1// Copyright (c) 2011 The Chromium Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
5// Brought to you by the letter D and the number 2.
6
7#ifndef NET_BASE_COOKIE_MONSTER_H_
8#define NET_BASE_COOKIE_MONSTER_H_
9#pragma once
10
11#include <map>
12#include <string>
13#include <utility>
14#include <vector>
15
16#include "base/basictypes.h"
17#include "base/gtest_prod_util.h"
18#include "base/memory/ref_counted.h"
19#include "base/memory/scoped_ptr.h"
20#include "base/synchronization/lock.h"
21#include "base/task.h"
22#include "base/time.h"
23#include "net/base/cookie_store.h"
24
25class GURL;
26
27namespace base {
28class Histogram;
29}
30
31namespace net {
32
33class CookieList;
34
35// The cookie monster is the system for storing and retrieving cookies. It has
36// an in-memory list of all cookies, and synchronizes non-session cookies to an
37// optional permanent storage that implements the PersistentCookieStore
38// interface.
39//
40// This class IS thread-safe. Normally, it is only used on the I/O thread, but
41// is also accessed directly through Automation for UI testing.
42//
43// TODO(deanm) Implement CookieMonster, the cookie database.
44//  - Verify that our domain enforcement and non-dotted handling is correct
45class NET_EXPORT CookieMonster : public CookieStore {
46 public:
47  class CanonicalCookie;
48  class Delegate;
49  class ParsedCookie;
50  class PersistentCookieStore;
51
52  // Terminology:
53  //    * The 'top level domain' (TLD) of an internet domain name is
54  //      the terminal "." free substring (e.g. "com" for google.com
55  //      or world.std.com).
56  //    * The 'effective top level domain' (eTLD) is the longest
57  //      "." initiated terminal substring of an internet domain name
58  //      that is controlled by a general domain registrar.
59  //      (e.g. "co.uk" for news.bbc.co.uk).
60  //    * The 'effective top level domain plus one' (eTLD+1) is the
61  //      shortest "." delimited terminal substring of an internet
62  //      domain name that is not controlled by a general domain
63  //      registrar (e.g. "bbc.co.uk" for news.bbc.co.uk, or
64  //      "google.com" for news.google.com).  The general assumption
65  //      is that all hosts and domains under an eTLD+1 share some
66  //      administrative control.
67
68  // CookieMap is the central data structure of the CookieMonster.  It
69  // is a map whose values are pointers to CanonicalCookie data
70  // structures (the data structures are owned by the CookieMonster
71  // and must be destroyed when removed from the map).  There are two
72  // possible keys for the map, controlled on a per-CookieMonster basis
73  // by expiry_and_key_scheme_/SetExpiryAndKeyScheme()
74  // (defaulted by expiry_and_key_default_):
75
76  // If expiry_and_key_scheme_ is EKS_KEEP_RECENT_AND_PURGE_ETLDP1
77  // (default), then the key is based on the effective domain of the
78  // cookies.  If the domain of the cookie has an eTLD+1, that is the
79  // key for the map.  If the domain of the cookie does not have an eTLD+1,
80  // the key of the map is the host the cookie applies to (it is not
81  // legal to have domain cookies without an eTLD+1).  This rule
82  // excludes cookies for, e.g, ".com", ".co.uk", or ".internalnetwork".
83  // This behavior is the same as the behavior in Firefox v 3.6.10.
84
85  // If use_effective_domain_key_scheme_ is EKS_DISCARD_RECENT_AND_PURGE_DOMAIN,
86  // then the key is just the domain of the cookie.  Eventually, this
87  // option will be removed.
88
89  // NOTE(deanm):
90  // I benchmarked hash_multimap vs multimap.  We're going to be query-heavy
91  // so it would seem like hashing would help.  However they were very
92  // close, with multimap being a tiny bit faster.  I think this is because
93  // our map is at max around 1000 entries, and the additional complexity
94  // for the hashing might not overcome the O(log(1000)) for querying
95  // a multimap.  Also, multimap is standard, another reason to use it.
96  // TODO(rdsmith): This benchmark should be re-done now that we're allowing
97  // subtantially more entries in the map.
98  typedef std::multimap<std::string, CanonicalCookie*> CookieMap;
99  typedef std::pair<CookieMap::iterator, CookieMap::iterator> CookieMapItPair;
100
101  // The key and expiry scheme to be used by the monster.
