Intention to Port eCos to STR9

Intention to Port eCos to STR9

[C. Adam Lamb]

Hi Gang,

I’m currently looking into porting eCos to the STR912 processor.  The
processor is on the bleeding edge (it just came out), so I’m still waiting
for my eval board.

My first goal in this message is to both let everyone know that I should be
starting to work on this soon and see if anyone else might be working on
this as well…

Second goal is to have someone with some eCos experience make sure my
porting approach isn’t too far off.  I’m definitely new to eCos, so any
pointers will be awesome (I have been reading some of the other “How do I
port to X” threads).  The STR9 is using an ARM966-E/S core, so my intention
is to start with the eCos port that is for that core and adapt it to the
STR9.  I will be using the all of the GNU tools for building and debugging. 
I intend to purchase either Signum System’s JTAG Jet with Trace or the PEEDI
JTAG Emulator if I can’t get my company to pony up for the slightly more
expensive device from Signum.  Both of those devices claim to be compatible
with the GNU debuggers.  Anyone with experience with either of those
devices?  Other suggestions?

By the time I am finished with the port (the complete port might take 6
months or more depending on my employers priorities), I hope to have drivers
for USB, Ethernet, Flash, SPI, …, and possibly GPIB.  I’m not sure my
employer will allow me to contribute the GPIB portion, but we’ll see…  I
intend to contribute everything else back that I can.

Should be fun getting to know you all!  I hope to be more of a help than a
hindrance in this forum when it’s all said and done.  Forgive me while I
spin up!

Thanks!

Adam Lamb
Sr. Firmware Engineer
ILX Lightwave Corp
Bozeman, MT

Re: Intention to Port eCos to STR9

[Andrew Lunn]

On Wed, Jun 14, 2006 at 02:17:37PM -0600, C. Adam Lamb wrote:
> Hi Gang,
> 
> I?m currently looking into porting eCos to the STR912 processor.? The
> processor is on the bleeding edge (it just came out), so I?m still waiting
> for my eval board.
> 
> My first goal in this message is to both let everyone know that I should be
> starting to work on this soon and see if anyone else might be working on
> this as well?
>
> Second goal is to have someone with some eCos experience make sure my
> porting approach isn?t too far off.?

O.K. First tip. Try not to have tunnel vision on the STR912. Try
taking a look at all the STR7 and STR9 devices. Work out what is
common and what is different. Try to make your port as generic as
possible. So if for example the UART is the same for all devices, just
write a STR UART driver, not a STR912 UART driver. Same goes for SPI,
FLASH controller, etc. If a peripheral for the STR9xx is different
from the STR7xx write an STR9xx driver. And lastly, if the STR912 has
something unque, write an STR912 unique driver. Same goes for the
HAL. Try to split it up into a varient HAL for STR9xx and a specific
HAL STR912. Maybe it is even possible to have a three layer HAL, with
STR at the top. The different between the STR7 and STR9 will be the
MMU setup. So maybe it is possible to put the MMU code into the STR9xx
HAL package and have most of the code in the STR HAL package.

The advantage of doing it like this is it makes it easier for others
to do ports to other members of the STR family building on what you
have done. I made a mistake here with the AT91SAM family. I did a port
to the AT91SAM7S and later found out that the AT91SAM7X is 99% the
same. I then ended up with a naming mess since the AT91SAM7S package
now also supports the AT91SAM7X devices. And i guess sometime in the
future the AT91SAM9 is going to be supported by the AT91SAM7S
package. Ugly.

> I will be using the all of the GNU tools for building and debugging.?

That is normal. Using anything else for compiling is virtually
impossible. You do have more flexibility with debugging, but i still
recommend gdb. You can add a GUI on top if you like, eg Insight, DDD,
etc.

> I intend to purchase either Signum System?s JTAG Jet with Trace or the PEEDI
> JTAG Emulator if I can?t get my company to pony up for the slightly more
> expensive device from Signum.? Both of those devices claim to be compatible
> with the GNU debuggers.? Anyone with experience with either of those
> devices??

I've had good experiance with PEEDI with the AT91SAM7S development
board. You need an up to date version of gdb. The version on the eCos
website as part of the ARM toolchain is too old for breakpoints to
work properly.

