summaryrefslogtreecommitdiffstats
path: root/net/socket/client_socket_pool_base.h
Commit message (Collapse)AuthorAgeFilesLines
* Turn off backup sockets for most tests in net_unittests.willchan@chromium.org2011-04-081-1/+1
| | | | | | | | | | BUG=78303 TEST=net_unittests Review URL: http://codereview.chromium.org/6814017 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@80972 0039d316-1c4b-4281-b951-d872f2087c98
* Rename a number of classes previously labeled "TCP" to "Transport" inmbelshe@chromium.org2011-04-071-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | preparation for non-TCP transports. This helps because the alternative is to either use non-TCP protocols (like SCTP) in classes which are called "TCPClientSocketPool", or to clone the code as "SCTPClientSocketPool", both of which are less than ideal. For now, we're just testing transports, so the TransportSocketPool classes will determine a single type of transport and just use them. In the future we'll likely need to figure out how to deal with runtime selection of transports, and probably support use of multiple transports either within the same pools or within subpools. But that is for the future. Note that the histograms have some "tcp" references to them. I didn't change these to "transport" yet, because it will effect existing histograms. Renames include: classes: TCPClientSocketPool -> TransportClientSocketPool MockTCPClientSocketPool -> MockTransportClientSocketPool TCPSocketParams -> TransportSocketParams methods (not the exhaustive list): CreateTCPClientSocket() -> CreateTransportClientSocket() DoTCPConnect() -> DoTransportConnect() BUG=none TEST=none Review URL: http://codereview.chromium.org/6804028 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@80781 0039d316-1c4b-4281-b951-d872f2087c98
* Add ignore limits flag and use for sync requestkristianm@google.com2011-03-301-1/+7
| | | | | | | | | | | | | Letting sync request ignore the socket and group limits when loading to prevent the WebCore thread from locking up. BUG=45986,58703 TEST=No new tests Review URL: http://codereview.chromium.org/6756004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@79817 0039d316-1c4b-4281-b951-d872f2087c98
* Move some files from base to base/memory.levin@chromium.org2011-03-281-3/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | raw_scoped_refptr_mismatch_checker.h ref_counted.cc ref_counted.h ref_counted_memory.cc ref_counted_memory.h ref_counted_unittest.cc scoped_callback_factory.h scoped_comptr_win.h scoped_handle.h scoped_native_library.cc scoped_native_library.h scoped_native_library_unittest.cc scoped_nsobject.h scoped_open_process.h scoped_ptr.h scoped_ptr_unittest.cc scoped_temp_dir.cc scoped_temp_dir.h scoped_temp_dir_unittest.cc scoped_vector.h singleton.h singleton_objc.h singleton_unittest.cc linked_ptr.h linked_ptr_unittest.cc weak_ptr.cc weak_ptr.h weak_ptr_unittest.cc BUG=None TEST=Compile Review URL: http://codereview.chromium.org/6714032 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@79524 0039d316-1c4b-4281-b951-d872f2087c98
* Reland rest of r77399.willchan@chromium.org2011-03-121-0/+2
| | | | | | | | | | | I had temporarily reverted it so I could break it up into 2 commits, so the first could be merged to 696. This is part 2. BUG=none TEST=none Review URL: http://codereview.chromium.org/6684019 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@77908 0039d316-1c4b-4281-b951-d872f2087c98
* Revert r77399 in preparation for merging a small portion of it to 696.willchan@chromium.org2011-03-111-2/+0
| | | | | | | | | | | | | | | | | | Reland r77075,r77077. They were reverted due to flaky tests, especially on valgrind. Basically, we kept hitting the backup socket timer (500ms) which would create another socket, which the tests don't expect, so they crash. I disabled the backup socket timer completely for the SPDY tests, because they make it too hard to handle the parallel alternate protocol jobs. I also deleted the HTTP fallback test from SpdyNetworkTransactionTest, because it had a similar problem. Also, it was already being tested in HttpNetworkTransactionTest. BUG=69688,75000 TEST=Try connecting to belshe.com with various proxy configurations. Should work still. Review URL: http://codereview.chromium.org/6635047 TBR=willchan@chromium.org Review URL: http://codereview.chromium.org/6681012 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@77864 0039d316-1c4b-4281-b951-d872f2087c98
* Coverity: Fix unnecessary call by values.vandebo@chromium.org2011-03-111-1/+1
| | | | | | | | | | | CID 13152, 14252 BUG=NONE TEST=NONE Review URL: http://codereview.chromium.org/6665016 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@77826 0039d316-1c4b-4281-b951-d872f2087c98
* Reland r77075,r77077.willchan@chromium.org2011-03-091-0/+2
