• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 // A Predictor object is instantiated once in the browser process, and manages
6 // both preresolution of hostnames, as well as TCP/IP preconnection to expected
7 // subresources.
8 // Most hostname lists are provided by the renderer processes, and include URLs
9 // that *might* be used in the near future by the browsing user.  One goal of
10 // this class is to cause the underlying DNS structure to lookup a hostname
11 // before it is really needed, and hence reduce latency in the standard lookup
12 // paths.
13 // Subresource relationships are usually acquired from the referrer field in a
14 // navigation.  A subresource URL may be associated with a referrer URL.  Later
15 // navigations may, if the likelihood of needing the subresource is high enough,
16 // cause this module to speculatively create a TCP/IP connection. If there is
17 // only a low likelihood, then a DNS pre-resolution operation may be performed.
18 
19 #ifndef CHROME_BROWSER_NET_PREDICTOR_H_
20 #define CHROME_BROWSER_NET_PREDICTOR_H_
21 #pragma once
22 
23 #include <map>
24 #include <queue>
25 #include <set>
26 #include <string>
27 #include <vector>
28 
29 #include "base/gtest_prod_util.h"
30 #include "base/memory/ref_counted.h"
31 #include "chrome/browser/net/url_info.h"
32 #include "chrome/browser/net/referrer.h"
33 #include "chrome/common/net/predictor_common.h"
34 #include "net/base/host_port_pair.h"
35 
36 class ListValue;
37 
38 namespace net {
39 class HostResolver;
40 }  // namespace net
41 
42 namespace chrome_browser_net {
43 
44 typedef chrome_common_net::UrlList UrlList;
45 typedef chrome_common_net::NameList NameList;
46 typedef std::map<GURL, UrlInfo> Results;
47 
48 // Note that Predictor is not thread safe, and must only be called from
49 // the IO thread. Failure to do so will result in a DCHECK at runtime.
50 class Predictor : public base::RefCountedThreadSafe<Predictor> {
51  public:
52   // A version number for prefs that are saved. This should be incremented when
53   // we change the format so that we discard old data.
54   enum { PREDICTOR_REFERRER_VERSION = 2 };
55 
56   // |max_concurrent| specifies how many concurrent (parallel) prefetches will
57   // be performed. Host lookups will be issued through |host_resolver|.
58   Predictor(net::HostResolver* host_resolver,
59             base::TimeDelta max_queue_delay_ms, size_t max_concurrent,
60             bool preconnect_enabled);
61 
62   // Cancel pending requests and prevent new ones from being made.
63   void Shutdown();
64 
65   // In some circumstances, for privacy reasons, all results should be
66   // discarded.  This method gracefully handles that activity.
67   // Destroy all our internal state, which shows what names we've looked up, and
68   // how long each has taken, etc. etc.  We also destroy records of suggesses
69   // (cache hits etc.).
70   void DiscardAllResults();
71 
72   // Add hostname(s) to the queue for processing.
73   void ResolveList(const UrlList& urls,
74                    UrlInfo::ResolutionMotivation motivation);
75   void Resolve(const GURL& url,
76                UrlInfo::ResolutionMotivation motivation);
77 
78   // Instigate pre-connection to any URLs, or pre-resolution of related host,
79   // that we predict will be needed after this navigation (typically
80   // more-embedded resources on a page).  This method will actually post a task
81   // to do the actual work, so as not to jump ahead of the frame navigation that
82   // instigated this activity.
83   void PredictFrameSubresources(const GURL& url);
84 
85   // The Omnibox has proposed a given url to the user, and if it is a search
86   // URL, then it also indicates that this is preconnectable (i.e., we could
87   // preconnect to the search server).
88   void AnticipateOmniboxUrl(const GURL& url, bool preconnectable);
89 
90   // Preconnect a URL and all of its subresource domains.
