[U-Boot-Users] Extremely tiny footprint

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

Hi everybody,

I would like to use U-Boot as a first-stage boot loader to program flash 
memory on the target board. The idea is to perform a two-steps procedure:

1. Program the first-stage boot loader in flash using JTAG (slow)
2. The first-stage boot loader downloads the final boot loader (full-featured 
U-Boot) using TFTP and writes it to flash (fast)

To minimize time spent in step one, I need to make the first-stage boot loader 
as small as possible (hopefully less than 32kB). Has anyone ever stripped 
U-Boot that much ? Is there any documentation available ?

Laurent Pinchart

[U-Boot-Users] Extremely tiny footprint

[Martin Krause]

Hi Laurent,

u-boot-users-bounces at lists.sourceforge.net wrote on :
> Hi everybody,
> 
> I would like to use U-Boot as a first-stage boot loader to program
> flash memory on the target board. The idea is to perform a two-steps
> procedure: 
> 
> 1. Program the first-stage boot loader in flash using JTAG (slow)

There are JTAG programmers, which could work very fast (e. g. BDI2000
with workspace enabled). The limiting factor is the flash itself,
not the programmer, or the JTAG interface.

> 2. The first-stage boot loader downloads the final boot loader
> (full-featured U-Boot) using TFTP and writes it to flash (fast)
> 
> To minimize time spent in step one, I need to make the first-stage
> boot loader as small as possible (hopefully less than 32kB). Has
> anyone ever stripped U-Boot that much ? Is there any documentation
> available ?

If programming speed is the only reason for your two-step approach,
then IMO this is not reasonable. It would be a waste of your time. Use
a faster JTAG programmer instead. Or you could buy the flashes 
preprogrammed.

Regards,
Martin

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

Hi Martin,

> > I would like to use U-Boot as a first-stage boot loader to program
> > flash memory on the target board. The idea is to perform a two-steps
> > procedure:
> >
> > 1. Program the first-stage boot loader in flash using JTAG (slow)
>
> There are JTAG programmers, which could work very fast (e. g. BDI2000
> with workspace enabled). The limiting factor is the flash itself,
> not the programmer, or the JTAG interface.
>
> > 2. The first-stage boot loader downloads the final boot loader
> > (full-featured U-Boot) using TFTP and writes it to flash (fast)
> >
> > To minimize time spent in step one, I need to make the first-stage
> > boot loader as small as possible (hopefully less than 32kB). Has
> > anyone ever stripped U-Boot that much ? Is there any documentation
> > available ?
>
> If programming speed is the only reason for your two-step approach,
> then IMO this is not reasonable. It would be a waste of your time. Use
> a faster JTAG programmer instead. Or you could buy the flashes
> preprogrammed.

Buying the flashes preprogrammed is not an option given the quantities.

We already have a BDI2000 which has issues programming the flash using the 
workspace, so it reverts to the slow pure JTAG mode. I still have to 
investigate this directly with Abatron, we haven't found a solution so far.

Even using the internal workspace, the BDI2000 is not a very practical 
solution, as it wouldn't be easy to integrate it in our test bench 
architecture. The price is also an issue. This is why I was looking for 
another solution. The alternative I thought of is using a pure JTAG 
programmer (slow) with a tiny ethernet-enabled boot loader. The Freescale 
JTAG commands used to control the processor are not documented, so I can't 
develop a faster JTAG-only alternative. Are you aware of any other solution ?

Laurent Pinchart

[U-Boot-Users] Extremely tiny footprint

[Amnon Cohen]

If you remove any commands you don't actually need from CONFIG_COMMANDS
you can strip you executable size down a fair bit. But if you need tftp
support,
you won't be able to get down to 32Kb.

