| Message ID | 20230807131256.254243-3-krzysztof.kozlowski@linaro.org |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v2,1/3] media: dt-bindings: samsung,exynos4212-fimc-is: replace duplicate pmu node with phandle | expand |
Hi Krzysztof, [...] > +static void __iomem *fimc_is_get_pmu_regs(struct device *dev) > +{ > + struct device_node *node; > + void __iomem *regs; > + > + node = of_parse_phandle(dev->of_node, "samsung,pmu-syscon", 0); > + if (!node) { > + dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); > + node = of_get_child_by_name(dev->of_node, "pmu"); > + if (!node) > + return IOMEM_ERR_PTR(-ENODEV); in my opinion this should be: ... if (!node) return IOMEM_ERR_PTR(-ENODEV); dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); Because if you don't have both "samsung,pmu-syscon and "pmu" then the warning should not be printed and you need to return -ENODEV. ... and... "*please* update your DTB", the user might get upset and out of sheer spite, decides not to do it – just because! :) Andi > + }
On 08/08/2023 01:13, Andi Shyti wrote: > Hi Krzysztof, > > [...] > >> +static void __iomem *fimc_is_get_pmu_regs(struct device *dev) >> +{ >> + struct device_node *node; >> + void __iomem *regs; >> + >> + node = of_parse_phandle(dev->of_node, "samsung,pmu-syscon", 0); >> + if (!node) { >> + dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); >> + node = of_get_child_by_name(dev->of_node, "pmu"); >> + if (!node) >> + return IOMEM_ERR_PTR(-ENODEV); > > in my opinion this should be: > > ... > if (!node) > return IOMEM_ERR_PTR(-ENODEV); > > dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); > > Because if you don't have both "samsung,pmu-syscon and "pmu" then > the warning should not be printed and you need to return -ENODEV. Why not? Warning is correct - the driver is trying to find, thus continuous tense "Finding", PMU node via old method. > > ... and... "*please* update your DTB", the user might get upset > and out of sheer spite, decides not to do it – just because! :) The message is already long enough, why making it longer? Anyone who ships DTS outside of Linux kernel is doomed anyway... Best regards, Krzysztof
Hi Krzysztof, On 08/08/2023 01:13, Andi Shyti wrote: > Hi Krzysztof, > > [...] > >> +static void __iomem *fimc_is_get_pmu_regs(struct device *dev) >> +{ >> + struct device_node *node; >> + void __iomem *regs; >> + >> + node = of_parse_phandle(dev->of_node, "samsung,pmu-syscon", 0); >> + if (!node) { >> + dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); >> + node = of_get_child_by_name(dev->of_node, "pmu"); >> + if (!node) >> + return IOMEM_ERR_PTR(-ENODEV); > > in my opinion this should be: > > ... > if (!node) > return IOMEM_ERR_PTR(-ENODEV); > > dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); > > Because if you don't have both "samsung,pmu-syscon and "pmu" then > the warning should not be printed and you need to return -ENODEV. I agree with Andi for this part. The only time you want to see this message is if samsung,pmu-syscon is missing AND pmu is present. If both are missing, then just return ENODEV as it was before. > > ... and... "*please* update your DTB", the user might get upset > and out of sheer spite, decides not to do it – just because! :) I don't care about this bit. I guess it doesn't hurt to add 'please', but I accept it either way. Regards, Hans > > Andi > >> + }
On 11/08/2023 11:49, Hans Verkuil wrote: > Hi Krzysztof, > > On 08/08/2023 01:13, Andi Shyti wrote: >> Hi Krzysztof, >> >> [...] >> >>> +static void __iomem *fimc_is_get_pmu_regs(struct device *dev) >>> +{ >>> + struct device_node *node; >>> + void __iomem *regs; >>> + >>> + node = of_parse_phandle(dev->of_node, "samsung,pmu-syscon", 0); >>> + if (!node) { >>> + dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); >>> + node = of_get_child_by_name(dev->of_node, "pmu"); >>> + if (!node) >>> + return IOMEM_ERR_PTR(-ENODEV); >> >> in my opinion this should be: >> >> ... >> if (!node) >> return IOMEM_ERR_PTR(-ENODEV); >> >> dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); >> >> Because if you don't have both "samsung,pmu-syscon and "pmu" then >> the warning should not be printed and you need to return -ENODEV. > > I agree with Andi for this part. > > The only time you want to see this message is if samsung,pmu-syscon is > missing AND pmu is present. If both are missing, then just return ENODEV as > it was before. OK, understood. I will send a v3. Best regards, Krzysztof
diff --git a/drivers/media/platform/samsung/exynos4-is/fimc-is.c b/drivers/media/platform/samsung/exynos4-is/fimc-is.c index 530a148fe4d3..c995b1226ca3 100644 --- a/drivers/media/platform/samsung/exynos4-is/fimc-is.c +++ b/drivers/media/platform/samsung/exynos4-is/fimc-is.c @@ -767,12 +767,32 @@ static void fimc_is_debugfs_create(struct fimc_is *is) static int fimc_is_runtime_resume(struct device *dev); static int fimc_is_runtime_suspend(struct device *dev); +static void __iomem *fimc_is_get_pmu_regs(struct device *dev) +{ + struct device_node *node; + void __iomem *regs; + + node = of_parse_phandle(dev->of_node, "samsung,pmu-syscon", 0); + if (!node) { + dev_warn(dev, "Finding PMU node via deprecated method, update your DTB\n"); + node = of_get_child_by_name(dev->of_node, "pmu"); + if (!node) + return IOMEM_ERR_PTR(-ENODEV); + } + + regs = of_iomap(node, 0); + of_node_put(node); + if (!regs) + return IOMEM_ERR_PTR(-ENOMEM); + + return regs; +} + static int fimc_is_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; struct fimc_is *is; struct resource res; - struct device_node *node; int ret; is = devm_kzalloc(&pdev->dev, sizeof(*is), GFP_KERNEL); @@ -794,14 +814,9 @@ static int fimc_is_probe(struct platform_device *pdev) if (IS_ERR(is->regs)) return PTR_ERR(is->regs); - node = of_get_child_by_name(dev->of_node, "pmu"); - if (!node) - return -ENODEV; - - is->pmu_regs = of_iomap(node, 0); - of_node_put(node); - if (!is->pmu_regs) - return -ENOMEM; + is->pmu_regs = fimc_is_get_pmu_regs(dev); + if (IS_ERR(is->pmu_regs)) + return PTR_ERR(is->pmu_regs); is->irq = irq_of_parse_and_map(dev->of_node, 0); if (!is->irq) {
Devicetree for the FIMC IS camera included duplicated PMU node as its child like: soc@0 { system-controller@10020000 { ... }; // Real PMU camera@11800000 { fimc-is@12000000 { // FIMC IS camera node pmu@10020000 { reg = <0x10020000 0x3000>; // Fake PMU node }; }; }; }; This is not a correct representation of the hardware. Mapping the PMU (Power Management Unit) IO memory should be via syscon-like phandle (samsung,pmu-syscon, already used for other drivers), not by duplicating "pmu" Devicetree node inside the FIMC IS. Backward compatibility is preserved. Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> --- Changes in v2: 1. Use IOMEM_ERR_PTR (Hans) --- .../platform/samsung/exynos4-is/fimc-is.c | 33 ++++++++++++++----- 1 file changed, 24 insertions(+), 9 deletions(-)