Perform ACM-triggered cloud connection migration without a disconnect

CHANGELOG.md

@@ -1,3 +1,27 @@
+## 6.3.0
+
+### FEATURES
+
+- New cloud event API that supports sending and receiving up to 16K of data in an event [#2840](https://github.com/particle-iot/device-os/pull/2840)
+- Electron 2 platform support [#2844](https://github.com/particle-iot/device-os/pull/2844)
+
+### ENHANCEMENTS
+- [wiring] Generate an error when using `EEPROM` with `String` objects as `String` contains dynamically allocated data on heap [#2855](https://github.com/particle-iot/device-os/pull/2855)
+- [network] Postpone `Ethernet` interface initialization to allow STARTUP() call to override pin configuration without a reset  [#2848](https://github.com/particle-iot/device-os/pull/2848)
+
+### BUGFIXES
+
+- Prevent low-level WiFI network buffer overflow [#2853](https://github.com/particle-iot/device-os/pull/2853)
+- [Gen 3] Increase main stack size from 2K to 4K [#2851](https://github.com/particle-iot/device-os/pull/2851)
+- Thread-safe `__cxa_guard_acquire`/`__cxa_guard_release` implementation [#2851](https://github.com/particle-iot/device-os/pull/2851)
+- [build] Fix section size calculations for paths containing section-like names [#2852](https://github.com/particle-iot/device-os/pull/2852)
+
+### INTERNAL
+- [test] wiring/no_fixture_long_running: update network tests to stress test large packets up to MTU size once max MTU is figured out [#2855](https://github.com/particle-iot/device-os/pull/2855)
+- [test] communication/long_publish: adds a long running publish test [#2854](https://github.com/particle-iot/device-os/pull/2854)
+- [build] Add asserts to linker files to make sure that stack sizes match the expected values [#2851](https://github.com/particle-iot/device-os/pull/2851)
+- Stub `atexit` to free up flash space used by global object destructors [#2851](https://github.com/particle-iot/device-os/pull/2851)
+
 ## 6.2.1
 
 ### FEATURES

build/release.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 set -o errexit -o pipefail -o noclobber -o nounset
 
-VERSION=${VERSION:="6.3.0-rc.1"}
+VERSION=${VERSION:="6.3.0"}
 
 function display_help ()
 {

build/version.mk

@@ -1,7 +1,7 @@
-VERSION_STRING = 6.3.0-rc.1
+VERSION_STRING = 6.3.0
 
 # PRODUCT_FIRMWARE_VERSION reported by default
 # FIXME: Unclear if this is used, PRODUCT_FIRMWARE_VERSION defaults to 65535 every release
-VERSION = 6300
+VERSION = 6302
 
 CFLAGS += -DSYSTEM_VERSION_STRING=$(VERSION_STRING)

communication/inc/dtls_session_persist.h

@@ -155,6 +155,9 @@ class __attribute__((packed)) SessionPersistOpaque : public SessionPersistData
 	void clear_use_count() { use_counter = 0; }
 	int use_count() { return use_counter; }
 	bool has_expired() { return use_counter >= MAXIMUM_SESSION_USES; }
+	void set_use_count_once() {
+		use_counter = MAXIMUM_SESSION_USES;
+	}
 
 	static const int MAXIMUM_SESSION_USES = 3;
 };

communication/inc/spark_protocol_functions.h

@@ -248,7 +248,8 @@ namespace ProtocolCommands {
     WAKE = 1, // Deprecated, use PING instead
     DISCONNECT = 2,
     TERMINATE = 3,
-    PING = 4
+    PING = 4,
+    MOVE_CONNECTION = 5
   };
 };
 

communication/src/dtls_message_channel.cpp

@@ -575,10 +575,17 @@ ProtocolError DTLSMessageChannel::command(Command command, void* arg)
 		reset_session();
 		return IO_ERROR_DISCARD_SESSION; //force re-establish
 
-	case MOVE_SESSION:
+	case MOVE_SESSION: {
+		bool once = false;
+		if (arg && *((bool*)arg)) {
+			once = true;
+		}
 		move_session = true;
+		if (once) {
+			sessionPersist.set_use_count_once();
+		}
 		break;
-
+	}
 	case LOAD_SESSION:
 		sessionPersist.restore(callbacks.restore);
 		break;