102  // EKS_KEEP_RECENT_AND_PURGE_ETLDP1 means to use
103  // the new key scheme based on effective domain and save recent cookies
104  // in global garbage collection.  EKS_DISCARD_RECENT_AND_PURGE_DOMAIN
105  // means to use the old key scheme based on full domain and be ruthless
106  // about purging.
107  enum ExpiryAndKeyScheme {
108    EKS_KEEP_RECENT_AND_PURGE_ETLDP1,
109    EKS_DISCARD_RECENT_AND_PURGE_DOMAIN,
110    EKS_LAST_ENTRY
111  };
112
113  // The store passed in should not have had Init() called on it yet. This
114  // class will take care of initializing it. The backing store is NOT owned by
115  // this class, but it must remain valid for the duration of the cookie
116  // monster's existence. If |store| is NULL, then no backing store will be
117  // updated. If |delegate| is non-NULL, it will be notified on
118  // creation/deletion of cookies.
119  CookieMonster(PersistentCookieStore* store, Delegate* delegate);
120
121  // Only used during unit testing.
122  CookieMonster(PersistentCookieStore* store,
123                Delegate* delegate,
124                int last_access_threshold_milliseconds);
125
126  // Parses the string with the cookie time (very forgivingly).
127  static base::Time ParseCookieTime(const std::string& time_string);
128
129  // Returns true if a domain string represents a host-only cookie,
130  // i.e. it doesn't begin with a leading '.' character.
131  static bool DomainIsHostOnly(const std::string& domain_string);
132
133  // Sets a cookie given explicit user-provided cookie attributes. The cookie
134  // name, value, domain, etc. are each provided as separate strings. This
135  // function expects each attribute to be well-formed. It will check for
136  // disallowed characters (e.g. the ';' character is disallowed within the
137  // cookie value attribute) and will return false without setting the cookie
138  // if such characters are found.
139  bool SetCookieWithDetails(const GURL& url,
140                            const std::string& name,
141                            const std::string& value,
142                            const std::string& domain,
143                            const std::string& path,
144                            const base::Time& expiration_time,
145                            bool secure, bool http_only);
146
147  // Returns all the cookies, for use in management UI, etc. This does not mark
148  // the cookies as having been accessed.
149  // The returned cookies are ordered by longest path, then by earliest
150  // creation date.
151  CookieList GetAllCookies();
152
153  // Returns all the cookies, for use in management UI, etc. Filters results
154  // using given url scheme, host / domain and path and options. This does not
155  // mark the cookies as having been accessed.
156  // The returned cookies are ordered by longest path, then earliest
157  // creation date.
158  CookieList GetAllCookiesForURLWithOptions(const GURL& url,
159                                            const CookieOptions& options);
160
161  // Invokes GetAllCookiesForURLWithOptions with options set to include HTTP
162  // only cookies.
163  CookieList GetAllCookiesForURL(const GURL& url);
164
165  // Deletes all of the cookies.
166  int DeleteAll(bool sync_to_store);
167  // Deletes all of the cookies that have a creation_date greater than or equal
168  // to |delete_begin| and less than |delete_end|
169  int DeleteAllCreatedBetween(const base::Time& delete_begin,
170                              const base::Time& delete_end,
171                              bool sync_to_store);
172  // Deletes all of the cookies that have a creation_date more recent than the
173  // one passed into the function via |delete_after|.
174  int DeleteAllCreatedAfter(const base::Time& delete_begin, bool sync_to_store);
175
176  // Deletes all cookies that match the host of the given URL
177  // regardless of path.  This includes all http_only and secure cookies,
178  // but does not include any domain cookies that may apply to this host.
179  // Returns the number of cookies deleted.
180  int DeleteAllForHost(const GURL& url);
181
182  // Deletes one specific cookie.
183  bool DeleteCanonicalCookie(const CanonicalCookie& cookie);
184
185  // Override the default list of schemes that are allowed to be set in
186  // this cookie store.  Calling his overrides the value of
187  // "enable_file_scheme_".
188  // If this this method is called, it must be called before first use of
189  // the instance (i.e. as part of the instance initialization process).
190  void SetCookieableSchemes(const char* schemes[], size_t num_schemes);
191
192  // Overrides the default key and expiry scheme.  See comments
193  // before CookieMap and Garbage collection constants for details.  This
194  // function must be called before initialization.
195  void SetExpiryAndKeyScheme(ExpiryAndKeyScheme key_scheme);
196
197  // Instructs the cookie monster to not delete expired cookies. This is used
198  // in cases where the cookie monster is used as a data structure to keep
199  // arbitrary cookies.
200  void SetKeepExpiredCookies();
201
202  // Delegates the call to set the |clear_local_store_on_exit_| flag of the
203  // PersistentStore if it exists.
204  void SetClearPersistentStoreOnExit(bool clear_local_store);
205
206  // There are some unknowns about how to correctly handle file:// cookies,
207  // and our implementation for this is not robust enough. This allows you
208  // to enable support, but it should only be used for testing. Bug 1157243.
209  // Must be called before creating a CookieMonster instance.
210  static void EnableFileScheme();
211
212  // Flush the backing store (if any) to disk and post the given task when done.
213  // WARNING: THE CALLBACK WILL RUN ON A RANDOM THREAD. IT MUST BE THREAD SAFE.
214  // It may be posted to the current thread, or it may run on the thread that
215  // actually does the flushing. Your Task should generally post a notification
216  // to the thread you actually want to be notified on.
217  void FlushStore(Task* completion_task);
218
219  // CookieStore implementation.
220
221  // Sets the cookies specified by |cookie_list| returned from |url|
222  // with options |options| in effect.
223  virtual bool SetCookieWithOptions(const GURL& url,
224                                    const std::string& cookie_line,
225                                    const CookieOptions& options);
226
227  // Gets all cookies that apply to |url| given |options|.
228  // The returned cookies are ordered by longest path, then earliest
229  // creation date.
230  virtual std::string GetCookiesWithOptions(const GURL& url,
231                                            const CookieOptions& options);
232
233  // Deletes all cookies with that might apply to |url| that has |cookie_name|.
234  virtual void DeleteCookie(const GURL& url, const std::string& cookie_name);
235
236  virtual CookieMonster* GetCookieMonster();
237
238  // Debugging method to perform various validation checks on the map.
239  // Currently just checking that there are no null CanonicalCookie pointers
240  // in the map.
241  // Argument |arg| is to allow retaining of arbitrary data if the CHECKs
242  // in the function trip.  TODO(rdsmith):Remove hack.
243  void ValidateMap(int arg);
244
245  // The default list of schemes the cookie monster can handle.
246  static const char* kDefaultCookieableSchemes[];
247  static const int kDefaultCookieableSchemesCount;
248
249 private:
250  // Testing support.
251  // For SetCookieWithCreationTime.
252  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest,
253                           TestCookieDeleteAllCreatedAfterTimestamp);
254  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest,
255                           TestCookieDeleteAllCreatedBetweenTimestamps);
256
257  // For gargage collection constants.
258  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, TestHostGarbageCollection);
259  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, TestTotalGarbageCollection);
260  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, GarbageCollectionTriggers);
261  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, TestGCTimes);
262
263  // For validation of key values.
264  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, TestDomainTree);
265  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, TestImport);
266  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, GetKey);
267  FRIEND_TEST_ALL_PREFIXES(CookieMonsterTest, TestGetKey);
268
269  // Internal reasons for deletion, used to populate informative histograms
270  // and to provide a public cause for onCookieChange notifications.
271  //
272  // If you add or remove causes from this list, please be sure to also update
273  // the Delegate::ChangeCause mapping inside ChangeCauseMapping. Moreover,
274  // these are used as array indexes, so avoid reordering to keep the
275  // histogram buckets consistent. New items (if necessary) should be added
276  // at the end of the list, just before DELETE_COOKIE_LAST_ENTRY.
277  enum DeletionCause {
278    DELETE_COOKIE_EXPLICIT = 0,
279    DELETE_COOKIE_OVERWRITE,
280    DELETE_COOKIE_EXPIRED,
281    DELETE_COOKIE_EVICTED,
282    DELETE_COOKIE_DUPLICATE_IN_BACKING_STORE,
283    DELETE_COOKIE_DONT_RECORD,  // e.g. For final cleanup after flush to store.
284    DELETE_COOKIE_EVICTED_DOMAIN,
285    DELETE_COOKIE_EVICTED_GLOBAL,
286
287    // Cookies evicted during domain level garbage collection that
288    // were accessed longer ago than kSafeFromGlobalPurgeDays
289    DELETE_COOKIE_EVICTED_DOMAIN_PRE_SAFE,
290
291    // Cookies evicted during domain level garbage collection that
292    // were accessed more recently than kSafeFromGlobalPurgeDays
293    // (and thus would have been preserved by global garbage collection).
