[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[Sander Vermin]

Hi All,

I have been struggling the last week to get the macb Ethernet driver in 
MII mode.

[in short]
I have communication over the MDIO interface, my activity led blinks but 
no ehternet.

[the long story]
I have a custom board build out of the schematics of the AT91SAM9260EK 
board. On my boar I did not use the DM9161 but the KS8721chip. On the 
ATMEL board the PHY works in RMII mode, on my board I used MII mode like 
this board from OLIMEX [1].

Olimex was kind enough to make Uboot nandflash build working with there 
board, And deliver sources I cant compile, due to an error: Hardware 
float vs software float. But I want a dataflash version because I am 
using a BGA chip with nandflash bug.

In my current setup I use the latest git of the AT91 tree with the 
following defines:
#define CONFIG_MACB            1
#define    CONFIG_MII                1
#undef CONFIG_RMII
#define CONFIG_NET_MULTI        1
#define CONFIG_NET_RETRY_COUNT        5000
#define CONFIG_RESET_PHY_R        1

With this config I have the MDIO interface working, when I try to ping 
or TFTP I only get time outs / host not alive errors.

Is there someone out there how has the MII mode working?

Best regards,

Sander


[1] http://www.olimex.com/dev/sam9-L9260.html (schematics on the bottom 
of the page)

[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[Stelian Pop]

Le lundi 14 avril 2008 ? 10:10 +0200, Sander Vermin a ?crit :
> Hi All,
> 
> I have been struggling the last week to get the macb Ethernet driver in 
> MII mode.
> 
> [in short]
> I have communication over the MDIO interface, my activity led blinks but 
> no ehternet.
> 
> [the long story]
> I have a custom board build out of the schematics of the AT91SAM9260EK 
> board. On my boar I did not use the DM9161 but the KS8721chip. On the 
> ATMEL board the PHY works in RMII mode, on my board I used MII mode like 
> this board from OLIMEX [1].

Do you have a working Linux kernel with a working network interface ? 
The Linux and U-Boot macb driver and PIO configuration is very close,
and if you manage to make it work under Linux you'll probably have
little trouble finding out the problem.

> Olimex was kind enough to make Uboot nandflash build working with there 
> board, And deliver sources I cant compile, due to an error: Hardware 
> float vs software float. But I want a dataflash version because I am 
> using a BGA chip with nandflash bug.

I don't see what NAND has to do with ethernet here.

> In my current setup I use the latest git of the AT91 tree with the 
> following defines:
> #define CONFIG_MACB            1
> #define    CONFIG_MII                1
> #undef CONFIG_RMII
> #define CONFIG_NET_MULTI        1
> #define CONFIG_NET_RETRY_COUNT        5000
> #define CONFIG_RESET_PHY_R        1
> 
> With this config I have the MDIO interface working, when I try to ping 
> or TFTP I only get time outs / host not alive errors.

Does the board send anything at all ? Try sniffing the net wire (using
tcpdump for example).

Also, what cross chain are you using ? I have seen strange network
errors when using recent compilers which after investigation were -Os
optimisation errors. (using the latest CodeSourcery toolchain for
example).

> Is there someone out there how has the MII mode working?

If I'm correct, Olimex has. So closer inspection of their code may
reveal some information.

Stelian.
-- 
Stelian Pop <stelian@popies.net>

[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[Sander Vermin]

