All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Michal Koutný" <mkoutny@suse.com>
To: cgroups@vger.kernel.org, linux-mm@kvack.org
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Michal Hocko <mhocko@kernel.org>,
	Roman Gushchin <roman.gushchin@linux.dev>,
	Shakeel Butt <shakeelb@google.com>,
	linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org,
	Richard Palethorpe <rpalethorpe@suse.de>,
	Muhammad Usama Anjum <usama.anjum@collabora.com>
Subject: [PATCH v3 3/5] selftests: memcg: Adjust expected reclaim values of protected cgroups
Date: Tue, 24 May 2022 18:29:53 +0200	[thread overview]
Message-ID: <20220524162955.8635-4-mkoutny@suse.com> (raw)
In-Reply-To: <20220524162955.8635-1-mkoutny@suse.com>

The numbers are not easy to derive in a closed form (certainly mere
protections ratios do not apply), therefore use a simulation to obtain
expected numbers.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
Acked-by: Roman Gushchin <roman.gushchin@linux.dev>
---
 MAINTAINERS                                   |  1 +
 .../selftests/cgroup/memcg_protection.m       | 89 +++++++++++++++++++
 .../selftests/cgroup/test_memcontrol.c        | 29 +++---
 3 files changed, 107 insertions(+), 12 deletions(-)
 create mode 100644 tools/testing/selftests/cgroup/memcg_protection.m

diff --git a/MAINTAINERS b/MAINTAINERS
index 78c57046fa93..b28b6aeb8636 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -5029,6 +5029,7 @@ L:	linux-mm@kvack.org
 S:	Maintained
 F:	mm/memcontrol.c
 F:	mm/swap_cgroup.c
+F:	tools/testing/selftests/cgroup/memcg_protection.m
 F:	tools/testing/selftests/cgroup/test_kmem.c
 F:	tools/testing/selftests/cgroup/test_memcontrol.c
 
