| Message ID | 1619586716-8687-1-git-send-email-sanm@codeaurora.org |
|---|---|
| Headers | show |
| Series | USB DWC3 host wake up support from system suspend | expand |
Hi, Sandeep Maheswaram <sanm@codeaurora.org> writes: > During suspend read the status of all port and make sure the PHYs > are in the correct mode based on current speed. > Phy interrupt masks are set based on this mode. Keep track of the mode > of the HS PHY to be able to configure wakeup properly. > > Also check during suspend if any wakeup capable devices are > connected to the controller (directly or through hubs), if there > are none set a flag to indicate that the PHY should be powered > down during suspend. > > Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> > --- > drivers/usb/dwc3/core.h | 3 +++ > drivers/usb/dwc3/host.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++++ > 2 files changed, 62 insertions(+) > > diff --git a/drivers/usb/dwc3/core.h b/drivers/usb/dwc3/core.h > index b1e875c..cecd278 100644 > --- a/drivers/usb/dwc3/core.h > +++ b/drivers/usb/dwc3/core.h > @@ -1123,6 +1123,9 @@ struct dwc3 { > > bool phys_ready; > > + unsigned int hs_phy_mode; > + bool phy_power_off; > + > struct ulpi *ulpi; > bool ulpi_ready; > > diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > index f29a264..527f04c 100644 > --- a/drivers/usb/dwc3/host.c > +++ b/drivers/usb/dwc3/host.c > @@ -11,6 +11,14 @@ > #include <linux/platform_device.h> > > #include "core.h" > +#include "../host/xhci.h" > +#include "../host/xhci-plat.h" > + > +static int xhci_dwc3_suspend_quirk(struct usb_hcd *hcd); > + > +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > + .suspend_quirk = xhci_dwc3_suspend_quirk, > +}; we're passing data using device_properties, why do you want this here? > @@ -115,6 +123,13 @@ int dwc3_host_init(struct dwc3 *dwc) > } > } > > + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > + sizeof(struct xhci_plat_priv)); > + if (ret) { > + dev_err(dwc->dev, "failed to add data to xHCI\n"); > + goto err; > + } > + > ret = platform_device_add(xhci); > if (ret) { > dev_err(dwc->dev, "failed to register xHCI device\n"); > @@ -127,6 +142,50 @@ int dwc3_host_init(struct dwc3 *dwc) > return ret; > } > > +static void dwc3_set_phy_mode(struct usb_hcd *hcd) > +{ > + > + int i, num_ports; > + u32 reg; > + unsigned int ss_phy_mode = 0; > + struct dwc3 *dwc = dev_get_drvdata(hcd->self.controller->parent); > + struct xhci_hcd *xhci_hcd = hcd_to_xhci(hcd); > + > + dwc->hs_phy_mode = 0; > + > + reg = readl(&xhci_hcd->cap_regs->hcs_params1); > + num_ports = HCS_MAX_PORTS(reg); there's a big assumption here that xhci is still alive. Why isn't this quirk implemented in xhci-plat itself? > +int xhci_dwc3_suspend_quirk(struct usb_hcd *hcd) who calls this?
Hi, Sandeep Maheswaram <sanm@codeaurora.org> writes: > Configure interrupts based on hs_phy_mode to avoid triggering of > interrupts during system suspend and suspend the device successfully. > > Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> > Reviewed-by: Matthias Kaehlcke <mka@chromium.org> > --- > drivers/usb/dwc3/dwc3-qcom.c | 26 ++++++++++++++++++++------ > 1 file changed, 20 insertions(+), 6 deletions(-) > > diff --git a/drivers/usb/dwc3/dwc3-qcom.c b/drivers/usb/dwc3/dwc3-qcom.c > index 66183c6..82125bc 100644 > --- a/drivers/usb/dwc3/dwc3-qcom.c > +++ b/drivers/usb/dwc3/dwc3-qcom.c > @@ -316,22 +316,36 @@ static void dwc3_qcom_disable_wakeup_irq(int irq) > > static void dwc3_qcom_disable_interrupts(struct dwc3_qcom *qcom) > { > - dwc3_qcom_disable_wakeup_irq(qcom->hs_phy_irq); > + struct dwc3 *dwc = platform_get_drvdata(qcom->dwc3); and who's using this? > static void dwc3_qcom_enable_interrupts(struct dwc3_qcom *qcom) > { > - dwc3_qcom_enable_wakeup_irq(qcom->hs_phy_irq); > + struct dwc3 *dwc = platform_get_drvdata(qcom->dwc3); or this?
