Calculation device training method and calculation system
By performing the second training of low-standard specifications on PCIe devices that have not passed the first link training, the problem that PCIe devices cannot be recognized by the host is solved, and the device recognition success rate and the normal operation of the computing system are improved.
Patent Information
- Application Number
- CN202411207733.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-12-27
AI Technical Summary
In a computing system, PCIe devices with qualified quality may not be properly recognized by the host, resulting in inability to cooperate with the system, and even the computing system cannot operate normally.
The second link training is performed to improve the device recognition success rate by setting the expansion slot of the PCIe device that has not passed the first link training to the low-end connection specification.
The PCIe device that has successfully passed the second training can be recognized and operated normally by the host, which improves the device's recognition success rate and ensures that the computing system can start normally.
Smart Images
Figure CN119201812A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to a computing device training method and a computing system, and in particular to a PCIe device training method and a computing system equipped with a PCIe device. Background Art
[0002] The computing system is configured to cooperate with peripheral devices and / or peripheral devices of various specifications to jointly provide the functions required by the computing system. During the startup process, the computing system performs link training on each peripheral device and / or peripheral device, and the devices that pass the training will be recognized by the system and can be used in conjunction with the system. In some cases, peripheral devices that are of qualified quality may not be properly recognized by the system, resulting in these peripheral devices being unable to be used with the system, and may even cause the computing system itself to not function properly. For example, for a server configured with one or more PCIe devices, the failure of the server host to link train the one or more PCIe devices may cause the PCIe devices that are of qualified quality to be unable to be recognized by the host, making the PCIe devices that are of qualified quality unable to be used in conjunction with the host. Therefore, it is necessary to provide a technical solution to improve the recognition success rate of peripheral devices such as PCIe devices. Summary of the Invention
[0003] According to one aspect, the present application provides a computing device training method for identifying a computing device connected to a host, such as one or more PCIe devices connected to an expansion slot of the host. The method according to this embodiment includes powering on the host and performing initial link training on the PCIe devices connected to the host according to preset specifications. If the initial training fails to identify one or more of the PCIe devices, the expansion slots where the one or more PCIe devices are located are set to a low-profile connection specification. For example, the expansion slots where the one or more PCIe devices are located are set to a first-generation PCIe specification.
[0004] The host then performs a second link training on the one or more PCIe devices that failed the first training according to the set low-profile specifications. The one or more PCIe devices that passed the second training can be recognized by the host, so that the one or more PCIe devices can immediately cooperate with the host to operate. At the same time, the link where the one or more PCIe devices that failed the first training but passed the second training according to the low-profile specifications are located is recorded as a low-profile specification. The host can then enter the normal startup step. According to the above method, the technical solution provided by the present application can re-identify PCIe devices that are of qualified quality but failed the first training by the host.
[0005] According to another aspect, the present application provides a computing system comprising a host, one or more computing devices, such as PCIe devices, and device links respectively connected between the host and each PCIe device. The host is configured to execute the following method for identifying the PCIe device: powering on the host, and performing a first training on each PCIe device according to a preset specification. If the PCIe device passes the first training, the method executes host startup. If the PCIe device fails the first training, the method executes the following steps: setting the device link to a low-profile specification; performing a second training on the PCIe device according to the low-profile specification; and if the PCIe device passes the second training, the method executes host startup. BRIEF DESCRIPTION OF THE DRAWINGS
[0006] The following provides a detailed description of the embodiments of the present application with reference to the accompanying drawings, in which:
[0007] Figure 1 is a flowchart of a computing device training method according to one embodiment of the present application;
[0008] Figure 2 is a flowchart of a computing device training method according to another embodiment of the present application;
[0009] Figure 3 This is a block diagram of a computing system architecture according to one embodiment of the present application. DETAILED DESCRIPTION
[0010] like Figure 1 As shown, a computing device training method 100 according to one embodiment of the present application includes, in step 112, powering on the host. In step 114, the host performs initial link training on one or more expansion cards connected to the expansion slot, such as PCIe devices, according to the preset specifications of each PCIe device. For example, for a fifth-generation PCIe device (PCIe5.0), the initial training is performed with the fifth-generation card as the preset specification, for a fourth-generation PCIe device (PCIe4.0), the initial training is performed with the fourth-generation card as the preset specification, and so on. If all PCIe devices connected to the host have passed the initial link training, the method 100 executes host startup in step 130.
