The patch titled
Subject: mm: hugetlb controller for cgroups
has been added to the -mm tree. Its filename is
mm-hugetlb-controller-for-cgroups-v2.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mm-hugetlb-controller-for-cgroups-v2.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mm-hugetlb-controller-for-cgroups-v2.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Giuseppe Scrivano <gscrivan@xxxxxxxxxx>
Subject: mm: hugetlb controller for cgroups
In the effort of supporting cgroups v2 into Kubernetes, I stumped on the
lack of the hugetlb controller.
When the controller is enabled, it exposes four new files for each hugetlb
size on non-root cgroups:
- hugetlb.<hugepagesize>.current
- hugetlb.<hugepagesize>.max
- hugetlb.<hugepagesize>.events
- hugetlb.<hugepagesize>.events.local
The differences with the legacy hierarchy are in the file names and using
the value "max" instead of "-1" to disable a limit.
The file .limit_in_bytes is renamed to .max.
The file .usage_in_bytes is renamed to .current.
.failcnt is not provided as a single file anymore, but its value can be
read through the new flat-keyed files .events and .events.local, through
the "max" key.
Link: http://lkml.kernel.org/r/20191216193831.540953-1-gscrivan@xxxxxxxxxx
Signed-off-by: Giuseppe Scrivano <gscrivan@xxxxxxxxxx>
Acked-by: Tejun Heo <tj@xxxxxxxxxx>
Cc: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
Cc: Michal Koutn <mkoutny@xxxxxxxx>
Cc: Li Zefan <lizefan@xxxxxxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Cc: Mina Almasry <almasrymina@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
Documentation/admin-guide/cgroup-v2.rst | 29 +++
include/linux/hugetlb.h | 3
mm/hugetlb_cgroup.c | 198 ++++++++++++++++++++--
3 files changed, 218 insertions(+), 12 deletions(-)
--- a/Documentation/admin-guide/cgroup-v2.rst~mm-hugetlb-controller-for-cgroups-v2
+++ a/Documentation/admin-guide/cgroup-v2.rst
@@ -61,6 +61,8 @@ v1 is available under Documentation/admi
5-6. Device
5-7. RDMA
5-7-1. RDMA Interface Files
+ 5-8. HugeTLB
+ 5.8-1. HugeTLB Interface Files
5-8. Misc
5-8-1. perf_event
5-N. Non-normative information
@@ -2056,6 +2058,33 @@ RDMA Interface Files
mlx4_0 hca_handle=1 hca_object=20
ocrdma1 hca_handle=1 hca_object=23
+HugeTLB
+-------
+
+The HugeTLB controller allows to limit the HugeTLB usage per control group and
+enforces the controller limit during page fault.
+
+HugeTLB Interface Files
+~~~~~~~~~~~~~~~~~~~~~~~
+
+ hugetlb.<hugepagesize>.current
+ Show current usage for "hugepagesize" hugetlb. It exists for all
+ the cgroup except root.
+
+ hugetlb.<hugepagesize>.max
+ Set/show the hard limit of "hugepagesize" hugetlb usage.
+ The default value is "max". It exists for all the cgroup except root.
+
+ hugetlb.<hugepagesize>.events
+ A read-only flat-keyed file which exists on non-root cgroups.
+
+ max
+ The number of allocation failure due to HugeTLB limit
+
+ hugetlb.<hugepagesize>.events.local
+ Similar to hugetlb.<hugepagesize>.events but the fields in the file
+ are local to the cgroup i.e. not hierarchical. The file modified event
+ generated on this file reflects only the local events.
Misc
----
--- a/include/linux/hugetlb.h~mm-hugetlb-controller-for-cgroups-v2
+++ a/include/linux/hugetlb.h
@@ -432,7 +432,8 @@ struct hstate {
unsigned int surplus_huge_pages_node[MAX_NUMNODES];
#ifdef CONFIG_CGROUP_HUGETLB
/* cgroup control files */
- struct cftype cgroup_files[5];
+ struct cftype cgroup_files_dfl[5];
+ struct cftype cgroup_files_legacy[5];
#endif
char name[HSTATE_NAME_LEN];
};
--- a/mm/hugetlb_cgroup.c~mm-hugetlb-controller-for-cgroups-v2
+++ a/mm/hugetlb_cgroup.c
@@ -3,6 +3,10 @@
* Copyright IBM Corporation, 2012
* Author Aneesh Kumar K.V <aneesh.kumar@xxxxxxxxxxxxxxxxxx>
*
+ * Cgroup v2
+ * Copyright (C) 2019 Red Hat, Inc.