294    DELETE_COOKIE_EVICTED_DOMAIN_POST_SAFE,
295
296    // A common idiom is to remove a cookie by overwriting it with an
297    // already-expired expiration date. This captures that case.
298    DELETE_COOKIE_EXPIRED_OVERWRITE,
299
300    DELETE_COOKIE_LAST_ENTRY
301  };
302
303  // Cookie garbage collection thresholds.  Based off of the Mozilla defaults.
304  // When the number of cookies gets to k{Domain,}MaxCookies
305  // purge down to k{Domain,}MaxCookies - k{Domain,}PurgeCookies.
306  // It might seem scary to have a high purge value, but really it's not.
307  // You just make sure that you increase the max to cover the increase
308  // in purge, and we would have been purging the same amount of cookies.
309  // We're just going through the garbage collection process less often.
310  // Note that the DOMAIN values are per eTLD+1; see comment for the
311  // CookieMap typedef.  So, e.g., the maximum number of cookies allowed for
312  // google.com and all of its subdomains will be 150-180.
313  //
314  // If the expiry and key scheme follows firefox standards (default,
315  // set by SetExpiryAndKeyScheme()), any cookies accessed more recently
316  // than kSafeFromGlobalPurgeDays will not be evicted by global garbage
317  // collection, even if we have more than kMaxCookies.  This does not affect
318  // domain garbage collection.
319  //
320  // Present in .h file to make accessible to tests through FRIEND_TEST.
321  // Actual definitions are in cookie_monster.cc.
322  static const size_t kDomainMaxCookies;
323  static const size_t kDomainPurgeCookies;
324  static const size_t kMaxCookies;
325  static const size_t kPurgeCookies;
326
327  // The number of days since last access that cookies will not be subject
328  // to global garbage collection.
329  static const int kSafeFromGlobalPurgeDays;
330
331  // Default value for key and expiry scheme scheme.
332  static const ExpiryAndKeyScheme expiry_and_key_default_ =
333      EKS_KEEP_RECENT_AND_PURGE_ETLDP1;
334
335  // Record statistics every kRecordStatisticsIntervalSeconds of uptime.
336  static const int kRecordStatisticsIntervalSeconds = 10 * 60;
337
338  ~CookieMonster();
339
340  bool SetCookieWithCreationTime(const GURL& url,
341                                 const std::string& cookie_line,
342                                 const base::Time& creation_time);
343
344  // Called by all non-static functions to ensure that the cookies store has
345  // been initialized. This is not done during creating so it doesn't block
346  // the window showing.
347  // Note: this method should always be called with lock_ held.
348  void InitIfNecessary() {
349    if (!initialized_) {
350      if (store_)
351        InitStore();
352      initialized_ = true;
353    }
354  }
355
356  // Initializes the backing store and reads existing cookies from it.
357  // Should only be called by InitIfNecessary().
358  void InitStore();
359
360  // Checks that |cookies_| matches our invariants, and tries to repair any
361  // inconsistencies. (In other words, it does not have duplicate cookies).
362  void EnsureCookiesMapIsValid();
363
364  // Checks for any duplicate cookies for CookieMap key |key| which lie between
365  // |begin| and |end|. If any are found, all but the most recent are deleted.
366  // Returns the number of duplicate cookies that were deleted.
367  int TrimDuplicateCookiesForKey(const std::string& key,
368                                 CookieMap::iterator begin,
369                                 CookieMap::iterator end);
370
371  void SetDefaultCookieableSchemes();
372
373  void FindCookiesForHostAndDomain(const GURL& url,
374                                   const CookieOptions& options,
375                                   bool update_access_time,
376                                   std::vector<CanonicalCookie*>* cookies);
377
378  void FindCookiesForKey(const std::string& key,
379                         const GURL& url,
380                         const CookieOptions& options,
381                         const base::Time& current,
382                         bool update_access_time,
383                         std::vector<CanonicalCookie*>* cookies);
384
385  // Delete any cookies that are equivalent to |ecc| (same path, domain, etc).
386  // If |skip_httponly| is true, httponly cookies will not be deleted.  The
387  // return value with be true if |skip_httponly| skipped an httponly cookie.
388  // |key| is the key to find the cookie in cookies_; see the comment before
389  // the CookieMap typedef for details.
390  // NOTE: There should never be more than a single matching equivalent cookie.