     Andrew

Re[2]: Intention to Port eCos to STR9

[Evgeny Belyanco]

Thursday, June 15, 2006, 1:09:42 AM, you wrote:

AL> STR at the top. The different between the STR7 and STR9 will be the
AL> MMU setup. So maybe it is possible to put the MMU code into the STR9xx
AL> HAL package and have most of the code in the STR HAL package.

STR9 has no MMU!


Evgeny.

Re: Intention to Port eCos to STR9

[Andrew Lunn]

On Thu, Jun 15, 2006 at 01:15:31AM +0400, Evgeny Belyanco wrote:
> Thursday, June 15, 2006, 1:09:42 AM, you wrote:
> 
> AL> STR at the top. The different between the STR7 and STR9 will be the
> AL> MMU setup. So maybe it is possible to put the MMU code into the STR9xx
> AL> HAL package and have most of the code in the STR HAL package.
> 
> STR9 has no MMU!

Oh, interesting. I thought all ARM9 had an MMU. The ARM920T and
ARM922T do which is what i have a passing knowledge of.

[Goes check the ARM website]

ARM926EJ-S has an MMU
ARM946E-S has an MPU - Memory Protection Unit
ARM966E-S has nothing!
ARM968E-S has DMA. DMA?
ARM996HS has MPU

         Andrew

Re: Intention to Port eCos to STR9

[Ilija Koco]

Andrew Lunn wrote:
> On Wed, Jun 14, 2006 at 02:17:37PM -0600, C. Adam Lamb wrote:
>   
>> Hi Gang,
>>
>> I?m currently looking into porting eCos to the STR912 processor.? The
>> processor is on the bleeding edge (it just came out), so I?m still waiting
>> for my eval board.
>>
>> My first goal in this message is to both let everyone know that I should be
>> starting to work on this soon and see if anyone else might be working on
>> this as well?
>>
>> Second goal is to have someone with some eCos experience make sure my
>> porting approach isn?t too far off.?
>>     
>
> O.K. First tip. Try not to have tunnel vision on the STR912. Try
> taking a look at all the STR7 and STR9 devices. Work out what is
> common and what is different. Try to make your port as generic as
> possible. So if for example the UART is the same for all devices, just
> write a STR UART driver, not a STR912 UART driver. Same goes for SPI,
> FLASH controller, etc.
Also check whether UART (or whatever) is same with some other (allready
ported) platform.
> If a peripheral for the STR9xx is different
> from the STR7xx write an STR9xx driver. And lastly, if the STR912 has
> something unque, write an STR912 unique driver. Same goes for the
> HAL. Try to split it up into a varient HAL for STR9xx and a specific
> HAL STR912. Maybe it is even possible to have a three layer HAL, with
> STR at the top. The different between the STR7 and STR9 will be the
> MMU setup. So maybe it is possible to put the MMU code into the STR9xx
> HAL package and have most of the code in the STR HAL package.
>
> The advantage of doing it like this is it makes it easier for others
> to do ports to other members of the STR family building on what you
> have done. I made a mistake here with the AT91SAM family. I did a port
> to the AT91SAM7S and later found out that the AT91SAM7X is 99% the
> same. I then ended up with a naming mess since the AT91SAM7S package
> now also supports the AT91SAM7X devices. And i guess sometime in the
> future the AT91SAM9 is going to be supported by the AT91SAM7S
> package. Ugly.
>
>   
>> I will be using the all of the GNU tools for building and debugging.?
>>     
>
> That is normal. Using anything else for compiling is virtually
> impossible. You do have more flexibility with debugging, but i still
> recommend gdb. You can add a GUI on top if you like, eg Insight, DDD,
> etc.
>
>   
>> I intend to purchase either Signum System?s JTAG Jet with Trace or the PEEDI
>> JTAG Emulator if I can?t get my company to pony up for the slightly more
>> expensive device from Signum.? Both of those devices claim to be compatible
>> with the GNU debuggers.? Anyone with experience with either of those
>> devices??
>>     
>
> I've had good experiance with PEEDI with the AT91SAM7S development
> board. You need an up to date version of gdb. The version on the eCos
> website as part of the ARM toolchain is too old for breakpoints to
> work properly.
>   
I'm using PEEDI with standard website gdb/Insight on LInux and it's ok.
But maybe I have newer firmware.

Ilija

RE: Intention to Port eCos to STR9

[C. Adam Lamb]

Thanks for all the input!

