| Message ID | 1532614438-7207-1-git-send-email-narmstrong@baylibre.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | arm64: dts: meson: Switch simple-mfd and syscon order | expand |
HI Neil On 07/26/2018 10:13 PM, Neil Armstrong wrote: > The order between "syscon" and "simple-mfd" is important because in these > particular cases, the node needs to be first a "simple-mfd" to expose > it's sub-nodes, and later on a "syscon" to permit other nodes to access > this register space through the "syscon" mechanism. > could you explain a bit which specific case that it need compatible "simple-mfd" come first? is this a problem in kernel side? because I see you have a comment below about problem in U-Boot.. also while Rob review the mmc-clkc patch, he raise a suggestion to drop "syscon" in drivers/mfd/syscon.c, see [1], will that solve or affect your problem? [1] https://lkml.kernel.org/r/CAL_Jsq+r+7BaJAPBn3L26sdxATVyDiuC=V3xE+yz0-27q0RJhA@mail.gmail.com > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > --- > This also has a consequence for U-Boot since the U-Boot DM is much simpler > and can only associate a single device to a node, thus here we want these > nodes to be an mfd bus to be able to probe the sub-nodes. > > arch/arm64/boot/dts/amlogic/meson-axg.dtsi | 4 ++-- > arch/arm64/boot/dts/amlogic/meson-gx.dtsi | 4 ++-- > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > index 6adc8d3..3f0de9d 100644 > --- a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > +++ b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > @@ -342,7 +342,7 @@ > ranges = <0x0 0x0 0x0 0xff63c000 0x0 0x1c00>; > > sysctrl: system-controller@0 { > - compatible = "amlogic,meson-axg-hhi-sysctrl", "syscon", "simple-mfd"; > + compatible = "amlogic,meson-axg-hhi-sysctrl", "simple-mfd", "syscon"; > reg = <0 0 0 0x400>; > > clkc: clock-controller { > @@ -1180,7 +1180,7 @@ > ranges = <0x0 0x0 0x0 0xff800000 0x0 0x100000>; > > sysctrl_AO: sys-ctrl@0 { > - compatible = "amlogic,meson-axg-ao-sysctrl", "syscon", "simple-mfd"; > + compatible = "amlogic,meson-axg-ao-sysctrl", "simple-mfd", "syscon"; > reg = <0x0 0x0 0x0 0x100>; > > clkc_AO: clock-controller { > diff --git a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi > index b8dc4db..d88699f 100644 > --- a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi > +++ b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi > @@ -344,7 +344,7 @@ > ranges = <0x0 0x0 0x0 0xc8100000 0x0 0x100000>; > > sysctrl_AO: sys-ctrl@0 { > - compatible = "amlogic,meson-gx-ao-sysctrl", "syscon", "simple-mfd"; > + compatible = "amlogic,meson-gx-ao-sysctrl", "simple-mfd", "syscon"; > reg = <0x0 0x0 0x0 0x100>; > > pwrc_vpu: power-controller-vpu { > @@ -431,7 +431,7 @@ > ranges = <0x0 0x0 0x0 0xc883c000 0x0 0x2000>; > > sysctrl: system-controller@0 { > - compatible = "amlogic,meson-gx-hhi-sysctrl", "syscon", "simple-mfd"; > + compatible = "amlogic,meson-gx-hhi-sysctrl", "simple-mfd", "syscon"; > reg = <0 0 0 0x400>; > }; > >
Hi, On 26/07/2018 16:41, Yixun Lan wrote: > HI Neil > > On 07/26/2018 10:13 PM, Neil Armstrong wrote: >> The order between "syscon" and "simple-mfd" is important because in these >> particular cases, the node needs to be first a "simple-mfd" to expose >> it's sub-nodes, and later on a "syscon" to permit other nodes to access >> this register space through the "syscon" mechanism. >> > > could you explain a bit which specific case that it need compatible > "simple-mfd" come first? is this a problem in kernel side? because I see > you have a comment below about problem in U-Boot.. It's not a problem for Linux, but Linux is just an implementation, the device tree should work on multiple OSs like U-Boot or BSD kernels. We should base ourselves on the bindings documentation and the Device tree specification to specify the HW. In this case, we need to switch the order to specify it correctly. > > > also while Rob review the mmc-clkc patch, he raise a suggestion to drop > "syscon" in