| Message ID | 1401829985-5212-2-git-send-email-lho@apm.com |
|---|---|
| State | Superseded, archived |
| Headers | show |
On 06/03/2014 11:13 PM, Loc Ho wrote: > This patch adds support for the APM X-Gene SoC SDHC controller > to Arasan SDHCI driver. > > Signed-off-by: Loc Ho <lho@apm.com> > --- > drivers/mmc/host/Kconfig | 4 +- > drivers/mmc/host/sdhci-of-arasan.c | 126 +++++++++++++++++++++++++++++++++--- > 2 files changed, 120 insertions(+), 10 deletions(-) > > diff --git a/drivers/mmc/host/Kconfig b/drivers/mmc/host/Kconfig > index 8aaf8c1..7ec5414 100644 > --- a/drivers/mmc/host/Kconfig > +++ b/drivers/mmc/host/Kconfig > @@ -108,9 +108,11 @@ config MMC_SDHCI_OF_ARASAN > tristate "SDHCI OF support for the Arasan SDHCI controllers" > depends on MMC_SDHCI_PLTFM > depends on OF > + select MMC_SDHCI_IO_ACCESSORS > help > This selects the Arasan Secure Digital Host Controller Interface > - (SDHCI). This hardware is found e.g. in Xilinx' Zynq SoC. > + (SDHCI). This hardware is found e.g. in Xilinx' Zynq and X-Gene > + SoC. > > If you have a controller with this interface, say Y or M here. > > diff --git a/drivers/mmc/host/sdhci-of-arasan.c b/drivers/mmc/host/sdhci-of-arasan.c > index f7c7cf6..4f1313c 100644 > --- a/drivers/mmc/host/sdhci-of-arasan.c > +++ b/drivers/mmc/host/sdhci-of-arasan.c > @@ -20,6 +20,8 @@ > */ > > #include <linux/module.h> > +#include <linux/dma-mapping.h> > +#include <linux/of_device.h> Keep headers sorted. > #include "sdhci-pltfm.h" > > #define SDHCI_ARASAN_CLK_CTRL_OFFSET 0x2c > @@ -34,6 +36,19 @@ > */ > struct sdhci_arasan_data { > struct clk *clk_ahb; > + struct platform_device *pdev; > + void __iomem *ahb_aim_csr; > + const struct sdhci_arasan_ahb_ops *ahb_ops; > +}; > + > +/** > + * struct sdhci_arasan_ahb_ops > + * @init_ahb Initialize translation bus > + * @xlat_addr Set up an 64-bit addressing translation This definitely breaking kernel-doc format. Please fix. > + */ > +struct sdhci_arasan_ahb_ops { > + int (*init_ahb)(struct sdhci_arasan_data *data); > + void (*xlat_addr)(struct sdhci_arasan_data *data, u64 dma_addr); > }; > > static unsigned int sdhci_arasan_get_timeout_clock(struct sdhci_host *host) > @@ -51,7 +66,21 @@ static unsigned int sdhci_arasan_get_timeout_clock(struct sdhci_host *host) > return freq; > } > > +static void sdhci_arasan_writel(struct sdhci_host *host, u32 val, int reg) > +{ > + struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host); > + struct sdhci_arasan_data *sdhci_arasan = pltfm_host->priv; > + > + if (reg == SDHCI_DMA_ADDRESS) { > + if (sdhci_arasan->ahb_ops && sdhci_arasan->ahb_ops->xlat_addr) just one if here. > + sdhci_arasan->ahb_ops->xlat_addr(sdhci_arasan, > + sg_dma_address(host->data->sg)); > + } > + writel(val, host->ioaddr + reg); > +} The same code was submitted in v1 and Arnd commented it but you are still keeping it here. Why? > + > static struct sdhci_ops sdhci_arasan_ops = { > + .write_l = sdhci_arasan_writel, > .get_max_clock = sdhci_pltfm_clk_get_max_clock, > .get_timeout_clock = sdhci_arasan_get_timeout_clock, > }; > @@ -121,13 +150,83 @@ static int sdhci_arasan_resume(struct device *dev) > static SIMPLE_DEV_PM_OPS(sdhci_arasan_dev_pm_ops, sdhci_arasan_suspend, > sdhci_arasan_resume); > > +static int