JP2010061364A - Method and device for storing trace data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce a possibility of causing deficiency of useful trace data in a failure analysis in storing the trace data in a nonvolatile memory when a failure occurs with respect to a method and device for storing the trace data. <P>SOLUTION: In saving the trace data stored in a main storage device (RAM) 20 in a nonvolatile memory 30 when any failure occurs, a log management part 11 divides the trace data stored in the main storage device (RAM) 20 into blocks of a prescribed size, and successively traces the trace data from the block (VI) of the trace data of the site just before the point of time when the save instruction is received due to the occurrence of the failure to the block of the site at the past point of time, that is, writes the trace data in the nonvolatile memory 30 by block units in the order of V to IV to III to II to I. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a trace data storage method and apparatus. The present invention relates to, for example, a storage method and apparatus for storing trace data when a failure occurs in a nonvolatile memory.

WiMAX (Worldwide Interoperability for Microwave Access) of the access point and S3G ^(Super 3 rd Generation) base band of mobile phone system (BB: Baseband) such as card card, an external non-volatile memory, which limits the number of rewriting times there (flash memory, etc.) In a device used as a storage device (storage), trace data frequently rewritten, such as a program execution history and invalid processing history, is usually saved in a file format in a main storage device (RAM) which is a volatile memory.

  FIG. 6 shows how trace data is output and acquired. The central processing unit (CPU) 10 includes a task A, a task B, a task C,... That performs various processes, a monitoring task, and a log management unit 11, and each task A, task B, task C,. Requests the log management unit 11 to output trace data during task execution (6-1).

  In embedded Linux (registered trademark) or the like, a part of the main storage device (RAM) 20 is treated as a trace log use area, and the log management unit 11 outputs trace data to the trace log use area of the main storage device (RAM) 20. (6-2).

  When the developer / operator makes an acquisition request (6-3) to the log management unit 11 from the external control device 200 via the network, the log management unit 11 acquires the trace data as needed. A request is made to acquire trace data from the trace log use area of the storage device (RAM) 20 (6-4), and the trace data is returned to the external control device 200 as response data. The developer / operator uses the trace data for program problem analysis and the like.

  When a system program or the like fails in the apparatus 100 and the system is restarted, the storage area of the main storage device (RAM) 20 is initialized, so the trace data is stored in the nonvolatile memory 30 before restarting. The operation to save (copy) to is required.

  This will be described with reference to FIG. 7. For example, when a failure occurs during execution of task A (7-1), an abnormality is detected in the monitoring task, and it is determined that restart is necessary in the determination of whether or not restart is required (7-2) The monitoring task sends a trace data save (copy) command to the log management unit 11 (7-3).

  In response to the save (copy) command, the log management unit 11 sends an instruction to save (copy) the trace data from the main storage device (RAM) 20 to the nonvolatile memory 30 to the data transfer unit 40 (7-4). . The data transfer unit 40 transfers the trace data stored in the main storage device (RAM) 20 to the nonvolatile memory 30 in accordance with the instruction.

The storage management of trace data is known, for example, from Patent Document 1 below.
JP-A-5-134895

  By saving (copying) the trace data to the nonvolatile memory 30 when a failure occurs, the trace data is saved without being lost due to the failure, but the operation speed of the nonvolatile memory 30 is slow, and the trace data to the nonvolatile memory 30 is saved. It takes a long time to write Therefore, system initialization / restart will start before all trace data is saved (copied) due to the influence of faulty tasks executed at the time of failure and shortening of service stop time. There is a case. In other words, the transfer process by the data transfer unit 40 may stop halfway.

  In such a case, the internal data and file configuration information in the trace data storage area of the nonvolatile memory 30 are lost, and the trace data cannot be read normally. Even if some trace data can be saved, the trace data file is saved from the beginning of the trace data, so it is necessary to analyze the characteristic parts related to the fault occurrence and the fault analysis. Saving part data may fail.

