dpdkbuild: add support for DPDK 20.05

EAL got a new dependency in 20.05: rte_telemetry.

Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>

(cherry picked from commit 3b99a376d5595f4d6a458e8221bfd0b6b6f07b83)
Change-Id: I43df7afe9a84e88f034a7f87fc6a299f0bbd8bac
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2899
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>

CHANGELOG.md

@@ -1,6 +1,15 @@
 # Changelog
 
-## v20.04: (Upcoming Release)
+## v20.04.2: (Upcoming Release)
+
+## v20.04.1:
+
+### dpdk
+
+Updated DPDK submodule to DPDK 19.11.2, which includes fixes for DPDK vulnerabilities:
+CVE-2020-10722, CVE-2020-10723, CVE-2020-10724, CVE-2020-10725, CVE-2020-10724.
+
+## v20.04:
 
 ### configuration
 
@@ -8,6 +17,24 @@ Legacy INI style configuration for SPDK applications has been deprecated and wil
 removed in future release.
 Please switch to JSON-RPC configuration files and/or RPC driven run-time configuration.
 
+### copy
+
+The copy engine library, modules and public APIs have been renamed. Use of the word `copy`
+has been replaced with the word `accel` short for accelerator in preparation for adding new
+capabilities in the future. Additionally, APIs for what was previously called the `memcpy`
+engine have been renamed to identify the engine as a software accelerator.
+
+### crypto
+
+Support for AES_XTS was added for the QAT polled mode driver (pmd).  The create RPC
+`bdev_crypto_create` has 2 new optional parameters: cipher and key2. Cipher can be either
+AES_CBC (default) or AES_XTS. AES_XTS is only valid when using the QAT polled mode driver.
+The key2 parameter is the second key required for AES_XTS.
+
+### event
+
+Reactors now accumulate CPU stats and they are retrieved by the RPC `framework_get_reactors`.
+
 ### idxd
 
 IDXD support was added in the form of a low level library that can directly
@@ -16,51 +43,77 @@ with the generic accel framework API. IDXD is the first in a family of offload
 engines that share the same interface, specifically DSA is added here. More info
 can be found here: https://01.org/blogs/2019/introducing-intel-data-streaming-accelerator
 
-Much of the implementation models IOAT, however their low level interfaces are very
+Much of the implementation models IOAT, however the low level interfaces are very
 different. The RPC to enable IDXD requires a configuration number as well. The
-code includes 2 pre-defined configurations of IDXD groups/work queues/engines. A future
+code includes two pre-defined configurations of IDXD groups/work queues/engines. A future
 version will provide an interface to allow for setting of individual configuration
 parameters.
 
 IDXD is not yet available so this feature should be considered experimental. It will
 be built up with additional documentation as an ongoing activity.
 
-### ocf
+### iscsi
-
-Update OCF submodule to OCF v20.03
-
-New version of OCF comes with API changes and bug fixes
-
-### nvme
-
-Export internal nvme_ctrlr_cmd_security_receive/send() APIs as public APIs with "spdk_"
-prefix.
-
-Added `priority` field in `spdk_nvme_transport_id`, this field is used to specify the priority
-of the NVMe-oF connection, and currently it is used for NVMe-oF tcp connection.
-
-### copy
-
-The copy engine library, modules and public APIs have been renamed. Use of the word `copy`
-has been replaced with the word `accel` short for accelerator in preparation for adding new
-capabilities in the future. Additionally, APIs for what was previously called the `memcpy`
-engine have been renamed to identify the engine as a software accelerator.
-
-### event
-
-Reactor now accumulates CPU stats and they are retrieved by the RPC `framework_get_reactors`.
-
-### iSCSI
 
 The iSCSI target now creates a lightweight thread per poll group instead of assuming a pool
 of lightweight threads already exist at start up time. A poll group is a collection of
 unrelated iSCSI connections. Each poll group is only accessed from the associated
 lightweight thread.
 
