CN1294488C

CN1294488C - Starting-up switching method of multi-processor computer system

Info

Publication number: CN1294488C
Application number: CNB200310124031XA
Authority: CN
Inventors: 李俊良
Original assignee: Inventec Corp
Current assignee: Inventec Corp
Priority date: 2003-12-31
Filing date: 2003-12-31
Publication date: 2007-01-10
Anticipated expiration: 2023-12-31
Also published as: CN1635472A

Abstract

The present invention relates to a start-up switching method of a multiprocessor computer system, which solves the problem of abnormal start-up of the multiprocessor computer system by a base board management controller (BMC). The present invention can switch to an available central processing unit (CPU) and an available basic input-output system (BIOS) to start a computer by a CPU switching program or an ROM switching program according to a using state of the central processing unit (CPU) of the computer system or a using state of the basic input-output system (BIOS) stored in a read only memory (ROM). The CPU switching program switches between a start-up CPU and at least one application CPU by the BMC to switch a start-up program executing BIOS from the start-up BIOS stored in a start-up ROM to a backup BIOS stored in at least one backup ROM; steps for carrying out the ROM switching program and restarting the computer are also involved when the computer still can not be started by the CPU switching program. The method has the advantage of effective system stability improvement.

Description

The start changing method of multiprocessor computer system

Technical field

The present invention relates to a kind of start exception management method of multiprocessor computer system, be meant especially and a kind ofly carry out CPU or ROM changeover program, with the method for management start abnormal problem by BMC (baseboard management controller).

Background technology

In computer system, design concept based on high availability (High Available) system, continue operation in order to keep system, and do not need any manual operation that fault is got rid of, the necessity that just has standby system to exist, this is one of reason of multiprocessor (multiple processor) system generation.Multiprocessor computer system such as server (sever) owing to have a plurality of CPU (central processing unit) (CPU), therefore can improve bulk treatment usefulness, and when specifying CPU to make a mistake as an alternative usefulness.

Generally speaking, boot program for multiprocessor computer system, the start CPU (Boot strap Processor) that is meant order one provides calculation function, the instruction of Basic Input or Output System (BIOS) (BIOS) when being responsible for handling start is to carry out computer system initialization operation and load operation system (OS); Wherein, start BIOS is on the ROM (read-only memory) (BIOS ROM) that is stored in Basic Input or Output System (BIOS), and when start other CPU be defined as using CPU (application processors), and be set and be in waiting status (wait state).

When using start CPU to start shooting, the existing practice is the program of CPU switching when writing start in BIOS, switches to other application CPU by start CPU, and its handover mechanism as shown in Figure 1.

Another contingent problem is that BIOS has switched to all CPU trials, but still can't starts shooting; This situation may be that BIOS ROM is out of joint.In order to solve the unusual problem of BIOS, one even a plurality of backup ROM (read-only memory) (backup ROM) are used as replacement scheme; Start BIOS is switched to the back-up BIOS that is stored on the backup ROM, proceed boot program, its handover mechanism as shown in Figure 2.

Yet, the shortcoming of the aforementioned practice is to use special BIOS to come CPU switching, perhaps does the ROM (read-only memory) start and switches (ROM Boot Swap) design, carries out the switching of BIOS ROM, its electronic circuit is a more complicated, so it too bothers and do not meet cost benefit.

Summary of the invention

Technical matters to be solved by this invention is in the common technology it is to switch mode such as the start unusual hand-off process of starting shooting to rewrite BIOS, design ROM, does not meet cost benefit and work requirements.

Problem in view of above known technology, the invention provides a kind of start changing method of multiprocessor computer system, utilize baseboard management controller BMC to manage the start judgement that CPU and BIOS switch when unusual and carry out operation, wherein comprise following steps: confirm that by a baseboard management controller start unusually; Carry out a CPU changeover program and start again; When the CPU changeover program is failed, also comprise the step of carrying out a ROM changeover program and starting shooting again; Wherein, the CPU changeover program switches between a start CPU and at least one application CPU by baseboard management controller, and wherein, this CPU changeover program comprises: change the system management interrupt state of each CPU, and this CPU and a cpu bus of preceding once start is isolated; And produce a CPU switching signal and by this baseboard management controller and reopen the machine signal to a start BIOS who is stored in a start ROM or be stored in the back-up BIOS of at least one backup ROM; And the ROM changeover program switches to this back-up BIOS with the BIOS that carries out boot program by this start BIOS by baseboard management controller.

