Improve the performance of detecting linux-musl platform by parsing ELF (#315)

Co-authored-by: Natalie Weizenbaum <[email protected]>

Author: ntkme

lib/src/compiler-path.ts

@@ -4,20 +4,29 @@
 
 import * as fs from 'fs';
 import * as p from 'path';
+import {getElfInterpreter} from './elf';
 import {isErrnoException} from './utils';
 
 /**
- * Detect if the current running node binary is linked with musl libc by
- * checking if the binary contains a string like "/.../ld-musl-$ARCH.so"
+ * Detect if the given binary is linked with musl libc by checking if
+ * the interpreter basename starts with "ld-musl-"
  */
-const isLinuxMusl = function () {
-  return fs.readFileSync(process.execPath).includes('/ld-musl-');
-};
+function isLinuxMusl(path: string): boolean {
+  try {
+    const interpreter = getElfInterpreter(path);
+    return p.basename(interpreter).startsWith('ld-musl-');
+  } catch (error) {
+    console.warn(
+      `Warning: Failed to detect linux-musl, fallback to linux-gnu: ${error.message}`
+    );
+    return false;
+  }
+}
 
 /** The full command for the embedded compiler executable. */
 export const compilerCommand = (() => {
   const platform =
-    process.platform === 'linux' && isLinuxMusl()
+    process.platform === 'linux' && isLinuxMusl(process.execPath)
       ? 'linux-musl'
       : (process.platform as string);
 

lib/src/elf.ts

@@ -0,0 +1,102 @@
+// Copyright 2024 Google LLC. Use of this source code is governed by an
+// MIT-style license that can be found in the LICENSE file or at
+// https://opensource.org/licenses/MIT.
+
+import * as fs from 'fs';
+
+/** Read a chunk of data from a file descriptor into a new Buffer. */
+function readFileDescriptor(
+  fd: number,
+  position: number,
+  length: number
+): Buffer {
+  const buffer = Buffer.alloc(length);
+  let offset = 0;
+  while (offset < length) {
+    const bytesRead = fs.readSync(fd, buffer, {
+      offset: offset,
+      position: position + offset,
+    });
+    if (bytesRead === 0) {
+      throw new Error(`failed to read fd ${fd}`);
+    }
+
+    offset += bytesRead;
+  }
+  return buffer;
+}
+
+/** Parse an ELF file and return its interpreter. */
+export function getElfInterpreter(path: string): string {
+  const fd = fs.openSync(path, 'r');
+  try {
+    const elfIdentification = new DataView(
+      readFileDescriptor(fd, 0, 64).buffer
+    );
+
+    if (
+      elfIdentification.getUint8(0) !== 0x7f ||
+      elfIdentification.getUint8(1) !== 0x45 ||
+      elfIdentification.getUint8(2) !== 0x4c ||
+      elfIdentification.getUint8(3) !== 0x46
+    ) {
+      throw new Error(`${path} is not an ELF file.`);
+    }
+
+    const elfIdentificationClass = elfIdentification.getUint8(4);
+    if (elfIdentificationClass !== 1 && elfIdentificationClass !== 2) {
+      throw new Error(`${path} has an invalid ELF class.`);
+    }
+    const elfClass32 = elfIdentificationClass === 1;
+
+    const elfIdentificationData = elfIdentification.getUint8(5);
+    if (elfIdentificationData !== 1 && elfIdentificationData !== 2) {
+      throw new Error(`${path} has an invalid endianness.`);
+    }
+    const littleEndian = elfIdentificationData === 1;
+
+    // Converting BigUint64 to Number because node Buffer length has to be
+    // number type, and we don't expect any elf we check with this method to
+    // be larger than 9007199254740991 bytes.
+    const programHeadersOffset = elfClass32
+      ? elfIdentification.getUint32(28, littleEndian)
+      : Number(elfIdentification.getBigUint64(32, littleEndian));
+    const programHeadersEntrySize = elfClass32
+      ? elfIdentification.getUint16(42, littleEndian)
+      : elfIdentification.getUint16(54, littleEndian);
+    const programHeadersEntryCount = elfClass32
+      ? elfIdentification.getUint16(44, littleEndian)
+      : elfIdentification.getUint16(56, littleEndian);
+
+    const programHeaders = new DataView(
+      readFileDescriptor(
+        fd,
+        programHeadersOffset,
+        programHeadersEntrySize * programHeadersEntryCount
+      ).buffer
+    );
+    for (let i = 0; i < programHeadersEntryCount; i++) {
+      const byteOffset = i * programHeadersEntrySize;
+      const segmentType = programHeaders.getUint32(byteOffset, littleEndian);
+      if (segmentType !== 3) continue; // 3 is PT_INTERP, the interpreter
+
+      const segmentOffset = elfClass32
+        ? programHeaders.getUint32(byteOffset + 4, littleEndian)
+        : Number(programHeaders.getBigUint64(byteOffset + 8, littleEndian));
+      const segmentFileSize = elfClass32
+        ? programHeaders.getUint32(byteOffset + 16, littleEndian)
+        : Number(programHeaders.getBigUint64(byteOffset + 32, littleEndian));
+
+      const buffer = readFileDescriptor(fd, segmentOffset, segmentFileSize);
+      if (buffer[segmentFileSize - 1] !== 0) {
+        throw new Error(`${path} is corrupted.`);
+      }
+
+      return buffer.toString('utf8', 0, segmentFileSize - 1);
+    }
+
+    throw new Error(`${path} does not contain an interpreter entry.`);
+  } finally {
+    fs.closeSync(fd);
+  }
+}