-### vmd
+### ftl
 
-A new function, `spdk_vmd_fini`, has been added. It releases all resources acquired by the VMD
+Several changes have been made to the `spdk_ftl_conf`, `spdk_ftl_dev_init_ops`, and
-library through the `spdk_vmd_init` call.
+`spdk_ftl_attrs` structs. Please see `include/spdk/ftl.h` for more details.
+
+### miscellaneous
+
+The `--json-ignore-init-errors` command line parameter has been added to ignore
+initialization errors on JSON config load.
+
+The public header file io_channel.h has been removed. Please use thread.h which has the
+exact same API.
+
+### nvme
+
+Exported internal nvme_ctrlr_cmd_security_receive/send() APIs as public APIs with "the spdk_"
+prefix.
+
+Added `priority` field in `spdk_nvme_transport_id`, this field is used to specify the priority
+of the NVMe-oF connection, and currently it is used for NVMe-oF tcp connection.
+
+A new poll group API has been added to allow for pooling of nvme qpairs across a single
+entity which can be polled for completions. This new API consists of the `spdk_nvme_poll_group`
+family of functions. As a result of this new API, all NVMe transports are expected to implement
+several poll group related functions.
+
+A new flag, `create_only`, has been added to the `spdk_nvme_io_qpair_opts` structure. This flag
+allows a user to call `spdk_nvme_ctrlr_get_default_io_qpair` without also connecting the qpair
+within the context of that call.
+
+As a result of the `create_only` flag, two new API functions, `spdk_nvme_ctrlr_connect_io_qpair`
+and `spdk_nvme_ctrlr_disconnect_io_qpair`, have been added to facilitate connecting newly created
+qpairs (for example, after they have been added to a poll group) and disconnecting qpairs without
+destroying them (for example to disconnect a qpair before migrating it to a new poll group and
+reconnecting it).
+
+The functions `spdk_nvme_ctrlr_alloc_cmb_io_buffer` and `spdk_nvme_ctrlr_free_cmb_io_buffer`
+have been changed to `spdk_nvme_ctrlr_map_cmb` and `spdk_nvme_ctrlr_unmap_cmb` respectively.
+
+An additional function, `spdk_nvme_ctrlr_reserve_cmb`, has been added to facilitate reserving
+the entire size of the controller memory buffer for data transfer.
+
+### nvme_cuse
+
+`spdk_nvme_cuse_get_ctrlr_name` now takes two additional parameters, `char *name` which
+stores the pointer to the controller name, and `size_t *size` which stores the length of
+the name. The return type has also been changed from char * to int.
+
+`spdk_nvme_cuse_get_ns_name` now takes two additional parameters, `char *name` which
+stores the pointer to the namespace name, and `size_t *size` which stores the length of
+the name. The return type has also been changed from char * to int.
+
+### nvme_opal
+
+Several public OPAL structure definitions have been changed since the last release.
 
 ### nvmf
 
@@ -71,36 +124,49 @@ of lightweight threads already exist at start up time. A poll group is a collect
 unrelated NVMe-oF connections. Each poll group is only accessed from the associated
 lightweight thread.
 
-### Miscellaneous
+A new struct, `spdk_nvmf_subsystem_listener`, has been added to encapsulate the subsystem specific
+nature of a listener object.
 
-`--json-ignore-init-errors` command line param has been added to ignore initialization errors
+`spdk_nvmf_tgt_listen` no longer accepts a callback function or argument. It also returns an
-on JSON config load.
+int to indicate the status of the listen call.
 
-The public header file io_channel.h has been removed. Please use thread.h which has the
+The execution of `spdk_nvme_poll_group_destroy` is now asynchronous and the function accepts
-exact same API.
+a cb_fn and cb_arg to call upon completion.
 
-### crypto
+The execution of `spdk_nvmf_subsystem_add_listener` is now asynchronous and the function accepts
+a cb_fn and cb_arg to call upon completion.
 
-Support for AES_XTS was added for the QAT polled mode driver (pmd).  The create RPC
+The `nvmf_transport.h` header has been made public to allow custom NVMe-oF transports to integrate
-`bdev_crypto_create` has 2 new optional parameters: cipher and key2. Cipher can be either
+with NVMe-oF libraries without using internal APIs.
-AES_CBC (default) or AES_XTS. AES_XTS is only valid when using the QAT polled mode driver.
-The key2 parameter is the second key required for AES_XTS.
 
-### util
+### ocf
 
-New functions `spdk_sn32_lt` and `spdk_sn32_gt` have been added. They compare two sequence
+Updated the OCF submodule to OCF v20.03
-numbers based on serial number arithmetic.
+
+New version of OCF comes with API changes and bug fixes
 
 ### rpc
 
 A new RPC `thread_set_cpumask` has been added to set the cpumask of the thread
 to the specified value.
 
-A new RPC `thread_get_pollers` has been added to retrieve pollers of SPDK threads.
+A new RPC `thread_get_pollers` has been added to retrieve pollers from SPDK threads.
 
-A new RPC `thread_get_io_channels` has been added to retrieve I/O channels of SPDK threads.
+A new RPC `thread_get_io_channels` has been added to retrieve I/O channels from SPDK threads.
 
-A new RPC `bdev_rbd_resize` has been added to resize the Ceph RBD bdev.
+A new RPC `bdev_rbd_resize` has been added to resize Ceph RBD bdevs.
+
+### sock
+
+The `spdk_sock_set_priority` function has been removed since the feature to set the sock priority
+will be contained in two new functions, i.e., `spdk_sock_listen_ext` and `spdk_sock_connect_ext`.
+Users may now specify the priority of the socket in the opts that they want to use.
+
+### spdk_top
+
+A new application, `spdk_top`, has been added which allows users to monitor resource consumption
+by a running SPDK application. More information on this application can be found in
+`app/spdk_top/README`.
 
 ### thread
 
@@ -113,7 +179,7 @@ Current SPDK operation types are `SPDK_THREAD_OP_NEW` and `SPDK_THREAD_OP_RESCHE
 The operation `SPDK_THREAD_OP_NEW` is called each time a new thread is created.
 The operation `SPDK_THREAD_OP_RESCHED` is called when SPDK thread needs to be rescheduled.
 
