Definition for flash w25q128 is wrong

[Robert Marko]

On Mon, 1 Apr 2024 at 11:25, e9hack <e9hack at gmail.com> wrote:
>
> Am 01.04.2024 um 11:06 schrieb Robert Marko:
> > On Mon, 1 Apr 2024 at 10:32, e9hack <e9hack at gmail.com> wrote:
> >>
> >> Am 01.04.2024 um 10:14 schrieb Robert Marko:
> >>> On Mon, 1 Apr 2024 at 06:29, e9hack <e9hack at gmail.com> wrote:
> >>>>
> >>>> Hi,
> >>>>
> >>>> I'm using a TP-LINK WDR3600 with a bigger flash. Since some time the router hangs in an endless boot loop. I see the following message:
> >>>>
> >>>> ...
> >>>> [    0.402716] spi-nor spi0.0: BFPT parsing failed. Please consider using SPI_NOR_SKIP_SFDP when declaring the flash
> >>>> [    0.413217] spi-nor: probe of spi0.0 failed with error -22
> >>>> ...
> >>>> [    0.926180] /dev/root: Can't open blockdev
> >>>> [    0.930427] VFS: Cannot open root device "(null)" or unknown-block(0,0): error -6
> >>>> [    0.938037] Please append a correct "root=" boot option; here are the available partitions:
> >>>> [    0.946520] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)
> >>>> [    0.954914] Rebooting in 1 seconds..
> >>>>
> >>>> It looks like the definition for the flash is wrong:
> >>>>
> >>>> --- a/drivers/mtd/spi-nor/winbond.c     2024-03-15 19:27:50.000000000 +0100
> >>>> +++ b/drivers/mtd/spi-nor/winbond.c     2024-04-01 05:59:17.166780732 +0200
> >>>> @@ -120,8 +120,8 @@ static const struct flash_info winbond_n
> >>>>                   NO_SFDP_FLAGS(SECT_4K) },
> >>>>           { "w25q80bl", INFO(0xef4014, 0, 64 * 1024,  16)
> >>>>                   NO_SFDP_FLAGS(SECT_4K) },
> >>>> -       { "w25q128", INFO(0xef4018, 0, 0, 0)
> >>>> -               PARSE_SFDP
> >>>> +       { "w25q128", INFO(0xef4018, 0, 64 * 1024, 256)
> >>>> +               NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ)
> >>>>                   FLAGS(SPI_NOR_HAS_LOCK | SPI_NOR_HAS_TB) },
> >>>>           { "w25q256", INFO(0xef4019, 0, 64 * 1024, 512)
> >>>>                   NO_SFDP_FLAGS(SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ)
> >>>>
> >>>> With these changes, the flash will be detected properly.
> >>>
> >>> Yeah, I am not so sure this is correct as all w25q128 versions have
> >>> SFDP table so skipping SFDP parsing
> >>> isn't really correct.
> >>> Can you check what is the exact model you have?
> >>
> >> The chip (SOIC8) is marked with:
> >>
> >> winbond
> >> 25Q128FVSG
> >> 1327
> >
> > If it is Genuine Winbond then it has an SFDP table according to the datasheet:
> > https://www.winbond.com/hq/support/documentation/downloadV2022.jsp?__locale=en&xmlPath=/support/resources/.content/item/DA00-W25Q128FV.html&level=1
> >
> > AFAIK, all Winbond NOR with datecode 1124 and onwards have SFDP tables.
> >
> > Has this happened with kernel 6.1 or been going on for a while?
>
> My build from October is using kernel 5.15.133. I assume it is an issue of kernel 6.1.
>
> It looks like a driver issue. A SOIC8 housing doesn't support dual/quad SPI. The flash will be detect, if I change
>
> PARSE_SFDP to NO_SFDP_FLAGS(SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ).

Hm, it really looks like your revision has broken SFDP but since
Winbond in their ultimate wisdom decided
that it was best to share the same JEDEC ID with other revisions SFDP
was enabled via:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/drivers/mtd/spi-nor/winbond.c?h=v6.1.83&id=7c6ba20a0b9aeb82a6c097c74ccbecdda8e9fc25

So you really need to report this to the linux-mtd crowd.
Can you check if SFDP table can be dumped on your NOR, cause there
should be one but most likely one of the tables is broken?

Regards,
Robert
>
> Regards,
> Hartmut
>
> >
> > Regards,
> > Robert
> >
> >>
> >>>
> >>> Regards,
> >>> Robert
> >>>>
> >>>> Regards,
> >>>> Hartmut
> >>>>
> >>>> _______________________________________________
> >>>> openwrt-devel mailing list
> >>>> openwrt-devel at lists.openwrt.org
> >>>> https://lists.openwrt.org/mailman/listinfo/openwrt-devel
> >>
>