From mboxrd@z Thu Jan 1 00:00:00 1970 From: tony@atomide.com (Tony Lindgren) Date: Thu, 2 Mar 2017 10:56:07 -0800 Subject: [PATCH DO NOT MERGE 5/6] ARM: dts: omap4: add some sample clkctrl data In-Reply-To: References: <1486992157-10673-1-git-send-email-t-kristo@ti.com> <1486992157-10673-6-git-send-email-t-kristo@ti.com> <20170302174503.GP20572@atomide.com> Message-ID: <20170302185607.GS20572@atomide.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org * Tero Kristo [170302 10:45]: > On 02/03/17 19:45, Tony Lindgren wrote: > > Hi, > > > > * Tero Kristo [170213 05:24]: > > > Adds clkctrl nodes for cm_l4per and cm_abe as example. Peripherals using > > > clocks from these nodes are modified also. > > > > Finally got around testing these. Looks like applying this patch > > breaks things for devices not using the new clocks entry? > > > > For example, SPI now breaks causing "imprecise external abort" during boot > > at least on droid 4. > > If SPI is under l4per, then yes, a breakage is expected. This will cause a > conflict with the existing hwmod data, and the new clock data, effectively > disabling the IP clocks during boot. This patch only converts part of the DT > data to the new format, and as such is only suitable for testing purposes. > > I will provide a full data conversion for the DT file for 4.11-rc. OK > > > --- a/arch/arm/boot/dts/omap4.dtsi > > > +++ b/arch/arm/boot/dts/omap4.dtsi > > > + cm_l4per: cm_l4per at 0 { > > > > Above should be cm_l4per: cm_l4per at 1400, right? > > Yea thats a bug in this test patch. Same issue with other nodes. DT compiler > isn't too picky about these so it works even with wrong node name. > > > > > > + compatible = "ti,omap4-cm"; > > > + reg = <0x1400 0x200>; > > > + #address-cells = <1>; > > > + #size-cells = <1>; > > > + ranges = <0 0x1400 0x200>; > > > + > > > + cm_l4per_clkctrl: cm_l4per_clkctrl at 20 { > > > + compatible = "ti,omap4-clkctrl"; > > > + reg = <0x20 0x1b0>; > > > + #clock-cells = <2>; > > > + }; > > > + }; > > > > You should update the binding doc accordingly if the "cm_l4per at 0" node > > there is not needed. I also noticed the binding doc still has > > "#clock-cells = <4>" while it should be 2. > > cm_l4per is somewhat redundant right now, but we want to add clockdomains > under that one in the future. Or, we could just add those directly under > parent node also (cm2 in this case.) OK so probably best to add it to avoid tweaking the dts files again later on. Regards, Tony