Good luck
   Amnon

On 3/2/07, Laurent Pinchart <laurent.pinchart@technotrade.biz> wrote:
>
> Hi Martin,
>
> > > I would like to use U-Boot as a first-stage boot loader to program
> > > flash memory on the target board. The idea is to perform a two-steps
> > > procedure:
> > >
> > > 1. Program the first-stage boot loader in flash using JTAG (slow)
> >
> > There are JTAG programmers, which could work very fast (e. g. BDI2000
> > with workspace enabled). The limiting factor is the flash itself,
> > not the programmer, or the JTAG interface.
> >
> > > 2. The first-stage boot loader downloads the final boot loader
> > > (full-featured U-Boot) using TFTP and writes it to flash (fast)
> > >
> > > To minimize time spent in step one, I need to make the first-stage
> > > boot loader as small as possible (hopefully less than 32kB). Has
> > > anyone ever stripped U-Boot that much ? Is there any documentation
> > > available ?
> >
> > If programming speed is the only reason for your two-step approach,
> > then IMO this is not reasonable. It would be a waste of your time. Use
> > a faster JTAG programmer instead. Or you could buy the flashes
> > preprogrammed.
>
> Buying the flashes preprogrammed is not an option given the quantities.
>
> We already have a BDI2000 which has issues programming the flash using the
> workspace, so it reverts to the slow pure JTAG mode. I still have to
> investigate this directly with Abatron, we haven't found a solution so
> far.
>
> Even using the internal workspace, the BDI2000 is not a very practical
> solution, as it wouldn't be easy to integrate it in our test bench
> architecture. The price is also an issue. This is why I was looking for
> another solution. The alternative I thought of is using a pure JTAG
> programmer (slow) with a tiny ethernet-enabled boot loader. The Freescale
> JTAG commands used to control the processor are not documented, so I can't
> develop a faster JTAG-only alternative. Are you aware of any other
> solution ?
>
> Laurent Pinchart
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share
> your
> opinions on IT & business topics through brief surveys-and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> U-Boot-Users mailing list
> U-Boot-Users at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/u-boot-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.denx.de/pipermail/u-boot/attachments/20070302/59b8f4ce/attachment.htm 

[U-Boot-Users] Extremely tiny footprint

[Martin Krause]

Hi Laurent,

Laurent Pinchart wrote on Friday, March 02, 2007 1:35 PM:
> > If programming speed is the only reason for your two-step approach,
> > then IMO this is not reasonable. It would be a waste of your time.
> > Use a faster JTAG programmer instead. Or you could buy the flashes
> > preprogrammed.
> 
> Buying the flashes preprogrammed is not an option given the
> quantities.

OK. But it seems, that the quantities are high enough, that a minute
ore two more programming time does matter :-)

> We already have a BDI2000 which has issues programming the flash
> using the workspace, so it reverts to the slow pure JTAG mode. I
> still have to investigate this directly with Abatron, we haven't
> found a solution so far.

I know workspace problems from our projects. Nearly always a wrong
board (SDRAM) settup was the reason. The Abatron support is very
helpful, though.

> Even using the internal workspace, the BDI2000 is not a very practical
> solution, as it wouldn't be easy to integrate it in our test bench
> architecture. The price is also an issue. This is why I was looking
> for another solution. The alternative I thought of is using a pure
> JTAG programmer (slow) with a tiny ethernet-enabled boot loader. The
> Freescale JTAG commands used to control the processor are not
> documented, so I can't develop a faster JTAG-only alternative. Are
> you aware of any other solution ? 

Possibly it makes sense to program the Flashes before soldering (with
a chip programmer)? This shouldn't be faster as a BDI2000 with a 
working) workspace configuration, but maybe cheaper?

Sorry, don't have other ideas.

Regards,
Martin

[U-Boot-Users] Extremely tiny footprint

[Wolfgang Denk]

In message <200703021334.31503.laurent.pinchart@technotrade.biz> you wrote:
> 
> We already have a BDI2000 which has issues programming the flash using the 
> workspace, so it reverts to the slow pure JTAG mode. I still have to 

That means you have not set up your board config file correctly.

Instead of wasting your time on a two-level bootloader approach,  fix
the config file instead.

> Even using the internal workspace, the BDI2000 is not a very practical 
> solution, as it wouldn't be easy to integrate it in our test bench 

The BDI2000 is perfectly scriptable with  expect,  so  what  are  you
missing?


Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH, HRB 165235 Munich, CEO: Wolfgang Denk
Office:  Kirchenstr. 5,       D-82194 Groebenzell,            Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
The IQ of the group is the lowest IQ of a member of the group divided
by the number of people in the group.

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

> > We already have a BDI2000 which has issues programming the flash using
> > the workspace, so it reverts to the slow pure JTAG mode. I still have to
>
> That means you have not set up your board config file correctly.
>
> Instead of wasting your time on a two-level bootloader approach,  fix
> the config file instead.

That's what I've been trying to do for the past 6 months ! Nobody have been 
able to help me, not even the Abatron distributor. I'm still trying to solve 
the issue, which involves sending hardware samples, NDAs, ... If I had known 
it would take so long for the support people to even work on my issue I would 
have gone for a two-level bootloader approach (or anything else that just 
works) right away.

> > Even using the internal workspace, the BDI2000 is not a very practical
> > solution, as it wouldn't be easy to integrate it in our test bench
>
> The BDI2000 is perfectly scriptable with  expect,  so  what  are  you
> missing?

