Ramzi Maalouf wrote:

> Hello -
> I am trying to use Embedded Linux 2.4.25 PPC port for proprietary
> hardware. It is very much like the walnut, except it does not have
> some of the things the walnut has, e.g., no NVRAM with TOD clock, no
> FPGA, etc. Can someone give me a quick tutorial or point to where I
> can find some information on adjusting the Kernel. I know that a
> minimum modification to board_setup_arch (platforms/walnut.c) has to
> get adjusted, but what else?

How could we know? Didn't you say proprietary?

List everything that you do not handle exaclty as walnut and you have got
your list...

> I am linux kernel newbie!

Then try to find someone that atleast have rebuilt x86 kernels...
Any additional embedded experience will help too (for toolchain issues).

/RogerL

--
Roger Larsson
Skellefteå
Sweden

Roger Larsson wrote in message news:<2P4xc.1556$dx3.14047@newsb.telia.net>...
> Ramzi Maalouf wrote:
>
> > Hello -
> > I am trying to use Embedded Linux 2.4.25 PPC port for proprietary
> > hardware. It is very much like the walnut, except it does not have
> > some of the things the walnut has, e.g., no NVRAM with TOD clock, no
> > FPGA, etc. Can someone give me a quick tutorial or point to where I
> > can find some information on adjusting the Kernel. I know that a
> > minimum modification to board_setup_arch (platforms/walnut.c) has to
> > get adjusted, but what else?
>
> How could we know? Didn't you say proprietary?
>
> List everything that you do not handle exaclty as walnut and you have got
> your list...
>
> > I am linux kernel newbie!
>
> Then try to find someone that atleast have rebuilt x86 kernels...
> Any additional embedded experience will help too (for toolchain issues).
>
> /RogerL

I have rebuilt the kernel several times on several different
platforms. But, I am not familiar with the inner workings of it.
As to the differences, Time Of Day Clock seems to be the big problem.
I tried to use the walnut kernel on our hardware. The xmon appears,
after going through the backtrace, and the system map, I discovered
the walnut tries to map the FPGA). I modified walnut.c, under the
platforms directory, to not map the FPGA, now the NVRAM (with the
Time of Day clock) is the problem. What should I do about the Time of
Day clock ?
-ramzi

Ramzi Maalouf wrote:

> Roger Larsson wrote in message
> news:<2P4xc.1556$dx3.14047@newsb.telia.net>...
>> Ramzi Maalouf wrote:
>>
>> > Hello -
>> > I am trying to use Embedded Linux 2.4.25 PPC port for proprietary
>> > hardware. It is very much like the walnut, except it does not have
>> > some of the things the walnut has, e.g., no NVRAM with TOD clock, no
>> > FPGA, etc. Can someone give me a quick tutorial or point to where I
>> > can find some information on adjusting the Kernel. I know that a
>> > minimum modification to board_setup_arch (platforms/walnut.c) has to
>> > get adjusted, but what else?
>>
>> How could we know? Didn't you say proprietary?
>>
>> List everything that you do not handle exaclty as walnut and you have got
>> your list...
>>
>> > I am linux kernel newbie!
>>
>> Then try to find someone that atleast have rebuilt x86 kernels...
>> Any additional embedded experience will help too (for toolchain issues).
>>
>> /RogerL
>
> I have rebuilt the kernel several times on several different
> platforms.

Then you are no linux newbie!

> But, I am not familiar with the inner workings of it.
> As to the differences, Time Of Day Clock seems to be the big problem.
> I tried to use the walnut kernel on our hardware. The xmon appears,
> after going through the backtrace, and the system map, I discovered
> the walnut tries to map the FPGA). I modified walnut.c, under the
> platforms directory, to not map the FPGA, now the NVRAM (with the
> Time of Day clock) is the problem. What should I do about the Time of
> Day clock ?

So - you do not need current time after a power cycle then?
If you have network connection you can use NTP to get
accurate time (or a simple query, add half round trip time before
setting). For kernel NTP see "linux/kernel/timer.c"
Well that is what the HW timer is used for, after that time is mostly
handled by counting interrupts (I think it can be used as a NTP source for
the kernel... but I might remember wrong)

So my guess is that you may remove all code that works with the Time of Day
clock - that will probably give you a clock that restarts 1 Jan 1970 at
each power on...

/RogerL

--
Roger Larsson
Skellefteå
Sweden