Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02342246 2001-02-28
WO 00/13085 PCT/US99/199b8
Interface Between a Control Process and a Target Process
Technical Field
The present invention relates to a method of indirectly accessing and
controlling one process from another within a multitasking computer operating
system environment.
Back~,round Art
In older 16-bit personal computer operating system environments such as
1 o Windows 3.2~' available from Microsoft Corporation, Redmond, Washington,
all
running programs-known as processes-share a single virtual address space.
Any process can read from and write to memory belonging to any other process,
even memory used by the operating system. If a first process inadvertently
overwrites data being used by another running second process, the second
process could well become unstable and crash. With the advent of 32-bit
personal computer operating systems, mare virtual address space is available
and a more robust environment may be created which avoids this problem.
In some 32-bit computer system operating environments such as
Windows 95~, Windows 9$~, and Windows N'I~', all available from Microsoft
2o Corporation, Redmond, Washington, each process is assigned a unique process
address space, typically, 4-GB. Such a process address space contains the
executable code and data necessary for the process, that is, the code fox the
EXE
and DLL files and their required data. In addition to an address space, a
process
also owns certain resources such as files, dynamic memory allocations, and
threads. Within a given process, when pointers are used to reference memory,
the pointer value refers to a memory address in that process's own assigned
memory space. It is not possible for one process to create a pointer which
references memory assigned to another process. As a result, the system is
stable
and robust, and less likely to crash due to memory conflicts. However, it is
also
3o very difficult for one process to interact with and affect another.
CA 02342246 2001-02-28
WO 00/13085 PCT/US99/19968
For example, when a user is prompted for an input by a dialog box, the
text from the user is typically stored in an edit control. An edit control is
simply
a window used to store free-form text input by a user. Within a process,
accessing a given edit control is almost trivial. Simply obtaining the handle
(or
name) of a given edit control window allows access via the operating system's
control window messages. However, the most common such control window
messages do not work across process boundaries in a 32-bit system with
separate
virtual address spaces for each process. This is because use of control
windows
messages typically requires that some message argument parameters be in the
~ o form of a memory address of a required data structure. As described above,
however, memory addresses are only meaningful within a given process. Thus,
user text within an edit control of one process is not available to another
process.
To enable a speech recognition system to control a target process when
both are running in an advanced 32-bit windowing computer, it is easy if the
target application is written with features of the controlling speech
recognition
system in mind. However, for a speech recognition system to control a target
process that is not thus speech-enabled requires utilizing whichever suite of
characteristics the target process exhibits in relation to the operating
system.
These characteristics often make sophisticated control by a speech recognition
2o system, at best, very difficult.
Summarv of the Invention
A preferred embodiment of the present invention provides a method for
interfacing between a control process and a target process in a multitasking
computer operating system environment. The method includes the steps of
a. establishing a region of shared memory between the control
process and the target process;
b. generating a request with the control process for a desired
operation by the target process and storing arguments and data for the request
3o in. the region of shared memory;
-2-
CA 02342246 2001-02-28
WO 00/13085 PCT/US99/19968
c. communicating the request for the desired operation to the target
process; and
d. effecting execution of the desired operation by the target process.
In another embodiment, the method includes:
a. establishing a region of shared memory between the control
process and the target process;
b. hooking an executable link module into an address space of the
target process;
c. generating a request for a desired operation by the target process
1 o and storing arguments and data for the request in the shared memory;
d. posting a private message from the control process to the target
process notifying the executable link module of the request;
e. receiving the private message in the executable link module;
f. executing the requested action in the target process; and
g. sending a private message acknowledgment from the executable
link module to the control process.
In such an embodiment, the step of receiving may further include
retrieving arguments and data for the request from the shared memory. In
addition, or alternatively, the step of sending may further include writing
new
2o data from the target process into the shared memory. In such an embodiment,
the step of sending may further include retrieving the new data by the control
process from the shared memory.
A preferred embodiment is also directed to an interface system used
between a control process and a target process in a multitasking computer
operating system environment. Such a system comprises:
a. a region of shared memory that is established between the control
process and the target process;
b. a request generator which generates a request from the control
process for a desired operation by the target process and stores arguments and
so data for the request in the region of shared memory;
-3-
CA 02342246 2001-02-28
WO 00/13085 PCT/US99/19968
c. a request communicator which communicates the request for the
desired operation to the target process; and
d. a request receiver which effects execution by the target process of
the desired operation.
Another preferred embodiment includes an interface system used
between a control process and a target process in a multitasking computer
operating system environment. Such a system comprises:
a. a region of shared memory that is established between the control
process and the target process;
b. an executable link module hooked into an address space of the
target process;
c. a request generator which generates a request for a desired
operation by the target process and stores arguments and data for the request
in
the shared memory;
d. a request communicator which communicates the request for the
desired operation to the executable link module;
e. receiving the private message in the executable link module;
f. means for executing the requested action in the target process; and
g. a message acknowledger within the executable link module which
2o sends a message acknowledging receipt and accomplishment of the request by
the target process to the control process.
