In general, I agree this solution that lets SMBIOS driver directly absorbs the SMBIOS table from PEI.
This can eliminate the needs of a separate driver that consumes the HOB and calls SMBIOS protocol to add the SMBIOS structures.
There are two options for the HOB design:
1. A single HOB that points to the SMBIOS table.
2. Multiple HOBs that each points to a SMBIOS structure.
In my opinion, option #2 is more flexible because it doesn't require the bootloader to consolidate all the SMBIOS structures together.
The CPU module in the bootloader can produce the type 4 and 7 structures.
The PCI module in the bootloader can produce the type 9 structures.
But, I am not sure if option #2 is conflict with what coreboot does. Does coreboot produce the whole SMBIOS table in a single buffer?
Option #2 also doesn't care whether it's a SMBIOS 3.0 table or 2.x table.
+ Status = EfiGetSystemConfigurationTable (1. Why don't you directly get the data from HOB list? This can eliminate the code in BlSupportDxe that gets data in HOB and publishes to
+ValidateSmbios20Table(2. I will defer to experts (Dandan, Star and Zhichao) to review whether the above two functions are implemented properly.
3. Please use SMBIOS_TYPE_END_OF_TABLE instead of hardcode 127.
4. Should we copy from Smbios30Table->TableAddress instead of Smbios30Table?
5. Can you explain in specific why SMBIOS table should be duplicated before parsing?