| | | | | | | | | | | | | They were reverted due to flaky tests, especially on valgrind. Basically, we kept hitting the backup socket timer (500ms) which would create another socket, which the tests don't expect, so they crash. I disabled the backup socket timer completely for the SPDY tests, because they make it too hard to handle the parallel alternate protocol jobs. I also deleted the HTTP fallback test from SpdyNetworkTransactionTest, because it had a similar problem. Also, it was already being tested in HttpNetworkTransactionTest. BUG=69688,75000 TEST=Try connecting to belshe.com with various proxy configurations. Should work still. Review URL: http://codereview.chromium.org/6635047 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@77399 0039d316-1c4b-4281-b951-d872f2087c98
* Plumb through NetworkChangeNotifier::IsOffline() to WebKit, enablingadamk@chromium.org2011-03-041-2/+2
| | | | | | | | | | | | | navigator.onLine and online/offline events. Only works on Windows at the moment, as IsCurrentlyOffline() is supported only by NetworkChangeNotifierWin. Most of the changes are due to the need to support two different kinds of NetworkChangeNotifier observers. Both observers currently happen to trigger on the same event, but that could change, e.g., if we store the previous online state and only notify on a change. Thus the need for two different observer interfaces, and associated Add/Remove methods. BUG=7469 TEST=Load https://bug336359.bugzilla.mozilla.org/attachment.cgi?id=220609, unplug network cable, reload, see that page changes to note offline status Review URL: http://codereview.chromium.org/6526059 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@76985 0039d316-1c4b-4281-b951-d872f2087c98
* Even more reordering the methods in headers and implementation in net/.erg@google.com2011-01-201-2/+3
| | | | | | | | | BUG=68682 TEST=compiles Review URL: http://codereview.chromium.org/6314010 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@71880 0039d316-1c4b-4281-b951-d872f2087c98
* More reordering the methods in headers in net/.erg@chromium.org2011-01-111-8/+8
| | | | | | | | | BUG=68682 TEST=compiles Review URL: http://codereview.chromium.org/6186005 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@71017 0039d316-1c4b-4281-b951-d872f2087c98
* Fix preconnect crash when we hit max socket limit.willchan@chromium.org2010-12-021-0/+5
| | | | | | | | | | | | | | | When we hit the max socket limit, we close an idle socket in order to make space for the new preconnecting socket. It's possible for the selected socket to belong to the same connection group as the one we're preconnecting a socket for. This is obviously broken. The bug currently results in us potentially deleting the ClientSocketPoolHelper::Group associated with that connection group, so the |Group* group| local variable in RequestSocketInternal() is now invalid. Any access to that variable later on in the function results in badness. This was safe before because we would never try to close an idle socket in the connection group we're request a socket for, because the first condition in RequestSocketInternal() checks to see if we can reuse an idle socket. In the preconnect case, since we want to warm up the number of sockets in the connection group, we bypass idle sockets. The solution is to create a new function: CloseOneIdleSocketExceptInGroup(const Group*). This way we avoid this problem. I provide a return value so that the caller can tell whether or not an idle socket was closed. If it was not closed (it's possible that the connection group we're requesting a socket for is the only one with idle sockets), then the caller can tell, so it can pass up the failure so RequestSockets() knows that we've hit the max socket limit and there's no point in trying to preconnect more sockets. BUG=64940 TEST=New unittest. Review URL: http://codereview.chromium.org/5549001 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@67953 0039d316-1c4b-4281-b951-d872f2087c98
* Fix clang build: remove a const on a return value.willchan@chromium.org2010-10-131-1/+1
| | | | | | | | | BUG=none TEST=builds Review URL: http://codereview.chromium.org/3768001 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@62370 0039d316-1c4b-4281-b951-d872f2087c98
* Add ClientSocketPoolBaseHelper support for preconnect.willchan@chromium.org2010-10-131-18/+90
| | | | | | | | | | | | | | | | | | | Adds a RequestSockets() API to ClientSocketPool interface. - no RequestPriority param, all requests default to LOWEST. - adds a |num_sockets| param to control how many sockets to try to ensure are connected. Adds an implementation for said function in ClientSocketPoolBaseHelper. Adds a new ClientSocketPoolBaseHelper::Flag type to modify socket request behavior. In this case, we bypass idle sockets. Adds a preconnect concept to ConnectJob. This lets normal requests hijack preconnect jobs. Modifies all ClientSocketPool subclasses to support new RequestSockets API(). Adds new tests. No client actually uses this API yet. We need to plumb it up to the preconnect system. BUG=54450 TEST=new tests in ClientSocketPoolBaseTest Review URL: http://codereview.chromium.org/3689004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@62365 0039d316-1c4b-4281-b951-d872f2087c98
