(2012/06/26 13:57), David Rientjes wrote:

On Mon, 25 Jun 2012, Glauber Costa wrote:

quoted

diff --git a/include/linux/thread_info.h b/include/linux/thread_info.h
index ccc1899..914ec07 100644
--- a/include/linux/thread_info.h
+++ b/include/linux/thread_info.h

@@ -61,6 +61,12 @@ extern long do_no_restart_syscall(struct restart_block *parm);
  # define THREADINFO_GFP		(GFP_KERNEL | __GFP_NOTRACK)
  #endif

+#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
+# define THREADINFO_GFP_ACCOUNTED (THREADINFO_GFP | __GFP_KMEMCG)
+#else
+# define THREADINFO_GFP_ACCOUNTED (THREADINFO_GFP)
+#endif
+

This type of requirement is going to become nasty very quickly if nobody
can use __GFP_KMEMCG without testing for CONFIG_CGROUP_MEM_RES_CTLR_KMEM.
Perhaps define __GFP_KMEMCG to be 0x0 if it's not enabled, similar to how
kmemcheck does?

I agree.

Thanks,
-Kame