From: Stefan Roesch <shr@devkernel.io>
To: kernel-team@fb.com
Cc: linux-mm@kvack.org, david@redhat.com, oliver.sang@intel.com,
shr@devkernel.io, ltp@lists.linux.it
Subject: [LTP] [PATCH v4 2/2] add ksm test for smart-scan feature
Date: Tue, 5 Dec 2023 10:09:52 -0800 [thread overview]
Message-ID: <20231205180952.3843711-3-shr@devkernel.io> (raw)
In-Reply-To: <20231205180952.3843711-1-shr@devkernel.io>
This adds a new ksm (kernel samepage merging) test to evaluate the new
smart scan feature. It allocates a page and fills it with 'a'
characters. It captures the pages_skipped counter, waits for a few
iterations and captures the pages_skipped counter again. The expectation
is that over 50% of the page scans are skipped (There is only one page
that has KSM enabled and it gets scanned during each iteration and it
cannot be de-duplicated).
Signed-off-by: Stefan Roesch <shr@devkernel.io>
---
runtest/mm | 1 +
testcases/kernel/mem/.gitignore | 1 +
testcases/kernel/mem/ksm/ksm07.c | 113 +++++++++++++++++++++++++++++++
3 files changed, 115 insertions(+)
create mode 100644 testcases/kernel/mem/ksm/ksm07.c
diff --git a/runtest/mm b/runtest/mm
index f288fed36..d859b331c 100644
--- a/runtest/mm
+++ b/runtest/mm
@@ -70,6 +70,7 @@ ksm05 ksm05 -I 10
ksm06 ksm06
ksm06_1 ksm06 -n 10
ksm06_2 ksm06 -n 8000
+ksm07 ksm07
cpuset01 cpuset01
diff --git a/testcases/kernel/mem/.gitignore b/testcases/kernel/mem/.gitignore
index 7258489ed..c96fe8bfc 100644
--- a/testcases/kernel/mem/.gitignore
+++ b/testcases/kernel/mem/.gitignore
@@ -53,6 +53,7 @@
/ksm/ksm04
/ksm/ksm05
/ksm/ksm06
+/ksm/ksm07
/mem/mem02
/mmapstress/mmap-corruption01
/mmapstress/mmapstress01
diff --git a/testcases/kernel/mem/ksm/ksm07.c b/testcases/kernel/mem/ksm/ksm07.c
new file mode 100644
index 000000000..1ae5e9cd8
--- /dev/null
+++ b/testcases/kernel/mem/ksm/ksm07.c
@@ -0,0 +1,113 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2010-2023 Red Hat, Inc.
+ */
+/*\
+ * [Description]
+ *
+ * Kernel Samepage Merging (KSM) for smart scan feature
+ *
+ * Test allocates a page and fills it with 'a' characters. It captures the
+ * pages_skipped counter, waits for a few iterations and captures the
+ * pages_skipped counter again. The expectation is that over 50% of the page
+ * scans are skipped. (There is only one page that has KSM enabled and it gets
+ * scanned during each iteration and it cannot be de-duplicated.)
+ *
+ * Smart scan feature was added in kernel v6.7.
+ *
+ * [Prerequisites]
+ *
+ * ksm and ksmtuned daemons need to be disabled. Otherwise, it could
+ * distrub the testing as they also change some ksm tunables depends
+ * on current workloads.
+ */
+#include <sys/wait.h>
+#include "mem.h"
+
+/* This test allocates one page, fills the page with a's, captures the
+ * full_scan and pages_skipped counters. Then it makes sure at least 3
+ * full scans have been performed and measures the above counters again.
+ * The expectation is that at least 50% of the pages are skipped.
+ *
+ * To wait for at least 3 scans it uses the wait_ksmd_full_scan() function. In
+ * reality, it will be a lot more scans as the wait_ksmd_full_scan() function
+ * sleeps for one second.
