fix(cloud-hypervisor/vsock): apply a workaround for notify sockets

Since systemd/systemd@13b67b6
systemd shutdown the write end of one end of the VSOCK.

cloud-hypervisor virtio-sock code does not handle well partial shutdown of the
stream.

socat `-T` flag can serve as a workaround to mask this bug.

> Total inactivity timeout: when socat is already in the transfer loop and
> nothing has happened for <timeout> [timeval] seconds (no data arrived, no
> interrupt occurred...) then it terminates. Useful with protocols like UDP that
> cannot transfer EOF.

Co-authored-by: Puck Meerburg <[email protected]>
Signed-off-by: Raito Bezarius <[email protected]>

Author: 2 people

lib/runners/cloud-hypervisor.nix

@@ -89,9 +89,7 @@ let
     vulkan = true;
   };
 
-  # systemd>=256 hangs at stage-2 on notifying X_SYSTEMD_HOSTNAME
-  supportsNotifySocket =
-    builtins.compareVersions pkgs.systemd.version "256" < 0;
+  supportsNotifySocket = true;
 
 in {
   inherit tapMultiQueue;
@@ -110,7 +108,9 @@ in {
 
     # Start socat to forward systemd notify socket over vsock
     if [ -n "$NOTIFY_SOCKET" ]; then
-      ${pkgs.socat}/bin/socat UNIX-LISTEN:notify.vsock_8888,fork UNIX-SENDTO:$NOTIFY_SOCKET &
+      # -T2 is required because cloud-hypervisor does not handle partial
+      # shutdown of the stream, like systemd v256+ does.
+      ${pkgs.socat}/bin/socat -T2 UNIX-LISTEN:notify.vsock_8888,fork UNIX-SENDTO:$NOTIFY_SOCKET &
     fi
   '' + lib.optionalString graphics.enable ''
     rm -f ${graphics.socket}