diff --git a/tools/testing/selftests/cgroup/memcg_protection.m b/tools/testing/selftests/cgroup/memcg_protection.m
new file mode 100644
index 000000000000..051daa3477b6
--- /dev/null
+++ b/tools/testing/selftests/cgroup/memcg_protection.m
@@ -0,0 +1,89 @@
+% SPDX-License-Identifier: GPL-2.0
+%
+% run as: octave-cli memcg_protection.m
+%
+% This script simulates reclaim protection behavior on a single level of memcg
+% hierarchy to illustrate how overcommitted protection spreads among siblings
+% (as it depends also on their current consumption).
+%
+% Simulation assumes siblings consumed the initial amount of memory (w/out
+% reclaim) and then the reclaim starts, all memory is reclaimable, i.e. treated
+% same. It simulates only non-low reclaim and assumes all memory.min = 0.
+%
+% Input configurations
+% --------------------
+% E number	parent effective protection
+% n vector	nominal protection of siblings set at the given level (memory.low)
+% c vector	current consumption -,,- (memory.current)
+
+% example from testcase (values in GB)
+E = 50 / 1024;
+n = [75 25 0 500 ] / 1024;
+c = [50 50 50 0] / 1024;
+
+% Reclaim parameters
+% ------------------
+
+% Minimal reclaim amount (GB)
+cluster = 32*4 / 2**20;
+
+% Reclaim coefficient (think as 0.5^sc->priority)
+alpha = .1
+
+% Simulation parameters
+% ---------------------
+epsilon = 1e-7;
+timeout = 1000;
+
+% Simulation loop
+% ---------------
+
+ch = [];
+eh = [];
+rh = [];
+
+for t = 1:timeout
+        % low_usage
+        u = min(c, n);
+        siblings = sum(u);
+
+        % effective_protection()
+        protected = min(n, c);                % start with nominal
+        e = protected * min(1, E / siblings); % normalize overcommit
+
+        % recursive protection
+        unclaimed = max(0, E - siblings);
+        parent_overuse = sum(c) - siblings;
+        if (unclaimed > 0 && parent_overuse > 0)
+                overuse = max(0, c - protected);
+                e += unclaimed * (overuse / parent_overuse);
+        endif
+
+        % get_scan_count()
+        r = alpha * c;             % assume all memory is in a single LRU list
+
+        % commit 1bc63fb1272b ("mm, memcg: make scan aggression always exclude protection")
+        sz = max(e, c);
+        r .*= (1 - (e+epsilon) ./ (sz+epsilon));
+
+        % uncomment to debug prints
+        % e, c, r
+
+        % nothing to reclaim, reached equilibrium
+        if max(r) < epsilon
+                break;
+        endif
+
+        % SWAP_CLUSTER_MAX roundup
+        r = max(r, (r > epsilon) .* cluster);
+        % XXX here I do parallel reclaim of all siblings
+        % in reality reclaim is serialized and each sibling recalculates own residual
+        c = max(c - r, 0);
+
+        ch = [ch ; c];
+        eh = [eh ; e];
+        rh = [rh ; r];
+endfor
+
+t
+c, e
diff --git a/tools/testing/selftests/cgroup/test_memcontrol.c b/tools/testing/selftests/cgroup/test_memcontrol.c
index 4924425639b0..dc2c7d6e3572 100644
--- a/tools/testing/selftests/cgroup/test_memcontrol.c
+++ b/tools/testing/selftests/cgroup/test_memcontrol.c
@@ -248,7 +248,7 @@ static int cg_test_proc_killed(const char *cgroup)
 /*
  * First, this test creates the following hierarchy:
  * A       memory.min = 50M,  memory.max = 200M
- * A/B     memory.min = 50M,  memory.current = 50M
+ * A/B     memory.min = 50M
  * A/B/C   memory.min = 75M,  memory.current = 50M
  * A/B/D   memory.min = 25M,  memory.current = 50M
  * A/B/E   memory.min = 0,    memory.current = 50M
@@ -259,10 +259,13 @@ static int cg_test_proc_killed(const char *cgroup)
  * Then it creates A/G and creates a significant
  * memory pressure in it.
  *
+ * Then it checks actual memory usages and expects that:
  * A/B    memory.current ~= 50M
- * A/B/C  memory.current ~= 33M
- * A/B/D  memory.current ~= 17M
- * A/B/F  memory.current ~= 0
+ * A/B/C  memory.current ~= 29M
+ * A/B/D  memory.current ~= 21M
+ * A/B/E  memory.current ~= 0
+ * A/B/F  memory.current  = 0
+ * (for origin of the numbers, see model in memcg_protection.m.)
  *
  * After that it tries to allocate more than there is
  * unprotected memory in A available, and checks
@@ -365,10 +368,10 @@ static int test_memcg_min(const char *root)
 	for (i = 0; i < ARRAY_SIZE(children); i++)
 		c[i] = cg_read_long(children[i], "memory.current");
 
-	if (!values_close(c[0], MB(33), 10))
+	if (!values_close(c[0], MB(29), 10))
 		goto cleanup;
 
-	if (!values_close(c[1], MB(17), 10))
+	if (!values_close(c[1], MB(21), 10))
 		goto cleanup;
 
 	if (c[3] != 0)
@@ -405,7 +408,7 @@ static int test_memcg_min(const char *root)
 /*
  * First, this test creates the following hierarchy:
  * A       memory.low = 50M,  memory.max = 200M
- * A/B     memory.low = 50M,  memory.current = 50M
+ * A/B     memory.low = 50M
  * A/B/C   memory.low = 75M,  memory.current = 50M
  * A/B/D   memory.low = 25M,  memory.current = 50M
  * A/B/E   memory.low = 0,    memory.current = 50M
@@ -417,9 +420,11 @@ static int test_memcg_min(const char *root)
  *
  * Then it checks actual memory usages and expects that:
  * A/B    memory.current ~= 50M
- * A/B/   memory.current ~= 33M
- * A/B/D  memory.current ~= 17M
- * A/B/F  memory.current ~= 0
+ * A/B/C  memory.current ~= 29M
+ * A/B/D  memory.current ~= 21M
+ * A/B/E  memory.current ~= 0
+ * A/B/F  memory.current  = 0
+ * (for origin of the numbers, see model in memcg_protection.m.)
  *
  * After that it tries to allocate more than there is
  * unprotected memory in A available,
@@ -512,10 +517,10 @@ static int test_memcg_low(const char *root)
 	for (i = 0; i < ARRAY_SIZE(children); i++)
 		c[i] = cg_read_long(children[i], "memory.current");
 
