Method for monitoring and displaying AI card information by server BMC

Abstract

The invention discloses a method for monitoring and displaying AI card information by a server BMC, and relates to the field of BMC monitoring and AI card information display, and the method comprises the following steps: when the timing time period of AI card information display arrives, reading the power supply voltage of each PCIe slot position by the server BMC, and judging whether the read power supply voltage of each PCIe slot position is in a normal range or not; the BMC reads a register of the AI card inserted in the PCIe slot position with each power supply voltage within the normal range through the multi-channel I2C switch to obtain information of the AI card inserted in the PCIe slot position with each power supply voltage within the normal range; and the BMC displays the information of the AI cards inserted in the PCIe slot positions with each power supply voltage within a normal range through an external interaction device. According to the method and the device, the information of the AI card inserted in each PCIe slot can be directly acquired and displayed without depending on a BIOS (Basic Input / Output System).

Description

Technical Field

[0001] The present invention relates to the field of monitoring of a server Baseboard Management Controller (BMC) and accelerating the display of information of a board (or called an AI card, an acceleration card device, an AI card device), and particularly relates to a method for a server BMC to monitor and display AI card information. Background Art

[0002] Generally, multiple AI cards are integrated in a server. When an AI card is in an abnormal state, or when it is necessary to view the health information of the AI card in a single-machine state, it is very difficult for a user to perceive and locate at the physical level, which brings great troubles to the user and the operation and maintenance personnel.

[0003] To solve the above problems, the existing BMC obtains the status information and slot information of the AI card. The specific method is that when the server performs power-on self-test, the BIOS reads the slot and AI card ID information according to the system configuration; the slot ID and the corresponding AI card ID information are sent to the BMC through the CPU. With this method, it is necessary to obtain the AI card information through the BIOS. Once the BIOS firmware is damaged, or the BIOS Flash fails, or the BIOS does not work, the BMC will not be able to obtain the AI card information and cannot monitor the AI card information. Summary of the Invention

[0004] The present invention provides a method for a server BMC to monitor and display AI card information, aiming to realize the monitoring and display of AI card information without relying on the BIOS.

[0005] A method for displaying the status of an AI card of a server provided by the present invention includes: when the timing time period for displaying AI card information arrives, the server BMC reads the power supply voltage of each PCIe slot and determines whether the power supply voltage of each read PCIe slot is within the normal range; the BMC reads the registers of the AI card plugged in each PCIe slot with a normal power supply voltage through a multi-path I2C switch to obtain the information of the AI card plugged in each PCIe slot with a normal power supply voltage; the BMC displays the information of the AI card plugged in each PCIe slot with a normal power supply voltage through an external interaction device.

[0006] Preferably, the information of the AI card plugged into any PCIe slot includes the in-position state and out-of-position state of the AI card plugged into the PCIe slot; correspondingly, the BMC reads the registers of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through a multi-path I2C switch, so as to obtain the information of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range, including: for any AI card plugged into a PCIe slot with the power supply voltage within the normal range, the BMC reads the registers of the AI card plugged into the PCIe slot through the multi-path I2C switch; the BMC determines whether the communication between the BMC and the AI card plugged into the PCIe slot is successful according to whether the information returned by the AI card plugged into the PCIe slot is received; when the communication between the BMC and the AI card plugged into the PCIe slot is successful, the BMC determines that the AI card plugged into the PCIe slot is in position; when the communication between the BMC and the AI card plugged into the PCIe slot fails and the number of communication failures reaches the maximum number of failures, the BMC determines that the AI card plugged into the PCIe slot is out of position.

[0007] Preferably, the information returned by the AI card plugged into the PCIe slot received by the BMC includes the board status of the AI card plugged into the PCIe slot; the in-position state of the AI card plugged into the PCIe slot includes the normal in-position state and abnormal in-position state of the AI card plugged into the PCIe slot; correspondingly, the method further includes: the BMC determines whether the AI card plugged into the PCIe slot is in the normal in-position state or abnormal in-position state according to the board status of the AI card plugged into the PCIe slot.

[0008] Preferably, the BMC determines whether the AI card plugged into the PCIe slot is in the normal in-position state or abnormal in-position state according to the board status of the AI card plugged into the PCIe slot, including: if the board status of the AI card plugged into the PCIe slot is the ready state or working state, the BMC determines that the AI card plugged into the PCIe slot is in the normal in-position state; if the board status of the AI card plugged into the PCIe slot is the alarm state, the BMC determines that the AI card plugged into the PCIe slot is in the abnormal in-position state.