drivers/mfd/syscon.c, see [1], will that solve or affect > your problem? Rob wants us to drop the "syscon" because he wants us to move all nodes depending of HHI as subnode, which is impossible for HDMI and the VPU power controller, they are only "syscon" clients, their own registers are not in HHI. Neil > > [1] > https://lkml.kernel.org/r/CAL_Jsq+r+7BaJAPBn3L26sdxATVyDiuC=V3xE+yz0-27q0RJhA@mail.gmail.com > >> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> >> --- >> This also has a consequence for U-Boot since the U-Boot DM is much simpler >> and can only associate a single device to a node, thus here we want these >> nodes to be an mfd bus to be able to probe the sub-nodes. >> >> arch/arm64/boot/dts/amlogic/meson-axg.dtsi | 4 ++-- >> arch/arm64/boot/dts/amlogic/meson-gx.dtsi | 4 ++-- >> 2 files changed, 4 insertions(+), 4 deletions(-) >> >> diff --git a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi >> index 6adc8d3..3f0de9d 100644 >> --- a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi >> +++ b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi >> @@ -342,7 +342,7 @@ >> ranges = <0x0 0x0 0x0 0xff63c000 0x0 0x1c00>; >> >> sysctrl: system-controller@0 { >> - compatible = "amlogic,meson-axg-hhi-sysctrl", "syscon", "simple-mfd"; >> + compatible = "amlogic,meson-axg-hhi-sysctrl", "simple-mfd", "syscon"; >> reg = <0 0 0 0x400>; >> >> clkc: clock-controller { >> @@ -1180,7 +1180,7 @@ >> ranges = <0x0 0x0 0x0 0xff800000 0x0 0x100000>; >> >> sysctrl_AO: sys-ctrl@0 { >> - compatible = "amlogic,meson-axg-ao-sysctrl", "syscon", "simple-mfd"; >> + compatible = "amlogic,meson-axg-ao-sysctrl", "simple-mfd", "syscon"; >> reg = <0x0 0x0 0x0 0x100>; >> >> clkc_AO: clock-controller { >> diff --git a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi >> index b8dc4db..d88699f 100644 >> --- a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi >> +++ b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi >> @@ -344,7 +344,7 @@ >> ranges = <0x0 0x0 0x0 0xc8100000 0x0 0x100000>; >> >> sysctrl_AO: sys-ctrl@0 { >> - compatible = "amlogic,meson-gx-ao-sysctrl", "syscon", "simple-mfd"; >> + compatible = "amlogic,meson-gx-ao-sysctrl", "simple-mfd", "syscon"; >> reg = <0x0 0x0 0x0 0x100>; >> >> pwrc_vpu: power-controller-vpu { >> @@ -431,7 +431,7 @@ >> ranges = <0x0 0x0 0x0 0xc883c000 0x0 0x2000>; >> >> sysctrl: system-controller@0 { >> - compatible = "amlogic,meson-gx-hhi-sysctrl", "syscon", "simple-mfd"; >> + compatible = "amlogic,meson-gx-hhi-sysctrl", "simple-mfd", "syscon"; >> reg = <0 0 0 0x400>; >> }; >> >> >
On Thu, 26 Jul 2018, Neil Armstrong wrote: > The order between "syscon" and "simple-mfd" is important because in these > particular cases, the node needs to be first a "simple-mfd" to expose > it's sub-nodes, and later on a "syscon" to permit other nodes to access > this register space through the "syscon" mechanism. > > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > --- > This also has a consequence for U-Boot since the U-Boot DM is much simpler > and can only associate a single device to a node, thus here we want these > nodes to be an mfd bus to be able to probe the sub-nodes. > > arch/arm64/boot/dts/amlogic/meson-axg.dtsi | 4 ++-- > arch/arm64/boot/dts/amlogic/meson-gx.dtsi | 4 ++-- > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > index 6adc8d3..3f0de9d 100644 > --- a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > +++ b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > @@ -342,7 +342,7 @@ > ranges = <0x0 0x0 0x0 0xff63c000 0x0 0x1c00>; > > sysctrl: system-controller@0 { > - compatible = "amlogic,meson-axg-hhi-sysctrl", "syscon", "simple-mfd"; > + compatible = "amlogic,meson-axg-hhi-sysctrl", "simple-mfd", "syscon"; I must profess, I haven't looked at the code recently so can't really comment on the semantics - I'll leave that to Rob et. al. However, if the ordering is important, you should add a comment to prevent anyone re-jigging it or reverting the patch for some reason.