391  bool DeleteAnyEquivalentCookie(const std::string& key,
392                                 const CanonicalCookie& ecc,
393                                 bool skip_httponly,
394                                 bool already_expired);
395
396  // Takes ownership of *cc.
397  void InternalInsertCookie(const std::string& key,
398                            CanonicalCookie* cc,
399                            bool sync_to_store);
400
401  // Helper function that sets cookies with more control.
402  // Not exposed as we don't want callers to have the ability
403  // to specify (potentially duplicate) creation times.
404  bool SetCookieWithCreationTimeAndOptions(const GURL& url,
405                                           const std::string& cookie_line,
406                                           const base::Time& creation_time,
407                                           const CookieOptions& options);
408
409
410  // Helper function that sets a canonical cookie, deleting equivalents and
411  // performing garbage collection.
412  bool SetCanonicalCookie(scoped_ptr<CanonicalCookie>* cc,
413                          const base::Time& creation_time,
414                          const CookieOptions& options);
415
416  void InternalUpdateCookieAccessTime(CanonicalCookie* cc,
417                                      const base::Time& current_time);
418
419  // |deletion_cause| argument is used for collecting statistics and choosing
420  // the correct Delegate::ChangeCause for OnCookieChanged notifications.
421  void InternalDeleteCookie(CookieMap::iterator it, bool sync_to_store,
422                            DeletionCause deletion_cause);
423
424  // If the number of cookies for CookieMap key |key|, or globally, are
425  // over the preset maximums above, garbage collect, first for the host and
426  // then globally.  See comments above garbage collection threshold
427  // constants for details.
428  //
429  // Returns the number of cookies deleted (useful for debugging).
430  int GarbageCollect(const base::Time& current, const std::string& key);
431
432  // Helper for GarbageCollect(); can be called directly as well.  Deletes
433  // all expired cookies in |itpair|.  If |cookie_its| is non-NULL, it is
434  // populated with all the non-expired cookies from |itpair|.
435  //
436  // Returns the number of cookies deleted.
437  int GarbageCollectExpired(const base::Time& current,
438                            const CookieMapItPair& itpair,
439                            std::vector<CookieMap::iterator>* cookie_its);
440
441  // Helper for GarbageCollect().  Deletes all cookies in the list
442  // that were accessed before |keep_accessed_after|, using DeletionCause
443  // |cause|.  If |keep_accessed_after| is null, deletes all cookies in the
444  // list.  Returns the number of cookies deleted.
445  int GarbageCollectDeleteList(const base::Time& current,
446                               const base::Time& keep_accessed_after,
447                               DeletionCause cause,
448                               std::vector<CookieMap::iterator>& cookie_its);
449
450  // Find the key (for lookup in cookies_) based on the given domain.
451  // See comment on keys before the CookieMap typedef.
452  std::string GetKey(const std::string& domain) const;
453
454  bool HasCookieableScheme(const GURL& url);
455
456  // Statistics support
457
458  // This function should be called repeatedly, and will record
459  // statistics if a sufficient time period has passed.
460  void RecordPeriodicStats(const base::Time& current_time);
461
462  // Initialize the above variables; should only be called from
463  // the constructor.
464  void InitializeHistograms();
465
466  // The resolution of our time isn't enough, so we do something
467  // ugly and increment when we've seen the same time twice.
468  base::Time CurrentTime();
469
470  // Histogram variables; see CookieMonster::InitializeHistograms() in
471  // cookie_monster.cc for details.
472  base::Histogram* histogram_expiration_duration_minutes_;
473  base::Histogram* histogram_between_access_interval_minutes_;
474  base::Histogram* histogram_evicted_last_access_minutes_;
475  base::Histogram* histogram_count_;
476  base::Histogram* histogram_domain_count_;
477  base::Histogram* histogram_etldp1_count_;
478  base::Histogram* histogram_domain_per_etldp1_count_;
479  base::Histogram* histogram_number_duplicate_db_cookies_;
480  base::Histogram* histogram_cookie_deletion_cause_;
481  base::Histogram* histogram_time_get_;
482  base::Histogram* histogram_time_load_;
483
484  CookieMap cookies_;
485
486  // Indicates whether the cookie store has been initialized. This happens
487  // lazily in InitStoreIfNecessary().
488  bool initialized_;
489
490  // Indicates whether this cookie monster uses the new effective domain
491  // key scheme or not.