On 4/28/2021 3:29 PM, Felipe Balbi wrote: > Hi, > > Sandeep Maheswaram <sanm@codeaurora.org> writes: >> Avoiding phy powerdown when wakeup capable devices are connected >> by checking phy_power_off flag. >> Phy should be on to wake up the device from suspend using wakeup capable >> devices such as keyboard and mouse. >> >> Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> >> Reviewed-by: Matthias Kaehlcke <mka@chromium.org> >> --- >> drivers/usb/dwc3/core.c | 7 +++++-- >> 1 file changed, 5 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c >> index b6e53d8..bb414c3 100644 >> --- a/drivers/usb/dwc3/core.c >> +++ b/drivers/usb/dwc3/core.c >> @@ -1738,7 +1738,7 @@ static int dwc3_suspend_common(struct dwc3 *dwc, pm_message_t msg) >> dwc3_core_exit(dwc); >> break; >> case DWC3_GCTL_PRTCAP_HOST: >> - if (!PMSG_IS_AUTO(msg)) { >> + if (!PMSG_IS_AUTO(msg) && dwc->phy_power_off) { > should be able to detect this generically, no? Shouldn't > device_may_wakeup() be valid here and give you the answer you want? I think device_may_wakeup() gives whether the controller is wake up capable or not. But we want to keep phy powered on only when some wakeup capable devices (eg:keyboard ,mouse ) are connected to controller.
Hi, Sandeep Maheswaram <sanm@codeaurora.org> writes: >>> @@ -11,6 +11,14 @@ >>> #include <linux/platform_device.h> >>> >>> #include "core.h" >>> +#include "../host/xhci.h" >>> +#include "../host/xhci-plat.h" >>> + >>> +static int xhci_dwc3_suspend_quirk(struct usb_hcd *hcd); >>> + >>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >>> + .suspend_quirk = xhci_dwc3_suspend_quirk, >>> +}; >> we're passing data using device_properties, why do you want this here? > Similar implemenation was done in > drivers/usb/cdns3/host.c<https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/drivers/usb/cdns3/host.c?h=next-20210503> then it seems we have two places to correct :-) >>> @@ -127,6 +142,50 @@ int dwc3_host_init(struct dwc3 *dwc) >>> return ret; >>> } >>> >>> +static void dwc3_set_phy_mode(struct usb_hcd *hcd) >>> +{ >>> + >>> + int i, num_ports; >>> + u32 reg; >>> + unsigned int ss_phy_mode = 0; >>> + struct dwc3 *dwc = dev_get_drvdata(hcd->self.controller->parent); >>> + struct xhci_hcd *xhci_hcd = hcd_to_xhci(hcd); >>> + >>> + dwc->hs_phy_mode = 0; >>> + >>> + reg = readl(&xhci_hcd->cap_regs->hcs_params1); >>> + num_ports = HCS_MAX_PORTS(reg); >> there's a big assumption here that xhci is still alive. Why isn't this >> quirk implemented in xhci-plat itself? >> >>> +int xhci_dwc3_suspend_quirk(struct usb_hcd *hcd) >> who calls this? > This will be called from > xhci_plat_suspend->xhci_priv_suspend_quirk->xhci_dwc3_suspend_quirk So xhci.ko calls a function from dwc3.ko? That sounds odd, doesn't it? It really looks like we're just finding ways to bypass the driver layering, rather than working with it.
