Logic analyzer having a disassembler employing symbol table information for identifying op-codes

Abstract

A logic analyzer according to the subject invention includes a disassembler for disassembling object code. The disassembler automates the process of locating op-code addresses by utilizing information derived from an object file corresponding to the code whose execution is being disassembled. The object file includes addresses for the starting location of op-codes corresponding to individual source code lines. When the disassembler cannot determine the correct starting location for an op-code, it uses information from the object file to obtain the address of an op-code within a specified range of interest.

Description

FIELD OF THE INVENTION

[0002] This invention relates to test and measurement systems and, more particularly, to a disassembler for use in a logic analyzer for more precisely determining the starting address of assembly language code statements running on a system under test.

BACKGROUND OF THE INVENTION

[0003] In earlier days, engineers typically wrote control code for test and measurement systems in assembly language, a low-level language. Today, however, to improve development time and to enable the creation of programs having greater functionality and complexity to be developed in shorter time, engineers typically employ high-level languages, for example C, C++, etc. The high-level language is converted into machine code for operation on the particular processor, by a compiler/assembly/linker system, which generates assembly language and/or machine code based on the high-level source code.

[0004] Unfortunately, when debugging the program, the engineer must work with the low-level machine code. The low-level machine code is a series of binary numbers, or hexadecimal codes (base 16) and is not easy to understand. Typically, the machine code is reconverted to assembly language by a disassembler program to aid the engineer to understand the code that he is viewing.

[0005] However, the engineer did not write the software in assembly language. Moreover, the engineer may be unfamiliar with the methods employed by the compiler to generate assembly language code corresponding to various source code statements. Thus, it becomes complicated for the engineer to understand the exact operation of the system during a debugging operation.

[0006] Substantially all computer programs must be debugged because of their complexity, and because human programmers make mistakes. Computer code that performed well in simulation may fail to perform adequately under “real world” conditions for a variety of reasons, such as the relative unpredictability of the timing and form of real world input signals. Traditional software debuggers perform disassembly of code residing on the system under test (also known as the target system) and aid in the troubleshooting of code that is not performing as expected. Disassembly of computer code is the process of converting machine code back into assembly language for ease of understanding by a programmer or engineer troubleshooting a problem. Software debugger based disassemblers typically have access to the target system memory by way of an emulator probe or serial connection, such as through a JTAG port. Using this access, they are able to perform disassembly based on the memory image in the target system, translating the code found there into higher-level assembly language.

[0007] Disassembler software can also be run on a logic analyzer, such as a TLA 700 Series Logic Analyzer, manufactured by Tektronix, Inc. of Beaverton, Oreg. Such a disassembler running on a logic analyzer is described in U.S. Pat. No. 6,016,556, entitled SYSTEM FOR IDENTIFYING AN ACQUISITION SAMPLE CORRESPONDING TO A SOURCE CODE STATEMENT, (Heath), issued 18 Jan. 2000, assigned to the same assignee as the subject invention, and hereby incorporated by reference. Disassembler software running on a logic analyzer translates low-level machine code, of the target system under test, into easy-to-understand higher-level language as a troubleshooting aid to the user of the logic analyzer. Unfortunately, a disassembler running in a logic analyzer does not have access to the memory of the system under test. It therefore cannot operate on an image of that memory to translate the code. Moreover, a logic analyzer cannot interrogate the system under test to determine the contents of a particular memory location. A logic analyzer can only monitor bus traffic on the target system, and attempt to interpret the binary data that it acquires from the bus.

[0008] In a logic analyzer, the process of disassembly involves converting acquired bus transaction records into a listing of assembly language instructions that, when executed, would have generated that bus activity. This conversion process is sometimes ambiguous as a result of a variety of circumstances such as insufficient transaction history, variable length instructions, and lack of an instruction fetch indicator. When any of these conditions occur, the conversion process will most likely suffer. That is, the conversion process sometimes makes an incorrect assumption about where a new assembly language instruction, known as an op-code, actually begins. The disassembler often mistakes data or address information for an op-code, and consequently incorrectly disassembles the given set of bus transactions, because the disassembler was pointed to an incorrect starting address.

[0009] To compensate for this ambiguity, disassemblers currently provide a mechanism known as “mark op-code”, which enables a user to manually indicate the correct starting address for an assembly language instruction. There are two significant problems with this solution. First, users must recognize when the displayed disassembly trace is incorrect and then figure out where it went wrong in order to know where to place the op-code mark. This can be confusing and error prone because the user is usually looking at unfamiliar code. Second, users must manually mark numerous such locations as they examine different parts of an acquisition trace. This manual marking procedure can be cumbersome and time consuming, especially because the user must frequently “jump” to new locations in the acquisition buffer that have not been previously disassembled. What is needed is a disassembler for a logic analyzer that more efficiently, more precisely, and automatically resolves most of these starting address ambiguities.

