For unification number of nodes is set based on environment. OS caches were not reliably dropped after execution of 'dd' on DATAFILE. And since 'dd' was executed in root cpuset the caches were spread among all available memory nodes. Because on consecutive reading of DATAFILE in cpuset_mem_hog didn't increase memory consumption. This situation caused random failures of the testcase.
In accordance to Documentation/sysctl/vm.txt to reliably drop caches we should invoke 'sync' before 'echo N > /proc/sys/vm/drop_caches'. Signed-off-by: Stanislav Kholmanskikh <[email protected]> --- .../cpuset_memory_spread_testset.sh | 14 +++----------- 1 files changed, 3 insertions(+), 11 deletions(-) diff --git a/testcases/kernel/controllers/cpuset/cpuset_memory_spread_test/cpuset_memory_spread_testset.sh b/testcases/kernel/controllers/cpuset/cpuset_memory_spread_test/cpuset_memory_spread_testset.sh index a357e22..ffce419 100755 --- a/testcases/kernel/controllers/cpuset/cpuset_memory_spread_test/cpuset_memory_spread_testset.sh +++ b/testcases/kernel/controllers/cpuset/cpuset_memory_spread_test/cpuset_memory_spread_testset.sh @@ -32,8 +32,8 @@ export TST_COUNT=1 exit_status=0 # must >= 3 for: 1-$((nr_mems-2)) -nr_cpus=4 -nr_mems=3 +nr_cpus=$NR_CPUS +nr_mems=$N_NODES # In general, the cache hog will use more than 10000 kb slab space on the nodes # on which it is running. The other nodes' slab space has littler change.(less @@ -223,6 +223,7 @@ general_memory_spread_test() fi # we'd better drop the caches before we test page cache. + sync /bin/echo 3 > /proc/sys/vm/drop_caches 2> $CPUSET_TMP/stderr if [ $? -ne 0 ]; then cpuset_log_error $CPUSET_TMP/stderr @@ -230,9 +231,6 @@ general_memory_spread_test() return 1 fi - # wait for droping the cache - sleep 10 - get_memsinfo /bin/kill -s SIGUSR1 $test_pid read exit_num < $FIFO @@ -344,12 +342,6 @@ if [ $? -ne 0 ]; then exit 1 fi -# drop page caches -/bin/echo 1 > /proc/sys/vm/drop_caches - -# wait for droping caches -sleep 10 - mkfifo $FIFO if [ $? -ne 0 ]; then rm -f DATAFILE -- 1.7.1 ------------------------------------------------------------------------------ October Webinars: Code for Performance Free Intel webinars can help you accelerate application performance. Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60134791&iu=/4140/ostg.clktrk _______________________________________________ Ltp-list mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ltp-list
