[PATCH 1/1] mtd: spi-nand: macronix: disable continuous read for MX35LFxGE4AD

[Miquel Raynal]

Hello,

+ Cheng Ming from Macronix

On 19/06/2025 at 22:00:03 +08, Chukun Pan <amadeus at jmu.edu.cn> wrote:

> Hi,
>
>> Please cc linux-mtd at lists.infradead.org for all patches and queries
>> related to MTD subsystems.
>> Run ./scripts/get_maintainers.pl on the patch for more information.
>
> Sorry, I forgot to add the following mailing list.
> Do I need to resend this patch?
>
>> Why? this datasheet says otherwise [1]:
>>
>> Would it be possible that SPI/QSPI controller of GL.iNet GL-MT3000
>> has a limitation?
>
> GL.iNet GL-MT3000 uses MediaTek MT7981 SoC, with SPI controllers
> based on IPM design. [1]
>
> I don't know what the difference is between the SPI controllers of
> MT7981 and the Zynq used in commit [2].
>
> In my case, ubi_io_read and ecc errors occurred when the openwrt
> attach ubi partition (NMBM has been disabled [3]):
>
> [    7.852000] ubi0 warning: ubi_io_read: error -5 while reading 126976 bytes from PEB 517:4096, read only 0 bytes, retry
> [    7.865374] ubi0 warning: ubi_io_read: error -5 while reading 126976 bytes from PEB 517:4096, read only 0 bytes, retry
> [    7.878742] ubi0 warning: ubi_io_read: error -5 while reading 126976 bytes from PEB 517:4096, read only 0 bytes, retry
> [    7.892104] ubi0 error: ubi_io_read: error -5 while reading 126976 bytes from PEB 517:4096, read 0 bytes
> [    7.901754] UBIFS error (ubi0:2 pid 687): ubifs_leb_read: reading 126976 bytes from LEB 1:0 failed, error -5
>
> After disabling continuous read, this flash works fine.
> So despite the datasheet stating support, I think it would be
> better to disable it first to avoid bricking MT7981 devices
> with MX35LFxGE4AD flash.

I am sorry but disabling the feature for every user of the chip does not
work if the problem comes from the SPI controller.

I just added Cheng Ming from Macronix. Cheng Ming, can you confirm these
chips support continuous read? Is there anything specific about them
that might cause this breakage?

Thanks a lot,
Miquèl