Hi, Philippe Mathieu-Daudé <phi...@linaro.org> writes: > On 18/12/22 08:12, Felipe Balbi wrote: >> STM32F405 has 128K of SRAM and another 64K of CCM (Core-coupled >> Memory) at a different base address. Correctly describe the memory >> layout to give existing FW images have a chance to run unmodified. >> >> Signed-off-by: Felipe Balbi <ba...@kernel.org> >> --- >> hw/arm/stm32f405_soc.c | 8 ++++++++ >> include/hw/arm/stm32f405_soc.h | 5 ++++- >> 2 files changed, 12 insertions(+), 1 deletion(-) >> >> diff --git a/hw/arm/stm32f405_soc.c b/hw/arm/stm32f405_soc.c >> index c07947d9f8b1..cef23d7ee41a 100644 >> --- a/hw/arm/stm32f405_soc.c >> +++ b/hw/arm/stm32f405_soc.c >> @@ -139,6 +139,14 @@ static void stm32f405_soc_realize(DeviceState *dev_soc, >> Error **errp) >> } >> memory_region_add_subregion(system_memory, SRAM_BASE_ADDRESS, >> &s->sram); >> >> + memory_region_init_ram(&s->ccm, NULL, "STM32F405.ccm", CCM_SIZE, > > Including the machine name in the memory description seems a bad > habit from old days. What do you think about renaming as > 'core-coupled-memory'?
I don't oppose it, but I was merely following the model from `netduino2plus'. >> + &err); >> + if (err != NULL) { >> + error_propagate(errp, err); >> + return; >> + } >> + memory_region_add_subregion(system_memory, CCM_BASE_ADDRESS, &s->ccm); >> + >> armv7m = DEVICE(&s->armv7m); >> qdev_prop_set_uint32(armv7m, "num-irq", 96); >> qdev_prop_set_string(armv7m, "cpu-type", s->cpu_type); >> diff --git a/include/hw/arm/stm32f405_soc.h b/include/hw/arm/stm32f405_soc.h >> index 5bb0c8d56979..249ab5434ec7 100644 >> --- a/include/hw/arm/stm32f405_soc.h >> +++ b/include/hw/arm/stm32f405_soc.h >> @@ -46,7 +46,9 @@ OBJECT_DECLARE_SIMPLE_TYPE(STM32F405State, STM32F405_SOC) >> #define FLASH_BASE_ADDRESS 0x08000000 >> #define FLASH_SIZE (1024 * 1024) >> #define SRAM_BASE_ADDRESS 0x20000000 >> -#define SRAM_SIZE (192 * 1024) >> +#define SRAM_SIZE (128 * 1024) >> +#define CCM_BASE_ADDRESS 0x10000000 >> +#define CCM_SIZE (64 * 1024) > > Since the CCM_SIZE won't be used elsewhere, we can simply use '64 * KiB' > in the memory_region_init_ram() in the source file. Up to the maintainer > :) Also not opposed. I'll wait a couple days before respinning the patch to give the rest of the community time to reply. -- balbi
signature.asc
Description: PGP signature