Reexamination Certificate
1995-05-31
2001-03-27
Beausoliel, Jr., Robert W. (Department: 2413)
Reexamination Certificate
active
06206584
ABSTRACT:
A microfiche appendix of c language source code for the preferred embodiment (©1991 Reed Hastings) is filed herewith. A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.
BACKGROUND OF THE INVENTION
The present invention relates generally to a method and apparatus for modifying relocatable object files. In particular, the present invention relates to a method for inserting additional instructions and data into an existing relocatable object file of a computer program, for any purpose. Most particularly, this purpose is to monitor memory access by the computer program.
Despite the recent increase in CPU speeds and software complexity, most programmers continue to rely on development tools that were designed over fifteen years ago and that have not changed significantly since then. These development tools have serious inadequacies that exacerbate the difficulties of developing large, complex programs.
Problems with developing applications in C/C++ are often more serious than with other programming languages, but are fairly typical. C/C++'s pointer and memory management facilities make it difficult to build large, robust programs. Prudent C/C++ programmers currently hesitate to use many commercial object code libraries because they are worried they may lose weeks of time later on in tracking down wild-pointer bugs introduced by their particular use of a given library. The difficulty in tracking down these kinds of programming bugs and many others is directly tied to the manner in which executable code is created from source code and to the inadequacies of current development tools.
The process of transforming source code into “executable” code is, briefly, as follows. The source code for a typical computer program is divided into many files. Some of these files may contain high-level language code, such as C, C++, Pascal, Fortran, Ada, or PL1, and some may contain assembly language code. Each high-level language file is translated by a language-specific compiler into either a relocatable object file, or into an assembly language file. An assembler translates the assembly language files into relocatable object files. A linker merges all of the relocatable object files into a single executable program.
As programs get larger and more complex, they become more difficult to test and debug. If one wants to monitor or analyze aspects of a program's behavior, the current practice is to have the compiler output the extra instructions required to implement the desired monitoring. One example of this exists in many Pascal compilers; there is typically a way to request the compiler to output the extra instructions required to check array bounds at run time, and to signal an error if there is a violation. Another example exists in many Unix/C compilers; most compilers will, upon request, output extra instructions to record how many times each function was called.
The approach of having the compiler output the extra instructions required to implement a monitoring or analysis scheme is, however, flawed in at least three significant ways: First, modifying the compiler to output new sequences is difficult, and in practive, nearly impossible, because most programmers don't have the source code to the compiler. Second, recompiling all of a program's files just to get the extra instructions inserted can be very time consuming and wasteful. Finally, not all code goes through a compiler; some is written in assembly language and does not get the new instructions inserted into it. Thus, any monitoring which requires complete coverage to work correctly cannot be implemented through only the compiler.
Some of the most vicious development problems relate to the difficulty in finding and eliminating a large class of memory-access related errors. Among the most important memory-access related errors that a programmer a programmer needs top detect are array bounds violations, uninitialized memory reads, free memory access, and data changing strangely.
Array bounds violations (where an array is any collection of data contiguous in memory) occur on those occasions when a program reads or writes past the end, or before the beginning, of an array and accesses whatever datum happens to be in that memory location.
Uninitialized memory reads happen when a program allocates some memory for data storage, but fails to initialize it completely. Later, an uninitialized portion is read, unintentionally providing a random value, which might sometimes cause to the program to fail, and sometimes not.
Free memory access describes the situation where a program deallocates some memory but incorrectly continues to use it. If the program reallocates that memory for another purpose, then it will be using the same memory for two different purposes, and the program will probably perform incorrectly.
“Data changing strangely” is a bit of a catch-all expression. Often there are many ways to change a datum, especially a “global” datum. The programmer can have a difficult time discovering which function is changing the datum incorrectly, in a given run of the program. What the programmer needs is to have a monitoring program tell him or her whenever a specified datum changes (this is called a watchpoint).
A comprehensive way to monitor the execution of today's and tomorrow's programs, in particular their memory access, is clearly needed by the program developer.
SUMMARY OF THE INVENTION
According to one aspect of the invention, an object code file is expanded by inserting additional instructions and/or data between preexisting instructions and data, which may also be modified in some ways. A basically unlimited variety of additional instructions and data can be inserted for an equally wide variety of purposes. After the insertion step, the offsets in the file are checked and modified, if necessary, to reflect the new positions of the preexisting instructions and data, so that the expanded code will execute properly. In the preferred embodiment additional offsets in symbol tables, data relocation tables and instruction relocation tables are updated in the same general manner as the other offsets. The basic method is as follows:
An old object code table is formed in memory space, containing the preexisting instructions and data. Space is also allocated for other tables: a new object code table, an inter-item offset table, a forward index tables, and a forward control index table. For each item in the old object code table (whether instruction or datum), the following four steps are performed: (1) making a new code block comprising any desired additional instructions and/or data and the item, and storing it into the new object code table; (2) storing the location of the item within the new object code table into the forward index table; (3) storing the location of the new code block within the new object code table into the forward control index table; and (4), for items that contain inter-item offsets, storing the location within the old object code table, of the offset and the base from which it is measured, into the inter-item offset table. Then, for each pair of offset/base locations in the inter-item offset table, the offset stored in the new object code table is updated using the index tables. Finally, the offsets in any symbol tables, instruction relocation structures, or data relocation structures in the old object code file are updated so that the new offset refers to the location in the new object code table to where the item referred to was moved.
According to a second aspect of the invention, all or substantially all the memory accesses of a given program are monitored (not including the memory accesses for instruction fetch), for the purposes of pe
Beausoliel, Jr. Robert W.
Palys Joseph E.
Rational Software Corporation
Ritter, Van Pelt & Yi LLP
LandOfFree
Method and apparatus for modifying relocatable object code... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for modifying relocatable object code..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for modifying relocatable object code... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2521855