JPH1185571A - System for storing log into flash memory - Google Patents

Abstract

PROBLEM TO BE SOLVED: To shorten log time, to prolong the service life of a flash memory while effectively utilizing a log area and to remove an illegal log record when write is not normally finished during logging. SOLUTION: A CPU 1, flash memory device(FMEM) 2 and real-time clock device(RTC) 3 are connected through a bus 5, one log header and plural log records are provided on the FMEM 2 and when an event to be logged occurs, the log record composed of time stamp information, which is read out of the RTC 3 when log data information is logged, the type information of the log record and the record length information of the log record is written in the FMEM 2. When the log header and the log record are periodically referred to and the log is newly added, the log record information is read out of the FMEM 2 and outputted to an external storage device 4.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a logging system in an information processing apparatus, and more particularly, to logging for a flash memory device.

[0002]

2. Description of the Related Art Generally, flash memory devices (F)
The MEM) is a device that can perform erasing in blocks and writing in bytes, and can log events occurring in the system on this device.

Conventionally, a nonvolatile random access memory (nvRA) has been used as a device for an event log of a system.
M) is often used. However, FMEM has recently begun to be used because FMEM is cheaper and has a larger capacity.

[0004] Since the FMEM or nvRAM is a device that can be read and written directly by the CPU, when a catastrophic failure occurs that cannot guarantee the operation of the system, the highest priority interrupt processing (generally maskable by software) Since logging is performed immediately in interrupt processing), there is a high possibility that logging can be performed. The logged data is transferred to an external storage in a state where the operation of the system is stabilized after the system is restarted.

[0005] Compared to external storage, FMEM and nvRAM
Since the unit price is higher than the storage capacity, efficient recording is required.

[0006] FMEM has a lower storage capacity unit price than nvRAM.
There is a limit of 0 times, and the writing time takes 10 to 160 microseconds per byte. On the other hand, nvRAM has no write restriction, and writing for 1 microsecond is sufficient.

Therefore, in consideration of the life of the FMEM and the logging performance for the FMEM, it is required to further reduce the number of bytes to be written for the FMEM as compared to the nvRAM.

Conventionally, as shown in FIG. 4, 6-byte time stamp data of year / month / date / hour / minute / second is fixedly used as time stamp information to be logged. As a method of compressing the information of the time stamp, for example, a method described in Japanese Patent Application Laid-Open No. 4-257936 is known. Here, only the difference information from the previously generated time is recorded as the year / month / day / hour / minute / second information. For example, if the year, month, and day are the same as the previous time stamp, only the hour, minute, and second are recorded.

[0009]

By the way, Japanese Patent Application Laid-Open No.
According to the method described in Japanese Patent No. 257936, the time stamp data can be compressed because only the difference is recorded, but the original year / month / day / hour / minute / There is a problem that seconds cannot be easily returned. In other words, since there is no information indicating how many bytes the time stamp information has been compressed, it is difficult to easily identify where the time stamp information is and from where the next event information is. When separating information, for example, it is necessary to rely on a skilled person.

Further, in the case of FMEM, if the processing is interrupted in the middle of writing each log record due to a power failure or the like, an incorrect log record is generated. In other words, once in FMEM, 1 →
If it is written to 0, it will not return to 0 → 1 unless it is erased in block units.
There is a problem that a new log record over the record cannot be erased by overwriting, and as a result, an incorrect log record is generated.

SUMMARY OF THE INVENTION An object of the present invention is to provide a log storage method capable of compressing and easily expanding a time stamp to be logged and using FMEM efficiently.

Another object of the present invention is to provide a log storage method capable of removing a bad log when writing is interrupted in FMEM.

[0013]

According to the present invention, in the log record writing means, the time stamp data of the real time clock device (RTC) is compressed from the difference between the time stamp data of the previous log record and the time stamp data and the flash memory is written. The data is written to a device (FMEM), and the log record reading means expands the compressed time stamp data and writes the data to an external storage device. As a result, the log data on the FMEM can be compressed and efficiently logged while guaranteeing the same log data as the conventional one for the external storage. In addition, a logging flag is provided in the log record,
In the above log record writing means, at the start of logging, the record length is written with the logging flag set to 1, and at the end of logging, the logging flag is set to 0 at the end of logging, and the record length is rewritten. In the record reading means, the logging completion flag is set to 1
In the case of, the log record is ignored as a bad record. Thereby, the failure log can be removed.

[0014] The log area initializing means is an F which is not initialized at all during the power-on self test (POST).
Only done once when MEM is detected.

When an event to be logged occurs, the log record writing means performs FME in the highest priority interrupt processing.
A log is written to M. Due to the highest priority interrupt processing, a log can be left even when the operating system does not operate properly. At this time, the time stamp information is compressed.

The log reading means on the operating system periodically refers to the flash log area on the FMEM, and when a new log record is found, transfers the log record to the extended storage as appropriate. This is done when the operating system is running normally. At this time, the compressed time stamp information is expanded. Also, incorrect log records are not transferred.

The log record re-initializing means is also performed when the operating system is operating normally.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings.