  FIG. 8 shows a conventional sequence example of the trace data storing process before restarting. (A) of the figure shows the trace data storage at the normal time. That is, when a failure occurs in task A and the failure is notified to the monitoring task (8-1), the monitoring task sends a trace data save (copy) command to the log management unit (8-2), Start the guard timer.

  When the log management unit receives the save (copy) command, it sends a trace data save (copy) instruction to the data transfer unit (8-3). The data transfer unit starts to transfer the trace data from the main storage device to the nonvolatile memory (8-4), and when saving (copying) of all the trace data is completed, the log management unit finishes saving (copying). Notify (8-5).

  Upon receiving the notification of the end of the saving (copying), the log management unit notifies the monitoring task of the end of the saving (copying) (8-6). The monitoring task stops the guard timer (8-7), and sends a stop command to task A and the log management unit (8-8, 8-9). Thereafter, a restart process is performed (8-10).

  On the other hand, when the guard timer of the monitoring task times out while the data transfer unit is transferring the trace data from the main storage device to the nonvolatile memory as shown in FIG. In order to send a stop command to the log management unit (8-11, 8-12), the log management unit sends a stop command to the data transfer unit and forcibly terminates it (8-13). Therefore, saving (copying) is interrupted in the middle, and the entire trace data file is damaged or data is lost.

  Also, as shown in FIG. 5C, when the data transfer unit is transferring the trace data from the main storage device to the nonvolatile memory, when the task is abnormally terminated due to a task error or the like, a restart process is executed. . Accordingly, the log management unit and the data transfer unit are also initialized by the restart, and the entire trace data file is damaged or data is lost.

  Here, the characteristics of the trace data will be described. As shown in FIG. 9, the trace data is sequentially stored on the main memory (RAM) as a trace log as time passes. When the apparatus detects a failure and starts saving (copying) the trace data to the nonvolatile memory, the trace data 9-3 indicating the failure is earlier than the trigger of the save (copy) command 9-4 (in the trace log). Appears above). In addition, there is a possibility that trace data 9-1 and 9-2 indicating a failure sign are also acquired, and they appear earlier than the trace data 9-3 indicating a failure (above the trace log).

  Since the trace log is saved including the trace data not related to the failure, the location of the failure cannot be specified by the program when the save (copy) command 9-4 occurs. However, the trace data associated with the failure is the more trace data that appears earlier than the trigger of the save (copy) instruction 9-4 (above the trace log) and closer to the time when the save (copy) instruction 9-4 occurs. Is likely to be included. However, the trace log storage area is limited, and old trace data is saved as a separate file on the main memory (RAM) or erased from time to time, so it may have already been deleted. .

  The saving (copying) of the trace data to the non-volatile memory is normally performed from the trace data (above the save area) previously (early) saved in the main memory (RAM). Therefore, when trace data is saved (copied) only halfway, there is a high possibility that the trace data at the failure occurrence location is not saved (copied) and only the trace data before the failure is saved.

  An object of this invention is to preserve | save trace data with high possibility concerning a failure more reliably.

  In a first proposal, the trace data storage method is a trace data storage method in which the trace data stored in the main storage device is saved in a nonvolatile memory when a failure occurs. The trace data stored in the main storage device is stored in the trace data storage method. Divide the data into blocks of a predetermined size and trace the trace data in units of blocks from the block of the trace data immediately before the time when the save command due to the failure is received to the block of the part at the past time To write to.

  Further, in the second proposal, the trace data storage device is a trace data storage device that saves the trace data stored in the main storage device in a nonvolatile memory when a failure occurs, and the trace data stored in the main storage device. The data is divided into blocks of a predetermined size, and the trace data is traced back to the block in the block unit from the block of the trace data immediately before the time when the save command due to the failure is received to the block of the part at the past time. A log management unit for sending a command to write to the nonvolatile memory; a data transfer unit for writing the trace data stored in the main storage device into the nonvolatile memory in units of blocks according to the command sent from the log management unit; It is equipped with.

