build: import manually minified Chrome URL lib. (#3)

This is a manually minified variant of
https://chromium.googlesource.com/chromium/src.git/+archive/74.0.3729.15/url.tar.gz,
providing just the parts needed for url::CanonicalizePath(). This is intended
to support a security release fix for CVE-2019-9901. Long term we need this to
be moved to absl or QUICHE for upgrades and long-term support.

Some specific transforms of interest:

* url_parse.h is minified to just Component and flattened back into the URL directory. It does not contain any non-Chromium authored code any longer and so does not have a separate LICENSE.
* envoy_shim.h adapts various macros to the Envoy context.
8 Anything not reachable from url::CanonicalizePath() has been dropped.
* Header include paths have changed as needed.
* BUILD was manually written.
* Various clang-tidy and format fixes.

Risk level: Low
Testing: Validated with WiP PR for CVE-2019-9901.

Signed-off-by: Harvey Tuch <htuch@google.com>

ci/run_clang_tidy.sh

@@ -21,15 +21,25 @@ function exclude_testdata() {
   grep -v tools/testdata/check_format/
 }
 
+# Do not run clang-tidy against Chromium URL import, this needs to largely
+# reflect the upstream structure.
+function exclude_chromium_url() {
+  grep -v source/common/chromium_url/
+}
+
+function filter_excludes() {
+  exclude_testdata | exclude_chromium_url
+}
+
 if [[ "${RUN_FULL_CLANG_TIDY}" == 1 ]]; then
   echo "Running full clang-tidy..."
   run-clang-tidy-7
 elif [[ -z "${CIRCLE_PR_NUMBER}" && "$CIRCLE_BRANCH" == "master" ]]; then
   echo "On master branch, running clang-tidy-diff against previous commit..."
-  git diff HEAD^ | exclude_testdata | clang-tidy-diff-7.py -p 1
+  git diff HEAD^ | filter_excludes | clang-tidy-diff-7.py -p 1
 else
   echo "Running clang-tidy-diff against master branch..."
   git fetch https://github.com/envoyproxy/envoy.git master
-  git diff $(git merge-base HEAD FETCH_HEAD)..HEAD | exclude_testdata | \
+  git diff $(git merge-base HEAD FETCH_HEAD)..HEAD | filter_excludes | \
     clang-tidy-diff-7.py -p 1
 fi

source/common/chromium_url/BUILD

@@ -0,0 +1,28 @@
+licenses(["notice"])  # Apache 2
+
+load(
+    "//bazel:envoy_build_system.bzl",
+    "envoy_cc_library",
+    "envoy_package",
+)
+
+envoy_package()
+
+envoy_cc_library(
+    name = "chromium_url",
+    srcs = [
+        "url_canon.cc",
+        "url_canon_internal.cc",
+        "url_canon_path.cc",
+        "url_canon_stdstring.cc",
+    ],
+    hdrs = [
+        "envoy_shim.h",
+        "url_canon.h",
+        "url_canon_internal.h",
+        "url_canon_stdstring.h",
+        "url_parse.h",
+        "url_parse_internal.h",
+    ],
+    deps = ["//source/common/common:assert_lib"],
+)

source/common/chromium_url/LICENSE

@@ -0,0 +1,27 @@
+// Copyright 2015 The Chromium Authors. All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//    * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//    * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//    * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

source/common/chromium_url/README.md

@@ -0,0 +1,15 @@
+This is a manually minified variant of
+https://chromium.googlesource.com/chromium/src.git/+archive/74.0.3729.15/url.tar.gz,
+providing just the parts needed for `url::CanonicalizePath()`. This is intended
+to support a security release fix for CVE-2019-9901. Long term we need this to
+be moved to absl or QUICHE for upgrades and long-term support.
+
+Some specific transforms of interest:
+* `url_parse.h` is minified to just `Component` and flattened back into the URL
+  directory. It does not contain any non-Chromium authored code any longer and
+  so does not have a separate LICENSE.
+* `envoy_shim.h` adapts various macros to the Envoy context.
+* Anything not reachable from `url::CanonicalizePath()` has been dropped.
+* Header include paths have changed as needed.
+* BUILD was manually written.
+* Various clang-tidy and format fixes.

