[U-Boot] mx23-olinuxino: occasional memory errors

[Tsvetan Usunov, OLIMEX Ltd <usunov@olimex.com>]

Hi Kiril
something in uboot/kernel 3.x is broken with the memory setup
we also never succeed to build stabile image from 3.x
the only image which work rock solid is this one with kernel 2.6.x / elftosb
someone who is more familiar with kernel/uboot should take look at the 
memory/clock/plls initializations in elftosb and set them in exactly same 
manner in uboot/kernel 3.x
we used kernel 3.x image with Maxi on Embedded World to demonstrate how 
imx233 interface A13-LCD4.3" and MAXI was freezing at random time after 
playing video for 5-10 minutes on the LCD.
Same board works and plays video 24/7 with the old kernel image and elftosb

Tsvetan
----- Original Message ----- 
From: "Kiril Zyapkov" <kiril.zyapkov@gmail.com>
To: <u-boot@lists.denx.de>; "Support OLIMEX LTD" <support@olimex.com>; 
<contact@archlinuxarm.org>
Cc: "Robert Nelson" <robertcnelson@gmail.com>
Sent: Wednesday, May 15, 2013 12:49 PM
Subject: mx23-olinuxino: occasional memory errors


> Hi all,
>
> I was told I should report this.
>
> The tests are performed on two imx233-olinuxino boards: maxi and mini.
> I've removed R17 (120 ohms between CLK and CLKN) from the mini board
> as advised in #olimex and the Olimex forum board, but kept it on the
> maxi board.
>
> I am testing U-Boot 2013-04 with a patchset for mx23, build recipe is
> at [1] with a 3.7.2 kernel [2]. I ran `memtester 24M 100` 3 times on
> each board, the memtester program is built using [3] on a computer
> running ArchlinuxARM with armv5 architecture.
>
> The maxi board shows no signs of errors -- somewhat strange, given the
> popular opinion that R17 introduces instability and should be removed.
>
> The mini board, despite having R17 removed, shows occasional memory
> errors, here are some:
>
> ------------------------------------------------------------------------
> Loop 3/100:
>  Stuck Address       : testing   6FAILURE: possible bad address line
> at offset 0x000392a0.
> Skipping to next test...
> ...
> Loop 15/100:
>  Stuck Address       : testing   8FAILURE: possible bad address line
> at offset 0x010321b0.
> Skipping to next test...
> ...
> FAILURE: 0x7e0ba0a0 != 0x7e4ba0a0 at offset 0x0082f8e0.
>  Compare OR          : FAILURE: 0x4e032000 != 0x4e432000 at offset 
> 0x0088bf50.
> ...
>  8-bit Writes        : /FAILURE: 0xfe0f01b0 != 0xfe4f01b0 at offset 
> 0x007de9a4.
> ------------------------------------------------------------------------
>
> As advised by Marek, I've repeated the test with disabled pullup on
> MX23_PAD_EMI_CLKN__EMI_CLKN, changing the line in
> board/olimex/mx23_olinuxino/spl_boot.c to:
>       MX23_PAD_EMI_CLKN__EMI_CLKN | (MXS_PAD_3V3 | MXS_PAD_12MA |
> MXS_PAD_NOPULL),
> , but this doesn't fix the errors on the mini board.
>
> I've prepared a rootfs tarball which can be used to replicate the
> tests. Please see [4].
>
> Thanks,
>
> -- Kiril
>
> [1] 
> https://github.com/kzyapkov/PKGBUILDs/blob/master/alarm/uboot-olinuxino/PKGBUILD
> [2] 
> https://github.com/kzyapkov/PKGBUILDs/blob/master/core/linux-olinuxino/PKGBUILD
> [3] https://aur.archlinux.org/packages/memtester/
> [4] http://promptly.undo.it/alarm-memtester/