-	if (!values_close(c[0], MB(33), 10))
+	if (!values_close(c[0], MB(29), 10))
 		goto cleanup;
 
-	if (!values_close(c[1], MB(17), 10))
+	if (!values_close(c[1], MB(21), 10))
 		goto cleanup;
 
 	if (c[3] != 0)
-- 
2.35.3


WARNING: multiple messages have this Message-ID (diff)
From: "Michal Koutný" <mkoutny-IBi9RG/b67k@public.gmane.org>
To: cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
	linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org
Cc: Andrew Morton
	<akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>,
	Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>,
	Michal Hocko <mhocko-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>,
	Roman Gushchin
	<roman.gushchin-fxUVXftIFDnyG1zEObXtfA@public.gmane.org>,
	Shakeel Butt <shakeelb-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
	linux-kselftest-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
	Richard Palethorpe <rpalethorpe-l3A5Bk7waGM@public.gmane.org>,
	Muhammad Usama Anjum
	<usama.anjum-ZGY8ohtN/8qB+jHODAdFcQ@public.gmane.org>
Subject: [PATCH v3 3/5] selftests: memcg: Adjust expected reclaim values of protected cgroups
Date: Tue, 24 May 2022 18:29:53 +0200	[thread overview]
Message-ID: <20220524162955.8635-4-mkoutny@suse.com> (raw)
In-Reply-To: <20220524162955.8635-1-mkoutny-IBi9RG/b67k@public.gmane.org>

The numbers are not easy to derive in a closed form (certainly mere
protections ratios do not apply), therefore use a simulation to obtain
expected numbers.

Signed-off-by: Michal Koutn√Ω <mkoutny-IBi9RG/b67k@public.gmane.org>
Acked-by: Roman Gushchin <roman.gushchin-fxUVXftIFDnyG1zEObXtfA@public.gmane.org>
---
 MAINTAINERS                                   |  1 +
 .../selftests/cgroup/memcg_protection.m       | 89 +++++++++++++++++++
 .../selftests/cgroup/test_memcontrol.c        | 29 +++---
 3 files changed, 107 insertions(+), 12 deletions(-)
 create mode 100644 tools/testing/selftests/cgroup/memcg_protection.m

diff --git a/MAINTAINERS b/MAINTAINERS
index 78c57046fa93..b28b6aeb8636 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -5029,6 +5029,7 @@ L:	linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org
 S:	Maintained
 F:	mm/memcontrol.c
 F:	mm/swap_cgroup.c
+F:	tools/testing/selftests/cgroup/memcg_protection.m
 F:	tools/testing/selftests/cgroup/test_kmem.c
 F:	tools/testing/selftests/cgroup/test_memcontrol.c
 