[0009] Preferably, the information returned by the AI card plugged into the PCIe slot received by the BMC further includes the board type, operating status, overall board power, overall board current, input voltage, temperature, and fan speed; the information of the AI card plugged into the PCIe slot further includes the loaded status of the AI card plugged into the PCIe slot and data related to the PCIe slot for plugging in the AI card, where the data related to the PCIe slot includes the slot number, board type, operating status, overall board power, overall board current, input voltage, temperature, and fan speed of the PCIe slot; correspondingly, the BMC reads the registers of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through a multi-path I2C switch to obtain the information of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range, including: after the BMC obtains the data related to the PCIe slot for plugging in the AI card, it determines the loaded status of the AI card according to the overall board power.

[0010] Preferably, the one external interaction device includes: an LCD for displaying the information of the AI cards plugged into each PCIe slot; an I2C-to-parallel port chip that receives the display data of the information of the AI cards plugged into each PCIe slot sent by the BMC through the I2C bus and transmits the display data of the information of the AI cards plugged into each PCIe slot in parallel to the LCD for the LCD to display.

[0011] Preferably, the one external interaction device further includes: a button; a parallel port-to-I2C chip for receiving the button input information when the user touches the button and transmitting the button input information to the BMC through the I2C bus.

[0012] Preferably, the method further includes: when the timing time period of button detection by the BMC arrives, the BMC scans and saves the button input information from the one external interaction device.

[0013] Preferably, the BMC displays the information of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through an external interaction device, including: during the BMC displays the in-position normal state, in-position abnormal state, or out-of-position state of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through the LCD, it determines whether button input information is received; if no button input information is received, the BMC keeps the display content of the LCD; if button input information is received, the BMC switches the display content of the LCD to the loaded status of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range or the data related to the PCIe slots with the power supply voltage within the normal range corresponding to a certain AI card according to the indication of the button input information.

[0014] Preferably, the one external interaction device is installed on the front panel or the rear panel of the chassis of the server and is connected to the BMC through the I2C bus.

[0015] The BMC of the present invention is connected to the AI cards plugged into each PCIe slot through a multi-path I2C switch, so as to directly obtain the information of the AI cards plugged into each PCIe slot by reading the registers of the AI cards plugged into each PCIe slot without relying on the BIOS. In addition, through the external interaction device, the AI card information can be displayed locally on the server without logging in to the web interface, which is convenient for positioning and viewing. Description of the Drawings

[0016] Figure 1 is the first flowchart of the method for the server BMC to monitor and display AI card information provided by the present invention;

[0017] Figure 2 is the schematic block diagram of the circuit principle involved in the method for the server BMC to monitor and display AI card information provided by the present invention;

[0018] Figure 3 is the reference effect diagram of the LCD display of the AI card information provided by the present invention;

[0019] Figure 4 is the first detailed flowchart of the method for the server BMC to monitor and display AI card information provided by the present invention;

[0020] Figure 5 is the second detailed flowchart of the method for the server BMC to monitor and display AI card information provided by the present invention;

[0021] Figure 6 is the third detailed flowchart of the method for the server BMC to monitor and display AI card information provided by the present invention. Detailed Description of the Invention

[0022] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0023] The present invention provides a method for a server BMC to monitor and display AI card information. The BMC directly communicates with the AI cards plugged into each PCIe slot through a multi-path I2C switch. In the case where the BIOS does not work, the BIOS firmware is damaged, the BIOS Flash fails, etc., the AI card information can still be monitored, and the AI card information is displayed through an external interaction device, which is convenient for positioning and viewing and has high reliability.

[0024] See Figure 1, a method for a server BMC to monitor and display AI card information may include the following steps:

[0025] Step S101: When the timing time period for AI card information display arrives, the server BMC reads the power supply voltage of each PCIe slot and determines whether the power supply voltage of each read PCIe slot is within the normal range.

[0026] Step S102: The BMC reads the registers of the AI cards plugged into the PCIe slots with normal power supply voltage through a multi-channel I2C switch to obtain the information of the AI cards plugged into the PCIe slots with normal power supply voltage.

[0027] Step S103: The BMC displays the information of the AI cards plugged into the PCIe slots with normal power supply voltage through an external interaction device.

