Increase default SMB connections to 8 and simplify setup action (#16)

- Change default SPICEIO_SMB_CONNECTIONS from 4 to 8 for better
  concurrent request throughput.
- Remove already-running instance detection from the setup action;
  always start a fresh instance (port auto-increment handles conflicts).
- Expose `pid` output from the setup action so callers can stop spiceio
  at the end of their workflow with `if: always()`.

Author: lukekim

.github/actions/setup/action.yml

@@ -35,6 +35,9 @@ outputs:
   endpoint:
     description: The S3-compatible endpoint URL
     value: ${{ steps.start.outputs.endpoint }}
+  pid:
+    description: PID of the spiceio process (use to stop it at the end of your workflow)
+    value: ${{ steps.start.outputs.pid }}
 
 runs:
   using: composite
@@ -80,31 +83,6 @@ runs:
       run: |
         set -euo pipefail
 
-        # Check if spiceio is already listening on the requested address
-        WANT_VERSION=$(spiceio --version | awk '{print $2}')
-        RUNNING_HEADER=$(curl -s -I "http://${SPICEIO_BIND}/" 2>/dev/null | grep -i '^server:.*spiceio' | tr -d '\r' || true)
-        if [[ -n "$RUNNING_HEADER" ]]; then
-          # Extract version from "Server: spiceio/X.Y.Z"; bare "Server: spiceio"
-          # (no slash) means a pre-versioned build — treat as outdated.
-          RUNNING_VERSION="${RUNNING_HEADER##*/}"
-          RUNNING_VERSION="${RUNNING_VERSION// /}"
-          if [[ "$RUNNING_HEADER" == */* && "$RUNNING_VERSION" == "$WANT_VERSION" ]]; then
-            echo "spiceio $WANT_VERSION already running at http://${SPICEIO_BIND} — skipping start"
-            echo "endpoint=http://${SPICEIO_BIND}" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-          # Old or mismatched version — kill it so we can replace it
-          mapfile -t OLD_PIDS < <(lsof -ti "tcp:${SPICEIO_BIND##*:}" -sTCP:LISTEN 2>/dev/null || true)
-          if (( ${#OLD_PIDS[@]} > 0 )); then
-            echo "replacing spiceio ${RUNNING_VERSION:-unknown} (PID(s) ${OLD_PIDS[*]}) with $WANT_VERSION"
-            kill "${OLD_PIDS[@]}" 2>/dev/null || true
-            for i in $(seq 1 10); do
-              lsof -ti "tcp:${SPICEIO_BIND##*:}" -sTCP:LISTEN >/dev/null 2>&1 || break
-              sleep 1
-            done
-          fi
-        fi
-
         # Parse smb://user:pass@server:port/share
         # Strips the smb:// prefix, then splits on :, @, /
         URL="${SMB_URL#smb://}"
@@ -163,6 +141,7 @@ runs:
           if [[ -n "$ENDPOINT" ]] && curl -fsS -o /dev/null "$ENDPOINT/" 2>/dev/null; then
             echo "spiceio ready at $ENDPOINT (PID $PID)"
             echo "endpoint=$ENDPOINT" >> "$GITHUB_OUTPUT"
+            echo "pid=$PID" >> "$GITHUB_OUTPUT"
             exit 0
           fi
           sleep 1

CLAUDE.md

@@ -34,7 +34,7 @@ The binary requires these environment variables:
 - `SPICEIO_SMB_DOMAIN` — SMB domain (default empty)
 - `SPICEIO_BUCKET` — virtual S3 bucket name (defaults to `SPICEIO_SMB_SHARE`)
 - `SPICEIO_REGION` — AWS region to advertise (default `us-east-1`)
-- `SPICEIO_SMB_CONNECTIONS` — number of SMB TCP connections in the pool (default `4`)
+- `SPICEIO_SMB_CONNECTIONS` — number of SMB TCP connections in the pool (default `8`)
 - `SPICEIO_SMB_MAX_IO` — max standalone read/write I/O size in bytes (default `65536`; raise for servers that handle larger I/O)
 - `SPICEIO_LOG_FILE` — append logs to this file in addition to stderr (optional; non-blocking, never stalls the proxy)
 
@@ -54,7 +54,7 @@ The codebase has three modules:
 
 - Zero external crypto dependencies — all crypto goes through `crypto::ffi` to CommonCrypto.
 - No `async-trait` — the SMB client uses `tokio::sync::Mutex` around the TCP stream with manual `async` methods.
-- Connection pool — N TCP connections (default 4) to the same SMB server, round-robin dispatched. Concurrent S3 requests fan out across connections instead of serializing on a single mutex. File handles are pinned to the connection that opened them.
+- Connection pool — N TCP connections (default 8) to the same SMB server, round-robin dispatched. Concurrent S3 requests fan out across connections instead of serializing on a single mutex. File handles are pinned to the connection that opened them.
 - Pipelined reads — streaming GetObject sends batches of 8 read requests before collecting responses, hiding per-request round-trip latency.
 - Configurable I/O cap — standalone read/write ops default to 64 KB (safe for commodity NAS); raisable via `SPICEIO_SMB_MAX_IO` for compliant servers. Compound operations always cap at 64 KB.
 - GetObject streams SMB read chunks directly to the HTTP response via `SpiceioBody::channel` — no full-file buffering.

src/main.rs

@@ -43,7 +43,7 @@ struct Config {
     bucket_name: String,
     /// AWS region to advertise
     region: String,
-    /// Number of SMB TCP connections in the pool (default 4)
+    /// Number of SMB TCP connections in the pool (default 8)
     smb_connections: usize,
     /// Max I/O size for standalone read/write operations (default 1MB)
     smb_max_io: u32,
@@ -72,7 +72,7 @@ impl Config {
             smb_connections: env::var("SPICEIO_SMB_CONNECTIONS")
                 .ok()
                 .and_then(|s| s.parse().ok())
-                .unwrap_or(4),
+                .unwrap_or(8),
             smb_max_io: env::var("SPICEIO_SMB_MAX_IO")
                 .ok()
                 .and_then(|s| s.parse().ok())