[U-Boot] raspberrypi compute module3 emmc not usable

[Jonathan Gray]

On Tue, Jul 10, 2018 at 10:37:19PM +0200, Belisko Marek wrote:
> Hello,
> 
> I'm trying to use u-boot on CM3 (which contains eMMC) to boot kernel.
> I'm using u-boot 2018.05 and it can start fine like:
> U-Boot 2018.05 (Jul 10 2018 - 20:32:05 +0000)
> 
> DRAM:  948 MiB
> RPI Compute Module 3 (0xa220a0)
> MMC:   sdhci at 7e300000: 0
> Loading Environment from FAT... unable to select a mode
> ** No partition table - mmc 0 **
> Failed (-5)
> In:    serial
> Out:   vidconsole
> Err:   vidconsole
> Net:   No ethernet found.
> starting USB...
> USB0:   scanning bus 0 for devices... 1 USB Device(s) found
>        scanning usb for storage devices... 0 Storage Device(s) found
> Hit any key to stop autoboot:  0
> U-Boot>
> U-Boot>
> U-Boot>
> U-Boot> mmc info
> Device: sdhci at 7e300000
> Manufacturer ID: 15
> OEM: 100
> Name: 4FPD3
> Bus Speed: 52000000
> Mode : MMC legacy
> Rd Block Len: 512
> MMC version 5.0
> High Capacity: Yes
> Capacity: 3.6 GiB
> Bus Width: 1-bit
> Erase Group Size: 512 KiB
> HC WP Group Size: 8 MiB
> User Capacity: 3.6 GiB WRREL
> Boot Capacity: 4 MiB ENH
> RPMB Capacity: 512 KiB ENH
> U-Boot> mmc part
> ## Unknown partition table type 0
> U-Boot> mmc read $loadaddr 0 2
> 
> but reading from eMMC fails with:
> 
> MMC read: dev # 0, block # 0, count 2 ... 0 blocks read: ERROR
> 
> so it looks like eMMC is not properly working with CM3 in this
> version? Any ideas what to check to have it working before I dig to
> debugging session ;). Thanks.

You should give 2018.07 a try, which includes a change for a problem
reported by someone with a CM3.

commit 79fd08f7456c7d12b04ef39e51d84d9981599c3a
Author: Alexander Graf <agraf at suse.de>
Date:   Wed May 23 22:24:51 2018 +0200

    mmc: Unirqify bcm2835_sdhost and fix writes
    
    The bcm2835 sdhost driver has a problem with "write multiple" commands.
    It seems to boil down to the fact that the controller dislikes its FIFO
    to get drained at the end of a block when a write multiple blocks command
    is in flight.
    
    The easy fix is to simply get rid of all the IRQ driven logic and make
    the driver push as much data into the FIFO as it can. That way we never
    drain and we never run into the problem.
    
    Reported-by: Jan Leonhardt <jan at cyberdesigner.net>
    Signed-off-by: Alexander Graf <agraf at suse.de>

 drivers/mmc/bcm2835_sdhost.c | 265 ++++++++++--------------------------------------------
 1 file changed, 47 insertions(+), 218 deletions(-)