MCF5475, Linux 2.6.25 & DDR > 256MB

[aeg]

Hi all,

I have a 5475 board with capability to add 1GB DDR via two 512MB modules. I've successfully used the entire range using CodeWarrior but when I tell the kernel that I have more than 256MB, it silently dies during boot up.

Has anyone else seen/tried this?

Best,
Andreas

[kmahan]

Hi Andreas,

The 547x/548x linux kernel has not been tested with that much memory.  I only have access to the LogicPD boards which don't have the ability to provide that much memory.

The kernel subsystem will have issues with more than 256M of memory due to how I allocated the ACRs.  The reasoning for this is to enhance the performance by minimizing the amount of TLBs used by the kernel.

What kind of board are you using?

--Kurt

[aeg]

Hi Wolfgang & Kurt.

I've verified that the DDR registers are not altered within the kernel so it is more likely an access issue somewhere. I will write a quick driver to see if I can access the memory at all.

Kurt,
I've rolled my own 5475 board with two DIMMS (512MB each). It is a mix of the LogicPD board and Verdi with a couple of other features I needed. Let me know if you're interested in more information about the board.

Finally, what is the latest status of the onboard USB?

Best,
Andreas

[kmahan]

Hi Andreas,

When you say "onboard BSP" what do you mean?  Are you talking about the USB on the silicon, or the ISP1561 attached via the PCI bus.  The ISP1561 is working in the 2.6.25 BSP subject to the limitations we document in the release notes.  Primarily that the LogicPD board design has an issue that causes instability.

I'm interested in the board you designed.  Will they be available?  I'd like to test the BSP on it.

--Kurt

[aeg]

Hi Kurt,

I should've said on-die, since I meant the silicon. Though, I have the ISP1561 and it works fine (except that the board house we're using seems to have a problem with it). I have the on-die brought out, but it is untested at this point.

The board itself is not available since it's part of another product but I'm willing to send you just the 5475 board for you testing as I think this is of value to the both of us.

Contact me on my email and we'll take it from there. (aengberg@linvatec.com)

Best,
Andreas

[aeg]

Solved.

-A

[w_wegner]

Hi Andreas,

how do you start the kernel, and how did you use the memory from CodeWarrior? Could it be the memory controller initialisation in the case of booting the kernel is still wrong?
(Depending on your setup, this should either be done in the kernel startup code or the bootloader, I guess - although I have to admit I only have experience with V3 devices (MCF532x, MCF537x) and uClinux.)

Best regards,
Wolfgang