JPH02230335A - Tracing system of farmware - Google Patents

Abstract

PURPOSE:To dynamically sample necessary information by rewriting the content of a memory area which a held address value shows and controlling a trace function when a service processor indicates the start or the termination of tracing. CONSTITUTION:When the indication of the start of tracing is detected, an address is read from a first holding area #n1, and an instruction stored in the address is read. Then, the instruction is written in a save area #g1, a first branch instruction branching into a first trace routine is read from a second holding area #p1..., and the first branch instruction is written into the address which is read from the first holding area #n1. When the first means detects the indication of the termination of tracing, or when the second means decides the stop of tracing, the instruction which has been written into the save area is read out, and is written into the area which the address held in the first holding area shows. Thus, the trace area can be efficiently used and trace data in much more states can be recorded.

Description

[Detailed Description of the Invention] [Summary] Regarding the collection of information for troubleshooting of firmware constituting at least a part of the hardware of a data processing device, there is a method that has little influence on the original operation of the firmware and dynamically The purpose is to provide a tracing means that can collect necessary information, and it has a save area where instructions are temporarily saved, a first holding area where the address of the firmware that starts the trace routine is held, and multiple types of trace routines. a second holding area for holding a plurality of branch instructions for branching to each branch; a plurality of types of trace routines; a first means for detecting when the service processor instructs to start or end tracing; and a second means for determining whether or not to stop subsequent tracing based on the contents of data collected by the trace routine, and writes the contents of the memory area indicated by the address value held in the first holding area. By changing the configuration, the trace function can be controlled.

[Industrial application fields] The present invention relates to a firmware tracing method.

In recent years, processing using firmware has been increasing, and with advances in semiconductor integration technology, control information and the like are often built into microprocessors and memories. Therefore, measuring instruments such as logic analyzers, which have been conventionally used for debugging hardware, are no longer usable. For this reason, it is essential for firmware to incorporate trace processing in order to debug itself. However, if the necessary trace data is not efficiently collected in the limited trace area, it may take a lot of time to analyze the trace data, or necessary trace data may not be recorded. .

Furthermore, during normal processing, processing by firmware affects the processing performance of the device, so it is necessary to prevent trace processing from delaying the original processing.

[Prior Art] Conventional tracing systems are configured to include an instruction to read a pointer and branch based on its value as an instruction to start tracing. When the trace routine is started to perform tracing, the pointer is rewritten to the start address of the trace processing routine, so that when the above instruction is executed, the trace processing is performed and then normal processing is executed. was.

FIGS. 4 and 5 are diagrams explaining such conventional tracing processing. That is, FIG. 4(a) shows the contents of the memory before and after starting tracing, and (8) shows the state before starting tracing, and (b) shows the state before starting tracing.
) shows the contents of memory when tracing is performed. Further, FIG. 5 shows the execution order of instructions, and FIG. 5(a) shows the 41st! This corresponds to I(a), and FIG. 4 et al.) corresponds to FIG.

Instruction 2 shown in FIGS. 4(a) and 4(b) is an instruction that reads the pointer and branches depending on its value. command 2
1 to 2n are trace processing routines. Instruction 2n is #
This is an instruction that branches to 3.

In the normal case (Figure 4(a)), the value of the #nm pointer is #3, so the trace processing routine is not executed as shown in Figure 5(a), and the next instruction 3.4...・・・・・・
Execute. On the other hand, when tracing (Fig. 4(b)
)) executes the trace processing routine as shown in Figure 5 etc. since the value of the pointer of #nffl is #2l,
Branches to #3 by instruction 2n and executes the next instructions 3, 4...
Execute...

[Problems to be Solved by the Invention] In the conventional trace data collection method as described above, even when there is no need to trace, a pointer is read and an instruction (instruction 2) that branches based on its value is executed. However, the time required for this process slows down the normal processing speed. Furthermore, since the address of the pointer (#nm) for starting the trace routine is always the same, only the same trace data can be obtained each time. Therefore, if the required trace data differs depending on the state, it is necessary to assume all the states and collect all the necessary trace data each time, which increases the amount of trace data. The problem is that it takes a lot of time to analyze.

Additionally, when the trace area becomes full, the trace data is rewritten from the beginning of the trace area, so if the amount of trace data increases, the trace data that was originally collected will be overwritten with new trace data, and the necessary trace data will not remain. There was also the problem that there were times when it was not available. This causes problems such as even if there is abnormal trace data, it may be overwritten by new trace data, making failure analysis impossible. was.

