Re: [BUG]odroid-c2 does not hotplug usb-devices
Hi Martin,
I did a lot of googling this weekend and found some interesting problems with odroid-c2.
1. booting from emmc stop after u-boot 2020.04 (this hits me also) 2. usb did not work with newer kernels.
This gaves me the idea that something in the configuration of the s905 soc is not ok.
here: https://forum.libreelec.tv/thread/12330-balbes150-le-images-with-kodi-19-for... and here: https://lists.denx.de/pipermail/u-boot/2020-November/thread.html
in https://groups.io/g/u-boot-amlogic/topic/77816805#661 some problems in pinctrl related to booting from emmc is discussed.
i found that commit dd5f2351e99aad8fcbedbc1305b8b51b09952336 of u-boot rendered the situation for odroid-c2 bad.
i gave the following patch on Top of commit ee1e04558ff8c8ed812b986939447f129bb0b0bb of u-boot a try ad it fixes the odroid-c2 boot issue.
--- u-boot.git/arch/arm/dts/meson-gxbb.dtsi.orig 2020-11-27 17:20:04.624193561 +0100 +++ u-boot.git/arch/arm/dts/meson-gxbb.dtsi 2020-11-27 17:23:18.472088934 +0100 @@ -403,36 +403,33 @@ gpio-ranges = <&pinctrl_periphs 0 0 119>; };
- emmc_pins: emmc { - mux-0 { - groups = "emmc_nand_d07", - "emmc_cmd"; - function = "emmc"; - bias-pull-up; - };
- mux-1 { - groups = "emmc_clk"; - function = "emmc"; - bias-disable; - }; - }; + emmc_pins: emmc { + mux { + groups = "emmc_nand_d07", + "emmc_cmd", + "emmc_clk"; + function = "emmc"; + }; + };
- emmc_ds_pins: emmc-ds { - mux { - groups = "emmc_ds"; - function = "emmc"; - bias-pull-down; - }; - }; + emmc_ds_pins: emmc-ds { + mux { + groups = "emmc_ds"; + function = "emmc"; + }; + };
- emmc_clk_gate_pins: emmc_clk_gate { - mux { - groups = "BOOT_8"; - function = "gpio_periphs"; - bias-pull-down; - }; - }; + emmc_clk_gate_pins: emmc_clk_gate { + mux { + groups = "BOOT_8"; + function = "gpio_periphs"; + }; + cfg-pull-down { + pins = "BOOT_8"; + bias-pull-down; + }; + };
nor_pins: nor { mux {
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
For me this make it work again. But i`m sorry, I don´t know what would be a real fix.
Am 05.12.20 um 20:55 schrieb Martin Blumenstingl:
Hi Otto,
On Sat, Dec 5, 2020 at 6:40 PM Otto Meier gf435@gmx.net wrote:
Hi Martin,
i have tried your second notice. but it does'nt solve the issue. if at leased on device is pluged in new devices are detected. if all devices have been removed, no new device will be detected anymore.
thanks for trying this out and for reporting back. can you please also test if running "lsusb -vvv" (as root) makes the devices show up?
background info: the Amlogic Meson GXBB SoC on your Odroid-C2 board uses a "dwc2" USB 2.0 controller. I don't have any GXBB based board, but my Odroid-C1+ uses the same controller and overall setup. There I can reproduce the problem you are seeing.
I am not sure which part of the "infrastructure" (on-board 4-port USB hub, dwc2 controller, ...) is causing this issue. My suggestion is to involve the linux-usb mailing list and dwc2 driver maintainer:
- Minas Harutyunyan hminas@synopsys.com (dwc2 driver maintainer)
- linux-usb@vger.kernel.org (usb mailing list)
In the past they provided information and helped to debug issues. Please also keep me Cc'ed so I can help with any Amlogic specific questions, drivers, etc.
Best regards, Martin
Hi,
(Please CC u-boot-amlogic@groups.io when U-Boot is concerned)
On 07/12/2020 13:19, Otto Meier wrote:
Hi Martin,
I did a lot of googling this weekend and found some interesting problems with odroid-c2.
- booting from emmc stop after u-boot 2020.04 (this hits me also)
Indeed, this already has been reported, but the root cause is unknown
- usb did not work with newer kernels.
This gaves me the idea that something in the configuration of the s905 soc is not ok.
here: https://forum.libreelec.tv/thread/12330-balbes150-le-images-with-kodi-19-for... and here: https://lists.denx.de/pipermail/u-boot/2020-November/thread.html
in https://groups.io/g/u-boot-amlogic/topic/77816805#661 some problems in pinctrl related to booting from emmc is discussed.
i found that commit dd5f2351e99aad8fcbedbc1305b8b51b09952336 of u-boot rendered the situation for odroid-c2 bad.
i gave the following patch on Top of commit ee1e04558ff8c8ed812b986939447f129bb0b0bb of u-boot a try ad it fixes the odroid-c2 boot issue.