source/common/chromium_url/envoy_shim.h

@@ -0,0 +1,17 @@
+#pragma once
+
+#include "common/common/assert.h"
+
+// This is a minimal Envoy adaptation layer for the Chromium URL library.
+// NOLINT(namespace-envoy)
+
+#define DISALLOW_COPY_AND_ASSIGN(TypeName)                                                         \
+  TypeName(const TypeName&) = delete;                                                              \
+  TypeName& operator=(const TypeName&) = delete
+
+#define EXPORT_TEMPLATE_DECLARE(x)
+#define EXPORT_TEMPLATE_DEFINE(x)
+#define COMPONENT_EXPORT(x)
+
+#define DCHECK(x) ASSERT(x)
+#define NOTREACHED() NOT_REACHED_GCOVR_EXCL_LINE

source/common/chromium_url/url_canon.cc

@@ -0,0 +1,16 @@
+// Envoy snapshot of Chromium URL path normalization, see README.md.
+// NOLINT(namespace-envoy)
+
+// Copyright 2017 The Chromium 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 "common/chromium_url/url_canon.h"
+
+#include "common/chromium_url/envoy_shim.h"
+
+namespace url {
+
+template class EXPORT_TEMPLATE_DEFINE(COMPONENT_EXPORT(URL)) CanonOutputT<char>;
+
+} // namespace url

