Extended the zcopy test with a step checking aborts. It's similar to `test/nvmf/target/abort.sh`, with the difference being that here the aborted requests are executed using zero-copy. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: Ia25936b7097a94d278c467ce07c009091ea3a55a Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10799 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com> Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
56 lines
1.9 KiB
Bash
Executable File
56 lines
1.9 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
|
|
testdir=$(readlink -f $(dirname $0))
|
|
rootdir=$(readlink -f $testdir/../../..)
|
|
|
|
source $rootdir/test/common/autotest_common.sh
|
|
source $rootdir/test/nvmf/common.sh
|
|
|
|
rpc_py="$rootdir/scripts/rpc.py"
|
|
|
|
nvmftestinit
|
|
nvmfappstart
|
|
|
|
if [ "$TEST_TRANSPORT" != tcp ]; then
|
|
echo "Unsupported transport: $TEST_TRANSPORT"
|
|
exit 0
|
|
fi
|
|
|
|
# Enable zero-copy and set in-capsule data size to zero to make sure all requests are using
|
|
# zero-copy
|
|
$rpc_py nvmf_create_transport $NVMF_TRANSPORT_OPTS -c 0 --zcopy
|
|
|
|
$rpc_py nvmf_create_subsystem nqn.2016-06.io.spdk:cnode1 -a -s SPDK00000000000001 -m 10
|
|
$rpc_py nvmf_subsystem_add_listener nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT \
|
|
-a $NVMF_FIRST_TARGET_IP -s $NVMF_PORT
|
|
$rpc_py bdev_malloc_create 32 4096 -b malloc0
|
|
$rpc_py nvmf_subsystem_add_ns nqn.2016-06.io.spdk:cnode1 malloc0 -n 1
|
|
|
|
# First send IO with verification
|
|
$rootdir/test/bdev/bdevperf/bdevperf --json <(gen_nvmf_target_json) \
|
|
-t 10 -q 128 -w verify -o 8192
|
|
|
|
# Then send IO in the background while pausing/resuming the subsystem
|
|
$rootdir/test/bdev/bdevperf/bdevperf --json <(gen_nvmf_target_json) \
|
|
-t 5 -q 128 -w randrw -M 50 -o 8192 &
|
|
perfpid=$!
|
|
|
|
while kill -0 $perfpid; do
|
|
# Add the same namespace again. It'll fail, but will also pause/resume the subsystem and
|
|
# the namespace forcing the IO requests to be queued/resubmitted.
|
|
$rpc_py nvmf_subsystem_add_ns nqn.2016-06.io.spdk:cnode1 malloc0 -n 1 &> /dev/null || :
|
|
done
|
|
|
|
wait $perfpid
|
|
|
|
# Verify that aborting requests serviced through zero-copy works too
|
|
$rpc_py nvmf_subsystem_remove_ns nqn.2016-06.io.spdk:cnode1 1
|
|
$rpc_py bdev_delay_create -b malloc0 -d delay0 -r 1000000 -t 1000000 -w 1000000 -n 1000000
|
|
$rpc_py nvmf_subsystem_add_ns nqn.2016-06.io.spdk:cnode1 delay0 -n 1
|
|
|
|
$SPDK_EXAMPLE_DIR/abort -c 0x1 -t 5 -q 64 -w randrw -M 50 -l warning \
|
|
-r "trtype:$TEST_TRANSPORT adrfam:IPv4 traddr:$NVMF_FIRST_TARGET_IP trsvcid:$NVMF_PORT ns:1"
|
|
|
|
trap - SIGINT SIGTERM EXIT
|
|
nvmftestfini
|