sdhci_arasan_xgene_init_ahb(struct sdhci_arasan_data *data) > +{ > + #define AIM_SIZE_CTL_OFFSET 0x00000004 > + #define AIM_EN_N_WR(src) (((u32) (src) << 31) & 0x80000000) > + #define ARSB_WR(src) (((u32) (src) << 24) & 0x0f000000) > + #define AWSB_WR(src) (((u32) (src) << 20) & 0x00f00000) > + #define AIM_MASK_N_WR(src) (((u32) (src)) & 0x000fffff) Remove that one more space between #define and name. I am not fan of having these defines just here - move them to the top. Also these macros are used just here at one location. Isn't it better just to define that BITS you want to setup instead of these macros which are hardly to read? > + > + struct sdhci_host *host = platform_get_drvdata(data->pdev); > + int ret; > + > + if (!data->ahb_aim_csr) > + return 0; > + > + /* > + * Setup AHB AIM windows ctrl register. The lower 32-bit is left > + * at 0 while the upper bit are programmed when the buffer address > + ( is set from function sdhci_arasn_writel. broken comment. > + */ > + writel(AIM_EN_N_WR(1) | ARSB_WR(1) | AWSB_WR(1) | AIM_MASK_N_WR(0), > + data->ahb_aim_csr + AIM_SIZE_CTL_OFFSET); > + > + /* Set DMA mask */ > + ret = dma_set_mask_and_coherent(&data->pdev->dev, DMA_BIT_MASK(64)); > + if (ret) { > + dev_err(&data->pdev->dev, "Unable to set dma mask\n"); > + return ret; > + } > + > + /* > + * This shouldn't be necessary. Just in case the FW doesn't > + * configure disable ADMA support as we can't support multiple > + * DMA buffer whose address is 64-bit. The AHB translation bridge > + * only has 8 entry max and that is required to be shared and > + * upper layer can pass more than 8 buffer pointers. > + */ > + host->quirks |= SDHCI_QUIRK_BROKEN_ADMA; > + > + return 0; > +} > + > +static void sdhci_arasn_xgene_xlat_addr(struct sdhci_arasan_data *data, > + u64 dma_addr) > +{ > + #define AIM_AXI_HI_OFFSET 0x0000000c > + #define AIM_AXI_ADDRESS_HI_N_WR(src) \ > + (((u32) (src) << 20) & 0xfff00000) ditto with indentation. 20 should be shift 0xfff00000 is mask. Also you should take this opportunity and add function description in kernel doc to be exactly clear what this function is doing. > + > + if (!data->ahb_aim_csr) > + return; > + > + writel(AIM_AXI_ADDRESS_HI_N_WR(dma_addr >> 32), > + data->ahb_aim_csr + AIM_AXI_HI_OFFSET); > +} > + > +static const struct sdhci_arasan_ahb_ops xgene_ahb_ops = { > + .init_ahb = sdhci_arasan_xgene_init_ahb, > + .xlat_addr = sdhci_arasn_xgene_xlat_addr, > +}; > + > +static const struct of_device_id sdhci_arasan_of_match[] = { > + { .compatible = "arasan,sdhci-8.9a" }, > + { .compatible = "apm,arasan,sdhci-8.9a", .data = &xgene_ahb_ops }, > + { } > +}; > +MODULE_DEVICE_TABLE(of, sdhci_arasan_of_match); > + > static int sdhci_arasan_probe(struct platform_device *pdev) > { > int ret; > - struct clk *clk_xin; > + struct clk *clk_xin = NULL; > struct sdhci_host *host; > struct sdhci_pltfm_host *pltfm_host; > struct sdhci_arasan_data *sdhci_arasan; > + const struct of_device_id *of_id = > + of_match_device(sdhci_arasan_of_match, &pdev->dev); > + struct resource *res; > > sdhci_arasan = devm_kzalloc(&pdev->dev, sizeof(*sdhci_arasan), > GFP_KERNEL); > @@ -136,8 +235,9 @@ static int sdhci_arasan_probe(struct platform_device *pdev) > > sdhci_arasan->clk_ahb = devm_clk_get(&pdev->dev, "clk_ahb"); > if (IS_ERR(sdhci_arasan->clk_ahb)) { > - dev_err(&pdev->dev, "clk_ahb clock not found.