Rob Herring [off-list ref] writes:

On Thu, Feb 18, 2021 at 5:34 AM Stephen Rothwell [off-list ref] wrote:

quoted

On Thu, 18 Feb 2021 21:44:37 +1100 Michael Ellerman [off-list ref] wrote:

quoted

I think it just needs this?

diff --git a/arch/powerpc/kexec/elf_64.c b/arch/powerpc/kexec/elf_64.c
index 87e34611f93d..0492ca6003f3 100644
--- a/arch/powerpc/kexec/elf_64.c
+++ b/arch/powerpc/kexec/elf_64.c

@@ -104,7 +104,7 @@ static void *elf64_load(struct kimage *image, char *kernel_buf,

      fdt = of_kexec_alloc_and_setup_fdt(image, initrd_load_addr,
                                         initrd_len, cmdline,
-                                        fdt_totalsize(initial_boot_params));
+                                        kexec_fdt_totalsize_ppc64(image));
      if (!fdt) {
              pr_err("Error setting up the new device tree.\n");
              ret = -EINVAL;

I thought about that, but the last argument to
of_kexec_alloc_and_setup_fdt() is extra_fdt_size and the allocation
done is for this:

fdt_size = fdt_totalsize(initial_boot_params) +
                   (cmdline ? strlen(cmdline) : 0) +
                   FDT_EXTRA_SPACE +
                   extra_fdt_size;

and kexec_fdt_totalsize_ppc64() also includes
fdt_totalsize(initial_boot_params) so I was not sure.  Maybe
kexec_fdt_totalsize_ppc64() needs modification as well?

You're both right. Michael's fix is sufficient for the merge. The only
risk with a larger size is failing to allocate it, but we're talking
only 10s of KB. Historically until the commit causing the conflict,
PPC was just used 2x fdt_totalsize(initial_boot_params). You could
drop 'fdt_size = fdt_totalsize(initial_boot_params) + (2 *
COMMAND_LINE_SIZE);' from kexec_fdt_totalsize_ppc64() as well, but
then the function name is misleading.

Lakshmi can send a follow-up patch to fine tune the size and rename
kexec_fdt_totalsize_ppc64.

Sounds good.

cheers