Add parsers for the Encryption and Encryption-Key HTTP headers.

components/gcm_driver/crypto/encryption_header_parsers.cc

+// Copyright 2015 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 "components/gcm_driver/crypto/encryption_header_parsers.h"
+
+#include <map>
+#include <string>
+#include <vector>
+
+#include "base/base64.h"
+#include "base/logging.h"
+#include "base/strings/string_number_conversions.h"
+#include "base/strings/string_tokenizer.h"
+#include "base/strings/string_util.h"
+#include "net/http/http_util.h"
+
+namespace gcm {
+
+namespace {
+
+const char kKeyIdName[] = "keyid";
+const char kSaltName[] = "salt";
+const char kRecordSizeName[] = "rs";
+const char kKeyName[] = "key";
+const char kDiffieHellmanName[] = "dh";
+
+using NameValueMap = std::map<std::string, std::string>;
+using NameValueMapVector = std::vector<NameValueMap>;
+
+// Parses |input| as a header value containing multiple name-value-lists.
+bool ParseMultipleNameValueListsHeader(const std::string& input,
+                                       NameValueMapVector* output) {
+  base::StringTokenizer tokenizer(input, ",");
+  tokenizer.set_quote_chars("'\"");
+
+  while (tokenizer.GetNext()) {
+    NameValueMap name_value_map;
+
+    net::HttpUtil::NameValuePairsIterator parser(tokenizer.token_begin(),

[johnme, 2015/07/21 14:34:43]

I notice that NameValuePairsIterator allows arbitrary LWS. But linear whitespace
is advised against by https://tools.ietf.org/html/std68 as "no longer legal in
mail headers and have caused interoperability problems in other contexts. Do not
use when defining mail headers and use with caution in other contexts."

So if the webpush spec allows LWS, it should perhaps stop doing so.
Or if it doesn't allow LWS, then this implementation doesn't strictly match the
webpush spec, and we should add support to NameValuePairsIterator for whatever
whitespace the webpush spec does allow.

Then again, there's value in consistency.

[Peter Beverloo, 2015/07/21 17:51:30]

I'd like to defer to Ryan for final judgement here, but as far as I can tell
Chrome isn't very strict in header parsing. It feels silly for these things to
be sensitive to whitespace. Mismatch between reality and spec theoretics? NVPI
has more relaxed semantics.

Additionally, by being able to use NameValuePairsIterator we avoid duplicating a
lot of logic.

+                                                 tokenizer.token_end(), ';');
+    if (!parser.valid())

[johnme, 2015/07/21 14:34:43]

It's not actually possible for valid to be false before you call GetNext.
Conversely, you don't check valid() after calling GetNext, so given partially
valid input like "salt=abcd;rs" your implementation will return true having
parsed salt only, even though the rs was invalid.

I suspect what you want here is to move the !parser.valid() check after your
parse.GetNext() while loop, so that you return false for inputs like
"salt=abcd;rs".

[Peter Beverloo, 2015/07/21 17:51:30]

Done.

+      return false;

[johnme, 2015/07/21 14:34:43]

It might also make sense to call output->clear() before returning false?

[Peter Beverloo, 2015/07/21 17:51:30]

Done.

+
+    while (parser.GetNext())
+      name_value_map[parser.name()] = parser.value();
+
+    output->push_back(name_value_map);
+  }
+
+  return true;
+}
+
+// TODO(peter): Generalize a URL-safe base64 implementation.

[johnme, 2015/07/21 14:34:43]

Please link to https://tools.ietf.org/html/rfc4648#section-5

[Peter Beverloo, 2015/07/21 17:51:30]

Done.

+bool Base64DecodeUrlSafe(const std::string& input, std::string* output) {
+  if (input.find_first_of("+/=") != std::string::npos)

[johnme, 2015/07/21 14:34:43]

Why do you disallow '='? Whilst padding is technically unnecessary,
https://tools.ietf.org/html/rfc4648#section-3.2 says "Implementations MUST
include appropriate pad characters at the end of encoded data unless the
specification referring to this document explicitly states otherwise.", and
https://tools.ietf.org/html/draft-thomson-http-encryption-01 does not say
otherwise. And if Martin updates the spec to make it optional, it might still
make sense to allow the padding even though redundant (unless the spec disallows
that).

[Peter Beverloo, 2015/07/21 17:51:30]

Done.

+    return false;
+
+  // Add padding.
+  size_t padded_size = (input.size() + 3) - (input.size() + 3) % 4;
+  std::string padded_input(input);
+  padded_input.resize(padded_size, '=');
+
+  // Convert to standard base64 alphabet.
+  base::ReplaceChars(padded_input, "-", "+", &padded_input);
+  base::ReplaceChars(padded_input, "_", "/", &padded_input);
+
+  return base::Base64Decode(padded_input, output);
+}
+
+// Parses the "salt" field of the Encryption header. Must be a base64 URL

[johnme, 2015/07/21 14:34:43]

Nit: "base64url" for consistency with the naming in the RFC (and also to make it
very clear that this != ordinary base64)?