+ * Author: Giuseppe Scrivano <gscrivan@xxxxxxxxxx>
+ *
* This program is free software; you can redistribute it and/or modify it
* under the terms of version 2.1 of the GNU Lesser General Public License
* as published by the Free Software Foundation.
@@ -19,18 +23,36 @@
#include <linux/hugetlb.h>
#include <linux/hugetlb_cgroup.h>
+enum hugetlb_memory_event {
+ HUGETLB_MAX,
+ HUGETLB_NR_MEMORY_EVENTS,
+};
+
struct hugetlb_cgroup {
struct cgroup_subsys_state css;
+
/*
* the counter to account for hugepages from hugetlb.
*/
struct page_counter hugepage[HUGE_MAX_HSTATE];
+
+ atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
+ atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
+
+ /* Handle for "hugetlb.events" */
+ struct cgroup_file events_file[HUGE_MAX_HSTATE];
+
+ /* Handle for "hugetlb.events.local" */
+ struct cgroup_file events_local_file[HUGE_MAX_HSTATE];
};
#define MEMFILE_PRIVATE(x, val) (((x) << 16) | (val))
#define MEMFILE_IDX(val) (((val) >> 16) & 0xffff)
#define MEMFILE_ATTR(val) ((val) & 0xffff)
+#define hugetlb_cgroup_from_counter(counter, idx) \
+ container_of(counter, struct hugetlb_cgroup, hugepage[idx])
+
static struct hugetlb_cgroup *root_h_cgroup __read_mostly;
static inline
@@ -178,6 +200,19 @@ static void hugetlb_cgroup_css_offline(s
} while (hugetlb_cgroup_have_usage(h_cg));
}
+static inline void hugetlb_event(struct hugetlb_cgroup *hugetlb, int idx,
+ enum hugetlb_memory_event event)
+{
+ atomic_long_inc(&hugetlb->events_local[idx][event]);
+ cgroup_file_notify(&hugetlb->events_local_file[idx]);
+
+ do {
+ atomic_long_inc(&hugetlb->events[idx][event]);
+ cgroup_file_notify(&hugetlb->events_file[idx]);
+ } while ((hugetlb = parent_hugetlb_cgroup(hugetlb)) &&
+ !hugetlb_cgroup_is_root(hugetlb));
+}
+
int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
struct hugetlb_cgroup **ptr)
{
@@ -202,8 +237,12 @@ again:
}
rcu_read_unlock();
- if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages, &counter))
+ if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages,
+ &counter)) {
ret = -ENOMEM;
+ hugetlb_event(hugetlb_cgroup_from_counter(counter, idx), idx,
+ HUGETLB_MAX);
+ }
css_put(&h_cg->css);
done:
*ptr = h_cg;
@@ -283,10 +322,45 @@ static u64 hugetlb_cgroup_read_u64(struc
}
}
+static int hugetlb_cgroup_read_u64_max(struct seq_file *seq, void *v)
+{
+ int idx;
+ u64 val;
+ struct cftype *cft = seq_cft(seq);
+ unsigned long limit;
+ struct page_counter *counter;
+ struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(seq_css(seq));
+
+ idx = MEMFILE_IDX(cft->private);
+ counter = &h_cg->hugepage[idx];
+
+ limit = round_down(PAGE_COUNTER_MAX,
+ 1 << huge_page_order(&hstates[idx]));
+
+ switch (MEMFILE_ATTR(cft->private)) {
+ case RES_USAGE:
+ val = (u64)page_counter_read(counter);
+ seq_printf(seq, "%llu\n", val * PAGE_SIZE);
+ break;
+ case RES_LIMIT:
+ val = (u64)counter->max;
+ if (val == limit)