[0028] In the first embodiment, the information of the AI card plugged into any PCIe slot may include the inserted state and non-inserted state of the AI card plugged into the PCIe slot. Correspondingly, for the AI card plugged into any PCIe slot with normal power supply voltage, step S102 may include: the BMC reads the registers of the AI card plugged into the PCIe slot through a multi-channel I2C switch; the BMC determines whether the communication between the BMC and the AI card plugged into the PCIe slot is successful according to whether the information returned by the AI card plugged into the PCIe slot is received. Generally, receiving the returned information indicates successful communication, otherwise it indicates communication failure; when the communication between the BMC and the AI card plugged into the PCIe slot is successful, the BMC determines that the AI card plugged into the PCIe slot is inserted; when the communication between the BMC and the AI card plugged into the PCIe slot fails and the number of communication failures reaches the maximum number of failures, the BMC determines that the AI card plugged into the PCIe slot is not inserted, where the maximum number of failures may need to be set in advance, for example, set to 5 times, 6 times, etc. Step S103 may include: the BMC displays the inserted state or non-inserted state of the AI card plugged into each PCIe slot through an external interaction device, which can be displayed in text mode. For example, "Slot 1: Inserted" displayed on the external interaction device indicates that the AI card plugged into slot 1 is inserted, and "Slot 6: Not Inserted" indicates that the AI card plugged into slot 6 is not inserted. It can also be displayed in image mode. For example, the circle corresponding to slot 1 displayed on the external interaction device is a solid black circle, indicating that the AI card plugged into slot 1 is inserted, and the circle corresponding to slot 6 is a hollow black circle, indicating that the AI card plugged into slot 6 is not inserted.

[0029] In the second embodiment, the information returned by the AI card plugged into the PCIe slot received by the BMC includes the board status of the AI card plugged into the PCIe slot; the information of the AI card plugged into the PCIe slot includes the in-position status and the out-of-position status, where the in-position status includes the normal in-position status and the abnormal in-position status of the AI card plugged into the PCIe slot. For any AI card plugged into a PCIe slot with a power supply voltage within the normal range, step S102 may include: the BMC reads the register of the AI card plugged into the PCIe slot through a multi-path I2C switch; the BMC determines whether the communication between the BMC and the AI card plugged into the PCIe slot is successful according to whether the information returned by the AI card plugged into the PCIe slot is received; when the communication between the BMC and the AI card plugged into the PCIe slot is successful, the BMC determines that the AI card plugged into the PCIe slot is in position, and at this time, the BMC determines whether the AI card plugged into the PCIe slot is in a normal in-position state or an abnormal in-position state according to the board status of the AI card plugged into the PCIe slot. Specifically, if the board status of the AI card plugged into the PCIe slot is the ready state or the working state, the BMC determines that the AI card plugged into the PCIe slot is in a normal in-position state; if the board status of the AI card plugged into the PCIe slot is the alarm state, the BMC determines that the AI card plugged into the PCIe slot is in an abnormal in-position state; when the communication between the BMC and the AI card plugged into the PCIe slot fails and the number of communication failures reaches the maximum number of failures, the BMC determines that the AI card plugged into the PCIe slot is out of position, where the maximum number of failures may need to be set in advance, for example, set to 5 times, 6 times, etc. Step S103 may include: the BMC displays the normal in-position state, abnormal in-position state or out-of-position state of the AI card plugged into each PCIe slot through an external interaction device, which can be displayed in text. For example, "Slot 1: In-position normally" displayed on the external interaction device indicates that the AI card plugged into slot 1 is in a normal in-position state, "Slot 5: In-position abnormally" indicates that the AI card plugged into slot 5 is in an abnormal in-position state, and "Slot 6: Out of position" indicates that the AI card plugged into slot 6 is out of position. It can also be displayed in an image. For example, the circle corresponding to slot 1 displayed on the external interaction device is a solid black circle, indicating that the AI card plugged into slot 1 is in a normal in-position state; the circle corresponding to slot 5 displayed on the external interaction device is filled with a black grid, indicating that the AI card plugged into slot 5 is in an abnormal in-position state; the circle corresponding to slot 6 is a hollow black circle, indicating that the AI card plugged into slot 6 is out of position.

[0030] In the third embodiment, based on the second embodiment, the information returned by the AI card plugged into the PCIe slot received by the BMC further includes the board type, operating status, overall board power, overall board current, input voltage, temperature, and fan speed; the information of the AI card plugged into the PCIe slot further includes the loaded status of the AI card plugged into the PCIe slot and data related to the PCIe slot for plugging the AI card, where the data related to the PCIe slot includes the slot number, board type, operating status, overall board power, overall board current, input voltage, temperature, and fan speed of the PCIe slot. Accordingly, step S102 may further include: after obtaining the data related to the PCIe slot for plugging the AI card, the BMC determines the loaded status of the AI card according to the overall board power. Step S103 may further include: the BMC displays the loaded status of the AI card plugged into each PCIe slot and the data related to the PCIe slot for plugging each AI card through an external interaction device. Among them, the loaded status of each PCIe slot and the plugged AI card may be displayed in text, such as fully loaded, heavily loaded, lightly loaded, etc., or may be displayed in graphical ways such as bar charts and pie charts. And the data related to the PCIe slot for plugging each AI card is displayed in text.