Stelian Pop schreef:
> Le lundi 14 avril 2008 ? 10:10 +0200, Sander Vermin a ?crit :
>   
>> Hi All,
>>
>> I have been struggling the last week to get the macb Ethernet driver in 
>> MII mode.
>>
>> [in short]
>> I have communication over the MDIO interface, my activity led blinks but 
>> no ehternet.
>>
>> [the long story]
>> I have a custom board build out of the schematics of the AT91SAM9260EK 
>> board. On my boar I did not use the DM9161 but the KS8721chip. On the 
>> ATMEL board the PHY works in RMII mode, on my board I used MII mode like 
>> this board from OLIMEX [1].
>>     
>
> Do you have a working Linux kernel with a working network interface ? 
> The Linux and U-Boot macb driver and PIO configuration is very close,
> and if you manage to make it work under Linux you'll probably have
> little trouble finding out the problem.
>   
Are all settings of uboot overwritten by linux? I am not to comfortable 
hacking in the Linux kernel and I had the focus on Uboot.
>> Olimex was kind enough to make Uboot nandflash build working with there 
>> board, And deliver sources I cant compile, due to an error: Hardware 
>> float vs software float. But I want a dataflash version because I am 
>> using a BGA chip with nandflash bug.
>>     
>
> I don't see what NAND has to do with ethernet here.
>   
The AT91SAM9260 BGA chip has a bug, that booting from NAND has problems.
>   
>> In my current setup I use the latest git of the AT91 tree with the 
>> following defines:
>> #define CONFIG_MACB            1
>> #define    CONFIG_MII                1
>> #undef CONFIG_RMII
>> #define CONFIG_NET_MULTI        1
>> #define CONFIG_NET_RETRY_COUNT        5000
>> #define CONFIG_RESET_PHY_R        1
>>
>> With this config I have the MDIO interface working, when I try to ping 
>> or TFTP I only get time outs / host not alive errors.
>>     
>
> Does the board send anything at all ? Try sniffing the net wire (using
> tcpdump for example).
>
> Also, what cross chain are you using ? I have seen strange network
> errors when using recent compilers which after investigation were -Os
> optimisation errors. (using the latest CodeSourcery toolchain for
> example).
>   
I am using the compilers from buildroot.
>   
>> Is there someone out there how has the MII mode working?
>>     
>
> If I'm correct, Olimex has. So closer inspection of their code may
> reveal some information.
>
> Stelian.
>   
Olimex has indeed, but that is a old version of uboot, the old ETHER 
driver and not the new. I inspected the initialization code on the ARM 
side, that was the same. The rest of te code is completely different, so 
spotting differences is difficult.

Sander

[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[Stelian Pop]

Le lundi 14 avril 2008 ? 11:19 +0200, Sander Vermin a ?crit :

> > Do you have a working Linux kernel with a working network interface ? 
> > The Linux and U-Boot macb driver and PIO configuration is very close,
> > and if you manage to make it work under Linux you'll probably have
> > little trouble finding out the problem.
> >   
> Are all settings of uboot overwritten by linux?

Almost all the settings are overwritten, yes.

>  I am not to comfortable 
> hacking in the Linux kernel and I had the focus on Uboot.

So you do not have a working Linux kernel. This was my question.

> >> Olimex was kind enough to make Uboot nandflash build working with there 
> >> board, And deliver sources I cant compile, due to an error: Hardware 
> >> float vs software float. But I want a dataflash version because I am 
> >> using a BGA chip with nandflash bug.
> >>     
> >
> > I don't see what NAND has to do with ethernet here.
> >   
> The AT91SAM9260 BGA chip has a bug, that booting from NAND has problems.

Ok, but this has nothing to do with the Ethernet, right ? If the
ethernet is supposed to work with their U-Boot version, it will probably
work as well if you configure U-Boot to boot from dataflash instead of
NAND flash.

> > Also, what cross chain are you using ? I have seen strange network
> > errors when using recent compilers which after investigation were -Os
> > optimisation errors. (using the latest CodeSourcery toolchain for
> > example).
> >   
> I am using the compilers from buildroot.

This doesn't learn us anything about the gcc version you're using.

The official toolchain for U-Boot is the DENX ELDK:
http://www.denx.de/wiki/DULG/ELDK . I don't really think this is your
issue, but when nothing works it may make sense to put yourself in a
well known configuration.

> Olimex has indeed, but that is a old version of uboot, the old ETHER 
> driver and not the new. I inspected the initialization code on the ARM 
> side, that was the same. The rest of te code is completely different, so 
> spotting differences is difficult.

Ah, I understand. However, there must be a difference somewhere. I'm not
sure about this board, but on the SAM boards a software reset has to be
performed once the PHY address is configured to activate the PHY (look
into at91sam9260.c). Maybe your board needs something equivalent ?

