Certificate Transparency: Code for unpacking EV cert hashes whitelist

chrome/browser/net/packed_ct_ev_whitelist.cc

+// Copyright 2014 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 "chrome/browser/net/packed_ct_ev_whitelist.h"
+
+#include <string.h>
+
+#include <algorithm>
+
+#include "base/big_endian.h"
+#include "base/files/file_util.h"
+#include "base/lazy_instance.h"
+#include "base/logging.h"
+#include "chrome/browser/net/bit_stream_reader.h"
+#include "content/public/browser/browser_thread.h"
+#include "net/ssl/ssl_config_service.h"
+
+namespace {
+const uint8_t kCertHashLengthBits = 64;  // 8 bytes
+const uint8_t kCertHashLength = kCertHashLengthBits / 8;
+const uint64_t kGolombMParameterBits = 47;  // 2^47
+
+void SetNewEVWhitelistInSSLConfigService(
+    const scoped_refptr<net::ct::EVCertsWhitelist>& new_whitelist) {
+  net::SSLConfigService::SetEVCertsWhitelist(new_whitelist);
+}
+
+int TruncatedHashesComparator(const void* v1, const void* v2) {
+  return *(static_cast<const uint64_t*>(v1)) <
+         *(static_cast<const uint64_t*>(v2));
+}
+}  // namespace
+
+void SetEVWhitelistFromFile(const base::FilePath& compressed_whitelist_file) {
+  VLOG(1) << "Setting EV whitelist from file: "
+          << compressed_whitelist_file.value();
+  std::string compressed_list;
+  if (!base::ReadFileToString(compressed_whitelist_file, &compressed_list)) {
+    VLOG(1) << "Failed reading from " << compressed_whitelist_file.value();
+    return;
+  }
+
+  scoped_refptr<net::ct::EVCertsWhitelist> new_whitelist(
+      new PackedEVCertsWhitelist(compressed_list));
+  if (!new_whitelist->IsValid()) {
+    VLOG(1) << "Failed uncompressing EV certs whitelist.";
+    return;
+  }
+
+  base::Callback<void(void)> assign_cb =

[Ryan Sleevi, 2014/10/27 17:09:48]

you can use base::Closure, but that's no big deal.

[Eran Messeri, 2014/10/28 11:43:58]

Done.

+      base::Bind(SetNewEVWhitelistInSSLConfigService, new_whitelist);
+  content::BrowserThread::PostTask(
+      content::BrowserThread::IO, FROM_HERE, assign_cb);
+}
+
+bool PackedEVCertsWhitelist::UncompressEVWhitelist(
+    const std::string& compressed_whitelist,
+    std::vector<std::string>* uncompressed_list) {
+  internal::BitStreamReader reader(base::StringPiece(
+      compressed_whitelist.data(), compressed_whitelist.size()));
+  std::vector<std::string> result;
+
+  VLOG(1) << "Uncompressing EV whitelist of size "
+          << compressed_whitelist.size();
+  uint64_t curr_hash(0);
+  if (!reader.ReadBits(kCertHashLengthBits, &curr_hash)) {
+    VLOG(1) << "Failed reading first hash.";
+    return false;
+  }
+  char hash_bytes[kCertHashLength];
+  base::WriteBigEndian(hash_bytes, curr_hash);
+  result.push_back(std::string(hash_bytes, kCertHashLength));
+  // M is the tunable parameter used by the Golomb coding.
+  static const uint64_t kGolombParameterM = static_cast<uint64_t>(1)
+                                            << kGolombMParameterBits;
+
+  while (reader.BitsLeft() > kGolombMParameterBits) {
+    uint64_t read_prefix = 0;
+    if (!reader.ReadUnaryEncoding(&read_prefix)) {
+      VLOG(1) << "Failed reading unary-encoded prefix.";
+      return false;
+    }
+    if (read_prefix > (UINT64_MAX / kGolombParameterM)) {
+      VLOG(1) << "Received value that would cause overflow: " << read_prefix;
+      return false;
+    }
+
+    uint64_t r = 0;
+    if (!reader.ReadBits(kGolombMParameterBits, &r)) {
+      VLOG(1) << "Failed reading " << kGolombMParameterBits << " bits.";
+      return false;
+    }
+    DCHECK_LT(r, kGolombParameterM);
+
+    uint64_t curr_diff = read_prefix * kGolombParameterM + r;
+    curr_hash += curr_diff;
+
+    base::WriteBigEndian(hash_bytes, curr_hash);
+    result.push_back(std::string(hash_bytes, kCertHashLength));

[Ryan Sleevi, 2014/10/27 17:09:48]

Why convert to an std::string, when you have it as a uint64_t here?

That is, why can't this just be an std::vector<uint64_t>? Did I miss something?

[Eran Messeri, 2014/10/28 11:43:58]

No reason - changed.

+  }
+
+  uncompressed_list->swap(result);
+  return true;
+}
+
+PackedEVCertsWhitelist::PackedEVCertsWhitelist(
+    const std::string& compressed_whitelist)
+    : is_whitelist_valid_(false) {
+  std::vector<std::string> new_whitelist;
+  if (!UncompressEVWhitelist(compressed_whitelist, &new_whitelist)) {
+    whitelist_.clear();
+    return;
+  }
+
+  whitelist_.resize(new_whitelist.size());
+  for (size_t i = 0; i < new_whitelist.size(); ++i) {
+    base::ReadBigEndian(new_whitelist[i].data(), &(whitelist_[i]));
+  }
+
+  is_whitelist_valid_ = true;
+}
+
+PackedEVCertsWhitelist::~PackedEVCertsWhitelist() {
+}
+
+bool PackedEVCertsWhitelist::ContainsCertificateHash(
+    const std::string& certificate_hash) const {
+  DCHECK(!whitelist_.empty());
+  return bsearch(certificate_hash.c_str(),

[Ryan Sleevi, 2014/10/27 17:09:48]

Isn't this where you should be ensuring big endian conversion? Otherwise, you're
just directly casting certificate_hash to a uint64_t

[Eran Messeri, 2014/10/28 11:43:58]

Good catch, done.

+                 &whitelist_[0],
+                 whitelist_.size(),
+                 kCertHashLength,
+                 TruncatedHashesComparator) != NULL;
+}
+
+bool PackedEVCertsWhitelist::IsValid() const {
+  return is_whitelist_valid_;
+}