--- u-boot.git/arch/arm/dts/meson-gxbb.dtsi.orig 2020-11-27 17:20:04.624193561 +0100 +++ u-boot.git/arch/arm/dts/meson-gxbb.dtsi 2020-11-27 17:23:18.472088934 +0100 @@ -403,36 +403,33 @@ gpio-ranges = <&pinctrl_periphs 0 0 119>; }; - emmc_pins: emmc { - mux-0 { - groups = "emmc_nand_d07", - "emmc_cmd"; - function = "emmc"; - bias-pull-up; - }; - mux-1 { - groups = "emmc_clk"; - function = "emmc"; - bias-disable; - }; - }; + emmc_pins: emmc { + mux { + groups = "emmc_nand_d07", + "emmc_cmd", + "emmc_clk"; + function = "emmc"; + }; + }; - emmc_ds_pins: emmc-ds { - mux { - groups = "emmc_ds"; - function = "emmc"; - bias-pull-down; - }; - }; + emmc_ds_pins: emmc-ds { + mux { + groups = "emmc_ds"; + function = "emmc"; + }; + }; - emmc_clk_gate_pins: emmc_clk_gate { - mux { - groups = "BOOT_8"; - function = "gpio_periphs"; - bias-pull-down; - }; - }; + emmc_clk_gate_pins: emmc_clk_gate { + mux { + groups = "BOOT_8"; + function = "gpio_periphs"; + }; + cfg-pull-down { + pins = "BOOT_8"; + bias-pull-down; + }; + }; nor_pins: nor { mux {
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
Neil
For me this make it work again. But i`m sorry, I don´t know what would be a real fix.
Am 05.12.20 um 20:55 schrieb Martin Blumenstingl:
Hi Otto,
On Sat, Dec 5, 2020 at 6:40 PM Otto Meier gf435@gmx.net wrote:
Hi Martin,
i have tried your second notice. but it does'nt solve the issue. if at leased on device is pluged in new devices are detected. if all devices have been removed, no new device will be detected anymore.
thanks for trying this out and for reporting back. can you please also test if running "lsusb -vvv" (as root) makes the devices show up?
background info: the Amlogic Meson GXBB SoC on your Odroid-C2 board uses a "dwc2" USB 2.0 controller. I don't have any GXBB based board, but my Odroid-C1+ uses the same controller and overall setup. There I can reproduce the problem you are seeing.
I am not sure which part of the "infrastructure" (on-board 4-port USB hub, dwc2 controller, ...) is causing this issue. My suggestion is to involve the linux-usb mailing list and dwc2 driver maintainer:
- Minas Harutyunyan hminas@synopsys.com (dwc2 driver maintainer)
- linux-usb@vger.kernel.org (usb mailing list)
In the past they provided information and helped to debug issues. Please also keep me Cc'ed so I can help with any Amlogic specific questions, drivers, etc.
Best regards, Martin
Am 07.12.20 um 13:29 schrieb Neil Armstrong:
Hi,
(Please CC u-boot-amlogic@groups.io when U-Boot is concerned)
On 07/12/2020 13:19, Otto Meier wrote:
Hi Martin,
I did a lot of googling this weekend and found some interesting problems with odroid-c2.
- booting from emmc stop after u-boot 2020.04 (this hits me also)
Indeed, this already has been reported, but the root cause is unknown
- usb did not work with newer kernels.
This gaves me the idea that something in the configuration of the s905 soc is not ok.
here: https://forum.libreelec.tv/thread/12330-balbes150-le-images-with-kodi-19-for... and here: https://lists.denx.de/pipermail/u-boot/2020-November/thread.html
in https://groups.io/g/u-boot-amlogic/topic/77816805#661 some problems in pinctrl related to booting from emmc is discussed.
i found that commit dd5f2351e99aad8fcbedbc1305b8b51b09952336 of u-boot rendered the situation for odroid-c2 bad.
i gave the following patch on Top of commit ee1e04558ff8c8ed812b986939447f129bb0b0bb of u-boot a try ad it fixes the odroid-c2 boot issue.
--- u-boot.git/arch/arm/dts/meson-gxbb.dtsi.orig 2020-11-27 17:20:04.624193561 +0100 +++ u-boot.git/arch/arm/dts/meson-gxbb.dtsi 2020-11-27 17:23:18.472088934 +0100 @@ -403,36 +403,33 @@ gpio-ranges = <&pinctrl_periphs 0 0 119>; };
- emmc_pins: emmc { - mux-0 { - groups = "emmc_nand_d07", - "emmc_cmd"; - function = "emmc"; - bias-pull-up; - };
- mux-1 { - groups = "emmc_clk"; - function = "emmc"; - bias-disable; - }; - }; + emmc_pins: emmc { + mux { + groups = "emmc_nand_d07", + "emmc_cmd", + "emmc_clk"; + function = "emmc"; + }; + };
- emmc_ds_pins: emmc-ds { - mux { - groups = "emmc_ds"; - function = "emmc"; - bias-pull-down; - }; - }; + emmc_ds_pins: emmc-ds { + mux { + groups = "emmc_ds"; + function = "emmc"; + }; + };
- emmc_clk_gate_pins: emmc_clk_gate { - mux { - groups = "BOOT_8"; - function = "gpio_periphs"; - bias-pull-down; - }; - }; + emmc_clk_gate_pins: emmc_clk_gate { + mux { + groups = "BOOT_8"; + function = "gpio_periphs"; + }; + cfg-pull-down { + pins = "BOOT_8"; + bias-pull-down; + }; + };
nor_pins: nor { mux {
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
Neil
The above patch fixes booting from emmc for me. Using a newer uboot then 2020.04 is mandatory for using the above mentioned kernel. The usb issue, i had with earlier kernels from the above mentioned tree.
Hope that clearifies it. I wanted to run the newest kernel on my system which boots from emmc. I think these issues are probaly independant but in my system they did interfere.
Best regards Otto
For me this make it work again. But i`m sorry, I don´t know what would be a real fix.
