lib/blob: prevent overflowing uint32_t when clearing cluster data

When blob is deleted or reduced in size, the trimmed clusters
are cleared using bs_batch_clear_dev().

blob_persist_clear_clusters() iterated over the clusters
and batched each contigous set of LBA.

The lba_count is of the uint32_t type, long enough
contigous set of LBA could overflow it.
As a result that range would only be cleared with the
overflown value.

This patch prevents that overflow from occuring by
verifying against UINT32_MAX.

This is already addressed with an API change in SPDK 21.10:
(f01146ae)blob: use uint64_t for unmap and write_zeroes lba count

Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I17282a2a81587c26b7e89a74dcc6a9c8a017ce15
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9971
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>

CHANGELOG.md

@@ -1,6 +1,39 @@
 # Changelog
 
-## v21.01: (Upcoming Release)
+## v21.01.2: (Upcoming Release)
+
+## v21.01.1:
+
+### dpdk
+
+Added `rte_ethdev` and `rte_net` dependencies for all builds with DPDK.
+
+Fixed compatibility issues with DPDK 19.11.
+
+### iscsi
+
+A security vulnerability has been identified and fixed in the SPDK iSCSI target.
+A TEXT PDU with no data, but CONTINUE flag set, would result in a NULL pointer dereference
+and crash the SPDK iSCSI target process. All users of the SPDK iSCSI target
+are recommended to update. All SPDK versions <= v21.01 are affected.
+
+### nbd
+
+Fixed kernel hang when bdev is removed by always setting NBD_SET_TIMEOUT.
+
+### nvme
+
+Fixed segfault when removing qpair when transport connection fails (issue #1777).
+
+### ocssd
+
+Fixed the bug that no media event is pushed to the target bdev.
+
+### sock
+
+Added `enable_quickack` and `enable_placement_id` when saving JSON configuration.
+
+## v21.01: Dynamic scheduler, vfio-user, ZNS Zone Append
 
 ### bdev
 
@@ -75,6 +108,10 @@ The `--pci-blacklist` command line option has been deprecated, replaced with
 The `--pci-whitelist/-W` command line options have been deprecated, replaced with
 `--pci-allowed/-A`.
 
+Added new experimental `dynamic` scheduler that rebalances idle threads, adjusts CPU frequency
+using dpdk_governor and turns idle reactor cores to interrupt mode. Please see
+[scheduler documentation](https://www.spdk.io/doc/scheduler.html) for details.
+
 ## ioat
 
 The PCI BDF whitelist option has been removed from the `ioat_scan_accel_engine` RPC.

CONFIG

@@ -121,7 +121,7 @@ CONFIG_VHOST=y
 CONFIG_VIRTIO=y
 
 # Build custom vfio-user transport for NVMf target and NVMe initiator.
-CONFIG_VFIO_USER=y
+CONFIG_VFIO_USER=n
 
 # Build with PMDK backends
 CONFIG_PMDK=n

app/spdk_top/spdk_top.c

@@ -2027,6 +2027,7 @@ show_thread(uint8_t current_page)
 
 	get_data();
 
+	assert(thread_number < g_threads_stats.threads.threads_count);
 	for (i = 0; i < g_threads_stats.threads.threads_count; i++) {
 		thread_info[i] = &g_threads_stats.threads.thread_info[i];
 	}
@@ -2067,6 +2068,7 @@ show_core(uint8_t current_page)
 
 	get_data();
 
+	assert(core_number < g_cores_stats.cores.cores_count);
 	for (i = 0; i < g_cores_stats.cores.cores_count; i++) {
 		core_info[i] = &g_cores_stats.cores.core[i];
 	}
@@ -2178,6 +2180,7 @@ show_poller(uint8_t current_page)
 	get_data();
 
 	prepare_poller_data(current_page, pollers, &count, current_page);
+	assert(poller_number < count);
 
 	poller_win = newwin(POLLER_WIN_HEIGHT, POLLER_WIN_WIDTH,
 			    (g_max_row - poller_counter) / 2, (g_max_col - POLLER_WIN_HOR_POS) / 2);
@@ -2300,6 +2303,7 @@ show_stats(void)
 			} else {
 				active_tab = THREADS_TAB;
 			}
+			g_selected_row = 0;
 			current_page = 0;
 			switch_tab(active_tab);
 			break;

autobuild.sh

@@ -12,6 +12,7 @@ rootdir=$(readlink -f $(dirname $0))
 
 source "$1"
 source "$rootdir/test/common/autotest_common.sh"
+source "$rootdir/scripts/common.sh"
 
 out=$output_dir
 if [ -n "$SPDK_TEST_NATIVE_DPDK" ]; then
@@ -50,7 +51,10 @@ function ocf_precompile() {
 function build_native_dpdk() {
 	local external_dpdk_dir
 	local external_dpdk_base_dir
+	local gcc_version
 
+	gcc_version=$(gcc -dumpversion)
+	gcc_version=${gcc_version//./}
 	external_dpdk_dir="$SPDK_RUN_EXTERNAL_DPDK"
 	external_dpdk_base_dir="$(dirname $external_dpdk_dir)"
 
@@ -64,9 +68,17 @@ function build_native_dpdk() {
 	git clone --branch $SPDK_TEST_NATIVE_DPDK --depth 1 http://dpdk.org/git/dpdk "$external_dpdk_base_dir"
 	git -C "$external_dpdk_base_dir" log --oneline -n 5
 
-	dpdk_cflags="-fPIC -g -Werror -fcommon"
+	dpdk_cflags="-fPIC -g -fcommon"
 	dpdk_ldflags=""
 
+	if [[ $gcc_version -ge 5 ]]; then
+		dpdk_cflags+=" -Werror"
+	fi
+
+	if [[ $gcc_version -ge 10 ]]; then
+		dpdk_cflags+=" -Wno-stringop-overflow"
+	fi
+
 	# the drivers we use
 	# net/i40e driver is not really needed by us, but it's built as a workaround
 	# for DPDK issue: https://bugs.dpdk.org/show_bug.cgi?id=576
@@ -112,19 +124,22 @@ function build_native_dpdk() {
 
 	cd $external_dpdk_base_dir
 	if [ "$(uname -s)" = "Linux" ]; then
-		dpdk_cflags+=" -Wno-stringop-overflow"
 		# Fix for freeing device if not kernel driver configured.
 		# TODO: Remove once this is merged in upstream DPDK
 		if grep "20.08.0" $external_dpdk_base_dir/VERSION; then
 			wget https://github.com/spdk/dpdk/commit/64f1ced13f974e8b3d46b87c361a09eca68126f9.patch -O dpdk-pci.patch
 			wget https://github.com/spdk/dpdk/commit/c2c273d5c8fbf673623b427f8f4ab5af5ddf0e08.patch -O dpdk-qat.patch
-		else
+		elif grep "20.11\|21.02" $external_dpdk_base_dir/VERSION; then
 			wget https://github.com/karlatec/dpdk/commit/3219c0cfc38803aec10c809dde16e013b370bda9.patch -O dpdk-pci.patch
 			wget https://github.com/karlatec/dpdk/commit/adf8f7638de29bc4bf9ba3faf12bbdae73acda0c.patch -O dpdk-qat.patch
+		else
+			wget https://github.com/karlatec/dpdk/commit/6fd2fa906ffdcee04e6ce5da40e61cb841be9827.patch -O dpdk-qat.patch
 		fi
 		git config --local user.name "spdk"
 		git config --local user.email "nomail@all.com"
-		git am dpdk-pci.patch
+		if [[ -f dpdk-pci.patch ]]; then
+			git am dpdk-pci.patch
+		fi
 		git am dpdk-qat.patch
 	fi
 
@@ -222,11 +237,15 @@ function test_make_uninstall() {
 }
 
 function build_doc() {
+	local doxygenv
+	doxygenv=$(doxygen --version)
+
 	$MAKE -C "$rootdir"/doc --no-print-directory $MAKEFLAGS &> "$out"/doxygen.log
 	if [ -s "$out"/doxygen.log ]; then
 		cat "$out"/doxygen.log
 		echo "Doxygen errors found!"
-		exit 1
+		eq "$doxygenv" 1.8.20 || exit 1
+		echo "Doxygen $doxygenv detected, all warnings are potentially false positives, continuing the test"
 	fi
 	if hash pdflatex 2> /dev/null; then
 		$MAKE -C "$rootdir"/doc/output/latex --no-print-directory $MAKEFLAGS &>> "$out"/doxygen.log
@@ -238,7 +257,8 @@ function build_doc() {
 	fi
 	$MAKE -C "$rootdir"/doc --no-print-directory $MAKEFLAGS clean &>> "$out"/doxygen.log
 	if [ -s "$out"/doxygen.log ]; then
-		rm "$out"/doxygen.log
+		# Save the log as an artifact in case we are working with potentially broken version
+		eq "$doxygenv" 1.8.20 || rm "$out"/doxygen.log
 	fi
 	rm -rf "$rootdir"/doc/output
 }

autopackage.sh

@@ -28,7 +28,7 @@ if [ $(git status --porcelain --ignore-submodules | wc -l) -ne 0 ]; then
 fi
 timing_exit porcelain_check
 
-if [[ $RUN_NIGHTLY -eq 0 && $SPDK_TEST_RELEASE_BUILD -eq 0 ]]; then
+if [[ $RUN_NIGHTLY -eq 0 ]]; then
 	timing_finish
 	exit 0
 fi

autorun.sh

@@ -4,7 +4,8 @@ set -e
 
 rootdir=$(readlink -f $(dirname $0))
 
-conf=~/autorun-spdk.conf
+default_conf=~/autorun-spdk.conf
+conf=${1:-${default_conf}}
 
 # If the configuration of tests is not provided, no tests will be carried out.
 if [[ ! -f $conf ]]; then

autotest.sh

@@ -150,16 +150,7 @@ timing_enter afterboot
 ./scripts/setup.sh
 timing_exit afterboot
 
-timing_enter nvmf_setup
-rdma_device_init
-timing_exit nvmf_setup
-
 if [[ $SPDK_TEST_CRYPTO -eq 1 || $SPDK_TEST_REDUCE -eq 1 ]]; then
-	# Make sure that memory is distributed across all NUMA nodes - by default, all goes to
-	# node0, but if QAT devices are attached to a different node, all of their VFs will end
-	# up under that node too and memory needs to be available there for the tests.
-	CLEAR_HUGE=yes HUGE_EVEN_ALLOC=yes ./scripts/setup.sh
-	./scripts/setup.sh status
 	if [[ $SPDK_TEST_USE_IGB_UIO -eq 1 ]]; then
 		./scripts/qat_setup.sh igb_uio
 	else
@@ -196,8 +187,18 @@ if [ $SPDK_RUN_FUNCTIONAL_TEST -eq 1 ]; then
 		run_test "blockdev_general" test/bdev/blockdev.sh
 		run_test "bdev_raid" test/bdev/bdev_raid.sh
 		run_test "bdevperf_config" test/bdev/bdevperf/test_config.sh
-		if [[ $(uname -s) == Linux ]]; then
+	fi
-			run_test "spdk_dd" test/dd/dd.sh
+
+	if [[ $(uname -s) == Linux ]]; then
+		if [[ $SPDK_TEST_BLOCKDEV -eq 1 || $SPDK_TEST_URING -eq 1 ]]; then
+			# The crypto job also includes the SPDK_TEST_BLOCKDEV in its configuration hence the
+			# dd tests are executed there as well. However, these tests can take a significant
+			# amount of time to complete (up to 4min) on a physical system leading to a potential
+			# job timeout. Avoid that by skipping these tests - this should not affect the coverage
+			# since dd tests are still run as part of the vg jobs.
+			if [[ $SPDK_TEST_CRYPTO -eq 0 ]]; then
+				run_test "spdk_dd" test/dd/dd.sh
+			fi
 		fi
 	fi
 
@@ -247,11 +248,17 @@ if [ $SPDK_RUN_FUNCTIONAL_TEST -eq 1 ]; then
 		# The NVMe-oF run test cases are split out like this so that the parser that compiles the
 		# list of all tests can properly differentiate them. Please do not merge them into one line.
 		if [ "$SPDK_TEST_NVMF_TRANSPORT" = "rdma" ]; then
+			timing_enter rdma_setup
+			rdma_device_init
+			timing_exit rdma_setup
 			run_test "nvmf_rdma" ./test/nvmf/nvmf.sh --transport=$SPDK_TEST_NVMF_TRANSPORT
-			run_test "spdkcli_nvmf_rdma" ./test/spdkcli/nvmf.sh
+			run_test "spdkcli_nvmf_rdma" ./test/spdkcli/nvmf.sh --transport=$SPDK_TEST_NVMF_TRANSPORT
 		elif [ "$SPDK_TEST_NVMF_TRANSPORT" = "tcp" ]; then
+			timing_enter tcp_setup
+			tcp_device_init
+			timing_exit tcp_setup
 			run_test "nvmf_tcp" ./test/nvmf/nvmf.sh --transport=$SPDK_TEST_NVMF_TRANSPORT
-			run_test "spdkcli_nvmf_tcp" ./test/spdkcli/nvmf.sh
+			run_test "spdkcli_nvmf_tcp" ./test/spdkcli/nvmf.sh --transport=$SPDK_TEST_NVMF_TRANSPORT
 			run_test "nvmf_identify_passthru" test/nvmf/target/identify_passthru.sh --transport=$SPDK_TEST_NVMF_TRANSPORT
 			run_test "nvmf_dif" test/nvmf/target/dif.sh
 		elif [ "$SPDK_TEST_NVMF_TRANSPORT" = "fc" ]; then

configure

@@ -454,36 +454,22 @@ else
 fi
 BUILD_CMD+=(-I/usr/local/include -L/usr/local/lib)
 
-function set_os_id_version() {
-	if [[ -f /etc/os-release ]]; then
-		source /etc/os-release
-	fi
-
-	OSID=$ID
-	OSVERSION=$VERSION_ID
-
-	echo "OS-ID: $OSID | OS-Version: $OSVERSION"
-}
-
 if [[ "${CONFIG[VFIO_USER]}" = "y" ]]; then
-	set_os_id_version
 
-	if [[ $arch != x86_64* ]] || [[ $sys_name == "FreeBSD" ]]; then
+	if ! hash cmake; then
-		echo "Non x86_64 and Linux platform, disable CONFIG_VFIO_USER"
+		echo "ERROR: --with-vfio-user requires cmake"
-		CONFIG[VFIO_USER]="n"
+		echo "Please install then re-run this script"
-	# disable tests on ubuntu16 due to lack of macro definition in pci_regs.h
+		exit 1
-	elif [[ $OSID == "ubuntu" ]] && [[ $OSVERSION == "16.04" ]]; then
+	fi
-		echo "ubuntu16 OS, disable CONFIG_VFIO_USER"
+	if [[ ! -d /usr/include/json-c ]] && [[ ! -d /usr/local/include/json-c ]]; then
-		CONFIG[VFIO_USER]="n"
+		echo "ERROR: --with-vfio-user requires json-c-devel"
-	elif ! hash cmake; then
+		echo "Please install then re-run this script"
-		echo "cmake not installed, disable CONFIG_VFIO_USER"
+		exit 1
-		CONFIG[VFIO_USER]="n"
+	fi
-	elif [[ ! -d /usr/include/json-c ]] && [[ ! -d /usr/local/include/json-c ]]; then
+	if [[ ! -e /usr/include/cmocka.h ]] && [[ ! -e /usr/local/include/cmocka.h ]]; then
-		echo "json-c-devel not installed, disable CONFIG_VFIO_USER"
+		echo "ERROR: --with-vfio-user requires libcmocka-devel"
-		CONFIG[VFIO_USER]="n"
+		echo "Please install then re-run this script"
-	elif [[ ! -e /usr/include/cmocka.h ]] && [[ ! -e /usr/local/include/cmocka.h ]]; then
+		exit 1
-		echo "libcmocka-devel not installed, disable CONFIG_VFIO_USER"
-		CONFIG[VFIO_USER]="n"
 	fi
 fi
 

doc/Doxyfile

@@ -835,6 +835,7 @@ INPUT                  += \
                          peer_2_peer.md \
                          pkgconfig.md \
                          porting.md \
+                         scheduler.md \
                          shfmt.md \
                          spdkcli.md \
                          spdk_top.md \

doc/general.md

@@ -1,5 +1,6 @@
 # General Information {#general}
 
 - @subpage event
+- @subpage scheduler
 - @subpage logical_volumes
 - @subpage accel_fw

doc/jsonrpc.md

@@ -78,7 +78,7 @@ A detailed description of each RPC method and its parameters is also available.
 scripts/rpc.py bdev_nvme_attach_controller --help
 ~~~
 
-### Generate JSON-RPC methods for current configuration
+### Generate JSON-RPC methods for current configuration {#jsonrpc_generate}
 
 An initial configuration can be specified for an SPDK application via the '-c' command line parameter.
 The configuration file is a JSON file containing all of the JSON-RPC method invocations necessary
@@ -206,6 +206,18 @@ Finally, call the rpc.py script with '--plugin' parameter to provide above pytho
 ./scripts/rpc.py --plugin rpc_plugin bdev_example_create 10 4096
 ~~~
 
+### Converting from legacy configuration {#jsonrpc_convert}
+
+Starting with SPDK 20.10, legacy configuration file support has been removed.
+Users with extensive configuration files already running in SPDK application,
+can [generate JSON-RPC for current configuration](@ref jsonrpc_generate).
+
+If binary for deploying the application is unavailable, the legacy configuration
+file can be converted to JSON-RPC using python tool:
+~~~
+./scripts/config_converter.py < config.ini > config.json
+~~~
+
 # App Framework {#jsonrpc_components_app}
 
 ## spdk_kill_instance {#rpc_spdk_kill_instance}

doc/nvmf.md

@@ -116,10 +116,6 @@ An NVMe over Fabrics target can be configured using JSON RPCs.
 The basic RPCs needed to configure the NVMe-oF subsystem are detailed below. More information about
 working with NVMe over Fabrics specific RPCs can be found on the @ref jsonrpc_components_nvmf_tgt RPC page.
 
