On Mon, Mar 15, 2021 at 10:46 AM Jagan Teki <ja...@amarulasolutions.com> wrote:
>
> Move the redundant config item like SPL, memory-related
> across all imx8mm config files in the common config header,
> imx8mm-common.h
>
> Verified the built files, seems almost the same as before.
>
> Cc: Tim Harvey <thar...@gateworks.com>
> Cc: Adam Ford <aford...@gmail.com>
> Cc: Peng Fan <peng....@nxp.com>
> Cc: Teresa Remmet <t.rem...@phytec.de>
> Cc: Igor Opaniuk <igor.opan...@toradex.com>
> Signed-off-by: Jagan Teki <ja...@amarulasolutions.com>
> ---
> Changes for v2:
> - add venice changes
>
> include/configs/imx8mm-common.h | 48 ++++++++++++++++++++++++++++++++
> include/configs/imx8mm_beacon.h | 48 ++++----------------------------
> include/configs/imx8mm_evk.h | 45 ++----------------------------
> include/configs/imx8mm_venice.h | 46 ++++--------------------------
> include/configs/phycore_imx8mm.h | 43 ++--------------------------
> include/configs/verdin-imx8mm.h | 42 +++-------------------------
> 6 files changed, 68 insertions(+), 204 deletions(-)
> create mode 100644 include/configs/imx8mm-common.h
>
> diff --git a/include/configs/imx8mm-common.h b/include/configs/imx8mm-common.h
> new file mode 100644
> index 0000000000..f1afa5c5a9
> --- /dev/null
> +++ b/include/configs/imx8mm-common.h
> @@ -0,0 +1,48 @@
> +/* SPDX-License-Identifier: GPL-2.0+ */
> +/*
> + * Copyright (c) 2019 NXP
> + * Copyright (c) 2020 Amarula Solutions(India)
> + */
> +
> +#ifndef __IMX8MM_COMMON_H
> +#define __IMX8MM_COMMON_H
> +
> +#include <linux/sizes.h>
> +#include <asm/arch/imx-regs.h>
> +
> +#ifdef CONFIG_SPL_BUILD
> +#define CONFIG_SPL_MAX_SIZE (148 * 1024)
> +#define CONFIG_SPL_STACK 0x920000
> +#define CONFIG_SPL_BSS_START_ADDR 0x910000
> +#define CONFIG_SPL_BSS_MAX_SIZE SZ_8K
> +#define CONFIG_SYS_SPL_MALLOC_START 0x42200000
> +#define CONFIG_SYS_SPL_MALLOC_SIZE SZ_512K
Jagan,
Seems like a good idea, thanks for your work here!
I ran into an issue with imx8mm-venice due to my FIT image being
larger than 512KiB causing me to bump CONFIG_SYS_SPL_MALLOC_SIZE (I
bumped to 1MiB).
I have a pending patch for that [1] which would collide with this and
I wonder what your feelings are about SYS_SPL_MALLOC_SIZE. When I
started work on imx8mm-venice support the SPL did not use a malloc
buffer for loading the FIT image so I never ran into an issue but that
changed with commit 03f1f78a9b44 ("spl: fit: Prefer a malloc()'d
buffer for loading images")' and after that I was not able to load my
FIT image. I'm honestly not sure how the other imx8mm boards are not
affected unless they just barely fit with a 512K malloc pool.
I'm not sure what the downside would be to doubling it to 1MiB or even
more and I would be boards using FIT with multiple dtb's would go over
512KiB pretty quickly (as opposed to having to undef and re-define it
per board).
Best regards,
Tim
[1]
http://patchwork.ozlabs.org/project/uboot/patch/20210308215236.27722-2-thar...@gateworks.com/
