Magnus,
--- On Wed, 10/22/08, Magnus Lilja lilja.magnus@gmail.com wrote:
From: Magnus Lilja lilja.magnus@gmail.com Subject: Re: [U-Boot] [PATCH 3/3] iMX31: Add support to iMX31PDK board boots from NAND Flash To: "Wolfgang Denk" wd@denx.de Cc: u-boot@lists.denx.de, "Alan Carvalho de Assis" alan.assis@freescale.com Date: Wednesday, October 22, 2008, 4:52 PM 2008/10/22 Magnus Lilja lilja.magnus@gmail.com:
Wolfgang,
2008/10/22 Wolfgang Denk wd@denx.de:
Dear Magnus,
In message
59b21cf20810221148u7afd8e19h148b496234b79274@mail.gmail.com you wrote:
I've updated the patch and it now almost
compiles. There's a problem
at the last link stage where gnu-ld complains
about multiple __udivsi3
definitions and incompatible EABI formats (see
below, modulo gmail
wrapping).
Try this patch which I posted today: ARM: Use do_div() instead of division for
"long long".
That patch doesn't seem to touch any files that
are involved in
building i.MX31 boards. I might try the equivalent
change though.
Although it is a bit puzzling that I can build the
i.MX31 Litekit
board with the same toolchain.
Ok, I've found what's triggering this. The PDK board has CONFIG_MX31_CLK32=32768 while Litekit has CONIFG_MX31_CLK32=32000. That difference nowadays causes problems when linking, I suppose that happened after the merging of CONFIG_MX31_TIMER_HIGH_PRECISION.
Changing the TICK_TO_TIME macro in arm1176/mx31/interrupts.c to a construction that uses do_div() solves the problem.
I don't have the PDK schematics here but since I decided to set CLK32 to 32768 I suppose the board uses that frequency.
Yes, I confirmed with the schematics. There is a 32768kHz crystal connected to the PMIC that drives the MX31 clock.
Regards,
Fabio Estevam
I haven't tried my new patch on real hardware yet, but I'll do that tomorrow and post a patch (without the do_div() fix though, perhaps someone else can do that?).
/Magnus _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot