Messages in this thread | | | Date | Mon, 29 Jul 2013 18:44:03 +0200 | From | boris brezillon <> | Subject | Re: [PATCH v2 00/42] ARM: at91: move to common clk framework |
| |
On 29/07/2013 17:09, Richard Genoud wrote: > 2013/7/29 boris brezillon <b.brezillon@overkiz.com>: >> On 29/07/2013 16:18, Richard Genoud wrote: >>> 2013/7/29 boris brezillon <b.brezillon@overkiz.com>: >>>> On 29/07/2013 10:00, Richard Genoud wrote: >>>>> 2013/7/26 boris brezillon <b.brezillon@overkiz.com>: >>>>>> On 26/07/2013 17:44, boris brezillon wrote: >>>>>> I found 2 bugs: >>>>>> 1) the main frequency ready test in recalc_rate function is wrong >>>>>> 2) the common clk framework uses the first match for dt clk >>>>>> registration >>>>>> and main clk has this compatible property: >>>>>> compatible = "atmel,at91rm9200-clk-main", "fixed-clock"; >>>>>> As a result the main clk is always bound to the fixed-clock driver >>>>>> which set the clock frequency to 0 if 'clock-frequency' property >>>>>> is >>>>>> missing. >>>>>> >>>>>> These bugs will be fixed in next version. >>>>>> Tell me if you want need thesse patches for testing purpose. >>>>> Hi Boris, >>>>> >>>>> I tested your patch, the clock is not at 0 anymore: >>>>> AT91: PIT: mck rate = 8333333 >>>>> >>>>> And it's booting correctly. >>>>> There's a small bug I found in the at91sam9x5 dtsi file (ssc id is 28, >>>>> not 27). (patch attached) >>>>> >>>>> And yes, I could use your patches for some more tests ! >>>>> >>>>> Thanks, >>>>> >>>>> Richard. >>>> Hi Richard, >>>> >>>> Thanks for the bug report. >>>> >>>> You'll find in attachements the pre-v3 series fixing the detected bugs >>>> (including the ssc clock id bug). >>>> >>>> If you need anything else please let me know. >>> Working on sam9g35ek ! >>> you can add my: >>> Tested-by: Richard Genoud <richard.genoud@gmail.com> >>> >>> >>> One remark though: >>> Maybe it would be interesting to keep using the defines in the device >>> tree for peripheral IDs. >>> For instance, in arch/arm/boot/dts/at91sam9x5.dtsi : (patch 33) >>> ssc_clk@28 { >>> id = <28>; >>> }; >>> May be more readable with the SSC id AT91SAM9X5_ID_SSC, as it was >>> defined in arch/arm/mach-at91/at91sam9x5.c (removed with patch 20) >>> >>> IMHO, all AT91SAM9X5_ID_xxx in >>> arch/arm/mach-at91/include/mach/at91sam9x5.h can me moved in >>> include/dt-bindings/ and used in at91sam9x5.dtsi >> This was requested by Jean-Christophe too. >> I'll add these macro files and make use of it in next version. >> >> BTW where should I put these macro files >> (include/dt-bindings/at91/'soc-name'/peripherals.h) ? > That sounds correct I suppose. I guess I should do the same for system clks :-). >> Should I use macros for address fied of clocks or just for id property (or >> drop the address) ? >> >> ssc_clk@AT91SAM9X5_ID_SSC { >> id = <AT91SAM9X5_ID_SSC>; >> }; >> >> or >> >> ssc_clk@28 { >> id = <AT91SAM9X5_ID_SSC>; >> }; >> >> or >> >> ssc_clk { >> id = <AT91SAM9X5_ID_SSC>; >> }; > I would drop the address, since the names are always different (for > instance, dma0_clk and dma1_clk are used). > And the define should also be used here: > ssc0: ssc@f0010000 { [...] > clocks = <&periph AT91SAM9X5_ID_SSC>; > [...] > }
| |