-Using .ini style configuration files for configuration of the NVMe-oF target is deprecated and should
-be replaced with JSON based RPCs. .ini style configuration files can be converted to json format by way
-of the new script `scripts/config_converter.py`.
-
 ## FC transport support {#nvmf_fc_transport}
 
 To build nvmf_tgt with the FC transport, there is an additional FC LLD (Low Level Driver) code dependency.

doc/scheduler.md

@@ -0,0 +1,82 @@
+# Scheduler {#scheduler}
+
+SPDK's event/application framework (`lib/event`) now supports scheduling of
+lightweight threads. Schedulers are provided as plugins, called
+implementations. A default implementation is provided, but users may wish to
+write their own scheduler to integrate into broader code frameworks or meet
+their performance needs.
+
+This feature should be considered experimental and is disabled by default. When
+enabled, the scheduler framework gathers data for each spdk thread and reactor
+and passes it to a scheduler implementation to perform one of the following
+actions.
+
+## Actions
+
+### Move a thread
+
+`spdk_thread`s can be moved to another reactor. Schedulers can examine the
+suggested cpu_mask value for each lightweight thread to see if the user has
+requested specific reactors, or choose a reactor using whatever algorithm they
+deem fit.
+
+### Switch reactor mode
+
+Reactors by default run in a mode that constantly polls for new actions for the
+most efficient processing. Schedulers can switch a reactor into a mode that
+instead waits for an event on a file descriptor. On Linux, this is implemented
+using epoll. This results in reduced CPU usage but may be less responsive when
+events occur. A reactor cannot enter this mode if any `spdk_threads` are
+currently scheduled to it. This limitation is expected to be lifted in the
+future, allowing `spdk_threads` to enter interrupt mode.
+
+### Set frequency of CPU core
+
+The frequency of CPU cores can be modified by the scheduler in response to
+load. Only CPU cores that match the application cpu_mask may be modified. The
+mechanism for controlling CPU frequency is pluggable and the default provided
+implementation is called `dpdk_governor`, based on the `rte_power` library from
+DPDK.
+
+#### Known limitation
+
+When SMT (Hyperthreading) is enabled the two logical CPU cores sharing a single
+physical CPU core must run at the same frequency. If one of two of such logical
+CPU cores is outside the application cpu_mask, the policy and frequency on that
+core has to be managed by the administrator.
+
+## Scheduler implementations
+
+The scheduler in use may be controlled by JSON-RPC. Please use the
+[framework_set_scheduler](jsonrpc.md/#rpc_framework_set_scheduler) RPC to
+switch between schedulers or change their options.
+
+[spdk_top](spdk_top.md#spdk_top) is a useful tool to observe the behavior of
+schedulers in different scenarios and workloads.
+
+### static [default]
+
+The `static` scheduler is the default scheduler and does no dynamic scheduling.
+Lightweight threads are distributed round-robin among reactors, respecting
+their requested cpu_mask, and then they are never moved. This is equivalent to
+the previous behavior of the SPDK event/application framework.
+
+### dynamic
+
+The `dynamic` scheduler is designed for power saving and reduction of CPU
+utilization, especially in cases where workloads show large variations over
+time.
+
+Active threads are distributed equally among reactors, taking cpu_mask into
+account. All idle threads are moved to the main core. Once an idle thread becomes
+active, it is redistributed again.
+
+When a reactor has no scheduled `spdk_thread`s it is switched into interrupt
+mode and stops actively polling. After enough threads become active, the
+reactor is switched back into poll mode and threads are assigned to it again.
+
+The main core can contain active threads only when their execution time does
+not exceed the sum of all idle threads. When no active threads are present on
+the main core, the frequency of that CPU core will decrease as the load
+decreases. All CPU cores corresponding to the other reactors remain at maximum
+frequency.

examples/nvme/fio_plugin/fio_plugin.c

@@ -1088,13 +1088,11 @@ spdk_fio_get_zoned_model(struct thread_data *td, struct fio_file *f, enum zbd_zo
 	struct spdk_fio_qpair *fio_qpair = NULL;
 	const struct spdk_nvme_zns_ns_data *zns_data = NULL;
 
-	*model = ZBD_IGNORE;
-
 	if (f->filetype != FIO_TYPE_FILE && \
 	    f->filetype != FIO_TYPE_BLOCK && \
 	    f->filetype != FIO_TYPE_CHAR) {
-		log_info("spdk/nvme: ignoring filetype: %d\n", f->filetype);
+		log_info("spdk/nvme: unsupported filetype: %d\n", f->filetype);
-		return 0;
+		return -EINVAL;
 	}
 
 	fio_qpair = get_fio_qpair(fio_thread, f);

examples/vmd/led/Makefile

@@ -39,6 +39,6 @@ APP = led
 
 C_SRCS := led.c
 
-SPDK_LIB_LIST = vmd log
+SPDK_LIB_LIST = vmd log util
 
 include $(SPDK_ROOT_DIR)/mk/spdk.app.mk

examples/vmd/lsvmd/Makefile

@@ -39,6 +39,6 @@ APP = lsvmd
 
 C_SRCS := lsvmd.c
 
-SPDK_LIB_LIST = vmd log
+SPDK_LIB_LIST = vmd log util
 
 include $(SPDK_ROOT_DIR)/mk/spdk.app.mk

include/spdk/pci_ids.h

@@ -52,6 +52,7 @@ extern "C" {
 #define SPDK_PCI_VID_VIRTIO		0x1af4
 #define SPDK_PCI_VID_CNEXLABS		0x1d1d
 #define SPDK_PCI_VID_VMWARE		0x15ad
+#define SPDK_PCI_VID_REDHAT		0x1b36
 
 #define SPDK_PCI_CLASS_ANY_ID		0xffffff
 /**

include/spdk/version.h

@@ -54,7 +54,7 @@
  * Patch level is incremented on maintenance branch releases and reset to 0 for each
  * new major.minor release.
  */
-#define SPDK_VERSION_PATCH	0
+#define SPDK_VERSION_PATCH	2
 
 /**
  * Version string suffix.

lib/blob/blobstore.c

@@ -1765,7 +1765,7 @@ blob_persist_clear_clusters(spdk_bs_sequence_t *seq, struct spdk_blob_persist_ct
 	spdk_bs_batch_t			*batch;
 	size_t				i;
 	uint64_t			lba;
-	uint32_t			lba_count;
+	uint64_t			lba_count;
 
 	/* Clusters don't move around in blobs. The list shrinks or grows
 	 * at the end, but no changes ever occur in the middle of the list.
@@ -1778,9 +1778,10 @@ blob_persist_clear_clusters(spdk_bs_sequence_t *seq, struct spdk_blob_persist_ct
 	lba_count = 0;
 	for (i = blob->active.num_clusters; i < blob->active.cluster_array_size; i++) {
 		uint64_t next_lba = blob->active.clusters[i];
-		uint32_t next_lba_count = bs_cluster_to_lba(bs, 1);
+		uint64_t next_lba_count = bs_cluster_to_lba(bs, 1);
 
-		if (next_lba > 0 && (lba + lba_count) == next_lba) {
+		if (next_lba > 0 && (lba + lba_count) == next_lba &&
+		    (lba_count + next_lba_count <= UINT32_MAX)) {
 			/* This cluster is contiguous with the previous one. */
 			lba_count += next_lba_count;
 			continue;
@@ -4340,6 +4341,12 @@ bs_load_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
 	}
 	ctx->bs->md_start = ctx->super->md_start;
 	ctx->bs->md_len = ctx->super->md_len;
+	rc = spdk_bit_array_resize(&ctx->bs->open_blobids, ctx->bs->md_len);
+	if (rc < 0) {
+		bs_load_ctx_fail(ctx, -ENOMEM);
+		return;
+	}
+
 	ctx->bs->total_data_clusters = ctx->bs->total_clusters - spdk_divide_round_up(
 					       ctx->bs->md_start + ctx->bs->md_len, ctx->bs->pages_per_cluster);
 	ctx->bs->super_blob = ctx->super->super_blob;
@@ -4911,7 +4918,7 @@ spdk_bs_init(struct spdk_bs_dev *dev, struct spdk_bs_opts *o,
 
 	lba = num_md_lba;
 	while (lba < ctx->bs->dev->blockcnt) {
-		lba_count = spdk_min(UINT32_MAX, ctx->bs->dev->blockcnt - lba);
+		lba_count = spdk_min(UINT32_MAX - 127, ctx->bs->dev->blockcnt - lba);
 		switch (opts.clear_method) {
 		case BS_CLEAR_WITH_UNMAP:
 			/* Trim data clusters */

lib/env_dpdk/env.mk

@@ -51,7 +51,7 @@ DPDK_INC := -I$(DPDK_INC_DIR)
 DPDK_LIB_LIST = rte_eal rte_mempool rte_ring rte_mbuf rte_pci rte_bus_pci rte_mempool_ring
 
 ifeq ($(OS),Linux)
-DPDK_LIB_LIST += rte_power
+DPDK_LIB_LIST += rte_power rte_ethdev rte_net
 endif
 
 # DPDK 20.05 eal dependency
@@ -101,7 +101,7 @@ endif
 LINK_HASH=n
 
 ifeq ($(CONFIG_VHOST),y)
-DPDK_LIB_LIST += rte_vhost rte_net
+DPDK_LIB_LIST += rte_vhost
 LINK_HASH=y
 ifneq ($(DPDK_FRAMEWORK),y)
 DPDK_LIB_LIST += rte_cryptodev

lib/env_dpdk/memory.c

@@ -1238,7 +1238,11 @@ vtophys_iommu_device_event(const char *device_name,
 		RTE_DEV_FOREACH(dev, "bus=pci", &dev_iter) {
 			if (strcmp(dev->name, device_name) == 0) {
 				struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev);
+#if RTE_VERSION < RTE_VERSION_NUM(20, 11, 0, 0)
+				if (pci_dev->kdrv == RTE_KDRV_VFIO) {
+#else
 				if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO) {
+#endif
 					/* This is a new PCI device using vfio */
 					g_vfio.device_ref++;
 				}
@@ -1267,7 +1271,11 @@ vtophys_iommu_device_event(const char *device_name,
 		RTE_DEV_FOREACH(dev, "bus=pci", &dev_iter) {
 			if (strcmp(dev->name, device_name) == 0) {
 				struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev);
+#if RTE_VERSION < RTE_VERSION_NUM(20, 11, 0, 0)
+				if (pci_dev->kdrv == RTE_KDRV_VFIO) {
+#else
 				if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO) {
+#endif
 					/* This is a PCI device using vfio */
 					g_vfio.device_ref--;
 				}
@@ -1362,7 +1370,11 @@ vtophys_iommu_init(void)
 	RTE_DEV_FOREACH(dev, "bus=pci", &dev_iter) {
 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev);
 
+#if RTE_VERSION < RTE_VERSION_NUM(20, 11, 0, 0)
+		if (pci_dev->kdrv == RTE_KDRV_VFIO) {
+#else
 		if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO) {
+#endif
 			/* This is a PCI device using vfio */
 			g_vfio.device_ref++;
 		}

lib/env_dpdk/pci.c

@@ -37,6 +37,7 @@
 #include <rte_devargs.h>
 #include "spdk/env.h"
 #include "spdk/log.h"
+#include "spdk/string.h"
 
 #define SYSFS_PCI_DRIVERS	"/sys/bus/pci/drivers"
 
@@ -63,6 +64,28 @@ static TAILQ_HEAD(, spdk_pci_device) g_pci_hotplugged_devices =
 	TAILQ_HEAD_INITIALIZER(g_pci_hotplugged_devices);
 static TAILQ_HEAD(, spdk_pci_driver) g_pci_drivers = TAILQ_HEAD_INITIALIZER(g_pci_drivers);
 
+struct env_devargs {
+	struct rte_bus	*bus;
+	char		name[128];
+	uint64_t	allowed_at;
+	TAILQ_ENTRY(env_devargs) link;
+};
+static TAILQ_HEAD(, env_devargs) g_env_devargs = TAILQ_HEAD_INITIALIZER(g_env_devargs);
+
+static struct env_devargs *
+find_env_devargs(struct rte_bus *bus, const char *name)
+{
+	struct env_devargs *da;
+
+	TAILQ_FOREACH(da, &g_env_devargs, link) {
+		if (bus == da->bus && !strcmp(name, da->name)) {
+			return da;
+		}
+	}
+
+	return NULL;
+}
+
 static int
 map_bar_rte(struct spdk_pci_device *device, uint32_t bar,
 	    void **mapped_addr, uint64_t *phys_addr, uint64_t *size)
@@ -448,6 +471,39 @@ pci_device_init(struct rte_pci_driver *_drv,
 	return 0;
 }
 
+static void
+set_allowed_at(struct rte_devargs *rte_da, uint64_t tsc)
+{
+	struct env_devargs *env_da;
+
+	env_da = find_env_devargs(rte_da->bus, rte_da->name);
+	if (env_da == NULL) {
+		env_da = calloc(1, sizeof(*env_da));
+		if (env_da == NULL) {
+			SPDK_ERRLOG("could not set_allowed_at for device %s\n", rte_da->name);
+			return;
+		}
+		env_da->bus = rte_da->bus;
+		spdk_strcpy_pad(env_da->name, rte_da->name, sizeof(env_da->name), 0);
+		TAILQ_INSERT_TAIL(&g_env_devargs, env_da, link);
+	}
+
+	env_da->allowed_at = tsc;
+}
+
+static uint64_t
+get_allowed_at(struct rte_devargs *rte_da)
+{
+	struct env_devargs *env_da;
+
+	env_da = find_env_devargs(rte_da->bus, rte_da->name);
+	if (env_da) {
+		return env_da->allowed_at;
+	} else {
+		return 0;
+	}
+}
+
 int
 pci_device_fini(struct rte_pci_device *_dev)
 {
@@ -468,7 +524,7 @@ pci_device_fini(struct rte_pci_device *_dev)
 
 	/* remove our allowed_at option */
 	if (_dev->device.devargs) {
-		_dev->device.devargs->data = NULL;
+		set_allowed_at(_dev->device.devargs, 0);
 	}
 
 	assert(!dev->internal.removed);
@@ -541,8 +597,8 @@ scan_pci_bus(bool delay_init)
 			rte_dev->device.devargs = da;
 		}
 
-		if (da->data) {
+		if (get_allowed_at(da)) {
-			uint64_t allowed_at = (uint64_t)(uintptr_t)da->data;
+			uint64_t allowed_at = get_allowed_at(da);
 
 			/* this device was seen by spdk before... */
 			if (da->policy == RTE_DEV_BLOCKED && allowed_at <= now) {
@@ -554,10 +610,10 @@ scan_pci_bus(bool delay_init)
 
 			if (delay_init) {
 				da->policy = RTE_DEV_BLOCKED;
-				da->data = (void *)(now + 2 * spdk_get_ticks_hz());
+				set_allowed_at(da, now + 2 * spdk_get_ticks_hz());
 			} else {
 				da->policy = RTE_DEV_ALLOWED;
-				da->data = (void *)(uintptr_t)now;
+				set_allowed_at(da, now);
 			}
 		}
 	}
@@ -638,8 +694,8 @@ spdk_pci_device_attach(struct spdk_pci_driver *driver,
 
 	rte_dev = dev->dev_handle;
 	da = rte_dev->device.devargs;
-	if (da && da->data) {
+	if (da && get_allowed_at(da)) {
-		da->data = (void *)(uintptr_t)spdk_get_ticks();
+		set_allowed_at(da, spdk_get_ticks());
 		da->policy = RTE_DEV_ALLOWED;
 	}
 

lib/event/reactor.c

@@ -1010,8 +1010,10 @@ void
 spdk_reactors_start(void)
 {
 	struct spdk_reactor *reactor;
+	struct spdk_cpuset tmp_cpumask = {};
 	uint32_t i, current_core;
 	int rc;
+	char thread_name[32];
 
 	g_rusage_period = (CONTEXT_SWITCH_MONITOR_PERIOD * spdk_get_ticks_hz()) / SPDK_SEC_TO_USEC;
 	g_reactor_state = SPDK_REACTOR_STATE_RUNNING;
@@ -1030,6 +1032,14 @@ spdk_reactors_start(void)
 				assert(false);
 				return;
 			}
+
+			/* For now, for each reactor spawn one thread. */
+			snprintf(thread_name, sizeof(thread_name), "reactor_%u", reactor->lcore);
+
+			spdk_cpuset_zero(&tmp_cpumask);
+			spdk_cpuset_set_cpu(&tmp_cpumask, i, true);
+
+			spdk_thread_create(thread_name, &tmp_cpumask);
 		}
 		spdk_cpuset_set_cpu(&g_reactor_core_mask, i, true);
 	}

lib/iscsi/param.c

@@ -315,6 +315,16 @@ iscsi_parse_params(struct iscsi_param **params, const uint8_t *data,
 	char *p;
 	int i;
 
