| Message ID | 20200513130842.24847-4-lars.povlsen@microchip.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | power: Adding support for Microchip Sparx5 SoC | expand |
On Wed, May 13, 2020 at 03:08:40PM +0200, Lars Povlsen wrote: > This documents the 'microchip,reset-switch-core' property in the > ocelot-reset driver. > > Signed-off-by: Lars Povlsen <lars.povlsen@microchip.com> > --- > .../devicetree/bindings/power/reset/ocelot-reset.txt | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > index 4d530d8154848..20fff03753ad2 100644 > --- a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > +++ b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > @@ -9,9 +9,15 @@ microchip Sparx5 armv8 SoC's. > Required Properties: > - compatible: "mscc,ocelot-chip-reset" or "microchip,sparx5-chip-reset" > > +Optional properties: > +- microchip,reset-switch-core : Perform a switch core reset at the > + time of driver load. This is may be used to initialize the switch > + core to a known state (before other drivers are loaded). How do you know when other drivers are loaded? This could be a module perhaps. Doesn't seem like something that belongs in DT. Can this behavior be implied with "microchip,sparx5-chip-reset"? Rob
Rob Herring writes: > On Wed, May 13, 2020 at 03:08:40PM +0200, Lars Povlsen wrote: >> This documents the 'microchip,reset-switch-core' property in the >> ocelot-reset driver. >> >> Signed-off-by: Lars Povlsen <lars.povlsen@microchip.com> >> --- >> .../devicetree/bindings/power/reset/ocelot-reset.txt | 6 ++++++ >> 1 file changed, 6 insertions(+) >> >> diff --git a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt >> index 4d530d8154848..20fff03753ad2 100644 >> --- a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt >> +++ b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt >> @@ -9,9 +9,15 @@ microchip Sparx5 armv8 SoC's. >> Required Properties: >> - compatible: "mscc,ocelot-chip-reset" or "microchip,sparx5-chip-reset" >> >> +Optional properties: >> +- microchip,reset-switch-core : Perform a switch core reset at the >> + time of driver load. This is may be used to initialize the switch >> + core to a known state (before other drivers are loaded). > > How do you know when other drivers are loaded? This could be a module > perhaps. Doesn't seem like something that belongs in DT. > The reset driver is loaded at postcore_initcall() time, which ensures it is loaded before other drivers using the switch core. I noticed other drivers do the same to do low-level system reset and initialization at early boot time. > Can this behavior be implied with "microchip,sparx5-chip-reset"? Since we need to cater for both modus operandi, I would need two driver compatible strings per platform, which scales worse than a single property. The "microchip,reset-switch-core" is a device configuration property which tells the system (driver) how the hw should be handled. Since you do not *always* want to reset the switch core (f.ex. when implementing systems with warm reboot), I think it makes perfect sense - but I may be biased off course :-) Thank you for (all) of your comments, by the way! ---Lars > > Rob
Hi, On Tue, Jun 02, 2020 at 11:49:08AM +0200, Lars Povlsen wrote: > Rob Herring writes: > > On Wed, May 13, 2020 at 03:08:40PM +0200, Lars Povlsen wrote: > >> This documents the 'microchip,reset-switch-core' property in the > >> ocelot-reset driver. > >> > >> Signed-off-by: Lars Povlsen <lars.povlsen@microchip.com> > >> --- > >> .../devicetree/bindings/power/reset/ocelot-reset.txt | 6 ++++++ > >> 1 file changed, 6 insertions(+) > >> > >> diff --git a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > >> index 4d530d8154848..20fff03753ad2 100644 > >> --- a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > >> +++ b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > >> @@ -9,9 +9,15 @@ microchip Sparx5 armv8 SoC's. > >> Required Properties: > >> - compatible: "mscc,ocelot-chip-reset" or "microchip,sparx5-chip-reset" > >> > >> +Optional properties: > >> +- microchip,reset-switch-core : Perform a switch core reset at the > >> + time of driver load. This is may be used to initialize the switch > >> + core to a known state (before other drivers are loaded). > > > > How do you know when other drivers are loaded? This could be a module > > perhaps. Doesn't seem like something that belongs in DT. > > > > The reset driver is loaded at postcore_initcall() time, which ensures it > is loaded before other