JP2007265157A - System and method for detecting fault of i/o device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an abnormality detection system of an I/O device capable of executing processing according to the abnormality of the I/O device even when operation of a processor stops. <P>SOLUTION: An external storage device management means 101 counts the number of normal terminations and the number of abnormal terminations of read/write of a magnetic disk device 130. The external storage device management means 101 compares the number of normal terminations and the number of abnormal terminations that are counted with a predetermined threshold, and determines whether abnormality occurs in the magnetic disk device 130. A watch dog timer management means 102 periodically resets a watch dog timer 111 when the magnetic disk unit 130 normally operates. When the watch dog timer 111 generates time out, a BMC (Base Management Controller) 110 generates NMI interruption and informs a processor 100 that the magnetic disk unit 130 is abnormal. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a failure detection system and method for an external device, and more particularly to a failure detection system and method for an external device that detects that a failure has occurred in an external device such as an I / O device.

  In general, a computer system includes a processor that performs system control and an I / O device such as an external storage device, and the processor and the I / O device are connected via a bus. As a method for detecting a failure of an I / O device in such a computer system, for example, there is a technique described in Patent Document 1. In this technique, occurrence of an I / O timeout is detected, and an I / O device that frequently generates an I / O timeout among a plurality of I / O devices is detected as a fixed failure.

JP-A-5-334205

  In the conventional computer system, for example, when the magnetic disk becomes unresponsive, the OS cannot be operated, and the user needs to perform a hard reset. Alternatively, when the OS becomes unresponsive for a certain time, the watchdog timer is used for resetting. However, when a failure is detected by detecting a non-response of the magnetic disk, the failure may be erroneously detected due to intermittent non-response of the magnetic disk, and an unnecessary reset may occur. There is also a problem that normal interrupt processing does not operate when the processor stops in a state in which interrupts are prohibited.

  The present invention provides an I / O device abnormality detection system and method that solves the above-described problems of the prior art and that can execute processing according to an abnormality of the I / O device even when the operation of the processor is stopped. For the purpose. It is another object of the present invention to provide an I / O device abnormality detection system and method that can prevent erroneous detection of an abnormality of an I / O device.

  In order to achieve the above object, a failure detection system for an I / O device according to the present invention is a failure detection system that detects that a failure has occurred in an I / O device. Monitors the data input / output between the I / O device and the timer means for generating a count-out when a predetermined value is reached, and counts the number of normal data input / output ends and the number of abnormal data input / output ends. An I / O device management means for judging whether or not the I / O device is operating normally based on at least one of the counted number of normal ends and the number of abnormal ends, and the I / O device When the management means determines that the I / O device is operating normally, timer management means for resetting the count value of the timer means to an initial value, and the timer means time-out. If raw, characterized in that it comprises an interrupt generating means for generating an NMI interrupt.

  The failure detection method of the present invention is a failure detection method for detecting that a failure has occurred in an I / O device. The output is monitored to count the number of normal terminations of data input / output and the number of abnormal terminations of data input / output. Based on at least one of the counted number of normal terminations and abnormal terminations, the I / O It is determined whether or not the device is operating normally. If it is determined that the I / O device is operating normally, the timer count value is reset to an initial value, and the timer count is a predetermined value. Then, an NMI interrupt is generated.

  In the I / O device failure detection system and method of the present invention, it is determined whether or not the I / O device is operating normally based on at least one of the normal termination number and abnormal termination number of data input / output. If it is determined that the timer is operating, the timer count is reset to the initial state. When an abnormality occurs in the I / O device, the timer is not reset, a timeout occurs, and the abnormality of the I / O device can be detected by the NMI interrupt on the processor side that inputs this interrupt. In the present invention, since at least one of the normal termination number and the abnormal termination number of data input / output is used for determining whether or not the I / O device is operating normally, an appropriate determination should be made based on this. Therefore, it is possible to determine whether the abnormal termination has just occurred accidentally or whether the apparatus has malfunctioned and the abnormal termination has occurred, and it is possible to prevent erroneous detection of a recoverable error. Here, if the I / O device becomes non-responsive, the means (processor) for counting the number of normal terminations and abnormal terminations may not operate at all. In such a case, the timer is not reset. When a time-out occurs and an NMI interrupt occurs, the processor can detect an abnormality in the I / O device and execute processing corresponding thereto.