The ability to make it fit in the bed of nails. It would involve designing a 
new mechanical test bed. We obviously can't integrate the BDI in the test bed 
PCB.

Laurent Pinchart

[U-Boot-Users] Extremely tiny footprint

[Wolfgang Denk]

In message <200703021629.19343.laurent.pinchart@technotrade.biz> you wrote:
>
> That's what I've been trying to do for the past 6 months ! Nobody have been 
> able to help me, not even the Abatron distributor. I'm still trying to solve 

I remember that this has been  discussed  before.  Don;t  talk  to  a
distributor,  talk  to Abatron support. Of course you are supposed to
have a thorough understanding of your own hardware.

Which CPU are we talking about, btw?

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH, HRB 165235 Munich, CEO: Wolfgang Denk
Office:  Kirchenstr. 5,       D-82194 Groebenzell,            Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Madness takes its toll. Please have exact change.

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

Hi Wolfgang,

> > That's what I've been trying to do for the past 6 months ! Nobody have
> > been able to help me, not even the Abatron distributor. I'm still trying
> > to solve
>
> I remember that this has been  discussed  before.  Don;t  talk  to  a
> distributor,  talk  to Abatron support.

I contacted them directly, and they redirected me to their French distributor. 
We are working through NDA details so that I can send them a hardware sample. 
I hope to get this resolved soon.

> Of course you are supposed to have a thorough understanding of your own
> hardware. 

Obviously :-)

> Which CPU are we talking about, btw?

MPC8248.

Best regards,

Laurent Pinchart

[U-Boot-Users] Extremely tiny footprint

[Andreas Schweigstill]

Dear Laurent!

Laurent Pinchart schrieb:
 >[...] not even the Abatron distributor.

You are talking to *sales people* when you have a technical problem???
Neither Abatron nor the BDI, JTAG, U-Boot or whatever is your problem...

I have to admit that I haven't worked with the BDI1000/2000 yet but I
know from many people that they use it very successfully.

 > I'm still trying to solve
> the issue, which involves sending hardware samples, NDAs, ...

Have you taken a look onto your JTAG signals? Are you using RTCK (on
ARM)? Have you configured it properly?

Have you configured the system and JTAG reset behaviour properly?

Did you know that an not properly terminated JTAG signals (especially
pull-up and pull-down resistors and reset connection) can cause LOTS of
problem in the field even when JTAG never seems to be used?

Do you use the same scan loop also for other devices on your board?
Are those devices' JTAG signals connected in parallel or correct serial
order? Did you configure such scan loop for your BDI?

And why don't you tell us which processor you are using?

>> The BDI2000 is perfectly scriptable with  expect,  so  what  are  you
>> missing?
> 
> The ability to make it fit in the bed of nails. It would involve designing a 
> new mechanical test bed. We obviously can't integrate the BDI in the test bed 
> PCB.

You could use another JTAG debugger, probably a much smaller device like
the Rowley CrossConnect Light.


With best regards
Andreas Schweigstill

-- 
Dipl.-Phys. Andreas Schweigstill
Schweigstill IT | Embedded Systems
Schauenburgerstra?e 116, D-24118 Kiel, Germany
Phone: (+49) 431 5606-435, Fax: (+49) 431 5606-436
Mobile: (+49) 171 6921973, Web: http://www.schweigstill.de/

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

Hi Andreas,

>  >[...] not even the Abatron distributor.
>
> You are talking to *sales people* when you have a technical problem???

The French distributor have a technical support service.

> Neither Abatron nor the BDI, JTAG, U-Boot or whatever is your problem...
>
> I have to admit that I haven't worked with the BDI1000/2000 yet but I
> know from many people that they use it very successfully.

I have used the BDI2000 successfully for several tasks, but it failed for 
flash programming.

> > I'm still trying to solve the issue, which involves sending hardware
> > samples, NDAs, ... 
>
> Have you taken a look onto your JTAG signals? Are you using RTCK (on
> ARM)? Have you configured it properly?
>
> Have you configured the system and JTAG reset behaviour properly?

I have (I'm using an MPC8248 processor by the way). JTAG works fine. Boundary 
scan works, debugging works, programming the flash memory through pure JTAG 
works. Only the fast flash programming mode fails.

> Did you know that an not properly terminated JTAG signals (especially
> pull-up and pull-down resistors and reset connection) can cause LOTS of
> problem in the field even when JTAG never seems to be used?
>
> Do you use the same scan loop also for other devices on your board?
> Are those devices' JTAG signals connected in parallel or correct serial
> order? Did you configure such scan loop for your BDI?

No other devices, just the MPC8248.

> And why don't you tell us which processor you are using?