Am 05.12.20 um 20:55 schrieb Martin Blumenstingl:
Hi Otto,
On Sat, Dec 5, 2020 at 6:40 PM Otto Meier gf435@gmx.net wrote:
Hi Martin,
i have tried your second notice. but it does'nt solve the issue. if at leased on device is pluged in new devices are detected. if all devices have been removed, no new device will be detected anymore.
thanks for trying this out and for reporting back. can you please also test if running "lsusb -vvv" (as root) makes the devices show up?
background info: the Amlogic Meson GXBB SoC on your Odroid-C2 board uses a "dwc2" USB 2.0 controller. I don't have any GXBB based board, but my Odroid-C1+ uses the same controller and overall setup. There I can reproduce the problem you are seeing.
I am not sure which part of the "infrastructure" (on-board 4-port USB hub, dwc2 controller, ...) is causing this issue. My suggestion is to involve the linux-usb mailing list and dwc2 driver maintainer:
- Minas Harutyunyan hminas@synopsys.com (dwc2 driver maintainer)
- linux-usb@vger.kernel.org (usb mailing list)
In the past they provided information and helped to debug issues. Please also keep me Cc'ed so I can help with any Amlogic specific questions, drivers, etc.
Best regards, Martin
Hi Otto,
On Mon, Dec 7, 2020 at 1:43 PM Otto Meier gf435@gmx.net wrote: [...]
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
if you have time it would be great if you could figure out which of the patches from Christian's tree fixes USB hotplugging for you. Or is it fixed in Linux 5.10-rcX even without any patches?
Best regards, Martin
Hi Martin,
Am 13.12.20 um 19:46 schrieb Martin Blumenstingl:
Hi Otto,
On Mon, Dec 7, 2020 at 1:43 PM Otto Meier gf435@gmx.net wrote: [...]
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
if you have time it would be great if you could figure out which of the patches from Christian's tree fixes USB hotplugging for you. Or is it fixed in Linux 5.10-rcX even without any patches?
The new mainline kernel 5.10.0 from Linus, without any other patches does detect USB hotpluging, when using u-boot DMI: Hardkernel Co., Ltd. ODROID-C2/ODROID-C2, BIOS 2021.01-rc3-00039-gec79f5ce22-dirty 12/08/2020 and the following u-boot patch:
--- u-boot.git/arch/arm/dts/meson-gxbb.dtsi.orig 2020-11-27 17:20:04.624193561 +0100 +++ u-boot.git/arch/arm/dts/meson-gxbb.dtsi 2020-11-27 17:23:18.472088934 +0100 @@ -403,36 +403,33 @@ gpio-ranges = <&pinctrl_periphs 0 0 119>; };
- emmc_pins: emmc { - mux-0 { - groups = "emmc_nand_d07", - "emmc_cmd"; - function = "emmc"; - bias-pull-up; - };
- mux-1 { - groups = "emmc_clk"; - function = "emmc"; - bias-disable; - }; - }; + emmc_pins: emmc { + mux { + groups = "emmc_nand_d07", + "emmc_cmd", + "emmc_clk"; + function = "emmc"; + }; + };
- emmc_ds_pins: emmc-ds { - mux { - groups = "emmc_ds"; - function = "emmc"; - bias-pull-down; - }; - }; + emmc_ds_pins: emmc-ds { + mux { + groups = "emmc_ds"; + function = "emmc"; + }; + };
- emmc_clk_gate_pins: emmc_clk_gate { - mux { - groups = "BOOT_8"; - function = "gpio_periphs"; - bias-pull-down; - }; - }; + emmc_clk_gate_pins: emmc_clk_gate { + mux { + groups = "BOOT_8"; + function = "gpio_periphs"; + }; + cfg-pull-down { + pins = "BOOT_8"; + bias-pull-down; + }; + };
nor_pins: nor { mux {
When i use the last unpatched emmc bootable u-boot 2020.04 the kernel boots, but usb hotplugging does not work.
Hope this describes my findings. If i can help further, please give me a note.
Best regards, Martin
Best regards Otto
Hi Otto,
On Mon, Dec 14, 2020 at 8:34 PM Otto Meier gf435@gmx.net wrote:
Hi Martin,
Am 13.12.20 um 19:46 schrieb Martin Blumenstingl:
Hi Otto,
On Mon, Dec 7, 2020 at 1:43 PM Otto Meier gf435@gmx.net wrote: [...]
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
if you have time it would be great if you could figure out which of the patches from Christian's tree fixes USB hotplugging for you. Or is it fixed in Linux 5.10-rcX even without any patches?
The new mainline kernel 5.10.0 from Linus, without any other patches does detect USB hotpluging, when using u-boot DMI: Hardkernel Co., Ltd. ODROID-C2/ODROID-C2, BIOS 2021.01-rc3-00039-gec79f5ce22-dirty 12/08/2020 and the following u-boot patch:
[...]
When i use the last unpatched emmc bootable u-boot 2020.04 the kernel boots, but usb hotplugging does not work.
Thank you for testing this!
Hope this describes my findings. If i can help further, please give me a note.
to be honest: I am a bit lost here. I don't understand how the BOOT_* pins interfere with USB. I also don't have any Odroid-C2 board myself so I cannot do any experiments myself. Neil, please let me know if you have any idea here.
Best regards, Martin
Hi Martin,
Am 19.12.20 um 14:58 schrieb Martin Blumenstingl:
Hi Otto,
On Mon, Dec 14, 2020 at 8:34 PM Otto Meier gf435@gmx.net wrote:
Hi Martin,
Am 13.12.20 um 19:46 schrieb Martin Blumenstingl:
Hi Otto,
On Mon, Dec 7, 2020 at 1:43 PM Otto Meier gf435@gmx.net wrote: [...]
