On Tue, 13 Mar 2012 14:33:16 -0700, Andrew Morton [off-list ref] wrote:

On Tue, 13 Mar 2012 12:37:06 +0530
"Aneesh Kumar K.V" [off-list ref] wrote:

quoted

+static int mem_cgroup_hugetlb_usage(struct mem_cgroup *memcg)
+{
+	int idx;
+	for (idx = 0; idx < hugetlb_max_hstate; idx++) {
+		if (memcg->hugepage[idx].usage > 0)
+			return memcg->hugepage[idx].usage;
+	}
+	return 0;
+}

Please document the function?  Had you done this, I might have been
able to work out why the function bales out on the first used hugepage
size, but I can't :(

I guess the function is named wrongly. I will rename it to
mem_cgroup_have_hugetlb_usage() in the next iteration ? The function
will return (bool) 1 if it has any hugetlb resource usage.

This could have used for_each_hstate(), had that macro been better
designed (or updated).

Can you explain this ?. for_each_hstate allows to iterate over
different hstates. But here we need to look at different hugepage
rescounter in memcg. I can still use for_each_hstate() and find the
hstate index (h - hstates) and use that to index memcg rescounter
array. But that would make it more complex ?

Upon return this function coerces an unsigned long long into an "int". 
We decided last week that more than 2^32 hugepages was not
inconceivable, so I guess that's a bug.

-aneesh