+	/* Spec does not disallow TEXT PDUs with zero length, just return
+	 * immediately in that case, since there is no param data to parse
+	 * and any existing partial parameter would remain as-is.
+	 */
+	if (len == 0) {
+		return 0;
+	}
+
+	assert(data != NULL);
+
 	/* strip the partial text parameters if previous PDU have C enabled */
 	if (partial_parameter && *partial_parameter) {
 		for (i = 0; i < len && data[i] != '\0'; i++) {

lib/nbd/nbd.c

@@ -50,6 +50,7 @@
 #define GET_IO_LOOP_COUNT		16
 #define NBD_BUSY_WAITING_MS		1000
 #define NBD_BUSY_POLLING_INTERVAL_US	20000
+#define NBD_IO_TIMEOUT_S		60
 
 enum nbd_io_state_t {
 	/* Receiving or ready to receive nbd request header */
@@ -922,6 +923,17 @@ nbd_start_complete(struct spdk_nbd_start_ctx *ctx)
 		goto err;
 	}
 
+#ifdef NBD_SET_TIMEOUT
+	rc = ioctl(ctx->nbd->dev_fd, NBD_SET_TIMEOUT, NBD_IO_TIMEOUT_S);
+	if (rc == -1) {
+		SPDK_ERRLOG("ioctl(NBD_SET_TIMEOUT) failed: %s\n", spdk_strerror(errno));
+		rc = -errno;
+		goto err;
+	}
+#else
+	SPDK_NOTICELOG("ioctl(NBD_SET_TIMEOUT) is not supported.\n");
+#endif
+
 #ifdef NBD_FLAG_SEND_TRIM
 	rc = ioctl(ctx->nbd->dev_fd, NBD_SET_FLAGS, NBD_FLAG_SEND_TRIM);
 	if (rc == -1) {

lib/nvme/nvme_ctrlr.c

@@ -418,8 +418,12 @@ spdk_nvme_ctrlr_alloc_io_qpair(struct spdk_nvme_ctrlr *ctrlr,
 	rc = spdk_nvme_ctrlr_connect_io_qpair(ctrlr, qpair);
 	if (rc != 0) {
 		SPDK_ERRLOG("nvme_transport_ctrlr_connect_io_qpair() failed\n");
+		nvme_robust_mutex_lock(&ctrlr->ctrlr_lock);
+		nvme_ctrlr_proc_remove_io_qpair(qpair);
 		TAILQ_REMOVE(&ctrlr->active_io_qpairs, qpair, tailq);
+		spdk_bit_array_set(ctrlr->free_io_qids, qpair->id);
 		nvme_transport_ctrlr_delete_io_qpair(ctrlr, qpair);
+		nvme_robust_mutex_unlock(&ctrlr->ctrlr_lock);
 		return NULL;
 	}
 

lib/nvme/nvme_internal.h

@@ -481,6 +481,7 @@ struct spdk_nvme_ns {
 	uint32_t			md_size;
 	uint32_t			pi_type;
 	uint32_t			sectors_per_max_io;
+	uint32_t			sectors_per_max_io_no_md;
 	uint32_t			sectors_per_stripe;
 	uint32_t			id;
 	uint16_t			flags;

lib/nvme/nvme_ns.c

@@ -64,6 +64,7 @@ nvme_ns_set_identify_data(struct spdk_nvme_ns *ns)
 	}
 
 	ns->sectors_per_max_io = spdk_nvme_ns_get_max_io_xfer_size(ns) / ns->extended_lba_size;
+	ns->sectors_per_max_io_no_md = spdk_nvme_ns_get_max_io_xfer_size(ns) / ns->sector_size;
 
 	if (nsdata->noiob) {
 		ns->sectors_per_stripe = nsdata->noiob;
@@ -556,6 +557,7 @@ void nvme_ns_destruct(struct spdk_nvme_ns *ns)
 	ns->md_size = 0;
 	ns->pi_type = 0;
 	ns->sectors_per_max_io = 0;
+	ns->sectors_per_max_io_no_md = 0;
 	ns->sectors_per_stripe = 0;
 	ns->flags = 0;
 	ns->csi = SPDK_NVME_CSI_NVM;

lib/nvme/nvme_ns_cmd.c

@@ -62,19 +62,27 @@ nvme_ns_check_request_length(uint32_t lba_count, uint32_t sectors_per_max_io,
 	return child_per_io >= qdepth;
 }
 
+static inline bool
+_nvme_md_excluded_from_xfer(struct spdk_nvme_ns *ns, uint32_t io_flags)
+{
+	return (io_flags & SPDK_NVME_IO_FLAGS_PRACT) &&
+	       (ns->flags & SPDK_NVME_NS_EXTENDED_LBA_SUPPORTED) &&
+	       (ns->flags & SPDK_NVME_NS_DPS_PI_SUPPORTED) &&
+	       (ns->md_size == 8);
+}
+
 static inline uint32_t
 _nvme_get_host_buffer_sector_size(struct spdk_nvme_ns *ns, uint32_t io_flags)
 {
-	uint32_t sector_size = ns->extended_lba_size;
+	return _nvme_md_excluded_from_xfer(ns, io_flags) ?
+	       ns->sector_size : ns->extended_lba_size;
+}
 
-	if ((io_flags & SPDK_NVME_IO_FLAGS_PRACT) &&
+static inline uint32_t
-	    (ns->flags & SPDK_NVME_NS_EXTENDED_LBA_SUPPORTED) &&
+_nvme_get_sectors_per_max_io(struct spdk_nvme_ns *ns, uint32_t io_flags)
-	    (ns->flags & SPDK_NVME_NS_DPS_PI_SUPPORTED) &&
+{
-	    (ns->md_size == 8)) {
+	return _nvme_md_excluded_from_xfer(ns, io_flags) ?
-		sector_size -= 8;
+	       ns->sectors_per_max_io_no_md : ns->sectors_per_max_io;
-	}
-
-	return sector_size;
 }
 
 static struct nvme_request *
@@ -393,7 +401,7 @@ _nvme_ns_cmd_rw(struct spdk_nvme_ns *ns, struct spdk_nvme_qpair *qpair,
 {
 	struct nvme_request	*req;
 	uint32_t		sector_size = _nvme_get_host_buffer_sector_size(ns, io_flags);
-	uint32_t		sectors_per_max_io = ns->sectors_per_max_io;
+	uint32_t		sectors_per_max_io = _nvme_get_sectors_per_max_io(ns, io_flags);
 	uint32_t		sectors_per_stripe = ns->sectors_per_stripe;
 
 	req = nvme_allocate_request(qpair, payload, lba_count * sector_size, lba_count * ns->md_size,

lib/nvme/nvme_pcie.c

@@ -918,7 +918,6 @@ nvme_pcie_qpair_build_contig_hw_sgl_request(struct spdk_nvme_qpair *qpair, struc
 
 	length = req->payload_size;
 	virt_addr = req->payload.contig_or_cb_arg + req->payload_offset;
-	mapping_length = length;
 
 	while (length > 0) {
 		if (nseg >= NVME_MAX_SGL_DESCRIPTORS) {
@@ -932,6 +931,7 @@ nvme_pcie_qpair_build_contig_hw_sgl_request(struct spdk_nvme_qpair *qpair, struc
 			return -EFAULT;
 		}
 
+		mapping_length = length;
 		phys_addr = spdk_vtophys(virt_addr, &mapping_length);
 		if (phys_addr == SPDK_VTOPHYS_ERROR) {
 			nvme_pcie_fail_request_bad_vtophys(qpair, tr);

lib/nvme/nvme_quirks.c

@@ -94,6 +94,9 @@ static const struct nvme_quirk nvme_quirks[] = {
 		NVME_INTEL_QUIRK_NO_LOG_PAGES |
 		NVME_QUIRK_MAXIMUM_PCI_ACCESS_WIDTH
 	},
+	{	{SPDK_PCI_CLASS_NVME, SPDK_PCI_VID_REDHAT, 0x0010, SPDK_PCI_ANY_ID, SPDK_PCI_ANY_ID},
+		NVME_QUIRK_MAXIMUM_PCI_ACCESS_WIDTH
+	},
 	{	{SPDK_PCI_CLASS_NVME, SPDK_PCI_VID_CNEXLABS, 0x1f1f, SPDK_PCI_ANY_ID, SPDK_PCI_ANY_ID},
 		NVME_QUIRK_IDENTIFY_CNS |
 		NVME_QUIRK_OCSSD

lib/nvmf/rdma.c

@@ -2,7 +2,7 @@
  *   BSD LICENSE
  *
  *   Copyright (c) Intel Corporation. All rights reserved.
- *   Copyright (c) 2019, 2020 Mellanox Technologies LTD. All rights reserved.
+ *   Copyright (c) 2019-2021 Mellanox Technologies LTD. All rights reserved.
  *
  *   Redistribution and use in source and binary forms, with or without
  *   modification, are permitted provided that the following conditions
@@ -944,6 +944,11 @@ nvmf_rdma_resize_cq(struct spdk_nvmf_rdma_qpair *rqpair, struct spdk_nvmf_rdma_d
 	}
 
 	if (rpoller->num_cqe != num_cqe) {
+		if (device->context->device->transport_type == IBV_TRANSPORT_IWARP) {
+			SPDK_ERRLOG("iWARP doesn't support CQ resize. Current capacity %u, required %u\n"
+				    "Using CQ of insufficient size may lead to CQ overrun\n", rpoller->num_cqe, num_cqe);
+			return -1;
+		}
 		if (required_num_wr > device->attr.max_cqe) {
 			SPDK_ERRLOG("RDMA CQE requirement (%d) exceeds device max_cqe limitation (%d)\n",
 				    required_num_wr, device->attr.max_cqe);
@@ -4106,19 +4111,24 @@ nvmf_rdma_qpair_abort_request(struct spdk_nvmf_qpair *qpair,
 	struct spdk_nvmf_rdma_transport *rtransport;
 	struct spdk_nvmf_transport *transport;
 	uint16_t cid;
-	uint32_t i;
+	uint32_t i, max_req_count;
-	struct spdk_nvmf_rdma_request *rdma_req_to_abort = NULL;
+	struct spdk_nvmf_rdma_request *rdma_req_to_abort = NULL, *rdma_req;
 
 	rqpair = SPDK_CONTAINEROF(qpair, struct spdk_nvmf_rdma_qpair, qpair);
 	rtransport = SPDK_CONTAINEROF(qpair->transport, struct spdk_nvmf_rdma_transport, transport);
 	transport = &rtransport->transport;
 
 	cid = req->cmd->nvme_cmd.cdw10_bits.abort.cid;
+	max_req_count = rqpair->srq == NULL ? rqpair->max_queue_depth : rqpair->poller->max_srq_depth;
 
-	for (i = 0; i < rqpair->max_queue_depth; i++) {
+	for (i = 0; i < max_req_count; i++) {
-		if (rqpair->resources->reqs[i].state != RDMA_REQUEST_STATE_FREE &&
+		rdma_req = &rqpair->resources->reqs[i];
-		    rqpair->resources->reqs[i].req.cmd->nvme_cmd.cid == cid) {
+		/* When SRQ == NULL, rqpair has its own requests and req.qpair pointer always points to the qpair
-			rdma_req_to_abort = &rqpair->resources->reqs[i];
+		 * When SRQ != NULL all rqpairs share common requests and qpair pointer is assigned when we start to
+		 * process a request. So in both cases all requests which are not in FREE state have valid qpair ptr */
+		if (rdma_req->state != RDMA_REQUEST_STATE_FREE && rdma_req->req.cmd->nvme_cmd.cid == cid &&
+		    rdma_req->req.qpair == qpair) {
+			rdma_req_to_abort = rdma_req;
 			break;
 		}
 	}

lib/sock/sock.c

@@ -832,6 +832,8 @@ spdk_sock_write_config_json(struct spdk_json_write_ctx *w)
 			spdk_json_write_named_uint32(w, "send_buf_size", opts.send_buf_size);
 			spdk_json_write_named_bool(w, "enable_recv_pipe", opts.enable_recv_pipe);
 			spdk_json_write_named_bool(w, "enable_zerocopy_send", opts.enable_zerocopy_send);
+			spdk_json_write_named_bool(w, "enable_quickack", opts.enable_quickack);
+			spdk_json_write_named_bool(w, "enable_placement_id", opts.enable_placement_id);
 			spdk_json_write_object_end(w);
 			spdk_json_write_object_end(w);
 		} else {

lib/util/fd_group.c

@@ -85,8 +85,8 @@ int
 spdk_fd_group_add(struct spdk_fd_group *fgrp,
 		  int efd, spdk_fd_fn fn, void *arg)
 {
-	struct event_handler *ehdlr;
+	struct event_handler *ehdlr = NULL;
-	struct epoll_event epevent;
+	struct epoll_event epevent = {0};
 	int rc;
 
 	/* parameter checking */

lib/vhost/rte_vhost_compat.c

@@ -46,6 +46,7 @@
 #include "spdk/barrier.h"
 #include "spdk/vhost.h"
 #include "vhost_internal.h"
+#include <rte_version.h>
 
 #include "spdk_internal/vhost_user.h"
 
@@ -331,7 +332,11 @@ vhost_register_unix_socket(const char *path, const char *ctrl_name,
 		}
 	}
 
+#if RTE_VERSION < RTE_VERSION_NUM(20, 8, 0, 0)
 	if (rte_vhost_driver_register(path, 0) != 0) {
+#else
+	if (rte_vhost_driver_register(path, RTE_VHOST_USER_ASYNC_COPY) != 0) {
+#endif
 		SPDK_ERRLOG("Could not register controller %s with vhost library\n", ctrl_name);
 		SPDK_ERRLOG("Check if domain socket %s already exists\n", path);
 		return -EIO;

lib/vhost/vhost.c

@@ -888,6 +888,7 @@ static int
 vhost_parse_core_mask(const char *mask, struct spdk_cpuset *cpumask)
 {
 	int rc;
+	struct spdk_cpuset negative_vhost_mask;
 
 	if (cpumask == NULL) {
 		return -1;
@@ -904,6 +905,16 @@ vhost_parse_core_mask(const char *mask, struct spdk_cpuset *cpumask)
 		return -1;
 	}
 
+	spdk_cpuset_copy(&negative_vhost_mask, &g_vhost_core_mask);
+	spdk_cpuset_negate(&negative_vhost_mask);
+	spdk_cpuset_and(&negative_vhost_mask, cpumask);
+
+	if (spdk_cpuset_count(&negative_vhost_mask) != 0) {
+		SPDK_ERRLOG("one of selected cpu is outside of core mask(=%s)\n",
+			    spdk_cpuset_fmt(&g_vhost_core_mask));
+		return -1;
+	}
+
 	spdk_cpuset_and(cpumask, &g_vhost_core_mask);
 
 	if (spdk_cpuset_count(cpumask) == 0) {

mk/spdk.fio.mk

@@ -39,6 +39,13 @@ FIO_PLUGIN := $(SPDK_ROOT_DIR)/build/fio/$(notdir $(FIO_PLUGIN))
 LIBS += $(SPDK_LIB_LINKER_ARGS)
 
 CFLAGS += -I$(CONFIG_FIO_SOURCE_DIR)
+# Compiling against fio 3.19 on latest FreeBSD generates warnings so we
+# cannot use -Werror
+ifeq ($(OS),FreeBSD)
+CFLAGS += -Wno-error
+else ifeq ($(CC_TYPE),clang)
+CFLAGS += -Wno-error
+endif
 LDFLAGS += -shared -rdynamic -Wl,-z,nodelete
 
 CLEAN_FILES = $(FIO_PLUGIN)

module/bdev/nvme/bdev_nvme.c

@@ -727,13 +727,6 @@ _bdev_nvme_submit_request(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_
 						     bdev_io->u.bdev.offset_blocks,
 						     bdev->dif_check_flags);
 