* Prefer used idle sockets over unused idle sockets.willchan@chromium.org2010-10-061-3/+4
| | | | | | | | | | | | * Pick used idle sockets LIFO. * In absence of used idle sockets, pick unused idle sockets FIFO. BUG=57491 TEST=ClientSocketPoolBaseTest.PreferUsedSocketToUnusedSocket Review URL: http://codereview.chromium.org/3539013 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@61710 0039d316-1c4b-4281-b951-d872f2087c98
* Adds the ability for a link to set the filter in the eventsmmenke@chromium.org2010-09-301-0/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | tab. Added two new filter types: "type" and "id". "Type" specifies source type, and "id" specifies source ID. Both only allow exact matches, and multiple values of both can be specified ("id:5 id:6" will match source IDs 5 and 6). Added links to the #dns/#sockets tabs to link to filters to display all outstanding DNS requests/live sockets. The idle and connecting socket counts also now link to source id filters, using ids which are now passed to Javascript along with the rest of the socket pool information. As socket pools do not seem to maintain a list of active sockets, the same is not yet done for active sockets. Also fixes a bug that resulted in handed out/connecting socket counts being reversed in the primary socket pool table. BUG=56437 TEST=manual Review URL: http://codereview.chromium.org/3436033 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@61069 0039d316-1c4b-4281-b951-d872f2087c98
* Stop refcounting ClientSocketPool.willchan@chromium.org2010-09-291-32/+27
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Establishes that HttpNetworkSession owns all the socket pools. Move out all the socket pools into a ClientSocketPoolManager. This is because of the dependency tree amongst socket pools, which dictates the order in which they must be constructed and destructed. In order to better establish it, I moved them out to their own class. HttpNetworkSession owns the ClientSocketPoolManager which owns the pools. We pass the pools as raw pointers everywhere. Note that ClientSocketPoolManager owns more pools than are publicly accessible via its interface. That's because some of them are wrapped by publicly exposed pools. Also, ClientSocketPoolHistograms used to be reference counted. That's because it can be shared by multiple ClientSocketPools. But it's effectively a global as well, so I make their lifetimes persist for the length of ClientSocketPoolManager too. I also removed internal refcounting in ClientSocketPoolBase. I had refcounted it before I knew about ScopedRunnableMethodFactory back when I first started. I cleaned up the unit tests a lot. Back when I was a young padawan, I didn't really know what I was doing, so I copy/pasted a metric asston of code. Turns out most of it was stupid, so I fixed it. I also stopped the use of implementation inheritance with ClientSocketPoolTest because it's discouraged by the style guide and more importantly because it caused the ClientSocketHandles within the TestSocketRequest vector to be destroyed _after_ the pools themselves were destroyed, which is bad since the handles will call pool_->Release() which blows up. BUG=56215,56215 TEST=Existing unit tests Review URL: http://codereview.chromium.org/3389020 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@60983 0039d316-1c4b-4281-b951-d872f2087c98
* Clean up socket backup job when there are no requests.willchan@chromium.org2010-09-271-1/+1
| | | | | | | | | BUG=56532 TEST=ClientSocketPoolBaseTest.CancelBackupSocketAfterFinishingAllRequests Review URL: http://codereview.chromium.org/3441034 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@60693 0039d316-1c4b-4281-b951-d872f2087c98
* Break reference cycle from HttpProxyClientSocket=>HttpNetworkSession=>...willchan@chromium.org2010-09-171-3/+1
| | | | | | | | | | | | Note that this undoes the fix for http://crbug.com/49387 which is now unnecessary without the cycle. Some other miscellaneous cleanup is thrown in here. BUG=55175 TEST=none Review URL: http://codereview.chromium.org/3418018 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@59873 0039d316-1c4b-4281-b951-d872f2087c98
* Add missing nested socket pools to socket pool info page. Each socket pool ↵mmenke@chromium.org2010-09-081-3/+4
| | | | | | | | | | | should be displayed once and only once. Also, if more than one proxy is in use, socket pools with the same proxy will appear adjacent to each other. BUG=39756 TEST=manual Review URL: http://codereview.chromium.org/3328009 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@58826 0039d316-1c4b-4281-b951-d872f2087c98
* Fix ClientSocketHandle reuse_type(). Correctly track socket use.willchan@chromium.org2010-09-021-5/+3
