Hello Christian,
Christian Riesch wrote:
Hello Heiko, Thank you for your reply!
On Wed, Nov 9, 2011 at 6:37 AM, Heiko Schocher hs@denx.de wrote:
Hello Christian,
Christian Riesch wrote:
Signed-off-by: Christian Riesch christian.riesch@omicron.at
Hello Heiko,
On my board I cannot use your code in arch_cpu_init() in da850_lowlevel.c since I have different versions of my board with different input clock frequencies. Here u-boot should first determine the board revision number and then configure the SoC accordingly. Therefore I would like to move all board-specific parts (and PLL and memory configuration is board-specific since it depends on the memory chips and oscillators deployed on the board) to board_early_init_f which is called right after arch_cpu_init() and keep only a few initializiation steps in arch_cpu_init().
Ok, if you have other needs on your board we must change something. But I want to prevent, that all the code you remove in your patch is moved to board code, so all boards must (copy?) it. Instead you should move it to a weak function, which you can replace for your specific needs. (I thought the code was common enough ...)
What do you think?
That's fine for me, I tried it now with
__attribute__((weak)) int arch_cpu_init(void)
and added my own arch_cpu_init() to my board specific file.
Of course I had to add all those CONFIG_SYS_DA850_PINMUXn with dummy values to my board config file to make it build... I don't like those defines because the values are difficult to read and we already have
Why? They are in sync with the doc ...
code for pinmuxing in board/davinci/common/davinci_pinmux.c so it is duplicate code. Couldn't we move this davinci_pinmux.c to arch/arm/cpu/arm926ejs/davinci and use it instead? What do you think?
Hmm.. if you use the http://www-s.ti.com/sc/techlit/spraba2.zip utility from ti, you exactly get the values for the defines ... so I prefered to go this way, but have no real objections against using the code from board/davinci/common/davinci_pinmux.c (Of course, it should be moved to arch/arm/cpu/arm926ejs/davinci)
added Sandeep to cc. Sandeep, what do you think?
I'll submit a patch for adding the weak attribute.
Thanks!
bye, Heiko