-	case SPDK_BDEV_IO_TYPE_WRITE_ZEROES:
-		return bdev_nvme_unmap(nvme_ns->ns,
-				       qpair,
-				       nbdev_io,
-				       bdev_io->u.bdev.offset_blocks,
-				       bdev_io->u.bdev.num_blocks);
-
 	case SPDK_BDEV_IO_TYPE_UNMAP:
 		return bdev_nvme_unmap(nvme_ns->ns,
 				       qpair,
@@ -837,20 +830,11 @@ bdev_nvme_io_type_supported(void *ctx, enum spdk_bdev_io_type io_type)
 		return cdata->oncs.dsm;
 
 	case SPDK_BDEV_IO_TYPE_WRITE_ZEROES:
-		cdata = spdk_nvme_ctrlr_get_data(ctrlr);
-		/*
-		 * If an NVMe controller guarantees reading unallocated blocks returns zero,
-		 * we can implement WRITE_ZEROES as an NVMe deallocate command.
-		 */
-		if (cdata->oncs.dsm &&
-		    spdk_nvme_ns_get_dealloc_logical_block_read_value(ns) ==
-		    SPDK_NVME_DEALLOC_READ_00) {
-			return true;
-		}
 		/*
 		 * The NVMe controller write_zeroes function is currently not used by our driver.
-		 * If a user submits an arbitrarily large write_zeroes request to the controller, the request will fail.
+		 * NVMe write zeroes is limited to 16-bit block count, and the bdev layer currently
-		 * Until this is resolved, we only claim support for write_zeroes if deallocated blocks return 0's when read.
+		 * has no mechanism for reporting a max write zeroes block count, nor ability to
+		 * split a write zeroes request.
 		 */
 		return false;
 

module/bdev/nvme/bdev_ocssd.c

@@ -967,7 +967,7 @@ bdev_ocssd_push_media_events(struct nvme_bdev_ns *nvme_ns,
 	TAILQ_FOREACH(nvme_bdev, &nvme_ns->bdevs, tailq) {
 		ocssd_bdev = SPDK_CONTAINEROF(nvme_bdev, struct ocssd_bdev, nvme_bdev);
 		if (bdev_ocssd_lba_in_range(ocssd_bdev, ocssd_ns, chunk_entry->lba)) {
-			return;
+			break;
 		}
 	}
 

pkg/spdk.spec

@@ -2,12 +2,12 @@
 %bcond_with doc
 
 Name: spdk
-Version: master
+Version: 21.01.x
 Release: 0%{?dist}
 Epoch: 0
 URL: http://spdk.io
 
-Source: https://github.com/spdk/spdk/archive/master.tar.gz
+Source: https://github.com/spdk/spdk/archive/v21.01.x.tar.gz
 Summary: Set of libraries and utilities for high performance user-mode storage
 
 %define package_version %{epoch}:%{version}-%{release}
@@ -22,7 +22,7 @@ License: BSD
 ExclusiveArch: x86_64
 
 BuildRequires: gcc gcc-c++ make
-BuildRequires: dpdk-devel, numactl-devel
+BuildRequires: dpdk-devel, numactl-devel, ncurses-devel
 BuildRequires: libiscsi-devel, libaio-devel, openssl-devel, libuuid-devel
 BuildRequires: libibverbs-devel, librdmacm-devel
 %if %{with doc}
@@ -32,6 +32,7 @@ BuildRequires: doxygen mscgen graphviz
 # Install dependencies
 Requires: dpdk >= 19.11, numactl-libs, openssl-libs
 Requires: libiscsi, libaio, libuuid
+Requires: python3-configshell, python3-pexpect
 # NVMe over Fabrics
 Requires: librdmacm, librdmacm
 Requires(post): /sbin/ldconfig

scripts/ceph/start.sh

@@ -15,6 +15,7 @@ mnt_dir=${base_dir}/mnt
 image=/var/tmp/ceph_raw.img
 dev=/dev/loop200
 
+modprobe loop
 umount ${dev}p2 || true
 losetup -d $dev || true
 

scripts/check_format.sh

@@ -1,7 +1,8 @@
 #!/usr/bin/env bash
 
-readonly BASEDIR=$(readlink -f $(dirname $0))/..
+rootdir=$(readlink -f "$(dirname "$0")")/..
-cd $BASEDIR
+
+cd "$rootdir"
 
 # exit on errors
 set -e
@@ -488,17 +489,19 @@ function check_bash_static_analysis() {
 		# SC2086: Double quote to prevent globbing and word splitting.
 		# SC2119: Use foo "$@" if function's $1 should mean script's $1.
 		# SC2120: foo references arguments, but none are ever passed.
+		# SC2128: Expanding an array without an index only gives the first element.
 		# SC2148: Add shebang to the top of your script.
 		# SC2153: Possible Misspelling: MYVARIABLE may not be assigned, but MY_VARIABLE is.
 		# SC2154: var is referenced but not assigned.
 		# SC2164: Use cd ... || exit in case cd fails.
 		# SC2174: When used with -p, -m only applies to the deepest directory.
+		# SC2178: Variable was used as an array but is now assigned a string.
 		# SC2206: Quote to prevent word splitting/globbing,
 		#         or split robustly with mapfile or read -a.
 		# SC2207: Prefer mapfile or read -a to split command output (or quote to avoid splitting).
 		# SC2223: This default assignment may cause DoS due to globbing. Quote it.
 		SHCK_EXCLUDE="$SHCK_EXCLUDE,SC1083,SC1090,SC1091,SC2010,SC2015,SC2016,SC2034,SC2046,SC2086,\
-SC2119,SC2120,SC2148,SC2153,SC2154,SC2164,SC2174,SC2001,SC2206,SC2207,SC2223"
+SC2119,SC2120,SC2128,SC2148,SC2153,SC2154,SC2164,SC2174,SC2178,SC2001,SC2206,SC2207,SC2223"
 
 		SHCK_FORMAT="tty"
 		SHCK_APPLY=false

scripts/perf/nvmf/run_nvmf.py

@@ -35,7 +35,7 @@ class Server:
         print("[%s] %s" % (self.name, msg), flush=True)
 
     def get_uncommented_lines(self, lines):
-        return [l for l in lines if l and not l.startswith('#')]
+        return [line for line in lines if line and not line.startswith('#')]
 
 
 class Target(Server):

scripts/perf/vhost/run_vhost_test.py

@@ -152,7 +152,7 @@ parser.add_argument('-d', '--max-disks', default=0, type=int,
                     each virtual machine gets it's own bdev to work on.")
 parser.add_argument('-v', '--vm-count', default=1, type=int,
                     help="How many VMs to run in test. Default: 1")
-parser.add_argument('-i', '--vm-image', default="$HOME/spdk_test_image.qcow2",
+parser.add_argument('-i', '--vm-image', default="$DEPENDENCY_DIR/spdk_test_image.qcow2",
                     type=str, help="VM image to use for running VMs.")
 
 subparsers = parser.add_subparsers()

scripts/rpc/client.py

@@ -14,6 +14,22 @@ def print_json(s):
     print(json.dumps(s, indent=2).strip('"'))
 
 
+def get_addr_type(addr):
+    try:
+        socket.inet_pton(socket.AF_INET, addr)
+        return socket.AF_INET
+    except Exception as e:
+        pass
+    try:
+        socket.inet_pton(socket.AF_INET6, addr)
+        return socket.AF_INET6
+    except Exception as e:
+        pass
+    if os.path.exists(addr):
+        return socket.AF_UNIX
+    return None
+
+
 class JSONRPCException(Exception):
     def __init__(self, message):
         self.message = message
@@ -54,23 +70,24 @@ class JSONRPCClient(object):
 
     def _connect(self, addr, port):
         try:
-            if os.path.exists(addr):
+            addr_type = get_addr_type(addr)
+
+            if addr_type == socket.AF_UNIX:
                 self._logger.debug("Trying to connect to UNIX socket: %s", addr)
                 self.sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
                 self.sock.connect(addr)
-            elif port:
+            elif addr_type == socket.AF_INET6:
-                if ':' in addr:
+                self._logger.debug("Trying to connect to IPv6 address addr:%s, port:%i", addr, port)
-                    self._logger.debug("Trying to connect to IPv6 address addr:%s, port:%i", addr, port)
+                for res in socket.getaddrinfo(addr, port, socket.AF_INET6, socket.SOCK_STREAM, socket.SOL_TCP):
-                    for res in socket.getaddrinfo(addr, port, socket.AF_INET6, socket.SOCK_STREAM, socket.SOL_TCP):
+                    af, socktype, proto, canonname, sa = res
-                        af, socktype, proto, canonname, sa = res
+                self.sock = socket.socket(af, socktype, proto)
-                    self.sock = socket.socket(af, socktype, proto)
+                self.sock.connect(sa)
-                    self.sock.connect(sa)
+            elif addr_type == socket.AF_INET:
-                else:
+                self._logger.debug("Trying to connect to IPv4 address addr:%s, port:%i'", addr, port)
-                    self._logger.debug("Trying to connect to IPv4 address addr:%s, port:%i'", addr, port)
+                self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
-                    self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+                self.sock.connect((addr, port))
-                    self.sock.connect((addr, port))
             else:
-                raise socket.error("Unix socket '%s' does not exist" % addr)
+                raise socket.error("Invalid or non-existing address: '%s'" % addr)
         except socket.error as ex:
             raise JSONRPCException("Error while connecting to %s\n"
                                    "Error details: %s" % (addr, ex))

scripts/rxe_cfg_small.sh

@@ -172,6 +172,10 @@ print_status() {
 	local field field_ref fieldidx
 	local pad
 
+	if [[ -n $NO_HEADER ]]; then
+		unset -v "lines[0]"
+	fi
+
 	for field_ref in "${lines[@]}"; do
 		printf '  '
 		fieldidx=0
@@ -243,23 +247,38 @@ link_up() {
 	echo $(($(< "$net/$1/flags") | 0x1)) > "$net/$1/flags"
 }
 
-collect_devices() {
+collect_net_devices() {
-	local net_dev rxe_dev
+	local net_dev
 
 	for net_dev in "$net/"!(bonding_masters); do
 		(($(< "$net_dev/type") != 1)) && continue
 		net_devices["${net_dev##*/}"]=$net_dev
-		for rxe_dev in "$infiniband/rxe"+([0-9]); do
-			if [[ $(< "$rxe_dev/parent") == "${net_dev##*/}" ]]; then
-				net_to_rxe["${net_dev##*/}"]=${rxe_dev##*/}
-				rxe_to_net["${rxe_dev##*/}"]=${net_dev##*/}
-				continue 2
-			fi
-		done
 	done
 }
 
-collect_devices
+collect_rxe_devices() {
+	local rxe_dev net_dev
+
+	for rxe_dev in "$infiniband/"*; do
+		if [[ -e $rxe_dev/parent ]]; then
+			# Soft
+			net_dev=$(< "$rxe_dev/parent")
+		elif [[ -e $rxe_dev/device/net ]]; then
+			# HW
+			net_dev=$(readlink -f "$rxe_dev/device/net/"*)
+			net_dev=${net_dev##*/}
+		else
+			continue
+		fi 2> /dev/null
+
+		[[ -n ${net_devices["$net_dev"]} ]] || continue
+		net_to_rxe["$net_dev"]=${rxe_dev##*/}
+		rxe_to_net["${rxe_dev##*/}"]=$net_dev
+	done
+}
+
+collect_net_devices
+collect_rxe_devices
 
 case "${1:-status}" in
 	start)
@@ -281,6 +300,9 @@ case "${1:-status}" in
 		)
 		status | grep -E "${match:-.}"
 		;;
+	rxe-net)
+		printf '%s\n' "${rxe_to_net[@]}"
+		;;
 	*)
 		printf 'Invalid argument (%s)\n' "$1"
 		;;

scripts/setup.sh

@@ -579,7 +579,9 @@ function status_linux() {
 	printf '\n%-8s %-15s %-6s %-6s %-7s %-16s %-10s %s\n' \
 		"Type" "BDF" "Vendor" "Device" "NUMA" "Driver" "Device" "Block devices" >&2
 
-	for bdf in "${!all_devices_d[@]}"; do
+	sorted_bdfs=($(printf '%s\n' "${!all_devices_d[@]}" | sort))
+
+	for bdf in "${sorted_bdfs[@]}"; do
 		driver=${drivers_d["$bdf"]}
 		if [ "$numa_nodes" = "0" ]; then
 			node="-"
@@ -611,7 +613,7 @@ function status_linux() {
 		printf '%-8s %-15s %-6s %-6s %-7s %-16s %-10s %s\n' \
 			"$desc" "$bdf" "${pci_ids_vendor["$bdf"]#0x}" "${pci_ids_device["$bdf"]#0x}" \
 			"$node" "${driver:--}" "${name:-}" "${blknames[*]:--}"
-	done | sort -bk2,2
+	done
 }
 
 function status_freebsd() {

scripts/vagrant/autorun-spdk.conf

@@ -9,6 +9,7 @@ SPDK_TEST_ISCSI_INITIATOR=0
 SPDK_TEST_NVME=0
 SPDK_TEST_NVME_CLI=0
 SPDK_TEST_NVMF=1
+SPDK_TEST_VFIOUSER=1
 SPDK_TEST_RBD=0
 SPDK_TEST_CRYPTO=0
 SPDK_TEST_OCF=0

test/common/autotest_common.sh

@@ -81,6 +81,8 @@ export SPDK_TEST_NVME_CLI
 export SPDK_TEST_NVME_CUSE
 : ${SPDK_TEST_NVMF=0}
 export SPDK_TEST_NVMF
+: ${SPDK_TEST_VFIOUSER=0}
+export SPDK_TEST_VFIOUSER
 : ${SPDK_TEST_NVMF_TRANSPORT="rdma"}
 export SPDK_TEST_NVMF_TRANSPORT
 : ${SPDK_TEST_RBD=0}
@@ -173,6 +175,8 @@ leak:$CONFIG_FIO_SOURCE_DIR/init.c
 leak:$CONFIG_FIO_SOURCE_DIR/filesetup.c
 leak:fio_memalign
 leak:spdk_fio_io_u_init
+# Suppress leaks in gperftools-libs from fio
+leak:libtcmalloc_minimal.so
 
 # Suppress leaks in libiscsi
 leak:libiscsi.so
@@ -186,7 +190,7 @@ export LSAN_OPTIONS=suppressions="$asan_suppression_file"
 export DEFAULT_RPC_ADDR="/var/tmp/spdk.sock"
 
 if [ -z "$DEPENDENCY_DIR" ]; then
-	export DEPENDENCY_DIR=/home/sys_sgsw
+	export DEPENDENCY_DIR=$HOME/spdk_dependencies
 else
 	export DEPENDENCY_DIR
 fi
@@ -201,10 +205,16 @@ if [ $SPDK_RUN_VALGRIND -eq 0 ]; then
 fi
 
 if [ "$(uname -s)" = "Linux" ]; then
+	export HUGEMEM=4096 CLEAR_HUGE=yes
+	if [[ $SPDK_TEST_CRYPTO -eq 1 || $SPDK_TEST_REDUCE -eq 1 ]]; then
+		# Make sure that memory is distributed across all NUMA nodes - by default, all goes to
+		# node0, but if QAT devices are attached to a different node, all of their VFs will end
+		# up under that node too and memory needs to be available there for the tests.
+		export HUGE_EVEN_ALLOC=yes
+	fi
+
 	MAKE="make"
 	MAKEFLAGS=${MAKEFLAGS:--j$(nproc)}
-	# Override the default HUGEMEM in scripts/setup.sh to allocate 8GB in hugepages.
-	export HUGEMEM=8192
 	if [[ $SPDK_TEST_USE_IGB_UIO -eq 1 ]]; then
 		export DRIVER_OVERRIDE=igb_uio
 		# Building kernel modules requires root privileges
@@ -422,6 +432,10 @@ function get_config_params() {
 		config_params+=' --with-raid5'
 	fi
 
+	if [ $SPDK_TEST_VFIOUSER -eq 1 ]; then
+		config_params+=' --with-vfio-user'
+	fi
+
 	# Check whether liburing library header exists
 	if [ -f /usr/include/liburing/io_uring.h ] && [ $SPDK_TEST_URING -eq 1 ]; then
 		config_params+=' --with-uring'
@@ -636,11 +650,6 @@ function process_shm() {
 	id=$2
 	if [ "$type" = "--pid" ]; then
 		id="pid${id}"
-	elif [ "$type" = "--id" ]; then
-		id="${id}"
-	else
-		echo "Please specify to search for pid or shared memory id."
-		return 1
 	fi
 
 	shm_files=$(find /dev/shm -name "*.${id}" -printf "%f\n")

test/common/config/README.md

@@ -17,7 +17,7 @@ to emulate an RDMA enabled NIC. NVMe controllers can also be virtualized in emul
 
 ## VM Specs
 
-When creating the user during the fedora installation, it is best to use the name sys_sgsw. Efforts are being made
+When creating the user during the fedora installation, it is best to use the name sys_sgci. Efforts are being made
 to remove all references to this user, or files specific to this user from the codebase, but there are still some
 trailing references to it.
 