  It is possible to reduce the possibility of loss of useful trace data in failure analysis or the like.

  FIG. 1 shows an example of the disclosed trace data storage process. In the figure, a main memory (RAM) 20 is a memory that stores trace data as a trace log, and is a high-speed volatile memory. The nonvolatile memory 30 is a memory that saves and saves the trace data stored in the main storage device (RAM) 20 when a failure occurs.

  When the trace data stored in the main storage device (RAM) 20 is saved in the nonvolatile memory 30, the log management unit 11 divides the trace data into blocks of a predetermined size and saves them in the nonvolatile memory 30. Then, after the apparatus is restarted or when a trace data browsing request is input by the external control apparatus, the divided trace data is reconstructed into the original trace data.

  Upon receiving a save (copy) command from the monitoring task, the log management unit 11 generates the number of blocks and block information necessary for save (copy) based on the data size of the trace data. In the illustrated example, the number of blocks is 6, and block information I to VI is generated. However, at this time, there is no trace data of the block information VII after the save (copy) instruction.

  The block information includes a division number (No) indicating the number of divided blocks, the size of the trace data stored in each block, and the address position in the trace data file of the start position of the data stored in each block. including. Further, the block information includes a write flag indicating whether saving (copying) to the nonvolatile memory 30 is completed. In the illustrated example, the write flag is indicated by “ON” for a block that has been saved (copied) to the nonvolatile memory 30 and “OFF” for an incomplete block.

  The nonvolatile memory 30 stores block information (division number, data size) and division trace data for each block. The maximum data size per block is a set of sizes suitable for recording in the non-volatile memory 30 (for example, a size that can be written in a batch operation, usually one sector depending on the medium).

  In response to the generation of a trace save (copy) command, the log management unit 11 goes back to the past time point from the block immediately before the point in time, that is, in the order of block units in the trace data file head direction. Then, the trace data is written to the nonvolatile memory 30. In a specific example, writing is performed in the order of block VI → block V → block IV → block III → block II → block I.

  When saving (copying) to the beginning of the trace data file is completed, it is checked whether additional trace data has been saved since the save (copy) command was generated. (Number, data size) is assigned, and trace data from the portion immediately after the save (copy) command is generated to the end is saved (copied). In a specific example, block information VII is generated at this point, the number of flocks is added to 7, and the trace data of block VII is saved (copied).

  The reconstruction of the trace data in units of blocks stored in the nonvolatile memory 30 is performed after the restart or when the developer / operator browses the trace data. The trace data is reconstructed based on the block information by constructing the trace data for the reconfigurable part. If there is a missing part, a notification to that effect is output.

  FIG. 2 shows a sequence example of the divided storage process of the trace data before restarting. (A) of the figure shows an example of dividing and saving the trace data at the normal time. When a failure occurs in task A and the failure is notified to the monitoring task (2-1), the monitoring task sends a trace data save (copy) command to the log management unit (2-2), and the guard timer Start up.

  Upon receiving the save (copy) command, the log management unit divides the trace data into blocks (2-3), and sends a save (copy) command for each block (for example, 1 sector) of the trace data to the data transfer unit. Send out (2-4). The data transfer unit performs a process of saving (copying) one block of trace data from the main storage device to the nonvolatile memory 30 (2-5), and when the saving (copying) is completed, the saving (copying) is completed in the log management unit. (2-6).

  When the log management unit receives a notification of completion of saving (copying) in units of blocks from the data transfer unit, it sends a saving (copying) command for the next block, and similarly, saving to the last block (copying) The command is sent (2-7), and the data transfer unit performs the saving (copying) process of the last block (2-8), and when finished, notifies the log management unit of the end of saving (copying) (2-9) ).

  When the log management unit receives notification of the end of saving (copying) of the last block, it notifies the monitoring task of the end of saving (copying) (2-10). The monitoring task stops the guard timer and sends a stop command to the task A and the log management unit (2-11, 12-12). Thereafter, a restart process is performed (2-13).

  On the other hand, if the guard timer of the monitoring task times out during the transfer of the trace data from the main storage device to the non-volatile memory in units of blocks, the monitoring task is changed to Task A and As shown in FIG. In order to send a stop command to the log manager (2-14, 2-15), the log manager sends a stop command to the data transfer unit and forcibly terminates it (2-16).

  Therefore, the trace data of the block (sector) that was saved (copied) when the save (copy) process was interrupted halfway and timed out, and the subsequent blocks (sectors) will be lost, but immediately before that, The trace data traced back is normally stored in the nonvolatile memory.

  In addition, as shown in FIG. 5C, when the trace data is being transferred from the main storage device to the non-volatile memory in units of blocks, if the process is abnormally terminated due to a task abnormality or the like, a restart process is executed. The log management unit and the data transfer unit are also initialized by the restart. However, the trace data of the block (sector) that was saved (copied) at the time of restart and the subsequent block (sector) are lost, but the trace data traced back to the past immediately before that is normally non-volatile memory Saved in.

The trace data to be saved can be as follows.
(1) Start saving from trace data that has a high possibility of failure occurrence, and save all trace data.
(2) Start saving from trace data with a high possibility of failure occurrence, and save trace data for the specified number of blocks.
(3) Saving is started from trace data having a high possibility of failure occurrence, and trace data for the number of blocks that can be saved within a specified time is saved. However, this specified time is set to be shorter than the guard timer of the monitoring task.

  FIG. 3 shows an example of a flow of processing for storing all the trace data (1). When the saving (copying) processing unit of the log management unit receives a saving (copying) command from the monitoring task (3-1), the size of the trace data is measured and substituted into the variable LogSz1 (3-2). Then, the nonvolatile memory writing process (subroutine) is started (3-3).

  When starting the nonvolatile memory writing process (subroutine), the additional log size is passed as an argument as the trace data size LogSz1, the save (copy) start position as the head of the trace data file, and the processing order as the descending order.

In the nonvolatile memory writing process (subroutine), the number of additional blocks is calculated by the following calculation formula based on the size LogSz1 of the trace data (3-4).
BlkNUm = (LogSz1 + SZ-1) / SZ (rounded down after the decimal point) (Formula 1)
Here, BlkNUm is the number of blocks to be additionally written, and SZ is a data size (excluding block information) that can be stored in one block.

For example, when SZ = 100 and LogSz1 is 700, BlkNUm may be 7 when the number of additional blocks BlkNUm is described. At this time,
BlkNUm = (700 + 100-1) / 100 = 799/100 = 7 (fractions are rounded down).

If LogSz1 is 701 to 799, BlkNUm is 8. When LogSz1 = 701,
BlkNUm = (701 + 100-1) / 100 = 800/100 = 8.
When LogSz1 = 799,
BlkNUm = (799 + 100-1) / 100 = 898/100 = 8 (fractions are rounded down).
Therefore, the number of additional blocks BlkNUm is calculated by the calculation of (Equation 1).

  Next, block information is generated. The block information is generated by repeatedly executing the process (3-5) for generating and saving the block information for the number of additional blocks BlkNUm. Here, it is assumed that the block information is stored from the empty area of the block information table of the processing area in the log management unit in the upper order.

  Next, in order to calculate the number of blocks of the entire stored data including the trace data after the generation of the save (copy) instruction, the additional block number BlkNum is additionally added as the total block number TotalBlkNum (3-6).

  Next, the nonvolatile memory is written. For writing to the nonvolatile memory, the following processing (3-7 to 3-11) is repeatedly executed for the number of additional blocks BlkNUm. First, the write flag is checked (3-7). If “OFF” is set in the write flag, block write data editing is performed (3-8).