I asked for help regarding the BDI200 problem in the past. This thread was 
initially about something completely different, and the processor type was 
irrelevant.

> > > The BDI2000 is perfectly scriptable with  expect,  so  what  are  you
> > > missing?
> >
> > The ability to make it fit in the bed of nails. It would involve
> > designing a new mechanical test bed. We obviously can't integrate the BDI
> > in the test bed PCB.
>
> You could use another JTAG debugger, probably a much smaller device like
> the Rowley CrossConnect Light.

As far as I know, it doesn't support the PowerPC architecture. Another problem 
is that our test bed currently has no USB connectivity.

Best regards,

Laurent Pinchart

[U-Boot-Users] Extremely tiny footprint

[Wolfgang Denk]

In message <200703021219.57763.laurent.pinchart@technotrade.biz> you wrote:
> 
> I would like to use U-Boot as a first-stage boot loader to program flash 
> memory on the target board. The idea is to perform a two-steps procedure:

That's not what U-Boot was designed for. You will run into problems.

> To minimize time spent in step one, I need to make the first-stage boot loader 
> as small as possible (hopefully less than 32kB). Has anyone ever stripped 
> U-Boot that much ? Is there any documentation available ?

Just use a real U-Boot. JTAG programming does not take so long. IMHO
you're wasting your time.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH, HRB 165235 Munich, CEO: Wolfgang Denk
Office:  Kirchenstr. 5,       D-82194 Groebenzell,            Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"There is nothing new under the sun, but there are lots of old things
we don't know yet."                                  - Ambrose Bierce

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

> > I would like to use U-Boot as a first-stage boot loader to program flash
> > memory on the target board. The idea is to perform a two-steps procedure:
>
> That's not what U-Boot was designed for. You will run into problems.
>
> > To minimize time spent in step one, I need to make the first-stage boot
> > loader as small as possible (hopefully less than 32kB). Has anyone ever
> > stripped U-Boot that much ? Is there any documentation available ?
>
> Just use a real U-Boot. JTAG programming does not take so long. IMHO
> you're wasting your time.

JTAG programming using the BDI2000 takes about 1s/kB. That's around 4 minutes 
for the real bootloader.

Laurent Pinchart

[U-Boot-Users] Extremely tiny footprint

[Andrew Dyer]

On 3/2/07, Laurent Pinchart <laurent.pinchart@technotrade.biz> wrote:
> > > I would like to use U-Boot as a first-stage boot loader to program flash
> > > memory on the target board. The idea is to perform a two-steps procedure:
> >
> > That's not what U-Boot was designed for. You will run into problems.
> >
> > > To minimize time spent in step one, I need to make the first-stage boot
> > > loader as small as possible (hopefully less than 32kB). Has anyone ever
> > > stripped U-Boot that much ? Is there any documentation available ?
> >
> > Just use a real U-Boot. JTAG programming does not take so long. IMHO
> > you're wasting your time.
>
> JTAG programming using the BDI2000 takes about 1s/kB. That's around 4 minutes
> for the real bootloader.

how fast does the bdi2000 go when accessing ram?  If it's fast enough
you could setup the ram via the bdi-2000 by doing register writes, and
use the bdi2000 to load a version of u-boot into ram, run it and then
use the now resident u-boot to load a copy of itself of something else
into flash.

-- 
Hardware, n.:
        The parts of a computer system that can be kicked.

[U-Boot-Users] Extremely tiny footprint

[Wolfgang Denk]

In message <200703021625.00237.laurent.pinchart@technotrade.biz> you wrote:
>
> > Just use a real U-Boot. JTAG programming does not take so long. IMHO
> > you're wasting your time.
> 
> JTAG programming using the BDI2000 takes about 1s/kB. That's around 4 minutes 
> for the real bootloader.

You must be doing something seriously wrong, then. 

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH, HRB 165235 Munich, CEO: Wolfgang Denk
Office:  Kirchenstr. 5,       D-82194 Groebenzell,            Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
A wise person makes his  own  decisions,  a  weak  one  obeys  public
opinion.                                           -- Chinese proverb

[U-Boot-Users] Extremely tiny footprint

[Laurent Pinchart]

> > > Just use a real U-Boot. JTAG programming does not take so long. IMHO
> > > you're wasting your time.
> >
> > JTAG programming using the BDI2000 takes about 1s/kB. That's around 4
> > minutes for the real bootloader.
>
> You must be doing something seriously wrong, then.

Remember I'm using pure JTAG mode (without workspace). Should I get faster 
programming in that mode (16bit data bus, MirrorBit flash) ?

Best regards,

Laurent Pinchart