-An unique ID has been added for each created SPDK thread, it is retrieved by a new function
+A unique ID has been added for each created SPDK thread, it is retrieved by a new function
 `spdk_thread_get_id`, and the SPDK thread which has the specific ID is got by
 a new function `spdk_thread_get_by_id`.
 
@@ -130,15 +196,19 @@ threads configuration, and a new function `spdk_thread_get_last_tsc` has been ad
 Voluntary termination of SPDK thread has been supported by refining the functions `spdk_thread_exit`
 and `spdk_thread_poll`.
 
+### util
+
+New functions `spdk_sn32_lt` and `spdk_sn32_gt` have been added. They compare two sequence
+numbers based on serial number arithmetic.
+
 ### vhost
 
 Poll groups per session have been replaced by SPDK threads per vhost controller.
 
-### sock
+### vmd
 
-Remove `spdk_sock_set_priority` function since the feature to set the sock priority will be
+A new function, `spdk_vmd_fini`, has been added. It releases all resources acquired by the VMD
-contained in two new functions, i.e., `spdk_sock_listen_ext` and `spdk_sock_connect_ext`.
+library through the `spdk_vmd_init` call.
-Users may now specify the priority of the socket in the opts that they want to use.
 
 ## v20.01
 

app/spdk_top/spdk_top.c

@@ -944,6 +944,10 @@ store_last_run_counter(const char *poller_name, uint64_t thread_id, uint64_t las
 	}
 
 	history = calloc(1, sizeof(*history));
+	if (history == NULL) {
+		fprintf(stderr, "Unable to allocate a history object in store_last_run_counter.\n");
+		return;
+	}
 	history->poller_name = strdup(poller_name);
 	history->thread_id = thread_id;
 	history->last_run_counter = last_run_counter;
@@ -983,8 +987,10 @@ sort_pollers(const void *p1, const void *p2, void *arg)
 			return strcmp(poller1->thread_name, poller2->thread_name);
 		case 3: /* Sort by run counter */
 			last_run_counter = get_last_run_counter(poller1->name, poller1->thread_id);
+			assert(last_run_counter != NULL);
 			count1 = poller1->run_count - *last_run_counter;
 			last_run_counter = get_last_run_counter(poller2->name, poller2->thread_id);
+			assert(last_run_counter != NULL);
 			count2 = poller2->run_count - *last_run_counter;
 			break;
 		case 4: /* Sort by period */
@@ -1021,6 +1027,7 @@ copy_pollers(struct rpc_pollers *pollers, uint64_t pollers_count, enum spdk_poll
 				last_run_counter = get_last_run_counter(pollers->pollers[i].name, thread->id);
 			}
 
+			assert(last_run_counter != NULL);
 			*last_run_counter = pollers->pollers[i].run_count;
 		}
 		pollers_info[*current_count] = &pollers->pollers[i];
@@ -1113,6 +1120,7 @@ refresh_pollers_tab(uint8_t current_page)
 
 		if (!col_desc[3].disabled) {
 			last_run_counter = get_last_run_counter(pollers[i]->name, pollers[i]->thread_id);
+			assert(last_run_counter != NULL);
 
 			snprintf(run_count, MAX_TIME_STR_LEN, "%" PRIu64, pollers[i]->run_count - *last_run_counter);
 			print_max_len(g_tabs[POLLERS_TAB], TABS_DATA_START_ROW + item_index, col,
@@ -1377,6 +1385,10 @@ draw_filtering_menu(uint8_t position, WINDOW *filter_win, uint8_t tab, MENU **my
 	elements = i;
 
 	my_items = (ITEM **)calloc(elements * WINDOW_COLUMNS + ADDITIONAL_ELEMENTS, sizeof(ITEM *));
+	if (my_items == NULL) {
+		fprintf(stderr, "Unable to allocate an item list in draw_filtering_menu.\n");
+		return NULL;
+	}
 
 	for (i = 0; i < elements * 2; i++) {
 		my_items[i] = new_item(col_desc[i / WINDOW_COLUMNS].name, NULL);
@@ -1441,7 +1453,7 @@ filter_columns(uint8_t tab)
 	PANEL *filter_panel;
 	WINDOW *filter_win;
 	ITEM **my_items;
-	MENU *my_menu;
+	MENU *my_menu = NULL;
 	int i, c, elements;
 	bool stop_loop = false;
 	ITEM *cur;
@@ -1552,6 +1564,10 @@ change_sorting(uint8_t tab)
 	elements = i;
 
 	my_items = (ITEM **)calloc(elements + 1, sizeof(ITEM *));
+	if (my_items == NULL) {
+		fprintf(stderr, "Unable to allocate an item list in change_sorting.\n");
+		return;
+	}
 
 	for (i = 0; i < elements; ++i) {
 		my_items[i] = new_item(g_col_desc[tab][i].name, NULL);

autopackage.sh

@@ -35,7 +35,12 @@ fi
 
 timing_enter build_release
 
-./configure $(get_config_params) --disable-debug --enable-lto
+if [ $(uname -s) = Linux ]; then
+	./configure $(get_config_params) --disable-debug --enable-lto --disable-unit-tests
+else
+	# LTO needs a special compiler to work on BSD.
+	./configure $(get_config_params) --disable-debug
+fi
 $MAKE ${MAKEFLAGS}
 $MAKE ${MAKEFLAGS} clean
 

configure

@@ -494,6 +494,11 @@ if [[ "${CONFIG[ISAL]}" = "n" ]] && [[ "${CONFIG[REDUCE]}" = "y" ]]; then
 		exit 1
 fi
 