+ */
+static void verify_ksm(void)
+{
+ int full_scans_begin;
+ int full_scans_end;
+ int pages_skipped_begin;
+ int pages_skipped_end;
+ int diff_pages;
+ int diff_scans;
+ unsigned long page_size;
+ char *memory;
+
+ /* Apply for the space for memory. */
+ page_size = sysconf(_SC_PAGE_SIZE);
+ memory = SAFE_MALLOC(page_size);
+ memory = SAFE_MMAP(NULL, page_size, PROT_READ|PROT_WRITE,
+ MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+#ifdef HAVE_DECL_MADV_MERGEABLE
+ if (madvise(memory, page_size, MADV_MERGEABLE) == -1)
+ tst_brk(TBROK|TERRNO, "madvise");
+#endif
+ memset(memory, 'a', page_size);
+
+ tst_res(TINFO, "KSM merging");
+
+ if (access(PATH_KSM "max_page_sharing", F_OK) == 0)
+ SAFE_FILE_PRINTF(PATH_KSM "run", "2");
+
+ /* Set defalut ksm scan values. */
+ SAFE_FILE_PRINTF(PATH_KSM "run", "1");
+ SAFE_FILE_PRINTF(PATH_KSM "pages_to_scan", "%ld", 100l);
+ SAFE_FILE_PRINTF(PATH_KSM "sleep_millisecs", "0");
+
+ /* Measure pages skipped aka "smart scan". */
+ SAFE_FILE_SCANF(PATH_KSM "full_scans", "%d", &full_scans_begin);
+ SAFE_FILE_SCANF(PATH_KSM "pages_skipped", "%d", &pages_skipped_begin);
+ wait_ksmd_full_scan();
+
+ tst_res(TINFO, "stop KSM");
+ SAFE_FILE_PRINTF(PATH_KSM "run", "0");
+
+ SAFE_FILE_SCANF(PATH_KSM "full_scans", "%d", &full_scans_end);
+ SAFE_FILE_SCANF(PATH_KSM "pages_skipped", "%d", &pages_skipped_end);
+ diff_pages = pages_skipped_end - pages_skipped_begin;
+ diff_scans = full_scans_end - full_scans_begin;
+
+ if (diff_pages < diff_scans * 50 / 100) {
+ tst_res(TINFO, "number of pages %d", diff_pages);
+ tst_res(TINFO, "number of scans %d", diff_scans);
+ tst_res(TFAIL, "not enough pages have been skipped by smart_scan.");
+ } else {
+ tst_res(TPASS, "smart_scan skipped more than 50%% of the pages.");
+ }
+
+#ifdef HAVE_DECL_MADV_MERGEABLE
+ if (madvise(memory, page_size, MADV_UNMERGEABLE) == -1)
+ tst_brk(TBROK|TERRNO, "madvise");
+#endif
+}
+
+static struct tst_test test = {
+ .needs_root = 1,
+ .options = (struct tst_option[]) {
+ {}
+ },
+ .save_restore = (const struct tst_path_val[]) {
+ {PATH_KSM "pages_skipped", NULL, TST_SR_TCONF},
+ {PATH_KSM "run", NULL, TST_SR_TCONF},
+ {PATH_KSM "sleep_millisecs", NULL, TST_SR_TCONF},
+ {PATH_KSM "smart_scan", "1",
+ TST_SR_SKIP_MISSING | TST_SR_TCONF},
+ {}
+ },
+ .needs_kconfigs = (const char *const[]){
+ "CONFIG_KSM=y",
+ NULL
+ },
+ .test_all = verify_ksm,
+};
--
2.39.3
--
Mailing list info: https://lists.linux.it/listinfo/ltp
WARNING: multiple messages have this Message-ID (diff)
From: Stefan Roesch <shr@devkernel.io>
To: kernel-team@fb.com
Cc: shr@devkernel.io, david@redhat.com, oliver.sang@intel.com,
linux-mm@kvack.org, ltp@lists.linux.it, pvorel@suse.cz,
liwang@redhat.com
Subject: [PATCH v4 2/2] add ksm test for smart-scan feature
Date: Tue, 5 Dec 2023 10:09:52 -0800 [thread overview]
Message-ID: <20231205180952.3843711-3-shr@devkernel.io> (raw)
In-Reply-To: <20231205180952.3843711-1-shr@devkernel.io>
This adds a new ksm (kernel samepage merging) test to evaluate the new
smart scan feature. It allocates a page and fills it with 'a'
characters. It captures the pages_skipped counter, waits for a few
iterations and captures the pages_skipped counter again. The expectation
is that over 50% of the page scans are skipped (There is only one page
that has KSM enabled and it gets scanned during each iteration and it
cannot be de-duplicated).