SUMMARY OF THE INVENTION

[0010] A logic analyzer according to the subject invention includes a disassembler for disassembling object code. The disassembler automates the process of locating op-code addresses by utilizing information derived from an object file corresponding to the code whose execution is being disassembled. The object file includes addresses for the starting location of op-codes corresponding to individual source code lines. When the disassembler cannot determine the correct starting location for an op-code, it uses information from the object file to obtain the address of an op-code within a specified range of interest.

BRIEF DESCRIPTION OF THE DRAWING

[0011] FIG. 1 is an illustration of a test setup, and for understanding the subject invention.

[0012] FIG. 2 shows a screen display from the test instrument of FIG. 1, overlaid by a screen display of the computer of FIG. 1, as known from the prior art.

[0013] FIG. 3 shows an acquisition buffer of the test instrument of FIG. 1, as known from the prior art.

[0014] FIGS. 4 a and 4b show function and variable symbol tables extracted from an object code file showing locations of functions and variables in accordance with the subject invention.

[0015] FIGS. 5 a and 5b show a source line symbol table extracted from an object file showing starting addresses of source line statements in accordance with the subject invention.

[0016] FIG. 6 is a simplified flow chart useful in understanding the subject invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Referring to FIG. 1, a logic analyzer 100 is coupled to a system under test 120 (also known as a “target system”, or simply “target”) by a cable and probe arrangement 110 for receiving binary signals corresponding to transactions appearing on the bus of the target system. One skilled in the art will quickly realize that cable and probe arrangement 110 is shown in a highly simplified manner, and in reality, may encompass a multitude of cables and probes coupled to hundreds of test points. Logic analyzer 100 may also be coupled to, and communicate with, a computer system 130, or may operate in a stand-alone mode.

[0018] Referring to FIG. 2, a screen display 200 of logic analyzer 100 is overlaid by a screen display 230 of computer 130. Individual portions of these screen displays are highlighted to show a correspondence of a line of assembly code 60 of a program running on target system 120 with a particular line of high-level source code 58 of that same program. It is important to note that the source code file was entered into computer 130 in the normal fashion, but the object code was disassembled in logic analyzer 100 from object code acquired from bus transactions of target system 120. That is, assembly code screen display 200 of logic analyzer 100 was built up over time by acquiring signals indicative of bus transactions and interpreting those bus transactions to be op-codes, addresses, and data.

[0019] A detailed discussion of the operation of disassembler software is beyond the scope of this application. However, a brief high level description with reference to FIGS. 1 and 3 may be appropriate, as an aid in understanding the subject invention. During disassembly, raw data is acquired from a system under test via probe and cable arrangement 110 and applied to an acquisition portion (i.e., front end) 300 logic analyzer 100, and stored in an acquisition buffer (i.e., an array of memory locations) 310. A pointer of the disassembler is pointed to a location 320 in buffer 310 and the contents of location 320 are read and analyzed. The contents of following memory locations may also be read out for use in the analysis of the contents of memory location 320.

[0020] Disassemblers include tables of op-codes and their corresponding hexadecimal representation (i.e., machine code). The disassembler analyzes the contents of memory location 320 based upon the characteristics of known op-codes. For example, an upper address byte and a lower address byte normally follow an op-code instructing a microprocessor to jump to a new location. Thus, if a binary word is acquired and stored in memory location 320, and is identical to the op-code value for a jump instruction, then the next two bytes 321322 will be examined. If the contents of the next two memory locations appear to constitute a valid address, then the analysis is confirmed. Consequently, the disassembler will assign that meaning to those three bytes, and move pointer to the next unanalyzed memory location 330.

[0021] As noted above, assembly code is written in human-readable form. Humans typically find it easier to relate to jump destinations and subroutines having names (i.e., labels) rather than relative addresses. For example, subroutine call to a subroutine labelled “light lamp” is easier to understand than a call to a subroutine at a particular numeric address. When code is assembled or compiled, resulting object code is generated. In addition, an object file called a symbol file is also generated and stored with the object code. The symbol file associates the label with the numeric address of the compiled or assembled code. Disassemblers make use of these symbol files to make the disassembled code more readable to humans by relabelling subroutines and variables after performing disassembly on the object code.

[0022] As already noted, the discussion above concerning the operation of disassemblers was necessarily simplified for ease of explanation. In fact, modern microprocessors do not fetch a single instruction at a time and do not necessarily operate on that instruction in the sequence in which it was fetched. For example, in an effort to increase operating speeds modern microprocessors fetch multiple instructions from memory at a time. It often happens that because of the outcome of a first instruction (perhaps a branch instruction wherein the branch was taken) the remaining fetched instructions will not be executed at all. Moreover, data resulting from an instruction may appear on the bus many cycles after the instruction that caused it to appear.

