On 11/8/21, Dave Hansen <dave.hansen@xxxxxxxxx> wrote: > But, you make a good point about the alternate approach. I don't have a > preference either way. I was going to do this, and remembered why I discarded this option. There is already a function to map EFI memmap entries to the e820_table called do_add_efi_memmap, but it is only called when you add the kernel param "add_efi_memmap", or when the system has some soft reserved entries in the EFI memmap. Do you know why there is such kernel param? Why would you not want to have also the EFI memmap? since e820 already has the mechanisms to deal with overlaps. I first thought that it was a size issue, maybe the e820_table couldn't hold both memmaps at the same time. But reading the comment where the e820_table is defined makes think that's not an issue. Am I missing something?