  Then, a save (copy) command is transmitted to the log transfer unit (3-9), and when the end of the save (copy) is received from the log transfer unit (3-10), the write flag of the block information table is set to “ON”. (3-11).

  When the above nonvolatile memory writing process (subroutine) ends, the process returns to the original main routine process. Then, the size of the trace data is measured again, and the size is substituted into the variable LogSz2 (3-12). This variable LogSz2 represents the data size including the trace data stored in the main storage device after the log management unit receives the save (copy) command from the monitoring task.

  Next, the variable LogSz2 and the variable LogSz1 are compared in size (3-13). If the variable LogSz2 is larger than the variable LogSz1, the trace data stored in the main storage device after the log management unit receives the save (copy) command from the monitoring task exists. Is saved (copied), the nonvolatile memory writing process (subroutine) is started again (3-14).

  At this time, as the arguments to be transferred to the nonvolatile memory writing process (subroutine), the additional log size is transferred (LogSz2-LogSz1), the save (copy) start position is (LogSz1 + 1), and the processing order is passed in ascending order. The nonvolatile memory writing process (subroutine) receives the argument and similarly executes the above-described processing flows 3-4 to 3-11.

  If the variable LogSz2 is not larger than the variable LogSz1, there is no trace data stored in the main storage device after the log management unit receives a save (copy) command from the monitoring task. do not do. Then, the monitoring task is notified of the end of saving (copying) (3-15).

  Next, FIG. 4 shows a flow example of processing for storing the trace data for the number of designated blocks (2) described above. In this process, saving is started from trace data having a high possibility of a failure occurrence site, and data corresponding to a predetermined number of designated blocks is saved.

  In the implementation of this processing, a write counter (Write_Counter) is provided in the log management unit, and when a save (copy) command is received from the monitoring task, the value of the write counter (Write_Counter) is set to 0 (4-1) Before performing the writing process to the nonvolatile memory, the value of the counter (Write_Counter) is compared with the designated number of blocks (MAX_BLK) (4-2).

  If the value of the counter (Write_Counter) does not exceed the specified number of blocks (MAX_BLK), after writing to the non-volatile memory, 1 is added to the write counter (Write_Counter) (4-3). If it exceeds, write processing is not performed. The other processing flow is the same as the processing flow described in FIG.

  Next, FIG. 5 shows a sequence example of processing for storing trace data for the number of blocks that can be stored within the specified time of (3) described above. In this process, saving is started from trace data having a high possibility of a failure occurrence site, and data corresponding to the number of blocks that can be saved within a predetermined designated time is saved. However, the specified time is a time shorter than the guard timer of the monitoring task.

  As shown in FIG. 5, for example, when a failure occurs in task A and the failure is notified to the monitoring task (2-1), the monitoring task sends a trace data save (copy) command to the log management unit. (2-2) Start the guard timer. The operation so far is the same as the operation described in FIG.

  Upon receiving the save (copy) command, the log management unit activates a timer for a specified time and generates (5-1) a child task. Similar to the processing described above, the child task divides the trace data into blocks (2-3), and sends a save (copy) command of the trace data in blocks to the data transfer unit (2-4).

  The data transfer unit performs a process of saving (copying) one block of trace data from the main storage device to the nonvolatile memory (2-5), and saving (copying) to the child task of the log management unit when the saving (copying) is completed. Is notified (2-6).

  When the child task of the log management unit receives a notification of completion of saving (copying) in units of blocks from the data transfer unit, it sequentially sends save (copying) instructions for the next block (2-7), and data transfer The section performs a saving (copying) process (2-8), and when finished, notifies the child task of the log management section of the end of saving (copying) (2-9).

  During this period, the log management unit monitors the timer for the specified time, and when all trace data has been saved (copied) by the completion of the timer, it sends a notice of completion of the save (copy) to the monitoring task. To finish the process. This case is equivalent to the flow of processing for storing all the trace data of (1) described above.