Referring to FIG. 1, the illustrated information processing apparatus includes a central processing unit (CPU) 1, and the CPU 1 includes an interrupt terminal (SMI) 101 having the highest priority. The CPU 1 is connected to a flash memory device (FMEM) 2, a real-time clock device (RTC) 3, and an external storage device 4 via a bus (BUS) 5. As the external storage device 4, for example, a fixed disk device is used. In the illustrated example, the bus 5 is represented by a single bus for simplification, but the bus 5 may be composed of a plurality of buses.

Referring to FIG. 2, in the illustrated example, FMEM
2 has four banks 21, 22, 23, and 24, and block erase is performed in units of the banks. The log area is placed on one of the banks (bank 22 in the illustrated example). At the head of the bank 22, there is a 32-byte log header 221. Then, a log record group 222 is placed behind it. Log record group 222
Are written as a log record 2221 and a log record 2222 in order from the top.
It is written until. Next, when an event to be logged occurs, a log record is written at the head of the free area 2224.

FIG. 3 shows the log header 221 in detail. As described later, the log header 221 is first initialized by the log area initializing means. Log header identifier,
The header ID, major revision, and minor revision log area length always store the same value after being initialized by the log area initialization means. In the lastly deleted year, month, day, hour, minute, and second, the latest date when the bank 22 was deleted by the log area initializing means or the log area reinitializing means described later is stored. The number of times of erasure is increased by one each time the log area re-initialization means erases a bank.

FIG. 4 shows each conventional log record. The first byte is the record length and contains a value of 0-7fH.
Since the maximum record of the log is 128 bytes, 0 is entered at that time.

The record type contains a code of 1-FEH. This makes it possible to identify what kind of system event has occurred. Depending on the record code, there are more fixed-length detailed data or variable-length data in the log body, and there is no log body.

Conventionally, the time stamp year / month / date / hour / minute / second always has 6 bytes. Therefore, the 8 bytes shown in FIG. 5 always existed in the log record.

Conventionally, the time stamp is BCD (binary 1).
0). Therefore, paying attention to the fact that the upper 4 bits of the YY portion can only contain values from 0 to 9, YY is defined as follows.

Upper 4 bits of YY, 0 to 9: YY represents 0 to 99 in BCD notation, the same as the conventional format.

0Ah: The lower 4 bits of YY represent the elapsed time (0 to 15 seconds) from the previous log, and the log body starts from MM (offset 3).

In the case of 0Bh: lower 4 bits of YY and M
M is the elapsed time from the previous log (0 to 0ffffh seconds)
, And the log body starts from DD (offset 4).

In the case of 0Ch: lower 4 bits of YY and M
M and DD are elapsed times from the previous log (0 to 0fff)
ffh seconds), and the log body starts from hh (offset 5). hh, mm, and ss are omitted.

In the case of 0Dh: lower 4 bits of YY and M
M, DD, and hh are elapsed times (0 to 0) from the previous log.
ffffh second), and the log body starts from mm (offset 6). Thus, the record header part can be compressed.

The log area initializing means will be described with reference to FIG. This log area initialization means is executed by a power-on self test (POST).
First, it is checked whether the log header identifier and the header ID are valid (step s1).
Subsequent processing is skipped.

If it is not valid, in this example, since the log area is in the bank 22 of FMEM, the bank 22 of FMEM is block erased (step s2). By this block erasure, all the data in the bank 22 is initialized to 0FFh.

Next, the log header, major revision,
Write the minor revision log area length (step s3). This value is fixed in the system.

Next, the current time is read from the RTC (step s4), and the value is read from the last deleted year, month, day,
The data is written in the hour / minute / second area (step s5). Then, the number of times of block erasure is set to 0 (step s).
6).

The log record writing means will be described with reference to FIG. The log record writing means is C
Executed when the highest priority interrupt enters the PU.

First, the record type is checked in order from the log record at the head of the log record group, and a record type with a record type of 0FFh is searched. When searching for 0FFh, the time stamp of each log record is also checked, and how many seconds have elapsed since the first “YYMMDDhhmmss” is calculated for the last collected log (step s7). After finding the one of 0FFh, it is confirmed that the log area has enough space to write the current log record (step s8). If there is not enough free area, the overflow flag in the log header is set (step s9), and the process ends. If there is enough free area, the current time is read from the RTC (step s10), and the current time read from the RTC and the first "YYM" calculated at the time of the search for 0FFh are read.
Compressed timestamp data is created from the number of seconds elapsed from the last collected log from MDDhhmmss ”(step s11).

Here, the compression of the time stamp data will be described with reference to FIG.

First, it is confirmed whether or not this is the first log record (step t1). In the case of the first log record, a time stamp is given with "YYMMDDhhmmss" (step t2). Logs after that are logged with the time difference from the previous time. That is, if it is not the first log record, the difference between the time stamp of the last log and the current time is calculated (step t3). The time difference between logs is 15
If the time is within seconds (step t4), the time stamp data is compressed to one byte of 0Axh (step t5).
If the difference time ≧ 15 seconds, 1 hour, 8 minutes, 15 seconds (0ff
(fh seconds) or less (step t6), the data is compressed to 2 bytes of 0Bxxxh (step t7). Similarly, 12 days, 3 hours, 16 minutes, and 15 seconds (0Cxxxxh seconds)
When the time is within (step t8), 3 bytes of 0Cxxxxxxh (step t9), 3106 days 21 hours 24 minutes 15
When the time is within seconds (0xdxxxxxxh seconds) (step t10), the length is 4 bytes (step t11).

