[BUG] mtd: rawnand: mircon: Cannot detect chip with commit 51f3b3970a8c

[Chris.Packham@xxxxxxxxxxxxxxxxxxx (Chris Packham)]

Hi Dinh, Boris,

On 07/07/18 05:21, Boris Brezillon wrote:
> On Fri, 6 Jul 2018 12:03:49 -0500
> Dinh Nguyen <dinguyen at kernel.org> wrote:
> 
>> Hi,
>>
>> It looks like commit: "51f3b3970a8c mtd: rawnand: micron: detect forced
>> on-die ECC" is causing the NAND chip on the SoCFPGA Arria10 devkit to
>> stop working on linux-next.
>>
>> If I revert the commit, then the NAND driver is able to detect the NAND
>> chip.
>>
>> The NAND chip is a Micron MT29F1G08ABBEAH4 part.
>>
>>
>> [    0.741984] nand: device found, Manufacturer ID: 0x2c, Chip ID: 0xa1
>> [    0.748336] nand: Micron MT29F1G08ABBEAH4
>> [    0.752330] nand: 128 MiB, SLC, erase size: 128 KiB, page size: 2048,
>> OOB siz
>> e: 64
>>   [    0.763358] ecc.mode = 3
>> [    0.766312] On-die ECC forcefully enabled, not supported
>> [    0.771654] denali-nand-dt: probe of ffb90000.nand failed with error -22
>>
>> With commit reverted:
>>
>> [    0.742481] nand: device found, Manufacturer ID: 0x2c, Chip ID: 0xa1
>> [    0.748833] nand: Micron MT29F1G08ABBEAH4
>> [    0.752828] nand: 128 MiB, SLC, erase size: 128 KiB, page size: 2048,
>> OOB siz
>> e: 64
>> [    0.760610] Bad block table found at page 65472, version 0x01
>> [    0.766433] Bad block table found at page 65408, version 0x01
>> [    0.772464] 2 fixed-partitions partitions found on MTD device denali-nand
>> [    0.779247] Creating 2 MTD partitions on "denali-nand":
>> [    0.784453] 0x000000000000-0x000001c00000 : "Boot and fpga data"
>> [    0.791241] 0x000001c00000-0x000008000000 : "Root Filesystem - JFFS2"
>>
>> Looks like if I comment out this part of the patch in
>> drivers/mtd/nand/raw/nand_micron.c
>>
>> +#if 0
>>          for (i = 0; i < ARRAY_SIZE(micron_on_die_ecc); i++)
>>                  if (chip->id.data[1] == micron_on_die_ecc[i])
>>                          return MICRON_ON_DIE_MANDATORY;
>> -
>> +#endif
> 
> Oops, I overlooked that detail. Testing only the ID bytes is not
> enough, because it does not encode whether the NAND has on-die ECC or
> not. We should really test the model name here...
> 
> Chris, can you prepare a fixup for that?
> 

With the latest set of changes is there anything left to do for this? I 
think we now have on-die ECC detection that doesn't require checking the ID.