| Message ID | 20211213070330.3351505-1-rong.chen@amlogic.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mmc: meson: initial ocr available by default value | expand |
Hello, On Mon, Dec 13, 2021 at 8:03 AM Rong Chen <rong.chen@amlogic.com> wrote: > > The patch will add a value of ocr supported by the controller, > to specify some of voltage values are supported. Can you please share some details on this topic? We already have many boards upstreamed which work fine without manually setting ocr_avail in the driver. In general there's two types of boards: 1) boards where the regulator can switch between different voltages. One example is the Odroid-C2, see meson-gxbb-odroidc2.dts [0] Here we specify: vmmc-supply = <&tflash_vdd>; vqmmc-supply = <&tf_io>; &tflash_vdd is either turned off or at 3.3V when turned on. &tf_io can be either turned off, 1.8V or 3.3V. 2) boards where the voltages are fixed. One example is Radxa Zero, see meson-g12a-radxa-zero.dts [1] For the SD card the board uses: vmmc-supply = <&vddao_3v3>; vqmmc-supply = <&vddao_3v3>; (meaning fixed 3.3V always-on power source for both VMMC and VQMMC) and for eMMC the board uses: vmmc-supply = <&vcc_3v3>; vqmmc-supply = <&vcc_1v8>; (meaning: 3.3V always-on power source for vmmc and 1.8V always-on power source for VQMMC) With the information from the regulators the MMC core will automatically detect ocr_avail in Linux. Or in other words: your .dts may be missing the link between regulator and MMC controller vmmc/vqmmc. Can you please double-check this? Best regards, Martin [0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts?id=2585cf9dfaaddf00b069673f27bb3f8530e2039c#n354 [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/amlogic/meson-g12a-radxa-zero.dts?id=2585cf9dfaaddf00b069673f27bb3f8530e2039c#n333
On Mon 13 Dec 2021 at 19:50, "rong.chen@amlogic.com" <rong.chen@amlogic.com> wrote: > Hello, > If 3.3V always-on power source for vmmc is supplied by hardware boards fixed, don't need regulator in software, so .dts unnecessary the link between regulator > and MMC controller vmmc/vqmmc. Controllable or not, your vmmc/vqmmc regulators should be described in DT. There are several examples of this in arch/arm64/boot/dts/amlogic/*.dts > > ------------------------------------------------------------------------------------------------------------------------------------------------------ > rong.chen@amlogic.com > > > From: Martin Blumenstingl > Date: 2021-12-13 16:00 > To: Rong Chen > CC: Ulf Hansson; Neil Armstrong; Kevin Hilman; Jerome Brunet; --to=1131046452; 45581586; linux-mmc; linux-arm-kernel; linux-amlogic; linux-kernel > Subject: Re: [PATCH] mmc: meson: initial ocr available by default value > [ EXTERNAL EMAIL ] > > Hello, > > On Mon, Dec 13, 2021 at 8:03 AM Rong Chen <rong.chen@amlogic.com> wrote: > > > > The patch will add a value of ocr supported by the controller, > > to specify some of voltage values are supported. > Can you please share some details on this topic? > > We already have many boards upstreamed which work fine without > manually setting ocr_avail in the driver. > In general there's two types of boards: > 1) boards where the regulator can switch between different voltages. > One example is the Odroid-C2, see meson-gxbb-odroidc2.dts [0] > Here we specify: > vmmc-supply = <&tflash_vdd>; > vqmmc-supply = <&tf_io>; > &tflash_vdd is either turned off or at 3.3V when turned on. > &tf_io can be either turned off, 1.8V or 3.3V. > > 2) boards where the voltages are fixed. > One example is Radxa Zero, see meson-g12a-radxa-zero.dts [1] > For the SD card the board uses: > vmmc-supply = <&vddao_3v3>; > vqmmc-supply = <&vddao_3v3>; > (meaning fixed 3.3V always-on power source for both VMMC and VQMMC) > and for eMMC the board uses: > vmmc-supply = <&vcc_3v3>; > vqmmc-supply = <&vcc_1v8>; > (meaning: 3.3V always-on power source for vmmc and 1.8V always-on > power source for VQMMC) > > With the information from the regulators the MMC core will > automatically detect ocr_avail in Linux. > Or in other words: your .dts may be missing the link between regulator > and MMC controller vmmc/vqmmc. > Can you please double-check this? > > > Best regards, > Martin > > > [0] > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts?id=2585cf9dfaaddf00b069673f27bb3f8530e2039c#n354 > > [1] > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/amlogic/meson-g12a-radxa-zero.dts?id=2585cf9dfaaddf00b069673f27bb3f8530e2039c#n333
On Mon, Dec 13, 2021 at 2:25 PM Jerome Brunet <jbrunet@baylibre.com> wrote: > > > On Mon 13 Dec 2021 at 19:50, "rong.chen@amlogic.com" <rong.chen@amlogic.com> wrote: > > > Hello, > > If 3.3V always-on power source for vmmc is supplied by hardware boards fixed, don't need regulator in software, so .dts unnecessary the link between regulator > > and MMC controller vmmc/vqmmc. > > Controllable or not, your vmmc/vqmmc regulators should be described in > DT. I agree with Jerome. Here are the two reasons why I think that it's best to describe these regulators in device-tree: - device-tree is there to describe the hardware. It's especially useful for non-discoverable information (USB and PCI IDs are discoverable examples), for example: voltage supply of an MMC controller, UART IRQ line, ... - reviewing board.dts would be a lot harder if we rely on defaults (defaults which cannot be described in a device-tree schema). For meson-g12a-radxa-zero.dts this means we could omit the vmmc/vqmmc regulators for the SD card slot but we need them for the eMMC (because vqmmc is fixed at 1.8V). Best regards, Martin
diff --git a/drivers/mmc/host/meson-gx-mmc.c b/drivers/mmc/host/meson-gx-mmc.c index 8f36536cb1b6..c339e12d5516 100644 --- a/drivers/mmc/host/meson-gx-mmc.c +++ b/drivers/mmc/host/meson-gx-mmc.c @@ -4,6 +4,9 @@ * * Copyright (c) 2016 BayLibre, SAS. * Author: Kevin Hilman <khilman@baylibre.com> + * + * Copyright (c) 2021 Amlogic, inc. + * Author: Rong Chen <Rong.Chen@amlogic.com> */ #include <linux/kernel.h> #include <linux/module.h> @@ -130,6 +133,7 @@ #define SD_EMMC_DESC_CHAIN_MODE BIT(1) #define MUX_CLK_NUM_PARENTS 2 +#define SD_EMMC_OCR_AVAIL 0x200080 struct meson_mmc_data { unsigned int tx_delay_mask; @@ -1150,6 +1154,7 @@ static int meson_mmc_probe(struct platform_device *pdev) host->dram_access_quirk = device_property_read_bool(&pdev->dev, "amlogic,dram-access-quirk"); + mmc->ocr_avail = SD_EMMC_OCR_AVAIL; /* Get regulators and the supported OCR mask */ host->vqmmc_enabled = false; ret = mmc_regulator_get_supply(mmc);
The patch will add a value of ocr supported by the controller, to specify some of voltage values are supported. Ocr_avail should place an initial value to avoid uncertain value for the platform that unsupport regulator such as S4. Signed-off-by: Rong Chen <rong.chen@amlogic.com> --- drivers/mmc/host/meson-gx-mmc.c | 5 +++++ 1 file changed, 5 insertions(+) base-commit: 141edd9e99eb91393e8a4d97742bd98328bff724