OLD | NEW |
---|---|
1 // Copyright 2017 The Chromium Authors. All rights reserved. | 1 // Copyright 2017 The Chromium Authors. All rights reserved. |
2 // Use of this source code is governed by a BSD-style license that can be | 2 // Use of this source code is governed by a BSD-style license that can be |
3 // found in the LICENSE file. | 3 // found in the LICENSE file. |
4 | 4 |
5 #include "chrome/browser/page_load_metrics/observers/ads_page_load_metrics_obser ver.h" | 5 #include "chrome/browser/page_load_metrics/observers/ads_page_load_metrics_obser ver.h" |
6 | 6 |
7 #include <string> | 7 #include <string> |
8 #include <utility> | 8 #include <utility> |
9 | 9 |
10 #include "base/feature_list.h" | 10 #include "base/feature_list.h" |
11 #include "base/logging.h" | 11 #include "base/logging.h" |
12 #include "base/memory/ptr_util.h" | 12 #include "base/memory/ptr_util.h" |
13 #include "base/strings/string_util.h" | 13 #include "base/strings/string_util.h" |
14 #include "chrome/browser/page_load_metrics/page_load_metrics_util.h" | 14 #include "chrome/browser/page_load_metrics/page_load_metrics_util.h" |
15 #include "content/public/browser/navigation_handle.h" | 15 #include "content/public/browser/navigation_handle.h" |
16 #include "content/public/browser/render_frame_host.h" | 16 #include "content/public/browser/render_frame_host.h" |
17 #include "content/public/browser/web_contents.h" | 17 #include "content/public/browser/web_contents.h" |
18 #include "url/gurl.h" | 18 #include "url/gurl.h" |
19 | 19 |
20 namespace { | 20 namespace { |
21 | 21 |
22 const base::Feature kAdsFeature{"AdsMetrics", | 22 const base::Feature kAdsFeature{"AdsMetrics", |
23 base::FEATURE_DISABLED_BY_DEFAULT}; | 23 base::FEATURE_DISABLED_BY_DEFAULT}; |
24 | 24 |
25 bool FrameIsAd(content::NavigationHandle* navigation_handle) { | 25 bool FrameIsAd(content::NavigationHandle* navigation_handle) { |
26 content::RenderFrameHost* current_frame_host = | |
27 navigation_handle->GetRenderFrameHost(); | |
28 DCHECK(current_frame_host); | |
29 const std::string& name = current_frame_host->GetFrameName(); | |
30 const GURL& url = navigation_handle->GetURL(); | |
31 | |
32 // Because sub-resource filtering isn't always enabled, and doesn't work | 26 // Because sub-resource filtering isn't always enabled, and doesn't work |
33 // well in monitoring mode (no CSS enforcement), it's difficult to identify | 27 // well in monitoring mode (no CSS enforcement), it's difficult to identify |
34 // ads. Google ads are prevalent and easy to track, so we'll start by | 28 // ads. Google ads are prevalent and easy to track, so we'll start by |
35 // tracking those. Note that the frame name can be very large, so be careful | 29 // tracking those. Note that the frame name can be very large, so be careful |
36 // to avoid full string searches if possible. | 30 // to avoid full string searches if possible. |
37 // TODO(jkarlin): Track other ad networks that are easy to identify. | 31 // TODO(jkarlin): Track other ad networks that are easy to identify. |
38 return base::StartsWith(name, "google_ads_iframe", | 32 |
39 base::CompareCase::SENSITIVE) || | 33 // In case the navigation aborted, look up the RFH by the Frame Tree Node |
40 base::StartsWith(name, "google_ads_frame", | 34 // ID. If there is no RFH remaining, then the frame is gone. We use the |
41 base::CompareCase::SENSITIVE) || | 35 // unsafe method of FindFrameByFrameTreeNodeId because we're not concerned |
42 (url.host_piece() == "tpc.googlesyndication.com" && | 36 // with which process the frame lives on (we're just measuring bytes and not |
43 base::StartsWith(url.path_piece(), "/safeframe", | 37 // granting security priveleges). |
44 base::CompareCase::SENSITIVE)); | 38 content::RenderFrameHost* current_frame_host = |
39 navigation_handle->GetWebContents()->UnsafeFindFrameByFrameTreeNodeId( | |
Bryan McQuade
2017/05/12 12:43:59
does this only return frames for committed loads?