[0031] The BMC of the present invention is connected to the AI card plugged into each PCIe slot through a multi-channel I2C switch, so as to directly obtain the information of the AI card plugged into each PCIe slot without relying on the BIOS by reading the registers of the AI card plugged into each PCIe slot. In addition, through an external interaction device, the AI card information is displayed locally on the server without logging in to the web interface, which is convenient for positioning and viewing.

[0032] In addition, in the prior art, the BMC sends the AI card status information to the CPLD through the I2C system bus, and the CPLD controls the indicator light to switch between different colors according to the status information and slot information received by the BMC. This method cannot display detailed data, and users or operation and maintenance personnel can only log in to the web interface through the network to view the detailed data. To display the status of the AI card and detailed data related to the PCIe slot on the server side, the present invention provides an external display device. The external interaction device is installed on the front panel or the rear panel of the chassis of the server and is connected to the BMC through the I2C bus to directly view the AI card information involved in the above three embodiments locally on the server. The external display device may include an LCD and an I2C-to-parallel port chip. Among them, the I2C-to-parallel port chip receives the display data of the AI card information plugged into each PCIe slot sent by the BMC through the I2C bus, and parallel-transmits the display data of the AI card information plugged into each PCIe slot to the LCD for the LCD to display the AI card information plugged into each PCIe slot.

[0033] In addition, limited by the size of the LCD, it is difficult to display the in-position normal state / in-position abnormal state / out-of-position state, loaded state, and data related to the PCIe slot of the AI card on a single display interface. Therefore, the external interaction device of the present invention further includes a button and a parallel port-to-I2C chip. Among them, the parallel port-to-I2C chip receives the button input information when the user touches the button, and transmits the button input information to the BMC through the I2C bus, so that the display content of the LCD can be switched according to the button input information. Specifically, when the timing time period of button detection arrives, the BMC scans and saves the button input information from the one external interaction device. Correspondingly, step S103 may include: the BMC displays the in-position normal state or in-position abnormal state or out-of-position state of the AI card plugged into each PCIe slot with the supply voltage within the normal range through the LCD. During this period, it is judged whether button input information is received. If no button input information is received, the BMC keeps the display content of the LCD. If button input information is received, the BMC switches the display content of the LCD to the loaded state of the AI card plugged into each PCIe slot with the supply voltage within the normal range or the data related to the PCIe slot corresponding to a certain AI card with the supply voltage within the normal range according to the indication of the button input information. The present invention can more intuitively display the AI card information at the physical level.

[0034] The following combines Figures 2 to 6 , and details the present invention.

[0035] In the prior art, the method for the BMC to obtain the status information and slot information of the AI card is as follows: When the server performs a power-on self-test, the BIOS reads the slot and AI acceleration card ID information according to the system configuration; the slot ID and the corresponding AI acceleration card ID information are sent to the BMC through the CPU. The BMC sends the AI card status information to the CPLD through the I2C system bus, and the CPLD controls the status indicator to switch between different colors according to the status information and slot information received by the BMC. This method has at least the following defects: 1. It is necessary to obtain AI card information through the BIOS. Once the BIOS firmware is damaged, or the BIOS Flash fails, or the BIOS does not work, the BMC will not be able to obtain the AI card information and cannot monitor the AI card information. 2. The indicator can only display some discrete states of the AI card and cannot display more data information such as the board type, power, fan speed, and real-time board temperature of the AI card, which is not convenient for operation and maintenance. For example, when it is seen in the background that an AI card has been running at high power for a year and needs to be replaced in advance, the AI card itself is not damaged and the indicator shows green to indicate that the AI card is normal. After the operation and maintenance personnel arrive at the scene, they cannot distinguish which AI card is in the heavy load mode. 3. One AI card corresponds to one multi-color indicator. The CPLD needs 1 pin to control one state of a multi-color indicator. For an 8-slot server, if the AI card in each slot needs to display 5 states, then 40 CPLD pins will be occupied. Therefore, the prior art is limited by the number of pins.