I'll summarize a few things (and correct me if I'm wrong or miss anything):


1) I should really try to make my port as generic as possible so things that
apply to all STR parts will be in one file with the parts that appear to be
specific to the STR912 will be in another file.  I'm not exactly sure at
this point how this will look (I need to start looking at some HAL code)
since the STR912 doesn't have an MMU and most of the others do...

2) PEEDI appears to be a good tool with good support.  I was leaning towards
the JTAG Jet tool from Signum because of the trace capability (I now have
the OK to buy that one, so I should be able to purchase whatever I want
now)...  Does anyone see any reasons not to get the Signum tool?  Maybe I
should just stick with PEEDI since it is proven out and they know what eCos
is?

3) Look into re-using drivers from other ports.

4) The closest port to start from will be the ARM Integrator Platform that
uses the ARM966E-S core since the core is the same as the STR9.  Again,
their might be drivers from other ports that I can pull in, but the starting
point will probably be the ARM966E-S port.

New Question:

I have heard a few people talk about not using Redboot, and instead just
using their JTAG tools to do the loading, flashing, debugging, etc.  I
assume that since I will have a JTAG tool I wouldn't necessarily need to use
Redboot, but I am guessing it still might make life easier for a couple of
reasons.  First, it will be an easier way to prove out that my HAL is
correct with the simpler set of code.  Second, the community will like me
slightly more if I do the work to provide a more complete port for the
STR9...

Thanks!
Adam

Re: Intention to Port eCos to STR9

[Andrew Lunn]

> New Question:
> 
> I have heard a few people talk about not using Redboot, and instead just
> using their JTAG tools to do the loading, flashing, debugging, etc.  I
> assume that since I will have a JTAG tool I wouldn't necessarily need to use
> Redboot, but I am guessing it still might make life easier for a couple of
> reasons.  First, it will be an easier way to prove out that my HAL is
> correct with the simpler set of code.  Second, the community will like me
> slightly more if I do the work to provide a more complete port for the
> STR9...

With a JTAG programmer, redboot is optional. Also, since your amount
of FLASH is limited, not using redboot does give your application more
room.

As for testing, redboot is actually quite complex. There are simpler
things to start with. Try some of the test programs in the HAL package
to start with. 

Once you have a basic port working getting redboot going is normally
trival.

        Andrew

RE: Intention to Port eCos to STR9

[Joakim Langlet]

Hi Adam & others,
I am currently looking at using the STR912FW44 and I have eCos in mind.
A evaluation board will turn up in a (hopefully) not too distant future.
Any progress on this port would be interesting to know about.
Hopefully, I will be able to contribute with something.
Regards
Joakim Langlet
Seaview AB

RE: Intention to Port eCos to STR9

[C. Adam Lamb]

I'm glad to hear there are some others with interest in the STR912 as well!
Welcome to the group Joakim!

I am also waiting for an evaluation board.  I have one on order from Hitex
(in Germany) that should be arriving on my doorstep at any time (it shipped
last week).  I am in the middle of studying the HAL right now, so I haven't
started any coding yet.  

I will post my progress here on the forum as I make it.

Although I heard a lot of great things about the PEEDI JTAG emulator, I have
decided to go with the Signum Systems JTAG-JET emulator because it has the
trace capabilities.  I realize that I probably won't use the trace all that
much -- but when you need a trace, they can save days and days of time...

Adam

-----Original Message-----
From: ecos-devel-owner@ecos.sourceware.org
[mailto:ecos-devel-owner@ecos.sourceware.org] On Behalf Of Joakim Langlet
Sent: Thursday, June 15, 2006 3:56 PM
To: ecos-devel@sourceware.org
Subject: RE: Intention to Port eCos to STR9

Hi Adam & others,
I am currently looking at using the STR912FW44 and I have eCos in mind.
A evaluation board will turn up in a (hopefully) not too distant future.
Any progress on this port would be interesting to know about.
Hopefully, I will be able to contribute with something.
Regards
Joakim Langlet
Seaview AB

RE: Intention to Port eCos to STR9

[fabio]

Joakim Langlet wrote:
> 
> Hi Adam & others,
> I am currently looking at using the STR912FW44 and I have eCos in mind.
> A evaluation board will turn up in a (hopefully) not too distant future.
> Any progress on this port would be interesting to know about.
> Hopefully, I will be able to contribute with something.
> Regards
> Joakim Langlet
> Seaview AB
> 
> 
I'm looking for the STR912FW44 too but with freeRtos. I don't known eCos:
what are the main differences ?
Do u know if a gcc compiler is available?
thanks.