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
if you have time it would be great if you could figure out which of the patches from Christian's tree fixes USB hotplugging for you. Or is it fixed in Linux 5.10-rcX even without any patches?
The new mainline kernel 5.10.0 from Linus, without any other patches does detect USB hotpluging, when using u-boot DMI: Hardkernel Co., Ltd. ODROID-C2/ODROID-C2, BIOS 2021.01-rc3-00039-gec79f5ce22-dirty 12/08/2020 and the following u-boot patch:
[...]
When i use the last unpatched emmc bootable u-boot 2020.04 the kernel boots, but usb hotplugging does not work.
Thank you for testing this!
Hope this describes my findings. If i can help further, please give me a note.
to be honest: I am a bit lost here. I don't understand how the BOOT_* pins interfere with USB. I also don't have any Odroid-C2 board myself so I cannot do any experiments myself. Neil, please let me know if you have any idea here.
The latest Patch from Neil in U-boot also fixes all the problems:
This fixes the wrong usage of clrsetbits_le32(), badly setting the set argument.
Fixes: c4c726c26b ("pinctrl: meson: add pinconf support") Reported-by: Anton Arapovarapov@gmail.com Reported-by: Otto Meiergf435@gmx.net Signed-off-by: Neil Armstrongnarmstrong@baylibre.com --- Hi Anton, Otto,
This should fix eMMC booting on Odroid-C2, could you have a quick try to confirm ?
Thanks, Neil
drivers/pinctrl/meson/pinctrl-meson.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/pinctrl/meson/pinctrl-meson.c b/drivers/pinctrl/meson/pinctrl-meson.c index d4539b02d8..5065b62436 100644 --- a/drivers/pinctrl/meson/pinctrl-meson.c +++ b/drivers/pinctrl/meson/pinctrl-meson.c @@ -216,13 +216,13 @@ static int meson_pinconf_bias_set(struct udevice *dev, unsigned int pin, }
/* othewise, enable the bias and select level */ - clrsetbits_le32(priv->reg_pullen + reg, BIT(bit), 1); + clrsetbits_le32(priv->reg_pullen + reg, BIT(bit), BIT(bit)); ret = meson_gpio_calc_reg_and_bit(dev, offset, REG_PULL, ®, &bit); if (ret) return ret;
clrsetbits_le32(priv->reg_pull + reg, BIT(bit), - param == PIN_CONFIG_BIAS_PULL_UP); + (param == PIN_CONFIG_BIAS_PULL_UP ? BIT(bit) : 0));
return 0; } -- 2.25.1
perhaps this gives an idea of what was wrong. But anyhow for me the issues with U-boot and usb hotplug is solved so far. For me this is even more i understand.
Best regards, Martin
Thanks for your help.
Otto
Hi Martin,
On Sat, 19 Dec 2020 at 19:29, Martin Blumenstingl martin.blumenstingl@googlemail.com wrote:
Hi Otto,
On Mon, Dec 14, 2020 at 8:34 PM Otto Meier gf435@gmx.net wrote:
Hi Martin,
Am 13.12.20 um 19:46 schrieb Martin Blumenstingl:
Hi Otto,
On Mon, Dec 7, 2020 at 1:43 PM Otto Meier gf435@gmx.net wrote: [...]
So with the latest u-boot and the kernel from https://github.com/chewitt/linux/tree/amlogic-5.10.y commit 725fc8df7898102f9031ba2075f763884ffa3ee8 everything is working again. USB does hotplugging as expected.
So, this fixes USB under Linux ?? It's not clear
if you have time it would be great if you could figure out which of the patches from Christian's tree fixes USB hotplugging for you. Or is it fixed in Linux 5.10-rcX even without any patches?
The new mainline kernel 5.10.0 from Linus, without any other patches does detect USB hotpluging, when using u-boot DMI: Hardkernel Co., Ltd. ODROID-C2/ODROID-C2, BIOS 2021.01-rc3-00039-gec79f5ce22-dirty 12/08/2020 and the following u-boot patch:
[...]
When i use the last unpatched emmc bootable u-boot 2020.04 the kernel boots, but usb hotplugging does not work.
Thank you for testing this!
Hope this describes my findings. If i can help further, please give me a note.
to be honest: I am a bit lost here. I don't understand how the BOOT_* pins interfere with USB. I also don't have any Odroid-C2 board myself so I cannot do any experiments myself. Neil, please let me know if you have any idea here.
Best regards, Martin
I was also looking into this issue so I made some changes in the phy driver to resolve the issue. Plz share your thoughts on the changes below.