+ seq_puts(seq, "max\n");
+ else
+ seq_printf(seq, "%llu\n", val * PAGE_SIZE);
+ break;
+ default:
+ BUG();
+ }
+
+ return 0;
+}
+
static DEFINE_MUTEX(hugetlb_limit_mutex);
static ssize_t hugetlb_cgroup_write(struct kernfs_open_file *of,
- char *buf, size_t nbytes, loff_t off)
+ char *buf, size_t nbytes, loff_t off,
+ const char *max)
{
int ret, idx;
unsigned long nr_pages;
@@ -296,7 +370,7 @@ static ssize_t hugetlb_cgroup_write(stru
return -EINVAL;
buf = strstrip(buf);
- ret = page_counter_memparse(buf, "-1", &nr_pages);
+ ret = page_counter_memparse(buf, max, &nr_pages);
if (ret)
return ret;
@@ -316,6 +390,18 @@ static ssize_t hugetlb_cgroup_write(stru
return ret ?: nbytes;
}
+static ssize_t hugetlb_cgroup_write_legacy(struct kernfs_open_file *of,
+ char *buf, size_t nbytes, loff_t off)
+{
+ return hugetlb_cgroup_write(of, buf, nbytes, off, "-1");
+}
+
+static ssize_t hugetlb_cgroup_write_dfl(struct kernfs_open_file *of,
+ char *buf, size_t nbytes, loff_t off)
+{
+ return hugetlb_cgroup_write(of, buf, nbytes, off, "max");
+}
+
static ssize_t hugetlb_cgroup_reset(struct kernfs_open_file *of,
char *buf, size_t nbytes, loff_t off)
{
@@ -350,7 +436,36 @@ static char *mem_fmt(char *buf, int size
return buf;
}
-static void __init __hugetlb_cgroup_file_init(int idx)
+static int __hugetlb_events_show(struct seq_file *seq, bool local)
+{
+ int idx;
+ long max;
+ struct cftype *cft = seq_cft(seq);
+ struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(seq_css(seq));
+
+ idx = MEMFILE_IDX(cft->private);
+
+ if (local)
+ max = atomic_long_read(&h_cg->events_local[idx][HUGETLB_MAX]);
+ else
+ max = atomic_long_read(&h_cg->events[idx][HUGETLB_MAX]);
+
+ seq_printf(seq, "max %lu\n", max);
+
+ return 0;
+}
+
+static int hugetlb_events_show(struct seq_file *seq, void *v)
+{
+ return __hugetlb_events_show(seq, false);
+}
+
+static int hugetlb_events_local_show(struct seq_file *seq, void *v)
+{
+ return __hugetlb_events_show(seq, true);
+}
+
+static void __init __hugetlb_cgroup_file_dfl_init(int idx)
{
char buf[32];
struct cftype *cft;
@@ -360,38 +475,93 @@ static void __init __hugetlb_cgroup_file
mem_fmt(buf, 32, huge_page_size(h));
/* Add the limit file */
- cft = &h->cgroup_files[0];
+ cft = &h->cgroup_files_dfl[0];
+ snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max", buf);
+ cft->private = MEMFILE_PRIVATE(idx, RES_LIMIT);
+ cft->seq_show = hugetlb_cgroup_read_u64_max;
+ cft->write = hugetlb_cgroup_write_dfl;
+ cft->flags = CFTYPE_NOT_ON_ROOT;
+
+ /* Add the current usage file */
+ cft = &h->cgroup_files_dfl[1];
+ snprintf(cft->name, MAX_CFTYPE_NAME, "%s.current", buf);
+ cft->private = MEMFILE_PRIVATE(idx, RES_USAGE);
+ cft->seq_show = hugetlb_cgroup_read_u64_max;
+ cft->flags = CFTYPE_NOT_ON_ROOT;
+
+ /* Add the events file */
+ cft = &h->cgroup_files_dfl[2];
+ snprintf(cft->name, MAX_CFTYPE_NAME, "%s.events", buf);
+ cft->private = MEMFILE_PRIVATE(idx, 0);
