ivampiresp/Spdk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nvmf,target: solve shutdown coredump issue

Browse Source 
It is caused by this commit:
4163626c5c

nvmf_tgt_delete_subsystem registers the poller
with function subsystem_delete_event.
subsystem_delete_event is called asynchronously,
the deletion should happen in this function.
Otherwise, with the current code,

g_subsystems_shutdown = true
TAILQ_EMPTY(&g_subsystems) = true
when subsystem_delete_event is firstly called.

If there are multiple subsystems, the logic is wrong.
Thus other subsystem will never be delete. since
we already execute shutdown_complete().

Also add related test scripts.

Change-Id: I3823563fc9e8611c11a6d798685ff64e2939842e
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
This commit is contained in:
Ziye Yang 2017-02-28 15:12:02 +08:00 committed by Ben Walker
parent b21fd06463
commit bfe2897da3
3 changed files with 45 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
app/nvmf_tgt/nvmf_tgt.c
View File

@ -69,6 +69,7 @@ subsystem_delete_event(void *arg1, void *arg2)
struct nvmf_tgt_subsystem *app_subsys = arg1;
struct spdk_nvmf_subsystem *subsystem = app_subsys->subsystem;
TAILQ_REMOVE(&g_subsystems, app_subsys, tailq);
free(app_subsys);
spdk_nvmf_delete_subsystem(subsystem);
@ -111,7 +112,6 @@ shutdown_subsystems(void)
g_subsystems_shutdown = true;
TAILQ_FOREACH_SAFE(app_subsys, &g_subsystems, tailq, tmp) {
TAILQ_REMOVE(&g_subsystems, app_subsys, tailq);
nvmf_tgt_delete_subsystem(app_subsys);
}
}
@ -286,7 +286,6 @@ nvmf_tgt_shutdown_subsystem_by_nqn(const char *nqn)
TAILQ_FOREACH_SAFE(tgt_subsystem, &g_subsystems, tailq, subsys_tmp) {
if (strcmp(tgt_subsystem->subsystem->subnqn, nqn) == 0) {
TAILQ_REMOVE(&g_subsystems, tgt_subsystem, tailq);
nvmf_tgt_delete_subsystem(tgt_subsystem);
return 0;
}

1
autotest.sh
View File

@ -118,6 +118,7 @@ run_test test/nvmf/fio/fio.sh
run_test test/nvmf/filesystem/filesystem.sh
run_test test/nvmf/discovery/discovery.sh
run_test test/nvmf/nvme_cli/nvme_cli.sh
run_test test/nvmf/shutdown/shutdown.sh
if [ $RUN_NIGHTLY -eq 1 ]; then
run_test test/nvmf/multiconnection/multiconnection.sh

43
test/nvmf/shutdown/shutdown.sh Executable file
View File

@ -0,0 +1,43 @@
#!/usr/bin/env bash
testdir=$(readlink -f $(dirname $0))
rootdir=$(readlink -f $testdir/../../..)
source $rootdir/scripts/autotest_common.sh
source $rootdir/test/nvmf/common.sh
MALLOC_BDEV_SIZE=128
MALLOC_BLOCK_SIZE=512
rpc_py="python $rootdir/scripts/rpc.py"
set -e
if ! rdma_nic_available; then
echo "no NIC for nvmf test"
exit 0
fi
timing_enter shutdown
# Start up the NVMf target in another process
$rootdir/app/nvmf_tgt/nvmf_tgt -c $testdir/../nvmf.conf &
pid=$!
trap "killprocess $pid; exit 1" SIGINT SIGTERM EXIT
waitforlisten $pid ${RPC_PORT}
# Create 12 subsystems
for i in `seq 1 12`
do
bdevs="$($rpc_py construct_malloc_bdev $MALLOC_BDEV_SIZE $MALLOC_BLOCK_SIZE)"
$rpc_py construct_nvmf_subsystem Virtual nqn.2016-06.io.spdk:cnode${i} "transport:RDMA traddr:$NVMF_FIRST_TARGET_IP trsvcid:$NVMF_PORT" '' -s SPDK${i} -n "$bdevs"
done
# Kill nvmf tgt without removing any subsystem to check whether it can shutdown correctly
rm -f ./local-job0-0-verify.state
trap - SIGINT SIGTERM EXIT
killprocess $pid
timing_exit shutdown