Best Regards -Anand
amoon@ThinkPad-T440s:~/mainline/linux-aml-5.y-devel$ git diff diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi index 7c029f552a23..363dd2ac17e6 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi @@ -20,6 +20,7 @@ usb0_phy: phy@c0000000 { #phy-cells = <0>; reg = <0x0 0xc0000000 0x0 0x20>; resets = <&reset RESET_USB_OTG>; + reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB0>; clock-names = "usb_general", "usb"; status = "disabled"; @@ -30,6 +31,7 @@ usb1_phy: phy@c0000020 { #phy-cells = <0>; reg = <0x0 0xc0000020 0x0 0x20>; resets = <&reset RESET_USB_OTG>; + reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB1>; clock-names = "usb_general", "usb"; status = "disabled"; diff --git a/drivers/phy/amlogic/phy-meson8b-usb2.c b/drivers/phy/amlogic/phy-meson8b-usb2.c index 03c061dd5f0d..31523becc878 100644 --- a/drivers/phy/amlogic/phy-meson8b-usb2.c +++ b/drivers/phy/amlogic/phy-meson8b-usb2.c @@ -143,14 +143,6 @@ static int phy_meson8b_usb2_power_on(struct phy *phy) u32 reg; int ret;
- if (!IS_ERR_OR_NULL(priv->reset)) { - ret = reset_control_reset(priv->reset); - if (ret) { - dev_err(&phy->dev, "Failed to trigger USB reset\n"); - return ret; - } - } - ret = clk_prepare_enable(priv->clk_usb_general); if (ret) { dev_err(&phy->dev, "Failed to enable USB general clock\n"); @@ -222,9 +214,23 @@ static int phy_meson8b_usb2_power_off(struct phy *phy) return 0; }
+static int phy_meson8b_usb2_reset(struct phy *phy) +{ + struct phy_meson8b_usb2_priv *priv = phy_get_drvdata(phy); + + if (priv->reset) { + reset_control_assert(priv->reset); + udelay(10); + reset_control_deassert(priv->reset); + } + + return 0; +} + static const struct phy_ops phy_meson8b_usb2_ops = { .power_on = phy_meson8b_usb2_power_on, .power_off = phy_meson8b_usb2_power_off, + .reset = phy_meson8b_usb2_reset, .owner = THIS_MODULE, };
@@ -271,6 +277,10 @@ static int phy_meson8b_usb2_probe(struct platform_device *pdev) return -EINVAL; }
+ priv->reset = of_reset_control_get_shared(pdev->dev.of_node, "phy-reset"); + if (IS_ERR(priv->reset)) + priv->reset = NULL; + phy = devm_phy_create(&pdev->dev, NULL, &phy_meson8b_usb2_ops); if (IS_ERR(phy)) { dev_err(&pdev->dev, "failed to create PHY\n");
linux-amlogic mailing list linux-amlogic@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-amlogic
Hi Anand,
On Sat, Dec 19, 2020 at 8:53 PM Anand Moon linux.amoon@gmail.com wrote: [...]
I was also looking into this issue so I made some changes in the phy driver to resolve the issue. Plz share your thoughts on the changes below.
first I have some questions :-) 1. do you see the same problem that Otto sees? this means: a) USB hotplug works as long as at least one device is plugged in at boot b) (if I understand Otto correctly then) it breaks once all USB devices have been removed 2. does the mainline u-boot patch mentioned by Otto fix the problem for you? according to him it fixes the problem and he did not have to modify the USB PHY driver
amoon@ThinkPad-T440s:~/mainline/linux-aml-5.y-devel$ git diff diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi index 7c029f552a23..363dd2ac17e6 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi @@ -20,6 +20,7 @@ usb0_phy: phy@c0000000 { #phy-cells = <0>; reg = <0x0 0xc0000000 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB0>; clock-names = "usb_general", "usb"; status = "disabled";@@ -30,6 +31,7 @@ usb1_phy: phy@c0000020 { #phy-cells = <0>; reg = <0x0 0xc0000020 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB1>; clock-names = "usb_general", "usb"; status = "disabled";
I don't see why the above two changes are needed see my comment about of_reset_control_get_shared below
diff --git a/drivers/phy/amlogic/phy-meson8b-usb2.c b/drivers/phy/amlogic/phy-meson8b-usb2.c index 03c061dd5f0d..31523becc878 100644 --- a/drivers/phy/amlogic/phy-meson8b-usb2.c +++ b/drivers/phy/amlogic/phy-meson8b-usb2.c @@ -143,14 +143,6 @@ static int phy_meson8b_usb2_power_on(struct phy *phy) u32 reg; int ret;
if (!IS_ERR_OR_NULL(priv->reset)) {ret = reset_control_reset(priv->reset);if (ret) {dev_err(&phy->dev, "Failed to trigger USB reset\n");return ret;}}ret = clk_prepare_enable(priv->clk_usb_general); if (ret) { dev_err(&phy->dev, "Failed to enable USB general clock\n");@@ -222,9 +214,23 @@ static int phy_meson8b_usb2_power_off(struct phy *phy) return 0; }
+static int phy_meson8b_usb2_reset(struct phy *phy) +{
struct phy_meson8b_usb2_priv *priv = phy_get_drvdata(phy);if (priv->reset) {reset_control_assert(priv->reset);udelay(10);reset_control_deassert(priv->reset);}return 0;+}
static const struct phy_ops phy_meson8b_usb2_ops = { .power_on = phy_meson8b_usb2_power_on, .power_off = phy_meson8b_usb2_power_off,
.reset = phy_meson8b_usb2_reset, .owner = THIS_MODULE,};
I tested this on my Odroid-C1: phy_meson8b_usb2_reset is never called after checking the dwc2 code this is expected: only in one very specific case the dwc2 driver calls phy_reset can you please find out how phy_meson8b_usb2_reset is called in your kernel?
@@ -271,6 +277,10 @@ static int phy_meson8b_usb2_probe(struct platform_device *pdev) return -EINVAL; }
priv->reset = of_reset_control_get_shared(pdev->dev.of_node,"phy-reset");
this causes a memory-leak upon driver removal also a few lines above we are already getting the reset line, so why is this needed?