\n"); > - return PTR_ERR(sdhci_arasan->clk_ahb); > + /* Clock is optional */ > + sdhci_arasan->clk_ahb = NULL; > + goto skip_clk; Clocks are optional for your SoC but they are necessary for Zynq. You can't just skip clocks for zynq because if DT is wrong clocks won't be enabled and even checked. Does it mean that there are no clocks for this IP? Or do you miss clock driver? Thanks, Michal -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, >> #define SDHCI_ARASAN_CLK_CTRL_OFFSET 0x2c >> @@ -34,6 +36,19 @@ >> */ >> struct sdhci_arasan_data { >> struct clk *clk_ahb; >> + struct platform_device *pdev; >> + void __iomem *ahb_aim_csr; >> + const struct sdhci_arasan_ahb_ops *ahb_ops; >> +}; >> + >> +/** >> + * struct sdhci_arasan_ahb_ops >> + * @init_ahb Initialize translation bus >> + * @xlat_addr Set up an 64-bit addressing translation > > This definitely breaking kernel-doc format. Please fix. I will add the missing ":". >> + sdhci_arasan->ahb_ops->xlat_addr(sdhci_arasan, >> + sg_dma_address(host->data->sg)); >> + } >> + writel(val, host->ioaddr + reg); >> +} > > The same code was submitted in v1 and Arnd commented it but you are still keeping > it here. Why? I responded back to Arnd. If I move this code to IO MMU, what do I do when we enable the actual IO MMU? The structure for the bus type only has one IO MMU pointer. We would need to IO MMU pointer and not sure how the IO MMU framework would handle this. > >> + >> static struct sdhci_ops sdhci_arasan_ops = { >> + .write_l = sdhci_arasan_writel, >> .get_max_clock = sdhci_pltfm_clk_get_max_clock, >> .get_timeout_clock = sdhci_arasan_get_timeout_clock, >> }; >> @@ -121,13 +150,83 @@ static int sdhci_arasan_resume(struct device *dev) >> static SIMPLE_DEV_PM_OPS(sdhci_arasan_dev_pm_ops, sdhci_arasan_suspend, >> sdhci_arasan_resume); >> >> +static int sdhci_arasan_xgene_init_ahb(struct sdhci_arasan_data *data) >> +{ >> + #define AIM_SIZE_CTL_OFFSET 0x00000004 >> + #define AIM_EN_N_WR(src) (((u32) (src) << 31) & 0x80000000) >> + #define ARSB_WR(src) (((u32) (src) << 24) & 0x0f000000) >> + #define AWSB_WR(src) (((u32) (src) << 20) & 0x00f00000) >> + #define AIM_MASK_N_WR(src) (((u32) (src)) & 0x000fffff) > > Remove that one more space between #define and name. > I am not fan of having these defines just here - move them to the top. > > Also these macros are used just here at one location. > Isn't it better just to define that BITS you want to setup instead of > these macros which are hardly to read? The only field that is single bit is AIM_EN_N_WR. All others are multiple bit fields. Either I write them in the code or use these defines. Would it be better if I just write them in the code? >> +static void sdhci_arasn_xgene_xlat_addr(struct sdhci_arasan_data *data, >> + u64 dma_addr) >> +{ >> + #define AIM_AXI_HI_OFFSET 0x0000000c >> + #define AIM_AXI_ADDRESS_HI_N_WR(src) \ >> + (((u32) (src) << 20) & 0xfff00000) > > ditto with indentation. > 20 should be shift > 0xfff00000 is mask. > > Also you should take this opportunity and add function description > in kernel doc to be exactly clear what this function is doing. The function is pretty small, simply and don't believe