[U-Boot] rpi: bcm2835_sdhost: occasional errors while writing

[Matthias Brugger]

On 05/03/2019 13:36, Zubair Lutfullah Kakakhel wrote:
> On Mon, Feb 25, 2019 at 10:43 AM Zubair Lutfullah Kakakhel
> <zubair at balena.io> wrote:
>>
>> Hi,
>>
>> I've encountered occasional glitches while writing to the sd card
>> via u-boot(2018.07) on a Pi 3 B+. The goal is to write the bootcount
>> in the boot partition. As a test, this is the loop I used to read, display
>> contents, write to the same file over and over again.
>>
>> ```
>> mw 200000 0
>> while itest *200000 == 0 ; do
>>     fatload mmc 0:1 $kernel_addr_r resin-image
>>     md $kernel_addr_r 18
>>     fatwrite mmc 0:1 $kernel_addr_r resin-image 18
>>     echo === done ====
>> done
>> ```
>> resin-image is a file in the boot partition that has some text in it.
>>
>> After a few minutes of this loop running, I 'sometimes' get this error
>> and the sd card stops working
>>
>> ```
>> writing resin-image
>> wait_transfer_complete - still waiting after 100001 retries
>> mmc write failed
>> Error: flush fat buffer
>>
>> ** Unable to write "resin-image" from mmc 0:1 **
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:376/bcm2835_prepare_data()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> ** Can't read partition table on 0:0 **
>> ** Invalid partition 1 **
>> 00080000: 4e204f44 5220544f 564f4d45 48542045    DO NOT REMOVE TH
>> 00080010: 46205349 0a454c49 0f030f03 0f030f03    IS FILE.........
>> 00080020: 0f030f03 0f030f03 01600103 03030300    ..........`.....
>> 00080030: 03030303 03030303 03030303 03030303    ................
>> 00080040: 03030303 03030303 03030303 03034103    .............A..
>> 00080050: 03030303 03030303 03030303 03030303    ................
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> ** Can't read partition table on 0:0 **
>> ** Invalid partition 1 **
>> ```
>>
>> What is even more concerning is that the device can't boot. Even if a
>> write failed.
>> Subsequent reads seem to fail as well.
>>
>> ```
>> U-Boot> mmc part
>>
>> Partition Map for MMC device 0  --   Partition Type: DOS
>>
>> Part    Start Sector    Num Sectors     UUID            Type
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> WARNING at /home/zubairlk/resin/yocto/resin-raspberrypi3/build/tmp/work/raspberrypi3-poky-linux-gnueabi/u-boot/1_2018.07-r0/git/drivers/mmc/bcm2835_sdhost.c:408/bcm2835_send_command()!
>> ** Can't read partition table on 0:0 **
>> U-Boot>
>> ```
>>
>> I can recover the device after a reboot.
>>
>> Any thoughts/ideas about where/what I should look for?
>> Or if I'm lucky, any wip patches floating around?
> 
> Any thoughts?
> 

I tried with master branch:
8303467e80 ("Merge git://git.denx.de/u-boot-fsl-qoriq")

And I wasn't able to reproduce the issue.
I see you are using a one year old u-boot. If you can't bump to a newer version,
then I would propose to find a version that doesn't show this bug and run
git-bisect between the two to find the commit that fixed the issue.

Best regards,
Matthias