Best regards, Martin
Hi Martin,
i think the the u-boot fix made the new u-boot for me usable. Because i boot from emmc it fixes a long standing problem since 2020.04 in not booting from emmc. The latest uboot, booting from emmc before the fix was 2020.04, but this old uboot does not funktion well (hotpluging and others) with newer kernel (5.8.10 was the latest i new of). Therefore the latest Uboot initialized the kernel in a way that new kernels work again.
I haven't run kernel in between, with newer uboots like 2020.07 or 2020.10, because they din't boot from emmc.
so if the uboot patch is the real reason, i don't think.
best regrads
Otto
Am 19.12.20 um 23:31 schrieb Martin Blumenstingl:
Hi Anand,
On Sat, Dec 19, 2020 at 8:53 PM Anand Moon linux.amoon@gmail.com wrote: [...]
I was also looking into this issue so I made some changes in the phy driver to resolve the issue. Plz share your thoughts on the changes below.
first I have some questions :-)
- do you see the same problem that Otto sees? this means: a) USB
hotplug works as long as at least one device is plugged in at boot b) (if I understand Otto correctly then) it breaks once all USB devices have been removed 2. does the mainline u-boot patch mentioned by Otto fix the problem for you? according to him it fixes the problem and he did not have to modify the USB PHY driver
amoon@ThinkPad-T440s:~/mainline/linux-aml-5.y-devel$ git diff diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi index 7c029f552a23..363dd2ac17e6 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi @@ -20,6 +20,7 @@ usb0_phy: phy@c0000000 { #phy-cells = <0>; reg = <0x0 0xc0000000 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB0>; clock-names = "usb_general", "usb"; status = "disabled";@@ -30,6 +31,7 @@ usb1_phy: phy@c0000020 { #phy-cells = <0>; reg = <0x0 0xc0000020 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB1>; clock-names = "usb_general", "usb"; status = "disabled";I don't see why the above two changes are needed see my comment about of_reset_control_get_shared below
diff --git a/drivers/phy/amlogic/phy-meson8b-usb2.c b/drivers/phy/amlogic/phy-meson8b-usb2.c index 03c061dd5f0d..31523becc878 100644 --- a/drivers/phy/amlogic/phy-meson8b-usb2.c +++ b/drivers/phy/amlogic/phy-meson8b-usb2.c @@ -143,14 +143,6 @@ static int phy_meson8b_usb2_power_on(struct phy *phy) u32 reg; int ret;
if (!IS_ERR_OR_NULL(priv->reset)) {ret = reset_control_reset(priv->reset);if (ret) {dev_err(&phy->dev, "Failed to trigger USB reset\n");return ret;}}ret = clk_prepare_enable(priv->clk_usb_general); if (ret) { dev_err(&phy->dev, "Failed to enable USB general clock\n");@@ -222,9 +214,23 @@ static int phy_meson8b_usb2_power_off(struct phy *phy) return 0; }
+static int phy_meson8b_usb2_reset(struct phy *phy) +{
struct phy_meson8b_usb2_priv *priv = phy_get_drvdata(phy);if (priv->reset) {reset_control_assert(priv->reset);udelay(10);reset_control_deassert(priv->reset);}return 0;+}
- static const struct phy_ops phy_meson8b_usb2_ops = { .power_on = phy_meson8b_usb2_power_on, .power_off = phy_meson8b_usb2_power_off,
.reset = phy_meson8b_usb2_reset, .owner = THIS_MODULE,I tested this on my Odroid-C1: phy_meson8b_usb2_reset is never called after checking the dwc2 code this is expected: only in one very specific case the dwc2 driver calls phy_reset can you please find out how phy_meson8b_usb2_reset is called in your kernel?
@@ -271,6 +277,10 @@ static int phy_meson8b_usb2_probe(struct platform_device *pdev) return -EINVAL; }
priv->reset = of_reset_control_get_shared(pdev->dev.of_node,"phy-reset");
this causes a memory-leak upon driver removal also a few lines above we are already getting the reset line, so why is this needed?
Best regards, Martin
Hi Martin,
Thanks for testing and sharing your view points.
On Sun, 20 Dec 2020 at 04:01, Martin Blumenstingl martin.blumenstingl@googlemail.com wrote:
Hi Anand,
On Sat, Dec 19, 2020 at 8:53 PM Anand Moon linux.amoon@gmail.com wrote: [...]
I was also looking into this issue so I made some changes in the phy driver to resolve the issue. Plz share your thoughts on the changes below.
first I have some questions :-)
- do you see the same problem that Otto sees? this means: a) USB
hotplug works as long as at least one device is plugged in at boot b) (if I understand Otto correctly then) it breaks once all USB devices have been removed
On C1/C2 issue is when single usb hdd is connected to board it will not get detected. unless you connect another usb keyboard or wireless dongle to the board.