The effect that the present invention reaches is and can manages the start abnormal problem by BMC that system bios and ROM all needn't do extra design, and can further improve system stability.

Description of drawings

Fig. 1, Fig. 2 are the start handover mechanisms of multicomputer system in the explanation prior art;

Fig. 3 is the start handover mechanism of explanation the present invention with BMC managing multiprocessor system;

Fig. 4 is explanation the present invention carries out the multicomputer system start with BMC a CPU switching flow; And

Fig. 5 is explanation the present invention carries out the multicomputer system start with BMC a ROM switching flow.

Wherein, description of reference numerals is as follows:

Step 110 BMC does not obtain the boot-strap information of start BIOS

The normal boot-strap operation of step 120 system

Step 130 is not finished CPU changeover program and ROM changeover program

The operation of can't starting shooting of step 140 system

Step 150 confirms not finish the CPU changeover program

Step 160 is carried out the CPU changeover program

Step 161 changes the SMI state of all CPU so that BSP CPU and cpu bus is isolated

Step 162 BMC produces a CPU switching signal and and reopens the machine signal to BIOS or the back-up BIOS of starting shooting

Step 170 ROM changeover program

The CPU CPU (central processing unit)

The BIOS Basic Input or Output System (BIOS)

The ROM ROM (read-only memory)

The BMC baseboard management controller

CPU SMI1, the SMI2 system management interrupt of the default start of BSP CPU

SWAP state switching state

The STBY_PGD open state of awaiting orders

ROM_SWAP ROM switching state

The STATE_CHANGE state exchange

The SYS_PGD system reopens the machine state

CPU_SWAP CPU switching state

The LOW low level

A HIGH high position

BACKUPROM backs up rom state

ROMswitch ROM switching state

Embodiment

The present invention relates to a kind of start changing method of multiprocessor computer system, mainly is to utilize baseboard management controller BMC (Baseboard Management Controller) to manage start CPU and BIOS switch when unusual judgement and execution operation.

BMC is applied to IPMI (IPMI), interface between the management software of control system and the platform hardware management, provide autonomic monitoring, logout and restoring control function, and the network gateway that can be used as between the system management software and Intelligent Platform Management Bus IPMB (Intelligent Platform Management Bus) and intelligent cabinet management bus ICMB (Intelligent Chassis Management Bus) interface uses.

Why can come the management system abnormal problem by BMC, be because system can obtain the condition information of system by low pin count LPC (Low Pin Count) interface from BMC.

The present invention is another brand-new application of BMC, below according to Fig. 3 explanation processing mode by BMC management start unusual condition.Execution priority is to make CPU switching, reopening machine earlier substantially, if not all right ROM switching, the reopening machine done again.

At first, after system power supply activates, confirm that BMC does not obtain the boot-strap information (step 110) of start BIOS; As obtain boot-strap information, expression system normal boot-strap is operated (step 120).BMC is with the power supply of system reserve power supply, so before the system power supply activation, BMC promptly is ready for, so could activate in system power supply one, just receive the boot program running status that BIOS transmits.

Then, confirm not finish CPU changeover program and ROM changeover program (step 130); As system finished CPU, the ROM changeover program but can't be started shooting, and represents all CPU all to make a mistake, system's operation (step 140) of can't starting shooting can only manually be fixed a breakdown, and for example changes CPU.

Secondly, confirm not finish CPU changeover program (step 150), carry out CPU changeover program (step 160) immediately.

The CPU changeover program of step 160 also comprises two detailed process; One for a change the SMI state of all CPU so that (so-called BSP CPU is meant boot strap processor, is meant when starting shooting at the beginning, the CPU that begins to move start earlier, just default CPU in order to start with BSP CPU; In for the second time later CPU changeover program, then be the preceding once CPU of start) completely cut off (step 161) with cpu bus, BMC produces a CPU switching signal and and reopens the machine signal to BIOS or the back-up BIOS (step 162) of starting shooting then.After the reopening machine, promptly get back to step 110 and confirm open state.

After the judgement of step 150, switched, promptly carried out ROM changeover program (step 170) as carrying out CPU.This program by the start BIOS among the BIOS ROM, switches to the back-up BIOS among the backup ROM, and reopens machine with back-up BIOS promptly at the BIOS that will carry out boot program; Know clearly it, produce ROM switching signal to a complex programmable logic device (CPLD) (Complex ProgrammableLogic Device) by BMC and back up ROM, and the generation system reopens the machine signal to back-up BIOS to switch to.After the reopening machine, also get back to step 110 and confirm open state.

