Hi Giulio, On Thu, 9 Jan 2020 at 06:53, Giulio Benetti <giulio.bene...@benettiengineering.com> wrote: > > Hi Simon, > > On 1/8/20 6:39 PM, Simon Glass wrote: > > Hi Giulio, > > > > On Tue, 7 Jan 2020 at 10:23, Giulio Benetti > > <giulio.bene...@benettiengineering.com> wrote: > >> > >> Hi Stefano, Simon, All, > >> > >> On 1/3/20 12:39 PM, Stefano Babic wrote: > >>> Hi Giulio, > >>> > >>> On 04/12/19 18:44, Giulio Benetti wrote: > >>>> This patchset add support for i.MXRT family starting from i.MXRT1050 SoC. > >>>> It provides: > >>>> - i.MXRT1050 SoC entry > >>>> - i.MXRT pinctrl driver > >>>> - i.MXRT serial driver tweaking > >>>> - i.MXRT sdram controller driver > >>>> - i.MXRT usdhc driver tweaking > >>>> - i.MXRT1050-evk initial support > >>>> > >>>> It uses all DM clocks all around and it loads correctly a basic Linux > >>>> zImage. > >>>> > >>> > >>> Lukasz has already reviewed most patches, and a V2 is surely WIP. FYI: I > >>> archive this first series, and I'll wait for your next V2 - thanks ! > >> > >> I really would like to, but I'm stuck on OF_PLATDATA issue for named > >> clocks in SPL, see: > >> https://lists.denx.de/pipermail/u-boot/2019-December/394944.html > >> > >> Is it enough for the moment to keep SPL *without* OF_PLATDATA and > >> implement it later? > > > > It seems fine to me, but the assert() idea should work too. > > With assert() idea you mean define a header(i.e. imx-clk-names-id.h) like: > ``` > #define IMX_CLK_NAME_PER 0 > #define IMX_CLK_NAME_IPG 1 > ... > ``` > > And then in non-OF_PLATDATA trying to: > ``` > struct clk per_clk; > int ret; > > ret = clk_get_by_index(dev, IMX_CLK_NAME_PER, &clk); > ``` > > And check if clock's name is "per"?
Well actually the assert would only happen in the non-of-platdata code, which would use clk_get_by_name() and then assert that clk.id is IMX_CLK_NAME_PER, for example. That would provide a check that the of-platdata code is getting the right clock. You can't do anything in the of-platdata code since you don't have the names. BTW I added noisy_assert() recently which is active even without DEBUG defined, so that might be useful. Regards, Simon