@@ -33,7 +33,7 @@ configuration file. For a full list of the variable declarations available for a
 
 1. Download a fresh Fedora 26 image.
 2. Perform the installation of Fedora 26 server.
-3. Create an admin user sys_sgsw (enabling passwordless sudo for this account will make life easier during the tests).
+3. Create an admin user sys_sgci (enabling passwordless sudo for this account will make life easier during the tests).
 4. Run the vm_setup.sh script which will install all proper dependencies.
 5. Modify the autorun-spdk.conf file in the home directory.
 6. Reboot the VM.
@@ -81,5 +81,5 @@ To create the VM image manually use following steps:
    ~~~
    ./spdk/test/common/config/vm_setup.sh -t 'fio'
    ~~~
-6. Place the guest VM in the host at the following location: `/home/sys_sgci/spdk_test_image.qcow2`.
+6. Place the guest VM in the host at the following location: `$DEPENDENCY_DIR/spdk_test_image.qcow2`.
 7. On the host, edit the `~/autorun-spdk.conf` file to include the following line: SPDK_TEST_VHOST=1.

test/common/config/pkgdep/git

@@ -203,7 +203,7 @@ function install_rocksdb() {
 function install_fio() {
 	# This version of fio is installed in /usr/src/fio to enable
 	# building the spdk fio plugin.
-	local fio_version="fio-3.19"
+	local fio_version="fio-3.28"
 
 	if [ ! -d /usr/src/fio ]; then
 		if [ ! -d fio ]; then

test/common/config/vm_setup.sh

@@ -185,6 +185,7 @@ SPDK_TEST_ISCSI_INITIATOR=1
 SPDK_TEST_NVME=1
 SPDK_TEST_NVME_CLI=1
 SPDK_TEST_NVMF=1
+SPDK_TEST_VFIOUSER=1
 SPDK_TEST_RBD=1
 SPDK_TEST_BLOCKDEV=1
 SPDK_TEST_BLOBFS=1

test/dpdk_memory_utility/test_dpdk_mem_info.sh

@@ -5,10 +5,9 @@ rootdir=$(readlink -f $testdir/../..)
 source $rootdir/test/common/autotest_common.sh
 
 rpc_py="$rootdir/scripts/rpc.py"
-SPDK_APP="$SPDK_BIN_DIR/spdk_tgt"
 MEM_SCRIPT="$rootdir/scripts/dpdk_mem_info.py"
 
-$SPDK_APP &
+"${SPDK_APP[@]}" &
 spdkpid=$!
 
 waitforlisten $spdkpid

test/iscsi_tgt/rpc_config/rpc_config.sh

@@ -32,13 +32,13 @@ $rpc_py framework_start_init
 echo "iscsi_tgt is listening. Running tests..."
 
 # RPC framework_wait_init should be already returned, so its process must be non-existed
-! ps $rpc_wait_pid
+NOT ps $rpc_wait_pid
 
 # RPC framework_wait_init will directly returned after subsystem initialized.
 $rpc_py framework_wait_init &
 rpc_wait_pid=$!
 sleep 1
-! ps $rpc_wait_pid
+NOT ps $rpc_wait_pid
 
 timing_exit start_iscsi_tgt
 

test/json_config/json_config.sh

@@ -300,8 +300,12 @@ function create_iscsi_subsystem_config() {
 function create_nvmf_subsystem_config() {
 	timing_enter "${FUNCNAME[0]}"
 
-	RDMA_IP_LIST=$(get_available_rdma_ips)
+	NVMF_FIRST_TARGET_IP="127.0.0.1"
-	NVMF_FIRST_TARGET_IP=$(echo "$RDMA_IP_LIST" | head -n 1)
+	if [[ $SPDK_TEST_NVMF_TRANSPORT == "rdma" ]]; then
+		rdma_device_init
+		NVMF_FIRST_TARGET_IP=$(get_available_rdma_ips | head -n 1)
+	fi
+
 	if [[ -z $NVMF_FIRST_TARGET_IP ]]; then
 		echo "Error: no NIC for nvmf test"
 		return 1
@@ -310,11 +314,11 @@ function create_nvmf_subsystem_config() {
 	tgt_rpc bdev_malloc_create 8 512 --name MallocForNvmf0
 	tgt_rpc bdev_malloc_create 4 1024 --name MallocForNvmf1
 
-	tgt_rpc nvmf_create_transport -t RDMA -u 8192 -c 0
+	tgt_rpc nvmf_create_transport -t $SPDK_TEST_NVMF_TRANSPORT -u 8192 -c 0
 	tgt_rpc nvmf_create_subsystem nqn.2016-06.io.spdk:cnode1 -a -s SPDK00000000000001
 	tgt_rpc nvmf_subsystem_add_ns nqn.2016-06.io.spdk:cnode1 MallocForNvmf0
 	tgt_rpc nvmf_subsystem_add_ns nqn.2016-06.io.spdk:cnode1 MallocForNvmf1
-	tgt_rpc nvmf_subsystem_add_listener nqn.2016-06.io.spdk:cnode1 -t RDMA -a $NVMF_FIRST_TARGET_IP -s "$NVMF_PORT"
+	tgt_rpc nvmf_subsystem_add_listener nqn.2016-06.io.spdk:cnode1 -t $SPDK_TEST_NVMF_TRANSPORT -a $NVMF_FIRST_TARGET_IP -s "$NVMF_PORT"
 
 	timing_exit "${FUNCNAME[0]}"
 }

test/make/check_so_deps.sh

@@ -49,17 +49,6 @@ function confirm_abi_deps() {
 	fi
 
 	cat << EOF > ${suppression_file}
-[suppress_type]
-	name = spdk_nvme_ctrlr_data
-[suppress_type]
-	type_kind = enum
-	changed_enumerators = SPDK_BDEV_NUM_IO_TYPES
-[suppress_type]
-	name = spdk_env_opts
-[suppress_type]
-	name = spdk_app_opts
-[suppress_type]
-	name = spdk_thread
 EOF
 
 	for object in "$libdir"/libspdk_*.so; do

test/nvme/hotplug.sh

@@ -54,7 +54,7 @@ function devices_delete() {
 }
 
 password=$1
-base_img=$HOME/spdk_test_image.qcow2
+base_img=$DEPENDENCY_DIR/spdk_test_image.qcow2
 qemu_pidfile=$HOME/qemupid
 
 if [ ! -e "$base_img" ]; then

test/nvmf/common.sh

@@ -71,7 +71,7 @@ function detect_nics_and_probe_drivers() {
 	fi
 }
 
-function pci_nics_switch() {
+function pci_rdma_switch() {
 	local driver=$1
 
 	local -a driver_args=()
@@ -103,26 +103,60 @@ function pci_nics_switch() {
 	esac
 }
 
+function pci_tcp_switch() {
+	local driver=$1
+
+	local -a driver_args=()
+	driver_args+=("Intel E810 ice")
+
+	case $driver in
+		ice)
+			detect_nics_and_probe_drivers ${driver_args[0]}
+			;;
+		*)
+			for d in "${driver_args[@]}"; do
+				detect_nics_and_probe_drivers $d
+			done
+			;;
+	esac
+}
+
 function detect_pci_nics() {
 
 	if ! hash lspci; then
 		return 0
 	fi
 
-	local rdma_drivers="mlx5_ib|irdma|i40iw|iw_cxgb4"
+	local nic_drivers
 	local found_drivers
 
-	# Try to find RDMA drivers which are already loded and try to
+	if [[ -z "$TEST_TRANSPORT" ]]; then
-	# use only it's associated NICs, without probing all drivers.
+		TEST_TRANSPORT=$SPDK_TEST_NVMF_TRANSPORT
-	found_drivers=$(lsmod | grep -Eo $rdma_drivers | sort -u)
+	fi
-	for d in $found_drivers; do
-		pci_nics_switch $d
-	done
 
-	# In case lsmod reported driver, but lspci does not report
+	if [[ "$TEST_TRANSPORT" == "rdma" ]]; then
-	# physical NICs - fall back to old approach any try to
+		nic_drivers="mlx5_ib|irdma|i40iw|iw_cxgb4"
-	# probe all compatible NICs.
+
-	((have_pci_nics == 0)) && pci_nics_switch "default"
+		# Try to find RDMA drivers which are already loded and try to
+		# use only it's associated NICs, without probing all drivers.
+		found_drivers=$(lsmod | grep -Eo $nic_drivers | sort -u)
+		for d in $found_drivers; do
+			pci_rdma_switch $d
+		done
+
+		# In case lsmod reported driver, but lspci does not report
+		# physical NICs - fall back to old approach any try to
+		# probe all compatible NICs.
+		((have_pci_nics == 0)) && pci_rdma_switch "default"
+
+	elif [[ "$TEST_TRANSPORT" == "tcp" ]]; then
+		nic_drivers="ice"
+		found_drivers=$(lsmod | grep -Eo $nic_drivers | sort -u)
+		for d in $found_drivers; do
+			pci_tcp_switch $d
+		done
+		((have_pci_nics == 0)) && pci_tcp_switch "default"
+	fi
 
 	# Use softroce if everything else failed.
 	((have_pci_nics == 0)) && return 0
@@ -131,7 +165,7 @@ function detect_pci_nics() {
 	sleep 5
 }
 
-function detect_rdma_nics() {
+function detect_transport_nics() {
 	detect_pci_nics
 	if [ "$have_pci_nics" -eq "0" ]; then
 		detect_soft_roce_nics
@@ -159,13 +193,17 @@ function get_available_rdma_ips() {
 }
 
 function get_rdma_if_list() {
-	for nic_type in /sys/class/infiniband/*; do
+	rxe_cfg rxe-net
-		[[ -e "$nic_type" ]] || break
+}
-		for nic_name in /sys/class/infiniband/"$(basename ${nic_type})"/device/net/*; do
+
-			[[ -e "$nic_name" ]] || break
+function get_tcp_if_list_by_driver() {
-			basename "$nic_name"
+	local driver
-		done
+	driver=${1:-ice}
-	done
+
+	shopt -s nullglob
+	tcp_if_list=(/sys/bus/pci/drivers/$driver/0000*/net/*)
+	shopt -u nullglob
+	printf '%s\n' "${tcp_if_list[@]##*/}"
 }
 
 function get_ip_address() {
@@ -284,6 +322,67 @@ function nvmf_veth_fini() {
 	ip netns del $NVMF_TARGET_NAMESPACE
 }
 
+function nvmf_tcp_init() {
+	NVMF_INITIATOR_IP=10.0.0.1
+	NVMF_FIRST_TARGET_IP=10.0.0.2
+	TCP_INTERFACE_LIST=($(get_tcp_if_list_by_driver))
+	if ((${#TCP_INTERFACE_LIST[@]} == 0)); then
+		nvmf_veth_init
+		return 0
+	fi
+
+	# We need two net devs at minimum
+	((${#TCP_INTERFACE_LIST[@]} > 1))
+
+	NVMF_TARGET_INTERFACE=${TCP_INTERFACE_LIST[0]}
+	NVMF_INITIATOR_INTERFACE=${TCP_INTERFACE_LIST[1]}
+
+	# Skip case nvmf_multipath in nvmf_tcp_init(), it will be covered by nvmf_veth_init().
+	NVMF_SECOND_TARGET_IP=""
+
+	NVMF_TARGET_NAMESPACE=$NVMF_TARGET_INTERFACE"_ns"
+	NVMF_TARGET_NS_CMD=(ip netns exec "$NVMF_TARGET_NAMESPACE")
+	ip netns del $NVMF_TARGET_NAMESPACE || true
+	ip -4 addr flush $NVMF_TARGET_INTERFACE || true
+	ip -4 addr flush $NVMF_INITIATOR_INTERFACE || true
+
+	trap 'nvmf_tcp_fini; exit 1' SIGINT SIGTERM
+
+	# Create network namespace
+	ip netns add $NVMF_TARGET_NAMESPACE
+
+	# Associate phy interface pairs with network namespace
+	ip link set $NVMF_TARGET_INTERFACE netns $NVMF_TARGET_NAMESPACE
+
+	# Allocate IP addresses
+	ip addr add $NVMF_INITIATOR_IP/24 dev $NVMF_INITIATOR_INTERFACE
+	"${NVMF_TARGET_NS_CMD[@]}" ip addr add $NVMF_FIRST_TARGET_IP/24 dev $NVMF_TARGET_INTERFACE
+
+	# Link up phy interfaces
+	ip link set $NVMF_INITIATOR_INTERFACE up
+
+	"${NVMF_TARGET_NS_CMD[@]}" ip link set $NVMF_TARGET_INTERFACE up
+	"${NVMF_TARGET_NS_CMD[@]}" ip link set lo up
+
+	# Accept connections from phy interface
+	iptables -I INPUT 1 -i $NVMF_INITIATOR_INTERFACE -p tcp --dport $NVMF_PORT -j ACCEPT
+
+	# Verify connectivity
+	ping -c 1 $NVMF_FIRST_TARGET_IP
+	"${NVMF_TARGET_NS_CMD[@]}" ping -c 1 $NVMF_INITIATOR_IP
+
+	NVMF_APP=("${NVMF_TARGET_NS_CMD[@]}" "${NVMF_APP[@]}")
+}
+
+function nvmf_tcp_fini() {
+	if [[ "$NVMF_TARGET_NAMESPACE" == "nvmf_tgt_ns" ]]; then
+		nvmf_veth_fini
+		return 0
+	fi
+	ip netns del $NVMF_TARGET_NAMESPACE
+	ip -4 addr flush $NVMF_INITIATOR_INTERFACE
+}
+
 function nvmftestinit() {
 	if [ -z $TEST_TRANSPORT ]; then
 		echo "transport not specified - use --transport= to specify"
@@ -291,22 +390,25 @@ function nvmftestinit() {
 	fi
 	if [ "$TEST_MODE" == "iso" ]; then
 		$rootdir/scripts/setup.sh
-		if [ "$TEST_TRANSPORT" == "rdma" ]; then
+		if [[ "$TEST_TRANSPORT" == "rdma" ]]; then
 			rdma_device_init
 		fi
+		if [[ "$TEST_TRANSPORT" == "tcp" ]]; then
+			tcp_device_init
+		fi
 	fi
 
 	NVMF_TRANSPORT_OPTS="-t $TEST_TRANSPORT"
-	if [ "$TEST_TRANSPORT" == "rdma" ]; then
+	if [[ "$TEST_TRANSPORT" == "rdma" ]]; then
 		RDMA_IP_LIST=$(get_available_rdma_ips)
 		NVMF_FIRST_TARGET_IP=$(echo "$RDMA_IP_LIST" | head -n 1)
 		NVMF_SECOND_TARGET_IP=$(echo "$RDMA_IP_LIST" | tail -n +2 | head -n 1)
 		if [ -z $NVMF_FIRST_TARGET_IP ]; then
-			echo "no NIC for nvmf test"
+			echo "no RDMA NIC for nvmf test"
 			exit 0
 		fi
-	elif [ "$TEST_TRANSPORT" == "tcp" ]; then
+	elif [[ "$TEST_TRANSPORT" == "tcp" ]]; then
-		nvmf_veth_init
+		nvmf_tcp_init
 		NVMF_TRANSPORT_OPTS="$NVMF_TRANSPORT_OPTS -o"
 	fi
 
@@ -334,20 +436,25 @@ function nvmftestfini() {
 	fi
 	if [ "$TEST_MODE" == "iso" ]; then
 		$rootdir/scripts/setup.sh reset
-		if [ "$TEST_TRANSPORT" == "rdma" ]; then
+		if [[ "$TEST_TRANSPORT" == "rdma" ]]; then
 			rdma_device_init
-		elif [ "$TEST_TRANSPORT" == "tcp" ]; then
-			nvmf_veth_fini
 		fi
 	fi
+	if [[ "$TEST_TRANSPORT" == "tcp" ]]; then
+		nvmf_tcp_fini
+	fi
 }
 
 function rdma_device_init() {
 	load_ib_rdma_modules
-	detect_rdma_nics
+	detect_transport_nics
 	allocate_nic_ips
 }
 
+function tcp_device_init() {
+	detect_transport_nics
+}
+
 function revert_soft_roce() {
 	rxe_cfg stop
 }

test/nvmf/nvmf.sh

@@ -41,7 +41,7 @@ run_test "nvmf_invalid" test/nvmf/target/invalid.sh "${TEST_ARGS[@]}"
 run_test "nvmf_abort" test/nvmf/target/abort.sh "${TEST_ARGS[@]}"
 run_test "nvmf_ns_hotplug_stress" test/nvmf/target/ns_hotplug_stress.sh "${TEST_ARGS[@]}"
 
-if grep -q '#define SPDK_CONFIG_VFIO_USER 1' $rootdir/include/spdk/config.h; then
+if [ $SPDK_TEST_VFIOUSER -eq 1 ]; then
 	run_test "nvmf_vfio_user" test/nvmf/target/nvmf_vfio_user.sh "${TEST_ARGS[@]}"
 fi
 

test/nvmf/target/rpc.sh

@@ -59,7 +59,7 @@ $rpc_py nvmf_subsystem_allow_any_host -d nqn.2016-06.io.spdk:cnode1
 $rpc_py nvmf_subsystem_add_listener nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a $NVMF_FIRST_TARGET_IP -s $NVMF_PORT
 
