Hi all, The numbers no longer make any sense since I either did not correctly understand the feedback being given, or dramatically changed the approach. This version introduces an architecture symbol: ARCH_HAS_PHYS_INITRD which indicates whether the architecture cares/supports parsing the physical address of the initrd. Currently ARM (32-bit), Unicore32 and now ARM64 support that. When that symbol is defined, we also have the generic FDT code populate the initrd physical address and size, and we can later make use of that within architecture specific code to populate the memblock regions and do the righ physical to virtual address conversion. Rob, hopefully this is what you had in mind. Previous discussions/submissions list here: v3: https://www.spinics.net/lists/arm-kernel/msg683566.html v2: https://lkml.org/lkml/2018/10/25/4 Florian Fainelli (6): nds32: Remove phys_initrd_start and phys_initrd_size arch: Make phys_initrd_start and phys_initrd_size global variables arch: Define ARCH_HAS_PHYS_INITRD for ARM and Unicore32 of/fdt: Populate phys_initrd_start/phys_initrd_size from FDT arm64: Utilize ARCH_HAS_PHYS_INITRD of/fdt: Remove definition check for __early_init_dt_declare_initrd() arch/Kconfig | 7 +++++++ arch/arm/Kconfig | 1 + arch/arm/mm/init.c | 6 +++--- arch/arm64/Kconfig | 1 + arch/arm64/include/asm/memory.h | 8 -------- arch/arm64/mm/init.c | 23 +++++++++++------------ arch/nds32/mm/init.c | 2 -- arch/unicore32/Kconfig | 1 + arch/unicore32/mm/init.c | 4 ++-- drivers/of/fdt.c | 6 ++++-- include/linux/initrd.h | 3 +++ 11 files changed, 33 insertions(+), 29 deletions(-) -- 2.17.1
