[PATCH v3 0/2] ARM: dts: aspeed: Add an alt 128M flash layout.

[PATCH v3 0/2] ARM: dts: aspeed: Add an alt 128M flash layout.

[Marc Olberding]

Add a layout for an alternate 128M flash layout. This is useful for the
aspeed ast2500[2] and ast2600[1] SoCs, where it has the ability to boot
from one of two spi chips, nominally the primary and backup SPIs. Adding
a layout allows userspace to flash the alternate spi by partition and
switch over to it, allowing lower downtime for firmware flash, or for
updating the primary flash from the golden backup flash. This is already
an established scheme for the 32M and 64M layouts, where the layout is
the same but each partition has `alt` prepended to its name. This allows
userspace to just prepend alt to the images in the update process to
specify which flash chip to update. This series also adds support for
the msx4 to use this layout.

References:

[1] https://www.aspeedtech.com/server_ast2600/
[2] https://www.aspeedtech.com/server_ast2500/

Signed-off-by: Marc Olberding <molberding@nvidia.com>
---
Changes in v3:
- Reworded commit adding alt flash layout and fixed formatting issues
- Expanded testing section 
- Readded the Changes in v2 section, somehow they got lost
- Link to v2: https://lore.kernel.org/r/20251218-alt-128-v2-0-10499a0b7eb9@nvidia.com
Changes in v2:
- Retargetted series for the bmc tree
- Added a usage of the dtsi in the MSX4 device tree
- Link to v1: https://lore.kernel.org/all/20251211-alt-128-layout-v1-1-100f57d2bc4a@nvidia.com/

---
Marc Olberding (2):
      ARM: dts: aspeed: add an alt 128M flash layout
      ARM: dts: aspeed: Add alt partition support to msx4

 .../boot/dts/aspeed/aspeed-bmc-nvidia-msx4-bmc.dts |  1 +
 .../dts/aspeed/openbmc-flash-layout-128-alt.dtsi   | 32 ++++++++++++++++++++++
 2 files changed, 33 insertions(+)
---
base-commit: 459a5aa171c0f13fcd78faa9594dc4aa5a95770b
change-id: 20251218-alt-128-19270d894872

Best regards,
-- 
Marc Olberding <molberding@nvidia.com>

[PATCH v3 1/2] ARM: dts: aspeed: add an alt 128M flash layout

[Marc Olberding]

Add a 128M layout for the BMC flash chip we didn't boot from. Including
this allows the user to write to each partition on the alternate spi
chip. This dtsi follows the existing standard of using the same layout
as non alt version and prepending `alt` to each partition's name.

Tested by running on target and catting sysfs nodes
    for devdir in /sys/class/mtd/mtd*; do
         [[ -d $devdir && -r $devdir/name ]] || continue
         name=$(<"$devdir/name")
         [[ $name == alt* ]] || continue

         size=$(<"$devdir/size")
         offset=0
         [[ -r $devdir/offset ]] && offset=$(<"$devdir/offset")

         dev=$(basename "$devdir")
         printf "%s name=%s offset=0x%X size=0x%X\n" "$dev" "$name" \
            "$offset" "$size"
    done

    mtd10 name=alt-rofs offset=0xA00000 size=0x5600000
    mtd11 name=alt-rwfs offset=0x6000000 size=0x2000000
    mtd7 name=alt-u-boot offset=0x0 size=0xE0000
    mtd8 name=alt-u-boot-env offset=0xE0000 size=0x20000
    mtd9 name=alt-kernel offset=0x100000 size=0x900000

Ran dtbs check, no errors from the msx4 dts

running diff between the main and alt flash layouts shows that the only
diff is alt prepended to the partition labels, as expected.

Signed-off-by: Marc Olberding <molberding@nvidia.com>
---
 .../dts/aspeed/openbmc-flash-layout-128-alt.dtsi   | 32 ++++++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/arch/arm/boot/dts/aspeed/openbmc-flash-layout-128-alt.dtsi b/arch/arm/boot/dts/aspeed/openbmc-flash-layout-128-alt.dtsi
new file mode 100644
index 000000000000..08ce2eab472c
--- /dev/null
+++ b/arch/arm/boot/dts/aspeed/openbmc-flash-layout-128-alt.dtsi
@@ -0,0 +1,32 @@
+// SPDX-License-Identifier: GPL-2.0+
+
+partitions {
+	compatible = "fixed-partitions";
+	#address-cells = <1>;
+	#size-cells = <1>;
+
+	u-boot@0 {
+		reg = <0x0 0xe0000>; // 896KB
+		label = "alt-u-boot";
+	};
+
+	u-boot-env@e0000 {
+		reg = <0xe0000 0x20000>; // 128KB
+		label = "alt-u-boot-env";
+	};
+
+	kernel@100000 {
+		reg = <0x100000 0x900000>; // 9MB
+		label = "alt-kernel";
+	};
+
+	rofs@a00000 {
+		reg = <0xa00000 0x5600000>; // 86MB
+		label = "alt-rofs";
+	};
+
+	rwfs@6000000 {
+		reg = <0x6000000 0x2000000>; // 32MB
+		label = "alt-rwfs";
+	};
+};

-- 
2.34.1

[PATCH v3 2/2] ARM: dts: aspeed: Add alt partition support to msx4

[Marc Olberding]

Add support for the alt 128M partition table to the msx4. This allows it
program the flash it didn't boot from by partition.

