at canonical.com>

-apw




More information about the kernel-team mailing list