Brief Description of the Drawings
The present invention will be more readily understood by reference to the
following detailed description taken with the accompanying drawings, in which:
Fig. 1 illustrates the logical flow associated with a preferred embodiment.
Fig. 2 illustrates the inter-relationships of various functional blocks within
the operating system environment according to a preferred embodiment.
Fig. 3 illustrates computer process arrangements in which a preferred
3o embodiment may be employed.
-4-
CA 02342246 2001-02-28
WO 00/13085 PCT/US99/19968
Detailed Description of Specific Embodiments
A preferred embodiment of the present invention, as shown in Figs. 1-3,
includes a technique for an originating control process 24 having generally a
non-keyboard input 32, to exchange detailed information with and send
commands from a command output block 31 to a separate target process 27.
Such techniques are useful, for instance, where the control process 24 is an
automatic speech recognition engine .33 having a speech input 34, and running
in
a 32-bit operating system environment 21, such as Windows NT~, and the target
process 27 is a non-speech aware application, such as an e-mail program,
1o running separately in the same operating system environment 21. It may be
desirable for the speech recognition control process 24 to obtain access to
and
control various forms of data in another target process 27 ranging from simple
text to more complex data structures such as in a database file. A preferred
embodiment may be advantageously employed in various operating system
environments 21, including, for instance, Windows N'I~', Windows 95~,
Windows 98~, X for Windows, and UNIX.
A preferred embodiment involves inserting a hook filter function 26 {also
known as a callback function) into 'the target process 27, and creating a
pointer
29b from the hook filter function 26 to an executable link module in shared
2o system memory, such as a dynamic linked library (DLL) file 25. A hook
filter
function 26 is a program mechanism which can intercept events in a process
message queue 28 such as messages, keystrokes, and mouse actions before they
reach the target process 27. To operate between separate processes, the filter
function invoked by the hook filter function 26 is placed within a DLL 25. The
DLL 25 then allows inter-process communication and control via shared
memory within the DLL 25. Thus, the shared memory 25 is common to the
operating system 21 and accessible to both the control process 24 via pointer
29a,
and the target process 27 via pointer 29b. This technique allows for very
efficient transfer of information between the control process 24 and the
target
-5-
CA 02342246 2001-02-28
WO 00/130$5 PCTIUS99/19968
process 27. In addition, the control process 24 may obtain access to
information
from the target process 27 which is not usually available via inter process
communication (IPC), such as user input text in the edit control of a dialog
box
of a target process 27. It also allows text from the control process 24,
e.g.,the text
output from an automatic speech recognition program, to be passed, via the
shared memory in the DLL 25, to an edit control in the target process 27 which
treats the text as if it were provided by the user from a keyboard.
Such inter-process text passing to a target process edit control may be
considered a short term or an immediately obtainable advantage of a preferred
embodiment of the present invention. More prolonged or longer term
usefulness may also be realized. For example, the hook filter function 26 may
be
employed for ongoing monitoring of the operation of the target process 27 for
the occurrence of target process events which are of interest to the control
process 24. Such interesting events may include, for example, opening a menu,
making a menu selection, or opening a dialog box within the target process 27.
A preferred embodiment may start with the control process 24 installing a
WH GETMESSAGE hook filter function 26 in the target process 27, step 11 in
Fig. 1. The hook filter function 26 is installed by calling SetWindoivsHookEx
as
follows:
HHOOK _hhook = SetWindowsHookEx(WH GETMESSAGE, GetMsgProc, hinstDll,
TargetThreadName);
The first parameter, WH_GETMESSAGE, indicates the type of hook filter
function 26 to install. WH_GETMESSAGE is a hook filter function 26 called
when the GetMessage or the PeekMessage function is about to return a message
to
the target process 27. The hook filter function 26 will receive a pointer 29b
to a
message structure within the shared memory of the DLL 25 which contains the
actual message. The second parameter, GetlVIsgProc, identifies the address of
a
function in the shared memory of the DLL 25 that the operating system 21
CA 02342246 2001-02-28
WO 00/13085 PCT/US99/19968
should call whenever a window in the target process 27 is about to process a
message in its message queue 28. The third parameter, hinstDll, identifies the
DLL 25 that contains the GetMsgProc function. In Windows NT, a DLL's hinstDll
value identifies the 32-bit virtual memory address where the DLL 25 is mapped
into the address space 22 of the control process 24. Since the memory address
space of the DLL 25 will be shared by both the control process 24 and the
target
process 27, it may be considered accessible to each process via a pointer, 29a
and
29b respectively. The fourth parameter, TargetThreadName, identifies the
thread to hook in the target process 27. A value of NULL may be passed for the
1 o fourth parameter to instruct the operating system 21 to hook all threads
in the
system.