492  ExpiryAndKeyScheme expiry_and_key_scheme_;
493
494  scoped_refptr<PersistentCookieStore> store_;
495
496  base::Time last_time_seen_;
497
498  // Minimum delay after updating a cookie's LastAccessDate before we will
499  // update it again.
500  const base::TimeDelta last_access_threshold_;
501
502  // Approximate date of access time of least recently accessed cookie
503  // in |cookies_|.  Note that this is not guaranteed to be accurate, only a)
504  // to be before or equal to the actual time, and b) to be accurate
505  // immediately after a garbage collection that scans through all the cookies.
506  // This value is used to determine whether global garbage collection might
507  // find cookies to purge.
508  // Note: The default Time() constructor will create a value that compares
509  // earlier than any other time value, which is is wanted.  Thus this
510  // value is not initialized.
511  base::Time earliest_access_time_;
512
513  std::vector<std::string> cookieable_schemes_;
514
515  scoped_refptr<Delegate> delegate_;
516
517  // Lock for thread-safety
518  base::Lock lock_;
519
520  base::Time last_statistic_record_time_;
521
522  bool keep_expired_cookies_;
523
524  static bool enable_file_scheme_;
525
526  DISALLOW_COPY_AND_ASSIGN(CookieMonster);
527};
528
529class NET_EXPORT CookieMonster::CanonicalCookie {
530 public:
531
532  // These constructors do no validation or canonicalization of their inputs;
533  // the resulting CanonicalCookies should not be relied on to be canonical
534  // unless the caller has done appropriate validation and canonicalization
535  // themselves.
536  CanonicalCookie();
537  CanonicalCookie(const GURL& url,
538                  const std::string& name,
539                  const std::string& value,
540                  const std::string& domain,
541                  const std::string& path,
542                  const base::Time& creation,
543                  const base::Time& expiration,
544                  const base::Time& last_access,
545                  bool secure,
546                  bool httponly,
547                  bool has_expires);
548
549  // This constructor does canonicalization but not validation.
550  // The result of this constructor should not be relied on in contexts
551  // in which pre-validation of the ParsedCookie has not been done.
552  CanonicalCookie(const GURL& url, const ParsedCookie& pc);
553
554  ~CanonicalCookie();
555
556  // Supports the default copy constructor.
557
558  // Creates a canonical cookie from unparsed attribute values.
559  // Canonicalizes and validates inputs.  May return NULL if an attribute
560  // value is invalid.
561  static CanonicalCookie* Create(const GURL& url,
562                                 const std::string& name,
563                                 const std::string& value,
564                                 const std::string& domain,
565                                 const std::string& path,
566                                 const base::Time& creation,
567                                 const base::Time& expiration,
568                                 bool secure,
569                                 bool http_only);
570
571  const std::string& Source() const { return source_; }
572  const std::string& Name() const { return name_; }
573  const std::string& Value() const { return value_; }
574  const std::string& Domain() const { return domain_; }
575  const std::string& Path() const { return path_; }
576  const base::Time& CreationDate() const { return creation_date_; }
577  const base::Time& LastAccessDate() const { return last_access_date_; }
578  bool DoesExpire() const { return has_expires_; }
579#if defined(ANDROID)
580  // Android can shut down our app at any time, so we persist session cookies.
581  bool IsPersistent() const { return true; }
582  bool IsSessionCookie() const { return !DoesExpire(); }
583#else
584  bool IsPersistent() const { return DoesExpire(); }
585#endif
586  const base::Time& ExpiryDate() const { return expiry_date_; }
587  bool IsSecure() const { return secure_; }
588  bool IsHttpOnly() const { return httponly_; }
589  bool IsDomainCookie() const {
590    return !domain_.empty() && domain_[0] == '.'; }
591  bool IsHostCookie() const { return !IsDomainCookie(); }
592
593  bool IsExpired(const base::Time& current) {
594    return has_expires_ && current >= expiry_date_;
595  }
596
597  // Are the cookies considered equivalent in the eyes of RFC 2965.
598  // The RFC says that name must match (case-sensitive), domain must
599  // match (case insensitive), and path must match (case sensitive).
600  // For the case insensitive domain compare, we rely on the domain
601  // having been canonicalized (in
602  // GetCookieDomainWithString->CanonicalizeHost).
603  bool IsEquivalent(const CanonicalCookie& ecc) const {
604    // It seems like it would make sense to take secure and httponly into
605    // account, but the RFC doesn't specify this.
