Intel Lynx Point LP Premium: W25Q64.V chip not detected correctly

[Denis Turischev]

Hi flashrom team,

I'm trying to read BIOS on SPI flash on Intel Lynx Point LP Premium chipset.
16M SPI flash consists of two Winbond chips 8M (W25Q64.V).

When I use ich_spi_mode=auto flash is not detected.
When I use ich_spi_mode=swseq detected only one 8M chip.

Can you help?

Denis

Intel Lynx Point LP Premium: W25Q64.V chip not detected correctly

[Stefan Tauner]

On Fri, 21 Mar 2014 14:00:10 +0200
Denis Turischev  wrote:
> Hi flashrom team,
>
> I'm trying to read BIOS on SPI flash on Intel Lynx Point LP Premium chipset.
> 16M SPI flash consists of two Winbond chips 8M (W25Q64.V).
>
> When I use ich_spi_mode=auto flash is not detected.
> When I use ich_spi_mode=swseq detected only one 8M chip.
>
> Can you help?

Hi,

swseq does not support the two-chip configuration (which is a hardware
limitation). In hwseq mode the software is not aware of any chips, hence
using the -c parameter leads to no detection. Leave out the -c
parameter and it will work... in general. Interestingly the flash
descriptor seems to be bogus in this case though. It specifies a wrong
density for the second chip, although the other addresses encoded in
there are according to your observation of two 8MB chips... either the
chipset is different or this is an error of the manu... oh, you ARE the
manufacturer ;)

Please check if the setting of FLCOMP in the flash descriptor is
according to the lynx point spi programming guide. I dont have access
to it. Either FLCOMP is set wrong or lynx point is different and this is
a bug in flashrom.

Intel Lynx Point LP Premium: W25Q64.V chip not detected correctly

[Denis Turischev]

Hi Stefan,

Thanks for the answer.

Without -c parameter it doesn't work too, it detects "opaque" chip of size 8M + 512K = 8704K.

I've checked density of second chip is set to 8M. Seems hardware is ok, we actually boot from this
16M bios, also we have Intel production tools for Windos/DOS, they do read/write all 16M.

How can I help you to resolve this?

Denis

On 03/22/2014 03:10 AM, Stefan Tauner wrote:
> On Fri, 21 Mar 2014 14:00:10 +0200
> Denis Turischev  wrote:
> > Hi flashrom team,
> >
> > I'm trying to read BIOS on SPI flash on Intel Lynx Point LP Premium chipset.
> > 16M SPI flash consists of two Winbond chips 8M (W25Q64.V).
> >
> > When I use ich_spi_mode=auto flash is not detected.
> > When I use ich_spi_mode=swseq detected only one 8M chip.
> >
> > Can you help?
>
> Hi,
>
> swseq does not support the two-chip configuration (which is a hardware
> limitation). In hwseq mode the software is not aware of any chips, hence
> using the -c parameter leads to no detection. Leave out the -c
> parameter and it will work... in general. Interestingly the flash
> descriptor seems to be bogus in this case though. It specifies a wrong
> density for the second chip, although the other addresses encoded in
> there are according to your observation of two 8MB chips... either the
> chipset is different or this is an error of the manu... oh, you ARE the
> manufacturer ;)
>
> Please check if the setting of FLCOMP in the flash descriptor is
> according to the lynx point spi programming guide. I dont have access
> to it. Either FLCOMP is set wrong or lynx point is different and this is
> a bug in flashrom.

Intel Lynx Point LP Premium: W25Q64.V chip not detected correctly

[Stefan Tauner]

On Mon, 24 Mar 2014 09:37:53 +0200
Denis Turischev  wrote:
> Hi Stefan,
>
> Thanks for the answer.
>
> Without -c parameter it doesn't work too, it detects "opaque" chip of size 8M + 512K = 8704K.
>
> I've checked density of second chip is set to 8M. Seems hardware is ok, we actually boot from this
> 16M bios, also we have Intel production tools for Windos/DOS, they do read/write all 16M.
>
> How can I help you to resolve this?

Hi,

I was able to obtain the respective SPI programming guide and
apparently Intel has changed the definition of (at least) the FLCOMP
descriptor register. This will require some changes in flashrom and I
don't have time for that currently, sorry.
It might be enough if you change the widths of the FLCOMP bitfields in
ich_descriptors.h corresponding to the densities (comp1_density and
comp2_density) to 4 (and the padding from 11 to 9) but there might be
more differences.