it needs function description. > >> + >> + if (!data->ahb_aim_csr) >> + return; >> + >> + writel(AIM_AXI_ADDRESS_HI_N_WR(dma_addr >> 32), >> + data->ahb_aim_csr + AIM_AXI_HI_OFFSET); >> +} >> + >> +static const struct sdhci_arasan_ahb_ops xgene_ahb_ops = { >> + .init_ahb = sdhci_arasan_xgene_init_ahb, >> + .xlat_addr = sdhci_arasn_xgene_xlat_addr, >> +}; >> + >> +static const struct of_device_id sdhci_arasan_of_match[] = { >> + { .compatible = "arasan,sdhci-8.9a" }, >> + { .compatible = "apm,arasan,sdhci-8.9a", .data = &xgene_ahb_ops }, >> + { } >> +}; >> +MODULE_DEVICE_TABLE(of, sdhci_arasan_of_match); >> + >> static int sdhci_arasan_probe(struct platform_device *pdev) >> { >> int ret; >> - struct clk *clk_xin; >> + struct clk *clk_xin = NULL; >> struct sdhci_host *host; >> struct sdhci_pltfm_host *pltfm_host; >> struct sdhci_arasan_data *sdhci_arasan; >> + const struct of_device_id *of_id = >> + of_match_device(sdhci_arasan_of_match, &pdev->dev); >> + struct resource *res; >> >> sdhci_arasan = devm_kzalloc(&pdev->dev, sizeof(*sdhci_arasan), >> GFP_KERNEL); >> @@ -136,8 +235,9 @@ static int sdhci_arasan_probe(struct platform_device *pdev) >> >> sdhci_arasan->clk_ahb = devm_clk_get(&pdev->dev, "clk_ahb"); >> if (IS_ERR(sdhci_arasan->clk_ahb)) { >> - dev_err(&pdev->dev, "clk_ahb clock not found.\n"); >> - return PTR_ERR(sdhci_arasan->clk_ahb); >> + /* Clock is optional */ >> + sdhci_arasan->clk_ahb = NULL; >> + goto skip_clk; > > Clocks are optional for your SoC but they are necessary for Zynq. > You can't just skip clocks for zynq because if DT is wrong clocks won't be enabled > and even checked. > Does it mean that there are no clocks for this IP? The clock for X-Gene is enabled by the time Linux boot. As the clock in programmed in the SDHC register, it knows the clock frequency. > Or do you miss clock driver? The clock will be configured by the FW and left out intentionally as this will also support ACPI boot. -Loc -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, On 06/05/2014 01:15 AM, Loc Ho wrote: > Hi, > >>> #define SDHCI_ARASAN_CLK_CTRL_OFFSET 0x2c >>> @@ -34,6 +36,19 @@ >>> */ >>> struct sdhci_arasan_data { >>> struct clk *clk_ahb; >>> + struct platform_device *pdev; >>> + void __iomem *ahb_aim_csr; >>> + const struct sdhci_arasan_ahb_ops *ahb_ops; >>> +}; >>> + >>> +/** >>> + * struct sdhci_arasan_ahb_ops >>> + * @init_ahb Initialize translation bus >>> + * @xlat_addr Set up an 64-bit addressing translation >> >> This definitely breaking kernel-doc format. Please fix. > > I will add the missing ":". > >>> + sdhci_arasan->ahb_ops->xlat_addr(sdhci_arasan, >>> + sg_dma_address(host->data->sg)); >>> + } >>> + writel(val, host->ioaddr + reg); >>> +} >> >> The same code was submitted in v1 and Arnd commented it but you are still keeping >> it here. Why? > > I responded back to Arnd. If I move this code to IO MMU, what do I do > when we enable the actual IO MMU? The structure for the bus type only > has one IO MMU pointer. We would need to IO MMU pointer and not sure > how the IO MMU framework would handle this. Then good time to start investigating this. >> >>> + >>> static struct sdhci_ops sdhci_arasan_ops = { >>> + .write_l = sdhci_arasan_writel, >>> .get_max_clock = sdhci_pltfm_clk_get_max_clock, >>> .get_timeout_clock = sdhci_arasan_get_timeout_clock, >>> }; >>> @@ -121,13 +150,83 @@ static int sdhci_arasan_resume(struct device *dev) >>> static SIMPLE_DEV_PM_OPS(sdhci_arasan_dev_pm_ops, sdhci_arasan_suspend, >>> sdhci_arasan_resume); >>> >>> +static int sdhci_arasan_xgene_init_ahb(struct sdhci_arasan_data *data) >>> +{ >>> + #define AIM_SIZE_CTL_OFFSET 0x00000004 >>> + #define AIM_EN_N_WR(src) (((u32) (src) << 31) & 0x80000000) >>> + #define ARSB_WR(src) (((u32) (src) << 24) & 0x0f000000) >>> + #define AWSB_WR(src) (((u32) (src) << 20) & 0x00f00000) >>> + #define AIM_MASK_N_WR(src) (((u32) (src)) & 0x000fffff) >> >> Remove that one more space between #define and name. >> I am not fan of having these defines just here - move them to the top. >> >> Also these macros are used just here at one location. >> Isn't it better just to define that BITS you want to setup instead of >> these macros which are hardly to read? > > The only field that is single bit is AIM_EN_N_WR. All others are > multiple bit fields. Either I write them in the code or use these > defines. Would it be better if I just write them in the code? >From my point of view will be the best to compose one macro like #define AIM_EN_N_WR BIT(31) ... #define AIM...INIT_VAL (AIM_EN_N_WR |... ) >>> +static void sdhci_arasn_xgene_xlat_addr(struct sdhci_arasan_data *data, >>> + u64 dma_addr) >>> +{ >>> + #define AIM_AXI_HI_OFFSET 0x0000000c >>> + #define AIM_AXI_ADDRESS_HI_N_WR(src) \ >>> + (((u32) (src) << 20) & 0xfff00000) >> >> ditto with indentation. >> 20 should be shift >> 0xfff00000 is mask. >> >> Also you should take this opportunity and add function description >> in kernel doc to be exactly clear what this function is doing. > > The function is pretty small, simply and don't believe it needs > function description. That documentation is for everybody not just for you who write the code. Simple description will be useful. >>> + >>> + if (!data->ahb_aim_csr) >>> + return; >>> + >>> + writel(AIM_AXI_ADDRESS_HI_N_WR(dma_addr >> 32), >>> + data->ahb_aim_csr + AIM_AXI_HI_OFFSET); >>> +} >>> + >>> +static const struct sdhci_arasan_ahb_ops xgene_ahb_ops = { >>> + .init_ahb = sdhci_arasan_xgene_init_ahb, >>> + .xlat_addr = sdhci_arasn_xgene_xlat_addr, >>> +}; >>> + >>> +static const struct of_device_id sdhci_arasan_of_match[] = { >>> + { .compatible = "arasan,sdhci-8.9a" }, >>> + { .compatible = "apm,arasan,sdhci-8.9a", .data = &xgene_ahb_ops }, >>> + { } >>> +}; >>> +MODULE_DEVICE_TABLE(of, sdhci_arasan_of_match); >>> + >>> static int sdhci_arasan_probe(struct platform_device *pdev) >>> { >>> int ret; >>> - struct clk *clk_xin; >>> + struct clk *clk_xin = NULL; >>> struct sdhci_host *host; >>> struct sdhci_pltfm_host *pltfm_host; >>> struct sdhci_arasan_data *sdhci_arasan; >>> + const struct of_device_id *of_id = >>> + of_match_device(sdhci_arasan_of_match, &pdev->dev); >>> + struct resource *res; >>> >>> sdhci_arasan = devm_kzalloc(&pdev->dev, sizeof(*sdhci_arasan), >>> GFP_KERNEL); >>> @@ -136,8 +235,9 @@ static int sdhci_arasan_probe(struct platform_device *pdev) >>> >>> sdhci_arasan->clk_ahb = devm_clk_get(&pdev->dev, "clk_ahb"); >>> if (IS_ERR(sdhci_arasan->clk_ahb)) { >>> - dev_err(&pdev->dev, "clk_ahb clock not found.