diff --git a/tools/testing/selftests/cgroup/memcg_protection.m b/tools/testing/selftests/cgroup/memcg_protection.m
new file mode 100644
index 000000000000..051daa3477b6
--- /dev/null
+++ b/tools/testing/selftests/cgroup/memcg_protection.m
@@ -0,0 +1,89 @@
+% SPDX-License-Identifier: GPL-2.0
+%
+% run as: octave-cli memcg_protection.m
+%
+% This script simulates reclaim protection behavior on a single level of memcg
+% hierarchy to illustrate how overcommitted protection spreads among siblings
+% (as it depends also on their current consumption).
+%
+% Simulation assumes siblings consumed the initial amount of memory (w/out
+% reclaim) and then the reclaim starts, all memory is reclaimable, i.e. treated
+% same. It simulates only non-low reclaim and assumes all memory.min = 0.
+%
+% Input configurations
+% --------------------
+% E number	parent effective protection
+% n vector	nominal protection of siblings set at the given level (memory.low)
+% c vector	current consumption -,,- (memory.current)
+
+% example from testcase (values in GB)
+E = 50 / 1024;
+n = [75 25 0 500 ] / 1024;
+c = [50 50 50 0] / 1024;
+
+% Reclaim parameters
+% ------------------
+
+% Minimal reclaim amount (GB)
+cluster = 32*4 / 2**20;
+
+% Reclaim coefficient (think as 0.5^sc->priority)
+alpha = .1
+
+% Simulation parameters
+% ---------------------
+epsilon = 1e-7;
+timeout = 1000;
+
+% Simulation loop
+% ---------------
+
+ch = [];
+eh = [];
+rh = [];
+
+for t = 1:timeout
+        % low_usage
+        u = min(c, n);
+        siblings = sum(u);
+
+        % effective_protection()
+        protected = min(n, c);                % start with nominal
+        e = protected * min(1, E / siblings); % normalize overcommit
+
+        % recursive protection
+        unclaimed = max(0, E - siblings);
+        parent_overuse = sum(c) - siblings;
+        if (unclaimed > 0 && parent_overuse > 0)
+                overuse = max(0, c - protected);
+                e += unclaimed * (overuse / parent_overuse);
+        endif
+
+        % get_scan_count()
+        r = alpha * c;             % assume all memory is in a single LRU list
+
+        % commit 1bc63fb1272b ("mm, memcg: make scan aggression always exclude protection")
+        sz = max(e, c);
+        r .*= (1 - (e+epsilon) ./ (sz+epsilon));
+
+        % uncomment to debug prints
+        % e, c, r
+
+        % nothing to reclaim, reached equilibrium
+        if max(r) < epsilon
+                break;
+        endif
+
+        % SWAP_CLUSTER_MAX roundup
+        r = max(r, (r > epsilon) .* cluster);
+        % XXX here I do parallel reclaim of all siblings
+        % in reality reclaim is serialized and each sibling recalculates own residual
+        c = max(c - r, 0);
+
+        ch = [ch ; c];
+        eh = [eh ; e];
+        rh = [rh ; r];
+endfor
+
+t
+c, e
diff --git a/tools/testing/selftests/cgroup/test_memcontrol.c b/tools/testing/selftests/cgroup/test_memcontrol.c
index 4924425639b0..dc2c7d6e3572 100644
--- a/tools/testing/selftests/cgroup/test_memcontrol.c
+++ b/tools/testing/selftests/cgroup/test_memcontrol.c
@@ -248,7 +248,7 @@ static int cg_test_proc_killed(const char *cgroup)
 /*
  * First, this test creates the following hierarchy:
  * A       memory.min = 50M,  memory.max = 200M
- * A/B     memory.min = 50M,  memory.current = 50M
+ * A/B     memory.min = 50M
  * A/B/C   memory.min = 75M,  memory.current = 50M
  * A/B/D   memory.min = 25M,  memory.current = 50M
  * A/B/E   memory.min = 0,    memory.current = 50M
@@ -259,10 +259,13 @@ static int cg_test_proc_killed(const char *cgroup)
  * Then it creates A/G and creates a significant
  * memory pressure in it.
  *
+ * Then it checks actual memory usages and expects that:
  * A/B    memory.current ~= 50M
- * A/B/C  memory.current ~= 33M
- * A/B/D  memory.current ~= 17M
- * A/B/F  memory.current ~= 0
+ * A/B/C  memory.current ~= 29M
+ * A/B/D  memory.current ~= 21M
+ * A/B/E  memory.current ~= 0
+ * A/B/F  memory.current  = 0
+ * (for origin of the numbers, see model in memcg_protection.m.)
  *
  * After that it tries to allocate more than there is
  * unprotected memory in A available, and checks
@@ -365,10 +368,10 @@ static int test_memcg_min(const char *root)
 	for (i = 0; i < ARRAY_SIZE(children); i++)
 		c[i] = cg_read_long(children[i], "memory.current");
 