Hi, Sandeep Maheswaram <sanm@codeaurora.org> writes: >> Sandeep Maheswaram <sanm@codeaurora.org> writes: >>> Avoiding phy powerdown when wakeup capable devices are connected >>> by checking phy_power_off flag. >>> Phy should be on to wake up the device from suspend using wakeup capable >>> devices such as keyboard and mouse. >>> >>> Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> >>> Reviewed-by: Matthias Kaehlcke <mka@chromium.org> >>> --- >>> drivers/usb/dwc3/core.c | 7 +++++-- >>> 1 file changed, 5 insertions(+), 2 deletions(-) >>> >>> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c >>> index b6e53d8..bb414c3 100644 >>> --- a/drivers/usb/dwc3/core.c >>> +++ b/drivers/usb/dwc3/core.c >>> @@ -1738,7 +1738,7 @@ static int dwc3_suspend_common(struct dwc3 *dwc, pm_message_t msg) >>> dwc3_core_exit(dwc); >>> break; >>> case DWC3_GCTL_PRTCAP_HOST: >>> - if (!PMSG_IS_AUTO(msg)) { >>> + if (!PMSG_IS_AUTO(msg) && dwc->phy_power_off) { >> should be able to detect this generically, no? Shouldn't >> device_may_wakeup() be valid here and give you the answer you want? > > I think device_may_wakeup() gives whether the controller is wake up > capable or not. Yes, but it's a bit more than that. Looking at devices.rst we read: If :c:func:`device_may_wakeup(dev)` returns ``true``, the device should be prepared for generating hardware wakeup signals to trigger a system wakeup event when the system is in the sleep state. For example, :c:func:`enable_irq_wake()` might identify GPIO signals hooked up to a switch or other external hardware, and :c:func:`pci_enable_wake()` does something similar for the PCI PME signal. So, if there is a condition where $this device has to, somehow, deal with wakeup, it should be configured accordingly. This ->phy_power_off flag is telling us the same thing. > But we want to keep phy powered on only when some wakeup capable devices > (eg:keyboard ,mouse ) are connected to controller. Understood, it could be that we're missing some method for propagating that state (i.e. keyboard with PM support) up to the parent device, but that's no excuse to bypass driver boundaries. Wouldn't you agree?
On Wed, Apr 28, 2021 at 12:55:21PM +0300, Felipe Balbi wrote: > > Hi, > > Sandeep Maheswaram <sanm@codeaurora.org> writes: > > During suspend read the status of all port and make sure the PHYs > > are in the correct mode based on current speed. > > Phy interrupt masks are set based on this mode. Keep track of the mode > > of the HS PHY to be able to configure wakeup properly. > > > > Also check during suspend if any wakeup capable devices are > > connected to the controller (directly or through hubs), if there > > are none set a flag to indicate that the PHY should be powered > > down during suspend. > > > > Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> > > --- > > drivers/usb/dwc3/core.h | 3 +++ > > drivers/usb/dwc3/host.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++++ > > 2 files changed, 62 insertions(+) > > > > diff --git a/drivers/usb/dwc3/core.h b/drivers/usb/dwc3/core.h > > index b1e875c..cecd278 100644 > > --- a/drivers/usb/dwc3/core.h > > +++ b/drivers/usb/dwc3/core.h > > @@ -1123,6 +1123,9 @@ struct dwc3 { > > > > bool phys_ready; > > > > + unsigned int hs_phy_mode; > > + bool phy_power_off; > > + > > struct ulpi *ulpi; > > bool ulpi_ready; > > > > diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > > index f29a264..527f04c 100644 > > --- a/drivers/usb/dwc3/host.c > > +++ b/drivers/usb/dwc3/host.c > > @@ -11,6 +11,14 @@ > > #include <linux/platform_device.h> > > > > #include "core.h" > > +#include "../host/xhci.h" > > +#include "../host/xhci-plat.h" > > + > > +static int xhci_dwc3_suspend_quirk(struct usb_hcd *hcd); > > + > > +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > > + .suspend_quirk = xhci_dwc3_suspend_quirk, > > +}; > > we're passing data using device_properties, why do you want this here? > > > @@ -115,6 +123,13 @@ int dwc3_host_init(struct dwc3 *dwc) > > } > > } > > > > + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > > + sizeof(struct xhci_plat_priv)); > > + if (ret) { > > + dev_err(dwc->dev, "failed to add data to xHCI\n"); > > + goto err; > > + } > > + > > ret = platform_device_add(xhci); > > if (ret) { > > dev_err(dwc->dev, "failed to register xHCI device\n"); > > @@ -127,6 +142,50 @@ int dwc3_host_init(struct dwc3 *dwc) > > return ret; > > } > > > > +static void dwc3_set_phy_mode(struct usb_hcd *hcd) > > +{ > > + > > + int i, num_ports; > > + u32 reg; > > + unsigned int ss_phy_mode = 0; > > + struct dwc3 *dwc = dev_get_drvdata(hcd->self.controller->parent); > > + struct xhci_hcd *xhci_hcd = hcd_to_xhci(hcd); > > + > > + dwc->hs_phy_mode = 0; > > + > > + reg = readl(&xhci_hcd->cap_regs->hcs_params1); > > + num_ports = HCS_MAX_PORTS(reg); > > there's a big assumption here that xhci is still alive. Why isn't this > quirk implemented in xhci-plat itself? That should work for determining which types of devices are connected to the PHYs, however IIUC the xhci-plat doesn't know about the PHY topology. Are you suggesting to move that info into the xhci-plat driver so that it can set the corresponding PHY modes?