| | | | | | | | | | | | | In particular, we used to consider that a socket had been used whenever it got returned to the ClientSocketPool. But, with preconnect, that is no longer true. Luckily, we now have UseHistory in the transport sockets. So, I create a WasEverUsed() method in ClientSocket, plumb this into all sockets, and use that in ClientSocketPoolBaseHelper instead of tracking whether or not the socket had been returned to the client or not. This ultimately will have two implications. We will record the correct values in Net.HttpSocketType histograms and we will use the correct timeout for preconnect sockets in ClientSocketPoolBaseHelper::CleanupIdleSockets(). BUG=none TEST=none Review URL: http://codereview.chromium.org/3353004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@58363 0039d316-1c4b-4281-b951-d872f2087c98
* Abort ConnectJobs with ERR_ABORTED when network changes.willchan@chromium.org2010-08-311-3/+6
| | | | | | | | | BUG=50665 TEST=ClientSocketPoolBaseTest.AbortAllRequestsOnFlush Review URL: http://codereview.chromium.org/3255005 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@58042 0039d316-1c4b-4281-b951-d872f2087c98
* Sockets page on net-internals now displays some information about the ↵mmenke@chromium.org2010-08-301-0/+8
| | | | | | | | | | current socket pool state. Table padding slightly increased for legibility. TEST=manual BUG=39756 Review URL: http://codereview.chromium.org/3267002 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@57869 0039d316-1c4b-4281-b951-d872f2087c98
* Reland r57100. Only create the backup ConnectJob when it is needed.willchan@chromium.org2010-08-261-42/+49
| | | | | | | | Includes a fix. I forgot that the iterator is invalidated after std::map::erase(). Review URL: http://codereview.chromium.org/3135047 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@57515 0039d316-1c4b-4281-b951-d872f2087c98
* Fix a crash in ClientSocketPoolBaseHelper where we double erase a Group from ↵willchan@chromium.org2010-08-251-0/+6
| | | | | | | | | | the GroupMap. BUG=49254 Review URL: http://codereview.chromium.org/3104026 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@57346 0039d316-1c4b-4281-b951-d872f2087c98
* A/B experiment for re-establishing TCP connections.ziadh@chromium.org2010-08-251-8/+9
| | | | | | | | | | | | | We would like to test the impact of automatic retries when a TCP connection exceeds a certain threshold before it gets back an ACK. We are observing a fair number of sockets where the connection had been established, but the sockets were not used thereafter. r=mbelche Review URL: http://codereview.chromium.org/3191019 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@57276 0039d316-1c4b-4281-b951-d872f2087c98
* Revert 57100 - Only create the backup ConnectJob when it is needed.thestig@chromium.org2010-08-231-49/+42
| | | | | | | | | | | | | | This reduces the unnecessary NetLog spam, since we would log to the NetLog everytime we created the backup ConnectJob, even though we usually wouldn't actually call Connect() on it. Now, we only create the backup ConnectJob when we intend to call Connect() on it. Includes some various cleanup necessary for this: * struct Group => class Group * method_factory moves from CSP to Group Review URL: http://codereview.chromium.org/3171017 TBR=willchan@chromium.org Review URL: http://codereview.chromium.org/3198009 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@57107 0039d316-1c4b-4281-b951-d872f2087c98
* Only create the backup ConnectJob when it is needed.willchan@chromium.org2010-08-231-42/+49
| | | | | | | | | | | This reduces the unnecessary NetLog spam, since we would log to the NetLog everytime we created the backup ConnectJob, even though we usually wouldn't actually call Connect() on it. Now, we only create the backup ConnectJob when we intend to call Connect() on it. Includes some various cleanup necessary for this: * struct Group => class Group * method_factory moves from CSP to Group Review URL: http://codereview.chromium.org/3171017 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@57100 0039d316-1c4b-4281-b951-d872f2087c98
* Reland 54771 (and 54795) To enable TCP Preconnection by defaultjar@chromium.org2010-08-061-6/+5
| | | | | | | | | | | | | | | | | | | | | | | I pulled out the code to update the socket connectivity stats. I added defensive code which should preclude the crash that was reported on the stability bot. I added a second call to update the stats from ~ClientSocketHandle to ensure that we updated the related ClientSocket when we are torn down. As noted in the original checkin: Enable speculative preconnection by default Added histogram to track preconnection utilization (vs waste). BUG=42694 r=mbelshe Review URL: http://codereview.chromium.org/3050040 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@55197 0039d316-1c4b-4281-b951-d872f2087c98
