| Message ID | 20190520032020.7920-1-peng.fan@nxp.com |
|---|---|
| State | Rejected, archived |
| Headers | show |
| Series | [RFC,1/2] dt-bindings: imx-ocotp: Add fusable-node property | expand |
| Context | Check | Description |
|---|---|---|
| robh/checkpatch | success |
On 20.05.2019 06:06, Peng Fan wrote: > Introduce fusable-node property for i.MX OCOTP driver. > The property will only be used by Firmware(eg. U-Boot) to > runtime disable the nodes. > > Take i.MX6ULL for example, there are several parts that only > have limited modules enabled controlled by OCOTP fuse. It is > not flexible to provide several dts for the serval parts, instead > we could provide one device tree and let Firmware to runtime disable > the device tree nodes for those modules that are disable(fused). > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > --- > > Currently NXP vendor use U-Boot to set status to disabled for devices > that could not function, > https://source.codeaurora.org/external/imx/uboot-imx/tree/arch/arm/mach-imx/mx6/module_fuse.c?h=imx_v2018.03_4.14.98_2.0.0_ga#n149 > But this approach is will not work if kernel dts node path changed. > > There are two approaches to resolve: > > 1. This patch is to add a fusable-node property, and Firmware will parse > the property and read fuse to decide whether to disable or keeep enable > the nodes. > > 2. There is another approach is that add nvmem-cells for all nodes that > could be disabled(fused). Then in each linux driver to use nvmem > api to detect fused or not, or in linux driver common code to check > device functionable or not with nvmem API. > > > To make it easy to work, we choose [1] here. Please advise whether > it is acceptable, because the property is not used by linux driver in > approach [1]. Or you prefer [2] or please advise if any better solution. Couldn't firmware parse nvmem-cells? Even without a full nvmem subsystem it would be possible for imx-specific code to walk the entire device tree, parse nvmem-cells and disable nodes which are disabled by fuse. -- Regards, Leonard
> Subject: Re: [RFC 1/2] dt-bindings: imx-ocotp: Add fusable-node property > > On 20.05.2019 06:06, Peng Fan wrote: > > Introduce fusable-node property for i.MX OCOTP driver. > > The property will only be used by Firmware(eg. U-Boot) to runtime > > disable the nodes. > > > > Take i.MX6ULL for example, there are several parts that only have > > limited modules enabled controlled by OCOTP fuse. It is not flexible > > to provide several dts for the serval parts, instead we could provide > > one device tree and let Firmware to runtime disable the device tree > > nodes for those modules that are disable(fused). > > > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > > --- > > > > Currently NXP vendor use U-Boot to set status to disabled for devices > > that could not function, > > > https://source.codeaurora.org/external/imx/uboot-imx/tree/arch/arm/mac > > h-imx/mx6/module_fuse.c?h=imx_v2018.03_4.14.98_2.0.0_ga#n149 > > But this approach is will not work if kernel dts node path changed. > > > > There are two approaches to resolve: > > > > 1. This patch is to add a fusable-node property, and Firmware will parse > > the property and read fuse to decide whether to disable or keeep > enable > > the nodes. > > > > 2. There is another approach is that add nvmem-cells for all nodes that > > could be disabled(fused). Then in each linux driver to use nvmem > > api to detect fused or not, or in linux driver common code to check > > device functionable or not with nvmem API. > > > > > > To make it easy to work, we choose [1] here. Please advise whether it > > is acceptable, because the property is not used by linux driver in > > approach [1]. Or you prefer [2] or please advise if any better solution. > > Couldn't firmware parse nvmem-cells? Even without a full nvmem subsystem > it would be possible for imx-specific code to walk the entire device tree, parse > nvmem-cells and disable nodes which are disabled by fuse. Firmware could parse nvmem-cells, but there are lots of modules that could be fused. If using nvmem-cells, that means need to add nvmem property for all the modules that could be fused. To make is looks cleaner, so introduced a fusable-node property. Regards, Peng. > > -- > Regards, > Leonard
Hi Rob, Srinivas > Subject: [RFC 1/2] dt-bindings: imx-ocotp: Add fusable-node property Do you have any comments about this patch? Thanks, Peng. > > Introduce fusable-node property for i.MX OCOTP driver. > The property will only be used by Firmware(eg. U-Boot) to runtime disable the > nodes. > > Take i.MX6ULL for example, there are several parts that only have limited > modules enabled controlled by OCOTP fuse. It is not flexible to provide several > dts for the serval parts, instead we could provide one device tree and let > Firmware to runtime disable the device tree nodes for those modules that are > disable(fused). > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > --- > > Currently NXP vendor use U-Boot to set status to disabled for devices that > could not function, > https://source.codeaurora.org/external/imx/uboot-imx/tree/arch/arm/mach > -imx/mx6/module_fuse.c?h=imx_v2018.03_4.14.98_2.0.0_ga#n149 > But this approach is will not work if kernel dts node path changed. > > There are two approaches to resolve: > > 1. This patch is to add a fusable-node property, and Firmware will parse > the property and read fuse to decide whether to disable or keeep enable > the nodes. > > 2. There is another approach is that add nvmem-cells for all nodes that > could be disabled(fused). Then in each linux driver to use nvmem > api to detect fused or not, or in linux driver common code to check > device functionable or not with nvmem API. > > > To make it easy to work, we choose [1] here. Please advise whether it is > acceptable, because the property is not used by linux driver in approach [1]. > Or you prefer [2] or please advise if any better solution. > > Thanks. > > Documentation/devicetree/bindings/nvmem/imx-ocotp.txt | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt > b/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt > index 7a999a135e56..e9a998588dbd 100644 > --- a/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt > +++ b/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt > @@ -21,6 +21,8 @@ Required properties: > > Optional properties: > - read-only: disable write access > +- fusable-node: array of phandles with reg base and bit offset, this > + property is used by Firmware to runtime disable nodes. > > Optional Child nodes: > > @@ -42,4 +44,7 @@ Example: > tempmon_temp_grade: temp-grade@20 { > reg = <0x20 4>; > }; > + > + fusable-node = <&usdhc1 0x10 4 > + &usdhc2 0x10 5>; > }; > -- > 2.16.4
On Mon, May 20, 2019 at 03:06:35AM +0000, Peng Fan wrote: > Introduce fusable-node property for i.MX OCOTP driver. > The property will only be used by Firmware(eg. U-Boot) to > runtime disable the nodes. > > Take i.MX6ULL for example, there are several parts that only > have limited modules enabled controlled by OCOTP fuse. It is > not flexible to provide several dts for the serval parts, instead > we could provide one device tree and let Firmware to runtime disable > the device tree nodes for those modules that are disable(fused). > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > --- > > Currently NXP vendor use U-Boot to set status to disabled for devices > that could not function, > https://source.codeaurora.org/external/imx/uboot-imx/tree/arch/arm/mach-imx/mx6/module_fuse.c?h=imx_v2018.03_4.14.98_2.0.0_ga#n149 > But this approach is will not work if kernel dts node path changed. Why would the path change? The DT should be stable. Rob
Hi Rob, > Subject: Re: [RFC 1/2] dt-bindings: imx-ocotp: Add fusable-node property > > On Mon, May 20, 2019 at 03:06:35AM +0000, Peng Fan wrote: > > Introduce fusable-node property for i.MX OCOTP driver. > > The property will only be used by Firmware(eg. U-Boot) to runtime > > disable the nodes. > > > > Take i.MX6ULL for example, there are several parts that only have > > limited modules enabled controlled by OCOTP fuse. It is not flexible > > to provide several dts for the serval parts, instead we could provide > > one device tree and let Firmware to runtime disable the device tree > > nodes for those modules that are disable(fused). > > > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > > --- > > > > Currently NXP vendor use U-Boot to set status to disabled for devices > > that could not function, > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fsour > > > ce.codeaurora.org%2Fexternal%2Fimx%2Fuboot-imx%2Ftree%2Farch%2Far > m%2Fm > > > ach-imx%2Fmx6%2Fmodule_fuse.c%3Fh%3Dimx_v2018.03_4.14.98_2.0.0_g > a%23n1 > > > 49&data=02%7C01%7Cpeng.fan%40nxp.com%7Ceb74876c943c4471d7f > f08d6f05 > > > 2fffb%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C63696063660 > 6051306& > > > amp;sdata=MrjCP0ufBu3uTb1ehTEu2g5eSQuYcho4UxuR9KUFEOA%3D& > reserved= > > 0 But this approach is will not work if kernel dts node path changed. > > Why would the path change? The DT should be stable. It changes sometimes, such as efb9adb27475 ("ARM: dts: imx6ul: Remove leading zeroes from unit addresses"), in bootloader, we are using node path, so if the node name changes, the bootloader will fail to disable the node. Thanks, Peng. > > Rob
diff --git a/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt b/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt index 7a999a135e56..e9a998588dbd 100644 --- a/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt +++ b/Documentation/devicetree/bindings/nvmem/imx-ocotp.txt @@ -21,6 +21,8 @@ Required properties: Optional properties: - read-only: disable write access +- fusable-node: array of phandles with reg base and bit offset, this + property is used by Firmware to runtime disable nodes. Optional Child nodes: @@ -42,4 +44,7 @@ Example: tempmon_temp_grade: temp-grade@20 { reg = <0x20 4>; }; + + fusable-node = <&usdhc1 0x10 4 + &usdhc2 0x10 5>; };
Introduce fusable-node property for i.MX OCOTP driver. The property will only be used by Firmware(eg. U-Boot) to runtime disable the nodes. Take i.MX6ULL for example, there are several parts that only have limited modules enabled controlled by OCOTP fuse. It is not flexible to provide several dts for the serval parts, instead we could provide one device tree and let Firmware to runtime disable the device tree nodes for those modules that are disable(fused). Signed-off-by: Peng Fan <peng.fan@nxp.com> --- Currently NXP vendor use U-Boot to set status to disabled for devices that could not function, https://source.codeaurora.org/external/imx/uboot-imx/tree/arch/arm/mach-imx/mx6/module_fuse.c?h=imx_v2018.03_4.14.98_2.0.0_ga#n149 But this approach is will not work if kernel dts node path changed. There are two approaches to resolve: 1. This patch is to add a fusable-node property, and Firmware will parse the property and read fuse to decide whether to disable or keeep enable the nodes. 2. There is another approach is that add nvmem-cells for all nodes that could be disabled(fused). Then in each linux driver to use nvmem api to detect fused or not, or in linux driver common code to check device functionable or not with nvmem API. To make it easy to work, we choose [1] here. Please advise whether it is acceptable, because the property is not used by linux driver in approach [1]. Or you prefer [2] or please advise if any better solution. Thanks. Documentation/devicetree/bindings/nvmem/imx-ocotp.txt | 5 +++++ 1 file changed, 5 insertions(+)