orion-nand: uncorrectable ECC error on v5.10-rc6

[Chris Packham]

Hi Miquel,

On 2/12/20 8:59 pm, Miquel Raynal wrote:
> Hi Chris,
>
> Chris Packham <Chris.Packham at alliedtelesis.co.nz> wrote on Wed, 2 Dec
> 2020 07:47:32 +0000:
>
>> Hi,
>>
>> I've just booted v5.10-rc6 on a kirkwood based board (which uses the
>> orion-nand driver) and I get the following errors reported. I haven't
>> started bisecting yet but v5.7.19 mounts the nand flash without any issue.
>>
>> ubi0: attaching mtd0
>> __nand_correct_data: uncorrectable ECC error
>> ubi0 warning: ubi_io_read: error -74 (ECC error) while reading 64 bytes
>> from PEB 0:0, read only 64 bytes, retry
>> __nand_correct_data: uncorrectable ECC error
>> ubi0 warning: ubi_io_read: error -74 (ECC error) while reading 64 bytes
>> from PEB 0:0, read only 64 bytes, retry
>> __nand_correct_data: uncorrectable ECC error
>> ubi0 warning: ubi_io_read: error -74 (ECC error) while reading 64 bytes
>> from PEB 0:0, read only 64 bytes, retry
>> __nand_correct_data: uncorrectable ECC error
>> ubi0 error: ubi_io_read: error -74 (ECC error) while reading 64 bytes
>> from PEB 0:0, read 64 bytes
>> CPU: 0 PID: 101 Comm: ubiattach Not tainted 5.10.0-rc6+ #1
>> Hardware name: Marvell Kirkwood (Flattened Device Tree)
>> [<8010ca64>] (unwind_backtrace) from [<80109bd0>] (show_stack+0x10/0x14)
>> [<80109bd0>] (show_stack) from [<8045f10c>] (ubi_io_read+0x184/0x304)
>> [<8045f10c>] (ubi_io_read) from [<8045f4ac>] (ubi_io_read_ec_hdr+0x44/0x240)
>> [<8045f4ac>] (ubi_io_read_ec_hdr) from [<80464db0>]
>> (ubi_attach+0x178/0x15fc)
>> [<80464db0>] (ubi_attach) from [<80458d8c>] (ubi_attach_mtd_dev+0x538/0xb48)
>> [<80458d8c>] (ubi_attach_mtd_dev) from [<8045a114>]
>> (ctrl_cdev_ioctl+0x170/0x1e0)
>> [<8045a114>] (ctrl_cdev_ioctl) from [<80203094>] (sys_ioctl+0x1f8/0x990)
>> [<80203094>] (sys_ioctl) from [<80100060>] (ret_fast_syscall+0x0/0x50)
>> Exception stack(0x87633fa8 to 0x87633ff0)
>> 3fa0:                   00000003 7e9b0c30 00000003 40186f40 7e9b0c30
>> 00000000
>> 3fc0: 00000003 7e9b0c30 000148f8 00000036 00014770 00013f90 76f3dfa4
>> 00000000
>> 3fe0: 76e936f0 7e9b0c1c 00011f68 76e936fc
> I recently contributed a pile of fixes to ensure DT parsing was not
> broken and this applies to Orion. Can you please check
>
> mtd: rawnand: orion: Move the ECC initialization to ->attach_chip()
That looks to be it. In Linus's tree commit 76dc2bfc2e1b ("Merge tag 
'mtd/fixes-for-5.10-rc6' of 
git://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux") seems to be 
the difference between working and not working.
> And tell me if you see something wrong there? I assumed this driver was
> not supporting on host ECC engines and only soft Hamming was used, is
> this assumption wrong?

Our dts has

         nand-ecc-mode = "soft";
         nand-ecc-algo = "bch";
         nand-on-flash-bbt;