  When a timer timeout for a specified time occurs, the log management unit performs an interrupt process for the child task (5-2) and stops the save (copy) process. The child task transmits a stop command to the data transfer unit by the interrupt processing (5-3). Further, the child task performs recovery processing such as initialization of a part in the middle of writing to the data transfer unit (5-4).

  The stop process by the child task is not due to the timeout of the guard timer of the monitoring task for performing the restart process, but is a stop process by the invoking task, so there is a possibility that data destruction or the like has occurred in the part being written In some cases, after performing recovery such as clear processing, it is possible to send a save (copy) end notification to the log management unit.

  Note that in the above-described form of saving trace data for the specified number of blocks in (2), it is a stop process by its own task, not a stop due to a guard timer time-out of the monitoring task, and therefore data destruction during the writing process, etc. It is less likely that it will occur, making it possible to perform more reliable storage.

  When the log management unit receives a notification of the end of saving (copying) from the child task, it notifies the monitoring task of the end of saving (copying) (2-10). As in the case of FIG. 2, the monitoring task stops the guard timer and sends a stop command to task A and the log management unit (2-11, 12-12). Thereafter, a restart process is performed (2-13).

It is a figure which shows an example of the trace data preservation | save process of an indication. It is a figure which shows the example of a sequence of the division | segmentation preservation | save processing of the trace data before restart. It is a figure which shows the example of a flow of the process which preserve | saves all the trace data. It is a figure which shows the example of a flow of a process which preserve | saves the trace data for the designated number of blocks. It is a figure which shows the example of a sequence of the process which preserve | saves the trace data for the number of blocks which can be preserve | saved within the designated time. It is explanatory drawing of the output of trace data, and its acquisition. It is explanatory drawing of the operation | movement which saves trace data to a non-volatile memory before restarting. It is a figure which shows the example of the conventional sequence of the trace data preservation | save process before restart. It is explanatory drawing about the characteristic of trace data.

Explanation of symbols

10 Central processing unit (CPU)
11 Log management unit 20 Main memory (RAM)
30 Non-volatile memory 100 Device 200 Developer / operator external control device

Claims (4)

In a trace data saving method for saving trace data stored in a storage device to a nonvolatile memory when a failure occurs,
Dividing the trace data stored in the storage device into blocks of a predetermined size,
Trace data is written to the non-volatile memory in units of blocks from the trace data block immediately before the time when the save command due to the failure is received to the block at the previous time. How to save trace data. Predetermining a designated block number that is an upper limit value of the number of blocks to be written in the nonvolatile memory, and comparing and collating a counter value for counting the number of blocks written in the nonvolatile memory with the designated block number
When the value of the counter does not exceed the specified number of blocks, write the block unit trace data to the nonvolatile memory,
2. The trace data storage method according to claim 1, wherein when the value of the counter exceeds the specified number of blocks, the trace data in units of blocks is not written to the nonvolatile memory. Predetermining a specified time that is an upper limit value of the time to write to the nonvolatile memory, when receiving a save command due to the occurrence of a failure, start a timer that measures the elapsed time until the specified time,
2. The trace data storage method according to claim 1, wherein when the timer for the specified time has timed out before the writing to the nonvolatile memory is finished, the writing to the nonvolatile memory is stopped. In the trace data storage device that saves the trace data stored in the storage device to the nonvolatile memory when a failure occurs,
Trace data stored in the storage device is divided into blocks of a predetermined size, going back in order from the block of the trace data of the part immediately before the time when the save command due to the failure is received to the block of the part of the past time point, A log management unit for sending an instruction to write trace data to the nonvolatile memory in units of the block;
In accordance with an instruction sent from the log management unit, a data transfer unit that writes the trace data stored in the storage device to the nonvolatile memory in units of blocks;
An apparatus for storing trace data, comprising:

Images