* Revert 55071 - Reland 54771 (and 54795) To enable TCP Preconnection by defaultdhollowa@chromium.org2010-08-051-5/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Leaks reported with this CL. http://build.chromium.org/buildbot/memory/builders/Linux%20Heapcheck/builds/6130/steps/heapcheck%20test:%20net/logs/stdio Eg. Leak of 24 bytes in 1 objects allocated from: @ 84aece net::SSLClientSocketNSS::BufferRecv @ 84b161 net::SSLClientSocketNSS::DoTransportIO @ 84ca1f net::SSLClientSocketNSS::DoHandshakeLoop @ 84ca6b net::SSLClientSocketNSS::OnHandshakeIOComplete @ 84cadc net::SSLClientSocketNSS::OnRecvComplete @ 84cbb0 net::SSLClientSocketNSS::BufferRecvComplete @ 84ea4b void DispatchToMethod @ 84ea7b CallbackImpl::RunWithParams @ 4b3a10 CallbackRunner::Run @ 853e7e net::TCPClientSocketLibevent::DoReadCallback @ 85426f net::TCPClientSocketLibevent::DidCompleteRead @ 856a5c net::TCPClientSocketLibevent::ReadWatcher::OnFileCanReadWithoutBlocking @ 93d8fd base::MessagePumpLibevent::FileDescriptorWatcher::OnFileCanReadWithoutBlocking @ 93d966 base::MessagePumpLibevent::OnLibeventNotification @ 9da639 event_process_active @ 9da923 event_base_loop @ 93dfd0 base::MessagePumpLibevent::Run @ 8f2873 MessageLoop::RunInternal @ 8f2893 MessageLoop::RunHandler @ 8f2938 MessageLoop::Run @ 44b7f9 TestCompletionCallback::WaitForResult @ 6a1ee6 SSLClientSocketTest_ConnectMismatched_Test::TestBody @ 961831 testing::Test::Run @ 965026 testing::internal::TestInfoImpl::Run @ 96515c testing::TestCase::Run @ 965bbe testing::internal::UnitTestImpl::RunAllTests @ 965d35 testing::UnitTest::Run @ 4a4bf7 TestSuite::Run @ 4a3b6d main @ 2adff5bb11c4 __libc_start_main Suppression: { <insert_a_suppression_name_here> Heapcheck:Leak fun:net::SSLClientSocketNSS::BufferRecv fun:net::SSLClientSocketNSS::DoTransportIO fun:net::SSLClientSocketNSS::DoHandshakeLoop fun:net::SSLClientSocketNSS::OnHandshakeIOComplete fun:net::SSLClientSocketNSS::OnRecvComplete fun:net::SSLClientSocketNSS::BufferRecvComplete fun:void DispatchToMethod fun:CallbackImpl::RunWithParams fun:CallbackRunner::Run fun:net::TCPClientSocketLibevent::DoReadCallback fun:net::TCPClientSocketLibevent::DidCompleteRead fun:net::TCPClientSocketLibevent::ReadWatcher::OnFileCanReadWithoutBlocking fun:base::MessagePumpLibevent::FileDescriptorWatcher::OnFileCanReadWithoutBlocking fun:base::MessagePumpLibevent::OnLibeventNotification fun:event_process_active fun:event_base_loop fun:base::MessagePumpLibevent::Run fun:MessageLoop::RunInternal fun:MessageLoop::RunHandler fun:MessageLoop::Run fun:TestCompletionCallback::WaitForResult fun:SSLClientSocketTest_ConnectMismatched_Test::TestBody fun:testing::Test::Run fun:testing::internal::TestInfoImpl::Run fun:testing::TestCase::Run fun:testing::internal::UnitTestImpl::RunAllTests fun:testing::UnitTest::Run fun:TestSuite::Run fun:main fun:__libc_start_main } I added defensive code in ClientSocketHandle::ReleaseSocket(), which should preclude the crash that was reported on the stability bot. I added a second call to ReleaseSocket() from ~ClientSocketHandle to ensure that we updated the related ClientSocket when we are torn down. r=mbelshe Review URL: http://codereview.chromium.org/3071022 TBR=jar@chromium.org git-svn-id: svn://svn.chromium.org/chrome/trunk/src@55090 0039d316-1c4b-4281-b951-d872f2087c98
* Reland 54771 (and 54795) To enable TCP Preconnection by defaultjar@chromium.org2010-08-051-6/+5
| | | | | | | | | | | | | | | I added defensive code in ClientSocketHandle::ReleaseSocket(), which should preclude the crash that was reported on the stability bot. I added a second call to ReleaseSocket() from ~ClientSocketHandle to ensure that we updated the related ClientSocket when we are torn down. r=mbelshe Review URL: http://codereview.chromium.org/3071022 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@55071 0039d316-1c4b-4281-b951-d872f2087c98
* Revert 54771 - Enable speculative preconnection by defaultjar@chromium.org2010-08-041-5/+6
| | | | | | | | | | | | | | Added histogram to track utilization (vs waste). [The stability bot was reporting problems, so I'm reverting] r=mbelshe Review URL: http://codereview.chromium.org/3026038 TBR=jar@chromium.org Review URL: http://codereview.chromium.org/3090011 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@54930 0039d316-1c4b-4281-b951-d872f2087c98
* Enable speculative preconnection by defaultjar@chromium.org2010-08-031-6/+5
| | | | | | | | | Added histogram to track utilization (vs waste). r=mbelshe Review URL: http://codereview.chromium.org/3026038 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@54771 0039d316-1c4b-4281-b951-d872f2087c98
* `#pragma once` for app, base, chrome, gfx, ipc, net, skia, viewsthakis@chromium.org2010-07-261-0/+1
| | | | | | | | | BUG=50273 TEST=everything still builds, build is 10% faster on windows, same speed on mac/linux TBR: erg git-svn-id: svn://svn.chromium.org/chrome/trunk/src@53716 0039d316-1c4b-4281-b951-d872f2087c98
* Fix idle socket cleanup code path.vandebo@chromium.org2010-07-221-0/+3
| | | | | | | | | | | | This reverses the order that sockets removed from the idle list and deleted. It also detects and handles recursive calls to CleanupIdleSockets. BUG=49387 TEST=none Review URL: http://codereview.chromium.org/2856049 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@53374 0039d316-1c4b-4281-b951-d872f2087c98
* Refactor how ClientSocketPoolBaseHelper avoids re-entrancy.willchan@chromium.org2010-07-151-21/+43
