There are cases where a strict upper limit on a memcg is required, for
example, when containers are rented out and interference between them
can not be tolerated.

Provide memory.max, a limit that can not be breached and will trigger
group-internal OOM killing once page reclaim can no longer enforce it.

This can be combined with the high limit, to create a window in which
allocating tasks are throttled to approach the strict maximum limit
gracefully and with opportunity for the user or admin to intervene.

Signed-off-by: Johannes Weiner <han...@cmpxchg.org>
---
 Documentation/cgroups/unified-hierarchy.txt |  4 ++++
 mm/memcontrol.c                             | 35 +++++++++++++++++++++++++++++
 2 files changed, 39 insertions(+)

diff --git a/Documentation/cgroups/unified-hierarchy.txt 
b/Documentation/cgroups/unified-hierarchy.txt
index fd4f7f6847f6..6c52c926810f 100644
--- a/Documentation/cgroups/unified-hierarchy.txt
+++ b/Documentation/cgroups/unified-hierarchy.txt
@@ -334,6 +334,10 @@ supported and the interface files "release_agent" and
 - memory.usage_in_bytes is renamed to memory.current to be in line
   with the new naming scheme
 
+- memory.max provides a hard upper limit as a last-resort backup to
+  memory.high for situations with aggressive isolation requirements.
+
+
 5. Planned Changes
 
 5-1. CAP for resource control
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 5a64fa96c08a..461834c86b94 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -6306,6 +6306,36 @@ static ssize_t memory_high_write(struct kernfs_open_file 
*of,
        return nbytes;
 }
 
+static u64 memory_max_read(struct cgroup_subsys_state *css,
+                          struct cftype *cft)
+{
+       struct mem_cgroup *memcg = mem_cgroup_from_css(css);
+
+       return res_counter_read_u64(&memcg->res, RES_LIMIT);
+}
+
+static ssize_t memory_max_write(struct kernfs_open_file *of,
+                               char *buf, size_t nbytes, loff_t off)
+{
+       struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
+       u64 max;
+       int ret;
+
+       if (mem_cgroup_is_root(memcg))
+               return -EINVAL;
+
+       buf = strim(buf);
+       ret = res_counter_memparse_write_strategy(buf, &max);
+       if (ret)
+               return ret;
+
+       ret = mem_cgroup_resize_limit(memcg, max);
+       if (ret)
+               return ret;
+
+       return nbytes;
+}
+
 static struct cftype memory_files[] = {
        {
                .name = "current",
@@ -6316,6 +6346,11 @@ static struct cftype memory_files[] = {
                .read_u64 = memory_high_read,
                .write = memory_high_write,
        },
+       {
+               .name = "max",
+               .read_u64 = memory_max_read,
+               .write = memory_max_write,
+       },
 };
 
 struct cgroup_subsys memory_cgrp_subsys = {
-- 
2.0.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Reply via email to