jkarlin
2017/05/12 14:31:42
Good question. It returns the committed RFH for th
| |
40 navigation_handle->GetFrameTreeNodeId()); | |
41 if (current_frame_host) { | |
42 const std::string& frame_name = current_frame_host->GetFrameName(); | |
43 if (base::StartsWith(frame_name, "google_ads_iframe", | |
44 base::CompareCase::SENSITIVE) || | |
45 base::StartsWith(frame_name, "google_ads_frame", | |
46 base::CompareCase::SENSITIVE)) { | |
47 return true; | |
48 } | |
49 } | |
50 | |
51 const GURL& url = navigation_handle->GetURL(); | |
52 return url.host_piece() == "tpc.googlesyndication.com" && | |
53 base::StartsWith(url.path_piece(), "/safeframe", | |
54 base::CompareCase::SENSITIVE); | |
45 } | 55 } |
46 | 56 |
47 } // namespace | 57 } // namespace |
48 | 58 |
49 AdsPageLoadMetricsObserver::AdFrameData::AdFrameData( | 59 AdsPageLoadMetricsObserver::AdFrameData::AdFrameData( |
50 FrameTreeNodeId frame_tree_node_id) | 60 FrameTreeNodeId frame_tree_node_id) |
51 : frame_bytes(0u), | 61 : frame_bytes(0u), |
52 frame_bytes_uncached(0u), | 62 frame_bytes_uncached(0u), |
53 frame_tree_node_id(frame_tree_node_id) {} | 63 frame_tree_node_id(frame_tree_node_id) {} |
54 | 64 |
(...skipping 15 matching lines...) Expand all Loading... | |
70 | 80 |
71 // The main frame is never considered an ad. | 81 // The main frame is never considered an ad. |
72 ad_frames_data_[navigation_handle->GetFrameTreeNodeId()] = nullptr; | 82 ad_frames_data_[navigation_handle->GetFrameTreeNodeId()] = nullptr; |
73 ProcessOngoingNavigationResource(navigation_handle->GetFrameTreeNodeId()); | 83 ProcessOngoingNavigationResource(navigation_handle->GetFrameTreeNodeId()); |
74 return CONTINUE_OBSERVING; | 84 return CONTINUE_OBSERVING; |
75 } | 85 } |
76 | 86 |
77 page_load_metrics::PageLoadMetricsObserver::ObservePolicy | 87 page_load_metrics::PageLoadMetricsObserver::ObservePolicy |
78 AdsPageLoadMetricsObserver::OnDidFinishSubFrameNavigation( | 88 AdsPageLoadMetricsObserver::OnDidFinishSubFrameNavigation( |
79 content::NavigationHandle* navigation_handle) { | 89 content::NavigationHandle* navigation_handle) { |
90 // Determine if the frame is part of an existing ad, the root of a new ad, | |
91 // or a non-ad frame. Once a frame is labled as an ad, it is always | |
92 // considered an ad, even if it navigates to a non-ad page. This function | |
93 // labels all of a page's frames, even those that fail to commit. | |
80 FrameTreeNodeId frame_tree_node_id = navigation_handle->GetFrameTreeNodeId(); | 94 FrameTreeNodeId frame_tree_node_id = navigation_handle->GetFrameTreeNodeId(); |
81 | |
82 if (!navigation_handle->HasCommitted() || | |
83 navigation_handle->IsSameDocument() || navigation_handle->IsErrorPage()) { | |
84 // We're not interested in tracking this navigation. In case we've seen a | |
85 // resource for the navigation before this message, clear it from | |
86 // ongoing_navigation_resources_. | |
87 ongoing_navigation_resources_.erase(frame_tree_node_id); | |
88 return CONTINUE_OBSERVING; | |
89 } | |
90 | |
91 content::RenderFrameHost* parent_frame_host = | 95 content::RenderFrameHost* parent_frame_host = |
92 navigation_handle->GetRenderFrameHost()->GetParent(); | 96 navigation_handle->GetParentFrame(); |
93 DCHECK(parent_frame_host); | |
94 | |
95 bool top_level_subframe = !parent_frame_host->GetParent(); | |
96 | 97 |
97 const auto& id_and_data = ad_frames_data_.find(frame_tree_node_id); | 98 const auto& id_and_data = ad_frames_data_.find(frame_tree_node_id); |
98 if (id_and_data != ad_frames_data_.end()) { | 99 if (id_and_data != ad_frames_data_.end()) { |
99 // An existing subframe is navigating again. | 100 // An existing subframe is navigating again. |
100 if (id_and_data->second) { | 101 if (id_and_data->second) { |
101 // The subframe was an ad to begin with, keep tracking it as an ad. | 102 // The subframe was an ad to begin with, keep tracking it as an ad. |
102 ProcessOngoingNavigationResource(frame_tree_node_id); | 103 ProcessOngoingNavigationResource(frame_tree_node_id); |