606    // NOTE: Keep this logic in-sync with TrimDuplicateCookiesForHost().
607    return (name_ == ecc.Name() && domain_ == ecc.Domain()
608            && path_ == ecc.Path());
609  }
610
611  void SetLastAccessDate(const base::Time& date) {
612    last_access_date_ = date;
613  }
614
615  bool IsOnPath(const std::string& url_path) const;
616  bool IsDomainMatch(const std::string& scheme, const std::string& host) const;
617
618  std::string DebugString() const;
619
620  // Returns the cookie source when cookies are set for |url|.  This function
621  // is public for unit test purposes only.
622  static std::string GetCookieSourceFromURL(const GURL& url);
623
624 private:
625  // The source member of a canonical cookie is the origin of the URL that tried
626  // to set this cookie, minus the port number if any.  This field is not
627  // persistent though; its only used in the in-tab cookies dialog to show the
628  // user the source URL. This is used for both allowed and blocked cookies.
629  // When a CanonicalCookie is constructed from the backing store (common case)
630  // this field will be null.  CanonicalCookie consumers should not rely on
631  // this field unless they guarantee that the creator of those
632  // CanonicalCookies properly initialized the field.
633  std::string source_;
634  std::string name_;
635  std::string value_;
636  std::string domain_;
637  std::string path_;
638  base::Time creation_date_;
639  base::Time expiry_date_;
640  base::Time last_access_date_;
641  bool secure_;
642  bool httponly_;
643  bool has_expires_;
644};
645
646class NET_EXPORT CookieMonster::Delegate
647    : public base::RefCountedThreadSafe<CookieMonster::Delegate> {
648 public:
649  // The publicly relevant reasons a cookie might be changed.
650  enum ChangeCause {
651    // The cookie was changed directly by a consumer's action.
652    CHANGE_COOKIE_EXPLICIT,
653    // The cookie was automatically removed due to an insert operation that
654    // overwrote it.
655    CHANGE_COOKIE_OVERWRITE,
656    // The cookie was automatically removed as it expired.
657    CHANGE_COOKIE_EXPIRED,
658    // The cookie was automatically evicted during garbage collection.
659    CHANGE_COOKIE_EVICTED,
660    // The cookie was overwritten with an already-expired expiration date.
661    CHANGE_COOKIE_EXPIRED_OVERWRITE
662  };
663
664  // Will be called when a cookie is added or removed. The function is passed
665  // the respective |cookie| which was added to or removed from the cookies.
666  // If |removed| is true, the cookie was deleted, and |cause| will be set
667  // to the reason for it's removal. If |removed| is false, the cookie was
668  // added, and |cause| will be set to CHANGE_COOKIE_EXPLICIT.
669  //
670  // As a special case, note that updating a cookie's properties is implemented
671  // as a two step process: the cookie to be updated is first removed entirely,
672  // generating a notification with cause CHANGE_COOKIE_OVERWRITE.  Afterwards,
673  // a new cookie is written with the updated values, generating a notification
674  // with cause CHANGE_COOKIE_EXPLICIT.
675  virtual void OnCookieChanged(const CookieMonster::CanonicalCookie& cookie,
676                               bool removed,
677                               ChangeCause cause) = 0;
678 protected:
679  friend class base::RefCountedThreadSafe<CookieMonster::Delegate>;
680  virtual ~Delegate() {}
681};
682
683class NET_EXPORT CookieMonster::ParsedCookie {
684 public:
685  typedef std::pair<std::string, std::string> TokenValuePair;
686  typedef std::vector<TokenValuePair> PairList;
687
688  // The maximum length of a cookie string we will try to parse
689  static const size_t kMaxCookieSize = 4096;
690  // The maximum number of Token/Value pairs.  Shouldn't have more than 8.