communication/src/dtls_protocol.h

@@ -115,6 +115,15 @@ class DTLSProtocol : public Protocol
 			}
 			return r;
 		}
+		case ProtocolCommands::MOVE_CONNECTION: {
+			LOG(INFO, "Moving connection (session move + ping)");
+			int r = ProtocolError::NO_ERROR;
+			bool once = true; // Use session only once, if migration fails it will be invalidated immediately
+			if (!(r = channel.command(Channel::MOVE_SESSION, &once))) {
+				return this->command(ProtocolCommands::PING, 0, nullptr);
+			}
+			return r;
+		}
 		default:
 			return ProtocolError::UNKNOWN;
 		}

modules/shared/system_module_version.mk

@@ -1,6 +1,6 @@
 # Skip to next 100 every v0.x.0 release (e.g. 108 for v0.6.2 to 200 for v0.7.0-rc.1)
 # Bump by 1 for every prerelease or release with the same v0.x.* base.
-COMMON_MODULE_VERSION ?= 6300
+COMMON_MODULE_VERSION ?= 6302
 SYSTEM_PART1_MODULE_VERSION ?= $(COMMON_MODULE_VERSION)
 
 RELEASE_080_MODULE_VERSION_BASE ?= 300

system/inc/system_version.h

@@ -225,7 +225,9 @@ extern "C" {
 #define SYSTEM_VERSION_v620         SYSTEM_VERSION_DEFAULT(6, 2, 0)
 #define SYSTEM_VERSION_v621         SYSTEM_VERSION_DEFAULT(6, 2, 1)
 #define SYSTEM_VERSION_v630RC1         SYSTEM_VERSION_RC(6, 3, 0, 1)
-#define SYSTEM_VERSION SYSTEM_VERSION_v630RC1
+#define SYSTEM_VERSION_v630RC2         SYSTEM_VERSION_RC(6, 3, 0, 2)
+#define SYSTEM_VERSION_v630         SYSTEM_VERSION_DEFAULT(6, 3, 0)
+#define SYSTEM_VERSION SYSTEM_VERSION_v630
 
 /**
  * Previously we would set the least significant byte to 0 for the final release, but to make
@@ -415,6 +417,8 @@ extern "C" {
 #define SYSTEM_VERSION_620
 #define SYSTEM_VERSION_621
 #define SYSTEM_VERSION_630RC1
+#define SYSTEM_VERSION_630RC2
+#define SYSTEM_VERSION_630
 
 typedef struct __attribute__((packed)) SystemVersionInfo
 {

system/src/system_cloud_connection.cpp

@@ -361,3 +361,22 @@ int system_cloud_get_inet_family_keepalive(int af, unsigned int* value) {
 
     return 0;
 }
+
+int system_cloud_move_connection(network_handle_t network) {
+    if (system_cloud_is_connected(nullptr)) { // XXX: this is not a bool!
+        return SYSTEM_ERROR_INVALID_STATE;
+    }
+
+    LOG(INFO, "Attempting to move cloud connection to interface %u", network);
+
+    int r = system_cloud_rebind(network);
+    if (r) {
+        LOG(WARN, "Cloud connection rebind to %u failed err=%d", network, r);
+        return r;
+    }
+    r = spark_protocol_command(spark_protocol_instance(), ProtocolCommands::MOVE_CONNECTION);
+    if (r) {
+        LOG(WARN, "Protocol connection move failed err=%d", r);
+    }
+    return r;
+}

system/src/system_cloud_connection.h

@@ -27,6 +27,7 @@
 #if HAL_PLATFORM_IFAPI
 #include "netdb_hal.h"
 #endif // HAL_PLATFORM_IFAPI
+#include "system_network.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -53,6 +54,8 @@ int system_multicast_announce_presence(void* reserved);
 int system_cloud_set_inet_family_keepalive(int af, unsigned int value, int flags);
 int system_cloud_get_inet_family_keepalive(int af, unsigned int* value);
 sock_handle_t system_cloud_get_socket_handle();
+int system_cloud_move_connection(network_handle_t network);
+int system_cloud_rebind(network_handle_t network);
 
 #if HAL_PLATFORM_IFAPI
 int system_cloud_resolv_address(int protocol, const ServerAddress* address, sockaddr* saddrCache, addrinfo** info, CloudServerAddressType* type, bool useCachedAddrInfo);

system/src/system_cloud_connection_compat.cpp

@@ -328,5 +328,8 @@ sock_handle_t system_cloud_get_socket_handle()
     return s_state.socket;
 }
 
+int system_cloud_rebind(network_handle_t network) {
+    return SYSTEM_ERROR_NOT_SUPPORTED;
+}
 
 #endif /* !HAL_USE_SOCKET_HAL_POSIX && HAL_USE_SOCKET_HAL_COMPAT */

system/src/system_cloud_connection_posix.cpp

@@ -382,4 +382,27 @@ sock_handle_t system_cloud_get_socket_handle()
     return s_state.socket;
 }
 