*USB hot plug only works if two ore more usb devices are connected.*
- does the mainline u-boot patch mentioned by Otto fix the problem
for you? according to him it fixes the problem and he did not have to modify the USB PHY driver
I have not check this out, I will verify this later.
amoon@ThinkPad-T440s:~/mainline/linux-aml-5.y-devel$ git diff diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi index 7c029f552a23..363dd2ac17e6 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi @@ -20,6 +20,7 @@ usb0_phy: phy@c0000000 { #phy-cells = <0>; reg = <0x0 0xc0000000 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB0>; clock-names = "usb_general", "usb"; status = "disabled";@@ -30,6 +31,7 @@ usb1_phy: phy@c0000020 { #phy-cells = <0>; reg = <0x0 0xc0000020 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB1>; clock-names = "usb_general", "usb"; status = "disabled";I don't see why the above two changes are needed see my comment about of_reset_control_get_shared below
Yep I borrowed this logic from rockchip usb phy controller. but I missed the action on reset.
err = devm_add_action_or_reset(base->dev, rockchip_usb_phy_action, rk_phy); if (err) return err;
diff --git a/drivers/phy/amlogic/phy-meson8b-usb2.c b/drivers/phy/amlogic/phy-meson8b-usb2.c index 03c061dd5f0d..31523becc878 100644 --- a/drivers/phy/amlogic/phy-meson8b-usb2.c +++ b/drivers/phy/amlogic/phy-meson8b-usb2.c @@ -143,14 +143,6 @@ static int phy_meson8b_usb2_power_on(struct phy *phy) u32 reg; int ret;
if (!IS_ERR_OR_NULL(priv->reset)) {ret = reset_control_reset(priv->reset);if (ret) {dev_err(&phy->dev, "Failed to trigger USB reset\n");return ret;}}ret = clk_prepare_enable(priv->clk_usb_general); if (ret) { dev_err(&phy->dev, "Failed to enable USB general clock\n");@@ -222,9 +214,23 @@ static int phy_meson8b_usb2_power_off(struct phy *phy) return 0; }
+static int phy_meson8b_usb2_reset(struct phy *phy) +{
struct phy_meson8b_usb2_priv *priv = phy_get_drvdata(phy);if (priv->reset) {reset_control_assert(priv->reset);udelay(10);reset_control_deassert(priv->reset);}return 0;+}
static const struct phy_ops phy_meson8b_usb2_ops = { .power_on = phy_meson8b_usb2_power_on, .power_off = phy_meson8b_usb2_power_off,
.reset = phy_meson8b_usb2_reset, .owner = THIS_MODULE,};
I tested this on my Odroid-C1: phy_meson8b_usb2_reset is never called after checking the dwc2 code this is expected: only in one very specific case the dwc2 driver calls phy_reset can you please find out how phy_meson8b_usb2_reset is called in your kernel?
Yep, Its not getting called on my board, l might be missing some thing.
@@ -271,6 +277,10 @@ static int phy_meson8b_usb2_probe(struct platform_device *pdev) return -EINVAL; }
priv->reset = of_reset_control_get_shared(pdev->dev.of_node,"phy-reset");
this causes a memory-leak upon driver removal also a few lines above we are already getting the reset line, so why is this needed?
I had a feeling that USB_OTG reset will be shared between both the phy controller.
Odroid C2 reset controller have additional RESET_USB_DDR_[0,3] reset, but Odroid C1 dose not have this feature.
Next time I will try to do more testing on both the devices C1/C2. before sending some code for review or testing.
Best Regards -Anand
Best regards, Martin
Hi Anand,
On Sun, Dec 20, 2020 at 2:46 PM Anand Moon linux.amoon@gmail.com wrote: [...]
On Sat, Dec 19, 2020 at 8:53 PM Anand Moon linux.amoon@gmail.com wrote: [...]
I was also looking into this issue so I made some changes in the phy driver to resolve the issue. Plz share your thoughts on the changes below.
first I have some questions :-)
- do you see the same problem that Otto sees? this means: a) USB
hotplug works as long as at least one device is plugged in at boot b) (if I understand Otto correctly then) it breaks once all USB devices have been removed
On C1/C2 issue is when single usb hdd is connected to board it will not get detected. unless you connect another usb keyboard or wireless dongle to the board.
*USB hot plug only works if two ore more usb devices are connected.*
I am not sure if that's really the same issue as described by Otto
[...]
amoon@ThinkPad-T440s:~/mainline/linux-aml-5.y-devel$ git diff diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi index 7c029f552a23..363dd2ac17e6 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi @@ -20,6 +20,7 @@ usb0_phy: phy@c0000000 { #phy-cells = <0>; reg = <0x0 0xc0000000 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB0>; clock-names = "usb_general", "usb"; status = "disabled";@@ -30,6 +31,7 @@ usb1_phy: phy@c0000020 { #phy-cells = <0>; reg = <0x0 0xc0000020 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB1>; clock-names = "usb_general", "usb"; status = "disabled";I don't see why the above two changes are needed see my comment about of_reset_control_get_shared below
Yep I borrowed this logic from rockchip usb phy controller. but I missed the action on reset.
err = devm_add_action_or_reset(base->dev, rockchip_usb_phy_action, rk_phy); if (err) return err;
(assuming that your comment relates to the of_reset_control_get_shared call below) I checked rockchip_usb_phy_action and it's not calling reset_control_put either this looks strange to me and I would say that there's a memory leak.
[...]
static const struct phy_ops phy_meson8b_usb2_ops = { .power_on = phy_meson8b_usb2_power_on, .power_off = phy_meson8b_usb2_power_off,
.reset = phy_meson8b_usb2_reset, .owner = THIS_MODULE,};
I tested this on my Odroid-C1: phy_meson8b_usb2_reset is never called after checking the dwc2 code this is expected: only in one very specific case the dwc2 driver calls phy_reset can you please find out how phy_meson8b_usb2_reset is called in your kernel?
Yep, Its not getting called on my board, l might be missing some thing.
are you sure that your patch fixes USB hotplug for you?
the new reset callback is a no-op because it's not called the code below (of_reset_control_get_shared) only fetches the same reset line that we are already using so I am not sure what changed behavior you are seeing - can you please explain this in more detail?