691  static const int kMaxPairs = 16;
692
693  // Construct from a cookie string like "BLAH=1; path=/; domain=.google.com"
694  ParsedCookie(const std::string& cookie_line);
695  ~ParsedCookie();
696
697  // You should not call any other methods on the class if !IsValid
698  bool IsValid() const { return is_valid_; }
699
700  const std::string& Name() const { return pairs_[0].first; }
701  const std::string& Token() const { return Name(); }
702  const std::string& Value() const { return pairs_[0].second; }
703
704  bool HasPath() const { return path_index_ != 0; }
705  const std::string& Path() const { return pairs_[path_index_].second; }
706  bool HasDomain() const { return domain_index_ != 0; }
707  const std::string& Domain() const { return pairs_[domain_index_].second; }
708  bool HasExpires() const { return expires_index_ != 0; }
709  const std::string& Expires() const { return pairs_[expires_index_].second; }
710  bool HasMaxAge() const { return maxage_index_ != 0; }
711  const std::string& MaxAge() const { return pairs_[maxage_index_].second; }
712  bool IsSecure() const { return secure_index_ != 0; }
713  bool IsHttpOnly() const { return httponly_index_ != 0; }
714
715  // Returns the number of attributes, for example, returning 2 for:
716  //   "BLAH=hah; path=/; domain=.google.com"
717  size_t NumberOfAttributes() const { return pairs_.size() - 1; }
718
719  // For debugging only!
720  std::string DebugString() const;
721
722  // Returns an iterator pointing to the first terminator character found in
723  // the given string.
724  static std::string::const_iterator FindFirstTerminator(const std::string& s);
725
726  // Given iterators pointing to the beginning and end of a string segment,
727  // returns as output arguments token_start and token_end to the start and end
728  // positions of a cookie attribute token name parsed from the segment, and
729  // updates the segment iterator to point to the next segment to be parsed.
730  // If no token is found, the function returns false.
731  static bool ParseToken(std::string::const_iterator* it,
732                         const std::string::const_iterator& end,
733                         std::string::const_iterator* token_start,
734                         std::string::const_iterator* token_end);
735
736  // Given iterators pointing to the beginning and end of a string segment,
737  // returns as output arguments value_start and value_end to the start and end
738  // positions of a cookie attribute value parsed from the segment, and updates
739  // the segment iterator to point to the next segment to be parsed.
740  static void ParseValue(std::string::const_iterator* it,
741                         const std::string::const_iterator& end,
742                         std::string::const_iterator* value_start,
743                         std::string::const_iterator* value_end);
744
745  // Same as the above functions, except the input is assumed to contain the
746  // desired token/value and nothing else.
747  static std::string ParseTokenString(const std::string& token);
748  static std::string ParseValueString(const std::string& value);
749
750 private:
751  static const char kTerminator[];
752  static const int  kTerminatorLen;
753  static const char kWhitespace[];
754  static const char kValueSeparator[];
755  static const char kTokenSeparator[];
756
757  void ParseTokenValuePairs(const std::string& cookie_line);
758  void SetupAttributes();
759
760  PairList pairs_;
761  bool is_valid_;
762  // These will default to 0, but that should never be valid since the
763  // 0th index is the user supplied token/value, not an attribute.
764  // We're really never going to have more than like 8 attributes, so we
765  // could fit these into 3 bits each if we're worried about size...
766  size_t path_index_;
767  size_t domain_index_;
768  size_t expires_index_;
769  size_t maxage_index_;
770  size_t secure_index_;
771  size_t httponly_index_;
772
773  DISALLOW_COPY_AND_ASSIGN(ParsedCookie);
774};
775
776typedef base::RefCountedThreadSafe<CookieMonster::PersistentCookieStore>
777    RefcountedPersistentCookieStore;
778
779class CookieMonster::PersistentCookieStore
780    : public RefcountedPersistentCookieStore {
781 public:
782  virtual ~PersistentCookieStore() {}
783
784  // Initializes the store and retrieves the existing cookies. This will be
785  // called only once at startup.
786  virtual bool Load(std::vector<CookieMonster::CanonicalCookie*>* cookies) = 0;
787
788  virtual void AddCookie(const CanonicalCookie& cc) = 0;
789  virtual void UpdateCookieAccessTime(const CanonicalCookie& cc) = 0;
790  virtual void DeleteCookie(const CanonicalCookie& cc) = 0;
791
792  // Sets the value of the user preference whether the persistent storage
793  // must be deleted upon destruction.
794  virtual void SetClearLocalStateOnExit(bool clear_local_state) = 0;
795
796  // Flush the store and post the given Task when complete.
797  virtual void Flush(Task* completion_task) = 0;
798
799 protected:
800  PersistentCookieStore() {}
801
802 private:
803  DISALLOW_COPY_AND_ASSIGN(PersistentCookieStore);
804};
805
806class NET_EXPORT CookieList : public std::vector<CookieMonster::CanonicalCookie> {
807};
808
809}  // namespace net
810
811#endif  // NET_BASE_COOKIE_MONSTER_H_
812