On Mon, May 03, 2021 at 02:20:23PM +0300, Felipe Balbi wrote: > > Hi, > > Sandeep Maheswaram <sanm@codeaurora.org> writes: > >> Sandeep Maheswaram <sanm@codeaurora.org> writes: > >>> Avoiding phy powerdown when wakeup capable devices are connected > >>> by checking phy_power_off flag. > >>> Phy should be on to wake up the device from suspend using wakeup capable > >>> devices such as keyboard and mouse. > >>> > >>> Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> > >>> Reviewed-by: Matthias Kaehlcke <mka@chromium.org> > >>> --- > >>> drivers/usb/dwc3/core.c | 7 +++++-- > >>> 1 file changed, 5 insertions(+), 2 deletions(-) > >>> > >>> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c > >>> index b6e53d8..bb414c3 100644 > >>> --- a/drivers/usb/dwc3/core.c > >>> +++ b/drivers/usb/dwc3/core.c > >>> @@ -1738,7 +1738,7 @@ static int dwc3_suspend_common(struct dwc3 *dwc, pm_message_t msg) > >>> dwc3_core_exit(dwc); > >>> break; > >>> case DWC3_GCTL_PRTCAP_HOST: > >>> - if (!PMSG_IS_AUTO(msg)) { > >>> + if (!PMSG_IS_AUTO(msg) && dwc->phy_power_off) { > >> should be able to detect this generically, no? Shouldn't > >> device_may_wakeup() be valid here and give you the answer you want? > > > > I think device_may_wakeup() gives whether the controller is wake up > > capable or not. > > Yes, but it's a bit more than that. Looking at devices.rst we read: > > If :c:func:`device_may_wakeup(dev)` returns ``true``, the device should be > prepared for generating hardware wakeup signals to trigger a system wakeup event > when the system is in the sleep state. For example, :c:func:`enable_irq_wake()` > might identify GPIO signals hooked up to a switch or other external hardware, > and :c:func:`pci_enable_wake()` does something similar for the PCI PME signal. > > So, if there is a condition where $this device has to, somehow, deal > with wakeup, it should be configured accordingly. This ->phy_power_off > flag is telling us the same thing. > > > But we want to keep phy powered on only when some wakeup capable devices > > (eg:keyboard ,mouse ) are connected to controller. > > Understood, it could be that we're missing some method for propagating > that state (i.e. keyboard with PM support) up to the parent device, but > that's no excuse to bypass driver boundaries. Wouldn't you agree? I'm not sure if device_may_wakeup() is really the right tool for the job. This is the current implementation: static inline bool device_may_wakeup(struct device *dev) { return dev->power.can_wakeup && !!dev->power.wakeup; } IIUC power.can_wakeup specifies whether the device is wakeup capable, primarily in physical terms and indicating that the driver is ready to handle wakeups, and power.wakeup represents the policy which can be changed by userspace. Supposing the hub is generally wakeup capable that flag shouldn't be changed. Neither should be the policy based on what is connected to the bus.
Hi, Matthias Kaehlcke <mka@chromium.org> writes: >> > @@ -127,6 +142,50 @@ int dwc3_host_init(struct dwc3 *dwc) >> > return ret; >> > } >> > >> > +static void dwc3_set_phy_mode(struct usb_hcd *hcd) >> > +{ >> > + >> > + int i, num_ports; >> > + u32 reg; >> > + unsigned int ss_phy_mode = 0; >> > + struct dwc3 *dwc = dev_get_drvdata(hcd->self.controller->parent); >> > + struct xhci_hcd *xhci_hcd = hcd_to_xhci(hcd); >> > + >> > + dwc->hs_phy_mode = 0; >> > + >> > + reg = readl(&xhci_hcd->cap_regs->hcs_params1); >> > + num_ports = HCS_MAX_PORTS(reg); >> >> there's a big assumption here that xhci is still alive. Why isn't this >> quirk implemented in xhci-plat itself? > > That should work for determining which types of devices are connected to > the PHYs, however IIUC the xhci-plat doesn't know about the PHY topology. > Are you suggesting to move that info into the xhci-plat driver so that it > can set the corresponding PHY modes? Yes, if xHCI needs to know about PHYs in order to properly configure the PHYs, so be it :-)