[0036] To solve the above problems, the present invention provides a method for a server BMC to monitor and display AI card information. The BMC is connected to each PCIe slot of the server through a multi-channel I2C switch (IIC SWITCH). There is a strict correspondence between the slot information of the PCIe slot and the hardware connection link. The BMC can directly obtain all the information of the AI cards on each PCIe slot through the I2C path, without relying on the BIOS to obtain the AI card information of the PCIe slot. In addition, the BMC is connected to the LCM (LCD display module, liquid crystal module) through the IIC interface, and sends the obtained AI card information (in-position normal state, out-of-position, in-position abnormal state, loaded state, data related to the PCIe slot) to the LCM module, and drives the LCM module to display it in real time. The LCM module can be installed and embedded in the front panel or the rear panel of the server chassis. Among them, the LCM module is an independent display interface module, which includes buttons, an LCD, and two IIC serial data-to-parallel port chips connected to the BMC through the I2C bus. One IIC serial data-to-parallel port chip (or I2C-to-parallel port chip) is connected to the LCD parallel port and is used for the display output function, that is, the BMC realizes the drive control of the LCD display through the I2C-to-parallel port chip. The other IIC serial data-to-parallel port chip (or parallel port-to-I2C chip) is connected with multiple (such as 4 or 5) independent buttons and is used for the button switching function. In addition, the server also includes a power supply module for powering the PCIe. Correspondingly, the LCM module also includes a power conversion module, which converts the voltage provided by the power supply module and provides the converted voltage to the two IIC serial data-to-parallel port chips and the LCD. The present invention adopts the LCM module, which can not only overcome the problem that the state of the LED lights can only be seen by opening the upper cover of the server chassis in the prior art, but also no longer requires the CPLD to drive and light the LED lights, solving the problem of insufficient hardware resources caused by excessive physical pins of the server CPLD being occupied. Moreover, the AI card status and data displayed on the LCD interface in the LCM module correspond one by one to the PCIe slot numbers of the AI cards. Through the button switching function, more real-time information of the AI cards can be displayed, facilitating the operation and maintenance personnel to understand more AI card information, such as load power consumption, board temperature, fan speed, etc., and then warning the health status of the AI cards and making corresponding countermeasures in advance.

[0037] See Figure 2 , the circuit includes a server and an external interactive device LCM.

[0038] In the server, the BMC and the multi-channel I2C switch (I2C SWITCH) are connected through the I2C interface. The multi-channel I2C switch and the AI cards on each PCIe slot are connected through the I2C interface. The internal power supply module of the server powers the PCIe board. The BMC periodically obtains the data of the internal registers of the AI card through the I2C interface to determine whether the AI card is in place, obtains the information of the AI card, and sends the obtained AI card information to the LCM module through the I2C interface for display output.

[0039] The LCM is an independent display interface module, which includes a device body. The device body integrates an LCD, buttons, an IIC serial data to parallel port chip (or I2C to parallel port chip) connected to the LCD in parallel, another IIC serial data to parallel port chip (or parallel port to I2C chip) connected to multiple independent buttons (such as 4 or 5), and a power conversion module for converting the voltage provided by the internal power supply module of the server into the voltage required by the LCM and supplying it to the two IIC serial data to parallel port chips and the LCD.

[0040] Among them, the BMC performs button input detection. For example, it periodically reads the button status of an I2C serial data to parallel port chip connected to the button through the I2C interface of the BMC, scans the button key value (or button input information), and based on the button key value, determines how to switch the display content of the LCD.

[0041] Among them, the display principle of the LCM module is to control the LCD to output the display according to the control timing of the LCD interface RW (read / write selection terminal), RS (data / command selection terminal), CS (enable signal), and D4-D7 (Date I / O). Correspondingly, the BMC drives the LCM module to perform display output, including: the BMC sends data to an I2C serial data to parallel port chip connected to the LCD through the I2C interface. The I2C serial data to parallel port chip connected to the LCD outputs a set of control IOs (P0-P7), and the parallel port IOs are respectively connected to the RW, RS, CS, and D4-D7 of the LCD to drive the LCD to output the display, and finally display the AI card information through the LCD.

[0042] See Figure 3 The reference effect diagram of the AI card information output and displayed by the LCM module shown. More effects can be displayed according to the different needs of users. Taking Figure 3 as an example, the AI card information displayed includes:

[0043] 1. Normal in-place state, abnormal in-place state, not in-place state

[0044] Figure 3Among them, the status of the AI cards inserted in each of the 8 PCIe slots is shown. Among them, the black-filled square indicates the normal in-position state, the grid-filled square indicates the abnormal in-position state, and the blank square indicates the out-of-position state.

[0045] In specific implementation, these three states can be distinguished by different colors, or by different graphics, or by different filling contents. The specific distinguishing and representation methods are not limited to this.

[0046] 2. Load state

[0047] Figure 3 Among them, a bar chart is used to display the load state of the AI card. The greater the power of the AI card, the higher the bar chart, and vice versa, the lower the bar chart.

