On Thursday 05 February 2009, Kevin Hilman wrote: > The /proc dump only has two types of clocks: PLL or PSC, and the > ref_clk technically is neither. You'll notice I removed the 'CLK_PLL' > flag from ref_clk because of the new init and handling of PLL-derived > clocks.
Ergo my suggestion to label only the PSC clocks in /proc ... it was a bit of a goof to define a CLK_PLL flag instead of a CLK_PSC. :( > NULL for non-PSC clocks would be ok, but the ARMCLK doesn't have an > LPSC associated with it either, and is not PLL derived. Huh? LPSC #31 on dm355 (says spru3fb table 7-1), and LPSC #0 on dm6467 (says sprs403d table 4-3). You may be right about dm6446 not having a documented LPSC; but it's certainly PLL-derived there too. > The other related thing left to fix is the fact that clk_enable calls > psc_init on !CLK_PLL clocks, and it probably should only be called for > clocks with valid lpsc field. i.e. some !PLL clocks like ARMCLK no not > actually have PSC, at least on some devices. Again that'd be easier if I had created a CLK_PSC not a CLK_PLL. :( On the other hand, having only two types of clock is clearly not quite right. We have Oscillator, PLL (and derived), and LPSC gates. But we don't yet cover things like the audio clock generator on dm6446evm, which is an external clock. Or even the non-LPSC gating for CLKOUTx on various chips (mux as GPIO vs mux as CLKOUT). - Dave _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
