Re: [PATCH v8 8/9] arm64: dts: ti: k3-am62p: Split r5f memory region

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Splitting the single firmware memory region into multiple smaller regions breaks kernel-side firmware loading for existing firmware binaries whose ELF segments do not perfectly align with the newly introduced exact boundaries.
--

commit 26f5fd0eab7dd6c1ef82f91d1fe6cdf1fe46183f
Author: Markus Schneider-Pargmann (TI) <msp@baylibre.com>

arm64: dts: ti: k3-am62p: Split r5f memory region

This commit splits the firmware memory region into multiple specific parts
so bootloader software can find necessary information like LPM metadata.
This replaces a single contiguous firmware region with several smaller ones.

> diff --git a/arch/arm64/boot/dts/ti/k3-am62p-ti-ipc-firmware.dtsi b/arch/arm64/boot/dts/ti/k3-am62p-ti-ipc-firmware.dtsi
> index 5d7f701420e2d..84dfbfd1986ce 100644
> --- a/arch/arm64/boot/dts/ti/k3-am62p-ti-ipc-firmware.dtsi
> +++ b/arch/arm64/boot/dts/ti/k3-am62p-ti-ipc-firmware.dtsi
> @@ -44,8 +44,14 @@ &wkup_r5fss0 {
>  &wkup_r5fss0_core0 {
>  	mboxes = <&mailbox0_cluster0 &mbox_r5_0>;
>  	memory-region = <&wkup_r5fss0_core0_dma_memory_region>,
> -			<&wkup_r5fss0_core0_memory_region>;
> -	memory-region-names = "dma", "firmware";
> +			<&wkup_r5fss0_core0_ipc_region>,
> +			<&wkup_r5fss0_core0_lpm_fs_stub_region>,
> +			<&wkup_r5fss0_core0_lpm_metadata_region>,
> +			<&wkup_r5fss0_core0_lpm_rest_region>,
> +			<&wkup_r5fss0_core0_dm_region>;

[Severity: High]
Does this fragmentation break kernel-side firmware loading for existing
firmware binaries?

When the remoteproc core attempts to load the firmware, it calls
rproc_elf_load_segments(), which requests a contiguous virtual address
pointer for each segment via k3_rproc_da_to_va().

Because k3_rproc_da_to_va() evaluates each rmem chunk independently, any
segment spanning the newly introduced boundaries fails the strict bounds
check:

drivers/remoteproc/ti_k3_common.c:k3_rproc_da_to_va() {
    ...
    if (da >= dev_addr && ((da + len) <= (dev_addr + size))) {
    ...
}

It does not coalesce contiguous chunks. Thus, any firmware segment that spans
across the newly introduced chunk boundaries will fail the check and
return NULL, rendering the device unable to boot.

Since device tree updates must not break backward compatibility with existing
firmware binaries, should the driver be updated to coalesce adjacent memory
regions before searching, or should this region split be reconsidered?

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260903-topic-am62a-ioddr-dt-v6-19-v8-0-74c650dfdea6@baylibre.com?part=8