[Bug 20548] New: Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

Summary: Memory leak in pending overlapped I/O?
           Product: Wine
           Version: 1.1.32
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: ntdll
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Valgrinding the chromium ipc tests shows definite leaks like this:

768 bytes in 32 blocks are definitely lost in loss record 812 of 907
   by RtlAllocateHeap (heap.c:1421)
   by NtWriteFile (file.c:987)
   by WriteFile (file.c:532)
   by IPC::Channel::ChannelImpl::ProcessOutgoingMessages
(ipc_channel_win.cc:351)

and

5,572 bytes in 199 blocks are definitely lost in loss record 884 of 907
   by RtlAllocateHeap (heap.c:1421)
   by NtReadFile (file.c:660)
   by ReadFile (file.c:424)
   by IPC::Channel::ChannelImpl::ProcessIncomingMessages
(ipc_channel_win.cc:255)

What is supposed to free that memory?

[Bug 20548] Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

--- Comment #1 from Dan Kegel   2009-11-01 15:38:13 ---
Here's a similar example, this time in ws2:

52 bytes in 1 blocks are definitely lost in loss record 431 of 983
   by RtlAllocateHeap (heap.c:1421)
   by WS2_register_async_shutdown (socket.c:1526)
   by WS_shutdown (socket.c:3468)
   by net::TCPClientSocketWin::Disconnect (tcp_client_socket_win.cc:345)
   by net::`anonymous namespace'::TCPClientSocketTest_Connect_Test::TestBody
(tcp_client_socket_unittest.cc:110)

Again, if the status is STATUS_PENDING, who's supposed to free that memory?

To reproduce the second example:
 mkdir demo
 cd demo
 wget -c http://kegel.com/wine/chromium/chromium-tests.tar.bz2
 tar -xjvf chromium-tests.tar.bz2
 valgrind --trace-children=yes --leak-check=full wine
src/chrome/Debug/net_unittests.exe --gtest_filter=TCPClientSocketTest.Connect

[Bug 20548] Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

--- Comment #2 from Dan Kegel   2009-11-09 15:31:46 ---
And another one in ConnectNamedPipe:

24 bytes in 1 blocks are definitely lost in loss record 218 of 920
   at RtlAllocateHeap (heap.c:1423)
   by server_ioctl_file (file.c:1216)
   by NtFsControlFile (file.c:1451)
   by ConnectNamedPipe (sync.c:1588)
   by IPC::Channel::ChannelImpl::ProcessConnection (ipc_channel_win.cc:202)

To reproduce,
 wget -c http://kegel.com/wine/chromium/chromium-tests.tar.bz2
 tar -xjvf chromium-tests.tar.bz2
 valgrind --trace-children=yes --leak-check=full wine
src/chrome/Debug/ipc_tests.exe

[Bug 20548] Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

--- Comment #3 from Alexandre Julliard   2009-11-11 05:39:32 ---
The memory is freed once the user APC routine is called. If the app exits
without doing an alertable wait then it won't get freed. Fixing this would be
quite complicated since we can't free memory inside a signal handler.

[Bug 20548] Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |WONTFIX

--- Comment #4 from Alexandre Julliard   2009-11-11 10:00:09 ---
Marking wontfix for now, reopen if it affects a real app.

[Bug 20548] Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

Dmitry Timoshkov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #5 from Dmitry Timoshkov   2009-11-16 06:42:20 ---
Closing WONTFIX.

[Bug 20548] Memory leak in pending overlapped I/O?

http://bugs.winehq.org/show_bug.cgi?id=20548

--- Comment #6 from Dan Kegel   2009-11-27 17:32:39 ---
FWIW, it looks like this also happens in the wine conformance
tests, in kernel32/tests/pipe.c:

24 bytes in 1 blocks are definitely lost in loss record 139 of 526
   at RtlAllocateHeap (heap.c:1697)
   by server_ioctl_file (file.c:1216) 
   by NtFsControlFile (file.c:1451)
   by ConnectNamedPipe (sync.c:1588)
   by serverThreadMain3 (pipe.c:551)

24 bytes in 1 blocks are definitely lost in loss record 140 of 526
   at RtlAllocateHeap (heap.c:1697)
   by server_ioctl_file (file.c:1216)
   by NtFsControlFile (file.c:1451)
   by ConnectNamedPipe (sync.c:1588)
   by overlapped_server (pipe.c:1406)

28 bytes in 1 blocks are definitely lost in loss record 163 of 526
   at RtlAllocateHeap (heap.c:1697)
   by NtReadFile (file.c:660)
   by ReadFile (file.c:424) 
   by serverThreadMain4 (pipe.c:697)

28 bytes in 1 blocks are definitely lost in loss record 164 of 526
   at RtlAllocateHeap (heap.c:1697)
   by NtReadFile (file.c:660)
   by ReadFile (file.c:424)
   by overlapped_server (pipe.c:1419)

192 bytes in 8 blocks are definitely lost in loss record 385 of 526
   at RtlAllocateHeap (heap.c:1697)
   by server_ioctl_file (file.c:1216)
   by NtFsControlFile (file.c:1451)
   by ConnectNamedPipe (sync.c:1588)
   by serverThreadMain4 (pipe.c:671)