diff --git a/hosts/skydick/datapool.nix b/hosts/skydick/datapool.nix
index d6e85ec..e7a5aa6 100644
--- a/hosts/skydick/datapool.nix
+++ b/hosts/skydick/datapool.nix
@@ -366,6 +366,36 @@
   # NFS — primary protocol for all datasets
   services.rpcbind.enable = true;
 
+  # NFS-over-RDMA listener (additive to TCP — clients pick one transport
+  # per mount). skydick's ConnectX-5 / mlx5_bond_0 RoCEv2 fabric is the
+  # transport; nfsd binds to UDP/TCP 2049 by default plus RDMA 20049
+  # when this oneshot writes to /proc/fs/nfsd/portlist. Clients use
+  # the RDMA path with `proto=rdma,port=20049` mount options. Saves a
+  # TCP/IP stack traversal + uses native NIC parallelism (no nconnect
+  # equivalent needed — RDMA QPs are intrinsically parallel).
+  boot.kernelModules = [ "rpcrdma" ];
+
+  systemd.services.nfsd-rdma-listener = {
+    description = "Enable NFS-over-RDMA listener on port 20049";
+    after = [ "nfs-server.service" ];
+    wants = [ "nfs-server.service" ];
+    wantedBy = [ "multi-user.target" ];
+    serviceConfig = {
+      Type = "oneshot";
+      RemainAfterExit = true;
+    };
+    # If the listener is already added (re-run after rebuild), nfsd
+    # returns EINVAL — that's fine, the listener stays up.
+    script = ''
+      ${pkgs.gnugrep}/bin/grep -q "^rdma 20049$" /proc/fs/nfsd/portlist && {
+        echo "rdma 20049 already listening"
+        exit 0
+      }
+      echo "rdma 20049" > /proc/fs/nfsd/portlist
+      echo "added RDMA listener on 20049"
+    '';
+  };
+
   services.nfs.server = {
     enable = true;
     nproc = 64;