[0048] In specific implementation, the display method of the load state can also adopt other methods such as a pie chart.

[0049] 3. Data related to the corresponding PCIe slot of the AI card (abbreviated as detailed data)

[0050] The data includes but is not limited to slot number (SLOT), board type (TYPE), operating status (STA), total board power (W), total board current (A), input voltage (V), temperature (T), fan speed (FS), etc.

[0051] See Figure 4 For the first detailed flowchart of, the steps of the method for the server BMC to monitor and display AI card information may include:

[0052] Step 1: After the BMC initializes the I2C, the BMC starts a timer.

[0053] Step 2: Determine whether the timing time period for displaying AI card information has arrived. If it has arrived, execute Step 3; otherwise, repeat Step 2.

[0054] Step 3: The BMC reads the power supply voltage of the server PCIe slot and determines whether the power supply of the server PCIe slot is normal. If it is in the normal PCIe power supply state, execute Step 4; if it is not in the normal power supply state, directly return and wait for the next timer startup.

[0055] In short, the present invention periodically determines whether the power supply of the server PCIe slot is normal through Step 2 and Step 3.

[0056] Step 4: The BMC reads the registers of the AI card through the I2C bus to determine whether the communication is successful by checking if data is returned. If data is returned, it indicates successful communication, meaning the AI card is present, and Step 6 is executed. If no data is returned, the command is resent to obtain data, and it is determined whether the maximum error retry count (or maximum communication failure count) has been reached. If data cannot be obtained multiple times, for example, if no data can be obtained even when the maximum error retry count is reached, it means the AI card is not present, and Step 5 is executed. Otherwise, this step is repeated;

[0057] Step 5: The BMC sends data indicating that the AI card is not present through the I2C interface to drive the LCD of the LCM module to display that the status of the AI card in the current slot is not present;

[0058] Step 6: The BMC sends data indicating that the AI card is present through the I2C interface to drive the LCD of the LCM module to display that the status of the AI card in the current slot is present.

[0059] See Figure 5 the second detailed flowchart of

[0060] Step 1: After the BMC initializes the I2C, the BMC starts a timer;

[0061] Step 2: Determine whether the timing period for displaying AI card information has arrived. If it has arrived, execute Step 3; otherwise, repeat Step 2;

[0062] Step 3: The BMC reads the power supply voltage of the server's PCIe slot and determines whether the server's PCIe slot is powered normally. If it is in a normal PCIe power supply state, execute Step 4; if it is not in a normal power supply state, return directly and wait for the next timer start;

[0063] In short, the present invention periodically determines whether the server's PCIe slot is powered normally through Step 2 and Step 3.

[0064] Step 4: The BMC reads the registers of the AI card through the I2C bus to determine whether the communication is successful by checking if data is returned. If data is returned, it indicates successful communication, meaning the AI card is present, and Step 6 is executed. If no data is returned, the command is resent to obtain data, and it is determined whether the maximum error retry count (or maximum communication failure count) has been reached. If data cannot be obtained multiple times, for example, if no data can be obtained even when the maximum error retry count is reached, it means the AI card is not present, and Step 5 is executed. Otherwise, this step is repeated;

[0065] Step 5: The BMC represents the data indicating that the AI card is not in place through the I2C interface, drives the LCD of the LCM module, and displays that the status of the AI card in the current slot is not in place.

[0066] Step 6: The BMC determines whether the AI card status data (i.e., the board status) represents that the AI card is in a normal state or an abnormal state. If the AI card status data represents that the AI card is in an abnormal in-place state, then execute Step 7. If the AI card status data represents that the AI card is in a normal in-place state, then execute Step 8.

[0067] Step 7: The BMC sends the data indicating that the AI card is abnormally in place through the I2C interface, drives the display of the LCM module, and displays that the status of the AI card is abnormally in place.

[0068] Step 8: The BMC sends the data indicating that the AI card is normally in place through the I2C interface, drives the LCD of the LCM module, and displays that the status of the AI card in the current slot is normally in place.

[0069] See Figure 6 the third detailed flowchart, the steps of the method for the server BMC to monitor and display AI card information may include:

[0070] Step 1: After the BMC initializes the I2C, the BMC starts two timers and executes Step 2 and Step 2A;

[0071] One timer is the timer for AI card information display, and the other timer is the key detection timer.

