[ECOS] Support Web Page? (MBX issue)

[ECOS] Support Web Page? (MBX issue)

[Jamie Guinan]

Hi all,

I'm supposed to ask this on the Cygnus support web page since
my current client has a support contract, but either 
  (a) I'm stupid
or 
  (b) the web pages at http://support.cygnus.com have changed
  enough recently that I can't navigate my way to the problem 
  submission form.
or 
  (c), both (a) and (b).


Failing my attempt to find the support form, I have a question
(I'll happily re-submit it to the problem support area if anyone
can point me to it).

I have a brand-new MBX board, and I am unable to get the GDB stubrom
image to work.  I tried both "powerpc-mbx/gdbload.bin" that came
with our supported distribution, and a "stubrom" image built according
to Jonathon Lamour's instructions,

 http://sourceware.cygnus.com/ml/ecos-discuss/1999-07/msg00023.html

I tftp'd the binary images to the target system using the EPPCBug
monitor, then used PFLASH to burn them into the 32-bit flash memory
(4 soldered chips).  EPPCBug is running from the 8-bit flash device
(1 socketed chip).

Anyway, when I switch J4 to boot from the 32-bit flash memory,
I get nothing.  I ran "powerpc-eabi-gdb -b 9600" and did a
"target remote /dev/ttyS0" but it failed to connect.  Browsing
the sources I expected to see a "+" sign emitted at boot time
but I saw nothing through my terminal emulator.

Any tricks to get the GDB stub to work on the MBX?

I tried the above several times, and even burned a copy of the 
EPPCBug monitor into the 32bit flash and that worked fine.

-Jamie

================================================================
Jamie Guinan                         Blue Button Solutions, Inc.
guinan@bluebutton.com                  http://www.bluebutton.com
================================================================

Re: [ECOS] Support Web Page? (MBX issue)

[Jamie Guinan]

On Thu, 2 Sep 1999, Jamie Guinan wrote:

> ... and a "stubrom" image built according
> to Jonathon Lamour's instructions,
> 
>  http://sourceware.cygnus.com/ml/ecos-discuss/1999-07/msg00023.html

Of course I used "--platform mbx" instead of "--platform cogent"
for the MBX.


================================================================
Jamie Guinan                         Blue Button Solutions, Inc.
guinan@bluebutton.com                  http://www.bluebutton.com
================================================================

Re: [ECOS] Support Web Page? (MBX issue)

[Hugo Tyson]

Jamie Guinan <guinan@bluebutton.com> writes:
> I have a brand-new MBX board, and I am unable to get the GDB stubrom
> image to work.  I tried both "powerpc-mbx/gdbload.bin" that came
> with our supported distribution, and a "stubrom" image built according
> to Jonathon Lamour's instructions,
> 
>  http://sourceware.cygnus.com/ml/ecos-discuss/1999-07/msg00023.html
> 
> I tftp'd the binary images to the target system using the EPPCBug
> monitor, then used PFLASH to burn them into the 32-bit flash memory
> (4 soldered chips).  EPPCBug is running from the 8-bit flash device
> (1 socketed chip).
> 
> Anyway, when I switch J4 to boot from the 32-bit flash memory,
> I get nothing.  I ran "powerpc-eabi-gdb -b 9600" and did a
> "target remote /dev/ttyS0" but it failed to connect.  Browsing
> the sources I expected to see a "+" sign emitted at boot time
> but I saw nothing through my terminal emulator.
> 
> Any tricks to get the GDB stub to work on the MBX?
> 
> I tried the above several times, and even burned a copy of the 
> EPPCBug monitor into the 32bit flash and that worked fine.

Sanity check: whichever of "ROM" or flash is selected for boot lives at
0xfe000000 and the other lives at 0xfc000000 (where ROM means the socketed
8-bit flash or PROM device), so the booting code shouldn't know any
difference between running from the 32-bit flash and the "ROM" - as
demonstrated by the fact the PPCBug indeed runs thus.  Right?

However, all the testing we have done for ROM start was using the 8-bit
device; we retained the copy of PPCBug in the 32-bit flash just-in-case.
That's what (I believe) we document: that's what we support.

So, to get started, acquire an AM29F040 (flash) or AM27F040 (PROM) and load
it with the stubrom image using a PROM programmer, y'know, outside of the
board.  Plug it in, select that socket for booting and all should be well.


Here's a guess at why it doesn't work: we program the UPMs during bootup to
deal with an 8-bit device on CS0 and 32-bit flash on CS7 (mbx.S in the mbx
component) - unconditionally.  So if it's trying to boot from flash
(ie. CS0 is the 32-bit device), the memory timing will be set up quite
wrongly (as soon as MM is enabled).

I guess there's some way to read the links through a status register and
find out whether you're starting from 32-bit space and swap these setups -
or some kludge involving just trying it - that must be in the PPCBug image.
A quick scan through the MBX doc reveals no such register. ;-(

If you want to try the experiment, and make a stubrom image that will ONLY
run from 32-bit flash, try swapping the setups for the BOOT ROM and FLASH
in the setup code in mbx.S: relevant fragments are below unchanged, so you
can find them unambiguously.  Ie. put the values for BOOT ROM in BR7,OR7
and the values for FLASH in BR0,OR0.  Does that make sense?

HTH,
	- Huge

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
	/*
	 * Base Register initialization.
	 */
	/* BOOT ROM */
        lwi	r3,0xfe000401	# 8-bit, GPCM
#ifdef __50MHZ
        lwi	r5,0xff800960 /* for 120nS devices, else 0xff800940 */
#else
        lwi	r5,0xff800930
#endif
        stw	r3,BR0(r4)
        stw	r5,OR0(r4)

.....

	/* FLASH */
        lwi	r3,0xfc000001	# 32-bit, GPCM
#ifdef __50MHZ
        lwi	r5,0xff800940
#else
        lwi	r5,0xff800930
#endif
        stw	r3,BR7(r4)
        stw	r5,OR7(r4)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

[ECOS] Re: Support Web Page? (MBX issue)

[Bart Veer]

>>>>> "Jamie" == Jamie Guinan <guinan@bluebutton.com> writes:

    Jamie> I'm supposed to ask this on the Cygnus support web page since
    Jamie> my current client has a support contract, but either 
    Jamie>   (a) I'm stupid
    Jamie> or 
    Jamie>   (b) the web pages at http://support.cygnus.com have changed
    Jamie>   enough recently that I can't navigate my way to the problem 
    Jamie>   submission form.
    Jamie> or 
    Jamie>   (c), both (a) and (b).

Thanks for reporting this. Your client will have been supplied with
full information on how to use the support web pages. I have asked the
appropriate person within Cygnus to follow up on this issue and
provide any assistance that may be required.

Bart Veer // eCos net maintainer