In view of these conventional problems, the present invention eliminates the delay caused by trace processing during normal processing, and
It is possible to execute trace processing only when the trace is started, and by determining the trace data and performing tracing efficiently, it is possible to record trace data corresponding to various conditions, and the trace data The purpose of this invention is to provide a means to facilitate failure analysis by stopping tracing when an abnormality is detected.

[Means for Solving the Problems] According to the present invention, the above objects are achieved by the means described in the claims.

That is, the first invention includes: a rewritable memory section that stores firmware and firmware trace results; an instruction processing section that reads firmware from the memory section and executes instructions; In a device that has a service processor capable of starting termination processing, a save area for temporarily saving instructions, a first holding area for holding the address of the firmware that starts the trace routine, and a plurality of storage areas are provided. a second holding area that holds multiple branch instructions that branch to different types of trace routines; and second means for determining whether or not subsequent tracing should be stopped based on the content of data collected by the tracing routine, when an instruction to start tracing is detected by the first means. , reads the address from the first holding area, reads the instruction stored at the address, writes the instruction to the save area, and branches from the second holding area to the first trace routine. A first trace routine can be executed by reading a branch instruction of the first holding area and writing the first branch instruction to the address read from the first holding area; When an instruction to end the trace is detected by the second means, or when it is determined to stop the trace by the second means, the instruction written in the save area is read out and stored in the first holding area. The second invention is a firmware tracing method that writes to an area indicated by an address, and a second invention adds means to the first invention to read and execute an instruction written in a save area during the processing of a trace routine. Furthermore, the third invention adds a third means for determining a subsequent change in the type of tracing based on the content of the collected trace data to the first and second inventions, and furthermore, the third invention adds a third means for determining a subsequent change in the type of tracing based on the content of the collected trace data, In order to change the type of trace to be taken next time by the third means according to the execution result of the process, the instruction at the address of the first holding area is branched to the second tracing routine of the second holding area. This is a firmware tracing method that rewrites the branch instruction to a second branch instruction.

[Operation] In the method of the present invention, the function of collecting trace data is realized by rewriting firmware instructions.

Therefore, when tracing is not activated, there is no time consumption due to trace-related processing, so there is no need to delay normal processing as in the past.

In addition, by determining the content of the collected data, it is possible to prevent unnecessary trace data from being collected thereafter, so the trace area can be used efficiently and more necessary trace data can be recorded. Is possible. Furthermore, failure analysis can be facilitated by detecting an abnormality in trace data and stopping the trace.

[Example] Figures 1 to 3 are diagrams for explaining an example of the present invention, in which Figure 1 shows the contents of the memory, and (a) shows the contents of the memory before starting tracing. (b) shows the state after tracing.

Moreover, FIG. 2 shows the order of execution of instructions, and (a)
(b) shows the state during normal operation, and (b) shows the state during tracing.

FIG. 3 shows a flowchart illustrating the control of the trace routine.

The operation of the embodiment will be described below based on these figures.

In the figure, RS represents a memory address (indicated by #), the contents are the contents of the memory at that address, and instructions 1, 3, 4, etc. are instructions to be executed during normal processing. , instruction 21. 22. ~2n is the first trace routine, instructions 31, 32. ~3n is the second trace routine, #nl is the first holding area, #P1. #P2...
. . . indicates the second holding area, and Itql indicates the save area. As mentioned above, FIG. 3 shows the flowchart of the trace routine. In the figure, condition A shows the value of the trace data when the firmware does not change from its current state, and condition B shows the value of the trace data when the firmware is in the current state. It shows the value of trace data when the armware changes from its current state.

A trace start interrupt is generated from the service processor to the instruction processing unit that activates the start of trace, and (1) the address (#2) for starting the first trace routine is read from the first holding area (#nl).

(2) Read the instruction (instruction 3) from the read address (#2).

■Write the read instruction (instruction 3) to the save area ($ql).

(2) Read the first branch instruction (instruction pi) that branches to the first trace routine (instructions 21 to 2n).

Write the first branch instruction (instruction pi) read in ■■ to the address (#2) read in ■■.

and ends the interrupt processing.