+int system_cloud_rebind(network_handle_t network) {
+    if (s_state.socket < 0) {
+        return SYSTEM_ERROR_INVALID_STATE;
+    }
+
+    if (network == NETWORK_INTERFACE_ALL) {
+        return SYSTEM_ERROR_INVALID_ARGUMENT;
+    }
+
+    // Bind to specific netif
+    struct ifreq ifr = {};
+    if_index_to_name(network, ifr.ifr_name);
+
+    auto sockOptRet = sock_setsockopt(s_state.socket, SOL_SOCKET, SO_BINDTODEVICE, &ifr, sizeof(ifr));
+    if (sockOptRet) {
+        LOG(WARN, "Failed to bind cloud socket to interface %u error: %d", network, sockOptRet);
+    } else {
+        LOG(INFO, "Re-bound cloud socket to iface %u (\"%s\")", network, ifr.ifr_name);
+    }
+
+    return !sockOptRet ? SYSTEM_ERROR_NONE : SYSTEM_ERROR_NETWORK;
+}
+
 #endif /* HAL_USE_SOCKET_HAL_POSIX */

system/src/system_connection_manager.cpp

@@ -319,6 +319,15 @@ int ConnectionManager::checkCloudConnectionNetwork() {
     }
     // If best candidate doesn't match current network interface - reconnect
     LOG(TRACE, "Best network interface for cloud connection changed (to %s) - move the cloud session", netifToName(best));
+    if (best != NETWORK_INTERFACE_ALL) {
+        int r = system_cloud_move_connection(best);
+        if (!r) {
+            return 0;
+        } else {
+            LOG(WARN, "Failed to move connection err=%d, fallback to reconnect", r);
+        }
+    }
+    // Fallback to disconnect/connect
     auto options = CloudDisconnectOptions().reconnect(true);
     auto systemOptions = options.toSystemOptions();
     spark_cloud_disconnect(&systemOptions, nullptr);

system/system-versions.md

@@ -200,6 +200,8 @@
 | 3100 | 6200 | 6.2.0         |      Argon, Boron, B SoM, B5 SoM, Tracker, Tracker M, E Som X, M SoM, P2 |
 | 3101 | 6201 | 6.2.1         |      Argon, Boron, B SoM, B5 SoM, Tracker, Tracker M, E Som X, M SoM, P2 |
 | 3101 | 6300 | 6.3.0-rc.1    |      Argon, Boron, B SoM, B5 SoM, Tracker, Tracker M, E Som X, M SoM, P2 |
+| 3101 | 6301 | 6.3.0-rc.2    |      Argon, Boron, B SoM, B5 SoM, Tracker, Tracker M, E Som X, M SoM, P2 |
+| 3101 | 6302 | 6.3.0         |      Argon, Boron, B SoM, B5 SoM, Tracker, Tracker M, E Som X, M SoM, P2 |
 
 [1] For 0.8.0-rc.1, The v101 bootloader was also released in the Github releases as v200. Thus the next released bootloader in the 0.8.x line should be v201. As of 4/5/2018: 22 device had v200 bootloaders.
 

user/tests/integration/slo/connect_time/connect_time.cpp

@@ -368,14 +368,16 @@ DEFINE_WARM_BOOT_TEST(08)
 
 test(publish_and_validate_stats) {
 #if Wiring_Cellular
-    Cellular.on();
-    assertTrue(waitFor(network().isOn, TEST_MAX_TIMEOUT));
-    // Exclude 2G and non-production devices from the SLO validation
-    CellularDevice devInfo = {};
-    devInfo.size = sizeof(devInfo);
-    assertEqual(cellular_device_info(&devInfo, nullptr), 0);
-    if (devInfo.dev == DEV_SARA_G350 || devInfo.dev == DEV_QUECTEL_EG91_NA) {
-        stats.excludeFromSloValidation = true;
+    if (network() == Cellular) {
+        Cellular.on();
+        assertTrue(waitFor(network().isOn, TEST_MAX_TIMEOUT));
+        // Exclude 2G and non-production devices from the SLO validation
+        CellularDevice devInfo = {};
+        devInfo.size = sizeof(devInfo);
+        assertEqual(cellular_device_info(&devInfo, nullptr), 0);
+        if (devInfo.dev == DEV_SARA_G350 || devInfo.dev == DEV_QUECTEL_EG91_NA) {
+            stats.excludeFromSloValidation = true;
+        }
     }
 #endif // Wiring_Cellular
     const size_t n = serializeStatsAsJson(nullptr /* buf */, 0 /* size */);