[autofill] Add support for PSL domain matching for password autofill.

chrome/browser/password_manager/login_database.cc

 // Copyright (c) 2012 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/password_manager/login_database.h"
 
 #include <algorithm>
 #include <limits>
 
+#include "base/command_line.h"
 #include "base/file_util.h"
 #include "base/files/file_path.h"
 #include "base/logging.h"
 #include "base/metrics/histogram.h"
 #include "base/pickle.h"
+#include "base/string_util.h"
 #include "base/strings/string_number_conversions.h"
 #include "base/time.h"
 #include "base/utf_string_conversions.h"
+#include "chrome/common/chrome_switches.h"
+#include "net/base/registry_controlled_domains/registry_controlled_domain.h"
 #include "sql/statement.h"
 #include "sql/transaction.h"
 
 using content::PasswordForm;
 
 static const int kCurrentVersionNumber = 3;
 static const int kCompatibleVersionNumber = 1;
 
 namespace {
 
@@ skipping 10 matching lines @@
   COLUMN_SSL_VALID,
   COLUMN_PREFERRED,
   COLUMN_DATE_CREATED,
   COLUMN_BLACKLISTED_BY_USER,
   COLUMN_SCHEME,
   COLUMN_PASSWORD_TYPE,
   COLUMN_POSSIBLE_USERNAMES,
   COLUMN_TIMES_USED
 };
 
+std::string GetPSLDomain(std::string signon_realm_str) {

[Ilya Sherman, 2013/06/06 09:25:35]

nit: Pass by const reference.

[Ilya Sherman, 2013/06/06 09:25:35]

Please spell out acronyms like PSL; or better yet, find a briefer alternative
that is still clear to someone unfamiliar with the acronym.

[palmer, 2013/06/06 21:16:42]

|GetPublicDomainSuffix|? |GetPublicSuffix|?

[nyquist, 2013/06/07 22:51:10]

Is |GetRegistryControlledDomain| OK? Since we technically get more than just the
suffix.

+  GURL signon_realm(signon_realm_str);

[palmer, 2013/06/06 21:16:42]

Declare this const if the callee takes a const &.

[nyquist, 2013/06/07 22:51:10]

Done.

+  return net::registry_controlled_domains::GetDomainAndRegistry(
+      signon_realm,
+      net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES);
+}
+
 }  // namespace
 
 LoginDatabase::LoginDatabase() {
 }
 
 LoginDatabase::~LoginDatabase() {
 }
 
 bool LoginDatabase::Init(const base::FilePath& db_path) {
   // Set pragmas for a small, private database (based on WebDatabase).
@@ skipping 35 matching lines @@
   if (!MigrateOldVersionsAsNeeded()) {
     LOG(WARNING) << "Unable to migrate database";
     db_.Close();
     return false;
   }
 
   if (!transaction.Commit()) {
     db_.Close();
     return false;
   }
+
+  psl_domain_matching_ = CommandLine::ForCurrentProcess()->HasSwitch(
+      switches::kEnablePasswordAutofillPSLDomainMatching);
+
   return true;
 }
 
 bool LoginDatabase::MigrateOldVersionsAsNeeded() {
   switch (meta_table_.GetVersionNumber()) {
     case 1:
       if (!db_.Execute("ALTER TABLE logins "
                        "ADD COLUMN password_type INTEGER") ||
           !db_.Execute("ALTER TABLE logins "
                        "ADD COLUMN possible_usernames BLOB")) {
@@ skipping 214 matching lines @@
   form->origin = GURL(tmp);
   tmp = s.ColumnString(COLUMN_ACTION_URL);
   form->action = GURL(tmp);
   form->username_element = s.ColumnString16(COLUMN_USERNAME_ELEMENT);
   form->username_value = s.ColumnString16(COLUMN_USERNAME_VALUE);
   form->password_element = s.ColumnString16(COLUMN_PASSWORD_ELEMENT);
   form->password_value = decrypted_password;
   form->submit_element = s.ColumnString16(COLUMN_SUBMIT_ELEMENT);
   tmp = s.ColumnString(COLUMN_SIGNON_REALM);
   form->signon_realm = tmp;
+  form->original_signon_realm = "";
+  form->is_psl_origin_match = false;
   form->ssl_valid = (s.ColumnInt(COLUMN_SSL_VALID) > 0);
   form->preferred = (s.ColumnInt(COLUMN_PREFERRED) > 0);
   form->date_created = base::Time::FromTimeT(
       s.ColumnInt64(COLUMN_DATE_CREATED));
   form->blacklisted_by_user = (s.ColumnInt(COLUMN_BLACKLISTED_BY_USER) > 0);
   int scheme_int = s.ColumnInt(COLUMN_SCHEME);
   DCHECK((scheme_int >= 0) && (scheme_int <= PasswordForm::SCHEME_OTHER));
   form->scheme = static_cast<PasswordForm::Scheme>(scheme_int);
   int type_int = s.ColumnInt(COLUMN_PASSWORD_TYPE);
   DCHECK(type_int >= 0 && type_int <= PasswordForm::TYPE_GENERATED);
   form->type = static_cast<PasswordForm::Type>(type_int);
   Pickle pickle(
       static_cast<const char*>(s.ColumnBlob(COLUMN_POSSIBLE_USERNAMES)),
       s.ColumnByteLength(COLUMN_POSSIBLE_USERNAMES));
   form->other_possible_usernames = DeserializeVector(pickle);
   form->times_used = s.ColumnInt(COLUMN_TIMES_USED);
   return true;
 }
 