Fabio Salvicchi
Firenze - Italy
-- 
View this message in context: http://www.nabble.com/Intention-to-Port-eCos-to-STR9-tf1788352.html#a6017671
Sent from the Sourceware - ecos-devel forum at Nabble.com.

RE: Intention to Port eCos to STR9

[Mohd_Arif]

Hi i m looking into porting ECOS in STR9 board. Can somebody send or arrange
the source code for CAN
driver. I've just read about Controller Area Network and i want to enable it
in my board. If somebody 
can arrange, i'll really be thankful him and to all of you.

fabio wrote:
> 
> 
> 
> Joakim Langlet wrote:
>> 
>> Hi Adam & others,
>> I am currently looking at using the STR912FW44 and I have eCos in mind.
>> A evaluation board will turn up in a (hopefully) not too distant future.
>> Any progress on this port would be interesting to know about.
>> Hopefully, I will be able to contribute with something.
>> Regards
>> Joakim Langlet
>> Seaview AB
>> 
>> 
> I'm looking for the STR912FW44 too but with freeRtos. I don't known eCos:
> what are the main differences ?
> Do u know if a gcc compiler is available?
> thanks.
> 
> Fabio Salvicchi
> Firenze - Italy
> 

-- 
View this message in context: http://www.nabble.com/Intention-to-Port-eCos-to-STR9-tp4871981p18684933.html
Sent from the Sourceware - ecos-devel mailing list archive at Nabble.com.

Re: Intention to Port eCos to STR9

[Ilija Koco]

C. Adam Lamb wrote:
> Thanks for all the input!
>
> I'll summarize a few things (and correct me if I'm wrong or miss anything):
>
>
> 1) I should really try to make my port as generic as possible so things that
> apply to all STR parts will be in one file with the parts that appear to be
> specific to the STR912 will be in another file.  I'm not exactly sure at
> this point how this will look (I need to start looking at some HAL code)
> since the STR912 doesn't have an MMU and most of the others do...
>   
Practically you need to make a variant port (STR) and platform port
(your board). then you may also need to make some serial driver which
can be independent from port.
> 2) PEEDI appears to be a good tool with good support.  I was leaning towards
> the JTAG Jet tool from Signum because of the trace capability (I now have
> the OK to buy that one, so I should be able to purchase whatever I want
> now)...  Does anyone see any reasons not to get the Signum tool?  Maybe I
> should just stick with PEEDI since it is proven out and they know what eCos
> is?
>   
PEEDI has Ethernet (TCP/IP) interface so it fits seemlesly (without need
for a driver) with any host OS (Linux, Windows, etc), that is not case
withUSB devices. Also you can remotely connect to it from everywhere in
yiour LAN. Actually by mans of Internet you can connect from everywhere
(dbg, telnet).
> 3) Look into re-using drivers from other ports.
>   
Yes, no need to duplicate code. Chack your onchip pripherals with
"ported" chips. If not same you may find something simillar.
> 4) The closest port to start from will be the ARM Integrator Platform that
> uses the ARM966E-S core since the core is the same as the STR9.  Again,
> their might be drivers from other ports that I can pull in, but the starting
> point will probably be the ARM966E-S port.
>   
I did my 1st port recently and I had same dillema. I choosed the one
with most simillar booting sequence and memory structure. I was looking
for peripheral simillarity too, but unfortunatelly i found no chips with
peripharals simillar to mine.

Porting eCos to STR9

[C. Adam Lamb]

> I did my 1st port recently and I had same dillema. I choosed the one
> with most simillar booting sequence and memory structure. I was looking
> for peripheral simillarity too, but unfortunatelly i found no chips with
> peripharals simillar to mine.

How long did it take you to get the basic part (enough to get an LED
blinking, for example) of your first port going?  A few weeks?  

Anybody have any idea about how much time I should expect to take to just
get the basics running (assuming I'm an average guy) before I start on all
of the peripherals?

