Re: Fwd: [linux-test-project/ltp] migrate_pages03 killed on aarch64, possibly because ksm takes too long (#431)
I was just looking at that - we don't run that test due to:
root@intel-corei7-64:/opt/ltp/testcases/bin# ./migrate_pages03 tst_test.c:770: CONF: require libnuma >= 2 and it's development packages
Daniel - is this something we should add?
Dan
On Tue, Nov 27, 2018 at 01:34:11PM -0500, Anmar Oueja wrote:
Saw this. Might be a kernel bug.
anmar
---------- Forwarded message --------- From: Jan Stancek notifications@github.com Date: Tue, Nov 27, 2018 at 1:09 PM Subject: Re: [linux-test-project/ltp] migrate_pages03 killed on aarch64, possibly because ksm takes too long (#431) To: linux-test-project/ltp ltp@noreply.github.com Cc: Subscribed subscribed@noreply.github.com
I overlooked that we wait for ksm scan in setup(), so that is not the issue. You can see in output that it completed.
This is looking more like kernel bug at the moment. Using same setup, but with 4.18 kernel I see it completing in couple seconds:
]# uname -r ; time ./migrate_pages03 4.18.0-12.el8.aarch64 tst_test.c:1085: INFO: Timeout per run is 0h 05m 00s ksm_helper.c:48: INFO: ksm daemon takes 1s to run two full scans migrate_pages03.c:138: PASS: migrate_pages() passed
Summary: passed 1 failed 0 skipped 0 warnings 0
real 0m9.826s user 0m0.002s sys 0m8.773s [root@huawei-t2280-04 migrate_pages]# numactl -H available: 4 nodes (0-3) node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 node 0 size: 64792 MB node 0 free: 59272 MB node 1 cpus: 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 node 1 size: 65404 MB node 1 free: 63229 MB node 2 cpus: 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 node 2 size: 65404 MB node 2 free: 64117 MB node 3 cpus: 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 node 3 size: 65402 MB node 3 free: 64653 MB node distances: node 0 1 2 3 0: 10 15 20 20 1: 15 10 20 20 2: 20 20 10 15 3: 20 20 15 10
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/linux-test-project/ltp/issues/431#issuecomment-442159766, or mute the thread https://github.com/notifications/unsubscribe-auth/AAnbfxxXJYRHl2428m1dvha9s9N2dWtuks5uzX_AgaJpZM4Y2A5T .
On 11/27/18 5:16 PM, Dan Rue wrote:
I was just looking at that - we don't run that test due to:
root@intel-corei7-64:/opt/ltp/testcases/bin# ./migrate_pages03 tst_test.c:770: CONF: require libnuma >= 2 and it's development packagesDaniel - is this something we should add?
Dan
I faced this recently in my LXC, the one generating the .deb packages. I needed libnuma1 and libnuma-dev in debian in order to have some tests successfully built.
My 5 cents...
Hello!
On Tue, 27 Nov 2018 at 13:22, Rafael David Tinoco rafael.tinoco@linaro.org wrote:
On 11/27/18 5:16 PM, Dan Rue wrote:
I was just looking at that - we don't run that test due to:
root@intel-corei7-64:/opt/ltp/testcases/bin# ./migrate_pages03 tst_test.c:770: CONF: require libnuma >= 2 and it's development packagesDaniel - is this something we should add?
I faced this recently in my LXC, the one generating the .deb packages. I needed libnuma1 and libnuma-dev in debian in order to have some tests successfully built.
We're currently shipping libnuma.so.1.0.0 from numactl 2.0.11+git (ea3a70681c2f523fe58e1d44527f478ca76db74e). What are we missing?
Greetings!
Daniel Díaz daniel.diaz@linaro.org
On 11/27/18 6:38 PM, Daniel Díaz wrote:
Hello!
On Tue, 27 Nov 2018 at 13:22, Rafael David Tinoco rafael.tinoco@linaro.org wrote:
On 11/27/18 5:16 PM, Dan Rue wrote:
I was just looking at that - we don't run that test due to:
root@intel-corei7-64:/opt/ltp/testcases/bin# ./migrate_pages03 tst_test.c:770: CONF: require libnuma >= 2 and it's development packagesDaniel - is this something we should add?
I faced this recently in my LXC, the one generating the .deb packages. I needed libnuma1 and libnuma-dev in debian in order to have some tests successfully built.
We're currently shipping libnuma.so.1.0.0 from numactl 2.0.11+git (ea3a70681c2f523fe58e1d44527f478ca76db74e). What are we missing?
Headers ? Possibly...
Greetings!
Daniel Díaz daniel.diaz@linaro.org
On Tue, Nov 27, 2018 at 06:53:21PM -0200, Rafael David Tinoco wrote:
On 11/27/18 6:38 PM, Daniel Díaz wrote:
Hello!
On Tue, 27 Nov 2018 at 13:22, Rafael David Tinoco rafael.tinoco@linaro.org wrote:
On 11/27/18 5:16 PM, Dan Rue wrote:
I was just looking at that - we don't run that test due to:
root@intel-corei7-64:/opt/ltp/testcases/bin# ./migrate_pages03 tst_test.c:770: CONF: require libnuma >= 2 and it's development packagesDaniel - is this something we should add?
I faced this recently in my LXC, the one generating the .deb packages. I needed libnuma1 and libnuma-dev in debian in order to have some tests successfully built.
We're currently shipping libnuma.so.1.0.0 from numactl 2.0.11+git (ea3a70681c2f523fe58e1d44527f478ca76db74e). What are we missing?
Headers ? Possibly...
I can't tell you what's missing, but it's checking HAVE_NUMA_V2 which is set @ https://github.com/linux-test-project/ltp/blob/master/m4/ltp-numa.m4
Dan
On 11/27/18 6:55 PM, Dan Rue wrote:
tst_test.c:770: CONF: require libnuma >= 2 and it's development packages
static void do_setup(int argc, char *argv[]) { if (!tst_test) tst_brk(TBROK, "No tests to run");
if (tst_test->tconf_msg) <-- HERE tst_brk(TCONF, "%s", tst_test->tconf_msg);
Initial part of "tst_run_tcases()", likely solving a requirement set in the main struct for the test description.
From the test.. they use this:
#ifdef HAVE_NUMA_V2 ... #else TST_TEST_TCONF("require libnuma >= 2 and it's development packages"); #endif
And it is set from the "configure" file (autogenerated from .in files). So configure is checking existence of headers (possibly trying to compile an abvious code) and defining its existence (or not).
On 11/27/18 7:00 PM, Rafael David Tinoco wrote:
On 11/27/18 6:55 PM, Dan Rue wrote:
tst_test.c:770: CONF: require libnuma >= 2 and it's development packagesstatic void do_setup(int argc, char *argv[]) { if (!tst_test) tst_brk(TBROK, "No tests to run");
if (tst_test->tconf_msg) <-- HERE tst_brk(TCONF, "%s", tst_test->tconf_msg);
Initial part of "tst_run_tcases()", likely solving a requirement set in the main struct for the test description.
From the test.. they use this:
#ifdef HAVE_NUMA_V2 ... #else TST_TEST_TCONF("require libnuma >= 2 and it's development packages"); #endif
And it is set from the "configure" file (autogenerated from .in files). So configure is checking existence of headers (possibly trying to compile an abvious code) and defining its existence (or not).
Yep, configure does:
#include <numa.h>
int main () {
#if LIBNUMA_API_VERSION < 2 # error Required numa headers >= 2 #endif
; return 0; }
To define it or not. So we are missing headers for sure.
-
Dan Rue -
Daniel Díaz -
Rafael David Tinoco