| | | | | | | | | Specifically, we defer asynchronous user callbacks to tasks. BUG=48861 Review URL: http://codereview.chromium.org/2994003 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@52509 0039d316-1c4b-4281-b951-d872f2087c98
* Implement SSLClientSocketPool.vandebo@chromium.org2010-07-141-0/+5
| | | | | | | | | | | | To support SSLClientSocketPool, ClientSocketPoolBase and ClientSocketHandle require a notion of additional error state reported from the pool. Overtime the error handling may get become more integrated, alleviating the need for some of the additional error state. To support getting Http Proxy credentials from the user, the SSLClientSocketPool will release unauthenticated HttpProxyClientSocket's into the pool as idle. However, it checks their authentication status when receiving one, completing the authentication once the user has provided the credentials. BUG=30357 TEST=existing unit tests, ClientSocketPoolBaseTest.AdditionalErrorState*, SSLClientSocketPoolTest.* Review URL: http://codereview.chromium.org/2870030 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@52275 0039d316-1c4b-4281-b951-d872f2087c98
* Make the various SocketParams reference counted.vandebo@chromium.org2010-07-121-4/+4
| | | | | | | | | | | This is so that the SSLSocketParam can hold one of any of the existing SocketParams. BUG=30357 TEST=existing unit tests Review URL: http://codereview.chromium.org/2848029 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@52107 0039d316-1c4b-4281-b951-d872f2087c98
* Reland 51081:mbelshe@chromium.org2010-07-111-61/+28
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is relandable now because we fixed a problem with the backup sockets, which was the real reason for initially reverting. We basically don't do late socket binding when a connect has already been started for a request, even if another socket frees up earlier. The reassignment logic was quite complicated, so I reworked it. Fixing this bug was easy by changing the way FindTopStalledGroup worked, but because that function is called in that loop, changing this case caused the loop to go infinitely in some cases. This led me to look into unwinding the loop. The problem really came down to ReleaseSocket/DoReleaseSocket. Because we allow for a pending queue of released sockets, we had to do this looping (which has been a source of bugs before). To fix, I eliminated the pending_releases queue. I also reworked the routes through OnAvailableSocketSlot to unify them and always run asynchronously. The result is that now we no longer have the loop. So when one socket is released, we hand out exactly one socket. Note also that this logic slightly changes the priority of how we recycle sockets. Previously, we always consulted the TopStalledGroup. The TopStalledGroup is really only interesting in the case where we're at our max global socket limit, which is rarely the case. In the new logic, when a socket is released, first priority goes to any pending socket in the same group, regardless of that group's priority. The reason is why close a socket we already have open? Previously, if the released socket's group was not the highest priority group, the socket would be marked idle, then closed (to make space for a socket to the TopStalledGroup), and finally a new socket created. I believe the new algorithm, while not perfectly matching the priorities, is more efficient (less churn on sockets), and also is more graceful to the common case. Finally OnAvailableSocketSlot does two things. First, it tries to "give" the now available slot to a particular group, which is dependent on how OnAvailableSocketSlot was called. If we're currently stalled on max sockets, it will also check (after giving the socket out) to see if we can somehow free something up to satisfy a stalled group. If that second step fails for whatever reason, we don't loop. In theory, this could mean that we go under the socket max and didn't dish out some sockets right away. To make sure that multiple stalled groups can get unblocked, we'll record the number of stalled groups, and once in this mode, OnAvailableSocketSlot will keep checking for stalled groups until the count finally drops to zero. BUG=47375 TEST=DelayedSocketBindingWaitingForConnect,CancelStalledSocketAtSocketLimit Review URL: http://codereview.chromium.org/2938006 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@52050 0039d316-1c4b-4281-b951-d872f2087c98
* Revert 51081 for causing crashes -willchan@chromium.org2010-07-071-28/+61