 bool LoginDatabase::GetLogins(const PasswordForm& form,
                               std::vector<PasswordForm*>* forms) const {
   DCHECK(forms);
   // You *must* change LoginTableColumns if this query changes.
-  sql::Statement s(db_.GetCachedStatement(SQL_FROM_HERE,
-      "SELECT origin_url, action_url, "
+  std::string signon_realm_operator = psl_domain_matching_ ? "regexp" : "==";
+  std::string sql_query = "SELECT origin_url, action_url, "

[palmer, 2013/06/06 21:16:42]

I think you can declare these strings const? I'm big on that because it's a nice
guarantee.

[nyquist, 2013/06/07 22:51:10]

Done.

       "username_element, username_value, "
       "password_element, password_value, submit_element, "
       "signon_realm, ssl_valid, preferred, date_created, blacklisted_by_user, "
       "scheme, password_type, possible_usernames, times_used "
-      "FROM logins WHERE signon_realm == ? "));
-  s.BindString(0, form.signon_realm);
+      "FROM logins WHERE signon_realm " + signon_realm_operator + " ? ";
+  sql::Statement s;
+  if (psl_domain_matching_) {
+    // TODO(nyquist) Re-enable usage of cached statements if possible.
+    // s.Assign(db_.GetCachedStatement(SQL_FROM_HERE, sql_query.c_str()));

[palmer, 2013/06/06 21:16:42]

Perhaps you still can? Have you tried it? It should be possible to cache the
statement and then bind new arguments in without invalidating the cache. I
think.

[nyquist, 2013/06/07 22:51:10]

Will try again before I submit this. I agree that it should work, but it didn't
for some very odd reason that I don't understand.

+    s.Assign(db_.GetUniqueStatement(sql_query.c_str()));
+    std::string domain = GetPSLDomain(form.signon_realm);
+    // We need to escape and . in the domain. Since the domain has already been
+    // sanitized using GURL, we do not need to escape any other characters.
+    ReplaceChars(domain, ".", "\\.", &domain);
+    // For a domain such as foo.bar, this regexp will match domains for any
+    // scheme and on the form: http://foo.bar/, http://www.foo.bar/,
+    // http://www.mobile.foo.bar/. It will not match http://notfoo.bar/.
+    std::string regexp = "(\\w+?:\\/\\/)([\\w\\-_]+\\.)*" + domain + "\\/$";

[palmer, 2013/06/06 21:16:42]

Can we instead use an explicit list of schemata, instead of \w+ ? There are some
new and exciting schemata like googlesearch and so on that I'd like to ensure
never get involved in this. Not that I have any specific fears, just that
explicitness is more safe in a thing like this.

[nyquist, 2013/06/07 22:51:10]

Done. Also added ^ at start of query. Still keeping any port though, since we
are regardless ignoring invalid SSL things later.

+    s.BindString(0, regexp);
+  } else {
+    s.Assign(db_.GetCachedStatement(SQL_FROM_HERE, sql_query.c_str()));
+    s.BindString(0, form.signon_realm);
+  }
 
   while (s.Step()) {
     scoped_ptr<PasswordForm> new_form(new PasswordForm());
     if (!InitPasswordFormFromStatement(new_form.get(), s))
       return false;
+    if (psl_domain_matching_) {
+      std::string found_psl_domain = GetPSLDomain(new_form->signon_realm);

[palmer, 2013/06/06 21:16:42]

More constiness, if possible

[nyquist, 2013/06/07 22:51:10]

Done.

+      std::string form_psl_domain = GetPSLDomain(form.signon_realm);
+      if (found_psl_domain != form_psl_domain) {
+        // The database returned results that should not match. Skipping result.
+        continue;
+      }
+      if (form.signon_realm != new_form->signon_realm) {
+        // This is not a perfect match, so we need to create a new valid result.
+        // We do this by copying over origin, signon realm and action from the
+        // observed form and setting the original signon realm to what we found
+        // in the database. We also set a flag to notify the caller that this
+        // is not a perfect match.
+        new_form->is_psl_origin_match = true;
+        new_form->original_signon_realm = new_form->signon_realm;
+        new_form->origin = form.origin;
+        new_form->signon_realm = form.signon_realm;
+        new_form->action = form.action;
+      }
+    }
     forms->push_back(new_form.release());
   }
   return s.Succeeded();
 }
 
 bool LoginDatabase::GetLoginsCreatedBetween(
     const base::Time begin,
     const base::Time end,
     std::vector<content::PasswordForm*>* forms) const {
   DCHECK(forms);
@@ skipping 70 matching lines @@
 std::vector<string16> LoginDatabase::DeserializeVector(const Pickle& p) const {
   std::vector<string16> ret;
   string16 str;
 
   PickleIterator iterator(p);
   while (iterator.ReadString16(&str)) {
     ret.push_back(str);
   }
   return ret;
 }