On 7/10/24 6:58 AM, Cédric Le Goater wrote:
On 7/9/24 11:39 PM, Philippe Mathieu-Daudé wrote:
On 9/7/24 17:58, Cédric Le Goater wrote:
On 7/9/24 5:25 PM, Philippe Mathieu-Daudé wrote:
Since v42:
- Stick to spec v4.3 (re-simplified EXT_CSD register & migrate)
- Fill CID register
- Few changes to CSD register
- Implement 'boot-mode' reset timing
- Add 'boot-size' property
Change required for aspeed branch:
-- >8 --
diff --git a/hw/arm/aspeed.c b/hw/arm/aspeed.c
index 8c0e36badd..563816b710 100644
--- a/hw/arm/aspeed.c
+++ b/hw/arm/aspeed.c
@@ -344,3 +344,3 @@ static void sdhci_attach_drive(SDHCIState *sdhci, DriveInfo
*dinfo, bool emmc,
if (emmc) {
- qdev_prop_set_uint8(card, "boot-config", boot_emmc ? 0x48 : 0x0);
+ qdev_prop_set_uint64(card, "boot-size", 1 * MiB);
}
(I'm still reluctant to merge patches 16-18)...
Then, please drop all changes related to the boot partitions. I will keep
the original patches in my tree and address the feature when I have time.
TYPE_EMMC is already great to have.
As you mentioned on today's community call, eMMC is a chipset soldered
on a board, so our boards exactly knows what size to expect on the card,
and whether boot partitions are present or not.
My remark was on the use of 3 blockdevs to represent the eMMC contents.
1 should be enough.
Having a "boot-size" property to set the BOOT_SIZE_MULT register is fine.
The Aspeed model does some assumption today when installing the first
boot area as a boot ROM :
aspeed_install_boot_rom(bmc, blk_by_legacy_dinfo(emmc0), 64 * KiB);
The "boot-size" property will clarify the settings. Please keep it.
64 * KiB is not related to the eMMC partition size. It is a SRAM
limit in which the SoC loads the boot data.
Thanks,
C.
