Hey guys,

Well I bricked a $15k board today :)

The Flash on this ARM926EJ-S core board is 2 512Mbit NOR Spansion
S29GL512N-11F chips arranged 32M x 32bit.

I was reading & writing to a blank sector just to test out my configs and I
noticed that when I did any fill command (b,h,w) to flash, when I read back
I didn't get the same results.

Here is what a flash fillb 0x20000000 0xaa 128 looks like:
0x20030000: ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa
ffffaaaa
0x20030020: ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa
ffffaaaa
0x20030040: ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa
ffffaaaa
0x20030060: ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa ffffaaaa
ffffaaaa

Uhhh, that don't look right!

I get the same thing if I use fillh or fillw.

My cfi config is:
flash bank cfi  0x20000000 0x04000000 4 4 0 (length was 0x08000000 but flash
probe said it was 0x04000000).  flinfo in uboot (when it was working)
displayed from 0x20000000 to three sectors shy of 0x27ffffff so don't know
if I believe the 0x04000000.

But that wasn't the bad news ... when I rebooted ... no u-boot.  Displaying
memory where uboot "was" showed that half the words of the address where it
lived were filled with 0xffff so something went crazy and hosed my flash in
areas I wasn't even messing with.

I have three versions from SVN I tried, 1411, 1470M and I can't remember the
other one.  They pretty much did similiar things.  I'm behind a firewall and
can't update until I leave work so I'll try that in case this is a result of
some of the recent churn.

Anyone seen anything like this before?

The good news is unlike my at91sam9260 board that I hosed last week, I can
still talk to the board so that is good!

Brain is fried ... looking for suggestions.  I'll go home and do a svn
update and pray that magically fixes it.

Regards,

Brian
_______________________________________________
Openocd-development mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to