
Hi Adam,
On Thu, May 23, 2019 at 4:11 PM Adam Ford aford173@gmail.com wrote:
Currently, when the spl_boot_device checks the boot device, it will only return MMC1 when it's either sd or eMMC regardless of whether or not it's MMC1 or MMC2. This is a problem when booting from MMC2 if MMC isn't being manually configured like in the DM_SPL case with SPL_OF_CONTROL.
This patch will check the register and return either MMC1 or MMC2.
Signed-off-by: Adam Ford aford173@gmail.com
diff --git a/arch/arm/mach-imx/spl.c b/arch/arm/mach-imx/spl.c index 9f1e0f6a72..1f230aca33 100644 --- a/arch/arm/mach-imx/spl.c +++ b/arch/arm/mach-imx/spl.c @@ -24,6 +24,7 @@ u32 spl_boot_device(void) { unsigned int bmode = readl(&src_base->sbmr2); u32 reg = imx6_src_get_boot_mode();
u32 mmc_index = ((reg >> 11) & 0x03); /* * Check for BMODE if serial downloader is enabled
@@ -84,11 +85,12 @@ u32 spl_boot_device(void) /* SD/eSD: 8.5.3, Table 8-15 */ case IMX6_BMODE_SD: case IMX6_BMODE_ESD:
return BOOT_DEVICE_MMC1;
/* MMC/eMMC: 8.5.3 */ case IMX6_BMODE_MMC: case IMX6_BMODE_EMMC:
return BOOT_DEVICE_MMC1;
if (mmc_index == 1)
return BOOT_DEVICE_MMC2;
else
return BOOT_DEVICE_MMC1;
I just got to test v2019.10-rc1, which includes this change, and I'm unable to boot SPL on my Hummingboard 2, which uses USDHC-2 for sdcard.
Looks like this change breaks devices that are not using device tree/dynamic mmc initialization, as the MMC index will not necessarily be correct.
Looking at mx6cuboxi.c in particular, fsl_esdhc_initialize will only be called once as it already knows which MMC device to initialize based on the BOOT_CFG register, causing the device mmc devnum to be 1 only. The issue shows up when booting SPL as find_mmc_device will look for a matching dev_num index, which gets automatically increased when fsl_esdhc_initialize gets called (and which is only called once in my case, for MMC2).
This is not an issue with imx6q_logic as at 8f4691e31a18254d225524a4b018b8cbcddc70b1 you removed fsl_esdhc_initialize, but all the other boards using MMC2 and doing only one single call will have the same problem.
Not sure if there is an easy fix here, but converting everything to dynamic mmc initialization will required quite a bit of effort.
Thanks,