When GetMessrzge or PeekMess~rge is about to return a message to the target
process 27 from the message queue 28, the operating system 21 checks whether a
WH GETMESSAGE hook 26 is installed on that thread in the target process 27,
and whether the DLL 25 containing the GetMsgProc function is mapped into the
address space 23 of the target process 27. If the DLL 2S has not been mapped,
the operating system 21 forces the DLL 25 to be mapped into the address space
23 of the target process 27 via pointer 29b and increments a lock count on the
mapping of the DLL 25 in the target process 27. The operating system 21 looks
2o at the hinstDll of the DLL 25 as it applies to the target process 27 and
checks to
see whether the hinstDll of the DLL 25 is at the same location as it applies
to the
control process 24. If the hinstDlls are the same, the memory address of the
GetMsgProc function is also the same in the two process address spaces. In
such
a case, the operating system 22 can simply call the GetMsgProc function in the
address space 22 of the control process 24.
If the hfnstDlls are different, the operating system 21 must determine the
virtual memory address of the GetMsgProc function in the address space 23 of
the target process 27 according to the following formula:
GetMsgProcTarget = hinstDllTarget + (GetMsgProcControl - hinstDllControl)
CA 02342246 2001-02-28
WO 00!13085 PCT/US99/19968
Subtracting hinstDllControl from GetMsgProcControl results in the offset in
bytes
for the GetMsgProc function. Adding this offset to hinstDtITarget gives the
location of the GetMsgProc function as it applies to the mapping of the DLL 25
in
the address space 23 of the target process 27. The operating system 21
increments a lock count on the mapping of the DLL 25 in the address space 23
of
the target process 27, and also calls the GetMsgProc function in the address
space
23 of the target process 27. When GetMsgProc returns, the operating system 21
decrements a lock count on the mapping of the DLL 25 in the target process 27.
Injecting or mapping the DLL 25 containing the hook filter function 26
1o maps the whole DLL 25 into the target process 27, not just the hook filter
function 26. Any and all functions contained in the DLL 25 now exist and can
be
called from threads running in the context of the target process 27. For
instance,
to subclass a window created by a thread in the target process 27, the
WH GETMESSAGE hook 2b is set on the thread that created the window, and
~ 5 then, when the GefMsgProc function is called, call SetWindozvLong to
subclass the
window. Of course, the subclass procedure must be in the same DLL 25 as the
GetMsgProc function.
The hook filter function 26 examines messages in the message queue 28
coming into the target process 27 for messages from the control process 24.
2o Thereafter, the control process 24, in step 12, may generate a request for
action or
information from the target process 27 and via the pointer 29a place in the
shared memory of the DLL 25 a description of the type of operation requested
and any arguments necessary. Then, in step 13, a message is sent from the
control process 24 to the target process 27 notifying the hook filter function
26 of
25 the request from the control function 24. This message includes an index
pointer, 29b in Fig. 2, to the shared memory location in the DLL 25 where the
required information and arguments are available. The hook filter function 26
in
the target process 27, in step 14, receives the message in the address space
23 of
the target process 27, recognizes the message as a command from the control
_g_
CA 02342246 2001-02-28
WO 00113085 PCT/US99/19968
process 24, and uses the index pointer 29 to obtain the necessary information
and
arguments from the shared memory of the DLL 25, step 15.
The target process 27, in step 16, then makes calls to the operating system
21 accessing the pertinent window to perform the required operation. As a
result of this sequence, the control process 24 enjoys the same level of
access
within the target process 27 as does the target process 27 itself. The control
process 24 may thereafter perform any actions within the target process 27
which
the target process 27 itself could do. Thus, the message may direct the target
process 27 to, for example, move the insertion point, or retrieve text or data
from
1 o an edit control in the target process 27 and, optionally as in step 17,
write the
data to the shared memory of the DLL 25.
After the control process 24 has sent its initial message to the hook filter
function 26, it waits for notification from the target process 27 that it has
accomplished the directed operation. This notification, according to step 18
of
Fig. 1, may be accomplished in various manners such as by posting a message in
the message queue of the control process 24, or by changing the state of an
event
object which managed by the operating system 21 and monitored by the control
process 24. Any text or data from the target process 27 may be posted to the
shared memory of the DLL where it may be retrieved by the control process 24,
2o step 19 of Fig. 1. If the control process 24 no longer requires the use of
the hook
filter function 26, the hook filter function 26 may be removed from the target
process 27 by the control process 24. This may be accomplished by use of the
LtnhookWindowsHookEx function.
-9-