+if [[ "${CONFIG[LTO]}" = "y" ]] && [[ "${CONFIG[UNIT_TESTS]}" = "y"  ]]; then
+		echo "ERROR Conflicting options: --enable-lto is not compatible with --enable-unit-tests."
+		exit 1
+fi
+
 if [ -z "${CONFIG[ENV]}" ]; then
 	CONFIG[ENV]=$rootdir/lib/env_dpdk
 	echo "Using default SPDK env in ${CONFIG[ENV]}"

dpdk

@@ -1 +1 @@
-Subproject commit 76f9669b1f4cdd52bae2c3408ad129272bf99d1f
+Subproject commit ef71bfaface10cc19b75e45d3158ab71a788e3a9

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/conf/conf.c

@@ -71,7 +71,6 @@ static struct spdk_conf *default_config = NULL;
 struct spdk_conf *
 spdk_conf_allocate(void)
 {
-	SPDK_ERRLOG("INI configuration has been deprecated and will be removed in a future release. Please switch to JSON-RPC.\n");
 	return calloc(1, sizeof(struct spdk_conf));
 }
 
@@ -611,6 +610,7 @@ spdk_conf_read(struct spdk_conf *cp, const char *file)
 	if (file == NULL || file[0] == '\0') {
 		return -1;
 	}
+	SPDK_ERRLOG("INI configuration has been deprecated and will be removed in a future release. Please switch to JSON-RPC.\n");
 
 	fp = fopen(file, "r");
 	if (fp == NULL) {

lib/env_dpdk/env.mk

@@ -78,6 +78,11 @@ ifneq (, $(wildcard $(DPDK_ABS_DIR)/lib/librte_bus_pci.*))
 DPDK_LIB_LIST += rte_bus_pci
 endif
 
+# DPDK 20.05 eal dependency
+ifneq (, $(wildcard $(DPDK_ABS_DIR)/lib/librte_telemetry.*))
+DPDK_LIB_LIST += rte_telemetry
+endif
+
 # There are some complex dependencies when using crypto, reduce or both so
 # here we add the feature specific ones and set a flag to add the common
 # ones after that.

lib/env_dpdk/env_internal.h

@@ -77,8 +77,8 @@ void pci_driver_register(struct spdk_pci_driver *driver);
 int pci_device_init(struct rte_pci_driver *driver, struct rte_pci_device *device);
 int pci_device_fini(struct rte_pci_device *device);
 
-void pci_init(void);
+void pci_env_init(void);
-void pci_fini(void);
+void pci_env_fini(void);
 int mem_map_init(bool legacy_mem);
 int vtophys_init(void);
 

lib/env_dpdk/init.c

@@ -499,7 +499,7 @@ spdk_env_dpdk_post_init(bool legacy_mem)
 {
 	int rc;
 
-	pci_init();
+	pci_env_init();
 
 	rc = mem_map_init(legacy_mem);
 	if (rc < 0) {
@@ -519,7 +519,7 @@ spdk_env_dpdk_post_init(bool legacy_mem)
 void
 spdk_env_dpdk_post_fini(void)
 {
-	pci_fini();
+	pci_env_fini();
 
 	spdk_free_args(g_eal_cmdline, g_eal_cmdline_argcount);
 }

lib/env_dpdk/pci.c

@@ -259,7 +259,7 @@ _get_alarm_thread_cb(void *unused)
 }
 
 void
-pci_init(void)
+pci_env_init(void)
 {
 #if RTE_VERSION >= RTE_VERSION_NUM(18, 11, 0, 0)
 	struct spdk_pci_driver *driver;
@@ -298,7 +298,7 @@ pci_init(void)
 }
 
 void
-pci_fini(void)
+pci_env_fini(void)
 {
 	struct spdk_pci_device *dev;
 	char bdf[32];

lib/ftl/ftl_core.c

@@ -1611,6 +1611,7 @@ ftl_update_l2p(struct spdk_ftl_dev *dev, const struct ftl_wbuf_entry *entry,
 	struct ftl_wbuf_entry *prev;
 	struct ftl_band *band;
 	int valid;
+	bool io_weak = entry->io_flags & FTL_IO_WEAK;
 
 	prev_addr = ftl_l2p_get(dev, entry->lba);
 	if (ftl_addr_invalid(prev_addr)) {
@@ -1618,14 +1619,7 @@ ftl_update_l2p(struct spdk_ftl_dev *dev, const struct ftl_wbuf_entry *entry,
 		return;
 	}
 
