| Message ID | 1456317210-42742-2-git-send-email-qiujiang@huawei.com |
|---|---|
| State | New |
| Headers | show |
On Wed, Feb 24, 2016 at 2:33 PM, qiujiang <qiujiang@huawei.com> wrote: > This patch switch device node to fwnode in dwapb_port_property, > so as to apply a unified data structure for DT and ACPI. > > This change also needs to be done in intel_quark_i2c_gpio driver, > since it depends on gpio-dwapb driver. > > Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com> > Signed-off-by: qiujiang <qiujiang@huawei.com> Yes, something like this. Though I have questions: - why do you use fwnode_*() instead of device_property_*() calls? What prevents us to move to device property API directly? > - gpio->domain = irq_domain_add_linear(node, ngpio, > - &irq_generic_chip_ops, gpio); > + gpio->domain = irq_domain_create_linear(fwnode, ngpio, > + &irq_generic_chip_ops, gpio); Are they equivalent? > @@ -415,7 +415,7 @@ static int dwapb_gpio_add_port(struct dwapb_gpio *gpio, > } > > #ifdef CONFIG_OF_GPIO > - port->gc.of_node = pp->node; > + port->gc.of_node = to_of_node(pp->fwnode); If fwnode is not OF one? Perhaps, something like ... = is_of_node() ? to_of_node() : NULL; > - node = dev->of_node; > - if (!IS_ENABLED(CONFIG_OF_GPIO) || !node) > + if (!IS_ENABLED(CONFIG_OF_GPIO) || !(dev->of_node)) > return ERR_PTR(-ENODEV); So, since you converted to fwnode, do you still need this check? > > - nports = of_get_child_count(node); > + nports = device_get_child_node_count(dev); > if (nports == 0) > return ERR_PTR(-ENODEV); ...I think this one fail if it will not found any child. > - if (of_property_read_u32(port_np, "reg", &pp->idx) || > + if (fwnode_property_read_u32(fwnode, "reg", &pp->idx) || device_property_*() ? > pp->idx >= DWAPB_MAX_PORTS) { > dev_err(dev, "missing/invalid port index for %s\n", > - port_np->full_name); > + to_of_node(fwnode)->full_name); If it's not OF? > - if (of_property_read_u32(port_np, "snps,nr-gpios", > + if (fwnode_property_read_u32(fwnode, "snps,nr-gpios", Ditto. > &pp->ngpio)) { > dev_info(dev, "failed to get number of gpios for %s\n", > - port_np->full_name); > + to_of_node(fwnode)->full_name); Ditto. > if (pp->idx == 0 && > - of_property_read_bool(port_np, "interrupt-controller")) { > - pp->irq = irq_of_parse_and_map(port_np, 0); > + of_property_read_bool(to_of_node(fwnode), > + "interrupt-controller")) { device_property_*() ? > + pp->irq = irq_of_parse_and_map(to_of_node(fwnode), 0); > if (!pp->irq) { > dev_warn(dev, "no irq for bank %s\n", > - port_np->full_name); > + to_of_node(fwnode)->full_name); > } > } > > pp->irq_shared = false; > pp->gpio_base = -1; > - pp->name = port_np->full_name; > + pp->name = to_of_node(fwnode)->full_name; > } > > return pdata;
在 2016/2/24 21:46, Andy Shevchenko 写道: > On Wed, Feb 24, 2016 at 2:33 PM, qiujiang <qiujiang@huawei.com> wrote: >> This patch switch device node to fwnode in dwapb_port_property, >> so as to apply a unified data structure for DT and ACPI. >> >> This change also needs to be done in intel_quark_i2c_gpio driver, >> since it depends on gpio-dwapb driver. >> >> Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com> >> Signed-off-by: qiujiang <qiujiang@huawei.com> > > Yes, something like this. > Though I have questions: > - why do you use fwnode_*() instead of device_property_*() calls? > What prevents us to move to device property API directly? Yes, it looks more reasonable by using devce_property. Howerver, device_get_child_node_count was used here to find each child node. This API output the fwnode_handle for each child node directly, but device property APIs need 'dev' data instead. Actually, the effects of fwnode_*() and device_*() are the same. So, I used fwnode_*() APIs here. If there is any other more way to traverse child nodes, let me know. Thank you. > >> - gpio->domain = irq_domain_add_linear(node, ngpio, >> - &irq_generic_chip_ops, gpio); >> + gpio->domain = irq_domain_create_linear(fwnode, ngpio, >> + &irq_generic_chip_ops, gpio); > > Are they equivalent? Yes, they are equivalent. > >> @@ -415,7 +415,7 @@ static int dwapb_gpio_add_port(struct dwapb_gpio *gpio, >> } >> >> #ifdef CONFIG_OF_GPIO >> - port->gc.of_node = pp->node; >> + port->gc.of_node = to_of_node(pp->fwnode); > > If fwnode is not OF one? > Perhaps, something like ... = is_of_node() ? to_of_node() : NULL; > The way you suggested is more resonable, I will fixed it in next version. > >> - node = dev->of_node; >> - if (!IS_ENABLED(CONFIG_OF_GPIO) || !node) >> + if (!IS_ENABLED(CONFIG_OF_GPIO) || !(dev->of_node)) >> return ERR_PTR(-ENODEV); > > So, since you converted to fwnode, do you still need this check? > Although this patch coverted device node to fwnode, only DTs binding was supported here, and patch2 support ACPI will remove this check. >> >> - nports = of_get_child_count(node); >> + nports = device_get_child_node_count(dev); >> if (nports == 0) >> return ERR_PTR(-ENODEV); > > ...I think this one fail if it will not found any child. This one fail? yes, it will return to failure. I am not very clear here. > >> - if (of_property_read_u32(port_np, "reg", &pp->idx) || >> + if (fwnode_property_read_u32(fwnode, "reg", &pp->idx) || > > device_property_*() ? > >> pp->idx >= DWAPB_MAX_PORTS) { >> dev_err(dev, "missing/invalid port index for %s\n", >> - port_np->full_name); >> + to_of_node(fwnode)->full_name); > > If it's not OF? This is checked above, and patch2 will remove it. > >> - if (of_property_read_u32(port_np, "snps,nr-gpios", >> + if (fwnode_property_read_u32(fwnode, "snps,nr-gpios", > > Ditto. > >> &pp->ngpio)) { >> dev_info(dev, "failed to get number of gpios for %s\n", >> - port_np->full_name); >> + to_of_node(fwnode)->full_name); > > Ditto. > >> if (pp->idx == 0 && >> - of_property_read_bool(port_np, "interrupt-controller")) { >> - pp->irq = irq_of_parse_and_map(port_np, 0); >> + of_property_read_bool(to_of_node(fwnode), >> + "interrupt-controller")) { > > device_property_*() ? > >> + pp->irq = irq_of_parse_and_map(to_of_node(fwnode), 0); >> if (!pp->irq) { >> dev_warn(dev, "no irq for bank %s\n", >> - port_np->full_name); >> + to_of_node(fwnode)->full_name); >> } >> } >> >> pp->irq_shared = false; >> pp->gpio_base = -1; >> - pp->name = port_np->full_name; >> + pp->name = to_of_node(fwnode)->full_name; >> } >> >> return pdata; > > -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Feb 25, 2016 at 1:58 PM, Jiang Qiu <qiujiang@huawei.com> wrote: > 在 2016/2/24 21:46, Andy Shevchenko 写道: >> On Wed, Feb 24, 2016 at 2:33 PM, qiujiang <qiujiang@huawei.com> wrote: >> - why do you use fwnode_*() instead of device_property_*() calls? >> What prevents us to move to device property API directly? > Yes, it looks more reasonable by using devce_property. Howerver, > device_get_child_node_count was used here to find each child node. This > API output the fwnode_handle for each child node directly, but device > property APIs need 'dev' data instead. Actually, the effects of fwnode_*() > and device_*() are the same. So, I used fwnode_*() APIs here. Right, looks okay then. >>> - node = dev->of_node; >>> - if (!IS_ENABLED(CONFIG_OF_GPIO) || !node) >>> + if (!IS_ENABLED(CONFIG_OF_GPIO) || !(dev->of_node)) >>> return ERR_PTR(-ENODEV); >> >> So, since you converted to fwnode, do you still need this check? >> > Although this patch coverted device node to fwnode, only DTs binding was > supported here, and patch2 support ACPI will remove this check. Yes, but like I said below device_get_child_node_count() will take care of that, will it? >>> >>> - nports = of_get_child_count(node); >>> + nports = device_get_child_node_count(dev); >>> if (nports == 0) >>> return ERR_PTR(-ENODEV); >> >> ...I think this one fail if it will not found any child. > This one fail? yes, it will return to failure. > I am not very clear here. See above.