Referring to FIG. 7 again, after the time stamp data is compressed as described above, the log type of the current log is written. Next, log record length bi
The data with t7 set to 1 is written, and the compressed time stamp data is written (step s12). Thereafter, the log body is continuously written (step s1
3) Finally, a value in which bit 7 in the record is set to 0 is written again (step s14).

By the above operation, the log record length b
After setting it7 to 1, bit7 in the log record
This is effective for invalidating the log in the middle of writing if the power is turned off or stalls before rewriting to 0.

Next, FIG. 9 shows a log record reading procedure periodically executed during the OS operation.

Referring to FIG. 9, the log type is checked sequentially from the beginning of the log record group, and a search is performed until a log type of 0FFh is found (step s15).
If a log that has not yet been transferred to the extended storage device is found (step s16), the record length bit7 is checked (step s17), and only a valid log whose bit7 is 0 is extracted, and the compressed time is extracted. The stamp data is expanded to the original 6-byte time stamp data of "YYMMDDhhmmss" (step s1).
8) The entire log record is transferred to the extended storage device (step s19). Then, the next log is searched (step s20), and the same processing is performed.

After all the logs have been transferred, the free space in the log area is checked. If the free space falls below a certain level (for example, less than 30% of the total), the log re-initializing means is thereafter executed. Be executed.

The log re-initialization means will be described with reference to FIG.

First, the number of times the log header has been erased is read (step s21), and then the block is erased from the bank 22 in the same manner as the log initialization means (step s2).
2) Write the log header identifier, header ID, major revision, minor revision, and log area length (step s23).

Next, the current time is read from the RTC (step s24), and its value is written with the last erased year, month, date, hour, minute, and second (step s25), and is read first. A value obtained by adding 1 to the value of the number of erases is written (step s26).

[0047]

As described above, according to the present invention, the FM
Since the time stamp data is compressed and stored in the EM, the FMEM can be used efficiently.
There is an effect that the frequency of erasing the block of the MEM is reduced and the life of the FMEM can be extended.

Further, in the present invention, since the entire length of the log record is shortened, the time required for writing to the FMEM by the highest priority interrupt can be reduced.

Further, according to the present invention, even if the power is interrupted during the log and the writing is interrupted on the way due to other factors, the defective log record can be detected and removed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of a system using a log storage method according to the present invention.

FIG. 2 is a diagram showing a bank configuration, a log area configuration, and a log record group configuration in the FMEM shown in FIG.

FIG. 3 is a diagram showing a log header in detail.

FIG. 4 is a diagram showing a conventional log record in detail.

FIG. 5 is a diagram showing a time stamp.

FIG. 6 is a flowchart showing a process in a log area initialization unit.

FIG. 7 is a flowchart showing a process in a log record writing unit.

FIG. 8 is a flowchart for explaining time stamp compression in log record writing means.

FIG. 9 is a flowchart illustrating a process in a log record reading unit.

FIG. 10 is a flowchart showing a process in a log record reinitialization unit.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 Central processing unit (CPU) 2 Flash memory device (FMEM) 3 Real-time clock device (RTC) 4 External storage device 5 Bus (BUS)

Claims (3)

[Claims] An information processing system in which a central processing unit, a flash memory device storing log data, and a real-time clock device indicating current year / month / date / hour / minute / second are connected via a bus. In the system, the central processing unit has one log header and a plurality of log records on the flash memory device, and initializes all the log headers and log records on the flash memory device in blocks to store the log header in the flash memory device. Log area initializing means for writing the log data information when an event to be logged occurs, time stamp information read from the real-time clock device when the event was logged, time information of the log record, and record length information of the log record. Log record to the flash memory device Log record writing means for writing to a device, and a log for periodically reading the log record information from the flash memory device and outputting the log record information to an external storage device when a new log is added by periodically referring to the log header and the log record. A record reading means, and a log area re-initializing means for block initializing a log header and a log record area when a free area that can be logged after reading the log reaches a predetermined extent or less, and writing a log header to the flash memory device again. A method of storing logs in flash memory, which is characterized in that: 2. The log storage method according to claim 1, wherein the log record writing means converts the time stamp data of the real-time clock device from a difference from a time stamp of a previous log record. A log storage method in a flash memory, wherein data is compressed and written to the flash memory device, and the log record reading means expands the time stamp data and writes it to an external output. 3. The log storage method according to claim 2, wherein a log flag is provided in the log record, and the log record writing unit performs the logging at the start of logging. The record length is written with the medium flag set to 1, and at the end of logging, the logging flag is set to 0 and the record length is rewritten. A log storage method in a flash memory, wherein the log record is ignored as a record.