Re: [Bug 39822] SDHC cards no longer recognized on AT91 based board

[Douglas Gilbert <dgilbert@interlog.com>]

On 12-01-29 12:22 PM, Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 19:37 Sat 28 Jan     , Douglas Gilbert wrote:
>> On 11-10-24 09:54 PM, Douglas Gilbert wrote:
>>> On 11-07-28 07:48 PM, Chris Ball wrote:
>>>> Hi, thanks for the bug report. Adding linux-mmc@ and some atmel driver
>>>> folks to CC.
>>>>
>>>> On Thu, Jul 28 2011, bugzilla-daemon@bugzilla.kernel.org wrote:
>>>>> Prior to 3.0.0 the FoxG20 board (Atmel AT91SAM9G20) could boot and use
>>>>> both micro SD and SDHC card to contain its root file system. With the
>>>>> advent of the 3.0.0 kernel the micro SDHC card is unable to mount the
>>>>> root file system.
>>>>>
>>>>> Comparing the relevant part of dmesg for 2.6.38.6 (and it worked on
>>>>> 2.6.39):
>>>>> ...
>>>>> Waiting for root device /dev/mmcblk0p2...
>>>>> mmc0: host does not support reading read-only switch. assuming write-enable.
>>>>> at91_mci at91_mci: Timeout waiting end of packet
>>>>> mmc0: new SDHC card at address 1234
>>>>> mmcblk0: mmc0:1234 SA04G 3.67 GiB
>>>>> mmcblk0: p1 p2 p3
>>>>> ...
>>>>>
>>>>> to the same dmesg section for 3.0.0 :
>>>>> ...
>>>>> Waiting for root device /dev/mmcblk0p2...
>>>>> mmc0: error -110 whilst initialising SD card
>>>>> at91_mci at91_mci: Timeout waiting end of packet
>>>>> mmc0: error -110 whilst initialising SD card
>>>>> [no further progress]
>>>>>
>>>>> Non "HC" SD cards work as expected in 3.0.0
>>>>
>>>> If no-one on the CC list has any ideas, would you be interested in
>>>> trying to bisect this change?
>>>>
>>>> Another possibility would be adding printks before all of the -ETIMEDOUT
>>>> lines in core/* and host/at*, in order to see which one you're hitting
>>>> (and therefore which operation is causing the initialization to fail).
>>>> But a full bisection would be more conclusive.
>>>
>>> Still broken in lk 3.1.0, same errors in dmesg.
>>
>> Still broken in lk 3.2.0, same errors in dmesg.
>>
>> The solution (hack) up to lk 3.1.0 has been to take
>> linux-2.6.38/drivers/mmc/core/sd.c file and overwrite
>> the defective sd.c . That worked in lk 3.0.* and 3.1.* .
>> In lk 3.2.0 the old sd.c no longer compiles but
>> by adding "#include<linux/stat.h>" is does and thereafter
>> that SDHC card can be read again.
>>
>> My test system is a FoxG20 card which is based on the
>> AT91SAM9G20 SOC. The card's BIOS can read the 4 GB
>> microSD (HC) card in all cases; the problem is with
>> SDHC cards and all drivers/mmc/core/sd.c from and
>> including lk 3.0.0
> can you test it on atmel_mci?

Nothing went into dmesg concerning mci and the boot up
froze at this point:
   Waiting for root device /dev/mmcblk0p2...

> at91_mci is deprecated and will be drop soon

Hmm. Doesn't seem to work for me so there would be no loss.
It could at least announce when it gets loaded and if there
are any errors.


My theory remains that the large rework done on
drivers/mmc/core/sd.c just prior to lk 3.0.0 has caused
this problem.

Doug Gilbert