  In the failure detection system for an I / O device according to the present invention, the timer means includes a watchdog timer that counts down a count value from a predetermined initial value as time elapses, and generates a timeout when the count value becomes 0. Configuration can be adopted. In this case, before the count value becomes 0, the I / O device management unit determines that the operation of the I / O device is abnormal, or the operation of the I / O device management unit itself is caused by some reason. By stopping, when the timer control means does not issue a reset, the countdown proceeds, the count value becomes 0, an NMI interrupt is generated, and an abnormality in the I / O device can be detected.

  In the failure detection system for an I / O device of the present invention, the I / O device management means determines that the I / O device is not operating normally when the number of abnormal terminations exceeds a predetermined threshold value. Can be adopted. For example, for a predetermined number of times of data input / output, a threshold value is set as a reference for determining that the I / O device is abnormal, and depending on whether the number of abnormal terminations exceeds the threshold value A configuration for determining whether or not the I / O device is abnormal can be employed.

  In the I / O device failure detection system according to the present invention, the I / O device management means determines that the I / O device is operating normally when the number of normal terminations exceeds a predetermined threshold. Can be adopted. For example, for a predetermined number of times of data input / output, a threshold value serving as a reference for determining that the I / O device is normal is set, and depending on whether the normal end count exceeds the threshold value A configuration for determining whether or not an I / O device is normal can be employed.

  In the failure detection system for an I / O device according to the present invention, the number of abnormal terminations includes the number of I / O timeout occurrences, and the I / O device management means includes the number of I / O timeout occurrences and the number of normal terminations. If the ratio exceeds a predetermined value, a configuration can be adopted in which it is determined that the I / O device is not operating normally. For example, a threshold value of the ratio between the number of I / O timeout occurrences and the number of normal terminations, which serves as a criterion for determining that an I / O device is abnormal for a predetermined number of data inputs / outputs, A configuration is adopted in which it is determined whether or not the I / O device is abnormal depending on whether the ratio between the number of I / O timeout occurrences and the number of normal terminations exceeds the threshold. it can.

  In the I / O device failure detection system according to the present invention, the I / O device may include an external storage device. Further, the I / O device may employ a configuration including a network device.

  In the I / O device failure detection system and method of the present invention, it is determined whether or not the I / O device is operating normally based on at least one of the normal termination number and abnormal termination number of data input / output. If it is determined that the timer is operating, the timer count is reset to the initial state. When an abnormality occurs in the I / O device, or when the operation of the processor is stopped due to no response from the I / O device, the timer is not reset, a timeout occurs, and an NMI interrupt is generated. . By inputting this NMI interrupt to the processor, the processor can execute processing according to the abnormality of the I / O device. In the present invention, since at least one of the normal termination number and the abnormal termination number of data input / output is used to determine whether the I / O device is operating normally, an appropriate determination is made based on this. By doing so, it can be determined whether the abnormal termination has just occurred accidentally or whether the apparatus has malfunctioned and the abnormal termination has occurred, and it is possible to prevent erroneous detection of a recoverable error.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows the configuration of an I / O device failure detection system according to an embodiment of the present invention. The failure detection system 10 includes a processor 100, a BMC (Base Management Controller) 110, and a SCSI controller 120. The processor 100 is connected to the SCSI controller 120 via the I / O bus 150. The SCSI controller 120 is connected to the magnetic disk device 130 via the SCSI bus 140.

  The processor 100 includes an external storage device management unit 101 and a watchdog timer management unit 102. The BMC 110 has a watch dog timer 111. Each means in the processor 100 and the BMC 110 is realized by a program operation. The external storage device management means 101 counts the number of normal and abnormal reads / writes with the magnetic disk device 130 and stores it in a memory (not shown), and starts up the processor 100, operates the system, When the system is shut down, the normality of the external storage device (magnetic disk device) 130 is confirmed.