在 2016/2/25 21:43, Andy Shevchenko 写道: > On Thu, Feb 25, 2016 at 1:58 PM, Jiang Qiu <qiujiang@huawei.com> wrote: >> 在 2016/2/24 21:46, Andy Shevchenko 写道: >>> On Wed, Feb 24, 2016 at 2:33 PM, qiujiang <qiujiang@huawei.com> wrote: > >>> - why do you use fwnode_*() instead of device_property_*() calls? >>> What prevents us to move to device property API directly? >> Yes, it looks more reasonable by using devce_property. Howerver, >> device_get_child_node_count was used here to find each child node. This >> API output the fwnode_handle for each child node directly, but device >> property APIs need 'dev' data instead. Actually, the effects of fwnode_*() >> and device_*() are the same. So, I used fwnode_*() APIs here. > > Right, looks okay then. > >>>> - node = dev->of_node; >>>> - if (!IS_ENABLED(CONFIG_OF_GPIO) || !node) >>>> + if (!IS_ENABLED(CONFIG_OF_GPIO) || !(dev->of_node)) >>>> return ERR_PTR(-ENODEV); >>> >>> So, since you converted to fwnode, do you still need this check? >>> >> Although this patch coverted device node to fwnode, only DTs binding was >> supported here, and patch2 support ACPI will remove this check. > > Yes, but like I said below device_get_child_node_count() will take > care of that, will it? Right, device_get_child_node_count() will take of it, this should be removed. > >>>> >>>> - nports = of_get_child_count(node); >>>> + nports = device_get_child_node_count(dev); >>>> if (nports == 0) >>>> return ERR_PTR(-ENODEV); >>> >>> ...I think this one fail if it will not found any child. >> This one fail? yes, it will return to failure. >> I am not very clear here. > > See above. Here, device_get_child_node_count will return ZERO if there is not any child. So, I think this will work ok, will it? > -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sat, Feb 27, 2016 at 9:15 AM, Jiang Qiu <qiujiang@huawei.com> wrote: >>>>> - node = dev->of_node; >>>>> - if (!IS_ENABLED(CONFIG_OF_GPIO) || !node) >>>>> + if (!IS_ENABLED(CONFIG_OF_GPIO) || !(dev->of_node)) >>>>> return ERR_PTR(-ENODEV); >>>> >>>> So, since you converted to fwnode, do you still need this check? >>>> >>> Although this patch coverted device node to fwnode, only DTs binding was >>> supported here, and patch2 support ACPI will remove this check. >> >> Yes, but like I said below device_get_child_node_count() will take >> care of that, will it? > Right, device_get_child_node_count() will take of it, this should be removed. >> >>>>> >>>>> - nports = of_get_child_count(node); >>>>> + nports = device_get_child_node_count(dev); >>>>> if (nports == 0) >>>>> return ERR_PTR(-ENODEV); >>>> >>>> ...I think this one fail if it will not found any child. >>> This one fail? yes, it will return to failure. >>> I am not very clear here. >> >> See above. > Here, device_get_child_node_count will return ZERO if there is not any child. > So, I think this will work ok, will it? I didn't check deeply, but I assume so.
