Timur Tabi wrote:
<improvements/fixes to patch and patch format>
Thank you, stay tuned for another attempt at the patch.
I'll test this code on my ITX, however, I'm curious about one thing. Can you explain why this patch is okay for *all* ITX boards? I remember something about your board having problematic DDR or something. With your patch, all ITX and ITX-GP boards will set sdram_clk_cntl to the new value when SPD is used.
I am not aware of my board having a problematic DDR. Perhaps this is the proposed fix to the issue you are thinking of? I think the big question is how many people using the 8349ITX are using git sources. Both Bruce and I experienced instability without this patch, to the point the system could not boot to console. After applying this patch, neither of us have witnessed any instability after banging on the system(*).
As I mentioned in a previous email, for the MPC8349ITX I am _reverting_ the configured value for the sdram_clk_cntl to what Freescale shipped in their patched u-boot 1.1.3 and to the value in the mpc83xx git sources prior to October 25, 2006. In fact, I think that a simliar patch might need to be applied for the MPC832XEMDS for exactly the same reasons. I think it would be helpful if you could look at the f6eda7f80ccc13d658020268c507d7173cf2e8aa commit to cpu/mpc83xx/spd_sdram.c and how the sdram_clk_cntl register was initialized before/after this commit. From that, you should (hopefully) either be able to confirm the proposed patch is doing the right thing or tell me that I am not understanding something. Thank you, Michael
* - I lied, I do have some issues with the serial layer that I am sorting through. However, I think this is unrelated. The instability prior to this patch was much wider in scope.