CONFIG_CGROUP_MEM_RES_CTLR=y hang on Nexus 4
Jani Monoses
jani at ubuntu.com
Mon Jun 3 10:44:10 UTC 2013
Hello,
If we want to set memory limits for LXC containers we need memcg turned
on, but enabling it in the Nexus 4 kernel causes a hang on boot and a
stacktrace can be produced [1]. The explanation I came up with is in the
second posting to the cgroup mailing list [2] . Since neither email got
an answer from the cgroup developers I hope someone here can offer an
opinion.
A workaround consists in enabling FLATMEM instead of SPARSEMEM
using the patch below. Can this have any ill-effects apart from
apparently using up 1M extra for cgroup accounting overhead?
commit 6422531ae224d73b55e8fa095d1c7405651eabac
Author: Jani Monoses <jani at ubuntu.com>
Date: Mon Apr 29 16:31:27 2013 +0300
Use FLATMEM to allow booting with memory cgroups on
The memory cgroups setup when sparsemem is enabled is not
working well with the layout of the mako hardware memory blocks
and causes a page fault when booting. This works around the bug
which may be either in the cgroups or mako hw setup code.
Change-Id: I3bf34bc5e2cff25e613595bc8bff06c81b711198
Signed-off-by: Jani Monoses <jani at ubuntu.com>
diff --git a/arch/arm/mach-msm/Kconfig b/arch/arm/mach-msm/Kconfig
index d982156..ae95fdc 100644
--- a/arch/arm/mach-msm/Kconfig
+++ b/arch/arm/mach-msm/Kconfig
@@ -163,9 +163,7 @@ config ARCH_MSM8960
select MSM_SPM_V2
select MSM_L2_SPM
select MSM_NATIVE_RESTART
- select DONT_MAP_HOLE_AFTER_MEMBANK0
select MSM_REMOTE_SPINLOCK_SFPB
- select ARCH_SPARSEMEM_ENABLE
select ARCH_HAS_HOLES_MEMORYMODEL
select CLEANCACHE
select QCACHE
@@ -199,9 +197,7 @@ config ARCH_MSM8930
select MSM_SPM_V2
select MSM_L2_SPM
select MSM_NATIVE_RESTART
- select DONT_MAP_HOLE_AFTER_MEMBANK0
select MSM_REMOTE_SPINLOCK_SFPB
- select ARCH_SPARSEMEM_ENABLE
select ARCH_HAS_HOLES_MEMORYMODEL
select MSM_ULTRASOUND
select MULTI_IRQ_HANDLER
[1] http://permalink.gmane.org/gmane.linux.kernel.cgroups/6590
[2] http://permalink.gmane.org/gmane.linux.kernel.cgroups/7003
More information about the kernel-team
mailing list