-	/* If the L2P's physical address is different than what we expected we don't need to */
-	/* do anything (someone's already overwritten our data). */
-	if ((entry->io_flags & FTL_IO_WEAK) && !ftl_addr_cmp(prev_addr, entry->addr)) {
-		return;
-	}
-
 	if (ftl_addr_cached(prev_addr)) {
-		assert(!(entry->io_flags & FTL_IO_WEAK));
 		prev = ftl_get_entry_from_addr(dev, prev_addr);
 		pthread_spin_lock(&prev->lock);
 
@@ -1634,12 +1628,33 @@ ftl_update_l2p(struct spdk_ftl_dev *dev, const struct ftl_wbuf_entry *entry,
 		prev_addr = ftl_l2p_get(dev, entry->lba);
 
 		/* If the entry is no longer in cache, another write has been */
-		/* scheduled in the meantime, so we have to invalidate its LBA */
+		/* scheduled in the meantime, so we can return to evicted path */
 		if (!ftl_addr_cached(prev_addr)) {
-			ftl_invalidate_addr(dev, prev_addr);
+			pthread_spin_unlock(&prev->lock);
+			goto evicted;
 		}
 
-		/* If previous entry is part of cache, remove and invalidate it */
+		/*
+		 * Relocating block could still reside in cache due to fact that write
+		 * buffers are independent for each IO channel and enough amount of data
+		 * (write unit size) must be collected before it will be submitted to lower
+		 * layer.
+		 * When previous entry wasn't overwritten invalidate old address and entry.
+		 * Otherwise skip relocating block.
+		 */
+		if (io_weak &&
+		    /* Check if prev_addr was updated in meantime */
+		    !(ftl_addr_cmp(prev_addr, ftl_get_addr_from_entry(prev)) &&
+		      /* Check if relocating address it the same as in previous entry */
+		      ftl_addr_cmp(prev->addr, entry->addr))) {
+			pthread_spin_unlock(&prev->lock);
+			return;
+		}
+
+		/*
+		 * If previous entry is part of cache and was written into disk remove
+		 * and invalidate it
+		 */
 		if (prev->valid) {
 			ftl_invalidate_addr(dev, prev->addr);
 			prev->valid = false;
@@ -1650,6 +1665,15 @@ ftl_update_l2p(struct spdk_ftl_dev *dev, const struct ftl_wbuf_entry *entry,
 		return;
 	}
 
+evicted:
+	/*
+	 *  If the L2P's physical address is different than what we expected we don't need to
+	 *  do anything (someone's already overwritten our data).
+	 */
+	if (io_weak && !ftl_addr_cmp(prev_addr, entry->addr)) {
+		return;
+	}
+
 	/* Lock the band containing previous physical address. This assures atomic changes to */
 	/* the L2P as wall as metadata. The valid bits in metadata are used to */
 	/* check weak writes validity. */
@@ -1660,7 +1684,7 @@ ftl_update_l2p(struct spdk_ftl_dev *dev, const struct ftl_wbuf_entry *entry,
 
 	/* If the address has been invalidated already, we don't want to update */
 	/* the L2P for weak writes, as it means the write is no longer valid. */
-	if (!(entry->io_flags & FTL_IO_WEAK) || valid) {
+	if (!io_weak || valid) {
 		ftl_l2p_set(dev, entry->lba, addr);
 	}
 

lib/ftl/ftl_init.c

@@ -923,6 +923,11 @@ ftl_dev_get_zone_info_cb(struct spdk_bdev_io *bdev_io, bool success, void *cb_ar
 				    "zone id: %"PRIu64"\n", init_ctx->zone_id);
 		}
 
+		/* Set write pointer to the last block plus one for zone in full state */
+		if (zone->info.state == SPDK_BDEV_ZONE_STATE_FULL) {
+			zone->info.write_pointer = zone->info.zone_id + zone->info.capacity;
+		}
+
 		if (zone->info.state != SPDK_BDEV_ZONE_STATE_OFFLINE) {
 			band->num_zones++;
 			CIRCLEQ_INSERT_TAIL(&band->zones, zone, circleq);

lib/nvme/nvme_rdma.c

@@ -1629,7 +1629,7 @@ static void
 nvme_rdma_ctrlr_disconnect_qpair(struct spdk_nvme_ctrlr *ctrlr, struct spdk_nvme_qpair *qpair)
 {
 	struct nvme_rdma_qpair *rqpair = nvme_rdma_qpair(qpair);
-	struct nvme_rdma_ctrlr *rctrlr;
+	struct nvme_rdma_ctrlr *rctrlr = NULL;
 	struct nvme_rdma_cm_event_entry *entry, *tmp;
 
 	nvme_rdma_unregister_mem(rqpair);
@@ -1658,8 +1658,7 @@ nvme_rdma_ctrlr_disconnect_qpair(struct spdk_nvme_ctrlr *ctrlr, struct spdk_nvme
 