[0011] If one or more of the PCIe devices fail the initial link training, the method 100 reports in step 116 that the one or more PCIe devices are in a state of failing the initial link training. In step 118, the method 100 sets the link where the one or more PCIe devices that failed the initial link training are located to a low-profile specification. For example, for PCIe devices of version 2.0 or above, i.e., PCIe devices of versions PCIe2.0 to PCIe5.0, or PCIe devices of version higher than PCIe5.0, the link where the one or more PCIe devices that failed the initial link training and are PCIe2.0 or above are located is set to a configuration specification that is one or more levels lower than the corresponding PCIe device version. For example, the link where all the one or more PCIe devices that failed the initial link training and are PCIe2.0 or above are located is set to a configuration that complies with the PCIe1.0 version specification (first generation PCIe specification). Alternatively, for a PCIe5.0 version card, the link where it is located is set to a configuration that complies with any version specification between PCIe1.0 and PCIe4.0. At step 120 , the method 100 marks the one or more PCIe devices that failed the initial link training as being in an initial link training failure state.
[0012] Method 100 then performs a second link training at step 122 for one or more PCIe devices that failed the initial link training, using the low-profile specifications set at step 118. Because the second link training utilizes the low-profile specifications, the link training pass rate for qualified PCIe devices connected to each link can be increased. PCIe devices that successfully pass the second link training can be recognized by the host and can operate with the host.
[0013] For one or more PCIe devices that failed the initial link training but passed the second link training according to the low-profile specification, the method 100 marks the one or more PCIe devices as being in the second link training success state at step 124, and records the one or more PCIe devices as being in the low-profile state at step 126. The method 100 then proceeds to the host startup step 130.
[0014] For a host connected to multiple PCIe devices, if one or more of the PCIe devices pass the first link training, and one or more of the other PCIe devices fail the second link training, the method 100 executes the subsequent process 200. Figure 2As shown, the subsequent process 200 includes, in step 212, performing a host hot reboot, followed by checking the PCIe device training status of each device link in step 214. A link with a training status marked "0" is a link where a PCIe device previously passed training, and a link with a training status marked "1" is a link where a PCIe device previously failed training. For PCIe devices connected to a link with a training status marked "0," the subsequent process 200 performs a third link training in step 218. For PCIe devices connected to a link with a training status marked "1," the subsequent process 200 sets the device link that failed the second training to a low-profile configuration in step 216, and then performs a third link training in step 218. If all PCIe devices pass the third link training, the subsequent process 200 performs a host boot in step 250.
[0015] For PCIe devices that fail the third link training, the subsequent process 200 reports the device that failed the third training in step 232, sets the link where the device failed the third training to a lower specification, such as the first generation PCIe specification, in step 234, and marks the device as a training failure in step 236. The subsequent process 200 then performs a fourth training on the PCIe device that failed the third training according to the set lower specification in step 238.
[0016] If the PCIe device that failed the third training passes the fourth training, the subsequent process 200 marks the device that passed the fourth training as a training success state in step 240, and records the device that passed the fourth training as a low-specification state in step 242, and then performs host startup in step 250.
[0017] For a PCIe device that still fails the fourth training, process 200 marks the device as faulty at step 262, preventing subsequent link training from being performed on the device. If the host has more devices connected to it besides the faulty device, such as PCIe devices connected to other expansion slots, process 200 returns to step 212 to repeat the previous link training steps for these PCIe devices connected to the other expansion slots.
