commit | 0dede335cd31592aa7e0299e9b4d40054e5ffd02 | [log] [tgz] |
---|---|---|
author | Patrick Venture <venture@google.com> | Mon Sep 17 15:28:35 2018 -0700 |
committer | Patrick Venture <venture@google.com> | Mon Sep 17 15:30:51 2018 -0700 |
tree | 17aa8afe64793eb6e30ec8562560c63b65462927 | |
parent | c7d5026c6a24904c3c036e0d846e0c2b245518b2 [diff] |
fixup: CI is x86-64, build is 32, cast down for debug The fprintfs were promoted to %lu to make CI work, which then broke the ARM build. Therefore need to cast to make it happy in both environments. Change-Id: Ibba14d8aa19aca44c9ce9b65b66dd79659ce126e Signed-off-by: Patrick Venture <venture@google.com>
There are and will be a variety of sys specific commands.
The cablecheck command checks whether the BMC is seeing traffic between itself and the host's NIC. Sys specifies which if_name is expected to be connected. The BMC presently only checks traffic on the interface specified. There are now ethernet statistics available over IPMI, which can be checked directly in lieu of this.
Request
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x00 | Subcommand |
0x01 | If_name length | Where you expect the cable, eth0 or eth1, etc |
0x02 ... | The name | The string, not null-terminated |
Response
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x00 | Subcommand |
0x01 | 0x00/0x01 | 0 for false, 1 for true |
Any CPLD on the system that can only be read directly via the BMC can have its version exported to Sys via the cpld version command.
Request
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x01 | Subcommand |
0x01 | CPLD ID | A one-byte identifier for the CPLD file to read, unsigned byte. |
Response
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x01 | Subcommand |
0x01 | Major | Major version |
0x02 | Minor | Minor Version |
0x03 | Sub 1 | Third version number |
0x04 | Sub 2 | Fourth version number |
Per the above, if the version number doesn't fit in a byte it'll be cast to size.
The BMC itself must have hard-coded into the image, which ethernet device is connected to the host NIC. This is true also in the mapping of ethernet device to channel number. The channel number is used to configure the ethernet device over IPMI, instead of the interface name. This is because we leverage the current IPMI command set to read and write the networking configuration. Sys can be programmed already to have this information in the board protobuf, however, this information -- can be read from the BMC over IPMI.
Request
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x02 | Subcommand |
Response
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x02 | Subcommand |
0x01 | Channel number | The IPMI channel number for use with the network configuration commands (such as reading the MAC or IP address of the BMC). |
0x02 | if_name length | The length of the if_name in bytes. |
0x03... | if_name | The interface name, not null-terminated |
Sys needs to be able to tell the BMC to reset the host but given a delay in seconds.
Request
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x03 | Subcommand |
0x01..0x04 | Seconds to delay (uint32) |
Response
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x03 | Subcommand |
It's a security requirement that the BMC must be able to persist a setting to disable plaintext console transmission via Nemora. The setting is stored in the BMC read-write filesystem, and Nemorad has a utility class to read and write that setting, which this OEM command uses. The command can be used for read-only as well as a read-then-write to ensure the state of the file.
For now, there is only one Subcommand; the byte is there in case we want to extend this command later.
Request
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x00 | Subcommand |
0x00 | want_to_set | 1 if a set operation is desired; 0 if only reading. |
0x01 | desired_value | 1 if plaintext should be allowed; 0 if plaintext should be disallowed; ignored if want_to_set is 0. |
Response
Byte(s) | Value | Data |
---|---|---|
0x00 | 0x00 | Subcommand |
0x00 | Current value | 1 if plaintext is currently allowed; 0 if plaintext is currently disallowed |