[Peter Beverloo, 2015/07/21 17:51:30]

Done.

+// encoded string that decodes to a string exactly 16 bytes in length.
+bool ParseSalt(const std::string& value, std::string* output) {
+  std::string decoded_value;
+  if (!Base64DecodeUrlSafe(value, &decoded_value))
+    return false;
+
+  if (decoded_value.size() != 16)
+    return false;
+
+  output->swap(decoded_value);
+  return true;
+}
+
+// Parses the "rs" field of the Encryption header. When present, the value must
+// be a positive decimal integer greater than 1.
+bool ParseRecordSize(const std::string& value, int64_t* output) {
+  int64_t decimal_value = 0;
+  if (!base::StringToInt64(value, &decimal_value))
+    return false;
+
+  if (decimal_value <= 1)
+    return false;
+
+  *output = decimal_value;
+  return true;
+}
+
+// Parses the "key" field of the Encryption-Key header. When present, the value
+// must be a base64 URL encoded string that decodes to a string at least 16
+// bytes in length.
+bool ParseKey(const std::string& value, std::string* output) {
+  std::string decoded_value;
+  if (!Base64DecodeUrlSafe(value, &decoded_value))
+    return false;
+
+  if (decoded_value.size() < 16)
+    return false;
+
+  output->swap(decoded_value);
+  return true;
+}
+
+}  // namespace
+
+bool ParseEncryptionHeader(const std::string& input,
+                           std::vector<EncryptionHeaderValue>* result) {
+  DCHECK(result);
+
+  NameValueMapVector parsed_input;
+  if (!ParseMultipleNameValueListsHeader(input, &parsed_input))
+    return false;
+
+  for (const auto& value_map : parsed_input) {
+    EncryptionHeaderValue value;
+
+    // Optional field: "keyid". May contain any string.
+    const auto& keyid_iter = value_map.find(kKeyIdName);
+    if (keyid_iter != value_map.end())
+      value.keyid = keyid_iter->second;
+
+    // Required field: "salt". Must contain a base64 URL-encoded string that
+    // decodes to a string that is exactly 16-bytes in length.
+    const auto& salt_iter = value_map.find(kSaltName);
+    if (salt_iter == value_map.end())
+      return false;

[johnme, 2015/07/21 14:34:43]

If value_map is completely empty, it should be ok to skip it and continue
(rather than returning false). See earlier comment about allowing "foo ,
,bar,charlie   ".

[Peter Beverloo, 2015/07/21 17:51:30]

Done.

+
+    if (!ParseSalt(salt_iter->second, &value.salt))
+      return false;
+
+    // Optional field: "rs". Must contain a positive decimal integer greater
+    // than one when supplied.
+    const auto& record_size_iter = value_map.find(kRecordSizeName);
+    if (record_size_iter != value_map.end()) {
+      if (!ParseRecordSize(record_size_iter->second, &value.rs))
+        return false;
+    }
+
+    result->push_back(value);
+  }
+
+  return true;
+}
+
+bool ParseEncryptionKeyHeader(const std::string& input,
+                              std::vector<EncryptionKeyHeaderValue>* result) {
+  DCHECK(result);
+
+  NameValueMapVector parsed_input;
+  if (!ParseMultipleNameValueListsHeader(input, &parsed_input))
+    return false;
+
+  for (const auto& value_map : parsed_input) {
+    EncryptionKeyHeaderValue value;
+
+    // Optional field: "keyid". May contain any string.
+    const auto& keyid_iter = value_map.find(kKeyIdName);
+    if (keyid_iter != value_map.end())
+      value.keyid = keyid_iter->second;
+
+    // Optional field: "key". Must contain a base64 URL-encoded string with the
+    // explicit encryption key.
+    const auto& key_iter = value_map.find(kKeyName);
+    if (key_iter != value_map.end()) {
+      if (!ParseKey(key_iter->second, &value.key))
+        return false;
+    }
+
+    // Optional field: "dh". Must contain a base64 URL-encoded string with the
+    // Diffie-Hellman share (either modp or elliptic curve).
+    const auto& dh_iter = value_map.find(kDiffieHellmanName);
+    if (dh_iter != value_map.end()) {
+      if (!Base64DecodeUrlSafe(dh_iter->second, &value.dh))
+        return false;
+    }
+
+    result->push_back(value);
+  }
+
+  return true;
+}
+
+}  // namespace gcm