diff --git a/drivers/gpio/gpio-dwapb.c b/drivers/gpio/gpio-dwapb.c index 597de1e..0ebbdf1 100644 --- a/drivers/gpio/gpio-dwapb.c +++ b/drivers/gpio/gpio-dwapb.c @@ -290,14 +290,14 @@ static void dwapb_configure_irqs(struct dwapb_gpio *gpio, struct dwapb_port_property *pp) { struct gpio_chip *gc = &port->gc; - struct device_node *node = pp->node; + struct fwnode_handle *fwnode = pp->fwnode; struct irq_chip_generic *irq_gc = NULL; unsigned int hwirq, ngpio = gc->ngpio; struct irq_chip_type *ct; int err, i; - gpio->domain = irq_domain_add_linear(node, ngpio, - &irq_generic_chip_ops, gpio); + gpio->domain = irq_domain_create_linear(fwnode, ngpio, + &irq_generic_chip_ops, gpio); if (!gpio->domain) return; @@ -415,7 +415,7 @@ static int dwapb_gpio_add_port(struct dwapb_gpio *gpio, } #ifdef CONFIG_OF_GPIO - port->gc.of_node = pp->node; + port->gc.of_node = to_of_node(pp->fwnode); #endif port->gc.ngpio = pp->ngpio; port->gc.base = pp->gpio_base; @@ -449,17 +449,16 @@ static void dwapb_gpio_unregister(struct dwapb_gpio *gpio) static struct dwapb_platform_data * dwapb_gpio_get_pdata_of(struct device *dev) { - struct device_node *node, *port_np; + struct fwnode_handle *fwnode; struct dwapb_platform_data *pdata; struct dwapb_port_property *pp; int nports; int i; - node = dev->of_node; - if (!IS_ENABLED(CONFIG_OF_GPIO) || !node) + if (!IS_ENABLED(CONFIG_OF_GPIO) || !(dev->of_node)) return ERR_PTR(-ENODEV); - nports = of_get_child_count(node); + nports = device_get_child_node_count(dev); if (nports == 0) return ERR_PTR(-ENODEV); @@ -474,21 +473,21 @@ dwapb_gpio_get_pdata_of(struct device *dev) pdata->nports = nports; i = 0; - for_each_child_of_node(node, port_np) { + device_for_each_child_node(dev, fwnode) { pp = &pdata->properties[i++]; - pp->node = port_np; + pp->fwnode = fwnode; - if (of_property_read_u32(port_np, "reg", &pp->idx) || + if (fwnode_property_read_u32(fwnode, "reg", &pp->idx) || pp->idx >= DWAPB_MAX_PORTS) { dev_err(dev, "missing/invalid port index for %s\n", - port_np->full_name); + to_of_node(fwnode)->full_name); return ERR_PTR(-EINVAL); } - if (of_property_read_u32(port_np, "snps,nr-gpios", + if (fwnode_property_read_u32(fwnode, "snps,nr-gpios", &pp->ngpio)) { dev_info(dev, "failed to get number of gpios for %s\n", - port_np->full_name); + to_of_node(fwnode)->full_name); pp->ngpio = 32; } @@ -497,17 +496,18 @@ dwapb_gpio_get_pdata_of(struct device *dev) * the IP. */ if (pp->idx == 0 && - of_property_read_bool(port_np, "interrupt-controller")) { - pp->irq = irq_of_parse_and_map(port_np, 0); + of_property_read_bool(to_of_node(fwnode), + "interrupt-controller")) { + pp->irq = irq_of_parse_and_map(to_of_node(fwnode), 0); if (!pp->irq) { dev_warn(dev, "no irq for bank %s\n", - port_np->full_name); + to_of_node(fwnode)->full_name); } } pp->irq_shared = false; pp->gpio_base = -1; - pp->name = port_np->full_name; + pp->name = to_of_node(fwnode)->full_name; } return pdata; diff --git a/drivers/mfd/intel_quark_i2c_gpio.c b/drivers/mfd/intel_quark_i2c_gpio.c index 0421374..265bd3c 100644 --- a/drivers/mfd/intel_quark_i2c_gpio.c +++ b/drivers/mfd/intel_quark_i2c_gpio.c @@ -227,7 +227,7 @@ static int intel_quark_gpio_setup(struct pci_dev *pdev, struct mfd_cell *cell) return -ENOMEM; /* Set the properties for portA */ - pdata->properties->node = NULL; + pdata->properties->fwnode = NULL; pdata->properties->name = "intel-quark-x1000-gpio-portA"; pdata->properties->idx = 0; pdata->properties->ngpio = INTEL_QUARK_MFD_NGPIO; diff --git a/include/linux/platform_data/gpio-dwapb.h b/include/linux/platform_data/gpio-dwapb.h index 28702c8..80954f2 100644 --- a/include/linux/platform_data/gpio-dwapb.h +++ b/include/linux/platform_data/gpio-dwapb.h @@ -15,7 +15,7 @@ #define GPIO_DW_APB_H struct dwapb_port_property { - struct device_node *node; + struct fwnode_handle *fwnode; const char *name; unsigned int idx; unsigned int ngpio;