 	if (rqpair->cm_id) {
 		rdma_disconnect(rqpair->cm_id);
-		if (qpair->ctrlr != NULL) {
+		if (rctrlr != NULL) {
-
 			if (nvme_rdma_process_event(rqpair, rctrlr->cm_channel, RDMA_CM_EVENT_DISCONNECTED)) {
 				SPDK_DEBUGLOG(SPDK_LOG_NVME, "Target did not respond to qpair disconnect.\n");
 			}

lib/vhost/vhost_scsi.c

@@ -1121,18 +1121,17 @@ spdk_vhost_scsi_dev_remove_tgt(struct spdk_vhost_dev *vdev, unsigned scsi_tgt_nu
 	svdev = to_scsi_dev(vdev);
 	assert(svdev != NULL);
 	scsi_dev_state = &svdev->scsi_dev_state[scsi_tgt_num];
+
+	if (scsi_dev_state->status != VHOST_SCSI_DEV_PRESENT) {
+		return -EBUSY;
+	}
+
 	if (scsi_dev_state->dev == NULL || scsi_dev_state->status == VHOST_SCSI_DEV_ADDING) {
 		SPDK_ERRLOG("%s: SCSI target %u is not occupied\n", vdev->name, scsi_tgt_num);
 		return -ENODEV;
 	}
 
 	assert(scsi_dev_state->status != VHOST_SCSI_DEV_EMPTY);
-	if (scsi_dev_state->status != VHOST_SCSI_DEV_PRESENT) {
-		SPDK_WARNLOG("%s: SCSI target %u has been already marked for hotremoval.\n",
-			     vdev->name, scsi_tgt_num);
-		return -EBUSY;
-	}
-
 	ctx = calloc(1, sizeof(*ctx));
 	if (ctx == NULL) {
 		SPDK_ERRLOG("calloc failed\n");

module/bdev/crypto/vbdev_crypto.c

@@ -1204,9 +1204,18 @@ _device_unregister_cb(void *io_device)
 	rte_cryptodev_sym_session_free(crypto_bdev->session_decrypt);
 	rte_cryptodev_sym_session_free(crypto_bdev->session_encrypt);
 	free(crypto_bdev->drv_name);
+	if (crypto_bdev->key) {
+		memset(crypto_bdev->key, 0, strnlen(crypto_bdev->key, (AES_CBC_KEY_LENGTH + 1)));
 		free(crypto_bdev->key);
+	}
+	if (crypto_bdev->key2) {
+		memset(crypto_bdev->key2, 0, strnlen(crypto_bdev->key2, (AES_XTS_KEY_LENGTH + 1)));
 		free(crypto_bdev->key2);
+	}
+	if (crypto_bdev->xts_key) {
+		memset(crypto_bdev->xts_key, 0, strnlen(crypto_bdev->xts_key, (AES_XTS_KEY_LENGTH * 2) + 1));
 		free(crypto_bdev->xts_key);
+	}
 	free(crypto_bdev->crypto_bdev.name);
 	free(crypto_bdev);
 }

module/sock/uring/uring.c

@@ -58,6 +58,7 @@
 enum spdk_sock_task_type {
 	SPDK_SOCK_TASK_POLLIN = 0,
 	SPDK_SOCK_TASK_WRITE,
+	SPDK_SOCK_TASK_CANCEL,
 };
 
 enum spdk_uring_sock_task_status {
@@ -82,6 +83,7 @@ struct spdk_uring_sock {
 	struct spdk_uring_sock_group_impl	*group;
 	struct spdk_uring_task			write_task;
 	struct spdk_uring_task			pollin_task;
+	struct spdk_uring_task			cancel_task;
 	int					outstanding_io;
 	struct spdk_pipe			*recv_pipe;
 	void					*recv_buf;
@@ -858,6 +860,26 @@ _sock_prep_pollin(struct spdk_sock *_sock)
 	task->status = SPDK_URING_SOCK_TASK_IN_PROCESS;
 }
 
+static void
+_sock_prep_cancel_task(struct spdk_sock *_sock, void *user_data)
+{
+	struct spdk_uring_sock *sock = __uring_sock(_sock);
+	struct spdk_uring_task *task = &sock->cancel_task;
+	struct io_uring_sqe *sqe;
+
+	if (task->status == SPDK_URING_SOCK_TASK_IN_PROCESS) {
+		return;
+	}
+
+	assert(sock->group != NULL);
+	sock->group->io_queued++;
+
+	sqe = io_uring_get_sqe(&sock->group->uring);
+	io_uring_prep_cancel(sqe, user_data, 0);
+	io_uring_sqe_set_data(sqe, task);
+	task->status = SPDK_URING_SOCK_TASK_IN_PROCESS;
+}
+
 static int
 spdk_sock_uring_group_reap(struct spdk_uring_sock_group_impl *group, int max, int max_read_events,
 			   struct spdk_sock **socks)
@@ -919,7 +941,6 @@ spdk_sock_uring_group_reap(struct spdk_uring_sock_group_impl *group, int max, in
 			if (spdk_unlikely(task->sock->outstanding_io > 0 &&
 					  TAILQ_EMPTY(&sock->base.pending_reqs))) {
 				if (--sock->outstanding_io == 0) {
-					sock->group = NULL;
 					/* Just for sock close case */
 					if (sock->base.flags.closed) {
 						spdk_uring_sock_close(&sock->base);
@@ -927,6 +948,11 @@ spdk_sock_uring_group_reap(struct spdk_uring_sock_group_impl *group, int max, in
 				}
 			}
 