[0023] As noted above, conventional disassemblers operate on an image of the object code in memory, whereas a disassembler running on a logic analyzer does not have that advantage. Thus, a disassembler running on a logic analyzer is subject to ambiguities that are not experienced by a debugger based disassembler that has access to the memory image residing on the target system. Not only must the disassembler running on a logic analyzer translate the acquired data into assembly code, it must also sort out data that was placed on the bus of the target system out of sequence.

[0024] It is herein recognized that the symbol files can be used by a disassembler running on a logic analyzer to resolve address ambiguities encountered during disassembly of target system object code. That is, the information is currently extracted from object files and stored in an application symbol database for uses such as source code correlation and symbolic substitution, has not previously been used to assist in the disassembly process. It is also recognized herein that it may not be the address that is in question but rather whether the contents of that address represent an instruction (op-code) or data. The disassembly technique described herein improves on existing disassemblers used in logic analyzers by providing an alternate, automated, means for a disassembler running in a logic analyzer to identify resolve these ambiguities in a program running on a target system.

[0025] FIGS. 4 a, 4 b, 5 a, and 5b form the symbol file (presented in a human readable form for purposes of explanation). This is the information that logic analyzer 100 uses at run-time to associate addresses with variable names, function names or source code lines. The subject invention makes use of this data to aid disassembly when traditional techniques fail.

[0026] The subject invention will now be described with respect to FIG. 3 and the simplified flowchart of FIG. 6. One skilled in the art will understand that the disassembler of logic analyzer 100 performs such functions as look ahead, and applies sophisticated heuristics when attempting to determine the starting address of an op-code. Such functions are not central to the subject invention and a description of them has been omitted for simplicity of explanation. The routine of FIG. 6 is entered at step 600 wherein a binary word is read from memory location 330. Assume that the disassembler recognizes the binary word as an op-code for a call to a subroutine (step 610). A high byte and a low byte of an address of the subroutine should follow such an op-code. Assume, for purposes of explanation, that the memory location 331 expected to contain the high byte of a subroutine actually contains a data word placed on the data bus between the op-code and its address bytes 332333. The symbol database file includes addresses for the starting location of op-codes corresponding to individual source code lines. When the disassembler cannot determine the correct starting location for an op-code (step 620), it uses information from the object file (step 630) to obtain the address of an op-code within a specified range of interest. The disassembler then knows where the new instruction begins and is able to disassemble that section of the code correctly without user intervention. For example, the disassembler knows the starting addresses of all of the source statements. By reading in the next two bytes and comparing them to each of the starting addresses, it can identify the proper address for the subroutine call. If the address were resolved at step 620, then that address can be used directly at step 640. In either case, the routine advances to step 650 to determine if the disassembly is finished. If it is not finished, then the next data word is fetched step 660, and the routine begins again. If it is finished, the routine is exited at step 670.

[0027] What has been described is a novel disassembler for a logic analyzer, or the like, that eliminates much of the need for users to manually mark op-codes, and eliminates errors inherent in the manual process.

[0028] One skilled in the art will realize that use of the subject disassembler is not a complete replacement for the existing manual marking method because there will still be situations wherein no symbolic information exists for a given range of addresses. It is also possible that errors may exist in the symbol database that could lead to incorrect op-code identification. For these reasons, the new mechanism is viewed as an adjunct to the existing manual marking method, not as a replacement. Therefore, the subject invention includes the further feature that a user can override the results of the subject disassembler by manually marking incorrect op-codes.

Claims

1. A logic analyzer, comprising: an acquisition arrangement for acquiring signals representative of bus transactions from a system under test; said logic analyzer storing, analyzing, and interpreting said signals as object code statements of a program running on said system under test; said logic analyzer storing a symbol database relating to said program running on said system under test; said logic analyzer disassembling said object code statements into assembly code; and said logic analyzer using said symbol database to resolve ambiguities encountered in said disassembly of said object code.

2. The logic analyzer of claim 1, further comprising: a display unit for displaying said assembly code; and a data input for receiving data entered by a user for manually correcting a result of said disassembly of said object code.

3. A method for disassembling object code in a logic analyzer, comprising the steps of: acquiring signals representative of bus transactions from a system under test; storing, analyzing, and interpreting said signals as object code statements of a program running on said system under test; storing a symbol database relating to said program running on said system under test; disassembling said object code statements into assembly code; and using said symbol database to resolve ambiguities encountered during said disassembly of said object code.

4. The method of claim 3, further comprising the steps of: displaying said disassembled code on a display screen of said logic analyzer; and manually correcting an incorrectly disassembled line of assembly code.