-- 
Stelian Pop <stelian@popies.net>

[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[Sander Vermin]

Stelian Pop schreef:
> Le lundi 14 avril 2008 ? 11:19 +0200, Sander Vermin a ?crit :
>
>   
>>> Do you have a working Linux kernel with a working network interface ? 
>>> The Linux and U-Boot macb driver and PIO configuration is very close,
>>> and if you manage to make it work under Linux you'll probably have
>>> little trouble finding out the problem.
>>>   
>>>       
>> Are all settings of uboot overwritten by linux?
>>     
>
> Almost all the settings are overwritten, yes.
>
>   
Oke, olimex has a demo linux with working ethernet.. I can try to put 
this on my board with my current uboot.
>>  I am not to comfortable 
>> hacking in the Linux kernel and I had the focus on Uboot.
>>     
>
> So you do not have a working Linux kernel. This was my question.
>
>   
>>>> Olimex was kind enough to make Uboot nandflash build working with there 
>>>> board, And deliver sources I cant compile, due to an error: Hardware 
>>>> float vs software float. But I want a dataflash version because I am 
>>>> using a BGA chip with nandflash bug.
>>>>     
>>>>         
>>> I don't see what NAND has to do with ethernet here.
>>>   
>>>       
>> The AT91SAM9260 BGA chip has a bug, that booting from NAND has problems.
>>     
>
> Ok, but this has nothing to do with the Ethernet, right ? If the
> ethernet is supposed to work with their U-Boot version, it will probably
> work as well if you configure U-Boot to boot from dataflash instead of
> NAND flash.
>
>   
>>> Also, what cross chain are you using ? I have seen strange network
>>> errors when using recent compilers which after investigation were -Os
>>> optimisation errors. (using the latest CodeSourcery toolchain for
>>> example).
>>>   
>>>       
>> I am using the compilers from buildroot.
>>     
>
> This doesn't learn us anything about the gcc version you're using.
>
> The official toolchain for U-Boot is the DENX ELDK:
> http://www.denx.de/wiki/DULG/ELDK . I don't really think this is your
> issue, but when nothing works it may make sense to put yourself in a
> well known configuration.
>
>   
Is there some pre compiled  arm version?
>> Olimex has indeed, but that is a old version of uboot, the old ETHER 
>> driver and not the new. I inspected the initialization code on the ARM 
>> side, that was the same. The rest of te code is completely different, so 
>> spotting differences is difficult.
>>     
>
> Ah, I understand. However, there must be a difference somewhere. I'm not
> sure about this board, but on the SAM boards a software reset has to be
> performed once the PHY address is configured to activate the PHY (look
> into at91sam9260.c). Maybe your board needs something equivalent ?
>
>   
Currently I am using the at91sam9260ek config with the options above edited.

Sander

[U-Boot-Users] Uboot (AT91 tree) macb in MII mode

[Stelian Pop]

Le lundi 14 avril 2008 ? 11:52 +0200, Sander Vermin a ?crit :
> >
> > The official toolchain for U-Boot is the DENX ELDK:
> > http://www.denx.de/wiki/DULG/ELDK . I don't really think this is your
> > issue, but when nothing works it may make sense to put yourself in a
> > well known configuration.
> >
> >   
> Is there some pre compiled  arm version?

ftp://ftp.denx.de/pub/eldk/4.1/arm-linux-x86/iso

ARM support in latest ELDK (4.2) seems to lag a bit ...

> >> Olimex has indeed, but that is a old version of uboot, the old ETHER 
> >> driver and not the new. I inspected the initialization code on the ARM 
> >> side, that was the same. The rest of te code is completely different, so 
> >> spotting differences is difficult.
> >>     
> >
> > Ah, I understand. However, there must be a difference somewhere. I'm not
> > sure about this board, but on the SAM boards a software reset has to be
> > performed once the PHY address is configured to activate the PHY (look
> > into at91sam9260.c). Maybe your board needs something equivalent ?
> >
> >   
> Currently I am using the at91sam9260ek config with the options above edited.

This means that you're doing the soft-reset sequence I told you above
(in boards/atmel/at91sam9260ek/at91sam9260ek.c). Maybe your board
doesn't need this, try commenting out the relevant code in
at91sam9260ek_macb_hw_init() (keep only the gpio set_A/B_periph().

-- 
Stelian Pop <stelian@popies.net>