91   void PreconnectUrlAndSubresources(const GURL& url);
92 
93   // Record details of a navigation so that we can preresolve the host name
94   // ahead of time the next time the users navigates to the indicated host.
95   // Should only be called when urls are distinct, and they should already be
96   // canonicalized to not have a path.
97   void LearnFromNavigation(const GURL& referring_url, const GURL& target_url);
98 
99   // Dump HTML table containing list of referrers for about:dns.
100   void GetHtmlReferrerLists(std::string* output);
101 
102   // Dump the list of currently known referrer domains and related prefetchable
103   // domains.
104   void GetHtmlInfo(std::string* output);
105 
106   // Discards any referrer for which all the suggested host names are currently
107   // annotated with negligible expected-use.  Scales down (diminishes) the
108   // expected-use of those that remain, so that their use will go down by a
109   // factor each time we trim (moving the referrer closer to being discarded in
110   // a future call).
111   // The task is performed synchronously and completes before returing.
112   void TrimReferrersNow();
113 
114   // Construct a ListValue object that contains all the data in the referrers_
115   // so that it can be persisted in a pref.
116   void SerializeReferrers(ListValue* referral_list);
117 
118   // Process a ListValue that contains all the data from a previous reference
119   // list, as constructed by SerializeReferrers(), and add all the identified
120   // values into the current referrer list.
121   void DeserializeReferrers(const ListValue& referral_list);
122 
123   void DeserializeReferrersThenDelete(ListValue* referral_list);
124 
125   // For unit test code only.
max_concurrent_dns_lookups()126   size_t max_concurrent_dns_lookups() const {
127     return max_concurrent_dns_lookups_;
128   }
129 
130   // Flag setting to use preconnection instead of just DNS pre-fetching.
preconnect_enabled()131   bool preconnect_enabled() const { return preconnect_enabled_; }
132 
133   // Put URL in canonical form, including a scheme, host, and port.
134   // Returns GURL::EmptyGURL() if the scheme is not http/https or if the url
135   // cannot be otherwise canonicalized.
136   static GURL CanonicalizeUrl(const GURL& url);
137 
138  private:
139   friend class base::RefCountedThreadSafe<Predictor>;
140   FRIEND_TEST_ALL_PREFIXES(PredictorTest, BenefitLookupTest);
141   FRIEND_TEST_ALL_PREFIXES(PredictorTest, ShutdownWhenResolutionIsPendingTest);
142   FRIEND_TEST_ALL_PREFIXES(PredictorTest, SingleLookupTest);
143   FRIEND_TEST_ALL_PREFIXES(PredictorTest, ConcurrentLookupTest);
144   FRIEND_TEST_ALL_PREFIXES(PredictorTest, MassiveConcurrentLookupTest);
145   FRIEND_TEST_ALL_PREFIXES(PredictorTest, PriorityQueuePushPopTest);
146   FRIEND_TEST_ALL_PREFIXES(PredictorTest, PriorityQueueReorderTest);
147   FRIEND_TEST_ALL_PREFIXES(PredictorTest, ReferrerSerializationTrimTest);
148   friend class WaitForResolutionHelper;  // For testing.
149 
150   class LookupRequest;
151 
152   // A simple priority queue for handling host names.
153   // Some names that are queued up have |motivation| that requires very rapid
154   // handling.  For example, a sub-resource name lookup MUST be done before the
155   // actual sub-resource is fetched.  In contrast, a name that was speculatively
156   // noted in a page has to be resolved before the user "gets around to"
157   // clicking on a link.  By tagging (with a motivation) each push we make into
158   // this FIFO queue, the queue can re-order the more important names to service
159   // them sooner (relative to some low priority background resolutions).
160   class HostNameQueue {
161    public:
162     HostNameQueue();
163     ~HostNameQueue();
164     void Push(const GURL& url,
165               UrlInfo::ResolutionMotivation motivation);
166     bool IsEmpty() const;
167     GURL Pop();
168 
169    private:
170     // The names in the queue that should be serviced (popped) ASAP.
