From c136b3146257d0f12d1347a9f1a4784372e19a56 Mon Sep 17 00:00:00 2001
From: Ryan Harrison <rharrison@chromium.org>
Date: Thu, 19 Oct 2017 14:37:06 -0400
Subject: Implement fuzzer for LZW decompressor

This adds a fuzzer that tests the LZW decompression code used by our
GIF decoder.

BUG=pdfium:908

Change-Id: I1381f3ebb2eddf8d2c6f0394b1bb00f67d64a600
Reviewed-on: https://pdfium-review.googlesource.com/16310
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Tom Sepez <tsepez@chromium.org>
---
 testing/libfuzzer/BUILD.gn          |  7 +++++
 testing/libfuzzer/pdf_lzw_fuzzer.cc | 58 +++++++++++++++++++++++++++++++++++++
 2 files changed, 65 insertions(+)
 create mode 100644 testing/libfuzzer/pdf_lzw_fuzzer.cc

diff --git a/testing/libfuzzer/BUILD.gn b/testing/libfuzzer/BUILD.gn
index 2952444421..d23fb88376 100644
--- a/testing/libfuzzer/BUILD.gn
+++ b/testing/libfuzzer/BUILD.gn
@@ -46,6 +46,7 @@ group("libfuzzer") {
       ":pdf_codec_tiff_fuzzer",
       ":pdf_css_fuzzer",
       ":pdf_fm2js_fuzzer",
+      ":pdf_lzw_fuzzer",
       ":pdf_xml_fuzzer",
     ]
   }
@@ -129,6 +130,12 @@ if (pdf_enable_xfa) {
     ]
   }
 
+  pdfium_fuzzer("pdf_lzw_fuzzer") {
+    sources = [
+      "pdf_lzw_fuzzer.cc",
+    ]
+  }
+
   pdfium_fuzzer("pdf_xml_fuzzer") {
     sources = [
       "pdf_xml_fuzzer.cc",
diff --git a/testing/libfuzzer/pdf_lzw_fuzzer.cc b/testing/libfuzzer/pdf_lzw_fuzzer.cc
new file mode 100644
index 0000000000..71c258969d
--- /dev/null
+++ b/testing/libfuzzer/pdf_lzw_fuzzer.cc
@@ -0,0 +1,58 @@
+// Copyright 2017 The PDFium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include <vector>
+
+#include "core/fxcodec/gif/cfx_lzwdecompressor.h"
+#include "third_party/base/numerics/safe_conversions.h"
+
+// Between 2x and 5x is a standard range for LZW according to a quick
+// search of papers. Running up to 10x to catch any niche cases.
+constexpr uint32_t kMinCompressionRatio = 2;
+constexpr uint32_t kMaxCompressionRatio = 10;
+
+void LZWFuzz(const uint8_t* src_buf,
+             size_t src_size,
+             uint8_t color_exp,
+             uint8_t code_exp) {
+  std::unique_ptr<CFX_LZWDecompressor> decompressor =
+      CFX_LZWDecompressor::Create(color_exp, code_exp);
+  if (!decompressor)
+    return;
+
+  for (uint32_t compressions_ratio = kMinCompressionRatio;
+       compressions_ratio <= kMaxCompressionRatio; compressions_ratio++) {
+    std::vector<uint8_t> des_buf(compressions_ratio * src_size);
+    // This cast should be safe since the caller is checking for overflow on
+    // the initial data.
+    uint32_t des_size = static_cast<uint32_t>(des_buf.size());
+    if (CFX_GifDecodeStatus::InsufficientDestSize !=
+        decompressor->Decode(const_cast<uint8_t*>(src_buf), src_size,
+                             des_buf.data(), &des_size))
+      return;
+  }
+}
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t* data, size_t size) {
+  // Need at least 3 bytes to do anything.
+  if (size < 3)
+    return 0;
+
+  // Normally the GIF would provide the code and color sizes, instead, going
+  // to assume they are the first two bytes of data provided.
+  uint8_t color_exp = data[0];
+  uint8_t code_exp = data[1];
+  const uint8_t* lzw_data = data + 2;
+  uint32_t lzw_data_size = static_cast<uint32_t>(size - 2);
+  // Check that there isn't going to be an overflow in the destination buffer
+  // size.
+  if (lzw_data_size >
+      std::numeric_limits<uint32_t>::max() / kMaxCompressionRatio) {
+    return 0;
+  }
+
+  LZWFuzz(lzw_data, lzw_data_size, color_exp, code_exp);
+
+  return 0;
+}
-- 
cgit v1.2.3