Signed-off-by: Stefan Roesch <shr@devkernel.io>
---
runtest/mm | 1 +
testcases/kernel/mem/.gitignore | 1 +
testcases/kernel/mem/ksm/ksm07.c | 113 +++++++++++++++++++++++++++++++
3 files changed, 115 insertions(+)
create mode 100644 testcases/kernel/mem/ksm/ksm07.c
diff --git a/runtest/mm b/runtest/mm
index f288fed36..d859b331c 100644
--- a/runtest/mm
+++ b/runtest/mm
@@ -70,6 +70,7 @@ ksm05 ksm05 -I 10
ksm06 ksm06
ksm06_1 ksm06 -n 10
ksm06_2 ksm06 -n 8000
+ksm07 ksm07
cpuset01 cpuset01
diff --git a/testcases/kernel/mem/.gitignore b/testcases/kernel/mem/.gitignore
index 7258489ed..c96fe8bfc 100644
--- a/testcases/kernel/mem/.gitignore
+++ b/testcases/kernel/mem/.gitignore
@@ -53,6 +53,7 @@
/ksm/ksm04
/ksm/ksm05
/ksm/ksm06
+/ksm/ksm07
/mem/mem02
/mmapstress/mmap-corruption01
/mmapstress/mmapstress01
diff --git a/testcases/kernel/mem/ksm/ksm07.c b/testcases/kernel/mem/ksm/ksm07.c
new file mode 100644
index 000000000..1ae5e9cd8
--- /dev/null
+++ b/testcases/kernel/mem/ksm/ksm07.c
@@ -0,0 +1,113 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2010-2023 Red Hat, Inc.
+ */
+/*\
+ * [Description]
+ *
+ * Kernel Samepage Merging (KSM) for smart scan feature
+ *
+ * Test allocates a page and fills it with 'a' characters. It captures the
+ * pages_skipped counter, waits for a few iterations and captures the
+ * pages_skipped counter again. The expectation is that over 50% of the page
+ * scans are skipped. (There is only one page that has KSM enabled and it gets
+ * scanned during each iteration and it cannot be de-duplicated.)
+ *
+ * Smart scan feature was added in kernel v6.7.
+ *
+ * [Prerequisites]
+ *
+ * ksm and ksmtuned daemons need to be disabled. Otherwise, it could
+ * distrub the testing as they also change some ksm tunables depends
+ * on current workloads.
+ */
+#include <sys/wait.h>
+#include "mem.h"
+
+/* This test allocates one page, fills the page with a's, captures the
+ * full_scan and pages_skipped counters. Then it makes sure at least 3
+ * full scans have been performed and measures the above counters again.
+ * The expectation is that at least 50% of the pages are skipped.
+ *
+ * To wait for at least 3 scans it uses the wait_ksmd_full_scan() function. In
+ * reality, it will be a lot more scans as the wait_ksmd_full_scan() function
+ * sleeps for one second.