[0072] Step 2: Determine whether the timing period for AI card information display has arrived. If it has arrived, then execute Step 3. Otherwise, repeat Step 2;

[0073] Step 2A: Determine whether the timing period for key detection has arrived. If it has arrived, then execute Step 3A. Otherwise, repeat Step 2A;

[0074] Step 3: The BMC reads the power supply voltage of the server PCIe slot and determines whether the power supply of the server PCIe slot is normal. If it is in a normal PCIe power supply state, then execute Step 4. If it is not in a normal power supply state, directly return and wait for the next timer start;

[0075] Step 3A: The BMC scans the key input, records the key value, and uses it to determine what kind of AI card information is displayed on the LCM module.

[0076] In short, the present invention periodically determines whether the power supply of the server PCIe slot is normal through Step 2 and Step 3, and periodically obtains the key input through the I2C interface through Step 2A and Step 3A.

[0077] Step 4: The BMC reads the registers of the AI card through the I2C bus. It determines whether the communication is successful by checking if data is returned. If data is returned, it indicates successful communication, meaning the AI card is present, and Step 6 is executed. If no data is returned, it resends the command to obtain data and checks if the maximum error retry count (or maximum communication failure count) has been reached. If data cannot be obtained multiple times, for example, when the maximum error retry count is reached and still no data is obtained, it means the AI card is not present, and Step 5 is executed. Otherwise, this step is repeated;

[0078] Step 5: The BMC indicates through the I2C interface that the AI card is not present, driving the LCD of the LCM module to display that the status of the AI card in the current slot is not present;

[0079] Step 6: The BMC determines whether the AI card status data (i.e., the board status) represents that the AI card is in a normal state or an abnormal state. If the AI card status data represents that the AI card is in an abnormal present state, Step 7 is executed. If the AI card status data represents that the AI card is in a normal present state, Step 8 is executed;

[0080] Step 7: The BMC sends data indicating an abnormal present state of the AI card through the I2C interface, driving the display of the LCM module to show that the status of the AI card is an abnormal present state, and Step 9 is executed;

[0081] Step 8: The BMC sends data indicating a normal present state of the AI card through the I2C interface, driving the LCD of the LCM module to display that the status of the AI card is a normal present state, and Step 9 is executed;

[0082] Step 9: Continue to check if there is a key value indicating that a key has been pressed. If there is no key value indicating that a key has been pressed, no display switch is made, and the current state is displayed. If there is a key value indicating that a key has been pressed, Step 10 is executed;

[0083] Four keys can be set. One key is used to switch the display of the loaded state of the AI card, and the other three keys include the left key, the right key, and the confirmation key, which are used to switch the display of the detailed data of the AI card. The detailed data of different AI cards can be switched by using the left and right keys.

[0084] Step 10: When the key value indicates the loaded state, the BMC sends data for the loaded state of the AI card through the I2C interface, driving the display of the LCM module to show that the status of the AI card is the loaded state of the AI card in this slot; when the key value indicates the slot and detailed data, the BMC sends the details of the AI card in this slot through the I2C interface, driving the display of the LCM module to show. The detailed data of the AI card in this slot includes: slot number, board type, operating status, overall board power, overall board current, input voltage, temperature, fan speed, etc.

[0085] The method for the server BMC to monitor and display the status of the AI card provided by the present invention has the following technical effects:

[0086] 1. In terms of hardware design, the BMC of the present invention is connected to the I2C of each PCIe slot through the I2C SWITCH one by one, and the corresponding relationship between the AI card and the PCIe slot can be directly determined. In this way, the internal registers of the AI card are sequentially read through the I2C interface to obtain the AI card information, without relying on the BIOS, avoiding the problem that the AI card information cannot be obtained due to the damage of the BIOS firmware or the BIOS flash, etc. And the AI card information obtained through the BIOS is limited. Therefore, the method of the present invention is more stable, reliable and easy to operate;

[0087] 2. The LCM can be installed on the front panel or the rear panel of the server chassis, which is convenient to install and easy to locate and view; the slot information displayed on the LCD of the LCM corresponds to the physical slot information one by one, and the AI card information can be more directly indicated at the physical level; the LCD can display more useful information, bringing great convenience and a better experience to the operation and maintenance personnel;

[0088] 3. The BMC of the present invention drives the LCM to display the AI card information through the I2C bus, occupying fewer IO ports of the BMC and solving the problem of limited hardware IO port resources.

[0089] The preferred embodiments of the present invention have been described above with reference to the accompanying drawings, and the scope of the rights of the present invention is not limited thereby. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and essence of the present invention shall be within the scope of the rights of the present invention.