 # This connect should fail - the host NQN is not allowed
-! nvme connect -t $TEST_TRANSPORT -n nqn.2016-06.io.spdk:cnode1 -q nqn.2016-06.io.spdk:host1 -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT"
+NOT nvme connect -t $TEST_TRANSPORT -n nqn.2016-06.io.spdk:cnode1 -q nqn.2016-06.io.spdk:host1 -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT"
 
 # Add the host NQN and verify that the connect succeeds
 $rpc_py nvmf_subsystem_add_host nqn.2016-06.io.spdk:cnode1 nqn.2016-06.io.spdk:host1
@@ -69,7 +69,7 @@ nvme disconnect -n nqn.2016-06.io.spdk:cnode1
 
 # Remove the host and verify that the connect fails
 $rpc_py nvmf_subsystem_remove_host nqn.2016-06.io.spdk:cnode1 nqn.2016-06.io.spdk:host1
-! nvme connect -t $TEST_TRANSPORT -n nqn.2016-06.io.spdk:cnode1 -q nqn.2016-06.io.spdk:host1 -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT"
+NOT nvme connect -t $TEST_TRANSPORT -n nqn.2016-06.io.spdk:cnode1 -q nqn.2016-06.io.spdk:host1 -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT"
 
 # Allow any host and verify that the connect succeeds
 $rpc_py nvmf_subsystem_allow_any_host -e nqn.2016-06.io.spdk:cnode1

test/ocf/management/remove.sh

@@ -57,7 +57,7 @@ $rpc_py bdev_ocf_delete ocfWT
 
 # Check that ocfWT was deleted properly
 
-! $rpc_py bdev_ocf_get_bdevs | jq -r '.[] .name' | grep -qw ocfWT
+[[ -z $("$rpc_py" bdev_ocf_get_bdevs | jq -r '.[] | select(.name == "ocfWT") | .name') ]]
 
 trap - SIGINT SIGTERM EXIT
 
@@ -73,7 +73,7 @@ waitforlisten $spdk_pid
 
 # Check that ocfWT was not loaded on app start
 
-! $rpc_py bdev_ocf_get_bdevs | jq -r '.[] .name' | grep -qw ocfWT
+[[ -z $("$rpc_py" bdev_ocf_get_bdevs | jq -r '.[] | select(name == "ocfWT") | .name') ]]
 
 trap - SIGINT SIGTERM EXIT
 

test/scheduler/common.sh

@@ -372,14 +372,15 @@ _get_thread_stats() {
 
 get_cpu_stat() {
 	local cpu_idx=$1
-	local stat=$2 stats
+	local stat=$2 stats astats
 
 	while read -r cpu stats; do
-		[[ $cpu == "cpu$cpu_idx" ]] && stats=($stats)
+		[[ $cpu == "cpu$cpu_idx" ]] && astats=($stats)
 	done < /proc/stat
 
 	case "$stat" in
-		idle) echo "${stats[3]}" ;;
+		idle) echo "${astats[3]}" ;;
+		all) printf '%u\n' "${astats[@]}" ;;
 		*) ;;
 	esac
 }
@@ -395,3 +396,124 @@ destroy_thread() {
 active_thread() {
 	"$rootdir/scripts/rpc.py" --plugin "$plugin" scheduler_thread_set_active "$@"
 }
+
+get_cpu_time() {
+	xtrace_disable
+
+	local interval=$1 cpu_time=$2 interval_count
+	shift 2
+	local cpus=("$@") cpu
+	local stats stat old_stats avg_load
+	local total_sample
+
+	# Exposed for the caller
+	local -g cpu_times=()
+	local -g avg_cpu_time=()
+
+	# cpu_time:
+	# 0 - user (time spent in user mode)
+	# 1 - nice (Time spent in user mode with low priority)
+	# 2 - system (Time spent in system mode)
+	# 3 - idle (Time spent in the idle task)
+	# 4 - iowait (Time waiting for I/O to complete)
+	# 5 - irq (Time servicing interrupts)
+	# 6 - softirq (Time servicing softirqs)
+	# 7 - steal (Stolen time)
+	# 8 - guest (Time spent running a virtual CPU)
+	# 9 - guest_nice (Time spent running a niced guest)
+
+	local -A cpu_time_map
+	cpu_time_map["user"]=0
+	cpu_time_map["nice"]=1
+	cpu_time_map["system"]=2
+	cpu_time_map["idle"]=3
+	cpu_time_map["iowait"]=4
+	cpu_time_map["irq"]=5
+	cpu_time_map["softirq"]=6
+	cpu_time_map["steal"]=7
+	cpu_time_map["guest"]=8
+	cpu_time_map["guest_nice"]=9
+
+	# Clear up the env
+	unset -v ${!stat_@}
+	unset -v ${!old_stat_@}
+	unset -v ${!avg_stat@}
+	unset -v ${!avg_load@}
+
+	cpu_time=${cpu_time_map["$cpu_time"]:-3}
+	interval=$((interval <= 0 ? 1 : interval))
+	# We skip first sample to have min 2 for stat comparision
+	interval=$((interval + 1)) interval_count=0
+	while ((interval_count++, --interval >= 0)); do
+		for cpu in "${cpus[@]}"; do
+			local -n old_stats=old_stats_$cpu
+			local -n avg_load=avg_load_$cpu
+			sample_stats=() total_sample=0
+
+			stats=($(get_cpu_stat "$cpu" all))
+			if ((interval_count == 1)); then
+				# Skip first sample
+				old_stats=("${stats[@]}")
+				continue
+			fi
+			for stat in "${!stats[@]}"; do
+				avg_load[stat]="stat_${stat}_${cpu}[@]"
+				sample_stats[stat]=$((stats[stat] - old_stats[stat]))
+				: $((total_sample += sample_stats[stat]))
+			done
+			for stat in "${!stats[@]}"; do
+				local -n avg_stat=stat_${stat}_${cpu}
+				avg_stat+=($((sample_stats[stat] * 100 / (total_sample == 0 ? 1 : total_sample))))
+			done
+			old_stats=("${stats[@]}")
+		done
+		sleep 1s
+	done
+
+	# We collected % for each time. Now determine the avg % for requested time.
+	local load stat_load
+	for cpu in "${cpus[@]}"; do
+		load=0
+		local -n avg_load_cpu=avg_load_$cpu
+		stat_load=("${!avg_load_cpu[cpu_time]}")
+		for stat in "${stat_load[@]}"; do
+			: $((load += stat))
+		done
+		cpu_times[cpu]=${stat_load[*]}
+		avg_cpu_time[cpu]=$((load / ${#stat_load[@]}))
+	done
+
+	xtrace_restore
+}
+
+collect_cpu_idle() {
+	((${#cpus_to_collect[@]} > 0)) || return 1
+
+	local time=${1:-5}
+	local cpu
+	local samples
+	local -g is_idle=()
+
+	printf 'Collecting cpu idle stats (cpus: %s) for %u seconds...\n' \
+		"${cpus_to_collect[*]}" "$time"
+
+	get_cpu_time "$time" idle "${cpus_to_collect[@]}"
+
+	for cpu in "${cpus_to_collect[@]}"; do
+		samples=(${cpu_times[cpu]})
+		printf '* cpu%u idle samples: %s (avg: %u%%)\n' \
+			"$cpu" "${samples[*]}" "${avg_cpu_time[cpu]}"
+		# Cores with polling reactors have 0% idle time,
+		# while the ones in interrupt mode won't have 100% idle.
+		# Work can be potentially be scheduled to the core by kernel,
+		# to prevent that affecting tests set reasonably high idle limit.
+		# Consider last sample
+		if ((samples[-1] >= 70)); then
+			printf '* cpu%u is idle\n' "$cpu"
+			is_idle[cpu]=1
+		else
+			printf '*cpu%u is not idle\n' "$cpu"
+			is_idle[cpu]=0
+		fi
+	done
+}

test/scheduler/governor.sh

@@ -86,7 +86,8 @@ verify_dpdk_governor() {
 		elif ((main_core_setspeed < old_main_core_setspeed)); then
 			dir=0
 		elif ((main_core_setspeed == old_main_core_setspeed)); then
-			# Frequency didn't change, skip
+			# Frequency didn't change, skip and wait for a bit
+			sleep 0.5s
 			continue
 		fi
 

test/scheduler/interrupt.sh

@@ -15,62 +15,6 @@ fold_list_onto_array cpus $(parse_cpu_list <(echo "$spdk_cpus_csv"))
 # Normalize the indexes
 cpus=("${cpus[@]}")
 
-collect_cpu_stat() {
-	local list=$1
-	local stat=$2
-
-	for cpu in "${cpus_to_collect[@]}"; do
-		eval "${list}[cpu]=\$(get_cpu_stat $cpu $stat)"
-	done
-}
-
-collect_cpu_idle() {
-	xtrace_disable
-
-	local sample_time=${1:-5} samples=0
-	local cpu bool inc user_hz
-
-	# idle scales to USER_HZ so we use that in order to determine the expected
-	# value it should have been increased to (more or less).
-	user_hz=100
-	# Expected increase of the idle stat
-	inc=$((user_hz * sample_time))
-
-	bool[0]="not" bool[1]="is"
-
-	init_idle_samples=() idle_samples=() is_idle=()
-
-	collect_cpu_stat init_idle_samples idle
-
-	printf 'Collecting cpu idle stats (cpus: %s) for %u seconds...\n' \
-		"${cpus_to_collect[*]}" "$sample_time"
-
-	while ((++samples <= sample_time)) && sleep 1s; do
-		collect_cpu_stat idle_samples idle
-	done
-
-	for cpu in "${!idle_samples[@]}"; do
-		# We start to collect after the spdk app is initialized hence if the interrupt
-		# mode is not working as expected, the idle time of given cpu will not have a
-		# chance to increase. If it does work correctly, then it should change even for
-		# a fraction, depending on how much time we spent on collecting this data.
-		# If idle time is over 70% of expected increase then we consider this cpu as
-		# idle. This is done in order to take into consideration time window the app
-		# needs to actually spin up|down the cpu. It's also taken for granted that
-		# there is no extra load on the target cpus which may be coming from other
-		# processes.
-		if ((idle_samples[cpu] > init_idle_samples[cpu] + (inc * 70 / 100))); then
-			is_idle[cpu]=1
-		else
-			is_idle[cpu]=0
-		fi
-		printf 'cpu%u %s idle (%u %u)\n' \
-			"$cpu" "${bool[is_idle[cpu]]}" "${init_idle_samples[cpu]}" "${idle_samples[cpu]}"
-	done
-
-	xtrace_restore
-}
-
 interrupt() {
 	local busy_cpus
 	local cpu thread

test/spdkcli/common.sh

@@ -43,3 +43,6 @@ function check_match() {
 	$rootdir/test/app/match/match $testdir/match_files/${MATCH_FILE}.match
 	rm -f $testdir/match_files/${MATCH_FILE}
 }
+
+# Allocate 5GB of hugepages to have some overhead for run_*()s
+HUGEMEM=5120 CLEAR_HUGE=yes "$rootdir/scripts/setup.sh"

test/spdkcli/match_files/spdkcli_details_vhost_ctrl.test.match

@@ -14,7 +14,7 @@
       }
     ]
   },
-  "cpumask": "$(S)",
+  "cpumask": "0x3",
   "ctrlr": "vhost_scsi1",
   "delay_base_us": 20,
   "iops_threshold": 1000000,

test/spdkcli/match_files/spdkcli_nvmf.test.match

@@ -7,9 +7,9 @@ o- nvmf ........................................................................
   | | o- hosts .......................................................................................................... [Hosts: 1]
   | | | o- nqn.2014-08.org.spdk:cnode2 ....................................................................................... [...]
   | | o- listen_addresses ........................................................................................... [Addresses: 3]
-  | | | o- $(N).$(N).$(N).$(N):4260 $(S) [RDMA]
+  | | | o- $(N).$(N).$(N).$(N):4260 $(S) [$(S)]
-  | | | o- $(N).$(N).$(N).$(N):4261 $(S) [RDMA]
+  | | | o- $(N).$(N).$(N).$(N):4261 $(S) [$(S)]
-  | | | o- $(N).$(N).$(N).$(N):4262 $(S) [RDMA]
+  | | | o- $(N).$(N).$(N).$(N):4262 $(S) [$(S)]
   | | o- namespaces ................................................................................................ [Namespaces: 4]
   | |   o- Malloc3 .................................................................................................... [Malloc3, 1]
   | |   o- Malloc4 .................................................................................................... [Malloc4, 2]
@@ -18,7 +18,7 @@ o- nvmf ........................................................................
   | o- nqn.2014-08.org.spdk:cnode2 ...................................................... [sn=$(S), st=NVMe, Allow any host]
   | | o- hosts .......................................................................................................... [Hosts: 0]
   | | o- listen_addresses ........................................................................................... [Addresses: 1]
-  | | | o- $(N).$(N).$(N).$(N):4260 $(S) [RDMA]
+  | | | o- $(N).$(N).$(N).$(N):4260 $(S) [$(S)]
   | | o- namespaces ................................................................................................ [Namespaces: 1]
   | |   o- Malloc2 .................................................................................................... [Malloc2, 1]
   | o- nqn.2014-08.org.spdk:cnode3 ...................................................... [sn=$(S), st=NVMe, Allow any host]
@@ -26,9 +26,9 @@ o- nvmf ........................................................................
   |   | o- nqn.2014-08.org.spdk:cnode1 ....................................................................................... [...]
   |   | o- nqn.2014-08.org.spdk:cnode2 ....................................................................................... [...]
   |   o- listen_addresses ........................................................................................... [Addresses: 2]
-  |   | o- $(N).$(N).$(N).$(N):4260 $(S) [RDMA]
+  |   | o- $(N).$(N).$(N).$(N):4260 $(S) [$(S)]
-  |   | o- $(N).$(N).$(N).$(N):4261 $(S) [RDMA]
+  |   | o- $(N).$(N).$(N).$(N):4261 $(S) [$(S)]
   |   o- namespaces ................................................................................................ [Namespaces: 1]
   |     o- Malloc1 .................................................................................................... [Malloc1, 1]
   o- transport ..................................................................................................... [Transports: 1]
-    o- RDMA .................................................................................................................. [...]
+    o- $(S) $(*) [...]

test/spdkcli/nvmf.sh

@@ -10,14 +10,16 @@ MATCH_FILE="spdkcli_nvmf.test"
 SPDKCLI_BRANCH="/nvmf"
 
 trap 'on_error_exit; revert_soft_roce' ERR
-rdma_device_init
 
 timing_enter run_nvmf_tgt
 run_nvmf_tgt
 timing_exit run_nvmf_tgt
 
-RDMA_IP_LIST=$(get_available_rdma_ips)
+NVMF_TARGET_IP="127.0.0.1"
-NVMF_TARGET_IP=$(echo "$RDMA_IP_LIST" | head -n 1)
+if [[ $TEST_TRANSPORT == "rdma" ]]; then
+	rdma_device_init
+	NVMF_TARGET_IP=$(get_available_rdma_ips | head -n 1)
+fi
 