Neil Armstrong <narmstrong@baylibre.com> writes: > The order between "syscon" and "simple-mfd" is important because in these > particular cases, the node needs to be first a "simple-mfd" to expose > it's sub-nodes, and later on a "syscon" to permit other nodes to access > this register space through the "syscon" mechanism. > > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Applied to v4.20/dt64, Kevin
On Thu, Jul 26, 2018 at 10:51 PM Neil Armstrong <narmstrong@baylibre.com> wrote: > > Hi, > > On 26/07/2018 16:41, Yixun Lan wrote: > > HI Neil > > > > On 07/26/2018 10:13 PM, Neil Armstrong wrote: > >> The order between "syscon" and "simple-mfd" is important because in these > >> particular cases, the node needs to be first a "simple-mfd" to expose > >> it's sub-nodes, and later on a "syscon" to permit other nodes to access > >> this register space through the "syscon" mechanism. > >> > > > > could you explain a bit which specific case that it need compatible > > "simple-mfd" come first? is this a problem in kernel side? because I see > > you have a comment below about problem in U-Boot.. > > It's not a problem for Linux, but Linux is just an implementation, the device tree > should work on multiple OSs like U-Boot or BSD kernels. We should base ourselves on > the bindings documentation and the Device tree specification to specify the HW. > > In this case, we need to switch the order to specify it correctly. syscon on Linux works because syscon is not a full blown device driver in the driver model. It is just a set of APIs wrapping regmap. > > > > > > also while Rob review the mmc-clkc patch, he raise a suggestion to drop > > "syscon" in drivers/mfd/syscon.c, see [1], will that solve or affect > > your problem? > > Rob wants us to drop the "syscon" because he wants us to move all nodes > depending of HHI as subnode, which is impossible for HDMI and the VPU power controller, > they are only "syscon" clients, their own registers are not in HHI. You can have whatever device driver this is bound to export regmaps for clients to consume. This gives you finer control over what registers can be accessed. See the following example. Provider: https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/clk/sunxi-ng/ccu-sun8i-r40.c#L1315 Client: https://elixir.bootlin.com/linux/v4.19-rc3/source/drivers/net/ethernet/stmicro/stmmac/dwmac-sun8i.c#L1035 Hope this helps. ChenYu > Neil > > > > > [1] > > https://lkml.kernel.org/r/CAL_Jsq+r+7BaJAPBn3L26sdxATVyDiuC=V3xE+yz0-27q0RJhA@mail.gmail.com > > > >> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > >> --- > >> This also has a consequence for U-Boot since the U-Boot DM is much simpler > >> and can only associate a single device to a node, thus here we want these > >> nodes to be an mfd bus to be able to probe the sub-nodes. > >> > >> arch/arm64/boot/dts/amlogic/meson-axg.dtsi | 4 ++-- > >> arch/arm64/boot/dts/amlogic/meson-gx.dtsi | 4 ++-- > >> 2 files changed, 4 insertions(+), 4 deletions(-) > >> > >> diff --git a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > >> index 6adc8d3..3f0de9d 100644 > >> --- a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > >> +++ b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi > >> @@ -342,7 +342,7 @@ > >> ranges = <0x0 0x0 0x0 0xff63c000 0x0 0x1c00>; > >> > >> sysctrl: system-controller@0 { > >> - compatible = "amlogic,meson-axg-hhi-sysctrl", "syscon", "simple-mfd"; > >> + compatible = "amlogic,meson-axg-hhi-sysctrl", "simple-mfd", "syscon"; > >> reg = <0 0 0 0x400>; > >> > >> clkc: clock-controller { > >> @@ -1180,7 +1180,7 @@ > >> ranges = <0x0 0x0 0x0 0xff800000 0x0 0x100000>; > >> > >> sysctrl_AO: sys-ctrl@0 { > >> - compatible = "amlogic,meson-axg-ao-sysctrl", "syscon", "simple-mfd"; > >> + compatible = "amlogic,meson-axg-ao-sysctrl", "simple-mfd", "syscon"; > >> reg = <0x0 0x0 0x0 0x100>; > >> > >> clkc_AO: clock-controller { > >> diff --git a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi > >> index b8dc4db..d88699f 100644 > >> --- a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi > >> +++ b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi > >> @@ -344,7 +344,7 @@ > >> ranges = <0x0 0x0 0x0 0xc8100000 