Message ID | 20210603101822.9645-1-steven_lee@aspeedtech.com |
---|---|
Headers | show |
Series | ASPEED sgpio driver enhancement. | expand |
On Thu, Jun 3, 2021 at 1:19 PM Steven Lee <steven_lee@aspeedtech.com> wrote: > > AST2600 SoC has 2 SGPIO master interfaces one with 128 pins another one > with 80 pins. > In the current driver, the maximum number of gpio pins of SoC is hardcoded > as 80 and the gpio pin count mask for GPIO Configuration register is > hardcode as GENMASK(9,6). In addition, some functions uses the hardcoded use > value to calculate the gpio offset. > The patch adds ast2600 compatibles and platform data that includes the > max number of gpio pins supported by ast2600 and gpio pin count mask for > GPIO Configuration register. > The patch also modifies some functions to pass aspeed_sgpio struct for > calculating gpio offset wihtout using the hardcoded value. without ... > +#include <linux/of_device.h> Why? ... > +#define GPIO_OFFSET(x) ((x) & 0x1f) GENMASK() ... > + pdata = of_device_get_match_data(&pdev->dev); device_get_match_data() I guess you may replace all those of_*() to the corresponding device_*() or fwnode_*() calls.
On Thu, 3 Jun 2021, at 19:48, Steven Lee wrote: > The current design initializes irq->chip from a global irqchip struct, > which causes multiple sgpio devices use the same irq_chip. > The patch moves irq_chip to aspeed_sgpio struct for initializing > irq_chip from their private gpio struct. > > Signed-off-by: Steven Lee <steven_lee@aspeedtech.com> Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
The 06/03/2021 19:05, Andy Shevchenko wrote: > On Thu, Jun 3, 2021 at 1:19 PM Steven Lee <steven_lee@aspeedtech.com> wrote: > > > > AST2600 SoC has 2 SGPIO master interfaces one with 128 pins another one > > with 80 pins. > > In the current driver, the maximum number of gpio pins of SoC is hardcoded > > as 80 and the gpio pin count mask for GPIO Configuration register is > > hardcode as GENMASK(9,6). In addition, some functions uses the hardcoded > > use > > > value to calculate the gpio offset. > > The patch adds ast2600 compatibles and platform data that includes the > > max number of gpio pins supported by ast2600 and gpio pin count mask for > > GPIO Configuration register. > > The patch also modifies some functions to pass aspeed_sgpio struct for > > calculating gpio offset wihtout using the hardcoded value. > > without > > ... > > > +#include <linux/of_device.h> > > Why? > > ... > I will remove it as of_device_get_match_data() will be replaced to device_get_match_data() > > +#define GPIO_OFFSET(x) ((x) & 0x1f) > > GENMASK() > > ... > Do you mean the macro should be modified as follows? #define GPIO_OFFSET(x) ((x) & GENMASK(4, 0)) > > + pdata = of_device_get_match_data(&pdev->dev); > > device_get_match_data() > > I guess you may replace all those of_*() to the corresponding > device_*() or fwnode_*() calls. > Thanks for the reviews, I will add a new patch for replacing all of_*() to device_*(). > -- > With Best Regards, > Andy Shevchenko
The 06/04/2021 07:25, Andrew Jeffery wrote: > Hi Steven, > > On Thu, 3 Jun 2021, at 19:48, Steven Lee wrote: > > sgpio-aspeed bindings should be converted to yaml format. > > > > Signed-off-by: Steven Lee <steven_lee@aspeedtech.com> > > --- > > .../bindings/gpio/aspeed,sgpio.yaml | 78 +++++++++++++++++++ > > .../devicetree/bindings/gpio/sgpio-aspeed.txt | 46 ----------- > > 2 files changed, 78 insertions(+), 46 deletions(-) > > create mode 100644 Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > delete mode 100644 Documentation/devicetree/bindings/gpio/sgpio-aspeed.txt > > > > diff --git a/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > b/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > new file mode 100644 > > index 000000000000..e7c2113cc096 > > --- /dev/null > > +++ b/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > @@ -0,0 +1,78 @@ > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) > > +%YAML 1.2 > > +--- > > +$id: http://devicetree.org/schemas/gpio/aspeed,sgpio.yaml# > > +$schema: http://devicetree.org/meta-schemas/core.yaml# > > + > > +title: Aspeed SGPIO controller > > + > > +maintainers: > > + - Andrew Jeffery <andrew@aj.id.au> > > + > > +description: > > + This SGPIO controller is for ASPEED AST2400, AST2500 and AST2600 SoC, > > + AST2600 have two sgpio master one with 128 pins another one with 80 > > pins, > > + AST2500/AST2400 have one sgpio master with 80 pins. Each of the > > Serial > > + GPIO pins can be programmed to support the following options > > + - Support interrupt option for each input port and various interrupt > > + sensitivity option (level-high, level-low, edge-high, edge-low) > > + - Support reset tolerance option for each output port > > + - Directly connected to APB bus and its shift clock is from APB bus > > clock > > + divided by a programmable value. > > + - Co-work with external signal-chained TTL components > > (74LV165/74LV595) > > + > > +properties: > > + compatible: > > + enum: > > + - aspeed,ast2400-sgpio > > + - aspeed,ast2500-sgpio > > + - aspeed,ast2600-sgpiom1 > > + - aspeed,ast2600-sgpiom2 > > You should have followed Rob's request here and made two patches for > the binding document: > > 1. A 1-to-1 conversion of the text file to dt-schema > 2. Add your new compatibles for the 2600. > Sorry I forgot to remove compatibles and move them to a new patch. > From a cursory glance it looks okay except for the new compatibles. > > Regarding the compatibles, I'd prefer we use something a bit more > meaningful. What do you think of these? > > - aspeed,ast2600-sgpiom-80 > - aspeed,ast2600-sgpiom-128 > Ok, I will change the name as you suggested. BTW, I and development team have an internal discussion about the current sgpio design. In the current design, the base offset of gpio input and output are calculated by the maximum number of gpio pins that SoC supported. For instance, in AST2500, max_ngpios is 80(defined in MAX_NR_HW_SGPIO), if ngpios is 16 in dts, gpio input pin id is from 0 to 15 and gpio output pin id is from 80 to 95. We are thinking of removing max_ngpios(and MAX_NR_HW_SGPIO) and corresponding design to make the gpio input and output pin base are determined by ngpios. For instance, in any AST SoC, if ngpios is 16 in dts, gpio input pin id is from 0 to 15 and gpio output pin id is from 16 to 31. Thus we don't need to care about the max_ngpios of SoCs, and needn't to add 2 compatibles for ast2600. However, it might affect users who update kernel/driver from the old kernel/driver as they may expect the gpio output pin base is start from 80(MAX_NR_HW_SGPIO). I was wondering if it is better to change the design as above. It would be great to have your suggestion. Thanks, Steven > Cheers, > > Andrew
On Fri, 4 Jun 2021, at 13:00, Steven Lee wrote: > The 06/04/2021 07:25, Andrew Jeffery wrote: > > Hi Steven, > > > > On Thu, 3 Jun 2021, at 19:48, Steven Lee wrote: > > > sgpio-aspeed bindings should be converted to yaml format. > > > > > > Signed-off-by: Steven Lee <steven_lee@aspeedtech.com> > > > --- > > > .../bindings/gpio/aspeed,sgpio.yaml | 78 +++++++++++++++++++ > > > .../devicetree/bindings/gpio/sgpio-aspeed.txt | 46 ----------- > > > 2 files changed, 78 insertions(+), 46 deletions(-) > > > create mode 100644 Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > > delete mode 100644 Documentation/devicetree/bindings/gpio/sgpio-aspeed.txt > > > > > > diff --git a/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > > b/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > > new file mode 100644 > > > index 000000000000..e7c2113cc096 > > > --- /dev/null > > > +++ b/Documentation/devicetree/bindings/gpio/aspeed,sgpio.yaml > > > @@ -0,0 +1,78 @@ > > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) > > > +%YAML 1.2 > > > +--- > > > +$id: http://devicetree.org/schemas/gpio/aspeed,sgpio.yaml# > > > +$schema: http://devicetree.org/meta-schemas/core.yaml# > > > + > > > +title: Aspeed SGPIO controller > > > + > > > +maintainers: > > > + - Andrew Jeffery <andrew@aj.id.au> > > > + > > > +description: > > > + This SGPIO controller is for ASPEED AST2400, AST2500 and AST2600 SoC, > > > + AST2600 have two sgpio master one with 128 pins another one with 80 > > > pins, > > > + AST2500/AST2400 have one sgpio master with 80 pins. Each of the > > > Serial > > > + GPIO pins can be programmed to support the following options > > > + - Support interrupt option for each input port and various interrupt > > > + sensitivity option (level-high, level-low, edge-high, edge-low) > > > + - Support reset tolerance option for each output port > > > + - Directly connected to APB bus and its shift clock is from APB bus > > > clock > > > + divided by a programmable value. > > > + - Co-work with external signal-chained TTL components > > > (74LV165/74LV595) > > > + > > > +properties: > > > + compatible: > > > + enum: > > > + - aspeed,ast2400-sgpio > > > + - aspeed,ast2500-sgpio > > > + - aspeed,ast2600-sgpiom1 > > > + - aspeed,ast2600-sgpiom2 > > > > You should have followed Rob's request here and made two patches for > > the binding document: > > > > 1. A 1-to-1 conversion of the text file to dt-schema > > 2. Add your new compatibles for the 2600. > > > > Sorry I forgot to remove compatibles and move them to a new patch. > > > From a cursory glance it looks okay except for the new compatibles. > > > > Regarding the compatibles, I'd prefer we use something a bit more > > meaningful. What do you think of these? > > > > - aspeed,ast2600-sgpiom-80 > > - aspeed,ast2600-sgpiom-128 > > > > Ok, I will change the name as you suggested. > > BTW, I and development team have an internal discussion about the > current sgpio design. > > In the current design, the base offset of gpio input and output > are calculated by the maximum number of gpio pins that SoC supported. > For instance, in AST2500, max_ngpios is 80(defined in MAX_NR_HW_SGPIO), > if ngpios is 16 in dts, gpio input pin id is from 0 to 15 and > gpio output pin id is from 80 to 95. > > We are thinking of removing max_ngpios(and MAX_NR_HW_SGPIO) and > corresponding design to make the gpio input and output pin base > are determined by ngpios. > For instance, in any AST SoC, if ngpios is 16 in dts, > gpio input pin id is from 0 to 15 and gpio output pin id is from 16 to 31. > Thus we don't need to care about the max_ngpios of SoCs, and needn't to > add 2 compatibles for ast2600. > > However, it might affect users who update kernel/driver from the > old kernel/driver as they may expect the gpio output pin base is start > from 80(MAX_NR_HW_SGPIO). > I was wondering if it is better to change the design as above. > It would be great to have your suggestion. Right, this breaks userspace. I don't think it's going to fly but I'm interested in feedback from Linus and Bartosz. If we were to break userspace, a scheme I'd consider with is to pair input/output GPIOs. For example, GPIO 0 is input, GPIO 1 is the associated output, GPIO 2 is input, GPIO 3 is output etc. That way you can increase/decrease the number of GPIOs without affecting userspace (after breaking it initially). Andrew
On Fri, Jun 4, 2021 at 5:14 AM Steven Lee <steven_lee@aspeedtech.com> wrote: > The 06/03/2021 19:05, Andy Shevchenko wrote: > > On Thu, Jun 3, 2021 at 1:19 PM Steven Lee <steven_lee@aspeedtech.com> wrote: > > > +#define GPIO_OFFSET(x) ((x) & 0x1f) > > > > GENMASK() > > > > ... > > > > Do you mean the macro should be modified as follows? > #define GPIO_OFFSET(x) ((x) & GENMASK(4, 0)) Yes. -- With Best Regards, Andy Shevchenko
On Fri, Jun 4, 2021 at 5:31 AM Steven Lee <steven_lee@aspeedtech.com> wrote: > However, it might affect users who update kernel/driver from the > old kernel/driver as they may expect the gpio output pin base is start > from 80(MAX_NR_HW_SGPIO). Why? What users? In-kernel, out-of-tree-kernel or userspace users? In-kernel users can be fixed, out-of-tree kernels we don't care about and userspace should be using the character device. Just change it. Yours, Linus Walleij