171     std::queue<GURL> rush_queue_;
172     // The names in the queue that should only be serviced when rush_queue is
173     // empty.
174     std::queue<GURL> background_queue_;
175 
176   DISALLOW_COPY_AND_ASSIGN(HostNameQueue);
177   };
178 
179   // A map that is keyed with the host/port that we've learned were the cause
180   // of loading additional URLs.  The list of additional targets is held
181   // in a Referrer instance, which is a value in this map.
182   typedef std::map<GURL, Referrer> Referrers;
183 
184   // Depending on the expected_subresource_use_, we may either make a TCP/IP
185   // preconnection, or merely pre-resolve the hostname via DNS (or even do
186   // nothing).  The following are the threasholds for taking those actions.
187   static const double kPreconnectWorthyExpectedValue;
188   static const double kDNSPreresolutionWorthyExpectedValue;
189   // Referred hosts with a subresource_use_rate_ that are less than the
190   // following threshold will be discarded when we Trim() the list.
191   static const double kDiscardableExpectedValue;
192   // During trimming operation to discard hosts for which we don't have likely
193   // subresources, we multiply the expected_subresource_use_ value by the
194   // following ratio until that value is less than kDiscardableExpectedValue.
195   // This number should always be less than 1, an more than 0.
196   static const double kReferrerTrimRatio;
197 
198   // Interval between periodic trimming of our whole referrer list.
199   // We only do a major trimming about once an hour, and then only when the user
200   // is actively browsing.
201   static const base::TimeDelta kDurationBetweenTrimmings;
202   // Interval between incremental trimmings (to avoid inducing Jank).
203   static const base::TimeDelta kDurationBetweenTrimmingIncrements;
204   // Number of referring URLs processed in an incremental trimming.
205   static const size_t kUrlsTrimmedPerIncrement;
206 
207   ~Predictor();
208 
209   // Perform actual resolution or preconnection to subresources now.  This is
210   // an internal worker method that is reached via a post task from
211   // PredictFrameSubresources().
212   void PrepareFrameSubresources(const GURL& url);
213 
214   // Only for testing. Returns true if hostname has been successfully resolved
215   // (name found).
WasFound(const GURL & url)216   bool WasFound(const GURL& url) const {
217     Results::const_iterator it(results_.find(url));
218     return (it != results_.end()) &&
219             it->second.was_found();
220   }
221 
222   // Only for testing. Return how long was the resolution
223   // or UrlInfo::kNullDuration if it hasn't been resolved yet.
GetResolutionDuration(const GURL & url)224   base::TimeDelta GetResolutionDuration(const GURL& url) {
225     if (results_.find(url) == results_.end())
226       return UrlInfo::kNullDuration;
227     return results_[url].resolve_duration();
228   }
229 
230   // Only for testing;
peak_pending_lookups()231   size_t peak_pending_lookups() const { return peak_pending_lookups_; }
232 
233   // Access method for use by async lookup request to pass resolution result.
234   void OnLookupFinished(LookupRequest* request, const GURL& url, bool found);
235 
236   // Underlying method for both async and synchronous lookup to update state.
237   void LookupFinished(LookupRequest* request,
238                       const GURL& url, bool found);
239 
240   // Queue hostname for resolution.  If queueing was done, return the pointer
241   // to the queued instance, otherwise return NULL.
242   UrlInfo* AppendToResolutionQueue(const GURL& url,
243       UrlInfo::ResolutionMotivation motivation);
244 
245   // Check to see if too much queuing delay has been noted for the given info,
246   // which indicates that there is "congestion" or growing delay in handling the
247   // resolution of names.  Rather than letting this congestion potentially grow
248   // without bounds, we abandon our queued efforts at pre-resolutions in such a
249   // case.