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | We basically don't do late socket binding when a connect has already been started for a request, even if another socket frees up earlier. The reassignment logic was quite complicated, so I reworked it. Fixing this bug was easy by changing the way FindTopStalledGroup worked, but because that function is called in that loop, changing this case caused the loop to go infinitely in some cases. This led me to look into unwinding the loop. The problem really came down to ReleaseSocket/DoReleaseSocket. Because we allow for a pending queue of released sockets, we had to do this looping (which has been a source of bugs before). To fix, I eliminated the pending_releases queue. I also reworked the routes through OnAvailableSocketSlot to unify them and always run asynchronously. The result is that now we no longer have the loop. So when one socket is released, we hand out exactly one socket. Note also that this logic slightly changes the priority of how we recycle sockets. Previously, we always consulted the TopStalledGroup. The TopStalledGroup is really only interesting in the case where we're at our max global socket limit, which is rarely the case. In the new logic, when a socket is released, first priority goes to any pending socket in the same group, regardless of that group's priority. The reason is why close a socket we already have open? Previously, if the released socket's group was not the highest priority group, the socket would be marked idle, then closed (to make space for a socket to the TopStalledGroup), and finally a new socket created. I believe the new algorithm, while not perfectly matching the priorities, is more efficient (less churn on sockets), and also is more graceful to the common case. Finally OnAvailableSocketSlot does two things. First, it tries to "give" the now available slot to a particular group, which is dependent on how OnAvailableSocketSlot was called. If we're currently stalled on max sockets, it will also check (after giving the socket out) to see if we can somehow free something up to satisfy a stalled group. If that second step fails for whatever reason, we don't loop. In theory, this could mean that we go under the socket max and didn't dish out some sockets right away. To make sure that multiple stalled groups can get unblocked, we'll record the number of stalled groups, and once in this mode, OnAvailableSocketSlot will keep checking for stalled groups until the count finally drops to zero. BUG=47375 TEST=DelayedSocketBindingWaitingForConnect,CancelStalledSocketAtSocketLimit Review URL: http://codereview.chromium.org/2861023 TBR=mbelshe@chromium.org BUG=48094 Review URL: http://codereview.chromium.org/2809052 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@51755 0039d316-1c4b-4281-b951-d872f2087c98
* We basically don't do late socket binding when a connect has alreadymbelshe@chromium.org2010-06-291-61/+28
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | been started for a request, even if another socket frees up earlier. The reassignment logic was quite complicated, so I reworked it. Fixing this bug was easy by changing the way FindTopStalledGroup worked, but because that function is called in that loop, changing this case caused the loop to go infinitely in some cases. This led me to look into unwinding the loop. The problem really came down to ReleaseSocket/DoReleaseSocket. Because we allow for a pending queue of released sockets, we had to do this looping (which has been a source of bugs before). To fix, I eliminated the pending_releases queue. I also reworked the routes through OnAvailableSocketSlot to unify them and always run asynchronously. The result is that now we no longer have the loop. So when one socket is released, we hand out exactly one socket. Note also that this logic slightly changes the priority of how we recycle sockets. Previously, we always consulted the TopStalledGroup. The TopStalledGroup is really only interesting in the case where we're at our max global socket limit, which is rarely the case. In the new logic, when a socket is released, first priority goes to any pending socket in the same group, regardless of that group's priority. The reason is - why close a socket we already have open? Previously, if the released socket's group was not the highest priority group, the socket would be marked idle, then closed (to make space for a socket to the TopStalledGroup), and finally a new socket created. I believe the new algorithm, while not perfectly matching the priorities, is more efficient (less churn on sockets), and also is more graceful to the common case. Finally OnAvailableSocketSlot does two things. First, it tries to "give" the now available slot to a particular group, which is dependent on how OnAvailableSocketSlot was called. If we're currently stalled on max sockets, it will also check (after giving the socket out) to see if we can somehow free something up to satisfy a stalled group. If that second step fails for whatever reason, we don't loop. In theory, this could mean that we go under the socket max and didn't dish out some sockets right away. To make sure that multiple stalled groups can get unblocked, we'll record the number of stalled groups, and once in this mode, OnAvailableSocketSlot will keep checking for stalled groups until the count finally drops to zero. BUG=47375 TEST=DelayedSocketBindingWaitingForConnect,CancelStalledSocketAtSocketLimit Review URL: http://codereview.chromium.org/2861023 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@51081 0039d316-1c4b-4281-b951-d872f2087c98
* Massively simplify the NetworkChangeNotifier infrastructure:pkasting@chromium.org2010-06-251-9/+4