drivers using the switch core. I noticed other > drivers do the same to do low-level system reset and initialization at > early boot time. > > > Can this behavior be implied with "microchip,sparx5-chip-reset"? > > Since we need to cater for both modus operandi, I would need two driver > compatible strings per platform, which scales worse than a single > property. > > The "microchip,reset-switch-core" is a device configuration property > which tells the system (driver) how the hw should be handled. Since you > do not *always* want to reset the switch core (f.ex. when implementing > systems with warm reboot), I think it makes perfect sense - but I may be > biased off course :-) > > Thank you for (all) of your comments, by the way! > > ---Lars > > Rob Is this series still needed? Did I miss a follow-up? -- Sebastian
Sebastian Reichel writes: > Hi, > > On Tue, Jun 02, 2020 at 11:49:08AM +0200, Lars Povlsen wrote: > > Rob Herring writes: > > > On Wed, May 13, 2020 at 03:08:40PM +0200, Lars Povlsen wrote: > > >> This documents the 'microchip,reset-switch-core' property in the > > >> ocelot-reset driver. > > >> > > >> Signed-off-by: Lars Povlsen <lars.povlsen@microchip.com> > > >> --- > > >> .../devicetree/bindings/power/reset/ocelot-reset.txt | 6 ++++++ > > >> 1 file changed, 6 insertions(+) > > >> > > >> diff --git a/Documentation/devicetree/bindings/power/reset/ocelot-rese= > t.txt b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > > >> index 4d530d8154848..20fff03753ad2 100644 > > >> --- a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > > >> +++ b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt > > >> @@ -9,9 +9,15 @@ microchip Sparx5 armv8 SoC's. > > >> Required Properties: > > >> - compatible: "mscc,ocelot-chip-reset" or "microchip,sparx5-chip-res= > et" > > >> > > >> +Optional properties: > > >> +- microchip,reset-switch-core : Perform a switch core reset at the > > >> + time of driver load. This is may be used to initialize the switch > > >> + core to a known state (before other drivers are loaded). > > > > > > How do you know when other drivers are loaded? This could be a module > > > perhaps. Doesn't seem like something that belongs in DT. > > > > > > > The reset driver is loaded at postcore_initcall() time, which ensures it > > is loaded before other drivers using the switch core. I noticed other > > drivers do the same to do low-level system reset and initialization at > > early boot time. > > > > > Can this behavior be implied with "microchip,sparx5-chip-reset"? > > > > Since we need to cater for both modus operandi, I would need two driver > > compatible strings per platform, which scales worse than a single > > property. > > > > The "microchip,reset-switch-core" is a device configuration property > > which tells the system (driver) how the hw should be handled. Since you > > do not *always* want to reset the switch core (f.ex. when implementing > > systems with warm reboot), I think it makes perfect sense - but I may be > > biased off course :-) > > > > Thank you for (all) of your comments, by the way! > > > > ---Lars > > > Rob > > Is this series still needed? Did I miss a follow-up? Hi Sebastian! Yes, the series is still needed, but the conversation died after my last message. If the DT-controlled reset property is too controversial, I am willing to drop that part. (Rob just reviewed the bindings). MCHP reference designs have GPIO resets, so we *could* get by without, but new designs may this feature. > -- Sebastian ---Lars
diff --git a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt index 4d530d8154848..20fff03753ad2 100644 --- a/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt +++ b/Documentation/devicetree/bindings/power/reset/ocelot-reset.txt @@ -9,9 +9,15 @@ microchip Sparx5 armv8 SoC's. Required Properties: - compatible: "mscc,ocelot-chip-reset" or "microchip,sparx5-chip-reset" +Optional properties: +- microchip,reset-switch-core : Perform a switch core reset at the + time of driver load. This is may be used to initialize the switch + core to a known state (before other drivers are loaded). + Example: reset@1070008 { compatible = "mscc,ocelot-chip-reset"; reg = <0x1070008 0x4>; + microchip,reset-switch-core; };
This documents the 'microchip,reset-switch-core' property in the ocelot-reset driver. Signed-off-by: Lars Povlsen <lars.povlsen@microchip.com> --- .../devicetree/bindings/power/reset/ocelot-reset.txt | 6 ++++++ 1 file changed, 6 insertions(+) -- 2.26.2