[PATCH] arm: Backward compatibility to U-Boot v2020.04

[Dorde Stojicevic]

Hi guys,

and thanks for your fast response.

Sorry for the formatting issue, I am bound to our company Outlook, which hard-codes our signatures, so this one will not be possible to take out. I am not quite sure git send-email would work, for the same reason + corporate stuff....

Anyhow, an update/clarification on the things going on here (I could/should have sent initially, sorry for that):

1) Lynx Toolchain compiles a FIT Image with the following structure:
	{
		Images{
			Kernel-1 {
				Data=Image;
			}
			Fdt-1 {
				Data=system.dtb;
			}
		}
		Configurations {
			Here kernel-1 and fdt-1 used as default conf-1
		}
	}
2) Data=Image does not have the Magic Code 0x644d5241
3) It is booting with U-Boot SPL 2020.04+g1ccc9d93576+p1 according to Lynx. I haven´t tested myself since we started with 2024.04, and we need it for some other features.
4) Then I investigated differences v2020.04 vs v2024.04, and found this part of code came in.
5) The finding is consistent with last error reported by U-Boot prior to reset "Bad Linux ARM64 Image magic!"

6) I 1st tried patching the Data=Image in Lynx FIT with the magic code, but it didn’t work.
7) once I patched bootm in U-Boot to skip this part and behaves as in v2020.04 (excluded all restructuring etc.) it booted the Lynx kernel

Basically, v2024.04 correctly detects "Bad Linux ARM64 Image magic!", as none is there at the moment, but in order to boot Lynx Hypervisor as-is this backward compatibility to v2020.04 is needed.

Does this shine more light on it?

Greetings
Dorde


Dorde Stojicevic

Networks and Cybersecurity

Rohde & Schwarz SIT GmbH
Hemminger Strasse 41 | 70499 Stuttgart-Weilimdorf | Germany
Phone: +4971169945195

Internet: www.rohde-schwarz.com



Geschäftsführer / Managing Director: Ralf Koenzen
Aufsichtsratsvorsitzender / Chair of the Supervisory Board: Mario Paoli
Sitz / Registered Office: Stuttgart
Handelsregister / Commercial Register: AG Stuttgart HRB 759 934
Umsatzsteuer-Identifikationsnummer (USt-IdNr.) / VAT Identification No.: DE 121 963 283
Elektro-Altgeräte Register (EAR) / WEEE Register No.: DE 877 727 67



-----Original Message-----
From: Tom Rini <trini at konsulko.com> 
Sent: Tuesday, February 3, 2026 6:00 PM
To: Stojicevic Dorde (11SIEPT1) <dorde.stojicevic at rohde-schwarz.com>
Cc: u-boot at lists.denx.de
Subject: *EXT* Re: [PATCH] arm: Backward compatibility to U-Boot v2020.04 ***CAUTION_Invalid_Signature***

On Tue, Feb 03, 2026 at 03:41:05PM +0000, Dorde Stojicevic wrote:

> From ae3f90922f5c31bd6198deb149edb9476ecfe4ef Mon Sep 17 00:00:00 2001
> From: Dorde Stojicevic 
> dorde.stojicevic at rohde-schwarz.com<mailto:dorde.stojicevic at rohde-schwa
> rz.com>
> Date: Tue, 3 Feb 2026 14:48:36 +0100
> Subject: [PATCH] arm: Backward compatibility to U-Boot v2020.04 
> Series-to: u-boot at lists.denx.de<mailto:u-boot at lists.denx.de> 
> Series-version: 1
> Signed-off-by: Dorde Stojicevic 
> dorde.stojicevic at rohde-schwarz.com<mailto:dorde.stojicevic at rohde-schwa
> rz.com>
> 
> Cover-letter:
> Backward compatibility with U-Boot v2020.04 Needed in order to boot 
> LynxSecure Hypervisor, otherwise U-Boot will fail at this position and 
> reset the controller END
> Commit-notes:
> Backward compatibility with U-Boot v2020.04 Needed in order to boot 
> LynxSecure Hypervisor, otherwise U-Boot will fail at this position and 
> reset the controller END
> ---
> boot/bootm.c | 27 ++++++++++++---------------
> 1 file changed, 12 insertions(+), 15 deletions(-)

This is an interesting find. And everything Quentin said applies. Also, the patch as-sent doesn't apply cleanly due to spacing issues, possibly from copy/paste to your email client?

Now, taking this patch and re-formatting things a bit, if do a "git diff
--ignore-all-space":

diff --git a/boot/bootm.c b/boot/bootm.c index 4bdca22ea8cf..9743051160ed 100644
--- a/boot/bootm.c
+++ b/boot/bootm.c
@@ -684,11 +684,7 @@ static int bootm_load_os(struct bootm_headers *images, int boot_progress)
 		int ret;
 
 		ret = booti_setup(load, &relocated_addr, &image_size, false);
-		if (ret) {
-			printf("Failed to prep arm64 kernel (err=%d)\n", ret);
-			return BOOTM_ERR_RESET;
-		}
-
+               if (ret == 0) {
                        /* Handle BOOTM_STATE_LOADOS */
                        if (relocated_addr != load) {
                                printf("Moving Image from 0x%lx to 0x%lx, end=0x%lx\n",

Is the interesting part. And you say this is on ARM, so we look at booti_setup in arch/arm/lib/image.c. The failure cases of the function is only "Bad Linux ARM64 Image magic!". So what are you trying to boot here, and is it really acting like a Linux Kernel "Image" file, which has a specific magic value? Or are we falling down to this case when we should not be?

Since it sounds like you can reproduce this rather easily, it might be good to use "git bisect" to narrow down which exact commit broke your use case, and we can work from there.

--
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 9150 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20260205/8c91b35d/attachment.png>