Claims

1. A method for a server BMC to monitor and display AI card information, characterized in that, The method includes: When the timing time period for AI card information display arrives, the server BMC reads the power supply voltage of each PCIe slot and determines whether the power supply voltage of each read PCIe slot is within the normal range; The BMC reads the registers of the AI cards plugged into the PCIe slots with normal power supply voltage through a multi-channel I2C switch to obtain the information of the AI cards plugged into the PCIe slots with normal power supply voltage; The BMC displays the information of the AI cards plugged into the PCIe slots with normal power supply voltage through an external interaction device.

2. The method according to claim 1, characterized in that, The information of the AI card plugged into any PCIe slot includes the present state and absent state of the AI card plugged into the PCIe slot; Correspondingly, the BMC reads the registers of the AI cards plugged into the PCIe slots with normal power supply voltage through a multi-channel I2C switch to obtain the information of the AI cards plugged into the PCIe slots with normal power supply voltage, including: For the AI card plugged into any PCIe slot with normal power supply voltage, the BMC reads the register of the AI card plugged into the PCIe slot through a multi-channel I2C switch; The BMC determines whether the communication between the BMC and the AI card plugged into the PCIe slot is successful according to whether the information returned by the AI card plugged into the PCIe slot is received; When the communication between the BMC and the AI card plugged into the PCIe slot is successful, the BMC determines that the AI card plugged into the PCIe slot is present; When the communication between the BMC and the AI card plugged into the PCIe slot fails and the number of communication failures reaches the maximum number of failures, the BMC determines that the AI card plugged into the PCIe slot is absent.

3. The method according to claim 2, characterized in that, The information returned by the AI card plugged into the PCIe slot received by the BMC includes the board status of the AI card plugged into the PCIe slot; The present state of the AI card plugged into the PCIe slot includes the normal present state and abnormal present state of the AI card plugged into the PCIe slot; Correspondingly, the method further includes: the BMC determines whether the AI card plugged into the PCIe slot is normally present or abnormally present according to the board status of the AI card plugged into the PCIe slot.

4. The method according to claim 3, characterized in that, The BMC determines whether the AI card plugged into the PCIe slot is normally present or abnormally present according to the board status of the AI card plugged into the PCIe slot, including: If the board status of the AI card plugged into the PCIe slot is the ready state or working state, the BMC determines that the AI card plugged into the PCIe slot is normally present; If the board status of the AI card plugged into the PCIe slot is the alarm state, the BMC determines that the AI card plugged into the PCIe slot is abnormally present.

5. The method according to claim 3, characterized in that, The information returned by the AI card plugged into the PCIe slot received by the BMC further includes the board type, operating status, overall board power, overall board current, input voltage, temperature, and fan speed. The information of the AI card plugged into the PCIe slot further includes the loaded state of the AI card plugged into the PCIe slot and data related to the PCIe slot for plugging in the AI card. Among them, the data related to the PCIe slot includes the slot number, board type, operating status, overall board power, overall board current, input voltage, temperature, and fan speed of the PCIe slot. Correspondingly, the BMC reads the registers of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through a multi-channel I2C switch, and the information obtained about the AI cards plugged into the PCIe slots with each power supply voltage within the normal range further includes: After the BMC obtains the data related to the PCIe slot for plugging in the AI card, it determines the loaded state of the AI card according to the overall board power.

6. The method according to claim 5, characterized in that, The one external interaction device includes: An LCD for displaying the information of the AI cards plugged into each PCIe slot. An I2C-to-parallel port chip that receives the display data of the information of the AI cards plugged into each PCIe slot sent by the BMC through the I2C bus and transmits the display data of the information of the AI cards plugged into each PCIe slot in parallel to the LCD for the LCD to display.

7. The method according to claim 6, characterized in that, The one external interaction device further includes: A key; A parallel port-to-I2C chip for receiving the key input information when the user touches the key and transmitting the key input information to the BMC through the I2C bus.

8. The method according to claim 7, characterized in that, The method further includes: When the timing time period of key detection by the BMC arrives, the BMC scans and saves the key input information from the one external interaction device.

9. The method according to claim 8, characterized in that, The BMC displays the information of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through an external interaction device, including: When the BMC displays the in-position normal state, in-position abnormal state, or out-of-position state of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range through the LCD, it determines whether it has received key input information. If no key input information is received, the BMC keeps the display content of the LCD. If key input information is received, the BMC switches the display content of the LCD to the loaded state of the AI cards plugged into the PCIe slots with each power supply voltage within the normal range or the data related to the PCIe slots with the power supply voltage within the normal range corresponding to a certain AI card according to the indication of the key input information.

10. The method according to any one of claims 1-9, characterized in that, The one external interaction device is installed on the front panel or the rear panel of the chassis of the server and is connected to the BMC through the I2C bus.

Images