Stefan Roese wrote:
On Friday 14 March 2008, Dave Littell wrote:
I'm working on an AMCC PPC440EPx platform that is similar to the AMCC Sequoia under U-Boot 1.3.1. I've copied the Sequoia board and configuration as a starting point, but I've run into a problem
with the
size of the flash-based portion of U-Boot. I've added code to initdram() that results in the addition of 3-4 dozen assembly instructions. Now the board hangs after a very few tftp (or even
ping)
commands.
Does it really hang, or do you get an exception?
Hi Stefan,
Thanks very much for the reply - I very much appreciate any help you can offer.
No exception appears. It hangs to the point where a JTAG debugger is unable to regain control of the processor.
However, if I remove the code I added, there's no problem with tftp, etc. I've narrowed it down to the point where I can
make the
difference between a working and non-working load by adding just a few instructions to initdram(). Some boundary or limit is being crossed somewhere...
I don't think so. It could be a problem with the DDR2 Denali setup
for your
board. What did you change in the DDR2 setup code? How similar is
your board
design to the Sequoia regarding DDR2 chips, termination etc?
From a hardware design perspective, much of the physical layout was
copied directly from the Sequoia (so I've been told), but it does use memory from a different manufacturer. The changes I added to initdram() were to use the DDR Controller register values suggested by the AMCC DDR Configuration Tool spreadsheet for two different PLB speeds. I felt the parameters were different enough to warrant using get_bus_speed() and a simple if() to determine which set of parameters were loaded to the DDR Controller.
If I instead use #ifdef's to determine at compile time which set of parameters are loaded, the board runs and I can tftp (or ping) all day. However, if I replace that with my if() logic and both sets of parameters (in the correct branch path - verified with the JTAG), the board will hang after only a few tftp's (or pings).
Hm.
I'm sure I've overrun something somewhere but I'm afraid I'm not literate enough in the magic of the .lds file (which is unmodified
from
the Sequoia platform) to understand what adding so few instructions to the flash-based portion of U-boot might have broken.
I've searched the FAQ and mailing list archives, all to no avail.
Does
anyone have any suggestions other that "write tighter code"? ;-)
Another idea comes to my mind here. It's on my todo list to implement a workaround for the 440EPx errata CHIP 11:
CHIP_11: End of memory range area restricted access. Category: 3 Overview: The 440EPx DDR controller does not acknowledge any transaction which is determined to be crossing over the end-of-memory-range boundary, even if the starting address is within valid memory space. Any such transaction from any PLB4 master will result in a PLB time-out on PLB4 bus. Impact: In case of such misaligned bursts, PLB4 masters will not retrieve any data at all, just the available data up to the end of memory, especially the 440 CPU. For example, if a CPU instruction required an operand located in memory within the last 7 words of memory, the DCU master would burst read 8 words to update the data cache and cross over the end-of-memory-range boundary. Such a DCU read would not be answered by the DDR controller, resulting in a PLB4 time-out and ultimately in a Machine Check interrupt. The data would be inaccessible to the CPU. Workaround: Forbid any application to access the last 256 bytes of DDR memory. For example, make your operating system believe that the last 256 bytes of DDR memory are absent. AMCC has a patch that does this, available for Linux.
Yes, I saw that and implemented the Linux kernel patch that AMCC provides. However, I didn't think that U-Boot ran up against the upper boundary of RAM.
One simple idea is to add :
#define CONFIG_PRAM 1
to your U-Boot config file.
I'll certainly give that a shot.
Thanks very much, Dave