Signed-off-by: Marc Olberding <molberding@nvidia.com>
---
 arch/arm/boot/dts/aspeed/aspeed-bmc-nvidia-msx4-bmc.dts | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm/boot/dts/aspeed/aspeed-bmc-nvidia-msx4-bmc.dts b/arch/arm/boot/dts/aspeed/aspeed-bmc-nvidia-msx4-bmc.dts
index 44f95a3986cb..a12d4164de4a 100644
--- a/arch/arm/boot/dts/aspeed/aspeed-bmc-nvidia-msx4-bmc.dts
+++ b/arch/arm/boot/dts/aspeed/aspeed-bmc-nvidia-msx4-bmc.dts
@@ -67,6 +67,7 @@ flash@1 {
 		spi-tx-bus-width = <1>;
 		spi-rx-bus-width = <1>;
 		status = "okay";
+		#include "openbmc-flash-layout-128-alt.dtsi"
 	};
 };
 

-- 
2.34.1

Re: [PATCH v3 1/2] ARM: dts: aspeed: add an alt 128M flash layout

[Krzysztof Kozlowski]

On 20/01/2026 20:26, Marc Olberding wrote:
> Add a 128M layout for the BMC flash chip we didn't boot from. Including
> this allows the user to write to each partition on the alternate spi
> chip. This dtsi follows the existing standard of using the same layout
> as non alt version and prepending `alt` to each partition's name.
> 
> Tested by running on target and catting sysfs nodes

You cannot test this patch that way - DTSI is not included by anything.
This probably should be squashed with next one because makes little
benefit on its own,

>     for devdir in /sys/class/mtd/mtd*; do
>          [[ -d $devdir && -r $devdir/name ]] || continue
>          name=$(<"$devdir/name")
>          [[ $name == alt* ]] || continue
> 
>          size=$(<"$devdir/size")
>          offset=0
>          [[ -r $devdir/offset ]] && offset=$(<"$devdir/offset")
> 
>          dev=$(basename "$devdir")
>          printf "%s name=%s offset=0x%X size=0x%X\n" "$dev" "$name" \
>             "$offset" "$size"
>     done
> 
>     mtd10 name=alt-rofs offset=0xA00000 size=0x5600000
>     mtd11 name=alt-rwfs offset=0x6000000 size=0x2000000
>     mtd7 name=alt-u-boot offset=0x0 size=0xE0000
>     mtd8 name=alt-u-boot-env offset=0xE0000 size=0x20000
>     mtd9 name=alt-kernel offset=0x100000 size=0x900000
> 
> Ran dtbs check, no errors from the msx4 dts

Drop this sentence, it's implied. There is no point to state obvious,
e.g. that you build source code, in the commit msg. It serves no purpose
in the git history. You also have changelog for such things, if you want
to emphasize.

> 
> running diff between the main and alt flash layouts shows that the only
> diff is alt prepended to the partition labels, as expected.

? Not sure what you want to say, but please do not explain us how DTS
works (in case this was about DTS).


Best regards,
Krzysztof

Re: [PATCH v3 1/2] ARM: dts: aspeed: add an alt 128M flash layout

[Marc Olberding]

On Tue, Jan 20, 2026 at 08:40:29PM +0100, Krzysztof Kozlowski wrote:
> On 20/01/2026 20:26, Marc Olberding wrote:
> You cannot test this patch that way - DTSI is not included by anything.
> This probably should be squashed with next one because makes little
> benefit on its own,
Ack, will do in the next patch series.

> > 
> > Ran dtbs check, no errors from the msx4 dts
> 
> Drop this sentence, it's implied. There is no point to state obvious,
> e.g. that you build source code, in the commit msg. It serves no purpose
> in the git history. You also have changelog for such things, if you want
> to emphasize.
ack, will drop.
> > 
> > running diff between the main and alt flash layouts shows that the only
> > diff is alt prepended to the partition labels, as expected.
> 
> ? Not sure what you want to say, but please do not explain us how DTS
> works (in case this was about DTS).

I'll drop this piece. This was a recommended piece of testing by Andrew,
the maintainer, just to show that the alt partition and main partition files
are only different by the label names. That said, its redundant with the testing
on target shown above.

Thanks for your feedback
Marc