  The watchdog timer management unit 102 issues a reset of the watchdog timer 111 to the BMC 110. When the watchdog timer 111 times out, the BMC 110 generates an NMI interrupt (non-maskable interrupt) 160 to the processor 100 and notifies the processor 100 that the magnetic disk device 130 is abnormal.

  FIG. 2 shows an operation procedure of the processor 100. The processor 100 issues a reset of the watchdog timer 111 by the watchdog timer management means 102 when the processor 100 is started or when an external storage device management command is issued, and resets the watchdog timer 111 to an initial state (step A1). The external storage device management unit 101 reads / writes data from / to the magnetic disk device 130 via the I / O bus 150, the SCSI controller 120, and the SCSI bus 140 (step A2). The external storage device management unit 101 determines whether or not the read / write has been normally completed (step A3). If the read / write has been completed normally, the normal number addition processing is performed to calculate the total of normal termination (step A4). .

  When the external storage device management unit 101 determines that the read / write has ended abnormally, it determines whether the abnormal end has occurred due to a timeout (step A5). When the external storage device management means 101 determines that the time-out is not caused, it performs an abnormal number addition process and calculates the cumulative number of abnormal terminations (step A6). If it is determined that the error occurred due to a timeout, a timeout number addition process is performed to calculate the total number of timeout occurrences (step A7). The external storage device management unit 101 determines whether or not read / write has been performed N times (step A8), and when it has not been performed N times, returns to step A2. As a result, the cumulative number of normal terminations, the cumulative number of abnormal terminations, and the cumulative number of timeouts when N times of reading and writing are performed are obtained.

  When the processor 100 executes read / write N times, the external storage device management means 101 performs normality determination processing (step A9) and determines the normality of the result (step A10). In step A9, for example, if the cumulative number of abnormal terminations exceeds a predetermined threshold value, it is determined that there is an abnormality. Further, it is determined whether or not it is normal based on the ratio between the cumulative number of timeouts and the cumulative number of normal terminations. Specifically, when the timeout number / normal termination number exceeds a predetermined threshold value, it is determined that there is an abnormality.

  If the external storage device management means 101 determines that it is normal at step A10, it performs a sleep process (step A11) and waits for a predetermined time, and then returns to step A1, and the watchdog timer management means 102 uses the watchdog timer. 111 is reset. Thereafter, Step A2 to Step A10 are executed. If it is determined that there is an abnormality, the processing is stopped (step A12).

  FIG. 3 shows an operation procedure of the BMC 110. The watchdog timer 111 resets the count value WDT to an initial value when the watchdog timer management means 102 issues a reset in step A1 of FIG. 2 (step B1). The count value WDT of the watchdog timer 111 is decremented by 1 (step B2), and it is determined whether or not the count value WDT has become 0 (step B3).

  If the count value WDT has not yet become 0, the process returns to step B2, and one more count value WDT is subtracted. As a result, the count value WDT approaches 0 over time. When a reset is issued by the watchdog timer management unit 102, the process proceeds to step B1 and the count value WDT is set to an initial value. When the count value WDT becomes 0, an NMI interrupt 160 is generated to notify the CPU that a timeout has occurred (step B4).

  In the failure detection system 10, if it is determined in step A10 in FIG. 2 that the operation of the magnetic disk device 130 is normal, the watchdog timer 111 is reset. On the other hand, if it is determined in step A10 that it is not normal, the process proceeds to step A12, and the process is stopped so that the watchdog timer 111 is not reset. For this reason, when a failure occurs in the magnetic disk device 130, the watchdog timer 111 times out and the NMI interrupt 160 is input to the processor 100. Further, even when the OS becomes inoperable, the processing shown in FIG. 2 is stopped, so that the watchdog timer 111 times out and the NMI interrupt 160 is generated.