 timing_enter spdkcli_create_nvmf_config
 $spdkcli_job "'/bdevs/malloc create 32 512 Malloc1' 'Malloc1' True
@@ -26,33 +28,33 @@ $spdkcli_job "'/bdevs/malloc create 32 512 Malloc1' 'Malloc1' True
 '/bdevs/malloc create 32 512 Malloc4' 'Malloc4' True
 '/bdevs/malloc create 32 512 Malloc5' 'Malloc5' True
 '/bdevs/malloc create 32 512 Malloc6' 'Malloc6' True
-'nvmf/transport create RDMA max_io_qpairs_per_ctrlr=4 io_unit_size=8192' '' True
+'nvmf/transport create $TEST_TRANSPORT max_io_qpairs_per_ctrlr=4 io_unit_size=8192' '' True
 '/nvmf/subsystem create nqn.2014-08.org.spdk:cnode1 N37SXV509SRW\
   max_namespaces=4 allow_any_host=True' 'nqn.2014-08.org.spdk:cnode1' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/namespaces create Malloc3 1' 'Malloc3' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/namespaces create Malloc4 2' 'Malloc4' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses create \
- RDMA $NVMF_TARGET_IP 4260 IPv4' '$NVMF_TARGET_IP:4260' True
+ $TEST_TRANSPORT $NVMF_TARGET_IP 4260 IPv4' '$NVMF_TARGET_IP:4260' True
 '/nvmf/subsystem create nqn.2014-08.org.spdk:cnode2 N37SXV509SRD\
  max_namespaces=2 allow_any_host=True' 'nqn.2014-08.org.spdk:cnode2' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode2/namespaces create Malloc2' 'Malloc2' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode2/listen_addresses create \
- RDMA $NVMF_TARGET_IP 4260 IPv4' '$NVMF_TARGET_IP:4260' True
+ $TEST_TRANSPORT $NVMF_TARGET_IP 4260 IPv4' '$NVMF_TARGET_IP:4260' True
 '/nvmf/subsystem create nqn.2014-08.org.spdk:cnode3 N37SXV509SRR\
  max_namespaces=2 allow_any_host=True' 'nqn.2014-08.org.spdk:cnode2' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode3/namespaces create Malloc1' 'Malloc1' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode3/listen_addresses create \
- RDMA $NVMF_TARGET_IP 4260 IPv4' '$NVMF_TARGET_IP:4260' True
+ $TEST_TRANSPORT $NVMF_TARGET_IP 4260 IPv4' '$NVMF_TARGET_IP:4260' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode3/listen_addresses create \
- RDMA $NVMF_TARGET_IP 4261 IPv4' '$NVMF_TARGET_IP:4261' True
+ $TEST_TRANSPORT $NVMF_TARGET_IP 4261 IPv4' '$NVMF_TARGET_IP:4261' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode3/hosts create \
  nqn.2014-08.org.spdk:cnode1' 'nqn.2014-08.org.spdk:cnode1' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode3/hosts create \
  nqn.2014-08.org.spdk:cnode2' 'nqn.2014-08.org.spdk:cnode2' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1 allow_any_host True' 'Allow any host'
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1 allow_any_host False' 'Allow any host' True
-'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses create RDMA $NVMF_TARGET_IP 4261 IPv4' '$NVMF_TARGET_IP:4261' True
+'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses create $TEST_TRANSPORT $NVMF_TARGET_IP 4261 IPv4' '$NVMF_TARGET_IP:4261' True
-'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses create RDMA $NVMF_TARGET_IP 4262 IPv4' '$NVMF_TARGET_IP:4262' True
+'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses create $TEST_TRANSPORT $NVMF_TARGET_IP 4262 IPv4' '$NVMF_TARGET_IP:4262' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/hosts create nqn.2014-08.org.spdk:cnode2' 'nqn.2014-08.org.spdk:cnode2' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/namespaces create Malloc5' 'Malloc5' True
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/namespaces create Malloc6' 'Malloc6' True
@@ -68,7 +70,7 @@ $spdkcli_job "'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/namespaces delete nsi
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/namespaces delete_all' 'Malloc4'
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/hosts delete nqn.2014-08.org.spdk:cnode2' 'nqn.2014-08.org.spdk:cnode2'
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode3/hosts delete_all' 'nqn.2014-08.org.spdk:cnode1'
-'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses delete RDMA $NVMF_TARGET_IP 4262' '$NVMF_TARGET_IP:4262'
+'/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses delete $TEST_TRANSPORT $NVMF_TARGET_IP 4262' '$NVMF_TARGET_IP:4262'
 '/nvmf/subsystem/nqn.2014-08.org.spdk:cnode1/listen_addresses delete_all' '$NVMF_TARGET_IP:4261'
 '/nvmf/subsystem delete nqn.2014-08.org.spdk:cnode3' 'nqn.2014-08.org.spdk:cnode3'
 '/nvmf/subsystem delete_all' 'nqn.2014-08.org.spdk:cnode2'

test/unit/lib/event/app.c/app_ut.c

@@ -120,7 +120,7 @@ test_spdk_app_parse_args(void)
 							"-d",
 							"-p",
 							"--single-file-segments",
-							"--silence-noticelog"
+							"--silence-noticelog",
 							"-R"
 						       };
 

test/unit/lib/idxd/idxd.c/idxd_ut.c

@@ -44,6 +44,9 @@
 #define NUM_ENGINES_PER_GROUP 1
 #define TOTAL_WQS (NUM_GROUPS * NUM_WQ_PER_GROUP)
 #define TOTAL_ENGINES (NUM_GROUPS * NUM_ENGINES_PER_GROUP)
+#define GRP_CFG_OFFSET 0x400
+#define MAX_TOKENS 0x40
+#define MAX_ARRAY_SIZE 0x20
 
 DEFINE_STUB(spdk_pci_idxd_get_driver, struct spdk_pci_driver *, (void), NULL);
 
@@ -108,6 +111,10 @@ test_idxd_wq_config(void)
 	SPDK_CU_ASSERT_FATAL(idxd.reg_base != NULL);
 
 	g_dev_cfg = &g_dev_cfg0;
+	SPDK_CU_ASSERT_FATAL(g_dev_cfg->num_groups <= MAX_ARRAY_SIZE);
+	idxd.groups = calloc(g_dev_cfg->num_groups, sizeof(struct idxd_group));
+	SPDK_CU_ASSERT_FATAL(idxd.groups != NULL);
+
 	idxd.registers.wqcap.total_wq_size = TOTAL_WQE_SIZE;
 	idxd.registers.wqcap.num_wqs = TOTAL_WQS;
 	idxd.registers.gencap.max_batch_shift = LOG2_WQ_MAX_BATCH;
@@ -138,12 +145,11 @@ test_idxd_wq_config(void)
 
 	free(idxd.queues);
 	free(idxd.reg_base);
+	free(idxd.groups);
 
 	return 0;
 }
 
-#define GRP_CFG_OFFSET 0x400
-#define MAX_TOKENS 0x40
 static int
 test_idxd_group_config(void)
 {

test/unit/lib/iscsi/iscsi.c/iscsi_ut.c

@@ -1993,6 +1993,43 @@ pdu_hdr_op_data_test(void)
 	g_task_pool_is_empty = false;
 }
 
+/* Test an ISCSI_OP_TEXT PDU with CONTINUE bit set but
+ * no data.
+ */
+static void
+empty_text_with_cbit_test(void)
+{
+	struct spdk_iscsi_sess sess = {};
+	struct spdk_iscsi_conn conn = {};
+	struct spdk_scsi_dev dev = {};
+	struct spdk_iscsi_pdu *req_pdu;
+	int rc;
+
+	req_pdu = iscsi_get_pdu(&conn);
+
+	sess.ExpCmdSN = 0;
+	sess.MaxCmdSN = 64;
+	sess.session_type = SESSION_TYPE_NORMAL;
+	sess.MaxBurstLength = 1024;
+
+	conn.full_feature = 1;
+	conn.sess = &sess;
+	conn.dev = &dev;
+	conn.state = ISCSI_CONN_STATE_RUNNING;
+
+	memset(&req_pdu->bhs, 0, sizeof(req_pdu->bhs));
+	req_pdu->bhs.opcode = ISCSI_OP_TEXT;
+	req_pdu->bhs.flags = ISCSI_TEXT_CONTINUE;
+
+	rc = iscsi_pdu_hdr_handle(&conn, req_pdu);
+	CU_ASSERT(rc == 0);
+	CU_ASSERT(!req_pdu->is_rejected);
+	rc = iscsi_pdu_payload_handle(&conn, req_pdu);
+	CU_ASSERT(rc == 0);
+
+	iscsi_put_pdu(req_pdu);
+}
+
 int
 main(int argc, char **argv)
 {
@@ -2024,6 +2061,7 @@ main(int argc, char **argv)
 	CU_ADD_TEST(suite, pdu_hdr_op_task_mgmt_test);
 	CU_ADD_TEST(suite, pdu_hdr_op_nopout_test);
 	CU_ADD_TEST(suite, pdu_hdr_op_data_test);
+	CU_ADD_TEST(suite, empty_text_with_cbit_test);
 
 	CU_basic_set_mode(CU_BRM_VERBOSE);
 	CU_basic_run_tests();

test/unit/lib/iscsi/param.c/param_ut.c

@@ -264,6 +264,14 @@ parse_valid_test(void)
 	EXPECT_VAL("F", "IIII");
 	CU_ASSERT_PTR_NULL(partial_parameter);
 
+	/* partial parameter: NULL data */
+	/* It is technically allowed to have a TEXT PDU with no data, yet
+	 * CONTINUE bit is enabled - make sure we handle that case correctly.
+	 */
+	rc = iscsi_parse_params(&params, NULL, 0, true, &partial_parameter);
+	CU_ASSERT(rc == 0);
+	CU_ASSERT_PTR_NULL(partial_parameter);
+
 	/* Second partial parameter is the only parameter */
 	PARSE("OOOO", true, &partial_parameter);
 	CU_ASSERT_STRING_EQUAL(partial_parameter, "OOOO");

test/unit/lib/nvme/nvme_ctrlr.c/nvme_ctrlr_ut.c

@@ -2162,6 +2162,27 @@ test_nvme_ctrlr_init_set_keep_alive_timeout(void)
 	nvme_ctrlr_destruct(&ctrlr);
 }
 
+static void
+test_alloc_io_qpair_fail(void)
+{
+	struct spdk_nvme_ctrlr ctrlr = {};
+	struct spdk_nvme_qpair *q0;
+
+	setup_qpairs(&ctrlr, 1);
+
+	/* Modify the connect_qpair return code to inject a failure */
+	g_connect_qpair_return_code = 1;
+
+	/* Attempt to allocate a qpair, this should fail */
+	q0 = spdk_nvme_ctrlr_alloc_io_qpair(&ctrlr, NULL, 0);
+	SPDK_CU_ASSERT_FATAL(q0 == NULL);
+
+	/* Verify that the qpair is removed from the lists */
+	SPDK_CU_ASSERT_FATAL(TAILQ_EMPTY(&ctrlr.active_io_qpairs));
+
+	cleanup_qpairs(&ctrlr);
+}
+
 int main(int argc, char **argv)
 {
 	CU_pSuite	suite = NULL;
@@ -2200,6 +2221,7 @@ int main(int argc, char **argv)
 	CU_ADD_TEST(suite, test_nvme_ctrlr_init_set_nvmf_ioccsz);
 	CU_ADD_TEST(suite, test_nvme_ctrlr_init_set_num_queues);
 	CU_ADD_TEST(suite, test_nvme_ctrlr_init_set_keep_alive_timeout);
+	CU_ADD_TEST(suite, test_alloc_io_qpair_fail);
 
 	CU_basic_set_mode(CU_BRM_VERBOSE);
 	CU_basic_run_tests();

test/unit/lib/nvme/nvme_ns.c/nvme_ns_ut.c

@@ -71,7 +71,8 @@ test_nvme_ns_construct(void)
 {
 	struct spdk_nvme_ns ns = {};
 	uint32_t id = 1;
-	struct spdk_nvme_ctrlr ctrlr = {};
+	struct spdk_nvme_ns_data nsdata = {};
+	struct spdk_nvme_ctrlr ctrlr = { .nsdata = &nsdata };
 
 	nvme_ns_construct(&ns, id, &ctrlr);
 	CU_ASSERT(ns.id == 1);

test/unit/lib/nvme/nvme_ns_cmd.c/nvme_ns_cmd_ut.c

@@ -240,6 +240,7 @@ prepare_for_test(struct spdk_nvme_ns *ns, struct spdk_nvme_ctrlr *ctrlr,
 	}
 	ns->md_size = md_size;
 	ns->sectors_per_max_io = spdk_nvme_ns_get_max_io_xfer_size(ns) / ns->extended_lba_size;
+	ns->sectors_per_max_io_no_md = spdk_nvme_ns_get_max_io_xfer_size(ns) / ns->sector_size;
 	ns->sectors_per_stripe = stripe_size / ns->extended_lba_size;
 
 	memset(qpair, 0, sizeof(*qpair));
@@ -909,11 +910,7 @@ test_nvme_ns_cmd_comparev_with_md(void)
 	 * Protection information enabled + PRACT
 	 *
 	 * Special case for 8-byte metadata + PI + PRACT: no metadata transferred
-	 * In theory, 256 blocks * 512 bytes per block = one I/O (128 KB)
+	 * 256 blocks * 512 bytes per block = single 128 KB I/O (no splitting required)
-	 * However, the splitting code does not account for PRACT when calculating
-	 * max sectors per transfer, so we actually get two I/Os:
-	 *   child 0: 252 blocks
-	 *   child 1: 4 blocks
 	 */
 	prepare_for_test(&ns, &ctrlr, &qpair, 512, 8, 128 * 1024, 0, true);
 	ns.flags |= SPDK_NVME_NS_DPS_PI_SUPPORTED;
@@ -923,18 +920,11 @@ test_nvme_ns_cmd_comparev_with_md(void)
 
 	SPDK_CU_ASSERT_FATAL(rc == 0);
 	SPDK_CU_ASSERT_FATAL(g_request != NULL);
-	SPDK_CU_ASSERT_FATAL(g_request->num_children == 2);
+	SPDK_CU_ASSERT_FATAL(g_request->num_children == 0);
-	child0 = TAILQ_FIRST(&g_request->children);
 
-	SPDK_CU_ASSERT_FATAL(child0 != NULL);
+	CU_ASSERT(g_request->payload.md == NULL);
-	CU_ASSERT(child0->payload_offset == 0);
+	CU_ASSERT(g_request->payload_offset == 0);
-	CU_ASSERT(child0->payload_size == 252 * 512); /* NOTE: does not include metadata! */
+	CU_ASSERT(g_request->payload_size == 256 * 512); /* NOTE: does not include metadata! */
-	child1 = TAILQ_NEXT(child0, child_tailq);
-
-	SPDK_CU_ASSERT_FATAL(child1 != NULL);
-	CU_ASSERT(child1->payload.md == NULL);
-	CU_ASSERT(child1->payload_offset == 252 * 512);
-	CU_ASSERT(child1->payload_size == 4 * 512);
 
 	nvme_request_free_children(g_request);
 	nvme_free_request(g_request);
@@ -1373,11 +1363,7 @@ test_nvme_ns_cmd_write_with_md(void)
 	 * Protection information enabled + PRACT
 	 *
 	 * Special case for 8-byte metadata + PI + PRACT: no metadata transferred
-	 * In theory, 256 blocks * 512 bytes per block = one I/O (128 KB)
+	 * 256 blocks * 512 bytes per block = single 128 KB I/O (no splitting required)
-	 * However, the splitting code does not account for PRACT when calculating
-	 * max sectors per transfer, so we actually get two I/Os:
-	 *   child 0: 252 blocks
-	 *   child 1: 4 blocks
 	 */
 	prepare_for_test(&ns, &ctrlr, &qpair, 512, 8, 128 * 1024, 0, true);
 	ns.flags |= SPDK_NVME_NS_DPS_PI_SUPPORTED;
@@ -1387,18 +1373,11 @@ test_nvme_ns_cmd_write_with_md(void)
 
 	SPDK_CU_ASSERT_FATAL(rc == 0);
 	SPDK_CU_ASSERT_FATAL(g_request != NULL);
-	SPDK_CU_ASSERT_FATAL(g_request->num_children == 2);
+	SPDK_CU_ASSERT_FATAL(g_request->num_children == 0);
-	child0 = TAILQ_FIRST(&g_request->children);
 
-	SPDK_CU_ASSERT_FATAL(child0 != NULL);
+	CU_ASSERT(g_request->payload.md == NULL);
-	CU_ASSERT(child0->payload_offset == 0);
+	CU_ASSERT(g_request->payload_offset == 0);
-	CU_ASSERT(child0->payload_size == 252 * 512); /* NOTE: does not include metadata! */
+	CU_ASSERT(g_request->payload_size == 256 * 512); /* NOTE: does not include metadata! */
-	child1 = TAILQ_NEXT(child0, child_tailq);
-
-	SPDK_CU_ASSERT_FATAL(child1 != NULL);
-	CU_ASSERT(child1->payload.md == NULL);
-	CU_ASSERT(child1->payload_offset == 252 * 512);
-	CU_ASSERT(child1->payload_size == 4 * 512);
 
 	nvme_request_free_children(g_request);
 	nvme_free_request(g_request);
@@ -1758,11 +1737,7 @@ test_nvme_ns_cmd_compare_with_md(void)
 	 * Protection information enabled + PRACT
 	 *
 	 * Special case for 8-byte metadata + PI + PRACT: no metadata transferred
-	 * In theory, 256 blocks * 512 bytes per block = one I/O (128 KB)
+	 * 256 blocks * 512 bytes per block = single 128 KB I/O (no splitting required)
-	 * However, the splitting code does not account for PRACT when calculating
-	 * max sectors per transfer, so we actually get two I/Os:
-	 *   child 0: 252 blocks
-	 *   child 1: 4 blocks
 	 */
 	prepare_for_test(&ns, &ctrlr, &qpair, 512, 8, 128 * 1024, 0, true);
 	ns.flags |= SPDK_NVME_NS_DPS_PI_SUPPORTED;
@@ -1772,18 +1747,11 @@ test_nvme_ns_cmd_compare_with_md(void)
 
 	SPDK_CU_ASSERT_FATAL(rc == 0);
 	SPDK_CU_ASSERT_FATAL(g_request != NULL);
-	SPDK_CU_ASSERT_FATAL(g_request->num_children == 2);
+	SPDK_CU_ASSERT_FATAL(g_request->num_children == 0);
-	child0 = TAILQ_FIRST(&g_request->children);
 