Adam

Re: Porting eCos to STR9

[Ilija Koco]

C. Adam Lamb wrote:
>> I did my 1st port recently and I had same dillema. I choosed the one
>> with most simillar booting sequence and memory structure. I was looking
>> for peripheral simillarity too, but unfortunatelly i found no chips with
>> peripharals simillar to mine.
>>     
>
> How long did it take you to get the basic part (enough to get an LED
> blinking, for example) of your first port going?  A few weeks?  
>   
Well' my biggest effort was to consume and compile vast amount of
documentation. While having experience with porting (proprietary) OS and
some other processors I was completely novice at ARM and eCos. So I
spent about 2 week in reading, waiting for my PEEDI.

Then I had to test my hardware (I hadn't off the shelf evaluation board)
during which time I also got familiar with ARM and found out that eCos'
downloadable GCC doesn't support big endian. You may not encounter this
issue but if your chip is BE you will need to build your GNU tool chain
<http://ecos.sourceware.org/build-toolchain.html>

Once I got this I started porting and within 3-4 weeks I got 1st message
at serial line through hal_diag functions (I was familiar with similar
serial port to one I ported).

Porting process is very well described in "eCos reference" and
"Component writer's guide" and also explained (with example) here
http://www.phptr.com/browse/product.asp?product_id={0BEE58C4-0812-4B13-9B8C-A0FC31A4C02F}

> Anybody have any idea about how much time I should expect to take to just
> get the basics running (assuming I'm an average guy) before I start on all
> of the peripherals?
>
> Adam
>
>
>   

Re: Porting eCos to STR9

[Ilija Koco]

Ilija Koco wrote:
> C. Adam Lamb wrote:
>   
>>> I did my 1st port recently and I had same dillema. I choosed the one
>>> with most simillar booting sequence and memory structure. I was looking
>>> for peripheral simillarity too, but unfortunatelly i found no chips with
>>> peripharals simillar to mine.
>>>     
>>>       
>> How long did it take you to get the basic part (enough to get an LED
>> blinking, for example) of your first port going?  A few weeks?  
>>   
>>     
> Well' my biggest effort was to consume and compile vast amount of
> documentation. While having experience with porting (proprietary) OS and
> some other processors I was completely novice at ARM and eCos. So I
> spent about 2 week in reading, waiting for my PEEDI.
>
> Then I had to test my hardware (I hadn't off the shelf evaluation board)
> during which time I also got familiar with ARM and found out that eCos'
> downloadable GCC doesn't support big endian. You may not encounter this
> issue but if your chip is BE you will need to build your GNU tool chain
> <http://ecos.sourceware.org/build-toolchain.html>
>
> Once I got this I started porting and within 3-4 weeks I got 1st message
> at serial line through hal_diag functions (I was familiar with similar
> serial port to one I ported).
>
> Porting process is very well described in "eCos reference" and
> "Component writer's guide" and also explained (with example) here
> http://www.phptr.com/browse/product.asp?product_id={0BEE58C4-0812-4B13-9B8C-A0FC31A4C02F}
>
>   
I placed wrong link, sorry!

It should be this
http://www.phptr.com/bookstore/product.asp?isbn=0130354732&redir=1

And here in PDF
http://www.informit.com/content/downloads/perens/0130354732.pdf

>> Anybody have any idea about how much time I should expect to take to just
>> get the basics running (assuming I'm an average guy) before I start on all
>> of the peripherals?
>>
>> Adam
>>
>>
>>   
>>     
>
>
>   

Re: Intention to Port eCos to STR9

[Evgeny Belyanco]

Thursday, June 15, 2006, 12:17:37 AM, you wrote:

CAL> I’m currently looking into porting eCos to the STR912 processor.  The
CAL> processor is on the bleeding edge (it just came out), so I’m still waiting
CAL> for my eval board.
...

CAL> By the time I am finished with the port (the complete port might take 6
CAL> months or more depending on my employers priorities), I hope to have drivers
CAL> for USB, Ethernet, Flash, SPI, …, and possibly GPIB.  I’m not sure my
CAL> employer will allow me to contribute the GPIB portion, but we’ll see…  I
CAL> intend to contribute everything else back that I can.

Are you plans actual? Is it possible to hope on eCos STR91x port near
Dec'06?

Evgeny Belyanko
**********************************
* E-mail: ea@kbkcc.ru
**********************************