250   // To do this, we will recycle |info|, as well as all queued items, back to
251   // the state they had before they were queued up.  We can't do anything about
252   // the resolutions we've already sent off for processing on another thread, so
253   // we just let them complete.  On a slow system, subject to congestion, this
254   // will greatly reduce the number of resolutions done, but it will assure that
255   // any resolutions that are done, are in a timely and hence potentially
256   // helpful manner.
257   bool CongestionControlPerformed(UrlInfo* info);
258 
259   // Take lookup requests from work_queue_ and tell HostResolver to look them up
260   // asynchronously, provided we don't exceed concurrent resolution limit.
261   void StartSomeQueuedResolutions();
262 
263   // Performs trimming similar to TrimReferrersNow(), except it does it as a
264   // series of short tasks by posting continuations again an again until done.
265   void TrimReferrers();
266 
267   // Loads urls_being_trimmed_ from keys of current referrers_.
268   void LoadUrlsForTrimming();
269 
270   // Posts a task to do additional incremental trimming of referrers_.
271   void PostIncrementalTrimTask();
272 
273   // Calls Trim() on some or all of urls_being_trimmed_.
274   // If it does not process all the URLs in that vector, it posts a task to
275   // continue with them shortly (i.e., it yeilds and continues).
276   void IncrementalTrimReferrers(bool trim_all_now);
277 
278   // work_queue_ holds a list of names we need to look up.
279   HostNameQueue work_queue_;
280 
281   // results_ contains information for existing/prior prefetches.
282   Results results_;
283 
284   std::set<LookupRequest*> pending_lookups_;
285 
286   // For testing, to verify that we don't exceed the limit.
287   size_t peak_pending_lookups_;
288 
289   // When true, we don't make new lookup requests.
290   bool shutdown_;
291 
292   // The number of concurrent speculative lookups currently allowed to be sent
293   // to the resolver.  Any additional lookups will be queued to avoid exceeding
294   // this value.  The queue is a priority queue that will accelerate
295   // sub-resource speculation, and retard resolutions suggested by page scans.
296   const size_t max_concurrent_dns_lookups_;
297 
298   // The maximum queueing delay that is acceptable before we enter congestion
299   // reduction mode, and discard all queued (but not yet assigned) resolutions.
300   const base::TimeDelta max_dns_queue_delay_;
301 
302   // The host resolver we warm DNS entries for.
303   net::HostResolver* const host_resolver_;
304 
305   // Are we currently using preconnection, rather than just DNS resolution, for
306   // subresources and omni-box search URLs.
307   bool preconnect_enabled_;
308 
309   // Most recent suggestion from Omnibox provided via AnticipateOmniboxUrl().
310   std::string last_omnibox_host_;
311 
312   // The time when the last preresolve was done for last_omnibox_host_.
313   base::TimeTicks last_omnibox_preresolve_;
314 
315   // The number of consecutive requests to AnticipateOmniboxUrl() that suggested
316   // preconnecting (because it was to a search service).
317   int consecutive_omnibox_preconnect_count_;
318 
319   // The time when the last preconnection was requested to a search service.
320   base::TimeTicks last_omnibox_preconnect_;
321 
322   // For each URL that we might navigate to (that we've "learned about")
323   // we have a Referrer list. Each Referrer list has all hostnames we might
324   // need to pre-resolve or pre-connect to when there is a navigation to the
325   // orginial hostname.
326   Referrers referrers_;
327 
328   // List of URLs in referrers_ currently being trimmed (scaled down to
329   // eventually be aged out of use).
330   std::vector<GURL> urls_being_trimmed_;
331 
332   // A time after which we need to do more trimming of referrers.
333   base::TimeTicks next_trim_time_;
334 
335   ScopedRunnableMethodFactory<Predictor> trim_task_factory_;
336 
337   DISALLOW_COPY_AND_ASSIGN(Predictor);
338 };
339 
340 }  // namespace chrome_browser_net
341 
342 #endif  // CHROME_BROWSER_NET_PREDICTOR_H_
343