[0018] According to another aspect, the present application provides a computing system. Figure 3As shown, a computing system 300 according to one embodiment of the present application includes a host 320 and one or more peripheral interfaces, such as eight expansion slots 331, 332, 333, 334, 335, 336, 337, and 338. Each expansion slot is connected to the host 320 via a respective link 341, 342, 343, 344, 345, 346, 347, and 348. One or more of the eight expansion slots, such as expansion slots 331, 332, and 333, is equipped with a computing device, such as a PCIe card 351, 352, and 353. As a specific non-limiting example, PCIe cards 351, 352, and 353 are PCIe cards of Generation 5 (PCIe 5.0), Generation 4 (PCIe 4.0), and Generation 3 (PCIe 3.0), respectively.
[0019] The computing system 300 is configured to execute the following method for enabling the host 320 to identify the PCIe cards 351, 352, and 353. Figure 1 and Figure 2 In the illustrated method, after the host is powered on, the host performs initial training on each PCIe card 351, 352, and 353 according to preset specifications. For example, the initial training of PCIe card 351 is performed according to the PCIe 5.0 specification, the initial training of PCIe card 352 is performed according to the PCIe 4.0 specification, and the initial training of PCIe card 353 is performed according to the PCIe 3.0 specification.
[0020] For PCIe card 351 that passes the initial training according to the PCIe 5.0 specification, system 300 records PCIe card 351 as being in the successful initial training state and records the expansion slot 341 in which it is located as being in the PCIe 5.0 specification state. If PCIe cards 352 and 353 fail the initial training, system 300 reports the initial training failure state of the PCIe cards 352 and 353 and sets the links 342 and 343 of the PCIe cards 352 and 353 to a lower specification, for example, setting the links 342 and 343 of the PCIe cards 352 and 353 to the PCIe 1.0 specification, and then performs a second training on the PCIe cards 352 and 353 that failed the initial training according to the lower specification.
[0021] Lowering the configuration specifications will increase the success rate of the second training. For example, if PCIe cards 352 and 353 are qualified devices, lowering the configuration specifications will increase the chances of each device passing the second training. If the second training is successful, the qualified PCIe cards 352 and 353 will be verified and can be successfully recognized by the host 320. Accordingly, the system 300 marks the links 342 and 343 of the PCIe cards 352 and 353 as having a lower configuration, so that the PCIe cards 352 and 353 operate according to the lower configuration specifications with the host 320, thus realizing the functions of the system 300.
[0022] If PCIe card 352 passes the second training according to the configured minimum specifications, but PCIe card 353 still fails the second training, system 300 can determine that PCIe card 353 itself is in a faulty state and mark it accordingly. After the fault in PCIe card 353 is corrected or a new PCIe card is replaced, the system can re-run the training according to the aforementioned steps. Alternatively, the system can continue to operate even if PCIe card 353 fails the second training while waiting for the fault to be corrected or replaced.
[0023] The present application has been presented above for purposes of illustration and description, but is not intended to be exhaustive or limiting. Many modifications and variations will be apparent and readily understood by those skilled in the art. The exemplary embodiments have been chosen and described in order to explain the principles and practical applications and to enable those skilled in the art to understand the various embodiments of the present application.
[0024] Therefore, although illustrative example embodiments are described herein with reference to the accompanying drawings, it should be understood that the description is not limiting and that those skilled in the art may conceive of various alternatives and / or modifications based on the described technical solutions without departing from the scope of the present disclosure.
Claims
1. A computing device training method, characterized in that: The method comprises: Power on the host; Performing an initial training on a computing device connected to the host, wherein the initial training is performed according to a preset specification; If the computing device passes the initial training, the method performs a host boot; If the computing device fails the first training, the method performs the following steps: Setting the device link of the computing device to a low-profile specification; performing a second training on the computing device according to the low-configuration specification; If the computing device passes the second training, the method performs host booting.
2. The method according to claim 1, characterized in that The method further includes marking the computing device as being in a first training failure state after the computing device fails the first training.