103 | 104 |
104 if (frame_tree_node_id == id_and_data->second->frame_tree_node_id) { | 105 if (frame_tree_node_id == id_and_data->second->frame_tree_node_id) { |
105 // This is the top-most frame in the ad. | 106 // This is the top-most frame in the ad. |
106 UMA_HISTOGRAM_BOOLEAN( | 107 UMA_HISTOGRAM_BOOLEAN( |
107 "PageLoad.Clients.Ads.Google.Navigations.AdFrameRenavigatedToAd", | 108 "PageLoad.Clients.Ads.Google.Navigations.AdFrameRenavigatedToAd", |
108 FrameIsAd(navigation_handle)); | 109 FrameIsAd(navigation_handle)); |
109 } | 110 } |
110 | |
111 return CONTINUE_OBSERVING; | 111 return CONTINUE_OBSERVING; |
112 } | 112 } |
113 // This frame was previously not an ad, process it as usual. If it had | 113 // This frame was previously not an ad, process it as usual. If it had |
114 // any child frames that were ads, those will still be recorded. | 114 // any child frames that were ads, those will still be recorded. |
115 UMA_HISTOGRAM_BOOLEAN( | 115 UMA_HISTOGRAM_BOOLEAN( |
116 "PageLoad.Clients.Ads.Google.Navigations.NonAdFrameRenavigatedToAd", | 116 "PageLoad.Clients.Ads.Google.Navigations.NonAdFrameRenavigatedToAd", |
117 FrameIsAd(navigation_handle)); | 117 FrameIsAd(navigation_handle)); |
118 } else if (top_level_subframe) { | 118 } else if (navigation_handle->IsParentMainFrame()) { |
119 top_level_subframe_count_ += 1; | 119 top_level_subframe_count_ += 1; |
120 } | 120 } |
121 | 121 |
122 // Determine who the parent frame's ad ancestor is. | 122 // Determine who the parent frame's ad ancestor is. |
123 const auto& parent_id_and_data = | 123 const auto& parent_id_and_data = |
124 ad_frames_data_.find(parent_frame_host->GetFrameTreeNodeId()); | 124 ad_frames_data_.find(parent_frame_host->GetFrameTreeNodeId()); |
125 DCHECK(parent_id_and_data != ad_frames_data_.end()); | 125 DCHECK(parent_id_and_data != ad_frames_data_.end()); |
126 AdFrameData* ad_data = parent_id_and_data->second; | 126 AdFrameData* ad_data = parent_id_and_data->second; |
127 | 127 |
128 if (!ad_data && FrameIsAd(navigation_handle)) { | 128 if (!ad_data && FrameIsAd(navigation_handle)) { |
129 // This frame is not nested within an ad frame but is itself an ad. | 129 // This frame is not nested within an ad frame but is itself an ad. |
130 ad_frames_data_storage_.emplace_back(frame_tree_node_id); | 130 ad_frames_data_storage_.emplace_back(frame_tree_node_id); |
131 ad_data = &ad_frames_data_storage_.back(); | 131 ad_data = &ad_frames_data_storage_.back(); |
132 } | 132 } |
133 | 133 |
134 ad_frames_data_[frame_tree_node_id] = ad_data; | 134 ad_frames_data_[frame_tree_node_id] = ad_data; |
135 | 135 |
136 if (top_level_subframe && ad_data) | 136 if (navigation_handle->IsParentMainFrame() && ad_data) |
137 top_level_ad_frame_count_ += 1; | 137 top_level_ad_frame_count_ += 1; |
138 | 138 |
139 ProcessOngoingNavigationResource(frame_tree_node_id); | 139 ProcessOngoingNavigationResource(frame_tree_node_id); |
140 return CONTINUE_OBSERVING; | 140 return CONTINUE_OBSERVING; |
141 } | 141 } |
142 | 142 |
143 page_load_metrics::PageLoadMetricsObserver::ObservePolicy | 143 page_load_metrics::PageLoadMetricsObserver::ObservePolicy |
144 AdsPageLoadMetricsObserver::FlushMetricsOnAppEnterBackground( | 144 AdsPageLoadMetricsObserver::FlushMetricsOnAppEnterBackground( |
145 const page_load_metrics::PageLoadTiming& timing, | 145 const page_load_metrics::PageLoadTiming& timing, |
146 const page_load_metrics::PageLoadExtraInfo& extra_info) { | 146 const page_load_metrics::PageLoadExtraInfo& extra_info) { |
(...skipping 22 matching lines...) Expand all Loading... | |
169 // Data uris should be accounted for in the generating resource, not | 169 // Data uris should be accounted for in the generating resource, not |
170 // here. Blobs for PlzNavigate shouldn't be counted as the http resource | 170 // here. Blobs for PlzNavigate shouldn't be counted as the http resource |