\n"); >>> - return PTR_ERR(sdhci_arasan->clk_ahb); >>> + /* Clock is optional */ >>> + sdhci_arasan->clk_ahb = NULL; >>> + goto skip_clk; >> >> Clocks are optional for your SoC but they are necessary for Zynq. >> You can't just skip clocks for zynq because if DT is wrong clocks won't be enabled >> and even checked. >> Does it mean that there are no clocks for this IP? > > The clock for X-Gene is enabled by the time Linux boot. As the clock > in programmed in the SDHC register, it knows the clock frequency. > >> Or do you miss clock driver? > > The clock will be configured by the FW and left out intentionally as > this will also support ACPI boot. Not a problem if this is what you like but you can't just break Zynq by this change which is what you are doing right now. Thanks, Michal -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wednesday 04 June 2014 08:09:12 Michal Simek wrote: > > + > > +static void sdhci_arasn_xgene_xlat_addr(struct sdhci_arasan_data *data, > > + u64 dma_addr) > > +{ > > + #define AIM_AXI_HI_OFFSET 0x0000000c > > + #define AIM_AXI_ADDRESS_HI_N_WR(src) \ > > + (((u32) (src) << 20) & 0xfff00000) > > ditto with indentation. > 20 should be shift > 0xfff00000 is mask. > > Also you should take this opportunity and add function description > in kernel doc to be exactly clear what this function is doing. Actually this should just be configured in the dma-ranges property I think. It's not the driver's business to do the dma translation. Arnd -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/Documentation/devicetree/bindings/mmc/arasan,sdhci.txt b/Documentation/devicetree/bindings/mmc/arasan,sdhci.txt index 98ee2ab..4dfcb9e 100644 --- a/Documentation/devicetree/bindings/mmc/arasan,sdhci.txt +++ b/Documentation/devicetree/bindings/mmc/arasan,sdhci.txt @@ -8,14 +8,22 @@ Device Tree Bindings for the Arasan SDHCI Controller [3] Documentation/devicetree/bindings/interrupt-controller/interrupts.txt Required Properties: - - compatible: Compatibility string. Must be 'arasan,sdhci-8.9a' - - reg: From mmc bindings: Register location and length. - - clocks: From clock bindings: Handles to clock inputs. - - clock-names: From clock bindings: Tuple including "clk_xin" and "clk_ahb" + - compatible: Compatibility string. Must be 'arasan,sdhci-8.9a' or + 'apm,arasan,sdhci-8.9a' + - reg: First resource is the SDHC register location and length. + Second optional resource is the bridge translation register location + and length if required. - interrupts: Interrupt specifier - interrupt-parent: Phandle for the interrupt controller that services interrupts for this device. +Optional Properties: + - clocks: For format, refer to clock bindings. It requires two clocks if + specified - AHB clock and SDHC clock. + - clock-names: For format, refer to clock bindings. The clock corresponding + to the AHBC clock must be named "clk_ahb". The clock + corresponding to the SDHC clock must be named "clk_xin". + Example: sdhci@e0100000 { compatible = "arasan,sdhci-8.9a";
This patch updates Arasan SDHC documentation for the APM X-Gene SoC SDHC controller DTS binding. Signed-off-by: Loc Ho <lho@apm.com> --- .../devicetree/bindings/mmc/arasan,sdhci.txt | 16 ++++++++++++---- 1 files changed, 12 insertions(+), 4 deletions(-)