3. The method according to claim 2, characterized in that Also includes: After the computing device passes the second training, the computing device is marked as being in a second training success state.
4. The method according to claim 3, characterized in that Also includes: After marking the computing device as being in the second training success state, marking the computing device as being in the low-specification state.
5. The method according to claim 2, characterized in that: After the computing device fails the second training, a warm reboot is performed on the host.
6. The method according to claim 5, characterized in that The computing device includes a first computing device and a second computing device, wherein the first computing device passes the first training state, and the second computing device fails the first training state and fails the second training state, and the method further includes: checking the training status of the first computing device and the second computing device; Performing a third training on the first computing device and the second computing device, wherein the third training on the first computing device is performed according to the preset specification, and the third training on the second computing device is performed according to the low-profile specification; If the first computing device and the second computing device pass the third training, the method executes host startup; If the first computing device fails the third training, the method performs the following steps: Setting the device link of the first computing device to a low-profile specification; performing fourth training on the first computing device according to the low-configuration specification; If the first computing device passes the fourth training, the method performs host booting.
7. The method according to claim 6, characterized in that The method further includes marking the first computing device as being in a training failure state after the first computing device fails the third training.
8. The method according to claim 7, characterized in that Also includes: After the first computing device passes the fourth training, the first computing device is marked as being in a training success state.
9. The method according to claim 6, characterized in that Also includes: After marking the first computing device as being in a training success state, the first computing device is recorded as being in a low-specification state.
10. The method according to claim 6, characterized in that The computing device further includes a third computing device, wherein the third computing device is in a state of passing the first training and the third training, and the method includes, after the first computing device fails the fourth training, performing a hot restart on the host.
11. A computing system, comprising a host, a computing device, and a device link connected between the host and the computing device, wherein the host is configured to perform the following method for identifying the computing device: Powering on the host; Performing an initial training on the computing device, wherein the initial training is performed according to a preset specification; If the computing device passes the initial training, the method performs a host boot; If the computing device fails the first training, the method performs the following steps: Setting the device link to a low-profile specification; performing a second training on the computing device according to the low-configuration specification; If the computing device passes the second training, the method performs booting of the host.
12. The computing system according to claim 11, characterized in that: The host is further configured to mark the computing device as being in a first training failure state after the computing device fails the first training.
13. The computing system according to claim 12, characterized in that: The host is further configured to mark the computing device as being in a second training success state after the computing device passes the second training.
14. The computing system according to claim 13, characterized in that: The host is also configured to mark the computing device as being in a low-specification state after marking the computing device as being in a second training success state.
15. The computing system according to claim 12, characterized in that: The host is further configured to perform a warm restart after the computing device fails the second training.
16. The computing system according to claim 15, characterized in that: The computing device includes a first computing device and a second computing device, wherein the first computing device passes the first training state, and the second computing device fails the first training state and fails the second training state, and the host is further configured to execute the following method: checking the training status of the first computing device and the second computing device; Performing a third training on the first computing device and the second computing device, wherein the third training on the first computing device is performed according to the preset specification, and the third training on the second computing device is performed according to the low-profile specification; If the first computing device and the second computing device pass the third training, starting the host; If the first computing device fails the third training, the method performs the following steps: Setting the device link of the first computing device to a low-profile specification; performing fourth training on the first computing device according to the low-configuration specification; If the first computing device passes the fourth training, the host is started.
17. The computing system according to claim 16, characterized in that: The host is further configured to mark the first computing device as being in a training failure state after the first computing device fails the third training.
18. The computing system according to claim 17, characterized in that: The host is further configured to mark the first computing device as being in a training success state after the first computing device passes the fourth training.
19. The computing system of claim 16, wherein: The host is further configured to record the first computing device as being in a low-specification state after marking the first computing device as being in a training success state.
20. The computing system of claim 16, wherein: The computing device further includes a third computing device, wherein the third computing device is in a state of passing the first training and the third training, and the host is further configured to perform a hot restart after the first computing device fails the fourth training.