0x0 0x100000>; > >> > >> sysctrl_AO: sys-ctrl@0 { > >> - compatible = "amlogic,meson-gx-ao-sysctrl", "syscon", "simple-mfd"; > >> + compatible = "amlogic,meson-gx-ao-sysctrl", "simple-mfd", "syscon"; > >> reg = <0x0 0x0 0x0 0x100>; > >> > >> pwrc_vpu: power-controller-vpu { > >> @@ -431,7 +431,7 @@ > >> ranges = <0x0 0x0 0x0 0xc883c000 0x0 0x2000>; > >> > >> sysctrl: system-controller@0 { > >> - compatible = "amlogic,meson-gx-hhi-sysctrl", "syscon", "simple-mfd"; > >> + compatible = "amlogic,meson-gx-hhi-sysctrl", "simple-mfd", "syscon"; > >> reg = <0 0 0 0x400>; > >> }; > >> > >> > > > > > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
On Wed, Sep 12, 2018 at 8:59 PM Kevin Hilman <khilman@baylibre.com> wrote: > > Neil Armstrong <narmstrong@baylibre.com> writes: > > > The order between "syscon" and "simple-mfd" is important because in these > > particular cases, the node needs to be first a "simple-mfd" to expose > > it's sub-nodes, and later on a "syscon" to permit other nodes to access > > this register space through the "syscon" mechanism. > > > > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > > Applied to v4.20/dt64, > I spoke too soon, this doesn't apply cleanly on top of Jeromes reorder/cleanup. Please respin/resend. Kevin
Hi Kevin, On 13/09/2018 06:20, Kevin Hilman wrote: > On Wed, Sep 12, 2018 at 8:59 PM Kevin Hilman <khilman@baylibre.com> wrote: >> >> Neil Armstrong <narmstrong@baylibre.com> writes: >> >>> The order between "syscon" and "simple-mfd" is important because in these >>> particular cases, the node needs to be first a "simple-mfd" to expose >>> it's sub-nodes, and later on a "syscon" to permit other nodes to access >>> this register space through the "syscon" mechanism. >>> >>> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> >> >> Applied to v4.20/dt64, >> > > I spoke too soon, this doesn't apply cleanly on top of Jeromes > reorder/cleanup. Please respin/resend. I'll respin. Neil > > Kevin >
diff --git a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi index 6adc8d3..3f0de9d 100644 --- a/arch/arm64/boot/dts/amlogic/meson-axg.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-axg.dtsi @@ -342,7 +342,7 @@ ranges = <0x0 0x0 0x0 0xff63c000 0x0 0x1c00>; sysctrl: system-controller@0 { - compatible = "amlogic,meson-axg-hhi-sysctrl", "syscon", "simple-mfd"; + compatible = "amlogic,meson-axg-hhi-sysctrl", "simple-mfd", "syscon"; reg = <0 0 0 0x400>; clkc: clock-controller { @@ -1180,7 +1180,7 @@ ranges = <0x0 0x0 0x0 0xff800000 0x0 0x100000>; sysctrl_AO: sys-ctrl@0 { - compatible = "amlogic,meson-axg-ao-sysctrl", "syscon", "simple-mfd"; + compatible = "amlogic,meson-axg-ao-sysctrl", "simple-mfd", "syscon"; reg = <0x0 0x0 0x0 0x100>; clkc_AO: clock-controller { diff --git a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi index b8dc4db..d88699f 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gx.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gx.dtsi @@ -344,7 +344,7 @@ ranges = <0x0 0x0 0x0 0xc8100000 0x0 0x100000>; sysctrl_AO: sys-ctrl@0 { - compatible = "amlogic,meson-gx-ao-sysctrl", "syscon", "simple-mfd"; + compatible = "amlogic,meson-gx-ao-sysctrl", "simple-mfd", "syscon"; reg = <0x0 0x0 0x0 0x100>; pwrc_vpu: power-controller-vpu { @@ -431,7 +431,7 @@ ranges = <0x0 0x0 0x0 0xc883c000 0x0 0x2000>; sysctrl: system-controller@0 { - compatible = "amlogic,meson-gx-hhi-sysctrl", "syscon", "simple-mfd"; + compatible = "amlogic,meson-gx-hhi-sysctrl", "simple-mfd", "syscon"; reg = <0 0 0 0x400>; };
The order between "syscon" and "simple-mfd" is important because in these particular cases, the node needs to be first a "simple-mfd" to expose it's sub-nodes, and later on a "syscon" to permit other nodes to access this register space through the "syscon" mechanism. Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> --- This also has a consequence for U-Boot since the U-Boot DM is much simpler and can only associate a single device to a node, thus here we want these nodes to be an mfd bus to be able to probe the sub-nodes. arch/arm64/boot/dts/amlogic/meson-axg.dtsi | 4 ++-- arch/arm64/boot/dts/amlogic/meson-gx.dtsi | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-)