  In the present embodiment, when the magnetic disk device 130 operates normally, the processor 100 periodically resets the watchdog timer 111, and when an abnormality occurs, the watchdog timer 111 is not reset so that a timeout occurs. Is generated. Further, when the magnetic disk device 130 becomes unresponsive, the processor 100 may not be able to operate. However, even in this case, a timeout occurs because the watchdog timer 111 is not reset. The BMC 110 inputs an NMI interrupt 160 to the processor 100 when the watchdog timer 111 times out. As a result, when an abnormality occurs in the magnetic disk device 130 or when the processor 100 cannot operate, the processor 100 can be notified that the magnetic disk device 130 is abnormal. In addition, the number of normal read / write terminations, the number of abnormal terminations, and the number of I / O timeout occurrences are observed, and based on these, it is possible to recover by determining whether or not the magnetic disk device 130 is operating normally. It is possible to prevent erroneous detection of errors.

  In the above embodiment, the magnetic disk device 130 is taken as an example of an I / O device that is a target of abnormality detection, but is not limited thereto. FIG. 4 shows the configuration of an abnormality detection system according to a modification of the present invention. The input / output management unit 101a of the abnormality detection system 10a of the modification corresponds to the external storage device management unit 101 of FIG. 1, and the I / O controller 120a corresponds to the SCSI controller 120. The magnetic disk device 130 and the LAN controller 180 are connected to the I / O controller 120a via the bus 170. The I / O device connected to the I / O controller 120a is not limited to the magnetic disk device 130 or the LAN controller 180, and can be various I / O devices.

  As described above, the present invention has been described based on the preferred embodiment. However, the abnormality detection system and method of the I / O device of the present invention are not limited to the above embodiment. Those in which various modifications and changes have been made to the configuration are also included in the scope of the present invention.

1 is a block diagram showing a configuration of a failure detection system for an I / O device according to an embodiment of the present invention. 4 is a flowchart showing an operation procedure of the processor 100. The flowchart which shows the operation | movement procedure of BMC110. The block diagram which shows the structure of the failure detection system of the I / O apparatus of the modification of this invention.

Explanation of symbols

10: Failure detection system 100: Processor 101: External storage device management means 102: Watchdog timer management means 110: BMC (Base Management Controller)
111: Watchdog timer 120: SCSI controller 130: Magnetic disk device 140: SCSI bus 150: I / O bus 160: NMI interrupt

Claims (8)

In a failure detection system that detects that a failure has occurred in an I / O device,
Timer means for proceeding counting over time and generating a count-out when the count value reaches a predetermined value;
Monitor data input / output with the I / O device, count the number of normal data input / output ends and the number of abnormal data input / output ends, and count the number of normal ends and abnormal end counts I / O device management means for determining whether or not the I / O device is operating normally based on at least one of the following:
Timer management means for resetting the count value of the timer means to an initial value when the I / O apparatus management means determines that the I / O device is operating normally;
An I / O device failure detection system comprising: an interrupt generation means for generating an NMI interrupt when the timer means generates a timeout.   2. The I / O device failure according to claim 1, wherein the timer means includes a watchdog timer that counts down a count value from a predetermined initial value as time elapses, and generates a timeout when the count value reaches 0. Detection system.   3. The I / O device according to claim 1, wherein the I / O device management unit determines that the I / O device is not operating normally when the number of abnormal terminations exceeds a predetermined threshold value. Device failure detection system.   3. The I / O device according to claim 1, wherein the I / O device management means determines that the I / O device is operating normally when the number of normal terminations exceeds a predetermined threshold value. 4. Device failure detection system.   The number of abnormal terminations includes the number of I / O timeout occurrences, and the I / O device management means, when the ratio between the number of I / O timeout occurrences and the number of normal terminations exceeds a predetermined value, The I / O device failure detection system according to claim 1, wherein the O device is determined not to operate normally.   The I / O device failure detection system according to claim 1, wherein the I / O device includes an external storage device.   The I / O device failure detection system according to claim 1, wherein the I / O device includes a network device. In a failure detection method for detecting that a failure has occurred in an I / O device,
While progressing the count of the timer as time passes,
Monitor data input / output with the I / O device, and count the number of normal data input / output ends and the number of abnormal data input / output ends,
Based on at least one of the counted normal end number and abnormal end number, it is determined whether or not the I / O device is operating normally,
When it is determined that the I / O device is operating normally, the count value of the timer is reset to an initial value,
A failure detection method, wherein an NMI interrupt is generated when the timer count reaches a predetermined value.

Images