+ cft->seq_show = hugetlb_events_show;
+ cft->file_offset = offsetof(struct hugetlb_cgroup, events_file[idx]),
+ cft->flags = CFTYPE_NOT_ON_ROOT;
+
+ /* Add the events.local file */
+ cft = &h->cgroup_files_dfl[3];
+ snprintf(cft->name, MAX_CFTYPE_NAME, "%s.events.local", buf);
+ cft->private = MEMFILE_PRIVATE(idx, 0);
+ cft->seq_show = hugetlb_events_local_show;
+ cft->file_offset = offsetof(struct hugetlb_cgroup,
+ events_local_file[idx]),
+ cft->flags = CFTYPE_NOT_ON_ROOT;
+
+ /* NULL terminate the last cft */
+ cft = &h->cgroup_files_dfl[4];
+ memset(cft, 0, sizeof(*cft));
+
+ WARN_ON(cgroup_add_dfl_cftypes(&hugetlb_cgrp_subsys,
+ h->cgroup_files_dfl));
+}
+
+static void __init __hugetlb_cgroup_file_legacy_init(int idx)
+{
+ char buf[32];
+ struct cftype *cft;
+ struct hstate *h = &hstates[idx];
+
+ /* format the size */
+ mem_fmt(buf, 32, huge_page_size(h));
+
+ /* Add the limit file */
+ cft = &h->cgroup_files_legacy[0];
snprintf(cft->name, MAX_CFTYPE_NAME, "%s.limit_in_bytes", buf);
cft->private = MEMFILE_PRIVATE(idx, RES_LIMIT);
cft->read_u64 = hugetlb_cgroup_read_u64;
- cft->write = hugetlb_cgroup_write;
+ cft->write = hugetlb_cgroup_write_legacy;
/* Add the usage file */
- cft = &h->cgroup_files[1];
+ cft = &h->cgroup_files_legacy[1];
snprintf(cft->name, MAX_CFTYPE_NAME, "%s.usage_in_bytes", buf);
cft->private = MEMFILE_PRIVATE(idx, RES_USAGE);
cft->read_u64 = hugetlb_cgroup_read_u64;
/* Add the MAX usage file */
- cft = &h->cgroup_files[2];
+ cft = &h->cgroup_files_legacy[2];
snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max_usage_in_bytes", buf);
cft->private = MEMFILE_PRIVATE(idx, RES_MAX_USAGE);
cft->write = hugetlb_cgroup_reset;
cft->read_u64 = hugetlb_cgroup_read_u64;
/* Add the failcntfile */
- cft = &h->cgroup_files[3];
+ cft = &h->cgroup_files_legacy[3];
snprintf(cft->name, MAX_CFTYPE_NAME, "%s.failcnt", buf);
cft->private = MEMFILE_PRIVATE(idx, RES_FAILCNT);
cft->write = hugetlb_cgroup_reset;
cft->read_u64 = hugetlb_cgroup_read_u64;
/* NULL terminate the last cft */
- cft = &h->cgroup_files[4];
+ cft = &h->cgroup_files_legacy[4];
memset(cft, 0, sizeof(*cft));
WARN_ON(cgroup_add_legacy_cftypes(&hugetlb_cgrp_subsys,
- h->cgroup_files));
+ h->cgroup_files_legacy));
+}
+
+static void __init __hugetlb_cgroup_file_init(int idx)
+{
+ __hugetlb_cgroup_file_dfl_init(idx);
+ __hugetlb_cgroup_file_legacy_init(idx);
}
void __init hugetlb_cgroup_file_init(void)
@@ -433,8 +603,14 @@ void hugetlb_cgroup_migrate(struct page
return;
}
+static struct cftype hugetlb_files[] = {
+ {} /* terminate */
+};
+
struct cgroup_subsys hugetlb_cgrp_subsys = {
.css_alloc = hugetlb_cgroup_css_alloc,
.css_offline = hugetlb_cgroup_css_offline,
.css_free = hugetlb_cgroup_css_free,
+ .dfl_cftypes = hugetlb_files,
+ .legacy_cftypes = hugetlb_files,
};
_
Patches currently in -mm which might be from gscrivan@xxxxxxxxxx are
mm-hugetlb-controller-for-cgroups-v2.patch