+			break;
+		case SPDK_SOCK_TASK_CANCEL:
+			if ((status == 0) && (sock->outstanding_io > 0)) {
+				sock->outstanding_io--;
+			}
 			break;
 		default:
 			SPDK_UNREACHABLE();
@@ -1151,32 +1177,15 @@ spdk_uring_sock_group_impl_add_sock(struct spdk_sock_group_impl *_group,
 	sock->pollin_task.sock = sock;
 	sock->pollin_task.type = SPDK_SOCK_TASK_POLLIN;
 
-	return 0;
+	sock->cancel_task.sock = sock;
-}
+	sock->cancel_task.type = SPDK_SOCK_TASK_CANCEL;
 
-static int
+	/* switched from another polling group due to scheduling */
-spdk_uring_sock_group_impl_remove_sock(struct spdk_sock_group_impl *_group,
+	if (spdk_unlikely(sock->recv_pipe != NULL &&
-				       struct spdk_sock *_sock)
+			  (spdk_pipe_reader_bytes_available(sock->recv_pipe) > 0))) {
-{
+		assert(sock->pending_recv == false);
-	struct spdk_uring_sock *sock = __uring_sock(_sock);
+		sock->pending_recv = true;
-	struct spdk_uring_sock_group_impl *group = __uring_group_impl(_group);
+		TAILQ_INSERT_TAIL(&group->pending_recv, sock, link);
-
-	if (sock->write_task.status != SPDK_URING_SOCK_TASK_NOT_IN_USE) {
-		sock->outstanding_io++;
-	}
-
-	if (sock->pollin_task.status != SPDK_URING_SOCK_TASK_NOT_IN_USE) {
-		sock->outstanding_io++;
-	}
-
-	if ((sock->recv_pipe != NULL) &&
-	    spdk_pipe_reader_bytes_available(sock->recv_pipe) > 0) {
-		TAILQ_REMOVE(&group->pending_recv, sock, link);
-		sock->pending_recv = false;
-	}
-
-	if (!sock->outstanding_io) {
-		sock->group = NULL;
 	}
 
 	return 0;
@@ -1220,6 +1229,41 @@ spdk_uring_sock_group_impl_poll(struct spdk_sock_group_impl *_group, int max_eve
 	return count;
 }
 
+static int
+spdk_uring_sock_group_impl_remove_sock(struct spdk_sock_group_impl *_group,
+				       struct spdk_sock *_sock)
+{
+	struct spdk_uring_sock *sock = __uring_sock(_sock);
+	struct spdk_uring_sock_group_impl *group = __uring_group_impl(_group);
+
+	if (sock->write_task.status != SPDK_URING_SOCK_TASK_NOT_IN_USE) {
+		/* For write, we do not need to cancel it */
+		sock->outstanding_io++;
+	}
+
+	if (sock->pollin_task.status != SPDK_URING_SOCK_TASK_NOT_IN_USE) {
+		sock->outstanding_io++;
+		_sock_prep_cancel_task(_sock, &sock->pollin_task);
+	}
+
+
+	/* Since spdk_sock_group_remove_sock is not asynchronous interface, so
+	 * currently can use a while loop here. */
+	while (sock->pollin_task.status != SPDK_URING_SOCK_TASK_NOT_IN_USE) {
+		spdk_uring_sock_group_impl_poll(_group, 32, NULL);
+	}
+
+	if (sock->recv_pipe != NULL) {
+		if (spdk_pipe_reader_bytes_available(sock->recv_pipe) > 0) {
+			TAILQ_REMOVE(&group->pending_recv, sock, link);
+			sock->pending_recv = false;
+		}
+		assert(sock->pending_recv == false);
+	}
+	sock->group = NULL;
+	return 0;
+}
+
 static int
 spdk_uring_sock_group_impl_close(struct spdk_sock_group_impl *_group)
 {
@@ -1242,7 +1286,13 @@ spdk_uring_sock_group_impl_close(struct spdk_sock_group_impl *_group)
 static int
 spdk_uring_sock_flush(struct spdk_sock *_sock)
 {
+	struct spdk_uring_sock *sock = __uring_sock(_sock);
+
+	if (!sock->group) {
 		return _sock_flush_client(_sock);
+	}
+
+	return 0;
 }
 
 static struct spdk_net_impl g_uring_net_impl = {

pkg/spdk.spec

@@ -2,12 +2,12 @@
 %bcond_with doc
 
 Name: spdk
-Version: master
+Version: 20.04.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/v20.04.x.tar.gz
 Summary: Set of libraries and utilities for high performance user-mode storage
 
