Re: [PATCH RESEND 0/9] ARM: stm32: add STM32H7R7 and ART-Pi2 support

[Alexandre TORGUE]

Hi Liu

On 7/26/26 17:49, Liu Changjie wrote:
> Resending to repair threading after Outlook rewrote the cover Message-ID
> during SMTP submission. Patch contents are unchanged.
> Add initial mainline support for the STM32H7R7 and the Ruiside ART-Pi2
> development board.
>
> The ART-Pi2 boots through first-stage firmware in the STM32 internal flash.
> That firmware configures the PLLs and memory interfaces before Linux
> executes in place from external XSPI NOR, with 32 MiB of HyperRAM used as
> system RAM.  Reprogramming the PLLs from Linux could therefore stop the
> instruction stream.
>
> The RCC driver in this series intentionally reconstructs the active clock
> rates from the firmware-programmed RCC registers.  It exposes the resulting
> SYSCLK, bus, timer, UART and SDMMC clocks, while Linux owns only the
> peripheral clock gates used by the initial device tree.  It also exposes
> the RCC reset registers through the existing reset-simple driver.  The
> first reset consumer is SDMMC1.
>
> The pinctrl driver covers GPIOA-H and GPIOM-P, including the AF0 trace
> functions described by the STM32H7RS pin-control data.  The board device
> tree enables UART4, TIM5, twelve GPIO banks, two LEDs, the user button and
> the four-bit SDMMC1 interface used by the microSD socket.  The socket has
> no card detect signal connected to the MCU, so the board uses broken-cd.
>
> ARM noMMU selects DMA_GLOBAL_POOL.  The board therefore reserves a 2 MiB
> shared DMA pool for the SDMMC internal DMA descriptor and bounce buffer.
> The pool is required before the mainline mmci-pl18x driver can probe.
Please, have a look on this series from Arnd:

https://lore.kernel.org/all/20260701212353.2196041-1-arnd@xxxxxxxxxx/

"Regarding actually removing the code, I would suggest linux-7.4 (January
2027), removing both the board files and any drivers that have either
become newly unused or were missed in a previous machine removal series,
followed by removing support for then unused CPUs a release later."

To be short, cortex-v7M CPUs are now deprecated and board support will be
deleting beginning of next year.

So, I'll not take new cortex-v7M board.

regards
Alex