@@ -271,6 +277,10 @@ static int phy_meson8b_usb2_probe(struct platform_device *pdev) return -EINVAL; }
priv->reset = of_reset_control_get_shared(pdev->dev.of_node,"phy-reset");
this causes a memory-leak upon driver removal also a few lines above we are already getting the reset line, so why is this needed?
I had a feeling that USB_OTG reset will be shared between both the phy controller.
indeed, the USB_OTG reset is shared
Odroid C2 reset controller have additional RESET_USB_DDR_[0,3] reset, but Odroid C1 dose not have this feature.
have you found where the patched Amlogic kernel/u-boot use these reset lines? if they need to be managed by us then we probably have a bug, because we're not currently using these anywhere
Next time I will try to do more testing on both the devices C1/C2. before sending some code for review or testing.
I am happy about everyone who helps fixing the USB hotplug issue! it's a tricky issue where I don't even know if the USB PHY, dwc2 controller or something else is the culprit so any help is appreciated :)
Best regards, Martin
Hi Martin,
On Sun, 20 Dec 2020 at 20:04, Martin Blumenstingl martin.blumenstingl@googlemail.com wrote:
Hi Anand,
On Sun, Dec 20, 2020 at 2:46 PM Anand Moon linux.amoon@gmail.com wrote: [...]
On Sat, Dec 19, 2020 at 8:53 PM Anand Moon linux.amoon@gmail.com wrote: [...]
I was also looking into this issue so I made some changes in the phy driver to resolve the issue. Plz share your thoughts on the changes below.
first I have some questions :-)
- do you see the same problem that Otto sees? this means: a) USB
hotplug works as long as at least one device is plugged in at boot b) (if I understand Otto correctly then) it breaks once all USB devices have been removed
On C1/C2 issue is when single usb hdd is connected to board it will not get detected. unless you connect another usb keyboard or wireless dongle to the board.
*USB hot plug only works if two ore more usb devices are connected.*
I am not sure if that's really the same issue as described by Otto
OK.
[...]
amoon@ThinkPad-T440s:~/mainline/linux-aml-5.y-devel$ git diff diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi index 7c029f552a23..363dd2ac17e6 100644 --- a/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi @@ -20,6 +20,7 @@ usb0_phy: phy@c0000000 { #phy-cells = <0>; reg = <0x0 0xc0000000 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB0>; clock-names = "usb_general", "usb"; status = "disabled";@@ -30,6 +31,7 @@ usb1_phy: phy@c0000020 { #phy-cells = <0>; reg = <0x0 0xc0000020 0x0 0x20>; resets = <&reset RESET_USB_OTG>;
reset-names = "phy-reset"; clocks = <&clkc CLKID_USB>, <&clkc CLKID_USB1>; clock-names = "usb_general", "usb"; status = "disabled";I don't see why the above two changes are needed see my comment about of_reset_control_get_shared below
Yep I borrowed this logic from rockchip usb phy controller. but I missed the action on reset.
err = devm_add_action_or_reset(base->dev, rockchip_usb_phy_action, rk_phy); if (err) return err;
(assuming that your comment relates to the of_reset_control_get_shared call below) I checked rockchip_usb_phy_action and it's not calling reset_control_put either this looks strange to me and I would say that there's a memory leak.
Ok, I will try to check this, my approach was just to get the reset code to get trigger, but is seem that it is not getting triggered,
[...]
static const struct phy_ops phy_meson8b_usb2_ops = { .power_on = phy_meson8b_usb2_power_on, .power_off = phy_meson8b_usb2_power_off,
.reset = phy_meson8b_usb2_reset, .owner = THIS_MODULE,};
I tested this on my Odroid-C1: phy_meson8b_usb2_reset is never called after checking the dwc2 code this is expected: only in one very specific case the dwc2 driver calls phy_reset can you please find out how phy_meson8b_usb2_reset is called in your kernel?
Yep, Its not getting called on my board, l might be missing some thing.
are you sure that your patch fixes USB hotplug for you?
the new reset callback is a no-op because it's not called the code below (of_reset_control_get_shared) only fetches the same reset line that we are already using so I am not sure what changed behavior you are seeing - can you please explain this in more detail?
I am looking into the phy-meson-gxl-usb2.c and I will try to model these changes and see if it works on my boards.
Thanks -Anand
@@ -271,6 +277,10 @@ static int phy_meson8b_usb2_probe(struct platform_device *pdev) return -EINVAL; }
priv->reset = of_reset_control_get_shared(pdev->dev.of_node,"phy-reset");
this causes a memory-leak upon driver removal also a few lines above we are already getting the reset line, so why is this needed?
I had a feeling that USB_OTG reset will be shared between both the phy controller.
indeed, the USB_OTG reset is shared
Odroid C2 reset controller have additional RESET_USB_DDR_[0,3] reset, but Odroid C1 dose not have this feature.
have you found where the patched Amlogic kernel/u-boot use these reset lines? if they need to be managed by us then we probably have a bug, because we're not currently using these anywhere
Next time I will try to do more testing on both the devices C1/C2. before sending some code for review or testing.
I am happy about everyone who helps fixing the USB hotplug issue! it's a tricky issue where I don't even know if the USB PHY, dwc2 controller or something else is the culprit so any help is appreciated :)
Ok,
Best regards, Martin
participants (4)
-
Anand Moon -
Martin Blumenstingl -
Neil Armstrong -
Otto Meier