-	SPDK_CU_ASSERT_FATAL(child0 != NULL);
+	CU_ASSERT(g_request->payload.md == NULL);
-	CU_ASSERT(child0->payload_offset == 0);
+	CU_ASSERT(g_request->payload_offset == 0);
-	CU_ASSERT(child0->payload_size == 252 * 512); /* NOTE: does not include metadata! */
+	CU_ASSERT(g_request->payload_size == 256 * 512); /* NOTE: does not include metadata! */
-	child1 = TAILQ_NEXT(child0, child_tailq);
-
-	SPDK_CU_ASSERT_FATAL(child1 != NULL);
-	CU_ASSERT(child1->payload.md == NULL);
-	CU_ASSERT(child1->payload_offset == 252 * 512);
-	CU_ASSERT(child1->payload_size == 4 * 512);
 
 	nvme_request_free_children(g_request);
 	nvme_free_request(g_request);

test/unit/lib/vhost/vhost.c/vhost_ut.c

@@ -267,28 +267,56 @@ create_controller_test(void)
 	int ret;
 	char long_name[PATH_MAX];
 
-	spdk_cpuset_set_cpu(&g_vhost_core_mask, 0, true);
+	spdk_cpuset_parse(&g_vhost_core_mask, "0xf");
+
+	/* Create device with cpumask implcitly matching whole application */
+	ret = alloc_vdev(&vdev, "vdev_name_0", NULL);
+	SPDK_CU_ASSERT_FATAL(ret == 0 && vdev);
+	SPDK_CU_ASSERT_FATAL(!strcmp(spdk_cpuset_fmt(spdk_thread_get_cpumask(vdev->thread)), "f"));
+	cleanup_vdev(vdev);
+
+	/* Create device with cpumask matching whole application */
+	ret = alloc_vdev(&vdev, "vdev_name_0", "0xf");
+	SPDK_CU_ASSERT_FATAL(ret == 0 && vdev);
+	SPDK_CU_ASSERT_FATAL(!strcmp(spdk_cpuset_fmt(spdk_thread_get_cpumask(vdev->thread)), "f"));
+	cleanup_vdev(vdev);
+
+	/* Create device with single core in cpumask */
+	ret = alloc_vdev(&vdev, "vdev_name_0", "0x2");
+	SPDK_CU_ASSERT_FATAL(ret == 0 && vdev);
+	SPDK_CU_ASSERT_FATAL(!strcmp(spdk_cpuset_fmt(spdk_thread_get_cpumask(vdev->thread)), "2"));
+	cleanup_vdev(vdev);
+
+	/* Create device with cpumask spanning two cores */
+	ret = alloc_vdev(&vdev, "vdev_name_0", "0x3");
+	SPDK_CU_ASSERT_FATAL(ret == 0 && vdev);
+	SPDK_CU_ASSERT_FATAL(!strcmp(spdk_cpuset_fmt(spdk_thread_get_cpumask(vdev->thread)), "3"));
+	cleanup_vdev(vdev);
+
+	/* Create device with incorrect cpumask outside of application cpumask */
+	ret = alloc_vdev(&vdev, "vdev_name_0", "0xf0");
+	SPDK_CU_ASSERT_FATAL(ret != 0);
+
+	/* Create device with incorrect cpumask partially outside of application cpumask */
+	ret = alloc_vdev(&vdev, "vdev_name_0", "0xff");
+	SPDK_CU_ASSERT_FATAL(ret != 0);
 
 	/* Create device with no name */
-	ret = alloc_vdev(&vdev, NULL, "0x1");
+	ret = alloc_vdev(&vdev, NULL, NULL);
-	CU_ASSERT(ret != 0);
-
-	/* Create device with incorrect cpumask */
-	ret = alloc_vdev(&vdev, "vdev_name_0", "0x2");
 	CU_ASSERT(ret != 0);
 
 	/* Create device with too long name and path */
 	memset(long_name, 'x', sizeof(long_name));
 	long_name[PATH_MAX - 1] = 0;
 	snprintf(dev_dirname, sizeof(dev_dirname), "some_path/");
-	ret = alloc_vdev(&vdev, long_name, "0x1");
+	ret = alloc_vdev(&vdev, long_name, NULL);
 	CU_ASSERT(ret != 0);
 	dev_dirname[0] = 0;
 
 	/* Create device when device name is already taken */
-	ret = alloc_vdev(&vdev, "vdev_name_0", "0x1");
+	ret = alloc_vdev(&vdev, "vdev_name_0", NULL);
 	SPDK_CU_ASSERT_FATAL(ret == 0 && vdev);
-	ret = alloc_vdev(&vdev2, "vdev_name_0", "0x1");
+	ret = alloc_vdev(&vdev2, "vdev_name_0", NULL);
 	CU_ASSERT(ret != 0);
 	cleanup_vdev(vdev);
 }

test/vhost/common.sh

@@ -8,7 +8,7 @@ VM_DIR=$VHOST_DIR/vms
 TARGET_DIR=$VHOST_DIR/vhost
 VM_PASSWORD="root"
 
-VM_IMAGE=$HOME/spdk_test_image.qcow2
+VM_IMAGE=${VM_IMAGE:-"$DEPENDENCY_DIR/spdk_test_image.qcow2"}
 
 if ! hash $QEMU_IMG_BIN $QEMU_BIN; then
 	error 'QEMU is not installed on this system. Unable to run vhost tests.'
@@ -27,22 +27,17 @@ source $rootdir/test/vhost/common/autotest.config
 function vhosttestinit() {
 	if [ "$TEST_MODE" == "iso" ]; then
 		$rootdir/scripts/setup.sh
+	fi
 
-		# Look for the VM image
+	if [[ -e $VM_IMAGE.gz ]]; then
-		if [[ ! -f $VM_IMAGE ]]; then
+		gzip -dc "$VM_IMAGE.gz" > "$VM_IMAGE"
-			echo "VM image not found at $VM_IMAGE"
-			echo "Download to $HOME? [yn]"
-			read -r download
-			if [ "$download" = "y" ]; then
-				curl https://ci.spdk.io/download/test_resources/vhost_vm_image.tar.gz | tar xz -C $HOME
-			fi
-		fi
 	fi
 
 	# Look for the VM image
-	if [[ "$1" != "--no_vm" ]] && [[ ! -f $VM_IMAGE ]]; then
+	if [[ ! -f $VM_IMAGE ]]; then
-		error "VM image not found at $VM_IMAGE"
+		[[ $1 != "--no_vm" ]] || return 0
-		exit 1
+		echo "$VM_IMAGE is missing" >&2
+		return 1
 	fi
 }
 

test/vhost/fiotest/fio.sh

@@ -119,7 +119,7 @@ rpc_py="$rootdir/scripts/rpc.py -s $(get_vhost_dir 0)/rpc.sock"
 
 for vm_conf in "${vms[@]}"; do
 	IFS=',' read -ra conf <<< "$vm_conf"
-	if [[ x"${conf[0]}" == x"" ]] || ! assert_number ${conf[0]}; then
+	if [[ -z ${conf[0]} ]] || ! assert_number ${conf[0]}; then
 		fail "invalid VM configuration syntax $vm_conf"
 	fi
 

test/vhost/hotplug/common.sh

@@ -88,7 +88,7 @@ function print_test_fio_header() {
 function vms_setup() {
 	for vm_conf in "${vms[@]}"; do
 		IFS=',' read -ra conf <<< "$vm_conf"
-		if [[ x"${conf[0]}" == x"" ]] || ! assert_number ${conf[0]}; then
+		if [[ -z ${conf[0]} ]] || ! assert_number ${conf[0]}; then
 			fail "invalid VM configuration syntax $vm_conf"
 		fi
 

test/vhost/initiator/blockdev.sh

@@ -19,7 +19,7 @@ function err_cleanup() {
 
 # start vhost and configure it
 trap 'err_cleanup; exit 1' SIGINT SIGTERM EXIT
-$SPDK_BIN_DIR/vhost &
+$SPDK_BIN_DIR/vhost -m 0xf &
 vhost_pid=$!
 waitforlisten $vhost_pid
 
@@ -35,17 +35,22 @@ rpc_cmd vhost_scsi_controller_add_target naa.Nvme0n1_scsi0.0 0 Nvme0n1p0
 rpc_cmd vhost_scsi_controller_add_target naa.Nvme0n1_scsi0.0 1 Nvme0n1p1
 rpc_cmd vhost_scsi_controller_add_target naa.Nvme0n1_scsi0.0 2 Nvme0n1p2
 rpc_cmd vhost_scsi_controller_add_target naa.Nvme0n1_scsi0.0 3 Nvme0n1p3
+[[ "$(rpc_cmd vhost_get_controllers -n naa.Nvme0n1_scsi0.0 | jq -r '.[].cpumask')" == "0xf" ]]
 
-rpc_cmd vhost_create_blk_controller naa.Nvme0n1_blk0.0 Nvme0n1p4
+rpc_cmd vhost_create_blk_controller naa.Nvme0n1_blk0.0 Nvme0n1p4 --cpumask 0xf
-rpc_cmd vhost_create_blk_controller naa.Nvme0n1_blk1.0 Nvme0n1p5
+[[ "$(rpc_cmd vhost_get_controllers -n naa.Nvme0n1_blk0.0 | jq -r '.[].cpumask')" == "0xf" ]]
+rpc_cmd vhost_create_blk_controller naa.Nvme0n1_blk1.0 Nvme0n1p5 --cpumask 0x1
+[[ "$(rpc_cmd vhost_get_controllers -n naa.Nvme0n1_blk1.0 | jq -r '.[].cpumask')" == "0x1" ]]
 
 rpc_cmd bdev_malloc_create 128 512 --name Malloc0
-rpc_cmd vhost_create_scsi_controller naa.Malloc0.0
+rpc_cmd vhost_create_scsi_controller naa.Malloc0.0 --cpumask 0x2
 rpc_cmd vhost_scsi_controller_add_target naa.Malloc0.0 0 Malloc0
+[[ "$(rpc_cmd vhost_get_controllers -n naa.Malloc0.0 | jq -r '.[].cpumask')" == "0x2" ]]
 
 rpc_cmd bdev_malloc_create 128 4096 --name Malloc1
-rpc_cmd vhost_create_scsi_controller naa.Malloc1.0
+rpc_cmd vhost_create_scsi_controller naa.Malloc1.0 --cpumask 0xc
 rpc_cmd vhost_scsi_controller_add_target naa.Malloc1.0 0 Malloc1
+[[ "$(rpc_cmd vhost_get_controllers -n naa.Malloc1.0 | jq -r '.[].cpumask')" == "0xc" ]]
 
 # start a dummy app, create vhost bdevs in it, then dump the config for FIO
 $SPDK_BIN_DIR/spdk_tgt -r /tmp/spdk2.sock -g &

test/vhost/manual.sh

@@ -6,7 +6,7 @@ source $rootdir/test/common/autotest_common.sh
 source $rootdir/test/vhost/common.sh
 
 CENTOS_VM_IMAGE="/home/sys_sgsw/spdk_vhost_CentOS_vm_image.qcow2"
-DEFAULT_FIO_BIN="/home/sys_sgsw/fio_ubuntu"
+DEFAULT_FIO_BIN="$DEPENDENCY_DIR/fio"
 CENTOS_FIO_BIN="/home/sys_sgsw/fio_ubuntu_bak"
 
 case $1 in
@@ -21,7 +21,7 @@ case $1 in
 		echo "  -h |--help                           prints this message"
 		echo ""
 		echo "Environment:"
-		echo "  VM_IMAGE        path to QCOW2 VM image used during test (default: $HOME/spdk_test_image.qcow2)"
+		echo "  VM_IMAGE        path to QCOW2 VM image used during test (default: $DEPENDENCY_DIR/spdk_test_image.qcow2)"
 		echo ""
 		echo "Tests are performed only on Linux machine. For other OS no action is performed."
 		echo ""
@@ -38,13 +38,7 @@ if [[ $(uname -s) != Linux ]]; then
 fi
 
 : ${FIO_BIN="$DEFAULT_FIO_BIN"}
-
+vhosttestinit
-if [[ ! -r "${VM_IMAGE}" ]]; then
-	echo ""
-	echo "ERROR: VM image '${VM_IMAGE}' does not exist."
-	echo ""
-	exit 1
-fi
 
 DISKS_NUMBER=$(lspci -mm -n | grep 0108 | tr -d '"' | awk -F " " '{print "0000:"$1}' | wc -l)
 

test/vhost/other/negative.sh

@@ -63,7 +63,7 @@ notice "==============="
 notice ""
 notice "running SPDK"
 notice ""
-vhost_run 0
+vhost_run 0 "-m 0xf"
 notice ""
 rpc_py="$rootdir/scripts/rpc.py -s $(get_vhost_dir 0)/rpc.sock"
 $rpc_py bdev_malloc_create -b Malloc0 128 4096
@@ -88,8 +88,13 @@ if $rpc_py vhost_delete_controller unk0 > /dev/null; then
 fi
 
 # SCSI
-notice "Trying to create scsi controller with incorrect cpumask"
+notice "Trying to create scsi controller with incorrect cpumask outside of application cpumask"
-if $rpc_py vhost_create_scsi_controller vhost.invalid.cpumask --cpumask 0x2; then
+if $rpc_py vhost_create_scsi_controller vhost.invalid.cpumask --cpumask 0xf0; then
+	error "Creating scsi controller with incorrect cpumask succeeded, but it shouldn't"
+fi
+
+notice "Trying to create scsi controller with incorrect cpumask partially outside of application cpumask"
+if $rpc_py vhost_create_scsi_controller vhost.invalid.cpumask --cpumask 0xff; then
 	error "Creating scsi controller with incorrect cpumask succeeded, but it shouldn't"
 fi
 
@@ -167,8 +172,13 @@ notice "Re-adding device 0 to naa.0"
 $rpc_py vhost_scsi_controller_add_target naa.0 0 Malloc0
 
 # BLK
-notice "Trying to create block controller with incorrect cpumask"
+notice "Trying to create block controller with incorrect cpumask outside of application cpumask"
-if $rpc_py vhost_create_blk_controller vhost.invalid.cpumask Malloc0 --cpumask 0x2; then
+if $rpc_py vhost_create_blk_controller vhost.invalid.cpumask Malloc0 --cpumask 0xf0; then
+	error "Creating block controller with incorrect cpumask succeeded, but it shouldn't"
+fi
+
+notice "Trying to create block controller with incorrect cpumask partially outside of application cpumask"
+if $rpc_py vhost_create_blk_controller vhost.invalid.cpumask Malloc0 --cpumask 0xff; then
 	error "Creating block controller with incorrect cpumask succeeded, but it shouldn't"
 fi
 

test/vhost/perf_bench/vhost_perf.sh

@@ -53,7 +53,7 @@ function usage() {
 	echo "    --vm-memory=INT         Amount of RAM memory (in MB) to pass to a single VM."
 	echo "                            Default: 2048 MB"
 	echo "    --vm-image=PATH         OS image to use for running the VMs."
-	echo "                            Default: \$HOME/spdk_test_image.qcow2"
+	echo "                            Default: \$DEPENDENCY_DIR/spdk_test_image.qcow2"
 	echo "    --vm-sar-enable         Measure CPU utilization in guest VMs using sar."
 	echo "    --host-sar-enable       Measure CPU utilization on host using sar."
 	echo "    --sar-delay=INT         Wait for X seconds before starting SAR measurement. Default: 0."

test/vhost/vhost.sh

@@ -13,17 +13,11 @@ if [[ $(uname -s) != Linux ]]; then
 fi
 
 CENTOS_VM_IMAGE="/home/sys_sgsw/spdk_vhost_CentOS_vm_image.qcow2"
-DEFAULT_FIO_BIN="/home/sys_sgsw/fio_ubuntu"
+DEFAULT_FIO_BIN="$DEPENDENCY_DIR/fio"
 CENTOS_FIO_BIN="/home/sys_sgsw/fio_ubuntu_bak"
 
 : ${FIO_BIN="$DEFAULT_FIO_BIN"}
-
+vhosttestinit
-if [[ ! -r "${VM_IMAGE}" ]]; then
-	echo ""
-	echo "ERROR: VM image '${VM_IMAGE}' does not exist."
-	echo ""
-	exit 1
-fi
 
 DISKS_NUMBER=$(lspci -mm -n | grep 0108 | tr -d '"' | awk -F " " '{print "0000:"$1}' | wc -l)
 

test/vhost/windows/windows_scsi_compliance.sh

@@ -9,7 +9,7 @@ source $rootdir/test/vhost/common.sh
 # and OS Version: 6.3.9600 N/A Build 9600
 # In order to run this test with windows vm
 # windows virtio scsi driver must be installed
-WINDOWS_IMG="/home/sys_sgsw/windows_scsi_compliance/windows_vm_image.qcow2"
+WINDOWS_IMG="$DEPENDENCY_DIR/windows_scsi_compliance/windows_vm_image.qcow2"
 aio_file="$SPDK_TEST_STORAGE/aio_disk"
 ssh_pass=""
 vm_num=1