-	if (!values_close(c[0], MB(33), 10))
+	if (!values_close(c[0], MB(29), 10))
 		goto cleanup;
 
-	if (!values_close(c[1], MB(17), 10))
+	if (!values_close(c[1], MB(21), 10))
 		goto cleanup;
 
 	if (c[3] != 0)
@@ -405,7 +408,7 @@ static int test_memcg_min(const char *root)
 /*
  * First, this test creates the following hierarchy:
  * A       memory.low = 50M,  memory.max = 200M
- * A/B     memory.low = 50M,  memory.current = 50M
+ * A/B     memory.low = 50M
  * A/B/C   memory.low = 75M,  memory.current = 50M
  * A/B/D   memory.low = 25M,  memory.current = 50M
  * A/B/E   memory.low = 0,    memory.current = 50M
@@ -417,9 +420,11 @@ static int test_memcg_min(const char *root)
  *
  * Then it checks actual memory usages and expects that:
  * A/B    memory.current ~= 50M
- * A/B/   memory.current ~= 33M
- * A/B/D  memory.current ~= 17M
- * A/B/F  memory.current ~= 0
+ * A/B/C  memory.current ~= 29M
+ * A/B/D  memory.current ~= 21M
+ * A/B/E  memory.current ~= 0
+ * A/B/F  memory.current  = 0
+ * (for origin of the numbers, see model in memcg_protection.m.)
  *
  * After that it tries to allocate more than there is
  * unprotected memory in A available,
@@ -512,10 +517,10 @@ static int test_memcg_low(const char *root)
 	for (i = 0; i < ARRAY_SIZE(children); i++)
 		c[i] = cg_read_long(children[i], "memory.current");
 
-	if (!values_close(c[0], MB(33), 10))
+	if (!values_close(c[0], MB(29), 10))
 		goto cleanup;
 
-	if (!values_close(c[1], MB(17), 10))
+	if (!values_close(c[1], MB(21), 10))
 		goto cleanup;
 
 	if (c[3] != 0)
-- 
2.35.3


  parent reply	other threads:[~2022-05-24 16:30 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-24 16:29 [PATCH v3 0/5] memcontrol selftests fixups Michal Koutný
2022-05-24 16:29 ` Michal Koutný
2022-05-24 16:29 ` [PATCH v3 1/5] selftests: memcg: Fix compilation Michal Koutný
2022-05-24 16:29   ` Michal Koutný
2022-05-24 16:29 ` [PATCH v3 2/5] selftests: memcg: Expect no low events in unprotected sibling Michal Koutný
2022-05-24 16:29   ` Michal Koutný
2022-05-25  2:27   ` Roman Gushchin
2022-05-25  2:27     ` Roman Gushchin
2022-05-24 16:29 ` Michal Koutný [this message]
2022-05-24 16:29   ` [PATCH v3 3/5] selftests: memcg: Adjust expected reclaim values of protected cgroups Michal Koutný
2022-05-24 16:29 ` [PATCH v3 4/5] selftests: memcg: Remove protection from top level memcg Michal Koutný
2022-05-24 16:29   ` Michal Koutný
2022-05-24 16:29 ` [PATCH v3 5/5] selftests: memcg: Factor out common parts of memory.{low,min} tests Michal Koutný
2022-05-24 16:29   ` Michal Koutný
2022-05-25  2:26   ` Roman Gushchin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220524162955.8635-4-mkoutny@suse.com \
    --to=mkoutny@suse.com \
    --cc=akpm@linux-foundation.org \
    --cc=cgroups@vger.kernel.org \
    --cc=hannes@cmpxchg.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mhocko@kernel.org \
    --cc=roman.gushchin@linux.dev \
    --cc=rpalethorpe@suse.de \
    --cc=shakeelb@google.com \
    --cc=usama.anjum@collabora.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.