 %define package_version %{epoch}:%{version}-%{release}

test/common/autotest_common.sh

@@ -827,7 +827,9 @@ function discover_bdevs()
 	local rootdir=$1
 	local config_file=$2
 	local cfg_type=$3
-	local wait_for_spdk_bdev=${4:-30}
+	shift 3
+	local bdev_svc_opts=("$@")
+	local wait_for_spdk_bdev=30
 	local rpc_server=/var/tmp/spdk-discover-bdevs.sock
 
 	if [ ! -e $config_file ]; then
@@ -841,7 +843,7 @@ function discover_bdevs()
 
 	# Start the bdev service to query for the list of available
 	# bdevs.
-	$rootdir/test/app/bdev_svc/bdev_svc -r $rpc_server -i 0 \
+	$rootdir/test/app/bdev_svc/bdev_svc -r $rpc_server -i 0 "${bdev_svc_opts[@]}" \
 		$cfg_type $config_file &>/dev/null &
 	stubpid=$!
 	while ! [ -e /var/run/spdk_bdev0 ]; do

test/spdkcli/virtio.sh

@@ -22,11 +22,11 @@ $spdkcli_job "'/bdevs/malloc create 32 512 Malloc0' 'Malloc0' True
 '/bdevs/malloc create 32 512 Malloc1' 'Malloc1' True
 "
 pci_blk=$(lspci -nn -D | grep '1af4:1001' | head -1 | awk '{print $1;}')
-if [ -n "$pci_blk" ]; then
+if [ -n "$pci_blk" ] && grep -Eq "DRIVER=(uio|vfio)" "/sys/bus/pci/devices/$pci_blk/uevent"; then
 	$spdkcli_job "'/bdevs/virtioblk_disk create virtioblk_pci pci $pci_blk' 'virtioblk_pci' True"
 fi
 pci_scsi=$(lspci -nn -D | grep '1af4:1004' | head -1 | awk '{print $1;}')
-if [ -n "$pci_scsi" ]; then
+if [ -n "$pci_scsi" ] && grep -Eq "DRIVER=(uio|vfio)" "/sys/bus/pci/devices/$pci_scsi/uevent"; then
 	$spdkcli_job "'/bdevs/virtioscsi_disk create virtioscsi_pci pci $pci_scsi' 'virtioscsi_pci' True"
 fi
 $spdkcli_job "'/vhost/scsi create sample_scsi' 'sample_scsi' True
@@ -63,10 +63,10 @@ $spdkcli_job "'/vhost/block delete sample_block' 'sample_block'
 '/vhost/scsi/sample_scsi remove_target 0' 'Malloc0'
 '/vhost/scsi delete sample_scsi' 'sample_scsi'
 "
-if [ -n "$pci_blk" ]; then
+if [ -n "$pci_blk" ] && grep -Eq "DRIVER=(uio|vfio)" "/sys/bus/pci/devices/$pci_blk/uevent"; then
 	$spdkcli_job "'/bdevs/virtioblk_disk delete virtioblk_pci' 'virtioblk_pci'"
 fi
-if [ -n "$pci_scsi" ]; then
+if [ -n "$pci_scsi" ] && grep -Eq "DRIVER=(uio|vfio)" "/sys/bus/pci/devices/$pci_scsi/uevent"; then
 	$spdkcli_job "'/bdevs/virtioscsi_disk delete virtioscsi_pci' 'virtioscsi_pci'"
 fi
 $spdkcli_job "'/bdevs/malloc delete Malloc0' 'Malloc0'

test/vhost/initiator/blockdev.sh

@@ -58,6 +58,7 @@ function run_spdk_fio() {
 function create_bdev_config()
 {
 	local vbdevs
+	local g_opt
 
 	if [ -z "$($RPC_PY bdev_get_bdevs | jq '.[] | select(.name=="Nvme0n1")')" ]; then
 		error "Nvme0n1 bdev not found!"
@@ -82,7 +83,13 @@ function create_bdev_config()
 	$RPC_PY vhost_create_scsi_controller naa.Malloc1.0
 	$RPC_PY vhost_scsi_controller_add_target naa.Malloc1.0 0 Malloc1
 
-	vbdevs=$(discover_bdevs $rootdir $testdir/bdev.json "--json")
+	# Check default size of host hugepages. If it's 2MB then we have to use
+	# bdev_svc "-g" option for virtio devices.
+	if (( $(grep "Hugepagesize" /proc/meminfo | grep -Eo "[[:digit:]]+") == 2048 )); then
+		g_opt="-g"
+	fi
+
+	vbdevs=$(discover_bdevs $rootdir $testdir/bdev.json "--json" $g_opt)
 	virtio_bdevs=$(jq -r '[.[].name] | join(":")' <<< $vbdevs)
 	virtio_with_unmap=$(jq -r '[.[] | select(.supported_io_types.unmap==true).name]
 	 | join(":")' <<< $vbdevs)