Fix notification-delivery crash under load (thread-unsafe gethostbyname)

.github/workflows/functional.yml

@@ -80,11 +80,6 @@ jobs:
     needs: [check-changes, build-test-image]
     if: needs.check-changes.outputs.should_run == 'true'
     services:
-      mosquitto:
-        image: eclipse-mosquitto:1.6
-        ports:
-          - 1883:1883
-          - 9001:9001
       postgres:
         image: postgis/postgis:17-3.5-alpine
         env:
@@ -126,11 +121,23 @@ jobs:
         run: |
           docker load --input /tmp/test-image.tar
 
-      - name: Start MQTTS broker
+      - name: Start MQTT + MQTTS brokers
         run: |
           sudo apt-get install -y mosquitto openssl
+          # Stop the systemd-launched mosquitto so we can run with our own
+          # config (mosquitto 2.0 defaults to allow_anonymous false, which
+          # breaks the test suite's assumption of an open broker).
+          sudo systemctl stop mosquitto || true
+
+          # Plain MQTT on 1883 — anonymous enabled
+          cat > /tmp/mosquitto-plain.conf << EOF
+          listener 1883
+          allow_anonymous true
+          EOF
+          mosquitto -c /tmp/mosquitto-plain.conf -d
+
+          # MQTTS on 8883 with self-signed certs
           mkdir -p /tmp/mosquitto-mqtts
-          # Generate self-signed certs for MQTTS testing
           openssl req -new -x509 -days 30 -nodes \
             -keyout /tmp/mosquitto-mqtts/ca.key -out /tmp/mosquitto-mqtts/ca.crt \
             -subj "/CN=MQTT Test CA" 2>/dev/null

CHANGES_NEXT_RELEASE

@@ -1,5 +1,6 @@
 ## Fixed Issues
   #1905 - TRoE: segfault (exit 139) connecting to TimescaleDB Cloud - added SSL support and fixed NULL pointer crash
+  #XXXX - Fixed crash under concurrent notification delivery by replacing thread-unsafe gethostbyname with getaddrinfo in orionldServerConnect
   #XXXX - Fixed query of many large entities (cfreyfh - Carsten Freyh at Fraunhofer IKTS)
   #XXXX - Wait for postgres instead of exiting after the first try - just like for mongo (1 try per second during 30 secs)
   #XXXX - Fixed crash in relationship endpoint when no relations found (cfreyfh)

doc/manuals-ld/installation-guide-centos7.md

@@ -253,6 +253,8 @@ sudo systemctl start mosquitto
 sudo systemctl enable mosquitto
 ```
 
+Note: Orion-LD's MQTT subscription support includes MQTT 5.0, which requires Mosquitto 2.0 or later. The EPEL package for CentOS 7 ships 1.4.x, so if you need MQTT 5.0 use the upstream repository from https://mosquitto.org.
+
 ### Fast-DDS (optional, for DDS support)
 
 Fast-DDS is eProsima's implementation of the DDS (Data Distribution Service) standard. It enables Orion-LD to communicate with DDS-based systems.

doc/manuals-ld/installation-guide-debian10.md

@@ -282,6 +282,8 @@ sudo systemctl start mosquitto
 sudo systemctl enable mosquitto
 ```
 
+Note: Orion-LD's MQTT subscription support includes MQTT 5.0, which requires Mosquitto 2.0 or later. The Debian 10 distro package ships 1.5.x, so if you need MQTT 5.0 use the upstream repository from https://mosquitto.org.
+
 ### Fast-DDS (optional, for DDS support)
 
 Fast-DDS is eProsima's implementation of the DDS (Data Distribution Service) standard. It enables Orion-LD to communicate with DDS-based systems.

doc/manuals-ld/installation-guide-debian9.md

@@ -257,6 +257,8 @@ sudo systemctl start mosquitto
 sudo systemctl enable mosquitto
 ```
 
+Note: Orion-LD's MQTT subscription support includes MQTT 5.0, which requires Mosquitto 2.0 or later. The Debian 9 distro package ships 1.4.x, so if you need MQTT 5.0 use the upstream repository from https://mosquitto.org.
+
 ### Fast-DDS (optional, for DDS support)
 
 Fast-DDS is eProsima's implementation of the DDS (Data Distribution Service) standard. It enables Orion-LD to communicate with DDS-based systems.

doc/manuals-ld/installation-guide-ubuntu-18.04.3.md

@@ -463,10 +463,12 @@ If you wish to enable `mosquitto` to have it start automatically on system reboo
 sudo systemctl enable mosquitto
 ```
 
-### Postgres 12
+Note: Orion-LD's MQTT subscription support includes MQTT 5.0, which requires Mosquitto 2.0 or later. The Ubuntu 18.04 distro package ships 1.4.x, so if you need MQTT 5.0 use the upstream PPA: `sudo add-apt-repository ppa:mosquitto-dev/mosquitto-ppa`.
+
+### Postgres 17
 Postgres is used as database for the Temporal Evolution of entities
 