+ */
+static void verify_ksm(void)
+{
+ int full_scans_begin;
+ int full_scans_end;
+ int pages_skipped_begin;
+ int pages_skipped_end;
+ int diff_pages;
+ int diff_scans;
+ unsigned long page_size;
+ char *memory;
+
+ /* Apply for the space for memory. */
+ page_size = sysconf(_SC_PAGE_SIZE);
+ memory = SAFE_MALLOC(page_size);
+ memory = SAFE_MMAP(NULL, page_size, PROT_READ|PROT_WRITE,
+ MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+#ifdef HAVE_DECL_MADV_MERGEABLE
+ if (madvise(memory, page_size, MADV_MERGEABLE) == -1)
+ tst_brk(TBROK|TERRNO, "madvise");
+#endif
+ memset(memory, 'a', page_size);
+
+ tst_res(TINFO, "KSM merging");
+
+ if (access(PATH_KSM "max_page_sharing", F_OK) == 0)
+ SAFE_FILE_PRINTF(PATH_KSM "run", "2");
+
+ /* Set defalut ksm scan values. */
+ SAFE_FILE_PRINTF(PATH_KSM "run", "1");
+ SAFE_FILE_PRINTF(PATH_KSM "pages_to_scan", "%ld", 100l);
+ SAFE_FILE_PRINTF(PATH_KSM "sleep_millisecs", "0");
+
+ /* Measure pages skipped aka "smart scan". */
+ SAFE_FILE_SCANF(PATH_KSM "full_scans", "%d", &full_scans_begin);
+ SAFE_FILE_SCANF(PATH_KSM "pages_skipped", "%d", &pages_skipped_begin);
+ wait_ksmd_full_scan();
+
+ tst_res(TINFO, "stop KSM");
+ SAFE_FILE_PRINTF(PATH_KSM "run", "0");
+
+ SAFE_FILE_SCANF(PATH_KSM "full_scans", "%d", &full_scans_end);
+ SAFE_FILE_SCANF(PATH_KSM "pages_skipped", "%d", &pages_skipped_end);
+ diff_pages = pages_skipped_end - pages_skipped_begin;
+ diff_scans = full_scans_end - full_scans_begin;
+
+ if (diff_pages < diff_scans * 50 / 100) {
+ tst_res(TINFO, "number of pages %d", diff_pages);
+ tst_res(TINFO, "number of scans %d", diff_scans);
+ tst_res(TFAIL, "not enough pages have been skipped by smart_scan.");
+ } else {
+ tst_res(TPASS, "smart_scan skipped more than 50%% of the pages.");
+ }
+
+#ifdef HAVE_DECL_MADV_MERGEABLE
+ if (madvise(memory, page_size, MADV_UNMERGEABLE) == -1)
+ tst_brk(TBROK|TERRNO, "madvise");
+#endif
+}
+
+static struct tst_test test = {
+ .needs_root = 1,
+ .options = (struct tst_option[]) {
+ {}
+ },
+ .save_restore = (const struct tst_path_val[]) {
+ {PATH_KSM "pages_skipped", NULL, TST_SR_TCONF},
+ {PATH_KSM "run", NULL, TST_SR_TCONF},
+ {PATH_KSM "sleep_millisecs", NULL, TST_SR_TCONF},
+ {PATH_KSM "smart_scan", "1",
+ TST_SR_SKIP_MISSING | TST_SR_TCONF},
+ {}
+ },
+ .needs_kconfigs = (const char *const[]){
+ "CONFIG_KSM=y",
+ NULL
+ },
+ .test_all = verify_ksm,
+};
--
2.39.3
next prev parent reply other threads:[~2023-12-05 18:10 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-05 18:09 [LTP] [PATCH v4 0/2] KSM: support smart-scan feature Stefan Roesch
2023-12-05 18:09 ` Stefan Roesch
2023-12-05 18:09 ` [LTP] [PATCH v4 1/2] mem: disable KSM smart scan for ksm tests Stefan Roesch
2023-12-05 18:09 ` Stefan Roesch
2023-12-05 18:09 ` Stefan Roesch [this message]
2023-12-05 18:09 ` [PATCH v4 2/2] add ksm test for smart-scan feature Stefan Roesch
2023-12-05 20:11 ` [LTP] " Petr Vorel
2023-12-05 20:11 ` Petr Vorel
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=20231205180952.3843711-3-shr@devkernel.io \
--to=shr@devkernel.io \
--cc=david@redhat.com \
--cc=kernel-team@fb.com \
--cc=linux-mm@kvack.org \
--cc=ltp@lists.linux.it \
--cc=oliver.sang@intel.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.