| | | | | | | | | | | | | | * Use a process-wide object (singleton pattern) * Create/destroy this object on the main thread, make it outlive all consumers * Make observer-related functions threadsafe As a result, the notifier can now be used by any thread (eliminating things like NetworkChangeObserverProxy and NetworkChangeNotifierProxy, and expanding its usefulness); its creation and inner workings are much simplified (eliminating implementation-specific classes); and it is simpler to access (eliminating things like NetworkChangeNotifierThread and a LOT of passing pointers around). BUG=none TEST=Unittests; network changes still trigger notifications Review URL: http://codereview.chromium.org/2802015 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@50895 0039d316-1c4b-4281-b951-d872f2087c98
* A/B test for determining a value for unused socket timeout. Currently theziadh@chromium.org2010-06-211-9/+5
| | | | | | | | | | | | | | timeout defaults to 10 seconds. Having this value set too low won't allow us to take advantage of idle sockets. Setting it to too high could possibly result in more ERR_CONNECT_RESETs, requiring one RTT to receive the RST packet and possibly another RTT to re-establish the connection. r=jar Review URL: http://codereview.chromium.org/2827016 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@50364 0039d316-1c4b-4281-b951-d872f2087c98
* Reland my close on idle socket change (r43882+r44150).willchan@chromium.org2010-06-101-5/+16
| | | | | | | | | | I reverted it the first time because it was suspected of causing a hang on the IO thread. A different CL caused that. When I relanded it previously, the fix for the hang on the IO thread broke this (changed an assertion). I've now deleted that assertion. BUG=32817 Review URL: http://codereview.chromium.org/2716004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@49444 0039d316-1c4b-4281-b951-d872f2087c98
* Revert "Revert an idle sockets change to trigger reliability bot. Will ↵willchan@chromium.org2010-06-081-16/+5
| | | | | | | | revert again soon." Review URL: http://codereview.chromium.org/2761001 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@49146 0039d316-1c4b-4281-b951-d872f2087c98
* Revert an idle sockets change to trigger reliability bot. Will revert again ↵willchan@chromium.org2010-06-081-5/+16
| | | | | | | | soon. Review URL: http://codereview.chromium.org/2760001 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@49145 0039d316-1c4b-4281-b951-d872f2087c98
* Do not attempt to reuse active sockets after a socket pool flush (usually a ↵willchan@chromium.org2010-06-071-5/+17
| | | | | | | | | | | network change). Implements this functionality by adding an |id_| field to ClientSocketPoolBaseHelper that is incremented on each Flush(). BUG=45872 Review URL: http://codereview.chromium.org/2647003 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@49076 0039d316-1c4b-4281-b951-d872f2087c98
* Rework the logging for sockets/connectjobs.eroman@chromium.org2010-06-031-10/+13
| | | | | | | | | | | | | | | | | | In particular, make it work better when using backup jobs / late binding (the display was very confused before because of how these asynchronous events would nest). Also changed the paradigm for how PassiveLogCollector preserves these async associations -- this fixes how it replays the events to net-internals. (Before we would collapse the event streams into the SOURCE_URL_REQUEST which lost some hiearchy.. now I keep the separate streams). Some of the particular changes to the event streams: * ConnectJobs now create their own source stream internally. * Sockets are now bounded by +SOCKET_ALIVE / -SOCKET_ALIVE events (removed the one-off SOCKET_DONE event). * The socket log streams contains +SOCKET_IN_USE / -SOCKET_IN_USE event blocks to show which URLRequest was controlling it at various points in time (this makes it much easier to understand which read/writes belonged to a particular network transaction when a socket gets re-used). * ConnectJobs are bounded by +SOCKET_POOL_CONNECT_JOB / - SOCKET_POOL_CONNECT_JOB events. * ConnectJobs log the net error they failed with. * Removed the SOCKET_BACKUP_TIMER_EXTENDED event. Review URL: http://codereview.chromium.org/2363003 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@48797 0039d316-1c4b-4281-b951-d872f2087c98
* Make ClientSocketPool histograms static so that they work properly.vandebo@chromium.org2010-05-201-5/+7
| | | | | | | | | | | | Also change their names so that they appear all together on the histograms page. BUG=43375 TEST=none Review URL: http://codereview.chromium.org/2029004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@47843 0039d316-1c4b-4281-b951-d872f2087c98
* Revert r47395. Looks like it still crashes =/willchan@chromium.org2010-05-201-4/+14
| | | | | | | | BUG=40455,40457 Review URL: http://codereview.chromium.org/2104013 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@47753 0039d316-1c4b-4281-b951-d872f2087c98
generated by cgit v1.1 at 2025-02-12 01:55:38 +0000