-#### Install Postgres 12
+#### Install Postgres 17
 ```bash
 wget --quiet -O - https://www.postgresql.org/media/keys/ACCC4CF8.asc | sudo apt-key add -
 echo "deb http://apt.postgresql.org/pub/repos/apt/ `lsb_release -cs`-pgdg main" | sudo tee  /etc/apt/sources.list.d/pgdg.list

doc/manuals-ld/installation-guide-ubuntu-20.04.1.md

@@ -459,10 +459,12 @@ If you wish to enable `mosquitto` to have it start automatically on system reboo
 sudo systemctl enable mosquitto
 ```
 
-### Postgres 12
+Note: Orion-LD's MQTT subscription support includes MQTT 5.0, which requires Mosquitto 2.0 or later. The Ubuntu 20.04 distro package ships 1.6.x, so if you need MQTT 5.0 use the upstream PPA: `sudo add-apt-repository ppa:mosquitto-dev/mosquitto-ppa`.
+
+### Postgres 17
 Postgres is used as database for the Temporal Evolution of entities
 
-#### Install Postgres 12
+#### Install Postgres 17
 ```bash
 wget --quiet -O - https://www.postgresql.org/media/keys/ACCC4CF8.asc | sudo apt-key add -
 echo "deb http://apt.postgresql.org/pub/repos/apt/ `lsb_release -cs`-pgdg main" | sudo tee  /etc/apt/sources.list.d/pgdg.list

doc/manuals-ld/installation-guide-ubuntu-22.04.md

@@ -404,7 +404,9 @@ If you wish to enable `mosquitto` to have it start automatically on system reboo
 sudo systemctl enable mosquitto
 ```
 
-### Postgres 12
+Note: Orion-LD's MQTT subscription support includes MQTT 5.0, which requires Mosquitto 2.0 or later. On Ubuntu 22.04 the default `mosquitto` package is already 2.x, so no extra step is needed.
+
+### Postgres 17
 Postgres is used as database for the Temporal Evolution of entities
 
 #### Install Postgres 17

src/lib/orionld/common/orionldServerConnect.cpp

@@ -22,9 +22,10 @@
 *
 * Author: Ken Zangelin
 */
-#include <string.h>                                              // strerror
+#include <stdio.h>                                               // snprintf
+#include <string.h>                                              // strerror, memset
 #include <unistd.h>                                              // close
-#include <netdb.h>                                               // struct hostent
+#include <netdb.h>                                               // getaddrinfo, freeaddrinfo, gai_strerror
 #include <errno.h>                                               // errno
 #include <sys/types.h>                                           // types
 #include <sys/socket.h>                                          // socket
@@ -46,36 +47,47 @@
 //
 int orionldServerConnect(const char* ip, uint16_t portNo)
 {
-  int                 fd;
-  struct hostent*     heP;
-  struct sockaddr_in  server;
+  // getaddrinfo is POSIX-mandated thread-safe; the older gethostbyname
+  // writes into a single process-wide static hostent and crashed under
+  // concurrent notification delivery when two MHD threads raced on that
+  // shared buffer.
+  struct addrinfo   hints;
+  struct addrinfo*  res = NULL;
+  char              portStr[16];
+  int               fd;
 
   KT_T(KtNotificationMsg, "Connecting to IP: '%s'", ip);
-  heP = gethostbyname(ip);
-  if (heP == NULL)
+
+  memset(&hints, 0, sizeof(hints));
+  hints.ai_family   = AF_INET;
+  hints.ai_socktype = SOCK_STREAM;
+  hints.ai_flags    = AI_NUMERICSERV;  // port is a numeric string — skip servent lookup
+
+  snprintf(portStr, sizeof(portStr), "%u", portNo);
+
+  int rc = getaddrinfo(ip, portStr, &hints, &res);
+  if (rc != 0 || res == NULL)
   {
-    KT_E("unable to find host '%s'", ip);
+    KT_E("unable to resolve host '%s': %s", ip, gai_strerror(rc));
     return -1;
   }
 
-  fd = socket(AF_INET, SOCK_STREAM, 0);
+  fd = socket(res->ai_family, res->ai_socktype, res->ai_protocol);
   if (fd == -1)
   {
     KT_E("Can't even create a socket: %s", strerror(errno));
+    freeaddrinfo(res);
     return -1;
   }
 
-  server.sin_family = AF_INET;
-  server.sin_port   = htons(portNo);
-  server.sin_addr   = *((struct in_addr*) heP->h_addr);
-  bzero(&server.sin_zero, 8);
-
-  if (connect(fd, (struct sockaddr*) &server, sizeof(struct sockaddr)) == -1)
+  if (connect(fd, res->ai_addr, res->ai_addrlen) == -1)
   {
-    close(fd);
     KT_E("Unable to connect to host/port: %s:%d", ip, portNo);
+    close(fd);
+    freeaddrinfo(res);
     return -1;
   }
 
+  freeaddrinfo(res);
   return fd;
 }