Skip slicer processing for unmodified DEX files

Slicer's `CreateFullIr` and `CreateImage` operations are highly CPU-intensive and cause significant load latency for large DEX files. We hence introduce a fast-path optimization using `memmem` to perform a raw binary scan of the DEX buffer for target signatures prior to processing.

If no matching signatures are found, the expensive IR parsing and image rebuilding phases are bypassed entirely, and the original file descriptor is safely returned.

Moreover, we update the logging tag to `VectorDaemon`.

Author: JingMatrix

daemon/src/main/jni/logging.h

@@ -25,7 +25,7 @@
 #include <errno.h>
 
 #ifndef LOG_TAG
-#define LOG_TAG    "LSPosed"
+#define LOG_TAG    "VectorDaemon"
 #endif
 
 #ifdef LOG_DISABLED

daemon/src/main/jni/obfuscation.cpp

@@ -221,6 +221,26 @@ extern "C" JNIEXPORT jobject JNICALL Java_org_lsposed_lspd_service_ObfuscationMa
         return nullptr;
     }
 
+    bool needs_obfuscation = false;
+    for (const auto &sig : signatures) {
+        if (memmem(mem, size, sig.first.c_str(), sig.first.length()) != nullptr) {
+            needs_obfuscation = true;
+            break;
+        }
+    }
+
+    if (!needs_obfuscation) {
+        LOGD("No target signatures found in fd=%d, skipping slicer.", fd);
+        munmap(mem, size);
+
+        // Wrap the duplicated FD into Java objects and return instantly
+        auto java_fd =
+            lsplant::JNI_NewObject(env, class_file_descriptor, method_file_descriptor_ctor, fd);
+        auto java_sm =
+            lsplant::JNI_NewObject(env, class_shared_memory, method_shared_memory_ctor, java_fd);
+        return java_sm.release();
+    }
+
     // Process the DEX and obtain a new file descriptor for the output
     int new_fd = obfuscateDexBuffer(mem, size);