Hi Geoff,
On 08/08/2016:10:19:33 PM, Geoff Levand wrote:
> Add kexec reboot support for ARM64 platforms.
>
> Signed-off-by: Geoff Levand <geoff at infradead.org>
> ---
> +
> + arm64_mem.page_offset = ehdr.e_entry - arm64_mem.text_offset;
> +
> + dbgprintf("%s: e_entry: %016llx -> %016lx\n", __func__,
> + ehdr.e_entry,
> + virt_to_phys(ehdr.e_entry));
> + dbgprintf("%s: p_vaddr: %016llx -> %016lx\n", __func__,
> + phdr->p_vaddr,
> + virt_to_phys(phdr->p_vaddr));
> + dbgprintf("%s: header_offset: %016lx\n", __func__,
> + header_offset);
> + dbgprintf("%s: text_offset: %016lx\n", __func__,
> + arm64_mem.text_offset);
> + dbgprintf("%s: image_size: %016lx\n", __func__,
> + arm64_mem.image_size);
> + dbgprintf("%s: phys_offset: %016lx\n", __func__,
> + arm64_mem.phys_offset);
> + dbgprintf("%s: page_offset: %016lx\n", __func__,
> + arm64_mem.page_offset);
Although it works on mustang, I would like to understand few things:
This is what is being printed by kexec-tools:
elf_arm64_load: e_entry: fffffc0008080000 -> 0000004000080000
elf_arm64_load: p_vaddr: fffffc0008080000 -> 0000004000080000
elf_arm64_load: header_offset: 0000000000000000
elf_arm64_load: text_offset: 0000000000080000
elf_arm64_load: image_size: 00000000013a0000
elf_arm64_load: phys_offset: 0000004000000000
elf_arm64_load: page_offset: fffffc0008000000
kernel is having 42 VA Bits, so a debug print in kernel says that
PAGE_OFFSET=fffffe0000000000. Notice the difference between the values
calculated by kernel and kexec-tools. Isn't it confusing? I think kexec-tools
page_offset is text_page_offset and not the one which kernel uses. So, probably
we can rename it accordingly. Otherwise, we may sync kexec-tools's
virt_to_phys() and vice-versa definitions with that of kernel definitions.
In my case, kernel has kimage_voffset=fffffbc008000000. So kernel translate
virtual address fffffc0008080000 as fffffc0008080000 - fffffbc008000000 =
0000004000080000 which is matching to the one calculated by kexec-tools (for
e_entry).
~Pratyush