171 // was already counted. Blobs for other things like CacheStorage or | 171 // was already counted. Blobs for other things like CacheStorage or |
172 // IndexedDB are also ignored for now, as they're not normal HTTP loads. | 172 // IndexedDB are also ignored for now, as they're not normal HTTP loads. |
173 return; | 173 return; |
174 } | 174 } |
175 | 175 |
176 const auto& id_and_data = | 176 const auto& id_and_data = |
177 ad_frames_data_.find(extra_request_info.frame_tree_node_id); | 177 ad_frames_data_.find(extra_request_info.frame_tree_node_id); |
178 if (id_and_data == ad_frames_data_.end()) { | 178 if (id_and_data == ad_frames_data_.end()) { |
179 // This resouce is for a frame that hasn't yet committed. It must be the | 179 // This resource is for a frame that hasn't ever finished a navigation. We |
180 // main document for the frame. Hold onto it and once it commits we'll run | 180 // expect it to be the frame's main resource but don't have enough |
181 // it in ProcessOngoingNavigationResource. | 181 // confidence in that to dcheck it. For example, there might be races |
182 // TODO(jkarlin): Plumb the resource type through and DCHECK that the type | 182 // between doc.written resources and navigation failure. |
183 // is document. | 183 // TODO(jkarlin): Add UMA to measure how often we see multiple resources |
184 auto it_and_success = ongoing_navigation_resources_.emplace( | 184 // for a frame before navigation finishes. |
185 ongoing_navigation_resources_.emplace( | |
185 std::piecewise_construct, | 186 std::piecewise_construct, |
186 std::forward_as_tuple(extra_request_info.frame_tree_node_id), | 187 std::forward_as_tuple(extra_request_info.frame_tree_node_id), |
187 std::forward_as_tuple( | 188 std::forward_as_tuple( |
188 extra_request_info.url, extra_request_info.frame_tree_node_id, | 189 extra_request_info.url, extra_request_info.frame_tree_node_id, |
189 extra_request_info.was_cached, extra_request_info.raw_body_bytes, | 190 extra_request_info.was_cached, extra_request_info.raw_body_bytes, |
190 extra_request_info.original_network_content_length, nullptr, | 191 extra_request_info.original_network_content_length, nullptr, |
191 extra_request_info.resource_type)); | 192 extra_request_info.resource_type)); |
192 DCHECK(it_and_success.second); | |
193 return; | 193 return; |
194 } | 194 } |
195 | 195 |
196 page_bytes_ += extra_request_info.raw_body_bytes; | 196 page_bytes_ += extra_request_info.raw_body_bytes; |
197 if (!extra_request_info.was_cached) | 197 if (!extra_request_info.was_cached) |
198 uncached_page_bytes_ += extra_request_info.raw_body_bytes; | 198 uncached_page_bytes_ += extra_request_info.raw_body_bytes; |
199 | 199 |
200 // Determine if the frame (or its ancestor) is an ad, if so attribute the | 200 // Determine if the frame (or its ancestor) is an ad, if so attribute the |
201 // bytes to the highest ad ancestor. | 201 // bytes to the highest ad ancestor. |
202 AdFrameData* ancestor_data = id_and_data->second; | 202 AdFrameData* ancestor_data = id_and_data->second; |
(...skipping 86 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... | |
289 void AdsPageLoadMetricsObserver::ProcessOngoingNavigationResource( | 289 void AdsPageLoadMetricsObserver::ProcessOngoingNavigationResource( |
290 FrameTreeNodeId frame_tree_node_id) { | 290 FrameTreeNodeId frame_tree_node_id) { |
291 const auto& frame_id_and_request = | 291 const auto& frame_id_and_request = |
292 ongoing_navigation_resources_.find(frame_tree_node_id); | 292 ongoing_navigation_resources_.find(frame_tree_node_id); |
293 if (frame_id_and_request == ongoing_navigation_resources_.end()) | 293 if (frame_id_and_request == ongoing_navigation_resources_.end()) |
294 return; | 294 return; |
295 | 295 |
296 ProcessLoadedResource(frame_id_and_request->second); | 296 ProcessLoadedResource(frame_id_and_request->second); |
297 ongoing_navigation_resources_.erase(frame_id_and_request); | 297 ongoing_navigation_resources_.erase(frame_id_and_request); |
298 } | 298 } |
OLD | NEW |