[PATCH V3 04/10] clk: imx: Support building SCU clock driver as module
dongas86 at gmail.com
Mon Jun 29 23:36:40 EDT 2020
On Tue, Jun 30, 2020 at 3:36 AM Arnd Bergmann <arnd at arndb.de> wrote:
> On Mon, Jun 29, 2020 at 2:53 PM Anson Huang <anson.huang at nxp.com> wrote:
> > > Subject: Re: [PATCH V3 04/10] clk: imx: Support building SCU clock driver as
> > > module
> > >
> > > On Mon, Jun 29, 2020 at 8:06 AM Anson Huang <Anson.Huang at nxp.com>
> > > wrote:
> > >
> > > > --- a/drivers/clk/imx/Makefile
> > > > +++ b/drivers/clk/imx/Makefile
> > > > @@ -21,9 +21,9 @@ obj-$(CONFIG_MXC_CLK) += \
> > > > clk-sscg-pll.o \
> > > > clk-pll14xx.o
> > > >
> > > > -obj-$(CONFIG_MXC_CLK_SCU) += \
> > > > - clk-scu.o \
> > > > - clk-lpcg-scu.o
> > > > +mxc-clk-scu-objs += clk-lpcg-scu.o
> > > > +mxc-clk-scu-objs += clk-scu.o
> > > > +obj-$(CONFIG_MXC_CLK_SCU) += mxc-clk-scu.o
> > >
> > > It looks like the two modules are tightly connected, one is useless without the
> > > other. How about linking them into a combined module and dropping the
> > > export statement?
> > >
> > From HW perspective, the SCU clock driver and LPCG SCU clock driver are different,
> > SCU clock driver is for those clocks controlled by system controller (M4 which runs a firmware),
> > while LPCG SCU clock is for those clock gates inside module, which means AP core can
> > control it directly via register access, no need to via SCU API.
> Sorry, I misread the patch in multiple ways. First of all, you already put
> clk-scu.o and clk-lpcg-scu.o files into a combined loadable module, and
> I had only looked at clk-scu.c.
> What I actually meant here was to link clk-scu.o together with clk-imx8qxp.o
> (and possibly future chip-specific files) into a loadable module and drop
> the export.
It sounds like a good idea to me.
Actually I planned to combine them into one driver in the future.
> > So, I think it is NOT that tightly connected, it is because they are both for i.MX8 SoCs with SCU
> > inside, so they are put together in the Makefile.
> > If the export statement is acceptable, I think it is better to just keep it, make sense?
> There is nothing wrong with the export as such, this was just an
> idea to simplify the logic.
More information about the linux-arm-kernel