On Mon, Mar 08, 2021 at 12:10:33PM +0100, Bjørn Mork wrote:
> Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx> writes:
>
> > On Sun, Mar 07, 2021 at 07:23:01PM +0100, Bjørn Mork wrote:
> >> The devicetree specification requires 8-byte alignment in
> >> memory. This is now enforced by libfdt since commit 79edff12060f
> >> ("scripts/dtc: Update to upstream version v1.6.0-51-g183df9e9c2b9")
> >> which included the upstream commit 5e735860c478 ("libfdt: Check for
> >> 8-byte address alignment in fdt_ro_probe_()").
> >>
> >> This broke the MIPS raw appended DTBs which would be appended to
> >> the image immediately following the initramfs section. This ends
> >> with a 32bit size, resulting in a 4-byte alignment of the DTB.
> >>
> >> Fix by padding with zeroes to 8-bytes when MIPS_RAW_APPENDED_DTB
> >> is defined.
> >>
> >> Cc: Rob Herring <robh+dt@xxxxxxxxxx>
> >> Cc: Frank Rowand <frowand.list@xxxxxxxxx>
> >> Signed-off-by: Bjørn Mork <bjorn@xxxxxxx>
> >> ---
> >> arch/mips/kernel/vmlinux.lds.S | 5 +++++
> >> 1 file changed, 5 insertions(+)
> >
> > thank you for your patch, but there already was a fix for the problem
> > pending from Paul, which I've applied to mips-fixes a few minutes ago.
>
> Yes, I see. That does look much nicer. But I don't think it addresses
> the problem with an uncompressed kernel? Could we have the padding in
> vmlinux.lds.S as well? Or some other solution to ensure that it is
> possible to cat the DTB to the end of vmlinux.bin without manually
> aligning it to 8 bytes?
I see
diff --git a/arch/mips/kernel/vmlinux.lds.S b/arch/mips/kernel/vmlinux.lds.S
index c1c345be04ff..4b4e39b7c79b 100644
--- a/arch/mips/kernel/vmlinux.lds.S
+++ b/arch/mips/kernel/vmlinux.lds.S
@@ -145,6 +145,7 @@ SECTIONS
}
#ifdef CONFIG_MIPS_ELF_APPENDED_DTB
+ STRUCT_ALIGN();
.appended_dtb : AT(ADDR(.appended_dtb) - LOAD_OFFSET) {
*(.appended_dtb)
KEEP(*(.appended_dtb))
@@ -172,6 +173,7 @@ SECTIONS
#endif
#ifdef CONFIG_MIPS_RAW_APPENDED_DTB
+ STRUCT_ALIGN();
__appended_dtb = .;
/* leave space for appended DTB */
. += 0x100000;
in that patch, and IMHO this does align the appended_dtb. What do I miss ?
Thomas.
--
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea. [ RFC1925, 2.3 ]