source/common/chromium_url/url_canon.h

@@ -0,0 +1,186 @@
+// Envoy snapshot of Chromium URL path normalization, see README.md.
+// NOLINT(namespace-envoy)
+
+// Copyright 2013 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef URL_URL_CANON_H_
+#define URL_URL_CANON_H_
+
+#include <stdlib.h>
+#include <string.h>
+
+#include "common/chromium_url/envoy_shim.h"
+#include "common/chromium_url/url_parse.h"
+
+namespace url {
+
+// Canonicalizer output -------------------------------------------------------
+
+// Base class for the canonicalizer output, this maintains a buffer and
+// supports simple resizing and append operations on it.
+//
+// It is VERY IMPORTANT that no virtual function calls be made on the common
+// code path. We only have two virtual function calls, the destructor and a
+// resize function that is called when the existing buffer is not big enough.
+// The derived class is then in charge of setting up our buffer which we will
+// manage.
+template <typename T> class CanonOutputT {
+public:
+  CanonOutputT() : buffer_(NULL), buffer_len_(0), cur_len_(0) {}
+  virtual ~CanonOutputT() {}
+
+  // Implemented to resize the buffer. This function should update the buffer
+  // pointer to point to the new buffer, and any old data up to |cur_len_| in
+  // the buffer must be copied over.
+  //
+  // The new size |sz| must be larger than buffer_len_.
+  virtual void Resize(int sz) = 0;
+
+  // Accessor for returning a character at a given position. The input offset
+  // must be in the valid range.
+  inline T at(int offset) const { return buffer_[offset]; }
+
+  // Sets the character at the given position. The given position MUST be less
+  // than the length().
+  inline void set(int offset, T ch) { buffer_[offset] = ch; }
+
+  // Returns the number of characters currently in the buffer.
+  inline int length() const { return cur_len_; }
+
+  // Returns the current capacity of the buffer. The length() is the number of
+  // characters that have been declared to be written, but the capacity() is
+  // the number that can be written without reallocation. If the caller must
+  // write many characters at once, it can make sure there is enough capacity,
+  // write the data, then use set_size() to declare the new length().
+  int capacity() const { return buffer_len_; }
+
+  // Called by the user of this class to get the output. The output will NOT
+  // be NULL-terminated. Call length() to get the
+  // length.
+  const T* data() const { return buffer_; }
+  T* data() { return buffer_; }
+
+  // Shortens the URL to the new length. Used for "backing up" when processing
+  // relative paths. This can also be used if an external function writes a lot
+  // of data to the buffer (when using the "Raw" version below) beyond the end,
+  // to declare the new length.
+  //
+  // This MUST NOT be used to expand the size of the buffer beyond capacity().
+  void set_length(int new_len) { cur_len_ = new_len; }
+
+  // This is the most performance critical function, since it is called for
+  // every character.
+  void push_back(T ch) {
+    // In VC2005, putting this common case first speeds up execution
+    // dramatically because this branch is predicted as taken.
+    if (cur_len_ < buffer_len_) {
+      buffer_[cur_len_] = ch;
+      cur_len_++;
+      return;
+    }
+
+    // Grow the buffer to hold at least one more item. Hopefully we won't have
+    // to do this very often.
+    if (!Grow(1))
+      return;
+
+    // Actually do the insertion.
+    buffer_[cur_len_] = ch;
+    cur_len_++;
+  }
+
+  // Appends the given string to the output.
+  void Append(const T* str, int str_len) {
+    if (cur_len_ + str_len > buffer_len_) {
+      if (!Grow(cur_len_ + str_len - buffer_len_))
+        return;
+    }
+    for (int i = 0; i < str_len; i++)
+      buffer_[cur_len_ + i] = str[i];
+    cur_len_ += str_len;
+  }
+
+  void ReserveSizeIfNeeded(int estimated_size) {
+    // Reserve a bit extra to account for escaped chars.
+    if (estimated_size > buffer_len_)
+      Resize(estimated_size + 8);
+  }
+
+protected:
+  // Grows the given buffer so that it can fit at least |min_additional|
+  // characters. Returns true if the buffer could be resized, false on OOM.
+  bool Grow(int min_additional) {
+    static const int kMinBufferLen = 16;
+    int new_len = (buffer_len_ == 0) ? kMinBufferLen : buffer_len_;
+    do {
+      if (new_len >= (1 << 30)) // Prevent overflow below.
+        return false;
+      new_len *= 2;
+    } while (new_len < buffer_len_ + min_additional);
+    Resize(new_len);
+    return true;
+  }
+
+  T* buffer_;
+  int buffer_len_;
+
+  // Used characters in the buffer.
+  int cur_len_;
+};
+
+// Simple implementation of the CanonOutput using new[]. This class
+// also supports a static buffer so if it is allocated on the stack, most
+// URLs can be canonicalized with no heap allocations.
+template <typename T, int fixed_capacity = 1024> class RawCanonOutputT : public CanonOutputT<T> {
+public:
+  RawCanonOutputT() : CanonOutputT<T>() {
+    this->buffer_ = fixed_buffer_;
+    this->buffer_len_ = fixed_capacity;
+  }
+  ~RawCanonOutputT() override {
+    if (this->buffer_ != fixed_buffer_)
+      delete[] this->buffer_;
+  }
+
+  void Resize(int sz) override {
+    T* new_buf = new T[sz];
+    memcpy(new_buf, this->buffer_, sizeof(T) * (this->cur_len_ < sz ? this->cur_len_ : sz));
+    if (this->buffer_ != fixed_buffer_)
+      delete[] this->buffer_;
+    this->buffer_ = new_buf;
+    this->buffer_len_ = sz;
+  }
+
+protected:
+  T fixed_buffer_[fixed_capacity];
+};
+
+// Explicitly instantiate commonly used instantiations.
+extern template class EXPORT_TEMPLATE_DECLARE(COMPONENT_EXPORT(URL)) CanonOutputT<char>;
+
+// Normally, all canonicalization output is in narrow characters. We support
+// the templates so it can also be used internally if a wide buffer is
+// required.
+typedef CanonOutputT<char> CanonOutput;
+
+template <int fixed_capacity>
+class RawCanonOutput : public RawCanonOutputT<char, fixed_capacity> {};
+
+// Path. If the input does not begin in a slash (including if the input is
+// empty), we'll prepend a slash to the path to make it canonical.
+//
+// The 8-bit version assumes UTF-8 encoding, but does not verify the validity
+// of the UTF-8 (i.e., you can have invalid UTF-8 sequences, invalid
+// characters, etc.). Normally, URLs will come in as UTF-16, so this isn't
+// an issue. Somebody giving us an 8-bit path is responsible for generating
+// the path that the server expects (we'll escape high-bit characters), so
+// if something is invalid, it's their problem.
+COMPONENT_EXPORT(URL)
+bool CanonicalizePath(const char* spec, const Component& path, CanonOutput* output,
+                      Component* out_path);
+
+} // namespace url
+
+#endif // URL_URL_CANON_H_