See also Fig. 4, illustrate that the execution flow process among the BMC can be in order to prove feasibility of the present invention when carrying out the CPU switching.SMI1 and SMI2 are two the system management interrupt SMI (SystemManagement Interrupt) on the BMC, SWAP state representation switching state, STBY_PGD, ROM_SWAP, STATE_CHANGE, SYS_PGD, CPU_SWAP etc. are functions of control program parameter among the BMC; STBY_PGD is the open state of awaiting orders, and ROM_SWAP is the ROM switching state, and STATE_CHANGE is a state exchange, and SYS_PGD is that system reopens the machine state, and CPU_SWAP is the CPU switching state.The switching of CPU comprises one of four states among the figure, can allow BMC know and switch to which CPU, and the execution content division under each state is as follows:

The execution content of the 1st state comprises: it is LOW (low level) that a. sets SMI1;

B. setting SMI2 is HIGH (high position);

C. set SWAP state to the 2 states;

D. setting STATE_CHANGE is CHANGE (conversion).

The execution content of the 2nd state comprises:

A. setting SMI1 is HIGH;

B. setting SMI2 is LOW;

C. set SWAP state to the 3 states;

D. setting STATE_CHANGE is CHANGE.

The execution content of the 3rd state comprises:

A. setting SMI1 is LOW;

B. setting SMI2 is LOW;

C. set SWAP state to the 4 states;

D. setting STATE_CHANGE is CHANGE.

The execution content of the 4th state comprises:

A. set SWAP state to the 4 states;

B. setting STATE_CHANGE is CHANGE.

Fig. 5 can illustrate the detailed process of the present invention when utilizing BMC to carry out ROM switching (ROM SWAP), also can be in order to verify feasibility of the present invention.Wherein BACKUPROM representative backup rom state in the present invention, backs up ROM and can be in normal condition (normal state) or stand-by state (backupstate); ROMswitch then represents the functional parameter of ROM switching state.

According to CPU and the ROM flow process of Fig. 4, Fig. 5, BMC can be according to the flow process of Fig. 3, and the CPU that carries out Fig. 4 when start is unusual earlier switches, and carries out Fig. 5 ROM when successfully not starting shooting again and switches, and confirms with BMC management start unusual condition truly feasible.

The above only for preferred embodiment of the present invention, is not in order to limiting scope of the invention process, to be familiar with this technician after announcement of the present invention, and change of being done and adjustment in view of the above all belongs in the scope of the technology of the present invention thought.

Therefore, the equalization of being done without departing from the spirit and scope of the present invention changes and modifies, and all should belong in claims of the present invention claim required for protection.

Claims

1. the start changing method of a multiprocessor computer system wherein comprises following steps:

Confirm that by a baseboard management controller start is unusual; And

Carry out a CPU changeover program and start again;

This CPU changeover program switches between a start CPU and at least one application CPU by this baseboard management controller, and wherein, this CPU changeover program comprises:

Change the system management interrupt state of each CPU, this CPU and a cpu bus of preceding once start is isolated; And

Produce a CPU switching signal and by this baseboard management controller and reopen the machine signal to a start BIOS who is stored in a start ROM or be stored in the back-up BIOS of at least one backup ROM;

In the time still can't starting shooting, also comprise the step of carrying out a ROM changeover program and starting shooting again by this CPU changeover program;

This ROM changeover program switches to this back-up BIOS with the BIOS that carries out boot program by this start BIOS by this baseboard management controller.

2. the start changing method of multiprocessor computer system as claimed in claim 1, wherein this ROM changeover program is to produce ROM switching signal to a CPLD switching to this backup ROM by this baseboard management controller, and the system that produces reopens the machine signal to this back-up BIOS.

3. the start changing method of multiprocessor computer system as claimed in claim 1, wherein start is meant that unusually this baseboard management controller do not receive the boot-strap information from this start BIOS or this back-up BIOS.

4. the start changing method of multiprocessor computer system as claimed in claim 1 wherein after confirming that start is unusual, also comprises the step that this CPU changeover program and this ROM changeover program are not finished in an affirmation.

5. the start changing method of multiprocessor computer system as claimed in claim 4 wherein before carrying out this CPU changeover program, also comprises the step that this CPU changeover program is not finished in an affirmation.