On Thu, May 13, 2021 at 04:46:41PM +0300, Felipe Balbi wrote: > > Hi, > > Matthias Kaehlcke <mka@chromium.org> writes: > >> > @@ -127,6 +142,50 @@ int dwc3_host_init(struct dwc3 *dwc) > >> > return ret; > >> > } > >> > > >> > +static void dwc3_set_phy_mode(struct usb_hcd *hcd) > >> > +{ > >> > + > >> > + int i, num_ports; > >> > + u32 reg; > >> > + unsigned int ss_phy_mode = 0; > >> > + struct dwc3 *dwc = dev_get_drvdata(hcd->self.controller->parent); > >> > + struct xhci_hcd *xhci_hcd = hcd_to_xhci(hcd); > >> > + > >> > + dwc->hs_phy_mode = 0; > >> > + > >> > + reg = readl(&xhci_hcd->cap_regs->hcs_params1); > >> > + num_ports = HCS_MAX_PORTS(reg); > >> > >> there's a big assumption here that xhci is still alive. Why isn't this > >> quirk implemented in xhci-plat itself? > > > > That should work for determining which types of devices are connected to > > the PHYs, however IIUC the xhci-plat doesn't know about the PHY topology. > > Are you suggesting to move that info into the xhci-plat driver so that it > > can set the corresponding PHY modes? > > Yes, if xHCI needs to know about PHYs in order to properly configure the > PHYs, so be it :-) Thanks for the confirmation, looks like we have a path forward here then :)
Hi Felipe, On 5/13/2021 3:17 AM, Matthias Kaehlcke wrote: > On Mon, May 03, 2021 at 02:20:23PM +0300, Felipe Balbi wrote: >> Hi, >> >> Sandeep Maheswaram <sanm@codeaurora.org> writes: >>>> Sandeep Maheswaram <sanm@codeaurora.org> writes: >>>>> Avoiding phy powerdown when wakeup capable devices are connected >>>>> by checking phy_power_off flag. >>>>> Phy should be on to wake up the device from suspend using wakeup capable >>>>> devices such as keyboard and mouse. >>>>> >>>>> Signed-off-by: Sandeep Maheswaram <sanm@codeaurora.org> >>>>> Reviewed-by: Matthias Kaehlcke <mka@chromium.org> >>>>> --- >>>>> drivers/usb/dwc3/core.c | 7 +++++-- >>>>> 1 file changed, 5 insertions(+), 2 deletions(-) >>>>> >>>>> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c >>>>> index b6e53d8..bb414c3 100644 >>>>> --- a/drivers/usb/dwc3/core.c >>>>> +++ b/drivers/usb/dwc3/core.c >>>>> @@ -1738,7 +1738,7 @@ static int dwc3_suspend_common(struct dwc3 *dwc, pm_message_t msg) >>>>> dwc3_core_exit(dwc); >>>>> break; >>>>> case DWC3_GCTL_PRTCAP_HOST: >>>>> - if (!PMSG_IS_AUTO(msg)) { >>>>> + if (!PMSG_IS_AUTO(msg) && dwc->phy_power_off) { >>>> should be able to detect this generically, no? Shouldn't >>>> device_may_wakeup() be valid here and give you the answer you want? >>> I think device_may_wakeup() gives whether the controller is wake up >>> capable or not. >> Yes, but it's a bit more than that. Looking at devices.rst we read: >> >> If :c:func:`device_may_wakeup(dev)` returns ``true``, the device should be >> prepared for generating hardware wakeup signals to trigger a system wakeup event >> when the system is in the sleep state. For example, :c:func:`enable_irq_wake()` >> might identify GPIO signals hooked up to a switch or other external hardware, >> and :c:func:`pci_enable_wake()` does something similar for the PCI PME signal. >> >> So, if there is a condition where $this device has to, somehow, deal >> with wakeup, it should be configured accordingly. This ->phy_power_off >> flag is telling us the same thing. >> >>> But we want to keep phy powered on only when some wakeup capable devices >>> (eg:keyboard ,mouse ) are connected to controller. >> Understood, it could be that we're missing some method for propagating >> that state (i.e. keyboard with PM support) up to the parent device, but >> that's no excuse to bypass driver boundaries. Wouldn't you agree? > I'm not sure if device_may_wakeup() is really the right tool for the > job. This is the current implementation: > > static inline bool device_may_wakeup(struct device *dev) > { > return dev->power.can_wakeup && !!dev->power.wakeup; > } > > IIUC power.can_wakeup specifies whether the device is wakeup > capable, primarily in physical terms and indicating that the > driver is ready to handle wakeups, and power.wakeup represents > the policy which can be changed by userspace. > > Supposing the hub is generally wakeup capable that flag > shouldn't be changed. Neither should be the policy based on > what is connected to the bus. Please suggest us how to proceed further with this patch.