In the case of normal processing (FIG. 1(a)), the trace processing routine is not executed and the next instruction 1. 3. 4. ．． Execute. On the other hand, when taking a trace (FIG. 1A), since instruction 3 has been rewritten to instruction pi, instruction 1 is executed, and then instruction pi is executed, thereby branching to #21. Then, command 21 of the first trace routine (flowchart shown in FIG. 3). 22. ．． ．． Execute and collect trace data.

Then, conditions A and B are used to determine whether the state of the firmware has changed. , and depending on the result,
If the trace data matches condition 8, the current state is the same, and the instruction (
Instruction 3'' is read out and executed, and a branch instruction (instruction 2n) that branches to the original routine is executed to end the trace routine. When instruction 2n is executed, the process branches to #3 and instruction 4 is executed.

If the trace data matches condition B, it means that the current state has changed, so the second holding area (lI
P2) reads the second branch instruction (instruction p2) that executes the second trace routine (instructions 31 to 3n), and stores this instruction (instruction p2) in the address (#2) of the first holding area (#il).
Write the instruction p2).

Then, the instruction (instruction 3) is read and executed from the save area ($ql), and the branch instruction (instruction 2n) that branches to the original routine is executed, and the trace routine ends.

When instruction 2n is executed, the process branches to #3 and instruction 4 is executed. Next, when p2 is executed, instruction p2 is executed and the second trace routine is executed.

If the trace data does not match any of the conditions, it is abnormal trace data, so read the instruction (instruction 3) from the save area (lql) and transfer the instruction (instruction 3) to the address of the first holding area ($nl). By writing this instruction (instruction 3) to (#2), even if #2 is executed next, instruction 3 will be executed and tracing will be stopped.

When the service processor instructs the trace to end,
A trace end interrupt is generated in the instruction processing unit, and (1) the address (#2) for starting the trace routine is read from the first holding area ($nl>).

■Read the instruction (instruction 3) written in ■■.

Write the instruction (instruction 3) read in ■ to the address (#2) read in ■■.

The above operations complete the interrupt processing.

In this state, even if #2 is executed next, instruction 3 will be executed and the trace will be in a stopped state.

[Effects of the Invention] As explained above, according to the present invention, when tracing is not started, processing for collecting trace data is not executed, so time is not wasted.
This has the advantage of not delaying normal processing.

Furthermore, when tracing is activated, the trace area can be used efficiently and trace data of more states can be recorded. Furthermore, by detecting an abnormality in the trace data and stopping the trace, failure analysis can be facilitated.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the contents of a memory according to an embodiment of the present invention;
FIG. 2 is a diagram showing the order of execution of instructions according to an embodiment of the present invention;
FIG. 3 is a flowchart showing the control of the trace routine, FIG. 4 is a diagram showing the contents of the memory before and after the start of tracing in the conventional system, and FIG. 5 is a diagram showing the order of execution of instructions in the conventional system.

Claims (1)

[Claims] 1. A rewritable memory unit that stores firmware and firmware trace results, an instruction processing unit that reads firmware from the memory unit and executes instructions, and a method for starting tracing by the instruction processing unit or In a device that has a service processor capable of starting termination processing, there are multiple types of storage areas, including a save area for temporarily saving instructions, and a first holding area for holding the address of the firmware that starts the trace routine. a second holding area that holds multiple branch instructions that each branch to a trace routine; multiple types of trace routines; and a first means for detecting when a service processor instructs the start or end of tracing. and second means for determining whether or not subsequent tracing should be stopped based on the content of the data collected by the tracing routine, and when an instruction to start tracing is detected by the first means, A first branch that reads the address from the first holding area, reads the instruction stored at the address, writes the instruction to the save area, and branches from the second holding area to the first trace routine. A first trace routine can be executed by reading an instruction and writing the first branch instruction to the address read from the first holding area, and the trace routine is executed by the first means. When an end instruction is detected or when it is determined by the second means to stop tracing, the instruction written in the save area is read out and the area indicated by the address held in the first holding area is read out. A firmware tracing method characterized by writing to. 2. The firmware tracing system according to claim 1, further comprising means for reading and executing instructions written in the save area during the processing of the tracing routine. 3.Equipped with a third means for determining whether to change the type of trace thereafter based on the content of the collected trace data, and depending on the execution result of the trace routine processing, the third means determines the next trace to be taken. To change the type, the instruction at the address of the first holding area is branched to the second trace routine of the second holding area. 3. The firmware tracing method according to claim 1, wherein the firmware tracing method is rewritten to a second branch instruction.