Method to coordinate system boot and reset flows and improve reliability, availability and serviceability (RAS) among multiple chipsets

Patent metadata

Field	Value
Publication number	US-10620966-B2
Application number	US-201715719340-A
Country	US
Kind code	B2
Filing date	Sep 28, 2017
Priority date	Sep 28, 2017
Publication date	Apr 14, 2020
Grant date	Apr 14, 2020

Abstract

Embodiments disclosed herein relate to coordinated system boot and reset flows and improve reliability, availability, and serviceability (RAS) among multiple chipsets. In an example, a system includes a master chipset having multiple interfaces, each interface to connect to one of a processor and a chipset, at least one processor connected to the master chipset, at least one non-master chipset connected to the master chipset, and a sideband messaging channel connecting the master chipset and the non-master chipsets, wherein the master chipset is to probe a subset of its multiple interfaces to discover a topology of connected processors and non-master chipsets, and use the sideband messaging channel to coordinate a synchronized boot flow.

First claim

What is claimed is:

1. A system comprising:
a master chipset having multiple interfaces, each interface to connect to one of a processor and a chipset;
at least one processor connected to the master chipset;
at least one non-master chipset connected to the master chipset; and
a sideband messaging channel connecting the master chipset and the non-master chipsets;
wherein the master chipset is to probe a subset of its multiple interfaces to discover a topology of connected processors and non-master chipsets, and use the sideband messaging channel to coordinate a synchronized boot flow, and wherein the master chipset is further to use the sideband messaging channel to coordinate a synchronized reset flow, the synchronized reset flow comprising a graceful reset when a reset scenario allows time for a handshake among the chipsets, and an ungraceful reset otherwise.

2. The system of claim 1 , wherein the master chipset is to coordinate the synchronized boot flow by:
receiving a synced wake signal on the sideband messaging channel;
preparing to boot;
using the sideband messaging channel to prompt the non-master chipsets to prepare to boot;
receiving indications of readiness to boot from the non-master chipsets on the sideband messaging channel;
using the sideband messaging channel to instruct the non-master chipsets to boot;
receiving indications of boot completion from the non-master chipsets on the sideband messaging channel; and
using the sideband messaging channel to issue a synced go message to the non-master chipsets;
wherein each of the non-master chipsets deasserts a partition reset signal.

3. The system of claim 1 , wherein, in the graceful reset scenario, the master chipset is to coordinate the synchronized reset flow among all chipsets and processors by:
receiving a synced reset request and a reset type on the sideband messaging channel;
using the sideband messaging channel to instruct the non-master chipsets to reset, with the reset type;
receiving indications of readiness to reset from the non-master chipsets on the sideband messaging channel; and
using the sideband messaging channel to send a synced reset message to the non-master chipsets;
wherein each of the non-master chipsets asserts a partition reset signal.

4. The system of claim 1 , wherein, in the ungraceful reset scenario, the master chipset is to immediately send the synced reset request to all chipsets on the sideband messaging channel in order to serve as a warning of an imminent reset.

5. The system of claim 1 , wherein the master chipset is further to use the sideband messaging channel to coordinate a synchronized power state transition flow by:
receiving a power state transition request and a target power state on the sideband messaging channel;
using the sideband messaging channel to instruct the non-master chipsets to do the power state transition, with the target power state;
receiving indications on the sideband messaging channel from the non-master chipsets of readiness to transition power state; and
using the sideband messaging channel to send a synced power state transition message to the non-master chipsets;
wherein each of the non-master chipsets implements the target power state.

6. The system of claim 5 , wherein the master chipset and its connected processors and non-master chipsets comprise a first partition, wherein the system further comprises at least one additional partition comprising an additional non-master chipset coupled to at least one additional processor, wherein the master chipset further coordinates the boot flow, the reset flow, and the power state transition flow of the additional non-master chipset.

7. The system of claim 1 , wherein the sideband messaging channel further comprises a multiple-driver sync wire comprising a wire coupled to a weak pull-up resistor and which is conditionally discharged by one of multiple open-drain transistors, multiple open-collector transistors, and multiple tristate buffers.

8. The system of claim 1 , wherein the master chipset, when coordinating the synchronized boot flow to boot from a warm reset state, is already aware of the topology and is to skip probing the subset of its interfaces to discover the topology.

9. The system of claim 1 , further comprising a printed circuit board with a fixed number of processor sockets, wherein the at least one processor is inserted into one of the fixed number of processor sockets, and wherein additional processors can be inserted into other ones of the fixed number of processor sockets.

10. A method comprising:
probing, by a master chipset having multiple interfaces, each interface to connect to one of a processor and a chipset, the master chipset further connected to one or more non-master chipsets by a sideband messaging channel, a subset of the multiple interfaces to discover a topology of connected processors and non-master chipsets; and
using, by the master chipset, the sideband messaging channel to coordinate a synchronized boot flow with the one or more non-master chipsets;
wherein using the sideband messaging channel to coordinate the synchronized boot flow comprises the master chipset:
receiving a synced wake signal on the sideband messaging channel;

preparing to boot;
using the sideband messaging channel to prompt the non-master chipsets to prepare to boot;
receiving indications of readiness to boot from the non-master chipsets on the sideband messaging channel;
using the sideband messaging channel to instruct the non-master chipsets to boot;
receiving indications of boot completion from the non-master chipsets on the sideband messaging channel; and
using the sideband messaging channel to issue a synced go message to the non-master chipsets.

11. The method of claim 10 , further comprising using, by the master chipset, the sideband messaging channel to coordinate a synchronized reset flow, the synchronized reset flow comprising a graceful reset when a reset scenario allows time for a handshake among the chipsets, and an ungraceful reset otherwise;
wherein coordinating the graceful synchronized reset flow comprises the master chipset receiving a synced reset request and a reset type on the sideband messaging channel, using the sideband messaging channel to instruct the non-master chipsets to reset, with the reset type, receiving indications of readiness to reset from the non-master chipsets on the sideband messaging channel, and using the sideband messaging channel to send a synced reset message to the non-master chipsets; and
wherein coordinating the ungraceful synchronized reset flow comprises the master chipset receiving a synced reset request and a reset type on the sideband messaging channel, and using the sideband messaging channel to immediately transmit the synced reset message to the non-master chipsets.

12. A method comprising:
using, by a non-master chipset connected to a master chipset and to at least one processor, a sideband messaging channel connected to the master chipset to exchange messages to coordinate a synchronized boot flow, by:
waiting for and receiving a prompt to prepare to boot;
preparing to boot and indicating readiness to boot;
waiting for and receiving a prompt to boot;
booting and indicating boot completion; and
waiting for and receiving a synced go message on the sideband messaging channel;
wherein the non-master chipset, in response to the synced go message, is to deassert a reset signal on its partition.

13. The method of claim 12 , further comprising:
using, by the non-master chipset, the sideband messaging channel to exchange messages with the master chipset to coordinate an ungrac

Assignees

• Intel Corp

Inventors

• Zhong Tina C
• Wunderlich Russell J
• Chen Chih-Cheh
• Trivedi Malay

Classifications

• G06F1/24 Primary

  Resetting means · CPC title

• G06F9/4418 Primary

  Suspend and resume; Hibernate and awake · CPC title